Evaluating the Relative Importance of Legacy Soil Sampling and Spatial Models in Digital Soil Mapping Performances: A Case Study in Languedoc-Roussillon (Southern France)

Abstract

A growing set of real-life applications of digital soil mapping (DSM) is now available across the planet. These DSM applications need to be thoroughly analyzed for identifying the corrective actions that will provide the best increase in performances. In Languedoc-Roussillon, the analysis of performances of three DSM models applied for mapping 29 soil properties showed that DSM performances were mainly driven by the ability of the spatial sampling to capture the variability of soil properties, itself driven by the sampling density and the intrinsic scale of the soil property variations. In this region, increasing the sampling density of soil measurements appeared therefore as the priority instead of looking for a more efficient DSM model. We recommend the extension of our approach for analyzing further DSM results.