Quantifying and Visualizing Uncertainty in Flood Inundation Forecasts

The ability to model flood inundation on a national scale using the National Water Model (NWM) will put invaluable information into the hands of decision makers. The communication of forecasted flood inundation typically uses deterministic model output to provide a visual representation of a single flood scenario; however, uncertainty exists in the atmospheric model precipitation estimates used to drive the hydrologic model. Thus, the flood inundation products derived from deterministic model output are lacking the communication of uncertainty, and therefore lacking a message of risk. To address this gap in the communication of uncertainty, this study will use precipitation estimates from the Global Ensemble Forecasting System (GEFS) to drive a hydrologic modeling framework similar to the NWM. Combining the water depth output from each ensemble and a Height Above Nearest Drainage (HAND) raster produces the associated flood extent for each ensemble. Quantifying the agreement between the flood extent ensemble rasters provides a measure of uncertainty and communicates the risk of flooding for each forecast grid. The results are incorporated into a web-based application which allows for real-time user interaction of the forecasted flood inundation scenario. The products produced from this study demonstrate how atmospheric model ensembles can be incorporated into the NWM to quantify and visualize uncertainty in flood inundation forecasts.