Integrating hydrological knowledge into deep learning for DEM super-resolution

Deep learning-based super-resolution methods have been successfully applied to DEM downscaling studies by designing structures and loss functions of the model. However, the design of super-resolution models with enhanced performance remains a challenge, particularly in the context of hydrological studies, which require high-resolution DEMs with correct hydrological characteristics. In this study, we introduce a super-resolution model that integrates hydrologic knowledge (HKSRCGAN), aiming for the model to effectively maintain topographic features as well as the hydrologic availability of the DEMs. FABDEM with 30 m spatial resolution，which removes buildings and vegetation, was used in the experiment to demonstrate the usability of the proposed method. The hydrological knowledge derived from surface flow direction and hydrological features are integrated into a deep learning algorithm to guide model training. The results show that the HKSRCGAN outperforms the bicubic interpolation, SRCNN, SRGAN, SRResNet methods in reducing topographic errors and maintaining hydrologic characteristics. In the test area, the entropy difference analysis shows that the DEM generated by HKSRCGAN is more similar to the information contained in the reference DEM. Furthermore, super-resolution models integrating hydrological knowledge are valuable for modeling terrain shaped mainly by gravity and surface water flows. In addition, deep learning-based models integrating hydrologic knowledge are expected to be applied in DEM upscaling to maintain consistent hydrological characteristics.