Ensemble Methods And Applications To Altimetry Data Assimilation In The Pacific

Ensemble data assimilation is at the intersection of ensemble forecasting methodologies and relatively independently developed data assimilation based on the theory of statistical estimation. The Ensemble Kalman Filter (EnKF) is a powerful data assimilation method and has proven its efficiency for strongly non-linear dynamical systems but is demanding in computing power. There are two major challenges for applying on a realistic ocean prediction.1. Two main statistical assumptions of the EnKF are the Gaussian distribution of state variables and the non-bias of error statistics (i.e., their mean and covariance are supposed to be exactly known). There are three different types of errors accounted in the sequential data assimilation algorithm: initial errors, model errors and observation errors. The first challenge is initial errors and model errors. The EnKF allows treating various model errors in flexible way. A successful implementation of the EnKF requires a good treatment of the model errors that can well represent the realistic model errors. The current implementation of the EnKF is still in its enfant stage in the term of treatment of the model errors.Ideally, the statistics of the initial ensemble represent the uncertainties of the initial guess for model states, in particular the ensemble standard deviation (or"spread") and the structure of the error covariance. In this paper, a method to initialize an ensemble, introduced by Evensen (1994, 2003), was applied to the Ocean General Circulation Model (OGCM) HYCOM (HYbrid Coordinate Ocean Model) for the Pacific Ocean. Taking advantage of the hybrid coordinates, an initial ensemble is created by first perturbing the layer interfaces and then running the model for a spin-up period of one month forced by randomly perturbed atmospheric forcing fields.In addition to the perturbations of layer interfaces, we implemented perturbations of the mixed layer temperatures. We investigate the quality of the initial ensemble generated by this scheme and the influence of the horizontal decorrelation scale and...