Covariates
The following covariates (Table 1) were used in analyses of hyena landscape use and navigation: normalized difference vegetation index (NDVI; 30m, Landsat 8 Surface Reflectance Tier 1, rainy and dry seasonal averages for 2019), slope (30m, Shuttle Radar Topography Mission [STRM]), elevation (30m, STRM), distance to rivers, distance to lakes, distance to boundaries, distance to verified livestock predation locations, distance to regions people perceive as being risky due to hyenas, and distance to participatory mapped livestock predation locations during the study period. The latter two variables were derived using participatory mapping data from communities living within 2 km of the protected area boundaries, while the verified predation dataset was from the local wildlife authority, Kenya Wildlife Service (Wilkinson et al. 2021a). Because the killing of or retaliation against wildlife is illegal in Kenya, participatory mapped livestock predation and participatory mapped risks from spotted hyenas can serve as proxies for spatially explicit intolerance or acceptance of spotted hyenas by local communities and are associated with the potential for deterrents and aversive behaviors toward hyenas. Euclidean distance was used for all distance layers, and road layers were derived through Open Street Maps and by hand tracing. Fences assessed in barrier analyses were mapped in person by driving and walking the boundaries of the protected areas. All distance covariates were created as rasters with 30m spatial resolution to match the NDVI, slope, and elevation rasters.