Combined Data Assimilation and Model Calibration in Water Resources Systems

The key issues in operational forecast system at the National Weather Service River Forecast System (NWSRFC) in US are the reliable quantification of forecast uncertainty which is dependent upon accurate parameter estimation (model calibration), state initializations and incorporation of historical climatological variability through Ensemble Streamflow Prediction (ESP) system. Ensemble-based data assimilation methods are becoming popular in water resources modeling largely because of their flexibility, capability and effectiveness. Emerging technologies in Bayesian estimation within the sequential Monte Carlo framework provides a platform for improved estimation of hydrologic model components and forecast uncertainty. In this presentation, the major effort goes into introducing the sequential Bayesian information fusion technique as an alternative approach to batch calibration to characterize and reduce the uncertainties associated with hydrologic model parameters and model state initialization while accounting for forcing data and observation uncertainties. For this purpose, two sequential ensemble paradigms are introduced and discussed, mainly Dual Ensemble Kalman Filter (DEnKF) and Dual Particle Filter (DPF). The applications of methods over an operational forecasting system are demonstrated while their efficiency and effectiveness in state-parameter estimation, ensemble streamflow forecasting and associated uncertainties are investigated.