Examining spatial nonstationarity in top-down control of mesocarnivores among disparate landscapes

Large predators can strongly influence carnivore communities through interspecific interactions, but their effects on mesocarnivores may vary across landscapes with changing ecological conditions. Wildlife management at the landscape level and centered on large predators may overlook spatial variation in species interactions and their conservation implications. We assessed spatial non-stationarity in the effects of tigers and leopards on the occurrence of striped hyaena and golden jackal across the Shivalik, Central India, and Western Ghats landscapes in India. Using occurrence records from secondary sources (e.g. All India Tiger Monitoring reports), we fitted Random Forest models separately for each landscape to model hyaena and jackal occurrence as a function of large predators, prey species, and environmental conditions. We also fitted global Random Forest models using pooled data from the three landscapes and evaluated their robustness in predicting regional responses. Regional models showed contrasting responses of hyaenas and jackals to large predators, prey, and geoclimatic covariates, indicating spatial non-stationarity in mesocarnivore responses across the landscapes. We found no consistent evidence of top-down control by large predators; instead, mesocarnivore occurrence was primarily associated with geoclimatic variables. Global Random Forest models captured broad regional response patterns and showed lower uncertainty than regional models, particularly in data-poor regions. This underscores the importance of flexible broad-scale models in generating regional inferences for effective wildlife management. Successful conservation of mesocarnivores in environmentally diverse countries such as India requires management at a finer scale, supported by a better understanding of the density-dependent effects of recovering large predator populations.