Comparing Subsurface Seasonal Deoxygenation and Acidification in the Yellow Sea and Northern East China Sea Along the North-to-South Latitude Gradient

To better understand the relationship between subsurface seasonal deoxygenation and acidification in the Yellow Sea and northern East China Sea (ECS), we examined carbonate system parameters and dissolved oxygen (DO) of seven field surveys conducted in 2017–2018, spanning all four seasons. Low pHT values of 7.71–7.80 and critically low aragonite saturation state (Ωarag) values of 1.07–1.40 along with undersaturated DO of mostly higher than 150 μmol O2 kg–1 occurred in the Yellow Sea Cold Water Mass area in summer and autumn, while hypoxic DO values of 49–63 μmol O2 kg–1 and extremely low pHT values of 7.68–7.74 as well as critically low Ωarag values of 1.21–1.39 were observed in the northern ECS in July 2018. At the beginning of warm-season stratification formation, the cold Yellow Sea waters had much higher DO but lower Ωarag values than those in relatively warmer ECS waters, while yearly initial pHT values rarely exhibited differences between the two coastal seas. During warm seasons, the central Yellow Sea accumulated respiration products beneath the thermocline in summer and autumn, while the northern ECS bottom waters preserved them only in summer. This study highlights fundamental roles of wintertime carbon dioxide (CO2) solubility along a north-to-south latitude gradient in the coastal acidification development. In comparison with the relatively low-latitude northern ECS subject to seasonal hypoxia, relatively high-latitude Yellow Sea exhibits higher CO2 solubility in winter and longer respiration-product accumulations in warm seasons, leading to lower Ωarag in the central Yellow Sea than those in the northern ECS. However, the present-day central Yellow Sea is free from hypoxia.


INTRODUCTION
The oceanic absorption of anthropogenic carbon dioxide (CO 2 ) has lowered sea surface pH and calcium carbonate (CaCO 3 ) mineral saturation state ( ) as compared with the preindustrial era, known as ocean acidification (Caldeira and Wickett, 2003;Orr et al., 2005;Doney et al., 2009). Here pH is the negative logarithm of the sum of the concentrations of hydrogen (H + ) and bisulfate (HSO 4 − ) ions, i.e., total hydrogen ion concentration scale, pH T = -log 10  2− ] are the concentrations of calcium and carbonate ions, respectively, and K sp * is the apparent solubility product for either calcite or aragonite. The declines in pH and could lead to CaCO 3 -undersaturated corrosive seawater conditions, affecting marine calcifying organisms and even the whole marine ecosystem (Fabry, 2008;Jin et al., 2015;Ravaglioli et al., 2020).
The anthropogenic CO 2 invasion has resulted in a decrease in pH by 0.1 unit (Orr et al., 2005) and a decline in [CO 3 2− ] by 30% (Sabine et al., 2004) in the upper ocean since the industrial revolution. The present-day open ocean pH T was detected at 8.023 ± 0.004 in the tropical Central Pacific (6.4 • N 162.4 • W), at 8.074 ± 0.004 in the subtropical Eastern Pacific (33.5 • N 122.5 • W), and at 8.020 ± 0.008 at an Antarctic site (77.6 • S 166.4 • W) (Hofmann et al., 2011). In the Pacific Ocean, the present surface arag values are 3-4.5 in low-latitude regions while only 1-2 in high-latitude regions (Feely et al., 2012). This latitude gradient of arag is largely attributed to higher solubility of CO 2 in colder seawaters of high-latitude regions. Moreover, seasonal aragonite undersaturation (i.e., arag < 1) has already been observed in surface and shallow subsurface waters of some northern polar seas (Bates et al., 2009;Fabry et al., 2009;Qi et al., 2017).
Chemically, arag > 1 indicates that the CaCO 3 mineral of aragonite is stable in the seawater, while arag < 1 indicates that the mineral is unstable. Although corals usually require much higher arag of > 3.0 for optimal growth (Eyre et al., 2018;, many researchers regarded an arag value of 1.5 as a critical threshold for marine shellfish development (Gruber et al., 2012;Ekstrom et al., 2015;Waldbusser et al., 2015), below which marine calcifying organisms may be under threat of acidified seawaters. On the Chinese side of the North Yellow Sea, the net community calcification rate in subsurface waters declined to zero when the arag value reached the critical level of 1.5-1.6 (Li, 2019;Li and Zhai, 2019).
In productive coastal zones, algae and other biogenic particles decompose in subsurface waters. Their respiration and/or remineralization processes consume dissolved oxygen (DO) and release a great deal of CO 2 into subsurface waters, leading to more rapid seawater acidification in coastal seas, compared with the open ocean (Feely et al., 2010;Cai et al., 2011;Melzner et al., 2013;Jiang et al., 2019). Moreover, the respirationinduced seasonal acidification is subject to uneven distributions of seawater temperature, salinity and alkalinity in coastal seas. For example, transregional carbonate studies along the U.S. East Coast have indicated that its northeast shelf region is more susceptible to CO 2 acidification than the southern region, along with a north-to-south increasing gradient in total alkalinity (TAlk) and limited geographical variation in dissolved inorganic carbon (DIC) Wanninkhof et al., 2015). Exposure, sensitivity and adaptive capacity to the acidification driven by atmospheric CO 2 absorption and local amplifiers (i.e., eutrophication, upwelling of CO 2 -riched waters and input of river water with low [Ca 2+ ] and subsequently arag ) were also assessed and compared throughout the U.S. coastal regions (Ekstrom et al., 2015). Along China's coasts, so far, the latitude gradient of coastal acidification and its controlling mechanisms are unclear because previous studies have been primarily confined to individual coastal seas (e.g., Cao et al., 2011;Chou et al., 2013a;Zhai, 2018). Ocean environments along China's coasts also exhibit complex local geological and hydrological characteristics, such as contrasting bottom topography, residence time and monsoon-driven coastal currents (e.g., Su and Yuan, 2005;Chen, 2009;Men and Liu, 2015).
In this study, we examined carbonate system parameters in the Yellow Sea and northern East China Sea (ECS) during 2017-2018, spanning all four seasons. For the first time, a north-tosouth gradient of carbonate system parameters in contrasting coastal seas along China's east coast was revealed. Together with hydrological data and DO measurements, the regional differentiation of the respiration-induced coastal acidification in the two coastal seas were investigated. This study provides the best understanding so far of the relationship between subsurface seasonal deoxygenation and acidification in the two coastal seas of both ecological and economic importance, which will assist future predictions of marine environmental changes under ocean acidification in the coming decades.

Study Area
The Yellow Sea and ECS, located on the China eastern shelf, are two major marginal seas of the western North Pacific. The boundary of the two coastal seas lies between the northern corner of the Changjiang Estuary and Jeju Island ( Figure 1A). The Yellow Sea is surrounded by mainland China to the west and the Korea Peninsula to the east. It is geographically divided into two basins, i.e., the North Yellow Sea and the South Yellow Sea. The former is connected to the Bohai Sea to the west, and the latter to the ECS to the south. The North Yellow Sea has an area of ∼7 × 10 4 km 2 with an average water depth of ∼38 m, while the South Yellow Sea has an area of ∼30 × 10 4 km 2 with an average water depth of ∼44 m. The more open ECS has a larger area of ∼77 × 10 4 km 2 with an average water depth of ∼370 m. The ECS inner shelf within the 50 m isobaths is quite broad. The climatic variations are primarily dominated by the East Asian Monsoon, with the rain-bearing southwest monsoon prevailing in summer (from June to early September) and a strong northeast monsoon lasting in winter (from December to early March of the next year, Chen, 2009).
Both of the ECS and the Yellow Sea are connected to the North Pacific via the Kuroshio intrusion, including those Kuroshioderived currents such as the TaiWan Warm Current (TWWC), the TsushiMa Warm Current (TMWC), and the Yellow Sea Warm Current (YSWC). Moreover, they are subject to freshwater discharges from the Changjiang and Yalu Rivers (Figure 1A), as well as several monsoon-driving coastal currents, including the northeastward-moving Changjiang Diluted Water (CDW) from late spring to early autumn, and the southward-moving Yellow Sea Coastal Current (YSCC) and Zhejiang-Fujian Coastal Current (ZFCC) in winter and early spring.
The Yellow Sea is semi-enclosed. Its summertime hydrography is characterized by a pronounced stratification in its deeper regions. A cold pool with water temperatures of 5-11 • C, the Yellow Sea Cold Water Mass (YSCWM), develops under the thermocline from late spring to autumn as the remnant of the previous winter cooling (Miao et al., 1990;Zhai et al., 2014b). The wintertime hydrography in the Yellow Sea is characterized by the southward-moving YSCC and the northward-moving YSWC ( Figure 1A). The YSWC is considered to be a compensating current to the monsoon-driven coastal current (Yuan et al., 2008), transporting warm and saline waters into the Yellow Sea. Based on 228 Ra/ 226 Ra measurements, the Yellow Sea hydraulic residence time has been estimated to be 5-6 years, while the residence time may only be 2-3 years or shorter on the ECS shelf (Nozaki et al., 1991;Men and Liu, 2015).
TAlk in the Yalu River is only 320-800 µmol kg −1 (Zhai et al., 2014b, approximately 1000 µmol kg −1 lower than the Changjiang TAlk value (1500-1900 µmol kg −1 , Xiong et al., 2019). As illustrated by Chen and Wang (1999) and Zhai et al. (2014a), the ECS offshore waters originate from the Kuroshio tropical water. The latter has a typical TAlk value of 2293 µmol kg −1 and the DIC value of 1994 µmol kg −1 (with the DIC:TAlk ratio of < 0.9) at a salinity of 34.9 (Chen and Wang, 1999). By comparison, the TAlk values in the semi-enclosed Yellow Sea were usually detected in a compact range of 2290 ± 25 µmol kg −1 (Zhai, 2018). In the North Yellow Sea, a usual water mixing model has been reported by Zhai et al. (2014b), i.e.,

TAlk North_Yellow_Sea
= 61.745 × Salinity + 320 (1) where 320 (µmol kg −1 ) represents the low TAlk feature of the mixture of rainwater and freshwater discharged from the Yalu River.

Sampling and Analyses
In this study, seven field surveys were conducted on the Chinese side of the Yellow Sea and the northern ECS during 2017-2018 (Supplementary Table S1 Figure S1A). Water samples were collected at two to seven different depths (including sea surface and the bottom water) using a rosette of 10 or 12 Niskin bottles, integrated with Conductivity-Temperature-Depth/Pressure (CTD) sensor packages. The ancillary data of in situ temperature (after the International Temperature Scale of 1990) and salinity (after the Practical Salinity Scale of 1978) were obtained primarily using the calibrated CTD sensor packages (SBE-19 plus in our October-November 2017 and May 2018 cruises, and SBE-911 plus during the other cruises, Sea-Bird Scientific, Bellevue, WA, United States). During the summertime estuarine survey conducted in July 2018, salinity values of discrete samples were also measured using a calibrated WTW's TetrCon925 probe.
DO samples were collected, fixed and titrated aboard following the Winkler procedure at an overall uncertainty level of <0.5%. A small quantity of sodium azide (NaN 3 ) was added during subsample fixation to remove possible interferences from nitrites (Wong, 2012). The DO saturation (DO%) was calculated from field-measured DO concentration divided by the DO concentration at equilibrium with the atmosphere which was calculated from temperature, salinity and local air pressure, as per the Benson and Krause (1984) equation. To quantify the effect of net community metabolism, apparent oxygen utilization (AOU) was also calculated by subtracting the field-measured DO concentration from the air-equilibrated DO. Assuming the water starts with a fully saturated state, and ignoring 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 analyses were also collected aboard. As recommended by Huang et al. (2012), water samples for DIC and TAlk were stored in 60 mL borosilicate glass bottles (for DIC, bubble free) and 140 mL high-density polyethylene bottles (for TAlk). There were no statistical differences between the measuring results from the above-mentioned sample storing procedure and from those procedure suggested by Dickson et al. (2007). Following filling procedure in Dickson et al. (2007), we filled these triple-rinsed sample bottles of DIC and TAlk smoothly from the bottom, and then immediately added 50 µL of saturated mercuric chloride (HgCl 2 ). Finally, water samples for DIC and TAlk were sealed and preserved at room temperature until determination. Note that the volume of saturated HgCl 2 added to the DIC samples exceeded the upper limit of recommended range (0.02-0.05% by volume), but was still below the maximum amount, i.e., 0.1% by volume (Dickson et al., 2007). Both DIC and TAlk samples were unfiltered but allowed to settle before measurement, although filtration techniques suitable for these samples were reported earlier by Bockmon and Dickson (2014). DIC was measured by an infrared CO 2 detector-based DIC analyzer (AS-C3, Apollo SciTech Inc., United States), and TAlk was determined at 25 • C by the Gran acidimetric titration using a semi-automated titrator (AS-ALK2, Apollo SciTech Inc., United States). DIC and TAlk determinations were referred to Certificated Reference Materials (CRM) from Andrew G. Dickson's lab at Scripps Institution of Oceanography at a precision of ± 2 µmol kg −1 (Dickson et al., 2007;Zhai et al., 2014b).

Calculation of Other Carbonate System Parameters
Seawater fugacity of CO 2 (f CO 2 ), pH T and arag were calculated from seawater temperature, salinity, and measured DIC and TAlk using the software CO2SYS.XLS (Version 24) (Pelletier et al., 2015), which is an updated version of the original CO2SYS.EXE (Lewis and Wallace, 1998). This program has been favorably evaluated by Orr et al. (2015) in a study comparing 10 packages of carbonate calculation program. The Millero et al. (2006) dissociation constants of carbonic acid were used in the calculation because they cover much broader applicable ranges of temperature (0-50 • C) and salinity (0-50). The Dickson (1990) dissociation constant was used for HSO 4 − ion. The phosphate and silicate values required by the program were usually unavailable and replaced by zero. The Ca 2+ concentrations were assumed to be proportional to salinity as presented in Millero (1979) and the values of apparent solubility product for aragonite (K sp * arag ) were taken from Mucci (1983).
To assess the quality of the carbonate system data, we calculated pH data using the National Bureau of Standards scale (pH NBS ) based on field-measured DIC and TAlk values. These data were compared with field-measured pH NBS data (see collection and analysis of pH NBS samples in Supplementary Material). Most measured and calculated values were consistent at a deviation level of ± 0.05 pH (Supplementary Figure S2A). To examine the possible existences of organic alkalinity in coastal waters within our study area, we also calculated TAlk values from field-measured DIC and pH NBS data. Most measured TAlk data and calculated results were consistent with each other at a deviation level of ± 20 µmol kg −1 (Supplementary Figure S2B). This deviation level was reasonably higher than the precision of TAlk determination (± 2 µmol kg −1 ). These comparisons suggested that the measured and calculated results of the carbonate system parameters were reliable. Due to accidentally insufficient addition of HgCl 2 , the North Yellow Sea DIC samples collected in April 2018 were damaged before determination. The Frontiers in Marine Science | www.frontiersin.org relevant DIC data were calculated from field-measured TAlk and pH NBS .
To quantify the effect of net community metabolism on DIC, we calculated the air-equilibrated DIC (corresponding to a mean air-equilibrated f CO 2 value of 415 ± 5 µatm during our seasonal cruises conducted in 2017-2018) from corresponding field-measured seawater temperature, salinity and TAlk (Zhai, 2018). The air-equilibrated f CO 2 was calculated from the flask analysis data of atmospheric CO 2 mole fraction at the adjacent Tae-ahn Peninsula (TAP) site (36 • 44'N 126 • 08'E), which varied from 406 ppm (ppm = parts of CO 2 per million dry air) in August to 419-421 ppm during January to May in 2017-2018 (Supplementary Figure S1B, data from NOAA/ESRL's Global Monitoring Division) 1 , and corrected to the survey-based barometric pressure and 100% humidity at water temperature and salinity . Similar to the definition of AOU (section "Sampling and Analyses"), the DIC departure from the air-equilibrated DIC was defined as the excess DIC (ExcessDIC). Assuming water starts with a fully saturated state, and ignoring effects of air-sea exchange, water mixing and CaCO 3 precipitation/dissolution, an ExcessDIC > 0 means net community respiration, while an ExcessDIC < 0 implies net community production.

Hydrological Settings
Generally, water temperature exhibited a north-to-southincreasing gradient from the North Yellow Sea, to the South Yellow Sea, and to the northern ECS in winter, spring and autumn (Supplementary Figures S3-S5). The only exception was sea surface temperature in summer, showing no latitude gradient in these sea areas. However, summertime temperature in subsurface and bottom waters exhibited a north-to-southincreasing gradient (Supplementary Figures S4, S5). During our winter and spring cruises, regionally survey-averaged temperatures were 4.8-7.8 • C in the North Yellow Sea, 7.2-11.7 • C in the South Yellow Sea and 9.6-16.7 • C in the northern ECS (Table 1).
Salinity also exhibited the north-to-south-increasing gradient, with annual mean values of 32.1 ± 0.3 in the North Yellow Sea, 32.2 ± 0.7 in the South Yellow Sea and 32.6 ± 1.9 in the northern ECS, based on data obtained from our seasonal cruises in 2017-2018. The Yellow Sea had relatively low salinity values and small salinity variations as compared with the northern ECS ( Figures 1B-G, 2A-C). In the Yellow Sea, relatively high salinity values of > 32 dominated the whole study area in winter and spring (Figures 1C-E). In summer and autumn, the relatively high salinity values of > 32 still dominated bottom waters in the central Yellow Sea (Figures 1B,F-G), i.e., the summertime YSCWM area ( Figure 1A). In the connection between the northern ECS and South Yellow Sea, several low salinity values of 29.7-31.5 were observed in late spring ( Figure 1E), likely due to the offshore transport of the CDW. In summer, the 1 http://www.esrl.noaa.gov/gmd/ CDW-affected sampling sites considerably increased, covering the northwestern ECS and the southern part of the South Yellow Sea ( Figure 1F and Supplementary Figure S1A). In autumn, low salinity values of 26.1-30.5 were observed at nearshore stations in the ECS (Figure 1G), indicating the effect of the southward-moving ZFCC during this northeast monsoon-driven season ( Figure 1A).
In the Yellow Sea, significant thermoclines and stratification occurred in summer and autumn (Figures 3A-C), with mean surface temperatures of 26.7 ± 2.3 • C and 19.2 ± 1.7 • C in summer and autumn, respectively, and with mean bottom-water temperatures of 12.5 ± 6.2 • C and 13.9 ± 4.7 • C in summer and in autumn, respectively (Supplementary Figure S6). In the central Yellow Sea, subsurface water had quite low temperature of ∼9 • C in warm seasons (Figures 2A,B), shaping the YSCWM area with large density difference between bottom and surface waters ( Density) of ∼5 kg m −3 in summer and ∼2 kg m −3 in autumn (Figures 3A-C). Compared with the Yellow Sea, the northern ECS had relatively small bottom-surface temperature differences (26.1 ± 1.8 • C versus 20.9 ± 1.7 • C) and Density

DO, fCO 2 , and Apparent DO Depletion Rate
During our winter and spring cruises, most DO values were at ∼100% saturations, while f CO 2 were close to the present-day airequilibrated f CO 2 of 415 µatm in the Yellow Sea and northern ECS ( Table 1 and Figures 2D-F), suggesting a vertically wellmixed situation during cold seasons. Exceptions to this were observed in the central part of the South Yellow Sea in winter, where relatively low DO% of 65-72% and supersaturated f CO 2 of 672-806 µatm occurred in bottom waters at four deep stations (Supplementary Figure S5). From late spring to autumn, the YSCWM bottom waters exhibited DO declines (from 97% or 285 µmol O 2 kg −1 in late spring to 87% or 248 µmol O 2 kg −1 in summer and 68% or 195 µmol O 2 kg −1 in autumn) and f CO 2 increases (from 440 µatm in late spring to 505 µatm in summer and 680 µatm in autumn) (Figures 2D,E, 3D-F). These low DO values in the YSCWM (mostly higher than 150 µmol O 2 kg −1 ) were still above the threshold of hypoxia (i.e., <63 µmol O 2 kg −1 ). The regional averaged apparent DO depletion rate in the YSCWM bottom waters was estimated to be 0.6 µmol O 2 kg −1 d −1 from late spring to autumn (∼150 days).
In the northern ECS, summertime bottom-water DO% values were averaged at only 56 ± 13% (with a range of 21-84%, having the DO concentrations of 49-185 µmol O 2 kg −1 ), while the autumnal bottom-water DO% increased to the airequilibrated level (Figures 2F, 3D-F). The three summertime hypoxic stations (with DO values of 49-63 µmol O 2 kg −1 ) off the Changjiang Estuary had extremely high f CO 2 values of ∼1000 µatm (Supplementary Figure S5). Since water stratification in the northern ECS was intensified from late spring to summer  (2019). c Calculated from field-measured TAlk and pH NBS data, since DIC samples collected during this survey were accidentally damaged before determination. (∼50 days), its bottom-water averaged apparent DO depletion rate (from 85% or 215 µmol O 2 kg −1 in late spring to 56% or 126 µmol O 2 kg −1 in summer) was estimated to be 1.8 µmol O 2 kg −1 d −1 .
In addition, moderately low bottom-water DO values of ∼140 µmol O 2 kg −1 were also observed in autumn at several southeastern stations (Figure 3F), where water temperature was moderately low (∼22 • C) (Supplementary Figure S6F), and salinity was quite high (∼34) (Figure 1G), likely resulted from the TWWC that intruded into the northern ECS ( Figure 1A).
TAlk versus salinity showed different relationships in the three regions (Figures 4A-C). In the North Yellow Sea, TAlk versus salinity roughly followed Eq. (1), although TAlk data obtained from our five cruises in 2018 were 10-30 µmol kg −1 higher than those values predicted by salinity and Eq. (1). In the South Yellow Sea, however, quite complicated water mixing behaviors were involved ( Figure 4B). In the northern ECS, many data points of TAlk versus salinity fairly followed a linear relationship (Figure 4C), i.e., Equation (2) indicated a two-endmember water mixing between the Changjiang freshwater (S = 0, TAlk = 1850 µmol kg −1 ) and the ECS offshore waters (S = 34.9, TAlk = 2266 µmol kg −1 ), as derived from our July 2018 cruise conducted off the Changjiang Estuary (Figure 1). This linear relationship also roughly characterized several ECS nearshore stations (along the China's east coast) sampled during  (Figure 4C), indicating that the southward-moving ZFCC was closely coupled with the CDW (Figure 1A). In winter and spring, however, quite high TAlk values of 2283-2333 µmol kg −1 at moderate salinity values of 31.3-32.8 were observed in the northern ECS (Figure 4C), showing the intrusion of the northeast monsoon−driven YSCC ( Figure 1A).

DIC and DIC:TAlk Ratio
The Yellow Sea exhibited higher DIC values than the northern ECS ( Figures 2G,H versus Figure 2I). Annual mean DIC values were 2145 ± 47 µmol kg −1 in the North Yellow Sea, 2119 ± 65 µmol kg −1 in the South Yellow Sea, and 2031 ± 65 µmol kg −1 in the northern ECS. Wintertime and springtime DIC values in the North Yellow Sea were averaged at 2168 ± 22 µmol kg −1 (Figure 4D), while wintertime and springtime DIC values in the South Yellow Sea were averaged at 2144 ± 36 µmol kg −1 (Figure 4E). In the Yellow Sea, DIC data showed greater vertical variations in summer and autumn than in winter and spring (Supplementary Figures S3-S5), and the YSCWM had relatively high DIC values of 2150-2270 µmol kg −1 in summer and autumn (Figures 4D,E). In the northern ECS, the low DIC values of this study of 1650-1950 µmol kg −1 were observed in the summertime ECS surface waters (Supplementary Figure S7F), while the ECS bottom-water DIC values were mostly 2050-2150 µmol kg −1 in summer ( Figure 4F).
The Yellow Sea usually had higher DIC:TAlk ratios than the northern ECS (Figures 2G-I). From early winter to late spring, survey-averaged DIC:TAlk ratio in the North Yellow Sea ranged between 0.930 ± 0.009 and 0.933 ± 0.005 (Table 1), usually at 0.930 ± 0.010 ( Figure 4G), while survey-averaged DIC:TAlk ratio in the South Yellow Sea varied from 0.924 ± 0.016 to 0.930 ± 0.009 (Table 1) Figure S5).
In the northern ECS, survey-averaged DIC:TAlk ratio values in early winter (0.920 ± 0.003) and early spring (0.930 ± 0.006) were much higher than the usual ratio of ∼0.9 in the ECS offshore waters ( Table 1), but quite close to the usual DIC:TAlk ratio of wintertime and springtime Yellow Sea waters (Figures 2G-I). This was likely because the YSCC transported the Yellow Sea waters into the northern ECS during the northeast monsoon season (Figure 1A). In late spring, the ECS DIC:TAlk ratio tended to show limited vertical gradient, with surface values of 0.891 ± 0.013 (Supplementary Figure S7I) and bottomwater values of 0.928 ± 0.009 (Figure 4I). In general, the ECS bottom waters increased their DIC:TAlk ratio values to 0.930-0.970 in summer, and then declined to < 0.9 in autumn (Figures 3G-I). At several nearshore stations affected by the ZFCC and those southeastern stations likely affected by the TWWC intrusion, relatively high bottom-water DIC:TAlk ratio values of 0.920-0.930 were also revealed during our autumn 2018 cruise (Figures 3I, 4I).

Bottom-Water AOU and Excess DIC From Late Spring to Autumn
In late spring, both bottom-water AOU and ExcessDIC varied around 0 in the North Yellow Sea (Figure 5A), while the South Yellow Sea bottom-water AOU and ExcessDIC  (Zhai et al., 2014b;Zhai, 2018). In (D-I), gray shaded areas represent mean ± SD of wintertime/springtime DIC or DIC:TAlk ratios in individual regions. In (J-L), the pH T of 8.0 is comparable with the present-day air-equilibrated pH T (corresponding to a mean air-equilibrated fCO 2 value of 415 ± 5 µatm during our seasonal cruises in 2017-2018), while the pH T of 7.7 shows the doubled concentration of total hydrogen ions. In (M-O), the arag of 1.5 shows a critical value that the net community CaCO 3 dissolution occurs in the North Yellow Sea (Li, 2019;Li and Zhai, 2019), while the arag of 1.0 indicates the critical value for the ideal aragonite dissolution.

Seasonal Variations in pH T and arag
In the Yellow Sea, wintertime and springtime pH T and arag were generally homogenous, while pH T and arag displayed vertical gradient in summer and autumn. Its early-winter pH T and arag were 8.04 ± 0.05 and 1.94 ± 0.22, respectively, while its earlyspring pH T and arag were 8.08 ± 0.05 and 1.86 ± 0.16, and the late-spring pH T and arag were 8.04 ± 0.06 and 1.98 ± 0.34 (Table 1). During our summer and autumn cruises, surface pH T ranged between 7.98 ± 0.07 and 8.04 ± 0.05 (Supplementary  Figures S7M,N), while surface arag varied from 2.58 ± 0.30 to 3.00 ± 0.36 (Supplementary Figures S7P,Q). The YSCWM pH T decreased to 7.97 ± 0.05 in summer (mostly lower than the present-day air-equilibrated level of 8.0-8.1) and 7.85 ± 0.04 in autumn. Note that very low pH T values of 7.71-7.80 occurred in the YSCWM in autumn (Figures 4J,K), suggesting that the concentration of total hydrogen ions should be twice the present-day air-equilibrated level of 8.0-8.1. The YSCWM arag declined to 1.59 ± 0.18 in summer and 1.28 ± 0.09 in autumn (Figures 2J,K). Its autumnal lowest arag value of 1.07 was detected in the northern area of the YSCWM (Figure 4M), which was even close to the critical value for the ideal aragonite dissolution. Even in early winter, very low bottom-water pH T values of 7.77-7.85 ( Figure 4K) and arag values of 1.13-1.29 ( Figure 4N) were also observed at the four deep stations (in the central South Yellow Sea) mentioned in section "DIC and DIC:TAlk Ratio." In the northern ECS, the early-winter pH T and arag were averaged at 8.03 ± 0.02 and 2.08 ± 0.07, respectively, and the early-spring pH T and arag were averaged at 8.03 ± 0.02 and 1.85 ± 0.12, respectively ( Table 1). In late spring, the ECS bottom-water pH T and arag were averaged at 7.96 ± 0.03 and 1.90 ± 0.21, respectively (Figure 2L), while its surface pH T and arag were averaged at 8.09 ± 0.05 (Supplementary Figure S7O) and 2.80 ± 0.30 (Supplementary Figure S7R), respectively. In summer, the ECS bottom-water pH T and arag were averaged at 7.85 ± 0.07 and 1.91 ± 0.32, respectively (Figure 2L), while the surface pH T and arag were averaged at 7.99 ± 0.16 (Supplementary Figure S7O) and 2.99 ± 0.99 (Supplementary Figure S7R), respectively. Near the riverine mouth of Changjiang, quite low summertime bottom-water pH T values of 7.68-7.74 and arag values of 1.21-1.39 were observed at five stations (Figures 3K,N, 4L,O). Three of these seriously acidified stations (pH T 7.68-7.71; arag 1.28-1.39) suffered from summertime hypoxia (with DO values of 49-63 µmol O 2 kg −1 ), while the other two stations (pH T 7.68 and 7.74; arag 1.21 and 1.39) were very shallow (with water depth of ∼10 m) and vertically well-mixed. In autumn, most pH T and arag in the northern ECS increased to nearly air-equilibrated values of ∼8.00 and ∼2.60 (Figures 2-3), except for several stations likely affected by the ZFCC and the TWWC intrusion, where low pH T values of 7.86-7.99 and low arag values of 1.97-2.60 were detected (Figures 4L,O).

Processes Driving Seasonal Variations of Bottom-Water DIC, pH T , and arag
Dynamics of DIC, pH T , and arag in coastal zones are subject to multi-drivers, including coastal upwelling (Feely et al., 2008), riverine freshwater inputs (Salisbury et al., 2008;Rheuban et al., 2019;Xiong et al., 2019), vertical and lateral water mixing Wanninkhof et al., 2015), and metabolic processes (Feely et al., 2010). In addition to these drivers, the seasonal temperature variability also affects the carbonate chemistry (Zhai et al., 2014b). In an open system, the temperature effect on carbonate system parameters includes the internal thermodynamic equilibrium shift and the air-sea CO 2 exchange caused by the solubility change Cai et al., 2020). These two temperature effects tend to cancel each other out for pH T , but to enhance each other for arag because of the dominant role of gas equilibrium (Cai et al., 2020).
To illuminate the temperature effect on seasonal changes of carbonate system parameters, air-equilibrated DIC, DIC:TAlk ratio, pH T and arag values (corresponding to a mean airequilibrated f CO 2 value of 415 ± 5 µatm in 2017-2018) in the Yellow Sea and northern ECS were plotted against the seawater temperature. Both of air-equilibrated DIC and DIC:TAlk ratio increased as temperature decreased (Figures 6A-C,E-G). However, air-equilibrated pH T varied limitedly with temperature change (Figures 6I-K), whereas air-equilibrated arag decreased as temperature decreased (Figures 6M-O).

Wintertime Re-equilibration
In shallow waters away from upwelling systems, the carbonate system within the water column re-equilibrated with atmospheric CO 2 every year between winter and spring (Figures 2D-F).
Comparing field-measured carbonate system parameters in the two coastal seas with their air-equilibrated levels, wintertime and springtime carbonate system parameters mostly varied around the corresponding air-equilibrated levels (Figure 6), showing the dominant role of air-sea re-equilibration at the beginning of warm-season stratification formation. At the four deep stations within the central South Yellow Sea, however, the early-winter DIC and DIC:TAlk ratio were much higher than their airequilibrated values (Figures 6B,F), while the early-winter pH T and arag were substantially lower than their air-equilibrated values (Figures 6J,N). Supersaturated bottom-water f CO 2 values of 672-806 µatm were also detected at these four stations, despite their vertical profiles of temperature and salinity were nearly homogenous (Supplementary Figures S3-S5). This is because that a longer period of time of 25-100 days was needed for the CO 2 -rich waters to equilibrate with the atmosphere (Zhai et al., 2014b;Li and Zhai, 2019). In the North Yellow Sea, the early-spring pH T (8.10 ± 0.04) and arag (1.81 ± 0.13) were slightly higher than those air-equilibrated levels estimated at a low temperature of 5 • C (Figures 6I,M). This was likely because water-cooling lowered f CO 2 to form an undersaturated level of ∼350 µatm in April 2018 ( Figure 2D). The similarly springtime undersaturated f CO 2 has been observed earlier in the North Yellow Sea in May 2011 and May 2012 (Zhai et al., 2014b).

Summertime Respiration Beneath Thermoclines
In summer, most bottom-water DIC and DIC:TAlk ratio in the Yellow Sea and northern ECS were higher than their air-equilibrated levels (Figures 6A-C,E-G). In the YSCWM waters, summertime DIC addition and DIC:TAlk ratio increase relative to the corresponding air-equilibrated values were 10-50 µmol kg −1 and 0.1-0.2, respectively (Figures 6A,B,E,F). In comparison, the northern ECS showed much greater summertime increases in bottom-water DIC and DIC:TAlk ratio, which were 50-100 µmol kg −1 and 0.2-0.7 over the air-equilibrated values (Figures 6C,G). An abundant supply of sinking organic matter induced by the eutrophic Changjiang plume and summertime strong stratification were key factors forming and maintaining very high bottom-water DIC and pCO 2 in the ECS shelf waters in summer (Chou et al., 2009(Chou et al., , 2013b. It is worthwhile to note that several warm southern stations in the South Yellow Sea also exhibited very high DIC:TAlk ratio (∼0.93) relative to their air-equilibrated levels (∼0.89) at seawater temperature of >20 • C (Figure 6F), likely due to the northeastward-flowing CDW in summer (Xu et al., 2016;Zhai, 2018).
To reveal the possible source processes dominating these DIC additions, we plotted bottom-water DIC additions relative to the corresponding air-equilibrated levels (i.e., ExcessDIC) against AOU (Figure 5). In both of the YSCWM and the northern ECS, ExcessDIC versus AOU was reasonably in agreement with the stoichiometry of marine phytoplankton organic matter oxidation, including the traditional Redfield ratio, i.e., C: (-O 2 ) = 106:138, and the more recent estimation of C: (-O 2 ) = 106:154 (Hedges et al., 2002). The C: (-O 2 ) ratio of 106/154 falls in the range of marine phytoplankton respiration quotients of 106/140 and 106/160 recommended by Anderson (1995) and Sarmiento and Gruber (2006). These facts indicate that both bottom-water DIC additions and DO depletions were dominated by the community respiration beneath thermoclines.
In the ECS shallow areas where wind-driven collapse of water stratification occasionally occurred (Figure 3B), the bottomwater ventilation might bias the relationship of ExcessDIC and AOU ( Figure 5C) due to the slower re-equilibration of CO 2 than O 2 (Zeebe and Wolf-Gladrow, 2001). In the South Yellow Sea and northern ECS, some late-spring plots also followed the Redfield line (Figures 5B,C), likely due to an earlier start of metabolic processes in warm southern waters, as supported by those sinking biogenic particles induced by frequent springtime blooms in the outer Changjiang Estuary and the South Yellow Sea (He et al., 2013).
The summertime increases in bottom-water DIC and DIC:TAlk ratio suppressed the arag (Figure 8), while the bottom-water pH T declines in summer and/or autumn (Figures 6I-K) also mirrored the seasonal increases in bottomwater DIC (Figures 6A-C) and DIC:TAlk ratio (Figures 6E-G).
Therefore, the community respiration under thermoclines was mainly responsible for summertime and/or autumnal low pH T and arag in these regions.
In comparison to the Yellow Sea, the northern ECS exhibited greater increases in bottom-water DIC and DIC:TAlk ratio in summer, thereby leading to larger decreases in the bottomwater pH T and arag (Figure 6). It is worthwhile to note that summertime bottom-water pH T values in the northern ECS were lower than those in the Yellow Sea (Figures 6I-K), while summertime bottom-water arag values were roughly comparable in the two coastal seas (Figures 6M-O). The latter was partially caused by the lower yearly initial arag in the colder Yellow Sea waters in winter and spring (section "Wintertime Re-equilibration Naturally Preconditions the Water Column").

Autumnal Collapse of Thermoclines and the Water-Column Overturning
In typical autumn months (October and November), the seasonal stratification was usually weakened and even collapsed, except for the central Yellow Sea and a small ECS region affected by the  TWWC intrusion (Figures 3A,C). Correspondingly, the earlier CO 2 -rich and DO-depleted subsurface and/or bottom waters tended to approach their air-equilibrated levels (Figure 6). The similar autumn CO 2 -releasing process has been reported earlier in the northern ECS (Shim et al., 2007;Zhai and Dai, 2009;Guo et al., 2015).
In summary, three primary processes controlled seasonal variations of bottom-water DIC, DIC:TAlk ratio, pH T and arag in the Yellow Sea and northern ECS (Figures 6D,H,L,P). The air-sea re-equilibration played a fundamental role in wintertime carbonate system parameters. In summer, the community respiration beneath thermoclines dominated the increases in bottom-water DIC and DIC:TAlk ratio and the declines in bottom-water pH T and arag . Moreover, the autumnal collapse of thermoclines driven by cooling-induced watercolumn overturning or episodic wind-driven mixing events led to bottom-water ventilation, causing these carbonate system parameters to approach their air-equilibrated levels.

Comparison With Other Studies Relating to Seasonal Acidification in the Yellow Sea and the Northern East China Sea
In the Yellow Sea, respiration-induced declines in subsurface pH T , arag and DO from spring to autumn were first reported by Zhai et al. (2014b). Low pH T values of 7.79-7.90, low arag values of 1.13-1.40 and low DO% of 57-66% dominated subsurface waters of the central North Yellow Sea in autumn 2011 (Zhai et al., 2014b). During a summer flood in 2013, the community respiration coupled with freshwater dilution contributed to the occurrence of corrosive waters ( arag < 1) along the north coast of the North Yellow Sea, near the Yalu River estuary . Stratification was an important factor affecting the accumulation of respiration-induced CO 2 in the Yellow Sea, especially in the YSCWM (Xu et al., 2018). The respirationinduced CO 2 was accumulated during summer and autumn in the YSCWM-dominated area, leading to one third of the surveyed areas in the Yellow Sea to have pH T < 7.9 and arag < 1.5 in subsurface and bottom waters (Zhai, 2018). Also, a significant DO depletion occurred in the summertime and autumnal YSCWM, with the lowest DO% of 45% observed in autumn (Xu et al., 2016;Zhai, 2018). In addition, the bottom water on the Korean side of the South Yellow Sea was already undersaturated with aragonite (with DO% of ∼60%) in autumn (Choi et al., 2020), presumably due to ocean dumping of organic materials in the central Yellow Sea (Choi et al., 2020).
In the northern ECS off the Changjiang Estuary, low pH T values of ∼7.8 and low arag values of ∼1.5 have been reported earlier in its summertime hypoxic zone (Cai et al., 2011;Wang B. et al., 2017). The term hypoxia refers to a DO threshold of lower than 2 mg L −1 , that is, 63 µmol O 2 L −1 or approximately 30% saturation (Dauer et al., 1992;Rabalais et al., 2010). The summertime hypoxic area in the ECS was estimated at greater than 12,000 km 2 (or 432 km 3 volume, Chen et al., 2007), which was comparable to the largest coastal hypoxic zones observed in the world (Diaz, 2001;Fennel and Testa, 2019). To date, aragonite undersaturation has not been observed in the northern ECS, although the lowest summertime bottom-water arag value in the outer Changjiang Estuary has showed an alarming decline in recent years from 1.70 in July 2009 (Chou et al., 2013a) to 1.52 in July 2016 and to 1.39 in July 2017 (Xiong et al., 2019). Chou et al. (2013a) predicted that the combination of intensifying eutrophication and increasing atmospheric CO 2 would push the bottom water of the Changjiang plume area toward undersaturated with respect to aragonite ( arag ∼0.8) by the end of this century.
The integrated results of these fragmented studies have suggested that the present-day lowest bottom-water arag in the Yellow Sea was usually lower than that in the northern ECS. The Yellow Sea might be more vulnerable than the adjacent ECS shelf to the potentially negative effects of CO 2 acidification (Zhai et al., 2014b). However, the difference in behaviors of bottom-water DO depletion and CO 2 acidification between the two coastal seas was not investigated in previous studies.
In this new research, the lowest regional bottom-water arag values were detected at 1.07 in the Yellow Sea and 1.21 in the northern ECS (Figures 6M-O), which were generally in agreement with the previous results. However, the lowest bottomwater pH T and arag in the northern ECS observed in this study (7.68 and 1.21 in July 2018) were lower than those reported in previous studies. This might suggest that the respiration-induced CO 2 acidification was developing rapidly in the northern ECS bottom waters off the Changjiang Estuary. Note that our results of this study were based on a latest transregional investigation, spanning all four seasons. Seasonal evolutions and controls of bottom-water DO and carbonate system parameters in both the Yellow Sea and northern ECS were detailed in this study (Figures 2, 6). Moreover, the different behaviors of bottom-water DO depletion and CO 2 acidification in the two coastal seas will be discussed in section "Different Behaviors of Bottom-Water DO Depletion and CO 2 Acidification in the Two Contrasting Coastal Seas."

Different Behaviors of Bottom-Water DO Depletion and CO 2 Acidification in the Two Contrasting Coastal Seas
The community respiration beneath thermoclines induces both DO depletions and CO 2 accumulations, thus depressing bottomwater pH T and arag . The following stoichiometric relationships (assuming the usual Redfield ratio) were used to characterize this respiration-driven process: All acid-base conversions were based on the dominant species in the saline aquatic environment at a pH of ∼8.
The coastal hypoxia is usually associated with the respirationinduced CO 2 acidification (Cai et al., 2011;Melzner et al., 2013). However, the Yellow Sea bottom-water DO values were always much higher than the threshold for hypoxia, despite the occurrence of seasonal acidification (Figures 7A,B,D,E). In contrast, the northern ECS off the Changjiang Estuary is a welldocumented summertime hypoxic zone (e.g., Li et al., 2002;Zhu et al., 2011). In the ECS hypoxic zone, quite low pH T values of ∼7.7 and critically low arag values of < 1.5 were observed in summer (Figures 7C,F).
To investigate the different behaviors of bottom-water DO depletion and CO 2 acidification in the Yellow Sea and the northern ECS, we compared their yearly initial conditions and respiration processes, and then modeled seasonal changes in bottom-water DO, pH T and arag in the two contrasting coastal seas.

Wintertime Re-equilibration Naturally Preconditions the Water Column
At the beginning of the warm-season stratification formation, yearly initial DO values (i.e., air-equilibrated wintertime/springtime values) in cold Yellow Sea waters (∼300 µmol kg −1 at 9 • C) were higher than those in the warmer northern ECS waters (∼250 µmol kg −1 at 20 • C) (Figure 7). The yearly initial pH T values in the two coastal seas were similar (8.0-8.1, Figures 7A-C), whereas yearly initial DIC:TAlk ratio and arag values were quite different. The Yellow Sea had higher yearly initial DIC:TAlk ratio (0.925-0.930) and lower yearly initial arag values (1.8-1.9) than those in the northern ECS (DIC:TAlk ratio ∼0.9 and arag ∼2.5) (Figures 6E-G, 7D-F).
In these cold seasons that were free from community respiration, the north-to-south decreasing gradient of yearly initial DIC:TAlk ratio was dominated by the CO 2 solubility. Seawater CO 2 solubility increased with the decrease of temperature (Weiss, 1974), resulting in lower [CO 3 2− ] and arag in colder waters compared with warmer waters. Note that the internal temperature effect only accounts for < 5% of the variations in arag (Zhai et al., 2014b). With respect to pH T , however, the CO 2 -dissolving induced [H + ] increase in cold waters was greatly canceled out by the [H + ] decrease due to the internal temperature effect of pH Cai et al., 2020). Therefore, the yearly initial pH T varied minimally along the north-to-south latitude gradient. Even over diverse ecosystems found between the poles and tropics, the air-equilibrated pH T values changed limitedly in a narrow range of 8.020-8.074 (Hofmann et al., 2011). In a word, the air-equilibrated arag was much more sensitive to seawater temperature than the air-equilibrated pH T . Wintertime re-equilibration naturally preconditioned the water column to have different yearly initial values of DO, pH T , arag and DIC:TAlk ratio.

Different Respiration Rates and Duration
Not only yearly initial values of DO and carbonate system parameters, but also respiration rates and duration were quite different between the two coastal seas. The Yellow Sea (especially within the YSCWM) showed higher bottom-water AOU in autumn than in summer (Figures 5A,B). In the northern ECS, bottom-water AOU reached the maximum value in summer and then sharply declined in autumn ( Figure 5C). Although the respiration duration beneath thermoclines in the northern ECS was shorter than the Yellow Sea, nearly one third of summertime bottom-water AOU values (with a range of 113-179 µmol kg −1 ) in northern ECS was about 24-97% higher than autumnal bottom-water AOU values in the Yellow Sea (91 ± 15 µmol kg −1 ). Also, the bottom-water averaged apparent DO depletion rate in the northern ECS (1.8 µmol O 2 kg −1 d −1 ) were about three times higher than that in the Yellow Sea (0.6 µmol O 2 kg −1 d −1 ), indicating a much higher net community respiration rate in the northern ECS as compared to the Yellow Sea.
In the Yellow Sea, the low bottom-water averaged apparent DO depletion rate of 0.6 µmol O 2 kg −1 d −1 was comparable to the net community respiration rate previously observed in the northern area of the YSCWM (Zhai et al., 2014b;Li and Zhai, 2019). In the northern ECS, however, its bottom-water averaged apparent DO depletion rate of 1.8 µmol O 2 kg −1 d −1 was lower than the lower limit of field-measured community respiration rates (∼3 µmol O 2 kg −1 d −1 ) on the basis of DO decreasing in dark incubation experiments (Chen et al., 2006;Zhu et al., 2016). A quick formation of bottom-water hypoxia was observed off the Changjiang Estuary within only 6 days after a typhoon disturbance (Wang B. et al., 2017), equivalent to an extremely high DO drawdown rate of 22 µmol O 2 kg −1 d −1 . Both bottle incubation results and the field data deduction suggest that the northern ECS should have relatively high net community respiration rates in its summertime oxygen-depleted bottom waters. However, the central Yellow Sea accumulated respiration products beneath the thermocline in summer and autumn, while the northern ECS bottom waters preserved them only in summer.

Coupling of Bottom-Water DO Depletion and CO 2 Acidification
To further explore the differentiation of bottom-water DO depletion and CO 2 acidification in the Yellow Sea and northern ECS, seasonal changes in bottom-water DO, pH T and arag were modeled based on Eq. (3) and different yearly initial values and respiration rates in the two coastal seas. When the respirationinduced DO depletion process started, DIC changed according the traditional Redfield ratio, i.e., DIC: AOU = 106:138, and the impact of respiration on TAlk ( TAlk: DIC = -17/106) was also considered. During calculation, as detailed above, yearly initial conditions in the North Yellow Sea were DO = 300 µmol O 2 kg −1 , DIC:TAlk ratio = 0.930; in the South Yellow Sea, yearly initial DO = 300 µmol O 2 kg −1 , yearly initial DIC:TAlk ratio = 0.925; in the northern ECS, yearly initial DO = 250 µmol O 2 kg −1 , yearly initial DIC:TAlk ratio = 0.900. The net community respiration rates were assumed to be 0.6 µmol O 2 kg −1 d −1 in the Yellow Sea and 2 µmol O 2 kg −1 d −1 in the northern ECS. Temperature, salinity and TAlk were set to the mean value in bottom waters in the given region: T = 9 • C, S = 32.1, and TAlk = 2316 µmol kg −1 in the North Yellow Sea; T = 9 • C, S = 32.1, and TAlk = 2305 µmol kg −1 in the South Yellow Sea; T = 20 • C, S = 33.0, and TAlk = 2243 µmol kg −1 in the northern ECS. The effect of mixing with the northeastmonsoon-driven ZFCC was not included (affected only several nearshore stations, Figure 1G), while the effect of the yearround TWWC intrusion was included. Starting conditions for the TWWC-affected offshore waters were DO = 200 µmol kg −1 and DIC:TAlk ratio = 0.890.
Field-measured bottom-water pH T versus DO and arag versus DO reasonably followed those Redfield lines in both the Yellow Sea and northern ECS (Figure 7). The Redfieldbased coupling of DO depletion and CO 2 acidification suggested again that both pH T and arag declines in the two coastal seas were controlled by the community respiration beneath thermoclines. At those low-oxygen levels, the Yellow Sea data showed consistently lower bottom-water pH T and arag values than those in the northern ECS (Figure 7). It is worthwhile to note that field-measured observations in the northern ECS fell between the local reaction in the northern ECS (solid gray curves) and the TWWC-affected reaction (solid black curves, Figures 7C,F), suggesting the both the local respiration and the TWWC intrusion influenced DO and carbonate system parameters in the northern ECS.
In summary, the Yellow Sea had higher yearly initial DO values, higher yearly initial DIC:TAlk ratio but lower yearly initial arag values than those in the northern ECS. However, yearly initial pH T values exhibited only a few differences between the two coastal seas. The different behaviors of yearly initial pH T and arag were consistent with their different responses to the temperature variability. Moreover, higher CO 2 solubility together with the respiration-induced CO 2 additions resulted in the colder Yellow Sea waters to have higher DIC:TAlk ratio and thus lower arag as compared with those in warm northern ECS waters. Assuming that the DIC:TAlk ratio difference (0.025) between 0.925 in wintertime/springtime Yellow Sea waters ( Figure 6F) and a typical open-ocean seawater DIC:TAlk ratio of 0.900 ( Figure 6G) was only induced by CO 2 solubility difference, and considering the excellent linear relationship between bottomwater arag and DIC:TAlk ratio, the contribution of the yearly initial high CO 2 solubility to summertime low arag in the Yellow Sea (with the averaged DIC:TAlk ratio of 0.944 in summer, section "DIC and DIC:TAlk Ratio") was estimated to be 0.025/0.044 × 100% = 57%. Similarly, the contribution of the yearly initial high CO 2 solubility to autumnal low arag in the Yellow Sea (with the averaged DIC:TAlk ratio of 0.959 in autumn, section "DIC and DIC:TAlk Ratio") was estimated to be 0.025/0.059 × 100% = 42%.

Concurrent Hypoxia and CO 2 Acidification in the Northern ECS off Changjiang Estuary
In 2018, we observed concurrent hypoxia (DO 49-63 µmol O 2 kg −1 ) and CO 2 acidification (pH T 7.68-7.71, arag 1.28-1.39) at three sampling sites in the northern ECS off the Changjiang Estuary in the middle of July, although the modeling prediction suggested that the bottom-water hypoxia (DO 61 µmol O 2 kg −1 ) and CO 2 acidification (pH T 7.59, arag 1.05) should concurrently occur in early August (Figures 7C,F). Apart from the intensive respiration rates (section "Different Respiration Rates and Duration"), the excess DO decline was likely attributed to those non-local drivers of summertime hypoxia in the northern ECS bottom waters, such as the TWWC intrusion (Figures 7C,F, Wei et al., 2015;Wang B. et al., 2017) and the nearshore upwelling of the Kuroshio branch (Yang et al., 2011(Yang et al., , 2012(Yang et al., , 2013Qian et al., 2017). Sediment oxygen consumption may also contribute to the hypoxia formation in the northern ECS (Zhang et al., 2017;Zhou et al., 2017).
In the Yellow Sea, declines in DO, pH T , and arag in the Yellow Sea were much less than those in the northern ECS (Figure 7). Especially the modeled October DO in Yellow Sea was ∼150 µmol O 2 kg −1 higher than the modeled August DO in the northern ECS. Even if the water stratification could have lasted in the central Yellow Sea (maintaining the YSCWM) until December, the modeling prediction of bottomwater DO in December (170 µmol O 2 kg −1 ) would be still quite high, despite that extremely low pH T and arag values (7.73 and 0.96, respectively) would dominate the YSCWM in this early winter month. The modeled results in December were comparable with those field-measured values at the four deep stations (in the central South Yellow Sea) during our early winter survey (Figures 7B,E). Therefore, the seasonal hypoxia that usually associated with the respiration-induced coastal CO 2 acidification did not occur in the present-day central Yellow Sea.
Recently, studies relating the concurrency of bottom hypoxia and CO 2 acidification were conducted in many other coastal zones. For example, summertime hypoxia and acidification (pH < 7.7) covered almost one-third of an aquaculture area in southern nearshore waters of the North Yellow Sea (Zhang et al., 2018). In the central Bohai Sea, the near-hypoxic (DO < 90 µmol O 2 kg −1 ) and concurrent low pH T (∼7.7) and arag (∼1.3) were observed in late summer Song et al., 2020). In the northern South China Sea shelf, summertime low pH values were mainly measured nearshore in the hypoxic zone, along with relatively high DIC concentrations . In the hypoxic zone of the northern Gulf of Mexico, the lowest bottom DO value obtained in summer 2010 (only at 11 µmol O 2 kg −1 ) matched with the lowest pH T (7.60) and arag (1.63) values (Hu et al., 2017). Even in the Chesapeake Bay, both anoxia and the production of hydrogen sulfide (H 2 S) were observed in summertime subsurface waters (with pH T < 7.6, Cai et al., 2017). Diverging from all the above-mentioned situations, in the well-buffered Corpus Christi Bay located along the Texas coast, relatively high pH T (>7.8) and arag (>2.6) values were revealed in its summertime hypoxia zone (McCutcheon et al., 2019). The high buffering capacity in the Corpus Christi Bay was attributed to abundant seagrass meadows and the strong evaporation that introduced low pCO 2 , high pH T and arag waters (McCutcheon et al., 2019). Hence, the concurrent occurrence of coastal hypoxia and CO 2 acidification does not always hold true in all situations. Such issues are worthy of further investigation in order to be better understood.

Correlations Between Bottom-Water arag Versus DIC and DIC:TAlk Ratio
We also found that bottom-water arag was negatively correlated with DIC (Figures 8A-C). The correlations in the Yellow Sea and northern ECS roughly fit in with several ideal relationships between arag and DIC that were estimated using CO2SYS.XLS. During simulation, regional mean TAlk with standard deviation ( Table 1 and Figures 4A-C) and typical wintertime and springtime DIC:TAlk ratio (Table 1 and Figures 6E-G) were set for starting conditions, while fixed regional mean salinity and temperature were used (Figures 2A-C). At a given DIC, arag decreased along with TAlk decrease. An early-spring plot near the Yalu river estuary with moderately low TAlk value of 2265 µmol kg −1 deviated from the three ideal relationships in the North Yellow Sea (Figure 8A). Also, several summertime plots affected by the CDW with low TAlk values of 2110-2183 µmol kg −1 and autumnal plots affected by the ZFCC with low TAlk values of 2138-2185 µmol kg −1 deviated from the two ideal relationships in the northern ECS ( Figure 8C).
By comparison, the correlation between arag and DIC:TAlk ratio (Figures 8D-F) was much tighter than that between arag and DIC. This was because that DIC and TAlk were inversely related to arag with higher DIC decreasing arag and higher TAlk increasing arag in seawater (Wanninkhof et al., 2015). Seawater DIC:TAlk ratio, by definition, could reflect the fraction of free CO 2 in the DIC pool (Cai et al., 2020), which essentially determined arag . When more CO 2 was dissolved in seawater (solubility and/or respiration induced CO 2 additions) as indicated by higher DIC:TAlk ratio, more CO 3 2− ions were titrated, resulting in lower arag values. On condition that the DIC:TAlk ratio varied between 0.83 and 0.95, the correlation was expressed linearly as arag = -24.08 × (DIC:TAlk)+24.3 (r = 0.992, n = 832), exhibiting little differences among those diverse waters involved in this study (Supplementary Figure S8). It might suggest that DIC:TAlk ratio was a qualified indicator of arag in seawater, even in contrasting coastal waters.

SUMMARY AND IMPLICATIONS
Seasonal variations and the controls of bottom-water carbonate system parameters and DO in the Yellow Sea and northern ECS were examined in this study. Wintertime air-sea reequilibration, summertime respiration under thermoclines and autumnal water-column overturning are three primary processes controlling seasonal variations in these parameters. At the beginning of warm-season stratification formation, the colder Yellow Sea waters had higher DO values but lower arag values than those in the relatively warmer northern ECS waters, while yearly initial pH T values were similar in the two coastal seas. During warm seasons, the Yellow Sea accumulated respiration products under the thermocline in summer and autumn, while the northern ECS bottom waters preserved them only in summer. Higher CO 2 solubility together with the respiration-induced CO 2 additions caused the colder Yellow Sea waters to have higher DIC:TAlk ratio and thus lower arag as compared with those in warm northern ECS waters. The seasonal hypoxia that usually associated with the respiration-induced coastal CO 2 acidification did not occur in the central Yellow Sea. In contrast, the summertime hypoxia and CO 2 acidification concurrently occurred in the northern ECS.
We could also evaluate how seawater acidity ([H + ]) in the Yellow Sea and northern ECS along the north-to-south latitude gradient respond to coastal acidification, considering starting conditions for the North Yellow Sea (T = 6 • C, S = 32.1, TAlk = 2316 µmol kg −1 , DIC:TAlk = 0.930), for the South Yellow Sea (T = 9 • C, S = 32.1, TAlk = 2305 µmol kg −1 , DIC:TAlk = 0.925), and for the northern ECS (T = 17 • C, S = 33.0, TAlk = 2243 µmol kg −1 , DIC:TAlk = 0.900). These starting conditions were those late spring levels in the given regions when water stratification was just formed. For an increase of seawater f CO 2 of 1 µatm, [H + ] increased by 1.97 × 10 −5 µmol kg −1 in the North Yellow Sea, by 1.94 × 10 −5 µmol kg −1 in the South Yellow Sea and by 1.90 × 10 −5 µmol kg −1 in the northern ECS, indicating a north-to-south decline in the sensitivity to coastal acidification stresses, related to atmospheric CO 2 intrusion and/or community respiration induced CO 2 addition. A similar north-to-southdecreasing gradient along the U.S. East Coast and Gulf of Mexico has been reported by Wang et al. (2013) based on a summertime cross-shelf observation. The [H + ] increase related to an increase of seawater f CO 2 of 1 µatm has been estimated to be 2.0 × 10 −5 µmol kg −1 in the Gulf of Maine, 11% greater than that in the Gulf of Mexico (1.8 × 10 −5 µmol kg −1 ). The northeastern U.S. shelf waters appear more susceptible to CO 2 acidification than their southern counterparts, likely due to relatively low temperature and thus high CO 2 solubility in the northeastern regions. Recent studies have predicted that waters unfavorable to calcification will exist year round in the subsurface waters within northern shelf waters along China's and U.S. eastern coasts, i.e., the North Yellow Sea  and Gulf of Maine (Wang Z. A. et al., 2017), by the middle of the century under the future ocean acidification. Consequently, cold northern shelf waters appear to be more vulnerable to the potentially negative effects of ocean acidification. In the future, more studies are needed to quantify ecological responses of different coastal zones to those acidification pressures, particularly in the relatively colder northern shelf waters.

DATA AVAILABILITY STATEMENT
The mapping dataset of dissolved oxygen and carbonate system parameters reported in this study is available at figshare.com via doi: 10.6084/m9.figshare.12630335.

AUTHOR CONTRIBUTIONS
WZ designed the research and performed one of the field surveys. TX performed three field surveys and analyzed the data set and drafted an early version of the manuscript. QW, CL, and YZ separately performed the other field surveys. SW, SL, SY, TX, and CL contributed to the data collection. All authors contributed to discussion and revision of the manuscript.