Hillslope surface classification from elevation models by using normal vectors

Abstract

Hillslope surface classification via digital terrain analysis (DTA) is a current research focus in geomorphology and geographic information science (GIS) studies. However, traditional methods are generally based on raster digital elevation models (DEMs) and window difference methods, which suffer from terrain description and analysis scale mismatch issues. In this study, we propose a hillslope surface classification approach based on the vector structure. Triangulated irregular networks (TINs) are used as an example. Benefiting from free sampling with TINs, we apply terrain surface reconstruction to the original point cloud or DEM and then calculate the terrain derivatives based on the normal vectors of the optimized TINs. Finally, the fuzzy inference method is used to classify hillslope surface elements. We select two cases to evaluate the proposed method: a small watershed with dense point cloud data and a large region with complex landforms and a 30 m resolution Copernicus DEM. The results show that the proposed approach can effectively reduce the influence of DEM errors on classification and mitigate the scale mismatching problem in terrain generalization and analysis. A novel hillslope surface classification method with a new data structure is proposed to extend the application of vector methods and structures in DTA and GIS.