Integrating drainage enforcement into existing raster Digital Elevation Models
Laing, Boyd R.
Initial evaluation of the "hydrologically correct" HYDRO1k (2001) data set, found that the drainage network derived from the topographic data using the TOPographic P ArameriZation (TOPAZ) software was inconsistent with the blue line data in the National Atlas of Canada data. Further analysis found minor elevation errors in the HYDRO1k DEM were causing drainage network inconsistencies in some relatively flat and low elevation areas. Implementation of the Australian National University Digital Elevation Model (Hutchinson 1989) algorithm for drainage enforcement corrected the drainage networks, however, it adversely affected the DEM characteristics by substantially lowering high elevation values and created a general smoothing effect across the DEM surface. It was felt that a method that captures ANUDEMs drainage enforcement capabilities, but limits the effect on the surrounding DEM outside the drainage network area would be beneficial. The focus of this project was to develop a new method for the development of a hydrologically correct Digital Elevation Model (DEM). Specifically, this project uses a presently accepted procedure, ANUDEM, developed by Hutchinson (1989), and builds upon it using a series of drainage network based buffers. The new method uses a simple distance weighting of elevation values based on proximity to the drainage network to weight elevation values closest to the drainage network more heavily to ANUDEM calculated values and to give ANUDEM values less weight as the distance moves further away towards the edge of the buffer. This approach effectively minimizes the effect of ANUDEM processing on areas away from the drainage networks. Three drainage basins DEMs were used to test the new procedure, each exhibiting a particular problem for hydrological modeling and automated delineation of drainage networks. The first DEM was extracted from the HYDRO1k database and represented the Saskatchewan River Basin at a resolution of 1 km. It contained minor elevation errors that gave rise to incorrect drainage delineation. The second DEM represented the Mackenzie River Basin at a resolution of 2 km and, as part of an earlier scaling study, had been derived by aggregation of an initial 1 km resolution DEM. The aggregation process introduced several drainage network errors. The third DEM represented the Snare River Basin at a resolution of 100 m. This basin had a very small elevation range nad chaotic drainage pattern which make it very difficult to model. Following implementation of the new procedure for the Saskatchewan River Basin and the Mackenzie River Basin. TOPAZ provided drainage networks that were consistent with the corresponding blue line networks in the National Atlas of Canada data set. Examination of the DEM characteristics found that effects of the new procedure on DEM characteristics were minimal. The range of elevations was maintained and the mean elevation of the DEM is statistically unchanged using a 1, 3 and 5 cell buffer width. Using 10 and 20 cell buffers, also provided correct drainage networks, however, effects on the DEM characteristics started to increase. In the case of the Snare River Basin, ANUDEM was unable to process the DEM for the Snare River Basin presumably because a threshold of drainage network complexity and topographic relief was exceeded.