Data mining techniques are studied to recover knowledge from geodatabases in order to improve updates of existing soil maps and to help in developing a preliminary soil map for neighboring unmapped areas. Classification tree, one of the most widely used inductive learning methods is used here to retrieve the expert knowledge embedded in the soil-landscape model used by the Harney County, Oregon soil survey (ca. 1980-2003). Spatial environmental data of geology, vegetation, precipitation, terrain attributes (elevation, slope, and aspect) and landsat ETM+ data at a resolution of 30 m were used to predict soil map units. The classification tree proved to be a powerful tool in retrieving the spatial relations between soil map units in the reference area of Harney County. By extrapolating the soil prediction model, a preliminary soil map of an adjacent unmapped area of Malheur County, Oregon is developed with reasonable accuracy.
See more of Soil-Landscape Modeling and Geospatial Analysis
See more of S05 Pedology
See more of The ASA-CSSA-SSSA International Annual Meetings (November 6-10, 2005)