Error due to DEM sources in catchment area and river network using D8 algorithm

Abstract

Digital elevation models (DEMs) are considered as an important spatial input parameter for hydrological models and have an inherent source of uncertainties which could be due to grid size or data sources. The acceptability of the DEM source for the research objective and appropriate resolution has to be properly examined before any hydrologic application. These impact the results of the SWAT hydrological model, which predominantly relies on DEMs for the hydrological variables derived. Four popularly used DEMs namely AW3D30 DEM, CartoDEM, SRTM DEM and ASTER GDEM have been considered for the present study in identifying the error in river network extraction using deterministic eight neighborhood (D8) method and error in automatic basin boundary delineation. It was found that newly released AW3D30 global DEM had the best river network extraction capability compared with the digitized river network from google earth imagery. For stream network extraction, the delineation accuracy for AW3D30 was best compared to other DEMs whereas ASTER GDEM showed poor river network extraction capability. On the other hand, SRTM and CartoDEM showed good river network extraction ability. For basin boundary delineation, ASTER GDEM showed the maximum total error of 167.46 km2 with AW3D30 basin boundary as reference boundary. The foremost question of applicability of dataset for the study area has been examined through this case study.