Improving Hydrologic Forecasting Using Ensemble Conditional Bias-Penalized Kalman Filter

We describe an application of ensemble conditional bias-penalized Kalman filter (CEnKF) to ensemble streamflow forecasting. CEnKF is an ensemble extension of conditional bias (CB)-penalized Kalman filter (CBPKF) which extends Kalman filter to account for Type-II CB in state estimation explicitly. Type-I and –II conditional CBs arise when falsely detecting an effect which does not exist and when failing to detect an existing effect, respectively. Whereas Type-I CB can be reduced via calibration, Type-II CB cannot. Therefore, reducing Type-II CB in ensemble data assimilation addresses an important gap in ensemble forecasting particularly of extremes.The premise of this work is that improved estimation of soil moisture states in the extremes translates into improved streamflow prediction particularly in flooding events. We apply CEnKF to the lumped Sacramento soil moisture accounting and unit hydrograph models of the NWS for assimilation of hourly streamflow observations for multiple headwater basins in Texas for a 10-yr period. In this presentation, we summarize the results and share issues and challenges.