3.1. Concentrations of Air Pollutants and Their Variations
In 2013, the annual mean concentration of PM
2.5 exceeded the Grade II NAAQS (GB3095-2012) limit value (35 μg m
−3) in each of the four cities. The annual mean was highest in Liuzhou (70 μg m
−3), followed by Guilin (66 μg m
−3), Nanning (57 μg m
−3), and Beihai (46 μg m
−3). The concentrations of SO
2 were highest in Liuzhou, followed by Guilin, Nanning, and Beihai, with 24 h mean 98th percentile values of 80 μg m
−3, 77 μg m
−3, 52 μg m
−3, and 38 μg m
−3, respectively. These values, however, were lower than the Grade II NAAQS (GB3095-2012) limit value (150 μg m
−3). NO
2 concentrations were highest in Nanning, followed by Liuzhou, Guilin, and Beihai, with 24 h average 98th percentile values of 85 μg m
−3, 74 μg m
−3, 72 μg m
−3, and 32 μg m
−3, respectively. Except for Liuzhou, NO
2 concentrations in the other three cities were lower than the Grade II NAAQS limit value (80 μg m
−3). CO concentrations showed a different trend compared to the other pollutants and were highest in Beihai and Guilin, followed by Nanning and Liuzhou, with 24 h average 95th percentile values of 2.1 mg m
−3, 2.1 mg m
−3, 1.7 mg m
−3, and 1.5 mg m
−3, respectively. All of these values were lower than the limit values (4 mg m
−3) for Grade II in NAAQS (GB3095-2012). O
3 concentrations were highest in Liuzhou, followed by Guilin and Beihai, and then Nanning, with 8 h average 90th percentile values of 148 μg m
−3, 147 μg m
−3, 147 μg m
−3, and 124 μg m
−3, respectively. Each of these values was lower than the Grade II NAAQS limit value (160 μg m
−3).
Table 1 shows descriptive statistics for air pollutants in the four cities in 2013.
Figure 2 presents the seasonal variation of pollutants in terms of box plots, including seasonal means, maxima, minima, 75th percentile, 50th percentile, and 25th percentile during spring (March to May), summer (June to August), fall (September to November), and winter (December to February) in each of the cities during 2013.
Variations in seasonal concentrations of the main air pollutants were similar amongst the four cities, except for CO and O
3. The concentrations of PM
2.5, PM
10, SO
2, and NO
2 were highest during fall and winter, followed by spring, and were at their lowest during summer. These results are similar to previous studies [
19,
20,
21]. Concentrations of CO were highest during spring, fall, and winter in Nanning and Liuzhou and were highest during spring and winter in Guilin and Beihai. For O
3, concentrations peaked during fall and winter in each of the cities. This result is similar to those from Pearl River Delta [
22], which showed that O
3 concentrations were highest during winter, followed by fall, summer, and spring. This may be due to the similar topographical and climatic characteristics of Guangxi and Guangzhou since both are parts of South China. The temperature remains relatively high in October and November in Guangxi, and people usually burn straw in their farmlands during this period. Additionally, the seasonal sugar industry, which is a key industry in Liuzhou and Nanning, begins production in November. Straw and other byproducts are used directly as fuel. The volatile organic compounds from straw burning perhaps lead to photochemical reactions and promote O
3 formation. In summertime, due to high cloud layers and the associated reduction in solar radiation influx, the O
3 formation process is suppressed, and therefore the concentrations of O
3 in this season are lower than those in regions without a rainy season. O
3 concentration changes are very different in Jing-Jin-Ji and the Yangtze River Delta [
19,
20,
21,
22,
23,
24], which exhibited the highest levels during summer and the lowest during winter. O
3 has a positive correlation with temperature [
25], and temperatures in Jing-Jin-Ji and Yangtze River Delta are highest in summer and lowest in winter, leading to suppression of the formation of O
3 during the latter season.
The ratio of PM
2.5/PM
10 at individual locations in the same city exhibited similar seasonal variation characteristics but showed different variation across cities. The average values of the PM
2.5/PM
10 ratio in Nanning, Liuzhou, Guilin, Beihai were 0.63 (range: 0.52 to 0.88), 0.71 (range: 0.68 to 0.88), 0.76 (range: 0.70 to 0.84), and 0.57 (range: 0.50 to 0.92), respectively, which were higher than those in Beijing (0.55) and Lanzhou (0.52) as per previous studies [
16,
26], and were similar to those in Guangzhou (0.65) and Chongqing (0.64) [
16]. The average values of PM
2.5/PM
10 ratio in Liuzhou and Guilin were the highest, followed by Nanning, and lowest in Beihai, indicating that PM
2.5 was the dominant form of particulate matter in Liuzhou and Guilin.
Figure 3 and
Figure 4 present diurnal variations of the main air pollutants during each season (see
Figure 3), and on weekdays (Monday to Friday) and weekends (Saturday and Sunday) (see
Figure 4), excluding government holidays, in the four cities during 2013.
Figure 3 shows that daily variations in the main air pollutant concentrations varied widely during fall and winter and to a lesser extent during spring and summer. The daily variations were unimodal for O
3 and bimodal for PM
10, PM
2.5, NO
2, SO
2, and CO.
Daily variations in the concentrations of PM10 and PM2.5 were similar. The first concentration peak appeared between 8:00 and 11:00. At sunrise, anthropogenic activities such as traffic emissions increased because of the morning rush hour, and pollutant emissions from factories also increased. With the increased solar radiation, photochemical reactions intensified and the concentrations of atmospheric particulate matter increased rapidly. Between 11:00 and 16:00, the concentrations of atmospheric particles decreased due to an increase in atmospheric dispersion. After 16:00, solar radiation and atmospheric dispersion decreased. During this time, traffic emissions increased because of the evening rush hour. During the evening, temperatures decreased, creating inversion conditions that accelerated the accumulation of atmospheric particulate matter. Particulate matter concentrations increased significantly after 22:00. The second peak in concentrations occurred between 23:00 and 24:00. The fact that the second peak did not appear during the evening rush hour showed that traffic emissions were not the main driver of particulate matter concentrations; rather, inversion conditions most likely played a much stronger role in increasing particulate matter concentrations than any one source.
Daily variations in the concentrations of SO2 and NO2 were similar to those of PM10 and PM2.5. The only differences were the times when the peak concentrations occurred and the peak values. For SO2 and NO2, the first concentration peak appeared at 8:00, and the second peak appeared at 20:00. For SO2, the morning peak was evident, and the evening peak was weak in Nanning and Liuzhou. Both peaks were clear in Guilin. Concentrations varied slightly, and the second peak was not clear in Beihai. For NO2, the two peaks were clearer in Guilin and Beihai compared with those in Nanning and Liuzhou. Concentrations were at a minimum at noon in each of the cities, which could have been due to the consumption of NO2 during photochemical reactions.
The concentrations of CO were low, and they varied bimodally, similar to NO2 and SO2 concentrations (except during summer in Beihai). The first peak appeared between 6:00 and 8:00, which was earlier than for other pollutants. The second peak occurred between 19:00 and 21:00, consistent with the evening rush hour.
The diurnal variation of O3 was unimodal, with a peak in the afternoon at 13:00–15:00 and lowest concentrations between 6:00 and 8:00, when solar radiation was weak and O3 formation from photochemical reactions was not intense. During the O3 concentration peak period, the concentrations of NO2, SO2, and CO were low, which was because NO2 and CO were precursors for O3 formation and SO2 was oxidized to form SO42−.
A previous study [
27] suggested that anthropogenic activities and traffic and industrial emissions were higher during weekdays than during weekends, and concentrations of NO
x were higher during weekdays than weekends. From
Figure 4, however, it is apparent that concentrations of NO
2 were slightly lower during weekdays than during weekends in Beihai and that there were almost no differences in Liuzhou and Guilin. As such, there was no obvious “weekend effect”. The concentrations of PM
10 and PM
2.5 also showed no obvious “weekend effect” in the four cities. The concentrations of CO in Guilin and those of O
3 in Nanning were slightly higher on weekends than on weekdays. Besides these minor differences, there was no obvious “weekend effect”. The “weekend effect” is known to be affected by the factors such as local emissions, meteorological conditions, photochemical reactions, the economic structure, and human activities, and its existence and strength can vary significantly among cities.
3.2. Special Case Analysis
There were 15, 11, 11, and 9 air pollution episodes (AQI ≥ 101) that occurred during 2013 in Liuzhou, Guilin, Nanning, and Beihai, respectively, and out of these, 7, 4, 3, and 2 events, respectively, were heavy air pollution episodes (201 ≤ AQI < 300). Air pollution episodes occurred in January, February, March, September, October, November, and December, while heavy air pollution episodes occurred in January and December in each of the cities, and in February, March, and October in Liuzhou, and in October in Guilin (see
Figure 5). In the four cities, the concentrations of SO
2, NO
2, PM
10, CO, and PM
2.5 increased slowly during the pollution episodes, possibly due to stagnant accumulated conditions, which has been described before [
28]. The pollution episode durations were similar in January, March, October, and December in the four cities (
Table 2).
Non-attainment and primary pollutants were similar during the air pollution episodes in the four cities. The non-attainment pollutants were PM10 and PM2.5, and the primary pollutant was PM2.5 in January, March, and December. O3 became one of the non-attainment and primary pollutants in October and November in each of the cities and was also one of the primary pollutants in December in Beihai. Pollution durations in January, October, and December were longer than those in March in each of the cities. Pollution levels were heaviest in Liuzhou and Guilin, followed by Nanning and then Beihai. From a meteorological standpoint, precipitation was extremely low during the pollution periods in all of the cities. The predominant wind direction was east in Nanning; south, southeast, and southwest in Liuzhou and Guilin; and west, southwest, and northwest in Beihai. Wind speeds were low. The average wind speed during each pollution episode was less than 1.5 m s−1 (0.7 m s−1 to 1.2 m s−1) in Nanning and Liuzhou, while that in Guilin and Beihai was from 0.8 m s−1 to 1.8 m s−1. The mean relative humidity was less than 85%. It ranged from 61% to 82% in Beihai, which was higher than that measured in Nanning, Liuzhou, and Guilin (from 48% to 76%).
The meteorological conditions were not good in Guangxi during the last 10 days of December in 2013. The humidity and wind speeds were low, and there was little rainfall. The conditions were not conducive to the dispersion of pollutants and caused low visibility in Guangxi. Air pollution episodes occurred around 20 December 2013 in the four cities. Heavy pollution appeared in Nanning (from 20 to 31 December 2013), Liuzhou (from 20 to 27 December 2013 ), and Guilin (from 19 to 26 December 2013). Moderate pollution occurred in Beihai (from 21 and 31 December 2013). During the pollution episodes, concentrations of the main air pollutants varied in a similar fashion (
Figure 6).
In Nanning, light pollution occurred on 20 December. The pollutant concentrations slowly increased, and heavy pollution formed on 22 December that lasted for 5 days. After this episode, the pollutant concentrations decreased to low levels on 27 December. On 30 December, however, the concentrations of PM10 and PM2.5 increased significantly, and the pollution level changed from light pollution to moderate pollution. Heavy pollution occurred again on 31 December. From a meteorological perspective, the dominant wind directions were southeast and east during the episode. Wind speeds were low, varying from 0.8 to 1.2 m s−1. During 20 and 26 December, the relative humidity was 50–54%. It decreased to 38% on 27 December and down to 30% on 29 December. It then increased to 46% on 30 December and to 54% on 31 December.
The pollution episode in Liuzhou was similar to that in Nanning. Light pollution occurred on 20 December. Pollutant concentrations slowly increased, and heavy pollution formed on 22 December and lasted for 2 days. After this, the pollutant concentrations decreased to low levels on 24 December and air quality became good or moderate on 28 December and lasted 2 days. During 30 and 31 December, however, the concentrations of PM10 and PM2.5 increased significantly and moderate pollution occurred. Visibility showed the opposite trend. From a meteorological perspective, the dominant wind direction was southwesterly during the episode. Wind speeds were low, varying from 0.6 m s−1 to 1.2 m s−1. On 20 and 23 December, the relative humidity ranged from 53 to 54%. It decreased to 38% on 24 December and then to 31% on 27 December.
In Guilin, the air pollution episode occurred 1 day earlier than in Nanning and Liuzhou. Light pollution occurred on 19 December, and pollutant concentrations increased until heavy pollution formed on 21 December, which lasted for 2 days. Pollutant concentrations then decreased to low levels on 24 December, and air quality was good or moderate on 27 December, which lasted for 3 days. During 30 and 31 December, however, the concentrations of PM10 and PM2.5 increased significantly and moderate pollution occurred. Visibility showed the opposite trend. From a meteorological perspective, the dominant wind directions were southeast and south during the episode. The wind speeds were higher than those in Nanning and Liuzhou. Speeds ranged from 1.7 to 2.3 m s−1 between 19 and 21 December and decreased to 1.2–1.6 m s−1 between 22 and 25 December. Wind speeds then increased to 2.6 m s−1 on 26 December, which helped to disperse pollutants and led to good air quality on 27 December. Between 19 and 22 December, the relative humidity ranged from 48 to 57%. It decreased to 44% on 23 December and then to 23% on 29 December. After this, it increased to 44% on 31 December.
Pollution levels were lower in Beihai, and the air pollution episode occurred 1 day later than in other cities. Light pollution occurred on 21 December, and pollutant concentrations slowly increased until heavy pollution formed on 22 December and lasted until 24 December. Pollutant levels then decreased, leading to light pollution episodes between 25 and 31 December. Visibility showed the opposite trend. From a meteorological perspective, the dominant wind direction was westerly, and the relative humidity ranged from 50% to 64% over the period.
3.3. Analysis of Factors Affecting Air Quality
Although air pollution showed regional variation characteristics in the four cities, the pollution duration and levels were different. Pollution was more serious in Liuzhou and Guilin than in Nanning and Beihai. This could be caused by the differences in pollution diffusion and air mass transport, which were impacted by the pollution emission and weather conditions in each city.
According to the Guangxi Statistical Yearbook 2013 (
http://tjj.gxzf.gov.cn/tjsj/tjnj/), Environmental Statistic Data 2013 and Nanning Ambient Air Quality Control Planning Report (
http://sthjt.gxzf.gov.cn/), SO
2, NO
2, and CO are mainly emitted from industrial and vehicular emissions in the four cities. Particulate matter (PM
10 and PM
2.5) mainly originated from dust and industrial emissions. Volatile organic compounds mainly originated from natural sources and industrial and vehicular emission. In 2013, emissions of SO
2, NO
x, and dust from industrial emission sources were highest in Liuzhou, followed by Guilin, Nanning, and Beihai. Emissions of particulate matter, SO
2, NO
x, and CO from vehicles were highest in Liuzhou, followed by Guilin, Nanning, and Beihai. Emissions of particulate matter (PM
10 and PM
2.5) from dust were highest in Nanning, followed by Liuzhou, Guilin, and Beihai. Pollutant emissions in Beihai were lower than those in the other cities.
Pollution levels are closely related to the prevailing meteorological conditions according to the findings of many previous studies [
19,
28,
29,
30,
31]. In this study, pollution episodes and clear days, along with precipitation, wind direction, wind speed, and relative humidity on polluted and non-polluted days during the three seasons (spring, fall, and winter) in the four cities were statistically analyzed to display the differences between heavily polluted episodes and normal days/clear days. The results showed that most rainfall was light during spring, fall, and winter. In Nanning, Liuzhou, Guilin, and Beihai, the percentage of rainy days was 23%, 29%, 27%, and 23%, respectively, during pollution periods (AQI ≥ 101), and 44%, 48%, 52%, and 37%, respectively, during non-pollution periods (AQI < 101). The percentage of precipitation during PM
2.5 pollution periods was low during spring, fall, and winter, with values of 4%, 16%, 9%, and 7% in Nanning, Liuzhou, Guilin, and Beihai, respectively. This highlights that precipitation had a large effect on PM
2.5 pollution. Humidity and wind speeds were lower during pollution periods than during non-pollution periods. Generally, pollutant concentrations peaked when wind speeds and humidity were lowest, and the opposite occurred when wind speeds and humidity were highest. Low wind speeds lead to reduced particulate matter dispersion and this allows particle concentrations to increase substantially. Temperatures are high during summer in Guangxi, and therefore the boundary layer is higher and it is easier for air pollutants to disperse. Therefore, there was less air pollution during summer. Pearson’s correlation raw analysis was carried out to decipher and quantify the possible relationships between different pollutants and meteorological parameters. Pearson correlation analysis between meteorological parameters and the main primary pollutant concentration (PM
2.5) showed that there were significant negative correlations (
p < 0.01) between wind speed, air pressure, relative humidity, precipitation, visibility, and the concentration of PM
2.5 (
Table 3).
Table 4 shows that wind directions were mainly southerly or southeasterly during pollution periods in Nanning, Liuzhou, and Guilin, and wind directions showed little change during non-pollution periods, except that the percentages of different wind directions were different. In Guilin, for example, the main wind directions were southerly and southeasterly during pollution and non-pollution periods during spring, fall, and winter. The percentage of southerly winds decreased from 38% to 36%, and the percentage of southerly and southeasterly winds increased from 17% to 36% during spring. During fall, the percentage of southerly winds decreased from 48% to 26% and the percentage of southerly and southeasterly winds slightly decreased from 28% to 26%. During winter, the percentage of southerly winds decreased from 40% to 17%, and the percentage of southerly and southeasterly winds increased from 28% to 45%.
The meteorological conditions in Beihai were different from those of the other three cities. As a coastal city, its wind directions were mainly westerly and southwesterly. Similarly to the other three cities, there was little change in the main wind direction during pollution and non-pollution periods, but the percentages of wind directions were different. This shows that wind direction can have important impacts on air pollutant concentrations.
Figure 7 shows the cluster analysis of trajectories and the analysis of concentration weighted trajectories (CWT) [
18] during the three pollution seasons (January in winter, March in spring, and October in autumn) in four cities. Results show that airflows reaching the four cities in the three seasons mainly come from non-local areas (outside Guangxi), with a total frequency of more than 50%, but this does not mean that long-distance transportation is the main factor causing pollution incidents. As shown in
Figure 7a–d, the occurrence frequency of air masses reaching the four cities from non-local areas in winter are about 50%, 60%, 50%, and 100%, respectively. Combined with the CWT analysis, it is shown that non-local airflows with higher AQI are the main cause of pollution in Liuzhou (about 50% in
Figure 7b) and Guilin (about 60% in
Figure 7c), while pollution in Nanning (
Figure 7a) and Beihai (
Figure 7d) is mainly due to local emissions. In spring (
Figure 7e–h), 90% of the airflow in Nanning City comes from non-local areas, but the AQI value of these airflows are low, whereas that of airflows from local areas are higher. The airflows in the Liuzhou and Guilin cluster trajectories are all AQI hotspots (high value more than 100), while those from the non-local areas of Liuzhou and Guilin are as high as 100% and 70%, which indicates that Liuzhou and Guilin are strongly affected by long-distance transportation in spring, mainly from in the northeast (e.g., Hunan and Hubei regions) and the western region of the Pearl River Delta. It can be seen from
Figure 7h that the AQI in Beihai is lower than in the other three cities in spring because the airflows come from a relatively clean sea area. In the autumn (
Figure 7i–l), the cluster trajectories of the four cities are associated with high AQI, and the airflows mainly come from Hunan and Hubei provinces in the northeast of Guangxi. There is another cluster trajectory in western Guangdong, which is also associated with high AQI. This shows that autumn pollution incidents in the four cities of Guangxi were greatly affected by regions outside Guangxi, which highlights the importance of joint prevention and control.
From satellite-based fires-pot analysis (
https://earthdata.nasa.gov/labs/worldview), it is apparent that there were large-scale fire points to the west (Yunnan) and southwest (Thailand, Myanmar, and Vietnam) of Guangxi in January (winter) 2013. In addition to Liuzhou, the other three cities have similar cluster trajectories through Yunnan and Guizhou, and the AQI associated with the trajectories is high. The fire points in March (spring) were located in the southwest of Guangxi, where the airflows have a higher AQI. In October (fall), they were centralized in Guangxi and located to the northeast, and AQIs in the airflows from the northeast were high. In general, air pollution during winter (January) in Nanning and Liuzhou is mainly influenced by local emission sources, while that in Liuzhou and Guilin is mainly influenced by long-distance transportation from the south and northeast of Guangxi. In spring, air pollution in cities is mainly influenced by local straw burning emission in the south of Guangxi. The air pollution in fall is affected by long-distance transportation from the northeast of Guangxi (e.g., Hunan and Hubei province).