Ochotona macrotis
Univariate models - Comparison of univariate models at different scales prior to model building revealed that the variables low percentage cover of vegetation, greater percentage of rocks, greater percentage of large sized rocks, and more number of large-sized rocks were important for predicting the presence of OM across all scales (Table C1). It also indicated unique variables that were important at each spatial scale (Table C1):
All the univariate models run are presented here (Table S3, Table S4, Table S5).
Multivariate models - At each spatial scale, models constructed were subjected to a stepwise regression analysis to pick the best additive models using different methods. The purposeful modeling approach (Model 2) consistently produced the best models across spatial scales. It included variables significantly associated with presence in univariate regressions at regional scales, such as the percentage cover of large-sized rocks and the number of large-sized rocks (Table 1). At the Ladakh spatial scale, the presence of OM was driven by a higher cover of large rocks, a greater number of large-sized rocks, and lower NDVI in summer (Table 1). At the South-East Ladakh spatial scale, the occurrence was primarily driven by a higher cover of large rocks, a greater number of large-sized rocks, higher elevation, lower cover of shrubs, and consistently low NDVI in winter (Table 1). At the North-West Ladakh scale, the occurrence was driven by many variables, with no variables related to rocks, vegetation, or geographic features strongly influencing the presence of the species (Table 1).