Disaggregation of Legacy Soil Maps to Create More Detailed Digital Soil Maps.

In efforts to create more detailed raster soil maps in areas with sparse point data, legacy soil maps are often the main source of soil-landscape information. In the United States, the Soil Survey Geographic (SSURGO) database is currently the most detailed and widely available soil information. SSURGO is published as a vector map product with an associated report and database for use in land management and planning. Each management unit (or map unit) has estimated tabular soil property data associated with distinct soil components present in each map unit. Problems arise because of spatial aggregation of components, qualitative methods of property estimation, and lack of a description of transitions between soil components. However, descriptions of within map-unit soil-landscape relationships are published in SSURGO databases and manuscripts. These descriptions include relationships of soil components to terrain positions, parent materials, vegetation, and other soil-forming environmental factors. These descriptions, along with the knowledge of the local soil scientists, can provide the framework to disaggregate these map units into raster maps of the distribution of soil components, which may then be translated into new updated soil maps. For this project expert rule-sets, decision tree ensemble models, and fuzzy membership methods were used to create updated soil series maps in Pocahontas and Webster Counties, West Virginia. Modeling results indicated good overall training accuracy (68%), and independent field validation also showed decent agreement. Pilot projects around the U.S. are showing potential for using this process in North America for soil survey update and Global Soil Map efforts in mapping soil properties.