The Changjiang River (also known as the Yangtze River) diluted water (CDW) has a significant impact on the circulation structure, water mass composition, and sediment deposition in the Yellow Sea (YS) and the East China Sea (ECS). The CDW generally enters the YS and ECS and spreads out in a plume and the surface salinity around the Changjiang River Estuary shows significant seasonal variation (Wu et al., 2020) although observations indicate that isolated low-salinity water lenses (LSWLs) occasionally detach from the main body of the CDW (Pu, 2002). These LSWLs are an important factor causing area eutrophication and hypoxia in the Changjiang River Estuary and its adjacent waters (Xuan et al., 2012); thus, this topic has attracted considerable attention from ecological dynamics scientists (Kim et al., 2009; Li et al., 2002, 2019; Zhang et al., 2019). Studies on the LSWL mechanism have been conducted. By analyzing observed data from the area east of Cheju Island, Lie et al. (2003) suggested that the separation of CDW patches from the shallow shelf area and upwelling favorable southerly winds in summer are the main contributors. Chen et al. (2008) used the unstructured-grid, Finite-Volume Coastal Ocean Model (FVCOM) to simulate a LSWL in the sea west of 123°E and proposed that the detachment of this LSWL is associated with eddies generated by baroclinic instability across the plume front. Using the Regional Ocean Modeling System (ROMS), Moon et al. (2010) simulated a LSWL under ideal wind, tide and runoff conditions and suggested that the strengthening of tidal mixing during the spring tide plays an important role in the detachment of the LSWL. Wu et al. (2011) indicated that strong upwelling during the spring tide can disconnect part of the Changjiang River plume. Xuan et al. (2012) used the MIT general circulation model (MITgcm) to analyze the dynamic mechanism responsible for LSWL detachment and indicated that wind mixing, wind-driven northward currents and wind-induced upwelling are the three driving forces underlying the detachment of the LSWL. Using the POM-$\sigma {\text{-}}z$ model, Zhang et al. (2014a, b) demonstrated that southerly winds push the CDW northeastward and cut it off, thereby detaching the LSWL, and they also showed that the intensified tidal mixing during the spring tide plays a major role in the LSWL detachment process. Peng et al. (2014) indicated that LSWL detachment depends on different combinations of wind and tide conditions and that southerly winds during the spring tide are the main contributors to LSWL detachment.

Both winds and tides are generally believed to have important influences on LSWL detachment. In recent years, upwelling water has been found in the detachment region (Moon et al., 2010; Wu et al., 2011; Xuan et al., 2012). Although the magnitude of the upwelling velocity is small, an upwelling can carry saline seawater from the bottom to the surface and therefore can significantly impact the distribution of salinity, which may be conducive to detachment. However, the role of upwelling water on detaching the LSWL has not been thoroughly explored; thus, the effects and dynamic mechanism of upwelling during the evolution of the LSWL need to be further studied. In this paper, the dynamic mechanism of the LSWL observed in August 2006 is studied with a focus on the effects of upwelling on the LSWL.

The observed data were obtained from a cruise survey carried out during July 27 to August 12, 2006. During the cruise, conductivity-temperature-depth (CTD) instruments were used in the field program to measure the vertical profiles of water temperature and salinity. The path of the cruise extended alongshore from 29.8°N to 32.3°N and offshore from 121°E to 128°E. The sea near the Changjiang River Estuary was the key observation area, and the minimum latitudinal resolution was (1/9)°.

The distribution of surface salinity in the survey area is shown in Fig. 1a. The salinity front (less than 22) near the Changjiang River Estuary extends toward the southeast, thus reflecting the jet pattern of the CDW (Wang et al., 2012), whereas the CDW from the Changjiang River Estuary extends northeastward in a plume pattern. The detachment of the LSWL occurs on the interior side of the river plume at approximately 31.8°N, 122.7°E. The LSWL isolated by the 22 isohaline has an elliptical shape, and the value of the central closed isohaline is 20. The size of the LSWL (the diameter of the long axis of the outermost closed isohaline) is 70 km. Figure 1b shows the vertical salinity distribution along Line A through the center of the LSWL. The upper salinity stratification structure is obvious, and the lower water is mixed uniformly. Additionally, the 32 isohaline obviously descends from west to east. This phenomenon was discussed by Zhao (1993) in consideration of upwelling.

The distribution of surface temperature is depicted in Fig. 2a. There is a clear, cold water belt between 122°E and 123°E that has two cold centers. The first cold center is located at the junction of the Changjiang River Estuary and Hangzhou Bay, while the second cold center is located northeast of the Changjiang River Estuary on the west side of the LSWL. Low surface temperature reveals the spread of cold water and thus the range of upwelling (Li et al., 2018). Figure 2b shows the vertical temperature distribution along Line A, and it consists of an upper mixed layer, a thermocline layer, and a lower mixed layer. The thicknesses of the upper mixed layer and the thermocline layer are both 5–10 m, and the isotherm below the thermocline gradually rises from east to west. Similar to the 32 isohaline, the 25°C isotherm rises from east to west. Both the isotherms and the isohalines suggest that cold and saline deep water upwells along the bottom slope.

The POM${\text{-}}\sigma {\text{-}} z$ model (Zhang et al., 2011), which is developed from the Princeton Ocean Model (Blumberg and Mellor, 1987), is used in the LSWL simulation. As shown in Fig. 3a, the $\sigma $ coordinate is employed for the current calculation, and the transform is

where $H$ is the bottom topography, $\eta $ is the surface elevation, and $D = H + \eta$ is the total water depth. As shown in Fig. 3b, the $\sigma {\text{-}} z$ coordinate is used for the salinity calculation. The seawater is vertically divided into two parts, where $z = - {H_0}$ is the interface. The $\sigma $ coordinate is used in the water above the interface, and the corresponding transform is expressed as

In contrast, the $z$ coordinate is used in the water below the interface.

The Eulerian-Lagrangian method (ELM) is used to address the effect of the current on salinity. The salinity equation used in the ELM in $\sigma {\text{-}} z$ coordinates is

where $\dfrac{{{\rm{d}}S}}{{{\rm{d}}t}}$ is the total derivative of salinity, which describes the change in salinity over time. $F$ is the turbulent mixing term. The expression of $F$ is

where ${F_S}$ is the horizontal turbulent mixing term. If a water particle goes from point $P$ at time $n\Delta t$ to point $Q$ at time $(n + 1)\Delta t$, then we have the expression as

The change of salinity at point $Q$ at time $(n + 1)\Delta t$ is caused by two kinds of dynamics. One is convection. The water particle at point $Q$ at time $(n + 1)\Delta t$ is substituted by one at point $P$ at time $n\Delta t$; therefore, the salinity at point $Q$ at time $(n + 1)\Delta t$ keeps the character at point $P$ at time $n\Delta t$. The other is eddy diffusion. To determine the effect of current on the salinity calculation, the location of point $P$ must be determined.

The method of determining the location of point $P$ is as follows:

where ${u^{n + 1}}(Q)$, ${v^{n + 1}}(Q)$, and ${\omega ^{n + 1}}(Q)$ are from the linear interpolation of the current at time $(n + 1)\Delta t$. ${S^n}(P)$ is calculated from the interpolation of salinity at time $n\Delta t$. Zhu et al. (2001) advanced this method by combining first-order and second-order Lagrange interpolations, and this advanced approach is used in this paper.

The computational domain covers the Bohai Sea, the YS and the ECS (Fig. 4a). The numerical grids are shown in Fig. 4b. The minimum and maximum distances between the grids are 1 540 m and 25 000 m, respectively. In the vertical direction, the current calculation uses 20 vertical $\sigma $ layers. The salinity calculation takes place in the $\sigma {\text{-}} z$ layer, which has 8 uniform $\sigma $ layers in the upper waters and 24 $z$ layers in the lower waters. The interface between these two parts is at a depth of 25 m.

For convenience, the numerical simulation of the LSWL in summer 2006 is called CASE06, which couples the tidal currents and circulation in consideration of runoff, winds, tides (the M₂, S₂, K₁ and O₁ tidal constituents), the Taiwan Warm Current (TWC), the Kuroshio, the Tsushima Warm Current, and the baroclinic effect. The Changjiang River discharge is obtained from the monthly averaged flux at the Datong hydrometric station, which is ~640 km upstream from the river mouth. The wind force is obtained from 6-hourly NCEP/NCAR reanalyzed wind data. The climatological averages of the salinity and sea temperature in summer are used as the original field data, and the sea temperature does not change with time in the calculation. The calculation coupled the tidal current and the circulation begins on July 2, 2006, and runs simultaneous with the salinity calculation on July 9, 2006. The numerical scheme of CASE06 is similar to that of a previous study (Zhang et al., 2011), which introduced more details of the model and its validation.

Figure 5 shows part of the simulated surface salinity. Beginning on August 3, low-salinity water (less than 30) extended northeastward from the Changjiang River mouth in a tongue shape. At 2:00 on August 7, the LSWL detached from the interior side of the river plume at approximately 32°N, 122.9°E and then gradually moved northeastward. After August 10, the central position of the LSWL was nearly fixed. Finally, the LSWL disappeared at 15:00 on August 19. During August 7 to 19, the value of the central closed isohaline increased from 18 to 28 and the size of the LSWL was generally 20–85 km. The LSWL disappeared briefly several times but consistently maintained a tongue shape.

Compared with the observed characteristics, the real-case simulation of the detachment of the LSWL in August 2006 shows that the shape of the LSWL remained basically the same. Numerical experiments to further reveal the mechanism of detachment are therefore justifiable.

In this section, seven experiments (Table 1) are designed to examine the influences of winds, tides, vertical mixing, and upwelling and the baroclinic effect on the LSWL. We take CASE06 as the control test. Based on CASE06, the arrangement of CASE06a removes the wind forcing to eliminate its influence. CASE06b uses the same model settings as CASE06, although the wind is changed to a steady southerly wind of 4 m/s. For CASE06c and CASE06d, the effects of tides and vertical mixing are removed, respectively, while all other physical processes remain intact. The vertical velocity in the $z$ coordinate (${w_z}$) of CASE06e is set to 0 (the setting method is explained in detail in Section 4.3). For CASE06f, the only modification to the model configuration is setting the baroclinic pressure gradient (BPG) term to 0; and for CASE06g, the BPG term is calculated by the climatological average for the salinity and sea temperature. In addition, we also verified the roles of the Changjiang River discharge and TWC in the occurrence of detachment (not shown). To discuss the effects of runoff and the TWC, experiments were conducted in which both the Changjiang River discharge and the TWC are both doubled and halved. The results (not shown) show that Changjiang River discharge and TWC have little effect on detachment; therefore, these experiments are not included in Table 1.

According to previous studies, the formation mechanisms responsible for upwelling are diverse. Lü et al. (2006) suggested that tidal mixing plays a dominant role in inducing upwelling. Additionally, topography exerts profound influences on upwelling. Zhu (2003) studied the upwelling off the Changjiang River Estuary by conducting numerical tests and proposed that in the northern submarine valley the baroclinic effect primarily accounts for upwelling. Zhao et al. (2003) found that the upwelling off the Changjiang River Estuary was induced by the interaction between the TWC and the slope bottom topography and that the southerly wind had an important impact on the upwelling in the upper layer. Ni et al. (2014) suggested that the Coriolis force and pressure gradient force were the main forcing factors for the upwelling off the Changjiang River Estuary.

As previously discussed, four experiments (Table 2) were designed to examine the influences of the baroclinic effect, wind forcing, and tides on upwelling during the evolution of the LSWL. The CASE06h experiment was designed to analyze the influences of the baroclinic effect on the upwelling, and the same model settings for CASE06 were used for CASE06h except that the salinity of this experiment was fixed and the BPG was set to 0. To analyze the influences of the wind forcing and tides on upwelling and to avoid the indirect effects of salinity changes, the arrangements of CASE06i and CASE06j, which were based on CASE06h, eliminated the wind forcing and tide components, respectively. For CASE06k, the effects of all dynamic factors except tides were removed. These four experiments were used to simulate the currents during August 2 to 21.

By combining observed temperature and salinity data, we found an isolated LSWL offshore from the Changjiang River Estuary in August 2006. The LSWL isolated by the 22 isohaline had an elliptical shape, and the value of the center closed isohaline was 20 psu. The vertical structures of the salinity and temperature along Line A indicated that cold and saline deep water upwells along the bottom slope.

In this study, the POM${\text{-}}\sigma {\text{-}} z$ model was used to simulate the LSWL in the summer of 2006 and the simulated results indicate that the LSWL became detached at the interior side of the river plume and persisted during August 7 to 19. At the beginning of the detachment, the LSWL had a low salinity, although as the LSWL moved offshore, the central salinity value increased due to mixing with high-salinity water.

Winds played important roles in the formation and evolution of the LSWL. During August 2 to 6, the persistent southeasterly wind induced the northwestward expansion of the LSWL. ERC vectors, which imposed an uneven northward force on the CDW, showed differences between the nearshore and far-shore detachment regions. Strong upwelling occurred along the Jiangsu coast and off the Changjiang River Estuary, and it carried high-salinity water to the surface layer and strengthened vertical mixing, which was favorable for the detachment of the LSWL. During August 7 to 11, with the continued easterly wind, the northeastward ERC in the detachment region was conducive to driving the LSWL offshore. A new upwelling area appeared along the 30 m isobath, which stabilized the LSWL. During August 12 to 16, the light wind in the study region weakened the ERC and the upwelling; hence, the LSWL remained steady. After August 17, the upwelling disappeared and the strong southwestward ERC induced by the strong northerly wind drove the LSWL onshore, which caused the LSWL area to decrease gradually until it disappeared.

Numerical experiments showed that tides are a necessary condition for the formation of LSWLs. Zhang et al. (2014a, b) indicated that tide-induced vertical mixing gradually strengthens from neap tide to spring tide, which is helpful for forming a LSWL. However, an interesting phenomenon presented herein is that the LSWL did not disappear when vertical mixing ceased. These results indicated that the tide affects detachment by physical processes other than tidal mixing.

Upwelling is crucial in detachment events. Furthermore, the influences of three dynamic mechanisms on upwelling, namely, the baroclinic effect, wind forcing and tides, were also discussed. The baroclinic effect played only a limited role in the upwelling along the Jiangsu coast and in the northeast sea off the Changjiang River Estuary. The persistent southeasterly wind during August 2 to 6 induced strong upwelling near the Changjiang River Estuary and along the Jiangsu coast; however, the direction and intensity of the wind during August 7 to 21 were not conducive to inducing upwelling. Tide-induced upwelling gradually increased during the transition from neap tide to spring tide. The bottom ERC near the coastal area with depths shallower than 30 m flowed offshore, and in the sea area north of 32°N, the bottom ERC flowed onshore across the 30 m isobath at spring tide. The tidal residual currents converged at the 30 m isobath, and the bottom topography provided an inclined plane that the bottom ERC could climb.

The distribution of nutrients and dissolved oxygen is greatly affected by LSWLs. In particular, detachment will change the local ecological environment. Our work can therefore provide theoretical suggestions for the protection of ecological environments and aquaculture in the Changjiang River Estuary. However, there are still many unclear aspects; for example, the TWC and Changjiang River discharge may affect the upwelling near the Changjiang River mouth. Therefore, more numerical simulations are needed for further study.