The Comparison Between Thermal-Optical Transmittance Elemental Carbon Andaethalometer Black Carbon Measuredat Multiple Monitoring Sites Cheol-Heon Jeonga,B, Philip K

ARTICLE IN PRESS

Atmospheric Environment 38 (2004) 5193–5204 www.elsevier.com/locate/atmosenv

The comparison between thermal-optical transmittance elemental carbon andAethalometer black carbon measuredat multiple monitoring sites Cheol-Heon Jeonga,b, Philip K. Hopkea,b,Ã, Eugene Kima,b, Doh-Won Leea,b

aDepartment of Civil Environmental Engineering, Clarkson University, Box 5708, Potsdam 13699-5708, NY, USA bDepartment of Chemical Engineering, Center for Air Resources Engineering and Science, Clarkson University, Box 5708, Potsdam 13699-5708, NY, USA

Received8 August 2003; receivedin revisedform 1 February 2004; accepted11 February 2004

Abstract

Continuous andsemi-continuous measurements of organic carbon (OC), elemental carbon (EC) andPM 2:5 were performed during the summer of 2002 in Rochester, NY and Philadelphia, PA. During the study period in Philadelphia, high concentrations of woodsmoke from a Canadianforest fire were transportedto the monitoring site. Two-hour integratedthermal EC (EC)/optical EC ðBCSÞ using a Sunset Lab OC/EC analyzer andcontinuous Aethalometer optical EC ðBCAÞ using a two-wavelength Aethalometer were compared in various environments. The weekdays diurnal average for EC peaked during the morning rush-hour and was much higher than the value during weekends, whereas the OC results showed no diurnal variation of OC during either weekdays or weekends at both sites. The diurnal variations of BCA were also strongly correlatedwith the rush-hour traffic. The correlation coefficient between EC and 2 2 BCA in Rochester ðr ¼ 0:84Þ was higher than in Philadelphia ðr ¼ 0:60Þ while the BCA/EC slopes were 3.3 and2.7 in Rochester and Philadelphia, respectively. The difference in the slopes indicates that the specific attenuation cross- section used for the optical carbon analysis depends on the physical and chemical characteristics of elemental carbon. During the Canadian forest fire, the BCA/EC slope dramatically dropped to 0.4 with a correlation coefficient of 0.60. The decrease of the proportionality between EC and BCA demonstrates the variability of the absorption coefficient. The level of UV absorbing organic compounds significantly increased during the fire aerosol episode suggesting the presence of abundant organic compounds in the forest fire smoke particles. r 2004 Elsevier Ltd. All rights reserved.

Keywords: Organic carbon; Elemental carbon; Black carbon; PM2:5; Sunset Lab OC/EC analyzer; Aethalometer; Speciﬁc attenuation cross-section; Canadian forest ﬁre

1. Introduction Carbonaceous particles are composedof two main fractions, organic carbon (OC) that is semivolatile and Carbonaceous aerosol is a complex mixture andone elemental carbon (EC) that is refractory andstrongly of the more abundant components in urban PM2:5. light absorbing. EC is formedby incomplete combustion andis nonvolatile. Gas to particle conversion of ÃCorresponding author. Fax: +1-315-268-4410. hydrocarbon gases, exhaust from combustion process, E-mail address: [email protected] (P.K. Hopke). boilers, residential wood burning, meat cooking, and

5194 C.-H. Jeong et al. / Atmospheric Environment 38 (2004) 5193–5204 roaddust( Pandis et al., 1992; Rogge et al., 1993a,b, different (Jennings andPinnick, 1980; Liousse et al., 1997, 1998) are known to be major sources of OC in 1993; Lavanchy et al., 1999; Hitzenberger et al., 1999). PM2:5. Also, emissions from diesel heavy trucks and Liousse et al. (1993) foundthat EC may exist both in diesel automobiles are known to be the main sources of external mixtures in which EC can constitute the shell or EC in urban areas (Gray andCass, 1998 ). Light the core of aerosols andin internal mixtures. The absorption by PM2:5 particles is also primarily due to various chemical andphysical properties of EC may also the carbonaceous species in the particles. The mass cause changes in the optical BC concentrations mea- relatedto light absorption is typically referredto as suredusing the optical attenuation methodeven if same black carbon (BC). Often BC andEC are used measurement system is applied. The intercomparison of interchangeably. However, as will be seen from the different analytical instruments at multiple sites can results presentedin this paper, these quantities are not provide a better understanding of the potential varia- measures of the same properties of PM2:5. Thus, careful tions of aerosol optical BC under various atmospheric definition of these quantities are required to permit conditions. proper discussion of their sources and how they behave Currently, there are few high time resolution measure- in the atmosphere. ments of OC, EC, andBC collocatedwith other These carbon fractions are typically defined opera- continuous chemical species measurements at various tionally by their measurement protocols. Since OC and monitoring sites. Continuous OC, EC, andBC measure- EC have thermally different properties, they are ments with higher time resolution than filter-based determined by using thermal optical protocols. One measurements are one of the important steps for such protocol is the National Institute of Occupational understanding aerosol behavior and supporting source Safety andHealth (NIOSH) method5040 with correc- apportionment studies since they could permit measure- tion by the thermal-optical transmission (TOT) (Birch ments to be on the time scale of dynamic atmospheric andCary, 1996; Babich et al., 2000; Chow et al., 2001; processes such as winddirectionshifts that affect the Watson andChow, 2002 ). The other protocol is the concentrations of atmospheric carbonaceous aerosol. interagency monitoring of protectedvisual environ- With the high-time resolution, it can be usedfor ments/thermal optical reflectance (IMPROVE/TOR) epidemiological studies to investigate the influence of (Chow et al., 1993, 2001). Both of these protocols carbonaceous materials on acute adverse effects of produce OC and EC values, but because of differences in human health. these protocols described by Chow et al. (2001), As parts of summer intensive Rochester PM study different values are obtained. There are also variations and the Philadelphia north east oxidant and particle in the measurement of BC depending on the wavelength study (NE-OPS) program, real-time PM2:5 mass, semi- of light andwhether reflectance or transmission is used. continuous organic andelemental carbon, andcontin- Comparing optical differences among carbonaceous uous black carbon were measuredsequentially at urban species, EC is generally characterizedas black carbon sites in Rochester andPhiladelphia.Semi-continuous (BC) (NARSTO, 2003). Because of its ability to absorb carbonaceous species andcontinuous PM 2:5 mass were light, BC reduces visibility by both scattering and measuredin two cities to determinethe relationships for absorbance andis important with respect to the earth’s OC, thermal EC, optical EC (BC) andPM 2:5 mass radiative balance (Wolff et al., 1981; Cruntzen and concentrations. This paper reports the results of the Andreae, 1990; Jacobson, 2000, 2001). In terms of intercomparison conducted to determine temporal analytical methodology, EC can be classified by two variations of carbonaceous aerosols andto evaluate categories, thermal EC andoptical BC. While thermal the variation of measurements of thermal EC and EC can be determined by TOT analysis or TOR optical BC in the various atmospheric conditions. (thermal-optical reflection) analysis as mentionedbe- fore, optical attenuation methods such as an Aethal- ometer have been usually usedto obtain BC ( Hansen 2. Experimental methods et al., 1984). At this time, there has not been many reports regarding the differences between thermal EC Studies were conducted at the University of Rochester andoptical BC concentrations. Even though reasonably Medical Center (Latitude 43 070 3000 N, Longitude goodagreements were reportedin a few intercomparison 77 370 1400 W) located5 km southeast of downtown studies of the Aethalometer with other carbon analytical Rochester, New York from 6 to 18 June 2002 andat methods. The calculated specific absorption cross the NE-OPS site (Latitude 40 020 1000 N, Longitude section (s, attenuation coefficient) for the Aethalometer 75 000 1400 W), approximately 20 km northeast of variedin differentenvironments ( Allen et al., 1999; downtown Philadelphia, Pennsylvania from 1 July to 2 Babich et al., 2000; Lim et al., 2003). In addition, the August 2002. The monitoring site in Rochester was thermal, optical, andchemical characteristics that adjacent to a road with moderate traffic and located include variable attenuation coefficients of BC are 5 km away from downtown Rochester. In Philadelphia, ARTICLE IN PRESS

C.-H. Jeong et al. / Atmospheric Environment 38 (2004) 5193–5204 5195 the monitoring site was located500 m away from a light transmittedwhile it is being analyzed.The light traffic roadand1 km to the east of a heavy traffic transmittance monitoredby the laser beam through highway (I-95). the filter was usedfor determiningthe split point Continuous PM2:5 mass concentrations were obtained between the pyrolizedcarbon formedfrom the organic with a 30C taperedelement oscillating microbalance carbon that was originally in the sample (Turpin et al., (TEOM) (Rupprecht andPatashnick, Albany, NY) with 1990; Birch andCary, 1996 ). During the oxidizing step, a sample equilibration system (SES) dryer every 10 min EC andpyrolytic carbon were oxidizedandthe light (Meyer et al., 2000). The 1400a TEOM was configured transmittance through the filter increasedto the back- with a PM10 head, a PM2:5 sharp cut cyclone (SCC), a groundlevels for a clean quartz filter. The split point is SES dryer, and a TEOM that directly measures particle the point in time during the analysis at which the laser mass collectedon a filter by drawingambient air value equals the initial signal. In order to reduce through a filter at a constant flow rate, continuously uncertainties due to low concentrations of elemental weighing the filter andcalculating rolling 10-min carbon, samples for the fieldOC/EC instrument were smoothedmass concentrations. The sample equilibra- collectedfor 105 min andwere analyzedfor 15 min tion system (SES) includes a Nafion dryer (Perma Pure every 2 h. Inc., Toms River, NJ) to condition the main and bypass The semi-continuous Sunset Lab OC/EC instrument sample streams to low humidity and fixed temperature. was externally andinternally calibratedby using sucrose The SES is designed to reduce the possibility of PM standard stock solution ð4:21 mgCmlÀ1Þ anda 4.97% mass concentrations being overestimateddueto the methane in helium mixture, respectively. Three sucrose moisture affinity of some types of particles. standard samples, micropipette deposits on clean À2 Carbonaceous materials in PM2:5 were measuredby punchedfilters (2.73, 27.34 and54 :68 mgCcm ), were using two analyzers, a semi-continuous Sunset Lab OC/ usedfor the external calibration of total carbon EC fieldinstrument (Sunset Laboratory, Forest Grove, concentrations every 3 days. Measured concentrations OR) for OC, thermal EC (EC), andSunset Lab optical of the standard samples were verified within 5% of the EC ðBCSÞ anda two-wavelength Aethalometer, AE-20 calculatedvalues. The internal calibration usedby the (Magee Scientific, Berkeley, CA) for Aethalometer methane mixture was automatically repeatedin a fixed optical EC (Aethalometer BC, BCAÞ. Samples intro- volume loop at the endof every analysis to compensate duced into the semi-continuous OC/EC field instrument the variability of the instrument’s FID responses. were segregatedby a PM 2:5 very sharp cut cyclone Unfortunately, there is no standard reference material (VSCC) (BGI Inc., Waltham, MA). The sampledaerosol for EC that can be usedas part of the quality assurance is then passedthrough a carbon impregnatedfilter (CIF) process. multichannel, parallel plate denuder to remove most of Sunset Lab ‘‘optical EC’’ ðBCSÞ was measuredby the the gas phase organic materials (Eatough et al., 1993). semi-continuous Sunset Lab OC/EC analyzer using a The semi-continuous OC/EC fieldinstrument is based 680 nm laser. The laser transmittance when a sample is on NIOSH Method5040. The temperature program of loaded and the final signal when the deposit has been the fieldOC/EC analyzer has only two steps insteadof analyzedandremovedare measuredto obtain absor- the basic NIOSH four steps during the OC analysis bance (attenuation, ATN) during sampling. Based on when the sample is heatedunderhigh purity He. The the absorbance andthe measuredthermal EC concen- time/temperature steps for the OC andthermal EC trations, the apparent attenuation coefficient is esti- analysis are given in Table 1. To correct for EC error matedusing a seconddegreepolynomial fit of the produced by OC pyrolysis, the darkening of the filter specific absorption cross section (s, attenuation coeffi- was monitoredby measuring the intensity of light cient) obtainedautomatically by the instrument. An example of the relationship between the attenuation coefficient andthe extent of the depositis shown in Table 1 Fig. 1 based on data obtained from measurements made Temperature andresidence time protocol of the semi-contin- in this study. The attenuation coefficient is related to the uous Sunset Lab OC/EC fieldanalyzer for 2-h integrated loading of elemental carbon on an unit filter area. As the samples in Rochester andPhiladelphia sample becomes darker, the attenuation coefficient Step Gas Residence time (s) Temperature ðCÞ decreases. At very low loadings, the attenuation coefficient is much higher. Using the absorbance a OC He 90 600 measuredevery 2 h, the mass concentration of BC S OC Hea 90 870 was calculatedusing an attenuation coefficient obtained Cooling Hea 60 640 from the equation developed by the instrument. b EC He þ O2 90 870 High time resolution Aethalometer optical EC a Ultra high purity grade, 499:99%. (Aethalometer BC, BCA) concentrations measuredby b10% oxygen. using a two-wavelength Aethalometer were obtainedat ARTICLE IN PRESS

5196 C.-H. Jeong et al. / Atmospheric Environment 38 (2004) 5193–5204

excluded from the calculations. Mean concentrations of thermal OC andthermal EC (EC) in Rochester were 8.6 and0: 4 mgmÀ3, respectively while the values in Phila- delphia were 5.2 and 0:4 mgmÀ3 during the study periods. Sunset Lab optical EC ðBCSÞ andAethalometer optical EC (Aethalometer BC, BCA) concentrations in Rochester rangedfrom 0.1 to 1: 8 mgmÀ3 with a mean of 0:3 mgmÀ3 andfrom 0.2 to 2: 4 mgmÀ3 with a mean À3 of 0:9 mgm , respectively. In Philadelphia BCS values rangedfrom 0.1 to 2: 1 mgmÀ3 with a mean of 0:4 mgmÀ3 À3 while BCA variedfrom 0.1 to 7: 2 mgm with a mean of 0:9 mgmÀ3. The average mass ratios andranges of OC, EC, BC S, andBC A to PM2:5 over the duration of field study are shown in Table 2 for each sampling site. The values in Fig. 1. Variations of the attenuation coefficients usedfor the Table 2 are calculatedbasedon the 2 h averages of each measurements as OC andEC were obtainedusing the determination of optical elemental carbon ðBCSÞ measuredby the semi-continuous Sunset Lab OC/EC fieldanalyzer. Sunset Lab analyzer every 2 h. As shown, average mass fraction of semi-continuous OC=PM2:5 in Rochester was much higher than the ratio in Philadelphia while there both sampling sites. Aethalometer BC ðBCAÞ was were no significant differences of EC fractions in PM2:5 measuredat l ¼ 880 nm andUV-absorbing aromatic between both sites. It clearly shows that the carbonac- organic material (UV-BC) was measuredat 370 nm. eous aerosol accountedfor 70% of PM 2:5 andOC is BCA mass was obtainedby determiningthe attenuation most abundant chemical species during the study period of light transmittedthrough a sampledfilter ( Hansen in Rochester. In Philadelphia, carbonaceous particles et al., 1984). In contrast to the non-fixedattenuation only constitutedapproximately 30% of the 2-h averaged coefficients of the Sunset Lab semi-continuous OC/EC PM2:5 concentrations. However, cases when the semi- analyzer, the attenuation coefficient of the Aethalometer continuous OC concentrations exceeded the PM2:5 were usedin this studyfor the calculation of BC A was observedat both sites. These appear to be either 16:6m2 gÀ1 as recommended by the manufacturer. The significant loss of semivolatile OC from the TEOM or attenuation coefficient is the critical parameter to a significant adsorption artifact on the quartz filter in the convert attenuation measurements (ATN) to BCA mass Sunset OC/EC instrument. Since monitoring room ð½BCAÞ using the relation: temperature exceeded 30 C, the TEOM was often operatedat more than 30 C for the cases andthus, loss ½BC ¼ATN=s: A of mass from the TEOM is the preferredhypothesis for Thus, variability is the attenuation coefficient would the lack of mass closure (Jeong et al., 2004). On average, result in errors in the estimatedblack carbon mass. The the semi-continuous EC fractions were low at both sites two-wavelength Aethalometer has UV wavelength ðl ¼ andthe contributions to PM 2:5 mass were small. The 370 nmÞ to identify the presence of certain organic ratio of BCA to PM2:5 in Rochester was somewhat compounds such as polycyclic aromatic hydrocarbons (PAH) that strongly absorb UV wavelengths. The enhancement in the UV absorbing signal indicates the Table 2 presence of additional quantities of non-black, UV- Average mass ratios of 2-h average OC, EC, BCS, andBC A to PM concentration obtainedduringthe period6–17 June, absorbing aromatic organic material. 2:5 2002 in Rochester andduringthe period1 July to 2 August, 2002 except the Canadian forest fire event in Philadelphia

3. Results and discussion Rochester Philadelphia

3.1. Variations of real-time PM and semi-continuous OC=PM2:5 Average 0.70 0.26 2:5 Range 0.32–1.36 0.10–1.39 carbonaceous compounds EC=PM2:5 Average 0.04 0.02 Range 0.01–0.14 0.01–0.12 OC/EC values in Rochester from 10 to 14 June were BCS=PM2:5 Average 0.03 0.02 lost due to the unexpected malfunction of the semi- Range 0.01–0.13 0.01–0.09 continuous OC/EC analyzer. For the initial discussion BCA=PM2:5 Average 0.08 0.04 of these measurements, the unusual Canadian forest ﬁre Range 0.02–0.29 0.01–0.27 episode observed in Philadelphia in early June was Validnumber 47 304 ARTICLE IN PRESS

C.-H. Jeong et al. / Atmospheric Environment 38 (2004) 5193–5204 5197 higher than the ratio in Philadelphia, whereas there was (NOAA, 2003) web server was also shown in Fig. 2.As no significant difference of BCS between Rochester and expected, the 3 h average mixing depths in the evening Philadelphia. rush-hour during July 2002 was approximately three time higher than the depths in the morning rush-hour. 3.2. Diurnal variations of carbonaceous aerosols In contrast to the diurnal trend of EC, semi- continuous OC slightly decreased for weekdays while Mobile sources, diesel and gasoline engines, are OC slightly increasedon the weekends.In general, known to be major sources of OC andEC. However, although OC measuredon weekdayswas somewhat OC concentrations may arise from a variety of source higher than the concentrations on weekends, the overall types (Gray andCass, 1998 ). Sources of OC andEC difference between weekdays and weekends was negli- may be potentially identified by using the temporal gible in the afternoon. It might be expectedthat sources trends of the concentrations. Diurnal variations of semi- of OC were less relatedto traffic sources andthey would continuous OC and EC during weekdays and weekends be linkedwith a variety of sources such as stationary in Philadelphia are shown in Fig. 2. As expected, the EC sources andformation of secondaryorganic aerosol in concentrations for weekdays peaked during the morning the afternoon. rush-hour from 7 to 9 a.m. andwere much higher than Similar to the diurnal trends shown in Fig. 2, the the EC concentrations measuredon the weekends.These diurnal variations of EC and OC measured in Rochester results strongly suggest that sources of EC were closely are also shown in Fig. 3. Note that due to limited data, relatedto motor vehicles. There were no evening rush- the differences between weekdays and weekends were hour peaks during both weekdays and weekends, not compared. The results strongly support the diurnal presumably because of the substantially higher bound- patterns of carbonaceous aerosols that show clear ary layer mixing depths during the late afternoon. The morning peaks of EC occurredduringrush-hour diurnal variation boundary layer mixing depths of whereas there was no significant temporal variation of obtainedon the NOAA Air Resources Laboratory semi-continuous OC concentrations. This lack of variation also suggests significant levels of transportedOC. Diurnal variations of Aethalometer BC ðBCAÞ obtained during weekdays (13 days for Rochester and 18 days for Philadelphia) and weekends (6 days for Rochester and 7 days for Philadelphia) in Rochester and Philadelphia are shown in Fig. 4. In Rochester, BCA concentrations on weekdays peaked around 9 a.m., and broadpeak was foundaround3–5 p.m. However, during the weekends, there was no peak as a function of time of day. It is assumed that the first peak was related to the morning rush-hour. The secondbroadpeak also appeared to be traffic related, but reduced because of the greater dilution in the afternoon periods. In Philadel- phia, the peak was observedduringthe morning rush- hour andthe highest concentration occurredat 6 a.m.

Fig. 2. Comparison of diurnal variations of elemental carbon, organic carbon, and boundary layer depths in Philadelphia, PA, Fig. 3. Diurnal variations of elemental carbon andorganic 10 July–2 August 2002. carbon measuredin Rochester, NY, 6–18 June 2002. ARTICLE IN PRESS

5198 C.-H. Jeong et al. / Atmospheric Environment 38 (2004) 5193–5204

peak offsets in Fig. 4. Similar feature was also foundin the diurnal trends of EC in Philadelphia and Rochester as shown in Figs. 2 and 3.

3.3. Comparison of thermal elemental carbon and optical elemental carbon

Fig. 5 show the correlations of 2-h average Sunset Lab thermal EC obtainedin Rochester from 7 to 19 June 2002 andin Philadelphiafrom 1 July to 2 August 2002 andSunset Lab optical EC ðBCSÞ andAethalometer BC ðBCAÞ concentrations, respectively. The number of pair- wisedvalidsamples were 60 and281 measurements in Rochester andPhiladelphia,respectively. Since a strong and distinct impact of the Canadian forest ﬁre from 6 to 9 July in Philadelphia was observed, these data that will

Fig. 4. Weekday and weekend diurnal variations of Aethal- ometer BC ðBCAÞ in Rochester, NY andin Philadelphia, PA. for weekdays. There was no significant diurnal variation of BCA during weekends. Thus, the variations of BCA were strongly relatedto motor vehicles combinedwith lower mixing height in the morning similar to the observedEC concentrations. In comparing the two morning peaks of BCS in Rochester andPhiladelphia,the peak time at the Philadelphia site occurred earlier than the peak at Rochester site. This is likely due to a result of near by surroundings of the sampling sites. As noted in the description of two monitoring sites, the Rochester site is locatedin the University of Rochester MedicalCenter, which is adjacent to a medium volume of traffic road. It is expectedthat most traffic sources might be concen- tratedin the late morning rush-hour, around9 a.m. In contrast, the immediate surroundings of the Philadel- phia site consistedof the suburban outskirts of Philadelphia and the busy I-95 highway, approximately 1 km from the site. It wouldbe expectedthat the morning rush-hour on the highway in Philadelphia may Fig. 5. Comparison between Sunset Lab thermal EC and start earlier than at the Rochester site. The difference in Sunset Lab optical EC ðBCSÞ in Rochester andPhiladelphia the morning rush-hour at two sites likely results in slight during the summer of 2002. ARTICLE IN PRESS

C.-H. Jeong et al. / Atmospheric Environment 38 (2004) 5193–5204 5199 be discussed in the next section were excluded in these figures. As presentedin Fig. 5, EC in Rochester values were strongly correlatedwith BC S values with a high correlation coefficient, r2 ¼ 0:95 andthe slope was 0:89 0:02. Similar goodagreement was also observed in Philadelphia although the correlation coefficient (r2 ¼ 0:73, andslope ¼ 0:99 0:07) was somewhat lower than the value in Rochester. In the bottom of Fig. 5 for Philadelphia, two extreme outliers were observed aroundin the early morning of August 2 with high BC S concentrations. Simultaneously with the peaks, the high concentrations of Aethalometer BC were also observed. It was probably due to unidentified particles having quite different attenuation coefficients compared to typical carbonaceous materials. The unusual cases with significantly high optical EC concentrations were mostly found during the Canadian forest fire episode. Fig. 6 shows the correlations between BCA measured using the Aethalometer andthermal EC in Rochester andPhiladelphia.CalculatedBC A concentrations were significantly higher than EC at both sites although they were highly correlated ðr2 ¼ 0:84 in Rochester and r2 ¼ 0:60 in Philadelphia). In these intercomparison studies, the slopes of collocatedBC A versus EC measurements were 3:3 0:2 and2: 7 0:1 in Rochester andPhiladel- phia, respectively. It shouldnote that the slopes were much higher than the comparison between BCS andEC while the correlation coefficients were relatively lower in the intercomparison of BCA andEC. As mentioned previously, the optical attenuation values measuredby the Aethalometer were convertedto Aethalometer BC mass concentration by using the specific absorption cross section, 16:6m2 gÀ1 (Hansen et al., 1984; Hansen andMcMurry, 1990 ). The obvious difference between the slopes at these sites suggests that the attenuation Fig. 6. Comparison between Sunset Lab thermal EC and coefficient of BCA is variable with time andlocation and Aethalometer optical EC (Aethalometer BC, BCA) in Rochester depends on the physical and chemical characteristics of and Philadelphia during the summer of 2002. light absorbing species that couldincludeOC as well as EC. Several studies found similar limitations for the phia, respectively. According to several measurements of Aethalometer andsuggestedthat optical analysis for the attenuation coefficients conducted in various loca- EC requires calibration for site-specific aerosol and tions (Jennings andPinnick, 1980; Heintzenberg, 1982; optical properties since the attenuation coefficient Liousse et al., 1993), the attenuation coefficients were depends on the size distribution, type of aerosol strongly relatedto the distanceof the site from the black mixtures, anddepositedmass per unit time ( Horvath, carbon sources. The attenuation coefficient at a remote 1993; Liousse et al., 1993; Petzoldet al., 1997; site was low while the value near a street was higher. In Hitzenberger et al., 1999; Babich et al., 2000; Ballach Rochester, the sampling site was locatedabout 3 m from et al., 2001). These results also suggest that the a local road, and the very high attenuation coefficient attenuation coefficients of the Aethalometer in Roche- might be due to the location of the Rochester sampling ster andPhiladelphiawere larger than the manufac- site. 2 À1 turer’s value of 16:6m g . Basedon the thermal EC The correlation coefficient of BCA andEC concentra- concentrations measuredusing the Sunset Lab OC/EC tions in Philadelphia was lower than the value in instrument, the fixedabsorption coefficient for the Rochester. The somewhat poorer correlation in Phila- Aethalometer can be modified as approximately delphia might have been due to the variable aerosol 54:8m2 gÀ1 in Rochester and44 :2m2 gÀ1 in Philadel- characteristics during several strong sulfate haze events ARTICLE IN PRESS

5200 C.-H. Jeong et al. / Atmospheric Environment 38 (2004) 5193–5204

(Jeong et al., 2004). During the haze episode, BCA may Average OC fraction of PM2:5 mass concentration become associatedother particles as well as some water was 12% higher during the forest fire period than during associatedwith the acidsulfate particles. There were five the rest of the period. The increased OC=PM2:5 ratio, sulfate haze episodes in Philadelphia and the most severe 0.38, is similar to the values foundby Patterson and haze episodes occurred from 2 to 4 July and from 17 to McMahon (1984). They reportedthat OC =PM2:5 ratios 19 July 2002 with the highest PM2:5 mass concentrations rangedfrom 0.41 to 0.61 for an open forest fire. In À3 of 92.8 and95 :4 mgm . During the severe haze events, addition, OC was more highly associated with PM2:5 2 the differences between BCA andEC concentrations during the episode ðr ¼ 0:95Þ than on the non-event 2 tended to increase because BCA sharply increased. The days ðr ¼ 0:68Þ. It clearly shows that OC strongly highest BCA concentration was observedon the morn- contributedto the increasedPM 2:5 mass during the ing of 3 July with a value of 8:0 mgmÀ3 whereas EC was forest fire event as expected. The mass ratio of À3 1:2 mgm . It can be concluded that BCA concentrations EC=PM2:5 slightly increasedwith a mean fraction of in Philadelphia might be overestimated due to the 4%, andthe correlation coefficient between EC and hygroscopic behavior of carbonaceous materials during PM2:5 also moderately increased from 0.30 to 0.58 the sulfate haze events. during the event. For the Aethalometer BC, the mass fraction BCA to PM2:5 mass was around3%. Although the mass fractions of EC andBC A were similar during 3.4. Influence of Canadian forest fire on carbonaceous the event, the correlation coefficient between BCA and compound measurements PM2:5 decreased from 0.33 to 0.13 and was lower than the value between EC andPM 2:5. There was a significant A Canadian forest fire started on 2 July and severe fluctuation of measuredBC A values during the event, smoke was transportedto the monitoring site from 6 to especially on 6 and7 July 2002. High values of BC A 9 July 2002. Table 3 shows summary of real-time PM2:5 were mainly observedaroundin the evening of 6 July mass andcarbonaceous species measuredin Philadel- andin the morning of 7 July whereas the strong peaks of phia during the Canadian forest fire episode. PM2:5 PM2:5 andEC were observedaroundmidnighton 7 July concentrations rangedfrom 12 to 164 mgmÀ3 with a andin the morning of 8 July 2002. À3 mean of 79:8 mgm with the highest PM2:5 mass being In order to better evaluate the effect of the forest fire observedaroundmidnighton 7 July 2002 (local time). on the carbon analysis methods for the determination of The mean PM2:5 value during the forest fire was elemental carbon, the correlations between thermal EC approximately 4 times higher than the mean during the andAethalometer BC (BC A) during the severe haze rest of the measurement period1 July to 2 August. The events occurredaroundfrom 2 to 4 July and18 to 19 highest OC andEC concentrations were observed July 2002, and during the Canadian forest fire episode in aroundmidnight,7 July andaround9 a.m. on 8 July Philadelphia are shown in Fig. 7. There are two distinct 2002, respectively. Comparedwith the normal ambient slopes for the events in the figure. One slope plotted conditions, the OC and EC concentrations dramatically as closedcircles is for the sulfate haze events when 2 increasedby factors of 7 and9, respectively. The mean EC andBC A are highly correlated( r ¼ 0:68, concentration of BCA was approximately four times slope ¼ 3:61 0:35). The BCA values during the haze higher than the mean observed during normal days. The events were typically higher than EC concentrations highest BCA value was also observedon the morning of with a goodcorrelation as mentionedin the previous 7 July 2002 with a value of 8:0 mgmÀ3. section. However, during the forest fire event, the trends

Table 3 Statistical summary of real-time PM2:5, semi-continuous OC/thermal EC, and continuous Aethalometer BC in Philadelphia during the Canadian forest ﬁre, 6–9 July, 2002

a b b a PM2:5 Organic carbon Elemental carbon Aethalometer black carbon ðmgmÀ3ÞðmgmÀ3ÞðmgmÀ3ÞðmgmÀ3Þ

Mean 79.8 30.6 2.9 2.3 Standard deviation 36.2 16.2 1.6 1.5 Standard error 4.0 2.5 0.2 0.2 Minimum 12.0 3.1 0.1 0.3 Maximum 164.0 67.1 6.2 8.0 Validnumber 82 41 41 82

aAverages of 1 h average samples. bAverages of 2 h integratedsamples. ARTICLE IN PRESS

C.-H. Jeong et al. / Atmospheric Environment 38 (2004) 5193–5204 5201

Fig. 7. Correlation between Sunset Lab thermal EC and Fig. 8. Correlation between Sunset Lab thermal EC andSunset Lab optical EC during the severe haze episodes and during the Aethalometer BC ðBCAÞ during the severe haze episodes and the Canadian forest ﬁre event in Philadelphia, PA. Canadian forest ﬁre event in Philadelphia, PA.

in EC andBC A changeddramaticallyon the afternoon Fig. 8 shows the correlations between EC andSunset of 7 July 2002 as is apparent in Fig. 7. The slope between Lab optical EC ðBCSÞ. The correlation coefficients EC andBC A also changedsignificantly from 3.61 to 0.35 during the haze and during the Canadian forest fire while the correlation coefficient decreased somewhat episode were 0.82 with a slope of 0:97 0:06 and0.80 ðr2 ¼ 0:60Þ. Note that the data from afternoon 6 July to with a slope of 0:31 0:02, respectively. As noted afternoon 7 July are in a transition region between two previously, both EC andBC S concentrations were slopes andare not shown in the Fig. 7. The result of the obtainedby a Sunset Lab OC/EC analyzer every 2 h. proportionality between EC andAethalometer BC As shown in Fig. 8, during the haze event the shows clearly that there was a significant change in the comparison showeda quite goodagreement as expected, attenuation coefficient in relation to the origin andthe whereas during the forest fire event BCS values were physical andchemical properties of carbonaceous aroundthree times lower than EC even though the BC S aerosols during the Canadian forest fire episode. The values were calculatedby flexible attenuation coeffi- variability of the attenuation coefficient is likely to be cients depending on the deposit loading of EC. The due to the presence of pyrolized organic matter and average attenuation coefficients appliedduringthe haze moisture emittedduringthe fire event ( Liousse et al., episode and the forest fire episode were 51.3 and 1993; Jacobson, 2000). In contrast to the correlation 19:1m2 gÀ1, respectively. The results indicate that fine between EC andBC A in the normal urban atmosphere, particles emittedandtransportedfrom the Canadian BCA concentrations calculatedby using a fixedattenua- forest fire resultedin the significant biases of the optical tion coefficient were appreciably lower than the thermal attenuation methods used to quantify EC. The absorp- EC values measuredby the Sunset OC/EC analyzer in tive capacity was probably underestimated since the the heavily pollutedatmosphere. Thus, this wouldresult attenuation coefficient usedfor the BC S variedas a in an apparent underestimate of BCA mass for a given function of the deposit loading. Based on the thermal attenuation coefficient of the Aethalometer. Reported EC values measured, the modified attenuation coeffi- absorption coefficients for laboratory fires rangedfrom cient was 5:9m2 gÀ1 on average during the Canadian 0.04 to 2:36 m2 gÀ1 (Patterson andMcMahon, 1984 ). forest fire episode. Note that the values are significantly smaller than the As described previously, the AE-20 Aethalometer has 16:6m2 gÀ1 being used in the study. By comparison, data two-wavelength sources, 880 and370 nm, andthe near of Patterson andMcMahon (1984) indicates that UV wavelength is usedfor identifyingthe presence of organic fraction during smoldering combustion pyroly- UV absorbing organic compounds ðBCUVÞ such as sis was larger than the during flaming combustion. polycyclic aromatic hydrocarbons (PAH) even if the ARTICLE IN PRESS

5202 C.-H. Jeong et al. / Atmospheric Environment 38 (2004) 5193–5204

0:89 0:02 while the correlation coefficient was 0.73 with a slope of 0:99 0:07 in Philadelphia. Comparing EC andBC A, the correlation coefficients at both sites were 0.84 with a slope of 3:3 0:2 in Rochester and0.60 with a slope of 2:7 0:1 in Philadelphia. The attenuation coefficients of optical elemental carbon analyses may vary anddependonthe sampling locations or atmospheric conditions. During severe haze events, the differences between BCA andEC concentrations tended to increase. During the Canadian forest fire, PM2:5 mass ranged aroundfrom 12 to 164 mgmÀ3 with a mean of À3 79:8 mgm while OC, EC, andBC A increasedby a factor of 7, 9, and4, respectively. While the OC =PM2:5 ratio during the forest fire episode was 0.38, OC and PM was strongly correlated ðr2 ¼ 0:95Þ. In addition, Fig. 9. Comparison of Aethalometer BC ðBCAÞ measuredusing 2:5 880 nm wavelength andAethalometer UV absorbing organic the proportionality between EC andBC A dramatically matters detected using 370 nm wavelength in Philadelphia, PA. decreased from 3.61 to 0.35 while there was no significant change of the correlation coefficient during the Canadian forest fire episode. Similar behavior was values of components measuredusing this second also foundin the comparison of EC andBC S. Although wavelength cannot be quantifiedas real mass concentra- EC andBC S calculatedby using multiple attenuation tions. The levels of BCUV were typically less than or coefficients showeda goodagreement, the proportion- similar to BCA concentrations during the normal ality was also changedfrom 0.97 to 0.31 dueto the measurement period. However, as shown in the Fig. 9, influence of the forest fire. The UV absorbing organic the UV absorbing organic compoundconcentration compounds level significantly increased during the fire increasedsignificantly to become much larger than BC A episode period. concentrations during the fire episode period. It can be It is clear from these results that there needs to be anticipatedthat some of the most abundantUV careful definitions of the operational carbon fractions. It absorbing organic species were PAHs since burning of is suggestedthat the term ‘‘black carbon’’ be reserved woodemits various hydrocarbons( McMahon and for optically basedmeasurements andthe term ‘‘ele- Tsoukalas, 1978; Jenkins et al., 1996). Thus, the mental carbon’’ be reservedfor thermal optical mea- instrument does respond to UV absorbing species and surements. Even within each of these categories, it is provides information that the mixture of sources of BC essential to carefully define the measurement protocols have changed. since differences among the available protocols will produce differences in the measured concentrations. Future work may resolve these differences or lead to 4. Conclusions more uniform methodology, but in the interim, care is needed in describing and reporting OC, EC, and BC The average semi-continuous OC=PM2:5 in Rochester values. was approximately 0.70 andwas much higher than in Philadelphia (0.26) while EC/PM2.5 in Rochester was slightly higher than in Philadelphia. The BCA ratio to PM2:5 in Rochester was two times higher than the ratio Acknowledgements in Philadelphia while BCS ratios to PM2:5 at two sites were similar with a range from 0.01 to 0.13. Parts of this work were financially supportedby The diurnal patterns of OC and EC concentrations EPA PM Center grant R827453 andthe Pennsylvania during weekdays and weekends suggest that the sources Department of Environmental Protection andthe US of EC were closely relatedwith traffic sources whereas Environmental Protection Agency though Cooperative there were no significant trends in OC concentrations Agreement CR 827591. We wouldlike to thank suggesting more effect of distant resources. The diurnal Dr. Gunter Oberdo¨ rster andRobert Gelein for their variations of the BCA obtainedshow that the variations assistance in making the measurements in Rochester of BCA were strongly correlatedwith motor vehicles andDr. C. Russell Philbrick of the Pennsylvania analogous to the behavior of EC. State University andRichardClark of Millersville Two hour average EC andBC S values were quite well University for their assistance with the Philadelphia correlatedin Rochester ðr2 ¼ 0:95Þ with a slope of measurements. ARTICLE IN PRESS

C.-H. Jeong et al. / Atmospheric Environment 38 (2004) 5193–5204 5203

References Jacobson, M.C., 2001. Strong radiative heating due to the mixing state of black carbon in atmospheric aerosols. Allen, G., Lawrence, J., Koutrakis, P., 1999. Fieldvalidationof Nature 409, 695–697. a semi-continuous methodfor aerosol blank carbon Jenkins, B.M., Jones, A.D., Turn, S.Q., Williams, R.B., 1996. (aethalometer) andtemporal patterns of summertime hourly Particle concentrations, gas-particle partitioning, andspe- black carbon measurements in southwestern PA. Atmo- cies intercorrelations for polycyclic aromatic hydrocarbons spheric Environment 33, 817–823. (PAH) emittedduringbiomass burning. Atmospheric Babich, P., Davey, M., Allen, G., Koutrakis, P., 2000. Method Environment 30, 3825–3835. comparisons for particulate nitrate, elemental carbon, and Jennings, S.G., Pinnick, R.G., 1980. Relationship between PM2:5 mass in seven US cities. Journal of Air andWaste visible extinction absorption andmass concentration of Management Association 50 (7), 1095–1105. carbonaceous smokes. Atmospheric Environment 14, Ballach, J., Hitzenberger, R., Schultz, E., Jaeschke, W., 2001. 1123–1129. Development of an improvedoptical transmission technique Jeong, C.-H., Lee, D.-W., Kim, E., Hopke, P.K., 2004. for black carbon (BC) analysis. Atmospheric Environment Measurement of real-time PM2:5 mass, sulfate, andcarbo- 35, 2089–2100. naceous aerosols at the multiple monitoring sites. Atmo- Birch, M.E., Cary, R.A., 1996. Elemental carbon-basedmethod spheric Environment, this volume. for monitoring occupational exposures to particulate diesel Lavanchy, V.M.H., Ga¨ a¨ ggeler, H.W., Nyeki, S., Baltensperger, exhaust. Journal of Aerosol Science andTechnology 25, U., 1999. Elemental carbon andblack carbon measurements 221–241. with a thermal methodandan aethalometer at the high- Chow, J.C., Watson, J.G., Pritchett, L.C., Pierson, W.R., alpine research station Jungfraujoch. Atmospheric Environ- Frazier, C.A., Purcell, R.G., 1993. The DRI thermal/optical ment 33, 2759–2769. reflectance carbon analysis system: description, evaluation Lim, H.J., Turpin, B.J., Edgerton, E., Hering, S.V., Allen, G., andapplications in US air quality studies.Atmospheric Maring, H., Solomon, P., 2003. Semicontinuous aerosol Environment 27A (8), 1185–1201. carbon measurements: comparison of Atlanta supersite Chow, J.C., Watson, J.G., Crow, D., Lowenthal, D.H., measurements. Journal of Geophysical Research 108 (D7), Merrifield, T., 2001. Comparison of IMPROVE and 8419 doi 10.1029/2001JD001214. NIOSH carbon measurements. Journal of Aerosol Science Liousse, C., Cachier, H., Jennings, S.G., 1993. Optical and andTechnology 34, 23–34. thermal measurements of black carbon aerosol content in Cruntzen, P.J., Andreae, M.O., 1990. Biomass burning in the different environments: variation of the specific attenuation Topics: impact on atmospheric chemistry andbiogeochem- cross-section sigma ðsÞ. Atmospheric Environment 27, ical cycles. Science 250, 1669–1678. 1203–1211. Eatough, D.J., Wadsworth, A., Eatough, D.A., Crawford, McMahon, C.K., Tsoukalas, S.N., 1978. Polynuclear aromatic J.W., Hansen, L.D., Lewis, E.A., 1993. A multiple-system, hydrocarbons in forest fire smoke. Carcinogenesis 3, multi-channel diffusion denuder sampler for the determina- 61–73. tion of fine-particulate organic material in the atmosphere. Meyer, M.B., Patashnick, H., Ambs, J.L., Rupprecht, E., 2000. Atmospheric Environment 27A, 1213–1219. Development of a sample equilibration system for the Gray, H.A., Cass, G.R., 1998. Source contributions to atmo- TEOM continuous PM monitor. Air andWaste Manage- spheric fine carbon particle concentrations. Atmospheric ment Association 50, 1345–1349. Environment 32, 3805–3825. NARSTO, 2003. Particulate matter science for policy makers— Hansen, A.D.A., McMurry, P.H., 1990. An intercomparison of a NARSTO assessment. Parts 1 and2. NARSTO Manage- measurements of aerosol elemental carbon during the 1986 ment Office, (Envair), Pasco, Washington. http:// carbonaceous species methodcomparison study.Journal of www.cgenv.com/NARSTO. Air andWaste Management Association 40, 894–895. National Oceanic andAtmospheric Administration (NOAA)— Hansen, A.D.A., Rosen, H., Novakov, T., 1984. The aethal- Air Resources Laboratory (ARL), 2003. ArchivedMeteor- ometer—an instrument for the real-time measurement of ological Products. NOAA ARL, Silver Spring, MD, Web optical absorption by aerosol particles. Science of the Total address: http://www.arl.noaa.gov/ready/amet.html. Environment 36, 191–196. Pandis, S.N., Harley, R.A., Cass, G.R., Seinfeld, J.H., 1992. Heintzenberg, J., 1982. Size-segregatedmeasurements of Secondary organic aerosol formation and transport. Atmo- particulate elemental carbon andaerosol light absorption spheric Environment 26A, 2269–2282. at remote Arctic locations. Atmospheric Environment 16, Patterson, E.M., McMahon, C.K., 1984. Absorption charac- 2461–2470. teristics of forest fire particulate mater. Atmospheric Hitzenberger, R., Jennings, S.G., Larson, S.M., Dillner, A., Environment 18, 2541–2551. Cachier, H., Galambos, Z., Rouc, A., Spain, T.G., 1999. Petzold, A., Kopp, C., Niessner, R., 1997. The dependence of Intercomparison of measurement methods for black carbon the specific attenuation cross-section on black carbon mass aerosols. Atmospheric Environment 33, 2823–2833. fraction andparticle size. Atmospheric Environment 31, Horvath, H., 1993. Comparison of measurements of aerosol 661–672. optical absorption by filter collection anda transmisso- Rogge, W.F., Hildemann, L.M., Mazurek, M.A., Cass, G.R., metric method. Atmospheric Environment 27, 319–325. Simoneit, B.R.T., 1993a. Sources of fine organic aerosol. 2. Jacobson, M.C., 2000. A physically-basedtreatment of Noncatalyst andcatalyst-equippedautomobiles andheavy- elemental carbon optics: implications for global direct duty diesel trucks. Environment Science and Technology 27, forcing aerosols. Geophysical Research Letters 27, 217–220. 636–651. ARTICLE IN PRESS

5204 C.-H. Jeong et al. / Atmospheric Environment 38 (2004) 5193–5204

Rogge, W.F., Hildemann, L.M., Mazurek, M.A., Cass, G.R., Turpin, B.J., Cary, R.A., Huntzicker, J.J., 1990. An in-situ Simoneit, B.R.T., 1993b. Sources of fine organic aerosol. 5. time-resolvedanalyzer for aerosol organic andelemental Natural gas home appliances. Environment Science and carbon. Aerosol Science andTechnology 12, 161–171. Technology 27, 2736–2744. Watson, J.G., Chow, J.C., 2002. Comparison andevaluation Rogge, W.F., Hildemann, L.M., Mazurek, M.A., Cass, G.R., of in situ andfilter carbonmeasurements at the Simoneit, B.R.T., 1997. Sources of fine organic aerosol. 8. Fresno Supersite. Journal of Geophysical Research 107, Boilers burning No. 2 distillate fuel oil. Environment JD000573. Science andTechnology 31, 2731–2737. Wolff, G.T., Countess, R.J., Groblicky, P.J., Ferman, M.A., Rogge, W.F., Hildemann, L.M., Mazurek, M.A., Cass, G.R., Cadle, S.H., Muhlbaier, J.L., 1981. Visibility-reducing Simoneit, B.R.T., 1998. Sources of fine organic aerosol. 9. species in the Denver ‘‘brown cloud’’—II. Source Pine, oak, andsynthetic log combustion in residential andtemporal patters. Atmospheric Environment 15, fireplaces. Environment Science andTechnology 13–22. 2485–2502.