Adding Gaussian noise to inaccurate digital elevation models improves spatial fidelity of derived drainage networks

Abstract

An economical approach to improving predictions of hydrological models produced highly accurate representations of ephemeral and perennial stream networks. Traditional drainage network extraction from digital elevation models (DEMs) often yields inaccurate and inconsistent results because of elevation errors. Topographic wetness index maps calculated from alternative terrain representations, produced by adding random errors to a DEM of a subwatershed with low relief, were combined to delineate a stream network that matches one produced by more time-intensive (and costly) differential Global Positioning System (GPS) field methods, particularly with respect to the ephemeral component of the drainage network.