The Yellow Sea is a semi-enclosed marginal sea, located on the western part of the Pacific Ocean, adjacent to China to the north and west and the Korean Peninsula to the east (Zhang and Chu, 2018; Wang and Zhai, 2021). It is a main pathway of air mass transport between the Asian continent and Pacific Ocean and can be divided into two basins: the northern Yellow Sea (NYS) and the SYS (Lyu et al., 2021). The SYS covers an area of about 10.8×104 km2, with an average depth of 44 m, and is strongly influenced by the EAM system (Zou et al., 2018). As shown in Fig. 1, to study the distributions of atmospheric CO2 and CH4 mixing ratios and their regulation mechanisms, two campaigns were conducted from 2 to 8 November 2012 and from 22 to 29 June 2013, respectively, the typical periods of the EAM (including the summer monsoon and winter monsoon). In order to ensure the comparability of observations, parallel observed CO2 and CH4 data from the three land (island)-based stations (LAN: Lin'an station; JGS: Jeju Gosan station; TAP: Tae-ahn Peninsula station) located in the vicinity are presented and studied in this study.

As shown in Fig. 2a, during the field surveys, the air inlet was fixed at the highest point of the bow, about 10 m a.s.l. (above sea level), and near the meteorological sensors to avoid anthropogenic contamination (Zang et al., 2017). Atmospheric CO2 and CH4 mixing ratios were measured by using a self-assembled Picarro system (G2301, Picarro Inc., USA). The Picarro analyser, which can acquire one measurement every 5 s and correct the measurements influenced by water vapour (Rella et al., 2013), has been proven to be excellent for measuring CO2 and CH4 with high precision and accuracy (Crosson, 2008; Fang et al., 2013).

As shown in Fig. 2b, ambient air was pumped via the dedicated tube by an external vacuum pump and passed through a membrane filter (1.0 µm, Whatman Inc., USA), a drying tube filled with magnesium perchlorate [Mg(ClO4)2] and another filter, respectively, to remove particles and water vapour. Then, it was regulated by valve sequence setting with a dry and clean air sample as well as the standard gases flowed into the CRDS analyser through an eight-port, multi-position valve (Valco Instruments Co. Inc. USA) with a flow rate of 200 mL min-1 controlled by a mass flow controller (Beijing Seven-star electronics Co. LTD. China). Before and after each campaign, the CRDS analyser was calibrated to guarantee its normal operation status. During field surveys, three standard gases were automatically measured in sequence each day, which was regulated by the CRDS analyser. Linear functions were yielded based on measurement results and standard values of three standard gases, i.e. 254.53 (±0.06) ppm, 365.14 (±0.06) ppm and 569.99 (±0.08) ppm for CO2 and 1601.0 (±0.8) ppb, 1925.5 (±0.8) ppb and 2317.7 (±0.5) ppb for CH4, respectively, which were used to calibrate the observed data. The standard gases used were propagated from the WMO primary standards (World Meteorological Organization, WMO, Global Atmospheric Watch, GAW, 2004 scale for CH4, 2007 scale for CO2) to guarantee the consistency, traceability and international comparability of observed data (Dlugokencky et al., 2005).

Both of the campaigns were conducted by a ship named RV Dongfanghong II, which was designed for research in multiple disciplines in marine environments with a ship-based atmospheric science lab. Meteorological data, including time, latitude, longitude, cruising speed and direction, wind speed, wind direction, relative humidity, air pressure, and temperature, were observed by the meteorological sensors (RM Young, USA) with a resolution of 10 s and were used to filter and flag the observed CO2 and CH4 mixing ratios and verify simulated wind fields.

HYSPLIT (Hybrid Single Particle Lagrangian Integrated Trajectory model) was developed by the National Oceanic and Atmospheric Administration's Air Resources Laboratory (NOAA-ARL) and the Bureau of Meteorology of Australia and can simulate the air mass transportation combined with the National Centers for Environmental Prediction (NCEP) reanalysis data. The principle of simulating the air mass transportation path is as follows: assuming that particles in the air are floating in the wind, their moving trajectory is the integral of their position vectors in time and space (Zhang et al., 2011; Xia et al., 2018). Backward trajectory analysis uses the mixed single-particle Lagrangian integral transport and diffusion model to calculate the transport route of air particles, analyse the influence of air mass transportation on the spatial and temporal distribution of atmospheric components in the observation area by tracking the transport path, and infer their potential sources. The main parameters required to calculate the backward trajectory are the altitude, latitude and longitude of the starting point. Generally, the calculation is carried for 72 h (Zhan et al., 2009; Zhang et al., 2017, 2019).

Generally, CO2 and CH4 mixing ratios decrease with increasing altitude and distance away from continent and decreasing latitude (Matsueda et al., 1996; Bartlett et al., 2003; Zang et al., 2017). Spatiotemporal distributions of atmospheric CO2 and CH4 mixing ratios in shelf seas suggest not only natural characteristics, but also multiple anthropogenic processes, such as marine oil and gas exploration (Nara et al., 2014; Zang et al., 2020), land–sea air mass transportation (Kong et al., 2010; Liu et al., 2018), and malfunction of observation instruments (Zang et al., 2017).

During the two field surveys, atmospheric CO2 mixing ratios ranged from 392.75 to 688.10 ppm in November 2012 (Fig. 3a and b) and ranged from 389.28 to 967.60 ppm in June 2013 (Fig. 3c and d), respectively. Atmospheric CH4 mixing ratios ranged from 1870.6 to 1986.0 ppb in November 2012 (Fig. 4a and b) and ranged from 1820.8 to 2179.0 ppb in June 2013 (Fig. 4c and d), respectively. Atmospheric CO2 and CH4 mixing ratios were comparable with the historical observation results of the Northern Hemisphere (Matsueda et al., 1996; Zang et al., 2017; Liu et al., 2018). Abnormally high observation values were attributed to exhaust gases of ships or anthropogenic interference of analysers.

Observed wind data were averaged to hourly data for subsequent analysis. As shown in Fig. 5a, during the survey of November 2012, hourly mean wind speed ranged from 0.05 to 20.46 m s-1 with an average value of 8.09 (±4.17) m s-1. The dominant wind direction was from the north and north-east, indicating that the air masses flowed from the Asian continent to the Pacific Ocean. As shown in Fig. 5c, during the survey of June 2013, hourly mean wind speed ranged from 0.08 to 9.42 m s-1 with an average value of 4.72 (±1.79) m s-1. Conversely, the predominant wind direction turned into south or south-east, which promoted air masses flowing from the Pacific Ocean to the Asian continent. In addition, the observed dominant wind directions (Fig. 5a and c) were consistent with the simulated wind fields (Fig. 5b and d), suggesting typical features of the winter and summer monsoon, which were ideal cases to study the effects of land–sea air mass transportation on the spatiotemporal variations in CO2 and CH4 mixing ratios in the MBL (marine boundary layer) of the SYS.

Although some empirically based data processes have been reported (Zang et al., 2017; Liu et al., 2018), a specific data-filtering approach for continuous shipborne observation needs to be optimised and established to distinguish impacts of multiple source–sink processes on observed shipborne atmospheric CO2 and CH4 mixing ratios along the cruise tracks, especially in the shelf seas.

Firstly, observed atmospheric CO2 and CH4 mixing ratios along the cruise tracks in November 2012 and June 2013 were calibrated by a linear function, averaged every 1 min and named “raw data” for the subsequent process.

Secondly, according to the voyage record, the abnormal values caused by malfunction of the instrument and impacted by manually refilling the drying tube were flagged (Zang et al., 2017).

Thirdly, when the ship stopped for oceanography investigation at discrete stations or cruised downwind with speeds lower than the wind speed, observed atmospheric CO2 and CH4 mixing ratios might be impacted by the ship's exhaust gas and human activities (Zang et al., 2017; Liu et al., 2018). Previous studies empirically considered 3 kn as the criterion to flag data influenced by the ship's exhaust gas and human activities (Zang et al., 2017; Liu et al., 2018).

In this study, two station measurements are taken as examples, as shown in Fig. 6a and b; when the ship speed slowed down from a normal cruising speed of 11 kn to less than 3 kn, the observed CO2 mixing ratio varied from a smooth pattern with a SD (standard deviation) value less than 0.10 ppm to an intensive fluctuation pattern with a SD value greater than 1.20 ppm due to influences of ship emissions and human activities. According to the quality control criteria of CO2 (±0.10 ppm), which was recommended by the WMO GAW (WMO, 2007), 3 kn was recognised as the threshold. Results showed that 15.5 % and 21.9 % of total observed data in November 2012 and June 2013, respectively, were flagged in this step.

Finally, the Pauta criterion (“3σ” method), a widely used data quality control approach in atmospheric greenhouse gas observation (Zhang et al., 2007; Fang et al., 2015; Zang et al., 2017), was introduced to filter and flag the non-background-measurement results. To optimise this process, observation data covered periods of 0.5, 1, 2 and 4 h as calculated, respectively. Any deviation between observed results and average values lying outside ±3 SD was considered to be non-background data and should be flagged. This procedure was repeated until no outliers were identified (Zhang et al., 2007). The results showed that data calculated hourly were optimal because not only could dispersed values be flagged, but the data could also be kept smooth.

As shown in Fig. 7, based on the optimised approach, observed data could be filtered and flagged. The remaining data accounted for 79.5 % and 75.7 % of raw data in November 2012 and June 2013, respectively, and were considered to be background and used for further analysis.

Observed mean CO2 mixing ratios were 403.94 (±13.77) ppm and 395.90 (±3.53) ppm in November 2012 and June 2013, respectively, which were slightly lower than previous studies' mean values of 405 and 410 ppm in the SYS and ECS in March 2013 and March 2017, respectively (Zang et al., 2017; Liu et al., 2018). Moreover, the observed mean atmospheric CO2 mixing ratio was almost equal to results observed at the TAP (401.37 ppm) and JGS (403.77 ppm) stations, but approximately 9 ppm higher than the MBL CO2 reference (394.41 to 394.78 ppm in the latitude zone of 30 to 37∘ N) (https://gml.noaa.gov/ccgg/mbl/data.php, last access: 10 October 2022) in November 2012 and almost equal to results observed at the LAN (396.43 ppm) and JGS (398.10 ppm) stations and the MBL CO2 reference (397.38 to 397.92 ppm in the latitude zone of 30 to 37∘ N) in June 2013.

Observed mean CH4 mixing ratios were 1924.8 (±27.8) ppb and 1918.0 (±25.7) ppb in November 2012 and June 2013, respectively, which were slightly higher than historical data of 1915.5 ppb in the SYS in March 2013 (Zang et al., 2017) and higher than the MBL CH4 references of November 2012 (1869.5 to 1880.3 ppb) and June 2013 (1835.3 to 1846.6 ppb).

Air–sea exchange is a dynamic process in which CO2 and CH4 molecules diffuse via the interface of surface seawater and overlying atmosphere. Sources and sinks of atmospheric CO2 and CH4 mean they were emitted from or absorbed by seawater. In fact, the magnitude of air–sea CO2 and CH4 exchange varied dramatically in spatial and temporal scale in coastal shallow seas (Yang et al., 2016; Gao et al., 2019). Generally, CO2 and CH4 emitted from the seawater into the air were difficult to trace by atmospheric measurements because they could dilute sharply (Schmale et al., 2005; Kourtidis et al., 2006; Zhai et al., 2014); only shallow seep areas and coastal regions could influence mixing ratios of local atmospheric CO2 and CH4 directly and be measured (Leifer et al., 2006; Luo et al., 2015). Even though the dissolved CO2 and CH4 were not observed in our field surveys, the published data showed that sea-to-air CO2 fluxes were 6.0 (±8.8) mmol m-2 d-1 in November 2012 and 2.6 (±4.3) mmol m-2 d-1 in June 2011 (Wang and Zhai, 2021), and sea-to-air CH4 fluxes were 6.4 µmolm-2d-1 in November 2002 and 15.7 µmolm-2d-1 in June 2006 (Zhang et al., 2008), respectively, in the SYS.

To estimate the effects of air–sea exchange on mixing ratios of atmospheric CO2 and CH4, we used a simple method described by Kourtidis et al. (2006) and optimised by Zang et al. (2020): assuming a box located above the survey area with a ceiling of 10 m, corresponding to the height of the air inlet in our field surveys. The contents of atmospheric CO2 and CH4 were only impacted by air–sea exchange. When CO2 and CH4 were vented into or absorbed from the box, their mixing ratios would increase or decrease homogeneously, caused by the mean calculated results of sea-to-air CO2 and CH4 fluxes.

Generally, coastal shallow seas are sources of atmospheric CH4, accounting for approximately 75 % of global ocean emissions (Bange et al., 1994). However, according to the calculation formula given by Zang et al. (2020), a sea-to-air CH4 flux of 50.8 µmolm-2d-1 could result in an increase of 2 ppb in the atmospheric CH4 mixing ratio in the MBL. Thus, the impacts of the reported mean sea-to-air CH4 fluxes (6.4 and 15.7 µmolm-2d-1 in November 2002 and June 2006) on the atmospheric CH4 would not exceed 1 ppb (Zhang et al., 2008). In addition, based on the same method, the impacts of the reported mean sea-to-air CO2 fluxes (Wang and Zhai, 2021) on the atmospheric CO2 mixing ratios were calculated and were no more than 14.1 ppb. Thus, it was reasonable to conclude that influences of air–sea exchange on the distribution of atmospheric CO2 and CH4 mixing ratios were low or negligible compared to the observed variability in atmospheric CO2 and CH4 (Figs. 3 and 4).

The EAWM is closely related to atmospheric-compound transportation from the Asian continent to the western Pacific (Yu et al., 2014). Since two surveys were conducted in November 2012 and June 2013, when the typical winter and summer monsoon seasons were in their early phases, respectively (Lyu et al., 2021), the observation data could give us an ideal opportunity to study the impacts of land-to-sea air mass transportation on spatiotemporal distribution of atmospheric CO2 and CH4 mixing ratios over the SYS. Observed atmospheric CO2 and CH4 mixing ratios were higher in November 2012 (Fig. 8) than in July 2013 (Fig. 9). Except for the Section 1 (S1) and the right end of Section 2 (S2), the spatial distributions showed a decreasing trend with the increase in offshore distance. (Fig. 8).

In-situ-observed data demonstrated that the dominant wind directions were W–NW–NNW for Sections S2, S3, S4 and S5 in November 2012, suggesting that the air masses were transported from the Asian continent to the Pacific Ocean (Fig. 5). Generally, CO2 and CH4 mixing ratios were higher on the continent than those of the MBL (Zhang et al., 2007; Zang et al., 2017; Liu et al., 2018). Land-to-sea air mass transportation driven by the EAWM could result in the horizontal transmission of greenhouse gases. Due to the subsequent mixing and dilution, CO2 and CH4 mixing ratios would decline along the route of wind transport (Bartlett et al., 2003; Kourtidis et al., 2006; Liu et al., 2018). Meanwhile, the mixing ratios of atmospheric CO2 and CH4 were low and homogeneous in Section S1 and the right end of Section 2 because the dominant wind directions were ENE–SE–S, indicating that air masses were transported from the open Pacific Ocean with low CO2 and CH4 content (Matsueda et al., 1996; Bartlett et al., 2003; Zang et al., 2017).

Furthermore, back-trajectory analysis showed that almost all the transport tracks originated from the Asian continent in November 2012 and the South China Sea and the western Pacific Ocean in July 2013 (typical characteristics of the early-summer monsoon) (Fig. 10), which resulted in higher atmospheric CO2 and CH4 mixing ratios in November 2012 (Fig. 8) than in July 2013 (Fig. 9). Seasonal variations in atmospheric CO2 and CH4 mixing ratios were consistent with the variations in atmospheric CO2 mixing ratios in the western Pacific Ocean, where atmospheric-component distributions were dominated by maritime air masses from the Pacific Ocean and polluted air masses from the Asian continent (Matsueda et al., 1996; Zhang et al., 2007; Liu et al., 2018).

As shown in Fig. 11, atmospheric CO2 and CH4 mixing ratios observed in November 2012 showed the same fluctuating feature versus wind direction, indicating that their variations were dominated by the land-to-sea air mass transportation, which was in agreement with previous studies (Zhang et al., 2007; Zang et al., 2017, 2020; Liu et al., 2018).

Simulation studies of gas seeps in the Black Sea and the Nord Stream pipeline gas leaks in the Baltic Sea showed that the atmospheric CH4 mixing ratio could be enhanced by the upwind emission source at a distance of 5 to 30 km (Kourtidis et al., 2006; Jia et al., 2022). The NOAA's MBL CO2 and MBL CH4 references were 394.56 ppm and 1875.4 ppb, respectively, in the same latitude zone as the survey area in November 2012. ΔCO2 and ΔCH4 represented deviations between observed atmospheric CO2 and CH4 mixing ratios and MBL CO2 and MBL CH4 references. As shown in Fig. 12, we assumed that the effects of mixing and dilution during the transportation were linear (Kourtidis et al., 2006). The further away the observation site is from the continent, the lower the ΔCO2 and ΔCH4 values are in each survey section. According to the calculated slope values, the gradient would be gradual at 123.30, 123.50 and 123.40∘ E for Section S3, S4 and S5, respectively. Moreover, the offshore distances away from the continent could be calculated as approximately 27.0, 26.3 and 11.7 km, respectively, with a mean value of 21.7 km. Thus, spatial distributions of atmospheric CO2 and CH4 mixing ratios in the China shelf sea could be impacted remarkably by land-to-sea air mass transportation during the early phase of the EAWM.