Spatiotemporal variations and controlling factors of the surface pCO2 in the northern South China Sea

The focal area of this study was the northern South China Sea (NSCS; 18.5°-22.5°N, 1.0°-8.0°E), located on the northwestern shelf of the SCS. The surface pCO2 (pCO2w), sea surface temperature (SST) and sea surface salinity (SSS) was measured by continuous measurement system during four field surveys to show their spatiotemporal variations in the NSCS. Low pCO2w in the southwestern region of PRE were observed in four seasons, particularly in spring and summer. The Pearl River diluted water (PRDW) discharged a large amount of nutrients into nearshore waters, which promoted phytoplankton propagation and CO2 uptake in spring and summer. On the contrary, primary productivity was low in winter and late autumn, runoff input of dissolved inorganic carbon (DIC) played important role in increasing the pCO2 in nearshore waters, so the PRDW controlled the spatiotemporal variations of pCO2w in the coastal and inner shelf. The pCO2w in the outer shelf and slope were relatively high in four seasons, and SST was the critical controlling factor of pCO2w, sea-air CO2 exchange also played role in the seasonal scales of pCO2w, the impact of weather and climate events on the variations of pCO2w and sea-air CO2 flux in the short term also were remarkable. Generally, the NSCS acted as sink of atmospheric CO2 in spring, late autumn and winter, particularly in latter two seasons, in contrast, it was weak CO21 source in summer.


Introduction
The emissions of a large amount of CO 2 and other greenhouse gases into the atmosphere leaded global warming and ocean acidification since the onset of the industrial revolution. The oceanic sink has absorbed ~40% of the CO 2 due to human activities, which was emitted from the burning of fossil and industrial production [1]. Marginal seas, despite comprising only 7% -8% of the global ocean, played critical role in global carbon cycle. First, marginal seas were important links of carbon exchange among sea, atmosphere and terrestrial ecosystem [2]. Second, marginal seas accounted for 15% -30% of primary productivity and ~80% of organic matter burial in the global ocean [3]. Lastly, because ~ 60% of the world's population lived near the 100km of coastlines [4], anthropogenic activities input abundant organic matter into coastal waters. Thus far, coastal waters have received increasing attention in studies of global carbon cycle [5][6][7]. However, the dynamic nature of coastal waters and the variations of pCO 2w were complicated, and most marginal seas were absent of sufficient field investigations. Hence, the controlling factors of pCO 2w and sea-air CO 2 flux should remain huge challenges in marginal seas. IOP Conf. Series: Earth and Environmental Science 569 (2020) 012092 IOP Publishing doi:10.1088/1755-1315/569/1/012092 2 The SCS covered an area of 3.5 × 10 6 km 2 and located in tropical and subtropical regions, was the third largest marginal sea in the world. It included central basins exceeding 5000 m in depth as well as wide northwestern and southern shelves. The northwestern shelf and slope regions were oligotrophic, had low productivity, and were net sources of atmospheric CO 2 , particularly in summer [8,9]. Carbon cycle in the SCS has been affected by other seas through water mass exchange. The deep water in the SCS and water at ~2000 m in the West Philippine Sea had similar characteristics [10]. The west area of Luzon Strait was affected by Kuroshio [11], the intrusion of the CO 2 -enriched Kuroshio from North Pacific deep water might have influenced the carbon cycle in the SCS. The SCS was connected with the Sulu Sea via the 420 m deep Mindoro Strait and with the East China Sea (ECS) via the Taiwan Strait [12]. Moreover, the SCS was fed by the two major Pearl River and Mekong River runoff. Thus, the SCS encompassed various physical and biogeochemical domains, high resolution surveys should be warranted to explore controlling factors of pCO 2w in the SCS.
The spatial and seasonal changes of the pCO 2 in the PRE were reported [13,14], the spatiotemporal distributions of pCO 2w and sea-air CO 2 fluxes were studied in different seasons and domains in SCS [15], but the controlling mechanisms of the PRDW on affecting the distribution of pCO 2ws in nearshore waters should need further studies. The nearshore waters had high productivity and served as a strong CO 2 sink because of the nutrient supply, and the productive estuarine plume might be extensive in flooding seasons [16]. However, primary productions promoted by runoff might be offset by heterotrophic activity enhanced by terrestrial organic carbon [17], and the runoff inorganic carbon input also directly elevated pCO 2w in nearshore waters [18]. In addition, the weather and climate events significantly have affected the distribution of pCO 2w and sea-air CO 2 fluxes in the short term. In this study, spatiotemporal variations of pCO 2w were investigated by using continuous and autonomous observation system and in-situ data from seasonal shipboard sampling, and the controlling dynamics of pCO 2w in the NSCS were explored relative to the diluted waters, thermohaline characteristic, and weather and climate events, especially the controlling mechanisms of the PRDW was discussed.

Materials and methods
The focal area of this study was the NSCS, which was located on the northwestern shelf of the SCS (Figure 1). The SCS monsoon was an important part of the East Asian monsoon system. The northeast monsoon began to bring dry air to the SCS in August and reached its peak in December, at which time the maximum wind speed appeared around the Taiwan  , and offshore (OS). The underway continuous measurement system (GO8050) was used to measure pCO 2 , the average collected frequency of datasets was one group per two minutes. The device was calibrated by zero gas and the highest CO 2 standard gas, the standard curve was drew by 4 standard gases at 0.00, 218.16, 346.40 and 584.75 ppm. These standard gases with < 0.03% uncertainty were made by the Chinese Academy of Meteorological Sciences. Seawater sample was continuously pumped from shaft in the fore deck and transported to the ship's laboratory at large flow and high speed to reduce heating. In order to avoid flue gas pollution and human disturbance, atmospheric air was sampling from front section and upper deck of ship. The temperature in the equilibration system was measured by platinum resistance thermometer, there was small difference of less than 0.2 °C on temperature between the ship's shaft in fore deck and the equilibrator. The pCO 2 in the equilibrator was calibrated to in situ pCO 2w by using saturated water vapor pressure and the temperature effect coefficient of 4.23% °C -1 [22]. SST and SSS was measured by CTD sensor (SBE21), SST and dissolved oxygen (DO) was measured by oxygen sensor (Aanderaa oxygen optode 3835), while SST, SSS and DO was also measured by multi-parameter water quality detector (YSI 6600 meter). All SST, SSS and DO data from different sensors were mutually corrected before or during the cruises, the results of different sensors were consistent. The error level between different SST sensors was less than 0.1 °C and the error between DO sensors was less than 1 μmol/L.
The formula for the sea-air CO 2 flux calculation was F = k×s×ΔpCO 2 , k (cm·h -1 ) was the gas transfer velocity of CO 2 and calculated by k = 0.31u 2 (Sc/660) -0.5 in the W92 formula [23]. It has been suggested that the parameterization of the W92 formula may overestimate the gas transfer velocity, as k = 0.27u 2 (Sc/660) -0.5 in the Sweeney equation [24]. However, for ease of comparison with most other studies, the k value was calculated on the basis of the W92 formula, s was the solubility of CO 2 in seawater [25], and ΔpCO 2 (pCO 2w -pCO 2a ) was the difference between pCO 2a and pCO 2w . Wind speed was the major factor driving sea-air CO 2 flux, wind speed, barometric pressure and relative humidity was collected by shipborne meteorological station at height of 10 m above the sea level. Thus, the mean of wind speed during every cruise was used in this study to estimate the sea-air CO 2 flux. A negative value indicated that the sea was sink of atmospheric CO 2 , a positive value represents that the sea was source of atmospheric CO 2 .

Spatiotemporal distributions of pCO 2w , SST, and SSS
The spatiotemporal distributions of pCO 2w , SST, and SSS in the NSCS for a composite year were shown in Figure 2. In order to better describe the distributions of pCO 2w , SST, and SSS and research controlling mechanisms of pCO 2w , the sea area influenced by PRDW was called the coastal area and inner shelf in this study where SSS was less than ~33, the other sea area was called the outer shelf and slope where SSS was more than ~33.
Low SST and high SSS in winter occurred in the entire observation zone (Figure 2a). The SST ranged from 14.85 °C to 24.59 °C, and the SSS was between 31.24 and 34.54 (Table 1). The low SST and high SSS can be attributed to the presence of cold air during the observation. The SST and SSS in the coastal area and inner shelf were both lower than other regions, averaging 15.99 °C and 32.09 (Table 1). The average of pCO 2w in the NSCS (except transect A) was (340±19) μatm ( Figure 2c) in winter, and among the average of pCO 2w (334±9) μatm to (367±11) μatm [15], but the average of pCO 2w was (381±3) μatm in transect A and higher than other transects, ΔpCO 2w between transect A and other transects reached ~41μatm.   (Table 1), and apparently higher than that in winter. The SSTs in transects NS, A, and B were lower than those in transects OS, C, and D (Figure 2d). This result was consistent with the changes of atmospheric temperature during the field investigation. The lowest SSS of 25.16 was observed in the southwestern region of the PRE (Table 1, Figure 2e), indicating the effect of Pearl River runoff on transect NS. The average of pCO 2w was (300±42) μatm and also low in the southwestern region of the PRE (Figure 2f), ranging from 198 μatm to 379 μatm (Table 1), and was obviously lower than the result off the Pearl River Estuary where the average of CO 2 was (361±10) μatm and ranged from 339 μatm to 378 μatm [15]. Interestingly, this result was consistent with the lowest SSS observed in the same region, which IOP Conf. Series: Earth and Environmental Science 569 (2020) 012092 IOP Publishing doi:10.1088/1755-1315/569/1/012092 5 could demonstrate the obvious effect of Pearl River runoff on the distribution of pCO 2w . However, pCO 2w was not uniform with SSS in other transects, and the pCO 2w was higher in transect A and B than C, D and OS. Moreover, the distribution of pCO 2w was similar to that of SST, which indicated that SST was a probable controlling factor in the outer shelf regions. The low pCO 2w in the coastal area and inner shelf accompanied by a ΔpCO 2 value of approximately -85 μatm indicated a remarkable CO 2 sink.
The SST and SSS in summer was the highest in all the four seasons, varying from 29.50 °C to 32.12 °C and from 28.66 to 33.83, respectively (Table 1), but the ranges of SST and SSS were relatively small. The SSS was slightly lower in the southwestern region of the PRE than other transects, and the plume of PRDW was smaller in summer than that in spring (Figure 2e, 2h). The range of pCO 2w was from 188 μatm to 435 μatm and the average of pCO 2w was (389±30) μatm (Table  1). The low endpoint appeared in the transect NS, which was consistent with low SSS (Figure 2i). In contrast, the average pCO 2w in the outer shelf and slope was as high as (397±10) μatm (Table 1), the ΔpCO 2 was ~25 μatm, and was within the previous studies results which was between (383±11) μatm and (404±6) μatm [15].
The SST was between 17.03 °C and 25.10 °C, the SSS ranged from 30.66 to 34.23 in late autumn, low SST and high SSS existed in NSCS and was similar to the results in winter (Figure 2j, 2k), SST and SSS was both lower in the transect NS affected by PRDW than other transects (Figure 2j, 2k). The pCO 2w sharply dropped down under the influence of cold air, the average of pCO 2w was (348±19) μatm with ranging from 292 μatm to 397 μatm (Table 1), that was similar to that in winter in the coastal area and extended to the entire shelf (Figure 2i). The pCO 2w in late autumn was lower than previous studies with the average of (358±4) μatm to (377±18) μatm [15].

Effect of diluted waters on pCO 2w in the coastal and inner shelf
The PRDW was the largest runoff in the NSCS, with an annual discharge of about 3.3×10 11 m 3 . The PRDW flowed to the southwest after exiting the PRE under effect of Coriolis force [26]. Only in summer, the surface of PRDW might shift to flow toward the northeast away from the coast under the influence of the southwest monsoon. The SSS in the four seasons 2006 and 2007 decreased gradually from northeast to southwest in PRE [27], and the bottom of PRDW expanded westward for the whole year [28]. The lowest SSS was in the southwestern region off the PRE (Figure 2), and the outer boundary of PRDW was SSS of ~33 [29].
Since the temperature effect was usually the first-order controlling factor on pCO 2w , temperaturenormalized pCO 2w (pCO 2ws ) was used to explore the effect of SSS and biological productivity on pCO 2w . In this study, pCO 2ws and SSS showed strong correlation when SSS was less than ~33 ( Figure  3a), the correlations in spring and summer were positive (Figure 3a) and the linear correlation coefficients R 2 were > 0.75, that indicated that lower SSS (more PRDW) related to lower pCO 2w . And we found relationships between pCO 2ws and DO were negative in spring and summer, the linear correlation coefficients R 2 were 0.77 and 0.58, respectively, so lower SSS coupled with lower pCO 2ws and higher DO (Figure 3b). In fact, The PRDW input a large amount of nutrients into nearshore waters, biological productivity could be enhanced in the plume, photosynthesis produced O 2 and removed CO 2 in spring and summer [30]. The diluted waters of Changjiang River was also the main factor of pCO 2w , and leaded to the decrease of pCO 2w and the increase of O 2 in summer, the variations of pCO 2w and DO generally mirrored in the Changjiang River plume [31]. And intensive primary productivity in spring owing to Changjiang diluted waters also induced the sharp drawdown of pCO 2w in the southern Yellow Sea [32]. The good negative correlations between pCO 2w and chlorophyll-a in spring were also found in the North Sea, and the pCO 2w was modulated by continental inputs and photosynthetic activity [33]. Generally, diluted waters carried nutrients into coastal sea, and promoted phytoplankton propagation in favorable season and conditions, phytoplankton absorbed CO 2 and released O 2 , finally primary productivity induced low pCO 2ws and high O 2 .
On the contrary, pCO 2ws and SSS showed a negative relationship in winter and late autumn (Figure  3a), the linear correlation coefficients R 2 were 0.71 and 0.48, respectively, that indicated that lower IOP Conf. Series: Earth and Environmental Science 569 (2020) 012092 IOP Publishing doi:10.1088/1755-1315/569/1/012092 6 SSS (more PRDW) related to higher pCO 2w , the PRDW could have contributed DIC into coastal area and inner shelf, and the variation of SSS was slight and the DO content was stable in late autumn and winter (Figure 3a). Because autumn and winter was dry seasons in South China, the Pearl River runoff relative decreased in dry seasons than wet seasons, but DIC in the freshwater end-members were >2700 μmol·kg -1 in autumn and winter, and higher at ~1000 μmol·kg -1 in spring and summer [34]. So runoff DIC input had significant impact on increasing CO 2 concentration in the plume, and directly elevated pCO 2w owing to terrestrial sources [35]. The DO was nearly steady (Figure 3b) and primary productivity was low in late autumn and winter. Low primary productivity didn't uptake abundant carbon and draw down pCO 2ws which the runoff delivered into nearshore waters, so pCO 2ws was high when SSS was low in late autumn and winter. The primary productivity in winter was only 1/10 of that in summer in ECS due to low temperature and light intensity, the Changjiang diluted water enhanced the pCO 2w in plume [36]. The Liaonan nearshore waters of the Northern Yellow Sea in autumn was a source under influencing of runoff input [37].

Controlling factors of pCO 2w in the outer shelf and slope
SST was the lowest in late autumn and winter, middle in spring and highest in summer when the SSS was more than ~33, and the relationships between pCO 2w and SST were strongly positive in the four seasons (Figure 4a), the all linear correlation coefficients R 2 were > 0.50 and the highest value of R 2 was 0.93 in February 2011, SST was the major controlling factor in the variations of pCO 2w in the outer shelf and slope [9,16]. However, pCO 2w in late autumn was higher than that in spring when the SST was around 25°C, which was mainly because vertical mixing induced by monsoon forced deep water with high pCO 2w to intrude into surface water and enhanced pCO 2w in late autumn [38]. In addition, pCO 2w in spring was higher than that in summer when the SST was around 28 °C, the offshore waters absorbed CO 2 from atmosphere in late autumn and winter, and had a lot of accumulations on CO 2 because the carbonate system had buffering effect. CO 2 efflux just began when SST rose in spring, then accumulative effect on pCO 2w decreased gradually with CO 2 release during three month period, so pCO 2w in summer was lower than in spring at the same SST, the sea-air CO 2 fluxes was also a main factor of modulating pCO 2w at seasonal time scales [39]. Clear differences were noted in late autumn and winter when the SSS was less than ~33 (Figure 4b).
The relationship between pCO 2w and SST was negative, although it was weak. The low-temperature diluted waters input abundant inorganic carbon into nearshore waters in late autumn and winter, but the biological productivity was not active, that resulted in relatively high pCO 2w and low SST. The relationship between pCO 2w and SST in spring and summer was not clear.

Impact of weather and climate events on pCO 2w
Cold air was encountered during the winter and late autumn observation. The average of pCO 2w and sea-air CO 2 flux in transect A was (381±3) μatm and -1.30 mmol·m -2 ·d -1 in winter, which was (340±19) μatm and -8.20 mmol·m -2 ·d -1 in other transects, the pCO 2w and sea-air CO 2 flux in transect A after the occurrence of cold air was obviously higher than that in other transects (Figure 2c), and the wind speed during the transect A was (11.8±2.0) m·s -1 , and bigger than (8.20±2.2) m·s -1 during the other transects observation. The cold air forced deep water with high pCO 2w to intrude into surface waters, and the vertical mixing in the transect A abated the decrease of pCO 2w through the cooling of the surface water.
SSTs in late autumn dropped sharply owing to the strong invasion of cold air prior to the investigation, and SSTs were close to those in winter (Table 1, Figure 4b). Intense cooling decreased the average of pCO 2w to (348±19) μatm, and the ΔpCO 2 was ~ -40 μatm. The sea-air CO 2 flux was -8.49 mmol·m -2 ·d -1 , the NSCS was a strong CO 2 sink in late autumn. Moreover, the wind speed also increased by 10.1 m·s -1 at the arrival of the cold air (Table 2), because the sea-air CO 2 flux was proportional to the square of wind speed, therefore, cold air not only influenced the distribution of pCO 2w , but also affected the sea-air CO 2 fluxes. In addition, the tropical depression also contributed to the high pCO 2w and CO 2 effluxes in SCS by uplifting CO 2 -rich deeper water to the surface due to winds accompanied by it, and the CO 2 effluxes during three typhoons were as high as 60% of the whole year in the ECS [40].
The El Niño also affected on the distributions of pCO 2w by modulating global climate system. Rainfall in Guangdong province increased during May 2015 due to the influence of El Niño, a large amount of diluted water was discharged into nearshore waters, and the average of SSS in the coastal and inner shelf region decreased to 28.05 ± 6.18, which was lower than that in May 2011 (29.65 ± 2.58), meanwhile the average of SST was 2.10 C higher in May 2015 than May 2011. The abundant nutrients which brought by PRDW and the higher SST jointly promoted phytoplankton propagation in May 2015, which absorbed a mounts of CO 2 from seawater and released O 2 in the coastal and inner shelf region. The average of pCO 2 in the coastal and inner shelf region was (286 ± 95) μatm in May 2015, which was lower than (300 ± 42) μatm in May 2011. SST was still the major controlling factor of the pCO 2w in the outer shelf and slope region, SST rose by 1.96 C during El Niño, the average of the pCO 2w in offshore waters was (421 ± 9) μatm in May 2015 and (386 ± 13) μatm in May 2011, the difference was 35 μatm. The influences of El Niño made carbon sink increase in the coastal and inner shelf region and carbon source enhance in the outer shelf and slope region, the results of two processes offset mostly each other. The sea-air CO 2 flux was (-0.40 ± 5.39) mmol·m -2 ·d -1 during El Niño, the difference was not significant comparing to -0.67 mmol·m -2 ·d -1 in May 2011 [41].

Sea-air CO 2 flux estimation
The NSCS was CO 2 source with the efflux of 1.40 mmol·m -2 ·d -1 and released CO 2 in summer. This result was consistent with 0 mmol·m -2 ·d -1 to 1.9 mmol·m -2 ·d -1 in the basin and 0.3 mmol·m -2 ·d -1 to 5.5 mmol·m -2 ·d -1 in the southern shelf of the SCS [42], but higher than previous estimations of -1.49 mmol·m -2 ·d -1 to 0.17 mmol·m -2 ·d -1 in the NSCS [16]. However, the sea-air CO 2 flux in summer was -2.82 mmol·m -2 ·d -1 in the coastal and inner shelf and nearshore water was a net sink, the sea-air CO 2 flux was 2.00 mmol·m -2 ·d -1 in the outer shelf and slope and offshore waters was a source.
Generally, the NSCS acted as net CO 2 sink in spring, late autumn and winter, particularly in latter two seasons, in contrast, it was weak CO 2 source in summer.

Conclusions
Based on high resolution field surveys in the NSCS by underway continuous measurement system (GO8050) in four seasons, the average of pCO 2w was lower in late autumn and winter, and higher in summer, which in spring was between them. The NSCS absorbed a lot of atmospheric CO 2 in winter and late autumn, but released CO 2 into atmosphere in summer, both was in dynamic equilibrium in spring. The average of pCO 2w in the coastal and inner shelf was low and high in the outer shelf and slope in all seasons, the coastal and inner shelf was CO 2 sink in all seasons, the outer shelf and slope was only CO 2 source in summer.
The PRDW was critical controlling factor of pCO 2ws in the coastal and inner shelf. The PRDW promoted biological productivity in spring and summer, phytoplankton photosynthesis absorbed CO 2 and released O 2 , pCO 2ws in the coastal and inner shelf sharply dropped down. However, runoff inorganic carbon input also directly elevated pCO 2w in winter and late autumn due to low biological productivity. SST was the major controlling factor of pCO 2w in the outer shelf and slope in all four seasons, sea-air CO 2 exchange also played role in the seasonal scales of pCO 2w , furthermore, the effects of weather and climate events on the variations of pCO 2w and sea-air CO 2 fluxes in the short term were remarkable.
We explored the controlling factors of pCO 2w in the NSCS by field data analysis and discussion, but the results of pCO 2w by underway continuous measurement were instantaneous. The controlling processes of pCO 2w needed to be explored by time-series measurement, including controlling mechanisms of key marine dynamic processes and weather and climate events on pCO 2w , it could quantitatively calculate various controlling processes to pCO 2w variation. And what was more, the SCS was one of the world's major marginal sea systems, it was very important to have comprehensive understandings of carbon cycle in the SCS, especially migration and transformation of different forms of carbon.