Digital Soil Mapping

Abstract

The Carbon/Nitrogen ratio (CN) of forest soils is one of the best predictors for evaluating the soil functions mainly involved in climate change issues. The CN ratio of forest litters depends generally on tree species and forest management which are local factors, but also on broader environmental factors. Thus, the European forest litter CN ratio map is predicted using: (a) punctual CN ratio measurements collected systematically every 16 km in European forests and analyzed according to a common European laboratory method; (b) spatially continuous information on tree species abundance (derived from interpolation) and climate, landform and lithology at 1 km resolution. The spatial modeling of the CN ratio is done according to complementary approaches: first, a classical kriging approach done on the CN ratio measurements; and second, a neural network approach using a set of nonlinear equations on the environmental predictors. Other multivariate geostatistical approaches were tested but not retained for final results due to lack of correlation between environmental factors. Twenty percent of CN ratio measurements are kept for validation purpose. The two approaches are compared using coefficient of determinations and Root Mean Square Errors on the validation dataset. Surprisingly, the best approach is the classical kriging, meaning that the spatial structure and variability of CN ratio cannot be explained by the environmental factors, which show high local variation. This leads to a discussion of the quality of the data and to envisage possible risks for global digital soil maps.