Characteristics of PM2.5 in an Industrial City of Northern China: Mass Concentrations, Chemical Composition, Source Apportionment, and Health Risk Assessment

Urban and suburban PM2.5 samples were collected simultaneously during selected periods representing each season in 2019 in Zibo, China. Samples were analysed for water-soluble inorganic ions, carbon components, and elements. A chemical mass balance model and health risk assessment model were used to investigate the source contributions to PM2.5 and the human health risks posed by various pollution sources via the inhalation pathway. Almost 50% of the PM2.5 samples exceeded the secondary standard of China’s air quality concentration limit (75 µg/m3, 24 h). Water-soluble inorganic ions were the main component of PM2.5 in Zibo, accounting for 50 ± 8% and 56 ± 11% of PM2.5 at the urban and suburban sites, respectively. OC and OC/EC decreased significantly in the past few years due to enhanced energy restructuring. Pearson correlation analysis showed that traffic emissions were the main source of heavy metals. The Cr(VI) concentrations were 1.53 and 1.92 ng/m3 for urban and suburban sites, respectively, exceeding the national ambient air quality standards limit of 0.025 ng/m3. Secondary inorganic aerosols, traffic emissions, and secondary organic aerosols were the dominant contributors to PM2.5 in Zibo, with the total contributions from these three sources accounting for approximately 80% of PM2.5 and the remaining 20% attributed to traffic emissions. The non-carcinogenic risks from crustal dust for children were 2.23 and 1.15 in urban and suburban areas, respectively, exceeding the safe limit of 1.0 in both locations, as was the case for adults in urban areas. Meanwhile, the carcinogenic risks were all below the safe limit, with the non-carcinogenic and carcinogenic risks from traffic emissions being just below the limits. Strict control of precursor emissions, such as SO2, NOx, and VOCs, is a good way to reduce PM2.5 pollution resulting from secondary aerosols. Traffic control, limiting or preventing outdoor activities, and wearing masks during haze episodes may be also helpful in reducing PM2.5 pollution and its non-carcinogenic and carcinogenic health impacts in Zibo.


Introduction
Particulate matter with an aerodynamic diameter ≤ 2.5 µm, known as PM 2.5 , is suspended in the air and is composed of various chemicals and particle sizes. PM 2.5 resides in the atmosphere for long periods and can penetrate deep into the lungs [1,2]. It has attracted significant attention due to its adverse effects on climate, visibility, and especially human health [3][4][5][6]. Long-term exposure to PM 2.5 has been proven to increase the risk of lung cancer, dementia, cardiovascular disease, chronic respiratory diseases, and depressive-like responses [7][8][9][10][11][12].
Because PM 2.5 poses human health risks, it is necessary to explore its toxicity. Carbonaceous species, water-soluble ionic species (WSI), and mineral dust are the major components

Gravimetric and Sample Analysis
Before sampling, quartz filters were baked in a muffle furnace at 550 °C for 3 h to remove carbonaceous pollutants. Before and after sampling, the filters were balanced under constant temperature (20 ± 1 °C) and humidity (50 ± 5%) for more than 24 h and weighed using an automatic filter-weighing system (AWS-1; Comde-Derenda GmbH, Germany, readability = 1 µg). The difference between two consecutive measurements was not more than 40 µg. After sampling, the filters were stored at -20 °C until analysis to prevent volatilization of volatile components [27]. The concentrations of PM2.5 were determined based on the values analysed from the Teflon filters.
Quartz filters were used to analyse the carbon contents and water-soluble inorganic ions (WSIIs). Organic carbon (OC) and elemental carbon (EC) were measured according A total of 45 PM 2.5 samples were simultaneously collected at the two sites each day during the periods of 8-22 January (winter), 16-25 April (spring), 30 July-8 August (summer), and 16-25 October (autumn) in 2019. Each day's sampling time was defined as running from 10:00 a.m. until 9:00 a.m. the next day (23 h). The samplers (ZR-3930; Zhongrui Inc., Qingdao, China) had two parallel channels and a flow rate of 16.7 L min −1 and used Teflon filters (diameter = 47 mm; Whatman Inc., Buckinghamshire, UK) and quartz filters (diameter = 47 mm; Pall, San Diego, CA, USA). The PM 2.5 manual sampling method was in accordance with the latest national standard, HJ 656-2013 [26].

Gravimetric and Sample Analysis
Before sampling, quartz filters were baked in a muffle furnace at 550 • C for 3 h to remove carbonaceous pollutants. Before and after sampling, the filters were balanced under constant temperature (20 ± 1 • C) and humidity (50 ± 5%) for more than 24 h and weighed using an automatic filter-weighing system (AWS-1; Comde-Derenda GmbH, Germany, readability = 1 µg). The difference between two consecutive measurements was not more than 40 µg. After sampling, the filters were stored at −20 • C until analysis to prevent volatilization of volatile components [27]. The concentrations of PM 2.5 were determined based on the values analysed from the Teflon filters.
Quartz filters were used to analyse the carbon contents and water-soluble inorganic ions (WSIIs). Organic carbon (OC) and elemental carbon (EC) were measured according to interagency monitoring of protected visual environments (IMPROVE) thermal optical reflectance (TOR; Model 2001A, Desert Research Institute, Reno, NV, USA). Because the oxidation temperatures of OC and EC are different, the IMPROVE temperature-programmed method was used to analyse the samples in this study. The heating program was divided into two stages: in the first stage, in the oxygen-free and pure-helium environment, the quartz filter was heated to 140 • C, 280 • C, 480 • C, and 580 • C to obtain the four components of OC (OC1, OC2, OC3, and OC4, respectively); in the second stage, in a helium atmosphere containing 2% oxygen, the temperature was gradually increased to 580 • C, 740 • C, and 840 • C to obtain the three components of EC (EC1, EC2, and EC3, respectively). Because OC forms optical pyrolyzed carbon (OPC) during the carbonization process, according to the IMPROVE analysis protocol, OC and EC were defined according to Equations (1) and (2), respectively [28]. OC = OC1 + OC2 + OC3 + OC4 + OPC (1) Water-soluble inorganic ions (Na + , NH 4 + , K + , Mg 2+ , Ca 2+ , Cl − , NO 3 − , and SO 4 2− ) were measured using ion chromatography analysers (DIONEX ICS-1100 and ICS-2100, Thermo Company, Waltham, MA, USA). The samples were soaked in deionized water (10 mL), shaken well, leached under ultrasonic vibration for 15 min, and let stand for 5 min; then, the supernatant was injected for ion chromatographic analysis [21].
To analyse the concentrations of elements, one 1/2 of a Teflon filter was cut into fragments and solubilized in a 5 mL HF acid solution in a PTFE crucible. The solution was then progressively heated to 160 • C on an electric stove for 2 h and leached in 10 mL of 2% HCl. After cooling, the solution was moved into a plastic colorimetric tube before analysis of eleven trace elements (As, Cd, Co, Cr, Cu, Mn, Ni, Pb, Sr, V, and Zn) using inductively coupled plasma-mass spectrometry (ICP-MS, Agilent 7500a, Santa Clara, CA, USA). The other 1/2 of Teflon filter was put in a Nickel crucible and progressively heated to 300 • C to ash in a muffle furnace and kept in the furnace under constant temperature for about 20 min. After cooling, pure ethanol and 0.2 g NaOH were added to crucible and melted for 10 min. Following the extraction from boiling water, the solution was moved into a plastic tube with 1 mL HCl (12 mol/L). After dilution, the other trace elements (Al, Mg, Ca, Fe, and Ba) were analysed by an inductively coupled plasma-optical spectrometer (ICP-OES, Agilent, Santa Clara, CA, USA) [29].

Source Apportionment Model
Source apportionment was conducted with a CMB model. Considering that a large number of samples (usually >100) is required for statistical analysis and there some subjectivity and uncertainty are involved in the determination of the number of source types and the discrimination in PMF model [30,31], we chose a CMB model rather than a PMF model. The theory behind the CMB model has is introduced in detail in [31]. The fundamental relationship between the concentration at a receptor site and source information in the CMB model can be expressed as Equation (3): where C i is the concentration (µg m −3 ) of elemental component i at a receptor site, p represents the number of sources, F ji is the fraction of the ith element from the jth source, and S j is the concentration of the contribution from the jth source to the receptor site. In our study, 90 ambient PM 2.5 samples and source profiles were input to CRAES CMB 1.0 to obtain the source apportionment results and contributions of various PM 2.5bound metals. CRAES CMB 1.0 is a Chinese-language version of the model derived from EPA CMB 8.2 by the Chinese Research Academy of Environmental Sciences, which includes an enumeration method [32]. Based on the emission inventory and industrial structure of Zibo, we chose the potential sources and conducted source sampling and analysis work. The squared regression coefficient (R 2 ), sum of residual squares value (X 2 ), degrees of freedom (DF), and percentage of explained mass to sample total mass were used to evaluate the validity of source contribution measures [33]. Table 1 summarizes the performance measures of the CMB model used in the present study.

Health Risk Assessment Model
The investigated model was introduced by the United States Environmental Protection Agency to quantify health risks (carcinogenic and non-carcinogenic) and human exposure to elements in dust [34]. The three main exposure pathways in humans are direct ingestion, inhalation, and dermal contact. PM 2.5 -bound metals cause human health hazards mainly via the respiratory system [31], so we only considered the inhalation exposure pathway in this study. Exposure is expressed in terms of a daily dose (mg kg −1 day −1 ), and the doses received through inhalation were calculated with Equations (4) and (5) [35,36]. Ba, Co, Cr(VI), and Mn exhibited non-carcinogenic risks, whereas As, Cd, Co, Cr(VI), and Ni exhibited carcinogenic risks in this study [37,38]. Because the toxicity of Cr is attributed to its hexavalent state, its concentrations were scaled by a factor of 1/7, given that atmospheric Cr(VI) is reported to have a ratio of 1/6 relative to Cr(III) [39,40]. where ADD inh (mg kg −1 day −1 ) is the dose obtained via inhalation, and LADD (mg kg −1 day −1 ) is the lifetime average daily dose of carcinogens. The exposure parameters are listed in Table 2. To assess the health risks due to trace element exposure in PM 2.5 , the hazard quotient (HQ), hazard index (HI), and carcinogenic risk (RI) techniques were employed and calculated with Equations (6)-(8) [35,36].
where RfD (mg kg −1 day −1 ) is the corresponding reference dose, and SF is the corresponding slope factor. The values of RfD and SF used in this study are listed in Table 3.  Figure 2 shows the PM 2.5 mass concentrations in winter, spring, summer, and autumn during the sampling period at the urban and suburban sites in Zibo. The minimum, median, and maximum concentrations of urban PM 2.5 were 20, 56, and 313 µg/m 3 , respectively, with 48.89% of the values exceeding the secondary standard of China's air quality concentration limit (75 µg/m 3 , 24 h) [44]. The suburban minimum, median, and maximum concentrations of PM 2.5 were 20, 65, and 212 µg/m 3 , respectively, with 42.22% of data representing concentrations > 75 µg/m 3 . Furthermore, 80% of the values at both sites exceeded the primary standard limit (35 µg/m 3 , 24 h), and 100% exceeded the WHO air quality guidelines for PM 2.5 (15 µg/m 3 , 24 h) [44,45]. By comparing the statistics of the two sites, we can see that the median and maximum concentrations at the urban site were greater than those at the suburban site, especially the maximum values. Compared with sites in the North China Plain (NCP; northwest of Zibo), the proportion of polluted weather (PM 2.5 > 75 µg/m 3 ) at the urban (67%) and suburban (67%) sites during the winter sampling period were greater than that at the NCP sites (50%). However, the overall concentrations in summer were <75 µg/m 3 at both the urban and suburban sites, whereas the NCP sites experienced haze pollution during their intensive summer observation period [46]. In other words, Zibo experienced greater winter pollution than the NCP but less summer pollution.
The ratio of NO3 − to SO4 2− ([NO3 − /SO4 2− ]) has been used as an indicator of the relat importance of mobile vs. stationary sources of nitrogen and sulphur in the atmosph

WSII Levels
The annual mean WSII concentrations of PM 2.5 at the urban and suburban sites ( Table 4) were 42.2 ± 35.0 and 44.0 ± 31.3 µg/m 3 , accounting for 50 ± 8% and 56 ± 11% of the total PM 2.5 , respectively, which is close to the results for 2018 [27]. This result shows that WSIIs were the main components of PM 2.5 in both urban and suburban areas. The secondary water-soluble inorganic ions composed of sulphate (SO 4 2− ), nitrate (NO 3 − ), and ammonium (NH 4 + ) aerosol (SNAs) were the major contributors to WSIIs, and their annual average concentrations were 37.8 ± 32.7 µg/m 3 (approximately 44% of the PM 2.5 in urban site) and 39.7 ± 29.8 µg/m 3 (approximately 50% of the PM 2.5 in suburban site). SNAs are formed in the atmosphere from gaseous precursors, for example, sulphur dioxide (SO 2 ), nitrogen oxides (NOx), and ammonia (NH 3 ), through complex chemical reactions [47]. The ratio of SNA/PM 2.5 was highest in autumn (56% for both urban and suburban sites), which is indicative of intense chemical reactions in this season.
Winter n = 15 Spring n = 10 Summer n = 10 Autumn n = 10 Annual n = 45 The ratio of NO 3 − to SO 4 2− ([NO 3 − /SO 4 2− ]) has been used as an indicator of the relative importance of mobile vs. stationary sources of nitrogen and sulphur in the atmosphere [48][49][50]. Ratios > 1.0 indicate that the particle sources at the sampling site are likely dominated by mobile sources, whereas stationary sources predominate at ratios < 1.0. In this study, the ratios of NO 3 − /SO 4 2− were 2.5, 2.2, 0.5, and 3.9 (2.5, 2.7, 0.8 and 4.2) in winter, spring, summer, and autumn, respectively, with an annual average ratio of 2.3 (2.5) at the urban site (suburban site, Table 4). Note that, in summer, a lower NO 3 − /SO 4 2− ratio may be ascribed to high temperatures, which lead to the evaporation of NH 4 NO 3 , whereas high ratios in other seasons are more likely related to traffic emissions from roads near the sampling sites [51].
There was a typical haze episode from 10 to 14 January 2019 in Zibo according to the definition of haze episode [27]; Figure 3 gives the daily concentrations of PM 2.5 , SNA, and NO 3 − /SO 4 2− during this episode. The average concentrations of PM 2.5 were 197 µg/m 3 and 180 µg/m 3 at the urban and suburban sites, respectively, and the peak values were 312 and 212 µg/m 3 , respectively. Obviously, SNAs were the main components in PM 2.5 during this episode, and the peak values of SNA/PM 2.5 were over 60%. SNA components ranked in the order of NO 3 − > SO 4 2− > NH 4 + , which was different from Beijing, where the SO 4 2− ranked first [50]. In addition, NO 3 − /SO 4 2− was above 1.0 and increased during the accumulation period. The above information demonstrates that NO 3 − played a predominant role in the formation of heavy haze, and mobile sources had a relatively important contribution.

OC and EC Levels
OC and EC are also major components of PM 2.5 , and their daily concentrations are shown in Figure 4. The annual concentrations of OC in PM 2.5 in urban and suburban areas were 10.9 and 11.2 µg/m 3 , respectively, which are lower than those in other Chinese cities, e.g., 13.8 µg/m 3 in Changzhou [51] and 14.4 µg/m 3 in Wuhan [52], and are comparable to those measured in Guangzhou (9.4 µg/m 3 ) [5]. The annual EC concentrations were 4.7 µg/m 3 in urban and 5.1 µg/m 3 in suburban areas, which are close to those of Changzhou (5.4 µg/m 3 ) [51], Wuhan (5.2 µg/m 3 ), and Guangzhou (6.2 µg/m 3 ) [5]. Obvious similar seasonal variations in OC and EC were observed in both urban and suburban areas, with the highest average concentrations occurring in winter: 18.6 and 6.8 µg/m 3 in urban and 19.1 and 8.2 µg/m 3 in suburban areas, respectively. Moreover, the ratios of OC/PM 2.5 were highest in winter, accounting for 19% and 20% of the PM 2.5 in urban and suburban areas, respectively. The lowest concentrations of OC and EC were observed in summer.

OC and EC Levels
OC and EC are also major components of PM2.5, and their daily concentrations are shown in Figure 4. The annual concentrations of OC in PM2.5 in urban and suburban areas were 10.9 and 11.2 µg/m 3 , respectively, which are lower than those in other Chinese cities, e.g., 13.8 µg/m 3 in Changzhou [51] and 14.4 µg/m 3 in Wuhan [52], and are comparable to those measured in Guangzhou (9.4 µg/m 3 ) [5]. The annual EC concentrations were 4.7 µg/m 3 in urban and 5.1 µg/m 3 in suburban areas, which are close to those of Changzhou (5.4 µg/m 3 ) [51], Wuhan (5.2 µg/m 3 ), and Guangzhou (6.2 µg/m 3 ) [5]. Obvious similar seasonal variations in OC and EC were observed in both urban and suburban areas, with the highest average concentrations occurring in winter: 18.6 and 6.8 µg/m 3 in urban and 19.1 and 8.2 µg/m 3 in suburban areas, respectively. Moreover, the ratios of OC/PM2.5 were highest in winter, accounting for 19% and 20% of the PM2.5 in urban and suburban areas, respectively. The lowest concentrations of OC and EC were observed in summer.
The OC/EC ratio was used to indicate the existence of secondary organic carbon (SOC). More than 60% of OC/EC ratios in this study exceeded 2.0, which reflects the fact that SOC is a significant component of PM2.5 [53]. However, the annual average OC/EC ratio in this study was only 2.4, which is much lower than that reported in a previous study (OC/EC = 6.68, OC = 31.97 ± 20.43 µg/m 3 , EC = 5.01 ± 2.68 µg/m 3 ; sampling time = 2006-2007) [25]. Clearly, the distinction between annual average OC concentrations caused this result. This might be attributed to the partial replacement of coal-fired energy with clean energy over the past decade, especially after the Chinese State Council implemented the Air Pollution Prevention and Control Action Plan in September 2013. According to a study by Li et al., study [54], OC emissions decreased by 31.7% in Beijing and surrounding areas (including Zibo) during the period of 2014-2017 as a result of strict governance of residential fuel and enhanced energy restructuring by using cleaner natural gas and electricity replacing coal.

Element Levels
The seasonal average concentrations of various elements in PM2.5 are shown in Figure  5. The total elemental concentrations accounted for 1.96% and 1.33% of the mean PM2.5 concentrations at the urban and suburban site, respectively. In urban PM2.5 samples, Ca exhibited the highest levels (453.  The OC/EC ratio was used to indicate the existence of secondary organic carbon (SOC). More than 60% of OC/EC ratios in this study exceeded 2.0, which reflects the fact that SOC is a significant component of PM 2.5 [53]. However, the annual average OC/EC ratio in this study was only 2.4, which is much lower than that reported in a previous study (OC/EC = 6.68, OC = 31.97 ± 20.43 µg/m 3 , EC = 5.01 ± 2.68 µg/m 3 ; sampling time = 2006-2007) [25]. Clearly, the distinction between annual average OC concentrations caused this result. This might be attributed to the partial replacement of coal-fired energy with clean energy over the past decade, especially after the Chinese State Council implemented the Air Pollution Prevention and Control Action Plan in September 2013. According to a study by Li et al., study [54], OC emissions decreased by 31.7% in Beijing and surrounding areas (including Zibo) during the period of 2014-2017 as a result of strict governance of residential fuel and enhanced energy restructuring by using cleaner natural gas and electricity replacing coal.

Element Levels
The seasonal average concentrations of various elements in PM 2.5 are shown in Figure 5. The total elemental concentrations accounted for 1.96% and 1.33% of the mean PM 2.5 concentrations at the urban and suburban site, respectively. In urban PM 2.5 samples, Ca exhibited the highest levels (453.  .7), and Co (0.2; medians, ng/m 3 ). The annual median concentrations of urban and suburban heavy metals Pb, Cd, and As were lower than the national ambient air quality standards limits of 500 ng/m 3 , 5 ng/m 3 , and 6 ng/m 3 , respectively. However, the Cr(VI) concentrations exceeded the limit of 0.025 ng/m 3 [45]. Shandong province, where the Zibo is located, has the highest Cr emissions among all provinces in China due to its high capacity for steel production and coal consumption of industrial boilers [55].     Figure 7. The urban and suburban sites have several common characteristics. Al, Ca, and Fe were significantly correlated with one another (p < 0.01, r > 0.48), and their concentrations increased significantly in spring, which is the season of frequent sand/dust storm weather in northern China [14], increasing these crustal elements. Additional, Mn, Co, Cu, Zn, Cd, and Pb were also significantly correlated with one another (p < 0.01, r > 0.86), all of which are typically elements of motor vehicle emissions sources or classic traffic-source-related elements [56,57]; most showed a significant increase in winter and autumn. The probable reasons are that vehicle emissions have strong time sensitivity, being highest in winter and lowest in summer [58], with clean travel modes, such as electric bicycles, being more commonly used in summer. Arsenic, a representative coal-combustion element, showed higher concentrations in winter than in other seasons due to increased demands for space heating [13]. However, there were some distinctions between the urban and suburban sites; e.g., V and Ni were not correlated with any of the other 15 elements at the suburban site, whereas the concentration of Ni was significantly lower in winter. This is probably due to the fact that agricultural and construction activities are more common in suburban areas than in urban areas and are less prevalent in winter, resulting in a significant decrease in the use of related off-road machinery and corresponding exhaust emissions [59,60]. Sr and Ba, which are typical alkaline-earth metals emitted by industrial processes such as alloy smelting, ceramic production, etc., were significantly related to each other in urban samples (p < 0.01, r = 0.81).

CMB Source Profiles
Chemical mass balance models for source apportionment require information about the chemical characteristics of the sources that are likely to affect pollutant concentrations at receptor sites [33]. The weight percentages of the PM 2.5 -bound chemical species in all sources (crustal dust, coal boilers, graphite production, aluminium smelting, glass manufacturing, iron smelting, traffic emissions, secondary inorganic aerosols, and secondary organic aerosols) are presented in Figure 8. amounts of Al (19.6%) and Fe (9.6%) [1], respectively. Carbon components are the most abundant species in traffic emissions. It is worth noting that the abundance of EC in traffic emissions is 14.3%, which is 5-27 times that in other profiles. EC was significantly correlated with traffic-source-related elements (Mn, Co, Cu, Zn, Cd, and Pb) (p < 0.01) in ambient PM2.5 samples. Hence, EC can be used as a key marker to distinguish traffic emissions from other sources, as it primarily arises from engines [13]. It is evident from the source profiles that the major components of crustal dust are Ca, OC, SO 4 2− , Si, and Al (Ca, Si, and Al are crustal elements). The high percentage of OC is an indication of the mixing of dust and organic matter during aging or by entrainment of organic materials from soils [61]. The coal boiler profile is dominated by SO 4 2− , OC, NH 4 + , Na, Cl − , Si, and Al, with abundances of 23.9%, 18.0%, 9.9%, 3.4%, 2.5%, 2.4%, and 2.0%, respectively. This is consistent with the results of a study by Han et al. [62] in which OC, EC, SO 4 2− , Ca, Si, Fe, and Al were found to be the abundant species in coal combustion fly ash from Xining City, China. The source profiles for graphite production and glass manufacturing are similar; therefore, SO 4 2− , NH 4 + , and OC were the major components, and Zn [63] and Na [64] were also enriched in graphite production and glass manufacturing, respectively. Aluminium smelting and iron smelting are characterized by large amounts of Al (19.6%) and Fe (9.6%) [1], respectively. Carbon components are the most abundant species in traffic emissions. It is worth noting that the abundance of EC in traffic emissions is 14.3%, which is 5-27 times that in other profiles. EC was significantly correlated with traffic-source-related elements (Mn, Co, Cu, Zn, Cd, and Pb) (p < 0.01) in ambient PM 2.5 samples. Hence, EC can be used as a key marker to distinguish traffic emissions from other sources, as it primarily arises from engines [13].

CMB Source Apportionment
The CMB results identified nine sources of PM 2.5 during the winter, spring, summer, and autumn sampling periods, including crustal dust (CD), coal boilers (CB), graphite production (GP), aluminium smelting (AS), glass manufacturing (GM), iron smelting (IS), traffic emissions (TE), secondary inorganic aerosols (SIA, including NO 3 − , SO 4 2− , and NH 4 + ), and secondary organic aerosols (SOA; Table 5). The top annual contributors were ranked as SIA > TE > SOA, with percentage contributions of 43.34%, 24.03%, and 9.78% in urban and 49.44%, 20.49%, and 10.57% in suburban areas, respectively. The percentage contributions of SIA were nearly 50%, which is consistent with the high concentrations of SNAs in PM 2.5 . The high contribution of traffic emissions might be attributable to the rapid increase in motor vehicle numbers over the past 10 years (from 443,179 to 1,136,377; Zibo Statistical Yearbook, 2010, 2019; http://tj.zibo.gov.cn/col/col886/index.html; accessed on 31 October 2021). Apart from TE and SIA, SOAs also played an important role as a PM 2.5 source in Zibo. Zibo is an important petrochemical industry base of China, and there were amounts of volatile organic compounds (VOCs) emissions (100 kt/year) [18] that could result in the formation of SOAs [25,65]. In addition, the percentage contribution of crustal dust was lower than that in in Tianjin (26.4%), Langfang (21.4%), Baoding (16.9%) [66], and Zibo in 2006 (29.2%) [25]. Combined with the low percentage of elements, the main reason should be that the Zibo government implemented many strict dust management policies in the last few years [67]. Although the percentage contributions of coal boilers, graphite production, aluminium smelting, glass manufacturing, and iron smelting were all lower than 3%, their contributions to gaseous precursors, such as SO 2 and NOx, could not be ignored because of the high contribution of SIA. The contribution rates were similar between the urban and suburban sites.

Risk Characterization
Because using the total concentration of heavy metals not only overestimated the human health risk but also inaccurately identified the high-risk pollution sources [68], we chose a source-specific human health risk assessment model based on the inhalation mode of exposure. The results are shown in Table 6.