A semi-automatic derivation of channel network from high-resolution DTMs: the example of an italian alpine region

High-resolution digital terrain models (HR-DTMs) of regional coverage open interesting scenarios for the analysis of landscape, including derivation and analysis of channel network. In this study, we present our experience in the derivation of the channel network from a HR-DTM for the Autonomous Province of Trento. This region (about 6200 km2) has a complex orography, with a high variety of lithotypes and different types of land use (forests, bare ground at the highest elevation, rocky outcrops and debris, mountain grasslands and agricultural areas mostly located in valley floors). A preliminary automatic extraction of the raw channel network was conducted using a curvature-based morphological algorithm (Tarboton and Ames, 2001) applied to a 2 m resolution DTM derived from an airborne LiDAR survey carried out in 2006. This algorithm shows a good performance in terms of drainage density and channel heads location when preliminarily compared to available cartography.The raw channel network automatically extracted from the HR-DTM needs a supervised control to check the spatial pattern of the channels and to classify every reach of the hydrographic network. This analysis was carried out by means of different informative layers grouped into three main classes: i) geomorphometric indexes derived from HR-DTM encompassing 2 m resolution shaded relief map, openness (Yokoyama et al., 2002), local anomalies, and plan curvature; ii) high-resolution orthophoto imagery (panchromatic and near-infrared); iii) technical cartography (1:10000, 1:25000 scale).The result is an accurate and scale independent channel network definition of the study area that therefore needs to be complemented and corrected via field surveys and expert supervision, especially in urban areas. Summary statistics are presented to relate the drainage density to different topological and geomorphological parameters.