Application of a Nonlinear and Non-Gaussian Sequential Estimation Method for an Ocean Mixed Laer Model

The strong nonlinearity and non-Gaussian statistics of an ocean mixed layer model, which is based on the second-moment closure of turbulence, render traditional filtering techniques (e.g., Kalman filter) impractical for data assimilation. To overcome this problem, the sampling-importance resampling filter is introduced in this study. This filter represents the required (non-Gaussian) probability density function as a set of samples for implementing recursive Bayesian inference. It is not restricted by the assumption of linearity or Gaussain statistics. The numerical experiments using real life data clearly demonstrate the validity of this filter for the estimation problem of the ocean mixed layer process.