The spatiotemporal characteristics of the air pollutants in China from 2015 to 2019

China’s rapid industrialization and urbanization have led to poor air quality, and air pollution has caused great concern among the Chinese public. Most analyses of air pollution trends in China are based on model simulations or satellite data. Studies using field observation data and focusing on the latest data from environmental monitoring stations covering the whole country to assess the latest trends of different pollutants in different regions are relatively rare. The State Council of China promulgated the toughest-ever Air Pollution Prevention and Control Action Plan (Action Plan) in 2013. This led to a major improvement in air quality. We use the hourly Air Quality Index (AQI) and mass concentrations of PM2.5, PM10, CO, NO2, O3, and SO2 in 362 cities from 2015 to 2019, obtained from the Ministry of Ecology and Environment, to study their temporal and spatial changes and assess the effectiveness of the policy on the atmospheric environment since its promulgation and implementation. We found that the national and regional air quality in China continues to improve, with PM2.5, PM10, AQI, CO, and SO2 exhibiting negative trends. However, O3 and NO2 pollution is an urgent problem that needs to be solved and the current control strategy for PM2.5 will only partially reduce the PM2.5 pollution in the western region. Although the implementation of "Action Plan" measures has effectively improved air quality, China’s air pollution is still serious and far from the WHO standard. Implementing measures for continuous and effective emissions control is still a top priority.


Introduction
To reduce pollution, considerable effort and resources must be invested to implement the measures in place. The evaluation of the effectiveness of the promulgated measures can provide important information for China and other developing countries and highly polluting countries to formulate high-efficiency air quality policies. Recent studies often use model simulation methods to study pollutant changes, for example, Zhang et al. quantified the emission reduction of air pollutants by examining each measure of the "Action Plan" [1] using models. a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 37.8% lower, respectively, than in 2012 [3]. Jiang et al. used model simulation to conclude that, from 2012 to 2017 in the Pearl River Delta, the "Action Plan" effectively reduced SO 2 by 34%, NOx by 28%, PM 2.5 by 26%, and volatile organic compounds by 10% [4]. Zheng et al. [5] and Xue et al. [6] used similar methods to obtain similar conclusions.
In addition, many previous studies have used satellite retrieved aerosol optical depth to estimate trends in PM 2.5 concentrations. Peng et al. reported that China's PM 2.5 concentration increased significantly between 1999 and 2011, especially in central and eastern regions of China. The proportion of areas with PM 2.5 concentrations higher than 35 μg/m 3 increased year by year, and the areas with PM 2.5 concentrations lower than the annual primary standard of 15 μg/m 3 decreased continuously [7]. Ma 3 ) concentration has been increasing steadily at a rate of 7% per year [10].
There are also many reports on the spatial and temporal changes in gas and particulate pollutants in China, but most of them are limited to a city or a certain type of pollutants [11][12][13][14][15][16]. For example, Zhou et al. studied the spatiotemporal changes in air quality in Beijing [11], Guo et al. used ground observations to assess the PM 2.5 concentration and exposure throughout China [12], and Chen et al. reported spatiotemporal changes in PM 2.5 and their relationship to meteorological factors in Nanjing City [13], among many other studies.
Some studies have used monitoring networks to study the spatiotemporal changes in multiple pollutants at multiple sites (urban). For example and Silver et al. reported that substantial changes in the air pollutants PM 2.5 , SO 2 , NO 2 and O 3 occurred across China from 2015-2017 according to data from 1689 sites [17]. Guo et al. studied the air quality of 366 cities in 2015-2017 and the spatiotemporal changes in six air pollutants, i.e., PM 2.5 , PM 10 , SO 2 , CO, NO 2 , and O 3 , and AQI [18]. The most recent trends in pollutant changes have also exhibited changes. For example, both the present study and that conducted by Guo et al. found that NO 2 hardly changed from 2015 to 2017.
Here, we use the latest data from environmental monitoring stations covering the whole country to evaluate the most recent trends of various pollutants in different regions of the country.

Materials and methods
The hourly data on national pollutants were downloaded from http://beijingair.sinaapp.com/, which were obtained from http://pm25.in. The URL http://pm25.in provides information identical to that provided by the official Ministry of Ecology and Environment (http://106. 37 [22]).
We conducted a statistical analysis of the PM 2.5 , PM 10 , SO 2 , CO, NO 2 , O 3 and AQI in different regions every year from 2015 to 2019.
The total number of monitoring points in different regions every year N is denoted by n. Let x i be the i th sample and w i be the i th weight; the annual mean of the pollutants in different regions every year is calculated using the following formula: If there is no weight variable, the formula is reduces as follows: The standard deviation of pollutants in different regions every year is calculated as follows: s ¼ ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffiffi The lower limit of the 95% confidence interval of the mean of the pollutants in different regions every year is calculated as follows: The upper limit of the 95% confidence interval of the mean of every region every year is calculated as follows: where t (1−α/2) is the (1 − α/2) critical value of the Student's t-statistic with n-1 degrees of freedom.
We counted the annual relative change of pollutants of every region as follows: where x i and x j represent the absolute value at time i and j, respectively; and dx represents the relative change.
The annual average relative change of pollutants of every region is calculated as follows: where x t . . . x t+n represents the absolute value at time t. . .t+n and d� x represents the annual average relative change.
We also counted the number (percentage) of cities with positive and negative annual relative changes and the number (percentage) of cities in different intervals of annual relative change.
was determined to be primarily the result of mineral dust from the Taklimakan Desert [18]. Fan et al. also found that air pollution was most serious in Xinjiang Province from 2014 to 2018 [19]. However, the distribution obtained in this study was somewhat different from Zhang et al. [2]. This may be attributed to model uncertainty or the limited number of monitoring stations available in Xinjiang in this study.
Among key provincial capitals and well-known tourist cities, only one city, Sanya, has an annual mean PM 2.5 of 14.6 μg/m 3 that meets standard I ( [20]. In addition, we found that, despite Altay (10.7 μg/m 3 ) being the city with the lowest concentration, the Altay region did not reach the annual average concentration of the WHO Air Quality Guidelines (AQG, 0-10 μg/m 3 ).
AQI. From 2015 to 2019, the national annual mean AQI was 72.9, from a regional perspective, Xinjiang (101.4), the Northern region (94.5), the Weihe River Basin (78.7), and the Central region (73.1) were the regions with the highest annual mean from 2015 to 2019 ( Table 1). The AQI in Xinjiang was 39% higher than the national annual mean, and Hetian  spatial clustering [23]. Fang et al. indicated that urbanization has played an important negative role in determining air quality in Chinese cities. The population, urbanization rate, automobile density, and secondary industry proportion were all found to have had a significant influence over air quality [24]. Wang et al. showed that three dust source areas, namely, 'Northwesterly Sources', 'Northerly Sources', and 'Loess Plateau Source', and an anthropogenic 'Southerly Source' contributed to the high particulate matter concentrations at Xian of the Weihe River Basin [25]. Inner Mongolia (67.5), the Eastern region (67.5), the Northeast region (66.9), the Sichuan Basin (65.6), and Qinghai-Tibet (63.6) (sorted by annual mean) have lower AQI values than the national annual mean (72.9); Yungui (47.9) and Southern (52.4) have the lowest annual mean values. Diqing (34.4) and Lijiang (39.3) in Yungui and Sanya (33.9), Haikou (38.2), and Shenzhen (49.5) in the Southern region are the cities with lowest annual mean. This finding is consistent with a previous study [26].  Table 1). The value in the Northern region was 51% higher than the national annual mean. Although the annual mean in Inner Mongolia was lower than that in the Northern region, it was still 17% higher than the national annual mean, indicating that the SO 2 pollution in the Northern region is serious. Jinzhong  (Table 1). The value in the Northern region was 26% higher than the national annual mean. Linfen NO 2 . From 2015 to 2019, the national annual mean NO 2 was 28.4 μg/m 3 , from a regional perspective, the Northern region (36.6 μg/m 3 ), Weihe River Basin (29.6 μg/m 3 ), and Eastern region (32.7 μg/m 3 ) were the regions with the highest annual mean from 2015 to 2019 ( Table 1). The value in the Northern region was 26% higher than the national annual mean. Tangshan (55.9 μg/m 3 ) and Xingtai (53.2 μg/m 3 ) in the Northern region are the cities with the highest annual mean in the country. Xian (50.0 μg/m 3 ) in the Weihe River Basin and Huzhou (49.3 μg/m 3 ) and Suzhou (46.2 μg/m 3 ) in the Eastern region are the cities with the highest annual mean values.
The fast developing resource and pollution intensive industries along with the 'Go West' movement and weak emission controls [28] contributed to the higher rate of increase in NO 2 over the Western region from 2005-2013 than over that over the Southwestern, Northern, Eastern, and Southern regions.
We found that NO 2 shows a different trend in cities in the same region. The concentration in certain cities has increased, while the concentration in nearby cities has decreased, although they are in the same region (Fig 1). Krotkov et al. also found that NO 2 has a large spatial heterogeneity [9]. Silver et al. suggested that the trend of spatial heterogeneity of NO 2 may be partly due to its relatively short lifespan [17].    [31][32][33]. From a regional perspective, the relative change in the annual mean (median) of PM 2.5 in all regions from 2015 to 2019 was negative. Qinghai-Tibet (-16.3%), Inner Mongolia (-9.9%), the Northeast region (-8.7%), the Eastern region (-8.4%), and the Sichuan Basin (-7.5%) are the regions with a rapid decline (sorted by relative change). The average change in Qinghai-Tibet was 120% higher than the national relative change. Ali (-20%) of the Qinghai-Tibet, Hegang (-33.9%) and Baicheng (-18.8%) in the Northeast region, and Haimen (-14.4%) and Jinhua (-12.7%) in the Eastern region are the cities with a faster decline.   N  68  3  66  5  66  5  71  0  66  5  53  18  6  65   X  14  2  14  2  14  2  16  0  16  0  10  6  5  11   E  48  0  48  0  48  0  47  1  47  1  43  5  9  39   S  37  0  37  0  37  0  34  3  37  0  29  8  12  25   SC  21  1  22  0  22  0  20  2  20  2  18  4  7  15   NE  37  1  38  0  38  0  38  0  35  3  37  1  11   Yin et al. showed that the 'Lhasa pattern' of Qinghai-Tibet may serve as a positive example for other regional hub cities. Effective air pollution control measures collectively contributed to the synchronous improvement of the economy and air quality in Lhasa, moreover, lower concentrations of air pollutants are observed in Lhasa except for O 3 because of the relatively isolated location, low air pollutant emissions associated with its industrial structure and renewable energy consumption [34]. Qiu et al. showed that in Baotou, which is a typical industrial city in Inner Mongolia for evaluating the current national control measures, that the total emissions of SO 2 , NO X , and PM 2.5 were 211.2 Gg, 156.1 Gg, and 28.8 Gg in 2013, respectively, and should be reduced to 39.0%, 32.0%, and 24.4% in 2020, respectively. Even for a typical industrial city, the reduction of PM 2.5 concentrations not only requires decreases in emissions from the industrial sector as well as residential sources [35]. Zheng et al. showed that the emission reduction rates markedly accelerated after the year 2013, thus confirming the effectiveness of China's Clean Air Action policy. From 2013-2017, China's anthropogenic emissions decreased by 59 % for SO 2 , 21 % for NOx, 23 % for CO, 36 % for PM 10 , and 33 % for PM 2.5 . Emission control measures are the main drivers of this reduction, and pollution controls on power plants and industries are the most effective mitigation measures [36].
The Weihe River Basin (-7.3%), the Northern region (-7.0%), the Central region (-6.8%), and the Southern region (-6.5%) have slightly lower values than the national relative change. Yungui has second lowest relative change (-4.9%) because it has the lowest concentration in the country, and there is little room for a decrease. Xinjiang has the lowest relative change (-3.0%), which is 59% lower than the national relative change, but Xinjiang has second highest concentration in the country. Xinjiang has the characteristics of high pollution and low improvement. This is at least partially due to the natural PM sources, such as dust. Lu et al. showed that the high PM 2.5 concentration was mainly affected by sand and dust in the northwest of China and by human activities in the eastern region [37]. Zhang et al. showed that in western China, dust particles are very important for PM 2.5 and the current control strategy of PM 2.5 (that is, reducing VOC and PM emissions from fossil / non-fossil combustion) will only partially reduce the pollution of PM 2.5 of the western region [15]. Cai et al. noted that with the implementation of the "Action Plan", the emissions of SO 2 , NO X , and PM 2.5 will decrease by 40%, 44%, and 40% in 2020, from the 2012 levels in Jing-Jin-Ji, respectively. Consequently, the ambient annual PM 2.5 concentration of 2020 will be 37.8% lower than that in 2012. Thus, the "Action Plan" provided an effective approach for alleviating PM 2.5 pollution levels in the Jing-Jin-Ji region [3].
From 2015 to 2019, the relative changes in the three major economic belts in the Northern, Eastern (Yangtze River Delta), and Southern regions (Pearl River Delta) are -26.6%, -30.3%, and -24.8%, respectively. Zhang et al. also noted that national and regional concentrations declined in all years from 2013 to 2017, the PM 2.5 of the Northern region (Beijing, Tianjin, Hebei and the surrounding areas), the Yangtze River Delta and the Pearl River Delta have decreased by 38%, 27%, and 21%, respectively [2]. Silver et al. found that the relative changes in the median PM 2.5 in all provinces except Shanxi and Jiangxi from 2015 to 2017 showed a negative trend [17]. At the same time, Lin et al. also estimated consistent trends in satellite data from 2011 to 2015 [30]. Jiang et al. estimated that the "Action Plan" would decrease the PM 2.5 by 26%, SO 2 by 34%, and NO 2 by 28% in the Pearl River Delta [4].  [18]. From a regional perspective, the relative change in the annual mean (median) PM 10 in all regions from 2015 to 2019 was negative. Qinghai-Tibet (-14.1%), the Sichuan Basin (-9.1%), Inner Mongolia (-7.3%), the Northeast region (-7.0%), and the Eastern region (-6.8%) are the regions with a rapid decline (sorted by relative change). The value in Qinghai-Tibet was 139% higher than the national relative change. Naqu (-24.1%) and Guoluo (-23.7%) in Qinghai-Tibet, Chengdu (-11.6%) and Zigong (-10.7%) in the Sichuan Basin, and Hulunbeier (-14.7%) in Inner Mongolia are the cities with the fasted decline.
The values in the Weihe River Basin (-5.9%), the Northern region (-5.6%), the Central region (-5.6%), and Yungui (-5.6%) are equal to or slightly lower than the national relative change (-5.9%). The Southern region has a lower relative change (-4.8%) because it has the lower concentration in the country and, thus, there is not much room for a decrease. Xinjiang has the lowest relative change (-0.7%), which is 88% lower than the national relative change, but Xinjiang has the highest concentration in the country. Xinjiang has the characteristics of high pollution and low improvement.
AQI. Nationally, the annual average relative change was -5.3% and the annual average relative change of 96% cities was negative from 2015 to 2019. The annual mean AQI decreased continuously year by year. The annual mean in 2019 was 20.5% (relative change) lower than that in 2015. The annual mean of 349 cities decreased in 2019 (Tables 2-5; S19-S27 Tables in S1 File ; Figs 1 and 2). From a regional perspective, the relative change in the annual mean (median) AQI in all regions from 2015 to 2019 was negative. Qinghai-Tibet (-9.4%), the Sichuan Basin (-6.9%), Inner Mongolia (-6.7%), the Northeast region (-6.6%), and the Eastern region (-6.3%) are the regions with a rapid decline (sorted by relative change). The value in Qinghai-Tibet was 74% higher than the national relative change.
The values in the Weihe River Basin (-5.1%), the Northern region (-4.7%), the Central region (-4.7%), and the Southern region (-4.7%) are slightly lower than the national relative change (-5.3%). Yungui has the lowest relative change (-4.1%) because it has the lowest concentration in the country and there is little room for a decrease. Xinjiang has the lowest relative change (-2.1%), which is 60% lower than the national relative change. Xinjiang also has the highest concentration in the country. Xinjiang has the characteristics of high pollution and low improvement.
From 2015 to 2019, the relative changes in the three major economic belts in the Northern region, the Yangtze River Delta, and the Pearl River Delta were -17.8%, -23.3%, and -18.6%, respectively. Beijing, Shanghai, Guangzhou, and Shenzhen exhibited changes of -36.7%, -29.7%, -18.9%, and -14.8%, respectively. Beijing and Shanghai exceeded the national relative change of -20.5%. SO 2 . Nationally, the annual average relative change was -16.3% and the annual average relative change of 96% cities was negative from 2015 to 2019. The annual mean SO 2 decreased continuously year by year. Silver et al. also found that the annual average relative change of 90% of monitoring stations was negative from 2015 to 2017 [17]. The annual mean in 2019 was 51.2% (relative change) lower than that in 2015. Compared with 2015, the annual mean of 354 cities decreased in 2019 (Tables 2-5; S28-S36 Tables in S1 File; Figs 1 [38]. The Weihe River Basin (-16.0%), Inner Mongolia (-14.3%), the Sichuan Basin (-13.5%), and Xinjiang (-13.3%) had slightly lower values than the national relative change (-16.3%). Yungui (-12.5%), Qinghai-Tibet (-11.8%) and the Southern region (-8.9%) had the lowest relative change because these regions have almost the lowest concentrations in the country, thus, there is little room for a decrease. Compared with other pollutants, it can be seen that the SO 2 in all regions of the country has decreased significantly, the minimum value is also -8.9%, and all regions have greatly improved.
From 2015 to 2019, the relative changes in the three major economic belts in the Northern region, the Yangtze River Delta, and the Pearl River Delta were -63.9%, -54.5%, and -32.2%, respectively. Huang et al. also reported that in the Beijing-Tianjin-Hebei region, reductions of 63.5% occurred for sulphur dioxide and 30.5% occurred for carbon monoxide from 2013 to 2017 [38].
Yungui (-3.8%), Qinghai-Tibet (-3.7%) had the lowest relative change because it has the almost lowest concentration in the country, thus, there is little room for a decrease.
From 2015 to 2019, the relative changes in the three major economic belts in the Northern region, the Yangtze River Delta, and the Pearl River Delta were -32.0%, -21.8%, and -19.5%, respectively.
Huang et al. reported that in the Beijing-Tianjin-Hebei region, reductions of 30.5% occurred for carbon monoxide from 2013 to 2017 [38]. Streets et al. showed that emissions of carbon monoxide are projected to decline from 115 mt in 1995 to 96.8 mt in 2020 due to more efficient combustion techniques, especially in the transportation sector, although if these measures are not realized, carbon monoxide emissions could increase to 130 mt by 2020 [39]. NO 2 . Zhao et al. found a nonlinear relationship between PM 2.5 and the precursor NO 2 and showed that the effects of strengthened vehicle emission standards on national air quality improvement were hindered by the complex nonlinear response of the PM 2.5 concentration to NOx emissions [40].
Nationally, the annual average relative change was only -2.5% and the annual average relative change of only 76% cities was negative from 2015 to 2019 (Tables 2-5; Tables S46-S54 [18]. These results are consistent with the trend we found. By 2019, our research found that the NO 2 concentration was reduced and pollution was improved. From a regional perspective, the annual average relative changes in the mean (median) in regions from 2015 to 2019 were negative. The Northeast (only -6.2%) and Eastern (only -3.8%) regions are the regions with a fast decline (sorted by relative change). The value in the Northeast region was 148% higher than the national annual average relative change. Songyuan (-15.1%) and Suihua (-13.6%) in the Northeast region and Shaoxing (-12.7%) and Quzhou (-10.0%) in the Eastern region are the cities with the fastest decline.
The Northern region (-2.5%) and Inner Mongolia (-2.5%) are equal to the national annual average relative change. Yungui (-1.9%), Sichuan Basin (-1.6%), the Central region (-1.5%), the Southern region (-1.5%), and Qinghai-Tibet (-1.4%) have lower values than the national annual average relative change (-2.5%). Xinjiang has a positive annual average relative change (+0.7%) because Hami (+20.7%) in the Xinjiang region is the city with the fastest increasing rate. However, Xinjiang has a high concentration throughout the country. Xinjiang has the characteristics of high pollution and low improvement.
From 2015 to 2019, the relative changes in the three major economic belts in the Northern, the Eastern (Yangtze River Delta), and Southern (Pearl River Delta) regions were -9.9%, -15.6%, and -7.3%, respectively.
Huang et al. reported that a key strength of the Air Pollution Prevention and Control Action Plan (APPCAP) is that it demonstrates China's ability to control air pollution. year −1 or 5.2% year −1 . A total of 55% of stations exhibited significant trends, and of these, 92% were positive. Across all stations, the percentage of days where the WHO AQG (100 μg/m 3 ) was exceeded for MDA8 rose from 9.8% in 2015 to 12.4% in 2017. The annual mean O 3 values showed similar relative and absolute trends, which is consistent with our results [17]. From a regional perspective, the annual average relative changes in the mean (median) in all regions from 2015 to 2019 were positive. Xinjiang (7.0%), the Central region (7.0%), the Northern region (6.7%), and Qinghai-Tibet (5.8%) were the regions with a rapid increase (sorted by relative change). Hami (22.6%) in Xinjiang and Chuzhou (37.1%) and Wuhu (19.4%) in the Central region were the cities with the fastest rates of increase.
Li et al. reported that a more important factor for ozone trends in the North China Plain is the 40% decrease of PM 2.5 over the 2013-2017 period, which slowed down the aerosol sink of HO 2 radicals and stimulated O 3 production [41].
Li et al. reported that from 2013-2017, increasing ozone trends of 1-3 ppbv a −1 occurred in the megacity clusters of eastern China, which we attribute to changes in anthropogenic emissions. Anthropogenic NOx emissions in China are estimated to have decreased by 21%, whereas volatile organic compounds (VOCs) emissions changed little. Decreasing NOx would increase ozone under the VOC-limited conditions thought to prevail in urban China [41].
From 2015 to 2019, the relative changes in the three major economic belts in the Northern region, the Yangtze River Delta, and the Pearl River Delta were 29.0%, 13.2%, and 9.5%, respectively.

Factors contributing to the decrease in air pollutant levels
A number of policy actions contributed to the decrease in PM 2.5 , SO 2 , NO 2 , and CO levels. Firstly, the emission standards of thermal power plants and all emission-intensive industrial sectors (such as steel and cement) have been strengthened. By the end of 2017, more than 95% of China's coal-fired power plants were equipped with flue gas desulfurization (FGD) and selective catalytic reduction (SCR) or selective non-catalytic reduction (SNCR) systems and 71% of the coal-fired power generation capacity reached the "ultra-low emission" standard. In addition, industrial boiler were upgraded and small coal-fired boilers were eliminated, which was important because large-scale operation boilers are widely equipped with SO 2 and particulate matter control devices. The elimination of backward industries phases out obsolete or inefficient technology in various industries and allows for structural adjustments. In addition, clean fuel was promoted in the residential sector and advanced stoves and clean coal nationwide were promoted from 2013 to 2016. In 2017, the use of natural gas and electricity to replace coal was further promoted, which affected 6 million households nationwide, of which 4.8 million were located in the Beijing-Tianjin-Hebei area and surrounding areas. The benefits of promoting clean fuels in the residential sector are also obvious throughout the country, and the transportation sector elevated emission standards and imposed mandatory elimination of old vehicles that do not meet emission standards [2].

Conclusion
We analyzed the spatiotemporal changes in six air pollutants and AQI in 362 cities in China from 2015 to 2019. The national and regional air quality in China continues to improve. PM 2.5 , PM 10 , AQI, CO, and SO 2 have exhibited negative trends; PM 2.5 is the most important air pollutant in most regions in China, particularly in the Northern China, Xinjiang, Central China, and the Weihe River Basin; the spatial distribution of NO 2 is heterogeneous, O 3 and NO 2 pollution is an urgent problem that needs to be solved; the main reason for the change in air quality is human activities; however, the current control strategy for PM 2.5 will only partially reduce the PM 2.5 pollution in the Western region. Although the implementation of the "Action Plan" measures has effectively improved air quality, China's air pollution is still serious and far from the WHO standard. Measures for continuous and effective emissions control are still a top priority.