Additionally, numerical results from JAMSTEC Eddy-resolving General Circulation Model (JEGCM, Ishida et al., 1998) are also used to understand the phenomena appeared in the observational data.
2. Data origin
2.1 Observation
Hydrographic observations were conducted by R/V Kaiyo at October 1992, February 1994, January 1995, July 1995 and June 1996. Observation parameters are temperature, salinity and pressure by CTD, current profile by shipboard ADCP, salinity, oxygen and nutrient from sampled water. These observations except that at July 1995 were carried out under WOCE project. The accuracy of the observations at February 1994, January 1995 and June 1996 are enough high to satisfy the requirements defined in WOCE Hydrographic Programme (e.g., accuracy of CTD salinity at June 1996 is ＜ 0.001 PSU).
Moreover, moorings were deployed at the southwestern boundary of the Philippine Sea (solid triangles in Figure 1) from February 1994 to June 1995 to understand the variability of the Indonesian Throughflow. An upward ADCP (250m depth), a CTD (260m) and three recording current meters (350m, 550m and 1050m) with some sensors were installed on each mooring (unfortunately, ADCP did not work well). Diurnal and semi-diurnal tidal signals were removed from the raw time series using a 48 hours tide-killer filter [Hanawa and Mitsudera, 1985].
2.2 Model
To study a role of ocean meso-scale eddies in the global ocean circulation, global ocean circulation model with enough high resolution, called JEGCM (JAMSTEC Eddy-resolving General Circulation Model) to resolve the meso-scale eddies was developed in JAMSTEC: 1/4 X 1/4 degree in horizontal and 55 levels in vertical.
The model is based on the Geophysical Fluid Dynamics Laboratry (GFDL) Modular Ocean Model ver.2 (MOM2) code [Pacanowski, 1995]. In this code, three dimensional primitive equations with hydrostatic and Bussinesq approximations formulated in the spherical coordinates are implemented with the grids by Bryan [1969].
Model domain is global ocean except for the Arctic Ocean. This model was driven by annual mean climatological wind [Hellerman and Rosenstein, 1983] and surface temperature/salinity from Levitus [1982] during first two model years, using harmonic viscosity/diffusivity. Then, the integration was performed by monthly mean wind and surface temperature/salinity in next 18 model years using bi-harmonic viscosity/diffusivity.
3. Results
At first, time series of parameters from moorings mentioned in 2.1 is shown in Figure 2. Most predominant phenomenon is strong northwestward flow at the southern mooring during boreal winter. Its speed reached 60cm/sec and the mooring largely inclined due to this flow as shown in pressure variability. For example, mean CTD depth increased about 100m than that during boreal summer.
