Soil property maps are important for land management and earth systems modeling. A new hybrid point-disaggregation predictive soil property mapping strategy improved mapping in the Colorado River Basin, and can be applied to other areas with similar data (e.g. conterminous United States). This new approach increased sample size ~6-fold over past efforts. Random forests related environmental raster layers representing soil forming factors to samples to predict 15 soil properties (pH, texture fractions, rock, electrical conductivity, gypsum, CaCO3, sodium adsorption ratio, available water capacity, bulk density, erodibility, organic matter) at 7 depths, depth to restrictive layer, and surface rock size and cover. Cross-validations resulted in coefficient of determinations averaging 0.52, with a range of 0.20 to 0.76; and mean absolute errors ranged from 3% to 98% of training data averages with a mean of 41%. Uncertainty estimates were also developed by creating relative prediction intervals (RPIs) for the entire study area, which allow end users to evaluate uncertainty relative to original data distributions. Average error increased with higher RPI values (higher uncertainty), and areas with the highest RPI are consistently under-sampled, suggesting that additional sampling in these areas may improve prediction accuracy. Greater uncertainty was also observed in areas with shale parent materials and physiographic settings uncommon relative to the broader study area.