Real-time Flash Flood Modeling over the Conterminous US

Flash flooding differs significantly from other natural hazards. Flash floods occur quite frequently during the warm season, they are more hazardous than tornadoes or lightning, and their impacts are strongly controlled by terrestrial and anthropogenic features. They are not only difficult to predict, they are challenging just to observe. This study presents a new paradigm in flash flood prediction called the Flooded Locations And Simulated Hydrographs (FLASH) project. FLASH utilizes 1 km/5-min rain rates from NSSL's NMQ/Q2 precipitation system input to an uncalibrated distributed hydrologic modeling framework. Forecast hydrographs are produced in real-time at each grid point in the conterminous US that has decent radar coverage. The forecast flows are then compared to their historical distributions, which were created using a hydrologic model reanalysis forced by archived NEXRAD data. This approach differs from the traditional method of calibrating model parameters. Instead, model parameters are based on measurable, geophysical properties of the Earth. Then, simulated flood frequencies are linked to observed flash flooding impacts using an a-priori database. This enables flooding impacts and magnitudes to be forecast with 6 hours of lead time, even at ungauged locations.

The study will present FLASH results that were produced in real-time during the deadly 2013 Oklahoma City flash flooding event, for a multi-year dataset in relation to the skill from operational tools used in the National Weather Service, and during the Flash Flood and Intense Rainfall (FFaIR) experiment that took place during July 2013 at the NCEP Weather Prediction Center. Future plans on system development will be presented, including the utilization of spaceborne remote-sensing data to improve rainfall estimation in areas of poor NEXRAD coverage and detection and assimilation of inundated regions.