An Assessment Of The Impact Of Dem Interpolation Technique, Resolution, And Terrain Type On The Extraction Of Drainage Network

Abstract

This research used points extracted from high-resolution DEMs (1m) to investigate the impact of resolution, interpolation method and topography on the accuracy of drainage network extraction. The investigation was conducted by evaluating the accuracy of the estimations of streams length, streams number, drainage density, and the Longitudinal Root Mean Square Error (LRMSE) of the extracted drainage networks from different DEMs interpolated using Topo to raster, Natural Neighbor (NN), kriging and IDW interpolation methods at 5, 10, 15 and 20m resolutions over moderate, steep, and gentle slope terrain. Each evaluation conducted yielded a different result, but the accuracy of the streams length estimation for most of the DEMs at all the sites increases with an increase in streams order. The total lengths of all the streams of each of the extracted networks at gentle and steep slope sites are shorter than those of the corresponding reference networks though, 15 and 20m kriging and IDW DEMs created longer streams at the moderate slope site. IDW DEMs have proven reliable for streams length estimation while Topo to raster 5, 10, and 15m for streams number estimation. In general, N.N. extracted networks are the only networks that show consistency in the streams length and number estimations, drainage density estimation as well as in LRMSE and DEM RMSE computation at all the resolutions and for all the sites. Therefore, the accuracy of N.N. DEMs and their derivatives do not rapidly change with change in resolution, especially between 5 and 20m at all (steep, gentle and moderate) terrain types.