Decades of intensive off-road vehicle use for border security, immigration, smuggling, recreation, and military training along the USA–Mexico border have prompted concerns about long-term human impacts on sensitive desert ecosystems. To help managers identify areas susceptible to soil erosion from anthropogenic activities, we developed a series of erosion potential models based on factors from the Universal Soil Loss Equation (USLE). To better express the vulnerability of soils to human disturbances, we refined two factors whose categorical and spatial representations limit the application of the USLE for non-agricultural landscapes: the C-factor (vegetation cover) and the P-factor (support practice/management). A soil compaction index (P-factor) was calculated as the difference in saturated hydrologic conductivity (K_s) between disturbed and undisturbed soils, which was then scaled up to maps of vehicle disturbances digitized from aerial photography. The C-factor was improved using a satellite-based vegetation index, which was better correlated with estimated ground cover (r² = 0·77) than data derived from land cover (r² = 0·06). We identified 9,780 km of unauthorized off-road tracks in the 2,800-km² study area. Maps of these disturbances, when integrated with soil compaction data using the USLE, provided landscape-scale information on areas vulnerable to erosion from both natural processes and human activities and are detailed enough for adaptive management and restoration planning. The models revealed erosion potential hotspots adjacent to the border and within areas managed as critical habitat for the threatened flat-tailed horned lizard and endangered Sonoran pronghorn.