Sensitivity of channel mapping techniques to uncertainty in digital elevation data

Abstract

Digital elevation data are now commonly used to map channel networks automatically. This paper evaluates the sensitivity of six methods of extracting channel networks from digital elevation models (DEMs) to elevation error. As such, the robustness of channel mapping techniques against noisy data was evaluated and not the accuracy of DEM extracted channel networks. Stochastic simulations were used in combination with a fine‐resolution LiDAR DEM of a small upland catchment to assess the robustness of each of the channel mapping algorithms. Findings showed that the four tested methods that were based on identifying patterns or positions in the surface morphology were more sensitive to elevation error than the two methods that were based on simulating overland flow and channelization processes, particularly at lower error magnitudes. While the morphology‐based channel‐mapping methods were highly sensitive to the degree of spatial autocorrelation in elevation error fields, the channelization based methods were relatively insensitive. Drainage network extent and the geometry of exterior stream channels were both found to be greatly influenced by low to moderate degrees of topographic error. Although LiDAR data were found to provide a sufficient resolution for mapping fine‐scale headwater channels, the greater surface roughness did present challenges for automated channel‐mapping techniques.