Impact of equatorial and continental airflow on primary greenhouse gases in the northern South China Sea

Four-year ground-level measurements of the two primary greenhouse gases (carbon dioxide (CO2) and methane (CH4)) were conducted at Dongsha Island (DSI), situated in the northern South China Sea (SCS), from March 2010 to February 2014. Their mean mixing ratios are calculated to be 396.3 ± 5.4 ppm and 1863.6 ± 50.5 ppb, with an annual growth rate of +2.19 ± 0.5 ppm yr–1 and +4.70 ± 4.4 ppb yr–1 for CO2 and CH4, respectively, over the study period. Our results suggest that the Asian continental outflow driven by the winter northeast monsoon could have brought air pollutants into the northern SCS, as denoted by significantly elevated levels of 6.5 ppm for CO2 and 59.6 ppb for CH4, which are greater than the marine boundary layer references at Cape Kumukahi (KUM) in the tropical northern Pacific in January. By contrast, the summertime CH4 at DSI is shown to be lower than that at KUM by 19.7 ppb, whereas CO2 is shown to have no differences (<0.42 ppm in July) during the same period. Positive biases of the Greenhouse Gases Observing Satellite (GOSAT) L4B data against the surface measurements are estimated to be 2.4 ± 3.4 ppm for CO2 and 43.2 ± 36.8 ppb for CH4. The satellite products retrieved from the GOSAT showed the effects of anthropogenic emissions and vegetative sinks on land on a vertical profiling basis. The prevailing southeasterly winds originating from as far south as the equator or Southern Hemisphere pass through the lower troposphere in the northern SCS, forming a tunnel of relatively clean air masses as indicated by the low CH4 mixing ratios observed on the DSI in summer.


Introduction
Carbon dioxide (CO 2 ) and methane (CH 4 ) are the most abundant anthropogenic greenhouse gases (GHGs) in the atmosphere. They contribute approximately 93.6% of the well-mixed GHGs to global radiative forcing in 2011 (IPCC 2013). An increase in GHG concentrations can affect the Earth's radiation balance, and, thus, may cause changes in climate. It is generally recognized that the current increase in CO 2 levels is driven by fossil fuel combustion and carbon release from the biosphere through changes in land use, such as deforestation. Fossil fuel combustion accounts for about 92% of total global emissions of the atmospheric CO 2 , excluding those from forest fires and the use of wood fuel (Olivier et al 2014). The impact of the changes in land use on CO 2 from 1850 to 2000 was evaluated to be 12 to 35 ppm (Brovkin et al 2004, Matthews et al 2004. CH 4 has an absolute global warming potential (AGWP) of 34 W m -2 yr kg -1 , normalized to CO 2 with the inclusion of climatecarbon cycle feedback over a 100-year period (IPCC 2013). Recent studies have indicated that global levels of CH 4 have risen because of addition in emissions that were likely provoked by meteorological causes, such as anomalously high temperatures in the Arctic and substantial precipitation in the tropics (Rigby et al 2008, Dlugokencky et al 2009, Nisbet et al 2014. Consequently, the radiative forcing and AGWP of CH 4 may decrease as its concentration increases because of its overlapping absorption spectra and changing atmospheric lifetimes (Reisinger et al 2011). Soil is a source and a sink of CH 4 in the balance of methanogenesis and methanotrophy because microbial processes depend on soil texture, latitude, climate, meteorology, and land use (Dutaur andVerchot 2007, Chen et al 2014). Changes in land use systems, such as the conversion of tropical forests to pasturelands and the initialization of fertilizer use, could reduce CH 4 uptake (Verchot et al 2000, Palm et al 2002, Veldkamp et al 2008, which is often seen in developing countries. This 'land effect' can affect the local air quality as well as that in remote downwind areas, reflecting the concentrations of air pollutants emitted through anthropogenic activity or deforestation.
The South China Sea (SCS) is the largest marginal sea in the world, covering an area from 99°E to 121°E and from the equator to 23°N. It is situated between the Tibetan Plateau and the western Pacific warm pool, featuring warm and wet weather with individual monthly total rainfall of approximately 300-500 mm throughout the region (Chang et al 2005). It is subject to physical disturbances during various periods of time in a range from short-term events (e.g. typhoons) to seasonal changes (e.g. alternating Asian monsoons) as well as inter-annual oscillations (e.g. El Niño and Southern Oscillation) (Fu et al 1983, Chao et al 1996, Zhang et al 1997, D'Asaro et al 2014. The prevailing winds in the SCS typically blow from the northeast in winter and from the southwest in summer (Liu et al 2001, Metzger 2003, Chang et al 2005. Various countries, including China, the Philippines, Vietnam, Malaysia and Indonesia, etc, surround the SCS, and it receives massive quantities of air pollutants from the surrounding land. Because of the increasing urbanization and industrialization in East Asia over the past several decades, increasing emissions of sulfur and nitrogen compounds have affected air quality over the North Pacific (Wild and Akimoto 2001, Akimoto 2003, Ohara et al 2007, particularly when their transport is driven by Asian continental outflow. Mounting evidences of the long-range transport of air pollutants from East Asia to as far as North America have been reported in the literature (Jaffe et al 2003, Liang et al 2004, Cooper et al 2010, Ambrose et al 2011. During winter, the cold surges driven by frontal passages can also transport polluted air masses into the SCS (Hsu et al 2007, Ou-Yang et al 2013, Ashfold et al 2015. Biomass burning activities are pronounced repeatedly in the Indochina region of peninsular Southeast Asia every year in spring (Liu et Reid et al 2013, Ou-Yang et al 2014, Wai and Tanner 2014. Substantial fire activities associated with agriculture begins in the maritime Southeast Asia in July and end at the onset of the winter monsoon during October to November (Moron et al 2009, Reid et al 2012, Reid et al 2013. Moreover, more than half of the world's annual tonnage carried by merchant fleets passes through the Straits of Malacca, Sunda, and Lombok, with the majority of them continuing on to the SCS, making the SCS one of the most crucial trade routes in the world (US EIA 2013). In addition, the SCS has been suggested to be a moderate source of CO 2 emitted to the atmosphere through sea−air exchanges (Zhai et al 2005, Dai et al 2013. These events, regardless of whether they are caused by anthropogenic or natural activities, may contribute additional constituents to the atmosphere in this region. To improve the scientific understanding of the role of Asian continental outflow and its effects on air quality, numerous intensive field campaigns have been conducted over the western Pacific (e.g. PEM-West B (Hoell et al 1997), BIBLE (Kondo et al 2002), TRACE-P , and EAREX (Nakajima et al 2007). More recently, the Seven Southeast Asian Studies (7-SEAS) program was launched to investigate the effect of biomass burning in the Indochina region on aerosol-environmental systems, cloud chemistry, atmospheric radiation, and regional climates , Reid et al 2013, Tsay et al 2013. Based on respective approaches, these studies have demonstrated that the anthropogenic emissions from Southeast Asia could influence the composition of the Pacific troposphere. However, few studies on the atmospheric constituents of GHGs have focused on the SCS.
Here, we present 4-year observational data on primary GHGs observed at DSI from the perspectives of their seasonality and average levels representative of the northern SCS. Seasonal cycles of CO 2 and CH 4 at DSI are carefully assessed in comparison with the marine boundary layer (MBL) reference in the subtropical north Pacific. To evaluate the effects of respective inflows on the GHGs in this region, we also investigate the characteristics of the two GHGs associated with their source regions through the cluster analysis of backward trajectories. The retrieval products of satellite observations are used in this research to characterize the vertical and horizontal structure of the atmospheric GHGs in the SCS as well as to evaluate the land effects resulted from different influences. In addition, the biases of the satellite data against the surface measurements are estimated.  Figure 1 illustrates the geographical location of DSI and the averaged backward trajectories based on cluster analysis (see further discussion in section 3.2). DSI is located on the pathway of the alternating Asian monsoons, and features a subtropical maritime climate. It serves as a remote downwind site of encountering frontal passages driven by the winter monsoon that originates from East Asia. While, in summer, DSI serves as a downwind site of monitoring the long-range transport of air masses originating from Indonesia or nearby areas. Detailed information on DSI can be also found in previous studies , Ou-Yang et al 2013.

Flask air sampling
As a part of the Carbon Cycle Greenhouse Gases Network (CCGG), a pair of 2.5 L flask air samples was collected once per week in the morning at DSI by using a portable sampling unit and then shipped to National Oceanic and Atmospheric Administration/Earth System Research Laboratory/Global Monitoring Division (NOAA/ESRL/GMD) for analysis. The CO 2 and CH 4 were measured using nondispersive infrared spectroscopy and gas chromatography/flame ionization detection, respectively. Typical precisions of the analytical methods for CO 2 and CH 4 are <0.1 ppm and <1.5 ppb, respectively (Thoning et al 1995, Dlugokencky et al 2005. Detailed information about the analytical techniques for each GHG has been documented (Dlugokencky et al 1994, Thoning et al 1995).
Standard scales used are the WMO X2007 CO 2 and NOAA04 CH 4 mole fractions. Any flask sample pair with a difference of 0.5 ppm or more in the CO 2 mixing ratio was flagged and not used in this study. To gain a wider view of the effects of Asian continental outflow on the GHGs in this region, Cape Kumukahi (KUM) (19.52°N, 154.82°E, 3 m a.s.l.), one of the CCGG sites and located in the mid-Pacific at a similar latitude to DSI, is used to represent the reference baseline GHG levels in the MBL in the Pacific. The analytical methods and data treatment are similar at DSI and KUM 6 .

Trajectory analysis
Five-day (120 h) backward trajectories were calculated using the NOAA Air Resources Laboratory HYbrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT, Version 4.9) model (Draxler and Rolph 2014). The meteorological grid data used are the Global Data Assimilation System archives provided by the National Center for Environmental Prediction, with a 3-hourly and 1°× 1°latitude−longitude resolution. Trajectory deviation to approximately 20% of the traveled distance on average was suggested (Stohl 1998). In this study, all backward trajectories of every flask air sample collected at DSI are computed with an initial sampling inlet height of 8 m a.s.l. We also use cluster analysis to classify these 6 Detailed descriptions of the sampling and measurement methods used are provided on the NOAA/ESRL/GMD website at www.esrl. noaa.gov/gmd/outreach/behind_the_scenes/measurementlab. html. The discrete flask air sampling data can also be found on their website at www.esrl.noaa.gov/gmd/dv/data/. backward trajectories associated with the GHG measurements to assess their respective contributions at DSI.

Greenhouse gases observing satellite data
Ground-based measurements and satellite retrievals can be highly effective in investigating the seasonal and spatial characteristics of air pollutants and can provide new insights into the location and magnitude of sources and sinks on a regional scale. However, the accuracy of retrieved data must be assessed using direct measurements, particularly for sites located in the continental interior. Substantial effort has been expended to validate the column-averaged GHG retrievals against ground-based measurements from the Total Carbon Column Observing Network

Results and discussion
3.1. Seasonality of CO 2 and CH 4 at Dongsha Island Figure 2 illustrates the time-dependent variations of CO 2 and CH 4 measured at DSI and KUM from March 2010 to February 2014. The mean mixing ratios of CO 2 and CH 4 at DSI are estimated to be 396.3 ± 5.4 ppm and 1863.6 ± 50.5 ppb, respectively, over the study period. Note that these uncertainties are calculated based on the standard deviation (1σ) of detrended data obtained by subtracting the increasing trends. Greater amplitudes within the CO 2 and CH 4 seasonal cycles can be found at DSI than those at KUM. The mean rates of annual growth of CO 2 and CH 4 at DSI are calculated to be +2.19 ± 0.5 ppm yr -1 and +4.70 ± 4.4 ppb yr -1 , respectively, during 2011 to 2014. The annual growth rate is determined by the average of November-February months in a given year subtracting the same four-month average centered on the previous January 1. The uncertainties of the CO 2 and CH 4 annual growth rates are calculated based on the standard errors (2σ) in slope of an average linear rate of increase fit to the CO 2 and CH 4 data at DSI over the study period as a crude estimate of the uncertainty. Note that the uncertainty should only be considered a lower bound to the actual uncertainty because it does not include a component of uncertainty for undersampling by collecting samples only once per week. The smoothed data illustrated as the lines in figure 2 is calculated by filtering out the residuals with a shortterm cutoff of 80 days. The algorithm is adopted by the NOAA/ESRL/GMD (Thoning et al 1989). During the last decade, the baseline CO 2 mixing ratios have also increased in the surrounding countries of SCS, for instance, at a rate of 1.8 ppm yr -1 at Bukit Kototabang  (2013), the global increases in CO 2 and CH 4 levels are reported to be 11.7 ppm and 29 ppb, respectively, during 2005 to 2011. Despite the possible differences or gradients in concentrations occurring within the lower troposphere, the GOSAT L4B data at 975 hPa (approximately 300 m a.s.l.) show seasonal patterns in the two GHGs similar to those at DSI with mean positive biases of 2.4 ± 3.4 ppm for CO 2 and 43.2 ± 36.8 ppb for CH 4 ; by contrast, relative to the surface measurements at KUM, the satellite data show smaller biases, exhibiting mean differences of less than 0.1 ± 1.0 ppm and 3.6 ± 10.5 ppb for CO 2 and CH 4 , respectively. Nevertheless, a larger degree of the data divergence around the regression line can be seen in the elevated range of CO 2 and CH 4 at DSI (see supplementary materials), which is likely due to relatively inhomogeneous atmospheric composition while encountering polluted air masses from temporal and spatial perspectives. A significant deviating relationship in CH 4 is evident for the regression line of NOAA/ESRL/GMD flasks versus the GOSAT L4B products (see supplementary materials available at stacks.iop.org/ERL/10/065005/mmedia).
Distinct seasonal features with wintertime maxima and summertime minima are found for surface 7 These satellite data can be requested on the official GOSAT website at https://data.gosat.nies.go.jp/gateway/gateway/MenuPage/open. do. Details on the data products and retrieval methods are also provided on their website at www.gosat.nies.go.jp/eng/gosat/info. observational data and satellite retrievals, which showed similar patterns at other remote sites in the Northern Hemisphere (Conway et al 1994, Dlugokencky et al 1994. Mean levels of CO 2 and CH 4 at DSI and KUM during specific seasons are listed in table 1. In winter, the Asian continental high-pressure system dominates the transport of the outflow plume, leading to a direct effect on the levels of GHGs as well as the air quality in the neighboring downwind areas (Ou-Yang et al 2013). Due to the influences by maritime southwest monsoon flows in summer, DSI experiences low and steady background levels of GHGs in the northern SCS. Here, we adopt the GHG measurements obtained at KUM as the MBL reference at approximately 20°N in the Pacific. Comparing the ground-level observational results at KUM shows that the excessive amounts of GHGs from the combustion of fossil fuels and burning of biomass permeate the northern SCS Basin in winter and spring when the air is leaving the Asian continent, indicating significant increases in the winter maxima of CO 2 and CH 4 at DSI (figure 2). In January, the differences in the GHG concentrations between DSI and KUM are estimated to be 6.5 ppm for CO 2 and 59.6 ppb for CH 4 . The cold fronts caused by the Siberian high pressure system in winter move southeastward, thus driving the Asian pollution outflow into the SCS (Zhang et al 1997, Liu et al 2003, Wang et al 2003. In summer, vegetation growth on land largely eliminates the CO 2 signal at DSI, which is as clear and stable as that measured at KUM. The CH 4 levels at DSI during the midsummer southwest monsoon period are lower than those at KUM by 19.7 ppb ( figure 2(b)), whereas CO 2 is shown to have no differences (<0.42 ppm in July) during the same period. In addition, the amplitude of the annual cycle of monthly averaged CH 4 at DSI is estimated to be 108.3 ppb, which is approximately three times greater than that of KUM at 36.1 ppb, indicating that surrounding lands exert a significant effect.

Source region characterization
Cluster analysis of backward trajectories is a common method for evaluating the contribution of air pollutants from respective source regions with respect to measurements at receptor sites. For the study period, 171 trajectories are computed and grouped into six clusters (figure 1). Each average trajectory represents a distinct route: C1 (western Pacific), C2 (east China), C3 (north China), C4 (local), C5 (SCS), and C6 (subsidence) (see supplementary materials for all trajectories in each cluster). The statistical frequencies for each cluster according to month are estimated as illustrated in figure 3, showing a clear seasonality of changes in air masses arriving at DSI. Our cluster analysis of the backward trajectories is similar to that of previous studies conducted at Hok Tsui, Hong Kong (Wang et al 2009, Ding et al 2013. However, no trajectory analyzed in this study passes directly through southern China. Air masses originating from China (C2 and C3) dominate during fall to late spring, accounting for 52.0% of the total trajectories. These air masses travel along the western Pacific coastline, a heavily polluted area in East Asia, showing relatively high levels of CO 2 and CH 4 . The results concur with previous studies which have indicated that episodes of enhanced CO 2 or other air pollutants over the SCS are driven by the Asian outflow traveling with the winter monsoon (Vay et al 2003, Ou-Yang et al 2013. C2 and C3 are distinguished by their heights and distances, representing various transport speeds and residue times of air masses. However, no significant differences in the levels of GHGs are found between them (table 2). C4 represents aged air masses that are transported from the Asian continent and mixed with those from the Pacific, showing medium levels of GHGs (table 2). The C6 air masses contribute only approximately 3.5% of all trajectories. They are conveyed along with the westerly winds in the free troposphere and are rapidly dragged down to the surface by the downward branch of the transient local East-West cell/circulation, which was induced by the subsidence of a cold surge anticyclone (Yen et al 2013). These quick-moving air parcels merge with the southwesterly winds at the end of their transport approximately 1 day before arriving at DSI, exhibiting the highest level of GHGs among all groups (table 2). The C1 group consists of the local air in the northern SCS and is primarily present during the transition period between spring and summer when the pathway of air masses rotates clockwise. In summer, the marine air primarily originates from the SCS (C5), accounting for 56.3% of all trajectories in July. Several crossequatorial airflows are present from Indonesia to the SCS (see supplementary materials). The C5 air masses exhibit the lowest levels of ΔCO 2 (−2.3 ± 3.0 ppm) and ΔCH 4 (−20.9 ± 30.2 ppb). Figure 4 shows a selected longitude−latitude slice of monthly average CO 2 and CH 4 mixing ratios in the GOSAT L4B data at latitude of 20°N in the Pacific atmosphere in winter (January 2011) and summer (July 2010). Elevated CO 2 levels are found in the surface layer (below 850 hPa) between 100°E and 120°E (figure 4(a), top panel), indicating that a concise air flow penetrates this region from the east edge of East Asia. The highly polluted continental air masses traveling with the winter monsoon spread as far as 150°E at a latitude of 20°N, as shown in the CO 2 and CH 4 cross sections. However, in contrast to CO 2 , a relatively high level of CH 4 is observed on the east coast of the Bay of Bengal and in Myanmar at this latitude in winter (figure 4(a), bottom panel) and summer ( figure 4(b), bottom panel). Stable and steady characteristics are observed for the two GHGs in the eastern part of the Pacific atmosphere (120°W-150°W) at the same time. No significant differences in CO 2 are evident in the summer between DSI and KUM ( figure 2(a)). The inconsistency in the levels of CO 2 in the slice over the entire Pacific is less than 3 ppm, which is close to the bias of column-averaged dry-air mixing ratios (1.48 ± 2.09 ppm) reported by Yoshida et al (2013). However, the land effect might partially remove atmospheric CO 2 in the Indochina region in summer ( figure 4(b), top panel). The CH 4 vertical profiles show a 'window' of relatively low concentrations below 700 hPa ( figure 4(b), bottom panel), implying that there is a tunnel in the northern SCS for the clean air masses from the lower latitudes or Southern Hemisphere to pass through. This phenomenon could be resulted from the migration of intertropical convergence zone occurred during summer (Lawrence 2004). A steep gradient in CH 4 mixing ratios from land to sea is observed near the surface at approximately 105°E in summer, forming a wall cut by the southwesterly inflows. This window characteristic is not clear for CO 2 because of the relatively higher CO 2 level in the air masses from the lower latitudes in summer. In addition, low spatial variability of GHGs above 700 hPa can be seen over the SCS, indicating minimal seasonal variations and unclear influences from land (see supplementary material available at available at stacks.iop.org/ERL/10/065005/mmedia). Figure 5 shows a time series of daily average vertical profiles of CO 2 and CH 4 mixing ratios at DSI extracted from the GOSAT L4B data from March 2010 to February 2011. Although the discrepancies between the L4B products and the surface measurements are negligible (<0.6% for CO 2 and <2.3% for CH 4 ), the GOSAT satellite has a 3-day repeat global cycle (Yokota et al 2009), which may induce several minor errors in the results within 1-2 days. As illustrated in figure 5, enhanced GHG levels are found within the lower troposphere (below approximately 700 hPa geopotential height) from early winter to late spring, particularly for the layers near the surface (below approximately 900 hPa), which is consistent with the results of the aforementioned cluster analysis of backward trajectories in association with surface measurements. During summer, no substantial concentration gradients in CO 2 are found in the air column within the entire troposphere. The 'window' opens to sweep the primary air pollutants, such as CH 4 , in the northern SCS from July to August, showing that the summertime CH 4 mixing ratio measured at DSI is lower compared with that at KUM in the mid-Pacific.

Conclusion
In this paper, we first present the 4-year ground-level measurements of primary GHGs (i.e. CO 2 and CH 4 ) at DSI as a representative site for monitoring background air quality in the northern SCS. Increased mixing ratios for the two compounds are observed at DSI and are identified to be caused by the Asian continental outflow. The enhanced levels are calculated to be 6.5 ppm and 59.6 ppb for CO 2 and CH 4 , respectively, in January. No significant difference (<0.42 ppm in July) in summertime CO 2 levels between DSI and KUM (which represents the MBL reference at approximately 20°N in the Pacific) is evident. However, the summertime CH 4 levels at DSI are relatively low and differ from those at KUM by 19.7 ppb in July, likely because of the inflows of maritime air masses originating at low latitudes of the SCS. By integrating the satellite retrieval products of the GOSAT, the land effects caused by respective sources are shown in association with backward trajectory clustering. The deviations of GOSAT L4B data from the surface measurements are assessed, and positive biases of 2.4 ± 3.4 ppm for CO 2 and 43.2 ± 36.8 ppb for CH 4 are found at DSI, showing greater discrepancies in the continental interior than those at KUM in the mid-Pacific. The results provided in this study may strengthen the understanding of how this relatively unpolluted region receives Asian continental outflow from higher latitudes and purged with maritime inflows from lower latitudes in respective seasons.