Modelling soil particle-size distribution in the region “Centre” (France)
Résumé
The objective of this study is to quantify the relationships between soil texture fractions and 1) other soil properties at the profile level; 2) environmental covariates at the regional scale (Region “Centre,”France). We used 2487 soil profiles and related soil analysis of the French soil database. The covariates we used refers to all the available data concerning: i) topography, using different DEMs [i.e., SRTM and ASTER DEM] to generate terrain parameters and topographic indexes; ii) soil cover, using Corine land cover and oth-er derived maps; iii) imagery, using gamma ray maps; iv) other thematic maps, using available geological and soil cartography of the region at scales 1M, 250k, and 50k. Particle-size data were modelled as compositional data operating alr-logarithmic transform in order to preserve dependence among the studied variables (e.g., clay, silt, and sand fractions should sum to 100%). For all the modelling procedures, we used boosting regres-sion trees (BRT). Results indicate that at the soil profile level, particle-size fractions are well explained by sample depth, soil type, and parent material, while at the geographical level, the best predictors are geological and soil cartographic units, gamma rays, and land cover