Flood early warning system with data assimilation enables site-level forecasting of bridge impacts

Abstract

Vehicle-related flood fatalities account for a majority of flooding deaths in the United States. As floods become more frequent and severe, emergency operators need accurate early warning systems to enact road closures and dispatch first responders. We present an operational flood forecasting framework that connects large-scale hydrologic predictions with site-level transportation impacts. This system integrates NOAA’s National Water Model (NWM) with a new data assimilation framework based on Kalman Filtering to generate improved discharge and stage predictions at progressive 12-hour forecast horizons. These discharge and stage forecasts are joined with a large-scale bridge infrastructure database to generate site-level probabilistic flood warnings. Tested across two major river basins in Texas, our data assimilation and forecasting framework outperforms the NWM’s existing nudging method at predicting bridge flooding impacts over all lead times considered. By enabling accurate site-level bridge warnings, the proposed framework will enable more targeted management of transportation systems during floods.