BACKGROUND GYRES AND MESOSCALE EDDIES
The survey in spring 1998 showed an elongated anticyclonic gyre, or warm tongue, north of about 20.5｡? (Fig. 3). South of this anticyclonic gyre there was a much larger cyclonic gyre occupying the area between 20.5｡? and 17｡?. Three eddies extending over 500m depth can be identified inside these two gyres, one warm eddy inside the anticyclonic gyre and two cold eddies inside the cyclonic gyre. The hydrographic data of summer 1994 (Fig. 2) also showed similar layout of the gyres, although only the northern part of the cyclonic gyre was evident because of the limitation of the observation area. Temperature distribution from the survey in March 1992, also limited to the north of 18｡?, showed a similar warm tongue (Fig. 4). South of the warm tongue there seemed to be a warm eddy in the upper layer at the western end inside the cyclonic gyre.This warm eddy was no longer evident in the temperature distribution at 500m depth.
The spring survey in 1998 also showed the presence of many eddies south of the cyclonic gyre below 17｡? (Fig. 3). Only the warm eddy next to Vietnam and the cold eddy east of it both extended to depths over 500m. The warm water in the upper layer next to Philippine was apparently a large fresh warm pool (Fig. 3a and 3c). Underneath this fresh warm pool the temperature distribution actually had a dome-like structure with a cold center (Fig. 3b). The temperature distribution at 500m depth (Fig. 3b) also indicated the likelihood of a basin-wide anticyclonic gyre south of 17｡? in the SCS.
The strong warm eddy next to Vietnam was also evident on the TOPEX/POSEIDON sea surface height map in the Monthly Ocean Report published by Japan Meteorological Agency. It seemed to have originated locally in the southern SCS. This warm eddy began to take shape on the map of 14 March 1998, centered near (10｡?, 112｡?). It became very strong on the map of 22 June, remained identifiable in July and disappeared after August. It may be pointed out that the cyclonic gyre between 17｡? and 20.5｡? was often present in these monthly maps. The large anticyclonic gyre south of 17｡? could also be identified in some months.
NUMERICAL RESULTS
2. 5-Layer Reduced Gravity Model
Previous numerical studies by either a 1.5-layer reduced gravity model (RGM) (Liu and Su, 1992) or atwo-layer mode (Cai, 1992) indicated that, unless its volume transport is rather small the Kuroshio cannot form a loop-like intrusion into the SCS. To see whether baroclinic instability can result in eddy-like intrusion of the Kuroshio into the SCS we employed a 2.5-layer RGM with a 18km grid-size.The model domain is limited to the north of 12｡?. We have run a variety of experiments. Anticyclonice ddies are indeed often found next to the northern boundary of the SCS in the model results (e. g., see Fig. 4)). However, none of these eddies seem to be the results of baroclinic instability. In addition, most of them are generated locally inside the SCS. Over the integration period of 13 years forone experiment, only once can we identify an anticyclonic eddy as being formed next to the Kuroshio at the Luzon Strait before moving westward into the SCS. The circulation in the SCS simulated in this 2.5-layer RGM is dominated by the same dynamics found in the 1.5-layer RGM (Liu and Su, 1992).
Fig. 4 shows four sea surface height (SSH) distributions in the SCS over 120days from one experiment of the 2.5-layer RGM. The scale for the y-axis has been reduced to two-thirds of that of the x-axis.
