Evaluation of global and continental scale soil maps for southern Africa using selected soil properties

Earth-systems modelling of the critical zone often require soil maps that represent in-field conditions. Some developing regions, such as southern Africa, do not have detailed soil data for all areas and are often reliant on data from global soil maps and data sets. This raises the question of global and regional soil map and soil data accuracy at local levels. Although previous studies have indicated that global soil maps do not always provide an accurate representation of in-field soil conditions, it is not clear how inaccurate these maps and data are for southern Africa. This study assessed the accuracy of widely used global and continental-scale soil maps for southern Africa. The accuracy of the soil maps was assessed by comparing profile and mapped soil data. Three soil maps of global, continental, and regional scale (Harmonized World Soil Database version 1.2, SOTER for southern African, and AfSoilGrids250m) were compared to observation data from two profile data sets (Africa Soil Profile Database and South African Soil Profile Database). Furthermore, maps of selected soil properties of the AfSoilGrids250m rasters were compared to the Harmonized World Soil Database map. Soil properties were statistically compared through Lin's Concordance Correlation Coefficient, Root Mean Square Error, and Mean Absolute Error. Investigated maps and point data generally correlated poorly and large Root Mean Square Error and Mean Absolute Error values were observed for most of the properties. Of the investigated soil maps, the digital AfSoilGrids250m soil property raster layers provided the most accurate estimate of in-field soil conditions for selected soil properties in the study region.