Documenting GlobalSoilMap.net grid cells from legacy measured soil profile and global available covariates in Northern Tunisia

摘要

In many regions of the planet, GlobalSoilMap.net grid cells can only be documented by Digital soil mapping (DSM) procedures using sparse sets of legacy measured soil profiles and landscape covariates available at the global level. This study aimed to test a DSM methodology for mapping from such an input, in the Cap Bon Region, northern Tunisia (2822 km~2), the soil properties according to the GlobalSoilMap.net specifications. Best possible maps of six primary soil properties at six intervals of depth were produced by selecting and applying, for each of the targeted soil properties, an appropriate DSM function among three candidates (linear regression, ordinary kriging and regression kriging). The selection was made from decision criteria that resulted from an exploratory analysis involving statistical tests. Only a small part of soil variability of the study area could be captured and it was at best obtained a decrease of one third of the widths of the 95% confidence intervals that expressed a prediction of soil property. These results could be explained by the low rate of variability of some properties and the predominance of short scale variations that could not be modeled from the sparse set of profiles. The merit of such a DSM approach is to identify the gaps that needs to be filled to provide data that will meet higher expectations, outlying the need for more investment in soil data.