DISTRIBUTION AND CONSERVATION STATUS OF THE CAUCASIAN PARSLEY FROG, PELODYTES CAUCASICUS (AMPHIBIA: ANURA)

Pelodytes caucasicus inhabits Turkey, Georgia, Abkhazia, South Ossetia, Azerbaijan, and six regions of Russia (226 localities). The forest cutting strongly threatens its populations. Therefore, the frog is listed in Red Data Books of Georgia, South Ossetia, Azerbaijan, and the Russian Federation. Additional factors influencing the decline of P. caucasicus populations are destruction and contamination of suitable water bodies, clearing of forests from fallen trees, destroying of litter, mortality on roads, and preying by the introduced North American raccoon. Using of MaxEnt, we developed a species distribution model based on climate, landscape and land cover data to estimate the potential distribution range, ecological preferences and conservation status of P. caucasicus. Two precipitation parameters, annual precipitation and precipitation seasonality, had the highest contribution percentage to the model (52% and 11% respectively). As a rule, suitable habitats for the species located in woodland mountain areas with annual precipitation ranged from 513 mm to 2376 mm. Drier regions to the north and south of the Caucasus limit its distribution.


Introduction
the range of Mertensiella caucasica (Waga, 1876) in the Minor and Turkish Caucasus is basically coinciding with the distribution of P. caucasicus. However, the first species is absent in the Great Caucasus (Tarkhnishvili et al., 2008;Gül et al., 2018). Ommatotriton ophryticus (Berthold, 1846) occurs in North Armenia, but does not penetrate in Azerbaijan and the central part of the North Caucasus Van Riemsdijk et al., 2017). Lissotriton lantzi (Wolterstorff, 1914) inhabits the Dagestan Republic of Russia, Armenia and the Lenkoran Lowland in Azerbaijan, but is lacking in the Black Sea coastal part of Turkey (Skorinov et al., 2014). Triturus karelinii (Strauch, 1870) occurs in the Dagestan Republic of Russia, the Crimea, Azerbaijan, Iran, but is absent in the Black Sea coastal part of Turkey (Litvinchuk & Borkin, 2009;Wielstra et al., 2013a,b). The Caucasian part of ranges of Bufo verrucosissimus (Pallas, 1814) and Pelodytes caucasicus are approximately the same. However, the first species probably inhabits western and southern Turkey, eastern Bulgaria, Lebanon and Syria (Litvinchuk et al., 2008;Garcia-Porta et al., 2012;Arntzen et al., 2013;our data). The Caucasian lineage of Hyla orientalis Bedriaga, 1890 (Dufresnes et al., 2016), which could be recognised as a separate subspecies, is also associated with the Caucasian forest belt. However, unlike P. caucasicus it is widely distributed in the eastern part of the Caucasus and is lacking in the Black Sea coastal part of Turkey. The GIS-based ecological niche modelling is a rapidly developing area of research. Recently, numerous studies were made by using this method (Tarkhnishvili et al., 2008;Garcia-Porta et al., 2012;Dufresnes et al., 2016). The obtained models of species distribution could be useful for study of many aspects of ecology and conservation, predicting of new localities for rare and threatened species, and invasive species spread (Bombi et al., 2009;Lyet et al., 2013;Vences et al., 2017). This approach appears a good and reliable tool for past and future climatic scenario studies, too (Brito et al., 2011;Tarkhnishvili et al., 2012;Litvinchuk et al., 2013;Duan et al., 2016;Iannella et al., 2017;Wetterings & Vetter, 2018). Therefore, the aim of this study was to apply the ecological niche modelling to predict potential distribution range, estimate ecological preferences and conservation status of Pelodytes caucasicus.

Material and Methods
To predict the potential distribution of the Caucasian parsley frog, we modelled the species distribution using MaxEnt (ver. 3.3.3k; Phillips et al., 2006). This algorithm combines environmental parameters with geographic co-ordinates and produces highquality predictions of species distribution, often more reliable when evaluated and compared with other predictive models (Hernandez et al., 2006). For the contemporary niche predictions, we used 226 localities, comprising own and previously published records (Tosonoğlu & Taskavak, 2004;Zinenko & Goncharenko, 2009;Iğci et al., 2013;Kuzmin, 2013;Gül, 2014;Tuniyev, 2018). To avoid spatial autocorrelation of occurrence points, we filter them by ENMTools 1.3 (Warren et al., 2010).
We used layers with 30 arc seconds spatial resolution because Pelodytes caucasicus inhabits very heterogeneous mountain landscapes. Searching for new localities and for verification of previously published records, we visited Abkhazia, Georgia, South Ossetia, as well as Krasnodarsky Krai, Karachay-Cherkessia, Kabardino-Balkaria, North Ossetia-Alania, and Adygea republics of Russia in 2006-2014. Co-ordinates and altitude for previously published records (if they were not specified) were determined by use of Google Earth, taking into account known ecological preferences of the species.
A total of 18 variables were used (Table). Model performance was measured using the Area Under the Curve (AUC) derived from the Receiver Operating Characteristic (ROC) plots. AUC values range from 0.5 to 1.0, with 0.5 indicating no greater fit than expected by chance and 1.0 indicating a perfect model fit. AUC values above 0.75 are considered useful and above 0.90 very good (Swets, 1988;Elith, 2002). To properly parameterise the model, we evaluated the performance of various combinations of ten regularisation multipliers (from 0.5 to 5.0, in increments of 0.5; see details in Vences et al., 2017). The best-fit model was parameterised with a regularisation multiplier of 1.0 (30 replicates). We used default settings in MaxEnt, i.e. all feature classes, maximum iterations 500 and maximum number of background points 10000 (Phillips & Dudík, 2008). We applied a jackknife analysis for estimating the relative contributions of variables to the MaxEnt model.  Fig. 2 and the Appendix. About half of the localities (n = 108) were found in the North Caucasus (Russia), where a majority was revealed in its westernmost part (98 localities in Krasnodarsky Krai and Adygea). In the eastern part of the North Caucasus, records of the species were single (10). In Transcaucasia, localities of P. caucasicus were numerous in Georgia (56), northeastern Turkey (25), Abkhazia (21), and South Ossetia (12 localities). Records of the species in northwestern Azerbaijan were few (4).

All defined records of Pelodytes caucasicus have been summarised in
The MaxEnt model for Pelodytes caucasicus had robust evaluation metrics. The average test AUC for the replicate runs was evaluated as 0.939 and the standard deviation was 0.031. The predicted potential distribution is shown in Fig. 3. Estimates of relative contributions of variables to the species are shown in the Table. Of the parameters included in the model, annual precipitation and precipitation seasonality were variables with the highest percentage contributions (52% and 11% respectively). Other parameters had no notable contribution (less than 10%). Suitable habitats for the species were located in areas with annual precipitation ranging from 513 mm in several localities in Turkey to 2376 mm in Batumi botanical garden in Georgia (mean 1068.7 mm; SD = 313.3; Fig. 4). As a rule, the species was revealed in mountain areas with an altitude ranging from sea level to about 2300 m a.s.l. in Tskhra-Tskharo Pass in Georgia (mean 842.3; SD = 588.9; Fig. 4), only occasionally penetrating to adjacent plains in the North Caucasus (Goryachiy Klyuch) and the Colchis lowland (Kulevi and Poti). Drier regions to the north and south of the Caucasus limit the species distribution.

Discussion
The annual precipitation, which had the highest contribution to the MaxEnt model obtained, plays a very important role for the species survival. According to our and previously published data (Golubev, 1980;Gül, 2014), Pelodytes caucasicus inhabits very wet terrestrial biotopes (Fig. 5), such as shores and banks of ponds and streams with a completely shaded, dense vegetation (trees, bushes and grasses). Usually, the frog was found in dark humid broad-leaved and mixed coniferous-deciduous forests, and it was only sometimes found in more open coniferous forests, as well as subalpine and alpine zones (Golubev, 1985;Dzuev & Ivanov, 2000;Tuniyev & Tuniyev, 2012;Lotiev & Tuniyev, 2017;Tuniyev, 2018). Pelodytes caucasicus breeds in pools along river banks, backwaters and creeks located deep in the forests, far from its boundaries (Golubev, 1980;Kuzmin, 2013). We have compared the ecological preferences of Pelodytes caucasicus with other Caucasian forest amphibian species studied with MaxEnt. The precipitation parameter has the greatest influence on the range of Lissotriton lantzi only (Skorinov et al., 2014). The precipitation of the warmest quarter strongly influences (58.2%) the distributional pattern of the species. A notable contribution of the precipitation of the driest month (20%) was revealed for Ommatotriton ophryticus as well . In contrast, only the thermal parameter (isothermality) has the highest percentage contribution for Mertensiella caucasica (Gül et al., 2018).
The distributional range of Pelodytes caucasicus is precisely restricted, being limited to woodland areas (Tarkhnishvili & Gokhelashvili, 1999). For example, Tuniyev (1990) noted that populations from Lagodekhi and Zakataly regions in eastern Georgia and western Azerbaijan are isolated from the main range. However, our results of the MaxEnt modelling (Fig.  3) shows that these populations are connected with the main range. On the other hand, we have found that populations which inhabit the central part of the North Caucasus (Kabardino-Balkaria, North Ossetia-Alania and Chechnya republics of Russia) could be isolated. Moreover, some of the local records (Khabaz in Kabardino-Balkaria, Martanka and Gekhi rivers in Chechnya) are placed in regions with little suitable habitats (probability of occurrence is 0.02-0.07).
It is important to note that a locality from the vicinities of Kars City in eastern Turkey (number 220 in the Appendix) seems to be wrong, being located under unsuitable environmental conditions (probability of occurrence is 0.01), which confirms the opinion previously expressed by Franzen (1999).
Undoubtedly, Pelodytes caucasicus should be included in Red Data Books of all regions which it inhabits (including Abkhazia and Turkey). The species needs protection in several nature reserves within all regions where it was found (Darevsky, 1987;Kaya et al., 2009;Kuzmin, 2013). For conservation of the species, it is necessary to identify spawning water bodies and organise micro-reserves (Tarkhnishvili & Gokhelashvili, 1999;. Puddles and ditches along poorly exploited forest roads are often used by the Caucasian parsley frog as breeding sites (Golubev, 1980). Therefore, to enlarge the population of this species, the best solution would be digging of new water bodies in such places. Additionally, it is necessary to create shelters from fallen cripples and trees near breeding sites and, perhaps, organise protection of spawning sites against raccoons with use of mesh fences.