Development of a Flexible, Multi-Model, Real Time, Compound Flood Forecasting System for Tropical and Non-tropical Events

The Floodwater Forecast System has been developed as a next generation coastal and compound flood forecast system to help inform decision makers during both tropical and non-tropical events with both deterministic and ensemble-based analysis. Floodwater utilizes the MetGet service as a meteorological engine to provide forecast meteorology to hydrodynamic models. The available types of meteorology available for use within Floodwater are the Global Forecast System (GFS), Global Ensemble Forecast System (GEFS), National Hurricane Center (NHC) forecast guidance, Weather Prediction Center Quantitative Precipitation Forecasts (WPC-QPF), Coupled Ocean/Atmosphere Mesoscale Prediction System for Tropical Cyclones (COAMPS-TC) and its ensemble component (CTCX), Hurricane Weather Research Forecast (HWRF), and High-Resolution Rapid Refresh (HRRR). Both the Floodwater Forecast System and MetGet are or are scheduled to become released under the MIT license publicly on Github.

The Floodwater system has been developed to be flexible and agnostic to the hydrodynamic models it uses and run in high performance computing (HPC) environments. It uses the ecFlow framework, developed by European Centre for Medium-Range Weather Forecasts (ECMWF), along with a python scripting to generate workflows which run unattended and in real time. At present, the Floodwater system supports ADCIRC, SWAN, HEC-RAS, and XBeach based simulations, including loose coupling between the models through input files. It is expected that additional models will be added in the future to support different use cases. The forecast system has been deployed on local systems, NSF funded supercomputing sites, and within the AWS cloud on both x86 and ARM processors.

The Floodwater system operates by interacting with MetGet through a client application which allows queries for data that has been made available by NOAA or other data providers and request that it be turned into model-specific input files. The system also prepares boundary conditions for the models based on observed and predicted riverine flows, which can come from the USGS or the River Forecast Center (RFC). Finally, when simulations are complete, the data is pushed to decision support dashboards for use by emergency managers, including dashboards which detail roadway inundation, exposure of critical infrastructure, and information necessary for search and rescue operations.

The Floodwater system was first deployed in 2022 and saw its first successful emergency usage during Hurricane Ian, where it oversaw ADCIRC simulations on 4 different model geometries using 4 different meteorological sources. Most recently, the system has so far been utilized during Hurricane Idalia to provide information both to emergency managers and for research purposes also on 4 different model geometries and with 4 types of meteorological forcing. Figure 1 shows two forecasts conducted prior to landfall using COAMPS-TC meteorological forcing. The left pane shows the results for simulations completed 32 hours prior to landfall and the right pane shows the forecast 8 hours prior to landfall. Both shows good agreement with the observation across the landfall region and can give decision makers critical information in advance of landfall.