Climate-Induced Shifts in the Niche Similarity of Two Related Spadefoot Toads (Genus Pelobates)

Org Divers Evol (2014) 14:397–408 DOI 10.1007/s13127-014-0181-7

ORIGINAL ARTICLE

Climate-induced shifts in the niche similarity of two related spadefoot toads (genus Pelobates)

Ruben Iosif & Monica Papeş & Ciprian Samoilă & Dan Cogălniceanu

Received: 12 March 2014 /Accepted: 28 July 2014 /Published online: 9 August 2014 # Gesellschaft für Biologische Systematik 2014

Abstract Of the four species encompassing the genus for P. fuscus, and Israel, southern Balkans, and Caucasus for Pelobates, only two overlap along a narrow contact zone, P. syriacus. Present potential distributions revealed a low i.e., Pelobates fuscus and Pelobates syriacus. Our study in- similarity of species’ ecological niches, comparable with Last vestigated the shifts in niche similarity of these two closely Interglacial, but projections towards 2080 revealed a sharp related species from the Last Interglacial towards the end of increase. the twenty-first century. We computed climatic suitability models using Maxent and projected them onto future and past Keywords Ecological niche . Climate change . Pelobates climates. We used fossil occurrences to test the predictive fuscus . Pelobates syriacus . Glaciations accuracy of past projections. Niche similarity was assessed between the studied species using Schoener’s D index and a background similarity test. Finally, we evaluated niche differ- Introduction entiation by contrasting the species occurrences using a logistic regression analysis. The ecological niches are slightly Climate change contributed to the decline and even extinction extended outside the present geographical ranges in the Cau- of many amphibian species throughout the world (Reading casus and the Balkans, south for P. fuscus and north and west 2007). Biological responses to climate change often result in for P. syriacus, suggesting that their present distribution is not range shifts (Parmesan 2006), usually towards higher altitudes at equilibrium with the climate. The Last Interglacial distribu- and latitudes, in accordance with the species’ thermal limits. tion of P. fuscus included British Isles and broad areas in For species to survive, they must keep pace with the climate western, central, and northern Europe, while P. syriacus ex- shifts (Loarie et al. 2009), but there are concerns that many tended northwards in the Balkans. The validation with fossil will not be able to establish populations in newly suitable records revealed good predictive performance (omission er- areas, particularly given the rapid rate of climate change and ror=4.1 % for P. fuscus and 16.6 % for P. syriacus). During man-made barriers to dispersal. the Last Glacial Maximum, climatic suitability persisted in Amphibians are declining worldwide at an alarming rate, a refugia in southern Europe, Pannonian Basin, and Caucasus phenomenon referred to as the sixth mass extinction (Wake et al. 2008). The synergistic effects of climate change are forecasted to produce major disturbances in the near future Electronic supplementary material The online version of this article (Araújo et al. 2006; Hof et al. 2011). Amphibians are not only (doi:10.1007/s13127-014-0181-7) contains supplementary material, which is available to authorized users. threatened by increases in temperature but also by changes in precipitation and hydrology since prolonged droughts may : ă : ă * R. Iosif C. Samoil D. Cog lniceanu ( ) lead to reproductive failure, shorter larval periods due to Faculty of Natural and Agricultural Sciences, Ovidius University Constanţa, 1 Aleea Univesitatii, Building B, Room P43, shorter hydroperiods, and lower fitness at metamorphosis Constanţa 900470, Romania (see Walls et al. 2013 for a review). e-mail: [email protected] Forecasting climate change and assessing its effects are complex research problems, especially because they represent M. Papeş Department of Zoology, Oklahoma State University, 501 LSW, an intergenerational issue (Wunsch et al. 2013), and few Stillwater, OK 74078, USA species have been the subject of long-term studies (Walls 398 R. Iosif et al. et al. 2013), yet we are now capable to rapidly address ques- experiencing hydrological stress (Forzieri et al. 2014). In the tions in biogeography and conservation due to the accumula- present study, we focused on identifying the potential changes tion of GIS and remotely sensed data and the tools associated in the two species’ distributions and particularly the spatial with modeling species’ ecological niches developed in the shifts of the overlapping zone. Our goals were to (i) compare past decade. One of the main issues in the long-term conser- the two species’ niche similarity over different time spans and vation strategies of amphibians is forecasting the range shifts scenarios, (ii) identify past climatic refugia, and (iii) identify to future changes, whether climate or land cover. To forecast the potential shifts induced by future climate changes in the these shifts, we need to understand the past events that shaped species’ distributions. Furthermore, we discussed the factors the species’ ranges. In this respect, it is relevant to investigate that may limit the two species’ range overlap. climate-driven past biogeographic events (Espregueira Themudo and Arntzen 2007). The Holarctic distribution patterns of amphibians were shaped by the Quaternary glaciations. The climatic refugia Materials and methods during the Last Glacial Maximum (LGM, ∼21 ka BP) and geographic barriers during the Last Interglacial (LIG, 120– Occurrence data 140 ka BP) influenced the size of the species’ ranges. The impact was stronger in Europe than in North America because We compiled a database of species occurrences (4,972 for of the east-west orientation of the mountain ranges that acted P. fuscus and 394 for P. syriacus) from published papers, more effectively as barriers to colonization (Gray 1912, natural history museum specimens, information available in Ibrahim et al. 1996, Hewitt 1999). Thus, many of the present online databases (www.gbif.org and http://nhm-wien.ac.at/), day widespread European amphibians are distributed mostly and from our own field surveys (Electronic Supplementary along longitudinal gradients rather than latitudinal ones, de- Material), covering the entire extent of the species’ ranges. If pending on the past glacial refugia and postglacial dispersal. the geographic coordinates were not available, the records This spatial pattern suggests their ranges are not at equilibrium from published papers were geo-referenced based on their with the current climate (Baselga et al. 2012) and that biotic textual description. Because the sampling effort differed interactions such as competition are acting as important range across the species’ ranges (Fig. 1a), we filtered the occur- determinants (Zeisset and Beebee 2008;Wiszetal.2013). rences to reduce the sampling bias and autocorrelation. First, The four species of the spadefoot toad genus Pelobates we trimmed duplicate occurrences within a single cell of the (i.e., Pelobates cultripes, Pelobates varaldii, Pelobates WorldClim climatic layers (i.e., 30 arcsec resolution, fuscus,andPelobates syriacus), distributed over parts of representing approximately 1 km2) and then we used an Europe, North Africa, Caucasus, and the Middle East, have arbitrary threshold of 0.5 occurrences/1,000 km2 for countries mostly allopatric ranges, except for P. fuscus and P. syriacus. with high density of occurrences (e.g., Austria for P. fuscus The ranges of the latter two species overlap along a narrow and Israel for P. syriacus). Our final dataset contained 1,200 contact zone in Dagestan and in the Balkans, along the Dan- occurrences of P. fuscus and 251 of P. syriacus (Fig. 1b). We ube between the Black Sea Coast and the Iron Gates. Thus, also manually relocated the coastal localities that did not P. fuscus and P. syriacus can be used as a case study to overlap with the climatic layers to the nearest cell. Recent understand the factors that shaped the present distribution of phylogeographic studies proposed to split the two species into closely related amphibians. There are few studies on the an eastern and a western clade for P. fuscus (Litvinchuk et al. spadefoot toads of the genus Pelobates because they are 2013) and a northern and southern clade for P. syriacus highly specialized species, with a narrow ecological niche, (Kieren et al. 2013). We decided to consider these clades as and are difficult to investigate: strictly nocturnal, obligatory single species for two reasons: (i) our database contains his- burrowing, with a weak underwater call (Nöllert 1990; torical records in all the proposed clades at a time when the Cogălniceanu et al. 2013). For these reasons, little is known taxonomic status of the two species was not under debate (i.e., about their biotic interactions and abiotic requirements. The the oldest localities were reported back in 1928 for P. syriacus interspecific competition can be enhanced in synergy with and 1921 for P. fuscus) and (ii) range limits for the proposed climate change or habitat alteration and, thus, disturb the clades cannot be established until further genetic analysis viability of the syntopic populations. Araújo et al. (2006) clarifies their distribution patterns. forecasted that the ranges of P. syriacus and P. fuscus will To validate the past projections of the ecological niche expand northwards while the range of P. cultripes will con- models, we used fossil records dating to the LIG obtained tract, as expected with higher frequencies of droughts in from Lisanfos KMS (Martín and Sanchiz 2013), Southern Europe already under hydrological stress. These FOSFARBASE (Böhme and Ilg 2003), and Holman (1998). results strengthen the hypothesis of reducing viability in the When geographic coordinates were unavailable, we manually syntopic populations as the region of range overlap is also geo-referenced the localities based on their textual description Climate-induced shifts in the niche similarity of two toads 399

Fig. 1 Presence datasets compiled for the two species studied: a raw species’ ranges explains 99.1 % of the climate variability. For P. fu sc us, data; b filtered data used to train present day models (density of points western and eastern climates can be delineated within its range (at reduced with an arbitrary threshold of 0.5 occurrences/1,000 km2 for each approximately 30° longitude), with a relatively constant climate in the country); the fossil records dating from the Last Interglacial used to eastern part and a highly variable one in the western part. For P. syriacus, validate the past projections; c the first axis of a principal component a high climatic heterogeneity is shown within its entire range analysis based on 19 climatic variables representing the extent of the whenever possible. The final dataset contained 50 fossil re- Furthermore, removing these two variables reduced the cords for P. fuscus and 6 for P. syriacus (Fig. 1b). models’ performance (i.e., overprediction and high omission error; data not shown). Finally, to capture the increased cli- Climatic space mate variability tolerated by the spadefoot toads, we selected the same seven variables for both species: annual mean tem- To represent the present climate conditions in the modeling perature, isothermality, temperature seasonality, minimum experiments, we used the WorldClim dataset containing 19 temperature of coldest month, precipitation seasonality, pre- variables of mean annual and seasonal temperature cipitation of warmest quarter, and precipitation of coldest and precipitation (Hijmans et al. 2005). First, to avoid quarter. multicollinearity in the dataset, we removed variables from The climate variables used to generate the models were highly correlated pairs based on all sample points (i.e., we clipped to an area merging both species ranges (i.e., calculated removed variables with Spearman’srankcorrelation>0.71for as minimum convex polygons based on present occurrences) P. fuscus; Guisan and Zimmermann 2000). Second, following with an additional buffer zone of 335 km to include all fossil Araújo et al. (2006), the a priori selection of variables took records that fell outside of the present distribution (hereafter into account the biology of spadefoot toads. Thus, we main- modeling extent). We used this modeling extent to capture the tained the minimum temperature of coldest month and tem- potential range shifts in different climate scenarios and to perature seasonality for P. syriacus, although they were highly compare the projections. Wider extents (700 and 1,000 km) correlated, as we expect the species ecology to be influenced included novel climatic conditions for both studied species by them (e.g., the species has seasonal activity and its northern (data not shown); in other words, some climatic variables were distribution may be limited by frost during winter). outside the range that characterized the region used to generate 400 R. Iosif et al. the models, thus rendering the unreliable projections (Thuiller operates with presence-only datasets and allows a rapid et al. 2004, Elith et al. 2010). For past climate conditions, we modeling workflow and transfer of the model output to statis- used two simulations for LGM, i.e., Model for Interdisciplin- tical tools for measuring niche similarity. Comparable ap- ary Research on Climate version 3.2 (MIROC) and Commu- proaches were used in recent ecology and biogeography stud- nity Climate System Model version 3 (CCSM; Collins et al. ies (e.g., Kozak and Wiens 2006;CostaandSchlupp2010; 2006), and one for LIG (Otto-Bliesner et al. 2008). To assess Nakazato et al. 2010). We used the auto features parameter in the future climate change effects, we used A1B IPCC climate the algorithm as proposed by Phillips and Dudik (2008)and scenario developed under the third Hadley Global Circulation cited in the study of Costa and Schlupp (2010), with default Model (HadCM3) for three-time frames (i.e., 2030, 2050, and options for model parameters. We used random partitions of 2080). This scenario describes a very rapid economic growth 75 % of the occurrence data for training and 25 % for testing but with a balance across all energy sources used the model accuracy. We performed 50 replicates for each of (Nakicenovic and Swart 2000). The past and future climate our modeling exercises using a bootstrapping procedure. datasets used for model projections were downloaded from Model performance was assessed via mean of area under WorldClim dataset (Hijmans et al. 2005). The raster resolution the receiver operating characteristic (ROC AUC) of the 50 was 30 arcsec for present climate and 2.5 arcmin for past and replicates for testing data. Warren and Seifert (2010)docu- future datasets, with consistency between scenarios. mented that for large datasets (as in our case), the performance Since most P. fuscus occurrences were clustered in the of information criterion approaches (i.e., Akaike information western half of the species’ range and scarce in the eastern criterion and Bayesian information criterion) is just slightly half (see Fig. 1a), we ran a principal component analysis higher than ROC AUC, but for small datasets, caution is (PCA) of climate variables across the species extent. The first needed when interpreting ROC AUC. Additionally, to evalu- PCA axis explained 99.1 % of the climate variability and ate model performance, we calculated the mean omission delineated a western and eastern climate regime (separated at error across 50 bootstrap replicates, that is, the proportion of approximately 30° longitude), suggesting a relatively constant presences in the testing dataset predicted absent by the climate in the eastern range (with a smooth ascending gradient model. from west to east), but high variability in the western part Finally, we used the species’ fossil datasets to assess the (determined by proximity to seas and the presence of moun- accuracy of potential distributions obtained when projecting tains; Fig. 1c). We assumed that since the climate is relatively the models onto LIG climate datasets. We calculated omission constant in the eastern part of the range, a smaller number of error after applying a minimum training presence threshold training points in that region compared to the number avail- (i.e., the minimum Maxent suitability value at which all able in the western part would not compromise the model’s training points are predicted present; Pearson et al. 2007)to ability to capture the climatic variation (Pearson et al. 2007), the model outputs to delineate regions with suitable and whereas a higher number of occurrences within the more unsuitable climates. This approach is considered suitable for climatically variable western part of P. fuscus range would presence-only data (Mateo et al. 2013). be necessary. The climate variability within P. syriacus range To compare the interpredictability between current and was also large, but no obvious general clustering pattern was fossil records, we calculated the omission error of one predic- present, the localities being scattered over the entire area of the tion based on the species’ occurrences from the other time minimum convex polygon (Fig. 1c). frame and vice versa. We used the same settings as for present day models to generate models based on the fossil records and Ecological niche modeling the LIG climate dataset and projected the models onto the present climate dataset. We calculated the omission error for We used the Maxent algorithm proposed by Phillips et al. current occurrences using the same minimum training pres- (2006) to generate models of the ecological niche for the two ence threshold described above to transform Maxent outputs spadefoot toad species and to project the models onto both from continuous suitability values to binary presence-absence. past and future climate scenarios. Maxent is a machine learn- We then contrasted the occurrences of the two ing method that uses the mathematical formulation of maxi- spadefoot toad species with the same seven climatic var- mum entropy to predict the species’ potential geographic iables used to train the present day Maxent models. We distribution (Phillips et al. 2004; Phillips et al. 2006). Al- performed a weighted logistic regression analysis with a though it was documented that this approach performed better forward stepwise addition of the climatic variables. The than standard statistical techniques (Elith et al. 2006), recent response variable is the presence of one species versus the criticism has been voiced not only related mainly to re- presence of the other species (see Arntzen and searcher’s complacent in training a proper model (Warren Espregueira Themudo 2008 for a complete description). and Seifert 2010) but also related to the algorithm itself This approach aims to highlight the different ecological (Renner and Warton 2013). We selected Maxent because it requirements of the species. Climate-induced shifts in the niche similarity of two toads 401

Niche similarity between the two species Mediterranean (e.g., Italian Peninsula) and northwards in the Pannonian Basin and north of Black Sea and of Caucasus. The study of ecological niche convergence processes can However, the species’ known range is limited in the Balkans provide information about corresponding distributional pat- along the Danube River, and a predicted climatic suitability terns (e.g., niche similarity or range overlap), and closely gap in western Balkans may act as a barrier to the species related species provide the framework for testing such hypoth- dispersal there (Fig. 3a). The species’ known presences— eses. A strong niche similarity between closely related species restricted to the southern slopes of Caucasus—support the is attributed to slow evolution of niche differences, while a hypothesis of dispersal barriers. The model for P. syriacus strong niche divergence is attributed to ecological differentia- had a high predictive accuracy (mean testing ROC AUC=0.960, tion after speciation (Nakazato et al. 2010). Here, we are SD=0.007). The mean omission error across replicates was referring to niche similarity as a property of niches to be 0.001 (<1 occurrence) for P. fuscus and 0.011 (1 occurrence) similar or different in geographical space (Peterson et al. for P. syriacus. 2011). The best model for the logistic regression analysis included We used the raw Maxent outputs to calculate Schoener’sD all seven climatic variables and predicted the outcome better index, a similarity index of niche space derived by Schoener than the constant model (Nagelkerke pseudo-R2 increased (1968) and implemented in ENMTools software (Warren et al. from 0.59 to 0.83). The occurrence probability of P. fuscus 2008). This index ranges between 0 (no similarity between versus P. syriacus (p) depends more on temperature rather niche models of a species pair) and 1 (identical niche). Recent than precipitation variables following the equation: p=1//[1+ studies found that Schoener’s D is best suited to compute exp (−2.623×annual mean temperature− 0.250× niche similarity from potential distributions (Rödder and isothermality+0.040×temperature seasonality+1.490×mini- Engler 2011). We calculated this statistic for the modeling mum temperature of coldest month−0.194×precipitation sea- extent that included both species ranges and for current, LIG, sonality+0.072×precipitation of warmest quarter+0.015× LGM, and future climates. precipitation of coldest quarter+0.412)]. Finally, we tested the hypothesis of niche similarity using the ENMTools background similarity test (Warren et al. 2008) Past climate potential distributions only for the present day climatic conditions. The background test compares the observed values of niche similarity metric When projecting the present climate ecological niche model of described above with the values obtained from random pre- P. fuscus onto LIG climatic conditions, we obtained a good dictions (i.e., niche models trained with randomly selected predictive performance of fossil records (omission error subsamples of occurrences; Nakazato et al. 2010). We per- 4.1 %). Similarly, when projecting the model obtained using formed 100 randomizations and assessed the niche similarity the LIG climate dataset and fossil records onto the present in climatic space in two directions: We compared the occur- climate, we obtained a good predictive performance of present rences of each species with the climatic background within the records (omission error 1.3 %). The LIG climatically suitable training region of the other one (see Warren et al. 2008 for a region was larger and included the British Isles, France, as complete description). A summary of the modeling steps is well as the vast areas in Northern Europe (e.g., Jutland Pen- presented in Fig. 2. insula; Fig. 3b). The climatically suitable areas in LGM MIROC scenario included the climatic refugia in Southern France, northern part of Italian Peninsula, Pannonian Basin, and Caucasus region (Fig. 3c). The results from LGM CCSM Results scenario differed in that the refugia were located in the north- eastern part of the extent, with just a few isolated spots in Present climate potential distributions southern Greece and north and east of the Black Sea (Fig. 3d). The predictive performance of the LIG fossil records was The minimum convex polygon based on the present records of similarly good for P. syriacus (omission error 16.6 %) when P. fuscus covered approximately 6,000,000 km2, but the po- projecting the model generated with present day records and tential ecological niche derived in the climatic space extended climate data onto LIG climate dataset. Compared with the outside the known range in the Dagestan Region, Georgia, present distribution, the LIG potential distribution extended and Azerbaijan (Fig. 3a). We obtained a moderate accuracy northwards in the Balkans and Caucasus (Fig. 3b). When for P. fuscus model (mean testing ROC AUC=0.878, SD= training the model with fossil occurrences and LIG climate 0.007). The minimum convex polygon based on the present dataset, the model projection onto the present climate dataset records of P. syriacus covered approximately 2,750,000 km2, had very low predictive power of present day species pres- but the potential ecological niche derived in the climatic space ences (omission error 81.2 %), most likely because of the was also extended outside the known range, westwards in the small sample size of fossil records used for training (n=6), 402 R. Iosif et al.

Fig. 2 Modeling stages used in the present study for investigating the factors limiting the current range overlap and possible changes in the geographic ranges of the two species studied under future climate change scenarios

all clustered in Israel. The climatic suitability in LGM was suitable environmental conditions for both species at higher restricted to three main refugia: Israel, Caucasus region, and latitudes (i.e., P. syriacus niche shifted northwards in the Southern Balkans, the results being comparable between Balkans and Dagestan Region; Fig. 3e). The climatic suitabil- CCSM and MIROC scenarios (Figs. 3c, d). ity of P. syriacus tended to decrease in the southern part of its range (25–30° N), with minor changes at midlatitudes Future climate potential distributions (Fig. 3e).

Models projected on climate conditions for the end of the twenty-first century indicated that the western part of the Niche similarity between the two species distribution of P. fuscus would preserve the climatic suitability, whereas a decrease would occur in the central part of the The ecological niche models from present and LIG had com- range (Fig. 3e). The climatic suitability was predicted to parable levels of similarity (DPRESENT=0.30, DLIG=0.26) significantly increase under all time frames of the A1B climate while lower values were revealed under LGM conditions (D- scenario east of approximately 20° longitude. At the eastern CCSM=0.15, DMIROC=0.20). The niche similarity was limit of this species range, the climatic suitability was predict- projected to increase from present towards 2080 (D2080= ed to remain stable (i.e., Dagestan Region). We obtained more 0.60; Table 1). Climate-induced shifts in the niche similarity of two toads 403

Fig. 3 The potential distribution of P. fu sc us and P. syriacus in: a present climate for the two species, b Last Interglacial climate, c Last Glacial Maximum MIROC Scenario, d Last Glacial Maximum CCSM Scenario, and e A1B scenario for 2080. The modeling extent merges the minimum convex polygons of the two species and an additional buffer zone of 335 km that includes all fossil records 404 R. Iosif et al.

Table 1 Similarity between predicted ecological niches using Schoener’s similar than expected by chance alone based on their climate D metric implemented in Warren et al. (2008). This metric takes values requirements (Fig. 4b). between 0 (no similarity between niche models of a species pair) and 1 (identical niche). The niche similarity assessments were carried out within the modeling extent (including the ranges of both species and a buffer zone of 335 km; see “Materials and methods”) Discussion Climatic scenario Schonner’sD

LIG 0.26 Present climate potential distributions LGM CCSM 0.15 The present climate ecological niche models of the two MIROC 0.20 spadefoot toads predicted potential distributions that extended Present 0.30 outside their known ranges, especially southwards for P.fuscus A1B and northwards for P.syriacus, suggesting that their geographic distributions are not at equilibrium with climate. We assume 2030 0.40 that other factors are responsible for limiting their distribution 2050 0.51 and range overlap, like competition and the presence of geo- 2080 0.60 graphic barriers. The climatic suitability of P. syriacus in Georgia showed a broader range northwards, but its distribution is probably restricted by competition with P. fuscus The background test comparing present ecological niche (Tarkhnishvili et al. 2009). Another study showed that indi- models of P. fuscus and P. syriacus showed that the observed viduals from populations of P. syriacus occurring in sympatry value of niche similarity was lower than expected under the with P. fuscus in the Balkans were significantly smaller, had null hypothesis, yet not significantly (Fig 4a). When compar- shorter life span, and had earlier sexual maturity than individ- ing in the opposite direction (i.e., P. syriacus with the climatic uals from allopatric populations (Rot-Nikčević et al. 2001). background of P. fuscus), the observed value of niche similar- The logistic regression analysis revealed that temperature ity was significantly higher than expected, indicating that the rather than precipitation variables are better determinants of niches of the two Pelobates species were significantly more species ranges, with annual mean temperature and minimum

Fig. 4 Background tests of niche similarity compare the observed value than expected under the null hypothesis, yet not significantly; b when of Schoener’s D index, a similarity index of niche space (dashed line), comparing in the opposite direction (i.e., P.syriacus with P.fuscus present with the distribution of 100 pseudoreplicate D values. a The background ecological niche), the observed overlap is higher (a more similar ecolog- test comparing P. fuscus occurrences with the background space of ical niche) P. syriacus revealed that the observed value of niche similarity is lower Climate-induced shifts in the niche similarity of two toads 405 temperature of the coldest month having the highest influence. barriers limited the distribution of other European amphibians A case study on the range limits of the newts Triturus (Arntzen et al. 2007). The hypothesis of biogeographic bar- marmoratus and Triturus pygmaeus revealed that range deter- riers is also supported by the presence of fossil records for minants differ not only in the climatic space but also in the P. fuscus in France and British Isles; their age suggesting that geographical space. The variables determining their range the climatic suitability in LIG scenario was extended west- limits vary along the range border from region to region, wards (see Fig. 3). favoring one species after the other in the competition process The climate of LGM shaped refugia for both species rough- between them (Arntzen and Espregueira Themudo 2008). The ly in the same regions as the overlapping zone of the present distribution of the spadefoot toads can have similar spatial day models (i.e., the lower Danube and the northern Cauca- variability of the range determinants. For example, P. syriacus sus), but the niche similarity values were slightly smaller than is locally better adapted to drought conditions and dominates in the present. The refugia for P. fuscus during the LGM were the arid areas in the overlapping zone in Romania (Székely estimated to be located in the Italian Peninsula, the Balkans, 2010; Székely et al. 2010). P. syriacus tadpoles from syntopic and the Caucasus Region in agreement with refugia predicted populations had a shorter larval period when constrained to for other amphibians (e.g., genus Triturus; Wielstra et al. metamorphose in the presence of P. fuscus tadpoles in ponds 2013). The populations located in the Italian Peninsula are with induced desiccation by decreasing water level; no signif- isolated by the Alps, and this may be responsible for the icant response was recorded for P. fuscus tadpoles (Székely genetic differentiation of P. f. insubricus in Po Valley (Crottini et al. 2010). Furthermore, the southern edge of the range of et al. 2007). This supports the hypothesis that the northern P. syriacus is determined by the limited osmoregulatory ca- species P. fuscus has changed its range more drastically than pacity of adults (Shpun et al. 1993) and the increased aridity the southern species P. syriacus during glaciations in accor- that does not allow for the tadpoles to metamorphose (Men- dance with the genetic analysis that revealed similar patterns delssohn and Steinitz 1944). Archaeological evidence from for other amphibians (Wielstra et al. 2013). the Bronze Age in Israel suggested that recent environmental changes (climatic fluctuations and/or anthropogenic impact) Forecasted distributional changes may be responsible for the present fragmented southern range of the species (Delfino et al. 2007). Under future climatic conditions, niche similarity showed a Both P. fuscus and P. syriacus are widely distributed, but rapid increase. For P. fuscus, the western part of the range was suffered a severe decline in the recent decades, most probably projected to maintain climatic suitability, the center part to due to human impacts (Andreone et al. 2004,Džukić et al. lose it, and the eastern part to increase it. The southern pop- 2005). For example, the subspecies Pelobates fuscus ulations were projected to lose their climatic space, a north- insubricus consists of no more than 15 populations scattered ward shift being expected. This result is in accordance with in northern Italy, with an estimated 50 % population loss in the Popescu et al. (2013), who evaluated the potential range shifts last 25 years (Giovannini et al. 2014). A case study focusing of Romanian amphibians and reptiles, showing that P. fuscus on peripheral populations of P. syriacus in Israel reported the will face range loss at the southern edge of its distribution in existence of additional genetic variability compared with the both dispersal and no dispersal scenarios. For P. syriacus,our core populations (Munwes et al. 2010). These peripheral models predicted that populations occupying the southern populations are becoming an important source of genetic limit will face range loss while those occupying the northern diversity that can be lost even with slight climate changes in edge will tend to expand northwards, along the Black Sea the extremely sensitive landscapes they inhabit. Thus, it is coast and lower Danube. Again, this result is in accordance important to evaluate whether these edge populations will face with the regional forecast of Popescu et al. (2013), where range loss by the end of the century. Romanian populations of P. syriacus were projected to gain suitable space northwards even within a limited dispersal Past climate potential distributions scenario. Although the climatic suitability is predicted to be preserved under future climate conditions in the western part The climate during the LIG period delineated a distribution of P. fuscus range, the pattern is challenged by the recent that extended westwards for P. fuscus and northwards for severe decrease of the western European populations. This P. syriacus, leading to a higher niche similarity and potential decline is well documented and is explained by high human range overlap (see Table 1). The northward extension of the impacts, mostly habitat destruction (Nyström et al. 2002). climatic suitability of P. syriacus under LIG scenario covered Thus, even if the climate conditions in the western part of the Pannonian plains, Ukraine, and Dagestan Region. The the range would allow the persistence of P.fuscus populations, Carpathian and Caucasus Mountains acted as geographic habitat loss will still threaten their persistence. Moreover, the barriers as no fossil records were found there. Our results are viable populations in the central part of P. fu sc us range, which consistent with phylogenetic studies revealing that mountain do not face important range loss at this moment (e.g., 406 R. Iosif et al. populations in southern Poland; Bonk and Pabijan 2010), are in Israel. This led to low predictive performance when facing higher risk from climate warming. Our analyses pre- training the models on past climate data with fossil records dicted an increase in the climatic suitability at the eastern edge and when testing the past projections of models trained of the P. fuscus range and a northward displacement of both with present occurrences and associated climate condi- species because of temperature rising and increasing deserti- tions. Fossil records can not only highlight regional cli- fication. In the region of range overlap, these changes can matic refugia but also may be the effect of the utilization of favor P.syriacus dispersal since the species takes advantage of microrefugia during the glaciations, i.e., local sites with agricultural fields and irrigation canals due to the availability favorable climate nested in a region with unfavorable con- of friable soils and breeding sites (Székely et al. 2013). In two ditions (Dobrowski 2011). This effect can create a biased sister species of marbled newts in Central Portugal, a rise in estimation of the past distribution, and for this reason, we temperature and increased desertification are accepted as suggest a cautionary interpretation of the power of fossil probable causes of shifts in their parapatric distribution records to validate broad scale predictions. (Espregueira Themudo and Arntzen 2007). It is probable that Niche similarity is a property of the biological organisms these environmental changes triggered competition distur- that can be quantified—at least to an extent—with the present bances at population scale; thus, the southern species climate distribution models, especially if proper biological T. pygmaeus expanded its range northwards and isolated data are used (Peterson et al. 2011). However, for past and populations of its relative T. marmoratus (Espregueira future projections, the niche similarity and overlap in geo- Themudo and Arntzen 2007). graphical space should be treated with caution especially when biotic interactions or dispersal measures are not taken Niche similarity into account.

Estimates of niche similarity based on climatic variables were slightly lower in the LGM compared with the present climate, although the climatic refugia mostly coincided Conclusions with the present overlapping zone. The range shifted northwards during LIG, but the niche estimates were compara- Both spadefoot toad species suffered recent range contrac- ble with the present climate. We propose that the overlap tions, especially the populations living at the edge of their zone has sustained viable populations from the Ice Ages, as range that are vulnerable to even slight climate changes. our models predicted high climatic suitability in all past The southern populations are exposed to increased aridity, and present scenarios in approximately the same regions and their capacity to keep the pace with climate change (e.g., lower Danube and Dagestan Region). This is in will depend not only on their dispersal abilities and lack of accordance with the molecular study of Litvinchuk et al. barriers to dispersal but also on habitat availability (sandy (2013) that revealed the same glacial refugia for P. fuscus. soils and aquatic habitats for reproduction). We found Towards the end of the twenty-first century, our results comparable niche similarity values between present and projected an increase in niche similarity. This test also LIG and less stringent values under LGM conditions. Fur- provide support for the hypothesis of a potential distur- thermore, the similarity was projected to rapidly increase bance due to the competition between the two related in less than 100 years from present, suggesting that these species in the region of range overlap, where edge popu- rapid climate changes have the potential to disturb the links lations barely coexist with the sister species (see between the abiotic and biotic spaces of the two species. Espregueira Themudo and Arntzen 2007 for an example Furthermore, the present climate models revealed a distri- on newts). Although we obtained a moderate value of bution that is not in equilibrium with climate at the south- niche similarity (D=0.30), the actual contact zone has an ern limit for P. fuscus and at the northern limit for average width of no more than 100 km. This can be caused P. syriacus. Towards the end of the twenty-first century, by (i) other limiting factors, most likely competition, as no northern territories were predicted to provide suitable cli- barrier can be identified at least in the Lower Danube basin matic space for both species. Competition disturbances and (ii) novel climate conditions producing a bias in the may be expected, but further ecological studies are needed models when selecting the spatial extent for calculating the to evaluate the dispersal ability and the adaptive capacity similarity index. of these two spadefoot toad species.

Limitations of the present study Acknowledgments This work was partly supported by a grant of the Romanian National Authority for Scientific Research PN-II-PCE-2011- 3-0173 and a Fulbright Senior Fellowship to DC. We thank Dr. Roberto The first caveat of our findings is related with the insuffi- Sindaco, Dr. Wieslaw Babik, Dr. Paul Székely, Dr. Laurenţiu cient fossil records for P. syriacus restricted to a small area Rozylowicz, and Dr. Jan Arntzen for sharing occurrence data with us. Climate-induced shifts in the niche similarity of two toads 407

We thank Florina Stănescu and Dr. Marton Venczel for the help in Forzieri, G., Feyen, L., Rojas, R., Flörke, M., Wimmer, F., & Bianchi, A. compiling the fossil record database. We appreciate the suggestions (2014). Ensemble projections of future streamflow droughts in provided by Dr. Jan Arntzen and an anonymous reviewer to improve this Europe. Hydrology and Earth System Sciences, 18,85–108. manuscript. Giovannini, A., Seglie, D., & Giacoma, C. (2014). Identifying priority areas for conservation of Pelobates fuscus insubricus (Cornalia, 1873) using a maximum entropy approach. Biodiversity and Conservation, 23,1427–1439. References Gray, A. (1912). Scientific papers of Asa Gray. Vol. 2. 1912. Reprint. London: Forgotten Books, 2013. 230–1. http://www. forgottenbooks.org/readbook_text/Scientific_Papers_of_Asa_ Andreone, F., Bergò, P. E., Bovero, S., & Gazzaniga, E. (2004). On the Gray_v2_000777018/235. Accessed 25 June 2014 edge of extinction? The spadefoot Pelobates fuscus insubricus in the Guisan, A., & Zimmermann, N. E. (2000). Predictive habitat distribution Po Plain, and a glimpse at its conservation biology. Bollettino di models in ecology. Ecological Modelling, 135,147–186. Zoologia, 71,61–72. Hewitt, G. M. (1999). Post-glacial re-colonization of European biota. Araújo, M. B., Thuiller, W., & Pearson, R. G. (2006). Climate warming Biological Journal of the Linnean Society, 68,87–112. and the decline of amphibians and reptiles in Europe. Journal of Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G., & Jarvis, A. Biogeography, 33, 1712–1728. (2005). Very high resolution interpolated climate surfaces for global Arntzen, J. W., Espregueira Themudo, G., & Wielstra, B. (2007). The land areas. International Journal of Climatology, 25,1965–1978. phylogeny of crested newts (Triturus cristatus superspecies): nucle- Hof, C., Araujo, M. B., Jetz, W., & Rahbek, C. (2011). Additive threats ar and mitochondrial genetic characters suggest a hard polytomy, in from pathogens, climate and land-use change for global amphibian line with the paleogeography of the centre of origin. Contributions diversity. Nature, 480,516–519. to Zoology, 76,261–278. Holman, J. A. (1998). Pleistocene amphibians and reptiles in Britain and Arntzen, J. W., & Espreguiera Themudo, G. (2008). Environmental Europe. USA: Oxford University Press. parameters that determine species geographical range limits as a Ibrahim, K. M., Nichols, R. A., & Hewitt, G. M. (1996). Spatial patterns matter of time and space. Journal of Biogeography, 35, 1177–1186. of genetic variation generated by different forms of dispersal during Baselga, A., Lobo, J. M., Svenning, J. C., & Araújo, M. B. (2012). Global range expansion. Heredity, 77,282–291. patterns in the shape of species geographical ranges reveal range Kieren, S., Crottini, A., De Pous, P., & Veith, M. (2013). Phylogeny and determinants. Journal of Biogeography, 39,760–771. phylogeography of the Syrian spadefoot toad (Pelobates syriacus). Bonk, M., & Pabijan, M. (2010). Changes in a regional batrachofauna in 17th European Congress of Herpetology, Programme & Abstracts, P south-central Poland over a 25 year period. North-Western Journal 129.Vesprem, Hungary. of Zoology, 6,225–244. Kozak, K. H., & Wiens, J. (2006). Does niche conservatism promote Böhme M., & Ilg A. (2003). fosFARbase, www.wahre-staerke.com/. speciation? A case study in North American salamanders. Evolution, Accessed 8 June 2012. 60, 2604–2621. Cogălniceanu, D., Székely, P., Székely, D., Roşioru, D. M., Băncilă,R.I., Litvinchuk, S., Crottini, A., Federici, S.,DePous,P.,Donaire,D.,Andreone, & Miaud, C. (2013). When males are larger than females in F., et al. (2013). Phylogeographic patterns of genetic diversity in the ecthotherms: reproductive investment in the Eastern Spadefoot common spadefoot toad, Pelobates fuscus (Anura, Pelobatidae), reveals Toad Pelobates syriacus. Copeia, 4,699–706. evolutionary history, postglacial range expansion and secondary contact. Collins, W. D., Bitz, C. M., Blackmon, M. L., Bonan, G. B., Bretherton, Organisms Diversity & Evolution, 13,433–451. C. S., Carton, J. A., et al. (2006). The Community Climate System Loarie, S. R., Duffy, P. B., Hamilton, H., Asner, G. P., Field, C. B., & Model Version 3 (CCSM3). JournalofClimate,19,2122–2143. Ackerly, D. D. (2009). The velocity of climate change. Nature, 462, Costa, G. C., & Schlupp, I. (2010). Biogeography of the Amazon molly: 1052–1055. ecological niche and range limits of an asexual hybrid species. Martín, C., & Sanchiz, B. (2013). Lisanfos KMS. Version 1.2. http:// Global Ecology and Biogeography, 19,442–451. www.lisanfos.mncn.csic.es/. Museo Nacional de Ciencias Crottini, A., Andreone, F., Kosuch, J., Borkin, L. J., Litvinchuk, S. N., Naturales, MNCN-CSIC. Madrid, Spain. Accessed 8 June 2012. Eggert, C., et al. (2007). Fossorial but widespread: the Mateo, R. G., de la Estrella, M., Felicísimo, Á. M., Muñoz, J., & Guisan, phylogeography of the common spadefoot toad (Pelobates fuscus), A. (2013). A new spin on a compositionalist predictive modelling and the role of the Po Valley as a major source of genetic variability. framework for conservation planning: a tropical case study in Molecular Ecology, 16,2734–2754. Ecuador. Biological Conservation, 160,150–161. Delfino, M., Bar-Oz, G., & Weissbrod, L. (2007). Recent shrinkage of the Mendelssohn, H., & Steinitz, H. (1944). Contribution to the ecological range of the Eastern spadefoot toad, Pelobates syriacus (Amphibia, zoogeography of the amphibians in Palestine. Revue de la Faculté Anura): archaeological evidence from the Bronze Age in Israel. des Sciences de l’Université d’Istanbul, 9,289–298. Zoology in the Middle East, 40,45–52. Munwes, I., Geffen, E., Roll, U., Friedmann, A., Daya, A., Tikochinski, Dobrowski, S. Z. (2011). A climatic basis for microrefugia: the influence Y., et al. (2010). The change in genetic diversity down the core-edge of terrain on climate. Global Change Biology, 17,1022–1035. gradient in the Eastern spadefoot toad (Pelobates syriacus). Džukić, G., Beskov, V., Sidorovska, V., Cogălniceanu, D., & Kalezić,M. Molecular Ecology, 19,2675–2689. (2005). Historical and contemporary ranges of the spadefoot toads Nakazato, T., Warren, D. L., & Moyle, L. C. (2010). Ecological and (Pelobates spp., Amphibia, Anura) in the Balkan Peninsula. Acta geographic modes of species divergence in wild tomatoes. American Zoologica Cracoviensia, 48,1–9. Journal of Botany, 97,680–693. Elith, J., Graham, C. H., Anderson, R. P., Dudík, M., Ferrier, S., Guisan, Nakicenovic, N., & Swart, R. (2000). Emission scenarios: a special A., et al. (2006). Novel methods improve prediction of species’ report of Working Group II of the Intergovernmental Panel on distributions from occurrence data. Ecography, 29,129–151. Climate Change. Cambridge, UK: Cambridge University. Elith, J., Kearney, M., & Phillips, S. (2010). The art of modelling range- Nöllert, A. (1990). Die Knoblauchkröte. Wittenberg Lutherstadt: shifting species. Methods in Ecology and Evolution, 1,330–342. Ziemsen. Espregueira Themudo, E., & Arntzen, J. W. (2007). Newts under siege: Nyström, P., Birkedal, L., Dahlberg, C., & Brönmark, C. (2002). The range expansion of Triturus pygmaeus isolates populations of its declining spadefoot toad Pelobates fuscus: calling site choice and sister species. Diversity and Distributions, 13,580–586. conservation. Ecography, 25,488–498. 408 R. Iosif et al.

Otto-Bliesner, B. L., Marshall, S. J., Overpeck, J. T., Miller, G. H., Hu, A., Shpun, S., Hoffman, J., Nevo, E., & Katz, U. (1993). Is the distribution of & CAPE Last Interglacial Project members. (2006). Simulating Pelobates syriacus related to its limited osmoregulatory capacity? arctic warmth and icefield retreat in the Last Interglaciation. Comparative Biochemistry and Physiology, 105,135–139. Science, 311, 1751–1753. Székely, P. (2010). Faunistical and ecological studies on amphibians Parmesan, C. (2006). Ecological and evolutionary responses to recent from Podişul Dobrogei. Unpubl. Ph.D. Thesis, Babeş–Bolyai climate change. Annual Review of Ecology, Evolution, and University, Cluj-Napoca. [In Romanian] Systematics, 37,637–669. Székely, P., Tudor, M., & Cogălniceanu, D. (2010). Effect of habitat Pearson, R. G., Raxworthy, C. J., Nakamura, M., & Peterson, T. A. drying on the development of the Eastern spadefoot toad (2007). Predicting species distributions from small numbers of (Pelobates syriacus)tadpoles.Amphibia-Reptilia, 31,425–434. occurrence records: a test case using cryptic geckos in Székely, P., Iosif, R., Székely, D., Stănescu, F., & Cogălniceanu, D. (2013). Madagascar. Journal of Biogeography, 34,102–117. Range extension for the Eastern spadefoot toad Pelobates syriacus Peterson, A.T., Soberón, J., Pearson, R.G., Anderson, R.P., Martínez- (Boettger, 1889) (Anura: Pelobatidae). Herpetology Notes, 6,481–484. Meyer, E., & Nakamura, N., et al. (2011). Ecological niches and Tarkhnishvili, D., Serbinova, I., & Gavashelishvili, A. (2009). Modelling geographic distributions. Monographs in Population Biology, 49. the range of Syrian spadefoot toad (Pelobates syriacus) with com- Princeton University Press. bination of GIS-based approaches. Amphibia-Reptilia, 30,401–412. Phillips, S.J., Dudik, M., & Schapire, R.E. (2004). A maximum entropy Thuiller, W., Brotons, L., Araújo, M. B., & Lavorel, S. (2004). Effects of approach to species distribution modeling. In Proceedings of the restricting environmental range of data to project current and future 21st International Conference on Machine Learning,p.83.ACM, species distributions. Ecography, 27,165–172. Banff, Alberta, Canada. Wake, D. B., & Vredenburg, V.T. (2008). Are we in the midst of the sixth Phillips, S. J., Anderson, R. P., & Schapire, R. E. (2006). Maximum mass extinction? Aview from the world of amphibians. Proceedings entropy modeling of species geographic distributions. Ecological of the National Academy of Sciences, 105, 11466–11473. Modelling, 190,231–259. Walls, S., Barichivich, W., & Brown, M. (2013). Drought, deluge and Phillips, S. J., & Dudik, M. (2008). Modeling of species distributions with declines: the impact of precipitation extremes on amphibians in a Maxent: new extensions and a comprehensive evaluation. changing climate. Biology, 2,399–418. Ecography, 31,161–175. Warren, D. L., Glor, R. E., & Turelli, M. (2008). Environmental niche Popescu, V. D., Rozylowicz, L., Cogălniceanu, D., Niculae, I. M., & equivalency versus conservatism: quantitative approaches to niche Cucu, A. L. (2013). Moving into protected areas? Setting conserva- evolution. Evolution, 62,2868–2883. tion priorities for Romanian reptiles and amphibians at risk from Warren, D. L., & Seifert, S. N. (2010). Ecological niche modeling in climate change. PLoS ONE, 8, e79330. Maxent: the importance of model complexity and the performance Reading, C. (2007). Linking global warming to amphibian declines of model selection criteria. Ecological Applications, 21,335–342. through its effects on female body condition and survivorship. Wielstra, B., Crnobrnja-Isailović,J.,Litvinchuk,S.,Reijnen,B., Oecologia, 151,125–131. Skidmore, A., Sotiropoulos, K., et al. (2013). Tracing glacial refugia Renner, I. W., & Warton, D. I. (2013). Equivalence of MAXENT and of Triturus newts based on mitochondrial DNA phylogeography and Poisson point process models for species distribution modeling in species distribution modeling. Frontiers in Zoology, 10,1–14. ecology. Biometrics, 69,274–281. Wisz, M. S., Pottier, J., Kissling, W. D., Pellissier, L., Lenoir, J., Damgaard, Rödder, D., & Engler, J. O. (2011). Quantitative metrics of overlaps in C. F., et al. (2013). The role of biotic interactions in shaping distribu- Grinnellian niches: advances and possible drawbacks. Global tions and realised assemblages of species: implications for species Ecology and Biogeography, 20,915–927. distribution modelling. Biological Reviews, 88,15–30. Rot-Nikčević, I., Sidorovska, V., Džukić,G.,&Kalezić, M. L. (2001). Wunsch, C., Schmitt, R. W., & Baker, D. J. (2013). Climate change as an Sexual size dimorphism and life history traits of two European intergenerational problem. Proceedings of the National Academy of spadefoot toads (Pelobates fuscus and P. syriacus) in allopatry and Sciences, 110,4435–4436. sympatry. Annales Series Historia Naturalis, 11,107–120. Zeisset, I., & Beebee, T. J. C. (2008). Amphibian phylogeography: a Schoener, T. W. (1968). The anolis lizards of Bimini: resource model for understanding historical aspects of species distributions. partitioning in a complex fauna. Ecology, 49,704–726. Heredity, 101, 109-119.