Propagation of Uncertainty in Multi-Source Elevation Models into Hydrological Modelling and Sediment Connectivity in a Loess Catchment

The accuracy and resolution of digital elevation models (DEMs) determine the reliability of outputs used in hydrological modeling. However, an assessment of their suitability should not be limited to vertical accuracy alone. Particularly in small, highly fragmented catchments, uncertainty in elevation representation can propagate to subsequent derivatives depending on the local topographic structure and, consequently, influence the interpretation of runoff organization and sediment transport. The aim of this study is to determine how differences between multi-source elevation models propagate from the underlying elevation representation through slope and runoff organization to the drainage network and the spatial pattern of sediment connectivity. In the Stocki Creek catchment in the Lublin Uplands (SE Poland), ASTER GDEM, SRTM, EU-DEM, and a local TIN model were compared. Aerial laser scanning (ALS) data were used as a high-resolution reference for validating elevation and derived outputs. Vertical accuracy was assessed using RMSE and NMAD, and then the representation of relief, slope, hydrologic routing outputs, the drainage network, and the Sediment Connectivity Index (SCI) were compared. The smallest discrepancies relative to ALS were obtained for the TIN (RMSE 1.48 m; NMAD 0.41 m), and the largest for ASTER (6.78 m; 2.37 m). Errors were highest in forested areas. All models reproduced the general elevation pattern of the watershed relatively well, whereas differences were much more pronounced in products dependent on local gradients and topographic continuity. Satellite models smoothed out the steep slopes of gullies, generalized short runoff lines, and led to a simplification of the spatial pattern of sediment connectivity. The results show that the usefulness of elevation models in hydrological applications should be assessed not only by their vertical accuracy but also by the propagation of uncertainty to subsequent processed products. Validation against ALS therefore allows for an assessment of the reliability of multi-source Earth Observation (EO) products throughout the entire analytical chain: from elevation, through local morphometry and runoff routing, to the drainage network and sediment connectivity.