Biological Conservation 167 (2013) 334338

Contents lists available at ScienceDirect

Biological Conservation
journal homepage: www.elsevier.com/locate/biocon

Comparing eld-based monitoring and remote-sensing, using

deforestation from logging at Important Bird Areas as a case study
Graeme M. Buchanan a,, Lincoln D.C. Fishpool b, Michael I. Evans b, Stuart H.M. Butchart b
a
b

Royal Society for the Protection of Birds, 2 Lochside View, Edinburgh Park, Edinburgh EH12 9DH, UK
BirdLife International, Wellbrook Court, Cambridge CB3 0NA, UK

a r t i c l e

i n f o

Article history:
Received 7 June 2013
Received in revised form 2 August 2013
Accepted 22 August 2013

Keywords:
Biodiversity
Logging
Forest loss
Land cover change
Monitoring
Site based conservation

a b s t r a c t
Monitoring sites of biodiversity conservation importance is essential for their conservation. It allows
threats to be identied and quantied, priorities to be set, responses to be developed, and facilitates
adaptive management. Field-based monitoring protocols need to be simple enough to be widely applied
in countries with limited capacity while being sufciently robust to provide widely reliable data. A simple, globally standardised monitoring protocol is now being implemented at thousands of sites of global
avian conservation signicance (Important Bird Areas, IBAs) worldwide, but the consistency of the
approach across sites, countries and regions remains untested. We tested the match between estimates
of the threat to IBAs from logging derived from such monitoring, with standardised deforestation rates
derived from remote sensing data for 20002005 to determine if the two were consistently related.
We found a signicant positive correlation between the impact of the threat from logging and the proportion of forest lost (although the gross forest loss did not differ systematically with the two components of the threat impact: scope and severity). The results give us some condence that the simple
eld-based protocol being implemented by a diversity of surveyors with varied technical capacity can
generate meaningful and consistent monitoring data across the globe.
2013 Elsevier Ltd. All rights reserved.

1. Introduction
Monitoring is essential to identify threats to biodiversity, to
develop appropriate responses to threats, and to assess the
effectivenessand facilitate adaptive managementof ensuing
actions (Jones et al., 2013). For sites of conservation signicance,
such as protected areas, monitoring should ideally evaluate the
state (condition) of conservation targets (populations, species,
habitats, etc.), the pressures (threats) affecting them, and the conservation actions (responses) in place. While remote sensing can be
useful for monitoring trends in land cover extent, at least for forest
(e.g. Achard et al., 2002; Hansen et al., 2010), it is less useful for
evaluating trends in habitat condition and population abundance.
Similarly, while it may be useful for monitoring some threats
(e.g. deforestation for oil palm plantations (e.g. Buchanan et al.,
2008), it cannot directly measure hunting pressures. Hence, remote
sensing often requires complementary eld-based monitoring to
provide adequate data for informed and comprehensive decisionmaking and adaptive management. However, capacity to
undertake such monitoring is limited, particularly in the poorest
Corresponding author. Tel.: +44 (0)131 317 4154.
E-mail addresses: graeme.buchanan@rspb.org.uk (G.M. Buchanan), lincoln.
shpool@birdlife.org (L.D.C. Fishpool), mike.evans@birdlife.org (M.I. Evans), stuart.
butchart@birdlife.org (S.H.M. Butchart).
0006-3207/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.biocon.2013.08.031

countries which often have the richest biodiversity and most

intense threats.
To address this, BirdLife International has developed a simple,
globally standardised protocol for monitoring Important Bird
Areas, or IBAs (BirdLife International, 2006). IBAs are sites of global
signicance for bird conservation, identied nationally through
multi-stakeholder processes using globally standardised criteria
for populations of threatened, restricted-range, biome-restricted
or congregatory species (BirdLife International, 2012). Over
12,000 have been identied worldwide (BirdLife International,
2013), and IBA monitoring is underway at over 25% of these. The
IBA monitoring protocol focuses on evaluating the state (condition
of), pressures (threats to), and response (management, mitigation,
etc.) at IBAs, each of which is scored on a simple 4-point scale (e.g.
Mwangi et al., 2010). It is typically implemented by staff of BirdLife
partner organisations, or members of local community conservation groups developed and supported by these partners, other
volunteers, or protected-area staff. These people are given basic
training, but many have relatively low technical capacity.
The consistency of the approach across sites and countries has
never been tested, partially due to the absence of an appropriate,
globally referenced dataset that could relate to aspects of the eld
assessments. Hansen et al. (2010) produced a coarse resolution
map of gross forest loss between 2000 and 2005 from remote
sensing data at a spatial resolution of 18.5 km by 18.5 km cells.

G.M. Buchanan et al. / Biological Conservation 167 (2013) 334338

To assess comparability between the two approaches, we matched

threat impact scores derived from eld monitoring for a sample of
IBAs with deforestation estimates based upon Hansen et al. (2010)
for these sites.
2. Methods
2.1. Field based monitoring
Data derived from eld-based monitoring assessments at IBAs
were extracted from BirdLifes World Bird Database (WBDB). Following a standardised protocol (BirdLife International, 2006),
scores for state (condition), pressure (threats) and responses
(action) are determined for each site. To evaluate pressure, the
principal threats affecting those species for which the site was
identied as an IBA (i.e. the trigger species) are classied using
the IUCN-CMP1 Threats Classication Scheme (Salafsky et al.,
2008; http://www.iucnredlist.org/technical-documents/classication-schemes/threats-classication-scheme). Each threat is then
scored for its impact, derived from separate assessments of timing,
scope, and severity of each threat type. Scope is dened as the proportion of the populationor the proportion of the key habitat for
each trigger speciesaffected, with scores assigned as 0 (<10%), 1
(1049%), 2 (5090%) or 3 (>90%). Severity is dened as the maximum rate of population decline of any trigger species caused by
the threat within its scope, with scores assigned as 0 (<1% over
10 years or three generations of the trigger species, whichever is
longer), 1 (110% over the same period), 2 (1130% over 10 years)
and 3 (>30% over 10 years). Timing is scored on the basis of whether
each threat is past and no longer limiting (score 0), likely in the long
term i.e. >4 years (1), likely in the short term or <4 years (2) or happening now (3). The overall threat impact is then calculated from the
sum of these scores. This resultant 09 scale can be converted back
to a four-point scale of 03 (with 3 indicating the highest level of
threat). The threat or threats with the highest score represent the
overall threat impact score for the site, following a weakest link approach (for further details see BirdLife International, 2006; Mwangi
et al., 2010).
These scores cover a wide range of threats (Salafsky et al., 2008),
but we were interested solely in the IBAs that were identied for
forest species and where logging was identied as a threat happening now. We identied IBAs identied for species estimated to
have high or medium dependency on forests (Buchanan et.al.
2011). Species with high forest-dependence are forest specialists
that are characteristic of the interior of undisturbed forest, rarely
occupy non-forest habitats, and almost invariably breed within forest; while they may persist in secondary forest and forest patches if
their particular ecological requirements are met, they are usually
less common in such situations. Species with medium forestdependence are forest generalists that breed in undisturbed forest
but are also regularly found in forest strips, edges and gaps and
secondary forest, where they may be commoner than in the interior of intact forest. There were 96 IBAs that had been monitored
between 2000 and 2005 and that had a threat of logging to forest-dependent species documented as happening now. The mean
size of these IBAs was 297.4 km2 (median = 381.3 km2) and 45
were smaller than 342 km2 (the area of a cell used by Hansen
et al. (2010)).
2.2. Remote sensing data
While a range of land-cover maps from multiple time periods
are available, only Hansen et al. (2010) has been produced in such
1

Conservation Measures Partnership.

335

a way that enables appropriate comparisons of land-cover change

over time in a way that is comparable worldwide. These authors
estimated forest cover and gross forest loss between 2000 and
2005 in 18.5 km by 18.5 km cells across most of the globe (many
oceanic islands excepted) using MODIS satellite image data at
500 m resolution, calibrated by analysis of higher resolution Landsat data. Digital boundaries were obtained for 10,253 IBAs (BirdLife
International, 2011), representing c. 95% of globally signicant IBAs
identied by 2011. Boundaries for many IBAs in USA, and for small
numbers in other countries, were unavailable, while IBAs had not
yet been identied in New Guinea and several other island Pacic
states at the time of the analysis. The forest loss and IBA maps were
projected using Berhmanns equal-area projection. IBAs smaller
than 1 km2 were scaled up to 1 km2, based on the Berhmann
projection. While some 7895 of these IBAs overlapped with the
deforestation map, only 4491 were designated for forest species.
The median size of these selected IBAs was 223 km2 (range
173,511 km2) while 2669 IBAs (59%) had an area less than
342 km2 (the cell size used by Hansen et al. (2010). Mean and maximum proportional forest loss within each IBA were determined by
overlaying. Digital IBA boundaries were overlaid onto the Hansen
et al. (2010) deforestation map in ArcMap 10, and the percentage
forest loss was calculated for each 18.5 km square cell used by
Hansen et al. (loc. cit.) that partially or wholly overlapped each
IBA. We then determined the mean percentage forest loss for all
cells overlapping each IBA, and the maximum percentage loss for
any cell overlapping the IBA. Due to the mismatch in size of IBAs
and spatial unit of forest loss, these estimates are best described
as loss in the location of IBAs. Additionally, mean forest loss in
the cells immediately surrounding, but not overlapping each IBA
(i.e. in the neighbouring locations) was calculated.
2.3. Analysis
To compare deforestation estimates from Hansen et al. (2010)
with the impact of logging as assessed on the ground, ranks were
calculated for each IBA based on (a) threat impact score (severity
score plus scope score) from eld monitoring, and (b) mean and
(c) maximum forest loss (from remote sensing). We compared
these ranks using a Spearman rank correlation. ANOVAs using generalised linear models (in Proc Genmod in SAS 9.2) were used to
test for systematic differences between IBAs in threat scores for
scope, severity and impact. Mean and maximum forest losses in
IBA locations was the dependant variable, and logging scope,
severity and impact scores were entered as xed effects. Log transformation was used to normalise the forest loss data for these
GLMs. The signicance of relationships was assessed by comparing
the explained deviance with the v2 distribution using the appropriate degrees of freedom (e.g. Crawley, 1993). Residuals were
examined to identify IBAs where loss fell outside the 95% CI around
the mean for focal scores. Forest loss in the location of IBAs was
compared to loss in cells in the neighbouring locations using a sign
test. All means presented SE.
3. Results
There was a signicant positive correlation between the threat
impact score for IBAs and both mean and maximum forest loss
(rho = 0.227, P = 0.026 and rho = 0.273, P = 0.007 respectively;
Fig. 1) in the location of IBAs for the 96 IBAs for which these data
were available. The signicant correlations across all IBAs indicates
that there was a broad agreement between remote sensing and
eld monitoring estimates of the intensity of pressures from forest
loss. However, neither the mean nor maximum rate of loss differed
systematically with threat impact score in an ANOVA (v26 6:07,

336

G.M. Buchanan et al. / Biological Conservation 167 (2013) 334338

there were signicantly more IBAs for which forest loss was higher
in areas in the neighbouring locations than in the locations themselves (sign test, P < 0.001).
The areas with the highest forest losses in the location of IBAs
were scattered across the globe, but there were notable concentrations in Central America, Amazonia, southern Africa, South-East
Asia, Borneo and Sumatra (Indonesia/Malaysia), Australia, and
parts of boreal Canada and Russia (Fig. 3 and Table A2).

4. Discussion

Fig. 1. Mean (dark grey) and maximum (light grey) forest loss during 20002005
for IBAs with different scores for the impact of logging derived from eld-based
monitoring.

P = 04 and v26 11:7, P = 0.07 respectively). Splitting the impact

score into its component parts indicated that mean forest loss in
the location of IBAs did not differ systematically with scores for
either scope or severity of logging (v23 4:55, P = 0.10 and
v23 3:95, P = 0.30 respectively.) Similarly, maximum forest loss
did not differ systematically from scores for severity of logging
(v23 5:25, P = 0.15), but there was a marginally signicant
relationship with the score for the scope of logging (v23 7:93,
P = 0.047; Fig. 2). Post hoc comparisons indicated the differences
were in the direction expected, with forest loss being signicantly
lower in the location of IBAs where the score for the scope of
logging was 0 than in those where the score was 1 or 2
(z = 2.27 and 2.07 respectively). No other comparisons revealed
signicant differences (z 6 0.63). The number of IBAs with scope
and severity scores of 0, 1, 2 and 3 was, respectively, 17, 62, 16,
1 and 6, 68, 20 and 2. None of these comparisons were signicant
when only IBAs with high forest dependency triggers were
analysed (Table A1).
Mean forest loss from 2000 to 2005 in the location of IBAs (i.e.
in cells in Hansen et al. that overlapped each IBA) was
0.943% 0.020, but with considerable variation between IBAs
(range 017.9%; Table A2). Mean forest loss in areas surrounding
IBAs (i.e. in cells in Hansen et al. (2010) immediately surrounding,
but not overlapping, each IBA) was 0.976% 0.020, which is, on
average, 0.044% 0.012 higher than the loss in the location of IBAs.
The number of IBAs for which loss was greater in the locations
neighbouring the IBAs than in the locations of IBA themselves
was 2181 IBAs (48.6%). There were 1704 IBAs (37.9%) where the
reverse was true. Loss was identical for 606 IBAs (13.4%). Hence,

Fig. 2. Mean (open bars) and maximum (lled bars) forest loss (1 SE) during
20002005 for IBAs with different scores for the scope of logging derived from eldbased monitoring during the same period. The single IBA with a scope score of 3 is
excluded.

The signicant positive correlation between coarse resolution

(18.5 by 18.5 km) estimates of forest loss in the location of an
IBA and the overall score for the impact of the threat from logging
at the site, as estimated through eld-based IBA monitoring,
indicates that there is a broad agreement between remote sensing
and eld-based monitoring. This suggests that eld-based IBA
monitoring assessments are being applied reasonably consistently
worldwide and that they provide meaningful data, despite the protocols simplicity in reducing threat scores to a 4-point scale, and
despite the fact that surveyors often lack considerable experience
and technical training and may only be able to visit part of a site.
However, while the relationship was signicant, there was still
considerable scatter around it, indicating variation in the eldbased monitoring scores worthy of further investigation. Further
validation is also required to assess the robustness of IBA monitoring data for other threats, and for aspects of the condition of sites
and responses in place.
We found a general lack of signicant systematic differences
between forest loss in the location of IBAs and the monitoring
scores for either scope or severity of logging. This may have several
possible non-exclusive explanations. First, we must consider the
coarse resolution of the remote-sensing assessment, which is
based on 342 km2 cells, and our amalgamating values for any cell
overlapping an IBA, no matter how small the overlap. Consequently, estimates of forest loss in the location of IBAs might not
represent the actual loss in the IBA, especially for smaller sites.
Additionally, it may be that neither of these components alone is
a good predictor of the overall impact of logging (because in some
cases a high impact results from intense logging within a small
proportion of the site, while in other cases it is a consequence of
less intense but more widespread logging), or that selective logging
recorded in the eld is not detected by the remote sensing data.
Logging is just one of several possible drivers of forest loss in
IBAs. Agriculture, residential or commercial development, energy
production, mining, and transport (including road and rail) disease
and natural or managed res, in combination and separately, may
drive loss of forest. Consequently, forest loss could be attributed to
other threats by the eld monitoring. People monitoring IBAs may
have applied methods inconsistently between sites, they may have
assessed threat severity over time frames that extend further into
the past than 2000, or they may have been poor at scoring the
individual elements of threat impact, but ensured that the product
of these generated the overall impact that they regarded as
appropriate.
Among the IBAs with the greatest discrepancies between forest
loss and scope or severity scores (i.e. where loss fell outside the
95% CI around the mean for that score), were the Central Cardamoms in Cambodia and Hwange National Park in Zimbabwe. These
sites had low scope and severity scores (all 61), but high forest loss
(4.0% and 18.9%), suggesting that the threat impact may have been
underestimated in the eld monitoring scores. In 2003, logging
was recognised as a threat to the Central Cardomons, with recommendations to prohibit activities that would have a negative affect
on biodiversity and discourage logging (Seng Kim Hout et al.,

G.M. Buchanan et al. / Biological Conservation 167 (2013) 334338

337

Fig. 3. Distribution of IBAs identied for forest-dependent species, showing forest loss during 20002005 (divided into quantiles, with black showing highest rates of loss and
light grey indicating the lowest loss).

2003). Thus, while the threat was recognised, it was apparently

underestimated. Logging was also recognised as a threat to
Hwange National Park in 1998 (Barnes, 1998), again suggesting
the threat score was an underestimate. In this site however strong
vegetation clearance around the periphery of the site, elephant
damage to trees and die back due to weather were also cited as
threats to the site (Barnes, 1998). All of these could result in a
change in vegetation in the location of the IBA that could have been
interpreted as forest loss from the remote sensing data, while not
being reected in the eld monitoring scores.
Almost half (25) of the 55 IBAs with high eld monitoring scope
scores (P2, suggesting high pressure), but low remote sensed forest loss were in Kenya. While some of these sites are threatened by
large potential developments that have not yet had a large impact
on the ground (Tana River Forests IBA and North Nandi Forest), this
is not the case for others (e.g. Arabuko-Sokoke.) This might suggest
systematic overestimation of the scope of logging threats to IBAs
from eld-based evaluations, methodological differences in the
way the data were collected and processed, or the driver of forest
loss being attributed away from another source, such as agricultural expansion. Of the 72 sites which have both logging and agricultural expansion threat scores, only one differed by more than
one point for severity and three for scope, suggesting the threats
from these two sources at least were generally similar. There
may of course also be variation in the accuracy of forest loss estimates from remote sensing data, although the calibration and error
checking of these maps (Hansen et al., 2010) mean that errors
should be comparable across the globe.
Overall rates of forest loss in the location of IBAs (and in the
locations neighbouring IBAs) were lower than the global mean of
3.1% (Hansen et al., 2010), although as this gure is a global average, and many IBAs will have lost more than the mean amount between 2000 and 2005. Forest loss in the location of IBAs tended to
be lower than loss in the immediately surrounding areas, hinting
that identication as an IBA may have helped to protect habitat
(Butchart et al., 2012; Beresford et al., 2013), or that IBAs may simply be located in areas that are more difcult to access or extract
timber from, consistent with results for protected areas (Joppa
and Pfaff, 2009). Identication of a site as an IBA does not automatically bring with it any degree of protection (which reduces forest
loss on IBAs Beresford et al., 2013), but half of all IBAs partially or
wholly overlap with protected areas (Butchart et al., 2012) while
IBA inventories in Europe have stimulated many IBAs to be
designated as Special Protection Areas under the European Unions
Birds Directive (BirdLife International, 2013). It is not possible to

determine from our data whether the higher forest losses around
IBAs are a consequence of leakage, i.e. displacement to areas bordering sites (Ewers and Rodrigues, 2008), a phenomenon for which
the evidence is equivocal (e.g. Oliveira et al., 2007; Andam et al.,
2008; Beresford et al., 2013). Clusters of IBAs with particularly high
rates of forest loss were found to be widely scattered in both the
tropics and temperate regions. Irrespective of the driver of loss,
all such sites are priorities for more detailed monitoring, including
assessment of threats and, potentially, for the identication and
implementation of conservation interventions to address the
drivers of deforestation.
Concluding, there is a recognised urgent need for improved
biodiversity monitoring across the world (Butchart et al., 2010;
Pereira et.al., 2013), and especially in the tropics (Collen et al.,
2008). IBA monitoring is now underway in around 25% of IBAs
worldwide. While not all IBAs have been assessed more than once,
this represents a unique and growing dataset. These data are being
increasingly used by governments to set priorities for action and to
track progress towards achieving targets for addressing biodiversity loss (Mwangi et al., 2010; BirdLife International, 2012). Our
results demonstrate the consistency and robustness of the data
generated by simple, eld-based monitoring, through showing that
they are consistent with remote-sensing data. However, not all
IBAs are monitored in the eld and hence integrating alternative
approaches is increasingly urgent. Remote sensing data are a
recognised way to augment assessments of pressures (e.g. Buchanan et al., 2009), and are especially useful if simple methods are
used (e.g. Bastin et al., 2013).
Acknowledgements
We are grateful to the many individuals who submit the eld
monitoring data. We thank the anonymous referees whose comments improved an earlier version of this manuscript.
Appendix A. Supplementary material
Supplementary data associated with this article can be found,
in the online version, at http://dx.doi.org/10.1016/j.biocon. 2013.
08.031.
References
Achard, F., Eva, H.D., Stibig, H.-J., Mayaux, P., Gallego, J., 2002. Determination of
deforestation rates of the worlds humid tropical forests. Science 297, 999
1002.

338

G.M. Buchanan et al. / Biological Conservation 167 (2013) 334338

Andam, K.S., Ferraro, P.J., Pfaff, A., Sanchez-Azofeifa, G.A., Robalino, J.A., 2008.
Measuring the effectiveness of protected area networks in reducing
deforestation. Proc. Nat. Acad. Sci. USA 105, 1608916094.
Barnes, K.N., 1998. The Important Bird Areas of southern Africa. BirdLife South
Africa, Johannesburg.
Bastin, L., Buchanan, G., Beresford, A., Pekel, J., Dubois, G., 2013. Open-source
mapping and services for web-based land-cover validation. Ecological
Informatics 14, 916.
Beresford, A.E., Eshiamwata, G.W., Donald, P.F., Balmford, A., Bertzky, B., Brink, A.B.,
Fishpool, L.D.C., Mayaux, P., Phalan, B., Simonetti, D., Buchanan, G.M., 2013.
Protection reduces loss of natural land-cover at sites of conservation
importance across Africa. PLoS ONE 8, e65370.
Birdlife International, 2006. Monitoring Important Bird Areas: a Global Framework.
BirdLife International, Cambridge.
BirdLife International, NatureServe, 2011. Bird species distribution maps of the
world. UK, Arlington, USA, Cambridge.
BirdLife International, 2012. State of the Worlds Birds. <http://www.birdlife.org/
datazone/sowb> (accessed 01.03.12).
BirdLife International, 2013. State of the Worlds Birds: Indicators for Our Changing
World. BirdLife International, Cambridge.
Buchanan, G.M., Butchart, S.H.M., Dutson, G., Pilgrim, J.D., Steininger, M.K., Bishop,
K.D., Mayaux, P., 2008. Using remote sensing to inform conservation status
assessment: estimates of recent deforestation rates on New Britain and the
impacts upon endemic birds. Biol. Conserv. 141, 5666.
Buchanan, G.M., Donald, P.F., Fishpool, L.D.C., Arinaitwe, J.A., Balman, M., Mayaux, P.,
2009. An assessment of land cover and threats in Important Bird Areas in Africa.
Bird Conservation International 19, 4961.
Buchanan, G.M., Donald, P.F., Butchart, S.H.M., 2011. Identifying priority areas for
conservation: a global assessment for forest-dependent birds. PLoS ONE 6,
e29080.
Butchart, S.H.M., Scharlemann, J.P.W., Evans, M.I., Quader, S., Arico, S., Arinaitwe, J.,
Balman, M., Bennun, L.A., Bertzky, B., Besancon, C., Boucher, T.M., Brooks, T.M.,
Bureld, I.J., Burgess, N.D., Chan, S., Clay, R.P., Crosby, M.J., Davidson, N.C., De
Silva, N., Devenish, C., Dutson, G.C.L., Fernandez, D., Fishpool, L.D.C., Fitzgerald,
C., Foster, M., Heath, M.F., Hockings, M., Hoffmann, M., Knox, D., Larsen, F.W.,
Lamoreux, J.F., Loucks, C., May, I., Millett, J., Molloy, D., Morling, P., Parr, M.,
Ricketts, T.H., Seddon, N., Skolnik, B., Stuart, S., 2012. Protecting important sites
for biodiversity contributes to meeting global conservation targets. PLoS ONE 7,
e32529.
Butchart, S.H.M., Walpole, M., Collen, B., van Strien, A., Scharlemann, J.P.W., Almond,
R.E.E., Baillie, J.E.M., Bomhard, B., Brown, C., Bruno, J., Carpenter, K.E., Carr, G.M.,
Chanson, J., Chenery, A.M., Csirke, J., Davidson, N.C., Dentener, F., Foster, M.,

Galli, A., Galloway, J.N., Genovesi, P., Gregory, R.D., Hockings, M., Kapos, V.,
Lamarque, J.-F., Leverington, F., Loh, J., McGeoch, M.A., McRae, L., Minasyan, A.,
Morcillo, M.H., Oldeld, T.E.E., Pauly, D., Quader, S., Revenga, C., Sauer, J.R.,
Skolnik, B., Spear, D., Stanwell-Smith, D., Stuart, S.N., Symes, A., Tierney, M.,
Tyrrell, T.D., Vi, J.C., Watson, R., 2010. Global biodiversity: indicators of recent
declines. Science 328, 11641168.
Collen, B., Ram, M., Zamin, T., McRae, L., 2008. The tropical biodiversity data gap:
addressing disparity in global monitoring. Tropical Conservation Science 1, 97
110.
Crawley, M.J., 1993. GLIM for Ecologists. Blackwell Publishing, Oxford.
Ewers, R.M., Rodrigues, A.S.L., 2008. Estimates of reserve effectiveness are
confounded by leakage. Trends Ecol. Evol. 23, 113116.
Hansen, M.C., Stehman, S.V., Potapov, P.V., 2010. Quantication of global gross
forest cover loss. Proc. Nat. Acad. Sci. USA 107, 86508655.
Jones, J.P.G., Asner, G.P., Butchart, S.H.M., Karanth, K.U., 2013. The why, what and
how of monitoring for conservation. In: MacDonald, D.W., Willis, K.J. (Eds.),
Key Topics in Conservation Biology 2. Wiley-Blackwell, Cambridge, pp. 327
343.
Joppa, L.N., Pfaff, A., 2009. High and far: biases in the location of Protected Areas.
PLoS ONE 4, e8273.
Mwangi, M.A.K., Butchart, S.H.M., Barasa, F., Bennun, L.A., Evans, M.I., Fishpool,
L.D.C., Kanyanya, E., Madindou, I., Machekele, J., Matiku, P., Mulwa, R., Ngari, A.,
Statterseld, A.J., Siele, J., 2010. Tracking trends in key sites for biodiversity: a
case study using Important Bird Areas in Kenya. Bird Conserv. Int. 20, 215
230.
Oliveira, P.J.C., Asner, G.P., Knapp, D.E., Almeyda, A., Galvan-Gildemeister, R., Keene,
S., Raybin, R.F., Smith, R.C., 2007. Land-use allocation protects the Peruvian
Amazon. Science 317, 12331236.
Pereira, H.M., Ferrier, S., Walters, M., Geller, G., Jongman, R.H.G., Scholes, R.J.,
Bruford, M., Brummitt, N., Butchart, S.H.M., Cardoso, A., Coops, N.C., Dulloo, E.,
Faith, D.P., Freyhof, J., Gregory, R.D., Heip, C., Hft, R., Hurtt, G., Jetz, W., Karp, D.,
McGeoch, M.A., Obura, D., Onoda, Y., Pettorelli, N., Reyers, B., Sayre, R.,
Scharlemann, J.P.W., Stuart, S.N., Turak, E., Walpole, M., Wegmann, M., 2013.
Essential biodiversity variables. Science 339, 277278.
Salafsky, N., Salzer, D., Statterseld, A.J., Hilton-Taylor, C., Neugarten, R., Butchart,
S.H.M., Collen, B., Cox, N., Master, L.L., OConnor, S., Wilkie, D., 2008. A standard
lexicon for biodiversity conservation: unied classications of threats and
actions. Conserv. Biol. 22, 897911.
Seng Kim Hout, Pech Bunnat, Poole, C.M., Tordoff, A.W., Davidson, P., Delattre, E.
2003. Directory of Important Bird Areas in Cambodia: key sites for conservation.
Department of Forestry and Wildlife, Department of Nature Conservation and
Protection, Phnom Penh.