Neural Net for Determining DEM-Based Model Drainage Pattern

Abstract

Manually determining drainage patterns from topographical maps for a grid-based model is time consuming and occasionally subjective. Eight methods including neural network are developed in this study to automatically determine the pattern from Digital Elevation Model (DEM) data. These methods are tested for a subwatershed located on Chin-Mei Creek, Taipei County, Taiwan, R.O.C. Results obtained using the neural network method are superior to those obtained using the drainage network method, which has performed the best among the other seven methods excluding the neural network method. The neural network method has a self-learning capability that could likely replace human assessment involved in the conventional approach. The implementation of the drainage network and neural network methods is described. Performances of the two methods are compared on the basis of their differences from the manually determined result.