The SYS is located on the eastern shelf of China, covers an area of 3.09 × 10⁵ km², and has an average depth of 44 m (mostly shallower than 100 m). It is adjacent to the North Yellow Sea to the north and is also adjacent to the East China Sea to the south. The SYS hydraulic residence time, based on ²²⁸Ra/²²⁶Ra measurements, has been estimated to be 5–6 years (Nozaki et al., 1991).

Climatic variations of the SYS are primarily dominated by the East Asian Monsoon (Chen, 2009). The rain-bearing southwest monsoon lasts from June to September, while the strong northeast monsoon prevails in winter, from November to March of the next year (Zhai et al., 2014). The winter circulation system consists of southward coastal currents and the northward Yellow Sea Warm Current (YSWC), while summer hydrography is characterized by the SYS cold water mass (SYSCWM). In winter and early spring, the Yellow Sea Coastal Current includes the LuBei Coastal Current and SuBei Coastal Current (Figure 1A). The LuBei Coastal Current is formed by the Bohai Sea outflow water (Chen, 2009; Zhai et al., 2014), which is transported outwards through the top of Laizhou Bay and into the North Yellow Sea. In the SYS, the coastal current from the north mixes with runoffs from the coast of China, forming the SuBei Coastal Current (Figure 1A). The YSWC originates from the Tsushima Warm Current, which is a branch of the Kuroshio. On the southwestern side of Jeju Island, it meets with the East China Sea shelf water and extends into the Yellow Sea (Figure 1A). The YSWC is a compensating current of the monsoon-driven Yellow Sea Coastal Current (Hsueh, 1988; Chen, 2009) and is the only passageway that introduces open-ocean waters and the Pacific Decadal Oscillation signals into the Yellow Sea (e.g., Li et al., 2022). It appears in late autumn and winter, recedes in late spring, and disappears in summer. In summer, a cold pool (i.e., the SYSCWM) develops below the thermocline in the central SYS from late spring to autumn (Yu et al., 2006), affecting biogeochemical processes there.

The SYSCWM is the remnant of winter cooling water (temperature < 12°C, salinity > 32), overlain by 20–25 m of warm water (Yu et al., 2006). In the SYSCWM, DO, pH_T, and Ω_arag decline from spring to autumn (Zhai, 2018; Xiong et al., 2020). In the central SYS, a springtime algal bloom develops every year (Tan and Wang, 2014), producing many biogenic particles. Field surveys along the 35°N section have shown that the particulate organic carbon (POC) in the central SYS varies from 140–150 μg·L⁻¹ in spring to 60–140 μg·L⁻¹ in summer to 80–100 μg·L⁻¹ in autumn and to 40–60 μg·L⁻¹ in winter (Cheng, 2011). The springtime biogenic particles presumably settle quickly and support subsequent oxygen consumption and CO₂ production in the SYSCWM in summer and autumn (Zhai, 2018; Xiong et al., 2020). According to Song et al. (2016), approximately 90% of the organic carbon obtained through the primary production process in the Yellow Sea is ingested by zooplankton and/or decomposed by microorganisms in the water column, and only 4% or even less may be buried in sediments.

Four field surveys were conducted in the SYS in 2019 (Figure 1), comprising spring and summer cruises (in April and in August, respectively, both on board the R/V Beidou) and two autumn cruises (in October on board the R/V Lanhai 101 and in November on board the R/VKexue III). Temperature and salinity in the water column were recorded with a calibrated conductivity, temperature, and depth/pressure (CTD) recorder (SBE 25-Plus on board the R/V Beidou, SBE 911 on board the R/V Lanhai 101, and SBE 911-Plus on board the R/V Kexue III). Water samples were collected to determine the dissolved oxygen (DO), DIC, and total alkalinity (TAlk) in the water column at two-to-four different depths using a rosette sampler fitted with 5 or 10 L Niskin bottles and mounted with the CTD units.

DO samples were collected, fixed, and titrated on board following the Winkler procedure (Knap et al., 1996), and 0.01% sodium azide (NaN₃) was added to eliminate the interference of nitrite in the water (Wong, 2012). The uncertainty of DO data was estimated to be <0.5% (Zhai et al., 2012). DO saturation (DO%)—the ratio of the measured DO concentration to the DO concentration at equilibrium with the atmosphere—was calculated as in Benson and Krause (1984) from temperature, salinity, and field atmospheric pressure. To quantify the effect of the net community metabolism, apparent oxygen utilization (AOU) was defined as the difference between the air-equilibrated DO concentration and the measured DO. When ignoring the effects of air–sea exchange and water mixing, an AOU >0 implies net community respiration, while an AOU <0 implies net community production.

Water samples for DIC and TAlk were stored in 60 ml screw-top borosilicate glass bottles (for DIC, bubble free) and 140 ml screw-top high-density polyethylene bottles (for TAlk), and 50 μl saturated HgCl₂ was added immediately. The bottles were sealed and preserved in the dark at room temperature and finally brought back to the laboratory for analysis (Huang et al., 2012). DIC was measured by a non-dispersive infrared CO₂ detector-based DIC analyzer (AS-C3; Apollo SciTech Inc., Newark, Delaware, USA) and TAlk was determined at 25°C by the Gran acidimetric titration (Gran, 1952) using a semi-automated titrator (AS-ALK2; Apollo SciTech Inc., USA). During DIC and TAlk determinations, certificated reference materials (CRMs) from A.G. Dickson’s laboratory were used for quality assurance at a precision of ± 2 μmol kg⁻¹ (Zhai et al., 2014).

To further assess the data quality of these carbonate parameters, we also collected water samples for pH analyses in 140 ml brown borosilicate glass bottles, preserved with 50 μl saturated HgCl₂. These samples were preserved at 25°C and measured within 6 h of sampling using a precision pH meter equipped with an Orion^® 8102BN Ross electrode against three standard buffers on the National Bureau of Standards scale (pH_NBS = 4.01, 7.00 and 10.01 at 25°C; Thermo Fisher Scientific Inc., Waltham, Massachusetts, USA). These pH_NBS data were only used for the quality assessment of our carbonate data and calculation (see below).

To better characterize seasonal carbonate dynamics in the SYS, we incorporated the historical data obtained in 2018 (Xiong et al., 2020) in the study. The 2018 data comprised a spring cruise in April, a late spring cruise in May, a summer cruise in July, and an autumn cruise in October.

Both Ω_arag and pH on the total hydrogen ion scale (pH_T) were calculated from seawater temperature, salinity, DIC, and TAlk data using the CO2SYS.XLS (Version 24) program (Pelletier et al., 2015), an updated version of the original CO2SYS.EXE program (Lewis & Wallace, 1998). The Millero et al. (2006) dissociation constants of carbonic acid were used in the calculation because they cover much broader applicable ranges of temperature (0°C–50°C) and salinity (0–50), and the dissociation constants for the HSO₄^– ion were taken from Dickson (1990). The values of K_sp*_arag followed Mucci (1983), while the Ca²⁺ concentrations were assumed to be proportional to salinity (Millero, 1979). In June, August, and November 2011, Ca²⁺ concentrations were measured in the Yellow Sea (Qi, 2013). The comparison between measured Ca²⁺ data and calculated Ca²⁺ values showed that salinity-derived Ca²⁺ concentrations were slightly lower than real data by 0%–2% (Zhai et al., 2014).

To further assess the quality of our carbonate system data, we also calculated pH_NBS based on field-measured DIC and TAlk values and compared the field-measured pH_NBS data with the calculated pH_NBS data. Most measured and calculated values were consistent at a deviation level of ± 0.05 pH (Figure 2A). In addition, TAlk–DIC ratios were predominantly between 1.05 and 1.15 (Figure 2B), in line with the seawater TAlk–DIC ratio in the mid-latitude regions (Zhai & Zhao, 2016). Both comparisons suggest that our measurements and calculation of the carbonate system parameters were reliable.

To quantify the effect of the net community metabolism on DIC, following Zhai (2018) and Xiong et al. (2020) we calculated the air-equilibrated DIC (DIC^equ, corresponding to a mean air-equilibrated fugacity of CO₂, f CO 2 air value of 415 μatm) from corresponding field-measured seawater temperature, salinity, and TAlk:

The DIC departure from the DIC^equ was defined as excess DIC:

The relationship between the excess DIC and the net community metabolism is similar to the relationship between AOU and the net community metabolism. That is, an excess DIC > 0 means net community respiration, while an excess DIC < 0 means net primary production.

Cruise-based distributions of seawater temperature in the SYS showed a remarkable seasonal variation (Figures 3a–h). From spring to summer, the sea surface temperature rose from 9.6–12.3°C in April to 22.0–28.8°C in August, while it dropped to 20.0–23.0°C in October and to 14.3–19.9°C in November. In contrast, the bottom-water temperature increased moderately from 10.1 ± 1.1°C in April to 16.6 ± 5.4°C in November. The vertical gradient of temperature in summer and mid-autumn brought about the development of a seasonal thermocline (Figure S1).

Salinity exhibited a small seasonal variation with mean values of 32.5 ± 0.5 in April, 31.6 ± 1.0 in August, 31.7 ± 0.7 in October, and 31.8 ± 0.5 in November (Table 1). In spring, relatively high salinity values (salinity > 32) dominated the whole SYS. In the central SYS, a high-salinity water tongue (salinity > 33) occurred in April (Figure 3m), indicating the YSWC. In summer, the relatively high salinity values of >32 dominated bottom waters in the central SYS (Figure 3n), where the summertime SYSCWM developed. The relatively low surface salinity (Figure 3j) may be caused by the abundant rainfall in this area from late spring to summer and/or by the Changjiang diluted water that flows toward Jeju Island, South Korea. In autumn, the salinity tended to be vertically mixed as the weak halocline that appeared in summer had almost disappeared in this transitional season.

In April, sea surface DO values varied between 269 and 359 μmol·kg⁻¹ (Figure 3q) and were always oversaturated. The highest value of surface DO saturation of 130% was observed in the central SYS (Figure 3y). In bottom waters in April, however, DO was determined at 274 ± 10 μmol·kg⁻¹, at ~100% saturation (Figure 3C). Vertical profiles of DO saturation (Figure S1) implied that the surface primary production affected approximately the upper 20 m. From August to November, sea surface DO concentration changed with temperature, but roughly at the air-equilibrated levels (Figures 3z, A, B). In the SYSCWM, bottom-water DO declined from 197 ± 20 μmol·kg⁻¹ in August to 154 ± 17 μmol·kg⁻¹ in November (Figures 3v–x). Nevertheless, the lowest DO value of 69 μmol·kg⁻¹ (with DO% of ~ 32%) was observed in the southern SYS (Figures 3v, D) that was affected by the Changjiang Diluted Water in August (Figures 3j, n). This DO value approached the critical value of hypoxia (i.e., DO < 63 μmol·kg⁻¹ or DO% < 30%).

For a quasi-conservative carbonate system parameter, survey-averaged TAlk in the SYS ranged between 2286 ± 55 μmol·kg⁻¹ in August and 2320 ± 40 μmol·kg⁻¹ in November. TAlk generally exhibited a descending trend from north to south and from nearshore areas to the central basin (Figures 4a–h). In contrast to TAlk, DIC was largely affected by metabolic processes. It was also subject to SYS circulation and the effect of temperature on the solubility of CO₂, showing striking seasonal variations (Figures 4i–p). Sea surface DIC varied between 2,044 and 2,229 μmol·kg⁻¹ in April, while the average DIC in bottom waters was 2139 ± 30 μmol·kg⁻¹. From August to November, relatively high DIC values of 2,170–2,284 μmol·kg⁻¹ were observed in the bottom waters of the SYSCWM.

Figure 4 shows cruise-based distributions of pH_T and Ω_arag in the SYS. In April, surface pH_T ranged from 7.92 to 8.22, while the corresponding Ω_arag was 1.76–2.80. In bottom waters, the April pH_T average was 8.04 ± 0.04, and the corresponding Ω_arag was 1.95 ± 0.13. From August to November, surface pH_T ranged between 7.90 and 8.17, slightly lower than that in April. However, the corresponding surface Ω_arag exhibited its maximum values (2.32 – 4.19) in August. In the SYSCWM, the average bottom-water pH_T dropped from 7.88 ± 0.04 in August to 7.84 ± 0.04 in November, while the corresponding Ω_arag dropped from 1.42 ± 0.12 to 1.32 ± 0.09. The lowest Ω_arag of 1.15 was observed in November at a deep-water station in the central SYS, while the lowest pH_T of 7.77 appeared in August at a station connecting the SYS and the northern East China Sea.

In April 2019, surface excess DIC ranged between –85 μmol·kg⁻¹ and –34 μmol·kg⁻¹. The most negative surface excess DIC values of –85 to –69 μmol·kg⁻¹ were located in the central SYS (Stations H14 – H17 in Figure 1B). The corresponding AOU ranged from –82 to –46 μmol·kg⁻¹. In April 2018, surface excess DIC ranged between –55 μmol·kg⁻¹ and –22 μmol·kg⁻¹. The most negative surface excess DIC values of –55 to –40 μmol·kg⁻¹ were also observed in the central SYS, with corresponding AOU values of –56 to –37 μmol·kg⁻¹. The negative excess DIC versus negative AOU on the surface roughly followed the traditional Redfield ratio (Figures 5A, B), showing that primary production synchronously modulated the carbonate system and DO on the sea surface. A small deviation of surface excess DIC : AOU from the Redfield ratio was observed in April 2019 (Figure 5A), presumably due to the slower air–sea re-equilibration of CO₂ than that of DO. This is because the marine carbonate system has a substantial chemical buffering capacity, while DO lacks any chemical buffering effect (Zeebe & Wolf-Gladrow, 2001).

From spring to autumn, excess DIC versus AOU in the bottom waters of the SYSCWM generally followed the traditional Redfield ratio (Figures 5C, D), suggesting that net community respiration dominated both DIC increase and DO decline. In 2018, the SYSCWM synchronously accumulated excess DIC and AOU by 0.32 and 0.47 μmol·kg⁻¹·d⁻¹, respectively (Figures 6A, B), from April to October. In 2019, however, the seasonal accumulations of excess DIC and AOU were not constant over time.

Excess DIC accumulated rapidly in the SYSCWM from mid-April to mid-August in 2019. During this period, excess DIC accumulated at a rate of 0.52 μmol·kg⁻¹·d⁻¹ [i.e., (60 – (–6))/127 = 0.52, Figure 6A], and the converted AOU accumulation rate was 0.68 μmol·kg⁻¹·d⁻¹ (based on the ratio of excess DIC : AOU = 106:138), approximately the same as the earlier reported thr DO consumption rate (ranging from 0.6 to 1 μmol·kg⁻¹·d⁻¹) in the North Yellow Sea cold water mass (Zhai et al., 2014). From mid-August to late November 2019, however, both the excess DIC and AOU accumulation rates declined (Figures 6A, B). The accumulation rate of excess DIC was estimated to be 0.17 μmol·kg⁻¹·d⁻¹ in the second half of 2019 [i.e., (77 – 60)/98 = 0.17, Figure 6A]. Correspondingly, the converted DO consumption rate was only 0.22 μmol·kg⁻¹·d⁻¹, one-third of that from mid-April to mid-August. In general, the SYSCWM exhibited a similar pattern of monthly variations in bottom-water excess DIC and AOU in the two years, although the surveying months were not identical between 2018 and 2019 (Figures 6A, B).

In connection with the strong spring primary production at the surface of the central SYS, the algae blooms in spring provided necessary biogenic debris for net community respiration in the SYSCWM in summer and autumn, making the quasi-synchronous accumulation of AOU and excess DIC in the SYSCWM. In 2019, the negative values of surface excess DIC and AOU were more remarkable in the SYS in comparison with those recorded in 2018 (Figures 5A, B). This may imply that the intensity of spring primary production in 2019 was stronger than that in 2018. The intense primary production in April 2019 may have introduced more biogenic debris to subsurface waters, inducing relatively higher decomposition rates of organic matter and releasing more CO₂ (coupled with greater oxygen consumption) in the bottom waters. From spring to autumn, however, earlier intense community respiration may have led to excessive consumption of degradable biogenic debris, weakening later community respiration and leading to a decrease in the accumulation rate of AOU (Figure 6B).

Figure 6 also shows seasonal variations in cruise-averaged bottom-water Ω_arag and pH_T in the SYSCWM. From mid-April to mid-August 2019, Ω_arag rapidly dropped by 0.55 units, while pH_T decreased by 0.17 units (Figures 6C, D). The corresponding rates of decline were estimated to be 0.0043 Ω_arag·d⁻¹ and 0.0013 pH·d⁻¹. From mid-August to late November 2019, however, the rates of decline were only 0.0010 Ω_arag·d⁻¹ and 0.0005 pH·d⁻¹. Since the net community respiration rate in the SYSCWM decreased from early spring to late autumn, the acidification indexes also exhibited a changing scheme of first decreasing rapidly and then slowing. In this study, logarithmic fitting curves were applied to characterize the Ω_arag and pH changes within the two years (Figures 6C, D).

The progress of acidification in the SYSCWM in 2018 was slightly different from that in 2019. Both the interannual variation in primary production (affecting sea surface hydrochemistry in April and May) and the early spring water temperature affected the annual initial Ω_arag values in subsurface waters. In this study, the annual initial average bottom-water Ω_arag of 1.82 at a temperature of 7.5°C was observed in April 2018, lower than that of 1.97 at a temperature of 9.9°C in April 2019. After 1 month of net community respiration, in May 2018, the average bottom-water Ω_arag had dropped to 1.63. In several deep stations (>70 m), bottom-water Ω_arag was observed to be as low as 1.42, even lower than the critical threshold of CaCO₃ dissolution of 1.5 in the Yellow Sea. For May 2019, we predicted a possible average bottom-water Ω_arag value based on the initial value in April 2019 and the rate of Ω_arag decline from mid-April to mid-August 2019. The Ω_arag was estimated to be 1.82 in May 2019, much higher than the data in May 2018. It was also much higher than the critical threshold of CaCO₃ dissolution of 1.5.

Because of the different annual initial Ω_arag, the appearance of CaCO₃-unsafe seawater occurred earlier in 2018 than in 2019. Benthic shellfish in the SYSCWM were exposed to CaCO₃-unsafe seawater since as early as May in 2018. However, the annual initial pH_T (in situ) values were approximately the same in both years (data not presented in this study). This is because, in an open system that re-equilibrates with atmospheric CO₂, the pH_T (in situ) changes little following changes in temperature (Cai et al., 2020; Xiong et al., 2020).

In the SYS, a strong thermocline remains stable in summer and autumn (e.g., Zhai, 2018; Guo et al., 2020; Xiong et al., 2020). In a recently published carbonate-related study, Wang and Zhai (2021) showed monthly variations in the density difference between surface and bottom layers (Δρ) in the SYS. Briefly, a remarkable stratification (of Δρ ≥ 1.5 kg·m⁻³) appears in late spring (May) every year, develops in summer and early autumn (from June to September), tends to collapse in mid-autumn and/or late autumn (October and November), and disappears in early winter (December). This is in good agreement with our results (Figure S1). The well-established stratification and the consequent cold water mass at the bottom layer in warm seasons (i.e., the SYSCWM) were distributed over a large area of the central SYS (Figures 1C–E), undoubtedly blocking vertical water mixing and gas exchanges between the bottom and surface waters. In the SYSCWM, excess DIC versus AOU exhibited a tight relationship with each other and satisfactorily followed the traditional Redfield relationship (Figures 5C, D). In surface waters, however, spring excess DIC versus AOU exhibited a loose relationship with each other, and, to some extent, deviated from the traditional Redfield relationship (Figures 5A, B). This comparison suggested that the SYSCWM served as an almost enclosed biogeochemical reactor on a time scale of 0.5 years. In this biogeochemical reactor, AOU and excess DIC were quasi-synchronously accumulated from late spring to autumn, leading to a decline in bottom-water Ω_arag and pH_T (Figure 6).

To examine the potential effects of East China Sea shelf waters on the SYSCWM carbonate system, the three-endmember water-mixing models in April 2018 and 2019 were also applied to the other surveys. Then, the non-conservative behaviors of DIC (i.e., ΔDIC) were estimated using Equation (7). The monthly relationships among ΔDIC, excess DIC, and AOU in the SYSCWM (Figures 11) were similar to those in the entire SYS in spring (Figure 10), indicating that the potential influences of horizontal mixing on the accumulation of inorganic carbon in SYSCWM in summer and autumn were negligible and suggesting that the simplified analyses based on excess DIC should be reasonable (see Results).

The data maps of water temperature, salinity, and TAlk (Figures 3, 4) provided more evidence of the negligible impacts of East China Sea shelf waters on the SYSCWM carbonate system in a given year. In August 2019, the potential East China Sea shelf water intrusion in the SYS was characterized as having a relatively low salinity of <30 (Figure 3j) and a low TAlk of < 2,210 μmol·kg⁻¹ (Figure 4b) at the surface layer. Both were substantially lower than the values generally found in SYS waters, indicating origins from rainfall and/or an estuarine plume (from the Changjiang diluted water). A similar water type also affected the SYS bottom waters along the China coast west of 122°E. In bottom waters east of 122°E, however, the only exogenous water with a relatively high temperature of >17°C (Figures 3f–h), low salinity of <32 (Figures 3n–p), and low TAlk of <2,210 μmol·kg⁻¹ (Figures 4f–h) was confined to the south of 33.5°N, exhibiting a clear boundary with the SYSCWM (Figures 1C–E). In summary, the East China Sea shelf water intrusion mainly affected the surface waters in the SYS in summer. Its impacts on the SYS bottom waters were located outside the SYSCWM area under study.

The DIC addition can decrease Ω_arag and pH values, and vice versa (Figure 12), resulting in near-linear relationships. The observed Ω_arag to ΔDIC relationship in the SYSCWM in 2018 was systematically lower than that found in 2019 (Figure 12E), while there was no statistical difference in the observed pH_T-to-ΔDIC relationships in both years (Figure 12F). The different intercepts of 1.85 in 2018 and 1.96 in 2019 (Figures 12A, B) denoted again the slightly different initial Ω_arag in spring 2018 and spring 2019 (see section 3.6) when the bottom-water net ecosystem metabolism (net ecosystem photosynthesis minus respiration) was zero. The slopes of the regression lines of Ω_arag versus ΔDIC ranged from −0.0088 to −0.0084 (Figures 12A, B). When ΔDIC was added with 100 μmol·kg⁻¹, Ω_arag decreased by 0.84–0.88. In other words, to reduce Ω_arag by one unit, it is necessary to increase ΔDIC by 114–119 μmol·kg⁻¹ in the SYSCWM.

A similar slope of 100 μmol·kg⁻¹ ΔDIC to 0.89 Ω_arag changes was obtained in the central North Yellow Sea (Li & Zhai, 2021). The value in the northern Gulf of Mexico (also a high productivity coastal region), however, differed from ours: Huang et al. (2021) found that each increase in the DIC removal of 100 μmol·kg⁻¹ increased Ω_arag by 1 unit. Therefore, the sensitivity of seawater Ω_arag changes to DIC addition varies in different systems.

These differences are mainly due to different water temperatures. Compared with the Yellow Sea cold water mass (where the water temperature was 7°C–12°C in this study), the water temperature in the northern Gulf of Mexico is much higher (20°C–30°C). Thus, the annual initial value of Ω_arag is also relatively high in the northern Gulf of Mexico, resulting in a relatively large space for the decline in Ω_arag. The higher the water temperature and the higher the annual initial Ω_arag, the greater the space for Ω_arag change, indicating that Ω_arag is more sensitive to the addition of DIC. In contrast, the lower the water temperature, the closer the annual initial Ω_arag to the mathematical limitation of Ω_arag (see Introduction for a definition of Ω_arag). The response of Ω_arag changes to the addition of DIC will be weaker.

To quantify this temperature-driven relationship, we calculated ideal relationships of Ω_arag and the sensitivity index of Ω_arag (to ΔDIC) versus ΔDIC in the Yellow Sea cold water mass and the surface of the open ocean (Figures 13A, B). The simulation was based on CO2SYS software, equation (1), region-specific water temperature, salinity, and TAlk. The initial f CO 2 air of 365 μatm was used to obtain the zero point of the ΔDIC. In the Yellow Sea cold water mass, the water temperature was fixed at 10°C, while TAlk = 2,313 μmol·kg⁻¹ and salinity = 32.5 (Table 2). On the surface of the open ocean, the water temperature was fixed at 25°C, TAlk = 2,300 μmol·kg⁻¹, and salinity = 35, very similar values to those found in the high-salinity region of the northern Gulf of Mexico, as reported by Huang et al. (2021). During the simulation, f CO 2 air was increased by 30 μatm at every step.

The results clearly revealed the region-specific sensitivity index of Ω_arag to ΔDIC (Figure 13A); that is, 100 μmol·kg⁻¹ ΔDIC to 0.88 Ω_arag changes in the Yellow Sea cold water mass (the same as our data-based results) and 100 μmol·kg⁻¹ ΔDIC to 1.00 Ω_arag changes at the surface of the open ocean (including the high-salinity region of the northern Gulf of Mexico) as reported by Huang et al. (2021). However, the simulation also clearly revealed that the sensitivity index declined when ΔDIC increased (Figure 13B). This was true in both of the two ideal systems, while the sensitivity decline was more pronounced in the Yellow Sea cold water mass than at the surface of the open ocean and in the high-salinity region of the northern Gulf of Mexico.

Water salinity also affects the relationship between Ω_arag changes and ΔDIC (Xiong et al., 2019). Therefore, there is no uniform proportional relationship between Ω_arag changes and DIC addition, and this relationship is greatly affected by the degree of CO₂ supplementation. In addition to the effects of the above annual initial cooling on Ω_arag, the sensitivity of Ω_arag changes to DIC addition in various sea areas may tend to decrease in the future as atmospheric CO₂ continuously increases and invades the ocean interior.

To further explore causes of the apparently region-specific performance of the sensitivity index, we plotted the sensitivity index of Ω_arag (to ΔDIC) against the TAlk : DIC ratio (Figure 13C). The apparent regional differences (Figures 13A, B) almost disappeared, suggesting that the temperature-driven decline in the sensitivity index of Ω_arag (to ΔDIC) in the Yellow Sea cold water mass—compared with that in open oceans and the Gulf of Mexico—was actually caused by cooling-promoted CO₂ dissolution and so by the decline in the annual initial TAlk : DIC ratio in the Yellow Sea in winter and early spring. Since the TAlk : DIC ratio is strongly correlated with the carbonate ion concentration (when the TAlk : DIC ratio is higher than 1 and the carbonate ion concentration is more than 50 μmol·kg⁻¹), and the Ω_arag variations primarily depended on carbonate ion concentrations (Li and Zhai, 2021), a significant correlation between Ω_arag and the TAlk : DIC ratio was revealed in our ideal simulation (Figure 13D). This correlation is almost linear (when the TAlk : DIC ratio is higher than 1), and exhibits ignorable regional differences based on our ideal settings. A data-based comparison between Ω_arag and the TAlk : DIC ratio has been carried out in the Gulf of Mexico and on the east coast of the United States by Wang et al. (2013).