Planning connectivity at multiple scales for large mammals in a human-dominated biodiversity hotspot

Abstract

Connectivity for large mammals across human-altered landscapes results from movement by individuals that can be described via nested spatial scales as linkages (or zones or areas) with compatible land use types, constrictions that repeatedly funnel movement (as corridors) or impede it (as barriers), and the specific paths (or routes) across completely anthropogenic features (such as highways). Mitigation to facilitate animal movement through such landscapes requires similar attention to spatial scale, particularly when they involve complex topography, diverse types of human land use, and transportation infrastructure. We modeled connectivity for Asian elephant (Elephas maximus) and gaur (Bos gaurus) in the Shencottah Gap, a multiple-use region separating two tiger reserves in the Western Ghats, India. Using 840km of surveys for animal signs within a region of 621km2, we modeled landscape linkages via resource selection functions integrated across two spatial resolutions, and then potential dispersal corridors within these linkages using circuit theoretical models. Within these corridors, we further identified potential small-scale movement paths across a busy transportation route via least-cost paths and evaluated their viability. Both elephants and gaur avoided human-dominated habitat, resulting in broken connectivity across the Shencottah Gap. Predicted corridor locations were sensitive to analysis resolution, and corridors derived from scale-integrated habitat models correlated best with habitat quality. Less than 1% of elephant and gaur detections occurred in habitat that was poorer in quality than the lowest-quality component of the movement path across the transportation route, suggesting that connectivity will require habitat improvement. Only 28% of dispersal corridor area and 5% of movement path length overlapped with the upper 50% quantile of the landscape linkage; thus, jointly modeling these three components enabled a more nuanced evaluation of connectivity than any of them in isolation.