Full text of DOI:10.1080/10473289.2001.10464254

This article was downloaded by: [Stony Brook University]
On: 30 October 2014, At: 07:28
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK
Journal of the Air & Waste Management Association
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/uawm20
Comparison of Real-Time Instruments Used To Monitor
Airborne Particulate Matter
Albert Chung a , Daniel P.Y. Chang a , Michael J. Kleeman a , Kevin D. Perry b , Thomas A.
Cahill b , Dabrina Dutcher b c , Eric M. McDougall d & Kenneth Stroud d
a Department of Civil and Environmental Engineering , University of California , Davis ,
California , USA
b DELTA Air Quality Group , University of California, Davis , California , USA
c Rupprecht & Patashnick, Inc , Albany , New York , USA
d California Air Resources Board , Sacramento , California , USA
Published online: 27 Dec 2011.
To cite this article: Albert Chung , Daniel P.Y. Chang , Michael J. Kleeman , Kevin D. Perry , Thomas A. Cahill , Dabrina
Dutcher , Eric M. McDougall & Kenneth Stroud (2001) Comparison of Real-Time Instruments Used To Monitor Airborne
Particulate Matter, Journal of the Air & Waste Management Association, 51:1, 109-120, DOI: 10.1080/10473289.2001.10464254
To link to this article: http://dx.doi.org/10.1080/10473289.2001.10464254
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions

TECHNICAL PAPER
ISSN 1047-3289 J. Air & Waste ManagCe. Ahsusonc.g51e:1t0a9-l1.20
Copyright 2001 Air & Waste Management Association
Comparison of Real-Time Instruments Used To Monitor
Airborne Particulate Matter
Albert Chung, Daniel P.Y. Chang, and Michael J. Kleeman
Department of Civil and Environmental Engineering, University of California, Davis, California
Kevin D. Perryand Thomas A. Cahill
DELTA Air Quality Group, University of California, Davis, California
Dabrina Dutcher
DELTA Air Quality Group, University of California, Davis, California, andRupprecht & Patashnick, Inc.,
Albany, New York
Eric M. McDougalland Kenneth Stroud
California Air Resources Board, Sacramento, California
ABSTRACT
made with a tapered element oscillating microbalance
(TEOM) were found to be lower than the reference filter-
based measurements by an amount approximately equal
to the concentration of NH4NO3 observed to be present
in the airborne particles. The performance of the Dusttrak
sampler and the integrating nephelometer was affected
by the size distribution of airborne particulate matter. The
performance of the BAM, the integrating nephelometer,
the CAMM, the Dusttrak sampler, and the TEOM was not
strongly affected by temperature, relative humidity, wind
speed, or wind direction within the range of conditions
encountered in the current study. Based on instrument per-
formance, the BAM, the integrating nephelometer, and the
CAMM appear to be suitable candidates for deployment in
a real-time continuous PM2.5 monitoring network in cen-
tral California for the range of winter conditions and aero-
sol composition encountered during the study.
Measurements collected using five real-time continuous
airborne particle monitors were compared to measure-
ments made using reference filter-based samplers at Bakers-
field, CA, between December 2, 1998, and January 31,
1999. The purpose of this analysis was to evaluate the
suitability of each instrument for use in a real-time con-
tinuous monitoring network designed to measure the mass
of airborne particles with an aerodynamic diam less than
2.5 µm (PM2.5) under wintertime conditions in the south-
ern San Joaquin Valley. Measurements of airborne par-
ticulate mass made with a beta attenuation monitor
(BAM), an integrating nephelometer, and a continuous
aerosol mass monitor (CAMM) were found to correlate
well with reference measurements made with a filter-based
sampler. A Dusttrak aerosol sampler overestimated air-
borne particle concentrations by a factor of ~3 through-
out the study. Measurements of airborne particulate matter
INTRODUCTION
Over the past decade, mounting evidence has suggested that
exposure to elevated concentrations of airborne particles with
an aerodynamic diam less than 2.5 µm (PM2.5) may pose a
significant human health risk.1-5 In response to this infor-
mation, regulatory agencies with a mandate to protect pub-
lic health must now consider how to implement monitoring
networks that will allow them to measure PM2.5 concentra-
tions. Ideally, these monitoring networks will allow regula-
tory agencies to issue advisories and to implement
appropriate restrictions in target geographical locations when
PM2.5 concentrations increase to unhealthy levels.
IMPLICATIONS
Regulatory agencies need instruments that can monitor
the concentration of PM2.5 in real time so that public health
advisories can be issued during severe pollution events.
Three types of continuous, real-time instruments have
been identified that can function in a routine PM2.5 moni-
toring network during winter conditions encountered in
central California: the BAM, the integrating nephelometer,
and the CAMM. Each instrument performed well within a
broad PM2.5 concentration range composed of both vola-
tile and semi-volatile particulate matter.
Volume 51 January 2001
Journal of the Air & Waste Management Association 109

Chung et al.
Traditional monitoring networks for airborne particu-
atmosphere. The purpose of this paper is to describe the
results of an experiment conducted to compare the re-
sponse of several real-time continuous PM2.5 monitoring
instruments to measurements made using standard fil-
ter-based sampling techniques. In the following sections,
the location and design of the intercomparison experi-
ment are given, a brief description of the operation of
each instrument is provided, and the continuous airborne
particle mass measurements made by different instru-
ments are discussed.
late matter have employed filter-based samplers. While
filter-based instruments have proven to be robust and
accurate for the study of the detailed nature of airborne
particles during intensive experiments, they have several
drawbacks when used in a routine monitoring network
for regulatory purposes. The chief disadvantage associated
with filter-based samplers for PM2.5 is the fact that they
do not provide information in real time. Often weeks or
even months pass between the time when samples are
collected and when PM2.5 data become available. This time
lag makes it impossible for regulatory agencies to react
quickly to changing air quality.
A second drawback related to filter-based techniques
is that they require integrated sample times of hours to
days to collect sufficient particulate matter mass for ac-
curate analysis. Long sampling times result in the prac-
tice of averaging periods of high and low PM2.5
concentrations, making it difficult to identify the con-
centration spikes of short duration that may be respon-
sible for some of the health effects that are observed to
be associated with increased concentrations of airborne
particulate matter. A third potential drawback when us-
ing filter-based sampling methods is the possibility of
sample contamination. Great care must be exercised
when using filter-based methods to measure certain
chemical components of airborne particles that are prone
to sampling artifacts. In addition to the technical diffi-
culties listed above, filter-based sampling techniques are
labor-intensive and relatively expensive when detailed
analysis is performed.
Because of the many challenges associated with the
use of filter-based samplers, implementation of filter-based
sampling techniques in a routine monitoring network is
a difficult task. Some of the problems associated with rou-
tine monitoring for PM2.5 could be avoided by the use of a
real-time continuous instrument capable of measuring
PM2.5 concentrations. The time resolution provided by real-
time PM2.5 monitors is a function of the instrument used,
but typically is on the order of minutes. With this de-
tailed time resolution, it would be possible for regulatory
agencies to monitor peak concentration events that could
then be correlated to health effects data. In addition to
these technical advantages, continuous real-time PM2.5
monitors are relatively labor-free, and so they are capable
of providing much more data at a reduced price relative
to filter-based techniques when employed in a routine
monitoring network.
SITE DESCRIPTION
The continuous monitoring instrument intercomparison
study described in this paper was conducted at Bakers-
field, CA, between December 2, 1998, and January 31,
1999. Bakersfield is a city with a population of ~300,000
located in the southern part of California’s San Joaquin
Valley. During the winter season, the temperature inver-
sion in the atmosphere at Bakersfield is typically quite
strong and much lower than in the surrounding moun-
tains. This atmospheric condition leads to a stable sur-
face layer where air pollutant concentrations quickly build
up to extremely high levels. The region around Bakers-
field includes notable sources of air pollutants such as
agricultural operations, oil production, and vehicular
emissions from traffic on the major highways that con-
nect northern and southern California along the valley
floor (CA Highway 99 and Interstate 5).
Meteorological conditions during the intercomparison
experiment varied greatly. The lowest temperature, rela-
tive humidity, and wind speed recorded during the study
were –0.7 °C, 21%, and 0 m/sec, respectively, while the
highest temperature, relative humidity, and wind speed
were 22 °C, 99%, and 16 m/sec, respectively. Hourly PM2.5
mass concentrations varied from 5 to 130 µg/m3, while
hourly PM10 concentrations varied from 10 to 150 µg/m3.
The broad range of meteorological conditions and par-
ticulate matter concentrations recorded at Bakersfield al-
lowed for an evaluation of instrument performance under
the wide variety of atmospheric conditions that can be
encountered when operating a routine monitoring net-
work in California.
INSTRUMENT DESCRIPTION
The continuous real-time airborne particle monitors
evaluated in the current study included a beta attenua-
tion monitor (BAM 1020, Met One Instruments), two
integrating nephelometers (Optec NGN-2 and Optec
NGN-3, Air Resource Specialists), a continuous aerosol
mass monitor (CAMM, Andersen Instruments), a
Dusttrak sampler (Dusttrak 8520, TSI Inc.), and a tapered
element oscillating microbalance (TEOM 1400, R&P Inc.).
Each instrument was equipped with either a PM2.5 inlet
Before continuous real-time PM2.5 monitors are widely
used in routine monitoring networks, confidence must
be established that these instruments are capable of accu-
rately measuring PM2.5 concentrations under the wide va-
riety of conditions typically encountered in the ambient
110 Journal of the Air & Waste Management Association
Volume 51 January 2001

Chung et al.
or a PM10 inlet to allow for sampling of a target size range
of airborne particles. Those instruments equipped with
PM10 inlets were considered to be “candidate” monitors
that could be adapted to measure PM2.5 concentrations
by changing the inlet configuration. Each of the con-
tinuous instruments was operated for the maximum
possible time during the study period subject to logisti-
cal and maintenance constraints. The recorded values
from each instrument were then summarized into daily
average concentrations that could be compared to the
reference sampler measurements.
Size Selective Inlet Sampler. The SSI uses a size-selective inlet
to remove PM10 from the sample stream. PM10 is collected
on an 8- × 10-in. quartz filter that offers high collection
efficiencies and is resistant to absorbing artifacts related to
the collection of sulfates and nitrates. In the present study,
the sampler operated at a flow rate of 40 cfm (1133 L/min)
to collect a large amount of particulate matter for precise
analysis. The SSI sampler used in the current study was
operated on the roof of the sampling site between Decem-
ber 2, 1998, and January 31, 1999.
The baseline against which all continuous, real-time
PM2.5 instruments were compared in the current study is
a Federal Reference Method (FRM) sampler for PM2.5 con-
centrations (RAAS 2.5-300, Andersen Instruments). Instru-
ments that were configured to measure PM10 were
compared to a size-selective inlet high-volume sampler
(SSI) (Sierra-Andersen, 1200). The FRM and SSI samplers
are filter-based instruments that are approved by the U.S.
Environmental Protection Agency (EPA) for the measure-
ment of PM2.5 and PM10 concentrations. In the present
work, duplicate FRM and SSI samplers were collocated at
the Bakersfield site to ensure that the baseline results were
reproducible. The measurement principles and the details
of operation for each of the instruments used in the
intercomparison study are described in the next sections.
Beta Attenuation Monitor. The automatic BAM consists of
a size-selective inlet, a filter tape, a beta radiation source,
and a beta radiation detector. Particles smaller than the
cut-diameter of the size-selective inlet are collected at a
single point on a length of filter tape. The difference in
the transmission of beta radiation through the filter tape
before and after a particulate sample has been collected
is measured and used to determine the mass of collected
particulate matter. The mass absorption coefficient for
beta radiation is determined through the measurement
of a series of known standards that bracket the mass range
of interest.6 Continuous operation is achieved by an au-
tomatic mechanism that advances the filter tape between
sampling events. Since the baseline attenuation signal
is measured before each sampling event, significant drift
in the baseline signal does not occur. In the current study,
each of the beta attenuation instruments was operated
with hourly time resolution.
Typical operation protocols for the BAM specify heat-
ing of the inlet line to a temperature of ~30 °C to reduce
relative humidity to below 60%. This methodology mini-
mizes particle-bound water, but it may also bias the par-
ticulate measurements when large amounts of volatile
particulate matter are present. Other potential sources of
measurement error associated with BAMs include a slight
sensitivity to hydrogen ion concentration present in air-
borne particles,6 and the fluctuations of the sample flow
rate due to pressure, relative humidity, and temperature
variations.7 Each of these sources of error are relatively
minor, and studies have shown that BAMs can measure
airborne particulate matter with accuracy similar to that
achieved by filter-based gravimetric methods.6,8,9
Federal Method Reference Sampler. The FRM sampler con-
sists of a well impactor ninety-six (WINS impactor) followed
by a Teflon filter. Particles in the sample stream with an
aerodynamic diam greater than 2.5 µm are captured in the
WINS impactor, while smaller particles are collected on the
downstream 47-mm Teflon filter. The concentration of air-
borne particulate matter is determined by pre- and post-
weighing the Teflon and then dividing the accumulated
mass by the amount of air that was sampled. Protocols for
the preparation and handling of filters during sampling
events followed the EPA PM2.5 guidelines referenced in the
Code of Federal Regulations (40 CFR, Part 50, Appendix L).
Filters were pre-weighed no earlier than 2 weeks before sam-
pling and loaded into the FRM sampler no earlier than
3 days before sampling. The temperature of each filter was
monitored while installed in the FRM sampler and did not
exceed the ambient temperature by more than 5 °C. Filters
were recovered from the FRM sampler within 96 hr after
sampling and were maintained at a temperature of 4 °C
during transportation and storage. After arriving in the lab,
filters were allowed to equilibrate with ambient air at a
temperature of 20–23 ºC and a relative humidity of
30–40% for 24 hr prior to post-weighing. Data from the
FRM-WINS sampler used in the current study were collected
from January 6 to January 31, 1999, using samplers located
on the roof of the study site.
It should be noted that beta radiation sources used in
BAMs may be subject to safety and environmental regu-
lations that restrict operation and disposal of the instru-
ment. Each BAM used in the current experiment was
equipped with a low-activity Carbon 14 beta isotope
source with a strength less than 60 µC. The half-life of
this beta particle source is over 5000 years, while the ex-
pected lifetime of the other sampler components is only
12–15 years; when a BAM is retired, the beta source can
be recycled for re-use in another unit.
Volume 51 January 2001
Journal of the Air & Waste Management Association 111

Chung et al.
Two automatic BAMs equipped with different inlets
The response time of the integrating nephelometer
can be adjusted to measure particle concentrations every
2 min, making it an attractive choice for real-time moni-
toring in situations where short time resolution is required.
In the present study, the measurements taken by the Optec
NGN-2 nephelometer were averaged over 1-hr intervals
and then summarized for comparison with the filter-based
reference measurements taken at Bakersfield during the
period of December 17, 1998, to January 31, 1999. Mea-
surements from the Optec NGN-3 nephelometer were only
available from December 17, 18, 23, and 24, 1998, so only
a limited analysis of this instrument is possible.
were evaluated during the intercomparison study at Bakers-
field. The first BAM was operated with a WINS PM2.5 im-
pactor inlet for the entire study period. The second BAM
was equipped with a PM10 inlet between December 15, 1998,
and January 6, 1999. After January 6, the second BAM was
retrofitted with a sharp cut cyclone (SCC 16.7, BGI Inc.)
calibrated to measure PM2.5. Both BAMs were located in-
side a sheltered environment and obtained samples of
ambient air through sample lines ~2.5 m in length.
Integrating Nephelometer. The integrating nephelometers
used in the current study measured particle concentra-
tions by intersecting an aerosol sample with light at sev-
eral wavelengths in the visible range (0.55 µm effective
center wavelength). The amount of laser light scattered
at angles between 5 and 175º by particles suspended in
the sample flow was measured and used to estimate a par-
ticle-scattering coefficient. This scattering coefficient was
then assumed to be proportional to the mass of airborne
particulate matter. The proportionality constant used to
calculate airborne particle concentrations during the study
period was calculated from historical data taken by the
Interagency Monitoring of Protected Visual Environments
(IMPROVE) sampling network throughout the United
States. It should be noted that the greatest problem asso-
ciated with the use of nephelometers in routine PM2.5
monitoring networks is the variability of light scattering
by particles with different sizes and compositions. The
uncertainty associated with nephelometer measurements
increases when a wide distribution of particle sizes is
sampled,10-12 and the proportionality constant used to
convert nephelometer scattering data to airborne particle
mass concentrations may need to be adjusted when aero-
sol size distributions and composition differ significantly
from the size and composition of the calibration aerosol.
Two integrating nephelometers with different inlet
configurations were operated on the roof of the sampling
site in the current study. The first nephelometer (Optec
NGN-2) employed a direct measurement of the airborne
particle-scattering coefficient with no selection to differ-
entiate between particles with different diameters. The
second nephelometer (Optec NGN-3) used a two-stage
impactor inlet (Airmetrics Co.) to remove particles with
an aerodynamic diam greater than 2.5 µm from the sample
stream. The sample chamber of both nephelometers was
maintained at a temperature ~10 ºC above ambient to
reduce relative humidity to less than 60%. This method-
ology minimizes interference in the technique associated
with airborne water vapor and reduces the amount of
particle-bound water, but it may bias the particulate mea-
surements when large amounts of volatile particulate
matter are present.
Continuous Aerosol Mass Monitor. The CAMM measures air-
borne particulate matter based on the relative difference
in pressure drop between a reference airstream channel and
an airstream channel where particulate matter is collected
on a filter tape. This pressure difference is then correlated
to the airborne particulate matter concentration. The in-
strument advances the filter tape in the sampling channel
to get continuous measurements with a time resolution
between 30 and 60 min. An impactor with a cut size of
2.5-µm aerodynamic particle diam was operated upstream
of the CAMM, and a diffusion dryer was employed in the
sample stream to reduce any particle-bound water. Past stud-
ies have shown good correlation between the CAMM and
conventional impactor samplers.13 In the current study, the
CAMM sampler was operated on the roof of the sampling
site from January 12 to January 28, 1999.
Dusttrak. The Dusttrak sampler is a compact commercial
instrument that includes an impaction inlet, laser diode,
focusing optics, photodetector, vacuum pump, and con-
trol electronics. The instrument draws a continuous aero-
sol stream through the impaction inlet where particles
with aerodynamic diam greater than 10 µm are removed
(impaction inlets with 1- and 2.5-µm cuts are also avail-
able). The sample stream then passes through a sensing
chamber where it is intersected by light (0.78-µm wave-
length) emitted from the laser diode. Particles in the
sample stream scatter light in all directions; the amount
of light measured by the photodetector oriented at an
angle of 90° to both the laser and the sample stream is
converted to a proportional mass concentration by inter-
nal electronics. The proportionality constant used by the
Dusttrak sampler is determined by calibration against a
gravimetric measurement of the International Organiza-
tion for Standardization (ISO) 12103-1, A1 test dust (Ari-
zona Test Dust). The Dusttrak cannot detect particles with
diam smaller than 0.1 µm, and the amount of light scat-
tered by particles with diam smaller than 0.25 µm is pro-
portional to particle diameter raised to the sixth power
(Dp6). These effects can cause the Dusttrak measurements
112 Journal of the Air & Waste Management Association
Volume 51 January 2001

Chung et al.
to differ from gravimetric measurements of airborne par-
ticulate matter when the size distribution of the airborne
particles differs significantly from the size distribution of
the test aerosol. The Dusttrak has a minimum time reso-
lution of 1 min; in the present study, measurements were
averaged to an hourly resolution. Dusttrak measurements
were taken on the roof of the sampling site between Janu-
ary 10 and January 28, 1999.
collocated instrument pair exhibiting a regression slope
and correlation coefficient very close to 1.00. These re-
sults demonstrate that the baseline FRM and SSI filter mea-
surements provide a stable data set for the evaluation of
the continuous instruments described in this paper. Two
of the candidate continuous aerosol measurement devices
evaluated in the current study also operated with collo-
cated pairs during portions of the experiment. A BAM
equipped with a WINS impactor inlet (BAM-WINS) was
operated alongside a BAM equipped with an SCC inlet
(BAM-SCC) between January 6 and January 31, 1999.
Collocated integrating nephelometers (Optec NGN-2 and
Optec NGN-3) also were operated on December 17, 18,
23, and 24, 1998. The regression statistics summarized in
the lower two rows of Table 1 indicate that on days when
data were available, the responses of the collocated BAMs
and integrating nephelometers were virtually identical,
with regression slopes and correlation coefficients very
close to 1.00. The CAMM, the TEOM, and the Dusttrak
operated independently without the implementation of
a collocated unit.
Tapered Element Oscillating Microbalance. The TEOM mea-
sures suspended particulate matter concentrations by pass-
ing sample air and particles through a hollowed tapered
channel where the particles are collected on a filter. The
tapered inlet tube oscillates at a frequency that is inversely
proportional to the amount of sample deposited on the
collection substrate. The sample area is maintained at 50 °C
to minimize thermal expansion of the tapered channel
that may affect the oscillation frequency and to reduce
the amount of particle-bound water. The minimum time
resolution that can be measured with the TEOM is 10
min.14 In the present study, hourly average TEOM mea-
surements were recorded.
In the present study, the FRM filter-based PM2.5 mea-
surements were used as the baseline for the evaluation of
the continuous real-time PM2.5 instruments. The continu-
ous measurements from the BAM, the nephelometer, and
the CAMM were summarized to 24-hr average PM2.5 con-
centrations that were then compared to the FRM mea-
surements. Figure 1 shows the response of the BAM, the
nephelometer, and the CAMM along with the FRM mea-
surements between January 6 and January 31, 1999, at
Bakersfield. Both the nephelometer and the BAM tracked
the response of the FRM PM2.5 measurements in early Janu-
ary when PM2.5 concentrations reached a peak daily value
of 117 µg/m3. CAMM measurements were lower than the
PM2.5 reference measurements during the early part of
January. Measurements made by the BAM and the CAMM
tracked the reference FRM readings during late January
when PM2.5 concentrations fell to 3.5 µg/m3. Nephelom-
eter measurements for PM2.5 were generally lower than
the reference FRM PM2.5 measurements during periods of
low concentrations, but fell within the bounds of error
reported in previous studies.9
The TEOM provides a direct measurement of airborne
particle mass and it is relatively easy to calibrate, making
it attractive for monitoring purposes. The principle dis-
advantage associated with the TEOM is the potential for
volatilization of collected material when heating the
sample to 50°C. Previous studies have shown that semi-
volatile particle components such as NO3– and secondary
organic compounds may be lost from the sample when it
is heated, leading to an underestimation of airborne par-
ticle concentrations.15-17 Other studies have shown that
in locations where the concentration of volatile particu-
late matter is low, the TEOM does not underestimate par-
ticulate matter concentrations.18
A single TEOM equipped with a Sierra-Andersen PM10
dichot head was used at the Bakersfield site to measure air-
borne particle concentrations from December 2, 1998, to
January 31, 1999. The TEOM was operated inside a shel-
tered environment and obtained samples of ambient air
through a sample line with a length of ~2.5 m. This instru-
ment has been included in the present analysis because it
can easily be converted to a PM2.5 configuration and because
TEOMs are commonly used for routine monitoring purposes.
Table 1. Regression statistics for collocated airborne particle sampling instruments.
RESULTS
Regression
Slope
Correlation
Coefficient
Intercept
The regression statistics comparing the response of collo-
cated airborne particle sampling instruments are summa-
rized in Table 1. The first two rows of Table 1 show that
measurements made by collocated FRM samplers (oper-
ated between January 6 and January 31, 1999) and collo-
cated SSIs (operated between December 2, 1998, and
January 31, 1999) are highly reproducible, with each
(µg/m3)
FRM/FRM
0.98
1.00
1.06
1.01
1.00
1.00
0.99
0.99
0.83
0.01
SSI/SSI
BAM-WINS/SCC
NEPH NGN-2/NGN-3
2.92
–1.64
Volume 51 January 2001
Journal of the Air & Waste Management Association 113

Chung et al.
24-hr Averaged PM2.5
Date
Figure 1. Time series of PM2.5 measurements made with a BAM, an integrating nephelometer, a CAMM, and an FRM sampler at Bakersfield
between January 6 and January 31, 1999. Measurements from each real-time instrument were averaged to 24-hr periods to match the resolution of
the reference FRM sampler.
Figure 2 illustrates the results of a regression calcula-
tion comparing the 24-hr averaged BAM, nephelometer,
and CAMM data to the reference FRM measurements. The
upper panel of Figure 2 shows that the BAM tracked the
FRM measurements well with a linear regression slope of
0.95, an intercept of 1.36 µg/m3, and a correlation coeffi-
cient of 0.99. The nephelometer measurements also com-
pared favorably to the FRM measurements, with a
regression slope of 1.03, an intercept of –6.41 µg/m3, and
a correlation coefficient of 0.99. The lower panel of Fig-
ure 2 shows that a comparison between the CAMM mea-
surements and the FRM measurements yields a regression
with a slope of 0.74, an intercept of 10.85 µg/m3, and a
correlation coefficient of 0.96. A more thorough analysis
of the time series and regressions are presented in the
analysis section.
For practical considerations, several candidate PM2.5
continuous monitors used in the current study were
equipped with PM10 inlets. These instruments are included
in the present analysis because they can easily be con-
verted to measure PM2.5. The ability of each continuous
monitor to measure airborne particle concentrations can
be evaluated independently of the inlet configuration by
comparing the averaged continuous results to the SSI PM10
reference filter sampler. Figure 3 shows the time series of
PM10 measurements recorded by the SSI, the BAM, the
TEOM, and the Dusttrak from December 2, 1998, to Janu-
ary 31, 1999, at Bakersfield. Measurements of PM10 recorded
by the BAM were only available between December 15,
1998, and January 6, 1999, since this instrument was
equipped with a PM2.5 inlet during other portions of the
experiment. Measurements of PM10 from the Dusttrak are
only available from January 9 to January 28, 1999, due to
maintenance difficulties. Agreement between the PM10
measurements taken by the different instruments is gen-
erally less favorable than the corresponding PM2.5 mea-
surements, but Figure 3 clearly shows that the limited BAM
measurements track the response of the filter-based refer-
ence method better than measurements taken with the
TEOM and the Dusttrak.
The linear regression of the TEOM measurements plot-
ted against the SSI reference measurements is shown in
the upper panel of Figure 4. The TEOM measurements
correlated well with the reference filter-based measure-
ments (correlation coefficient = 0.95), but the linear re-
gression slope of 0.37 indicates that the TEOM consistently
underestimated PM10 concentrations throughout the en-
tire study period. The middle panel of Figure 4 shows the
regression obtained between the BAM PM10 and the refer-
ence SSI data. Good agreement is observed between these
two sets of measurements with a linear regression slope
of 1.01, an intercept of 1.85 µg/m3, and a correlation co-
efficient of 0.99. Good agreement between PM10 BAM and
SSI measurements is consistent with the results of previ-
ous experiments.8 The lower panel of Figure 4 shows that
PM10 measurements made using the Dusttrak sampler were
114 Journal of the Air & Waste Management Association
Volume 51 January 2001

Chung et al.
highly correlated with the reference measurements, but
the Dusttrak consistently overestimated PM10 concentra-
tions in the current study.
Average BAM vs. FRM
ANALYSIS
The EPA defines PM2.5 as the mass of airborne particles
with aerodynamic diam less than 2.5 µm as measured by
the FRM. The ability of a continuous monitor to make
PM2.5 measurements that are equivalent to FRM measure-
ments can be gauged by the absolute instrument error
(defined as the difference between the instrument read-
ing and the corresponding FRM measurement) and the
relative instrument error (defined as the absolute instru-
ment error divided by FRM measurement). A strong cor-
relation between a meteorological variable and the
instrument error indicates that the continuous PM2.5 moni-
tor in question may be incapable of making measurements
that are equivalent to FRM measurements under certain
atmospheric conditions. Likewise, a strong correlation
between instrument error and particle composition indi-
cates that the continuous PM2.5 monitor in question may
not be able to properly measure the mass of certain aero-
sol components.
y = 0.95x + 1.36
R2 = 0.99
FRM (µg/m3)
Nephelometer vs. FRM
In the current study, the absolute and relative error
for each of the continuous aerosol monitors were corre-
lated with measurements of ambient temperature, rela-
tive humidity, wind speed, wind direction, and the
concentration of particulate NO3–, NH4+, SO42–, total car-
bon (TC), organic carbon (OC), and elemental carbon (EC).
The concentrations of airborne particulate matter com-
ponents used in this analysis were determined from
samples collected using the spiral ambient speciation sam-
pler (SASS, Met One Instruments). It has been determined
that the spiral inlet used by the original SASS design to
remove particles larger than 2.5 µm from the sample
stream does not provide a sharp cut at the design flow
rate of 7 L/min.19 While this issue does not affect the ac-
curacy of the chemical analysis performed on the samples
after collection, it does suggest that under certain atmo-
spheric conditions, the original version of the SASS col-
lected particles larger than 2.5 µm. A statistical comparison
between the PM2.5 measurements taken with the SASS and
the FRM samplers used in the current study yielded a cor-
relation slope of 1.00, a correlation coefficient of 0.98,
and an intercept of 1.16 µg/m3. These results compare fa-
vorably to the variability shown between the collocated
FRM samplers themselves (see Table 1), indicating that
the SASS correctly sampled the PM2.5 concentrations un-
der the atmospheric conditions experienced in the cur-
rent study. The chemical composition measurements
made by the SASS compared favorably with other mea-
surement techniques,20 once again demonstrating that
SASS measurements were not biased.
y = 1.03x – 6.41
R2 = 0.99
FRM (µg/m3)
CAMM vs. FRM
y = 0.74x + 10.85
R2 = 0.96
FRM (µg/m3)
Figure 2. Comparisons of PM2.5 measurements made by real-time
instruments and a reference FRM sampler at Bakersfield. The upper
panel shows the performance of the BAM between January 6 and
January 31, 1999. The center panel shows the performance of the
integrating nephelometer (Optec NGN-2) between December 17, 1998,
and January 31, 1999. The lower panel shows the performance of the
CAMM between January 12 and January 28, 1999.
Volume 51 January 2001
Journal of the Air & Waste Management Association 115

Chung et al.
24-hr Averaged PM10
Date
Figure 3. Time series of PM10 measurements made by a TEOM, a BAM, a Dusttrak sampler, and an SSI at Bakersfield between December 2, 1998,
and January 31, 1999. Measurements from each real-time instrument were averaged to 24-hr periods to match the resolution of the reference SSI.
Before performing a detailed analysis of instrument
error, we must first consider possible collinearity between
the independent meteorological and particle composition
variables. Table 2 summarizes the correlation coefficients
(R2) between average daily ambient temperature, relative
humidity, wind speed, wind direction, and the difference
between the maximum and minimum daily values for
temperature and relative humidity. Also included in Table
2 are the correlation coefficients between the concentra-
tions of major aerosol chemical components including
NO3–, SO42–, NH4+, TC, OC, and EC. The results shown in
Table 2 indicate that a high degree of collinearity exists
between the concentrations of particulate NO3–, NH4+, and
SO42–. This dependency may result from the fact that SO42–
formed in fog droplets persists in the particle phase after
a fog event ends. The equilibrium vapor pressure of NH3
and HNO3 above SO42– particles is significantly lower than
the equilibrium vapor pressure of these components over
a pure NH4NO3 particle.21 As a result, airborne SO42– par-
ticles tend to accumulate more NH4NO3 than particles
Table 2. Correlation coefficients (R2) for linear regression between meteorological variables and airborne particle composition at Bakersfield between December 2, 1998, and
January 31, 1999.
–
+
–
Temperature
Relative
Humidity
Wind
Wind
∆ Temperature
∆ Relative
NO3
NH4
SO4
TC
OC
EC
Speed
Direction
Humidity
Temperature
1.000
0.199
1.000
0.064
0.011
1.000
0.049
0.087
0.222
1.000
0.104
0.063
0.006
0.005
1.000
0.195
0.658
0.021
0.018
0.029
1.000
0.009 0.007 0.119 0.011 0.009 0.016
0.119 0.102 0.023 0.243 0.227 0.261
0.007 0.015 0.035 0.032 0.026 0.048
0.001 0.000 0.035 0.000 0.001 0.001
0.298 0.287 0.278 0.468 0.466 0.424
0.085 0.086 0.048 0.304 0.300 0.285
1.000 0.992a 0.773a 0.334 0.324 0.324
1.000 0.826a 0.308 0.306 0.280
1.000 0.285 0.317 0.189
Relative Humidity
Wind Speed
Wind Direction
∆ Temperature
∆ Relative Humidity
–
NO3+
NH42–
SO4
TC
1.000 0.988a 0.931a
OC
1.000 0.867a
EC
1.000
aStatistically significant pairwise correlation.
116 Journal of the Air & Waste Management Association
Volume 51 January 2001

Chung et al.
containing large amounts of hydrophobic material dur-
ing the same event.
TEOM vs. SSI
The results summarized in Table 2 also show that con-
centrations of particulate TC, OC, and EC are all highly
correlated. TC is the sum of OC and EC. Under typical
atmospheric conditions, EC concentrations are relatively
low, resulting in a large correlation between TC and OC.
The significant correlation between TC and EC could re-
sult from one large source of carbonaceous aerosol in the
study region or from relatively constant emissions from
numerous smaller sources of carbonaceous aerosol.
Table 3 summarizes the correlation between the ab-
solute and relative error of each of the continuous par-
ticle sampling instruments and the independent sets of
variables shown in Table 2. All available sample pairings
were used in this analysis to provide the most robust sta-
tistical analysis possible from the data set. Consequently,
each entry in Table 3 represents a different number of
sample points with a different level for statistical signifi-
cance. The results shown in Table 3 indicate that there is
no statistically significant correlation between the instru-
ment error (absolute or relative) and the meteorological
conditions encountered in the current study. A statisti-
cally significant correlation was found between the abso-
lute error of the Dusttrak and TEOM and the concentration
of particulate NO3–. A statistically significant relationship
also was observed between the relative error of the inte-
grating nephelometer and the concentration of particle
NO3–.
y = 0.37x + 6.63
R2 = 0.95
SSI (µg/m3)
BAM vs. SSI
y = 1.01x + 1.55
R2 = 0.99
Figure 5 summarizes the statistically significant in-
teractions between instrument error and particulate NO3–
concentrations that were identified in Table 3. The upper
panel of Figure 5 shows that the Dusttrak absolute error is
positively correlated with the aerosol NO3– concentration.
Impactor data collected during the study period show that
on days when aerosol NO3– concentrations were large, the
mass distribution of ambient aerosol peaked strongly at
~0.7-µm aerodynamic particle diam.21 Since the Dusttrak
sampler is calibrated using a test aerosol that is composed
of dust particles that have an aerodynamic diam greater
than 1.0 µm, it is not surprising that the error of the
Dusttrak sampler is correlated with particulate NO3– con-
centrations, because this instrument was not calibrated
to measure submicron particles.
SSI (µg/m3)
Dusttrak vs. SSI
y = 3.00x – 13.60
R2 = 0.92
The center panel of Figure 5 shows that the relative
error of the integrating nephelometer was negatively cor-
related with the concentration of particulate NO3– mea-
sured during the study period. As described above,
impactor data collected during the study period show that
on days when aerosol NO3– concentrations were low, the
ambient aerosol size distribution was dominated by par-
ticles with an aerodynamic diam greater than 1.0 µm.21
The integrating nephelometer measurements were converted
SSI (µg/m3)
Figure 4. Comparison between PM10 measurements made by real-
time instruments and a reference SSI at Bakersfield. The upper panel
shows the performance of the TEOM between December 2, 1998,
and January 31, 1999. The center panel shows the performance of
the BAM between December 15, 1998, and January 6, 1999. The
lower panel shows the performance of the Dusttrak sampler between
January 9 and January 28, 1999.
Volume 51 January 2001
Journal of the Air & Waste Management Association 117

Chung et al.
to PM2.5 concentrations using a calibration coefficient cal-
culated based on data collected from the IMPROVE PM2.5
sampling network. A single coefficient value cannot trans-
form the response of the integrating nephelometer to PM2.5
concentrations when the shape of the aerosol size distri-
bution changes drastically. Based on the results of the
current analysis, it appears that the nephelometer cali-
bration coefficient used in the present study accurately
transforms the light-scattering signal from particles smaller
than 1.0-µm aerodynamic diam to PM2.5 concentrations.
The absolute error of the integrating nephelometer was
small during the study period because PM2.5 concentra-
tions were lower when the airborne particle size distribu-
tion was dominated by particles with aerodynamic diam
greater than 1.0 µm. Under conditions when PM2.5 con-
centrations are large and the airborne particle size distri-
bution is dominated by particles greater than 1.0-µm
aerodynamic diam, the absolute error of the integrating
nephelometer would likely be greater.
The lower panel of Figure 5 shows that the absolute
error of the TEOM was positively correlated with the con-
centration of particulate NO3–. Data acquired during the
study period indicate that the atmospheric aerosol at
Bakersfield contained more than 65 µg/m3 of particulate
NH4NO3 during the second week of January, representing
over half of the measured PM2.5 mass at that time. The
TEOM heats the captured particle sample to a greater ex-
tent than the other instruments used in the current study
do, and so it is reasonable to expect that losses of volatile
aerosol components such as NH4NO3 would be greater in
the TEOM. To investigate the performance of the TEOM
in the current study, the amount of particulate NH4NO3
contained in particles less than 2.5 µm in aerodynamic
diam (measured using the SASS) was added to the raw
TEOM measurements to produce a “corrected” set of
TEOM measurements. This method assumes that the
majority of NH4NO3 was present in particles with an aero-
dynamic diam less than 2.5 µm, a result that is consistent
with cascade impactor measurements made at Bakersfield
during the study period.21
NO3– (µg/m3)
NO3– (µg/m3)
NO3– (µg/m3)
Figure 6 shows the time series of corrected TEOM PM10
measurements along with the SSI and BAM measurements
between January 7 and January 31, 1999. By comparison
to Figure 3, it can be seen that the corrected TEOM mea-
surements in Figure 6 track the response of the SSI refer-
ence measurements much more closely than the response
of the uncorrected TEOM measurements. A linear regres-
sion analysis between the corrected TEOM measurements
and the reference SSI measurements reveals a correlation
slope of 0.85 and a correlation coefficient of 0.98. The
improved agreement between the corrected TEOM data
and the reference measurements suggests that the error
observed in the TEOM measurements is consistent with
Figure 5. Correlation between the Dusttrak (upper panel),
nephelometer (center panel), and TEOM (lower panel) instrument error
and the concentrations of particulate NO3– at Bakersfield during the
study period. The error bars shown correspond to a confidence interval
of 99%.
118 Journal of the Air & Waste Management Association
Volume 51 January 2001

Chung et al.
Table 3. Correlation coefficients (R2) for linear regression between instrument error and independent sets of meteorological variables and particle composition variables at
Bakersfield between December 2, 1998, and January 31, 1999. NO3– and TC are used as representative variables in collinear independent variable sets.
Absolute Error
NEPH
Relative Error
BAM- BAM- BAM NEPH CAMM Dusttrak
WINS SCC NGN-2
BAM-
WINS
BAM-
SCC
BAM
CAMM Dusttrak TEOM
TEOM
NGN-2
(PM2.5
)
(PM2.5
)
(PM10)
(PM2.5
)
(PM2.5
)
(PM10) (PM10) (PM2.5) (PM 2.5) (PM10) (PM2.5) (PM2.5
)
(PM10)
(PM10)
Temperature
0.097
0.002
0.193
0.000
0.045
0.057
0.036
0.040
0.072
0.023
0.032
0.019
0.085
0.353
0.204
0.315
0.052
0.067
0.155
0.000
0.002
0.216
0.091
0.002
0.035
0.073
0.217
0.036
0.000
0.039
0.018
0.006
0.003
0.020
0.034
0.001
0.393
0.165
0.677
0.723
0.346
0.004
0.093
0.191
0.257
0.315
0.062
0.000 0.107 0.008 0.156 0.088
0.010 0.276 0.032 0.031 0.088
0.018 0.000 0.050 0.066 0.035
0.003 0.036 0.360 0.023 0.000
0.001 0.001 0.048 0.234 0.111
0.004 0.027 0.003 0.063 0.116
0.000
0.009
0.108
0.065
0.388
0.021
0.090
0.181
0.038
0.091
0.004
0.004
0.032
0.065
0.078
0.012
Relative Humidity
Wind Speed
Wind Direction
∆ Temperature
∆ Relative Humidity 0.003
–
+
2–
NO3 /NH4 /SO4
TC/OC/EC
0.016
0.016
0.956a 0.885a
0.672 0.145
0.092 0.116
0.168 0.137
0.610a 0.269
0.301 0.263
aStatistically significant pairwise correlations.
the concentration of particulate NH4NO3. The remaining
difference between the adjusted TEOM data and the ref-
erence measurements may be due to the loss of other semi-
volatile aerosol components such as organic compounds.
and January 31, 1999, at Bakersfield. Measurements from
a Dusttrak aerosol monitor overestimated PM10 by about
a factor of 3 throughout the study. Measurements of PM10
made using a TEOM were lower than the reference filter-
based PM10 measurements by an amount approximately
equal to the measured concentration of particulate
NH4NO3. The performances of the Dusttrak and integrat-
ing nephelometer aerosol monitors were affected by the
size distribution of the airborne particulate matter. The
CONCLUSIONS
PM2.5 measurements made using a BAM, an integrating
nephelometer, and a CAMM correlated well with FRM
filter-based PM2.5 measurements taken between January 6
24-hr Averaged PM10
Date
Figure 6. Time series of corrected PM10 measurements made by the TEOM at Bakersfield between December 2, 1998, and January 31, 1999. Raw
TEOM measurements are corrected by adding the amount of NH4NO3 (measured using a filter-based sampler) to the base TEOM reading.
Volume 51 January 2001
Journal of the Air & Waste Management Association 119

Chung et al.
14. Patashnick, H.; Rupprecht, E.G. J. Air & Waste Manage. Assoc. 1991,
41, 1079-1083.
15. Meyer, M.; Lijek, J.; Ono, D. J. Air & Waste Manage. Assoc. 1992, TR-
22, 1, 24-38.
16. Allen, G.; Sioutas, C.; Koutrakis, P. J. Air & Waste Manage. Assoc. 1997,
47, 682-689.
17. Ayers, G.P.; Keywood, M.D.; Gras, J.L. Atmos. Environ. 1999, 33, 3717-
3721.
18. Ono, D.; Hardebeck, E.; Parker, J.; Cox, B. Systematic Differences in
PM10 Values Measured with TEOM, Wedding, Dichot, Partisol, and Graseby
Samplers at Owens Lake, California; Great Basin Unified Air Pollution
Control District: Bishop, CA, July 1999.
19. Merrifield, T. Met One Instruments, Inc., Grants Pass, OR. Personal
communication, 2000.
performances of the BAM and the CAMM were not
strongly affected by particle composition or meteorologi-
cal conditions. The BAM, the integrating nephelometer,
and the CAMM appear to be suitable candidates for de-
ployment in a real-time continuous PM2.5 monitoring net-
work in central California for the range of winter
conditions and aerosol composition encountered during
the study.
20. Dutcher, D.; Chung, A.; Kleeman, M.J.; Miller, A.E.; Perry, K.D.; Cahill,
T.A.; Chang, D.P.Y. Instrument Intercomparison Study, Bakersfield, CA
1998–1999; Final report, California Air Resources Board, Contract No.
97-536, July 1999.
ACKNOWLEDGMENTS
This research was supported by the California Air Resources
Board under Contract No. 97-536. The authors also thank
the sponsors and participants of the California Regional
PM10/PM2.5 Air Quality Study for their support. Thanks are
due to Pamela Morgan for help with the statistical analy-
sis shown in this paper.
21. Stelson, A.W.; Senfeld, J.H. Atmos. Environ. 1982, 16, 983-993.
DISCLAIMER
The statements and conclusions in this paper are those of
the authors and not necessarily those of the California
Air Resources Board, the San Joaquin Valleywide Air Pol-
lution Study Agency, or its Policy Committee, their em-
ployees, or their members. The mention of commercial
products, their source, or their use in connection with
material reported herein is not to be construed as actual
or implied endorsement of such products.
About the Authors
Albert Chung is a graduate student in the Department of
Civil and Environmental Engineering at the University of
California at Davis, One Shields Ave., Davis, CA 95616.
Dabrina Dutcher is a staff scientist at Rupprecht &
Patashnick, Inc., 25 Corporate Cir., Albany, NY 12203. Kevin
D. Perry is an assistant professor in the Department of Me-
teorology at San Jose State University, One Washington Sq.,
San Jose, CA 95192-0104. Eric M. McDougall is an associ-
ate air pollution specialist and Kenneth Stroud is a man-
ager for the Special Purpose Monitoring and Data Support
Section at the California Air Resources Board, P.O. Box 2815,
Sacramento, CA 95812. Thomas A. Cahill is a professor in
the Department of Land, Air, and Water Resources at the
University of California at Davis, One Shields Ave., Davis,
CA 95616. Daniel P.Y. Chang is a professor and Michael J.
Kleeman is an assistant professor in the Department of Civil
and Environmental Engineering at the University of Califor-
nia at Davis, One Shields Ave., Davis, CA 95616.
REFERENCES
1. Dockery, D.W.; Pope, C.A.; Xu, X.P.; Spengler, J.D.; Ware, J.H.; Fay,
M.E.; Ferris, B.G.; Speizer, F.E. N. Engl. J. Med. 1993, 329, 1753-1759.
2. Dockery, D.W.; Pope, C.A., III. Rev. Public Health 1994, 15, 107-132.
3. Schwartz, J. Environ. Res. 1994, 64, 36-52.
4. Pope, C.A., III; Thun, M.; Namboodiri, M.N.; Dockery, D.W.; Evans,
J.S.; Speizer, F.E.; Heath, C.W., Jr. Am. J. Respir. Crit. Care Med. 1995,
151, 669-674.
5. Pope, C.A., III; Schwartz, J. Inhal. Toxicol. 1995, 7, 1-18.
6. Jaklevic, J.M.; Gatti, R.C.; Goulding, F.S.; Loo, B.W. Environ. Sci. Technol.
1981, 15, 680-686.
7. Spagnolo, G.S. J. Aerosol Sci. 1989, 20, 19-27.
8. Wedding, J.B.; Weigand, M.A. J. Air & Waste Manage. Assoc. 1993, 43,
475-479.
9. Arnold, S.; Hague, W.; Pierce, G.; Sheetz, R. In PM Standards and Non-
Traditional Particulate Source Controls; Chow, J.C.,1O0 no, D.M., Eds.; Air
& Waste Management Association: Pittsburgh, PA, 1992; pp 13-23.
10. Rabinoff, R.A.; Herman, B.M. J. Appl. Meterol. 1973, 12, 184-186.
11. Anderson, T.L.; Covert, D.S.; Marshall, S.F.; Laucks, M.L.; Charlson,
R.J.; Waggoner, A.P.; Ogren, J.A.; Caldow, R.; Holm, R.L.; Quant, F.R.;
Sem, G.J.; Weidensohler, A.; Ahlquist, N.A.; Bates, T.S. J. Atmos. Oce-
anic Technol. 1996, 13, 967-986.
12. Reed, G.T.; Howser, P. Meas. Sci. Technol. 1981, 6, 422-428.
13. Babich, P.; Wang, P.Y.; Allen, G.; Sioutas, C.; Koutrakis, P. Aerosol Sci.
Technol. 2000, 32, 309-324.
120 Journal of the Air & Waste Management Association
Volume 51 January 2001