Fine-scale habitat suitability and connectivity analysis for the core populations of Yellow-spotted mountain pond-breeding newt (Neurergus derjugini) in the west of Iran and east of Iraq

The evolutionary characteristics of amphibians, especially those more water-dependent, have faced the study of habitat connectivity the challenge of Movement Context (MC), a context for which water and moisture are an integral part. This ne-scale study was conducted to evaluate the distribution of N. derjugini and identify potential displacement paths between the population cores, using the physical elements of the landscape playing an MC role. The distribution modeling was performed by MaxEnt and the inverse of the habitat suitability map was used as the cost map. The connection of population cores was modeled by the Linkage Mapper toolbox. The MC of the population cores was set in the drainage basins regarding the slope position and landform classes. Three scenarios were considered. In Scenario 1, drainage basins, in Scenario 2, the valleys, and in Scenario 3, canyons, shallow valleys, headwaters, and u-shaped valleys were considered MCs. According to the results, slope and altitude variability had the highest effect on the distribution of N. derjugini. The results showed that the proxy did not work well in Scenario 1. Compared to Scenario 2, the connection paths of Scenario 3 were more compatible with the ecological and biological characteristics of the species. ne-scale modeling could lead to reliable results for the displacement of MC-dependent species. The connectivity of the physical structure of landscapes plays a critical role in the connectivity of population cores. Proper distribution of population cores adjacent to each other in valleys and waterways can be considered a successful step in communication.


Introduction
With changes in the landscape, the protection of endangered species depends on a clear awareness of the species and their habitat.Human manipulation in the wilderness areas has led to a sharp decline in many wildlife species' populations and distribution ranges (Mills, 2012; Paudel and Kindlmann, 2012; Morrison et al., 2012).Amphibian populations are currently declining dramatically in most parts of the world (Greenberg et al., 2018), and more than 2,000 amphibians are at risk of extinction (IUCN, 2018).Species with a small distribution range do not show diverse responses to climate change due to local adaptations, so they will have little chance of adapting to different environmental changes (Rehm et al., 2015).Moreover, many habitats are not covered by protected areas (Nori et al., 2015), so they experience local extinctions within their distribution range (Hecnar and M'Closkey, 1996;Trenham, 2003).The restrictions of the species on access to new habitats can largely be attributed to their physiological characteristics.During their life cycle, most amphibian species depend on aquatic environments in the stages of reproduction, spawning, larvae, and metamorphosis, and terrestrial environments for distribution and feeding.Therefore, it would be of utmost importance to keep the connection between aquatic and terrestrial habitats (Becker et al., 2007).Reducing genetic diversity in a fragmented landscape (Pittman et al., 2014) reduces genetic connections, perpetuates population threats, and reduces the adaptive potential of species (Cameron et al., 2019).Amphibians are characterized by two common characteristics of inactivity and loyalty to the site of birth (Duellman and Trueb, 1994); thus, these species are generally believed to have little mobility and are often metapopulations (Marsh and Trenham, 2001).Meta-population explains population dynamics in a fragmented landscape.It is essential to identify the environmental factors that affect population connectivity.These factors can impede gene ow between populations and threaten the survival of the remaining populations in the future (Cushman et al., 2012).Protecting the species' populations in a fragmented landscape requires knowledge of their distribution extent (Campos et al., 2017).Ecological nest models (ENMs), also known as species distribution models (SDMs) (Campos et al., 2020), are one of the e cient tools to gain knowledge of the distribution range of a species.SDM includes methods that, using a mathematical algorithm, establish the relationship between presence data and environmental variables to introduce ecological niches or distribution ranges (Fišer, 2012;Peers et al., 2014).SDMs are used on broad-and ne scales.Fine-scale studies allow the physical and biological characteristics of species to be revealed using distribution models, as well as identify necessary management actions that are not achievable on a broad scale (Soares and Brito, 2007) because it is on a small scale that conservation measures can be executed not over the distribution range of species (Ceballos et al., 2017).Therefore, examining the species with a limited displacement range can provide valuable information.By knowing the species distribution range, habitat connection can also be examined.
The assumption that individuals in transit pathways pass through suitable habitats has been the subject of many landscape connectivity studies (Ahmadi et al., 2017;Iannella et al., 2021;Eslamlou et al., 2022).
Ecological corridors preserve gene ow and genetic diversity by facilitating the exchange of individuals between populations (Iannella et al.,2021).Ecological cohesion is a structural feature of the landscape that is generally de ned as the degree to which the landscape can facilitate the movement of species between their habitats (Urban and Keitt, 2001).Landscape characteristics affect the functional connectivity of habitats, migration, and gene ow of individuals within populations (Keller et al., 2015).These structures can be living, non-living, natural, or man-made (anthropogenic), or as a result of the creation of compounds such as non-human, bio-human, natural and non-living, and nally, natural and biological (Goudarzi et al., 2019).The concept of physiographic units can be considered one of the main focal points de ning the relationships between different variables with landscape (Bailey, 2009;Miliaresis and Argialas, 1999).Landforms are one of the most important non-living structures in shaping many important landscape processes (Corenblit et al., 2007).However, in the connectivity study of the population cores of salamanders with a small size and high dependence on water and wet areas, the role of this important physical parameter as an MC in landscape connectivity has received less attention than other landform structures, and N. derjugini is no exception to this rule (Afroosheh et al., 2016).
The genus Neurergus has 4 different species (Shari and Afroosheh, 2014), each of which has a speci c distribution range.Neurergus salamanders are scattered in northeastern Iraq and western Iran and are part of the Zagros fauna.N. derjugini is an important species of this genus, which according to the IUCN classi cation, is a Critically Endangered (CR) species.It has traveled from Europe to Iran throughout its evolutionary history and spread over time to its current distribution range (Öz et al., 2002).At present, the Zagros Mountains, as the distribution area of this species, play a vital role in preserving the remaining populations.Therefore, studying the distribution of species and the connections of the population cores in the species distribution range can provide a clear view for better managing the remaining populations.
Various studies have been performed on the landscape connections (Hosseini et al., 2019;Ahmadi et al., 2017), and among the studies on salamander species, we can mention the following studies.Gherghel and Papeş (2015) studied the distribution pattern and the connection path of the two populations in the west and east of the Danube Crested Newt.The study designed the cost map based on elevation and slope classes and used cost distance analysis.According to the results, the landscape has facilitated further the connection between the western and the eastern populations.The study detected two migration routes along the Iron Gate Canyon and the Timiş -Cerna Gap, separated by the Carpathian Mountains.Ahsani et al. (2018) studied the habitat suitability of Salamandra infraimmaculata in Kurdistan Province, western Iran.The model was implemented with 25 species presence points, 19 climatic habitat variables, and 10 runs in MaxEnt.The results showed that the Bio4 variable had the greatest effect on the species' presence.It is necessary to develop an oak forest cultivation program for the western slope of the Zagros Mountains to protect this species.Vaissi and Shari (2019) conducted a study to identify suitable areas for the protection of Neurergus kaiseri.In this study, various criteria, including climatic and human variables and the maximum entropy model, were entered into the multicriteria decision-making analysis (MCDM).At the end of the study, suitable patches were identi ed to protect this species in its distribution range.Ashrafzadeh et al. (2019) studied the habitat suitability and landscape connectivity for Neurergus kaiseri under the different climate change scenarios in 2050 and 2070.The species' habitat suitability in different climate change scenarios was investigated using different habitat modeling methods.Then, suitable habitat patches were identi ed using the TSS, and electrical circuit theory was used to connect the habitat patches.The results showed that the species' habitat suitability shifts towards higher geographical latitudes in climate change scenarios.Most of the studies on N. derjugini (Afroosheh et al., 2019;Khwarahm et al., 2021), N. kaiseri (Ashrafzadeh et al., 2019), and Salamandra infraimmaculata (Ahsani et al., 2018;Khwaram et al., 2021) have been using climatic variables with low spatial resolution.This is while the relationship between organisms and habitat variables at the bioclimatic scale is approximately 1000 times larger than the living surface (Beck et al., 2012;Potter et al., 2013).
Accordingly, this study aimed to model the habitat suitability and species connection patterns using maximum entropy methods and electrical circuit theory, considering MC on a ne scale.

Study area and presence points
Selecting the modeling boundary was one of the most challenging parts of the study.Unlike broad-scale studies on this species (Shari et al.,2017;Vaissi et al., 2019;Malekoutian et al., 2020;Afroosheh et al., 2019), in this ne-scale investigation, the similar environmental conditions of the land landscape in the species distribution range caused the study areas to be small and not around the presence points that reveal the characteristics and differences of the land landscape.Therefore, the study boundary was chosen so as to maximize heterogeneity and include all potential habitats, taking into account the high cost of the model implementation.Accordingly, an area of 442.9724549 ha in Iran was considered the nal border area for modeling (Fig. 1).The points of presence were identi ed using eld monitoring, previous studies (Naja majd and Kaya, 2010; Afroosheh et al., 2016;Karami, 2021), and the information available in the Department of Environment of Kermanshah and Kurdistan Provinces.These points included most of the reported populations.

Habitat variables
Due to the presence of amphibians in terrestrial and aquatic habitats (Swanson et (Afroosheh et al., 2016;Shari et al., 2017), the variables of elevation, elevation variability, the distance from the rst-order streams, cluster hillshade, slope percentage, Normalized Different Vegetation Index (NDVI), Compound Topographic Index (CTI) (Gessler et al., 1995;Moore et al., 1993), Heat Load Index (HLI), and slope position (6 classes) were selected.The elevation map was obtained from https://dwtkns.comwith an accuracy of approximately 30 m.The slope percentage map was prepared from the elevation variable.The distance from the rst-order streams, due to its importance for the species (Vaissi et al., 2019;Afroosheh et al., 2016), was developed using the Surface and Hydrology commands of the Spatial Analyst Tools in ArcGIS10.4.1 by applying a threshold limit on the FlowAccumulation and the relevant instructions.The HLI map was obtained by Eq. 1, in which θ is azimuth in degrees.
HI=[1 − cos(θ − 45)] / 2 (1) The index range is between zero for the northeast direction (the coolest hillside) and one for the southwest direction (the warmest hillside) (McCune and Keon, 2002).The CTI is an expression of microclimate (Kramer -Schadt et al., 2013).The CTI is a function of both the slope and the upstream contributing area per unit width orthigonal to the ow direction.This index and HLI were prepared using Geomorphometry and Gradient Metrics toolbox.The K-Mean method cluster was used to prepare the hillshade index variable, and the output provided an integrated distribution of radiation to the terrain (Veronesi and Hurni, 2014).The areas with less radiation were darker, and the regions that received more radiation were brighter.The hillshade map was prepared by the "3 × 3" lter in Terrain Mapping software.The Topographic Position Index (Jenness, 2006) in the Topography Tools was used to obtain the slope position variable (Dilts, 2015).According to the classi cation of this command, the position of the land slope in the study area was divided into 6 classes, including valley, toe slope, at, midslope, upper slopes, and ridges.This variable was entered into the model as strati ed.The density of ridges was prepared as an independent variable from the land slope map.First, the slope position map was reclassi ed.The value of 1 was given to the class of ridges and 0 to the other 5 categories.Then, the density of this class was measured using the Focal Statistics toolbox.The NDVI variable was also prepared from Landsat 8 satellite images in the Google Earth Engine (Gorelick et al., 2017) captured from early 2020 to early 2022.All the variables mentioned were prepared in ArcGIS10.4.1 (ESRI, 2011).

Distribution
The maximum entropy algorithm in MaxEnt software was used to model the suitability and distribution range of the species (Phillips et al., 2006).The reason for choosing this model was "its less sensitivity to the number of presence points," "the resolution of input data (Merow and Silander, 2014;Wisz et al., 2008)", and "its high forecasting performance (Phillips et al., 2017; Du ot et al., 2018)".All mentioned made it one of the most robust models for presence-only data (Qin et al., 2020;Elith et al., 2011).Out of 71 points in the modeling process, 70% were considered for training and 30% for testing (Afroosheh et al., 2019).In the implementation of the model, the defaults of the entropy model were used.The relative importance of each variable was investigated using the Jackknife test.The model accuracy was assessed by the Area Under the Curve (AUC), related to the Receiver Operating Characteristic (ROC) curve.The AUC value ranges between 0.5 and 1.The value of 0.5 means the accuracy is not better than the random chance, and the values over 0.7 show acceptable performance.The values greater than 0.8 and 0.9 reveal the model's good and excellent performance, respectively.A correlation analysis was established between the input variables of each run, and the variables with a correlation value over 0.75 did not enter the modeling process.

Connectivity scenarios
Newts cannot travel long distances on land; therefore, the elevation (Igawa et al., 2011) and related parameters such as slope are considered barriers to their movement (Arntzen, 2003;Arntzen et al., 1997;Gherghel et al., 2015).The limited distribution can be attributed to the physiological characteristics of the species, such as the lack of lungs.Dryness of the gills and cutaneous respiration limit the possibility of moving long distances.The small body size can also be one of the reasons for the limited displacement of these species, as there is a relationship between the amphibians' displacement and their body size (Watling and Braga, 2015).To detect the role of topographic (Gherghel and Papeş, 2015) and hydrological factors (Säumel and Kowarik, 2010) in the connectivity of population cores, the boundaries of the basin were de ned so that all areas with a shared water ow are considered as a unit with potential for connecting population cores.The identi cation of the basins was performed using DEM and the ow direction map.The spatial position of the population patches was required to implement the connectivity model.Some studies use thresholds to connect habitat patches that have been the focus of attention of many scholars (Almasieh et  ).Since N. derjugini species, as pond-breeding newts, are present in multiple groups, to create the connecting cores, the home range area of the species with an approximate area of 230 km2 (Shari And Afroosheh, 2014) was used as a basis for forming the core area.A buffer with a length of 8.55 m was considered from each of the points of presence.The cost maps in all scenarios were prepared by reversing the habitat suitability maps (Almasieh et al.,2019;Ahmadi et al., 2017;Ashrafzadeh et al.,2019).Different scenarios were developed to investigate the connection between the population patches by the basins.
In Scenario 1, population cores and cost maps were entered into modeling without considering the physical structures of the landscape.Only the MCS connected the aquatic and terrestrial habitats on the ne scale (Allen, 2016) were selected for displacement in the subsequent connectivity scenarios.The physical structures of the landscape in question include speci c classes of the slope position and landform maps prepared from DEM by the toolbox of Topography Tools (Dilts, 2015).The different classes of the slope position map include valley, top slope, at, midslope, upper slopes, and ridges.The landform classes also include canyons/deeply incised streams; midsole drainages/shallow valleys, upland drainages/ headwaters, u-shaped valleys, plains, open slopes, upper slopes/ mesas, local ridges/hills in valleys, midslope ridges/ small hills in plains, and mountain tops/high ridges.The distribution of the presence points in different classes was rst investigated using the Extract Values to Points tool to select MC.Considering that high mountainous ridges are an obstacle to the spread and distribution of amphibians (Giordano et al., 2007), only those MCs with high water permeability (Allen, 2016) were selected.Accordingly, only the valleys were selected from the slope position map for modeling as scenario 2. The classes of 1 = canyons (deeply incised streams), 2 = midslope drainages (shallow valleys), 3 = upland drainages (headwaters), and 4 = u-shaped valleys were also selected from the landform map as scenario 3 (Shari And Afroosheh, 2014; Vaissi and Shari .,2019).The cost map was extracted only for these areas.Other "slope position and landform" classes were excluded from the analysis using the Set Null command in the ArcGIS10.4.1.

Electrical circuit theory and cost map
Identi cation of population spots for connection based on Electronic Circuit Theory and Least cost paths was performed in the Linkage Mapper toolbox (McRae and Kavanagh, 2011).Electronic Circuit Theory and least-cost paths analysis in the Linkage Mapper toolbox (McRae and Kavanagh, 2011) were used to identify the connection paths of the population cores.The model integrates the least cost analysis and electrical circuit theory.The applicability of electrical circuit theory in ecological problems is due to the similarities between ecological and electrical connections.This theory considers electrical nodes as habitat spots, the movement path of the individuals as current, and resistance between nodes as habitat paths or corridors.The quality of each linkage was determined by two metrics.The rst metric was the cost-weighted distance divided by the Euclidean distance (CWD:ED).The cost-weighted distance was considered equal to the Euclidean distance to have the highest possible quality linkage, in which case, the ratio is 1.The second metric was the ratio of cost-weighted distance to the length of the least-cost path (CWD:LCP), which indicates the average landscape resistance encountered along the optimal path.After drawing the corridors, Pinch-point Mapper (McRae, 2012) was used, which uses circuit theory in the Circuitscape program (McRae et al., 2008) to estimate centrality metrics and identify pinch points within the least-cost corridors.Pinch points represent areas within least-cost corridors with high current densities and can act as bottlenecks for movement or refer to areas where there are no alternatives for displacement (Dutta et al., 2016).Drawing pinch points requires estimating the 'width' of the costweighted corridor.The 'wide' corridors must be within the cost-weighted distance (CWD) (WHCWG, 2010).
Since the maximum mileage distance for N. derjugini has been reported to be 49.19 ± 71.75 m (Shari and Afroosheh, 2014), this study considered the CWD cutoff distance of 50.

Modeling
The model assessment results based on the AUC metric showed that the model was successfully implemented.The value of this parameter for the test and training data was measured to be 0.90.The results of the JackKnife test showed that among the habitat variables used in this study, the slope, height variability, hillshade, elevation, distance from the rst-order waterway, and slope position classes have the highest impact, respectively.Figure 2 shows the habitat suitability map by the basins.Based on the results, habitat suitability is high in most parts of Basin 1.In Basins 2 and 3, the western parts have higher habitat suitability.

Basins and the cost map
By verifying the outputs of the maximum entropy model, the cost of the species relocation in 3 scenarios was prepared separately for 3 basins (Fig. 3).In Scenario 1, the habitat suitability map was reversed.The areas with the most friction had the lowest habitat suitability.In Scenario 2, the cost map was calculated only for the valleys.The brown color on the map represents the lowest cost of displacement and blue the highest cost.The color blue was assigned to the shallow valleys in the western parts of the distribution range.As moving eastward and higher altitudes, the variations in elevation increase, and valleys with low displacement costs are found.In Scenario 3, the costs are more stringent than in the previous two scenarios, and narrower ranges have been identi ed as areas with passing potential.In Scenario 1, the population cores in basins 1, 2, and 3 were 6, 20, and 45, respectively.This number was 4, 13, and 31 for Scenarios 2 and 4, 10, and 24 for Scenario 3, respectively, based on the order of the mentioned basins.

Scenario 1
Figure 4 shows the connection of Scenario 1.In the top 3 maps, the pinch points, and in the bottom 3 maps, the current ows of each corridor are displayed separately for the 3 basins.In Basins 1, 2, and 3, a total number of 3, 11, and 62 corridors were modeled to the lengths of 81.72, 438.28, and 1197.54km, respectively.In Basin 1, the population patches of 2, 3, and 6 are not connected to the other population cores.In Basin 2, the population cores 2, 3, 5, 6, 10, 17, 19, and 20 have no connection, and the patches 19 and 20 are located in deep valleys and waterways.In Basin 3, the population patches 2, 7, 18, 26, 36, 40, 41, and 44 are outside the network.Pinch points show the quality of the movement path so that the areas with a higher value have poor quality for the species' mobility.Pinch points are found in all basins along the route.Current ow displays the current passing along the corridor.In Basin 1, there is no high current intensity along the corridor route.In Basin 2, the value of current ow in the connection path of "population cores 9 and 11 to the next population cores " and up to the population core of 18 is high, indicating the quality of these areas.The highest current intensity occurs from the population core 9 to 12.In Basin 3, the highest current intensity is in the path of population cores 11,12,15,19,20,21,24,26, and 32 to 35.The characteristics of the three quality corridors in each basin based on the EWD: ED and CWD: LCP metrics are shown in Table 1.Based on the results, in Basin 1, 2 habitat corridors with lengths of 13934 and 56613 m were identi ed, connecting the population core of 1 to 4 and 5 to 4, respectively.The values of EucD, CWD, and LCP metrics in the connecting path of 1 to 4 cores are less than in other paths.In addition, the CWD: EucD metric in the corridor that connects the 1 to 4 population cores has a lower value, which indicates its better quality than the second corridor.However, the current intensity in both corridors is 2 Amps.The metric shows the di culty of newts in moving along the path.The metric CWD: LCP (addressing low resistance to move along the least-resistance path) is also less for the 1-to-4 core connection path.According to the interpretation of the metrics mentioned for Basin 2, the corridor connecting the population score of 8 to 13 has the highest quality, and the current intensity in this corridor is much higher than other population cores (Amps 7.02).In Basin 3, the highest current intensity is in the 14 to 20 corridor.However, the quality of the corridor that connects the population cores of 13 to 21 is very small compared to other corridors.and 30 are out of the connection network.In Basin 1, the pinch points are available along the entire length, revealing the low quality of the connection route.There is also a low current ow in this corridor.In Basin 2, pinch points are found in most movement routes.In the north of Basin 3, the population cores can be connected to the other population cores at a high cost.The highest current intensity in the drawn corridors occurs at the link path of 28 to 31 population cores.
According to the results (Table 2), in Basin 1, only one corridor is established between the population cores 4 and 1, and its current intensity is equal to 1 Amp.The low quality of the corridor is also shown in Fig. 5.In Basin 2, the connection route of 4 to 7 population cores, based on CWD: EucD (0.51) and CWD: LCP (0.35) metrics, is of higher quality than the other 3 routes.However, there is no high current intensity among these 3 corridors.In Basin 3, the quality of the corridors is higher than in the other two basins, and the highest current intensity (24.29 Amps) is in the path of connecting 9 to 14 cores.

Scenario 3
Figure 6 shows the results of investigating the habitat corridors in Scenario 3. The identi ed corridors in Basins 1, 2, and 3 are 1, 5, and 29, the length of which is 55.56, 318.24, and 686.45 km, respectively.In Basin 1, the population cores 2 and 3, in Basin 2, the population cores 6, 9, and 10, and in Basin 3, the population cores 10, 11, 12, 19, 20, 21, 22, and 23 are outside the connection area.The pinch points in Basin 1 are found along the entire pathway length; therefore, connecting these population cores would be very di cult.In Basin 2, the connecting pinch points are between the 3 to 5 and 4 to 5 population cores.The highest current ow in Basin 2 is between the 5 to 10 population cores.In Basin 3, the highest pinch points are between the cores 4, 5, 6, 7, and 10.The most intensive current ow also occurs in this region.
Table 3 shows the results of the metrics used for the habitat corridors in Scenario 3. As in Scenario 2, in Basin 1, there is only one connection path with low current intensity (Amps 1).In Basin 2, there is no high current intensity in 3 corridors (less than Amps), and the 3 to 5 core connection path, based on CWD: EucD (0.32) and CWD: LCP (0.45) metrics, is of better quality.Compared to the other two basins, in Basin 3, there is more current intensity in the corridors.Also, CWD: LCP and CWD: EucD are lower than the other two basins, indicating the higher quality of the corridors.

Ground trust
Figure 7 shows the spatial compliance of the identi ed corridors to ground trust for a small area (black box).The corridors in all three basins are depicted in green in Scenario 1, red in Scenario 2, and blue in Scenario 3. In Basin 1 and Scenario 1, the connection route of population cores is direct and away from waterways and valleys.In Scenarios 2 and 3, the identi ed corridors are consistent with MCs.In Basin 1, the connection path between the population cores was utterly different from that predicted under all three scenarios.In Basin 2, there is an overlap between the passing routes of the scenarios in the northwestern parts.However, the corridor in Scenario 1 passes through highlands in some sections.The differences become more pronounced in Basin 3.

Discussion
This study investigated the habitat suitability and connectivity pathways of N. derjugini population cores on a ne scale using the maximum entropy model.The results show that the model could predict the species' habitats, so it is proposed as a suitable model for modeling on such a scale.The ancestors of the genus Neurergus were distributed in Europe and the Mediterranean.Then, due to the favorable climate, some of this genus moved to the southern parts.The displacement to the south has expanded the species' distribution range to the habitats in the Zagros forest and adjacent areas in Turkey and Iraq (Steinfartz et al., 2000).The current distribution range of N. derjugini is enclosed in the Zagros highlands.
By preserving the connections made between the population cores in this area and due to the effect of high topographic diversity (Odunuga and Badru, 2015) and vegetation cover (Kalota, 2017) on moderating temperature, these areas can act as a suitable climatic refuge not only for this species but also for N. kaiseri (Ashrafzadeh et al., 2019) and Salamandra infraimmaculata (Ahsani et al., 2018) and even plant species (Olea europaea and Myrtus communis) (Malekoutian et al., 2020).These cases support the view that uneven topography, which causes species to displace at short distances and thus modify changing conditions, can inhibit the biotic effects of climate change (Ackerly et al., 2010).Among the habitat variables, slope percentage, altitude variation, and cluster hillshade had the highest impact, and CTI, landform dominance, and NDVI had the lowest impact.Based on the research ndings on this study scale, "slope" is the most critical parameter for N. derjugini.This is while this habitat variable has been used neither in other studies on this species (Vaissi and Shari , 2019; Afroosheh et al., 2019) nor even in the release studies (Vaissi, 2021).With increasing the slope up to 400%, the habitat suitability increases, and from this point on, the suitability declines.Habitat suitability rises with increasing altitude diversity.This factor is a signi cant stimulus in the formation of microclimates and even landforms.
Landform diversity has even been recognized as one of the components of temperature reduction (Rastandeh et al., 2019).Therefore, in combination with the two criteria of slope and elevation, this parameter plays a critical role in the habitat suitability of N. derjugini.Many studies have identi ed the relationship between altitude and slope as a descriptor of salamanders' habitat (Gherghel and Papeş, 2015).The cluster hillshade showed that habitat suitability decreases as radiation to the surface increases.
Plain areas will not have high habitat suitability for the species due to lower elevation, and more radiation received.Direct sunlight in areas without topography can cause higher temperatures and thus reduce the amount of moisture available, which is different from the favorable conditions for this and other salamander species (Hernandez et al., 2017).Moreover, low radiation can affect the durability of snow at higher altitudes and thus the water supply of streams and springs (Shari et al., 2009).Increasing the height up to 1500 m increases the habitat suitability of N. derjugini, which is emphasized in similar studies (Naja majd and Keya, 2010; Afroosheh et al., 2016).However, from this point on, the suitability declines.In research by Shari et al. (2017) on the same species, the results showed that the optimum height for the species is up to 2057 m, which is different from the ndings of this study and may be due to the spatial resolution of the data used.Areas with altitudes above 1,500 m may be unfavorable to the species for the reasons such as extreme cold and frost conditions.Height is generally considered a limiting factor for distribution (Lomolino, 2001).For the eastern populations of Plethodon salamanders, altitude is also a factor limiting the distribution range.The N. derjugini populations during the mid-Holocene and LGM periods tended to lower altitudes for nding glacial refugees (Afroosheh et al., 2019).
The CTI in the response curves showed that by decreasing the value of this index from 2 to lower values, the species' habitat suitability would fall.Then, by increasing the value from 4 onwards, the suitability would rise again.This shape of the response curve shows the difference in the environmental conditions of the two distribution ranges.Parts of the presence points in the Zagros highlands have higher humidity and altitude than in Iraq, which was previously mentioned in the study by Shari and Vaissi (2014) (Ficetola et al., 2018).However, the modeling with broad-scale data for specialized species such as salamanders may not include many species characteristics in the predictions.These specialized features have not been considered in the mentioned studies and even in the study of creating protected areas (Vaissi, 2021).In a study by Ficetola et al. (2018) to examine the differences between broad-and ne-scales, the results showed that the microhabitat scale is more appropriate to use for the study on salamanders.In a multi-scale study of Neurergus kaiseri by Goudarzi et al. (2021), the results showed that ne-scale models effectively reduce uncertainty.For species with local distributions, local factors play a vital role in their habitat and displacement.Sometimes, these factors create constraints for species displacement that can only be detected on a ne scale.In the different scenarios, regardless of the physical structures of the landscape, there were found populations with no connection, indicating the absence of a favorable MC.This distribution limit can only be detected on a ne scale (Fig. 4).This study used drainage basins to investigate the potential of population cores for binding connected.Drainage basins are a suitable study unit for water-dependent amphibians due to a single out ow and the collection of water from all waterways.Since the boundaries between these basins are determined by parameters such as rainfall, water ow, and altitude, gene ow and the formation of local populations of N. derjugini are possible to occur in these basins.In a study by Malekoutian et al. (2020) on this species using Mitochondrial DNA sequence, the results showed that 3 populations could be identi ed for this species.Consistency of these results with the present study's ndings, reporting the northern population in Basin 1, the central population in Basin 2, and the downstream population in Basin 3, indicates that the population connection trend in the basins can be independent.In Scenario 1, where there was no emphasis on MC, many of the corridors crossed areas that may pose a threat to the species (Fig. 7).
Crossing  2016) showed that using the habitat suitability map as a proxy performed better for the generalized species (Ovis canadensis nelsoni) than the specialized species (Cervus canadensis).Therefore, the reverse of habitat suitability is a weak proxy for specialized species such as N. derjugini.According to the metrics of ow intensity, EucD, and CWD, Basin 3 has more suitable corridors containing the highest current ow among the population cores.This could be attributed to the closer proximity of the population cores and the greater roughness of the basin.The proximity of population cores and suitable MCs such as valleys and waterways have led to a high quality of the corridors.The population cores in Basin 1 under all studied scenarios did not have suitable corridors due to the long distance between the population cores, reducing the possibility of genetic exchange.Therefore, this basin is more at risk than the other two basins.Afroosheh et al. (2019) reported little correlation between the northern populations of N. derjugini, which is similar to the ndings of this study.Scenario 3 investigated the connection paths only in the landform classes of canyons (deeply-incised streams), mid-slope drainages (shallow valleys), upland drainages (headwaters), and u-shaped valleys that, compared to Scenario 2, provided a more limited context for the species movement.In this scenario, the corridors identi ed along the transit routes are more in compliance with the species' habitats (Fig. 7).Based on the research ndings, landform and then slope position classes can quantify the species' movement paths more e ciently, in a way that is more in line with the ecological and biological characteristics of the species.In most areas identi ed as corridors, there was suitable vegetation cover in the context.Although the NDVI had little effect in this study, Scenario 3, by compliance with the four landform classes, included this variable appropriately in the modeling.Areas such as waterways and valleys in mountainous regions can be considered a kind of riparian ecosystem.Vegetation cover, water, and moisture in these areas can play a signi cant role in facilitating movements (Gregory et al., 1991; Naiman and Decamps, 1997).Considering the maximum distance that can be traversed on land (300 to 1000 m) for different amphibian species (Pittman et al., 2014), the identi ed corridors (Tables 1 to 3) have the potential to be used in terms of length.However, the distance for N. derjugini was measured to be 49.19 ± 71.75 m (Shari and Afroosheh, 2014), which is less than the mentioned mileage distance.So what is clear is that the potential of using the corridors identi ed in the scenarios depends to a large extent on the existence of stepping stones along the corridors.The pools along the potential corridors of Scenarios 2 and 3 can be considered stepping stones, providing the best conditions for connecting the population cores.Also, the species has adapted to use these areas (Öz et al ., 2002).Rainfall (Mazerolle, 2001;Todd and Winne, 2006) and snowmelt, and water ow in canals can all contribute to creating these areas, which differ from the role of runoff (Costa et al., 2016).The displacement between these habitat spots is considered part of the away mod stage (Pittman et al., 2014), in which the species shows the least response to terrestrial habitat quality.Accordingly, the presence of these stepping stones can help increase the distribution range at this stage.These conditions have also been observed in the distribution of ringed salamanders (Ambystoma annulatum) (Boone et al., 2006).In a study by Ribeiro et al. (2011), the results also showed that the relationship between pool networks was effective on amphibian richness and communication and the presence of some species.In all scenarios used, there were detected populations that had no connection with other population cores.These populations are usually present in springs (Afroosheh et al., 2016), and due to the lack of communication and unfavorable environmental conditions, they may become extinct.This is not only experienced by N. derjugini (Shari and Vaissi, 2014) but also by many amphibian populations (Hecnar and M'Closkey, 1996;Trenham, 2003).Therefore, the protection of populations in this basin can be by integrating in-situ and ex-situ conservation methods and even captive breeding.Considering the limitations that amphibians have for propagation and communication between habitat cores, the role of structural connectivity is prominent, which can be de ned based on the degree of specialization in maintaining habitat connectivity.
If the suitability map can be interpreted as a re ection of the habitat needs of the species and the degree of specialization of functional connection (Valerio et al., 2019), then it can be said that the approach of this study is to integrate the structural and functional features of the landscape to facilitate the connection between population cores.

Conclusion
Newts are dependent on aquatic environments.This degree of specialization has made the study of their habitat connections different from other animal species.Habitat connectivities of newts occur in a wet context.Therefore, maintaining habitat connections for the species depends on conserving moist environments in the landscape structures (characterized by preserving moisture and water ow).More emphasis can be placed on waterways and valleys in which upstream currents feed on seasonal rainfall or snowmelt water.This study was performed using data with a resolution of 30 m.Using accurate data can lead to results different from those studies conducted with less spatial resolution data.The results showed that physical structures of landscape and habitat suitability models could help evaluate the role of landscape structures in facilitating functional habitat connections of the species with low ability and

Figures
Figures

Figure 3 Cost
Figure 3

Table 1
Characteristics of the corridors identi ed between the population cores under Scenario 1Figure5shows the results of the calculation of connection corridors under Scenario 2. In this scenario, the cost map is calculated only for the valleys.The number of corridors identi ed in Basins 1, 2, and 3 is 1, 6, and 40, respectively, 72.48, 261.93, and 1655.79 km in length.In Basin 1, the population cores 2 and 3 are not connected to the other cores, and in Basin 2, the population cores 1, 2, 3, 5, 8, 12, and 13 are not on the connection network pathway.In Basin 3, the population cores 2,10, 11, 12, 16, 17, 21, 26, 27, 29,

Table 2
Characteristics of the identi ed corridors between the population cores in Scenario 2

Table 3
Characteristics of the identi ed corridors between the population cores in Scenario 3 . High humidity is essential for most salamander species in this mountain range, such as Salamandra infraimmaculata and N. derjugini (Ahsani et al., 2018).According to Ahsani et al. (2018) and Ashrafzadeh et al. (2019), physical parameters play an in uential role in preventing distribution.
(Pope et al., 2000;Mazerolle ,2005)e loss as areas with sparse vegetation cover typically have high surface temperatures.Moreover, due to sparse vegetation cover, open spaces can increase the risk of hunting by birds, reptiles, and mammals.Any response from the species is somehow a cost that affects habitat use and movement potential(Winandy et al., 2017).Three parameters of humidity, temperature, and predation are considered the primary factors that affect amphibians' movement(Joly, 2019).However, amphibians use the narrow structural elements of the landscape, such as hedgerows, ditches, eld margins, road verges, and channeled agricultural and headwater streams, to move.(Popeetal., 2000;Mazerolle ,2005).In the studies by Afroosheh et al. (2019) on N. derjugini and Ashrafzadeh et al. (2019) on Neurergus kaiseri, no particular emphasis is made on the MC.Keeley et al. (