Multivariate assimilation into layered ocean models of the North Atlantic using the SEEK filter

The objective of the DIADEM and TOPAZ projects is the development of a pre-operational forecasting prototype for the North Atlantic Ocean, based on advanced assimilation methods. During DIADEM, a demonstration experiment has already been set up, assimilating satellite data into the Miami isopycnic c...

Full description

Bibliographic Details
Main Authors: Birol, Florence, Brankart, Jean-Michel, Brasseur, Pierre, Verron, Jacques
Other Authors: Laboratoire des Écoulements Géophysiques et Industriels Grenoble (LEGI), Université Joseph Fourier - Grenoble 1 (UJF)-Institut polytechnique de Grenoble - Grenoble Institute of Technology (Grenoble INP )-Centre National de la Recherche Scientifique (CNRS)
Format: Conference Object
Language:English
Published: HAL CCSD 2003
Subjects:
Online Access:https://hal.science/hal-00230238
Description
Summary:The objective of the DIADEM and TOPAZ projects is the development of a pre-operational forecasting prototype for the North Atlantic Ocean, based on advanced assimilation methods. During DIADEM, a demonstration experiment has already been set up, assimilating satellite data into the Miami isopycnic coordinate ocean (MICOM) model. The new system (TOPAZ) is based on the hybrid coordinate, primitive equation ocean model HYCOM developed at RSMAS, on a 1/3º horizontal grid. The vertical discretization is isopycnal in the ocean interior and z-levels in the mixed layer and shallow-water regions. This new coordinate system improves the discretization of the physical variables and a larger part of the observed signal can then be taken into account by the assimilation process. The assimilation scheme is the SEEK filter (Singular Evolutive Extended Kalman Filter), a reduced order approximation of the Kalman filter, in which only the most significant directions of forecast error are controlled by the observations. The SEEK assimilation scheme has been upgrated to also incorporate hydrographic in situ measurements in addition to surface observations. In theory this only requires a new observation operator and three-dimensional, multivariate error covariances. However a keypoint is the choice of a new estimation space ensuring the robustness of the method. In this contribution, we analyse preliminary results obtained with the upgrated system. The attention will be focused on the statistical behavior of the filter and the impact of different surface data sets in the assimilation process.