due to the inconsistency in the mean field and the time-varying part of the altimetric observation which increases the error in the analysis field (ha in Eq. (14)). The mean interface depth field at the scale of the gyre circulation cannot be corrected in this case and this result is consistent with Marshall (1985) that assimilation of the altimetric data alone cannot separate the error in the model and in the mean field.
The generation of gravity waves shown in Exp. 1 is reduced by the correction of the mean interface depth. The time evolution of the velocity error at 142.5｡?, 35｡? in Exp. 2-2 is shown in Fig. 15. It appears that the gravity waves are reduced especially in the latter half of the experiment compared to those in Exp. 1-2 (Fig. 13). This result shows that the inconsistency between the observed time-varying part and the estimated mean field is reduced through the experiment. Some effects of gravity waves remain because the assumptions in the error covariance matrices (namely, a Gaussian distribution and the geostrophic balance) are insufficient and introduce the noise to the analysis field especially in the strong current region.
6 Summary
Our study has successfully obtained the global surface current (and its kinetic energy distribution, though not shown here) with (2° × 2°) grids (with (1° × 1°) grids in specified regions) using the archived drifting buoy data from the MEDS and the JODC. The number of the data is rather great enough to make the finer map than in the previous studies (e.g., Wyrtki et al., 1976). Since the distribution of the mean and the eddy kinetic energy is significantly affected by spatial resolution (grid size), our finer maps provide more realistic features of the global surface circulation. For example, the jet-like structures of the western boundary and equatorial currents, which cannot be resolved with the coarse grid in the previous studies, are well reproduced.
The features of seasonal current field have also been examined. The result revealed that outstanding seasonal variabilities such as flow reversal appear especially in the Indian ocean and the tropical Pacific. The circulation in the western tropical Pacific shows quite complex seasonal variabilities. The seasonal variability of the western boundary current of the subtropical gyre such as the Kuroshio and the Gulf stream is not active so much whereas the western boundary current of the subpoler gyre such as the Oyashio and the Labrador Current is suggested to show a significant seasonal variability.
The successive correction model of the mean SSH field is confirmed to work well with the simultaneous assimilation of drifting buoy and altimetric data. In particular, the velocities derived from the drifting buoys play an essential role in correcting the mean SSH field. As pointed out by Marshall (1985), the altimetric data assimilation alone cannot correct the mean SSH field. This problem is avoided in this study by the simultaneous assimilation of the drifting buoy data that is independent of the altimetric data. In addition, the correction of the mean SSH minimizes the inconsistency between the time-varying component derived from the altimetric data and the mean SSH field derived from the model average or climatological data. It also leads to the reduction of the gravity wave generation which tends to contaminate model results. Although the mean SSH field is corrected by the simultaneous assimilation of the drifting buoy data and the altimetric data, our present model is very simple and should be improved in the future.
Acknowledgment. We express our hearty thanks to the Marine Environmental Data Service in Canada and the Japan Oceanographic Data Center (especially the efforts of the drifter community) for their offers of drifting buoy data.
