We used 2071 forest soil profiles to model a wide range of soil properties for the forested area of Switzerland. The spatial prediction is based on the principle of «digital soil mapping». This involves linking soil profiles with soil forming factors using statistical or machine learning methods.
A quantile regression forest (QRF) approach was applied to predict the following soil properties at six depth ranges: clay, gravel, sand, fine earth density, SOC. The depth ranges correspond to the standard depths of the GlobalSoilMap.Net specification: 0–5, 5–15, 15–30, 30–60, 60–100 and 100–200 cm. In addition, the total soil depth down to a non-root-permeable layer or solid rock soil thick was predicted. To quantify the uncertainty for each predicted pixel, the upper and lower limit of the 90% prediction interval derived from QRF was calculated. More details on the methods and results are described in Baltensweiler et al. 2021 and Baltensweiler et al 2022.
The soil property maps, and the uncertainty maps are provided as a GeoTIFF files at 25 m resolution. The excel file (xlsx) provides a short description of the raster layers.
The soil and the uncertainty maps can be viewed in a simple web-GIS application available at:
www.wsl.ch/soilmaps.