Atmos Chem Phys 14 3047ndash3062 2014wwwatmos-chem-physnet1430472014doi105194acp-14-3047-2014copy Author(s) 2014 CC Attribution 30 License

Atmospheric Chemistry

and PhysicsO

pen Access

Understanding primary and secondary sources of ambient carbonylcompounds in Beijing using the PMF model

W T Chen1 M Shao1 S H Lu1 M Wang1 L M Zeng1 B Yuan1 and Y Liu2

1State Joint Key Laboratory of Environmental Simulation and Pollution Control College of Environmental Sciences andEngineering Peking University Beijing China2Chinese Research Academy of Environmental Sciences Beijing China now at Earth System Research Laboratory Chemical Sciences Division NOAA 325 Broadway Boulder Colorado 80305USA

Correspondence toM Shao (mshaopkueducn)

Received 19 April 2013 ndash Published in Atmos Chem Phys Discuss 13 June 2013Revised 11 February 2014 ndash Accepted 15 February 2014 ndash Published 26 March 2014

Abstract Carbonyl compounds are important intermediatesin atmospheric photochemistry To explore the relative con-tributions of primary and secondary carbonyl sources car-bonyls and other volatile organic compounds (VOCs) weremeasured at an urban site in both winter and summer inBeijing The positive matrix factorization (PMF) model wasused for source apportionment of VOCs As VOCs undergophotochemical processes in the atmosphere and such pro-cesses may interfere with factor identification the relation-ships between the contributions of the resolved PMF factorsto each non-methane hydrocarbon (NMHC) species and itskOH value were used to distinguish fresh factors and photo-chemically aged factors As the result of PMF five factorswere resolved in winter and two of them were identifiedas photochemically aged emissions In summer four factorswere resolved including one aged factor Carbonyls abun-dances from aged factors were simulated by VOCs consump-tion and the corresponding carbonyl production yields andthe simulated abundances agreed well with the results ob-tained by the PMF model The source apportionment resultsindicated that secondary formation was the major sourceof carbonyls in both winter and summer with the respec-tive contributions of 512 and 460 For the three ma-jor carbonyl species primary anthropogenic sources con-tributed 289 and 323 to ambient formaldehyde 537 and 416 to acetaldehyde 681 and 562 to acetone inwinter and summer respectively

1 Introduction

Carbonyls including aldehydes and ketones are a group ofimportant oxygenated volatile organic compounds (OVOCs)As the key intermediates in the photo-oxidation of hydrocar-bons carbonyls are major sources of free radicals and areprecursors of ozone peroxyacyl nitrates and secondary or-ganic aerosol (Singh et al 1995 Finlayson-Pitts and Pitts1997 Liggio et al 2005) Not only being produced fromphoto-oxidation of hydrocarbons carbonyls also can be di-rectly emitted into the atmosphere from a variety of nat-ural and anthropogenic sources The major anthropogenicemission sources of carbonyls include incomplete combus-tion of fuels industrial processes and solvent usage (Zhangand Smith 1999 Ban-Weiss et al 2008 Kim et al 2008)Carbonyls can be lost from the atmosphere through reactionswith the hydroxyl radical (OH) photolysis and deposition(Lary and Shallcross 2000) Due to the complexity of car-bonyl sources and sinks it is a challenge to quantifying therelative contributions of primary emissions and secondaryformation

Previous studies have applied several techniques for thesource apportionment of carbonyls The multi-linear regres-sion method is a commonly used technique which separatesprimary and secondary carbonyls based on their correlationswith markers for primary emission (eg CO ethyne) andsecondary production (eg O3 glyoxal) (Friedfeld et al2002 Garcia et al 2006) However this approach heavilydepends on the selection of markers and improper mark-ers may bias source apportionment results For primary

Published by Copernicus Publications on behalf of the European Geosciences Union

3048 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

emissions some markers might be appropriate for certainemission sources (Rappengluumlck et al 2010) and thus the useof a single marker for all primary emissions could be veryrisky (Friedfeld et al 2002) In addition reactive VOCs thatare co-emitted with a primary emission marker can form sec-ondary carbonyls during air mass transport Hence the cor-relation will also exist between secondary carbonyls and theprimary marker For secondary formation the correlation co-efficient and the slope between carbonyls and the secondarymarker vary significantly depending on the VOCs NOx ratioand the degree of photochemical processing Furthermoreprimary carbonyls also can be precursors of these secondarymarkers which may also lead to a correlation between them(Parrish et al 2012)

To improve the multi-linear regression method de Gouwet al (2005) developed a parameterization method based onthe degree of photochemical processing (ie photochemicalage) with the consideration of the production of carbonyls bytheir precursors and the chemical removal of carbonyls dur-ing transport This method made several assumptions whichshould be carefully considered (1) anthropogenic emissionsof carbonyls and their precursors are proportional to a pri-mary marker (2) The removal of VOCs is governed by thereactions with OH radicals (3) Biogenic sources of car-bonyls are proportional to the emission of isoprene (4) Thephotochemical ages for sampled air masses can be deter-mined (de Gouw et al 2005) Positive matrix factorization(PMF) a receptor model has also been used to separatesources of carbonyls (Bon et al 2011 Buzcu Guven andOlaguer 2011 Yuan et al 2012 Zheng et al 2013) Com-pared with the multi-linear regression method which em-ploys one species for a certain source the PMF model canuse a number of species for source identification Most pre-vious studies identified several primary emission sources anda mixed secondary source among the PMF-resolved factorsHowever Yuan et al (2012) found that the PMF-resolvedfactors could be attributed to a common source at differentstages of photochemical processing rather than several inde-pendent sources In addition to the various source apportion-ment approaches based on carbonyl measurements Parrish etal (2012) compared the primary emission flux of formalde-hyde with fluxes of its precursors to quantify the relative con-tributions of primary and secondary formaldehyde sourcesHowever this approach relies on the accuracy of emissioninventories and reaction mechanisms

Large differences have been found among the above-mentioned methods used to calculate sources of carbonylsFor example several studies were conducted in the Hous-ton area TX USA to determine the contributions of pri-mary anthropogenic sources of formaldehyde Based on themulti-linear regression using CO as a primary marker Fried-feld et al (2002) reported that primary emissions contributed36 of atmospheric formaldehyde In a later study Rap-pengluumlck et al (2010) calculated the contributions of 385 and 89 for formaldehyde from primary vehicular emis-

sions and industrial emissions using CO and SO2 as primarymarkers respectively By using the PMF Buzcu Guven andOlaguer (2011) estimated that mobile sources and industrialsources contributed 23 and 17 of formaldehyde respec-tively In contrast by comparing emissions of formaldehydeand its precursors Parrish et al (2012) found that only 1 and 4 of formaldehyde were respectively emitted from ve-hicle exhaust and industrial sources and that secondary for-mation accounted for the overwhelming majority of atmo-spheric formaldehyde The source apportionment studies forcarbonyls were also conducted in Beijing Li et al (2010) at-tributed 76 of atmospheric formaldehyde to direct anthro-pogenic emissions using the multi-linear regression methodwhile Yuan et al (2012) calculated that this contributionwas only 22 using the parameterization method Comparedwith the number of formaldehyde studies there were fewersource apportionment studies for other carbonyls and mostof these studies were conducted at different sites (Goldsteinand Schade 2000 de Gouw et al 2005 Liu et al 2009Yuan et al 2012 Guo et al 2013) The only two studies thatcould be used for inter-comparison were taken in the sum-mer in Beijing and the apportioned primary contributions ofacetaldehyde were 16 and 46 in 2005 and 2010 respec-tively (Liu et al 2009 Yuan et al 2012) It was unclearthat such difference was caused by the variation of carbonylsources or the uncertainty of source apportionment

Though the State Council of China released mandatory re-quirement to improve air quality all over China the way toachieve the goal could be very tough (Liu et al 2013) Bei-jing is the capital of China and suffers from serious air pollu-tion problems characterized by high levels of ground-levelozone and PM25 (Shao et al 2006 Wang et al 2012a)Ambient carbonyls were detected at high levels in Beijing(Liu et al 2009 Zhang et al 2012) and formaldehyde dis-played a significant positive trend from 1997 to 2010 (DeSmedt et al 2010) However ambient concentrations of non-methane hydrocarbon (NMHC) were found to decrease from2003 (Wang et al 2012b) It is essential to quantify contribu-tions of different sources to carbonyls for understanding suchtrends and improving air quality in Beijing Yuan et al (2012)used the PMF model for carbonyl source apportionment in2010 summer and found that factors in PMF were actuallythe results of different degrees of photochemical process-ing rather than individual sources In this study online mea-surements of carbonyls and other VOCs were conducted atPeking University (PKU) site in the winter and summer Weused PMF to apportion carbonyl sources in these two sea-sons with different degrees of photochemical processing Thedegradation of NMHCs and production of carbonyls in theaged factors were analyzed based on the processes of photo-chemical aging to verify the PMF results Differences in car-bonyl sources between winter and summer were discussedand compared with previous studies

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3049

2 Methods

21 VOCs measurements

Online VOCs measurements were conducted at an urban sitein Beijing from 4 August 2011 to 9 September 2011 and from29 December 2011 to 18 January 2012 The sampling sitewas on the top of a six-floor building (sim 20 m above ground)on Peking University campus The PKU campus is locatedin the northwest of Beijing about 500 m north of the FourthRing Road This site was described in detail elsewhere (Songet al 2008 Liu et al 2009)

A custom-built online gas chromatography systemequipped with a mass spectrometry and a flame ionizationdetector (GC-MSFID) was used to measure C2ndashC10 hydro-carbons C3ndashC6 carbonyls C1ndashC4 alkyl nitrates halocar-bons and methyl tert-butyl ether (MTBE) with a time reso-lution of 1 h CFC-113 (112-trichloro-122-trifluoroethane)was used as a natural internal standard from ambient air dueto its long lifetime and minimal manmade emissions (W-T Liu et al 2012) Daily calibrations were performed everyday and the variations of target species responses were re-quired to be withinplusmn20 from the calibration curve Thedetection limits for each species varied from 1 to 20 ppt andthe precisions ranged from 1 to 6 The detailed informa-tion of this system can be found in Yuan et al (2012)

A commercial high-sensitivity proton-transfer-reactionmass spectrometry (PTR-MS) (Ionicon Analytik InnsbruckAustria) was used to measure 28 masses including C1ndashC4carbonyls and C6ndashC9 aromatics with a time resolution ofabout 30 s Background signals were measured for 15 min ev-ery 25 h by diverting the ambient air sample flow througha Pt-coated quartz wool converter at 370C (Yuan et al2013) Calibrations were done every two or three days andthe response factors varied within 20 The detection lim-its of each species varied from 40 to 200 ppt except forformaldehyde As the proton affinity of formaldehyde is onlyslightly higher than water the back reaction of proton trans-fer becomes significant The kinetic of PTR-MS formalde-hyde detection was found to be mainly influenced by humid-ity (Vlasenko et al 2010 Warneke et al 2011) Thereforewe calibrated our formaldehyde measurement by using gasstandards provided by a permeation tube (Kin-Tek USA) atthe humidity from 0ndash30 mmol molminus1 to obtain the curve offormaldehyde response factor on humidity During the entireambient measurement formaldehyde signals were correctedaccording to the response curve Due to the different ambi-ent humidity between winter and summer the detection lim-its of formaldehyde were 022ndash034 ppb in winter and 045ndash080 ppb in summer at a time resolution of 30 s Among allour data only less than 02 of formaldehyde concentra-tions in summer fell below the detection limit

During these two campaigns C3ndashC4 carbonyls C6ndashC9aromatics isoprene and MVK+MACR (methyl vinyl ke-tone+ methacrolein) were measured simultaneously by on-

line GC-MS and PTR-MS Good agreements were found be-tween these two systems for most species with correlationcoefficients larger than 090 and slopes ranging from 08 to12 Formaldehyde and acetaldehyde were only measured byPTR-MS However inter-comparison between the PTR-MSand 24-dinitrophenyl hydrazinendashhigh-pressure liquid chro-matography (DNPHndashHPLC) methods (EPA TO-11A) wasperformed before the field measurements By correcting theinfluence of humidity on formaldehyde response a correla-tion coefficient of 093 and a slope of 106 were obtained be-tween measurements by PTR-MS and DNPHndashHPLC meth-ods The detailed results of inter-comparison were shown inthe Supplement

22 Positive matrix factorization (PMF)

The US Environmental Protection Agencyrsquos PMF 30 recep-tor model was used for VOCs source apportionment in thisstudy The PMF is a multivariate factor analysis tool whichassumes that measured concentrations at receptor sites arelinear combinations of contributions from different sources(Paatero and Tapper 1994) Generally an ambient data setcan be viewed as a data matrixx with i byj dimensions inwhich i samples andj chemical species are measured Thegoal of multivariate receptor modeling is to identify the num-berp of factors the species profilef of each source the massg contributed by each factor to each individual sample andthe residuale (Eq 1)

xij =

sump

k=1gikfkj + eij (1)

The PMF solution minimizes the object functionQ basedupon the uncertainties (Eq 2)

Q =

sumn

i=1

summ

j=1

[xij minus

sump

k=1gikfkjuij

]2

(2)

whereu represents the uncertainty of each data The theo-retical Q (Qtheoretical) can be calculated as Eq (3) and thebest PMF solution should haveQQtheoreticalwith the valueof sim 1

Qtheoretical= i times j minus p times (i + j) (3)

The basic assumption of the PMF model is that the fittingspecies are not allowed for chemical losses or productionThus apportioning carbonyl sources using the PMF modelshould be approached with care As mentioned in the in-troduction several studies apportioned carbonyl sources byusing the PMF model (Bon et al 2011 Buzcu Guven andOlaguer 2011 Zheng et al 2013) However the study toinvestigate the validity of the results was not available yetYuan et al (2012) proved the capacity of the PMF approachin identifying the role of chemical aging for better under-standing the PMF factors The VOC emission ratios derivedfrom the PMF fresh factors agreed well with the ones calcu-lated based on photochemical ages indicating that the PMF

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3050 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Table 1The sum of formaldehyde and acetaldehyde concentrationsin Beijing (unit ppb)

Year Summer Winter

2005 217a 70a

2006 1312b

2008 133cd 66e

2009 114e 59e

2010 126e

2011 80f 81f

a(Pang and Mu 2006)b(Shao et al 2009)c(Xu et al 2010)ddata before traffic restriction in 2008e(Zhang et al 2012)f this study

approach could identify the contributions from primary emis-sions reasonably Additionally the abundances of NMHCsin the PMF-aged factors could be reproduced by the photo-chemical aging of fresh factors In this study we run PMFmodels using two VOC data sets from different seasons anddiscussed the apportionment results to investigate whetherthe PMF approach can separate carbonyl sources well

For the PMF analysis uncertainties of online GC-MSFIDand PTR-MS measurements were calculated using themethod by Yuan et al (2012) Generally for the species mea-sured by the GC-MSFID system the uncertainties were cal-culated as the sum of 10 of measured concentrations andone-third of detection limits For the species measured bythe PTR-MS the uncertainties were calculated as the sum of10 of concentrations and the errors from mass spectrome-ter measurements Finally only species with high signal-to-noise ratios and few missing data values were used for thePMF analysis In winter 10 carbonyls 26 NMHCs 2-butylnitrate (2-BuONO2) tetrachloroethylene and MTBE wereused In summer most of the values of cistrans-2-butene andcistrans-2-pentene fell below the detection limits so these 4species were not used for the PMF analysis

In this work the PMF factor numbers were explored from3 to 8 for the best solution The final numbers of factors de-pended on two aspects (1)QQtheoretical which was usedto characterize the quality of the reconstruction and (2) thephysical plausibility of the factors (Bon et al 2011) Four-and five-factor resolutions were selected in summer and win-ter respectively The free rotation of the PMF factors wasinvestigated by adjusting the Fpeak parameter betweenminus5and 5 The results with non-zero Fpeak values were gen-erally consistent with the runs with zero Fpeak values andthus the results used in this study were from the runs withnon-rotation

As the PMF-resolved factors might be influenced by pho-tochemical processes we used the relationship between thecontribution of one factor to each NMHC species and itschemical reactivity (kOH) to distinguish fresh and aged fac-

tors as performed by Yuan et al (2012) Generally if allPMF-resolved factors were originated from fresh emissionsthe distribution of each species would not correlate with itschemical reactivity As the air mass from a source was age-ing the NMHCs underwent photochemical reactions and themore reactive species would be more largely consumed As aresult a negative correlation could be seen between an agedfactor contribution to each NMHC species and itskOH valueConsidering that such complex relationships were difficult toexpress in mathematical formulae Spearmanrsquos coefficient ofrank correlation (rs) was used to represent the relationship

3 Results and discussion

31 Characteristics of ambient carbonyls in Beijing

The average concentration of the 10 measured carbonyls inBeijing was 132plusmn 79 ppb (average valueplusmn standard devi-ation) in winter In summer the average concentration wentup to 163plusmn 74 ppb Formaldehyde acetone and acetalde-hyde were the three major species contributing more than80 of the total measured carbonyls in both seasons Am-bient concentrations of carbonyls in Beijing were measuredby several previous studies from 2005 (Pang and Mu 2006Shao et al 2009 Xu et al 2010 Zhang et al 2012) Dur-ing the recent years VOC sources changed remarkably inBeijing resulting from quick increment of vehicular pop-ulation stricter standards for vehicle emissions (Zhang etal 2012) and solvent compositions (Yuan et al 2010) anddramatic changes in industries Table 1 shows the sum ofmeasured formaldehyde and acetaldehyde concentrations inBeijing during 2005ndash2011 The measured concentrations insummer showed a significant decrement from 2005 to 2006and from 2010 to 2011 while the wintertime concentra-tions did not exhibit a significant change Some of the abovestudies provided daily average concentrations of formalde-hyde and acetaldehyde but some measurements were con-ducted only in daytime To check the effect of diurnal varia-tions we compared the daytime average concentration (from0800 to 2000 LT) and daily average concentration of thesetwo carbonyls in 2011 The result showed that the relativedifferences between the daytime and daily average concen-trations were very low with the values of 7 in summerand 1 in winter It seemed the daytime and daily aver-age concentrations of these carbonyls provided by previ-ous studies can be compared to investigate carbonyl trendsIn contrast with the decreasing trend of carbonyl levels de-rived from ground-based measurements satellite measure-ments suggested that formaldehyde columns showed an in-creasing trend (De Smedt et al 2010) As the satellites mea-surements were conducted in midday and covered a largerspatial scale the ground-based measurements were consid-ered to better represent the trend of 24 h carbonyl concentra-tions in the urban area of Beijing The decreasing trend for

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3051

Fig 1 Diurnal variations of measured ratios of formaldehyde ethene acetaldehyde propene and acetone ethane at the PKU site duringwinter and summer Black lines represent the average values and grey shaded areas indicate standard deviations

summed concentrations of formaldehyde and acetaldehydeindicated that the emission controls on these carbonyls ortheir precursors in Beijing possibly made progress during thepast years However the flat trend during winter suggestedthat there might be some important sources for carbonyls inwinter were not under effective control

Figure 1 shows the diurnal variations of ratios of formalde-hyde ethene acetaldehyde propene and acetone ethane atthe PKU site during winter and summer Ambient concentra-tions of carbonyls were influenced by meteorological con-ditions and the emission strengths of carbonyls and theirprecursors Assuming that the direct emissions of carbonylsfrom anthropogenic sources were proportional to the emis-sions of NMHCs the diurnal variations of carbonyl NMHCratios could reflect the importance of secondary formationto ambient carbonyls These three pairs of carbonyls andNMHCs were chosen because the two species in each pairhad similar reaction rates with OH radicals (Atkinson et al2006) As a result the two species in the atmosphere wereremoved approximately equally by reactions with OH radi-cals and therefore the effect of photochemical degradationon their ratios can be neglected As shown in Fig 1 themeasured carbonyl NMHC ratios showed similar daily vari-ations both in summer and winter they were stable dur-ing nighttime increased after sunrise (about 0800 LT) andreached a maximum in the early afternoon (about 1300ndash1500 LT) then decreased to low values at night This varia-tion indicated an important contribution from secondary pro-duction during the daytime both in winter and summer Thediurnal variations of the acetone ethane ratios were less thanthose for formaldehyde ethene and acetaldehyde propeneSuch difference was consistent with previous studies whichfound that secondary production of acetone was less impor-tant than that of formaldehyde and acetaldehyde (de Gouw etal 2005 Yuan et al 2012)

Though the diurnal patterns of carbonyl NMHC ratioswere similar in winter and summer the values of car-bonyl NMHC ratios were approximately 3ndash5 times higher in

summer than those in winter during both day and night Asmentioned above ambient concentrations of carbonyls didnot show significant discrepancies between these two sea-sons The seasonal differences of carbonyl NMHC ratioswere mainly due to the much higher NMHC levels in win-ter This also indicated the significant differences in VOCsources between these two seasons Considering the effectsof photochemical processing on VOC ratios were minor atnight nighttime VOC ratios were close to their emission ra-tios (Wang et al 2013) Therefore the higher nighttime car-bonyl NMHC ratios in summer indicated that the fractionof NMHCs in primary emitted VOCs was higher in winterthan that in summer whereas carbonyls fraction was higherin summer

Previous studies usually supposed that carbonyls in win-ter mainly came from direct anthropogenic emissions (Pangand Mu 2006 Ceroacuten et al 2007) From our measure-ments the differences of carbonyl NMHC ratios betweennoon and night in summer were only slightly larger thanthe differences in winter The noon night ratios were 26and 31 for formaldehyde ethene 26 and 35 for acetalde-hyde propene and 14 and 15 for acetone ethane in winterand summer respectively This indicated that secondary for-mation was still an important source of carbonyls in winterIt should be noted that photolysis removal of carbonyls wasalso considered here besides the removal by reactions withOH radicals According to our estimation reaction with OHwas the main pathway of carbonyl removal The photolysiscontributed 33 1 and 12 to the losses of formalde-hyde acetaldehyde and acetone at daytime and these contri-butions showed no significant difference between winter andsummer The details for the calculation of carbonyl removalrates were shown in the Supplement As a result neglect-ing the photolysis of carbonyls had little effect on comparingthe contributions of secondary formation between winter andsummer and the variation of carbonyl NMHC ratios couldreflect the importance of secondary formation For furtherdiscussion the PMF model was used for carbonyls source

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3052 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 2Profiles of the four factors resolved by the PMF model and the distributions of species among these factors in summer

apportionment to determine relative contributions from pri-mary and secondary sources

32 Identifying PMF factors

The PMF model is a mathematical tool which separates pol-lutants according to the correlations among species and theunderstanding of PMF factors is subjective to a certain ex-tent We were aware of the deficiency of the PMF modelwhen apportioning VOC sources In this section we com-pared the profiles derived by the PMF model with previouslymeasured ones to check the reliability of resolved fresh fac-tors In the next section we tried to identify aged factorsby investigating the relationship between degradation of pri-mary emitted VOCs and production of carbonyls

For summertime measurements 593 samples were usedfor the PMF analysis and four factors were identified Theprofiles of resolved PMF factors and the distributions ofspecies among the factors were shown in Fig 2 The profileswere expressed as the mass percentage of individual VOCspecies in each factor and the distributions represented thecontributions of individual factors to the measured level ofeach species

Factor 1 made an important contribution to our measuredalkanes and alkenes The profile of this factor was dominatedby C2ndashC5 alkanes C2ndashC3 alkenes ethyne and aromaticsThese species were mainly emitted from exhaust of vehiclesand evaporation of gasoline (Liu et al 2008) This factor ex-plained 381 of the measured MTBE which was com-monly used as a tracer of gasoline vehicle emissions (Changet al 2006) Previous studies also reported that traffic-relatedemissions were the most important NMHCs sources in sum-mer of Beijing (Song et al 2008 Wang et al 2010) Wetherefore attributed this factor to traffic-related emissions

In factor 2 the loadings of aromatics including benzenetoluene ethylbenzene and xylenes were high Aromatics

were reported to be major constituents of solvents (Yuan etal 2010) which were widely used in industrial processes aswell as some household products Thus we considered thisfactor to be related to industry and solvent usage

Factor 3 had contributions to almost all measured VOCspecies except for some very reactive species This factor ac-counted for more than 60 of our measured carbonyls and875 of our measured 2-BuONO2 levels Alkyl nitrateswere believed to be mainly from secondary production in ur-ban regions and had relatively long lifetimes (Simpson et al2003) Thus we identified this factor as aged emissions

Factor 4 contributed 815 of measured isoprene whichwas often considered as a tracer for biogenic emissions (Situet al 2013) so we attributed this factor to biogenic emis-sions High loadings of formaldehyde acetaldehyde andacetone were also found in this factor This is possibly be-cause these carbonyls could also be emitted from biogenicsources (Winters et al 2009) and they were important ox-idation products of other biogenic VOCs (Carter and Atkin-son 1996)

Based on the derived profiles of each factor and the dis-tributions of species among these factors we identified thattwo of the four PMF-resolved factors could represent pri-mary emission sources another related to biogenic emis-sions and the other was an aged emission factor As dis-cussed in Sect 22 the relationship between the contributionof one factor to each NMHC species and its chemical reactiv-ity could be used to distinguish factors related to photochem-ically aged emissions from those for fresh emissions Such arelationship was shown in Fig 3 where each NMHC specieswas shown as a circle Positive correlations were identifiedfor factors 1 and 2 and thus these two factors were consid-ered to be related to fresh emissions However a significantnegative correlation was identified for factor 3 indicatingthat this factor represented photochemically aged emissions

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3053

Fig 3 Relationships between the factor contributions to each NMHC species andkOH values for NMHC species in summer Each circlerepresents one compound Carbonyls are shown as solid triangles in the graph but they were not used for correlation analysis due to theirsecondary productionlowastindicates a significant correlation at a confidence level of 005lowastlowastindicates a significant correlation at a confidencelevel of 001

The distributions of carbonyls in PMF-resolved factorswere shown in Fig 3 by solid triangles If all factors were re-lated to fresh emissions the distributions of carbonyls wouldbe similar to the distributions of NMHCs in each factor Dueto the influence from photochemical aging the distribution ofcarbonyls could be higher in aged factors owing to their sec-ondary production and meanwhile the distributions of car-bonyls would be lower in fresh factors In factors 1 and 2 thedistributions of carbonyls were similar or lower than those ofNMHCs whereas the distributions of carbonyls in factor 3were higher than those of NMHCs Such appearance agreedwith our above identifications on PMF-resolved factors

In this study two fresh factors and an aged factor wereidentified in the summer of 2011 However Yuan etal (2012) identified one mixed fresh factor and two aged fac-tors with different photochemical ages using a similar anal-ysis approach based on measurements at the same site in thesummer of 2010 It was interesting to get such different re-sults in two consecutive years Yuan et al (2012) concludedthat PMF results depended on the degree of photochemicalprocessing and the differences in emission compositions ofvarious sources Assuming that the differences in VOC emis-sions were not significant between 2010 and 2011 the PMFresults were mainly influenced by the difference in degrees

in photochemical aging between these two years We usedthe ratio of o-xylene to ethylbenzene as an indicator of thedegree of photochemical processing As seen in Fig S4 therelative standard deviations of o-xylene ethylbenzene ratiosin 2010 were 30 larger than those in 2011 It indicated thatthe range of photochemical processing degrees was larger in2010 As a result the PMF factors in 2010 were extractedmainly according to different degrees of photochemical pro-cessing whereas the PMF factors in 2011 could be extractedmainly based on individual sources

In winter 341 samples were used for the PMF analysisand five factors were identified (Fig 4)

Factor 1 and factor 3 contributed to most of our measuredalkanes and alkenes but there were some differences be-tween these two factors Factor 1 had higher contributions toC2ndashC3 NMHCs while factor 3 had higher contributions toC4ndashC5 NMHCs These light hydrocarbons were mainly gen-erated by incomplete combustion processes such as vehicleexhaust and coal burning (Liu et al 2008 Wang et al 2013)The measured source profiles in China showed that C2ndashC3NMHCs contributed more than half of NMHCs that emit-ted from coal burning (Liu et al 2008 Wang et al 2013)Benzene and toluene also made important contributions tofactor 1 These two species were also found to be important

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3054 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 4Profiles of the five factors resolved by the PMF model and the distributions of species among these factors in winter

components in NMHCs emissions from coal burning (Liu etal 2008) The ratio of benzene to toluene (231 ppb ppbminus1)

in this factor fell between the ratios measured by Liu etal (2008) for residential coal burning (181 ppb ppbminus1) andindustrial coal burning (262 ppb ppbminus1) but much higherthan that measured in the tunnel experiment (070 ppb ppbminus1)

(Liu et al 2008) therefore factor 1 was identified as coalburning C4ndashC5 NMHCs were important species from trafficrelated emissions and factor 3 showed similar characteris-tic with factor 1 in summer In addition factor 3 explained379 of the measured MTBE Therefore factor 3 was at-tributed to traffic related emissions

Factor 2 had high contributions to aromatics and showedsimilar characteristic with the resolved factor 2 for sum-mertime measurements Besides aromatics this factor con-tributed about 30 of measured long-chain alkanes (C7ndashC10 alkanes) The long-chain alkanes were also consid-ered to be important components of printing VOC emissions(Yuan et al 2010) Thus this factor was identified as indus-try and solvent usage

Both factor 4 and factor 5 had important contributionsto carbonyls but little contribution to reactive NMHCs andtherefore these two factors were inferred to possibly not rep-resent fresh emissions The major NMHC species in factor 4were C2-C5 alkanes ethene ethyne benzene and tolueneThese species could be emitted from all of the three primaryfactors discussed above and thus we considered this factor tobe aged emissions from these anthropogenic sources TheNMHCs in factor 5 were dominated by unreactive alkanessuch as ethane and propane so factor 5 was at a more aged

stage than factor 4 Factor 5 accounted for 677 of the mea-sured 2-BuONO2 levels and 197 of the measured carbonyllevels According to the abundances of alkyl nitrates and un-reactive alkanes we considered that the loadings of VOCspecies in this factor were related to secondary productionand background levels

Similarly with summertime measurements the relation-ships between the contribution of one factor to each NMHCspecies andkOH values for each NMHC species were ana-lyzed As shown in Fig 5 no significant correlation was ob-served for factor 1 and factor 2 while a positive correlationwas found for factor 3 and thus these three factors were con-sidered to be fresh emissions Significant negative correla-tions were identified for factor 4 and factor 5 indicating thatthese two factors were influenced by photochemical aging

33 Exploring the aged and fresh emission factors

In the last section we distinguished photochemically agedfactors from fresh factors To better understand the relation-ships between aged and fresh factors and the role of photo-chemical aging in PMF analysis we explored the relation-ships between VOC consumption and carbonyl formation inthe aged factors

For two factors representing different photochemicalstages of the same emissions the ratios of NMHC abun-dances in these two factors should follow the equation below(Yuan et al 2012)

[NMHC]aged

[NMHC]fresh= D times eminuskOHNMHC[OH]1t (4)

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3055

Fig 5 Relationships between the factor contributions to each NMHC species andkOH values of NMHC species in winter Each circlerepresents one compound Carbonyls are shown as solid triangles in the graph but they were not used for correlation analysis due to theirsecondary productionlowastlowastindicates a significant correlation at a confidence level of 001

where [NMHC]agedand [NMHC]fresh(ppb) refer to the abun-dances of NMHC species in the aged and fresh factor respec-tively D is a scaling factor which normalizes the NMHCabundances to unitykOHNMHC is the OH rate constant forthe NMHC [OH] is the average concentration of OH rad-ical and1t is the difference of reaction time between theaged and fresh factors In this study the OH exposure (iethe time integrate of [OH]) expressed by [OH]1t is calcu-lated as a whole

As discussed in Sect 32 for wintertime measurementsthe [NMHC]agedrefers to the abundances of NMHC in fac-tor 4 while the [NMHC]fresh refers to the sum of the abun-dances of NMHC in factors 1ndash3 For summertime measure-ments the [NMHC]agedrefers to the abundances of NMHC

in factor 3 while the [NMHC]fresh refers to the sum of theabundances of NMHC in factors 1 and 2 Figure 6 showedthe dependence of [NMHC]aged [NMHC] fresh (circles in thefigure) onkOH values with the lines being the fitted resultsfrom Eq (4) The values ofD and [OH]1t were estimated tobe 043 and 299times 1010 molecule cmminus3 s in winter and 126and 105times 1011 molecule cmminus3 s in summer respectivelyReferring to the calculated OH radical concentrations in thesummer in Beijing (Z Liu et al 2012) and the differencesof measured OH concentrations in Tokyo between summerand winter (Kanaya et al 2007) we assumed that the aver-age daytime concentration of OH in Beijing was 15times 106

and 6times 106 molecule cmminus3 in winter and summer thus the

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3056 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 6 Relationships between the ratios of VOCs abundances inthe aged emissions factors with those in fresh emission factors andkOH values for each VOC species Each data point represents onecompound The lines are the fitted results from Eq (4)

photochemical ages of the aged factor in winter and summerwere estimated to be 55 and 49 h respectively

The [carbonyl]aged [carbonyl]fresh ratios (Fig 6 solidtriangles) were significantly higher than the ratios of[NMHC]aged [NMHC] fresh because secondary productionenhanced the abundances of carbonyls in aged factors Asshown in Eq (5) the abundances of carbonyls in aged fac-tors can be separated into two parts The first part standsfor aged primary emissions and the second part stands forsecondary formation calculated based on the consumption ofother VOCs[carbonyl

]aged= D times

[carbonyl

]freshtimes eminuskOHcarbonyltimes[OH]1t (5)

+D times

sum([VOC]freshtimes

(1minus eminuskOHVOCtimes[OH]1t

)times YVOCcarbonyl

)wherekOHcarbonyl is the OH rate constant for the carbonylYVOCcarbonyl (ppbppb) refers to the carbonyl productionyield from the oxidation of particular VOC which canbe estimated using the Leeds master chemical mechanism(MCM v32 httpmcmleedsacukMCM) The estimationof YVOCcarbonyl is based on the following assumptions theremoval of VOC is governed by the reaction with OH radi-cals and the removal of alkyl peroxyl radicals (RO2) is gov-erned by the reaction with NO TheYVOCcarbonylvalues usedin this study are listed in Table 2 The VOC species that makeno contributions to the production of our focused carbonylswere not shown in this study The values ofD and [OH]1t

were fitted from Eq (4)The carbonyl abundances in aged factors were calculated

by Eq (5) and agreed well with the PMF-resolved abun-dances (Fig 7) In winter all data points were distributednear the 1 1 line In summer though the data points showedsome variability but still agreed within a factor of two Theagreement between the calculated and PMF-resolved car-bonyl abundances indicated that our understanding on thephotochemical aging of PMF factors was acceptable Thedifferences between the calculated and PMF-resolved car-bonyl abundances may be due to one or some of the fol-lowing reasons (1) the PMF analysis andYVOCcarbonyl es-timations have their own uncertainties (2) Only reactions of

VOCs with OH radicals are considered in our analysis (3)Further reactions of secondary carbonyls are not consideredin our analysis (4) Some VOC species are not applied in ourPMF analysis but they might be precursors of carbonyls ThePMF-analyzed VOC species accounted for 90 and 87 oftotal concentrations of all measured VOCs species in winterand summer respectively However the contribution of eachVOC to the formation of carbonyls was not only dependenton its concentration but also its ability to produce carbonyls

In this work we identified the PMF factors in summerand winter as several individual fresh emissions and sev-eral mixed aged emissions with different degrees of photo-chemical processing Beijing was a large city with extensiveand continuous local emissions and therefore our measuredair mass was in fact a mixture of fresh and aged plumesThe fresh emissions were separated by the PMF model ac-cording to the correlations among VOC species and the dif-ferent emission characteristics of these sources Howeverthe resolved factors by the PMF model cannot represent theemissions at different aged stages from individual sourcesAs discussed in this section the respective photochemicalages of the aged factors in winter and summer were respec-tively about 55 and 49 h and the factor representing sec-ondary production and background levels in winter should bemore aged We conjectured that VOCs from different sourceshave been well mixed during such long aging process and itwas hard to identify the aged contributions from individualsources Likewise previous studies also used averaged emis-sion ratios to characterize the emission and aging process ofurban VOC emissions (de Gouw et al 2005 Warneke et al2007 Yuan et al 2012 Borbon et al 2013) In this sec-tion we proved the reasonability of the PMF-derived agedfactors However one problem remained when apportioningreactive VOCs using the PMF model Actually the photo-chemical reaction of VOCs in the atmosphere is a continuousprocess but the PMF model is not able to describe such kindof process In this work several different aged stages wererecognized by the PMF model and the PMF model used thelinear combination of these stages to describe the continuousphotochemical process This might be an important issue forthe PMF analysis and the influence of such approximationrequired future researches

34 Primary and secondary sources of carbonyls

The calculated relative contributions of carbonyl sources ap-portioned by the PMF model in Beijing were shown in Fig 8Previous studies usually attributed all carbonyls in aged fac-tors as secondary formation As discussed in Sect 33 theabundances of carbonyls in aged factors could be separatedinto two parts that is the aged direct emissions and theproduction from VOC consumption and the former partshould be treated as primary emissions In this study theaged direct emission part was further separated into eachfresh factor considering their relative contributions For all

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

3048 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

emissions some markers might be appropriate for certainemission sources (Rappengluumlck et al 2010) and thus the useof a single marker for all primary emissions could be veryrisky (Friedfeld et al 2002) In addition reactive VOCs thatare co-emitted with a primary emission marker can form sec-ondary carbonyls during air mass transport Hence the cor-relation will also exist between secondary carbonyls and theprimary marker For secondary formation the correlation co-efficient and the slope between carbonyls and the secondarymarker vary significantly depending on the VOCs NOx ratioand the degree of photochemical processing Furthermoreprimary carbonyls also can be precursors of these secondarymarkers which may also lead to a correlation between them(Parrish et al 2012)

To improve the multi-linear regression method de Gouwet al (2005) developed a parameterization method based onthe degree of photochemical processing (ie photochemicalage) with the consideration of the production of carbonyls bytheir precursors and the chemical removal of carbonyls dur-ing transport This method made several assumptions whichshould be carefully considered (1) anthropogenic emissionsof carbonyls and their precursors are proportional to a pri-mary marker (2) The removal of VOCs is governed by thereactions with OH radicals (3) Biogenic sources of car-bonyls are proportional to the emission of isoprene (4) Thephotochemical ages for sampled air masses can be deter-mined (de Gouw et al 2005) Positive matrix factorization(PMF) a receptor model has also been used to separatesources of carbonyls (Bon et al 2011 Buzcu Guven andOlaguer 2011 Yuan et al 2012 Zheng et al 2013) Com-pared with the multi-linear regression method which em-ploys one species for a certain source the PMF model canuse a number of species for source identification Most pre-vious studies identified several primary emission sources anda mixed secondary source among the PMF-resolved factorsHowever Yuan et al (2012) found that the PMF-resolvedfactors could be attributed to a common source at differentstages of photochemical processing rather than several inde-pendent sources In addition to the various source apportion-ment approaches based on carbonyl measurements Parrish etal (2012) compared the primary emission flux of formalde-hyde with fluxes of its precursors to quantify the relative con-tributions of primary and secondary formaldehyde sourcesHowever this approach relies on the accuracy of emissioninventories and reaction mechanisms

Large differences have been found among the above-mentioned methods used to calculate sources of carbonylsFor example several studies were conducted in the Hous-ton area TX USA to determine the contributions of pri-mary anthropogenic sources of formaldehyde Based on themulti-linear regression using CO as a primary marker Fried-feld et al (2002) reported that primary emissions contributed36 of atmospheric formaldehyde In a later study Rap-pengluumlck et al (2010) calculated the contributions of 385 and 89 for formaldehyde from primary vehicular emis-

sions and industrial emissions using CO and SO2 as primarymarkers respectively By using the PMF Buzcu Guven andOlaguer (2011) estimated that mobile sources and industrialsources contributed 23 and 17 of formaldehyde respec-tively In contrast by comparing emissions of formaldehydeand its precursors Parrish et al (2012) found that only 1 and 4 of formaldehyde were respectively emitted from ve-hicle exhaust and industrial sources and that secondary for-mation accounted for the overwhelming majority of atmo-spheric formaldehyde The source apportionment studies forcarbonyls were also conducted in Beijing Li et al (2010) at-tributed 76 of atmospheric formaldehyde to direct anthro-pogenic emissions using the multi-linear regression methodwhile Yuan et al (2012) calculated that this contributionwas only 22 using the parameterization method Comparedwith the number of formaldehyde studies there were fewersource apportionment studies for other carbonyls and mostof these studies were conducted at different sites (Goldsteinand Schade 2000 de Gouw et al 2005 Liu et al 2009Yuan et al 2012 Guo et al 2013) The only two studies thatcould be used for inter-comparison were taken in the sum-mer in Beijing and the apportioned primary contributions ofacetaldehyde were 16 and 46 in 2005 and 2010 respec-tively (Liu et al 2009 Yuan et al 2012) It was unclearthat such difference was caused by the variation of carbonylsources or the uncertainty of source apportionment

Though the State Council of China released mandatory re-quirement to improve air quality all over China the way toachieve the goal could be very tough (Liu et al 2013) Bei-jing is the capital of China and suffers from serious air pollu-tion problems characterized by high levels of ground-levelozone and PM25 (Shao et al 2006 Wang et al 2012a)Ambient carbonyls were detected at high levels in Beijing(Liu et al 2009 Zhang et al 2012) and formaldehyde dis-played a significant positive trend from 1997 to 2010 (DeSmedt et al 2010) However ambient concentrations of non-methane hydrocarbon (NMHC) were found to decrease from2003 (Wang et al 2012b) It is essential to quantify contribu-tions of different sources to carbonyls for understanding suchtrends and improving air quality in Beijing Yuan et al (2012)used the PMF model for carbonyl source apportionment in2010 summer and found that factors in PMF were actuallythe results of different degrees of photochemical process-ing rather than individual sources In this study online mea-surements of carbonyls and other VOCs were conducted atPeking University (PKU) site in the winter and summer Weused PMF to apportion carbonyl sources in these two sea-sons with different degrees of photochemical processing Thedegradation of NMHCs and production of carbonyls in theaged factors were analyzed based on the processes of photo-chemical aging to verify the PMF results Differences in car-bonyl sources between winter and summer were discussedand compared with previous studies

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3049

2 Methods

21 VOCs measurements

Online VOCs measurements were conducted at an urban sitein Beijing from 4 August 2011 to 9 September 2011 and from29 December 2011 to 18 January 2012 The sampling sitewas on the top of a six-floor building (sim 20 m above ground)on Peking University campus The PKU campus is locatedin the northwest of Beijing about 500 m north of the FourthRing Road This site was described in detail elsewhere (Songet al 2008 Liu et al 2009)

A custom-built online gas chromatography systemequipped with a mass spectrometry and a flame ionizationdetector (GC-MSFID) was used to measure C2ndashC10 hydro-carbons C3ndashC6 carbonyls C1ndashC4 alkyl nitrates halocar-bons and methyl tert-butyl ether (MTBE) with a time reso-lution of 1 h CFC-113 (112-trichloro-122-trifluoroethane)was used as a natural internal standard from ambient air dueto its long lifetime and minimal manmade emissions (W-T Liu et al 2012) Daily calibrations were performed everyday and the variations of target species responses were re-quired to be withinplusmn20 from the calibration curve Thedetection limits for each species varied from 1 to 20 ppt andthe precisions ranged from 1 to 6 The detailed informa-tion of this system can be found in Yuan et al (2012)

A commercial high-sensitivity proton-transfer-reactionmass spectrometry (PTR-MS) (Ionicon Analytik InnsbruckAustria) was used to measure 28 masses including C1ndashC4carbonyls and C6ndashC9 aromatics with a time resolution ofabout 30 s Background signals were measured for 15 min ev-ery 25 h by diverting the ambient air sample flow througha Pt-coated quartz wool converter at 370C (Yuan et al2013) Calibrations were done every two or three days andthe response factors varied within 20 The detection lim-its of each species varied from 40 to 200 ppt except forformaldehyde As the proton affinity of formaldehyde is onlyslightly higher than water the back reaction of proton trans-fer becomes significant The kinetic of PTR-MS formalde-hyde detection was found to be mainly influenced by humid-ity (Vlasenko et al 2010 Warneke et al 2011) Thereforewe calibrated our formaldehyde measurement by using gasstandards provided by a permeation tube (Kin-Tek USA) atthe humidity from 0ndash30 mmol molminus1 to obtain the curve offormaldehyde response factor on humidity During the entireambient measurement formaldehyde signals were correctedaccording to the response curve Due to the different ambi-ent humidity between winter and summer the detection lim-its of formaldehyde were 022ndash034 ppb in winter and 045ndash080 ppb in summer at a time resolution of 30 s Among allour data only less than 02 of formaldehyde concentra-tions in summer fell below the detection limit

During these two campaigns C3ndashC4 carbonyls C6ndashC9aromatics isoprene and MVK+MACR (methyl vinyl ke-tone+ methacrolein) were measured simultaneously by on-

line GC-MS and PTR-MS Good agreements were found be-tween these two systems for most species with correlationcoefficients larger than 090 and slopes ranging from 08 to12 Formaldehyde and acetaldehyde were only measured byPTR-MS However inter-comparison between the PTR-MSand 24-dinitrophenyl hydrazinendashhigh-pressure liquid chro-matography (DNPHndashHPLC) methods (EPA TO-11A) wasperformed before the field measurements By correcting theinfluence of humidity on formaldehyde response a correla-tion coefficient of 093 and a slope of 106 were obtained be-tween measurements by PTR-MS and DNPHndashHPLC meth-ods The detailed results of inter-comparison were shown inthe Supplement

22 Positive matrix factorization (PMF)

The US Environmental Protection Agencyrsquos PMF 30 recep-tor model was used for VOCs source apportionment in thisstudy The PMF is a multivariate factor analysis tool whichassumes that measured concentrations at receptor sites arelinear combinations of contributions from different sources(Paatero and Tapper 1994) Generally an ambient data setcan be viewed as a data matrixx with i byj dimensions inwhich i samples andj chemical species are measured Thegoal of multivariate receptor modeling is to identify the num-berp of factors the species profilef of each source the massg contributed by each factor to each individual sample andthe residuale (Eq 1)

xij =

sump

k=1gikfkj + eij (1)

The PMF solution minimizes the object functionQ basedupon the uncertainties (Eq 2)

Q =

sumn

i=1

summ

j=1

[xij minus

sump

k=1gikfkjuij

]2

(2)

whereu represents the uncertainty of each data The theo-retical Q (Qtheoretical) can be calculated as Eq (3) and thebest PMF solution should haveQQtheoreticalwith the valueof sim 1

Qtheoretical= i times j minus p times (i + j) (3)

The basic assumption of the PMF model is that the fittingspecies are not allowed for chemical losses or productionThus apportioning carbonyl sources using the PMF modelshould be approached with care As mentioned in the in-troduction several studies apportioned carbonyl sources byusing the PMF model (Bon et al 2011 Buzcu Guven andOlaguer 2011 Zheng et al 2013) However the study toinvestigate the validity of the results was not available yetYuan et al (2012) proved the capacity of the PMF approachin identifying the role of chemical aging for better under-standing the PMF factors The VOC emission ratios derivedfrom the PMF fresh factors agreed well with the ones calcu-lated based on photochemical ages indicating that the PMF

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3050 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Table 1The sum of formaldehyde and acetaldehyde concentrationsin Beijing (unit ppb)

Year Summer Winter

2005 217a 70a

2006 1312b

2008 133cd 66e

2009 114e 59e

2010 126e

2011 80f 81f

a(Pang and Mu 2006)b(Shao et al 2009)c(Xu et al 2010)ddata before traffic restriction in 2008e(Zhang et al 2012)f this study

approach could identify the contributions from primary emis-sions reasonably Additionally the abundances of NMHCsin the PMF-aged factors could be reproduced by the photo-chemical aging of fresh factors In this study we run PMFmodels using two VOC data sets from different seasons anddiscussed the apportionment results to investigate whetherthe PMF approach can separate carbonyl sources well

For the PMF analysis uncertainties of online GC-MSFIDand PTR-MS measurements were calculated using themethod by Yuan et al (2012) Generally for the species mea-sured by the GC-MSFID system the uncertainties were cal-culated as the sum of 10 of measured concentrations andone-third of detection limits For the species measured bythe PTR-MS the uncertainties were calculated as the sum of10 of concentrations and the errors from mass spectrome-ter measurements Finally only species with high signal-to-noise ratios and few missing data values were used for thePMF analysis In winter 10 carbonyls 26 NMHCs 2-butylnitrate (2-BuONO2) tetrachloroethylene and MTBE wereused In summer most of the values of cistrans-2-butene andcistrans-2-pentene fell below the detection limits so these 4species were not used for the PMF analysis

In this work the PMF factor numbers were explored from3 to 8 for the best solution The final numbers of factors de-pended on two aspects (1)QQtheoretical which was usedto characterize the quality of the reconstruction and (2) thephysical plausibility of the factors (Bon et al 2011) Four-and five-factor resolutions were selected in summer and win-ter respectively The free rotation of the PMF factors wasinvestigated by adjusting the Fpeak parameter betweenminus5and 5 The results with non-zero Fpeak values were gen-erally consistent with the runs with zero Fpeak values andthus the results used in this study were from the runs withnon-rotation

As the PMF-resolved factors might be influenced by pho-tochemical processes we used the relationship between thecontribution of one factor to each NMHC species and itschemical reactivity (kOH) to distinguish fresh and aged fac-

tors as performed by Yuan et al (2012) Generally if allPMF-resolved factors were originated from fresh emissionsthe distribution of each species would not correlate with itschemical reactivity As the air mass from a source was age-ing the NMHCs underwent photochemical reactions and themore reactive species would be more largely consumed As aresult a negative correlation could be seen between an agedfactor contribution to each NMHC species and itskOH valueConsidering that such complex relationships were difficult toexpress in mathematical formulae Spearmanrsquos coefficient ofrank correlation (rs) was used to represent the relationship

3 Results and discussion

31 Characteristics of ambient carbonyls in Beijing

The average concentration of the 10 measured carbonyls inBeijing was 132plusmn 79 ppb (average valueplusmn standard devi-ation) in winter In summer the average concentration wentup to 163plusmn 74 ppb Formaldehyde acetone and acetalde-hyde were the three major species contributing more than80 of the total measured carbonyls in both seasons Am-bient concentrations of carbonyls in Beijing were measuredby several previous studies from 2005 (Pang and Mu 2006Shao et al 2009 Xu et al 2010 Zhang et al 2012) Dur-ing the recent years VOC sources changed remarkably inBeijing resulting from quick increment of vehicular pop-ulation stricter standards for vehicle emissions (Zhang etal 2012) and solvent compositions (Yuan et al 2010) anddramatic changes in industries Table 1 shows the sum ofmeasured formaldehyde and acetaldehyde concentrations inBeijing during 2005ndash2011 The measured concentrations insummer showed a significant decrement from 2005 to 2006and from 2010 to 2011 while the wintertime concentra-tions did not exhibit a significant change Some of the abovestudies provided daily average concentrations of formalde-hyde and acetaldehyde but some measurements were con-ducted only in daytime To check the effect of diurnal varia-tions we compared the daytime average concentration (from0800 to 2000 LT) and daily average concentration of thesetwo carbonyls in 2011 The result showed that the relativedifferences between the daytime and daily average concen-trations were very low with the values of 7 in summerand 1 in winter It seemed the daytime and daily aver-age concentrations of these carbonyls provided by previ-ous studies can be compared to investigate carbonyl trendsIn contrast with the decreasing trend of carbonyl levels de-rived from ground-based measurements satellite measure-ments suggested that formaldehyde columns showed an in-creasing trend (De Smedt et al 2010) As the satellites mea-surements were conducted in midday and covered a largerspatial scale the ground-based measurements were consid-ered to better represent the trend of 24 h carbonyl concentra-tions in the urban area of Beijing The decreasing trend for

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3051

Fig 1 Diurnal variations of measured ratios of formaldehyde ethene acetaldehyde propene and acetone ethane at the PKU site duringwinter and summer Black lines represent the average values and grey shaded areas indicate standard deviations

summed concentrations of formaldehyde and acetaldehydeindicated that the emission controls on these carbonyls ortheir precursors in Beijing possibly made progress during thepast years However the flat trend during winter suggestedthat there might be some important sources for carbonyls inwinter were not under effective control

Figure 1 shows the diurnal variations of ratios of formalde-hyde ethene acetaldehyde propene and acetone ethane atthe PKU site during winter and summer Ambient concentra-tions of carbonyls were influenced by meteorological con-ditions and the emission strengths of carbonyls and theirprecursors Assuming that the direct emissions of carbonylsfrom anthropogenic sources were proportional to the emis-sions of NMHCs the diurnal variations of carbonyl NMHCratios could reflect the importance of secondary formationto ambient carbonyls These three pairs of carbonyls andNMHCs were chosen because the two species in each pairhad similar reaction rates with OH radicals (Atkinson et al2006) As a result the two species in the atmosphere wereremoved approximately equally by reactions with OH radi-cals and therefore the effect of photochemical degradationon their ratios can be neglected As shown in Fig 1 themeasured carbonyl NMHC ratios showed similar daily vari-ations both in summer and winter they were stable dur-ing nighttime increased after sunrise (about 0800 LT) andreached a maximum in the early afternoon (about 1300ndash1500 LT) then decreased to low values at night This varia-tion indicated an important contribution from secondary pro-duction during the daytime both in winter and summer Thediurnal variations of the acetone ethane ratios were less thanthose for formaldehyde ethene and acetaldehyde propeneSuch difference was consistent with previous studies whichfound that secondary production of acetone was less impor-tant than that of formaldehyde and acetaldehyde (de Gouw etal 2005 Yuan et al 2012)

Though the diurnal patterns of carbonyl NMHC ratioswere similar in winter and summer the values of car-bonyl NMHC ratios were approximately 3ndash5 times higher in

summer than those in winter during both day and night Asmentioned above ambient concentrations of carbonyls didnot show significant discrepancies between these two sea-sons The seasonal differences of carbonyl NMHC ratioswere mainly due to the much higher NMHC levels in win-ter This also indicated the significant differences in VOCsources between these two seasons Considering the effectsof photochemical processing on VOC ratios were minor atnight nighttime VOC ratios were close to their emission ra-tios (Wang et al 2013) Therefore the higher nighttime car-bonyl NMHC ratios in summer indicated that the fractionof NMHCs in primary emitted VOCs was higher in winterthan that in summer whereas carbonyls fraction was higherin summer

Previous studies usually supposed that carbonyls in win-ter mainly came from direct anthropogenic emissions (Pangand Mu 2006 Ceroacuten et al 2007) From our measure-ments the differences of carbonyl NMHC ratios betweennoon and night in summer were only slightly larger thanthe differences in winter The noon night ratios were 26and 31 for formaldehyde ethene 26 and 35 for acetalde-hyde propene and 14 and 15 for acetone ethane in winterand summer respectively This indicated that secondary for-mation was still an important source of carbonyls in winterIt should be noted that photolysis removal of carbonyls wasalso considered here besides the removal by reactions withOH radicals According to our estimation reaction with OHwas the main pathway of carbonyl removal The photolysiscontributed 33 1 and 12 to the losses of formalde-hyde acetaldehyde and acetone at daytime and these contri-butions showed no significant difference between winter andsummer The details for the calculation of carbonyl removalrates were shown in the Supplement As a result neglect-ing the photolysis of carbonyls had little effect on comparingthe contributions of secondary formation between winter andsummer and the variation of carbonyl NMHC ratios couldreflect the importance of secondary formation For furtherdiscussion the PMF model was used for carbonyls source

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3052 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 2Profiles of the four factors resolved by the PMF model and the distributions of species among these factors in summer

apportionment to determine relative contributions from pri-mary and secondary sources

32 Identifying PMF factors

The PMF model is a mathematical tool which separates pol-lutants according to the correlations among species and theunderstanding of PMF factors is subjective to a certain ex-tent We were aware of the deficiency of the PMF modelwhen apportioning VOC sources In this section we com-pared the profiles derived by the PMF model with previouslymeasured ones to check the reliability of resolved fresh fac-tors In the next section we tried to identify aged factorsby investigating the relationship between degradation of pri-mary emitted VOCs and production of carbonyls

For summertime measurements 593 samples were usedfor the PMF analysis and four factors were identified Theprofiles of resolved PMF factors and the distributions ofspecies among the factors were shown in Fig 2 The profileswere expressed as the mass percentage of individual VOCspecies in each factor and the distributions represented thecontributions of individual factors to the measured level ofeach species

Factor 1 made an important contribution to our measuredalkanes and alkenes The profile of this factor was dominatedby C2ndashC5 alkanes C2ndashC3 alkenes ethyne and aromaticsThese species were mainly emitted from exhaust of vehiclesand evaporation of gasoline (Liu et al 2008) This factor ex-plained 381 of the measured MTBE which was com-monly used as a tracer of gasoline vehicle emissions (Changet al 2006) Previous studies also reported that traffic-relatedemissions were the most important NMHCs sources in sum-mer of Beijing (Song et al 2008 Wang et al 2010) Wetherefore attributed this factor to traffic-related emissions

In factor 2 the loadings of aromatics including benzenetoluene ethylbenzene and xylenes were high Aromatics

were reported to be major constituents of solvents (Yuan etal 2010) which were widely used in industrial processes aswell as some household products Thus we considered thisfactor to be related to industry and solvent usage

Factor 3 had contributions to almost all measured VOCspecies except for some very reactive species This factor ac-counted for more than 60 of our measured carbonyls and875 of our measured 2-BuONO2 levels Alkyl nitrateswere believed to be mainly from secondary production in ur-ban regions and had relatively long lifetimes (Simpson et al2003) Thus we identified this factor as aged emissions

Factor 4 contributed 815 of measured isoprene whichwas often considered as a tracer for biogenic emissions (Situet al 2013) so we attributed this factor to biogenic emis-sions High loadings of formaldehyde acetaldehyde andacetone were also found in this factor This is possibly be-cause these carbonyls could also be emitted from biogenicsources (Winters et al 2009) and they were important ox-idation products of other biogenic VOCs (Carter and Atkin-son 1996)

Based on the derived profiles of each factor and the dis-tributions of species among these factors we identified thattwo of the four PMF-resolved factors could represent pri-mary emission sources another related to biogenic emis-sions and the other was an aged emission factor As dis-cussed in Sect 22 the relationship between the contributionof one factor to each NMHC species and its chemical reactiv-ity could be used to distinguish factors related to photochem-ically aged emissions from those for fresh emissions Such arelationship was shown in Fig 3 where each NMHC specieswas shown as a circle Positive correlations were identifiedfor factors 1 and 2 and thus these two factors were consid-ered to be related to fresh emissions However a significantnegative correlation was identified for factor 3 indicatingthat this factor represented photochemically aged emissions

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3053

Fig 3 Relationships between the factor contributions to each NMHC species andkOH values for NMHC species in summer Each circlerepresents one compound Carbonyls are shown as solid triangles in the graph but they were not used for correlation analysis due to theirsecondary productionlowastindicates a significant correlation at a confidence level of 005lowastlowastindicates a significant correlation at a confidencelevel of 001

The distributions of carbonyls in PMF-resolved factorswere shown in Fig 3 by solid triangles If all factors were re-lated to fresh emissions the distributions of carbonyls wouldbe similar to the distributions of NMHCs in each factor Dueto the influence from photochemical aging the distribution ofcarbonyls could be higher in aged factors owing to their sec-ondary production and meanwhile the distributions of car-bonyls would be lower in fresh factors In factors 1 and 2 thedistributions of carbonyls were similar or lower than those ofNMHCs whereas the distributions of carbonyls in factor 3were higher than those of NMHCs Such appearance agreedwith our above identifications on PMF-resolved factors

In this study two fresh factors and an aged factor wereidentified in the summer of 2011 However Yuan etal (2012) identified one mixed fresh factor and two aged fac-tors with different photochemical ages using a similar anal-ysis approach based on measurements at the same site in thesummer of 2010 It was interesting to get such different re-sults in two consecutive years Yuan et al (2012) concludedthat PMF results depended on the degree of photochemicalprocessing and the differences in emission compositions ofvarious sources Assuming that the differences in VOC emis-sions were not significant between 2010 and 2011 the PMFresults were mainly influenced by the difference in degrees

in photochemical aging between these two years We usedthe ratio of o-xylene to ethylbenzene as an indicator of thedegree of photochemical processing As seen in Fig S4 therelative standard deviations of o-xylene ethylbenzene ratiosin 2010 were 30 larger than those in 2011 It indicated thatthe range of photochemical processing degrees was larger in2010 As a result the PMF factors in 2010 were extractedmainly according to different degrees of photochemical pro-cessing whereas the PMF factors in 2011 could be extractedmainly based on individual sources

In winter 341 samples were used for the PMF analysisand five factors were identified (Fig 4)

Factor 1 and factor 3 contributed to most of our measuredalkanes and alkenes but there were some differences be-tween these two factors Factor 1 had higher contributions toC2ndashC3 NMHCs while factor 3 had higher contributions toC4ndashC5 NMHCs These light hydrocarbons were mainly gen-erated by incomplete combustion processes such as vehicleexhaust and coal burning (Liu et al 2008 Wang et al 2013)The measured source profiles in China showed that C2ndashC3NMHCs contributed more than half of NMHCs that emit-ted from coal burning (Liu et al 2008 Wang et al 2013)Benzene and toluene also made important contributions tofactor 1 These two species were also found to be important

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3054 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 4Profiles of the five factors resolved by the PMF model and the distributions of species among these factors in winter

components in NMHCs emissions from coal burning (Liu etal 2008) The ratio of benzene to toluene (231 ppb ppbminus1)

in this factor fell between the ratios measured by Liu etal (2008) for residential coal burning (181 ppb ppbminus1) andindustrial coal burning (262 ppb ppbminus1) but much higherthan that measured in the tunnel experiment (070 ppb ppbminus1)

(Liu et al 2008) therefore factor 1 was identified as coalburning C4ndashC5 NMHCs were important species from trafficrelated emissions and factor 3 showed similar characteris-tic with factor 1 in summer In addition factor 3 explained379 of the measured MTBE Therefore factor 3 was at-tributed to traffic related emissions

Factor 2 had high contributions to aromatics and showedsimilar characteristic with the resolved factor 2 for sum-mertime measurements Besides aromatics this factor con-tributed about 30 of measured long-chain alkanes (C7ndashC10 alkanes) The long-chain alkanes were also consid-ered to be important components of printing VOC emissions(Yuan et al 2010) Thus this factor was identified as indus-try and solvent usage

Both factor 4 and factor 5 had important contributionsto carbonyls but little contribution to reactive NMHCs andtherefore these two factors were inferred to possibly not rep-resent fresh emissions The major NMHC species in factor 4were C2-C5 alkanes ethene ethyne benzene and tolueneThese species could be emitted from all of the three primaryfactors discussed above and thus we considered this factor tobe aged emissions from these anthropogenic sources TheNMHCs in factor 5 were dominated by unreactive alkanessuch as ethane and propane so factor 5 was at a more aged

stage than factor 4 Factor 5 accounted for 677 of the mea-sured 2-BuONO2 levels and 197 of the measured carbonyllevels According to the abundances of alkyl nitrates and un-reactive alkanes we considered that the loadings of VOCspecies in this factor were related to secondary productionand background levels

Similarly with summertime measurements the relation-ships between the contribution of one factor to each NMHCspecies andkOH values for each NMHC species were ana-lyzed As shown in Fig 5 no significant correlation was ob-served for factor 1 and factor 2 while a positive correlationwas found for factor 3 and thus these three factors were con-sidered to be fresh emissions Significant negative correla-tions were identified for factor 4 and factor 5 indicating thatthese two factors were influenced by photochemical aging

33 Exploring the aged and fresh emission factors

In the last section we distinguished photochemically agedfactors from fresh factors To better understand the relation-ships between aged and fresh factors and the role of photo-chemical aging in PMF analysis we explored the relation-ships between VOC consumption and carbonyl formation inthe aged factors

For two factors representing different photochemicalstages of the same emissions the ratios of NMHC abun-dances in these two factors should follow the equation below(Yuan et al 2012)

[NMHC]aged

[NMHC]fresh= D times eminuskOHNMHC[OH]1t (4)

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3055

Fig 5 Relationships between the factor contributions to each NMHC species andkOH values of NMHC species in winter Each circlerepresents one compound Carbonyls are shown as solid triangles in the graph but they were not used for correlation analysis due to theirsecondary productionlowastlowastindicates a significant correlation at a confidence level of 001

where [NMHC]agedand [NMHC]fresh(ppb) refer to the abun-dances of NMHC species in the aged and fresh factor respec-tively D is a scaling factor which normalizes the NMHCabundances to unitykOHNMHC is the OH rate constant forthe NMHC [OH] is the average concentration of OH rad-ical and1t is the difference of reaction time between theaged and fresh factors In this study the OH exposure (iethe time integrate of [OH]) expressed by [OH]1t is calcu-lated as a whole

As discussed in Sect 32 for wintertime measurementsthe [NMHC]agedrefers to the abundances of NMHC in fac-tor 4 while the [NMHC]fresh refers to the sum of the abun-dances of NMHC in factors 1ndash3 For summertime measure-ments the [NMHC]agedrefers to the abundances of NMHC

in factor 3 while the [NMHC]fresh refers to the sum of theabundances of NMHC in factors 1 and 2 Figure 6 showedthe dependence of [NMHC]aged [NMHC] fresh (circles in thefigure) onkOH values with the lines being the fitted resultsfrom Eq (4) The values ofD and [OH]1t were estimated tobe 043 and 299times 1010 molecule cmminus3 s in winter and 126and 105times 1011 molecule cmminus3 s in summer respectivelyReferring to the calculated OH radical concentrations in thesummer in Beijing (Z Liu et al 2012) and the differencesof measured OH concentrations in Tokyo between summerand winter (Kanaya et al 2007) we assumed that the aver-age daytime concentration of OH in Beijing was 15times 106

and 6times 106 molecule cmminus3 in winter and summer thus the

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3056 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 6 Relationships between the ratios of VOCs abundances inthe aged emissions factors with those in fresh emission factors andkOH values for each VOC species Each data point represents onecompound The lines are the fitted results from Eq (4)

photochemical ages of the aged factor in winter and summerwere estimated to be 55 and 49 h respectively

The [carbonyl]aged [carbonyl]fresh ratios (Fig 6 solidtriangles) were significantly higher than the ratios of[NMHC]aged [NMHC] fresh because secondary productionenhanced the abundances of carbonyls in aged factors Asshown in Eq (5) the abundances of carbonyls in aged fac-tors can be separated into two parts The first part standsfor aged primary emissions and the second part stands forsecondary formation calculated based on the consumption ofother VOCs[carbonyl

]aged= D times

[carbonyl

]freshtimes eminuskOHcarbonyltimes[OH]1t (5)

+D times

sum([VOC]freshtimes

(1minus eminuskOHVOCtimes[OH]1t

)times YVOCcarbonyl

)wherekOHcarbonyl is the OH rate constant for the carbonylYVOCcarbonyl (ppbppb) refers to the carbonyl productionyield from the oxidation of particular VOC which canbe estimated using the Leeds master chemical mechanism(MCM v32 httpmcmleedsacukMCM) The estimationof YVOCcarbonyl is based on the following assumptions theremoval of VOC is governed by the reaction with OH radi-cals and the removal of alkyl peroxyl radicals (RO2) is gov-erned by the reaction with NO TheYVOCcarbonylvalues usedin this study are listed in Table 2 The VOC species that makeno contributions to the production of our focused carbonylswere not shown in this study The values ofD and [OH]1t

were fitted from Eq (4)The carbonyl abundances in aged factors were calculated

by Eq (5) and agreed well with the PMF-resolved abun-dances (Fig 7) In winter all data points were distributednear the 1 1 line In summer though the data points showedsome variability but still agreed within a factor of two Theagreement between the calculated and PMF-resolved car-bonyl abundances indicated that our understanding on thephotochemical aging of PMF factors was acceptable Thedifferences between the calculated and PMF-resolved car-bonyl abundances may be due to one or some of the fol-lowing reasons (1) the PMF analysis andYVOCcarbonyl es-timations have their own uncertainties (2) Only reactions of

VOCs with OH radicals are considered in our analysis (3)Further reactions of secondary carbonyls are not consideredin our analysis (4) Some VOC species are not applied in ourPMF analysis but they might be precursors of carbonyls ThePMF-analyzed VOC species accounted for 90 and 87 oftotal concentrations of all measured VOCs species in winterand summer respectively However the contribution of eachVOC to the formation of carbonyls was not only dependenton its concentration but also its ability to produce carbonyls

In this work we identified the PMF factors in summerand winter as several individual fresh emissions and sev-eral mixed aged emissions with different degrees of photo-chemical processing Beijing was a large city with extensiveand continuous local emissions and therefore our measuredair mass was in fact a mixture of fresh and aged plumesThe fresh emissions were separated by the PMF model ac-cording to the correlations among VOC species and the dif-ferent emission characteristics of these sources Howeverthe resolved factors by the PMF model cannot represent theemissions at different aged stages from individual sourcesAs discussed in this section the respective photochemicalages of the aged factors in winter and summer were respec-tively about 55 and 49 h and the factor representing sec-ondary production and background levels in winter should bemore aged We conjectured that VOCs from different sourceshave been well mixed during such long aging process and itwas hard to identify the aged contributions from individualsources Likewise previous studies also used averaged emis-sion ratios to characterize the emission and aging process ofurban VOC emissions (de Gouw et al 2005 Warneke et al2007 Yuan et al 2012 Borbon et al 2013) In this sec-tion we proved the reasonability of the PMF-derived agedfactors However one problem remained when apportioningreactive VOCs using the PMF model Actually the photo-chemical reaction of VOCs in the atmosphere is a continuousprocess but the PMF model is not able to describe such kindof process In this work several different aged stages wererecognized by the PMF model and the PMF model used thelinear combination of these stages to describe the continuousphotochemical process This might be an important issue forthe PMF analysis and the influence of such approximationrequired future researches

34 Primary and secondary sources of carbonyls

The calculated relative contributions of carbonyl sources ap-portioned by the PMF model in Beijing were shown in Fig 8Previous studies usually attributed all carbonyls in aged fac-tors as secondary formation As discussed in Sect 33 theabundances of carbonyls in aged factors could be separatedinto two parts that is the aged direct emissions and theproduction from VOC consumption and the former partshould be treated as primary emissions In this study theaged direct emission part was further separated into eachfresh factor considering their relative contributions For all

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3049

2 Methods

21 VOCs measurements

Online VOCs measurements were conducted at an urban sitein Beijing from 4 August 2011 to 9 September 2011 and from29 December 2011 to 18 January 2012 The sampling sitewas on the top of a six-floor building (sim 20 m above ground)on Peking University campus The PKU campus is locatedin the northwest of Beijing about 500 m north of the FourthRing Road This site was described in detail elsewhere (Songet al 2008 Liu et al 2009)

A custom-built online gas chromatography systemequipped with a mass spectrometry and a flame ionizationdetector (GC-MSFID) was used to measure C2ndashC10 hydro-carbons C3ndashC6 carbonyls C1ndashC4 alkyl nitrates halocar-bons and methyl tert-butyl ether (MTBE) with a time reso-lution of 1 h CFC-113 (112-trichloro-122-trifluoroethane)was used as a natural internal standard from ambient air dueto its long lifetime and minimal manmade emissions (W-T Liu et al 2012) Daily calibrations were performed everyday and the variations of target species responses were re-quired to be withinplusmn20 from the calibration curve Thedetection limits for each species varied from 1 to 20 ppt andthe precisions ranged from 1 to 6 The detailed informa-tion of this system can be found in Yuan et al (2012)

A commercial high-sensitivity proton-transfer-reactionmass spectrometry (PTR-MS) (Ionicon Analytik InnsbruckAustria) was used to measure 28 masses including C1ndashC4carbonyls and C6ndashC9 aromatics with a time resolution ofabout 30 s Background signals were measured for 15 min ev-ery 25 h by diverting the ambient air sample flow througha Pt-coated quartz wool converter at 370C (Yuan et al2013) Calibrations were done every two or three days andthe response factors varied within 20 The detection lim-its of each species varied from 40 to 200 ppt except forformaldehyde As the proton affinity of formaldehyde is onlyslightly higher than water the back reaction of proton trans-fer becomes significant The kinetic of PTR-MS formalde-hyde detection was found to be mainly influenced by humid-ity (Vlasenko et al 2010 Warneke et al 2011) Thereforewe calibrated our formaldehyde measurement by using gasstandards provided by a permeation tube (Kin-Tek USA) atthe humidity from 0ndash30 mmol molminus1 to obtain the curve offormaldehyde response factor on humidity During the entireambient measurement formaldehyde signals were correctedaccording to the response curve Due to the different ambi-ent humidity between winter and summer the detection lim-its of formaldehyde were 022ndash034 ppb in winter and 045ndash080 ppb in summer at a time resolution of 30 s Among allour data only less than 02 of formaldehyde concentra-tions in summer fell below the detection limit

During these two campaigns C3ndashC4 carbonyls C6ndashC9aromatics isoprene and MVK+MACR (methyl vinyl ke-tone+ methacrolein) were measured simultaneously by on-

line GC-MS and PTR-MS Good agreements were found be-tween these two systems for most species with correlationcoefficients larger than 090 and slopes ranging from 08 to12 Formaldehyde and acetaldehyde were only measured byPTR-MS However inter-comparison between the PTR-MSand 24-dinitrophenyl hydrazinendashhigh-pressure liquid chro-matography (DNPHndashHPLC) methods (EPA TO-11A) wasperformed before the field measurements By correcting theinfluence of humidity on formaldehyde response a correla-tion coefficient of 093 and a slope of 106 were obtained be-tween measurements by PTR-MS and DNPHndashHPLC meth-ods The detailed results of inter-comparison were shown inthe Supplement

22 Positive matrix factorization (PMF)

The US Environmental Protection Agencyrsquos PMF 30 recep-tor model was used for VOCs source apportionment in thisstudy The PMF is a multivariate factor analysis tool whichassumes that measured concentrations at receptor sites arelinear combinations of contributions from different sources(Paatero and Tapper 1994) Generally an ambient data setcan be viewed as a data matrixx with i byj dimensions inwhich i samples andj chemical species are measured Thegoal of multivariate receptor modeling is to identify the num-berp of factors the species profilef of each source the massg contributed by each factor to each individual sample andthe residuale (Eq 1)

xij =

sump

k=1gikfkj + eij (1)

The PMF solution minimizes the object functionQ basedupon the uncertainties (Eq 2)

Q =

sumn

i=1

summ

j=1

[xij minus

sump

k=1gikfkjuij

]2

(2)

whereu represents the uncertainty of each data The theo-retical Q (Qtheoretical) can be calculated as Eq (3) and thebest PMF solution should haveQQtheoreticalwith the valueof sim 1

Qtheoretical= i times j minus p times (i + j) (3)

The basic assumption of the PMF model is that the fittingspecies are not allowed for chemical losses or productionThus apportioning carbonyl sources using the PMF modelshould be approached with care As mentioned in the in-troduction several studies apportioned carbonyl sources byusing the PMF model (Bon et al 2011 Buzcu Guven andOlaguer 2011 Zheng et al 2013) However the study toinvestigate the validity of the results was not available yetYuan et al (2012) proved the capacity of the PMF approachin identifying the role of chemical aging for better under-standing the PMF factors The VOC emission ratios derivedfrom the PMF fresh factors agreed well with the ones calcu-lated based on photochemical ages indicating that the PMF

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3050 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Table 1The sum of formaldehyde and acetaldehyde concentrationsin Beijing (unit ppb)

Year Summer Winter

2005 217a 70a

2006 1312b

2008 133cd 66e

2009 114e 59e

2010 126e

2011 80f 81f

a(Pang and Mu 2006)b(Shao et al 2009)c(Xu et al 2010)ddata before traffic restriction in 2008e(Zhang et al 2012)f this study

approach could identify the contributions from primary emis-sions reasonably Additionally the abundances of NMHCsin the PMF-aged factors could be reproduced by the photo-chemical aging of fresh factors In this study we run PMFmodels using two VOC data sets from different seasons anddiscussed the apportionment results to investigate whetherthe PMF approach can separate carbonyl sources well

For the PMF analysis uncertainties of online GC-MSFIDand PTR-MS measurements were calculated using themethod by Yuan et al (2012) Generally for the species mea-sured by the GC-MSFID system the uncertainties were cal-culated as the sum of 10 of measured concentrations andone-third of detection limits For the species measured bythe PTR-MS the uncertainties were calculated as the sum of10 of concentrations and the errors from mass spectrome-ter measurements Finally only species with high signal-to-noise ratios and few missing data values were used for thePMF analysis In winter 10 carbonyls 26 NMHCs 2-butylnitrate (2-BuONO2) tetrachloroethylene and MTBE wereused In summer most of the values of cistrans-2-butene andcistrans-2-pentene fell below the detection limits so these 4species were not used for the PMF analysis

In this work the PMF factor numbers were explored from3 to 8 for the best solution The final numbers of factors de-pended on two aspects (1)QQtheoretical which was usedto characterize the quality of the reconstruction and (2) thephysical plausibility of the factors (Bon et al 2011) Four-and five-factor resolutions were selected in summer and win-ter respectively The free rotation of the PMF factors wasinvestigated by adjusting the Fpeak parameter betweenminus5and 5 The results with non-zero Fpeak values were gen-erally consistent with the runs with zero Fpeak values andthus the results used in this study were from the runs withnon-rotation

As the PMF-resolved factors might be influenced by pho-tochemical processes we used the relationship between thecontribution of one factor to each NMHC species and itschemical reactivity (kOH) to distinguish fresh and aged fac-

tors as performed by Yuan et al (2012) Generally if allPMF-resolved factors were originated from fresh emissionsthe distribution of each species would not correlate with itschemical reactivity As the air mass from a source was age-ing the NMHCs underwent photochemical reactions and themore reactive species would be more largely consumed As aresult a negative correlation could be seen between an agedfactor contribution to each NMHC species and itskOH valueConsidering that such complex relationships were difficult toexpress in mathematical formulae Spearmanrsquos coefficient ofrank correlation (rs) was used to represent the relationship

3 Results and discussion

31 Characteristics of ambient carbonyls in Beijing

The average concentration of the 10 measured carbonyls inBeijing was 132plusmn 79 ppb (average valueplusmn standard devi-ation) in winter In summer the average concentration wentup to 163plusmn 74 ppb Formaldehyde acetone and acetalde-hyde were the three major species contributing more than80 of the total measured carbonyls in both seasons Am-bient concentrations of carbonyls in Beijing were measuredby several previous studies from 2005 (Pang and Mu 2006Shao et al 2009 Xu et al 2010 Zhang et al 2012) Dur-ing the recent years VOC sources changed remarkably inBeijing resulting from quick increment of vehicular pop-ulation stricter standards for vehicle emissions (Zhang etal 2012) and solvent compositions (Yuan et al 2010) anddramatic changes in industries Table 1 shows the sum ofmeasured formaldehyde and acetaldehyde concentrations inBeijing during 2005ndash2011 The measured concentrations insummer showed a significant decrement from 2005 to 2006and from 2010 to 2011 while the wintertime concentra-tions did not exhibit a significant change Some of the abovestudies provided daily average concentrations of formalde-hyde and acetaldehyde but some measurements were con-ducted only in daytime To check the effect of diurnal varia-tions we compared the daytime average concentration (from0800 to 2000 LT) and daily average concentration of thesetwo carbonyls in 2011 The result showed that the relativedifferences between the daytime and daily average concen-trations were very low with the values of 7 in summerand 1 in winter It seemed the daytime and daily aver-age concentrations of these carbonyls provided by previ-ous studies can be compared to investigate carbonyl trendsIn contrast with the decreasing trend of carbonyl levels de-rived from ground-based measurements satellite measure-ments suggested that formaldehyde columns showed an in-creasing trend (De Smedt et al 2010) As the satellites mea-surements were conducted in midday and covered a largerspatial scale the ground-based measurements were consid-ered to better represent the trend of 24 h carbonyl concentra-tions in the urban area of Beijing The decreasing trend for

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3051

Fig 1 Diurnal variations of measured ratios of formaldehyde ethene acetaldehyde propene and acetone ethane at the PKU site duringwinter and summer Black lines represent the average values and grey shaded areas indicate standard deviations

summed concentrations of formaldehyde and acetaldehydeindicated that the emission controls on these carbonyls ortheir precursors in Beijing possibly made progress during thepast years However the flat trend during winter suggestedthat there might be some important sources for carbonyls inwinter were not under effective control

Figure 1 shows the diurnal variations of ratios of formalde-hyde ethene acetaldehyde propene and acetone ethane atthe PKU site during winter and summer Ambient concentra-tions of carbonyls were influenced by meteorological con-ditions and the emission strengths of carbonyls and theirprecursors Assuming that the direct emissions of carbonylsfrom anthropogenic sources were proportional to the emis-sions of NMHCs the diurnal variations of carbonyl NMHCratios could reflect the importance of secondary formationto ambient carbonyls These three pairs of carbonyls andNMHCs were chosen because the two species in each pairhad similar reaction rates with OH radicals (Atkinson et al2006) As a result the two species in the atmosphere wereremoved approximately equally by reactions with OH radi-cals and therefore the effect of photochemical degradationon their ratios can be neglected As shown in Fig 1 themeasured carbonyl NMHC ratios showed similar daily vari-ations both in summer and winter they were stable dur-ing nighttime increased after sunrise (about 0800 LT) andreached a maximum in the early afternoon (about 1300ndash1500 LT) then decreased to low values at night This varia-tion indicated an important contribution from secondary pro-duction during the daytime both in winter and summer Thediurnal variations of the acetone ethane ratios were less thanthose for formaldehyde ethene and acetaldehyde propeneSuch difference was consistent with previous studies whichfound that secondary production of acetone was less impor-tant than that of formaldehyde and acetaldehyde (de Gouw etal 2005 Yuan et al 2012)

Though the diurnal patterns of carbonyl NMHC ratioswere similar in winter and summer the values of car-bonyl NMHC ratios were approximately 3ndash5 times higher in

summer than those in winter during both day and night Asmentioned above ambient concentrations of carbonyls didnot show significant discrepancies between these two sea-sons The seasonal differences of carbonyl NMHC ratioswere mainly due to the much higher NMHC levels in win-ter This also indicated the significant differences in VOCsources between these two seasons Considering the effectsof photochemical processing on VOC ratios were minor atnight nighttime VOC ratios were close to their emission ra-tios (Wang et al 2013) Therefore the higher nighttime car-bonyl NMHC ratios in summer indicated that the fractionof NMHCs in primary emitted VOCs was higher in winterthan that in summer whereas carbonyls fraction was higherin summer

Previous studies usually supposed that carbonyls in win-ter mainly came from direct anthropogenic emissions (Pangand Mu 2006 Ceroacuten et al 2007) From our measure-ments the differences of carbonyl NMHC ratios betweennoon and night in summer were only slightly larger thanthe differences in winter The noon night ratios were 26and 31 for formaldehyde ethene 26 and 35 for acetalde-hyde propene and 14 and 15 for acetone ethane in winterand summer respectively This indicated that secondary for-mation was still an important source of carbonyls in winterIt should be noted that photolysis removal of carbonyls wasalso considered here besides the removal by reactions withOH radicals According to our estimation reaction with OHwas the main pathway of carbonyl removal The photolysiscontributed 33 1 and 12 to the losses of formalde-hyde acetaldehyde and acetone at daytime and these contri-butions showed no significant difference between winter andsummer The details for the calculation of carbonyl removalrates were shown in the Supplement As a result neglect-ing the photolysis of carbonyls had little effect on comparingthe contributions of secondary formation between winter andsummer and the variation of carbonyl NMHC ratios couldreflect the importance of secondary formation For furtherdiscussion the PMF model was used for carbonyls source

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3052 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 2Profiles of the four factors resolved by the PMF model and the distributions of species among these factors in summer

apportionment to determine relative contributions from pri-mary and secondary sources

32 Identifying PMF factors

The PMF model is a mathematical tool which separates pol-lutants according to the correlations among species and theunderstanding of PMF factors is subjective to a certain ex-tent We were aware of the deficiency of the PMF modelwhen apportioning VOC sources In this section we com-pared the profiles derived by the PMF model with previouslymeasured ones to check the reliability of resolved fresh fac-tors In the next section we tried to identify aged factorsby investigating the relationship between degradation of pri-mary emitted VOCs and production of carbonyls

For summertime measurements 593 samples were usedfor the PMF analysis and four factors were identified Theprofiles of resolved PMF factors and the distributions ofspecies among the factors were shown in Fig 2 The profileswere expressed as the mass percentage of individual VOCspecies in each factor and the distributions represented thecontributions of individual factors to the measured level ofeach species

Factor 1 made an important contribution to our measuredalkanes and alkenes The profile of this factor was dominatedby C2ndashC5 alkanes C2ndashC3 alkenes ethyne and aromaticsThese species were mainly emitted from exhaust of vehiclesand evaporation of gasoline (Liu et al 2008) This factor ex-plained 381 of the measured MTBE which was com-monly used as a tracer of gasoline vehicle emissions (Changet al 2006) Previous studies also reported that traffic-relatedemissions were the most important NMHCs sources in sum-mer of Beijing (Song et al 2008 Wang et al 2010) Wetherefore attributed this factor to traffic-related emissions

In factor 2 the loadings of aromatics including benzenetoluene ethylbenzene and xylenes were high Aromatics

were reported to be major constituents of solvents (Yuan etal 2010) which were widely used in industrial processes aswell as some household products Thus we considered thisfactor to be related to industry and solvent usage

Factor 3 had contributions to almost all measured VOCspecies except for some very reactive species This factor ac-counted for more than 60 of our measured carbonyls and875 of our measured 2-BuONO2 levels Alkyl nitrateswere believed to be mainly from secondary production in ur-ban regions and had relatively long lifetimes (Simpson et al2003) Thus we identified this factor as aged emissions

Factor 4 contributed 815 of measured isoprene whichwas often considered as a tracer for biogenic emissions (Situet al 2013) so we attributed this factor to biogenic emis-sions High loadings of formaldehyde acetaldehyde andacetone were also found in this factor This is possibly be-cause these carbonyls could also be emitted from biogenicsources (Winters et al 2009) and they were important ox-idation products of other biogenic VOCs (Carter and Atkin-son 1996)

Based on the derived profiles of each factor and the dis-tributions of species among these factors we identified thattwo of the four PMF-resolved factors could represent pri-mary emission sources another related to biogenic emis-sions and the other was an aged emission factor As dis-cussed in Sect 22 the relationship between the contributionof one factor to each NMHC species and its chemical reactiv-ity could be used to distinguish factors related to photochem-ically aged emissions from those for fresh emissions Such arelationship was shown in Fig 3 where each NMHC specieswas shown as a circle Positive correlations were identifiedfor factors 1 and 2 and thus these two factors were consid-ered to be related to fresh emissions However a significantnegative correlation was identified for factor 3 indicatingthat this factor represented photochemically aged emissions

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3053

Fig 3 Relationships between the factor contributions to each NMHC species andkOH values for NMHC species in summer Each circlerepresents one compound Carbonyls are shown as solid triangles in the graph but they were not used for correlation analysis due to theirsecondary productionlowastindicates a significant correlation at a confidence level of 005lowastlowastindicates a significant correlation at a confidencelevel of 001

The distributions of carbonyls in PMF-resolved factorswere shown in Fig 3 by solid triangles If all factors were re-lated to fresh emissions the distributions of carbonyls wouldbe similar to the distributions of NMHCs in each factor Dueto the influence from photochemical aging the distribution ofcarbonyls could be higher in aged factors owing to their sec-ondary production and meanwhile the distributions of car-bonyls would be lower in fresh factors In factors 1 and 2 thedistributions of carbonyls were similar or lower than those ofNMHCs whereas the distributions of carbonyls in factor 3were higher than those of NMHCs Such appearance agreedwith our above identifications on PMF-resolved factors

In this study two fresh factors and an aged factor wereidentified in the summer of 2011 However Yuan etal (2012) identified one mixed fresh factor and two aged fac-tors with different photochemical ages using a similar anal-ysis approach based on measurements at the same site in thesummer of 2010 It was interesting to get such different re-sults in two consecutive years Yuan et al (2012) concludedthat PMF results depended on the degree of photochemicalprocessing and the differences in emission compositions ofvarious sources Assuming that the differences in VOC emis-sions were not significant between 2010 and 2011 the PMFresults were mainly influenced by the difference in degrees

in photochemical aging between these two years We usedthe ratio of o-xylene to ethylbenzene as an indicator of thedegree of photochemical processing As seen in Fig S4 therelative standard deviations of o-xylene ethylbenzene ratiosin 2010 were 30 larger than those in 2011 It indicated thatthe range of photochemical processing degrees was larger in2010 As a result the PMF factors in 2010 were extractedmainly according to different degrees of photochemical pro-cessing whereas the PMF factors in 2011 could be extractedmainly based on individual sources

In winter 341 samples were used for the PMF analysisand five factors were identified (Fig 4)

Factor 1 and factor 3 contributed to most of our measuredalkanes and alkenes but there were some differences be-tween these two factors Factor 1 had higher contributions toC2ndashC3 NMHCs while factor 3 had higher contributions toC4ndashC5 NMHCs These light hydrocarbons were mainly gen-erated by incomplete combustion processes such as vehicleexhaust and coal burning (Liu et al 2008 Wang et al 2013)The measured source profiles in China showed that C2ndashC3NMHCs contributed more than half of NMHCs that emit-ted from coal burning (Liu et al 2008 Wang et al 2013)Benzene and toluene also made important contributions tofactor 1 These two species were also found to be important

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3054 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 4Profiles of the five factors resolved by the PMF model and the distributions of species among these factors in winter

components in NMHCs emissions from coal burning (Liu etal 2008) The ratio of benzene to toluene (231 ppb ppbminus1)

in this factor fell between the ratios measured by Liu etal (2008) for residential coal burning (181 ppb ppbminus1) andindustrial coal burning (262 ppb ppbminus1) but much higherthan that measured in the tunnel experiment (070 ppb ppbminus1)

(Liu et al 2008) therefore factor 1 was identified as coalburning C4ndashC5 NMHCs were important species from trafficrelated emissions and factor 3 showed similar characteris-tic with factor 1 in summer In addition factor 3 explained379 of the measured MTBE Therefore factor 3 was at-tributed to traffic related emissions

Factor 2 had high contributions to aromatics and showedsimilar characteristic with the resolved factor 2 for sum-mertime measurements Besides aromatics this factor con-tributed about 30 of measured long-chain alkanes (C7ndashC10 alkanes) The long-chain alkanes were also consid-ered to be important components of printing VOC emissions(Yuan et al 2010) Thus this factor was identified as indus-try and solvent usage

Both factor 4 and factor 5 had important contributionsto carbonyls but little contribution to reactive NMHCs andtherefore these two factors were inferred to possibly not rep-resent fresh emissions The major NMHC species in factor 4were C2-C5 alkanes ethene ethyne benzene and tolueneThese species could be emitted from all of the three primaryfactors discussed above and thus we considered this factor tobe aged emissions from these anthropogenic sources TheNMHCs in factor 5 were dominated by unreactive alkanessuch as ethane and propane so factor 5 was at a more aged

stage than factor 4 Factor 5 accounted for 677 of the mea-sured 2-BuONO2 levels and 197 of the measured carbonyllevels According to the abundances of alkyl nitrates and un-reactive alkanes we considered that the loadings of VOCspecies in this factor were related to secondary productionand background levels

Similarly with summertime measurements the relation-ships between the contribution of one factor to each NMHCspecies andkOH values for each NMHC species were ana-lyzed As shown in Fig 5 no significant correlation was ob-served for factor 1 and factor 2 while a positive correlationwas found for factor 3 and thus these three factors were con-sidered to be fresh emissions Significant negative correla-tions were identified for factor 4 and factor 5 indicating thatthese two factors were influenced by photochemical aging

33 Exploring the aged and fresh emission factors

In the last section we distinguished photochemically agedfactors from fresh factors To better understand the relation-ships between aged and fresh factors and the role of photo-chemical aging in PMF analysis we explored the relation-ships between VOC consumption and carbonyl formation inthe aged factors

For two factors representing different photochemicalstages of the same emissions the ratios of NMHC abun-dances in these two factors should follow the equation below(Yuan et al 2012)

[NMHC]aged

[NMHC]fresh= D times eminuskOHNMHC[OH]1t (4)

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3055

Fig 5 Relationships between the factor contributions to each NMHC species andkOH values of NMHC species in winter Each circlerepresents one compound Carbonyls are shown as solid triangles in the graph but they were not used for correlation analysis due to theirsecondary productionlowastlowastindicates a significant correlation at a confidence level of 001

where [NMHC]agedand [NMHC]fresh(ppb) refer to the abun-dances of NMHC species in the aged and fresh factor respec-tively D is a scaling factor which normalizes the NMHCabundances to unitykOHNMHC is the OH rate constant forthe NMHC [OH] is the average concentration of OH rad-ical and1t is the difference of reaction time between theaged and fresh factors In this study the OH exposure (iethe time integrate of [OH]) expressed by [OH]1t is calcu-lated as a whole

As discussed in Sect 32 for wintertime measurementsthe [NMHC]agedrefers to the abundances of NMHC in fac-tor 4 while the [NMHC]fresh refers to the sum of the abun-dances of NMHC in factors 1ndash3 For summertime measure-ments the [NMHC]agedrefers to the abundances of NMHC

in factor 3 while the [NMHC]fresh refers to the sum of theabundances of NMHC in factors 1 and 2 Figure 6 showedthe dependence of [NMHC]aged [NMHC] fresh (circles in thefigure) onkOH values with the lines being the fitted resultsfrom Eq (4) The values ofD and [OH]1t were estimated tobe 043 and 299times 1010 molecule cmminus3 s in winter and 126and 105times 1011 molecule cmminus3 s in summer respectivelyReferring to the calculated OH radical concentrations in thesummer in Beijing (Z Liu et al 2012) and the differencesof measured OH concentrations in Tokyo between summerand winter (Kanaya et al 2007) we assumed that the aver-age daytime concentration of OH in Beijing was 15times 106

and 6times 106 molecule cmminus3 in winter and summer thus the

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3056 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 6 Relationships between the ratios of VOCs abundances inthe aged emissions factors with those in fresh emission factors andkOH values for each VOC species Each data point represents onecompound The lines are the fitted results from Eq (4)

photochemical ages of the aged factor in winter and summerwere estimated to be 55 and 49 h respectively

The [carbonyl]aged [carbonyl]fresh ratios (Fig 6 solidtriangles) were significantly higher than the ratios of[NMHC]aged [NMHC] fresh because secondary productionenhanced the abundances of carbonyls in aged factors Asshown in Eq (5) the abundances of carbonyls in aged fac-tors can be separated into two parts The first part standsfor aged primary emissions and the second part stands forsecondary formation calculated based on the consumption ofother VOCs[carbonyl

]aged= D times

[carbonyl

]freshtimes eminuskOHcarbonyltimes[OH]1t (5)

+D times

sum([VOC]freshtimes

(1minus eminuskOHVOCtimes[OH]1t

)times YVOCcarbonyl

)wherekOHcarbonyl is the OH rate constant for the carbonylYVOCcarbonyl (ppbppb) refers to the carbonyl productionyield from the oxidation of particular VOC which canbe estimated using the Leeds master chemical mechanism(MCM v32 httpmcmleedsacukMCM) The estimationof YVOCcarbonyl is based on the following assumptions theremoval of VOC is governed by the reaction with OH radi-cals and the removal of alkyl peroxyl radicals (RO2) is gov-erned by the reaction with NO TheYVOCcarbonylvalues usedin this study are listed in Table 2 The VOC species that makeno contributions to the production of our focused carbonylswere not shown in this study The values ofD and [OH]1t

were fitted from Eq (4)The carbonyl abundances in aged factors were calculated

by Eq (5) and agreed well with the PMF-resolved abun-dances (Fig 7) In winter all data points were distributednear the 1 1 line In summer though the data points showedsome variability but still agreed within a factor of two Theagreement between the calculated and PMF-resolved car-bonyl abundances indicated that our understanding on thephotochemical aging of PMF factors was acceptable Thedifferences between the calculated and PMF-resolved car-bonyl abundances may be due to one or some of the fol-lowing reasons (1) the PMF analysis andYVOCcarbonyl es-timations have their own uncertainties (2) Only reactions of

VOCs with OH radicals are considered in our analysis (3)Further reactions of secondary carbonyls are not consideredin our analysis (4) Some VOC species are not applied in ourPMF analysis but they might be precursors of carbonyls ThePMF-analyzed VOC species accounted for 90 and 87 oftotal concentrations of all measured VOCs species in winterand summer respectively However the contribution of eachVOC to the formation of carbonyls was not only dependenton its concentration but also its ability to produce carbonyls

In this work we identified the PMF factors in summerand winter as several individual fresh emissions and sev-eral mixed aged emissions with different degrees of photo-chemical processing Beijing was a large city with extensiveand continuous local emissions and therefore our measuredair mass was in fact a mixture of fresh and aged plumesThe fresh emissions were separated by the PMF model ac-cording to the correlations among VOC species and the dif-ferent emission characteristics of these sources Howeverthe resolved factors by the PMF model cannot represent theemissions at different aged stages from individual sourcesAs discussed in this section the respective photochemicalages of the aged factors in winter and summer were respec-tively about 55 and 49 h and the factor representing sec-ondary production and background levels in winter should bemore aged We conjectured that VOCs from different sourceshave been well mixed during such long aging process and itwas hard to identify the aged contributions from individualsources Likewise previous studies also used averaged emis-sion ratios to characterize the emission and aging process ofurban VOC emissions (de Gouw et al 2005 Warneke et al2007 Yuan et al 2012 Borbon et al 2013) In this sec-tion we proved the reasonability of the PMF-derived agedfactors However one problem remained when apportioningreactive VOCs using the PMF model Actually the photo-chemical reaction of VOCs in the atmosphere is a continuousprocess but the PMF model is not able to describe such kindof process In this work several different aged stages wererecognized by the PMF model and the PMF model used thelinear combination of these stages to describe the continuousphotochemical process This might be an important issue forthe PMF analysis and the influence of such approximationrequired future researches

34 Primary and secondary sources of carbonyls

The calculated relative contributions of carbonyl sources ap-portioned by the PMF model in Beijing were shown in Fig 8Previous studies usually attributed all carbonyls in aged fac-tors as secondary formation As discussed in Sect 33 theabundances of carbonyls in aged factors could be separatedinto two parts that is the aged direct emissions and theproduction from VOC consumption and the former partshould be treated as primary emissions In this study theaged direct emission part was further separated into eachfresh factor considering their relative contributions For all

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

3050 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Table 1The sum of formaldehyde and acetaldehyde concentrationsin Beijing (unit ppb)

Year Summer Winter

2005 217a 70a

2006 1312b

2008 133cd 66e

2009 114e 59e

2010 126e

2011 80f 81f

a(Pang and Mu 2006)b(Shao et al 2009)c(Xu et al 2010)ddata before traffic restriction in 2008e(Zhang et al 2012)f this study

approach could identify the contributions from primary emis-sions reasonably Additionally the abundances of NMHCsin the PMF-aged factors could be reproduced by the photo-chemical aging of fresh factors In this study we run PMFmodels using two VOC data sets from different seasons anddiscussed the apportionment results to investigate whetherthe PMF approach can separate carbonyl sources well

For the PMF analysis uncertainties of online GC-MSFIDand PTR-MS measurements were calculated using themethod by Yuan et al (2012) Generally for the species mea-sured by the GC-MSFID system the uncertainties were cal-culated as the sum of 10 of measured concentrations andone-third of detection limits For the species measured bythe PTR-MS the uncertainties were calculated as the sum of10 of concentrations and the errors from mass spectrome-ter measurements Finally only species with high signal-to-noise ratios and few missing data values were used for thePMF analysis In winter 10 carbonyls 26 NMHCs 2-butylnitrate (2-BuONO2) tetrachloroethylene and MTBE wereused In summer most of the values of cistrans-2-butene andcistrans-2-pentene fell below the detection limits so these 4species were not used for the PMF analysis

In this work the PMF factor numbers were explored from3 to 8 for the best solution The final numbers of factors de-pended on two aspects (1)QQtheoretical which was usedto characterize the quality of the reconstruction and (2) thephysical plausibility of the factors (Bon et al 2011) Four-and five-factor resolutions were selected in summer and win-ter respectively The free rotation of the PMF factors wasinvestigated by adjusting the Fpeak parameter betweenminus5and 5 The results with non-zero Fpeak values were gen-erally consistent with the runs with zero Fpeak values andthus the results used in this study were from the runs withnon-rotation

As the PMF-resolved factors might be influenced by pho-tochemical processes we used the relationship between thecontribution of one factor to each NMHC species and itschemical reactivity (kOH) to distinguish fresh and aged fac-

tors as performed by Yuan et al (2012) Generally if allPMF-resolved factors were originated from fresh emissionsthe distribution of each species would not correlate with itschemical reactivity As the air mass from a source was age-ing the NMHCs underwent photochemical reactions and themore reactive species would be more largely consumed As aresult a negative correlation could be seen between an agedfactor contribution to each NMHC species and itskOH valueConsidering that such complex relationships were difficult toexpress in mathematical formulae Spearmanrsquos coefficient ofrank correlation (rs) was used to represent the relationship

3 Results and discussion

31 Characteristics of ambient carbonyls in Beijing

The average concentration of the 10 measured carbonyls inBeijing was 132plusmn 79 ppb (average valueplusmn standard devi-ation) in winter In summer the average concentration wentup to 163plusmn 74 ppb Formaldehyde acetone and acetalde-hyde were the three major species contributing more than80 of the total measured carbonyls in both seasons Am-bient concentrations of carbonyls in Beijing were measuredby several previous studies from 2005 (Pang and Mu 2006Shao et al 2009 Xu et al 2010 Zhang et al 2012) Dur-ing the recent years VOC sources changed remarkably inBeijing resulting from quick increment of vehicular pop-ulation stricter standards for vehicle emissions (Zhang etal 2012) and solvent compositions (Yuan et al 2010) anddramatic changes in industries Table 1 shows the sum ofmeasured formaldehyde and acetaldehyde concentrations inBeijing during 2005ndash2011 The measured concentrations insummer showed a significant decrement from 2005 to 2006and from 2010 to 2011 while the wintertime concentra-tions did not exhibit a significant change Some of the abovestudies provided daily average concentrations of formalde-hyde and acetaldehyde but some measurements were con-ducted only in daytime To check the effect of diurnal varia-tions we compared the daytime average concentration (from0800 to 2000 LT) and daily average concentration of thesetwo carbonyls in 2011 The result showed that the relativedifferences between the daytime and daily average concen-trations were very low with the values of 7 in summerand 1 in winter It seemed the daytime and daily aver-age concentrations of these carbonyls provided by previ-ous studies can be compared to investigate carbonyl trendsIn contrast with the decreasing trend of carbonyl levels de-rived from ground-based measurements satellite measure-ments suggested that formaldehyde columns showed an in-creasing trend (De Smedt et al 2010) As the satellites mea-surements were conducted in midday and covered a largerspatial scale the ground-based measurements were consid-ered to better represent the trend of 24 h carbonyl concentra-tions in the urban area of Beijing The decreasing trend for

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3051

Fig 1 Diurnal variations of measured ratios of formaldehyde ethene acetaldehyde propene and acetone ethane at the PKU site duringwinter and summer Black lines represent the average values and grey shaded areas indicate standard deviations

summed concentrations of formaldehyde and acetaldehydeindicated that the emission controls on these carbonyls ortheir precursors in Beijing possibly made progress during thepast years However the flat trend during winter suggestedthat there might be some important sources for carbonyls inwinter were not under effective control

Figure 1 shows the diurnal variations of ratios of formalde-hyde ethene acetaldehyde propene and acetone ethane atthe PKU site during winter and summer Ambient concentra-tions of carbonyls were influenced by meteorological con-ditions and the emission strengths of carbonyls and theirprecursors Assuming that the direct emissions of carbonylsfrom anthropogenic sources were proportional to the emis-sions of NMHCs the diurnal variations of carbonyl NMHCratios could reflect the importance of secondary formationto ambient carbonyls These three pairs of carbonyls andNMHCs were chosen because the two species in each pairhad similar reaction rates with OH radicals (Atkinson et al2006) As a result the two species in the atmosphere wereremoved approximately equally by reactions with OH radi-cals and therefore the effect of photochemical degradationon their ratios can be neglected As shown in Fig 1 themeasured carbonyl NMHC ratios showed similar daily vari-ations both in summer and winter they were stable dur-ing nighttime increased after sunrise (about 0800 LT) andreached a maximum in the early afternoon (about 1300ndash1500 LT) then decreased to low values at night This varia-tion indicated an important contribution from secondary pro-duction during the daytime both in winter and summer Thediurnal variations of the acetone ethane ratios were less thanthose for formaldehyde ethene and acetaldehyde propeneSuch difference was consistent with previous studies whichfound that secondary production of acetone was less impor-tant than that of formaldehyde and acetaldehyde (de Gouw etal 2005 Yuan et al 2012)

Though the diurnal patterns of carbonyl NMHC ratioswere similar in winter and summer the values of car-bonyl NMHC ratios were approximately 3ndash5 times higher in

summer than those in winter during both day and night Asmentioned above ambient concentrations of carbonyls didnot show significant discrepancies between these two sea-sons The seasonal differences of carbonyl NMHC ratioswere mainly due to the much higher NMHC levels in win-ter This also indicated the significant differences in VOCsources between these two seasons Considering the effectsof photochemical processing on VOC ratios were minor atnight nighttime VOC ratios were close to their emission ra-tios (Wang et al 2013) Therefore the higher nighttime car-bonyl NMHC ratios in summer indicated that the fractionof NMHCs in primary emitted VOCs was higher in winterthan that in summer whereas carbonyls fraction was higherin summer

Previous studies usually supposed that carbonyls in win-ter mainly came from direct anthropogenic emissions (Pangand Mu 2006 Ceroacuten et al 2007) From our measure-ments the differences of carbonyl NMHC ratios betweennoon and night in summer were only slightly larger thanthe differences in winter The noon night ratios were 26and 31 for formaldehyde ethene 26 and 35 for acetalde-hyde propene and 14 and 15 for acetone ethane in winterand summer respectively This indicated that secondary for-mation was still an important source of carbonyls in winterIt should be noted that photolysis removal of carbonyls wasalso considered here besides the removal by reactions withOH radicals According to our estimation reaction with OHwas the main pathway of carbonyl removal The photolysiscontributed 33 1 and 12 to the losses of formalde-hyde acetaldehyde and acetone at daytime and these contri-butions showed no significant difference between winter andsummer The details for the calculation of carbonyl removalrates were shown in the Supplement As a result neglect-ing the photolysis of carbonyls had little effect on comparingthe contributions of secondary formation between winter andsummer and the variation of carbonyl NMHC ratios couldreflect the importance of secondary formation For furtherdiscussion the PMF model was used for carbonyls source

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3052 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 2Profiles of the four factors resolved by the PMF model and the distributions of species among these factors in summer

apportionment to determine relative contributions from pri-mary and secondary sources

32 Identifying PMF factors

The PMF model is a mathematical tool which separates pol-lutants according to the correlations among species and theunderstanding of PMF factors is subjective to a certain ex-tent We were aware of the deficiency of the PMF modelwhen apportioning VOC sources In this section we com-pared the profiles derived by the PMF model with previouslymeasured ones to check the reliability of resolved fresh fac-tors In the next section we tried to identify aged factorsby investigating the relationship between degradation of pri-mary emitted VOCs and production of carbonyls

For summertime measurements 593 samples were usedfor the PMF analysis and four factors were identified Theprofiles of resolved PMF factors and the distributions ofspecies among the factors were shown in Fig 2 The profileswere expressed as the mass percentage of individual VOCspecies in each factor and the distributions represented thecontributions of individual factors to the measured level ofeach species

Factor 1 made an important contribution to our measuredalkanes and alkenes The profile of this factor was dominatedby C2ndashC5 alkanes C2ndashC3 alkenes ethyne and aromaticsThese species were mainly emitted from exhaust of vehiclesand evaporation of gasoline (Liu et al 2008) This factor ex-plained 381 of the measured MTBE which was com-monly used as a tracer of gasoline vehicle emissions (Changet al 2006) Previous studies also reported that traffic-relatedemissions were the most important NMHCs sources in sum-mer of Beijing (Song et al 2008 Wang et al 2010) Wetherefore attributed this factor to traffic-related emissions

In factor 2 the loadings of aromatics including benzenetoluene ethylbenzene and xylenes were high Aromatics

were reported to be major constituents of solvents (Yuan etal 2010) which were widely used in industrial processes aswell as some household products Thus we considered thisfactor to be related to industry and solvent usage

Factor 3 had contributions to almost all measured VOCspecies except for some very reactive species This factor ac-counted for more than 60 of our measured carbonyls and875 of our measured 2-BuONO2 levels Alkyl nitrateswere believed to be mainly from secondary production in ur-ban regions and had relatively long lifetimes (Simpson et al2003) Thus we identified this factor as aged emissions

Factor 4 contributed 815 of measured isoprene whichwas often considered as a tracer for biogenic emissions (Situet al 2013) so we attributed this factor to biogenic emis-sions High loadings of formaldehyde acetaldehyde andacetone were also found in this factor This is possibly be-cause these carbonyls could also be emitted from biogenicsources (Winters et al 2009) and they were important ox-idation products of other biogenic VOCs (Carter and Atkin-son 1996)

Based on the derived profiles of each factor and the dis-tributions of species among these factors we identified thattwo of the four PMF-resolved factors could represent pri-mary emission sources another related to biogenic emis-sions and the other was an aged emission factor As dis-cussed in Sect 22 the relationship between the contributionof one factor to each NMHC species and its chemical reactiv-ity could be used to distinguish factors related to photochem-ically aged emissions from those for fresh emissions Such arelationship was shown in Fig 3 where each NMHC specieswas shown as a circle Positive correlations were identifiedfor factors 1 and 2 and thus these two factors were consid-ered to be related to fresh emissions However a significantnegative correlation was identified for factor 3 indicatingthat this factor represented photochemically aged emissions

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3053

Fig 3 Relationships between the factor contributions to each NMHC species andkOH values for NMHC species in summer Each circlerepresents one compound Carbonyls are shown as solid triangles in the graph but they were not used for correlation analysis due to theirsecondary productionlowastindicates a significant correlation at a confidence level of 005lowastlowastindicates a significant correlation at a confidencelevel of 001

The distributions of carbonyls in PMF-resolved factorswere shown in Fig 3 by solid triangles If all factors were re-lated to fresh emissions the distributions of carbonyls wouldbe similar to the distributions of NMHCs in each factor Dueto the influence from photochemical aging the distribution ofcarbonyls could be higher in aged factors owing to their sec-ondary production and meanwhile the distributions of car-bonyls would be lower in fresh factors In factors 1 and 2 thedistributions of carbonyls were similar or lower than those ofNMHCs whereas the distributions of carbonyls in factor 3were higher than those of NMHCs Such appearance agreedwith our above identifications on PMF-resolved factors

In this study two fresh factors and an aged factor wereidentified in the summer of 2011 However Yuan etal (2012) identified one mixed fresh factor and two aged fac-tors with different photochemical ages using a similar anal-ysis approach based on measurements at the same site in thesummer of 2010 It was interesting to get such different re-sults in two consecutive years Yuan et al (2012) concludedthat PMF results depended on the degree of photochemicalprocessing and the differences in emission compositions ofvarious sources Assuming that the differences in VOC emis-sions were not significant between 2010 and 2011 the PMFresults were mainly influenced by the difference in degrees

in photochemical aging between these two years We usedthe ratio of o-xylene to ethylbenzene as an indicator of thedegree of photochemical processing As seen in Fig S4 therelative standard deviations of o-xylene ethylbenzene ratiosin 2010 were 30 larger than those in 2011 It indicated thatthe range of photochemical processing degrees was larger in2010 As a result the PMF factors in 2010 were extractedmainly according to different degrees of photochemical pro-cessing whereas the PMF factors in 2011 could be extractedmainly based on individual sources

In winter 341 samples were used for the PMF analysisand five factors were identified (Fig 4)

Factor 1 and factor 3 contributed to most of our measuredalkanes and alkenes but there were some differences be-tween these two factors Factor 1 had higher contributions toC2ndashC3 NMHCs while factor 3 had higher contributions toC4ndashC5 NMHCs These light hydrocarbons were mainly gen-erated by incomplete combustion processes such as vehicleexhaust and coal burning (Liu et al 2008 Wang et al 2013)The measured source profiles in China showed that C2ndashC3NMHCs contributed more than half of NMHCs that emit-ted from coal burning (Liu et al 2008 Wang et al 2013)Benzene and toluene also made important contributions tofactor 1 These two species were also found to be important

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3054 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 4Profiles of the five factors resolved by the PMF model and the distributions of species among these factors in winter

components in NMHCs emissions from coal burning (Liu etal 2008) The ratio of benzene to toluene (231 ppb ppbminus1)

in this factor fell between the ratios measured by Liu etal (2008) for residential coal burning (181 ppb ppbminus1) andindustrial coal burning (262 ppb ppbminus1) but much higherthan that measured in the tunnel experiment (070 ppb ppbminus1)

(Liu et al 2008) therefore factor 1 was identified as coalburning C4ndashC5 NMHCs were important species from trafficrelated emissions and factor 3 showed similar characteris-tic with factor 1 in summer In addition factor 3 explained379 of the measured MTBE Therefore factor 3 was at-tributed to traffic related emissions

Factor 2 had high contributions to aromatics and showedsimilar characteristic with the resolved factor 2 for sum-mertime measurements Besides aromatics this factor con-tributed about 30 of measured long-chain alkanes (C7ndashC10 alkanes) The long-chain alkanes were also consid-ered to be important components of printing VOC emissions(Yuan et al 2010) Thus this factor was identified as indus-try and solvent usage

Both factor 4 and factor 5 had important contributionsto carbonyls but little contribution to reactive NMHCs andtherefore these two factors were inferred to possibly not rep-resent fresh emissions The major NMHC species in factor 4were C2-C5 alkanes ethene ethyne benzene and tolueneThese species could be emitted from all of the three primaryfactors discussed above and thus we considered this factor tobe aged emissions from these anthropogenic sources TheNMHCs in factor 5 were dominated by unreactive alkanessuch as ethane and propane so factor 5 was at a more aged

stage than factor 4 Factor 5 accounted for 677 of the mea-sured 2-BuONO2 levels and 197 of the measured carbonyllevels According to the abundances of alkyl nitrates and un-reactive alkanes we considered that the loadings of VOCspecies in this factor were related to secondary productionand background levels

Similarly with summertime measurements the relation-ships between the contribution of one factor to each NMHCspecies andkOH values for each NMHC species were ana-lyzed As shown in Fig 5 no significant correlation was ob-served for factor 1 and factor 2 while a positive correlationwas found for factor 3 and thus these three factors were con-sidered to be fresh emissions Significant negative correla-tions were identified for factor 4 and factor 5 indicating thatthese two factors were influenced by photochemical aging

33 Exploring the aged and fresh emission factors

In the last section we distinguished photochemically agedfactors from fresh factors To better understand the relation-ships between aged and fresh factors and the role of photo-chemical aging in PMF analysis we explored the relation-ships between VOC consumption and carbonyl formation inthe aged factors

For two factors representing different photochemicalstages of the same emissions the ratios of NMHC abun-dances in these two factors should follow the equation below(Yuan et al 2012)

[NMHC]aged

[NMHC]fresh= D times eminuskOHNMHC[OH]1t (4)

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3055

Fig 5 Relationships between the factor contributions to each NMHC species andkOH values of NMHC species in winter Each circlerepresents one compound Carbonyls are shown as solid triangles in the graph but they were not used for correlation analysis due to theirsecondary productionlowastlowastindicates a significant correlation at a confidence level of 001

where [NMHC]agedand [NMHC]fresh(ppb) refer to the abun-dances of NMHC species in the aged and fresh factor respec-tively D is a scaling factor which normalizes the NMHCabundances to unitykOHNMHC is the OH rate constant forthe NMHC [OH] is the average concentration of OH rad-ical and1t is the difference of reaction time between theaged and fresh factors In this study the OH exposure (iethe time integrate of [OH]) expressed by [OH]1t is calcu-lated as a whole

As discussed in Sect 32 for wintertime measurementsthe [NMHC]agedrefers to the abundances of NMHC in fac-tor 4 while the [NMHC]fresh refers to the sum of the abun-dances of NMHC in factors 1ndash3 For summertime measure-ments the [NMHC]agedrefers to the abundances of NMHC

in factor 3 while the [NMHC]fresh refers to the sum of theabundances of NMHC in factors 1 and 2 Figure 6 showedthe dependence of [NMHC]aged [NMHC] fresh (circles in thefigure) onkOH values with the lines being the fitted resultsfrom Eq (4) The values ofD and [OH]1t were estimated tobe 043 and 299times 1010 molecule cmminus3 s in winter and 126and 105times 1011 molecule cmminus3 s in summer respectivelyReferring to the calculated OH radical concentrations in thesummer in Beijing (Z Liu et al 2012) and the differencesof measured OH concentrations in Tokyo between summerand winter (Kanaya et al 2007) we assumed that the aver-age daytime concentration of OH in Beijing was 15times 106

and 6times 106 molecule cmminus3 in winter and summer thus the

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3056 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 6 Relationships between the ratios of VOCs abundances inthe aged emissions factors with those in fresh emission factors andkOH values for each VOC species Each data point represents onecompound The lines are the fitted results from Eq (4)

photochemical ages of the aged factor in winter and summerwere estimated to be 55 and 49 h respectively

The [carbonyl]aged [carbonyl]fresh ratios (Fig 6 solidtriangles) were significantly higher than the ratios of[NMHC]aged [NMHC] fresh because secondary productionenhanced the abundances of carbonyls in aged factors Asshown in Eq (5) the abundances of carbonyls in aged fac-tors can be separated into two parts The first part standsfor aged primary emissions and the second part stands forsecondary formation calculated based on the consumption ofother VOCs[carbonyl

]aged= D times

[carbonyl

]freshtimes eminuskOHcarbonyltimes[OH]1t (5)

+D times

sum([VOC]freshtimes

(1minus eminuskOHVOCtimes[OH]1t

)times YVOCcarbonyl

)wherekOHcarbonyl is the OH rate constant for the carbonylYVOCcarbonyl (ppbppb) refers to the carbonyl productionyield from the oxidation of particular VOC which canbe estimated using the Leeds master chemical mechanism(MCM v32 httpmcmleedsacukMCM) The estimationof YVOCcarbonyl is based on the following assumptions theremoval of VOC is governed by the reaction with OH radi-cals and the removal of alkyl peroxyl radicals (RO2) is gov-erned by the reaction with NO TheYVOCcarbonylvalues usedin this study are listed in Table 2 The VOC species that makeno contributions to the production of our focused carbonylswere not shown in this study The values ofD and [OH]1t

were fitted from Eq (4)The carbonyl abundances in aged factors were calculated

by Eq (5) and agreed well with the PMF-resolved abun-dances (Fig 7) In winter all data points were distributednear the 1 1 line In summer though the data points showedsome variability but still agreed within a factor of two Theagreement between the calculated and PMF-resolved car-bonyl abundances indicated that our understanding on thephotochemical aging of PMF factors was acceptable Thedifferences between the calculated and PMF-resolved car-bonyl abundances may be due to one or some of the fol-lowing reasons (1) the PMF analysis andYVOCcarbonyl es-timations have their own uncertainties (2) Only reactions of

VOCs with OH radicals are considered in our analysis (3)Further reactions of secondary carbonyls are not consideredin our analysis (4) Some VOC species are not applied in ourPMF analysis but they might be precursors of carbonyls ThePMF-analyzed VOC species accounted for 90 and 87 oftotal concentrations of all measured VOCs species in winterand summer respectively However the contribution of eachVOC to the formation of carbonyls was not only dependenton its concentration but also its ability to produce carbonyls

In this work we identified the PMF factors in summerand winter as several individual fresh emissions and sev-eral mixed aged emissions with different degrees of photo-chemical processing Beijing was a large city with extensiveand continuous local emissions and therefore our measuredair mass was in fact a mixture of fresh and aged plumesThe fresh emissions were separated by the PMF model ac-cording to the correlations among VOC species and the dif-ferent emission characteristics of these sources Howeverthe resolved factors by the PMF model cannot represent theemissions at different aged stages from individual sourcesAs discussed in this section the respective photochemicalages of the aged factors in winter and summer were respec-tively about 55 and 49 h and the factor representing sec-ondary production and background levels in winter should bemore aged We conjectured that VOCs from different sourceshave been well mixed during such long aging process and itwas hard to identify the aged contributions from individualsources Likewise previous studies also used averaged emis-sion ratios to characterize the emission and aging process ofurban VOC emissions (de Gouw et al 2005 Warneke et al2007 Yuan et al 2012 Borbon et al 2013) In this sec-tion we proved the reasonability of the PMF-derived agedfactors However one problem remained when apportioningreactive VOCs using the PMF model Actually the photo-chemical reaction of VOCs in the atmosphere is a continuousprocess but the PMF model is not able to describe such kindof process In this work several different aged stages wererecognized by the PMF model and the PMF model used thelinear combination of these stages to describe the continuousphotochemical process This might be an important issue forthe PMF analysis and the influence of such approximationrequired future researches

34 Primary and secondary sources of carbonyls

The calculated relative contributions of carbonyl sources ap-portioned by the PMF model in Beijing were shown in Fig 8Previous studies usually attributed all carbonyls in aged fac-tors as secondary formation As discussed in Sect 33 theabundances of carbonyls in aged factors could be separatedinto two parts that is the aged direct emissions and theproduction from VOC consumption and the former partshould be treated as primary emissions In this study theaged direct emission part was further separated into eachfresh factor considering their relative contributions For all

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3051

Fig 1 Diurnal variations of measured ratios of formaldehyde ethene acetaldehyde propene and acetone ethane at the PKU site duringwinter and summer Black lines represent the average values and grey shaded areas indicate standard deviations

summed concentrations of formaldehyde and acetaldehydeindicated that the emission controls on these carbonyls ortheir precursors in Beijing possibly made progress during thepast years However the flat trend during winter suggestedthat there might be some important sources for carbonyls inwinter were not under effective control

Figure 1 shows the diurnal variations of ratios of formalde-hyde ethene acetaldehyde propene and acetone ethane atthe PKU site during winter and summer Ambient concentra-tions of carbonyls were influenced by meteorological con-ditions and the emission strengths of carbonyls and theirprecursors Assuming that the direct emissions of carbonylsfrom anthropogenic sources were proportional to the emis-sions of NMHCs the diurnal variations of carbonyl NMHCratios could reflect the importance of secondary formationto ambient carbonyls These three pairs of carbonyls andNMHCs were chosen because the two species in each pairhad similar reaction rates with OH radicals (Atkinson et al2006) As a result the two species in the atmosphere wereremoved approximately equally by reactions with OH radi-cals and therefore the effect of photochemical degradationon their ratios can be neglected As shown in Fig 1 themeasured carbonyl NMHC ratios showed similar daily vari-ations both in summer and winter they were stable dur-ing nighttime increased after sunrise (about 0800 LT) andreached a maximum in the early afternoon (about 1300ndash1500 LT) then decreased to low values at night This varia-tion indicated an important contribution from secondary pro-duction during the daytime both in winter and summer Thediurnal variations of the acetone ethane ratios were less thanthose for formaldehyde ethene and acetaldehyde propeneSuch difference was consistent with previous studies whichfound that secondary production of acetone was less impor-tant than that of formaldehyde and acetaldehyde (de Gouw etal 2005 Yuan et al 2012)

Though the diurnal patterns of carbonyl NMHC ratioswere similar in winter and summer the values of car-bonyl NMHC ratios were approximately 3ndash5 times higher in

summer than those in winter during both day and night Asmentioned above ambient concentrations of carbonyls didnot show significant discrepancies between these two sea-sons The seasonal differences of carbonyl NMHC ratioswere mainly due to the much higher NMHC levels in win-ter This also indicated the significant differences in VOCsources between these two seasons Considering the effectsof photochemical processing on VOC ratios were minor atnight nighttime VOC ratios were close to their emission ra-tios (Wang et al 2013) Therefore the higher nighttime car-bonyl NMHC ratios in summer indicated that the fractionof NMHCs in primary emitted VOCs was higher in winterthan that in summer whereas carbonyls fraction was higherin summer

Previous studies usually supposed that carbonyls in win-ter mainly came from direct anthropogenic emissions (Pangand Mu 2006 Ceroacuten et al 2007) From our measure-ments the differences of carbonyl NMHC ratios betweennoon and night in summer were only slightly larger thanthe differences in winter The noon night ratios were 26and 31 for formaldehyde ethene 26 and 35 for acetalde-hyde propene and 14 and 15 for acetone ethane in winterand summer respectively This indicated that secondary for-mation was still an important source of carbonyls in winterIt should be noted that photolysis removal of carbonyls wasalso considered here besides the removal by reactions withOH radicals According to our estimation reaction with OHwas the main pathway of carbonyl removal The photolysiscontributed 33 1 and 12 to the losses of formalde-hyde acetaldehyde and acetone at daytime and these contri-butions showed no significant difference between winter andsummer The details for the calculation of carbonyl removalrates were shown in the Supplement As a result neglect-ing the photolysis of carbonyls had little effect on comparingthe contributions of secondary formation between winter andsummer and the variation of carbonyl NMHC ratios couldreflect the importance of secondary formation For furtherdiscussion the PMF model was used for carbonyls source

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3052 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 2Profiles of the four factors resolved by the PMF model and the distributions of species among these factors in summer

apportionment to determine relative contributions from pri-mary and secondary sources

32 Identifying PMF factors

The PMF model is a mathematical tool which separates pol-lutants according to the correlations among species and theunderstanding of PMF factors is subjective to a certain ex-tent We were aware of the deficiency of the PMF modelwhen apportioning VOC sources In this section we com-pared the profiles derived by the PMF model with previouslymeasured ones to check the reliability of resolved fresh fac-tors In the next section we tried to identify aged factorsby investigating the relationship between degradation of pri-mary emitted VOCs and production of carbonyls

For summertime measurements 593 samples were usedfor the PMF analysis and four factors were identified Theprofiles of resolved PMF factors and the distributions ofspecies among the factors were shown in Fig 2 The profileswere expressed as the mass percentage of individual VOCspecies in each factor and the distributions represented thecontributions of individual factors to the measured level ofeach species

Factor 1 made an important contribution to our measuredalkanes and alkenes The profile of this factor was dominatedby C2ndashC5 alkanes C2ndashC3 alkenes ethyne and aromaticsThese species were mainly emitted from exhaust of vehiclesand evaporation of gasoline (Liu et al 2008) This factor ex-plained 381 of the measured MTBE which was com-monly used as a tracer of gasoline vehicle emissions (Changet al 2006) Previous studies also reported that traffic-relatedemissions were the most important NMHCs sources in sum-mer of Beijing (Song et al 2008 Wang et al 2010) Wetherefore attributed this factor to traffic-related emissions

In factor 2 the loadings of aromatics including benzenetoluene ethylbenzene and xylenes were high Aromatics

were reported to be major constituents of solvents (Yuan etal 2010) which were widely used in industrial processes aswell as some household products Thus we considered thisfactor to be related to industry and solvent usage

Factor 3 had contributions to almost all measured VOCspecies except for some very reactive species This factor ac-counted for more than 60 of our measured carbonyls and875 of our measured 2-BuONO2 levels Alkyl nitrateswere believed to be mainly from secondary production in ur-ban regions and had relatively long lifetimes (Simpson et al2003) Thus we identified this factor as aged emissions

Factor 4 contributed 815 of measured isoprene whichwas often considered as a tracer for biogenic emissions (Situet al 2013) so we attributed this factor to biogenic emis-sions High loadings of formaldehyde acetaldehyde andacetone were also found in this factor This is possibly be-cause these carbonyls could also be emitted from biogenicsources (Winters et al 2009) and they were important ox-idation products of other biogenic VOCs (Carter and Atkin-son 1996)

Based on the derived profiles of each factor and the dis-tributions of species among these factors we identified thattwo of the four PMF-resolved factors could represent pri-mary emission sources another related to biogenic emis-sions and the other was an aged emission factor As dis-cussed in Sect 22 the relationship between the contributionof one factor to each NMHC species and its chemical reactiv-ity could be used to distinguish factors related to photochem-ically aged emissions from those for fresh emissions Such arelationship was shown in Fig 3 where each NMHC specieswas shown as a circle Positive correlations were identifiedfor factors 1 and 2 and thus these two factors were consid-ered to be related to fresh emissions However a significantnegative correlation was identified for factor 3 indicatingthat this factor represented photochemically aged emissions

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3053

Fig 3 Relationships between the factor contributions to each NMHC species andkOH values for NMHC species in summer Each circlerepresents one compound Carbonyls are shown as solid triangles in the graph but they were not used for correlation analysis due to theirsecondary productionlowastindicates a significant correlation at a confidence level of 005lowastlowastindicates a significant correlation at a confidencelevel of 001

The distributions of carbonyls in PMF-resolved factorswere shown in Fig 3 by solid triangles If all factors were re-lated to fresh emissions the distributions of carbonyls wouldbe similar to the distributions of NMHCs in each factor Dueto the influence from photochemical aging the distribution ofcarbonyls could be higher in aged factors owing to their sec-ondary production and meanwhile the distributions of car-bonyls would be lower in fresh factors In factors 1 and 2 thedistributions of carbonyls were similar or lower than those ofNMHCs whereas the distributions of carbonyls in factor 3were higher than those of NMHCs Such appearance agreedwith our above identifications on PMF-resolved factors

In this study two fresh factors and an aged factor wereidentified in the summer of 2011 However Yuan etal (2012) identified one mixed fresh factor and two aged fac-tors with different photochemical ages using a similar anal-ysis approach based on measurements at the same site in thesummer of 2010 It was interesting to get such different re-sults in two consecutive years Yuan et al (2012) concludedthat PMF results depended on the degree of photochemicalprocessing and the differences in emission compositions ofvarious sources Assuming that the differences in VOC emis-sions were not significant between 2010 and 2011 the PMFresults were mainly influenced by the difference in degrees

in photochemical aging between these two years We usedthe ratio of o-xylene to ethylbenzene as an indicator of thedegree of photochemical processing As seen in Fig S4 therelative standard deviations of o-xylene ethylbenzene ratiosin 2010 were 30 larger than those in 2011 It indicated thatthe range of photochemical processing degrees was larger in2010 As a result the PMF factors in 2010 were extractedmainly according to different degrees of photochemical pro-cessing whereas the PMF factors in 2011 could be extractedmainly based on individual sources

In winter 341 samples were used for the PMF analysisand five factors were identified (Fig 4)

Factor 1 and factor 3 contributed to most of our measuredalkanes and alkenes but there were some differences be-tween these two factors Factor 1 had higher contributions toC2ndashC3 NMHCs while factor 3 had higher contributions toC4ndashC5 NMHCs These light hydrocarbons were mainly gen-erated by incomplete combustion processes such as vehicleexhaust and coal burning (Liu et al 2008 Wang et al 2013)The measured source profiles in China showed that C2ndashC3NMHCs contributed more than half of NMHCs that emit-ted from coal burning (Liu et al 2008 Wang et al 2013)Benzene and toluene also made important contributions tofactor 1 These two species were also found to be important

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3054 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 4Profiles of the five factors resolved by the PMF model and the distributions of species among these factors in winter

components in NMHCs emissions from coal burning (Liu etal 2008) The ratio of benzene to toluene (231 ppb ppbminus1)

in this factor fell between the ratios measured by Liu etal (2008) for residential coal burning (181 ppb ppbminus1) andindustrial coal burning (262 ppb ppbminus1) but much higherthan that measured in the tunnel experiment (070 ppb ppbminus1)

(Liu et al 2008) therefore factor 1 was identified as coalburning C4ndashC5 NMHCs were important species from trafficrelated emissions and factor 3 showed similar characteris-tic with factor 1 in summer In addition factor 3 explained379 of the measured MTBE Therefore factor 3 was at-tributed to traffic related emissions

Factor 2 had high contributions to aromatics and showedsimilar characteristic with the resolved factor 2 for sum-mertime measurements Besides aromatics this factor con-tributed about 30 of measured long-chain alkanes (C7ndashC10 alkanes) The long-chain alkanes were also consid-ered to be important components of printing VOC emissions(Yuan et al 2010) Thus this factor was identified as indus-try and solvent usage

Both factor 4 and factor 5 had important contributionsto carbonyls but little contribution to reactive NMHCs andtherefore these two factors were inferred to possibly not rep-resent fresh emissions The major NMHC species in factor 4were C2-C5 alkanes ethene ethyne benzene and tolueneThese species could be emitted from all of the three primaryfactors discussed above and thus we considered this factor tobe aged emissions from these anthropogenic sources TheNMHCs in factor 5 were dominated by unreactive alkanessuch as ethane and propane so factor 5 was at a more aged

stage than factor 4 Factor 5 accounted for 677 of the mea-sured 2-BuONO2 levels and 197 of the measured carbonyllevels According to the abundances of alkyl nitrates and un-reactive alkanes we considered that the loadings of VOCspecies in this factor were related to secondary productionand background levels

Similarly with summertime measurements the relation-ships between the contribution of one factor to each NMHCspecies andkOH values for each NMHC species were ana-lyzed As shown in Fig 5 no significant correlation was ob-served for factor 1 and factor 2 while a positive correlationwas found for factor 3 and thus these three factors were con-sidered to be fresh emissions Significant negative correla-tions were identified for factor 4 and factor 5 indicating thatthese two factors were influenced by photochemical aging

33 Exploring the aged and fresh emission factors

In the last section we distinguished photochemically agedfactors from fresh factors To better understand the relation-ships between aged and fresh factors and the role of photo-chemical aging in PMF analysis we explored the relation-ships between VOC consumption and carbonyl formation inthe aged factors

For two factors representing different photochemicalstages of the same emissions the ratios of NMHC abun-dances in these two factors should follow the equation below(Yuan et al 2012)

[NMHC]aged

[NMHC]fresh= D times eminuskOHNMHC[OH]1t (4)

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3055

Fig 5 Relationships between the factor contributions to each NMHC species andkOH values of NMHC species in winter Each circlerepresents one compound Carbonyls are shown as solid triangles in the graph but they were not used for correlation analysis due to theirsecondary productionlowastlowastindicates a significant correlation at a confidence level of 001

where [NMHC]agedand [NMHC]fresh(ppb) refer to the abun-dances of NMHC species in the aged and fresh factor respec-tively D is a scaling factor which normalizes the NMHCabundances to unitykOHNMHC is the OH rate constant forthe NMHC [OH] is the average concentration of OH rad-ical and1t is the difference of reaction time between theaged and fresh factors In this study the OH exposure (iethe time integrate of [OH]) expressed by [OH]1t is calcu-lated as a whole

As discussed in Sect 32 for wintertime measurementsthe [NMHC]agedrefers to the abundances of NMHC in fac-tor 4 while the [NMHC]fresh refers to the sum of the abun-dances of NMHC in factors 1ndash3 For summertime measure-ments the [NMHC]agedrefers to the abundances of NMHC

in factor 3 while the [NMHC]fresh refers to the sum of theabundances of NMHC in factors 1 and 2 Figure 6 showedthe dependence of [NMHC]aged [NMHC] fresh (circles in thefigure) onkOH values with the lines being the fitted resultsfrom Eq (4) The values ofD and [OH]1t were estimated tobe 043 and 299times 1010 molecule cmminus3 s in winter and 126and 105times 1011 molecule cmminus3 s in summer respectivelyReferring to the calculated OH radical concentrations in thesummer in Beijing (Z Liu et al 2012) and the differencesof measured OH concentrations in Tokyo between summerand winter (Kanaya et al 2007) we assumed that the aver-age daytime concentration of OH in Beijing was 15times 106

and 6times 106 molecule cmminus3 in winter and summer thus the

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3056 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 6 Relationships between the ratios of VOCs abundances inthe aged emissions factors with those in fresh emission factors andkOH values for each VOC species Each data point represents onecompound The lines are the fitted results from Eq (4)

photochemical ages of the aged factor in winter and summerwere estimated to be 55 and 49 h respectively

The [carbonyl]aged [carbonyl]fresh ratios (Fig 6 solidtriangles) were significantly higher than the ratios of[NMHC]aged [NMHC] fresh because secondary productionenhanced the abundances of carbonyls in aged factors Asshown in Eq (5) the abundances of carbonyls in aged fac-tors can be separated into two parts The first part standsfor aged primary emissions and the second part stands forsecondary formation calculated based on the consumption ofother VOCs[carbonyl

]aged= D times

[carbonyl

]freshtimes eminuskOHcarbonyltimes[OH]1t (5)

+D times

sum([VOC]freshtimes

(1minus eminuskOHVOCtimes[OH]1t

)times YVOCcarbonyl

)wherekOHcarbonyl is the OH rate constant for the carbonylYVOCcarbonyl (ppbppb) refers to the carbonyl productionyield from the oxidation of particular VOC which canbe estimated using the Leeds master chemical mechanism(MCM v32 httpmcmleedsacukMCM) The estimationof YVOCcarbonyl is based on the following assumptions theremoval of VOC is governed by the reaction with OH radi-cals and the removal of alkyl peroxyl radicals (RO2) is gov-erned by the reaction with NO TheYVOCcarbonylvalues usedin this study are listed in Table 2 The VOC species that makeno contributions to the production of our focused carbonylswere not shown in this study The values ofD and [OH]1t

were fitted from Eq (4)The carbonyl abundances in aged factors were calculated

by Eq (5) and agreed well with the PMF-resolved abun-dances (Fig 7) In winter all data points were distributednear the 1 1 line In summer though the data points showedsome variability but still agreed within a factor of two Theagreement between the calculated and PMF-resolved car-bonyl abundances indicated that our understanding on thephotochemical aging of PMF factors was acceptable Thedifferences between the calculated and PMF-resolved car-bonyl abundances may be due to one or some of the fol-lowing reasons (1) the PMF analysis andYVOCcarbonyl es-timations have their own uncertainties (2) Only reactions of

VOCs with OH radicals are considered in our analysis (3)Further reactions of secondary carbonyls are not consideredin our analysis (4) Some VOC species are not applied in ourPMF analysis but they might be precursors of carbonyls ThePMF-analyzed VOC species accounted for 90 and 87 oftotal concentrations of all measured VOCs species in winterand summer respectively However the contribution of eachVOC to the formation of carbonyls was not only dependenton its concentration but also its ability to produce carbonyls

In this work we identified the PMF factors in summerand winter as several individual fresh emissions and sev-eral mixed aged emissions with different degrees of photo-chemical processing Beijing was a large city with extensiveand continuous local emissions and therefore our measuredair mass was in fact a mixture of fresh and aged plumesThe fresh emissions were separated by the PMF model ac-cording to the correlations among VOC species and the dif-ferent emission characteristics of these sources Howeverthe resolved factors by the PMF model cannot represent theemissions at different aged stages from individual sourcesAs discussed in this section the respective photochemicalages of the aged factors in winter and summer were respec-tively about 55 and 49 h and the factor representing sec-ondary production and background levels in winter should bemore aged We conjectured that VOCs from different sourceshave been well mixed during such long aging process and itwas hard to identify the aged contributions from individualsources Likewise previous studies also used averaged emis-sion ratios to characterize the emission and aging process ofurban VOC emissions (de Gouw et al 2005 Warneke et al2007 Yuan et al 2012 Borbon et al 2013) In this sec-tion we proved the reasonability of the PMF-derived agedfactors However one problem remained when apportioningreactive VOCs using the PMF model Actually the photo-chemical reaction of VOCs in the atmosphere is a continuousprocess but the PMF model is not able to describe such kindof process In this work several different aged stages wererecognized by the PMF model and the PMF model used thelinear combination of these stages to describe the continuousphotochemical process This might be an important issue forthe PMF analysis and the influence of such approximationrequired future researches

34 Primary and secondary sources of carbonyls

The calculated relative contributions of carbonyl sources ap-portioned by the PMF model in Beijing were shown in Fig 8Previous studies usually attributed all carbonyls in aged fac-tors as secondary formation As discussed in Sect 33 theabundances of carbonyls in aged factors could be separatedinto two parts that is the aged direct emissions and theproduction from VOC consumption and the former partshould be treated as primary emissions In this study theaged direct emission part was further separated into eachfresh factor considering their relative contributions For all

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

3052 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 2Profiles of the four factors resolved by the PMF model and the distributions of species among these factors in summer

apportionment to determine relative contributions from pri-mary and secondary sources

32 Identifying PMF factors

The PMF model is a mathematical tool which separates pol-lutants according to the correlations among species and theunderstanding of PMF factors is subjective to a certain ex-tent We were aware of the deficiency of the PMF modelwhen apportioning VOC sources In this section we com-pared the profiles derived by the PMF model with previouslymeasured ones to check the reliability of resolved fresh fac-tors In the next section we tried to identify aged factorsby investigating the relationship between degradation of pri-mary emitted VOCs and production of carbonyls

For summertime measurements 593 samples were usedfor the PMF analysis and four factors were identified Theprofiles of resolved PMF factors and the distributions ofspecies among the factors were shown in Fig 2 The profileswere expressed as the mass percentage of individual VOCspecies in each factor and the distributions represented thecontributions of individual factors to the measured level ofeach species

Factor 1 made an important contribution to our measuredalkanes and alkenes The profile of this factor was dominatedby C2ndashC5 alkanes C2ndashC3 alkenes ethyne and aromaticsThese species were mainly emitted from exhaust of vehiclesand evaporation of gasoline (Liu et al 2008) This factor ex-plained 381 of the measured MTBE which was com-monly used as a tracer of gasoline vehicle emissions (Changet al 2006) Previous studies also reported that traffic-relatedemissions were the most important NMHCs sources in sum-mer of Beijing (Song et al 2008 Wang et al 2010) Wetherefore attributed this factor to traffic-related emissions

In factor 2 the loadings of aromatics including benzenetoluene ethylbenzene and xylenes were high Aromatics

were reported to be major constituents of solvents (Yuan etal 2010) which were widely used in industrial processes aswell as some household products Thus we considered thisfactor to be related to industry and solvent usage

Factor 3 had contributions to almost all measured VOCspecies except for some very reactive species This factor ac-counted for more than 60 of our measured carbonyls and875 of our measured 2-BuONO2 levels Alkyl nitrateswere believed to be mainly from secondary production in ur-ban regions and had relatively long lifetimes (Simpson et al2003) Thus we identified this factor as aged emissions

Factor 4 contributed 815 of measured isoprene whichwas often considered as a tracer for biogenic emissions (Situet al 2013) so we attributed this factor to biogenic emis-sions High loadings of formaldehyde acetaldehyde andacetone were also found in this factor This is possibly be-cause these carbonyls could also be emitted from biogenicsources (Winters et al 2009) and they were important ox-idation products of other biogenic VOCs (Carter and Atkin-son 1996)

Based on the derived profiles of each factor and the dis-tributions of species among these factors we identified thattwo of the four PMF-resolved factors could represent pri-mary emission sources another related to biogenic emis-sions and the other was an aged emission factor As dis-cussed in Sect 22 the relationship between the contributionof one factor to each NMHC species and its chemical reactiv-ity could be used to distinguish factors related to photochem-ically aged emissions from those for fresh emissions Such arelationship was shown in Fig 3 where each NMHC specieswas shown as a circle Positive correlations were identifiedfor factors 1 and 2 and thus these two factors were consid-ered to be related to fresh emissions However a significantnegative correlation was identified for factor 3 indicatingthat this factor represented photochemically aged emissions

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3053

Fig 3 Relationships between the factor contributions to each NMHC species andkOH values for NMHC species in summer Each circlerepresents one compound Carbonyls are shown as solid triangles in the graph but they were not used for correlation analysis due to theirsecondary productionlowastindicates a significant correlation at a confidence level of 005lowastlowastindicates a significant correlation at a confidencelevel of 001

The distributions of carbonyls in PMF-resolved factorswere shown in Fig 3 by solid triangles If all factors were re-lated to fresh emissions the distributions of carbonyls wouldbe similar to the distributions of NMHCs in each factor Dueto the influence from photochemical aging the distribution ofcarbonyls could be higher in aged factors owing to their sec-ondary production and meanwhile the distributions of car-bonyls would be lower in fresh factors In factors 1 and 2 thedistributions of carbonyls were similar or lower than those ofNMHCs whereas the distributions of carbonyls in factor 3were higher than those of NMHCs Such appearance agreedwith our above identifications on PMF-resolved factors

In this study two fresh factors and an aged factor wereidentified in the summer of 2011 However Yuan etal (2012) identified one mixed fresh factor and two aged fac-tors with different photochemical ages using a similar anal-ysis approach based on measurements at the same site in thesummer of 2010 It was interesting to get such different re-sults in two consecutive years Yuan et al (2012) concludedthat PMF results depended on the degree of photochemicalprocessing and the differences in emission compositions ofvarious sources Assuming that the differences in VOC emis-sions were not significant between 2010 and 2011 the PMFresults were mainly influenced by the difference in degrees

in photochemical aging between these two years We usedthe ratio of o-xylene to ethylbenzene as an indicator of thedegree of photochemical processing As seen in Fig S4 therelative standard deviations of o-xylene ethylbenzene ratiosin 2010 were 30 larger than those in 2011 It indicated thatthe range of photochemical processing degrees was larger in2010 As a result the PMF factors in 2010 were extractedmainly according to different degrees of photochemical pro-cessing whereas the PMF factors in 2011 could be extractedmainly based on individual sources

In winter 341 samples were used for the PMF analysisand five factors were identified (Fig 4)

Factor 1 and factor 3 contributed to most of our measuredalkanes and alkenes but there were some differences be-tween these two factors Factor 1 had higher contributions toC2ndashC3 NMHCs while factor 3 had higher contributions toC4ndashC5 NMHCs These light hydrocarbons were mainly gen-erated by incomplete combustion processes such as vehicleexhaust and coal burning (Liu et al 2008 Wang et al 2013)The measured source profiles in China showed that C2ndashC3NMHCs contributed more than half of NMHCs that emit-ted from coal burning (Liu et al 2008 Wang et al 2013)Benzene and toluene also made important contributions tofactor 1 These two species were also found to be important

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3054 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 4Profiles of the five factors resolved by the PMF model and the distributions of species among these factors in winter

components in NMHCs emissions from coal burning (Liu etal 2008) The ratio of benzene to toluene (231 ppb ppbminus1)

in this factor fell between the ratios measured by Liu etal (2008) for residential coal burning (181 ppb ppbminus1) andindustrial coal burning (262 ppb ppbminus1) but much higherthan that measured in the tunnel experiment (070 ppb ppbminus1)

(Liu et al 2008) therefore factor 1 was identified as coalburning C4ndashC5 NMHCs were important species from trafficrelated emissions and factor 3 showed similar characteris-tic with factor 1 in summer In addition factor 3 explained379 of the measured MTBE Therefore factor 3 was at-tributed to traffic related emissions

Factor 2 had high contributions to aromatics and showedsimilar characteristic with the resolved factor 2 for sum-mertime measurements Besides aromatics this factor con-tributed about 30 of measured long-chain alkanes (C7ndashC10 alkanes) The long-chain alkanes were also consid-ered to be important components of printing VOC emissions(Yuan et al 2010) Thus this factor was identified as indus-try and solvent usage

Both factor 4 and factor 5 had important contributionsto carbonyls but little contribution to reactive NMHCs andtherefore these two factors were inferred to possibly not rep-resent fresh emissions The major NMHC species in factor 4were C2-C5 alkanes ethene ethyne benzene and tolueneThese species could be emitted from all of the three primaryfactors discussed above and thus we considered this factor tobe aged emissions from these anthropogenic sources TheNMHCs in factor 5 were dominated by unreactive alkanessuch as ethane and propane so factor 5 was at a more aged

stage than factor 4 Factor 5 accounted for 677 of the mea-sured 2-BuONO2 levels and 197 of the measured carbonyllevels According to the abundances of alkyl nitrates and un-reactive alkanes we considered that the loadings of VOCspecies in this factor were related to secondary productionand background levels

Similarly with summertime measurements the relation-ships between the contribution of one factor to each NMHCspecies andkOH values for each NMHC species were ana-lyzed As shown in Fig 5 no significant correlation was ob-served for factor 1 and factor 2 while a positive correlationwas found for factor 3 and thus these three factors were con-sidered to be fresh emissions Significant negative correla-tions were identified for factor 4 and factor 5 indicating thatthese two factors were influenced by photochemical aging

33 Exploring the aged and fresh emission factors

In the last section we distinguished photochemically agedfactors from fresh factors To better understand the relation-ships between aged and fresh factors and the role of photo-chemical aging in PMF analysis we explored the relation-ships between VOC consumption and carbonyl formation inthe aged factors

For two factors representing different photochemicalstages of the same emissions the ratios of NMHC abun-dances in these two factors should follow the equation below(Yuan et al 2012)

[NMHC]aged

[NMHC]fresh= D times eminuskOHNMHC[OH]1t (4)

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3055

Fig 5 Relationships between the factor contributions to each NMHC species andkOH values of NMHC species in winter Each circlerepresents one compound Carbonyls are shown as solid triangles in the graph but they were not used for correlation analysis due to theirsecondary productionlowastlowastindicates a significant correlation at a confidence level of 001

where [NMHC]agedand [NMHC]fresh(ppb) refer to the abun-dances of NMHC species in the aged and fresh factor respec-tively D is a scaling factor which normalizes the NMHCabundances to unitykOHNMHC is the OH rate constant forthe NMHC [OH] is the average concentration of OH rad-ical and1t is the difference of reaction time between theaged and fresh factors In this study the OH exposure (iethe time integrate of [OH]) expressed by [OH]1t is calcu-lated as a whole

As discussed in Sect 32 for wintertime measurementsthe [NMHC]agedrefers to the abundances of NMHC in fac-tor 4 while the [NMHC]fresh refers to the sum of the abun-dances of NMHC in factors 1ndash3 For summertime measure-ments the [NMHC]agedrefers to the abundances of NMHC

in factor 3 while the [NMHC]fresh refers to the sum of theabundances of NMHC in factors 1 and 2 Figure 6 showedthe dependence of [NMHC]aged [NMHC] fresh (circles in thefigure) onkOH values with the lines being the fitted resultsfrom Eq (4) The values ofD and [OH]1t were estimated tobe 043 and 299times 1010 molecule cmminus3 s in winter and 126and 105times 1011 molecule cmminus3 s in summer respectivelyReferring to the calculated OH radical concentrations in thesummer in Beijing (Z Liu et al 2012) and the differencesof measured OH concentrations in Tokyo between summerand winter (Kanaya et al 2007) we assumed that the aver-age daytime concentration of OH in Beijing was 15times 106

and 6times 106 molecule cmminus3 in winter and summer thus the

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3056 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 6 Relationships between the ratios of VOCs abundances inthe aged emissions factors with those in fresh emission factors andkOH values for each VOC species Each data point represents onecompound The lines are the fitted results from Eq (4)

photochemical ages of the aged factor in winter and summerwere estimated to be 55 and 49 h respectively

The [carbonyl]aged [carbonyl]fresh ratios (Fig 6 solidtriangles) were significantly higher than the ratios of[NMHC]aged [NMHC] fresh because secondary productionenhanced the abundances of carbonyls in aged factors Asshown in Eq (5) the abundances of carbonyls in aged fac-tors can be separated into two parts The first part standsfor aged primary emissions and the second part stands forsecondary formation calculated based on the consumption ofother VOCs[carbonyl

]aged= D times

[carbonyl

]freshtimes eminuskOHcarbonyltimes[OH]1t (5)

+D times

sum([VOC]freshtimes

(1minus eminuskOHVOCtimes[OH]1t

)times YVOCcarbonyl

)wherekOHcarbonyl is the OH rate constant for the carbonylYVOCcarbonyl (ppbppb) refers to the carbonyl productionyield from the oxidation of particular VOC which canbe estimated using the Leeds master chemical mechanism(MCM v32 httpmcmleedsacukMCM) The estimationof YVOCcarbonyl is based on the following assumptions theremoval of VOC is governed by the reaction with OH radi-cals and the removal of alkyl peroxyl radicals (RO2) is gov-erned by the reaction with NO TheYVOCcarbonylvalues usedin this study are listed in Table 2 The VOC species that makeno contributions to the production of our focused carbonylswere not shown in this study The values ofD and [OH]1t

were fitted from Eq (4)The carbonyl abundances in aged factors were calculated

by Eq (5) and agreed well with the PMF-resolved abun-dances (Fig 7) In winter all data points were distributednear the 1 1 line In summer though the data points showedsome variability but still agreed within a factor of two Theagreement between the calculated and PMF-resolved car-bonyl abundances indicated that our understanding on thephotochemical aging of PMF factors was acceptable Thedifferences between the calculated and PMF-resolved car-bonyl abundances may be due to one or some of the fol-lowing reasons (1) the PMF analysis andYVOCcarbonyl es-timations have their own uncertainties (2) Only reactions of

VOCs with OH radicals are considered in our analysis (3)Further reactions of secondary carbonyls are not consideredin our analysis (4) Some VOC species are not applied in ourPMF analysis but they might be precursors of carbonyls ThePMF-analyzed VOC species accounted for 90 and 87 oftotal concentrations of all measured VOCs species in winterand summer respectively However the contribution of eachVOC to the formation of carbonyls was not only dependenton its concentration but also its ability to produce carbonyls

In this work we identified the PMF factors in summerand winter as several individual fresh emissions and sev-eral mixed aged emissions with different degrees of photo-chemical processing Beijing was a large city with extensiveand continuous local emissions and therefore our measuredair mass was in fact a mixture of fresh and aged plumesThe fresh emissions were separated by the PMF model ac-cording to the correlations among VOC species and the dif-ferent emission characteristics of these sources Howeverthe resolved factors by the PMF model cannot represent theemissions at different aged stages from individual sourcesAs discussed in this section the respective photochemicalages of the aged factors in winter and summer were respec-tively about 55 and 49 h and the factor representing sec-ondary production and background levels in winter should bemore aged We conjectured that VOCs from different sourceshave been well mixed during such long aging process and itwas hard to identify the aged contributions from individualsources Likewise previous studies also used averaged emis-sion ratios to characterize the emission and aging process ofurban VOC emissions (de Gouw et al 2005 Warneke et al2007 Yuan et al 2012 Borbon et al 2013) In this sec-tion we proved the reasonability of the PMF-derived agedfactors However one problem remained when apportioningreactive VOCs using the PMF model Actually the photo-chemical reaction of VOCs in the atmosphere is a continuousprocess but the PMF model is not able to describe such kindof process In this work several different aged stages wererecognized by the PMF model and the PMF model used thelinear combination of these stages to describe the continuousphotochemical process This might be an important issue forthe PMF analysis and the influence of such approximationrequired future researches

34 Primary and secondary sources of carbonyls

The calculated relative contributions of carbonyl sources ap-portioned by the PMF model in Beijing were shown in Fig 8Previous studies usually attributed all carbonyls in aged fac-tors as secondary formation As discussed in Sect 33 theabundances of carbonyls in aged factors could be separatedinto two parts that is the aged direct emissions and theproduction from VOC consumption and the former partshould be treated as primary emissions In this study theaged direct emission part was further separated into eachfresh factor considering their relative contributions For all

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3053

Fig 3 Relationships between the factor contributions to each NMHC species andkOH values for NMHC species in summer Each circlerepresents one compound Carbonyls are shown as solid triangles in the graph but they were not used for correlation analysis due to theirsecondary productionlowastindicates a significant correlation at a confidence level of 005lowastlowastindicates a significant correlation at a confidencelevel of 001

The distributions of carbonyls in PMF-resolved factorswere shown in Fig 3 by solid triangles If all factors were re-lated to fresh emissions the distributions of carbonyls wouldbe similar to the distributions of NMHCs in each factor Dueto the influence from photochemical aging the distribution ofcarbonyls could be higher in aged factors owing to their sec-ondary production and meanwhile the distributions of car-bonyls would be lower in fresh factors In factors 1 and 2 thedistributions of carbonyls were similar or lower than those ofNMHCs whereas the distributions of carbonyls in factor 3were higher than those of NMHCs Such appearance agreedwith our above identifications on PMF-resolved factors

In this study two fresh factors and an aged factor wereidentified in the summer of 2011 However Yuan etal (2012) identified one mixed fresh factor and two aged fac-tors with different photochemical ages using a similar anal-ysis approach based on measurements at the same site in thesummer of 2010 It was interesting to get such different re-sults in two consecutive years Yuan et al (2012) concludedthat PMF results depended on the degree of photochemicalprocessing and the differences in emission compositions ofvarious sources Assuming that the differences in VOC emis-sions were not significant between 2010 and 2011 the PMFresults were mainly influenced by the difference in degrees

in photochemical aging between these two years We usedthe ratio of o-xylene to ethylbenzene as an indicator of thedegree of photochemical processing As seen in Fig S4 therelative standard deviations of o-xylene ethylbenzene ratiosin 2010 were 30 larger than those in 2011 It indicated thatthe range of photochemical processing degrees was larger in2010 As a result the PMF factors in 2010 were extractedmainly according to different degrees of photochemical pro-cessing whereas the PMF factors in 2011 could be extractedmainly based on individual sources

In winter 341 samples were used for the PMF analysisand five factors were identified (Fig 4)

Factor 1 and factor 3 contributed to most of our measuredalkanes and alkenes but there were some differences be-tween these two factors Factor 1 had higher contributions toC2ndashC3 NMHCs while factor 3 had higher contributions toC4ndashC5 NMHCs These light hydrocarbons were mainly gen-erated by incomplete combustion processes such as vehicleexhaust and coal burning (Liu et al 2008 Wang et al 2013)The measured source profiles in China showed that C2ndashC3NMHCs contributed more than half of NMHCs that emit-ted from coal burning (Liu et al 2008 Wang et al 2013)Benzene and toluene also made important contributions tofactor 1 These two species were also found to be important

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3054 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 4Profiles of the five factors resolved by the PMF model and the distributions of species among these factors in winter

components in NMHCs emissions from coal burning (Liu etal 2008) The ratio of benzene to toluene (231 ppb ppbminus1)

in this factor fell between the ratios measured by Liu etal (2008) for residential coal burning (181 ppb ppbminus1) andindustrial coal burning (262 ppb ppbminus1) but much higherthan that measured in the tunnel experiment (070 ppb ppbminus1)

(Liu et al 2008) therefore factor 1 was identified as coalburning C4ndashC5 NMHCs were important species from trafficrelated emissions and factor 3 showed similar characteris-tic with factor 1 in summer In addition factor 3 explained379 of the measured MTBE Therefore factor 3 was at-tributed to traffic related emissions

Factor 2 had high contributions to aromatics and showedsimilar characteristic with the resolved factor 2 for sum-mertime measurements Besides aromatics this factor con-tributed about 30 of measured long-chain alkanes (C7ndashC10 alkanes) The long-chain alkanes were also consid-ered to be important components of printing VOC emissions(Yuan et al 2010) Thus this factor was identified as indus-try and solvent usage

Both factor 4 and factor 5 had important contributionsto carbonyls but little contribution to reactive NMHCs andtherefore these two factors were inferred to possibly not rep-resent fresh emissions The major NMHC species in factor 4were C2-C5 alkanes ethene ethyne benzene and tolueneThese species could be emitted from all of the three primaryfactors discussed above and thus we considered this factor tobe aged emissions from these anthropogenic sources TheNMHCs in factor 5 were dominated by unreactive alkanessuch as ethane and propane so factor 5 was at a more aged

stage than factor 4 Factor 5 accounted for 677 of the mea-sured 2-BuONO2 levels and 197 of the measured carbonyllevels According to the abundances of alkyl nitrates and un-reactive alkanes we considered that the loadings of VOCspecies in this factor were related to secondary productionand background levels

Similarly with summertime measurements the relation-ships between the contribution of one factor to each NMHCspecies andkOH values for each NMHC species were ana-lyzed As shown in Fig 5 no significant correlation was ob-served for factor 1 and factor 2 while a positive correlationwas found for factor 3 and thus these three factors were con-sidered to be fresh emissions Significant negative correla-tions were identified for factor 4 and factor 5 indicating thatthese two factors were influenced by photochemical aging

33 Exploring the aged and fresh emission factors

In the last section we distinguished photochemically agedfactors from fresh factors To better understand the relation-ships between aged and fresh factors and the role of photo-chemical aging in PMF analysis we explored the relation-ships between VOC consumption and carbonyl formation inthe aged factors

For two factors representing different photochemicalstages of the same emissions the ratios of NMHC abun-dances in these two factors should follow the equation below(Yuan et al 2012)

[NMHC]aged

[NMHC]fresh= D times eminuskOHNMHC[OH]1t (4)

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3055

Fig 5 Relationships between the factor contributions to each NMHC species andkOH values of NMHC species in winter Each circlerepresents one compound Carbonyls are shown as solid triangles in the graph but they were not used for correlation analysis due to theirsecondary productionlowastlowastindicates a significant correlation at a confidence level of 001

where [NMHC]agedand [NMHC]fresh(ppb) refer to the abun-dances of NMHC species in the aged and fresh factor respec-tively D is a scaling factor which normalizes the NMHCabundances to unitykOHNMHC is the OH rate constant forthe NMHC [OH] is the average concentration of OH rad-ical and1t is the difference of reaction time between theaged and fresh factors In this study the OH exposure (iethe time integrate of [OH]) expressed by [OH]1t is calcu-lated as a whole

As discussed in Sect 32 for wintertime measurementsthe [NMHC]agedrefers to the abundances of NMHC in fac-tor 4 while the [NMHC]fresh refers to the sum of the abun-dances of NMHC in factors 1ndash3 For summertime measure-ments the [NMHC]agedrefers to the abundances of NMHC

in factor 3 while the [NMHC]fresh refers to the sum of theabundances of NMHC in factors 1 and 2 Figure 6 showedthe dependence of [NMHC]aged [NMHC] fresh (circles in thefigure) onkOH values with the lines being the fitted resultsfrom Eq (4) The values ofD and [OH]1t were estimated tobe 043 and 299times 1010 molecule cmminus3 s in winter and 126and 105times 1011 molecule cmminus3 s in summer respectivelyReferring to the calculated OH radical concentrations in thesummer in Beijing (Z Liu et al 2012) and the differencesof measured OH concentrations in Tokyo between summerand winter (Kanaya et al 2007) we assumed that the aver-age daytime concentration of OH in Beijing was 15times 106

and 6times 106 molecule cmminus3 in winter and summer thus the

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3056 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 6 Relationships between the ratios of VOCs abundances inthe aged emissions factors with those in fresh emission factors andkOH values for each VOC species Each data point represents onecompound The lines are the fitted results from Eq (4)

photochemical ages of the aged factor in winter and summerwere estimated to be 55 and 49 h respectively

The [carbonyl]aged [carbonyl]fresh ratios (Fig 6 solidtriangles) were significantly higher than the ratios of[NMHC]aged [NMHC] fresh because secondary productionenhanced the abundances of carbonyls in aged factors Asshown in Eq (5) the abundances of carbonyls in aged fac-tors can be separated into two parts The first part standsfor aged primary emissions and the second part stands forsecondary formation calculated based on the consumption ofother VOCs[carbonyl

]aged= D times

[carbonyl

]freshtimes eminuskOHcarbonyltimes[OH]1t (5)

+D times

sum([VOC]freshtimes

(1minus eminuskOHVOCtimes[OH]1t

)times YVOCcarbonyl

)wherekOHcarbonyl is the OH rate constant for the carbonylYVOCcarbonyl (ppbppb) refers to the carbonyl productionyield from the oxidation of particular VOC which canbe estimated using the Leeds master chemical mechanism(MCM v32 httpmcmleedsacukMCM) The estimationof YVOCcarbonyl is based on the following assumptions theremoval of VOC is governed by the reaction with OH radi-cals and the removal of alkyl peroxyl radicals (RO2) is gov-erned by the reaction with NO TheYVOCcarbonylvalues usedin this study are listed in Table 2 The VOC species that makeno contributions to the production of our focused carbonylswere not shown in this study The values ofD and [OH]1t

were fitted from Eq (4)The carbonyl abundances in aged factors were calculated

by Eq (5) and agreed well with the PMF-resolved abun-dances (Fig 7) In winter all data points were distributednear the 1 1 line In summer though the data points showedsome variability but still agreed within a factor of two Theagreement between the calculated and PMF-resolved car-bonyl abundances indicated that our understanding on thephotochemical aging of PMF factors was acceptable Thedifferences between the calculated and PMF-resolved car-bonyl abundances may be due to one or some of the fol-lowing reasons (1) the PMF analysis andYVOCcarbonyl es-timations have their own uncertainties (2) Only reactions of

VOCs with OH radicals are considered in our analysis (3)Further reactions of secondary carbonyls are not consideredin our analysis (4) Some VOC species are not applied in ourPMF analysis but they might be precursors of carbonyls ThePMF-analyzed VOC species accounted for 90 and 87 oftotal concentrations of all measured VOCs species in winterand summer respectively However the contribution of eachVOC to the formation of carbonyls was not only dependenton its concentration but also its ability to produce carbonyls

In this work we identified the PMF factors in summerand winter as several individual fresh emissions and sev-eral mixed aged emissions with different degrees of photo-chemical processing Beijing was a large city with extensiveand continuous local emissions and therefore our measuredair mass was in fact a mixture of fresh and aged plumesThe fresh emissions were separated by the PMF model ac-cording to the correlations among VOC species and the dif-ferent emission characteristics of these sources Howeverthe resolved factors by the PMF model cannot represent theemissions at different aged stages from individual sourcesAs discussed in this section the respective photochemicalages of the aged factors in winter and summer were respec-tively about 55 and 49 h and the factor representing sec-ondary production and background levels in winter should bemore aged We conjectured that VOCs from different sourceshave been well mixed during such long aging process and itwas hard to identify the aged contributions from individualsources Likewise previous studies also used averaged emis-sion ratios to characterize the emission and aging process ofurban VOC emissions (de Gouw et al 2005 Warneke et al2007 Yuan et al 2012 Borbon et al 2013) In this sec-tion we proved the reasonability of the PMF-derived agedfactors However one problem remained when apportioningreactive VOCs using the PMF model Actually the photo-chemical reaction of VOCs in the atmosphere is a continuousprocess but the PMF model is not able to describe such kindof process In this work several different aged stages wererecognized by the PMF model and the PMF model used thelinear combination of these stages to describe the continuousphotochemical process This might be an important issue forthe PMF analysis and the influence of such approximationrequired future researches

34 Primary and secondary sources of carbonyls

The calculated relative contributions of carbonyl sources ap-portioned by the PMF model in Beijing were shown in Fig 8Previous studies usually attributed all carbonyls in aged fac-tors as secondary formation As discussed in Sect 33 theabundances of carbonyls in aged factors could be separatedinto two parts that is the aged direct emissions and theproduction from VOC consumption and the former partshould be treated as primary emissions In this study theaged direct emission part was further separated into eachfresh factor considering their relative contributions For all

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

3054 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 4Profiles of the five factors resolved by the PMF model and the distributions of species among these factors in winter

components in NMHCs emissions from coal burning (Liu etal 2008) The ratio of benzene to toluene (231 ppb ppbminus1)

in this factor fell between the ratios measured by Liu etal (2008) for residential coal burning (181 ppb ppbminus1) andindustrial coal burning (262 ppb ppbminus1) but much higherthan that measured in the tunnel experiment (070 ppb ppbminus1)

(Liu et al 2008) therefore factor 1 was identified as coalburning C4ndashC5 NMHCs were important species from trafficrelated emissions and factor 3 showed similar characteris-tic with factor 1 in summer In addition factor 3 explained379 of the measured MTBE Therefore factor 3 was at-tributed to traffic related emissions

Factor 2 had high contributions to aromatics and showedsimilar characteristic with the resolved factor 2 for sum-mertime measurements Besides aromatics this factor con-tributed about 30 of measured long-chain alkanes (C7ndashC10 alkanes) The long-chain alkanes were also consid-ered to be important components of printing VOC emissions(Yuan et al 2010) Thus this factor was identified as indus-try and solvent usage

Both factor 4 and factor 5 had important contributionsto carbonyls but little contribution to reactive NMHCs andtherefore these two factors were inferred to possibly not rep-resent fresh emissions The major NMHC species in factor 4were C2-C5 alkanes ethene ethyne benzene and tolueneThese species could be emitted from all of the three primaryfactors discussed above and thus we considered this factor tobe aged emissions from these anthropogenic sources TheNMHCs in factor 5 were dominated by unreactive alkanessuch as ethane and propane so factor 5 was at a more aged

stage than factor 4 Factor 5 accounted for 677 of the mea-sured 2-BuONO2 levels and 197 of the measured carbonyllevels According to the abundances of alkyl nitrates and un-reactive alkanes we considered that the loadings of VOCspecies in this factor were related to secondary productionand background levels

Similarly with summertime measurements the relation-ships between the contribution of one factor to each NMHCspecies andkOH values for each NMHC species were ana-lyzed As shown in Fig 5 no significant correlation was ob-served for factor 1 and factor 2 while a positive correlationwas found for factor 3 and thus these three factors were con-sidered to be fresh emissions Significant negative correla-tions were identified for factor 4 and factor 5 indicating thatthese two factors were influenced by photochemical aging

33 Exploring the aged and fresh emission factors

In the last section we distinguished photochemically agedfactors from fresh factors To better understand the relation-ships between aged and fresh factors and the role of photo-chemical aging in PMF analysis we explored the relation-ships between VOC consumption and carbonyl formation inthe aged factors

For two factors representing different photochemicalstages of the same emissions the ratios of NMHC abun-dances in these two factors should follow the equation below(Yuan et al 2012)

[NMHC]aged

[NMHC]fresh= D times eminuskOHNMHC[OH]1t (4)

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3055

Fig 5 Relationships between the factor contributions to each NMHC species andkOH values of NMHC species in winter Each circlerepresents one compound Carbonyls are shown as solid triangles in the graph but they were not used for correlation analysis due to theirsecondary productionlowastlowastindicates a significant correlation at a confidence level of 001

where [NMHC]agedand [NMHC]fresh(ppb) refer to the abun-dances of NMHC species in the aged and fresh factor respec-tively D is a scaling factor which normalizes the NMHCabundances to unitykOHNMHC is the OH rate constant forthe NMHC [OH] is the average concentration of OH rad-ical and1t is the difference of reaction time between theaged and fresh factors In this study the OH exposure (iethe time integrate of [OH]) expressed by [OH]1t is calcu-lated as a whole

As discussed in Sect 32 for wintertime measurementsthe [NMHC]agedrefers to the abundances of NMHC in fac-tor 4 while the [NMHC]fresh refers to the sum of the abun-dances of NMHC in factors 1ndash3 For summertime measure-ments the [NMHC]agedrefers to the abundances of NMHC

in factor 3 while the [NMHC]fresh refers to the sum of theabundances of NMHC in factors 1 and 2 Figure 6 showedthe dependence of [NMHC]aged [NMHC] fresh (circles in thefigure) onkOH values with the lines being the fitted resultsfrom Eq (4) The values ofD and [OH]1t were estimated tobe 043 and 299times 1010 molecule cmminus3 s in winter and 126and 105times 1011 molecule cmminus3 s in summer respectivelyReferring to the calculated OH radical concentrations in thesummer in Beijing (Z Liu et al 2012) and the differencesof measured OH concentrations in Tokyo between summerand winter (Kanaya et al 2007) we assumed that the aver-age daytime concentration of OH in Beijing was 15times 106

and 6times 106 molecule cmminus3 in winter and summer thus the

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3056 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 6 Relationships between the ratios of VOCs abundances inthe aged emissions factors with those in fresh emission factors andkOH values for each VOC species Each data point represents onecompound The lines are the fitted results from Eq (4)

photochemical ages of the aged factor in winter and summerwere estimated to be 55 and 49 h respectively

The [carbonyl]aged [carbonyl]fresh ratios (Fig 6 solidtriangles) were significantly higher than the ratios of[NMHC]aged [NMHC] fresh because secondary productionenhanced the abundances of carbonyls in aged factors Asshown in Eq (5) the abundances of carbonyls in aged fac-tors can be separated into two parts The first part standsfor aged primary emissions and the second part stands forsecondary formation calculated based on the consumption ofother VOCs[carbonyl

]aged= D times

[carbonyl

]freshtimes eminuskOHcarbonyltimes[OH]1t (5)

+D times

sum([VOC]freshtimes

(1minus eminuskOHVOCtimes[OH]1t

)times YVOCcarbonyl

)wherekOHcarbonyl is the OH rate constant for the carbonylYVOCcarbonyl (ppbppb) refers to the carbonyl productionyield from the oxidation of particular VOC which canbe estimated using the Leeds master chemical mechanism(MCM v32 httpmcmleedsacukMCM) The estimationof YVOCcarbonyl is based on the following assumptions theremoval of VOC is governed by the reaction with OH radi-cals and the removal of alkyl peroxyl radicals (RO2) is gov-erned by the reaction with NO TheYVOCcarbonylvalues usedin this study are listed in Table 2 The VOC species that makeno contributions to the production of our focused carbonylswere not shown in this study The values ofD and [OH]1t

were fitted from Eq (4)The carbonyl abundances in aged factors were calculated

by Eq (5) and agreed well with the PMF-resolved abun-dances (Fig 7) In winter all data points were distributednear the 1 1 line In summer though the data points showedsome variability but still agreed within a factor of two Theagreement between the calculated and PMF-resolved car-bonyl abundances indicated that our understanding on thephotochemical aging of PMF factors was acceptable Thedifferences between the calculated and PMF-resolved car-bonyl abundances may be due to one or some of the fol-lowing reasons (1) the PMF analysis andYVOCcarbonyl es-timations have their own uncertainties (2) Only reactions of

VOCs with OH radicals are considered in our analysis (3)Further reactions of secondary carbonyls are not consideredin our analysis (4) Some VOC species are not applied in ourPMF analysis but they might be precursors of carbonyls ThePMF-analyzed VOC species accounted for 90 and 87 oftotal concentrations of all measured VOCs species in winterand summer respectively However the contribution of eachVOC to the formation of carbonyls was not only dependenton its concentration but also its ability to produce carbonyls

In this work we identified the PMF factors in summerand winter as several individual fresh emissions and sev-eral mixed aged emissions with different degrees of photo-chemical processing Beijing was a large city with extensiveand continuous local emissions and therefore our measuredair mass was in fact a mixture of fresh and aged plumesThe fresh emissions were separated by the PMF model ac-cording to the correlations among VOC species and the dif-ferent emission characteristics of these sources Howeverthe resolved factors by the PMF model cannot represent theemissions at different aged stages from individual sourcesAs discussed in this section the respective photochemicalages of the aged factors in winter and summer were respec-tively about 55 and 49 h and the factor representing sec-ondary production and background levels in winter should bemore aged We conjectured that VOCs from different sourceshave been well mixed during such long aging process and itwas hard to identify the aged contributions from individualsources Likewise previous studies also used averaged emis-sion ratios to characterize the emission and aging process ofurban VOC emissions (de Gouw et al 2005 Warneke et al2007 Yuan et al 2012 Borbon et al 2013) In this sec-tion we proved the reasonability of the PMF-derived agedfactors However one problem remained when apportioningreactive VOCs using the PMF model Actually the photo-chemical reaction of VOCs in the atmosphere is a continuousprocess but the PMF model is not able to describe such kindof process In this work several different aged stages wererecognized by the PMF model and the PMF model used thelinear combination of these stages to describe the continuousphotochemical process This might be an important issue forthe PMF analysis and the influence of such approximationrequired future researches

34 Primary and secondary sources of carbonyls

The calculated relative contributions of carbonyl sources ap-portioned by the PMF model in Beijing were shown in Fig 8Previous studies usually attributed all carbonyls in aged fac-tors as secondary formation As discussed in Sect 33 theabundances of carbonyls in aged factors could be separatedinto two parts that is the aged direct emissions and theproduction from VOC consumption and the former partshould be treated as primary emissions In this study theaged direct emission part was further separated into eachfresh factor considering their relative contributions For all

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3055

Fig 5 Relationships between the factor contributions to each NMHC species andkOH values of NMHC species in winter Each circlerepresents one compound Carbonyls are shown as solid triangles in the graph but they were not used for correlation analysis due to theirsecondary productionlowastlowastindicates a significant correlation at a confidence level of 001

where [NMHC]agedand [NMHC]fresh(ppb) refer to the abun-dances of NMHC species in the aged and fresh factor respec-tively D is a scaling factor which normalizes the NMHCabundances to unitykOHNMHC is the OH rate constant forthe NMHC [OH] is the average concentration of OH rad-ical and1t is the difference of reaction time between theaged and fresh factors In this study the OH exposure (iethe time integrate of [OH]) expressed by [OH]1t is calcu-lated as a whole

As discussed in Sect 32 for wintertime measurementsthe [NMHC]agedrefers to the abundances of NMHC in fac-tor 4 while the [NMHC]fresh refers to the sum of the abun-dances of NMHC in factors 1ndash3 For summertime measure-ments the [NMHC]agedrefers to the abundances of NMHC

in factor 3 while the [NMHC]fresh refers to the sum of theabundances of NMHC in factors 1 and 2 Figure 6 showedthe dependence of [NMHC]aged [NMHC] fresh (circles in thefigure) onkOH values with the lines being the fitted resultsfrom Eq (4) The values ofD and [OH]1t were estimated tobe 043 and 299times 1010 molecule cmminus3 s in winter and 126and 105times 1011 molecule cmminus3 s in summer respectivelyReferring to the calculated OH radical concentrations in thesummer in Beijing (Z Liu et al 2012) and the differencesof measured OH concentrations in Tokyo between summerand winter (Kanaya et al 2007) we assumed that the aver-age daytime concentration of OH in Beijing was 15times 106

and 6times 106 molecule cmminus3 in winter and summer thus the

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3056 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 6 Relationships between the ratios of VOCs abundances inthe aged emissions factors with those in fresh emission factors andkOH values for each VOC species Each data point represents onecompound The lines are the fitted results from Eq (4)

photochemical ages of the aged factor in winter and summerwere estimated to be 55 and 49 h respectively

The [carbonyl]aged [carbonyl]fresh ratios (Fig 6 solidtriangles) were significantly higher than the ratios of[NMHC]aged [NMHC] fresh because secondary productionenhanced the abundances of carbonyls in aged factors Asshown in Eq (5) the abundances of carbonyls in aged fac-tors can be separated into two parts The first part standsfor aged primary emissions and the second part stands forsecondary formation calculated based on the consumption ofother VOCs[carbonyl

]aged= D times

[carbonyl

]freshtimes eminuskOHcarbonyltimes[OH]1t (5)

+D times

sum([VOC]freshtimes

(1minus eminuskOHVOCtimes[OH]1t

)times YVOCcarbonyl

)wherekOHcarbonyl is the OH rate constant for the carbonylYVOCcarbonyl (ppbppb) refers to the carbonyl productionyield from the oxidation of particular VOC which canbe estimated using the Leeds master chemical mechanism(MCM v32 httpmcmleedsacukMCM) The estimationof YVOCcarbonyl is based on the following assumptions theremoval of VOC is governed by the reaction with OH radi-cals and the removal of alkyl peroxyl radicals (RO2) is gov-erned by the reaction with NO TheYVOCcarbonylvalues usedin this study are listed in Table 2 The VOC species that makeno contributions to the production of our focused carbonylswere not shown in this study The values ofD and [OH]1t

were fitted from Eq (4)The carbonyl abundances in aged factors were calculated

by Eq (5) and agreed well with the PMF-resolved abun-dances (Fig 7) In winter all data points were distributednear the 1 1 line In summer though the data points showedsome variability but still agreed within a factor of two Theagreement between the calculated and PMF-resolved car-bonyl abundances indicated that our understanding on thephotochemical aging of PMF factors was acceptable Thedifferences between the calculated and PMF-resolved car-bonyl abundances may be due to one or some of the fol-lowing reasons (1) the PMF analysis andYVOCcarbonyl es-timations have their own uncertainties (2) Only reactions of

VOCs with OH radicals are considered in our analysis (3)Further reactions of secondary carbonyls are not consideredin our analysis (4) Some VOC species are not applied in ourPMF analysis but they might be precursors of carbonyls ThePMF-analyzed VOC species accounted for 90 and 87 oftotal concentrations of all measured VOCs species in winterand summer respectively However the contribution of eachVOC to the formation of carbonyls was not only dependenton its concentration but also its ability to produce carbonyls

In this work we identified the PMF factors in summerand winter as several individual fresh emissions and sev-eral mixed aged emissions with different degrees of photo-chemical processing Beijing was a large city with extensiveand continuous local emissions and therefore our measuredair mass was in fact a mixture of fresh and aged plumesThe fresh emissions were separated by the PMF model ac-cording to the correlations among VOC species and the dif-ferent emission characteristics of these sources Howeverthe resolved factors by the PMF model cannot represent theemissions at different aged stages from individual sourcesAs discussed in this section the respective photochemicalages of the aged factors in winter and summer were respec-tively about 55 and 49 h and the factor representing sec-ondary production and background levels in winter should bemore aged We conjectured that VOCs from different sourceshave been well mixed during such long aging process and itwas hard to identify the aged contributions from individualsources Likewise previous studies also used averaged emis-sion ratios to characterize the emission and aging process ofurban VOC emissions (de Gouw et al 2005 Warneke et al2007 Yuan et al 2012 Borbon et al 2013) In this sec-tion we proved the reasonability of the PMF-derived agedfactors However one problem remained when apportioningreactive VOCs using the PMF model Actually the photo-chemical reaction of VOCs in the atmosphere is a continuousprocess but the PMF model is not able to describe such kindof process In this work several different aged stages wererecognized by the PMF model and the PMF model used thelinear combination of these stages to describe the continuousphotochemical process This might be an important issue forthe PMF analysis and the influence of such approximationrequired future researches

34 Primary and secondary sources of carbonyls

The calculated relative contributions of carbonyl sources ap-portioned by the PMF model in Beijing were shown in Fig 8Previous studies usually attributed all carbonyls in aged fac-tors as secondary formation As discussed in Sect 33 theabundances of carbonyls in aged factors could be separatedinto two parts that is the aged direct emissions and theproduction from VOC consumption and the former partshould be treated as primary emissions In this study theaged direct emission part was further separated into eachfresh factor considering their relative contributions For all

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

3056 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 6 Relationships between the ratios of VOCs abundances inthe aged emissions factors with those in fresh emission factors andkOH values for each VOC species Each data point represents onecompound The lines are the fitted results from Eq (4)

photochemical ages of the aged factor in winter and summerwere estimated to be 55 and 49 h respectively

The [carbonyl]aged [carbonyl]fresh ratios (Fig 6 solidtriangles) were significantly higher than the ratios of[NMHC]aged [NMHC] fresh because secondary productionenhanced the abundances of carbonyls in aged factors Asshown in Eq (5) the abundances of carbonyls in aged fac-tors can be separated into two parts The first part standsfor aged primary emissions and the second part stands forsecondary formation calculated based on the consumption ofother VOCs[carbonyl

]aged= D times

[carbonyl

]freshtimes eminuskOHcarbonyltimes[OH]1t (5)

+D times

sum([VOC]freshtimes

(1minus eminuskOHVOCtimes[OH]1t

)times YVOCcarbonyl

)wherekOHcarbonyl is the OH rate constant for the carbonylYVOCcarbonyl (ppbppb) refers to the carbonyl productionyield from the oxidation of particular VOC which canbe estimated using the Leeds master chemical mechanism(MCM v32 httpmcmleedsacukMCM) The estimationof YVOCcarbonyl is based on the following assumptions theremoval of VOC is governed by the reaction with OH radi-cals and the removal of alkyl peroxyl radicals (RO2) is gov-erned by the reaction with NO TheYVOCcarbonylvalues usedin this study are listed in Table 2 The VOC species that makeno contributions to the production of our focused carbonylswere not shown in this study The values ofD and [OH]1t

were fitted from Eq (4)The carbonyl abundances in aged factors were calculated

by Eq (5) and agreed well with the PMF-resolved abun-dances (Fig 7) In winter all data points were distributednear the 1 1 line In summer though the data points showedsome variability but still agreed within a factor of two Theagreement between the calculated and PMF-resolved car-bonyl abundances indicated that our understanding on thephotochemical aging of PMF factors was acceptable Thedifferences between the calculated and PMF-resolved car-bonyl abundances may be due to one or some of the fol-lowing reasons (1) the PMF analysis andYVOCcarbonyl es-timations have their own uncertainties (2) Only reactions of

VOCs with OH radicals are considered in our analysis (3)Further reactions of secondary carbonyls are not consideredin our analysis (4) Some VOC species are not applied in ourPMF analysis but they might be precursors of carbonyls ThePMF-analyzed VOC species accounted for 90 and 87 oftotal concentrations of all measured VOCs species in winterand summer respectively However the contribution of eachVOC to the formation of carbonyls was not only dependenton its concentration but also its ability to produce carbonyls

In this work we identified the PMF factors in summerand winter as several individual fresh emissions and sev-eral mixed aged emissions with different degrees of photo-chemical processing Beijing was a large city with extensiveand continuous local emissions and therefore our measuredair mass was in fact a mixture of fresh and aged plumesThe fresh emissions were separated by the PMF model ac-cording to the correlations among VOC species and the dif-ferent emission characteristics of these sources Howeverthe resolved factors by the PMF model cannot represent theemissions at different aged stages from individual sourcesAs discussed in this section the respective photochemicalages of the aged factors in winter and summer were respec-tively about 55 and 49 h and the factor representing sec-ondary production and background levels in winter should bemore aged We conjectured that VOCs from different sourceshave been well mixed during such long aging process and itwas hard to identify the aged contributions from individualsources Likewise previous studies also used averaged emis-sion ratios to characterize the emission and aging process ofurban VOC emissions (de Gouw et al 2005 Warneke et al2007 Yuan et al 2012 Borbon et al 2013) In this sec-tion we proved the reasonability of the PMF-derived agedfactors However one problem remained when apportioningreactive VOCs using the PMF model Actually the photo-chemical reaction of VOCs in the atmosphere is a continuousprocess but the PMF model is not able to describe such kindof process In this work several different aged stages wererecognized by the PMF model and the PMF model used thelinear combination of these stages to describe the continuousphotochemical process This might be an important issue forthe PMF analysis and the influence of such approximationrequired future researches

34 Primary and secondary sources of carbonyls

The calculated relative contributions of carbonyl sources ap-portioned by the PMF model in Beijing were shown in Fig 8Previous studies usually attributed all carbonyls in aged fac-tors as secondary formation As discussed in Sect 33 theabundances of carbonyls in aged factors could be separatedinto two parts that is the aged direct emissions and theproduction from VOC consumption and the former partshould be treated as primary emissions In this study theaged direct emission part was further separated into eachfresh factor considering their relative contributions For all

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3057

Table 2Production yields of carbonyls from VOCs (unit ppbppb)httpmcmleedsacukMCM

NMHC Formaldehyde Acetaldehyde Acetone Propanal Butanal

Ethane 0 0991 0 0 0Propane 0 0 0705 0259 0i-Butane 0773 0 0774 0 0n-Butane 0 0581 0 0 0024i-Pentane 0 0606 0611 0 0n-Pentane 0 0114 0 0116 022-dimethylbutane 0282 0289 0282 0 023-dimethylbutane 0 0 1638 0 02-methylpentane 0 0035 0195 0192 03-methylpentane 0 0491 0 0 0Ethene 16 0 0 0 0Propene 0979 0979 0 0 0trans-2-Butene 0 1918 0 0 01-Butene 0961 0 0 0961 0i-Butene 0988 0 0988 0 0cis-2-Butene 0 1918 0 0 013-Butadiene 073 0 0 0 01-Pentene 0941 0 0 0 0941cis-2-Pentene 0 0936 0 0936 0trans-2-Pentene 0 0936 0 0936 0Isoprene 0709 0 0 0 0Styrene 1 0 0 0 0Acetaldehyde 0999 0 0 0 0Acetone 1998 0 0 0 0Propanal 0 0991 0 0 0MEK 1390 0540 0 0 0n-Butanal 0013 0013 0 0832 0MACR 0978 0 0 0 0MVK 0994 0 0 0 0n-Pentanal 0 0 0 0 0154

Fig 7 Comparison of carbonyl abundances calculated based on VOCs consumptions with PMF-resolved abundances in aged factor Thegrey shaded area shows an agreement within a factor of two

the measured carbonyls secondary formation was the majorsource in both winter and summer with relative contributionsof 512 and 460 respectively In winter the three pri-mary emission sources that is coal burning industry andsolvent usage and traffic related emissions had similar con-

tributions to ambient carbonyl levels (153ndash169 ) In sum-mer 170 and 266 of carbonyls were attributed to in-dustry and solvent usage and traffic related emissions re-spectively In addition 104 of carbonyls in summer wereconsidered to be from biogenic emissions or their oxidation

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3058 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Fig 8The relative contributions of carbonyl sources in Beijing de-rived by the PMF model during winter and summer

The relative contributions of carbonyls from primary emis-sions and from secondary formation did not change muchbetween winter and summer which agreed with the diurnalvariation characteristics of carbonyl NMHC ratios as dis-cussed in Sect 31 However such results differed from thoseof Possanzini et al (2002) and Bakeas et al (2003) whofound great differences between carbonyl sources in winterand summer Hence we used the carbonyl production rate(Fp) to compare carbonyl formation between these two sea-sons TheFp can be calculated using Eq (6) (Lin et al2012)

Fp =

sum([VOC]i times kOHVOC times [OH] timesYVOCcarbonyl) (6)

whereYVOCcarbonyl is the carbonyl formation yield of VOC(Table 2) [OH] and [VOC]i are concentrations of OH radi-cals and each VOC species respectively The values of [OH]in winter and summer were assumed to be 15times 106 and6times 106 molecule cmminus3 as discussed in Sect 33 The cal-culatedFp was 23 and 36 ppb hminus1 for formaldehyde 09and 13 ppb hminus1 for acetaldehyde 03 and 04 ppb hminus1 foracetone in winter and summer respectively Though the OHconcentrations in summer were much higher than the val-ues in winter concentrations of NMHCs especially alkeneswere much higher in winter As a resultFp in summer wasonly 31ndash53 higher than that in winter For primary emis-sions though the levels of NMHCs were higher in winterthe emission ratios of carbonyl to ethyne in summer derivedfrom the PMF analysis were 2ndash5 times higher than those inwinter Therefore the contributions from primary emissionsappeared to be similar between winter and summer

Though primary and secondary sources had similar rela-tive contributions to ambient carbonyls in winter and sum-mer there were obvious differences for primary sources be-tween these two seasons Coal burning was an important pri-mary source for carbonyls in winter However it was notidentified in summer In Beijing coal was widely used forheating in winter while it was greatly reduced in summerThe emission strengths of biogenic VOCs favored the higherambient temperature and light intensity in summer (Guen-ther et al 2012) and thus the biogenic-related factor wasonly identified for summertime measurements

Formaldehyde is one of the most abundant and importantcarbonyls in the atmosphere Table 3 listed the source contri-butions of formaldehyde obtained in this study and comparedwith the results from previous studies in Beijing and otherregions The contributions of primary anthropogenic sourcesto formaldehyde were similar in the two seasons with thevalues of 289 in winter and 323 in summer Biogenicsources contributed 118 of the measured formaldehyde insummer but their contributions were not identified in win-ter The contribution of primary anthropogenic sources toformaldehyde obtained in this study was a bit higher thanthat reported by Yuan et al (2012) whereas the contribu-tion of biogenic sources was much lower than that reportedby Yuan et al (2012) Based on our PMF results in sum-mer the biogenic factor (ie factor 4) contributed less than30 of the measured MVK and MACR (products of iso-prene oxidation) and therefore this factor represented rela-tively fresh emissions from biogenic sources However thecontribution calculated by Yuan et al (2012) was for bothprimary and secondary biogenic sources and thus the pre-sented value was much higher Li et al (2010) calculatedthat 76 of formaldehyde was emitted from primary an-thropogenic sources which was much higher than the valuesfrom other studies There were two possible reasons (1) only2-days of data were analyzed in the study by Li et al (2010)and they might not represent the actual source characteris-tics of formaldehyde in Beijing and (2) the concentration ofozone the secondary marker used in the study might be in-fluenced by the titration of NO New York City and MexicoCity were two mega cities with large populations and largeamounts of vehicles which were similar to the situation ofBeijing The contribution of primary anthropogenic sourcesin Beijing was similar to that for New York City (Lin et al2012) and a bit lower than that for Mexico City (Garcia et al2006)

The contributions of primary anthropogenic sources to thelevels of other carbonyls were shown in Table 4 Primary an-thropogenic sources had higher contribution in winter dueto less secondary production and lower biogenic emissionsThe contributions from anthropogenic emissions were moreimportant for ketones than those for aldehydes with the val-ues larger than 50 in both winter and summer Comparedwith previous studies in Beijing the contribution from an-thropogenic emissions for acetaldehyde in 2011 was similarto the result of 2010 (Yuan et al 2012) but much higherthan 2005 (Liu et al 2009) This indicated a possible changein acetaldehyde sources between 2005 and 2011 Thoughthe concentrations of acetaldehyde decreased from 36 ppbin 2005 (Shao et al 2009) to 23 ppb in 2011 the contri-bution of anthropogenic emissions increased from 06 ppbto 10 ppb from 2005 to 2011 As the specific sources ofprimary acetaldehyde were not distinguished in 2005 thereasons for such change were unclear Further research ontemporal trends of acetaldehyde levels and sources wereneeded in the future For acetone propanal and butanal

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3059

Table 3Comparison of source apportionment results of ambient formaldehyde obtained in this study with previous studies

City season Primary anthropogenic Biogenic Secondary Background

Beijing wintera 289 ndash 711Beijing summera 323 118 559Beijing summerb 22 36 28 15Beijing summerc 76 ndash 18 5New York City summerd 30 ndash 70Mexico City springe 42 ndash gt 38 lt 21

Values are presented as percentages () of formaldehyde concentrationaThis studyb(Yuan et al 2012)c(Li et al 2010)d(Lin et al 2012)e(Garcia et al 2006)

Table 4Percentages of carbonyl concentrations from primary anthropogenic sources in this study and previous studies in Beijing

Carbonyl 2011ndash2012 2011 2005 2010Wintera Summera Summerb Summerc

Acetaldehyde 537 416 16 46Acetone 681 562 40 38Propanal 339 158 14 3MEK 710 622 47 80Butanal 490 155 8 1

Values are presented as percentages () of carbonyl concentration from primaryanthropogenic sourcesaThis studyb(Liu et al 2009)c(Yuan et al 2012)

our results calculated a bit higher contributions from anthro-pogenic emissions than that of previous studies

4 Summary

In this study online VOC measurements were conducted atthe urban site of Beijing during winter and summer The PMFreceptor model was used as a technique for source apportion-ment of VOCs In winter five factors were identified namelycoal burning industry and solvent usage traffic related emis-sions aged emissions and secondary formation plus back-ground levels In summer four factors were identified whichwere traffic related emissions industry and solvent usageaged emissions and biogenic emissions

Relationships between the factor contributions to eachNMHC species andkOH values for each NMHC specieswere analyzed to distinguish photochemically aged emis-sions from fresh emissions In both winter and summer afactor accounted for aged emissions was identified The agedfactor corresponded to the photochemical stages of mixedfresh emissions and the relationship between the aged andfresh factors was investigated The results indicated that theformation of carbonyls in the aged factor could be explainedby the consumption of VOCs and their carbonyl production

yields In winter besides the factor for aged emissions an-other factor was attributed to secondary formation and back-ground level This result demonstrated that the PMF analy-sis could be influenced by the photochemical processing offresh emissions Secondary carbonyls might be resolved tomore than one factor according to the different degrees ofphotochemical processing When using the PMF model forcarbonyl source apportionment the scientific understandingof each factor should be considered with care

Both PMF results and diurnal variations of car-bonyl NMHC ratios indicated that relative contributionsfrom primary and secondary sources in Beijing did notchange much between winter and summer Such a resultwas proved by calculating carbonyl production rates whichshowed that carbonyl production rates in summer were only31ndash53 higher than the rates in winter due to higher VOCsconcentrations in winter Secondary formation was the ma-jor source of carbonyls in Beijing with the relative contri-butions of 512 and 460 in winter and summer respec-tively Traffic related emissions industry and solvent usagewere the main sources of primary anthropogenic carbonylsin Beijing Coal burning was found to be an important an-thropogenic source of carbonyls in winter associated withthe heating activities During summertime biogenic sourcescontributed 104 of carbonyl levels For the three major

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3060 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

carbonyl species ndash formaldehyde acetaldehyde and ace-tone ndash the contribution of primary anthropogenic sourceswas 289 and 323 537 and 416 and 681 and562 in winter and summer respectively These results de-rived from source apportionment could be used to check thesource inventories of carbonyls which were in the process ofbeing updated in follow-up studies

Supplementary material related to this article isavailable online athttpwwwatmos-chem-physnet1430472014acp-14-3047-2014-supplementpdf

AcknowledgementsThis study was funded by the Natural ScienceFoundation for Outstanding Young Scholars (grant no 41125018)

Edited by X Tie

References

Atkinson R Baulch D L Cox R A Crowley J N Hamp-son R F Hynes R G Jenkin M E Rossi M J Troe Jand IUPAC Subcommittee Evaluated kinetic and photochemi-cal data for atmospheric chemistry Volume II ndash gas phase re-actions of organic species Atmos Chem Phys 6 3625ndash4055doi105194acp-6-3625-2006 2006

Bakeas E B Argyris D I and Siskos P A Carbonylcompounds in the urban environment of Athens GreeceChemosphere 52 805ndash813 doi101016s0045-6535(03)00257-1 2003

Ban-Weiss G A McLaughlin J P Harley R A Kean A JGrosjean E and Grosjean D Carbonyl and nitrogen diox-ide emissions from gasoline- and diesel-powered motor vehiclesEnviron Sci Technol 42 3944ndash3950 doi101021es80024872008

Bon D M Ulbrich I M de Gouw J A Warneke C KusterW C Alexander M L Baker A Beyersdorf A J Blake DFall R Jimenez J L Herndon S C Huey L G KnightonW B Ortega J Springston S and Vargas O Measurementsof volatile organic compounds at a suburban ground site (T1)in Mexico City during the MILAGRO 2006 campaign mea-surement comparison emission ratios and source attributionAtmos Chem Phys 11 2399ndash2421 doi105194acp-11-2399-2011 2011

Borbon A Gilman J B Kuster W C Grand N Chevaillier SColomb A Dolgorouky C Gros V Lopez M Sarda-EsteveR Holloway J Stutz J Petetin H McKeen S BeekmannM Warneke C Parrish D D and de Gouw J A Emissionratios of anthropogenic volatile organic compounds in northernmid-latitude megacities Observations versus emission invento-ries in Los Angeles and Paris J Geophys Res-Atmos 1182041ndash2057 doi101002jgrd50059 2013

Buzcu Guven B and Olaguer E P Ambient formaldehyde sourceattribution in Houston during TexAQS II and TRAMP AtmosEnviron 45 4272ndash4280 doi101016jatmosenv2011040792011

Carter W P L and Atkinson R Development and eval-uation of a detailed mechanism for the atmospheric re-actions of isoprene and NOx Int J Chem Kinet 28497ndash530 doi101002(sici)1097-4601(1996)287lt497aid-kin4gt30co2-q 1996

Ceroacuten R M Ceroacuten J G and Muriel M Diurnal and sea-sonal trends in carbonyl levels in a semi-urban coastal site inthe Gulf of Campeche Mexico Atmos Environ 41 63ndash71doi101016jatmosenv200608008 2007

Chang C C Wang J L Liu S C and Lung S C C Assess-ment of vehicular and non-vehicular contributions to hydrocar-bons using exclusive vehicular indicators Atmos Environ 406349ndash6361 doi101016jatmosenv200605043 2006

de Gouw J A Middlebrook A M Warneke C Goldan PD Kuster W C Roberts J M Fehsenfeld F C WorsnopD R Canagaratna M R Pszenny A A P Keene W CMarchewka M Bertman S B and Bates T S Budget of or-ganic carbon in a polluted atmosphere Results from the NewEngland Air Quality Study in 2002 J Geophys Res-Atmos110 D16305 doi1010292004jd005623 2005

De Smedt I Stavrakou T Muumlller J F van der A R J andVan Roozendael M Trend detection in satellite observationsof formaldehyde tropospheric columns Geophys Res Lett 37L18808 doi1010292010gl044245 2010

Finlayson-Pitts B J and Pitts J N Tropospheric air pol-lution Ozone airborne toxics polycyclic aromatic hy-drocarbons and particles Science 276 1045ndash1052doi101126science27653151045 1997

Friedfeld S Fraser M Ensor K Tribble S Rehle D LeleuxD and Tittel F Statistical analysis of primary and secondaryatmospheric formaldehyde Atmos Environ 36 4767ndash4775doi101016s1352-2310(02)00558-7 2002

Garcia A R Volkamer R Molina L T Molina M J Samuel-son J Mellqvist J Galle B Herndon S C and Kolb C ESeparation of emitted and photochemical formaldehyde in Mex-ico City using a statistical analysis and a new pair of gas-phasetracers Atmos Chem Phys 6 4545ndash4557 doi105194acp-6-4545-2006 2006

Goldstein A H and Schade G W Quantifying biogenic and an-thropogenic contributions to acetone mixing ratios in a rural en-vironment Atmos Environ 34 4997ndash5006 doi101016s1352-2310(00)00321-6 2000

Guenther A B Jiang X Heald C L Sakulyanontvittaya TDuhl T Emmons L K and Wang X The Model of Emissionsof Gases and Aerosols from Nature version 21 (MEGAN21) anextended and updated framework for modeling biogenic emis-sions Geosci Model Dev 5 1471ndash1492 doi105194gmd-5-1471-2012 2012

Guo H Ling Z H Cheung K Wang D W Simpson I J andBlake D R Acetone in the atmosphere of Hong Kong Abun-dance sources and photochemical precursors Atmos Environ65 80ndash88 doi101016jatmosenv201210027 2013

Kanaya Y Cao R Akimoto H Fukuda M Komazaki Y Yok-ouchi Y Koike M Tanimoto H Takegawa N and KondoY Urban photochemistry in central Tokyo 1 Observed andmodeled OH and HO2 radical concentrations during the winterand summer of 2004 J Geophys Res-Atmos 112 D21312doi1010292007jd008670 2007

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3059

Table 3Comparison of source apportionment results of ambient formaldehyde obtained in this study with previous studies

City season Primary anthropogenic Biogenic Secondary Background

Beijing wintera 289 ndash 711Beijing summera 323 118 559Beijing summerb 22 36 28 15Beijing summerc 76 ndash 18 5New York City summerd 30 ndash 70Mexico City springe 42 ndash gt 38 lt 21

Values are presented as percentages () of formaldehyde concentrationaThis studyb(Yuan et al 2012)c(Li et al 2010)d(Lin et al 2012)e(Garcia et al 2006)

Table 4Percentages of carbonyl concentrations from primary anthropogenic sources in this study and previous studies in Beijing

Carbonyl 2011ndash2012 2011 2005 2010Wintera Summera Summerb Summerc

Acetaldehyde 537 416 16 46Acetone 681 562 40 38Propanal 339 158 14 3MEK 710 622 47 80Butanal 490 155 8 1

Values are presented as percentages () of carbonyl concentration from primaryanthropogenic sourcesaThis studyb(Liu et al 2009)c(Yuan et al 2012)

our results calculated a bit higher contributions from anthro-pogenic emissions than that of previous studies

4 Summary

In this study online VOC measurements were conducted atthe urban site of Beijing during winter and summer The PMFreceptor model was used as a technique for source apportion-ment of VOCs In winter five factors were identified namelycoal burning industry and solvent usage traffic related emis-sions aged emissions and secondary formation plus back-ground levels In summer four factors were identified whichwere traffic related emissions industry and solvent usageaged emissions and biogenic emissions

Relationships between the factor contributions to eachNMHC species andkOH values for each NMHC specieswere analyzed to distinguish photochemically aged emis-sions from fresh emissions In both winter and summer afactor accounted for aged emissions was identified The agedfactor corresponded to the photochemical stages of mixedfresh emissions and the relationship between the aged andfresh factors was investigated The results indicated that theformation of carbonyls in the aged factor could be explainedby the consumption of VOCs and their carbonyl production

yields In winter besides the factor for aged emissions an-other factor was attributed to secondary formation and back-ground level This result demonstrated that the PMF analy-sis could be influenced by the photochemical processing offresh emissions Secondary carbonyls might be resolved tomore than one factor according to the different degrees ofphotochemical processing When using the PMF model forcarbonyl source apportionment the scientific understandingof each factor should be considered with care

Both PMF results and diurnal variations of car-bonyl NMHC ratios indicated that relative contributionsfrom primary and secondary sources in Beijing did notchange much between winter and summer Such a resultwas proved by calculating carbonyl production rates whichshowed that carbonyl production rates in summer were only31ndash53 higher than the rates in winter due to higher VOCsconcentrations in winter Secondary formation was the ma-jor source of carbonyls in Beijing with the relative contri-butions of 512 and 460 in winter and summer respec-tively Traffic related emissions industry and solvent usagewere the main sources of primary anthropogenic carbonylsin Beijing Coal burning was found to be an important an-thropogenic source of carbonyls in winter associated withthe heating activities During summertime biogenic sourcescontributed 104 of carbonyl levels For the three major

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3060 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

carbonyl species ndash formaldehyde acetaldehyde and ace-tone ndash the contribution of primary anthropogenic sourceswas 289 and 323 537 and 416 and 681 and562 in winter and summer respectively These results de-rived from source apportionment could be used to check thesource inventories of carbonyls which were in the process ofbeing updated in follow-up studies

Supplementary material related to this article isavailable online athttpwwwatmos-chem-physnet1430472014acp-14-3047-2014-supplementpdf

AcknowledgementsThis study was funded by the Natural ScienceFoundation for Outstanding Young Scholars (grant no 41125018)

Edited by X Tie

References

Atkinson R Baulch D L Cox R A Crowley J N Hamp-son R F Hynes R G Jenkin M E Rossi M J Troe Jand IUPAC Subcommittee Evaluated kinetic and photochemi-cal data for atmospheric chemistry Volume II ndash gas phase re-actions of organic species Atmos Chem Phys 6 3625ndash4055doi105194acp-6-3625-2006 2006

Bakeas E B Argyris D I and Siskos P A Carbonylcompounds in the urban environment of Athens GreeceChemosphere 52 805ndash813 doi101016s0045-6535(03)00257-1 2003

Ban-Weiss G A McLaughlin J P Harley R A Kean A JGrosjean E and Grosjean D Carbonyl and nitrogen diox-ide emissions from gasoline- and diesel-powered motor vehiclesEnviron Sci Technol 42 3944ndash3950 doi101021es80024872008

Bon D M Ulbrich I M de Gouw J A Warneke C KusterW C Alexander M L Baker A Beyersdorf A J Blake DFall R Jimenez J L Herndon S C Huey L G KnightonW B Ortega J Springston S and Vargas O Measurementsof volatile organic compounds at a suburban ground site (T1)in Mexico City during the MILAGRO 2006 campaign mea-surement comparison emission ratios and source attributionAtmos Chem Phys 11 2399ndash2421 doi105194acp-11-2399-2011 2011

Borbon A Gilman J B Kuster W C Grand N Chevaillier SColomb A Dolgorouky C Gros V Lopez M Sarda-EsteveR Holloway J Stutz J Petetin H McKeen S BeekmannM Warneke C Parrish D D and de Gouw J A Emissionratios of anthropogenic volatile organic compounds in northernmid-latitude megacities Observations versus emission invento-ries in Los Angeles and Paris J Geophys Res-Atmos 1182041ndash2057 doi101002jgrd50059 2013

Buzcu Guven B and Olaguer E P Ambient formaldehyde sourceattribution in Houston during TexAQS II and TRAMP AtmosEnviron 45 4272ndash4280 doi101016jatmosenv2011040792011

Carter W P L and Atkinson R Development and eval-uation of a detailed mechanism for the atmospheric re-actions of isoprene and NOx Int J Chem Kinet 28497ndash530 doi101002(sici)1097-4601(1996)287lt497aid-kin4gt30co2-q 1996

Ceroacuten R M Ceroacuten J G and Muriel M Diurnal and sea-sonal trends in carbonyl levels in a semi-urban coastal site inthe Gulf of Campeche Mexico Atmos Environ 41 63ndash71doi101016jatmosenv200608008 2007

Chang C C Wang J L Liu S C and Lung S C C Assess-ment of vehicular and non-vehicular contributions to hydrocar-bons using exclusive vehicular indicators Atmos Environ 406349ndash6361 doi101016jatmosenv200605043 2006

de Gouw J A Middlebrook A M Warneke C Goldan PD Kuster W C Roberts J M Fehsenfeld F C WorsnopD R Canagaratna M R Pszenny A A P Keene W CMarchewka M Bertman S B and Bates T S Budget of or-ganic carbon in a polluted atmosphere Results from the NewEngland Air Quality Study in 2002 J Geophys Res-Atmos110 D16305 doi1010292004jd005623 2005

De Smedt I Stavrakou T Muumlller J F van der A R J andVan Roozendael M Trend detection in satellite observationsof formaldehyde tropospheric columns Geophys Res Lett 37L18808 doi1010292010gl044245 2010

Finlayson-Pitts B J and Pitts J N Tropospheric air pol-lution Ozone airborne toxics polycyclic aromatic hy-drocarbons and particles Science 276 1045ndash1052doi101126science27653151045 1997

Friedfeld S Fraser M Ensor K Tribble S Rehle D LeleuxD and Tittel F Statistical analysis of primary and secondaryatmospheric formaldehyde Atmos Environ 36 4767ndash4775doi101016s1352-2310(02)00558-7 2002

Garcia A R Volkamer R Molina L T Molina M J Samuel-son J Mellqvist J Galle B Herndon S C and Kolb C ESeparation of emitted and photochemical formaldehyde in Mex-ico City using a statistical analysis and a new pair of gas-phasetracers Atmos Chem Phys 6 4545ndash4557 doi105194acp-6-4545-2006 2006

Goldstein A H and Schade G W Quantifying biogenic and an-thropogenic contributions to acetone mixing ratios in a rural en-vironment Atmos Environ 34 4997ndash5006 doi101016s1352-2310(00)00321-6 2000

Guenther A B Jiang X Heald C L Sakulyanontvittaya TDuhl T Emmons L K and Wang X The Model of Emissionsof Gases and Aerosols from Nature version 21 (MEGAN21) anextended and updated framework for modeling biogenic emis-sions Geosci Model Dev 5 1471ndash1492 doi105194gmd-5-1471-2012 2012

Guo H Ling Z H Cheung K Wang D W Simpson I J andBlake D R Acetone in the atmosphere of Hong Kong Abun-dance sources and photochemical precursors Atmos Environ65 80ndash88 doi101016jatmosenv201210027 2013

Kanaya Y Cao R Akimoto H Fukuda M Komazaki Y Yok-ouchi Y Koike M Tanimoto H Takegawa N and KondoY Urban photochemistry in central Tokyo 1 Observed andmodeled OH and HO2 radical concentrations during the winterand summer of 2004 J Geophys Res-Atmos 112 D21312doi1010292007jd008670 2007

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

3060 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

carbonyl species ndash formaldehyde acetaldehyde and ace-tone ndash the contribution of primary anthropogenic sourceswas 289 and 323 537 and 416 and 681 and562 in winter and summer respectively These results de-rived from source apportionment could be used to check thesource inventories of carbonyls which were in the process ofbeing updated in follow-up studies

Supplementary material related to this article isavailable online athttpwwwatmos-chem-physnet1430472014acp-14-3047-2014-supplementpdf

AcknowledgementsThis study was funded by the Natural ScienceFoundation for Outstanding Young Scholars (grant no 41125018)

Edited by X Tie

References

Atkinson R Baulch D L Cox R A Crowley J N Hamp-son R F Hynes R G Jenkin M E Rossi M J Troe Jand IUPAC Subcommittee Evaluated kinetic and photochemi-cal data for atmospheric chemistry Volume II ndash gas phase re-actions of organic species Atmos Chem Phys 6 3625ndash4055doi105194acp-6-3625-2006 2006

Bakeas E B Argyris D I and Siskos P A Carbonylcompounds in the urban environment of Athens GreeceChemosphere 52 805ndash813 doi101016s0045-6535(03)00257-1 2003

Ban-Weiss G A McLaughlin J P Harley R A Kean A JGrosjean E and Grosjean D Carbonyl and nitrogen diox-ide emissions from gasoline- and diesel-powered motor vehiclesEnviron Sci Technol 42 3944ndash3950 doi101021es80024872008

Bon D M Ulbrich I M de Gouw J A Warneke C KusterW C Alexander M L Baker A Beyersdorf A J Blake DFall R Jimenez J L Herndon S C Huey L G KnightonW B Ortega J Springston S and Vargas O Measurementsof volatile organic compounds at a suburban ground site (T1)in Mexico City during the MILAGRO 2006 campaign mea-surement comparison emission ratios and source attributionAtmos Chem Phys 11 2399ndash2421 doi105194acp-11-2399-2011 2011

Borbon A Gilman J B Kuster W C Grand N Chevaillier SColomb A Dolgorouky C Gros V Lopez M Sarda-EsteveR Holloway J Stutz J Petetin H McKeen S BeekmannM Warneke C Parrish D D and de Gouw J A Emissionratios of anthropogenic volatile organic compounds in northernmid-latitude megacities Observations versus emission invento-ries in Los Angeles and Paris J Geophys Res-Atmos 1182041ndash2057 doi101002jgrd50059 2013

Buzcu Guven B and Olaguer E P Ambient formaldehyde sourceattribution in Houston during TexAQS II and TRAMP AtmosEnviron 45 4272ndash4280 doi101016jatmosenv2011040792011

Carter W P L and Atkinson R Development and eval-uation of a detailed mechanism for the atmospheric re-actions of isoprene and NOx Int J Chem Kinet 28497ndash530 doi101002(sici)1097-4601(1996)287lt497aid-kin4gt30co2-q 1996

Ceroacuten R M Ceroacuten J G and Muriel M Diurnal and sea-sonal trends in carbonyl levels in a semi-urban coastal site inthe Gulf of Campeche Mexico Atmos Environ 41 63ndash71doi101016jatmosenv200608008 2007

Chang C C Wang J L Liu S C and Lung S C C Assess-ment of vehicular and non-vehicular contributions to hydrocar-bons using exclusive vehicular indicators Atmos Environ 406349ndash6361 doi101016jatmosenv200605043 2006

de Gouw J A Middlebrook A M Warneke C Goldan PD Kuster W C Roberts J M Fehsenfeld F C WorsnopD R Canagaratna M R Pszenny A A P Keene W CMarchewka M Bertman S B and Bates T S Budget of or-ganic carbon in a polluted atmosphere Results from the NewEngland Air Quality Study in 2002 J Geophys Res-Atmos110 D16305 doi1010292004jd005623 2005

De Smedt I Stavrakou T Muumlller J F van der A R J andVan Roozendael M Trend detection in satellite observationsof formaldehyde tropospheric columns Geophys Res Lett 37L18808 doi1010292010gl044245 2010

Finlayson-Pitts B J and Pitts J N Tropospheric air pol-lution Ozone airborne toxics polycyclic aromatic hy-drocarbons and particles Science 276 1045ndash1052doi101126science27653151045 1997

Friedfeld S Fraser M Ensor K Tribble S Rehle D LeleuxD and Tittel F Statistical analysis of primary and secondaryatmospheric formaldehyde Atmos Environ 36 4767ndash4775doi101016s1352-2310(02)00558-7 2002

Garcia A R Volkamer R Molina L T Molina M J Samuel-son J Mellqvist J Galle B Herndon S C and Kolb C ESeparation of emitted and photochemical formaldehyde in Mex-ico City using a statistical analysis and a new pair of gas-phasetracers Atmos Chem Phys 6 4545ndash4557 doi105194acp-6-4545-2006 2006

Goldstein A H and Schade G W Quantifying biogenic and an-thropogenic contributions to acetone mixing ratios in a rural en-vironment Atmos Environ 34 4997ndash5006 doi101016s1352-2310(00)00321-6 2000

Guenther A B Jiang X Heald C L Sakulyanontvittaya TDuhl T Emmons L K and Wang X The Model of Emissionsof Gases and Aerosols from Nature version 21 (MEGAN21) anextended and updated framework for modeling biogenic emis-sions Geosci Model Dev 5 1471ndash1492 doi105194gmd-5-1471-2012 2012

Guo H Ling Z H Cheung K Wang D W Simpson I J andBlake D R Acetone in the atmosphere of Hong Kong Abun-dance sources and photochemical precursors Atmos Environ65 80ndash88 doi101016jatmosenv201210027 2013

Kanaya Y Cao R Akimoto H Fukuda M Komazaki Y Yok-ouchi Y Koike M Tanimoto H Takegawa N and KondoY Urban photochemistry in central Tokyo 1 Observed andmodeled OH and HO2 radical concentrations during the winterand summer of 2004 J Geophys Res-Atmos 112 D21312doi1010292007jd008670 2007

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014

W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds 3061

Kim K-H Hong Y-J Pal R Jeon E-C Koo Y-S and Sun-woo Y Investigation of carbonyl compounds in air from var-ious industrial emission sources Chemosphere 70 807ndash820doi101016jchemosphere200707025 2008

Lary D J and Shallcross D E Central role of carbonyl com-pounds in atmospheric chemistry J Geophys Res-Atmos 10519771ndash19778 doi1010291999jd901184 2000

Li Y Shao M Lu S Chang C-C and Dasgupta P KVariations and sources of ambient formaldehyde for the 2008Beijing Olympic games Atmos Environ 44 2632ndash2639doi101016jatmosenv201003045 2010

Liggio J Li S M and McLaren R Reactive uptake of glyoxalby particulate matter J Geophys Res-Atmos 110 D10304doi1010292004jd005113 2005

Lin Y C Schwab J J Demerjian K L Bae M-S Chen W-N Sun Y Zhang Q Hung H-M and Perry J Summertimeformaldehyde observations in New York City Ambient levelssources and its contribution to HOx radicals J Geophys Res-Atmos 117 D08305 doi1010292011jd016504 2012

Liu H Wang X M Pang J M and He K B Feasibility anddifficulties of Chinarsquos new air quality standard compliance PRDcase of PM25 and ozone from 2010 to 2025 Atmos ChemPhys 13 12013ndash12027 doi105194acp-13-12013-2013 2013

Liu Y Shao M Fu L Lu S Zeng L and Tang DSource profiles of volatile organic compounds (VOCs) mea-sured in China Part I Atmos Environ 42 6247ndash6260doi101016jatmosenv200801070 2008

Liu Y Shao M Kuster W C Goldan P D Li X H Lu S Hand De Gouw J A Source identification of reactive hydrocar-bons and oxygenated VOCs in the summertime in Beijing Envi-ron Sci Technol 43 75ndash81 doi101021es801716n 2009

Liu Z Wang Y Gu D Zhao C Huey L G Stickel RLiao J Shao M Zhu T Zeng L Amoroso A CostabileF Chang C-C and Liu S-C Summertime photochemistryduring CAREBeijing-2007 ROx budgets and O3 formation At-mos Chem Phys 12 7737ndash7752 doi105194acp-12-7737-2012 2012

Liu W-T Hsieh H-C Chen S-P Chang J S Lin N-H Chang C-C and Wang J-L Diagnosis of air qualitythrough observation and modeling of volatile organic com-pounds (VOCs) as pollution tracers Atmos Environ 55 56ndash63doi101016jatmosenv201203017 2012

Paatero P and Tapper U Positive matrix factorization ndasha nonnegative factor model with optimal utilization oferror-estimates of data values Environmetrics 5 111ndash126doi101002env3170050203 1994

Pang X and Mu Y Seasonal and diurnal variations of carbonylcompounds in Beijing ambient air Atmos Environ 40 6313ndash6320 doi101016jatmosenv200605044 2006

Parrish D D Ryerson T B Mellqvist J Johansson J FriedA Richter D Walega J G Washenfelder R A de Gouw JA Peischl J Aikin K C McKeen S A Frost G J Fehsen-feld F C and Herndon S C Primary and secondary sourcesof formaldehyde in urban atmospheres Houston Texas regionAtmos Chem Phys 12 3273ndash3288 doi105194acp-12-3273-2012 2012

Possanzini M Palo V D and Cecinato A Sources and pho-todecomposition of formaldehyde and acetaldehyde in Rome am-

bient air Atmos Environ 36 3195ndash3201 doi101016s1352-2310(02)00192-9 2002

Rappengluumlck B Dasgupta P K Leuchner M Li Q and LukeW Formaldehyde and its relation to CO PAN and SO2 inthe Houston-Galveston airshed Atmos Chem Phys 10 2413ndash2424 doi105194acp-10-2413-2010 2010

Shao M Tang X Zhang Y and Li W City clusters in China airand surface water pollution Front Ecol Environ 4 353ndash361doi1018901540-9295(2006)004[0353CCICAA]20CO22006

Shao M Lu S Liu Y Xie X Chang C Huang S and ChenZ Volatile organic compounds measured in summer in Beijingand their role in groundlevel ozone formation J Geophys Res-Atmos 114 D00G06 doi1010292008jd010863 2009

Simpson I J Blake N J Blake D R Atlas E FlockeF Crawford J H Fuelberg H E Kiley C M MeinardiS and Rowland F S Photochemical production and evolu-tion of selected C2ndashC5 alkyl nitrates in tropospheric air influ-enced by Asian outflow J Geophys Res-Atmos 108 8808doi1010292002jd002830 2003

Singh H B Kanakidou M Crutzen P J and Jacob D J Highconcentrations and photochemical fate of oxygenated hydrocar-bons in the global troposphere Nature 378 50ndash54 1995

Situ S Guenther A Wang X Jiang X Turnipseed A WuZ Bai J and Wang X Impacts of seasonal and regionalvariability in biogenic VOC emissions on surface ozone in thePearl River delta region China Atmos Chem Phys 13 11803ndash11817 doi105194acp-13-11803-2013 2013

Song Y Dai W Shao M Liu Y Lu S H Kuster W andGoldan P Comparison of receptor models for source apportion-ment of volatile organic compounds in Beijing China EnvironPollut 156 174ndash183 doi101016jenvpol200712014 2008

Vlasenko A Macdonald A M Sjostedt S J and Abbatt J P DFormaldehyde measurements by Proton transfer reaction ndash MassSpectrometry (PTR-MS) correction for humidity effects At-mos Meas Tech 3 1055ndash1062 doi105194amt-3-1055-20102010

Wang B Shao M Lu S H Yuan B Zhao Y Wang M ZhangS Q and Wu D Variation of ambient non-methane hydrocar-bons in Beijing city in summer 2008 Atmos Chem Phys 105911ndash5923 doi105194acp-10-5911-2010 2010

Wang M Shao M Chen W Yuan B Lu S Zhang QZeng L and Wang Q Validation of emission inventories bymeasurements of ambient volatile organic compounds in Bei-jing China Atmos Chem Phys Discuss 13 26933ndash26979doi105194acpd-13-26933-2013 2013

Wang Q Geng C Lu S Chen W and Shao M Emissionfactors of gaseous carbonaceous species from residential com-bustion of coal and crop residue briquettes Front Environ SciEng 7 66ndash76 doi101007s11783-012-0428-5 2013

Wang Y Konopka P Liu Y Chen H Muumlller R PloumlgerF Riese M Cai Z and Luuml D Tropospheric ozonetrend over Beijing from 2002ndash2010 ozonesonde measurementsand modeling analysis Atmos Chem Phys 12 8389ndash8399doi105194acp-12-8389-2012 2012a

Wang Y S Ren X Y Ji D S Zhang J Q Sun J and Wu FK Characterization of volatile organic compounds in the urbanarea of Beijing from 2000 to 2007 J Environ Sci-China 2495ndash101 doi101016s1001-0742(11)60732-8 2012b

wwwatmos-chem-physnet1430472014 Atmos Chem Phys 14 3047ndash3062 2014

3062 W T Chen et al Understanding primary and secondary sources of ambient carbonyl compounds

Warneke C McKeen S A de Gouw J A Goldan P DKuster W C Holloway J S Williams E J Lerner B MParrish D D Trainer M Fehsenfeld F C Kato S At-las E L Baker A and Blake D R Determination of urbanvolatile organic compound emission ratios and comparison withan emissions database J Geophys Res-Atmos 112 D10S47doi1010292006jd007930 2007

Warneke C Veres P Holloway J S Stutz J Tsai C AlvarezS Rappenglueck B Fehsenfeld F C Graus M GilmanJ B and de Gouw J A Airborne formaldehyde measure-ments using PTR-MS calibration humidity dependence inter-comparison and initial results Atmos Meas Tech 4 2345ndash2358 doi105194amt-4-2345-2011 2011

Winters A J Adams M A Bleby T M Rennenberg HSteigner D Steinbrecher R and Kreuzwieser J Emissions ofisoprene monoterpene and short-chained carbonyl compoundsfrom Eucalyptus spp in southern Australia Atmos Environ 433035ndash3043 doi101016jatmosenv200903026 2009

Xu Z Liu J Zhang Y Liang P and Mu Y Ambi-ent levels of atmospheric carbonyls in Beijing during the2008 Olympic Games J Environ Sci-China 22 1348ndash1356doi101016s1001-0742(09)60261-8 2010

Yuan B Shao M Lu S and Wang B Source pro-files of volatile organic compounds associated with solventuse in Beijing China Atmos Environ 44 1919ndash1926doi101016jatmosenv201002014 2010

Yuan B Shao M de Gouw J Parrish D D Lu S WangM Zeng L Zhang Q Song Y Zhang J and Hu MVolatile organic compounds (VOCs) in urban air How chem-istry affects the interpretation of positive matrix factoriza-tion (PMF) analysis J Geophys Res-Atmos 117 D24302doi1010292012jd018236 2012

Yuan B Hu W W Shao M Wang M Chen W T Lu S HZeng L M and Hu M VOC emissions evolutions and con-tributions to SOA formation at a receptor site in eastern ChinaAtmos Chem Phys 13 8815ndash8832 doi105194acp-13-8815-2013 2013

Zhang J and Smith K R Emissions of carbonyl compounds fromvarious cookstoves in China Environ Sci Technol 33 2311ndash2320 doi101021es9812406 1999

Zhang Y J Mu Y J Liu J F and Mellouki A Levelssources and health risks of carbonyls and BTEX in the ambi-ent air of Beijing China J Environ Sci-China 24 124ndash130doi101016s1001-0742(11)60735-3 2012

Zheng J Garzoacuten J P Huertas M E Zhang R LevyM Ma Y Huertas J I Jardoacuten R T Ruiacutez L GTan H and Molina L T Volatile organic compounds inTijuana during the Cal-Mex 2010 campaign Measurementsand source apportionment Atmos Environ 70 521ndash531doi101016jatmosenv201211030 2013

Atmos Chem Phys 14 3047ndash3062 2014 wwwatmos-chem-physnet1430472014