Ecological corridors for the amphibians and reptiles in the Natura 2000 sites of Romania

Landscape heterogeneity and fragmentation are key challenges for biodiversity conservation. As Earth’s landscape is increasingly dominated by anthropogenic land use, it is clear that broad-scale systems of nature reserves connected by corridors are needed to enable the dispersal of flora and fauna. The European Union currently supports a continent-wide network of protected areas, the Natura 2000 program, but this program lacks the necessary connectivity component. To examine whether a comprehensive network could be built in order to protect amphibians and reptiles, two taxonomic groups sensitive to environmental changes due to their physiological constrains and low dispersal capacity, we used species’ distribution maps, the sites of community interest (SCIs) in Romania, and landscape resistance rasters. Except Vipera ursinii rakosiensis, all amphibians and reptiles had corridors mapped that, when assembled, provided linkages for up to 27 species. Natura 2000 species were not good candidates for umbrella species as these linkages covered only 17% of the corridors for all species. Important Areas for Connectivity were identified in the Carpathian Mountains and along the Danube River, further confirming these regions as hot spots for biodiversity in Europe, where successful linkages are most likely. In the end, while such corridors may not be created just for amphibians and reptiles, they can easily be incorporated into more complex linkages with corridors for more charismatic species, therefore enhancing the corridors’ value in terms of quality and structure.


Results
Single species corridors. Corridor maps for each species are available in the Supplementary Material S2-Maps ( Figures S1-S41). The length of the mapped corridors ranged from 200 m to 114 km for amphibians, and from 200 m to 89 km for reptiles. Mean corridor length was ~ 7.6 km for amphibians and ~ 8.8 km for reptiles. Among amphibian species, Bufo bufo had the longest mapped continuous habitat corridor at 114 km, while the reptile with the longest link (89 km) was Natrix tessellata. Connectivity ranged from 33 to 80% for amphibians and from 0 to 100% for reptiles (Table 1). In the case of amphibians, Triturus dobrogicus had the lowest connectivity (33%), followed by Bombina bombina (36%), while the sites for Ichthyosaura alpestris were best connected by habitat corridors (80%). Among reptiles, vipers' sites were both the best and least connected: Vipera ammodytes ammodytes obtained a perfect connectivity between sites (100%), while V. ursinii moldavica and V. nikolskii Scientific Reports | (2020) 10:19464 | https://doi.org/10.1038/s41598-020-76596-z www.nature.com/scientificreports/ sites had among the lowest connectivity of all reptiles (22% and 18%); no sites were connected through habitat corridors for the other meadow viper (V. u. rakosiensis).

Multispecies corridors. Assembled, the corridors provided linkages for up to 27 species if all herpetofauna
was taken into account ( Fig. 1), 16 species when reptiles were considered separately (Supplementary Material S2-Maps, Figure S44) and 13 taxa for amphibians (Supplementary Material S2-Maps, Figure S43); 63% of the area covered by the corridors was shared by more than one species if all amphibians and reptiles were combined, 46% when amphibians were considered separately, and 53% for only reptiles. When taking into account only species from Annex II of the Habitats Directive, the maximum number of species in the habitat corridors was 6 for all amphibians and reptiles (Supplementary Material S2-Maps, Figure S45), 5 when assembled only for amphibians and 3 if assembled only for reptiles. The overlap between the corridors involving only Natura 2000 species and corridors where all species were included was 17%; the overlap between the corridors for the amphibians from Annex II and the corridors for all amphibians was 8%, and when analyzing reptiles, the overlap was ~ 20%. Connectivity between Natura 2000 sites was about 60% when links were considered for all herpetofauna species (Table 1) or only for amphibians, and 54% when connectivity was explored only for reptiles.

SCI connectivity.
When data from all herpetofauna taxa were pooled together, the number of connections for the protected areas ranged from 0 (no connectivity) to 126 connections (Fig. 2). SCIs that lacked any kind of connectivity for amphibians or reptiles were generally located outside the range of Carpathian Mountains, inside the Transylvanian Plateau and in southern Dobrogea, while the core areas with the greatest number of connections were located in the Carpathian Mountains and along the lower Danube River. A similar pattern was obtained when each group was treated individually, with isolated core areas located outside Carpathian Mountains, but the number of connections per core ranged from 0 to 66 for amphibians and from 0 to 62 for reptiles (Supplementary Material S2-Maps, Figures S47-S48).

Important areas for connectivity (IAC).
When pooled together for all amphibians and reptiles, the LCPs showed important areas for connectivity mainly in the Carpathian Mountains and sporadically along the Danube River (Fig. 3). In the Carpathians, most of the important areas for connectivity were identified in the Western Carpathian Range (and surrounding hills) and in the Eastern Carpathian Range, and some patches in Lotrului Mountains, in Retezat-Godeanu Massif and in Șureanu Mountains from the Southern Carpathians. Along the Danube River, the important areas for connectivity were located downstream from the city of Drobeta Turnu Severin, near Balta Mică a Brăilei Natural Parc and between Brăila and Galați.

Discussion
Our study explored whether habitat corridors could be a viable option for connecting the existing network of Romanian Natura 2000 sites for the amphibian and reptile species that occupy the sites. During the study design we elected to use only the fragments of the landscape that can be considered "natural" (such as forests, shrubs, rivers etc.) although animals, even amphibians and reptiles, are known to move through semi-natural or even heavily modified landscapes [73][74][75][76][77] . We had multiple reasons in choosing this approach. Foremost, natural habitats have exceptional value for both connectivity and biodiversity 78,79 -they maximize animal mobility, contain higher taxonomic diversity, higher genetic diversity, and meta-population retention, thus represent refugia for biodiversity. In contrast, areas impacted by human activities are viewed as unlikely to hold sufficient connectivity value for many key elements of biodiversity 78 . Human-modified landscapes are also susceptible to frequent land use changes, one of the primary sources of decline for both amphibians and reptiles 55,80,81 . It is generally agreed that avoiding degradation of existing habitat is a better strategy than restoration, as the latter implies increased costs and is unlikely to lead to full recovery 78,82 ; corridors generally imply substantial financial efforts and, as a result, their implementation has been met with resistance throughout Europe 83 . Human-modified landscapes could be included within the network in a stepping-stone manner 84 , but this would require a case-by-case analysis and consultations with the stakeholders, both of which are beyond the scope of this manuscript.

Natura 2000 connectivity and species migration corridors.
Our results showed that, at least for some SCIs and species, connectivity can be achieved using only natural patches of habitat; for other species, semi-natural areas will have to be taken into account, or landscape restoration will be necessary to promote connectivity. The taxa most affected by poor connectivity seem to be those with very specific habitat requirements, such as species at the edge of their distribution range or endemics (e.g. Eryx jaculus, Elaphe sauromates, V. u. moldavica, V. u. rakosiensis). Species occupying sites located at low elevations (e.g. Pelobates syriacus, Eremias arguta) also displayed poor connectivity, as those regions of the Romanian landscape have been affected by intense agricultural activities 68 . Some of the linkages mapped are accessible through a one-time dispersal event, while others would require multiple generations during which the species become corridor dwellers, but in fact this increases the overall occupancy and metapopulation connectivity of the landscape matrix 85 . The main issue is that, because of the high cost associated with allocating land for corridors, linkages for single species are an unrealistic undertaking, so possible alternatives are to (a) either find corridors common for as many species as possible, (b) use umbrella species to cover the needs of other species/groups or (c) merge the mapped corridors for amphibians and reptiles within existing/future planned linkages for other taxa. By examining the multi-species corridors, we observed that the Carpathian Mountains are the most likely landscape unit where these linkages can be created with great success. The Carpathians are a well-known hot spot for biodiversity in Europe 86 , with large continuous natural areas 87 , and projects involving connectivity in the Carpathians have been developed in recent years 88 . Areas along the Danube River also showed great connectivity,

Scientific Reports
| (2020) 10:19464 | https://doi.org/10.1038/s41598-020-76596-z www.nature.com/scientificreports/ where almost continuous corridors between SCIs were obtained (more so for amphibians than reptiles). These results are supporting the ongoing project to create the Lower Danube Green Corridor, which aims to enforce existing protected areas, create new ones, as well as restore natural floodplains 89 . Moreover, the corridors further emphasize the importance of riparian habitats as natural linkages 90 . Sites of community interest located within the Carpathian Mountains or along the lower Danube River were also likely to exhibit the highest degree of connectivity, as evidenced by our results, while many sites within intense anthropogenic settings lacked any kind of connectivity and will likely require habitat restoration measures or land use regulation policies to facilitate future corridors. Overall, more than half of the species of amphibians and reptiles analyzed were grouped in shared corridors (more than one species) when linkages were pooled together and core connectivity was above average, making the respective swathes of land more valuable for nature conservation, especially taking into account that those are exclusively natural fragments and so would necessitate only minor investments.
Conservation implications. Because the species listed in Annex II of the Habitats Directive are given higher priority (with more funds being allocated for their conservation) compared to taxa from other annexes, we explored the possibility of using them as "umbrella species" (sensu Breckheimer et al. 91 ) but the results were not very encouraging, as the corridors for the Natura 2000 species of herpetofauna covered only a small fraction of the corridors for all species. This can be explained by the fact that, with few exceptions, the species from the Annex II generally have specific habitat requirements and are themselves declining. Our method was to simulate corridors for each species of amphibian and reptile and combine them into general habitat corridors, but this will not be possible in other situations, such as working with tight deadlines, fixed budgets, an array of species from different groups, and much greater species diversity or little knowledge of the organisms for which connectivity is to be provided. Also, the topic of umbrella species remains intensely debated [92][93][94] with no right or wrong answer, mostly because species (even from the same group) can differ greatly in resource utilization, habitat requirements, and dispersal capability 92 . One convenient method would be to use a small number of widedispersing and habitat generalist species 91,95 as umbrella species in the context of connectivity, but some results have shown that, if less than three species are used, an indirect approach to connectivity using habitat characteristics is more effective 92 . A more viable method is to use a number of 5 96 to 9 92 surrogate species; however, our results show that even in this case the species should be chosen carefully, by applying dimensionality reduction or grouping 97 . We used 12 species from the Annex II of the Habitats Directive as surrogates for the Romanian herpetofauna, but the corridors only fitted the needs of a small number of species and the overlap with the multispecies corridor was reduced; the same was true even if the corridors were pooled based on taxonomic groups.
By pooling the corridors for all species we noticed that "intersections" would form where many corridors merged, these areas being located outside the boundaries of designated SCIs. As such, using the LCPs we isolated these "hot-spots" which we termed "Important Areas for Connectivity"; these are landscape features where the density of LCPs is the greatest and therefore the highest number of corridors pass through those specific locations. These features could be proposed as SCIs in the future, therefore creating connectivity in a "stepping-stone" manner 14 , or they could be given enhanced priority because of their importance in promoting linkages. Future research should compare these areas to those obtained for other species or taxonomic groups, with the goal of finding common ground that would further consolidate the corridor status of these areas and make them more likely to be adopted by national authorities.
Concluding remarks. Amphibians and reptiles are not the most popular model organisms for conservation 55,60 and studies on corridors generally take into account mammalian and avian taxa, with socalled "flagship species" (generally large carnivores) 98 being at the forefront. However, through their intrinsic characteristics, amphibians and reptiles make very good model organisms for connectivity analyses 65,70 . The landscape linkages obtained in this study raise the alarm on the fact that the complex network of sites created through the Natura 2000 program does not function like a system; there are sites and species that cannot be connected through areas of natural habitat, so these sites are of utmost importance and should be prioritized for conservation 78 . Some of the species that had low connectivity or no connectivity (e.g. Elaphe sauromates, V. u. moldavica, V. u. rakosiensis) are precisely those prioritized for conservation through the Habitats Directive. Moreover, these linkages can be used when mapping complex corridors (covering a range of plants and animals) or when corridors focused on large carnivores (the brown bear, the grey wolf and the Eurasian lynx) are created, as amphibian and reptile species value habitat structure over size 71 , therefore enhancing their quality and value and resulting in more species inhabiting the linkages 14,99 .
Last but not least, the IACs identified in our study show that there are areas of utmost importance to the functionality of amphibian and reptile communities (i.e. providing connectivity) that are outside any type of protection; these areas are composed entirely of natural blocks of land that, in the context of today's anthropogenic intervention, may soon disappear. These IACs could also be vital for amphibians and reptiles for adaptation under anthropogenic climate change, as they would allow populations to track their preferred microclimates 38,78 . Distribution data. We collated the baseline distribution for the herpetofauna of Romania from two peerreviewed papers 100,101 . The distribution data presented in the two papers were not publicly available, thus we manually georeferenced the distribution maps in ArcGIS 10.4 72 . The resulting distribution grids were further revised with records collected by three of us (T. C. S., A.S., and I.G.) (see Supplementary Material S3-Additional methods information for more data).

Methods
Core areas. We considered the Sites of Community Interest (SCIs) as core areas for mapping landscape linkages for amphibians and reptiles. Each site has a standard form where protected species are mentioned, but which is not always updated; we therefore chose to create an updated list for each site (see Supplementary Material S3-Additional methods information).

Environmental variables and resistance layers.
The resistance rasters used in the analysis were created through the traditional method of assigning scores to different variables based on experts' opinion regarding the ability of animals to negotiate certain environmental features [102][103][104][105] , with the modification that we used information from the species' current range (in the form of spatial statistics for each environmental variable) to reduce subjectivity. In ArcGIS, we extracted basic statistics (min, max, 1st percentile, 3rd percentile) of seven environmental variables (Table 2) from the species' distribution grids and established five classes for scoring: (C1) raster minimum to species' distribution minimum, (C2) minimum to 25%, (C3) 25-75%, (C4) 75% to species' distribution maximum, and (C5) species' distribution maximum to raster maximum. The goal was to score different values of an environmental variable depending on how readily available they are for the species, but only when also accounting for what is known about the species' current range through the distribution grids (see Supplementary Material S3-Additional methods information). The score values ranged from one to five, where one represented areas readily available for the species, and five signaled areas difficult to navigate (Supplementary Material S1-Resistance rasters scoring). A value of 1000 was used to show barriers or features which we considered impossible to pass. Some intervals included values outside the species' range in Romania but which are not outside the species' environmental limits; as such these areas were judged based on the known biology of the species and scored accordingly. Moreover, values from the first or the last interval almost always received a high (5) or a low (1) score, or a value of 1000 if they were considered barriers, because they usually reflect conditions towards the limits of what the species is able to tolerate. The landcover variable was treated separately as it consisted of categorical data; the human-modified landcover classes were given a score value of 1000 because the goal was to connect core areas using existing natural patches of habitat, resulting in 13 landcover categories capable of sustaining habitat corridors (Supplementary Material S1-Resistance rasters scoring). We reclassified the variables based on scoring categories in ArcGIS 10. Corridor model. We used the Linkage Pathways Tool of the Linkage Mapper Toolbox for ArcGIS 106 to create corridors for each of the 42 species of amphibians and reptiles included in the analysis. Similarly to other connectivity modeling tools, Linkage Mapper uses as inputs a set of core areas to be connected and a resistance layer. The result is a raster surface of cost-weighted distances normalized by least-cost paths (LCPs) between core areas. When working with large datasets, Linkage Mapper requires less processing time than alternative mapping software 90 . The toolbox allows the user to limit the maximum length of a corridor based on a cost-weighted or Euclidean distance or other user-defined criteria. First, we created a model without any limitations, but corridors were identified spanning very large distances (> 100 km) with unrealistic chances of connectivity. Therefore, on the second run we took into account a time-frame of 50 years as the maximum time span during which connectivity could be achieved, as data regarding the dispersal capacity of amphibians and reptiles in Romania were available from Popescu et al. 107 . Corridors were discarded if the LCPs intersected an intermediate core area; the number of neighboring core areas connected was not limited. Post-modeling, we clipped off corridors to a cutoff width equal to twice the size of the yearly dispersal range of species to include space for a main corridor and a buffer zone (see Supplementary Material S3-Additional methods information).

Multispecies corridors.
Post-mapping analysis explored the idea of multi-species corridors as linkages for individual species are difficult to justify and biological conservation is most effective when the strategies applied meet the needs of many species 108 . We reclassified the truncated corridors for each species to a value of one and we mosaicked linkages based on the following strategies: (1) corridors for all herpetofauna taxa; (2) corridors for all species of amphibians; (3) corridors for all species of reptiles; (4) corridors for the Annex II Natura 2000 species. Besides exploring the viability of multi-species corridors, we also analyzed the possibility of umbrella species, i.e. whether corridors for a limited number of species could cover the space necessary for a whole group.

SCI connectivity.
We also aimed to evaluate the contribution of the sites (SCIs) to the overall ecological network and we did this by counting the number of times each core area was successful in achieving a connection to www.nature.com/scientificreports/ another core area. We obtained the information from LCPs and we created connectivity counts for all taxa, and for amphibians and reptiles separately.   www.nature.com/scientificreports/ Important areas for connectivity. After creating the multispecies corridors, it became clear that some areas are very important for the dispersal of several taxa and when overlapping with the SCI network, we realized that those features were generally located outside protected areas. Therefore, using the least cost paths (LCPs) and the line density function in ArcGIS, we mapped areas of high value for the connectivity of herpetofauna.

Data availability
The datasets generated during and/or analyzed during the current study are not publicly available due to very large size, but are available from the corresponding author on reasonable request. The rest of data generated or analyzed during this study are included in this published article and its supplementary information files.