Atmospheric Pollution Research 8 (2017) 1151e1159

HOSTED BY Contents lists available at ScienceDirect

Atmospheric Pollution Research

journal homepage: http://www.journals.elsevier.com/locate/apr

Spatio-temporal variation and influence factors of PM2.5 concentrations in from 1998 to 2014

* * Debin Lu a, b, , Jianhua Xu a, , Dongyang Yang a, Jianan Zhao a a School of Geographic Sciences, Normal University, , 200241, China b Department of Tourism and Geography, Tongren University, Tongren, Guizhou Province, 554300, China article info abstract

Article history: Based on the remote sensing retrieval of PM2.5 concentration data in the long-time series, both the linear Received 17 March 2017 regression and grey system correlation analysis methods were employed to analyze the spatial and Received in revised form temporal pattern, variation trend and the main influencing factors of PM2.5 concentration in China from 15 May 2017 1998 to 2014. The results showed that only 16.21%e24.67% of the land area in China PM concentrations Accepted 21 May 2017 2.5 reached the annual average criterion value of 10 mg/m3 set by the World Health Organization (WHO) in Available online 27 May 2017 3 1998e2014; the PM2.5 concentrations were greater than 95 mg/m mainly in Xinjiang Taklimakan Desert, west of and the central region of . PM concentration was less than 10 mg/m3 mainly in Keywords: 2.5 Tibet, western , northeastern Yunnan, Taiwan, northern Xinjiang, northern and PM2.5 pollution Spatio-temporal variation northwest of . High PM2.5 concentration in the northwest of China was mainly affected by Grey correlation analysis sand and dust, while it was mainly caused by human activities in the eastern region. Except for Taiwan, Influence factors low PM2.5 concentration areas were mainly located in the economically backward regions. The positive China indicators in highly correlation with PM2.5 concentration include the average temperature, the propor- tion of primary and secondary industry to GDP, industrial consumption, the proportion of fulfilled amount of investment in real estate development to GDP, SO2 emissions and population density. The negative indicators in highly correlation with PM2.5 concentration include the average precipitation, the average wind velocity, the proportion of the tertiary industry to GDP, and the greening coverage rate of the built-up areas. © 2017 Turkish National Committee for Air Pollution Research and Control. Production and hosting by Elsevier B.V. All rights reserved.

1. Introduction has also become a social concern and the focus of public opinion and academic research. . With the rapid development of economy and urbanization in At present, the ground observation (Wang et al., 2014a,b; Zhang China, a series of environmental problems have been brought and Cao, 2015; Sun et al., 2016) and satellite remote sensing about, especially the pollution of PM2.5 (particulate matter with retrieval of data (Han et al., 2015; Zheng et al., 2016; Lin et al., 2013; aerodynamic2.5 mg/m3) (Geng et al., 2015; Tian et al., 2014; Wang Peng et al., 2016) are applied for the study of PM2.5 concentration. et al., 2014a,b) has seriously affected the human health (Guan et al., The accuracy and time resolution of ground observation data are 2014) and socio-economic development (Xie et al., 2016). It has high, but the data from monitoring stations often cannot reflect the become a serious obstacle to the sustainable development of Chi- spatial variability of PM2.5 concentration in the large area, espe- nese cities, forcing the government to take control measures such cially in the mountainous areas with complex topography and as traffic restrictions (Wang et al., 2015), class suspension for stu- highly populated cities with complicated underlying surface. The dents, factory production limitation, or even stopping production. It mean value of the data from the decentralized monitoring stations as the PM2.5 concentration in the area may lead to systematic errors in assessing population exposure and health effects (Lin et al., * Corresponding authors. School of Geographic Sciences, East China Normal 2016). Compared with the ground observation, the satellite University, Shanghai, 200241, China. remote sensing can quickly obtain PM2.5 concentration of contin- E-mail addresses: [email protected] (D. Lu), [email protected] (J. Xu). uous surface in the large area, especially it has the vital significance Peer review under responsibility of Turkish National Committee for Air Pollu- tion Research and Control. for understanding the PM2.5 pollution level in areas where the http://dx.doi.org/10.1016/j.apr.2017.05.005 1309-1042/© 2017 Turkish National Committee for Air Pollution Research and Control. Production and hosting by Elsevier B.V. All rights reserved. 1152 D. Lu et al. / Atmospheric Pollution Research 8 (2017) 1151e1159 monitoring stations have not been established. By the end of 2012, sources, urbanization and industrial structure, corporate pollution the national air pollution monitoring network has been established, control and technological progress, according to data availability but few monitoring stations, short construction time and lack of principle as well as its pollution sources and influencing fac- long-term observation data have been shown. Therefore, the sat- tors(He et al., 2016). Meteorological conditions include precipi- ellite remote sensing retrieval of data is still an effective means to tation, average wind velocity, average temperature and average evaluate the PM2.5 pollution level in the large area (Ma et al., 2015), relative humidity to reflect the influence on PM2.5 mass concen- especially for the variation trend analysis and evaluation of the tration; PM2.5 pollution source indicators include industrial elec- PM2.5 concentration, it is indispensable. The research on PM2.5 tricity consumption, proportion of fulfilled amount of investment concentration both at home and abroad mainly focused on element in real estate development to GDP, sulfur dioxide emissions, composition (Shaltout et al., 2013), source analysis (Huang et al., smoke (dust) emissions, the total volume of annual bus passenger 2014; Jansen et al., 2014), health effect (Song et al., 2016), pollu- traffic, total road freight and public transport vehicles, reflecting tion control (Zhang et al., 2013) and so on. Although some scholars the industrial production process and traffic-generated road dust, used the annual average or multi-year average to study PM2.5 exhaust emissions. Urbanization and industrial structure include pollution pattern, there are still relatively fewer studies and anal- population density, area of built-up districts, per capita GDP, ysis on PM2.5 concentration pattern, variation trend and influencing proportions of primary, secondary and tertiary industry to GDP, factors, because the PM2.5 concentration data are difficult for total green area and greening coverage rate of built-up area, to acquisition in a large scale for a long term or the continuous reflect the influence of urbanization process and socio-economic monitoring has not been performed for the data at all. PM2.5 activities on PM2.5 concentration. Corporate pollution control pollution process is a very complex system process, the pollution and technological progress include SO2 removal, smoke (dust) source diversity is remarkable, which contains a large number of removal and proportion of R&D expenditure to GDP, reflecting the known and unknown information state, and it is a typical grey intensity of corporate emission reduction and pollution control system (He et al., 2016). technology. Therefore, based on the remote sensing retrieval of PM2.5 concentration data in the long-time series, data extraction and partition calculation, the spatial pattern of PM2.5 concentration in 2.2. PM2.5 concentrations data China from 1998 to 2014 was analyzed. The linear regression method was employed to reveal the variation trend of PM2.5 Currently, there is a lack of publically available global remote concentration. Finally, the grey correlation analysis method was sensing data relating to PM2.5. van Donkelaar et al., (2006, 2010, fi adopted to analyze the correlation between PM2.5 concentration 2015) estimate ground-level ne particulate matter (PM2.5)by and influencing indicators. Through the above methods, the main combining Aerosol Optical Depth (AOD) retrievals from the NASA factors influencing PM2.5 concentration change can be found in MODIS, MISR, and SeaWIFS instruments with the GEOS-Chem order to make targeted control and improve urban air quality, chemical transport model, and subsequently calibrated to global providing a scientific basis for China to choose a scientificpathof ground-based observations of PM2.5 using Geographically new urbanization development and make policy on PM2.5 pollu- Weighted Regression (GWR), The global PM2.5 concentration tion control. dataset had a spatial resolution of 10 km from 1998 to 2014. It is by far the most accurate PM2.5 remote-sensing data set with the largest coverage and longest time span that is available, it has been 2. Index system construction and data sources validated and can be effectively applied on a national scale (Lee et al., 2012; de Sherbinin et al., 2014). The dataset can be directly 2.1. PM2.5 concentration impact index system construction download from Atmospheric Composition Analysis Group at Dal- housie University (http://fizz.phys.dal.ca/~atmos/martin/?page_ PM2.5 is mainly originated from coal combustion, motor id¼140). We used a subset of global PM2.5 concentration dataset vehicle exhaust, road dust and biomass combustion (Cao, 2014), that covered China from 1998 to 2014 extracted from.nc format by and its formation and changes are also affected by meteorological ArcGIS Soft. conditions and other factors. Therefore, 22 indexes (Table 1)were Although, van Donkelaar et al. has validated the accuracy of the selected from four aspects of meteorological conditions, pollution global annual PM2.5 concentrations grids dataset based on the

Table 1

Influence index system of PM2.5 concentration in China.

þ 2 Meteorological conditions Average relative humidity (1%) þ (x1) Urbanization and Population density (person/km )(x12) þ 2 Precipitation (mm) (x2)industrial structure Area of built-up districts (km )(x13) þ Average wind velocity (m/s) (x3) Per capita GDP (x14) þ Proportion of primary industry to GDP (x15) þ þ Average temperature ( C) (x4) Proportion of secondary industry to GDP (x16) PM2.5 pollution sources Total volume of annual bus passenger Proportion of tertiary industry to GDP (x17) þ traffic (10,000 person) (x5) þ 2 Total road freight (t) (x6) Total green area (km )(x18) þ Public transport vehicles (x7) þ Industrial electricity consumption Greening coverage rate of built-up area(x19) (100 million kwh)

(x8) Proportion of fulfilled amount of Corporate pollution control SO2 removal (t) (x20) investment in real and technological progress þ estate development to GDP (x9) þ SO2 emissions (t) (x10) Smoke (dust) dust removal (t) (x21) þ Smoke (dust) emissions (t) (x11) Proportion of R&D expenditure to GDP (x22) Note: þ denotes the positive indicators and the rest are negative indicators. D. Lu et al. / Atmospheric Pollution Research 8 (2017) 1151e1159 1153 agreement between satellite-derived estimates and ground-based P P P n $ 1 n n measurements (van Donkelaar et al., 2010). However, considering i PM2:5i i PM2:5i Slope i¼1 n i¼1 i¼1 (1) ¼ P P 2 that the relationship between AOD-PM2.5 can differ by space and n 2 1 n i¼1i n i¼1i countries, it was still necessary to evaluate the reliability of the dataset for the specific area of China (Luo et al., 2017). The Ministry In which PM2.5 is the grid unit PM2.5 concentrations, n is the > of Environmental Protection of China (MEP) published the PM2.5 time span, and i is the time unit. slope 0, meaning that PM2.5 concentration over ground observation sites from 2012. We increased over time; slope<0, meaning that PM2.5 decreased over fl collected Chinese ground-based PM2.5 measurements of 68 sample time; The size of slope could re ect the speed of increase or points in 2013 year and spatial distribution of these sample points decrease. F test was used to conduct significance test. The increased are shown in Fig.1 (a). Furthermore, according to the sample points, or decreased trend was significant if the slope p < 0.05, and if < fi the corresponding satellite-derived values of PM2.5 concentrations p 0.01, it would be highly signi cant. data were calculated. Also, the linear correlation between “PM2.5 Ground-based Values” and “PM Satellite-derived Values” is 2.5 3.2. Grey correlation analysis method shown in Fig. 1(b). A significant overall agreement is found (R2 ¼ 0.817), which indicated that the satellite-derived PM con- 2.5 The grey correlation analysis method is proposed by Ju-Long, centration data were reliable. (1982), the calculation steps go as follows: First, specify the reference sequence x (t) and compared 2.3. Socioeconomic data 0 sequence xi(t), t represents point or time, they are given by the followed formula respectively: In order to ensure the continuity of data, the data of PM2.5 pollution sources, and urbanization and industrial structure in- x0ðtÞ¼fx0ð1Þ; x0ð2Þ; /; x0ðnÞg; t ¼ 1; 2; /; m (2) dicators from 274 cities in China during 2003e2014 were collected on a prefecture-level city scale. The data are sourced from China x ðtÞ¼fx ð1Þ; x ð2Þ; /; x ðnÞg; i ¼ 1; 2; /; n (3) City Statistical Yearbook (2003e2015) (http://tongji.cnki.net/ i i i i kns55/Navi/NaviDefault.aspx). Public transport vehicles ¼ Buses where m is the number of point or time, n is the number of operated at the end of the year þ Number of taxis at the end of the compared sequence. year. Given that there may be different dimension between reference sequence and compared sequences, the standardization processing 2.4. Meteorological data is need before going to the next step. Essentially, correlation coefficient is the geometrical differences The meteorological data are derived from China Meteorological between curves. Hence, the size of differences becomes the mea- Data Network (http://data.cma.cn), including four indicators of surement scale of correlation. Correlation coefficient calculate average daily temperature, cumulative precipitation, average wind correlation coefficient and correlation degree. Correlation coeffi- velocity and average relative humidity. cient is calculated by the followed formula: r 3. Methods min minjx0ðkÞ xiðkÞj þ max maxjx0ðkÞ xiðkÞj i k i k xiðkÞ¼ r (4) jx0ðkÞxiðkÞj þ max maxjx0ðkÞxiðkÞj 3.1. Unitary linear regression analysis i k

Unitary linear regression analysis was used to examine the where, xiðkÞ is the relative difference between reference sequence variation trend of PM2.5, and the fitting slope could reflect the (or curve) and compared sequence (or curve), which was regarded variation trend. The formula is: as the correlation coefficient between xi and x0 at point k (or time

Fig. 1. (a) The ground-based PM2.5 measurements in 2013 collected from urban air quality monitoring data undertaken by China's National Environmental Monitoring Centre and (b) linear correlation of the two datasets (PM2.5 ground-based values and satellite-derived values). 1154 D. Lu et al. / Atmospheric Pollution Research 8 (2017) 1151e1159 k). r is resolution coefficient, whose range is 0e1. Generally, it is 0.5 4.2. Spatial distribution characteristics of PM2.5 concentrations in use. Correlation degree is the average value of correlation coefficient, Fig. 3 shows the spatial distribution of the maximum, minimum calculated by the followed formula: and average of PM2.5 concentration in China from 1998 to 2014. From the maximum figure, the regions with high PM2.5 concen- 1 Xn tration were seriously polluted areas, mainly in the Taklimakan g ¼ x ðkÞ (5) i n i Desert of Xinjiang and the -Tianjin-Hebei Economic Zone in k¼1 3 . PM2.5 concentration was greater than 95 mg/m in the Xinjiang Taklimakan Desert, west of Tianjin and the central Hebei. where g is the correlation degree between reference sequence x (t) i 0 Although the minimum concentration of PM in these areas did and compared sequence x (t). g . The bigger theg , the stronger the 2.5 i i i not exceed 95 mg/m3, it was also higher than that in other areas, impact of factor and the more significant the contribution, and vice ranging from 35 to 75 mg/m3. In the maximum figure, 10 mg/m3 and versa. Generally, the correlation degree is low if 01 24.67% reached the annual average criterion value of 10 mg/m3 by for severe increase. As shown in Fig. 4, slope values were the World Health Organization. between 1.1 and 2.2 in the past 17 years, and the variation trend of PM2.5 concentration in China exhibited significant spatial distri- bution difference, with slope>0 accounting for 76.86% of the total area in North China Plain, River Delta region, , southwest region, Xinjiang Taklimakan Desert and , while slope<0 accounting for 23.14% of the total area in the Loess Plateau, northeast of Heilongjiang and Inner Mongolia, the southeast coastal areas and Hainan. PM2.5 concentration presented a rapid growth in Xinjiang Taklimakan Desert, Qaidam Basin and the eastern part of the North China Plain. The spatial distribution and area of different growth types are as show Fig. 4 and Table 2. The increasing trend of PM2.5 concentration from 1998 to 2014 was much greater than the decreasing trend. The F-test method was employed to examine the significant change of PM2.5 concentration (P < 0.05). As shown in Fig. 5, the change trend of PM2.5 value was statistically significant in only

Fig. 2. Area proportion of China under different PM2.5 concentrations from 1998 to 27.3% of the land area, and the increasing trend of PM2.5 concen- 2014. tration was subject to significance testing in Qinghai, Tibet, D. Lu et al. / Atmospheric Pollution Research 8 (2017) 1151e1159 1155

Fig. 3. Spatial distribution of the maximum, minimum and mean values of PM2.5 concentrations in China from 1998 to 2014.

Xinjiang, , , , Hebei, Beijing, southwest of regions as the same in the eastern developed areas to adopt low- Yunnan and the central region of Guizhou, with the decreasing carbon and environmentally friendly economic development trend in Hainan, Fujian, Ningxia, Loess Plateau at junction of model to prevent the deterioration of PM2.5 pollution wherein. In Shaanxi and Inner Mongolia. Except for Hebei, Shandong, Beijing addition, compared with other regions, the North China Plain was a and Tianjin in the eastern North China Plain, the PM2.5 concentra- typical high-value region of PM2.5 concentration in China, and the tion increase was significant in the mostly central and western analysis results showed a significant upward trend. Therefore, the regions, indicating that the rising trend of PM2.5 concentrations in PM2.5 pollution control efforts shall be further strengthened. It can the central and western parts of China had a statistical significance. also be seen from the above analysis, the rising PM2.5 concentration Government policy-makers should be alert to this phenomenon, trend was shown on a regional, or even national scale. Thus, control and take emphasis on PM2.5 pollution in the central and western policies shall be developed not only for a particular city but also 1156 D. Lu et al. / Atmospheric Pollution Research 8 (2017) 1151e1159

Fig. 4. Trend of PM2.5 concentrations in the China during 1998e2014.

Table 2 Proportion of provinces in China under different slope classification (%).

Slope 1 1 < Slope 0.5 0.5 < Slope 00< Slope 0.5 0.5 < Slope 1 Slope > 1

China 0.07 1.07 22.08 58.09 12.69 6 Beijing ee e e 0.15 0.03 Tianjin ee e e e 0.06 Hebei ee 0.08 0.70 0.78 0.48 ee 0.11 0.41 1.07 0.09 Shandong ee 0.01 0.23 0.65 0.72 Inner Mongolia e 0.43 8.36 4.66 ee Heilongjiang e 0.01 3.19 2.41 0.01 e ee 0.51 1.69 0.02 e ee 0.67 0.95 ee ee 0.01 0.80 0.20 e Shanghai ee e 0.07 ee ee 0.48 0.49 ee ee 0.49 0.90 0.02 e Fujian ee 1.12 0.04 ee ee 0.63 0.96 ee Hubei ee 0.17 1.38 0.27 e ee e 1.74 0.16 e ee 0.75 0.87 ee ee e 2.06 ee Hainan ee 0.30 ee Guizhou ee 0.05 1.61 ee Yunnan ee 0.06 3.41 0.11 e ee e 0.79 0.01 e Sichuan e 0.02 0.53 4.04 0.14 e Tibet ee 0.45 11.04 0.07 e Shaanxi 0.06 0.11 0.50 0.57 0.93 0.01 ee 0.06 0.87 0.80 e e 0.15 1.11 2.63 0.43 e Ningxia 0.01 0.36 0.11 ee Qinghai ee e 3.74 3.17 0.50 Xinjiang ee 1.82 8.65 3.71 4.11 Taiwan ee 0.28 0.05 ee D. Lu et al. / Atmospheric Pollution Research 8 (2017) 1151e1159 1157

Fig. 5. Significance of PM2.5concentrations in the China during 1998e2014. from the regional or even national perspective. population density, greening coverage rate of the built-up areas, per capita GDP, area of built-up districts and total green area.

4.4. Correlation analysis of PM2.5 concentrations and influence Construction and emissions from freight automobiles are the main index sources of dust emissions from transportation. Industrial electricity consumption and the proportion of secondary industry to GDP

There are many intricate factors influencing PM2.5 concentra- represent the scale and volume of industrial production. Emissions tion. According to the index system constructed in Table 1, the grey of industrial exhaust gas and smoke (dust) are one of the main correlation between PM2.5 concentration and influence index was sources of PM2.5 pollution, while the green area can inhibit in calculated. The correlation (gi) values between 22 indexes and reducing the concentration of PM2.5. The correlation of corporate PM2.5 concentration are shown in Table 3, and it can be seen that pollution control and technological progress indexes is exhibited their correlation values are larger at moderate or strong level, fully below from high to low: the proportion of R&D expenditure to GDP, indicating the evaluation factors have an important impact on the SO2 removal and removal of industrial dust. PM2.5 concentration change, and also showing the selection of each In order to avoid the selection of a large area, resulting in the index factor was relatively reasonable. The meteorological condi- instability of system analysis, the national seven regions were tion is the external cause to the formation of PM2.5 pollution, and subject to grey correlation calculation according to the traditional the correlation of indexes from high to low is as follows: the Chinese regional zoning (Luo et al., 2017), with the results as shown average precipitation, average wind velocity, average air tempera- in Table 3. From Table 3, it can be seen that the total precipitation, ture and average relative humidity. The research has shown that the average wind velocity, the proportions of primary, secondary fi the precipitation has the effect of flushing on dust particles in PM2.5 and tertiary industry to GDP, the proportion of ful lled amount of (Cao, 2014), the wind velocity has a significant effect on the diffu- investment in real estate development to GDP, population density, sion of PM2.5 (Han et al., 2016), the temperature has a catalytic SO2 emission and greening coverage rate of built-up areas have effect on the chemical reactions of PM2.5 precursor pollutants strong correlation with PM2.5 concentration in seven regions. Per (Zhang and Cao, 2015), and high humidity conditions will promote capita GDP and relative humidity are also highly correlated with the formation of nitrate and secondary organic aerosols (Wang PM2.5 concentration in the other six regions except for the north- et al., 2016). Excessive emission is the root cause of PM2.5 pollu- west region. There are differences in correlation between the tion. The correlation of PM2.5 pollution source indexes is listed remaining indexes and PM2.5 concentration in different regions, below from high to low: the proportion of fulfilled amount of in- fully showing that the meteorological conditions, PM2.5 pollution vestment in real estate development to GDP, SO2 emission, indus- sources, urbanization process, efforts of corporate pollution emis- trial electricity consumption, public transport vehicles, the total sion control, technology, etc. have different impact on PM2.5 con- volume of annual bus passenger traffic, total road freight and centration in different regions. smoke (dust) emissions; the correlation of urbanization and in- The results of the above analysis show that the grey correlation dustrial structure indexes is shown as follows from high to low: the model analysis method can effectively analyze the main index proportion of tertiary industry to GDP, the proportion of secondary factors that influence the spatial distribution of PM2.5 concentra- industry to GDP, the proportion of primary industry to GDP, tion. The external causes, such as the total precipitation and 1158 D. Lu et al. / Atmospheric Pollution Research 8 (2017) 1151e1159

average wind velocity are the main influencing factors of PM2.5 concentration. The internal causes, such as proportions of primary, secondary and tertiary industry to GDP, the proportion of fulfilled amount of investment in real estate development to GDP, popula- tion density, SO2 emission and greening coverage rate of the built- up areas are the main influencing factors of PM2.5 concentration. Therefore, the regional joint prevention and control (Liu et al., 2015), rational control of the urban population size, adjustment of the industrial structure, reduction of corporate pollution emis- sions, improvement of urban greening coverage rate of built-up areas and technological level should be emphasized in the treat- ment and control of PM2.5 concentration.

5. Discussion

Social and economic development is a key factor in the deteri- oration of air quality, and meteorological conditions have a signif- icant effect on its formation, distribution, maintenance and change. A broad consensus has been reached, and many studies have confirmed this (Zhang et al., 2013; Lin and Wang, 2016; Luo et al., 2017). From the analysis results of this study, the growth trend of PM2.5 concentration in China was regional and even national, while a significant growth trend was exhibited in the North China Plain and the central & western regions. This result was consistent with the statistics by employing monitoring site data of Greenpeace (http://www.greenpeace.org.cn/city-ranking-2016-q1/). PM2.5 data of remote sensing inversion were adopted herein, so that the time series were longer and more persuasive. The PM2.5 concentration was significantly increased in the North China Plain and the central & western regions, because Beijing-Tianjin-Hebei, Yangtze River Delta and the were key areas in China's action plan of air pollution prevention and control, and more strict PM2.5 governance goals have been set up. On the other hand, many re- strictions imposed by the central government on haze in the east would lead to the transfer of some polluting enterprises to the central and western regions that were not subject to heavy pollu- tion. The results of grey relational analysis also showed stronger correlation between PM2.5 concentration and the secondary in- dustry proportion of GDP in southwest, northwest and central China than in and east China. In the traditional heavy industry base in north and , the proportion of the 0.58 0.58 0.55 0.58 0.59 0.57 0.54 0.58 0.62 0.60 0.65 0.60secondary 0.67 industry 0.60 has been 0.64 relatively 0.66 large. As can be seen from Table 3, except for the average wind velocity and population den- sity, the grey correlation values are basically the same in the southwest and northwest regions, but the PM2.5 concentration in the northwest region is obviously higher than that in the southwest region, indicating a considerable portion from the dust aerosol in the northwest region. Through the observation data analysis of 88 monitoring stations in Xinjiang from 2000 to 2011, the Taklimakan lled amount D expenditure to GDP 0.58 0.64 0.67 0.60 0.57 0.55 0.60 0.53 fi Desert and the south source had a high incidence of sand and dust c & fi weather, and the number of annual dusty days in southern Xinjiang was 2.7 times that of northern Xinjiang (-An et al., 2013), so it was China's most serious sandstorm area. As the economy of South emissionsremoval 0.62 0.6 0.57 0.61 0.58 0.69 0.63 0.56 0.60 0.68 0.69 0.58 0.71 0.54 0.61 0.55 uence index Whole country South China Central China East China North China Northeast China 2 2 China, Central China, East China, North China and Northeast China fl Smoke (dust) emissionsPopulation densityArea of built-up districtsPer capita GDPTotal green areaProportion of secondary industry toProportion GDP of tertiary industry toTotal GDP green area 0.76 0.57Greening coverage rate of built-upSO area 0.77 0.58Smoke (dust) 0.66 dust removal 0.63Proportion of R 0.74 0.55 0.60 0.71 0.68 0.58 0.8 0.75 0.57 0.58 0.74 0.57 0.65 0.69 0.77 0.60 0.66 0.74 0.57 0.56 0.66 0.67 0.58 0.7 0.77 0.58 0.64 0.62 0.58 0.59 0.72 0.62 0.71 0.61 0.6 0.74 0.74 0.72 0.68 0.68 0.62 0.58 0.64 0.63 0.6 0.71 0.62 0.68 0.72 0.53 0.59 0.61 0.7 0.57 0.80 0.61 0.72 0.64 0.74 0.64 0.60 0.56 0.72 0.62 0.60 0.63 0.72 0.75 0.54 0.56 0.61 0.62 0.59 0.52 in PrecipitationAverage wind velocityAverage temperaturepassenger traf Total road freightPublic transport vehiclesIndustrial electricity consumptionProportion of ful 0.77 0.67 0.60 0.77 0.59 0.65 0.57 0.62 0.57 0.77 0.58 0.58 0.75 0.62 0.58 0.72 0.60 0.53 0.72 0.72 0.61 0.73 0.59 0.74 0.58 0.62 0.61 0.71 0.59 0.67 0.67 0.63 0.57 0.65 0.70 0.56 0.59 0.60 0.77 0.62 0.55 0.70 0.71 0.66 0.62 0.68 0.61 0.52 of investment in real estate development to GDP SO was relatively developed in eastern and central regions, the overall concentration of PM2.5 in South China was lower than the above areas. From the industrial structure, the proportion of secondary industry to GDP in North China, Central China and Northeast China was highly correlated, followed by East China, with the minimum in South China, while the grey correlation of SO2 and smoke (dust) removal in South China was higher than that in the other areas mentioned above, indicating that the proportion decrease of sec- ondary industry and emission reduction by the corporate could concentrations and impact indicators factor correlation.

industrial structure and technological progress fi 2.5 signi cantly reduce the regional PM2.5 concentration. Urbanization and Corporate pollution control PM2.5 pollution sources Total volume of annual bus Meteorological conditions Average relative humidity 0.61 0.58 0.66 0.70 0.62 0.66 0.67 0.67

Table 3 PM The above analysis results also showed that the different regions D. Lu et al. / Atmospheric Pollution Research 8 (2017) 1151e1159 1159 were affected by the complex factors such as the regional resources concentrations in China using satellite-based aerosol optical depth and a e and environment, and the influence direction and intensity of social chemical transport model. Remote Sens. Environ. 166, 262 270. Guan, D.B., Su, X., Zhang, Q., et al., 2014. The socioeconomic drivers of China's & economic development and meteorological conditions on PM2.5 primary PM2.5 emissions. Environ. Res. Lett. 9, 1e9. were also different significantly. Although the current economic Han, L., Zhou, W., Li, W., 2015. Increasing impact of urban fine particles (PM2.5)on areas surrounding Chinese cities. Sci. Rep. 5. development in China will increase the concentration of PM2.5, our Han, L., Zhou, W., Li, W., 2016. Fine particulate (PM2.5) dynamics during rapid ur- discussion on the means to curb the deterioration of air quality, banization in Beijing, 1973e2013. Sci. Rep. 6. does not mean to fully curb economic development. Therefore, He, X., Lin, Z.H., Liu, H.Y., et al., 2016. Analysis of the driving factors of PM2.5 in through the comparative analysis of socio-economic factors and Jiangsu province based on grey correlation model. J. Geogr. 71 (7), 1119e1129. Huang, R.J., Zhang, Y., Bozzetti, C., et al., 2014. High secondary aerosol contribution PM2.5, the regional differences and the common development to particulate pollution during haze events in China. Nature 514 (7521), model of population, social economy and ecological environment 218e222. should be taken into account comprehensively in the development Jansen, R.C., Shi, Y., Chen, J., et al., 2014. Using hourly measurements to explore the role of secondary inorganic aerosol in PM2.5 during haze and fog in , of PM2.5 emission reduction policy on the basis of regional devel- China. Adv. Atmos. Sci. 31 (6), 1427e1434. opment status and resource environment, to make a reasonable Ju-Long, D., 1982. Control problems of grey systems. Syst. Control Lett. 1 (5), control of urban population size, adjust the industrial structure, 288e294. reduce enterprise pollution emissions and increase the green Lee, S.J., Serre, M.L., van Donkelaar, A., et al., 2012. Comparison of geostatistical interpolation and remote sensing techniques for estimating long-term exposure coverage in urban built-up areas. In the process of pollution man- to ambient PM2.5 concentrations across the continental United States. Environ. agement and control, the regional joint prevention and control health Perspect. 120 (12), 1727. should be emphasized (Liu et al., 2015), to minimize the negative Lin, X., Wang, D., 2016. Spatiotemporal evolution of urban air quality and socio- economic driving forces in China. J. Geogr. Sci. 26 (11), 1533e1549. impact caused by the economic development. Lin, G., Fu, J., Jiang, D., et al., 2013. Spatio-temporal variation of PM2.5 concentrations and their relationship with geographic and socioeconomic factors in China. Int. e 6. Conclusions J. Environ. Res. public health 11 (1), 173 186. Lin, C., Li, Y., Lau, A.K.H., et al., 2016. Estimation of long-term population exposure to PM2.5 for dense urban areas using 1-km MODIS data. Remote Sens. Environ. 179, Based on the remote sensing retrieval of PM2.5 concentration 13e22. data in the long-time series, both the linear regression and grey Liu, Y., Li, Y., Chen, C., 2015. Pollution: build on success in China. Nature 517 (7533), 145e145. system correlation analysis methods were employed to analyze the Luo, J., Du, P., Samat, A., et al., 2017. Spatiotemporal pattern of PM2.5 concentrations spatial and temporal pattern, variation trend and the main influ- in mainland China and analysis of its influencing factors using geographically weighted regression. Sci. Rep. 7. encing factors of PM2.5 concentration in China from 1998 to 2014. Ma, Z.W., Hu, X.F., Sayer, A.M., et al., 2015. Satellite-based spatiotemporal trends in PM2.5 concentrations: China, 2004e2013. Environ. Health Perspect. (1) 61.95%e78.93% of the land area reached the annual average Peng, J., Chen, S., Lü, H., et al., 2016. Spatiotemporal patterns of remotely sensed 3 limit of level II (35 mg/m ) according to the ambient air PM2.5 concentration in China from 1999 to 2011. Remote Sens. Environ. 174, e quality standard (GB3095-2012), and only 16.21%e24.67% 109 121. Shaltout, A.A., Boman, J., Al-Malawi, D.R., et al., 2013. Elemental composition of m 3 reached the annual average criterion value of 10 g/m by the PM2.5 particles sampled in industrial and residential areas of Taif, Saudi Arabia. World Health Organization from 1998 to 2014. Aerosol Air Qual. Res. 13 (4), 1356e1364. (2) PM concentration was greater than 95 mg/m3 in the Xin- Song, Y., Wang, X., Maher, B.A., et al., 2016. The spatial-temporal characteristics and 2.5 health impacts of ambient fine particulate matter in China. J. Clean. Prod. 112, jiang Taklimakan Desert, west of Tianjin and the central 1312e1318. 3 “ ” Hebei. PM2.5 concentration was less than 10 mg/m mainly in Sun, Y., Wang, Z., Wild, O., et al., 2016. APEC blue : secondary aerosol reductions Tibet, western Sichuan, northeastern Yunnan, Taiwan, from emission controls in beijing. Sci. Rep. 6. Tian, S., Pan, Y., Liu, Z., et al., 2014. Size-resolved aerosol chemical analysis of northern Xinjiang, northern Inner Mongolia and northwest extreme haze pollution events during early 2013in urban Beijing, China. of Heilongjiang. J. Hazard. Mater. 279, 452e460. Van Donkelaar, A., Martin, R.V., Park, R.J., 2006. Estimating ground-level PM2.5 (3) The areas presenting a rising trend of PM2.5 concentrations using aerosol optical depth determined from satellite remote sensing. were much larger than those presenting a decrease trend. J. Geophys. Res. Atmos. 111 (D21). PM2.5 concentration exhibited a significant rising trend in van Donkelaar, A., et al., 2010. Global estimates of ambient fine particulate matter central and western regions as well as North China Plain. concentrations from satellite-based aerosol optical depth: develope and application. Environ. Health Perspect. 118, 847e855. (4) The positive indicators in highly correlation with PM2.5 van Donkelaar, A., Martin, R., Brauer, M., Boys, B., 2015. Use of satellite observation concentration include the average temperature, the propor- for long-term exposure assessment of global concentration of fine particulate tion of primary and secondary industry to GDP, industrial matter. Environ. Health Perspect. 123, 135e143. consumption, the proportion of fulfilled amount of invest- Wang, Y., Yao, L., Wang, L., et al., 2014. Mechanism for the formation of the January 2013 heavy haze pollution episode over central and eastern China. Sci. China ment in real estate development to GDP, SO2 emissions and Earth Sci. 57, 14e25. population density. The negative indicators in highly corre- Wang, Y., Ying, Q., Hu, J., et al., 2014. Spatial and temporal variations of six criteria e lation with PM concentration include the average precip- air pollutants in 31 provincial capital cities in China during 2013 2014. Environ. 2.5 Int. 73, 413e422. itation, the average wind velocity, the proportion of the Wang, Z.B., Fang, C.L., Xu, G., Pan, Y.P., 2015. Spatial-temporal characteristics of the tertiary industry to GDP, and the greening coverage rate of PM2.5 in China in 2014. J. Geogr. 70 (11), 1720e1734. the built-up areas. Wang, G., Zhang, R., Gomez, M.E., et al., 2016. Persistent sulfate formation from London Fog to Chinese haze[J]. Proc. Natl. Acad. Sci. 113 (48), 13630e13635. Xie, Y., Dai, H., Dong, H., et al., 2016. Economic impacts from PM2.5 pollution-related Acknowledgments: health effects in China: a provincial-level analysis. Environ. Sci. Technol. 50 (9), 4836e4843. Yuan-An, J., Ying, C., Yi-Zhou, Z., et al., 2013. Analysis on changes of basic climatic This work was supported by the Guizhou Province Science and elements and extreme events in Xinjiang, China during 1961e2010. Adv. Clim. Technology fund (LKT[2012]07;25). Change Res. 4 (1), 20e29. Zhang, Y.L., Cao, F., 2015. Fine particulate matter (PM2.5) in China at a city level. Sci. Rep. 5. References Zhang, X.Y., Sun, J.Y., Wang, Y.Q., et al., 2013. Factors contributing to haze and fog in China (in Chinese). Chin. Sci. Bull. Chin. Ver. 58, 1178e1187. Cao, J., 2014. J. PM2.5 and Environment. Science Press, Beijing. Zheng, Y., Zhang, Q., Liu, Y., et al., 2016. Estimating ground-level PM2.5 concentra- de Sherbinin, A., Levy, M.A., Zell, E., et al., 2014. Using satellite data to develop tions over three in China using satellite-derived aerosol optical environmental indicators. Environ. Res. Lett. 9 (8), 084013. depth measurements. Atmos. Environ. 124, 232e242. Geng, G., Zhang, Q., Martin, R.V., et al., 2015. Estimating long-term PM2.5