Particle Kalman Filtering for Ocean State Estimation

Abstract

Abstract : The long-term scientific objective is to develop a fully nonlinear Bayesian approach that generalizes the optimality of ensemble Kalman filter methods to nonlinear systems and can be suitable for large dimensional data assimilation problems. The approach will be tested with realistic applications to ocean data assimilation problems. The goal of this research is to explore a new direction that can lead to fully nonlinear filters that can perform better than the ensemble Kalman filter (EnKF) methods with highly nonlinear systems at reasonable computational requirements. We aim at proposing, implementing and testing new nonlinear Kalman filters with ocean data assimilation problems in mind. Simple nonlinear dynamical models will be first considered to evaluate the behavior of these new filters and assess their efficiency compared to the EnKF methods.