NATOScienceSeries

ASeries presentingthe results ofactivitiessponsoredby theNATO ScienceCommittee.

The seriesispublishedbylOS Press and K1uwerAcademicPublishersinconjunctionwith the NATO
ScientificAffairs Division.
General Sub-Series
A Life Sciences lOS Press
B Physics K1uwerAcademicPublishers
C Mathematicaland PhysicalSciences K1uwerAcademicPublishers
D Behaviouraland Social SCiences K1uwerAcademicPublishers
E AppliedSciences K1uwerAcademicPublishers
F Computerand SystemsSciences lOS Press
Partnership Sub-Series
1 DisarmamentTechnologies K1uwerAcademic Publishers
2 EnvironmentalSecurity Kluwer AcademicPublishers
3 High Technology K1uwerAcademicPublishers
4 ScienceandTechnologyPolicy lOS Press
5 ComputerNetwor1(ing lOS Press
ThePartnershipSub-Seriesincorporatesactivitiesundertakenincollaborationwith NATO'sPartnersin
the Euro-AtlanticPartnershipCouncil - countriesofthe CIS and Central and Eastern Europe - inPriority
Areas ofconcernto those countries.
N T ~ O D T BASE
The NATOSCience5eriescontinuesthe series of books publishedformer1yinthe NATO ASI Series.
An electronicindex tothe NATO ASI seriesprovides fullbibliographicalreferences(with keywords
and/orabstracts)to more than 50000contributionsfrom internationalscientistspUblishedinan
sectionsofthe NATOASI Series.
Accesstothe NATO-PCO-DATABASE ispossible intwo ways:
- via onlineFILE 128(NATO-PCo-DATABASE) hosted by ESRIN.
Via GalileoGalilei, 1-00044Frascati, Italy;
- via CD-ROM"NATo-PCO-DATABASE" with user-friendtyretrievalsoftwareinEnglish. French and
German(@WTVGmbHand DATAWARETechnologiesInc.1989).
TheCD-ROMofthe NATOASlseriescan beordered from PCO, Overijse. Belgium.
~
seriesA: Life Sciences- Vol. 321 ISSN: 1387-6686
Beyond the Map
ArchaeologyandSpatialTechnologies
Editedby
GaryLock
InstituteofArchaeology, University of Oxford, Oxford, UnitedKingdom
IDS
Press
Otmsha
Amsterdam.Berlin.Oxford.Tokyo.Washington, DC
Publishedincooperationwilh NATO ScientificAffairs Division
ProceedingsoftheNATOAdvancedResearchWorkshopon
BeyondtheMap:ArchaeologyandSpatialTechnologies
Ravel1o. Italy
1-2October.1999
e 2000.lOSPress
AI1 rights reserved.No part ofIbis book maybereproduced. stored in a retrieval system. ortransmitted.
inanyformorbyanymeans.withouttheprior writtenpennissionfromthepublisher.
ISBN I58603021 3(lOSPress)
ISBN4274903877C3045(Ohrnsha)
LibraryofCongressCatalogCardNwnber:00-106857
Publisher
lOSPress
NieuweHemweg6B
1013BG Amsterdam
Netherlands
fall:+31206203419
e-mail:order@iospress.nl
DistribuJor in the USA and Canada
Distributorin the UK and Ireland
lOSPress,Inc.
lOSPress/LavisMarketing
5795-0BurkeCentrePartway
73LimeWalk
Burke,VA22015
Headington
USA
Oxford0X37AD
fall: +I7033233668
England
e-mail:iosbooks@iospress.com
fall:+44 1865750079
Distributorin Japan
Distributorin Germany
Ohmsha, Ltd.
lOSPress
3-1KandaNishiki-cho
SpandauerStrasse2
Chiyoda-ku, Tokyo10I
0.10178Berlin
Japan
Germany
fall:+81332332426
fall:+49302423113
LEGALNOTICE
TIlepublisherisnot responsiblefortheusewllichmightbe made ofthefol1owinginformation.
PRINTEDINTIlENlITHERLANDS
cc 80. LJ
P-4
a
'X I , v
J ~ P O
Contents
Listofcontributors ; vii
Preface
Gary Lock xi
Introduction: Return toRavello
Gary Lock and Trevor Harris ~ xiii
Session1
Vision, perceptionandGIS: developingenrichedapproachestothestudyof
archaeologicalvisibility
David Wheatley and Mark Gillings 1
Perceptionand viewsheds: are they mutually inclusive?
Andre P. Tschan, W/odzimierz Raczkowski and Ma/gorzata Latalowa 28
Understandingwetlandarchaeologicallandscapes:GIS,environmental
analysis andlandscapereconstruction;pathways and narratives
Henry Chapman , 49
Session Idiscussion:
Aparticularview
Gary Lock : 60
Session2
Understandingmovement: apilotmodel towards thesociologyofmovement
Marcos Uobera 65
Topographicandcultural influenceson walking'the Ridgeway inlater
prehistorictimes
Tyler Bell and Gary Lock . : 85,
Quantifyingthe non-quantifiable:studying hunter-gathererlandscapes
Joel Boaz and Espen Uleberg , 101
Session 2discussion:
MovingGIS:exploringmovementwithin prehistoricculturallandscapesusing GIS
Trevor Harris : 116
vi
vii
Session 3
Land evaluationas predictivemodel1ing:adeductiveapproach
Hans Kamermans 124
UnderstandingRoman settlementpatterns throughmultivariatestatisticsand predictive
modelling
Zoran StanCicand Tatjana Yeljanovski 147
AGIS-basedanalysis oftheEtruscancemeteryofPontecagnanousing fuzzy logic
Marco Crescioli, Andrea D'Andrea and Franco Niccolucci 157
Session3discussion:
Commentson archaeologicalprediction
Martin Kuna 180
Session 4
Ancientand new Pompeii:a projectfor monitoringarchaeologicalsites in
denselypopulatedareas
Federica Giannini. Maria Teresa Pareschi, Grete Stefani and Marina Bisson 187
Archaeologyand virtualmicro-topography:the creationofOEMs for reconstructing
fossil landscapesbyRemoteSensingandGIS applications
Maurizio Forte 199
Beyond the map: harmonisingresearchand CulturalResourceManagement
Neil Lang 214
Session4discussion:
Archaeology.GIS andCulturalResourceManagement
Philip Verhagen 229
Authorindex 236
ListofContributors
TylerBell
Institute ofArchaeology
UniversityofOxford
36,BeaumontStreet
Oxford
OXI2PG
UK
tyler.bell@arch.ox.ac.uk
MarinaBisson
CNR-CentroStudioperlaGeologia
StrutturaleeDinamicadel1'Appennino
ViaS. Maria 53
1-56126Pisa
Italy
JoeJBoaz
InstituteforArchaeology
Art History andConservationStudies
UniversityofOslo
Norway
joel.boaz@Ulkn.uio.no
HenryChapman
Centre for Wetland Archaeology
UniversityofHull
Hull
HU67RX
UK
H.P. Chapman@geo.hull.ac.uk .
Marco Crescioli
Unirel sri
Florence
Italy
marco@unirel.it
AndreaD'Andrea
IstitutoUniversitarioOrientale
VicoltettoI,S. Maria adAgnone
1-80143Naples
ltalia
dandrea@iuo.it
Maurizio Ferte,
CNR-ITABC
ViaSalariakm 29,300, c.p. 10
()()()16MonterotondoStazione
Rome
Italy
forte@m1ib.cnr.it
FedericaGiannini
CNR-CentroStudioperlaGeologia
Strutturale eOinamicadell'Appennino
ViaS.Maria 53 .
1-56126Pisa
Italy
gionnini@dst.unipi.it
Mark Gillings
SchoolofArchaeologicalStudies
UniversityofLeicester-
UniversityRoad
LE17RH
Leicester.
UK
mg41@le.ac;uk
viii
ix
TrevorHarris
DepartmentofGeologyand
Geography
West Virginia University
425,White Hall
Morgantown
WV 26506-6300
USA
Tharris2@wvu.edu
HansKamermans
Faculty ofArchaeology
UniversityofLeiden
POBox9515
2300RA Leiden
The Netherlands
H.Kamermans@Arch.LeidenUniv.nl
MartiDKuoa
Archeologickyustav AVCR
Letenska4
CS-1l801
Prague
Czech Republic
kuna@arup.ca.cz
NeilLang
English Heritage
National MonumentsRecord Centre
Kemble Drive
SN22GZ
Swindon
UK
neil.lang@rchme.co.uk
MalgonataLatalowa
LaboratoryofPalaeoecologyand
Archaeobotony
DepartmentofPlantEcologyand
Nature Protection
UniversityofGdaDsk
80-441 Gdansk
Poland
bioml@monika.univ.gda.pl
MarcosLlobera
PittRivers Museum
UniversityofOxford
60BanburyRoad
0X26PN
Oxford
UK
marcos.llobera@arch.ox.ac.uk
GaryLock
Institute ofArchaeology
UniversityofOxford
36,BeaumontStreet
Oxford
OXI2PG
UK
gary.lock@arch.o:x.ac.uk
FrancoNiccolucci
University ofFlorence
Florence
Italy
archimed@ats.it
MariaTeresaPareschi
CNR-CentroStudioperlaGeologia
StrutturaleeDinarnicadell' Appennino
ViaS.Maria 53
1-56126Pisa
Italy
pareschi@dst.unipi.it
WlodzimierzRaczkowski
Institute ofPrehistory
AdamMickiewiczUniversity
sw.Marcin 78
61709 Poznan
Poland
wlodekra@amu.edu.pl
ZonmStao&
ScientificResearch Centre ofSASA
Gosposka 13
1000Ljubljiana
Slovenia
zoran@zrc-sazu.si
GreteStefani
SoprintendenzaArcheologicadi
Pompeii
Pompeii
Naples
Italy
Andre Tscban
Instituteof Archaeology
UniversityofOxford
36,BeaumontStreet
Oxford
OXI2PG
UK
andre.tschan@arch.o:x.ac.uk
Espen meberg
UniversityCulturalHistoricMuseums
UniversityofOslo
Frederiksgt.3
N-DI64
Norway
euleberg@ialcn.uio.no
TatjaoaVeIjanovski
ScientificResearchCentreofSASA
Gosposka13
1000Ljubljiana
Slovenia
tajanav@alpha.zrc-sQZu.si
Pbilip Verhagen
RAAP
Zeeburgerdijk54
1094ARAmsterdam
The Netherlands
ph.verhagen@raop.nl
David Wheatley'
DepartmentofArchaeology
UniversityofSouthampton
Highfield
Southampton
S017IB1
UK
D.W. Wheatley@soton.ac.uk
xi
Preface
These are the proceedings of the second meeting that I have organised at Ravello. The
academic justifications for the return visit are discussed in the Introduction to this volume.
Here I would simply like to record thanks to the people and organisations that made the
meeting possible and so enjoyable. As previously, the meeting was hosted by the European
University Centre for CultUral Heritage at the Villa Rufolo. Many thanks to the Scientific
Committee for looking favourably on my application and providing funding and support
for the infrastructure of the meeting. I am grateful to Eugenia Apicella and her team at the
Villa who were again responsible for all of the arrangements and handling the inevitable
complications of bringing together. thirty. people from ten different countries with their
customary style and humour. I would also like to thank Lidia Lozano, Institute of
Archaeology, Oxford University, for dealing with the financial affairs.
The bulk of the funding was provided by the North Atlantic Treaty Organization
(NATO), Scientific and Environmental Affairs Division, as an Advanced Research
Workshop (Ref. 975817). Many thanks to Trevor Harris for suggesting this source.
ARW funding includes tight deadlines for publication so thanks are due to all of the
authors for being so prompt with written versions and for quickly sorting out the various
problems that arose. Special thanks are due to Jenny Cashman of Oxford University who
produced the final camera-ready manuscript required by the publishers, and to Ron Jansen
at lOS Press for his guidance through the publication process.
A final thank-you to the speakers, discussants and participants who made it such a
successful meeting. I hope you all agree that this volume is. a faithful record of events, or at
least some of them.
Gary Lock
Oxford, March 2000
Sponsored by English Heritage
- -- .... -_.- . -,----751 it . . . . eo.
.-----
xiii
BeyondtheMap
G. Lock (Ed.)
lOS Press, 2000
Beyond the Map:
Archaeologyand
Spatial Technologies
Introduction: Return to Ravello
Gary Lock
Institute.0/Archaeology
UniversityofOxford
Trevor Harris
DepartmentofGeologyandGeography
WestVirginia University
1. Welcome
The first Ravello conference on Archaeology and Geographical Information Systems: A
European Perspective [I) took: place six years ago to the very day of this conference. That
event had several aims, not least to establish a European 'presence' within a developing
domain of archaeological GIS that in mid-I993 was looking heavily biased towards North
America. Perhaps the most interesting result c;>f the 1993 meeting was the focus that
emerged on the relationship, and especially the tensions, between theory and practice.
Although these issues encompassed wider notions such as scientism versus humanism, and.
the constraints of quantification and working with qualitative data, .the tensions were
manifested in the issues of Environmental Determinism, and as a split between Cultural
Resource Management (CRM) and 'Iaddscape archaeology'. .
It was with these issues in mind, and how they had developed over the intervening
six years, that the second meeting was organised. The general Call for Papers asked for
contributions that specifically considered the theory and practice of spatial archaeology:
not just descriptive applications of drawing maps electronically but a consideration of the
implications and potential of going beyond the map .. No area of application was favoured
over others: Cultural Resource Management, landscape analysis, applications
were all as welcome, and the technological base was widened from GIS to Spatial
Technologies in recognition of the increasing integration taking place. In terms of the .
number of papers submitted the response was surprisingly disappointing. Twelve papers
were accepted and three rejected. It is difficultto say whether this is significant or not, but
considering the increasing utilisation of GIS and related technologies it could be construed
as a lack: of interest in, or awareness of, the more thooretical aspects and implications of
their application in archaeology.
This is not to denigrate the quality of the papers that comprised the meeting and are
presented here. The structure of this volume mirrors that of the meeting with four sessions
xiv xv
Beyondth Map
of three papers each and a period of discussion. In each case the discussion was led and the
discussant has written a contribution that pulls together themes from the session within a
wider context. Themed sessions were not planned at the outset but 'emerged' from the
papers offered, albeit some more convincingly than others. Sessions 1 and 2, the first day,
form a fairly coherent group concerned with humanistic approaches to landscape and
especially how spatial technologies can relate to notions of visibility and movement, of
being 'situated within', of dwelling. The second day, sessions 3 and 4, offered an
interesting contrast to the first and includes papers that can be loosely grouped as CRM
including predictive modelling, and/or as technologically-methodologically focussed.
There are obvious echoes here of 1993, and it is to a more detailed consideration of the last
six years that we now tum.
2. From 1993 to 1999 - themes of continuity and change
In the proceedings of the first Ravello meeting we attempted to evaluate GIS in European
archaeology by considering the past, present and future of theory and applications [2J. The
intention here is to briefly review that paper, to identify continuing themes and update
them where relevant. Following this, the final three sections explore in more detail three
issues that we feel are fundamental to current and future GIS usage in archaeology. Firstly
the socio-theoretic critique of GIS, a debate that is well established within Geography and
should be explored within an archaeological context, although it was first mentioned in
1995 under the sub-heading of 'issues of data and knowledge distortion'. This includes the
Environmental Determinism argument which has now moved on to the more productive
considerations of how to incorporate 'non-environmental', or social/cultural information.
Indeed, the first two sessions of this meeting were concerned with trying to extend the
limitations of essentially quantitative software to incorporate qualitative experiences.
Section 2.1 below extends this debate, challenges the unquestioned acceptance of the
technology, and places it within a wider inter-disciplinary arena.
Scale arose as a second major issue. Again it is one that we first mentioned six
years ago and we return to it below in section 2.2. Whereas our initial concerns were based
on the scale and accuracy of archaeological representations due to pointllinelpolygon
limitations, here we widen the debate to include a critique of source data mis-use and the
implications for multi-scalar analysis. We also include the issues of ecological fallacy and
the Modifiable Areal Unit Problem which are little discussed in archaeology and yet pose
significant threats to the validity of GIS-based analysis.
The third and final extended discussion below, section 2.3, concerns technological
advances which have been considerable since the first conference. Here we are not so
much interested in the massively increased accessibility to GIS-type software and the
flourishing of desk-top applications due to the ever-increasing speed and power of Pes, but
rather with the increasing integration of spatial technologies. These include scientific
visualisation, Virtual Reality, and GPS which when integrated with GIS and spatial
analysis have given rise to the term Geographic Information Science. Combined with
delivery over the Internet, GIScience offers opportunities to explore alternative concepts of
space and place that archaeology is already beginning to utilise. Six years ago we
identified one of the functional limitations of GIS as being the inability of the technology
to deal with the multi-dimensionality of archaeological requirements. There are still very
few examples of 'multi-dimensional' GIS as then described and work in that area has
progressed little, although the potential of the new integrated technologies has now shifted
the focus somewhat.
G. [,oel and T. Harris / Introduction
Our 1995 paper started by looking at the development of GIS in archaeology and
identifying three 'traditional', i.e. pre-GIS, areas of activity that accounted for early GIS
emphasis, namely CRM, statistical spatial analysis and landscape analysis. The GIS boom
was charted by using various publications as indicators, particularly the annual Computer
Applications in Archaeology conference (CAA). Not surprisingly the adoption of GIS in
all areas of archaeology has continued. to rise dramatically as shown 'by Scollar in his CAA '
25'" anniversary review [3], and at CAA in Barcelona the following year, 1998, where GIS
papers accounted for nearly 40% of the total published [4J. The impact is much wider than
the somewhat self-perpetuating momentum of CAA, for example the Xlllth Congress of
the International Union of Pre- and Proto-Historic Sciences in Forli, Italy, in 1996 ,
contained a colloquium on GiS applications that has produced proceedings containing 35
papers [5]. An interesting secondary product of the Forli Colloquium is the paper by
Bampton [6], an American geographer who attended the meeting and claims that
archaeological GIS applications are being championed by a small cadre of specialists while
the vast majority of archaeologists still have serious reservations and misconceptions about
the technology.
In contrast to Bampton's minimalist view, the CAERE Survey conducted and
published in 1998 as a special edition of thejournal Archeologia e Calcolatori (volume 9),
suggests that a deeper understanding qf the technology and its implications for areas of
archaeological endeavour are beginning to mature in many countries. In the UK, for' .
example, several issues have been forced into focus through the work of the Archaeology
Data Service (ADS) in establishing mechanisms for digital archiving in archaeology. Their
GIS Guide to Good Practice [7] is an essential reference work and has been invaluable in
raising awareness of wider issues such as developing international spatial data standards .
and the importance of metadata. ,
Also discussed in our 1995 paper as being significant in the differing developments
of GIS in European and North American archaeology, was the role of predictive .
modelling. This was seen as central to North American applications and peripheral to
European ones based largely on different traditions of 'landscape' archaeology and the
different characteristics of the archaeological record in North America and in Europe. With
the recent publication of Practical Applications of GIS for Archaeologists. A Predictive
Modeling Kit [8] it seems that the gulf has widened even further. In the preface to this
collection of nine papers based on an American conference held in 1996, the editors think
that 'the archaeological community would benefit from a book with a narrower and more .
refined focus [to] help steer GIS applications-in archaeology on a more solid
course' [ibid: xiv]. There follows various case-studies based on statistical methods
developed in the mid-1980s and then a critique of the basic assumptions of inductive '
predictive modelling concluding that it is a deeply flawed methodology [9J. The final
chapter, entitled GIS Applications in Archaeology: Method in Search of Theory [10],
suggests the way forward is through the adoption of explanatory models within a landscape
perspective based on the theoretical framework' of human ecology developed by Karl .
Butzer and others in the early 19805 but using modem technology. The differences
between this volume and the landscape approach espoused here in sections 1 and 2, is
shown by the material in the bibliographies which reinforces the different traditions, from,
which the works emanate. This difference is brought into sharp focus by the fact that
Tilley's The Phenomenology of Landscape is not cited once in any' of the eight :
bibliographies, despite being the most referenced book: in post-processual landscape
studies. '
The following three sections elucidate several of the themes introduced so far and
situate them within wider networks of discourse.
xvi xvii
Beyond the Map
2.1 The socio-theoretic critique ofGIS
As GIS becomes more prevalent and embedded within archaeological analysis, so latent
issues associated with the technology and fundamentally linked to the methodological,
epistemological and ontological traditions of archaeology are increasingly coming to the
forefront. GIS brings with it certain assumptions, principles, and practices about the way
in which analysis and the acquisition of knowledge is pursued. Archaeology, like many
other social science disciplines has undergone philosophical debates concerning the way in
which archaeological research is undertaken and the relationships between data and theory.
In archaeology, these debates are especially significant because it is now accepted to be a
discipline based on interpretation [l l], Material culture, whether artifacts, sites or
landscapes do not carry inherent meaning, but only make sense within interpretative, and
often contested, frameworks and the advent of GIS technology brings its own set of issues
to those debates. To date, the use of GIS has largely represented the positivist traditions of
archaeology in which the spatial analytic approach to archaeological phenomena has
predominated. Some would argue that this return or re-emergence of positivist approaches
and the quantitative treatment of data has been to the detriment of advances made by post-
processualist thinking. These debates have certainly come to the fore within the discipline
of Geography and are worthy of evaluation in the context of archaeological studies. This
largely social-theoretic critique of GIS, broadly defined under the label of 'GIS and
Society', provides valuable insight into the potential inherent bias of the technology. But it
also provides valuable pointers to the way in which GIS itself might respond to redress the
sole reliance on a positivist approach and to seek alternative uses of GIS in a manner that
extends beyond its spatial analytic positivist traditions. As shown in sections I and 2 of
this volume, this is certainly an issue when seeking to explore or use GIS in the context of
post-processualist debates and discourse. In the context of the overall concept of looking
beyond the map, these themes appear to be especially relevant and worthy of further
discussion.
It is perhaps ironic that as the advent of GIS during the 1980s and 1990s raised the
profile of maps, spatial analysis and the overall role of geographical space in numerous
studies, that the earliest criticisms of the technology should in fact come from the
discipline of Geography. With the possible exception of Chrisman's paper in 1987 [12],
and his call that GIS design be based on social and cultural goals and the pursuit of equity,
much of the early focus in GIS was on predominantly technical issues or on demonstration
application projects. In the early 1990s several discussions ensued that brought to the fore
how space is conceived and represented in GIS and of the perceived positivism and
hegemonic power relations embedded in the technology. The early public debate was
initiated by Taylor and others [13], [14], [15], [16], [17] and began to shift discussion
toward consideration of the relations of GIS to ongoing theoretical advances in geography.
A number of issues were raised in these early discussions concerning the role of GIS, not
the least of which was the perceived objective, value neutral nature of GIS itself. In this
discourse several points are raised that in many ways relate to the use of GIS by
archaeologists. The publication of Ground Truth: the Social lmplicaiions of Geographic
Information Systems [18] and the Friday Harbor Workshop [19], The NCGIA Initiative 19
and Varenius workshops, and the specialist issue of Cartography and Geographic
Information Systems, all prepared the basis for the debate. Two aspects of this debate are
detailed here, as they pertain to the use of GIS in archaeology.
Firstly, data is a fundamental component of GIS. Without data, the richness of the
technology is depleted to that of an empty shell. Indeed, the metaphor is often made that a
GIS without data is equivalent to an empty refrigerator. Numerous studies have repeatedly
G. Lock and T. Harrisllntroduction
recorded the effort and expense required to populate a GIS with -useful data. Generating,
capturing, encoding, and sharing a multiplicity of spatial data types has become a
predominant pastime of archaeological researchers as well many other GIS users. Many
valuable GIS publications have dwelt upon the nature and utility of a multitude of spatial
data and sources and how such data contributes, or may contribute, to GIS analysis. Issues
about data. and the political economy' of data, permeate much of the critique of GIS.
Taylor, for example, claimed that with the advent of GIS, 'facts' had risen to the top of the
agenda followed by a retreat from knowledge to data [20]. As data is a fundamental
component of GIS, concern was expressed that the technology would find application more
readily in data rich environments or countries, and that data poor topics or regions would
be relegated to a lesser priority. "Goodbye Africa - little data little geography' chided
Taylor and Overton [21: 1088]. In archaeology, data and GIS has already undergone one
series of heated debate in the previous Ravello meeting [22]. The greater availability of
contemporary physical environmental information in the form of data on soils, terrain, and'
hydrology and the inherent difficulty of generating or re-creating cultural or societal
information, has provided a natural advantage to physically-oriented GIS interpretations of
landscape and cultural processes. In this respect the first Ravello conference raised
concern about the perceived GIS-instigated re-emergence of environmental determinism
and such an interpretation gained occasional open ac1herents [ibid.]. Data, of course, do not .'
exist, they have to be created, and as many have pointed out, data area social construction
- the 'for whom, by whom, and for what purpose' is based upon a mix of social, political,
and economic contexts [21]. In the context of such data inequity, the rise of GIS in
archaeology poses several questions as to the universality of its use in archaeological
analysis. The GIS and Society critique takes this discussion a stage further within the
context of democratic use of the technology and the power relations arising from
differential access to the hardware, software, and human-ware components of GIS, its
surveillant capability, privacy and confidentiality, and the implications of data
commodification in the context of the growing bureaucratic-informational complex. GIS is
not an inexpensive undertaking and those who are excluded by reasons of cost or access to
data are in essence socially or spatially marginalized from the decision making process.
A second point concerns issues of how other differing representations of .
knowledge are potentially excluded from GIS. Much information generated is of
qualitative interpretation and traditionally this has been very difficult to include within
GIS. GIS in this respect is spatially deterministic, in that data that cannot be captured as. a
spatial primitive comprising point, line, polygon, or pixel, is essentially excluded from the
database. But what of other qualitative' forms of knowledge. such as sketch maps, or the'
artifacts which provide meaning to the analysis. Essentially qualitative or alternative
forms of knowledge representation that are crucial to understanding the nature of a place
are largely excluded from a GIS. Similarly, understanding and representing the multiple.
realities of space and environment is a crucial component of the GIS and Society
discourse. It is traditional for GIS to 'capture' one official view of reality that is heavily
biased toward a scientific data-driven representation. Mark raiSes several fascinating
aspects concerning whose view of the world is being represented in a GIS [23]. It has' been
suggested that one reason why GIS has been so spectacularly successful to date is because
it represents one non-contradictory view of the world [24]. This understanding is
particularly important for regional and national CRM systems within current debates of
alternative views of thepast. The often posed question 'who owns thepast' is likely to be
channeled into a single answer through a seemingly inflexible GIS rather than into multiple.
interpretations that take into account the beliefs of indigenous peoples and alternative
religious groups. The contemporary claims on Stonehenge [25], perhaps England's most
famous archaeological site, is a good example of how the rigid recording systems of state
xviii xix
Beyond the Map
archaeology fail to incorporate alternative explanations. This single official view is more
likely to become fossilised rather than more flexible, through the application of GIS. GIS is
also fundamentally based upon the Boolean logic of the Turing machine [26] in the
analysis of geographical space thus imposing a way of understanding and representing
nature and society.
Post-processual approaches in archaeology would thus appear to be directly at odds
with the use of GIS. Some of these issues became apparent at this meeting during a
valuable exchange of views related to the use of GIS in CRM. Essentially the question
was posed as to how the meaning or symbolic nature of a site could be incorporated or
explored in a GIS predictive model of archaeological resources. GIS and the way it is used
tends to treat phenomena as discrete points or lines in Euclidian space, upon which spatial
analytic procedures, in this instance, seek to tease out a predictive map for management
purposes. Thus distance from water, or road, or rock quarry are explored as predictors of
human activity or settlement patterns, However, what is there to differentiate the sites and
how is the archaeology of a site incorporated into this 'homogenous' set of point data?
How is the symbolism of a site or landscape, the culturally important aspects that could
affect locational decisions, captured and incorporated into the system. Much as the earlier
re-occurrence of environmental determinism arose from the facilitation of GIS, so too
might the pendulum swing too far toward 'spatialization' of archaeology to the detriment
of ignoring, avoiding, or abandoning alternative but very rich ways of knowing and
understanding archaeological phenomena.
Thus, despite an interval of six years between the Ravello meetings, one
fundamental point remains: that data availability and the nature of that data remains both a
core component and a point of debate in terms of the use of GIS in archaeology and its
utility as a tool in archaeological analysis. The development of object-oriented GIS and
the availability of multi-media based Internet GIS may go some way toward resolving
these questions and providing a mechanism whereby both qualitative and quantitative data
can be combined within a GIS and enable a cross fertilization of both geographical science
and theoretical developments in post-processual archaeology
2.2 Scale as artifact
Important as the above points are, a deeper underlying generic component of geographical
and indeed, archaeological data, is that of scale. Most insights about the importance of
scale came to the fore many years ago and are largely understood in the context of paper
maps. The advent and rapid diffusion of GIS has acted to dramatically emphasize the
importance of scale in GIS analyses. However, it is highly debatable whether, in the flood
of GIS usage, the complexities of scale issues are understood or even acknowledged as a
crucial aspect of data analysis. Concern about the use, misuse, and abuse of spatial data in
GIS analyses, oftentimes linked to scale issues, represents an important but little discussed
component of current GIS usage in archaeology.
Many disciplines share with geography and archaeology the need to operate across
a scale continuum ranging from the micro-level to national, international or even global
macro-level. GIS greatly facilitates the almost effortless transition between either end of
this continuum. In that ease of use, however, lays a very real concern for scale issues
associated with multi-scale data integration and analysis and the risk of invalidating results
because of ecological fallacy. Ecological fallacy can occur when data that are captured at
one scale are used to infer processes or patterns that are occurring at other scales of
analysis. Because archaeological patterns, spatial heterogeneity, and understanding are
dependent on scale, concern about ecological fallacy is rarely raised in archaeological
analyses.
G. Lock and T. Harris Ilniroduction
Generating data takes extensive effort and introduces a slew of issues associated
with data capture, accuracy, resolution, error estimation, metadata, storage, data structures
and compression, and data sharing to name but a few. Of concern to many working in
GIS, however, are the problems that arise from working with digital data inherited from
elsewhere and over which an investigator has had very little, if any, control in its
generation. Fundamentally coupled to the generation and use of spatial data is the issue of
scale, or perhaps more precisely, appropriate scale. .
Scale refers to a level of spatio-temporal representation. Scale can refer to the
geographical extent of a study or to the degree of detail or content exhibited. At I: 10;000,
archaeological sites may well be represented. in a GIS as polygonal units whereas at
I :50,000 scale these same sites would be represented as point data thus representing an
effective change in data type. In the latter, scale is inferred to represent a level of
accuracy. Scale is intertwined with issues of data measurement, data accuracy, and data
resolution. Spatio-temporal scaling is a conceptual and methodological 'problem that
applies to nearly all social sciences though GIS is increasingly exacerbating the issues
because of its ability to move beyond the restrictions previously imposed by the static and
controlled environment of the paper map. Representation of the real world is based on
spatial and temporal data usually captured at a limited range of scales and problems.
Temporal and spatial scale issues often plague data collection efforts as well as subsequent
analyses. Data at 'sub-optimal' scales are sometimes all that is available and
archaeologists must be very conscious of the implications of this constraint on the validity
and quality of results.
There are several other aspects associated with scale issues but two are singled out
for attention here as they pertain to archaeological analyses: First, because our
understanding of scale largely originates from our historical use of the traditional paper
map, the flexibility and dynamism of digital GIS represents significant challenges to our
understanding of data and scale issues. On the paper map scale remains constant and fixed
and represents the level at which the original information was captured and represented on
the map. The ability of a user to alter or change that ratio on a paper map is severely
limited. However, GIS and its component digital mapping capability now provide
considerable opportunity for the misuse of scale, Many will have probably heard the
horror stories of those who, in their search for supposedly 'larger scale' data, photocopied
and enlarged paper maps thinlcing they were changing the scale of .the rnap.. Of course
while the ratio between map units and real distance units had changed, the original Scale at
which the data had been captured remains unaltered The ease with which GIS essentially
enables the important ratio of scale to be changed and the difficulty of identifying data
scale in GIS products, demands that care and attention must be exercised regarding the
correct use of scale.
A second issue regarding the potential misuse of scale arises from the need for most
archaeological GIS applications to use available digital spatial data. A great variety of
digital spatial data is now available, often at nominal cost. Importantly. it is crucial that
the scale and purpose for which that dataset was created must be borne in mind by
potential users. Metadata is a vital element in providing the necessary information
regarding the dataset being used. Archaeological research' is often undertaken by
researchers using digital spatial data created by others and invariably the researcher had
little control and sometimes only limited understanding (or concern) about the limitations
or purpose of that data. Despite this, decisions or interpretations are made based on
geographic scales and units determined by other groups and invariably based on
contemporary political spatial units. Thus the scale of dataused in a GIS is often outside
the control of the archaeologist and is determined and inherited from data collection efforts
or activities designed for' entirely different purposes. Given the substantial cost of
i
.j
I
xx x;li
Beyond the Map
obtaining GIS data this trend will likely continue. Archaeologists are unlikely to be able to
control the entire process of data collection, compilation, management, analysis and
modeling. It is possible then, that the digital spatial data available for many GIS-
archaeological projects will only rarely perfectly match the scales at which processes that
are being studied actually operate. Research in this respect will become both more
complex and scale dependent. GIS does not distinguish between appropriate or
inappropriate use of data Orscale.
Thirdly, GIS has the innate capacity to integrate multi-scale data captured at a
variety of scales. But this implies that the implications of using data captured at different
scales must be fully understood. Given that scale represents the extent of detail in which
the digital features reflect the real world, the issue of generalization has significant impact
on the integration and use of multi-scale data. A well-known question to ask geography
students studying the fractal nature of landscape is to determine the length of the coastline
of the US. The answer is, of course, that it is infinite - depending on the scale at which it
is viewed. From space the distance will be much less given the need to generalize than if
the coastline were measured from a lower altitude. If the viewing focus were to shift
closer still then the distance will vary depending on whether a cliff edge is followed for
example, or each boulder is circumscribed, or eventually, whether every grain of sand is
measured. This continuum of scales enforces the generalization or simplification of
features as one transitions from one scale to another. Overlaying one data scale on to
another is prone to produce scale artifacts and produce significant error if considerable care
is not exercised.
Over and above the questions raised above, a fourth factor comprising ecological
fallacy and the Modifiable Areal Unit Problem represents significant underlying, but as yet
unresolved, issues concerning the analysis and interpretation of geographical and
archaeological phenomenon. In an earlier study [27] we questioned how several known
archaeological sites under review should be represented within a GIS - either as points or
as polygons. Scale was clearly a crucial question but so too were the larger spatial units to
which the sites were to be aggregated and the concern for capturing the processes under
investigation. Identifying and determining patterns and the causal linkages or mechanisms
behind them is essentially related to the scale at which information is collected. No single
mechanism explains pattern at all scales, and these mechanisms typically operate at
different scales than those at which the patterns are observed [28]. Importantly for
archaeologists, certain archaeological patterns only become apparent at a certain scale and
when 'viewed' at a certain 'distance'. As studies move progressively from fine scale to
coarse scale, it is unknown whether patterns and processes change smoothly and gradually,
or abruptly. To move from the individual micro-organism to the macro-scale it is crucial
to understand how information becomes discernible as scale changes from fine to coarse
[29]. The key to understanding how information is transferred across scales is to
determine what information is preserved and what information is lost as one moves from
one scale to the other [28]. The effects of moving from one scale to another has received
some discussion [30], [31], [32] but central to the concern about scale change is the extent
to which patterns and processes at one level may be invalid at another level. This is the
basis of ecological fallacy.
Scale then, is central in capturing and interpreting archaeological patterns and
process in the landscape. Scale and the determination, analysis, and interpretation of
pattern are interwoven. Central to any resolution of ecological fallacy is to recognize that
the problem exists and to be sensitive to analyses that involve the potential for fallacy. For
archaeologists, this poses several core questions not the least of which is what is an
appropriate scale of analysis: whether it be global, regional or local as defined by Allen
perhaps [33]? Furthermore, what are the effects of the chosen scale on the results of the
G. Lock and T. Harris / Introduction
archaeological analysis and can the results achieved at one scale be extrapolated to another
scale? These are significant issues for archaeologists using GIS in their analyses.
2.3 Virtual worlds: Geographic Information Science. the Internet, and archaeology
Another aspect to emerge from the '1999 Ravello meeting was the increasing focus and
discussion on visualization, and in particular Virtual Reality, and the linkage of lmmersive
Technologies to GIS and archaeological analysis. In the 1995 conference some focus fell
upon the linkage of GIS with multi-media as a way of integrating qualitative and
quantitative information. Despite the fact that this union has yet to be fully consummated,
discussions are already striving to move beyond the constraining boundaries of a two-
dimensional static representation of the world. While the iconography of GIS continues to
rest on the series of stacked 2D coverages, in many respects this icon exposes the many
functional limitations that continue to exist in the software. The ability to undertake
spatio-temporal analysis or multidimensional analysis in which aspects of depth or height
can be explored as independent axes and not as graphical derivatives of wire frame models,
are still very difficult to accomplish in contemporary GIS environments.. In many respects
the dynamism and intricacies of spatial data demand that GIS .!D0ve beyond the two-
dimensional world of the map. .
This convergence of technological trajectories is just one of many in which related
spatial data technologies are increasingly complementing and augmenting GIS. As the
twenty-first century begins the breadth of technological advances in these related areas,
their availability to archaeological researchers, and their rapid advancement is remarkable,
Indeed, such is the pace of integration of these technologies such as scientific visualization,
Virtual Reality, the Internet, spatial analysis, and Global Positioning Systems that
increasingly Geographic Information System is giving way to the term Geographic
Information Science. Integral to this term, however, must be the commonality of the
underlying principles, concepts and understanding of space that accompanies this
technology. The integration of the graphical and virtual capabilities of Immersive
Technologies with the powerful spatial data manipulation and spatial analytic capabilities
of GIS promises substantial scientific benefits. These benefits, to archaeologists as well as
the broad user community, will increase significantly if such a coupling can be achieved
and made available and distributed through an Internet environment.
A virtual world is a deliberately designed place. sometimes representing real space,
sometimes abstract conceptualspace, in which objects 'live' through their relationships,
under conditions specified by theworlds' space they cohabit [34: 37]. Virtual worlds can
manipulate and display spatial information in a dynamic environment that enhances rapid
comprehension of complex data. The feeling of immersion is achieved through feeding a
user's senses with images and sound thereby permitting the user total control over the
viewpoint [35: 1009]. Draped Digital Elevation Models area basic component of virtual
worlds and the texture can be derived from map images or aerial photographs or satelJite
imagery. Significantly for archaeologists, virtual worlds combine aspects of human
cognition with the visual auditory images created by computers. As such, Virtual Reality
enables a researcher to 'experiencethe qualities and contents of space..The possibility of
a convergence of the technologies of GIS and VR to create a 'hybrid technology' is
intriguing but brings with it a slew of questions - many of which were raised at the
conference. A central one is how, and whether if at all, these technologies assist in gaining
meaning of the archaeology. This raises again the question of 'whose eyes' we look
through regardless of the sophistication of the images being looked at. As Goodrick and
Gillings have recently pointed out [36] we need to understand what reality is, before we
think about the virtual version.' .
xxii
xxiii
Beyond the Map
Immersive Visualization offers substantial advantages over traditional two-
dimensional computer graphics. The nature of Immersive VisualizationIVR is to
encapsulatethe user inan artificial visual environment. The technologyused tocreate this
illusion can range in scale and sophistication from a pairofgoggles all the way to entire
rooms with massivecomputersdedicated tocomplexrendering and display. Dependingon
the scale and quality ofthe immersion theuser can experiencea variety ofscenarios, from
the visualization of hyper dimensional datasets, complex medical imagery, or even a
virtual recreation ofa Victorian parlor. One unifying factorinall of these scenariosisthat
they all require a massive amount of spatially referenced data. There are enormous
scientific and social benefits to be gained from linking GIS and Immersive Visualization
which would providea powerfulalternative graphical environment heretofore unavailable
to GIS researchers and users for analyzing complex, interrelated spatial datasets. GIS
presents Immersive Visualization with a rich and efficient resource of external spatially
referencedinformation.
Increasingly, it can be expected that the Internet will considerably modify the way
in which archaeological analysis using GIScience will be undertaken. GIS developments
havequicklyfollowed therapid expansionofthe Internet and Internet GIS is already well
established. The opportunities for distributed database access, distributed processing, and
the field acquisition of data through wireless technologies combineto make GIS a rapidly
mobilizing technology [37], [38], [39], [40]. The challenge within the context of this
research willbe toaddress thepotential benefitsofVR-type visualizationas itpertains toa
platform-neutral, Internet-based environment [41], [42], [43]. It is not too far fetched to
already conceive of the coupling ofGIS and VR within a collaborative Internet-based
environment.
The data models and data structure involved in establishing connections between
VR and GIS lead to two broad model groups comprising loosely coupled and tightly-
coupled models [44]. In a loosely coupled scenario, the GIS graphics engine is isolated
from the display engines ofthe VR scenario. This leads to a very difficult and two-fold
interface that challenges the user to master, and the developer to create within. The two
units are inherently separate. The tightly coupled model presents a much more pleasing
situation where the user cannottell the differencebetween the GIS that is providingdata,
and the VR system that is displaying it. In this scenario, the best of both systems
complement one another. The GIS provides data seamlessly by direct access to the
graphics engine while the graphical interfacefrom the VR system provides access toGIS
functionalitytotheuser inthevirtualenvironment.
An integrated GIS- VR system would considerablyfacilitate the spatial referencing
of virtual display components; the delivery of multi-scale, multi-thematic geographical
information; landscape and feature depiction; analytical and spatial search capability;
dynamic modeling and simulation capability; and the inclusionofreal world connectivity.
The virtual navigationcapabilityinherent toVR providesa powerful means to interact and
explore the informationstored within the GIS. The imrnersive nature ofVR considerably
augments the graphical display of aGIS to providea more intuitive portrayal of complex
spatial and archaeological information. However, in the use of VR the line between real
and artificial worlds becomes increasingly fuzzy and difficult to define. Considerable
debate ensuedat the meeting regarding the 'accuracy' or more precisely, the 'inaccuracy'
ofVirtual Reality indescribing archaeologicalworlds.
3.Conclusion
As shown, the use of VR in archaeological analysis has generatedconsiderable debate in
this context and yet, to summarise much of the above, little has been raised to date
G. U;lclc and T. Harris / Introduction
concerning the assumptions underpinning the use of GIS in archaeological analysis.
Conclusions regardingcontinuityandchangeover thelast six yearsmust be that while the
applications of GIS and related spatial technologies are flourishing in several different
areas of archaeology, this is at a superficial level of understanding. As the discussion on
data and scale issues indicates, archaeologists have far from addressed the fundamental
issues raised six years ago surroundingGIS interpretationsofthe world, irrespectiveof the
complementarydebate concerningthe 'accuracy'ofvirtual worlds.
This returns us full circle to the philosophical debates generated by GIS and
Society issues in that archaeologists.must be cognizant of the assumptions embedded in
GIS and the unintended outcomes ofwhat at first appears to be an objective science.
Considerationofthe theoreticalissues associatedwithGIS is anecessaryprecursorto wise
useof thetechnologyinarchaeologicalanalysis.
References
[I) Lock.G. andSlarlCic,Z (eds). Archaeology and GeographicallnformaJion Systems: a European Perspective,
Taylor andFrancis. London, 1995. .
(2) Harris, T. and Lock. G. Toward an evaluation of GIS in European the past, present and
future of theory and applications: in Lock. G. and StanCic,Z (eds). Archaeology and Geographical
In/onna/ion SySUms: a European Perspective, Taylor andFrancis,London,1995,pp.349-365.
(3) Scollar, I. Twenty Five years of Computer Applications to Archaeology, in Dingwall, L., Exon, S.,
Gaffney, v. Lallin, S. and van Leusen, M. (eds), Archaeology in the Age of the Internet. CAA97.
Complller Applications and QUQIIJitative Methods in Archaea/ogy. Proceedings of the 25'" Anniversary
Conferetu:e University of BirminghQm. April 1997. Oxford: BARInlernationalSeries7'50,1999,pp. 5-10.
(4) Barcelo,J.A., Briz, I.And Vila, A.(eds), New Techniques for Old Times. CAA98.Computer Applications
and QUQIIJitative Methods in Archaeology. Proceedings of the26'" Con/eretu:e. Barcelona. March 1998.
Oxford: BARInternationalSeries 757, 1999.
[5] Johnson, I. and North, M. (eds). Archaeological Applications of GIS. Proceedings of Colloquium ll,
UISPP Xlllth Congress, Forti, Italy, September 1996. Sydney University Archaeological MethodsSeries
5,Sydney, 1997. CD-ROM.
[6] Bampton.M. Archaeologyand GIS: The view fromoutside,Archeologia e Calcolatori, 8,1997,pp9-26.
[7] Gillings, M. and Wise, A. (eds). Archaeology Data Service GIS Guide to Good Practice. AHDS and
OxbowBooks, Oxford, 1999. .
[8] Westcott, K.L. andBrandon, R.J. (eds), P.;actical applications of GIS for Archaeologists. A Predictive
Modeling Kit. London: Taylorand Fr.incis, 2000. .
[9] Ebert, 1.1. TIle state of theartin 'inductive' predictivemodeling: seven big mistakes (and lots of smaller
ones), in Westcott, K.L. and Brandon, RJ. (eds). Practical applications of GIS for Archaeologists: A
Predictive Modeling Kil. London: Taylorand Francis, 2000, pp.I29-134.
(10) Church, T., Brandon, R.J. and Burgett, G.R. GIS.applications in archaeology: method in search of
theory, in Westcott, K.L.and Brandon, RJ. (eds). Practical applications of GIS for Archaeologists. A
Predictive Modeling Kit. London: TayloraDdFrancis, 2000, . .
[11] Hodder, I., Shanks, M.,A1exandri, A., Buchli,V., Cannan,J., Last, 1. and G. Lucas(eds). ItIle'Pmlng
Arc:lrMoiogy. Flllding meaning In Wpasl, Loodon:Root1edge, I99S.
[12] Chrisman, N. R. Design of information systems based on social and cultural goals, Photogrammetric
Engineering and Remote Sensing, 53, 1987,pp. 1367-1370.
[13] Openshaw, S. A view on iheGIS crisis. in geography, or, Using GIS to pUIHumpty-Dumpty back
togetheragain, Environment and PlDnning A,23, 1991, pp.621-628.
[14] Edney, M.H. Strategies for maintaining the democratic utlD'e of information systems,
Papers and Proceedings of theApplied Geography Con/eretu:es, 14,1991, pp. 100-108.
[15] Goodchild,M.F.Just the facts, Politicai Geography Quarterly, 10(4), 1991, pp. 335-337..
xxiv
Beyond the Map
[16] Pickles, J. Geography, GIS, and the surveillant society, Papers and Proceedings of the Applied
Geography Conferences. 14, 1991, pp. 8091.
(17) Lake, R. W. Planning and applied geography: positivism, ethics, and geographic information systems,
Progress in HUI1UllI Geography, 17,3, 1993, pp. 404-413.
(18) Pickles, J. (ed). Ground Truth. The social implications of Geographic Informalion Systems, The
GuildfordPress, New York, 1995.
(19) Friday Harbor Workshop, Geographic Information and Society: a workshop, National Center for
Geographic Information and Analysis. Unpublished proceedings of a workshop held at Friday Harbor,
Washington,Nov 11-14, 1993.
(20) Taylor, P.J., GKS, Political Geography Quarterly, 3, 1990, pp. 211-212.
(21) Taylor, PJ. and Overton, M. Further thoughts OnGeographyandGIS, Environment and Planning A, 23,
1991,pp. 1087-1094.
(22) Gaffney, V. and van Leusen, P.M. Postscript - GIS, environmental determinism and archaeology, in
Lock. G. and ~ Z. (eds), Archaeology and Geographical InjrJrmaJiqn Syskms: a Europea
Perspective, TaylorandFrancis, London, 1995, pp. 367-382.
(23) Mark, D. M., On the Ethics of Representation: or wlKtseworld is it anyway?, Geographic Infonnation
and Society: a workshop, National Center for Geographic Information and Analysis. Unpublished
proceedingsofa workshop heldatFriday Harbor, Washington, Nov 1114, 1993,unnumbered.
(24) Harris, T.M., Weiner, D., Warner, T.A. and Levin, R. Pursuing social goals through participatory
Geographic Information Systems. Redressing South Africa's historical political ecology, in Pickles, J.
(ed). Ground Truth. The social implications of Geographic InjrJrmalion Systems, The Guildford Press,
New York, 1995, pp' 196-222.
(25) Bender, B.Stonehenge: making space, Berg,Oxford, 1998
(26) Sheppard, E., GIS and Society: ithal and reality, Geographic Informalion and Society: a workshop,
National Centerfor Geographic Infonnationand Analysis. Unpublishedproceedingsofa workshop held
atFriday Harbor, Washington, Nov 11-14,1993, unnumbered.
127)Harris,T.M.andLock.G.R.TowardaregionalGISsiteinfonnalion retrievalsystem:the OxfordsltireSites
andMonumentsRecord(SMR)prototype,inC.larsen(ed),Sisesand MtJtUunmts: NalionolArchaeological
Records, The NationalMuseumofDenmad,DKC, Copenhagen,1992, pp.185-199.
[28) Levin, S. A.,The problem ofpaltemandscaleinecology, Ecology, 73, (6),1992,pp. 1943-1967.
(29)Jarvis, P.G.and McNaughton, K.G. Stomatal control of transpiration: scaling up from leaf to region,
Advances in Ecological Research, 15, 1986, pp.1-49.
[30) Haggett, P., Scale components in geographical problems, in Chorley, R. J. and Haggett, P. (eds)
Frontiers in geographical teaching, London, Methuen, 1965, pp. 164-85
[31) Harvey, D. W., Pattern, process and the scale problem in geographical research, Transactions of the
Instilute ofBritish Geographers, 45,1968,pp.71-8.
[32) Hudson, J. Scale, space and time, in Abler, R.E, Marcus, M. G. and Olsen, J. M. (cds) Geography's
inner worlds: pervasive themt!s in contemporary Amt!ricon Geography, Rutgers University Press, New
Brunswick,NJ, pp. 280-297.
(33) Allen, K.M.S. Considerationsof scale in modeling settlementpatterns using GIS: an Iroquois example,
inWestcott, K.L. and Brandon, R.I. (eds), Practical applications ofGISfor Archaeologists. A Predictive
Modeling Kit. London: Taylorand Francis, 2000,pp. 101-112.
(34) Jacobson, R.Virtual worlds capturespatial reality,GIS World, Dec 1994, pp.36-39.
(35) Neves, J.N. andCamara. A.Virtual environmentsand GIS, inMaguire, D., M. Goodchild, and D.Rhind
(Eds.)Geographical Infonnation Systems, 1999, pp.557565,Longman, London.
(36) Goodrick, G.and Gillings, M. ConslruCts,simulations and hyperreal worlds: the role of Virtual Reality
(VR) in archaeological research, in Lock,G. and Brown, K. (eds) On the Theory and Practice of
Archaeological Computing, Oxford UnivecsityCommilleeforArchaeology, Oxford, 2000.
(37) Gardels. K., A Comprehensive Data Model for Distributed, Heterogeneous Geographic Infonnation,
httpJlwww.regis.betke1sy.ec!ulllanje!slgeomo<!eldeChqpl., 1997.
xxv
G. Lock and T. Harris Iintrodu<:tion
(38) Laurini R. Distributed Geographic Databases: An Overview. The AGI Source Book for Geographic
Informalion Systems, The AssociationforGeographic Information. 1995,pp.45-55..
(39) Onsrod, H. J., and G. Rushton, (eds.)Sharing Geographic Infonnation Systems. New Brunswick, NJ
Center forUrban Policy Research Center, 1995.
(40) Plewe, B., GIS Online: InfOrmalion retrieval, mapping, and the Internet, 1997, Onword Press.
(41) Annitto, R., and B.Paaerson.,Anewparadigm'forGIS dataCommunications'-Joumalofthe Urbanand
Regional Informalion Systems Association 7 (I),1995,pp.64-67.
(42) Rodriguez, 0., Integrating Visualization into !he EnvironmentalInformation Flow. Geo Info Systems.
7(1),1997. .
(43) Shiffer, M. J.,MultimediaGIS forplanning supponand public discourse. Cartography and Geographic
Informalion Systems. 25(2), 1998, pp.89-95.
(44)1ltorpe,S. R.and H.A.Karimi, On tightly couplingmodels with visualization: thepackage for analysis
and visualization of environmental data. Qlrtogra{ihy and Geographic Infomialion Systems. 24(4),
1997,pp.195-202.
Beyond the Map
G. Lock (Ed.)
lOS Press, 2000
Vision, perception and GIS:
developing enriched approaches to
the study of archaeological visibility
David Wheatley
DepartmentofArchaeology
UniversityofSouthampton
. Mark Gillings
School ofArchaeologicalStudies
ofLeicester
Abstract: At present GIS-based visibility studies run the very real risk of becoming
an unquestioned, and deeply uncritical,orthodoxy within archaeological practice.
An orthodoxy whose true utility to the discipline may well be obscured by an almost.
formulaic application. The purpose of the present discussion is to address this .
imbalance. In the first section the major critiques and considerations involved in
thinking ahout and analysing visibility are assembled and reviewed. Having
explored these issues, discussion then moves beyond critique to explore in detail
how these identified shortcomings may be addressed through the development of a
number of enriched approaches to the generation and analysis of viewsheds.
Through the development of a new archaeological application, the 'Higuchi
viewshed', issues such as directionality and clarity will be explored building upon a
large body of research which has been undertaken in the fields of Landscape
Planning and Cognition. .
1. Introduction
"There is more to seeing than meets the eyeball" [8:5].
Over the last decade, the rise in popularity of GIS-based approaches to the analysis and
exploration of past landscapes has been dramatic. If one unique characteristic can be
identified in this process it has been the application of visibility and viewshed analyses.
These have become the most frequently cited 'response to those who question whether GIS
represents a true methodological advance, or simply increased efficiency in the spreading
of dots across maps. The suggestion is that it is through viewshed analysis that the GIS
makes its most unique and valuable contribution to landscape study. .
It is important to acknowledge that visibility-based approaches have a long
pedigree in archaeology and are not exclusive to GIS. A considerable body of analyses had
been undertaken long before the appearance' of GIS, varying from the observations' of
antiquarians, to the rigorous quantitative methodologies proposed by researchers such as
Renfrew, Davidson, and Fraser [46]; [19]. More recently, the importance of vision - in the
sense of visual perception - has been fore-grounded within cognitive archaeology [<J7] and
2
3
Beyond the Map
in theoretical debates concerning the relationship between bodily experience and
understanding, in which the role of vision has been re-examined in the context of narrative
and framing devices. Whilst GIS-based approaches to visibility have been highlighted as a
dramatic methodological advancement, it should be acknowledged that the solution offered
by the GIS as to the problem of how to quantify and represent visibility mirrors precisely
that adopted by archaeologists in the early 1970's. Like GIS-based solutions, these were
based on the measurement and analysis of line-of-sight (intervisibility) and field of view
(viewshed) elements. In drawing attention to this heritage, the aim is not to deny the utility
of such analyses or dilute their impact. Rather, it is to emphasise that the GIS-based
analyses of visibility currently being developed are not 'new' in any theoretical sense, and
that the explosion of interest in visibility, sparked by the widespread adoption of GIS, has
not taken place in a disciplinary vacuum.
A considerable body of criticism and issues arose in response to these earlier
studies, all of which have equal validity to the more recent GIS-based work. What is
intriguing is that there has been a marked lack of interest in building upon these pioneering
studies or addressing the (often substantial) bodies of criticism which grew up with respect
to them. Instead the relative ease with which such analyses can be undertaken within the
GIS has led to a familiar sense of push-button functionality with no underlying
archaeological purpose, as ever more papers become choked with ever more viewsheds.
We firmly believe that rather than simply repeating or routinising earlier
methodologies, GIS-based approaches to the study of vision do have the potential to
revolutionise our understandings of past landscapes. To realise this we must regard generic
GIS routines as stepping-stones towards the development of enriched, and specifically
archaeological approaches to the study of visibility: approaches that acknowledge and
consider the varied critiques that have arisen in response to attempts to explore the role of
visibility in the past. As a result the aims of this paper are twofold. First we offer a
thorough review of the varied critiques and issues which have arisen in response to
attempts to integrate visibility-based approaches into archaeological research. Secondly we
suggest how we may begin to address and constructively incorporate these factors into the
creation of enriched visibility studies. The second section is not intended to address all of
the issues raised in the first, but instead to offer a single example of how such new
approaches must stem from archaeological theories and questions.
2. Issues in GIS-based visibility-viewshed analyses
As intimated earlier, a considerable body of critique and criticism has developed over the
years, as concerted efforts have been made to incorporate visual phenomenon into
archaeological analyses. Some of these criticisms are undoubtedly debatable, others largely
intractable, whilst some are easily remedied. In various combinations they comprise the
raw materials from which the more ill-informed, disingenuous, and generally unpublished,
criticisms (for example' why bother to use a GIS when you can go and look yourself) have
been assembled. For the purposes of discussion, we class these issues into three groups,
though it should be noted that considerable overlap exists between them. We will term
these categories pragmatic. procedural and theoretical. Pragmatic concerns are those that
arise when discussing visibility itself and are consequently equally applicable to GIS and
non-GIS studies. These often date back to the earliest attempts to explore the phenomenon
of visibility in a quantifiable way. In contrast, procedural issues refer to those additional
concerns that arise as a product of operationalising the analysis of vision within GIS. The
final category, theoretical, arises from debates both in archaeology and neighbouring
disciplines such as humanistic geography and the philosophy of science. Some of these
D. Wheatley and M. Gillings I Vision. perception and GIS
have already been isolated and discussed by the authors in a broader exploration of the
historical relationship between archaeology, visibility and GIS [26]. Whilst not claiming to
be exhaustive, the present discussion seeks to elaborate and expand upon these points
through a more thorough review of the issues that have been raised. Taken as a block, these
form a formidable body of critical considerations that must be considered carefully if GIS-
based visibility studies are to develop and proceed.
3. What is visibility?
Like many terms in routine archaeological usage, 'visibility' has a 'variety of meanings,
and it is important to clarify precisely what we are referring to in the context of the present
discussion. Here the term visibility refers to past cognitive/perceptual acts that served to
not only inform, structure and organise the location and form of cultural features, but also'
to choreograph practice within and around them. In keeping with the principal focus of
existing investigations into the heuristic utility of visibility-based approaches, the spatial
context is that of the inhabited landscape'. This serves to distinguish it directly from any
use of the term to indicate survival into the present within the archaeological record. Less
directly, it is also to distinguish it from visual observations whose immediacy and
relevance is solely to our contemporary experience. The latter ambiguity derive's from an
acceptance of the complex interplay and inevitable interpretative relationship that exists
between such visual perceptions in the present and those we seek to identify and explain in
the past [53:49].
The visual organisation and structuring engendered by these past acts of seeing and
looking can be with respect to (i) other contemporaneous or pre-existing sites and
monuments (ii) natural components of the environment (iii) the positions of heavenly
bodies and astronomical phenomenon or (iv) all of the above. It can be manifested in any,
number of ways, from a suggested emphasis towards visually prominent locations, for
example the construction of a monument on a hilltop or crest, to an explicit concern with
the referencing of other features whether individually or as part of a complex schema of
visual cues and pathways. Examples of the latter may include the siting ofa watchtower so
as to provide a link in a chain of similar structures or deliberate visual association sought
between monuments, themselves perhaps hidden from view, and highly visiblenatural
formations such as rock outcrops. The key factor is that such visual phenomena had
meaning to the societies responsible for the construction of landscape features, and
incorporation of existing natural and cultural features into their conceptual schema'. In this
sense our interest is as much with what has been termed the 'will to visibility' exhibited by
social groups as with the raw physical location of individual features. Here the
effectiveness of a given social act is determined to a degree by its visual impact. It has
been claimed that such a will to visibility manifests itself through a number of 'strategies
of visibility' such as monument construction and the ascription of meaning to natural
places, as outlined above [10]. '
At this point it is interesting to look at how visibility has beeri operationalised
within mainstream and GIS-based archaeological research. In the pioneering quantification
work undertaken in the 1970's, visibility was portrayed as an attribute of the environment.
Particular topographical configurations yielded large and impressive view-fields, others
less so. Visibility was an essentially environmental variable. A second strand in debates
concerning the role of visibility in archaeological research emerged as a result' of
developments in archaeological theory in the late 1980's and early 1990's, broadly termed
post-structuralist. Central to these theoretical debates was a growing appreciation of the
4
Beyond the Map
role of past peoples as active, purposeful agents situated within a meaningful world. In
such formulations, emphasis was given over to experience on the human scale. Visibility
was seen as a fundamentally experiential variable, a component of perception which was
bodily centred and culturally embedded. As a result the choices appear to be rather stark:
visibility as an abstract consequence of the environment; or visibility as a human
perceptual act. What is interesting is that the same fundamental techniques, field-of-view
(Viewshed) and line-of-sight (inter-visibility) have been applied to each of these contexts-
compare Fraser's use of inter-visibility patterns [l9:38IJ with, for example, that of Tilley
[54: 156].
We argue that general or formal methods for analysis of visibility must take
account of both of these and, in doing so, we would point to the work of the psychologist
James J Gibson [22]; [23]; [24]. Gibson rejected traditional theories of visual perception
which assume that the brain processes static, two-dimensional retinal images into a three-
dimensional model of the world - 'computational' theories of vision that draw on image
processing and computer-based image analysis for analogies. Instead, Gibson proposed
that visual perception does not proceed from the interpretation of a static retinal image but
begins with the situation of the perceiver (human or animal) at the centre of an ambient
optical array [23]. This is composed of the entire array of ambient light arriving at a
particular observer after structuring by the surfaces and objects in the world. For Gibson,
the optical array contains higher-order invariant information, and the observer actively
samples the optical array to detect this infonnation
2
A full discussion of the implications
of Gibson's theories of perception to archaeology is beyond the scope of this text, but two
factors at least are of immediate concern to us in the construction of methods for the
analysis of archaeological visibility.
Firstly, Gibson's view of visual perception is far more holistic than traditional
approaches. He rejects the use of simplified schematic images to elucidate responses as
irrelevant to human understanding. Instead of starling from simplified pictures which can,
for example, fool an observer into perceiving a small object nearby in the same way as a
large object far away, he pointed out that observers are situated in complex textural arrays
and are almost never presented with such a problem.' Moreover, Gibson recognised that
visual perception cannot be understood unless we also take account of movement.
Movement of the observer, surfaces or objects is manifested in the ways that textural
elements flow through the optical array. Rather than converting this into the differences
that would be evident between two static images, he saw these flows as primal to the way
in which the observer obtains information about the world. Observer movement is so
central to Gibson's theories that he uses the term 'perceptual systems' [24:53] in
preference to 'senses' to indicate that sense organs cannot be properly understood if they
are separated from the moving, active observer.
Secondly, Gibson considers that the information within the optical array presents
affordances to observers. Thus, to an able-bodied human actor, a chair presents the
affordance of being available to be sat on. It does not necessarily present the same
affordances to other observers (consider a horse, an insect or a wheelchair user, for
example). A key aspect of affordance is that it is dependent on both the perceiver and on
the environment. Objects in the world only afford properties in the context of practical
action. To illustrate with an example, it is undoubtedly true that certain topographical
configurations afford more or less panoramic, directionally constrained etc. views.
However, they only afford these properties in the context of practical action - the selection
of such locales and acts of seeing and looking that take place at these locations are
essentially human perceptual acts, driven by a will to make visible [54:204]. Such an
D. Whealley and M. Gillings I Vision, perception and GIS
integrated conceptualisation is already implicit within a series of recent studies regarding
the landscape context of rock-art (e.g. the recent work ofBradley [6]).
With these general points in mind, we now tum to a consideration of some of the
criticisms that have been aimed at Archaeological visibility analysis, with the hope of
suggesting some of the ways in which we might develop enriched forms of analysis that
address both the specific issues, and our wider theoretical concerns ..
4. Pragmatic critiques and issues
As mentioned earlier, pragmatic critiques are those which apply equally to GIS and non-
GIS based visibility studies. They comprise some of the more frequently cited factors in
any consideration of visibility based analyses and we consider each of them in tum.
Pragmatic Issues:
Palaeoenvironment/palaeovegetation
Object-Background clarity
Mobility
Temporal and Cyclical dynamics
View reciprocity
Simplification
4.J Palaeoenvironment/palaeovegetation
"It is simply impossible to know exactly where the trees and bushes were in relation -to
sites and monuments, where the flowers bloomed and the rushes sighed in .the wind"
[54:73].
One group of criticisms raised in connection with visibility analyses hinge on the fact that
analyses are based upon modem landscape topography and take little or no account of
palaeovegetation. Analysis invariably begins with the topography that is recorded on the
detailed modem base-maps that underlie much traditional work and. provide the
foundational data layers in the GIS. It must be acknowledged that the. landscape form may
have been very different in antiquity. To give a crude example, if you are' interested in
quantifying the visual prospect of a numberof suggested Mesolithic coastal foraging Sites,
it is important to establish whether the coastline of the Mesolithic bore any relation to that
visible today. Although the issue' has been discussed within a number of Gls-based studies
it has rarely featured in debates concerning viewsheds (for a rare discussion in the context
of viewsheds see Lock and Harris [38:217-221]).
The issue of intervening vegetation, what Wheatley has recently referred to as the
'tree factor', is one of the most established critiques within visibility studies (e.g. [49]).
Stated simply, past patterns in the presence and absence of trees and other vegetation could
have had a dramatic effect on patterns of visibility and invisibility. The OEM is a smooth
surface and although techniques exist for the introduction of blocking features, such as
stands of woodland, the analyst is faced with the problem of determining the extent and
height of vegetation cover in antiquity ([38:221-2J; [41:8J; [59:97-8J). The most oft-cited
solution is based around the careful examination of palaeo-environmental evidence (e.g.
Fraser [19]) though this is often difficult to obtain with sufficient spatial resolution.
Rather than a problem, we see these issues as opportunities to refocus the analysis
of visibility in more interesting directions. Itis entirely obvious that individual, calculated
6
Beyond the Map
lines-of-sight in prehistoric landscapes cannot be assumed to be actual lines of sight in
prehistory. Thus, instead of building interpretations from individual observations we must
be specific in the meaning ascribed to line-of-sight maps - often that they represent a
possible line of sight and tum instead to the stochastic patterns of visibility and
intervisibility that are produced through the accumulation of these effects. Additionally, it
must be noted that the effects of vegetation are often seasonal, with the permeability of
vegetation to vision changing over time, so that even if accurate maps of vegetation were
obtained they might add little to our interpretative method. Taken together these factors
serve to introduce a critical temporal dynamic into the patterns of visibility/invisibility for
a certain locale. One final, rather contingent scenario where this limitation can be seen as a
strength is in the use of the GIS to strip away modem tree obstructions enabling views to
be assessed for which modem observation is impossible [40: III].
4.2 Object-background clarity
"Sheppey (n.) Measure of distance (equal to approximately seven eighths of a mile),
defined as the closest distance at which sheep remain picturesque" [I: 122].
Being theoretically able to see something is very different from actually being able to
recognise what it is that you are looking at. Assumptions of 20:20 vision in antiquity are
implicit in the majority of visibility studies. This was a point noted by Fraser when
exploring the intervisibility of prehistoric cairns between the islands of Orkney. He noted
that long sightlines, whilst eminently feasible, often required some degree of prior
knowledge as to the existence and location of the target point, unless marked by a highly
visible feature such as a beacon [19]. In the context of rock-art studies, it is often the case
that visibility factors are calculated for features which are themselves visually elusive (20:
55].
This issue of clarity is important, as a number of archaeologists are beginning to
note the deliberately enhanced visual characteristics of certain structures. For example, in a
recent study of an Iron Age pit alignment, Taylor noted that it was only as one got to
within 200 metres of what appeared to be a single, impermeable barrier, that the individual
components of the alignment would make themselves clear and the permeable nature of the
structure become apparent [51: 197]. Another good example can be seen in the clean chalk
and gravel banks of many Prehistoric monumental structures, which one researcher has
likened to acting as a form of prehistoric neon [II :84]. That such an enhanced visual
impact would have been long lived has been suggested by experimental work [4]. The
issue of object-background clarity and the artificiality of the 'infinite' view has been
commented on by a number of GIS-based researchers (e.g. Wheatley (59]) with a possible
solution suggested in the field of fuzzy viewsheds [50: 142], where clarity decreases with
increased distance [13]; [15]; [16]. Though often cited such approaches have yet to be
applied in the archaeological context.
4.3 Mobility
"I should note in passing that horizons, since they are relative to a place and move as
people move, do not cut the land into pieces. Hence they mark the limits of perception, but
they do not enclose ..." [32:31].
A fundamental implication which comes with the acceptance of visibility as an embodied
perceptual act, is the issue of mobility. We have already drawn attention to the Gibsonian
view of perception that holds movement to be a central component of the 'perceptual
7
D. Wheatley and M. Gillings I Vision. perception and GIS
system' and it is clear that views can change dramatically as a viewer moves through their
environment [54:74]. In all archaeological visibility studies the viewpoint taken has been
static. Despite the fact that care is often taken to ensure that the viewpoint is taken from the
height of an optimum adult (usually between 1.6 and 1.75 metres) in defining the viewshed
that adult is set spinning around a rigidly fixed point. It is important to accept that any vista
is dependant upon bodily facing and our physiologically constrained field-of-view.
Attempts to address the issue of mobility have so far followed two distinct
trajectories. The first is to generate multiple viewsheds or determine intervisibility indices
for points on a pre-determined pathway. Once generated these can be integrated into a.
dynamic sequence [61]. The second is to integrate GIS with other technologies such as
virtual reality modelling [25] which enable a degree of mobility and interaction within the
study area, although at present the link in such studies between quantitative GIS and
qualitative VR is at best tenuous.
4.4 Temporal and cyclical dynamics
As well as observer movement, changes to the ambient optical array are effected by diurnal
cycles, seasonal cycles, climate and weather. As a result, the places that are visible from
(and the affordances presented 'by) a given place change in a partly predictable, partly
random way through time. The predictable element of temporal change to visibility might
be modelled as a series of cycles, beginning with the daily round in which the following
effects recur:
night (none) -7 morning (haze) -7 noon (ideal-visibility) -7 evening (low sun)
This is nested within other rhythms such as the annual, seasonal cycle in which summer
typically provides better visibility, while autumn and then winter provide less ideal
conditions. Overlain on these more-or-less predictable cyclic variations will also be the
less predictable effects of weather.
One possible prototype for such a model might be the regionalised variable theory
of Matheron [42], which models spatial variables in terms of a combination of three
components: (i) a structural component, described by a function m(x) (ii) a random
component that is spatially correlated with this. called e'(x) and a random noise or residual
error not spatially correlated with (i) called e"(x); A mathematical description of the
changes in visibility through time might likewise be composed of three components: (i)
one or more structural functions that describe diurnal and seasonal variations (ii) stochastic
components that are correlated to these just as e'(x) is correlated with m(x) and (iii) a
residual noise similar to Matheron'se'{x).
The parameters for such a model would have to derive from field observations or
consultation of meteorological records, and carry assumptions about projecting these
backwards. Nevertheless, if is possible that this kind of exercise would make patterns of
changing visibility from archaeological sites amenable to more SUbstantiveanalysis just as
regionalised variable theory permits geostatistics using variograms and Kriging.
4.5 Views-to and views-from
Analyses tend to assume perfect reciprocity between a viewpoint and what is viewed, i.e.
all points in a viewshed are equally able to see the viewpoint. This is particularly acute in
the case of GIS-based studies which derive intervisibility uncritically from viewsheds. That
this is not the case has been elegantly demonstrated by Fisher [17:1298] and was noted
archaeologically as early as 1983 by Fraser in his pioneering worlc: on Orkney. In the
'1
8 9
Beyond the Map
archaeological-GIS context, Loots has stressed this differentiation through the use of two
distinct terms to describe a given viewshed: projective (views-from) and reflective (views-
to) [39].
This failure to explicitly differentiate not only serves to raise questions regarding
the validity of many existing GIS-based intervisibility studies, but has important
implications for studies where issues of reciprocity are a significant factor. For example,
the deliberate location of prehistoric tomb sites amidst trees and natural rocks which have a
hidden quality, "sites from which one could view but not be viewed" [53:69]. It could
equally apply to potential hunting sites, where the desire to observe without being observed
is paramount. Confusion between 'views-to' and 'views-from' abound in the
archaeological-GIS literature (e.g. [59:88]; [36:619]. This is compounded by the lack of
many popular commercial GIS packages to differentiate [18:35-6].
Although this is a valid criticism, it is also true that the difference between
projective and reflective viewsheds becomes rapidly less as the distance of view increases
as a proportion to the viewer height. Where the viewer height is negligible with respect to
the distance of the calculation then reciprocity may be an acceptable assumption. In
general, though, we see this issue not as a problem but as another opportunity to develop
enriched methods of analysis. For example, the difference between projective and
reflective views from a given location may itself prove to be a valuable and interesting
analytical product.
4.6 Reduction and simplification
This criticism concerns the implicit assumption of temporal concordance, and the tendency
to group monuments together on broad typological grounds for the purposes of viewshed
and intervisibility study. The result is often a collection of features, spanning many
centuries, which are investigated for visual patteming as if designed and executed
according to a single, coherent plan. This can lead to complex, diachronic landscapes being
reduced to simple synchronic layers. Although monuments undoubtedly persist as
significant landscape and social features, such assumptions must be tested rather than
accepted uncritically. Whilst such questions have been raised in the GIS literature [38:220]
few attempts have been made to give adequate consideration as to their implications
[59:84]; [36:614].
To some extent, of course, this is a reflection of a wider problem within
archaeology: how to explain large-scale spatial and temporal patterns and processes
without divorcing them from the smaller-scale actions that created them. We might also
draw attention to Wheatley [59] to illustrate how an apparently synchronic, and
reductionist method can be used to develop diachronic and historical forms of analysis.
This is because, in developing a method for inferring intervisibility among groups of
archaeological sites, the general pattem is recognised to be the result of individual choices
of location for the long barrows. The method chooses to analyse the compound effects of
choice of monument location over many centuries with the aim of making clear how
individual choices were made: no attempt is made to offer an explanation of the pattern
itself, rather the interpretation is of the individuals and groups who constructed each
monument. Also, while being superficially synchronic, the analysis is implicitly diachronic
because it stems from an assumption that the monuments represent a sequence of
construction. Choice of location for monument construction is understood to have been
influenced by existing monuments within the landscape. Although no specific sequence is
proposed, the method is meaningless unless a sequence existed. What this shows, is that
methods must be firmly grounded in theory. Simplification is necessary and inevitable in
the analysis of archaeological materials because we rarely find evidence for single actions,
D. Wheatleyand M. Gillings/ Vision. perceptionand GIS
and are compelled to interpret these from the (often unintended) consequences of multiple
actions over time and space. At the risk of sounding repetitive, this should be seen not as a
problem but as an opportunity for GIS-based studies to contribute to general archaeological
theorising.
5. Procedural critiques and issues
Within disciplines such as Geography, Cartography and Geomatics, a considerable body of
critical work has developed which is exclusive to studies of visibility undertaken in a
digital environment. The key areas for discussion are as follows:
Procedural Issues:
The Digital Elevation Model (DEM)
Algorithms used to calculate line-of-site and viewshed
The undifferentiated nature of the viewshed
Quantitative rigour
Robustness and sensitivity
Edge effects
5.1 TheDEM
Any formal analysis of visibility undertaken in a GIS is predicated upon the prior existence
of a Digital Elevation Model (DEM) or Digital Terrain Model (DTM). It is important to
realise that in each case we are dealing with a formal model of a terrain prototype. With the
DEM, by far the most common model type used in archaeological visibility studies, this
takes the form of a raster grid (altitude matrix) in which the values of the individual cells
represent regularly spaced samples taken from the prototype terrain. A DTM differs insofar
as it incorporates an explicit representation of the form of the landscape, e.g. cliffs and
ridge lines. The most cornmon form of DTM is the Triangulated Irregular Network (TIN).
This is a vector representation of elevation that models the prototype in the form of
triangular faces derived from the Delauney triangulation of spot-heights or survey points.
The point to emphasise here is that the formal analysis' of visibility within a GIS comprises
the analysis of a model rather than a prototype. In the vast majority of archaeological .
studies this is a DEM. As a result, the; quality and accuracy of the elevation model is a
critical concern in deriving reliable, quantified estimates about the visibility and
intervisibility of archaeological sites. Uncertainties in the DEM are a function' of the
inherent limitations of the model used coupled with the quality of the data used in its
construction [60: 21].
The issue of DEM error, its representation and propagation has been explored in
detail within the general GIS literature [12]; [43]; [60]. Tire more specific effectsof errors
and uncertainty on the derivation of viewsheds have been discussed in a number of
important papers by Fisher [15]; [16]. To give an example of some of the issues that arise..
Woods has pointed out that the DEM is a setof discrete measurements of elevation taken
from a continuous surface. As a result, concordance between model and prototype is
dependant upon the resolution of the DEM and roughness of the prototype surface '[60:9].
As the prototype will always contain detail finer than the resolution of the DEM, the latter
serves to model topography at a certain scale [ibid]. That changing the resolution of a
given DEM has an important impact upon the form of a viewshed determined froni it has
been illustrated in studies undertaken by Fisher [17]. . .
10
Beyond the Map
It is important to acknowledge that the accuracy of the elevation model is not
related in any simple way to the accuracy of visibility predictions. This is for two reasons.
Firstly, small variations in topography near to the viewer are far more likely to have large
effects than similar variations further away. Secondly, hilltops and crests are far more
important in the determination of visibility than valley bottoms or the sides of hills. This
has implications both for the choice of elevation model and the methods used to process it.
There is a clear need to ensure that the elevation values recorded in a given model are
particularly accurate at hill crests and this, in turn, implies that mean-filtering of elevation
matrices to reduce noise is highly undesirable when visibility analysis is to be undertaken.
Although the attenuation of 'noise' in an interpolated terrain model is beneficial in many
situations (derivation of aspect, for example) most noise attenuators (including mean
filters) have the effect of attenuating the areas that most characterise hillcrests. Put simply,
mean filtering elevation matrices tends to lower the tops of hills and raise the bottoms of
valleys - precisely the worst possible outcome for inferences about visibility.
For visibility analysis, it has been suggested by a number of researchers that the
TIN should be preferred over an altitude matrix (e.g. (34]). The TIN allows the forms of
crests and hilltops to be more faithfully represented and allows the sample density (number
of facets and points) to be increased in regions of high topographic variation. Despite this,
published archaeological examples of visibility analysis have almost invariably used
altitude matrices and raster-based GIS for analysis. This is unfortunate, but probably
inevitable, given that raster-based GIS (such as GRASS and IORISI) are more widely
available than vector-based systems with the capability to work with TINs (the most
notable of these being Arc/lnfo and Arcview).
5.2 The Algorithms used to delineate viewsheds
As well as factors related to the choice of surface model, Fisher has shown that variations
in the algorithm chosen to derive a given Viewshed can result in significant differences in
the final form of the mapped field-of-view. In a study of the algorithms implemented
within different GIS packages, Fisher found variations of up to 50% in the area of a given
viewshed, caused by differences in the way in which different implementations determined
factors such as the method used to infer elevation for the DEM , how viewer and target
locations are specified and treated, and how elevations are compared between locations on
the DEM (14]. The importance of accurate and comprehensive metadata accompanying
any visibility study cannot be emphasised enough.
5.3 The undifferentiated nature ofthe viewshed
The traditional result of a viewshed analysis is a simple binary data layer, coded 'I' for
areas in-view and '0' for those out-of-view. The result is extreme and emphatic and the
cut-off between the two zones is clearly demarcated: you can either see a portion of the
land surface or not. In the viewshed there appears to be no room for ambiguity and
uncertainty.
Following on from our discussion ofDEM and algorithm issues, the one factor that
should be clear is that there are potentially enormous levels of uncertainty inherent in any
viewshed calculation. As a result, Fisher has suggested that the simple binary viewshed
should be replaced with what he has termed a 'probable' viewshed (16]. The probable
viewshed is based upon the summation of repeated viewshed calculations taken from the
same viewpoint. These are based upon versions of the DEM which utilise a monte-carlo
simulation approach to incorporate errors taken from the overall error range resulting from
the DEM creation process. The probable viewshed technique offers a methodology for
II
D. Wlreatley and"M Gillings / Vision, perception and GIS
accommodating and representing the errors inherent in the DEM creation process directly
within the viewshed itself.
A further weakness of the traditional, binary viewshed is that it fails to take into
consideration more perceptual aspects of visibility, such as the observed fall-off in object-
background clarity that occurs with increased distance. One solution has been to.
incorporate fuzzy set theory into the representation of a given viewshed (15]. Rather than a
clear cut edge, values at the interface between 'in-view' and 'out-of-view' are graded
between I and 0 depending upon distance from the viewpoint. As Fisher has pointed out,
by combining both the probable and fuzzy Viewshed a much more satisfactory product can
be obtained than that offered by the traditional binary product.
It is fair to say that the time constraints involved in the generation of the multiple
viewsheds required for such representations would, in the past, have limited the routine
applicability of such techniques. However, increasing processing power coupled with
optimised techniques for the rapid calculation of viewsheds (e.g. (33]), are serving to
increase greatly the accessibility of such representations. As a result, applications of the
probable viewshed are beginning to appear in the specialist 4.
5.4 The lack ofquantitative rigour
In highlighting this, it is not our intention to denigrate the more qualitative use of
viewsheds as heuristic devices. It is merely to point out that in GIS-based archaeological
studies purporting to be analytical, little quantitative rigour is exercised in assessing the
validity of claims based upon the study of viewsheds. As Aldenderfer has pointed out, the .
study of multiple viewsheds 'could' be seen as a form of exploratory data analysis, but
more often 'the calculation of viewsheds can be used in lieu of thinking about the problem'
(2: 16]. An exception has been the work of Wheatley on the analysis of cumulative
viewsheds. These summed viewsheds can be treated as a- statistical population of the
number of visible sites of all of the cell locations in the study area, and the individual sites.
as samples of this population. As a result one-sample statistical tests can be undertaken
comparing the site intervisibility data against a background standard .. In practice the
Kolmogorov Smimoffgoodness of fit test has been utilised (58].
5.5 Robustness and sensitivity
This concerns the need to repeat visibility analyses to investigate the effect of factors such .
as changes in viewer and target heights and intervening vegetation. For example, if the aim
was to investigate the issue of intervisibiliiy between supposed signal stations, given
uncertainties as to the height of the original structures and role of factors such as smoke, it
would be prudent to repeat the analysis for a range of viewer and target heights. The
importance of this factor was noted by Lock and Harris intheir detailed study of visibility
patterns in the Danebury region ofSouthern England [38}.
5.6 Edge effects
This important factor in the interpretation of viewshed and visibility patterns has been
noted by Martijn van Leusen. Since viewsheds are generally large relative to the study
region, they tend to 'fall off the edge' of the study area. Conversely, viewsheds of sites
lying outside the region will fall partly within the study area, but those sites are not part of
the analysis so their viewsheds are never calculated. As a result, a degree of caution must
be exercised in the interpretation of apparent visibility patterns (for a detailed discussion
12
r
Beyond the Map
see van Leusen [55). Edge effects are particularly important where cumulative or multiple
viewsheds are generated, and they can be mitigated in a number of ways. The most
obvious solution is to exclude the edge of a study region from the formal statistical
analysis: in other words to generate a 'buffer' around the actual area of interest that ensures
minimal edge impact within the area under consideration. Of course, in situations where
the entire archaeological populations are available - such as islands or discrete spatial
groups of monuments - then the impact of edge effects is minimal.
6. Theoretical critiques and issues
"The eyes have been used to signify a perverse capacity - honed to perfection in the
history of science tied to militarism, capitalism, colonialism and male supremacy - to
distance the knowing subject from everybody and everything in the interests of unfettered
power" [28: 188).
Discussion in this section concentrates upon the more general cntiques raised in the
context of post-structuralist developments in archaeological theory, themselves arising
from wider debates in the humanities as a whole. Although widely discussed within the
context of archaeological theory, such issues have rarely been fore-grounded and explored
within GIS-based studies (for a notable exception see Llobera [36]). As will quickly be
appreciated, although three groupings have been identified for the purposes of discussion,
considerable overlap exists between them:
Theoretical Issues:
Meta-critique of technological determinism
Occularism, visual ism and the nature of perception
The map perspective and the viewing platform
6./ Meta-critique oftechnological determinism
"Multiple viewshed analysis is more the product of the methodological possibilities of a
GIS than of archaeological theory"[56:61).
This comprises a series of well rehearsed cntiques regarding the theoretical
impoverishment of GIS and the primacy afforded to certain theoretical perspectives over
others. This can be attributed to the push-button functionality of GIS and the notion that
the GIS is little more than a neutral, atheoretical tool. Taken together these factors can
result in the inherent strengths and weaknesses of the GIS exerting a strong influence upon
the lines of enquiry followed. A subset of this can be seen in the debates concerning
environmental determinism that dominated discussion in the mid 1990's.
Although this is now largely the province of unhelpful polemic and rhetoric, a
number of researchers have pointed out that in the case of viewsheds there does seem to be
a tendency to regard the only limiting factor as being the time it takes to generate them.
Coupled to this is a sense that if you throw enough accumulated viewsheds at a problem it
is bound to go away. One of the aims of this paper and recent discussions regarding the
historical development of visibility analyses (26) is to establish a constructive framework
for addressing precisely this issue.
13
D. Wheal/ey and M. Gillings / Vision, perception and GIS
6.2 Occularism and visualism and the nature ofperception
"The prioritisation of vision, its separation from and privileging over the other senses, can
be detected in other areas of life, substantiating the claim that we live in a 'specular
civilisation"[52:22).
Drawing on Gibson's ideas, we have already commented on the limitations a
view of visual perception. However, another critique that can be made of visibility analysis
concerns the primacy placed upon vision' over and abo've the other senses. This,' it is
argued, is more a legacy of enlightenment ideals and a Renaissance reformulation of
perception away from the other senses and towards vision, than any a priori hierarchical
ordering of the senses. The question posed is whether in claiming to explore visual
phenomenon in the past we are instead projecting a particular type.of visual understanding,
peculiar to the present? Why, for example, is it assumed that a modem appreciation of,
panoramic and dramatic vistas was shared by the past communities under study. In such
critiques, far from depicting vision as a timeless and neutral act, the act of looking in the
Neolithic being the same as looking today, it is depicted instead as being highly contingent,
imbued with power and deeply political [52); [31: 15-20). Returning .to Gibson, we cannot
assume that the ambient optical (or haptic, or auditory) array presents the same atrordances
to us as it did to prehistoric inhabitants.
It can also be argued that a notion of perception based solely upon what can and
cannot be seen, is a gross and inadequate simplification. A number of researchers have
argued that perception is better conceptualised as a synergistic relationship between visual
(sight), auditory (sound), olfactory (smell) and haptic (touch) phenomena (e.g. Rodaway,
[48]). Reliance upon vision at the expense of the other senses, or perhaps more importantly
believing that there is a reliance upon the visual, can be termed 'visualism'. As has been
noted, this notion of vision as the dominant mode of perception is situated both culturally ,
and historically within contemporary society. What has to be emphasised is that even when
a single sense is claimed as somehow dominant or primary, for example sound to the
forest-dwelling Umeda of New Guinea (21) or vision to modem western Europeans,it
never functions independently of the other, senses, nor is it in any way objective (5).. ,
6.3 The Map perspective and the viewing platform '
"Distance therefore is llilt a line endwise to the eye... To think so is to confuse abstract
geometrical space with the living space 'of the environment. It is to confuse the z-axis of a
cartesian coordinate system with the number of paces along the ground to afixed object"
[24: 117).
Building upon the above, having noted the unique nature of modern visual perception,
many researchers have gone on to highlight a concern that the representational schema'
chosen to represent and depict the world, serve to enforce and legitimate very particular
ways of seeing. Such critiques are principally focused upon the map, but given its reliance
upon the map interface, have begun 'recently to explicitly target GIS. Central to' these
critiques is the idea that a de-corporealisation of vision is enshrined and perpetuated by
such techniques [27:65). In studies of visual phenomena, and in particular those reliant
upon GIS, the perceptual act is reduced to the delineation of a viewshed, which is in tum
presented as an aseptic zone on a flat projected mapsheel. It bas been argued that the
adoption of such a perspective serves to enforce a rigid removal of the observer from the
frame. They become a hovering voyeur, able to see everything from nowhere in particular
14
r-
Beyond the Map
and in no way taking part. Such a perspective has been referred to by a number of writers
as a "God-Trick" [28: 189].
In addition, the objective nature of mapsheets has been brought into question by a
number of authors, e.g. [29]; [45]. In an archaeological context this can be illustrated by
the problems facing landscape archaeologists in Poland during the 1970's and 1980's. As a
result of the pervasive climate of cold-war paranoia, the basic map coverage of the country
was full of deliberate distortions, such as artificially steepened slopes and the addition of
fictional areas ofland [3].
7. Developing enriched approaches to the study ofviewsheds
The aim of the previous section has been to highlight and assemble the principal issues,
limitations and criticisms that have emerged as concerted efforts have been made to
incorporate visual phenomenon into archaeological analyses. Taken as a block they form a
formidable body of critical considerations that must be addressed if GIS-based visibility
studies are to develop and proceed. Whilst not all are applicable to each and every analysis,
we feel strongly that all should be considered before embarking upon a GIS-based
visibility study.
So far so good, but at this point it could be argued that whilst critique is
undoubtedly useful, on its own it is only of limited use. If visibility is to be adequately
incorporated into archaeological studies, and GIS is to provide the ideal mechanism within
which to develop such approaches, then the critiques listed above must be fore-grounded,
confronted and/or embraced. In short, archaeologists must be encouraged to ask whether
undifferentiated viewsheds and calculations of intervisibility are adequate to explore
visibility. Ifthe answer is no then they must be willing to work towards the development of
new approaches.
This process is not new within archaeology. Both Evans and Fraser proposed the
use of distance classes and visual ranges to break down the undifferentiated field-of-view,
and suggested techniques for exploring networks of intervisibility [II]; [19]. Looking
specifically at the GIS, researchers have not been slow in suggesting ways in which to
elaborate and differentiate the Viewshed. This has been through techniques such as
cumulative viewshed analysis and the gradient viewshed [58]; [36].
Although developments are currently taking place, for example the pioneering
work of Llobera in his analyses of the innate affordances of landscape configurations
towards visibility through the study offactors such as prominence and hiding [37], it is in a
rather ad-hoc fashion. What is lacking is any clear and coherent conceptual framework
within which to situate such developments. In the remainder of this paper we seek to
identify precisely such a framework through an introduction to the work of the landscape
planner Tadahiko Higuchi. Using the work of Higuchi as a base, we would also like to take
the first steps towards addressing some of the issues highlighted above, through the
introduction of a series of enhancements to traditional viewshed analysis. The aim is to
replace the undifferentiated viewshed with a field-of-view calculation that is structured
around a number of quantifiable view-distance classes. In so doing, we hope to address
aspects of the criticism of the undifferentiated nature of the binary viewshed highlighted
earlier. In addition, we would like to further investigate structure within a given viewshed
through a simple technique for quantifying any inherent directionality.
In each case an important consideration has been that of accessibility. Both of the
techniques rely upon the application of straightforward map algebra, and techniques
present in the generic GIS toolbox. Development was undertaken using the relatively low-
cost raster system, Idrisi.
IS
D. WhIMlley and M j l i n ~ ~ / Vision, perception and GIS
7.I The Higuchi viewshed
In his study of the prehistoric monuments of Orkney, Fraser offered a series of distance
classes that could be used to decompose a given field-of-view into bands related to the fall
off of visual clarity with distance [19]. Although his technique has found application in a
number of studies of rock-art. [7] this has been to an extent uncritical, as the distance
criteria used by Fraser were purely qualitative and specific to the island environment of .
Orkney.
In seeking to establish a series of distance classes for the decomposition of the
binary viewshed, our aim has been to identify a more objective measure for the
classification of a viewshed into visual ranges. One that can be applied reliably in any
environmental context. Our solution builds directly upon Higuchi's pioneering work on the
visual and spatial structure of landscapes, a concern with the visual structure of landscapes,
i.e. the factors affecting the .appearance of a scene from a freely chosen point of
observation. His aim was to establish a series of indexes that could be used to quantify and
characterise a given visual environment. In practice eight indexes were identified [30: 4],
these are listed as follows: .
The VISUal[;,dexes proposed by Higuchi
I. Visibility/Invisibility
2. Distance
3. Angle of incidence
4. Depth of invisibility
5. Angle of depression
6. Angle of elevation
7. Depth and 3-Dimensionality
8. Light
It can be argued that a number of these indexes have already been investigated in the
context of GIS. For example, the issue of visibility/invisibility, arguably the simplest
index, is already being addressed by the binary viewshed. In addition, the work of Fisher
on horizon, reverse and offset viewsheds, coupled with Llobera's studies of prominence
and hiding, have begun to explore what Higuchi identified as indexes 4 and 7. In ibis paper
we intend to concentrate on the second of Higuchi's indexes, that of distance, elaborating.
Higuchi's description to also incorporate direction, '.
Higuchi acknowledged that the way in which a view unfolds and the quality or
texture of the object viewed varies with (and depends upon) distance from the observer.
Rather than attempt to evaluate this phenomenon on wholly arbitrary or' contingent
grounds, Higuchi sought to determine a standard index that quantified and corrununicated
this effect of longer/shorter distances. One of the major problems faced was in the
selection of a standard object for observation, i.e. an object whose behaviour .in a' given.
view could be used to quantitatively establish a series of classes. Prior to Higuchi's work,
formal studies of distance had been concentrated within the built environment. where
distance was corrunonly restricted by structures and buildings. These studies had tended to
use aspects of the human body as a scale, for example the distance at which a human face
is recognisable.
As Higuchi acknowledged, natural landscapes are much more complex as they.
extend much farther and engage a host of additional complicating factors such as haziness
and object background clarity. As a result, the standard objects used in urban contexts
. i
I

17
16
Beyondthe Map D. Wheatley andM. Gillings ( Vision,perceptionand GIS
could not be applied uncritically to non-urban situations. In response to this, Higuchi
selectedthetreeashisstandardobject.
In practice a given view is divided into 3 components short-distance view
(foreground); middle-distanceview(middle ground) andlong-distanceview(background)
[30: 13-14).
Intheshort-distance rangetrees are recognised as individualentities with discrete
leaves and branches. You can hear the wind blowing through them and see the fluttering
leaves. Objects in the short-distance view are perceived as being immediate and close to
the viewer, engaging all oftheir senses. In quantitative terms the maximum distance at
which visibility can be regarded as short-distance is equivalent to a horizontal angle of
steadygazeof Idegree,orapproximately60timesthesizeofthedominanttreespeciesfor
thearea.
The middle-distance range is that which we most commonly associate with a
pictorial landscape.Theoutlineoftreetopsisvisible butnotthedetailsofindividual trees.
Allsense impressionsotherthan visionplay nopart. Toquote Higuchi"One doesnot feel
but merely views, and the variations in the shape of the terrain become important
compositional elements. We see the forest rather than the trees" [ibid:12). The
juxtaposition oftopographical patterns in the middle-distance serves to give it a strong
sense ofdepth and the effects of factors such as mist and haze also begin play a part in
.
At a horizontal angle of gaze of 3 minutes, equal to a distance of I,I00 times the
sizeofthetree,wemoveintothelong-distancerange.Hereaviewercantellthatanareais
wooded but littlemore. The contours of the treetops cannot beperceived, but the overall
textureisuniformandcolouronlyfunctionsas lighterordarkerpatchesinanoverallblur.
Oneofthemostimportantfactorsisthehorizon.Thereisnolongerasenseofdepth.What
isviewedactsasaverticalbackdrop.
It is important to emphasise that the precise distances involved in distinguishing
betweentherangesaredictatedbythetypicaltreesizefortheareaunderstudy.Asaresult
the extent of the ranges will change depending upon the precise location and prevailing
environmental conditions.
7.2Arecipe: how togeneratean Higuchi viewshed
Generation ofan Higuchi viewshed is a very straightforward operation. The stages are
summarised below. Rather than a full case-study, a sample data set has been used to
develop the Higuchi viewshed, This comprises a section of the route of the prehistoric
track known as the Ridgeway, as it runs along the Wiltshiredowns of Southern England
(Figure I). Inpracticeaseriesofviewsheds havebeengenerated atpointsalong thetrack
whereitshowsclearchangesindirection.
Stagesin the Generationofan HiguchiViewshed
I. Calculatebinaryviewshedforviewpoint
2. Calculateadistancelayerfromtheviewpoint
3. Reclassify the distance layer according to the Higuchi class criteria
(determinedbythetypicaltreeheightfortheareaandperiodunderstudy)
4. Overlay the binary viewshed upon the distance layer to extract the distance
rangevalueswhichfallwithinthein-viewzone
Thefirststage intheprocessistogenerateasimplebinaryviewshedfromtheviewpointof
interest.Adistance layeristhencreatedfromtheobservationpoint.Thelatterencodes
o I 15 2 25
A
"
Figure I. The Ridgeway studyarea.The viewpoints usedintheevaluationare shown numbered.
Also marked isthelocationoftheWorld Heritage Site ofAvebury (A)
eachcellintherastergridwithavalueequaltoitsdistance awayfromtheviewpoint.This
distance layer is then reclassified intoshort,middle and long-distance zonesbased on the
valuescalculatedforthetypicaltree-heightintheareafortheperiod'understudy.
Thefinalstageistouseanoverlayoperationtreatedthebinaryviewshedasamask..
or 'cookie-cutter,' to extract the relevant information from the distance layer, i.e. the
distance bands falling within the areas of the original Viewshed. The process is
summarisedinFigure2, . .
Depending upon theprecise GIS being used to undertake the operation,a number
ofsimple enhancements can be made- to this basic process. For example, in situations
where view reciprocity is deemed important, systems.such as GRASS, ArcInfo and,
ArcViewcan be used to generatethe initialviewshed.Ifyou would liketo'emphasisethe
uncertainty involved in the calculation a probable viewshed can form the basis for the
generation of the Higuchi viewshed, with.fuzzy classificatory criteria used to create
gradualzonesoftransitionbetweeneachoftheclasses.
7.3ThearchaeologicalusesofHiguchi viewshed
By decomposing the Viewshed we can now begin to examine the qualitative visual
relationship between the viewpoint and the features and objects within the field-of-view.
Featureswhichareintheshort-distancerangecanbethoughtofasintegralandimmediate
totheeveryday livesoftheoccupantsoftheviewpoint.Incontrast, featuresinthemiddle-.
distanceformwhatwemightthinkofasthesceniclandscapesettingforagivenviewpoint,
repletewithspatialandtemporaldepthandactingasbothreferentandcontextofmeaning
18
19
,.....
Beyond the Map D. Wheatley and M. Gillings I Vision; perception and GJ!;
The distance layer re-classified into the Higuchi
distance ranges
The distance layer genera led with respect to
viewpoint 3
The binary viewshed overlain upon the Higuchi
distance ranges to yield the final Higuchi
viewshed.
Figure 2. Generating an Higuchi Viewshed.
for a given locale. Features in the long-distance category are those which may be visible
but are not readily identifiable, having lost any distinctive and individual identity.
Let us illustrate these ideas through a brief example concerning the visual
choreography of Bronze Age burial mounds in the siting of a later monument. If an
emphasis upon a direct lineage is important a location may be selected so as to include a
specific mound in the short-distance range. It may then be desirable to ensure that the
burial mounds of the more general ancestors are prominent in the middle-range view,
selling a broader temporal' context for the monument. Acknowledgement of a deeper,
mythical time, may take the form of much earlier long mounds punctuating the horizon of
the long-distance backdrop. .
The Higuchi Viewshed also enables locations at the transition points between the
distance classes to be highlighted, enabling areas of possible Iiminality and transition to be
identified. One challenge will be to incorporate the biasing effects of certain monuments
and topographic configurations upon this simple clarity-distance based calculation. For
example Evans' notion of Neolithic and Bronze Age earthworks acting as a form of
Prehistoric neon [II]. .
In quantitative terms, the Higuchi viewshed also enables a given viewshed to be
characterised according to its overall character. Is it dominated by short, medium or long
views and how do the relative proportions of each class relate? It also has the potential to
provide a much more useful basis for comparison between views taken from different
viewpoints than mere area.
In practice we have quantified the proportions of each class in a given viewshed in
terms of percentage-of-possible-view rather than absolute area. Assuming a tree height of6
metres, the possible view in the short-distance range will always bepi x 0.36 km
2
, or 0.407
km
2
The middle-range will then be pi x 6.60 km
2
- 0.407 km
2
and the. long distance,
assuming a maximum visible distance of l8km, will correspond to pi x 18.00 1an
2
- the
values for the short and middle .distance bands. You are effectively expressing the
proportion of each class as a percentage of a constant and the result can easily be
summarised using a histograrn. Sample histograms for the Ridgeway viewsheds are given
in Figure 3. .
This percentage method could also proveuseful in minimising edge effects
whereby the longer distance classes for a given viewpoint fall beyond the limits ofthe
HiguchiViewshedsforRidgewayViewpoints
100
E
iJ
80
.Std-Short
:::E CD 60
05
~ Std-Middle
~ 40
i1Std-long
it 20
l.
0
2 3 4 56 7 8 9
Viewpolnta
Figure 3. Proportions of viewsbed in each of the Higuchi bands at locations on the Ridgeway
r
21
20
D. Wheatley and M. Gillings / Vision, perception and GIS
Beyond the Map
studyarea.Insuchcases thearea-of-possibleviewwouldnolongerbeaconstant, though
some degree ofcaution should be exercised in determining the inferential value of the
resultsgenerated.
One important considerationis that there willinvariably bea higher percentage of
view in the nearer than farther bands. Comparing, for example, 60% of possible terrain
visible in the short-distance band with 60% in the long-distance as if that constituted an
equality, would be problematic. A solution would be to generate (or estimate) the
population parameters, i.e. generate a series ofpercentage-of-possible view values for
every location(orarandom sampleoflocations)toderivea 'typical' diagram forthestudy
area,Thiscanthenbeusedtoassessdeviation inthearchaeological viewsheds.Thedegree
ofdeviation(statistical significance) canthen becalculated using a one-sample test(in the
caseofapopulation)ortwo-sample test(forasampleofthepopulation).
7.4 Assessing Directionality in a Viewshed
As workprogressed on the Ridgeway viewsheds itrapidly became clear thatcertain ofthe
fields-of-view exhibited a very pronounced directionality (Figure 4). In an attempt to
investigate thisinamore quantitative fashion,wehavedeveloped asimple techniqueonce
again based upon the generic techniques available in any simple raster GIS. The various
stagesintheprocessare summarised below:
Assessing Directionalityina given Viewshed
I. Generateasimple binaryviewshed fromtheobservation pointofinterest
2. Calculateadistance layerfromtheviewpoint
3, Reverse the distance layer so that cells record a decreasing value from the
observation point
4. Derivean aspect layer from the reversed distancelayer, treating it ineffect as a
cone-likeDEM
5. Reclassifytheaspect layer intodirectional zonesasrequired (e.g. N;NE;E;SE;
S;SW;W;NW)
6. Overlay thebinary viewshedoverthereclassifiedaspect layerusing itasamask
toextracttherelevant informationfallingwithinthein-viewarea
7. Produce a histogram or summary statistics for proportion ofcells in the 'in-
view' areawhich fallwithineachdirectionzone.
Thefirsttwostagesare identicaltothoseusedinthegenerationofanHiguchi viewshed.A
simple binary viewshed is calculated along with a distance layer centred upon the
viewpoint location. The next stage is to reverse the distance layer. This was achieved in
Idrisibysubtractingthemaximumcell value(rounded uptoaconvenient integer)fromthe
layer. The result was a series ofnegative values increasing as you move away from the
viewpoint.Bymakingthelayerabsolute(i.e.removingthenegative sign)theresultisan
inverse ofthe original distance layer (i.e, cell values increase as you move closer to the
viewpoint).
The nextstage is totreat thislayer as a DEM(in effect acone with the viewpoint
on top) and derive an aspect layer from it. Once generated, this aspect layer can be
reclassifiedintothe directional zones of interest. In thestudyofthe Ridgeway viewsheds
we reclassifiedthe aspect layer into 8 zones corresponding to N; NE; E; SE; S; SW; W;
andNW.Thebinary viewshed isonceagain usedas amaskin asimple overlay operation
to extract the relevant aspect information for the 'in-view' zones. The process is
summarisedinFigures5aand5b.
Figure4. Viewshedsgeneratedfrom selectedviewpointsalong thecourseofthe Ridgeway
The final stage is toderivesummarystatistics ora histogram fortheproportionof cells of
the 'in-view'area which fallwithineachofthedirection zones.Theresultisaquantitative
summary of the directional trends inherent in.the Viewshed that can be used in a
qualitative orquantitative fashion(Figures6aand6b).
. I
I
22
Beyond the Map
The Aspect layer calculated using the reverse
distance layer as a surrogate DEM. The aspect layer
has been re-elassified into 8 zones: N; NE; E; SE; S;
SW;W;NW
The distance layer for viewpoint 3.
=
The reversed distance layer (i.e. cell values increase
as you approach the viewpoint).
23
D. Wheatley and M. Gillings / Vision. perception and GIS
The binary viewshed is then overlain onto the aspect layer to yield the Directional Viewshed
Figure 5b. Creating a Directional Viewshed

<i'>t- ,
;;-
s,
<v"'" e,,'';;'
forViewpoint3
66
50
"i
1
40
30

o
10
0
#"

,/


Direction.
Figure 5a. Creating a Directional Viewshed
Figure 6a. Quantifying Directionality
.,....
24
25
Beyond the Map
D. Wheatley and M Gillings / Vision, perception and GIS
Viewshed Direction
:90
\
.2 60
-Viewpoint3
! 70
- -
'0
50
/ - - - - - Viewpoint5
&40
- . - -Viewpoint7
.!I 30
-, " .-
- - - - Viewpoint9
i 20 ".-
10 _......

Q. 0
_,o<;
s,

....... 0 o'1>' ,,'Ii o'1>' *"

"'""
c:f
e:,0
e:,0
Bearings
The directionalcharacteristicsof selected viewpointscan bedirectlycompared.
Figure6b. Comparingviewsheds
8. Conclusions
Itisclear that, ifvisibilityisto beadequatelyincorporatedintoarchaeological studies, then
a number of developments must take place. Archaeologists must be encouraged to work
towards the development of new approaches, geared around the specific requirements of
archaeology. These can involve the design and implementation of novel new approaches
(e.g. the work of Llobera, [36]; [37])ortheweaving togetheroftechniques already present
within thecapabilitiesof theGIS (theapproach followed here).
The aims of the present discussion have been twofold. Firstly, tooffer areasonably
thorough critique of current, GIS-based viewshed applications, In doing so, we
acknowledgethat critiquealone isoflimited use. As aresult, thespatial indexes proposed
by Higuchi have been advocated as a framework for the further elaboration of the
Viewshed function. The second aim has been to suggestsome positive ways forward. As
an accessible 'first-step' towards addressing some of the base criticisms of traditional
binary viewshed analysis, the Higuchi and directional viewshed techniques have been
offered. It is hoped that these will not only serve as a set of useful heuristics, but more
importantlyactascatalysts for further development.
Notes
I. It is perhaps 0.01 surpnsmg that the majority of visual approaches are concerned with landscape
archaeologygiven that a direct equation is often drawn between the conceptof'landscape' and that of
'field-of-view'. This has much 10 do with the origin ofthe term 'landscape' within art, a topic much
discussed by archaeologists and geographers (for a recent discussion see Lemaire [35J; Cosgrove [9]:
Ingold [32]).
2. Note that, unlike Webster [57], we see that it might be possible tofollow a Gibsonian view ofvisual
perception without necessarily subscribing wholesale to his controversial, theory ofdirect perception.
This was only one componentofhis 'EcologicalApproach to Visual Perception' [24] and proposes that
the information in the optical array evokes some kind ofresonation in the animal, rather than being
sampledfrom the optical array and then processed into meaningful information.
3. To illustratethe comedyvalue ofan observer who actuallycould confusethese situations, we needonly
toconsiderthe scene inChannel4'spopularTV series'FatherTed'.When Father Dougal isconfusedby
the view from acaravanwindow,FatherTed replies along the lines of'Howmanytimes do I have totell
you, Dougal.Big cows: nearby. Small cows: far away'.
4. Rather confusingly the authors refer to the technique as fuzzy viewshed analysis, though it should be
acknowledgedthat Fisherhimselfconfused the two inearly discussions[16: footnote 1].
References
[I] Adams,D.and J.Lloyd, The Meaning ofLiff. Faber & Faber, London, 1983.
[2] Aldenderfer, M., Introduction. In: M. Aldenderfer and Hp.G. Maschner (eds), Anthropology, Space.
and Geographic Information Systems. Oxford UniversityPress, New York, 1996, pp.3-18.
[3] Barford,P., W. Brzezinskiand Z. Kobylinski,The 'past, present and future ofthe Polish Archaeological
Record project. In: J. Bintliff, M. Kuna and N. Venclova (eds), The future of Archaeological Field
Survey in Europe. Inpress.
[4] Bell, M., PJ.Fowler and S.W. Hillson, The experimental earthwork project 1960-1992. CBA, London,
1996.
[5] Bender, 8., (ed.), Landscape - Politics and Perspectives. Berg, London, 1993.
[6] Bradley, R.,Rock art and the prehistory ofAtlantic Europe: signing the land. Routledge,London, 1997.
[7] Bradley, R., J. Harding, S. Rippon, S. and M. Mathews, M, A field method for investigating the
disuibutionof rock art. Oxford Journal ofArchaeology 12(2) 1993,pp. 129-143.
[8] Chalmers,A., What is this thing called science? Open UniversityPress, Buckingham, 1999.
[9] Cosgrove,D., Inhabitingmodemlandscapes.Archaeological Dialogues 1997-1, 1997,pp. 23-26.
[10] Criado, F., The visibility ofthe archaeological record and the interpretation ofsocial reality. Iii: I.
Hodder, M. Shanks, A. Alexandri, V. Buchli, J. Carman, J. Last and G. Lucas (eds), Interpreting
Archaeology: finding meaning in the past. Routledge,London, 1995pp. 194-204.
[II] Evans,C;Tradition and the cultural landscape: an archaeologyofplace, Archaeological Review from
Cambridge 4(1),1985,pp. 80-94.
[12] Fisher, P., First experiments in view;hed uncertainty: the accuracy of the viewable area,
Photogrammetric Engineering and Remote Sensing 57, 1991,pp. 1321-1327. .
[13] Fisher, P., First experimentsin viewsheduncertainty: simulating the fuzzy viewshed,Photogrammetric
Engineering and Remote Sensing 58,1992,pp. 345-352.
[14] Fisher, P., Algorithm and implementation uncertainty in viewshed analysis, International Journal of
Geographical Information Systems 7 (4), 1993, .pp. 331- 347:
[15] Fisher, P., Probable and fuzzy models ofthe viewshedoperation. In: M.F. Worboys (ed.), Innovations
in GIS: selected papers from the First national Conference on GIS Research UK. Taylor & Francis,
London, 1994,pp. 161-175.
[16] Fisher, P., Anexploration ofprobableviewsheds in landscapeplanning,Environment and Planning B:
Planning and Design 22, 1995, pp. 527-546. .
[17] Fisher, P., Extending the applicability of viewsheds in landscape planning, Photogrammetric
Engineering & Remote Sensing 62 (11), 19968,pp. 1297-1302. .
[18] Fisher, P., Reconsideration ofthe viewshed function in terrainmodelling, Geographical Systems 3,
I996b,pp. 33-58.
. I
I
r
27
26
D. Wheatley and M. Gillings / Vision, perception and GIS
Beyond the Map
[19] Fraser, D., Land and Society in Neolithic Orkney. British Archaeological reports (British Series 117),
Oxford,1983.
[20] Gaffney, V., Z. StanCicand H. Watson,Movingfrom catchmentsto cognition: tentativesteps towards a
largerarchaeologicalcontextforGIS, Scottish Archaeological Review 9/I0, 1995, pp. 41-64
[21] Gell, A., The Languageofthe Forest: Landscape and Phonological Iconism in Umeda. In: E. Hirsch
and M. O'Hanlon(eds), The Anthropology ofLandscape. ClarendonPress,Oxford, 1995, pp. 232-254.
[22] Gibson,J., The Perception ofthe Visual World. Greenwood Press, WestportCT, 1950.
[23] Gibson,J., The senses considered as perceptual systems. Houghton Mifflin,BostonMA, 1966.
[24] Gibson,J., The Ecological Approach to Visual Perception. LawrenceErlbaumAssociates,HillsdaleNJ,
1979.
[25] Gillings, M. and G.T. Goodrick, Sensuous and reflexive GIS: exploring visualisation and VRMI..
Internet Archaeology I, 1996.
[26] Gillings, M. and D. W. Wheatley, Seeingisnot believing: unresolved issues in archaeological visibility
analysis. In: B.Slapsak(ed.), On the good use ofGIS in Ancient Landscape Studies. Ljubljana, inpress.
[27] Gregory,D., Geographical Imaginations. Blackwells,London, 1994.
[28] Haraway, D., Simians. Cyborgs. and Women: the reinvention of nature. Free Association Books,
London, 1991.
[29] Harley, J., Maps, Knowledge, and power. In: D. Cosgroveand S. Daniels (eds), The iconography of
landscape. CUP,Cambridge, 1988, pp. 277-312.
(30] Higuchi,T., Visual and spatial structure oflandscapes. MIT, Massachusetts, 1983.
(31] Ihde, D., Postphenomenology: Essays in the Postmodern Context. Northwestern University Press,
Evanston, Illinois, 1993.
(32] Ingold, T., The picture is not the terrain: maps, paintings and the dwelt in world, Archaeological
Dialogues 1997-1, 1997,pp. 29-3I.
(33] Lake, M., P.E. Woodmanand SJ. Mithen, TailoringGIS Software for archaeological applications: An
ExampleConcerningViewshed Analysis,Journal ofArchaeological Science 25, 1998, pp. 27-38.
(34] Lee, J., Analyses ofvisibility sites on topographic surfaces, International Journal of Geographical
Information Systems 5(4),1991,pp. 413-429.
(35] Lemaire, T., Archaeology between the invention and the destruction ofthe landscape, Archaeological
Dialogues 1997-1, 1997,pp. 5-21.
(36] L1obera,M., Exploringthe topographyofmind: GIS, social space and archaeology,Antiquity 70, 1996,
pp. 612-622.
(37] L1obera, M., Landscapes of Experiences in Stone: notes on a humanistic use of a Geographic
Information System to study ancient landscapes. Unpld. PhD Thesis. Oxford, 1999.
[38] Lock, G. and T.M. Harris, Daneburyrevisited: an English Iron Age hillfort in a digital landscape. In:
M. Aldenderfer and RD.G. Maschner, (eds). Anthropology. Space. and Geographic Information
Systems. OxfordUniversityPress,New York, 1996,pp. 214-240.
(39] Loots, 1..,The use ofProjective and Reflective Viewsheds in the Analysis ofthe Hellenistic City
DefenceSystemat Sagalassos,Turkey,Archaeological Computing Newsletter 49, 1997, pp. 12-16.
[40] Madry, S. and I.. Rakos, Line-of-Sight and Cost-Surface Techniques for Regional research in the
Arroux River Valley. In: H.D.G. Maschner (ed.), New Methods. Old Problems: Geographic
Information Systems in Modern Archaeological Research. Center for Archaeological Investigations
Occ,Paper No. 23,SouthernIllinoisUniversityatCarbondale, 1996, pp. 104-126.
[41] Maschner, R, Geographic Information Systems in Archaeology. In: H.D.G. Maschner (ed.), New
Methods. Old Problems: Geographic Information Systems in Modern Archaeological Research. Center
for Archaeological InvestigationsOcc.Paper No. 23, SouthernIllinois University at Carbondale, 1996.
pp.I-24.
[42] Matheron, G., The theory of regionalized variables and its applications.Les Cahiers du centre de
morphologie mathematique de Fontainebleu, Paris: Ecole Nationale Superieure des Mines de Paris,
1971.
[43] Monckton,C; An Investigation into the spatial structure of error in digital elevation data. In: M.F.
Worboys (ed.), Innovations in GIS: selected papers from the First national Conference on GIS
Research UK. Taylor& Francis,London, 1994,pp. 201-211.
[44] Nackaerts, K. and G. Govers, A non-deterministic use of a DEM in the calculation ofviewsheds,
Archaeological Computing Newsletter 49, 1997,pp. 3-11.
[45] Pickles, J., Texts,Hermeneulicsand Propaganda Maps. In: TJ.Barnesand J.s. Duncan (eds), Writing
Worlds: discourse. text and metaphor in the representation oflandscape. Routledge, London, 1992, pp.
191-228.
[46] Renfrew,Ci.Lnvestigations in Orkney. SocietyofAntiquaries, London, 1979.
[47] Renfrew, C., Introduction. In: C. Renfrew and E.B.W. Zubrow. (eds), The Ancient Mind: Elements of
Cognitive Archaeology. CambridgeUniversityPress, Cambridge:Cambridge, 1995,pp. 3-12.
[48] Rodaway,P., Sensuous Geographies. Routledge,London, 1994.
[49] Royal Commission on Historic Monuments England, An inventory of historical monuments in the
county ofDorset, VolumeTwo, South-East,Part 3. HMSO,London, 1970.
[50] Ruggles,C. and DJ.Medyckyj-Scott,Site Location, LandscapeVisibility,and SymbolicAstronomy: A
ScottishCase Study. In: H.D.G. Maschner(ed.), New Methods, Old Problems: Geographic Information
Systems in Modern Archaeological Research. Center for Archaeological Investigationsace.Paper No.
23,SouthernIllinois Universityat Carbondale, 1996, pp. 127-146.
[51] Taylor, J., Space and Place: some thoughts on Iron Age and Romano-British landscapes. In: A. Gwilt
and C. Haselgrove(eds), Reconstructing Iron Age societies: new approaches to the British Iron Age.
Oxbow,Oxford, 1997,pp. 192-204.
[52] Thomas,J., The PoliticsofVision'and theArchaeologiesofLandscape. In:B.Bender (ed.), Landscape;
Politics and Perspectives. Berg, Oxford, 1993,pp. 19-48.
[53] Tilley, C., Art, Architecture, Landscape [Neolithic SWeden]. In:B.Bender(ed.), Landscape; Politics
and Perspectives. Berg,Oxford, 1993,pp.49-84.
[54] Tilley,c.,A Phenomenology ofLandscape. Berg, Oxford, 1994.
[55] Van Leusen, M., Viewshed and Cost Surface Analysis using GIS (Cartographic Modelling in a Cell-
Based GIS II). In: J.A. Barcelo, I. Briz and A. Vila (eds), New Techniquesfor Old Times - CAA 98-
Computer Applications and Quantitative Methods in Archaeology: proceedings ofthe 2f!' Conference,
Barcelona 1998. British Archaeological Reports S757, Oxford, 1999, pp. 215-223.
[56] Wansleeben,M. and I.. Verhart, Geographical information Systems: methodological progress and .
theoreticaldecline,Archaeological Dialogues 1997-1, 1997,pp. .
[57] Webster,D.,The conceptofaffordanceand GIS: anote on L1obera,,4ntiquity 73, 1999,pp. 915-917.
[58] Wheatley,D.,CumulativeViewshedAnalysis: aGIS-basedmethod for investigatingintervisibility,and
its archaeological application. In: G. Lock and Z. Stancic (eds), Archaeology and Geographical
Information Systems: a European Perspective. Taylor& Francis,London, 1995,pp. 171-186.
[59] Wheatley,D., The Use of GIS to UnderstandRegionalvariation inearlier Neolithic Wessex. In;H.D.G.:
Maschner (ed.), New Methods. Old Problems: Geographic Information 'Systems in Modern
Archaeological Research. Center for Archaeological Investigations Occ, Paper No. 23, Southern
Illinois Universityat Carbondale, 1996,pp.75-103.
[60] Wood, J., The Geomorphological Characterisation of Digital Models. Unpld PhD Thesis.
Leicester, 1996.
[61] Yorston, R., Presenting Archaeological Information with Java Applets. In: I.. Dingwall, S.Exon, V.
Gaffney, S. Lafflinand M. van Leusen (cds), Computer ApplicatiOns and Quantitative Methods in
Archaeology: Proceedings of the 25th Anniversary Conference. University ofBinningham. April 1997.
BritishArchaeologicalReportsS750 [CD-ROM],Oxford, 1999. .
r
i
28 29
Beyond theMap
G.Lock (Ed.)
lOS Press, 2000
Perceptionandviewsheds:
aretheymutuallyinclusive?
AndreP.Tschan
Institute ofArchaeology
University of Oxford
Wlodzimierz Raczkowski
Institute of Prehistory
AdamMickiewicz University
Malgorzata Latalowa
LaboratoryofPalaeoecologyandArchaeobotany
DepartmentofPlant EcologyandNature Protection
Universityof Gdansk
Abstract: One of the main themes of this work is the notion that the human
experience, together with the world that surrounds us. forms a conceptual whole
according to an intrinsic and reciprocal relationship of interaction and impaction
between people and their environment. However, in order to achieve any level of
understanding/meaning based on this all-inclusive worldview, itdemands a different
and potentially more complex theoretical framework than is available through
traditional archaeological research. This need for additional theory-driven and
enhanced holistic tools/methods pertains especially to the concept of landscape and
thearchaeologicalinterpretationsofcultural/natural phenomena. This paper attempts
to expand on some of the current methodology using Geographic Information
Systems (GIS) technology by evaluating the potential of Viewshed Analysis in
conjunction with all forms of human perception. The aim is to highlight a hitherto
rarely exploited mutual inclusiveness between GIS and humanistic approaches in
archaeology inpursuit ofanenrichedfutureforapplication developments.
1. Introduction
Anytool available to archaeology requires aclear understanding of its appropriate use, as
well as its overall function in order to achieve its full potential. This basic truism is
especially relevant when considering the increasingly sophisticated electronic means
employed inanalysingthe archaeological record. Unsurprisingly, significant problemscan
arise for anyone who opts to use or include powerful Geographic Information Systems
eGIS)technologyfor hisor her research basedsolelyontheability topress an appropriate
sequence of keys.Inother words, it is substantiallyeasier toacquire the necessary skill to
make a GIS work and produce some results, than it is to fully appreciate the cumulative
processes required in order to arrive at meaningful interpretations or conclusions.
Essentially, it is imperative to keep in mind that each successive step when engaging in
A.Tschan e/at. /Perceptionandviewsheds
GIS-based research is the ultimate consequence of previous decisions and actions,
especially concerning the data selection. At the same time all new choices and processes
directlyaffect thesubsequent results. .
The specific approach that we propose looks at the archaeological record using a
framework where, based on past vegetation characteristics, thechanges inflora represent
protracted timeline stages. This is an important methodological construct since plant life,
unlike other more predictable and static environmental variables, can drastically alter the
outcome of visibility studies using GIS. It applies particularly to applications involving
archaeological viewshedanalysis'wherethe line-of-sight mighteither beclearor restricted
depending onthedifferentfloralspeciesoccupyinga givenspace atagiventime.
Inaddition, plantsalso have adirect bearing on other perceptual senses,especially
one'sability to hearandsmell, Figure 1,andforthatreason vegetation mustbeconsidered
highly rewarding for the purpose of human landscape interpretations, This means that, by
introducing plant lifedistribution patterns as an extra variable beyond the standard binary
requirements ofviewsheds(i.e.,aviewpointandatopographical background),awhole new
set of tools for researching the archaeological record becomes available. The anticipated
outcome of this vegetation-inclusive courseof action is, therefore, instark contrast to any
potentialrepresentational worksoffictionthatengagein'barren landscape' approaches.
2. Istherereally a problem?
The question arises whether or not there is anything wrong at present considering the
recognisable increase in the application of computers in archaeology..Judging from the
consistent level of archaeological computing publications, however, the appeal of digital
tools does notseem to have translated intoa wider proliferation regardingresearch. Ifthe
discipline appropriates technology merely on the principle of availability it may well
enhance the general quantity but might actually affect the qualitydetrimentally. The
sophistication of computer equipment, and especially current GIS-based spatial analysis
techniques, doesnotresultbydefaultinanincreasedunderstandingofthepast bysimply
Figure I. Meansofhumanperception and aviewfro",
White HorseHill.Oxfordshire, England
T
30 31
Beyond the Map
A. Tschan et 01./ Perception and viewsheds
incorporating the latest 'gadgets' [I: 303). Foremost, there is the clear need for a
conceptual understanding of the archaeology in order to translate this infonnation into
meaningful digital data sets. From this purposeful collection of information, results may be
produced bearing in mind that a solid comprehension of the various spatial transformation
algorithms is also required.
The other problem is that recent developments in archaeological theory have
reached daunting levels and may actually impede the possibility of designing
methodologies. This is particularly the case in relation to those works that heavily
emphasise humanistic theories or post-modem philosophy (2) and it would seem that a
rigorous tool such as the computer is poorly suited to answer questions according to these
theory-based approaches of archaeological interpretation. However, there is always the
potential to overcome adversity by bridging this gap and developing generic, as well as
specific methodological solutions that can address and incorporate disciplinary theory
regardless of which 'flavour of the month' currently garners the most popularity.
From the above it is clear that archaeological theory, by either acting as the main
impetus to answer a specific research question or by directly influencing the selection of
the data, lends itself ideally to the crucial stages concerning the collection and preparation
of archaeological information. Within this context, it would be rewarding to investigate all
the different theoretical approaches best suited to viewshed analysis and other GIS
visibility tools used in archaeology. Of much greater interest and importance though is
identifying theory-inspired cultural and environmental variables, the mutual effect they
might have through new combinations, and how this might change the understanding and
interpretation potential for landscapes using GIS-based applications, for example, line-of-
sight tools.
3. The current setting
The focus of this work deals with archaeological landscapes in order to investigate how our
understanding of the past can be enhanced by using all forms of human perception such as
touch, taste, sight, smell and hearing. This obviously requires a substantially more complex
theoretical setting due to the necessity of introducing increased 'realism' that reflects
human experience in the form of an intrinsic and reciprocal relationship of interaction and
impaction with the world that surrounds us.
For the purpose of methodology we look at whether or not the combination of
viewshed analysis and perceptional variables can achieve an enhanced understanding of the
past in contrast to those conclusions generally obtained by more traditional spatial research.
The success of this line of investigation represents a highly desirable recognition of a
hitherto poorly exploited mutual inclusiveness for research concerning past/present cultural
landscapes. However, before presenting a case study from Poland it is important to
highlight some issues pertaining to the status quo and conceptual matters for archaeological
computing, archaeological theory, viewshed analysis and the notion of human landscapes.
3.1 Archaeological computing
The following is by no means a detailed biography of archaeological computing. More
often than not, however, topical aspects or specific tools that are part of this line of
research are very much the product of an historic context. In other words, some identified
shortcomings can be closely linked to a broader set of circumstances and this also applies
to spatial investigations in archaeology. For example, recent critical discussions of GIS'
approaches (3), (4) are very much tied in with the overall developments of archaeological
computing during the 1990's involving a greater emphasis on understanding the concept of
space and human landscapes (5), (6).
This is in contrast to the earliest beginnings in the late "1960's and early 1970's
witnessing the first fruitful attempts at combining digital technology with research in
archaeology [7: 14). What is critical to recognise is the character of this pioneering stage
inasmuch as it heavily emphasised solving technological shortcomings. It means that'
archaeological computing not only dealt with many first-time design issues relating to the
selection and construction of digital data sets, but there was also a struggle with the rather
brutish and awkward nature of the available hardware/software. Consequently, this period
in archaeological computing, time and again, required solutions based on extensive
programming and substantial technical skills.
Continuing this highly generalised history it becomes clear that despite all the
accumulated knowledge and more sophisticated methodologies the above trend by and
large prevailed until the middle of the 1980' s. At this point in time, there is a recognisable
shift toward the application phase with regard to computer usage in archaeology. This
change is clearly related to the increased availability and affordability of the personal
computer (PC) and the subsequent proliferation of ever-increasing powerful software tools
[ibid.). The result was a move toward investigating highly specific archaeological problems
by using commercially available products and to abandon, more and more, the
development of original software programs or hardware-specific solutions. The inevitable
consequence was that archaeology started to give up creating its own set of tools while
siphoning off the technology and associated methodologies from many other sources (8).
Nonetheless, in the wake of this market-driven appropriation of computer
technology there have been many useful advances in archaeological computing. There is,
however, also a sizeable emphasis on replication inasmuch as the same computer-based
approaches in archaeology are often used inter-changeably for entirely different geographic
locations or culture periods/groups. In other words, there has been a marked increase of
computing case-studies that employ digital tools for specific archaeological investigations
but essentially only set out to duplicate or copy already established methods (e.g., the
seemingly insatiable appetite expressed nowadays for electronic topographical surveys or.
Digital Elevation Models (OEMs. .
In many ways this is a. healthy trend when' looked upon as an .indicator of. the
general effectiveness and popularity of computers in archaeology. However, as a reflection
of the present situation it also poses an interesting question with regard to the relationship
between the discipline and digital technology. For example, GIS are highly specific to
investigations concerning geographical data/space and reflect this primary functional
purpose in their overall design. This means that geographers and related earth science
disciplines can advance their knowledge using a powerful tool that is tailor-made for their
areas of research. Ideally, archaeologists should have the same capacity using discipline-
specific software or hardware, although in reality there is no such thing as an
'Archaeological Information System'. Consequently, we are, more often than not, only the
borrowers of technology and methods in the hope that such appropriated tools can be
tweaked or adjusted to fit our particular needs or detailed archaeological research agendas.
Arriving at conceptual and practical dead-ends is, therefore, not an uncommon
outcome and it is clear that archaeology, not the technical limitations of digital tools, has to
be the driving force in the discipline's computing' efforts. 'There is a need for new
application solutions based on an unmistakable archaeological emphasis in order 'to avert
32
33
r
Beyond the Map A. Tschan et 01./ Perception and viewsheds
technology prescribing research parameters. The discipline as a whole needs to further its
own cause through the creation of tools that more specifically address circumstances
pertaining to the archaeological record and not fall prey to the continued indiscriminate
borrowing of gadgets and methods from other sources.
In addition, archaeological computing as part of a larger developmental process
does not adhere to finite levels of technological achievements, particularly when
considering the ever-changing and rapid advancements of software/hardware. The wider
success of computer tools throughout archaeology may reflect a popular functional (and
affordable) application standard while some of the methodological approaches employed
do not actually make use of the fullest potential according to the original design purposes
of a tool. The example used here again involves GIS, which by definition is an active
research tool [5], [6J. Undoubtedly, GIS are also highly effective visual data management
tools because of their design architecture and overall functional characteristics. However,
to relegate a potentially expensive GIS product to just storing data with the added bonus of
a graphical front-end surely minimises the overall power utilisation inherent in these
products.
The current perspective needs to change because many of the computing tools
employed in archaeology will work in some form or another while there is still plenty of
room to maximise efficiency through a more concerted general engagement. This means
that archaeology as a whole should abandon any misconceived beliefs that if something is
unattainable at present then maybe new technological advances in the future will solve the
problem. For many solutions will only come by virtue of a wider discipline-specific
initiative and home-grown ingenuity rather than attempting to absorb the generic products
designed by large international companies appealling to the broadest commercial
denominator.
This very short and broad account of the last four decades of archaeological
computing not only serves to highlight some of the contextual background but also outlines
the next step forward. It seems that all of the above describes a logical growth sequence
and the suggestion now might be a return to the spirit of those early pioneering days
coupled with the vastly improved tools presently at hand. Rather than resting on the
accomplishments already achieved, there is a clear need to harness any established
expertise by designing tools and methods in accordance with the major advances in
computer technology and other academic research fields, as well as developments
throughout the discipline itself, particularly, archaeological theory. This could hail the
meaning phase for archaeological computing where new design efforts, in conjunction with
tested application functionality, achieve enhanced interpretations as the result of computer-
based analyses.
3.2 Archaeological theory and archaeological computing
Archaeological theory has also progressed dramatically over the same stretch of time as
archaeological computing. The earliest conceptions of computer tools for archaeology,
therefore, owed much to the processual paradigm. This theoretical focus was closely
associated with archaeological computing by virtue of the overwhelming quantitative
capacity inherent in computer technology. Clearly, this is neither the place nor time for a
detailed account of all the significant advances in archaeological theory over the last four
decades. However, one consequence needs to be highlighted as many of the changing
theoretical 'flavours' (in particular post-processual) While enhancing the interpretational
understanding sought from the archaeological record also pose a considerable dilemma.
Many digital tools are not readily useful in conjunction with humanistic theories and
philosophies [9], also many recent theoretical 'advances' are entirely devoid of any current
methodologies or even the potential of such. .
At the same time, the importance of archaeological theory can be in no doubt
considering Peer's [IOJ'suggestions with respect to the field of geography:
"... the philosophy of a discipline constantly interacts with philosophy as a discipline and
thus with current ideas in, general. Yet this 'interaction' is most productive when
geographers interpret philosophy to apply its summarized knowledge in furthering the
system of ideas formed through contemplating discrete forms of theoretical practice."
Clearly a discipline-transcending fact, this implies that there will be an ever-increasing
dependence/reliance on the inventiveness of archaeologists trained in computer application,
development and archaeological theory to arrive at new and improved methodological
solutions. And, some recognisable progress in this direction has been made, particularly
since the mid-1990's.
What is important is to design archaeological computing tools that do not just
answer a highly detailed aspect of some theory-based inquiry but create unified application
development frameworks. As mentioned before, 'it is essential to recognise that by
engaging in GIS-based research, the successive steps can be cause and effect
simultaneously. Consequently, there is a clear need for more holistic conceptual and
practical approaches that can describe and explain the transformation and changes
involving the originally selected archaeological information at any given moment of the
study. The idea is that a methodology incorporates archaeological theory as an integral part
of the data selection, processing/manipulation, and analysis/interpretation' stages
respectively. Archaeological data, by nature, can only benefit from this mutual theoretical
and digital compositional understanding during the data collection, transformation, and,
finally, meaning stages through application development. Being able to comprehend
structural and representational changes that affect the data set, by virtue of a
comprehensive understanding of the algorithms applied, aids in the final interpretation of
the results and supports any discoveries regarding the overall meaning.
3.3 Yiewshed analysis
As a fairly recent arrival, GIS-based view'shed analysis has had a very bumpy history to say
the least. The fact that this tool is closely associated with the human capacity of sight
together with the relative ease application results can be achieved has clearly appealed to
many different archaeological research agendas over the last decade. Hence, many diverse'
methodological and theoretical approaches have already been employed for generating
viewsheds (Wheatley and Gillings, this volume). At the same time, because of this generic
suitability for seemingly irreconcilable ideas,' sometimes originating from distant or
opposite ends of the spectrum of archaeological knowledge and practice, viewshed analysis
does not rely on any standardised methodological framework or formalised means for the
interpretation of results.
Rather than presenting the detailed history Of viewshed analysis in archaeology
here, it is sufficient to briefly highlight the essential compositional structures of the two
most common techniques currently applied. In general, viewshed tools either work on a .
purely binary or on an indexed principle and the tYPe of cultural or natural data under
investigation, specific to any.predetermined research question, clearlyaffects the ineaning
34
Beyond the Map
of the results. However, a simple binary line-of-sight analysis does not allow inferences
beyond those areas that can be seen (I) and those that can not (0), radiating outward in all
directions from a selected viewpoint, Figure 2. The second technique, also known as
cumulative viewshed analysis, describes visible locales as the sum of overlying visible
areas according to an index where multiple overlaps are reflected by higher values, Figure
3.
One of the major problems with line-of-sight applications in archaeology is even
more generic than the above structural differences inasmuch as the variable selection for a
viewshed analysis neglects one of the most fundamental and omnipresent features of
landscapes, namely, vegetation. Whether or not this has led to an implicit geographical or
cultural selection emphasis for those areas where the overall a priori knowledge of floral
distribution characteristics (i.e., height, density) suggests no impedance on visibility, may
not be established with certainty. What can be identified though is that many of the major
works on line-of-sight tools clearly ignore the presence of vegetation (Wheatley and
Gillings, this volume).
By assuming that plant life has no, or at best only a trivial, impact on the overall
outcome of a viewshed analysis is to reduce the means of human perception solely to the
capacity for sight. Clearly, by the inclusion of information on past/present floral patterns
the interpretation potential for any body of research is enhanced even when dealing with
the most sparsely vegetated landscapes. This improvement goes beyond just the ability to
see and, if considered as part of any analytical process a greater quality based on more
'realistic' representations showing the various distributions of plants can be achieved.
Viewsheds without vegetation characteristics will, therefore, more often than not, merely
reflect the physical shape of the surface and not access the more rewarding detailed
information that could be extracted as part of the wider landscape context, Figure 4. These
O=Sile O=Site
Light areas = visible from site (I) Lightest areas = Visible from 3 sites (3)
Dark areas = nol visible (0) Light areas = Visible from 2 sites (2)
Dark areas =Visible from 1 site (1)
Dark areas = Not visible (0)
Figure 2 (left). A simple viewshed analysis
Figure 3 (right). A cumulative viewshed analysis
r
35
A.Tschan et 01./Perception and viewsheds
Figure 4. A 'banen landscape' viewshed analysis for a river valley,
plan above and profile below
'barren landscape' approaches are also particularly problematic when there is an obvious
chance that in the past the vegetation clearly played an important role, Figure S. Hence,
when the outcome of a viewshed analysis is interpreted, if environmental data has been left
out, any interpretation must reflect this detrimental omission.
As mentioned above there have been attempts that emphasised the introduction of
more subjective human criteria in relation to wider landscape definitions and settings,
However, any success for these approaches hinges on the ability to maintain the theoretical
impetus throughout the study. When transforming variables based on humanistic er any
other criteria, they can change their representational characteristics as well as their
overall meaning according to the specific GIS algorithm that is applied: This suggests that
the final outcome of a viewshed analysis, or any other computer-based spatial analysis,
needs to account for these changes and the specific processes that acted on and produced
the intermediate stages during the application development, It would also ensure that any
final interpretations are based on a theoretical approach that has been carried through as
part of the whole application development process rather than solely applying some
notional ideas at the beginning (selection) or end (analysis) of a study.
4. A human landscape approach
What does it actually mean to engage in or conduct a huinan landscape investigation?
Essentially, the difficulty lies in the fact that landscapes are' the product of people's
understanding and actions, or as Bender Ill: 1] argues: .
,....
36
37
Beyond the Map
O=Site
Areas without black border = Nolvisible
~ ~ j
Figure 5.Avegetation-inclusiveviewshedanalysis fora river valley,
planabove andprofile below
"Landscapes are created by people - through their experience and engagement with the
worldaround them.They maybeclose-grained, worked-upon,lived-in places, or theymay
bedistantandhalf-fantasised."
In other words,oursurroundings arealreadybestowedwith inherent human definition and
what makes humanistic approaches inarchaeologydifferent is theemphasis on long-gone
peoples to which our current understanding is applied. Maybe it is more appropriate to
refer to such investigations as 'human landscape studies with a predominantly modem
interpretativeemphasisasexpressedbyarchaeologists'.
However, in the absence of any direct means of verification for the conclusions
drawn from the archaeological record, it is clear that humanistic attempts to foster
understanding ofthe past haveconsiderable validityoverthose practices that merelyserve
purposes of acquisition and are devoid of explanations for past ways of life. We need to
look at whether wecan achieve insights on long-gonepeoples beyond merely cataloguing
and classifying their remains. Since the data maynotalwaysspeak for itself, a newset of
tools,suchashumanisticapproachesinlandscapeinvestigations,areclearlyrequired.
4.1 Concepts ofhuman landscape
Archaeologists startedto introduce space intotheir thoughts on pastpeoples and societies
with theonset ofthe20
th
century,although throughtheworks of Bender [12],Tilley [13],
[14]andThomas [15], [16]asignificantjuncture hasbeenreached recently.This ispartof
A.Tschan et al, / Perception and viewsheds
a gradual growth of philosophy-inspired critiques concerning the teachings and methods
offered by geography and archaeology [17], [18]. This advancement advocates the
investigation of space by means of phenomenology in order to achieve insights and
descriptions of a surrounding based on individual human experience. It is, therefore, the
contextual character ofspacethatrepresentsthemainelement forpotentialandmeaningful
'understanding' ofthepast. .
At the same time space, in the absence of a social context, is devoid of meaning
since it can only be established through individuals and/or groups. Inother words, space
can not exist without a human presence and active interaction between a'setting and
between people. Space becomes a meaningful construct when it acts as a resource for
human activities and as the result or consequence thereof. This inseparable relationship
between our surroundings and the other people within it leads to the perpetual cultural
reproduction ofsocialspace(i.e.,contextualising)througheverydayhumanendeavours and
commonrelations.
Cultural reproduction as a process, aims to make an exact copyofsocial space on
each and every occasion. However, changes can occur that affect the way space and the
contents within it are understood or experienced by an individual: As mentioned above,
space has no meaning in itself without the relations created between place and people.
Hence, any perception of space is also entirely subjective because individuals and social
groupswill continual1yconstruct and alter its meaning, resulting innosingle meaning for
space because of the vast differences in experiential understanding attained by human
beings.Once recognised inthis way,space represents aholistic construct thatincorporates
allformsof human cognitivecapacity, physiological condition of thebody, naturalsetting
andopportunity fordirectional movement.
Another important issue relates tomodernist theories and theirtendencyto employ
a prism of history where time is considered to be ful1 of richness, life and dialectics
. [19:70].Space, on theother hand, isoften viewed asstatic, immovable, non-dialectic and
essential1y dead, in stark contrast to post-modem philosophies. As'part of this new
philosophical discourse, there maynolonger be this artificial separation betweentime and
space, considering the growing interest in the latter. These holistic approaches represent
importanttoolswithregardtotheassessmentandinterpretation ofcultural 'reality'.
Space is also intrinsical1y.linkedto time because it is always'created, reproduced,
and transformed in relation to the space that has been created in the past. Freud [20:19]
suggeststhatthe representationof historicalsequences inspatial termscanonly-bedone by
juxtaposition in space since the same space cannot have two different contents. Hence,
space is closely associated with the conception of indiv'idualand social life experiences
withinrespective temporalsettingsforindividuals andgroupsofpeople.
4.2 The case study setting for Wrzesnica, Poland
The following information describing an archaeological case study applies some ofthe
above theoretical concepts by defining and giving some qualitative.meaning to a given
space. At the same time, there is an understanding that the' specific contents and
characteristics aresolelytheinterpretationofarchaeologists with regardtothepeoples that
livedthere in the past. Itis,therefore, worthwhile tohighlight someof thegeneralaspects
of Polish archaeology in conjunction with the specific data from the Wrzesnica region,
colIectedundertherulesof anational programmecalled thePolish ArchaeologicalRecord
(AZP). These detailed standards and definitions, employed by archaeologists across
Poland,werethe productof discussions throughout the mid-1970's'whichresultedin their'
38
r
3\l
Beyond the Map A.Tschan et al. / Perception and viewsheds
implementation in 1978. For the purpose of the AZP, Poland is divided into rectangles of
approximately 37.5 km
2
resulting in each area conveniently fitting onto an A4 sheet of
paper when mapped at a 1:25,000 scale. Any identified or retrieved information on
sites/artefacts is then recorded on 1:25,000 and I: 10,000 maps and in a catalogue system
known as the Archaeological File Card or KESA.
A total of 101 archaeological sites, Figure 6, from several different periods were
discovered in the Wrzesnica research area and most of this data was obtained through
fieldwalking carried out in 1982 and 1983. Whether surface finds were present as a result
of erosion or ploughing was usually not accounted for nor investigated further as part of
this survey work. A key definition that was applied to the Wrzesnica data defined sites as
being spaces with clusters of archaeological material and a context that is isolated from
others by spaces without any archaeological remains [21:19]. This means that the overall
surface distribution of archaeological remains (i.e., pieces of pottery, flints, slag etc.)
identified the perimeter of a site as drawn on maps.
Based on this, the chronology and function of sites/artefacts was assessed while the
interpretation of major earthworks including strongholds, hillforts or burial mounds was an
obviously more clear-cut endeavour. Sophisticated methods like geophysics or aerial
photography have unfortunately not been employed so far and these resources could
therefore not be drawn upon here. Additional research materials for this area which
contributed to the overall data set include archives, museums and the wider literature.
The final results, by and large, also reflect the modem land-use practices at
Wrzesnica where agricultural fields dominate the western and southern parts of the
research area allowing for fewer archaeological sites to be discovered. A main problem is
the damage due to deep heavy ploughing introduced in this area of Poland and linked with
.
.
.-
5'
\9l
I
.
e: \b .
.'.
.
.
8"
T.

Figure 6. Archaeological sites in theWrzesnica region
so-called Prussian Capitalism [22]. It is, therefore, more than likely that the surface
distribution of archaeological remains does not accurately describe the true extent of
settlement. On the other hand, there is a rich forest cover that prevails throughout the
region's east-lying area containing several clusters of early medieval burial mounds. The
years that followed the initial 1982/1983 surveys also allowed those areas of forest culled
for recultivation to be successfully investigated, contributing an additional 32'
archaeological sites spanning many periods.
Overall, twelve sites have been sufficiently dated to suggest some Mesolithic
association while attempts for a more precise chronological identification have been
curtailed due to the character of the collected artefacts (i.e., microliths, debitage). At the
same time, the western and southern agricultural areas have produced no materials so far
for this period. It is difficult to accept that during the Mesolithic there would have been
some kind of barrier (e.g., geographic, social, symbolic, etc.) that could explain this
absence of sites. Furthermore, the slope gradient is rather shallow suggesting that erosion
played a fairly limited role with regard to the archaeological record at Wrzesnica. The
likelihood, therefore, prevails that deep ploughing destroyed the remains and the sites
found in the eastern part only exist because of a prolonged forest cover. At present, none of
these Mesolithic sites have been excavated mainly due to the heavy forestation and dense
undergrowth.
4.2.1 Reconstructing past vegetation pasterns
The Wrzesnica region has always been subject to some form of vegetation distribution and
this supported the idea that floral representation should be included as part of a more'
'realistic' viewshed analysis. At the same time, the obvious impact of vegetation on the
human ability to perceive landscapes based on sight as well as all other means of.
perception was another major factor in the overall decision to integrate this type of
environmental data. However, the goal was not to recreate the actual provenience of
individual trees or shrubs, an obviously impossibletask, but to show the predominant plant
life and its effect on the interpretation of the landscape when using a GIS-based line-of-
sight tool such as viewshed.
In the absence of actual pollen diagrams for the early Holocene of Wrzesnica, the
pollen data of the Baltic Coastal Zone of Poland [23], [24], [25], [26] is acceptable as a
reference for reconstructing the past vegetation patterns for the area period c. 9,500
radiocarbon BP was chosen for the following reasons:
Mesolithic communities/groups were present in the Wrzesnica .area;
The species composition of woodland was relatively simple and
enables a clear reconstruction of the vegetation;
In addition to the pollen diagrams the distribution of surface Quaternary deposits. the
modem soil types and the geomorphological features, were all taken into consideration for
this vegetation reconstruction. However, recognising the substantial environmental changes
during the Holocene, it is clear that the contemporary evidence, especially on the
morphology of the area and on the soils, does not accurately illustrate the conditions fOT the
Quaternary epoch. Hence, with respect to all the above physical landscape criteria.. the
following reservations should be made,in identifying past plant communities:'
"
40
r
41
Beyondthe Map A.Tschan et0/.IPerception and viewsheds
Theriverbeds oftheWieprzaandothersmallriverswereprobably
largerandfollowed slightlydifferentpaths thanatpresent[27], [28];
Theareawascovered byalargenumberofkettle-holes thatwere
eitherstillfilled withwaterorcovered intheinitialstagesof mire
vegetation;
Theuplandandvalleyedgeshadamoredistinct reliefandsomeof
thestillactiveslopesenabledthespreadofpioneervegetation;
Thesoils werenotfullydevelopedandtheywouldhavebeen
relativelyrichincalciumcarbonate.
Any reconstruction of vegetation histories isbasedon pollen diagrams inorder to identify
the presence or absence of particular pollen types and to suggest proportions of different
tree and herb taxa in a research area. Pollen diagrams, however, do not present a
straightforward picture that can be read directly, but rather it requires a more specialist
analysis based on specific palynological and ecological knowledge. The interpretation of
pollen types in terms of plant taxa as suggested by Berglund and Ralska-Jasiewiczowa
[29], together with the identification of the processes of pollen production, dispersal and
preservation affecting theirpresenceinsediments[30],constitutes oneof themostdifficult
palynological problems. However, once retrieved from the fossil record this data on the
ecological requirements for particular species, with an emphasis on competition between
plant communities and their succession, represents the essential basis for understanding
anychanges invegetation patterns.
4.2.2 Theplantcoveratc. 9,500radiocarbonyears BP(c.8,500cal. Be)
In this period rather open pine-birch forests with aspen (Populus tremula) expanded into
the area, Figure 7. On the uplands certainly,Pinus sylvestris was dominant, whileon the
active slopes and valley edges Betula pendula and Populus tremula formed thickets and
patches of open woodland. With regard to the flood terraces, forest communities with
Betula pubescens, Populus tremula and willows (Salix spp.) were probably common. At
the bottomofriver valleyspatches ofozier hopeswithSalix spp. andrushes of tallsedges
andreed(Phragmitesaustralis) startedtodevelop.
These pine-dominated woodlandsweremostlyalooseand relatively open structure
aiding in the development of light-demanding plants including the rejuvenation of pine
and, in places, birch and aspen. The tree stands were characterised by a presence of
successive generations withamaximumheightofabout30-40mand,ingeneral, thisforest
covermade asubstantial case for good visibilityduring boththe summer and winter.This
also means that this setting was very favourable for hunting activities for local
hunter/gatherer groupsduringtheMesolithic.
The plant communities dominated by birches and aspen probably provided a
differentsetofcircumstances withregardtolocalwildlifepopulations, yetdecent visibility
would almost certainly have been ensured due to small and delicate leaves. In other
words,despitethemaximum heightofthesetreesbeingclose to20m,therewasgoodlight
penetration through the forest canopy to the understorey allowing for lower levels of
vegetation todevelopandflourish.
Different circumstances apply to the flood terraces where rich forests of tree
willows,birchand aspenformeddensecurtainsallalongtheriverbed. Attaining heightsof
upto30mthisvegetationwassubjecttoamuchmorepronounced dependence onseasonal
factorstoenabletheobservationofgame.Therelativestateoftreeleafcoverage, therefore,
/
.';t.... r .',............*-1
.. . 'r--.
Jj
c1i <, I

'. .

e->
..
.- .r
Lightest areas =Sedges, Reeds (34m)
Lightareas =Pine (3040m)
Dark areas =Birch, Aspen, Willow (20-30-m)
Darkest areas =Water
Figure 7.The distributionofdominant plants and their average heights for
theWrzesnicaregion c.9,500radiocarbon years BP, 8,500 cal. Be.
must have had a crucial and direct bearing on the employed hunting strategy and in
accordance withrespective seasons theseareas mayhave been utilised todifferent extents
duetovisibility.
The plant communities at the bottom of river valleys and, particularly, any
overgrown paleo-meanders or oxbow lakes, formed dense thickets attaining heights of
around3-4m.These areas, in manyparts, would havebeendifficult tocross andcertainly
restricted the possibility to'recogniseand pursue wildlife. Although, tracesof regular fires
inrushesduring thetime underdiscussion andevident in thepalaeo-ecological sites in the
region,indicate thatMesolithic groupsburnedthis kind of vegetationpossibly tofacilitate
huntingconditions {31],[32].Atthesame time, these areasexhibited thegreatestseasonal
fluctuation for the purpose of visibility as the vegetation wasmuch more open 'during
wintertimewhenreedsdriedoutandlosttheirdensity.
4.2.3 Viewshed limitations
Having set thescene by explaining thecontextual background and the nature/meaningof
the archaeological and ecological evidence, the spatial investigation follows using a
viewshed analysis. Because this case study involves a line-of-sight tool there is a
requirement to integrate some environmental"data with the bare minimum of the'
topography for the purpose of showing any sight-based.effects. The major plant
.. -
r
43
42
A.Tschan et al. / Perception and viewsheds
Beyond the Map
distributions areincluded, based on theabove vegetation reconstruction for theWrzesnica
region,thusgoing beyondthegeomorphological circumstances thatdefine theshapeofthe
landscape.
Itis not necessary to highlight orexplain the detailed GIS application itself since,
for the most part, well-established and proven techniques were applied. This includes
digitising archaeological and ecological data using AutoCAO and employing the
VIEWSHEO routine in Idrisi for Windows with a maximum search distance across the
diagonal length of the research area. What may be slightly different, however, is the
subsequent map algebra manipulations to introduced the plant layer. The Idrisi manual
statesthefollowing:
"To take account of obstructions such as trees and buildings, create images that indicate
heights and add these to topographic surface using OVERLAY module. An image mapof
forests can be reclassified with each species stand assigned the average height of that
species.This imageisaddedtothetopographicsurface."
However, this does not mention the fact that a further adjustment is required involving
subtracting the overlaid vegetation value from the specific viewpoint in order to set a
correctviewer heightwithreferencetothesurface.FortheWrzesnica case studythisvalue
was1.60meters aboveground level,emulatingapersonstanding erect inthemiddleofthe
selectedMesolithic campsiteandsurroundedbyaforestconsisting of20to30mtallbirch
trees.Allother vegetation information wassimplyaddedtothepreviously createdOEMto
describethetopographical andecologicalcircumstances.
The actual resultsobtained arebasedon a 'barren landscape', Figure 8, inaddition
to thevegetation-inclusive viewshed, Figure9, forcomparative purposes and theoutcome
ofeach is fairly unsurprising to saytheleast.Whereas the former identifies a hugevisible
area restricted merely by the physical shape of the landscape, the latter shows no visible
areasbeyondtheimmediate vicinity.
This impenetrable wallofbirchforestismuchlesstheconsequence oftheenhanced
ecological context itself but rather a deliberate methodological limitation. The way the
information onflora wasdesigned foruseintheGISincluded noconceptual refinementof
vegetational distributions apart fromtheidentification of specific locales forthedominant
species.
The reason forthis was tohighlighttwo problems atonce. Firstly, itmust beclear
thattoomit vegetation willproduce impressiveresults but, atthesame time, alwaysleaves
some doubt as to the reliability and persuasiveness of the results obtained. Secondly, the
fact that we can not see anything at all demonstrates an indiscriminate and deterministic
reduction of complex environmental data forthe purpose of a nicelyquantifiable data set.
Intricate vegetation characteristics were reduced to huge solid blocks with a single
representational meaning and placed on the surface without taking into account any
qualifyingpotential accordingtoactualhumanperception andunderstanding.
S. Human landscapes and perception
The suggestionthat fortheinvestigatedMesolithiccampof theWrzesnica casestudythere
isnovisibility beyonditsimmediateperimeterduetothepresenceofbirchtreesis
~
~
~
. : ......
, "-":'/
~
'", J,
... "'A
.. "", .. ,""
.. ~
~
-
",...

= Site .=Site
Dark areas = Visible fromsite Dark areas = Visible fromsite
Light areas =Not visible Light areas = Not visible'
Figure 8(left). Viewshed analysis forWrzesnicawithout vegetation
Figure 9(right). Viewshed analysis forWrzesnicawith vegetation included
certainly not a supportable interpretation of past conditions. A human understanding of
walking through many forests in Central Europe and elsewhere should soon prove that
one's sight is not as comprehensively restricted as proposed by the above vegetation-
inclusiveviewshed analysis. However, as pointed out, this shortcoming is-nottheproduct
of any technical limitations regarding GIS line-of-sight tools but rather the result of the
purposely-broad conceptual character of the incorporated plant data. Evidently, a more
detailedappreciationisthereforerequiredwithregardtothesignificantrolethatperception
playsindefiningexperiential landscapesandtheobjectsthatarecontained'withinthem. At
the same time, additional conceptual definitions need tobe presented concerning the idea
of space and the many forms of human relations that will constitute it.This represents a
backgroundconducive tofurther understanding theroleof perception andthewaysensory
informationcan beintegrated in an attempt toarrive at morehumanistic interpretations of
ourimmediatesurroundings, theworldasweknowit.
5.1 The role a/space
Spacetakes its meaning from places and space does not exist without place. Place is the
keyfocusinourdiscussion on space sincespace identifies .thelocationalcontextfor place.
In other words, place being located inspace is known through everydayand accumulated
experiences, symbolism and meaning. Place, therefore,plays a pivotal role with regard to.
r
45
44
Beyond the Map A.Tschan et al. / Perception and viewsheds
human activity, meaning and emotional relations by providing the context for people's
knowledge, movements and memories.
However, places are experienced and conceptualised at different spatial levels,
ranging from the individual (private) to the regional/global (communal/social) space.
Hence, places are both internal and external when they relate to human consciousness of
place and the physical elements of space respectively. Individuals and social groups
therefore construct their identities in relation to place and the experience of an individual or
group 'geography' will start with place and continually grow to the eventual creation of
landscape. This means that places are crucial in the formation of individual and social
group identities and this relationship between humans and place, therefore, tells 'stories'
and constructs a history.
Names, on the other hand, will relate both to the environment (e.g., mountains,
rivers, trees, etc.) and places created or transformed by human activities (e.g., settlements,
burial mounds, field systems, etc.) or historical events in the form of stories that are
anchored spatially [33:41]. Named places will, therefore, attain meaning through many
forms of social relations. The cultural significance of a name is important because it
converts purely physical dimensions into objects with meaning as part of a historic and
social context. The 'outside world' composed of a myriad of random physical objects that
are devoid of meaning constitutes the raw material from which humans define space.
Physical objects must be perceived and named to be included in an ordered and human
space. Perception of the 'world' will constitute it. The emphasis expressed by an individual
or group is therefore highly relevant as argued by Bender [11:1]:
"In the contemporary western world we 'perceive' landscapes, we are the point from which
the 'seeing' occurs. It is thus an ego-centred landscape, a perspectival landscape, landscape
of views and vistas. In other times and other places the visual may not be the most
significant aspect, and the conception of the land may not be ego-centred."
Very complex structures of objects and relations can, therefore, exist and for the purpose of
a landscape definition Tilley [14:25] refers to:
" ... the physical and visual form of the earth as an environment and as a setting in which
locales occur and in a dialectical relation to which meanings are created, reproduced and
transformed."
5.2 Perception through direct and indirect engagement
Space being socially created depends on perception and through it results in the definition
and understanding of landscape while at the same time perception defines the world and its
contents. It is therefore important to understand what is actually meant by the act of
perceiving since it nowadays may be misinterpreted as solely visual cognition. By
definition perception is actually an all-inclusive act and described as:
'The process by which an organism detects and interprets information from the external
world by means of the sensory receptors." (Collins English Dictionary).
The specific means of sensory capabilities for humans, providing there is no physiological
restriction, will, therefore, allow experiences of the world based on the ability to touch,
taste, see, hear and smell. Our individual capacity to make sense of our surroundings
consequently hinges greatly on these perceptional tools.
There is, however, the possibility to distinguish the five senses utilising a useful
separation into two main modes of perception. The first set involving touch and taste, for
the most part, depends on actual physical contact in the form of direct engagement between
us and our surroundings. The ..other group of perceptional tools,' however, offers
considerable promise with regard to humanistic landscape analysis and other forms of
spatial interpretation. These are.the sensory experiences leading to an understanding of the
world through indirect engagement and include the ability to see, hear and smell. In
contrast to the direct modes of perception these three can potentially cover large areal
expanses without the need for any close physical contact.
With regard to spatial analysis using GIS-based tools the above separation could
prove highly rewarding when looking at perception based on indirect engagement. Whereas
touch and taste seem limited because their sphere of influence is so tightly prescribed to the
human experience within the immediate vicinity, seeing; hearing and smelling could be
subject to spatial analysis using already existing technology/methods such as the line-of-
sight algorithms. The key is to understand that all three are subject to distance as a process
and based on this fact there is no reason why viewshedanalysis parameters for visibility
could not be adopted to also identify sound, odour or cumulative combinations, Figure 10.
5.3 Future trends
The earlier vegetation-inclusive viewshed result from Wrzesnica was unsatisfactory due to
the definition of the different plant distributions as solid objects. In reality there is the
= See, Hear, Smell
= See
= Hear
= Smell
= No direct perception
o =Site
Lighest area
Light area
Dark area
Darker area
Darkest area
Figure 10. A perception-shed model
'1
46
47
r-
Beyond the Map
potential to see considerable distances within many different categories of forests
depending on the prevailing tree and undergrowth species. At the same time, this ability
might be enhanced due to seasonal factors. For the Wrzesnica Mesolithic site this could
mean extensive room for representational and interpretational improvements based on the
potential visibility in line with the dominantplant life and seasonal characteristics as well
as the human eye's ability to interpret landscape features and their meaning differentially
over alargedistance.
The application of indirect engagement mentioned above might also prove fruitful
when considering issues of mobility within the landscape. Although concentrating mainly
onsight, Thomas[16:30] suggests:
"Vision isimportant ... butin termsof the interplay between what isopen toviewand what
isconcealed, which is instrumental in the reproductionof privileged positions with respect
to social and cultural knowledge. Thus what is seen has to be understood in the context of
movementfrom place toplace, and thatof non-visual experience:'
These non-visualexperiencesareparticularlyinteresting when looking at movementacross
heavily vegetated landscapes and where there would be an added emphasison other forms
of perception. A spatial analysis that combines sight, sound and odours might be highly
revealing where settlements would have been hidden from general view but could have
been identified by virtue of the noises and smells (e.g., smoke) generated by a local
population or domesticated animals. Beyond seasonal criteria that affect perception there
are also climatic and meteorological variables such as prevailing wind directions,
precipitation or other forms of dominant recurrences of weather-related effects. Clearly,
this list could be extended forever and is limited only to the human imagination for
understandingthe world.
6. Conclusion
The aim of this work was to investigate the methodological potential for an enhanced
human understanding of archaeologicaVcultural landscapes. For this purpose an increased
representational form of 'realism' was sought that would allow a greater capacity for
interpretation than is currently available through the tools and methods employed in GIS-
based research. We strongly advocate the development of new discipline-specific
approaches for landscape studies that integrate concepts of space based on human
perception. The example of a viewshed analysis using reconstructed vegetation data
highlights the significance of complex environmental variables in relation to the
understandingthatcanbeachieved and, inparticular, thedramatic effect itmayhaveonthe
outcomeofspatial analyses that ignore such valuable sourcesofinformation.
Finally, webelieve that viewsheds andperceptionare not automaticallyor mutually
inclusive because of the often narrow or even reductionist agendas currently applied to
line-of-sightanalysis.This isan unsatisfactorysituationsince itisclear thatsuch aresearch
focus lacks the perceptional and humanistic approachesnecessary toelucidateanenhanced
understandingof the archaeological record. The currentsteps aimed at turning quantitative
data into qualitative information should be augmented to further include a series of
application developmentstages that lead fromqualitativedata to meaning, Whetheror not
line-of-sight tools in archaeology are actually still feasible when they continue to ignore
omnipotent environmental information, such as vegetation, is only one ofa great many
A.Tschan et al. / Perception and viewsheds
conceptual issues that needs'to be addressed further with regard to archaeological
knowledge andpractice.
Acknowledgements
Gary Lock deserves our express gratitude for all his support before, during and after the
conference. Special thanks also go to Professor Wadaw Florek for the geomorphologic
dating and interpretation and Rafal Zaplata for his data entry computer work. Additionai
credit goes to Patrick Daly, Tom Evans, Sheila Raven and all the other Oak regulars for
various discussions. Finally, once again many thanks toColleen for her, as always, highly
valuable input and seemingly shatterprooflevels oftolerance-bjednabonbonierkal
References
[I] Tschan, A.P. An Introduction to Object-Onented GIS in Archaeology. In: J. A. Barcelo et al. (eds.),
New Techniques for Old Times CM 98 - Computer Applications and Quantitative Methods in
Archaeology. BARInternationalSeries757, Oxford, t999,pp. 3 ~ : 3 1 6
[2] Raczkowski, W. Power of image: some ideas on post-processual aerial archaeology, AARGnews t9,
September1999, pp. 10-14.
[3] Kvamme, K. Ranter's Comer: bringing the camps together: GIS AND ED, Archaeological Computing
Newsletter, Vol. 47 (1997) pp. 1-5.
[4] Wheatley, D. Ranter's Corner; Keeping the camp fires burning: the case for pluralism, Archaeological
Computing Newsletter, Vol. 50(1998)pp. 2-7. .
[5] Allen, K.M.S., S. Green and E. Zubro;" (eds.), Interpreting Space: GIS and Archaeolog. London:
Taylor& Francis, 1990.
[61 Lock, G. and Z. Stancic,Archaeology and Geographicallnfromation Systems - A European Perspective.
London: Taylor& Francis, 1995. .
17] Lock, G. Archaeological computing, archaeological theory and moves towardscontextualisrn, In:
J.Huggell and N. Ryan (eds.), Computer Applications and QUlJntitative Methods in Archaeology i994.
BAR InternationalSeries600,Oxford, 1995,pp. 13-18. .
[81 Tschan,A.P. and P.T. Daly, Is there really such a thing as"ComputerArchaeology", In:G. Lock andK.
Brown (eds), On the Theory and Practice of Archaeological Computing.' University of Oxford
Committeefor ArchaeologyMonograph,Oxford,2000.
[91 L1obera, M. Exploring the topographyofmind: GIS, social space and archaeology, Antiquity 70, 1996,
pp. 612-622.
(10) Peel, R. Modem Geographical Thought. BlackwellPublishers,Oxford, 1998.
[IIIBender,B. Introduction: Landscape- Meaning and Action. In: B. Bender(ed.), Landscape: politics and
perspectives. Berg, Oxford, 1993, pp. 1-17.
[12] Bender,B. (ed.), Landscape: Politics Dn4 Perspectives. Berg, Oxford, 1993'.
(13) Tilley, C. Art, Architecture, LandscapeofNeolithic Sweden. In: B. Bender (ed.), Landscape: politics
and perspectives. Berg, Oxfo!lL.1993,pp. 49-84.
(14) Tilley,C. APhenomenology ofLandscape: Places. Paths and Monuments. Berg, Oxford, 1994.
[l5JThomas, J. The hermeneutics ofmegalithic space. In: C. Tilley(ed) interpretative Archaeology. Berg,
Oxford, 1993, pp.73-97. .
[161Thomas. J. The politics of vision and the archaeologies of landscape. 10:B. Bender(ed.). Landscape:
politics and perspectives. Berg, Oxford,-I993,pp. 1948. .
(17)Cosgrove,D.Social Formation andSyonbolic Landscape. Croom Helm, London, 1984. .
r
48
49
BeyondtheMap
[18] Wagstaff, J.(ed.), Landscape and Culture. Geographicaland ArchaeologicalPerspectives. Blackwell,
Oxford, 1987.
[19] Foucault,M.Power/Knowledge. Selected Interviewsand Other Writings 1972-1977. ed. by C. Gordon.
TheHarvesterPress, Brighton, 1980.
[20) Freud,S. Civilization and ItsDiscontents.W.W. Norton& Company, New York, 1989.
[21] Mazurowski, R. Metodyka archeologicznych badan powierzchniowych [Methods offieldwalking
survey).PanstwoweWydawnictwoNaukowe, Warszawa,Poznan, 1980.
[22] Burszta,J. Kulturaludowa Wielkopolski,1.1, Poznan, 1960.
[23) Latalowa, M. TypeRegion pot:BalticCoastalZone. ActaPalaeobotanica 29 (2), 1989, pp. 103-
108.
[24] Latalowa, M.Type Region P-u: Baltic shore, W part Wolin Island. Acta Palaeobotanica 29 (2), 1989,
pp. ll5-120.
[25] Latalowa, M. andK. Tobolski, Type Region P-u: Baltic shore. Acta Palaeobotanica 29 (2), 1989, pp.
109-ll4.
[26] Ralska-Jasiewiczowa, M. and M. Latalowa, Poland. In: B.E. Berglund, H.J.B. Birks, M. Ralska-
Jasiewiczowa & H. Wright (eds.), Palaeoecological Events during the Last 15,(}()() years. Regional
Syntheses of PalaeoecologicalStudies inLakesandMiresinEurope.J.Wiley& Sons Ltd, Chichester,
1996, pp. 403-472.
[27] Florek, W.Postglacjolny rozw6j dolin rzeksrodkowej czesci pO/nocnegosklonu Pomorza [Postglacial
development ofriver valleys in the middle part ofthe northern slope of Pomerania). Wyisza Szkola
PedagogicznawSlupsku,Slupsk, 199I.
[28) Florek, W. Gl6wneetapyrozwojukoryt idolin rzecznychsrodkowegoPrzyrnorzawp6tnyrn vistulianiei
holocenie [The main development phases ofthe river valleys on the South Baltic Coast in the Late
Vistulian and Holocene]. In: W. Florek (ed.), Geologia i Geomorfologia Srodkowego Pobrzeza i
PoludniowegoBaltyku.WyiszaSzkolaPedagogicznawSlupsku, Slupsk, 1993, pp. 279-290.
[29] Berglund, B.E. and M. Ralska-Jasiewiczowa, Pollen analysis and pollen diagrams. In: B.E. Berglund
(ed), Handbook of Holocene Palaeoecology and Palaeohydrology, John Wiley& Sons Ud., Chichester-
Singapore, 1986, pp. 455-484.
[30] Moore,P,D. et al., Pollenanalysis. BlackwellPublications,Oxford, 1991.
[31] Tobolski, K. Holocene vegetational development basedon the Kluki reference site in the Gardno-Leba
Plain. ActaPalaeobotanica 27 (1),1987,pp. 179-222.
[32] Latalowa,M. Gospodarka mezolityczna i poez atk!rolnictwa naobszarzepolskiegoPobrzeia Baltyku w
swietle danychpalinologicznych[The Mesolithiceconomyand the beginningofagriculturein the Polish
Baltic coastal zone in the light of palynological data]. Polish Botanical Studies, Guidebook Series ll,
1994, pp. 135-153.
[33] Bender,B. SUbvertingthe WesternGaze: mapping alternative worlds. In: P.J.Uckoand R.Layton (eds.),
TheArchaeologyandAnthropologyofLandscape. Routledge,London, 1999, pp. 31-45.
Beyondthe Map
G. Lock (Ed.)
lOS Press, 2000
Understanding wetland archaeological
landscapes: GIS, environmental analysis
and landscape reconstruction; .
pathways and narratives
HenryChapman
Centrefor WetlandArchaeology
University ofHull
Abstract: Archaeological landscape theory provides a framework for studying and
understanding sites within their surroundings. However, multiple phasing. visual
obstructions and environmental change can make the experience and interpretation
of a landscape more limited. By reconstructing elements of both the destroyed
archaeology and the palaeoenvironment it is possible to provide a better foundation
from which to beginto interpretthe landscape. This paperpresentsan applicationof
GIS to reconstruct elements of the archaeological and palaeoenvironmental
landscapeas a basis of iii interpretation.
1. Introduction
The development of landscape'phenomenology has provided a framework for studying
archaeological sites within their landscapes. Dualistic concepts such as space/place [I]
have been expanded to provide frameworks of understanding of locations, and for
movement between them [2].Such analyses within landscape archaeology ~ commonly
approached through narration [3];highlighting elements of visibility' and inter-visibility
fromdifferent points along a route [4]. Observations are then classified into processes of
observeranticipation andarrival[2].
Twoproblemsexist withinthisapproach tolandscape. Firstly,themultiplephasing
of landscapes, visual obstructions and environmental change can make the experience of
landscape more limited. The application or"landscape reconstruction through GIS has
provided a way of recreating these past archaeological landscapes and has also enabled
additionalanalysesofabstractedperceptions of these constructs to beevaluated.Secondly,
thedescription of theobservations within narrative frameworks provides limitedscope for
re-interpretation,repeatabilityorqualification(5]. .
Within wetland archaeological landscapes the picture of environmental change is
often highly variable with both hydrological and vegetational 'changes through time
influencing the overall perception of the landscape [6]. With this increased number of
contributing factors, manyof which are unknowable, perceiving past landscapes becomes
increasingly difficult. Therefore, if we can successfully model the' landscape using the
enhanceddataset availablefromtheiricreasedpotential for preservationofferedby wetland
r 50 51
Beyond the Map H. Chapman / Wetland archaeological landscapes
sites, then we can hopefully produce a more realistic model of the contemporary
environmentforuseinour analyses.
2.Thesite
Sutton Common, near Doncaster, England, lies on the western edge of the Humberhead
Levels;anareaofreworked earlypost-glacial lakesediments (Figure 1).Inthis area raised
sea levels associated with the melting icesheetscreated flooding in themain river valleys
that have maintained a high ground water table [7]. The value of the subsequent buried
anoxic environments that reduce biological activity, thereby enabling the preservation of
biogenic remains, has been well documented [6], [8] and [9]. Such remains can include
archaeologicalandpalaeoenvironmentalsourcematerial.
On Sutton Common, a pair of early Iron Age enclosures, defined by earthwork
banks and ditches, was noted on either side of the relict palaeochannel of the Hampole
Beck [10], [11] and [12]. Limited excavations and architectural qualities indicate that the
two enclosures were contemporaneous. Further, it has been revealed that a causeway
constructed ofsandand linedby atimberpalisaderan acrossthechannel connecting them.
It appears that by this time the channel had silted up, presenting an area of increased
surfacewetness[13].
Until the late 1970s the site remained as a common, undrained and unfarmed.
However,in 1979muchof thearea wasflattened bybulldozer tofacilitate ploughing.The
completeareaof thelargerofthetwoenclosureswasdestroyed andthesouthernportionof
~ ~ r ~
.'jlf/ '
...
~
Figure 1.LocationmapofSuttonCommonon the
western edge oftheHumberheadlevels
the smaller enclosure was damaged. This activity coincided with the implementation of
landdrainage through underfielddrains,andthiswascloselyfollowed bywaterextraction.
Thisactivitydramatically altered theappearanceof both thearchaeological and thenatural
wetlandlandscape, renderingthesitemuchlesscomprehensible ontheground.
The site has been investigated through a number of excavations that have either
attempted to understand the archaeology, e.g. [12], or to assess its state of organic
preservation, e.g. [14]. On the basis of this work it has been suggested that the smaller
enclosureacted asagatehouseorannexeforthelargerenclosure [ibid.]. This interpretation
maybe understood on the basis of two factors. Firstly, if it is assumed that"a defensive
strUcturewill have aninternalbank andexternalditches,then thesmallerenclosure would
need to be approached from the west. The western bank displays an arrangement ofan
external ditch with an internal bank. A break in this arrangement is visible in the
southwestern corner. The eastern side of the enclosure shows an internal ditch and an
external bank indicating that it was approached from the interior. Therefore, if a real or
symbolicallydefensive roleisassumed for thesite, then the smaller enclosure would need
tobeapproachedfromthewestandtravelledthrough,creating twolinesofdefence.
Secondly, the largerenclosure appears that it should be approached from the west
also.Thewesternbankisrevettedexternallywithlimestone, andthisistheonlysidewhere
this isthe case. It seems therefore that, from both a defensiveand a visual point of view,
thelargerenclosure wouldhavebeenapproachedfromthewest.
Following theseobservations, a, routewassuggested wherebythesmaller enclosure
wasapproached fromthewestwhereonewas metby adefensive ditch-bank arrangement.
The enclosure was entered and traversed offering a second defensive ditch-bank
arrangement as itwas exited. The presumablyboggy areaof therelict Hampole Beck was
crossedoverapostulatedcauseway,asdemonstrated bylaterexcavation [13]. .
A number of questions remain to be answered in order to understand the
applicability of this theory, but which cannot be understood on.the ground due to the
comprehensive destruction of much of the archaeology and the alteration of the natural
environmental context. The first question concerns the arrangement of banks defining the
largerenclosure. A singlebank defines itswestern and northwestern edges, adouble bank
structure defines its eastern and northern sides and a triple bank defines ItS northeastern
corner. The largest earthwork structures define an area that is not 'a part of the route
outlinedabove.The secondquestion concernsthesmaller enclosure whereanextensionof
the banks and ditches defining its southeastern corner have never been explained, despite
being morphologically distinctive and'unusual. The interpretation of this feature is
increasinglydifficultsincebulldozingdestroyedit.
3.Thesurveyanddataprocessing
3.1. Survey
Thesitewas initially surveyedto provideaframework upon whichfurtheranalyses would
bepossible, particularly in relation to the management of the site, but equally useful as a
basis for the reconstruction of the archaeological landscape and selected elements ofthe
palaeoenvironment. The topographic survey of the site, and all subsequent survey, was
undertaken using a Geotronics System 2000 Ll - RTK differential Global Positioning
System(GPS).Thisequipment receivesradiosignalstransmittedfromaconstellation of24
satellitescontrolled bytheUSAirForceatColorado.Thesesignalsprovidedatarelatingto
H : ~ ~ - > :
r 52 53
Beyond the Map
time, takenfromtheatomic clocks aboard thesatellites, and velocity. Each satellite canbe
trackedover timeand so its orbit.spherecan becalculated from predictions of its position
within agivenepoch. When data arecollected fromseveral satellites at thesame time, the
intersections of these orbit spheres can becalculated to provide a position on the Earth's
surface[15].
Inaccuracieswithin thesystemstemfromanumberofsources including multi-path
and selective availability, the latteradded bythe USAF to provide security. Anallowance
can bemadeforthesebycombining tworeceiversthatrecorddata fromthe same groupof
satellites atthesame time, knownasdifferential GPS. Bykeeping abase station stationary
throughout the survey, data from the roving receiver can be calibrated to provide
centimetre accuracy relative to it. Absolute coordinates (relative to GB National Grid _
1936)can beobtained through atransformation of theGPSgrid (which is recorded inthe
spherical WGS-84 coordinate system) and by locating points on the ground of known
positionsuchasbenchmarks.
The site wassurveyed in paced transectssighted using ranging-rods. The distances
between the points surveyed varied across the site from less than a metre, to inexcess of
five metres.Changes insurveyresolution werereactionarytothenatureofthearchaeology
being recorded so that it was increased in theareas of morecomplex earthwork features,
but less densein the ploughed areas. Allother featuresand specific data wererecorded in
thesamemanner.
3.2. Data-processing methods
TheGPS surveydata were recorded inacoded formatthat wascorrected toNationalGrid
valuesand converted toaCSV (comma separated value) file,consisting of x-coordinates,
y-coordinates and z-coordinates, using software developed by Richard Middleton
(University of Hull). These data were processed to generate a digital elevation model
(OEM)withinARC/INFO version 7.2.I GeographicalInformation System(GIS)software
runthroughaUNIXplatform.
The variably spaced point data were converted to form a triangulated irregular
surface(TIN)usingtheCREATETIN command.Theaccuracyofsurfaces interpolated asa
TIN model is dependent upon the function of the triangulation process. ARCIINFO
employs a process of Delaunay Triangulation, which dictates the size and shape of the
trianglesformedinthegeneralised surface[16].Thecriterionofthisprocess isthatacircle
drawnthroughthevertexesofeach trianglewillcontainnoothercoordinate points.
Acell-based surface wascreated from theTIN using the TINLATTICE command
which interpolates a continuous grid using theTIN as a reference. The function places a
grid of cells at a pre-determined density across the area covered by the TIN that are
referenced in terms of x- and y-coordinates. A height attribute for each cell is then
interpolated from the TIN. The flat planes of the TIN were smoothed using a quintic
function. This method appreciates that the surface being represented is without the harsh
breaks in slope represented through any linear interpolation, instead a continuous surface
iscreated that runs through the nodes and formsasmooth interpolation through the areas
between. The resulting surface is not only more realistic aesthetically, but is potentially
also more accurate. The resulting topographical lattice surface was interpolated at a
resolutionof0.5m.Theresulting OEMisshowninFigure2asanexaggerated perspective
modelwithalightsourceappliedfromthenorthwest(azimuth315,altitude45).
H. Chapman / Wetland archaeological landscapes
Figure 2.Basic perspective hillshaded modelof SuttonCommon
4.Archaeologicalreconstruction
The model displayed the preservation of the earthworks relating tophase two of the site
and the damage that had been done to them. Fortunately the position of the banks were
recordedon surveysconducted byBennett and Hill inthe 1930swhichwere subsequently
published with theexcavation reports [12].Theexact positions ofthese featurescould not
be ascertained, but when the GIS surface was scaled to match, the relationship between
visible features and the removed fence marked on the early survey, and grid north meant
thatthetwocould bebrought together.
Themainproblemwithunderstandingthesitewasthedestruction oftheearthworks
defining the larger enclosure. The banks were therefore digitised as polygons from the
earlier survey and overlain onto the model. These were then converted to a grid with the
sameresolution as theOEM. Localmemoryrecords the banks as havingbeen 20' (c. 6 m)
high before they were bulldozed. No' accurate measurement exists and the published
photographs fromthe 1930sdo notshow the banks clearly. Aminimum valueof I m was
chosen on the basis of the minimum height of banks of the larger enclosure for the
reconstruction. Bychoosing such a lowheight no exaggerated results could be generated
fromtheanalysis,but itcould beconsidered thatthe banks wereoriginallycertainly much
higher than this, therefore the relationship between the positions of the.banks andthe
natural topography could be understood better, The areas of the banks on the newly
generatedDEMweregiven avalueof I mwiththesurrounding areagivenavalueof0 m.
This surface was then added to the basic OEM to provide a reconstructed archaeological
surface.
The ditches of the smaller enclosure were reconstructed. This was done for two
reasons. Firstly, ditches were considered important in relation to the function of the
enclosuresandevidencefromexcavationhaddemonstrated thattheyhadbeenconsiderably
deep - generallyat least \.5 m. However,excavation of ditches had been limitedwith the
larger enclosure and so it was n?t considered appropriate to reconstruct .thesefeatures.
~
54
Beyond the Map
Secondly, the reconstruction of the ditches enabled an understanding of the extension
feature on the southern comer of the site that had also been bulldozed. The reconstruction
of the ditch in this area would have the added advantage of' providing a clearer
understanding of the earthworks in this area despite their differential preservation. The
ditches were reconstructed in the same way as the banks. They were converted to a grid,
but provided with a value of -1.5 m for those areas within the ditch polygons while the
surrounding area was given a value of 0 m. This was also added to the surface model of the
site.
5. Environmental reconstruction
The key to interpreting and understanding a lowland wetland site such as this lies in its
environment. This is partly because wetlands provide the potential for environmental
reconstruction through the preservation of organic palaeoenvironmental remains, but also
because the potential for localised and dramatic environmental conditions can have a large
effect on the archaeological understanding of a site.
The ground water table and its fluctuations were being monitored as part of a
separate project investigating the effect of drainage upon the buried organic archaeological
remains [13]. This was being conducted through a network of 50 piezometers arranged
across the site spaced at 50 m intervals in a grid. The data collected from the piezometers
had been interpolated into time-slice surfaces surveyed to 00 using GPS and compared
with the depth of buried archaeological remains across the site. The model of the water
table, constructed within the GIS in the same way as the topographic surface, showed it to
be domed, approximating the topographic surface, with higher areas reflecting the positions
of the sandy islands.
The period of highest water table recorded during the monitoring period was on the
21" January 1999. By subtracting the interpolated groundwater surface generated using this
data from the topographic surface, areas of standing water could be observed by a positive
value within the resulting third surface. These areas were converted to polygon coverages
to facilitate overlaying them on the topographic model. A ground truthing exercise at this
time demonstrated that the pattern appeared to be correct.
The contemporary level of ground water is not known, although it may be assumed
that it was higher annually and more stable than currently within the modem drainage
regime. Exceptional preservation within the basal 20 em of the ditches of the smaller
enclosure, coupled with the continued preservation of pollen within the upper section
indicated that an increase in water table of at least 20 cm was considerable. A largely
arbitrary value of 30 cm was therefore chosen for the reconstruction of the water table. This
value provided a greater amount than the minimum increase but there were other factors
that could not be modelled. Firstly the seasonal fluctuation could not be understood,
although it is understood that the dramatic lowering of summer levels would not have been
echoed before drainage. Secondly, the growth of peat within the channel area and the
subsequent shrinkage of it through desiccation meant that the effect of ground water could
also not be modelled very accurately.
6. Static results
The results from the combined reconstruction of the archaeology and hydrology of the site
show a number of features that demonstrate architectural choice, particularly in relation to
55
H. Chapll1/JZl / Wetland archaeological landscapes
the areas of standing water and the position of the in-filled channel, Figure '3, In terms of
defence you can see a relationship between the distribution of surface water and a reduced
number of banks. On the western side where the standing water was modelled in the
palaeochannel only one bank was constructed, with a similar scenario on the northern side
of the site. On the eastern side of the site standing water was more limited. This is reflected
by an increase in the number of banks to two. Finally, in the northeast comer of the larger
enclosure a triple bank was recorded from within an area that displayed no standing water
at all. It seems therefore that a relationship exists between wetness and the number of
banks constructed. The reasons for this could be defensive, either physically or
perceptually, or symbolic as part of a larger significance with the wetlands further to the
east.
In terms of visibility the enigmatic structure on the southeastern comer of the
smaller enclosure, bulldozed in 1979, may be explained in relation to the digital re-cutting
of the ditches and the raising of the water table. It is possible to see a continuous 'moat'
surrounding the site that links into the standing water in the southern part, of the
palaeochannel. ITthe site was approached from the west this would have the twofold effect
of extending the length of the southern earthworks to be similar to those to the north of the
entrance, and further to demonstrate a perceived moat, even if not actually complete.
/ Edge of monitoredarea
o Areas of modelledsurface water
Figure 3. Combined model of archaeological and palaeoenvironmental reconstruction
7. Pathways and narratives
Attempts to understand and interpret archaeological landscapes within the framework of
phenomenology have usually focused on what is visible from given points within the
landscape as visibility is understood to be the primary sense when dealing with landscape
perception [17]. This has also been used in similar studies that have applied GIS in
attempting to understand the past landscape in terms of post-processual concepts, e.g. [18].
The previous interpretation of a route through the' site has been outline above,
suggesting a journey that approached the site from the west, travelling .through the smaller
enclosure to arrive at the larger enclosure [14]. This theory may be rationalised in three
r
57
56
H. Chapman / WeI/and archD.e%gica//antbcapes
Beyond the Map
main ways. Firstly, approachingfrom the west wouldprovide accessfrom the dryland- the
positions of the main areas of wetland (the Hampole Beck channel and the wetland of
ShirleyWood) would have restricted access from the east. Secondly, the arrangementof
the ditches defining the smallerenclosure indicates that the site was approached from the
west. On its western side there is an external ditch and internal bank. On its eastern edge,
the position is reversed so that there is an internal ditch and an external bank. Assuming
that an external ditch is more defensible, either physically or psychologically, then an
approach from the west would be met by the typicalarrangementof a ditchbeforea bank.
As the enclosurewas traversedasecondditch/bankarrangementwouldbe met. Finally, the
western bank of the larger enclosure was faced with limestone quarried from the
Magnesian limestone ridge at least five kilometres to the west. This would have been
visuallyvery impressivewhen approachedfrom the west, but not visiblefrom the east.
The route was followed using the GIS by calculating the visibility from five
selectedareas along it,each atan observerheightof 1.7m (Figure4).
Position 1 - arriving at the entrance to the smaller enclosure (Figure 4a)
The approach on the ground is obstructed by the position of the modem drain, though the
above analysishas suggestedarouteacross it.Fromthis position the area up to the smaller
enclosureis very visible, as is the majorityof its interior. Whilemuchof the internal ditch
on the easternside of the enclosurecannotbeseen, this wouldhave highlightedthe eastern
bank that could be seen. The palaeochannel was not visible, although the revetted western
bank of the larger enclosurewas clearlyvisible. The southern banks of the larger enclosure
were also visible from this approach, as was the inside edge of the eastern bank. This
would have provided a sequence of three banks that would have formed a series of
intermediatehorizons.Fromapsychologicalperspective, from this pointthe destinationof
the largerenclosurecould be anticipatedbut not reached, the banksdisplayingstrength.
Position 2 . approaching the exit of the smaller enclosure (Figure 4b)
The route across the smaller enclosure is defined by a hollow, visible on the model,
extending between the two entrancesobservedabove, although it has been noted that this
feature may be theresult of later activityand vegetation.From the start of the route through
the smaller enclosure the majority of its interior is visible, including the nature of the
internal ditches on the eastern side. The banks of the larger enclosure were also visible,
particularlyon its western side where they dominate the horizon. As the enclosureexit is
approacheditisnotable that, althoughthe majorityofthepalaeochannelis not visiblefrom
this position,the area of thecausewaycould beseen. Movingfurtherthroughthesmaller
enclosure the position of the causeway and the surrounding palaeochannel became more
visible with no break up to the western edge of the larger enclosure. If the wetness of the
channel is interpreted as visuallyor perceptuallydefensive, then from this pointthe strong
nature of the largerenclosure graduallycomes into view, with the channeland the revetted
western bank. Its interior remains invisible, as does the area behind it. This approach
towards what was the distant horizon is not replacedby anotherdistanthorizon, but rather
loomsas itisapproached,blockingout furtherunderstandingofcontext.
Position 3 - the causeway (Figure 4c)
The route across the causeway was more complex than for stages I and 2 because
of the more variable topographyinthis area with sharp changesinelevationacross the
Areas of surface water
Vrewshed
III Observer101T"" 1.7m)
Figure4(atoe),Movingthrough thelandscape
palaeochannel. A principlepointapproachingthe centreof the causewayis illustrated, but
must beseen inconjunctionwith thefourth stageof the route.
Upon leavingthe smallerenclosurethe view behindbecomesgradually less visible.
Ahead the view remains largelythe samewith the visual dominance of the palaeochannel
and therevettedwesternedgeof the larger enclosure. Still the interiorand area beyondthe
largerenclosurecannotbeseen. .
Position 4 - the entrance to the larger enclosure (Figure 4d)
Continuingacross the causewaythe viewshed changes further. Upon meetingthe entrance
ofthe larger enclosurethe view infront remainsvery restrictedwith limited visibilityofits
interiorand no visibilityof the area to the east. However, there is quitea dramatic change
inthe viewshedover therest of the landscape. From this pointthe view behind isextremely
limited, with minimal visibility of the smaller enclosure or the area further west. The
viewshed is focused inwards, concentrated. upon the lowerarea ofthe palaeochannel. This
;=-. ..-
T
.59
!
58
Beyond the Map
point marks the distinction between anticipation of arrival and arrival - little can be
understood of the area behindor in front of the observer. This scenario may be compared
with the role of a shallow valley along the route of the avenue at Avebury that may have
had asimilarfunction [19].
Position 5 - entering the larger enclosure (Figure 4e)
The final posrtion of this route was calculated from within the centre of the larger
enclosure. What is striking from the resulting viewshed is that, whilst almost the entire
interiorof the enclosure is visible from this position, very little of the area outsideof the
enclosurecan beseen. Fromoutside, in the previousfour positions,only partsof the larger
enclosure could be seen whereas once inside, the whole of the interior can be
comprehended but nothing beyond its boundaries. Following the development of
anticipationduringthe approachatthis pointthe observerhas finally arrived, demonstrated
byan inwardlyfocused viewshed.
Starting at the entrance to the smaller enclosure, and extending up to its exit, the
view of the larger enclosure constantly forms the most distant horizon, and the third of a
sequenceof ramparts. As the exit of the smallerenclosure is approached the contextof the
site, including the palaeochannel, is gradually revealed with an ongoing increase in areas
visible, whilst maintaining the separation from the interior of the larger enclosure. This
pattern is then reversed as the larger enclosure is approached with a reduction in visibility
of all but the immediate vicinity of the causeway and palaeochannel. Once the larger
enclosure is entered the view changes again, becoming internalised but complete,
highlightingthe senseof placeand arrival.
8. Conclusions
The work at Sutton Common has demonstrated an application of GIS for reconstructing
complex elements of archaeological landscapes and palaeoenvironments. This alone can
assist in the interpretation of the site by presenting multiple layers of data together. It can
also form the basis for further analysis using other GIS tools such as visibility analysis.
Theories of movement through landscapes can then be tested through demonstration, to
understand how the site architecture and the environmental context worked together from
the perspectiveof the individual.
This work has also demonstrated that the value of data collected for a single
purpose is not necessarily limited to that purpose and that the application of different
techniquesto themcan providegreater information, Here data obtainedduring attempts to
understandhow the site might be managedhave been used toassistinits interpretation.
Acknowledgements
The topographic survey of Sutton Common was jointly funded by the Carstairs
CountrysideTrustand the Centre for Wetland Archaeologyat the University of Hull. The
monitoring of the groundwater table was funded by English Heritage as part of a wider
evaluationof the site.
H. Chapman / Wetland archaeological landscapes
References
11] Tuan, Y.,Space and place. The perspective ofexperience. Arnold, London, 1977.
[21Tilley, c.,Aphenomenology of landscape - places, paths and monuments. BERG, Oxford, 1994.
[3]Bender, B., S. HamiltonandC.Tilley, Leskernick: stone worlds; alternativenarratives; nested landscapes.
Proceedings of the Prehistoric Society 63,1997,pp. 147-78. .
[4]Devereux,P.,Three-dimensionalaspects ofapparentrelationshipsbetween selected natural andartificial
features within the topographyofthe Avebury complex. Antiquity 65,1991,pp.894-8.
[5]Bell, J., Interpretationand testability intheories about prehistoricthinking. In:C. Renfrew andE.B.W.
Zubrow (ed.), The ancient mind. Elements of cognitive archaeology. CambridgeUniversityPress,
Cambridge 1994, pp. 15-21.
[6]Coles, B., Wetland management: a survey for English Heritage. WetlandArchaeologyResearch Project,
Exeter, 1995. ..
[7]Dinnin, M., The palaeoenvironmentalsurveyof the Rivers Idle, Tomeand Old River Don. In: R. Vande
Noon & S. Ellis (ed.), Wetland heritage of the Humberhead levels: an archaeological survey. Humber
WetlandsProject,University of Hull, 1997,pp.81-150.
[81 Coles, J. and B. Coles, Enlarging the past. The contribution of wetland archaeology. Society of
Antiquariesof Scotland,Edinburgh, and ArchaeologyResearchProject, Exeter, 1996. .
[9] Bernick. K., Hidden dimensions. The cultural significance of wetland archaeology. University of British
ColumbiaPress, Vancouver, 1998.
[101 Sunees,S., Footprints of Roman occupation in the southern part of North Humber Land. Baines, Leeds,
1868.
[111Allcroft, A.,The earthworks of England. Macmillan,London, 1908.
[121Whiting, c., Excavations at Sutton Conunon 1933, 1934, 1935. Yorkshire Archaeological Journal 33,
1936,pp. 57-80.
[I3J Van de Noon,R. and H.Chapman,An archaeological assessment in preparation of a management plan
at Sutton Common, Sutton, South Centre for Wetland Archaeology Research Report
CWAIRESIEH-Sutonl99-I,Hull, 1999. .
[I4] Parker Pearson, M. and R.E. Sydes, The Iron Age enclosures and prehistoric landscape of Sulton
Conunon,South Yorkshire. Proceedings of the Prehistoric Society 63, 1997,pp. 221-59. . .
[151Leick, A.,GPS satellite surveying. JohnWiley, New York, 1995.
[I61 Voigtmann, A., L. Becker and K. Hinrichs, A hierarchical model for multi resolution surface
reconstruction. Graphical models and iniageprocessirig 59, 1997, pp. 333-48.
[171Thomas, J.,The politics of vision and the archaeologies of landscape. In: B. Bender (ed.) Landscape:
politics and perspectives. BERG, Oxford, 1993,pp. 19-48. . .
[I8]Gaffney, V., Z. Staneitand H. Watson, Moving from catchments to cognition: tentativesteps towardsa
larger archaeological context for GIS. In: M. A1denderfer and H.D.G. (ed.), Anthropology,
space, and geographic information systems. Oxford UniversityPress, Oxford, 1996, pp. 132-54.
[191 Barrell, J., Fragments from antiquity. An archaeology of social life in Britain, 2900-1200 ec:
Blackwell,Oxford, 1994.

60 61
T G.Lock!Session J discussion
Beyondthe Map
G. Lock (Ed.)
lOS Press, 2000
Session 1discussion:
aparticular view
GaryLock
InstituteofArchaeology
University ofOxford
The three papers in this first session are concerned with issues ofvisibility, vision and
perception within a general application context ofunderstanding past landscapes. What
follows isnotadetailedaccountofthediscussionperiodbutapersonal viewonwhatIfeel
are the twomainpoints thatemerged from it.The first isthe importance and the natureof
modelling, and secondly how modelling vision/visibility/view helps or doesn't help in
understanding past landscapesandthoselociwithinthem thatwecallarchaeological sites.
It was generally agreed that modelling is central to the process ofresearch as a
quest for new understanding. Whether modelsare formal and explicit or more informally
implicit within ideas, they are extensions of thinking about a problem. As analytical
constructs designed toprogress understanding, models takeon an extra dimension andan
added significance within adigital environment.The very actofcodifying data fordigital
use isan act of modelling witheven thesimplest database being a representation of real-
world entities and relationships. Structured intoalgorithms thatwe may have littlecontrol
over when performing an analysis, are assumptions andprocesses thatare fundamental to
themodelbeingworkedwith.
Generating a viewshed istheoperationalisationofamodel that iswithin aspecific
process of questioning. Like most other computer-based actions it will always produce a
'result' and it ishere the problems begin. The meeting agreed that the modelling process
should be aniterative onethat stimulates thinking and new ideas so that aresult isnotan
absolute conclusion but an interim stage of understanding and a platform from which to
launch new work. This is in the tradition ofstatistical Exploratory Data Analysis and the
oftenquoted 'what ifanalysisencouragedbyGIS.Re-worked bypost-processual thinking
this iterative process of knowledge acquisition has become the hermeneutic circle, or
spiral, where 'pre-understandings' are placed in constant contention with data and their
interpretation isup-datedaccordingly.
This creates a situation of 'playing' with data, of 'teasing out' patterns and
relationships that register within a wider framework ofinterpretation. It also makes the
research process a creative activity, an activity devoid of a rigid procedure aimed at
objectivity, anactivity thatdevelops itsownaesthetic(setof social values) as more people
become involvedinaddressingthesameproblemarea.It isherethat theproblem hintedat
above begins because notions of creativity and aesthetics do not sit easily with those of
validation and substantiation. It seems to me that viewshed studies have come up against
this impasse. Initially driven by a combinationof technological determinism as most GIS
include a viewshed routine, together with the irresistible creative playing involved in
viewshedgeneration, thevalueofviewshedsrapidlygained currency. This wasshort lived,
however, mainly for the reasons detailed by Dave Wheatley and Mark Gillings in their
paperabove, whichshow thatour understanding and application ofvision and visibility to
archaeological studies are theoretically impoverished and that the standard algorithms
providedbyGISsoftwarearenaively inadequate.
These failings have been brought into focus through a process of validation, a
process ofcomparison with something else to establishvalue. Not 'hard' validation as in
thequantified acceptance orrejectionofstatistical hypothesis testing, buta 'softer' version
that can be defined as 'sustainable under criticism'. This is criticism from both the
immediateacademy involved in that particular research interest,and also thecurrent state
ofthewideracademy intowhich itfits.Inwhatfollows,T would liketoexplore thenature
of this validation process further and question to what extent viewshed analysis will ever
beabletoprovideanunderstanding oflandscapewithinanacceptablevalidationscheme.
Central to this will be how the relationship between people and landscape can be
modelledwithin GIS, aconsiderationnotonly fundamental to viewshed studies butto the
application of GIS more widely. This reflects the GIS and Society debate outlined in the
Introduction tothis volumeand iscaptured inCurry'sclassification of GIS into PaleoGIS
and GISz [I]. The former, which is most of present GIS applications and certainly all
archaeological ones, are defined by their under-representation of the basic elements of
humanexperience thatgivemeaning totheworld. WithinCurry'slistofwhatisneeded to
overcome this under-representation in the construction of GIS'2 is 'the creation of
community', a return to a pre-modem past and an appreciation of community based on
sharedconnections toplace.Curry also includes anabilitytorepresent time asanessential
formovingbeyond PaleoGIS, not timeasmeasured chronology which isalready possible,
but 'lived time' as an experiential narrative. Implicit within any snapshot of a landscape
(i.e.a GIS coverage) is time as past,structured through spatial relationships, and time as
futurerepresentedbypotential.
This notion of community has great resonance within archaeological applications,
especiallyinthecontextofrecentdiscussionsofspace,placeandhumanised landscapes.A
community isheld together by a shared set of constructed categories for making sense of
the material and social worlds, categories that also allow for individual expression and
development. These categories include spatial ones so that places become lociofhuman
actions where individual and group social relationships are negotiated and re-negotiated
time after time. This stands in sharp contrast to the traditional concept oflandscape as
developedby 16
th
century artists, where.landscape isobserved fromadistance rather than
livedin,aviewreinforced bythedevelopmentofcartographyandcontinued byGIS.
Thistraditionofspatial representation, asmeasured andmapped detachedspace, is
whatconstrains Curry'sPaleoGIS andwhatisso difficult togetbeyond. One illuminating
insightinto the complexities involved in taking this step isthe Parish Mapping Project of
the environmental group Common Ground [2J. Through the construction of maps as
perceptions of individualplaces infinitely linked into awebof local meaning, thisproject
shows how individual understanding is negotiated toproducegroup connections with the
landscape.Theaimof theprojectisto'encouragecommunities tochart thefamiliarthings
whichtheyvalue intheirownsurroundings, andtogive activeexpression totheiraffection
fortheeveryday andcommonplace' [ibid:236].The resultsare 'maps' intheloosestsense
of the word that incorporate. drawings and stories and give meaning to. places for
individualsand groupsofindividuals connected in various ways.It isinherent within this
that a single place can mean many different things to many different people. Places are
fluidand take on meaning based in the experiences and knowledge thatpeople bring to
them. Members of a particular group, whether the Girl Guides 'or the local Police, will
bring another layer of meaning toaplace on top of the individual'sbased on thegroup's
ethics, values and interests. Developing an understanding of this complexity shows how
landscape as lived in stands in stark contrast to the cartographic representationof diose
63
62
Beyond the Map
same places as presented in the maps of the Ordnance Surveyor any other traditional
mapping agency.
An emphasis that comes through time and again, and one that is central to this
discussion, is on the interaction between places, people and meaning. Much of this is based
in the writings of recent social theorists and is detailed and referenced in several of the
papers in the first two sections of this volume. Bourdieu's logic of practice, for example,
sees landscape and habitus as mutually creative and the structuration theory of Giddens has
space as being produced by, while at the same time producing, social action. Connections
between understanding place and cartographic representation are stretched even further in
the phenomenology of Heidegger, another important figure who looms large in
archaeological writings of the last decade. Of particular interest is his essay Building
Dwelling Thinking [3] in which he accounts how dwelling is the basic characteristic of his
often quoted notion of 'being' in the world, and, in fact, the relationship between humans
and space is none other than dwelling. His description of what it is to dwell is, of course,
not simple and revolves around the idea of the fourfold, that is earth and sky, divinities and
mortals. Through their relationships with locales, both topographical and through building,
people create places where the fourfold is safeguarded, where the four 'unfold' allowing a
primal 'oneness' with a place which produces the sense of dwelling or of being at home
there. Locales, therefore, house human life by bringing together all or some of its aspects
in that place; aspects of earth and sky, the physical and spiritual needs of divinities and
mortals.
So the complexity of places increases and it is not surprising that PaleoGIS are
unable to come anywhere near offering a meaningful digital representation. It seems to me
that the main reason for this is because we are trying to incorporate social/cultural
information into the landscape itself whereas it actually resides within people. A particular
locale does not carry meaning inherent within it but takes on meaning for an individual or
group when he/she/they enter into it, engage with it, think or talk about it. As suggested
above, that same locale can take on different meaning and values to different people and
can change with time and with context for the same person/people. This can be thought of
in the wider context of Jean-Paul Sartre's ideas of how people interact with the physical
world, with material objects including places [4]. He suggests that things act as reflectors
in the sense of returning the light to where it was received from. Substituting meaning for
light, the assumption is that material objects can only mean what we already know and
through a continuous process of interaction that meaning is altered and enriched. Put
another way, our understanding of a place is within us and surfaces when we encounter
that place. Indeed, according to Sartre, the emotional connections to objects and places can
become so embedded that it becomes 'impossible to distinguish between what is felt and
what is perceived".
There are, of course, many other aspects of places which enrich their meaning and
contribute towards that vague notion of 'a sense of place'. Historical depth and connections
with past people and events can be fundamental in constructing meaning. A place may take
on additional/different meaning through having sacred connotations and, perhaps, being
involved in ritual, not necessarily as formal organised large events but ritual practices on
J It is not difficult to translate these ideas into personal experience. My own favourite is the football stadium
which I have been going to regularly for the last thirty-five years with my father and brother. The emotional
surge upon entering the ground brings to the surface family bonds, connections with birthplace, a floodof
memories (Coventry City won the FA Cup in 1987, all three ofus went andstill talk about it), temporary
allegiance to a group (home supporters versus visitors), negotiations with power structures (constant booing
of a poor player can convince the manager to drop him), not to mention the whole range of structured
practices which are continued and reinforced by group involvement (wearing certain clothes, singing certain
songs, performing certain actions etc).
G. Lock / Session I discussion
an individual or small group level. This introduces the potential importance of context and
how a place can adapt and change according to the different context of practices being
carried out therein. Other aspects of material culture are crucial here, the positioning and
use of portable artefacts, the consumption and disposal of food and drink, the erection of
temporary structures, the digging and filling of holes in the ground, the decoration of
natural features. All of these can both take meaning from, and add meaning' to, the context
of human action and the process of understanding the material and social world.
A remarkably powerful account of the power of places is Keith Basso's book
Wisdom Sits in Places, an ethnographic study of the Western Apache and the role of places
in their daily lives [5]. Named places in their landscape. are not only fundamental to their
understanding of history, time and practical activities like the logistics of hunting, but also
provide a means of re-enforcing and transmitting the moral framework within which the
tribe lives. Places have specific stories attached to them concerning activities of the
ancestors and the stories contain coded messages about the 'correct' way to behave. By
telling the story, or sometimes by just speaking the place-name,. a mental image of the
place and a re-enactment of the story is triggered in the listeners, a mnemonic which is
only possible because people know their own local landscape intimately.
Faced with this richness and complexity what can we hope to achieve through the
digital representation of places? Even with the full force of new irnmersive VR
technologies it is unlikely that many of the subtleties of a place outlined above could be
captured. Using 2-dimensional GIS and viewsheds the task seems almost impossible.
However, research usually progresses through accumulated small steps and we can see
from the three papers in this section that progress is being made. It is no longer acceptable
for viewsheds to be applied unthinkingly within an apparent theoretical void and these
papers offer a sound framework within which to place any application. Between them the
three papers also offer a range of interesting ideas that can be pursued to create richer .
viewshed analyses more fully integrated into attempts to understand past landscapes.
In my view, however, the greatest challenge is in going beyond the patterns
produced by these techniques to possible meaning. All viewsheds, whether Higuchi, fuzzy,
probabilistic and/or directional, as well as smellsheds, noisesheds and any other quantified
representation of human senses used within current 2D GIS, will ultimately produce
minimalist patterns. Those patterns will contain and enable spatial relationships with other
aspects of landscape and archaeology worthy of interpretation, but how can interpretation
incorporate humanistic aspects like those above? As already suggested, it seems to me that
if understandings of a place can be captured they need to be.coded into the virtual people
who inhabit the landscape rather than into the virtual landscape itself.
Obviously this is not going to be easy although two possibilities are already at hand
and worth flying as kites. Firstly the potential of Object-Oriented GIS is almost completely
untapped in archaeology. Utilising 00 functionality, individual places (objects) could
adopt different 'behaviour' (i.e, characteristics) that is triggered by different people
entering them. As a simple' example, -if the area 'of study was thought to contain two
kinship groups a particular place within it may have historical connections for one group
and not the other. The place, therefore, has two different sets of 'behaviour' depending on
the affiliation of the person entering it. The second possible area of development is similar
but based on simulation so that actors are encoded with 'cultural' knowledge and as they
move through the landscape so places take on meaning by reading it from them. The
interesting aspect of both of these scenarios is the resonance with Sartre's notion of
reflection and the emphasis on places taking meaning from people.
There is no doubt that within the interpretation of archaeological landscapes the
constraints of PaleoGIS are well documented, and have been for some time. The
metamorphosis into GIS2 is proving difficult and any progress that has been made so far
':__ , ~ , , ~ c -
64
Beyond the Map
has been based on attempts to understand the social and culturalcomponentsoflandscape
and to integrate these within the functionality of the technology. It is not new for
archaeology to be pushing at the peripheryofwhat the technology was designed for, and
while landscapesremain being about people itseems likely tostaythat way.
References
(I] Curry, M.R. Digltal Places. Living with geographic information technologies, Routledge, London, 1998.
[2]Crouch,D. and Matless, D. Refiguring geography: Parish maps of CommonGround, Transactions ofthe
Institute ofBritish Geographers, NS21, 1996,pp. 236-255.
[3] Heidegger, M. Building Dwelling Thinking, in D. F. Krell (ed) Basic Writings. Martin Heidegger.
Revisedand expandededition,Routledge,London, 1993,pp. 347-363.
[4] Sartre, J-P. Thephilosophy ofJean-Paul Sartre, R.Cumming(ed), VintageBooks,New York, 1965.
[5] Basso, K. H. Wisdom Sits in Places. Landscape and language among the Western Apache, Albuquerque.
UniversityofNew MexicoPress, 1996.
65
Beyond the Map
G. Lock (Ed.)
lOS Press. 2000
Understanding movement:.
apilotmodel towards the
sociologyofmovement
Marcos Llobera
Institute ofArchaeology, University College London .
Pitt Rivers Museum, University of Oxford
Abstract:In recent years, new approaches have emerged (especially ~ t n the
EuropeanarchaeologicaLconlext)that rightfullyaddress thecomplexitiesof studying
ancient landscapes. They represent, in a sense, a reaction to the spatial models
(originatingin EconomicGeography) heavily used and developedduring processual
archaeology. Unfortunately. these approaches rest on observations that are hardly
validated. In fact, they seem tobe opposed to any form of methodology. It is within
this contextthat the application and further developmentof GIS is vital in order to
study past social and experiential processes that occurred within a landscape. In'
particular. this paper guides the reader through the construction of a model to
explore the process of movement within a 'natural' landscape (as opposed 10an
urban one). The two most importantcharacteristicsof this model are: the retrieval of
'general,' patternsof movementrather than specific paths and the notion that these
patterns(theircreation,evolution,etc) are a vital piecein the socialisationof peOple
into a landscape. The model does not seek to produce objective results ('TIlE
RESULT') for it is finnly believed (as shown throughout the paper) that models with
'humanistic' aims do not produce such results. however, they do produce
descriptions of the world in which biases are well identified, can be re-evaluated
continuously and have enormous'potential to precipitate new insights. The paper
concludes with a description of the results so far obtained, some ofthe difficulties
encounteredand ideas for future developments.
1, Introduction
"It follows that the construction of a monument guides movement, conditions the way in
whichaplace is 'read',and itisinstrumentalinthecreationofsubjectivity"[43].
This and similar statements are often found in archaeological accounts attempting to
describe the social construction of being within the landscape of a certain time period.
However convincing, these allegations are hardly ever pursuedanyfurther. Questionssuch
as: howdomonumentsguide movement?Howdoes the introductionof alandscape feature
interact with existingpatterns of movement? are seldom discussedwithin 'thespecifics ofa
study area. These descriptions are usually based on the researcher's.appreciation, which
often remains at a theoretical level. They are incorporated within a narrative and lack any
formofvalidationormeansbywhich theycanbefaithfully explored,
This paper guides the reader through the creation ofa new GIS routine, named
66
Beyond the Map
move, aimed at tackling some of the criticisms mentioned above. Itsgoal is to explore, to
attemptatleast,thedynamicsof movementinanatural,asopposed toanurban,landscape.
During the process of design many questions emerged for which no immediate answer,
eitherat the theoretical orpractical level,could be found.The model is still in its infancy,
muchdiscussion and further development need to be undertaken, but its implications and
potential are well illustrated here. The aim of this paper, more specifically, is to fully
describe the model (i.e. the rationale behind it, its implications), to forward several
proposalsonhow to modelaspects relatedto movement thathave neverbeen addressedin
arigorousmanner,andfinallytopresent theresultssofarobtained.
2. Aboutmovement
2.1 Theoretical fabric
Manysubject areas have made direct or indirect reference tothe movement of people ina
landscape,e.g. ethnography [7], anthropology [13], [271,[381,[471,geography [18], [201,
[40], philosophy [33], sociology [51,[151,experimental and social psychology [141, [171
[39],urbanism[251, andsocialphysics [25].
Tacit within thecontext of current archaeological accounts (see [21,[61,[191,[431,
[44]) is the constant reference to movement as part of the socialisation process of an
individual. It is within this background that the following model must be situated. In
archaeology, this may translate into many questions like: how did the incorporation of
certainfeaturessuchasfunerarymonuments orsettlements affectand reflect thepatternsof
movementinalandscape? whatarethesocialimplicationsofchanges inmovements?
The sociological background forthis model canbefound in Pred'sPlace, practice
and structure: social and spatial transformation of southern Sweden [40]. This work
represents the successful marriage of two distinct fields, Hagerstrand's Time-Geography
and Structuration theory. It includes a useful synthesis of the ideas and goals found in
currentarchaeological approaches tolandscape. Inspiteof itsnon-archaeological nature,it
providesanexceptional example oftheuseofsocial theorytogether withconcrete data.
AccordingtoPred [ibid:51:"the structuringofspaceisinseparable fromtheprocess
of social structuring." He stresses the idea of investigating the interrelation of
geographicallyspecific practices with social and spatial structures. In his work, Pred sets
the goal of presenting "a theory of place as a historically contingent process that
emphasises institutional and individual practices as well as the structural features with
whichthosepractices areinterwoven." [ibid: 6-7].Particularlyimportant forarchaeologyis
the acknowledgement of the role of natural and human-made objects in the creation of
time-space practices. In setting the background for his work, he outlines the main ideas
includedwithinstructuration:
1. "For anygivenarea,social reproduction isaconstantlyongoing process
that is inseparable from the everyday performance of institutional
activities, including those mundane practices associated with the
institution of the family or household. More precisely, such activity
performance results in the perpetuation or modification of the
institutions themselves, of the knowledge necessaryto repeat or create
activitiesandalreadyexistingstructural relationships [i.e.habitus],
2. Through socialisation, rules of behaviour are absorbed and become
taken for granted and the skills and competencies appropriate to given
67.
M. Llobera I Understanding movement
social context areacquired andreinforced. Socialisation occurs through
participating in institutional activities. Socialisation persists through
adulthood. for societal institutions are not a 'static or finished order,'
butarethemselves inastateofbecomingthroughhumanactivity.
3. In the simultaneous unfolding of socialisation and social reproduction
the individual andhisorherconsciousness areshaped bysociety, while
society isunintentionally andintentionally shapedbytheindividualand
hisor herconsciousness, i.e.socialisation andsocial. reproductions (and
transformations)alwaysbecomeoneanother. .
4. Inorder todeal withthedialectical relationships betweentheindividual
and society, the constant becoming of both, one must really deal with
material continuity and the dialectics of practice and structure. which
aresynonymous withtheunfoldingof structuration processes.
5. Asanysetofstructuration processesrelentlesslyproceeds, thestructural
properties of a social systemexpress themselves through theoperation
of routine and non-routine daily practices. At the same time, daily
practices generate and reproduce micro- and macro level structural
properties of the social system inquestion. The shaping of practiceby
structure andthereproductionortransformation of structure bypractice
are alwayspartofthesameprocess"[ibid: 8-9].
Predpairsthese ideaswithconceptsoftime-geography,whosetwomaintenetsaretheidea
of time and space as scarce resources andthat the constraints to which humanbeings are
subject.especiallyinthephysicalenvironment, dictatemuchofthepatternofeveryday life
[45].The incorporation oftime-geography,isnotonlybeneficial'atatheoreticallevel(due
to the syncretism that emerges from this union) but also at a methodological one, as it
providesasetofpracticalconcepts thatcanbereadilyappliedtospecificcontexts.
Relevant to this work is the time-geography discussion on constraints, which.are.
groupedasfollows:
Capability constraints' limit the activities of the individual through both hislher own
biological make-up andthetoolshe/shecanuse.
Coupling constraints arise from the necessity that individuals, tools and materialsare
boundtogetheratcertainplacesandtimes. ...
Authority constraints imposelimitedaccesstoeither spaceortimelocations.
Movement can clearly be related to everyone of ihe above. It is obvious that slope, soil
type and vegetation cover, are factors that influence the movement of the individual
throughalandscape (i.e.capabilityconstraints).Thealterationofthelandscapethrough the
construction of a newlandscapefeature(e.g.aditch), orbythe modificationof landcover
willaffectexistingcouplingconstraints, modifying;creatingOf rendering someimpossible
to maintain. Finally, there is no betterexample of the exertion of power bysome regime
thanitscontrol over themovementpatternsofpeople,see[12]. .
Pred maintains thattheconceptsofpath and(institutional'')project makethenexus'
with structuration possible [40: 101.Here,a path refers to each of the actions andevents
I Some authors, especially phenolllCllOlogists, see [11], [361,and (42), wouldsee theseas the basis. foc our
experiences,andwouldarguetballlIlherthanconstraintstheycanbe viewedase....blingfaclOrS, e.g.[15].
Z Institutionalprojects or practicesan: those embeddedin theexistingsocial structures, someof whichare
dominantintermsof theirinfluenceon thedailypaths and lifepathsofspecificpeopleand,therefore,upon
the fonnation of an individual'sconsciousnessdevelopment and socialisation [40:121.
T
68
69 .
Beyond the Map M. Llobera / Understanding movemens
that makes up the existence of an individual, each of which possesses spatial and temporal
attributes. Human beings must be seen as purposeful agents who possess their Own
objectives. In order to attain these, they engage in projects, a series of simple or complex
tasks needed to complete any goal-oriented behaviour. In the end:
"The integration of time-geography with the theory of structuration becomes possible when
it is recognised that each of society's component institutions does not exist apart from the
everyday and long-term production, consumption or other projects for which it is
responsible. If alI of society's formal and informal institutions are project-bound and if all
projects require some human participation, then the folIowing propositions may be stated:
The detailed situations and material continuity of interpenetrating structuration processes
are perpetually spelIed out by the intersection of particular paths with particular
institutional projects occurring at specific temporal and spatial locations" [ibid.].
The above sets the arena in which the folIowing model is constructed. OveralI, emphasis is
put here on general patterns of movement. Movements to which no particular, or conscious
purpose, is attached for they are imbedded in everyday life. These are patterns that have
been welI 'sedimented' through time, and form part of the habitus of a group. It is by virtue
of these that an area of land becomes familiar to an individual and a community. In
addition, they also relate to other patterns, those of co-presence [17] that set the scenario
for possible encounters with other people and resources. Thus, they are necessary in order
to maintain those modes of colIaboration and production that characterise a social group. If
these patterns are modified, or transformed, changes to the rate and conditions of social
interaction will folIow. It is clear, therefore, that movement throughout a landscape holds a
direct relation with the creation, maintenance, and transformation of social structures.
2.2Formal stratum
Formal approaches to movement cover a wide range of disciplines, from biomechanics to
social physics. Each of these covers different aspects, aims and scales of human movement
and not all of them make use of GIS. Current GIS applications only target specific cases,
particularly those where the origin and destination (or simply the origin) are known. Hence,
they cannot inform us about the potential and importance of other origins and destinations,
that is, about more general movement patterns.
Within an archaeological context, movement as an element in the socialisation
process has never been addressed directly. Ancient road networks and settlement
hierarchies in the New World, edited by Trombold [46] is one of the few archaeological
monographs where movement is inherent in all of the case studies presented. Still,
emphasis is put on roads and trackways as artefacts, on obtaining their typology and
establishing a hierarchy among them, rather than concentrating on the process of movement
per se. No attempt is made to consider social or more experiential aspects, to relate road
morphology with its social impact on the traveller or to study his or her interaction with the
surrounding landscape. In addition, case studies are predominately synchronic; alterations
made on paths through time are not related to social changes but incorporated into an
evolutionary perspective, e.g. [46: 10]. Kantner [29] provides one of the few GIS examples
where a GIS is used to calculate the optimal paths based on minimal time travel efficiency,
that connect several Anasazi communities. These are later compared to existing trackways
and evaluated in relation to three general models of Chacoan road function. Kantner's work
highlights some of the simple possibilities that are readily available within most GIS.
Unfortunately, his model requires the presence of known locations and remnant ancient
tracks, and presents a synchronic view, a view that does not consider previous and
subsequent archaeological information. It is necessary to move Into the areas of social
physics and urban planning in order to find something closer to our interests.
Briefly, social physics is based on the application of models derived from the
physical world to social phenomena. This is not a foreign field. to archaeologists (early
spatial gravitational models derived from this area). In this particular case, principles found
in fluid dynamics are applied to the dynamics of pedestrian movement [23], [25]. So far,
these works are characterised by their micro-scale and have yet to be applied to a large
landscape. According to Batty [I], one of the hardest issues, whenever quantitative
methods are applied to human systems, is the inability of these models to link spatial
structure to social behaviour. An exception is the work conducted by Helbing et al. [22],
which ilIustrates how spatial structure, i.e. trackways, emerge from social behaviour. In
doing so, Helbing provides an exciting alternative which could potentially be adapted to
archaeology. In urbanism, the work of Hillier and Hanson [25], especialIy Hillier's most
recent work [24], stand out as an important contribution to understanding the dynamics of .
movement. Though their work is mainly developed Within an urban context, many of the
concepts and explanations proposed are potentialIy applicable to natural landscapes.
Particularly relevant is Hillier's discussion on the impact of an urban layout on patterns of
pedestrian mobility. He claims a cause effect relationship between the index of integration
a location has within a spatial arrangement and movement through this location. Though
undoubtedly very insightful, Hillier's work does not escape criticism. BasicalIy the fact that
movement cannot be reduced to a measure of urban fonn;as new work done at the Centre
for Advanced Spatial Analysis (CASA) currently maintains [41].
3. Modelling Movement
3.1Aim
Before digressing into details of the model, its aim needs to be further clarified.
Formulating a strict definition is often seen as the major 'limitation' of models and
information systems in archaeology. This is due, primarily, to their inability to incorporate
information that lacks precise numerical expression. Although true to a certain extent, more
often than not this handicap is a reflection'of poor modelling skills found in over-ambitious
(usually top-down) models. The necessity for precise wording is seen here as a strength,
rather than a limitation, as it provides a mechanism to formalise biases within research
(Hodder's pre-understandings [26]). Furthermore, it is argued here that GIS are optimal
tools for incorporating these biases.
Because of the way information 'is stored and handled.in GIS, it is essential to
express our original question (what .are the dynamics of pedestrian movement inuie
landscape?) in GIS raster terms. So by 'dynamics of pedestrian movement' reference is
made to changes in patterns of movement undergone through time. By patterns, we refer to
those patterns created by locations (i.e. celIs) that possess a similar (i.e, high) probability of
an individual moving into or through them. Ultimately, the model seeks to obtain a raster
image in which each celI contains an index (called here accessibility) that represents the
probability of an individual moving into, or through, that cell.
With a few possible exceptions, it is very difficult to imagine any comprehensive
archaeological record that would' describe movement patterns in the past. However,
r
71
70
I
M. Llobera I Understanding movement
Beyond me Map
knowing exactly where people moved is not necessary for we can consider how the
potential (i.e. probability) for movement, as defined above, changed by the constant
construction and re-interpretation of landscape features. An added advantage of focusing
on changes of movement potential isthe fact that all periods and archaeological data sets
canbeincluded,andthatthelandscapeasawholeissubjecttoscrutiny.
Thelikelihoodof anindividual movingintoalocation isequated withthepotential
(i.e. probability) of a location 'to be moved into'. This potential isdirectly relatedtoease
of movement, that is to the easiness of moving into a location from any other given the
spatialconfigurationofthelandscape.Itcanalsobeinterpretedasthedegreeofintegration
atanylocation;locations thatare 'more integrated' are likelytobemoreoftenthetargetof
dailymovementthanthosethataresegregated.
3.2 Movement Cost
In thismodel the likelihood of an individual to move intoa specific location isdefined in
relation to the cost of reaching that location from any other one. The larger the cost of
moving into that location, the lower the probability that an individual will tend to move
intoit.
Costs can be of various kinds and origins (some of them social). For the sake of
clarity,heretheyareseparated intotwo:thecostdue to topography(topographic cost) and
thecostdue to thepresence of anylandscapefeature/s(landscape feature cost), seeFigure
I.
3.2.1 Topographic cost
The shape of the landscape (i.e. its topography) provides the starting point from where
movement patterns originate.Topography isbasicfor itisalwayspresent andits influence
is continuous in every landscape. Hence it constitutes the background on which other
factorsoperate.Inthissense,itsroleisanalogous totheoneperformed bytheurban
e.g.WIumc:e or inlIuma:
of the:pratnU of awJIIIImCfIl,
od><n
In <>1ubbng """'"""'"'
COR INIIJ'othafIctor:IadI
Q; slope
u uorion/sedimaDbon or
IOiJ typt coWot beintrodua::ed
r. U')'ururoftime(e.s- limeperiod. aninIbnce withinconcin\lOl,ll cfuonolociaI Wne,ett:)
Figure 1:Deriving theAccessibilityIndex
layout(see Hillier'sprinciple of natural movement, [24: 166]).Yet inthecaseofanatural
landscape,the impact of itsshapeon movement patterns maynot.always beas significant.
Gentle and undifferentiated topography plays a minor role in the formation of movement
patterns.Hence, thestudyoftheeffect oftopographymustbeviewedasasensiblestarting
pointfrom whichother morecomplicated realitiescanbetackled.
The basic assumption used here isthat ifeverything is'removed from a landscape,
leaving only its shape intact (similar to a 'moonscape'), patterns of movement will be
determined, at least initially, bythe 'easiness' to reacha locationfromany otherone. The
easiness will be described by the level of metabolic energy expenditure per unit"distance
travelled on an incline [28], [32], [34], [35]. The relationship between slope and energy
consumed isdescribed inFigure 2,whereeachpointcorresponding toaslopeangle/energy
expenditure pairhasbeenplotted.
The diagram shows that energy expenditure occurs when moving uphill (from
liftingone'sown weight) as well asdownhill (from constant breaking inorder to stop the
forceofgravity). However, the rate ofchange isobviously higherforuphillmotion.Italso
showsthattheoptimum gradient forwalkingisnot0but-10%(=-5.71).
Information from a Digital Elevation Model (OEM) can be used to calculate the
topographiccost incurred when movingfrom one cell (i.e. location) to anyof its adjacent
ones.Topographiccostisderived fromvaluesgiverrbyatopographic cost file. Inthiscase,
thetopographiccostfilecontains thesamevaluesusedtocreateFigure2.
3.2.2 The effect of landscape features
Thoughnotstraightforward, itispossible toconceiveofawayoftransformingtopographic
information into some sort of energy cost given our current knowledge of biomechanics.
What is less foreseeable is the impact that some landscape features, such,as funerary
monumentsmayhavehad.Thisisaddressedinthissection.The work,stillatapreliminary
stage,representsafirstattempt toexplorethiselusive issue.
Topographic Cost
.., . . ,.
. , :
Mi_tti 1995 .r.
.,.
.,. .,.
r
Created
.,.
. ".
I
(I_d
. \. .
L-gWllJi,.I99n
., . . ,
"
.
.
.;I.
. '\ '
I.:
: ".
.-...";;,,
-70 -60 -50 -40 -30 -20 -10 0 10 20 30 ) 50 60 70
Slope(.....>
Figure 2:Values from+30
0
to +49
0
arederivedforanindiVidualclimbing anescalator
(extractedfromFigure 2BinMinetti1995 (35))
72
Beyond the Map
Within the context of this work, the term landscape feature refers to any element of the
landscape that has some effect on the movement of an individual, i.e,that alters the basic
cost due to topography. Itcan be represented by a point (e.g. a burial site), a line (e.g. an
existing trailoraditch), andanarea(e.g,patchofvegetation).
Given the apparent lack of any discussions on this matter, the following
descriptionsaresuggested. Landscapefeatures haveaneffect:
At a physical level. The individual does not have anyparticularinterestinthe feature butis
still affected byits presence. The featurecan act asa barrier (absoluteor with some degree
of permeability)e.g. a fence or ditch. It can also facilitate movement, as in the case of a
trackway.
At a mental level. The materialityof thefeature iscoupled with the 'intentionality' ofthe
individual. Intentionality is viewed as a social construct inherent in the formation of the
individual as a social being. What, when and how any feature affects an individual's
movement fonns partofthisconstruct.
Obviously, itispossibleforafeature tobeoperating atboth levels. Inaddition, within each
landscape feature wemaydistinguishthreeelements:
Amagnitude. The impact on, or importance of, the movementprocess may vary from one
landscapefeature toanother (i.e. theexistence ofahierarchy?). This maybe establishedby
one or a combination of criteria, e.g. friction, the presence ofprestigious artefacts or the
prominenceof the location in the landscape. Ultimately, the magnitude may be thought of
asacost; thecost given bythepresence ofalandscape feature.
An area of influence. If acting at a physical level, the area corresponds to the physical
dimensions, if any, of the feature. At a mental level this correspondence does not
necessarily holdtrue. Afeature maybe thoughtofashaving anarea thatextendsbeyond its
physicality andover whichitcommandsitsinfluence. The magnitudeofthefeature (i.e. its
effect) may,or maynot,bespread uniformly withinthearea.
Abehaviour. Threeofthem aredistinguishedhere:
1) Repelling movement. In this case, an individual avoids moving near the feature. An
example of this behaviour is the avoidance of burial sites that some societies profess
during the liminal phase.
2) Attracting movement.
3) Being neutral.
Each of these elements interacts with each other and may be described ultimately, as
functions oftime (among otherparameters).
In order to illustrate the above, and their description within a raster GIS, an
example with two point features (e.g. tombs) has been worked out in the following
paragraphs. On this occasion the viewshed of each feature is used to define the area of
influence, Figure 3, so that any location in sight is considered to be affected by the
monuments. A different colour grading is used to indicate the change ofimpact on each
location (note that the monuments have different levels of importance). The behaviourof
themonument canbe modelled usingpositive valuestodescribeattractionand negative
73
M. Llobera / Understanding movement

.0lJ:0I...".",
V... 1un fealiM1
ViewIaI\ talIAe 2
_ \few"0Ift JeaIue 1.wd2
Dr..... '
III r.-.2'
Figure 3: Sphereofinfluence(for each feature)based on visibility
onesto indicate repulsion. Ineach oftheexamples,thedistanceaway from thefeature has
adifferenteffect on theoriginal impact ofthefeature.
Withoutfurther informationwearefaced withseveral questions.
How, if at all, does the significance ofeach monument spread (vary) throughout its areaof
influence? Here, several alternatives areprovidedgiven the lack of data on which to base
an answer. Each of them shows a different impact as the distance away from the feature
increases.
Figure 4shows what itwould belike ifthecost (e.g.significanceofthe monument)
was 'spread' linearly throughout the viewsheds. The cost decreases linearly, using the
followingformula, from thefeature outtowards the rimoftheviewshed:
c,(d)
-magnitude. d+magnitUde
d
max
where,
magnitude: rankof thefeature
d: distance awayfrom thefeature
dmax: maximum distance away from feature within theviewshed
Unear Decay
80 __ ____ 1: 60 ==----
o 40
'""I
U I
20 .4. ... g
o
C _ _
Hilflo ....... L......' Oi.lance!'wnIty(M)
Figure4. Linear decay impact Figure 5. Transect showingLinear Decay
- - -
74 75
r
BeyondlheMap M.LIoberaI Understandingmovement
In this case both features would be attracting movement. Figure 5 shows a transect for
feature one, inwhich the impactofthefeature decreaseslinearlythroughoutits viewshed.
Figures6 and7 show how the impactofthe features would spreadif anexponential
decay function was used.
ExponentialDecay
80
: 6 ~
0.40
-20
o :;==;::.=_---.
o ~ g coo 0
N en ~ 8 ~ ~
~
Dis11Ulce Away (m)
HI"'....,... ......
Figure 6.Exponentialdecay impact Figure7.Transectshowing Exponential Decay
CI (d)= magnitudee( -k.d)
where,
magnitude: rank of thefeature
d: distanceaway from thefeature
Ir(O.OOY. gn;tude)
k: constant, k = - ma kiscalculatedso that the magnitude
d
max
iscloseto0at therimoftheviewshed
dma.: maximumdistanceaway from feature within theviewshed
The exponentialfunction would beutilised when the effect of the monument isfelt only at
averycloserange. In the final example,stepwisedecay, Figures 8and 9,the full impact of
the feature (i.e. all its magnitude) is felt within the immediate surroundings of the feature.
However, at the border, the impact depends on the number ofadjacent locations that are
out ofsight. This isdescribedthus,
magnillJde. for any locationfully within the viewhsed (i.e. n = 0)
cf (d) = { magnillJde, for any locationonthe edge of the viewshed
1+n
where,
magnitude: rank ofthe feature
n: number oflocations out ofsight
How to combine the effect ofvarious landscape features? What if the features have
differentbehaviour?Theseare two unresolvedquestionsprecipitatedby the model. In this
case, the spread ofeach feature was calculated separately and then combined together.
Those locationsthat fell within the viewshedof both features were assigned the maximum
cost, inthiscase out ofthe possibletwo. Another alternative,rather than choosingthe
StepwiseDecay
.. 1l
':1
t; : _. _
o I I I I
C ~ ~ ~ ~ ~
L.,M1..... 0;...."""""'" (..)
H'I" lm,...
Figure 8.Stepwise decay impact Figure9.Transectshowing stepwise decay
maximum,wouldhave been toaverage them out.
Finally, the way in which the effect oflandscapefeatures is combined with the
topographic cost, is also arbitrarily chosen. This points. out, once again, our lack of
knowledgeon the specificssurroundingthese matters (which has more;weight, topography
or the presence ofa feature?). A simple solution is offered here. The cost given by the
landscape feature is comparedwith that ofthe topographyand the maximumis selected if
the feature is attracting otherwise, the minimum is selected (i.e. when the feature is
repelling).
All the above hints at the complexityinvolved in simulating the'impact of a series
of monuments on a landscape. A complexity not represented in.current archaeological
narratives.
3.3Therepresentationoftheculturallandscape
At this point, we are ready to combine the above ideas in order to determine the initial
parameters of the model. To ibis end, the OEM information is read, combined with any
existing landscape feature information(whether operating at a physical.or mental level),
and finally transformedin order tocreatea bi-directionalmathematical graph. The graph
contains for each location(i.e.cell),the cost of movingfrom it to any of itsadjacentcells.
This isaccomplished by means ofaroutine that creates a graph using adjacency lists (i.e.
linked lists) after reading the OEM, and combining it with.any other rasterirnage
describingthe impactofoneor several landscape features stored ascell values.
The goal ofthe entire routine is to obtain a value that expresses the probabilityof
moving into a location (i.e. a node inthe graph). The costs stored in the adjacency list ofa
node (i.e. location)are not theonesobtainedbymovingfrom one node toits-adjacent ones
butthereverse,ofmoving from anyadjacentone into the node inquestion.Thisdistinction
is important given that the functions for some movementcosts (e.g. topographic) are not
symmetrical(i.e, itisnot the same tomovefrom AtoB,asto move from II toA).
3.4The movementprocess
The mathematical graph, as described above, must be thought of as a simplified
representation of the cultural landscape during a period of time, after the effect of
topography and extant features have been combined. Once it has been constructed. the
76
r
Beyond the Map
movement of an individual can be simulated in order to attain, ultimately, the accessibility
index for each location.
Accessibility: So far, only the cost of moving from one node to any adjacent one has
been discussed. Dijkstra's shortest-path algorithm [9] implemented using a binomial heap as a
priority queue, is used to derive the cost of moving between nodes separated by one or
more intermediate ones (needed to calculate the accessibility). Using this method, the
minimum cost path of going from one node to any other node on the graph is calculated. If
additional information beside the topography has been included in the construction of the
graph, the minimum cost path can no longer be thought of as being the most efficient from
an energetic point of view. To calculate the accessibility, this calculation is repeated until
all paths going into a particular node (source) have been obtained. These lire then added
together and divided by the number of paths, giving the average path cost of moving from
any node in the graph to the source.
A radius of search, centred at each node (i.e. location), Figure 10, is also introduced
in order to limit the number of locations included in the average path cost calculation. This
allows us to calculate the accessibility associated with different ranges of movement, from
local to global. Thus formally:
The accessibility a; for any location I given a radius r, is defined as the reverse to the
average path cost c of moving from any location within r to location I.
r
a',r =100 - c'.r
vt6I:Qet bOde (i.e locatJ.an)
..
~

.\1 ...
. *\ 'y
pnodeS
..-1
. ;

. ." 'f' 'W'

r..... . . ./j.' 'K
.v. .. . . .......
"-.
. . qlobal nd1us
local [MillS
.. .

. . .
. . .
D locaI..-..nI
~ t
edoeat I:UUI: 1.6Ot:
~
Figure 10. Using a radius to determine Accessibility Indexes for different ranges
The collection of parts so far described (created as a series of routines), together with other
I/O and subsidiary routines, make up the GIS routine move (i.e. the formal expression of
the movement model). Initially, the intention was to calculate the accessibility index within
the program, however, this was not possible. The main problem was the standardisation or
normalisation of each of the path costs. The path uniting any two nodes is made up of any
number of intermediate nodes; the only known value that can be used to normalise each
path is the total number of nodes in the graph. But this number proved to be too large, it
77
M. Llobera I Understanding m o ~ m n t
was decided that standardisation (i.e. obtaining the average path cost) and the later
derivation of the accessibility would be done using map algebra. Thus
the move routine returns a raster image in which each cell contains the total path cost of
going from each cell to each other cell within a series of radii.
4. Assumptions
The model incorporates the following assumptions:
Patterns of movement are identified with those locations into which an individual is
more likely to move, i.e. with high accessibility. Together, they define stretches in the
landscape that are likely to become pathways, [22].
A sense of most 'efficient' behaviour is implicit in the model. This behaviour,
however, should not be compared with ecological efficiency. Agents here are efficient
insofar as they choose the best moving option given the existing cultural landscape. They
are most competent within their system of values, expressed through the spatial
configuration of the landscape. .
In order to understand movement at various scales (from local to global), radius
was used to limit the number of nodes included in the calculation of the average path cost.
In this case, the selection of such nodes was done after the shortest path was calculated for
all other nodes but the source. The result could be potentially different from that obtained if
the selection of nodes was done before the calculation of the shortest path (i.e. by using a
subset of the original graph). The latter scenario is far more restrictive and implies
ignorance of other locations (i.e. nodes)that lay beyond the radius, while the formerallows
for paths that may include some node outside the radius.
5. Promenades
An area of the Yorshire Wolds in north eastern England was used to illustrate model,
Figure II. It is 36km
2
, with a minimum altitude of 66m and a maximum of 210m.
The results of executing themove routine are presented in Figures 12 to 14 which
show the accessibility for three radii considering the topography only. The accessibility was
calculated by subtracting 100 from the average cost path, which in turn was obtained when
Figure 11. The case-study area viewed from thewest
78
Beyond the Map
An ..-,..:T.......... ... cc..-.,: T... t...)
......
.-
... ...
,.
_

@J
-
'0""11
""'n
-(1)- -
'JIll'"
-
Figure 12,Topographic Accessibility at 100m Figure 13,Topographic Accessibility at I,OOOm
Aft""-": ,,-
...
\\IY'
-..
-
ID'"
Figure 14,TopographicAccessibility at6,OOOm
the total path cost image (i.e. output of the move routine) was standardised by the
maximum total path cost. Tbese images, derived for each radius, represent background
valuesthatcanbeusedlatertonormalisesubsequentaccessibility images.
The accessibility for r=IOOmis relatively uninformative, showing only that spurs
and ridgesare lessaccessible than the restof the landscape, Figure 12.Ata middle range
(r=I,0000), the highest accessibility area isconcentrated on aslight depression locatedat
the north eastern comer of the test area, Figure 13. At a global scale (r=6,000m) this
depression loses importance while slacks(dry valleys)become more accessible, especially
themiddleone, andamongthese, thoseslacksthataresurrounded bylowergroundshowa
higherindex,Figure 14.
79
M, Llobera / Understanding movement
Figures15to 17show theresults aftertheintroductionofdouble andmultiplelaterBronze
Age linear ditches. In this case, the incorporation of different types of ditch (e.g. the
multiple Huggate Dykes shown in the north of the area) allowed the possibility of
observing differences in their impact on surrounding locations. Bevan [3] refers to the
important significance that may have been attached to multiple ditch systems, a
significancethat islikelytobeof asymbolicnature,giventheamount ofeffort invested in
themandtheirapparent limited functionality.Thevalueassociated withtheditches wasset
arbitrarilyto 100andcombined withthevaluesof thetopographic cost image byselecting
themaximum value(inthiscasethosecomingfromtheimageoftheditches), .
An..""":1AI_ ........ I")
......
...
.......
. ...
-
.
'...... 1800 80
Figure 15,Accessibility withLinearDitches Figure 16,AccessibilitywithLinearDitches
at 100m at I,OOOm .

....
...

'.DOl
Figure 17,Accessibility withLinear Ditches at6,OOOm
80
Beyond the Map
NI""o-,. "'" A4Q ....,a-. ....


L
L
',IIIUII -
-
-I
-
AlII'"
Figure 18. Accessibility Change at 100m
Figure 19. Accessibility Change at I,OOOm

L
-
'/JJ}'"
Figure 20. Accessibility Change at 6,OOOm
Changes in accessibility are not very noticeable in these images. This is mostly a
consequence of the way in which the linear ditches are combined with the topographic cost
cost inside the model. In the following section, several alternatives will be discussed,
aimed at palliating this shortcoming. In contrast to above, the impact of the linear ditches
on nearby locations is clearly displayed in Figures 18 to 20. These images result from
subtracting the
accessibility image attained for topography from the one produced by the introduction of
the linear ditch system.
As these figures show, the effect of each linear ditch on the landscape is
proportional to its magnitude, that is, the effect due to the presence of multiple ditches is
much more noticeable than that for single ditches. In addition, the impact of linear ditches
would have had on movement within a radius of lkm is strikingly similar to the ones
'{D-
-CD-
-(1;-
81
M. Llobera / Understanding movement
obtained for more global scales (i.e. 6,OOOm). This similarity helps to illustrate a point
mentioned earlier, that the incorporation of monuments, or any type of landscape feature, in
the landscape can affect differently the various scales of movement, local to global, that
become part of a cultural landscape.
6. Discussion: further developments
Given time and resource constraints, the routine has not been executed as often as is
required. Archaeological interpretations are alsoomitted and only comments that relate
directly to the model are offered here.
Undoubtedly the model raises many new questions and directions in which further
research may need to be undertaken. It also points towards the necessity to develop new
strategies that might be used in order to calibrate this and other 'exploratory' models. In
this case, applying the model to a real landscape and obtaining comparative measures
would be a way forward.
Overall, the description of the model helps to illustrate' various points, especially
the benefits of engaging in modelling within a more 'humanistic' context (e.g. precipitating
new ideas, challenging existing concepts, providing solid 'guidelines for research), as well
as the fallacy of considering models as providing 'objective' views of the world. Plenty of
evidence throughout this paper shows just how many choices a researcher faces when
developing a model, here in the form of a new GIS routine. Thus models aimed at
exploring any aspect of humanity, do not, cannot, and should notaspire to provide an
'objective' view of the phenomena they aim to understand. Their value in that they
provide a formal, reproducible description of these phenomena, at times the only A
description that, because of its sensible construction, may be used to corroborate certain
interpretations, but that ultimately may be challenged by another perspective, new
information or a new modelling philosophy.
The results so far, have shown that several limitations exist, some of which may be
amended through minor modifications, while others point towards a possible change in
modellingstrategy.' ..
Minor problems include the inability, in the current version, to incorporate real
barriers, those with an infinite cost. This handicap could be fixed if the mathematical graph
representing the original landscape did not store cost information but accessibility indexes
instead. This would obviously entail changing the shortest-path procedure into one thar
would find maximum accessibility paths rather than minimum cost ones. This .change
would allow assigning, quite easily, a low or zero accessibility index to any path between
two locations representing a barrier (whether physical or mental). The second alteration
required refers to the limited impact that the introduction of the linear ditches seems to
have had on the original topographic accessibility images; this isIikely to be related to the
way in which the accessibility index is computed. Currently, it is based on averaging out.
multiple path costs and it is precisely this averaging that minimises.the impact, in this case,
of the linear ditch. It is necessary, therefore, to experiment with other mathematical
expressions which might be more sensitive.
Major limitations of the current model include the intensity of computation and the
limited scope for interaction. So far, incorporating and modelling the impact of any
landscape feature is quite limited and, ideally, we would want a system that allows us to
observe the effects of incorporating features quickly, although not necessarily real-time. It
seems quite obvious that such a modelling technique would require a complexity beyond
82
Beyond the Map
anordinaryGISapplication andthis suggests that futureworkaimed atexploring thetopic
ofmovement maybechannelled inatleasttwodirections:
I) Bycreating a simulation engine, possibly as an object oriented application. A system
thatwould also incorporate several of theelements discussed byHelbing et al. (22] in
theirworkonpedestrian trails.
2) Byadopting anautonomous agent approach. The necessityfor modelsof this sort has
already been recognised elsewhere, particularly in urbanism and recreational planning
(41], (16]. This model1ing strategy may, ultimately, represent more faithfulIy the
dialectic relationshipexistentbetweenlandscape andpeople.
The way the model is specified here, priority is given to the role of the landscape rather
thantheeffectofpeopleon thesame. Thedialectic relationship betweenthelandscape and
theindividuaUsisaddressedbutinapassive way.Theimpactofpeopleonthelandscape is
integrated into the model by incorporating the monuments they built, and observing how
these affected the accessibility index. By using autonomous-agents, (10], this dialectic
would be integrated in a moredynamic manner although would alsoprecipitate a newset
of difficulties. Finally, the model may be further developed by adopting a probability
inferenceframework(e.g.Bayesian), whereprobabilities are manipulated in thesame way
asrasterlayers.
Independent of whichever direction is followed, it is true to say that all of these
options would utilise GIS as an interface from where data can be extracted and used to
'populate' a procedure, or simulation engine. This procedure may be more or less
complicated and it is up to the user to consider it as an extension to the GIS or as an
externalsystem.
References
[I]Bally, M. PredictingWhereWe Walk, Nature. Vol. 388,1998,pp. 19-20.
[2] Barrell, 1. C. Fragmentsfrom Antiquity - An Archaeology ofSocial Ufe in Britain, 2900-1200 Be.
Oxford: BlackwellPublishers, 1994.
[3] Bevan, B. Bounding the Landscape: Place and Identity during lite Yorkshire Wolds Iron Age. In
Reconstructing Iron Age Societies(eds. Gwill, A andHaselgrove. C.). OxbowMonograph71, 1997, pp.
181-191.
[4] Bourdieu,P. Outline ofa Theory of Practice. Cambridge:CambridgeUniv. Press, 1977.
[5]Bourdieu,P. The Logic ofPractice. Cambridge: Polily Press, 1990.
[6] Bradley,R., CriadoBoado,F and FAbregasValcarge. Rock Art andResearchas LandscapeArchaeology:
APilol Study inGalicia,NorthwestSpain,WorldArchaeology,25(3), 1994, pp. 317-391.
[7] Brody, H. Maps and Dreams: Indians and the British Colombia Frontier. Hannondsworlh: Penguin
Books, 1983.
[8]Bultimer,Aand Seamon,D. The HumanExperienceof Place andSpace. London:Crom Helm, 1980.
[9]Cormen,T. H., Leiserson,C. E. and Rivest, R.L. IntroductiontoAlgorithms.Cambridge:The MIT Press,
1990.
[10] Doran, 1.andGilbert,N. (eds.) SimulatingSocieties.London: UCL Press, 1994.
[II]Evans, 1. R. and Minieka, E. Optimitation Algorithmsfor Networks and Graphs. New York: Dekker,
1992.
[12] Foucault, M.Disciplineand Punish: the Birthofthe Prison. Hannondsworlh: PenguinBooks, 1979.
83
M. LloberaI Understanding movement
[13]Gell, A. How 10 Read a Map: Remarks on the PracticalLogicof Navigation,Man (N.S.), 20, 1985, pp.
271-280. .
[14]Gibson,1.1.The EcologicalApproachto VisUllIPerception. New lersey: LawrenceErlbaum Associates
Inc., 1986.
[15]Giddens,A. The ConstitutionofSociety. Cambridge: Polity Press, 1984.
[16] Gimblell, H. R. B, Durnota, B.and Itami, R. M. Spatially-Explicil Autonomous Agents for Recreation
Use of ComplexWildernessLandscapes,ComplexityInternational3,Amsterdam:lOS Press, 1996.
[I7J Goffman, E. Encounters: Two Studiesin the SoFiology ofInteraction. London: Allen Lanethe Penguin
Press, 1972.
[18] Golledge, R. G. and Stimson, R.I. Spatial Behavior: A Geographic' Perspective. New York:: The
GuildfordPress, 1997.
[19]Gosden,C. SocialBeing and Ti,,!e.Oxford: Blackwell, 1994.
[201Hligerslrand, T. Whal about Peoplein Regional Science? Papers ofthe RegionalScience Association, .
24, 1970, pp. 7-21.
[21] Helbing, D. and Molnar, P. Social Force For Pedestrian Dynamics, PhysicalReview E. 51(5), 1995, pp.
4282-4286.
[221Helbing, D., Keltsch,1.and Molnar,P. Modellingthe Evolutionof Human Trail Systems, Nature. (388),
1997, pp. 47-50.
123)Henderson,L. F. The StatisticsofCrowdfluids,Nature. (229),1971,pp. 381-383.
[241Hillier, B. Spaceisthe Machine. CambridgeUniv. Press, 1996.
[251Hillier, B. and Hanson,1.TheSocialLogic ofSpace. Cambridge:CambridgeUniv. Press, 1984.
[26] Hodder,I. The ArchaeologicalProcess: an imroduction.Oxford: Blackwell, 1999.
[271Ingold, T. The AppropiationofNature - Essays on Human Ecologyand Social Relations. Mancheste;:
ManchesterUniv. Press, 1986.
[28] Kamon, E. Negative and Positive Work in Climbing a Laddermill, Journal ofAppliedPhysiology, 29,
1970, pp. 1-5. .
[29] Kainer, 1.An Evaluationof Chaco AnasaziRoadways. Paperpresentedinthecurrenttechnologyapplied
10archaeologypostersession,61" SAA Annual Meeting.New Orleans,Lousiana, 1996.
[30] Kohler, T. A. and Carr, E. Swarm-basedModelling of Prehistoric SettlementSystems in Southwestern
North America, in ArchaeologicalApplications ofGIS: Proceedings ofCol/0'luium II, UISp,P Xlllth
Congress, Forly.Jtaiy, Sept 1996 (eds. Bieui,A., Cazzella, A.,lohnson,'"and Maclaren,N.), 1997, pp.
10I-I06 (hllp:/lwww.archaeology.usyd.edu.au/resQurcesidocumentslkohlerlkohler_l.hlml/).
[31] Langmuir, E. Mountaincraft and Glasgow: The Scoltish S"';rt Council - The Mounlain
LeaderTrainingBoard, 1997.
[32] Margaria, R. Sulla Fisiologia e Especialmente sui Consumo Energetico della Marcia e della Corsa a
varia Velocitaed Inclinazionedel Terreno, inAliiAccQd Naz:Lincei Memorie, 1938, pp. 299-368.
[33] Merleau-Ponty, M. The PhenomenologyofPerception. London:Roulledge,1989.
(34) Minelli, A. E. OptimumGradienlofMountain Paths, Journal of AppliedPhysiology, 79 (5), 1995:pp.
1698-1703.
[35] Minelli, A. E. Ardigb, L. P. and Saibene, F.Mechanical Detenninantsof GradientWalking, Journalof
Physiology,471,1993,pp.725-735. . . .
[36] Mugerauer, R.and Seamon,D.Dwelling, Placeand Environment: Towards a PhenomenologyofPerson
and World. New York: ColombiaUniv. Press, 1989.
[38] O'Frake, C. Cognitive Maps of Timeand Tide among Medieval Seafarers; Man (N.S.), 20, 1985, pp.
254-270.
[39] Piaget, 1.and Inhelder, B.TheChild'sConception ,;,fSpace. London:RouUedge and Kegan Paul,
I
84 .85
r
Beyondthe Map
Beyondthe Map
)
[40] Pred, A. Place, Practice and Structure: Social and Spatial Transformation ofSouthern Sweden,
Cambridge:PolityPress, 1986.
I
[41] Schelhom, T., O'Sullivan, D., Haklay, M. and Thurstain-Goodwin, M., STREETS: an
pedestrian model, paper presented at the Computers in Urban Planning and Urban Management
conference, Venice, 8-10September, 1999.
[42] Seamon, D.AGeographyofthe Lifeworld: Movement, Rest andEncounter.London: Crom Helm, 1979.
[43J Thomas, 1. The Hermeneutics of Megalililic Space, in Inserpretative (ed. Tilley C.)
ExplorationsinAnthropology. Oxford: Berg, 1993,pp.73-98.
[44]Tilley, C. Phenomenologyof Landscape.Oxford: Berg, 1994.
[45]Thrift, N. Time and Theory in Human Geography. Part II, in Progress in Human Geography. London:
Edward Arnold, 1977,pp.413-457.
[461Trombold, C. D. RoodNetworks and Settlement Hierarchies in the New World. Cambridge:
CambridgeUniversity Press,1991.
[47] Widlok, T. Orientation inthe Wild: theSharedCognition of HAIIIOMBushpeople,Journalof the Royal
AnthropologicalInstitute. incorporatingMan (N.S.), 3,1997,pp.317-332.
G.Lock(Ed.)
lOSPress, 2000
Topographicandculturalinfluences
onwalking theRidgeway
inlaterprehistorictimes
TylerBellandGaryLock
Institute ofArchaeology
University of Oxford
Abstract: This paper explores the origins and route of the prehistoric Ridgeway in
Oxfordshire, England. and'its spatial and temporal relationships with aseries'ofIron
Age hillfortslocated on and close to it. Byconsideringmovementanddevelopinga
newapproach tocost surface and leastcostpath analyses. we suggest anearly origin
for the trackway based primarily on topographic considerations. Interpreting the
leastcost paths suggests thatwhen firstconstructed, the hillforts appear tohavebeen
located on the existing Ridgeway to establish a direct relationship with travellers
along the track. Developing a visibility index based on moving along the Ridgeway
suggests thatthese locations werenotchosen tomaximise local visibility.
1. Why the Ridgewayandmovement?
Theworkof Marcos Llobera hasbeen important inraising issues concerning thetheory of .
movementandmethodologies for modelling movement through alandscape (thisvolume;
[I]). While this paper picks up on some of the points hemakes, its aims are much more
specificbybeingrooted intheinterestsof afieldwork project. Rather thanconsidering the
general accessibility of an area as being a determinantofmovement, here we assess the
influenceof asingletrackway anditsrelationship withaseries ofprehistoric sitessituated
alongits route. This isthe Ridgeway, along distance track that extends through much of
centralsouthernEngland acrossthechalkdownlands, While there isnodirectevidence for
theRidgeway originating in prehistoric times, it is generally claimed tohaveexisted then
becauseof the proximity ofprehistoric monuments toit.These range from Neolithiclong
barrows to Iron Age hillforts and include impressive cemeteries of Bronze Age round
barrowsthatappeartobealignedalongitsroute [2].Thispaperconsiders theoriginsofthe
Ridgeway by examining the relationship between movement and the.topograpliy ofthe
studyarea,andthendiscusses thelocationofaseriesofhillforts alongitsrouteinterms of
movementbetweenthem.
Many recent writers have emphasised the socialisation of landscape and the
importanceof landscape as a cultural construct (see Wheatley and Gillings, Llobera, this
volumeforexample).Whileweareinbroadsympathy withthesearguments wewould like
to shift the focus more towards the importance of the historical. By historical wedo not
meanformal written history, but a personal and group understanding of the past and its
influence on the practices of both individuals and the social group.'Whether this
understanding isconscious or subconscious it is woven into the everyday praxisof living
86
Beyond the Map
so that moments and events are contingent on the past. The present and the future emerge
from the past through social practice and through negotiation with it at the personal and
group level. In very simplistic terms, we either continue doing things as 'they have always
been done' or we decide to do something different and new, even if only slightly different.
Either way, practice is defined in relation to the past and long term social change is a result
of accumulated decision making.
Of course the past has depth, and something belonging to just a few years ago can
have a different resonance in the present to something that is very old. Concerning the
landscape, places can have their own biographies which give them meaning and
significance through the past people and events associated with them. The older the place,
the richer the potential for biography. For non-literate peoples, such biographical locations
within a landscape can act as mnemonics for tribal histories, whether genealogical or
mythical or both, acted out through storytelling and ritual to reproduce social and personal
relations [3]; [4]. This is history written through marks in the landscape and given meaning
through social memory and connections with the physicality of the present. While it is
dangerous to stretch cross-cultural analogies too far, Basso [5] has shown how
relationships with places are not only used by the Western Apache to give meaning to
history but also to enforce a moral code and a way of acting. Through this the landscape
becomes central in ensuring social continuity and, ultimately, the survival of the group.
Equally, movement is socially embedded [6]. Movement is an integral part of being
in a place, it involves encountering and interpreting information as a sequential spatial
narrative. As Chapman has shown (this volume), movement links places relationally so
that approaching, arriving and leaving a locale are important elements of understanding it.
Again this is in relation to the past so that the narrative of movement includes a level of
expectation based on familiarity, or conversely, a very different experience based on
unfamiliarity. Trackways also have biographies based on people, events and places
associated with them, endowing the route with cultural meaning and significance that
elevates it above the purely functional requirement of getting from A to B.
These ideas are central to the work of the Hillforts ofthe Ridgeway Project which is
investigating the changing social landscape through the later prehistoric and Romano-
British periods. The study area encompasses a 15 x 25 km area of the Ridgeway in south-
western Oxfordshire, England, which includes the hillforts of Liddington Castle, Hardwell
Camp, Uffington Castle, Rams Hill and Segsbury Camp, and although it may not be
associated with the Ridgeway and the other hillforts directly, Alfred's Castle (Figure 1).
The intention in this paper is to investigate movement along this stretch of the Ridgeway,
specifically exploring the possible influences of topography and the idea that social
memory may be expressed through the location and use of the hillforts.
2. Quantifying movement using cost surfaces
To date, GIS and cost surface generation have been the basic tools used in attempting to
model movement across a landscape, although there are inherent difficulties. A cost
surface is a computer-generated model of the landscape in which each part of the surface
(i.e. each cell within the raster structure) is assigned a 'cost', this is effectively the price one
must 'pay' to reach that point from a predetermined point called the 'target'. The cost is
calculated by an algorithm that traditionally incorporates both the distance of that point
from the target, as well as an additional relative cost based on a particular quantifiable
element of the landscape. This second cost is user-defined, and is usually derived from a
87
T. Bel{ and G. Lock / Walking the Ridgeway
.41"":'
rrft

Figure I. 'Thestudy area in southern Oxfordshire, England, showing the Ridgeway and Iron
Age hillforts (0 Crown Copyright Ordnance Survey. An EDINA DigimapiTISC supplied
service)
classification system in which each cell of the. landscape model is assigned a cost value.
that is relative to a base value.
One of the first examples of using a cost surface in archaeology still stands as the
'classic' approach, the Hvar study, in which Gaffney and Stancic developed a cost model
based on the angle of slope [7]. From this, general accessibility of the landscape was
established by calculating walking times from specific sites. Similar cost surfaces have
been used in other ways in archaeology such as the gravity/topography model used to.
investigate colluvium movement downslope and its impact on the archaeological landscape
[8]. Rather than general landscape accessibility,' specific least cost paths between two'
points can be established from the cost surface which can then be combined' into
cumulative least cost paths and least cost networks [9]. Such models need not be limited to .
the single variable of slope, as Madry and Rakos [10] have shown by generating paths
between hillforts based on maintaining the highest possible elevation and using the lowest
angles of slope while remaining in view of the maximum number of hillforts.
As with GIS-based visibility studies (Wheatley and Gillings, this volume),
modelling movement introduces fundamental technical concerns involving the use of
DEMs and their accuracy [11]. This is obviously not just the concern of archaeology and
research of relevance to archaeological applications is progressing elsewhere, for example
on the use of fuzzy optimal paths based on the spatial error and uncertainty of DEM data,
and using animation for displaying that uncertainty [12]. Of central importance to cost
surface applications are the algorithms used .to generate the cost surface itself and to
determine the route of movement that results in the optimal path, these have been recently
reviewed by van Leusen [13].
As Llobera (this volume) has indicated, any methodology based entirely on a -slope-
dependent cost gives primacy to characteristics of topography at the expense of any
cultural attractors or detractors that may affect movement across the landscape. Even so it
must be acknowledged that topography is influential and this is also our starting point. This
reflects the influence of Tilley's so-called bones of a landscape [6] and echoes Muir's [14] .

. 89
88
Beyond the Map
statement that the contradiction inherent within the structure of the landscape is that it is at
the same time both essential and always incomplete. Topography is a fundamental
component of the mechanics of movement and yet is by no means the complete
explanation of it.
Our analysis begins by considering the topographic influences on the location of the
Ridgeway. We attempt to move beyond the traditional approach to optimal paths by
considering two important characteristics of slopes that must be understood before any
attempts can be made to model human movement across them. First, it is misleading to
assume that the cost of climbing a slope is directly proportional to the degree of slope, thus
surmounting a 45 slope is not simply 45 times as difficult as moving on the level, a 0
slope. Taking this to its logical conclusion would suggest that climbing a vertical slope of
90 is ninety times as difficult as walking on the level, an absurd simplification which
would not stand up to scrutiny. Secondly, it is equally naive to assume that movement
across a slope will always be in a uniform direction as movement across the landscape
entails crossing hillsides at different angles, not simply 'up slope' or 'down slope' as
assumed in the classic model. It follows from this that varied direction of movement in
relation to the slope will incur different costs. These two ideas are presented in more detail
in the next two sections and then applied to the Ridgeway.
2.1 Different slopes - different costs
The actual relationship between slope and the relative effort required is slightly more
complex than a simple linear scale, and can be expressed mathematically. Figure 2 shows
two slopes in profile, their angles represented by 0, and O:z, and their heights represented by
y/ and Y2 respectively. X is the horizontal distance the traveller must cross to ascend the
slope.
Before integrating this comparison into a cost surface one must first establish the
relative cost of ascending these slopes. Discovering the change in potential energy after
ascending the slope is comparatively straightforward: mass *gravity *height ascended,
and the ratio between the two changes in potential energy is equal to rngy, : mgy2. Because
a relative cost surface assumes the mass of the traveller to be the same, and gravity is a
constant, we can reduce the ratio to Y/ : Y2, the ratio of their heights. The problem here is
that slope quantification in a GIS does not allow for absolute heights within a pixel, but
only the angles of slope. However, because y/=xtan(O/) and Y2=xtan(Oz), the same ratio
can be expressed as xtan(O/) : xtan(O:z). Furthermore, because x (which is the distance
travelled, or the size of one pixel) remains the same, the ratio can be further simplified to
tanO/ : tanO:z. The relative cost of ascending slopes can, therefore, be correctly expressed
as the ratio of the tangents of the slope angles, tanil, : tan02, rather than a ratio using the
angle of slope itself, Figure 3.
It is necessary to use a base cost tan(l) slope to avoid the division by zero which
would result if we were to use 0 (a level surface). The final step in generating a coverage
from which all costs will be calculated is to take the tangents of all slopes within the
landscape, and then dividing that coverage by 0.01745, the tangent of 1. The resulting
cost surface reflects a more accurate, non-linear progression which better describes the
comparative difficulty in traversing steep terrain. As a result, the relative cost of
surmounting a 60 slope, for example, is 100 units, rather than the 60 units of the
conventional linear scale.
T. Belt and G. Lock / Walking the Ridgeway
y.
y,
Figure 2. Two slopes in profile
CoqIeralIveEnergyfor CaIcuI8Iing MagnItude
TAN(SIopeVTAN(1)
700 , I
600
-_.---
11
8
400

1300
I
2CJ) .... - - -
100 .. - - - - -
01 , ii' , ;:::=, , , I I , I I I I
1
AngIeolS,,-
Figure 3. The non-linear relationship of cost, based on the tangent of slope
2.2 Direction of movement - Anisotropic Cost Surfaces
Another important feature of topography, but one that is often neglected in GIS modelling,
is that a slope does not exert a force uniformly in alldirections, but exerts 100% of its
force in one direction only, namely 'downhill'. While the slope will offer full resistance to a
traveller moving directly uphill, it will offer comparatively little impedance against a
traveller moving perpendicular to the slope. This physical principle underlies 'the 's-curves'
of mountain roads, and the sheep trails running across hillsides. Applied within a GIS, the.
relative friction of a pixel based on slope is, therefore, wholly dependent upon the direction
of travel across it. This is in contrast to the traditional slope-based cost surface which uses
uniform slope values so that there is little analytical difference between modelling
movement down a 20 slope and up the same. While traditional slope-based landscape
analyses are valid in so much as they ascribe low-costs to level land and high-costs to all
slopes, they do not, by their very nature, vary the cost of a pixel based on the direction of
travel across it.
A solution to this problem is the application of an anisotropic cost algorithm, one
which incorporates both the magnitude and direction of a force to generate a cost surface
T
90
91
Beyond the Map
T. Bell and G. Lock i Walking the Ridgeway
Upslope
"5- uphill:50'4 of force against .er
...u... 1
/
Trayerse: noforceaglin,' .. 11k.r
Trev."': notorc. Iglinst .I.er
Directionof
.. ..
Tr.wel Across Slop.
....._...K j
Downslope
Figure 4. Anisotropic Cost Surfaces: a diagram demonstrating the algorithm's
application of force vectors upon movement across a slope detennined by the
walker's direction of travel
dependent on the direction of travel across various vectors from point A to B, Figure 4.
Applied to a landscape, the model employs slope data (magnitude) and aspect data
(direction in which the slope faces) to produce a directionally-based cost surface. This
applies the full force against the traveller when travelling uphill, with the traveller when
walking downhill, and in proportional increments relative to the direction of travel when
walking at acute angles to the direction of force.
In modelling human movement this algorithm will still present problems when
applied to steep terrain, in part because it 'assumes' that moving down a 50 slope is very
easy (low-cost). In reality, even slopes above 25 can be nearly as tiring to descend as
ascend, and the descent of a 50 slope can be more of a tumble than a controlled
perambulation. In an attempt to solve this difficulty, slope data were reclassified so that
extremely steep slopes were given proportionally high costs reflecting the difficulty in
traversing them from any direction. A further difficulty is the assumption that a direct
traverse of the hillside - moving perpendicular to the force - would incur simply the base
cost, no more than if the walker were moving on the level. This apparent discrepancy can
be accommodated by using a user-defined function in the algorithm that incorporates
'corrections' for traversing a hillside at angles near-perpendicular to the slope [15].
It is necessary to point out the inherent cultural bias within these discussions and
that different groups of people can perceive and react to topography and slopes in different
ways. The slopes of the Ridgeway and the chalk escarpment would have been basically the
same 2,500 years ago as they are now although the reaction to them could have been very
different. These days hardly anybody walks up or down the escarpment slope and only
'serious walkers' venture along the Ridgeway for more than a few hundred metres from the
car parks. In later prehistoric times walking considerable distances would have been part of
the regular practice of life, especially if it involved the procurement of essentials such as
water from the spring line at the bottom of the scarp slope. Accepting that the
'interpretation' of topography and slopes in terms of the physical effort required to move
across them is culturally situated, it is clear that the arguments presented here, including
definitions of 'steepness', are based on our own interpretation and are not intended to be
absolute.
2.3 Up and down slopes
The raster surface model was generated at 10m pixel resolution from 5m Ordnance Survey
digital contours available to UK Higher Education Institutions through the Digimap
Projed. Slope and aspect values were derived directly from this surface using Idrisi.
Our primary investigations are focussed just on topography, to investigate how the
bones of the landscape influence movement The resulting least cost path that can then be
compared with the route of the modem Ridgeway to establish the 'topographic component'
within the current route. It is important to appreciate the character of the chalk downlands
in this area which have a strong north-south. 'grain' making movement in those directions
comparatively easy along the dry valleys. The interesting aspect of the Ridgeway is that it
goes against the topographic grain, running from west to east, hugging the edge of the
chalk escarpment and providing spectacular views over the much lower Vale along its
length to the north.
The initial least cost path was generatedbetween the hillforts of Barbury Castle an.!!
Segsbury Camp, Figure 5, (the former is actually to the west of the case-study area to
minimise 'edge effects' and to enable the inclusion of Liddington Camp in the analysis).
This immediately highlights problems inherent within the technique. The algorithm
conceives of downward movement, especially on steep slopes, as both preferable and
entailing less cost. Thus, when starting from Barbury Castle, the resultant pathway to
Segsbury would always produce a path running down slope into the Vale, and then
continuing on level ground until it was forced to ascend the scarp to reach Segsbury.
Clearly this path is both unrealistic and extremely costly.
i _ ....... _. .._',-_. .- .._'" _...... ". 1
Figure 5. The initial least cost pathway belween BarburyCastle andScgsbury Camp,
andtheeffect of 'falling' off theridge (C> Crown Copyright Ordnance Survey. An
EDINADigimapiRSC supplied service)
I http://edina.ac.ukldigimapl
T
92 93
Beyond the Map T. Bell aNi G. Lock I Wa/Jcing the Ridgeway .
It can be seen that the entire course of this pathway is determined at its outset by the steep
slope directly to the north of Barbury Castle. The algorithm seizes on this steep,
presumably low cost area without understanding that it will eventually have to incur
further cost in surmounting the downs again to arrive at Segsbury. The problem is that the
slope-based algorithm does not take into account the fact that both the starting and
finishing points are on the same ridge so that the total cost of the journey will be less if
relatively level ground is kept to, and that going downhill is not always beneficial in the
long term.
This initial solution highlights the current inadequacies of existing cost surface and
pathway-generating algorithms, but also serves to remind us that the combination of
observation, interpretation, and a priori knowledge that facilitate human movement from
one place to another are not so easily reproduced in mathematical formulae. An important
factor is the a priori intention of destination that is inherent within human movement and
becomes scribed on the landscape by the persistence of pathways. Both individual and
social memory would incorporate the knowledge that the Ridgeway connects various
destinations along the ridge, and this would be utilised within the intentionality of
movement.
In an attempt to correct the tendency to 'falloff' the Ridgeway a topographical bias
was introduced into the algorithm's interpretation of slope and aspect. By proportionally
inverting the north - south aspect data we can generate a topography in which the descent
into the Vale is no longer a descent but an ascent; the resultant pathway will adhere to the
'ridge', only surmounting or descending slopes when they exist in the east - west
directions. Figure 6 shows the least cost pathway generated using the east-west directional
bias algorithm. In effect this is a simulated walk along the edge of the chalk escarpment
based on the 'easiest' route but assuming that the start and end points are known and are
located on the Ridgeway.


Figure 6. The least cost pathway generatedafter introducing a topographical bias. The
current Ridgeway is the chick line (Copyright Ordnance Survey. An EDINA
DigimapinSC supplied service)
The relationships between the modem Ridgeway, the hillforts and the generated Ridgeway
deserve comment. Firstly it can be seen that the generated path fits remarkably well with
the existing one with interesting deviations occurring at the hillforts. Looking at these in
detail and working from the west, Figure 7 shows that the generated path runs to the south
of the modem Ridgeway and through the hillfort of Liddington, before returning to the
present route. A similar deviation happens at Hardwell Camp, Figure 8, and Rams Hill,
Figure 9, so that the pathway makes contact with the hillforts. Uffington Castle is different,
Figure 10, and shows the generated pathway running to the north of the hillfort, between it
and the Uffington White Horse chalk figure, whereas the modem Ridgeway exists to the
south of the hillfort.
One possible interpretation of this, and the one to be explored further here, is that
the Ridgeway originated before the hillforts were constructed and was based on a least
effort ridge route across this area of downland. This could date back to earlier prehistoric
times, probably pre-Neolithic as monuments of that period are associated with it:
Environmental evidence suggests that even before clearance of the chalk downs began in
the Neolithic, these areas were drier and less densely wooded than the lower clay Vale
[16]. Movement across the downs would have been easier than in the Vale and the
Ridgeway could have originated as an animal trackway, particularly for herd animals, in
Mesolithic or even later Palaeolithic times that became utilised by humans. It follows, then,
that when the hillforts were first constructed in the 8
ch
_7
ch
centuries cal BC they were
located on a well established long distance trackway running across the chalk downlands in
a landscape already substantially cleared of woodland when compared to the Vale. It is
worth mentioning here that until very recent times with the introduction of stretches of
surfacing and consolidation, the Ridgeway was not a 'fixed' trackway but wandered within
a general 'corridor of movement'. This is still the case for much of it, particularly in the
Figure 7. The area around Liddington Castle showing the generated Ridgeway running
through the hiUfOl1 and the modemU1ICk to the notth (Cl Crown Copyright Ordnance
Survey. An EDINA DigimapinSC supplied service)
T 95 .
94
Beyond the Map T. Bell and G. Lock'/ Walking the Ridgeway
Figure 8. Tbe area around Hardwell Camp showing the generated Ridgeway running to
the hillfort and the modem Ridgeway in the south eastern cornet of the figure ( Crown
Copyright Ordnance Survey. An EDINA DigimapinSC supplied service)
Figure 9. The area around Rams Hill showing the generated Ridgeway leaving the
modem Ridgeway to go through the hillfort ( Crown Copyright Ordnance Survey. An
EDINA DigimapinSC supplied service)
Figure 10. Tbe area around Uffington Castle showing the generated Ridgeway running
to the north of the hillfort and the modem Ridgeway to the south. The linear ditch exists
as a fence line running south from hillfort (0 Crown Copyright Ordnance Survey. An
EDINA Digimal'inSC supplied service)
winter when conditions get very muddy and new routes are being constantly created. The
implication of this is that both the modern and generated Ridgeway shown here are
generalised routes and not exact constants. .
The generated pathway also provokes interesting questions concerning the
entrances of the hillforts. Liddington Castle conforms to the common pattern for Wessex
th
hillforts of having two opposing entrances in its first phase, 8
th
_7 century cal BC, with one
th
subsequently being blocked during the second phase rampart re-modelling in the Sib4
centuries. Liddington's opposing entrances are oriented east-west with the western one
becoming blocked [17], although the generated least cost pathway goes through both,
Figure 7. As with several of the other hiliforts this suggests that Liddington was sited on
the pre-existing Ridgeway so that the track passed through its entrances, a situation
preserved until the present day at Barbury Castle [18). The blocking of a hillfort's entrance
would obviously have repercussions for movement into and around the enclosed space and
would end the control of movement along the track that the original siting established, It
could be that with the blocking of the western entrance at Liddington the Ridgeway moved
downslope to the north to establish its current route. The change in entrance configuration .
could also reflect a change in function for the hillfort.no longer directly based on contact
with people moving along the Ridgeway.
Hardwell Camp is the least well understood of all the hillforts in the study area due
to a complete lack of any archaeological work having been done there. It has one very
obvious entrance through the southern ramparts looking towards the current Ridgeway and
it is relevant to note that the generated pathway deviates to access it, Figure 8. .
Rams Hill has a complex temporal Sequence and consists of three different
enclosures dating to the Late Bronze Age, Iron Age and Romano-British periods [19]. The
Iron Age hillfort, although completely ploughed-out, appears to have opposing east-west
96
Beyond the Map
the east, Segsbury Camp is the finishing point for the generated path so by definition it
runs to it althoughit is worth noting that a blockedentranceexistsin the western ramparts
[20].
The final site inthe group, UffingtonCastle, standsout as beingdifferentbecauseit
is part of acomplexof prehistoric monumentson WhiteHorseHill locatedon the highest
hilltop in the area and incorporatingan unusual and spectaculararea of natural landscape.
Several parts ofthis complex have been excavated recently so the sequence is relativelr
well understood[16]. Like Liddington,the first phase hillfortwas constructedin the 8"'-7
centuriescal BC and had opposingentrances to the east and west. A linearditchran from
the southern ramparts southwards acrossthe downswhich wouldhaveacted as a barrierto
east-west movementsouth of the hill fort. This suggests that the Ridgewayran throughthe
hillfort itself. It is worth noting that several linear ditches also exist around Liddington
Castle although none have been excavated and their direct relationships with the hillfort
are not known, they could have been for channelling movement as suggested for
Uffington. In the 4'"centurycal BC the easternentranceat Uffington was blockedand the
ramparts re-designed with the hillfort appearing to be abandoned shortly afterwards and
not used again for seven hundredyears (at least inany way that left a materialrecord). The
linear ditch remained SUbstantially open for this same period finally being intentionally
filled and levelled in the 3
n1
centuryAD at the same time as the hillfort was re-used.This
suggests that for this period, 4'" century BC to 3
n1
century AD, the Ridgeway ran to the
north ofthe hillfortalongthe route taken bythe least cost path.
3
n1/4'"
In the centuries AD White Horse Hill resumed importance as a
ritual/ceremonial centre with various activities including the re-use of prehistoric
monuments for burial. Part of this activity was the opening-up of the hillfort by breaking
gaps throughthe rampartstoprovideaccess and the filling inofthe linearditchto establish
a new route for the Ridgeway to the southof the hillfort, the route still used today. The
continuingimportanceof WhiteHorseHill was almostcertainlybasedon the existenceof
the chalk-cutUffingtonWhiteHorse which, according to recentOSLdating, was possibly
1,000years old and alreadyanancientmonumentby Romano-Britishtimes.
Another remarkable aspect ofUffington Castle, and indeed many other Iron Age
hillforts, is its visibility. It is situated on the highest point of the northern downs and is
visiblefrom avery wide area around,and itistoquestionsof visibilitythat we now tum.
3. Qualifyingmovement- seeiogthehillforts
The argumentso far has suggestedthatthe Ridgewaytrack pre-datesthe hillforts and was
probably established as an animal route seemingly based on a least cost principle for
moving east-west across the chalk downs. When the hillforts were first constructed they
appear to have been locatedon the Ridgeway to ensureaccess into and through them for
peopletravellingalongthe track. Thereasons forthis are not knownand couldbe variable.
The hillfort occupants could have exerted 'control' over use of the Ridgeway by
monitoring movementthroughthe enclosures. If the Ridgewaywas used for long distance
movementthe hillforts could have acted as stop-overs providing shelter. The building of
these massive structures in the 8"'n'" centuries cal BC and the organisation and effort
involved,suggeststhatduringatime of social changehillfortscouldhave been placesthat
symbolised social cohesion and identity within an otherwise unstable social landscape.
This wouldgive themimportanceand meaningabovethe purelypracticaland partof their
functioncouldhave beentoprovidespiritualrejuvenation for travellers as well as physical
sustenance. Recentexcavations at UffingtonCastle [16] have certainlysuggestedthat this
was not asimplesettlementbut probablyfunctioned asaceremonialcentreas well.
1)7
T. Bell and G.Lock / Walking the Ridgeway
sustenance. Recentexcavations at Uffington Castle [16} have certainlysuggested that this
was not asimplesettlementbutprobablyfunctioned as aceremonialcentreas well.
The actual location of the hillforts on the Ridgeway would have determined their
visual impact on travellers along the trackway and in this section we explore this idea
further. Wheatleyand Gillings(this volume) have shownthat visibilityis a complex issue
and thatthe simpleapplication of viewshed analysis is a naive approach. One of the most
basic complicating factors is that of distance of view and the effects of distance on the
perception of what is in view. This is relevant here as we differentiate between viewing
what is immediateto the Ridgewaywhen travellingalongitand long distanceviews across
the downsto the south and the Vale tothe north.
The visual importance of Uffington mentioned above is based on'long distance
visibility and checking this with viewshed analysis is unnecessary and would be nothing
otherthan establishingthe obvious. Buildingon the argumentsabove for the hillfortsbeing
located so as to make them an integral partof the experienceof travellingthe Ridgeway,
here we are interested in establishing the visibility of each hillfort from different points
along the track. Were they locatedto be highlyvisibleto travellersalongthe Ridgewayor
were the visibilityconsiderationsbasedontheirwiderregionalimpact,or both? .
Atotal of sixty3km viewshedswere generated,one every250malong the lengthof
the Ridgeway from Barbury Castle to Segsbury Camp, Figure 11. The 3km maximum
distanceis basedon empiricalexperienceof walking the Ridgewayand comparingwhatis
visible alongitineitherdirectionunderdifferentatmosphericconditionswith known
Figure II.Clockwisefromtopleft: fo.or consecutive viewshodsat 2SOmintervalsalong
the Ridgeway, moving east towards Segsbury (from a total of sixty), 3km maximum
distance based on a 1.8m tall observer (0Crown Copyright Ordnance Survey. An
EDINADigimap/JISCsupplied service)
98
Beyond the Map
distances read from a map. While no single value will satisfy the complexities of reality,
3 Icm is a reasonable average based on our experience. The other important consideration is
past vegetation cover and, as stated earlier, at the time of hillfort construction the
environmental evidence suggests that large areas of the downs would have been cleared
grassland with fields, and an unknown proportion of managed areas of woodland.
To establish a 'visibility index' for walking the Ridgeway the sixty individual
viewshed binary coverages (in or out of view) were combined to produce a single
cumulative viewshed coverage. Based on the 3km maximum distance and the 250m
spacing between viewsheds, each point (pixel) along the Ridgeway can accumulate a
maximum value of 25 on the visibility index. This is the number of viewshed origins
located within 31cm in each direction from the point.
Figure 12 shows the combined coverages as the visibility index for the Ridgeway
corridor and the locations of the hillforts. It can be seen that they are not convincingly
positioned to maximise visual impact for travellers, in fact the individual hillforts have the
following visibility index scores:
Liddington Castle 8
Hardwell Camp 0
Uffington Castle 7
Rams Hill 5
Because the resolution of the OEM is lOrn and the hillforts are several hundred metres in
size the calculation of these figures was not straightforward, and represent the highest
value for any point of the hillfort. Even so, they are surprisingly low and reinforce the
claim that the hillforts were not positioned to be visible from many points along the
Ridgeway. Even Liddington Camp and Uffington Castle which display remarkable long
distance visibility from large areas to the north and south, command limited near-distance
views from the track itself.
>T"'-."-;f.\
r. ./c%
,- (Y. // r' .: f Y ...
'1:1 __ I) ,;/ ...
!loi ....ji' -' '''.i( " '"' (.!' "'l.

4G-
v v
/ ... -.?j 'it; '; L> fi ('"
}
>t&;., _r
.. r/...
Figure 12. Thecombined viewshed coverages shown as a visibility
index for the Ridgeway,with the location of the hillforts. NOIe that
darker colours equal a higher index value.
99.
T. Sell and G. Lock I WalJcjng rheRidgeway
4. Conclusions
From the outset, this work was intended to address wider questions of interpretation based
on theon-going fieldwork of the Hillforts a/the Ridgeway Project. In general terms it was
intended to explore the relationship between the Ridgeway and the hillforts and it has
generated some interesting results. It is suggested that the Ridgeway is considerably older
than the hillforts and its route is based on least cost principles. The hiIlforts seem to have
been located on the Ridgeway to establish a close relationship with the practice
of travelling along it, although not to maximise local visibility. Long distance visibility is a
different concern and certainly for Uffington Castle this was an important factor
contributing to its positioning.' . .
One of the other outcomes is the bringing into focus the important relationship
between methodology and application. It is only through the medium of application that we
were confronted with the shortcomings of existing methodology. By thinking through the
practicalities of travelling the Ridgeway the problems inherent within least cost procedures
became apparent. This triggered an iterative procedure that linked archaeological
interpretation and the logic of algorithms, eventually resulting in interpretations that were
intrinsically acceptable and methods that are validated by the interpretations they produce.
This is obviously not an objective process, and should not be. The whole approachjs
situated within the school of exploratory analysis and modelling with the aim of
stimulating ideas rather than providing answers.
Every analysis is started within a framework of prior knowledge and intentions,
some may call these biases although that carries rather negative connotations, and we have
tried to be explicit about ours. An underlying interest has been in the idea of prehistoric
history being structured within the landscape and the influence ofpast events and places.
The important point about establishing pathways at both regional and local Scales is that
they are contingent on what went before. Landscape features are part of the social fabric
that current practices are embedded within, stretching back in time and incorporating
history through stories of people and events. Movement cannot be considered as something
situated within an atemporal landscape but must be part of an attempt to understand the
contemporary historicity of that landscape. These 'big' ideas may seem remote from the
details of least cost algorithms but it is only by linking the two that computer-based
archaeology retains a relevance to today's research agenda. .
Acknowledgement
We would like to thank Thorpe, Sarah Ruth and Paul Morris for their helpful
comments on constructing a mathematically modelled cost surface - any errors or
misunderstandings in its application are, of course, entirely ours.
References
[l] M. Uobera, Landscapes of Experiences inStone: notes on a humanisricuse of a Geographic Information
System to study ancien/landscapes. Unpld. PhD Thesis. Oxford, 1999.
[2] Woodward, PJ. The Sourh Dorset Ridgeway. Survey and ucaWliions 1977-<14. Dorset Natural History
andArchaeological Society Monograph 8, 1991.
[3] Gosden, C. and Lock, G. Prehistoric Histories. World Arr:haeology, 30 (I), pp. 2-12, 1998.
100
. I
Beyondthe Map
(4) Barrell, J.e. The mythical landscapes of the British Iron Age, in Ashmore, W. and Knapp, A.B. (eds)
Archaeologiesof Landscapes. ContemporaryPerspectives.Oxford: Blackwells,pp. 253-65, 1999.
(5) Basso, K.Wisdom Sits in Places. LandscapeandLanguage among the Western Apache. Albuquerque:
University of New MexicoPress, 1996.
(6)Tilley, C.A Phenomenologyof Landscape. Places, pathsandmoruunents. Oxford: Berg, 1994.
17] Gaffney, V. and StanCic, Z. GIS approaches to regional analysis: A case study ofthe island ofHvar,
Znanstveni Filozofskefakultete, UniversityofLjubijana,Yugoslavia,(reprinted 1996),1991.
(8) Vermeulen, F.,DeDapper,D. and Brackman, P. A GIS-based geo-archaeological approach to survey in a
centralAnatolian landscape, inJ.Peterson(ed) Theuse of GeographicInfonnationSystemsin the study of
ancientlandscapesandfeatures related to ancientland use, European CommissionCOST ACTION 02,
1998,pp.63-80.
(9) Bell, T. and Wilson, A. Tracking the Samnites: landscape and communications routes in the Sangro
Valley, Italy (in prep).
[10] Madry, S.L.Rand Rakos,L Line-or-siteand Cost Surface techniques for regional research in the Arroux
Valley,inRD.G.Maschner (ed) New Methods, Old Problems. Geographic InformalionSystemsin Modem
Archaeological Research, Southern Illinois University at Oubondale: Center for Archaeological
investigations Occasional Monograph No.23, 1996, pp. 104-126.
[11) Hageman,J.B. and Bennett, D.A. Construction ofDigital Elevation Models forarchaeological applications,
in K.L. Westcott and RJ. Brandon (eds), Practical Applicationsof GIS for Arr:haeologists. A predictive
modelingtoolkit,London:Routledge, 2000, pp.113-128. .
[12] Ehlschlaegtt, CR, Shortridge, A.M. and Goodchild, M.F. Visualizing spatial data uncertainty using
animation. Computersin GeoSciences, 1997,Vol.23,No.4,pp.387-395.
(13) M. vanLeusen,Viewshed and Cost Surface AnalysisusingGIS (Cartographic modelling in cell-based GIS
D), in lA. Barcelo, 1.Brizand A. Vila(eds) New Techniques for Old Tunes. CAA98. ComputerAppli<:alions
andQuantilaliveMethodsin Archaeology.Oxford:BARInternationalSeries757. 1999,pp.215-223.
(14)Muir,R.Approaches10 landscape. London:MacMillanPress. 1999.
(15) Eastman, J.R. ldrisiforWindowsUser'sGuide, 1997,Clark University.
(16) Cromarty, A.M., Gosden, C; Lock, G., Miles, D. and Palmer, S. Uffington White Horse and its
Landscape: investigations at While Horse Hill, Uffington, 1989-95and Tower Hill Ashbury, 1993-4,
Osfordshire.Thames Valley MonographSeries, OAU and OUCA, Oxford, inprep.
(17) Hirst, S. and P. Rahtz. Liddington Castle and the battle of Badon: excavations and research 1976.
Archaeologi<:alJoumal153,1996,pp. 1-59.
(18) Gingell, e. The Marlborough Downs: A Later Bronze Age Landscape and its Origins. Wiltshire
Archaeologicaland Natural History Society Monograph I, 1992.
(19) Bradley, R. and A. Ellison. RamsHill: a Bronze Agedefendedenclosure and irs landscape. Oxford:
British ArchaeologicalReports 19, 1975.
(20) Lock,G.and Gosden,e.The HillfortsoftheRidgewayProject:excavations at Segsbury Camp1996,South
MidiondsArr:haeolagy,27,1997,pp.69-77.

101
Beyond the Map
G. Lock(Ed.)
lOS Press, 2000
Quantifying the non-quantifiable:
studying hunter-gatherer landscapes
JoelBoaz
InstituteforArchaeology
AnHistory andConservationStudies
University of Oslo
EspenUleberg
University Cultural HistoricMuseums
University'of Oslo .
Abstract: In this paper we attemptto demonstrate that the mythological cultural
landscapes constructed by prehistoric hunter-gatherers are amenable to studies
utilizing GIS technologies. Tbese technologies also provide the opportunity to
addressquestionsthatare not possible using traditional methodology. Inaddition
to mythological descriptions of how the world came into being, hunler-gatherer
cultural landscapes also define and describe the proper behaviour of people
within the landscape. In this way, changing interpretations of the cultural
landscapemaybe reflected inchangesinsite locations.
We presenttwo preliminarystudies aimed attheexaminationofchangeand
continuityincultural landscapesduring the Mesolithicof southernNorway. Both
examples focus upon variation in thetypes of locationschosen for occupation.
lbefirst is from the dynamic coastal landscapes, while the second examines the
distribution of sites in the more stabile interior landscapes. The preliminary
results are encouraging and open a number of new possibilities in the study of
culturallandscapes. .
1. Introduction
The vast majority of GIS-based studies,of prehistoric cultural landscapes have
focussed upon monumental constructions. This focus is understandable, these
constructions provide clearly defined focal points for analysis; and allow the
measurement of a number of variables which can be used to quantify spatial
relationships between monuments. In this paper we hope to show that there are
to basing landscapestudiesupon monuments. Wefocusuponsocieties
that seldom created monuments, hunter-gatherers. While' hunter-gatherers, with
notableexceptions,rarelycreatelongtermmodificationsto theirphysicallandscape,
thisdoes not mean that theydo not constructcultural landscapes. Hunter-gatherers
construct and occupy cultural landscapes containing mythological and historical
descriptionsof their environment [c.f, 7], [19], [21]. While traditional explanatory.
frameworks have tended to treat such cultural constructions as anecdotal, in this
paper we argue that the study of the patterning that results from these social
constructionsprovidesnewavenuesforthestudyofprehistoricculturallandscapes.
Mythological landscapes are often considered to be non-quantifiable, and
thereforenotsuitedtoGISbasedstudies.While theprecise understandingofhowa
T
103
102
J. Boa: and E. Uleberg / HlinterGatherer landscapes
Beyond IiIe Map
I
given Mesolithic hunter-gatherer perceived the landscape is not a realistic goal of
study,webelieve thataspects ofchangeandcontinuityintheconstructionofcultural I
landscapescan bestudiedthrough theuseofGISmethodology.The methodologywe
I
proposeisnotmerelyanadaptationofexistingmethodstoGIS,butratherattemptsto
!
usethecapabilitiesofGIStoaddressquestionsthatcannotbestudied withtraditional
penandpapermethods.
In thefirstsectionof this paperweexplore anumberof topics relatingto the
studyof hunter-gatherer landscapes and how these directly influence the patterning
foundinthearchaeological record. Followingthis wepresenttwoexamples fromthe
Mesolithic of southern Norway that demonstrate why we believe that GIS is the
methodologicaltoolthatisbestsuitedtothestudyofthesequestions.
2. Hunter-GathererCulturalLandscapes
Within hunter-gatherer studies discussions of mythologyand religion are presented
as topics isolated from the more substantial topics of kinship, material culture and
settlement andSUbsistence. Instudies ofprehistorichunter-gatherers,mythologyand
religion are topics that are rarely broached. This may not be too surprising, since
hunter-gatherermythologiesanddescriptionsoflandscapesoccupiedbysupernatural
beings are so clearly fantastic, that it is difficult for us to believe that any rational
personcould takethemseriously.
This separation between the material and religious is a classic example of
archaeologists and anthropologists transferring their understanding of the world to
the object of their study. In western culture, supernatural forces and religion are
clearlydistinct from,andlargely irrelevantto,theproductiveactivities thatstructure
ourlives.Religion,supernaturalpowers,magicandotherfantasticphenomenaareso
obviously transparent and irrelevant to our daily life that they, by and large, have
been relegatedtoformsofentertainment.Thisdistinctionissoselfevidenttousthat
weacceptitasagiven. andreproduceitinourstudiesofthepast.Itisassumed that
the rituals and beliefs described inethnographies are merelyobligatory actions, and
that if we dig deep enough there is a functional explanation that can be related to
whatweassumetobehunter-gatherersprimeconcern,subsistence.
While it is easy for us to dismiss fanciful creation myths of giant turtles
flipping lobs of mud here and there to create the first humans. it is important to
accept that such cultural constructions played an important part in the beliefs and
actions of people in the past. The arena where humans interacted with such
supernatural forces was in the landscape. In the following we will discuss the
relationship between these myths,landscape and the activerole of these landscapes
inthelivesofhunter-gatherersinthepast.
2.1 LivingLandscapes
Hunter-gatherer cultural landscapes consist of locations that are linked to
mythologicaland/orhistorical events. The locationscan rangefrom a waterbody, a
mountain, or an occupation site; essentially any place, location, etc. that can be
namedoridentifiedmayhaveadirectorindirectassociation.These associationscan
beeithermythologicalorhistorical,oracombinationthereof[19].
Anexampleofamythologicalassociationbetweenasupernaturaleventanda
definitelandscape formationisprovided ina RockCree myth[6:43-44], thatrelates
events in the time before human beings were created. A group of animals were
escaping..the monstrousandpredatoryGreatSkunk." Atonepoint,
..... Weasel reports to the other animals that Great Skunk is close behind them.
Wolverine volunteers tofacehimdownatabeaverlodge.HetricksGreatSkunkinto
discharging allhis 'medicine'andthenimmobilizes himwhiletheotheranimalstear
him apart; the fragments become the ancestors of modem skunks. Wolverine
however is blinded by a last spray of Great Skunk 'medicine' and must travel to
HudsonBaytowash;thereafter,theBayissaltyandunfittodrink."(Ibid.)
Another example from the Koyukon in Alaska relates the story of the Raven,
disguisedasaman,fleeingfromanunsuccessfulmarriageproposal: .
"She refused to be his wife,so he made her sink into the mud and disappear; and
thenhe beganpaddlingbackhome.The woman's motherkepttwobrownbears, and
inheranger shetold themtodrowntheyoungman.Theydugfuriously atthe lake's
edge.makinghuge waveseverywhereonthewater. ButRaven calmedanarrowpath
beforehim andpaddled00. Eventuallyhebecame exhausted, sohe threwa harpoon
thatstruck the crest of awave. At thatmoment he fainted from the intensityof his
concentration, and whenhe awoke a forested land had replaced the water. He saw
thatthefirst wavehisharpoonstruckhadbecomeasmallmountain. Thenitglanced
off,eventually strilcinga huge wavethat solidified into another mountain - theone
nowcalledDeenaalee,orMountMcKinley"[15:34]. .
Typicalexamples ofhistoricalassociationsaremuch lesselaborate.The Koyukonin
Alaskaavoided acertainsectionofa riverbecauseof thepresence of thegraveof a
shaman and because two children had also drowned in the area [15:35]. Another
exampleisanareathatisunusuallyrichingame.Theexplanation forthisisthatlong
ago this area was a shamans trappingarea, and that he used magic to increasethe
areasproductivity,and theeffectsofthisarestillvisibletoday[15:205].
2.2 Interconnections
Perhaps the single most importantconcept in coming to grips with hunter-gatherer
cultural landscapes is that everything is interconnected. Creation myths are not
simplyexplanations of the surroundinggeomorphological features, but alsoprovide
linkageswithinthesupernaturalworldaswellasbetweenthat worldandhumans.In'.
the example of the Great Skunk cited earlier, the Wolverine, who has the trickster
role in Cree mythology [6:44], and in most cases displays solitary and selfish
behavior, is often contrasted to the wolf,who displays a group ethic that ishighly
valuedbytheCree.Withinthese landscapesmythologicalassociations arenotmerely
between a single creation myth and a single.object or place in the landscape, but
rather situates the location within a complex mythology, and simultaneously the
locationprovidesavisibleand tangiblelinkto mythology.
Theseinterconnectionsarealsoillustratedinthehistoricalassociations,which
mostoften seem to relateto improperbehaviorby humans resulting in locationsor
objects acquiring supernaturalpowers that can directly influence human existence.
104
~ o n the Map
However, the relationship goes both ways; humans can have direct influence on
supernatural, as well as natural events [c.f. 18:133-135]. The best examples of this
are the numerous methods of controlling weather, or the complex set of rules
regulating the preparations for the hunt and the proper treatment ofthe spoils of the
hunt. These rules, rituals and prohibitions define the interaction between human
beings and their world, as well as defining how the world interacts with human
beings. According tooneKoyukon:
"The country knows. Ifyou do wrong things toit, the whole country knows. Itfeels
what's happening to it. I guess everything is connected together somehow, ..."
[15:241].
The mythicofreligious nature ofhunter-gatherer landscapes is aunifying elementthat
links and intertwines the landscape with religion and production, and there is little
doubt that it was as an importantelement for prehistoric hunter-gatherers as it was
for the hunter-gatherers of the ethnographic present. However by their very nature
such mythical landscapes would seem to be unapproachable in the archaeological
record. In the following we will attempt to demonstrate that this is not the case and
we will use the placement ofoccupation sites within the landscape to provide an
avenue from which to study change and continuity in hunter-gatherer cultural
landscapes.
2.3 Implications
While thespatial and temporal distribution ofresources,forexamplethe presenceor
absence of fish, will have a direct influence on whether or not a given lake will be
utilized, the presence ofthat resourcedoes not determine precisely where occupation
sites will be located. While not trying to argue that considerations related to
subsistence were unimportant in the choice of site locations, it is important to
consider that the choice ofprecisely where a site was located was a result of both
practical considerations and relevant social rules regarding what types of qualities
were acceptable for thelocation ofa site, and therefore the placementofsites in the
cultura1 landscape. For example, the linkage of certain types of water bodies or
specific geographic formations, with positive, negative or neutral associations are
considerations that could have directly affected which locations were chosen for
occupation sites. More generally, the considerations can be studied in terms of the
topographic qualities and the history of the location, or the location's role in the
cultural landscape.
The term 'topographicqualities' ismeant toimplyconsiderationssuch as the
microtopography of the location, its view over the surroundings, its placement in
regards to sunlight and water bodies etc. Among the Mistassini Cree, Tanner
[15:101-105] provides asolid example ofhow the placement of sites on east facing
shores meets a number of practical considerations, as well as providing links to
supernatural forces associated with cardinal directions. It is the interplay between
practical considerations and understanding of the landscape that detennine where
siteswill belocated andwhere theyarenot.
The use of a location alters its status in the landscape and gives new and
addedmeaning and history tothe location.Theevents that occurred atagiven place,
positive andnegative, mythological andhistorical becomeapartofthehistory of the
105
J. Boa; and E, Uleberg / Hunter-Gatherer landscapes
location and its role in the cultura1 landscape [12]. For example, among the
Mistassini, there are clear rules for the cleaning of occupation sites following
abandonment, so as to appease thespirits ofsites, as improper treatment ofthe site
would result in poor hunting in the future [18:74-75]. In another example from the
Koyukon, abandoned village sites were avoided because ofthe presence of ghosts
andevilspirits [15:242-245]. The useofspecific locationsresults inthecreation ofa
historyand a place in thecultural landscape that is unique to the location and atthe
sametimealters locations, andtheirrolewithin thelandscape.
2.4 Spaceand Time
Hunter-gatherer cultural landscapes are not fixed, unchanging creations. One of the
most important aspects ofthese landscapes is the active role that they'canplay. The
dailyinterpretation ofthe interplaybetween humans and the natural and supernatural
landscape provides both challenges and reinforcement for existing ideologies and
practices.
"Actors draw upon their settings; andthe manner inwhich theydosodepends upon
the specificity oftheir relationship to the place. In this manner locales, in the'most
general sense, can be defined as a presencing of potentialitieson which actors draw
inthedailyconduct oftheiractivities" [19:19].
The activerole of thelandscape within hunter-gatherersocieties is seated intheprocessof
social mitigation and social mechanisms for reaching consensus. New elements of the
landscape, observations of new phenomena, or alternative interpretations of existing
mythological or historical relationships are all events that need to be placed or related to
existing mythological and religious observations. The methods for reaching consensus in
hunter-gatherer societies are many, and there are an equal number of methods for
challenging the consensus. Breaking taboos is a common method to challenge existing
interpretations. For example, the Koyukon have extensive rules and taboos regarding the
treatmentofanimals. OneofNelson's [15:137]informants stated that"My brotherdoesn't
follow the taboos. When he cut a beaver's throat in front of me,itjust hurt-to see it
happen" (original italics). Breaking taboos is a challenge to the existing order in any
society. Whetheror not the challenge is,effectiveis dependent upon contexts and results.
FornativeAmericans, perhaps the mosteffective challenge wasat the timeofcontact with
Europeans, who were able to'move through the landscape, ignoring 'or intentionally
violatingtaboos, withoutretribution fromsupernatura1forces [7:159-163].
Changes or absence of change in the norms regarding the characteristics of
locations that are suitable for occupation sites can therefore be taken as an indication of
change or continuity in the construction ofthesocial landscape. For example, norms that
describe the types ofplaces that are considered to be acceptable locations for long-term
OCCupation can bereinterpreted andlocations with different typesofcharacteristics will be
chosen. Changes in patterns of site location may bedue to changes in theconstruction of
thelandscape,and neednotnecessarily berelated solelytochangingsubsistence strategies.
While the preceding has demonstrated the relevance of cultura1 landscapes to
hunter-gatherers of the ethnographic present, the question .remains of how these
considerations can be approached in the study of prehistoric hunter-gatherers, In the
following, twoexamples willbepresentedwhich proposeapproaches thatattempt toutilize
the capability of GIS technology in the study of prehistoric hunter-gatherer cultural
r
106
I
Beyond til< Map
landscapes. However, these examples are works in progress and can largely beconsidered
tobefirst attempts at approachestoproblems.
3.MobileLandscapes
From the retreat of the Pleistocene ice sheet until the appearance of agricultural
adaptations,the Oslofjord area of eastern Norway(Figure 1) was occupied by a variety of
coastallyadaptedhunter-gatherers[5, 14].One constantand uniquefactor inthe prehistory
of this area is the extremityof isostaticrebound [8, 16]. Duringthe Pleistocene, icesheets
on the Hardangervidamountain plateauin southern Norway are estimatedto'have been c.
2500-2800 m thick [20]. The isostatic rebound following the retreat ofthis ice sheet has
been extreme. For examplein the central areas ofthe Oslofjord, the marine boundary, the
areas above sea level after deglaciation are over c. 200 m above the modem sea level
(Figure 2). The rates ofcoastalreboundwere particularlydramaticduring the earliestpart
ofthe Mesolithic.
Between 10,000 and 8000BP the shoreline dropped from 200 to 70 masl (meters
above sea level). This 130 m drop in sea level over 2000 years averages to a sea level
change ofc. 6.5 ern peryear. While rates slowed following 8000 BP, the shore level
changeswere considerableandcontinuousthroughoutthe Mesolithicand Neolithic.
Figure1. Mapof Norway withstudy areas nwted.
J. Boaz and E. UlebergI Husuer-Gasherer ltuuiscopes
Meters above
...-
""
''''
.OJ
so
I.
calibrated
"uncalibrated
V""",BP
llXJOO
...,
~ I 2M
107
Fig=2. Coastalreboundin the SkianaoftheOsloljord(After n s e n 1979, 1996)
Given the low and irregular topography in the Oslofjord area, this extreme isostatic
rebound has lead to dramaticchangesin coastal landscapesthroughoutthe Mesolithic. At a
regional scale there aredramaticdifferencesbetweenthemodemcoastlineandthat of 5000
years ago. The changes in the coastlines in this area can not be described as asimple
gradual, linearadvance of the coastlineover time, but ratherinvolvedramaticchanges in
the topographic structure of the area; islands rise above sea level and eventually are
connected to the mainland, fjords and other channels of communication appear and
disappearovertime (Figure3). When viewed onanevensmallerscale (Figure 4),it isclear
that the coastal changes would have also Involved changes that would have directly
affected the details of the world in which people lived; existing bays gradually change
character and become shallower and smaller. Inechanges in the structure of thefjord
systems would result inchangesin movementofwater throughthesystem.
Rather than the dominant model 'of conservative hunter-gatherers that live in
equally conservative environments, the hunter-gatherers living in these environments
wouldhave had todeal with arapidlyandcontinuouslychanginglandscape.Interms ofthe
constructionofthe cultural landscape, thisimplies thenecessityfortheconstructionof
descriptions/explanations of new elements and a constant re-evaluation. of existing
elements. This couldhave involvedboth theprocessesresponsiblefor.thechanges, as well
as theresult of the process. For example,theappearanceofa small bedrock outcropin the
bayand its gradual transition into an island couldbeexplained/describedbothin termsof
Which processes were involved in its transformation into an island, as well as the
incorporation into the cultural landscape ofthe events that occurred on the island..Also
changes in existing, named landscape fonnations would have had to be negotiated and
explainedinrelationtoexistingunderstandingsof mythologicalevents and relationships.
1
109 108 1
Beyond the Map
T'
9000BP
\ ... .
\ \ ,I, r I I 'j
N \ \ 1'\ -.:.
\\;' \,";;
,,'
, f(
}:
;;i ;Li)
I.-::."r .... ,
"ifl\
\ h."'f/\, ,!;, .,'
\t" }", . I
JC
j
, \
'\. '"
__...

\. 5 000 BP
7000BP
II \}I
\'\\ ",
\ ll: I'
\l l. " .
\ '-, "
l ''-1\
''''\

))
. i Ie' l <',.,-"
'<'\'(
"l .,!
'\.... c..", I j.,;, I' ! 0':
\)'f 'r
\hf ,". '.
l
l.
if
v c \ )\ 'j i
'l t \ I. ),'\
,,"). \\ \
'i'\ \,JI.,\ ,I. .'.
'1-' \1\ Vi
r \!ll':4'- 1'<) , o 2.5 5 K_...
\ . \\1 . L J\'-6'
..\'.:nf\__ .\ 01"-1
1\ _ .. \" 9,' " =-----
Figure 3.Changes inshorelines inbetween 9000 - 5000 B.P: ina28kmx28kmsection of thestudyarea
thisareabut ratherin smallerdikesscatteredthroughoutthearea, atdifferentheights
abovesea level.Over time, newdiabasedikes would appear from the ocean, while
older ones would, in effect, move further away from the coast. The use of the raw
materialsfrom these dikes in the production of axes provides a clear link between
materialcultureusedindailyactivitiesandthechanginglandscapes.
3.1 Site Location andCultural Change
InthefollowingwesuggestGISmethodologydirectedtowardstheidentificationand
chronological placement of cultural change within these continually changing
landscapes. The chronological divisions utilized for the identification of cultural
changeinMesolithic easternNorwayisbaseduponchronologicalsequencesofaxes
and microliths that are used todivideprehistory into phases with absolute temporal
boundaries. These boundaries define the transitions that are understood as
representing significant culturalchange. In the following we propose an alternative
approach; basing identification of cultural change upon changing landscape
perceptions.
Throughout theMesolithictherewouldhavebeenacontinual needtorelocate
sites, due to the changing topography. As such the norms governing site locations
would have been under constant renegotiation, changes in the types of locations
chosen for occupation should be indicative of changes in the understanding of the
J.Boat and E. Uleberg I Hunter-Gatherer landscapes
500

N
N
SOmas!.
Figure4.Coastal changesonalocalscale,elevationsare marlredin5m
culturallandscape and associaied mythology.We assume that cultural change will
alsobereflectedin changingpatternsoflandscapeuse, and that thelocationofsites
in relation to topographic features is not solely a reflection of subsistence
considerations,but also reflects norms or changes in norms related to sitelocation.
These changes mayor may not occur'simultaneously with changes in artifact
assemblages. .
In this there is clearly overlap with concerns of subsistence, and as
demonstrated by Akerlund [22:115-119] changing sea levels can have significant
effectsongeographicvariables relevantto subsistence. It is not meant to implythat
concernsofsettlement andsubsistenceareirrelevantinthechoiceof sitelocation.In
thecase of hunter-gatherer cultural landscapes, the importance of preyanimals in
mythologyclearly demonstratesthatsubsistence; landscapeandmythology are linked
andinter-related.Alloftheseelementsareimportantconsiderations inchoiceofsite
locations.
In this we suggest the useof viewshedsto categorize the types of landscape .
elements thatare visible from the sites, for example the different types of water
bodies(bays, inlets, open ocean, etc.) or tecrestrialfeatures (mountains, mainland,
islands,etc.).Allofthesecharacteristicsarerelevanttotheplacement ofagivensite
within the cultural landscape, as well as relating to factors of settlement and
subsistence.'Throughthe comparison of sites of different types at different heights
'I
III .
110
Beyond lh. Map
above sea level, it should be possible to identify periods of change and continuity,
which then can be compared to traditional, artifact based definitions of change.
Preliminary efforts towards this have been promising, and are illustrated in
Figures 5-8. These figures show the viewsheds from eight sites from about a 20 km
2
area in Borre township in Vestfold County.
Only one site from our study area is dated to the earliest Mesolithic (c. 9000-
8(00). The viewshed from this site shows a clear maritime orientation, no mainland
areas and only the outer coastal islands are visible from this site (Figure 5). In the
Middle Mesolithic a completely different pattern is visible. There are four sites at c.
80 masl, c. 7000 BP. As shown in Figure 6, the viewsheds from these sites cover
virtually every possible area, from mainland to the outer islands, however there is a
clear dominance of overlapping viewsheds in the area between mainland and coastal
islands. In contrast, only a single site is dated to the period at c. 6000 BP (Figure 7),
however it is also the only excavated site in the sample, the site of Frebergsvik [13].
The viewshed from this site shows a clear orientation toward the ocean, which
coincides well with the faunal assemblage from the site, which is dominated by sea
mammal bone. In contrast the sites dated to the end of the Mesolithic at c. 5000 BP
(Figure 8), while located on islands or peninsulas have viewsheds that are clearly
oriented towards the mainland with virtually no open ocean areas included in the
viewsheds.
These examples clearly suffer from the limited sample size, however, the clear
patterning and variability over time is promising for further analysis. In addition to
sample size, two additional problems remain. The first of these is the ability to model
marine hydrodynamics within a GIS, in order to adequately characterize the types of
water bodies that were present within these viewsheds (currents, tidal movements,
etc.). Especially for marine hunter-gatherers the characteristics of the nearby marine
environment would have important implications for both subsistence and the cultural
landscape
The other major problem is that inherent with the use of a 2D GIS in a 3D
world [9]. The compression of time into slices, in this case 5000 thousand years into
four static diagrams is clearly inadequate, and the use of 3D GIS to model change
over time would clearly provide a more adequate understanding of the location of
sites within the changing landscapes.
4. StableLandscapes
Our second suggestion for the application of GIS to the study of hunter-gatherer
cultural landscapes is from a completely different region, the high mountain areas of
southern Norway. In contrast to coastal areas, these areas above the treeline are
marked by considerable environmental and geographical stability.
The Mesolithic and Neolithic sites in these areas have long provided
significant challenges for archaeologists. In most cases there are clear indications of
multiple occupations, although stratigraphically separated occupations are virtually
absent. Faunal remains are rarely encountered and the assemblages are characterized
by a lack of chronological indicators and retouched artifacts. As such the placement
of these sites in chronological sequences andsettlement and subsistence patterns has
been difficult [4], [10].
J. Boa; and E. Uleberg / Hunter-Gatherer landscapes
Figure 5. The study area at c. 9000 B.P. (one site)
Figure 7. The study area at c. 7000 B.P. (one site)
Figure 6. The study ";,,eaat 8000 B.P. (four sites)

u
I
Figure 8. The studyarea at c. 5000 B.P. (two sites) A .
0"1 Hm W.
N
r
112 113 .
Beyond the Map
Discussions of these sites have almost exclusively focussed upon the
extraction of resources, more specifically, the location of the sites in relation to
reindeer hunting and reindeer migration routes [1], (10]; [II]. However, associations
with the migration routes have been difficult to demonstrate, either in terms of sites
directly associated with the routes or higher densities of sites in the lakes in the
vicinity of these routes. In short, processual approaches focussing on subsistence and
resource utilization have generally failed in these areas.
We believe that by changing focus, and by asking different questions, these
intractable sites can provide new information. Rather than focussing upon resources,
we suggest that a consideration of the role of these locations within the cultural
landscape provides avenues to ask a number of new questions. In the following
example we suggest a focus upon assessing the importance of the choice of site
location.
Again, as was the case in the coastal areas, the questions that we are
interested in addressing are the identification of change. In contrast to the coastal
areas, where choices would have had to have continually been made regarding where
new occupation sites would be located, in the interior areas the choice of locations
for occupation sites could involve the reuse of sites used earlier, sites with traditions,
or the creation of new occupation sites.
4. J Analysis
We propose approaching this problem from two directions. The first is from the
characteristics of the locations chosen for occupation sites. The question we are
asking is not what types of artifacts the site contains but rather, what are the
microtopographic and macrotopographic characteristics that distinguish this site from
areas not chosen for occupation. By macrotopographic we mean considerations such
as the relative location of the site, for example on the shore of a bay, on a bluff
overlooking a lake or on a peninsula or promontory. By microtopographic
considerations we refer to the characteristics of the individual location, for example,
the presence and abundance of bedrock outcrops which both provide protection from
the elements and divide and structure the space at these locations. The second is
attempting to assess the degree to which the use of certain locations focuses activities
with those areas, that is, how does the history of repeated use of a certain locale focus
activities upon the same location or within the same general area [19].
Both of these considerations can be studied through the comparison of the
features that characterize sites and randomly chosen, non-site locations. Through this
it should be possible to define the characteristics that are unique to the locations
chosen for occupation. Through the comparison of variability in choices in site
location over time, in the face of environmental stability, we should be able to
describe the presence or absence of changing norms regarding the factors involved in
which locations were chosen for settlement. For example, are there sites at all
locations which provide a commanding view of the nearby surroundings, or are there
concentrations of sites in certain areas? Through the use of GIS it will also be
possible to compare these variables on a regional scale, which would provide the
ability to address questions that have not been addressed before.
Our preliminary efforts towards this are interesting. Figure 9 shows the location of
eight sites at lake Eldrevatn in Lardal township, in the county of Sogn og Fjordane.
The viewsheds from these sites (Figure 10) provide somewhat unexpected results in
J. Boaz and E. Uleberg / Hunter-Gatherer landscapes
,.
D-
,N
Figure 9. Location of sites and lakes discussed in the text
'j Hm
D-
N
Figure 10. Viewsheds from three sites at take 01jusjllen
that with a single exception, the viewsheds from thesites provide visual reference to
the same areas of the lake, irregardless of the precise location of the site. While the'
limited sample clearly restricts the utility of generalizing about this pattern, the
patterning is intriguing and bears further study.
r
!
114 115
Beyond the Map
S.ConclusioDS
Anoften neglected consideration in archaeology is that most often when we begin
asking new questions of the archaeological record, we continueto use data that was
collected for other purposes. The classic example of this is artifact or antiquity
typologies, which were originallydesignedfor classificatorypurposes, howeverwith
changing paradigms, the sub-division ofthese old classificatory schemes have now
been assumed to have social meaning. New questions call for new data and new
methodsofanalysis, and webelievethat GIS is well suitedto this. Ratherthan solely
focussing on the same monuments that have beenstudied throughout the historyof
archaeology,we have attemptedtoemphasizethe importanceofthe non-monumental
in the constructionofculturallandscapes.
In the preceding we have attempted to demonstrate the potential ofthe use of
GIS in the analysis ofprehistoric hunter-gatherercultural landscapes. Ouremphasis
has been upon suggesting methodologies to identify cultural change based upon
changes in landscape perception, rather than basing our understanding ofchange
uponartifacttypologies.
We have also attempted to emphasize the importance ofincluding 'natural'
elements in cultural landscapes. By shifting the focus from monumental
constructions to the less tangible social constructions it is possible to utilize the
geographiccapabilitiesof GIS to ask new questionsofthe archaeologicalrecord.
References
[1] Bang-Andersen, S. The first reindeer hunters in the Southern Norwegian High Mounlains. In
Tbevenin, A. (ed) L'Europe des demiers chasseurs: Epipaliolithique et Misolilhique, pp. 341-
346.Actes du 5e ColloqueInternationalUISPP, CorrunissionXII.erHS,Paris, 1999.
[2] Berg,E.Mesolitiske Boplosser ved Af'IUIgen i Asog Frogn; AlershllS: DObbeitsporlE6 .prosjek:
1996. Varia34. UniversitetsOldsaJcsamling,Oslo, 1997.
[3] Berg, E. Rawmaterialuseandaxe productionintheMesolithicof SoutheasternNorway. In Boaz,
J. (ed.) The Mesolithic of Central Scandinavia, pp. 267-282. Universitetets OldsaJcsamlings
SkrifterNy 22, Oslo, 1999.
(4) Boaz, J. HlUl1er-gatherer site variability: changing patterns of site utilization in the interior of
eastern Norway, between 8000 and 2500 B.P. OldsaJcsamlingens Skrifter,Ny Nr. 20, Oslo,
1998.
[5] Boaz,J. The Mesolithic of Central Scandinavia: slaIUs andperspectives. In Boaz, J. (ed.) The
Mesolilhic of Central Scandinavia. pp. 125-152. UniversitetetsOJdsaksamlingsSkrifterNy
22, Oslo, 1999.
[6] Brightman, R. Grateful. prey: Rock Cree hlUflQ1l-<mima1 relationships. University of California
Press,Berkeley,1993.
(7) Burch,E.The nonempiricalenvironmentof the Arctic AlaskanEskimos. Southwestern Journal of
AtIlhropology 27,1971,pp.148-165.
(8) Hafsten, U. Shore-level changes in south Norway during the last 13,000 yean, traced by
biostratigraphical methodsandradiometricdatings. Norst Geographisk Tidskrift 37, 1983, pp. 63-
79.
(9) Harris, T. and G. Lock. Multi-dimensional GIS: exploratory approaches to spatial and tempora1
relationships within archaeological stratigraphy. In Kamermans, H. and K. Fennema (eds),
ftllerfacing the past: computer applications and qlUJllJilalive methods in archaeology CAA 95, pp.
297-306.Paleohistoria28, Leiden, 1996.
J. Boaz and E. Uleberg I Humer-Gasherer landscapes
[10] Indrelid, S. Fangstlolk og i Fjellet. Bidrag til Hardangerviddas 8500-2500
Ar Ndtid. 01dsaJcsam1ingens rekke Nr. 17.Oslo, 1994. .
[II)Johansen,A. ved Lardals Yassdraget. Bind 1/. Oslo, 1978.
[12] KUchler,S.Landscapeasmemory:.The mappingofprocessandits represeotatiooin Melanesian
Society. In Bender, B. (ed) Landscape: Politics and Perspectives, pp. 85-106. Berg, Oxford,
1993. . .
[l3) E. Frebergsvik. Et mesolitisk boplassomrad ved Oslofjorden. Universititets
OldsaksamlingenskrifterNy 1.Oslo, 1975. .
[14] E. Seasonality and mesolithic adaptation in Norway. In New directions. in
Scandinavian Archaeology, edited by K. Kristiansen and C. Paludan-Muller, pp. 79-119.
Copenhagen, 1978.
(15) Nelson, R. Make prayers to the Raven: A Koyukon. view of the Northern Forest. University of
ChicagoPress,Chicago, 1983.' .
(16) Sllrensen,R. LateWeische1ian deglacitaioninthe Oslofjordarea, South Norway. Boreas 8, 1979,
pp.241-246.
[17]Sllrensen,R.Frogn Bygdebok. Bind f. Fro urtid til ca. 1550.Frogo Kommune, Drebak, 1996.
[18]Tanner. A. Bringing home animals: Religious ideology and mode ofproduction of the Mistassini
Cree hlUl1ers. C;Hurst &Company.London, 1979. .
(19) Tilley, C. A phenomenology of landscape: places. paths and _tIls. Berg Publishers,
Oxford, 1994.
[20] Vorreo,T. Weichselianice movementsin South Norway andadjaceolareas.Boreas 6, pp. 247-
257,1977-
[21]Zvelebi1,M. Hunter-gathererritual landscapes: spatial organizatioo,social slr\JCture and ideology
among hunlac-galhererof NonhemEuropeandWesternSiberia.InvanGijn, A.andM. Zvelebil
(eds.) Ideology and social structure of Stone Age Communities in Europe. pp. 33-50..Analectica
PraehistoricaLeideosia29, Leiden, 1997. .
[22] Akerlund, A. HIUPIQJt responses to share displacemetll: Living 'by the sea in Eastern Middle
Sweden @ring the Stone Age. Studier fdn UV Stockholm, Arkeologiska undersOkningar,
Slcrifter 16,1996.
,.,.-
117
116
BeyondtheMap
G. Lock(Ed)
lOSPress.2000
Session2discussion:
MovingGIS: exploring
movementwithin prehistoric
culturallandscapesusingGIS
Trevor Harris
Department of Geologyand Geography
WestVirginia University
1. Introduction
lri a recent discussion broadcast on national radio in the United States, one of the
discussants stated "Life is about movement. Without movement there is no life."
Althoughused in thecontextof internationalmigration, this statementisrelevant toalmost
any discussion of 'movement' whether it be vehicular traffic, information flow, the
transportationof goods. or the movementofpeople, or even animals. The examinationof
the movementof people at an aggregatedlevel has taken place since the census returns of
the nineteenth century. More recently, mapping individual time-space paths sought to
capture thedynamicsofthespecific movementby people through time. The re-creationof
individual space-time paths through the construction of sibling time paths based on the
nineteenth century census enumerator's returns is one example of seeking to analyze the
historicalmovementofindividuals[1]. Itisclear that without examiningmovementor the
activity patterns of peoplein time, the geographyofsociety isfossilized and non-dynamic.
Despitebeing afundamental ofeveryday life, it is interesting to note how difficult it isto
exploreissues ofmovementinaGIS. Areview oftheprimarytext onGIS indicates onlya
smatteringofreferences to movement using GIS and the GIS chapterdedicated to this is
primarilyfocused on transportation and transportation networks [2]. When considered in
the contextof archaeology and the primary reliance upon non-textual, non-statistical data
then suchdifficultiesaresubstantiallycompounded.
There has been some attempt to explore movement across a landscape in an
archaeological setting using GIS, but such studies are few. Lock and Harris [3], for
example,in their GIS study of the Danebury hillfort sought to recreate the social relations
that linked the centralizedhillfort tothe surrounding farmsteads and settlements. lri doing
so it is easy to fall into the trap ofperforming an analysis as if the landscape were an
isotropic surface and uniformly accessible from any direction. Personal experience
alone indicates that going uphill is more difficult thantraveling downhill, that crossing a
stream can cause additional delays that may even force a traveler to seek a detour to a
bettercrossingpoint, that marshland ismoredifficult totraverse than dry land and soforth.
In their study ofDanebury, Lock and Harris sought to model the relative difficulty, or
ease, of crossing a landscape based upon a GIS model using distance, slope, soil type,
and the absence or presence of standing water bodies. The development of such a
T.HarrisISession2discussion
'friction' surface, representing the ease of access to various parts of the landscape
surrounding Danebury, was importantfor identifying relative access across the landscape.
Treating accessibility or movement across a landscape as being merely a measure of
absolute distanceignoresthe realcomplexitiesinvolved inanalyzingmovementin the real
world. It also indicates that without the.availability of contemporary accounts or other
measurable criteria, that archaeologists are forced to adopt a terrain-based approach to
analysis. invariably using contemporary environmental data. Alternative approaches to
gaining insight into patterns of historical or pre-historic movement within a cultural
landscape are thus intriguing to explore. The papers presented at the conferenceprovide
three such alternative approaches and are valuable as much for their contribution In
identifying significantissues about movementwithin prehistoriclandscapesastotheactual
resultsthat theypresented.
2. Thestudies
The conference presentations by Llobera, Bell and Lock, and Boaz and U1eberg, are .
valuable for their focus on seeking to understand movement within prehistoric cultural
landscapes using GIS and for doing so in differing archeeological and regional contexts.
The paper by Boaz and Uleberg on Quantifying the non-quantifiable: Studying hunter-
gatherer landscapes seeks to address the many issues associated with the GIS analysis of
prehistoric culturallandscapeshead-on. Theirintent ischallengingand fascinating in that
they deliberately seek not to focus on clearly identifiable monumental constructions and
the analysis of the relationships between them, the core ofGIS studies in archaeology to
date, but rather to focus on societies that seldom created monuments and represent the
most difficult of groups to model in GIS, that of hunter-gatherers. Their analysis thus
shifts the focus away from monuments, the formation of elite groups, ritual and
domination, and focuses instead on what must be one of the most challengingoftopics -
the examination of groups which by their very nature are nomadic and transitory.
Examining such groups through the use ofGIS is truly challenging to the researcher.
Hunter-gatherersocieties,the authors claim, rarely created long-term modification to their
physical landscapes but did construct cultural landscapes comprising mythological and
historical descriptions oftheir environment. Nature and cuiture are thus symbiotically
entwined and any attempt to separate them as components ofa study is invalid. The.
cultural landscapes ofhunter-gatherers consist of locations that are intricately linked to
mythological and/or historical associatiorls that explain and reinforce the individual's or
group'spositionin theworld. The landscape isanactive participantin the daily activities
ofhunter-gatherers. To undertakesuchastudy must imply shifting GIS from itspositivist,
processualistorigins, into the realms ofpost-processual application..To perform this, the
theoretical assumptions of GISmust be challenged and the technology itselfadapted to
confront the tendency toward.physical representations ofa cultural landscape to one in'
which cognitive, non-quantitative forms are significant factors. This is certainly a
challenge toarchaeologistsbutespeciallysoto those wishing to take aGIS approach. It is
intriguing to speculate as to how GIS might be used to support such an unorthodox
endeavor in which the analysis is so heavily laden with qualitative and non-quantitative
data.
Boaz and U1eberg provide two examples as to how such a study might be
undertaken. These comprise the examination of settlement location and pattems of raw
material use. It is not a criticism of the authors that havingset out to undertake such a
complex and difficulttask it isnot surprising that the examplecase studies fall somewhat
short of the loftygoals. lri theexaminationof settlementlocationarather'traditional useof
118
Beyond the Map
GIS is proposed to explain settlement in the context of available natural resources.
Because location decision making is not solely a function of resource distribution it is
proposed to study the characteristics of settlement locations by drawing upon the type of
landscape formation that is visible, as well as the micro-topography. Continuity of use, or
breaks in the use of selected sites within the same region through time, would indicate
elements of continuity or discontinuity of the cultural landscape.
The second example draws upon changing patterns of lithic raw material use.
Again, I would emphasize that the value of this study lays not so much in the attempted
case studies but in the nature of the problem that is proposed to be solved using GIS.
Examining nomadic movement, and inferring landscape use through mythological and
historical associations is complex in the extreme but it exposes a trait in the use of GIS in
archaeological studies to date. GIS has largely been used to explore those parts of
archaeology in which the problem or the available data are more easily addressed. 'The
more retractable questions, however, are no least important despite the difficulty of
modeling them in a GIS environment. The use of GIS for archaeological analysis thus
brings with it silences - those problems, regions, or topics for which the use of GIS
becomes more retractable. Such is the case of the hunter-gatherer societies. This is an
important point to note, for powerful as GIS is, it has the potential to significantly impact
the archaeological research agenda. As noted in the introductory chapter, without data GIS
is a hollow shell and can contribute little and it is lileely that data rich environments will
tend to predominate in GIS studies. It is in this context that the study of hunter-gatherer
societies using GIS poses more questions than this preliminary study by Boaz and Uleberg
could begin to address. How then might other prehistoric societies be studied using GIS?
How is non-documented movement through a landscape to be examined and explored
using GIS? How might other landscape elements, other than cultural monuments, be
included in the GIS? How might the non-quantifiable and cognitive aspects of a culture be
captured and explored using GIS? How can GIS be used in the environmental
reconstruction of prehistoric landscapes and what level of detail is required to support such
work? What is an appropriate scale of analysis to be used in such studies? Clearly, this
study is already moving toward one of a heavily interpretive nature and it must be
questionable as to what extent a GIS would or should contribute to that exposition.
Certainly, this question requires serious thought to warrant the significant data and
processing requirements needed for a GIS analysis. The interaction between society and
nature also poses significant challenges. Despite these questions and doubts it is to be
hoped that many are willing to pursue such points for the depth of GIS-based analysis and
innovative uses will only increase as a result. If nothing else the limits to GIS use will be
defined and researchers must then be cognizant of the inherent biases resulting from the
use of the technology.
The paper by Bell and Lock, Topographic and cultural influences on walking the
Ridgeway in later prehistoric times, seeles to directly examine the process of movement
along the ancient Ridgeway path that straddles the challe downlands of central England.
The authors seek to gain a humanistic understanding of landscape through an integration of
cultural landscape features with an optimal path analysis of the Ridgeway and its hillforts.
Movement in this study is seen as a central explanatory leey to understanding the form,
function, and placement of the hillforts from the perspective of actually moving along the
Ridgeway. Seelcing to explain monumental cultural features from the perspective of
movement along an optimal path is an innovative approach to the use of GIS in
archaeological analysis. Using a lO-metre OEM, an anisotropic cost surface is generated
based on slope and aspect. The resultant surface is a directionally based cost surface, the
traversal of which is heavily influenced by the direction of travel across the surface from
one point to another. Using this cost surface, optimal paths are generated between two
119
T. Harris I Session 2 discussion
known points. A raster approach is used to generate a network of optimal paths that it is
hoped will add insight to the location oftracleways, field boundaries, and the location of
monuments within the cultural landscape. Adding a culturally constituted notion of place
to this heavily reductionist approach is pursued through a reinterpretation of the Ridgeway
hillforts as focal points of social identity. The question of scale is brought to the fore by
focusing not only on a coarse regional level but in using such a technique to explore
blocleed entrances to hillforts and micro deviations in route as evidenced by a comparison
of the optimal path with the known remains of the trail. .
This paper, then, deals directly with movement and the identification of optimal
paths along a well-known ridgeline. The use of an anisotropic surface is interesting in the
way in which the direction of travel is constrained in order to maximize forward progress
between two points or in a mown general direction. In the task that the authors set
themselves, the results are impressive for the optimal path generated closely matches the
lenown route even to the point of identifying minor deviations or 'detours' to take in the
hillforts. However, given the distinctive landform of the region in which this study was
undertaken, it would perhaps be surprising if the optimally generated paths did not closely
mirror the Ridgeway path. The significant question is perhaps more to see jf such a model
would have validity if the landform of the study area were less well defined. Furthermore,
the authors were able to identify the success, or failure, of the model. because much of the
original Ridgeway path can be clearly' identified even today. It does pose the' question,
however, as to how the success of the model could be tested or validated if such evidence
were not available for corroboration of the results. Many studies using automated or semi-
automated routing techniques have generated routes subsequent to a 'manual' or non-
automated study. Verification of the accuracy of the automated routing procedure has been
possible through a comparison with the results of the manual siting study. The enthusiasm
for the 'accurate' results of the automated approach has then been vocal. However, it is
open to speculation as to how such verification or validation of the model could be.
undertaken in instances where landscapes are much less well defined geomorphologically
or where, in essence, the 'answer' is actually unknown. How then might the 'success' of
such an optimal routing be evaluated?
The objective of Llobera's paper, Understanding movement: a pilot model towards
the sociology of movement, is an attempt to move beyond the economic spatial models of
processual archaeology used in landscape studies and to understand movement and,
significantly, the sociology of movement ina 'natural' landscape. The focus is on
identifying general patterns of movement rather than specific paths. Importantly, how-the
notion of these patterns of movement might include perspectives on their creation,
evolution and use, and how they carne about is central to the question posed: The model is
humanistic and does not claim to be an objective result. In this way he seeles to explore a
vital process in the socialization of people into a landscape. The emphasis on the
sociology of movement and of the cognitive underpinnings of movement is important and I
will dwell in greater detail on this point later in this discussion.
3. Common themes and insights
While each of the studies discussed above is set in a different regional and archaeological
context there would appear to be some important common themes between these three
studies. Firstly, they represent some of the first studies to directly address issues of
movement across a prehistoric cultural landscape. Most archaeological studies to date
have focused on an almost static landscape - static both in terms of time and space. GIS
has traditionally struggled with such dynamic' analyses. By posing the questions as tliey
I
,..,.
120 121
Beyond Ihe Map
have the authors have identified several fascinating research tracks for subsequent
exploration. Secondly, and in similar manner, because these prehistoric landscapes and
their artifacts must speak to us about the logics, symbolism, reasoning, and experiences of
people long gone and for whom almost no written evidence exists, the studies again
identify and pose a series of complex questions and issues for archaeological GIS analysts.
Given that examining movement in contemporary society with GIS is, in itself, a complex
process this problem would initially appear to be insurmountable. That is why the real
value of these three papers is not in the 'answers' that they provide but in the questions and
problems they pose, the insight that they provide, and examples as to how these questions
might be pursued through these innovative studies.
Thirdly, as mentioned above, GIS brings with it its own issues. In all three papers
it is notable how the authors sought to move beyond a purely spatial analytic analysis and
explore ways to integrate or address qualitative aspects of the cultural landscape. As
discussed in the Introduction to this volume, while the power of GIS is readily
acknowledged and demonstrated, the technology has yet to be questioned by archaeologists
for the assumptions, constraints, and biases that it brings to archaeological studies. Not
least, there are many instances in which GIS is not the best methodology or cannot readily
be applied to particular questions or areas of study. These silences, whether data induced
or methodologically induced, are important for archaeologists to acknowledge and be
aware of. The need demonstrated by these three studies has been to explore aspects of the
cultural landscape by linking both qualitative and quantitative aspects of nature and society
within a GIS. In this they have sought to explore GIS in the context of philosophical
developments arising out of post-processualist theory. GIS does not lend itself easily to
such interpretations. A number of these issues are worthy of further investigation. The
notion that GIS should not give a 'result' but should be used as a tool for interpretation was
made in all three papers. What data is available for inclusion within the GIS is necessarily
a significant factor in determining the types of questions that can be posed or addressed
using the technology. GIS imposes the Boolean logic of the Turing machine and this, in its
own right, will influence the nature of GIS studies. How other logics and epistemologies
might be explored in a GIS environment remains open to question. The concern not just
for environmental determinism but also for spatial determinism brought about by the need
to link data to the spatial primitives of point, line, polygon, or raster will be discussed
subsequently in the context of how qualitative aspects of archaeological study can be
incorporated within a GIS analysis. Many of these points are discussed in some detail in
the Introduction and will not reproduced here.
Fourthly, there are several points of interest with regard to the actual procedures as
to how optimal paths are generated across a cost surface. It is one of the strengths of GIS
that the raster and vector data models each provide a different strength in the approach that
can be taken to determine shortest paths or optimal paths. Vector systems of course draw
heavily upon networks defined by connecting arcs or segments and for which attributes
concerning those segments can be used to constrain or influence flows through the
network. Most well known are the transportation networks that model flows of vehicles
through a network that can be constrained by the flow capacity of each segment, the rate of
flow or speed of the movement, the direction of flow, and the extent of inter-connectivity
to other parts of the network. In the studies outlined above, and indeed possibly for most
prehistoric landscape studies, it would seem that the raster data model provides greater
opportunity to explore archaeological questions than the vector data model. However, a
shortest path algorithm applied across a discrete cost surface. is influenced by an infinite
spatial variation and an infinite permutation of directional moves. The assignment of
suitability scores to the rasters to denote both the landscape characteristic and the
T. Harris/ Session 2 discussion
characteristics of movement across the cost surface are important in determining the
delineation of the paths.
A number of other issues concerning the GIS assessment of movement across a
surface were only cursorily discussed in the three papers. In the same way that data is
significant to the outcome of a GIS analysis, so to is the selection of the. optimizing
algorithm and the assumptions that lie behind the various optimal routing procedures to be
used. Different algorithms, differing cell sizes, and the number of the directional moves
permitted, will influence the outcome of an optimal routing study and their importance
should not be minimized. Several points are made here to indicate that undertaking a
routing analysis in GIS is not so straight forward as it might initially appear. Firstly, the
algorithm to be used and the nature of the minimization/maximization procedure must be
carefully identified [4]. The specific algorithms used for shortest path algorithms will
influence the outcome. Dijkstra's shortest path algorithm, for example, works well with
larger grid data sets whereas the Bellman Ford algorithm works well with smaller grid data .
sets [5]. Both algorithms are available in the cost-distance function of ESRI's ArcInfo .
software for example. In addition, Eastman's push broom algorithm is yet another
available for possible selection in the routing study. [6]. Particularly valuable is the
Dijkstra-Lombard gateway algorithm that supports cell based route modeling and uses a.
gateway path generation method to force a route to travel through selected intermediate.
points [7].
Secondly, it should be noted that given the almost infinite number of directional
moves possible across a cost lattice that the nature of these moves will also influence the
outcome. Huber and Church [8] define a value system based on the number of linking
nodes to the origin cell in which R=O when the four adjacent closest cells are linked; R= 2
when the nearest 16 nodes are linked; R=3 when 32 nodes are linked; and R=4 when 64
nodes are linked. Thus as R increases and the number of non-adjacent closest nodes
increases, the total number of directions available to move from a cell also increases.
Huber and Church found that including diagonal links and increasing the value of R
markedly improves the routing accuracy. The elongation error of geometric distortion was
as much as 41% when R=Ocompared to less than 1% when the 48 nearest cells were used
[9],[10].
Cell size is also a somewhat arbitrary factor in. optimal path studies. Cell size
should be of sufficient resolution to capture the accuracy of the original source data.
However, most GIS databases are usually made up of multiple sources captured at different
scales, spatial resolution, and positional accuracy. Empirical studies have noted that
increasing cell size has a significant effeCt in reducing map accuracy. As a rule, cell size
or resolution is usually determined such that the cell captures 50% of the space occupied
by the smallest polygon identified. In tests using Idrisi and ArcInfo software on a database
variously rasterized using I, 5, 25, 50, and 100 m cell resolutions it was found that Idrisi
generated five different shortest paths albeit within a single clearly defined corridor.
However, Arc1nfo generated two very different corridors that were clearly dependent 0"
the differing cell resolutions. Differing cell sizes can thus produce different routes. Added
to this the orientation of the raster grid, usually arbitrarily made parallel to the coordinate
system, can generate distortion; the actual process of rasterization Whether using the
central point or the dominant areal unit; and the determination of cell scores and wei ghts
can all influence the outcome of the optimal path analysis [11]. 1be point to be made from
this discussion is that there is not just one method for defining and modeling an optimal
path of movement across a landscape. The 'route selection process is sensitive to the
algorithm selection, the availability of different directional moves, and the selection of cell
resolution to name but a few.
T
122 123
BeyondtheMap T. Harris/ Session 2discussion
The concluding point is a fundamental one that addresses the important post-
processualist approach sought in the use of GIS for exploring movement within a
prehistoriccultural landscape and of the necessary cognitive and qualitative aspects of a
culturallandscapethat are neededtosupportsuch an interpretation. This pointunderlayall
three papers, and is especially related to the sociology ofmovement in Llobera's paper.
All three paperswere painstakingin seeking to move the discussion beyond optimal path
generationbasedupon physicalparameterssuch asslope, toaddressfundamentalquestions
about prehistoric societies and the role of movement. One basic question raised in each
paper was how to assess how space is perceived and used by people and how these
cognitiveaspects ofbehaviorcan be capturedand modeledin a GIS. Understanding how
peopleperceive,learn, use, and navigatealandscape iscentralto understanding landscape
use. Peopleundertake complexmental operationsto store, retrieve, interpret, and analyze
landscape information. This mental model of realityis then usedin theirspatialdecision-
making. As Boazand Ulebergnote, self-identityisinherentlytied to knowledgeofnature
and the spatialenvironment. A mental map acts as a filing systemofspatialenvironment
memoriesand isusedto navigate across alandscapeor interactingwithin the environment.
Mental maps form the backdropof everydaylife and reflect notions of safety, security,of
the familiarand unfamiliar. Evenemotionsare encodedon thesemental maps becausethe
maps are created through the senses of sight, sound, smell, touch, taste - the direct
experience of life. A cartographic map represents a greatly simplified version of that
realityand is rarelythe sameas these distorted maps that begin at birthand continueto be
refined throughlife. Babieshave primitivemaps that are egocentricand relatethe world to
the child's own position. As knowledge of the environment grows through sequential
exploration, so familiar pathways and features are formed. Knowledge is greatest about
local surroundings but large gaps can occur. Adults connect information to remember
routes and landmarks and this information is obtained from direct experience as well as
indirect experiences such as through conversation. The egocentric view of the child is
replaced over time by a geocentric view in which individuals orient themselves with
respect to the external environment and structure their mental maps in such a way as to
envisageand connectthesepathways. Mostpeopleare very adeptat this and hardlynotice
the mental exercises. Thisinnateabilityserveswell infamiliarsurroundingsand generally
awareness onlysurfaces when one is lost or a sense of direction has failed. In this case
peoplereturn toamore primitivepathwaystructuringfor minimizingdisorientation.
"Tobecomecompletely lost is perhaps a ratherrare experiencefor most people....we are
supported by the presence ofothers and by special way finding devices: maps, street
numbers, route signs, bus placards. But let the mishap of disorientation occur, and the
sense ofanxietyand eventerrorthat accompaniesitrevealstous how closelyitis linked to
our senseofbalance and well-being. The very word 'lost' in our language means more
than simplegeographicaluncertainty:itcarriesatone ofutterdisaster"[12].
Given that people's viewofthe worldand the environmentis non-Euclidian it is difficult
to predictfrom oneto another. In the contextof exploringcognitiveaspects of movement
and interaction within a cultural landscape thisaspect causes considerable difficulties in
the modeling process. Information also originates from our imaginaries, fantasies,
hallucinationsoverwhichwe have little controlyetthey form ourview ofthe world. Thus
we rarely see thingsas they are but as weperceivethem to be and perceptioniscloudedby
past experience, biases and limitations - a conglomeration of fact and fiction gathered
through a haphazard combination ofdirect, indirect and extra perceptual experience. If
mental maps are influenced by extraperception- information that originates in themind
ratherthan with externalstimuliand this knowledgeis modified throughcontinuousdirect
and indirectexperienceand unsupportedinferentialinformation- then how can aGIS hope
to model such complex forms of reasoning? How then can these complexities ofhuman
decision-makingbecapturedfor prehistoricpeopleand modeledin aGIS?
This is perhaps the root question that pervades all three papers. Certainly, the
ability to simulate differing scenarios within GIS is a major advantage. The ability to
recreate something of the cognitive symbolism ofthe cultural landscape through the re-
creationof viewsheds,for example,takes ussomeway towardacknowledgingthe complex
cognitive aspects influencing spatial decision-making. Furthermore, the experiential
realism ofmultiple media and of virtual reality promises considerable potential to link
human cognitionwith the environmental world contained in a GIS and thereby moves the
interpretation beyond the current two-dimensional representation of the world. A virtual
world is a deliberately designed place (sometimes representing real space, sometimes
abstract conceptual space) in which objects 'live'. through their relationships, under
conditions specified by the world's space they cohabit [13]. This includes the
organization, presentation, and manipulation of multidimensional data and the
representation and experiencesof the qualities and contentsof space. Thuswe may seea
convergenceoftechnologies emerging, linking the virtual to the goo-spatial world ofthe
GIS. Importantly, these immersive technologies in which the full senses such asvision,
hearing, and even touch,smell and balanceprovidean importantexperienceoflandscape.
In these instances the researcher is not a passivebut an active participant. This may take
us a long way toward recreating a semblance of the emotional and cognitive worlds of
prehistoricsocietiesand oflinkingadvancesin post-processualtheorytoGIS andthe study
of movementwithincultural landscapes.
References
[I)Harris, T. M. Sibling time-paths: an examination ofnineteenth century migration to a dockyard town,
OccasionalPaper No.6, 1982,Departmentof Geography,PortsmouthPolytechnic,
[2J Maguire,0,Goodchild,M, and Longley,P. Geographicalln!omuuionSystems.Longman, 1999.
[3] Lock. G. and Harris.T. M. Daneburyrevisited:an Iron Age hillfonin adigital landscape.inA1denderfer,
M. and Maschner.D. G.(eds) Anthropology,Spac;and GeographicIn/ormo.tio';and Analysis. 1996, pp.
214-240,OxfordUniversityPress.
(4) Dreyfuss. S. E.An appraisal of someshortestpath algorithms. OperationsResearch, 17. 1969, pp. 395-
~ .
(5) Dijkstra, E. W. A note on two problems 10connec! with gniphs, Numerische MatMmatiJc. 1959, I,pp.
269-271. . ..
(6)Eastman, J. R. A pushbroomalgorithm for Calculating distances in rastergrids,Auto-Cano9;1989.pp.
288-297.
(7) Lombard, K. and Church,R. L. 'Thegatewayshortest path problem: generating alternative routes for a
corridorlocationproblem,GeographicalSystetils, I, 1993:pp.25-45.
(8) Huber. D. L. and O1urch, R.L TranSmission corridor location modeling, Journalo/Environmelllal
Engineering, 1985, pp. 114-130.
(9) Church, R. L.An inlafacefor exploring spatial alternatives for a corridor.location problem, Computers
and Geosciences 18(8), 1992,pp. 1095-1105.
(10) Goodchild,M. F. An evaluationof lattice solutions10theproblemofcorridorlocation,Environmeruand
Planning A,9. 1977,pp. 727-738.
[lIJCarver, S. J. and Brunsdon,C. F. Vector10raster CQDverUonerrorsand feature complexity:an empirical
study USingsimulateddata, llIlemationaJJournal ofGeograPhicalInfonrllJlionSystems, 8(3), 1994, pp.
261-270.
(12) Lynch, K. Mental Maps, 1960.
[I 3)Jacobson,R.Virtualworlds capture t i l reality, GIS World, Dec1994.36-39.'
124
Beyond the Map
G. Lock (Ed.)
lOS Press, 2000
Land evaluation as predictive modelling:
a deductive approach
Hans Kamennans
Faculty ofArchaeology
Leiden University
Abstract: Inductive predictive modelling is widely used in cultural research
management to predict archaeological site location. In most cases this is done on the
basis of a simple correlation between known archaeological sites and attributes from the
current physical landscape. Inductive predictive modelling has been met by severe
criticism. Not only the practice of equating correlation with causality, but also the fact
that distorting factors have not been taken into account and that in most cases no
distinction has been made between archaeological time periods, economic systems and
the difference in types of find spots.
Land evaluation as deductive predictive modelling shows a more promising
approach to predicting archaeological site location. Originally from soil science, land
evaluation generates different models for land use on the basis of ecological and socio-
economic data. Confronted with the archaeological data it predicts activity areas. In this
article data from a long term international survey project in the Agro Pontino (Lazio,
Italy) is used to illustrate and evaluate both approaches.
1. Introduction
As a result of modem developments the European landscape is changing so rapidly that in
many areas the archaeological record is under heavy threat. Recently a number of European
countries signed the treaty of Valletta (Malta) to protect the European archaeological heritage.
In order to guide future developments in the landscape, the government wants archaeologists
to identify areas with a 'high risk' of archaeological finds in order to protect the archaeology
that is left, and to record and study what is in threat of being destroyed. In order to do this
archaeologists have to reconstruct or, in post-processual terms construct, the original spatial
patterning of the material culture of the past. One way of doing this is by means of predictive
modelling.
1.1. Predictive modelling
Predictive modelling is a technique used to predict archaeological site locations in a region on
the basis of observed patterns or assumptions about human behaviour [I, 2, 3]. There are two
different approaches to predictive modelling - inductive or deductive.
With the inductive approach a model is constructed based on the correlation between
known archaeological sites and attributes from (mostly) the current physical landscape. On the
basis of correlation, causality is assumed, and the model is then used to predict unknown site
location. These predictions in tum can be used for planning purposes. Often external expert
125
Kamermans/ Landevaluationaspredictivemodelling
knowledge is used to evaluate and adjust the models. In the deductive approach, which is
much more rarely used, the model is constructed on the basis of aprioriknowledge (social,
mainly anthropological, historical and archaeological) and the known sites are then used to
evaluate themodel. An example of thedeductive approach is the technique of archaeological
land evaluation [4], demonstrated in a hybrid application by Verhagen [5,6].
Archaeological predictive modelling was first developed in the USA during the second
half of the1970s and theearly half of the1980s (see overviews in [1] and [3]). It evolved from
govemmentalland management projects and gave rise to considerable academic debate [7, 8].
At first theemphasis was on thestatistics to evaluate the correlation between parameters and
thephysical landscape [9, 10], although later researchers have tried to incorporate social data
into their applications [II, 12].
Ken Kvamme introduced inductive predictive modelling into the Netherlands [13] and
since then it has been widely used for Cultural Resource Management (CRM) purposes [14], .
although not without criticism [15, 16, 17, 18]. It seems that archaeological predictive
modelling has not gained a strong foothold in other parts of Europe.
This article will demonstrate the two approaches of predictive modelling using data
collected in the Agro Pontino (Lazio, Italy).
1.21heAgroPontinosurvey
The Agro Ponti no is a coastal plain c. 80 km south-east of Rome. It is 60 km long and 15 km
wide and surrounded by the Monti Lepini, the Monti Ausoni, the Tyrrhenian Sea and the low
foothills of the Colli Albani (Figure 1).
. , I'.
I I .-1
I!I!!I- ~ ...-
Figure I. The Agro Pontino.
126
Beyond the Map
Between 1966and1984theLaboratoryforPhysicalGeographyand SoilScienceof
theUniversityofAmsterdam.theNetherlands,hadaresearchprojectinSouthernLazio[19].
During a soil survey in the Agro Pontino in1978 archaeologicalmaterial was encountered[20]
andin 1979theInstituutvoorPre- enProtohistorischeArcheologieAlbertEggesvanGiffen
(IPP)oftheUniversityofAmsterdamstartedastudyprojectintheAgroPontinodirectedby
S.H.Loving,A.Voorripsand myself[21.22]. Duringthenexttenyearstheprojectcarriedout
sevensurveys:three smalloneswithtwotofourpeople(1979. 1980and1989)andfourlarger
oneswithacrewuptotwentyscholarsandstudents(1982, 1984.1986and 1988).Duringthe
sametime(1980 and 1981)material forpalynologicalresearchwascolIected.Themethods
usedandpreliminaryresultswerepublishedin1991[22]andvariousotherpublications[23].
[24].[25],[26].ThetwomainresearchthemesoftheprojectwerethetransitionfromMiddle
toUpperPalaeolithic [27], [28], [29]. andtheapplicationoflandevaluation inarchaeology
[4],[30], [31]. [32].
The total area ofthe Agro Pontino is approximately750km
2
,andalthoughacomplete
surveywasplannedinitially,itbecameapparentthatneitherthemoneynorthepersonpower
wouldbeavailable tosurveytheentire area.so it wasdecided tosample [33]. Inthethen
(1982) prevailing traditionofprocessualarchaeology.amultistageapproach [34] wasused
consistingofthreestages:anexploratoryphase,aprobabilisticphaseandaproblem-oriented
phase. withtheresultsofonephase beingusedformakingdecisions about thenext phase.
Duringtheexploratoryphasefieldsweresurveyedinordertoprovideanimpression ofthe
presentarchaeologicalmaterialandthedistributionoverthedifferentsoilunitsasdefinedby
thesoil survey.Theinformationcollectedfromthefieldswasused todevelop methodsfor
assessing factors affecting visibility [35] andto estimate the size of a randomly selected
samplerequiredtomakestatementsaboutthepopulationoffieldsintheentireAgroPontino.
Fortheprobabilitysamplingphaseasystematicnon-alignedtransectdesignwasselectedto
ensure asufficient samplesizefor making probabilitystatements about the archaeological
populationsintheAgroPontinoasawhole.Duringtheproblem-orientedphase, additional
materials werecollectedtohelp accomplishspecificresearchgoals. Together,allphases of the
surveyyieldedatotalof360findspots;mostofthem,289,weremulti-period.
FortheapplicationofpredictivemodelIingintheAgroPontino,findspotdensityper
landunitwasusedinsteadofartefactdensity,althoughFoley[36]convincinglydemonstrates
thatartefactdensityshouldbegivenpreference.Thiswasnotpossibleinthe AgroPontino.
however,becausetouseartefactdensitythedateofindividualartefacts isneededandhere
manyareundiagnosticandfromwidelydatedassemblages.
2.The inductive approach
Much has been written about inductive predictive modelIing [3].The inductive approach
constructs amodel basedonthecorrelationbetweenknownarchaeological sitesandattributes
from(mostly) thecurrentphysicallandscape.Oftenreadilyavailable maps of thephysical
landscapeareused, such as soil maps or geologicalmaps. Ingeneral, all known archaeological
sites,regardlessofdateorfunction,areusedtosearchforacorrelationbetweenfindspotsand
landscape.Variousstatisticshavebeenusedtotestthiscorrelation[I],[3].
FortheAgroPontinostudythefolIowingmapswereused:thegeologicalmapsofFrosinone,
Latina and Terraeina (1961-66). scale 1:25.000;a geomorphological map constructedby Segre
[37];asoilmapandacombinationmap,bothfromSevinket al. [38].
Thereare II landunitsonthegeologicalmap.Table I shows theareaofeachunit,
togetherwiththeobservedandexpectednumberoffindspotsperlandunitifthefindspots
127
Kamermans / Land evaluation as predictive modelling
hadbeenrandomlydistributedoverthelandscape'.Totestthespatialrelationshipbetweenthe
findspotsandthelandunits,theAttwell-Fletchertestwasused[39]. [40].[31], [41].[17].
Table I.Attwell-F1etcher test tocomparethefind spot densityand the geology oftheAgro Pontino.
Numberofsites;338. number ofcategories; It.number.ofsimulations;200.
95th percentile;0.43 0.059. 5th percentile;0.000.000.
landUDit
nwnberor
sites
""peeled
DrooortiOIlS
observed
DroDOrtions
category
wei2ht
Holocenedunes 1 0.02 0.00 0.04
HolocenedrainedIagoon 0 0.00 0.00 0.00
Holocenepeat deposits 92 0.27 0.27 0.21
Holocenehuman influenceddeposits 26 0.07 0.08 0.25
Holocenedepositswithshells 44 0.17 0.13 0.16
UooerPleistoceneold dunes 171 0.43. 0.51 0.25
UDDerPleistocenetravertine deposits 0 0.01 0.00 0.00
Upper Pleistocenetravertine deposits
withbones and shells 4 0.03 0.01 0.10
Pleistocenevolcanic deoosits(sandv) O' 0.00 0.00 '0.00
Pleistocenevolcanic deoosits(zrev) 0 0.01 0.00
.'
0.00
Karstic depressions 0 0.01 0.00 0.00 .
Thevery conservativeAnwell-Fletchertest showsnocorrelationbetweenfindspot densityand
landunits;nolandunithasahighercategoryweightthanthetluesholdof0.43 0.059.Using
the chi-squared test is not valid because more than 20% of the cells have an expected
frequencyof less than 5and one cell has an expected frequencyof less thanI. Consequently,
Ihaveused an arbitrarythresholdof 3 for (O-E)2/Eto assume apositiveornegativecorrelation
between find spot density and land units. As shown in Table 2 and Figure 2, the upper .
Pleistoceneolddunesshowahigherthanexpected density,whiletheHolocenedunes, the
HolocenedepositswithshellsandthePleistocenevolcanicdeposits(grey)showalowerthan
expectedsitedensity.
Thegeomorphologicalmapconsistsof7landunits.Againtheuseofthechi-squared
testisnotvalidandthe Attwell-fletchertestshowsnocorrelation(Table3).althoughusing
the(O-E)2/Ethresholdof3,landunit I hasahigherthanexpecteddensityandlandunit 3a
lowerthanexpected(Table4andFigure3).
Table 2.Comparison between the lind spotdensityand thegeologyoftheAgro Pontino.'
laud UDit
area prec. osites E sites (OE)
SIE
Holocenedunes 13..9125 1.74 1 5.87 4.0366
HolocenedrainedIazoon 2.4725 0.31 0 1.04 1.0425
Holoceneoeat deoosits 213.225 26.60 92 89.90 0.0488
Holocenehuman influenced deposits 52.7475 6.58 26 22.24 0.6354
Holocenedeoositswithshells 137.63 17.17 44 .58.D3 3.3924
UonerPleistoceneolddunes 341.3875 42.59 171 143.94 5.0856
UooerPleistocenetravertine deposits 5.8925 0.74 0 2.48 2.4845
Upper Pleistocenetravertinedeposits
withbones and shells 20.6525 2.58 4 8.71 . 2.5454
Pleistocenevolcanic denosits (sandy) 0.305 0.04 0 0.13 0.1286
Pleistocenevolcanicdeoosits <JUeY) . 9.43 1.18 0 3.98. 3.9761
Karstic depressions 3.97 0.50 0 1.67 1.6739
801.6250. 100 3.3.8 338 25.0499-
128
Beyond the Map
o nodata
as expected
lessthanexpected
morethan expected
Figure2.Inductivepredictivemodelling. Comparisonbetween thefind spot density
and the geology of theAgro Pontino.
Table3.Attwell-Fletchertest tocomparethe findspot density andthegeomorphologyof the Agro Pontino.
Number ofsites =335, number ofcategories=7, number ofsimulations=200.
95th percentile= 0.63 0.113, 5th percentile= 0.00 0.000.
land unit
numberor
sites
expeded
OrooortiollS
observed
OrooortiOIlS
calegory
weint
Beach ridges 92 0.19 0.27 0.30
I Lagoonal deposits 178 0.54 0.53 0.20
Basin andvalley fills 56 0.23 0.17 0.15
Alluvial fans 0 0.01 0.00 0.00
TravertinedeoosilS 4 0.01 0.01 0.17
Volcanicdeoosits 5 0.02 0.01 0.17
Limestonerocks 0 0.00 0.00 0.00
Table4. Comparisonbetween findspot density and thegeomorphologyof the Agro Pontine.
land unit area orec. osites Esites (o-E)SIE
Beach ridzes 140.7025 18.98 92 63.57 12.7097
Lae:oonaldeposits 399.305 53.86 178 180.42 0.0325
Basin and vallev fills 171.24 23.10 56 77.37 5.9036
Alluvial fans 5.8425 0.79 0 2.64 2.6398
Travertinedeposits 10.51 1.42 4 4.75 0.1181
VolcanicdeoosilS 13.5675 1.83 5 6.13 0.2084
Limestonerocks 0.2525 0.03 0 0.11 0.1141
741.42 100 335 335 21.7262
129
Kamermans I Land evaluation as predictive modelung
DnOdala
asexpected
lessthan expected
more thanexpected
Figure3. Inductivepredictivemodelling. Comparisonbetweenthe findspot density
andthe geomorphologyof the Agro Pontino.
The soil map has 20 different soil types.The threshold of 3 gives a higher than expected
densityforthe HaplicPhaeozems,GleysolLuvisolsandChromicLuvisolsandalowerthan
expected density for the EutricHistosols,CalcaricRegosols, Solodic Planosols andVertic
Luvisols (fable5andFigure4).
Table5. Comparisonbetween thefindspotdensity and soil typein theAgro Pontine.
soUtvne area orec. osites E sites CO-E) SIE
Eutric Histosol 41.8175 5.64 9 18.89 5.1816
l l i ~ Vertisol 0.5075 0.D7 0 0.23 0.2293
ChromicVertisol 138.7425 18.71 61 62.69 0.0455
CalcaricF1uvisol 0.7575 0.10 0 0.34 0.3423
Eutric Fluvisol 21.22 2.86 11 9.59 0.2080
Mollie Glevsol 1.615 0.22 0 0.73 0.7297
Calcaric Glevsol 21.925 2.96 11 9.91 0.1207
Eulric Glevsol 34.0075 4.59 18 15.37 0.4516
Carnbic Arenosol 30.435 4.10 17 13.75 0.7673
CalcaricRezcsol 18.6675 2.52 I .8.43 6.5532
Eutric Rezosol 7.3625 0.99 4 3.33 0.1363
Rendzina 3.9375 0.53 I 1.78 0.3412
SolodicPlanosol 60.0425 8.10 16 27.13 4.5656
Hanlic Phaeozem 36.2775 4.89 28 16.39 8.2212
Glevic Luvisol 68.02 9.17 43 30.73 4.8955
Vertic Luvisol 91.6775 12.37 28 41.42 4.3498
ChromicLuvisol 86.315 11.64 53 39.00 5.0255
Orthic Luvisol 3.945 a.53 0 1.78 1.7825
G1evicCambisol 73.895 9.97 34 33.39 0.0112
othersoils 0.2525 0.03 0 0.11 0.1141
741.42 100 335 535 44.0"720
130
131
Kamermans / Land evaluation as predictive modelling
Beyond the Map
o nodata
III as expected
III less than expected
II morethanexpected
Figure 4. Inductive predictive modelling. Comparison between the find spot density
and soil typein the Agro Pontino.
If attributes from the physical landscape were predicting archaeological site location the
predictive modelling based on the three different maps of physical characteristics of the Agro
Pontino would produce the same pattern or at least show the same trends. This is, however,
not the case. The prediction based on the geological map contradicts that based on the
geomorphological map in almost 57% of the area of the Agro Pontino (Figure 5). This pattern
is the same for the other combinations, the predictions contradict each other in around 50%
of the area.
Modem applications of inductive predictive modelling use a combination of variables
of the current landscape to predict archaeological site location, for example, the map published
by Sevink et aI. [38J. The different land units are based on a combination of textural classes,
drainage classes, slope classes, deposit type, soil type and some special features like the
admixture of tuff. Table 6 gives a cross tabulation between all known archaeological find spots
in the Agro Pontino and the land units from the combination map.
Again, using the arbitrary threshold of 3 for (O-E)2/E, the following soil types have a
higher than expected frequency for archaeological find spots: the association of Haptic
Phaeozems and Cambic Arenosols on the Early Holocene to WUnnian sands (c4); the Chromic
Luvisols on the Borgo Ermada beach ridge (dl and d3); the association of Gleyic Cambisols
and Gleyic Luvisols on the Borgo Ermada lagoonal and circum-lagoonal deposits (e3); the
Chromic Luvisols on the Minturno beach ridge and aeolian sands (fl and 0); the Gleyic
Luvisols on the Mintumo lagoonal and circum-lagoonal deposits (g2); the Gleyic Luvisols on
the Latina lagoonal deposits (j I); and the Orthic to Chromic Luvisols on the volcanic rocks
(rl). The following soils have a less than expected frequency of archaeological find spots: the
Calcaric Regosols on the Terraeina beach ridge (a1); the Eutric Histosols on the Terracina
lagoonal deposits (bt); the Solodic Planosols on the Latina lagoonal deposits (j2); and the
Chromic Luvisols on the fluvio-colluvial basin fill (m5).
Dnodata
lIIIas expected
less than expected
more lhan expected
III contradktion
Figure 5. The contradiction between the prediction based on lhe geological map and that
based on lhe geomorphological map in the Agro Ponti no.
o nodata
as expected
leISthan expected
more than expected
Figure 6. Inductive predictive modelling. Comparison between fwd spot density
andthe combination map of the Agro PontillO.
Figure 6 visualises this, and, again. it can be seen that a different picture emerges with around
50% of the area in this prediction contradicting the others. Only the predictions based on the
132 133
Beyond the Map
Table 6. Comparison between find spot density and the combination map of the Agro Pontino.
1aad
unit
area prec. o sites E sites (O-E)
SIE
al 18.6675 2.52 I 8.43 6.5532
bl 40.0400 5.40 9 18.09 4.5687
b2 1.7775 0.24 0 0.80 0.8031
b3 0.7575 0.10 0 0.34 0.3423
b4 6.1375 0.83 4 2.77 0.5428
b5 1.6150 0.22 0 0.73 0.7297
b6 20.7500 2.80 11 9.38 0.2814
b7 34.0075 4.59 18 15.37 0.4516
b8 3.7600 0.51 I 1.70 0.2875
b9 0.1775 0.02 0 0.08 0.0802
blO 34.8350 4.70 15 15.74 0.0348
cl 7.3625 0.99 4 3.33 0.1363
c3 27.2475 3.68 14 12.31 0.2316
c4 33.1825 4.48 25 14.99 6.6791
c5 3.0950 0.42 3 1.40 1.8342
dl 7.4500 1.00 7 3.37 3.9228
d2 3.9450 0.53 0 1.78 1.7825
d3 5.4725 0.74 6 2.47 5.0318
el 58.4125 7.88 33 26.39 1.6540
e2 6.7125 0.91 4 3.03 0.3083
e3 3.0400 0.41 4 1.37 5.0220
e4 7.2825 0.98 3 3.29 0.0256
e5 16.2100 2.19 8 7.32 0.0623
e6 19.7000 2.66 5 8.90 1.7098
n 7.1075 0.% 6 3.21 2.4214
f2 3.1875 0.43 3 1.44 1.6892
f3 20.6075 2.78 22 9.31 17.2916
,,:1 5.9825 0.81 3 2.70 0.0326
~ 2.3225 0.31 4 1.05 8.2964
~ 5.7425 0.77 I 2.59 0.9801
hi 3.3775 0.46 I 1.53 0.1814
il 59.7150 8.05 36 26.98 3.0145
i2 52.7600 7.12 13 23.84 4.9281
i3 17.0600 2.30 6 7.71 0.3786
j4 0.5075 0.07 0 0.23 0.2293
11 15.0825 2.03 7 6.81 0.0050
12 4.5300 0.61 0 2.05 2.0468
ml 1.1750 0.16 0 0.53 0.5309
m2 47.9325 6.46 18 21.66 0.6177
m3 48.6325 6.56 17 21.97 1.1259
m4 12.4850 1.68 5 5.64 0.0729
m5 34.9575 4.71 8 15.80 3.8469
m6 3.6575 0.49 0 1.65 1.6526
m7 2.5175 0.34 I 1.14 0.0166
n2 5.3675 0.72 0 2.43 2.4252
n4 0.4750 0.06 0 0.21 0.2146
pI 0.2700 0.04 0 0.12 0.1220
al 5.6875 0.77 2 2.57 0.1263
a2 4.8225 0.65 2 2.18 0.0147
rl 1.9750 0.27 3 0.89 4.9778
r2 11.5925 1.56 2 5.24 2.0016
s 0.2525 0.03 0 0.11 0.1141
741.42 100 335 335 102.4306
Kamermans / Land evaluationas predictive modelling
o no data
...expected
less than expected
- more than expected
II contradiction
Figure 7. The contradiction between the prediction based on thesoil mapand that
based on the combination map in theAgro Pontine.
soil map (Figure 4) and on the combination map (Figure 6) look reasonably the same (Figure
7), although this was to be expected since the combination map is based on the soil map (it is,
in fact, an 'improved'version of the soil map).
3. The deductive approach: landevaluation
There are not many examples of deductive predictive modelling. Kamermans et al. [4]
introduced the land evaluation approach in archaeology as a framework for integrating
palaeoecological data and analysing archaeological data to study the relationship between.
humans and their natural environment. This technique, generates different models for land use
on the basis of ecological and socio-economic data. These models are then confronted with the
archaeological database in order to identify, or predict, activity areas [31], [32], [4], [30].
Other examples of the use of land evaluation as predictive modelling in archaeology are by
Boerma [42] and Finke et al. [43].
Land evaluation is a technique developed by soil scientists and widely used in the
19708 and 19808 by the FAO (Food and Agriculture Organisation of the United Nations) in
Third World countries for estimating the potential of land for alternative kinds of use [44],
[45], [46], [47], [48]. The definition of Land Evaluation is: "the process of collating and
interpreting basic inventories of soil, vegetation, climate and other aspects of land in order to
identify and make a first comparison of promising land use alternatives in simple
socio-economic terms" [44:7]. The basic feature otland evaluation is the comparison of the.
requirements of land use with the resources offered by the land, it requires information from
three sources: land, land use, and economics [47].
There are some important differences between theapplications by the FAO and using
134
Beyond the Map
land evaluation in archaeology (Figure 8). Firstly, it is, ofcourse, impossible to measure
prehistoricland qualitydirectly;it has to be reconstructedfromdataobtainedbysurveysof
recentland characteristics.Secondly,theeconomicand socialanalysis of the approachasused
inphysicalgeographyhas tobereplacedbymodelsofprehistoricsocio-economicsituations.
To construct these models, information on the ecological and technical requirements of
different kinds ofland use as well as data on the economic and social contexthas to be
generated from ethnographic, archaeological, and historical sources. The outcome is an
expectedform oflanduse for everychosensocio-economicmodel.Thirdly, the purposeof
using thelandevaluationapproachinarchaeologyistoevaluateour modelsandnotthecurrent
real-world. The comparison ofthe expected form ofland use with the archaeologically
recordedlanduse providesabasisfor modifyingthe models.We repeatthis procedureuntil
the outcomefits bestwiththe archaeologicalrecord.
In the applicationofland evaluationthere are five steps to be considered(Figure8):
Step l. An inventoryofthe naturalenvironmentcollected by field surveys and reviews.
These data form the basis forareconstructionofthenatural environmentatdifferenttimes
in the past.
5
3
2
4
step
archaeology soilscience
Figun::8. The fivesteps inland evaluation.
135
KamermansflAnd evaluationas predictivemodelling
Step 2. Theconstruction ofsocio-economic models for early forms ofland use from
ethnographic,historicand archaeologicaldata.
Step 3. Theclassification ofthe area into different land mapping units on the basis of
physical factors. These units are described in terms oftheir properties to provide a
qualitativelandclassification.
Step4. A semi-quantitativeland classification:Themeasurementofthe suitabilityofan
areafor a certain typeof land use on the basis of the requirements for that type of land use.
Step5.An expectedformoflanduse for everychosensocio-economicmodelbasedon
resultsfrom steps 2through5.Thecomparisonoftheexpectedform oflandusewith the
archaeologicallyrecordedland useprovidesabasisfor modifyingthemodel andrepeating
steps 2through5.
3.1 Step 1:the physical and bioticfactors
Geologicallyspeaking, the Agro Pontino consistsoftwo parts,alow lying graben, mainly
filled withpeat, and adune areaalongthecoast,bothdating from theQuaternary.Sevink'ssoil
surveys distinguished four marineterraces along the coast: the Latinalevel at c. 25 m.a.s.
(metresabovesea level), the Minturnolevel at c. 16m.a.s., the BorgoErmadalevel at c. 6
m.a.s, and the youngest,still activemarinecomplex,theTerracinalevel. Estimateddatesare
560,000 BP (Tyrrhenian I), 125,000 BP (Tyrrhenian ll), 90,000 BP (Tyrrhenian III), and
post-glacial(pre-Neolithic).Eachterraceconsistsofasandybeach ridge and aclayey lagoon.
The graben is partlycoveredwith Holocenealluvial and slope deposits [38). The calcaric
mountainsalongthe north ofthe Agro Pontino, the Monti Lepiniand the MontiAusoni, and
MonteCirceointhe extremesouthan isolatedpartofthe Apennines,were formedduringthe
Mesozoic. In the northern part ofthe area are volcanic rocks, a result ofvolcanic activity
700,000until 10,000years ago [37), [49).
Palynological research of a nine metre sediment core from the graben area near
Mezzaluna, close to the fault along the Monti Lepini, provided a well dated record of
environmentalchangefrom the FullGlacial torecent times [50). [51].Thepollendiagram and
the interpretationofthe core aspresentedbyHunt and Eisner[52) is not entirelycorrectand
a revision is in progress [53]. Duringthe Full Glacial the landscape was dominatedby dry
grass-steppe withsome Quercus and Pinus. In the Late Pleistocene all arboreal elements
disappeared.TheEarly Holoceneischaracterisedbyarapidexpansionof theforest vegetation,
and the disappearanceoftypical steppe taxa. Laterthevegetationchangedfrom analdercarr
and woodlandswampvegetation to a dry'bracken Woodland. The Late Holocenesees the
return ofthe wet aldercarrwithdry Mediterraneanvegetationelementsas well as possible
humaninfluenceon the landscape.
Figure9showsareconstructionofthe Agro Ponti no vegetationduringthe Neolithic
with the dominanttype on the beach ridges and theeoliansedimentsbeing adunevegetation
ofPinus, PoaceaeandPistacia. DryopteriswasgrowingintheTerracinainlandlagoonal area.
The tuffand the alluviuaVcolluvialdepositssupporttwo types ofvegetationalassemblages:
Quercusilex forest and macchia-garriguewith Vitissp., causedbydeforestationofevergreen
woods.On the travertinedepositsCarpinus betulusandPistaciawere growing.TheTerracina
coastallagoon was an open lagoon and the othercoastallagoonal depositshad a dry open
vegetation. On the mountainslopeswas amixedoak forestbelt.
3.2 Step 2:the socio-economicfactors
In ordertomodel land utilisationlYPesfor agriculturalsocietiesinthepast, itisfirst necessary
to identifythe variablesuseful for describingland use and to list the differentlYPeS of
136
Beyond the Map
o nodata
III pine,grasses,Ptstacia
lagoon
Cern
III dry, openvegetation
III Quf!rt:W ilex or macchia
beechandPistacia
Figure 9. Reconstruction ofthe vegetationduringthe Neolilhic in the Agro Pontine.
agriculturalland use. Forthis there are three sources: presentday land use, ethnographic
sources, and historical sources. It is necessaryto selectfrom the identifiedvariables those
which can be used toreconstructpastland use.
3.2.1 Land utilisation types
It iscustomarytouse the FAOterminologytodefinepresentday land use. Beekintroduced
theconceptoftheland utilisationtype(LUT) in 1972[54] todescribeaspecific kind ofland
use and the FAOadoptedit inA framework for land evaluation [55]. The key attributesof
LUIsareproduce,labour, capital,management,technology,andthe scaleofoperations[46].
Rutheberg[56] providesausefulclassificationofpresentdaycultivationsystems:
1. basedon typeofrotation: fallow system; ley system; field system;systemwith perennial
crop.
2. based on intensityofrotation: shiftingcultivation;semi-permanentcultivation/stationary
cultivationwith fallowing;permanentfarming.
3. basedon watersupply: irrigationfarming; rainfedfarming.
4. basedoncroppingpatternand animal activities:differentmain crops;differentlivestock
activities.
5. basedon implementsused forcultivation: pre-technicalmethods(no implements;fire and
zero tillage;cattle treading;planting/diggingstick); hoefarming orspade farming; farming
with ploughsand animaltraction; farming with ploughsand tractors.
6. basedon degreeofcommercialisation: subsistencefarms; partly commercialisedfarms;
semi-commercialisedfarms; highly commercialisedfarms.
137
Kamermans / Land evaluation as predictive modelling
A good starting point for developing socio-economic models of prehistoric agricultural
societies from ethnographic examples is provided by Wolf. In his book Peasants [57], he
defines ecotypeasthesystem ofenergytransfers fromenvironmenttomanand foragricultural
societieshemakes adistinctionbetweenpalaeotechnicandneoteehnic ecotypes. The first type.
palaeotechnic, is based mainlyon human and animal organisms, and the second relies on
"energy supplied by combustible fuels and skills supplied by science" [57:19]: For our
purposesonly thepalaeotechnictype isrelevant.
Wolfsclassificationofthe palaeotechnicpeasantecotypesisbasedon:
1.the degreeof use of agiven pieceofland over time.
2. theamountofland used.
3.the labourrequirement.
4. the degreeto which occupancyofapieceof land requiresagiven input of labour.
On the basis of these criteriaWolfdistinguishesfive ethnographicecotypes:
1. Long-termfallowingsystemsorswidden systems. Fields are clearedbyfireand cultivated
with the hoe. They are used upto apointofdecreasingyields, thenabandoned.to regain
fertility. .
2. Sectorialfallowingsystems. The landisdividedinto two or moresections.each isplanted
fortwotothree years, and then leftfallow for three or four. The dominanttool isthe hoe
or diggingstick.
3. Short-termfallowingsystems(Eurasiangrain farming). The land iscultivatedfor one or
two years, and reoccupiedafter one yearof regeneration.The dominanttool istheplough:
4. Permanentcultivationor hydraulic systems. Land usesystems with technicalarrangements
for apermanentwatersupply. The dominanttools arehoe and plough.
5. Permanentcultivationof favoured plots. (mostlyrainfed)combinedwith the sporadicuse
of irrigated,otherwiseunpromising,plots in thehinterland(infield-outfieldsystem).
Wolfsecotypescan beconsideredthesameastheFAO's land utilisationtypes Uffs[55] and
arecovered byRutheberg's sixcriteria fortheclassificationofcultivationsystems. Wolfstype
1 is a fallow system with shifting cultivation, rainfed water supply, and the use of a
planting/digging stick. Type 2 isalso a fallow system, but stationary cultivation w.ith
fallowing, itisrainfed farming andtheimplementsusedforcultivationare thehoeordigging
stick.Type 3hasthesame criteria astype2exceptfortheimplements. Type4isafield system
with permanentfarming andirrigation. Type 5isalso afield systemwith permanentfanning,
but with arainfedwatersupplyand Wolfgives no informationabout the implements used.
Also of interestisWhite'suse,amongothers,of historicalsourcesfor reconstructing'
agriculture in the Roman world [58], [59], [60]. It has also been suggested that duringthe
Italian Bronze Age two LUTs were common; the 'a campi ed 'erba' system involving
abandoninglandforsometimeafterharvesting toregainfertility, andthe 'debbio'systemusing
swidden,orslashand bum,[61].
3.2.2 Archaeological applications
Itseems that archaeologistscan usemostof theattributes withinthedefinitionsofLUTtypes
to model them for pre- and proto-historicagricultural societies. The physicalattributes are
measurable from the presentday landscape requiringdataaboutgeology, geomorphology,
soils, and climaticconditionsand vegetationin"thepast. Assumptionswill have to be made
139 .
138
Beyond the Map Kamermans/ Land evaluation as predictive modelling
about the technological andsocio-economicaspects,suchasthedegree ofcommercialisation.
Descriptions of present day land use, ethnographicdescriptions such as Wolfs, or historic
descriptionscan be used toguideour modelling.
For step 2 of the land evaluation, five socio-economic models for agriculture are
considered, mainly based onWolf[57]: shiftingcultivation with longterm fallow; shifting
cultivationwith asectorial fallowing system; ashort-termfallowing system; and two systems
of permanentcultivation,Type I with or withoutsome relianceon a hydraulic system, and
Type 2with the cultivationoffavoured plots.
3.3. Steps 3and 4:land classification
Inorder toclassify andcompare itisnecessary todivide. Thedefinitionofland units forland
evaluationofthe Agro Pontinois done on the basis of Sevinksenhanced soil map [38] in
which two variableswere used: physiographyand time. The variabilityinthe landscapeisa
result ofdifferencesinsedimentand relief, factors which arealso responsiblefordifferences
in flora and.fauna. Timeis importantbecausesome land units simplydid not existduring
certain time periods.Figure 10shows the 16differentland units defined for the application
oflandevaluation.
3.3.1 Qualitative land classification
Qualitativeland classificationdescribesland units interms of landqualities. Brinkmanand
Young[45:16]give a listofland qualitiesrelated tothe productivityofcrops or otherplant
growth. For prehistoric palaeotechnic agricultural societies only a selection of their land
qualitiescan be used:
1.moistureavailability
2. nutrientavailability
3.oxygenavailabilityinthe root zone
4. adequacyoffootholdforroots
5. salinityoralkalinity
6. floodinghazard
7. workabilityofthe land
Themoistureavailabilitydependson the amount of rainfall,potential evapo-transpiration,
available water capacityof the soil and the presence or absenceof irrigation. Sevinksmap
makes a distinction between seven drainage classes. The two highest classes (somewhat
excessivelydrained andexcessivelydrained) can notretain enough water for thecultivation
ofcrops.
Thecurrentsoiltypegivesanindicationofthenutrientavailabilityinthepast.Luvisols
and especially Chromic Luvisols are the most fertile soils. Bad soils in this respect are
Arenosols,Regosolsand Planosols.
Theoxygen availability intherootWile islinked with thedrainageofthesoil. This is
not only indicatedbythedrainage class but alsobythepresenceorabsenceof hydromorphic
properties. Gleysolsorsoils with the prefix Gleyic are and were inthe past not suitablefor
agriculture.
Thetextureofasoilisimportant forenabling plants totakeroot. Sevinkdistinguishes
three textureclasses with a heavy texture hindering rooting. Vertices within a soil are also
important,asyoungplantsareverysensitivetoswellingandshrinking. The salinityofthesoil
Doodala
Terraciaabeachridge
Terracina coastallagoon
IITcnacina"inland lagoon
BorgoErmada beach ridge
BorgoEnnadacoastal lagoon
Borgo Ermada inland Iagooo
I11III MinhJrno beachridge
III MinbJrno coastal lagoon
III Mintumoinlandlagoon
Latinabeachridge
IILatina lagoonal
IIeoliandeposits
IIvolcanicdeposits
III travertine deposits
III alluviaUcolluvial deposits
IIrecentdeposits
Figure 10.The 16different land units in theAgro Pontinodefmed
fortheapplicationoflandevaluation.
wasonlyaproblem inthe landunitsalongthecoast whereflooding wasaproblem inthe low-
lyinggraben.
Table 7 and Figure 11 show the qualitative land classification for palaeoteehnic
farming inthe AgroPontino. Land unit 16,therecentfluvial andfluvio-colluvialdeposits,is
notincluded.
Table 7.Qualitative landclassificationfor palaeotechnic peasant ecotypes.
score:+ = good;- = bad
totalscore: - = notsuitable,+= mediumsuitability,++= suitable.
IaDdlIIIit moisture avallabWty
nutrimt
oxygen
adequacy of
foothold for
roots/workability
salinity tolal score
I - .+ + + -
2 + + - - -
3 + + - - + -
4 + +
.+
+ ++
5 + - + + +
6 + + + +
7 + + + + ++
8 + + - - + +
9 + + - + +
10 + + + + + ++
11 + - - - + -
12 - + + + -
13 + + + + -
14 - + + "" - + +
15 + - - + + -
140
141
Beyond the Map
Kamermans /Land evaluation as predictive modelling
o no data
III not suitable
I11III medium suitabje
I11III suitable
o no data
I11III not suitable
_ medium suitable
.I11III suitable
Figure II. Qualitative land classification for palaeotechnic peasant ecotypes in the Agro Pontino.
3.3.2 Semi-quantitative land classification
Due 10the uncertainties of the reconstructed environmental data, in archaeology only a semi-
quantitative evaluation can be made involving the suitability for a certain land utilisation type
being established for each land unit. This is done using the requirements for the LlIT which
results in the land units being ranked according to their suitability (Table 8 and Figure 12).
Table 8. Semi-quantitative land classification for palaeotechnic farming.
Figure 12. Semi-quantitative land classification for permanent cultivation Type I in tbeAgroPontino.
3.4. Step 5: archaeology and land evaluation
The rank order of the different land units, based on find spot density data collected during the
survey, was compared with the expected rank order of the land units for the different models
(Table 9). The survey didn't collect data from land units 10 and 15.
Two tests (Spearman and Kendall) were used to establish which predicted ranking
Table 9. Comparison of predicted and"observed preferences of fanners during the Neolitllic.
\aDd unit swiddeo systems faDow
permaneot cultivatioo
tvpe I tYDe2
predicted raokiog obserYed .
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
3
3
3
I
2
2
1
2
2
1
3
3
1
2
3
3
3
3
I
1
1
I
I
1
1
3
3
I
2
3
3
3
3
I
2
2
I
2
2
1
3
2
1
2
3
3
3
3
2
3
3
2
3
3
1
3
3
1
3
3
Iaod UBit
1
2
3
4
5
6
7
8
9
II
12
13
14
swidden SYstems
11
11
11
3
6
6
3
6
6
II
11
3
6
fallow
11
11
11
4,5
4,5
4,5
4,5
4,5
4,5
II
II
4,5
8
permaoeot cullivatioo .
type I lYpe2
11,5 7,5
11,5 7,5
u.s 7,5
3 2,5
6,5 7,5
6,5 7,5
3 2,5
6,5 7,5
11,5 7,5
u.s 7,5
6,5 7,5
3 1
6,5 7,5

Neelitbie
11,9
11,9
11,9
2
8
9
.4
11,9
1
6
3
5
7
142
143
Beyond the Map
correlatedbest with theobservedranking.The predictedrank:orderfor permanentcultivation
withor without some reliance on a hydraulic system (permanent cultivation Type I), for
example,correlatessignificantlywith the observedsite densityrankorder.
4.Discussion
Nowadaysarchaeologyplays arole in environmentalplanning,the spatialorganisationof the
presentlandscape,and so predictivemodellinginarchaeologyisahot topic asone ofthetools
that isused tosolve socially relevantproblems.Itiscritical, therefore,that the techniquesused
topredictsite locationare thoroughlytested anddiscussed.ArchaeologistsinvolvedinCRM
and academic archaeologists should search together for the most optimal ways ofdoing
predictivemodelling.
It is not difficultto criticisepredictivemodelling.To startwitha general point,for
example, all predictionsare rathergeneral, based mainlyon proxydata, and the distorting
factors areverydifficulttocontrol. Bothapproaches topredictivemodelling,the inductiveand
the deductive,have theirstrongand their weak points. Inductivepredictivemodellingbased
on observedpatternsbetweencurrentlyknownsite locationand differentattributesfrom the
currentphysical landscapecan producevariableresults(as shownabove).This factis quite
easyto explain.
I. Thereare distortingfactors, often physical(forinstancegeological)that explainwhy some
land units producealessthan expectedfrequency forarchaeologicalfindspots. Thismeans
that a geological map will by definition produce different results to a soil map or a
geomorphological map
2. Differenteconomicsystemshave differentrequirementsofthe landscape.One canexpect
arelationshipbetweensoil type and Neolithicfanningbut it is less obviousbetweensoil
type andPalaeolithichunter-gatherersforwhom arelationshipbetweengeomorphologyand
site locationismore likely. Lumpingallarchaeologicalfind spots togethermeans lumping
differenteconomicsystemstogether.
3. The sameargumentholds for differenttypesof find spots. Habitationsites and burial sites
do not necessarilyhave the samerequirementsofthe landscape.
Withinductivepredictive modelling in general no effortis madeto try and understandthe
culturalorenvironmentalmechanismsthat arecausingcorrelationsor totakedistortingfactors
into accountsufficiently.You cannotput all archaeologicalfind spotsfrom aregion together
inabox witharandomselectionof presentdayphysicalcharacteristicsofthe landscapeand
expectthat the resultis aviablepredictionofarchaeologicalsitelocation.
Mostapplicationsof inductivepredictivemodellingdo not take any social variables
regardingsite locationinto account,forexample,Ellen [62] lists nightmares,social conflict,
illness,and deathasmotivesinvolvedinsite location.Inductivepredictivemodellingisvery
susceptibleto ecologicalfallacyand environmentaldeterminism. Another, and perhapsthe
most serious, disadvantage of inductive predictive modelling is that it uses known
archaeologicalsites tocreatethe model,leavingnodata left to test it!
Deductive predictive modelling also has its disadvantages. The application ofland
evaluationin archaeologyrequiressome(unfortunatelyarguable)assumptions:
I. Past humanexploitationofthe environmentwas accordingtothe principleofleast effort.
2. The combinationofenvironmentand human behaviourcreatesaspecificspatialpatternin
Kamermans I Land evaluation as predictive modelling
particulartypesofareas.
3. Thereis arelationshipbetweenprehistoriclanduse and artefactor site density.
4. The economicsystemduringeacharchaeologicallydistinctperiodwas, broadlyspeaking,
constant.
Furthermoreland evaluationismore time consumingand thus more expensivethan inductive
predictivemodelling.
The advantages ofinductive predictive modelling are that it is cheap and easy to
perform. The advantages ofland evaluation are that it is capable ofincorporating social
variables,that itmakesadistinctionbetweeneconomicsystemsand that itis possibleto test
the modelsusing the archaeologicaldata. .
5.Conclusions
Landevaluation as an archaeological techniquefor predictive modelling works, and looks
more promising for the purpose ofenvironmental planning than the lraditionalforms of.
inductivepredictivemodelling. Itseems tobebestsuitedtosimpletypesofagriculture(which
itwas designedfor), butboththe geographicalareaand thearchaeologicaldata setmust satisfy
a large numberof conditions. The areamust be vast and have considerable environmental and
physiographic variety with good possibilities for collecting palaeoecological data. The
archaeologicaldatamust besufficientinnumberandpreferablybecollectedinathoroughand
controlledway.
Acknowledgements
The Agro Pontino survey project was a joint venture by the Instituut voor Pre- en
ProtohistorischeArcheologieAlbertEgges van Giffen(lPP), ofthe UniversityofAmsterdam
and theInstituutvoorPrehistorieoftheStateUniversityofLeiden(both inThe Netherlands).
Itwas underthe auspicesofthe IstitutoOlandeseinRomeand supportedby researchgrants
from theNetherlandsOrganisationforScientificResearch(NWO).The followinginstitutions
co-operated in the project: Soprintendenza Archeologica del Lazio, the Department of
Anthropology of the Universita di Roma "La Sapienza", the Laboratory for Physical
Geographyand Soil ScienceoftheUniversityofAmsterdam,theMuseoNazionale Preistorico
Etnografico"Luigi Pigorini"inRome,the IstitutoItalianodi PaleontologiaUmanain Rome,
DlGITERgoo-informaticain Romeand the Hugode Vries Laboratoryofthe Universityof
Amsterdam.
Iwouldlike to thankProf. A.P.A.Vinkand Prof. J. Sevinkfrom the Laboratoryfor
PhysicalGeographyand Soil Scienceofthe UniversityofAmsterdamfor their inspirationto
use land evaluationinArchaeologyandProf. Sevinkforhisenthusiasticand ongoingsupport.
OfcourseIwantto thank myco-directorsSusanLovingand Bert Voorripsand all the other
participantsin thesurveys.FinallyIwouldlike tothankGaryLock for correctingmyEnglish
and JoanneMol for assistancewiththe figures.
References
[1)Kohler, T.A.andS.C.Patter. Predictive modelsforan:haeologicaln:sourcelocations.In:M.B. Schiff.,..(ed.),
AdvQflces in Archaeological Method and Theory. Vol9.AcademicPress,NewYork,1986, pp. 397-452.
144 145
the Map
[2) Kvamme,K.L.Development andtesting of quantitative models. In: W.L. Judge and L. Sebastian (eds),
Quantifying the Present and Predicting the Past: Theory, Method and Application of Archaeological
Predictive Modeling. Denver,USBureauofLandManagement, 1988,pp. 324428.
[3) Kvamme, K.L. The fundamental principles and practice of predictive archaeological modeling. In: A.
Voonips(ed.) Mathematics and Information Science In Archaeology: A Flexible Framework. StudiesIn
ModernArchaeology,Vol.3.Holos,Bonn, 1990,pp.257-295.
[4)Kamermans,H.,S.H.Loving& A.Voorrips.ChangingpatternsofprehistoriclanduseintheAgroPontino.
In:C. Malone& S.Stoddart (eds). Papers inItalian Archaeology IV. Part i TheHuman Landscape. BAR
InternationalSeries243, 1985,pp.53-68.
[5)Verhagen,P.,J.McGlade,S.Gili& R.Risch. Some CriteriaforModelling Socio-EconomicActivities inthe
Bronze Age of south-east Spain. In: G. Lock and Z. (eds). Archaeology and. Geographical
Information Systems: A European Perspective, Taylor& Francis,London, 1995,pp. 187-210.
[6]Verhagen.P.The use of GIS as a tool for modellingecological change andhuman occupation in the Middle
Aguas Valley (S.E. Spain). In: H. Kamermans & K. Fennema (eds). llIlerfacing the Past. Computer
Applications and Quanritative Methods inArchaeology CAA95. AnalectaPraehistoricaLeidensia28,1996,
pp.317-324.
[7)Carr,C.lntroductoryremarksonRegionalAnalysis.In:C.Carr(ed.).For Concordance inArchaeological
Analysis. Bridging Da/Q Structure, Quanrila/ive Technique, and Theory. WestportPublishers, KansasCity,
1985,pp.114-127.
[8)Savage, S.H. GIS in archaeological research. In: K.M.S. AIlen,S.W. Green and E.B.W. Zubrow (eds).
IlIlerpreting Space: GIS and archaeology. Taylor& Francis,London,1990,pp.22-32.
[9]Parker,S.Predictive modellingofsitesettlementsystemsusingmultivariatelogistics. InC.Carr(ed), For
Concordance inArchaeological Analysis. Bridging Data Structure, Quantitative Technique, and Theory.
WestportPublishers, KansasCity, 1985,pp. 173-207.
[10) Kvamme,K.L.Determiningempirical relationshipsbetweenthe naturalenvironment andprehistoricsite
locations: a hunter-gatherer example. In C. Carr (ed), For Concordance in Archaeological Analysis.
Bridging Data Structure, Quanritative Technique, and Theory. WestportPublishers,KansasCity,1985,pp.
208-238.
[11] Wheatley, D.W. Between the lines: the roleof GIS-based predictive modelling in the interpretation of
extensivesurveydata. In:H.Kamennans& K.Fennema(eds). IlIlerfacing the Past. Compiaer Applications
and QuanrilaJive Methods inArchaeology CAA95. AnalectaPraehistoricaLeidensia28,1996,pp.275-292.
[12J Z. & K.L. Kvamme. Settlement Pattern Modelling through Boolean Overlays of Social and
Environmental Variables. In:J.A. Barcelo, I.Briz and A. Vila (eds). New Techniques for Old Times -
CAA98. Computer ApplicatiOns and QllQ/ltitativeMethods in Archaeology. BARInternationalSeries757,
1999,pp.231-237.
[13)Brandt,R.,B.1.GroenewoudtandK.L.Kvamme.Anexperimentinarchaeologicalsitelocation:modeling
inthe Netherlands usingGIStechniques.World Archaeology 24,1992,pp.268-282.
[14]Deeben,J,D.P. Hallewas,J.Kolen,R.Wiemer.Beyondthe CrystalBall.PredictiveModellingasaTool
in Archaeological Heritage Management and Occupation History. In: W.1.H. Willems, H. Kars,D.P.
Hallewas(eds)Archaeological Heritage Managemelll in the Netherlands. VanGorcum,Assen, 1997,pp.
76-118.
[15)Leusen,P.M.van.GISandarchaeologicalresourcemanagement:aEuropeanagenda.In:G.LockandZ.
(eds). Archaeology and Geographical Information Systems: A European Perspective, Taylor&
Francis,London, 1995,pp.2741.
[16) Leusen, P.M., van. GIS and Locational Modelingin Dutch Archaeology: A Reviewof Current Approaches.
In:H. D. G. Maschner(ed.) New Methods, Old Problems: Geographic Information Systems in Modem
Archaeological Research, CenterforArchaeologicalInvestigations,Occasional Paper No. 23, 1996,pp.
177-197.
[17)Kamermans. H.& E. Rensink. GISin palaeolithicarchaeology. Acase study fromthe southern Netherlands.
In:L.Dingwall,S.C.Exon,V.L.GalTney,S.Laflin& P.M. vanLeusen(eds).Archaeology inthe Age of
lraemet - CAA97. Computer Applications and Quantitative Methods inArchaeology. BARInternational
Series750,1999,81& CD-ROM.
Kamermans I Landevalu<i/ionas predictive modelling
[18] Kamermans,H.& M.Wansleeben. Predictive modelling inDutch archaeology,joining forces. In:J.A.
Barcelo, I.BrizandA.Vila(eds). New Techniques for Old Times - CAA98. Computer Applications and
Quantitative Methods inArchaeology. BARInternationalSeries757,1999,pp.225-230.
[19] Sevink,J., A.Remmelzwaal & O.C. Spaargaren. The soils of southern Lazio and adjacent Campania.
Amsterdam. 1984.
[20)Sevink,J.,P.Vos,W.E.Westerhoff,A.Stiennan& H.Karnerrnans. Sequenceofmarineterraces nearLatina
(AgroPontino,CentralItaly).Catena 9,1982,pp.361-378.
[21)Voorrips,A.,S.H.Loving& H.Kamermans.AnarchaeologicalsurveyoftheAgroPontino(prov.ofLatina,
Italy). In: D.R. Keller & D.W. Rupp (eds). Archaeological survey in the MediterraMan area. BAR
InternationalSeries 155,1983,pp. 179-181.
[22] Voomps, A., S.H. Loving & H. Karnermans (eds). The Agro Pontino Project. Methods and
preliminary results. StudiesinPrae- enProtohistorie 6.Amsterdam,1991.
[23] Kamermans, H. Survey sampling, right or wrong? In:J. Huggett & N. Ryan (eds), 1995.Computer
Applications and Quanritative Methods in Archaeology 1994. BAR International Series 600,1995, pp_ 123-
126.
[24) Kamennans, H & A. Voorrips. Digging for Data. The Useof Fieldcomputers at IPP Excavations and
Surveys.In:L.H.vanWijngaarden-Bakker(ed.), Database Managemelll and Zooarchaeology. PAC[Vol.
14.CouncilofEurope,Strasbourg, 1986,pp. 109-114.
[25)Voornps,A, S.H.Loving& J.Strackee.Thegamma!'nixdensity:Anewstatisticalmodelforsomeaspects
ofsystemorganization.in:A.Voorrips & S.H.Loving(eds). To Pattern the Past. Pact II,1986,pp. 323-
346.
[26)Voorrips,A, S.H.Loving& H.Kamermans.1nfonnationScienceinArchaeologicalSurvey.in:T.Hackens
& U.Miller(eds),Geology and Palaeoecology for Archaeologists. PAIT24, 1989,pp.189-210.
[27) Loving,S.H. Estimatingthe ageof stoneartifacts usingprobabilities. In: H.Karnermans& K.Fennema
(eds). Interfacing the Past. Computer Applications and Quantitative Methods inArchaeology CAA95.
AnalectaPraehistoricaLeidensia28, 1996,pp.251-261.
[28) Loving,S.H.Discard ofcoresassignedtotheMiddlePaleolithicinthePontineregionofWestCentral Italy.
Quaternaria Nova VI,1996,pp.505-532.
[29)Loving,S.H.,A.Voorrips,C.W.Koot& H.Kamermans. The Pontinianontheplain:Someresultsfromthe
AgroPontinosurvey.Quaternaria Nova 1,1990191, pp.453477.
[30)Kamermans,H.,S.H.Loving& A.Voorrips.Archaeologyandlandevaluationinthe AgroPontino.In:M.
Cherchi(ed.). La valle pomina nell'antichita. Atti del convegno. Roma, 1990,pp.9-11.
[31)Kamermans,H.Archeologie en landevaluatie in de Agro Pontirro(Laslo, Italie). Academischproefscluift.
Amsterdam.1993.
[32)Kamermans, H.TheAgroPontinoSurveyproject:asummaryofthe applicationoflandevaluation.In: S.
Milliken& C.Pereuo(eds), Archaeology, Methodology and tireorganisation of research. Forli,1996,pp.
51-60.
[33]Loving,S.H., H.Kamermans& A.Voorrips.Randomizingourwalks:theAgroPontinosurveysampling
design. In: A. Voorrips, S.H. Loving & H.Kamennans (eds), 199\. The Agro Pontino Project.
Methods and preliminary results. StudiesinPtae- enProtohistorie6.Amsterdam, 1991,pp.61-78.
[34) Redman,C.L.MUltistagefieldworkandanalyticaltechniques.American AlIliquity 38,1973,pp.
[35) Verhoeven, A.A.A Visibilityfactors affectingartifact recoveryin the AgroPontinosurvey.In:A.Voonips,
S.H.Loving& H.Kamermans (eds). The Agro Pontino Project. StudiesinPrse- enProtohistorie6,
Amsterdam. 1991,pp.87-97.
[36) Foley,R Off-site Archaeology and Human Adoption in Eastem Africa. BARInternationalSeries97, 1981.
[37)Segre,A.G.NotasuirilevamentiesguitinelFoglio 158Latinadellacartageologicad'Italia.Bolletino, del
servizio geologico d'ltalia 73, 1957,pp.569-84.
147 146
BeyondtM Map
[38JSevink, J., J.Duivenvoorden& H. Kamerrnans. The soils oftheAgro Pontino. In:A. Voonips,S.H. Loving
& H.Karnermans(005),ThAgro Pontine Survey Project.M.thodsandpreliminaryresults.Studies inPrae-
en Protohistorie6. Amsterdam, 1991, pp. 31-47.
[39] Attwell, M. and M. Fletcher, A new techniquefor investigatingspatial relationships: Significancetesting.
In: A. VoonipsandS. H. Loving (eds) To Pattern the Past, PACT, Conseilde l'Europe,Strasbourg, 1985,
pp. 181-189.
[40] Attwell, M. R. and M. Fletcher, An AnalyticalTechniquefor InvestigatingSpatialRelationships,Journal
ofArchaeologicalScience, 14, 1987, pp. 1-11.
[41] Wansleeben, M. and L.B.M. Verhart. MeuseValley Project: GIS and Site LocationStatistics,Analecta
PraehistoricaLeidensia 25, Leiden, 1995, pp. 99-108.
[42] Boerma, J.A.K. Land evaluationin prehistoricperspective: Some observations. In: O.M.C. Haex, H.H.
Curvers& P.M.M.G. Akkermans(eds). To theEuphratesandbeyond.Rotterdam, 1989, pp. 17-28.
[43] Finke,P., J. Harding,J.Sevink,R. Sewuster& S. Stoddart.The dissectionofa BronzeandearlyIron Age
landscape.Siteterritoriesand landuse. In: C.Malone& S. Stoddart(eds). Territory, time and state. The
Archaeologicaldevelopmentofthe GubioBasin.Cambridge, 1994, pp. 81-93.
[44] Brinkman,R. & A.J. Smyth (eds). Landevaluationforruralpurposes.ILRJ Publ. 17.Wageningen, 1973.
[45] Brinkman,R. & A. Young. AFrameworkforLAndEvaluation. ILRJ Publication22. Wageningen, 1976.
[46] Beek,K.J. LAndevaluationfor agriculturaldevelopment.ILRJ Publication23. Wageningen, 1978.
[47J Dent,D.& A. Young. SoilSurveyand LandEvaluation.London, 1981.
[48] McRae,S.G. & C.P. Burnham.Landevaluation. Oxford, 1981.
[49] Kamermans, H.. Faulted land: the geology of the Agro Pontino. In: A. Voonips, S.H. Loving &
H.Kamermans(eds), 1991. TheAgro Pontino Survey Project.Methods andpreliminaryresults.Stndiesin
Prae- en Protohistorie6. Amsterdam, 1991, pp. 21-30.
[50) Eisner, W., H. Kamermans & S.H. Loving. Resultati preliminari di una ricerca palinologica nell'Agro
Pontine. Atti della XXIVriunione scientifica dell'lstituto Italiano di preistoria eprotostoria nel Lazio.
Firenze, 1984,pp. 207-211.
[51] Eisner, W., H. Kamermans & TJ. Wyrnstra. The Agro Pontinosurvey: results from a first pollencore.
DialoghidiArcheologia2,1986,pp. 145-153.
[52] Hunt,CO & W.R. Eisner. PalynologyoftheMezzalunacore. In:A.Voonips,S.H. Loving & H.Kamermans
(005).TheAgro PontinoSurvey Project.Methods andpreliminaryresults.StudiesinPrae- en Protohistorie
6.Amsterdam, 1991, pp. 49-59.
[53] Eisner, W.R. & H.Karnermans.LateQuaternaryVegetationHistoryofLatina.Italy:AFinal Report, inprep.
[54] Beek, K.J. The conceptof land utilizationtypes. In:Approachesto land classification.FAOSoils Bull. 22,
1972,pp. 103-120.
[55] FAO. Aframeworkfor landevaluation.ILRJ Pub122.Wageningen, 1976.
[56] Rutheberg.Fanningsystems in the tropics. Oxford, 1971.
[57] Wolf, E.R.. Peasants. New Jersey, 1966.
[58] White,K.D. AgriculturalImplementsof theRoman World. Cambridge, 1967.
[59] White, K.D. ABibliographyonRoman Agriculture.Reading, 1970.
(60) White, K.D. Farm EquipmentoftheRomanWorld.Cambridge, 1975.
[61] Ampolo,C. Lecondizionimaterialidellaproduzione.Agricolturae paesaggioagrario.In: La formazione
dellaciuanel Lazio.DwloghiIii Archeologia2, 1980,pp. 15-46.
[62] Ellen, R. Environmefll, subsistenceandsystem. Cambridge,1982.
Beyondthe Map
G. Lock(Ed.)
lOS Press, 2000
Understanding Roman settlement
patterns through multivariate statistics
and predictive modelling.
ZoranStancic '
Tatjana Veljanovski
CentreforScientificResearchofthe
SloveneAcademyofSciencesandArts
Ljubljana
Abstract: Predictivemodels are becomingused increasingly often in archaeological
Cultural Resource Management. In addition to tilis extremely successful and
productiveapplication, predictive models can be an effective tool in archaeological
site locationexplanation. The principal objectiveof this paperis 10 demonstrate an
application of archaeological predictive modelling and to outline the potential of
multivariate statistics for the analysis of settlement patterns. The first part of the
paperprovides a history of predictive modelling, while the second part focuses on
the analysesofRomansettlementpatternson the islandofBraein CentralDalmatia.
After describing the island of Brae and the data used in the analyses, special
attention is givento the locationalanalysisof Romansettlementsites datedfrom the
2nd centuryBC to the 2nd centuryAD. The results of these analysesare then applied
within a predictive model representing the distribution ofRoman settlement sites.
The limitedperformanceof this model is followed by the applicationof more refined
multivariate statistical methods, the results of which suggest more meaningful
clustersofRomansettlement.
1. Introduction and background
Predictive models have a fairly long tradition in American archaeology, having been
intensivelyused forCultural Resource Managementin theUSA. The principal reason behind
their use lies in the fact that, legally,American archaeologists areonly allowed to carry out
archaeological survey and excavationon state or federally owned lands. On the basis of the
limited areas available fordirectstudypredictive models wereused togeneralisesite locations'
over wider areas [1],[2],[3]. Generally, the legal basis underwriting archaeological fieldwork
is different in European countries and allows archaeologiststo carry out work on privately
owned landwith relativeease. '
As a consequenceof this situation archaeological predictivemodels were not utilised
inEurope until the later 1980s,Whilst thedevelopmentof North American predictive models
was mainly related tothepragmatic aspectof modelling (i.e. mapping sensitive areas withina
landscape), within Europe modelling wasdirected towards models which would increaseour
understandingof settlement on the basisof prior knowledge of locational determinants, Van
Leusen noted: "locationalfactors used in the CRM approach relate to factors that have a
significantstatistical correlation withaspecificsetofsites, while the locational choicefactors
148
Beyond the Map
used in the academic approach refers to factors that were involved in the original process of
site selection" (for details see reference [4]).
Several publications attempt to describe what an archaeological predictive model is, or
at least what it should be. Archaeological predictive models are essentially based on the
fundamental assumption that our knowledge of known archaeological sites allows us to
establish which factors influenced their location in the landscape and to use this data in
empirical testing. Prediction is merely the elucidation of settlement 'rules' in a form that
allows us to map locations which conform to the 'conditions' predicted by the model for
settlement. To achieve this we analyse the relationship between the natural and social
environment and archaeological site location. In analysing the natural environment we most
frequently analyse topographic variables (elevation, slope, surface geology, quality of soils,
aspect, climate, existence of water sources, vegetation etc.), The analysis of the social
environment might include those variables which are understood as descriptive of cultural,
religious, ideological or economical relationships with the landscape. The American
archaeologist Julian Steward deserves particular credit for his attempts to incorporate the
human component within predictive modelling. His pioneering research, based on the
theoretical principles of cultural ecology and the link between the cultural and natural
environments, as well as his interest in the characteristics of site location, led to wider interest
in the analysis of settlement patterns within archaeology [5J. Even today Steward's extensive
approach to analysis is a fundamental principle in locational predictive modelling.
One of the goals of locational modeiling is to establish the determinants which
influenced settlement behaviour in the past. Following this, a given archaeological site can be
understood as a reflection of certain decisions and represents the final choice within a certain
system. The selection of a dwelling-place is, then, a complex reflection of such decisions
which can also be influenced by personal preference, in accordance with the potential offered
by the natural environment. The incorporation of a human component within locational
modelling acknowledges the role of subjective judgement, but at the same time models
become more and more cognitive.
It is not within the scope of this paper to discuss in detail theoretical approaches to
predictive modelling. However, two different general approaches can be defined: inductive
and deductive. In the inductive model, one starts with the basic archaeological data and tries
to build some conclusions based on this knowledge. The deductive approach starts with
theoretical knowledge at the synthetic level and tries to deduce relevant conclusions related to
the logic of settlement patterns and land use in the past. More discussion of these issues can
be found elsewhere, for example in Dalla Bona [6] and Kamermans (this volume). The
technological problems when applying any of theabove theoretical approaches also should be
briefly mentioned. Generally speaking, one can apply a number of alternative techniques. The
Boolean overlay of variables for which it has been proved have in some way influenced
location patterns may be utilised along with multivariate statistical regression. Although the
latter approach is very powerful and gives a detailed insight into the relationship between
individual variables analysed, regression analyses can only be used when the number of sites
available for analysis is large.
It seems reasonable to suggest that while each archaeological predictive model is
unique in some way, many of the problems faced by modellers are similar in any environment.
Consequently, much can be learned from the experience gained in implementing a model and
the decisions which were made during the process. The following sections describe such a
process.
149
Z Slana" and T. Yelianovski I Roman settlement pauems
2. Introduction to the study area and database
2.1 General introduction to the island ofBrae
The data set used in this case study is from the island of Brat in Central Dalmatia, Croatia,
The Adriatic Islands Project [7],[8] has conducted field surveys and excavations on several
Central Dalmatian islands including the small island of Palagruza in the centre of the Adriatic
Sea, the islands of Solta, Vis, Hvar, and the island of Brac., the largest island in the region.
The island of Brae has a total swface area of 395 km
2
, and is elliptical in shape, with
the longer axes oriented east-west and measuring around 36 km length and 12 krn in width
(Figure I). The whole of Central Dalmatia is characterised by dramatic relief and Brae is no
exception to this. The highest peak on Brae is Vidova gora, at 778 m above sea level, which is
also the highest peakof all the Adriatic islands. The geology of the island on the other hand is
fairly monotonous. It is essentially comprised of cretaceous limestone and dolomite. The most
important of these, in economic terms, is a thin layer of SennoneIirnestone which was and
still is used for quarrying. Soft Eocene deposits can be found only in small areas on the
southern coast, while quaternary deposits can be found in most valleys and numerous karst
dolinas. The best soils developed on these two geological bases.
The climate of the island is best described as typically' Mediterranean with mild
winters and hot summers. The average summer temperature is 16
0
C, however, the
temperature drops about 0.6
0
C with every 100 metres of elevation. Topographical relief has a
similar impact on precipitation, which is almost exclusively rain. While average rainfall is 799
mm per square metre on the western tip of the island, the highest locations such as Praznica
record up to 1320 mm of precipitation yearly on average.
i>
o 5
Figure 1; The island of ~ modern settlemenlS (named) and the distribution of Roman Senlements,
150
Beyond the Map
The vegetation of the island is characterised by black pine in the higher locations,
while lowerlocationsare oflen covered with red pine or denseMedilerraneanscrub. The best
qualitysoils in alluvialvalleys and dolinasare used for viticultureand agriculture,while many
mild slopes are terraced and used for olive groves. Due to the drastic depopulation ofthe
island which started at the end of the 19th centuryand has continued, many fields have not
been used fordecadesand are overgrownwith Mediterraneanscrub.
The followingphysiographicregionscan be definedon the basis ofthe environmentaldata:
coastlandand low alluvial valleys,
intermediatehills along with the westernplateauand the Nereziskopolje,
higherhills with the uppereasternplateaubetweenPraznicaand Selea.
Itseems that eachof these physiographicregionsattracteddifferenthumanactivity. Coastland
and alluvialvalleysare mostintensivelyused today for agriculture,and itcomesas nosurprise
that the majority ofRoman settlement sites are located there as well. In the Bronze Age,
however, the higher hills with the uppereastern plateau were the focus of human activities.
This microregion is used today mostly for pasture for sheep and goat with some limited
agriculturein smallerkarst dolinas.
2.2 Archaeological and environmental data used inthe study
After the successful completion of field survey projects on the islands of Hvar, Vis and
Palagruza,it was decided to apply the same survey techniques to the island of Brae. The
survey was essentially based on field observation and surface artefact collection. A
considerable numberofarchaeological sites on Brae were excavatedor recorded between the
end ofthe 19thcenturyand between the 19505and 1960s [9]. Unfortunatelythese sites were
ratherpoorly recorded,and locations were frequentlyonly describedor locatedvia toponymic
reference. As a consequencethe actual positionof known archaeological localitiescould not
often be easilydetermined or used in analysis, therefore, all known sites were revisited and
properly recorded. Duringthe 1994 field surveya total of nearly600sites were documented
and input into adatabase.The numberofsites more than doubledthosepreviouslyknown.
Sites from all archaeologicalperiods were recorded, including industrial sites such as
lime kilns or stone quarries. A total of more than 90 Roman sites were recorded, of which
around one third were settlements. In the following analyses only the sites which were
interpreted as settlement sites and which contain evidence of structures dated to the 2nd
centuryBC- 2nd centuryAD were used. Atotal of 29sites fulfilled these conditions.
The environmentaldataused in the analysesisthat of the presentnatural environment.
One of the basic questions faced when the decision is taken to use modem data is how
representative such information can be for past conditions. For some natural data, general
comparison can be made on the basis of comparative analysis of data from other Central
Adriatic Islands.The reliefand geologyhas notchangedsubstantiallysince the end of the last
glaciation. Therefore, it can be claimed that the modem relief and geological data are fair
representationsofthe reliefand geology in the Roman Period. However, this can not be said
for allenvironmentaldata,
Comparative analysis on the mainland [10] suggests that soils have changed
significantly since the Bronze Age, although they were supposed to be fairly stable in the
Roman Period, with further changes in the 19th and 20 century. Theavailable soil data for
Brae was not reliableas the availablesoil map was very poor, the printedmap was atthe scale
151
Z StQllQC and T.Yeljanovski I RonuJIIsettlemen: panems
I :200,000, whichis barely usable for any kind of regional analysis. Thereforesoil datacan
onlybeused with extremecaution.
Similarproblems were encountered with the hydrology data. Although data on water
springs and natural ponds was recorded during the field survey, the preliminary analysis of
theirdistributionwas shownto have very limi1edpotentialfor our analysis. The distributionof
waterresourceswas concentratedinseverallargerkarstdolinasand alluvial valleys. It isclear
that a substantial number of present water resources remained unrecorded, or that their
distributionmusthave changed.The majorreasonfor thechangeof waterresourcesliesin the
history of the vegetationcover which has been dramatically transformed during prehistory
throughtill today. This can be demonstratedby considerationof data from the end ofthe last
century. At the end of the nineteenth century, the Central Dalmatian Islands had been
intensivelyused for viticultureand agriculture.Inthe WesternMediterraneanthe winedisease
phyloxeradestroyed the vineyards, althoughoriginallythe Central Adriatic was spared from
this disease. Due to the increaseddemandfor local wine, previouslyunusedland was cleared,
terraced, and convertedinto vineyards. The vegetationcoverwas dramaticallyreduced.Since
vegetation is a very importantfactor for controlling water during periods of extreme rainfall
storms resulted in the flooding of previously dry creeks and, during the 19th century, the
deaths of several people. Today the land on the island of Brae is extensively covered with
dense Mediterranean scrub. This has resulted in the lack of surface water and it is hard to
believethat whatappears today as a dry valley with no water at all through the year, could
have been a substantialstreamduringthe 19th centurystorms. However,on thebasisofthese'
preliminary analyses it was decided, that data on water resources would not be used in the
analysis.
3. Locationalanalysisof theRomansettlementsites
An important step in analysing settlement patterns is to show that the characteristic of site
locations differs significantly from general locations across the landscape. We used a
Geographic Information System (GIS) for the exploration of this spatial relationship, and
slatistical software to develop a regression-based predictive model. Such linear regression
modelsmakeitpossibletopredictthe potentialasite possessesatagiven location.
The initial stage of the locational analysis was to analyse the relationship between
individual natural and social environmenialdataalong with the known site locations [11]. At
this stage of the researchit was hopedto get someinsight Intothe general distributionof sites
and to understandhow individual variables influenced-sitelocation. However, one shouldbe
awarethat the list of variablesused at this stagewas based on two basic facts. First, themodel
is always asimplificationof the real world, therefore, inany model some variablesare always
omitted because we do not know what would have influenced the real world. It also worth
stressing that certain variables are not used in the model because they would be' hard or
impossibletoobtain.
The first step in variable selection was based both on previous experience and the
empirical measurement of data, Therefore the initial set incorporated eight variables:
elevation, measure of surface slope, measure of surface aspect, quality of soils, land use,
erosion,distancefrom the coastand distancefrom Sennonelimestone, which is viewedas an
extension of surface geology. While all these variables are easy to understand, we need to
explain aspect in some detail. While aspect is generally given as the direction that a 'slope
faces, we decided to re-cede aspect so thatit represents a differenceof orienlation from the
south-westfacing slopes.
153
152
Beyondlhe Map Z Sianac and T. Veljanovski/Roman settlement patterns
The known sites and the background area were compared for each variable using a
difference of frequencies test, t. This is somewhat unusual in that we use a varietyof both
categorical and continuous independent variables, so to meet the requirements of thet test
wereclassified continuous variables intoappropriateclasses.Asalready noted, weusedfairly
simple univariate statistical techniques to extract a formal decision rule from the sample of
observations, the t technique is simple and transparent while at the same time provides
important insights into distributional relationships. A more refined insight into the
relationships betweenvariables wasobtainedbyregressionanalyses,seebelow.
The results of univariate statistical tests indicate that there are significant
environmental differences between site and background locations across five of the eight
independent variables (Table 1): elevation, slope, aspect, distance from Sennone limestone
anddistancefromthecoast.Sites tendtobelocated,ontheaverage,closer toBrae's 'marble.'
They also have a tendency to be located rather closer to the coast then the interior. As
expected sites also occur in low-lying areas(up to400m), on gentle to steep terrain and on
south-west facing slopes. These five variables identified in the univariate analyses were then
used in a multiple linear regression, the remaining variables were dropped because of their
statistically poor performance andtheirusemightproducespurious results.Onthebasisofthe
field survey and theoretical knowledge of Roman provincial archaeology, some correlation
betweenthedistribution ofRomansettlementsitesandsoils,erosion andpresentlandusewas
expected. Unfortunately, the relationship was statistically rather poor, probably due to the
inappropriate scale of the soil mapandchanges inerosion and land usedata due todifferent
vegetation.
Thepromising resultsof theinitialanalysisledustothecreationof apredictive model
of Roman settlement. Due to the fairly large number of sites, it was hoped that regression
analysis mightbeapplied toprovide amoredetailed insightintothe importance of individual
variablesand theircontributiontotheaccuracyofthepredictive model. Itisalsoimportant to
stress that regression analysis produced a continuous surface for the whole area and thus is
most suitable for producing archaeological predictive models. However, the results of the
regression analysis were not as good as anticipated. Some analyses were repeated using a
multivariate linear regression procedure since all the measurements of distance were made
intwoways:Euclidean andcostdistance. Regressionsolutionsandtheresultingfunctions
Table1:Results ofunivariate descriptivestatisticaltesting.
Variable SitesMean BackgronndMean
t t criI (<<=0.05)
Degreesof
freedom
Elevation 175m 265 m 9.85 9.49 4
Slope 7.5
0
10.9
0
7.90 7.81 3
Aspect 205
0
163
0
11.35 9.49 4
Land use (4classes) 2.6 2.7 1.24 7.81 3
Soils (4classes) 2.3 2.8 3.75 7.81 3
Erosion(4classes) 3.2 2.9 2.75 7.81 3
Disl. from Sennone
limestone(Euclid.)
1591 m 2771 m lUI 11.10 5
Disl. from Sennone
Limestone(cost)
39 min 178 min 20.76 14.1 7
Disl. from coast
i (Euclid.)
1952 m 2245 m 2.09 12.6 6
Dist. fromcoast
I (cost)
48 min 105min 9.75 12.6 6
relating site potential on 29 observations of five variables were not very promising in any
combination. For the distance from the coast it was realised that it has a strong correlation
withtheelevation soitwasexcluded fromfurther regressionanalyses(seealso Figure2).The
adjusted R-square, interpreted alsoasapartof theexplained variance,did notexceed 48 %in
anyof thesecases. Despite thepoor performanceof the predictive model, regression analysis
stillprovided ausefulinsightintorelationshipsbetweentheindependentvariables.
Itseems thatthethinlayerof Brae 'marble' reallyaffected locational decision making
on theisland (seeTable 2: lJ =0.540), theother threeimportant factorsfor Romansettlement
seems tobe aspect(south-west facingterrain) with lJ = 0.335, elevation (lJ=0.284) and slope
(lJ =0.137).
Table2: Results ofmultiplelinear regression.
MODEL
R'=0.467 Unstaodardised Standardised
F=5.262 Coeff'lCients Coeflidents
(Sil!,0.003)
6 Std.Error '6 t Sig.
(Constant) 11.117 4.387 2.534 0.Ql8
Aspect225 -0.062 0.028 -0.335 -2.168 0.040
Elevation 0.019 0.011 0.284 1.738 0.095
Disl. from S. I. 0.074 0.021 . 0.540 3.460 0.002
Slope 0.261 0.301 0.137 0.868 0.394
c:l ::;-___
I
,.e..
10_1
1211.'"
.11CI*'270....
t:l'.Hil
D"'-_
....,
....
......
5111.

- 0.1111",
.'-1
.'-4
."-'
.'-14
.H-
c0-1I ....
Figure 2: Setofvariablesiocluded intheregressionanalyses: a)cost
distaocefrom Sennonelimestone, b)aspect,c) elevationandd) slope.
(" ""-!,?"", ..
154
155
Beyond the Map
When the results (see Figure 3) were analysed in more detail it was realised that the
reason for the poor performance may have been in the archaeological data. What we have
interpreted as a homogenous type of site, i.e. Roman settlements dated from the 2nd century
BC - 2nd century AD, could have been in fact a mixture of several types of settlement. Some
settlement sites could have been positioned with the objective to perform agricultural
activities, while others might have been controlling other resources like Sennone limestone or
pasture. Therefore it was decided to carry out a more refined analysis of all Roman settlement
sites in order to provide ideas on possible clusters of sites on the basis of their relationships
with the natural and social environment.
Cluster Analysis of the sites seemed to be the most appropriate technique for this
objective. We applied the K-Means Cluster Analysis algorithm with the main objective of
identifying groups of sites with similar spatial characteristics. As this algorithm requires the
number of clusters to be specified, a preliminary Hierarchical Agglomerative Cluster Method
was applied and using the Ward criterion function, four clusters were identified. Elevation,
slope, aspect, cost distance from Sermon limestone and cost distance from the coast were the
variables used in the cluster analyses, the results are given in Table 3.
The first cluster incorporates nine sites. They occur on lower elevation (an average of
50 rn), they provide the highest values of aspect (meaning mainly north-east locations), they
are close to, but not on, the coast (average 15 min. walk away) and on average 22 minutes
away from outcrops of Brae's 'marble.' Slope values on these sites are usually fairly high. On
the basis of these characteristics, we defined this cluster as a seaport type of Roman site. It
seems the sites lie mainly on the north part of the island. The reason for the 15 minutes walk
distance from the coast is probably due to the need for shelter from the north-eastern wind.
The second cluster comprised 10 sites. They appear at a slightly higher elevation
(average 200 m) and on slightly colder north-eastern locations. At the same time these sites
are on average 45 minutes away from the coast. Because the smallest distances from the area

Potential
050%
5170 %

71- 100 %
o 5
Figure 3: Linear regression predicti ve model
Y = II. 117 + 0.074 (disl. from 'marble') - 0.062 (aspect) + 0.019 (elevation) +0.261 (slope).
Z S_iU and T. Yeljanovski / Roman settlement patterns
Table 3: Final cluster centres.
Ouster Elevation Slope Asoec:t 'Marble' Coast No. or Cases
I 47,78 9,33 83,56 21,78 14,56 9
2 196,20 6,50 80,70 12,00 44,40 10
3 10,50 8,50 50.00 248,50 2,00 2
4 331,88 6,38 40,88 38,13 100,50 8
of 'marble' (average 12 min.) seems to dominate the cluster, we denoted the cluster as a
quarry type of Roman site.
The third cluster has only two Roman settlements, both located on the lowest elevation
(average 10 m), on the coast (both sites are only a couple of minutes walk from the coast) and
far away from the Sennone limestone area. They also occur on south-west facing and quite
steep slopes. It is pretty obvious that these two sites are isolated Roman settlements on the
southern part of the island. Giving a name to this cluster is rather more difficult since the two
dominating characteristics are that this cluster is different from the first one mainly in its very
great distance from the Sennone limestone, and in orientation. On-the other hand, despite
these differences, it could still be incorporated in the cluster of the seaport type ofRoman site.
lbe fourth cluster is comprised of eight sites with the following common
characteristics: compared to other clusters, these occur on highest elevations (average 330 m),
on gentle slopes (average 6) and on the sunny hillsides. They are more than one hour and a
half walking distance away from coast _and on average 40 minutes away from the area of
Sennone limestone. This group does not overlap with the other clusters, and we are sure that
with better soil data there would have been a strong correlation with good agricultural soils. It
is also worth emphasising that relatively great elevation is not a restriction for viticulture. As
distinctive characteristics suitable for farming activity were expected at these sites, we named
this cluster simply as an agriculture type.
4. Summary and conclusions
In this paper we have presented one approach to the creation of predictive modelling for
Roman settlement sites. In the initial stage of the locational analysis we discovered a number
of interesting patterns of settlement location. Unfortunately due to the poor quality of the
archaeological data the resulting predictive model derived from the linear regression was not
satisfactory. In the more refined analysis it was proven that there are three different types of -
sites on the island of Brae. For each of these different types of sites different strategies and
patterns of control over natural resources and site locations were isolated. The research will
continue and will result in the- production of predictive models of individual Roman
settlement site types, however, the research presented here has shown the great potential of
applying regression procedures to predictive modelling of this kind.
Acknowledgements
European Commission, DG XII financed the field survey on the island of Croatia.
Computerisation of the site and monument records was partially supported through the
Research Support Scheme of the OSIIHESP, .grant No. 1363/1997. Special thanks go to the
..
156 157
BeyondthMap
Ministry ofScienceand technology of theRepublic ofSloveniafor the supportofbilateral co-
operation with the Republic of Croatia. Many thanks to John Hayes and Branko Kirigin for
their expert work on establishing the chronological sequence ofsites. Niksa Vujnovic was
together with Branko Kirigin and ZoranStancicco-directing the field survey on the island of
Brae in 1994. Vince Gaffney provided stimulating discussion, he has also linguistically
correctedthe text. Anymistakes are,ofcourse, ourown.
References
[I] Allen, K.M., S.W. Green, E.B.W. Zubrow (eds.), Interpreting Space: GIS andarchaeology. Taylor &
Francis,London, 1990.
[21 Judge, W.L., L. Sebastien (eds.), Quantifying rhepresentandpredicting the past: Theory, method, and
applicationofarchaeologicalpredictivemodelling. U.S. GovernmentPrintingOffice, Washington, D.e.,
1988.
[3] Kvanune, K.L. A predictivesite location model on the High Plains: Anexamplewith an independenttest.
Plains Anthropologist37(138),1992,pp. 19-40.
[4] van Leusen,P.M.GIS andlocationalmodelingin DutchArchaeology: A reviewofcurrentapproaches.In:
H.D.G. Maschner(ed),New Methods, Old Problems: GIS inmodemarchaeologicalresearch. Centerfor
ArchaeologicalInvestigations,OccasionalPaperNo. 23,SouthernIllinoisUniversity, 1996,pp. 177-197.
[5] Kohler,T.A.andS.c.Parker, PredictiveModelsfor ArchaeologicalResourceLocation. In: M.B. Schiffer
(ed.),AdvancesinArchaeologicalMethodandTheory9,Orlando,AcademicPress,1986,pp.397-452.
[6] Dalla Bona, L. Cultural heriklge resource predictive modeling project, Volume 3, Merhodological
considerations.Centrefor ArchaeologicalResourcePrediction,LakeheadUniversity,ThunderBay, 1994.
[7] Gaffney, V. and Z. Staneie GIS approaches to regional analyses: A case study ofrhe islandofHvar,
Znanstveniinstitutfilozofskefakultete, Ljubijana, 1991.
[8] Gaffney, V. B. Kirigin, M. Petrie and N. Vujnovie, The Adriatic Islands Project, Volume l. Tempus
Reparatum,BARInternationalSeries660, 1997.
[9] Vrsalovic,D. PovjestotokaBrae. Brackizbomik6,Supetar, 1968.
[10] Shiel, R.and J.e.Chapman,Theextentofchangein the agricultural landscapeofDalmatia, Yugoslavia,
asa resultof7,000years ofland management.In:J. Chapman,J. Bintliff, V. Gaffney, B. l p ~ (eds.),
Recentdevelopments in Yugoslav Archaeology. British Archaeological Reports International Series431,
1998,pp. 32-44.
[II] Kvamme, K.L. Ranters comer: Bringing the camps together: GIS and ED. Archaeological Computing
Newsletter47, 1997,pp. 1-5.
Beyondthe Map
G. Lock(Ed.)
lOS Press,2000
AGIS-basedanalysis ofthe
EtruscancemeteryofPontecagnano
usingfuzzy logic
MarcoCrescioli
Unirelsrl
Florence
Andrea D'Andrea
lstuuioUniversitarioOrientale
Naples
Franco Niccolucci
University of Florence
Florence
Abstract.Thispaperexaminesthe differentpositionsin the presentdebateonGIS
use in archaeologyand the useof fuzzy systemsin this field. After recalling some
concepts from fuzzy set theory and giving some necessary definitions, a'general
model for the implementationof a FuzzySpatialDatabaseManagementSystemor
fuzzy GIS is proposed.Then we considerthe case study, the Etruscancemeteryof
Pontecagnano,and apply the fuzzy model toit.
1.Introduction
This contributionis the convergenceof differentavenues ofresearch. First, theauthorsare.
aware of some inadequaciesin the present use (or misuse, or absence of use) of spatial
databases in many archaeological applications,beingat the same time convinced that most
professionals underestimate the potential ofthese tools. Also, we want to understand if
there isa"culturalmarket"forsoftwaretools that can innovate, inour opinion, the way in
which archaeological research iscurrently performed,especially- but not only- inItaly.
We also askwhy there isso little diffusion ofopen, free-of-chargesystems incomparison
with commercial packages that are often less powerful and always more expensive, and
verify if the availability of special features, possibly solutions to the first problem, may
significantly improve the appeal ofsuch systems. Hence this paper wavers (hopefully,
balances)betweentheoretical questionsoncomputationalarchaeology(sections2,3and 6,
mainly due to Andrea D'Andrea) and more technical aspects (sections 4, 5 and7, mainly
due to Franco Niccolucci) that are necessaryto explainour proposal: both are considered
as essential, the former to give the necessary cultural background to technicalities, the'
latter toavoid apurely ideologicaldiscussion.
. i
I
158
Beyond Ihe Map
2. GIS and archaeological theory
Does there exist a relationship between GIS and archaeological theory? And how does GIS
technology influence archaeological research? These questions raised by many authors
([I], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11]) still await, perhaps, a final answer,
notwithstanding the fact that GIS represent today the band-wagon of computational
applications in Archaeology. If then there is an open question concerning the assumed
existence of a relationship between GIS and archaeological knowledge, our discussion
cannot be limited by the argument, noticed by Gillings [7] that, after Wheatley's [I]
contribution aimed at "re-addressing" the debate on GIS applications, no archaeologist can
any more consider GIS as devoid of significant theoretical connections. Our discussion
must, on the contrary, address the relationship that GIS may generate in each applied field
(CRM, Spatial Analysis, Locational Analysis, etc.) as far as archaeological theory is
concerned. GIS are not impartial or neutral tools: they "represent the social reproduction of
knowledge and, as such, the development of a GIS methodology cannot be divorced from
the development of the theory needed to sustain it" [4].
In recent years, the adoption of information systems in many operational fields has
been followed by a number of papers and conferences devoted to the problems raised in
the context of archaeological research by the use of such an innovative technology. The
debate appears to be focused on two threads: a) the mutual relationship between GIS and
theoretical models, with a specific attention to the critical points of interrelationship
between spatial information and archaeological interpretation; b) the operational-
methodological circuit, with an interest on the limits and the future of GIS.
Voorrips [10] noticed that predictive models, the most popular sector, represent a
kind of empirical generalisation of experience, thus loosing any explanatory function and
power; spatial statistics produce models (maps and databases) based on precise selections,
which end by reducing or, worst, deleting, variability. GIS applications have often caused a
reduction and simplification of archaeological reality [7]. In this "normalising" perspective
GIS do not seem to radically innovate the theoretical panorama. TIley maintain some
utility, on the contrary, if they can open new paths to the analysis of space/time variability,
thus favouring a cognitive process that increases the quality of archaeological record. This
record, to be explained, needs new developments in theory [10], otherwise the simple use
of the tool and the relative easiness of manipulating environmental variables leads, even
unconsciously, to stressing the importance of environmental factors over cultural ones.
According to Voorrips, GIS have nothing to offer to studies that are not holistic and do not
try to explain archaeological phenomena by means of wide generalisations, looking for
space and time patterns. But for van Leusen the simple action of producing predictive
models for CRM does not need "histories" to be justified; it becomes useful for urban
planning, since the financial resources for archaeological research are limited [3].
Gaffney replies to this assertion that seems to assign some "neutral" value to GIS
application; he states that the description of simulation models is not necessarily predictive
and that there is the risk of destroying the archaeological heritage not included in the
model, if the model is inadequate [3].
For some authors, it is poor awareness of the potential of GIS use and the absence
of adequate theoretical reflection that limits GIS application, and not software limitations.
In the CRM field, the use of GIS is mainly related to the need of processing large
quantities of data (graphical and alphanumerical): the need of converting information from
analog to digital format represents the first step for GIS use in archaeology, which replaces
and improves operations and procedures often performed manually [4]. In this perspective,
GIS have often been used as simple tools to manage and visualise data, thematic maps and
queries. Limp [8] notes that the lack of analytical functions impoverishes the use of GIS in
159
M. Crescioli et al. / Etruscan cemetery of Pontecagnano and fuzzy logic
archaeology, in most applications; on the contrary, the use of spatial technologies may lead
to the discovery of patterns and thus produce new theoretical opportunities. Archaeological
theory seems, therefore, intimately connected to methodologies capable of "measuring" the
models and hence the archaeological phenomena, as has already happened with CI4 dating
techniques [8], [I].
As regards the initial question, it seems that the emerging opinion wavers between
a "good" GIS, capable of creating a new time/space metaphor by innovating theory and
explaining the variability and/or themodels, and a "bad" GIS, assimilated to a simple data
manager or, worst, produces an eco--functionalist determinism.
In the debate on the function of GIS, as seen above, there are different positions
that often diverge, but they coincide as far as they underline the need of always considering
the theoretical connections derived by the use of technology in archaeology together with
the technological tools. Anyhow, the discussion on GIS is influenced by the more general
debate on archaeological theory; it appears still strongly conditioned by the criticism that
some tendencies in European archaeology, mainly Post-Processual Archaeology,
addressed to the mathematical-statistical approach and more generally to computer use in
archaeology. Such a negative judgement, that involved several theoretical and
methodological features of New and ProcessualArchaeology, addressed the rhetoric of '.
quantification, namely the generalised adoption of mathematical-statistical techniques,
charged with being a substitute for archaeological interpretation. Post-Processual
Archaeology rejects the computational approach in the context of a more general
relativistic atmosphere: computer science is again relegated to the role of "ancilla" in
archaeology (on these problems see [12] and [13]). It maintains some utility only when it
allows "objective" and "neutral" analysis of large quantities of data In order to generate
statistical indexes to support analysis [14]. In conclusion, notwithstanding their irresistible
development, GIS still originate strongly contrasting reactions, wavering between a total
scepticism and a sometimes acritical enthusiasm.
However, these positions, with their diversity, reflect the level of maturity achieved
by GIS use in archaeology. TIley correspond to two different opinions: one expresses doubt
and perplexity on the value of this technology in the context of the theoretical tradition, the
other exalts the analytical virtues of this revolutionary instrument. The first position, with
strong roots in the academic world, includes those who look curiously but critically at the
value and the impact of technology on traditional archaeological theories. The diffidence
toward GIS hides the "fear" of the introduction of any innovation in research. It appears to
be generated by the suspicion that the contribution of GIS as a substitute for interpretation
rather than a tool capable of improving the investigative ability of researchers and
therefore guarantee a wider knowledge of archaeological phenomena. In their opinion, the
extraordinary analytical power offered by new technologies led the archaeologists, in many
cases, to progressively abandon the goal of their studies and to exalt the instrument and its
power. Thus, the application of computational instruments .replaces the aini of
investigation that should be to generate social significance from the archaeological record
to produce interpretation. The only escape offered to GIS is in the possibility of
incorporating cultural (space, time) or subjective (visibility, perception) variables in the
study of the territory; these should reduce the risk of generalisation and deterministic
statements through a cognitive or contextual approach to "social" landscape ([2], [3], [4],
[Ill). However, if such an analysis derives only through viewshed studies, substantially
involving only environmental factors, the research does not differ from the most
inadequate eco--functionalist determinism [3]. TIle only adoption of viewshed analysis,
namely to consider vision instead of perception as an act of cultural relationship in man-
environment interaction, risks reproducing the same determinism ofpredictive models [7].

160 161
Beyond Ihe Map
To avoid such damage it is necessary to add simulation and interaction (3D, VR) to land
study, for the possibility they offer to investigate and realise the relationship between a
culture and the ancient landscape in which it developed [9].
A more open-minded opinion is supported by scholars who maintain that
introducing information systems in Archaeology represents by itself an event that may
transform the entire body of archaeological science ([8], [15]). 3D landscape analysis
allows, for instance, to build what has been named a new perception of space, determining,
by means of viewshed and cost-surface analysis, the birth of a new space/time metaphor
which goes beyond any traditional two-dimensional paper approach. Reconstructing such
a space-time frame eventually produces important consequences on methodology and
epistemology [6].
Therefore, being technology neither neutral nor harmful but only badly used, the
archaeologist must try to "humanise" the computational tool by fully integrating it in the
theory and in the methodological reflections, cutting out any neo-positivistic approach and
the pretended objectivity of archaeological facts. The use of GIS has already oriented
investigations towards new paths, otherwise impracticable, creating a new cognitive
perspective of space which traditional methods could not explore. The use of economic,
social and political models as alternate perspectives to the cultural study of the territory
helps in giving a social explanation of archaeological phenomena rather than the simple
description of observed structures [4]. Using software that can store and analyse 3-
dimensional forms offers to the archaeologist possibilities of knowledge that up to now the
traditional cartographic systems did not offer, increasing and improving traditional
instruments such as the stratigraphic matrix.
After a long period reserved almost completely for the more promising applications
side, recently much attention has been devoted to the study of theoretical problems and
questions. Besides precisely defining the scope of GIS application, investigation concerns
which studies may take an advantage by such an innovative technology. Several
contributions interpret the need of understanding in a more analytical and detailed way
than just application fields, for example the scientific problems raised by GIS use and
above all the limits that their bad use may induce in research. The analysis of the potential
of raster and vector GIS use in inter- and intra-site applications is accompanied by the
interest in the transfer of such technology in Cultural Resource Management, in particular
at a regional or local level (for a recent and complete survey of CRM applications of GIS
see [16], which contains also some interesting considerations of general character). This
trend reflects the need to preserve archaeological data in digital format, also to preserve
their integrity for future use and it determined very recently a substantial increase in
applications dedicated to inventory, to CRM and to the creation of simulation models and
archaeological impact maps. Around the increasing diffusion of GIS it seems that a new
science of geographic information is originating: this, as Kvamme noticed at the very
beginning of GIS introduction in Archaeology [17], is one of the primary components of
archaeological recording and, therefore, may greatly benefit from computerised
procedures. In conclusion, the development of GIS use in Archaeology cannot be
considered only a technological issue, limited to a small circle of followers, interested only
in the technical novelty. If its diffusion is still not uniform, this may be only attributed to
different research and study traditions determining a different level of awareness of the
methodological and epistemological value of the new instrument.
In this debate it clearly emerges that all reflections consider only spatial analysis at
an inter-site level, including "economic" questions, as an answer to physical-natural
adaptive processes, and the investigation of cultural facts through perception. The same
attention has not been devoted to intra-site applications and to the possible impact on ritual
analysis, on the study of subjectivity and on cultural context analysis, especially
M. Crescioli et al. / Etruscan cemetery ofPontecagnano andfuuylogic
cemeteries. A limiting factor to the development in this direction might be the restrictions
of GIS functionality as far as 3D spatial analysis and time are concerned [4]. Probably the
greatest obstacle to GIS introduction in the study of funerary habits is the impossibility of
investigating cultural facts, which are not directly connected to the spatial and physical
dimension but are also mediated by cultural and social factors.
Our investigation is an attempt to redress the potential of GIS in this area, still difficult and
little explored, in order to acknowledge and reconstruct the functional and social
organisation of an ancient community, from funerary evidence ([18], [13], [19], [20]).
Integrating fuzzy logic in GIS essentially aims at incorporating in the data structure the
uncertainty of data, which derives from imperfect knowledge or statistical derivation. To
preserve the vague character of information may indubitably avoid improper
generalisation.
As far as the relationship between GIS and archaeological theory is concerned, the
cemetery is a privileged observatory: since the relationship between social organisation
and its mirror in funerary habits is complex and is never immediate [19], [20], we want to
investigate if the use of a fuzzy GIS in a funerary context may make a significant
contribution to the concepts of spatial and cultural analysis. Ofcourse, this paper is only a
first step in this direction.
3. Fuzzy Logic applications in archaeology
Fuzzy theory was created by Zadeh in 1965 [21] with the aim of rationalising all the
situations in which the traditional, bi-modal logic was clearly inadequate. The idea is to
characterise imprecise statements assigning a degree of fulfilment which is conventionally
defined as a number between 0 and 1,0 corresponding to "False" and I to "True". As for
traditional Boolean logic, this has a counterpart in set theory, which may bethought of as
the degree of truth of the assertion "x belongs to X": this leads to the definition of fuzzy
sets, which are an extension of traditional sets. Fuzzy sets are thus meant to describe the
situations in which it is difficult to draw a crisp border between the set and the world
outside of it: a "mountain" is a geographic example of such a set, because it is almost
impossible to establish where the mountain ends and the plain starts. We shall recall the
main definitions and results of fuzzy theory in the next section in a more formal way, but
these intuitive concepts are sufficient to give a short account of the applications of fuzzy
theory in the archaeological field, which may be useful - although it does not pretend to be
exhaustive.
Fuzzy logic was introduced into archaeology at the end of the 19808, particularly in
mathematical and statistical applications; the use of this theory in data management is
relatively recent, the first mention due to Thaller in 1981 [22] concerns specifically the
possible use of fuzzy logic in historic data management (see also, by thesame author, [23] .
and [24]).
First applications concern chronology, some developed in Italy [25], [26], the latter
aiming at dating ancient lagoon settlements to a period varying from 1000 to 1400 AD.
This investigation concerns 44 sites and is based on a c-mean classification of bricks found
at the sites; the goal is to recognise clusters based on selected materials, to propose a
chronology for settlements for which dating is not certain .: Using fuzzy logic, sites are
classified according to their degree of membership and not in a rigid way as produced by
multivariate statistics.
A distinguished Russian historian, Leonid Borodkin, has devoted several papers to
similar problems, developing ian algorithm [27] for pattern recognitiori; one of his

162
Beyondthe Map
applications to archaeological problems concerns the analysis of data coming from burials
[28]. In this paper they examine data coming from 23 cemeteries in Asian Sarmatia that
can be dated between the Bronze Age and the first Sauromatian period; their method is
based upon the definition of a weight relative to the centre of the cluster, for each
cemetery.
Other works related to fuzzy classification are due to Horia Pop, with applications
to archaeological problems [29], [30], [31]; up to now, he has continued working on
similar methods but addressing different fields of application (see, for instance, [32]). Two
papers by Nick Ryan and Janet Bagg address chronological issues and, according to the
authors, solve the problems of time interval types in relational databases with an innovative
solution [33], [34]. Two early papers by one of the present authors address chronological
problems [35] and historic boundary questions [36], both dealing with archaeological
applications. They apply fuzzy set theory to some cases, in which limited information does
not allow the determination precisely of the values of some attributes of data (time of
foundation of a village in the first case, places belonging to a dioceses in the second),
which are supposed to be stored in a database. The methods suggested in these earlier
papers are the starting point for some proposals in the current one.
Another research field concerns the development of systems able to analyse
complex situations basing on a set of rules which the system itself, to some extent, can
generate, so-called expert systems [37]. Expert systems have been applied in archaeology
to discover patterns and obtain automatic responses to investigations. A study by Barcel6
[38] attempts to reconstruct the attribution of a ceramic fragment by determining the
degree of membership of single variables defining the typical object. As in neural
networks, the selected parameters describe a sequence of nodes that connect the descriptive
variables. This investigation aims at creating an expert system for the study and analysis of
Phoenician ceramics (800 - 500 BC), to derive from it the chronology, the production
features and the exchange patterns of this particular ceramic class.
At a different scale - and this introduces the subject of fuzzy theory applications in
GIS - Claire Reeler [39] has developed an expert system for the analysis of the remains of
pa sites, Maori settlements in New Zealand before thearrival of the Europeans in the late
1700s. This work is logically connected through a module, named fuzzyCOPE, a hybrid
system incorporating neural networks and fuzzy logic, designed to look for patterning
within data and explain the patterning, developed at the Department of Information
Science, University of Otago, New Zealand.
Although considering this approach interesting, we do not feel comfortable with it,
since we have the impression that the set of rules necessary to draw conclusions from
given data is so complex that except in very simple cases it is not worth the effort.
Moreover, the peculiarity of archaeological situations could lead, in many cases, to the
reformulation of a significant part of these rules. Last but not least, hiding part of the
discovery process within some software program might add a brick to the wall of data
privatisation, in which an intensive use of Internet is now beginning, on the contrary, to
produce the first cracks.
The acknowledgement of the potential of fuzzy logic in geographical applications
is marked by the publication of the proceedings of a meeting organised by GlSDATA [40].
In the same period, a paper by Nick Ryan [41] evidences the need of incorporating fuzzy
theory in GIS data structures. In the context of a detailed discussion of the need of
introducing fuzzy concepts into archaeological GIS, Harris and Lock [4] maintain that
there is an open question in the accuracy and precision of archaeological information
stored in a GIS: the problems arising from imprecise chronology, uncertain location and
vague borders, impose the necessity of incorporating this uncertainty level within the
J63
M. Crescioliet al. / Etruscancemeteryof Pontecagnanoand fuzzy logic
architecture of the data management system. Similar problems about uncertainty in
archaeological GIS are addressed by Kirkinen [42].
Some steps in this direction were made by Wheatley [43] who pointed out, in an
innovative way but starting from some principles worked out by Fisher [44], the need of
introducing into cumulative viewshed analysis the uncertainty deriving from the way the
OEM is constructed (interpolation, height accuracy, etc.) and from the site location
inaccuracy. These hints towards a fuzzy viewshed have been followed by a Belgian team,
which has applied a Monte-Carlo algorithm to viewshed estimate [45] and, to the external
defence system of the Hellenistic town of Sagalassos [46]. The viewshed sector seems,
therefore, a promising application field ([7], [11D, together with the above mentioned
clustering techniques.
Geographers have continued studying the problems concerning precise definition of
geographical objects, which often involve deep mathematical concepts. Very recently the
6
1h
Symposium on Spatial Databases (SSD) has faced questions that might have important
consequences in GIS applications in archaeology: as far as the scope of this paper is
concerned, Schneider's contribution [47] has great relevance. It may be interesting to note
that in the title they speak of "Spatial databases" namely of databases in which data,
attributes or relations have a spatial character and are processed accordingly. Maps - which
too often seem to be the unique goal of many applications - remain in the background. The
same happens in the work performed by the Chorochronos research project (see the Web
site of the project [48] for a list of related work and, for instance, [49] for previous work by
some Chorochronos participants), which might have a great potential in archaeological
applications.
In conclusion, fuzzy logic has so far received a sporadic attention and none of the
very few fuzzy database management systems up to now implemented have been applied
to archaeological problems. Apart from Underlining the questions arising when dealing
with imprecise data, the highly specific applications of fuzzy set theory have left little or
no space to discuss the theoretical implications of a method that makes explicit the level of
data uncertainty or imprecision. The common attitude toward databases usually does not
take into account any critical analysis of archaeological deductions drawn from searching
collections of data, which may be highly reliable singly but lose reliability as a whole.
Fuzzy logic may be not the solution, but boolean logic is in most cases too narrow a cage.
4, Fuzzy concepts
According to Zadeh [21]. the formal definition of a fuzzy set is the following:
Given a set U (called the domain or universe), afuzzy set Q is a function Q: U [0, 3],
namely a function that associates a number in [0, 3] to each element of U, Figure I.
(Xx)
---l I I I \ .U
Figure 1. Graphof a fuzzy set

165
164
Beyond the Map
This concept is a generalisationofordinary(crisp) sets: for these, the associatedfuzzyset
is the characteristicfunctionx(x), which is afunction that has value0 outsidethe set, 1on
each elementof the set.
A fuzzy set can be regarded - as can an ordinary set - as being characterised by
some propertyP affecting elementsof U. For a crisp set X, an elementx belongs to X if
and onlyifx has the propertyP, namely,Xx<x) =1ifandonlyifx has the propertyP. For a
fuzzyset A, the value ofA(x) expresses the degree of satisfaction ofpropertyP, in a scale
from 0 to 1.If U is a discreteset, the fuzzy set is also called afuzzy label; if U C R, the
fuzzyset isalso calledafuzzy number orfuzzy value.
For fuzzy sets, it is possible to define operations which are similar to set union and
intersection; in terms of the definingproperties, these are the equivalentsof AND and OR
logical operations: ifA and B aretwo fuzzy sets,
(A v B)(x) = Max (A(x),B(x
(A n B)(x) = Min (A(x), B(x
It is easy to give examplesof common sentences in which fuzzy properties are involved:
for instance, any predicatewith tall, young, long defines a fuzzy setsince it is impossible
to separate objects that fulfil these properties from objects that do not. In practice,
assigning the value of a fuzzy set is not automatic and involves a certain degree of
subjectivity: in some senseitgives an explicitformulation toconceptsthat usuallyhave an
implicit meaning and facilitates the communication by assigning a numerical value to
expressionsthat otherwisewouldremainindeterminate.
Fuzzy logic, that is the logic based on the use offuzzy concepts, is generallymore
adequate than Boolean logic to express mathematically concepts and expression of
commonuse. To force fuzzy conceptstoBooleanlogic, usuallythe associatedfuzzy set is
rounded to a crisp set, defined as the set of the elements x having A(x) il., il. being a
predetermined value. For instance, one can put il. =180 ern and state that tall individuals
are those whoseheight is greaterthan it: in this way the fuzzy conceptof"tall"is rounded
to the crispconcept "having a height greater than 180 ern". This may obviously lead to
confusion: for instance another person may associate the concept of "tall" to a height
greaterthan 170em and so an individual 175em tall can be a sourceof misunderstanding.
This iseasilyavoidedusingfuzzy sets instead oftheir roundedcrispapproximation
Another possible consequence of rounding is the deduction of a false conclusion
derivedby achainofimplicationsinwhich the startingpointis afuzzyset and each step is
based on a fuzzy property, each with a high degree of satisfaction. This means that a
syllogismmust assume that the premisehas a 100% certainty and also the deduction rule
has 100%effectiveness:so that, from the rule "everyA is B" and the premise"x isA" one
can deducethat "x is B." If, on the contrary, we know only that, for instance, "80% Aare
B", we can chooseoneof the following:
round "80%A are B" to "allA are B" which may soundreasonablesincethe level
of satisfaction is quite high and hence deduce that "x is B" from the premise"x is
An; or
evaluatethat "X is Bwith aprobabilityof80%."
The second choice inhibits the use of Boolean logic, so usually the first is preferred,
especiallywhen computersare used. This may produceconsiderableerrorsif the rounding
offis implicitand the reasoning is chained: if we know also that "80% B are C","80% C
are D" and soon, roundingtells us that "X isC", "X isD" etc. but this may be false - or, at
least very unlikely- after a few passes. By using probability rules, it is simpleto verify
M. Crescioli et ai.1Etruscan cemetery of Pontecagnano andfuzzy logic
that after threepassesthe probabilityofthe deduction is about50% and after seven passes
itcollapsesto20%.
This may be very dangerousif part of the reasoning is not under the user'scontrol.
Consider, for instance, the attribution of gender to deceased found in graves during the
excavationofa cemetery.Thismay be veryreliably done by osteological methods, butit
must be kept in mind that it is a statistical methodthat maintains a degree of uncertainty,
even though very small and ignoring this fact may lead to completely wrong deductions.
Unfortunately, in ordinarydatabases, Boolean logic is a necessityand there is no way of
avoidingsucherrors.
Another very common problem arises with chronology, which onlyexceptionally
may be preciseso that the orderednatureof time may be used in selections.In most cases,
attribution has wide ranges of oscillation and thus chronological order is unavailable.
Choosingobjects belonging to a certain period is possible only if the hitter is so wide to
allow for fluctuations due to fuzzy chronological assignment. Indirect, but precise, dating
(for instance as a consequence of additional information as an epigraph or historical
documents) may lead to even worse mistakes, if improperlyused: for. instance, in a
medievalcemeterywith dated gravestonesfrom the 12'''and 13
th
century,inorderto select
burial objectsfrom the 12
th
century,criteriasuchas "date<1201"may excludesomething
dated 1201 and very likely madebeforethat year, namelyin the 12
th
century(see'[35] for a
more detaileddiscussion).
Similarproblems arise in making statementson the cardinalityof subsets that are
only vaguely defined, as saying "most of', that is using quantifiers. The classical
quantifiers "all"and "atleast one" may be extendedto include a set of conceptscurrently
used suchas "almostall", "few", "many","most","aboutone half',"atleast 20%",and so
on. These operators were called by Zadeh (see [50], p. 60 or directly [51]) linguistic
quantifiers. We shall deal with linguisticquantifiersinfuture developmentsofour model.
Finally, spatial concepts such as near to or oriented towards are hard to define
using booleanlogic sincethey naturallypertainto the fuzzy realm.If aspatialconcept may
be referred to the (fuzzy) evaluationof a one-dimensional variable, the theory developed
up to here is sufficient to solve the problem in a fuzzy context: this is the case of
proximity, which uses someform ofdistancebetweenobjects, and orientation which uses
some angular value. Also the definition of geographical objects often involves fuzzy
concepts.We intend to developadetailedstudyoffuzzy spatialconceptsin thefuture, and
here we shallconsideronly fuzzy extensionsofcrispgeographical objects. Advancements
inthis direction,atapurelytheoreticallevel, are presentedin [52] and[47].
We shall start by introducing a fuzzy representation of "near'! andthe concept of
fuzzrborder. Afuzzy set in R
2
is a function fX.x, y): R
2
[0, 3]. Let R be a (crisp) region
in R (we use the conceptofregion intuitively: for adiscussion on this see [47]). Given a
fuzzyset f3 such that fX.x, y) = 1for (z, y)e R, thefuzzy border ofR is the closure B ofthe
set (x, y) E R
2
: 0 < fX.x, y) <1). If it is useful to underline the fact that B is the fuzzy
border of R, denoted BR. Thus the concept offuzzy border is strictly connected to the
functionused todefineit,the fuzzy borderfunction.
For practicalreasons, it is convenientthat B is a limited set. A very simplewayof
buildingthe fuzzyborderof a(crisp)polygonalregion isthe following, devised firstly in a
slightlydifferentway in [36]. EnlargethepolygonR (aquadrilateralinFigure 2)by afixed
ratio /p, obtaining a similar polygon R' with the same centroid C. This will keep into
account the proximity region, that is the part R' - R ofthe plarie in which objects are
considered"near,"to asmalleror largerdegree,to theregion R.
J""". ....
166
I
Beyond the Map
Figure 2. The "neighbourhood" of a plane region
To compute the value of the fuzzy border function f3, let us assume that it is 1 for the points
of R, 0 outside R' and linear on R' - R. Representing the values of f3 on the z axis, we
obtain a truncated pyramid as in Figure 3.
IfIf! is the fixed value determining the extension of the proximity region, out of
which the points are considered as not near to R, and p is given by the following formula
p = Y. -m.... m. Y.., - Y. where =
yp -m..xp X!+I -Xj
where (x., Y') are the vertices of the polygon R, it can be shown that the value of the fuzzy
boundary function f3 for any point P =(x,' y,) in the boundary region R' - R is gi ven by
13(""Y,)= If'p-I
If'p-p
Thus the algorithm to evaluate the degree of proximity f3 of the point P to the region R
may be stated as follows:
I. Determine the side k of R next to P. For each couple of consecutive vertices compute
A ".Y,-",Y.
..,(y..,-y.)-(...., - ... P,
and find the (only) side k for which o:s A. :s 1
2. Evaluate
13("" y,)= mi{: 'I}O)
The last formula accounts for the fact that f3 must be I inside R and 0 outside R'.
z = fk..xy)


Figure 3. Determining the value of the proximity fW\Ction
16.7
M. Crescioli et aI.1 Etruscan cemetery ofPoniecagnano and futxylogic
It may be interesting to consider the proximity degree of two regions R1 and R2;
there are several ways to accomplish this. One considers both fuzzy borders, and the
proximity is given by the quote of the highest point of the intersection of the surfaces
representing the fuzzy borders. Another method considers one of the two regions, R2, as
subordinate to the other Rio which has a fuzzy border Bix, y). The proximity 7l( R2, R1) of
the secondary region R2to the primary one R. is evaluated as the highest value attained by
B(x, y) in the points of R2. These concepts too will be developed in subsequent studies.
To compare fuzzy sets, it is necessary to introduce relational operators among
fuzzy sets that generalise the usual equality or inequality operators. Of course, the result of
the comparison will not necessarily be True (1) or False (0), but most probably it will
assume some value between 0 and 1. The similarity s(x) between the fuzzy sets A and B,
defined over the same domain U, is the function s(x) = min (A(x), B(x. A graphical
representation of s(x) is shown in Figure 4.
j .)
Y \ A(x)-B(x) / U.
Figure 4. Graph of similarity function
To compare globally the two fuzzy sets A and B, the maximum of s over U is taken: in this
way, we define a fuzzy operator, that is a function associatinga number in [0, 3] to each
couple of fuzzy sets. We shall use the symbol - to denote this operator, calledjUzzy equal
to. In Figure 4, the value of A(x) - B(x) is given by the ordinate of the marked point. In
particular, if one of the two sets is the neighbourhood (with given radius) of a given
number, we obtain the definition of fuzzy equality to a number. Ifthe domain Uis ordered,
it makes sense to consider the operatorsjUzzygreater and jUzzy less; we shall deal in detail
with these in a forthcoming paper, since the ones defined in [53] are, in our opinion, not
completely satisfactory.
S. Fuuy archaeological GIS
The argument developed above has, as has been pointed out, a very simple and intuitive
interpretation in terms of ordinary statements, relationships and concepts. Our goal is now
to introduce the concept of fuzziness in database management and to apply this extension
of traditional databases and GIS to archaeological problems.
Fuzzy theory generally aims at determining the best strategies 10 control systems that
behave in a fuzzy way, de-fuzzifying their behaviour. Defuzzification is outside our goals:
our aim is, on the contrary, to preserve all the uncertainty related to the information and to
determine how this uncertainty propagates to the data. In our opinion, it looks promising to
introduce fuzziness into database management systems, leading to so-called fuzzy
databases. Fuzziness may reside in the data themselves, in the relationships between data .
or in the search system, which can allow queries based OIl discursive, unstructured
language. At present we shall be limited to the first two instances, thinking that an
"intelligent" system of queries may be a secondary concern for a DBMS addressing
scientific research.

168
Beyond Ihe Map
Fuzzy databases recei ved significant consideration in the early nineties and are still
under investigation. The most interesting results have been obtained, in our opinion, by a
Spanish research team working at the University of Granada which has developed a
theoretical model, named GEFRED [54], which later they implemented as a Fuzzy
Relational Database Management System [53]. The Granada system has no provision for
spatial relationships, which is, therefore, our original contribution. Moreover, since our
needs are more limited, we are able to manage uncertainty with less technical apparatus
and we introduce some changes to the Granada system; some of these, which we hope may
be considered as improvements, are possible since we use a more powerful DBMS, as
described below.
We propose an implementation based on the public domain RDBMS PostgreSQL
linked to the public domain GIS GRASS and endowed with an extensibility feature that
makes it the ideal candidate for this purpose. Moreover, a Tclffk interface has been
produced for both, so menu and mouse addicts are not frightened by command-line
software. As it happens, this free software is better than costly ones (see www.postgres.org
for a comparison between PostgreSQL and other RDBMS). The RDBMS PostgreSQL has,
moreover, the possibility of defining new data types and this allows the introduction of a
FUZZY data type. Operators and functions may be user-defined, and the system has among
its native data types a number of geometric types and functions: thus it is the ideal
environment to develop a Fuzzy Database Management System, on which to base a Fuzzy
GIS. To develop new operators and functions, it is possible to use different languages,
among these SQL queries. So far, optimisation has not been a primary concern for us,
therefore data storage and computation hardly ever follow the most efficient way.
Recalling the concepts introduced above, first of all we need to define a FUZZY data
type and the functions to operate with this type of data. To take into account the significant
difference between the discrete, unordered domain of a fuzzy label and the continuous,
ordered domain of a fuzzy number, there will be two different fuzzy data types, a
FUZZY_LABEL type and a FUZZY_VALUE type.
For FUZZY_LABEL, it will be necessary to store somewhere the domains, namely the
possible labels, which arethe same for all data of the same type; so theonly variable data
are the values of the membership function, which can be stored in an array.
For FUZZY_VALUE, we shall generalise the method used in [53]. In this paper the
fuzzy data types were introduced as simple trapezoidal fuzzy sets, thus needing only four
parameters to be stored, that is the lower and upper limits for the value 0 and thelower and
upper limits for the value I; as a particular case, the latter may coincide (triangular fuzzy
set) or be identical to the two other limits (rectangular fuzzy set). Even if this approach
optimises storage, we think that it has a serious drawback: logical operations with two
fuzzy items may give the result of a fuzzy set that is not one of the above types. Since
fuzzy sets are in some cases a generalisation of Boolean variables, this may introduce
unnecessary limitations. Therefore our FUZZY_VALUE will store an array of couples, the
values where the membership function changes the slope, assuming it is piecewise linear in
the intervals within, and the corresponding values of the membership function.
In conclusion, both our FUZZY data types will consist of a sequence of couples, the
first being the "value" and the second its membership degree in the fuzzy set. For fuzzy
labels, the first element of the couple will be implied.
It may be useful to define constants for any type of fuzzy data, which will be stored
in a separate system table. Due to SQL naming rules, and to make constants in queries
easier to use, these will be denoted as functions with no parameter, as for instance
YOUNG().
169
M. Crescioliet al. / Etruscancemeteryof Pontecagnanoandfuzzy logic
The process of fuzzy data type creation may be standardised and then a simple
procedure may be prepared which creates the necessary tables in an interactive way, asking
the user for the necessary parameters and labels. Also a graphical way to input/output the
values of fuzzy attributes might make their use easier. Windows, buttons and the like may
be easily created using a TcVfk interface to PostgreSQL, named Pgaccess.
On the GIS side, we chose GRASS, a well-known GIS package. In the past, using
GRASS had been discouraging for its command-line interface, but recently a menu
version has become available, and it makes working' with GRASS quite easy as with any
commercial GIS package. The command-line version may be used concurrently for special
tasks, which are not yet supported by the TcVfk interface. Moreover, there are functions
that connect PostgreSQL and GRASS and allow the visualisation of query results
(d.vect.pg) or interrogating the database from the graphical interface (d.what.s .pg),
thus making the PostgreSQL - GRASS system one of the best spatial databaselGISs
available for desktop computers. The potential of this system is emphasised by the
existence of PostgreSQL geometric native types, which allow complex queries and
processing of geometric data.
6. The case-study
The case-study data come from the Etruscan-Campanian centre of Pontecagnano, a
modem settlement placed about 70 krn south of Naples, which incorporates the area of the
ancient cemetery. In this place thirty years of excavations have revealed about 100
funerary areas, with more than 8000 burials dating from the period between the First Iron
1h
Age (9 century BC) and the Hellenistic period (beginning of the 3
rd
century BC). To
store, analyse and process the information retrieved during thirty years' investigations,
spatial and topological data on tombs have been linked with descriptive data on funerary
context in a GIS [55], [56] by one of the authors of the present paper. Then archaeological
records have been superimposed on modem maps to reconcile the interest of scientific
users, addressing the reconstruction of the social, economic and cultural organisation of the
ancient community, with the aims of planners, integrating land planning with
archaeological heritage preservation..
The present data are derived from the sample studied by Antonia Serritella [57].
His work concerns, in particular, Ponteeagnano cemeteries of the 4
th
and 3
rd
century BC. It
includes a philological study on grave-goods and pottery production and offers significant
reflection on funerary habits;giving interesting hints on how to reconstruct the behaviour
of human communities from Pontecagnano in the transition period from the Etruscan-
Campanian centre to the Latin colony. The areas examined by Serritella are a
topographically distinct set from theremainder of the 4
th
century cemetery: they insisted on
areas that were not previously occupied and thus represent aprivileged observatory for the
investigation on the ancient community. These funerary areas differ from others in
Pontecagnano cemetery as they were occupied with continuity since the Orientalising
Period to the 4
th
century BC. . .
The book considers 164 burials from 7 areas explored during preliminary
investigations performed before building on them. To examine funerary habits, the author
used, for each funerary' sector, palaeo-demographic and palaeo-pathological
investigations, performed by palaeo-anthropologists [58], [59]. The availability and
publication of these anthropological data, a remarkable example of data diffusion that must
be positively ascribed to the author's credit, were crucial to our research.
~ . . . . ~ ~ ~ : ; ~ ~ ~ ~
170
Beyond the Map
To give a summarydescription of the areas, identified by the name of the modern
owners, the 34 tombs from the Rossomando-Sica excavations were part of the same
cemetery,occupied between the secondhalfofthe 4
th
centuryand the first decadesofthe
3
rd
; they are placed at the northwest limit of the urban cemetery, north of the oldest
Picentine cemetery. Fifty tombs come from the Forte-Bassano area, also settled in the
urban cemeterysouth ofan area intensely occupiedsince the 6
th
centuryBC, in which the
funerary sector was in use since the second quarter of the 4
th
century BC to the first
decades of the3
rd
century.Twoburial zones were discoveredsouth of the ancienttown, in
an area outside the urban cemetery; these were separated by a sterile area. The first one
(Maisto-Boccia)presented55 tombs dating between the secondquarterofthe 4
th
century
and the beginningofthe 3
rd
century Be.The second, placedeastofthe former,included25
tombsdating betweenthesecondquarterand theend ofthe4
th
centuryBe.
According to Serritella, the tombs are widely distributed in space, thus differing
from otherdensely occupied cemeteries, and painted pottery is almostcompletely absent
from grave-goods,differingfrom othercontemporaryburials. Inall, 65% ofthe tombshad
grave-goods and 35% were void: of these, 7.4% had been disturbed and the grave-goods
probably dispersed and the remaining 27.6% had no grave-goods, though they had not
been disturbed. Each nucleus presents different behaviour and values as far as funerary
ritual isconcerned,even ifthey are homogeneous asregards chronologyand topographical
and spatial organisation.
Of164 tombs, 161had a single skeletonand only three were multiple: we decided
thus to simplify the structureof the database excluding the three multipletombs from our
data and obtaining a I-to-I correspondence between the tombs and the deceased.
Considering the multiple tombs, which moreover brought very little information
concerning the attributes relevant to our case-study, would require splicing data into two
tables which presentedno theoreticalproblembut introducedunnecessarycomplicationsin
the manipulationofqueriesthat have notyetapracticalinterface.
Data were originally entered into a database consisting of two tables: the tombs
(which include information on the deceased) and the grave-goods. Our attention
concentrates,atpresent,on thefirst, whichpresentsthe followingfields:
I. the IDof the tomb (fieldsMAP_IDand TOMB_ID, the formerused to link datato the
graphicalrepresentation)
2. the area, denotedby thename oftheowners(field PROPERTY)
3. the typologyof the tomb (field TYPOLOGY)
4. the presenceof "controfossa"(booleanfield CONTROFOSSA)
5. the presenceand type ofacover(fieldCOVER)
6. thedepth, lengthand width of the tomb, incm (numeric fields DEPTH, LENGTH, and
WIDTH)
7. theorientation,adescriptivefield wekept fordocumentation(field ORIENTATION)
8. the positionofthedeceased'sbody inthe tomb (field BODY_POSITION)
9. the rite, inhumationorcremation(field RITE).
10.the gender, distinguishing attribution based on osteological or archaeological
factors (fields GENDER_OSTand GENDER_ARCH)
II.the age,distinguishing betweenan archaeological attributionto large periodisation
(infant, child, young, adult, elderly) and a more preciseosteological attribution to
narrowerperiods,availableonly forapartoftheset (fieldsAGEand YEARS_OST)
12.thechronologicalinterval attributed,in25-yearsperiods(fieldCHRONOLOGY)
13.the presenceof grave-goods, to be used to avoid joining with grave-goodtable if
not necessary(booleanfield GOODS)
171
M. Crescioli et al. / Etruscan cemetery of Pontecagnano and fuzzy logic
Someofthe aboveattributesshow afuzzy nature derivingfrom the methodused toacquire
data or inherentto the nature ofthe attribute itself. Thus, genderand age are derivedbased
on osteological statistical methods and thus are worth fuzzifying; chronology has an
inherentlyfuzzy nature and must befuzzified also.
The assignment ofgender is mostly based on a statistical index that takes into
accountseveral factors and may vary from -2(female) to +2 (male). When it is basedon
archaeological considerations, the decision is made taking into account the grave-goods
andtheexperienceofthe researcheron similarexcavationsin Pontecagnano.
We decidedto transfertheavailabledataintoour fuzzy model using thefollowingrules:
convertthe osteological main index into a 0 - I value, to be assigned to female if
negativeand tomale ifpositive . .
give a conventional (low) value to the fuzzy function if the anthropologist
determined the gender based on other factors, when the main index gave result 0
(undecidable)
give a conventional (middle) value to gender assignment based on archaeological
factors.
It might be interestingtoask theresearchertostate directlythelattervalue, thus expressing
confidencein theassignmentof anumericscale. Obviouslythis was impossibleinourcase
and so we substituted the original researcher giving each assignment the same flat
evaluation. This procedureled to the introductionofa.newfield F_AGE of type fuzzy label,
with two possibleoptions,that is male andfemale, resuming two previousagefields in
the above described way. This simplification made age selections possible across the
originalderivationoftheir assignment,but preservingcompletelythe uncertainty.
The assignment of age also derived from two sources, with the complication that
anthropologistshad used twodifferentmethods leadingsometimestocontradictoryresults:
inthecase 4027,the ageinterval was20-25forone method, 40-46for theother..
Sothe rules forage fuzzificationwere thefollowing:
assign a fuzzy function with value I on the interval determined by osteology,
descendingto0onboth sideswithaslack of25% of.the ageinterval oneach side; .
if osteology gives two intervals, compute the fuzzy function as above for each and
take the highest value ofthefunction thus obtained
if the age is based on archaeological factors, assign conventional ranges (according
to Serritella'sassumptions) to each descriptive assignment and behave the sameas
in l.
Alsointhe case ofage, thefuzzy field F_AGE unifies two fields thatotherwisecould not be
comparedas aqueryselectioncondition.
The last field to be fuzzified is chronology, obtaining the field F_CHRONOLOGY. The rules
for thisoperationare verysimple:
establishchronological intervalscorrespondingto each description andassign value
I to thecorrespondingfuzzy function in thisinterval
give aslackof25years on both sides ofthe interval.
172 173
Beyond the Map
M. Crescioli et al. I Etruscan cemetery of Pontecagnano and fuzzY logic
Itmust beunderlined thattheF_AGEfieldcould notbeprocessed in the Fuzzy Relational
Database Management System (53], which provided only for "trapezoidal" fuzzy
attributes.
So,thefuzzyattributesarethefollowing:
Attribute name Attribute type Attribute content
F_GENDER FUZZY_LABEL Fuzzygender
F_AGE FUZZY_VALUE Fuzzyage
F_CHRONOLOGY FUZZY_VALUE Fuzzychronology
To make queries easier, we defined the following constants for each fuzzy attribute built
intothedatabase; thecouplesrepresenteachstate andthecorrespondingvalue ofthefuzzy
function.
Constants for the attribute F_GENDER
MALE()
FEMALE!)
Constants for the attribute F_AGE
INFANT(}
CHILD!)
YOUNG()
ADULT(}
ELDERLY(}
Constants for F_CHRONOLOGY
SECOND_QUARTER-4TH(}
THIRD_QUARTER-4TH(}
LAST_QUARTER_4TH()
FIRST_QUARTER_3RD(}
FIRST_HALF_4TH()
349.99,O}}
SECOND_HALF_4TH!}
299.99,O}}
FIRST_HALF_3RD(}
249.99,O}}
Constants for F_CHRONOLOGY
FOURTH(}
299.99,O}}
THIRD(}
199.99,O}}
FOURTH_THIRD()
275,O}}
7. Implementing tbefuzzyGIS
Definition
{{M,I}, {F,O}}
{{M,O}, {F,l}}
Definition
({O,O}.(O,l).{6,l},{B,O}}
{{5,0}.{7,1}.{14,1}.{16,O}}
({13,O},{15,1}.{20,l}.{22,0)}
({lB,0}.{21,1).{40,l},{43,O})
((35,0).{4l,I} {lOO,I}.{120,O}}
Definition
{{-3B5,0}.{-375,l}.{-350,l},(-340.0))
{{-360,0},(-350,l).(-325,1).(-315,0)}
{{-335,O}.(-325,l).{-300,1}.(-290,O)}
({-310.0).{-300,ll,{-275,l}.{-265,O}}
({-400,0),{-400.l},(-350,1),{-
{{-350.A}. (-350,l).{-300,I},{-
({-300,O),(-300,l).(-250,1).{-
Definition
({-400,O).(-400,1),(-300,1),{-
{{-300,0}.{-300,l}.{-200,l}.{-
({-325.0l,0).{-310,1}.{-290,1}.{-
Asdescribed above,thefuzzyGIShasbeenimplemented usingPostgreSQL andGRASS.
Thefirststageisthesetupofthedatabase:
1. create thefuzzydatatype
2. create thefuzzyequalityfunctionforbothfuzzytypes
3. create thecorresponding fuzzyoperator.
ThesestepsareeasilyperformedusingSQLstatements; onlythedefinitionofthefunctions
requiressomeCprogramming.
Loading data into the database is a straightforward matter: the geographic base is
converted from a .DXF file, using the GRASS function v.in.dxf, then every operation
canbeperformed throughtheTclfllc: interface.
After these initial steps, the system can be queried using SQL statements. While
awaiting more articulated reports, the HTML built-in output is quite satisfactory to
visualise theresults. Visualisation oftheresults ontheGIS still requires somegymnastics,
since neither the authors nor we have yet automated the operation and.:moreover, the
functiond .vect.pg hasnotyetagraphical interface.This taskwill beaccomplishedin the .
near future if the system proves to be useful, since it is quite simple to write a Tcl/Tk
interface using thedevelopment toolsdistributed withLinux. Somesample queriesand the
corresponding resultsfollow.
7.1 Search the database and retrieve similarity with "about 40 years old"
Wehavedecided tointerpretthelocution"about40yearsold" asfollows:


35 38 42 45 U
Figure 5.Graphofthe fuzzy set "about40 years old"
The SQLquery isthefollowing:
select tomb_id, Cage - '{{35.A}. {3B,I},{42,l},(45,O)}' from tomb;
The query retrieves all 161 records and gives for each the degree of similarity of F_AGE
with"about40yearsold", evenwhenthisisO. Wegiveonlyapartoftheresult:
Retrieved 161 rows 2 fields
toE> i4 ?coluaa?
4023 0
4024 0
4025 ' 0.466667
4026 1
4027 1
402B 0.25
4029 0
4030 0.555556
4031 0
Figure 6. Result ofthequery "about40 years old" (abridged)
7.2 Search tombs where the degree ofsimilarity to "about 40 years old" is greater than 0
This query differs from the previous one since it does not retrieve elements with 0
similarity.
TheSQLqueryis:
select tomb_id , {Cage - '({35,0).{3B,l),{42,l).{45,0}}') from tomb
where (f_age - '({35.0).{3B,i}.(42,l).(45,O})'}>O;

175
174
Beyond the Map
M. Crescioli et al. I Etruscan cemetery ofPontecagnano andfuzzy logic
Theresultisthefollowing: dated, andmay be easily omittedfromthe result; we avoidedthis togive asimpler query
example.
Retrieved 38 rows 2 fields
Retrieved 7 rows 2 fields
t.-b14 ?col....?
4012 1
4013 1
4014 1
4015 1
4016 1
4018 1
4020 1
4021 1
4025 0.466667
4026 1
4027 1
4028 0.25
4030 0.555556
4033 0.238095
4034 1
4035 1
4037 1
4038 0.555556
4039 1
4040 1
4042 1
4043 1
4045 0.555556
4047 1
4051 1
4058 1
4065 1
4083 0.238095
4084 1
4357 0.238095
4358 0.238095
5754 1
5760 1
5761 1
5762 1
5763 1
5765 1
5766 1
Figure 7.Resultofthe query"about40 yearsold" > 0
7.3 Group the distribution oftombs selected in the previous query in the different
chronological phases
Thequeryisthefollowing:
select chronology, count(tomb_id) from tomb
where (f_age - '{{35,Ol,{38,ll,{42,ll,{45,Oll'0
group by chronology;
The results are grouped by chronological phases and show the count of the number of
tombs foreach period. The first rowconcernstombs without grave--goodsthatcannot be
chroDology count
15
2nd quarter 4th cent. BC 5
Half 4th cent. BC 1
3rd quarter 4th cent. BC 6
4th quarter 4th cent. BC 4
End 4th - beg. 3rd cent. BC 3
1st quarter 3rd cent. BC 4
Figure 8.Resultofpreviousquery grouped bychronology
7.4 Group the distribution of "isfants" with respect to chronology .
This query uses thefuzzyconstant INFANT() andchecks for adegreeofsimilaritygreater
than0(actually,greaterthan0.01toavoiderrorsduetoround-off)
select chronology, count(tomb_id) from tomb
where (f_age - INFANT()1>0.01
group by chronology;
Theresultsare(asbefore,thefirstrowrelatestotombswithnograve-goods):
Retrieved 6 rows 2 fields
chronology count
Half 4th cent. BC
3rd quarter 4th cent. Be
15
5
2
End 4th cent. BC.
End 4th - beg. 3rd cent. BC
1
4
1st quarter 3rd ~ n t BC 4
I
. I
Figure9.Resultof "infants"grouped bychronology
This approach can be used to.ascertain whether the sample is representative, since it is
necessary to verify the composition of the burial group and especially'of the adult/child
ratio, according to theparameterstypicalof pre-industrial societies [18]. Asimilar query
caneasily give thisratio foreachof theperiods. The grave-goodsdatabase could then be
usedtoverifythecriterionof"mortuaryvariability" [60].
7.5 Maps derived from queries
Toderivemapsfromqueries,theprocedureisstill manual,although automatingitis nota
difficult task, usingabuilt-infunctionrequiring someparameters thatarefixed or can be
user-selected by meansof a windowinterface (for instance, thecolourof the lines to be
plotted). In a similar way, the GIS can be endowed with any sort of spatial analysis
functionsthatcanbeimplementedwithoutdifficultyandundercompleteusercontrol.
176
177
Beyond the Map
We havenotyet implemented theproximity function defined above, but since the
algorithmhasbeenstated,thistaskshouldalsopresentnodifficulty andwillbeperformed
inthenearfuture.
Graphical interrogationofthedatabase (i.e."point-and-dick")is,on thecontrary,
already available invoking a built-in function, named d.what.s.p9, which obtains the
featuresof anydisplayeditem simplybyclickingonit.Also thisfunction, notyetpresent
inthetcltkgrassmenu(themenudrivenversionofGrasswehaveusedforthismodel),can
be included in the menusystem to invokeit using the mouse and avoiding the command
line.
8. Conclusionsand perspectives
This paper concerns a methodology thathas been applied to a funerary context as a case
study. We intend to explore also other archaeological applications of FGIS. The method
lookspromising,and weare going todevelop itbothon the technicalside, implementing
more functions and a more complete graphical user interface, and on the archaeological
side to obtain further insight of social and cultural dynamics that can be gathered by
observing funerary data and their spatial properties, as, for instance, proximity and
orientation.
Another field under development concerns Web access to data, which will allow
disseminationofourspatialdatabases.
References
[I] D. W. Wheatley,Going overold ground: GIS, archaeologicaltheoryand the act of perception. In:J.
Andersen, T. Madsen and I. Scollar (eds.), Computing the past: Computer Applications and
Quantitative Methods in Archaeology, CAA92. Aarhus UniversityPress,Aarhus, 1993, pp. 133-138.
[2] V. Gaffney, Z.Staneic and H. Watson, The impactof GIS on archaeology: a personal perspective.
In: G. Lock and Z. Stancic (eds.), Archaeological and Geographical Information Systems: a
European Perspective. Taylor& Francis,London, 1995, pp.319-334.
(3) V. Gaffney and P. M. van Leusen, Postscript- GIS, Environmentaldeterminismand archaeology.
In: G. Lock and Z. Stancic (eds.), Archaeological and Geographical Information Systems: a
European Perspective. Taylor& Francis,London, 1995, pp.367-382.
[4] T. M. Harris and G. R. Lock, Toward an evaluation of GIS in European archaeology. The past,
presentand future of the theory and applications. In: G. Lock and Z.Stancic(eds.),Archaeological
and Geographical Information Systems: a European Perspective. Taylor& Francis, London, 1997,
pp.349-365.
(5) J. A. Barcelo, M. Pal/ares, A CritiqueofG.I.S. in Archeology. From Visual Seduction to Spatial
Analysis,Archeologia e Calcolatori, 7, 1996,pp. 313-326.
[6] M. Barnplon, Archaeology and GIS: the view from outside,Archeologia e Calcolatori 8, 1997, pp.
9-26.
(7) M. Gillings, Not drawing but waving? - ReHumanising GIS, the Titsa floodplain revisited. In: A.
Bietti, A. Cazzella, I. Johnson and A. Voorrips (eds.), Theoretical and Methodological Problems,
Proceedings of the Xl/lth International Congress of Prehistoric and Protohistoric Sciences, vol.l ,
Forll, 1m,pp.69--M.
(8) W. F.Limp, Developingmethodologiesin theanalysisof spatiallyreferenceddata and their impacts
on archaeological method and theory. In: A. Bietti,A. Cazzella,I. Johnson and A. Voorrips (eds.),
Theoretical and Methodological Problems, Proceedings of the Xlllth International Congress of
Prehistoric and Protohistoric Sciences, vol. I ,ForD, 1996, pp. 115-125.
M. Crescioli et al. i Etruscan cemetery ofPontecagnano andfuzzy logic
[9] J. McGlade, GIS and integrated archaeological knowledge systems- In: M. North and I. Johnson
(eds.), Archaeological Applications of GIS. Proceedings of Colloquium I/, UlSPP Xl/lth Congress
(Forti, Italy, September 1996). Sydney University Archaeological MethodsSeries, n. 5 (CD-ROM),
1997.
[101 A. Voorrips, Archaeological theoryand GIS, any relations? In; A. Bietti, A. Cazzella, I. Johnson
and A. Voorrips (eds.), Theoretical and Methodological Problems. Proceedings of the Xlllth
lsuemational Congress of Prehistoric and Protohistoric Sciences, vol. I, Forll, 1997,pp.209-214.
[III P. M Van Leusen, Viewshed and costsurface analysis using GIS (Cartographicmodellingin a cell-
based GIS Il), In: J. A. Barcelo, I. Briz and A. Vila (eds.), New Techniques for Old Time.
Proceedings ofthe CAA98 Conference. BAR InternationalSeries757. Archaeopress, Oxford. 1999,
pp.215-223.
[12] M. Cuozzo, Prospettive teoriche e metodologiche nell'interpretazione delle necropoli: la post-
processual archaeology, AlONArcStAntn.s.,3, 1996, pp.I-39.
[131 A. De Guio, "ArcheologiadellaComplessita"ecalcolatori: un percorsodi sopcavvivenzafra teorie
del caos, "attrattori strani", frattali e... frattaglie del postmoderno (I). In: M. Bernardi (ed.)
Archeologia del paesaggio, IV ciclo di Lezioni sulla Ricerca applicata in Archeologia; Certosa di
Pontignano (Siena), /4--26 gennaio /99/.All'insegnadelgiglio,Firenz.e, 1992,pp.305-389-.
[141 C. Parceo-Oubina, Deconstructing the Land: the Archaeology of sacred geographies. In:.-L.
Dingwall, S.c.Exon, V. L Gaffney,S. Laflin and P. M. Van Leusen (eds.), Archaeology in the Age
ot the lntemet, Proceedings ofthe CAA97 Conference. BAR InternationalSeries 750. Archaeopress,
Oxford,1999, pp.73-79.
[15] P. Moscati,GIS applicationsin Italian archaeology,Archeologia e Calcolatori 9, 1998,pp.191-236.
[161 L. GarciaSanjuan and D. W. Wheatley, The state ofthe Arc: differential rates of adoption ofGIS
for Europeanheritage management,European Journal ofArchaeology 2, 1999,pp.2,201-228.
[171 K. L. Kvamme, Geographic Information Systems in regional archaeological research and data
management. In: M. B. Shiffer (ed.) Method and Theory in Archaeology. University ofArizona
Press,Tucson, 1989,pp. 139-202.
[181 M. Cuozzo, Patterns of organisation and funerary customs in the cemetery of Ponteeagnano
(Salerno)duringtheorientalisingperiod, Journal of European Archaeology 2, 1994,pp.2, 263-298.
[l9J B.d'Agostino,Societadei vivi, comunitadei morti: un rapportodifficile, DialArch 1.3, 1985, ills.,
pp.47-58.
[20J B. d'Agostino, Problemi di interpretazione delle necropoli. In: R. Francovich and D. Manacorda
(eds.), Lo scavo archeologico dalla diagnosi all'edizione, l/l ciclo di Lezioni sulla Ricerca applicata
in Archeologia; Certosa di Pontignano (Siena), 6-18 novembre 1989. A1I'insegna del giglio,
Firenze, 1990, pp.401-420.
[21J L.A.Zadeh,Fuzzy Sets, In[onna/ion and control 8, 1965, pp.338-353.
[22) M. Thaller, Tbe need for aTheory of Historical Computing. In: P. Denley, S. Fogelvik and C.
Harvey(eds.), History and Computing ll, Manchester, 1989.
(23) M. Thaller,Tbe Winds of Change.Problemsofa DatabankOrientedSystemUsing the Conceptof
Fuzzy Sets. Paper presentedat the 1981Joint Conferenceof IFDOand IASSIST, Grenoble, 14-18
September1981. ..
[24J M. Thaller,The HistoricalWorkstationProject,Gottingen, 1990.
[25J O. Signore and R. Bartoli, Managing art history fuzzy dates; an application in historical-
geographicalauthority,Historical Social Research 14, 1989,p.3.
[26] E. Canal and S. Cavazzoni, Antichi insediamenli antropici nella laguna di Venezia: Analisi
multivariataditipo"FuzzyC-MeansClustering",Archeologia e Calcolatori I, 1990,pp.l65-177.
[27J L. I. Borodkin, Fuzziness in Multivariate Arialysis ofHistorical DataUsing a Pattern Recognition
AlgorithmBasedon Fuzzy Set Concepts.In: L.Borodkin,M. Thallerand J.Turner(eds.), Statistics
for Historians: Standard Packages and Specific Historical Software. Max-Planck-Institut tur
Geschichte,ScriptaMercalUracVerlag,St.Katharinen, 1995, pp. 95-104..
.''''''''''';''_'
178 179
Beyond the Map
[28] L. I. BorodkinandI. M.Garskova, Analytical procedure formultidimensional hierarchical data. In:
B. Genito and M. G. Moskova (eds.), Statistical Analysis of Burial Customs of the Sauromatian
Period in Asian Sarmata (6'"- 4" Centuries BC). Napoli, 1995,pp.63-114.
[29] D. Dumitrescu, Gh. Lazarovic and H. F. Pop, Fuzzy clustering in archaeology (in Romanian),
Symposium of Computer Science Applied in Socio--Human Sciences, Cluj-Napoca, September
1989.
[30] F. L. MedeletandH.F.Pop,Fuzzyclusteringofmonetarydiscoveries of lateneepoquein Banat(in
Romanian),FourthNationalSymposiumofArchaeometry,C1uj-Napoca, September 19-20, 1991.
[31] D. Dumitrescu and H. F. Pop, Qualitative fuzzy clustering in archaeology (in Romanian), 5"
NationalSymposiumofArchaeometry,Cluj-Napoca,March 1992.
[32) H. F. Pop and C. Sarbu, The fuzzy hierarchical cross-clustering algorithm. Improvements and
comparative study,Journal of Chemical Iflformatiofl and Computer Sciences 37, 1997,pp.51Q.-516.
[33] J. Bagg, Working with Uncertainty, Paper presented at the IX International AHC Conference,
Moscow, 1996.
[34) J. Bagg and Nick Ryan, Modeling historical changes in Southern Corsica. In: Z. Kemp (ed.),
Innovations in GIS. Taylor & Francis,London, 1997,pp.42-55.
[35] F. Niccolucci, G. Pardi and T. Zoppi, Un archivio georeferenziato di insediarnenti archeologici,
Archeologia e Calcolatori 7,1996,pp.I61-177.
[36] A. Benvenuti andF. Niccolucci, Historical objects withindeterminate boundaries. Paper presented
atthe IXInternational AHCConference,Moscow,1996.
[37] J. Puyol-Gruart, Computer Science, Artificial Intelligence and Archaeology. In: J. A. Barcelo, I.
BrizandA.Vila(eds.),New Techniques for Old Time, Proceedings of the CAA98 Conference. BAR
InternationalSeries757. Archaeopress,Oxford, 1999,pp. 19-27.
[38] J.A.Barcelo,Heuristicclassification and fuzzysets.Newtools forarchaeological typologies.In:H.
Kamermansand K.Fennema (eds.),Interfacingthe Past,CAA95,Analecta Praehistorica Leidensia
28, 1996,pp.I55-164.
[39] C. Reeler, Fuzzy logic, neural networks and the analysis ofpa sites. In: M. North and I. Johnson
(eds.),Archaeological Applications of GIS, Proceedings of Colloquium ll, UISPP X///th Congress
(Forti, Italy, September 1996). SydneyUniversityArchaeological Methods Series, n.5(CD-ROM),
1997.
[40] P. A. Burrough and A. U. Frank, Geographic Objects with Indeterminate Boundaries. Taylor &
Francis,London, 1996.
[41] N.Ryan,ManagingComplexity: Archaeologicalinformationsystems past, presentand future.Paper
presented at the British Association Annual Festival of Science, University of Birmingham, 8-13
September, 1996.
[42] T. Kirkinen,GIS-assistedDataAnalysis- FindingMeanings inComplex SpatialDataSets.ln:J.A.
Barcelo, I. Briz and A. Vila (eds.), New Techniques for Old Time, Proceedings of the CAA98
Conference. BARInternational Series757.Archaeopress,Oxford, 1999,pp.255-258.
[43) D. W. Wheatley, Cumulative Viewsbed Analysis: a GIS-based method for investigating
intervisibility,andits archaeological application. In:G. Lock and Z. Staucic (eds.),Archaeological
and Geographical Information Systems: a European Perspective. Taylor & Francis, London, 1995,
pp. 171-186.
[44] P.Fisher,Firstexperiments in viewsbeduncertainty:simulating fuzzy viewsbeds, Photogrammetric
Eflgineering Sensing 58, 1992,pp.3,345-352.
[45] K. Nackaerts, G. Govers and L. Loots,The use of Monte Carlo techniques for the estimation of
visibility. In: L. Dingwall, S. C. Exon, V. L. Gaffney, S. Laflin and P. M. Van Leusen (eds.),
Archaeology in the Age ot the lnsernet, Proceedings of the CAA97 conference. BAR International
Series750.Archaeopress,Oxford, 1999, pp.63-{i5.
[46] L. Loots,K. Naclcaertsand M. WaelkensFuzzy viewsbed analysis of the hellenistic City.Defence
system at Sagalassos. In: L. Dingwall, S.C. Exon, V.L. Gaffney, S. Laflin and P.M. Van Leusen
M. Crescioli et al. / Etruscan cemetery of Pontecagnano and fuzzy logic
(eds.), Archaeology in the Age of the Internet, Proceedings of the CAA97 Conference. BAR
International Series750.Archaeopress,Oxford, 1999,pp.63-{i5.
[47] M. Schneider, Uncertainty Management.for Spatial Data in Databases: Fuzzy SpatialData Types.
In:R. H.Guting, D.PapadiasandF. Lochovsky (eds.),Advances in SpaJwl Databases, Proceedings
of the 6" lnsemational Symposium, SSD '99, Hong Kong, Lecture Notes in Computet Sciences
1651.Springer, Berlin, 1999.
[48] Chorochronos site(December 1999):http://www.dbnet.ntua.gr/-choros/
[49] E. Stefanakis, M. Vazirgiannis and T. Sellis, Incorporating Fuzzy Logic Methodologies into GIS
Opera/ions, Proceedings of the lnterruuional COnference in Urban, Environmental and Regional
Planning, Athens, 1996.
[50] R. R. YagerandD.P.Filev,Essentials ofFIi<zy Modeliflg and Control. Wiley,NewYork,1994.
[51] L. A. Zadeh, A computational approach to fuzzy quantifiers in natural languages, Computing and
Mathemalics with Applications 9,1983,pp.149-184.
[52] E.L. Usery, AConceptual Framework and FuzzySet Implementation forGeographic Features. In:
P. A.Burrough andA.U. Frank(eds.), Geographic Objects with Indetermiruue Boundaries. Taylor
& Francis, London, 1996,pp.71--86. .
[53] J. Galindo, J. M. Medina, O. Pons and J. C. Cubero, A Server for Fuzzy SQL Queries. In: T.
Andreasen, H. Christiansen, and H. L Larsen (eds.), Flexible Query Aflsweriflg Systems,
Proceedings of the Third lntemationai Conference. FQAS '98, Roskilde. LectureNotesinArtificial
Intelligence 1495.Springer,Berlin, 1998,pp. 164-174
[54] J. M. Medina, O. Pons and M. A. Vila, GEFRED. A Generalised Model of Fuzzy Relational
Databases, lnformauon Sciences 76, 1994, pp, 1-2,87-109
[55] A. D'Andrea, R. De Nicola and A.Giordano, The Eurialo project: a vectorGIS for the integrated
management ofthe archaeological dataofPontecagnano, Italy.In:J. A.Barcelo, I. BrizandA.Vila
(eds.), New Techniques for Old Time, Proceedings of the CAA98 Conference. BAR International
Series757. Archaeopress,Oxford, 1999,pp. 145-148.
[56] A. D'Andrea, II GIS nella produzione delle carte dell'impallo archeologico: l'esempio di
Pontecagnano Archeologia e Calcolatori 10,1999,in press. .
[57] A. Serritella, Pontecagnano. //.3. Le fluove aree di necropoli del IV e /II sec. a. C. Istituto
Universitario Orientale, Napoli. 1995.
[58] P.P. Petrone,Analisipaleodemografica e paleopatologica tombeinproprieta Rossomando. In
[57],Appendix I,pp.
[59] C.ScarsiniandR. Bigazzi,Studioantropologico deirestiumani.In[57],AppendixIl,pp. 135-148.
[60] J.O'Shea,Mortuary variability. AcademicPress, NewYork,1984.

180
181
BeyondtheMap
G. Lock (Ed.)
lOS Press, 2000
Session3discussion:
Commentsonarchaeologicalprediction
MartinKuna
ArcheologickyustavAVeR
Prague
1. Introduction
The ideaofprediction inarchaeology resultedfrom theobservation thatthearchaeological
recordisusually notscattered randomly overthelandscape, butispatterned with respectto
its geographic setting, and to settlement organisation. Although this observation was first
made along time ago, itonly became attractive inthe 1960sand 1970s when archaeology
started topay more attention to structural aspects ofthepast (in contrast to the search for
'events' typical of the previous historical paradigm [11]. Research on settlement patterns
and locational analysis was recognised as one of the keys to economy, ecology,
demographyandotherstructural aspects ofpastsocial systems. The recognition ofpatterns
orrules inthedistributionofpast settlement hasbecome equally attractive inthesphere of
archaeological heritage management offering a chance of making it cheaper and more
effective.
From the theoretical point of view, the use of predictive modelling in
archaeologicalheritage managementisquitejustified.It would obviously not becorrect to
only look after sites thatare already known andcan belisted. Some attention mustalso be
paid to the archaeological record that isnot known and cannot be accurately located since
this contains new and potentially even more important information. Archaeological
heritage management must, therefore, employ means other than just the care of known
historic monuments andnaturalprotected areas,etc.
A real boom in predictive modelling and its use for heritage management came
with the appearance of geographical information systems in the middle 19808[1]. Since
then,fascination with the new technology anditspotential for thetreatment of spatial data
has become so wide spread and so strong that many archaeologists worried about a new
wave of environmental determinism penetrating archaeology along with it. Most GIS-
basedstudies, itis true,still work witha limitedrangeof geographical variables and some
fear of environmental simplification is, hence, understandable. It is, however, doubtful if
the technology itself is the only reason for this unsatisfactory situation. Rather than in
methodology, we should look for a deficit in archaeological theory and, of course, in the
lack ofawider rangeof relevant geographicaldataavailable indigital format.
Geographical information systems are indisputably an extremely efficient tool to
cope with various types of spatial data. To fully exploit their potential for predictive
modelling archaeology needs to develop appropriate theory defining the wide range of
variables describing the natural and social world that are responsible for the behaviour of
M. KIUIQ/ Session 3discussion
people in their geographical environment. Only then will it be possible to avoid both an.
unreasoned optimism and a hypercritical scepticism about the use and reliability of
prediction inarchaeology.
2. Whattopredict?
The first problem with archaeologicalprediction, surprisingly, concerns the veryobjectof
prediction - the definition of the units to be predicted. In Central Europe, modern field
surveys (mainly using aerial photography and systematic field walking) invariably show
thatthe archaeologicalrecordinalandscape isspread muchmore widely andstructuredin
a more complex way than archaeologists used to believe. As a consequence of modern
surveys, archaeological information often covers the cultural landscape in a continuous
way over many square kilometres, although thedensity ofdata may vary locally,Pigure 1.
This kind ofdistribution results from both the continuous way of usinglandscapes in the
past (landscape is a continuum of activity areas of different functions (10)) and from the
specific structuring andre-useofsettlementareasinprehistory.
Archaeological prediction cannot provide reasonable results if it is not- using
appropriateoperational concepts. Forthesituationdescribed above ithasbecome clear that
-CD-
o
~
o
D,
o

&2:
" 0
81
o
o
a
a
Figure I. Prehistoricdata from Central VinofBkjpotok valley.
Sizeof squaresmarksclasses1-10,11-25,26-50and>50fragments
perI hectare. Walkedwith20metre spacing
Hatchedareas:buill-upareas(villages)
= findsbefore 1980.;
o= finds from 1980. (mostlyinpipe-lines)
0=prehistoricpotteryinsurfaceartefactsurvey 1996-1998.
-, I
I
183 182
Beyond the Map
archaeological landscape data can hardly be classified using the traditional concept of 'the
site' [5], [6], [7], [8]. Sites as remarkable concentrations of finds are usually not discrete
spatial units and they do not occur in an empty landscape but in a continuum of
archaeological information. Further, they are units of the archaeological record, the pattern
of which has been mainly created by archaeological transformations and not by the past
behaviour of people. In particular, sites usually represent a very biased sample of activity
areas, both chronologically and functionally. The rest of the record still contains a lot of
information, e.g. the activity areas of those archaeological cultures that did not produce
many durable remains, areas which were used not so often, or areas that were used for
reasons other than residential or funerary activities.
The continuous distribution of the archaeological record has, in fact, a complex
internal structure. Predictive models must pay attention to this and be able to reflect the
high variability and the wide spread of archaeological information over space. They can
hardly be very effective unless it is lcnownquite precisely which category of data is to be
predicted and how it can be defined within the overall continuous distribution of
archaeological data [12].
3. Source data
The most usual way to perform archaeological prediction is to measure find densities
within various categories of environment in a landscape sample, and to extrapolate this
information onto other parts of the territory. Sampling the archaeological record, however,
usually brings its own problems and does not always produce unambiguous results. The
source of the bias may be in both the archaeological transformations [II] and in the
method of collecting the archaeological sample itself.
A typical example of the former situation may be archaeological evidence in valley
bottoms, which although intensively used for various purposes throughout prehistory
usually only produce a very low density of finds. Most of the floodplain was either covered
by later deposits or eroded away by the river (e.g. more than 60% of the Mesolithic surface
within the Labe river floodplain in Central Bohemia has been removed by erosion [4]). If
prediction is based upon these data it may result in a correct estimation of the surviving
archaeological sites in similar eroded environments but it can hardly estimate the original
role of floodplain areas in the past.
Data from the available sample always depend on the method of sampling used.
Information from earlier finds and rescue excavations, for example, usually concentrate at
built-up areas (villages) and their immediate surroundings (industrial zones like sandpits,
clay-pits, etc.), and these areas have been intensively occupied since prehistory but are not
available for survey today. Data from recent surveys may often relate to a completely
different part of the landscape, spatially distinct, and, therefore, showing different
chronological and functional patterns due to the methods used, Figure I. In the end, it may
not be easy to decide which aspects of the observed pattern reflects past reality and may be
confidently used as a basis for prediction.
4. Inductive or deductive approach?
The impact of post-depositional transformations may be avoided if prediction is drawn not
from empirical data but from a theoretical model based upon the expected needs of past
communities in their environment, a possibility developed by Hans Kamermans in this
volume. Past settlement systems certainly paid attention to landscape variables (like the
M. Kuna I Session 3 discussion
quality of soil, climate, etc.) which can be defined by observing and generalising either
from societies which still exist, or from those in the past which are better recorded. Our
experience shows, however, that these variables rarely work entirely in the context of the
past and there is always some part of the data that can not be explained in this way. The
main problem of the deductive approach is that any deduction can hardly predict spatial
rules resulting from specific cultural relations and symbolic meanings created by people in
the past. Such relations certainly existed in prehistoric Europe but they need not have any
analogies within the world today. Let us consider, for example, the location of prehistoric
hill-forts and ritual enclosures in' relation to their surrounding territory and the changing
location of settlements with regard to local relief, etc. These can be. explained by specific
symbolic structures of the past social world rather than by practical aspects of the
environment. Specific cultural meanings of landscape may have been stronger, of course,
on the local scale and their impact may have faded out in the context of larger geographical
units (regions). This difference in scale may, therefore, contribute to a general framework
for the credibility of the inductive/deductive approach.
5. Factors of settlement location
Predictive models often consider just basic factors of the natural 'environment: altitude,
distance from a water source, surface geology, slope gradient and aspect, etc. Even the
application of these is not easy and does not always produce unambiguous results. Only the
first two of the variables listed above are generally valid, although as predictive factors
they are still too general. Other factors usually lead to the recognition. of certain trends but
rarely any strict rules, and in the case of some of them (e.g. slope and aspectjno significant
results have been obtained so far. This is supported by Stancic and Veljanovski for the
island of Brae (this volume) although the pattern of significant factors in their case is
slightly different from that in our data (6], [8], [12]. There is, however, one potential
problem in analyzing such landscapes, that is the possibility of mutual correlation among
variables making it difficult to understand what the observed patterns really mean. In the
case of Brae, for example, the preference for north-eastem slopes by seaport sites may
result not from the aspect itself but from the direction in which the mainland is situated
and, therefore, where the most intensive communication was practiced. Similarly, the
higher values of slope gradient at the same sites may not reflect greater preference for
steeper slopes by their inhabitants but merely the character of the coast, the distance from
which was the vital factor. .
It is, hence, quite probable that prehistoric activities were less dependent upon the
attributes of the environment than it is generally supposed [II], (12]. This corresponds to
the presupposition that there might have been other relevant attributes of the landscape that
were paid attention to not because of their practical aspects but because of their symbolic
meaning. We can see in Table 1, for example, that Bronze Agesettlement.tends to occupy
elevated valley edges whereas during the Roman Period the opposite is clearly visible.
Most settlements of the later period appear on the lowest parts of slopes, quite close
to streams or river floodplains. In some periods settlements prefer places delimited by
terrain edges on more than one side (promontories, hill tops), in other times such places are
avoided. Usually, we do not have any economic or ecological explanations for these
patterns and it is not sure we will ever find them. It is possible, however, that these patterns
result from purely cultural preferences or symbolic reasons [9].
The location of prehistoric. activity areas was also influenced by the. complexity of
past landscapes. In the record we can usually only identify certain
184 185
Beyond the Map M. Kuna/ Session3 discussion
Table I. Vertical distance of prehistoric settlements from the valley bottom in (a) metres
and (b) percentages of the maxima available. Given in numbers of settlements. Table 2. Spatial correlation among settlements of the Late Bronze Age to Roman Periods.
Vinoisky potok project, Central Bohemia [6]. Vinoisky potok project, Central Bohemia [6].
Venial distance LBA -FBA H.lIstatt - L8 Tene ROlDlloPeriod
0-5m 4 10 11
5 -10m II 6 5
10 - 15 m 7 9 -
15-20m 3 2 1
>20m - 4
0-25 % 3 10 II
25-50% 6 5 5
50-75 % 6 7 -
75-100% 10 9 1
L8teBrooze
Age
1300-1000 BC
Final Bronze
Age
10l10-750BC
Hallstatt
period
(7S0400BC)
L8Tene
period
(44lOBC- en
ROlDllo
period'
(0-400 AD)
Late Bronze Aae 1.00 0.66 0.25 0.26 0.13
Final Bronze Aze 0.66 1.00 0.35 0.43 0.13
Hallstatt oeriod 0.25 0.35 1.00 0.53 0.13
La TeneOeriod 0.26 0.43 0.53 1.00 0.18
Roman oeriod 0.13 0.13 0.13 0.18 1.00
activities, mainly residential and funerary ones. We should, however, also try to see other
types of activities that are necessary within every community area, for example fields,
pastures, and woodland as a source of fodder and fuel [10]. These less visible activities
must have occupied even larger areas in the landscape than the more obvious ones and they
probably were even more stable through space and time than settlement areas themselves.
The patterning of these less visible activities might have been the constraint for the spread
of settlement and, hence, the reason why the culturally homogeneous remains are usually
spatially clustered. A large part of the landscape always appears as empty if the remains of
only one archaeological period are considered, even in an optimal natural environment.
This need not be interpreted in terms of low population density or in a lack of
archaeological data but rather in terms of functional variability and complexity within the
prehistoric landscape.
The expected stability of the less visible activity areas (prehistoric fields, pastures,
ritual areas etc.) may also explain the long-term continuity of residential areas which is
well recorded in the archaeological data, Table 2. Although prehistoric residential areas
were often rebuilt and relocated in space their shifts were over short distances, being
limited by the structure of the surrounding landscape. This is why residential areas usually
show a strong correlation with areas of the immediately preceding periods. The longer the
time span between two periods, the lower the spatial correlation between their activity
areas. Looking for locational factors of prehistoric settlement we have, therefore, to also
consider the history of the settlement system, the factors of landscape antecedent [3], [13]
created by the previous use of the landscape by its inhabitants.
6. Approaches to environmental variables
Individual factors of the natural environment used for prediction differ in their general
character, e.g. they have a different spatial impact. Some are strictly local (e.g. slope
gradient), others may work within a radius of several hundred metres (the quality of soil
-
used for arable fields) or even over distances of kilometres (the sources of some raw
materials).
According to the scale of impact of these factors the accuracy of prediction will be
different (here accuracy can be thought of as the cell size in a raster data structure). The
ability of some factors to influence activity patterns over large distances may introduce
some ambiguities into locational analyses: is it, for example, more appropriate to evaluate
a locus according to the soil type in that exact place or according to the optimum available
within a reasonable radius?
In prehistory, characteristics of landscape were not perceived quantitatively in a linear way
as that used in predictive procedures. It is doubtful if our quantification corresponds to the
way the landscape was understood and classified in the past. It is not certain, for example,
if the distance of one kilometre from a stream was understood as being twice as good as a
distance of two kilometres. Rather than measuring on a linear scale people in the past
probably made simple distinctions between 'good' and 'bad' place's.There must have been,
of course, a zone of transition between the two that cannot have been quite clearly
definable. For this, the theory of fuzzy sets may bring new insights since this concept is
equally valid both in formal and in geographical space. It is very useful, therefore, that this
topic is discussed by Crescioli, D'Andrea and Niccolucci in this volume.
The evaluation of loci to be used as activity areas was also dependent on their
wider landscape context, particularly regarding the availability of relevant attributes in the
surrounding area. We know, for example, that most Bronze Age settlements in Bohemia
were preferably located in higher positions, mostly on elevated valley edges. In a typical
landscape of Central Bohemia this usually means a vertical distance of 10-20 metres from
the valley bottom. There are, however, many areas in Central Bohemia where loci like this
do not exist because of the flat relief, although these areas were, of course, also occupied
during the Bronze Age. The prediction rule for the occurrence of Bronze Age settlements
should not, therefore, be defined by absolute values of vertical distance but rather by a
relative value relating the actual vertical distance to the maximum value available in the
vicinity. 'Bronze Age settlements were located 'as high as possible' in the particular
landscape, this may mean an elevation oftwenty, or sometimes only two metres, Table I.
7. Why predict?
In theory archaeological prediction may save on costs of building activities and introduce
better efficiency to archaeological heritage management. Predictive models can also be
used in research projects of settlement and landscape archaeology-as tools to malce
archaeological surveys cheaper and' more effective. It is necessary, however; to evaluate
the potentials and risks correctly.
First, the results of prediction must be understood for what they are, i.e.
probabilistic models, not as maps of past settlement systems or the archaeological record.
It is always necessary to maintain a feedback, testing and correcting methods of prediction
used in any particular case.
It is also necessary to admit ~ the areas predicted as archaeologically rich must
contain many places which are quite empty, at least in terms of what archaeology can
recognise. In the same way, archaeological data will always be found outside the predicted
areas. From the point of view of modem industrial activities, predictive models- may divide
-; ----_..- - ~ = 7 ~ - - : - - - . ~
187 186
BeyondtheMap
the landscape into safer and more dangerous areas, this division must not, however, be
translated in terms of the importance of these areas for archaeology. A high density of
finds does not automatically mean a greater importance,justthe opposite may be true, in
fact, as a discoveryof a rare archaeological situationoutside the predicted territorycould
be more importantthan many otherfinds in areas where their occurrence is expected (2],
(12]. Fromthis pointof view, prediction is not an end in itself, it isjusta methodological
tool tobe used withinaparticularframeworkof acertain theoreticalapproach.
References
[I] Allen, K.M.S., Green, S.W. and Zubrow, E.B.W. (eds), Interpreting space: GIS and archaeology,
London,Taylor&Francis, 1990.
[2J Allschul, J.H. Red flag models: the use of modelling in managementcontexts, in Allen, K.M.S., Green,
S.W. and Zubrow, E.B.W. (eds), Interpreting space: GIS and archaeology, London,Taylor& Francis,
1990, pp. 226-238.
[3] Benes, J. Ke koncepci krajinne archeologie - Towards a conception of landscape archaeology,
Archeologickirozhledy45, No.3, 1993,pp. 404-417.
[41Dresterova,D.Theprehistoryofthe Middle Labe (Elbe) floodplain in the light ofarchaeologicalfinds -
Vyvoj nivy stfcdnlho Labe ve svetle arcbeologickych mllezll, Pamatky archeologicke 86 [21, 1995,
pp.105-145.
[51 Kuna, M. Archeologicky prdzkum povrchovymi sbby - Archaeological survey by surface collection,
Praha,Zprllvy CAS - Supplement23, 1994.
[6] Kuna, M. Geograficky informacni system a vyzkum pravW sidelni struktury, in Machacek, J. (ed.),
Po(!{taoovapodporavarcheologii,Bmo(MU), 1997, pp.173-192.
[7] Kuna, M. Methodofsurvey in landscapestudies, in: Neustupny, E. (ed.) Space inprehistoric Bohemia,
Praha,lnstituteofArchaeology, 1998, pp.77-83.
[8] Kuna, M. The memory oflandscapes, in: Neustupny, E. (cd) Space in prehistoric Bohemia, Praha,
InstituteofArchaeology, 1998, pp. 106-115.
[91Kuna, M. and Adelsbergerova, D. Prehistoric location preferences: an applicationofGIS to the Vinofsky
potok project, Bohemia, the Czech Republic, in Lock, G. and Stancic, Z. (eds), Archaeology and
GeographicalInformation Systems, London,Taylorand Francis, 1995, pp.II7-131.
[101Neustupny, E. SidelniarelI.Iypravekych zemedelcu - Settlementareas ofprehistoric farmers, Pamatlcy
archeologicke77, 1986, pp. 226-234.
[IIINeustupny,E. (ed.) Space inprehistoricBohemia, Praha, Instituteof Archaeology, 1998.
[121 Neustupny, E. Predikce ared/11archeologickeho ztijmu. Monumentum Jan RulJ, Praha, Institute of
Archaeology, inpress.
[13] Roberts, B.K. Landscape archaeology, in: Wagstaff, J.M. (ed.), LAndscape andculture. Geographical
and archaeologicalperspectives, 1987, pp.77-95.
Beyondthe Map
G. Lock (Ed.)
lOS Press, 2000
AncientandnewPompeii: aproject
formonitoringarchaeological
sitesindensely populatedareas
FedericaGiannini I,MariaTeresaPareschi I, GreteStefani",MarinaBissonI,
I CNR-CentrodiStudio perlaGeologia
StrutturaleeDinamica dell'Appennino, Pisa
2SoprimendenzaArcheologicadi Pompei. Naples
Abstcact: In this paperweproposea GIS for the ancient city of Pompeii, destroyed
by the eruptionofVesuviusin 79 AD.Itis based ona photogranunetric survey(13
stereoscopic models at a scale of I :3,(00)of the area together with fieldwork. The
absoluteerrorinthe location of geometric features is less than 100m.TheGIS ofthe
ancient city is integrated with that ofthe new town. The GIS ofancient Pompeii
includes regions, insulae and buildings, toponyms, streets, pavements, walls, and
green areas. It enables: i) the establishment of accurate dimensions andlocations
without the limitationsimposedbyscale; ii) the combinationoflayers ofinterestinto
single thematic maps; iii) the storage ofalphanumeric andgeographicdata; and iv)
topologicalquerying.
1. Introduction
The incomparable value of the archaeological site of Pompeii makes itessentialto protect
this area. This can only be done by means of a surveyand a highlyaccurateplan ofall the
currently knownarchaeologicalfeatures and protected areas. This shouldalso include the
previousdiscoveries,someof which have,beenre-buriedfor reasonsofconservationordue
to a lack offunds. Ironically, the AD 79 eruption deposits that buried Pompeii, for mariy
centuries actually guaranteed the integrity ofits fragile structures. Today, modem urban
growthand increasingpopulationdensity, togetherwith a lack ofplanningregulations, are
threatening the excavated area ofPompeii. Moreover, as we shall see later, due to its
proximityto Vesuvius,Pompeiistill lies ina regionof volcanicrisk (Figure I).
The monitoringofthis area is based' on the evaluation ofvolcanic risk and onthe
identification and classification ofcurrently visible archaeological features. All of this
information is then digitised and organised in a GIS, a particularly flexible and easily
updateabletool. The GIS for the archaeologicalarea ofPompeii,is intum integrated intoa
largerscale GIS of the Pompeiimunicipalityand of the Vesuvianarea. Thevisualisationof
the territory,necessaryto monitoritsevolutionthroughtime, iscarriedout throughthe GIS
which contains more than 400 natural and artificial thematic layers, and about 200
archaeological layers (all in vector form). These layers can all be superimposed on the
shadowedimageofaDigital ElevationModel(OEM)ofthe area.
189
188
F. Gi<UUlini et 01. I Ancient and new Pompeii
Beyond the Map
1971to26,1l8inJanuary 1999.The areaof themunicipal district is 1,244hectares; built-
up zones (including civil and industrial buildings, houses, minor buildings, etc.) cover an
area of 1I0.51 hectares, an increase from about 90 hectares in 1987. Besides the
excavations of theancienttown,thereareabout fortyknown archaeologicalsiteswithin the
municipal district, covering an area of 62hectares, of which 51 hectares have beeri
excavated (fableI).
TableI. The Pompeiimunicipaldistrict
Area
(Km
')
Average density
(iobIkm')
Buildings
(1991)
BuDding
.ctivitie;
11991l
PopUlation
(1971)
.Population
(1981)
PopaJ.tion
(1994)
Population
(1999)
12 2,036 7,770 1,800 21,547 22,896 25,177 26,118
Figure 1.The territoryaround Vesuvius
In this paper, we shall briefly outline the particular events of the AD 79 eruption and its
deposits,thepopulation factors, thevolcanicriskelementsofthisareaandtheGIS.
2. Theplinianeruptionof AD79
On the 24
th
August, AD 79, Vesuvius suddenly awoke and the deposits of a plinian
eruption buried the towns of Herculaneum, Oplontis, Stabiae and Pompeii (a plinian
eruption isahighenergyexplosive eruption). Arainoflapilli and ash fellandcovered the
landwithathickblanketofpyroclastic deposits.Threemainphasescan bedistinguished in
the AD 79 eruption, corresponding to changes in eruptive dynamics and, consequently,
variationsinthedeposits [I],[2],[3].Thethreephaseswere:
I) aphreatomagmatic openingphase;
2)amainplinianphase;
3)aphreatomagmatic finalphase.
Each of these phases is characterised by different deposits and only some of the
phreatomagmaticunitsoftheAD79depositsreachedPompeii[4].
3. Populationdensity
The plinian deposits of theAD 79 eruption destroyed thecity of Pompeii but, at thesame
time, have protected it from the degradation of following centuries. Today, however, the
increasing population density in the areas surrounding Vesuvius and the ever expanding
urbanisation of the area, together with the lack of a specific land use plan, threatens the
archaeological features within the Pompeii municipal district. The population of new
Pompeii has increased by more than 20% in the last 20 years,from 21,547 inhabitants in
4.Vokanicrisk
Asmentioned previously,theproximity toVesuvius meansthat Pompeiiisverymuch ina
volcanic risk area.The lasteruption wasin 1944and since thattimeithas beenquiet with
noevidence of renewedactivity.However, the volcano ischaracterised by longperiods of
rest followed by eruptions of increasingly greater energy, depending on the length of the
restperiod. ThisputstheareaaroundVesuvius atgreatriskgiving risetoconsiderationsof
amaximum expected event andtheeruptivescenario associated withsuchanevent.
4.1 The maximum expected event
After the 1944 eruption Vesuvius changed in character from an open conduit to an
obstructed conduit condition. Usually, the renewal of activity after obstructed conduit
periodshasfollowed alongrestperiodandresultsinmainlyexplosive eruptionsofvarying
size andenergy. If we presume there has been a constant feeding of the magmachamber
since thelast eruption, theestimated volumeof magmaentering themagmachamber until
3
thepresent dayshouldbe2x10
8
m . If this volume of magma wereejected during a single
event it would constitute a subplinian eruption. The eruption of AD 1631 had similar
characteristics tosuchanhypothesizedevent, socan beconsidered asrepresentative of the
maximumexpectedevent[5].
4.2 The reconstructed scenario
By studying the AD 1631deposits, Figure 2, and theresults of computer simulations, a
scenariofor themaximumexpectedeventhasbeendefined.Santacroceetal.,P], proposed'
thefollowing sequenceofevents:
1) aphreatomagmatic opening
2) asustainedplinianphase
3) theemplacementofpyroclasticflows
4) aphreatomagmatic waning
The reconstructed scenario leads to thedefinition of areas subjected. to different kinds of
phenomenaand,consequently,todifferentkindsofrisk,Figure3.Pompeiilies ~ thearea
~ , P ~
191
190
Beyond the Map
Figure 2.The AD 1631eruption deposits. The dotted areacorresponds tothe
isopachs ofthe fallout, theother areas correspondtothepyroclastic flows.
Figure3.The reconstructederuptive scenario showing the threezones potentially
affected by different kinds of phenomena: I) horizontal hatching =fallout; 2)
vertical hatching=proximal phenomena including pyroclastic flews: and3)cross
hatching = debris flows. The pyroclastic fallout area was estimated as having a
probability greater than 10%ofafalloutof morethan400 Kglm'.
F. Giannini et al. I Ancienl and new Pompeii
potentially affected by pyroclastic flows and with a 10% probability that more than 400
Kglm
2
ofpyroclasticfallout wouldcoverit[6].[7].
5.The ancientPompeii GIS
The managementof different typesof data and the large amounts of information involved
require asuitablesetof tools. Geographical Information Systems provide a whole new set
of techniques and represent a great improvement byenabling the rapid storage, retrieval,
manipulationandanalysisofdata inaverypractical way.Moreover. theprecise location of
every artefact, architectural unit, site and archaeological feature is guaranteed and the
superimposition of the georeferenced layers allows analysis by means of different
categories ofitem. TheGISprovidesthefollowing basicoperations:
I) theaccurate maintenance ofdatadimensions andlocations without scalerestrictions; it
is, therefore, possible to magnify anyview to provide more detail as one zooms inon
objects
2) thecreationofoverlays bycombiningalllayersofinterest inasinglethematic map
3) thestorage ofalphanumeric andspatialinformation withsimultaneous retrieval
4) topologicalquerying
Inparticular, theGISofPompeii hasbeendesigned:
l) todeal with features recorded atdifferent degreesof precision andto integratethe data
(thearchaeologicalinformation hasdetailstoafewcentimetres inspatialprecision, and
actualbuildingsandtransport networks toafewdecimetres);
2) to store and manage spatio-temporal stratified information by the use of layers and
relational databases.
6.TheGISofPompeii municipality
The GIS of thePompeii municipal district is apartof the larger GISof the Vesuvian area,
basedonthe integration ofdifferentdatasourcescollected atdifferentscales.Theseare:
I) the photogramrnetric plan of the Vesuvian territory based ona 9 ~ flightat arelative
altitude ofapproximately 2,000 mand ascale of 1:5,000.The scaleof the photograms
isapproximately I:12-13,000[8],[9]
2) thephotogrammetricplanof thePompeii municipality basedonaflightinMarch1997
at a relative altitude of approximately 1,200 m. The scale of thephotograms is
approximately 1:8,000 and the nominal scale of the photogramrnetric plan derived is
1:2,000
3) the IGM (MilitaryGeographical Institute) maps of the surrounding areas at a scaleof
1:25,000
4) information about the population with more than 300 attributes, including information
aboutsocial/publicandeconomicactivities;sourceisthe 1991ISTATCensus
5) the AlMA orthophotos from a flight in May 1997. These digital images have a
resolution of 1metreon thegroundand anominalscaleof I:10,000,aflightaltitudeof
6,000 m,and anaverage scaleforthephotograms of1:40,000. .
193
192
Beyond the Map
All the information collected from these different sources is organised in georeferenced
layers, both vector and raster, and in linked database tables.
6. I Raster data: the Digital Elevation Model
Digital Elevation Models (OEMs) are continuous raster layers in which cell values
represent elevation and can be constructed from a Triangulated Irregular Network (TIN)
where the vertices of triangles are points of a known elevation, for example the points
derived from the photogrammetric plans.
The OEM for Vesuvius covers an area of 48 km x 50 km and is based on elevation
points from photogrammetry and the IGM 1:25,000 contour lines. From the OEM
geomorphological features such as the watershed, drainage network and slope matrix have
been calculated. Using the elevation matrix a 30 representation was made giving a realistic
view of the topography allowing us to see structures which are not visible in traditional
two-dimensional plans, Figure 4. A more detailed OEM has been computed for the
Pompeii municipality and the ancient city.
6.2 Vector Data
The vector layers for the GIS of the Vesuvian area are: contour lines, spot heights,
hydrology, buildings, roads (including footpaths, roads and motorways), railways, place
names and administrative boundaries.
At present a new GIS is in preparation, at a scale of I:2,000, for the municipality of
Pompeii, to include more than 400 layers. It is planned that all the municipalities within the
area surrounding Vesuvius will be covered by this level of detail, as shown in Figure 5.
As with GIS generally, the number of features and layers displayed depends on the
scale, a smaller scale shows more detail and a larger number of features. The scale also
influences the way in which a single element is drawn, for example, either as a point at a
Figure 4 . A 3D view of Vesuvius, offering a very realistic image
F. Giannini et aI. / Ancienl and new Pompeii
Figure 5. The GIS of Pompeii at a scale of 1:2;000. The shaded area (centre left) is the excavated-area
large scale or as a polygon at a smaller scale. This is an important consideration here
because of the detailed classification used. which is organised in hierarchical classes.
Therefore, when necessary (for example, when zooming to a smaller scale), grouping the
features may reduce the number of layers and create a more understandable view.
7. The GIS and archaeological features
The various archaeological features scattered throughout the territory were captured by the
digitisation of maps (at the scales of 1:100 and 1:200) belonging' to the archaeological.
Superintendency and georeferenced at ground level by GPS measurements. For the ancient
city of Pompeii the photogrammetric plan was constructed from 13 stereoscopic models (at
a scale of 1:3,(00) produced from a low level flight of 450 m altitude using a wiLD Re8
with a principal focal length of 151.77 mm.
The survey was georeferenced by using the grid vertices of the Italian National
Geodetic System (lGM) photogrammetric points, and those of the previous survey done
during the Pompeii excavations (the Neapolis Survey at a scale of 1:500). To produce the
two-dimensional plan a rototational translation was carried out using the IGM
transformation of trigonometric points belonging. to the new IGM 95 network, which is .
194
Beyond the Map
high precision for the reference co-ordinates used (trigonometric points 184802 - 185903 -
185904). To calculate altitudinal points the interpolating plane method was used for the
transformation of geodetic heights into ellipsoidal heights. In the GPS network some points
where the actual height was known were measured (points belonging to the IGM geometric
levelling network line 83 Cs n" 48- 49-51). Overall this produced 3,142 land spot heights
(2.2 - 51.3 metres above sea level) and 6,822 spot heights referring to man-made edifices.
The features of the photogrammetric plan are organised into many layers and subdivided
into sub-layers, Figure 6.
The whole archaeological area is subdivided into a hierarchical scheme of Regions,
Insulae, Buildings and Rooms as shown in Figure 7 and Table 2. For the identification of
features inside the excavated buildings we used the RICA maps which refer to Pompeii
Insulae at a scale of I:500, derived from an enlargement of an aerial survey at a scale of
I: 1,000. Unfortunately, this publication only contains the Insulae of the first five Regions,
and for the rest, sixth to ninth, we used the maps of the Neapolis publications. Compared
with the RICA maps, these have fewer details and have not been verified by survey. For
tabular data we referred to the Neapolis Publication, Indirizzario, which is a table that lists
each building and its house number.
This classification was also used for the different layers of the GIS. In order to do
this, a vector polygonal feature representing the entire excavated area was positioned
beneath the photogrammetric plan and then divided into Regions and Insulae according to
the centre lines of the roads and the perimeter city wall. The boundaries of the buildings
were drawn using the external perimeter walls and the centre lines of the internal walls,
[lOl, [11].
195
F. Giannini et al. /Ancienl and new Pompeii
Figure 6. The dark area represents !he phologranunelric plan of ancienl Pompeii, Figure 7, (a) The nine Regions of Pompeii, (b) !he Insulae subdivisions, and(c) the buildings.
surrounded by features in !he GIS of new Pompeii The code for each building identifies the Insula and Ihe Region in whiclrit is located. Building
boundaries are drawn Ihrough the external walls andthe centre line of i.ntemal walls.
_._"'-----_.. ...-r-:..---:----_.---;--.
197
196
Beyond the Map
Table2.The compositionofidentificationcodes
No. of Regions
Regions X No. of Insulae
Insulae X X No. of Buildinesl
Buildings X X X INo. of ROOIDS I
Rooms X X X I X I
Roads X X
Sidewalks X X
For the definitionof the Regions,Insulaeand Buildings,the detailsofthe photogrammetric
plan were sufficient although to draw the room areas and shapes, Figure 8, we needed to
georeferencethe RICAmapsfor the first five Regions,and the Neapolis maps for theother
four. This was sometimes impossible, however, because sufficient control points in the
photograrnmetric plan to locate the room could not be identified, for example, when the
view is obstructed by tree foliage or a roof. Whenever possible though, the
photograrnmetricplan was used ratherthan the RICA and Neapolismaps becausethe scale
and positioningis more precise.
The databasetablesconnectedto the graphicalfiles containacode, i.e. akey,which
uniquely identifies each element. The code is hierarchical so that a smaller element is
identifiedbythe higher-ordercodeelements. For example,aroom isidentifiedby the code
of the Region, Insula and Building to which the room belongs, plus an element for the
room itself. The buildings are identified by a main civic number with a secondary civic
numbersometimesbeingused in theroomclassification.
The roads and sidewalksare polygonalfeatures with the formeridentifiedby acode
that is linked to theirplace-nametable. Both are also identifiedby the codeofthe Regions
and theInsulaeinwhich they are located.
Regie 5
Insula 1
N
A
o 10 10 MuU$
Figure 8.Room subdivisionsinthebuildings ofInsula I, Region 5
F. Giannini et al. / Ancient andnew Pompeii
The perimeter wall of the town is in polygonal format and classified into three
categories: (a) features that are present, i.e, are defined on the photogrammetric plan, (b)
hypothesizedfeatures, i.e, not on the plan, and (c) towers.
Each graphicalfeature has a code that links it to a text table, for the buildings this
uses the Neapoliscodification that includes the use of the building, the specific name, the
main and the secondary civic numbers. For the rooms, the ICCD classification is used
althoughthis only applies tothose alreadyphotographicallyarchived..
8. Conclusion
This workshowsthat GIS isapowerfultool thatcan be used inmany differentways:
I) for the compilation of suitable maps, in which, depending on theirscaleand purpose,
differentlayerscan be included,and featureswithinthe layers can beswitchedon or off
2) for combiningdifferentcategoriesofdata collectedasdifferentlayers, and carryingout
analysesacrossthedifferentlayers .
3) for the construction and management ofan archive of archaeological features which
can beeasilyupdatedwith new information .
4) to aid in the work ofarchaeologicalrisk assessment
Acknowledgments
The work was carried out with the financial support of the Italian Civil Protection
Department (law n.74, 26-2-1996, Art. 15 quater, to the Osservatorio Vesuviano), the
Italian National Groupfor Volcanology-CNR and the Progetto Finalizzato Beni Culturali
ofCNR. We also thank two unknownreferees for theircomments.
References
[I] H. Sigurdsson, S. Carey, W. Cornell and T. Pescatore, The eruption of Vesuvius in AD 79. National
Geographic Res. I, 1985,pp. 332-387.
[2] R. Cioni, P. Marianelli and A. Sbrana, Dynamics of the A.D. 79 eruption: stratigraphic, sedimentologic
and geochemical data on the successions of the southeniand eastern sectors, Acta
Yulcanologica.Z, 1992,pp. 109-124.
[3] R. Cioni, L Civetta, P. Marianelli, N.Metrich, R.Santacroceand A. Sbrana,Compositionallayeringand
syneruptive mixing of a periodically refilled shallow magma chamber: the AD 79 plinian eruption of
Vesuvius, Journal ofPetralogy, 36, 3, 1995,pp.739-776.
[4] H. Sigurdsson, S. Cashdollarand R.S J. Sparks, The eruptionof Vesuviusin A. D. 7'): Reconstruction
from historicaland volcanologicalevidence,American Journal ofArchaeology, 86, i 982, pp. 39-51.
[5] R.Santacroce, Preparing Naples for Vesuvius, WodshopHandbook: Vesuvius decade Volcano,IAVCEI
Napoli, 17-22September, 1996,pp. 34-39.
[6] F.Barberi,G. Macedonio,M.T. Pareschi andR.Santacroce,Mappingthe tephra fallout risk:an example
from Vesuvius(Italy), Nature, 334, 1990,pp.142-144.
[7] R. Santacroce,D. Andronico, L. Cav8rT8,R.Cioni, M. Favalli, A. GiOncada, A. Longo, G. Macedonio,
M.T Pareschi., M. Ranci, M. Rosi, A. Sbrana, R. Sulplizioand G. Zanchetta, Aggiomamento della carta
di zonazione della pericolositl dell'eruzione massima attesa aIVesuvio, Ani del Congresso Scieruifico
AMuole del Gruppo Nazionale perla Vulcanologia, Catania 24Marzo, 1998, p.143..
198
199
the Map
[8J M.T. Pareschi,R. Santacroce, M.Bisson, L. Cavarra, M. Favalli, F. Giannini and A. Un GIS
dell'area vesuviana per /a mitigazione del rischio vulcanico. GNY - Gruppo Nazionale per la
Vulcanologia,FeliciArtigrafiche,Pisa,inpress.
[9] M.T. Pareschi, L. Cavarra, M. Favalli and F. Giannini, GIS and Volcanic Risk: Management, Natural
Hazard, inpress.
[10]F.Giannini,M:T.PareschiandG.Stefani, AGISforthearchaeologicalsitesindensely populated areas.
The case of the ancient and new Pompeii. Proceedings of the 2'"'International Congress, Science and
Technology for the safeguard of cultural heritage in the mediterranean basin, 5-9 July Paris, France
1999,inpress.
[II]M.T. Pareschi,G. Stefani,A.Varone,L. CavarraandF.Giannini, AGIS forthe archaeological areaof
Pompeii,VolcanoesEarthqualces& Archaeology,Journal ofthe Geological Society of LoruJon, inpress.
Beyondthe Map
G.Lock:(Ed.)
IDSPress,2000
Archaeologyandvirtual micro-topography:
thecreationofDEMsfor reconstructing
fossil landscapesbyRemoteSensing
and GISapplications
MaurizioForte
Institutefor Technologies Applied to Cultural Heritage
Rome
Abstract: Micro-topography' is the ability to produce detailed cartographic
information in micro-scale using contour lines and/or elevation points with small
intervals. This often involves creating DEMs by laser Total Station; centimetre-
accurate GPS or usingrnicro-cartographicdata usually collected by geologistsand
archaeologists. Inprocessing these micro-data it is possible toanalyse, to visualise
and to interpret archaeological features of the landscape not otherwise visible. In
fact, this is the only way to reconstruct invisible cognitive archaeological
landscapes. Also, bycomparing the verticalexaggeration of each micro-OEMwith
raster images (aerial photos) and vector data (cartography, interpretation,
geomorphology, etc.), it is possible to obtain further information about the
diachronicevolutionofthesite,itsgeomorphologyand palaeo-environment.Finally,
itisalsopossibleto predictthepositionoffeatures andancientanthropicstructures.
The fossillandscapeofthecentralPo Plainduring the MiddleandLateBronzeAge
ispresentedasacase-study
1.Introduction
This paper attempts to address importaatquestions regardinglandscape archaeology and
theuseof archaeologicalcartographyintegrated with GIS andremote sensing applications.
What scale ismost useful forinterpretation?How and whatcan weperceive ofthe original.
archaeological landscape? What are therelationships betweenaerial-photo interpretations,
digitalclassifiedimages, and the terrain micro-relief? How can the thirddimensionhelpin
reconstructingand interpretingarchaeologicallandscapes?
Usingamicro-topographicalapproachwecandistinguishfourtypesof investigation:
1) thereal landscape oftoday - howdoweperceiveitandexplore it?
2) the40virtual landscape, where the fourth dimensionischangethroughtime - how has
it been modified, what is the visibility on the ground year by year (Figures 4,5, 6 and
7)?
3) the 30 enhanced pseudo-virtual landscape, specifically the OEM with vertical
exaggeration- can weestablishcognitive perceptionthrough virtual navigation (Figure
8)?
200
Beyond the Map
4) the 3D virtual paleo-environmental reconstructed landscape - how much can we
reconstruct, how can wepreservearchaeologicalinformation?
Due to the complexity of these questions it is important to emphasise that without an
approach based on micro-topography we are unable to interpret an archaeological
landscape indetail and changescale from inter-site to intra-site. As shown specifical1y in
this case-studyofthe Po plain, this approach has been fundamental for the perception of
thelandscape.
Micro-topography in archaeology is the possibility ofcreating and representing
cartographic and landscapedata on a micro-scale. This isdone by creatinggee-referenced
virtual models based on DEMs (Digital Elevation Models) using data from laser Total
Stations, from GPS or from detailed cartography, i.e. contour lines from altitude points.
We can obtain new analytical perspectivesofthe archaeological landscape by combining
different types of DEM interpolationand bychangingmodel details and resolution.In fact,
itisoften the case that traditionalcartographicinformation (i.e. at the scale of1:10,000or
1:50,(00)is useless for meaningful and useful representation of archaeological data. The
first experiments in micro-topographywere undertakentowards the end of the 19808,[I),
(2), and the results were very promising; today, with the possibilityof visualising 3D data
inrealtime, thereisenormouspotential forthestudy ofvirtual landscapes.
Here we define representative cartography as general1y available paperor digital
2D mapping, and cognitive cartographyas that specifical1yproducedfor the digital virtual
visualisation and interpretation of archaeological sites and landscapes. Cognitive
cartography increases the functionality of 3D GIS visualisation and processing, in
particularfor topographic reconstruction of micro-reliefusing the detail shown on aerial
photos (taken atdifferenttimes), including textures, geomorphological and archaeological
features.
Moreover, these results are often extremely helpful in planing excavation
campaigns or for monitoring visibility and conservation ofarchaeological features. This
approach also increases knowledge of sites that haven't been investigated by generating
detailed 3Dmodels and describingtheir morphologydigitally.
2.ArchaeologicalDEMs
Digital ElevationModelsaredigital representationsofelevation organisedasaregular grid
of numbers so that the spacing between grid elements represents the interval between
samples. For example, a 30-metre grid spacing means one elevation sample every 30
metres. The numerical value ineach gridelementrepresentsthe elevationat that point. In
micro-topographya DEM isthefundamental basis for studyingthe morphologicalfeatures
of archaeological sites. The results of different interpolation methods between data points
produces different terrain visualisations,here wehave used TIN andGrid:
1) a TIN, Triangulated Irregular Network, represents swfaces using contiguous, non-
overlapping triangles. A surface value anywhere in the triangulation can be estimated
by averaging node values ofnearby triangles, giving more weight and influence to
thosethatarecloser
2) incontrast, Grid interpolationisa representationofasurface using amesh of regularly
spacedpoints.
20;
M. Forie /Archaeology virtual micro-topography
Othermethodsofinterpolationare:
lOW, Inverse Distance Weighted interpolator, assuming that each input point has a
local influencethatdiminisheswithdistance
Spline, a general purposeinterpolation method that fits a minimum-curvature surface
throughthe input points .
Kriging, where the distance or directions between sample points reflects spatial
correlationthatcan beused toexplain variation inthesurface.
3.TheDEM:creation,processingandvisualisation
Thefol1owingbasicdigital sources areavailable forconstructingandprocessingaDEM:
ASCndata from laserTotal Stations
ASCndata from GPS instruments (thisproject usescentimetre-accurateGPS)
Rasterdata(elevationgrid)
Vector data(contourlines from whichtointerpolateinraster format).
In archaeologyitisvery importanttochoose the best method forprocessingthe DEM after
itsinterpolationasgood visualisationcorrespondswiththequalityof the DEM.The image
processingtechniquesuseful formicro-scaleprocessingare:
Edge enhancement, contrast enhancement, edge detection and other segmentation
filters
Sun angle filters, that is, edge filters that highlightstructures in a specific direction to
enable the detailed detection of geomorphological and archaeological features in real
time basedon thedirectionofsunlight .
Averaging filters that blur the data by taking an average ofneighbouring cel1s,useful
for smoothing noisy DEMs prior to sun-shading. Average filters also show variation .
betweenhighfrequencydata intheDEM
Ranking filters that rank data in various ways, these may be used to clean up noisy
DEMs, forexampleifsome values aretaken from treecrown height and somefrom the
ground. . .
Otherraster DEM processingapplications are:
Generatingslope from aDEM, a measure of the how steep any given point is.Slopeis
computed by comparing a cel1value with values of neighbouring cel1s,the result is
then normalisedsothat slopeismeasured regardless ofdirection
Generating aspect from aDEM,a measure of the directiona slopeis facing. Aspectis
measured from 0to360degrees, with0being Nonh,90 being East,etc.
Finally, imageprocessingtechniquesuseful forthe analysis ofarchaeologicalDEMsare:
Pseudo colourprocessing. A false colourimage is similarto a greyscaleimage except
thatapaletteofcolours isallocated tothedifferent DEM values
Real-timesun-shadingisapowerfultechnique forhighlightingstructure withinaDEM
by simulating how asurface would look if the sun were in different positions (defined
asazimuth rotation fromNorth, andelevation above thehorizon)
Combinedreal-timesunand falsecolourdraped images
202
Beyond the Map
Colour draping whereby afilter isdraped overtheDEM asanIntensity Layer, usually
viewedin3D,toseethecorrelation betweentheoverlaindataandtheDEMheights
RGB-Height display, a combination often used when a colour image is to be shown
over a DEM. Red, Green and Blue layers(from thecolour image), and a Height layer
showntogether
Classification-Heightdisplay, inthis combination imagesthathave been classified are
displayed overaDEMasanIntensityorHeightlayer.
The results of such DEM processing applied to the terramare case-study are very
informative, Figures 3 to 11,mainly through comparing raster and vector overlays in 3D
including aerial photos, geomorphological evidence, diachronic change in the visibility of
sites, and filtered digital images. In this multi-layered and integrated analysis, the overall
levelofinformation isincreasedincrementally bycheckingthedigitalresults withthoseof
fieldsurvey. In allthesites investigated (Case Cocconi, S. Rosa,Monticelli, and Case del
Lago) this approach has given the scientific community new and unexpected knowledge
aboutthestateofpreservation, theshapeandconfigurationofthesettlements.
FortheDEMvisualisation andinterpretation theprincipalstageshavebeen:
The digital monitoring of the territory through panchromatic analysis of aerial photos
overtime,Figures 3to7
Theintegration andoverlay ontheDEMof allvectorandrasterdata,Figures 2,5,6,7
and8
The texturemappingofaerialphoto-mosaics ontotheDEM,Figures5,7,and 10
The multi-dimensional and multi-layered analysis of the 3D models, at macro and
micro-scales, Figures 3toII
The vertical exaggeration of the micro-DEM based on the elevation range (it is
important to select thedegree of exaggeration with thepreservation of archaeological
features inmind).
Table I summarises the main results of the micro-topographical analyses. Note that with
suchalowelevation range, the3Dre-constructions of thesites increase their visualisation
considerably.
4.Seeing theinvisiblelandscapethroughvirtualinteractions
Of the archaeological sites subjected to these GlS- and remote sensed-based methods,
CaseCocconi hasbeen chosenforcomparing groundlevelvisibilitywiththatproduced by
the 3D draping of aerial photos onto a DEM, Figures 3 to 7. The archaeological
interpretation of thissite usingtraditional methods is,infact,verydifficult due to thelack
ofsystematicsurveyandexcavation.
Theexperiment hasinvolvedthefollowingsteps:
a QTVR (Quick Time Virtual Reality) reconstruction of site panoramas starting from
different observation points using Photovista software and 2D interactive technology.
Thisenables virtualexplorationofthesiteusing360
0
stitchedphotographic panoramas
3D modelling of the sitebased on horizon perception with thenaked eye, recognising
shapesin3D.Acomputer graphicanimationshowingthemorphologicalfeaturesofthe
sitepriortothemicro-reliefestablishedbylaserTotalStationsurvey
203 .
M. Forie /ArcNuiology virtual micro-topography
Table1.Micro-topographicalanalysisof the sites inthe-terramarecase-study
Mootkdli CaseCocconi Case del
LIll!O
S. Rosa
Surface area 7 hectares 35 hectares 7hectares 10bectares
Dala points 837 2408 900 600
Elevationranee -0.613to 1.68 -0.925 to 1.33 -1.20to 1.20 oto0.93
Embankment
visibilitv
Yes Yes Yes Yes
Grid--
DEM
interoolation
Kriging. TIN Grid
Structuredsites Yes Yes Yes Yes
Aerial visibility 1955, 1981,
1990
1955,1977,
1979,1981,
1988,1991
1955, 1977,
1981,1988
1955 to 1998
Quadrangulat
shape
Yes Yes Yes Yes
Quarry works
visible
No No Yes Yes
Aerial
photographic
rectification
Yes Yel!. Yes Yes
a3DDEMreconstructionandvisualisationbased onlaserTotal Station surveyand 3D
interpolation. A fly-through navigation of the site with vertical scale exaggeration to
help identify the original anthropic deposits, using the 3D real-time interface ofER
MapperandtheVRMLbrowserCosmoPlayer. .
Finally wehave integrated into a single html document a virtual photographic panorama,
ananimation filereconstructing howthesite isseenfrom ground level, anda VRMLfile.
The result is very interesting and surprising, because by offering multiple views it is
possible to identifyfeatures not visiblein the real world.This approach, therefore,shows
thepotentialofdevelopingGISonamultimediaplatform enabling newinterpretations that
aretheproduct ofthecombinationofmultimediatools.
5.TheTerramareProject
"An inventory of the Terramare in the central Po Plain: physiographic context,
stratigraphic andstructuralcharacteristics,stateofpreservation", isthetitleoftheCultural
HeritageFinalised Project,sponsoredbytheItalianNational Research Council(CNR) and
directed by Professor Mauro Cremaschi; University. of Milan.. Its main objective is the
production of digital thematic cartography of the central Po Plain during the Middle and
Late Bronze Age, integrating different 2D and 3D geo-referenced data including sites,
excavations, surveys,landscapeandterritorialdata.
Terramare arearchaeological sitesdating to theMiddle andLate BronzeAge (16
th
to 11
th
centuries BC) [3], [4],[5] whosepreserved characteristics werealtered in the 19
lh
centurybybeingusedasfertiliserquarriesbecauseoftheirrichorganicdeposits.The main
structures of the terramareconsist of villages, roughly quadrangular inshape, surrounded
by aditch and earthramparts. The enclosed space suggests urban planning withhouses,
roads,canals, andcowshedsalonggriddedorientations: .
204 205
Beyond the Map M. Forte/Archaeology virtual micro-topography
An analysis of site distribution, Figure I, with one site per 15-25 km
2
, suggests a
hierarchical relationship between the structured urban sites, described above, and less
structured rural sites which were small settlements. The subsistence strategy of the
terramare consisted of intensive agriculture, deforestation and pastoralism, organised over
a very wide territory representing a good case of a ranked landscape of power [6].
Through detailed study of the archaeological deposits it was surprising to discover
that the landscape is still partially and meaningfully preserved, although invisible from the
surface. By 20 and 30 digital processing it is possible to recover much more
archaeological information, making virtual cognitive models, reconstructing memories of
past landscapes.
Geo-morphological evidence shows that the shift of the main river course and
alluvial deposition since the Roman period has buried part of the Bronze Age landscape of
the Po Plain. Consequently, we have created a OEM with 1m contour lines, integrated it
with other raster and vector data (aerial photos, digital cartography, etc.), to verify the
micro-evolution of the landscape and its ancient settlements. The main palaeo<hannel of
the Enza River, Figures 2 and 9, is cut in a Late Pleistocene alluvial fan, the northern fringe
of which was delimited by a palaeo-channel of the Po river. Both were active in the Middle
and Late Bronze Age and in this area some of the largest sites of the whole terramare
system (Case Cocconi, Case del Lago, Monticelli, S. Rosa, Motta Balestri) are found
connecting the Apennine fringe to the River Po. The orientation of these sites, along the
Enza palaeo-channel to the Po palaeo-channel, is towards the north-east and a network of
minor meandering water-courses fed their ditches.
The main trend of field orientation in the area is north-east to south-west and reflects
the Roman centuriation [6], the terramare structures are also aligned north-east the same as
the minor water course pattern. Centuriation has sometimes covered and surrounded
former agricultural field systems although they can still be identified as at Case Cocconi
where a large rectangular area delimited by ditches and about 60 hectares wide surrounds
Figure I. Bronze Age sites in the Po valley, northern Italy
. .
the site which is seven hectares wide, Figure II. A similar pattern was also recovered
around the Terramara Castello del Tartaro, north of the Po river [4]. These patterns occur
mainly around the largest terramare and can be interpreted as relict features of a structured
agrarian landscape. Their relationship with the highest ranked sites in the distribution
indicates that the Bronze Age landscape in Northern Italy was mainly cultural and the
product of power strategies, a landscape of power [4].
The integration of traditional research, geo-morphology, palaeo-pedology, surveys
and excavation, with Remote Sensing and GIS techniques has been made it possible to
reconstruct the diachronic landscape"surrounding the terramare, the structure of the sites
themselves and their relationships with surrounding territories.
The largest sites of the whole terramare system, from 7 to 14 hectares, are found in
the northern valley of the River Enza, Reggio Emilia-Parma, Figure 2. Using Remote
Sensing and GIS applications a range of data have been processed in 20 and 30 including
geo-morphological, pedological (Holocene alluvium) and archaeological, to produce an
integrated 30 reconstruction of the ancient landscape.
6. Cartographic information and the project stages
To enable the production of cartography and micro-models of the central Po plain
landscape, two different levels of representation have been used:
micro-scale (level of analysis = site, intra-site); OEM produced from cartography at
1:10,000, contour lines digitised at I m interval
macro-scale (level of analysis = territory, inter-site); OEM created on the ground by laser
Total Station survey, grid size of I m, also centimetre accurate GPS, contour lines
interpolated at 20 cm intervals.
Geomorphology and BronzeAge sites
in Ihe northernElIZa basin, Emilia
Romagna, Italy:
I) 0 S. lIario AllUvial fan (Lale
Pleistocene);
2) paloochlllll1el;
3) 0 Praticello paleochamtel deeply
cut down, adive during Ihe Bronze
Age;
4). fluvial ridge;
5) small meandering waler courses;
6) _low lying flood basins;
7) WIStnIctured Bronze Age sites (a:
on swface; b: buried); .
8) 0 lemlmare (s1ructured sites);
.1 9>++ mainaxes of the Roman
cenluria1ion;
10) cootour lines, eq. I m..
Figure 2. Geol1lOlJlhology and Age siies in the Enzabasin, northernItaly .
.
"',1
206
Beyond the Map
The stages of digital processing have been:
histogram matching across aerial photo-mosaics to eliminate chromatic discontinuities
and produce an homogeneous raster base
DEM creation from geo-morphological and cartographic data
DEM creation from laser Total Station survey data using a grid of about 1000-7000
points for each site
geo-referencing and rectification of raster data and aerial photos, Figure 3
digital analysis of change through time in the territory using panchromatic analysis of
aerial photos, Figures 4, 5 and 6
Figure 3. The integration of cartographic and aerial photo
interpretation data for the site of Case Cocconi
Figure 4. Tracking landscape change through aerial photographic evidence for 1977 (bottom),
1982 (top right), and 1991 (top left) at Case Cocconi
207
M. Forie / Archaeology virtual micro-topography

Not corresponding
parts between aerial
photo-interpretation
(1979)
and micro-DEM
(1998)
Figure 5. Comparing archaeological features at Case Cocconi between aerial photographic
interpretation and the micm-DEM, the {hree areas of light shading are areas of difference:
Figure 6. An example of aerial photographic. diachronic overlay at Case Cocconi
1:10.000 map, OEM, '55 aerial pboto
3D OEM and aerial photos
Figure 7. The integration of I: 10,000 mapdata, the OEM and
aerial photos at Case Cecconi
, 1
209
208
Beyond the Map
Monticelli:DEM
Figure8.AverticallyexaggeratedOEMforthesiteofMonticelli
Figure 9.Theintegrationof aOEMand aerial photographicdata
integration and overlayofallvector andrasterdataonthe DEM, Figure 7
texturemappingofthe photo-mosaicsontheDEM
multi-dimensional and multi-layered analysis of the 3D models at macro and micro-
scales
interactive 3Dnavigationthoughthemodelofthearchaeologicallandscape
interactive 3D visualisation ofexcavated stratigraphy, including raster data (vertical
photos ofthe excavated areas) and vector data (digitised finds, stratigraphical units,
etc.)
M. Forte / Archaeology virtual micro-topography
creation of a GIS specifically for the archaeological excavations with digital
acquisitionofdataonsite
developmentof3DGIS models using Java applets and VRML for on-lineapplications,
Figure12
virtual communicationand integration of the whole projectfrom the GIS to the virtual
theatre (see below). .
Figure 11. Surface featuresatCaseCocconiappearontheOEM
Figure 10. Applyingan aerial photographas texture toaOEM
211 210
Beyond the Map
7.Theuseof GPS
Theuse of GPS with centimetreprecision has great potentialfor archaeologicalfieldwork
for the following reasons: firstly the speedof data acquisition with only one operator(one
point per second), secondly topographic survey data automatically has U1M co-ordinate
registration, and thirdly because of the integration of 20 spatial co-ordinates with 3D
altitude data. This project is experimenting with the possibility of creating micro-OEMs
directly from GPS data instead of using the laser Total Station, due to co-operation with
theCNR-ITABCof Rome(InstituteofTechnologiesAppliedtoCulturalHeritage).
A LeikaGPS receiverprocesses the signals received by the antenna and data are
stored for post-processing on a personal computer. Baselines can be calculated with a
precisionof about5-1Omm+2ppm, and, with a radio modemattached, the receivercan be
used for real-time measurements accepting RTCM code corrections. Co-ordinates can be
calculated with aprecisionof about O.5m.Parametersin the receivercan be set to steerthe
GPS measurementoperation. Data input is via a fully alphanumeric QWERTY keyboard
and an LCDdisplayof 32 x 12characters which may be illuminated. Baselineprecisionis
dependent upon various factors including the number of satellites tracked, constellation
geometry, observation time, ephemeris accuracy, ionospheric disturbance, multi-path and
resolved ambiguities. The first experiments show that in a day's work it is possible to
collectmorethan 10,000points.
8. VRGISandcognitiveinteractions
VRGlS is a class of software applications that combines Virtual Reality (VR) and
GeographicallnfonnationSystem (GIS) [8]. By integrating a real-time three-dimensional
visual simulationwith aGIS it is possibleto explore3D landscapesby flying and walking
throughthem [10], [11].
The value of a VR model is in its cognitive interaction, a process that is also
important in the reconstruction of archaeological sites and landscapes. "Theenvironment
acts on the system and the system in tum reacts with an output. The behaviour of the
system therefore can be considered as the interaction between input and output. The
behaviourdepends in the first place on the structureof the system, by which is meant the
organisationoflinks betweenitselements"[6].
The growth of scientific and didactic results obtainable by using VRGIS can be
summarisedasfollows:
3Dreal-timeinteractionand visualisation
cognitivearchaeologicallandscapes,or mindscapes[12]
visual interactiveinformationand dynamicinterpretation
multi-dimensionaland multi-layeredanalysis
diachronicanalysisand monitoring
virtual communication
on-lineeditingand publication
multimediaintegrationofdifferenttechnologies
3Dvirtual reality,virtual navigation
VRMLmodel editing,OEMsand models
VRMLinteractivesurfaces,new tools forinteractiveOEM reconstructionand analysis.
M. Forte /Archaeology virtual micro-topography
Human perception of objects and/or three-dimensional virtual ecosystems results in
different types ofinteraction with virtual space. Perceived phenomena providea level of
interaction,or moreappropriatelyanexchange,evenif this is anabsenceof behaviour. We
can identifytwo main levels ofinteractioninreal-time:
a passive cognitive interaction where navigation/movement in 3D space takes place
withoutbehaviourexchangesbetweenobjectsorevents
an active cognitive interaction where navigation in 3D space takes place with.
behaviouralexchangesbetweenthe environmentand events. This type of interactionis
only possible in the case of three-dimensional virtual models and represents the most
advancedlevelof perceptive-cognitivegain, still rare inarchaeologicalapplications.
9.A 3D Java-VRMLapplication
For this specific project, and with the co-operationof the Universityof Bologna, we have
created a Java-VRMLinterface in orderto visualise, modify and enhancethe on-line 3D
GIS archaeological models available on the Internet, Figure 12. Using this dedicated
interface it is possible to load.the VRMLGIS models (created and exported by Arcview
3.1 with 3D Analyst), and process textures (raster images) and OEMs. The interface
includesa3D menufor navigationand a20windowfor loadingand modifyingthedata.
The main interfacemenu includes:
a window for loading the VRMLsurface model (OEMs constructed from laserTotal
Stationsurvey)and the texture(aerial photos)
a directions tool for choosingthe positionof the pointeron the grid surface(a micro-
ball)
a Ztool for choosingthe verticalexaggerationfactorin 3D
agrid tool for incrementingthe gridsize
a Views menu for savingdifferentperspectiveviews of the samemodel inVRML
aninterpolation commandfor choosingthe type of interpolation,cubic, bilinear, etc.
Figure 12.TheJava-VRMLinterface
213 212
Beyond the Map
Data-entry involves a VRML model imported as a surface (OEM) and an image imported
for texture (aerial photo). When a VRML model is loaded (as a OEM with aerial
photographicoverlay) itispossible tomanipulatethefeaturesofthemodelin3D,and
compare in real-time the results of different interpolations and modifications. Moving the
pointeronthesurface,everygridpointcanbeselectedandmodifiedonitsverticalaxis,a
different gridsizeandinterpolation methodcanalsobechosen.Thedensityofpointscan
also be increased to produce a more detailed model and the results visualised in 3D real-
time.Forinstance,thegridsurface for reconstructing theoriginal archaeological landscape
can bechanged resulting in moving rivers, artificial hills, ramparts and settlement and to
aidintheanalysisofthestateofpreservationofeachsite.
10. The virtual theatre: towards an immersive virtual environment for landscape
archaeology
Otherpossibledevelopments oftheprojectconcern avirtual museumofthe archaeological
landscape, that is, the possibility to visualise these models with 3D interaction by an
audience of up to 50people. The term 'virtual theatre' isused to denote a physical space
whereagroupofpeoplecannavigate inavirtualrealityenvironment, ifnecessary theycan
beprovided withpolarisedeyeglasses forastereophonicview.
The virtual theatre can be designed to accommodate different types of user as an
archaeologist will need a more detailed 3-D visualisation than the average virtual touris!.
Integrated communication will provide two different ways of accessing the Virtual
Environment, byInternet andinanimmersiveenvironment, although thedata visualisedin
both models will be equally validated by archaeological research. The first Italian
experiments of thiskind arebeingcarried out atCINECA (Inter-universityConsortium of
Supercomputing, Bologna),wherethefirst virtual theatreislocated.The digital equipment
includes: 8 R 10,000 processors, 4 Gigabyte of Ram, three Infinite Reality2 graphic
pipelines with 10rastermanagers, andthreeBarcovideo-projectors [13].
In order to allow more complex behaviour dependent upon external data or user
interaction, itispossible touseanExternal AuthoringInterface (EAI) toexternally control
theVRML world through aJava applet,The choice of VRML andJava guarantees cross-
platformaccess.Fromthetechnological aspectthetwomostimportantcharacteristics tobe
considered are immersion and interaction, and the high-quality graphics with photo-
realisticcharacteristics.
There isreason tobelieve that the virtual theatre could bethe natural development
for archaeological high resolution models.This project intends todevelop communication
bybuildingVRGISmodelsforthevirtualtheatreinordertopermit thecoJlectivereal-time
interactionbetweenthescientificcommunity andourapplications.
Acknowledgements
Thanks are due to Dr.Maria Elena Bonfigli, Dr. Andrea Gaudenzi and Professor Aldo
Paolo Palaretiof BolognaUniversity for thedevelopment ofJava andVRML interfaces. I
especially thank the Director of the Terramare Project, Professor Mauro Cremaschi,
UniversityofMilan, andhisteam for thefundamental supportof myapplications. Finally,
a special mention forDr.Gianna Ayala andDr.Savino Di Lernia for their friendship and
cooperation.
M. Forie / Arcbaeologyvirtual micro-topography
References
[I] M. Forte, Image processing applications inarchaeology: classification systems of archaeological sites in
the landscape, in T.Madsen, I.Scol1ar (eds.) Computing the Past CAA92. C<Jmputer Applications and
QlUJIIluativeMethods in ArcMeology, Aarhus, 1993,pp.5 3 ~
[2] M. Forte, Scientific visualization and archaeologicallandscape: the case study of a terramara, Italy, in
Archaeology and Geographical Information Systems. G. Lock and Z. S ~ i c (eds), London, Taylorand
Francis, 1995, pp.231-238.
[3] M. Cremaschi (ed.), Ambiente, insediamento, ecoilomia durante la media etAdel Bronzo. L'ltalia
settentrionale,Rassegna di Archeologia, 10,1991-92,pp.145-188;
[4] M.Cremaschi, A.Ferretti, andM.Forte, Tecniche digitali edi visualizzazioneingeoarcheologia:it caso di
studio del1aTerramaradiS.Rosa diPoviglio,Archeologia e Calcolasori, 5, 1994,pp.305-316.
[5]M. Forte, and A.Guidazwli,Archaeology, GISand DesktopVirtual Reality: theARCTOS Project,in H.
Kamermans, K. Fennema (eds.), Inleifacing the past: Computer Applications and QUOIItitative Methods
in Archaeology, CAA95, Analecta PrehistoricaLeidensia, n.28,Leiden, 1996,pp.439-451
[6]MalinaJ., andVasicek Z.,Archeologia. Storia, problemi, metodi, Milano, 1997.
[7] Cremaschi M., Ferretti A., and Ravazzi C, Geomorphological evidence for land surfaces "clearedfrom
forest inthecentral Po Plain (NorthernItaly)during theRoman period, inB.Frenzel(ed.), Evaluation of
Land Surfaces Clearedfrom Forests in the Mediterranean Region during the Time ofthe Roman Empire,
ESFProjectEuropeanPaleoclimateandMan,5, 1994,pp. 119-132
[8]hllp:/lwww.casa.ucl.ac.uklvirginlcatl
[91htlp:/lwww.gsaup.ucla.edul-robinlESRIlp308.htrnl
[101Liggett, R,andW. Jepson. An integrated environmentfor urban simulation, inProceedings ofthe' Third
lntemational Conference on Computers in Urban PIQIl/Iing and Managemenl, Atlanta, 1993, pp. 565-
5n .
[III Liggett, R., and W. Jepson, Implementing an integrated ,environment for urban simulation: CAD,
visualizationandGIS. inA.Koutamanis (ed) Visual Data Bases, Avebury, 1995.
[12] MaruyamaM,Mindscapesandscience theories, Current Anlhropology, 21,1980,pp.589-599
[13] hltD:/lwww.cineca.itlvisitlvirtualtheatre.htrnl
.'j/. "'"
214
Beyond theMap
G. Lock(Ed.)
lOS Press, 2000
BeyondtheMap:
harmonisingresearchand
CulturalResourceManagement
NeilLang
English Heritage
National Monuments RecordCentre
Swindon
Abstract:Inthe last fifteen years, impressive progress hasbeen made inthecreation
of text databases of the historic environment in England. Typically, their contents
have been used to support Cultural Resource Management (CRM) applications.
Although 'researchers' have made use of these, this has not lead to a positive
understanding of the respective information needs of research and CRM projects.
This paperargues foran improvedrelationship,based on commonstandanls,and an
agreed cycle forinformationexchange. It isargued that there are substantialbenefits
for both parties which are not being realised, becauseinformation is not being used
and sharedeffectively.
Although text dalabases are now well developed, investment in spatial
infonnation systems has often lagged behind. Exchange of information between
'researchers' and 'culturalresource managers' is desirable irrespectiveof the 'type'
of information (text, images etc.) concerned. However, geographic information is
particularly interesting because, as a relatively new area for CRM, there is wide
scope to influence it, while the analytical capability of GIS offers much greater
potentialasaresearchtool thantext alone Using acase study based on the Heritage
SpatialInformationService (HSIS), this paperpresents adynamic model for relating
research to CRM (and vice versa) and explores the benefits which flow from
adoptingthis forgeographicherilage information.
1.Introduction
"...the academic community of archaeologistsstill has a responsibility to help ensure the
future of the archaeological heritage in Europe. One important way it can do this is by
using its research and education potential in the development of GIS applications for
ArchaeologicalResourceManagement."[1]
A paperdiscussing the application of GISto themanagementof the historicenvironment
might,atfirstsight,seemoutofplaceinavolumeconcernedwithnewresearchintospatial
analysis in archaeology. However, I argue that there should be a natural relationship
between these two activities, which is fundamental to the future health of research and
conservation. It is a relationship that does not work satisfactorily at present but, with
furtherattention, could radicallyimproveboth areasof interest.In this paper,threetopics
willbeaddressed:
215
N.lAngIHarmonising research andCRM
I) TherelationshipbetweenresearchandCRMuse
2) Tbe roleofstandardsinsupportingthisrelationship
3) A case study, illustrating this relationship based on the new Heritage Spatial
Information Service(HSIS)
In a relativelyshort periodof time, there has been an impressive rangeof applications to
which GIS has been applied in Europe and the USA. GIS has had such successful
penetration that, on occasions, it has been mandated for use in projects where it actually
adds little value. Despite issues of environmental determinism, and the availability of
suitable data for certain research applications, its value in research has been widely
demonstrated inagrowingbodyofdocumentedcasestudies. In theUK,latterly,there has
been increasingtake-up byorganisations responsible for heritage managementbut to date
therehavebeenfewlargescale 'mature' applications.
The GIS market has changed rapidly over the last five years. The distinctions
betweenCADandGISarebeingeroded,andbasicfacilities forspatialanalysisanddisplay
areevenbeingincorporatedwithinPCofficesuites. Forthepurposesofthispaper,GIS is
taken to mean a collection of hardware, software, rules and techniques for data
managementandproceduresforderivinginformation.
Archaeological applications of GIS technology have tended to divide into two
principalthemes:
ResearchUse, wheretheprincipalaimhasbeentotesthypothesesandfindout new
information, often heavily influenced by geographical.models. Typically, it has
been conducted in an 'academic' University environment, and on comparatively
smalldatasets.
Cultural Resource Management, where the emphasis has been on supporting the
managementof heritageassets through moreeffective organisation anddisplay of
the informationresource. Usersareoriented towards infonnation management for
large datasets, normally compiled originally as text databases, than with 'new
discovery' oranalysis.
As thesetwoareashavematured,thedistinctionbetweenthem is-startingtoblur. Indeed,
the very idea that there shouldbe adistinction is hotlycontested by some, but an aspect
which helpsdifferentiatethemistheextent towhichanalysisordata'managementis their.
primaryconcern.
2.Research and CulturalResourceManagement
Cultural ResourceManagement(CRM)offers the single strongest growthareaforGIS in
archaeology[1], [2],[3], [4]. Ithastheadvantageofbeingable tointegratedisparate. data
typesintoasingleenvironment,distributionmapping,images,textdatabasesetc.,reducing
time to retrieve informationand make decisions. Itcan also integrate separate, spatially
referenced data sets whichcan be visualised into a common environment. GIS can be a
useful catalyst to review the way information flows are managed, and modelling these
againstbusinessprocesseswithintheorganisation.
TheadvantagesofaGISovertraditionalprocedurescanbeconsiderable. In a UK
local authority, for example, mostplanningenquiries start with the identificationofthe
areaof landtowhichtheapplicationrefers.This isthencompared toII: hardcopymapbase
- ~ - ~ - - - - - - . ~ - - - - , - - , , - - - . - . . _ - ~ ~ ~ - -
-
-
216 217
Beyond 1MMap N. Lang / Harmonising research. and CRM
showing the location of sites and monuments and the unique identifiers of any which are
likely to be affected are recorded. These are then checked on the text database, and perhaps
against secondary record files to examine archive related to them, photographs, plans,
reports and so on. There are obvious benefits to being able to view these elements within a
single spatial interface including speed, ease of handling, requiring less office space, and
preparation of subsequent reports.
GIS are becoming increasingly common tools for the national and local authority
inventory records (Sites and Monuments Records, or SMRs (3]), and are essential elements
of the 30 or so Urban Archaeological Databases developed in England (intensive records
for the management of historic urban areas (5]). Given this widespread take-up, GIS could
play an important role in more closely aligning the needs of research and cultural resource
management. I will use the term 'inventories' here to describe the databases, both text and
spatial, commonly used by CRM.
I have suggested above that research and CRM are separate functions. This is not
put forward as a theoretical ideal but reflects the reality of how these two areas have
developed in the UK. CRM applications have sometimes been seen as a 'poor relation' of
the more exciting world of research. In fact, this is an unfair perception. Researchers often
lack the large datasets which are held by curators, and their research potential remains
largely underdeveloped. Conversely, the curators lack the resources (and sometimes
expertise) to exploit the research potential of their data holdings.
Ideally, CRM and research should have a symbiotic relationship, in which each
feeds upon the other. This is set out in Figure 1, below. Although there are exceptions, on
the whole, it is not happening in the UK, or, apparently, elsewhere in Europe as a regular
process. In part this stems from distrust and misunderstanding about each other's aims and
needs. There are also barriers created through issues such as copyright, charging regimes
and concerns over plagiarism. Changing the perception of each other's needs, establishing
na_x-w.. no_x:-w,.
Figure I. A schematic diagram depicting an idealised relationship between
Cultural Resource Management inventories and research.
a sounder infrastructure (through standards) and agreements on the generation and
exchange of archaeological data would greatly enhance the value of research to CRM and
vice-versa.
The cultural resource manager is often under considerable pressure operating within
the UK local authority planning system. SMR functions do not leave abundant time to
conduct research to enhance the dataset upon which planning and other decisions are based.
Also, software developed as a tool for data management may not be ideal to support the
complex analyses required by some research fields, even though this could greatly facilitate
the work of the cultural resource manager. . .
At present, a 'passive' relationship exists between the researcher and the cultural
resource manager although this does not have to be the case. The steps in a more positive
dynamic relationship may be set out as follows. The first point of research is consultation
of the permanent knowledge-base, i.e. one or more repositories, or 'data stores', making up
an inventory of the historic environment for a given administrative area. The 'data stores' .
making up the knowledge base may be linked, as part of a network, or entirely separate. In
the UK, for example, there are 'official' national and local inventories (the National
Monuments Record and Sites and Monuments Records), as well as a number of databases
in the voluntary sector, or iii private hands. Contents of these will vary from place to place,
but typically include historic monuments (including buildings), events occurring on land or
property units (such as excavations, site visits, observations, and surveys, including
'negative' surveys which find nothing of interest), and the locations of archive materials
(including bibliographic references).
All inventories are likely to be an imperfect dataset for the needs of the researcher.
By their nature, they are designed for the needs of general management and to facilitate
planning or decisions to afford statutory protection to sites, and therefore may not be
capable of answering the detailed questions of the academic researcher without further
enhancement. But this is not a reason to discount them as a research tool. The next step in
focusing research is the gathering of data to form the specific research database (a database
of middle Saxon pottery for example). This will almost certainly incorporate new
information from secondary records, not yet included in the inventories, and may involve
original fieldwork which discovers other monuments or events which have not been
previously recorded. It is also likely to correct, or place a different interpretation upon
some inventory entries.
Following this, the material will be synthesised, and may result in general or specific
modifications to the original data. For example, the conventional chronology for a whole
group of cultural material may bechanged, or particular items may be re-classified, The
final phase involves dissemination of the results, submission to the sponsoring institution
where appropriate, publication of the research, and the preparation of data for passing back
to the information repository to be used in managing the resource, This enables future
researchers to benefit from the newly discovered or synthesised knowledge and to re-
analyse the material. This is the ideal process which should be worked through.
In practice, many researchers have been dismissive of the inventories, either because
of the breadth of content, or the level of detailed indexing. For example, inventories do not
normally record artefacts from sites, except where these are significant evidence for the
site's existence. They will not be able to answer specific requests such as "provide a list of
all settlement sites which have produced middle Saxon pottery" (though they would be able
to provide a list of the settlements, with details of excavation reports etc. which could be
checked by the researcher). On the other hand, curators have expressed dissatisfaction with
researchers. This can stem from the common practice of framing very broad requests fer .
~ : - :
218 219
Beyond the Map
information (e.g. everything indexed as Bronze Age for someone researching Bronze Age
logboats which Can produce several feet of wasted printout). Another problem is if
information is not returned at the end of the research, or is returned in a form which
requires a lot of work to assimilate (e.g. a printed thesis, entries in which may not even be
cross-referenced to the unique identifier from the inventory database used as a source).
The relationship of inventories and research requires re-definition based upon a
formal business model of reciprocal benefit. Inventories often hold information vital to the
beginning and progress of research, and research is the key to the refinement and
empowering of the dataset to make it effective for CRM. Access to the dataset given by the
cultural resource manager to the researcher should come with conditions attached, key
deliverables expected back from the researcher in exchange for the basic data.
Data standards, format, content and timetable should all be agreed in advance. The
validitylrelevance of the research also needs to be considered (whether it is valid, has it
been checked, is it correct, is it sufficiently reliable to be incorporated into the central
archive?). The incorporation of metadata is becoming increasingly important, some of
which can be extensive. This might, for each record answer questions like who created the
record? How senior/ experienced was he/she? Who checked it? Where did it come from?
What map-scale and product was the record captured against? In return for co-operation
from the researcher, data should not be unreasonably withheld together with support (e.g.
by providing personal knowledge which would assist the researcher which may not be held
within the record).
3. The role orstandards
.....before the powerful new engine of GIS is hitched to a rickety old cart, there is a need for
more strategic thinking by national bodies, greater co-operation and discussion between the
creators, curators and users of the records, and fresh research anddevelopment" (6].
This quote was written in the context of the difficulties of assimilating local and national
datasets into a coherent single view for the national UK Monuments at Risk Survey
(MARS). It is illustrative of the difficulties research users face, but also of the expectations
which are held and often require qualification. If we are to forge a more successful
relationship between research and CRM, there must be a solid platform for the structure
and interchange of information, and greater understanding of respective needs. An
important contribution towards this will come from the promotion and application of
standards.
Standards specify or provide rules and guidelines for commonly repeated processes
to arrive at consistent results which can be communicated between supplier and user. For
geographical information, examples of the application of standards include:
1) data capture
2) data depiction
3) data quality
4) transfer of information
5) metadata
Successful standards fulfil a genuine need within a user community. They must be
communicated effectively, and become widely used, so that their benefits are generally
N. Lang / Harmonisingresearchand CRM
realised. Standards initiatives can draw together a fragmented user community, improving
co-ordination and awareness of relevant developments. For example, resource discovery
metadata can help users to find useful data, and can inform about quality and
suitability/fitness for purpose. This can avoid needless duplication of effort, sometimes
within a single organisation! So, rather than being an optional chore, standards should be
seen as a necessary part of good data management practice.
In the last ten years, a range of formal standards has emerged for geographical
information developed and endorsed by national and international standardisation bodies,
These include the Committee European de Normalisation (CEN), the International
Standards Organisation (ISO), and in Britain, the British Standards Institute (BSI). In the
UK, the development of the National Street Gazetteer (NSG) is an example of a standard
developed to meet a practical need. The scope of standards includes:
I) How the real world is represented and described in data - data models and structures,
data dictionaries, encoding, metadata
2) The selection of real world objects which will be captured on maps or in data - a
catalogue of representation. . .
3) How data is captured - specification for capture, the type of locational referencing
4) The quality of data - positional accuracy, attribute accuracy, currency (how up-to-date),
content, logic consistency
5) How the data is accessed and updated
6) The testing of data - quality evaluation procedures
7) Thetransfer of data - transfer formats
. .
There is also agreement emerging, at a high level, on the underlying data model for
Inventory databases (e.g. throughCIDOC, CAA), Figure 2. However, this model is still
vested in the realm of text databases and spatial representation of the three principal
entities, authorities, observations and interpretations, needs to be incorporated.
Co-ordination of standards across the archaeological communities will be an
important area if greater harmonisation between research and CRM applications is to be
achieved, particularly given the expense and potential for duplication of data capture.
Standards for GIS have been slow to gather momentum in the industry as a whole, though
there are welcome initiatives.. For the researcher, it is usually easier to apply standards.
within the context of the individual project. However, a dialogue is necessary to ensure
standards cover the entire process from database design, capturing the data, to ensuring that
it can be transferred to other systems.
4. A case-study - The Heritage Spatial Information Service (HSIS)
English Heritage, is. at the time of writing, in the closing stages of implementing .the
Heritage Spatial Information Service (HSIS). HSIS is a partnership between English
Heritage and ffiM to develop and provide a managed geographical intormation service for
heritage data. This has been developed under the government's Private Finance Initiative
(PFI). A contract, which will last for five years from the date when the services become
operational, was negotiated and signed with ffiM in 1998. theservice is due to go live in
Spring, 2000. It is described here as an illustration of the potential for successful
integration of research users and curators of data.
-, .... - "'\
220 221
Beyondthe Map
OUTLINE DATAMODEL
11" _
Figure 2. A high level data model showingthe principalelementsof an Inventory database
4.1 Background to the Project
In April, 1999, The Royal Commission on the Historical Monuments of England
(RCHME) and English Heritage (EH) joined forces to become the Government's new lead
body for the historic environment, the 'new' English Heritage. It is responsible for field
survey of monuments, maintaining a large, permanent archive of the historic environment,
developing, and advising on conservation policy, managing a substantial portfolio of
historic properties and collecting, managing and disseminating information on the historic
environment.
Each year, through its survey and sponsorship programmes, English Heritage
creates the largest number of heritage records of any single organisation in the country. It is
also the UK's largest single source of heritage information, which is held in the National
Monuments Record Centre (NMRC) in Swindon, the public archive of English Heritage.
The heritage data sets have, at their heart, a geographical component, which can be
efficiently managed through a GIS. Although CAD and GIS-based mapping systems have
been used within different parts of the organisation for many years, these have been
physically separated from one another (often distributed across different offices in the
country).
There are now a substantial number of 'heritage' databases, both large and small,
which have been developed to meet different needs. Collectively these comprise around I
million records. Spatial information is essential for many of the key business operations at
EH, and yet few sections have had access to digital spatial information systems. Many
sections of English Heritage use maps in their everyday work to identify specific sites, such
as listed buildings and scheduled monuments, to record sites relating to the historic
environment. Most of the mapping has been in paper form, with the inherent disadvantages
of maintaining, editing, copying, accessing, fragility, storage space, and currency. In
parallel, the full exploitation of the heritage records has been limited by the lack of
N. Lang/ Harmonisingresearchand CRM
integration between them, and the difficulty of re-sizing paper maps (e.g. to produce a user-
defined scale of distribution map).
The techniques and processes involved in the management of heritage information
have shaped and defined the organisation in a distinctive way. There have been three main
activities. The first consists of identification (either field or desk-based), the second
involves the process of protection (of affording statutory protection to a monument), and
the third is the long term curation of the records as a resource to support internal business
activities and external enquiries. The first group of users may broadly be equated with
researchers. The second and third groups may, taken together, be equated with the cultural
resource managers (combining the functions of support for the conservation process and
records management).
4.2 Description ofthe system:
The HSIS system will incorporate a series of heritage layers including (Figure 3):
I) Scheduled Ancient Monuments (protected archaeological remains)
2) Historic Battlefields
3) Historic Parks and Gardens
4) World Heritage sites
5) Information on terrestrial and maritime sites (NewHIS)
6) Listed buildings, (provision is being made for their inclusion following on from a
successful project to computerise the text records of the statutory lists).
and in time,
7) Events (records of excavations and surveys carried out on monuments) both of which
are held in the NMRC's NewHiS database.
8) Conservation Areas (designated areas in cities, towns and villages where special
planning controls apply to buildings).
Collectively, when all listed buildings have been captured, these layers will contain over a
million records. HSIS is important both for the scope and scale on which it is being
developed. It will be the single largest spatial database for the heritage in the UK. .
The HSIS system is built using.Integraph's GeoMedia Pro software, Figures 4 and
6. It is a desktop GIS, designed .as a data management tool, which can read a number of
GIS data formats without the need for translation into a proprietary Intergraph data-type.
GeoMedia Pro is part of a suite of programs, which is designed to be scalable across
organisations, catering for different types of user. Intergraph see changing needs in the
marketplace for GIS software. Their model characterises three types of user in a typical
organisation, doers, users, and viewers, Figure 5.
Their products range from GeoMedia Pro software (for doers creating spatial data),
through GeoMedia as a user tool (for analysis), to GeoMedia Web as a viewing tool (basic
distributions etc.). In EH, initially, the GIS is being deployed mainly for the doers,
professional GIS users within the organisation. In the medium to long term, web-based
delivery is likely to be more important in delivering the application to more general users
within the organisation. Ultimately, it is our intention to disseminate on-line spatial
information to the general public (viewers). The ability to share information to meet these
different needs is starting to bring data creators and data managers and data users closer
together, and to recognise a common purpose ..
222
Figure 3. Schematic diagram of the HSIS architecture. Maps and data are held as replicated sets on data
servers at Swindon and London. An interface is provided between HSIS and the three large external
ORACLE databases(RSM. NewHISand LBS) andan interface fonn enables further details of HSIS objects
tobe viewed.
. 223
N. Lang/ Harmonising research and CRM Beyond the Map
The Heritage Spatial I.formatioD Service
__n
Intergraph's Model for GIS Usage
-------. 'DOERS'
8SIB
lAY1lIlS
'USERS'
-

'YJEWIRS'
.............
....-....
....T......
RlTUFAoC& lorUIU""L
DolT..-
_ACIS
POIIM(I,. DAT4UUS
Figure 5. Intergraph's model for usage of GIS suggests that as the price/performance of the technology has
changed,successivelylarger marketsforUsersandViewers of spatialdata havedeveloped.
Figure4. The HSIS GUl Interface. This isa standard GeoMedia Pro interface, with a customised menu bar
and functions. In this instance. a proposedlineardevelopmenthas beenbuffered to identify sites which will
beaffected bythe developmentcorridor.
!
.
IN
I i
! I
L ..-
:e:==.
,...

i
":"":' :'="-

-...;...,-..... __M_"
__'_I"
--"
_a''-f'

_.-...
-_.-,.,
Figure 6. TheHSIS User Interface. In this example, a series of monuments have been selected (highlighted
onscreen). 1beirHSIS attributes arehighlighted inatext windowatthe bottom of thescreen. Further details
ofeach monument an:displayed inanOracle interface Form, bottomright.overlying the textwindow.
---
224 225
BeyondtheMap
ThebenefitsofHSIShavebeenenhanced throughtheprovision ofacomprehensivesetof
Ordnance Survey digital mapping through the central government Service Level
Agreement.Previously,thehighcostoflarger-scaledigital mappinghadbeenasignificant
handicapinrestrictingthedevelopmentofspatialtechnologies.
ForEH,thiswillprovide:
I) BaseDataGB( I:625,000vectormapproduct)
2) Strategi(1:250,000vectormapproduct)
3) Complete cover of England at 1:10,000 scale using current OS raster maps. A select
range of Landline data (Ordnance Survey's large-scale vector mapping at 1:1,250
(urbanareas)and I:2,500scale(ruralareas
4) BoundaryLine(adefinitive setofadministrativeboundaries)
4.3Thebusinessneedsitwasdesignedtoaddress
The new facilities, which HSIS will deliver, even in its first phase of operation, will
provide an extensive new resource for users. In EH and the former RCHME, heritage
information systems were often developed in isolation to serve the needs of individual
sectionson different platforms. often to different data standards, whilst duplicating large
amounts of data between them. The need for a morecorporate approach to information
management has long been recognised, but the cost of integration has beenprohibitively
expensive.
Therehasalsobeenalackofintegrationat thelevelof its businessprocesses. The
starting point for field projects (researchers) has rarelybeen a search of whatknowledge
has beencommitted tothe various heritagedatabases. Intum, these havenot necessarily
provided the field teams with the information they would find most useful. In
consequence,puttingbacktheproductsoftheresearchhasbeenseenasanoverhead,rather
than a contribution to the next cycle of investigation. Data collection has rarely been
optimised for ease of transfer to the permanent record. Those responsible for protection
and records management (CRM Managers) have, in tum, failed to deliver the sort of
information service required by the field teams, or to undertake the levels of analysis
withinthedatabaseswhich mightspark newlinesofresearchenquiry. Theyhavebecome
datagatherers,ratherthandataanalysers.
Becausedatabases havehadnocommon enquirytool,theyrequiretimeconsuming
multiplesearchestoanswereven basicquestions.Enquiries totheNMR'sPublicServices
section,forexample,areoftenspatiallybased(e.g.'find meallheritageitemsonapieceof
land'). Just tocover the majorheritagedatabases,thiswould requiresix separatesearches
(theListedBuildingsSystem,Recordof ScheduledMonuments, NewIDS,Photonet aerial
photographindex,excavations index, andarchivecatalogue).To provideacomprehensive
answertothequestionwouldrequirefarmore(includingmanualsearches).
Progressonintegrationis,ofcourse,dependentondatahavingbeensuitablyspatially
referenced, and some substantial databases (e.g. listed buildings containing 370,000
records) are currently only referenced to the local administrative area (the parish).
Capturing very precise references for each listed object (through an Ordnance Survey
nationalgridreference)willbeacostlyexercise.
N. Lang/ Harmonising research andCRM
4.4Benefitsof the HSIS System
TherearefivemajoranticipatedbenefitsinthedevelopmentoftheHSIS:
I) Improvingdataquality:a keypurposeinimplementingthesystemwastoimprove the
locational accuracyandrepresentationofheritageobjects(forexample,thedepiction of
visible and probable extents of monuments,constraint areas and linear features) both
for internal use and to enable highquality information to be incorporated into other
systems
2) Enabling better integration of data sets: through viewing in a common spatial
environment, currently, many of thenational datasets sit on independentplatforms to
different standards. This willenablethe majorstatutoryandquasi-statutoryrecords to
be viewed against the totality of heritage records held in the National Monumeots
Record. Thiswillhaveasignificantinputtofuturepolicyandselection.
3) Automating the reflection ofchanges in spatial information in text databases: when
information in the real world changes, for example the boundaries of administrative
areas, the processof identifyingandcorrecting recordswithinatext database is slow;
expensive and subject to human error. Close attention has been paid to using the
spatial engine of the GIS to control the spatial information held in the external
databases.
4) Improving data retrieval:despite a great deal of effort on methods of indexing text
databases, therearestilla numberof typesofenquirywhichcannotbeeasily resolved
without effective tools for spatial capture and analysis (for example, irregular
developmentcorridors).
5) Improving understandingof data.'in common with many'other records management
organisations, farmoreefforthasbeenexpendedongatheringdatatogether,organising
it, and indexing it than on understanding what it means. The research potential for
many suchrecordsremains largelyuncharted territory. Althoughdataqualitymay not
always fully meetthe needs of researchers,the scale of the datasets allows questions
and manipulation which can rarelybe performed within the scope,for example, of a
typical PhD thesis. HSIS should become an integtal tool for the field teams within
English Heritage and it is also hoped that it will be used extensively by external
researchers. There should be a cycle between 'what we know' as expressed in the
heritage datasetsand 'what wewanttoknow',ourfuture. researchagendas..Currently, .
thereisanuncomfortablegapbetween 'current knowledge' and'deposited knowledge'
(i.e.knowledgeexpressedinacommonlyaccessibledatabank).
The HSIS Service, then, can be seen as not merelythe provision of a GIS. It is about
providing a common organisational framework through which data can be collected,
analysed against a range of other geographical information.and disseminated'to a wide
range of user communities, including the public. English Heritage seeks to help public
accessibility to information about the historic environment, both physically (by making
informationreadilyavailable)andintellectually(byaidingunderstanding).
To illustrate the changes whichcould be brought about bythe implementation of
HSIS,Figure7showsatypicalflowofinformationwithintheorganisation. Priorto HSIS,
mostdata capture, whetherin the fieldor desk-based involved manual mappingand data.
entryforms.Increasingly,thisisnowcaptured digitallyviaGPSandamobiledatabase ina
form directly ready for transfer to HSIS. Prior to HSIS, paper 'maplets' would be
circulatedtoassistindecisionsonconservationandprogressivelyitwillbepossible for

226 227
Beyond the Map
B
o

C
R
E
A
T
I
o
II
DigitalInformationFlowline
Figure 7. Illustration of an 'end to end' digital flow from datacapture, support for conservation decisions,
pennanentrecord curationand public use.
these tobe replacedbyon-lineaccess. Prior to HSIS, the heritagedatasets playeda minor
role indevelopingheritagepolicy, and selectionof sitesor topics for survey. We anticipate
HSIS will progressively become an essential service to support these functions in the
future.
This is not merelya question of one technology (digital) replacing another(paper-
based). It is a question ofbringing together related processes and information flows in a
positivecycle, toenabledifferentbusinessactivitiestobenefitfrom one another. HSIS will
enableEnglishHeritagetomove on from using the heritagedatasets as asimplerepository
for collecteddata, to makinganalysisand understandingof the dataa fundamental process
in its own right. Throughdoing this, it will feedback into the processes through which we
selectfuture topics and themesfor research.
5,Conclusions
In this paper, Ihave suggestedthat there shouldbe a natural relationshipbetweenresearch
and culturalresourcemanagement. At the very minimum, this shouldconsistof apassive
relationship, which should acknowledge the permanent knowledge base, the totality of
knowledge expressed in the heritage data sets held by CRM managers, as being a key
startingpointfor research. Attheend of theresearch.afeedbackloop providingenhanced
or new data to the permanent knowledgebasein a suitable format will completea single
cycleofknowledgetransferand development.
The benefitscan be enhancedthrough a dynamicdialogue betweenresearchers and
cultural resource managers. This will include active collaboration on projects, and using
CRM databases as a test bed for existing hypotheses, and the generation of new research
questions. Spatial technologies can be highly suitable for supporting this, because they
enablerapidquestioningofdatawith avisibleoutput.Fromthis, theresearchercan start to
N. LAng / Harmonising research and CRM
searchfor patterns, and investigatewhetherthese have any meaning. Althoughsuch spatial
analysis has often been criticised as environmentally determined, the scope of spatial
research has moved beyond physical geography to the cognitive and the symbolic. Since
CRM managers are often responsible for very large datasets, they are able to offer the
means for analysis at the regional or national level that would be impossible if the
researcherwere tohave togatherthe informationfrom scratch. Culturalresourcemanagers
shouldbe able to suggestavenuesof researchwhich wouldbe suitablefor everythingfrom
an undergraduatedissertationtoadoctoralthesis.
In this way, researchers can provide a very positive service to cultural resource
management, by improving the data sources supporting decision making. In tum,
researchers are able to gain access to a rich information resource which they would not
normallyhave the resources tocompilethemselves. Cultural resourcemanagersoften lack
the time (and do not necessarilyhave the technical expertise) to performdetailed analysis
on datasets themselves. For example, most CRM databases reniain as two-dimensional
maps, whereas researchers could expand the scope of this data either by adding the
dimensionof height (e.g. throughaterrain model),or time (e.g. through atemporal GIS).
Therewill probablyalways be groups who refuseto engagein this dialogue. Either
it is too time consuming, too laborious, Or there is not enough incentive to engage.
However, an areaon which we can all agree is the importance of making our results
publicly available. There should be no dispute on the importance of disseminating the
results ofarchaeological research to a public, as well as a professional audience. GIS
shouldbe animportantcomponentin makinginformationpubliclyaccessible.
Perhaps the most exciting recent development.is the application of spatial
technologiesto the Internet. Along with spatiallyenabled web sites, the developmentofe-
commerceseems set to have a.significantinfluenceon the future publicinvolvement with
theheritage.E-rommercewillopen upthe marketfor internetaccess, so that devices which
can access the web will becomeas commonas telephones. Theweb will then becomean
importantsourceofinformation on leisure pursuits. A service offering suitablypackaged
information could provide, for example, information on sites within a particularjourney
time from the home, associated transport information, opening times and access
arrangements plus downloadable maps and information about the site. The current
popularity ofTVArchaeology, and the various activities. stemming from it (such as the
Time Team Club) indicate there is not only an audience of tele-consumers, but also a
publicthat wants toget involvedwith their historicenvironment. And apublicthat engages
with itsheritagewill bethe surestwayofprotectingarchaeologyfor the future.
References
(I] van Leusen, P.M. GIS and Archaeological Resource Management: AEuropean Agenda, in G.R. Lock
and Z.Staneic,ArchaeokJgy and Geographic Tnfonnotion Systems: A European Perspective. Taylorand
Francis, London, 1995, pp. 27-41.
(2) Kvamme, K. Geographic Infonnation Systems andArchaeology, in G. Lock andJ. Moffett (eds.),
Computer ApplictUions and Quantitative Methods in Archaeology 1991, International Series 577,
TempusReparatum,Oxford. BritishArchaeologicalReports, 1992.pp.77-84.
(3] Lang, N.A.R. From model to machine: proclU"eJneDt and implementation ofGeographical Information
Systems for County Sites andMonuments Records, in J. Andresen, T. Madsen and1.Scollar (eds.),
Computing the {HUt. CAA 92CompUler Applications and Quantitative Methods inArchaeokJgy 1992,
Aalbus, 1993, pp.167-176. -
229 228
Beyond the Map
[41 Miller, A.P. How to Look Good and Influence People: Thoughts on the Design and Interpretation of an
Archaeological GIS, in G.R. Lock and Z. Stanei':, Archaeology and Geographic Information Systems: A
European Perspective. Taylor andFrancis, London, 1995, pp. 319-333.
[51Miller, A.P. The York Archaeological Assessment: Computer Modelling of Urban Deposits in the City of
York, in J Wilcock & K. Lockyear (eds.), Computer Applications and Quantiuuive Methods in
Archaeology 1993, International Series 598, Tempus Reparatum, Oxford. British Archaeological
Reports, 1995, pp. 149-154.
[6J Darvill, T. Researching Archaeology? ThaI'S tough, British Archaeology 9, York, CBA, 1995, p. 6.
Beyond the Map
G. Lock (Ed.)
lOS Press, 2000
Session 4 discussion:
Archaeology, GIS and
Cultural Resource Management
Philip Verhagen
RAAP
Amsterdam
1. Introduction
The papers presented in the session on archaeology, 'GIS and Cultural Resource
Management have little in common at first glance. Nevertheless, on closer inspection three
issues emerge that are of concern to all three contributors.
11Je first of these is scale, a recurring topic in many discussions on GIS inside and
outside of archaeology. In the context of Archaeological/Cultural Resource Management
(CRM) it is often seen as an issue of data quality; in fact, this is only part of the story.
Scale is also closely related to our definition of archaeological entities, and the way in
which we use these for CRM purposes.
The second issue is raised by Lang (this volume), but will be of concern to anyone
working in GIS and CRM: the uneasy relationship between archaeological research and
CRM, and its implications for the collection and management of large national and
regional archaeological databases. Lang states that well-devised data standards may be
helpful in bridging the gap between research and CRM. There may be more to it than that,
however.
The third issue (that was not raised as such during the workshop) concerns the
future of archaeological GIS. What are the opportunities for GIS in CRM now that GIS is
more and more becoming a tool for' information dissemination, rather than just data
storage?
2. Scale issues
A recurring topic in discussions on the use of GIS in CRM is the issue of scale-.Usually,
the term scale in this context refers to the level of detail at which the information is
collected. However, scale can have a number of different meanings; Lam and Quattrochi
[IJ provide a useful classification:
Cartographic scale refers to the geometric component of scale, the ratio of real world to
map co-ordinates. Cartographic scale defines the level of detail at which data is depicted.
Goodchild and Quattrochi [2J note that cartographic scale helps to define the content (the
type of features likely to be shown), as well as to define the positional accuracy of the
231 230
Beyond the Map
mapped geographical features. Scale is often used as a surrogate for positional accuracy
and provides content for many digital databases that never existed as paper products.
Geographic scale is seen as the extent of the study area. A large-scale study usually
implies cartography covering a large area.
Resolution is the actual precision of the measurements involved.
Operational scale, finally, is the scale at which a certain geographical phenomenon is
operating. The operational scale is logically the scale at which one would like to observe
the phenomenon, and as such, it is directly related to the cartographic scale.
It is this operational scale (or level of observation) which is of most interest, as it
determines what should be mapped. In papers on the integration of archaeological data in
large geographical databases however, little attention is usually paid to scale. If attention is
paid at all, much of it is devoted to the issue of resolution (e.g. [3]; Giannini et al., this
volume), as archaeological data stored in CRM databases are notorious for their positional
inaccuracies. Scale (or resolution) in this context is seen as a measure of data quality. For
some reason however, there is little appreciation of the fact that the choice of the features
to be mapped has a direct relationship with the cartographic scale applied.
A clear example is the mapping of site distributions based on field survey data [4],
[5]. Archaeological databases are usually filled with point locations of archaeological
observations (find spots), however, the basic entity commonly used for archaeological
resource management is 'the site'. Obviously, an archaeological site (or complex, to use a
more neutral term [6]) may contain many observations: various people may visit the same
site, find different materials at different locations, and possibly document these at different
resolutions. When trying to define the actual delimitation and position of the
archaeological site the observed point data has to be aggregated to a new level of
observation. This can be done by finding a new central co-ordinate, or by aggregating the
observations in grid cells or land parcels. The resulting loss of information on how the
information was actually obtained, may lead to highly variable degrees of accuracy of site
locations, unless systematic surveying was done at all sites. This simple conclusion has
severe consequences for CRM: curators will only be able to determine if there is a threat to
a known archaeological site if they can be reasonably confident of the site's location on
their maps. In practice they are not, and will advise to carry out an additional survey in
order to determine the actual extent of the site. Obviously, one of the key geographical
phenomena of archaeology, the site, is thus defined so that it cannot be easily mapped
without detailed and systematic observation. The operational scale of the site is not the
same as the operational scale of the find spot, and a clear and accepted methodology of
aggregation from find spot to site does not exist.
An interesting approach to tackle this problem is presented by Lang & Stead [4] in
which they try to define a more object-oriented approach, where data entities are defined in
a manner that allows for hierarchical grouping of features. This hierarchical classification
scheme is of course commonplace in other disciplines: social scientists are often dealing
with data collected at the level of municipalities that can be aggregated to provincial or
national levels. Similarly, soil scientists have devised a number of hierarchical
classification schemes in order to describe soil types. In essence, the technique is equally
applicable to archaeological data as shown by Giannini et al. (this volume), where the
sequence from individual buildings to insulae provides a perfect example of moving
between operational scales without changing positional accuracy. In essence, this approach
requires the database builder to choose a basic level of information that can actually be
P. Verhogen I Session4 discussion
collected, and think of possible aggregates that can be built from there. This also means
that if these aggregates can not be made, they should not be considered. As an example, we
can return to the find spot/site problem: in most cases, we will have great difficulty in
constructing sites from individual find spots. We can, however, create a number of other
aggregate maps from the find spots, like number of observations per parcel, or
classifications by archaeological period. These will not lead to the immediate definition of
site boundaries, but will at least be able to inform the curators about what to expect in a
less ambiguous way.
A second scale-related issue that tip to now has received little attention is the
integration of non-archaeological data sets and archaeological information systems. It
seems as if it is assumed that non-archaeological data is more reliable and more accurate,
than the archaeological information itself. This may be true for modem day topographical
maps, but is certainly not true for a lot of the environmental mapping that exists. This.
difference in accuracy is closely related to scale. Threeexamples will be given to illustrate .
this effect.
The effect of operational scaler- French geological maps show a surprising number of
differences between map .sheets - even though they are all published by the same
organisation. It almost seems that for every new sheet published a new hierarchical
classification system is devised. Of course some general similarities can be observed but
there are substantial differences in classification schemes, the boundaries between units
and the interpretation of the geological units. This is obviously a consequence of changing
opinions on the operational scale to be chosen for geological mapping. Basically, this
means that French geological map sheets are in varying degrees 'unreliable' and don't fit.
This has direct consequences for the use of these maps in CRM: what if we try to locate the
areas where sites may be buried under alluvial deposits and the maps don't show the right
information?
The effect of resolution 1: In the Netherlands, soil maps are available for the whole
country. Fortunately, no classification problems exist here, as the system used was agreed
upon in the 1960's and has been in use for every new soil map produced since. The
operational scale of soil mapping has therefore been fixed for a long time. On the other
hand, the resolution of the mapping is causing accuracy problems. According to the Dutch
Soil Survey Institute the chance of actually finding (at a random location) the soil type
mapped is in the order of 70%. This, isa direct consequence of the mapping procedure
followed: soil boundaries are drawn On the basis of 10 to 25 observation points per 100
hectares, and are generalised from 1:25,000 to 1:50,000 scale before publication of the
maps. Again, this may have serious consequences for CRM, for example when site
location predictions are made based on the presence or absence of certain soil types.
The effect of resolution 2: DBMs 'are among' the most popular add-ons to archaeological
information systems and are always created using an interpolation method. The accuracy
of the DBM not only depends on the resolution of the base data but also on the
interpolation method chosen. Forte (this volume) gives some clear examples of the
consequences of choosing a different interpolation algorithm using the same base data.
Common accuracy problems withDEMs include: 'terraces' on the position of the contour-
lines used for the interpolation, 'undershoots' and 'overshoots' in valleys and on ridges,
smoothing of sharp terrain transitions, and lack of data points in low-lying areas, resulting'
in large areas with a monotonous slope.
....-1.'
232
Beyond ~ Map
One of the problems encountered here is that the desired resolution of non-
archaeological data is often finer than the resolution of the available digital information.
This is also a question of being able to afford the acquisition of more detailed maps.
Chartrand et al. [7J have shown how this may influence the choices to be made: in order to
have a suitable digital map base they compared prices of the digital Ordnance Survey maps
and a commercially available product. For reasons of pricing, they chose the latter
although this was at a scale of I:250,000. Even though the data were bought for use in a
non-urban area, it seems a questionable choice - not to mention the fact that the data was
digitised from 1928 base maps. Similarly, the Dutch State Archaeological Service decided
not to buy the desired I: 10,000 digital topographic maps, but instead settled for 1:25,000
land use maps which only contain a limited number of topographical features.
In short, it seems that archaeologists should not only be more critical about the
contents of the archaeological databases they are working with, but also about the non-
archaeological digital mapping they are using. In order to do this they should better define
what they will actually be using the non-archaeological data for. This will also make it
easier to generate the funds necessary for the acquisition of more detailed mapping.
3. Cultural Resource Management, research and data standards
Lang (this volume) makes a strong plea for the better integration of research archaeology
and Cultural Resource Management based on standards of data collection and exchange.
Until now, it seems that existing data standards have mainly been defined for the benefit of
CRM purposes. The anticipated inflow of new archaeological data from non-CRM oriented
archaeologists has, however, been minimal and Lang raises an interesting point: is this
really the consequence of badly designed, cumbersome standards and accompanying
software that puts off anyone who is willing to enter their archaeological data into the
system?
In general, it is well known that the documentation of archaeological finds has
always been problematic. The amount of data that has never been published in an
accessible form is probably staggeringly large, possibly to do with expected benefits of
standardised documentation that have not materialised. For commercial archaeological
firms, this is not a big issue: if their contractor obliges them to deliver the data in some
standard format then they will do that - for a fee. In a situation where archaeological data
has acquired commercial value, the need for data standards is quickly recognised. For non-
commercial organisations (universities, but also amateur organisations), the benefit of
committing to data standards is less clear. Certainly when the available data in the 'central
repository' is difficult to access, and its quality is not regarded sufficient for research
purposes, these organisations will not feel that they benefit directly from contributing their
data to the central database. Apart from that, lack of time when it comes to documenting
excavation records or survey results is often mentioned as a reason for not submitting data.
It seems that there is a concern among Cultural Resource Managers that they will
gradually loose contact with the research community and Lang's approach is interesting,
therefore, as it tries to reactivate the dialogue between the two. One of the key issues is the
benefit to researchers of submitting their data to a central database. Experience in the
Netherlands has shown, however, that the fact that universities were involved in defining
the data standards for the national archaeological information system is not enough to
guarantee the inflow of data from them. It is up to the researchers to inform the Cultural
Resource Managers what they would like to see coming back to them from the central
database, and obviously a more user friendly interface, with the option to do intelligent
querying of the data is one of those issues. Another is the management of the database
. 233
P. verhagen / Session 4 discussion
itself. If a researcher submits data in the right format.jt is very disappointing to find out a
few years later that this data has not yet been entered into the central database - especially
as, in theory, this could only take a few minutes.
This is a basic dilemma: a central archaeological data repository will only be
valuable to each of the data providers if it is reliably comprehensive. As long as
considerable amounts of data are either not submitted or not entered and checked for their
quality, the actual utility of the repository will be questionable. It is hard to see how to
overcome this problem without a legislative framework. In the Netherlands for example,
there is not even a political consensus that it is a government task to maintain a central
archaeological database, let alone that archaeological researchers are obliged to submit the
results of their studies to it. Secondly, it seems that lack of money is one of the enduring
problems in archaeology. When universities complain that they don't have the time to enter
their data in a compatible format for use in a central repository, this basicallymeans that
they cannot make the funds available. This is of course different for commercial
archaeological firms, who will include the time needed for data entry in budget proposals -
when the obligation is the same for each company, there will be little opportunity for
competition on data entry rates. A logical step.would therefore be to apply for government
funds for entry of archaeological data from non-commercial organisations. This also
implies that the data entry doesn't have to be carried out by the uriiversities themselves if
they find this unattractive: if they have money available, they could hire trained people
from a government organisation or a commercial firm, for example. In order to achieve
this, the status of archaeological data, i.e. the importance that is attributed to a high quality
national archaeological database, should be much higher than it currently is. However,
given the trend of increasing commercial value of archaeological data for planning
purposes, it may not take that long before this is realised.
4. Tbe future
Even though archaeologists are more concerned with predicting the past, a few words
should be dedicated to the future of GIS technology in archaeology. The small commercial
market that archaeology presents has led to a situation that is slightly different from the
'main-stream' GIS market. To give one example: in Dutch archaeology, MapInfo is the
main GIS package used instead of ArcView, which is preferred by. most non-
archaeological organisations. This is a direct consequence of the fact that archeology has
never been able to afford buying an expensive high-end product like ArclInfo and the
training of people to work with it. Furthermore, there is a lack of data standards and
exchange formats for archaeological data, as there has been little need to exchange
archaeological data with non-archaeological organisations. As was noted above, this
situation is about to change rapidly. The data standard and exchange problem is not unique
to archaeology and at the moment we can see a lot of effort in the GIS world being devoted
to the development of data standards and formats (especially through the OpenGiS
consortium). One of the most interesting aspects is the development of format independent
GIS software. This development will lead within a few years to the proliferation or'
packages that can easily access data from all major GIS and database software across
platforms and networks, thereby reducing data redundancy without forcing the user to
migrate to a completely new software and hardware environment. In the longer term, 'it
also means that data and data-types that were hitherto inaccessible in a GIS environment
may be linked. 3D-modelling and virtual reality, for example, are at the moment
technologies of only minor importance to CRM, but the linking of these techniques to
conventional GIS'will certainly be valuable, as is demonstrated by Forte (this volume).
235
234
Beyond the Map
Apartfrom that,thereisagrowingawareness intheGIScommunity thatthevalue
ofGIS isintheinformation itself,andthataccessto thisinformation should beaseasyas
possible.Afewyearsago, Intergraphpresenteda 'data user' modelthatdistinguishedthree
typesofGISusers:
doers, thespecialistswhocollectgee-information,processitandmakeitavailableto
thepublic
users, theresearcherswhowanttoanalysegeo-information
viewers, whoonlywanttoseetheresultsofdatacollectionandanalysis
Untilafew yearsagoGIS wasamarketfor 'doers', withhigh-endpackages likeArclInfo
dominating the market. It is only over the last four or five years that the 'users' have
become more attracted to GIS, through the distribution of packages like ArcView and
MapInfo. However, the 'viewers' constitute the largest market opportunity, and currently
ESRI is giving away its 'viewer' package ArcExplorer for free in order to enter the
growing market of e-eommerce. Although it is currently hard to imagine archaeological
data being sold on the Internet,developments in the UK indicate that this may not be a
fictioninthenearfutureLang,thisvolume).
Even without (e-)commercial aims in mind, the accessibility of national and
regionalarchaeologicaldatabasesimprovesgreatlythroughtheuseofweb-basedbrowsing
and querying utilities. At the moment, the Dutch national archaeological information
system ARCHIS (access to which is currently restricted to a limited number of
archaeological organisations) is being transformed to a web-based system. Currently, it
offers the opportunity to do on-line selection of find spots and monument records in an
HlML-form, as well as graphical display and selection of find spots on a topographical
background.As aconsequence,theactualuseofthedatabasesystemisincreasingwiththe
increasingcapabilitiesofweb-basedquerying.
Coupled to this development, the labour market in GIS is changing from a
specialist-oriented market to a market where 'geographical information managers' will
become more important. Even though the actual amount of 'doers' may not drastically
diminish, therewillbelittlegrowthinthissegmentofGIS.Instead,people whoknowhow
tocouple thedesiresofuserstotheavailablegeographical informationwill be indemand.
In CRM this will not be different: decision-makers will demand that archaeological
informationiseasilyaccessibleandispresentedinsuchawaythattheycanworkwithit.
It is clear that CRM will follow this development and that archaeological risk
assessment will becomeastandard procedure inplanning after the implementation of the
Valletta treaty. Furthermore, many planning decisions will be based on available digital
information, and non-archaeologists willbe involved moreclosely in taking decisions on
managing archaeological resources.Mostarchaeological GISdatabases are currently data
storage and retrieval engines for archaeologists, whereas, in the near future they will be
transformed into integrated decision support tools. The challenge for archaeologists,
therefore, is to present the information needed for informed decision-making in an
unambiguous way.This meansthatthearchaeologicalcommunity willhave togatherand
disseminate not only the available archaeological information, but also its inherent
uncertainties.
References
[I)Lam, N. & D.A. Quattrochi, 1992. 'On the issues of scale, resolutionand fractal analysis in the mapping
sciences'. Professional Geographer 44, pp.88-98.
P. Verhagen I Session 4discussion
[2) Goodchild, M.F. & D.A. Quattrochi, 1997. 'Scale, Multiscaling, Remote Sensing and GIS'. In:
Quattrochi, D.A. & M.F. Goodchild (eds.): Scale in Remote Sensing and GIS. CRC Press I Lewis
Publishers,BocaRaton, pp. 1-12.
[3] Miller, P., 1995. 'Howtolook good and influence people: thoughtson thedesign and interpretationof an
aschaeological GIS'. In: Lock, G. & Z. t o c i ~ (eds.): Arcllileology and Geographical Information
Systems. Taylorand Francis,London, pp. 319-334.
[4] N.Lang & S. Stead, 1992. 'Sites and monuments records in England- Theoryand Practice'. In: Lock,
G. & J. Moffet (eds.): Compiuer Applications and QUQII/ilalive Methods in Arcllileology 1991. BAR
InternationalSeries S577. TempusRepasatum, Oxford.pp. 69-76.
[5] Robinson, H., 1993. 'The Archaeological Implications of a Computerised Integrated National Heritage
Information System'. In: J. Andresen, T. Madsen & 1.Scollas{eds.): Computing the Past. Compll.ler
Applu:ations and QlI.a1ltitative Methods in Arcllileology CAA92. Aarhus University Press,Aashus, pp.
139-150.
/61RoonIa, I. &R. Wiemer, 1992. 'Towardsanew ArchaeologicalInfonnationSysteminthe Netherlands'.
In: Lock, G. & J. Moffet (eds.): Computer ApplU:alions and Quantitalive Metllilds in Archaeology 199/.
BAR InternationalSeries S577. TempusRepasaturn, Oxford,pp. 85-88.
[7] Chastrand,J.,J. Richasds & B. Viner, 1993. 'Bridging the wban-ruralgap: GIS and the York Environs
Project'.In:J. Andresen, T. Madsen & I.Scollas (eds.): Computing the Past. Computer Applications and
QUQII/italive Methods in Archaeology' CAA92. Aarhus UniversityPress, Aarhus, pp. 159-166.
236
Author Index
Bell, Tyler
Bisson,Marina
Boaz, Joel
Chapman,Henry
Crescioli,Marco
D'Andrea, Andrea
Forte, Maurizio
Giannini,Federica
Gillings,Mark
Harris, Trevor
Kamermans, Hans
Kuna, Martin
Lang,Neil
Latalowa,Malgorzata
L1obera,Marcos
Lock, Gary
Niccolucci,Franco
Pareschi,MariaTeresa
Raczkowski,Wlodzimierz
Stancic, Zoran
Stefani,Grete
Tschan,Andre P.
U1eberg,Espen
Veljanovski,Tatjana
Verhagen,Philip
Wheatley,David
85
187
101
49
157
157
199
187
I
xiii, 116
124
180
214
28
65
xi, xiii, 60, 85
157
187
28
147
187
28
101
147
229
I
----.--,---- - - --
'..