Introduction

Air pollution is a worldwide environmental issue which can lead to significant ecological and environmental effects and threaten human health, especially the atmospheric fine particulate matters (PM2.5) in many Asian cities1,2,3,4. Typically in China, associated with the rapid industrialization and urbanization and massive energy consumption, air quality was deteriorated5,6 and haze pollution has frequently occurred7,8. In 2016, 75.1% cities in China exceeded the annual ambient air quality guidelines, and PM2.5 was the primary pollutant during most pollution days9. Specifically, the population-weighted mean PM2.5 in Chinese cities was 61 μg m−3, three times higher than the global mean10. Therefore, air pollution control polices and measures have also been gradually emphasized and strengthened, for which understanding the spatial and temporal distribution of air pollutants and related sources was the key step11,12,13. However, the PM2.5 distribution was significantly influenced by local terrain features, meteorological conditions, city characteristics and economic levels, local emission sources and regional pollution transport14,15,16,17,18. For instance, the PM2.5 concentrations in 20 monitoring sites of California, USA change daily and such change was cyclical and changed with the season in the corresponding cycle19. Therefore, it is desired to study the large-scale and long-term spatio-temporal distribution of PM2.5 and corresponding mechanisms. Nowadays, there were some high-resolution air pollution estimation methods, including interpolation method20 and the satellite top-down approach based on ground-based fixed-site air pollution monitoring networks21,22,23.

In this study, the 5-year high-resolution distributions of PM2.5 concentrations in a province with 13 cities in eastern China were investigated with air pollution sources and land uses by Geographic Information System (GIS) and backward trajectory clustering analysis. The main objectives were: (1) to explore the spatial and temporal distribution characteristics of PM2.5 in different geographical areas and under varied environmental managements; and (2) to illuminate the influence mechanisms of air pollution source, economic, topographic, and meteorological factors on PM2.5 pollution patterns.

Results and Discussions

Spatio-temporal distributions of the 5-year provincial air PM2.5 pollution

Intra-annual and inter-annual variations of PM2.5 in the overall province

The large-scale and long-term PM2.5 pollution was significant in Jiangsu. Compared with the Chinese Ambient Air Quality Standards (CAAQS) (Table A1), the average annual PM2.5 concentration of each city during 2013–2017 exceeded the Grade II guideline (Fig. A1). Intra-annually, the PM2.5 levels showed significant seasonal variations (Table 1), which was a U-shaped pattern of high in winter and low in summer24 (Fig. 1). According to the proportions of days with different PM2.5 levels (Fig. A2e), there were 57 days in 2017 exceeded the national Grade II standard, up to 42 days of which were in winter. The PM2.5 concentrations in winter and summer of 2017 were 72.5 and 32.3 μg m−3, respectively (Fig. 2). These seasonal phenomena were attributed to pollution sources and meteorological conditions. In winter, anthropogenic emissions related to heating demand were increased, and a combination of persistent temperature inversions and low mixed boundary layer was unfavorable to the atmospheric pollutant dispersion25,26,27,28. While in summer, influenced by the monsoon climate, the precipitation was high and could significantly reduce the particulates in atmosphere, and the prevailing east wind and the clean air from ocean has also dilution effects on the air pollution in Jiangsu province.

Table 1 Significant differences of PM2.5 levels among different seasons in 2013–2017.
Full size table
Figure 1
figure 1

Daily and annual variations of PM2.5 in Jiangsu province, China from 2013 to 2017.

Full size image
Figure 2
figure 2

Spatial distribution of PM2.5 concentrations in summer (a) and winter (b) of 2017 for Jiangsu province.

Full size image

Inter-annually, the average annual concentration of the overall Jiangsu province has decreased significantly from 2013 to 2017 (Fig. 1), which was 71.8, 66.3, 57.7, 50.3, and 49.6 μg m−3, respectively. The proportion of pollution days has decreased simultaneously, that was 34.2%, 29.5%, 22%, 17.1%, and 15.6%, respectively. These air quality improvements were closely related to the environmental protection measures taken by the provincial government in recent years. It can be seen from Fig. 3 that the pollutants in various regions of Jiangsu Province have been reduced to varying degrees from 2013 to 2017, indicating that the government’s environmental protection measures were effective. As a key area for environmental governance, the pollution concentration of provincial capital Nanjing decreased significantly, while the air quality improvement in the northern Jiangsu was slight, which mainly due to the local heavy industry.

Figure 3
figure 3

Differences of annual PM2.5 concentrations between 2013 (a) and 2017 (b) in the provincial distribution.

Full size image

Spatial distribution of the provincial PM2.5

Using the average PM2.5 values at all 110 monitoring sites in 2013 and 2017, the Kriging interpolation analysis was performed to obtain the provincial simulated distribution map of PM2.5 concentrations (Fig. 3). In the whole studied area, the PM2.5 levels decreased gradually from west to east. According to the geographical location, urbanization level, population density, civilian car ownership, total GDP and per capita GDP, all cities of this province can be divided into 4 regions (Fig. 4), including western inland, eastern coastal, northern heavy industry and south developed regions. The industrial enterprises of XZ (Xuzhou) and HA (Huai’an) were dominated by heavy industries including mining, so they were divided into northern heavy industrial area. The specific features of each city with PM2.5 concentrations were showed in Table 2. The PM2.5 level was in the order of heavy industrial > inland > developed > coastal regions. It showed a pattern that the PM2.5 pollution in densely populated and developed areas with more emission sources was higher than those with better vegetation coverage and self-purification capacity28,29. Besides the effects of urbanization, the industry and geographic location also showed significant impacts. Figure 2 showed that the pollution level in developed areas was lower than that in the underdeveloped northern regions with heavy industrial pollution emissions. Impacted by the ocean atmospheric transmission, the pollution in coastal areas with stronger self-purification capabilities was lower than that in the inland regions.

Figure 4
figure 4

Locations of Jiangsu province in China and the 110 scattered monitored sites (solid triangles) covering all 13 cities.

Full size image
Table 2 Regional divisions of Jiangsu province and the corresponding PM2.5 concentrations in 2017.
Full size table

Because the economic development between southern and northern regions of Jiangsu province was seriously unbalanced (GDP in Table 2), to balance the huge gap, developments of heavy industry in the northern region led by machinery, electricity, chemistry, building materials and energy were vigorously promoted by governments. For instance, of the 221 key enterprises with exhaust emissions in the 13 cities of Jiangsu Province in 2018, there were up to 30 in a north heavy industry city (XZ). Moreover, in order to accelerate the process of urbanization in the northern region, the municipal construction was in the peak period, corresponding dusts were substantial and aggravated air quality.

Of course, compared with the satellite-based top-down approach based ground fixed-site air pollution monitoring networks, Kriging interpolation method has limitations. Remote sensing provides the detailed information in space and time not only from accessible areas but also from inaccessible areas30. Satellite-retrieved aerosol optical depth has been increasingly utilized for the mapping of fine particulate matter concentrations23,31. Due to the lack of actual landform data, the spatial distribution of PM2.5 was mainly simulated by interpolation. The future further study about satellite mapping of fine particulates should be based on geographically weighted regression.

Effects of source emissions on spatial and temporal PM2.5 distributions

The sources of urban PM2.5 were complex, including exhaust of motor vehicles, emissions of power plants and industrial boilers, combustions of household coal and biomass, open waste incineration and the dusts32,33,34. Clarifying the contributions of main pollutant sources would be beneficial to the environmental managements and air pollution control measures.

Table 3 showed the detailed air pollutant emissions of Jiangsu from 2011 to 2015, that the emission of soot was increased by 127 thousand tons, while SO2 and NOx reduced by 219 thousand and 468 thousand tons. Because NOx and SO2 were the two key air pollutants for producing secondary PM2.5, they were of great significance to the decrease of PM2.5 pollution in this study area35. According to the structure of exhaust emissions, the percentage of industrial SO2 and NOx emissions were decreased due to vigorously promoting clean energy and renovating coal fired boilers. In 2015, the total emissions of soot, NOx, and SO2 in Jiangsu province were 654, 1068, and 835 thousand tons, of which industrial emissions contributed 612, 754, and 795 thousand tons and accounted for 93.6%, 70.6%, and 95.2% of corresponding total emissions, respectively. In addition to industrial emissions, vehicle exhaust contributed significantly to NOx emissions, accounting for 28.6%. As to the regional distribution, the anthropogenic emissions of soot, SO2 and NOx at the heavy industrial city were higher than coastal city. To effectively control the urban air pollution, reductions of local coal, vehicular and industrial emissions, and the regional joint pollution prevention and control policies are necessary.

Table 3 Annual detail emission information of air pollution sources in Jiangsu from 2011 to 2015.
Full size table

Regional air pollutant transport

Besides local pollution sources, the typical monsoon climate plays an important role in the long distance transport of aerosols18,36. The typical life cycle of aerosols in the atmosphere was 3–10 days. To explore the impact of inter-regional air transport on pollutant concentrations, 72 h backward trajectory cluster and PSCF analysis were conducted for four typical regions in Jiangsu province.

Figure 5 compared the cluster analyses of them in summer and winter. Air pollution was found to be affected by the monsoon climate. In summer, the clustering percentage of air transport from nearby regions accounted for 65.9%, 59.1%, 73.9%, and 63.6%, respectively. Long-range air mass transport clusters account for a less proportion, and the directions were mainly from the southeast coast and the southern region. In winter, the air masses mainly come from the inland areas of the northwest, representing long-distance air mass transport, the transmission distance and vertical span were significantly larger than those in summer. During the process of winter air pollutants transmission, the study areas were impacted by the northwest polluted air masses originated from Inner Mongolia and along Shanxi, Hebei, Henan, Anhui and Shandong provinces37. The trajectory of the airflow was faster, indicated that the air mass rises to the free troposphere at the source, and moved to the downstream at a faster speed under the action of the wind speed. After reaching the research area, it entered the boundary layer after vertical mixing and sinking. In addition, pollution level was influenced by the monsoon climate and Siberia high pressure, the downdraft and stably stratified atmosphere help to increase PM2.5 pollution level in these cities38,39. These results of backward trajectory clustering analysis implied the significant effects of regional dispersion and meteorological conditions on regional air quality.

Figure 5
figure 5

The 72-h backward trajectories clustering for four representative cities (XZ, northern heavy industrial city; NT, eastern coastal city; TZ, inland city; and NJ, developed city) during summer and winter.

Full size image

PSCF analysis was applied to explore the potential source region distribution of PM2.5. A PSCF analysis of PM2.5 combined with atmospheric data based on the backward trajectory for four representative cities in 2017 was showed in Fig. 6. The spatial patterns were similar, and the source areas with higher contribution rate were mainly distributed in neighboring provinces such as Anhui, Shandong, Zhejiang provinces, etc., indicating the significance of pollution in the nearby area37. In addition, due to the long-range air mass transport, the contribution values of Inner Mongolia, Hebei, Gansu, Ningxia, Guangdong and Fujian provinces reached 0.5–0.8, which increased the pollutant concentrations in the study area40. It should be noted that the PSCF analysis did not estimate the spatial distribution of all sources. The high potential contribution source area may coincide with the regional emission sources, but the low contribution value does not necessarily indicate low emissions in the area.

Figure 6
figure 6

The 72-h backward trajectories clustering and PSCF for four representative cities (XZ, northern heavy industrial city; NT, eastern coastal city; TZ, inland city; and NJ, developed city) during 2017 based on PM2.5 concentration.

Full size image

Conclusions

Air PM2.5 pollution was still a significant environmental issue with varied spatial and temporal distributions. Typically in eastern China, 5-year data of 110 monitoring sites from all the 13 cities of a province were compared by statistics and illustrated by GIS. According to the intra-annual and inter-annual variations of PM2.5 in large-scale and long-term, the PM2.5 pollution contributed by industrial and traffic emissions was still obvious, but the level has reduced significantly in recent 5 years owing to the strengthened pollution control measures. PM2.5 concentrations were significantly higher in winter than summer, and were in an order of heavy industrial area > inland area > developed area > coastal area, due to different emission sources and meteorological conditions. Besides local primary emissions and secondary aerosol formation, the backward trajectory clustering analysis and potential source contribution factor (PSCF) analysis indicated that the typical monsoon climate played an important role in the aerosol transport. In summer, it was mainly affected by the updraft from near region, while in winter, the long-distance transport from the northwest had a significant impact on air pollution. Moreover, the nearby provinces were important potential pollution sources.

Data and Methods

Study areas

Jiangsu Province is an important part of the developed Yangtze River Delta located along the eastern coast of China (Fig. 4). It covers an area of 103 thousand km2 and has a population of 8.03 million. With Yellow Sea to its east, Jiangsu adjoins Anhui and Shandong provinces in the west and north respectively, with Zhejiang province and Shanghai city as neighbors in the southeast. It has a large area of plain as typical topography and dotted with two top-5 largest lakes in China. Situated in a transition belt from a subtropical to temperate zone, this province has a typical monsoon climate. Approximately demarcated by the Huai River, the south is subtropical monsoon climate and the north is warm moist monsoon climate. Generally, it is mild with moderate rainfall and four distinct seasons. Its economy is dominated by industrial activities with a GDP of 859 billion yuan (top 2 in China), and coal is the main energy source with an annual consumption of 258 million tons. Moreover, the number of motor vehicles reached 17.3 million in 2015, and 1232 polluting enterprises were monitored41. As an industry and transportation dense region, it was the highest PM2.5 emission province (0.28 million tons) in the year of 201442,43. Associated with the economic developments, the problem of air pollution has become a key environmental issue that regulatory policies and pollution control measures were implemented and strengthened by recent years44,45. Therefore, it provided an ideal area to study the profound characteristics of air pollution in a large-space scale and long-term scale.

To characterize the air PM2.5 pollution patterns and mechanisms, this study analyzed air pollution and related factor data from all the 13 cities of Jiangsu Province, including Nanjing (NJ), Suzhou (SZ), Wuxi (WX), Changzhou (CZ), Nantong (NT), Xuzhou (XZ), and Lianyungang (LYG), Zhenjiang (ZJ), Huai’an (HA), Yancheng (YC), Taizhou (TZ), Suqian (SQ) and Yangzhou (YZ). Figure 4 showed the location of study area and air monitoring sites.

Data sources

Based on the monitoring site information provided by the China Environmental Protection Monitoring Bureau, 110 national air quality monitoring stations in Jiangsu Province were used for spatial study (Table A2). Daily average of 24 h PM2.5 concentration and real-time concentration data (i.e., hourly average) from 18 January, 2013 to 31 December, 2017 were collected for intra-annual and inter-annual variation analyses. The meteorological data used for backward trajectory analysis was from the simultaneous global data assimilation system (GDAS) provided by the National Center for Environmental Prediction (NCEP) of USA, including temperature, relative humidity, surface precipitation, horizontal and vertical wind speeds. Data of pollution sources, emissions and economic development were obtained from the Statistical Yearbook of Jiangsu Province41. The numbers of companies with emissions were from the Self-Monitoring Information Release Platform of Focused Enterprises in Jiangsu Province (http://218.94.78.61:8080/newPub/web/home.htm).

The GIS spatial analysis by Kriging interpolation method

The GIS is a comprehensive subject combining geography and cartography, remote sensing and computer science. GIS technology integrates the unique visual effect and geo-analysis function of maps with general database operations46, and has been widely applied in environmental science47,48,49. Interpolation method is an important content of spatial statistical analysis in GIS. Among the various interpolation methods, the Kriging interpolation is flexible and can fully utilize the data exploratory analysis tools to improve the efficiency of spatial analysis effectively20. It can use the statistical characteristics of known samples to quantify the spatial autocorrelation between measurement points, highlighting the overall distribution trend, increasing the data fidelity, and has the highest prediction accuracy for normal data50. In this study, the Kriging interpolation method was applied to explore the spatial distribution of PM2.5 concentration data of 110 stations during 2013 and 2017, thus reveal the PM2.5 pollution patterns in the overall study area intuitively.

Long-range air mass transportation by trajectory calculation

As a long-range source analysis technique based on observational difference or simulated meteorological field51,52, the 72-h backward air trajectories clustering and PSCF analysis were calculated using the HYSPLIT model (http://ready.arl.noaa.gov/HYSPLIT.php). The meteorological data as 1° × 1°GDAS data, and the trajectory calculation points were NJ (Nanjing, 32.06N, 118.78E, 1000 m), NT (Nantong, 31.99N, 120.88E, 1000 m), XZ (Xuzhou, 34.25N, 117.21E, 1000 m) and TZ (Taizhou, 32.49N, 119.90E, 1000 m) were used in the trajectory calculation, and two trajectory were calculated every day at 0:00 and 12:00 (UTC), starting at: 2017.1.1 0:00 to 2018.1.1 0:00.

PSCF was a trajectory-based gridded statistical analysis method that can obtain the spatial distribution of pollution sources in a semi-quantitative manner40. The basic concept of PSCF was to combine the air mass trajectory with the atmospheric component data to generate a conditional probability in a given region which was divided into i*j small grids. The conditional probability was combined with the air mass trajectory to describe the possible spatial distribution of geographic source locations. The PSCFij value of the ij grid point was

$$PSC{F}_{ij=\frac{{m}_{ij}}{{n}_{ij}}=\frac{thenumberoftrajectoryendpoints{\rm{reaching}}{\rm{the}}{\rm{pollution}}{\rm{threshold}}{\rm{in}}{\rm{the}}{\rm{ij}}{\rm{grid}}{\rm{points}}}{thenumberofendpointsintheijgridpoints}}$$

In order to reduce the uncertainty caused by the small nij value, the weight function wij was introduced, and when the number of trajectory endpoints in the ij grid points was less than about three times the average number of trajectory endpoints, wij decreased the PSCFij value. The PSCF method was widely used in atmospheric chemistry research, such as analysis of particulate matter in the atmosphere and potential source distribution of inorganic components in the particles. It was generally considered that when PSCF ≤ 0.5, it was a low contribution source region, and conversely, a high contribution source region. The area analyzed in this paper was located in 85–135°E, 15–60°N, and the grid resolution was 0.5° × 0.5°, Weight function

$${\rm{wij}}=\{\begin{array}{cc}1.00\, & 12 < {n}_{ij}\\ 0.70\, & 6 < {n}_{ij}\le 12\\ 0.42\, & 4 < {n}_{ij}\le \,6\\ 0.17\, & {n}_{ij}\le 4,\end{array}$$

the pollution threshold was set to the primary standard of PM2.5 concentration (35 μg.m−3).