Spatially referenced density estimates are essential for understanding population dynamics and informing conservation planning for wide-ranging carnivores. This dataset compiles georeferenced leopard (Panthera pardus) population density estimates aggregated from peer-reviewed literature and gray literature spanning sub-Saharan Africa. The dataset contains 115 individual density observations drawn from 45 independent sources, including camera-trap studies, mark-recapture analyses, and spatially explicit population models. Each record includes a density estimate (individuals per 100 km²), geographic coordinates (decimal degrees longitude and latitude), and the associated source citation or DOI.

We extracted density data from published studies that incorporated spatial components, specifically camera trap-based SECR, due to their demonstrated robustness and narrower confidence intervals compared to non-spatial methods (Sutherland et al., 2019). Most studies estimated density across all habitats or available habitats within a defined survey state-space, but some studies constrained activity centres to suitable habitats using habitat masks (Table S4 in the manuscript). We extracted the density estimate along with the coordinates of the study area. For studies where no coordinates were provided, we extracted the coordinates of the study area's centroid. The published papers did not always use the same definition for the sampling area. Hence, we interpolated the centroid of the area either from the map, if provided (which was most of the studies), or we took the centroid of the study area extent that was provided. The density values were then log-transformed for further analysis.

Raster files for our candidate variables were procured through both the USGS Earth Explorer platform (Interface and Documentation 2000) and the Google Earth Engine (Gorelick et al. 2017) with a pixel resolution of 1 km x 1 km (Table S5 in the manuscript).