Fine particulate matter pollution characteristics and source apportionment of Changchun atmosphere

In order to study the pollution characteristics and main sources of fine particulate matter in the atmosphere of the city of Changchun, PM2.5 samples were collected during the four seasons in 2014, and representative months for each season are January, April, July, and October. Sample collection was carried out on 10 auto-monitoring stations in Changchun, and PM2.5 mass concentration and its chemical components (including inorganic elements, organic carbon, elemental carbon, and water-soluble ions) were measured. The results show that the annual average mass concentration of PM2.5 in Changchun in 2014 was about 66.77 μg/m3. Organic matter was the highest component in PM2.5, followed by secondary inorganic ions (SNA), mineral dust (MIN), elemental carbon (EC), and trace elements (TE). Positive matrix factorization (PMF) results gave seven factors, namely, industrial, biomass and coal burning, industrial and soil dust, motor vehicle, soil and secondary ion, light industrial, and hybrid automotive and industrial sources in PM2.5, with contributing values of 18.9%, 24.2%, 5.7%, 23.0%, 11.5%, 13.0%, and 3.6%, respectively.


Introduction
Fine particulate matter (PM 2.5 , particle size 2.5 microns or less) is defined as a major global urban air pollutant and a major health risk factor (Lelieveld et al., 2015;Mukherjee and Agrawal, 2018). In December 2012, more than 300 international research institutions, including the World Health Organization, jointly released the Global Burden of Disease 2012 (Domhasa et al., 2011). Results showed that PM 2.5 could globally cause the premature deaths of approximately 3.2 million people, including China's death toll of 1.23 million (Zhao et al., 2017). Therefore, the problem of fine particulate matter pollution has attracted wide attention at home and abroad. PM 2.5 sources are series of complex mixture of natural and artificial sources. Chemical components mainly include water-soluble ions, carbon components (OC, EC), and inorganic elements (Wu et al., 2016). Primary sources include vehicle exhaust pipes (Gertler et al., 2000), coal (Cheng et al., 2017), and biomass combustion (Singh et al., 2021). Secondary sources include secondary sulfate, nitrate, and organic matter (Sun et al., 2019). In order to effectively control PM 2.5 concentrations in the urban atmosphere, scientific methods must be used for PM 2.5 source analysis. Because the receptor model does not require detailed emission intensity of pollution sources and does not rely on emission conditions, weather, topography, and other information of emission sources, it has become one of the most commonly used methods for source apportionment of particulate matter. It mainly includes chemical mass balance (CMB) and factor analysis (FA) (Feng, Yinchang, 2017). CMB model requires detailed source profile data but does not require a large number of receptor data. FA model does not require profile data but needs a large number of receptor chemical component data, which includes PCA and positive matrix factorization (PMF). Due to the lack of detailed source profile data, the CMB model cannot be applied accurately. Previous studies have shown that PMF can deal with missing values and outliers in data more reasonably than  (Pentti et al., 1993) and reduce the uncertainty of matrix rotation (Huang et al., 1999). Therefore, based on a large number of PM 2.5 chemical composition data, we chose PMF model as the PM 2.5 source apportionment method in this study.
As the core city of Jilin Province's economic zone, Changchun is fast-developing. In recent years, due to a series of reasons such as city-scale expansion, and population (7,545,472 persons) and vehicle (1,500,551 unit) increase, the pollution level of PM 2.5 has been increasing, and the outlook on air quality is not optimistic (Statistics, 2015). Many observational PM 2.5 studies have been extensively carried out in China to understand the characteristics and sources of different chemical components in PM 2.5 . Many scholars have used the PMF model to analyze the main pollution sources of PM 2.5 in Tai'an, Beijing, Tianjin, and Shenzhen, but currently research on PM 2.5 is mainly concentrated in the Beijing-Tianjin-Hebei and the Pearl River Delta regions (Hong et al., 2010). Previous studies on atmospheric particulate matter in Changchun have focused on PM 10 and TSP. There are few studies on PM 2.5 sources in Changchun, and there is a lack of long-term systematic-observation data. So, this paper selected 10 sampling sites in Changchun, sampled during the four seasons, obtained PM 2.5 concentrations and its components in the representative sites in Changchun, and analyzed the sources with the PMF model in order to provide technical support for improving the air quality of Changchun. This paper fills the gap in the research of PM 2.5 in Changchun, as it is of great significance to control the concentration of PM 2.5 .

Sampling sites
Changchun is located at 43°05′N-45°15′N and 124°18′E-127°05E′, with a total area of 20,565 km 2 (Fig. 1). It has a temperate continental humid climate and is the central city of northeastern Asia's economic circle. Ten automonitoring stations were set up in Changchun, all of which are national controlling stations.
Among the 10 sites in Changchun, nine are located in the built-up area of Changchun: Daishan Park (DP), High-Tech  Table 1.

Receptor sample collection
In 2014, PM 2.5 samples were collected in Changchun. In each season, representative months were selected to be January, April, July, and October. The specific sampling dates were April 1-April 15, July 1-July 15, and October 8-October 22 in 2014 and January 5-January 19 in 2015. Chemical components (inorganic elements, carbonaceous components, and water-soluble ions) were measured. A medium-volume air sampler (KC-120 Particulate Sampler, China) at a flow rate of 100 L/min was used to collect PM 2.5 samples for 22 h daily (Fig. 2). In order to evaluate the error of the sampling process, a blank filter membrane was arranged during the sampling process. The blank filter membrane was weighed alone and with the sampling membrane, underwent the whole process of sampling, and was then shipped back to the laboratory with the sampling membrane. The difference between blank filter mass before and after sampling should have been much smaller than the particle mass on the sampling filter; otherwise, batch sampling monitoring data were invalid.
In order to reduce the effect on results, we handled the filter membranes used for sampling before and after use. The handling process of the filter membrane is as follows: (1) All filter membranes have been inspected in detail without any defects. (2) Blank tests were carried out on the selected organic and inorganic filter membranes, and the blank concentrations of the tested substances were all below the detection limit. (3) All filter membranes were baked in the oven before sampling to ensure that the accuracy of analysis was not affected. (4) Put the filter membrane into the dryer for 3 days after baking for full drying. (5) The filter membrane is weighed and the mass is recorded. (6) The weighing was carried out under the condition of constant temperature and humidity, and the balance meets the standard. (7) After weighing, the filter membrane was placed in the preservation box containing the filter membrane, which avoided the bending and folding of the filter membrane.
Samples were collected on a quartz fiber filter (90-mm diameter) or a cellulose filter (Teflon, 90-mm diameter) (Fig. 3). The filter was weighed before and after the sampling process with a microbalance MX5 Mettler Toledo. We weighed the filter in the weighing chamber for 48 h. Balance conditions were temperature, which was controlled at 15-30 °C; accuracy, controlled at 1 °C; humidity, controlled at 50% ± 5% RH; and balanced room temperature humidity that was consistent with the constant temperature and humidity equipment.
To assess the effects of unstable conditions, blank filters were periodically used during each sampling process, and their values were used to correct the measurements of other filters. The hourly PM 2.5 concentrations used in this paper were monitored by the Changchun central monitoring station.

Analysis of water-soluble ions
One-fourth of the filter membrane was cut and put into a sample tube. Then, 10 mL of ultrapure water was added; the sample was placed in an ultrasonic cleaner, sonicated for 20-25 min to dissolve the ions, and then placed in centrifuge for 5 min; the extract was filtered through a filter (0.45 μm). Mass concentrations of Na + , K + , Mg 2+ , and cations were analyzed with an ion chromatograph (DIONEX ICS-2000, USA), and the anion mass concentrations, such as Cl − , NO − 3 , and SO 2− 4 , were analyzed with an ion chromatograph (DIONEX ICS-3000, USA). Use the standard sample to calibrate the standard curve, and add a standard sample for every 10 samples for calibration. In the measurement timing of water-soluble ions, each sample is measured > 2 times, and the same product has a continuous measurement error < 5%.

Organic and elemental carbon analysis
The results of the OC/EC analysis were based on the NIOSH5040 method. The samples are collected on a quartz fiber filter and placed in an instrument, and the furnace is cleaned with helium gas. The furnace is heated to 850℃ through a temperature gradient, and organic compounds are released by thermal decomposition, which entered the MnO 2 furnace. When carbon particles enter the MnO 2 furnace, they are quantitatively converted into CO 2 gas. The CO 2 is washed out of the furnace by a stream of helium gas and measured directly by a non-diffusion infrared (NDIR) detection system. A second temperature gradient is then initiated in the oxidation stream, and all elemental carbon is oxidized away from the filter membrane and into the oxidation furnace and NDIR. Elemental carbon is eventually detected in the same way as organic carbon. Organic carbon and elemental carbon in this study was performed using a semi-continuous OCEC carbon aerosol analyzer (Sunset Laboratory Inc.).

Inorganic element analysis
Cut Teflon filter into small pieces with ceramic scissors and put it in the digestion tank/Teflon beaker. A total of 10 mL of nitric acid-hydrochloric acid mixture was added to immerse the filter membrane, and heat and reflux at 100±5℃ for 2.0 hours, and then cool. Rinse the inner wall of the beaker with reagent water, and leave it for half an hour for leaching. Bring the volume to 50.0mL. Then inorganic elements were determined by inductively coupled plasma mass spectrometry (ICP-MS, Agilent, USA).

PMF source apportionment model
In this study, a PMF receptor model was used to analyze the main sources of PM 2.5 in Changchun. The receptor model is a mathematical model for analyzing the chemical composition and physical properties of atmospheric particulate matter. It can quantitatively identify the contribution ratios of every kinds of pollution sources on the basis of factor eigenvalues or source profiles. The PMF model is a commonly used model for the source apportionment of atmospheric particulate matters. Compared with other source analytical methods, it needs not to measure the source composition profile, it does not require to input the pollutant discharge inventory and photochemical reaction mechanism equations, and it can simultaneously determine the source composition profiles and source contribution ratios. The PMF model decomposes observed component concentration matrix X ij into source concentration profile matrix F kj and factor contribution ratio matrix G ik (Formula 1), where E ij is the residual matrix, the difference between model simulation and actual observation values. The sum of the squares of the residuals of the observation data and its error estimation ratio was defined as Q, and the solution process of the PMF model is the process of minimizing the value of Q.
(1) The basic input to the PMF model is the mass concentration and uncertainty of the chemical components of the sample. Basic output was (a) the share and uncertainty of each chemical component in the source spectrum, (b) the contribution of each factor (source) to the overall concentration of the particulate matter, and (c) the time series of the contribution of each factor (source).
Uncertainty detection method: Concentration below detection limit Unc =

Concentration above detection limit Unc =
This study used EPA PMF 5.0 model software released by the US Environmental Protection Agency in the calculations.

Sample mass closure
Atmospheric particulate matter mass closure can estimate the impact of aerosols from different sources on ambient air quality on the basis of the proportion of different constituent compounds. Mineral dust (MIN), elemental carbon (EC), trace elements (TE), organic matter (OM), secondary inorganic ions (SNA; SO 2− 4 , NO − 3 , NH + 4 ), and the rest are the mass reconstitution of PM 2.5 . Three kinds of water-soluble ions, EC, and TE were directly measured, and OM and MIN were calculated with many of the measured components as follows.
It was generally assumed that there was 0.2-0.4g of other elements (e.g., O, H, and N) per gram of carbon in the OM in atmospheric particulate matters, so 1.2-1.4 represents the times of the weight of OM to OC (Yang et al., 2004). Turpin and Lim recently tested the rationality of this approach and considered that it is appropriate to assume 1.4 for aerosols (2) in urban areas, but it is more suitable for nonurban areas with more bioaerosols or secondary oxidized aerosols with 1.9-2.3 (Turpin and Lim, 2001). In this paper, we took 1.4 as the OC multiplier. Soil dust in atmospheric particulate matter is usually estimated by the sum of the oxide concentrations of specific elements. It is generally assumed that soil dust is composed of the oxides of six elements (i.e., SiO 2 , Al 2 O 3 , TiO 2 , CaO, FeO, Fe 2 O 3 , and K 2 O). Those oxides belong to the eight most important compounds that construct the continental crust, and the two other oxides are Na2O and MgO, accounting for about 3% of the mass of Earth's crust. Since Ti concentration was lower than the detection limit, MIN

Descriptive statistics
The long-term average temperature of Changchun in 2014 was 4.75 ℃, the highest temperature was 36 ℃, and the lowest temperature was −35 ℃. Annual precipitation was 1480.5 mm, and average daily precipitation is 4.1 mm. The main wind direction is southwestern, western, and southern (meteorological data from the website: http:// data. cma. cn/ data/ detail/ dataC ode/A. 0012. 0001. html). The average concentration of PM 2.5 in Changchun in 2014 was 66.77 μg/m 3 , which was 1.9 times the annual average primary limit value (35μg/m 3 ) in the Ambient Air Quality Standard (GB 3095-2012) (Ministry of Ecology and Environment of the People, 2012). Monthly mean values of PM 2.5 in Changchun in 2014 are shown in Figure 4. The PM 2.5 data peaked in October, and the monthly average concentration was as high as 143.76 µg/m 3 . The reason is that the gas pollutant emissions are high during the start-up and

Concentration[μg/m³]
Month commissioning of the heating boiler at the end of October. And adverse meteorological conditions in autumn such as low temperature (Li et al., 2019), high pressure (Li et al., 2017), and inversion weather (Yang et al., 2017), are unfavorable for the spread of gas pollutants.
The heating period in Changchun is from mid to late October to early April of the following year. It can be seen from Figure 5 that hourly PM 2.5 concentrations in Changchun basically showed a bimodal change, and the heating period was significantly higher than the nonheating period. The first peak appeared in the morning from 6:00 to 9:00, during early traffic peak. The number of road vehicles increased gradually, and emitted pollutants increased. The second peak appeared at 19:00 to 21:00. The reason may be the outdoor human activities. Barbecues are a common dietary feature in the northeast. During this period, residents like to have a barbecue, and grilled fumes increase PM 2.5 concentrations. Due to the large amount of coal used during the heating period, PM 2.5 concentrations significantly increased compared to the non-heating period.

Mass closure
From the perspective of the whole year, OM was the highest proportion of Changchun PM 2.5 , accounting for 38.2% (Fig. 6). From the four seasons, the seasonal variation trend for OM was winter > summer > autumn > spring, and OM proportion was the highest in winter contributed 60%. The reason is that heating in Changchun begins in winter, and the amount of burnt coal increases, resulting in an increase in discharged pollutants. The reason why summer OM proportion is higher than that in spring and autumn is because of the increasing of barbecues in the summer night and the emission of a large amount of oil smoke. The seasonal trend for MIN is spring > summer > winter > autumn. Wind speed in spring (4.41 m/s) was higher than that in other seasons, which rolls up soil dust. The trend of EC in PM 2.5 is autumn > spring > summer > winter, and EC proportion in autumn is significantly higher than that of the other seasons, which is due to large-scale straw burning in autumn and a relatively high motor flow. EC produced by biomass burning in the autumn and motor vehicle exhausts have great impact on PM 2.5 (Ravindra et al., 2019 ) . The seasonal change of SNA is spring > autumn > winter > summer,  and the formation of secondary particulate matter is not only related to pollutant discharge, but also wind speed, temperature, humidity, and other meteorological conditions that have related impact on the formation of secondary particulate matter. During the winter-spring transition, the increase in the number of activated molecules which is due to the gradual increase in temperature makes it easier for SO 2 to be converted into SO 2− 4 (Zheng et al., 2015). Straw burning in autumn has great influence on NO X , resulting in an increase in nitrate ion content, so SNA proportion in spring and autumn is higher.
Mass closure results in Changchun are similar to those in Beijing (Yang et al., 2004) and Shenzhen (Huang et al., 2014), but OM proportion in Changchun PM 2.5 was higher and SNA proportion was lower; the main components of PM 2.5 in Shenyang are SNA, OM, and MIN (Tian et al., 2019). This may be related to the industrial structure and meteorological conditions in Shenyang. Changchun's MIN accounts for 12%. According to the study of Hasheminassab et al. (2014), MIN in Los Angeles accounts for 6% for its PM 2.5 , which is half of that in Changchun. In addition, Los Angeles is located on the coast and is affected by sea salt sources. This is where characteristics of fine particulate matter pollution in Changchun and cities of the USA are different.

Enrichment factors
The enrichment factor (EF) was used to determine the natural or believed sources of the 21 elements (Fig. 7). The enrichment factor for element X is defined as where C x is the concentration of element X, C ref is the concentration of the reference element, (C x ∕C ref ) PM is the ratio of the two in PM 2.5 , and (C x /C ref ) crust is the concentration ratio of the two in the crust. This paper chose Al as the reference element = 1. The background value of soil elements was taken from background and reference values of soil chemical elements in Chinese cities (Cheng et al., 2014). Results are shown in Fig. 4. Studies have shown that an EF < 1 indicates that elements are not enriched mainly from natural sources such as the crust, and EF > 10 indicates that elements are enriched mainly from anthropogenic sources. According to the results, Co (<1) was mainly from crust source, while Mg, K, Ca, Mn, Fe, and Si (1-10) were from both anthropogenic and crustal sources. For B, V, Cr, Ni, Cu, Zn, As, Se, Mo, Cd, Sn, Pb, and Si (>10), they were mainly from anthropogenic source. The higher the EF, the stronger the enrichment degree. The EF of B, Cr, Zn, As, Se, Mo, Cd, Sn, Pb, and Hg was greater than 100, indicating strong enrichment.

Source apportionment
The model ran in a factor of 3-9 to find the best results (USEPA, 2014). The most stable solution was found when the factor number was 7. Seven factors were industrial, biomass and coal burning, industrial and soil dust, motor vehicle, soil and secondary ion, light industrial, and hybrid automotive and industrial sources (Fig. 8). The contribution rate of each factor must be greater than 0.05%; otherwise, it will not be included in the determined factor (Fig. 9).
The contribution rate of the first factor was 18.9%, from which the specific gravity of Zn, Cd, Cu, and Mn was large. Zn is often derived from rolling mills, Cu and Mn are related to metal smelting, and Changchun is used as a hub for automobile manufacturing, which has great demand for steel smelting. Hence, judgment factor 1 is industrial sources.
The contribution rate of the second factor was 24.2%. From this, Cl − and K + had large specific gravity. Cl − can be used as a characteristic component of coal-fired emissions; and K + is a characteristic component of biomass combustion, and it is converted from NO x . Straw burning has a great The contribution rate of the third factor was 5.7%, in which Ni and Al were relatively large. Ni is mainly from artificial sources and related to metal smelting; they belong to industrial dust, and Al mainly comes from Earth's crust. Hence, judgment factor 3 is industrial and soil dust.
The contribution rate of the fourth factor was 23.0%, in which the contributions of EC, Pb, and Se were more prominent. EC is the main emission of motor fuel. Pb production is related to the wear of the brake components of motor vehicles. Hence, judgment factor 4 is motor vehicle sources. In Changchun, the number of motor vehicles is high, and traffic is prone to congestion. The motor vehicle exhausts have great impact on the atmosphere.
The contribution rate of the fifth factor was 11.5%, in which Co, Mg 2+ , and SO 2− 4 had a high load. Co is a soil element, Mg 2+ may be derived from road dust, and SO 2− 4 is secondary ion; so, judgment factor 5 is a soil and secondary ion sources.
The sixth factor contributed 13.0%, in which F − and Na + had large specific gravity, while atmospheric perfluorinated and polyfluoro-organic compounds (PFCs) are usually used as surfactants in nonstick coatings, paper, textiles, and other coatings. In this production, judgment factor 6 is light industrial sources.
The seventh factor contributed 3.6%, of which Hg, V, and Cr had higher load rates. Studies showed that oil and gas used in motor vehicles contain Hg. After combustion in motor vehicle engines, exhaust gas containing Hg is discharged into the atmosphere. In addition, V is often added as an auxiliary material to oil products. Cr mainly comes from industrial production such as electroplating, battery manufacturing, and stainless steel production. Hence, judgment factor 7 is hybrid automotive and industrial sources.
This study compared the results of PM 2.5 source apportionment in Changchun with other important cities and found that the contribution rate of coal-fired sources in Changchun is similar to that of Zhengzhou (Geng et al., 2013) and slightly higher than that of Beijing (Wang et al., 2008;Zhou et al., 2021) and Tianjin (Kong et al., 2010), which may be related to coal-firing demands in the Beijing-Tianjin region (Table 2). Compared with Hangzhou (Wang et al., 2016), the contribution rate of coal-fired sources in Changchun is much higher than that in Hangzhou, which is mainly related to the non-heating season in Hangzhou, resulting in less contribution from coal-fired sources. The secondary ion contribution rate of Changchun is significantly lower than that of Beijing (Wang et al., 2008), Tianjin, Shenzhen (Huang et al., 2014), Hangzhou, and other cities. This may be related to the relatively stable meteorological conditions in Changchun during the sampling period, which is not conducive to the formation of secondary ions. From the perspective of the contribution rate of motor vehicles, Changchun is similar to Shenzhen and Hangzhou, lower than that of Beijing (Zhou et al., 2021), which may be related to the motor vehicle limit policy in various regions. Due to energy structure differences of each region, industrial structure, economic development status, and natural conditions, the results of each study are uncertain (Table 3). Therefore, the source apportionment results need further examination in the future.

Conclusion
1. The PM 2.5 concentrations in the heating period were significantly higher than those in the non-heating period. For the hourly concentration, the hourly PM 2.5 concentrations in Changchun basically showed double peak changes that have certain regional characteristics. Changchun should pay attention to controlling the impact of coal burning during winter heating. 2. From the perspective of the mass closure of Changchun PM 2.5 , OM had the largest proportion of PM 2.5 , accounting for 38.2%. Changchun should make efforts to control the emission of organic pollutants in the air. SNA in PM 2.5 should not be ignored, accounting for 13.3%. MIN accounted for 12%, which is twice the city center of Los Angeles. This is the difference between characteristics of fine particulate pollution in Changchun and big cities in the USA, which may be related to the degree of urban greening. PM 2.5 quality reconstruction in various regions has certain regional characteristics. 3. According to the PMF results, PM 2.5 sources of Changchun are divided into industrial, biomass and coal burning, industrial and soil dust, motor vehicle, soil and secondary ion, light industrial, and hybrid automotive and industrial sources. Compared with other cities' source apportionment results, it was found that winter heating in northern regions has increased the proportion of coal burning sources. Due to the low winter temperature and stable meteorological conditions in Changchun, the proportion of secondary ion sources is relatively low compared to that in other cities. Motor vehicle limit policy is implemented in the Beijing-Tianjin-Hebei region, so the vehicle sources in Changchun are relatively high.
Therefore, Changchun should improve the utilization efficiency of coal combustion and vigorously develop clean energy; actively promote green travel, encouraging citizens to use public transport to reduce vehicle emissions; strengthen environmental management and straw remediation actions while encouraging comprehensive straw utilization; increase urban greening areas and regularly carry out ground-sprinkling work, reducing the impact of ground dust on the atmosphere; and increase the management of street barbecue merchants, strictly prohibit open-air barbecue, and impose corresponding punishments on violators.