Comparative Analysis of 30-m DEM Products for Hydrological Applications: A Case Study in the Flinders Catchment Australia

Digital Elevation Models (DEMs) are fundamental to hydrological modelling, watershed delineation, flood hazard assessment, and resource management. However, the reliability of these applications depends heavily on the vertical accuracy of the DEMs. Although several global DEM products with 30-m spatial resolution are widely available, variations in sensor technology, data acquisition methods, and surface characteristics can significantly influence their accuracy and suitability for hydrological studies. This research provides a comparative evaluation of five commonly used global DEMs—TanDEM-X, ASTER GDEM, SRTM, Copernicus DEM, and ALOS World 3D—by assessing their vertical accuracy against high-resolution airborne LiDAR data and ICESat-2 ATL06 measurements. The findings aim to inform best practices for selecting DEMs in hydrological modelling and catchment-scale applications, particularly in data-scarce regions.

The Flinders River catchment in northern Queensland was selected as the critical test area for evaluating how DEM errors propagate into hydrological calculations. This region is characterised by low rainfall and pronounced topographic variability, encompassing flat lowland plains, dissected upland terrain, and localised areas of steep slopes. All DEMs were standardised to a common horizontal and vertical reference framework and co-registered with the test datasets to eliminate systematic discrepancies. ICESat-2 ATL06 data were rigorously filtered to retain only the highest-quality measurements, based on a combination of quality flags, topographic slope thresholds, and signal strength criteria in vegetated areas.

Elevation differences were computed at matched locations, and DEM performance was evaluated using key statistical metrics, including bias, root mean square error (RMSE), mean absolute error (MAE), median error, and standard deviation. To provide a more comprehensive assessment, error behaviour was analysed in relation to terrain slope and catchment characteristics, highlighting zones most vulnerable to error propagation in flow routing and watershed delineation. Systematic patterns in DEM error were further examined with respect to sensor characteristics under varying landscape conditions.

Results indicate that TanDEM-X and Copernicus DEM exhibit the highest vertical accuracy, closely aligning with ICESat-2 and LiDAR observations, whereas ASTER GDEM and SRTM show larger mean errors, particularly in dissected or mountainous terrain. These findings suggest that TanDEM-X and Copernicus DEM are preferable for hydrology-focused applications in semi-arid basins, while ASTER and SRTM should be used cautiously where precise modelling is required. The study underscores the importance of DEM accuracy evaluation in relation to basin characteristics, as errors can significantly influence hydrological modelling outcomes.