Soil thickness determines the soil productivity in the black soil region of northeast China, which is important for national food security. Existing information on the spatial variation of black soil thickness is inadequate. In this paper, we propose a model framework for spatial estimation of the black soil thickness at the watershed scale by integrating field observations, unmanned aerial vehicle variations of topography, and satellite variations of vegetation with the aid of random forest. We sampled 141 sample profiles over a watershed and identified the black soil thickness based on indices of the mollic epipedon. Topographic variables were derived from a digital elevation model and vegetation variables were derived from Landsat 8 imagery. Random forest was used to determine the relationship between black soil thickness and environmental variables. The resulting model explained 61% of the black soil thickness spatial variation, which was more than twice that of traditional interpolation methods (ordinary kriging, universal kriging and inverse distance weighting). Topographic variables contributed the most toward explaining the thickness, followed by vegetation indices. The black soil thickness over the watershed had a clear catenary soil pattern, with thickest black soil in the low depositional areas and thinnest at the higher elevations that drain into the low areas. The proposed model framework will improve estimates of soil thickness in the region of our study.