The true North is 20° to the west of the y-axis of the undistorted domain. In the upper layer the Kuroshio has an inflow of 22 Sv, 80% of which exits east of Taiwan and 10% of which each leaves the SCS through the Taiwan Strait and the Mindoro Strait, respectively. These two straits are not depicted explicitly in Fig. 4. The Kuroshio transport in the lower layer is 6 Sv, all of which leaves the model domain east of Taiwan. No wind forcing is imposed and the horizontal eddy viscosity coefficient is chosen to be 450m2s-1.
Like the 1.5-layer RGM (Liu and Su, 1992) the northern half of the SCS is dominated by a fluctuating cyclonic gyre (Fig. 4). As was demonstrated by Liu and Su (1992) positive vorticity derived from the western edge of the Kuroshio is being constantly advected into the SCS. As a cyclonic eddy grows, it starts to more westward and continues to grow at the same time. The eddy sheds to the southwest periodically at about 4.5 months intervals (half-yearly for the 1.5-layer RGM). This cyclonic gyre has an intensified northern/northwestern segment, which may be the observed Dongsha Current (Su,1998). In a realistic model the Kuroshio water may be entrained at the Luzon Strait by this gyre and then advected westward into the SCS. North of the cyclonic gyre there is an anticycionic gyre of smaller scale, likely to be induced by the cyclonic gyre. Fig. 4 also shows the pinch-off of an anticyclonic eddy from the anticyclonic gyre, associated with the passing of a growing cyclonic eddy just south of it. Although our model is highly idealized, this scenario may indeed contribute partly to the generation of the warm eddies southwest of Taiwan. In our model the anticyclonic gyre extends all the way southwestward to the Hainan Island south of China. For a realistic model the plateau near the Dongsha Islands may serve as the western limit for the anticyclonic gyre.
Single-Layer and Two-Layer Coupled Model
To model the circulation in the SCS under the combined influence of winds and the Kuroshio, we employed a single-layer and two-layer coupled model. The single-layer model is applied over the continental shelves and the two-layer model over the basin of the SCS. The continental slope is modified so that the interface fluctuates over a vertical section. The domain is flattened beyond the 3000m depth and the interface is initially at the 200m depth. In this model the Mindoro Strait is closed and zero normal gradient condition for both the sea-level and velocity is imposed at the southern boundary. The Kuroshio has an inflow of 18 Sv in the upper layer and 9 Sv in the lower layer. The model has a 40km grid-size.
Fig.5 shows three SSH distributions in the SCS over 40 days from an experiment with a constant Hellerman-Rosenstein December winds. The true North is also 20° west of the y-axis. The circulation in the northern half of the SCS has a similar pattern as that in Fig. 4, i.e., a fluctuating cyclonic gyre with warm eddies northwest of the gyre. In the southern half of the SCS there is another large fluctuating cyclonic gyre over the deep plateau west of the Nansha Trough. Taking together, the mean state of these two cyclonic gyres may be regards as one basin-wide cyclonic gyre. According to Yang Haijun (private comminication) this is consistent with the field of the mean winter wind-stress curl.
While the fluctuation of the northern cyclonic gyre is associated with the Kuroshio (see Fig. 4), the oscillation of the southern cyclonic gyre is closely associated with an intense anticyclonic eddy (Fig. 5). It is generated over the Nansha Trough, propagates northwestward, seemingly advected southward along the Vietnamese coast by the strong western boundary current there, and dissipates near the broad shelf in the south. Its period is, for this experiment, about 50 days.
