High-resolution modeling of Hurricane Harvey Flooding for Harris County, TX using a calibrated GPU-accelerated 2D Flood Model

With advancement in computational power, reconstruction of historic flood events through high-resolution hydrodynamic model has become more affordable in the recent decade. Hurricane Harvey that made landfall in the southern Texas was one of the most destructive hurricanes in U.S. history. The downpour from Hurricane Harvey has caused significant flooding resulting in about 77 casualties, displacing more than 30,000 people, and causing more than $70 billion in direct damage. Harris County, which covers over two HUC-8 drainage basins ( 2702 mi²), has experienced more than 80% of its annual average rainfall during this event. The objective of this case study is to reconstruct flooding caused by Hurricane Harvey in Harris County, Texas using a calibrated model. The flood simulation was performed at a 30-m spatial resolution using a graphic processing unit (GPU) accelerated 2D flood model (Flood2D-GPU) at through high performance computing (HPC). The hourly radar-based rainfall estimates from the National Center for Environmental Prediction Stage IV Quantitative Precipitation Estimate (Stage IV QPE) were fed into a calibrated Variable Infiltration Capacity (VIC) hydrologic model to generate runoff estimates which were then routed through the Routing Application for Parallel computation of Discharge (RAPID) to obtain streamflow hydrographs at 69 locations for Flood2D-GPU simulation. Using 4 GPU compute nodes, a total of 26 machine hours was consumed to simulate a complete 5-day flood hydrograph. Preliminary results indicate that the maximum flood extent from Hurricane Harvey has resulted 453mi² of flooded area just in Harris County alone, which is 60 mi² more than the FEMA derived floodplain. A calibrated model run will help in revising these estimates. This study will include model calibration and validation for this event, and ultimately demonstration of the capability of the modeling framework.