Since 2001, CLS has been leading studies to develop a model based on satellite and in situ observations, (called MOSIS). The first model, set up in 2001 during ATOS project, covers the North-West of the Indian Ocean, and the model was refined during AGANPANTHE study. Finally, the same kind of model was developed for the Mediterranean sea in MOSIS study. These models share a common operating principle, based initially on the vertical projection of surface information provided by satellite observations. These projected observations are then combined with in situ observations.

The aim is to estimate the ocean's thermohaline state from both kinds of observations. Prediction models can thus be compared with the ocean field deduced from these observations and subsequently validated. MOSIS also offers an alternative if the forecasting system fails. The work achieved over the past years (2001-2003) has led to develop an efficient system.

CLS has been responsible for:

• developing a MOSIS model for the North-West Indian Ocean.
  A first model was set up in 2001, in the frame of ATOS study, for the North-West Indian Ocean, mainly represented by the Arabian Sea. The model is based on reduced-order optimal analysis using SOFA assimilation software. Only altimetry data were used.

• improving the model's vertical projection of satellite observations.
  With AGAPANTHE project, led in 2002, the model has been refined. The improvement was based on multiple linear regression combining altimetry data and sea surface temperature data.

• developing a model for the Mediterranean Sea.
  The system is being upgraded in MOSIS project in 2003 and was now based on a vertical projection of satellite observations (using multiple linear regression) and a combination of these projected observations (synthetic observations) with in situ observations (using multivariate optimal interpolation).

Users: