You are on page 1of 13

Journal of Archaeological Science 32 (2005) 1252e1264 http://www.elsevier.

com/locate/jas

Backscatter responses and resolution considerations in archaeological side-scan sonar surveys: a control experiment
Rory Quinn a,*, Martin Dean b, Mark Lawrence b, Steve Liscoe b, Donal Boland c
a

Centre for Maritime Archaeology, Centre for Coastal and Marine Research, School of Environmental Sciences, University of Ulster, Coleraine BT52 1SA, Co. Derry, Ireland b Archaeological Diving Unit, University of St. Andrews, Fife, KY16 9AJ, Scotland c Boland Archaeological Services, Arden Road, Tullamore, Co. Oaly, Ireland Received 7 July 2004; received in revised form 9 March 2005

Abstract The backscatter responses of submerged archaeological material remain poorly understood. To address this, a control experiment was conducted in Belfast Lough during July and August 2001. A selection of material and targets was arranged on the seabed in varying orientations, in an attempt to replicate the situation found on a range of archaeological sites where material is exposed on the seaoor. Repeat side-scan sonar surveys of the control array were conducted at variable ranges and sh heights to examine the backscatter responses and resolution of three commercially available side-scan sonar systems. Results indicate that the electro-acoustic specication of the pulse length and the beam angle of the acoustic sources govern the theoretical resolution of sidescan data. Backscatter data from the control experiment indicate both organic (wood and leather) and inorganic (metals, ceramics, glass and varying aggregate and ballast grades) natural and man-made materials can be imaged using side-scan sonar and a wide variety of archaeological material types and concentrations may be dierentiated by their backscatter responses. 2005 Elsevier Ltd. All rights reserved.
Keywords: Side-scan sonar; Control experiment; Underwater archaeology; Shipwreck

1. Introduction Side-scan sonar is widely regarded as the instrumentof-choice in academic research programmes and predevelopment surveys for reconnaissance scale mapping of the seaoor. Over the past decade, as the cost of sidescan instrumentation has decreased and ease of operation increased, Universities, Government Agencies and seabed developers have adopted this technique for maritime archaeological surveys in an attempt to locate shipwrecks and inundated former habitation sites [1,2,4,5,12,16e18,20e22,25].

* Corresponding author. Tel.: C353 27 70 324 884; fax: C353 28 70 324911. E-mail address: rj.quinn@ulster.ac.uk (R. Quinn). 0305-4403/$ - see front matter 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.jas.2005.03.010

Unfortunately, a black-box approach is often adopted in the deployment of these techniques, where the users accept the resultant sonar imagery as a denitive snapshot of the seaoor. However, very little is understood regarding the backscatter characteristics of archaeological and man-made materials in the marine environment, and this often leads to misinterpretation of sonar data acquired for archaeological intent. A second factor that hinders the interpretation of these data, and therefore the application of this technique to maritime archaeology, is the comprehension of the resolution of o-the-shelf sonar systems. Manufacturers manuals often quote theoretical resolutions for such instrumentation, typically of the order of 0.05 m. This implies that an object measuring 0.05 m or greater on the seaoor will be imaged by the specic sonar device. However, the actual geophysical signature and practical resolution of the recorded data are

R. Quinn et al. / Journal of Archaeological Science 32 (2005) 1252e1264

1253

dependent upon many other factors including transmission methodology, transducer ringing, material types insonifed, range, survey conditions, grazing angle and post-acquisition data processing. Until the early 1990s, the majority of marine archaeological side-scan survey was reconnaissance in nature [1,4,20], where the principal objective of the investigation was to locate a shipwreck site close to navigational hazards or ports [1,20,21]. Results from these and more recent marine geoarchaeological surveys suggest that images of large intact objects and artefact assemblage, such as upstanding wrecks, can be insonied with side-scan sonar and readily interpreted [2,4,9,10,18,25], but diculties in imaging and interpreting smaller object types and scattered wreck material exist [17,19]. Exploitation of side-scan sonar by those interested in imaging man-made objects on the seabed resulted in the publication of numerous images, but the majority of these are of sites larger than 5 m in length. Many of the published images are of upstanding hulls of ships or complete aeroplanes [9,10] that are readily identiable and interpretable by the observer. The detail within these images often includes relatively small items, such as blades of aeroplane propellers. However, due to their relative position to other identiable components, they are generally easy to interpret. Isolated small items without any reference to their original context are inevitably much more dicult to identify. Metrical data can be recovered from sonar images, with target height calculations from shadow lengths a standard technique [9] that has been applied to archaeological features [7,8,18]. However, the increased availability of software which allows geo-referencing of side-scan sonar data has led to the ability to produce scalable site images that can form the basis for accurate overall site plans [16,18] although, in many instances they lack quantiable small-scale detail. This approach is widely used in checking the veracity of conventional surveys undertaken in near-zero underwater visibility [18] and for monitoring changes in wreck structures and site development with reference to the natural environment [16]. It has also proved invaluable in making quick assessments of new sites about which there is little quantied information but, even then, it is possible to interpret appropriately-shaped natural phenomena as wreck-sites [19]. The application of such techniques has become increasingly important as computing power and software developments allow quantitative data analysis from a wider range of better-designed instruments. The success of any marine archaeo-geophysical survey is largely dependent upon the experience of the surveyors and interpreters. Although some research has focused on the geophysical signatures of archaeological materials submerged and buried in the marine environment

[3,6,14,15], it remains a poorly understood subject. This often leads to the misinterpretation of acoustic data acquired for both commercial and research-driven archaeological purposes. In near-shore waters misinterpretation of side-scan sonar data is often compounded by the presence of relatively modern man-made objects and marine debris on the seaoor. Furthermore, the resolution of commercially available side-scan systems and the understanding of small-target insonifcation are poorly understood. The majority of recent research on small-target insonication has been conducted by the military interested in identifying mines and unexploded submerged ordnance [24]. In order to address the paucity in knowledge of the reverberation signatures of archaeological material in the marine environment, a control experiment was conducted in Belfast Lough during July and August 2001 by the Archaeological Diving Unit (University of St. Andrews), the Centre for Maritime Archaeology (University of Ulster) and Boland Archaeological Services Ltd. (Fig. 1). A selection of material and target types were arranged on the seabed in varying orientations, in an attempt to replicate the situation found on a range of archaeological sites where material is exposed on the seaoor. Repeat side-scan sonar surveys (using three commercially available side-scan sonar systems) were conducted over the control array and data compared in terms of theoretical- and practical resolutions.

2. Methodology 2.1. Survey and dive equipment 2.1.1. Positional data Positional data were provided by a Trimble DSM-Pro dierential GPS from SV Scimitar. Throughout this survey, diver, ROV and side-scan positioning was provided by an LXT ultra-short baseline (USBL) acoustic tracking system mounted directly on each instrument. The position of the USBL beacon was interrogated by three transducers housed in a unit on the starboard side of the survey vessel, located beneath the DGPS antenna. Positional data were acquired in latitudeelongitude using the WGS-84 reference ellipsoid. The methodology for underwater positioning outlined above negated the requirement for any towsh-antenna layback corrections. 2.1.2. Diver operations The grid was laid out by divers working on surfacedemand from SV Scimatar on a 60%N2:40%O2 gas mix. This Nitrox mix allowed for relatively generous bottom times without decompression stops. Dives were controlled by the diving supervisor, with one diver in the water at a time and another acting as a standby

1254

R. Quinn et al. / Journal of Archaeological Science 32 (2005) 1252e1264

Fig. 1. Location map of the test site located in Smelt Mill Bay on the southern shore of Belfast Lough, Ireland. Bathymetric contours are shown at 2 m intervals below Belfast Datum.

throughout. Diver 1 wore a Pullnix colour camera mounted on a Kirby Morgan Superlight 27B helmet, while Diver 2 wore a Kirby Morgan KMB10 Band mask. Diver 1s umbilical contained a gas hose, multi-core electronics cable, and a pneumofathomometer hose. The electronics cable carried a video signal from a Pullnix video camera on the divers helmet to the video recorder and display unit. The electronic cable also carried the hard-wire voice communication between the diver and the supervisor. All dialogue and imagery were recorded on Mini DV tape and archived. The divers real-time position was monitored using the LXT USBL acoustic tracking system. The position of the diver was logged to screen in the wheelhouse and the co-ordinates were overlain on the video image. Using this methodology, the diver was accurately guided to grid positions. Additionally, a VideoRay ROV was used to check the test array at various times throughout the experiment. 2.1.3. Side-scan instrumentation Three commercially available side-scan sonar systems were used in the control experiment (Table 1). The cost of purchase of System 1 (EdgeTech) was approximately twice that of System 2 (GeoAcoustics), which in turn cost

approximately twice System 3 (Imagenex). Although their operation frequencies varied from 390 to 675 kHz, they all operated in the industry standard high frequency band. Side-scan data were acquired without slant range corrections. It should be noted that although the electro-acoustic specications are set for each commercially available side-scan system (Table 1), some equipment manufacturers will change appropriate settings on request. 2.2. The control array An array of targets were positioned on a at ne-sand substrate (grain size: 0.129e0.164 mm) in an average water depth of 10 m in a sheltered area of Smelt Mill Bay on the southern shore of Belfast Lough, Ireland (Fig. 1). The choice of water depth was important, suciently shallow not to inhibit dive-times, but suciently deep to allow a range of sonar geometry through dierent towsh heights and ranges. The overall morphology of the array was designed to reect the basic dimensions of a wreck, and the shape was a function of having a sucient number of targets suitably spaced for insonication, without upstanding

Table 1 Electro-acoustic parameters of the side-scan systems used in the control experiment Parameter Operating frequency Horizontal beam angle Pulse length Pulse repetition rate No. of pixels in display System 1 390 G 20 kHz 0.5  0.010 ms 75 ms 1024/channel (50 m range) System 2 410 G 1% kHz 0.5  0.088 ms 75 ms 1240/channel (57 m range) System 3 675 kHz 0.7  0.100 ms 109 ms 240/channel (40 m range)

Data are from the manufacturers specications.

R. Quinn et al. / Journal of Archaeological Science 32 (2005) 1252e1264

1255

objects resulting in consequent concealing acoustic shadows (Fig. 2). These requirements resulted in a design of two parallel lines of material, one with predominantly upstanding components (northern baseline), the other with at-lying distributions of material on the seaoor (southern baseline). The 70 m northern baseline was laid eastewest along the prevailing seabed contour, along which were distributed the vertical components of the array (Fig. 2). The origin of the 50 m south baseline was oset 5 m to the south and west of the north baseline. Horizontal distributions of material types and targets were laid along the length of the line. Within the array, ve points were marked with objects expected to provide strong targets to act as checks against the veracity of post-acquisition image geo-rectication (Fig. 2). These ve control points comprised the terminations of the north (N1 and N50) and south (S1 and S26) baselines and a single point 5 m north of the north baseline (NN1).

2.3. Material and target types Material types used in the experiment were chosen on the basis of two parameters e archaeological materials most frequently found on wreck-sites and man-made materials often found associated with inshore wrecksites. Some of the material types used were manufactured for other purposes, but successfully replicated a range of archaeological material found on submerged archaeological sites. This included pottery, glass and wicker, and some modern items unfortunately intrusive on many inshore archaeological sites, such as rubber tyres. Easily-handled aluminium tubing of dierent diameters was used to replicate heavier metal objects such as cannons and metal fastenings. Flint nodules and aggregate were included on the basis that they appear on inundated terrestrial sites or as a result of dispersal of ship ballast. Other targets included arrays of aluminium
10m
10

tubes up to 1 m long and of 0.025, 0.050, 0.100 and 0.200 m diameter, with wall thickness of 0.002 and 0.003 m. The aluminium tubing was primarily included in the array to assess the transverse and range resolutions of the side-scan systems. Diameter of 0.006 m solid steel rods were used as supports for some of the upstanding targets and were also arranged vertically and horizontally. On site-specic historic wreck-sites wood (used in hull construction) is sometimes still abundant. Consequently, a range of oak beams from a late 18th Century wreck were used in the control array, complimented by softwood spars of 0.10 ! 0.10 m and 0.05 ! 0.03 m section, and 0.022 m diameter wooden dowel sections (broom handles). A full coded list of the targets and material types used in the experiment is provided in Table 2. Along the north baseline, from positions N1 to N37, aluminium tubes, iron rods, softwood posts and broom handles were arranged vertically upstanding from a height equalling each individual cross-sectional measurement, grading up to circa 0.70 m in height above the seabed (Table 2; Fig. 2). This maximum was dependent on the extent that each 1 m long object needed to be pushed into the seabed to maintain rmness. Also included in the vertical array was one tube of each diameter angled at 45  to the south and an equivalent to the north. On the south side of the north baseline three tubes were laid horizontally, one perpendicular to the baseline, one parallel but 1 m to the south (touching the south end of the perpendicular) and the one at 45  from the other end of the parallel tube, angled to the north (Table 2; Fig. 2). From the 45 m position on the north baseline, the array consisted of two 0.65 m diameter car tyres, one vertical (N38) and one horizontal (N39); the neck and shoulder of a ceramic amphora (N40); a 0.26-m diameter ceramic sphere (N41); an upturned hanging basket of woven twigs (N42); a woven willow garden ornament in the shape of a conical sh basket (N43); a rectangular woven grass basket (N44); a leather jacket set upright

Northings (m)

5 0 -5

-10 0 10 20 30 40 50 60 70

Eastings (m)
Fig. 2. Layout of the array used in the control experiment with coded targets and material types. Refer to Table 2 for further details.

1256

R. Quinn et al. / Journal of Archaeological Science 32 (2005) 1252e1264

Table 2 ID labels, descriptions, dimensions and osets for the material types and objects on the north and south baselines ID N1 N2 N3 N4 N5 N6 N7 N8 N9 N10 N11 N12 N13 N14 N15 N16 N17 N18 N19 N20 N21 N22 N23 N24 N25 N26 N27 N28 N29 N30 N31 N32 N33 N34 N35 N36 N37 N38 N39 N40 N41 N42 N43 N44 N45 N46 N47 N48 N49 N50 SN1 SN2 SN3 SN4 S1 Type 0.2 m diameter Aluminium tube 0.2 m diameter Aluminium tube 0.2 m diameter Aluminium tube 0.2 m diameter Aluminium tube 0.2 m diameter Aluminium tube 0.2 m diameter Aluminium tube 0.1 m diameter Aluminium tube 0.1 m diameter Aluminium tube 0.1 m diameter Aluminium tube 0.1 m diameter Aluminium tube 0.1 m diameter Aluminium tube Green softwood post Green softwood post 0.05 m diameter Aluminium tube 0.05 m diameter Aluminium tube 0.05 m diameter Aluminium tube 0.05 m diameter Aluminium tube 0.05 m diameter Aluminium tube 0.05 m diameter Aluminium tube Green softwood post Green softwood post 0.025 m diameter Aluminium tube 0.025 m diameter Aluminium tube 0.025 m diameter Aluminium tube 0.025 m diameter Aluminium tube 0.025 m diameter Aluminium tube 0.025 m diameter Aluminium tube 0.025 m diameter Aluminium tube 0.025 m diameter wood broom handle 0.025 m diameter wood broom handle 0.025 m diameter wood broom handle 0.025 m diameter wood broom handle 0.025 m diameter wood broom handle 0.006 m steel rod 0.006 m steel rod 0.006 m steel rod 0.006 m steel rod 0.65 m diameter car tyre 0.65 m diameter car tyre Amphora shoulder and neck Ceramic ball Upturned hemispherical twig basket Willow sh basket Woven grass basket Upstanding leather jacket Green softwood trellis 0.05 m diameter Aluminium radar reector 0.10 m diameter Aluminium radar reector 0.30 m diameter Aluminium foil covered oat 0.20 m diameter Aluminium tube 3 ! 0.20 m diameter Aluminium tubes arranged horizontally: one NeS, one EeW, one SWeNE 3 ! 0.10 m diameter Aluminium tubes arranged horizontally: one NeS, one EeW, one SWeNE 3 ! 0.05 m diameter Aluminium tubes arranged horizontally: one NeS, one EeW, one SWeNE 3 ! 0.025 m diameter Aluminium tubes arranged horizontally: one NeS, one EeW, one SWeNE 0.30 m diameter earthenware ower pots standing end-on-end Dimensions Vertical; H Z 0.64 m 45  to S; H Z 0.42 m 45  to N; H Z 0.48 m Vertical; H Z 0.40 m Vertical; H Z 0.20 m Vertical; H Z 0.78 m 45  to S; H Z 0.40 m Vertical; H Z 0.70 m 45  to N; H Z 0.45 m Vertical; H Z 0.21 m Vertical; H Z 0.08 m 0.1 ! 0.1 m; H Z 0.10 m 0.1 ! 0.1 m; H Z 0.425 m 45  to S; H Z 0.53 m Vertical; H Z 0.55 m 45  to N; H Z 0.52 m Vertical; H Z 0.80 m Vertical; H Z 0.80 m Vertical; H Z 0.53 m 0.05 ! 0.03 m; H Z 0.19 m 0.05 ! 0.03 m; H Z 0.42 m 45  to S; H Z 0.43 m Vertical; H Z 0.568 m 45  to N; H Z 0.41 m Vertical; H Z 0.40 m Vertical; H Z 0.08 m Vertical; H Z 0.06 m Vertical; H Z 0.025 m Vertical; H Z 0.04 m Vertical; H Z 0.07 m Vertical; H Z 0.12 m Vertical; H Z 0.23 m Vertical; H Z 0.21 m Horizontal 45  to S; H Z 0.21 m Vertical; H Z 0.38 m 45  to N; H Z 0.24 m Vertical; H Z 0.60 m Horizontal; H Z 0.20 m H Z 0.35; Diameter at mouth Z 0.18 m Diameter Z 0.26 m Diameter Z 0.17 m H Z 0.38 m; L Z 1.2 m 0.40 ! 0.24 ! 0.16 m H Z 0.70 m; L Z 1.0 m H Z 0.90 m; L Z 1.5 m Vertical; H Z 0.90 m o bed Vertical; H Z 0.88 m o bed Vertical; H Z 0.66 m o bed Vertical; H Z 0.83 m 1.0 m ! 2.0 m 1.0 m ! 2.0 m 1.0 m ! 2.0 m 1.0 m ! 2.0 m H Z 0.60 m X Oset from (0,0) (m) 0 1 2 3 4 10 11 12 13 15 16 18 20 21 22 23 24 25 26 28 29 31 32 33 33.75 34.25 34.75 35.25 36.5 37.0 37.5 38.0 38.5 40 41 42 43 45 46 47 48 49 50.0e51.2 51.4e51.8 52.3e53.4 54.3e55.8 57 58 59 70 0 10 20 30 5 (continued on next page)

R. Quinn et al. / Journal of Archaeological Science 32 (2005) 1252e1264 Table 2 (continued) ID S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13 S14 S15 S16 S17 S18 S19 S20 S21 S22 S23 S24 S25 S26 Type 100% coverage 0.15 m diameter ower pots 50% coverage 0.15 m diameter ceramic pots 100% coverage glazed pot sherds 50% coverage glazed pot sherds 4 ! waterlogged oak ships timbers 100% Flint cobbles No material 100% coverage glass bottles 40% coverage glass bottles 100% coverage pebbles 50% coverage pebbles 100% coverage 10 mm aggregate 50% coverage 10 mm aggregate Articulated human skeletal remains 30% coverage dis-articulated skeletal remains 95% coverage 12.5 mm iron nails 40% coverage 10 mm iron nails Sand mound Shallow pit EeW orientated rack of 0.05 m diameter Aluminium tubes 3 mm Aluminium sheet Eight-limb star shape formed by 0.20 m Aluminium tubes 2 1.0 ! 1.0 m Aluminium angle frames EeW orientated rack of 0.20 m and 0.10 m diameter Aluminium tubes 0.30 m diameter ceramic ower pot Dimensions H Z 0.30 m H Z 0.25 m e e 1.5 m; 1.5 m; 1.5 m; 1 m 0.01e0.06 m; 1.0 m2 e Horizontal Horizontal 1.0 m2 1.0 m2 1.0 m2 1.0 m2 1.0 m2 1.0 m2 1.0 m2 1.0 m2 H Z 0.10 m; Diameter Z 0.227 m D Z 0.15 m; Diameter Z 0.234 m 1.0 m2 1.0 m2 with missing corner 1.3 m2 1.0 m ! 2.0 m 1.0 m ! 1.6 m H Z 0.30 m

1257

X Oset from (0,0) (m) 6 7 8 9 12e13.5 15 17 18 19 20 21 22 23 24 25 26 27 28 29 32 34.5e35.8 38 45.75e48.8 55

Refer to Fig. 1 for the relative positions within the control array.

with outstretched arms, supported internally with broom sticks (N45); a light wooden trellis (N46); and three dierent articial reectors (N47eN49). These comprised 0.05 m and 0.10 m diameter cylindrical-shaped radar reectors, and a plastic spherical shing oat of 0.30 m diameter covered in aluminium foil. Along the 50 m south baseline, a series of alternating patterns of material dened within 1 m squares were set out (Table 2; Fig. 2). The rst grid square (S2) comprised 100% coverage of glazed ceramic pots and dishes averaging 0.15 m diameter and 0.15 m high, succeeded by a square containing the same material but with an approximate 50% coverage (S3). S4 and S5 comprised corresponding grid squares of 100% and 50% coverage of broken glazed pot sherds. Target S6 comprised four 2.40 ! 0.20 ! 0.20 m late 18th Century oak ships timbers, forming a diamond array centred on the baseline. The ships timbers were succeeded to the east by six pairs of metre squares (S7eS18) with 100% and 50% coverage of material (Table 2; Fig. 2). These squares were lled respectively with 0.01e0.06 m int nodules; glass wine bottles; 0.03e0.04 m pebbles; 0.01 m aggregate; skeletal remains (long bones, pelvic girdle and vertebra of a human skeleton in a crouched inhumation arrangement, with an adjacent low-density random bone distribution) and 0.10 m wire nails. S19 comprised an excavated hole circa 0.50 m in diameter

and 0.15 m deep, with the spoil heaped into an adjacent mound to a height of 0.10 m (S20). S21 consisted of 1 m2 of 0.05 m diameter tubes racked parallel to the baseline whilst S22 comprised a 1-m2 aluminium sheet with one corner missing. S23 comprised eight 0.20 m diameter tubes arranged in a star pattern. S24 consisted of two 1 m2 aluminium angle frames placed either side of the south baseline. The two nal arrays were thirteen 0.10 m diameter aluminium tubes laid parallel and arranged in a rectangle with a 0.20-m tube at the outer north and south ends (S25) and a 0.30-m diameter ceramic ower pot marking the western end of the south baseline (Table 2; Fig. 2). 2.4. Sonar geometry A total of 12 parallel survey lines (BLEW-1 to BLEW-12) were employed in the experiment (Fig. 3), spaced at 5 m, 10 m, 20 m, 40 m and 80 m north and south of the baselines on the seabed (Fig. 3). Repeat side-scan surveys were conducted at sh altitudes of 2 m, 4 m and 8 m above the seaoor. A ray-path diagram of the geometry of each pass shows that the grazing angles varied between a maximum of 47  and minimum of 11  at the centre of the north and south baselines (Fig. 4). Although the survey was conducted in calm conditions with no appreciable natural wave action, the site

1258
100

R. Quinn et al. / Journal of Archaeological Science 32 (2005) 1252e1264

BLEW-1

50
BLEW-2

Northings (m)

BLEW-3 BLEW-4 BLEW-5 BLEW-6 BLEW-7 BLEW-8 BLEW-9 BLEW-10

-50

BLEW-11

50m

BLEW-12

-100 -100 -50

50

100

150

Eastings (m)
Fig. 3. Plan view of survey lines BLEW-1 to BLEW-12 superimposed on the control array.

was subject to noticeable, intermittent swell produced by fast ferries which passed within 1.5 km of the control site as they slowed down to enter Belfast ferry terminal. The eect was noticeable to those on the boat and also to divers on the seabed, who lmed the eect this movement had on the large diameter tubes. Despite pinning all objects to the seabed, some of the horizontal 0.20 m diameter tubes moved from their set position during the survey period. For example, only two tubes from the star array (S23) remained in-situ at the end of the sonar passes. This resulted in the re-distribution of a number of 0.20 m diameter tubes through and beyond the control site, although the loose tubes migrated in
10m
0m

a general NNE direction. To maintain consistency in data interpretation, the site was surveyed and recorded preceding each sonar run using the VideoRay ROV.

3. Theoretical resolution considerations In general terms, side-scan data can be described as the convolution of the transmitted sonar pulse with the scattering function of the seaoor [23]. In practice, this is an over-simplication of sonar imagery as the nal display data are not only a function of the convolution process, but is a graphic representation of interactions

10m BLEW-11

BLEW-10

BLEW-9 BLEW-8 BLEW-7 BLEW-6 South Baseline North Baseline

Fig. 4. Ray-path diagram of the sonar geometry used in the control experiment. The maximum grazing angle of 47  (b1) and minimum of 11  (b2). Grazing angles are calculated for the central line of the array (midpoint of BLEW-6 and BLEW-7).

R. Quinn et al. / Journal of Archaeological Science 32 (2005) 1252e1264

1259

of the outgoing pulse with the water column and the seaoor and the transmission, processing and display mechanisms of the system. The resolution of the nal side-scan image is therefore a function of many factors including pulse specications, beam spreading, acoustic velocity, seaoor morphology and post-acquisition processes that can at times decimate original data quality. It is important to understand the resolving capabilities of a side-scan system (towsh and topside) in order to calculate expected theoretical image resolution at various positions within a sonograph. The distinction between resolution in the x-axis of the plot and resolution in the y-axis is variably dened. Generally, the cross-range (x-axis) footprint is described by the beam angle of the side-scan system and the down-range ( y-axis) footprint is described by the projection of the band-limited pulse on the seaoor. Transverse resolution (Drt ) is the minimum distance between two objects parallel to the line of travel that are displayed on the sonar as separate objects. This is equivalent to the intersection of the horizontal beam angle (which broadens with increasing range) with any point on the seabed. Transverse resolution is calculated using the equation: Drt Z sinqha R; 1

the y-axis. However, Eq. (3) is the only equation that accounts for the eect of grazing angle, a very important facet in the resolving capabilities of a side-scan sonar. Although the quality of sonar imagery is therefore dependent upon the electro-acoustic parameters of the transceivers (specically the beam angle and eective pulse length), the actual recorded data is a rectangular grid of pixels, each representing a sample of the georeferenced convolved eld. Each pixel occupies the same size on the imagery, and the pixel size denes the actual range resolution of the displayed imagery by: R Drr Z ; N 4

where R is range and N is the number of pixels in the display [11]. It should also be noted that the range resolution of acquired data is also heavily dependent upon the survey speed and pulse rate of the sonar system.

4. Results and interpretation 4.1. Theoretical resolutions of the side-scan suites used in this experiment For the control experiment, an average VPW is calculated as 1500 ms1 using the Wong and Zhu method [26], where water temperature is 13  C, salinity is 35 parts per thousand and average water depth is 10 m. Theoretical range and transverse resolutions for the three systems operating in industry standard 500 kHz mode are calculated using Eqs. (1) and (3), respectively, and the electro-acoustic parameters outlined in Table 1. The results of these calculations are outlined in the plot of range against resolution in Fig. 5. This plot indicates that the theoretical resolutions of the three systems decrease with increasing range. Furthermore the transverse and range resolutions of System 1 are highest and System 3 lowest, whilst the transverse resolution of Systems 1 and 2 are equal. Regrettably, it is often assumed amongst sonar users that the operating frequency of the instrument is the dominant factor that dictates the nal resolution of the acquired data. That is, a general rule of thumb is accepted that the higher the operating frequency, the higher the resolution of the data. However, the relationships presented range and resolution in Fig. 5 indicate that the dominant controls on the theoretical resolution of the data are the electro-acoustic specications of the pulse length and the beam angle of the transmitted pulse, not the actual operating frequency. Table 3 lists the lower limits of range resolution for each side-scan system as dened by Eq. (4). By this criterion, System 2 oers the optimum display

where qha is the horizontal beam angle and R is the range of the target [11]. As target range increases the minimum resolvable distance also increases i.e. resolution decreases. Range resolution Drr is the minimum distance between two objects perpendicular to the line of travel that are displayed as separate objects. Although this is a function of both the display system and seaoor topography, the pulse length of the sonar is the overriding characteristic that determines the lower theoretical limits of resolution. Range resolution is expressed by researchers in a number of ways, although commonly as: Drr Z Vpw te ; 2 2

where te is the eective pulse length and Vpw is the compressional wave velocity of the acoustic pulse in the water column [23] and as Drr Z Vpw te sec b; 2 3

where b is the grazing angle [13]. Eq. (3) assumes some knowledge of seaoor topography, and therefore problems exist in the use of this equation for the majority of side-scan applications as seaoor topography is not always clearly dened or known. Of the above equations, Eq. (2) is the most commonly used to dene the theoretical resolution of the side-scan footprint in

1260
0.35

R. Quinn et al. / Journal of Archaeological Science 32 (2005) 1252e1264

tr system 1

0.30

tr system 3 tr system 2 rr system 1 rr system 3 rr system 2

0.25

Resolution (m)

0.20

0.15
System 3: minimum rr

0.10

0.05

System 1: minimum rr System 2: minimum rr

0.00 0 5 10 15 20 25

Range (m)
Fig. 5. Transverse resolution (rt, Eq. 1) and range resolution (rr, Eq. 3) plots for the three side-scan systems used in the control experiment. The range resolution calculated for the three side-scan systems using Eq. 4 are also displayed. Refer to Table 3 for further details.

resolution, whilst System 3 oers the worst even though System 3 operated at the highest frequency (Table 1).

4.2. Results of the control experiment The results of the side-scan test runs are summarised in the three sonographs in Fig. 6, where a single pass along line BLEW-9 is presented for each of the three systems. These data were acquired at a range of 10 m from the southern baseline. Each sonograph is corrected for Time Varied Gain (TVG). No other lters have been applied. The data quality and resolution characteristics of these data are in agreement with the predictions outlined by the plots of theoretical resolution (Fig. 5), in that the data acquired using System 1 (Fig. 6a) is superior to that of System 2 (Fig. 6b), which is in turn better than System 3 (Fig. 6c). As all of these sonographs were taken under identical conditions, the resultant image quality is interpreted as being primarily dependent upon the beam angle and pulse length of the side-scan systems. Again, this interpretation is probably an oversimplication of the factors controlling data quality. As noted previously, the digitisation process of the incoming analogue signal can aect the resolution of stored and displayed side-scan data. Several manufacTable 3 Range resolutions calculated for the three side-scan systems using Eq. (4) and the data outlined in Table 1 Parameter Range resolution System 1 0.049 m System 2 0.046 m System 3 0.125 m

turers now digitise the analogue signal at a xed rate that can lead to over-sampling in the near ranges and under-sampling in the far ranges (John Fish, pers comm.). Furthermore, the quality of the nal image is a direct function of the number of pixels in the display and the pulse repetition rate. The number of pixels across the screen automatically set a lower limit to the range resolution, whilst the pulse rate sets an automatic lower limit to the range-dependent transverse resolution. With one exception, all of the inorganic and organic material and target types used in the control experiment were imaged with each side-scan system during the control experiment. The exception is the 1 m2 grids of skeletal remains, which were only imaged using System 1 at a range of 5 m from the southern baseline. The fact that the skeletal remains were not imaged in the majority of passes may be a reection of their irregular outline. Experience has proven that it is easier to identify and interpret regularly shaped targets and reectors, specically if they have straight edges. Fig. 7 presents two expanded sections of data from Fig. 6a, displaying targets S6eS18 along the south baseline and N38eN45 on the north baseline (refer to Table 2 for more details). These two sections contain the majority of organic and inorganic material types and targets contained within the array. In Fig. 7a, N38 and N39 are the vertical and horizontal car tyres, respectively, with the shadow from the upstanding car tyre dominating the backscatter response of N38. The problem in image interpretation is highlighted by the responses from targets N40eN42, where each of the targets is imaged as individual high backscatter

R. Quinn et al. / Journal of Archaeological Science 32 (2005) 1252e1264

1261

(a)

(b)

(c)

10

Northings (m)

5 0
-5 -10 0 10 20 30 40 50 60 70

Eastings (m)
Fig. 6. (a) System 1 sonograph of the test site; (b) System 2 sonograph of the test site; (c) System 3 sonograph of the test site. All sonographs are reproduced from line BLEW-9 at a sh height of 4 m above the seaoor. Refer to Fig. 2 for target and material ID numbers.

features with associated acoustic shadows. Although the three targets comprise three dierent material types (pottery, ceramic and twig), and are clearly imaged as three separate targets, they are impossible to dierentiate by eye. This is an important factor in image

interpretation, when objects are imaged out of context, they are almost impossible to interpret accurately. Target S43 represents a slightly dierent scenario, in that the conical sh basket is easier to identify as such, as the shape of the overall target is reected in the

1262

R. Quinn et al. / Journal of Archaeological Science 32 (2005) 1252e1264

(a)

(b)

Fig. 7. Expanded sections of data from Fig. 6a displaying targets (a) N38eN45 along the northern baseline and (b) S6eS18 along the southern baseline.

backscatter response and the acoustic shadow is concentrated in the area where the basket is heavily woven with twigs. However, if imaged in isolation, and without prior knowledge, it would be almost impossible to interpret the feature as a sh basket. Target N44, a woven grass basket, is similar in both shape and backscatter response to targets N40eN42. The upstanding leather jacket (N45) is represented by a high backscatter response with associated acoustic shadow. The backscatter characteristics of leather material are important as it has similar characteristics to hide which was regularly used as a hull material in boat building in the past. Additionally, more detail regarding the morphology of the target can be gleaned from the acoustic shadow than from the backscatter response of the primary target. The wooden trellis represented by N46 is again imaged as a high backscatter signature in the sonar data, although it lacks any detail. This is interpreted as being due to the fact that the individual trellis members (0.02 m) are of proportions smaller than the resolution of the side-scan system and so lateral smearing occurs. In Fig. 7b, the four late 18th Century oak ships timbers are clearly imaged in the sonar data as high backscatter linear features with associated acoustic shadows. Importantly for maritime archaeology, the backscatter responses of the ships timbers and the green wood sections used in the experiment are high, reinforcing the interpretation of sonar data acquired from a number of historic wreck-sites [15,16,18]. S7eS21 predominantly represent 1 m2 cells of full and partial coverage of various material types ranging from glass bottles to int nodules (refer to Table 2 for more details). All of the material types used in this section of the array were imaged. Target S22, a at aluminium sheet, represents a unique target in the

control array, in that there is no backscatter response from the material. Instead, the sheet is represented by a white square (with missing corner) as pure reection has occurred on the plate, and no acoustic energy has been scattered back to the towsh. Although aluminium sheets are not normally found on archaeological sites, this response indicates that highly polished or at surfaces associated with more modern wreck-sites may not be imaged successfully in side-scan surveys. However, the total absence of any backscatter response may therefore indicate the presence of such materials. Finally, S23 represents the remains of the eight-armed star formation from the end of the experiment, with only two 0.20 m diameter aluminium tubes remaining. One dominant consideration throughout the exercise was how to interpret backscatter data from targets which remain out of context. This scenario is common in midto high-energy wreck-sites, where material is scattered by natural processes over time. One approach to address this is an attempt to characterise the backscatter response of the material type through reverberation characteristics of the material as dened by its tonal response. Fig. 8 displays the backscatter response from a variety of material types and concentrations. For example, Fig. 8a displays the response for 100% (S9) and 40% (S10) glass coverage as simple grey level histograms, quantifying the global distribution of grey levels within each 1 m2 grid. Statistical information extracted from the histograms indicates the 100% glass coverage is characterised by a mean grey level of 4.86 with a standard deviation of 9.25 and the 40% coverage is characterised by a mean level of 74.74 with an associated standard deviation of 36.62. The examples in Fig. 8 represent a basic rst order approach towards texture quantication of material types, and

R. Quinn et al. / Journal of Archaeological Science 32 (2005) 1252e1264

1263

Fig. 8. (a) Backscatter response from 100% glass coverage (mean Z 4.86, 0e255 greyscale); (b) backscatter response from 40% glass coverage (mean Z 36.69, 0e255 greyscale); (c) backscatter response from 100% pebble coverage (mean Z 7.37, greyscale Z 0e255); (d) backscatter response from 50% pebble coverage (mean Z 51.38, greyscale Z 0e255); (e) backscatter response from 95% 12.5 mm iron nail coverage (mean Z 2.67, greyscale Z 0e255) and (f) backscatter response from 40% 10 mm iron nail coverage (mean Z 38.52, greyscale Z 0e255).

demonstrate how material types and coverage density can be dierentiated in the post-processing stage. A full discussion of this technique is beyond the scope of this paper. However, it may form the focus of future research.

5. Discussion and conclusions The results from this control experiment indicate a very wide range of submerged organic and inorganic materials can be imaged using side-scan sonar. However, individual targets may be dicult (or impossible) to resolve when they are out of context. It is apparent that many of the materials deployed could not be interpreted on the basis of their backscatter response alone. The problem in image interpretation is highlighted by the responses from targets N40eN42 (pottery, ceramic and twig basket), where each of the targets is imaged as individual high backscatter features with associated acoustic shadows. Although the three targets comprise three dierent materials, and are clearly imaged as three separate targets, they are impossible to dierentiate by eye. This is an important factor in image interpretation, when objects are imaged out of context, they are almost impossible to interpret accurately. Furthermore, it is clear that acoustic shadows can be misleading when interpreting the nature of individual artefacts (e.g. the woven sh basket). One square metre grids of skeletal remains were only imaged using System 1 at a range of 5 m from the southern baseline. The fact that the skeletal remains

were not imaged in the majority of passes may be a reection of their irregular outline. The results of this experiment imply that it is easier to identify and interpret regularly shaped targets and reectors, specifically if they have straight edges. From the experimental results it is clear that the best imagery for all systems was acquired at a range of 10e 15 m from the array. Also, it is clear that the eective range of the high-resolution sources for marine archaeo-geophysical work is circa 40 m. This has important implications for both reconnaissance and site-specic surveys. For reconnaissance scale surveys, if the object of the exercise is to acquire data for the location of submerged archaeological sites, then the most eective swath width should be set at a maximum of 80 m. For site-specic research, lane spacing should be set at 5e10 m in order to acquire eective data sets. Fish altitude above the seaoor also has a profound eect on data quality and interpretability. From the results of the control experiment and previous survey experience, it is clear that eective sh altitude is highly site-specic. What may work for one site will not work on another. The lower the sh altitude, the more profound the acoustic shadows, and therefore the easier it is to detect upstanding objects. However, this may cause serious masking eects and cause the surveyor/ interpreter to actually overlook or completely fail to image a certain section of the seaoor and/or wreckage. One complex aspect of side-scan sonar data is evident from the control experiment, that is, there is an obvious dierence in resolution categories of side-scan sonar data,

1264

R. Quinn et al. / Journal of Archaeological Science 32 (2005) 1252e1264 [7] R.W. Duck, J. McManus, A side-scan sonar survey of a previously drawn-down reservoir: a control experiment, The International Journal of Remote Sensing 6 (1985) 601e609. [8] R.W. Duck, W.M. Dow, Side-scan sonar reveals submerged remains of the rst Tay Railway Bridge, International Journal of Geoarchaeology 9 (1994) 139e153. [9] J.P. Fish, H.A. Carr, Sound Underwater Images: a Guide to the Generation and Interpretation of Side Scan Sonar Data, Lower Cape Publishing Co, Boston, MA, 1990. [10] J.P. Fish, H.A. Carr, Sound Reections: Advanced Applications of Side Scan Sonar, Lower Cape Publishing Co, Boston, MA, 2000. [11] B. Flemming, Side-scan sonar: a practical guide, International Hydrographic Review 53 (1976) 65e91. [12] C.H. Hobbs, D.B. Blanton, R.A. Gammisch, J. Broadwater, A marine archaeological reconnaissance survey using side-scan sonar, Jamestown, Virginia, USA, Journal of Coastal Research 10 (1994) 351e359. [13] E.J.W. Jones, Marine Geophysics, John Wiley and Sons Ltd., Chichester, UK, 1999. [14] P. Newell, Ground-truthing the eective resolution of high frequency side-scan sonar data, unpublished MSc Thesis, University of Ulster, Coleraine, 1999. [15] R. Quinn, J.M. Bull, J.K. Dix, Imaging wooden artefacts using Chirp sources, Archaeological Prospection 4 (1997) 25e35. [16] R. Quinn, J.R. Adams, J.K. Dix, J.M. Bull, The Invincible site e an integrated geophysical assessment, International Journal of Nautical Archaeology 27 (1998) (1758) 126e138. [17] R. Quinn, J.A.G. Cooper, B. Williams, Marine geophysical investigation of the inshore coastal waters of Northern Ireland, The International Journal of Nautical Archaeology 29 (2000) 294e298. [18] R. Quinn, C. Breen, W. Forsythe, K. Barton, S. Rooney, D. OHara, Integrated geophysical surveys of The French Frigate La Surveillante (1797), Bantry Bay, Co. Cork, Ireland, Journal of Archaeological Science 29 (2002) 413e422. [19] R. Quinn, W. Forsythe, C. Breen, M. Dean, M. Lawrence, S. Liscoe, Comparison of the maritime sites and monuments record with side-scan sonar and diver surveys: a case study from Rathlin Island, Ireland, International Journal of Geoarchaeology 17 (2002) 441e451. [20] M. Redknap, G. Fleming, The Goodwins Archaeological Survey: towards a regional marine site register in Britain, World Archaeology 16 (1985) 312e328. [21] M. Redknap, Surveying for underwater archaeological sites: signs in the sands, The Hydrographic Journal 58 (1990) 11e16. [22] H.D. Smith, A.D. Couper, The management of the underwater cultural heritage, Journal of Cultural Heritage 4 (2003) 25e33. [23] M.L. Somers, Resolving the issue, a look at resolution and related topics in sonar, Manmade Objects on the Seaoor, Society for Underwater Technology (1995) 51e58. [24] D.C. Summey, J.D. Lathrop, J.F. McCormick, P.J. Carroll, Locating UXO using the mobile underwater debris survey system, Sea Technology 43 (2001) 33e42. [25] A.J. Wheeler, Environmental controls on Shipwreck preservation: the Irish context, Journal of Archaeological Science 29 (2002) 1149e1159. [26] G.S.K. Wong, S. Zhu, Speed of sound in seawater as a function of salinity, temperature and pressure, The Journal of the Acoustical Society of America 97 (1995) 1732e1736.

namely theoretical-, display- and storage resolution. The dominant factors controlling the resolution types are the electro-acoustic specications of the transmitted pulse and the digitisation of the backscattered analogue signal. These conclusions are clearly at odds with the generally accepted rule that the operating frequency is the dominant control on data resolution and hence target identication. Although the responsibility for system design and specications lies with equipment manufacturers, it is nevertheless imperative that the users understand the distinctions between these parameters. Furthermore there is clearly a distinction between resolving and/or imaging an object or material and identifying the material type, especially on archaeological sites where the range of material types within close proximity can be decidedly varied. Acknowledgements The authors would like to thank David Burden, skipper of SV Scimitar; the Crown Estate for permission to conduct the experiment; Wes Forsythe, Declan Hurl, Peter Moore, Sian Keith, Robert White and David Kilner for their assistance in the experiment. We are very grateful to John Fish for feedback and discussion. This research was supported by the Environment and Heritage Service (Department of the Environment for Northern Ireland), the Advisory Committee on Historic Wreck Sites and the Department for Culture, Media and Sport (UK). Reviews by two anonymous reviewers improved an earlier version of this manuscript. References
[1] J.B. Arnold, R.S. Weddle, The Nautical Archaeology of Padre Island, Academic Press, New York, 1978. [2] C. Beltrame, D. Gaddi, Report on the rst research campaign on the Napoleonic brick, Mercure, wrecked o Lignano, Udine, Italy in 1812, The International Journal of Nautical Archaeology 31 (2002) 60e73. [3] J.M. Bull, R. Quinn, J.K. Dix, Reection coecient calculation from marine high resolution seismic reection (Chirp) data and application to an archaeological case study, Marine Geophysical Researches 20 (1998) 1e11. [4] G.F. Caston, Wreck marks: indicators of net sand transport, Marine Geology 33 (1979) 193e204. [5] M. Dean, A review of marine archaeology, Marine archaeology in the Irish Sea, Irish Sea Forum Seminar Report No. 21, University of Liverpool, 1999, 32e40. [6] J.K. Dix, S. Arnott, I.A. Best, D. Gregory, The acoustic characteristics of archaeological wood, Underwater Acoustics Group Conference on Acoustical Oceanography, 9e12 April, Southampton, UK, 2001.

You might also like