Assimilation of temperature data has improved the quality of ocean state estimates used to initialize the coupled model forecasts conducted by the NASA Seasonal-to- Interannual Prediction Project (NSIPP). Nonetheless, deficiencies in the density field are still present. One of the main reasons for such deficiencies is related to the fact that salinity is not analyzed and its effect on the density field is not negligible. In this work we analyze how the ocean initial conditions are affected when not only temperature but also salinity is updated at assimilation times. However, only observed sub-surface temperature is used. Salinity increments are derived from a scheme that uses the analyzed temperature along with the model temperature and salinity to preserve the model's water masses.

Results show that the salinity scheme is effective at maintaining the halocline and thermocline structures especially in tropical regions, structures that would otherwise be corrupted by spuriously enhanced vertical mixing caused by the univariate assimilation of subsurface temperature. The effect of these subsurface modifications on seasonal forecasts will be discussed.