Persistent Continental Shelf Carbon Sink at the Ieodo Ocean Research Station in the Northern East China Sea

Hourly (2017–2021) to seasonal (2015–2021) inorganic C data were collected at the Ieodo Ocean Research Station (32.07°N and 125.10°E) in the northern East China Sea (ECS), located under the influence of the nutrient-rich Changjiang Diluted Water (CDW). An increase in phytoplankton biomass from April to mid-August (the warming period) equalized much of the temperature-driven increase in the surface pCO2 and thus, made the northern ECS a moderate sink of atmospheric CO2. From November to March (the cooling period), a large pCO2 reduction, driven by a temperature reduction, and a high air–sea CO2 exchange rate, because of high windspeeds, transformed the basin into a substantial CO2 sink, yielding an annual net C uptake of 61.7 g C m–2 yr–1. The effects of biological production and temperature change on seawater pCO2 (and thus, the net air–sea CO2 flux) were decoupled each season and acted in concert to increase the net annual CO2 sink by the region. The present study provided the observational and mechanistic lines of evidence for confirming “continental shelf C pump”—a mechanism in the shallow waters of the continental shelves that accumulate a significant amount of C (via reinforced cooling and promoted biological C uptake) that is transported from the basin surface waters to the interior of the adjacent deep ocean. In the future, an increasing input of anthropogenic nutrients into the northern ECS is likely to make the region a stronger CO2 sink.


INTRODUCTION
Unlike major ocean basins, the role of coastal and marginal seas as an anthropogenic CO 2 reservoir has been studied less because the small surface area of the global marginal seas (only 7% of the world's ocean area) may only store small amounts of anthropogenic CO 2 . Contrary to this paradigm, the marginal seas can absorb substantial amounts of anthropogenic CO 2 because the primary producers in those locations use high nutrient loads from adjacent continents and hence, accelerate CO 2 transfer by substantially reducing the concentration of surface C (Lee et al., 2011;Najjar et al., 2018). Moreover, the interaction between the resultant planktonic organic matter deposited onto the shallow sediments and the overlying seawater can generate alkalinity, facilitating the transfer of additional anthropogenic CO 2 (Thomas et al., 2009). Intrinsically, the air-sea CO 2 flux in the coastal and marginal seas is subject to large variabilities with time and location because the complex interactions between natural (e.g., tidal cycle, upwelling, wind-driven mixing, freshwater input, eutrophication, and biological activity) and anthropogenic (e.g., inputs of anthropogenic C and nutrients and ocean temperature increase) factors make estimating the air-sea CO 2 flux challenging and often unsuccessful (Park et al., 2006;Dai et al., 2013;Cao et al., 2019). A major hurdle in carrying out such estimates is our inability to capture all major variabilities in the air-sea CO 2 flux over the extensive periods covering all seasons.
No exception to this generalization is the East China Sea (ECS), which has an extensive continental shelf covering twothirds of its area and is surrounded by populated countries (i.e., China, Korea, and Japan). The CO 2 emissions from these countries fall within the top tier; however, the successful execution of reforestation in these countries has led to a substantial C uptake by the terrestrial biosphere (Choi et al., 2002;Piao et al., 2009). The ECS is teeming with plankton as a result of nutrient input by the Changjiang River. The addition of these river-borne nutrients has divided the ECS into two distinct regions in terms of nutrient concentration: the northern ECS, which is characterized more by the nutrient input from the Changjiang River (anthropogenic influence), and the central and southern basins, which are more influenced by water exchange with the Kuroshio Current, which lacks nutrients (natural influence) . Tropical cyclones are also common in the ECS in summer and fall. To determine if the entire ECS is a net sink or source of atmospheric CO 2 , the determination of the amount of CO 2 that enters the northern ECS (i.e., the areas located between 28.5-33.0°N and 122.5-126.0°E) is critical because this area can be a large C sink as a result of the proliferation of phytoplankton.
Another factor equally acute in determining seawater pCO 2 (and thereby net air-sea CO 2 flux) in the northern ECS is the seasonal temperature change. The seasonal temperature extremes in the ECS and the adjacent Yellow Sea ranged from 10°-20°C (colors in Figure 1), which are likely the largest seasonal temperature swings in the world's oceans and arise from a seasonal switchover between a hot and humid North Pacific air mass governing the summertime temperature of the ECS (also the Yellow Sea) and a cold and dry Siberian air mass governing the winter temperature of the same basins (Chang, 2004). From a thermodynamics perspective, a pCO 2 change of up to 190 μatm can occur when transitioning from summer to winter.
This study reports the results of 5 years (2017-2021) of highfrequency (obtained at 1 h intervals) C data collected at a midpoint (the Ieodo Ocean Research Station, also known as Socotra Rock and Suyan Islet) between the mouth of the Changjiang River and Jeju Island, located downstream of the Changjiang freshwater plume in the northern ECS. The results of 7 years (2015-2021) of seasonal data obtained between Jeju Island and the Ieodo Ocean Research Station were added to substantiate the interpretation of the high frequency data. In parallel, we examined published data covering the northern ECS and concluded a basin-wide status of C sink. We also explored the key drivers of the C-sink status.

Measurements of Total Alkalinity (A T ), Total Dissolved Inorganic Carbon (C T ), and pH
During the study period (2015)(2016)(2017)(2018)(2019)(2020)(2021), at more than 50 locations between Jeju Island and the Ieodo Ocean Research Station (32.07°N and 125.10°E; halfway between the Changjiang River estuary, China and Jeju, Korea; hereafter referred to as "Ieodo"; Figure 1), we collected surface samples at seasonal intervals (April to December with no samples taken from January to March) for measurements of seawater C parameters. The measured C parameters were  total alkalinity (A T = ½HCO − 3 + 2½CO 2− 3 + ½B(OH) − 4 + proton acceptors − ½H + − proton donors), total dissolved inorganic carbon (C T = ½CO 2 + ½HCO − 3 + ½CO 2− 3 ), and pH. The concentrations of A T and C T were determined using potentiometric and coulometric titrations, respectively. The precision of the A T and C T measurements were determined to be ± 1.8 mmol kg −1 for A T and ± 2.2 mmol kg −1 for C T (Supplementary Figure 1) by titrating reference materials with certified A T and C T values (provided by A. Dickson, Scripps Institution of Oceanography, USA). Small deviations of the measured A T and C T values on the reference materials from their certified values were either subtracted or added for sample measurements conducted during the same period in which the reference materials were analyzed. The pH was measured at 25°C spectrophotometrically using the m-cresol purple indicator following the procedure documented elsewhere (Clayton and Byrne, 1993;Lee et al., 1996).
Other CO 2 parameters (i.e., CO 2 partial pressure, pCO 2 ; seawater saturation state with respect to aragonite, W arag ) were calculated from the values of A T and C T using the set of the thermodynamic constants, including the carbonic acid dissociation constants reported by Mehrbach et al. (1973) and other ancillary constants proposed by Millero (1995). The thermodynamic model, including these dissociation constants, led to the best agreement between measurements and calculations involving all three measured parameters (A T , C T , and pH), as demonstrated in previous studies (McElligott et al., 1998;Lee et al., 2000;Lueker et al., 2000;Millero et al., 2006;Fong and Dickson, 2019). The agreement was compelling evidence for insignificant biases in A T derived from organic acids (Cai et al., 1998;Kim and Lee, 2009;Ko et al., 2016) and phytoplankton and bacteria cells . The B:Cl ratio of 0.2414 confirmed for the open ocean (Lee et al., 2010) was shown to be accurate in the study area and in experiments designed to predict the borate contribution to A T (Lee et al., 2019).

Continuous Seawater pH Measurements
Beginning in September 2017, seawater pH was measured at a depth of 4 m at 1 h intervals using a pH sensor package equipped with an ion-sensitive field effect transistor (ISFET), which produces stable pH signals; the sensor package also included a data logger based on a Honeywell Durafet pH sensor. The ISFET sensor was determined to be precise within ±0.005 pH units over periods of weeks to months. The pH sensor package (including temperature and salinity) was protected by a perforated copper guard, which deterred biofouling. The pH values were determined on the total scale at in situ temperature and salinity. During the period of deployment, we visited the site at seasonal intervals to check for signs of sensor drift. At each visit, we sampled seawater in close proximity (within the 10 m) of the sensor package 3 to 5 times per day over 2 days and analyzed their pH and salinity in the laboratory. A total of 70 samples were collected at Ieodo during a total of 8 visits over a 4 year period. A comparison between the sensor-based and laboratory-based pH measurements indicated that the sensor-based pH data showed no sign of drift over 6 months and were consistent with the laboratory measurements to within ±0.03 pH units (Supplementary Figure 2). During each site visit, we downloaded the pH, salinity, and temperature data from the data logger, thoroughly cleaned the sensor package, and then redeployed it.

Polynomial Model for Predicting A T
Empirical algorithms that relate A T to sea surface salinity (S) and sea surface temperature (T) have been used to construct the distribution of A T in regions  and have greatly facilitated predictions of the CO 2 flux across the air-sea interface, where it combines with another carbon parameter. A functional form for fitting A T data was derived from datasets obtained within the 10 km of Ieodo (n = 162) by testing its polynomial form (first or second order) with two predictor variables (S and T) which gave a fit. (1) where 31.2 and 21.8 represent the mean values of S and T (°C) values, respectively. We found that the second-order polynomial model yielded the lowest value of the root-mean-square error (RMSE: the standard deviation of the predicted errors; 9.8), the highest value of r 2 (0.92), and the lowest value of p (< 0.05) ( Figure 2). Because the A T data (used to derive the fit) covered four seasons and seven years, the fit would represent seasonal and interannual variations in A T .

Air-Sea CO 2 flux
The air-sea CO 2 flux (mol C area −1 time −1 ) across the air-water interface was calculated using the formula: K 0 · k(pCO air 2 − pCO sw 2 ). In this formula K 0 (mol L −1 atm −1 ) is the CO 2 gas solubility, and k (cm hr −1 ) is equal to 0.251 U 2 10 (Sc/660) −0.5 , where U 10 (m s −1 ) is the windspeed at 10 m above sea level and Sc is the Schmidt number (Wanninkhof, 2014). Windspeed data measured at 33.08°N and 126.03°E, in the vicinity of the sampling site, are available at the National Climate Data Center. Air pCO air 2 data were obtained from the observatory at Anmyeondo (36.54°N, 126.33°E), which is operated by the Korea Meteorological Administration. Seawater pCO sw 2 data were calculated from pH measurements and A T values derived from Equation (1), and thus had a probable error of < 30 μatm because of pH and A T errors (±0.03 and ±10 μmol kg −1 , respectively). The resulting pCO sw 2 error was directly translated into a flux error of ±0.002 g C m −2 hr −1 .

Variations in Seawater C Parameters between Ieodo and Jeju Island
A notable feature found between Ieodo and Jeju Island in the northern ECS, located downstream of the Changjiang Diluted Water (CDW), is the large seasonal variability in salinity (26-35) ( Figure 3A). Much of the seasonal variability is directly associated with changes in the intrusion of the CDW sourced from the Changjiang Estuary. Specifically, the values of salinity at Ieodo decreased by as much as S = 9, which corresponds to the maximum intrusion of the CDW into the study area between July and August. The decrease in salinity also varied considerably interannually and was found to be the highest at Ieodo and to decrease toward Jeju Island. This salinity decrease was broadly proportional to the increase in intrusion of the CDW. The intrusion of the CDW into the study area was particularly pronounced in the summer of 2017, when the CDW propagated eastward and even reached the coastal waters near Jeju Island, close to southern Korea.
The intrusion of CDW into the northern ECS affected not only the salinity but also all of the C parameters, including in situ pCO 2 (175-490 μatm), pH (8.0-8.3), and W arag (2.0-5.8) ( Figures 3D-F). The gradients in four C parameters observed between Ieodo and Jeju Island were particularly pronounced during the summer months (June-August). The summertime pCO 2 values at Ieodo often remained less than 300 μatm, considerably lower than the current atmospheric pCO 2 of~420 μatm and thus, the air-sea pCO 2 disequilibrium was highest at Ieodo and decreased toward Jeju Island ( Figure 3D). The summertime pH distribution was opposite to that of pCO 2 because both parameters changed in opposite directions in response to changes in the same environmental factors (e.g., temperature, salinity, photosynthesis, and respiration) ( Figure 3E). The highest W arag values were observed at Ieodo and decreased toward Jeju Island ( Figure 3F). One striking characteristic of the CDW was unusually high values of the salinity-normalized A T (nA T ) and salinity-normalized C T (nC T ) ( Figures 3B, C), which were a direct consequence of high A T and C T values of the Changjiang River at salinity ≈ 0. The greater contribution of the Changjiang River led to the higher nA T and nC T values of the CDW. Conversely, with increasing distance from the Changjiang Estuary, the values of nA T and nC T in the CDW rapidly decreased as a result of mixing with the warm current of Tsushima with low nA T and nC T values. In other months, the Ieodo-Jeju gradients in salinity and all of the CO 2 parameters either substantially decreased or disappeared completely.
Seawater pCO 2 and Net Air-Sea CO 2 Flux at Ieodo The low levels of summertime pCO 2, between Ieodo and Jeju, were further substantiated by continuous pCO 2 measurements at Ieodo ( Figure 4C). During the warming period (from April to mid-August), the surface pCO 2 values at Ieodo remained~125 μatm lower than the atmospheric pCO 2 . The magnitude of the air-sea disequilibrium differed slightly by year. However, the basin switched to a neutral condition in terms of the seawater pCO 2 level from September to October, with large temporal variability, which made the status of the basin as a net C sink or a net C source ambiguous. Moreover, the timing of the switchover from the lower-than-atmospheric pCO 2 condition to the higherthan-atmospheric pCO 2 condition differed by year. The switchover occurred in early September in 2019 ( Figure 5C); however, it occurred late August in 2020, when the site was hit by Typhoon Bavi (Figure 5D).
Following this short transition period, the surface pCO 2 levels progressively decreased as the season transitioned to a full winter condition and were persistently 35-130 μatm lower than the atmospheric pCO 2 levels during the cooling period (November-March), with small temporal variations ( Figure 4C). The measured decrease in surface pCO 2 from the late-summer condition to the full winter condition was approximately consistent with thermodynamic predictions based on the seasonal temperature decrease ( Figure 6B). Both the undersaturated pCO 2 condition and high windspeeds led to a persistent C sink status of the basin ( Figure 4A, D). Collectively, the year-round continuous data (2020-2021) show that the perennial C sink during both the warming (0.13 g C m −2 d −1 ; from April to mid-August) and the majority of the cooling periods (0.32 g C m −2 d −1 ; from November to March), along with a small source for a short transition period, led the region to a strong annual sink of atmospheric CO 2 (61.7 g C m −2 yr −1 ).

Effects of Typhoons on Net Air-Sea CO 2 Flux at Ieodo
In the period 2019-2020, Ieodo was hit by seven major typhoons (four typhoons in 2019 and three in 2020). The ocean CO 2 system responded to those typhoons differently, depending on the timings of the typhoons' passage. The three typhoons that struck the site during the warming period accelerated ocean CO 2 uptake during their passages (see Figures 5A, B). The warmingperiod enhancement of ocean CO 2 uptake during the typhoon passage was an order of magnitude greater than those for periods either before or after the typhoon passage. The greater ocean CO 2 uptake during the typhoon passage was a consequence of the combined effects of sustained low pCO 2 levels and a high rate of CO 2 exchange. To our surprise, typhoons exerted the opposite effect on the net air-sea CO 2 flux during the cooling period, when oceanic pCO 2 levels were higher than the atmospheric levels ( Figures 5C, D). The typhoon-induced high surface pCO 2 levels resulted in unusually high effluxes (~7 times greater than the CO 2 fluxes that occurred during the typical cooling period).

Status of C Uptake by the ECS
In the ECS, the major controls of the seasonality in terms of the air-sea CO 2 flux differed regionally: the northern ECS was dominated by biological activities and ventilation of water, whereas the central and southern ECS were mainly determined by seawater temperature (Guo et al., 2015;Deng et al., 2021). With regard to the status of net air-sea CO 2 flux across the entire ECS, mixed conclusions have been reported in literature, in part because the datasets that led to differing conclusions did not cover time and space extensively (Deng et al., 2021 and references therein). Specifically, the datasets were skewed to either the Chinese coastal waters under the direct influence of the nutrient-rich CDW (Li et al., 2021;Wu et al., 2021) or the offshore waters distant from the CDW influence (Shim et al., 2007;Kim et al., 2013). Overall, the entire ECS appeared to be a slight C sink on an annual scale (Wang et al., 2000;Zhai and Dai, 2009;Guo et al., 2015;Wu et al., 2021), although it can be a weak C source when seasonal dynamics of river input and physical mixing occasionally induce C release in summer and fall in the inner shelves of the ECS (the Changjiang freshwater plume).
In comparison with the previous estimates based on either the data in a specific month or an empirical model (Wang et al., 2000;Tseng et al., 2011;Guo et al., 2015), our year-round measurements (2020)(2021) in the northern ECS yielded the annual net air-sea CO 2 flux of 61.7 g C m −2 yr −1 (= 14.1 mmol C m −2 d −1 from air to sea), which was 2-3 times higher than the values reported for the adjacent regions. The air-sea CO 2 influx in summer (4.7 mmol C m −2 d −1 ) from these year-round measurements agreed with the reported results of 6.5 to 4.6 mmol C m −2 d −1 (4.9 to 3.4 mmol C m −2 d −1 using the same k used in this study) associated with the river plum and the outer estuary, respectively (Guo et al., 2015). However, we found large differences in the net air-sea CO 2 fluxes for spring and winter: the net air-sea CO 2 fluxes for spring (20.3 mmol C m −2 d −1 ) and winter (31.5 mmol C m −2 d −1 ) were 2-4 times higher than the previous estimates Guo et al., 2015).

Mechanisms Responsible for C Uptake by the ECS
Both high plankton biomass in spring and summer and a large temperature drop from summer to winter, collectively, contributed to the substantial C sink by the northern ECS; their effects on the net air-sea CO 2 flux were decoupled in terms of season. An increase in planktonic biomass during the growing season spread vertically into the shallow shelf water (< 70 m depth) and the resultant organic matter was either buried into the shallow sediments or moved laterally. These biological activities lowered the water-column pCO 2 levels, which were lower than the atmospheric pCO 2 levels throughout the warming months (April-August), except for a short transition time to the cooling period, during which the surface pCO 2 reached levels similar to or slightly greater than the atmospheric levels. In particular, a rapid increase in planktonic biomass during the warming period counteracted nearly all the temperature-driven pCO 2 increase (~85%, Figure 6A). As the season developed into a full winter condition, the water column was substantially cooled by as much as 17°C (August-March, Figure 4B) and this temperature drop lowered the water column pCO 2 values by as much as 175 μatm at Ieodo. The observed reduction of the surface pCO 2 during the cooling months was surprisingly consistent with the thermodynamic expectations (> 75%, Figure 6B). In particular, during the winter, both the low pCO 2 levels and high air-sea CO 2 exchange rates, driven by a large temperature drop and high windspeeds, respectively, acted in concert to increase the C uptake by the northern ECS and made the basin an even stronger C sink. Reported water properties of the northern ECS appeared to be an additional favorable factor for the northern ECS to be a perennial C sink-the continental shelf pump (Tsunogai et al., 1999). Isopycnal mixing (via diffusive and advective flows) in the ECS tends to transport the shelf waters to the deeper layers of the Kuroshio region, which transports dissolved inorganic C to the open ocean. A large temperature drop from summer to winter in the shallow continental shelf zone, including our study area, resulted in greater CO 2 flux into the ocean (resulting from high CO 2 solubility at low temperatures) and the rapid formation of dense water (resulting from rapid heat loss while cooling). Both of these factors enable the effective transport of C from the shallow northern ECS to the open ocean. The C export from the northern ECS to the open ocean was also reported by Chen and Wang (1999) to be in the form of both particulate and dissolved organic C. The organic C produced by phytoplankton on the ECS shelf was exported as DOC (8%) to the offshore water column or exported as POC (10%) to the shelf and offshore sedimentary environments to the shelf and offshore sedimentary environments. The downslope transport of POC accounted for~40% of the inorganic C transport to offshore. The transport of the shelf water to the open ocean via isopycnal flow greatly facilitated the transport of both inorganic and organic C to the open ocean, which contributed to substantial uptake of atmospheric CO 2 by the northern ECS shelf region.

Future Projections of C Uptake by the ECS
How the C uptake status of the northern ECS has evolved and how it will change are noteworthy. Because the northern ECS has increasingly received anthropogenic nutrients over the past 40 years via the CDW and, to a lesser extent, atmospheric deposition (Kim et al., 2020;Zheng and Zhai, 2021), the region has grown fertile in phytoplankton and thus, removed surface C in the form of organic matter. Much of the organic matter has likely been buried in the shallow marine sediments there. Increasing organic C production and the corresponding burial of organic C in shallow sediments have probably strengthened the northern ECS as a C sink over the past 40 years, particularly in summer (Chou et al., 2013). Moreover, the nutrient-reinforced C-rich bottom water may have affected the interannual variations in the winter C uptake status (Chou et al., 2011). However, the C sink of the northern ECS will not increase indefinitely in proportion to the increasing input of nutrients from the Changjiang River but is likely to increase until nitrate (N) acts as a limiting nutrient for organic C production (i.e., until N increases to the levels estimated on the basis of phosphate (P) concentrations multiplied by 16) (Moon et al., 2021). Further increase in seawater N concentration exceeding an N:P ratio of 16 will no longer enhance the biological production in the northern ECS. Instead, excess N (= [N] -16 × [P]) spreads into adjacent regions that are still N deficient and progressively makes them fertile. Eventually, it could expand the C sink area in the ECS. The increasing input of nutrients will make the entire ECS a stronger C sink. Conversely, a scenario of the C sink weakening by the northern ECS is equally feasible because of a reduction of nutrient supply to this area by controlling the Changjiang river discharge via the operation of the Three Gorges Dam .
Another factor that has yet to be fully explored is the effect of typhoons on the net air-sea CO 2 flux in the entire ECS. Yearround data collected from a mooring at the central ECS located outside the influence of the CDW showed that a large CO 2 efflux induced by typhoons dictated the overall C status of the central and southern ECS by switching the basin from a weak C sink to a weak C source (Wu et al., 2021). Unlike the typhoons in the central and southern ECS, those in the northern ECS exerted contrasting effects on the net air-sea CO 2 flux, depending on the levels of seawater pCO 2 at the time of typhoon passage ( Figure 5). When the typhoon passage occurs during the warming period, low pCO 2 levels (resulting from phytoplankton growth) and a high rate of CO 2 exchange (because of high windspeeds) would greatly enhance CO 2 uptake. By contrast, during the cooling period, higher oceanic pCO 2 levels (because of the addition of CO 2 -rich bottom water to the surface by the increased vertical mixing) would result in considerable CO 2 effluxes. Over the two years (2019-2020), two competing effects of typhoon passage on the net air-sea CO 2 fluxes approximately canceled out each other because typhoons meeting the two different cases were equal in number during the observational period. However, the entire ECS will likely be struck by typhoons more frequently in the future as the sea surface temperature in the region increases at an unprecedented rate. With the increasing frequency of typhoon passage, the future status of the ECS as a C sink or a C source will depend on the timing of typhoon passage.

CONCLUSION
Using discrete surface C measurements between Ieodo andJeju Island (2015-2021) and in situ continuous measurements at Ieodo (seasonal coverage for 2017-2019 and the year-round coverage for 2020-2021), we concluded that the northern ECS was a substantial sink of atmospheric CO 2 during most seasons (December to August). Such large C uptake by the northern ECS shifted the entire ECS toward a moderate C sink on an annual basis. The competing effects (reinforcing the C-sink vs. inducing the C-source) of episodic typhoon passage depending on the bottom water C accumulation by the timing of typhoon passages transiently deviated the CO 2 system. Nonetheless, the northern ECS is likely to remain as a large C sink unless the current input of anthropogenic nutrients to the northern ECS shows a sudden change. The present study provides observational and mechanistic lines of evidence for confirming the continental shelf C pump (Tsunogai et al., 1999).

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: Time series data on temperature, salinity, and seawater pH are available at Global Ocean Acidification Observing Network (http://portal.goa-on. org/Explorer?action=oiw:mobile_platform: STS_667: observations:). Data on daily mean air pCO 2 (https://data.kma. go.kr/data/gaw/selectGHGsRltmList.do?pgmNo=587) and hourly mean windspeeds (https://data.kma.go.kr/data/sea/ selectBuoyRltmList.do?pgmNo=52&tabNo=1) are available online at the National Climate Data Center supported by the Korea Meteorological Administration. The seasonal sea surface temperature data for 2020 were derived from the MODIS-AQUA satellite remote sensing maintained by the National Aeronautics and Space Administration (https://oceandata.sci.gsfc.nasa.gov).

AUTHOR CONTRIBUTIONS
KL formulated the research question. KL, J-MK and G-SL analyzed the data, and J-MK and KL wrote the paper together. G-SL, EL, J-YJ, JL and I-SH contributed to C measurements for this paper. All authors contributed to the article and approved the submitted version.