Estimation of the concentrations of primary and secondary organic carbon in ambient particulate matter: Application of the CMB-Iteration method

Estimation of the concentrations of primary and secondary organic carbon in ambient particulate matter: Application of the CMB-Iteration method

Atmospheric Environment 45 (2011) 5692e5698 Contents lists available at ScienceDirect Atmospheric Environment journal homepage: www.elsevier.com/loc...

248KB Sizes 3 Downloads 53 Views

Atmospheric Environment 45 (2011) 5692e5698

Contents lists available at ScienceDirect

Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv

Estimation of the concentrations of primary and secondary organic carbon in ambient particulate matter: Application of the CMB-Iteration method Guo-Liang Shi a, Ying-Ze Tian a, Yu-Fen Zhang a, *, Wen-Yuan Ye a, Xiang Li a, b, Xue-Xi Tie c, d, Yin-Chang Feng a, *, Tan Zhu a a

State Environmental Protection Key Laboratory of Urban Ambient Air Particulate Matter Pollution Prevention and Control, College of Environmental Science and Engineering, Nankai University, Tianjin 300071, China Department of Computer Science, University of Georgia, Athens, GA, USA c National Center for Atmospheric Research, Boulder, CO, USA d Institute of Earth and Environment, Chinese Academy of Science, Xi’an, China b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 14 April 2011 Received in revised form 14 July 2011 Accepted 18 July 2011

A new method (the CMB-Iteration method) was developed and applied to estimate primary organic carbon (POC) and secondary organic carbon (SOC) concentrations in ambient particulate matter. In addition to the concentrations, this model also calculates the source contributions to POC and SOC. For each source category, the estimated source contribution is continuously calculated until the iteration minimum is achieved, i.e., the ratio of the estimated source contribution at the final step iteration to previous step iteration is less than 0.01. The application of this method is used to analyze a reported database. The result of the CMB-Iteration method is consistent with in the reported result in the literature. Additionally, synthetic datasets were analyzed using the CMB-Iteration method, and acceptable results were obtained. Finally, this method is used for a dataset collected from Taiyuan, China across different seasons (winter and summer). The calculated concentrations of SOC are 14.3 and 10.7 mg/m3 in winter and summer, respectively, suggesting that there are high secondary OC particles in this highly polluted city (Taiyuan). Ó 2011 Elsevier Ltd. All rights reserved.

Keywords: CMB-Iteration method Secondary organic carbon Source apportionment Receptor model

1. Introduction Atmospheric aerosol is one of the major pollutants in urban areas and contains a significant fraction of carbonaceous material. Carbonaceous species are generally grouped into two main fractions, elemental carbon (EC) and organic carbon (OC) (Schauer et al., 1999; Pandis et al., 1992). EC is essentially a primary pollutant, generated directly during the incomplete combustion of fossil and biomass carbonaceous fuels. OC can be directly emitted during combustion as well as from other sources (referred to as primary OC (POC)) or produced from atmospheric chemical reactions involving gaseous organic precursors (referred to as secondary OC (SOC)). SOC is produced either from gas to particle conversion due to the condensation of low vapor pressure products, when concentrations exceed saturation levels, or from physical/chemical adsorptive processes of gaseous species on particle surfaces, which can happen at

* Corresponding authors. Tel./fax: þ86 22 23503397. E-mail addresses: [email protected] (G.-L. Shi), [email protected] (Y.-F. Zhang), [email protected] (Y.-C. Feng). 1352-2310/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.atmosenv.2011.07.031

sub-saturation levels (Pankow,1987; Plaza et al., 2006). Photochemical pollution became severe in recent years, and SOC has contributed significantly to the fine particulate matter (PM) in the atmosphere (Yu et al., 2007; Ding et al., 2008). However, there is a lack of direct measurement methods for POC and SOC. Several methods have been applied to estimate SOC concentrations: 1) empirical methods, mainly based on measurement, such as the OC/EC ratio method; 2) environmental models, such as chemical mass balance (CMB) models; and 3) semi-empirical methods, combining measurements with environmental models (Ding et al., 2008; Odum et al., 1996; Kleindienst et al., 2006; Takahama et al., 2006; Lane and Pandis, 2007; Lewandowski et al., 2008; Murphy and Pandis, 2009; Pachon et al., 2010). For the calculation of the CMB model, some researchers used this receptor model to study the SOC concentration (Marmur et al., 2005; Chen et al., 2010). However, some issues arise when using the CMB model to estimate SOC concentration. First, the CMB model needs source profiles, and there was no actual SOC profile. Thus, we cannot obtain the SOC contribution by using the CMB model without information from the SOC profile. Second, in previous studies, a “pure” SOC source profile was assumed and introduced into the CMB model (Marmur et al., 2005; Chen et al., 2010).

G.-L. Shi et al. / Atmospheric Environment 45 (2011) 5692e5698

However, this method led to collinearity problems with OC dominant sources (such as vehicle exhaust) (Lee et al., 2007). In this study, a new method is proposed to study SOC by using the iterative method of the CMB model (the CMB-Iteration method). For this model, SOC and POC can be estimated without adding information from the SOC profile into the CMB model. In addition to the estimate of SOC concentrations, the source contributions can be estimated simultaneously. This model will be tested in this study by using ambient (PM2.5 and PM10) datasets. 2. The CMB-Iteration method 2.1. Principle of the traditional CMB model The chemical mass balance model is a widely used receptor model (Watson et al., 1984; Watson et al., 2002; Chen et al., 2007). The CMB model establishes a balance between sources and the receptor, and it can be used to estimate the source contributions to the receptor (USEPA, 2004; Chow et al., 2007; Watson et al., 2008). The CMB model has been applied to estimate the primary emission source contributions to organic carbon (OC) (Lee et al., 2007; Schauer et al., 1996; Zheng et al., 2002, 2007; Yan et al., 2009; Lee et al., 2009). The CMB model can be described as the following equation:

Cðn1Þ ¼ FðnmÞ *Sðm1Þ

(1)

where C(n  1) is the receptor species concentration vector (mg/m ); F(n  m) is the source profile matrix (mg/mg); S(m  1) is the source contribution vector (mg/m3); n is the number of species measured; m is the number of source categories; and 1 is the number of receptor samples. If we account for the contribution of SOC separately, Eq. (1) can be rewritten as:

5693

POC* ¼ TOC  SOC

(6)

In Eq. (6), the concentration of SOC is unknown and needs to be estimated by the CMB-Iteration method. 2.3. The procedure of the CMB-Iteration method The calculation procedure of the CMB-Iteration method is shown below. In the procedure, the superscript k represents the kth iteration. 1) As discussed above, to introduce the revised receptor into the CMB8.2 model the SOC concentration should be removed from the original receptor. However, the SOC concentration is initially unknown. Thus, the SOC should be set to zero at the beginning of the iteration:

SOC0ðn1Þ ¼ 0

(7)

2) Next, the revised receptor and source profiles are introduced into the CMB8.2 model. The relationship of the revised receptor and sources is established at the kth iteration,

ðC*Þkðn1Þ ¼ FðnmÞ  Skðm1Þ

(8)

3

Cðn1Þ ¼ FðnmÞ *Sðm1Þ þ SOCðn1Þ where SOC(n expressed by

 1)

SOCðn1Þ ¼ ½ OC

(2)

is the profile of secondary OC, which can be

0

0

.

0 Tðn1Þ

(3)

where OC in the vector is the concentration of SOC (mg/m3). The superscript “T” calls for the transposition of the vector, and “0” is the concentrations (mg/m3) for the species excluding OC. According to Eq. (3), if no SOC profile was introduced into the CMB model, the sources and receptor in Eq. (1) would be incompatible, and the SOC concentration cannot be estimated. 2.2. CMB-Iteration method description For the CMB-Iteration method, we do not attempt to use the SOC profile in the CMB model. Thus, the relationship of the original receptor and sources is established as follows:

Cðn1Þ  SOCðn1Þ ¼ FðnmÞ *Sðm1Þ

(4)

where C(n1) is the original receptor, SOC(n1) is the SOC receptor, F(nm) is the source profile, and S(m1) is the source contributions vector. In Eq. (4), the left term of C e SOC represents the concentration * of primary OC, which is expressed by the revised receptor of Cðn1Þ and expressed by Eq. (5); * Cðn1Þ ¼ FðnmÞ *Sðm1Þ

(5)

If we use TOC (mg/m3) to represent the total OC and to replace * , the calculated POC*(mg/m3) can be expressed by: Cðn1Þ

3) The source contributions skj at the kth iteration are estimated by the CMB model (USEPA, 2004). In this step, the revised k

receptor ðC * Þn1 and source profile matrix F(nm) (excluding SOC profile) are introduced into the CMB model. skj is the estimated contribution of the jth source to the revised receptor. For each iterative solution, the performance indices of the result should meet the requirement of the CMB model (USEPA, 2004). 4) The POC* concentration is estimated at the kth iteration. In the studies of Zheng et al. (2002, 2007), PM mass apportionment can be calculated from POC source apportionment results based on the ratio of POC to total particle mass from source tests (Zheng et al., 2002, 2007).

OCj ¼ Sj  ðOC=TOTÞj

(9)

where OCj is the contribution (mg/m3) of the jth primary emission source to total OC in the receptor; Sj is the estimated contribution (mg/m3) of the jth primary emission source; (OC/ TOT)j is the ratio of OC to total particle mass in the jth source profile (mg/mg).

OCprimary ¼

j X

OCm

(10)

m¼1

where OCm is the contribution of the mth source to OC and j is the number of source categories. 5) The SOC concentration is estimated at the kth iteration. The kth iterated concentration of SOC can be estimated as follows (Russell and Allen, 2004):

SOCk ¼ TOC  POC*k

(11)

where, TOC is the concentration of total OC in original receptor and SOCk is the estimated SOC concentration at the kth iteration.

5694

G.-L. Shi et al. / Atmospheric Environment 45 (2011) 5692e5698

6) The iteration ratios are calculated for sj, POC*, and SOC at the kth iteration against those at the (k  1)th iteration (USEPA, 2004) by following the steps:

    k   if  skj  sk1 sj >0:01;  POCk  POCk1 POCk >0:01; j   j SOCk  SOCk1 SOCk j>0:01; (12) and return to step 2 for next iteration; (j ¼ 1, 2, 3.m);

    k   if  skj  sk1 sj  < 0:01;  POCk  POCk1 POCk  < 0:01; j   (13) j SOCk  SOCk1 SOCk j < 0:01; and then the iteration is stopped. 7) The final results are estimated, and as a result skj is the estimated source contribution, POC*k is the estimated POC concentration, and SOCk is the estimated SOC concentration. The CMB-Iteration method is based on the EPACMB 8.2 model. The source contributions are calculated by effective variance weighted least squares solution. So, the same to EPACMB 8.2 model (Watson et al., 2001; Christensen and Gunst, 2004), the source contributions solution of CMB-Iteration method is approximately unbiased.

3. Testing of the CMB-Iteration method 3.1. Application of CMB-Iteration method for a reported dataset In this study, a set of ambient receptor and source datasets were applied to study the results of the CMB-Iteration method. These actual ambient and source profiles were provided by Marmur et al. (2005). The PM2.5 ambient dataset was from Atlanta, GA. The source categories used in this work were the following: LDGV e light-duty gasoline vehicles; HDDV e heavy-duty diesel vehicles; SDUST e fugitive soil dust; BURN e vegetative burning; CFPP e coal-fired power plants; and AMSULF e ammonium sulfate. In this test, the iterative solution was achieved by the 5th iteration. The results of the CMB-Iteration method in each iterative solution are shown in Table S1 in the Supplementary Material File. From the 1st to 4th iterations, the ratios of results at the kth iteration to those at the (k  1)th iteration were higher than 0.01 (the ratios were calculated and shown in Table S1b), which indicates that the iterative processing should be continued. At the 5th iteration, SOC4 was 2.46 mg/m3 at the beginning of the iteration (obtained at the 4th iteration); and the (POC*)5 is TOC2.46 mg/m3. After this iteration, the iteration ratios in Table S1b are all less than 0.01, which show that the iteration minimum has been reached, and the results at the 5th iteration are considered to be the final results of the CMB-Iteration method. Additionally, the performance indices (c2, R2, PM) (listed in Table S1c) for each iteration meet the requirement of the CMB model, which indicate that the results at each iteration can be accepted by the CMB model. The final results of the CMB-Iteration method are listed in Table 1. The estimated contributions from different sources to the concentrations of POC and SOC are LDGV d1.57 mg/m3, HDGV d1.96 mg/m3, SDUST d 0.28 mg/m3, BURN d 1.07 mg/m3, CFPP d 0.15 mg/m3, AMSULF d 8.00 mg/m3, AMNITR d 1.46 mg/m3, and SOC d2.47 mg/m3. In the study of Marmur et al. (2005), the dataset was studied by another method (CMB-LGO: CMB-Lipschitz global optimizer). The results of the two different models are compared in Table 1. The

Table 1 Results of different methods for the Ambient dataset given by Marmur et al. (2005). Source

LDGV HDDV SDUST BURN CFPP AMSULF AMNITR SOC

CMB-Iteration method

CMB-LGO*

Contribution (mg/m3)

Contribution (mg/m3)

1.57 1.96 0.28 1.07 0.15 8.00 1.46 2.47

 1.77  1.73  0.49  1.19  0.33  3.84  1.11  2.22

1.28 1.96 0.39 1.13 0.15 7.03 1.60 2.59

 0.90  1.63  0.48  0.69  0.12  5.12  1.34  1.64

LDGV: light-duty gasoline vehicles; HDDV: heavy-duty diesel vehicles; SDUST: fugitive soil dust; BURN: vegetative burning; CFPP: coal-fired power plants; AMSULF: ammonium sulfate; AMNIT: ammonium nitrate; SOC secondary other OC. The results of CMB-LGO (Lipschitz global optimizer) were given in the reference (Marmur et al., 2005).

comparison indicates that the result of the CMB-Iteration method was consistent with the CMB-LGO model. In addition to the CMB-Iteration method, the traditional CMB model is used to study this dataset. In the traditional CMB model, we introduced a pure SOC profile (for the pure SOC profile, OC fraction is 100%, and other species fractions are 0%) into the model. However, the collinearity problem still remains. The source profiles and receptor significantly affect the CMB results and lead to large uncertainties (Lee and Russell, 2007). In the Supplementary Material File, the collinearity problem among the source profile is discussed in Table S2. The result shows that the collinearity problem would be present when adding the SOC profile into the CMB model. The estimated source contributions, including the pure SOC profile, were also presented in Table S3. 3.2. Application of CMB-Iteration method for synthetic datasets Additionally, the CMB-Iteration method was applied to calculate the synthetic datasets. For each synthetic dataset, the source contribution values were known (can be considered as true values). The estimated contributions would be compared with those true values. (1) Synthetic Receptor Dataset Development The synthetic data were generated according to Eq. (1). The synthetic receptor dataset was composed of chemical species contributions from eight source categories: LDGV, HDDV, SDUST, BURN, CFPP, AMSULF, AMNITR (these seven sources were given by Marmur et al. (2005)), and a “pure” SOC source profile. The simulated eight source contributions are compared to the results in Table 1: LDGV e 1.57 mg/m3, HDDV e 1.96 mg/m3, SDUST e 0.28 mg/m3, BURN e 1.07 mg/m3, CFPP e 0.15 mg/m3, AMSULF e 8.00 mg/m3, AMNITR e 1.46 mg/m3, and SOC e 2.47 mg/m3. The method for the synthetic dataset was developed according to a previous methodology (Javitz et al., 1988; Lowenthal et al., 1992). The source and receptor profiles were disturbed within the range of random uncertainties. The random uncertainties were generated from a lognormal distribution. The coefficient of variation (CV) is applied to calculate the uncertainties of the source profiles and the receptor concentrations. The CV is defined as the uncertainty of the species concentration divided by the mean of the species contribution (Javitz et al., 1988). This method for synthetic receptors development was described clearly in a previous methodology (Javitz et al., 1988; Lowenthal et al., 1992). According to Javitz et al. (1988), several simulation runs were conducted in the synthetic data tests. In each simulation run, 100

G.-L. Shi et al. / Atmospheric Environment 45 (2011) 5692e5698

days data were generated. And for each source, the simulated source contribution was set at a constant value (Javitz et al., 1988). The CVs of the source profile were set to a constant value, and the CVs of receptor were set to another constant value in each simulation run. The values of the CVs were also compared to those in Javitz et al. (1988).

100 L/min and run continuously for 24 h. Two parallel mediumvolume air samplers were used to obtain PM10 data with polypropylene membrane filters and quartz fiber filters. The sampling process was referred to in the literature (Shi et al., 2009; Bi et al., 2007). 4.3. Source sampling

where AAEj is the value for the jth source category; n is the number of the days of data subjected to the model (in this study, n ¼ 100) for each simulation run; Ei is the estimated contribution for a particular source on the ith day; and Ti is the true contribution for a particular source on the ith day. The AAE values for eight sources are listed in Table S4. According to the study of Javitz et al. the CMB yields acceptable accuracy and precision (an AAE of 50% or less) even when the CV of the source profiles is 25% and the CV of the measurement error is 10%, and AAEs less than 50% would represent acceptable precision. The AAE values in Table S4 were all less than 50%, which indicates that the results from the CMB-Iteration method are acceptable. From the above analysis, the CMB-Iteration method can give reasonable results compared to other methods for calculating POC and SOC. In the next section, this model will be applied to the measurements collected in a polluted city, Taiyuan, China in summer 2001 and winter 2002. 4. Application of CMB-Iteration method in Taiyuan In this study, the CMB-Iteration method is applied to measurements, which were made in a polluted city, Taiyuan, China in the summer and winter of 2001 and 2002, respectively. During two sampling campaigns, 30 PM10 samples were obtained from August to September 2001, and 35 PM10 samples were obtained from December 2001 to January 2002. 4.1. Sampling site description

fraction (g/g)

0.4 0.3

coal combustion

0.2 0.1 0.0 0.3

fraction (g/g)

(14)

According to the previous studies and emission inventories, seven potential source categories are identified for the CMB model,

crustal dust 0.2

0.1

0.0 0.4

fraction (g/g)

The synthetic datasets were analyzed using CMB-Iteration method. The SOC profile was not introduced into the model. In each simulation run, 100 results were obtained. To compare the estimated results with the true value and to evaluate the accuracy of the obtained estimates, the average absolute error (AAE) for each of the source categories was calculated for each simulation run (Javitz et al., 1988; Lowenthal et al., 1992).

0.3

cement dust

0.2 0.1 0.0 0.4

fraction (g/g)

(2) Results for synthetic datasets

n 1 X AAEj ¼  ðjE  Ti j=Ti Þ n i¼1 i

5695

0.3

steel manufacture

0.2 0.1



4.2. Particulate matter sampling The ambient PM10 samples were collected by filtration with a medium-volume air sampler (Wuhan Tianhong Intelligence Instrumentation Facility, TH-150 Medium Volume Sampler) situated about 5 m from the ground. The pump was set at a rate of

0.0 0.5 0.4

fraction (g/g)

Taiyuan (110 300 -113 090 E, 37 270 - 38 250 N) is the capital city of the Shanxi province and is located in northern China and has a population of approximately 3.4 million. The climate belongs to the continental temperate climate. It is dry and sandy in spring and is rainy in summer. Taiyuan is one of the Chinese oldest national heavy industrial regions and is one of the important coal production cities in China. The main industries in Taiyuan are energy, metallurgy, machinery, and chemical industry.

vehicle exhaust

0.3 0.2 0.1 0.0

Na Al P Ca V Mn Ni Cu Br Pb EC Cl- SO42Mg Si K Ti Cr Fe Co Zn Ba OC NH4+ NO3Fig. 1. Source profiles in Taiyuan.

5696

G.-L. Shi et al. / Atmospheric Environment 45 (2011) 5692e5698

including crustal dust, coal combustion, cement, vehicle exhaust, steel manufacturing, secondary sulfate, and nitrate. Samples of the crustal dust source were swept from representative portions of the ground surface with a plastic brush and tray (Baldwin et al., 1994). Coal combustion was collected from particulate pollution control devices (electrostatic precipitators, fabric filters, or wet scrubbers) or sampled by a dilution stack sampler. Cement source samples were obtained from a cement plant and building construction sites (Bi et al., 2007). Steel manufacturing dust and vehicle exhaust dust were sampled from the steel industries and exhaust pipes, respectively. All powder samples were sieved and suspended in a resuspension chamber or separated by a Bahco centrifugal instrument for PM10 (Bi et al., 2007). 4.4. Chemical analysis TC (total carbon), EC (element carbon), and OC (organic carbon) were analyzed by a carbon elemental analyzer (VarioE1, Elementar Analysensysteme Gmbh, Hanau, Germany)(Bi et al., 2007). To analyze TC, the carbon species on the clip were first oxidized into CO2 at 980  C for 90 s and then reduced to 600  C in 1 atmosphere of oxygen. The quantity of TC was determined through the detection of CO2 by a thermal conductivity detector (TCD). The analysis procedure of the OC fraction was similar to the TC analysis with the only difference of an oxidizing temperature of 450  C (when the temperature reached 450  C the oxygen was provided). In the analysis process, helium gas was used as the carrier gas. EC is derived as the difference between TC and OC (Duan et al., 2005). To construct the receptor and source profiles, other species were also analyzed. A range of elements (Na, Mg, Al, Si, P, K, Ca, Ti, V, Cr, Mn, Fe, Ni, Co, Cu, Zn, Br, Ba, and Pb) was analyzed by ICP (IRIS Intrepid II, Thermo Electron) (Baldwin et al., 1994; Watson et al.,  2  1999). Water-soluble NHþ 4 , Cl , NO3 , and SO4 were extracted by an ultrasonic extraction system (AS3120, AutoScience) and analyzed using ion chromatography (DX-120, DIONEX) (Carvalho et al., 1995; Chow and Watson, 1999). These methods of sampling, treatment and analysis of source and ambient samples were referred to in prior work (Shi et al., 2009). The potential source categories and the chemical profiles of selected sources are shown in Fig. 1. 5. Results and discussion 5.1. PM10 and OC contributions The species concentrations in two ambient receptors are listed in Table 2. Reported values of PM10 and OC concentrations in other cities in China are listed in Table S5 of the Supplementary Material File (Cao et al., 2004; Duan et al., 2005; Gnauk et al., 2008; Ho et al., 2003). The values of PM10 and OC concentrations in this study are higher than those in the cities listed in Table S5. Compared with other cities in China, the values of PM10 and OC concentrations in different seasons in this study are closer to those in Beijing and Guangzhou. 5.2. Estimating the source contributions, POC and SOC concentrations The final results of summer and winter samples were shown in Table 3. During summer, crustal dust (29.29 mg/m3) had the highest contribution to PM10. The other source contributions to the concentration of PM10 were coal combustion (27.18 mg/m3), secondary sulfate (18.55 mg/m3), vehicle (17.08 mg/m3), cement dust (15.50 mg/m3), steel manufacturing (9.71 mg/m3), and secondary

Table 2 Ambient Receptors (mg/m3) in Taiyuan.

Na Mg Al Si P K Ca Ti V Cr Mn Fe Ni Co Cu Zn Br Ba Pb OC EC NHþ 4 Cl  NO3 SO4 2 total mass

Summer

Winter

1.77 2.79 6.80 14.35 0.89 2.89 9.00 0.33 0.04 0.12 0.62 3.78 0.05 0.06 0.46 0.79 0.23 0.20 0.23 25.89 6.82 5.55 0.56 4.01 19.57 146.36

2.84 3.35 9.82 21.00 1.35 3.53 11.07 0.53 0.07 0.27 0.85 4.83 0.06 0.09 0.73 0.57 0.65 0.43 0.53 40.98 12.07 9.47 3.99 8.14 25.94 214.62

nitrate (4.58 mg/m3). The estimated POC concentration was 15.21 mg/m3. The SOC concentration was 10.68 mg/m3, which shows a large contribution to the total concentration of PM10. During winter, coal combustion was the largest contributor to the concentration of PM10 (48.67 mg/m3). This result is because coal combustion is used in residential heating during winter in northern China. The other source contributions to the concentration of PM10 were crustal dust (38.25 mg/m3), vehicle (31.88 mg/m3), secondary sulfate (26.75 mg/m3), cement dust (17.17 mg/m3), steel manufacturing (11.16 mg/m3), and secondary nitrate (9.54 mg/m3). The estimated POC concentration was 26.66 mg/m3, and the SOC concentration was 14.32 mg/m3. Table 4 describes the characteristics of the calculated concentrations of TOC, POC, and SOC in different seasons in Taiyuan. The estimated SOC concentration in winter was relatively higher than that in summer. However, the ratio of SOC/TOC in summer was higher than in winter, showing that the formation of SOC formation is more active in summer than in winter due to a higher photochemical activity in

Table 3 Results of Ambient Receptor from Taiyuan. Sources

Contributions (mg/m3)

s

to PM10

to TOC

to PM10

to TOC

Coal combustion Crustal dust Cement dust Steel manufacturing Vehicle exhaust Secondary sulfate Secondary nitrate POC SOC

27.18 29.29 15.50 9.71 17.08 18.55 4.58 15.21 10.68

6.96 1.16 0.12 0.64 6.33 0 0

48.67 38.25 17.17 11.16 31.88 26.75 9.54 26.66 14.32

12.46 1.51 0.14 0.74 11.81 0 0

Summer

Calculated PM10 Measured PM10

c2

R2 PM (%)

132.57 146.36 0.21 0.04 90.58

Winter

197.74 214.62 0.99 1.00 92.13

G.-L. Shi et al. / Atmospheric Environment 45 (2011) 5692e5698

5697

Table 4 Concentrations (mg/m3) of TOC, POC and SOC in different seasons in some studies. Location

Season

TOC

POC

SOC

SOC/TOC

References

Taiyuan Chinaa

summer winter

25.89  6.64 40.98  15.56

15.21  1.65 26.66  2.61

10.68  6.84 14.32  15.77

0.41 0.35

This study

Beijing Chinaa

Autumn winter

16.4 25.6

7.2 14.9

9.2 10.7

0.57 0.40

Duan et al., 2005

Beijing Chinab

summer winter

17.10  4.1 41.20  20.8

6.70 24.00

10.40 17.20

0.61 0.42

Dan et al., 2004

Birmingham UKa

summer winter

4.77 3.71

1.67 3.08

3.10 0.63

0.65 0.17

Castro et al., 1999

Coimbra Portugala

summer winter

5.16 9.76

2.58 6.15

2.58 3.61

0.50 0.37

a b

PM10 samples. PM2.5 samples.

summer (Duan et al., 2005). Table 4 also shows the concentrations of TOC and SOC from other studies (Castro et al., 1999; Duan et al., 2005; Dan et al., 2004), which are consistent with our result that the SOC/ TOC ratios were higher in summer than in winter. 6. Conclusion A new receptor CMB-iterative method was developed in this study. This model can directly estimate the SOC concentrations in the ambient receptor, without introducing the SOC profile into the model. Initially, the model was used to study reported ambient dataset. The obtained result approached to the reported values. In addition, an ambient dataset measured from Taiyuan city in China was introduced into the CMB-Iteration method. Acceptable results were obtained. The above analysis suggests that the newly developed CMB-Iteration method is a good tool to estimate the concentrations of POC and SOC. Acknowledgments This study is supported by Special Funds for Research on Public Welfares of the Ministry of Environmental Protection of China (201109002), Science & Technology Fund Planning Project of Tianjin (09ZCGYSF02400), and National Natural Science Foundation of China (20877042). The National Center for Atmospheric Research is sponsored by the National Science Foundation. Appendix. Supplementary data Supplementary data associated with this article can be found in the online version, at doi:10.1016/j.atmosenv.2011.07.031. References Baldwin, D.P., Zamzow, D.S., D’Silva, A.P., 1994. Aerosol mass measurement and solution standard additions for quantization in laser ablation-inductively coupled plasma atomic emission spectrometry. Analytical Chemistry 66, 1911e1917. Bi, X.H., Feng, Y.C., Wu, J.H., Wang, Y.Q., Zhu, T., 2007. Source apportionment of PM10 in six cities of northern China. Atmospheric Environment 41, 903e912. Cao, J.J., Lee, S.C., Ho, K.F., Zhou, S.C., Fung, K., Li, Y., Watson, J.G., Chow, J.C., 2004. Spatial and seasonal variations of atmospheric organic carbon and elemental carbon in Pearl River delta region, China. Atmospheric Environment 38, 4447e4456. Carvalho, L.R.F., Souza, S.R., Martinis, B.S., Korn, M., 1995. Monitoring of the ultrasonic irradiation effect on the extraction of airborne particle matter by ion chromatography. Analytica Chimima Acta 317, 171e179. Castro, L.M., Pio, C.A., Harrison, R.M., Smith, D.J.T., 1999. Carbonaceous aerosol in urban and rural European atmospheres: estimation of secondary organic carbon concentrations. Atmospheric Environment 33, 2771e2781.

Chen, L.W.A., Watson, J.G., Chow, J.C., Magliano, K.L., 2007. Quantifying PM2.5 source contributions for the San Joaquin Valley with multivariate receptor models. Environmental Science & Technology 41, 2818e2826. Chen, L.W.A., Watson, J.G., Chow, J.C., Dubois, D.W., Herschberger, L., 2010. Chemical mass balance source apportionment for combined PM2.5 measurements from U.S. non-urban and urban long-term networks. Atmospheric Environment 44, 4908e4918. Chow, J.C., Watson, J.G., 1999. Ion chromatography in elemental analysis of airborne particles. In: Landsberger, S., Creatchman, M. (Eds.), Elemental Analysis of Airborne Particles. Gordon and Breach, NewarkNJ, pp. 539e573. Chow, J.C., Watson, J.G., Lowenthal, D.H., Chen, L.W.A., Zielinska, B., Mazzoleni, L.R., Magliano, K.L., 2007. Evaluation of organic markers for chemical mass balance source apportionment at the Fresno Supersite. Atmospheric Chemistry and Physics 7, 1741e1754. Christensen, W.F., Gunst, R.F., 2004. Measurement error models in chemical mass balance analysis of air quality data. Atmospheric Environment 38, 733e744. Dan, M., Zhuang, G.S., Li, X.X., Tao, H.R., Zhuang, Y.H., 2004. The characteristics of carbonaceous species and their sources in PM2.5 in Beijing. Atmospheric Environment 38, 3443e3452. Ding, X., Zheng, M., Edgerton, E.S., Jansen, J.J., Wang, X.M., 2008. Contemporary or fossil origin: split of estimated secondary organic carbon in the southeastern United States. Environmental Science & Technology 42, 9122e9128. Duan, F., He, K., Ma, Y., Jia, Y., Yang, F., Lei, Y., Tanaka, S., Okuta, T., 2005. Characteristics of carbonaceous aerosols in Beijing, China. Chemosphere 60, 355e364. Gnauk, T., Müller, K., Pinxteren, D., He, L.Y., Niu, Y., Hu, M., Herrmann, H., 2008. Sizesegregated particulate chemical composition in Xinken, pearl river delta, China: OC/EC and organic compounds. Atmospheric Environment 42, 6296e6309. Ho, K.F., Lee, S.C., Chan, C.K., Yu, J.C., Chow, J.C., Yao, X.H., 2003. Characterization of chemical species in PM2.5 and PM10 aerosols in Hong Kong. Atmospheric Environment 37, 31e39. Javitz, H.S., Watson, J.G., Robinson, N., 1988. Performance of the chemical mass balance model with simulated local-scale aerosols. Atmospheric Environment 22, 2309e2322. Kleindienst, T.E., Edney, E.O., Lewandowski, M., Offenberg, J.H., Jaoui, M., 2006. Secondary organic carbon and aerosol yields from the irradiations of isoprene and alpha-pinene in the presence of NOx and SO2. Environmental Science & Technology 401, 3807e3812. Lane, T.E., Pandis, S.N., 2007. Predicted secondary organic aerosol concentrations from the oxidation of isoprene in the Eastern United States. Environmental Science & Technology 41, 3984e3990. Lee, D., Balachandran, S., Pachon, J., Shankaran, R., Lee, S., Mulholland, J.A., Russell, A.G., 2009. Ensemble-trained PM2.5 source apportionment approach for health studies. Environmental Science & Technology 43, 7023e7031. Lee, S., Russell, A.G., Baumann, K., 2007. Source apportionment of fine particulate matter in the southeastern United States. Journal of the Air & Waste Management Association 57, 1123e1135. Lee, S., Russell, A.G., 2007. Estimating uncertainties and uncertainty contributors of CMB PM2.5 source apportionment results. Atmospheric Environment 41, 9616e9624. Lewandowski, M., Jaoui, M., Offenberg, J.H., Kleidnienst, T.E., Edndy, E.O., Sheesley, R.J., Schauer, J.J., 2008. Primary and secondary contributions to ambient PM in the Midwestern United States. Environmental Science & Technology 42, 3303e3309. Lowenthal, D.H., Chow, J.C., Watson, J.G., Neuroth, G.R., Robbins, R.B., 1992. Shafritz, B.P.; Countess, R.J. the effects of collinearity on the ability to determine aerosol contributions from diesel- and gasoline- powered vehicles using the chemical mass balance model. Atmospheric Environment 26A, 2341e2351. Marmur, A., Unal, A., Mulholland, J.A., Russell, A.G., 2005. Optimization-Based source apportionment of PM2.5 incorporating gas-to-particle ratios. Environmental Science & Technology 39, 3245e3254. Murphy, B.N., Pandis, S.N., 2009. Simulating the formation of semivolatile primary and secondary organic aerosol in a regional chemical transport model. Environmental Science & Technology 40, 4722e4728.

5698

G.-L. Shi et al. / Atmospheric Environment 45 (2011) 5692e5698

Odum, J.R., Hoffman, T., Bowman, F., Collins, T., Flagan, R.C., Seinfeld, J.H., 1996. Gasparticle partitioning and secondaryorganic aerosol yields. Environmental Science & Technology 30, 2580e2585. Pachon, J.E., Balachandran, S., Hu, Y., Weber, R.J., Mulholland, J.A., Russell, A.G., 2010. Comparison of SOC estimates and uncertainties from aerosol chemical composition and gas phase data in Atlanta. Atmospheric Environment 44, 3907e3914. Pandis, S.N., Harley, R.H., Cass, G.R., Seinfeld, J.H., 1992. Secondary organic aerosol formation and transport. Atmospheric Environment 26A, 2269e2282. Pankow, J.F., 1987. Review and comparative analysis oftheories of partitioning between the gas and aerosolparticulate phases in the atmosphere. Atmospheric Environment 21, 2275e2283. Plaza, J., Gómez-Moreno, F.J., Núñez, L., Pujadas, M., Artíñano, B., 2006. Estimation of secondary organic aerosol formation from semi-continuous OCeEC measurements in a Madrid suburban area. Atmospheric Environment 40, 1134e1147. Russell, M., Allen, D.T., 2004. Seasonal and spatial trends in primary and secondary organic carbon concentrations in southeast Texas. Atmospheric Environment 38, 3225e3239. Schauer, J.J., Rogge, W.F., Hildemann, L.M., Mazurek, M.A., Cass, G.R., 1996. Source apportionment of airborne particulate matter using organic compounds as tracers. Atmospheric Environment 30, 3837e3855. Schauer, J.J., Kleeman, M.J., Cass, G.R., Simoneit, B.T., 1999. Measurement of emissions from air pollution sources. 2. C1 through C30 organic compounds from medium duty diesel trucks. Environmental Science & Technology 33, 1578e1587. Shi, G.L., Feng, Y.C., Zeng, F., Li, X., Zhang, Y.F., Wang, Y.Q., Zhu, T., 2009. Use of a Nonnegative Constrained Principal Component Regression chemical mass balance model to study the contributions of Nearly Collinear sources. Environmental Science & Technology 43, 8867e8873. Takahama, S., Davidson, C.I., Pandis, S.N., 2006. Semicontinuous measurements of organic carbon and acidity during the Pittsburgh air quality study: implications for acid-catalyzed organic aerosol formation. Environmental Science & Technology 40 (17), 2191e2199.

U.S. Environmental Protection Agency, 2004. EPA CMB8.2 User’s Manual. Office of Air Quality Planning and Standards, Research Triangle Park NC 27711. Watson, J.G., Cooper, J.A., Huntzicker, J.J., 1984. The effective variance weighting for least squares calculations applied to the mass balance receptor model. Atmospheric Environment 18, 1347e1355. Watson, J.G., Chow, J.C., Frazier, C.A., 1999. X-ray fluorescence analysis of ambient air samples. In: Landsberger, S., Creatchman, M. (Eds.), Elemental Analysis of Airborne Particles. Gordon and Breach, Newark, NJ, pp. 67e96. Watson, J.G., Chow, J.C., Fujita, E.M., 2001. Review of volatile organic compound source apportionment by chemical mass balance. Atmospheric Environment 35, 1567e1584. Watson, J.G., Zhu, T., Chow, J.C., Engelbrecht, J., Fujita, E.M., Wilson, W.E., 2002. Receptor modeling application framework for particle source apportionment. Chemosphere 49, 1093e1136. Watson, J.G., Chen, L.W.A., Chow, J.C., Doraiswamy, P., Lowenthal, D.H., 2008. Source apportionment: findings from the U.S. supersites program. Journal of the Air & Waste Management Association 58, 265e288. Yan, B., Zheng, M., Hu, Y., Ding, X., Sullivan, A.P., Weber, R.J., Baek, J., Edgerton, E.S., Russell, A.G., 2009. Roadside, urban, and rural comparison of primary and secondary organic molecular markers in ambient PM2.5. Environmental Science & Technology 43, 4287e4293. Yu, S.C., Bhave, P.V., Dennis, R.L., Mathur, R., 2007. Seasonal and regional variations of primary and secondary organic aerosols over the continental United States: semi-empirical estimates and model evaluation. Environmental Science & Technology 41, 4690e4697. Zheng, M., Cass, G.R., Schauer, J.J., Edgerton, E.S., 2002. Source apportionment of PM2.5 in the southeastern United States using solvent-extractable organic compounds as tracers. Environmental Science & Technology 36, 2361e2371. Zheng, M., Cass, G.R., Ke, L., Wang, F., Schauer, J.J., Edgerton, E.S., Russell, A.G., 2007. Source apportionment of daily fine particulate matter at Jefferson Street, Atlanta, GA, during summer and winter. Journal of the Air & Waste Management Association 57, 228e242.