Comparison of Digital Elevation Models, Stream Networks, and Land Surface Model Input Datasets in Flood-Mapping Case Studies: The 2022 Pakistan and 2024 North Carolina Floods

Flood forecasting and predicting the sustaining impacts is vital in protecting people and infrastructure, particularly as increased intensities and frequencies of precipitation events continue to threaten populations globally. Flood maps are an important component of these predictive efforts and communicating potential impacts of a given flood event. While advanced tools and frameworks exist to conduct large-scale forecasts, there are several variations and limitations that challenge the accurate mapping of global flood events. Key among these are uncertainties in precipitation forcings and accurate streamflow estimates—in both timing and magnitude and capturing effects of existing infrastructure. Digital elevation model (DEM) accuracy is another critical component in flood extent modeling. With growing investments in remote sensing, new elevation datasets are becoming more readily available. Therefore, in this study, we test three global DEM products and two stream networks, derived from the DEM products, to ascertain their effectiveness in flood extent mapping using a model cascade simulating a forecast system design to produce regional streamflow estimates and flood extents. Simulations of recent large flooding events in Pakistan in 2022 and North Carolina in 2024 served as the test domain for this study. An increase in DEM resolution proved to be the most significant factor in improving overall accuracy, with simulations using the TanDEM-X 12 m DEM producing the highest accuracy in both study areas (Pakistan: 90.84%, North Carolina: 85.31%), though it held consistent bias to underprediction in Pakistan (E=0.28–0.65). Impacts from different sources of streamflow and hydrography datasets were minimal in the two study domains, but coupling stream networks derived from the same DEM product used in the mapping improved accuracy within each DEM subset. This research provides insight into the effectiveness of these datasets for simulating large-scale flood events by replicating the data, methods, and workflows leveraged in a developing forecast system.