Chemical Compositions and Source Analysis of PM2.5 during Autumn and Winter in a Heavily Polluted City in China

As one of the biggest cities in North China, Jinan has been suffering heavy air pollution in recent decades. To better characterize the ambient particulate matter in Jinan during heavy pollution periods, we collected daily PM2.5 (particulate matter with aerodynamic diameters equal to or less than 2.5 μm) filter samples from 15 October 2017 to 31 January 2018 and analyzed their chemical compositions (including inorganic water-soluble ions (WSIs), carbonaceous species, and inorganic elements). The daily average concentration of PM2.5 was 83.5 μg/m3 during the sampling period. A meteorological analysis revealed that both low wind speed and high relative humidity facilitated the occurrence of high PM2.5 pollution episodes. A chemical analysis indicated that high concentrations of water-soluble ions, carbonaceous species, and elements were observed during heavy pollution days. The major constituents of PM2.5 in Jinan were secondary aerosol particles and organic matter based on the results of mass closure. Chemical Mass Balance (CMB) was used to track possible sources and identified that nitrate, sulfate, vehicle exhaust and coal fly ash were the main contributors to PM2.5 during heavy pollution days in Jinan, accounting for 25.4%, 18.6%, 18.2%, and 13.3%, respectively.


Introduction
With the rapid development of industrialization and urbanization, the Chinese economy has progressed quickly over the last few decades. However, the economic boom has imposed tremendous pressure on the environment, especially the ambient atmosphere in Northern China [1][2][3]. The frequent severe haze episodes after the year 2010 have aroused worldwide concern. PM 2.5 (particulate matter with aerodynamic diameters equal to or less than 2.5 µm) is regarded as the main contributor to heavy air pollution [4,5]. A previous study showed that cardiovascular morbidity and mortality are associated with long-term exposure to high PM 2.5 [6]. Another study found that 3.45 million premature deaths around the world were linked to PM 2.5 pollution in 2007 [7]. PM 2.5 was the primary air pollutant in 80% of Chinese cities in 2016 [8].
To date, the chemical compositions, potential sources, and formation mechanisms of PM 2.5 have been widely investigated in economically developed and densely populated areas [9][10][11][12][13][14][15]. However, the mass concentration and chemical composition characteristics of PM 2.5 vary greatly between different

Chemical Analysis
Both the Teflon and quartz filters were prebaked before use to remove any absorbed organic material. The samples and blank filters were weighed before and after sampling at least three times using a millionth automatic balance (Commodore AWS-1, German) with a precision of 0.04 mg, and the differences in the filter weights were divided by the sampling volume in order to calculate the PM2.5 mass concentrations.
One fourth of each quartz filter sample was put into the sampling tube, and 10 mL of ultrapure water was added, ultrasonicated for 20 min, and centrifugalized for 10 min. Then, the extract was filtered with a 0.45 μm filter membrane to remove any insoluble species. Finally, the supernatant solution was taken into a clean bottle to analyze water-soluble ions. Dionex Inc., ICS-2000 was used to analyze the cations (Na + , NH4 + , K + , Mg 2+ and Ca 2+ ) and ICS-3000 for anions (Cl − , NO3 − and SO4 2− ). The detection limits of these water-soluble ions ranged from 0.005 to 0.085 μg/m 3 , and the precisions

Chemical Analysis
Both the Teflon and quartz filters were prebaked before use to remove any absorbed organic material. The samples and blank filters were weighed before and after sampling at least three times using a millionth automatic balance (Commodore AWS-1, German) with a precision of 0.04 mg, and the differences in the filter weights were divided by the sampling volume in order to calculate the PM 2.5 mass concentrations.
One fourth of each quartz filter sample was put into the sampling tube, and 10 mL of ultrapure water was added, ultrasonicated for 20 min, and centrifugalized for 10 min. Then, the extract was filtered with a 0.45 µm filter membrane to remove any insoluble species. Finally, the supernatant solution was taken into a clean bottle to analyze water-soluble ions. Dionex Inc., ICS-2000 was used to analyze the cations (Na + , NH 4 + , K + , Mg 2+ and Ca 2+ ) and ICS-3000 for anions (Cl − , NO 3 − and SO 4 2− ).
For the element analysis, half of one Teflon filter sample was extracted by hydrofluoric acid and diluted chloroazotic acid in order to analyze Na, K, Cr, Ni, Cu, Zn, As, Mo, Cd, Sn, Sb, Tl, Pb and Bi using the inductively coupled plasma-mass spectrometry analysis method (Agilent 7500 a, Agilent Technologies, CA, USA). The other half was first ashed and then extracted by solid sodium hydroxide to analyze Al, Mg, Ca, Fe, and Si using the inductively coupled plasma optical emission spectrometry method (Optima 4300 DV, PerkinElmer, MA, USA). The method detection limit (MDL) of the elements ranged from 0.02 to 30 ng m −3 , and the precisions were <10% [26,27].
Meteorological parameters, such as wind speed, temperature (T), and the relative humidity (RH) were selected from the website of the Weather Underground) [28]. Hourly SO 2 and NO 2 data were obtained from [29] 2.3. Date Analysis

Ionic Balance
Understanding the acidity of the PM is important since it is a key factor that affects the heterogeneous reactions, hygroscopic growth, and toxicity of aerosols [30,31]. Ion balance calculations were useful for studying the acid-base balance of the aerosol. The ratios of AE (anion equivalent) and CE (cation equivalent) were used to indicate the acidity of the atmospheric aerosol. The calculations of the anion and cation are as follows [30,32] of these ion compositions in the PM 2.5 samples. Particles were in an acidic condition when the AE/CE ratio was larger than 1, while a ratio lower than 1 indicated an alkaline condition [33].

Oxidation Ratio
Previous articles have proven that the oxidation degree of gaseous pollutants is reflected by the sulfur oxidation ratio (SOR) and the nitrogen oxidation ratio (NOR), respectively, to the extent that SO 2 and NO x are oxidized to form SO 4 2− and NO 3 − [34]: where n(SO 4 2− ), n(SO 2 ), n(NO 3 − ), n(NO 2 ) represent the molar concentrations of sulfate, sulfur dioxide, nitrate, and nitrogen dioxide, respectively.

Enrichment Factor
The enrichment factor (EF) method can be used to evaluate the sources of elemental compositions in PM 2.5 . The method is as follows [35][36][37]: where (C i /C R ) sample represents the ratio of the concentration of the targeted element to the concentration of a reference element in the aerosol; (C i /C R ) crust represents the ratio of the target element to a reference element in the Earth's crust [38]. Al was chosen as a reference element due to its stability in most anthropogenic contaminants in this study. EF < 10 indicates that the EFs are mainly from natural sources; EF ≥ 10 suggests that the EFs are mostly from anthropogenic sources.

Mass Reconstruction
In this study, the PM 2.5 mass was reconstructed by organic matter (OM), mineral dust, trace elements (TE), SO 4 2− , NO 3 − , NH 4 + , and EC. The OM mass was calculated by the organic carbon (OC) mass concentration multiplied by 1.6 for urban areas [39], as follows: The mineral dust composition is the sum of the mass concentration of the elemental oxides in the crust [40], as in the following equation: [Mineral] = 1.89 (Al) + 2.14 (Si) + 1.21 (K)+1.4 (Ca)+ 1.66 (Mg) + 1.7 (Ti) + 1.43 (Fe) TE is the sum of the mass concentrations of the trace elements other than crust and sea salt. The concentrations of SO 4 2− , NO 3 − , NH 4 +, and EC were directly measured [32].

Source Apportionment
Source contribution estimates (SCEs) were the main outputs of the chemical mass balance (CMB) model, which represents the fractional contribution to the ambient PM 2.5 concentration by each source profile used in the model [32]. The principle can be expressed as where C represents the total mass measured at the receptor, µg/m 3 ; S j represents the contribution from an individual source, µg/m 3 ; and j represents the number of sources. The concentration of elemental composition i, C i , will be where F ij represents the fraction of source contribution S j composed of element i, g/g; S j is the contribution from an individual source, µg/m 3 ; j represents the number of sources; and i represents the number of chemical compositions. Therefore, the contribution concentration and total contribution rate of each pollution source to the different chemical compositions in atmospheric particulate matter can be obtained as long as the composition spectrum of each pollution source and the compositional spectra of the acceptor's atmospheric particulate matter are known.

Establishment of Source Profile
Fugitive dust was collected on windowsills positioned at heights of 5-15 m in residential areas. The soil dust mainly originated from bare croplands, hilly country, dry riverbeds in the vicinity of Jinan, and exposed land within urban areas. Cement dust was collected from the roofs of residential buildings around the building site or from production lines of nearby cement factories. Coal combustion fly ash was collected from the stacks of different coal burning facilities by using a dilution system. Vehicle exhaust dust was obtained from 2 buses, 4 taxis, 3 small trucks, 2 medium-sized passenger cars, 2 large passenger vehicles, 15 small passenger vehicles, and 2 large-sized trucks. Iron making dust was obtained from the industrial furnaces of top companies (utilizing different processes) with the highest total pollutant emissions in Jinan. Biomass emission samples were collected from the biomass combustion boiler. The source profiles of secondary sulfate and nitrate used the constitutions of (NH4) 2 SO 4 and NH 4 NO 3 , respectively. All of the sampling methods and source profile constructions followed the Technical Guideline of Source Apportionment for Particulate Matter (Chinese version available [41]).

PM 2.5 Mass Concentration
The trends of PM 2.5 mass concentrations in the four different functional areas of Jinan were relatively similar during the sampling period. Therefore, the data of the four sites were pooled together for discussion to more objectively represent the pollution status of Jinan. The time-series data of the concentrations of PM 2.5 and meteorological parameters are shown in Figure 2. The mass concentration of PM 2.5 , water-soluble inorganic ions (WSIs), inorganic elements, OC, and EC at different pollution levels during the sampling period are listed in Table 2. In this study, the mass concentrations of PM 2.5 ranged from 20.7 µg/m 3 (11 January) to 237.2 µg/m 3 (17 January), with an average of 83.5 (Standard deviaton: 49.59) µg/m 3 , which is higher than the secondary limit of the National Ambient Air Quality Standard of PM 2.5 (75 µg/m 3 ). As shown in Table S1, the mass concentrations of PM 2.5 were found to be lower than those in north China, such as Lanzhou [12], Xi'an [42], Beijing [43], as well as one previous study on Jinan [19], but higher than those in south cities, such as Nanjing [44]. The trends of PM2.5 mass concentrations in the four different functional areas of Jinan were relatively similar during the sampling period. Therefore, the data of the four sites were pooled together for discussion to more objectively represent the pollution status of Jinan. The time-series data of the concentrations of PM2.5 and meteorological parameters are shown in Figure 2. The mass concentration of PM2.5, water-soluble inorganic ions (WSIs), inorganic elements, OC, and EC at different pollution levels during the sampling period are listed in Table 2. In this study, the mass concentrations of PM2.5 ranged from 20.7 μg/m 3 (11 January) to 237.2 μg/m 3 (17 January), with an average of 83.5 (Standard deviaton: 49.59) μg/m 3 , which is higher than the secondary limit of the National Ambient Air Quality Standard of PM2.5 (75 μg/m 3 ). As shown in Table S1, the mass concentrations of PM2.5 were found to be lower than those in north China, such as Lanzhou [12], Xi'an [42], Beijing [43], as well as one previous study on Jinan [19], but higher than those in south cities, such as Nanjing [44].
In this study, we divided the mass concentrations of PM2.5 into four grades according to the Technical Regulations of the Environmental Air Quality Index (AQI) (HJ 633-2012): good (≤75 μg/m 3 ), light (75-115 μg/m 3 ), moderate (115-150 μg/m 3 ), and heavy (≥150 μg/m 3 ). During the sampling period, the mass concentrations of PM2.5 during heavy pollution days was approximately 4.3, 2.1, and 1.5 times higher than those during good, light, and moderate pollution days, respectively. The average wind speed during the sampling period was 1.3 m/s, with the highest speed observed on good days and the lowest observed on heavy pollution days. In contrast, the relative humidity followed the order: heavy pollution days > moderate pollution days > light pollution days > good days. The stable atmosphere and low mixed boundary layer height, which frequently occur when the wind speed is low, are adverse to the diffusion of air pollutants and consequently lead to the accumulation of air pollutants. Furthermore, high relative humidity accelerates the formation of new particles and the growth of aerosol particles, which results in the aggravation of the pollution level in the atmosphere [45].   Figure 2. Time series of daily PM2.5 and meteorological date (blue column represents good days, yellow column represents light days, orange column represents moderate days, red column represents heavy days).

Figure 2.
Time series of daily PM 2.5 and meteorological date (blue column represents good days, yellow column represents light days, orange column represents moderate days, red column represents heavy days).  In this study, we divided the mass concentrations of PM 2.5 into four grades according to the Technical Regulations of the Environmental Air Quality Index (AQI) (HJ 633-2012): good (≤75 µg/m 3 ), light (75-115 µg/m 3 ), moderate (115-150 µg/m 3 ), and heavy (≥150 µg/m 3 ). During the sampling period, the mass concentrations of PM 2.5 during heavy pollution days was approximately 4.3, 2.1, and 1.5 times higher than those during good, light, and moderate pollution days, respectively. The average wind speed during the sampling period was 1.3 m/s, with the highest speed observed on good days and the lowest observed on heavy pollution days. In contrast, the relative humidity followed the order: heavy pollution days > moderate pollution days > light pollution days > good days. The stable atmosphere and low mixed boundary layer height, which frequently occur when the wind speed is low, are adverse to the diffusion of air pollutants and consequently lead to the accumulation of air pollutants. Furthermore, high relative humidity accelerates the formation of new particles and the growth of aerosol particles, which results in the aggravation of the pollution level in the atmosphere [45].

Water-Soluble Ions
WSIs are important components of atmospheric particles. Figure 3 shows the mass concentrations of the major chemical compositions of PM 2.5 during the sampling period. The mass concentrations of WSIs accounted for 43.5%, 50.6%, 52.1%, and 55.1% of PM 2.5 under good, light, moderate, and heavy pollution levels, respectively. The mass concentrations of the WSIs in the four pollution levels followed the same order:  [45], Shanghai [11], and Chongqing [46], but the concentrations of SO 4 2− were lower than those in the above cities. Other WSIs (Na + , K + and Cl − ) were observed to be higher in concentration during polluted days, likely due to industrial emissions, coal combustion, or biomass burning. The concentrations of Ca 2+ and Mg 2+ were higher on moderate and heavy pollution days and mainly came from soil dust, road dust, and construction dust [22,47,48].
Atmosphere 2020, 11  The relationships of SOR/NOR to temperature/RH during the sampling period are shown in Figure 4. The original sulfate and nitrate can absorb water and enlarge the surface area of the particles with an increasing RH, which favors the heterogeneous reactions of NO3 − and SO4 2− , thereby increasing the SOR and NOR [49]. However, the rising trend of SOR was more obvious than that of NOR, especially when the RH was greater than 60%. At the same time, the Spearman correlation coefficients of SOR and NOR with relative humidity were 0.7 and 0.5, respectively, which also indicated that the influence of relative humidity on SOR was greater than that on NOR. Moreover, SOR showed a weaker correlation with temperature, and NOR decreased with an increase in temperature, which is related to the volatilization of NH4NO3 at higher temperatures [50].   The relationships of SOR/NOR to temperature/RH during the sampling period are shown in Figure 4. The original sulfate and nitrate can absorb water and enlarge the surface area of the particles with an increasing RH, which favors the heterogeneous reactions of NO 3 − and SO 4 2− , thereby increasing the SOR and NOR [49]. However, the rising trend of SOR was more obvious than that of NOR, especially when the RH was greater than 60%. At the same time, the Spearman correlation coefficients of SOR and NOR with relative humidity were 0.7 and 0.5, respectively, which also indicated that the influence of relative humidity on SOR was greater than that on NOR. Moreover, SOR showed a weaker correlation with temperature, and NOR decreased with an increase in temperature, which is related to the volatilization of NH 4 NO 3 at higher temperatures [50]. The anion and cation equivalents discussed in this study have strong correlations with the four pollution levels, as well as the major ions (NH 4 + and NO 3 − + SO 4 2− ) ( Figures 5 and 6), demonstrating the reliability of the analytical results. The mean equivalent ratios of AE/CE for all pollution days was 0.9, which indicates that the collected PM was slightly alkaline, which is consistent with other studies [30,45]. On the other hand, the AE and CE were strongly correlated among the four pollution levels (good: R 2 = 0.90, light: R 2 = 0.92, moderate: R 2 = 0.91, heavy: R 2 = 0.98). Furthermore, the AE/CE ratios on moderate and heavy pollution days were 1.15 and 1.22 times higher than those on good days (good: 0.86, light: 0.96, moderate: 0.99, heavy: 1.05), which suggests that the particles on more polluted days were more acidic than those on good days; this is consistent with previous studies [51][52][53]. A high concentration of secondary inorganic ions (NO 3 − and SO 4 2− ) not only has a significant impact on the formation of pollutants but also could enhance the acidity of particles.
thereby increasing the SOR and NOR [49]. However, the rising trend of SOR was more obvious than that of NOR, especially when the RH was greater than 60%. At the same time, the Spearman correlation coefficients of SOR and NOR with relative humidity were 0.7 and 0.5, respectively, which also indicated that the influence of relative humidity on SOR was greater than that on NOR. Moreover, SOR showed a weaker correlation with temperature, and NOR decreased with an increase in temperature, which is related to the volatilization of NH4NO3 at higher temperatures [50].

Carbonaceous Aerosols
The mean concentrations of OC and EC in PM 2.5 were 11.5 and 5.0 µg/m 3 , accounting for 13.8% and 6.0% of PM 2.5 , respectively, which is higher than Shanghai [11]. The mass concentrations of OC were higher than those of EC in each polluted level ( Table 2). The proportion of OC and EC decreased from clean to polluted periods with an increase in PM 2.5 concentration (OC: 16.7% on good days and 12.5% on heavy pollution days; EC: 7.3% on good days and 5.4% on heavy pollution days), which is consistent with previous studies [54,55].
The OC/EC ratio can be used to characterize the emission and conversion characteristics of carbon aerosols [56]. In Jinan, the daily average ratio of OC/EC was 2.5, which is lower than some large cities, such as Chengdu [57], but higher than those in some industrial cities, such as Tangshan [14] and Lanzhou [58], as shown in Table S2. In this study, the daily OC/EC ratio ranged from 1.8 to 5.7, with an average of 2.3 on good days, from 1.8 to 5.5, with an average of 2.3 on light pollution days, from 1.9 to 5.6, with an average of 2.5 on moderate pollution days, and from 1.9 to 2.8, with an average of 2.6 on heavy pollution days. The OC/EC ratios were higher on heavy pollution days than on other pollution days, and the same trends were found for K + and Cl − , showing that biomass burning, fossil fuel combustion, and vehicular emissions played a significant role in the high OC/EC ratios [59,60]. The anion and cation equivalents discussed in this study have strong correlations with the four pollution levels, as well as the major ions (NH4 + and NO3 − + SO4 2− ) ( Figure 5, Figure 6), demonstrating the reliability of the analytical results. The mean equivalent ratios of AE/CE for all pollution days was 0.9, which indicates that the collected PM was slightly alkaline, which is consistent with other studies [30,45]. On the other hand, the AE and CE were strongly correlated among the four pollution levels (good: R 2 = 0.90, light: R 2 = 0.92, moderate: R 2 = 0.91, heavy: R 2 = 0.98). Furthermore, the AE/CE ratios on moderate and heavy pollution days were 1.15 and 1.22 times higher than those on good days (good: 0.86, light: 0.96, moderate: 0.99, heavy: 1.05), which suggests

Inorganic Elements
The daily average concentration of the thirty-seven elements was 7.1 µg/m 3 , accounting for 8.5% of the PM 2.5 fraction in this study ( Table 2). The crustal elements (Ca, Al, Na, Mg, K, and Fe) had a total concentration of 5.5 µg/m 3 , accounting for 77.3% of the total mass of the analyzed elements. The concentrations of the trace elements (Bi, Sn, V, Cd, Sb, Zn, Pb, Tl, Cr, Ni, As, and Cu) occupied a small part of PM 2.5 . Particularly, in these trace elements, the daily average concentration of Cr was 0.02 µg/m 3 , exceeding the Ambient Air Quality Standards in China by 751.2 times (0.025 ng/m 3 , GB3095-2012). Figure 7 presents the concentrations of the major element compositions at four pollution levels. The concentrations of the elements were higher on heavy pollution days (10.7 µg/m 3 ) and moderate pollution days (9.4 µg/m 3 ) than on light pollution days (6.8 µg/m 3 ) and good days (6.0 µg/m 3 ), and the mass concentrations of total elements accounted for 5.8% during heavy days, which were 57.5%, 19.9%, and 118.3% lower than those during good, light, and moderate episodes, respectively. Notably, the mass concentrations of most crustal elements and anthropogenic pollution elements during heavy pollution days were substantially higher than those on other pollution days. However, the proportion of all crustal elements, such as Al, Na, Mg, K, Fe, and Ca, on heavy pollution days were 63.0%, 52.6%, 55.8%, 31.2%, 59.3%, and 72.7%, lower than those on good days. These elements, which mainly originate from natural sources such as road dust or construction dust, are difficult to raise up into the atmosphere and contribute little to the PM 2.5 on heavy days due to the low wind speed and high RH.  Figure 8 shows the EF values of elements in different pollution levels. The EF values of the elements Na, K, Al, Ca, Fe, Mg were lower than 10 during the sampling period, suggesting that these elements were mostly from mineral dusts. In contrast, the EF values for As, Cu, Cr, Tl, Pb, Sn, Zn, Bi, Sb, Si, and Cd were above 10 in all pollution levels. Moreover, the EF values were higher than 10 on good days for Ni; on light and moderate pollution days for W and Ni; and on heavy pollution days for W, As, Cu, and Mo, which suggests that they came mostly from anthropogenic sources [61]. For example, Cr, Pb, As, Cd, and Sb accumulated from coal combustion; Cu, Zn, and Tl originated from vehicle emissions; Ni was produced by oil combustion; and Sn, Bi, and Mo were associated with industrial production [62,63]. Notably, the EFs of Zn, Bi, Sb, and Cd presented the highest EF values on each pollution day, especially on heavy pollution days (188.6, 314.0, 451.7, and 1784.1), which was probably due to the coal combustion and vehicle emissions contributing to metal emissions. Overall, these elements in PM2.5 in Jinan were mainly produced by human activities, such as coal combustion, vehicle emissions, and industry.  Figure 8 shows the EF values of elements in different pollution levels. The EF values of the elements Na, K, Al, Ca, Fe, Mg were lower than 10 during the sampling period, suggesting that these elements were mostly from mineral dusts. In contrast, the EF values for As, Cu, Cr, Tl, Pb, Sn, Zn, Bi, Sb, Si, and Cd were above 10 in all pollution levels. Moreover, the EF values were higher than 10 on good days for Ni; on light and moderate pollution days for W and Ni; and on heavy pollution days for W, As, Cu, and Mo, which suggests that they came mostly from anthropogenic sources [61]. For example, Cr, Pb, As, Cd, and Sb accumulated from coal combustion; Cu, Zn, and Tl originated from vehicle emissions; Ni was produced by oil combustion; and Sn, Bi, and Mo were associated with industrial production [62,63]. Notably, the EFs of Zn, Bi, Sb, and Cd presented the highest EF values on each pollution day, especially on heavy pollution days (188.6, 314.0, 451.7, and 1784.1), which was probably due to the coal combustion and vehicle emissions contributing to metal emissions. Overall, these elements in PM 2.5 in Jinan were mainly produced by human activities, such as coal combustion, vehicle emissions, and industry.  9%). Additionally, the proportion of mineral dust showed a marked decline; the maximum value of mineral dust was observed on good days (20.2%), while the minimum value was observed on heavy pollution days (only 7.5%). However, minimum amounts of SO 4 2− and NH 4 + were noted on heavy pollution days, accounting for 114.1% and 12.1% of PM 2.5 , respectively, which increased with an aggravation of pollution. EC and TE comprise a relatively small percentage, with the maximum on good days (7.3% and 1.6%) and the minimum on heavy pollution days (5.4% and 0.8%), respectively. The contribution of unidentified constituents reached its maximum (19.0%) on heavy pollution days and its minimum (12.3%) on good days, which needs to be further explored in future studies.

Chemical Mass Closure
Atmosphere 2020, 11, x FOR PEER REVIEW 12 of 20 Figure 8. Enrichment factors of elements in four pollution days (red column ≥10, blue column＜10; The x-axis, y-axis, and z-axis represent different pollution levels, elements, and enrichment factor (EF) values, respectively).

Chemical Mass Closure
PM2.5 mass concentrations were reconstructed in four pollution levels using the method described in Section 2.4.4, compared against the measured PM2.5 mass concentrations. The proportion of major species in PM2.5 are shown in Figure 9. The major constituents were NO3 -and OM. Specifically, the proportions of NO3 -during the four periods was 16.4%, 23.6%, 24.1%, and 23.7%, respectively. The percentage of OM showed a change under different pollution levels, with the maximum on good days (23.2%), the minimum on heavy pollution days (17.5%), and intermediate levels on light (19.0%) and moderate pollution days (17.9%). Additionally, the proportion of mineral dust showed a marked decline; the maximum value of mineral dust was observed on good days (20.2%), while the minimum value was observed on heavy pollution days (only 7.5%). However, minimum amounts of SO4 2-and NH4 + were noted on heavy pollution days, accounting for 114.1% and 12.1% of PM2.5, respectively, which increased with an aggravation of pollution. EC and TE comprise a relatively small percentage, with the maximum on good days (7.3% and 1.6%) and the minimum on heavy pollution days (5.4% and 0.8%), respectively. The contribution of unidentified constituents reached its maximum (19.0%) on heavy pollution days and its minimum (12.3%) on good days, which needs to be further explored in future studies.

Source Apportionment Analysis
In this study, fugitive dust, soil dust, cement dust, coal fly ash, vehicle exhaust dust, steel dust, biomass dust and secondary inorganic aerosols were determined to be the most important pollutants. The source profile data were collected from the Environmental Protection Science Research Institute of Jinan. Figures 10-12 present the source profiles of PM2.5 for different types of sources. According to Figure 10, abundant amounts of Al, Si, Ca, Fe, and OC were found in mobile sources, with

Source Apportionment Analysis
In this study, fugitive dust, soil dust, cement dust, coal fly ash, vehicle exhaust dust, steel dust, biomass dust and secondary inorganic aerosols were determined to be the most important pollutants. The source profile data were collected from the Environmental Protection Science Research Institute of Jinan. Figures 10-12 present the source profiles of PM 2.5 for different types of sources. According to Figure 10, abundant amounts of Al, Si, Ca, Fe, and OC were found in mobile sources, with proportions of 8.6%, 8.0%, 5.6%, 2.2%, and 7.8% in fugitive dust samples, 8.9%, 10.2%, 6.0%, 3.8%, and 8.2% in soil dust samples, and 4.7%, 10.8%, 20.3%, 2.1%, and 5.7% in cement dust samples, respectively. The major components of coal fly ash were OC, Si, Al, and SO 4 2− , with proportions of 11.6%, 10.4%, 7.7%, and 4.7%, and the highest content of PM 2.5 in the vehicle exhaust was OC (25.8%), followed by EC (22.7%) ( Figure 11). The ironmaking dust samples contained the most Fe, with proportions of 33.1% by weight, Ca and Al also comprised a certain proportion, and the chemical profile of the biomass dust in PM 2.5 was found to have major components: OC (28.5%), Cl (21.2%), K (19.1%), and EC (12.7%) ( Figure 12).

Source Apportionment Analysis
In this study, fugitive dust, soil dust, cement dust, coal fly ash, vehicle exhaust dust, steel dust, biomass dust and secondary inorganic aerosols were determined to be the most important pollutants. The source profile data were collected from the Environmental Protection Science Research Institute of Jinan. Figures 10-12 present the source profiles of PM2.5 for different types of sources. According to Figure 10, abundant amounts of Al, Si, Ca, Fe, and OC were found in mobile sources, with proportions of 8.6%, 8.0%, 5.6%, 2.2%, and 7.8% in fugitive dust samples, 8.9%, 10.2%, 6.0%, 3.8%, and 8.2% in soil dust samples, and 4.7%, 10.8%, 20.3%, 2.1%, and 5.7% in cement dust samples, respectively. The major components of coal fly ash were OC, Si, Al, and SO4 2-, with proportions of 11.6%, 10.4%, 7.7%, and 4.7%, and the highest content of PM2.5 in the vehicle exhaust was OC (25.8%), followed by EC (22.7%) (Figure 11). The ironmaking dust samples contained the most Fe, with proportions of 33.1% by weight, Ca and Al also comprised a certain proportion, and the chemical profile of the biomass dust in PM2.5 was found to have major components: OC (28.5%), Cl (21.2%), K (19.1%), and EC (12.7%) ( Figure 12).  Figure 10. Fugitive dust and soil dust source profile. Figure 10. Fugitive dust and soil dust source profile. Proportion of components(%) Figure 11. Coal fly ash and vehicle exhaust source profile. Proportion of components(%) Figure 11. Coal fly ash and vehicle exhaust source profile.
The source apportionment results of PM 2.5 from the CMB receptor model are presented in Figure 13. Nitrate, vehicle exhaust, sulfate, and coal fly ash were the main contributors of PM 2.5 in Jinan. The contribution of nitrate was the highest in each pollution level, with 18.4% on good days, 20.5% on light pollution days, 23.4% on moderate pollution days, and 25.4% on heavy pollution days. Nitrate is mainly converted from NOx, which is derived from coal and motor vehicle exhaust. Vehicle exhaust dust was found to be the second largest contributor of PM 2.5 , which showed a large difference in its contributions during the good and heavy pollution days (14.9% for good days and 18.2% for heavy pollution days). The number of motor vehicles in Jinan reached 2.06 million in 2017, an increase of 13.2% over the previous year, which also indicates that the emissions of vehicle exhaust contributed greatly to the growth of nitrogen in the atmosphere. Sulfate was the third biggest source in Jinan, the contributions from sulfates (which are the precursors of SO 2 and mainly come from the combustion of coal) were 16.7% and 18.6% on moderate and heavy pollution days, respectively. Coal fly ash also contributed significantly to the PM 2.5 in Jinan, with the highest proportion on heavy pollution days (13.3%), 1.6 times higher than on good days, due to the operation of a heating boilers increasing the coal burned for heating. In addition, the contribution of open sources (fugitive dust, soil dust, and cement dust) to PM 2.5 should not be ignored, especially on good days. Proportion of components(%) Figure 12. Steel dust and biomass dust source profile.
The source apportionment results of PM2.5 from the CMB receptor model are presented in Figure 13. Nitrate, vehicle exhaust, sulfate, and coal fly ash were the main contributors of PM2.5 in Jinan. The contribution of nitrate was the highest in each pollution level, with 18.4% on good days, 20.5% on light pollution days, 23.4% on moderate pollution days, and 25.4% on heavy pollution days. Nitrate is mainly converted from NOx, which is derived from coal and motor vehicle exhaust. Vehicle exhaust dust was found to be the second largest contributor of PM2.5, which showed a large difference in its contributions during the good and heavy pollution days (14.9% for good days and 18.2% for heavy pollution days). The number of motor vehicles in Jinan reached 2.06 million in 2017, an increase of 13.2% over the previous year, which also indicates that the emissions of vehicle exhaust contributed greatly to the growth of nitrogen in the atmosphere. Sulfate was the third biggest source in Jinan, the contributions from sulfates (which are the precursors of SO2 and mainly come from the combustion of coal) were 16.7% and 18.6% on moderate and heavy pollution days, respectively. Coal fly ash also contributed significantly to the PM2.5 in Jinan, with the highest proportion on heavy pollution days (13.3%), 1.6 times higher than on good days, due to the operation of a heating boilers increasing the coal burned for heating. In addition, the contribution of open sources (fugitive dust, soil dust, and cement dust) to PM2.5 should not be ignored, especially on good days.   Figure 13. Source contribution to ambient PM2.5 in four pollution days in Jinan.
Overall, the contribution of sulfates, nitrates, vehicle exhaust, and coal fly ash increased with an aggravation of pollution (Figure 14), while biomass dust and cement dust contributed less to PM2.5 at each pollution level. Therefore, we suggest that the government should strengthen the control of sulfate and nitrate precursors in heavy pollution and vigorously control the emissions of coal and motor vehicle exhaust.  Overall, the contribution of sulfates, nitrates, vehicle exhaust, and coal fly ash increased with an aggravation of pollution (Figure 14), while biomass dust and cement dust contributed less to PM 2.5 at each pollution level. Therefore, we suggest that the government should strengthen the control of sulfate and nitrate precursors in heavy pollution and vigorously control the emissions of coal and motor vehicle exhaust. Overall, the contribution of sulfates, nitrates, vehicle exhaust, and coal fly ash increased with an aggravation of pollution (Figure 14), while biomass dust and cement dust contributed less to PM2.5 at each pollution level. Therefore, we suggest that the government should strengthen the control of sulfate and nitrate precursors in heavy pollution and vigorously control the emissions of coal and motor vehicle exhaust.  Figure 14. Trend of source contribution in four pollution days in Jinan. Figure 14. Trend of source contribution in four pollution days in Jinan.

Conclusions
Atmospheric PM 2.5 and its major components were measured from 15 October 2017 to 31 January 2018 in Jinan, China. In this study, the daily average concentration of PM 2.5 was 83.5 µg/m 3 , which is about 1.1 times the National Ambient Air Quality Standard of PM 2.5 (75 µg/m 3 ). The concentration of PM 2.5 on heavy pollution days was approximately 4.3 times higher than that on good days, which was related to the lower wind speed and higher relative humidity on heavy pollution days. High concentrations of WSIs, carbonaceous species, and elements were observed on heavy pollution days. In addition, more NO 3 − , SO 4 2− , and NH 4 + were formed during more polluted days, and the ion balance showed that the PM 2.5 in Jinan was slightly alkaline. The percentage of OC and EC in PM 2.5 decreased by 4.2% and 1.8% from good days to heavy pollution days, respectively. The enrichment factor showed that Zn, Bi, Sb, and Cd were contributed to by coal combustion and vehicle emission. The major constituents of Jinan were secondary aerosol particles and OM. The CMB model showed that sulfate, vehicle exhaust dust, nitrate, and coal fly ash were the main contributors of PM 2.5 in Jinan.
Although the air quality of PM 2.5 has improved slightly compared to prior years, it is still far behind the World Health Organization's recommendations and Chinese national standards. This paper studied the rapid and significant changes in the chemical composition and sources of PM 2.5 on heavy pollution days compared to on good days, revealing that motor vehicles should be better controlled to effectively reduce the concentrations on heavy pollution days.