Suitability of LiDAR point density and derived landform curvature maps for channel network extraction

Abstract

This study uses landform curvature as an approach for channel network extraction. We considered a study area located in the eastern Italian Alps where a high-quality set of LiDAR data was available and where channel heads and related channel network were mapped in the field. In the analysis, we derived 1-m DTMs from different ground LiDAR point densities, and we used different smoothing factors for the landscape curvature calculation in order to test the suitability of the LiDAR point density and the derived curvature maps for the recognition of channel network. This methodology is based on threshold values of the curvature calculated as multiples (1–3 times) of the standard deviation of the curvature. Our analyses suggested that (i) the window size for curvature calculations has to be a function of the size of the features to be detected, (ii) a coarse ground LiDAR point density could be as useful as a finer one for the recognition of main channel network features and (iii) rougher curvature maps are not optimal as they do not explore a sufficient range at which features occur, while smoother curvature maps overcome this problem and are more appropriate for the extraction of surveyed channels. Copyright © 2010 John Wiley & Sons, Ltd.