Potential Source Regions and Transportation Pathways of Reactive Gases at a Regional Background Site in Northwestern China

Reactive gases (O3, CO, NO2, and SO2) were collected hourly at the Akedala regional background station in northwestern China during September 2017 to August 2018. Wind rose, cluster analysis, potential source contribution function (PSCF), and concentration-weighted trajectory (CWT) methods were adopted for identifying the transport pathways and potential source regions of these atmosphere components at Akedala. +e average O3, CO, NO2, and SO2 mixing ratios detected were 29.65± 11.44 ppb, 123.78± 73.35 ppb, 3.79± 0.98 ppb, and 4.59± 0.88 ppb during the observation period, and the statistical results of the monthly mean values revealed that there were differences during the highest pollution period, with O3 and COmainly peaking in February, with mixing ratios of 38.03± 7.10 ppb and 208.50± 106.50 ppb, respectively. Meanwhile, NO2 peaked in March (4.51± 0.54 ppb) and SO2 in January (5.70± 1.92 ppb). +e most obvious diurnal variation of CO and SO2 was observed in the winter, with maximum levels reaching between 13 : 00 and 14 : 00. +e diurnal variations of O3 exhibited low values during the night and maximum values in the afternoon (16 : 00–18 : 00). Diurnal variation was not significant in the case of NO2. Cluster analysis showed that six main paths affected the Akedala atmosphere. In turn, the PSCF and CWT analysis results indicated that the Akedala reactive gas was affected by both local and foreign sources. +e high PSCF value of the reactive gas potential source areas appeared in eastern Kazakhstan, northern Xinjiang, Western Mongolia, and Southern Russia. +e WCWT (weighted concentration-weight trajectory) values of CO and SO2 in winter were the highest, totaling 180–240 ppb and 5–6.5 ppb, respectively. +eWCWTvalue of O3 in the spring and summer was higher than that in the autumn and winter.+emain source area of O3 was about 32–36 ppb in the spring and summer, and the main source area of NO2 in the summer had a lowWCWTvalue of 3–3.5 ppb, whereas the NO2 WCWT value was concentrated at 4–4.5 ppb in the other seasons.


Introduction
Reactive gases are strong chemical reaction substances in the Earth's lower atmosphere, with lifetimes between a few hours and a few months [1]. Among these, ozone (O 3 ), sulfur dioxide (SO 2 ), carbon monoxide (CO), and nitrogen dioxide (NO 2 ) are important reactive gases that present dangers for eco-environments and human health [2,3], and can also catalyze atmospheric chemical reactions, which are an important source of highly reactive atoms and free radicals [4,5]. ey are also the preferred gaseous pollutants recommended by the World Meteorological Organization and China [6,7]. In addition to the effects of natural and manmade sources on atmospheric concentrations of SO 2 , NO 2 , O 3 , and CO, regional-and global-scale atmospheric transport also affect atmospheric concentrations of these species, which can lead to changes in the climate system. For these reasons, these are currently popular research topics in the fields of Earth and environmental science [8][9][10].
With the rapid economic development and the acceleration of industrialization, man-made emissions of reactive gases have sharply increased. However, growth contribution to different regions also differs, and reactive gas levels and potential sources vary greatly among geographical locations in Asia, as shown in Table 1. For China, specifically, many scholars have studied reactive gases in areas with highly polluting emissions, such as cities [11,12], road traffic [13,14], and ports [15,16]. ere are relatively few reports on the background concentration of reactive gases compared to highly polluting areas such as cities, and field research on atmospheric constituents of regional background sites can not only provide valuable information on the impact of human activities, but also help illuminate the scope of regional air pollution. In addition, the general population around the background station is small, and the man-made source and sink densities are scarce. Reactive gases in these environments are determined by natural processes, such as turbulent exchange and advective transport [17]. ese background sites are also recognized as climate-sensitive hotspots, and they offer a better perception of changes to the surrounding environment [18], so it is a good choice to use background stations to track the source of the reactive gas. Many methods are used to investigate the location or direction of emission sources relative to observatories, wind roses, or CBPF (conditional bivariate probability function); matching wind data (wind speed, wind direction) with air pollutants values is superior to identifying local pollution sources [19][20][21][22]. Mangia et al. [23] identified areas of air pollutants emitted from different sources by analyzing prevailing winds and measurements of NO 2 and SO 2 in Taranto, Italy, finding that higher values of NO 2 and SO 2 were observed when the station was downwind of the industrial site. Pollution wind roses indicated the two prevalent wind directions that influenced the dispersion of air pollutants (CO, NO 2 , CH 4 , and SO 2 ) in Eleme, Rivers State, Nigeria. e higher pollutant values were associated with northeast wind in the dry season and southwest wind in the rainy season [24]. Li et al. [25] studied regional transportation of air pollutants in Northeast China Plain. eir studies revealed that high O 3 mixing ratios occurred in presence of strong winds (>3 m/s) but were independent of wind direction at different cities of Northeast China. Lagrangian-integrated trajectory (HYSPLIT) based on numerical simulation has been widely used for evaluating longrange transported pollutants and shows excellent potential [26]; on the basis of this, cluster analyses have been conducted, along with potential source contribution factor analyses (PSCF), concentration-weighted trajectory (CWT) method, and other trajectory analysis methods [27,28]. Ou-Yang et al. [29] use backward trajectory software to reveal the source of air masses from alpine atmospheric background stations in central Taiwan, indicating that air masses from the westerly belt are predominant in spring and winter, leading to higher CO concentrations. Vellingiri et al. [30] calculated the air quality transmission path in Seoul, South Korea, by using the TrajStat trajectory statistical function of HYSPLIT, and found that the high mixing ratio of NO 2 (greater than 60 × 10 −9 ) emanated from Mongolia, Russia, and northern China. Kasparoglu et al. [31] employed cluster analysis to evaluate air quality in the Marmara region of Turkey; high O 3 concentrations often seemed to appear in the trajectory clusters of air masses, mainly from the northeast and northern directions, accounting for about 80% of the totals.
In this study, drawing on the field observation data of reactive gases (O 3 , SO 2 , CO, and NO 2 ) at the Akedala Atmospheric Background Station in Xinjiang, China, from September 2017 to August 2018, the change characteristics of reactive gases on multiple time scales are analyzed and wind rose, cluster analysis, potential source contribution factor analysis, and concentration-weighted analysis employed. e wind rose model was applied to identify possible directions of local pollution sources. e cluster analysis revealed the characteristics of the Akedala reactive gases across different seasons and the main transportation routes, whereas PSCF and CWT were primarily used to identify the potential pollution source areas of reactive gases in northwestern China and their contribution levels.

Site and Data.
e Akedala Regional Atmospheric Background Station (47°10′N, 87°58′E, 562 m) is part of the GAW program, which is managed by the China Meteorological Administration, and is located in western Fuhai County, in the Altay region of Xinjiang in northwestern China, as is shown in Figure 1. Alluvial flatlands surround the site for approximately 50 km, along with the Gobi valleys of the Irtysh and Ulungur Rivers. e terrain is wide and flat, with the tall Altai Mountains in the northeast, the Sawyer Mountains in the southwest, and Junggar Basin in the south. In terms of climate characteristics, the area features a typical continental temperate and semiarid climate, with northwesterly and easterly winds prevailing throughout the year, with hot and dry conditions in summer, cold and dry ones in the winter, and large temperature differences between the day and night. O 3 , NO 2 , CO, and SO 2 have been routinely measured at Akedala since 2009, offering basic information on the sources and sinks of reactive gases in the area. e monitoring instruments are installed in an airconditioned laboratory on the top floor of a building 10 m in height. e observation period spanned August 1st, 2017, to September 31st, 2018, totaling 365 days, and a valid monthly average was required to be based on at least 27 valid daily averages over one month (with at least 25 valid daily averages in February); valid daily averages of air pollutants at each site were required to be based on at least 20 valid hourly averages in one day. e seasons were divided as follows: autumn (September-November), winter (December-February), spring (March-May), and summer (June-August). e instruments used for measuring the concentrations of SO 2 , NO 2 , O 3 , and CO were for continuous automatic analysis, were produced by Australia Ecotech Inc., and corresponded to EC/ML 9850, 9841B, 9810B, and 9830B, respectively.
Strict quality control was implemented during the instrument observation period, such as zero/span checks and multipoint calibrations. Zero/span checks were used to regularly check the drift in the instrument response. Multipoint calibration established the response relationship between the standard gas concentration and analytical instrument. e correct observed value of the instrument in terms of the concentration was based on standard gas. In addition, the quality of the original collected observation data was further controlled, and the blank value and negative concentration data were eliminated, effectively guaranteeing the quality of the monitoring data. e meteorological data for the backward trajectory was obtained from the global Data Assimilation System (GDAS) provided by the  [45] Advances in Meteorology 3 National Center for Environmental Forecasting (NCEP) of the United States (interval: 1 h, horizontal resolution 0.5°× 0.5°), and the meteorological elements included temperature, air pressure, relative humidity, and vertical and horizontal wind speed.

Backward Trajectory Model and Clustering
Analysis. e HYSPLIT model is a professional model system jointly developed by the National Oceanic and Atmospheric Administration (NOAA) Air Resources Laboratory (ARL) to calculate and analyze the source, transmission, and diffusion trajectories of atmospheric pollutants. e model is capable of processing multiple meteorological element input fields, multiple physical processes, and functions of different types of emission sources [46]. In this study, the backward trajectory analysis was performed using the TrajStat software (http:// www.meteothinker.org/), which is an application developed for air mass trajectory visualization and statistical analysis. e TrajStat uses a geographical information system technique for data management and includes the Hybrid Single-Particle Lagrangian-Integrated Trajectory (HYSPLIT) model [28]. To determine the passing routes of air masses before they reach Akedala, the 48 h backward trajectories were calculated. e NCEP's GDAS half-degree outputs from NOAA's ARL were used to drive our backward trajectory calculation. e simulations were run at every hour. e end point was set to 500 m above sea level due to the friction and turbulence effects near the ground, and the choice of a low reaching height may lead to inaccuracies in the trajectory calculation [47], and the height of 500 m can reflect the regional fluidity characteristics of the airflow and can also reduce the influence of the surface friction.
is height layer can effectively represent a well-mixed convective boundary layer and is suitable for regional transportation investigations [48].
Cluster analysis is a multivariate statistical analysis technique that utilizes similarity or difference indicators to mathematically determine the relationship between samples and divide the trajectory data into different clusters according to this. Trajectory cluster analysis is carried out on the basis of the TrajStat software, which features two clustering methods, namely, angular distance and Euclidean distance. Euclidean distance is often used to define latitude and longitude positions as variables for the distance between two trajectories. Its main disadvantage is that if two backward trajectories have the same path of motion but differing speeds, they will be divided into two different categories. In this study, we utilize the angular distance algorithm to cluster the airflow trajectory to Akedala and calculate the reactive gas concentration corresponding to the airflow over different seasons. e main reason for using  Advances in Meteorology angular distance clustering was to distinguish the direction of the reactive gas arriving at Akedala.

Potential Source Contribution Function (PSCF).
PSCF, combining airflow trajectory and hourly pollutant concentrations, was measured at the receptor region and used in this study. is technique has frequently been used to identify the probable locations of emission sources that affect pollutant loadings in the study region [49]. As Akedala is a relatively clean background station in western China, the mean value of hourly data on reactive gases is taken as the threshold value [50]. In other words, when the mixing ratio of the reactive gas corresponding to the airflow trajectory through the grid exceeds the hourly mean value, the trajectory is considered a pollution trajectory, and the PSCF value is calculated by determining the number of pollution trajectories. e analytical field is divided into small equal grid cells (i, j). e PSCF value for a single grid cell is then calculated by counting each trajectory segment endpoint that terminates within that grid cell [51] as follows: where i and j denote the latitude and longitude, respectively; m ij is the number of back-trajectory endpoints at the grid cell (i, j) corresponding to the measured mixing ratios of reactive gases higher than a user-determined threshold criterion; and n ij is the total number of end points that fall at the i and j cell. e higher the PSCF value, the higher the possibility of the areas being potential sources of reactive gases at the receptor sites. In order to reduce the uncertainty in a grid cell with a small number of endpoints, an arbitrary weight function W ij , defined by (2), was multiplied into the PSCF value [52]: 0.70, 20 < n ij ≤ 80, 0.42, 10 < n ij ≤ 20, 0.05, n ij ≤ 10.

Concentration-Weighted Trajectory (CWT)
. PSCF can only reflect the proportion of each grid pollution trajectory, and it is somewhat difficult to separate moderate sources from strong ones for a PSCF method in which grid cells can have the same PSCF value when sample mixing ratios are either only slightly higher or much higher than the criterion. In order to overcome this limitation, this paper employs the concentration-weighted trajectory analysis method (CWT) to calculate the value weight of the airflow trajectory in the potential source area [27] and then quantitatively analyzes the degree of pollution for different trajectories and potential source areas. e CWT calculation method is shown as follows: where C ij is the mean weighted value in the (i, j) cell, n is the index of the trajectory, M represents the total number of trajectories, C n is the measured reactive gas mixing ratio observed at the arrival of trajectory n, and β ijn is the time spent in the grid cell (i, j) by trajectory n. A high C ij value indicated that the high potential emission sources in the grid cell can result in high polluted values at the receptor site. Like the PSCF method, the CWT method also employs the arbitrary weight function to eliminate grid cells with few endpoints.

Variation Characteristics of Reactive Gas Mixing Ratio in the Akedala Atmospheric Background Station.
As is shown in Figure 2, in general, the hourly average maximum mixing ratio of O 3 did not exceed 70 ppb, the overall average was 29.65 ± 11.44 ppb, and the median was 29.87 ppb. Compared to other background sites in China, the O 3 level was lower than that of Wariguan and Shangdianzi and close to the Jinsha background site [53,54]; the main reason for this is that Akedala and Jinsha are remote rural background sites, to which pollutant transport contributed equally, and Shangdianzi is located in an economically developed area of China, where industrial and vehicle emissions make an important contribution, whereas Wariguan is located on the Qinghai-Tibet plateau, which features high ozone levels associated with strong radiation. e highest hourly average of Akedala's CO concentration was below 700 ppb and was at its highest level during the winter, with an average of 123.78 ± 73.35 ppb and a median of 102.75 ppb. Compared to other local stations, it was lower than the CO concentration level of the Lin'an (372.5 ± 0.6 ppb) [55] and Shangdianzi background stations (159.4 ± 0.4 ppb) [53] and was similar to Shangri-La and Wariguan in western China [56]. e main reason for this could be that the atmospheric CO at the eastern stations is primarily affected by the air masses emitted by nearby megacities and the monsoons, reflecting the impacts of man-made pollution. Shangri-La and Wariguan are representative of the CO background level of the Qinghai-Tibet Plateau. e air mass is blocked by the plateau's topography, and the long-distance airflow is relatively clean. e average value of NO 2 was concentrated between 2 and 8 ppb and was 3.79 ± 0.98 ppb, whereas the maximum value was 11.13 ppb and the median value 3.74 ppb in Akedala, which is closer to the NO 2 level at Longfeng Mountain [57]. e latter represents the agricultural area of the Northeast Great Plain, and Akedala the agricultural plain area of northern Xinjiang. Both can reflect the background of NO 2 emissions from agricultural production, and the average value of the SO 2 was also concentrated between 2 and 8 ppb, whereas the maximum was more than 16 ppb. e average value was 4.59 ± 0.88 ppb and the median was 4.37 ppb in Akedala. e concentration of SO 2 was high in the winter, affected by heating period in Xinjiang, and was significantly lower than that at Lin'an station [58] and Shangdianzi [59] and higher than that at Wariguan and Jinsha [60]. e main reason for this might depend on the difference in fuel burning sulfur emissions.

Advances in Meteorology
As is shown in Figure 3, the monthly statistical results of the hourly average value of the reactive gases revealed that there were certain differences in the mixing ratio of each reactive gas from month to month. e mixing ratio of O 3 in the autumn season was relatively low, and the minimum ozone value of Akedala occurred in December (18.50 ppb), whereas the maximum was measured in February (38.03 ppb). A decreasing trend was apparent from September to November (autumn). Meanwhile, the O 3 mixing ratio in February (winter) showed an increasing trend, and March-May (spring) and June-August (summer) were at a higher level, which differed from those of other background stations in China, such as the O 3 peak in Shangdianzi and Lin'an in the summer, driven by the Asian summer monsoon [61]. However, the Akedala background station was influenced by the westerly circulation, and there was a relatively strong stratospheric ozone intrusion from high latitudes [62]. e high value region of monthly CO variation in Akedala is concentrated in winter (December-February), and the maximum value appeared in February, which was 208.49 ppb, and slightly earlier than that at the northern hemisphere site [63]. Due to heating, the contribution of CO emission sources in winter is higher than in the other seasons. e winter spans December to February and is the main period for coal-fired heating. e incomplete combustion of coal is an important source of CO. e electricity and thermal energy produced by Central Asian countries such as Kazakhstan are mainly based on coal, and the high-emission CO in winter was also transported to the background station through the northwest airflow [64]. In November, the CO mixing ratio showed a slight upward trend, and that in the spring (March-May) showed a slight downward trend, with the minimum value appearing in August, at 72.52 ppb, in Akedala. Compared with other background stations, this emerging seasonal maximum and minimum was close to those in Lin'an and Longfengshan [65] but tended to differ from Shangdianzi [66] and was affected by emissions as well as special weather conditions (wind direction, wind speed, air quality transport). e distribution of NO 2 mixing ratio showed a "fluctuation change" in autumn and winter. e NO 2 concentration level in autumn was the highest, and there could be large nitrogen oxide emissions produced by biomass combustion in Akedala, as agricultural crop residues are burned not only as household fuel, but also directly in the fields, especially during the autumn harvest season, which may result in serious local and regional pollution [67]. NO 2 mixing ratios exhibited a downward trend in the spring and summer from a monthly point of view, and the average NO 2 value was the highest, at 4.51 ppb, in March, and lowest, at 2.28 ppb, in August. Construction resumed in the spring, machinery and vehicle use began to increase, and road traffic and agricultural activities emitted relatively more NO 2 . In Kazakhstan, NO 2 emissions were fairly high in March, when coal-fired heating was still being supplied in power plants; NO 2 in the atmosphere has a long lifespan. Cold air from Siberia moves eastward and transports NO 2 southwards to the Akedala background station [68].
Compared with Akedala, Shangdianzi, Wariguan, and Longfengshan all showed high NO 2 levels during winter and low NO 2 in the summer [59]. is type of seasonal cycle can be explained as the higher heat supply leading to high NO 2 emissions in winter, and stronger removal through photochemical conversion and wet deposition in the summer, with seasonal changes at the height of the mixed layer. e monthly mixing ratio of SO 2 underwent little fluctuation. e SO 2 level in the winter is relatively high and is similar in other seasons, reaching its maximum in January, which totaled 5.70 ppb in Akedala. is is similar to that of Longfeng Mountain, which had higher levels of SO 2 mixing ratio in the cold seasons [65]. Due to the influence of atmospheric diffusion conditions and seasonal changes in different regions, the month with the highest mixing ratio of Akedala, compared with that of Lin'an, and the highest SO 2 mixing ratio at the Lin'an background station was December [69]. e reasons for the higher SO 2 level in the winter were that coal-fired heating activities were more frequent, the removal of SO 2 by chemical conversion and precipitation of quantity was minimal, and the winter has a lower boundary layer height, which is conducive to the accumulation of pollutants. Figure 4 shows the average daily changes of CO, O 3 , NO 2 , and SO 2 mixing ratios in Akedala. In terms of CO, the average mixing ratio was the highest in the winter and lowest in the summer. e average daily range across the four seasons (which refers to the difference between the maximum and minimum values) was distributed at 40.65, 130.86, 62.89, and 51.99 ppb. e difference value was the largest in the winter because the winter photochemical reaction was weak, due to the heating and the burning of fossil fuels and biomass combustion, with relatively stable changes in the spring, summer, and autumn. e peak value reached around 13 : 00 during the four seasons, and the peak value in winter was the largest, at 285.95 ppb. e peak value in the autumn, spring, and summer was 129.20 ppb, 122.09 ppb, and 100.06 ppb, respectively. e CO mixing ratio values of the four seasons all appeared at night, among which the minimum values of spring, summer, and autumn occurred at 03 : 00, and the minimum CO mixing ratio in winter at 01 : 00. For O 3 in Akedala, the mixing ratio in the spring and summer was significantly higher than that in the autumn and winter, and the average daily range during the four seasons of autumn, winter, spring, and summer was 16 e daily range of O 3 was the largest in the summer and the lowest in the winter. e reason for this was that solar radiation was strong in summer. e photochemical reaction was intense, the height of the mixed layer was high, the air within was well mixed and the abovementioned characteristics generated obvious differences during the day and night; the atmospheric conditions in winter contrasted with those in summer, the height of the mixed layer was low, the quiet and stable weather increased, and the air pollutants on the ground were not easily diffused, gathering on the surface and becoming coupled with low temperatures, solar radiation, 6 Advances in Meteorology and weak photochemical reactions in the winter [53].  Advances in Meteorology ratio gradually decreased, and the night level was the smallest, concentrated between 01 : 00 and 03 : 00 in Akedala. With respect to NO 2 , the NO 2 mixing ratio level in summer was significantly lower than in the other seasons in Akedala; the average daily range in the four seasons of autumn, winter, spring, and summer was 1.31, 1.09, 1.11, and 0.71 ppb, respectively; and the daily variation was generally stable, with the high value periods being concentrated at about 20 : 00, 19 : 00, 1 : 00, and 23 : 00 in autumn, winter, spring, and summer and the peaks being 4.93 ppb, 4.59 ppb, 4.82 ppb, and 3.37 ppb, respectively, indicating that both solar radiation and NO 2 consumption decreased. With respect to SO 2 , the winter diurnal changes were more significant, and the mixing ratio level was high. SO 2 is a primary pollutant and is especially affected by man-made and meteorological conditions. e atmospheric stratification temperature in the winter was prone to inversion, and SO 2 easily accumulated [70]. Similar to CO, SO 2 fluctuated smoothly before 11 : 00, showing a significant upward trend at 12 : 00-15 : 00 and reaching a peak at 15 : 00, 5.58 ppb, and then the average mixing ratio of SO 2 exhibited a downward trend in Akedala. e low mixing ratio of SO 2 at night could be due to the removal of the high humidity environment in the area at night during the winter. In the spring, summer, and autumn, the fluctuation is severe from 0 : 00 to 4:00 and reaches its maximum at night. e daily range of SO 2 in the four seasons of autumn, winter, spring, and summer was 0.39, 0.79, 0.31, and 0.50 ppb, respectively.

Identification of Local Sources.
Wind field can affect the diffusion and the local and transboundary transportation of pollutants [71]. In general, wind speed has a considerable effect on pollutant transportation and removal [72,73]. During environmental processes, the difference in wind direction reflects the path of pollutant transmission [74,75]. Figure 5 shows the near-surface winds, and their effect on mixing ratios of reactive gases was analyzed using the hourly observation of reactive gases and wind data from September 2017 to August 2018. Figures 5(a)

Advances in Meteorology
show that CO was highly influenced by wind direction in the winter. Winds from the southwest (SW), south-southwest (SSW), south (S), and south-southeast (SSE) corresponded to higher CO mixing ratios, which were above 400 ppb. e wind speeds in these sectors were close to 4-6 m/s, and their main sources might be heating and combustion activities in Fuhai County near the southwest of the background station and regional emissions from Karamay Oilfield [76,77]. e corresponding CO mixing was relatively high, at 150-250 ppb, in the SSE and west-northwest (WNW) wind directions. In the WNW direction, the CO high mixing ratio appeared in the wind speed range of 4-6 m/s, while in the SSE direction, the CO high mixing ratio corresponded to a high wind speed (greater than 8 m/s). In spring, the CO mixing ratios corresponding to each sector had a small difference. e region was industrially well developed and produced high ozone precursors [45], which corresponded to the SW, SSW, S, SSE, and southeast (SE) sectors with O 3 mixing ratio of more than 40 ppb. e high contribution areas of O 3 sporadically appeared in the SW, west (W), and WNW wind directions during the spring, which corresponded to the O 3 mixing ratio of 50-60 ppb. In the northwest (NW), SW, and northeast (NE) directions, the mixing ratios of O 3 had a significant increase in summer, which was above 50 ppb. e wind speed ranges corresponding to O 3 high values in the above three directions were 6-10, 8-14, and 4-6 m/s, respectively.
Figures 5(i)-5(l) show that NO 2 was not significantly affected by wind direction in spring, summer, and autumn. e mixing ratio of NO 2 corresponding to each wind direction was concentrated at 3-5 ppb. e S, SW, and W wind directions corresponded to a higher mixing ratio of NO 2 (5-7 ppb), and wind speed ranges from 4 to 6 m/s in winter, which might be mainly contributed by frequent traffic activities in the north of Tianshan Mountain [78,79].

Advances in Meteorology
Figures 5(m)-5(p) show that the wind directions of WNW, NW, and east-southeast (ESE) significantly uplifted the SO 2 mixing ratio in autumn, and the corresponding SO 2 value was 6-7 ppb. e wind speed ranges of the three high value areas were 8-10, 10-12, and 4-6 m/s, respectively. e high value areas of SO 2 appeared in the NW and S of the background station in winter. ese two wind directions had high SO 2 mixing ratios, concentrated at 7-13 ppb, and their corresponding wind speeds were 6-10 and 4-6 m/s, respectively. e SO 2 mixing ratios in the northwest might be derived from the contribution of long-range transport air masses, for example, in Kazakhstan. e high SO 2 mixing ratio in the south of the background station might came from the local transportation of Urumqi in the southern margin of the basin, where the demand and consumption rate of fossil fuels were high in winter, and a large amount of SO 2 was emitted by human activities [80]. e north (N), NE, and E directions had a slight lifting effect on the SO 2 mixing ratios in spring. e high value was concentrated at 6-8 ppb, corresponding to the wind speed range of 6-8 m/s, while those in other directions were concentrated at 4-5 ppb.
e effect of wind direction on SO 2 mixing ratio was not significant in summer, and the high value areas had a pointed shape with corresponding mixing ratios of 6-8 ppb.

Cluster Analyses.
Akedala is a low-altitude environmental background area in northwestern China and is more sensitive to regional environmental changes. e average value of reactive gases corresponding to various airflow trajectories had significant differences (see Figures 6 and 7  and Table 2). e trajectory clustering calculation was undertaken by adding the hourly average data of reactive gases to the backward trajectory and then dividing the seasonal airflow trajectories of Akedala into six categories in accordance with the total spatial variance method. Based on the results of the cluster analysis of the airflow trajectories in each season, the cluster statistics module was used to analyze the average value of reactive gases corresponding to different types of trajectory (Table 2) in order to quantitatively characterize the mixing ratio level characteristics of reactive gases under the influence of different types of flow trajectory.
In autumn, the frequency of various trajectories was as follows: 4 > 2 > 5 > 1 > 6 > 3 ( Figure 6). e airflow trajectories (2, 4, 5) from the northwest of Kazakhstan accounted for 74.54% of the total trajectories, which was the most significant airflow direction that affected the Akedala background station in autumn, and this type of airflow trajectory was primarily affected by the westerly circulation system [81]. Trajectory 2 derived from Balkhash Lake in Kazakhstan; passed through the northern border of Xinjiang, China; and then extended to the northeast, finally moving east from the Irtysh River Valley to Akedala. Trajectory 4 originated in northeastern Kazakhstan, passing through the Kazakh Uplands and Zaysan Lake and progressing downstream of the Irtysh River in Kazakhstan and finally to northern Xinjiang. Track 4 corresponded to the highest average value of O 3 , of 24.84 ± 10.70 ppb. ere are vast agricultural planting areas in eastern Kazakhstan and northern Xinjiang, which also coincide with the autumn harvest. e contribution of autumn straw burning in these areas cannot be ignored [82]. Trajectory 5 originated from the Tacheng area in Xinjiang and passed through the Tacheng air opening area and on to the Junggar Basin. When the pollutants in the Tacheng area were affected by the narrow tube and landslide effects, the airflow was easily accelerated.
e first type of trajectory mainly originated from the Gurbantunggut Desert in northern Xinjiang. e average CO mixing ratio corresponding to this type of airflow was the highest, at 105.13 ± 37.31 ppb. e desert is strewn with oilfield operation areas, and crude oil extraction processes such as seismic drilling and steam soaking would release toxic gases such as CO [83]. Trajectory 3 originated in Southern Russia, crossed the Altai Mountains, and entered Akedala from the southeast through Xinjiang. e sixth type of trajectory primarily originated in the Zhundong area of Xinjiang, where large coal fields were distributed, and the corresponding NO 2 and SO 2 of this type of airflow were the highest, at 4.42 ± 1.41 and 5.02 ± 1.22 ppb, respectively. ese two pollutants were affected by the emission of mining activities, mainly from mobile sources, fixed sources, and scattered sources. Mobile sources include vehicle transportation and excavating equipment, whereas fixed sources include large-scale operations such as drying, roasting, and smelting; dissipation sources mainly comprised plumes produced by open blasting [84].
In winter, the occurrence frequency of various kinds of tracks from high to low was as follows: 2 > 5 > 6 > 1 > 3 > 4. Mongolia's high pressure constitutes a weather system that significantly affects Xinjiang. e region was affected by the Xinjiang high-pressure ridge and East Asian trough, and cold air moved southward [85]. e second and fifth types of airflow trajectories from Kazakhstan in the winter accounted for the highest proportion, totaling 44.03%. Among these, trajectory 5 in the winter was similar to trajectory 4 in the autumn, which originated in the eastern Kazakh city of Oskemen. Nonferrous metal smelting in this area is relatively developed, and the production process could emit more gaseous pollutants [86]. e second type of airflow trajectory originated in the east of Balkhash Lake and Tacheng City, crossing the Saur Mountains, with the average value of O 3 , CO, and NO 2 corresponding to this type of airflow trajectory being the highest, at 31.24 ± 11.90, 251.72 ± 122.13, and 4.26 ± 1.11 ppb, respectively; the main polluted area in this trajectory was Tacheng City, which was mainly influenced by coal-fired heating in the winter. Trajectories 1 and 6 originated from the border areas of Mongolia, accounting for 35.46% of the total trajectories. e highest SO 2 average value corresponding to the airflow trajectory was 5.63 ± 2.00 ppb. Western Mongolia is dominated by agriculture and animal husbandry, and the family lifestyle is relatively undeveloped, which is likely to result in the emission of pollutants. Due to the increase in heating demand in winter, a large amount of coal and wood would be consumed. According to the statistics, each family in Mongolia used about 5.0 tons of coal and three cubic meters of wood per year for heating and cooking purposes [87]. Trajectory 3 passed through the Karamay Tuyere, which is located to the west of Baikouquan, in Urho. e unique topography and the effect of cold air crossing the mountains and depressions could easily cause strong winds and carry pollutants.
In the spring, the frequency of various trajectories from high to low was as follows: 2 > 5 > 4 > 3 > 6 > 1. is season was mainly affected by the eastward movement of cold air from Siberia to the south, and the airflow trajectory from Kazakhstan and northwest of the Russian border (2,4,5) accounted for the largest proportion, totaling 73.78%. Track 2 originated to the north of Balkhash Lake and then progressed along the northern border of Xinjiang. is type of airflow trajectory corresponded to the highest NO 2 mixing ratio, which was 4.25 ± 0.79 ppb. Fishing activities around Lake Balkhash are relatively developed and ship exhaust gas emissions cannot be ignored [88]. NO 2 would be emitted from the agriculture and animal husbandry developed in northern Xinjiang, including nitrogen fertilizer applications in the spring, accumulation of animal manure, and exhaust emissions from transportation equipment [89]. Trajectory 5 originated in Russia, passed through the Pavlodar region, and then entered Akedala from the Irtysh Valley in China.
is type of trajectory corresponded to the highest O 3 mixing ratio, which was 36.25 ± 7.59 ppb.
e Pavlodar region is a major industrial center; is engaged in the following industrial activities: bauxite mining, alumina and aluminum production, thermal power production, oil refining, and mechanical engineering; and might produce greater ozone pollutants [90]. Track 4 originated in the eastern Kazakhstan Region and then entered Akedala through Jeminay County and the Wulungu River. Eastern Kazakhstan is rich in minerals. e metallurgical industry is the pillar industry in the area, and industrial furnaces are used to add fuel. During smelting, a large amount of harmful waste gases such as CO is produced [91]. e first type of airflow trajectory originated in the west of Mongolia, crossing the Altai Mountains and moving northwestward to Akedala. e third type of airflow trajectory originated from the Alataw Pass. is airflow then easily moved eastward, passing through Karamay, and was transported to the background station. Trajectory 6 originated from the Tacheng area. e westerly airflow had a convergence effect and was transported to the southeast through the Karamay Oilfield. Oilfield extraction, refining, and processing would release a large amount of SO 2 [76]. e average SO 2 value corresponding to this trajectory was 4.44 ± 0.77 ppb.
In the summer, the frequency of occurrence of the various airflow trajectories was as follows: e airflow trajectory was relatively concentrated, and the northwest and westerly winds prevailed in the summer. e frequency occupied by 1, 4, and 5 was, at most, 74.05%. Both 1 and 4 originated in Advances in Meteorology eastern Kazakhstan. Trajectories 1 and 4 started at Zaysan Lake, passed through cities around the Irtysh Valley such as those in Habahe and Burqin counties, and entered the background station. Trajectory 5 started from Balkhash Lake, moved eastward to Zaysan Lake, and then entered Xinjiang. Trajectories 2 and 6 mainly originated from cities in northern Xinjiang. Trajectory 2 started in Tacheng City and entered the northern part of the Junggar Basin. e maximum average SO 2 value corresponding to track 2 was 4.70 ± 0.48 ppb. ere are many oil and gas fields in the northern area of the Junggar Basin. Drilling, production, and exhaust gas combustion are important sources of SO 2 [92]. Trajectory 6 started at the Alataw Pass, moved east to Karamay, and was then transported in a northeastern direction to Akedala. e average O 3 value corresponding to this type of airflow trajectory was the highest, at 36.23 ± 11.42 ppb. Karamay Oilfield is an important petrochemical center in western China, producing petrochemicals, and petroleum transportation vehicles would emit a large amount of ozone precursors (VOC and NOx), which is conducive to the generation of ozone in the area [93]. Track 3 originated in Russia's Altai borders.
After passing through the Oskemen, the capital of East Kazakhstan, and then entering the background station along the Irtysh River Valley, the average mixing ratio of CO and NO 2 corresponding to this type of airflow trajectory was the highest, at 93.67 ± 32.54 and 3.28 ± 0.75 ppb, respectively. Oskemen in Kazakhstan has one of the country's worst air qualities. It is an important railway hub and river port, as well as a nonferrous metal smelting industrial center. Its nitrogen oxide and carbon monoxide emissions are primarily affected by road and industrial sources [64].

PSCF and CWT Analyses.
Trajectory clustering analysis can be utilized to judge the main migration path of pollutants in the Akedala background station, but the relative contribution level of the potential source area around the reactive gas cannot be identified [32]. As is shown in Figure 8, using WPSCF to analyze the potential source distribution of reactive gas, red could relate to high mixing ratios of reactive gas, whereas blue reflects low mixing ratios in the source area. In order to better describe the distribution   In general, the potential source area of the reactive gas had a large spatial distribution, and the contribution level of reactive gas in the potential source area also had a certain difference across the different seasons. In terms of CO, the potential sources of CO were concentrated in northern Xinjiang, and its WPSCF values were between 0.4 and 0.7 in the autumn. e main potential sources of CO indicated four major strip regions in the winter. e first is the Turpan Basin, where oilfield operations are located [94]. e second largest area originated around Lake Balkhash, affected by the cold air, and pollutants flowed eastward through the tuyere [95]. e third area was the Siberian Plain in Russia, and the fourth was mainly located in Western Mongolia. Its WPSCF value was between 0.5 and 0.8. e main potential source of CO in the area is around the Junggar Basin, and this is affected by the westward airflow outside the country in spring. e WPSCF's high value areas appear in the eastern and southern regions of Kazakhstan. e potential source of CO primarily originated in the Altai border region of Russia and then entered Xinjiang through a large agricultural area in central East Kazakhstan in the summer [96]. e WPSCF value of the O 3 was concentrated in 0.4-0.6. For O 3 , the main potential source areas were certain parts of Kazakhstan in the autumn, such as the area around Balkhash Lake and the cities along the Irtysh River. e value was between 0.6 and 0.7. e eastern part of the basin also exhibited a moderately polluted distribution. In the winter, northern Hami and Karamay in Xinjiang were heavily polluted, and both places are in windy areas, where pollutants are easily transported by strong winds [97]. In general, there are three main strips of moderate O 3 pollution outside the country in autumn, which originated in the eastern part of the Aral Sea, the Novosibirsk region of Russia, and the western region of Mongolia. In the spring, the area of the medium potential source of O 3 abroad had increased and was widely distributed in East Kazakhstan, Pavlodar, Karaganda, and Almaty. In the summer, the source of O 3 was primarily the Altai Territory of Russia, passing through eastern Kazakhstan. e territory was mostly distributed among the surrounding cities of the Irtysh River Valley, and the WPSCF was between 0.5 and 0.6. For NO 2 , the moderately polluted areas were concentrated on the northern slopes of the Tianshan Mountains in autumn. is type of area is the most economically developed in northern Xinjiang and features a high degree of industrialization, and thus it makes an important contribution to regional NO 2 emissions [98]. In the winter, Zhundong, Oskemen (Kazakhstan), and Balkhash Lake (Kazakhstan) were heavily polluted. e potential source of NO 2 indicated a path conforming to the westerly wind. e NO 2 from eastern Kazakhstan is transported eastward through the northern vents of Xinjiang. In addition to the cities on the northern slopes of the Tianshan Mountains, there are moderately polluted areas in the territory in the spring. Heavy pollution sources arise in the northwest of East Kazakhstan and the Altai region in summer. e northwestern part of East Kazakhstan is an industrial agglomeration area; NO 2 emissions in this area could be affected by industrial emissions [91]. e WPSCF of the two regions was concentrated at 0.7-0.8, and the main area was moderately polluted in northern Xinjiang, with the WPSCF value being 0.4-0.5. As far as SO 2 is concerned, in autumn there was a moderately polluted source area in northeastern Kazakhstan and the eastern part of the Junggar Basin, whereas other sources in Kazakhstan were only slightly polluted. e main source of SO 2 came from external sources in the winter, such as Western Mongolia, eastern Kazakhstan, and Russian border regions.
e WPSCF contribution value of this type of region was above 0.5. In the spring, the potential source of SO 2 outside was concentrated in Almaty. Pollutants entered China through the northern vents of Xinjiang, and the source of heavy sulfur dioxide pollution appeared in the background station during the summer. e southern part of Akedala was primarily around the Gurbantunggut Desert area, which could be affected by local oil production emissions [99], and the northern part was mostly affected by the city of Altay. Contaminated areas abroad were mostly transported by cities around the Irtysh River, and their WPSCF values were between 0.4 and 0.6.
In summary, the WPSCF distribution of reactive gases in different seasons was more consistent with the spatial distribution of the probability of trajectory clusters in Section 3.2. Visibly, the source simulated by the PSCF analysis method contributions had greater credibility and could be used for determining potential source contributions.  Reactive gases were affected by both local emissions and external sources. Sources of pollution in Xinjiang, eastern Kazakhstan, Southern Russia, and Western Mongolia all made important contributions to Akedala. e PSCF analysis method can only reflect the contribution rate of potential source regions. In other words, the proportion of the pollution trajectory in each grid could not quantitatively reflect the reactive gas mixing ratio [49]. e WCWT was used to simulate the potential source region affecting the Akedala reactive gas mixing ratio value, the results of which are shown in Figure 9.
In terms of CO, the high WCWT areas were concentrated in the Junggar Basin and eastern Kazakhstan in the autumn and spring. e corresponding WCWT values were between 90 and 120 ppb. e CO mixing ratio in winter was higher than that during other seasons. e domestic areas were concentrated in northern Xinjiang, where the CO value was greater than 180 ppb. e CO contribution value in the eastern part of the Junggar Basin was higher, at over 210 ppb. e area with a higher WCWT abroad appeared in Lake Balkhash, Kazakhstan, and was greater than 300 ppb. e contribution of CO in the Irtysh River Basin is equivalent to that of Western Mongolia, which was 180-240 ppb. In the summer, the contribution of the CO potential source area was the smallest, and the WCWT value was concentrated at 60-90 ppb. For O 3 , the main source area contribution of O 3 was 32-36 ppb in the spring and summer, and the distribution area of high value areas in the summer was smaller than in the spring and was concentrated at 24-28 ppb in the autumn and winter. For NO 2 , the WCWT value of the main source area of NO 2 was relatively low in the summer, ranging from 3 to 3.5 ppb, and the NO 2 WCWT value of other seasons was between 4 and 4.5 ppb. For SO 2 , the contribution value of the source area in winter was relatively high. e WCWT values in other cities in eastern Kazakhstan and the northern slopes of the Tianshan Mountains were in the range of 5-6.5 ppb. In   addition, Kobdo in Western Mongolia featured the highest SO 2 , and its contribution was greater than 6.5 ppb. e contribution of WCWT in the SO 2 source area in other seasons was 4-5 ppb.

Discussion
A large number of studies have combined various methods such as trajectory statistics, PSCF, and CWT to extensively investigate the potential source areas and transport paths of gaseous pollutants at urban and rural observation sites [100,101], revealing their potential contributing areas and providing important reference for the development of regional air pollution control measures [9,102,103]. However, the current research community has paid relatively little attention to the sources of pollution in background areas [1], which are suitable for tracking the sources of atmospheric pollutants because of their sensitivity to environmental changes [18]. Specifically, for the Chinese background stations, the combination of airflow trajectory statistics, PSCF, and CWT reveals that the air trajectory of Shangdianzi GAW station shows a southeastern trend in summer and is mainly affected by the northwestern path in other seasons, which are mainly reflected in the background information of Beijing, Hebei, Tianjin, and Inner Mongolia [104]. Moreover, the Mt Waliguan GAW station mainly reflected the pollution sources of Qinghai, Tibet, and Gansu [56,57]. e results of WPSCF show that the overseas potential source areas of CO 2 and CO in Shangri-La are mainly located in South Asia, such as Myanmar, Bangladesh, and northern India [105]. e WCWT analysis of atmospheric pollutants (i.e., SO 2 , CO, NO x , O 3 ) in the Lin'an area indicates that their major sources may be in the big cities around the observation sites, such as Hangzhou, Shanghai, and Nanjing [106]. Based on WCWT, Luan [107] found that the potential sources of gaseous pollutants in Longfeng Mountain, e.g., CO 2 , CH 4 , and CO, are Harbin and Qiqihar, located in its northwest, and Changchun, Jilin, and Shenyang in its southwest. However, the analysis of the sources of gaseous pollutants in past study was mainly for single pollutants and lacked systematic revelation of the sources of SO 2 , O 3 , CO, and NO 2 , the four gaseous pollutants recommended by China Ambient Air Quality Standard (GB3095-2012) [108][109][110]. Furthermore, several studies have focused on background pollutants in Xinjiang; for example, Wang et al. [111] investigated the pollutants of black carbon, Wang et al. [112] studied the mixing ratio characteristics of ozone, and Liu et al. [53] and Han et al. [113] used satellite product data to invert the aerosol optical thickness and tropospheric ozone column concentrations. However, current studies have paid less attention to the sources of background reactive gases. Based on the reactive gas mixing ratio data observed in situ at the background station and introducing the backward trajectory model, this study systematically investigated the sources of CO, SO 2 , NO 2 , and O 3 in the background region of Xinjiang, China, and foreign pollutants were found to contribute significantly to them, such as Mongolia, Kazakhstan, and Russia. e reason for this phenomenon may lie in the unique geographical location and meteorological conditions of Akedala, which is in the Irtysh River Valley, where northwesterly winds prevail all year round and pollutants are more easily transported across the borders [114,115]. Considering the limitation of this study, i.e., investigating only potential sources of reactive gases, multisource remote sensing data and source emission inventories can be introduced in the future, and further studies on cross-border transport of pollutants in the Xinjiang region can be conducted [116,117]. is study provides a valuable reference for monitoring and controlling of the air pollution prevention in Xinjiang region of China by providing an innovative and feasible method.

Conclusions
In this study, cluster trajectory, PSCF, and CWT methods were used to identify the potential source regions and transportation pathways with reactive gases from August 2017 to September 2018 at Akedala station. On the whole, the annual average (median value) of O 3 , CO, NO 2 , and SO 2 mixing ratios was 29.65 (29.87) ppb, 123.78 (102.75) ppb, 3.79 (3.74) ppb, and 4.59 (4.37) ppb, respectively. e monthly mean values indicated that the high O 3 period mainly occurred in February (38.03 ppb), and the high CO and SO 2 monthly mean values were related to the heating and coal burning in the winter and peaked in February (208.49 ppb) and January (5.70 ppb). e mixing ratio of NO 2 was the highest in March (4.51 ppb).
e cluster analysis showed that the Akedala atmosphere was mainly influenced by air masses from the northwest. e trajectory of airflow outside of China could be divided into three main paths. One originated near Lake Balkhash, the second originated from the eastern Kazakhstan, and the third comprised a shortdistance air mass from the southeast that crossed the Altay Mountains and entered Xinjiang.
According to the results of the PSCF and CWT analysis methods, the high PSCF values of reactive gases were located in eastern Kazakhstan, northern Xinjiang, Western Mongolia, and Southern Russia. e WCWT value of CO was concentrated between 180 and 240 ppb in the winter, and WCWT value of CO was mostly between 90 and 120 ppb in the spring and autumn, whereas the contribution value of the main source region of CO was the smallest in the summer, at about 60-90 ppb. e WCWT value of O 3 in the spring and summer was higher than that in the autumn and winter. e main source area contribution of O 3 in the spring and summer was 32-36 ppb and was concentrated at 24-28 ppb in the autumn and winter. e WCWT value of the main source area of NO 2 was low in the summer, at 3-3.5 ppb, whereas the WCWT value of NO 2 was 4-4.5 ppb during the other seasons. For SO 2 , the WCWT was highest, at 5-6.5 ppb, in the winter. e contribution of WCWT to SO 2 sources in the area during the other seasons was 4-5 ppb. 16 Advances in Meteorology

Data Availability
Data used in this paper can be obtained from Quanwei Zhao (zqw311403020128@163.com) upon request.

Conflicts of Interest
e authors declare that they have no conflicts of interest.