Techniques for measuring particle size distribution of ...

Atmospheric Environment 66 (2013) 25e32

Contents lists available at SciVerse ScienceDirect

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

Techniques for measuring particle size distribution of particulate matter emitted from animal feeding operations Lingjuan Wang-Li a, *, Zihan Cao a, Michael Buser b, Derek Whitelock c, Calvin B. Parnell d, Yuanhui Zhang e a

Department of Biological and Agricultural Engineering, North Carolina State University, Raleigh, NC 27695, USA Department of Biosystems and Agricultural Engineering, Oklahoma State University, USA c Agricultural Research Service, U.S. Department of Agriculture, Las Cruces, NM, USA d Department of Biological and Agricultural Engineering, Texas A&M University, USA e Department of Agricultural and Biological Engineering, University of Illinois at Urbana-Champaign, USA b

h i g h l i g h t s < Particle size distributions of PM samples were analyzed by four analyzers. < Significant differences in PSD measurements were observed. < Laser diffraction analyzers provided greater and broader PSDs than ESZ analyzer. < Measured PM2.5 mass fractions differed from the lognormal fitting PM2.5 fractions.

a r t i c l e i n f o

a b s t r a c t

Article history: Received 3 April 2012 Received in revised form 21 August 2012 Accepted 25 August 2012

While various techniques for measuring particle size distributions (PSD) of particulate matter (PM) exist, there is no a single agreed upon standard or reference method for PM with different characteristics. This study investigated differences in the PSD measurements by four PSD analyzers: LS13 320 multi-wave length laser diffraction particle size analyzer, LS230 laser diffraction particle size analyzer, LA-300 laser scattering particle size analyzer, and Coulter Counter Multisizer3 (CCM3). Simultaneously collected total suspended particulate (TSP) samples in a commercial egg production house were analyzed by the four analyzers for PSDs. In addition, four types of testing powders (limestone, starch, No.3 micro aluminum, and No.5 micro aluminum) were also analyzed by these four PSD analyzers. The results suggest when comparing measured mass median diameters (MMDs) and geometric standard deviations (GSD) of the PSDs, the laser diffraction method (LS13 320, LS230 and LA-300) provided larger MMDs and broader distributions (GSDs) than the electrical sensing zone method (CCM3) for all samples. When comparing mass fractions of PM10 and PM2.5 between the measured values and the lognormal fitting values derived from the measured MMDs and GSDs, lognormal fitting method produced reasonably accurate PM10 mass fraction estimations (within 5%), but it failed to produce accurate PM2.5 mass fraction estimations. The measured PM2.5 mass fractions significantly differed from the lognormal fitting PM2.5 fractions and the mean differences reached as high as 95%. It is strongly recommended that when reporting a PSD of certain PM samples, in addition to MMD and GSD, the mass fractions of PM10 and PM2.5 should also be reported. Ó 2012 Elsevier Ltd. All rights reserved.

Keywords: Particulate matter Animal feeding operation Particle size distribution analyzer Laser diffraction Electrical sensing zone PM10 mass fraction PM2.5 mass fraction

1. Introduction As a criteria pollutant, particulate matter (PM) has been a research topic for numerous studies in the context of air pollution. The studies of health impacts, emission estimation of PM, and * Corresponding author. Tel.: þ1 919 515 6762; fax: þ1 919 515 7760. E-mail address: [email protected] (L. Wang-Li). 1352-2310/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.atmosenv.2012.08.051

development of new control technologies require knowledge of PM characteristics. Among these PM characteristics, the particle size distribution (PSD) is perhaps the most important physical parameter governing particle behavior. Various methods and techniques are available for conducting PSD analyses. Advantages and disadvantages associated with each method exist. Unfortunately, there is no single agreed upon method to determine the PSD of PM emitted from difference sources.

26

L. Wang-Li et al. / Atmospheric Environment 66 (2013) 25e32

In the literature, the commonly used techniques for PSD measurements can be classified into five categories based upon the principals applied in particle size measurements (Hinds, 1999): (1) aerodynamic method, including aerodynamic particle sizer (APS) and various cascade impactors; (2) optical method, including various optical particle counters, light scattering and laser diffraction particle size analyzers; (3) electrical sensing zone method, i.e. Coulter Counter; (4) electrical mobility and condensation method, e.g. differential mobility analyzer (DMA) plus condensation nuclei counter (CNC); (5) electron microscopy. The APS measures the aerodynamic equivalent diameter (AED) of individual particles using the time-of-flight principle (Hinds, 1999; Mitchell and Nagel, 1999; Peters and Leith, 2003). It has a particle size measurement range of 0.5e20 mm. The cascade impactor method uses the principle of inertia to separate PM into different particle size ranges based upon different cut-off sizes at different impaction stages connected in series (Hinds, 1999). The impactor method provides onsite PSD measurement of airborne particles, but the PSD classification of this method can only include a small number of size classes. The inertial removal may cause inter-stage loss due to sharp bends in the inter-stage flow path. The optical particle counters (OPCs) are used most frequently for measuring number concentrations corresponding to different size classes. While they can provide real-time concentrations and size measurements of airborne particles, they are not intended for use on larger particles or in hostile environments with high levels of PM and gaseous pollutants (Parker et al., 2009). In light scattering particle size analyzers, for particles larger than the light wavelength, Mie scattering theory is applied to determine the angular distribution of scattered lights by particles suspended in a solvent. This angular distribution of the scattered lights is a very sensitive indicator of PSD (Hinds, 1999). In this method, knowledge of the refractive index of the particle and the solvent is required to solve Mie equations for determination of the scattered light angular distribution. It was reported (Xu and Guida, 2003) that the light scattering method may provide more reliable PSD measurements on account of ease of use and broad size ranges, from sub-micrometers to millimeters. On the other hand, the light scattering methods are typically limited to measuring particles greater than 0.3 mm due to reduced detection efficiency with smaller particle size and errors caused by particle shape and refractive index variations. In particle size measurements of airborne particles, the light scattering analyzers measure PSDs from PM samples taken using gravimetric collection techniques. Thus, the light scattering analyzers don’t provide real-time PSD measurements. In this method, PM samples need to be extracted from filter media into a solvent for PSD analysis. Consequently, this method is only suitable for insoluble particles. All the optical analyzers provide PSD measurements in equivalent spherical diameter (ESD), not AED. Information of particle shape and density is needed to convert ESD to AED. The electrical sensing zone (ESZ) method is also known as the Coulter Counter method (Xu and Guida, 2003; Lines et al., 1996). In a Coulter Counter, the particles suspended in an electrolyte solution are forced to pass through a small aperture where an electric field is applied. The solution’s conductance changes as the particles pass through the aperture. The change in conductance is a function of particle size. When particles pass through the aperture, the particles’ individual volumes are directly measured. This method measures a single particle’s volume and provides high resolution and reproducibility for individual particle size assessment in ESD (Lines et al., 1996). However, particles that can be analyzed are restricted to those that can be dispersed in an electrolyte solution and still retain their original integrity. Like the light scattering particle size analyzers, the Coulter Counter also analyze PSDs of

airborne particles from samples taken using gravimetric collection techniques. Thus, this method does not provide real-time measurements of PSDs either. The electrical mobility depends on the electric mobility of the particles for PSD measurement and it can only work well for particles with good mobility. The electron microscopy method is capable of providing both particle size and morphology information. However, this method could not provide sufficient statistical representation of particle measurements to derive a PSD for a PM sample. It is not appropriate for continuous or long term PSD analysis. Due to lack of a standardized method for PSD measurements in different applications, efforts have been made to compare PSD results measured by some aforementioned methods. When both cascade impactor and laser diffraction particle size analyzer (one type of light scattering analyzers) were used to evaluate PSDs of particles generated by nebulizers (Ziegler and Wachtel, 2005; Smyth and Hickey, 2003), the results of PSDs measured by both of these two techniques correlated with each other very well. In comparison of the light scattering method and the ESZ method, Xu and Guida (2003) reported that the laser diffraction particle size analyzer provided much larger mean sizes and broader distributions for irregular particles when compared with the Coulter Counter. Similar finding was also reported by Jerez et al. (2011) in a study of PSDs in a swine building. When compared the ESZ with other two methods, McClure (2009) found that the Coulter Counter (ESZ method) compared well to APS, whereas Xu and Guida (2003) discovered that the ESZ produced compatible results with dynamic image analysis, which are much less affected by particle shape (Xu and Guida, 2003). In general, particle size results of non-spherical particles measured by different instruments are often less consistent when compared with each other (Xu and Guida, 2003). The objectives of this reported study were to (1) investigate differences of PSDs measured by different particle size analyzers using either laser diffraction or ESZ techniques for PM from animal feeding operations with large mass median diameters (MMDs) and geometric standard deviations (GSDs); (2) compare mass fractions of PM10 and PM2.5 between the measured values and the lognormal fitting values derived from the measured MMDs and GSDs. For future relevant studies, it is recommended that limitations of PSD measurements by a given method should be recognized and measured PM10 and PM2.5 mass fraction should be simultaneously reported when reporting a MMD and a GSD for a PSD measurement. 2. Material and methods 2.1. Particle size analyzers The particle size analyzers for the study include (1) a LS13 320 multi-wave length laser diffraction particle size analyzer (Beckman Coulter Inc., Miami, FL) owned by the Research Group at North Carolina State University (NCSU); (2) a LA300 laser scattering particle size analyzer (Horiba Instruments Inc., Irvine, CA) owned by the Research Group at University of Illinois at UrbanaChampaign (UIUC); (3) a Coulter Counter Multisizer3, CCM3 (Beckman Coulter Inc., Miami, FL) owned by the Center for Agricultural Air Quality Engineering and Science at Texas A&M University (TAMU); (4) a Coulter Counter Multisizer3, CCM3 (Beckman Coulter Inc., Miami, FL) and a LS230 laser diffraction particle size analyzer (Beckman Coulter Inc., Miami, FL) owned by the research group at USDA-ARS Cotton Production and Processing Research Unit in Lubbock, TX, hereby also known as USDA.


Since the selection of refractive index is critical for the light scattering analyzers, a preliminary investigation was conducted for unknown refractive index of the PM samples taken from the animal house. The final selection of the refractive index was 1.34 for LA300 at UIUC (swine PM-based), and 1.5 for LS13 320 at NCSU (organic matter based). The PSDs measured by all these the analyzers were volume distributions in ESD. Since the U.S. EPA regulates PM by mass in two size ranges (10 mm, or 2.5 mm) in the form of AED, measured PSDs by volume in ESD were then converted to PSDs in AED using the following equation (Hinds, 1999):

rffiffiffiffiffi AED ¼ ESD

rP c

(1)

Where rp is the particle density, measured by a pycnometer (AccuPyc1330, Micromeritics, Norcross, GA), c is the shape factor of a particle and was assumed to be 1 in this study (Cao, 2009). It was assumed that densities of particles in different sizes of the same kind PM were the same; hence the volume-based PSD was considered as the mass-based PSD. The volumetric median diameters were treated as MMD.

2.2. Field PM sample collection and sample allocation Samples of total suspended particulate (TSP) were collected using six low-volume (LV) TSP samplers (Wang-Li et al., 2013). This LV-TSP sampler was originally designed and manufactured by the research group at Texas A&M University following the EPA’s specifications of the engineering design parameters for a high volume TSP sampler in 40 CFR Part 50 Appendix B (1987). Detailed design and performance of the sampler were reported in Wangjura et al. (2005). The field PM sampling was conducted in a high-rise tunnelventilated egg production (layer) house. The detailed information about the farm and the layer houses is reported in Wang-Li et al. (submitted for publication). In brief, the sampling house has dimensions of 175 m long by 18 m wide with approximately 95,000 laying hens. In this high-rise house, laying hens were housed in the upper floor (referred to as 2nd floor). Manure fell into the pit (referred to as 1st floor). Ventilation air entered the 2nd floor of the house through 36.5 m long air inlets centered on two side walls of the house. Each endwall was equipped with seventeen, 1.2-m (48in) diameter, 480 VAC, 3-phase, belt-driven ventilation fans (Choretime). Among these seventeen fans, eight fans were located in the 2nd floor, and nine fans were in the 1st floor. Each house had 34 fans in total, and was ventilated in 11 stages. Three (3) LV-TSP

27

samplers were co-located on the first floor and the other three LV-TSP samplers were co-located on the second floor of the same house immediately upstream of the exhaust fans on the east endwall. The placement of the TSP samplers is illustrated in Fig. 1 and more detailed information about the PM sampling is reported in Wang-Li et al. (2013). The co-located sampling design generated three replicates in TSP samples on each floor for each sampling event. Thirteen (13) sampling events were carried out through 4 consecutive days in winter of 2008e2009, spring of 2009, respectively. Thus, each season produced 78 samples in total. After sampling, the three sample replicates on each floor were randomly selected and distributed to three research groups having different instruments at three locations. For each season, 26 samples were allocated for each of the three locations. The PM samples assigned to NCSU, UIUC and TAMU or USDA were analyzed for PSDs in terms of MMDs and GSDs by LS13 320, LA300, CCM3 and LS230 particle size analyzers. 2.3. Filter-based PM sample extraction To measure PSDs of the PM sample collected on the TSP filters, 47 mm (2.5 mm) ZefluorÔ supported PTFE filters (Pall Corporation, Ann Arbor, Michigan), it is required to extract PM from the filters and disperse them into a solvent that could be introduced into the analyzers for PSD analysis. When selecting a suitable solvent as a dispersant for the PM samples, properties of the samples (i.e., solubility, reactivity, and suspendibility) governed the choice of the medium. Through a chemical speciation analysis study, it was discovered that the PM in the egg production houses was organic in nature mainly generated from chicken manure and feed (Li, 2012), and formation of secondary inorganic PM was insignificant due to very short residence time of pollutant gases in the animal houses and lack of acid gases to react with NH3 (Li, 2012). To minimize soluble loss, the organic solvent, ethanol was used as the PM sample extraction medium. For laser diffraction analyzers, the PM samples were extracted from filters into ethanol using ultrasonic bath method; whereas for Coulter Counter Multisizer, the PM samples were extracted from filters into electrolyte solution, a 5% lithium chloride/methanol solution. A preliminary study on determining optimum ultrasonic time was conducted to obtain the best sample extraction and dispersion without particle breakdown (Cao, 2009). It was discovered that fifteen minutes was the time when the most coagulated particles were separated and yet no break-down of the particles was observed. As a result, all the PM samples were extracted through ultrasonic bath for 15 min.

Fig. 1. Placement of the LV-TSP samplers inside the high-rise layer house. [A]: Three LV-TSP samplers collocated on the 1st floor; [B]: Three LV-TSP samplers on the 2nd floor.

28


2.4. Testing powders for laboratory comparisons

RD ¼

In order to further investigate the difference in PSDs measured by different instruments, four types of testing aerosols were used to reduce the measurement uncertainty possibly caused by field sampling and/or sample extraction at three different locations (NCSU, UIUC and USDA). These four types of testing aerosols include limestone, starch, No.3 micro aluminum, and No.5 micro aluminum. These testing aerosols were prepared using the same operation procedure before they were analyzed by the different analyzers. The testing aerosol samples were analyzed for PSDs in terms of MMDs and GSDs by LS13 320 at NCSU, LA300 at UIUC, and CCM3 and LS230 at USDA, respectively.

2.5. Data analysis 2.5.1. Cumulative lognormal distribution forms of PSDs The lognormal distribution is the most commonly used PSD model (Hinds, 1999). Typically, a lognormal distribution describing a PSD in mass is characterized by a MMD and a GSD. By definition, MMD is the particle size where 50% of PM mass is larger or smaller than this diameter. The GSD is defined as the ratio of particle sizes corresponding to cumulative mass fraction of 84.1% and 50%, or 50% and 15.9%, or the square root of the ratio of 84.1% and 15.9% (Hinds, 1999). The mathematical expression of a fractional PSD is as following:

" 2 # lndp lnðMMDÞ ddp df ¼ pffiffiffiffiffiffiffi exp 2p*dp *lnðGSDÞ 2ðlnðGSDÞÞ2 1

(2)

The PSD can also be described as a cumulative distribution Fa, which gives the mass fraction of all the particles with diameters less than a, thus F10 ¼ PM10 and F2.5 ¼ PM2.5. The cumulative distribution function is presented as:

Za

1

pffiffiffiffiffiffiffi exp 2p*dp *lnðGSDÞ ¼ F dp ; MMD; GSD

" 2 # lndp lnðMMDÞ

Fa ¼

0

2ðlnðGSDÞÞ2

ddp (3)

2.5.2. Comparisons of PM10 and PM2.5 mass fractions: lognormal fitting values vs. measured values As indicators for PM in the National Ambient Air Quality Standards (NAAQS), mass fractions of the PM10 and PM2.5 are usually of interests. When a PSD was measured by an analyzer, the mass fractions of PM10 and PM2.5 were also measured by the instrument. In the literature, the PSD has often been reported by MMD and GSD assuming that a PSD followed a lognormal distribution. Based upon the measured MMD and GSD, the mass fractions of PM10 and PM2.5 may be calculated using cumulative lognormal distribution function (Equation (2)). These calculated PM10 and PM2.5 mass fraction are hereby defined as lognormalfitting values. The lognormal-fitting values of PM10 and PM2.5 are most probably what could be cited by other researchers if only measured MMD and GSD were reported. It is expected that the measured & the fitting values of PM10 or PM2.5 agree only if a PSD follows the lognormal distribution well. To investigate possible errors associated with the lognormal fitting method for PM10 and PM2.5 mass fraction determination, the relative differences (RDs) between the measured and the fitting values were defined for comparison:

Lognormal Measured 100% Measured

(4)

Where, Measured is the mass fraction of PM10 or PM2.5 measured by the analyzer; Lognormal is the mass fraction of PM10 or PM2.5 calculated using the lognormal distribution formula (Equation (3)) with measured MMD and GSD. 2.5.3. Statistic tests For comparisons of measured PSDs by the different analyzers for field samples taken in winter and spring, or for testing powders samples, single factor ANOVA tests were performance to compare mean MMDs and PSDs. TukeyeKramer test, a multiple comparison procedure was also applied to further check whether measured MMDs and GSDs from different instruments were significantly different from each other. In addition, paired-t tests were conducted on means to identify significant difference between two means.

3. Results and discussion 3.1. PSDs of the field PM samples Since it is not feasible to report all the original PSD measurement curves for all the 156 field samples (78 samples per season for two seasons), one set of measured fractional PSD curves by the different analyzers is illustrated in Fig. 2 as an example. Figs. 3 and 4 show the comparisons of the cumulative PSD curves by the different analyzers for the field TSP samples. These PSD curves were constructed using Equation (2) and the measured mean MMDs and mean GSDs listed in Tables 1 and 2. The statistic tests (ANOVA and TukeyeKramer tests) suggest that there were significant difference in MMDs (p < 0.0001) and GSDs (p < 0.0001) obtained by these three instruments. In general, LA-300 at UIUC provided the largest MMDs; whereas CCM3 at TAMU gave the smallest MMDs. LS13 320 provided the largest GSDs; whereas CCM3 gave the smallest GSDs. As it may be observed in Fig. 2, the size measurement ranges of the analyzers were different. While the laser diffraction analyzers provided large size ranges (0.04e 2000 mm for LS13 320; 0.1e600 mm for LA300), ESZ analyzer, i.e. the CCM3 with 100 mm aperture provided a measurement range in 2.95e60 mm. The differences in size measurement ranges is suspected to be the major source causing measurement difference between light scattering method and ESZ methods for this study. In comparison to the literature, the observation of smaller MMD and GSD measurements by the ESZ method (CCM3) as compared to the laser diffraction method is in agreement with the findings by Jerez et al. (2011) in a study of PSDs of the PM samples taken from a swine building. There was no significant difference in MMDs (p ¼ 0.400) and GSDs (p ¼ 0.297) between LS13 320 at NCSU and LS230 at USDA. Significant differences in MMDs (p < 0.0001) and GSDs (p < 0.0001) from between LS13 320 and LA-300 were observed. Using different refractive indices is suspected to be one cause of different PSD measurement results. In addition, LA300 is a laser scattering instrument with much simpler optics and fitting algorithm, compared with LS13 320, although both of these two instruments are laser diffraction particle size analyzers. If there are small peaks in the PSD, LA300 may not be able to detect. As shown in Fig. 5, there were some small peaks missed by the LA300 when compared with the LS13302’s measurement. The MMDs obtained by LA-300 were larger than that measured by LS13 320, whereas GSDs obtained by LA-300 were smaller than that measured by LS13 320 for all PM samples. These differences may be considered as


29

systematic error due to different characteristics of these two laser diffraction particle size analyzers. 3.2. PSDs of the testing powder samples Figs. 6e9 show the comparisons of the cumulative PSD curves by the different analyzers for the testing powder samples, i.e. limestone, starch, No.3 micro aluminum and No.5 micro aluminum. The measured mean MMDs and GSDs by the different analyzers are listed in Table 3. There was significant difference in MMDs obtained from LA-300 at UIUC and the other three instruments (LS13 320, LS230 and CCM3). Two laser diffraction type analyzers (LS13 320 at

Fig. 2. (continued).

Fig. 3. Comparison of the cumulative PSD curves by the different analyzers for the winter PM samples (Note: the PSD curves were constructed using Equation (2) and the measured mean MMDs and mean GSDs in Table 1).

Fig. 2. Illustration of the measured PSD curves by the different analyzers for the field PM samples.

Fig. 4. Comparison of the cumulative PSD curves by the different analyzers for spring PM samples (Note: the PSD curves were constructed using Equation (2) and the measured mean MMDs and mean GSDs in Table 2).

30


Table 1 Summary of the MMDs (mm) and GSDs measured by the three analyzers for winter samples. PSDa

MMD (mean SD) GSD (mean SD) a

NCSU

UIUC

TAMU

LS13 320

LA-300

CCM3

17.13 0.81 2.63 0.04

22.71 1.43 2.02 0.11

13.94 1.00 1.85 0.04

Mean MMDs and GSDs are the average values of replicates of 26 PM samples.

Table 2 Summary of the MMDs (mm) and GSDs measured by the three analyzers for spring samples. PSDa

MMD (mean SD) GSD (mean SD) b

NCUS

UIUC

USDA

LS13 320

LA-300

LS230

CCM3

18.44 1.44 2.67 0.11

22.62 2.68 1.99 0.15

18.47 1.38 2.65 0.22

13.99 0.74 1.84 0.04

Mean MMDs and GSDs are average values of replicates of 26 PM samples.

Fig. 6. Comparison of the cumulative PSD curves by the different analyzers for limestone samples (Note: the PSD curves were constructed using Equation (2) and the measured mean MMDs and mean GSDs in Table 3).

NCSU and LS230 at USDA) provided very similar results for MMDs. In general, LA-300 provided larger MMDs than the other three instruments. No significant difference in MMDs among the other three instruments (LS13 320, LS230 and CCM) was observed. Differences in GSDs obtained from different instruments were observed. In general, laser diffraction series particle size analyzers (LS13 320 at NCSU and LS230 at USDA) provided quite similar GSDs. Small but consistent difference in GSDs was observed between CCM3 at USDA and LA-300 at UIUC. However, significant

Fig. 7. Comparisons of the cumulative PSD curves by the different analyzers for starch samples (Note: the PSD curves were constructed using Equation (2) and the measured mean MMDs and mean GSDs in Table 3).

Fig. 5. Measured PSD curves for the limestone sample by the two light-scattering analyzers.

Fig. 8. Comparisons of the cumulative PSD curves by the different analyzers for No.3 aluminum samples (Note: the PSD curves were constructed using Equation (2) and the measured mean MMDs and mean GSDs in Table 3).


Fig. 9. Comparisons of the cumulative PSD curves by the different analyzers for No.5 aluminum samples (Note: the PSD curves were constructed using Equation (2) and the measured mean MMDs and mean GSDs in Table 3).

Table 3 Summary of PSDsa of the testing aerosols measured by the four analyzers. Testing aerosols

LS13 320 MMD (mm)

GSD

MMD (mm)

GSD

MMD (mm)

GSD

MMD (mm)

GSD

Limestone Starch No.3 micro aluminum No.5 micro aluminum Average

7.50 13.31 5.28

3.07 1.59 1.98

12.29 16.78 7.62

1.83 1.50 1.56

8.11 14.38 5.37

3.15 1.55 1.93

8.56 14.32 5.03

1.72 1.33 1.42

7.09

1.69

8.38

1.49

7.21

1.71

6.31

1.39

8.30

2.08

11.27

1.60

8.77

2.09

8.56

1.47

a

LA-300

LS230

CCM3

MMDs and GSDs are average values of three replicates.

differences in GSDs were observed between LS series and CCM3. In general, LS series provided larger GSDs than CCM3 and LA-300. A T-test was performed on the data to further determine if there were significant differences in MMDs and GSDs measured by four different instruments. The differences in MMD measurements by LS13 320 at NCSU and LA-300 at UIUC were significant (p ¼ 0.014), but the GSDs were not (p ¼ 0.079) at the 0.05 level of significance (a). No significant differences in MMDs and GSDs of PSDs were detected between LS13 320 at NCSU, LS230 at USDA and CCM3 at USDA. The comparative PSD results of testing powder samples measured by the different instruments were generally consistent with that of the filter-based samples (field samples). Differences in PSDs for both filter-based samples and testing powders measured by the four different instruments at four different locations (LS13 320 at

31

NCSU, LA-300 at UIUC, CCM3 at TAMU, CCM3 and LS230 at USDA) are statistically significant. In general, LA-300 provided the largest MMDs and LS230 gave the largest GSDs; the smallest MMDs and GSDs were provided by LS13 320 and CCM3, respectively. LS230 and LS13 320 provided very similar results for both MMDs and GSDs. Both of LS13 320 and CCM3 were produced by Beckman Coulter Inc. Xu and Guida (2003) reported that laser diffraction provided much larger MMDs and broader distributions (GSDs) than the ESZ. In this study, LS13 320 provided larger MMDs and GSDs than CCM3, which is consistent with Xu and Guida’s report. These differences may be systematic errors due to different principles that these two instruments rely on, and different size measurement range. Specifically, the laser diffraction particle size analyzer (e.g. LS13 320, LS230, LA300) applies aerosol optical property for size measurements. As it is well known, element carbon is a light absorber, whereas organics, sulfates, nitrates and soil are light scatters. Therefore, PM chemical compositions have substantially impact on LS13 320/LS230/LA300, but not on CCM3. On the other hand, the Coulter Counter particle sizer (e.g. CCM3) operates on the electrical resistance principle and uses an apparatus to count and size particles suspended in electrolytes. Irregular shapes and conductivity of the particles passing through the aperture would have substantial impact on measurements. Knowing the chemical and physical properties of PM being studied is essential for selection of an appropriate PSD analyzer. 3.3. Mass fractions of PM10 and PM2.5: measured values vs. lognormal-fitting values The mean mass fractions of PM10 and PM2.5 obtained from measurements of PSDs by the different analyzers and from lognormal-fitting values using the measured MMD and GSDs of the corresponding PSD measurements are summarized in Table 4. When examining the PM10 mass fractions, no significant difference was observed between lognormal-fitting values and the measured values. The mean (SD) of the RDs for PM10 were in range of 0.39% (0.80%) to 3.34% (5.34%). The lognormal method produced reasonably accurate estimation of PM10 mass fraction. On the other hand, as shown in Table 4, lognormal-fitting method generated significant smaller PM2.5 mass fractions than the measured fractions by the different analyzers (laser diffraction or ESZ method) for all the samples. This might be due to the fact that the mass fraction of PM2.5 is considerably away from medium mass (2.5 mm