Skill of ensemble flood inundation forecasts at short- to medium-range timescales

Abstract

We investigate the ability to improve flood inundation forecasts at short- to medium-range (1–7 days) timescales through weather ensembles and statistical water surface elevation (WSE) postprocessing. To generate the flood inundation forecasts, a one-dimensional hydraulic model, namely the Hydrologic Engineering Center’s River Analysis System (HEC-RAS), is coupled to a regional hydrological ensemble prediction system (RHEPS). The RHEPS is comprised of: i) hydrometeorological observations; ii) weather ensembles from the National Centers for Environmental Prediction Global Ensemble Forecast System Reforecast version 2 (GEFSRv2); and iii) Hydrology Laboratory Research Distributed Hydrologic Model (HL-RDHM) as the hydrological model. The coupled RHEPS-hydraulic system is evaluated along the tidal Delaware River near the City of Philadelphia, Pennsylvania, U.S. For the evaluation, emphasis is placed on the tidal-riverine transitional zone of the Delaware River and the downstream propagation of hydrometeorological uncertainty through the flood inundation forecasts. The coupled system is used to generate hourly flood inundation forecasts at lead times from 1 to 7 days, over the period 2008–2013. Additionally, WSEs from the coupled system are statistically postprocessed using quantile regression (QR). Results show that the raw flood inundation ensemble forecasts exhibit higher skill than the deterministic ones. We also find that statistical postprocessing improves the skill of the raw flood inundation ensemble forecasts, with greater improvements at the longer lead times (>3 days). Overall, we find that both weather ensembles and statistical WSE postprocessing can be used to enhance the skill of flood inundation forecasts at short- to medium-range timescales. In turn, this may serve to enhance the spatial representation of flood forecasts several days in advance, which could contribute in the future to making flood forecast communication and warnings more effective.