Quantifying Precipitation, Streamflow, and Floodplain Forecasting Skills During Extreme Weather Events in Brays Bayou, Houston, Texas

Abstract

Extreme precipitation and increased urban land cover have increased the frequency and severity of urban flooding events in recent years. Accurate precipitation, streamflow, and floodplain inundation forecasts are necessary to decrease the damage from these events via reservoir operations planning, evacuation of residents, and mobilization of relief efforts. In this study, Quantitative Precipitation Forecasts (QPFs) developed by the National Weather Service (NWS) were analyzed for their skill in predicting precipitation in Brays Bayou in Houston, Texas. This forecasted data were used to force the Distributed Hydrological Soil and Vegetation Model (DHSVM), a physically-based, distributed hydrological model, and the resulted streamflows were assessed for accuracy. Then, a 2-dimensional hydraulic model, Flood2D-GPU, was employed to produce forecasted floodplains, also with skill assessment. This study focuses on three major flood events in the last decade with an emphasis on Hurricane Harvey. Results were focused on three aspects: 1) identifying changes in forecast accuracy with increased lead time; 2) quantifying skill scores of the forecasts through the flood forecasting system; and 3) comparing DHSVM forecasts with those used by the West Gulf River Forecasting Center (WGRFC) to identify optimal forecasting lead time during extreme events.