Effects of digital elevation model accuracy on hydrologic predictions

The effect of vertical accuracy of digital elevation models (DEMs) on hydrologic prediction accuracy was evaluated by comparing three DEMs and associated stream-flow simulations for the 7.2 km² USDA-ARS watershed at Mahantango Creek, PA. The DEMs were the standard 30 m USGS 7.5′ DEM, a 5 m product derived from low altitude aerial photography, and a 30 m product derived from interferometric processing of Spaceborne Imaging Radar-C (SIR-C). Statistical analysis of the DEMs showed that the USGS DEM had the typical stipling errors resulting from processing of the source digital contour maps, and a vertical error structure related to the topographic attributes of the watershed. The SIR-C DEM had a vertical offset of about 50 m from the high resolution and USGS DEMs, as well as error features that were somewhat related to topographic features. Inaccuracies in both the USGS and SIR-C DEMs were apparent in the drainage network, as well as in spatial images of elevation, slope; and contributing area. Comparisons of runoff predicted using a hydrologic model based on the three DEMs showed that mean annual predicted runoff volumes were 0.3% and 7.0% larger for the USGS and SIR-C DEMs, respectively, as compared to the reference DEM. Much larger differences were apparent in individual hydrographs; and the USGS and SIR-C DEMs predicted lower peaks, and higher base flows, than did the reference.