Streamflow and precipitation data assimilation into the National Water Model : an investigative study for statewide flood forecasting

Flooding is a dangerous natural disaster that poses risk to property and personal safety world-wide. The state of Texas is especially vulnerable to floods; a region known as “Flash flood Alley” runs through central Texas, and the coastal region is prone to severe tropical storms and hurricanes. Flood warning systems exist in large cities throughout Texas, but there is no coordinated state-wide flood warning network. In this study, we investigate the feasibility of creating a locally intelligent, state-wide flood forecasting system by using local observational streamflow data to better inform a national forecasting model. A method is developed to integrate streamflow sensors and precipitation products into short-range, 18-hour National Water Model (NWM) forecasts through data assimilation (DA). Four-dimensional variational data assimilation is coupled with a mass-conservative Muskingum-Cunge flow routing scheme to propagate streamflow corrections upstream and downstream from the sensor locations. The model is applied to a rural study area in the Texas Hill Country, and the streamflow DA model creates improved forecasts products at the sensor location and over 15 miles upstream and downstream of the sensor. Proof of concept results from a simplified surface runoff model using precipitation corrections indicate improved streamflow profiles at the upstream location. With further validation and development, there is real potential in assimilating local data into the NWM to create a statewide flood forecasting network.