Abstract
Vipera ursinii graeca is a restricted-range, endemic snake of the Pindos mountain range in the southwestern Balkans. The subspecies was previously reported from eight localities in Greece and one locality in southern Albania. We used species distribution modelling based on climate data from known localities in Greece to estimate the potential distribution of the subspecies. The model predicted suitable areas for eleven mountains in southern Albania, which we visited in ten field expeditions in four years. Based on 78 live individuals and 33 shed skins, we validated the presence of the snake on eight of the eleven mountains. Six populations (Dhëmbel, Llofiz, Griba, Shendelli, Tomorr and Trebeshinë Mountains) are reported here for the first time. Morphological characters undoubtedly supported that all individuals found at these new localities belong to V. u. graeca. Genetic analysis of mitochondrial DNA sequences also confirmed the identity of the snakes as V. u. graeca and a low number of identified haplotypes suggested low genetic variability among populations despite significant spatial isolation. All localities were subalpine-alpine calcareous meadows above 1600 m. These high montane habitats are separated by deep valleys and are threatened by overgrazing, soil erosion, and a potential increase in the elevation of the tree line due to climate change. Our surveys increased the number of known populations by 60% and the known geographical range of the subspecies by approximately 30%. Our study serves as a baseline for further ecological research and for conservation measures for one of the least known European viperid snakes.
Introduction
An increasing number of species are threatened by habitat changes (Brooks and Kennedy, 2004) and those with restricted distributions face an elevated risk of extinction (Malcolm et al., 2006). Montane species are especially vulnerable in this regard, as their distribution is often not only restricted, but also highly fragmented and isolated (Ehrich et al., 2007). Therefore, precise information on their distribution is fundamental for the effective conservation of their populations (Johnson and Gillingham, 2005).
Taxa in the Vipera ursinii species complex (Acridophaga subgenus, Reuss, 1927, hereafter referred to as meadow vipers) are especially vulnerable, because most have a highly fragmented distribution, ranging from Eastern France to Western China (Nilson and Andrén, 2001). Members of the V. ursinii complex are among the most endangered viperid species in Europe (Nilson and Andrén, 2001). V. ursinii is a species listed in CITES (Appendix I), in the Bern Convention (Appendix II) and in the European Union’s Habitats Directive (Appendix II and IV), and also protected by national legislation in most of its distribution range, except in Albania, Bosnia-Herzegovina and Montenegro. Some taxa are classified in the IUCN Red List of Threatened Species as vulnerable (V. ursinii including all subspecies), endangered (V. u. rakosiensis) or critically endangered (V. u. moldavica).
Meadow vipers live in highly fragmented lowland steppes or subalpine-alpine grasslands. Much of their typical lowland steppe habitats have been lost due to human activity (e.g. crop production, grazing) in the 20th century and, as a consequence, meadow vipers have gone extinct in the lowland plains of Austria, Bulgaria and possibly Moldova (Nilson and Andrén, 2001). Subalpine-alpine meadow habitats are less affected by habitat alteration, even though many are used for intensive grazing by sheep, cattle and goats. The remaining European populations survive in small, isolated habitat patches that are often suboptimal for long-term persistence (Nilson and Andrén, 2001; Újvári et al., 2002; Filippi and Luiselli, 2004; Edgar and Bird, 2005).
Meadow vipers typically exhibit a low level of morphological variability between taxa, thus, proper identification of specimens may be difficult, especially in the field (Nilson and Andrén, 2001). The phylogeny of meadow vipers has not been fully resolved (Ferchaud et al., 2012; Gvoždik et al., 2012; Zinenko et al., 2015), due to limited knowledge on certain taxa and inadequate sampling of taxa or geographical regions. In the last two decades, some of the formerly recognised subspecies of the complex were elevated to full species status based on morphological, allozyme or immunological analyses (V. eriwanensis: Nilson et al., 1995; V. renardi: Kotenko et al., 1999; Nilson and Andrén, 2001; V. anatolica and V. ebneri: Nilson and Andrén, 2001).
The Greek meadow viper (Vipera ursinii graeca Nilson and Andrén, 1988) is one of the rarest and least-known meadow vipers. V. u. graeca represents a distinctly divergent evolutionary lineage sister to all other clades of V. ursinii (Ferchaud et al., 2012; Zinenko et al., 2015), and therefore, a possible candidate for full species status. All the more so because other, less divergent taxa have been recognised as full species within the complex. V. u. graeca is considered endemic to the Pindos mountain range in Greece (Nilson and Andrén, 1988) and southern Albania (Korsós, Barina and Pifkó, 2008; Mizsei and Üveges, 2012). Greek meadow vipers inhabit high elevation subalpine-alpine meadows, which normally occur above the tree line, beginning at altitudes of 1600 m a.s.l. and extending upwards of 2200 m unless meadows are interrupted by rocky slopes and peaks (Nilson and Andrén, 1988; Mizsei and Üveges, 2012).
Prior to our investigation, V. u. graeca was known from only eight mountains in Greece (Dimitropoulos, 1985; Nilson and Andrén, 1988, 2001; Chondropoulos, 1989) and had one record in southern Albania (Korsós, Barina and Pifkó, 2008). Former studies mentioned V. u. macrops as the only subspecies occurring in Albania (Kopstein and Wettstein, 1920; Bruno, 1989; Haxhiu, 1998) until Korsós, Barina and Pifkó (2008) reported the occurrence of V. u. graeca from one site in Nemerçkë Mountain in southern Albania on the basis of a single photo. This observation raised the possibility that V. u. graeca also occurs on other mountains of southern Albania.

Geographic area and presence points of Vipera ursinii graeca used in species distribution modelling. Grey shading indicates the area for model training, black outline indicates projection area. The distribution of V. u. macrops is also presented due to its proximity to the study taxon.
Citation: Amphibia-Reptilia 37, 1 (2016) ; 10.1163/15685381-00003031

Geographic area and presence points of Vipera ursinii graeca used in species distribution modelling. Grey shading indicates the area for model training, black outline indicates projection area. The distribution of V. u. macrops is also presented due to its proximity to the study taxon.
Citation: Amphibia-Reptilia 37, 1 (2016) ; 10.1163/15685381-00003031
Geographic area and presence points of Vipera ursinii graeca used in species distribution modelling. Grey shading indicates the area for model training, black outline indicates projection area. The distribution of V. u. macrops is also presented due to its proximity to the study taxon.
Citation: Amphibia-Reptilia 37, 1 (2016) ; 10.1163/15685381-00003031
Given the scarcity of information about the distribution of V. u. graeca and uncertainty surrounding the identity of Vipera ursinii populations in southern Albania, our study had three aims: (i) to model the potential distribution of V. u. graeca across the Balkan Peninsula using species distribution modelling (SDM); (ii) to empirically test the model results by field surveys in the potential habitats identified by the SDM exercise with the aim of detecting meadow viper populations, and (iii) to confirm the identity of meadow vipers in southern Albania using morphological and molecular data.
Material and methods
Species distribution modelling
We used MaxEnt (Phillips, Anderson and Schapire, 2006) to predict the potential distribution of Vipera ursinii graeca. MaxEnt is a species distribution modelling approach that performs consistently comparable with the highest performing methods (Elith et al., 2006; Merow, Smith and Silander, 2013) even when only a few data points are available for modelling (Hernandez et al., 2006; Wisz et al., 2008).
Presence data of V. u. graeca were collected from scientific literature (Dimitropoulos, 1985; Nilson and Andrén, 1988, 2001; Chondropoulos, 1989) and personal communications (see Acknowledgements). At the time of creating our model (2010) only one presence data was known from Albania (Korsós, Barina and Pifkó, 2008), which we considered uncertain due to the lack of detailed morphological measurements, the reported extraordinary size of the specimen and contradictions with previous findings (Kopstein and Wettstein, 1920; Bruno, 1989; Haxhiu, 1998), which reported only V. u. macrops form Albania. Therefore, we omitted this occurrence point from the model. In total, we used 20 non-overlapping occurrence points from eight mountains in the model, including all published locations of the subspecies from Greece (fig. 1): Koziakas (), Lakmos (Peristeri) (), Metsovon (), Oiti (), Tsouka Karali (), Tzoumerka (Athamanika) (), Tymfristos (), Vardoussia (Akamanika) (). Because geographic coordinates were not available, we georeferenced these presence data using satellite imagery (Google Earth, Google Inc.) by choosing a random point within habitats based on descriptions and elevations above sea level as given in the literature. This method has been used before to accurately establish coordinates for sensitive, e.g. endangered, taxa for which accurate occurrence data are not typically given in the literature (Boakes et al., 2010). All data were entered and stored in a GIS using Quantum GIS 2.6 (http://www.qgis.org/), which was also used to prepare data for analysis and to visualise results.
We used the MaxEnt 3.3.3e (Phillips, Anderson and Schapire, 2006) software for modelling the potential distribution based on basic Bioclim variables (Busby, 1991). Climate data were obtained from the WorldClim database, version 1.4 (http://www.worldclim.org, Hijmans et al., 2005), at a resolution of 30 arc seconds. We selected climatic variables based on the ecology of alpine reptiles (Monasterio et al., 2009) and on correlation tests between potential predictors. Although MaxEnt is more robust in controlling for correlations between variables than stepwise regression (Elith et al., 2011), strongly correlated variables () were excluded to avoid multicollinearity in the models (Stohlgren et al., 2010; Elith et al., 2011). For example, altitude appeared to be an important predictor, because the known habitats of V. u. graeca are restricted to elevations above the tree line, we excluded it from the model due to its strong correlation with BIO1. We selected four variables for modelling: (i) annual mean temperature (BIO1), (ii) temperature seasonality (BIO4), (iii) annual mean precipitation (BIO12) and (iv) precipitation seasonality (BIO15). Temperature is a limiting factor for the annual activity period of ectothermic animals. Precipitation was considered important due to its strong influence on the vegetation type and structure.
Since all collected points of presence were located in Greece, we first used the area of that country as a background for model training to counterbalance our sampling bias. In the second step, the model was projected to the Balkan Peninsula (Croatia, Bosnia and Herzegovina, Montenegro, Serbia, Former Yugoslav Republic of Macedonia and Albania and the western two-thirds of Bulgaria; fig. 1). The discrimination ability of the model is tested by the Area Under Curve (AUC) metric, which is a measure of model performance in correctly distinguishing between sites associated to presence and background. The value of AUC varies between 0.0 and 1.0, where 1.0 is considered perfect prediction and 0.5 or less is considered no better than random (Fielding and Bell, 1997; Franklin, 2009). We calculated the importance of predictor variables by jackknife tests using MaxEnt. MaxEnt was run using the default settings except the fade by clamping function, which was applied for projection. Logistic output was used to visualise the climate suitability for the studied subspecies. We calculated the planimetric area of the predicted suitable habitats using Quantum GIS 2.6 (table 1).
Model testing by field surveys
Based on the model predictions (see Results section), we selected 10 separate mountains in southern Albania to test the subspecies distribution model and search for previously unrecorded populations (fig. 2). None of these mountains were known to harbour meadow viper populations prior to our study. Furthermore, we included the Nemerçkë Mountain in our field surveys in order to confirm the existence of V. u. graeca. We visited these potential habitats between 2010 and 2014 (table 1).
With consideration to the low detectability of vipers in the field, we also estimated detection probability to quantify how long one must search in order to obtain proof of presence of the subspecies and to eventually prove the model right or wrong for a given locality. We estimated the detection probability of vipers by tracking the movements of observers in potential habitats using GPS receivers (Garmin Ltd. Schaffhausen, Switzerland) as the searches were being made. We recorded track length, as well as the number of individual meadow vipers found on each track (Garmin BaseCamp 4.4.2). To test how track length predicts detection probability, we used a generalised linear mixed model (GLMM) with binomial error distribution, constructed using the lme4 package (Bates et al., 2014) in the R statistical environment (R Core team, 2013). Presence-absence data were used as the binary dependent variable, while the length of track was included as fixed explanatory variable. Observer and locality (mountain) were included in the GLMM as random factors to control for the non-independence of observations in these categories.
Subspecific identification
When we found a viper, the date, time, GPS coordinates of capture, sex, total body length and body mass were recorded in the field; furthermore, all specimens were photographed in order to collect more morphological data (scale counts) in the laboratory. For morphological identification, only dorsal scales were counted on site to minimize the time of handling and disturbance of the snakes. Tissue samples (a small piece from a ventral scale or tip of the tail) were also collected from each individual for genetic analyses and stored in 96% ethanol. After handling, animals were released where they were caught. In addition, shed skins of the target subspecies were also collected, which were insufficient for morphological measurements, but useful for genetic data collection. Morphological identification was determined based on the subspecific description (Nilson and Andrén, 1988). Meristic characters (preventrals, ventrals, subcaudals, apicals, supralabials, sublabials, circumoculars, loreals, crown scales) were counted using the photos in the program ImageJ 1.47v (Rasband, 2012).
Localities (mountains) predicted by species distribution modelling and visited in this study with details and results of surveys to detect Vipera ursinii graeca in southern Albania.



The prediction of MaxEnt model for Vipera ursinii graeca throughout the range and in southern Albania (inset). Field surveys were conducted on all mountains shown. Filled acronyms indicate mountains where the subspecies was found, open letters indicate those where no evidence of presence was found. Abbreviations: Bureto (BU), Çika (CI), Dhëmbel (DH), Griba (GR), Kulmak (KU), Llofiz (LL), Lunxhërisë (LU), Nemerçkë (NE), Shendelli (SH), Tomorr (TO), Trebeshinë (TR).
Citation: Amphibia-Reptilia 37, 1 (2016) ; 10.1163/15685381-00003031

The prediction of MaxEnt model for Vipera ursinii graeca throughout the range and in southern Albania (inset). Field surveys were conducted on all mountains shown. Filled acronyms indicate mountains where the subspecies was found, open letters indicate those where no evidence of presence was found. Abbreviations: Bureto (BU), Çika (CI), Dhëmbel (DH), Griba (GR), Kulmak (KU), Llofiz (LL), Lunxhërisë (LU), Nemerçkë (NE), Shendelli (SH), Tomorr (TO), Trebeshinë (TR).
Citation: Amphibia-Reptilia 37, 1 (2016) ; 10.1163/15685381-00003031
The prediction of MaxEnt model for Vipera ursinii graeca throughout the range and in southern Albania (inset). Field surveys were conducted on all mountains shown. Filled acronyms indicate mountains where the subspecies was found, open letters indicate those where no evidence of presence was found. Abbreviations: Bureto (BU), Çika (CI), Dhëmbel (DH), Griba (GR), Kulmak (KU), Llofiz (LL), Lunxhërisë (LU), Nemerçkë (NE), Shendelli (SH), Tomorr (TO), Trebeshinë (TR).
Citation: Amphibia-Reptilia 37, 1 (2016) ; 10.1163/15685381-00003031
Vipera ursinii graeca specimens sampled and sequenced from southern Albanian mountains and additional mtDNA ND4 sequences retrieved from GenBank from Ferchaud et al. (2012) from Greek localities.


Climatic variables used in the species distribution model.


Because some of the mountains predicted by the SDM as potential occurrences of the subspecies are relatively close (c. 80 km) to the southernmost known occurrences of V. u. macrops, we also used genetic data to confirm subspecies identity. We sequenced the mitochondrial DNA marker NADH dehydrogenase subunit 4 (ND4) of 20 specimens collected from Albania (table 2) since it had been successfully used to discriminate V. ursinii subspecies (Ferchaud et al., 2012). Three additional ND4 sequences of V. u. graeca specimens, published by Ferchaud et al. (2012), were used from GenBank. We used the DNeasy Blood & Tissue Kit (Qiagen) and the NucleoSpin Tissue kit (Macherey-Nagel) for extracting genomic DNA. Polymerase chain reaction (PCR) conditions followed the protocol of Ferchaud et al. (2012). PCR products were cleaned with High Pure PCR Product Purification Kit (Roche) or on NucleoFast 96 PCR plates (Macherey-Nagel) using vacuum filtering. We used BigDye v1.1 for cycle sequencing reactions. DNA sequencing was performed on an ABI 3130 (xl) capillary sequencer (Life Technologies). Sequences were assembled and aligned using CodonCode Aligner v5 (CodonCode Corp.), and chromatograms were checked manually in order to clean the sequences. Sequences were then deposited in the European Nucleotide Archive (ENA, http://www.ebi.ac.uk/ena) (table 2). A median-joining haplotype network was calculated with the software Network v4.6 (Bandelt et al., 1999).
Results
Potential distribution
The species distribution model predicted suitable habitats for V. u. graeca in treeless alpine meadows and mountaintops in the Pindos mountain range in Greece and Albania (fig. 2) with high support values (AUC = 0.997). The distribution predicted by the model fits well with the known occurrences of the subspecies but also identified additional potential habitats in Greece. More importantly for this study, the model predicted potential occurrences for 11 mountains in Albania, ten of which meadow vipers were not known from (fig. 2).
The analysis of variable contribution showed that annual mean temperature (BIO1) had the highest importance (78%), whereas the precipitation variables (BIO12 and BIO15) contributed less, while seasonality of temperature (BIO4) did not contribute to the model at all (table 3). The permutation test showed an even higher importance of annual mean temperature (98%) to the model than the percentage of contribution by the same variable (table 3). The jackknife-regularized training provided additional support for this result, as BIO1 alone gave the most useful information for predicting the potential distribution of V. u. graeca, and the performance of the model was the lowest without BIO1.
Model testing by field surveys and discovery of new populations
During ten field expeditions in four years to the 11 mountains predicted by the model, we found one or more live individuals or shed skins of the Greek meadow viper on eight mountains (table 1, fig. 2). Six populations (Dhëmbel, Llofiz, Griba, Shendelli, Tomorr and Trebeshinë Mountains) are described here for the first time. We also verified the occurrence of the subspecies in Nemerçkë Mountain and found additional occurrences on Lunxhërisë Mountain (Mizsei and Üveges, 2012). Meadow vipers were not found on three of the 11 mountains visited (Çika, Bureto and Kulmak Mountains, table 1, fig. 2).
In total, we caught 78 vipers and collected 33 shed skins at eight different locations. During the field surveys, the observers covered a total of 457.2 km by walking. The probability of detecting at least one individual (or shed skin) increased with track length (GLMM, , , ). Detection probability reached 50% after 4.9 km distance covered by a single observer on one track. In localities where no meadow vipers were found, track length was only 3.3 ± 0.45 km (mean ± SE, tracks) due to logistical or weather constraints, and so the detection probability predicted for this distance was only 42.0%.
All of the visited Greek meadow viper habitats were characterised by a mosaic of open or closed grass and shrub communities formed on limestone. The annual mean temperature, averaged from BioClim data for all localities where the viper was observed, was 5.9 ± 0.5°C (mean ± SE), and meadows were partially covered by snow until mid-summer. The south-facing slopes were usually more open and rocky than north-facing slopes. The open grasslands were dominated by different Festuca, Poa and Sesleria species, and characteristic shrubs were Juniperus sabina, Daphne oleoides and Astragalus creticus. Most of the observed vipers were found close to shrubs or piles of stone in these south-facing habitat patches. In northern or north-eastern exposed slopes, open grasslands were replaced by closed grassland, where Agrostis species were dominant. All habitats were found to be used as sheep and goat pastures.
Subspecific identification
Specimens captured were identified as V. u. graeca based on the following morphological characteristics: number of ventrals (119-133), ventral base colouration white or pale brown, parietals frequently fragmented, generally fewer than 17 scale rows on posterior part of the body, 7 or 6 supralabial scales on one side of head, third supralabial scale always under the eye, posterior supralabials markedly smaller than anterior ones, no dark labial sutures, and nasal plate often partially divided (tables 4 and 5).
Quantitative characteristics of Vipera ursinii graeca from seven Albanian localities (mountains) where live specimens were found and from the terra typica (Lakmos, Greece). Values are given as ranges and means ± SE. Values for Lakmos based on Nilson and Andrén (1988) and additional data cordially provided by G. Nilson (pers. comm.).


All obtained mtDNA ND4 gene sequences (, 683 bp aligned length) were very similar, most were identical, including the reference V. u. graeca sequences from GenBank, which confirmed the morphological identification (fig. 3). The observed haplotype diversity was limited (fig. 3): sixteen (70%) of our ND4 sequences shared the same haplotype with those V. u. graeca sequences published by Ferchaud et al. (2012), including the sequences which originated from the two shed skins collected in Griba Mountain (table 2). Four specimens belonged to two slightly different haplotypes, differing only in single nucleotide substitutions. The specimens from Shendelli Mountain had a single T → C substitution in position 222, while two of the four sequences from Lunxhërisë Mountain had a G → A substitution in position 384 (fig. 3).
Discussion
Our study verified the presence of the Greek meadow viper on eight mountains of southern Albania and expanded its known distribution by ca. 100 km to the north. The populations reported in this study double the number of populations previously known and the areas reported here comprise approximately one-third of the distribution area of the subspecies. We found evidence of the presence of Greek meadow vipers in seven previously unknown locations (including Lunxhërisë Mountain which has been mentioned in Mizsei and Üveges, 2012) and verified the former record of Korsós, Barina and Pifkó (2008) in Nemereçkë Mountain. The high proportion of the occurrences confirmed relative to the predicted ones (8 of 11 mountains or 73%) indicated that the prediction of our model was generally reliable for the northern part of the range of the subspecies.
Our distribution model for V. u. graeca based on macroclimatic factors (temperature, precipitation) performed well and highlighted the overwhelming importance of annual mean temperature. Annual mean temperature is known to have a direct effect on ectothermic organisms (Scali et al., 2011) and was recently demonstrated to play a key role in defining suitable habitat types for the closely related V. u. ursinii, while factors such as shelter or food availability have strong influence in determining the best patches within the habitat (Lyet et al., 2013). The availability of shelter and food, i.e., the vegetation and arthropod fauna of the suitable habitats, is in turn largely determined by precipitation (Marini et al., 2008), which may explain why annual precipitation and seasonality of the precipitation contributed more than 20% to our distribution model.
The prevalence of discriminating morphological characters in the examined Vipera ursinii graeca specimens.



Median-joining haplotype network of Vipera ursinii graeca based on mitochondrial ND4 sequences. Numbers 222 and 384 indicate the relative positions of single nucleotide substitutions. ∗See table 2.
Citation: Amphibia-Reptilia 37, 1 (2016) ; 10.1163/15685381-00003031

Median-joining haplotype network of Vipera ursinii graeca based on mitochondrial ND4 sequences. Numbers 222 and 384 indicate the relative positions of single nucleotide substitutions. ∗See table 2.
Citation: Amphibia-Reptilia 37, 1 (2016) ; 10.1163/15685381-00003031
Median-joining haplotype network of Vipera ursinii graeca based on mitochondrial ND4 sequences. Numbers 222 and 384 indicate the relative positions of single nucleotide substitutions. ∗See table 2.
Citation: Amphibia-Reptilia 37, 1 (2016) ; 10.1163/15685381-00003031
The match between model predictions and localities of observations suggested that the distribution model accurately predicted the macrohabitats such as mountain ranges, but did not differentiate at the microhabitat level such as among major plant associations. The microhabitats actually used by individuals varied among mountain ranges. For example, in the only habitat (Tomorr Mountain) where Juniperus species are present and common, vipers used this shrub for shade and hiding, similarly as V. u. ursinii uses junipers in France (Baron et al., 1966). On other montane habitats where junipers are rare or missing, we usually found vipers close to shrubs (mostly Astragalus spp., Daphne spp.) which they used similarly as in the case of junipers. However, in localities where shrubs are absent (e.g. Trebeshinë Mountain), probably due to long-term grazing activity, the snakes were frequently observed in habitat patches characterized by Festuca spp. tussocks. These observations suggest that the biotic characteristics of the vegetation on a fine spatial scale within suitable habitats and their effects on the microhabitat choice and activity of snakes should be taken into account in future ecological studies.
The fact that we did not detect V. u. graeca on three of the 11 mountains does not necessarily mean that the subspecies is not present there. Usually, perfect detection of a secretive snake species is not possible during short visits (Kéry, 2002; MacKenzie et al., 2002). Our analysis of detection also suggests that our effort was insufficient to exclude viper occurrence (or confirm viper absence) at these locations, thus, further field surveys at these sites are needed in order to confirm the presence or absence of populations (see Lyet et al., 2013).
Our morphological and genetic identification of all southern Albanian populations of meadow vipers confirmed them being V. u. graeca, as proposed by Korsós, Barina and Pifkó (2008). Although the morphological characteristics reported here agree well with the original description of V. u. graeca (Nilson and Andrén, 1988, 2001), we also found some inconsistencies: none of the specimens displayed a nasal scale fused with nasorostralia (table 5), and we found a broader range of numbers of ventral, subcaudal and crown scales (table 4). Such differences in pholidotic characters can possibly be attributed to geographic variation or to greater sample sizes in the current study than in previous ones. Furthermore, two specimens out of 72 displayed brownish ventral colouration, whereas six other specimens showed grey ventral colouration caused by a densely spotted pattern, which was not reported as being typical for V. u. graeca. The latter specimens were all small juveniles (mean SVL ± SE: 137.3 ± 3.20 mm), thus this colouration probably represents an adolescent character state.
Mitochondrial sequences showed that the sampled V. u. graeca populations have low ND4 haplotype diversity (fig. 3). Only three different ND4 haplotypes were found: N5 (Ferchaud et al., 2012) was detected in the entire distribution of V. u. graeca from the southernmost (Stavros Mountain, Greece) to the northernmost (Tomorr Mountain, Albania) population; LU (this study), was so far found only in Lunxhërisë Mountain, where the N5 haplotype was also present; and SH (this study) was found in the Shendelli Mountain population, the smallest known habitat for this taxon (table 1; 1.38 km2). These results do not warrant strong conclusions with regard to conservation genetics, but may show that the number of ND4 haplotypes can be lower in V. u. graeca than in other subspecies of V. ursinii, with the exception of the critically endangered V. u. moldavica (Ferchaud et al., 2012).
Despite our discovery of new populations, the known range of V. u. graeca is still narrowly confined to central and north-western Greece and southern Albania (high elevation meadows in the Pindos mountain range). Moreover, its potential distribution shows a highly fragmented pattern, with patch-like suitable locations separated from each other by tens of kilometres of unsuitable habitat. These potential populations are currently isolated by deep valleys with unsuitable forested habitats that most likely hinders gene flow between the high elevation meadows (>2000 m a.s.l.). This is especially relevant for a snake species that has limited movement capabilities, such as V. ursinii (Lyet et al., 2009). As V. u. graeca is constrained to specific habitats located on mountain tops, global warming could result in serious loss of its habitats, due to tree encroachment and upward movement of the tree line (Grace, Berninger and Nagy, 2002; Galbreath, Hafner and Zamudio, 2009). Extensive grazing could also be a significant threat as it can negatively influence the vegetation structure of habitats (Wilson, 1994; Beever and Brussard, 2004). Furthermore, the direct killing of snakes by local shepherds might also contribute to the decline of isolated and vulnerable populations.
With regard to these complex threats, the Greek meadow viper should be considered an important target for conservation research. Vital information for its conservation includes knowledge on the distribution in Greece and Albania, along with knowledge on habitats (vegetation types, grazing pressure, soil erosion etc.) including microhabitats (vegetation structure, food/shelter availability etc.) and population characteristics and dynamics (population genetics, local threat factors, sex ratio and demographic structure, dispersion ability, etc.). Increased conservation attention is also warranted by its phylogenetic distinctiveness from other members of the Acridophaga subgenus (Ferchaud et al., 2012). Future international cooperation between experts is strongly recommended to conserve Europe’s least known viperid snake.
Acknowledgements
We thank M. Schweiger and J. Spreyboeck for distribution data from Greece; P. Zaránd, G. Kulcsár and J. Jakó (Hungarian Land Rover Association) for logistic help, 24 volunteers of the Greek Meadow Viper Working Group for fieldwork; A.-L. Ferchaud, B. Halpern and J. Vörös for primers and sequences; O. Vincze, Z. Barina, Z. Korsós, G. Kardos and Y. Ioannidis for advice; G. Nilson for data on the population at the type locality and Paul Lambourne for proof-reading the manuscript. Financial support was provided by the CSEMETE Environmental and Nature Conservation Association; the Serpentes Foundation; the Department of Evolutionary Zoology, University of Debrecen; the Balassi Institute, Hungary (B2/1CS/11521/1, B2/1SZ/12851), the Hungarian Meadow Viper LIFE+ programme (LIFE07 NAT/H/000322), and a grant from OTKA (K106133). Permits were provided by the Ministry of Environment of Albania (Research Permit Request No. 6584). We also thank the reviewers, S. Roussos and P. Tarroso for their useful comments and suggestions.
References
Bandelt H.J., Forster P., Röhl A. (1999): Median-joining networks for inferring intraspecific phylogenies. Mol. Biol. Evol. 16: 36-48.
Baron J.P., Ferrière R., Clobert J., Saint Girons H. (1996): Life history of Vipera ursinii ursinii at Mont-Ventoux (France). Comptes rendus de l’Acadèmie des sciences. Sèrie 3, Sciences de la vie 319: 57-69.
Bates D., Mächler M., Bolker B.M., Walker S. (2014): lme4: Linear mixed-effects models using Eigen and S4. R package version 1.0-5. Available at http://CRAN.R-project.org/package=lme4.
Beever E.A., Brussard P.F. (2004): Community- and landscape-level responses of reptiles and small mammals to feral-horse grazing in the Great Basin. J. Arid. Environ. 59: 271-297.
Boakes E.H., McGowan P.J.K., Fuller R.A., Chang-qing D., Clark N.E., O’Connor K., Mace G.M. (2010): Distorted views of biodiversity: spatial and temporal bias in species occurrence data. PLoS Biol. 8: e1000385.
Brooks T., Kennedy E. (2004): Conservation biology: biodiversity barometers. Nature 431: 1046-1047.
Bruno S. (1989): Introduction to a study of the herpetofauna of Albania. Herpetol. Bull. 29: 16-41.
Busby J.R. (1991): BIOCLIM: a bioclimate analysis and prediction system. Plant Protection Quarterly (Australia).
Chondropoulos B.P. (1989): A checklist of Greek reptiles. II. The snakes. Herpetozoa 2: 3-36.
Dimitropoulos A. (1985): First records of Orsini’s viper, Vipera ursinii (Viperidae) in Greece. Ann. Mus. Goulandris 7: 319-323.
Edgar P., Bird D.R. (2005): Action plan for the conservation of the meadow viper (Vipera ursinii) in Europe. Convention on the Conservation of European Wildlife and Natural Habitats Standing Commitee, Council of Europe, Strasbourg, 32 pp.
Ehrich D., Gaudeul M., Assefa A., Koch M.A., Mummenhoff K., Nemomissa S., Consortium I., Brochmann C. (2007): Genetic consequences of Pleistocene range shifts: contrast between the Arctic, the Alps and the east African mountains. Mol. Ecol. 16: 2542-2559.
Elith J., Graham C.H., Anderson R.P., Dudík M., Ferrier S., Guisan A., Hijmans R.J., Huettmann F., Leathwick J.R., Lehmann A., Li J., Lohmann L.G., Loiselle B.A., Manion G., Moritz C., Nakamura M., Nakazawa Y., Overton J.M.M., Townsend Peterson A., Phillips S.J., Richardson K., Scachetti-Pereira R., Schapire R.E., Soberón J., Williams S., Wisz M.S., Zimmermann N.E. (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. Divers. Distrib. 17: 43-47.
Ferchaud A.-L., Ursenbacher S., Cheylan M., Luiselli L., Jelić D., Halpern B., Major Á., Kotenko T., Keyan N., Crnobrnja-Isailović J., Tomović L., Ghira I., Ioannidis Y., Arnal V., Montgerald C. (2012): Phylogeography of the Vipera ursinii complex (Viperidae): mitochondrial markers reveal an east-west disjunction in the Palaearctic region. J. Biogeogr. 39: 1836-1847.
Fielding A.H., Bell J.F. (1997): A review of methods for the assessment of prediction errors in conservation presence/absence models. Environ. Cons. 24: 38-49.
Filippi E., Luiselli L. (2004): Ecology and conservation of the meadow viper, Vipera ursinii, in three protected mountainous areas in central Italy. Ital. J. Zool. Suppl. 2: 159-161.
Franklin J., Miller J.A. (2009): Mapping Species Distributions: Spatial Inference and Prediction. Cambridge University Press, Cambridge, England.
Galbreath K.E., Hafner D.J., Zamudio K.R. (2009): When cold is better: climate-driven elevation shifts yield complex patterns of diversification and demography in an alpine specialist (American pika, Ochotona princeps). Evolution 63: 2848-2863.
Grace J., Berninger F., Nagy L. (2002): Impacts of climate change on the tree line. Ann. Bot. 90: 537-544.
Gvoždik V., Jandzik D., Cordos B., Rehák I., Kotlík P. (2012): A mitochondrial DNA phylogeny of the endangered vipers of the Vipera ursinii complex. Mol. Phylogenet. Evol. 62: 1019-1024.
Haxhiu I. (1998): The reptilia of Albania: species composition, distribution, habitats. Bonn. zool. Beitr. 48: 35-57.
Hernandez P.A., Graham C.H., Master L.L., Albert D.L. (2006): The effect of sample size and species characteristics on performance of different species distribution modeling methods. Ecography 29: 773-785.
Hijmans R.J., Cameron S.E., Parra J.L., Jones P.G., Jarvis A. (2005): Very high resolution interpolated climate surfaces for global land areas. Int. J. Climatol. 25: 1965-1978.
Johnson C.J., Gillingham M.P. (2005): An evaluation of mapped species distribution models used for conservation planning. Environ. Cons. 32: 1-12.
Kéry M. (2002): Inferring the absence of species: a case study of snakes. J. Wildl. Manage. 66: 330-338.
Kopstein F., Wettstein O. (1920): Reptilien und Amphibien aus Albanien. Weh. Zool.-Bot. Ges. 70: 387-457.
Korsós Z., Barina Z., Pifkó D. (2008): First record of Vipera ursinii graeca in Albania (Reptilia: Serpentes, Viperidae). Acta Herpetol. 3: 167-173.
Kotenko T., Morozov-Leonov S.Y., Mezhzherin S.V. (1999): Biochemical genetic differentiation of the steppe viper (Vipera ursinii group) in Ukraine and Romania. In: 10th Ordinary General Meeting of the Societas Europaea Herpetologica, p. 88-90. Natural History Museum of Crete, Irakleio.
Lyet A., Cheylan M., Prodon R., Besnard A. (2009): Prescribed fire and conservation of a threatened mountain grassland specialist: a capture-recapture study on the Orsini’s viper in the French Alps. Anim. Conserv. 12: 238-248.
Lyet A., Thuiller W., Cheylan M., Besnard A. (2013): Fine-scale regional distribution modelling of rare and threatened species: bridging GIS tools and conservation in practice. Divers. Distrib. 19: 651-663.
MacKenzie D.I., Nichols J.D., Lachman G.B., Droege S., Andrew Royle J., Langtimm C.A. (2002): Estimating site occupancy rates when detection probabilities are less than one. Ecology 83: 2248-2255.
Malcolm J.R., Liu C., Neilson R.P., Hansen L., Hannah L. (2006): Global warming and extinctions of endemic species from biodiversity hotspots. Conserv. Biol. 20: 539-548.
Marini L., Fontana P., Scotton M., Klimek S. (2009): Vascular plant and Orthoptera diversity in relation to grassland management and landscape composition in the European Alps. J. Appl. Ecol. 45: 361-370.
Merow C., Smith M.J., Silander J.A. (2013): A practical guide to MaxEnt for modeling species’ distributions: what it does, and why inputs and settings matter. Ecography 36: 1058-1069.
Mizsei E., Üveges B. (2012): Novel defensive behaviours of both sexes of Vipera ursinii graeca (Serpentes: Viperidae). Herpetol. Notes 5: 481-483.
Monasterio C., Salvador A., Iraeta P., Díaz J.A. (2009): The effects of thermal biology and refuge availability on the restricted distribution of an alpine lizard. J. Biogeogr. 36: 1673-1684.
Nilson G., Andrén C. (1988): A new subspecies of the subalpine meadow viper, Vipera ursinii (Bonaparte) (Reptilia, Viperidae), from Greece. Zool. Scr. 17: 311-314.
Nilson G., Andrén C., Joger U. (1993): A re-evaluation of the taxonomic status of the Moldavian steppe viper based on immunological investigations, with a discussion of the hypothesis of secondary intergradation between Vipera ursinii rakosiensis and Vipera (ursinii) renardi. Amphib.-Reptil. 14: 45-57.
Nilson G., Andrén C. (2001): The meadow and steppe vipers of Europe and Asia – the Vipera (Acridophaga) ursinii complex. Acta Zool. 47: 87-267.
Phillips S.J., Anderson R.P., Schapire R.E. (2006): Maximum entropy modeling of species geographic distributions. Ecol. Model. 190: 231-259.
R Core Team (2014): R: a Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.
Rasband W. (2012): ImageJ. Version 1.47t. Available at http://imagej.nih.gov/ij/ (accessed 25 April 2013).
Reuss T. (1927): Sechs europäische Giftschlangengattungen. Zool. Anz. 73: 124-129.
Scali S., Mangiacotti M., Sacchi R., Gentilli A. (2011): A tribute to Hubert Saint Girons: niche separation between Vipera aspis and V. berus on the basis of distribution models. Amphib.-Reptil. 32: 223-233.
Stohlgren T.J., Ma P., Kumar S., Rocca M., Morisette J.T., Jarnevich C.S., Benson N. (2010): Ensemble habitat mapping of invasive plant species. Risk Anal. 30: 224-235.
Sweets J.A. (1988): Measuring the accuracy of diagnostic systems. Science 240: 1285-1293.
Újvári B., Madsen T., Kotenko T., Olsson M., Shine R., Wittzell H. (2002): Low genetic diversity threatens imminent extinction for the Hungarian meadow viper (Vipera ursinii rakosiensis). Biol. Conserv. 105: 127-130.
Wilson S.D. (1994): The contribution of grazing to plant diversity in alpine grassland and heath. Aust. J. Ecol. 19: 137-140.
Wisz M.S., Hijmans R.J., Peterson A.T., Graham C.H., Guisan A. (2008): Effects of sample size on the performance of species distribution models. Divers. Distrib. 14: 763-773.
Zinenko O., Stümpel N., Mazanaeva L., Bakiev A., Shiryaev K., Pavlov A., Kotenko T., Kukushkin O., Chikin Y., Duisebayeva T., Nilson G., Orlov N.L., Tuniyev S., Ananjeva N.B., Murphy R.W., Joger U. (2015): Mitochondrial phylogeny shows multiple independent ecological transitions and northern dispersion despite of Pleistocene glaciations in meadow and steppe vipers (Vipera ursinii and Vipera renardi). Mol. Phylogenet. Evol. 84: 85-100.
Footnotes
Associate Editor: Sylvain Ursenbacher.