Distribution modeling of the long-tailed marmot (Marmota caudata) for objectives of directing field surveys and ground validation of the snow leopard (Panthera uncia) habitat quality

Marmots form a part of the diet of some endangered species such as the snow leopard (Panthera uncia), therefore the knowledge on their distribution and habitat preferences are crucial to the interest of conservation and management of carnivores at high altitudes. Considering this, within a Snow Leopard Project run by Biosphere Expeditions and NABU (Kyrgyzstan), surveys were carried out in summer field seasons of 2014–2019 to assess the distribution of the long-tailed marmots (Marmota caudata) in an area centred around the Karakol Mountain Pass (polygon centroid 74.83°E, 42.37°N) in the Kyrgyz Ala-Too Range. The presence of occupied marmot burrows was recorded using the location (cell) given by a grid, the code of which was displayed in a GPS. Using cells allows examination of data at a wider scale, so information is collected from different cells that are spread from each other, avoiding data autocorrelation. Environmental factors that may affect the spatial distribution of burrow systems were considered: land surface temperature (LST) in winter and summer, summer normalized difference vegetation index (NDVI), a Digital Elevation Model (DEM), and soil type data. The relationship between environmental factors and burrow records was analysed using ecological niche models (Maxent) to predict the distributions of marmot burrows. The models performed well with average test AUC values of 0.939. The contribution orders of the variables in the models were summer NDVI and DEM, winter LST, summer LST, and soil type. The distribution of the suitable areas was largely (up to 38 % permutation importance) affected by summer NDVI. NDVI is an indicator of the feeding conditions of marmots and most of the records were distributed in areas with NDVI in summer ranging from 0.5 to 0.7. According to the prediction maps, suitable marmot habitat (>0.5 predicted probabilities of occurrence) can occupy up to 40 % of study area. These maps are used to direct sampling efforts to areas on the landscape that tend to have greater predicted probabilities of occurrence and accomplish ground validation of snow leopard habitat quality