Advanced Hydrologic Conditioning of the TanDEM-X DEM to Produce Version 2 of the HydroSHEDS Database

Kurzfassung

For over a decade, the HydroSHEDS database has provided foundational global hydrographic data to support large-scale hydroecological research and applications. Providing seamless global river networks and catchment delineations at multiple scales, the HydroSHEDS database frequently forms the basis of regional and global models, assessments, and management applications across a wide range of disciplines including environmental, conservation, socioeconomic, human health, and sustainability studies. HydroSHEDS v1 was derived from Shuttle Radar Topography Mission (SRTM) elevation data at 3 arc-second resolution, using a combination of automated processes and extensive manual corrections. HydroSHEDS v1 has important limitations, particularly in areas above 60˚ of northern latitude as SRTM data were not available in this region. The TanDEM-X Digital Elevation Model (DEM) provides an opportunity for an update to the HydroSHEDS database which will offer significantly improved hydrographic data, while displaying the same structure and characteristics as HydroSHEDS v1. Version 2 of the HydroSHEDS database, derived from the 0.4 arc-second TanDEM-X DEM, is currently being created in a collaboration between the German Aerospace Centre (DLR), McGill University, Confluvio Consulting, and the World Wildlife Fund. HydroSHEDS v2 is produced at 1 arc-second resolution with full global coverage, using the same basic specifications as HydroSHEDS v1. The high-resolution TanDEM-X DEM is pre-processed by DLR at its original 0.4 arc-second resolution before being upscaled to a preconditioned 1 arc-second DEM for HydroSHEDS which preserves detailed features of the original data. The preconditioned DEM is then refined using customized, advanced conditioning algorithms incorporating recent, high-resolution auxiliary data products and manual corrections to ensure the highest level of accuracy and quality control. The conditioning process of HydroSHEDS v2 transforms the preconditioned 1 arc-second DEM into a hydrologically conditioned DEM which can be used to derive accurate drainage pathways. As a first step, surface corrections are applied to vegetated and urban areas using a combination of high-resolution global forest height maps, NDVI values, and World Settlement Footprint data to reduce distortions caused by vegetation and built-up areas. The resultant DEM is modified using filters designed to remove spurious errors in the DEM, breach obstructions such as bridges, and maintain connectivity within river channels. Finally, open water areas are ‘burned' (i.e., systematically lowered) using a global map which comprises a combination of high-resolution terrestrial water surface maps complemented by a water body map developed for HydroSHEDS by DLR using a machine learning (ML) detection approach. In all open water areas, refined smoothing and centre-line tracing algorithms are applied to ensure logical and consistent flow through the water bodies, including rivers and lakes. Drainage pathways are derived from this treated DEM using a custom algorithm which traces flow paths through nested depression structures, such as wetlands, to preserve terrain details within complex landscapes. In this process, candidate depressions of potential inland drainage (sinks) are identified and added to the sink locations already identified in HydroSHEDS v1. In combination with the automatic conditioning algorithms, iterative manual stream-burning and sink identification corrections are applied to the DEM in a final quality control process. The resulting, fully conditioned DEM is used to produce the core gridded flow direction data by tracing the steepest descents within 3x3 grid cell kernels which form the foundation of the global, seamless river networks and catchment boundaries of the HydroSHEDS v2 database.