Diurnal profiles of

(cid:1) Isoprene primarily originated from local biogenic emissions. (cid:1) Higher isoprene levels were observed on days when average daily temperature was above 30 (cid:3) C. (cid:1) Traf ﬁ c emissions and biogenic emissions were the major contributors to MACR/MVK. (cid:1) The use of MACR/MVK to represent the isoprene oxidation rate is inappropriate in urban Hong Kong.


Introduction
Biogenic volatile organic compounds (BVOCs) are emitted in substantial quantities from certain types of terrestrial vegetation, and are believed to play an important role in ozone (O 3 ) chemistry in forests (Makar et al., 1999), mountains (Dreyfus et al., 2002), semi-rural (Starn et al., 1998) and urban areas (Biesenthal and Shepson, 1997;Fuentes et al., 2000).Isoprene (2-methyl-1,3butadiene, C 5 H 8 ), with an annual global emission of about 500e 750 Tg, is the single highest VOC emission in the troposphere (Guenther et al., 2006).Based on measurements in 2006, biogenic sources are the biggest emission source category for VOCs in the Pearl River Delta (PRD) region of China, where air pollution has been severe due to the rapid growth of industries and population since the 1980s (Zheng et al., 2009).In particular, isoprene is the single highest VOC contributor to ozone formation potential, accounting for 15%.In Hong Kong, a highly urbanized and densely populated city, BVOCs have been shown to account for 8.8 mg m À3 (49%) of ambient PM 2.5 -bounded organic carbon content on days under regional transport influences, compared to 0.99 mg m À3 (21%) on days under mainly local emission influences (Hu et al., 2008).Given the significant contrition of BVOCs to both O 3 and SOA formation in urban areas and on a regional scale, the implications of photochemical oxidation of BVOCs need to be considered and understood for the development of effective air quality regulations.
The main removal pathway of isoprene is reaction with OH radicals during daytime and with O 3 and NO 3 radicals at night (Brown et al., 2009).In high NO x environments, formaldehyde, MVK and MACR are the major primary oxidation products of isoprene, accounting for more than 50% of the carbon yield (Carter and Atkinson, 1996;Miyoshi et al., 1994;Zhao et al., 2004).Many previous studies have used ratios of isoprene and its oxidation products, such as MVK/MACR and [MVK þ MACR]/isoprene, to investigate the magnitude and location of isoprene emission sources (Guo et al., 2012;Karl et al., 2007;Barket et al., 2004;Yokouchi, 1994;Stroud et al., 2001).In particular, MVK and MACR have been used to estimate isoprene's contribution to O 3 formation at a semi-rural site in British Columbia, Canada (Biesenthal et al., 1997) and at an urban forested site in Nashville, Tennessee, USA (Starn et al., 1998).The potential use of MVK/MACR as a direct measurement of the actual oxidation rate of isoprene allows the estimation of isoprene's contribution to O 3 production (Guo et al., 2012;Biesenthal and Shepson, 1997).However, MVK and MACR can also originate from primary emission sources including automobile exhaust (Yokouchi, 1994;Biesenthal and Shepson, 1997).Therefore, in urban areas with strong anthropogenic sources, the use of MVK/MACR to represent the isoprene oxidation rate may not be reliable due to the additional contributions of MVK/MACR from primary emissions.
Diurnal variations and species correlations at a particular site can contain information about the dominant local sources and chemical processes.Although many studies have reported timeseries profiles of BVOCs in the PRD region (Li and Wang, 2012;Tang et al., 2007), limited information is available on the diurnal profile of isoprene oxidation products, particularly in urban areas in close proximity to fresh primary emissions.In this paper, the diurnal profiles of isoprene, MVK and MACR are studied at an urban site in Hong Kong and used to identify their sources and formation mechanisms.The goal of the paper is to examine the sources of MVK and MACR, and determine whether MVK/MACR is a reliable tool to represent the oxidation rate of isoprene in an urban area with local primary emissions.

Sampling description
The Hong Kong Environmental Protection Department (HKEPD) air quality monitoring station at Tsuen Wan (TW) is used to characterize VOCs in an urban area (Fig. 1).Tsuen Wan District, located in the New Territories of Hong Kong, has an area of 60.7 km 2 with a population of around 300,000 in 2011; it is a mixed residential, commercial and light industrial district.The sampling site (22.373 N, 114.112E) is adjacent to major roadways and surrounded by residential and industrial blocks.Previous studies have shown that vehicular exhaust is the predominant emission source of polycyclic aromatic hydrocarbons (Sin et al., 2003) and total nonmethane hydrocarbons (NMHCs) (Guo et al., 2004b) at this site.
The sampling campaign lasted from September to November 2010, the season when O 3 levels are generally highest due to the long range transport of pollution-laden continental air masses, strong photochemical activity, and/or meteorological conditions that favor the accumulation of atmospheric pollutants (Chan et al., 1998;Leung and Zhang, 2001).In this study, hourly VOC samples were collected on selected O 3 episode days (Oct. 24,29e31;Nov. 1e 3,9 and 19) and non-episode days (Sep. 28,Oct. 2,8,14,18e19,27e 28;Nov. 20e21).During the sampling period, O 3 levels were predicted based on weather forecasts and meteorological data such as temperature, wind speed and vertical mixing conditions.An episode day was defined when the highest hourly O 3 level exceeded 100 ppbv on a regional level.VOCs samples were collected using conditioned 2-L electropolished stainless steel canisters.The canisters were developed and prepared by the Rowland/Blake group at the University of California, Irvine (UCI).More details about the design and specification of the canisters can be found in Simpson et al. (2010).A stainless steel flow controlling device was used to control the sampling flow at 0.033 lpm for the collection of 2-L of air in one hour.During O 3 episode days, hourly samples were collected from 9 a.m. to 4 p.m. (one sample every hour) with some additional samples collected at 7 a.m., 6 p.m. and 9 p.m. On non-O 3 episode days, hourly VOC samples were collected from 7 a.m. to 7 p.m. (one sample every two hours).A total of 161 VOC samples were collected at the sampling site.

Chemical analysis of VOCs
The canister samples were analyzed by the UCI group.73 VOCs were quantified using multicolumn gas chromatography (GC).The details of the analytical procedures were described in Colman et al. (2001).In brief, for each canister sample, a 1520 cm 3 aliquot was introduced into the system's manifold and split into five streams.Each stream was chromatographically separated on an individual column and sensed by a single detector.Two flame ionization detectors (FIDs), two electron capture detectors (ECDs) and a quadrupole mass spectrometer detector (MSD) were used to measure hydrocarbons, halocarbons and sulfur compounds respectively.VOCs were then identified by their retention times and mass spectra.The mixing ratios of target VOCs were quantified using multipoint external calibration curves, National Bureau of Standards, Scott Specialty Gases and UCI-made standards.For quality control and assurance, working standards were analyzed every four samples and absolute standards were analyzed twice a day.The limit of detection is 5 pptv for MVK and MACR, and 3 pptv for NMHCs including isoprene.

Continuous monitoring of O 3 , CO, SO 2 and NOeNO 2 eNO x
The sampling site is one of the 14 air quality monitoring stations maintained and operated by the HKEPD.O 3 , CO, SO 2 , NOeNO 2 eNO x and meteorological parameters were monitored continuously by the HKEPD, and the hourly data were obtained from their online database (http://epic.epd.gov.hk/ca/uid/airdata).O 3 was measured by UV absorption (Advanced Pollution Instrumentation (API), model 400), whereas CO and SO 2 were monitored by a gas filter correlation, non-dispersive infrared analyzer (API, model 300) and a pulsed UV fluorescence analyzer (API, model 100E), respectively.A commercial chemiluminescence analyzer (API, Model 200A) was used to quantify the levels of NO, NO 2 and NO x .More details on the sampling instruments, measurement protocols, and quality assurance and control procedures can be found in HKEPD (2012).

Results
Table 1 shows overall statistics of SO 2 , O 3 , CO and NO, NO 2 , isoprene, MVK and MACR at TW.The average mixing ratios of SO 2 , O 3 , CO and NO, NO 2 were calculated using the hourly data which correspond to the hour when VOC canister samples were simultaneously collected.In general, significant levels of SO 2 (overall average ¼ 6.21 AE 2.85 ppbv, hourly maximum ¼ 16.9 ppbv) and NO 2 (overall average ¼ 31.6 AE 9.4, hourly maximum ¼ 66.5 ppbv) were observed.These levels were in line with other urban cities such as Los Angeles and Pittsburgh in the United States (http://www.epa.gov/airtrends/), highlighting the impact of anthropogenic emissions at this site.The overall average isoprene mixing ratio was 252 AE 204 pptv, and the overall average MACR and MVK levels were 101 AE 85 pptv and 175 AE 131 pptv, respectively.The average O 3 level during episode days (39.2 AE 7.2 ppbv) was significantly higher than on non-episode days (24.3 AE 8.2 ppbv) (p < 0.001).Nonetheless, the maximum hourly O 3 level on any given sampling day at this site was lower than 100 ppbv e the level defining an O 3 episode according to the Ambient Air Quality Standard (Grade II) in China.As mentioned previously, the TW sampling site is located in an urban environment surrounded by major roadways as well as residential and industrial blocks.Due to the proximity to anthropogenic emission sources, the average NO mixing ratio was significant at this site (24.9AE 17 ppbv).Reaction with NO is known to be the primary removal in environments with strong primary emissions (Ghim and Chang, 2000;Tan et al., 2009).A strong negative correlation between daily NO and O 3 (R 2 ¼ 0.64, p < 0.001), together with troughs of O 3 levels during morning traffic hours (6e9 a.m.; Fig. 2), confirm the titration of O 3 by NO at this urban site.

Sources of isoprene
The top panel of Fig. 2 shows the diurnal profiles of isoprene on the 19 sampling days.A typical bell-shaped distribution, with peaks occurring between 11 a.m. and 3 p.m., was observed on most days with the exception of Oct. 24, Nov. 20 (peaks at 9 or 10 a.m.) and Nov. 19 (peak at 5 p.m.).On Oct. 24, solar radiation was strongest (daytime maximum ¼ 851 W m À2 ) of the study, facilitating the photo-oxidation of isoprene.The daytime average solar radiation was 413 W m À2 on Oct. 24, compared to an overall average of 297 AE 68 W m À2 on the rest of the sampling days.This probably led to lower levels of isoprene as the day progressed.The isoprene peak on Nov. 19 coincided with the peaks of limonene and a-pinene, suggesting a common biogenic origin.Higher levels of isoprene (p < 0.001) were observed in the beginning of the sampling campaign (Sep.28 to Oct. 24) compared to the latter part of the sampling period (Oct.27 to Nov. 21), when the temperature was higher (average daily temperatures of 30.6 AE 1.6 C).From Oct. 27 onwards, the temperature dropped to an average daily value of 23.8 AE 1.2 C. Indeed, the association between isoprene and temperature was fair (R 2 ¼ 0.54).The emission rate of isoprene from vegetation largely depends on the ambient temperature.For example, an in-vivo study demonstrated that isoprene emission increases with leaf temperature almost linearly from 15 C to around 40 C, with a maximum at 39 C (Rasulov et al., 2010).Therefore, the higher levels of isoprene observed in the beginning of the study are consistent with expectation.
Although isoprene is primarily emitted by vegetation, vehicular exhaust is also a source of isoprene, as demonstrated in several dynamometer, tunnel and city studies (Chiang et al., 2007;Borbon et al., 2001;Dai et al., 2010).Since the sampling site is located adjacent to roadways with a high traffic flow (annual average daily traffic rate ¼ 29,980 vehicles per day in 2010; The Annual Traffic Census (2011)), the assumption that isoprene originated purely from biogenic sources was examined using a linear regression analysis.Highest associations were found with dimethyl sulfide (DMS) (R 2 ¼ 0.24, p < 0.001), followed by the biogenic tracer limonene (R 2 ¼ 0.22, p < 0.001).Some correlation was found with the tailpipe exhaust tracers for n-decane (R 2 ¼ 0.15, p < 0.001).Together with the bell-shaped diurnal variation of isoprene, it is suggested that isoprene predominately arose from biogenic emissions at this sampling location during the study period.The specific source contributions to isoprene are quantified using positive matrix factorization (PMF) in Section 3.3.
Based on a consecutive reaction scheme of isoprene oxidation by OH radicals under NO x -rich environments, Stroud et al. (2001) developed an expression for the time rate of change in the MACR/ isoprene and MVK/isoprene ratios as a function of the rate coefficients, processing time and average OH concentrations.Since then, the ratios of MACR/isoprene and MVK/isoprene have been widely used to estimate the photochemical age of isoprene in an air mass (Apel et al., 2002;Guo et al., 2012;Roberts et al., 2006).Fig. 3 shows MACR/isoprene versus MVK/isoprene, based on daytime data from 9 a.m. to 4 p.m. (n ¼ 104), when isoprene photochemistry was strong.Theoretical ratios were calculated using an average daytime OH concentration of 8 Â 10 6 molecules cm À3 , as obtained in a rural area about 60 km northwest of Guangzhou in the PRD in July 2006 (Hofzumahaus et al., 2009).As seen in the Figure, the measured data fit the predicted line well, with most data points corresponding to isoprene photochemical ages between 12 and 42 min, averaging to around 25 min.With an average wind speed of 2.1 AE 0.9 m s À1 from 9 a.m. to 4 p.m., the average distance between the center of the isoprene emitting source and the sampling site was about 3.2 km.Three large country parks, namely Tai Mo Shan Country Park, Kam Shan Country Park and Shing Mun Country Park are located 2e5 km from the sampling site (Fig. 1).With wind originating from southeast or east half of the sampling time, these country parks could be the dominant emission source of isoprene measured at TW.Note that the isoprene transport time calculated using this method is very sensitive to the assumed OH levels, which could vary spatially and temporally.Nonetheless, it is reasonable to conclude that the isoprene measured at TW is not significantly impacted by regional air mass transport from the PRD region of China, where the nearest city (Shenzhen) is located >20 km away  from the sampling site, and an air mass originating from Shenzhen would take 1.9 h to arrive the sampling site even at a wind speed of 3 m s À1 (average wind speed þ one standard deviation).On the other hand, most of the measured data fall above the predicted line, consistent with some previous observations, suggesting that in addition to the photochemical oxidation of isoprene, anthropogenic sources could contribute to the daytime levels of MVK (Apel et al., 2002;Guo et al., 2012;Stroud et al., 2001).The correlation between MVK and MACR is R 2 ¼ 0.70, suggesting predominant common origins but also the presence of additional, uncorrelated sources.The specific sources of MACR and MVK during the study period are discussed and quantified below.

Sources of MACR and MVK
Fig. 4 shows the time-series plot of MACR and MVK.The diurnal profiles of MACR and MVK followed each other fairly well, with the exception of Oct. 29e31 and Nov 2. As stated in Section 3.1, the association between MACR and MVK was good (R 2 ¼ 0.70), suggesting their common origins.Most of the time, levels of MVK were higher than MACR; in only 12 of the 161 sampling hours were the levels of MACR higher than those of MVK.The peaks of MACR and MVK usually occurred between 11 a.m. and 3 p.m., except on Nov. 9 (peak at 9 a.m.) and Nov. 21 (peak at 7 a.m.).This is consistent with the isoprene peaks observed from 11 a.m. to 3 p.m., suggesting that the peak concentrations of MACR and MVK in the middle of the day were the results of photo-degradation of isoprene.On 8 out of 19 sampling days (Oct. 8,14,30,31,and Nov. 2,3,9,21), morning levels of MACR/MVK were comparable to/higher than the afternoon ones.Increased levels of MACR/MVK were also observed in the morning rush hours on 7 additional days (Oct. 2,27,28,29,and Nov. 1,19 and 20), although their levels were not necessarily as high as their respective afternoon peaks.Low levels of MACR/MVK were observed on 4 sampling days (Sep. 28 and Oct. 18,19 and 24) during the morning sampling hours.On three of these days (Sep. 28,Oct. 18 and Oct. 19) northeasterly to northwesterly wind prevailed in the early morning.As seen in Fig. 1, to the north of the site is Tai Mo Shan Country Park.The northerly wind likely brought air masses that were relatively free of primary emissions to the site.On the fourth day (Oct.24), westerly/northwesterly wind dominated from midnight until early afternoon.Also, of the 19 sampling days, the strongest solar radiation was observed on Oct. 24, reaching 111 and 577 W m À2 at 7 and 9 a.m.respectively.This strong solar radiation likely resulted in a higher mixing depth, which might account for the lower levels of anthropogenic MACR/MVK in the morning rush hour on Oct. 24.While increased levels of MACR and MVK in both morning and midday were not observed in previous studies in rural or forested sampling sites (Apel et al., 2002;Montzka et al., 1995;Jordan et al., 2009;Nouaime et al., 1998;Spaulding et al., 2003), such diurnal trends were observed in metropolitan Houston, Texas and the morning peak was attributed to tail-pipe emission sources in the morning rush hour (Park et al., 2011).Given the low levels of isoprene in the early morning on all sampling days in this study (Fig. 2), it is possible that there were additional morning emission sources of MACR and MVK.
The top panel of Fig. 4 shows the ratio of the sum of [MACR þ MVK] to [isoprene].In areas where the three species originate from biogenic emissions, this ratio is mainly driven by the dominant oxidant chemistry that destroys isoprene and produces and destroys MACR and MVK during daytime.The ratio also depends on the degree of atmospheric mixing, distance from isoprene emitters, and the concentration of NO x (Biesenthal et al., 1998).In this study, the average [MACR þ MVK]/[isoprene] ratio was 1.69 AE 1.58, which is higher than those reported in other studies (0.23e1.0) (Montzka et al., 1995;Yokouchi, 1994;Apel et al., 2002;Biesenthal et al., 1998).Note that higher ratios are expected when the sampled air mass has aged under high NO x conditions (Biesenthal et al., 1998), which could be our case because NO x levels were relatively high (Section 3) with an average photochemical age time of w25 min (Fig. 3).Even higher ratios (>6) were observed in the sampling hours of 7 a.m., 5 p.m. or 9 p.m., which further suggests the direct emissions of MACR and MVK at this site.In contrast to the fair association between isoprene and temperature (R 2 ¼ 0.54), there was no clear relationship for MACR and MVK with temperature (R 2 ¼ 0.027 and 0.009, respectively).Fig. 5 shows the linear regression between the mixing ratios MACR/MVK and isoprene.The very low correlation coefficients (R 2 ¼ 0.066 for MACR and R 2 ¼ 0.048 for MVK) suggest that local isoprene was not responsible for all of the measured MACR and MVK; that MACR and MVK could be formed at a distance and transported to the site; and/ or that local primary MACR and MVK emissions might be present.The fact that the associations were not improved after lagging the isoprene concentrations relative to MACR and MVK (R ¼ 0.036 and 0.020 respectively), coupled with the generally low wind speeds in the early morning (average ¼ 1.21 AE 0.38 m s À1 from 5 to 8 a.m.), suggest that the transport of the isoprene oxidation products might not explain the MACR and MVK levels in the early morning.
A linear regression analysis was performed to better understand the sources of MACR and MVK.For MVK, its highest association was found with acetaldehyde (R 2 ¼ 0.39, p < 0.001) followed by HCFC-141b (R 2 ¼ 0.38, p < 0.001) (Fig. 6a, b).MACR experienced some correlation with acetaldehyde (R 2 ¼ 0.49) and 1-butene (R 2 ¼ 0.42) (Fig. 6c, d).HCFC-141b is used in metal degreasing and dry cleaning agents in Hong Kong (Guo et al., 2009).Previous studies at TW and in other urban parts of Hong Kong have associated acetaldehyde and 1butene with vehicular emissions (Ho et al., 2002b(Ho et al., , 2002a;;Guo et al., 2004a).In a tunnel study conducted in California, MVK was identified in the emission of light-duty gasoline vehicles (emission factor ¼ 0.67 AE 0.13 mg kg À1 in 2001 and 0.26 AE 0.02 mg kg À1 in 2006), as well as mid-duty and heavy-duty diesel trucks (emission factor ¼ 3.0 AE 1.2 mg kg À1 in 2006) (Ban-Weiss et al., 2008).MVK is also listed as a constituent of gasoline exhaust in the USEPA Master List of Mobile Source Air Toxics (http://www.epa.gov/OMSWWW/toxics.htm).Similarly, MACR was identified in a study of gasoline exhaust emissions using different fuels, with emission rates of 0.001 mg mile À1 , 0.03 mg mile À1 and 0.001 mg mile À1 for gasoline, 10% ethanol blend and 15% ethanol blend, respectively (Fanick, 2011).City studies have also associated MACR and MVK with vehicular emissions in Toronto and Houston (Biesenthal and Shepson, 1997;Park et al., 2011).Thus, the moderate correlations between MACR and MVK with acetaldehyde, HCFC-141b and 1butene suggest anthropogenic contributions to the measured MACR and MVK levels.To examine the impact of vehicular exhaust, a restricted analysis was conducted using data from morning rush hours (7 a.m. and 9 a.m., n ¼ 34).Indeed, improved associations were observed using this data subset (R 2 > 0.77 for all regressions, Fig. 6aed), further suggesting a significant contribution of traffic emissions to the observed MACR and MVK levels.

Source apportionment
To quantitatively apportion the sources of isoprene, MACR and MVK, the U.S. Environmental Protection Agency (EPA) PMF model was applied to the dataset.PMF is a multivariate factor analysis tool based on the explicit least-squares technique, and it has been widely used to apportion the source contributions of ambient VOCs in various environments (Anderson et al., 2001;Pindado and Perez, 2011;Kim et al., 2005).More details about PMF modeling can be found in the literature (Paatero, 1997).Based on the calculated statistical parameters and prior knowledge about the known emission source profiles, a five-factor model best reproduced the observed VOC concentrations (Table 2).The first factor has high loadings of i-and n-pentane, as well as n-hexane and n-heptane, and is attributed to fuel evaporation.For both MACR and MVK, 8e 13% was associated with this factor.Factor 2, with a dominance of propene, i/n-butane, as well as n-nonane and n-decane, was likely derived from liquefied petroleum gas (LPG) and diesel emissions.Around 2.5% of MVK lay in this factor.The factor with high loadings of benzene, toluene, ethylbenzene, and xylenes (BTEX) (Factor 3) was thought to be associated with gasoline emissions, and MACR and MVK respectively contributed 55 and 69% to this factor.Previous studies have identified both MACR and MVK in the composition of fuel and lubricating oil, and in tunnel studies (Ban-Weiss et al., 2008;Magnusson et al., 2002).The significant contribution of MACR and MVK from vehicular emissions observed in this study is consistent with the observations in other urban environment (Park et al., 2011).Note that approximately one-fourth of isoprene was associated with this factor, suggesting a non-trivial influence of vehicular emissions on isoprene levels at TW.The factor with dominant isoprene presence (69%; Factor 4) was identified as biogenic emissions.The associated MACR and MVK contributions were 29 and 20% respectively, suggesting a significant fraction of the measured MACR and MVK arose from isoprene oxidation.Note that reactive species including isoprene, ethylene, and 1,3butadiene have short lifetimes, typically a few hours.Yet, due to their usefulness, and sometimes exclusiveness in source identification, they have been included in many previous source apportionment studies (Brown et al., 2007;Kim et al., 2005;Xie and Berkowitz, 2006).Since reactive compounds are degraded on the way from the source to the receptor, the concentrations used in the PMF analysis can be considered the lower limit of the concentrations that would be measured at the sources.Therefore, although the sources of isoprene were likely emissions from nearby (w3.2 km at the wind speed of 2.1 m s À1 ; Section 3.1) country parks, the results of the PMF analysis should be interpreted with caution considering the possibility of underestimation due to the depletion from source to receptor.Factor 5, which contained most of the measured ethylbenzene and xylenes, was identified as solvent use emissions.4% of MACR was associated with this factor.Thus, gasoline exhaust and biogenic emission were the two most important contributors to the observed MACR and MVK at the sampling site, with minor contributions from fuel evaporations.

Conclusions
In summary, the diurnal profiles of isoprene and its oxidation products (MACR and MVK) were examined at an urban site with close proximity to vehicular and industrial emissions.Isoprene, which primarily originated from local biogenic emissions at this site, showed a typical bell-shaped diurnal distribution with higher levels on days when average temperature was above 30 C. Levels of MACR and MVK, on the other hand, were not correlated with temperature or isoprene concentrations.Although MACR and MVK usually peaked from noontime to early afternoon, high levels were often observed in the morning or afternoon rush hours.The low correlations between isoprene and MACR and MVK, together with high (MVK þ MACR)/isoprene ratios during rush hour, suggested either that MACR and MVK were not formed from local isoprene oxidation, and/or that vehicular emissions potentially contributed.Furthermore, high associations were found between MACR and MVK and tracers of anthropogenic emissions (acetaldehyde, 1butene, HCFC-141b, etc), particularly when the analysis was restricted to morning rush hour data.Positive matrix factorization analysis showed that traffic-related emissions and biogenic emissions were the two dominant contributors to MACR and MVK.As a result, we conclude that the use of MACR and MVK to represent the oxidation rate of isoprene is inappropriate in areas impacted by strong local anthropogenic emissions.

Fig. 1 .
Fig. 1.Map of the sampling site (TW) and its surrounding environments.The highly developed areas are in gray, and Kowloon is located 7e15 km southeast of TW.

Fig. 2 .
Fig. 2. Diurnal profiles of isoprene (top), ozone and temperature (bottom) at the TW site in autumn, 2010.Ozone episode days are marked with an asterisk.

Fig. 3 .
Fig. 3. Measured ratios of MVK/isoprene versus MACR/isoprene (black circles) together with calculated ratios (white triangles) based on a consecutive reaction scheme model (see text).

Table 1
Overall statistics of SO 2 , O 3 , CO and NO, NO 2 , isoprene, MVK & MACR at the Tsuen Wan (TW) sampling site for data collected between September and November, 2010.Ozone episode days (see Section 2.1) are highlighted in bold.

Table 2
Positive Matrix Factorization (PMF) extracted source profiles for air samples collected at Tsuen Wan (TW) from September to November, 2010 (n ¼ 161) (unit: % of species).