Assimilation

We use 4D-var data assimilation in a 1/3° North Atlantic model to describe and understand the circulation and water mass properties variability since the beginning of OVIDE in 2002. The model is OPA 8.2. We assimilate altimetry and temperature and salinity profiles from Argo. The assimilation produces analyses of the hydrology and circulation coherent with the primitive equation model dynamics. Those analyses will allow us to quantify ocean variability from 2002 to present, a time period when there is a large number of in situ observations, thanks to Argo.

Ferron, B., 2011: A 4D-variational approach applied to an eddy-permitting North Atlantic configuration: Synthetic and real data assimilation of altimeter observations. Ocean Modelling, 39, 370-385, doi:10.1016/j.ocemod.2011.06.001

Forget, G., B. Ferron, H. Mercier, 2008: Combining ARGO profiles with a general circulation model in the North Atlantic. Part I: estimation of hydrography and circulation anomalies from synthetic profiles over a year. Ocean Modelling, doi:10.1016/j.ocemod.2007.06.001 or Open Access Version http://archimer.ifremer.fr/doc/2008/publication-3721.pdf.

Forget, G., B. Ferron, H. Mercier, 2008: Combining ARGO profiles with a general circulation model in the North Atlantic. Part II: realistic transports and improved hydrography between spring 2002 and spring 2003. doi:10.1016/j.ocemod.2007.06.002 or Open Access Version http://archimer.ifremer.fr/doc/2008/publication-3920.pdf.

Model set-up:

  • Domain: 20°S - 70°N, 98°W - 14°E.
  •  Horizontal resolution: 1/3° x 1/3° cos(latitude) (280 x 361 grid points). 
  • Vertical resolution: 43 vertical levels. Time step: 2880 s 
  • Biharmonic horizontal diffusivity - TKE in the vertical 
  • Bathymetry from Smith and Sandwell. 
  • Daily atmospheric forcings from NCEP reanalysis. 
  • Observations: altimetry, temperature and salinity profiles from Argo profilers. 
  • Diagonal error covariance matrices for background and observations. 
  • Data assimilation window from 30 to 90 days. 
  • 4D-var assimilation is non incremental.

As shown below, the twin experiments, which have been performed to evaluate the performances of the assimilation system, emphasize the improvement due to in situ data assimilation compared to assimilation of altimetry only. Assimilation of real data will start in 2008.

Figure 1: Results from twin experiments with a 30-day window. Rms differences between the 4D-var estimate and the true ocean state as a function of depth for a) temperature (0 to 1000 m), b) meridional velocity component (surface to bottom). Three experiments were performed: assimilation of Sea Surface Height (SSH) only; assimilation of SSH and T-S profiles with no influence radius (local); assimilation of SSH and T-S profiles with influence radius (extrap.). Distribution of the differences in temperature at 100 m between the 4D-var estimation and the truth for c) SSH only, d) SSH and T-S profiles with influence radius.

Contact: Bruno Ferron