This study was performed as part of the Antarctic Cruise program (ANA08D) of the Korea Polar Research Institute (KOPRI) in autumn from March 31, 2018 to April 27, 2018 using the icebreaker, ARAON, in the Pacific sector of the Southern Ocean ( Figure 1 ). To compare the regional characteristics of the variations in surface pCO₂ and air–sea CO₂ flux in the Southern Ocean, the study area was divided into three regions, under the domains of the commission for the conservation of Antarctic marine living resources, as follows: the north of Ross Sea (NRS), Amundsen–Bellingshausen Sea (ABS), and WAP with Weddell Sea (WAP/WS) ( Figure 1A ). Our ship moved from northwest to southeast in the Pacific sector of the Southern Ocean. Thus, unlike the other two subregions, the NRS survey covered the meridional band between 60°S and 65°S and did not include areas that interact with the sea ice marginal zone of the Ross Sea. The WAP includes a Confluence Zone (CZ) between 67°W and 59°W, which is the transition zone between the Bellingshausen Sea and the Weddell Sea, and between the southern and northern parts of the WAP (S-WAP/WS and N-WAP/WS, respectively) ( Figure 1B ). In addition, two repeated cruises were conducted to survey between the Trinity peninsula and the King George Island in the Bransfield strait in order to evaluate the temporal variations in pCO₂. The first (BS1) and second (BS2) surveys in the Bransfield strait had a time interval of approximately two weeks (from April 11 to April 12 and from April 25 to April 26, respectively).

During the cruise, the underway sea surface salinity (SSS) and temperature (SST) were measured using the SEB45 and SEB38 sensors, respectively. Further, data on wind speed were collected every 10 s using an automatic weather station installed at approximately 19 m above sea level. Additionally, the underway pCO₂ ( p CO 2 sens ) was continuously determined during the ANA08D cruise using a carbon dioxide sensor (CONTROS HydroC CO₂, Germany). This sensor uses optical non-dispersive infrared gas detection to measure the oceanic pCO₂ in different environments (Fietzek et al., 2014; Marrec et al., 2014; Totland et al., 2020; Macovei et al., 2021). Zeroing was conducted every 6 h to adjust the drifting of the baseline during the observation. An equilibrator-type commercial underway pCO₂ system (GO system; General Oceanics, USA) was also installed in the ARAON. However, the equilibrator-based pCO₂ data, for comparison with p CO 2 sens , could not be acquired due to an unexpected system malfunction. Instead, surface seawater samples were collected every day during the cruise to calibrate p CO 2 sens obtained from the sensor (Marrec et al., 2014). The seawater samples were collected into 500 mL borosilicate bottles and immediately mixed with 200 μL of saturated mercury (II) chloride (HgCl₂) solution to inhibit biological activities. The mixed solution was stored at room temperature until the analysis. The total alkalinity (TA) and dissolved inorganic carbon (DIC) were measured using Versatile Instrument for the Determination of Titration Alkalinity 3C (VINDTA 3C, Germany). Routine analyses performed using certified reference materials (obtained from Prof. A. Dickson, Scripps Institute of Oceanography (USA)) ensured that the analytical precisions for the TA and DIC measurements were approximately 2 and 1 μmol kg⁻¹, respectively. The CO2SYS program developed by Lewis and Wallace (1998) was used to calculate pCO₂ ( p CO 2 calc ) from the measured SSS, SST, TA, and DIC using the carbonate dissociation constants provided by Mehrbach et al. (1973) refitted by Dickson and Millero (1987) and the ancillary thermodynamic constants reported by Millero (1995). We averaged the p CO 2 sens data collected for 20 min before and after collecting the surface seawater samples for the determination of TA and DIC. The resulting mean p CO 2 sens values were directly compared with the p CO 2 calc values calculated from the corresponding TA and DIC values, and a robust linear fit between p CO 2 sens and p CO 2 calc was observed ( Figure S1A ). The regression equation was applied to the 10-min mean p CO 2 sens values to obtain the pCO₂ data ( p CO 2 obs ) used in this study ( Figure S1B ). Before this calibration procedure, the p CO 2 sens values were corrected for the difference between in situ SST and shipboard laboratory temperature. The p CO 2 calc was calculated using in situ SST. As indicated, the p CO 2 sens -based p CO 2 obs data were used in this study because of malfunction of the high-precision GO system, which may affect the quality of the individual pCO₂ values. However, on a regional scale, we believe that the mean p CO 2 obs values can represent the overall regional conditions owing to the data calibration based on the bottle data. In addition, our data were sufficient to examine the spatiotemporal variations (i.e., relative changes) in pCO₂, because p CO 2 obs showed significant correlations with the related variables (see discussion).

The underway analysis of the gas ratios between oxygen and argon (O₂/Ar) were performed using a membrane inlet mass spectrometer (Hiden Analytical, UK) to estimate the degree of oxygen supersaturation by biological activities (see below). Kim et al. (2017) presented a more detailed measurement method for O₂/Ar. Finally, the difference between the O₂/Ar (ΔO₂/Ar) values of the sample and air-saturated water was used to isolate the O₂ produced by the biological activities.

where the superscripts “sample” and “sat” represent the O₂/Ar values in the sample and air-saturated water, respectively. The O₂/Ar can be altered by biological activities (e.g., photosynthesis and respiration) and physical processes (e.g., mixing of water masses) (Eveleth et al., 2014).

In addition to the field survey data, we used the data obtained from the SOCCOM floats (data available in soccom.princeton.edu/content/data-access). Five floats (5904184, 5904185, 5905075, 5905099, and 5905635) were found over our study area and period. Their pCO₂ (with an accuracy of ±11 μatm; Williams et al., 2018) and DIC (with an accuracy of ±4 μatm; Williams et al., 2018) values were derived from in situ pH sensor (with an accuracy of ±0.01; Johnson et al., 2016) in the floats and algorithm-based predicted TA (with an accuracy of ±5.4 μmol kg⁻¹; Carter et al., 2017). The determined SOCCOM pCO₂ values were compared with our ship-based pCO₂ data to evaluate their consistency. Additionally, we utilized SOCCOM float data to evaluate the effects of temporal variations in environmental variables such as SSS, SST, TA, and DIC on pCO₂ values. To maintain consistency with our field survey period, we used data collected in April 2018. To understand the overall changes occurring during the given month, we incorporated the first and last data recorded in April into our calculations. Following the approach proposed by Williams et al. (2018), and using the CO2SYS software (Lewis and Wallace, 1998) in conjunction with elemental stoichiometric relationships, we were able to investigate the influence of monthly shifts in each variable on pCO₂ during April.

The effect of temperature variation (i.e., thermal effect) on pCO₂ can be distinguished from the effects of other environmental factors (i.e., non-thermal effects; mainly physical mixing and biology). pCO₂ (4.23% °C^–1, Takahashi et al., 2002) values vary in response to changes in temperature, which originate from the temperature dependence of the gas solubility of CO₂ as well as the dissociation constants of the dissolved inorganic carbon species. However, non-thermal effects include alternations in TA and DIC caused by the physical mixing of water masses and biological activities (e.g., organic matter formation and degradation). The thermal and non-thermal effects on the variations in pCO₂ can be derived using the following equations (Takahashi et al., 2002):

where p CO 2 th and p CO 2 nt indicate the thermal and non-thermal pCO₂ components, respectively; the superscript “obs” represents the observed individual data; the subscript “mean” indicates a mean value for the study period; p CO 2 mean and SST^mean were 424 μatm and 1.1°C during our field survey period, respectively, and 395 μatm and 0.7°C in the SOCCOM data, respectively. In the data analysis elucidated below, the 10-min averaged values have been used for p CO 2 obs and SST obs . Separation of the contributions of the thermal and non-thermal effects in pCO₂ is generally applied to determine the dominant factors that affect the seasonal pCO₂ variations in a fixed location (e.g., Takahashi et al., 2002; Ko et al., 2022). However, in this study, this method was primarily used to correct for thermal effects associated with spatial variations in SST.

The following parameters were used to calculate air–sea CO₂ flux :   p CO 2 obs , SSS, SST, atmospheric pCO₂ ( p CO 2 air ) and the shipboard wind speed, which was adjusted for a height of 10 m (U₁₀) following the study of Thomas et al. (2005). Because wind speed could not be measured directly at the SOCCOM float locations, the mean wind speed observed during the cruise (approximately 11.9 m s^–1) was used for the floats. p CO 2 air was obtained from the US Palmer station located on the Anvers Island in WAP, and the monthly mean p CO 2 air was approximately 404 μatm in April 2018 (Dlugokencky et al., 2021; available at https://www.esrl.noaa.gov/gmd/dv/data/).

The following equation was used to calculate the air–sea CO₂ flux (in mmol m⁻² d⁻¹):

where s is the solubility of CO₂ in seawater (mol L⁻¹ atm⁻¹) and is a function of salinity and temperature (Weiss, 1974); k is the CO₂ gas transfer velocity (cm h⁻¹), which was calculated using the following equation (Wanninkhof, 2014):

where 0.251 (cm h⁻¹) (m s⁻¹) ⁻² is the optimal coefficient of CO₂ gas transfer, and Sc is the Schmidt number reported by Wanninkhof (2014).

The daily polar gridded sea ice concentration obtained from the near-real-time Defense Meteorological Satellite Program with the Special Sensors Microwave Imager/Sounder daily polar gridded sea ice concentration on April 5, 2018 was used to distinguish the marginal sea ice zone during the cruise (data available in https://nsidc.org/data/NSIDC-0081/versions/1; Figure 1A ) (Maslanik and Stroeve, 1999) as well as the changes in the sea ice concentration between April 15 and April 30 in the N-WAP/WS region ( Figure 1B ). Further, the marginal sea ice zone was estimated to be up to 200 km from the edge of the ice (Wadhams, 1986).

The overall SST in the Pacific sector of the Southern Ocean ranged from −0.7 to 3.7 °C during the period of our study ( Figure S2 ). The mean ( ± 1σ) values of the SST in the NRS, ABS, and WAP/WS were 1.9 ± 0.7 °C, 0.8 ± 0.2 °C, and 1.0 ± 1.1 °C, respectively. A relatively low SST was found near the marginal sea ice zone in the ABS and WAP/WS. The lowest SST, ranging from −0.7 to 2.8°C, was observed in the WAP/WS. The mean SST in the western part of the CZ (2.0 ± 0.4 °C) was higher than that in the N-WAP/WS (1.5 ± 1.0 °C). The SST in the CZ ranged from 1.0 to 2.5 °C and showed the greatest variability during short observation durations (1.9 days). In particular, the SST differed significantly between the N-WAP/WS and S-WAP/WS. The SST in the N-WAP/WS was clearly higher than that in the S-WAP/WS (0.0 ± 0.4 °C). Finally, significant changes in the SST were observed during the repeated observations in the Bransfield strait (i.e., between the Trinity peninsula and the King George Island). The mean SST values were 1.1 ± 0.8 °C and 1.3 ± 0.5 °C during the BS1 (from April 11 to April 12) and BS2 (from April 25 to April 26) observations, respectively.

During the study, the SSS ranged from 33.06 to 34.25 psu ( Figure S2 ). The mean SSS values in the NRS, ABS, and WAP/WS were 33.56 ± 0.09 psu, 33.24 ± 0.08 psu, and 33.71 ± 0.25 psu, respectively. The SSS in the ABS and the western part of the CZ were relatively lower than that in the other regions. In the CZ, the SSS rapidly increased from 33.28 to 33.73 psu. The SSS difference between N-WAP/WS (33.81 ± 0.14 psu) and S-WAP/WS (33.86 ± 0.17 psu) was smaller than the temporal variations observed between the BS1 and BS2 surveys (33.62 ± 0.04 and 33.93 ± 0.06 psu, respectively).

Two SOCCOM floats (No. 5904184 and 5905635) were in the south of the SB, while the other floats were between the PF and SB ( Figure 1A ). The SST recorded by the SOCCOM floats showed a decreasing trend during April ( Table S1 ). The SST near the marginal sea ice zone in the Ross Sea (float No. 5905635) and ABS (float No. 5905075) showed subzero temperatures during April. However, the SSS recorded by the SOCCOM floats showed a small variation during April, indicating a mean ( ± 1σ) SSS of 33.9 ± 0.1 psu.

The overall p CO 2 obs (i.e., 10-min average of p CO 2 sens ) ranged from 366 to 467 μatm in our study ( Figures 2A–C ). The mean p CO 2 obs values were 431 ± 6, 403 ± 18, and 429 ± 16 μatm in the NRS, ABS, and WAP/WS, respectively. In particular, p CO 2 obs in the western part of the WAP (i.e., from 85°W to 60°W) displayed a high variability, ranging from 375 to 430 μatm in the western part of the CZ and 384 to 467 μatm in the CZ. Observed p CO 2 obs values showed a similar range in the Bransfield Strait for two weeks. The p CO 2 obs values ranged from 403 to 454 μatm and from 406 to 461 μatm during the BS1 and BS2 observations, respectively. The spatiotemporal variations in the SST significantly affected the sea surface pCO₂. Accordingly, we separated p CO 2 th and p CO 2 nt for describing the pCO₂ variations in the subregions investigated in this study.

The p CO 2 th data, in which the effects of nonthermal factors (e.g., SSS, TA, DIC) were removed, gradually decreased from 473 to 424 μatm in the NRS ( Figure 2A ). Although p CO 2 th exhibited small variations in the ABS (420 ± 4 μatm), p CO 2 th was found to be reduced in the seawater around sea ice (i.e., from April 4 to April 6 in Figure 1A ). Finally, the p CO 2 th in the WAP/WS showed very large variability in time and space ( Figure 2C ). The mean p CO 2 th was 406 ± 7 μatm in the S-WAP/WS, and 424 ± 14 μatm and 428 ± 9 μatm during the BS1 and BS2 observations, respectively. Obviously, those changes in p CO 2 th were the effect of SST variations.

The p CO 2 nt data, in which thermal effects were removed, gradually increased in the NRS, ranging from 386 to 437 μatm ( Figure 2A ). The distribution of p CO 2 nt was similar to that of p CO 2 obs in the ABS, entrance of the WAP/WS, and CZ because of the small SST variation in these areas ( Figures 2A–C ). The p CO 2 nt value ranged from 365 to 439 μatm in the ABS, 354 to 418 μatm in the western part of the CZ, and 368 to 442 μatm in the CZ. In addition, the distribution of p CO 2 nt in the northern tip of the Antarctic peninsula (i.e., N-WAP/WS and S-WAP/WS) showed a high spatiotemporal variation, although it was opposite to the p CO 2 th distribution ( Figure 2C ). The p CO 2 nt value ranged from 388 to 475 μatm in the N-WAP/WS and from 433 to 472 μatm in the S-WAP/WS. In general, changes in p CO 2 nt are mainly associated with biological carbon fixation and vertical mixing reducing and enhancing pCO₂, respectively.

Considering the five SCCCOM floats, pCO₂ ranged from 358 to 421 μatm during April ( Table S1 ). A relatively low pCO₂ was found in the floats located near the sea ice (i.e., float No. 5905075 and 5905635). In general, pCO₂ increased with time during April, except for the float in the WAP (i.e., float No. 5904185). Float No. 5905635 showed the broadest pCO₂ range from 358 to 386 μatm. p CO 2 th gradually decreased from 433 to 360 μatm with time during April because of the decreasing SST. In contrast, p CO 2 nt generally increased from 373 to 420 μatm, except for the float No. 5904185.

The overall air–sea CO₂ flux in the study area ranged from −59.4 to 140.8 mmol m⁻² d⁻¹ during the cruise ( Figures 2D–F ). The air–sea CO₂ flux in the NRS and ABS ranged from 3.1 to 18.8 mmol m⁻² d⁻¹ and −12.7 to 17.3 mmol m⁻² d⁻¹, respectively. The air–sea CO₂ flux in the WAP/WS ranged from −59.4 to 140.8 mmol m⁻² d⁻¹; however, it showed a high spatial variability depending on the local environmental condition. The air–sea CO₂ flux in the CZ, N-WAP/WS, and S-WAP/WS ranged from −33.4 to 117.6 mmol m⁻² d⁻¹, 0.0 to 82.4 mmol m⁻² d⁻¹, and 0.0 to 140.8 mmol m⁻² d⁻¹, respectively. Finally, a significant change in the air–sea CO₂ flux was observed in the Bransfield strait, with a mean of 21.3 ± 14.9 mmol m⁻² d⁻¹ and 8.4 ± 9.2 mmol m⁻² d⁻¹ during the BS1 and BS2 observations, respectively.

The air–sea CO₂ flux of the five SOCCOM floats ranged from −13.9 to 5.0 mmol m⁻² d⁻¹ during April ( Table S1 ). Notably, two floats adjacent to sea ice exhibited a significant uptake of atmospheric CO₂. The mean air–sea CO₂ fluxes for these floats were −12.6 ± 1.2 mmol m⁻² d⁻¹ and −8.3 ± 4.9 mmol m⁻² d⁻¹ for floats NO. 5905075 and No. 5905635, respectively. In contrast, the other three floats showed a weak efflux of CO₂ to the atmosphere. The mean values for these floats were 2.0 ± 0.5 mmol m⁻² d⁻¹, 0.7 ± 1.3 mmol m⁻² d⁻¹, and 3.9 ± 1.2 mmol m⁻² d⁻¹ for float No. 590418, 5904185, and 5905099, respectively.

During the cruise, the ΔO₂/Ar in the study area ranged from −1.8 to 4.9% ( Figure S3 ). The mean ΔO₂/Ar values in the NRS and ABS were 0.9 ± 1.2 and −0.3 ± 2.3%, respectively, whereas that the WAP/WS was −3.7 ± 1.9%. The highest ΔO₂/Ar (4.9%) was observed in the western part of the CZ, while the lowest ΔO₂/Ar (−9.5%) was shown in the northern tip of the Antarctic peninsula. The ΔO₂/Ar in the CZ ranged from −8.8 to 4.9%, and the ΔO₂/Ar in the N-WAP/WS (−2.8 ± 1.6%) was higher than those in the S-WAP/WS (−4.2 ± 1.9%). In the Bransfield strait, the mean ΔO₂/Ar during the BS1 observation (−2.0 ± 0.7%) was relatively higher than that during the BS2 observation (−3.0 ± 2.0%).

To assess the consistency between the shipboard underway and SOCCOM float pCO₂ data, which has an 11 μatm of uncertainty at 400 μatm (Williams et al., 2017), two SOCCOM floats (No. 5905099 and 5905075) were selected because of their proximity with our cruise track. Nonetheless, the distances from our cruise track to the two floats were ~145 km (No. 5905099) and ~60 km (No. 5905075) ( Table 1 ). In addition, there was a 3-day difference in the observation date between the shipboard and float pCO₂ data. The pCO₂ differences between the two platforms were ~19 μatm (No. 5905099) and ~33 μatm (No. 5905075). In the case of the 5905075 float, the pCO₂ difference reduced to ~4 μatm if the effect of the SST difference was corrected (i.e., p CO 2 nt was compared). In contrast, for the No. 5905099 float, there was still a difference of 20 μatm even after the temperature correction was applied. This discrepancy may be attributed to the considerable distance (~145 km) between the two platforms. Fay et al. (2018) also conducted a comparison between shipboard pCO₂ data and data collected within three days from SOCCOM floats located within a 75 km radius of their ship’s track. Their findings also indicated a similar difference (~ 24 μatm) to those observed in this study, which ranged between 19 and 33 μatm.

To identify the factors that control pCO₂ in the study area, we determined the influence of each factor (e.g., SST, SSS, TA, and DIC) using CO2SYS and SOCCOM float pCO₂ data ( Table 2 ). CO₂ solubility in seawater is a function of salinity, temperature, and pressure (Weiss, 1974). Under a constant pressure (surface), a decrease in salinity and temperature can enhance the CO₂ solubility. The SST-driven change in pCO₂ ( Δ p CO 2 Δ SST = [pCO₂ at the last observation in April] – [pCO₂ at the first observation in April]) ranged from −12.7 to −6.5 μatm in the study area because of the decreasing SST in April. However, the SSS-driven changes ( Δ p CO 2 Δ SSS ) were much smaller ( Δ p CO 2 Δ SSS = ± 0.4 μatm) because of the relatively low variations in the SSS ( Δ SSS= ~0.05). The changes in pCO₂ ( Δ p CO 2 Δ TA ) associated with TA ranged from −12.6 to 7.6 μatm, while those in pCO₂ ( Δ p CO 2 Δ DIC ) associated with DIC ranged from 5.9 to 30.5 μatm in April 2018. Finally, residual in pCO₂ ( Δ p CO 2 Δ RESID ) showed a variation from 0.0 to 2.3 μatm after the removal of the Δ p CO 2 Δ SST , Δ p CO 2 Δ SSS , Δ p CO 2 Δ TA , and Δ p CO 2 Δ DIC effects. These results suggest that the effects of the SST, TA, and DIC are significant in all the floats; however, the SSS exudes the least impact on pCO₂ due to small variation of SSS in April 2018.

ΔTA and ΔDIC (estimated from SOCCOM data) were decomposed to understand the processes causing their variations, such as freshwater input, organic matter, calcium carbonate (CaCO₃), and gas exchange, using a formula reported by Williams et al. (2018) ( Table 3 ). In their approach, the SSS, nitrate ( NO 3 − ) , DIC : TA ratio were used to separate the effect of freshwater, organic matter decomposition/synthesis, and calcification. Because the suggested method requires nitrate ( NO 3 − ) concentration, we could use three floats except for the 5904184 and 5904185 floats, in which ( NO 3 − ) concentrations were not available. The analysis suggested that freshwater and CaCO₃ dissolution were the dominant factors for ΔTA. In contrast, the variations in DIC were not associated with the freshwater input, biological activities, and gas exchanges, and thus, the largest values of the residual term were obtained in this case. Since we did not measure the wind speed directly at the SOCCOM float locations, the gas exchange term could not be accurately determined. The gas exchange term did not affect other terms accompanying the changes in TA. Williams et al. (2018) studied the eastern Antarctic Ocean region and suggested that freshwater input and CaCO₃ dissolution/formation govern the TA variations in the ASZ (i.e., between PF and SB), while the TA in the SSIZ is substantially influenced by the freshwater input (i.e., south of SB). In addition, they suggested that organic matter production/remineralization is the most dominant factor that influences the changes in DIC, followed by the freshwater input. Our results for TA were consistent with those of the previous study, whereas those of DIC were not identical.

The p CO 2 obs and p CO 2 nt values in the NRS increased as the icebreaker moved southeast through the PF and SB, while the p CO 2 th in this region decreased during the cruise ( Figures 1A , 2A ), indicating that the non-thermal factors control the p CO 2 obs variations in the NRS. According to a previous study, p CO 2 obs between the PF and SB ranged from 331 to 354 μatm in December 2005 (Guéguen and Tortell, 2008). Even considering the atmospheric CO₂ growth from 378 to 404 μatm, the value observed in summer was lower than those shown in our autumn survey. The relatively high p CO 2 obs in this study might have resulted from the increase in the mixed layer depth (MLD) due to strong winds and the remineralization of the organic matter produced in summer. The mean U₁₀ in the NRS was 9.9 ± 1.5 m s⁻¹, and ranged from 7.7 m s⁻¹ to 13.0 m s⁻¹ in April, which was relatively greater than that in summer (ranging from 4.0 to 7.9 m s⁻¹) (Guéguen and Tortell, 2008; DeJong and Dunbar, 2017). Along the survey track, the surface water in the SB (62.5°S) was the most saline and had a relatively high temperature, indicating an influence from the upwelling of or enhanced mixing with the sub-surface water, which is relatively rich in CO₂ and nutrients ( Figure S2A ). In addition, a correlation analysis indicated that the main controlling factors of p CO 2 nt were SSS (e.g., physical effects such as changes in mixing and sea ice) and ΔO₂/Ar (i.e., biological effects) in the north and south of the SB, respectively ( Figure 3 ). Guéguen and Tortell (2008) found no significant correlation between pCO₂ and the SSS over the ocean from New Zealand to the Ross Sea during summer. Thus, the significant correlation between SSS and p CO 2 nt was a local characteristic limited to the 62.5 to 65°S zone or April (autumn). It is well known that CDW upwelling occurs near the SB, and the upwelled CDW diverges from the SB to the PF and to the Antarctic shelf region (Speer et al., 2000). At the SB, high biological activities are found because of the upwelling of the nutrient-rich subsurface water (Sokolov and Rintoul, 2007). In addition, Rigual-Hernández et al. (2015) reported a relatively high concentration of chlorophyll-a in the Southern ACC Front (SACCF, located in ASZ) in the Antarctic coastal area, while low chlorophyll-a concentrations were found from the SACCF to New Zealand. Therefore, p CO 2 nt in seawater, which was transferred from the SB to coastal areas, might have been altered by biological activities. p CO 2 nt decreases because of the net biological CO₂ uptake (i.e., ΔO₂/Ar > 0) when the phytoplankton production surpasses the ecosystem respiration. In contrast, p CO 2 nt increases when ΔO₂/Ar< 0. Accordingly, it was likely that p CO 2 nt was largely controlled by the upwelling at the SB and the following biological activities in the NRS in April 2018. Our data clearly highlighted a shift in the ecosystem dynamics associated with the inorganic carbon cycle across the SB.

The NRS acted as an overall CO₂ source (8.8 ± 3.5 mmol m⁻² d⁻¹) for atmospheric CO₂ during April. The amount of CO₂ released from the ocean to the atmosphere was increased in the eastern part of the NRS, which had a relatively higher p CO 2 obs (than ABS), high wind speeds, and a heterotrophic ecosystem (i.e., negative ΔO₂/Ar) in the observation period ( Figures 2A, C and S3A ). The NRS and the Ross Sea exhibited different patterns with respect to air–sea exchange of CO₂. Most previous studies on air–sea CO₂ flux were conducted during spring and summer in the southwestern Ross Sea, and suggest that the air–sea CO₂ flux ranges from −132 to 10 mmol m⁻² d⁻¹ (Arrigo and Van Dijken, 2007; DeJong and Dunbar, 2017). Thus, Ross Sea has been considered to act as a CO₂ sink during spring and summer because of the enhanced biological activities (DeJong et al., 2017; Rivaro et al., 2017). In particular, the Terra Nova Bay in the Ross Sea acts as a strong sink for atmospheric CO₂ (up to −132 mmol m⁻² d⁻¹) (DeJong and Dunbar, 2017). The discrepancies between the findings in the NRS and the previous research results from the Ross Sea can likely be attributed to the differences in productivity between the two regions and the variations in the timing of the respective investigations. In general, the rate of photosynthesis by the phytoplankton increases from spring to summer because of the abundance of sunlight and nutrients due to the shrinking of the sea ice (Zhang et al., 2010; Hill et al., 2018; Lannuzel et al., 2020). This effect can help elucidate the strong CO₂ sink observed in studies conducted in the Ross Sea, where most previous research is primarily focused during the summer. In autumn, pCO₂ likely increases due to enhanced vertical mixing and the decomposition of organic matter that has accumulated from biological activities in previous seasons. These factors are likely to outweigh CO₂ fixation in the pelagic water (i.e., NRS) with relatively lower nutrient concentrations, especially during autumn when light availability is reduced. In addition, a relatively high pCO₂ in surface seawater and high wind speed in autumn might facilitate CO₂ emissions from the NRS. Therefore, the disagreement (CO₂ source vs. sink) between our results and those previously reported results may be attributed to the meridional variations in the biogeochemical properties.

While moving through the ABS region, the icebreaker was over the south of the SB for the first three days, and then crossed the SB to sail to the north of the SB for the next three days ( Figure 1A ), during which the marginal sea ice zone was also encountered (i.e., from April 4, 17:20 UTC to April 7, 05:10 UTC). A correlation analysis showed that the SSS could most significantly explain the variations in p CO 2 nt in the ABS, with some exception data collected in the east of 90°W ( Figure 4A ). The significant correlation between the SSS and p CO 2 nt in this region possibly resulted from the CO₂-rich saline CDW upwelling at the SB as well as the brine rejection effect in the marginal sea ice zone. When sea ice begins to form in early autumn, most of the DIC are rejected from the sea ice structure (Moreau et al., 2016; Mo et al., 2022). Thus, the surrounding seawater has high DIC and salinity, which in turn directly enhance p CO 2 nt . Discrete profiles of the ocean inorganic carbon parameters were observed in the ABS from April 14 to May 06 in 2018 with the aid of the RV Nathaniel B. Palmer, which significantly overlapped with our cruise track (Macdonald et al., 2021). During this observation, the SST (−1.4 ± 0.3 °C) and SSS (33.7 ± 0.2) were found to be colder and saltier compared to our observations. The mean pCO₂ (358 ± 10 μatm) was lower than our result due to the lower SST than in this study, whereas the mean p CO 2 nt (398 ± 12 μatm) was similar to our result ( Table 4 ). Another observation study, whose track partially overlapped with our cruise track, was conducted in the Bellingshausen Sea from February 30, 2019 to March 22, 2019 using the RV Nathaniel B. Palmer (Sutherland et al., 2019). In their study, the mean pCO₂ and p CO 2 nt were 355 ± 26 and 384 ± 29 μatm, respectively ( Figure 4B ). However, in our case, the mean pCO₂ and p CO 2 nt were 403 ± 18 and 407 ± 20 μatm, respectively, which are ~20 μatm higher than p CO 2 nt reported by Sutherland et al. (2019). This significant difference may be due to the seasonal changes in the sea ice conditions (dilution and concentration because of sea ice melting and formation, respectively) and biological activities (ΔO₂/Ar< 0 in our study) between late summer and early autumn. The brine rejection from sea ice formation during autumn has a high pCO₂ value, causing an increase in p CO 2 nt . A decrease in ΔO₂/Ar at the surface was caused by an increase in respiration relative to production by phytoplankton during autumn, as well as the mixing of surface water with the subsurface water, which has lower oxygen concentration (Eveleth et al., 2014).

The air–sea CO₂ flux in the ABS ranged from −12.7 to 17.3 mmol m⁻² d⁻¹. Overall, the marginal sea ice zone of the ABS region acted as a weak CO₂ sink (−1.8 ± 4.2 mmol m⁻² d⁻¹) for atmospheric CO₂, whereas the open area served as a weak CO₂ source (0.9 ± 5.1 mmol m⁻² d⁻¹) during the study period ( Figure 2E ). According to previous studies, the air–sea CO₂ fluxes in the Amundsen Sea and Bellingshausen Sea range from −16 to 14 mmol m⁻² d⁻¹ (Tortell et al., 2012; Mu et al., 2014; Kim et al., 2018) and from −6.5 to 0 mmol m⁻² d⁻¹ (Álvarez et al., 2002; Ito et al., 2018), respectively, during summer. In particular, previous studies identified strong atmospheric CO₂ uptake in polynyas or areas with diatom blooms (Álvarez et al., 2002; Tortell et al., 2012). However, it is important to emphasize that in this study, significant CO₂ fluxes were observed due to high wind speeds, even when the pCO₂ difference between the atmosphere and ocean was small. ( Figure 2E ).

A large part of the WAP region belongs to the Bellingshausen Sea. However, in this study, we separated the WAP region from the rest of the Bellingshausen Sea to highlight the complex pCO₂ variations in the WAP/WS region. First, at the entrance (66°W–68°W in this study) of the WAP before arriving at the CZ zone, we found a sudden increase in the ΔO₂/Ar around the SB ( Figure S3C ). Thus, we expected that biological activities governed the pCO₂ variations in this region. Correlation analysis showed that p CO 2 nt was correlated with the SSS and ΔO₂/Ar in the south (66°W–68°W) and north (66°W–65.5°W) of the SB at the entrance of the WAP, respectively ( Figure 5 ), similar to the results obtained from the NRS. The SSS– p CO 2 nt relationship unlikely originated from the sea ice formation, because the entrance of the WAP was ice-free during the study period ( Figure 1B ). Shadwick et al. (2021) suggested that the MLD in the WAP gradually deepens from summer (~20 m in January) to winter (~100 m in November), and reported that the MLD is approximately 60 m in April (Panassa et al., 2018). Therefore, the high p CO 2 nt values in this region result from the intensification of the vertical mixing, which leads to the upwelling of CO₂-riched seawater from the subsurface layer.

In the CZ, various water masses originating from the Bellingshausen Sea, WS, and meteoric water are usually mixed (Meredith et al., 2013; Cook et al., 2016), resulting in the greatest variation in the SST and SSS compared to those in the other subregions. p CO 2 nt in this region was strongly inversely correlated with the ΔO₂/Ar ( Figure 6 ). Despite overall heterotrophic condition (i.e., mean ΔO₂/Ar< 0), the CZ area showed the broadest variations (−8.8 to 3.7%) in ΔO₂/Ar, among the three subregions, indicating correlations between pCO₂ and the ΔO₂/Ar. The pCO₂ sensitivity to a given ΔO₂/Ar variation (i.e., a regression slope between ΔO₂/Ar and pCO₂) was similar in the CZ and NRS. Previous studies have demonstrated that the concentrations of dissolved organic carbon during summer range from 35 to 127 μmol kg^–1 in the Gerlache strait (between 64°S – 65°S and between 61°W – 64°W) (Doval et al., 2002; da Cunha et al., 2018) and from 44 to 64 μmol kg^–1 in the Bransfield Strait (Ruiz-Halpern et al., 2014). These values are relatively higher than or similar to those in the other regions; 38–56 μmol kg^–1 in the Ross Sea (Ogawa et al., 1999; Carlson et al., 2000; Bercovici et al., 2017) and approximately 44–63 μmol kg^–1 in the Amundsen Sea (Fang et al., 2020). The high accumulation of dissolved organic carbon suggests a potential for respiration-based CO₂ production in the CZ area.

The transition between the N-WAP/WS and S-WAP/WS exhibited a sharp shift in pCO₂ ( Figure 2C ). p CO 2 nt ( p CO 2 th ) in the N-WAP/WS was lower (higher) than that in the S-WAP/WS (mainly northwestern part of the Weddell Sea). The mean p CO 2 obs ( p CO 2 nt ) values were 432 ± 14 (425 ± 25) and 435 ± 11 (454 ± 8) μatm in the N-WAP/WS and S-WAP/WS, respectively. In contrast, Ito et al. (2018) reported that pCO₂ in the N-WAP/WS (361 to 392 μatm) was higher than that in the Weddell Sea (343 to 376 μatm) during summer in 2008–2010. This inconsistency possibly results from the seasonal changes in the sea ice melting/formation and biological processes from summer to autumn. The chlorophyll-a concentration in the Weddell Sea was higher than that in the N-WAP/WS during the summers of 2008 and 2009 (Ito et al., 2018). However, from April 15 to April 30, the sea ice over most of the Weddell Sea increased by up to ~80% ( Figure 1B ). Consistently, the S-WAP/WS showed a slight SSS increase in April, relative to that observed in the study period of Ito et al. (2018). Combined with the reduced light availability, associated with the increased sea ice cover, brine rejection in the Weddell Sea during autumn likely caused the relatively high p CO 2 obs in the S-WAP/WS.

Two observations were conducted with a time interval of 14 days from the Trinity peninsula to the Maxwell Bay in the Bransfield Strait ( Figures 1B , 7 ) to evaluate the mechanism underlying the effect of ice formation in the Weddell Sea on the pCO₂ variation in Bransfield Strait. Although the ocean current system in the Bransfield Strait is complex (Krechik et al., 2021), a relatively warm water mass that originates from the ACC and Bellingshausen Sea is generally introduced into the northern part of the Bransfield Strait. In contrast, relatively cold water from the Weddell Sea flows into the southern part of the Bransfield Strait. The north-to-south shift in p CO 2 obs and p CO 2 nt could be explained using the above-mentioned current system in the Bransfield Strait (Green shadings in Figure 2C ). p CO 2 nt in the southern Bransfield Strait was lower than those in the northern Bransfield Strait during the first observation (BS1) ( Figure 7A ). However, during the second visit (BS2), p CO 2 nt was drastically elevated from< 400 μatm to >450 μatm, particularly in a part of the southern Bransfield Strait (near 63°S and 58°W) ( Figure 7B ). The SSS at 63°S and 58°W were 33.67 and 34.00 during the first and second observations, respectively. The SSS increase possibly originated from the corresponding increase in the SSS due to the enhanced sea ice concentration in the Weddell Sea. The values of p CO 2 nt in the northern boundary of the Bransfield Strait (i.e., close to the King George Island) during BS2 were higher than those found in the southern boundary of the Bransfield Strait, possibly due to the lower ΔO₂/Ar in the northern edge of the Bransfield Strait ( Figure S3C ). The mean ΔO₂/Ar in the northern (black box in Figure 7B ) and southern (red box in Figure 7B ) boundaries of the Bransfield Strait were –2.8 ± 1.6 and –4.2 ± 1.9%, respectively. The contribution of the water masses flowing from the ACC, Bellingshausen Sea, and Weddell Sea to the Bransfield Strait could be affected by the Southern Annual Mode (SAM). During the positive SAM period, contributions from the ACC and Bellingshausen Sea were more dominant, while the influence of the water masses from the Weddell Sea was dominant during the negative SAM period in the Bransfield strait (Vorrath et al., 2020). We speculate that the current from the Weddell Sea to the Bransfield strait increased in April 2018 because of the negative SAM index (–1.66) (Marshall, 2003).

Although the air–sea CO₂ flux in the WAP/WS considerably varied depending on the subregions, the WAP appeared to act as a CO₂ source (18.1 ± 27.6 mmol m⁻² d⁻¹) during this study ( Table 4 ). The mean air–sea CO₂ flux was estimated to be 5.0 ± 20.8 mmol m⁻² d⁻¹ at the entrance of the WAP, 32.5 ± 35.6 mmol m⁻² d⁻¹ in the CZ, 9.4 ± 10.0 mmol m⁻² d⁻¹ in the N-WAP, and 23.8 ± 33.3 mmol m⁻² d⁻¹ in the S-WAP. In particular, the strong CO₂ emissions in the WAP/WS were caused by a high wind speed (13.5 ± 7.8 m s⁻¹). Álvarez et al. (2002) reported that a strong oceanic CO₂ uptake (−6.5 mmol m⁻² d⁻¹) occurred during summer due to the diatom bloom in the western WAP region (the northern part of CZ). A few researches also reported that the Bransfield Strait acted as an atmospheric CO₂ sink during spring and summer, with air–sea CO₂ fluxes ranging from −3.4 to 1.6 mmol m⁻² d⁻¹ (Ito et al., 2018; Caetano et al., 2020). Shadwick et al. (2021), who estimated the annual air–sea CO₂ flux using a mooring system suggested that the entrance of the WAP region acted as a CO₂ absorber in April, with an air–sea exchange of approximately −2 mmol m⁻² d⁻¹, which indicates potentially large temporal variations in this region.

Our data indicate that non-thermal factors predominantly controlled pCO₂ in the Pacific sector of the Southern Ocean in April 2018, with the exception of the N-WAP/WS and S-WAP/WS ( Figure 2 ). In the N-WAP/WS and S-WAP/WS, both thermal and non-thermal factors significantly influenced pCO₂. Williams et al. (2018) reported that biological processes primarily governed summer pCO₂ in the ASZ and SSIZ. Nevertheless, such predominant biology-driven control of pCO₂ was found in limited subregions, namely the southern SB of the NRS and the CZ of the WAP/WS regions. The SSS- p CO 2 nt correlation observed in the NRS, ABS, and entrance of WAP resulted from upwelling, sea ice formation, and enhanced vertical mixing in the NRS, ABS, and entrance of WAP, respectively. Our findings, distinct from those reported from summertime investigations, highlight the importance of cold-season observations. Our high-resolution data revealed a complex mechanism controlling pCO₂ in the Pacific sector of the Southern Ocean, suggesting that ship-based underway observations are essential for understanding the carbon cycle in the Antarctic waters.