Accuracy assessment of ASTER, SRTM, ALOS, and TDX DEMs for Hispaniola and implications for mapping vulnerability to coastal flooding

Digital elevation models (DEMs) derived from remote sensing data provide a valuable and consistent data source for mapping coastal flooding at local and global scales. Mapping of flood risk requires quantification of the error in DEM elevations and its effect on delineation of flood zones. The ASTER, SRTM, ALOS, and TanDEM-X (TDX) DEMs for the island of Hispaniola were examined by comparing them with GPS and LiDAR measurements. The comparisons were based on a series of error measures including root mean square error (RMSE) and absolute error at 90% quantile (LE90). When compared with > 2000 GPS measurements with elevations below 7 m, RMSE and LE90 values for ASTER, SRTM, ALOS, TDX DEMs were 8.44 and 14.29, 3.82 and 5.85, 2.08 and 3.64, and 1.74 and 3.20 m, respectively. In contrast, RMSE and LE90 values for the same DEMs were 4.24 and 6.70, 4.81 and 7.16, 4.91 and 6.82, and 2.27 and 3.66 m when compared to DEMs from 150 km 2 LiDAR data, which included elevations as high as 20 m. The expanded area with LiDAR coverage included additional types of land surface, resulting in differences in error measures. Comparison of RMSEs indicated that the filtering of TDX DEMs using four methods improved the accuracy of the estimates of ground elevation by 20-43%. DTMs generated by interpolating the ground pixels from a progressive morphological filter, using an empirical Bayesian kriging method, produced an RMSE of 1.06 m and LE90 of 1.73 m when compared to GPS measurements, and an RMSE of 1.30 m and LE90 of 2.02 m when compared to LiDAR data. Differences in inundation areas based on TDX and LiDAR DTMs were between - 13% and - 4% for scenarios of 3, 5, 10, and 15 m water level rise, a much narrower range than inundation differences between ASTER, SRTM, ALOS and LiDAR. The TDX DEMs deliver high resolution global DEMs with unprecedented elevation accuracy, hence, it is recommended for mapping coastal flood risk zones on a global scale, as well as at a local scale in developing countries where data with higher accuracy are unavailable.