In global ecological protection and urban development, air pollution has become a crucial issue affecting human health and the environment, along with climate change, which has attracted increasing attention from scholars and government administrators1,2,3,4. The number of air pollutants has grown from dozens to hundreds now. Major phenomena, such as acid rain, ozone holes, and urban heat islands, continuously challenge the airways of urban dwellers and sustainable urban development5,6. After decades of efforts by developed countries in Europe and America, sulfur pollution and soot pollution in urban air have basically been solved, and the environmental air quality has been greatly improved. However, the contribution of NOx to environmental acidification is growing as the tropospheric ozone problem is aggravated by the emission of nitrogen oxides (NOx) and volatile organic compounds (VOCs)7. In recent decades, the enormous emissions of airborne particulate pollution have become the primary cause of climate change and air pollution in China8. Air pollution has gradually changed from a local problem to a regional problem and may be either emitted directly (primary pollutants) or formed in the atmosphere9. Major metropolises have invested large human and financial resources to control particulate pollution, but the effect is still unsatisfactory. CO, NOx, hydrocarbons and photochemical smog pollution remained serious.

Urban agglomeration (UA) is a crucial phenomenon at the urban scale and form in the process of globalization. In different stages of socioeconomic and human development, scholars use different terms to describe this urban landscape phenomenon, such as megalopolis10, urban clusters11, metropolitan chain areas12, urban agglomerations13, metropolitan interlocking regions14 and conurbations15. Despite the inconsistency in terminology, this extensive, multicentered, multicity urban landscape has been well recognized. Urban agglomerations (UAs) in China are dynamic regions with immense potential for regional development, which play a vital role in global competition and the international division of labor and gather many industries, transportation and residents, and the mechanisms of environmental pollution are more complicated8,16. Recently, in China, air pollution incidents have changed from frequent occurrence to visible improvement. Meanwhile, the characteristics and influencing factors of air quality have attracted the attention of many researchers17,18,19. Together these results indicate that the spatiotemporal patterns of air pollution are not only related to meteorological conditions such as terrain20, temperature21, and humidity22, but also to the emissions of local and surrounding cities, exhibiting distinct regional features 23. For example, source contributions to PM2.5 indicate that coal24, biomass burning25, transport26 and industry27 are the main sources.

Only three air pollutants, SO2, NO2 and PM10, were monitored, and the air pollution index (API) was calculated by the China National Environmental Monitoring Centre before 2013. Then, China promulgated the Ambient Air Quality Standard (GB3095-2012), adding three indicators, PM2.5, CO, and O3, for air quality evaluation. The air quality index (AQI) was used to evaluate the air quality, which makes the air quality evaluation more accurate and strict. At present, China's air quality evaluation indicators are mainly PM2.5, PM10, SO2, NO2, CO and ozone. However, in terms of identifying regions with similar air pollution behaviors and locating emission sources, few studies have comprehensively considered the spatiotemporal comparison of primary pollutants among UAs and detailed emission sources. For many countries, without this comprehensive assessment, it is impossible to understand the synergies and interrelationships between energy-related urban sprawl and air pollution. Previously published studies on spatiotemporal variations in PM2.5 and O3 have mainly focused on a specific region in China (Beijing-Tianjin-Hebei, Yangtze River Delta, or Pearl River Delta) and shorter observation times, ignoring UAs with weak economic development, such as northwest and central China28,29,30. Few studies have compared the changes in pollutant concentrations and primary pollutants in multiple UAs. Six air pollutants and 19 UAs were considered comprehensively in our work. Regional joint prevention and control of air pollution should be strengthened. As China has entered the 14th Five-Year Plan period, it is necessary to review and summarize the air quality situation of China's key UAs in the 13th Five-Year Plan (FYP) and study the impact of anthropogenic emission sources generated in urban construction on air pollution. This is a research highlight of this paper. Given these facts and based on air quality monitoring data, this study aims to investigate (1) the temporal variation characteristics of primary air pollutants in 19 UAs during China's 13th Five-Year Plan period, (2) the spatial distribution features of the barycenter of six ambient air pollutants from 2015 to 2019, (3) the dissimilarity of the contribution of atmospheric pollutants produced by anthropogenic sources at the UA scale.

This paper is organized as follows. In the first stage, we described the study area and the datasets used in the study. Then, we will present the spatial–temporal variation characteristics of primary pollutants based on UAs and anthropogenic source emissions. Finally, we summarized key research findings and drew conclusions.

Study area and methods

China’s urban agglomerations

Due to the compound and regional features of air pollution, UAs are the main unit of joint prevention and control of air pollution. UA development planning in China can be divided into world-class, national and regional levels. The 19 selected Chinese UAs from south to north and from west to the east are Beibu Gulf, Pearl River Delta, Central Yunan, Central Guizhou, Chengdu-Chongqing, Triangle of Central China, Central Plains, West Side of the Straits, Yangtze River Delta, Shandong Peninsula, Tianshan North-Slope, Lanzhou-Xining, Guanzhong Plain, Ningxia Yanhuang, Hohhot-Baotou-Ordos-Yulin, Taiyuan-Jinzhong, Beijing-Tianjin-Hebei, Mid-southern Liaoning and Harbin-Changchun UAs. Nineteen UAs locations and names are annotated with colors and numbers in Fig. 1. Moreover, for convenience of expression, Table 1 briefly lists the basic indicators of the 19 UAs in China.

Figure 1
figure 1

Geographical locations of 19 UAs in China. This figure was generated with python 3.7.6 based on geopandas and contextily packages. The scope of 19 UAs were obtained according to the Chinese planning documents. The source of the base map from contextily package is ‘http://{s}{z}/{x}/{y}.jpg’.

Table 1 Basic indicators of UAs in China.

Data sources and statistical analyses

Real-time daily hourly concentrations in each monitoring site covering 19 UAs (Table 2) were derived from China's National Environmental Monitoring Centre. In addition, city daily data (excluding detailed sites) were also used for verification. In this study, the monitoring data were pretreated with python language programming based on Spider platform to ensure the effectiveness of air pollutant concentration data. First, some outliers, such as hourly pollutants concentrations in each station less than zero and missing values, were removed. Secondly, the daily average value of each pollutant in each station was only calculated when the valid data of the day in each station is greater than or equal to 20 h. And the number of effective monitoring points shall not be less than 75% of the total number of urban points. Then the daily mean value (i.e., arithmetic mean of a 24-h monitoring value on a natural day) of urban pollutants can be obtained by calculating the average of each pollutant at each station. Finally, month and annual mean pollutants concentrations for the monitoring sites and UAs were obtained according to the arithmetic mean method31. Comparing the annual average concentration of the city with the government report, our calculation results were found to be reliable.

Table 2 The raw data example of cities.

It is well known that 2020 is the year for China to fight theCOVID-19 pandemic with all their strengths32. Home isolation was adopted to reduce mobility, and social and economic development stagnated. Meanwhile, anthropogenic activities such as transportation and industry are directly proportional to air pollution. Therefore, the study period was conducted from January 1, 2015, to December 31, 2019, avoiding the impact of the COVID-19 epidemic. The air quality index (AQI) is a quantitative description of air quality data, which can denote the short-term air quality status and trends in a city33. The overall AQI represents the maximum of the sub-AQI of all pollutants, where when the AQI is higher than 50, the highest sub-AQI contributor is defined as the primary pollutant on that day34. The calculation of the AQI and primary pollutants is as follows:

$$ IAQI_{P} = \frac{{IAQI_{Hi} - IAQI_{Lo} }}{{BP_{Hi} - BP_{Lo} }} \times (C_{P} - BP_{Lo} ) + IAQI_{Lo} $$
$$ {\text{AQI}} = {\text{Max}}\left\{ {IAQI1, \, IAQI2, \ldots , \, IAQIp, \ldots ,IAQIn} \right\} $$

where IAQIP is the air quality sub-index of pollutant item P. Cp is the concentration value of P. BPHI and BPLO are the high and low values of the pollutant concentration limit, respectively, similar to Cp35. IAQIHI and IAQILO are air quality sub-indexes corresponding to BPHI and BPLO. After calculating the air quality sub-index of various pollutants, the maximum value is the air quality index (AQI). Primary pollutants are the largest corresponding pollutant items of IAQIP.

Illustration of spatiotemporal variations in sector emissions

The emission source inventory used in the experiment is from the MEIC (Multi-resolution Emission Inventory for China) 2017 version developed by Tsinghua University, with a spatial resolution of 0.5° × 0.5° ( MEIC is a bottom-up emission inventory model based on a cloud computing platform that covers ten air pollutants and carbon dioxide emissions (SO2, NOx, CO, NMVOC, NH3, PM2.5, PM10, BC, OC and CO2) . It should be noted that the emission inventory contains the location and the amount of emission sources information, but does not establish a direct relationship with ambient air quality. The emission source inventory here is mostly used as input data of various chemistry transport model36. Since pollutants go through a series of complex physical and chemical processes after entering the atmosphere, the contribution of various industries to the environmental concentration of pollutants is highly nonlinear and constantly changing. That is, emissions from MEIC are not equal to ground monitoring values.

The emission data are divided into five sectors: industry, power, residential, transportation and agriculture. Annual grid emission inventories of 0.25°, 0.5° and 1.0° spatial resolutions are provided. Currently, MEIC data are widely used by more than 100 research institutions and business units to forecast air quality and plan air pollution standards36,37.

Herein, in our research, two mathematical statistical methods, including the average industrial emission of pollutants (IEP) and growth rate, were used to evaluate spatiotemporal variations in IEP in 19 UAs. The formulas of these indexes are specified as follows:

$$ AI_{x,j} = \frac{1}{n}\sum\limits_{i}^{n} {IEP_{i,j} } $$
$$ GI_{x,j} = \frac{{AI_{x,j}^{2017} - AI_{x,j}^{2015} }}{{AI_{x,j}^{2015} }} \times 100\% $$

where AIx,j is the average industrial emission of j pollutants in city x, IEPi,j is the estimated IEP of pixel i of j pollutants, and GIx,j stands for the growth rate of IEP of j pollutants. To further explore the spatial and temporal variations in UAs, the spatial changes in the industrial emissions of pollutants at the UA scale were calculated by averaging the AI value in the inner city. Specific experiments were completed by zonal statistics function on ArcGIS 10.6 platform. Zoning statistical tools are also widely used in scientific researches, which can summarize the values of raster within the zones of another dataset (either raster or vector). The explanation link for this tool is as follows:

Besides, Mann–Kendall trend test and spatial centroid calculation were used to explore the spatiotemporal variation patterns of UAs on pollutant concentration.

Results and discussion

Daily change in primary pollutants

To elucidate the change trend of primary pollutants under the 13th Five-Year Plan, we calculated the daily primary pollutants in 2015 and 2019 based on formula (1) and formula (2). Such diurnal comparisons can reduce the effects of seasonal weather to some extent. From the 19 UAs (224 prefecture-level cities), the heat diagram of the daily change transfer matrix of primary pollutants from 2015 to 2019 is shown in Fig. 2, including six primary pollutants and clean day conditions.

Figure 2
figure 2

Transfer change matrix heatmap of primary pollutants from 2015 to 2019.

From the sum of the diagonal numbers, 37% of the primary pollutants had no shift during the 13th Five-Year Plan period. PM2.5, PM10 and O3 were the main primary pollutants, especially PM2.5. More primary pollutants were diverted to ozone pollution, indicating that the proportion of O3 as the primary pollutant is gradually increasing. In addition, the proportion of clean air has increased significantly, which shows that pollution control has been effectively reflected during the 13th Five-Year Plan period. However, the proportion of NO2 before and after metastasis was approximately the same, with approximately 5% NO2 pollution. This may imply that the governance of NO2 pollution was rendered nonsignificant. It is noteworthy that ozone pollution in China has become an increasingly prominent task in recent years. Similar to Xiao’s16 research on ozone pollution, they argue that present-day ozone levels in major Chinese cities are comparable to or even higher than the 1980 levels in the United States. Taken together, ozone and PM2.5 have become the top two air pollution pollutants in China.

Monthly distribution of primary pollutants

To further explore the spatiotemporal distribution of the primary pollutants across the UAs, we obtained the most primary pollutants per month by dividing the number of days with the most pollutants by the number of cities in each UA from the 2019 data. In Fig. 3, the UAs location was plotted on the abscissa, and the monthly variance of the primary pollutant was plotted on the ordinate. As shown in Fig. 3, PM2.5 appeared as dark green, PM10 appeared as light green, O3 appeared as orange, NO2 appeared as yellow, and clean days appear as dark blue. The main pollutants in the 19 UAs are PM2.5, PM10 and O3. NO2, as the primary pollutant, only appeared in the HBOY UA in January. Ordos, located in HBOY, possess rich oil and coal resources, with coal mining as its leading industry38. According to the China Energy Statistical Yearbook 2019, nearly 250 million tons of raw coal were used for thermal power generation in Inner Mongolia Autonomous Region, making it the region with the largest amount of raw coal for thermal power generation in China39. To a certain extent, the increase of heating40 and the imperfect denitration technology41 are both contributing to the increase of NO2 pollution in the atmosphere. CO and SO2 did not become major pollutants. Clean days (where AQI < 50) occurred mainly in South and Southwest China during summer. As the primary pollutant, PM2.5 is mainly concentrated in December, January and February, that is, it occurs in winter. As the primary pollutant, PM10 is mainly concentrated in October–November and March–April, that is, autumn and spring. As the primary pollutant, O3 is mainly concentrated from May to September, that is, summer. These results are in accord with recent studies indicating that they showed a strong seasonality but there were small differences6,42,43.

Figure 3
figure 3

Monthly calendar of primary pollutants in UAs in 2019.

From the perspective of UAs, monthly primary pollutants in different UAs are different44. The O3 pollution in the PRD and BG, as representatives of UAs in southern China, mainly concentrated from August to November. In addition to human factors, this may be associated with lower latitudes, higher temperatures and stronger solar radiation19,45. The northern, northwestern, northeastern, central and eastern UAs have similar monthly primary pollutant distributions, with PM2.5, PM10 and O3 pollution distributed in the 12 months of 2019. Near surface O3 is mainly produced by the complex photochemical reactions between nitrogen oxides, volatile organic compounds, carbon monoxide, methane and other precursors19. O3 was a major pollutant in most UAs from May to September, which may be the result of the comprehensive action of anthropogenic factors and meteorological conditions.

Annual patterns of ambient air pollutants

As shown in Fig. 4, unary linear regression was used to extract the time variation trend of PM2.5, PM10 and O3 concentrations in each UA. The slope value depicted a decreasing trend in the annual mean PM2.5 and PM10 during 2015–2019, suggesting that the emissions of particulate matter were effectively controlled. Meanwhile, ozone pollution has become a new environmental challenge in most UAs. The non-parametric Mann–Kendall trend test and Sen’s slope estimator (MKTT-SSE) confirmed these findings (Table 3).

Figure 4
figure 4

Annual variation trends (slope values) of three major pollutants in each UA of China from 2015 to 2019. This figure was made in the ArcGIS 10.6 platform.

Table 3 MKTT-SSE for annual mean concentration of air pollutants in 19 UAs from 2015 to 2019.

In more details in UAs, the average annual PM2.5 and PM10 concentrations significantly decreased, mostly in the BTH, SP, PRD and CC UAs, which are economically developed regions in China46,47 (Fig. 4, Table 3). Among them, BTH region had the largest reduction of PM2.5 concentrations in the 10th, 12th and 13th FYP period48. In contrast, the annual mean ozone pollution displayed the greatest enhancement while PM2.5 and PM10 showed a minimal decrease in the TJ UA in northern China. Regarding MKTT-SSE, annual mean concentration of SO2 decreased statistically significant (P < 0.05) in most UAs, especially in TJ, which affiliated Shanxi Province, with nearly 12 μg/m3 per year during the 13th Five-Year Plan period. As reported by People's Government of Shanxi Province, the main measures to the reduction the concentration of SO2 include the designation of "no-coal areas", raising the standard for eliminating excess capacity and increasing railway freight49. However, the concentration of NO2 was still increasing with 3.10 μg/m3 per year, so it is imminent to adjust the industrial structure and control the discharge of pollutants. Regional emergency linkage can effectively deal with the heavily polluted weather. For CO pollution, it decreased statistically significant (P < 0.05) with the slope of 0.07 mg/m3 per year in all UAs (Table 4).

Table 4 Recommended AQG levels and interim targets.

At the same time, by checking the annual average AQI in 2019, we showed that the CP UA has the highest annual average AQI, and the particulate matter pollution and ozone are both high, suggesting that the control effect is not obvious in the central region. We found that the annual average AQI showed a gradually decreasing trend outward, taking the CP UA as the center. In addition, the UAs in the southern and southwestern regions showed both good annual average AQI and low fine particulate pollution while high ozone concentration pollution. The average annual AQI, fine particulate matter and ozone pollution in the TNS are also at a high level in 19 UAs. This is consistent with the level of the monthly primary pollutants mentioned above.

The World Health Organization (WHO) published new Global Air Quality Guidelines (AQGs), which recommend new air quality levels on 22 September 2021. As shown in Table 4and Fig. 5, almost half of the UAs have reached the second-level national standards of China and the first-stage interim target of the World Health Organization.

Figure 5
figure 5

Numerical distribution of three major pollutant concentrations in UAs in 2019.

Hourly variations in PM2.5, PM10 and O3

As seen from the above analysis, the spatial distribution of changes and the concentrations of air pollutants shared a location similarity between adjacent UAs. Furthermore, to profile the diurnal changes in pollutant concentrations during the 13th Five-Year Plan, the 24-hourly variations were calculated in seven directions. The results are shown in Fig. 6. From 2015 to 2019, the rates of O3 concentration in any hour significantly increased compared to particulate matter (PM) pollution, which was consistent with the average annual variation trends above. Overall, diurnal trends in O3 concentration at UAs in all directions also showed a single peak, with the largest increments concentrating between 13:00 and 16:00, due to more frequent heat waves in recent times50 and the fact that higher temperature, lower relative humidity, and stronger solar radiation at 16:00 favor more secondary pollutant production9,51. A similar hourly variation, which daytime O3 increased more significantly than that of nighttime in Beijing, has been reported5. However, the peak O3 pollution spatially varied, exhibiting a relatively earlier trend in the southeastern region than in the northwestern region because of the Earth's rotation, making the eastern area receive direct sunlight first. The distance between the whiskers of the box plot in the southern city is the largest compared with the others, indicating that ground-level ozone concentration variations presented great dissimilarity in southern cities.

Figure 6
figure 6

Box plots of hourly three major pollutant rates across UAs.

For PM pollution, there was a significant downward trend in all UAs. PM2.5 reduction is more obvious in rush hours (10:00 and 22:00), which reveals the response to aggressive actions by the government to restrict motor vehicles51. In addition, we noticed that the ‘coal-to-gas and coal-to-electricity’ transformation has substantially improved air pollution, particularly in northern China. It seems clear from these figures that the changing rate of PM2.5 in northern regions was significantly higher than that of PM10.

Spatial centroid variations of air pollutants

As far as pollution centroid are concerned, their trends indicate that the ambient air pollutants are gradually moving southward and mainly concentrated in the Central Plains UA from 2015 to 2019 (Fig. 7). We noticed that the centroids of PM2.5 and PM10 showed similar movement trends, which were both firstly located in the north of the Zhumadian city in Henan Province, and following to the southwest sharply about 30 km, and then slightly shifted towards the southeast, and finally located in the south of the starting centroid. In other words, PM pollution have moved towards the south during China's 13th Five-Year Plan period, indicating that the north/northeast/northwest UAs have dropped significantly than that in the southern UAs. Consistent with the results of the Shi et al. (2019), our results show that the PM governance in China is indeed effective52 (Fig. 7a).

Figure 7
figure 7

Moving path of the centroids of six air pollutants. (We marked the centroid of PM2.5 pollution with circle, PM10 with star, O3 with square, CO with triangle_up, NO2 with triangle_down and SO2 with hexagon.). This figure was generated with python 3.7.6 based on geopandas and contextily packages. The source of the base map from contextily package is ‘http://{s}{z}/{x}/{y}.jpg’.

With respect to the SO2 and NO2 pollution, they were offset southwest by nearly 100 and 50 km, respectively (Fig. 7b). These two pollutants have the largest offset, which may suggest that the decline of SO2 and NO2 pollution in northeast is greater than that of other pollutants. These findings have also been evidenced by Zhang et al. (2019)49 and Cui et al. (2021)53.

With regard to the O3 pollution, the centroid spatial movement trend was weak and mainly located in the north of Xinyang city in Henan Province. During the 13th Five-Year Plan period, the control of ozone pollution is not obvious. Therefore, ozone pollution become the focus of the 14th Five-Year Plan exceeded PM2.554. In term of CO pollution, it moved sharply south in 2017, indicating that CO pollution in the north has been effectively controlled since 2017 (Fig. 7c).

Sector emission source impacts on primary pollutants

Due to the availability of data, statistics on industry emissions are limited to 2015–2017. The average annual emission value of agricultural sources was too small to provide statistical significance. Figure 8 only shows the contribution of four sector sources (industry, power, residential and transportation) to pollutants in 19 UAs from 2015 to 2017, where the dotted line represents the average of total emissions. As a result, near-surface ozone pollution is a secondary pollutant produced by the photochemical reaction of a series of precursors under the action of solar radiation. In particular, VOCs and NOx are important precursors for ozone pollution19. Therefore, we explore the emissions of these two important precursors in different UAs.

Figure 8
figure 8

Average sector emission of pollutants in 19 UAs.

Sector pollution sources directly cause increases in PM2.5, PM10, nitrogen oxides, volatile organic compounds and other pollutants in the atmosphere55. As seen from Fig. 8a, PM2.5 is mainly derived from industrial sources, and the main high value areas are TJ, YRD and SP UAs. Meanwhile, in the three UAs, the total contribution of anthropogenic emission sources to particulate pollution declined the fastest, which is largely related to government intervention, industrial transformation and environmental governance. Approximately 50–75% of PM2.5 in the northern UAs comes from industrial sources, while PM2.5 in the southwestern UAs mainly comes from residential sources. This result is similar to the analysis of PM2.5 sources in BTH in 2013 by Li et al. Regarding PM10, industrial sources have the greatest influence on PM10 in all UAs (Fig. 8b). There is no significant regional difference in the impact of traffic sources on PM2.5 and PM10. The contribution of traffic sources to PM2.5 in 19 UAs is 5–7% on average, and PM10 tends to 0. By using the receptor model method, Huang et al.56 found that the contribution of gasoline dust to total suspended particulate matter (TSP) in Changzhou was less than 1%, confirming that exhaust gas was not the main contribution source of PM2.5. In each UA, the cumulative emission of PM2.5 is consisted of the four average emissions caused by anthropogenic sources, which is approximately 2.5 times of that of PM10 (Fig. 8a,b). Here the emissions of pollutants from different sources were calculated, and the cumulative emission were all decreasing. However, we did not use the atmospheric transmission model to simulate pollutant concentration, so different from Zhang’s research57 on PRD, we could not get how various control measures and policies affect the monitoring concentration of different urban agglomerations.

Among the 19 UAs, NOx and VOCs mainly come from industrial and traffic sources, with less contribution from residential and power sources (Fig. 8c,d). During the study period, the NOx emissions of most UAs decreased, but the CY, CG and CC UAs in the southwest, WSS in the eastern and BG UAs all showed an increasing trend of NOx. This is mainly concentrated in the increase in industrial sources, which can also be seen from the increase in industrial production in the south. The contribution rate of power sources to VOCs in UAs is low and can be ignored.

From the perspective of time series changes, the contribution rate of anthropogenic emission sources to particulate pollution is decreasing, but for gas pollutants, nearly half of UAs show an increasing trend of VOCs and NOX. In addition, VOCs high value areas are mainly concentrated in the YRD, PRD, BTH and other state-level UAs. In regional joint prevention and control, different UAs have different priorities to prevent and control sector emission of pollutants.

Limitations of this study

As mentioned in the literature reviews, air pollution is not only affected by social activities and atmospheric emissions, but also by the impacts of meteorological factors6,44. In previous studies, we have specially studied the impact of meteorological elements on the YRD UA and found that the winds blowing to YRD (southeasterly & northwesterly) have opposite effects on air quality. This study only visualized the emission source inventory of pollutants at urban agglomeration scale. A number of work remain to be done in future research. Meteorological factors (temperature, wind speed, wind direction and etc.58,59) and chemistry transport model (such as integrated source apportionment method60,61,62) were not taken into account, so the contribution of anthropogenic sources to air pollutant concentration and trans-regional transmission of pollution were not realized in this study. Thus, it is necessary to further explore the impact of the combination of social and meteorological activities on air pollution.


Other studies conducted on the temporal and spatial variations in air primary pollutants in 19 UAs during the 13th Five-Year Plan are not sufficient. Based on air quality monitoring data and using mathematical statistics and cartography, this paper analyzed the spatiotemporal characteristics of primary pollutants over 19 Chinese UAs. Although we have achieved remarkable results in pollution control during the 13th Five-Year Plan period, PM and O3 pollution are still serious problems in China. Returning to the question posed at the beginning of this study, it is now possible to state that the main conclusions are as follows:

  1. (1)

    Generally, from the daily changes of all primary pollutants, 37% of the primary pollutants had no shift during the 13th Five-Year Plan period. PM2.5, PM10 and O3 were the main primary pollutants, especially PM2.5. The proportion of clean air days has increased significantly. NO2 pollution was rendered nonsignificant.

  1. (2)

    From the perspective of monthly distribution of primary pollutants, clean days (where AQI < 50) occur mainly in South and Southwest China during summer. As the primary pollutant, PM2.5 is primarily concentrated in December, January and February, and PM10 is primarily concentrated in October–November and March–April. The O3 pollution in the PRD and BG is mainly concentrated from August to November. The northern, northwestern, northeastern, central and eastern UAs have similar monthly primary pollutant distributions.

  1. (3)

    In terms of air quality differences in UAs, the annual average AQI showed a gradually decreasing trend outward, taking the Central Plain as the center. The Central Plain has the highest annual average AQI, and the particulate matter pollution and ozone are both high, suggesting that the control effect is not obvious in the central region. The average annual PM2.5 and PM10 concentrations decreased, most significantly in HC and MSL in the northeast, while there was little difference in annual ozone concentration. In contrast, the annual mean ozone pollution displayed the greatest enhancement, while PM2.5 and PM10 showed a minimal decrease in TJ UA in northern China. BTH UA had the largest reduction of PM2.5 concentration and the ambient air pollutants are gradually moving southward and mainly concentrated in the Central Plain from 2015 to 2019.

  1. (4)

    There were significant differences in the intensity and direction of each sector emission source in nineteen UAs. The total amount of PM discharged from the emission sources of Taiyuan-Jinzhong, Yangtze River Delta and Shandong Peninsular was much higher than those of others. PM2.5 and PM10 are mainly derived from industrial sources. The contribution rate of anthropogenic emission sources to particulate pollution is decreasing, but for gas pollutants, nearly half of UAs show an increasing trend of VOCs and NOX.

Based on the above findings, this paper proposes the following policy recommendations. First, it is necessary to pay more attention to emission reduction efforts in national UAs. Second, PM and ozone pollution are important pollution types in UAs in China, and they should be coordinated in prevention and control. The government should strengthen regional joint prevention and control and establish a compensation and incentive mechanism for cities with weak economies. Third, different UAs should formulate differentiated PM and O3 emission reduction strategies according to their own social and economic conditions. Industrial emissions are still an important source of air pollution, and the industrial structure should be optimized based on local resources. We will encourage energy enterprises to develop energy and introduce new energy technologies. We will improve cleaner production technology and build a monitoring platform for urban greening.