Towards user needs assessment for DSM products

Digital Soil mapping (DSM) provides standardised information layers. The recent availability of global and continental remote sensing-derived products coupled with the ease-of-access to computational resources has made the production of such layers easier. It is ever to characterise and evaluate such DSM-derived products, in particular the type of actual information they can provide to users.  

 

DSM studies commonly assess prediction uncertainty using various approaches, including multiple simulations or quantile random forests. These studies provide measures of accuracy derived from statistical (cross-)validation, often based on non-probability and non-representative observations. However, these accuracy metrics and uncertainty assessments do not encompass all the potential elements that could be used to characterise a DSM product, and they do not directly address the needs of the users. We assessed maps based on area of applicability (i.e., the area in covariate space where the model learns about relationships based on the training data), the landscape heterogeneity both in the landscape itself and in covariate space, and the local influence of the covariates on the final products.  

 

We present examples of continental and global mapping products, highlighting main accuracy, uncertainty and interpretability aspects and how these influence their suitability for intended use by stakeholders, decision makers and users in general at the given resolution. The results permit some practical reflections on how to integrate all the above elements to identify regions where the confidence in the predictions is highest and the associated uncertainty lowest, but also, where the product is not considered fit for the intended use.