Abstract. Temperature and salinity sampling strategies are studied and compared by means of the Observing System Simulation Experiment technique in order to assess their usefulness for data assimilation in the framework of the Mediterranean Forecasting System. Their impact in a Mediterranean General Circulation Model is quantified in numerical twin experiments via bivariate data assimilation of temperature and salinity profiles in summer and winter conditions, using the optimal interpolation algorithm implemented in the System for Ocean Forecasting and Analysis. The data impact is quantified by the error reduction in the assimilation run relative to the free run.

The sampling strategies studied here include various combinations of temperature and salinity profiles collected along Volunteer Observing Ship (VOS) tracks, by Mediterranean Multi-sensor Moored Arrays (M3A), a Glider and ARGO floating profilers. Idealized sampling strategies involving VOS data allow to recognize the impact of individual tracks. As a result, the most effective tracks are those crossing regions characterized by high mesoscale variability and the presence of frontal structures between water masses.

Sampling strategies adopted in summer–autumn 2004 and winter 2005 are studied to assess the impact of VOS and ARGO data in real conditions. The combination of all available data allows to achieve up to 30% error reductions. ARGO data produce a small impact when alone, but represent the only continuous coverage of the basin and are useful as a complement to VOS data sets.

Localized data sets, as those obtained by M3As and the Glider, seem to have an almost negligible impact in the basin-scale assessment, and are expected to be more effective at regional scale.