Spatial predictions of soil properties at regional to continental scales are almost exclusively derived from small scale polygon class maps created for general purpose soil survey programs. These maps represent variability in the soil cover with discrete boundaries that do not reflect the continuous spatial distribution of soils and their properties. Component soils within these map units are assigned representative values for selected soil properties, but the location of these component soils are not explicitly identified. These double-crisp maps may present challenges for use of soils data for regional and continental assessment of climate change, carbon sequestration, and ecosystem services. As an alternative, raster-based continuous soil property maps that represent the spatial variability at diverse scales is a desired products for modeling and conservation communities. The objective of this research is to develop methods for creating continuous raster soil property maps from the disaggregation of current legacy soil polygon maps. At a regional level, topography is one of the major drivers for soil prediction. Watersheds were grouped based on a fuzzy K means classification using mean, range, and standard deviation values for selected terrain attributes. The grouping of similar watersheds provided for stratification into “rulesheds” where we could consistently apply soil-terrain attribute relationships for soil prediction. Establishment of soil-terrain attribute relationships allowed for numerical quantification of soil-landscape models that were used to generate fuzzy soil class maps and continuous soil property maps.  Soils belonging to a landform unit or slope position were further correlated based on taxonomic similarity overcoming the county line boundaries present in the current soil polygon maps.