a novel method to calculate climatic niche similarity among … · 2018-02-05 · graphic ranges...

METHODS AND APPLICATIONS

A novel method to calculate climatic niche similarityamong species with restricted ranges—the case of terrestrialLycian salamanders

Dennis Rödder & Stefan Lötters & Mehmed Öz &

Sergé Bogaerts & Karolos Eleftherakos & Michael Veith

Received: 4 April 2011 /Accepted: 26 August 2011 /Published online: 22 September 2011# Gesellschaft für Biologische Systematik 2011

Abstract Within the framework of the present study wetest whether climatic niche similarity can be identified in amonophyletic group of species inhabiting remarkablyrestricted ranges by pooling presence data of all speciesinto a single concatenated data set and subsequentlyjackknifing single species. We expect that, when thejackknifed species differs markedly in its climatic nichefrom all other species, this approach will result in increasedniche homogeneity, allowing assessments of niche diver-gence patterns. To test our novel jackknife approach, wedeveloped species distribution models for all members ofLycian salamanders (genus Lyciasalamandra), native toTurkey and the adjacent Aegean islands using Maxent.Degrees of niche similarity among species were assessed

using Schoener’s index. Significance of results was testedusing null-models. The degree of niche similarity wasgenerally high among all seven species, with only L.helverseni differing significantly from the others. Carsticlime stones providing specific microhabitat features mayexplain the high degree of niche similarity detected, sincethe variables with the highest explanative power in ourmodels (i.e. mean temperature, and precipitation of thecoldest quarter) corresponded well with salamander naturalhistory observations, supporting the biologically plausibilityof the results. We conclude that the jackknife approachpresented here for the first time allows testing for nichesimilarity in species inhabiting restricted ranges and with fewspecies records available. Our results strongly support theview that detailed natural history information and knowledgeof microhabitats is crucial when assessing possible climatechange impacts on species.

Keywords Bioclimate . Lyciasalamandra . Maxent .

Salamandridae . Species distribution model

Introduction

The question of whether or not species’ climate niches areconservative over time periods of several decades to manymillennia has recently become a powerful field of research(e.g. Peterson et al. 1999, 2008; Wiens 2004; Wiens andGraham 2005; Broennimann et al. 2007; Fitzpatrick et al.2007; Warren et al. 2008; Beaumont et al. 2009; Rödderand Lötters 2009; Rödder et al. 2009b). It has beenproposed that niche conservatism may be a driving forcein speciation (e.g. Kozak and Wiens 2006; Kozak et al.2008) and may help to explain (Hawkins et al. 2006) orpredict biogeographical patterns via statistical species

D. Rödder (*)Zoological Research Museum Alexander Koenig,Adenauerallee 160,53113 Bonn, Germanye-mail: [email protected]

D. Rödder : S. Lötters :M. VeithDepartment of Biogeography, Trier University,54286 Trier, Germany

M. ÖzDepartment of Biology, Faculty of Science and Education,Akdeniz University,Antalya, Turkey

S. BogaertsLupinelaan 25,NL-5582CG Waalre, The Netherlands

K. EleftherakosSection of Zoology-Marine Biology, Department of Biology,University of Athens,Panepistimioupolis,157 84 Athens, Greece

Org Divers Evol (2011) 11:409–423DOI 10.1007/s13127-011-0058-y

distribution modelling (SDM) approaches (e.g. Martínez-Meyer et al. 2004; Pearman et al. 2008). Several SDMapproaches can be applied to characterize a species’environmental niche. The basic idea behind correlativeSDMs is to combine species records (longitude/latitudecoordinates of presence) with climatic information and, ifnecessary, add environmental data as predictors to develop aspatially explicit predictive model (Guisan and Zimmermann2000; Jeschke and Strayer 2008). SDMs have beensuggested to be useful in assessing whether a species’potential distribution might increase or decrease underclimate change (e.g. Araújo et al. 2004, 2006; Rödder et al.2009a) or whether it may even be driven to extinction(Thomas et al. 2004; Wake and Vredenburg 2008).

Projecting SDMs through space and time inherentlyassumes that the environmental niche of a species remains‘stable’ over time (Heikkinen et al. 2006; Jeschke andStrayer 2008). This basic assumption of niche stasis hasbeen challenged recently by mismatches between invasivespecies’ realised niches in their native and invasive ranges(e.g. Broenimann et al. 2007; Fitzpatrick et al. 2007;Beaumont et al. 2009). These observations are beingdiscussed controversially since they may either be tracedback to evolutionary change or may represent methodicalartefacts (Fitzpatrick et al. 2008; Pearman et al. 2008;Peterson and Nakazawa 2008; Rödder and Lötters 2009).Therefore, it has been advocated that a better understandingof the degree of niche conservatism among related speciesor among native and invasive ranges of one species shouldbe tested with SDMs by applying niche overlap indices (e.g.Warren et al. 2008; Rödder and Lötters 2009; Rödder et al.2009b; Broennimann et al. 2011; Rödder and Engler 2011).

Apart from issues related to parameter selection andspatial properties of species records (Dormann et al. 2007;Graham et al. 2004; Peterson and Nakazawa 2008; Phillips2008; Phillips et al. 2009), the validity of a SDM necessaryfor such tests depends on the question of whether thenumber of species records used for modelling providesenough statistical power (Hernandez et al. 2006; Schwartzet al. 2006; Wisz et al. 2008). The minimum number ofspecies records necessary depends on the applied SDMalgorithm and the ecological traits of the species at hand, i.e. ifthe species occupies a broad or narrow environmental niche(Wisz et al. 2008). Among algorithms requiring presence/pseudo-absence data or presence-only data, Maxent (Phillipset al. 2006; Phillips and Dudík 2008) has been shownfrequently to outperform other methods (see Elith et al. 2006;Hernandez et al. 2006; Wisz et al. 2008; Giovanelli et al.2010). Maxent also serves best among available algorithmswhen the number of species records is low (Hernandez et al.2006; Pearson et al. 2007; Wisz et al. 2008). However, insuch cases, predictions of potential distributions should beregarded as guidance for further field surveys rather than

being interpreted as a robust estimation of a species’potential distribution (Pearson et al. 2007; Weinsheimer etal. 2010). This raises an additional problem. Species knownnaturally from only small distribution areas are less suited toSDM applications due to the low number of availablespecies records. However, such species are often expected tobe especially vulnerable to anthropogenic climate change (e.g.Thomas et al. 2004; Thuiller et al. 2005; Schwartz et al.2006) and thus are being focussed on by conservationbiology. In other words, prognostic applications, which arecommonly performed with SDMs, are required for suchspecies.

Although similarity in realized niches can be measuredusing niche overlap indices (e.g. Warren et al. 2008;Broennimann et al. 2011; Rödder and Engler 2011), anotherdilemma is that the proposed techniques are aimed at singlespecies and cannot be applied reliably when the number ofpresence records is limited (Hernandez et al. 2006; Wisz etal. 2008). In this paper, we test whether climatic nichesimilarity can be identified by pooling presence data of allspecies of a monophyletic group inhabiting small geo-graphic ranges into a single concatenated data set andsubsequently jackknifing single species (i.e. a repeated‘leave-one-species-out’ method). We expect that, if ajackknifed species differs markedly in its climatic nichefrom all other species, this approach will result in increasedniche homogeneity wherein significance can be assessedusing null models.

We apply this approach to Lycian salamanders(Lyciasalamandra; Amphibia: Salamandridae), whichcomprise a monophyletic group of seven viviparousspecies. Lycian salamanders are distributed narrowly inrelatively small patches within a strip approximately30 km broad along the Mediterranean Turkish coast andsome adjacent Aegean islands (Fig. 1). With only oneexception (L. helverseni), Lyciasalamandra populationsare restricted to boulder fields at the foot of carsticlimestone formations where the salamanders live partiallyunderground (e.g. Steinfartz and Mutz 1998; Veith et al.2001; Veith and Steinfartz 2004). They are apparently‘trapped’ in these localized habitats, which are usuallysurrounded by xeric landscapes lacking the extensivesystems of crevices that are needed to escape from theoverall xeric habitat conditions throughout major parts ofthe year (Weisrock et al. 2001; Eleftherakos et al. 2007).Consequently, the number of available presence records islow for each of the seven species. Due to the smallgeographic range encompassed by these amphibians, weexpect highly restricted climatic niches within the entiregenus Lyciasalamandra. Analyzing the degree of climaticniche similarity among all seven Lycian salamanderspecies using our novel approach, we study the relativeexplanative power of climatic predictors for the spatial

410 D. Rödder et al.

distribution of these salamanders and link these predictorsto natural history properties of the species.

Materials and methods

Species records

For all seven species of Lyciasalamandra, only 80 geo-referenced presence records are available (Veith et al. 2001;Johannesen et al. 2006; Eleftherakos et al. 2007; Beukemaet al. 2009; F. Pasmans and S. Bogaerts, personalcommunication; J. Speybroeck, personal communication;M.Ö., unpublished data). None of the species fits the 30-record criterion for a meaningful SDM (Hernandez et al.2006), with 11 records obtained for Lyciasalamandraantalyana, 8 for L. atifi, 6 for L. billae, 15 for L. fazilae,

3 for L. flavimembris, 15 for L. helverseni and 22 for L.luschani (Fig. 1a).

Environmental data

Information on current climate within the region wasobtained from the Worldclim database (version 1.4) with agrid cell resolution of 30 arc sec (Hijmans et al. 2005a).Climate data include monthly mean variables of minimumand maximum temperature and precipitation for the period1950–2000. From these data, bioclimatic variables werecalculated with DIVA-GIS 7.3.0 (Hijmans et al. 2005b).These bioclimatic variables are more suitable predictorsthan monthly values and have been used widely in SDMapproaches (Nix 1986; Busby 1991; Beaumont et al. 2005).In order to avoid ‘over-fitting’ and problems caused bymulticolinearity of predictor variables (Heikkinen et al.

Fig. 1 a Presence of carsticlimestone formations (lightgrey) and species records usedfor species distribution model-ling (SDM). White dotsLyciasalamandra antalyana,black dots L. atifi, black squaresL. billae, grey squares L. fazilae,black triangles L. flavimembris,white triangles L. helverseni,grey triangles L. luschani. bMean of 100 Maxent SDMscomputed with pooled speciesrecords of all Lyciasalamandraspecies. c Maxent SDMs trainedwith all pooled records omittingL. helverseni. Areas climaticallysuitable for salamanders andalso suitable due to the presenceof carstic limestone formationsare indicated. Dark grey Maxentvalues above the minimumtraining presence, black valuesabove the minimum 10% train-ing omission threshold. Areasclimatically unsuitable to Lyciansalamanders are indicated inlight grey where carsticlimestone formations are presentand white where they are not

Climatic niche similarity in salamanders 411

2006), it is useful to restrict the number of variables tothose that are biologically meaningful for the study speciesat hand. This can greatly improve SDM results (e.g. Rödderet al. 2009a). Acknowledging published natural historyproperties of Lyciasalamandra (following Özeti and Atatür1979; Klewen 1991; Steinfartz and Mutz 1998; Veith et al.2001; Gautier et al. 2006; Tok et al. 2009), we selected fivebioclimatic variables as informative predictors for SDMbuilding: ‘mean temperature of the wettest quarter’ (BIO8),‘mean temperature of the coldest quarter’ (BIO11), ‘annualprecipitation’ (BIO12), ‘precipitation of the wettest quarter’(BIO16) and ‘precipitation of the coldest quarter’ (BIO19).These variables reflect those climatic conditions to whichLyciasalamandra species are exposed during their overgroundactivity periods. We consider them crucial for long-termpersistence of populations.

Since, as mentioned above, Lycian salamanders occuralmost exclusively on boulder fields at the foot of carsticlimestone formations, we used a shapefile provided by theCentral Energy Resource Team of the US GeologicalSurvey (http://energy.cr.usgs.gov/oilgas/wep/; downloaded20 August 2009) to highlight these geomorphologicalproperties.

SDM computation

For SDMbuilding, we usedMaxent 3.3.3e (Phillips et al. 2006;Phillips and Dudík 2008; Elith et al. 2011). This softwareemploys a machine-learning algorithm for SDMs withenvironmental predictors and commonly reveals better resultsthan comparable methods (e.g. Elith et al. 2006; Hernandez etal. 2006; Wisz et al. 2008; Giovanelli et al. 2010), especiallywhen assessing niche overlaps (Broennimann et al. 2011).When training Maxent, the algorithm calculates a probabilitydistribution over the study area that satisfies a set ofconstraints derived from environmental conditions detectedat species records. Subsequently, Maxent chooses a distribu-tion closest to uniform by maximizing entropy (Jaynes 1957)within all distributions fulfilling these conditions (Phillips etal. 2006; Phillips and Dudík 2008; Elith et al. 2011). The finaloutput is a map interpreted as the potential distribution of thestudied species derived from its realized niche. Maxent usesrandom background points as contrast for model computation(‘pseudo-absence’). Ideally, these are obtained within an areathat is accessible and can be potentially colonised by the studyspecies (Phillips 2008). The background points we used werechosen randomly within the general area of occurrence of thegenus Lyciasalamandra, enclosing all species records in aminimum rectangular polygon.

Araújo and New (2007) suggested that ensembles ofmodels may help to evaluate model uncertainties and thatthe combined results of multiple models may provide morerobust predictions. We therefore computed 100 Maxent

models each using the procedure described above based oneither all species records or all records with one speciesbeing omitted, whereby each of these SDMs was developedwith 25% of the records chosen randomly omitted frommodel training and used as test points (Phillips et al. 2006).Subsequently, the average of all 100 models automaticallycomputed by Maxent was taken for further processing andthe standard deviation per grid cell was assessed in order toevaluate the degree of uncertainty among different areas.All runs were conducted using the default values for allprogram settings. Using the logistic output format, continuousprobability surfaces ranging from 0 (unsuitable) to 1(suitable), were transformed into presence/absence mapsfor illustrative purposes using two different thresholdsfollowing Liu et al. (2005), i.e. the minimum trainingpresence and the 10 percentile training omission thresholds.

To test for model performance, Maxent allows for modeltesting by calculation of the area under the receiveroperation characteristics curve (AUC) based on trainingand test data (Fielding and Bell 1997; Phillips et al. 2006),with AUC values range from 0.5 (no better discriminationthan random) to 1.0 (perfect discrimination). Training AUCvalues represent the ability of the model to distinguishpresence data from random background data, whereby testAUC values represent the model’s ability to predictrandomly chosen test points omitted from model training(herein 25%) (Phillips et al. 2006).

Test for niche similarity

To compare quantitatively the salamander species’ potentialdistributions derived from their realized climatic niches, weused Schoener’s index for niche overlap (D; Schoener1968). This index describes the degree of similarity ofpotential distributions by comparing corresponding valuesper cell of two grids and range from 0 (no similarity) to 1(two grids are identical). First, we computed niche identitytests as proposed by Warren et al. (2008, 2010) between allspecies with more than ten records, i.e. Lyciasalamandraantalyana, L. fazilae, L. helverseni and L. luschani. Toovercome the limited availability of species records in theremaining species, in a second step we pooled all records ofthe seven Lyciasalamandra species (nall) in a jackknifingapproach. We calculated D and 1−D values based on allrecords compared to nall-speciesx (repeated ‘leave-one-species-out’). Significance of D was evaluated using null-models; 100 SDMs were computed based on nall minus anumber of records randomly selected across all species inthe amount of speciesx. D values obtained by exclusion ofrecords of speciesx were compared to the null distribution.This allowed us to test whether the observed degree ofoverlap between the climatic niche of the omitted speciesand all other species was significantly more different. We


http://energy.cr.usgs.gov/oilgas/wep/

expected that if a species occupies a niche different to thatoccupied by the other species of this particular group,omitting it from the entire data set will result in a strongerdecrease in the observed overlap than expected by chance.To summarize the P-values of all tests into a single measureof statistical significance, we applied Fisher’s omnibus test(Haccou and Meelis 1994), adjusting the probability ofalpha-errors. 1−D scores allow a relative ranking of thespecies, with higher values indicating more divergentniches. However, this method requires that the entireenvironmental space suitable for a species is representedin its records, which perhaps is easily warranted in speciesinhabiting small geographic ranges such as Lycian sala-manders. When interpreting the results, it is pivotal toacknowledge the different questions asked by differenttests, i.e. the test design proposed previously by Warrenet al. (2008) and our novel jackknifing approach (cf.Fig. 2). Hence, Schoener’s D and 1−D scores should beused as a ranking tool within one data set only andabsolute scores are not necessarily comparable acrossdifferent studies.

Results

Following Swets (1988), we derived high AUC values ineach of the 100 computed SDMs based on either onlyspecies-specific, all pooled records or all records minus onespecies. The only exception was Lyciasalamandra ant-alyana (Figs. 3, 4). Training AUC values were slightlybetter than test AUC values, with the exception thatomission of L. atifi was not significantly different. Thisindicates an absence or low degree of over-fitting. The‘mean temperature of the coldest quarter’ (BIO11) followedby the ‘precipitation of the wettest quarter’ (BIO16) had thehighest explanatory power, whereas the other variablescontributed less than 10% in SDMs, except in those inwhich L. helverseni was omitted (Fig. 3) or when only L.luschani records were used for model training (Fig. 4).Here, the contribution of the ‘precipitation of the wettestquarter’ (BIO16) was much lower and the ‘precipitation ofthe coldest quarter’ (BIO19) more important. Variations ineach threshold were negligible between all model runs (fordetails see Figs. 3, 4).

In the climate space spanning the two most importantenvironmental variables (Figs. 3, 4), all species but L.helverseni overlapped considerably, with only some mar-ginal species records falling outside a core area at 350–620 mm in the ‘precipitation of the coldest quarter’(BIO19) and at 6.5–12°C in the ‘mean temperature of thecoldest quarter’ (BIO11) (Fig. 5). This indicates that muchof the available climatic space actually is exploited byLyciasalamandra, despite a few areas with a ‘mean

temperature of the coldest quarter’ (BIO11) below 3°C.As a consequence, within its western and eastern margins,the realised distribution of the genus matches its potentialdistribution, when mapped SDMs are overlaid with carsticlimestone formations to which these amphibians areassociated (Veith et al. 2001, 2008). As seen in Fig. 1,only a few carstic mountain blocks near the coast areapparently not inhabited by Lycian salamanders.

When assessing niche overlap between two species asproposed by Warren et al. (2008), Schoener's D valuesranged from 0.14 to 0.65. According to the classificationproposed by Rödder and Engler (2011), we detected ‘no orvery limited’ to ‘low’ overlap when L. helverseni was

Pro

babi

lity

Pro

babi

lity

Pro

babi

lity

Environmental gradient

Pro

babi

lity

Pro

babi

lity

Species 1Species 2Species 3

1 2

Overlap species 1/2

Pair-w

ise niche overlap

3

Overlap species 1/3

Overlap all / all - species 3

Overlap all / all - species 2

Jackknife niche overlap

Fig. 2 Conceptual differences among proposed niche overlapanalyses. Imagine a set of three species with different preferencesalong an environmental gradient. Using SDM projections, it ispossible to derive for each species a probability distribution acrossthe gradient. Comparing two species as proposed by Warren et al.(2008), the overlap (grey area in the middle panel) of their respectiveprobability distributions in geographic space is computed (pair-wiseniche overlap). In our jackknife approach, a probability distributionderived from all species records is compared to a probabilitydistribution derived from all minus one species. The 1−the resultingoverlap value reflects the relative contribution of the omitted speciesto the entire probability distribution (grey area in the lower panel) andcan be used as measure to rank all species when omitting themiteratively. Note that these specific indices are comparable only acrosseach method but not among the different approaches


All

Spe

cies

With

out

L. a

nta

lyan

aW

ithou

t L

. ati

fiW

ithou

t L

. bill

ae

Fig. 3 Characterization of 100 Maxent models either trained withrecords of all species or omitting one species in terms of training AUC(train) and test AUC (test; 75% training/25% test) (left side), variableimportance (BIO8 = ‘mean temperature of the wettest quarter’; BIO11 =

‘mean temperature of the coldest quarter’; BIO12 = ‘annual precipita-tion’; BIO16 = ‘precipitation of the wettest quarter’; BIO19 =‘precipitation coldest quarter’) (middle) and minimum training presence(Min) and minimum 10 percent training presence (10%) (right side)


involved in the comparisons, but ‘moderate overlap’ amongall other species. The niche identity tests indicate that most

species except L. antalyana/L. luschani differ significantlyin their climatic niches (Table 1).

With

out

L. f

azill

aeW

ithou

t L

. fla

vim

emb

ris

With

out

L. h

elve

rsen

iW

ithou

t L

. lu

sch

ani

With

out

Fig. 3 (continued)


Applying our jackknife approach to realise climatic nicheoverlap tests revealed high values in terms of Schoener’s D(i.e. > 0.8) and low deviances (1−D ≤0.13) between thepotential distributions derived from pooled species records,as well as those derived from pooled species records witheach one species omitted (Table 2; Figs. 1b,c, 6, 7). Thissuggests that L. helverseni is the most deviant of all

Lyciasalamandra species, followed by L. atifi. Looking atthe null-models, we obtained non-significant results for L.antalyana, L. atifi, L. fazilae, L. helverseni and L. luschaniand significant ones when L. billae or L. flavimembris wereomitted (Table 2). Overall, the Fisher omnibus test revealed ageneral effect of niche differentiation in Lycian salamanders(χ2=25.33, df=14, P=0.0314).

L. a

ntal

yana

L. fa

zila

eL.

hel

vers

eni

L. lu

scha

ni

0.5

0.6

0.7

0.8

0.9

1

AU

C

0.5

0.6

0.7

0.8

0.9

1

AU

C

0.5

0.6

0.7

0.8

0.9

1

AU

C

Train AUC Test AUC

Train AUC Test AUC

Train AUC Test AUC

Train AUC Test AUC

0.5

0.6

0.7

0.8

0.9

1

AU

C

BIO8 BIO11 BIO12 BIO16 BIO190

20

40

60

80

100


20

40

60

80

100


20

40

60

80

100


20

40

60

80

100

Var

iab

le C

on

trib

uti

on

[%

]V

aria

ble

Co

ntr

ibu

tio

n [

%]

Var

iab

le C

on

trib

uti

on

[%

]V

aria

ble

Co

ntr

ibu

tio

n [

%]

Fig. 4 Characterization of 100Maxent models trained withspecies-specific records in termsof training AUC (train) and testAUC (test; 75% training/25%test) and variable importance(BIO8 = ‘mean temperature ofthe wettest quarter’; BIO11 =‘mean temperature of the coldestquarter’; BIO12 = ‘annualprecipitation’; BIO16 =‘precipitation of the wettestquarter’; BIO19 = ‘precipitationcoldest quarter’)


Discussion

Previous studies have found that SDMs may becomeunreliable when too few species records are available formodel training (Hernandez et al. 2006; Wisz et al. 2008), asevident in L. antalyana indicated by low AUC scores(Fig. 4). Hence, these results need to be interpreted withcaution, and niche overlap estimates for species inhabitingvery small ranges are not possible using standard methods.Therefore, here we applied, for the first time, a jackknife

approach to test for among-species niche similarity prior tocomposite modelling of potential distributions. If signifi-cant niche similarity can be proved, pooling of distributiondata of several species would be possible and allow SDMprediction even if the data base for each single species isinsufficient. Our approach clearly identified a high degreeof niche similarity among Lyciasalamandra species, withonly one (L. helverseni) covering a divergent climatic niche(Tables 1, 2; Fig. 5). This pattern was also confirmed whencomputing niche overlaps and identity tests as proposed byWarren et al. (2008).

Climatic niche conservatism among Lycian salamanderspecies contrasts with the repeatedly changing climatichistory of the Mediterranean region. Molecular data suggestthat Lycian salamanders started radiating at least 5.9 mya(7.9–5.9 mya, as estimated by Weisrock et al. 2001; 20.9–12.6 mya, as estimated by Zhang et al. 2008). Since thattime, the climate has oscillated repeatedly, with cold periods(stitials) alternating with warmer periods (interstitials)(Andersen and Borns 1994). During cold periods, circum-Mediterranean habitats have been remarkably dry due tothe huge amounts of water bound in the immense polar iceshields and worldwide glaciers (Wilson et al. 1999).

100

200

300

400

500

600

700

-5 -3 -1 1 3 5 7 9 11 13 15

Pre

cipi

tatio

n of

the

cold

est q

uart

er [m

m]

Mean temperature of the coldest quarter [°C]

Background L. antalyana L. atifi L. billae L. fazilae L. flavimembris L. helverseni L. luschani

Fig. 5 Relative position of species records in environmental niche space. Note the decreasing number of species records with decreasingtemperature and precipitation. This pattern corroborates well field observations by Klewen (1991) and Steinfartz and Mutz (1998)

Table 1 Niche overlaps in terms of Schoener’s index (D) amongLyciasalamandra species with > 10 species records

Species pair Overlap Significance level

L. antalyana - L. fazilae 0.65 0.05*

L. antalyana - L. helverseni 0.31 < 0.01*

L. antalyana - L. luschani 0.62 0.88

L. fazilae - L. helverseni 0.14 < 0.01*

L. fazilae - L. luschani 0.57 0.02*

L. helverseni - L. luschani 0.17 < 0.01*

* P>0.05


Therefore, southern Anatolian environments wheretoday’s Lyciasalamandra populations live were possiblyeven more hostile for amphibians than they are today.

Usually, species distributions tend to shift withchanging climate, either latitudinally or longitudinally(Parmesan and Yohe 2003; Root 2003; Parmesan 2006),so that survival of a species depends mainly on its optionsto “move out of trouble”. The extent to which the spatialdistribution of a species can decrease or increase dependson its specific dispersal abilities in concert with thecontemporary landscape connectivity (e.g. Araújo et al.2004; 2006; Rödder et al. 2009a). Since the landscapeadjacent to most known localities of Lycian salamanders isnot only dry and hot, but also lacks deep-reachingcrevices, dispersal of these strictly terrestrial, moisture-dependent animals must have been severely hampered oreven impossible during major phases of Late Pliocene andPleistocene climate oscillations.

The lack of options to react to climate change with rangeshifts inevitably calls for either high niche plasticity or forevolutionary niche shifts to parallel deteriorating environ-

Fig. 6 Mean of 100 MaxentSDMs computed with pooledspecies records of all Lyciasala-mandra species omitting L.antalyana (a), L. atifi (b), andL. billae (c). Areas climaticallysuitable for salamanders andsuitable due to the presence ofcarstic limestone formations areindicated (dark grey Maxentvalues above the minimumtraining presence, black valuesabove the minimum 10%training omission threshold).Areas climatically unsuitable toLycian Salamanders are indicatedin light grey where carstic lime-stone formations are present andwhite where they are not

Table 2 Niche overlap in terms of Schoener’s index (D) of mappedSDMs of all seven Lyciasalamandra species combined against all butone species; each model excludes another species. 1−D allows a relativeranking of the absolute niche deviance per species from all other species

Entire species clade but excluding… D 1−D P

L. antalyana 0.93 0.07 0.10

L. atifi 0.91 0.09 0.44

L. billae 0.96 0.04 < 0.01*

L. fazilae 0.92 0.08 0.39

L. flavimembris 0.96 0.04 0.05*

L. helverseni 0.87 0.13 0.90

L. luschani 0.92 0.08 0.41

* P>0.05


mental conditions. Due to the pronounced climatic nicheconservatism among Lyciasalamandra species as shown inthis study, and due to the lack of any options for alatitudinal and longitudinal response, we assume that theseamphibians were forced into an in situ response. Naturalhistory properties may provide the key to understanding thesuccess of Lycian salamanders. Today’s climate at siteswhere Lycian salamanders occur is highly unfavourable foramphibians throughout almost all the year (e.g. Klewen1991; Polymeni 1994; Steinfartz and Mutz 1998; Veith etal. 2001). As a result, extant Lyciasalamandra specimensspend substantial parts of their lives deep inside humidcrevices of boulder fields at the foot of carstic limestoneslopes. In these crevices, unfavourable above-groundmacroclimatic conditions are effectively buffered. Theactivity period of specimens outside their limestone shelteris therefore restricted to autumn and winter, when humidity

is relatively high and the air temperature is below 20°C(Steinfartz and Mutz 1998). Captive breeding has con-firmed these physiological constraints (Bogaerts 2004),which were also well identified by our SDM approachwherein the ‘mean temperature of the coldest quarter’ andthe ‘mean precipitation of the wettest quarter’ bestexplained the potential distribution of all Lycian salaman-ders. It can therefore be concluded that a temporaryadoption of an underground, cavernicole lifestyle partiallydecouples Lycian salamanders from their macroclimate.The latter defines species conditions for above-groundactivity for Lyciasalamandra, as confirmed by the relativelyvariable contributions in our SDMs. Those variablesdescribing the macroclimate within the species above-ground activity periods had the highest explanative power(Fig. 3), suggesting that they in fact restrict the realiseddistributions of the species.

Fig. 7 Mean of 100 MaxentSDMs computed with pooledspecies records of all Lyciasala-mandra species omitting L.fazilae (a), L. flavimembris (b),and L. luschani (c). Areasclimatically suitable for sala-manders and suitable due tothe presence of carstic limestoneformations are indicated(dark greyMaxent values abovethe minimum training presence,black values above the mini-mum 10% training omissionthreshold). Areas climaticallyunsuitable to Lycian salamandersare indicated in light grey wherecarstic limestone formationsare present and white where theyare not


Lyciasalamandra—a special case?

Holt et al. (2005) hypothesized that isolated populationsliving in habitats close to their environmental tolerancelimits may be most prone to evolving new physiologicalproperties that result in a shift of their fundamental niche.Since population densities in Lycian salamanders are oftenremarkably high (5,000–10,000 specimens per hectare;Veith et al. 2001), it can be expected that perhaps a highnumber of specimens from these source populationsmigrates to sink habitats just outside their climatictolerance. This should theoretically enhance the chance ofselection of novel adaptations (e.g. Holt et al. 2005; Jakobet al. 2010). However, this is contraindicated by ouranalysis, which showed that the realised niches of mostspecies are quite similar (Tables 1, 2), although populationslive from 0 to 1,000 m above sea level (Veith et al. 2001).One explanation may be that environmental conditionsduring summer times are too harsh to allow colonisation inthe absence of the retreat possibilities provided by carsticlimestone cavities. These harsh conditions may cause astabilising selection on the species’ climate niches byrestricting their range.

It should be noted that the detected niche constancyacross species does not necessarily indicate that mostLyciasalamandra species exhibit a climatic niche that isclose to that of a common ancestor, i.e. one that has notchanged much during evolution. Alternatively, all speciesmay have changed their niches in the same direction due toa pronounced similarity of climatic parameters of theirrespective habitats. Unfortunately, there is no a priorireason to decide between these two hypotheses. However,genetic variability, which is thought to facilitate niche shiftthrough local level adaptation (Holt and Gomulkiewicz2004; Holt et al. 2005; Jakob et al. 2010) is rather low inLyciasalamandra (Veith et al. 2008). This may advocate fora hampered evolutionary response to changing climate,and hence argues for niche conservatism over time inLyciasalamandra.

Methodological implications and limitations

Some aspects need consideration when applying our noveljackknife approach. First of all, although Maxent has beenshown to perform well when the available number ofspecies records is rather limited (Elith et al. 2006;Hernandez et al. 2006; Wisz et al. 2008), a great varietyof SDM algorithms are available, and applying differentalgorithms may yield varying results (e.g. Araújo and New2007). Selection of the appropriate algorithm may dependon the data at hand, i.e. the number of records and whetherabsence data is available or not (Guisan and Zimmermann2000). However, in the case of Lycian salamanders, Maxent

is apparently the best choice since it frequently outperformsother methods when the sample size is low.

Our approach most easily identifies species that differmost from all others analysed. However, depending on theniche breadths and niche positions of the species at hand,one could imagine a variety of possible scenarios (Fig. 2),which can be assessed in niche space using multivariatestatistical methods. Imagine that some species are found incolder climates and a couple of species found in hotterenvironments. Pooling all species but one will perhapsresult in a quite large overall niche breadth and, dependingon the relative positions of the species omitted in nichespace, niche similarity patterns may vary. A secondscenario may describe a species that occupies a niche spacenested in the middle of another species’ niches. Dependingon the algorithm applied, the overall niche envelope maynot change too much when removing it, even though it canoccupy a niche space distinct from all other species.

Fortunately, these difficulties can be circumvented viacomparisons of probability surfaces derived from SDMs.The relative occurrence probabilities at a given combinationof environmental variables depend on the density of speciesrecords in niche space (Fig. 2). As a consequence, multipleobservations of different species and/or within singlespecies in similar or identical parts of the ecological spacelead to locally higher probabilities. Therefore, even in theunlikely case that two species occupy exactly the same partof the ecological space, removal of one of them will cause alocal decrease in the probability, and a comparison of bothprobability distributions via Schoener’s D will still identifya deviance. In order to detect such patterns, interpretation ofresults should always include inspection of the relativepositions of the species in niche space spanning along themost important axes and the overall available niche space(e.g. as shown in Fig. 3 for a two-dimensional climate space).Note that in more complex cases, principal component plotsmay be used (e.g. as shown in Broennimann et al. 2007;2011). These additional analyses become most importantwhen assessing niche conservatism in a phylogeneticcontext. Here, we may consider jackknifing entire cladesinstead of single species in order to obtain a strongersignal.

Further application of our jackknife approach to a varietyof taxa will need to show if the identification of specieswith deviant climatic niches and their subsequent omissionfrom a pooled data set will help to derive robust SDM or,more accurately, clade distribution models (CDM), even inthe absence of large sets of distribution data. Facing theongoing habitat deterioration due to climate changeurgently requires reliable tools for incorporation of predictivedistribution analyses for conservation planning, especially ingeographically restricted and thus potentially vulnerablespecies.


Acknowledgements We are grateful to Dan Warren, who helped uswith statistics and provided us with the Perl script referred to inWarren et al. (2008). The manuscript benefitted from the valuablecomments of G. Francesco Ficetola and Jan O. Engler. Part of ourwork was funded by the 'Forschungsinitiative' of the Ministry ofEducation, Science, Youth and Culture of the Rhineland-Palatinatestate of Germany 'Die Folgen des Global Change für Bioressourcen,Gesetzgebung und Standardsetzung'. Thanks to Jerome Speybroeckfor providing unpublished distribution records and Ursula Bott forvaluable suggestions.

References

Andersen, B. G., & Borns, H. W. (1994). The ice age world. Oslo:Scandinavia University Press.

Araújo, M. B., & New, M. (2007). Ensemble forecasting of speciesdistributions. Trends in Ecology & Evolution, 22, 42–47.

Araújo, M. B., Cabeza, M., Thullier, W., Hannah, L., & Williams, P.H. (2004). Would climate change drive species out of reserves?An assessment of existing reserve-selection methods. GlobalChange Biol, 10, 1618–1626.

Araújo, M. B., Thuiller, W., & Pearson, R. G. (2006). Climatewarming and the decline of amphibians and reptiles in Europe.Journal of Biogeography, 33, 1712–1728.

Beaumont, L. J., Hughes, L., & Poulsen, M. (2005). Predicting speciesdistributions: use of climatic parameters in BIOCLIM and itsimpact on predictions of species' current and future distributions.Ecological Modelling, 186, 250–269.

Beaumont, L. J., Gallagher, R. V., Thuiller, W., Downey, P. O.,Leishman, M. R., & Hughes, L. (2009). Different climaticenvelopes among invasive populations may lead to under-estimations of current and future biological invasions. Diversityand Distributions, 15, 409–420.

Beukema, W., de Pous, P., & Brakels, P. (2009). Remarks on themorphology and distribution of Lyciasalamandra luschanifinikensis with the discovery of a new isolated population.Zeitschrift für Feldherpetologie, 16, 115–127.

Bogaerts, S. (2004). Temperature tolerance of captive salamandersduring a heat wave. Podarcis, 5, 15–22.

Broennimann, O., Treier, U. A., Müller-Schärer, H., Thuiller, W.,Peterson, A. T., & Guisan, A. (2007). Evidence of climatic nicheshift during biological invasion. Ecology Letters, 10, 701–709.

Broennimann, O., Fitzpatrick, M. C., Pearman, P. B., Petitpierre, B.,Pellissier, L., Thuiller, W., et al. (2011). Measuring ecologicalniche overlap from occurrence and spatial environmental data.Global Ecology and Biogeography. doi:10.1111/j.1466-8238.2011.00698.x.

Busby, J. R. (1991). BIOCLIM—a bioclimatic analysis and predictionsystem. In C. R. Margules & M. P. Austin (Eds.), Natureconservation, cost effective biological surveys and data analysis(pp. 64–68). Melbourne: CSIRO.

Dormann, C. F., McPherson, J., Araújo, M. B., Bivand, R., Bollinger,J., Carl, G., et al. (2007). Methods to account for spatialautocorrelation in the analysis of species distributional data, areview. Ecography, 30, 609–628.

Eleftherakos, K., Sotiropoulos, K., & Polymeni, R. M. (2007).Conservation units in the insular endemic salamander Lyciasala-mandra helverseni (Urodela, Salamandridae). Annales ZoologiciFennici, 44, 387–399.

Elith, J., Graham, C. H., Anderson, R. P., Dudik, M., Ferrier, S., Guisan,A., et al. (2006). Novel methods improve prediction of species´distributions from occurrence data. Ecography, 29, 129–151.

Elith, J., Phillips, S. J., Hastie, T., Dudík, M., Chee, Y. E., & Yates, C.J. (2011). A statistical explanation of MaxEnt for ecologists.Diversity and Distributions, 17, 43–57.

Fielding, A. H., & Bell, J. F. (1997). A review of methods for theassessment of prediction errors in conservation presence/absencemodels. Environmental Conservation, 24, 38–49.

Fitzpatrick, M. C., Weltzin, J. F., Sanders, N. J., & Dunn, R. R.(2007). The biogeography of prediction error, why does theintroduced range of the fire ant over-predict its native range?Global Ecology and Biogeography, 16, 24–33.

Fitzpatrick, M. C., Dunn, R. R., & Sanders, N. J. (2008). Data setsmatter, but so do evolution and ecology. Global Ecology andBiogeography, 17, 562–565.

Gautier, P., Olgun, K., Üzüm, N., & Miaud, C. (2006). Gregariousbehaviour in a salamander: attraction to conspecific chemicalcues in burrow choice. Behavioral Ecology and Sociobiology, 59,836–841.

Giovanelli, J. G. R., Siqueira, M. F., Haddad, C. F. B., & Alexandrino,J. (2010). Modelling a spatially restricted distribution in theNeotropics, how the size of calibration area affects the perfor-mance of five presence-only methods. Ecological Modelling,221, 215–224.

Graham, C. H., Ron, S. R., Santos, J. C., Schneider, C. J., & Moritz,C. (2004). Integrating phylogenetics and environmental nichemodels to explore speciation mechanisms in dendrobatid frogs.Evolution, 58, 1781–1793.

Guisan, A., & Zimmermann, N. (2000). Predictive habitat distributionmodels in ecology. Ecological Modelling, 135, 147–186.

Haccou, P., Meelis, E. (1994). Statistical analyses of behavioural data.Oxford University Press.

Hawkins, B. A., Diniz-Filho, J. A. F., Jaramillo, C. A., & Soeller, S.A. (2006). Post-Eocene climate change, niche conservatism, andthe latitudinal diversity gradient of New World birds. Journal ofBiogeography, 33, 770–780.

Heikkinen, R. K., Luoto, M., Araújo, M. B., Virkkala, R., Thuiller,W., & Sykes, M. T. (2006). Methods and uncertainties inbioclimatic envelope modelling under climate change. ProgressPhysical Geography, 30, 751–777.

Hernandez, P. A., Graham, C. H., Master, L. L., & Albert, D. L.(2006). The effect of sample size and species characteristics onperformance of different species distribution modelling methods.Ecography, 29, 773–785.

Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G., & Jarvis, A.(2005a). Very high resolution interpolated climate surfaces forglobal land areas. International Journal of Climatology, 25,1965–1978.

Hijmans, R.J., Guarino, L., Jarvis, A., O'Brien, R., Mathur, P.,Bussink, C., Cruz, M., Barrantes, I., Rojas, E. (2005b). Diva-GIS version 5.2 manual. International Potatoe Center, Lima,Peru.

Holt, R. D., & Gomulkiewicz, R. (2004). Conservation implications ofniche conservatism and evolution in heterogeneous environ-ments. In R. Ferriere, U. Dieckmann, & D. Couvet (Eds.),Evolutionary conservation biology (pp. 244–264). Cambridge,UK: Cambridge University Press.

Holt, R.D., Barfield, M., Gomulkiewicz, R. (2005). Theories of nicheconservatism and evolution, could exotic species be potentialtests? In D. Sax, J. Stachowicz, S.D. Gaines, (Ed.), Speciesinvasions, insight into ecology, evolution, and biogeography.Sinauer, Sunderland, MA, pp. 259–290.

Jakob, S. S., Heibl, C., Rödder, D., & Blattner, F. R. (2010).Population demography influences climate niche evolution,evidence from diploid American Hordeum species (Poaceae).Molecular Ecology, 19, 1423–1438.

Jaynes, E. T. (1957). Information theory and statistical mechanics.Physics Review, 106, 620–630.


http://dx.doi.org/10.1111/j.1466-8238.2011.00698.x

http://dx.doi.org/10.1111/j.1466-8238.2011.00698.x

Jeschke, J. M., & Strayer, D. L. (2008). Usefulness of bioclimaticmodels for studying climate change and invasive species. Annalsof the New York Academy of Sciences, 1134, 1–24.

Johannesen, J., Johannesen, B., Griebeler, E. M., Baran, I., Tunc, M.R., & Veith, M. (2006). Distortion of symmetrical introgressionin a hybrid zone, evidence for locus-specific selection and uni-directional range expansion. Journal of Evolutionary Biology, 19,705–716.

Klewen, R. (1991). Die Landsalamander Europas, Teil 1. Die NeueBrehm-Bücherei, Hohenwarsleben.

Kozak, K., & Wiens, J. J. (2006). Does niche conservatism promotespeciation? A case study in North American salamanders.Evolution, 60, 2604–2621.

Kozak, K. H., Graham, C. H., & Wiens, J. J. (2008). Integrating GIS-based environmental data into evolutionary biology. Trends inEcology & Evolution, 23, 141–148.

Liu, C., Berry, P. M., Dawson, T. P., & Pearson, R. G. (2005).Selecting thresholds of occurrence in the prediction of speciesdistributions. Ecography, 28, 385–393.

Marténez-Meyer, E., Peterson, A. T., & Hargrove, W. W. (2004).Ecological niches as stable distributional constraints on mammalspecies, with implications for Pleistocene extinctions and climatechange projections for biodiversity. Global Ecology Biogeography,13, 305–314.

Nix, H. (1986). A biogeographic analysis of Australian elapid snakes.In R. Longmore (Ed.), Atlas of elapid snakes of Australia (pp. 4–15). Canberra: Bureau of Flora and Fauna.

Özeti, N., & Atatür, M. K. (1979). A preliminary survey of the serumproteins of a population of Mertensiella luschani finikensisBaşoğlu and Atatür, 1975 from Finike in southwestern Anatolia.Istanbul UniversityJournal of the Faculty of Science, 44B, 23–29.

Parmesan, C. (2006). Ecological and evolutionary responses to recentclimate change. Annual Review of Ecology, Evolution, andSystematics, 37, 637–669.

Parmesan, C., & Yohe, G. (2003). A globally coherent fingerprint ofclimate change impacts across natural systems. Nature, 421, 37–42.

Pearman, P. B., Guisan, A., Broennimann, O., & Randin, C. F. (2008).Niche dynamics in space and time. Trends in Ecology &Evolution, 23, 149–158.

Pearson, R. G., Raxworthy, C. J., Nakamura, M., & Peterson, A. T.(2007). Predicting species distributions from small numbers ofoccurrence records, a test case using cryptic geckos in Madagascar.Journal of Biogeography, 34, 102–117.

Peterson, A. T., & Nakazawa, Y. (2008). Environmental data setsmatter in ecological niche modelling, an example with Solenopsisinvicta and Solenopsis richteri. Global Ecology & Biogeography,17, 135–144.

Peterson, A. T., Soberon, J., & Sánchez-Cordero, V. (1999).Conservation of ecological niches in evolutionary time. Science,285, 1265–1267.

Phillips, S. J. (2008). Transferability, sample selection bias andbackground data in presence-only modelling, a response toPeterson et al. (2007). Ecography, 31, 272–278.

Phillips, S. J., & Dudík, M. (2008). Modelling of species distributionswith Maxent, new extensions and comprehensive evaluation.Ecography, 31, 161–175.

Phillips, S. J., Anderson, R. P., & Schapire, R. E. (2006). Maximumentropy modelling of species geographic distributions. EcologicalModelling, 190, 231–259.

Phillips, S. J., Dudík, M., Elith, J., Graham, C. H., Lehmann, A.,Leathwick, J., et al. (2009). Sample selection bias andpresence-only distribution models, implications for back-ground and pseudo-absence data. Ecological Applications,19, 181–197.

Polymeni, R. M. (1994). On the biology of Mertensiella luschani(Steindachner, 1981): a review. Mertensiella, 4, 301–314.

Root, T. L., Price, J. T., Hall, K. R., Schneider, S. H., Rosenzweig, C., &Pounds, A. (2003). Fingerprints of global warming on wild animalsand plants. Nature, 421, 57–60.

Rödder, D., & Engler, J.O. (2011). Quantitative metrics of overlaps inGrinnellian niches, advances and possible drawbacks. GlobalEcology & Biogeography doi:10.1111/j.1466-8238.2011.00659.x

Rödder, D., & Lötters, S. (2009). Niche shift versus niche conserva-tism? Climatic characteristics within the native and invasiveranges of the Mediterranean housegecko (Hemidactylus turcicus).Global Ecology & Biogeography, 18, 674–687.

Rödder, D., Schlüter, A., & Lötters, S. (2009a). Is the 'Lost World'lost? High endemism of the South American tepuis in a changingclimate. In J. C. Habel & T. Assmann (Eds.), Relict species.Phylogeography and conservation biology (pp. 401–416). Berlin:Springer.

Rödder, D., Schmidtlein, S., Veith, M., & Lötters, S. (2009b). Alieninvasive slider turtle in unpredicted habitat, a matter of nicheshift or of predictors studied? PloS One, 4, e7843.

Schoener, T. W. (1968). Anolis lizards of Bimini: resource partitioningin a complex fauna. Ecology, 49, 704–726.

Schwartz, M. W., Iverson, L. R., Prasad, A. M., Matthews, S. N., &O’Connor, R. J. (2006). Predicting extinctions as a result ofclimate change. Ecology, 87, 1611–1615.

Steinfartz, S., & Mutz, T. (1998). Mertensiella luschani (Steindachner,1891) - Lykischer Salamander, Kleinasiatischer Salamander, InK. Grossenbacher, B. Thiesmeier (Ed.), Handbuch der Reptilien undAmphibien Europas, vol. 4/1, Schwanzlurche. Aula: Wiesbaden,pp. 367–397.

Swets, K. (1988). Measuring the accuracy of diagnostic systems.Science, 240, 1285–1293.

Thomas, C. D., Cameron, A., Green, R. E., Bakkenes, M., Beaumont,L. J., Collingham, Y. C., et al. (2004). Extinction risk fromclimate change. Nature, 427, 145–148.

Thuiller, W., Lavorel, S., Araújo, M. B., Sykes, M. T., & Prentice, I.C. (2005). Climate change threats to plant diversity in Europe.Proceedings of the National Academy of Science USA, 102,8245–8250.

Tok, C. V., Mosungolu, M., Ayaz, D., Cicek, K., & Gül, C. (2009).Hematology of the Lycian salamander, Lyciasalamandra fazilae.North-West Journal of Zoology, 5, 321–329.

Veith, M., & Steinfartz, S. (2004). When non-monophyly results intaxonomic consequences - the case of Mertensiella within theSalamandridae (Amphibia, Urodela). Salamandra, 40, 67–80.

Veith, M., Baran, I., Godmann, O., Kiefer, A., Öz, M., & Tunc, M. R.(2001). A revision of population designation and geographicdistribution of Mertensiella luschani (Steindachner, 1891).Fauna in the Middle East, 22, 67–82.

Veith, M., Lipscher, E., Öz, M., Kiefer, A., Baran, I., Polymeni, R. M.,et al. (2008). Cracking the nut, geographical adjacency of sistertaxa supports vicariance in a polytomic salamander clade in theabsence of node support. Molecular Phylogenetics & Evolution,47, 916–931.

Wake, D. B., & Vredenburg, V. T. (2008). Are we in the midst of thesixth mass extinction? A review from the world of amphibians.Proceedings of the National Academy of Science USA, 105,11466–11473.

Warren, D. L., Glor, R. E., & Turelli, M. (2008). Environmental nicheequivalency versus conservatism. Quantitative approaches toniche evolution. Evolution, 62, 2868–2883.

Warren, D. L., Glor, R. E., & Turelli, M. (2010). ENMTools: atoolbox for comparative studies of environmental niche models.Ecography, 33, 607–611.

Weinsheimer, F., Mengistu, A. A., & Rödder, D. (2010). Potentialdistribution of threatened Leptopelis spp. (Anura, Arthroleptidae) inEthiopia derived from climate and land-cover data. EndangeredSpecies Research, 9, 117–124.


http://dx.doi.org/10.1111/j.1466-8238.2011.00659.x

Weisrock, D. W., Macey, J. R., Ugurtas, I. H., Larson, A., &Papenfuss, T. J. (2001). Molecular phylogenetics and historicalbiogeography among salamandrids of the "true" salamanderclade, rapid branching of numerous highly divergent lineages inMertensiella luschani associated with the rise of Anatolia.Molecular Phylogenetics and Evolution, 18, 434–448.

Wiens, J. J. (2004). Speciation and ecology revisited, phylogenetic nicheconservatism and the origin of species. Evolution, 58, 193–197.

Wiens, J. J., & Graham, C. H. (2005). Niche conservatism, integratingevolution, ecology, and conservation biology. Annual Review ofEcology and Systematics, 36, 519–539.

Wilson, R. C. L., Drury, S. A., & Chapman, J. L. (1999). The GreatIce Age. London: The Open University.

Wisz, M. S., Hijmans, R. J., Peterson, A. T., Graham, C. H., Guisan,A., & NPSDW Group. (2008). Effects of sample size on theperformance of species distribution models. Diversity andDistributions, 14, 763–773.

Zhang, P., Papenfuss, T. J., Wake, M. H., Qu, L., & Wake, D. B.(2008). Phylogeny and biogeography of the family Salaman-dridae (Amphibia, Caudata) inferred from complete mito-chondrial genomes. Molecular Phylogenetics and Evolution,49, 586–597.


a novel method to calculate climatic niche similarity among … · 2018-02-05 · graphic ranges...

Documents