Scenario-Based Validation of Moderate Resolution DEMs Freely Available for Complex Himalayan Terrain

Abstract

Accuracy of the Digital Elevation Model (DEM) affects the accuracy of various geoscience and environmental modelling results. This study evaluates accuracies of the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global DEM Version-2 (GDEM V2), the Shuttle Radar Topography Mission (SRTM) X-band DEM and the NRSC Cartosat-1 DEM V1 (CartoDEM). A high resolution (1 m) photogrammetric DEM (ADS80 DEM), having a high absolute accuracy [1.60 m linear error at 90 % confidence (LE90)], resampled at 30 m cell size was used as reference. The overall root mean square error (RMSE) in vertical accuracy was 23, 73, and 166 m and the LE90 was 36, 75, and 256 m for ASTER GDEM V2, SRTM X-band DEM and CartoDEM, respectively. A detailed error analysis was performed for individual as well as combinations of different classes of aspect, slope, land-cover and elevation zones for the study area. For the ASTER GDEM V2, forest areas with North facing slopes (0°–5°) in the 4th elevation zone (3773–4369 m) showed minimum LE90 of 0.99 m, and barren with East facing slopes (>60°) falling under the 2nd elevation zone (2581–3177 m) showed maximum LE90 of 166 m. For the SRTM DEM, pixels with South-East facing slopes of 0°–5° in the 4th elevation zone covered with forest showed least LE90 of 0.33 m and maximum LE90 of 521 m was observed in the barren area with North-East facing slope (>60°) in the 4th elevation zone. In case of the CartoDEM, the snow pixels in the 2nd elevation zone with South-East facing slopes of 5°–15° showed least LE90 of 0.71 m and maximum LE90 of 1266 m was observed for the snow pixels in the 3rd elevation zone (3177–3773 m) within the South facing slope of 45°–60°. These results can be highly useful for the researchers using DEM products in various modelling exercises.