Authors: Aline Falck, Kristen Sanfilippo and Abby Frazier – Water Resources Research Center, University of Hawai’i at Mãnoa (WRRC-UHM); Andy Wood – Colorado School of Mines
Title: Probabilistic Flood Forecasting for Tropical Islands Using the NextGen NWM and Downscaled Ensemble Forcing Data
Presentation Type: Poster Presentation
Abstract: High-resolution ensemble-based meteorological datasets are essential for advancing hydrological modeling and hazard assessment, particularly in regions characterized by complex terrain and strong spatial variability. The Hawaiian Islands exhibit some of the most pronounced precipitation gradients globally, driven by the interaction of trade winds with steep topography. Deterministic forecasts are often insufficient because they do not fully capture the intensity, timing, and uncertainty of short-duration precipitation extremes that govern flood response. This study presents a probabilistic flood-forecasting framework for tropical islands by integrating statistically downscaled ensemble meteorological forcing with the NextGen National Water Model. We leverage operational ensemble forecasts from the European Centre for Medium-Range Weather Forecasts, selected for their long historical record, large ensemble size, and high spatial and temporal resolution. These data will be statistically downscaled using Generalized Analog Regression Downscaling (GARD) to produce hourly ensemble precipitation fields over Hawaiʻi. This high-resolution ensemble-based forcing is critical for representing spatial scales and uncertainty relevant to flash flooding. The resulting ensemble forcing dataset will be used to drive the NextGen NWM to generate probabilistic streamflow forecasts. Model performance will be evaluated across historical flood events, focusing on peak flow magnitude, timing, and probabilistic reliability. This work aims to evaluate a fully integrated, ensemble-based forecasting pipeline that combines downscaled ensemble precipitation forcing with a next-generation hydrologic model. The proposed framework is expected to advance probabilistic flood prediction capabilities for tropical islands and provide a transferable methodology for other complex and data-limited regions where high-resolution ensemble forcing is essential for reliable flood forecasting.