Disaggregation of the 1:100,000 Reconnaissance soil map of the Busia Area, Kenya using a soil landscape rule-based approach

Abstract

A 1:100,000 soil map of the Busia area in western Kenya published in 1991 is the main source of soil information for the area, but the map is outdated and unsuitable for new emerging demands. We disaggregated the Reconnaissance Soil Map of the Busia Area (quarter sheet No. 101) into individual soil classes to produce a soil class map that may better meet current needs. A soil landscape rule-based model was used to disaggregate the soil map units by exploiting information in the map legend and the map unit descriptions. These descriptions were used to generate rules that were applied to a fuzzy soil class map generated from a parent material map and a K-means cluster map generated from six terrain attributes, namely, multiresolution ridgetop flatness (MRRTF), multiresolution valley bottom flatness (MRVBF), topographic wetness index (TWI), topographic position index (TPI), planform curvature, and profile curvature. The result was a soil class map with a spatial resolution of 30 m with an overall accuracy of 58% and a Kappa coefficient of 0.54. The soil landscape rule-based approach provides an opportunity to disaggregate traditional soil maps in cases where there is no adequate dataset. The drawback with this approach, however, is that it relies heavily on the resolution of the DEM and the type of information contained within the soil survey report describing the soil classes within the soil map units.