Ann. Occup. Hyg., Vol. 53, No. 6, pp. 585–594, 2009� The Author 2009. Published by Oxford University Press

on behalf of the British Occupational Hygiene Societydoi:10.1093/annhyg/mep021

Field and Wind Tunnel Comparison of Four AerosolSamplers Using Agricultural DustsSTEPHEN J. REYNOLDS1*, JASON NAKATSU1, MARVIN TILLERY1,THOMAS KEEFE1, JOHN MEHAFFY1, PETER S. THORNE2,KELLEY DONHAM2, MATTHEW NONNENMANN3, VIJAY GOLLA2 andPATRICK O’SHAUGHNESSY2

1Department of Environmental and Radiological Health Sciences, Colorado State University,Ft Collins, CO 80523, USA; 2Department of Environmental and Occupational Health, University ofIowa, Iowa City, IA 55337, USA; 3Department of Occupational Health Sciences, University of TexasHealth Science Center at Tyler, Tyler, TX 75708, USA

Received 16 July 2008; in final form 30 March 2009; published online 14 May 2009

Occupational lung disease is a significant problem among agricultural workers exposed to or-ganic dusts. Measurements of exposure in agricultural environments in the USA have tradition-ally been conducted using 37-mm closed-face cassettes (CFCs) and respirable Cyclones.Inhalable aerosol samplers offer significant improvement for dose estimation studies to reducerespiratory disease. The goals of this study were to determine correction factors between theinhalable samplers (IOM and Button) and the CFC and Cyclone for dusts sampled in livestockbuildings and to determine whether these factors vary among livestock types. Determination ofthese correction factors will allow comparison between inhalable measurements and historicalmeasurements. Ten sets of samples were collected in swine, chicken, turkey, and dairy facilitiesin both Colorado and Iowa. Pairs of each sampling device were attached to the front and backof a rotating mannequin. Laboratory studies using a still-air chamber and a wind tunnel pro-vided information regarding the effect of wind speed on sampler performance. Overall, theIOM had the lowest coefficient of variation (best precision) and was least affected by changesin wind speed. The performance of the Button was negatively impacted in poultry environ-ments where larger (feather) particulates clogged the holes in the initial screen. The CFC/IOM ratios are important for comparisons between newer and older studies. Wind speedand dust type were both important factors affecting ratios. Based on the field studies (Table6), a ratio of 0.56 is suggested as a conversion factor for the CFC/IOM (average for all environ-ments because of no statistical difference). Suggested conversion factors for the Button/IOMare swine (0.57), chicken (0.80), turkey (0.53), and dairy (0.67). Any attempt to apply a conver-sion factor between the Cyclone and inhalable samplers is not recommended.

Keywords: aerosol samplers; agriculture; Button sampler; correlation analysis; inhalable dust; IOM sampler; organicdust

INTRODUCTION

Occupational lung disease resulting from exposure tocomplex agricultural dusts may affect .1 million USagricultural workers (Donham et al., 1989, 1995;Donham, 1990; Heederik et al., 1991; Reynoldset al., 1993, 1996; Rylander and Jacobs, 1994;Merchant et al., 1995; Kullman et al., 1998; Merchantet al., 2002; Kirychuk et al., 2006). Organic dusts

found in livestock facilities are complex mixturescomposed of feed components, urine, fecal material,dander, feathers, insect parts, bacteria, fungi, pollengrains, (1 / 3)-b-D-glucans, endotoxins, and otherbioactive materials (Donham, 1986; Sonesson et al.,1990; Rylander and Jacobs, 1994; Popendorf andReynolds, 1997). Sampling to determine airbornedust exposure levels in livestock facilities in theUSA has largely relied on gravimetric methods thatutilize the 37-mm closed-face cassette (CFC) whilefollowing National Institute for Occupational Safetyand Health (NIOSH) method 0500 for ‘total dust’

*Author to whom correspondence should be addressed.Tel: þ970-491-3141; fax: þ970-491-2940;e-mail: stephen.reynolds@colostate.edu

585

determination (NIOSH, 1994b). Cyclones have alsobeen used in these facilities to measure the respirablefraction of dust as described by method 0600 (NIOSH,1994a). However, some evidence suggests that respira-tory ailments developed while working in livestockfacilities may result in airway effects rather than pul-monary effects and are more closely associated withexposure to total dust than exposure to respirable dust(Donham et al., 1989, 1995; Donham, 1990).

The only threshold limit value (TLV) for agricul-tural dusts is the 4 mg m�3 total dust value for(oat, wheat, and barley) grain dust as promulgatedby the American Conference of Governmental Indus-trial Hygienists (ACGIH). Regulatory organizations,such as the US Occupational Safety and Health Ad-ministration and the UK Health and Safety Executive(HSE), have occupational exposure limits of 10 mg m�3

for grain dust, primarily including barley, oats, andwheat (HSE, 2007; Occupational Safety and HealthAdministration, 2009). The HSE and other Europeanagencies have also established a similar standard forflour (wheat) dust. Therefore, other agricultural dusts(e.g. from livestock facilities) fall under guidelinesestablished for ‘particulates not otherwise specified’,which have an inhalable dust TLV of 10 mg m�3

(ACGIH, 2008). Furthermore, this TLV is high rela-tive to a proposed guideline of 2.4 mg m�3 for totaldust that has been recommended to prevent adversework-related health effects in swine and poultry pro-duction environments (Donham et al., 1995, 2000,2002; Reynolds et al., 1996). However, the estab-lishment of a total dust guideline specific to agricul-tural facilities is complicated by the current use of‘size-specific’ sampling devices designed to collectin a way compatible with the definition for inhalableparticles adopted by the ACGIH (2008). In fact, theACGIH TLV committee has expressed the intent toreplace the existing total particulate TLVs with in-halable, thoracic, and respirable particulate TLVs(ACGIH, 2008).

When characterizing the sampling behavior of theCFC in a wind tunnel, Mark et al. (1994) demon-strated that the collection efficiency of the CFC fallsbelow the line defining the inhalable fraction, espe-cially for particles .20 lm. Therefore, the expecta-tion is that a guideline value for agricultural dustsbased on the use of an inhalable sampler should behigher than that proposed for the CFC. For example,Predicala and Maghirang (2003) compared the IOMand the CFC while assessing airborne dust in swineconfinement barns and determined that the IOM pro-vided a greater overall geometric mean concentra-tion. Mahar et al. (1999) found that IOM and CFCmeasurements were strongly correlated (r 5 0.92,P , 0.001) when used to measure organic dust aero-sols in refuse-derived fuel plants with inhalable sam-plers again consistently providing higher measures ofdust concentration than total dust samplers. Kullman

et al. (1998) also found good correlation betweeninhalable and total dust samplers (r 5 0.76, P ,

0.001) used during a study of dairy workers with geo-metric mean concentrations of inhalable dust consis-tently higher than those determined using total dustsampling methods.

The degree to which the measurement made withthe CFC will differ from that of an IOM is primarilybased on the size distribution of the aerosol beingsampled (O’Shaughnessy et al., 2007). Because theCFC collects fewer of the larger particles than theIOM, there will be a greater discrepancy betweenthe two measurements as the median of the size dis-tribution increases. Comparing open-faced 37-mmfilter cassettes with IOM samplers, Liden et al.(2000) found a wide range of ratios between thetwo devices when compared for seven types of or-ganic dust (textile, wood, paper, cereals, flour, food,and animal breeding) with the differences largely de-pendent on differences in size distribution. Tatumet al. (2001) studying wood dust also found that therelative performance of the inhalable samplers ascompared to the total dust sampler may vary as afunction of particle size distribution in the work-place. A number of other studies have also been con-ducted, both under controlled laboratory conditionsand in field settings, to determine the relationship be-tween IOM samplers and other aerosol-sampling de-vices (Vaughn et al., 1990; Vincent, 1995; Kennyet al., 1997; Li et al., 2000; Demange et al., 2002;Teikari et al., 2003; Witschger et al., 2004; De Vochtet al., 2006). These studies typically demonstratedwide variation in sampler-to-sampler ratios (or corre-sponding correlations) suggesting that other factorsbesides size distribution may also cause the collec-tion efficiency curve for either device to differ,thereby resulting in a change in the ratio betweenmeasurements made by the two devices in the sameenvironment. These factors include sampler orienta-tion relative to a moving air source and air speed aswell as physical and environmental factors, such asparticle shape and electrostatic effects.

Given the wide variety of factors that may causechanges in the collection characteristics of an aerosol-sampling device, there is a need for site-specificsampler comparisons. Because of the concern forthe inducement of adverse health effects by agricul-turally derived dusts in particular, there is a needfor a rigorous comparison of these devices when usedin livestock buildings. It is apparent that workers inthese facilities are at elevated risk of developing air-way inflammation and accelerated declines in lungfunction. Thus, guidance on safe levels of airbornedust is needed to protect these workers. However,in the USA, a guidance value is currently applicableto only measurements taken with the CFC. There-fore, this study was conducted to determine a correc-tion factor between the CFC and the inhalable

586 S. J. Reynolds et al.

samplers for dusts sampled in livestock buildings andto determine if this factor varies between livestocktype. It is worth noting that Bartley (1998), in ananalysis of inhalable samplers, concluded that theIOM was the best reference sampler to be used incomparisons like these. In addition to field studies,laboratory studies were conducted to assess the effectof wind speed on sampler collection. Since a numberof previous studies conducted in animal rearingfacilities included the use of a respirable Cyclone(Donham et al., 1995; Reynolds et al., 1996), we uti-lized this opportunity to determine whether or notthere is a correlation between results from an inhal-able sampler and those from a respirable sampler.To our knowledge, this is the first field study compar-ison including the Button sampler.

METHODS

Two commonly used inhalable aerosol samplerswere compared to the CFC: the IOM (plastic) andthe Button. Measurements were also made with theSKC aluminum respirable Cyclone (all availablefrom SKC Inc., Eighty Four, PA, USA). Flow ratesfor each sampler were adjusted with a needle valveand calibrated to within 5% (plus or minus) of thesuggested flow rate before each trial using an elec-tronic soap bubble flow meter (Gilibrator�, Sensi-dyne, Clearwater, FL, USA). Flow rates of 2.0, 4.0,and 2.5 l min�1 were set according to the manu-facturers’ instructions for the IOM, Button, andCyclone, respectively. A flow rate of 2.0 l min�1

for the CFC was chosen from the suggested rangeof 1–2 l min�1 (NIOSH, 1994b). Polyvinyl chloridefilters (5-lm pore size) were weighed with the useof a six-place balance (Model MT5, Mettler-ToledoInc., Columbus, OH, USA).

Laboratory trials

Laboratory studies were conducted using three ag-ricultural dusts—swine, chicken, and turkey—in twodifferent chambers, one to mimic ‘still-air’ condi-tions and the other a wind tunnel in which two airspeeds were applied. The controlled laboratory stud-ies did not include dust from dairies due to the diffi-culty in collecting sufficient dust in the field forgeneration of this aerosol. Field studies were con-ducted in all four agricultural settings.Still-air chamber. Samplers were attached to a sta-

tionary mannequin (torso and head) in a 1-m3

Rochester-style chamber (Hinners et al., 1968) oper-ated with a downward speed of air of only 0.07 cm s�1,equivalent to the settling speed of a 5-lm particle andless than a criterion of ,0.1 m s�1 used by others for‘very slow-moving air’ (Witschger et al., 2004).Aerosols were produced from agricultural dusts col-lected from settled dust in the field. The settled dustwas assumed to contain a distribution of particle

sizes larger than the airborne dust due to preferentialsettling of the larger particles and exhausting of thesmaller particles via the ventilation system. There-fore, the collected dust was sieved to remove par-ticles (such as feather debris) .420 l using a40-mesh sieve to minimize the difference in size dis-tributions between the chamber and the field settings.The aerosols were generated using a Wright dustfeed (BGI Inc., Waltham, MA, USA) This aerosol-generating device was designed for inhalation studiesto produce a consistent amount of micrometer-sizedparticles (Hinds, 1980). An eight-stage cascadeimpactor (Marple 290, Thermo-Electron Corp.,Waltham, MA, USA) was used to determine the sizedistribution of the aerosol during each trial. The massmedian aerodynamic diameter (MMAD) and geo-metric standard deviation (GSD) values were deter-mined by first applying correction factors toaccount for stage losses (Rubow et al., 1987) andthen plotting the cumulative frequency of mass oneach stage versus stage cut diameter (Hinds, 1982).

A pair of each sampler type was attached to themannequin during each trial. One of each of the foursamplers was randomly located across the front ofthe mannequin, and the corresponding sampler wasplaced on the other side of the mannequin and di-rectly opposite to its pair. The SKC device was usedto hold the face of the CFC downward. Ten 4-h trialswere conducted for each dust type: swine, chicken,and turkey. The mannequin was turned 45� every15 min (1 revolution h�1).

An initial analysis (unreported) was conducted todetermine the spatial uniformity of dust concentra-tions inside the chamber. Four trials were conductedusing six replicate samplers (CFCs), by producingaerosols from International Organization for Stan-dardization (ISO) test dust (ISO, 1997) as describedby O’Shaughnessy et al. (2007). The averageMMAD and GSD of the aerosol produced in the fourtrials was 4.5 and 2.9 l, respectively. An analysis ofvariance (ANOVA) of the four trials by location andday showed that there were no significant differencesbetween locations (P 5 0.187) and that the overallcoefficient of variation (CV) was 8%. These resultsshowed that the aerosol concentration produced inthe chamber was spatially uniform and thus that itwas not necessary to rotate the mannequin duringthese experiments.Wind tunnel. A wind tunnel with a 1-m2 cross-

sectional area (Buchan et al., 1986) was used to eval-uate sampler performance under horizontal windvelocities of 0.2 and 1.0 m s�1. Portions of the samedusts dispersed in the still-air chamber were redis-persed in the wind tunnel with the use of a NationalBureau of Standards dust feeder (Branford VibratorCo., New Britain, CT, USA) (Hinds, 1980). Samplerswere placed on a revolving mannequin (torso andhead) in the wind tunnel to integrate the effect of

Field and wind tunnel comparison using agricultural dusts 587

wind direction. Rotation was at 1 revolution h�1. Sam-pler flow rates were identical to those described above.An eight-stage cascade impactor (Series 20-800 MarkII, Thermo-Electron Corp.) was used during windtunnel experiments to determine the aerosol size distri-bution. Ten trials were conducted for each wind speed.Preliminary trials were conducted to optimize homoge-neity of aerosol concentration across the test section ofthe wind tunnel. Coefficients of variation for the actualtrials were calculated and reported here.

Field trials

Field trials were conducted in four agriculturalenvironments—swine, chicken, turkey, and dairybarns—with 10 trials each in Iowa and in Colorado.Regional differences in climate were anticipated.Sampling times ranged from 8 to 12 h for each trial.A movable cart was outfitted with sampling pumpsand a slowly rotating mannequin form (head andtorso). A motorized turning plate was equipped witha timer to turn the mannequin 1 revolution h�1 to in-tegrate the effect of unidirectional drafts in a building.Two of each sampler type were attached to the man-nequin and operated as described above. Detailed de-scriptions of the study barns are presented in Table 1.Sampling devices were located as close to the centerof buildings as possible, depending on access to elec-trical power sources. An eight-stage cascade impac-tor (Marple 290, Thermo Scientific, Franklin, MA,USA) was used to determine the size distribution ofthe aerosol during each trial.

Statistical analysis

Data were analyzed using Microsoft Excel andSAS v.9.1 (SAS Institute, Cary, NC, USA). TheUNIVARIATE procedure was used to evaluate distri-butions of data (both concentrations and ratios ofconcentrations from different samplers) sorted by

dust type, wind speed, and sampling device. Dustconcentrations (milligrams per cubic meter) werefound to be normally distributed (Kolmogorov–Smirnov statistic). The IOM was used as the basisfor calculating ratios of the dust concentrations deter-mined by the different samplers—i.e. CFC/IOM,Button/IOM, and Cyclone/IOM. The ratios of dustconcentrations for the CFC/IOM, Button/IOM, andCyclone/IOM were found to be lognormally distrib-uted; as a result, the log10 transforms of these ratioswere used in statistical analyses. CVs of dust concen-trations for replicates of each sampler were calcu-lated to evaluate reproducibility. CV values werecomputed for each pair of samplers as the standarddeviation divided by the arithmetic mean of the twosampler concentrations. Pearson correlations for dustconcentrations determined among the four samplingdevices were calculated for field and laboratory stud-ies separately. General linear model (GLM) ANOVAwas used to statistically evaluate factors contributingto the variation in both the dust concentrations andthe ratios of the dust concentrations. ANOVAs wererun separately for each sampler ratio type. For labo-ratory trials, main effects in the ANOVA includeddust type and wind speed; for field trials, main effectsincluded dust type and location (Iowa versus Colora-do). Interactions of main effects were included inall ANOVA models. Tests for specific differencesamong least-square means in both analyses usedthe Tukey–Kramer adjustment for multiple compari-sons. Linear regressions of dust concentrations forthe CFC, Button, and Cyclone versus the IOM wereconducted by dust type and wind speed for the labo-ratory trials and by dust type for the field trials.Regressions were forced through the origin in orderto determine a single ‘slope factor’ relating the twosamplers compared (Eisenhauer, 2003). Based ondata from the laboratory studies, stepwise linear

Table 1. Building characteristics in the field studies of dust samplers in Colorado and Iowa

Building Type Number ofanimals/birds

Ventilation Feeddispensing

Manure handling Sampler location

Swine Colorado Finishing 1000 Natural Mechanical Shallow pit 2 m from feeddispensing unit

Swine Iowa Finishing 1100 Mechanical Mechanical Deep pit Middle of building

Chicken Colorado Egg laying 375 000 Mechanical Mechanical Rotating beltcollection

2 m from pens

Chicken Iowa Egg laying 100 000 Natural Mechanical Slurry pit Centered near oneend of building

Turkey Colorado Finishing 7500 Natural withmechanical shades

Mechanical Flooring, scrapedbetween cycles

Middle of building

Turkey Iowa Finishing 6500 Natural withmechanical shades

Mechanical Flooring withwood chips scraped

Centered near oneend of the building

Dairy Colorado Milkingand heifers

1500 Natural withmechanical in parlors

Via tractorto pens

Flooring, flushedwith water

2 m from one stall

Dairy Iowa Milking 50 Natural Manual Flooring, scraped Center of building

Wind speed in all buildings was estimated to be close to 0.2 m s�1. Approximate dimensions of the buildings were as follows:turkey and chicken facilities, 100 � 30 m; swine facilities, 50 � 8 m; dairies, 100 � 50 m.

588 S. J. Reynolds et al.

regression (0.05 in/out criteria) was also used todetermine models for predicting CFC, Button, andCyclone concentrations based on IOM concentra-tions while accounting for the effects of wind speed.

RESULTS

Pooled CV values indicating intrasampler preci-sion for the four sampling devices are presented inTable 2 for the laboratory and field settings. PooledCV values ranged from 0.052 to 0.465 across all con-ditions analyzed. In general, however, no noticeablydifferent trends in CV values were evident eitheramong dust types or among sampler types. CV valuespooled across all samples by dust type showed valuesranging between 0.150 for chicken and 0.225 fordairy. Likewise, overall pooled values by samplertype ranged between 0.132 for the Button and0.240 for the Cyclone.

Mean dust concentrations for the field studies arepresented in Table 3. Dust type, sampler type, andtheir interaction were all statistically significant inthe model. Under all conditions, field and laboratory(data not shown), the IOM yielded the highest esti-mate of dust concentrations as anticipated(P , 0.0001, R2 5 0.60 based on GLM ANOVA).The Button consistently showed higher concentra-tions than the CFC in the laboratory, but the two sam-plers provided similar concentrations in the fieldswine environments. The dust concentrations pro-duced in the laboratory studies were generally higherthan those found in the field but were within the sameorder of magnitude. The MMAD produced in the lab-oratory was lower than the MMAD found in the fieldmost likely because the larger particles were cap-tured in the pipework leading from the generator tothe main volume of the chamber (Table 4). However,there was fairly good agreement for the wind tunnelstudies wherein the particles were introduced directlyinto the tunnel and the wind speed prevented thelarger particles from settling. Some differences inthe MMAD measured in Iowa and Colorado for thesame type of livestock environment were observed.Data on MMAD were not available for the Coloradoswine building because the limited electric power inthis building had to be used to operate the mannequinor for the Iowa dairy building because dust levels werebelow detection limits on the stages of the impactor.

Pearson correlation coefficients among samplingdevices varied depending on both dust type andchamber air speed in the laboratory (Table 5). Forthe still-air conditions, moderate to strong positivecorrelations were found for all four devices with allthree dust types. Under conditions of increasing windspeed, far fewer consistent relationships were found.In the field, strong positive correlations were foundfor the IOM, CFC, and Button in most environments(Table 5). Correlations with the Cyclone were lessconsistent.

The linear regressions of the dust concentrationsfor the CFC, Button, and Cyclone versus the IOMare presented by dust type in Fig. 1 for the field stud-ies as an example of the relationships between thesamplers. Regressions were forced through the ori-gin. Because the R2 is not strictly applicable to re-gressions when the intercept is forced through zero,R2 values are not provided in these figures (Eisenhauer,

Table 2. Pooled coefficients of variation for dustconcentrations from paired samplers

Laboratory Studies (n 5 10 pairs)

Sampler Swine Chicken Turkey

0 m s�1

CFC 0.208 0.190 0.157

Cyclone 0.377 0.251 0.122

Button 0.143 0.149 0.052

IOM 0.288 0.091 0.208

0.2 m s�1

CFC 0.119 0.187 0.162

Cyclone 0.146 0.187 0.330

Button 0.125 0.184 0.269

IOM 0.133 0.094 0.465

1.0 m s�1

CFC 0.232 0.134 0.212

Cyclone 0.396 0.095 0.179

Button 0.145 0.143 0.124

IOM 0.092 0.108 0.327

Field studies (n 5 20 pairs)

Sampler Swine Chicken Turkey Dairy

CFC 0.119 0.129 0.105 0.260

Cyclone 0.332 0.256 0.138 0.308

Button 0.092 0.090 0.104 0.096

IOM 0.063 0.116 0.168 0.234

Table 3. Mean and (standard deviation) of dust concentrations (mg m�3) in field studies

Sampler Swine (n 5 20) Chicken (n 5 21) Turkey (n 5 20) Dairy (n 5 21)

IOM 2.97 (1.04) 2.66 (1.86) 3.52 (1.53) 0.32 (0.33)

CFC 1.52 (0.64) 1.59 (0.97) 2.05 (0.67) 0.12 (0.08)

Button 1.87 (1.03) 1.97 (1.30) 1.83 (0.63) 0.17 (0.11)

Cyclone 0.20 (0.18) 0.39 (0.45) 0.43 (0.22) 0.08 (0.09)

GLM ANOVA—R2 5 0.60, P , 0.0001. Model main effects included dust type, sampler type, and their interaction. All weresignificant. The CFC and Button were not significantly different.

Field and wind tunnel comparison using agricultural dusts 589

2003). Comparison of the regressions providesa clear comparison of sampling devices and showsthe effect of wind speed. The slopes for the variousdevices varied depending on dust type and speed.Increased wind speed resulted in smaller slopesfor all three devices relative to the IOM for allthree dusts included in the laboratory studies, al-though this effect was less pronounced for chickenand turkey dust at the highest speed (1.0 m s�1).Regression analysis found a significant interactionbetween speed and MMAD, both of which contrib-uted significantly to the prediction of Button andCassette concentrations from IOM concentrations.

The geometric means of the ratios of the dust con-centrations for the CFC, Button, and Cyclone sam-plers relative to the IOM sampler (denoted as CFC/IOM, Button/IOM, and Cyclone/IOM) are presentedin Table 6. In the laboratory studies, the ratios varied

by dust type and wind speed, with a trend toward thelowest ratios at the highest speed (1.0 m s�1) for alldevices and dust types. In general, the ratios for theButton/IOM and the CFC/IOM were similar, withthe Button/IOM ratio being consistently higher.GLM ANOVA confirmed (similar to regressions) thatall three of the ratios (CFC/IOM, Button/IOM, andCyclone/IOM) varied significantly (P , 0.0001) byboth dust type and wind speed (R2 5 0.52, 0.50,and 0.84, respectively) for field studies. The patternseen in the field studies was similar to that in the lab-oratory, except for the turkey environment for whichthe CFC/IOM ratio was higher than the Button/IOMratio. For field studies, there was no significant differencein ratios of the CFC/IOM by dust type or location (P 5

0.33). The Button/IOM ratio varied by dust type (P 5

0.0056, R2 5 0.23), and the Cyclone/IOM ratio variedby dust type and location (P , 0.0001, R2 5 0.56).

Table 4. Mean MMAD (GSD) for dust from the laboratory and field studies

Laboratory Studies (n 5 10)

Speed (m s�1) Swine Chicken Turkey

0 10.3 lm (8.6) 7.6 lm (2.6) 3.9 lm (3.1)

0.2 17.9 lm (5.8) 28.8 lm (3.8) 11.2 lm (4.3)

1.0 10.0 lm (5.6) 33.0 lm (4.2) 13.6 lm (4.3)

Field Studies (n 5 10)

Swine Chicken Turkey Dairy

Colorado Not sampled 29.1 lm (4.5) 12.0 lm (2.9) 7.7 lm (7.3)

Iowa 10.3 lm (3.3) 13.2 lm (6.6) 10.4 lm (6.4) Not sampled

Table 5. Pearson correlation coefficients for dust concentrations

Laboratory studies (n 5 10) Field studies (n 5 20)

Sampler Still chamber 0.2 m s�1 speed 1.0 m s�1 speed CFC Button Cyclone

CFC Button Cyclone CFC Button Cyclone CFC Button Cyclone

Swine

IOM 0.67* 0.86** 0.53 �0.45 0.3 �0.36 0.12 0.09 �0.43 0.90** 0.79** 0.10

CFC 0.76** 0.62* 0.44 0.76** 0.43 0.70* 0.87** 0.22

Button 0.74* 0.13 0.36 0.40

Chicken

IOM 0.81** 0.66* 0.68* 0.43 0.50 0.82** �0.07 0.42 �0.02 0.92** 0.96** 0.87**

CFC 0.87** 0.43 0.41 0.24 0.64* 0.75** 1.00 0.93** 0.79**

Button 0.53 0.37 0.65* 1.00 0.82**

Turkey

IOM 0.61 0.77** 0.28 0.63* 0.29 0.47 0.51 �0.14 0.57 0.81** 0.81** 0.32

CFC 0.80** 0.72* 0.15 0.22 0.62 �0.11 0.70** 0.44*

Button 0.56 1.00 �0.03 �0.52 0.40

Dairy

IOM 0.13 0.46* 0.14

CFC 0.87** 0.90**

Button 0.90**

*P � 0.05, **P � 0.01.

590 S. J. Reynolds et al.

DISCUSSION

This study evaluated and compared the perfor-mance of the 37-mm CFC, Button, and Cyclonerelative to the IOM in four agricultural livestock en-vironments. Aerosols from these environments weregenerated in the laboratory to evaluate the effects ofwind speed on sampler performance. Dairies werenot included in the laboratory studies because ofthe difficulty in collecting enough settled dust inbuildings to generate aerosols. The IOM was usedas the basis for comparison with the other samplers,as it is the best characterized aerosol sampler. TheCFC and Cyclone are commonly used standardmethods specified in NIOSH Methods 0500 and0600 (NIOSH, 1994a,b). Comparison of the IOMwith the CFC and Cyclone addresses the need to un-derstand the relationship between inhalable, total,and respirable size fractions, especially for applica-tions to previous studies.

The IOM and Button should be directly compara-ble since they are both reported by the manufacturerto follow the ACGIH/ISO sampling criteria for theinhalable fraction. However, the developers of theButton sampler concluded that the Button under-sampled relative to the sampling criteria under

moving-air conditions, whereas the IOM followedthe criteria well under those same conditions (Aizenberget al., 2000). Furthermore, the Button was shownto follow the sampling criteria more accuratelythan the IOM under slow-moving air conditions,whereas the IOM oversampled relative to the crite-ria so that the Button undersampled relative to theIOM in this case also (Witschger et al., 2004). Ourresults are consistent with these findings, with theButton yielding measured dust concentrationslower than the IOM. However, our results showedthe Button undersampling to the point of being morecomparable to the CFC than to the IOM. The overallaverage Button/IOM ratio for all field studies was0.65 compared with an overall average CFC/IOM ra-tio of 0.57. The large particulates (feather material) inthe chicken and turkey facility also created problemsby plugging the holes in the cap of the Button. Basedon this experience, use of the Button in some environ-ments, such as poultry production, is not recommen-ded. This problem was eliminated in the laboratorybecause settled dust was sieved to remove largerparticles before aerosolization.

Ideally, any two colocated samplers should obtainthe same concentration value. However, differencesbetween paired samplers can be attributed to even

Fig. 1. Relationship of samplers to IOM for field studies—regression (through 0) (n 5 20).

Field and wind tunnel comparison using agricultural dusts 591

slight changes in the homogeneity of the aerosol con-centration, as well as variations caused by unknownchanges in sample flow rate, differences in the con-struction of the devices by the manufacturer, and er-rors in the filter weighing process (Liden and Kenny,1992). Our results showed intrasampler differencesthat varied by CV factors as high as 46% but wereotherwise relatively consistent between 15 and 24%averaged across all sampler and dust types.

Wind speed affected sampling devices differently,as demonstrated by changes in both the slopes of re-gressions and geometric means of the ratios of dustconcentrations from samplers. In general, the geo-metric means of ratios of devices to IOMs decreased,sometimes dramatically as wind speed increased.Wind speeds as extreme as the maximum used inthese experiments (1.0 m s�1) are unlikely to befound in most enclosed livestock environments (datanot reported) but are common in hoop buildings, dou-ble curtain-walled buildings, and tunnel-ventilatedbuildings under maximum fan speed (Thorne et al.,2009). The experiments using 0.2 m s�1 wind speedare more likely to match field conditions as evi-denced by the similarity in geometric mean ratiosfor field studies and those in the 0.2 m s�1 laboratorystudies.

The poor correlations between the Cyclone and theother samplers are consistent with the pronounced bi-as found in our earlier experiments (O’Shaughnessyet al., 2007). Laboratory results also demonstratedthat higher wind velocities decreased these correla-tions. Higher wind speeds also dramatically reducedthe Cyclone/IOM ratio. Kenny et al. (1997) demon-strated that the IOM collection efficiency increasesacross all particle sizes with a decrease in wind speedand therefore the Cyclone/IOM ratio should have in-creased with wind speed. However, in this case,where two samplers with very different collection ef-ficiencies were compared, a shift in the size distribu-tion toward larger particles with an increase in windspeed apparently had a more pronounced effect onthe resulting ratio than did changes in collection effi-ciency. This shift was caused by the increased pro-pensity for larger particles to be carried to thesampler locations in the wind tunnel rather than set-tling. Therefore, one should expect a ‘decrease’ inthe ratios during conditions, such as in summermonths, when air movement through agriculturalbuildings is highest, given that the sampler that con-sistently collects the largest particles with the mostefficiency, the IOM, is used in the denominator ofthe ratio.

The Cyclone/IOM geometric mean ratios (but notother sampler ratios) were significantly higher inColorado field studies (tested in GLM). Again, thisre-emphasizes that factors in addition to the MMADare important in affecting sampler performance inthese environments. The CFC/IOM ratios were notT

able

6.

Geo

met

ric

mea

nan

d(G

SD

)fo

rra

tios

of

sam

ple

rco

nce

ntr

atio

ns

Lab

ora

tory

stu

die

s

Sam

ple

rra

tio

Sw

ine

(n5

10

each

)C

hick

en(n

51

0ea

ch)

Tu

rkey

(n5

10

each

)G

LM

AN

OV

A0

ms�

10

.2m

s�1

1.0

ms�

10

ms�

10

.2m

s�1

1.0

ms�

10

ms�

10

.2m

s�1

1.0

ms�

1R

2F

CF

C/I

OM

0.5

9(1

.4)

0.3

1(1

.6)

0.1

3(1

.6)

0.4

4(1

.4)

0.5

1(1

.4)

0.5

0(2

.2)

0.4

8(1

.5)

0.4

2(1

.3)

0.2

8(1

.3)

0.5

2,

0.0

001

Bu

tto

n/I

OM

1.0

9(1

.3)

0.4

1(1

.4)

0.2

6(1

.4)

0.8

8(1

.4)

0.6

6(1

.5)

0.6

5(1

.8)

0.6

6(1

.3)

0.5

5(1

.7)

0.3

9(1

.4)

0.5

0,

0.0

001

Cy

clo

ne/I

OM

0.2

7(1

.5)

0.0

3(2

.0)

0.0

1(2

.3)

0.2

2(1

.4)

0.0

3(1

.3)

0.0

4(2

.2)

0.2

5(1

.6)

0.1

2(1

.4)

0.0

5(1

.2)

0.8

4,

0.0

001

Fie

ldst

ud

ies

Sam

ple

rra

tio

Sw

ine

(n5

20

)C

hick

en(n

52

1)

Tu

rkey

(n5

20

)D

airy

(n5

21

)G

LM

AN

OV

AR

2F

CF

C/I

OM

0.5

0(1

.2)

0.6

7(1

.5)

0.6

0(1

.3)

0.4

9(2

.7)

0.3

3B

utt

on

/IO

M0

.57

(1.6

)0

.80

(1.4

)0

.53

(1.3

)0

.69

(2.1

)0

.23

0.0

056

Cy

clo

ne/I

OM

0.0

5(2

.0)

0.0

8(2

.3)

0.1

2(1

.7)

0.2

2(3

.8)

0.5

6,

0.0

001

592 S. J. Reynolds et al.

affected by geographic location and thus can be com-bined for prediction of ratios by building type.

Limitations

Because the mass of dust collected by a size-selective aerosol-sampling device is dependent onthe size distribution of the aerosol measured, the lackof exact concordance between the MMADs producedin the laboratory and in the field limited our ability touse the laboratory results to determine sampler-to-sampler ratios for the various dust types analyzedhere. However, the sieving process used to removethe largest particles still retained other unique quali-ties such as density and shape. Therefore, we limitedthe results from the laboratory studies to an investi-gation of the effects of wind speed on the dust types.In addition, we found that the determination of theMMAD for these aerosols is problematic becausethe cascade devices used may not accurately accountfor the large particles found in these environmentsbecause a majority of the mass collected falls inthe upper stages and therefore reduces the accuracyof an estimate of the overall size distribution. Im-proved methods for measuring the distribution ofthese types of aerosols are needed.

CONCLUSIONS

This study evaluated the performance of fourcommonly used aerosol samplers in four agriculturallivestock environments. Laboratory studies usinga still-air chamber and a wind tunnel provided impor-tant information regarding the effect of wind speedon sampler performance. In contrast to several stud-ies in other environments wherein CFC showed thelowest variability, here the IOM had the lowest CV(best precision) overall. The performance of the But-ton was also good but was negatively impacted inpoultry environments where larger (feather) particu-lates clogged the holes in the screen, causing pumpstoppages. The CFC/IOM ratios are important forcomparisons between newer and older studies. Windspeed and dust type were important factors affectingthe ratios. Based on the field studies (Table 6), a ratioof 0.56 is suggested as a conversion factor for theCFC/IOM (average for all environments since nostatistical difference). Conversion factors for theButton/IOM are swine (0.57), chicken (0.80), turkey(0.53), and dairy (0.67). Any attempt to apply a con-version factor between the Cyclone and inhalablesamplers is not recommended.

FUNDING

Centers for Disease Control and Prevention/National Institute for Occupational Safety and Health(R01 OH007841); National Institutes of Health(P30-ES05605).

Acknowledgements—The authors would like to thank SKC Inc.for support with samplers.

REFERENCES

Aizenberg V, Grinshpun SA, Willeke K et al. (2000) Perfor-mance characteristics of the Button personal inhalable aero-sol sampler. Am Ind Hyg Assoc J; 61: 398–404.

American Conference of Governmental Industrial Hygienists(ACGIH). (2007) Threshold limit values for chemical sub-stances and physical agents and biological indices. Cincin-nati, OH: ACGIH.

Bartley DL. (1998) Inhalable aerosol samplers. Appl OccupEnviron Hyg; 13: 274–8.

Buchan RM, Soderholm SC, Tillery MI. (1986) Aerosol sam-pling efficiency of 37 mm filter cassettes. Am Ind Hyg AssocJ; 47: 825–31.

De Vocht F, Huizer D, Prause M et al. (2006) Field comparisonof inhalable aerosol samplers applied in the European rubbermanufacturing industry. Int Arch Occup Environ Health; 79:621–9.

Demange M, Gorner P, Elcabache J-M et al. (2002) Field com-parison of 37-mm closed-face cassettes and IOM samplers.Appl Occup Environ Hyg; 17: 200–8.

Donham KJ. (1986) Hazardous agents in agricultural dusts andmethods of evaluation. Am J Ind Med; 10: 205–20.

Donham KJ. (1990) Health effects from work in swine confine-ment buildings. Am J Ind Med; 17: 17–25.

Donham KJ, Cumro D, Reynolds SJ. (2002) Synergistic effectsof dust and ammonia on the occupational health effects ofpoultry production workers. J Agromedicine; 8: 57–76.

Donham KJ, Cumro D, Reynolds SJ et al. (2000) Dose-response relationships between occupational aerosol expo-sures and cross-shift declines of lung function in poultryworkers: recommendations for exposure limits. J OccupEnviron Med; 42: 260–9.

Donham KJ, Reynolds SJ, Whitten P et al. (1995) Respiratorydysfunction in swine production facility workers: dose-response relationships of environmental exposures and pul-monary function. Am J Ind Med; 27: 405–18.

Donham KP, Haglind Y, Peterson R et al. (1989) Environmen-tal and health studies of farm workers in Swedish swine con-finement buildings. Br J Ind Med; 46: 31–7.

Eisenhauer JG. (2003) Regression through the origin. TeachStat; 25: 76–80.

Health and Safety Executive. (2007) EH40 workplace exposurelimits, London: Available at http://www.hse.gov.uk/coshh/table1.pdf. Accessed 28 March 2009.

Heederik D, Brouwer R, Biersteker K et al. (1991) Relationshipof airborne endotoxin and bacteria levels in pig farms withthe lung function and respiratory symptoms of farmers. IntArch Occup Environ Health; 62: 595–601.

Hinds WC. (1980) Dry-dispersion aerosol generators. In Willeke K,editor. Generation of aerosols and facilities for exposureexperiments. Ann Arbor, MI: Ann Arbor Science PublishersInc. pp. 171–87.

Hinds WC. (1982) Aerosol technology: properties, behavior,and measurement of airborne particles. New York: JohnWiley & Sons. pp. 87–90.

Hinners RG, Burkart JK, Punte BS. (1968) Animal inhalationexposure chambers. Arch Environ Health; 16: 194–206.

International Organization for Standardization. (1997) Interna-tional Standard ISO 12103-1:1997. Road vehicles—test dustfor filter evaluation—Part 1: Arizona test dust. Geneva,Switzerland: International Organization for Standardization.

Kenny LC, Aitken R, Chalmers C et al. (1997) A collaborativeEuropean study of personal inhalable aerosol sampler perfor-mance. Ann Occup Hyg; 41: 135–53.

Kirychuk SP, Dosman JA, Reynolds SJ et al. (2006) Total dustand endotoxin in poultry operations: comparison between

Field and wind tunnel comparison using agricultural dusts 593

cage and floor rearing and respiratory effects of workers.J Occup Environ Med; 48: 741–8.

Kullman G, Thorne PS, Waldron PF et al. (1998) Organic dustexposures from work in dairy barns. Am Ind Hyg Assoc J;59: 403–13.

Li S-N, Lundgren DA, Rovell-Rixx D. (2000) Evaluation of sixinhalable aerosol samplers. Am Ind Hyg Assoc J; 61:506–16.

Liden G, Kenny LC. (1992) The performance of respirable dustsamplers: sampler bias, precision and inaccuracy. AnnOccup Hyg; 36: 1–22.

Liden G, Melin B, Lidblom A et al. (2000) Personal samplingin parallel with open-face filter cassettes and IOM samplersfor inhalable dust—implications for occupational exposurelimits. Appl Occup Environ Hyg; 15: 263–76.

Mahar S, Reynolds SJ, Thorne PS. (1999) Worker exposures toparticulates, endotoxins, and bioaerosols in two refuse-derived fuel plants. Am Ind Hyg Assoc J; 60: 679–83.

Mark D, Lyons CP, Upton SL et al. (1994) Wind tunnel testingof the sampling efficiency of personal inhalable aerosol sam-plers. J Aerosol Sci; 25 (Suppl 1): 339–40.

Merchant JA, Reynolds SJ, Zwerling C. (1995) Assessment ofworkers. In McDonald JC, editor. The epidemiology ofwork-related diseases. London, UK: BMJ Publishing Group.

Merchant JA, Zwerling C, Burmeister LF et al. (2002) Chronicdisease and injury in an agricultural county: The KeokukCounty Rural Health Cohort Study. J Rural Health; 18: 521–35.

National Institute for Occupational Safety and Health(NIOSH). (1994a) Particulates not otherwise regulated, re-spirable (0600), NIOSH manual of analytical methods. 4thedn. Cincinnati, OH: NIOSH.

National Institute for Occupational Safety and Health(NIOSH). (1994b) Particulates not otherwise regulated, total(0500), NIOSH manual of analytical methods. 4th edn. Cin-cinnati, OH: NIOSH.

Occupational Safety and Health Administration. (2009) Per-missible exposure limits. Washington, DC: USDOL.

O’Shaughnessy PT, Lo J, Golla V et al. (2007) Correction ofsampler-to-sampler comparisons based on aerosol size distri-bution. J Occup Environ Hyg; 4: 237–45.

Popendorf WJ, Reynolds SJ. (1997) Industrial hygiene evalua-tion of the agricultural environment. In Langley R, editor.

Agriculture, forestry, and fishing. Rockville, MD: Govern-ment Institutes. pp. 439–78.

Predicala BZ, Maghirang RG. (2003) Field comparison of in-halableand totaldust samplers for assessing airborne dust in swineconfinement barns. Appl Occup Environ Hyg; 18: 694–701.

Reynolds SJ, Donham KJ, Whitten P et al. (1996) Longitudinalevaluation of dose response relationships for environmentalexposure and pulmonary function in swine production work-ers. Am J Ind Med; 29: 33–40.

Reynolds SJ, Parker D, Vesley D et al. (1993) Cross-sectionalepidemiological study of respiratory disease in turkey farm-ers. Am J Ind Med; 24: 713–22.

Rubow KL, Marple VA, Olin J et al. (1987) A personal cascadeimpactor: design, evaluation and calibration. Am Ind HygAssoc J; 48: 532–8.

Rylander R, Jacobs R. (1994) Organic dusts-exposure, effects,and prevention. Boca Raton, FL: Lewis Publishers and CRCPress Inc. pp. 219–46.

Sonesson A, Larsson L, Schutz A et al. (1990) Comparison ofthe limulus amebocyte lysate test and gas chromatography-mass spectrometry for measuring lipopolysaccharides (endo-toxins) in airborne dust from poultry processing industries.Appl Environ Microbiol; 56: 1271–8.

Tatum VL, Ray AE, Rovell-Rixx DC. (2001) The performanceof personal inhalable dust samplers in wood-products indus-try facilities. Appl Occup Environ Hyg; 16: 763–9.

Teikari M, Linnainmaa M, Laitinen J et al. (2003) Laboratoryand field testing of particle size-selective sampling methodsfor mineral dusts. Am Ind Hyg Assoc J; 64: 312–8.

Thorne PS, Ansley A, Perry SS. (2009) Concentrations of bioaer-osols, odors and hydrogen sulfide inside and downwind fromtwo types of swine livestock operations. J Occup Environ Hyg;6: 211–20.

Vaughn NP, Chalmers CP, Botham RA. (1990) Field compari-son of personal samplers for inhalable dust. Ann Occup Hyg;34: 553–73.

Vincent JH. (1995) Aerosol science for industrial hygienists.Tarrytown, NY: Elsevier Science Inc. pp. 278–9.

Witschger O, Grinshpun SA, Fauvel S et al. (2004) Perfor-mance of personal inhalable aerosol samplers in very slowlymoving air when facing the aerosol source. Ann Occup Hyg;48: 351–68.

594 S. J. Reynolds et al.