AbstractRooting in deep regolith enables forests to withstand seasonal and annual precipitation shortfalls. Despite its ecological importance, spatial patterns in regolith thickness within forest ecosystems are scarcely documented. Regolith thickness was estimated at 66 sites throughout a 543‐ha watershed in the southern Sierra Nevada by hand auger to point of failure or a maximum depth of 7.5 m, describing a minimum thickness estimate. Regolith consists of 1–2 m of soil overlying thick and porous weathered granodiorite. Depth to auger failure ranged from 1.52 to an indeterminate depth beyond 7.5 m. A total of 27 points exceeded 7.5 m depth. Normal, lognormal, and gamma data distribution models were fitted to observations to extrapolate thickness across the watershed and estimate thicknesses beyond the measurement limitation. Predictions for the 95th percentile of regolith thickness varied substantially; 26.05 m for lognormal, 16.87 m for gamma, and 9.56 m for normal. Considering any best fit model, >55% of the watershed area was deeper than 5 m. Depth classes were formed to evaluate the extent to which topography is associated with spatial trends in regolith thickness. Spatial patterns were related to two covariate proxies (distance from stream channel and topographic wetness) with the general landscape trend of shallow depth classes (<3.3 m) in lowlands and deeper regolith classes (>7.5 m) in uplands. The normalized difference vegetation index signatures over the late stages of a 5‐year drought were greener in the lowlands. In contrast, upland forests displayed widespread tree die‐off, suggesting deep water storage does not maintain forests over long‐term drought.