Prediction of soil depth using environmental variables in an anthropogenic landscape, a case study in the Western Ghats of Kerala, India

Soil (regolith) depth is a crucial input for modeling earth surface phenomena. However, most studies ignore its spatial variability. Techniques that map the spatial variability of soil depth are of three types: (1) physically-based; (2) empirico-statistical from environmental correlates; and (3) interpolation from point observations. In an anthropogenic landscape, soil depth does not depend primarily on natural processes, making it difficult to apply a physically-based approach. The present study compares empirico-statistical methods with geostatistical methods for predicting soil depth in such a landscape: Aruvikkal catchment (9.5 km2) in the Western Ghats of Kerala, India. Regression kriging applied on blocks of 20 m by 20 m using the environmental covariates elevation, slope, aspect, curvature, wetness index, land use and distance from streams, proved to be the best predictor of soil depth. This model explains 52% of the variability of soil depth in the catchment; with a prediction variance of 0.05 to 0.19. A Gaussian simulation was attempted for a more realistic visualization of the depth, as opposed to the smooth kriging prediction. The most important explanatory variable of soil depth in this landscape is land use, as expected from the strong human intervention.