The Maximum Likelihood Ensemble Filter development at the Colorado State University

Milija Zupanski

Colorado State University 

The Maximum Likelihood Ensemble Filter (MLEF) is an ensemble filter which combines ensemble filtering and control theory. It calculates the maximum likelihood analysis solution (as in the variational methods), rather than the minimum variance estimate (as in the Kalman filtering and the ensemble Kalman filtering). The analysis error covariance, which is used to define the initial perturbations for the ensuing ensemble forecast, is obtained as a minimization by-product, employing a relation between the analysis error covariance and the inverse Hessian matrix.
The new method is being developed with NOAA THORPEX support and will be tested in the NCEP GFS system. Preliminary results of the MLEF will be presented, in various applications ranging from the one-dimensional models to the primitive equation models. Implications of a double-resolution MLEF to a potential operational use will also be discussed.