A Flood Inundation Prediction for Hurricane Harvey Using a Hyperresolution Hydrologic & Hydraulic Model Driven by the Remote Sensing Observations and Precipitation Forecasting Products

Abstract:

As climate change will increase the frequency, slow down the speed and intensify the precipitation of tropical cyclones from the Atlantic Ocean, there is a clear need for a system that provides comprehensive information about each flood event for risk minimization. Flood modeling has been improved dramatically in recent years due to the improvement of computational capability and the merging of machine learning. This study provides a hydrologic and hydraulic coupled modeling system, that is based on Coupled Routing and Excessive Storage (CREST) model and the Australia National University- Geophysics Australia (ANUGA) model, to simulate and predict the flood inundation during Hurricane Harvey. The near-real-time MRMS precipitation observation, the short-range precipitation forecast from HRRR and machine-learning-based nowcasting were executed to simulate and forecast the streamflow, the flood extent, and the inundation depth over the study region. The results were compared with post-event survey data by the United States Geological Survey (USGS) and the Federal Emergency Management Agency (FEMA) flood insurance claims. The near-real-time simulation indicated that the near-real-time simulation could capture 87% of all flood insurance claims within the study area and the overall error of water depth was 0.38 meters. The forecasting products showed comparable results. This study proves that the hydrologic and hydraulic approach has the potential to operationally provide three dimensional (streamflow, flood extent, inundation depth) information for decision-makers to reduce the flood damage and risk.