Accounting for the uncertainty of precipitation forecasts and its impacts on probabilistic flood inundation mapping skill

Uncertainty in operational weather forecasts, particularly in predicting storm location and precipitation patterns, presents challenges for flood inundation mapping. As hurricane forecasts evolve, spatial discrepancies in precipitation estimates lead to misalignment between forecasted and observed rainfall, affecting flood prediction accuracy. This study presents a methodology for addressing storm mispositioning using an integration of the High Resolution Rapid Refresh (HRRR) NWP data and a 2-D hydrodynamic model to generate flood inundation maps. The analysis, focused on Hurricane Beryl (July 2024), evaluates the impact of multiple storm location scenarios over 24-hour forecast periods with 6-hour intervals. Quantitative Precipitation Forecast (QPF) fields are used as input to the Super-Fast INundations of CoasT (SFINCS) model, applied to the highly urbanized central Houston area. Results show that incorporating HRRR forecasts and spatial displacement of QPF fields improves the correlation with in-situ meteorological and water elevation observations. This method provides a more accurate flood inundation mapping by accounting for uncertainties in the precipitation forecasts. SFINCS model performance metrics, including Kling-Gupta Efficiency (KGE), improved from 0.632 using only forecast data to 0.70 when incorporating the WSE ensemble mean generated from modified QPF fields.