Practical experiences with errorcovariance parameter estimation using real data

Fran Verter and Dick Dee

Data Assimilation Office, NASA/GSFC


In this informal seminar we will describe some of the lessons we havelearned at the DAO from our experiences with the estimation of error covariance parameters based on observed-minus-forecast residuals obtained in data assimilation. We will briefly review the methodology, based on maximum-likelihood parameter estimation, and discuss some of the problems we have encountered. We then describe preliminary experiences with the Bayesian technique proposed by Purser and Parrish at NCEP.