Comparative evaluation of ensemble Kalman filter, particle filter and variational techniques for river discharge forecast

Hydrologic data assimilation techniques provide a means to improve river discharge forecasts through updating hydrologic model states and correcting the atmospheric forcing data via optimally combining model outputs with observations. The performance of the assimilation procedure, however, depends on the data assimilation techniques used and the amount of uncertainty in the data sets. To investigate the effects of these, we comparatively evaluate three data assimilation techniques, including ensemble Kalman filter (EnKF), particle filter (PF) and variational (VAR) technique, which assimilate discharge and synthetic soil moisture data at various uncertainty levels into the Sacramento Soil Moisture accounting (SAC-SMA) model used by the National Weather Service (NWS) for river forecasting in The United States. The study basin is Greens Bayou watershed with area of 178 km2 in eastern Texas. In the presentation, we summarize the results of the comparisons, and discuss the challenges of applying each technique for hydrologic applications.