This research talk focuses on efficient uncertainty quantification (UQ) and uncertainty reduction (UR) techniques. In the UQ part, the intrusive (Galerkin) polynomial chaos (PC) method was modified to improve the efficiency of UQ applied to stiff chemical models. The non-intrusive collocation least-squares method was proposed to simplify the UQ process. The numerical tests were carried out on the Sulfur Transport Eulerian Model (STEM) to enhance the regional ozone prediction.
In the UR part, the focus is on theoretical exploration of 4D-Var and Ensemble Kalman Filter (EnKF) method hybridization. A hybrid ensemble method was used to improve and update the background error covariance matrix estimation between the 4D-Var assimilation windows. The theoretical equivalency between the 4D-Var and the EnKF was investigated. A novel hybridization scheme is proposed. Numerical tests on a nonlinear model indicate the improvement over regular EnKF results.