Improving National Water Model ensemble streamflow predictions using post-processed precipitation forecasts

Accurate ensemble streamflow forecasts are important for providing impact-based decision support services to water resource managers and emergency personnel, especially for characterizing the uncertainty in possible flood impacts leading up to an extreme-precipitation event. In this presentation, results will be presented that demonstrate the improvements that can be made to the NOAA National Water Model streamflow forecasts when forced with post-processed NOAA Global Ensemble Forecast System version 11 (GEFSv11) precipitation forecasts. The post-processing of the precipitation forecasts was done based on the censored, shifted, gamma distribution and then reordering using an ensemble copula coupling technique. The forecast analysis period will focus on the 2016-2017 wet season (October-March) over the state of California when 45 atmospheric rivers made landfall on the United States west coast. An 11-member ensemble of streamflow forecasts forced with the precipitation forecasts from the raw and post-processed GEFSv11 will be compared over several rain-fed basins in northern California. Preliminary results show that forcing the NWM with the post-processed precipitation results in larger ensemble spread and improved representation of the streamflow relative to analysis streamflow values.