Impact of Marine Heatwaves on Air‐Sea CO2 Flux Along the US East Coast

Marine heatwaves (MHWs) are extremely warm ocean temperature events that significantly affect marine environments, but their effects on the coastal carbonate system are still uncertain. In this study, we systematically quantify MHWs' impacts on air‐sea carbon dioxide (CO2) flux anomalies (FCO2′) in the Mid‐Atlantic Bight (MAB) and South Atlantic Bight (SAB) from 1992 to 2020. During the longest MHW in both regions, oceanic CO2 uptake capabilities substantially decreased, primarily due to significant increases in the seawater partial pressure of CO2 (pCO2sea). For all cases, MHWs played a more significant role in driving pCO2sea changes in the MAB than the SAB, where non‐thermal drivers dominated pCO2sea variability. In the MAB, weakened wind speeds related to wintertime atmospheric perturbations increase ocean temperatures and pCO2sea, further reducing CO2 uptake during winter MHWs. This work is the first to connect extreme temperatures to coastal air‐sea CO2 fluxes. The reduction in CO2 absorption noted during MHWs in this study has important implications for coastal regions to act as continued sinks for excess CO2 emissions in the atmosphere.

• Marine heatwaves (MHWs) primarily generated positive sea surface pCO 2 (pCO 2sea ) anomalies in the Mid-Atlantic Bight (MAB) and South Atlantic Bight (SAB) but had a larger impact on air-sea CO 2 flux anomalies in the MAB • Reduced wind speeds amplified MHW contributions during CO 2 sink months and counteracted them during CO 2 source months • In the MAB, wintertime atmospheric perturbations related to zonal shifts in the jet stream produce slower wind speeds which aid in generating air-sea heat flux type MHW events that ultimately reduce oceanic CO 2 uptake Supporting Information: Supporting Information may be found in the online version of this article. 10.1029/2023GL105363 2 of 12 of the United States (Chen et al., 2014(Chen et al., , 2015;;Gawarkiewicz et al., 2019).However, their influence on carbonate system parameters like seawater partial pressure of CO 2 (pCO 2sea ) and, therefore, air-sea CO 2 flux in this region remains insufficiently investigated.
Air-sea CO 2 flux (FCO 2 ) quantifies the exchange of CO 2 between the atmosphere and ocean.The difference between seawater (pCO 2sea ) and the atmospheric partial pressure of CO 2 (pCO 2air ), ΔpCO 2 , determines whether the ocean is taking up CO 2 from the atmosphere (negative ΔpCO 2 ) or emitting CO 2 to the atmosphere (positive ΔpCO 2 ).Spatial and temporal variations in pCO 2air are small, so ΔpCO 2 variability is primarily controlled by variations in pCO 2sea (Sarmiento & Gruber, 2006).Spatiotemporal pCO 2sea changes are driven by thermodynamics, mixing, biological activities, and air-sea gas exchange.Since MHW events are extremely warm SST events, it is conceivable that MHWs could impact pCO 2sea and therefore FCO 2 via thermodynamics.
Despite this, literature connecting MHWs and FCO 2 is limited.One study found that prolonged MHWs reduced CO 2 uptake in major North Pacific open-ocean uptake regions and, due to linkages with the El Nino Southern Oscillation (ENSO), reduced CO 2 release in major tropical Pacific open-ocean outgassing areas (Mignot et al., 2022).However, the influence of MHWs on nearshore carbonate systems in continental shelves needs to be better understood.Despite the coastal ocean constituting only 7%-10% of the world's oceans, it is argued to play a disproportionately large role in the uptake of CO 2 by the ocean (Dai et al., 2022;Gattuso et al., 1998;Le Quéré et al., 2009, 2010;Liu et al., 2010;Najjar et al., 2012).Therefore, understanding how coastal air-sea CO 2 fluxes are potentially modulated by MHW events is not only important for furthering our comprehension of coastal biogeochemical cycles, but also for improving climate models and predictions of future climate change impacts.
This study focuses on the Mid-Atlantic Bight (MAB, Figure 1a) and South Atlantic Bight (SAB, Figure 1b), two regions on the U.S. East Coast that are well-studied and have decades of inorganic carbonate data available (e.g., Cai et al., 2020;Li et al., 2022).Though regional differences in physical and biological processes exist, both the MAB and SAB have been consistently estimated as net sinks of atmospheric CO 2 , with the MAB's CO 2 flux density estimated as −0.73 to −1.90 mol C m −2 yr −1 (Cahill et al., 2016;DeGrandpre et al., 2002;Fennel et al., 2008;Laruelle et al., 2018;Signorini et al., 2013), and the SAB's as −0.48 to −0.75 mol C m −2 yr −1 (Cahill et al., 2016;Jiang et al., 2008;Signorini et al., 2013).Cai et al. (2020) found that pCO 2sea variations along the US East Coast reflect local, short-term modifications by coastal physical and biological processes.Under extreme cases like MHW events, pCO 2sea and, therefore, FCO 2 , may be more sensitive to significant increases in SST and potentially switch coastal CO 2 sinks to sources for the atmosphere.As such, this study aims to quantify the changes in pCO 2sea and FCO 2 during MHWs in the MAB and SAB and understand the underlying dynamics.
To achieve this goal, we first examined each region's most prolonged MHW event from 1992 to 2020.We then investigated the common patterns and mechanisms underlying changes in pCO 2sea and FCO 2 during MHWs in each region to identify the conditions under which CO 2 flux will substantially change.The following section describes the data and methods used for MHW detection and FCO 2 calculations.Section 3 presents and discusses the results, and Section 4 reiterates the conclusions.

Marine Heatwave Detection
We utilized the Daily Optimum Interpolation Sea Surface Temperature v2 (OISST) data set from the National Oceanic and Atmospheric Administration (NOAA) (Reynolds et al., 2002) to detect MHWs within the period 1992/01-2020/12.OISST has a spatial resolution of 0.25° × 0.25°.The MAB and SAB regions (depth < 200 m) were analyzed separately using this daily gridded data.The MAB is composed of 193 grid cells, spanning from Cape Hatteras, North Carolina (35° north) to Cape Cod, Massachusetts (41.5°N), while the SAB consists of 163 grid cells ranging between Cape Canaveral, Florida (26°N) to Cape Hatteras, North Carolina (35°N).
Using the daily gridded SSTs, MHWs were detected in each grid cell following the definition from Hobday et al. (2016), that is, the period that SSTs are above the 90th percentile threshold of the climatology.For this study, the climatology was computed from 1992 to 2020.The CO 2 data used in this study is limited to monthly resolution, so we cannot determine the impact of individual MHW events by day.Instead, we examine the influence of MHW months.The number of days a MHW was detected in each grid was summed for each month (348 in total) within the 28-year study period.Individual months with 15 or more MHW days are considered MHW months because we expect that the impact of extreme temperatures that endure for at least half a month should be reflected in the pCO 2 and CO 2 flux data.The remaining months are defined as non-MHW months.

Air-Sea CO 2 Flux Calculation
Monthly air-sea CO 2 flux (FCO 2 ) from January 1992 to December 2020 was calculated using the gas exchange formula (Wanninkhof et al., 2009), where k is the gas transfer velocity, K 0 is CO 2 solubility in seawater, pCO 2sea and pCO 2air are the partial pressure of CO 2 in the ocean and atmosphere, respectively.A positive or negative FCO 2 value indicates the ocean is acting as a CO 2 source or sink to the atmosphere.Gas transfer velocity (k) is calculated using wind speed at 10-m (U 10 ) above the sea surface (Wanninkhof, 2014), where U 10 is monthly 10-m wind speeds from the fifth generation ECMWF high-resolution global reanalysis data set, ERA5 (Hersbach et al., 2023), which contains monthly gridded atmospheric data at 0.25° × 0.25° resolution from 1959 to present.Sc is the Schmidt number and is specific to CO 2 (Wanninkhof, 2014), In situ SST, SSS, and sea surface fCO 2 were obtained from the gridded Surface Ocean CO 2 Atlas (SOCAT) version 2022 (0.25° × 0.25°).fCO 2 was converted to sea surface pCO 2 (Takahashi et al., 2019).Atmospheric CO 2 mole fractions in dry air (xCO 2air ; in ppm) were downloaded from NOAA's zonally averaged Greenhouse Gas Marine Boundary Layer Reference (MBL Reference, Conway et al., 1994).The MBL reference is composed of weekly air samples collected at a subset of sites globally from NOAA's Cooperative Air Sampling Network and has a spatial resolution of 0.05 sine of the latitude.Monthly air samples were obtained by averaging all weekly values for each month.pCO 2air was calculated from xCO 2air using the conversion equation (Sarmiento & Gruber, 2006): where p w is the water vapor pressure, which is assumed to be at saturation in the vicinity of the air-sea interface (Weiss & Price, 1980): (5) ) is a function of absolute SST and SSS (Weiss, 1974): )2 ] (6) where A 1 = −58.0931,A 2 = 90.5069,A 3 = 22.2940, B 1 = 0.027766, B 2 = −0.025888,and B 3 = 0.0050578 are CO 2 -specific coefficients.The SOCAT database is comprised of in situ observations, which means data gaps are present.Months without data were excluded from analysis.

Thermal and Non-Thermal Components of pCO 2sea Anomalies
Since pCO 2sea is the dominant factor influencing CO 2 flux, changes (i.e., anomalies) in this parameter would likely induce CO 2 flux anomalies.To isolate the influence of SSTs (i.e., MHWs) on pCO 2sea anomalies (denoted using a prime, pCO 2sea ′), we calculate the thermal ( where  CO2 and  SST are the monthly pCO 2sea and SST climatology, respectively.The temperature sensitivity of pCO 2sea is set as 0.0423°C −1 (Takahashi et al., 1993(Takahashi et al., , 2002)).The thermal component represents the pCO 2sea ′ driven by SST change and the non-thermal component represents pCO 2sea ′ driven by other non-temperature factors, including dissolved inorganic carbon (DIC), total alkalinity (TA), and SSS change.Processes effecting non-thermal parameters include biological activities (Cao et al., 2020;Signorini et al., 2013), physical transport and mixing (Cao et al., 2020;Jiang et al., 2008Jiang et al., , 2013)), and air-sea gas exchange (Cai et al., 2020;Xu et al., 2020).In this work, a student's t-test is used to test whether anomalies are significantly different from zero at a 95% confidence interval (i.e., statistically signifigant at p-value of 0.05 or 5%).

Taylor Expansion of Air-Sea CO 2 Flux
SST is positively correlated with gas transfer velocity (k) and negatively correlated with CO 2 solubility (K 0 ).As a result, the variability of the gas transfer coefficient (Γ), the product of k and K 0 , is almost independent of SST and is instead predominately controlled by wind speed (Wanninkhof & Triñanes, 2017).To understand which factor (ΔpCO 2 or wind) is driving FCO 2 anomalies during marine heatwave and non-marine heatwave months, a Reynolds decomposition of the CO 2 flux anomaly (FCO 2 ′) is computed.This expansion reveals the relative contributions of the two flux components (i.e., ΔpCO 2 and Γ) in the MAB and SAB.Each flux component (example: ΔpCO 2 ) is considered the sum of their long-term monthly mean (i.e., climatology;  ∆CO2 ) and anomaly (ΔpCO 2 ′).Thus, the decomposition of FCO 2 begins as: Rearrangement of Equation 9into their zero, first, second, and higher order terms produces: where the first term on the right-hand side is the zero-order term (i.e., the FCO 2 climatology), the next two terms are the first-order terms, and the last term is the second-order term.Since CO 2 flux anomaly (FCO 2 ′) is the focus of this analysis, Equation 10 can be consolidated into: where the terms on the right-hand side of Equation 11 represent the contribution to FCO 2 ′ from oceanic and atmospheric pCO 2 anomalies (housed within the ΔpCO 2 term) and wind speed anomalies (housed within the gas transfer coefficient, Γ), and (O 2 ) is the higher-order residual term that can be neglected.The component with the largest contribution is the dominant term driving the FCO 2 anomalies during each month in the MAB and SAB. 10.1029/2023GL105363 5 of 12

Case Studies: Two Most Prolonged MHW Events in the MAB and SAB
To investigate the response of sea surface pCO 2sea and CO 2 flux to MHWs, we begin by examining the two most prolonged MHWs in the MAB and SAB, respectively.These two events provide a valuable opportunity to examine changes in pCO 2 and FCO 2 in response to the longest exposure period of the sea surface to extreme temperatures.Insights from these two case study events may be generally representative of the impact of all prolonged MHWs on pCO 2 and FCO 2 .The longest event in the MAB lasted over 200 days (25 November 2011-15 June 2012), named the "2012 MAB event."Analysis focused on the overlapping period between the MHW event and available SOCAT observations from February 2012 to June 2012 (5 months, Figure 1a).The longest MHW event in the SAB, referred to as the "2017 SAB event," lasted for 44 days (21 March 2017-05 May 2017).To maintain the definition of a MHW month, analysis focused on April 2017 (Figure 1b), as the event lasted fewer than 15 days in March and May, preventing these months from being considered MHW months in this analysis.These two MHW events greatly exceed the average MHW duration (in days) in both regions: 16.2 ± 9.3 days in the MAB and 13.9 ± 7.8 in the SAB (Figure S1 in Supporting Information S1).
During the 2012 MAB event, SST anomalies exhibited a significant increase in the initial 3 months (February-April), followed by a gradual decline in May and June (Figure 2a).Therefore, we divided the 2012 MAB event into the early period from February to April 2012 and the later period from May to June 2012.During the early period, pCO 2sea anomalies experienced no significant difference, but a substantial surge of approximately 46 μatm above the climatology in May and remained notably positive until June (+36.9± 12.7 μatm) (Figure 2c).This increased pCO 2 impedes air-sea CO 2 exchange since the air-sea pCO 2 gradient becomes smaller (Equation 1).Additionally, the wind speed was −0.42 ± 0.34 m s −1 lower than the climatology throughout the entire event (Figure 2e), which also impedes the CO 2 uptake.Consequently, the average CO 2 flux anomaly is positive (+0.59 ± 0.55 mol C m −2 yr −1 ), causing a 26% reduction of the MAB's average CO 2 uptake from the climatology (−2.31 ± 1.90 mol C m −2 yr −1 ) during this event (Figure 2i).This uptake reduction is equivalent to 0.82 Tg C yr −1 , which reduces the MAB's 2012 annual net carbon uptake (in mol C yr −1 ) by 13.7%.Although it lasted only 44 days, the 2017 SAB event provides a valuable snapshot of CO 2 flux changes during MHW events in this region.Both SST and pCO 2sea exhibited positive anomalies before and after the MHW (Figure 2b).In April, however, both parameters showed a notable increase: the SST and pCO 2sea were +1.47°C and +99.2 μatm higher than climatology, respectively (Figures 2b and 2d).Thermodynamically, such a temperature increase alone accounted for a one-fourth rise in pCO 2sea of +25.3 μatm (Takahashi et al., 1993), while the non-thermal component contributed an additional +69.5 μatm.This highlights the significant amplification of the influence by non-thermal drivers during the 2017 SAB event, contrasting with the counteracting influence observed in the early period of the 2012 MAB event.Concurrently, wind speed slightly decreased (−0.35 m s −1 ) in April.On average, climatological sea surface pCO 2 in the SAB tends to reach equilibrium with the atmosphere in April, with a flux of +0.10 ± 1.4 mol C m −2 yr −1 .However, during the 2017 event, the SAB transformed from CO 2 neutral into a significant CO 2 source with a flux anomaly of +2.07 mol C m −2 yr −1 (Figure 2j).This increased outgassing added an additional 2.66 Tg C yr −1 to the atmosphere, thereby reducing the SAB's 2017 annual carbon uptake (in mol C yr −1 ) by 165%.
In both the 2012 and 2017 MHW events, flux anomaly was anomalously positive.Simultaneously, pCO 2sea was anomalously positive while wind speed was anomalously low.The first-order terms from the Taylor expansion of CO 2 flux during both the 2012 MAB and 2017 SAB events (Equation 11) reveal whether ∆pCO 2 (i.e., pCO 2sea ) or wind speed (Γ in Figure S2 of the Supporting Information S1) was the primary driver of the flux anomaly during both events (Figure S2 in Supporting Information S1).A monthly breakdown of these contributions during the 2012 MAB event (Figure S2b in Supporting Information S1) shows that ΔpCO 2 was the primary contributor during 3 of the 5 months during the 2012 MAB event (i.e., February, May, and June), while wind speed dominated March and April.As such, ΔpCO 2 is generally considered the primary factor controlling FCO 2 ′ during the entire 2012 MAB event.Similarly, ΔpCO 2 was the overwhelming contributor to FCO 2 ′ during the 2017 SAB MHW event, with wind speed contributing inconsequently (Figures S2c and S2d in Supporting Information S1).
Understanding whether thermal (i.e., MHW) or non-thermal drivers of pCO 2sea were controlling pCO 2sea anomalies during each MHW event will determine whether the MHW or non-temperature factors were responsible for inducing the positive FCO 2 ′ experienced during both case study events.During the 2012 MAB event, non-thermal drivers of pCO 2sea counteracted the thermal contribution from the MHW during the early period of the event but were unable to do so in the later period (Figure S3a in Supporting Information S1).Thus, processes controlling non-thermal pCO 2sea change in the MAB are potentially important buffers against MHW events because they can counteract either all or part of a MHW's thermal influence on FCO 2 .During the 2017 SAB event, the MHW was not primarily responsible for the large magnitude of FCO 2 ′ in April 2017 (Figure S3b in Supporting Information S1).Rather, non-thermal parameters dominated pCO 2sea ′ during the event.This highlights the increased importance of non-thermal drivers in the SAB compared with the MAB since the large magnitude of the non-thermal pCO 2sea ′ suggests these drivers may be capable of markedly amplifying or reducing MHW influences on pCO 2sea ′.
With non-thermal pCO 2sea changes proving to be important in both regions, understanding the processes driving this parameter would be useful.However, without water-column total alkalinity and dissolved inorganic carbon data, a definitive answer regarding the mechanism(s) responsible for driving non-thermal pCO 2sea changes is beyond this study.However, work by Jones et al. (2014) on the CO 2 gas exchange timescale allows us to broadly hypothesize possible mechanisms.Assuming that open ocean timescales are the same in adjacent coastal regions, in the SAB region, the gas exchange timescale is longer than the mixed layer residence time.So, in this case, mixing processes are vital controls of pCO 2sea changes.In the MAB region, the opposite occurs; the gas exchange timescale is shorter than the mixed later residence time.Consequently, mixing is not a dominant control on gas exchange in this region.Instead, our results suggest that the timescale of thermally induced pCO 2sea changes resulting from MHWs in the MAB is shorter than that of air-sea gas exchange.This further highlights the importance of temperature as a dominant control in air-sea gas exchange in the MAB compared with non-thermal drivers.Non-thermal pCO 2sea changes in the MAB during the 2012 MHW event were likely biologically driven, as the bulk of the event occurred during the time period of the spring bloom (Cao et al., 2020;Signorini et al., 2013).

MHW Impacts on pCO 2sea Change
To assess whether the significant changes in pCO 2sea and CO 2 flux observed during the most prolonged events are representative of all MHW events, we statistically compared the pCO 2sea and flux anomalies between MHW and non-MHW months in the two regions.Over the 28-year study period (1992-2020), the ensemble mean flux anomalies during MHW months in the MAB and SAB (+0.13 ± 0.63 and +0.26 ± 0.82 mol C m −2 yr −1 , respectively) showed only slight differences compared to non-MHW months (−0.12 ± 0.81 and −0.14 ± 0.96 mol C m −2 yr −1 , respectively).While the ensemble mean flux anomaly during MHW months is statistically different from non-MHW months in both regions (p-values of 0.03 and 0.01 in the MAB and SAB, respectively), the large standard deviations indicate that a MHW alone is not a sufficient condition for the occurrence of positive flux anomalies.To identify conditions under which MHWs will associate with significant positive flux anomalies, we further divided MHW months into two categories: those above and below the 75th percentile FCO 2 ′ values (+0.54 mol C m −2 yr −1 in the MAB and +0.84 mol C m −2 yr −1 in the SAB).This helps to distinguish whether MHWs or non-temperature factors are responsible for producing large, positive flux anomaly values, that is, those that fall above the 75th percentile.Both the 2012 and 2017 events fell above the 75th percentile threshold.
In the MAB, temperature exerts a more significant influence on pCO 2sea changes than other drivers during all MHW events (Figure 3a).Above the 75th percentile, both thermal and nonthermal drivers are important for producing a large positive flux anomaly (Figure 3b).Thermal contributions are usually larger than non-thermal influences, but the means of the two components are not statistically different (p-value = 0.10; Figure 3b).Nevertheless, temperature remains the primary contributor to pCO 2sea anomalies during these months, accounting for approximately 68% of the total pCO 2sea changes (+17.1 ± 12.6 μatm).Like the early period of the 2012 MAB event, nonthermal drivers tend to counterbalance the impact of temperature on pCO 2sea during all MHW months (Figure 3a), especially in MHW months with flux anomalies below the 75th percentile (Figure 3c).
However, in MHW months above the 75th percentile in the MAB, nonthermal drivers amplify thermally induced positive pCO 2sea anomalies rather than offsetting the temperature's impact (Figure 3b).Furthermore, the importance of non-thermal processes as drivers of flux anomalies increases relative to the thermal component because, unlike MHW months below the 75th percentile, there is not a statistical difference between the two components during MHWs above the 75th percentile (Figure 3e).In general, MHWs produce a temperature-induced increase in pCO 2sea that is somewhat offset by non-temperature factors in the MAB.This is consistent with Mignot et al.'s (2022) result in the North Pacific subtropical gyre.In the SAB, initial comparisons of all MHWs in this region (Figure 3d) to those in the MAB (Figure 3a) indicate that the magnitudes of thermally induced pCO 2sea changes in the SAB are similar to those in the MAB, with nonthermal drivers also playing a reduced role compared to MHWs.However, statistical analysis reveals that the thermal and non-thermal components in the SAB are not statistically different (p-value = 0.17), meaning non-thermal processes exert a greater influence in the SAB during MHW months (Figure 3d), as observed in the 2017 SAB event.Like the MAB, nonthermal drivers in months below the 75th percentile offset the impact of high temperatures (Figure 3f), resulting in pCO 2sea anomalies close to zero.However, for months above the 75th percentile (Figure 3e), the main contribution to pCO 2sea anomalies comes from nonthermal drivers (+49.7 ± 43.1 μatm) rather than extremely warm temperatures.This suggests that high pCO 2sea values during MHW events in the SAB are not solely caused by temperature itself, so it is inappropriate to conclude that MHW events have significant impacts on CO 2 flux in this region, unless the nonthermal drivers also produce a positive pCO 2sea during MHW events.
While MHWs appear to play a larger role driving air-sea gas exchange in the MAB, prolonged MHW events, like the 2012 MAB and 2017 SAB MHW events, can still produce a noticeable impact on interannual scales.The impact of these two events on the interannual variability of air-sea CO 2 flux anomalies is evident in Figure S4 of the Supporting Information S1.The years in which the two most prolonged MHWs occurred correspond to some of the highest annual flux anomalies over the 28-year study period.So, while non-thermal flux drivers are vital to produce large flux anomalies, especially in the SAB, prolonged MHWs are still a necessary condition to produce large changes in air-sea CO 2 flux.

Wind Speeds Change During All MHW Months
We also examined wind speed changes during MHW months since wind speed significantly impacts the magnitude of air-sea CO 2 flux (FCO 2 ), with faster winds generally resulting in higher gas transfer velocities (k) and larger FCO 2 values.Wind speeds were below the climatological average during the two most prolonged MHW events in the SAB and MAB.These slower wind speeds amplified MHW influences on FCO 2 ′ by further decreasing the CO 2 sink in wintertime months but counteracting MHW effects in summertime months.
Reduced wind speeds were typical during many MHW months (Figure 4).Yet, like pCO 2sea , further investigation revealed that wind speeds did not significantly deviate from the climatology during all MHW events in the MAB and SAB (Figures 4a and 4d).However, among the 20 MHW months above the 75th percentile in the MAB (Figure 4b), 12 occurred during the wintertime when the MAB typically acts as a CO 2 sink.During these events, wind speeds were slower by −0.36 ± 0.50 m/s compared to the climatology, which may link to zonal shifts in the jet stream (Chen et al., 2014), significantly reducing the CO 2 sink in the MAB.
In the SAB, wind speeds were generally slower than the climatological average during all MHW events (Figures 4d, 4e, and 4f).This indicates that wind speeds tend to decrease CO 2 uptake during CO 2 -sink months but counteract the impact of MHWs on flux during CO 2 -source months.However, the large standard deviations in wind speeds in both regions suggest that the effects of wind speed vary depending on the individual event.

Underlying Dynamics Between MHWs and CO 2 Flux on the US East Coast
During the 2012 MAB event, the extremely warm SSTs resulted in a significant increase in pCO 2sea , ultimately reducing the MAB's ability to take up CO 2 .In contrast, despite the extremely high temperature during the 2017 SAB event, non-thermal drivers dominated the pCO 2sea increase.In April 2017, the increase in pCO 2sea was accompanied by a SSS change of −0.95 psu in the SAB, corresponding to a change in TA of −44.1 μmol kg −1 using the linear relationship (TA = 46.56× SSS + 688.24) established by Xu et al. (2020).Given a climatological DIC value of 2015 μmol kg −1 (Xu et al., 2020), this reduction in TA would lead to a +73.8 μatm increase in pCO 2sea .This value is comparable to the non-thermal pCO 2sea increase of +69.5 μatm and accounts for 74.4% of the total pCO 2sea increase (+99.2 μatm) during the 2017 SAB event.
In the MAB, CO 2 flux into the ocean significantly decreased during MHW months, especially under weakwind conditions in winter.This is particularly true for the atmospheric type of MHWs that were driven by anomalous air-sea heat flux, according to the classification of Oliver et al. (2021).In the case of atmospheric-type MHWs, perturbations in the atmosphere can lead to decreased wind speeds and downward 10.1029/2023GL105363 9 of 12 heat flux, both resulting in abnormally warm SST.These phenomena have been attributed to the zonal shift of the jet stream (Chen et al., 2014).Considering that winter MHW events above the 75th percentile also coincide with reduced wind speeds, it is plausible that this mechanism contributes to the occurrence of these MHW months.We, therefore, expanded upon the mechanism proposed by Chen et al. (2014) to explain how MHW events impact CO 2 flux in the MAB during winter months (Figure 5).In the MAB, the northward shift of the jet stream during winter causes a reduction in wind speeds, leading to a decrease in latent and sensible heat fluxes.This hinders heat loss from the ocean and results in a warmer sea surface that induces the MHW event.These elevated temperatures increase pCO 2sea , and the concurrence of weakened wind speeds further reduce CO 2 uptake by the ocean.

Summary and Conclusions
We investigated the impact of MHW events on air-sea CO 2 flux in the SAB and MAB over the past three decades.The sensitivity of the carbonate system to SST variability indicates that extreme temperatures during MHW events can influence carbon transfer between the atmosphere and the ocean.Interestingly, we found that while the two most prolonged MHW events in both regions resulted in large positive CO 2 flux anomalies (FCO 2 ′), MHWs alone were insufficient to guarantee positive flux anomalies.By analyzing MHW months that exceeded the 75th percentile of CO 2 flux anomalies, we identified specific conditions that lead to substantial flux changes.
In the SAB, MHWs were not directly responsible for flux changes, and pCO 2sea variations during MHW events were attributed to factors other than temperature.In contrast, MHWs played a more significant role in the MAB, though non-thermal drivers were still vital to produce significant positive pCO 2sea changes.The prevalence of reduced wind speeds during MHW months above the 75th percentile in the MAB led us to append the influence of MHWs on pCO 2sea to Chen et al.'s (2014) mechanism for wintertime air-sea heat flux MHW induction.We add that the northward shift of the wintertime jet stream in the MAB results in weakened wind speeds and downward heat flux, leading to increased SST and, thus, pCO 2sea .This combination of factors further reduced the ocean's ability to uptake CO 2 .

Figure 1 .
Figure 1.Average monthly sea surface temperature anomalies during (a) the 2012 MAB event from February 2012 to June 2012 and (b) the 2017 SAB event in April 2017.

Figure 2 .
Figure 2. Monthly values (blue), climatology (black), and anomalies (red) of (a and b) SST, (c and d) sea surface pCO 2 , (e and f) wind speed, (g and h) SSS, and (i and j) FCO 2 during the 2012 MAB event (left panels) and the 2017 SAB event (right panels).Blue squares in each plot highlight the 2012 and 2017 MHW events in the MAB (left) and SAB (right).Black error bars are one standard deviation associated with the monthly climatology in each panel.

Figure 3 .
Figure 3. Mean and one standard deviation (error bars) of thermal (light blue bars) and non-thermal (dark blue bars) pCO 2sea anomalies in the (a through c) MAB and (d through f) SAB during MHW months with FCO 2 anomalies during (a and d) every MHW month and (b and e) above and (c and f) below the 75th percentile of CO 2 flux anomalies.p-values from a student's t-test are presented in red text, determining whether thermal and non-thermal components are statistically different from each other in each panel.

Figure 4 .
Figure 4. Wind speed anomalies during MHW months that are climatologically CO 2 sinks and sources during (a and d) all MHW months, (b and e) above the 75th percentile of flux anomalies, and (c and f) below the 75th percentile in the MAB (top panels) and SAB (bottom panels).

Figure 5 .
Figure 5. Conceptual diagram of mechanisms connecting MHWs and CO 2 flux in the MAB.The solid arrows indicate increasing vectors or fluxes, and the dashed arrows represent decreasing vectors.The numbers represent the order of occurrence for the respective processes.