Troccoli et al. 2002

Salinity adjustments in the presence of temperature data assimilation

In this paper we evaluate the role of salinity in the framework of temperature data assimilation in a global ocean model that is used to initialise seasonal climate forecasts. It is shown that the univariate assimilation of temperature profiles, without attempting to correct salinity, can induce first order errors in the subsurface temperature and salinity fields. A recently developed scheme by Troccoli and Haines (1999) is used to improve the salinity field. In this scheme, salinity increments are derived from the observed temperature, by using the model temperature and salinity profiles, assuming that the T-S relationship in the model profiles is preserved. In addition the temperature and salinity fields are matched below the observed temperature profile by vertically displacing the original model profiles.
Two data assimilation experiments were performed for the 6-year period 1993-1998. These show that the salinity scheme is effective at maintaining the haline and thermal structures at and below thermocline level, especially in tropical regions, by avoiding spurious convection. In addition to improvements in the mean state, the scheme allows more temporal variability than simply controlling the salinity field by relaxation to climatological data. Some comparisons with sparse salinity observations are also made which suggest that the subsurface salinity variability in the western Pacific is better reproduced in the experiment in which the salinity scheme is used. The salinity analyses might be improved further by use of altimeter sea level or sea surface salinity observations from satellite.