RS and GIS Applications

7/24/2019 RS and GIS Applications

1/22

Abstract

Predictive models of animal distributions based on habitat can be used to assess the likely effects

of changes in landuse on a species. In this study we developed a model of the distribution of

buzzard nests in part of Argyll, Scotland. The model was tested on a further study site. . !abitat

was described in terms of vegetation cover, derived from satellite imagery, and topography, using

a digital terrain model to classify altitude, slope, aspect and ruggedness. This data base was

incorporated into a "eographical Information System. #. $nvironmental data, in the form of

areas and boundary lengths of vegetation types and landscape classifications, were e%tracted

from the data base for circular areas of various radii from the centre of &'' m grid cells covering

each study area. (e also included counts of buildings and lengths of roadways. ). *oth logistic

regression analysis and discriminant function analysis were used to produce classification

models, which assigned each grid cell a probability that it contained a buzzard nesting area. The

best predictive model was based on median altitude, total boundary length between all vegetation

categories, the amount of moorland and the length of boundary between pre+thicket forestry and

open ground. &. This model successfully reclassified -./ of grid cells in the areas from

which it was developed and .#&/ in a test area. Previous studies have fre0uently predicted the

distribution of species within the environment, but here we were able to predict the distribution

of nesting areas within the distribution of a species.

7/24/2019 RS and GIS Applications

2/22

7/24/2019 RS and GIS Applications

3/22

Modelling landscape-scale habitat use using GIS and remotesensing: a case study with great bustards

Summary

1

Many species are adversely affected by human activities at large spatial scales and their conservation requires detailed information on

distributions. Intensive ground surveys cannot keep pace with the rate of land-use change over large areas and new methods are needed for

regional-scale mapping.

2

We present predictive models for great bustards in central pain based on readily available advanced very high resolution radiometer

!"#$%%& satellite imagery combined with mapped features in the form of geographic information system !'I& data layers. "s "#$%%

imagery is coarse-grained( we used a 12-month time series to improve the definition of habitat types. )he 'I data comprised measures of

pro*imity to features likely to cause disturbance and a digital terrain model to allow for preference for certain topographies.

+

We used logistic regression to model the above data( including an autologistic term to account for spatial autocorrelation. )he results from

models were combined using ,ayesian integration( and model performance was assessed using receiver operating characteristics plots.

ites occupied by bustards had significantly lower densities of roads( buildings( railways and rivers than randomly selected survey points.

,ustards also occurred within a narrower range of elevations and at locations with significantly less variable terrain.

/ogistic regression analysis showed that roads( buildings( rivers and terrain all contributed significantly to the difference between occupied

and random sites. )he ,ayesian integrated probability model showed an e*cellent agreement with the original census data and predicted

suitable areas not presently occupied.

0

)he great bustards distribution is highly fragmented and vacant habitat patches may occur for a variety of reasons( including the species

very strong fidelity to traditional sites through conspecific attraction. )his may limit recoloni3ation of previously occupied sites.

4

We conclude that "#$%% satellite imagery and 'I data sets have potential to map distributions at large spatial scales and could be applied

to other species. While models based on imagery alone can provide accurate predictions of bustard habitats at some spatial scales( terrain

and human influence are also significant predictors and are needed for finer scale modelling.

7/24/2019 RS and GIS Applications

4/22

Introduction

Many wild species are adversely affected by human-induced changes in land use that operate over very large spatial scales. 5or e*ample( in

6urope agricultural policy change and its consequent effects on farming practice have profoundly influenced many bird species !78onnor 9

hrubb 1:;0

7/24/2019 RS and GIS Applications

5/22

are needed for every grid square or pi*el covering the species distribution. $owever( it is likely to detect more subtle changes in distributions

than interpolation methods( providing the predictor variables are reasonably correlated with the habitat features chosen by the species being

mapped. 5ortunately( the digital data sets now available reasonably appro*imate some ecological requirements of the great bustard and

probably other species too. 7ur emphasis throughout was on the use of readily available data sets that later would permit scaling-up to the

national or regional scale.

7ur starting premise was that vegetation type( terrain characteristics and human disturbance determine bustard distributions in pain( factors

that may apply equally to other species. 'reat bustards favour open( steppe-like( landscapes comprising cerealBfallow rotations( a habitat

that is particularly under threat of intensification through irrigation under 6uropean Anion agricultural policy. Cumerous studies !/yon

1:;+>1&reported the absence of flocks within a band of

about 1 km around villages and busy roads. 5or other birds( strong effects of roads on breeding density and performance have been noted

!%ei@nen 9 5oppen 1::

7/24/2019 RS and GIS Applications

6/22

Methods

Study site

)he pilot study area measured c.120 K 1+2 km( centred on Madrid province( pain( with the lower right co-ordinate at 2L W( +:L: C

!5ig. 1&. It was chosen because accurate great bustard census data( geographic information system !'I& data coverages and satellite

imagery were readily available. )he area comprised N:O agricultural land( 2:N2O natural scrub or wooded cover( ;N2O forestry( N0O built

environment and the remaining 1N1O bare ground or open water !6uropean Anion 8orine /and 8over =ro@ect&. )he ma@ority of analyses

were confined to Madrid province itself because this was where bustard censuses were conducted but( where possible( models were

e*trapolated to the full study area !see later&.

Figure 1."ppro*imate location of the study site !bo*ed area& measuring 120 K 1+2 km in central pain( centred on Madrid province.

Gis coverages

?igiti3ed infrastructure maps were available from "utonomous 8ommunity of Madrid 8artographic ervice at 1 P 1>> >>> scale. )he data

were separated into four layers !roads( buildings( railways and river systems& using "rc#iew software !6%I 1::0&and then rasteri3ed to ;>-

m pi*els in Idrisi !6astman 1::&. )his resolution was chosen because a digital terrain model !?)M& was also available at this scale for the

province. We created new variables from each of the infrastructure layers by replacing the central pi*el of a 1+ K 1+-cell moving window with

the proportion of pi*els recording the feature of interest !)able 1&. )his is equivalent to calculating the percentage land cover of the feature at

;>-m resolution within a c.1-km2quadrat.

Variable Defnition

GIS layers

RoadsProportion of 80-m pixels in a 13 13-array containing roads. E!i"alent to t#e density of roads at 1$1-

%m resol!tion

&!ildings Proportion of 80-m pixels in a 13 13-array containing '!ildings or large '!ilt str!ct!res s!c# as air(elds

Rail)ays Proportion of 80-m pixels in a 13 13-array containing rail)ay trac%s

Ri"ers Proportion of 80-m pixels in a 13 13-array containing ri"ers

*ltit!de +#e altit!de in m recorded on t#e 80-m digital terrain model

+errain

"aria'ility ,

oe/cient of "ariation in altit!de in a -pixel array of 80-m pixels. eas!res "ariation in altit!de at

0$1-%m,resol!tion

7/24/2019 RS and GIS Applications

7/22

Variable Defnition

GIS layers

+errain

"aria'ility 81

oe/cient of "ariation in altit!de in a 2 2-pixel array of 80-m pixels. eas!res "ariation in altit!de at

0$,-%m,resol!tion

+errain

"aria'ility 12

oe/cient of "ariation in altit!de in a 13 13-pixel array of 80-m pixels. eas!res "ariation in altit!de at

1$1-%m,resol!tion

Slope slope. +#e maxim!m of eit#er t#e nort#4so!t# or east4)est slopes across a 3 3-pixel array

Satellite

imagery

567I mont#9+#e "al!e of t#e normali:ed di;erence "egetation index for eac# mont# 'ased on a maxim!m "al!e

composite of *7

7/24/2019 RS and GIS Applications

8/22

Bird census data

" complete great bustard census was conducted in March 1::4 for the whole of Madrid province. )he distribution of the species in the study

area was known from previous censuses !"lonso 9 "lonso 1::>( 1::0&. ?uring 1 week three teams each of two e*perienced observers

counted bustard flocks at the known breeding areas and also searched all other potential sites. In practice( birds sighted @ust beyond the

province boundary were also recorded. 7ne-hundred and three flocks !:0> birds& were first marked on field maps( then digiti3ed and the

point coverage rasteri3ed to ;>-m and 1N1-km resolutions( recording the presence of bustards in the pi*el. )his resulted in 41 pi*els with one

or more flocks at 1N1-km resolution( and :2 pi*els at ;>-m resolution. 5or comparison( we generated equivalent random point coverages(

stratifying them geographically both to sample the whole province and to reduce spatial autocorrelation !i.e. to reduce the probability of using

ad@acent pi*els&. )he number of random points selected is important because prevalence !i.e. the ratio of positive to negative pi*els& affects

the outcome of model performance testing in logistic regression as used here !5ielding 9 ,ell 1::4

7/24/2019 RS and GIS Applications

9/22

%esults of logistic regression models are often @udged as successful if predicted probabilities R >N correspond with observed occurrences

and values S >N with absences. $owever( this dichotomy is arbitrary and lacks any ecological basis< patches rated with a >N0 probability of

occurrence may in fact be unsuitable. )he more powerful approach used here is to assess model success across the full range of

dichotomies using receiver operating characteristics !%78& plots. %78 plots are widely used in clinical chemistry !,eck 9 hult3 1:;0N !for

a chance performance& to 1N> for a perfect fit. We generated %78 plots using = software and calculated the "A8 and its standard error

using a non-parametric approach. )he results are reported here as the "A8 U its standard error together with the significance of a test that

the area F >N( i.e. that the model results do not differ from chance.

)he probability surfaces derived from the two independent logistic regression models were combined using ,ayesian inference after first

resampling the 1N1-km surface to ;>-m resolution. )he technique permits prior probabilities !for e*ample derived from one model& to be

revised on the basis of new probabilities calculated from a second model. )he appropriate formula !from=ereira 9 Itami 1::1&isP

where PC?#Iis the probability derived from the C?#I model( and P'Iis the probability derived from the model based on 'I data layers.

,ayesian approaches to decision-making have previously been used as here by=ereira 9 Itami !1::1&( and in other ways in wildlife

distribution modelling by"spinall 9 #eitch !1::+&and)ucker et al. !1::4&.

Results

Univariate analyses

"t ;>-m resolution( neither the bustard locations nor the random points e*hibited significant spatial autocorrelation !Morans I S G>N>>>1 in

both cases&. 5ollowing 8liff 9 7rd !1:;1&we therefore compared site characteristics using standard t-tests ad@usted for unequal variance.

ites occupied by bustards had significantly lower densities of roads( buildings( railways and rivers than random points !)able 2&. 7f these(

the effect of buildings was particularly strong< bustards occurred at sites with a mean of only >N;O land cover by buildings( whereas random

sites averaged 1>NO !PS >N>>1&. )able 2gives the ranges for each variable within which bustards occurred and these are combined as a

threshold mask in 5ig. +a. )his clearly shows the elimination of the radiating network of roads and buildings from Madrid 8ity.

Variable Bustard sites Random points Adusted t!test Ran"e used by bustards

Roads 0$0,, = 0$0> 0$0>3 = 0$080 ,$,8? 041$0

&!ildings 0$008 = 0$0, 0$10 = 0$,1 >$3,??? 0418$3

Rail)ays 0$00 = 0$0,, 0$018 = 0$0, ,$0,? 0411$,

7/24/2019 RS and GIS Applications

10/22

Variable Bustard sites Random points Adusted t!test Ran"e used by bustards

Roads 0$0,, = 0$0> 0$0>3 = 0$080 ,$,8? 041$0

Ri"ers 0$0,8 = 0$0>> 0$03 = 0$0 3$11?? 041$

*ltit!de A3$2 = $2 A22$A = ,8$> >$1A??? 4A80 m

+errain "aria'ility , 0$>0 = 0$>13 0$2>2 = 0$81> >$1??? 041$0

+errain "aria'ility 81 0$8A1 = 0$ 1$18 = 1$0,A $1??? 04,$3A

+errain "aria'ility 12 1$13 = 0$>2 1$22 = 1$,A0 $>8??? 0$034,$81

Slope 3$08 = 3$0 $0 = A$2, 3$8??? 041,$2

Table #. $omparison o% %eatures around sites occupied by "reat bustards and &' random points (except

%or terrain variability and slope based on )' and )* sites+ respectively+ to eliminate ed"e e,ects-. Values

are means standard deviations and all t!tests are adusted %or si"nifcant une/ual variance

Figure . )hreshold masks for areas suitable for bustards !in black& based on !a& constraints of roads( buildings( railways and rivers(

and !b& with the addition of altitude and terrain variability.

,ustards occurred within a narrow range of elevations from 00 to 4;> m a.s.l.( whereas random points covered the range 2>B21: m a.s.l.

)he terrain surrounding bustard sites was also significantly less variable than that around random sites at all three scales e*amined

!)able 2&. )here was no obvious trend for a difference between the scales( although the significance of the difference between bustard and

random sites increased within window si3e. Asing 1+ K 1+ cells !about 1 km2&( bustards occurred at sites with up to 2N;O coefficient of

variation in altitude !mean 1N2O&( whereas random sites had up to N:O variability !mean 2N>O&. 8ombining the results for altitude and

terrain variability with the mask in 5ig. +ayields5ig. +b. )his indicates that only 21++ km2of the 0> km2!+2N0O& studied meets the

characteristics of infrastructure( elevation and river networks at sites used by bustards. 5urthermore( this area is fragmented into blocks by

the radiating networks from Madrid 8ity.

Predictive modelling using logistic regression

,uilding threshold masks !e.g.5ig. +&is a convenient way to define areas meeting criteria but suffers from failing to take into account

interactions between variables. Asing the 'I variables in )able 2!selecting terrain variability 10: for variation in elevation&( we built

predictive models for bustard presence using forward stepwise logistic regression. 7nly railways and slope were not included in the model

!at PS >N>& and all other variables were significant at PS >N>1 !)able +&. )he greatest contributions came from terrain variability

7/24/2019 RS and GIS Applications

11/22

!PS >N>>>1& and housing density !PS >N>>>2&. 7verall the %78 plot for the model !5ig. a& had an "A8 of >N;:; U >N>2+ and was highly

significant !PS >N>>1&. " simplified probability surface for bustard occurrence based on the significant 'I variables is shown in 5ig. . )here

are obvious similarities with5ig. +bbut the probability plot shows far more te*ture and greater weighting in the east-central and e*treme

south-west parts of the province.

0odel Predictor variable $oecient 2tandard error

GIS data layers 80-m resol!tion9 Roads B10$2A?? 3$A8

&!ildings B,$20??? A$13

Ri"ers B2$8?? 3$A>

*ltit!de B0$002??? 0$003

+errain "aria'ility 12 B1$1>??? 0$,2

onstant9 2$,??? ,$08

567I data 1$1-%m resol!tion9 567I Can!ary9 0$1,? 0$0

567I 5o"em'er9 B0$31??? 0$08

567I *pril9 4 567I C!ly9 0$08?? 0$03

onstant9 ,3$,?? 8$2,

Table '. 2ummary results o% the lo"istic re"ression analyses. The si"nifcance o% the coecients was

assessed usin" the 3ald statistic

Figure !. %78 plots for !a& logistic regression model with five 'I variables( !b& logistic regression model based on C?#I( !c&

autologistic regression model based on C?#I( and !d& ,ayesian integrated model incorporating both the 'I variables and the

autologistic C?#I model.

7/24/2019 RS and GIS Applications

12/22

Figure ". implified probability surface for the occurrence of great bustards based on logistic regression analysis of five 'I

variables.

We built a separate model based on the satelli te data at 1N1-km resolution. )he stepwise inclusion of the e*planatory variables resulted in the

selection of three C?#I variables for January( Covember and the contrast "pril minus July !)able +&. Cone of the quadratic terms was

significant. )he model %78 plot !5ig. b&had an "A8 of >N:+ U >N>22 and was highly significant !PS >N>>1&. )he resultant probability surface

is shown in5ig. 0. "lthough this model of habitat suitability was derived by analysing only the presence or absence of bustards( the

probabilities also relate to the numbers of birds present !5ig. 4&. Most points lie beneath the diagonal from bottom left to top right( indicating

that large numbers of birds only occur in pi*els with high predicted probabilities.

Figure #. implified probability surface for the occurrence of great bustards based on logistic regression analysis of temporal C?#I

signatures.

Figure $. %elationship between the predicted probability of occurrence generated by logistic regression analysis of presenceB

absence data and the number of bustards recorded.

We took account of autocorrelation in the model based on the satellite data by including an autologistic term< the regression coefficients

stabili3ed within five iterations to produce5ig. ;.8omparing this with5ig. 4reveals the down weighting of isolated pi*els previously defined

as suitable( and the consolidation of the larger suitable habitat blocks. )he %78 plot for the autologistic model !5ig. c&differed very little

from the standard C?#I model !"A8 F >N:+ U >N>22( PS >N>>1&. )he 1N1-km2resolution probability surface in5ig. ;was resampled to ;>-m

resolution in order to combine it with the thresholds imposed by infrastructure and natural features !i.e.5ig. +&. )aking an arbitrary lower

threshold of >O probability of occurrence from the habitat suitability map !5ig. ;&alone yielded an area of :42 km2or 1N:O of the province.

)his provides an estimate of the area of the province with vegetation inde* signatures comparable to those used by bustards. When

combined with5ig. +b( however( this falls to 4> km2or ;N4O of the province !5ig. :&. )he difference between these two values !>> km2& is

the amount of apparently suitable habitat that bustards are unlikely to use due to pro*imity to unfavourable landscape

elements. 5igure :!area 4> km2& is the area of Madrid province that cannot be distinguished from sites occupied by bustards using 'I and

remote sensing( based on simple threshold mapping.

Figure %. implified probability surface for the occurrence of great bustards based on autologistic regression analysis of temporal

C?#I signatures. Cote how in comparison with5ig. 0the image shows better definition of core areas.

Figure &. )hreshold map for the occurrence of great bustards based on the limits from5ig. +band habitat suitability R >N from 5ig. 4.

?efined in this way( ;N4O of the province could be used by great bustards.

7/24/2019 RS and GIS Applications

13/22

"n alternative approach is to combine the two probability surfaces based on feature data !5ig. & and C?#I !5ig. ;&using ,ayesian

integration !see the Methods&. "s5ig. represents largely fi*ed features of the landscape today( we regarded these as prior probabilities that

may be refined by consideration of 5ig. ;which represents land cover based on green biomass. )he ,ayesian integrated probability model

!5ig. 1>&showed an e*cellent agreement with the original census data !simplified in 5ig. 11for ease of display& and had an %78 "A8 value

of >N:0: U >N>1+( PS >N>>1 !5ig. d&.

Figure 1'. ,ayesian integrated probability map for the occurrence of great bustards in central pain.

Figure 11. =redicted probabilities of occurrence R >N4 from the ,ayesian integrated model overlaid with the recorded locations of

great bustard flocks. Cote that flocks observed beyond the province boundary are ad@acent to areas predicted to be highly suitable

for bustards.

(iscussion

We started with a premise gleaned from the literature that great bustard distributions may be related to vegetation( topography and human

influence( and sought digital data sets that could characteri3e these features. "t the landscape scale( our analyses successfully predicted the

occurrence of bustards around Madrid province and all these features were significant predictors.

)he negative effect of human disturbance on bustard occurrence is not surprising but is interesting because at first sight it appears that the

birds are well integrated into the urban setting in Madrid province. )here is growing evidence that roads impact on large mammals !Mech

1:;:

7/24/2019 RS and GIS Applications

14/22

unoccupied sites and indicates the potential of using C?#I time series to predict avian habitats. We believe that our model largely detected

the sharp change in C?#I brought about by rapid crop growth in cereal fields in spring followed by biomass loss through harvesting in July.

$owever( based on local knowledge( it appears that the model is not simply predicting all cereal areas( because some large areas were

e*cluded. $ow the subset of more suitable fields is recogni3ed is less clear.Maurer !1::&has previously shown a relationship between

avian abundance and C?#I( but the correlation of abundance with probabilities of occurrence generated by temporal C?#I curves is new.

)he relationship here indicates that sites defined as suitable have the potential to hold most birds( although they do not necessarily do so.

8onversely( sites defined as unsuitable appear capable of holding only few birds.

?espite the coarse-grained nature of "#$%% data( models based on C?#I alone do provide significant predictions of bustard occurrence. In

some situations( remotely sensed data may be all that are available and our results indicate that they are a useful first step in model building.

$owever( we conclude that while models based on vegetation alone can provide accurate predictions of bustard habitats at some spatial

scales( terrain and human influence are also significant predictors and are needed for finer scale modelling. 5or e*ample( the breeding site in

the centre of the south-east part of the province !5ig. 11& was not predicted by the C?#I probability surface !5ig. ;& but appeared after

integration with the landscape 'I variables. We thus concur withManel et al. !1:::&that prediction success in distribution models may be

enhanced when local data are available( despite the apparent success of coarse-grained models.

)he use of an autologistic term based on a modified 'ibbs sampler !"ugustin( Mugglestone 9 ,uckland 1::0&in the logistic regression

models proved very effective in sharpening the definition between occupied and unoccupied patches. "lthough similar visual effects can be

achieved more simply in image processing through 'aussian or median fil ters( they lack an ecological basis( whereas the 'ibbs sampler

uses information on the pro*imity of occupied pi*els. We chose to consider neighbours within N km of the target pi*el !i.e. a : K : window

at 1N1-km resolution& because this seemed an appropriate scale for breeding great bustards( comparable with the average area of influence

of a lek."ugustin( Mugglestone 9 ,uckland !1::0&tried a range of window si3es to km distance for red deer Cervus elaphusbut found that

+ km gave optimum results in terms of the balance between model accuracy and computation time. 5urther work is needed on the selection

of optimum neighbourhood distances for a range of species when trying to account for spatial autocorrelation in distribution models.

'enerally( the prediction of occupied sites was more successful than the prediction of absences. 5or e*ample( the ,ayesian integrated

model !5ig. 1>& correctly predicted :+N>O of occupied sites and 4;N:O of unoccupied sites( based on a >O probability dichotomy. )his is

the opposite result to that reported byManel et al. !1:::&for si* bird species and reinforces their view that more work is needed on model

performance indicators and their determinants. ?ifferences in performance at predicting presence and absence may be due to a number of

factors !5ielding 9 ,ell 1::4& but we offer the following ecological e*planations. 5irst( the bustard census data were a single-day snapshot of

occurrences( taken when flocking was at its ma*imum at lek sites. Ander these conditions( ad@acent areas could be mistakenly regarded as

unsuitable whereas with greater flock dispersion they would be occupied. econdly( the panish great bustard population is fragmented into

numerous habitat patches and( due to high site f idelity !"lonso et al. 1::&( movement between patches for breeding may be scarce. )ilman(

/ehman 9 Eareiva !1::4&predict that these conditions will lead to the non-occupancy of some suitable sites even at population equilibrium.

)hus we might e*pect some unoccupied sites to be ecologically inseparable from occupied sites. )hirdly( recent field studies by/ane( "lonso

9 Martin !2>>1&could not discriminate occupied from unoccupied but apparently suitable areas for bustards in central pain( and it is

unreasonable to e*pect large-scale models to outperform intensive field investigations. )heir study compared 1+ areas occupied by great

bustards with a matched set of 12 unoccupied areas nearby. )hese sites could not be distinguished using discriminant analysis on a set of

variables defining relevant habitat characteristics such as crop type( substrate heterogeneity( field si3e( presence of roads( villages and

7/24/2019 RS and GIS Applications

15/22

powerlines. ,oth their study and ours indicate that not all potentially suitable areas are occupied and that great bustards show fidelity to sites

regardless of the availability of suitable habitat elsewhere. We believe that this may be due to a combination of a series of local e*tinction

processes in recent decades due to human-induced habitat deterioration and hunting( and the very low re-coloni3ation capability of the

species which arises from its complicated lek breeding system. ettlement patterns are probably determined by the presence of conspecifics

rather than habitat cues !J.8. "lonso( unpublished data&. )his means that conservation efforts must be directed towards protecting traditional

lek sites( and that once a lek is e*tinct the site will probably remain empty in the future.

)he limited number of historically documented e*tinctions !e.g. the small patch predicted as suitable west of Madrid city was occupied

2> years ago and 11& and circumstantial evidence !e.g. place names in the larger area with high probability of occurrence south of

the city& suggest that5ig. 11shows the likely distribution of great bustards some 2>B> years ago. ince then many of the predicted areas

have become vacant due to local e*tinction processes. It is also important to remember that the locations in5ig. 11were based on breeding

birds and that these perform seasonal movements and e*pand the occupied area throughout the year. 8onservation based on breeding sites

alone would almost certainly be ineffective( and a landscape-scale approach incorporating seasonal changes in locations is required.

7ur study was preliminary in developing methods for landscape-scale models but the intention is to e*amine distributions at the national

scale. We envisage two ma@or challenges in relation to scaling-up to larger study areas. 5irst( as mentioned earlier( the selection of ma*imum

value C?#I data may not lead to adequate standardi3ation over large areas due to a level-of-view and solar-angle bias. In practice( this

means that a given surface may generate a different C?#IBM#8 score at different locations on the image. )he main factor causing this is the

inherent variability of surfaces at different look angles. " traditional solution to this problem is to first standardi3e images against a common

target such as a large body of water or bare ground !annier et al. 1::;&. " better approach( however( may be to take into account the

bidirectional reflectance distribution function !,%?5& of the surfaces being sensed !8ihlar( Manak 9 #oisin 1::&. 8orrection for these effects

can be undertaken using linear semi-empirical kernel-driven models !%ou@ean( /eroy 9 ?eschamps 1::2& that ad@ust image data for ,%?5

variability and e*tract surface information. Init ial results indicate that the new models provide enormous scope for improving data quality and

also provide useful additional information !8hopping 1::;&.

)he second challenge to scaling-up is that animals may not choose habitats according to absolute needs but may adopt a comparative

approach. =articularly on the edge of a species range( occupied sites are likely to be far from the ideal habitat with poorer breeding

performance to match !/awton et al. 1::&. "lso( several habitat types with different spectral signatures may be equally suitable at large

geographical scales. /arge-scale models must therefore include samples from across the species range and analysis may be better

partitioned spatially prior to modelling.

)c*nowledgements

We are grateful to ?r "ntonio DagVe of Infocarto( .".( for supplying the C7"" data( and the "utonomous 8ommunity of Madrid 8artographic

ervice for the 'I coverages. John Mc"rthur and "lison Martin helped prepare the data for analysis. ?r 6nrique Martn( ?r Manuel Morales(

8arlos Martn( ?r imon /ane and ?r Javier ". "lonso assisted with the collection of bird data. )his research was @ointly funded by ,ritish

8ouncilHMinistry of 6ducation and 8ulture "ccion Integrada $,:0-;2 and $,:4-0: awards( ?'I8D) pro@ect =,:->>0;( and by the

?irecciXn 'eneral del Medio Catural of Madrid 8ommunity.

7/24/2019 RS and GIS Applications

16/22

Re+erences

"lonso( J.8. 9 "lonso( J.". !1::>& Parmetros demogrficos, seleccin de hbitat y distribucin de la avutarda en tres regiones

espaolas. I87C"( Madrid( pain.

"lonso( J.8. 9 "lonso( J.". !1::0& )he great bustard Otis tardain painP present status( recent trends and an evaluation of earlier

censuses. Biological Conservation( $$( 4:B;0.

o 8ross%ef(

o Web of cienceY

"lonso( J.8.( "lonso( J.".( Martn( 6.( Morales( M. !1::& %ange and patterns of great bustard movements at #illafZfila( CW

pain.Ardeola( !2( 0:B40.

"lonso( J.8.( Morales( M.,.( "lonso( J.". !2>>>& =artial migration( lek fidelity and nesting area fidelity in female great

bustards Otis tarda. Condor( 1'2( 124B1+0.

o 8ross%ef(

o Web of cienceY

"spinall( %. 9 #eitch( C. !1::+& $abitat mapping from satellite imagery and wildlife survey data using a ,ayesian modeling

procedure in a 'I. Photogrammetric ngineering and !emote "ensing( "&( +4B+.

o Web of cienceY

"ugustin( C.$.( Mugglestone( M.".( ,uckland( .). !1::0& "n autologistic model for the spatial distribution of wildlife. #ournal of

Applied cology( ( ++:B+4.

o 8ross%ef(

o Web of cienceY

"ustin( '.6.( )homas( 8.J.( $ouston( ?.8.( )hompson( ?.,.". !1::0& =redicting the spatial distribution of bu33ard Buteo

buteonesting areas using a geographic information system and remote sensing. #ournal of Applied cology( ( 11B1>.

o 8ross%ef(

o Web of cienceY

"very( M.I. 9 $aines-Doung( %.$. !1::>& =opulation estimates of the dunlin Calidris alpinaderived from remotely sensed satellite

imagery of the 5low 8ountry of northern cotland. $ature( !!( ;0>B;02.

o 8ross%ef(

o Web of cienceY(

o "?

,eck( J.,. 9 hult3( 6.E. !1:;0& )he use of relative operating characteristic !%78& curves in test performance

evaluation.Archives of Pathology and %aboratory &edicine( 11'( 1+B2>.

o =ubMed(

o Web of cienceY

,uckland( .). 9 6lston( ?.". !1::+& 6mpirical models for the spatial distribution of wildlife. #ournal of Applied cology( '( 4;B

:.

o 8ross%ef(

7/24/2019 RS and GIS Applications

17/22

o Web of cienceY

8aldow( %.W.'. 9 %acey( =.". !2>>>& Introduction. /arge-scale processes in ecology and hydrology. #ournal of Applied

cology( $!upplement 1&( 0B12.

?irect /inkP

o "bstract

o 5ull "rticle !$)M/&

o

o =?5!;;E&

o %eferences

o Web of cienceY )imes 8itedP 1:

8hopping( M.J. !1::;& 6valuation of the performance of linear semi-empirical ,%?5 models in a semi-arid grassland

biome.Proceedings of the '(th Annual Conference of the !emote "ensing "ociety!eds =.J.".,urt( 8.$.=ower 9 =.M.Tukowsky@&(

pp.0;B4. %emote ensing ociety( Cottingham( AE.

8ihlar( J.( Manak( ?.( #oisin( C. !1::& "#$%% bidirectional reflectance effects and compositing. !emote "ensing of the

nvironment(!%( 44B;;.

o 8ross%ef(

o Web of cienceY

8liff( ".?. 9 7rd( J.E. !1:;1& "patial Processes) &odels and Applications. =ion( /ondon( AE.

8ollar( C.J. 9 "ndrew( =. !1:;;& Birds to *atch) the +CBP *orld Chec-list of hreatened Birds. I8,= )echnical =ublication ;.

I8,=( 8ambridge( AE.

?el $oyo( J.( 6lliot( ".( argatal( J. !1::0& /andboo of the Birds of the *orld, #ol. +. /yn*( ,arcelona( pain.

6astman( J.%. !1::& +drisi for *indo0s. 1ser2s 3uide( #ersion 1.>. 8lark /abs( 8lark Aniversity( Worcester( A".

6%I !1::0&Arc4ie0 3+". 6nvironmental ystems %esearch Institute( %edlands( 8".

5arina( ". !1::4& /andscape structure and breeding bird distribution in a sub-Mediterranean agro-ecosystem. %andscape

cology(12( +0B+4;.

o 8ross%ef(

o Web of cienceY )imes 8itedP 02

5ielding( ".$. 9 ,ell( J.5. !1::4& " review of methods for the assessment of prediction errors in conservation presenceHabsence

models. nvironmental Conservation( 2!( +;B:.

o 8ross%ef(

o Web of cienceY

5@eldsa( J.( 6hrlich( ?.( /ambin( 6.( =rins( 6. !1::4& "re biodiversity hotspots correlated with ecoclimatic stability[ " pilot study

using the C7""-"#$%% remote sensing data. Biodiversity and Conservation( #( >1B22.

o 8ross%ef(

o Web of cienceY

7/24/2019 RS and GIS Applications

18/22

'aston( E.J. 9 ,lackburn( ).M. !1::& Mapping biodiversity using surrogates for species richness B macro-scales and Cew World

birds. Proceedings of the !oyal "ociety of %ondon, "eries B, Biological "ciences( 2#2( ++B+1.

o 8ross%ef(

o Web of cienceY )imes 8itedP +(

o "?

$eredia( ,.( %ose( /.( =ainter( M. !1::0& 3lobally hreatened Birds in urope. 8ouncil of 6urope( trasbourg( 5rance.

$idalgo de )rucios( .J. !1::>& World status of the great bustard !Otis tarda& with special attention to the Iberian peninsula

populations. &iscelanea 5oologica( 1!( 104B1;>.

$ill( ?.( $ockin( ?.( =rice( '.( )ucker( %.( Morris( %.( )reweek( J. !1::4& ,ird disturbanceP improving the quality and utility of

disturbance research. #ournal of Applied cology( !( 24B2;:.

o 8ross%ef(

o Web of cienceY

$olben( ,. !1:;0& 8haracteristics of ma*imum value composite images from temporal "#$%% data. +nternational #ournal of

!emote "ensing( $( 141B1+.

o 8ross%ef(

o Web of cienceY )imes 8itedP 4:(

o "?

Johnsgard( =.". !1::1& Bustards, /emipodes and "andgrouse. Birds of 6ry Places . 7*ford Aniversity =ress( 7*ford( AE.

o Web of cienceY )imes 8itedP 1

Eremer( %.'. 9 %unning( .W. !1::+& 8ommunity type differentiation using C7""H"#$%% data within a sagebrush-steppe

ecosystem. !emote "ensing of the nvironment( !#( +11B+1;.

o 8ross%ef(

o Web of cienceY

/ane( .J.( "lonso( J.8.( Martn( 8.". !2>>1& $abitat preferences of great bustard Otis tardaflocks in the arable steppes of central

painP are potentially suitable areas unoccupied[ #ournal of Applied cology( %( 1:+B2>+.

?irect /inkP

o "bstract

o 5ull "rticle !$)M/&

o

o =?5!014E&

o %eferences

o Web of cienceY

/avers( 8.=.( $aines-Doung( %.$.( "very( M.I. !1::0& )he habitat associations of dunlin Calidris alpinain the 5low 8ountry of

northern cotland and an improved model for predicting habitat quality. #ournal of Applied cology( ( 24:B2:>.

o 8ross%ef(

7/24/2019 RS and GIS Applications

19/22

o Web of cienceY

/awton( J.$.( Cee( .( /etcher( ".J.( $arvey( =.$. !1::& "nimal distributionsP patterns and processes. %arge-"cale cology and

Conservation Biology!eds =.J.6dwards( %.M.May 9 C.%.Webb&( pp. 1B;. ,lackwell cientific =ublications( 7*ford( AE.

/egendre( =. 9 /egendre( /. !1::;& $umerical cology. 6evelopments in nvironmental &odelling '7. 6lsevier( "msterdam( the

Cetherlands.

/yon( J./. !1:;+& /"C?")-derived land-cover classifications for locating potential kestrel nesting habitat . Photogrammetric

ngineering and !emote "ensing( !&( 2B2>.

o Web of cienceY

Mace( %.?.( Walker( J..( Manley( )./.( /yon( /.J.( Tuuring( $. !1::0& %elationships among gri33ly bears( roads and habitat in the

wan Mountains( Montana. #ournal of Applied cology( ( 1+:B1>.

o 8ross%ef(

o Web of cienceY

Manel( .( ?ias( J.M.( ,uckton( .).( 7rmerod( .J. !1:::& "lternative methods for predicting species distributionP an illustration

with $imalayan river birds. #ournal of Applied cology( #( 4+B44.

?irect /inkP

o "bstract

o 5ull "rticle !$)M/&

o

o =?5!2>2E&

o %eferences

o Web of cienceY

MarQal( ".%.. 9 Wright( '.'. !1::4& )he use of overlapping C7""-"#$%% C?#I ma*imum value composites for cotland and

initial comparisons with the land cover census on a cottish regional and district basis. +nternational #ournal of !emote

"ensing( 1%(:1B>+.

o 8ross%ef(

o Web of cienceY(

o "?

Mather( =.M. !1:::& Computer Processing of !emotely "ensed +mages. An +ntroduction. John Wiley and ons( 8hichester( AE.

Maurer( ,.". !1::& 3eographical Population Analysis ools for the Analysis of Biodiversity. ,lackwell cientific =ublications(

7*ford( AE.

Mech( /.?. !1:;:& Wolf population survival in an area of high road density.American &idland $aturalist( 121( +;4B+;:.

o 8ross%ef(

o Web of cienceY

Mladdenoff( ?.J.( ickley( ).".( $aight( %.'.( Wydeven( ".=. !1::& " regional landscape analysis and prediction of favorable gray

wolf habitat in the Corthern 'reat /akes region. Conservation Biology( &( 24:B2:.

7/24/2019 RS and GIS Applications

20/22

?irect /inkP

o "bstract

o

o =?5!112+4E&

o Web of cienceY

Morales( M.,.( "lonso( J.8.( "lonso( J.".( Martin( 6. !2>>>& Migration patterns in male great bustards. Au( 11$( :+B:;.

o 8ross%ef(

o Web of cienceY

78onnor( %.J. 9 hrubb( M. !1:;0& 8arming and Birds. 8ambridge Aniversity =ress( 8ambridge( AE.

7nrubia( ".( aen3 de ,uruaga( M.( 7sborne( =.( ,aglione( #.( =urroy( 5.J.( /ucio( ".J.( 8ampos( M.". !1::;& 4iabilidad de la

poblacion $avarra de avutardas. 8onsultora de %ecursos Caturales( #itoria( pain.

7rmerod( .J.( =ienkowski( M.W.( Watkinson( ".%. !1:::& 8ommunicating the value of ecology. #ournal of Applied

cology( #(;4B;.

?irect /inkP

o "bstract

o 5ull "rticle !$)M/&

o

o =?5!2++E&

o %eferences

o Web of cienceY

7sborne( =.6. 9 )igar( ,.J. !1::2& Interpreting bird atlas data using logistic modelsP an e*ample from /esotho( outhern

"frica.#ournal of Applied cology( 2&( B02.

o 8ross%ef(

o Web of cienceY

=ain( ?.J. 9 =ienkowski( M.W. !1::4& 8arming and Birds in urope. "cademic =ress( /ondon( AE.

=alma( /.( ,e@a( =.( %odrigues( M. !1:::& )he use of sighting data to analyse Iberian lyn* habitat and distribution. #ournal of

Applied cology( #( ;12B;2.

?irect /inkP

o "bstract

o 5ull "rticle !$)M/&

o

o =?5!+10E&

o %eferences

7/24/2019 RS and GIS Applications

21/22

o Web of cienceY

=aruelo( J.M. 9 /auenroth( W.E. !1::& %egional patterns of normali3ed difference vegetation inde* in Corth "merican shrublands

and grasslands. cology( $#( 1;;;B1;:;.

o 8ross%ef(

o Web of cienceY

=ereira( J.M.8. 9 Itami( %.M. !1::1& 'I-based habitat modeling using logistic multiple regressionP a study of the Mt. 'raham red

squirrel. Photogrammetric ngineering and !emote "ensing( "$( 14B1;0.

o Web of cienceY

%eed( ,.8.( ,rown( J.5.( #ander3ee( ?.( /oveland( ).%.( Merchant( J.W.( 7hlen( ?.7. !1::& Measuring phenological variability

from satellite imagery. #ournal of 4egetation "cience( "( 4>+B41.

?irect /inkP

o "bstract

o

o =?5!21:E&

o %eferences

o Web of cienceY

%ei@nen( %. 9 5oppen( %. !1::& )he effects of car traffic on breeding bird populations in woodland. I. 6vidence of reduced habitat

quality for willow warblers Phylloscopus trochilusbreeding close to a highway. #ournal of Applied cology( 1( ;B:.

o 8ross%ef(

o Web of cienceY

%ei@nen( %.( 5oppen( %.( )er ,raak( 8.( )hissen( J. !1::& )he effects of car traffic on breeding bird populations in woodland. III.

%eduction in density in relation to the pro*imity of main roads. #ournal of Applied cology( 2( 1;4B2>2.

o 8ross%ef(

o Web of cienceY

%ogers( ?.J. 9 Williams( ,.'. !1::& )setse distribution in "fricaP seeing the wood and the trees. %arge-"cale cology and

Conservation Biology!eds =.J.6dwards( %.M.May 9 C.%.Webb&( pp. 24B242. ,lackwell cientific =ublications( 7*ford( AE.

%ogers( ?.J.( $ay( .I.( =acker( M.J.( Wint( '.%.W. !1::4& Mapping land-cover over large areas using multispectral data derived

from the C7""-"#$%%P a case study of Cigeria. +nternational #ournal of !emote "ensing( 1%( +2:4B++>+.

o 8ross%ef(

o Web of cienceY(

o "?

%ou@ean( J./.( /eroy( M.( ?eschamps( =.D. !1::2& " bidirectional reflectance model of the earths surface for the correction of

remote sensing data. #ournal of 3eophysical !esearch( &$( 2> B2> 2> 0;.

?irect /inkP

7/24/2019 RS and GIS Applications

22/22

o "bstract

o "rticle

o

o =?5!112+E&

o %eferences

annier( 8.".?.( )aylor( J.8.( ?u =lessis( W.( 8ampbell( E. !1::;& %eal-time vegetation monitoring with C7""-"#$%% in outhern

"frica for wildlife management and food security assessment. +nternational #ournal of !emote "ensing( 1&( 021B0+:.

o 8ross%ef(

o Web of cienceY(

o "?

= !1::4& "P"" for *indo0s, 9.7.7. = Inc.( Michigan( A".

toms( ?.M.( ,ueno( M.J.( ?avis( 5.W. !1::4& #iewing geometry of "#$%% image composites derived using multiple

criteria.Photogrammetric ngineering and !emote "ensing( #( 0;1B0;:.

o Web of cienceY

)ilman( ?.( /ehman( 8./.( Eareiva( =. !1::4& =opulation dynamics in spatial habitats. "patial cology!eds ?.)ilman 9 =.Eareiva&(

pp.+B2>. =rinceton Aniversity =ress( =rinceton( CJ.

)ucker( E.( %ushton( .=.( anderson( %.".( Martin( 6.,.( ,laiklock( J. !1::4& Modelling bird distributions B a combined 'I and

,ayesian rule-based approach. %andscape cology( 12( 44B:+.

o 8ross%ef(

o Web of cienceY

Walker( %.6.( toms( ?.M.( 6stes( J.6.( 8ayocca( E.?. !1::2& Improved modelling of biological diversity with multitemporal

vegetation inde* data. echnical Papers of the :;;' Annual &eeting of the A"P!"