LAUSR

Abstract Parent material is a crucial co-variate in predicting soil properties using digital soil mapping (DSM) methods. This spatial information can be obtained using available lithology maps, or using proxies such as gamma-ray spectroscopic maps. In this study, we used random forests to predict topsoil texture (clay, silt, and sand in grams per kilogram) in a French sub-region using a high density of soil measurements and available co-variates including climate, topography, land use, and satellite data. Then, we tested the value of adding a lithology map at the 1:50,000 scale and/or an airborne gamma-ray spectroscopy map in a French region characterised by a considerable contrast in geology and lithology. We showed that adding airborne gamma-ray spectroscopic data substantially increased the indicators of prediction performance and led to less noisy and more interpretable maps for this region. These results suggest that airborne gamma-ray spectroscopy can be a very useful co-variate to predict these topsoil properties.