Geomorphographic terrain classification for predicting forest soil properties in Northwestern Switzerland
Herbst, Philipp; Gross, Jens; Meer, Uwe; Mosimann, Thomas
published: Mar 1, 2012
ArtNo. ESP022005601002, Price: 29.00 €
Digital terrain classification is of elementary importance for model-based prediction of the spatial distribution of soil properties and local water balance. This study describes a terrain analysis method that combines modeled and defined geomorphographic terrain attributes with detailed field knowledge and local mapping experience. 450 forest soil profiles were used to statistically analyze the correlation to soil properties. The result is a two-stage hierarchical terrain classification approach that provides a differentiated and satisfactory reproduction of characteristic terrain attributes in the primary landscape units. As field surveys show, modeled steep slopes, ridges, slope edges, convex and concave slope formations correlate well to local landform elements. Up to now, the results for mid slope modeling have been unsatisfactory. Furthermore, non-relevant or incorrectly delimitated landform elements were also modeled, making manual correction indispensible. In order for terrain modeling to be feasible it is necessary to entail detailed terrain knowledge of the region to be modeled. In this study, statistical analyses of the correlation between soil properties and modeled landform elements were conducted. The results show that steep slopes, convex and concave slope formations, steep stream channels and rock faces correlate well to soil properties. A few single relations were found for plateaus and ridges and trough-shaped valleys. Slope edges and mid slopes show almost no characteristic soil properties. Modeled landform elements are the most essential predictor for the development of prediction models for forest soil properties. The complex morphographic terrain classification approach, described in this study, is a feasible spatial and hierarchical basis for decision-based prediction models.