Bringing Theory to Practice: Digital Soil Mapping in the Ochoco and Malheur National Forests.

Soil resource information is currently in the process of being updated across the forests of the Pacific Northwest. In an effort to shorten the duration of the survey efforts, digital soil mapping is being considered a necessary approach for advancing the completion date of the soil surveys.  Here we focus on the assessment of geomorphometric variable analysis in the context of predictive soil mapping. Geomorphometrics are readily derived from digital elevation layers and present an opportunity to capture the landscape dynamics influencing the distribution of soils, thus are expected to become increasingly important for fine scale DSM. Furthermore, we recognize and explore a significant hurdle which all digital soil mappers must grapple with if we are to move this technology into the realm of widespread practicality; how to best identify soils which do not exist in our training data set and create meaningful class delineations for the land manager when moving int ounmapped areas? We suggest that subgroup represents the finest unit of prediction as it reflects process whereas Series reflects subjective land use interpretations and as such are fundamentally unpredictable.  These questions are explored in the diverse soilscape of a Central Oregon forest using a Random Forest algorithm to predict explicit soil classes to the subgroup level at 10 meter resolution. A conceptual model is presented which could streamline the prediction to publication process.