Flächenhafte Modellierung von waldbodeneigenschaften in der Nordwestschweiz

Abstract

Forest soils are an important natural resource. However, up to now almost no area-wide forest soil information was available for Switzerland. Since 2006, model-based, high-resolution maps of forest soil properties in the cantons Basel-Landschaft and Basel-Stadt were generated, depicting soil depth, stone content, soil wetness, acidity and water storage capacity. These maps are based on all available point information on soils, and on 95 recently analyzed pedological forest soil profiles. Two different methods were applied in parallel: 1) the development of decision trees based on frequency statistics combined with expertise and 2) the semi-automated Random Forest modeling approach. Highly branched hierarchical decision trees were used to derive soil properties from 24 predictors (relief forms, parent material, vegetation, forest type, location, climate, etc.). This article describes the approaches and portrays the mapped results of soil depth, top soil acidity and water storage capacity. Our project shows that it is basically feasible to predict soil properties with a high spatial resolution, classifying them into 4-5 categories. However, depending on soil type, for 10-30% of the area no predictions are possible, especially because of high soil heterogeneity, inadequate morphographic slope differentiation in the terrain models and the implausibility of predictor information. Soil property maps provide basic information for set up forestry maps for forest development, forest management and risk assessment.