Quantification of Outcomes Generated Using Multi-Scale Geomorphic Classification Systems in Predictive and Update Modes of Digital Soil Mapping.

We assessed terrain attributes and data mined soil survey databases in an effort to develop a robust, multi-scale, multi-process geomorphological database to be tested and used in predictive soil mapping. The “R” factor of relief (topography, terrain, etc) was isolated by ignoring or holding the other factors (climate, organisms, parent material, time) as constants to predict the soil at any given point purely as a function of the community of the underlying terrain attributes. Information on the geological nature of the interior of the landform, however, was used, for populating the geomorphic process or environment of the geomorphic matrix for said landform. We are attempting to improve spatial disaggregation of soil-area class maps by incorporating multi-scale geomorphic information derived from existing SSURGO data sources. Our approach involves the spatial disaggregation of soil map units into their identified components. Geomorphology data and terrain attributes derived from terrain attribute generators were used as dependent and independent layers, respectively, in implementation of Random Forests. We hypothesize that (i) the application of a geomorphic classification system enhances the predictability of natural soil bodies in landscapes with complex landforms and (ii) outcomes of digital soil mapping are significantly improved where the intrinsic variability of a geomorphic landscape classification system is properly matched to the appropriate spatial scale and complexity of the environment which it is attempting to classify.