Particle-bound PAHs and Chemical Composition, Sources and Health Risk of PM2.5 in a Highly Industrialized Area

PM2.5 monitoring campaigns were conducted in 2006, 2010, and 2011 in Tula, Hidalgo, Mexico, a highly industrialized area which includes a refinery, a thermoelectric power plant, five cement plants, limestone mining, and industrial waste combustion. These data establish baselines and trends against which later concentrations can be compared as emission reduction plans are implemented. PM2.5 mass, chemical composition, and 15 particle-bound polycyclic aromatic hydrocarbons (PAHs) were measured at two sites. PM2.5 masses ranged from 26 to 31 μg m–3. Carbonaceous aerosols were the largest PM2.5 component, accounting for 47–57% of the mass. Approximately 40–51% of the carbonaceous aerosol was attributed to secondary organic carbon. Ionic species accounted for 40–44% of PM2.5, with sulfate being the dominant ion. The sum of particle-bound PAH concentrations ranged from 14–31 ng m–3. Six factors derived from Principal Component Analysis (PCA) explained ~85% of the PM2.5 variance. The derived factors were associated with sources based on marker species resulting in heavy-oil combustion (22% of variance), vehicle engine exhaust (13–19% of variance), fugitive dust (18% of variance), biomass burning (9–13% of variance), secondary aerosols (14% of variance), and industrial emissions (6– 10% of variance). Combustion of solid waste (e.g., tires and industrial waste) of the recycling cement kilns and incinerators resulted in elevated toxic species such as, Cd, and Sb in the range of 0.02–0.3 μg m–3. A health risk assessment of carcinogenic trace elements was performed showing that the total cancer risk decreased for both children and adults in 2010/2011 (ranging from 3.5 × 10–6 to 6.0 × 10–5) as compared to 2006 (ranging from 8.6 × 10–7 to 5.7 × 10–6). The inhalation life-time cancer risk (ILCR) for particle-bound PAHs ranged from 8.6 × 10–5 to 1.2 × 10–4. Air quality can be improved by switching to cleaner fuels and benefit from the use of natural gas instead of fuel oil in the power plant.


INTRODUCTION
Industrial activities are important contributors to poor air quality, particularly in developing countries (Silva et al., 2021;Taiwo et al., 2014) and when many of them are clustered together. Past studies find that air pollution increases the risk of respiratory and cardiovascular diseases (Du et al., 2016), decreases the quality of life (Lee et al., 2014), alters productivity (Zivin and Neidell, 2012), escalates medical costs (van Essen et al., 2018), and generates adverse birth outcomes (Ha et al., 2014). Industrial emissions of gases, particulate matter (PM), and toxic compounds are

Sampling and Analytical Methods
The first field campaign took place from 22 March to 22 April 2006 as part of the Megacity Initiative Local and Global Research Observations (MILAGRO). For comparison, additional 24-h PM2.5 samples were also collected every third day at the same locations from 28 April to 28 of May 2010, and from 02 March to 30 April 2011.
U.S. EPA guidelines (http://sor.epa.gov) quantify the 16 PAHs as priority pollutants that are considered to be possible or probable human carcinogens. However, only 15 PAHs were reported due to the high uncertainties associated with the quantification of benzo [b]fluoranthene. For PAH analysis, microfiber quartz-fiber filters were ultrasonically extracted three times for 20 minutes each. The extract was concentrated to 1 mL in an ultra-pure nitrogen stream. Acetonitrile was added, filtered, and concentrated to 0.5 mL. A total of 120 samples were analyzed for 15 PAHs using Gas Chromatograph/Mass Spectrometry (Model 6890N GC/MS and 5973N Agilent Technologies, San José Calif. USA). Identification was confirmed by deuterated PAHs (Cambridge Isotope Laboratories, CIL) and quantification was performed using authentic standards (ChemService), with the addition of an internal standard of fluoranthene d-12. Data validation included examination of linearity with standards (r > 0.99) with a relative standard deviation (RSD) < 3%. Recovery efficiencies were between 89% and 106% by using internal standards of deuterated PAHs.

Quality Assurance/Quality Control (QA/QC)
The analytical limits of detection (LODs) ranged from 0.45 ng m -3 (Li) to 128.40 ng m -3 (Hg) for ICP/MS; 6.30 ng m -3 (Na + ) to 24.53 ng m -3 (Ca 2+ ) for IC; 0.109 µg C m -3 for carbon analysis, and 19.6 ng m -3 (BkF) to 107.93 ng m -3 (NAP) for GC/MS. Twenty field blanks were collected and analyzed, representing 14% of the total samples. Average blank values for all species were below LODs. Analytical results were corrected by subtracting the average blank. Replicates were performed for at least 10% of the samples with results of less than 5% differences. Calibration curves were acceptable when correlation coefficients were greater than 0.99. QC standards were analyzed before each sample run, after each group of 10 analyses, and at the end of each set of analysis. Precision was verified by analyzing a sample 10 times with a mixture of species. Data were submitted to three levels of data validation, as documented in Chow et al. (2002).

Mass Closure and Secondary Organic Carbon (SOC)
PM2.5 reconstruction requires indirect estimates of unmeasured species to achieve closure between measured gravimetric mass and sum of components as shown in Eq. (1) (Chow et al., 2015): PM = Inorganic ions + 1.4 × OC + EC + geological minerals + salts + non-geological elements + others (1) To examine the extent of inorganic ion neutralization, ammonium (NH4 + ) concentrations were estimated based on the stoichiometric ratios of ammonium salts and compared to the IC measurements. It is assumed that particulate nitrate is present as ammonium nitrate (NH4NO3) and sulfate is present in the form of ammonium sulfate ((NH4)2SO4), or ammonium bisulfate (NH4HSO4). Therefore, calculated ammonium equals 0.192 × sulfate + 0.29 × nitrate (Chow et al., 2015). The ion balance showed that there was not enough NH4 + to fully neutralize SO4 2-, suggesting the presence of NH4HSO4 and sulfuric acid (H2SO4). This is typical of local, rather than regional, SO4 2-, as a longer residence time usually allows the non-neutralized SO4 2to come into contact with ammonia (NH3).
OC may be of primary origin from combustion processes (e.g., wildfires, fireplaces, waste burning, and vehicles emissions) and secondary when formed in the atmosphere by photochemical reactions involving VOCs (Castro et al., 1999). EC is a primary emission from incomplete combustion of fossil fuels or biomass burning and the OC/EC ratio is often used to separate primary from secondary organic carbon (SOC) according to Eq. (2) (Turpin and Huntzicker, 1995;Ramírez et al., 2018): where organic mass (OM) is 1.4 × OC to account for unmeasured oxygen and hydrogen associated with carbon); and (OC/EC)min is the minimum ratio observed, representing primary aerosol. A (OC/EC)min ratio of 1.2 was used as it was the average of the minimum ratios observed. This value agrees with the (OC/EC)min = 1.1 measured for diesel engine exhaust (Viidanoja et al., 2002).

Source Identification
Principal Component Analysis (PCA) was applied to identify associations among the measured components and possible PM2.5 sources for samples collected in 2006 and 2011. A correlation matrix (IBM SPSS STATISTICS software V. 26) of the ambient concentrations was examined ChooChuay et al., 2020). The Varimax rotation was used to maximize (or minimize) loading factors of each species, leading to large eigenvector values (loading) toward one and small loadings toward zero. Species with communality > 0.70 and values > 0.32 (Han et al., 2006) were considered for association of factors with sources.

Health Risk Estimates
Risk analysis uses human equivalent concentrations to develop inhalation unit risks (IUR) and inhalation cancer slope factors (U.S. EPA, 2009). This approach is based on a linear extrapolation from exposures observed in animal and human occupational studies and is conservatively protective of public health (U.S. EPA, 2005). IUR is defined as the upper-bound excess lifetime cancer risk estimated from continuous exposure to an agent at a concentration of 1 µg m -3 in air (U.S. EPA, 2008). This approach has been used in recent studies on health risks of PM2.5 (Nirmalkar et al., 2021;Xu et al., 2021).
The health risk associated with carcinogenic contaminants was estimated based on the assumption that inhalation is the major exposure pathway to trace elements following the criteria established by the U.S. EPA (2011). The selected toxic elements include As, Cd, Co, Cr (VI), Ni, and Pb, which are considered possibly/probably carcinogenic to humans by the International Agency for Research on Cancer (IARC) (1990, 2006a, b, 2012). It assumed that the proportion of carcinogenic Cr (VI) to non-carcinogenic Cr (III) concentrations in ambient air is 1:6 , and the concentration of the Cr (VI) was 1/7 of the total Cr concentration (Park et al., 2008;Ramírez et al., 2020). Adults and children living in the study area were considered potential receptors. The Individual Cancer Risk (ICR) calculation is shown in Eq. (3): where ICRinhalation (dimensionless) represents individual lifetime cancer risk through inhalation of carcinogenic elements; ExCo is exposure concentration (µg m -3 ); and IUR is the Inhalation Unit Risk (µg m -3 ) -1 provided by the Integrated Risk Information System (IRIS) (https://www.epa.gov/iris), the Office of Environmental Health Hazard Assessment (OEHHA) (https://oehha.ca.gov), and the National Research Council (2000). Generally, acceptable or tolerable ICRinhalation for regulatory purposes range between 1 × 10 -6 and 1 × 10 -4 (U.S. EPA, 2011). ExCo were calculated with Eq. (4): where Ct is the carcinogenic element concentration in PM2.5 (µg m -3 ); ET is the exposure time set at 24 hours day -1 ; EF is the exposure frequency set at 365 days year -1 ; ED is the exposure duration set at 6 and 24 years for children and adults, respectively ; and AET is the  (Nisbet and LaGoy, 1992;Shen et al., 2019). URBaP is the inhalation cancer unit risk of BaP (8.7 × 10 -5 ng m -3 ) for a lifetime of 70 years exposure to 1 ng m -3 of BaP (WHO, 2000;Mehmood et al., 2020).

Meteorological Aspects
Meteorological conditions during the MILAGRO campaign have been reported by Sosa et al. (2013). Surface meteorological variables of T, RH, WP, and P showed similar patterns at the two sites with maximum T (24°C) reached in the afternoon (16:00-17:00 h, local time) and minimum T (8.5-12°C) during morning (07:00-08:00 h) (Supplemental Fig. S1). The large thermal oscillation of ~16°C is typical of the area. Solar radiation was at its maximum at 14:00-15:00 h, just before the maximum temperature occurred. Solar radiation at TEP was twice that at JAS. Maximum RH occurred at night with the minimum at 16:00-17:00 h. The minimum WS was found at night and early morning, with maxima during 19:00-20:00 h, and WS was two times higher at JAS as compared to TEP.
Maximum RH occurred at night with the minimum at 15:00-17:00 h. The minimum WS was found at night and early morning, with maximum during 15:00-18:00 h. Average WS was twofold higher in TEP as compared to JAS. Table S1 shows a decreasing trend of PM2.5 concentrations from 2006 to 2011, with a threeyear average of 26.8 and 24.0 µg m -3 at the JAS and TEP sites, respectively. Average 24-hr PM2.5 concentrations at JAS (31 µg m -3 ) and TEP (26 µg m -3 ) for 2006 were below the Mexican National Ambient Air Quality Standard (NAAQS) of 65 µg m -3 . Concentrations in JAS were reduced by 10% (2010) and 30% (2011) at JAS and 1% (2010) and 19% (2011) at TEP. During 2006, PM2.5 concentrations were associated with temperature and inversely correlated with wind speed (Sosa et al., 2013). Low WS (< 2 m s -1 ) from ENE and ESE winds at the TEP and moderate WS (2.6 m s -1 ) from the SW at the JAS were associated with the highest PM2.5concentrations. Murillo-Tovar et al. (2018) found that lower photochemical activities during cold-dry season inhibit atmospheric chemical reactions and enhance ambient PAHs concentrations, partially due to the stronger tendency of PAH to bond particles. Reduction of PAHs has been found during the rainy season due to wet deposition. In addition, it may alter the irradiance and ozone concentrations that effect photo-oxidation during the cold-dry season (Amador et al., 2020).

Chemical Composition and Material Balance
Mass and chemical composition of ~200 PM2.5 samples from the 2006 campaign were used to identify emission sources. High correlations (r = 0.9) were found between water soluble ions and the sum of measured species. PM2.5 calcium (Ca) includes both water-soluble and -insoluble oxides resulting in a low anion versus cation correlation (r = 0.62). The correlation improved considerably (r = 0.94) when Ca 2+ was removed. Tables S2 and S3 show average and maximum 12-h and 24-h PM2.5 mass and chemical composition of PM2.5 concentrations in JAS and TEP, respectively. Yearly (2006, 2010, and 2011) 10.5 26.7 10.9 784.8 OC concentration reductions were found for 2010 (65% in JAS and 70% in TEP) as compared to 2011 (40% in JAS and 13% in TEP). Reductions in EC concentrations were similar at both sites with > 30% in 2010 and ~45% in 2011. This is probably due to the compulsory switching from fuel oil to natural gas during 2010 at the electric power plant (Sosa et al., 2020).

Water-soluble ions
Ionic species were higher at JAS than at TEP, accounting for 44% of the PM2.5 in 2006 (with ~24% of anions and ~20% of cations). This fraction was reduced to 27% of the PM2.5 in 2011 (with ~19% of anions and 8% of cations). Year-to-year variations of ionic species were not found at the TEP site. SO4 2was the dominant ion, with a maximum of 14.9 µg m -3 at JAS during daytime on 09 April 2006 (accounting for 33% of the PM2.5 mass) and 15.4 µg m -3 at TEP during daytime on 11 April 2006 (accounting for 44% of the PM2.5 mass). The reduction of 2-5 µg m -3 of SO4 2from 2010 to 2011 reflect changes from fuel oil to natural gas combustion.
Average 24-h SO4 2concentrations (6.4-7.5 µg m -3 ) at TIC in 2006 were higher than those found in the highly industrialized areas of Salamanca (5.3 µg m -3 ), and Cadereyta (4.3 µg m -3 ) in Central and Northern Mexico (Vega et al., 2007). This is consistent with the transport and transformation of SO2 to SO4 2from north to south.
At JAS, the carbonaceous fraction in PM2.5 consisted of 25.6% OC and 11.4% EC, over twofold higher during the daytime (9.6 µg m -3 ) than at nighttime (4.1 µg m -3 ). Average 24-h OC and EC concentrations were 8.0 and 3.4 µg m -3 , similar to those at TEP, with OC/EC ratios of 2.7. The extent of secondary aerosol formation and industrial emissions may affect the OC:EC ratios. Table 3

Geological material, salts, and trace elements
In 2006, geological material accounted for ~18.5% of 12-h PM2.5 (daytime and nighttime) at JAS, consistent with nearby quarrying activities. Higher daytime (8% of PM2.5) than nighttime (3%) geological contributions were found at TEP. Contributions from geological material were reduced to ~10% in 2010 and ~8% in 2011 at both sites.
Average Ca in the form of calcium oxide (CaO), attributed to the nearby limestone mining and quarries, accounted for 62% of 24-h PM2.5 geological material. The highest 12-h CaO concentrations (7.5 and 7.3 µg m -3 ) were found for daytime on 11 April and nighttime on 7 April 2006 at the JAS, accounting for > 31% of the geological material. This was likely due to prevailing northerly to northeasterly winds (4-9 m s -1 ) during daytime hours with direct influence from upwind cement processes. Salts contribution were low, ranging from 0.38 to 0.71% of PM2.5. Trace elements contributions ranged from 0.3 to 0.5% of 24-h PM2.5 mass at the JAS and 0.9-1.7% at the TEP (Table 4).
Since 1996, the cement industries were allowed to burn both recycled alternative fuels and industrial wastes (e.g., coke, coal, tires, and solid and liquid hazardous waste), leading to higher emissions of metals and organic compounds (Jacott et al., 2003;Arfala et al., 2018). The use of alternative fuels reached 5% of the total fuel consumption in 2006 (Semarnat, 2008(Semarnat, , 2016. Coprocessing of industrial hazardous waste (e.g., tires and infectious biological waste) were authorized to 30 plants in 2010 (Semarnat, 2010). These toxic emissions may create health risks to local populations and to the environment (Cangialosi et al., 2008).
Under the regulation of the Ministry of the Environment and Natural Resources (DOF, 2004), NOM-040-SEMARNAT-2002, the maximum air emission levels are 700 ng m -3 for the following ten elements: Sb, As, Se, Ni, Mn, Cd, Hg, Pb, Cr and Zn from a cement plant. Tables S2 and S3 show that maximum 24-h PM2.5 concentrations were 2,926 ng m -3 for Cr and 6,388 ng m -3 for Fe  (Lucarelli et al., 2019), although Pb may also be originated from other sources such as leaded gasoline, gasoline spills, lead smelting, coal combustion, and paint materials (Das et al., 2018). In 2006, maximum 24-h Zn concentrations were 16 and 24 ng m -3 , while maximum Pb concentrations were 60 and 785 ng m -3 at the JAS and TEP sites, respectively. Large reductions of toxic species were found from 2006 to 2011: from 1,015 to 9 ng m -3 for Cr; from 1,990 to 6 ng m -3 for Pb in TEP; and from 15 to 5 ng m -3 for Cr; and from 20 to 3 ng m -3 for Pb in JAS. However, average 24-h Zn concentrations increased from 17 ng m -3 in 2006 to 26 and 31 ng m -3 in 2011 at the JAS and TEP sites, respectively.
Concentrations of other toxic air pollutants such as Cd, As, and Sb were low. Cd is a known human carcinogen, it causes damage to the lungs, kidneys, and bones (Satarug et al., 2010). Average 24-h Cd concentrations were 8 ng m -3 in 2006 and decreased to 0.3 and 0.2 ng m -3 at the JAS and TEP sites, respectively. Maximum 24-h Hg concentrations from hazardous waste incinerator and a petrochemical plant reached 17.5 ng m -3 at JAS and 33.9 ng m -3 at TEP. Brake wear contains Fe, Cu, Ba, and Sb (Hagino et al., 2016). A Cu/Sb ratio of 3.8 in this study, is consistent with the ratio of 4.6 by Hagino et al. (2016) used to identify brake wear.

Comparison with nearby measurements
Figs. 2 and 3 compare the major PM2.5 ionic and metal concentrations, respectively, from the three sampling campaigns with those reported for PM10 from the TIC (Querol et al., 2008;Martínez-Carrillo et al., 2010). Collected PM10 samples were simultaneously collected at the JAS and TEP sites during the 2006 campaign (Querol et al., 2008). Samples from Tlaxcoapan (TLA, a small agricultural town located 7.5 km NE from the TIC) were collected during July-December 2007 (Martinez-Carrillo, 2009). In addition, PM10 samples were collected at the T2 site (in a ranch ~50 km east of TIC at an elevated height [200 m]) from the MILAGRO campaign (Querol et al., 2008).
The upper panel of Fig. 2 shows that the PM2.5 fraction was mostly composed of Si and SO4 2-, ~one to three orders of magnitude higher than K and Ca. The relatively high concentrations of the crustal element K in PM2.5 at both sites is consistent with fugitive dust contributions. In contrast, Ca is the major species in the PM10 fraction at TLA and JAS, consistent with the operation of limestone processing. The relatively high abundance of Ca in PM10 suggests that coarse particle (PM10 minus PM2.5) Ca is mostly produced by limestone quarries and cement industry.
The metal components of both PM2.5 and PM10 samples in Fig. 3 at JAS showed elevated V and Zn. At TEP, Pb and Cr were abundant in both the coarse and fine particle sizes, consistent with anthropogenically derived activities. PM10 ionic and metal concentrations from site T2 (Querol et al., 2008) were low, reflecting less impact from the anthropogenic activities.

Average PM2.5 PAH Concentrations
The average of total PAHs (sum of 15 PAHs) at JAS (17.5 to 185 ng m -3 ) was 18% higher than that at TEP. Table 5 shows that total PAHs were lower than 50 to 310 ng m -3 reported at trafficintersections in Mexico City (Marr et al., 2004). Elevated PAHs were found for acenaphthylene (ACY), fluorene (FLU), anthracene (ANT), naphthalene (NAP), and acenaphthene (ACE). Maximum benzo[a]pyrene (BaP), a marker for fossil fuels combustion, petroleum cracking, and vehicular emissions (Zhang et al., 2019), was 3.8 and 2.8 ng m -3 at JAS and TEP, respectively. On-road studies in Mexico City showed good associations between benzo[ghi]perylene (BgP) from gasoline vehicles emissions and morning rush hours (Dzepina et al., 2007). The results found in this study are lower than those reported in MILAGRO, with an average total PAHs concentration of 114.0 ng m -3 north of Mexico City (site T0, Instituto Mexicano del Petróleo), and 7.0 ng m -3 South of Mexico City (Pedregal) (Thornhill et al., 2008). Table 6 compares PAH concentrations with those from urban or industrial areas in U.S.A., China, India, Spain, and Kenya. Except for the elevated concentrations at Nairobi, Kenya, PAHs from this study are one to several orders of magnitude higher than the industrial areas in other countries.
The  Table 7 summarizes the eight factors identified by PCA, explaining 91-92% of the total PM2.5 variance, with the inclusion of PAHs in 2006. Cement production emissions were the major factor representing 29% of the total variance, on average. This factor is characterized by the presence of OC, EC, Co, and Ni as markers of heavy-oil combustion processes . Recycled alternative fuels, including tires and industrial wastes, explain the high factor loadings of Hg, Cd, Sb, Cr, Pb, Cu, and Zn (Hua et al., 2016). The second factor (15-17% of variance) is associated with mixed industrial emissions marked by PAHs such as FLU, PHE, and ANT, representing emission from coking, steelworks, and use of gas . The third factor (14% of variance) is characterized by markers of soil and fugitive dust such as Fe, Mg, Si, Ca and Al. The presence of Ca 2+ indicates contributions from limestone quarries (Galindo et al., 2011). The fourth factor is associated with vehicle engine exhaust emissions (11-17% of variance), characterized by PAHs such as FLA, BaA, PYR, CHR, and BkF, indicative of vehicle emissions . The fifth  Yang et al., 2010;c Rajput et al., 2009;. d Ramirez et al., 2011;e Muendo et al., 2006. factor (6-7% of variance) is characterized by secondary aerosols mixed with markers of oil combustion such as V and Ni at Jasso (Corbin et al., 2018). The sixth factor (5-6% of the variance) is associated with biomass combustion with markers such as NO3 -, BaP, BkF, Cl -, Pb, and K + at Jasso (Manousakas et al., 2017;Wang et al., 2020), and OC, EC, and K + at TEP (Bernardoni et al., 2011). The seventh factor (3-4% of variance) is characterized by crude oil with markers such as V and Ni (Manousakas et al., 2017). The eighth factor (3-6% of variance) represents the metal manufacturing industry, associated with As and Cu at JAS, and Pb, Cl -, and Sn at TEP (Hedberg et al., 2005;Taiwo et al., 2014).

Principal Component Analysis
When PAHs are deleted from the PCA analysis, Table 8 shows that five factors were identified, explaining 86-88% of the total PM2.5 variance in 2006. Cement kiln emissions remain as the major factor, explaining 36% of the total variance, with markers of heavy-oil combustion (Co, Ni, OC and EC) and recycled alternative fuels (Hg, Cd, Sb, Cr, Pb, Cu, and Zn) . The second factor (21% of variance) is fugitive dust with markers such as Fe, Mg, Si, Ca 2+ , and Al. The third factor (15% of variance) is heavy-oil combustion processes with markers such as SO4 2-, V, and Ni (Manousakas et al., 2017;Corbin et al., 2018). The fourth factor (5-9% of variance) is characterized by Cland Pb, associated to recycled alternative fuels and metal manufacturing industries at TEP (Galindo et al., 2011;Taiwo et al., 2014), and Cu related to metal -mechanic industries at JAS (Taiwo et al., 2014). The fifth factor (8% of variance) is biomass combustion characterized by OC, EC, and K + at TEP (Bernardoni et al., 2011), and NO3 -, Cl -, Pb, and K + at JAS (Manousakas et al., 2017;Wang et al., 2020). Table 9 summarizes PCA identified factors, explaining 84-85% of the total PM2.5 variance in 2011. Oil combustion is the major factor, representing on average 22% of the total variance. This factor is characterized by V and Ni (Manousakas et al., 2017;Corbin et al., 2018;Khan et al., 2021), originating from the refinery and thermoelectric plants. The presence of Mo at JAS and TEP suggests emissions from catalysts in the petrochemical process (Minocha and Goyal, 2013).
The presence of Ca 2+ shows contributions of cement plants, particularly at TEP (Gupta et al., 2012). The second factor (13-19% of variance), is associated with vehicle engine exhaust emissions, marked by the high contribution of OC and EC. Zn is used as an additive in lubricating oil (Liu et al., 2017). The presence of Zn, Sb, Cd, and Cu are associated with non-exhaust emissions from road traffic (Taiwo et al., 2014;Hsu et al., 2016), and elements such as Fe, Al, Si, and Ca are related to road dust resuspension (Bernardoni et al., 2011). The third factor (17-18% of variance) is characterized by markers of soil and fugitive dust such as Si, Mg, Al, and Fe (Martínez-Carrillo et al., 2010;Khan et al., 2021). The presence of Ca 2+ indicates contributions from limestone  quarries, particularly at JAS (Sharma and Pervez, 2004). The fourth factor (9-13% of variance) represents biomass burning. This factor was mainly associated with K + , along with As, Rb, NO3 -, and Zn (Hedberg et al., 2005;Manousakas et al., 2017;Lin et al., 2018;Wang et al., 2020). The fifth factor (8%) in TEP is related to cooking activities. The presence of OC could be attributed to the fumes emitted from the oil-based cooking, whereas SO4 2and NO3could be related to fuel combustion (Sun et al., 2020). The presence of Na + has been associated with salts used in cooking (Sun et al., 2020). In JAS, the fifth factor (14%) was related to secondary inorganic aerosols characterized by NH4 + , SO4 2-, and NO3 - (Khan et al., 2021). These elements may trace vehicular and industrial emissions, which undergo atmospheric transport and transformations from gas to particle (Liu et al., 2017). The sixth factor (6-10%) is associated with industrial emissions. Pb, Cr, and Sb can be released by cement plant and metal manufacturing processes in TEP (Taiwo et al., 2014;Hua et al., 2016), while Cr, Co, NO3 -, and Clmay originate from waste incinerator plants and metal industries in JAS (Taiwo et al., 2014;Liu et al., 2017;Lucarelli et al., 2019).
Separation of a mixture of pollution sources presents a challenge for factor analysis methods, including PCA and Positive Matrix Factorization (PMF) owing to profile mixing when measured compounds are highly correlated (Watson et al., 2016). Vehicle engine exhaust is clearly identified with the inclusion of PAHs, along with a higher percentage (91-92%) of variance explained (Table 7). Other factors contributing to air pollution at TIC for the 2006 and 2011 samples includes cement kiln emissions with heavy-oil and recycled alternative-fuel combustion, biomass burning, metal industry, and fugitive dust.

Health Risk Assessment
A health risk assessment was conducted with average As, Cd, Co, Cr (VI), Ni, and Pb, concentrations representing carcinogenic elements in PM2.5 (Table 10)  reported in areas influenced by industrial emissions (Cheng et al., 2017;Liu et al., 2018;Ramírez et al., 2020).
The ICR values for Cd (for both children and adults) and Co (only for adults) were also slightly higher than the minimal acceptable risk level, ranging between 1.1 × 10 -6 and 4.3 × 10 -6 in 2006 (Table 10). However, exposure concentrations of Cd and Co were noticeably reduced during 2010/2011, resulting in negligible values of carcinogenic risk (below 1 × 10 -6 ) (U.S. EPA, 2011). Overall, the total cancer risk decreased for both children and adults in 2010/2011 (between 3.5 × 10 -6 and 6.0 × 10 -5 ) compared to 2006 (between 8.6 × 10 -7 and 5.7 × 10 -6 ), which reflects the health benefits from switching of fuel oil to natural gas in the power plant.
BaPeq, a parameter used to evaluate the human health risk (Shen et al., 2019), exhibited 35% higher value in JAS as compared to TEP (Table 11). These BaPeq values (0.99-1.34 ng m -3 ) are similar to those measured in Shanghai during winter (0.916-1.86 ng m -3 ) (Yang et al., 2021), but higher than values found in cities of Pakistan (< 0.24 ng m -3 ) (Ishtiaq et al., 2021). Average BaPeq in TIC were lower than that reported in northern China during winter (3.16-120 ng m -3 ) (Shen et al., 2019) and in Islamabad (5.19-10.61 ng m -3 ) (Mehmood et al., 2020). Average BaPeq of 1.34 ng m -3 in JAS was above the standard of 1 ng m -3 defined by WHO, representing a health risk for the exposed population.
In addition, the ILCR was 8.6 × 10 -5 and 1.2 × 10 -4 at TEP and JAS (Table 11), respectively, indicating ~9 cancer cases can occur per 100,000 inhabitants at TEP and 2 cases can occur per 10,000 inhabitants at JAS. Both ILCR values exceeded the threshold value (1 × 10 -6 ) suggested by U.S. EPA, but the risk in JAS stands out since it is two orders of magnitude higher than the U.S. EPA acceptable cancer risk level. The results suggest the importance of periodically monitoring the risk by inhalation of atmospheric PAHs in the TIC.

CONCLUSIONS
This study characterized over ~320 PM2.5 mass and chemical composition samples acquired at the two Central Mexico Tula Industrial Corridor sites (Jasso [JAS] and Tepeji [TEP] during the drywarm months (March-May) in 2006, 2010. A decreasing trend was found for PM2.5 mass with 30% and 19% reduction from 2006 to 2010 at the JAS and TEP sites, respectively. PM2.5 mass in 2006 were 31.03 ± 0.93 µg m -3 (JAS) and 25.72 ± 0.26 µg m -3 (TEP), consisting of 47-57% of carbonaceous aerosol (OC × 1.4 + EC) and 40-44% of ionic species demonstrated by SO4 2-. While most of the SO4 2was in the form of ammonium sulfate, the ion balance showed that there was not enough NH4 + to neutralize SO4 2-, suggesting the presence of ammonium bisulfate and sulfuric acid of nearby origin.
The sum of the 15 PAHs ranged from 14-18 µg m -3 , lower than the 50-310 µg m -3 range reported in Mexico City and orders of magnitude lower than those found in U.S.A., China, India, and Spain. Geological minerals contributed ~8-10% of PM2.5 mass with abundant (> 60%) Ca in JAS, attributing to nearby limestone mining and quarries.
Air pollution concentrations were reduced from 2006 to 2010/2011 as a result of implementing air pollution control measures, such as the substitution of industrial fuel to natural gas that reduces industrial emissions and minimize health risks of the exposed population. Thus, the cancer risk of toxic trace elements in 2010/2011 was reduced by an order of magnitude relative to 2006. However, it is crucial to monitor the health risk from inhalation of PAHs, especially in JAS where a higher risk was found.
This study showed that heavy-oil combustion and vehicle engine exhaust are major sources of PM2.5. Therefore, it is necessary to use cleaner fuels to improve air quality. Implementing control measures at the TIC would lead to less transport downwind that impact Mexico City and the State of Mexico.
Although PM2.5 concentrations at JAS and TEP during the study periods were within the Mexican National Ambient Air Quality Standard, it is recommended to continuously monitor the chemical composition and conducts source apportionment of PM2.5, since elevated PAHs and toxic compounds such as Cd, Cr, and Pb consistently posing health risks.