Organic Aerosol Characterization and Source Identification in Karachi , Pakistan

With its rapidly growing population and large industrial base, the megacity of Karachi, Pakistan, has been subjected to an increasing amount of ambient particulate matter (PM). Fine particulate matter (PM2.5) in Karachi was collected every 24 hours from January 8 to January 29, 2006. The samples were extracted and analyzed by gas chromatography-mass spectrometry (GC-MS) and ultra-performance liquid chromatography (UPLC) coupled with a triple quadrupole mass spectrometric detector (TQD) and time-of-flight (TOF) mass spectrometer. The daily PM2.5 levels ranged from 99.5 to 251.1 μg m with a mean of 177.4 μg m, averaging 4–10-fold higher than the WHO guideline for 24-hour averaged PM2.5 (25 μg m). We found that the day-of-the-week variations of PM2.5 demonstrate Sundays have significantly lower concentrations (with a t-based confidence level of 95%), indicating that weekly behavioral patterns affect local PM2.5 concentrations. A significant negative correlation was found between the daily concentrations of levoglucosan, a biomass burning tracer, and the average daily temperatures (r = –0.589, p = 0.004), implicating heating as a major source of biomass burning emissions. Results indicate that polyaromatic hydrocarbons, hopanes, steranes, and alkanes are mainly emitted from fossil fuels and the combustion of carbonaceous materials. Organosulfates (OSs) and sulfonates were also quantified; significant correlations between OSs and sulfonates indicate a common source and/or similar formation mechanisms, while correlations with hopanes, steranes, and levoglucosan suggest their emission from primary sources such as fossil fuels and biomass burning. Qualitative analysis suggests the presence of a C6–C9 alkyl sulfate series. Through this study, the chemical composition and origin of the organic fraction of PM2.5 in Karachi has been evaluated for the first time, and the results indicate a strong anthropogenic influence on combustion emissions, particularly those from biomass and fossil fuel burning.


INTRODUCTION
Air pollution has been a growing problem in megacities undergoing urban development and industrialization.Contributing to air pollution is fine particulate matter (PM 2.5 ), which pertains to solid or liquid particles suspended in the atmosphere equal to or less than 2.5 microns in aerodynamic diameter.Studies have demonstrated that there is a causal link between higher levels of air pollution (specifically total suspended particulate matter, ozone, and on source apportionment of PM 2.5 in Karachi during January 2006-January 2008 using a Positive Matrix Factorization (PMF) based on trace metal species; they identified five main factors (reported with their estimated contributions to PM 2.5 ): soil and road-side dust (16.1%); industrial emissions including iron and steel industry or ferrous smelters (13.3%); vehicle sources containing two-stroke engines and battery manufacturing cottage metal sources (18.5%);secondary aerosols (12.4%); and oil/coal burning in industries, vehicles, and thermal power plants (39.7%).Parekh et al. (1987) also used a factor analysis based on trace elements to identify the sources of ambient air aerosol at a suburban site of Karachi in July 1985.They identified four major sources of PM 2.5 including sea spray (12%), soil dust (28%), cement (20%), and fossil fuel sulfate (3%), while 36% of the mass was unassigned.A study by Begum and Tabassum (1991) attributed sources of PM in Karachi by using dry deposition to sea spray, solid waste disposal incineration, fuel combustion, industrial effluents, soil dusts, and locomotive exhaust smoke, utilizing water-soluble inorganic ions as tracers.They found the concentration of sodium ion to be the highest, indicating a contribution from sea spray.Alam et al. (2011) identified combustion, re-suspension of road dust, and sea salt aerosols as sources of PM in Karachi during March and April 2010 by the analysis of crustal elements (Al, Fe, Si, Mg, and Ca) and trace elements (B, Ba, Cr, Cu, K, Na, Mn, Ni, P, Pb, S, Sr, Cd, Ti, Zn, and Zr).These studies of PM sources in Karachi, focused mainly on determining crustal and trace elements of ambient air and using them to identify various sources of PM which include road dust, steel industry vehicles, secondary aerosols, and industrial oil burning as dominant sources of the inorganic PM fraction.
Current knowledge on organic PM in Karachi is limited to one study by Shahid et al. (2016).The organic carbon (OC) concentrations in Karachi were 9.7 µg m -3 in PM 2.5 and 21.8 µg m -3 in PM 10 , accounting for 13% of PM 2.5 mass and 5% of PM 10 mass.Elemental carbon concentrations were also 1.93 µg m -3 and 4.72 µg m -3 for PM 2.5 and PM 10, respectively, accounting only for 3% PM 2.5 mass and 1% PM 10 mass.The PM 2.5 /PM 10 ratio for OC in Karachi is 0.44, indicating that approximately 40% of OC in Karachi is within the fine particle fraction (Shahid et al., 2016).Using levoglucosan as a tracer, biomass burning was estimated to account for only 2.9% of PM 2.5 and 1% of PM 10 .Arabitol, a tracer for fungal spores contributed only 0.2% of PM 2.5 and 0.1% of PM 10-2.5 .The major contributor to PM mass was mineral dust, which contributed 46% of PM 2.5 and 78% PM 10-2.5 .Because of high levels of organic matter (21% of PM 2.5 ) and elemental carbon (6.5%), other fossil fuel combustion sources were expected to be significant.Sea salt contributions were found to be minor, around 2% (Shahid et al., 2016).Further elucidation of the sources of organic carbon in PM 2.5 is the focus of this study.
As shown, the majority of research on PM in Karachi has been conducted on the metals component of PM 2.5 and PM 10 , with fewer studies conducted on the organic fraction.Indeed, organic matter is an important component of PM, with an OC fraction of 13% in PM 2.5 and 5.0% in PM 10 as seen in Shahid et al. (2016).While the importance of combustion and fossil fuel in PM has been established, this study will link to investigate fossil fuel and biomass burning sources by way of organic tracers.The measurement of organic tracers can provide insight into the sources of particlephase organic matter (Zhang et al., 2010).This study mainly focuses on detecting and measuring the organic components of PM 2.5 and identifying their sources in Karachi, Pakistan.Our first objective is to use specific PM 2.5 organic markers to identify sources and analyze their day-to-day variations.This includes levoglucosan, a tracer for biomass fuel burning (Simoneit et al., 1999); polycyclic aromatic hydrocarbons (PAH), tracers for combustion and carbonaceous fuel (Moeinaddini et al., 2014); and hopanes and steranes, which are fossil fuel markers (Schauer et al., 1999;Lough et al., 2007).Additionally, organosulfates (OSs) and sulfonates are a class of organic compounds considered to make a fractional contribution to total PM 2.5 levels, and whose sources, formation mechanisms, composition, and secondary organic aerosol contribution still need to be further explored.Therefore, our second objective is to perform various statistical and correlation analyses between aromatic OSs, sulfonates and various organic tracers as well as between OSs and sulfonates themselves.In addition, high-resolution time-of-flight (ToF) mass spectrometry (MS) was used to identify the major sulfur-containing organic species in Karachi.Understanding the OS and sulfonate contribution to PM 2.5 mass and various correlational significances will help us to better understand their sources and formation mechanisms.

Sample Collection
The megacity Karachi (Fig. 1) is the most urbanized, industrialized, and affluent city in Pakistan.The population density varies from the central city (33,014 persons km -2 ) to the outskirts (433 persons km -2 ).Being Pakistan's largest port city and its business and manufacturing capital, Karachi is also the home to the largest number of vehicles (> 3.6 million).It has a large industrial base in and around the city.Three major industrial areas, Landhi, Korangi, and Sindh Industrial Trading Estates (S.I.T.E), are shown on the map (Fig. 1).Sampling site was M. A. Jinnah Road (commercial/residential) in "Saddar" (Fig. 1), which is located in the midst of a large central business district.Approximately 300,000 vehicles pass M. A. Jinnah Road daily.Sampling was carried out on the curb of this road at a height of 2 m above the ground, i.e., almost the breathing level of a pedestrian.
Samples of PM 2.5 were collected on 47 mm quartz fiber filters using a cyclone separator.The PM 2.5 sampler consisted of a gooseneck, 5.72 cm inner diameter rubber stopper, 47 mm filter holder, pump, and an aluminum cyclone separator (URG Corporation, Chapel Hill, NC, USA) with a cut size of 2.5 µm operated at a flow rate of 16.7 L min -1 .The resulting total sampled air volume was ~24 m 3 .Prior to sampling, 47 mm quartz fiber filters were baked at 550°C for 24 hours to remove all organic materials.PM 2.5 samples were collected for every 24 hours starting from January 8-29, 2006.At the end of each 24-hour period, filters were carefully removed from the sampling device, wrapped in aluminum foil, and were placed inside individual polyethylene petri-dishes.After sample collection, filters were stored refrigerated (4°C) in the dark until shipped to the University of Iowa for analysis in November 2014 after which they were stored frozen (-20°C) in the dark until extraction for GC-MS analysis in March 2014 and extraction for LC-MS analysis in June 2015.There were twenty-two filter samples collected and processed in total.Three field blanks were collected.Four laboratory filter blanks were analyzed for organic species.

PM Mass Determination
The laboratory for weighing PM 2.5 was maintained under "clean room" conditions to minimize contamination, airflow interferences, and electrostatic charges.The quartz fiber filters, both before and after sampling, were conditioned for a minimum of 24 hours in a temperature (20-23°C) and relative humidity (RH, 20-30%) controlled room.The filter weights were obtained on a Micro-Analytical Balance (Model C-44, Mettler).Owing to the fragile nature of quartz filters, stringent quality control/quality assurance guidelines were followed for weighing, sampling, and storage of filter samples.PM 2.5 mass concentration was calculated in µg m -3 as the difference in filter weight before and after sampling divided by the total sampled air volume.Field blanks did not show any mass loading, so blank subtraction was negligible.

Analysis of Organic Species by Gas Chromatography-Mass Spectrometry (GC-MS)
The quartz filter samples were cut in half, spiked with isotopically-labeled internal standards, and then extracted sequentially into two 20 mL aliquots of hexane (Optima, 99.9%) and acetone (CHROMASOLV ® for HPLC, 99.9%) with the assistance of ultra-sonication (Barnstead 5510).Each extract was concentrated to 100 µL using a Turbovap LV Concentration Workstation (Caliper Life Sciences).Instrumental analysis by GC-MS followed the methods described by Stone et al. (2012).Briefly, PAH, alkanes, hopanes, and steranes were quantified directly from the final extract using electron ionization, while levoglucosan and cholesterol were quantified following silylation derivatization in positive chemical ionization mode with methane as the reagent gas (Airgas, 99.999%).
Fourteen PM 2.5 samples from Karachi, Pakistan, from a total of twenty-two sampling days in January 2006 were successfully extracted according to the method described in Hettiyadura et al. (2015).This includes samples taken on January 8-10, 12-16, 24-29.Briefly, the filter samples were extracted into 10 mL of 95:5 acetonitrile: ultra-pure (UP) water by sonication for 20 minutes.Extracts were then filtered through PTFE syringe filters (0.2 µm pore size) then blown to ~500 µL under ultra-high purity nitrogen gas (5 psi) at 50°C using a Turbovap LV Concentration Workstation (Caliper Life Sciences).These evaporation conditions have previously been demonstrated to be effective in removing solvent while providing recoveries of organosulfates in the range of 83-121% (Hettiyadura et al., 2015).The extracts were then transferred to liquid chromatography vials with the addition of ~450 µL of the 95:5 acetonitrile:UP water solution and then blown to dryness at 50°C under a light stream of nitrogen in a Reacti-Therm heating module (Thermo Fisher Scientific).Extracts were re-constituted in acetonitrile:UP water (95:5) to a final volume of 300 µL.
The sample extracts were analyzed using ultra-performance liquid chromatography (UPLC, Waters ACQUITY) coupled with a triple quadrupole mass spectrometer (TQ MS, Waters ACQUITY).Analytes were separated on a reversed phase C 18 column (ACQUITY UPLC HSS T3, 2.1 mm ID × 75 mm length, 1.8 µm particle size) at 45°C using aqueous (100% water; A and C) and organic mobile phases (100% acetonitrile; B).The mobile-phase conditions were adapted from a previous method (Kundu et al., 2013).The mobile phase was generated from three channels: A) 50 mM aqueous pH 5 buffer made from ammonium acetate and acetic acid, B) acetonitrile, and C) ultrapure water.Channel A continuously comprised 20% of the mobile phase in order to maintain a constant buffer concentration, while channels B and C varied.Channel B was held at 0% until 2 minutes, then increased linearly to 70% at 10 minutes, held at 70% until 10.2 minutes, and then decreased linearly to 0% at 12 minutes, and was held until 16 minutes.Channel C was held at 80% for two minutes, then decreased linearly to 10% at 10 minutes, held at 10% until 10.2 minutes, increased linearly to 80% at 12 minutes, and then held until 16 minutes for column reequilibration.Sample injection volume was 10.0 µL.Flow rate was 0.3 mL min -1 .
The organosulfur compounds were detected using a TQ MS operated in electrospray ionization (ESI) negative mode.ESI parameters include capillary voltage of 2.70 kV, cone voltage of 50.00 V, source temperature of 150°C, desolvation temperature of 450°C, cone gas flow rate of 100 L hr -1 , and desolvation gas flow rate of 900 L hr -1 .Sample extracts were analyzed in multiple reaction monitoring (MRM) mode.In MRM, the first quadrupole scan deprotonated molecular ions ([M-H] -) of interest, the second quadrupole is the collision cell in which [M-H] -fragmented into product ions by colliding with a neutral gas (Ar), and the third quadrupole scan specific product ions.The utilized MRM transitions and optimized MS conditions (cone voltages and collision energies) for each organosulfur compound used in this study are summarized in Table 1.Fragmentation patterns for the sulfonates are shown in Fig. S1.The transition with the greatest detector response was used for quantitation, and a second transition was used qualitatively.All data were processed using MassLynx (v.4.1).The instrument was calibrated each time using freshly prepared calibration standards.The calibration curve consisted of 7 points ranging from 1 to 80 µg L -1 .A 5-point curve from 1 to 30 µg L -1 was used for analytes that proved to be nonlinear at the extended range.Limits of detection (LOD; 3 times the standard deviation) and limit of quantitation (LOQ; 10 times the standard deviation) were determined via 7 consecutive injections of a 2 µg L -1 standard.Detection limits for each organosulfur compound used in this study are given in Table 1.All compounds were detectable to 2 µg L -1 or lower, and quantifiable as low as 1 µg L -1 (benzyl sulfate), and as high as 8 µg L -1 (p-toluene sulfonate, and 3-methylbenzyl sulfate).
The qualitative analysis of organosulfur compounds was performed using a UPLC coupled with a ToF MS (Waters ACQUITY).The nonpolar C 18 column, materials, sample preparation, extraction, and separation methods used for the qualitative analysis were essentially identical to that of TQD analysis.The ToF MS was operated in the ESI negative mode.Instrumental parameters were capillary voltage 2.8 kV, cone voltage 35 V, source temperature 110°C, desolvation temperature 400°C, cone gas flow rate 30 L hr -1 , desolvation gas flow rate 700 L hr -1 .Data were collected on a mass range of 100 to 400 in V geometry.Val-Tyr-Val (a small peptide) was used for mass correction.MassLynx software was utilized in peak identification.

Statistical Analysis
Statistical and correlation analysis involved subjecting each data set to an Anderson-Darling test for normality.Statistics were considered significant at the 95% confidence interval (p ≤ 0.05).As many data sets were not normally distributed, data were log-transformed to produce a normal distribution.The Pearson correlation test was then used to evaluate correlations.The degree of correlation was classified so that 0.0-0.3 was considered negligible, 0.3-0.5 low, 0.5-0.7 was considerate moderate in correlation strength, while 0.7-0.9 was considered high/strong, and 0.9-1.0very high (Mukaka, 2012).In addition, t-tests were performed to determine differences between sets of data.

RESULTS AND DISCUSSION
PM 2.5 Mass PM 2.5 concentrations ranged from 99.5 to 251 µg m -3 with an overall mean of 177 µg m -3 .Daily levels of PM 2.5 are high with a significant temporal variability, where some days have sharp peaks of pollution levels exceeding 200 µg m -3 (Fig. 2(a)).PM 2.5 concentrations indicate a general weekly pattern, with maxima occurring on weekdays and minima occurring on weekends (Sundays: January 8, 15, 22 and 29).Extremely elevated PM 2.5 concentration was observed on Saturday, January 28, which may be linked to heavy vehicular traffic and a hazy day.Table 2 summarizes all analytes and their measured concentrations.These levels exceed the WHO and EPD guidelines for 24hour averaged PM 2.5 of 25 and 35 µg m -3 , respectively, by a factor of more than seven and ten times and also exceed reported PM 2.5 24-hour averages from March-April 2009 in Shahid et al. (2016) by one and a half times.This comparison accordingly gives rise to the conclusion that seasonal changes (with PM mass being more predominant in the dry winters) may potentially influence PM mass concentrations.As expected, our measurement averages exceed those reported by Alam et al. (2011) by almost three times the concentration as seen at a site near the sea coast in Karachi and almost one-and-a-half times greater than at another site located on a rooftop at a less urban location in Karachi.In contrast, a reported PM 2.5 24-hour average near a high-traffic site exceeds our study's measurement by only 7 µg m -3 , suggesting that PM mass varies directly with the levels of nearby sources of anthropogenic activity.Lastly, compared to PM 2.5 24-hour averages reported in Mansha et al. (2012), our reported average is a factor of about two higher.These samples were collected at ground pedestrian level in an urban site in Karachi, compared to a high-traffic site where the samples of our study were collected.Levels of PM mass were also higher than PM 2.5 daytime average reported in Wang et al. (2007) by a factor of almost one-and-a-half times higher on a rooftop in Nanjing, China (Wang et al., 2007).

Primary Sources
Amongst the organic species that are associated with primary sources are polycyclic aromatic hydrocarbons (PAHs), hopanes, steranes, alkanes, and levoglucosan (Figs.2(b)-2(f)).These molecular markers of primary sources were quantified to identify sources of PM 2.5 in Karachi.Foremost, PAHs are a class of compounds that are formed by incomplete combustion of carbon-containing materials (Fig. 2(b)) (Collins et al., 1998).Compounds that are classified as PAHs are tracers for combustion and carbonaceous fuel (Moeinaddini et al., 2014).Select species of PAHs have been established as human carcinogens, specifically, benzo(a)pyrene (Collins et al., 1998).For this reason, benzo(a)pyrene will be of specific focus in this study, including the degree to which its concentration varies in comparison to the similar bustling city of Lahore, Pakistan, and Nanjing, China.
Benzo(a)pyrene 24-hour average in Karachi was measured to be 2.67 ng m -3 , with a range from 1.37-6.41ng m -3 .Measurements by Stone et al. (2010) indicate that benzo(a)pyrene in Lahore was found at three times the concentration as compared to Karachi, with reported 24hour average concentrations to be around 7.70 ng m -3 in January of 2007, and annual monthly average concentrations ranged from 2.00-13.0ng m -3 .These particular samples were collected over a 24-hour sampling period on four days in January (Stone et al., 2010).Similarly, reported average concentrations in Nanjing, China, during winter also show higher levels of benzo(a)-pyrene by a factor of  about two times compared to Karachi (Wang et al., 2007).These measurements show that Karachi has lower ambient levels of this particular PAH compound relative to other similarly sized cities, though.Hopanes are pentacyclic triterpanes usually containing 27 to 35 carbon atoms with four six-membered rings and one five-membered ring (Fig. 2(c)) (Prince and Walters, 2007).They are characteristic of petroleum because of their thermodynamic stability (Prince and Walters, 2007).Similarly, steranes are a class of tetra-cyclic compounds derived from steroids or sterols (Fig. 2(d)).Hopanes and steranes are important classes of compounds as they have been established as fossil fuel markers (Schauer et al., 1999;Lough et al., 2007).Hopane and sterane concentrations in Karachi are reported in Table 2. Compared to Nanjing, China, in wintertime, the concentrations of 17α(H)-22,29,30-trisnorhopane are 25 times less than the average concentration in Karachi, although Nanjing had similar PM 2.5 mass concentrations of approximately 120 µg m -3 (Wang et al., 2007).For 17α(H)-21β(H)-hopane the concentrations are also seen less in Houston during wintertime as compared to Karachi, by a factor of 100-400 times at two different sites in Houston where annual average PM 2.5 mass concentrations were in the range of 20 µg m -3 (Fraser et al., 2002).This indicates that for the most part, there is a greater abundance of hopanes in Karachi relative to studies completed in similarly sized megacities, suggesting a higher contribution of fossil fuel to PM 2.5 .Organosulfates and sulfonates have altered sampling dates as described in Methods.n = 14 for organosulfates and sulfonates, n = 22 for all other compounds.n-Alkanes are long chain hydrocarbons that are used as tracers for anthropogenic or biogenic sources (Fig. 2(e)).Even-numbered n-alkanes are tracers for fossil fuel, gaseous emissions, and tailpipe emissions (Wang et al., 2009).The carbon preference index (CPI) has been established as a definitive measurement indicating strong odd-carbon preference of greater than 6 for plant wax distributions, while a CPI closer to 1 indicates vehicular emissions (Cox et al., 1982;Mazurek et al., 1989).CPI is calculated by dividing the total mass of odd homologues over the total mass of even homologues (Mazurek et al., 1989).The CPI for Karachi n-alkanes C 21 -C 34 is 1.21, indicating a weak odd-carbon preference, suggesting there is a strong influence of alkanes emitted from fossil fuel combustion and evaporation in Karachi.Therefore, the n-alkanes found in Karachi are consistent with the findings of hopanes and steranes, showing a strong contribution of fossil fuel to PM 2.5 .
Levoglucosan is an organic tracer for biomass fuel burning (Simoneit et al., 1999).24-hour levoglucosan averages produced a range from 69.8-1571 ng m -3 and a 24-hour average of 535 ng m -3 (Fig. 2(f)).Compared to other cities, these values measure relatively high, with one study of three megacities in China reporting 24-hour levoglucosan concentrations ranging from 10.1-19.0ng m -3 , and another study of a megacity in China reporting PM 10 daytime levoglucosan concentrations of 52.3 ng m -3 , and an average of 251.3 ng m -3 during a four-day biomass burning episode (Jiang et al., 2009;Zhang et al., 2010).There is about four times as much levoglucosan reported in Lahore as compared to Karachi (Stone et al., 2010).In contrast, another study reported five times less levoglucosan concentration in a sampling period in March-April 2009 (Shahid et al., 2016).A significant negative correlation was found between the daily levoglucosan values and daily average temperatures (r = -0.589,p = 0.004), indicating that with warmer temperatures, there is less of a need to burn biomass fuel for heat in homes.This finding is also supported by a significant negative correlation that was observed between daily 24-hour average levoglucosan concentrations and daily minimum temperature value (r = -0.606,p = 0.003).Similarly, this negative correlation between levoglucosan and minimum temperature values indicates that with cooler temperatures, there would be a greater need to burn more fuel for heat; the nature of these fuels could be biomass or perhaps biomass mixed with waste, which is shown to emit levoglucosan (Jayarathne et al., 2018).No significant correlation was observed between levoglucosan levels and daily maximum temperature values.Retene, another PAH, is a tracer for softwood burning (Simoneit et al., 1999).A correlation analysis performed between retene and levoglucosan revealed a moderate correlation (r = 0.468, p = 0.028), indicating that there is some softwood burning contributing to the total biomass fuel burning.Most likely, there is a combination of various hardwood and softwood biomass burning contributing to the overall PM 2.5 , and it is likely that biomass is co-burned with other materials in some cases.

Day-of-the-Week Variation
Variation in primary sources was evaluated by comparing t-tests to examine significant differences between weekends (Sundays: January 8, 15, 22, and 29) and weekdays (all other sampling days excluding January 10-12, as these were Eid-ul-Azha holidays).The discussion will primarily focus on one analyte from each class of compounds observed in this study including 17α(H)-22,29,30-trisnorhopane, αββ-20(R)-C29-sitostane, triacontane, levoglucosan, benzo(a)pyrene and PM 2.5 mass.We would expect that human activity related to vehicular traffic and industrial activity would be greater on weekdays as compared to weekends.Generally, minimum PM 2.5 mass concentrations were observed on Sundays, averaging 123 µg m -3 compared to weekdays, averaging around 190 µg m -3 .The difference was found to be significant at the 95% confidence interval and is most likely attributed to anthropogenic activity related to higher traffic, more industrial oil burning and other pollution emitted by factories during work hours, thus leading to the conclusion that anthropogenic activity directly influences the levels of organic PM 2.5 in Karachi.
Similarly, 17α(H)-22,29,30-trisnorhopane and αββ-20(R)-C29-sitostane were also found to have a significant difference of weekday concentrations greater than weekend, with 17α(H)-22,29,30-trisnorhopane having an average concentration of 14.1 ng m -3 on weekdays versus 7.68 ng m -3 on weekends, and αββ-20(R)-C29-sitostane having 12.6 ng m -3 average concentration on weekdays versus 6.63 ng m -3 on weekends (Figs.2(c)-2(d)).This implies that anthropogenic activity significantly influences the PM 2.5 concentrations observed in Karachi, showing higher amounts of tracers for fossil fuels on weekdays versus weekends, as well as PM 2.5 mass in general.In contrast, benzo(a)pyrene, triacontane, and levoglucosan did not show a significant difference at the 95% confidence interval in weekday versus weekend concentrations in Karachi.

Quantitative Results
Of the quantified organosulfur compounds, 3-methylbenzyl sulfate exhibited the highest average concentration (0.917 ng m -3 ) with respect to OSs and benzene sulfonate exhibited the highest average concentration (0.304 ng m -3 ) with respect to sulfonates (Fig. 3, Table 2).Benzyl sulfate concentrations ranged from 0.023-0.310ng m -3 with an overall mean of 0.182 ng m -3 .Benzyl sulfate was shown to be higher in Lahore as compared to Karachi by a factor of three according to measurements collected in January 2007 reported in a study by Kundu et al. (2013).Additionally, samples collected in July were shown to have the lowest concentrations of benzyl sulfate (0.05 ng m -3 ), suggesting a seasonal influence on OS concentrations.
In a previous study by Riva et al. (2015), photooxidation of naphthalene resulted in the formation of aromatic sulfonates instead of organosulfates.While naphthalene was not measured, it is interesting to note that our results show this was not the case as sulfonates made a significantly smaller contribution to PM 2.5 than did organosulfates in Karachi.
The correlations between aromatic OSs and sulfonates, as well as aromatic OSs and sulfonates were analyzed in order to evaluate if they have similar sources.Within the OSs, phenyl sulfate has shown moderate positive correlations with 3-methylphenyl sulfate (r = 0.79, p = 0.001) and with benzene sulfonate (r = 0.69, p = 0.006), and there is a strong correlation between the benzene sulfonate and 4-ethyl benzene sulfonate (r = 0.75, p = 0.003).Benzyl sulfate exhibits a moderate positive correlation with 2-methylbenzyl sulfate (r = 0.56, p = 0.036) and strong correlations with 3methybenzyl sulfate (r = 0.84, p = 0.000) and 2-formyl benzene sulfonate (r = 0.73, p = 0.003).2-formyl benzene sulfonate exhibits moderate positive correlation with 2methybenzyl sulfate (r = 0.54, p = 0.045), and strong positive correlations with benzyl sulfate (r = 0.73, p = 0.003) and 3-methylbenzyl sulfate (r = 0.693, p = 0.006).The above moderate and strong correlations suggest that those organosulfur compounds originate from some common sources.Moreover, the fact that there are many weak intercorrelations between organosulfur species that are statistically significant at the 95% confidence interval suggests that there are multiple sources or formation mechanisms of aromatic OSs and sulfonates in the atmosphere in Karachi.

Qualitative Results
Qualitative analysis of a single sample collected from Karachi on January 15, 2006, using UPLC-ToF MS indicates the presence of an alkyl sulfate series starting from C 6 H 13 O 4 S -to C 9 H 19 O 4 S -(Table 4, Fig. 4).In a previous study by Riva et al. (2016) photooxidation of C 10 -C 12 alkanes in the presence of ammonium sulfate seed aerosol yielded an alkyl sulfate series that were also found in SOA detected in major cities.This, in combination with the study by Riva et al. (2015) mentioned earlier where photooxidation of naphthalene yields OSs and sulfonates, support the possibility of formation of alkyl series from the reaction with anthropogenic sources present in Karachi.Likely precursors for the formation of the alkyl sulfate series would be similar cyclic alkane-based compounds such as hopanes, suggesting the formation of the alkyl series from petroleum-based sources.A similar study showed that photooxidation of diesel and biodiesel fuels formed OSs and sulfonates (Blair et al., 2017).

CONCLUSIONS
In summary, the PM 2.5 concentration measured in Karachi, Pakistan, in 2006 was shown to regularly exceed WHO and EPD guidelines and varied according to the time/season as well as anthropogenic activity.The total contribution of organics observed in the PM 2.5 mass was 1%.Benzo(a)pyrene was shown to be present in lower concentrations in Karachi compared to other cities.In contrast, hopanes were present in higher concentrations than in other cities.This, in combination with the finding that there is a weak odd-carbon preference for n-alkanes, indicates a greater contribution from fossil fuel combustion to the overall PM mass in Karachi.The correlation between biomass tracers and the temperature, in addition to the lower levels of PM during weekends, suggests an anthropogenic influence as well.Based on this work, it can be concluded that specific aromatic sulfonates and aromatic OSs are potentially being emitted from a common source or being produced by similar formation mechanisms.Furthermore, while specific OSs are emitted from fossil fuel combustion, others are connected to biomass burning, whereas sulfonates are emitted solely from fossil fuel burning.

Fig. 3 .
Fig. 3. Concentration time series of organosulfates and sulfonates in Karachi, Pakistan from January 8 to 29, 2006.Data are shown only for days data was recorded.See Experimental and Statistical Methods.

Dibenz
for organosulfates and sulfonates, n = 22 for all other compounds.

Fig. 4 .
Fig. 4. Overlaid extracted ion chromatograms obtained for alkyl sulfate series (m/z ± 0.01 Da) in fine particulate matter collected from Karachi, Pakistan on January 15, 2006 using ultra-performance liquid chromatography and time-of-flight mass spectrometry.As indicated by '*' are the peaks corresponding to alkyl sulfates.

Table 1 .
The organosulfur compounds analyzed by ultra-performance liquid chromatography and tandem mass spectrometry including their characteristic product ions, optimized mass spectrometric conditions, retention time (t

Table 3 .
Pearson correlation coefficients (r) for primary tracers, organosulfates, and sulfonates (number of samples = 14) that are significant at or above 95% confidence interval.Bolded values are 'r' that are significant at 95% confidence interval.

Table 4 .
Molecular characterization of alkyl sulfates identified in fine particulate matter collected from Karachi, Pakistan on January 15, 2006 using ultra-performance liquid chromatography and time-of-flight mass spectrometry.