Mixing ratios of volatile organic compounds (VOCs) in the atmosphere of Karachi, Pakistan

Mixing ratios of carbon monoxide (CO), methane (CH 4 ), non-methane hydrocarbons, halocarbons and alkyl nitrates (a total of 72 species) were determined for 78 whole air samples collected during the winter of 1998–1999 in Karachi, Pakistan. This is the ﬁrst time that volatile organic compound (VOC) levels in Karachi have been extensively characterized. The overall air quality of the urban environment was determined using air samples collected at six locations throughout Karachi. Methane (6.3 ppmv) and ethane (93ppbv) levels in Karachi were found to be much higher than in other cities that have been studied. The very high CH 4 levels highlight the importance of natural gas leakage in Karachi. The leakage of liqueﬁed petroleum gas contributes to elevated propane and butane levels in Karachi, although the propane and butane burdens were lower than in other cities (e.g., Mexico City, Santiago). High levels of benzene (0.3–19ppbv) also appear to be of concern in the Karachi urban area. Vehicular emissions were characterized using air samples collected along the busiest thoroughfare of the city (M.A. Jinnah Road). Emissions from vehicular exhaust were found to be the main source of many of the hydrocarbons reported here. Signiﬁcant levels of isoprene (1.2 ppbv) were detected at the roadside, and vehicular exhaust is estimated to account for about 20% of the isoprene observed in Karachi. 1,2-Dichloroethane, a lead scavenger added to leaded fuel, was also emitted by cars. The photochemical production of ozone (O 3 ) was calculated for CO and the various VOCs using the Maximum Incremental Reactivity (MIR) scale. Based on the MIR scale, the leading contributors to O 3 production in Karachi are ethene, CO, propene, m -xylene and toluene. r 2002 Elsevier Science Ltd. All rights reserved.


Introduction
Urban air pollution is a problem in most megacities. Volatile organic compounds (VOCs) coupled with nitrogen oxides (NO x ) are the most important chemical precursors of tropospheric ozone (O 3 ) formation. Nonmethane hydrocarbons (NMHCs) are especially important among VOCs in urban areas because their reaction with the hydroxyl radical (OH) plays a critical role in atmospheric photochemical reactions (Derwent, 1995;Finlayson-Pitts and Pitts, 1999). The production of VOCs and NO x in urban environments is often dominated by vehicular emissions involving gasoline distribution, evaporation and automobile exhaust (Watson et al., 2001). Ethene and acetylene are characteristic products of internal combustion engines, while C 5 -C 8 NMHCs are generally associated with unburned vehicular emissions (Mayrsohn and Crabtree, 1976). Alkanes are emitted from gasoline evaporation, liquefied petroleum gas (LPG) leakage and natural gas leakage (e.g., Blake and Rowland, 1995;Chen et al., 2001).
Natural biological sources of hydrocarbons and leakage of fossil fuels other than gasoline can also be important contributors for some species (Roselle et al., 1991;Benjamin et al., 1997). In order to determine the relative importance of various hydrocarbons to O 3 production, detailed NMHC speciation in urban atmospheres is necessary, particularly in developing countries (Haszpra, 1991;Mohan et al., 1997;Sharma, 1997;Singh et al., 1997;Grosjean et al., 1998aGrosjean et al., , 1999Hao et al., 2000;Padhy and Varshney, 2000;Sharma et al., 2000).
In this study, detailed NMHC speciation was performed to determine the overall sources and types of emissions for Karachi, southeastern Pakistan. The city of Karachi has a rapidly expanding population of over 14 million and a large industrial base including oilfired power plants, cement factories, a steel mill, textile plants, two major refineries, heavy petrochemical industries and several light industries. A recent study reports one of the highest total suspended particulate (TSP) loadings recorded in any megacity of the world, with an average daily TSP concentration range of 627-938 mg m À3 (Parekh et al., 2001). Air pollution has reached alarming levels in major cities of Pakistan and the city dwellers are being continuously exposed to its harmful effects.
A major focus of this study is the contribution of vehicular emissions to air pollution in Karachi. Vehicular traffic can be classified into three categories: gasoline fueled (e.g., cars, jeeps), diesel fueled (e.g., buses, trucks), and vehicles with two-stroke engines (e.g., motor rickshaws). The number of vehicles in Pakistan (1 per 143 people in 1997) has swelled in recent years from 1.1 million in 1980 to 4.2 million in 1999 (Federal Bureau of Statistic, Government of Pakistan, 1990;Turner, 2001). Although the per capita vehicle usage is still well below that of the US (1 per 1.3 people), the average Pakistani vehicle emits 25 times as much carbon monoxide (CO) as the average US vehicle, and about 20 times more NMHCs (United States Energy Information Administration, 2000). A major problem is that most of the vehicles in Pakistan are old, illmaintained and they have poor or no pollution control devices. Furthermore, the lead concentrations in gasoline in Pakistan are in the range of 1.5-2.0 g/l (Murley, 1991), which is well above the WHO maximum allowable concentration of 0.15 g/l.
Thirty percent of Pakistan's automobile traffic is concentrated in Karachi. While Karachi is about 50 km across, the location of commercial and business areas that have the greatest impact on transportation has remained the same since the 1970s. The peak hour vehicle speed has fallen from 15 km/h in 1983 to 10 km/h today. Numerous traffic lights lead to frequent stop/ start traffic patterns which in turn result in poor and incomplete combustion. Gases and particulates originating in automobile engine emissions are among the most significant sources of air pollution in the congested and poorly ventilated streets of the city. Certain areas (e.g., M.A. Jinnah Road) are badly affected by automotive exhaust. The projected increase in the use of gasoline and diesel powered automobiles will further increase this problem.

Experimental
Karachi is situated on the Arabian Sea (251N; 671E; 22 m elevation), with a subtropical climate and little rainfall (mean annual rainfall 169 mm). Between 10 December 1998 and 11 January 1999, two sets of whole air samples were collected in evacuated 2 l electropolished stainless steel canisters throughout the city of Karachi. The sampling height was about 1.5 m above the ground. The preparation and pre-conditioning of the canisters prior to sampling is described in Blake et al. (1994).
In the first sampling campaign, 28 air samples were collected at the side M.A. Jinnah Road (Fig. 1 and 14-17 December were normal working days and 13 December was Sunday, a non-working day in Pakistan. Heavy traffic is expected on all working days on the M.A. Jinnah Road, in particular during the commuting hours of 07:30-10:30 and 15:00-20:00. During the sampling period the weather was average for that time of year (15-311C and cloudless). All of the samples were collected at the same location, close to the traffic. The purpose was to have a detailed time series to define which compounds are primarily emitted from Karachi's motor vehicles and to determine their daily and weekly variations.
In the second sampling campaign, Karachi was divided into grids in order to determine the overall CO and hydrocarbon mixing ratios in the Karachi urban area. Fifty samples were collected at six sampling sites located throughout the city: two industrial areas (Landhi industrial area and Korangi), three sites relatively well separated from local sources (Karachi University area, North Karachi and Mauripur) and the M.A. Jinnah Road (Fig. 1). The first sample from each site was collected at 00:00 on Saturday, 9 January 1999. Samples were then collected every 4 h beginning at 04:00 on Sunday, 10 January 1999 and ending at 08:00 on Monday, 11 January 1999.
During sampling the canister valve was slightly opened, allowing about 2 min for the collection of the ''integrated'' samples. The canisters were then shipped to our laboratory at the University of California, Irvine (UCI) and analyzed for CO, methane (CH 4 ), NMHCs, halocarbons and alkyl nitrates. Carbon monoxide analyses were carried out using a gas chromatograph (HP 5890) equipped with a flame ionization detector (FID) and a 3 m molecular sieve column. The samples were injected and then converted to CH 4 in a catalyst (2% coating of nickel on Chromosorb G) after the chromatographic separation. Methane was also analyzed using an HP 5890 GC equipped with an FID detector. The samples were injected into an 1/8 00 stainless steel 0.9 m column packed with 80/100 mesh Spherocarb.
The analytical system used to analyze NMHCs, halocarbons and alkyl nitrates involved the cryogenic pre-concentration of 217 cm 3 of air sample in a stainless steel loop (1/4 00 O.D.) filled with 3 mm glass beads and immersed in liquid nitrogen (À1961C). The pre-concentrated sample was subsequently vaporized with hot water and split into five different streams directed to three GCs (all HP 6890). The first GC was equipped with two different detector/column combinations: (1) a 60 m, 0.32 mm I.D., 1 mm film thickness, 1701 column (Restek) output to an ECD for the detection of halogenated hydrocarbons and C 1 -C 4 alkyl nitrates, and (2) a GS-Alumina PLOT, 30 m, 0.53 mm I.D. (J&W Scientific) output to an FID measuring C 2 -C 7 NMHCs. The second GC was equipped with a 60 m, 0.32 mm I.D., 1 mm film thickness, DB-1 column (J&W Scientific) output to an FID providing data on C 3 -C 10 hydrocarbons. The third GC was equipped with two different detector/column combinations: (1) a 60 m, 0.25 mm I.D., 0.5 mm film thickness, DB5-MS column (J&W Scientific) output to a quadrupole mass spectrometer detector (HP-5973) working in selected ion monitoring (SIM) mode, and (2) a 30 m, 0.25 mm I.D., 0.5 mm film thickness, DB-5 column (J&W Scientific) output to an ECD for the detection of halocarbons and C 1 -C 4 alkyl nitrates. Additional analytical details are given by Blake et al. (2001) and Colman et al. (2001).
The measurement precision, detection limits and accuracy vary by compound and are quantified for each species in Colman et al. (2001). Briefly, the detection limit is 5 ppbv for CO, 3 pptv for NMHCs, 0.02 pptv for alkyl nitrates, and 0.01-10 pptv for halogenated species (CH 4 is always above its detection limit). The accuracy of our measurements is 1% for CH 4 , 5-7% for CO, 5% for NMHCs, 10-20% for alkyl nitrates, and 2-20% for halogenated species. The measurement precision is 2 ppbv for CH 4 , 4 ppbv for CO, and ranges from 0.5-16% for NMHCs, 1-2% for alkyl nitrates, and 0.7-9% for halogenated species. The measurement precision improves with increasing mixing ratio, and the precision of measurements reported in this city study is expected to be better than those reported for less polluted air by Colman et al. (2001).

Composition of Karachi's urban air
The samples collected at the six urban sampling sites throughout the city were used to make a detailed characterization of the composition of Karachi's urban air (Table 1). Methane (77% of total measured compounds) and CO (19.5%) were dominant among the gases. The remainder comprised alkanes (2.6%), alkenes (0.4%), alkynes (0.2%), aromatic compounds (0.2%), halogenated hydrocarbons and alkyl nitrates (o0.1%). We note that oxygenated compounds (e.g., alcohols, aldehydes and ketones) were not measured during this study. As will be shown in Section 3.3, a compound's concentration coupled with its reactivity determines its contribution to O 3 production in the city.
Comparison with NMHC mixing ratios from other cities indicates that Karachi has some of the highest NMHC mixing ratios of anywhere in the world (Table 2). In particular, CH 4 (6.3 ppmv) and ethane (93 ppbv) levels in Karachi are substantially elevated compared to other cities that have been studied, and propane mixing ratios are also high (41 ppbv). For these three compounds, a decrease in their mixing ratios was observed during the daytime, leading to a midafternoon Table 1 Mean, minimum and maximum mixing ratios for the compounds identified in the Karachi urban area. Correlation coefficients ðR 2 Þ with CO and acetylene are also shown. SD-standard deviation A sample of natural gas used in Karachi in January 1999 confirms that CH 4 dominates the natural gas composition (99.9%), followed by ethane (0.06%) and propane (0.01%). Whereas natural gas is the main source of CH 4 in Karachi, the source of the elevated ethane throughout Karachi is less clear. Because natural gas in Karachi contains only 0.06% ethane, natural gas leakage can only account for a few of the 93 ppb of ethane observed in Karachi (Table 2). A second known source of light alkanes is the leakage of unburned LPG from household uses (Blake and Rowland, 1995;Chen et al., 2001). In Karachi, the hydrocarbon composition of LPG samples reveals n-butane as main constituent (43.4%), followed by propane (28.9%), i-butane (15.1%), CH 4 (7.8%) and ethane (2.9%). However, in all 50 samples collected at the six different urban locations, the ethane mixing ratio was always found to be higher than propane (ethane/propane ratio of 2.370.4), even though ethane is a minor component of LPG compared to propane. That is, natural gas and LPG leakage do not explain the high levels of ethane found in Karachi, suggesting either an additional unknown source, or that the natural gas source sample was not representative of ethane levels normally found in Karachi's natural gas. In the latter case, natural gas containing a maximum of 1.5% of ethane would be required to explain the observed levels of ethane in Karachi.
Although the CH 4 and ethane levels in Karachi are much higher than in other cities (Table 2), Karachi shows lower levels of propane (41 ppbv) and n-butane (20 ppbv) compared to Mexico City (158 and 70 ppbv, respectively; Blake and Rowland, 1995) and Santiago (137 and 27 ppbv, respectively; Chen et al., 2001). For example, whereas ethane is the most abundant NMHC in Karachi (Table 2), propane was found to be the most abundant NMHC in Santiago's air (Chen et al., 2001). These results highlight the lesser importance LPG leakage on NMHC levels in Karachi compared to cities such as Mexico City and Santiago. By contrast, the very high levels of CH 4 in Karachi show the strong influence of natural gas leakage in Karachi compared to other cities.
Benzene, ethylbenzene, n-hexane, toluene, xylenes (meta-, para-and ortho-) and 1,3-butadiene are in the Urban Air Toxics list of the Clean Air Act of the United States, as amended in 1990. The regulatory limit for benzene in several countries is 10 mg/m 3 (3.1 ppbv) and sometimes less. In Pakistan, air quality standards have not yet been set for organic pollutants. However, the maximum benzene mixing ratio measured in Karachi (19.3 ppbv; Table 1) exceeds the 3.1 ppbv regulatory limit in other countries, indicating that high benzene levels are a concern in Karachi.  Sharma et al. (2000). b Blake and Rowland (1995). c Chen et al. (2001). d Moschonas and Glavas (1996). e Ding and Wang (1998). f Tsujino and Kuwata (1993 Nelson et al. (1982). i Aronian et al. (1989). j Derwent et al. (2000).
Significant levels of isoprene were detected in Karachi's urban area (Table 1). The mean isoprene mixing ratio (0.8 ppbv) is higher than that reported in Kathmandu (0.3 ppbv; Sharma et al., 2000) and Mexico City (0.1 ppbv; Blake and Rowland, 1995) (Table 2). A correlation plot of isoprene versus CO is shown for the 50 samples collected throughout the urban area in Fig. 2a. The crosses correspond to samples collected at sites well removed from downtown and the main traffic artery. In these samples, no correlation exists between isoprene and CO, suggesting that biogenic isoprene emissions are significant. By contrast, samples collected at the other four sampling sites throughout Karachi correlate well with CO, suggesting an anthropogenic source (Section 3.2).
Most of the hydrocarbons identified in the Karachi urban area correlate well with CO and acetylene, two markers for vehicular fuel combustion (Table 1). These results suggest that vehicular exhaust is an important source for numerous hydrocarbons in Karachi, though we note that if two sources have nearly parallel time patterns of usage in a city, then correlations may not be useful for discriminating between sources (Blake and Rowland, 1995). The poorer correlations of isoprene, CH 4 , ethane, propane and i-butane with CO and acetylene in the urban area samples suggest other/ additional sources. The lack of correlation for the reported alkyl nitrates and the halogenated hydrocarbons, with the exception of 1,2-dichloroethane, indicates that these gases have primary sources other than vehicular emissions. Although the alkyl nitrate and halocarbon data are presented here for completeness, their sources are not discussed in this paper.

Composition of Karachi's roadside air
Karachi vehicular emissions were characterized using the air samples collected at the side of M.A. Jinnah Road, the busiest route in Karachi. Time series of toluene, benzene and i-pentane illustrate the daily and weekly mixing ratio trends (Fig. 3). In particular, traffic levels and pollutant concentrations were significantly lower on Sunday (13 December). The working day hydrocarbon mixing ratios show the strongest increases during the evening rush hour (15:00-20:00), despite the expanded boundary layer at this time of day. Smaller increases were observed during the morning rush hour (07:30-10:00), with the exception of 11 December in which the peak diurnal value occurred during the morning rather than afternoon commute. The day-today variations in the NMHC mixing ratio levels and patterns are expected to reflect changes in meteorology and vehicle traffic density.
Mixing The number of vehicles in Karachi equipped with catalytic converters is quite small. The ethene/acetylene ratio is well-known to be an indicator of the presence of a catalytic device: emissions from non-catalyst vehicles have an ethene/acetylene ratio close to 1 (Hoekman, 1992), while catalyst equipped cars have a ratio of 3 or greater (Duffy and Nelson, 1996). In addition, benzene/ toluene ratios of about 0.4 have been reported for noncatalyst cars, whereas a ratio of about 2 indicates cars equipped with catalytic converters (Hebb et al., 2000). Plots of ethene versus acetylene and of benzene versus toluene are shown for the 28 samples collected beside the M.A. Jinnah road in Figs. 5a-b. Ethene/acetylene and benzene/toluene ratios were 0.95 ðR 2 ¼ 0:88Þ and 0.43 ðR 2 ¼ 0:97Þ; respectively, reflecting the high number of vehicles without catalytic converters.
Karachi's gasoline contains high levels of lead (United States Energy Information Administration, 2000), and 1,2-dichloroethane (1,2-DCE, a known carcinogenic and mutagenic agent) is a leaded gasoline additive that is added on a mole-to-mole basis with lead (Clark et al., 1984). Among the halogenated hydrocarbons that were measured in the M.A. Jinnah Road samples, only methyl bromide (CH 3 Br) and 1,2-DCE correlated well with both CO and acetylene, suggesting that vehicular emissions contribute to elevated levels of these gases in Karachi (Table 3). Other studies have also found a small global source of CH 3 Br resulting from vehicular emissions (e.g., Thomas et al., 1997;Baker et al., 1998;Chen et al., 1999).
Isoprene mixing ratios in Karachi's urban area suggested the presence of both biogenic and anthropogenic sources ( Fig. 2a; see Section 3.1). For the traffic samples collected along the M.A. Jinnah road, a strong correlation was found between isoprene and both CO and acetylene (R 2 ¼ 0:91 and 0:78; respectively). Moreover, the diurnal variations of isoprene were similar to other hydrocarbons with vehicular sources, showing a higher mean daytime value (1.6 ppbv) with a peak during the late afternoon from 18:00 to 20:00, and a lower mixing ratio at night (0.7 ppbv) ( Table 3) Rasmussen, 1992;Guenther et al., 1993;Jobson et al., 1994) but the above observations help confirm that isoprene is also emitted during combustion (Jonsson et al., 1985;McLaren and Singleton, 1996;Starn et al., 1998). The percentage contribution of isoprene from vehicular exhaust can be estimated from the slope of isoprene versus CO for the samples collected at the M.A. Jinnah Road (0.1 ppbv/ppmv, Fig. 2b) and the mean CO value throughout Karachi (1.6 ppmv; Table 1). We estimate that about 20% of the observed 0.8 ppbv mean isoprene comes from vehicular exhaust.
i-Butane and CO were better correlated in the M.A. Jinnah roadside samples compared to the Karachi urban area (Figs. 6a-b). The same behavior was found for n-butane, which showed an excellent correlation with CO for the samples collected at the M.A. Jinnah Road ðR 2 ¼ 0:96Þ and a much weaker correlation for the samples collected throughout Karachi ðR 2 ¼ 0:67Þ: From this we conclude that an additional source of butanes must be considered. The average i-butane/nbutane ratio calculated for the samples collected at the M.A. Jinnah Road was 0.36, while the ratio for air  samples collected throughout Karachi was 0.53. From source samples, the ratio of i-butane/n-butane from natural gas leakage is 1. However, natural gas in Karachi contains o0.03% butanes (Section 3.1), and the leakage of butanes from natural gas can contribute only minimally to the observed butane levels in Karachi. Butanes are a major component of LPG in Karachi (Section 3.1), and LPG leakage is expected to be a source of butanes in Karachi. However, LPG leakage does not explain the higher ratio of i-butane/n-butane in Karachi compared to the M.A. Jinnah road samples, because the i-butane/n-butane ratio for LPG (0.35) is almost the same as that for vehicular emissions (0.36). Instead, because n-butane reacts more quickly with OH than does i-butane (Atkinson, 1997), the higher ratio of i-butane/ n-butane throughout Karachi could be in part the result of preferential removal n-butane during oxidation in more aged Karachi air, as compared to a lower ratio for fresh vehicular emissions along the M.A. Jinnah Road.

Photochemical reactivity
To estimate the amount of O 3 produced in Karachi by CO and each VOC, all reactions following the initial OH attack on CO or the VOC must be considered. In this study, the contributions of CO and each hydrocarbon to local O 3 production was approximated using the Maximum Incremental Reactivity (MIR) estimates of Carter (1994). This dimensionless coefficient (grams of O 3 produced per gram of VOC added to an initial VOC-NO x mixture), multiplied by the measured VOC concentration gives the expected contribution of each compound to local ozone formation (Grosjean et al., 1998b; Table 2). When applying the MIR scale, it is important to be aware that it does not perfectly fit the Karachi situation, and we use these calculations to assess the general significance of our Karachi observations.
Based on the MIR scale, the largest contributors to O 3 production in Karachi are ethene, CO, propene, m-xylene and toluene (Table 4). Ethene and propene are characteristic emissions from vehicular exhaust, and vehicular emissions are seen to be a leading source of O 3 production in Karachi. Methane, usually a very minor contributor to urban O 3 production, is sixth overall in Karachi. The wintertime mixing ratios of CH 4 in Karachi (6.3 ppmv) are more than triple current background levels (Dlugokencky et al., 2001;Simpson et al., 2002), and they are by far the highest levels we have measured in a wintertime city study (Blake and Rowland, 1995;Chen et al., 2001; Table 2). A reduction in natural gas emissions would benefit Karachi air quality and have a positive effect on both local and regional scales.

Conclusions
Two sampling campaigns were conducted in Karachi from December 1998 to January 1999. The first campaign characterized vehicular emissions next to the busiest thoroughfare in Karachi (the M.A. Jinnah Road), and the second campaign included air sampling throughout the Karachi urban area. This is the first time that VOC levels in the urban environment in Pakistan have been extensively characterized.
Many of the detected hydrocarbons were emitted from vehicular exhaust, as suggested by linear regression analysis carried out on the samples collected at the M.A. Jinnah Road and throughout the city. About 20% of the isoprene observed in Karachi was found to come from vehicular emissions. During working days, hydrocarbon mixing ratios typically showed the strongest increase during the afternoon commute (15:00-20:00) rather than in the morning. Much lower daytime mixing ratios were observed along the roadside on Sunday, a non-working day in Pakistan.
The measurements taken throughout the Karachi urban area allow us to compare differences in NMHC emissions and abundances among various cities. In Karachi, CH 4 and ethane levels were much higher than in other cities that have been studied, highlighting the strong influence of natural gas leakage on Karachi's hydrocarbon composition. LPG leakage was found to contribute to the propane and butane burden in Karachi, though propane and butane levels in Karachi were less than in other cities such as Mexico City and Santiago. Benzene levels in Karachi sometimes exceeded recommended levels in other countries, suggesting that high benzene levels are of concern for Karachi air quality.
The VOC data were used to estimate the amount of O 3 produced in Karachi by each VOC. Based on the Maximum Incremental Reactivity (MIR) scale, the leading contributors to O 3 production in Karachi are ethene, CO, propene, m-xylene and toluene. Ethene and propene are signature gases from vehicular exhaust, and vehicular emissions are a leading source of O 3 production in Karachi. In addition to photochemical reactivity and O 3 production calculations, the VOC data obtained in this study will serve as inputs to toxicity (human exposure) and air quality and fuel policy related studies of air pollution control in the urban centers of Pakistan. Improvement in fuel quality and implementation of emission controls are needed in order to significantly improve Karachi air quality.