Road-fill failure is a common and potentially catastrophic hazard on mountainous roads. Land managers need methods for assessing road-fill failure hazards that minimize time in the field and that use readily available data. We developed a probabilistic hazard assessment method using virtual fieldwork and LiDAR-derived slope distributions over a 97-km length of the Blue Ridge Parkway in North Carolina. The locations of arc-shaped pavement cracks and previous road-fill failures were mapped using field-verified VisiData video reconnaissance, and the slopes surrounding failed-or-cracked sites were extracted from a digital elevation model and compared to uncracked portions of road. The site-median slopes of the failed-or-cracked sites and uncracked sites are normally distributed and significantly different (P < 0.01) from one another. This observation provides the basis for a probabilistic, hazard assessment model that uses the site-median slope of a 30-m–diameter buffer to categorize currently uncracked sites as either stable or at risk of road-fill failure during future high-intensity rainfall events.