Visual and statistical comparison of ASTER, SRTM, and Cartosat digital elevation models for watershed

Abstract

Accurate delineation of watershed and drainage networks is crucial for hydrological and geomorphological models, water resource management, change of floodplains, flood risk management, and surface water mapping. Since high-resolution digital elevation models (DEMs) are often not available, it is necessary to evaluate open source products. Various statistical measures were used to estimate the vertical accuracy of these freely available DEMs. Moreover, DEM products from Shuttle Radar Topography Mission (SRTM), Advanced Thermal Emission and Reflection Radiometer (ASTER), and Cartosat data were also compared. The study areas are located in the Vadodara district of Gujarat State of India. A comparison of SRTM-, ASTER-, and Cartosat-derived DEMs allowed a qualitative assessment of the vertical component of the error, whereas statistical measurements were used to estimate their vertical accuracy. In order to compare the frequency histograms of the elevation distributions in the DEMs in the study area, skewness and kurtosis were determined. Further, to obtain the degree of relationship between the DEMs, scatterplots, as well as correlation coefficients, were used. The results showed that all DEMs have imperfections in the delineation of a small river like Vishwamitri, and the comparison showed that SRTM 30 m and ASTER 30 m failed to delineate proper main drainage for the river. Cartosat 30 m DEM exhibited better results. The root mean square error (RMSE) was calculated as 7.21 m for ASTER and 3.24 m for SRTM. The correlation value of 0.94 indicates the existence of a strong positive linear correlation between SRTM and Cartosat. The study shows that for the study area ASTER elevation data were highly underestimated, whereas SRTM elevation data were slightly overestimated.