Almost five of TOPEX/POSEIDON (T/P) and ERS-1/2 altimeter data were analysed in the Mediterranean Sea as part of the European MAST MATER project. The project aims to improve our knowledge of the circulation and marine ecosystems in the Mediterranean. Our contribution focuses on analyzing T/P and ERS-1/2 data. The main results (response of the mean sea level to atmospheric pressure , analysis of mean sea level variations , ocean circulation and combination of T/P and ERS-1/2) are presented below.

Response of the mean sea level to atmospheric pressure

In the open ocean, the response of sea level to atmospheric pressure is close to the well-known inverse barometer (IB) effect (e.g. Gaspar and Ponte, 1997). The IB correction is simply given by - 1/rg (P-Pref) where Pref is the mean pressure over the ocean which, in practice, is often assumed to be constant (r is density, g is gravity). A 1-mbar increase in atmospheric pressure thus induces a sea-level depression of about 1 cm. In a closed basin, because mass must be conserved, Pref must be made equal to the mean pressure over the basin; the mean sea level does not thus respond to the mean pressure and only adjustments of sea level to pressure anomalies (e.g. due to spatial variations of pressure) are possible. In a semi-enclosed sea such as the Mediterranean, the response is more complex. The sea level can respond to the mean pressure over the basin if there is an adjustment through the straits : the presence of a mean anticyclonic pattern will only induce a fall in mean level if the water can "escape" via the Straits. In practice, at high frequencies the adjustment is limited by the Straits narrowness and friction. The high-frequency response will therefore be noticeably different from a simple inverse barometer effect (it would be the same for the lowest frequencies , particularly at the seasonal frequency discussed below).

Le Traon and Gauzelin (1997) used T/P data to analyze the response of the Mediterranean mean sea level to atmospheric pressure. Coherence analysis between mean sea level and atmospheric pressure showed a significant departure from a standard Inverse Barometer (IB) effect at frequencies higher than 20 days-1. At high frequencies, the phase difference between mean sea level and pressure is about 100° while it should be 180° for a perfect IB response. This result confirmed the role of the straits of Gibraltar in limiting the water exchange at high frequencies. The analytical model proposed by Candela (1991), which includes friction in the Straits of Gibraltar and Sicily, was also successfully applied to T/P data. The model explained a large part of the variance in T/P mean sea level variations. Compared to an inverse barometer correct ion, it gave a smoother response with a phase delay at high frequencies. It also explained more variance in T/P mean sea level variations. This demonstrated that this simple model provides an improved correction of atmospheric pressure effects in T/P data. Future works will deal with the implementation of a realistic barotropic model in the Mediterranean sea to better estimate the response of sea level to atmospheric pressure forcing.

Mean sea levels variations

Satellite altimetry and, in particular, T/P is, for the first time, providing a way of characterizing changes in the mean level of the Mediterranean (Larnicol et al., 1995). This is illustrated on Figure 1 which shows the Mediterranean mean sea level variations calculated from almost five years of TOPEX/POSEIDON data (October 1992 to September 1997, i.e. from cycles 2 to 181 - T/P cycle duration is 10 days). There is strong seasonal variation, on the order of 20 cm, peaking in October-November, falling quickly in January/February, and reaching a minimum in March/April. There is also interannual variability. Note, in particular, the mean sea level rise during the fourth year. Roughly half the seasonal mean sea level variations is related to the steric effect of contraction/dilation of the sea surface driven by heat exchanges at the ocean-atmosphere interface. The other half is probably due to a slight seasonal imbalance between the inflow and outflow at Gibraltar, and to evaporation and precipitation. This suggests that mass in the Mediterranean is not conserved at seasonal scale. The exact mechanism is not clear. Larnicol et al. (1995) suggests that it may be related to a baroclinic response to E-P and to the Gibraltar Straits dynamics (hydraulic control and switching from maximal to submaximal exchange ) (Garrett et al., 1990; Bryden and Kinder, 1991).

Figure 1

A mean sea level drift of about 1 cm/year can also be derived from Figure 1 results. However, given the large seasonal and interannual variability, it cannot be considered as a long term drift and a much longer time series is needed. The difference in mean sea level between the Atlantic and the Mediterranean and the western and eastern Mediterranean were also analyzed. They can provide an indirect estimation of the flux of water through the Gibraltar and Sicily straits.

Sea level variability and oceanic circulation

Merging T/P and ERS-1/2

Figure 1 Figure 2

Merging multiple altimeter data sets is not an easy task. It requires homogenous and inter-calibrated data sets. Homogeneous means that the same altimetric corrections should be used for the different satellites. To obtain inter-calibrated data sets, we used T/P as a reference for the less precise ERS-1/2 missions (Le Traon and Ogor, 1997 ). We then extract Sea Level Anomaly relative to a 4-year mean (October 1992 - October 1996) using a conventional repeat-track analysis. A specific processing is performed to have consistent mean between T/P and ERS-1/2. This yields consistent along-track Sea Level Anomaly data sets. To map sea level anomalies, we used an improved space-time objective analysis method which takes residual long wavelength errors (e.g. due to tides, inverse barometer, residual orbit errors) into account (Le Traon et al., 1997). This methodology was shown to provide a much better mapping of sea level variations and ocean circulation in the Mediterranean sea (Ayoub et al., 1997; Ayoub, 1997 ). Sea Level Anomaly maps were systematically calculated from T/P and ERS-1/2 separately and T/P and ERS-1/2 combined. Figure 1 shows the rms difference between T/P and T/P and ERS-1/2 combined maps. Between T/P tracks, the differences can be as large as the signal itself (Figure 2). They reveal the sampling differences between the satellites. This shows that merging the data is vital for monitoring mesoscale signals in the Mediterranean. The merging of T/P and ERS-1/2 allows, for example, the detection of Algerian eddies, their detachment from meanders and their propagation. These features were not resolved by T/P only. A much more coherent picture of the evolution of the eddy field is thus obtained.

---

Seasonal and interannual variations of the circulation

Figure 1 Figure 2

Figure 1 represents the seasonal variations in sea level anomaly (anomaly of sea level relative to a 4-year mean) observed by the combination of T/P and ERS-1 in winter 1993 and fall 1993. The map is an average over a three-month period. This figure and the figures for the other 3-month periods show the main characteristics of the Mediterranean circulation (Alboran gyres; Algerian eddies; Ionian, Ierepetra, Mersa-Matruh and Shikmona gyres, etc.) and the seasonal variations in these features (strengthening of cyclonic circulation in winter, strengthening of anticyclonic Alboran and Ierepetra gyres in summer and fall). T/P and ERS-1/2 reveal some of the particularly strong signals very well : Alboran gyres east of Gibraltar and Ierepetra gyre south-east of Crete. The strong seasonal signal from the Ierepetra gyre is probably linked to direct forcing by strong Etesian winds, which interact with the Cretan topography. There is also a large interannual variability, in particular in the Levantine basin. This is exemplified in Figure 2 which shows the combination of TOPEX/POSEIDON and ERS-1/2 in summer 1993 and summer 1996. The interannual variability in the Levantine might be related to a change in the Etesian winds and the switching from a state with well developed Ierepetra gyre to a state with a large anticyclonic system in the central Levantine basin (with the development of the Mersa-Matruh and Shikmona anticyclonic gyres) (Ayoub, 1997). These results agree well with what we know about circulation in the Mediterranean. They provide, however, for the first time, a global view of the seasonal and interannual variations of the circulation in the Mediterranean sea.

T/P and ERS-1/2 data are now being used to validate Mediterranean circulation models. Comparison with high resolution primitive equation models will also be used to quantify the main forcing mechanisms of the circulation...