Pacific Ocean–originated anthropogenic carbon and its long-term variations in the South China Sea

Abstract

Coastal oceans, traditionally seen as a conduit for transporting atmospheric carbon dioxide (CO₂)–derived anthropogenic carbon (C_ANT) to open oceans, exhibit complex carbon exchanges at their interface. South China Sea (SCS) exemplifies this complexity, where interactions with the Pacific, particularly through Kuroshio intrusion, challenge the understanding of C_ANT source and variability in a coastal ocean. Contrary to prevailing paradigm expectations, our high-resolution, long-term data reveal that C_ANT in the SCS primarily originates from Pacific water injection across the Luzon Strait rather than atmospheric CO₂ invasion. Over the past two decades, the SCS has experienced increasing C_ANT levels, with notable interannual fluctuations driven by El Niño and La Niña events influencing Kuroshio intrusion, generating anomalously high and low C_ANT inventories, respectively. This highlights an overlooked C_ANT transport pathway from open to coastal oceans, responsible for cumulative ocean acidification that has already affected coral reefs enriched in the SCS located west of the Coral Triangle.

Anthropogenic carbon is transported from open to coastal oceans through boundary exchanges associated with the ENSO cycle.

INTRODUCTION

The global ocean annually absorbs ~25% of the anthropogenic carbon (C_ANT) emitted from both fossil fuel burning and land use changes (1), representing a sustained atmospheric CO₂ sink and an important climate mitigation pathway. Penetration of C_ANT in the ocean, however, causes ocean acidification, which directly influences marine calcifiers by interfering with the formation of their calcium carbonate skeleton (2, 3). Such acidification has been an important driver of coral reef degradation both globally and regionally (4, 5).

Coastal oceans, defined as ocean margins or continental margins (6), are relatively small in surface area but store disproportionally more C_ANT than the open ocean. Some coastal systems are currently undergoing acidification faster than the open ocean (7, 8), thus posing a substantial threat to their ecosystem services. The relatively strong atmospheric CO₂ invasion in coastal oceans is mainly ascribed to elevated biological production under eutrophic conditions and enhanced buffer capacity due to benthic inputs of alkalinity (9). Moreover, a prevailing paradigm is that coastal oceans are a “pipeline” transporting C_ANT to the open ocean (9–18).

However, the exchanges between the coastal and open ocean are often three-dimensional and interactive under the influence of climate variability and change (6, 19–21). Such complexity is also temporally and regionally variable, creating a daunting challenge in constraining the C_ANT source, storage, and evolution in a coastal ocean. Emerging evidence suggests that coastal oceans could incorporate C_ANT originating in the open ocean, thus helping to maintain and/or enhance the latter’s buffer capacity. On the basis of a mass balance analysis, the Mediterranean Sea receives net C_ANT from the North Atlantic across the Strait of Gibraltar (22), the process of which and its impact has not been fully explored.

The South China Sea (SCS), the world’s largest tropical-subtropical ocean margin, is an overall source of atmospheric CO₂. The degassing flux was initially estimated to be 33.6 ± 51.3 Tg C year⁻¹ (23, 24) and has been recently updated to be 13.3 ± 18.5 Tg C year⁻¹ (25). The SCS also features strong upwelling (26, 27) and, therefore, was previously believed to play a minor role in absorbing and storing C_ANT compared to other coastal oceans and the open ocean (28, 29). On the other hand, the SCS actively exchanges with the western North Pacific (wNP) through the 2200-m-deep Luzon Strait, whose circulation displays a “sandwich-like” structure comprising a net inflow from the wNP both above ~600 m, including Kuroshio intrusion (30), and below ~1500 m, and a net outflow from the SCS between 600 and 1500 m (fig. S1) (27, 31). Given the complex vertical circulation structure, how C_ANT exchanges between the SCS and the wNP with water depth is unknown.

Here, contrary to the prevailing paradigm, we show that instead of absorption from the atmosphere, C_ANT in the SCS originates in the Pacific Ocean. This overlooked pathway of C_ANT transport from open to coastal oceans is attested by the long-term variability and change of SCS C_ANT that are associated with the El Niño–Southern Oscillation (ENSO).

RESULTS

Distinct spatial gradients of C_ANT corresponding to circulation fields

Available C_ANT data in the SCS are sparse and limited to only a few cruise investigations (28, 29, 32). In the present study, we use a large field dataset to estimate SCS C_ANT concentrations (see the “Estimation of C_ANT concentrations” section in Materials and Methods; fig. S2) spanning the 1997–2018 period (Fig. 1 and table S1). We examine C_ANT patterns in space and C_ANT evolution over time in the SCS, which document the control of boundary exchanges associated with the ENSO cycle on C_ANT concentrations, fluxes, and storage in the coastal ocean.

Fig. 1. Map of the SCS showing the sampling stations.

Colored symbols indicate the number of visits to each station. Arrows schematically denote the pathway of the North Equatorial Current (NEC), Kuroshio Current (KC), and Kuroshio intrusion (KI) in the wNP. A total of 452 visits were conducted at 252 stations during 16 cruises over the 1997–2018 period (table S1).

In the SCS, C_ANT concentrations generally decrease from 60 to 70 μmol kg⁻¹ within the surface mixed layer with increasing depth and attain extremely low levels of <5 μmol kg⁻¹ at 1500 m (Fig. 2A). A latitudinal pattern is characterized by a north-to-south decrease in C_ANT above 600 m, showing concentration gradients along the stream of the cyclonic circulation in the SCS (Figs. 2B; 3, A, D, and G; and 4, A and B) (27). Longitudinal gradients of C_ANT above 600 m, though relatively small, exist in both the northern (18°N to 24°N) and southern (10°N to 17°N) SCS. In the former, high C_ANT concentrations near the Luzon Strait decrease westward along the pathway of intruding Kuroshio waters (Figs. 2C; 3, B, E, and H; and 4, A and B) (27). This is consistent with the finding that below the surface layer, C_ANT contours generally shoal from the wNP to the SCS across the Luzon Strait (33). In the southern SCS, upwelling off the Vietnam coast (34) brings C_ANT-poor deep water upward and induces lower concentrations in the west than in the east (Figs. 2D; 3, F and I; and 4, A and B).

Fig. 2. C_ANT in the SCS generally decreases along the circulation pathway.

(A) Overall pattern of C_ANT and circulation above 1500 m in the SCS. Black solid, dashed, and dotted arrows indicate the climatological mean circulation fields over 0 to 100 m, 100 to 600 m, and 600 to 1500 m, respectively (27). Red solid arrows schematically indicate the pathway of the NEC, KC, and KI in the wNP. (B) Latitudinal distributions of C_ANT concentrations averaged across longitudes of 109°E to 120°E. (C and D) Longitudinal distributions of C_ANT concentrations averaged across latitudes of 18°N to 24°N and 10°N to 17°N, respectively. In (B) to (D), white dashed lines indicate the C_ANT contour (50 μmol kg⁻¹), while black dashed arrows denote climatological mean circulation fields (27). Both latitudinal and longitudinal gradients of C_ANT correspond well to the circulation fields; in particular, C_ANT concentrations generally decrease westward along the pathway of intruding Kuroshio waters from the Luzon Strait.

Fig. 3. Evolution of time-averaged C_ANT concentrations in the SCS over the years.

(A, D, and G) along the vertical section, (B, E, and H) along the northern horizontal section, and (C, F, and I) along the southern horizontal section (Fig. 2) over the periods 1997–2005, 2006–2012, and 2013–2018. In (A) to (I), white dashed lines indicate the C_ANT contour (50 μmol kg⁻¹), while black dashed arrows denote climatological mean circulation fields (27). (J) Vertical distribution of time-averaged C_ANT concentrations above 1500 m of the entire SCS. Error bars indicate uncertainties of the C_ANT estimation method, and shading represents one SD of the mean. C_ANT concentrations gradually increase over the three stages in the SCS.

Fig. 4. C_ANT fluxes in three depth layers of the SCS.

(A) 0 to 100 m. (B) 100 to 600 m. (C) 600 to 1500 m. Red/blue arrows and numbers indicate the influx/outflux of C_ANT across the Luzon (LZ), Taiwan (TW), Mindoro and Balabac (MB), and Karimata (KAR) straits, as well as riverine inputs and those via sea-air exchange in (A). In (B), black arrows and numbers indicate C_ANT upward and downward fluxes mainly driven by upwelling and downing from depths of 100 and 600 m, respectively. The green circled number indicates the net C_ANT influx in each layer. Also shown is the distribution of C_ANT concentrations at the base of each layer. Black dashed arrows schematically denote climatological mean circulation fields (27). C_ANT in the SCS originates from Pacific water injection, including the KI above 600 m rather than atmospheric CO₂ invasion, indicating that a coastal ocean acts as a buffer container of the open ocean C_ANT.

Fast C_ANT storage resulting from Pacific water injection

According to the gradient in depth-dependent C_ANT concentrations, we divide the water column above 1500 m into three layers to examine C_ANT exchange fluxes of the SCS with surrounding waters, as well as in the SCS interior (see the “Estimation of C_ANT fluxes” section in Materials and Methods; fig. S3). In the upper 0- to 100-m layer, a total C_ANT influx of ~181 kmol s⁻¹ mainly arises from Kuroshio intrusion through the Luzon Strait and upwelling of deeper waters, while riverine inputs contribute only a small fraction. Meanwhile, a total C_ANT outflux of ~167 kmol s⁻¹ is primarily transported out of the SCS through the Taiwan, Mindoro and Balabac, and Karimata straits. The higher influx than outflux (by ~14 kmol s⁻¹) represents a C_ANT storage rate of ~6 Tg C year⁻¹ (Fig. 4A).

In the intermediate 100- to 600-m layer, inflow from the wNP including Kuroshio intrusion is the only C_ANT influx source, while outfluxes include inputs to both the upper and lower layers, mainly driven by upwelling and downwelling, and outflow through the Mindoro and Balabac Strait. A net influx of ~41 kmol s⁻¹ equals a C_ANT storage rate of ~16 Tg C year⁻¹ (Fig. 4B). In the lower 600- to 1500-m layer, a C_ANT influx of ~18 kmol s⁻¹ is generated solely from downwelling of overlying waters, while a single outflux of ~15 kmol s⁻¹ occurs via outflow of SCS waters through the Luzon Strait. Their difference yields a C_ANT storage rate of ~1 Tg C year⁻¹ (Fig. 4C).

Together, C_ANT in the SCS originates from Pacific water injection across the Luzon Strait rather than atmospheric CO₂ invasion. Instead, a sea-air C_ANT flux of very low value, though with considerable uncertainty, was estimated using a first-order mass balance (see the “Estimation of C_ANT fluxes” section in Materials and Methods), which is minor compared to the horizontal C_ANT influx/outflux through various straits in the upper layer (Fig. 4A). The total C_ANT storage rate in the SCS (23 Tg C year⁻¹) is comparable to that in the Arctic Ocean (25 Tg C year⁻¹) (17), whereas the latter is primarily governed by deep convection. Moreover, the C_ANT storage rate per unit volume above 1500 m in the SCS (8 × 10⁻¹⁵ Tg C year⁻¹ m⁻³) is nearly three times that of the entire Pacific (3 × 10⁻¹⁵ Tg C year⁻¹ m⁻³) (35). We thus suggest that through boundary exchanges, a coastal ocean acts as an efficient “buffer container” of C_ANT absorbed by the open ocean.

Interannual fluctuations of C_ANT superimposed on an overall increasing trend

In addition to the presence of characteristic spatial gradients, long-term variation of C_ANT is also distinct in the SCS. C_ANT concentrations generally increase over time, in particular above 100 m, where high values of ~72 μmol kg⁻¹ were observed in 2016–2018 (Fig. 5A). In contrast, long-term variations of dissolved inorganic carbon (DIC), total alkalinity (TA), oxygen (O₂), and temperature are obscure above 1500 m (figs. S4 to S7). Depth-averaged C_ANT concentrations in both the upper and intermediate layers also show an overall increasing trend from 1997 to 2018 (Fig. 5B). This increase is well observed along each specific section shown in Fig. 2, while time-averaged C_ANT concentrations of the entire SCS are notably higher in the latest stage, 2013–2018, than in the previous two stages (Fig. 3).

Fig. 5. Trend of increasing C_ANT modulated by dominant interannual fluctuations in the South China Sea.

(A) Above 1500 m. White dashed lines indicate the C_ANT contour (50 μmol kg⁻¹). (B) Depth-averaged concentrations at 0 to 100 m and 100 to 600 m, respectively. Error bars indicate one SD of the mean. Solid lines indicate a general trend of increasing C_ANT during 1997–2018. (C) Interannual variation of the Niño 3.4 index calculated as the mean of October to December in each year. Red circles indicate years when C_ANT data are available, while green ones indicate no C_ANT data. While C_ANT concentrations show an overall increasing trend over the past two decades, interannual fluctuations do occur and generally correspond to the temporal variation of the Niño 3.4 index, highlighting the control by circulation on C_ANT in the SCS.

On the basis of the evolution of depth-integrated C_ANT inventory above 1500 m over the time series, the C_ANT storage rate can be calculated as 1.0 ± 0.2 mol m⁻² year⁻¹ (Fig. 6). Note that we excluded the apparently low value in 2007 mainly collected from the western SCS off the Vietnam coast, where prominent upwelling occurred (34) bringing C_ANT-poor deep water upward (Figs. 2D; 3, F and I; and 4, A and B). This C_ANT storage rate is essentially consistent with that of 1.0 ± 0.1 mol m⁻² year⁻¹ previously obtained in the SCS using tracer-derived C_ANT data (28, 36), and within errors comparable to that of 0.7 to 1.0 mol m⁻² year⁻¹ in the wNP with close latitudes (37–40). Moreover, the SCS C_ANT storage rate is slightly higher than recent estimations for the entire Pacific (0.6 ± 0.2 mol m⁻² year⁻¹) and the global ocean as a whole (0.6 ± 0.1 mol m⁻² year⁻¹) (39, 40). This is reasonable given that both the C_ANT inventory and storage in the wNP, providing the majority of C_ANT in the SCS, represent an intermediate to high level from a Pacific or global perspective.

Fig. 6. Interannual variation of the depth-integrated inventory of C_ANT above 1500 m of the South China Sea.

The solid line and equation indicate the results of the linear regression analysis excluding the 2007 data point representing an apparently low C_ANT inventory indicated by a blue circle. R², P, and n denote the coefficient of determination, probability value, and sample size, respectively. The slope value indicates a C_ANT storage rate of 1.0 ± 0.2 mol m⁻² year⁻¹ over the past two decades.

In comparison to estimates obtained in some high-latitude coastal oceans, the C_ANT storage rate above 1500 m in the SCS is clearly lower than in the Labrador Sea (1.5 ± 0.2 mol m⁻² year⁻¹) (13), within errors comparable to in the Arctic Ocean (0.9 ± 0.1 mol m⁻² year⁻¹) (17), and slightly higher than in the Japan/East Sea (0.5 ± 0.1 mol m⁻² year⁻¹) (41). Extracting data from a global ocean estimation also yields higher C_ANT storage in the SCS than in the Japan/East Sea (39). Despite rapid ventilation, the absence of overturning circulation, i.e., no outflow of intermediate and deep waters to the adjacent open ocean, may have resulted in a relatively slow annual accumulation of C_ANT in the enclosed Japan/East Sea (9).

However, interannual fluctuations of C_ANT clearly occur and generally correspond to the temporal variation of the Niño 3.4 index (Fig. 5, B and C). Such consistency indicates a control of circulation beyond atmospheric CO₂ forcing. In addition, the consistency is greater at 0 to 100 m than at 100 to 600 m, likely reflecting the larger influence of the ENSO cycle on the upper water column. A nonsteady evolution of C_ANT was also observed in the Labrador and Nordic seas and has been linked to variability in the ventilation state (13, 14). In the present study, we propose that during El Niño/La Niña with a high/low Niño 3.4 index, strengthened/weakened Kuroshio intrusion (42, 43) brings more/less C_ANT from the wNP to the SCS, and the shift between the two events results in the long-term variation pattern observed in the SCS.

DISCUSSION

Kuroshio intrusion control on C_ANT affected by ENSO

To test our hypothesis, we take a closer look at the relationship between the SCS C_ANT inventory and the C_ANT influx from the Pacific through the Luzon Strait above 100 m. A significant positive linear relationship based on data for individual years (R² = 0.55, P = 0.001, n = 15; Fig. 7A) suggests that increased C_ANT influx from the wNP is associated with enhanced C_ANT inventory in the SCS, thus demonstrating the control of Kuroshio intrusion on C_ANT distributions in the upper water column of the SCS.

Fig. 7. Control by KI on the distribution of C_ANT in the SCS.

(A) Relationship of the SCS C_ANT inventory to the Pacific C_ANT influx through the Luzon Strait above 100 m. The solid line and equation indicate the results of the linear regression analysis; R², P, and n denote the coefficient of determination, probability value, and sample size, respectively. (B and C) C_ANT inventory anomaly (defined as the departure from the mean) above 100 m during El Niño (Niño 3.4 index > +0.5) and La Niña (Niño 3.4 index < −0.5). Arrows schematically indicate the pathway of the NEC, KC, and KI. Changes in the magnitude of KI associated with El Niño and La Niña events dominate the interannual fluctuations of C_ANT in the SCS, showing an overall positive and negative C_ANT inventory anomaly, respectively.

This effect of boundary exchanges is further depicted during El Niño and La Niña (Fig. 7, B and C). The onset of El Niño events strengthens the North Equatorial Current and drives its northward movement, resulting in a weakened Kuroshio Current mainstream favoring its intrusion into the SCS (42, 44). Consequently, the increased C_ANT influx from the wNP contributes to a higher-than-mean C_ANT inventory above 100 m in the SCS (Fig. 7B). In contrast, during La Niña, the stronger Kuroshio Current leads to weaker Kuroshio intrusion, which corresponds to a lower-than-mean C_ANT inventory above 100 m in the SCS (Fig. 7C). It is noteworthy that the C_ANT inventory anomaly (i.e., departure from the mean) off Hainan Island (Fig. 1) is slightly negative during El Niño and positive during La Niña. This opposite result is probably related to the Qiongdong upwelling, which is stronger during El Niño bringing more C_ANT-poor deep water upward (45, 46).

Ocean acidification affected by C_ANT

Since C_ANT storage in seawater directly induces ocean acidification, we evaluate this effect by examining the variation of both pH and aragonite saturation state (Ω_arag) in the SCS. Both depth-averaged pH and Ω_arag values above 100 m generally decrease over 1997–2018 (Fig. 8, A and B), opposite to the long-term trend of C_ANT (Fig. 5B). The pH evolution over the time series points to an annual acidification rate of 0.0019 ± 0.0006 in the SCS (Fig. 8A), which is comparable to that observed at stations ALOHA and BATS in the Pacific and Atlantic, respectively (47, 48). Moreover, in the SCS, pH and C_ANT above 100 m show a significant negative linear relationship (R² = 0.53, P = 0.002, n = 15; Fig. 8C), indicating that pH decreases by 0.0012 ± 0.0005 per addition of 1 μmol kg⁻¹ of C_ANT. Similarly, a significant negative linear relationship between Ω_arag and C_ANT (R² = 0.52, P = 0.002, n = 15; Fig. 8D) suggests that Ω_arag decreases by 0.0069 ± 0.0031 per addition of 1 μmol kg⁻¹ of C_ANT. These relationships point to ongoing ocean acidification resulting from increasing C_ANT in the SCS.

Fig. 8. Effect of ocean acidification driven by the accumulation of C_ANT in the SCS.

(A and B) Interannual variation of depth-averaged pH and aragonite saturation state (Ω_arag) above 100 m during 1997–2018. (C and D) Relationship of depth-averaged pH to C_ANT and Ω_arag, respectively, above 100 m. Error bars indicate one SD of the mean. The solid line and equation indicate the results of the linear regression analysis; R², P, and n denote the coefficient of determination, probability value, and sample size, respectively. Both pH and Ω_arag show a generally decreasing trend over the past two decades, suggesting ongoing ocean acidification resulting from the increase in C_ANT in the SCS.

The SCS hosts 571 known reef coral species, a number comparable to that found inside the Coral Triangle (49), and the former’s coral reef area represents 5% of that in the global ocean (50), providing valuable ecosystem services. Since the 1960s, however, both the corals’ skeletal density and calcification rate in the northern SCS have shown a gradual decline corresponding to continuous pH reduction (51). Ocean acidification, acting synergistically with warming, is harming coral reefs in the SCS. This effect would be exacerbated as the SCS receives net C_ANT from the Pacific, thus continuously lowering pH and Ω_arag, particularly if the C_ANT influx from the Pacific through the Luzon Strait increases in the future. This is highly likely because (i) C_ANT concentrations in the wNP are expected to increase for a long period of time mainly driven by atmospheric CO₂ forcing (52), (ii) most Coupled Model Intercomparison Project Phase 6 (CMIP6) models (53) simulate stronger Kuroshio intrusion under either SSP1-2.6 (carbon neutrality) or SSP5-8.5 (high emission) scenarios (Fig. 9), and (iii) ENSO variability has increased and is projected to increase even further (54–56), favoring extraordinary Kuroshio intrusion and acidification in the SCS. Although these predictions need further examination, the threat of C_ANT-induced ocean acidification is real.

Fig. 9. Magnitude of the KI over the 2020–2100 period under SSP1-2.6 and SSP5-8.5 scenarios simulated by various CMIP6 models.

The KI speed anomaly for a specific year is calculated as the departure from the climatological mean value over 2000–2020, which is estimated using Archiving, Validation and Interpretation of Satellite Oceanographic (AVISO) data (www.aviso.altimetry.fr/en/data.html). Deduced anomaly values for all years are averaged and plotted, with positive/negative values indicating stronger/weakened KI. Nearly all models project an overall increase in KI from the wNP to the SCS in the future.

Summary and implications

Our finding that C_ANT in the SCS originates from Pacific water injection, rather than atmospheric CO₂ invasion, is underpinned by the high resolution of our observational data in both space and time. The process responsible for the trend of increasing C_ANT concentrations over the past two decades is similar to that responsible for its interannual fluctuations, as both are linked to the Kuroshio intrusion in the upper water column. During El Niño, the Kuroshio intrusion intensifies, and C_ANT is high in the SCS. This unique process creates an overlooked pathway of C_ANT transport from open to coastal oceans. Therefore, coastal oceans without deep convection can also be an important buffer container of C_ANT via boundary exchanges with the open ocean.

A direct consequence of the above is a faster rate of C_ANT storage in the SCS than in the Pacific, with potentially devastating ecosystem impacts. There is already evidence showing that C_ANT-induced ocean acidification has been impairing SCS corals (51). Future extreme acidification in the SCS is highly likely, as the projected enhancement of ENSO variability (54–56), strengthening Kuroshio intrusion, and increasing C_ANT in the wNP (52) all conspire to the increased occurrence of high C_ANT in the SCS. Such extreme acidification, combined with elevating ocean temperatures due to future global warming, could result in a loss of reef resilience below levels from which the prospect of recovery is diminished (4, 57). Our finding also highlights the need to improve spatiotemporal coverage of observations in ocean-dominated margins similar to the SCS (6, 21, 58), as well as the need to use high-resolution modeling to capture the intrusion process of open ocean waters and simulate the risk of extreme acidification in coastal oceans.

MATERIALS AND METHODS

Estimation of C_ANT concentrations

C_ANT in the SCS during the fall of 2016 was calculated using the field chlorofluorocarbon data (36) based on the TTD (transit time distribution) method (59). Subsequently, O₂, DIC, and TA data collected during the same cruise were used to obtain optimized values of the coefficients in Eq. 1 (i.e., a = 1.15, b = 7.67, c = 0.008, d = 2.35 × 10⁶; R² = 0.98, P < 0.001, n = 152), which illustrates the TrOCA (tracer combining O₂, DIC, and TA) method (60). The criterion is that the C_ANT concentration values obtained by Eq. 1 have minimum deviations from the TTD-derived results (fig. S2). This optimized equation was further applied to field data collected during 16 cruises conducted in the SCS from 1997 to 2018, producing a comprehensive dataset of C_ANT with high spatiotemporal coverage in a large coastal ocean. Note that we used C_ANT concentrations at the base of the surface mixed layer to represent those within the waters above. A further comparison with the eMLR (extended multiple linear regression) method (61) highlights the applicability of our optimized TrOCA method in calculating C_ANT in a coastal ocean. Details of the C_ANT concentration estimations including error evaluations are provided in the Supplementary Materials.

$${\mathrm{C}}_{\text{ANT}}=\left[{\mathit{O}}_2+\mathit{a}\left(\text{DIC}-0.5\text{TA}\right)-\text{exp}\left(\mathit{b}-\mathit{c}\mathrm{\theta }-\mathit{d}/\mathrm{T}{\mathrm{A}}^2\right)\right]/\mathit{a}$$ (1)

Estimation of C_ANT fluxes

The SCS exchanges with the Pacific and surrounding seas through the Luzon, Taiwan, Mindoro and Balabac, and Karimata straits. The C_ANT flux (F) across these straits was calculated by

$$\mathit{F}=\overline{{\displaystyle \underset{0}{\overset{\mathrm{s}}{\int }}}\left(\mathit{u}\times \mathit{c}\right)\mathit{ds}}$$ (2)

where u is the current velocity, c is the C_ANT concentration, and s is the integrating length or area. The overbar denotes time-averaged fluxes. Information on u, c, and s for a specific strait is provided in the Supplementary Materials. Equation 2 was also applied to calculate the vertical C_ANT flux in the SCS interior, in which case u is the total water transport velocity mainly driven by upwelling or downwelling and s is the integrated depth (fig. S3).

The riverine C_ANT input flux was estimated by multiplying the nearshore C_ANT concentration with salinities <30 (on average, 55 μmol kg⁻¹ in the present study) by the total freshwater discharge entering the SCS (3.3 × 10³ km³ year⁻¹) (62). The C_ANT flux via sea-air exchange was estimated as the difference between the total and natural CO₂ flux, with the former based on field observations in the SCS (25) and the latter obtained using a quasiconservative tracer (63, 64). Details of this C_ANT flux estimation are provided in the Supplementary Materials.

Supplementary Materials

This PDF file includes:

Supplementary Methods

Figs. S1 to S7

Table S1

References