Correcting Aethalometer Black Carbon Data for Measurement Artifacts Byinter-Comparison Using Methodology Based on Two Di Atmos

Home , Aethalometer

Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 1 − . Field 1 − ects on BC mea- ff erent light attenuation (ATN) ff . In the absence of sampling artifacts, 1 by approximately 13 and 9 % in summer − erent sampling flow rates, and the actual 1 ff − 2852 2851 ects on BC measurements could be corrected ff can be exactly equal to the corrected BC for 2 Lmin cient. Therefore, the correction algorithm can be applied to 1 ffi − ects varied with the aerosol deposition rate. If the true ATN value ff erent light attenuation ff This discussion paper is/has beenTechniques under (AMT). review Please for refer the to journal the Atmospheric corresponding Measurement final paper in AMT if available. Department of Safety, Health and Environmental Engineering, Ming Chi University of can be found, then BCATN change measurements rate. can be Therefore,tive corrected determining for of the artifacts this true by study. ATN usingATN value The value the was proposed from true the ATN correction values primary algorithm acquired objec- can at be di used to obtain the true Abstract In black carbon (BC) measurementsartifacts obtained are using a the filter-based majortifacts optical problem. to technique, Recently, a itschemes certain has have their extent become advantages and possible by disadvantagesa under to using field new correct conditions. numerical correction In these this model methods. study, ar- that Nevertheless, can all be correction usedincreasing for rates. determining Two aethalometers artifactaerosol were e used sampling to flow measure ratesthe ATN of values ratio 6 in of and parallel ATNratio 2 at values Lmin of measured the by samplingters. flow the In rates two practice, (or aethalometers the aerosolthe ratio should deposition of same be rates) ATN as of equal values these measured theaerosol to two by loading the ratio aethalome- the e of two the aethalometers was sampling not flow rates of the aethalometers because the BC mass concentrations canBC be correction, the determined BC from concentration the measured true at ATN the change sampling rate. flow Before rate of 6 Lmin surements was proposed; the model is based on two di test results demonstrated that loading e accurately by using theATN proposed caused model. by Additionally, light theany scattering problem light at of scattering the enhanced coe unloadeda light newly filter designed can instrument be to determine overcome actual without real-time using BC concentrations by using was smaller than that measured at 2 Lmin Technology, 84 Gungjuan Rd, Taishan, New Taipei 24301, Taiwan Received: 10 February 2015 – Accepted: 5Correspondence March to: 2015 Y.-H. Cheng – ([email protected]) Published: 17 MarchPublished 2015 by Copernicus Publications on behalf of the European Geosciences Union. increasing rates Y.-H. Cheng and L.-S. Yang and winter seasons, respectively. Aftercorrected BC BC correction for by 6 Lmin using the true ATN value, the Correcting aethalometer black carbon data for measurement artifacts byinter-comparison using methodology based on two di Atmos. Meas. Tech. Discuss., 8, 2851–2879,www.atmos-meas-tech-discuss.net/8/2851/2015/ 2015 doi:10.5194/amtd-8-2851-2015 © Author(s) 2015. CC Attribution 3.0 License. 5 15 10 25 20 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | er- ect ff ff cients. This pro- ffi ect, the measured BC con- ff erence value of 0.001 could cause ff erent particle deposition rates for di ff ers high time resolution (Watson et al., 2005; 2854 2853 cients, which are necessary for use in correc- ff ffi erent aerosol deposition rates. Moreover, a simple empirical cients are considered in some of these correction methods, ff ffi ects on climate and health (Ramanathan and Carmichael, 2008; Suglia et al., ff erent environments. ects on aethalometer data without using the aerosol scattering coe Recently, it has become possible to correct these artifacts to a certain extent by using Recently, a new generation aethalometer (AE33; Magee Scientific) based on a dual- The most common method used to measure BC involves collecting aerosols on a fil- ff ff the relationship between ATNseveral change reasons, and one of BC which concentrationcollected is is on that not the both light-scattering filter linear and1993; alter because light-absorbing the Petzold of particles internal et reflection al.,2003). properties Measurement 1997; of artifacts the Reid resulting filter from etlationship (Liousse the have al., et nonlinearity shown al., 1998; of that the Bond because aforementioned of et re- the al., filter 1999; loading Weingartner e et al., Collaud Cone et al., 2010). Nevertheless,and all disadvantages correction under methods field have their conditions.ing advantages For and example, although absorption both coe aerosol scatter- numerical methods (Weingartner et al., 2003; Arnott et al., 2005; Virkkula et al., 2007; centration decreases with ancauses increase scattering in aerosols on the the filter filter load, to and increase the the measured sample BC matrix concentration. e most BC measurements performed usingneously the acquire aethalometer aerosol scattering in coe thetion field methods. Virkkula do et not al. simulta- e (2007) proposed a simple procedure to correct the loading spot method has been developed to obtain di an error of approximatelyVirkkula 4 % et in al. the (2007) BCand is is concentrations used. dependent when on Moreover, the the the correction density correction model of factor of the is particles not deposited on a the constant filter. value cedure is based on the originalit measurement is results assumed of two that continuousand the filter that spots, the BC and BC concentration value measuredcentration remains with value. A stable a correction lightly during factor loaded is the filterBC then is filter data. determined the for However, spot closest each the to change filter assumption theenvironment spot real that during to con- the correct the the filter BC spot concentrationthat change is the is sensitivity stable not of in always true. the the Järvi correction ambient et factor al. with (2008) a showed di two sampling spots for di correction scheme for post-processing foris correcting the also existed presented. aethalometer BC Whiletype, data it this can simple be used correction toconditions scheme correct are BC is data similar when dependent to thewith on primary those appropriate the source in of flow aerosol this BC control anddi study. the Furthermore, can weather be two existed used aethalometers to create correction schemes suitable for ative e ter and measuring the1984). reduction in The light aethalometer transmission (AE;vices through Magee used the for Scientific) filter measuring is (Hansen BC, anddevice et one it has al., of is been based the used on currentlybecause the extensively it filter-based available to can optical de- monitor technique. be This operated environmental easily BC and mass o concentrations 1 Introduction Black carbon (BC) aerosolsceous are fuels produced and are by a the crucial incomplete atmospheric combustion constituent because of of carbona- their potentially neg- 2008; Jacobson, 2010; PowerHence, the et measurement al., of BC 2011;in has urban, Cornell become rural, increasingly et and crucial background al., over areas the 2012; worldwide. past decade Bond et al., 2013). Park et al., 2006;et Järvi al., et 2014). al., In 2008; theare aethalometer, collected Chow an on et the air al., filter. sampleaerosols Subsequently, 2009; is the on Cheng drawn decrease the in through and light a filter Lin,attenuation transmission filter is 2013; (ATN). through It and the measured. Cheng is aerosols assumed A thatby decrease any BC in ATN increase aerosols the is accumulated solely transmissioncalculated on due implies to from the increased light filter, the absorption and rate the of BC ATN concentration change. can However, therefore previous be studies have shown that 5 5 10 15 25 20 10 15 25 20 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | (2) (4) (5) (3) (1) erent is the ff ATN σ ). Therefore, t erent aerosol ect on the BC ff d ff / erent sets of ATN mea- ff can be expressed as erent aerosol deposition rates t ff ect on BC can be corrected by ff 0) to time ) is used to estimate the BC concentration is the sampling flow rate, and t = d Q 0 / erent aerosol deposition rates. Hence, a sim- t 2856 2855 ( ff 0 t 1 ATN σ · A Q · ects on BC measurements can be corrected accurately erent ATN increasing rates was developed in the current 0 ff t ff er significantly because of di − ff ATN t = is a correction factor. On the basis of the same principle as Eq. (2), 1 ects resulting from di ect. Cheng and Lin (2013) noted that di ATN 1 ATN 0 I ff k ff I erences in the sampling flow rates or aerosol deposition areas among σ BC) from initial time σ · ff · ATN) are the intensities of light transmitted through a reference blank spot and · ATN) A Q ln A Q I · · · and ATN represent corrected and measured light ATN values, respectively. k · ATN and BC represent the corrected and measured black carbon concentration, BC k c − c t − is the area of the sample spot, cient. Field test results can also further provide users with a simple correction t and 100 (1 ATN ∆ d 0 ffi A (1 − I ∆ dATN = = = c = = c Drinovec et al. (2014) showed that the loading e In practice, if the true ATN value can be found, then the artifact e measurement can be correcteddetermining from the the true true ATN ATN change value was rate the (dATN main objective of this study. According to the using the following equation, which is similar to that proposed byBC Virkkula et al. (2007): black carbon optical mass cross section. scheme during post-processing for correcting existing aethalometer BC data. 2 Methods 2.1 Correction model For measuring BC with the filter-based optical technique, the opticalATN ATN is defined as aerosol loading e ple algorithm, similar to theDrinovec correction methods et developed al. by Virkkula (2014), etdemonstrated was al. that used (2007) loading and to e determineby the using the true proposed ATN values.of model. light Field Moreover, scattering the test at problem the results of unloadeding filter coe enhanced can light be overcome ATN without because using any light scatter- where ATN sampling flow rates in parallel. Invalues the measured absence by of these sampling two artifacts,pling the aethalometers flow ratio rates should of (or be the the aerosol equal ATN the deposition to rates) ratio the of of these ratio the two of aethalometers. ATNthe the In values practice, sam- ratio measured by of these the two sampling aethalometers was flow not rates identical of to these two aethalometers because of di BC Then, Eq. (3) can be written as where ATN where a spot of aerosol on theThe filter, respectively. ATN The change factor in 100 a is time introducedas for interval follows: convenience. (dATN BC where BC respectively, and the average BC ( the used instruments. Therefore, a correctionmeasurements model based for determining on the di artifactsstudy. on Two BC aethalometers were used to measure the ATN values at di resulting from di surement results could di ent sampling flow conditionsa (Drinovec real-time model et for al., determiningfor 2014). the the temporal This variation loading dual-spot of method e a involves compensation parameter 5 5 20 15 10 15 10 20 25 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | ), c. (9) (6) (7) (8) F6 ffi and Q erent usion ( value. ff F1 ff 3 Q − /Q erent ATN ff ). The filter F2 Q ( 3 − erent flow rates can be ex- ff ect, the ratio of true ATN values usion dryer was not used, negative ff ff ect is a cumulative property of the cumu- ff F2 is assumed to be fixed for an ATN 2857 2858 , respectively. k F2 ATN ects on ATN measurement results. can be determined on the basis of two measured · 1 ff Q F2 F k F1 F1 Q Q and = ATN ATN t t · · · · F1 erent aerosol sampling flow rates in the same deposition Q F2 F2 ff F2 F1 represent the mass of black carbon aerosol deposited on the Q Q represent the ATN measured at sampling flow rates Q Q · · ) ) − − F2 F1 F2 erent loading e F2 F2 erent flow rates should be equal to the ratio of the two di BC BC ff BC,F2 erent BC aerosol deposition rates can correspond to di ff ff F1 F2 m = erent sampling flow rates. Subsequently, the true ATN value can be ATN ATN ff ATN ATN · · · · k k ATN ATN and and ATN F1 F1 BC,F2 BC,F1 − − Q F1 m m (1 (1 ATN BC,F1 = can be solved using Eqs. (6)–(8): · = = m k F1 F2,c F1,c Q F1,c F2,c , respectively. In the absence of the artifact e According to Eq. (5), the ATN correction equation at di In this model, the correction factor = F2 ATN ATN ATN values for di Therefore, the temporal variation of pressed as ATN area could lead to di obtained from Eq. (5). Moreover,the the true actual ATN BC change concentration rate by can using be Eq. determined (2). from 2.2 Sampling equipment and data collection In this study,Scientific); BC one (or aethalometer ATN) was was operated measured at using a sampling two flow aethalometers rate (AE31; of Magee 6 Lm sampling flow rates: lative deposit of aerosolment on the results. filter, Therefore, and di it also directly influences the ATN measure- increasing rates. Nevertheless, the loading e Q measured at two di tape in these twoon aethalometers the filter was at shiftedhad the automatically same a to time fixed every expose starting eightternal a and timer hours pristine ending and to sampling time spot ensure flow ofsampling that rate the site the of was the sampling two located aethalometer sampling for on werehan, spots internal the checked New campus every comparison. week. of Taipei The City, The Ming Taiwan. in- Chi TheSometimes, University main wood of source combustion Technology at of could Tais- BC beposal. at observed During this on the sampling the sampling siteto campus periods, was each for tra the waste other two wood inmately dis- aethalometers a 2 were m sampling above positioned cabin the ground and adjacent operated level the outside between the inlets 21 sampling December of cabin. The 2013, bothtween aethalometers and aethalometers were 5 24 were July January approxi- 2014 2014,for (winter and all season), 26 measurements and be- was Septemberter set 2014 (Pallflex at Q250F), (summer which 5 min. season). waswater suggested The The by vapor filter logging the condensing tape aethalometer interval on manufacturer. used thedryer To avoid was deposited with a silica aerosols quartz gel in was the materialery installed two summer fil- on days season, during the a sampling sampling periods. di values line. When could The the silica be di gel observedature was in replaced and ev- BC high data relativewhen humidity. sets Negative sampling at values noon was were under alsoearly performed morning recorded conditions at hours, in of especially very BC at highin low low data temper- sampling the sets ambient flow data rates. concentrations set If negative of at values a appeared midnight sampling spot, and the entire data for the sampling spot was excluded definition of light ATN,position. the That ATN is, value di is dependent on the amount of BC aerosol de- where ATN where whereas the other was operated at a sampling flow rate of 2 Lm ATN filter at sampling flow rates Then, k 5 5 20 10 15 20 25 10 15 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | F6 F2 ect and (10) ff 1 − : 0.307 for 1 − , indicating = F2 p 18.119 (95 % , for 2–3 days. , especially for − 1 − F2 ATN , respectively, and 1 × value to the ATN − ATN F6 × 0.019 for 6 Lmin = 0.001). The average rela- 5.411) and p − p < by approximately 13 and 9 % in F2 C( ◦ being determined from a cumulative weather station (Davis Instruments). k could be exactly equal to the corrected 56.091 to − TM F6 0.001 for both). These comparison results erent from zero ( 2860 2859 ff value estimated using the model of Drinovec p < value determined using the proposed model ap- 0.001 for both), respectively. After BC correction k erential ATN value. k ff p < ect on the ATN measurement result. When the ATN F2 ff value was smaller than the value of 3 ATN · F6 er significantly between the summer and the winter seasons ff ). The slopes determined for these two flow rates were 0.991 F6 F6 1 erent from 1.0 ( − ff BC BC − − F6 F2 before and after correction in the summer and winter seasons. Before 0.915 for winter). The 1 BC = − ATN · p value in the proposed model was due to F2 30.751 (95 % confidence interval (CI): k was significantly smaller than the measured BC − after correction 43.841–7.603) for the sampling flow rates of 6 and 2 Lmin . Figure 3 presents a comparison of BC corrected using the proposed model and 0.086). BC 0.163 for 2 Lmin − F6 F2 = Before beginning field sampling, the performance of the two aethalometers was com- = = p value could fall toalgorithms 3 proposed : 1. in Thatcould this be is, study. estimated the from Subsequently, the true the trueBC ATN ATN actual change values rate. BC Before could BC mass correction, be the concentration determined measured using the value increased, it deviatedthe considerably 370 nm from wavelength. the The measurement valueon results of demonstrated the that 3 ATN the value loadingAfter e at ATN short correction wavelengths using was Eqs. stronger (6)–(9), than that the at ratio long of wavelengths. the ATN the summer and winter seasons ( k by using the true ATN value,BC the corrected BC that corrected usingmodel the of correction Drinovec model etthis of al. model Drinovec (2014) can et ismeasurement be al. results presented easily of (2014). in BC estimated The Eq. and from ATN (3), correction at the sampling and flow following the rates equation of correction 6 on factor and the 2 in Lmin basis of the proached a constant value steadily whenof the the ATN value increased. The steady variation ATN value, rather than an instant di The corrected BC estimatedwith using that the obtained model using proposedsummer; the in correction this model study was ofet comparable Drinovec al. et (2014) al. exhibitedthe (2014) significant entire ( variations data set. between By negative contrast, and the positive values for the presence of the artifact e CI: the intercepts were not considerably di pared at the sampling site for two sampling flow rates, 6 and 2 Lmin from further treatment. Local meteorologicalmidity data such were as recorded temperature and using relative a hu- Vantage Pro 2 tive humidity did not( di Table 1 shows the ambient temperatureods. and The relative average humidity temperature during during the thethat sampling summer during peri- season was the significantly winter higher than season by approximately 15 The hourly measured results obtained usingThe the two hourly aethalometers are average shown BCwere in Fig. calculated mass 1. using concentrations 5were in min raw units data. of Statistical nanogram results per indicated cubic that meter the intercepts (95 % CI: 0.985–0.997) and 0.988were (95 % significantly CI: 0.980–0.997), di respectively, and the slopes p Figure 2 showsand the 2 relationships Lmin betweenATN ATN correction, values the for ATN sampling flow rates of 6 3 Results and discussion 3.1 ATN values and black carbon mass concentrations before and indicated that the performance ofsidered both to aethalometers be used similar in to each this other. study could be con- 5 5 20 25 10 15 10 15 25 20 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | for /Q (11) (12) value /Q /Q value was , as shown k cient. F2 ffi and ATN 0.33 (estimated ± k values, which was could be determined /Q and ATN was set as 2.14 for the cient amount of aerosol /Q F6 ffi C ) and 1.11 cients of aerosol at 470 and F6 cient of aerosol can be deter- is shown in Fig. 2, and it can ffi ffi F2 and ATN value increased rapidly as ATN k k value being fixed for an ATN k and ATN F6 2862 2861 was similar, especially for long wavelengths. can be used as an index to determine the type 1 − values. The α c and wood burning, respectively. A high absorption 0.44 (estimated from corrected ATN) in the summer /Q ffi are the absorption coe ± and from corrected ATN in the summer and winter sea- initially increased as a positive value and then decreased 950 = F2 λ k 950 value could be observed at small ATN = abs, λ k ) to 1.17 b 0.22 (estimated from measured ATN F2 abs, ± 470 and ATN 1 C and / /b · F6 A Q 470 470 · 950 = = λ λ ect also strongly influences the calculation of the absorption Ångström expo- ln t is a light enhancement parameter. It is associated with multiple scattering of ff abs, d abs, b dATN b C = ln Kirchstetter and Novakov (2004) noted that the absorption Ångström exponent value = abs using Eq. (9) and the relationship between measured ATN in Fig. 4. Analytical results showed that the relationship between sampling flow rates of 6 and 2 Lmin This result indicated that the assumption of the in the proposed model was reasonable.negative Analytical results at indicated extremely that the small ATN reasonable in terms of the lightsampling scattering spot, behavior especially of at the short nearly wavelengths unloadedfound in filter that the at with winter a increasing season. new In aerosolbehavior this load study, of on it was the the filter, filter the matrix influence could of the be light eliminated. scattering When a su 3.3 Absorption Ångström exponent and emissionThe source absorption of black Ångström carbon exponent gradually and steadily to aproposed constant model value. could These overcome observation the results problemscattering indicated of that at enhanced light the a ATN new resulting sampling from light spot, without using any light scattering coe increased, and then approached amultiple constant value. scattering Weingartner et in al. theTherefore, (2003) noted a nearly that negative unloaded fiber filter could enhance light absorption. was sampled on the filter, of BC emission sources (Sandradewi et al., 2008), and it is computed from the aerosol 950 nm wavelengths, respectively. The absorption coe mined from the change rate of ATN: b where where from measured ATN the light beam atthe the filter filter fibers material in (Weingartner the et unloaded al., filter, and 2003). it In is this strongly study, dependent on light absorption between 470 and 950 nm wavelengths as follows: α ues were 1.1 and 1.8–1.9 for tra measured ATN sons. Analytical results showed thatimproved the from absorption 1.01 Ångström exponent value could be ranged between 0.8 andsorption Ångström 1.1 exponent for values dieselaerosol, were which soot. between strongly 0.9 Day and depended etdradewi 2.2 on et al. for the al. (2006) fresh type (2008) wood showed of also smoke that wood demonstrated the and that burning ab- the conditions. absorption San- Ångström exponent val- nent value. Table 2 shows the absorption Ångström exponent values calculated from quartz filter (Weingartner et al., 2003; Drinovec et al., 2014). Ångström exponent value could beto observed the for compounds wood of combustionloading the aerosol e aerosol that is with due strong absorption in the UV. However, the filter be expressed as abasis power of analytical law results, relationship the relationship through between measurement data fitting. On the 3.2 Influence of light scattering behaviorThe on relationship unloaded filter between measured ATN 5 5 15 20 10 15 20 10 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | (13) ATN) value × k erent sampling 0.20 (estimated ff − ± erent aerosol deposi- ATN) was significantly ff erences in the compo- ff ect. According to the es- erent from the corrected × ff ff k values could be predicted ) and 1.10 − . For the sampling flow rate k 1 F6 erent weather conditions. − value showed a power law re- ff 0.594 for winter), indicating that before and after correction. The k = was then estimated using the pro- 1 ATN) and ATN at di p − k × k , the − /Q 2863 2864 ectively used to correct BC data for the loading 1 (14) ff 1.21 and 1.14 in the summer and winter sea- erent sampling flow rates, the (1 3 and 10 in the summer and winter seasons, re- ff + = ) and measured ATN. Finally, the corrected BC could f ) 0.24 (estimated from corrected ATN). The absorption k 10 0.734 for summer; ( ect on BC measurement results. Moreover, the aerosol 10 ± /Q > ( ff = ln ln p − c. − ) ) ffi were relatively lower than those estimated from the corrected value for small ATN for ATN 50 1 ) to 1.15 ( ATN k − ( F2 0.18 (estimated from measured ATN ln /Q ln ± ect on the unloaded filer was neglected in the post-processing model. ff 1 mass concentrations were not significantly di − ) F6 erence between seasons could be because of di f 1 ff , BC corrected using the post-processing model was comparable with that 1 ect on BC measurement results in the summer season was greater than that ATN 0.001 for both), indicating that the absorption Ångström exponent value could − is a fit parameter dependent on the aerosol type. BC corrected using the BC ff ( = ATN) approached 1.0. For di f ) R mass concentrations ( × p < black carbon data = k F2 c ered significantly for the same ATN conditions because of di ATN ected by the ATN and sampling flow rate. When the ATN was very small, the value of ect. − Figure 6 shows the relationship between (1 Figure 7 presents the results of a comparison between 5 min BC concentrations ( ff ff ff tion rates, indicating that the aerosol deposition density on the filter could influence the a di posed empirical equations in Table 3.Eq. Second, (5) the by corrected using ATN was the determined determined from (1 spectively (Fig. 5). Analyticalusing the results empirical indicated equations presented thatof in aethalometer Table the 3. AE31 Furthermore, could the beferent existed post-corrected sampling data using flow sets these rates. empiricalAE31 First, equations were the for divided dif- ATN by values the in sampling flow existed rate, data and sets of aethalometer be estimated from thelight change scattering e rate of the corrected ATN. Itflow should rates. be Analytical noted results that showed the that the value of (1 lationship with ATN Despite the negative 3.4 Simple correction scheme for post-processing for correcting be significantly underestimated because of thetimated filter absorption loading Ångström e exponentthe values, sampling the site primary was tra emission source of BC at post-processing model was comparable withof that corrected Weingartner using the et correctionsons, al. model respectively, (2003) at the for of sampling 2 Lmin flow rate of 6 Lmin extent of the aerosol loadingloading e e where measured at sampling flow rates of 6 and 2 Lmin from measured ATN Ångström exponent values computedrates using of 6 the and ATN 2ATN measured ( Lmin at the sampling flow in the winter seasonrate. by This approximately di 1–6 %, and it depended on the sampling flow BC the post-processing model could bee e 3.5 Comparison with a previous correctionFigure model 8 presents the results ofprocessing a model comparison and of the BC correction corrected modelmodel using of developed the Weingartner by proposed et Weingartner post- al. et (2003).basis al. The correction of (2003) is the widely measured used ATN value, to and correct it BC can data be on expressed the BC as R BC was corrected using thecorrected proposed post-processing BC model. Results showed that the sition, source, and age of aerosols, in addition to the di season. In the wintercreased from season, 1.01 the absorption Ångström exponent value could be in- 5 5 25 15 20 20 10 15 10 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | in 1 ects − ff erent ff 10. In other in the correc- < 1.25 and 1.15 f = f AE51 and the e r ect on BC measurement ff c site, Aerosol Air Qual. Res., 13, erent light ATN increasing rates ff ffi erent environments. ff 50. Moreover, the parameter 2865 2866 > ect and aerosol loading e ff 0.019), except for the sampling flow rate of 6 Lmin ≤ p 0.123). BC corrected using the correction model of Weingart- = 1.0. Otherwise, the BC corrected using the correction model of p The authors would like to thank the Ministry of Science and Technology > 20, especially for ATN > cult to determine from field sampling data. ffi c sites, Atmos. Environ., 90, 78–86, 2014. ffi of aerosol loading on its measurement results at a tra 1853–1863, 2013. monitoring the diurnal characteristicstra of atmospheric black carbon size distribution at urban Park, K., Arnott, W. P., andtal Motallebi, N.: carbon Aerosol at light the absorption, Fresno black Supersite, carbon, California, and Atmos. elemen- Res.,tje, 93, 874–887, H., 2009. Henzing,tensperger, J. U.: S., Minimizing light Jennings, absorptionuation S. measurement of five artifacts G., correction of algorithms, Moerman, the Atmos.2010, Meas. aethalometer: M., 2010. Tech., 3, eval- Petzold, 457–474, doi:10.5194/amt-3-457- A., Schmid, O., and Bal- light-absorption measurements with a 7-wavelengthcoustic aethalometer: instrument evaluation and with 3-wavelength a nephelometer, photoa- Aerosol Sci. Tech., 39, 17–29,measurements 2005. of visible light absorption by aerosols, Aerosol Sci. Tech., 30, 582–600,ner, 1999. M. G., Ghan, S., Karcher,Schultz, B., Koch, M. D., Kinne, G., S.,tikunda, Kondo, Schulz, Y., Quinn, S. P. M., K., K., Sarofim, Venkataraman,Schwarz, M. Hopke, C., C., J. P., P. Zhang, Shindell, K., D.,role H., Storelvmo, Jacobson, of black Zhang, T., carbon M. Warren, in S., S. the Z.,118, G., climate Bellouin, 5380–5552, system: Kaiser, and 2013. a N., Zender, J. scientific C. assessment, Gut- W., J. S.: Geophys. Klimont, Bounding Res.-Atmos., the Z., Lohmann, U., aerosol deposition rates. Moreover,scheme this for study post-processing provided for correcting athis the simple simple existed correction aethalometer empirical scheme BC correction is data.rect dependent Although BC on data the when aerosol thethose type, primary in it this source can study. of Moreover, be two BC usedcan existed and to be aethalometers the cor- under used weather appropriate to conditions flow create are control correction similar schemes to for di to improve the measurement ofcan aerosol overcome the black light carbon. scattering The e proposed correction model Cheng, Y. H., Liao, C. W., Liu, Z. S., Tsai, C. J., and Hsi,Chow, J. H. C., C.: Watson, A J. size-segregation G., method Doraiswamy, P., Chen, for L. W. A., Sodeman, D. A., Lowenthal,Collaud D. Coen, H., M., Weingartner, E., Apituley, A., Ceburnis, D., Fierz-Schmidhauser, R., Flen- results simultaneously, and it can bethe used actual in a BC newly concentration designed instrument in to real determine time by using two sampling spots under di Acknowledgements. Bond, T. C., Anderson, T. L., and Campbell, D.: CalibrationBond, and T. intercomparison C., of Doherty, filter-based S. J., Fahey, D. W., Forster, P. M., Berntsen, T., DeAngelo, B. J., Flan- Cheng, Y. H. and Lin, M. H.: Real-time performance of the microAeth of the Republic ofMOST China, 102-2221-E-131-002-MY3. Taiwan, for financially supporting this research under Contract No. References Arnott, W. P., Hamasha, K., Moosmüller, H., Sheridan, P. J., and Ogren, J. A.: Towards aerosol This study developed a new method based on two di ner et al. (2003)model was significantly when lower ATN than wascorrection that small, model corrected and of using this Weingartner the et is al. post-processing possibly (2003) was because larger the than function 1.0 for R(ATN) ATN in the 4 Conclusions in the summer and winter seasons,processing respectively. However, model BC was corrected using slightly the higher post- Weingartner than et that al. corrected (2003) using the ( the correction winter model season of ( tion model of Weingartner etand it al. was (2003) di is significantly dependent on the aerosol type, corrected using the correction model of Weingartner et al. (2003) for Weingartner et al.model (2003) for was ATN higher than that corrected using the post-processing words, the corrected BCmodel was for smaller R(ATN) than that corrected using the post-processing 5 5 25 20 20 25 30 10 15 15 10 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | c- ffi c-related ffi erent time scales cient by means of ff ffi c emission contributions to particulate matter, ffi ects of aethalometer data, J. Air Waste Manage., ff 2868 2867 , S. E., Coull, B. A., Spiro., A., and Schwartz, J.: Tra ff ects of controlling fossil-fuel soot, biofuel soot and gases, and erent environments: variation of the specific attenuation cross- ff ff , H., Schnaiter, M., Streit, N., Bitnar, B., and Baltensperger, U.: Ab- ff ected by organic carbon, J. Geophys. Res., 109, D21208, 2004. ff nik, G., Zotter, P., Prévôt, A. S. H., Ruckstuhl, C., Coz, E., Rupakheti, M., č related air pollution and119, cognitive 682–687, function 2011. in a cohort of older men,Nat. Environ. Geosci., Health 1, Persp., 221–227, 2008. parisons of techniquesaerosols for from measuring biomass burning shortwave in Brazil, absorption J. and Geophys. Res., blackner, 103, E., 32031–32040, carbon and 1998. content Baltensperger, U.: of tative Using determination aerosol of light wood absorption burningEnviron. measurements and Sci. for Technol., tra 42, the 3316–3323, quanti- 2008. black carbon exposure and lung1333–1337, function 2008. among urban women, Environ. Health Persp., 116, A simple procedure for correcting loading57, e 1214–1222, 2007. carbon analysis methods and intercomparisons, Aerosol Air Qual. Res.,sorption 5, 65–102, of 2005. light byaethalometers, J. soot Aerosol particles: Sci., 34, determination 1445–1463, of 2003. the absorption coe carbon aerosol content in di section, sigma, Atmos. Environ., 27, 1203–1211, 1993. Bowers, K., Bode, R., Petzold, A.,based and carbon Hansen, A. measurements D. at A.:2006. Comparison the of Fresno continuous Supersite, and J. filter- Air Waste Manage., 56,section 474–491, on black carbon mass fraction and particle size, Atmos. Environ., 31, 661–672, 1997. in Helsinki during 1996–2005, Atmos.2008, Chem. 2008. Phys., 8, 1017–1027, doi:10.5194/acp-8-1017- by aerosols is a methane on climate, Arctic,doi:10.1029/2009JD013795 2010. ice, and air pollution health, J.Kulmala, Geophys. M.: Res., Temporal 115, variations D114209, in black carbon concentrations with di Sciare, J., Müller, T., Wiedensohler, A.,an and improved Hansen, A. measurement D.Atmos. of A.: Meas. The aerosol Tech. “dual-spot” Discuss., aethalometer: black 7, 10179–10220, carbon, doi:10.5194/amtd-7-10179-2014 with 2014. real-timetime loading measurement compensation, of optical196, absorption 1984. by aerosol particles, Sci. Total Environ., 36, 191– Divjan, A., Olmedo, d. O.Goldstein, E., I. Lopez-Pintado, F., Rundle, S., A. Kinney,exhaled G., P. L., nitric and Perera, oxide Perzanowski, F. in M. P., Jacobson, children S.: J. in Domestic S., NYC, airborne J. black Expo. carbon Sci.from and Env. laboratory Epid., studies 22, of 258–266, fresh2006. 2012. smoke from biomass fuels, J. Geophys. Res., 111, D22202, Reid, J. S., Hobbs, P. V., Liousse, C., Vanderlei Martins, J., Weiss, R. E.,Sandradewi, and J., Eck, Prévôt, T. A. F.: Com- S. H., Szidat, S., Perron, N., Alfarra, M. R., Lanz, V.Suglia, A., S. Weingart- F., Gryparis, A., Schwartz, J., and Wright, R. J.: Association between tra Watson, J. G., Chow, J. C., and Chen, L. W. A.: Summary of organic and elemental carbon/black Virkkula, A., Mäkelä, T., Hillamo, R., Yli-Tuomi, T., Hirsikko, A., Hämeri, K., and Koponen, I. K.: Weingartner, E., Saatho Ramanathan, V. and Carmichael, G.: Global and regional climate changes due to black carbon, Park, K., Chow, J. C., Watson, J. G., Trimble, D. L., Doraiswamy, P., Arnott, W. P., Stroud, K. R., Petzold, A., Kopp, C., and Niessner, R.: The dependence of the specific attenuation cross- Power, M. C., Weisskopf, M. G., Alexee Liousse, C., Cachier, H., and Jennings, S. G.: Optical and thermal measurements of black Kirchstetter, T. W. and Novakov, T.: Evidence that the spectral dependence of light absorption Järvi, L., Junninen, H., Karppinen, A., Hillamo, R., Virkkula, A., Mäkelä, T., Pakkanen, T., and Hansen, A. D. A., Rosen, H., and Novakov, T.: The aethalometer – anJacobson, instrument M. for Z.: the Short-term real- e Cornell, A. G., Chillrud, S. N., Mellins, R. B., Acosta, L. M., Miller, R. L., Quinn,Day, J. D. W., Yan, E., B., Hand, J. L., Carrico, C. M., Engling, G.,Drinovec, and L., Malm, Mo W. C.: Humidification factors 5 5 10 15 20 25 30 20 15 10 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 0.44 0.24 ± ± SD Min–Max corrected ATN ± F2 1314 37–93 34–91 ± ± 0.330.20 1.17 1.15 ± ± measured ATN 2870 2869 C Relative Humidity, % ◦ F6 SD Min–Max Average ± 0.18 1.10 0.22 1.11 ± ± 3 10–28 69 3 25–39 68 ± ± measured ATN Average SD ± Winter 1.01 SeasonSummer 1.01 Ångström exponent estimated from Average Winter 17 SeasonSummer Temperature, 32 Absorption Ångström exponent values in the summer and winter seasons. Ambient temperature and relative humidity during the sampling periods. Table 2. Table 1. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

values. k b 0.6768 0.6835 0.6968 0.7532 0.8370 0.5664 0.6623 0.8383 0.7602 0.8127 0.7843 0.7491 0.8099 0.8569 Q − − − − − − − − − − − − − − ATN × a = entration measurements obtained with entration measurements a b λ k

27 2872 2871 590 nm660 nm880 nm950 nm 0.0089 0.0087 0.0087 0.0101 470 nm520 nm 0.0100 0.0092 470 nm520 nm590 nm660 nm 0.0168 880 nm 0.0167 950 nm 0.0149 0.0132 0.0149 0.0148 Winter 370 nm 0.0088 Season Wavelength Summer 370 nm 0.0263 Comparison of hourly BC mass concentration measurements obtained with the two Parameters of empirical equations used for predicting the Figure 1. aethalometers used in this study. Table 3.

-1 (b)

winter season. summer season and (b) the proposed model with that corrected the proposed model for sampling flow rates of 6 and 2 L min flow rates of 6 and 2 L for sampling . (2014) for the (a) summer season and (b) . (2014) for the (a) summer (a) 28 29 2874 2873 summer season and

(a) (a) (a) values Figure 2. Relationship between ATN (b) (b)

before and after correction for the (a) summer season and (b) winter season. before and after correction for the (a) summer of BC corrected using Figure 3. Comparison (b) (b)

using the correction model of Drinovec et al using the correction model winter season. winter season. 2 1 3 4 5 6 2 1 3 4 5 6 7 Relationship between ATN values for sampling ﬂow rates of 6 and 2 Lmin Comparison of BC corrected using the proposed model with that corrected using the Figure 3. correction model of Drinovec et al. (2014) for the and after correction for the Figure 2. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | winter (b) 3 in the summer /Q >

summer season and when ATN (a) (a) /Q for the N/Q for the (a) season and (b) summer k and ATN/Q (a) when ATN/Q > 3 in the (a) when ATN/Q k and ATN/Q 30 31 /Q and ATN 2876 2875 k and ATN k 10 in the winter season.

/Q > (a) (a) (b) (b) Figure 4. Relationships between k and AT

winter season. winter season. (b) (b) Figure 5. Power law relationship between > 10 in the winter season. ATN/Q summer season and (b) when

erent sampling ﬂow rates in the ﬀ summer season and (a) concentrations measured at sampling at sampling measured concentrations N) and ATN at different sampling flow sampling at different N) and ATN 32 33 before and after correction for the (a) summer (a) summer for the after correction before and -1 2878 2877 ATN) and ATN at di × k and 2 L min and 2 L − -1

winter season. (a) (a) (b) (b) (b) Figure 6. Relationship between (1 – k × AT rates in the (a) summer season and (b) winter season. winter season. season and (b) (a) summer rates in the

(b) (b) season and (b) winter season. BC mass of the 5-min Figure 7. Comparison min flow rates of 6 L

2 3 4 5 6 1 4 5 6 7 2 3 1 before and after correction for the 1 − Relationship between (1 Comparison of the 5 min BC mass concentrations measured at sampling ﬂow rates summer season and Figure 7. of 6 and 2 Lmin (a) Figure 6. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | in 1 − 2 Lmin (c) in the winter -1 oposed postprocessing in the winter season. season. in the winter ingartner et al. (2003) for -1 in the winter season, 1 −

6 Lmin (d) (b) obtained using the pr 34 (b) 2879 in the summer season, (b) 6 L min 6 L season, (b) in the summer -1 in the winter season. 1 − ing the correction model of We 2 Lmin in the summer season, and (d) 2 L min season, and (d) 2 L in the summer -1 (d) in the summer season, 1 −

6 Lmin Comparison of the corrected BC obtained using the proposed postprocessing model season, (c) 2 L min L 2 season, (c) (c)

Figure 8. Comparison of the corrected BC model with that obtained us (a)

sampling flow rates of (a) 6 L min flow rates of (a) 6 L sampling (a) 5 2 3 1 6 7 8 4 Figure 8. with that obtained usingrates the of correction model of Weingartner et al. (2003) for sampling ﬂow the summer season, and