Using Soil Properties in Digital Soil Mapping of the Malheur National Forest.

Digital soil mapping is a reiterative process whereby predictions of soil variables made based on landscape variables are validated in the field. The end result is a complete map of a region with estimated margins of error in prediction accuracy, and an increase in accuracy after each successive prediction-validation cycle. Soil physical and chemical properties can be mapped in addition to soil type, and the resulting map used for land management planning. The Malheur National Forest in Oregon is partially mapped using traditional soil survey methods, and the known information can be integrated into the predictive model to generate a digital soil map of the region. The predictive model for estimating soil properties across the region was generated using RandomForest, a program for making a bundle of decision trees trained using existing information. The final map of predicted soil properties can be used to dissaggregate complex soil map units and be accessible to land managers.