Model Setting

We used the Regional Ocean Modeling System (ROMS) [Shchepetkin and McWilliams, 2005] to model the three-dimensional, time-dependent circulation governed by the hydrostatic primitive equations. For the vertical mixing parameterization, we adopted the local closure scheme of Mellor and Yamada [1982] which is based on the level-2.5 turbulent kinetic energy equations.

The model domain extended from 0.95oN, 99oE in the southwest corner to the northeast corner of the Sea of Japan . This computation domain provided a high degree of dynamic freedom for the western boundary current (Kuroshio) and for transport/exchange flow through the major straits/slope surrounding the CS to be developed by the model physics of the primitive equations. We adopted a curvilinear grid with a (430, 550) dimensional array for the horizontal coordinates (x, y). The horizontal size of this grid array decreased gradually from 10 km in the southern part to 7 km in the northern part of the domain. Vertically, we adopted a 30-level stretched generalized terrain-following coordinate (s). The vertical grid spacing had a higher resolution in the surface and bottom boundary layers and avoided artificial diffusion in large water depths. We obtained water depths h(x, y) by merging ETOPO5 (1/12o) data from the National Geophysical Data Center in the U.S. with digitized water depths from navigation maps published by the China Maritime Safety Administration. Our limited-area simulation considered the physical and numerical needs of resolving circulation dynamics in the CS which would have been difficult for a global-scale model to achieve.

We forced the model with wind stress derived from climatological (averaged from 1988 to 2013) monthly Reanalysis of 10 m Blended Sea Winds released by the National Oceanic and Atmospheric Administration (, based on the bulk formulation by Fairall et al. [2003]. The data set has 0.25o0.25o resolution. We calculated the atmospheric heat and salt fluxes from climatological monthly mean NCEP (National Centers for Environmental Prediction) Reanalysis 1 meteorological variables using bulk formula. The Reanalysis has ~1.875o resolution.

We applied active open boundary conditions (OBCs), which integrated the active OBCs of Gan and Allen [2005] and the Flather-type OBCs [Flather, 1976], to accommodate concurrently tidal and subtidal forcing along the eastern and southern open boundaries of the model domain. The OBCs linked the CS processes with the remote forcing, and they allowed disturbances from the interior of the CS to propagate outward freely. The OBCs also separated all subtidal components into local (forced) and global (unforced) components such that inflow/outflow conditions can be calculated with the unforced Orlanski-type OBCs using the global component. The local components of eastward and westward depth-integrated velocities (U, V), depth-dependent velocities (u, v), temperature (T), and salinity (S) along the open boundaries were the monthly mean solutions derived from the Ocean General Circulation Model for the Earth Simulator (OFES) global model [Sasaki et al., 2008]. This global model has 10 km 3 10 km horizontal resolution and is forced with climatological monthly mean atmospheric fluxes. The same OBCs were applied to U, V, u, v, T, and S for dynamic consistency.

Tidal forcing exerts its effect on the CS through the tidal wave propagation from the WPO. We applied the tidal forcing, with eight harmonic constituents (M2, S2, N2, K2, K1, O1, P1, and Q1), derived from the Pacific Ocean and Indian Ocean subdomains of the Inverse Tide Model (ITM) [Egbert et al., 1994] as boundary forcing in our model. The ITM is constrained by TOPEX/Poseidon altimetry data and uses an advanced and effective data assimilation technique. In addition, in our model, we added tidal potential to a pressure term in momentum equation, following Ray [1998], to account for the self-attraction and loading effect of the ocean. We used the climatological monthly discharges from major rivers (e.g., the Mekong River, Pearl River, and Changjiang River) and other smaller rivers in the Gulf of Beibu, BHS, and YS as lateral buoyancy fluxes. We obtained the discharge data from the Information Center of Water Resources (Bureau of Hydrology, Ministry of Water Resources of P. R. China) and from those reported in literatures [e.g., Dai and Trenberth, 2002]. 

We initialized our model with winter climatological WOA13 (World Ocean Atlas 2013) potential temperature and salinity, and spun the model up for 50 years. To avoid a drifting (T) and (S) in the deep ocean [Gan et al., 1998], we relaxed (T, S) in the bottom layers of the deep basin (h > 200 m) to the seasonally averaged values from WOA13 in 200 days. We also applied same relaxation to S in the upper layer of the deep basin to compensate for the uncertainty in surface salinity flux arising from NCEP Reanalysis precipitation data.

For details of CMOMS implementations, please refer to Gan et al. (2016).