Comparison of watershed boundaries derived from SRTM and ASTER digital elevation datasets and from a digitized topographic map
J.K. Pryde; J. Osorio; M.L. Wolfe; C. Heatwole; B. Benham; A. Cardenas
Type of Document:
Conference Proceeding or Document
Date of Publication:
Place of Publication:
Abstract: Watersheds are natural integrators of hydrological, biological, and geological processes and as such require an integrated approach to data analysis and modeling, which usually starts delineating accurately a polygon vector layer of watershed boundaries as input. In that way, the Rio Illangama watershed in Alto Guanujo, Ecuador, had been isolated with the objectives of evaluate the accuracy of watershed boundaries derived from three different sources: One was delineated by hand and the other two were derived from a 30-m ASTER DEM and a 90-m SRTM DEM, using the Spatial Analyst extension of ArcGIS. Visually, there are small differences between the manually-delineated and the SRTM-based boundaries, while the ASTER-based varies from the manually-delineated one. The area of the watershed delineated manually is 13,061.3 ha, while the SRTM-based and the ASTER-based watershed are 0.66% and 2.6% larger. The regression analyses comparing the complete boundaries yielded an R2 of 0.999 between the SRTM and manual boundaries and the 0.988 for the ASTER and the manual boundaries. The t-test comparing DEMs indicated a significant difference (p< 0.001) in the distance differences. To determine the cause of the errors in the ASTER DEM, map algebra was used to define where the "Fill" tool had filled the sinks finding that the errors in the stream network occurred where some especially large (60 to 100 m) fills had occurred. Then the ASTER-DEM was corrected and processed to obtain a new watershed boundary with almost no difference with the Hand-drawn boundaries. Therefore, the accuracy of the watershed delineation depends on the first place on the accuracy of the Digital Elevation Model available (DEM).