Accuracy Comparison of State Estimation and Parameter Identification by Particle and Ensemble Kalman Filters for Nonlinear Observation System

• Activity Number: 271
• Type: Contributed
• Date/Time: Tuesday, August 9, 2005 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Kazuyuki Nakamura*+ and Tomoyuki Higuchi
• Companies: The Institute of Statistical Mathematics and Institute of Statistical Mathematics
• Keywords: Particle Filter ; Ensemble Kalman Filter ; Nonlinear observation ; Data assimilation

Abstract

Data assimilation technique, which is developed in meteorology and oceanography, aims at accommodating states of a physical simulation model to observations. It is motivated to provide the good initial condition for the nonlinear simulation model and to realize the online model parameter estimation. To achieve these purposes, sequential data assimilation such as the Ensemble Kalman Filter (EnKF) is used. The EnKF is based on the second-order statistics and cannot deal with the systems directly if observed data are given through nonlinear transformation of states. On the other hand, it is well known that the Particle Filter (PF) can deal with higher-order statistics and nonlinear transformed states. Both are the ensemble-based filtering methods and can be extended to the fixed lag smoothers (the Ensemble Kaman Smoother [EnKS] and the Particle Smoother [PS]). This research demonstrates that the PF and PS are superior to the EnKF and the EnKS in assimilating nonlinear observations by numerical experiments.