Authors: Sudip Pathak, Dan Ames, Riley Hales – Brigham Young University; Iman Maghami – Aquaveo
Title: Multi-Metric Evaluation of NWM Medium-Range and GEOGLOWS Streamflow Forecasts for Peak Events Across Lead Times in Humid Subtropical (Cfa) Regions
Presentation Type: Poster Presentation
Abstract: Accurate streamflow forecasting is critical for effective flood warning and water resources management. This study evaluates the NOAA National Water Model (NWM) and GEOGLOWS across 131 unregulated USGS stream gauges in the Cfa (humid subtropical) Köppen climate zone, using the Peak-Over-Threshold (POT) method to identify 252 flood peak events. Results reveal that both models struggle to capture observed peak flows within their ensemble range, with median coverage below 15% and predominantly negative KGE values across all lead times. Despite this shared weakness, GEOGLOWS shows a modest coverage advantage owing to its larger 51-member ensemble, and better peak timing (Tp) at longer lead times, while NWM improves at shorter lead times, likely aided by its hourly Analysis and Assimilation (AnA) cycle. Both models consistently underpredict peak flow magnitude (Qp) and flood volume (FV) regardless of lead time. Performance becomes more nuanced with catchment complexity, higher-order streams (SO ≥ 5), and larger watersheds (1,000–10,000 km²) show highly variable KGE, while Tp, Qp, and FV bias improve, and coverage remains consistently low across all stratifications. These results provide actionable insights for forecasters and decision-makers, offering a foundation for improving ensemble calibration, bias correction, and the reliability of operational flood early warning systems.