THETIS-2 STUDY
The THETIS-2 study (Menemenlis et al. 1997) was carried out using data provided by Uwe Send from the University of Kiel. The data consisted of depth-averaged temperature change along three sections, climatological temperature and salinity, and meteorological surface forcing fields. The model used was a primitive equation model developed by Marshall and others at the Massachusetts Institute of Technology(Marshall et al, 1997a, 1997b). It was integrated in the Mediterranean basin at 1/4 degree grid-spacing in the horizontal and 20 vertical levels. Surface boundary conditions were from ECMWF meteorological analyses.
Altimetric and tomographic data were compared under the assumption that changes in sea level measured by the altimeter were caused by uniform heating or cooling of the top 50m of water. Differences between the tomographic and altimetric heat content estimates result from physical processes other than heating or cooling. In particular, changes of mass due to barotropic transports through the Straits of Gibraltar and Sicily have significant contributions to sea level anomaly. Therefore a key contribution of the tomographic data is the ability to isolate the heat content part of the sea level anomaly signal.
Estimates of the near surface circulation were obtained which exhibited several known features of the circulation, for example, the cold waters in the Gulf of Lyon in the Spring, the gradual warming of the basin towards Fall, and the intrusion of warm salty Levantine intermediate water from the Strait of Sicily. The estimates were obtained by essentially solving a large least squares problem (Menemenlis and Wunsch, 1997)whereby a quadratic function of data and model errors is minimized after weighting each error term by the inverse of its respective uncertainty. There are several known methods for solving this problem. Here we used the so-called reduced state Kalman filter and smoother approach. When the statistical models and assumptions made are correct, data assimilation provides estimates of the circulation which are more complete than those from data alone and closer to reality than estimates from the numerical model.
In summary, the key contributions of data and model were the following : hydrographic data provided the time-mean baroclinic circulation, acoustic tomography the time-evolving temperature, altimetry provided horizontal resolution, and the numericai model determined the effect of observed winds and was used to interpolate the data to regions without measurements.
