Numerical Study of Circulation and Seasonal Variability in the Southwestern Yellow Sea

: A nested-grid ocean circulation modelling system (NGMS-swYS) is used for examining the impact of tides and winds on the three-dimensional (3D) circulation, hydrography and seasonal variability over the southwestern Yellow Sea (swYS). The modelling system is based on the Princeton Ocean Model (POM) and uses a nested-grid setup, with a ﬁne-resolution (~2.7 km) inner model nested inside a coarse-resolution (~9.0 km) outer model. The domain of the outer model covers the China Seas and adjacent deep ocean waters. The domain of the ﬁne-resolution inner model covers the swYS and adjacent waters. The NGMS-swYS is driven by a suite of external forcings, including the atmospheric forcing, tides, freshwater discharge and currents speciﬁed at the lateral open boundaries. A comparison of model results with observations and previous numerical studies demonstrates the satisfactory performance of the NGMS-swYS in simulating tides, seasonal mean circulation and distribution of temperature and salinity. Five additional numerical experiments were conducted using NGMS-swYS with different combinations of external forcing. Analysis of model results demonstrates that the monthly mean circulation over the swYS is affected signiﬁcantly by tides and winds, with large seasonal variability. The northward Subei Shoal Current occurred in both winter and summer months in 2015, with persistent strong southeastward mean currents induced by tides along the 50 m isobath. Model results also demonstrated that strong wind-induced currents occurred with large sea surface cooling during Typhoon Chan-Hom.


Introduction
The southwestern Yellow Sea (swYS, see Appendix A for all abbreviations and symbols used in this paper) is a semi-enclosed marginal sea in the northwest Pacific Ocean, with an average water depth of about 44 m (Figure 1). The topography in the swYS is characterised by rugged coastlines and highly variable sea bottoms, with the Radial Sand Ridges (RSR) in the inner shelf region and the Yellow Sea Trough in the offshore region. Another important geographic feature in the swYS is the Changjiang Bank (CJB), which is a flat and broad shallow area located at the junction of the Yellow Sea and the East China Sea [1].
The hydrodynamics in the swYS was found to have significant seasonal variability [2][3][4]. The monthly mean oceanographic conditions in February and August are used in this paper to represent the winter and summer conditions. The February mean sea surface temperatures (SSTs) over the swYS (Figure 2a), inferred from the MURSST satellite remote sensing data over a 3-year period from 2014 to 16, vary from about 4 • C over the coastal waters off Shandong to about 14 • C in the deep waters around Jeju Island, with isotherms following approximately isobaths. The August mean SSTs during these 3 years are warm and in a range of 25-27 • C in the swYS (Figure 2b), with several patches of cold waters over the Subei Shoal (SS), HZB and Zhoushan Island [5,6]. The salinity over the swYS is The general circulation in the swYS includes the Yellow Sea Coastal Current (YSCC), Taiwan Warm Current (TWC), Yellow Sea Warm Current (YSWC) and Changjiang Plume (CJP). The circulation in the region is affected by the seasonally varying atmospheric forcing associated with monsoons [8][9][10][11], with the strong northeast monsoon in winter and the rain-bearing southwest monsoon in summer [4]. In winter, the shelf circulation system in the swYS normally features two major currents: a southeastward coastal current (known as the YSCC) flowing approximately along the 50 m isobath from the Lunan Coast Water (LNCW) to the CJB and a northwestward current (known as the YSWC). The latter separates from the Kuroshio Current and runs from the southwest of Jeju Island to the northern YS [10][11][12]. The circulation in the northern ECS features the weakly seaward CJP and northward TWC [9]. In summer, the YSCC in the swYS and the northward Korea Coastal Current (KCC) form the cyclonic circulation over the southern YS, while the YSWC disappears due to the adjustment of the circulation in the ECS. The CJP and TWC are enhanced in summer due mainly to the increase of the CRD and the effect of the southwest monsoon, respectively [13][14][15][16]. The general circulation and hydrography in the swYS (and also in the CJE) are significantly modulated by tides. The tidal residual currents were found to contribute more than 50% to the time-mean flow over the Changjiang Bank [1]. Over the RSR, the maximum tidal range reaches up to 9.4 m, and the tidal currents reach up to 2.5 m s −1 [17]. Over the SS, the tidally induced residual currents compensate for the wind-driven currents, leading to the counter-wind transport in winter months [1,8,13]. An upwelling branch over the swYS is caused by the large baroclinic gradient across the strong tidal mixing fronts (TMF), which is induced by tidal mixing over sloping topography in the swYS [18]. As a result, several cold water patches were observed in summer in this region [19].
The general circulation, hydrography and seasonal variability in the swYS were studied previously based on observations and numerical results [5,13,[20][21][22][23][24][25]. The unfavourable natural and political conditions, however, have posed a great challenge to making successive and synchronous in situ observations over this region. With the advent of computer resources and advanced numerical methods, numerical ocean circulation models have increasingly been used in understanding the roles of tides and winds in the temporal and spatial variability of hydrography and circulation over the region. Naimie et al. [25] simulated the seasonal mean circulation in the YS and analyzed their dominant dynamics. Xia et al. [24] studied the three-dimensional (3D) structure of the summertime circulation of the Yellow Sea using the Princeton Ocean Model (POM). Wu et al. [26] examined the tidal effects on the CJP using the Estuarine, Coastal and Ocean Model (ECOM). Xuan et al. (2016) investigated the effect of tidal residual current on the mean flow over the CJB using the Finite Volume Coastal Ocean Model (FVCOM) [1]. Most of the previous studies focused on the hydrography and circulation in summer. Very limited studies, however, were made on wintertime hydrodynamics over the region. In this study, a nested-grid ocean circulation model was used to examine the impacts of tides and wind forcing on the hydrography, circulation and seasonal variability in the swYS. Our results will be useful in improving our knowledge of temporal and spatial variability of circulation and the main processes affecting the macroalgal blooms of Ulva prolifera in this region [27,28]. As suggested by Zhou et al. [20] and Wang et al. [21], the hydrographical structure, horizontal advection and vertical mixing play substantial roles in the phytoplankton bloom in the swYS. This paper is arranged as follows. Observational and reanalysis data used in this study are introduced in Section 2. The nested-grid ocean circulation modelling system, external forcing and design of experiments are presented in Section 3. Assessment of the model performance is given in Section 4. The impacts of tides and winds on seasonal variations of hydrography and circulation are studied in Section 5. A conclusion is provided in Section 6.

Observational and Reanalysis Data
Four types of data are used in this study to drive the circulation model and to assess the model's performance. These four types of data include satellite remote sensing data, in situ oceanographic observations, atmospheric reanalysis and hydrographic climatology. The satellite remote sensing data include the Multi-scale Ultra-high Resolution (MUR) SST data generated by the Jet Propulsion Laboratory of National Aeronautics and Space Administration [29]. The MURSST data are based upon the nighttime GHRSST L2P skin and subskin SST observations and in situ SST observations. The gridded MURSST data have a spatial resolution of 0.01 • in longitude and latitude and at a daily frequency.
The in situ oceanographic data used to assess the model performance included the time series of sea surface temperature (SST) and sea surface salinity (SSS) observed at three stations ( Figure 1b) over the Chinese coastal waters of the YS, namely the Xiaomaidao station (XMD), Lianyungang station (LYG) and Lvsi station (LVS). There were six hourly observations at 08:00, 14:00 and 20:00 (UTC + 8 h) at locations XMD and LYG and hourly observations at LVS in 2015. It should be noted that only the observed SSS at site XMD in 2015 were available for this study. These in situ data were provided by the National Marine Data Center (NMDC). The observed hourly surface elevations at Pingdao, Sheyang, Waikejiao and Lianyungang stations in July 2016 provided by the State Oceanic Administration were also used. The observed vertical profiles of temperature and salinity along transect HH (Figure 1b) were also used, which were conducted by Hohai University on 21 July 2016 using conductivity-temperature-depth (CTD) [5]. The observed vertical profiles of horizontal currents at station DD (its position is shown in Figure 1b) were also used, which were also conducted by Hohai University from 06:00 on 23 July to 08:00 on 24 July 2016 using the ADCP, with the sampling interval being 1 s. The ADCP data were averaged over 60 s time windows.
The monthly climatology of temperature and salinity digitised from the Marine Atlas of Bohai Sea, Yellow Sea and East China Sea (Hydrology) was also used to assess the model performance. This Marine Atlas was published by the Editorial Board for Marine Atlas based on the collection of historical observations between 1970 and 1992 [7].
The hourly CFSv2 atmospheric reanalysis data produced by the National Centers for Environmental Prediction (NCEP) of the United States [30] from 1 January 2014 to 31 December 2016 were used to force the ocean circulation model. The horizontal resolution of the CFSv2 data was~0.2 • ×~0.2 • .

Nested-Grid Modelling System and Forcing
The ocean circulation model used in this study is the nested-grid modelling system for the southwestern Yellow Sea (hereafter NGMS-swYS) based on the POM [31]. The modelling system uses the nested-grid configuration similar to a triply-nested coastal circulation forecast system for the Pearl River Estuary developed by Sheng et al. [32]. The NGMS-swYS has two components (  (Figure 1b), with a horizontal resolution of about 2.7 km. The 30 arcsec General Bathymetric Chart of the Oceans (GEBCO) bathymetry data are used in the inner model, except for the SS and CJE. Over areas of the SS and CJE, the fine-resolution topography in the model is based on the Marine Atlas published by the Ministry of Transport of the People's Republic of China. Both the outer and inner models are 3D circulation models with 41 sigma levels for the vertical coordinates. The horizontal viscosity and diffusivity coefficients (A m and A h ) are calculated in the model using the scheme suggested by Smagorinsky [33], with the ratio of A h to A m for the scheme being 0.1. The vertical viscosity and diffusivity coefficients K m and K h are calculated using the modified Mellor-Yamada 2.5 turbulence closure scheme [34,35].
The initial conditions for temperature, salinity and velocity in the outer and inner models are taken from the daily mean products produced by an East Asian Marginal Seas model, with a horizontal resolution of 1/12 • × 1/15 • [36]. The external forcing for driving the outer and inner models of the NGMS-swYS in the control run (CR) includes the atmospheric forcing, tides, freshwater discharge and boundary forcing specified at the model lateral open boundaries. The atmospheric forcing includes the wind field taken from hourly NCEP atmospheric reanalysis. The bulk formula of Kondo [37] is used to convert the wind speed to wind stress. The net heat flux at the sea surface is calculated using shortwave radiation, long-wave radiation and sensible and latent heat fluxes in the model. All these components are calculated using the empirical formulas of Hirose et al. [38]. The net freshwater flux at the sea surface is calculated based on differences between precipitation and evaporation. The CRD is specified based on the monthly mean data published in the Chinese River Sediment Bulletin [39].
The tidal forcing at open boundaries of the outer model is specified using the radiation conditions for the depth-averaged currents and tidal surface elevations using the boundary condition suggested by Davies and Flather [40]. Four major tidal constituents (M 2 , S 2 , K 1 and O 1 ) are used based on the dataset produced by the TPXO ocean tidal circulation model [41]. The non-tidal components of the open boundary conditions are taken from the monthly mean reanalysis of the 1/2 • × 1/2 • surface elevations, currents, temperature and salinity produced by the Geophysical Fluid Dynamics Laboratory CM2.5 coupled circulation-ice model with a simple ocean data assimilation [42]. The GFDL CM2.5 coupled model has 1440 × 1070 eddy-permitting quasi-isotropic horizontal grid cells. The observed temperature and salinity assimilated into the GFDL CM2.5 model include the World Ocean Database of historical hydrographic profiles, in situ observations and satellite remotely sensed SST. More details of the model are reported by Carton et al. [42].
The surface elevations and depth-mean currents of tidal forcing at open boundaries of the inner model are specified using the same open boundary condition as in the outer model [40]. For the 3D currents, temperature and salinity, the adaptive open boundary condition is used at the open boundaries of the inner model based on the direction of currents (or waves) through the open boundary [43]. If the open boundary is active (i.e., Model results in five numerical experiments (Control Run, NoTide, NoWind, NT_CH, and NT_CH, see Table 1) are analyzed to examine the effects of tides and winds on seasonal variability in circulation and hydrography over the swYS. In the Control Run (CR), the NGMS-swYS is driven by the full suite of external forcing discussed above. In the case of NoTide (NT), the model setup and external forcing are the same in the CR, except for the exclusion of tidal forcing at the open boundaries of the outer model. In the case of NoWind (NW), the model setup and external forcing are the same in the CR, except that the wind forcing in the momentum equation and vertical turbulent kinetic energy equation is set to zero in the model. The NGMS-swYS in these three cases is integrated for three years from 1 January 2014 to 31 December 2016 (UTC). The model results of the last two years are used in this study. Two additional numerical experiments (cases NT_CH and NW_CH, Table 1) were conducted to examine the effects of tides and winds in circulation and hydrography over the swYS during Typhoon Chan-Hom. In the cases of NT_CH and NW_CH, the external forcing is the same in the CR, except for the exclusion of tidal forcing and wind forcing between 9 July and 12 July 2015. The model in cases NT_CH and NW_CH are integrated from 1 January to 31 December 2015. For convenience, the swYS is divided into the coastal region with water depths shallower than 50 m and the offshore region with water depths deeper than 50 m. The coastal region includes the LNCW, SS, CJE and CJB. The offshore region consists of the YS Trough and adjacent deep water regions. Circulation and hydrography have significantly different features in these two regions.

Model Performance
We first assess the performance of the NGMS-swYS in simulating the tide elevations over the coastal and shelf regions of the NWPO.  [44], Guo and Yanagi [45] and Zhang and Sheng [46], and also tidal charts included in the Marine Atlas for the Bohai Sea, Yellow Sea and East China Sea [7].
The co-amplitudes and co-phases of M 2 tidal elevations produced by the outer model in the case of CR (Figure 3a) are also compared to the counterparts from the TPXO ocean tidal model ( Figure A1a   The horizontal distribution of simulated S 2 tidal waves is very similar to the M 2 tidal waves, except for relatively smaller amplitudes of S 2 tidal elevations than M 2 tidal elevations. The simulated S 2 amplitudes are about 0.5 m over coastal waters of southeastern China, about 0.2 m over the SS and larger than 0.6 m in Gyeonggi Bay. The simulated S 2 tidal elevations also have four amphidromic points at locations very similar to the M 2 tidal elevations. The general features of co-amplitudes and co-phases for S 2 tidal elevations produced by the outer model agree very well with the previous model studies [7,45]. The S 2 tidal wave propagations and locations of four amphidromic points produced by the outer model ( Figure 3b) are also in very good agreement with the TPXO results ( Figure A1b). The simulated S 2 amplitudes produced by the outer model are slightly smaller than the TPXO results, with a maximum difference of about 0.2 m over Korea Bay.
The simulated K 1 tidal waves produced by the outer model propagate northward from the deep ocean waters to the south of western Ryukyu Islands into the ECS, the YS and the BS (Figure 3c Figure A1d) and previous numerical studies [44][45][46].
We next assess the model performance by comparing the simulated surface elevations produced by the inner model in the case of CR ( Figure 4  We next examine the model performance in simulating the hydrography over the swYS. Figure 5 shows the monthly mean SSTs in February and August 2015 produced by the inner model in the case of CR (right panels) and derived from the satellite remote sensing data (left panels). In February (Figure 5b), the simulated monthly mean SSTs are cool and about 2 • C over the SS and LNCW; they are warm and about 16 • C or higher near the southwestern and southeastern areas of the inner model domain. The simulated February mean SSTs shown in Figure 5b indicate the influence of two warm currents, namely the TWC and the YSWC, which transport the high temperature (about 18 • C) and high salinity (about 34 psu) waters to the swYS in wintertime. In August 2015, the surface water is significantly warmer than in February over the whole inner model domain ( Figure 5d). The simulated August mean SSTs are about 26 • C over most areas of the swYS, except for a cold water patch of about 24 • C located off the SS [18]. This cold water patch indicates an upwelling off the SS in the summer months, as discussed in Lü et al. [18] and Wang et al. [23].
A comparison of the monthly mean SSTs produced by the inner model and satellite data (MURSST) demonstrates that the inner model reproduces well the SSTs over the inner model domain. The differences in the monthly mean SST between the inner model results and MURSST are less than 1.0 • C in both February and August over most areas of the swYS but relatively large and up to about 2.5 • C over the inner shelf of the SS in August (Figure 5c,d). The simulated August mean SSTs produced by the inner model ( Figure 5d) are significantly warmer than the MURSST (Figure 5c) over the inner shelf due mainly to the fact of less reliable SSTs inferred from the satellite remote sensing data over coastal areas. In comparison, the simulated August mean SSTs produced by the inner model agree well with the previous numerical results produced by Huang et al. [19] over the inner shelf of the swYS. Furthermore, a comparison of the in situ hydrographic observations and simulated SSTs produced by the inner model at three Chinese coastal stations (to be shown later) also demonstrates that the NGMS-swYS reproduce well the SSTs over the inner shelf of the swYS. Figure 6 presents the monthly mean sea bottom temperatures (SBT) in February and August produced by the inner model in the case of CR (right panels) and derived from the Marine Atlas for the Bohai Sea, Yellow Sea and East China Sea (Hydrology) [7] (left panels). In February, the distribution of the SBT produced by the inner model ( Figure 6b) is very similar to the SST, as discussed above. This indicated that, in February 2015, the high-temperature TWC and YSWC flowed northward into the swYS in the whole water columns. The general features of the SBT produced by the inner model agree with the Marine Atlas ( Figure 6a). However, the simulated SBT produced by the inner model implies that the TWC flows more northward to the eastern CJE and that the YSWC is weaker than the counterparts in the Marine Atlas. In August 2015, the bottom layer of the YS Trough in the inner model was occupied by the relatively cool water lower than 8 • C (Figure 6d), known as the Yellow Sea Cold Water Mass (YSCWM) [48] extending from the central YS to the ECS. The August mean SBT produced by the inner model reaches up to 28 • C over the inner shelf, with sharp temperature gradients between the coastal warm water and YSCWM. The distribution of the simulated YSCWM in the YS Trough produced by the inner model agrees with the Marine Atlas (Figure 6c), except that the fringe of the YSCWM is significantly affected by the topography in the model results. Furthermore, the simulated SBT produced by the inner model is higher than the Marine Atlas over the coastal region.
The vertical distribution of observed temperature and salinity on 21 July 2016 along transect HH made by Hohai University is presented in Figure 7a,c. This transect extends from SS to YS Trough (marked in Figure 1b). Over the inner section of the transect (on the SS), the vertical stratification of observed temperature and salinity were vertically uniform and weakly stratified horizontally with vertically well-mixed temperature (salinity) varying from 26.2 • C (28.6 psu) at site HH1 near the coast to about 22.5 • C (31.3 psu) at site HH3. The observed low-salinity (<30 psu) water over the coastal area of transect HH (Figure 7c) originated from the CRD [13].  The simulated daily mean temperature and salinity on 21 July 2016 along transect HH produced by the inner model in the case of CR are compared with the hydrographic observations in Figure 7. The inner model reproduces very well the vertical distribution of observed temperature and salinity over the inner section of the transect (Figure 7b,d), with the simulated temperature (salinity) varying from 27.5 • C (29.2 psu) at HH1 to 23.2 • C (31.5 psu) at HH3 over the inner section of transect HH. The inner model also satisfactorily reproduces the observed upward climbing of bottom cold waters and the observed vertically uniform and horizontally stratified temperature and salinity below 20 m over the central section of the transect. Over the outer section of transect HH, the inner model reproduces reasonably well the uniform temperature and salinity in the surface layer of fewer than 5 m and the subsurface layer below 30 m. It should be noted that the thermocline in the model is too diffusive in comparison with observations, due mainly to a model deficiency [49].  To assess the performance of the NGMS-swYS in simulating the seasonal cycle and synoptic variability of temperature and salinity, the simulated temperature and salinity produced by the inner model in the case of CR were compared with the hydrographic observations at three sites (XMD, LYG and LVS marked in Figure 1) over the Chinese coastal waters of western YS in 2015 provided by the NMDC (Figure 9). The observed SSTs at these three sites had large seasonal cycles, which increased from about 5 • C in the winter to about 30 • C in the summer in 2015 (Figure 9a,c,d). The observed SSTs at the three sites also had significant synoptic variability associated mainly with wind-driven upwelling and downwelling and other dynamic processes.
The simulated SSTs at three sites (XMD, LYG and LVS) produced by the inner model agree in general with the in situ observations, particularly the seasonal cycle (Figure 9a,c,d).
It should be noted that the simulated SST at site XMD in the period between May and September 2015 was warmer than the observations, with differences reaching up to 3.

Physical Processes Affecting Seasonal Circulation and Hydrography
The three-dimensional (3D) circulation and hydrography in the swYS have significant seasonal variability, as shown above. The model results in 2015 in cases CR, NT and NW are examined in this section to investigate the main physical processes affecting the circulation and hydrography, with a special emphasis on the roles of wind and tidal forcing on the monthly mean fields.
We follow Wang et al. [52] and quantify the impacts of tides (and winds) on the seasonal circulation using the impact index (P i ) defined as where U CR and U NT represent the monthly mean horizontal velocity vectors produced by the inner model in cases CR and NT, respectively, and |A| is the amplitude of the vector. It should be noted that (U CR − U NT ) includes the currents driven directly by tidal forcing and the nonlinear interactions of tidal currents with wind-driven currents and densitydriven currents. The above-mentioned nonlinear interactions could not be produced by the circulation model with the tidal forcing only. The impact index for wind forcing can be calculated in the same way as in Equation (1) by replacing U NT with U NW .  (Figure 10d).

Impacts of Tides and Winds on Seasonal Circulation
The effect of tidal forcing on the February mean circulation in terms of the tidal impact index (TII) is shown in Figure 10b for the upper layer and Figure 10e for the lower layer. In the upper layer, the effect of tidal forcing on the February mean circulation is significantly large, with the TII values greater than 0.6 over the inner shelf of the swYS, the YS Trough and the ZJCW. The tidal effect is also large for the February mean circulation in the upper layer, with the TII values between 0.2 and 0.6 over the eastern CJE. The TII values are relatively small for the main pathway of the TWC in the upper layer (Figure 10b).
The tidally induced monthly mean currents (TIMMCs, i.e., the monthly mean tidal residual currents) in the upper layer can be approximated by the differences (∆ U tide UL ) in the monthly mean currents between cases CR and NT. The ∆ U tide UL distribution in February 2015 (arrows in Figure 10b      The effect of tidal forcing on the August mean circulation in terms of the TII is shown in Figure 11b for the upper layer and in Figure 11e for the lower layer. In the upper layer, the effect of tidal forcing on the August mean circulation is significantly large, with the TII values greater than 0.6 over the northern and western YS Trough, the CJE, the QTB and the ZJCW. The tidal effect is also large for the August mean circulation in the upper layer, with the TII values between 0.2 and 0. 6  In the lower layer, the effect of tidal forcing on the August mean circulation is significantly large, with the TII values greater than 0.6 over the LNCW and the YS Trough (Figure 11e). Over other areas, the tidal effect for the August mean circulation in the lower layer is relatively weak. The ∆ U tide LL distribution in August (arrows in Figure 11e) demonstrates that the lower-layer TIMMCs mainly move along the YS Trough (up to 0.18 m/s) and have a mesoscale eddy (up to 0.07 m/s) off the CJE in the lower layer.
The TIMMCs in August have flow directions similar to the February condition in both upper and lower layers. However, the magnitudes of the southeastward TIMMCs along the western slope of the YS Trough in August are larger than that in February in both upper and lower layers, due to the interaction between tidal currents and baroclinicity, since the temperature and salinity are highly-stratified in summer months in the swYS. The above analyses demonstrate that, in February 2015, the WIMMCs flow against the northward SCC over the SS, and the northward YSWC along the YS Trough in the lower layer was enhanced by the wind-induced barotropic pressure gradient. By comparison, in August 2015, the southwesterly summer monsoon was generally weak and had minor impacts on the circulation in the lower layer. The TIMMCs were the dominant currents over the coastal region of the swYS and YS Trough in both February and August 2015.

Impacts of Tides and Winds on Seasonal Hydrography
The differences in the monthly mean SST (∆T tide S ) and SSS ∆S tide S between cases CR and NT were used to quantify the effects of tidal mixing and advection on the monthly mean hydrography over the swYS. In February 2015, the ∆T tide S values (Figure 12a) were relatively large and up to 2.2 • C over the deep waters off the SS and along the 50 m isobaths, due mainly to the transport of cold water by the TIMMCs from the northern YS to the swYS. In comparison, the ∆T tide S values are very small over the SS in this month. The ∆S tide S values in this month (Figure 12c) are significantly negative over the CJE and the coastal water off Sheyang of China, with maximum negative values of about −7.8 psu over the CJE and about −6.5 psu over the coastal water off Sheyang of China, which are mainly due to the effect of the unique tidal wave system over the swYS [8], which means that the northward TIMMCs from the CJE and the southward TIMMCs from the LNCW merge over the coastal water off Sheyang of China, then turn eastward to flow offshore, and the TIMMCs over the RSR spread seaward [51]. The tidal residual currents transport high salinity water from the offshore area to the RSR; thus, positive ∆S tide S values (up to 6.5 psu) occur in this region. ) in the monthly mean SST and SSS between cases CR and NW are used to quantify the effect of wind forcing on the monthly mean hydrography at the sea surface. It should be noted that the simulated SSTs in the case of NW have similar seasonal cycles as the counterparts in the case of CR since wind forcing is used in calculating the net surface heat flux in case NW. If the wind forcing is set to zero in the calculation of the net surface heat flux, the model does not generate the seasonal cycles in the SST. Therefore, only the role of advection and vertical mixing generated by wind forcing on the hydrographic distribution over the swYS is considered based on differences in simulated hydrography between cases CR and NW.
The    (Figure 15d). Based on differences in SST, SBT, SSS and SBS between cases CR and NT in February and August 2015, tidal residual currents are found to play an important role in the northward movement of the CRD over the coastal region and southeastward movement of waters from northern YS to the swYS, which have significant effects on the water exchange between different regions in both seasons. The strong tidal stirring also enhances the mixing of the surface and subsurface waters over the regions between 30 m and 50 m isobaths in February and August. The northerly winter monsoon transports cold water from the northern YS to the swYS and blocks the northward advection of CRD, while the upwelling-favourable southwesterly wind in summer enhances the northward advection of CRD and inshore transport of bottom waters in the swYS. The wind-induced mixing also contributes to the exchange of surface and subsurface waters in both seasons. We next examine the effects of tides and winds on the vertical distribution of monthly mean temperature and salinity along transect HP (shown in Figure 5a) based on the inner model results in the case of CR. In February 2015 (Figure 16a,b), the vertically uniform cold (about 5.7 • C) and low-salinity (about 29.3 psu) waters occurred in the inner section of transect HP, while vertically uniform warm (about 9.3 • C) and high salinity (about 34.1 psu) water covered the YS trough of transect HP.
The differences in the February mean temperature and salinity between CR and NT (∆T tide HP , ∆S tide HP ) along transect HP are presented in Figure 16c  The vertical distribution of the August mean temperature and salinity produced by the inner model in case CR is presented in Figure 17a,b. The monthly mean temperature and salinity in this month are vertically uniform and weakly stratified horizontally over the inner section of transect HP. The vertically well-mixed temperature and salinity in this month vary from 28.  in the bottom layer, which is generated by the secondary vertical circulation induced by the strong baroclinic pressure gradients in the TMF [18]. Over the swYS, the tidal forcing generates strong vertical eddy viscosity in the near-bottom layer. The strong eddy viscosity mixes the warm water close to the thermocline and the cold water in the bottom layer; as a result, the lower mixed layer below the strong thermocline is maintained. Therefore, the significant negative values of the ∆T tide HP for the August mean temperature (up to −3.5 • C) in the intermediate layer between 20 m and 50 m are generated by the tidal mixing over the outer section. The positive values of ∆S tide HP in Figure 15d demonstrates that the tidal forcing increases the salinity over the entire transect HP in this month. The maximum value of ∆S tide HP occurs in the upper layer of the central section and up to 2.6 psu since the tidal residual currents transport high-salinity water from the northern YS to this region.  (Figure 17f), the monthly mean salinity in this month is slightly increased by up to 0.7 psu due to the wind-induced transport of high salinity water. The cool and high salinity water over the central section of transect HP originates from the subsurface water in the YS Trough, which needs further study.

Circulation and Hydrography during a Storm
The 3D circulation and hydrography over the swYS are also affected by typhoons [53,54]. In this section, we examine the model results in three different experiments (CR, NT_CH and NW_CH) over the swYS during Typhoon Chan-Hom [50]. Chan-Hom was formed in the western Pacific Ocean on 30 June 2015 and passed the Miyako Strait on 10 July 2015. The storm swept Zhejiang Province of China on 11 July 2015 and then moved into the YS on 12 July with a maximum wind speed of about 40 m/s [54]. Chan-Hom passed through the YS and caused economic losses exceeding 8.5 billion Chinese Yuan. Figure 18 presents the simulated SST, SSS and surface currents produced by the inner model in the case of CR and the daily mean wind vectors at 10 m derived from the NCEP atmospheric reanalysis data before, during and after the passage of Typhoon Chan-Hom. Before the arrival of Chan-Hom on 3 July 2015, the daily mean surface currents (DMSCs) produced the inner model (arrows in Figure 18d) flow northward along the Jiangsu and Lunan Coast over the SS, HZB and LNCW and southwestward along the Zhejiang Coast over the ZJCW under the influence of the easterly and northeasterly winds (arrows in Figure 18a). On the same day, the simulated temperature and salinity (Figure 18a,d) have typical hydrographic features in summer months, with relatively warm and low-salinity waters over coastal waters and relatively cool and high-salinity waters in the middle and outer shelf regions of the swYS. The simulated daily mean SSTs on 3 July vary between 18 • C and 27 • C, with several cold water patches occurring between the 30 m and 50 m isobaths, western area of Jeju Island and southern area of the Shandong peninsula. Previous studies suggest that these cold water patches are induced by strong tidal mixing and wind-induced upwelling [5,18,46]. The CJP on this day mainly flowed northward with the 31 psu isohaline extending northward to 36 • N and eastward to 123 • E. The simulated daily mean SSTs on this day were significantly reduced and low to 16.9 • C over the middle and outer shelf regions of the swYS, especially over areas with water depths deeper than 50 m. The differences in the daily mean SST between coastal waters of the SS and YS Trough are very large and greater than 8 • C. The daily mean SSS on 12 July 2015 was confined to the inner shelf of the SS, the CJE and the ZJCW (Figure 18e) in comparison with the typical SSS distribution in summer months [26], due mainly to the strong southward currents induced by Typhoon Chan-Hom.
After the passage of Typhoon Chan-Hom on 27 July 2015, the simulated surface circulation and hydrography over the swYS returned gradually to the typical summertime patterns (Figure 18c,f). On this day, the daily mean southwesterly winds over the swYS were weak and less than 5 m/s. The strong TWC flowed northeastward over the ZJCW on this day, with speeds up to 1.1 m/s. The CJP (up to 0.4 m/s) mainly flows eastward over the CJE, and the DMSCs (up to 0.6 m/s) flow southeastward over the central swYS. The daily mean SSTs on this day were warm and up to 30 • C over the coastal region of the swYS and about 24.5 • C over the offshore regions of the swYS. The eastward CJP transports the low-salinity water to the offshore area with the 31 psu isohaline extending eastward to 124 • E.
The differences in the simulated SST (∆T CH s ), SSS (∆S CH s ) and surface currents (∆U CH s ) between cases CR and NW_CH and cases CR and NT_CH on 12 July 2015 ( Figure 19) were used to examine the impacts of tides and winds on the hydrography and circulation over the swYS during Typhoon Chan-Hom. The   Figure 19i) featured a cyclonic circulation (up to 0.53 m/s) over the swYS and seaward currents (up to 0.41 m/s) over the offshore region of the CJE and QTB. Similar to previous studies, the spread of low-salinity water over the inner shelf of the SS was controlled by tidal residual current [8,54], the tide-induced ∆S CH s were positive and up to 6.5 psu over the inner shelf of the SS on this day. The tide-induced ∆S CH s are negative and up to −10.8 psu over the eastern CJE and positive and up to 10.5 psu over the eastern QTB, due mainly to the strong tidal mixing and tidal residual currents, respectively.

Discussion and Conclusions
A nested-grid ocean circulation modelling system was used to examine the impacts of tides and winds on the circulation, hydrography and associated seasonal variability over the southwestern Yellow Sea (swYS). The nested-grid modelling system (NGMS-swYS) is based on the Princeton Ocean Model (POM) and has two components, with a fine-resolution inner model embedded inside a coarse-resolution outer model. The domain of the outer model with a horizontal resolution of about 9.0 km covers the China Seas and adjacent deep ocean waters of the northwestern Pacific Ocean. The domain of the inner model with a horizontal resolution of about 2.7 km covers the swYS and adjacent shelf waters. Five numerical experiments, namely the control run (CR), NoTide (NT), NoWind (NW), NoTide_Chan-Hom (NT_CH) and NoWind_Chan-Hom (NW_CH), were conducted using the NGMS-swYS with different combinations of external forcing. The model performance in the case of CR was assessed using the MURSST data, Marine Atlas and in situ oceanographic data over the study region. It was found that the NGMS-swYS has satisfactory skills in reproducing the four major tides (M 2 , S 2 , K 1 and O 1 ), three-dimensional (3D) circulation and hydrography over the swYS.
The monthly mean model results in cases CR, NT and NW produced by the inner model in February and August 2015 were analyzed to examine the 3D circulation and hydrography and the roles of tidal and wind forcings in the winter and summer months over the swYS. The monthly mean circulation in case CR featured a northward Subei Coastal Current and persistent southeastward mean currents along the 50 m isobaths in both February and August. These currents were stronger in August than in February of the same year.
The roles of tidal forcing on the 3D circulation and hydrography over the swYS were quantified based on differences in monthly mean results produced by the inner model between cases CR and NT. It was found that the tidally induced monthly mean currents (TIMMCs) are similar over the SS and western slope of the YS Trough in both February and August, indicating the barotropic characteristics of the tidal residual currents over these areas. By comparison, the TIMMCs differ significantly over the LNCW, HZB and CJB in February and August, and the southeastward TIMMCs along the 50 m isobaths are stronger in August than in February, indicating the role of baroclinic tides over these areas.
The tidal residual currents also play an important role in the northward transport of the Changjiang River Discharge (CRD) over the coastal region and the southeastward moving of waters along the western slope of the YS Trough from northern YS to the swYS, which have significant effects on the water horizontal exchange between different regions in both February and August. The southeastward TIMMCs bring the cool and high-salinity water from the northern YS to the swYS, while the northward TIMMCs transport warm and lowsalinity CRD from the CJE to the SS in both February and August. The strong tidal mixing enhances the vertical mixing in the subsurface layer contributing to the maintenance of the bottom mixed layer below the strong thermocline over the YS Trough in the summertime and the maintenance of the vertically uniform water in the wintertime.
The wind-induced monthly mean currents (WIMMCs) feature a classical two-layer circulation in the vertical direction with southward currents in the surface layer and northward compensating currents flowing along the YS trough in the deep layer in February. In August, the weak wind forcing enhances mainly the coastal currents over the SS and HZB to move northward in both upper and lower layers and has a weak influence on the circulation over the YS Trough.
The southward WIMMCs transport cold water from the northern YS to the swYS over the coastal region and block the northward advection of the CRD in February, while the upwelling-favourable southwesterly wind in August enhances the northward advection of the CRD and inshore transport of bottom waters in the swYS. The wind-induced vertical mixing plays an important role in forming the vertically uniform temperature and salinity in February and maintaining the highly stratified vertical pattern in August over the swYS.
It was demonstrated that strong winds significantly changed the circulation and hydrography over the swYS during Typhoon Chan-Hom. The coastal currents over the HZB and SS turn to flow southward, and the surface currents feature a strong cyclonic circulation over the swYS during Typhoon Chan-Hom. The SST is significantly reduced in the offshore region, and the SSS is confined to the inner shelf of the coastal region during the same time.
The main focus of this study was on the impact of tides and winds on the 3D circulation, hydrography and associated seasonal variability over the swYS. Hydrodynamics of the swYS are also affected by other dynamical factors, however, including the Changjiang River Discharge, the Taiwan Warm Current and other basin-scale oceanic circulation. The 3D hydrodynamics of the swYS are also affected by the surface waves. Further research on the influences of these dynamic factors on the seasonal variability of circulation and hydrography over the swYS is needed. Furthermore, comprehensive oceanographic observations of temperature, salinity and currents, especially during extreme weather conditions, are needed to validate further the performance of the NGMS-swYS.  where n is the total number of observations, M and O are the simulated and observed values, and var stands for the variance, the overbar indicates the temporal mean. The smaller the values of RMSE and γ 2 are, the better the model performance is.