The ECS is one of the major marginal seas located in the temperate northwestern Pacific (Figure 1). It covers a surface area of 1.25 × 10⁶ km², with >70% of continental shelf shallower than 200 m (Wang et al., 2000). Changjiang, the fourth largest river in the world, delivers 940 km³ of freshwater (Dai and Trenberth, 2002), 300 × 10⁴ t of N, 8 × 10⁴ t of P, and 100 × 10⁴ t of Si to the ECS annually (Zhang et al., 2007; Gao et al., 2009; Li et al., 2014). The ECS is connected to the YS in the north (Figure 1). The climate is modulated by the East Asian monsoon, and the Kuroshio Current and Changjiang plume mainly dominate mixing on the ECS shelf. The Changjiang plume flows northeastward in summer but southwestward along the China Mainland coastline in winter (Lee and Chao, 2003). The northward flowing Kuroshio follows the isobaths beyond the shelf break at ∼200 m (Lee and Chao, 2003; Liu and Gan, 2012). On the northern ECS shelf, the Yellow Sea Coastal Current (along the coast of Jiangsu Province) flows southward year-round, which brings YS water to the ECS, except in summer, under the influence of the strong southwest monsoon (Yuan et al., 2017).

The sea surface temperature (SST) in the ECS is low in winter and early spring but high in summer and early fall (Gong et al., 2003; Guo et al., 2015). In addition to the Changjiang input, upwelling of Kuroshio subsurface water is also an important nutrient source to the ECS shelf (Chen and Wang, 1999). In general, productivity on the ECS shelf is low in winter due to cold temperatures, but high (primary production up to > 1 g C m^–2 d^–1) during warm seasons, especially in the Changjiang plume (Gong et al., 2003). Regulated by temperature and phytoplankton productivity, the partial pressure of CO₂ (pCO₂) shows strong seasonal and spatial variations. On the outer ECS shelf, the seasonal variability of surface water pCO₂ is dominated by seasonal variations in SST (Guo et al., 2015), while in the Changjiang plume, freshwater discharge and biological productivity mainly control the surface water pCO₂ and air-sea CO₂ fluxes (Tseng et al., 2014). Overall, pCO₂ variability on the ECS shelf is characterized by low values (under-saturation with respect to atmospheric CO₂) during cold seasons (temperature-dominated) and in productive areas during warm seasons (biology-dominated), while high pCO₂ values are observed on the outer shelf in summer (temperature-dominated) (Chou et al., 2009a,2011; Tseng et al., 2014; Guo et al., 2015).

In summer, in the productive Changjiang plume, strong stratification occurs with low pCO₂ and high dissolved oxygen (DO) concentrations in the surface water vs. high pCO₂ and low DO concentrations in the subsurface water (Chou et al., 2009b). Hypoxia in the bottom water occurs in summer in this area (Li et al., 2002; Wang et al., 2012). Strong remineralization of marine-sourced organic matter is the major cause of the summer bottom hypoxia (Wei et al., 2007; Wang et al., 2016, 2017), although remote drivers, including reduced DO input from Kuroshio Intermediate Water to the ECS bottom water, also play a role (Lui et al., 2014; Qian et al., 2017). Hypoxia in bottom water is accompanied by enhanced acidification in summer than the impact of atmospheric CO₂ invasion (Cai et al., 2011).

During December 20–30 of 2015 (winter cruise), August 3–13 of 2016 (summer cruise), November 30–December 4 of 2018 (late fall cruise), and April 29–May 5 of 2019 (spring cruise), 4–7 transects were visited on the ECS shelf off the Changjiang estuary. At each station, salinity and temperature-depth profiles were recorded with a Seabird^® SBE 917 Conductivity-Temperature-Depth/pressure (CTD) sensor package (Sea Bird Electronics, Washington, United States). At each station, discrete water samples were collected with 5-L Niskin bottles mounted on a rosette sampler.

Subsamples for DO, dissolved inorganic carbon (DIC), and total alkalinity (TA) were taken with Tygon^® tubing free of air bubbles, with ample sample overflow in order to minimize any contamination from atmospheric O₂ or CO₂. DO samples were taken with 60 mL biological oxygen demand bottles and fixed with Winkler reagents (Carpenter, 1965). Samples for DIC and TA measurements were taken into 250 mL borosilicate bottles with ground-glass stoppers and poisoned with 200 μL of saturated HgCl₂ solution, then stored in the dark until analysis.

Dissolved Oxygen samples were measured with the titration method (Dickson and Goyet, 1994) using a Micrometer Burette (IA281 ABC, 2 mL, Gilmont instruments, IL, United States). A standard KIO₃ solution provided by the Standard Material Center of the Second Institute of Oceanography (Ministry of Natural Resource of the People’s Republic of China) was used to standardize the measurements. The precision of the DO measurements was better than ±2 μmol kg^–1. Analyses of DIC and TA were done following the methods of Cai et al. (2004). DIC was measured by collecting and quantifying the CO₂ released from the sample upon acidification with a non-dispersive infrared detector (Li-Cor 7000) using a DIC Analyzer (Apollo SciTech model AS-C3; Apollo SciTech, DE, United States) with a precision of better than ± 2 μmol kg^–1. TA was determined by Gran titration with hydrochloric acid using an automated Alkalinity Titrator (Apollo SciTech model AS-ALK1+; Apollo SciTech, DE, United States) with a precision better than ±2 μmol kg^–1. Both DIC and TA were calibrated with the certified reference material provided by Dr. A. G. Dickson of the Scripps Institution of Oceanography to an accuracy of better than ±2 μmol kg^–1.

The saturation state index of CaCO₃ (Ω) was defined as the product of concentrations of Ca²⁺ and carbonate (CO₃^2–) ions divided by the apparent solubility product of CaCO₃ (K_sp*) (Equation 1).

The K_sp* of CaCO₃ (aragonite and calcite) was calculated according to Mucci (1983). The value, Ω > 1 favors precipitation, while Ω< 1 favors dissolution. As aragonite is more soluble than calcite and their distribution patterns are similar (Millero, 2013), we only report the saturation state index of aragonite (Ω_Ar) in this study.

The program CO2SYS (Version 14, Pierrot et al., 2006) was used to calculate the Ω_Ar, pH at 25°C and at in situ temperatures at the total hydrogen scale (pH₂₅ and pH_in–situ hereafter), and CO₃^2– concentrations. The dissolution constants of carbonic acid are derived from Millero et al. (2006); the CO₂ solubility coefficient is derived from Weiss (1974); the sulfate dissociation constant is derived from Dickson (1990). The relationship of total boron with salinity is derived from Uppstrom (1974), and the PO₄^3– and SiO₂ data are derived from Guo et al. (unpublished data).

DIC, pH₂₅, and Ω_Ar at CO₂ equilibration with the atmosphere were calculated from TA and atmospheric pCO₂ in cold seasons using the CO2SYS program (Version 14, Pierrot et al., 2006). The constant selections were the same as used in the calculations above. Excess DIC (DIC_excess) is defined as the difference between the observed DIC (DIC_obs) concentration and the DIC concentration when CO₂ is in equilibrium with the atmosphere (DIC_equ) (Cai et al., 2020). Similarly, the pH₂₅ and Ω_Ar changes due to biogeochemical processes (ΔpH₂₅ and ΔΩ_Ar) are defined as the differences between the observed values and the values at equilibrium with atmospheric CO₂. The average atmospheric pCO₂ (390 μatm) during the late fall and winter cruises was taken as the atmospheric pCO₂ during cold seasons before water column stratification (Guo et al., unpublished data).

In the above equations, the subscript “equ” means CO₂ at equilibrium with the atmosphere.

Apparent oxygen utilization (AOU) is defined as the difference between saturated oxygen concentrations (DO_sat) and the observed DO (DO_obs) concentrations. DO_sat was calculated according to the empirical formula of Benson and Krause (1984).

Freshwater discharge from the Changjiang varied seasonally between high discharge (up to > 40,000 m³ s^–1) in summer and low discharge (<20,000 m³ s^–1) in winter, with spring and fall as transitional seasons. During our summer and winter cruises, Changjiang water discharge rates were 40 and 63% higher than the long-term averages, respectively, while discharge rates during the spring and late fall cruises were similar to those of the long-term averages (Figure 2). River discharge has a large influence on the seasonal variability and spatial variations of primary production and CO₂ parameters on the ECS shelf (Gong et al., 2000, 2006; Tseng et al., 2014).

Generally, SST was the lowest in winter and the highest in summer, with spring and fall as transitional seasons. The Changjiang freshwater discharge and the monsoon-dominated coastal current influenced the seasonal variations in sea surface salinity (SSS), with high values in winter but relatively lower values in summer.

In winter, salinity ranged from 25.6 to 34.5 and water temperatures ranged from 10.7 to 20.2°C. Lower temperatures were measured in the nearshore area and temperatures increased southeastward (Figure 3A-1). As winter was dominated by a strong northeastern monsoon, the river plume was oriented southwest along the China Mainland coast. Salinity increased southeastward and the highest values were observed in the southeastern area (Figure 3A-2), indicating the influence of the Kuroshio Current. Vertically, the water was well-mixed over almost the entire inner ECS shelf. Both temperature and salinity increased southward and eastward (Figures 4A-1,B-1).

In spring, SSTs increased to 16.8–24.9°C with higher values in the southeastern area (Figure 3B-1). The plume extended offshore and covered a larger area than in winter. SSS decreased in the area off the Changjiang Estuary and the inshore area to less than 32. A high salinity greater than 33 was observed only in the southeastern area (Figure 3B-2). Vertically, the water was partially stratified. The temperature of the surface water was 17.5–18.1°C, which was conspicuously higher than the subsurface water (<17.5°C, Figure 4A-2). Salinity was low in the whole water column in the inshore area (<30), but was stratified in the offshore area where salinity was 30–32.5 at the surface and increased with depth to 34.2–34.6 at 20 m (Figure 4B-2).

In summer, SST further increased from 24.1 to 30.3°C, with lower values near the Changjiang Estuary (Figure 3C-1). SSS further decreased in the northern ECS with the lowest value of 18.6 observed near the Changjiang Estuary. The ECS water was strongly stratified with a high salinity of 33.5–34.6 in the bottom beyond the Estuary (Figure 5B-1). The temperature of the bottom water was 10.4–24.5°C, which was generally much lower than the surface water (Figure 5A-1). The relatively low temperature of the bottom water in the center of the northern ECS was consistent with the observations of Wang and Chen (1998), who suggested that it was a remnant winter water.

As our late fall cruise was conducted in late November and early December, the hydrologic characteristics were similar to those in winter but with relatively higher temperatures and weak stratification (Figures 3D-1,D-2, 5A-2,B-2). Temperatures ranged from 15.2 to 21.3°C and salinity ranged from 20.6 to 33.9 during our late fall cruise.

On the ECS shelf, pH₂₅ and Ω_Ar decreased along with a decreasing ratio of TA/DIC throughout the whole water column during all seasons (Figures 6A,B). From Equation 1, Ω_Ar is dominated by the concentrations of Ca²⁺ and CO₃^2– and K_sp*. On the ECS shelf, Ca²⁺ concentrations and K_sp* increased synchronously along with the decrease in the temperature and increase in the salinity and/or depth. Using the summer dataset as an example, the relative variability of K_sp* was found to be 13.7%, and the relative variability of Ca²⁺ concentration was 11.1%. As Ca²⁺ concentration and K_sp* increase or decrease synchronously, the relative variability of [Ca²⁺]/K_sp* ratios were as low as 3.4%. However, the relative variability of [CO₃^2–] was 38.4%. Therefore, the [Ca²⁺]/K_sp* ratio was relatively invariant compared to the variability of [CO₃^2–], and the variability of Ω_Ar depended on the variability of [CO₃^2–]. Additionally, [CO₃^2–] positively correlates with the TA/DIC ratio (Figure 6C). Therefore, Ω_Ar showed a very good positive correlation with the TA/DIC ratio (R2 = 0.99) (Figure 6B).

The linear relationship between [CO₃^2–] and TA/DIC ratios in the ECS is not unique. Similar cases were reported in the Gulf of Mexico and along the eastern and western coasts of the United States (Wang et al., 2013; Cai et al., 2020). As TA/DIC ratios are indicators of the relative abundances of the carbonate species, they imply the buffering capacity of the carbonate system. Buffering capacity is the lowest when TA/DIC is ∼1. Therefore, variations in the TA/DIC ratio provide a direct indicator of changes in the sensitivity of seawater pH, Ω_Ar, and other properties of carbonate system perturbations (Wang et al., 2013; Cai et al., 2020).

We will discuss the processes hereafter from the view of the relative changes in the TA/DIC ratio. Processes or factors influencing the TA/DIC ratio (and therefore pH₂₅ and Ω_Ar) include air-sea CO₂ exchange, temperature variations, inputs of river water, or YS Coastal Water (water mixing), and biological processes.

During the relatively low productivity of late fall and winter (Gong et al., 2000), the TA/DIC ratio ranged from 1.07 to 1.14. However, the DIC additions due to remineralization of organic matter in the bottom water decreased the TA/DIC ratio to 1.04 during summer (Figure 6). As discussed above, the strong remineralization in the bottom water enhanced acidification. In the bottom water of the ECS, the pH₂₅ minimum was 7.73, and the Ω_Ar minimum was 1.27. An Ω_Ar of 1.5 is considered to be the minimum threshold for marine shellfish development (Ekstrom et al., 2015; Waldbusser et al., 2015). Therefore, Ω_Ar in ECS bottom water in summer was below this threshold. However, the lowest Ω_Ar value in the ECS was higher than the values in the adjacent YS, where Ω_A may be close to 1 (Xiong et al., 2020). Compared with other marginal seas, pH₂₅ and Ω_Ar values in the bottom water in summer were similar to those of the surface waters of fragile Polar Regions with regards to OA (Sabine et al., 2004; Anderson et al., 2010).

Within the context of global CO₂ increases and ocean acidification, Ω_Ar in the ECS will continuously decrease. Using this information, we can estimate pH₂₅ and Ω_Ar in ECS shelf bottom water in the near future (by the mid-century) under the influence of the same biogeochemical processes as observed during our surveyed periods.

According to the IPCC (2014) prediction, atmospheric pCO₂ will increase to ∼500 μatm (PR6, medium emissions) or 550 μatm (PR8.5, high emissions) by the middle of this century. Assuming that salinity and TA remain invariant, we first calculate the DIC_equ under the two scenarios (PR6 and PR8.5) and then add the DIC_excess to DIC_equ to get the DIC under the influence of organic carbon remineralization. Subsequently, pH₂₅ and Ω_Ar were calculated with CO2SYS (Pierrot et al., 2006).

The results show that by mid-century, the pH₂₅ and Ω_Ar of the bottom water in summer will be 7.46–7.82 and 0.94–1.99, respectively, under the medium emissions scenario, and 7.41–7.78 and 0.84–1.83, respectively, under the strong emissions scenario (Figure 13). The values of pH_in–situ will be 7.51–7.90 (medium) and 7.45–7.86 (high) under two emissions scenarios. In spring, the lowest values of Ω_Ar will be 1.48 (medium) and 1.36 (high) under the two scenarios (Figure 13), slightly higher than in summer. The results suggest that under the combined influences of atmospheric CO₂ invasion and strong seasonal organic matter remineralization, aragonite in the ECS will be under-saturated in summer and slightly oversaturated in spring within 30 years (by the middle of this century). The conceptual model of the combined forcing of atmospheric CO₂ invasion and strong organic matter mineralization on the OA in the bottom water on the ECS shelf is shown in Figure 14.

The projected pH₂₅ and Ω_Ar minima in spring would be 7.71–7.75 and 1.36–1.48, respectively. These projections are very conservative due to the following factors. First, the surveyed area of our spring cruise was beyond the most productive areas of the Changjiang Estuary and plume and hence experienced weaker organic matter remineralization in the bottom water. Second, the influence of potentially enhanced eutrophication occurring from the decadal variability of KIW (Lui et al., 2014) was not considered. Therefore, it is reasonable to speculate that aragonite might be under-saturated in the bottom water from spring to summer by the middle of this century. As discussed above, Ω_Ar is not highly sensitive to temperature changes. Although Ω_Ar increases slightly under global warming, it cannot offset the effect of OA. The enhanced acidification observed in this study is more severe and the calculated aragonite saturation levels occur two decades earlier than that predicted by Chou et al. (2013). This can be attributed to the fact that the summer of 2015 was in El Nino, resulting in more severe DO depletion and acidification.