The role of spatial resolution and vertical accuracy of global DEMs in delineating stream networks

The efficacy of global digital elevation models (DEMs) derived from satellite observations is of paramount importance for hydrological applications, including stream network delineation. However, their effectiveness is constrained by spatial resolution and vertical biases introduced by vegetation and man-made structures. This study evaluates the performance of various global DEMs, including bare-earth models, in a Central European mountainous region. Utilizing lidar-derived digital terrain models (DTMs) and national stream networks as reference data, this study shows that bare-earth DEMs have higher vertical accuracy than conventional global DEMs across all land cover types. However, it is concluded that vertical accuracy alone is not a reliable indicator of a DEM's capacity to delineate stream networks. Instead, terrain slope and land cover exert a more significant influence on the accuracy of the delineated stream network. While the use of higher-resolution DEMs (e.g., 12 m TanDEM-X) has been shown to improve stream delineation, it has also been demonstrated to amplify errors related to vegetation bias. The study emphasizes that DEM selection should be based on their performance in delineating stream networks, rather than solely on vertical accuracy. The findings underscore the necessity of enhanced vegetation and building offset removal in high-resolution DEMs to optimize their applications in hydrology.