To observe this Kuroshio recirculation as well as the Kuroshio itself, the ASUKA line was designed to extend far beyond the Kuroshio. Activities of propagating cyclonic and anticyclonic mesoscale eddies are also very high in this region, and introduce large fluctuations in the estimated transport. Actually, most of the fluctuations shown above are associated with those mesoscale eddies.
The transport of the Kuroshio as a throughflow shows the smaller mean and smaller fluctuations with no apparent seasonal signals. The transport of the Kuroshio recirculation is estimated in a similar way using the altimeter data and a relationship between the transport and SSDT difference obtained for the Kuroshio recirculation. The throughflow transport is estimated by subtracting this recirculation transport from the above-estimated Kuroshio transport. The result is shown in Fig. 2b. The transport and its variability of the Kuroshio as a throughflow are reduced, compared with those of the Kuroshio as an entire eastward flow. In the present analysis, the southern extent of the integral in estimating the recirculation transport is fixed at 26｡?. There are several options for choosing this southern extent. In any case, it is rather difficult to exclude the effect of transport of local mesoscale eddies satisfactorily. Here the latitude of 26｡? is chosen partly because it is close to the southern end of the local stationary anticyclonic warm eddy in the climatological mean field (8) and partly because eddy activities are lowest at this latitude on the ASUKA line for this time period of 1992-1997, which is deduced from the altimetry data.
For comparison, the transport estimated by a simple linear theory for the wind-driven ocean (so-called Sverdrup theory) using surface wind data prepared by the European Centre for Medium-Range Weather Forecasts (K. Kutsuwada, personal communication) is shown in Fig. 2b; here the integral is made over the entire width of the basin. The theoretical estimates show a very clear seasonal cycle, in contrast to the observed fluctuation. This contrast is further confirmed by comparison of the five-year mean monthly values between the observed Kuroshio transport as a throughflow and the theoretically estimated transport (Fig. 3). The observed transport coincides with the theoretical tranport during winter and spring, but they are quite different from each other during summer (from July to September). The seasonal signal of the observed transport is small compared with the theoretical one, although there may be some seasonality that it is low in September and high in December, similar tendency as the theoretical transport.
DISCUSSIONS
The presently estimated mean transport of the Kuroshio is fairly large compared with the transport of the Kuroshio in the upstream region. The Kuroshio mean transport east of Taiwan is recently estimated to be about 23 Sv from moored current meter data and repeated hydrography data (9). The mean transport in the East China Sea is estimated to be 24, Sv as a long-term average from hydrographic data and surface velocity data (10). The increase of the Kuroshio transport between the East China Sea and south of Japan suggests that a considerable amount of water is flowing northward outside of the East China Sea, probably east of the Ryukyu Ridge as suggested by previous studies (11).
