You must have a javascript-enabled browser and javacript and stylesheets must be enabled to use some of the functions on this site.
 
   

 

Recent advances in data assimilation in the MERSEA project

Sergey Skachko(1) , Léo Berline(1) , Laurent Bertino(2) , Jean-Michel Brankart(1) , Pierre Brasseur(1) , Yann Ourmières(1) , Jens Schroter(3) , Peter Jan van Leeuwen(4) , and Jacques Verron(1)

(1) LEGI, CNRS, BP 53X, 38041 Grenoble, France
(2) NERSC, Thormohlensgate 47, 5006 Bergen, Norway
(3) AWI, Postfach 120161, 27515 Bremerhaven, Germany
(4) IMAU, Utrecht University , 3508 TA Utrecht, Netherlands

Abstract

The MERSEA European project aims at developing a European system for operational monitoring and forecasting on global and regional scales of the ocean physics, bio-geochemistry and ecosystems. The purpose of this paper is to review the recent advances of data assimilation in the MERSEA project. Fundamental studies are lead with the Ensemble Kalman Filter (EnKF) and the Sequential Importance Resampling (SIR) filter to deal with biases and non-liearities of the ocean model, while more applied studies are lead with the Singular Evolutive Extended Kalman (SEEK) filter to control the mixed layer, and to explore the impact of the control on an ecosystem model.

First of all, the EnKF developments are aimed at conditions where bias and non-linearities are not negligible. The approach is to modify statistical estimators based on geostatistical methods (Gaussian anamorphosis, bias estimation) with applications to coupled ice-ocean systems and ecosystems.

Another fundamental study similar to work with the EnKF filter is carried out with the SIR filter. In contrast with the EnKF, the SIR filter updates probabilities of the ensemble members and not the ensemble states themselves. This di erence makes the SIR filter a truly variance minimizing scheme, which can be easily applied to strongly non-Gaussian ecosystem models that MERSEA needs to investigate. Several tests on such a nonlinear model have been performed, estimating model state, model parameters and even model noise strength. We found that by making the strength of the model noise variable in time the model parameters became much less time dependent, which leads to a much more realistic ecosystem model.

The SEEK filter development consists of two parts. The first one is to explore the problem of estimation of turbulent momentum, heat and fresh water fluxes, one of the main sources of ocean model errors that strongly penalize the operational capacity to provide realistic forecasts of the thermohaline characteristics of the mixed layer and of the surface ocean currents. The idea is to augment the control space of the filter to include, in addition to the state variables, information about the air-sea fluxes. For this purpose the turbulent coe cients CE and CH of latent and sensible heat flux were chosen among parameters which contribute mainly to the errors of heat and fresh water fluxes. The possibility to control errors from atmospheric forcing by assimilation is shown.

Another part of SEEK filter development work is to quantitatively improve the representation and space-time variability of marine ecosystems in ocean basins, by coupling with multi-data assimilative eddy-resolving circulation models, focusing on key coupling mechanisms (mixed layed dynamics, diapycnal mixing, eddy activity, and vertical advection). To serve this purpose, a prototype of a coupled physical-biological assimilation system is developed. The data assimilation system mainly uses satelite derived products such as Sea Surface Temperature and Sea Surface Height. The impact of the assimilation of multivariate data sets on the coupling and ecosystem response is studied.

A synthesis of the results obtained during the first phase of the MERSEA project will be presented at the Symposium.

 

 

                 Last modified: 07.10.03