Mapping soil depth classes in dry Mediterranean areas using terrain attributes derived from a digital elevation model

Modern land management increasingly demands quantitative information on spatially variable soil properties. Traditional soil survey maps do not provide this information. In this paper we report on the application of terrain attributes to mapping soil depth classes at high spatial resolution over large areas under dry Mediterranean conditions. Soil data were collected in III georeferenced field plots of 30 m x 30 m, more or less equally distributed over three lithological units, phyllites, shales and limestones. Topographic attributes were computed from a digital elevation model at 30 m resolution. A principal components analysis was carried out on the map overlays of the terrain attributes in order to obtain uncorrelated topographic factors that enabled us to apply a probability approach. Using the maximum likelihood classifier with the field plots as a dispersed training area, predictions were made of mean soil depth class and the probability of occurrence of shallow or deep soils. A cross- validation revealed a 65%, 81% and 61% accuracy for the three maps of the shale area, a 50%, 55% and 40% for the maps of the phyllite area, and a 78%, 72% and 75% accuracy for the maps of the limestone area. Explanations for both the good results in the shale and limestone areas, and the poor results in the phyllite area, focus on the effects of the spatial scale of topographic variation, sediment transport mechanisms, and the impact of land use.