Soil-landscape modelling to quantify spatial variability of soil texture

Soil-landscape models were developed to predict the spatial distribution of soil texture at the surface horizon across a catchment in southeastern Nig eria. A discretised thin-plate spline technique, in conjunction with a conn ected drainage-enforcement algorithm supplemented with the incorporation of ridge and stream-line data, was used to ensure proper hydrogeomorphic prop erties of the output Digital Elevation Model (DEM). Terrain attributes were derived from the resulting DEM. Stepwise multiple-linear regression was pe rformed on the normalised terrain attributes and on the principal component s constructed from the normalised terrain attributes to avoid multi-colline arity. The derived soil-landscape models were used to predict clay, silt, s and, ironstone and thickness of the surface horizon from the original terra in attributes for the entire study area (R-2 = 0.41 to 0.75). The models we re further validated using statistical criteria. Only for the soil variable clay did the soil-landscape model improve after stratification according t o geological formation (R-2 increased from 0.47 to 0.76). Cell-based algori thms were used to map the soil-landscape models spatially. The resulting sp atial patterns correctly showed a significant relationship with the terrain attributes. This relationship is useful when studying patterns of sediment movement. (C) 2000 Elsevier Science Ltd. All rights reserved.