Designing monitoring networks to represent outdoor human exposure
-
Upload
independent -
Category
Documents
-
view
0 -
download
0
Transcript of Designing monitoring networks to represent outdoor human exposure
Designing monitoring networks to represent outdoorhuman exposure
Judith C. Chow a,*, Johann P. Engelbrecht a, John G. Watson a,William E. Wilson b, Neil H. Frank c, Tan Zhu d
a Desert Research Institute, 2215 Raggio Parkway, Reno, NV 89512, USAb US Environmental Protection Agency, National Center for Environmental Assessment, MD 52,
Research Triangle Park, NC 27711, USAc US Environmental Protection Agency, Office of Air Quality Planning and Standards, MD C304-01,
Research Triangle Park, NC 27711, USAd Nankai University, Environmental Science & Engineering Institute, Tianjin 300071, China
Received 12 January 2000; accepted 28 November 2001
Abstract
Measurements of outdoor human exposure to suspended particulate matter (PM) are always constrained by
available resources. An effective network design requires tradeoffs between variables measured, the number of sampling
locations, sample duration, and sampling frequency. Sampling sites are needed to represent neighborhood and urban
spatial scales with minimal influences from nearby sources. Although most PM measurements for determining com-
pliance with standards are taken over 24-h periods every third to sixth day, outdoor human exposure assessment re-
quires measurements taken continuously throughout the day, preferably over durations of 1 h or less. More detailed
particle size and chemistry data are also desirable, as smaller size fractions and specific chemicals may be better in-
dicators of adverse health effects than total mass samples.
� 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Network design; Air quality; PM2:5; PM10; Human exposure
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 962
2. Sampling objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 962
3. Sampling uncertainties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 964
4. Environmental sampling theories. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 964
5. Spatial scales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 966
6. Temporal scales . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 967
6.1. Sampling period . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 967
6.2. Sampling duration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 968
6.3. Sampling frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 969
Chemosphere 49 (2002) 961–978
www.elsevier.com/locate/chemosphere
*Corresponding author. Tel.: +1-775-674-7050; fax: +1-775-674-7009.
E-mail address: [email protected] (J.C. Chow).
0045-6535/02/$ - see front matter � 2002 Elsevier Science Ltd. All rights reserved.
PII: S0045-6535 (02 )00239-4
7. Designing an outdoor exposure network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 971
8. Particulate monitor siting criteria. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 972
9. Evaluating zones of representation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 974
10. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 974
Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 974
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 974
1. Introduction
Integrated human exposure to suspended particulate
matter (PM) is a combination of indoor and outdoor
concentrations weighted by the time spent in each en-
vironment (Moschandreas et al., 2002). Although many
people spend a significant portion of their lives indoors,
they are still exposed to ambient concentrations when
they leave a building or when outdoor PM infiltrates.
Ambient PM (especially the size range of 0.1–1.0 lm inaerodynamic diameter) can constitute a large fraction of
indoor levels (e.g., Chao and Tung, 2001; Koponen et al.,
2001; Liu and Nazaroff, 2001) that are enhanced by
contributions from cooking, smoking, house dust, and
bioaerosols. While outdoor concentrations can be reg-
ulated by public policy, increments from indoor emis-
sions are controlled by the resident and are not subject
to external standards. Empirical relationships between
adverse respiratory and circulatory effects and outdoor
PM have been observed (Vedal, 1997; Watson et al.,
1997; Gordon et al., 1998; US Environmental Protection
Agency, 2002) and are the basis for National Ambient
Air Quality Standards (NAAQS) in the United States
(US Environmental Protection Agency, 1997) and else-
where. It is important to install and operate monitoring
networks that represent the exposure of large popula-
tions to outdoor air. This paper specifies the issues that
must be considered in implementing such a network.
Outdoor PM measurements must consider sampler
locations, sample durations, sampling frequencies, and
the variables measured. Sampling networks are con-
strained by access, available technology, and monitoring
resources. Tradeoffs must be made to obtain the most
information relevant to specific objectives for a given
resource expenditure. There is no simple formula ap-
plicable to all monitoring objectives and resource levels.
This discussion draws on theoretical concepts and past
experience to provide guidance on where and when to
monitor, what to measure, and how to make tradeoffs
for a variety of sampling objectives. The examples apply
to measurements of PM2:5 and PM10 (particles with 50%
cut points at aerodynamic diameters of 2.5 and 10 lm,respectively) that are regulated by NAAQS. Current PM
NAAQS use community-representative monitors and
spatial and temporal averaging to better represent
exposure and health effects (Watson et al., 1997). Al-
though general concepts and definitions related to en-
vironmental sampling have been published (Wilson and
Spengler, 1996; Baron and Willeke, 2001; Cohen and
McCammon, 2001), these are not consistent or neces-
sarily applicable to PM.
2. Sampling objectives
Ambient and workplace air quality measurements are
made for several purposes, and these purposes influence
sampling locations, durations, frequencies, variablesmea-
sured, and number of samples required to represent a
population:
• Compliance with air quality standards: The United
States and other countries have established ambient
air quality standards for PM in outdoor air to protect
public health. Indoor air standards are available
for industrial workplaces (Samet and Spengler, 1991;
Yocom and McCarthy, 1991) but are not common
for household and office air. Particle measurements
for compliance are intended to determine if PM
mass concentrations exceed existing standards and
if trends show improvements in air quality. US PM
standards currently require PM2:5 and PM10 mass
concentrations taken on filters by officially desig-
nated measurement systems (Watson and Chow,
2001a). Samples of 24-h duration are typically taken
every third or sixth day to estimate annual averages
and 98th percentile or one expected exceedance per
year concentrations. Sampling locations are intended
to represent large populations rather than nearby
source contributions or regional background levels.
Compliance monitoring requires long-term measure-
ments over many years, depending on the form and
values of the standards. US PM standards require
a minimum of three years to determine attainment
or non-attainment. Since public policy, financial or
criminal penalties, or litigation may result from these
measurements, they must be well documented and of
unquestionable consistency.
• Alerts: Pollution levels approaching or exceeding con-
centrations that cause immediate and adverse health
consequences require decisions to issue public warn-
ings or to impose temporary emissions reductions.
Many communities in the US curtail wood-burning
during winter when concentrations appear to be ris-
962 J.C. Chow et al. / Chemosphere 49 (2002) 961–978
ing. The real-time network in Mexico City is used to
shut down industries and reduce traffic during high
pollution episodes. In situ continuous monitors with
fast response times and telemetry to a central loca-
tion are required for alerts. High accuracy and preci-
sion are not required, as long as a high danger
threshold can be quickly detected. The data returned
by alert networks must be monitored, either auto-
matically or manually, and disseminated to officials
and the public by news broadcasts or over the inter-
net. While quantitative, continuous, real-time moni-
tors have been used for several gaseous pollutants
and particle measurements, they have not been com-
pletely perfected for some PM mass surrogates (Wat-
son et al., 1998; McMurry, 2000).
• Implementation of a standard: The causes of excessive
concentrations are not always evident from emissions
inventories or compliance measurements. Observ-
ables other than mass, such as PM chemical composi-
tion and size distribution, must be measured for these
purposes. Measurements are needed with 5 min to
hourly time resolution to observe the effects of diur-
nal changes in emissions and the surface mixing layer.
Multi-year monitoring periods are needed to detect
the effects of emissions reductions that may be ob-
scured by year-to-year meteorological variability.
Table 1 summarizes several of the useful PM observ-
ables and methods available for their measurement.
• Atmospheric process research: Excessive particle con-
centrations result from complex interactions involv-
ing emissions and meteorology. Primary particles
are directly emitted from sources, while secondary
particles form in the atmosphere from gaseous pre-
cursor emissions. Networks for process research
Table 1
Measurement parameters for PM
Observable Filter-based integrated method Commercially available in situ continuous method
Mass Gravimetry (Chow, 1995) Beta attenuation monitor (BAM)
(Nader and Allen, 1960; Spurny and Kubie, 1961; Lilienfeld and Dulchinos,
1972; Husar, 1974; Cooper, 1975; Lilienfeld, 1975; Sem and Borgos, 1975;
Cooper, 1976; Macias and Husar, 1976; Jaklevic et al., 1981; Courtney et al.,
1982; Klein et al., 1984; Wedding and Weigand, 1993; Speer et al., 1997)
Tapered element oscillating microbalance (TEOM) (Patashnick and
Hemenway, 1969; Patashnick, 1987; Patashnick and Rupprecht, 1991;
Meyer et al., 1992; Rupprecht et al., 1992; Allen et al., 1997)
Continuous ambient mass monitoring system (CAMMS, a pressure drop
tape monitor) (Babich et al., 2000)
Elements X-ray fluorescence
(Watson et al., 1999)
Semicontinuous elements in aerosol system (SEAS) for elements of Al, Ca,
Fe, Cu, Cr, Mn, Zn, Cd, As, Sb, Pb, Ni, V, and Se (Kidwell and Ondov,
2001; Ondov and Kidwell, 2002a,b)
Inductively coupled plasma–mass
spectrometry (Bettinelli et al., 1989;
Tanaka et al., 1998; Chin et al.,
1999)
Inductively coupled plasma–emis-
sion spectrometry (Harman, 1989)
Ions Ion chromatography
(Chow and Watson, 1999)
Flash volatilization for sulfate or nitrate (Watson et al., 1998; Stolzenburg
and Hering, 2000)
Automated colorimetry Particulate sulfate by chemical reduction (Watson, 2002)
Atomic absorption spectrophoto-
metry (Chow et al., 2002c)
Continuous gas and particle speciation monitor by ion chromatography for
particulate sulfate, nitrate, and ammonium, and for gaseous hydrochloric
acid, nitrous oxide, nitric acid, sulfur dioxide, and ammonia
Carbon Thermal optical reflectance
(Chow et al., 1993, 2001)
Ambient particulate carbon monitor for organic and elemental carbon
(Rupprecht et al., 1995)
Thermal optical transmission
(Birch and Cary, 1996)
Thermal combustion for organic and elemental carbon (Sunset Laboratory,
Forest Grove, OR)
Thermal manganese oxidation
(Fung, 1990)
Aethalometer (Hansen et al., 1984, 1988, 1989; Hansen and Rosen, 1990)
and Radiance particle soot absorption photometer (PSAP, Bond et al., 1998;
Quinn et al., 1998) for black carbon
Differential thermal combustion or
desorption (Cachier et al., 1989;
Schmid et al., 2001)
J.C. Chow et al. / Chemosphere 49 (2002) 961–978 963
require simultaneous gas and particle measurements
at the surface and aloft with novel measurement
devices. Simultaneous surface and upper air meteoro-
logical monitoring is also needed. Research networks
are often operated for a few days or weeks during
pollution episodes and need sampling locations rep-
resenting various spatial scales.
• Determination of health effects: PMmeasurements are
needed to estimate exposure for use in epidemiologi-
cal studies, to assess human exposure, and to identify
toxic components of PM mass. Particle size, chemical
composition, and particle number are needed with
other variables that might cause respiratory and car-
diovascular distress. For acute epidemiology and ex-
posure studies, 1-h or continuous measurements may
be needed as well as 24-h measurements. For studies
of chronic epidemiology, measurements that inte-
grate over a week to a month may be more cost-effec-
tive than hourly or daily concentration averages, but
these must continue for many years. For dosimetric
studies and modeling, particle size distributions and
their changes with relative humidity and temperature
must be known. Health outcomes are affected by
non-pollution variables, such as temperature and
weather, that are often unquantified.
• Determination of ecological effects: Chemical concen-
trations in rain, fog, and dew are required to under-
stand the contributions of PM to soiling of surfaces,
damage to materials, and wet and dry deposition of
acidity and toxic substances to surface water, soil,
and plants. Some differentiation into particle size is
needed to determine dry deposition.
• Determination of radiative effects: PM directly affects
climate and visibility by scattering and absorbing
light. PM indirectly affects climate by providing cloud
condensation nuclei, thereby changing the albedo
and lifetimes of clouds. To understand these effects,
information on refractive index (including the ratio
of scattering to absorption), size distribution, and
changes in particle size with relative humidity is re-
quired. Measurements are needed that correspond
to daylight hours rather than over an entire 24-h
period (Watson, 2002).
As noted above, measurement attributes appropriate
for one of these objectives are not necessarily adequate
for others, even though the same variables are quanti-
fied.
3. Sampling uncertainties
The first uncertainty is sampling error due to popu-
lation variability, which can be assessed from a random
sample, while the second uncertainty is ascribed to ex-
perimental error from inadequate sampling procedures
and imprecise monitoring or analytical equipment. Bi-
ases may arise from:
• Sample number: Monitoring results, expected to be
free of bias, are obtained by randomly selecting
samples (observations) or by measuring the total
population, as in continuous monitoring. Sampling
conducted at set intervals, such as a specific time of
the day or day of the week, may be biased even
though a set sampling schedule is more practical to
implement. US PM networks have typically applied
a sixth-day sampling strategy to randomize effects
of emissions that differ by day of the week.
• Particle size fraction: Fine particles (PM2:5) are more
uniformly distributed over a region. PM10, PMcoarse
(difference between PM10 and PM2:5), and total
suspended PM (TSP, particles with aerodynamic
diametersK 40 lm) contain larger particles that candeposit close to the source (Kotchmar et al., 1987;
Wilson and Suh, 1997). PM2:5 also contains second-
ary sulfates and nitrates that form from sulfur diox-
ide and oxides of nitrogen gases emitted into the
atmosphere and tend to be more uniformly distrib-
uted in space than directly emitted particles com-
posed of carbon and geological materials.
• Particle composition:Water-soluble and volatile com-
pounds can bias PM measurements depending on the
temperature and relative humidity history of the
sample. Special problems are caused by: (1) water ab-
sorbed by the particles that increase PM mass; (2)
volatile and semivolatile organic compounds (VOCs
and SVOCs) that evaporate from particles, thereby
reducing mass; (3) semivolatile inorganic compounds
such as ammonium nitrate that evaporate and reduce
mass; and (4) inorganic and organic gases such as
ammonia, sulfur dioxide, oxides of nitrogen, and
polar organic compounds that adsorb onto particles
or filter media, thereby increasing mass. Non-volatile
particles include most fugitive dust, sulfates, and sta-
ble organic compounds.
4. Environmental sampling theories
Environmental sampling networks have been studied
in hydrology (Andricevic, 1990; Kassim and Kottegoda,
1991; Woldt and Bogardi, 1992; Meyer et al., 1994),
meteorology (Gandin, 1970), and the geological sci-
ences (Camisani-Calzolari, 1984; de Marsily et al., 1984;
Russo, 1984). Only a few of these concepts have been
adapted to air quality networks (Munn, 1981). Statisti-
cal methods take advantage of the fact that most air
quality measurements are correlated either in time at the
same location or in space with other monitors in a net-
work. Networks can be optimized by examining time
series correlations from long measurement records or
964 J.C. Chow et al. / Chemosphere 49 (2002) 961–978
spatial correlations among measurements from many
nearby monitors (Munn, 1975; Elsom, 1978; Hands-
combe and Elsom, 1982). Munn (1981) identifies
four types of correlation analysis: (1) time correlation
(autocorrelation) at one site; (2) cross-correlation of
several pollutant concentrations at one site; (3) spatial
correlations among simultaneous measurements at dif-
ferent sites; and (4) spatial correlations among different
sites with time lags. Other statistical measures include:
(1) the coefficient of geographic variation (Stalker and
Dickerson, 1962a,b; Stalker et al., 1962), (2) structure
functions (Goldstein et al., 1974; Goldstein and Lan-
dovitz, 1977a,b), (3) cluster analysis (Sabaton, 1976),
(4) principal component analysis (Peterson, 1970; Sab-
aton, 1976), (5) the variational principle (Wilkins, 1971),
and (6) linear programming (Darby et al., 1974; Hou-
gland, 1977).
Statistical sampling includes four approaches: (1)
random sampling, (2) systematic sampling, (3) judg-
mental sampling, and (4) geostatistical sampling. Ran-
dom sampling locates sites by chance, without taking
into consideration the sources of pollutants (Nesbitt and
Carter, 1996). Random placement is accomplished by
specifying boundaries of a rectangular domain, gener-
ating x and y coordinates from a uniform-distribu-
tion random number generator limited by the domain
boundaries, and placing samplers as close to these co-
ordinates as practical. Random sampling designs have
several advantages: (1) sampling bias is eliminated or at
least minimized; (2) simplicity of implementation with
no knowledge assumed about the spatial and temporal
distribution of concentrations; and (3) objective selec-
tion of sampling locations. The disadvantages are that:
(1) many sampling locations are needed for an accept-
able sampling error; (2) there is large potential for re-
dundancy in a network with many locations; and (3)
there is large risk of poorly representing exposures in a
network with few locations.
Borgman et al. (1996) show that for a 95% confidence
interval, an accepted error of 1 lg/m3 and a variance, r2,of 6.5 (lg/m3)2, the estimated minimum number of ob-servations (n) can be calculated from:
ffiffiffi
np
¼ 1:965r ð1Þ
yielding n ¼ 25. This is reasonable if an annual averageis being estimated, if there are a large number of PM
sampling sites, samplers are randomly distributed in
space, and the measurements are independent.
Random network siting is not usually practical for
air quality monitoring. Prior knowledge, though some-
times incomplete, is always available concerning the
sources and meteorology that affect PM concentrations
in an area. Sampler siting constraints of power, security,
accessibility, and minimum separations from nearby
emitters and obstructions impose logistical constraints
that prevent a purely random selection of measurement
locations. A community exposure monitoring philoso-
phy is not served by a random-sampling network design.
Systematic sampling overlays a grid of squares with
equal dimensions and spacing on a study domain with
one sampler assigned to each grid cell. This ‘‘area
method’’ (Noll and Miller, 1977) is most applicable in
flat terrain with a few large point sources or with nu-
merous area sources. Samplers are placed as close to the
center of the cell as practical. This method minimizes
sampling bias because of its regular sampler spacing.
However, systematic sampling requires a substantial
number of samplers, depending on the size of the study
domain, and most of these samplers supply redundant
information where PM concentrations are spatially
uniform. Owing to the large number of sites, costs of
random and systematic sampling designs may be pro-
hibitively high except for short periods during which
spatial uniformity is being evaluated.
Judgmental sampling (Nesbitt and Carter, 1996) uses
knowledge of source emissions and sensitive recep-
tor locations, coupled with mechanisms for pollutant
transport, to locate measurement sites. Noll and Miller
(1977) call this the ‘‘source orientation method’’. Air
pollution models can be used to assist in this judgment,
but this requires exceptional accuracy of the model
formulation and input data. Few areas have good esti-
mates of particle and precursor gas emissions, especially
from mobile and area sources. Complex terrain and
meteorology, as well as simulating secondary aerosol
formation, also present challenges to currently available
models for suspended particles.
Judgmental sampler locations may be determined by
data from an existing monitoring network or by iden-
tifying the locations of pollutant sources and inferring
pollutant transport from data analysis of emissions and
wind measurements. Short-term experiments involving
spatially dense measurements and modeling may assist
in making or verifying judgments. Compliance moni-
toring networks almost always use judgmental sampling
strategies that consider where source emissions are lo-
cated in relation to population centers and meteorology.
Judgmental sampling may be used to monitor hot spots
for compliance monitoring or to identify sites repre-
sentative of community exposure for epidemiology.
Nesbitt and Carter (1996) combine judgmental and
systematic sampling by applying the following steps: (1)
identify potential sources of contamination or hot spots
using existing measurements or models, (2) place a grid
system over these areas, (3) sample at these grid points,
(4) define a systematic grid at points that yield positive
contamination, and (5) use the systematic grid to assess
the remainder of the study area.
Fig. 1 shows how a judgmental sampling strategy
compares with a combined judgmental and systematic
sampling strategy. The concentration isopleths can
J.C. Chow et al. / Chemosphere 49 (2002) 961–978 965
be interpolated from spatially dense measurements or
produced by an air quality model. The judgmental
strategy, by itself, misses areas of significant concentra-
tions, while the combined judgmental and systematic
strategy covers the areas of significant concentration
that have not previously been monitored.
Another hybrid method for determining potential
locations of PM samplers is based on geostatistical
sampling (Journel, 1980; Russo, 1984; Kassim and Kot-
tegoda, 1991; Trujillo-Ventura and Ellis, 1991; Rouhani
et al., 1992a,b; Borgman et al., 1996). Kriging is a
common method for interpolation to estimate unknown
values from existing spatial data (Volpi and Gambolati,
1978; Lefohn et al., 1987; Venkatram, 1988). Kriging
uses the correlation structure to produce an estimator
with the least mean square error and results in reduced
sample size compared to other methods.
Guidance for PM compliance networks (Watson
et al., 1997) is based on judgmental network design.
These PM networks should be periodically evaluated
and redesigned to better define PM concentration gra-
dients, human exposure, and source contributions.
Because of their nature and sources, fine particles
such as PM2:5 are difficult to model (Seigneur et al.,
1997; Seigneur, 2001). Numerical models for PM are still
under development and can only be used qualitatively
for network design. Source-oriented models simulate
atmospheric diffusion or dispersion and estimate concen-
trations at defined receptors. Numerical source models
can be grouped as kinematic, first-order closure, or
second-order closure models (Bowne and Londergan,
1983). Kinematic models are the simplest both mathe-
matically and conceptually. These models simplify the
non-linear equations of turbulent motion, thereby per-
mitting a closed analytical approximation to describe
pollutant concentration (Green et al., 1980). First-order
closure models are based on the assumption of an iso-
tropic pollutant concentration field. Consequently, tur-
bulent eddy fluxes are estimated as being proportional
to the local spatial gradient of the transport quantities.
Eulerian grid models, Lagrangian particle models, and
trajectory puff/plume models are included in this cate-
gory. Second-order closure models involve a series of
algorithm transformations of the equations of state,
mass continuity, momentum, and energy by using the
Boussinesque approximation and Reynolds decomposi-
tion theory (Holton, 1982; Stull, 1988).
5. Spatial scales
Monitoring sites should be selected to represent
several spatial scales as defined below (US Environ-
mental Protection Agency, 1997). Distances indicate the
diameter of a circle, or the length and width of a grid
square, with a monitor at its center:
• Collocated or indoor scale or ducted emissions (1–10
m): Collocated monitors are intended to measure
the same air. Collocated measurements are often used
to define the precision of the monitoring method.
Different types of monitors are operated on collo-
cated scales to evaluate the equivalence of mea-
surement methods and procedures. The distance
between collocated samplers should be large enough
to preclude the air sampled by any of the devices
from being affected by any of the other devices but
small enough so that all devices obtain air containing
the same pollutant concentrations. Effluent pipes and
smoke stacks duct emissions from industrial sources
to ambient air. Pollutants are usually most concen-
trated in these ducts and are monitored to create
emissions inventories as well as to determine compli-
ance with emissions standards. There may be some
variability across the duct that can be compensated
for by longer averaging times and traverses to obtain
a composite sample.
• Microscale (10–100 m): Microscale monitors are
most often used to assess human exposure. These
monitors show differences from compliance monitors
when the receptor is next to a low-level emissions
source, such as a busy roadway. Ambient compliance
monitoring site exposure criteria avoid microscale in-
fluences even for source-oriented monitoring sites,
while source emissions monitors avoid them because
they represent emissions from a variety of sources.
Microscale sites are usually operated for short peri-
ods to define the zones of representation for other
sites and to estimate the zones of influence for ducted
and non-ducted emitters. These sites are also used to
estimate emission rates and compositions for non-
ducted sources such as suspended road dust (e.g.,
Gillies et al., 1999).
• Middle scale (100–500 m):Middle-scale monitors are
also used for human exposure studies (e.g., Engelbr-
echt and Swanepoel, 1998; Watson and Chow, 2001b;
Chow et al., 2002a), to evaluate contributions from
Fig. 1. Examples of judgmental and hybrid sampling strategies.
966 J.C. Chow et al. / Chemosphere 49 (2002) 961–978
large industrial facilities, and to evaluate the zones of
representation of compliance sites. They are also used
for process research to examine rapid changes in pol-
lutant composition, dilution, and deposition. For air
quality research, vertical resolution of pollutant con-
centrations on this scale (e.g., measurements on roofs
of tall buildings or hilltops) elucidates mechanisms
of day-to-day carryover, long-range transport, and
nighttime chemistry that cannot be understood by
surface measurements.
• Neighborhood scale (500 m to 4 km): Neighborhood-
scale monitors are used for compliance to protect pub-
lic safety and show differences that are specific to
activities in the district being monitored. The neigh-
borhood-scale dimension is often the size of emissions
and modeling grids used for air quality source appor-
tionment in large urban areas, so this zone of repre-
sentation of a monitor is the only one that should be
used to evaluate such models. Sources affecting neigh-
borhood-scale sites typically consist of small individ-
ual emitters, such as clean, paved, curbed roads,
uncongested traffic flow without a significant fraction
of heavy-duty vehicles, or neighborhood use of resi-
dential heating and cooking devices (Chow et al.,
1992, 1999, 2002b; Chow and Watson, 2001).
• Urban scale (4–100 km): Urban-scale monitors are
most common for ambient compliance networks
and are intended to represent the exposures of large
populations. Urban-scale pollutant levels are a com-
plex mixture of contributions from many sources that
are subject to areawide control. Urban-scale sites
are often located at higher elevations or away
from highly traveled roads, industries, and residential
wood-burning appliances. Monitors on the roofs of
two- to four-story buildings in the urban core area
are often good representatives of the urban scale
(Engelbrecht et al., 1998).
• Regional-scale background (100–1000 km): Regional-
scale monitors are typically located upwind of urban
areas and far from source emissions. Regional mon-
itors are not necessary to determine compliance,
but they are essential for determining emissions
reduction strategies. A large fraction of certain pollu-
tants detected in a city may be due to distant emitters,
and a regional (rather than local) control strategy
may be needed to reduce outdoor exposure. Re-
gional-scale concentrations are a combination of nat-
urally occurring substances as well as pollutants
generated in urban and industrial areas that may be
more than 100 km distant (Chow et al., 1996). Re-
gional-scale sites are best located in rural areas away
from local sources, and at higher elevations.
• Continental-scale background (1000–10 000 km):
Continental-scale background sites are located hun-
dreds of kilometers from emitters and measure
a mixture of natural and diluted manmade source
contributions. Anthropogenic components are at
minimum expected concentrations. Continental-scale
monitors determine the mixture of natural and an-
thropogenic contaminants that can affect large areas.
The InteragencyMonitoring of PROtected Visual En-
vironments (IMPROVE) network has been used ex-
tensively in the US to determine PM2:5 compositions
and their effects on visibility in National Parks and
Wilderness Areas (Eldred et al., 1997).
• Global-scale background (>10 000 km): Global-scale
background monitors quantify concentrations trans-
ported between different continents as well as natu-
rally emitted particles and precursors from oceans,
volcanoes, and windblown dust. These are located
in isolated spots such as McMurdo in Antarctica
(Mazzera et al., 2001), Mauna Loa in Hawaii
(Holmes et al., 1997), and Barrow, Alaska (Polissar
et al., 1999).
The zone of representation for a monitoring site is
often not evident in the absence of measurements from
nearby locations. Chow et al. (1999) show that larger
concentrations within urban areas are often represented
by samplers located near roadways or in neighborhoods
with construction. Most of the sites complying with the
urban-scale criteria denoted above showed similar
concentrations. Regional concentrations were achieved
within distances outside the urban area that are com-
parable to the dimensions of the city and are fairly
similar even over large distances. Zones of representa-
tion are different for different pollutants.
6. Temporal scales
Assessment of outdoor human exposure requires
time scales ranging from minutes to years. These are
defined by sampling period, sample duration, and sam-
pling frequency.
6.1. Sampling period
PM concentrations vary from year to year and season
to season, and multi-year monitoring is needed to detect
some of these changes. Fig. 2 shows annual average and
maximum PM2:5 concentrations from 1991 through 1998
at Fresno, CA, measured with a dichotomous sampler
operating on a sixth day sampling frequency. Substan-
tial differences in annual average concentrations occur
from year to year at the Fresno site. The average of the
yearly 98th percentiles (the second-highest value corre-
sponds to the 98th percentile for the sixth-day sampling
schedule) exceeded the 24-h PM2:5 NAAQS of 65 lg/m3
between 1991 and 1995. For the more recent period
between 1996 and 1998, this average did not exceed 65
lg/m3, although several individual values were in excess
J.C. Chow et al. / Chemosphere 49 (2002) 961–978 967
of the 24-h NAAQS. These extreme values have been
relatively rare occurrences, as the 75th percentile is
substantially lower than the highest values. The high
values tend to influence the average more than the me-
dian (Fig. 2).
Different seasons may also show substantial concen-
tration variability. Fig. 3 shows that the highest seasonal
average concentrations occur during winter at Fresno.
These seasonal averages are lower during fall (with the
exceptions of 1993 and 1995), but still substantially
higher than those for spring and summer. Maximum 24-
h PM2:5 concentrations are found during both winter
and fall. With rare exceptions, spring and summer
experience low averages of 10 lg/m3 or less and lowmaximum concentrations. Fig. 3 also shows that PM2:5
concentrations vary among winter seasons of different
years.
6.2. Sampling duration
PM2:5 and PM10 have typically been measured as 24-
h averages because this was the approximate time nee-
ded to extract enough particles onto a filter for accurate
gravimetric analysis. In situ particle measurement
technologies (Klein et al., 1984; Patashnick, 1987; Chow,
1995; Tsai and Cheng, 1996; Watson et al., 1998;
McMurry, 2000) have been used to approximate hourly
particle concentrations in some networks and pro-
vide detailed information about how outdoor exposure
changes throughout the day and at different locations.
Fig. 4 illustrates diurnal PM10 variations between
weekdays and weekends at various locations in Las
Vegas, NV, during 1995 (Chow and Watson, 1997).
Median (i.e., 50th percentile) PM10 concentrations for
weekdays and weekends in the north (Craig Road),
southeast (East Charleston), east downtown (City Cen-
ter), and west (Walter Johnson) sides of the Las Vegas
Valley are shown. For both weekdays and weekends, the
two sites on the fringes of Las Vegas (Craig Road and
Walter Johnson) exhibit similar patterns. The two urban
sites (East Charleston and City Center) exhibit similar
patterns. All four sites show a pronounced decrease in
the size of the morning peak on weekends. The urban
sites show concentrations peaking later than the other
two sites with persistence into the early morning hours
on both weekdays and weekends. In the early afternoon
on weekends, all sites have similar 50th percentile con-
centrations of about 20 lg/m3. PM10 decreases quickly
during late evening at the less-urban sites, which is ex-
pected when people return to their homes after work. At
the downtown urban sites, however, traffic near casinos
and restaurants continues through the late evening.
Traffic, in conjunction with low mixing heights and light
winds at this time of day, is largely responsible for the
high PM10 concentrations into the early morning hours
at these sites. Fig. 5 shows a different temporal pattern
resulting from residential coal combustion in South
African townships during the early mornings and late
afternoons when cooking fires in the braziers and stoves
are lit (Engelbrecht and Swanepoel, 1998). Peak pollu-
tion levels vary greatly from background values to
nearly 1400 lg/m3 depending on wind speed and direc-tion.
Chow (1995) observed that US particulate standards
are based on health relationships with measurements
Fig. 2. Annual average and maximum PM2:5 from 1991 to 1998 at the California Air Resources Board�s dichotomous sampler networksite in Fresno, CA. Whiskers and boxes in plot represent the 10th, 25th, median, mean (thick line), 75th, and 90th percentiles. Circles
represent data points above the 90th and below the 10th percentiles.
968 J.C. Chow et al. / Chemosphere 49 (2002) 961–978
acquired from networks designed to determine compli-
ance with previous standards. For suspended particles,
this practice has perpetuated 24-h filter sampling
that does not recognize some of the potentially high
short-term concentrations that may be a hazard. Shorter-
duration measurements that are collocated with long-
duration measurements for a limited time are needed to
evaluate the extent to which the longer-duration mea-
surements are achieving their desired objectives. The
hourly measurements shown in Figs. 4 and 5 can also be
used for alerts, even though 24-h measurements are
adequate for determining compliance with current
NAAQS.
6.3. Sampling frequency
Long-term averages need less frequent samples than
monitoring for extreme events. Table 2 compares the
Fig. 3. Seasonal summaries of PM2:5 from 1991 to 1998 at the California Air Resources Board�s dichotomous sampler network site inFresno, CA. Whiskers and boxes in the plot represent the 10th, 25th, median, mean (thick line), 75th, and 90th percentiles. Circles
represent data points above the 90th and below the 10th percentiles.
Fig. 4. Diurnal variations of hourly PM10 mass concentrations acquired with beta attenuation monitors (BAM) during winter 1995 at
several locations in Las Vegas, NV, for: (a) weekday, and (b) weekend (Chow and Watson, 1997).
Fig. 5. Diurnal variations of 30-min PM10 concentrations ac-
quired with tapered-element oscillating microbalances (TEOM)
in Qalabotjha, South Africa, during stagnant atmospheric
conditions.
J.C. Chow et al. / Chemosphere 49 (2002) 961–978 969
annual average and standard deviation, highest, second-
highest, and several upper percentiles of 24-h PM10
concentrations for samples taken at daily and second-,
third-, sixth-, twelfth-, and thirtieth-day intervals. US
PM2:5 standards require sampling frequencies of every
day to every sixth day, with more frequent sampling in
communities with concentrations near or in excess of the
standard.
Table 2 shows that annual averages for PM10 are
reasonably consistent with sampling frequencies up to
one in twelve days. However, the maximum and second-
highest concentrations are very sensitive to sampling
frequency, regardless of the lag between samples. The
upper percentiles also have the possibility of significant
error, especially when relatively few samples are avail-
able. The US PM2:5 and PM10 air quality standards are
expressed in terms of percentiles or single expected ex-
ceedance concentrations over three-year intervals to in-
crease the consistency of the measurement. If values for
maximum concentrations and precise peak value statis-
tics are desired, however, sampling frequencies must be
continuous and complete. Extreme values are rare events
that do not always obey standard statistical distributions
(de Nevers et al., 1977). These generalizations from
Table 2 for PM10 are consistent with those found from
other locations and time periods (Watson et al., 1981).
Distributions of environmental data can often be
summarized in terms of a simple mathematical function
with two parameters (Walpole and Meyers, 1978). The
lognormal distribution is most applicable to underlying
populations of pollutant concentrations (de Nevers et al.,
1979). The lognormal distribution describes many
naturally occurring processes such as gold ore concen-
tration in mines (Krige, 1978), geology, agriculture,
medicine, psychophysical phenomena, and even literary
research (Johnson and Kotz, 1970). Fig. 6 illustrates a
lognormal distribution for PM10 concentrations from a
South African region characterized by heavy industry,
coal mines, coal-fired power generation and residential
coal-burning smoke. The normal distribution also de-
scribes various naturally occurring processes (Albert and
Horwitz, 1996). Many such continuous distributions of
random variables can be represented by the character-
istic bell-shaped curve (Walpole and Meyers, 1978) as
illustrated in Fig. 7. The statistics for this distribution
are relatively simple to calculate and have been applied
widely to naturally occurring processes. Statistical in-
ference, which deals with the estimation of population
parameters and tests of hypotheses, is readily executed
on samples that show at least approximate Gaussian
distributions (Albert and Horwitz, 1996). Samples dif-
fering from a Gaussian distribution can be dealt with by
applying non-parametric statistics (Siegel, 1956; Ott,
1995) or transforming the data (Box et al., 1978). The
Central Limit Theorem states that the averages of ran-
dom sample sets taken from any population, regardless
of distribution, will show an approximate Gaussian
distribution in the limit as the number of sample sets
approaches infinity (Williams, 1978). Fig. 7 shows the
distribution of 100 average values from Leandra, South
Africa. From a sample average and a known standard
deviation, a confidence interval for the population mean
(l) at various levels of confidence can be calculated(Walpole and Meyers, 1978) as:
Table 2
Effects of sampling frequency on statistical indicators for daily PM10 concentrations (lg/m3) acquired with beta attenuation monitors
Sampling
frequency
Annual average
� standard deviationMin Max Second max 99th% 98th% 95th% 90th%
Every day 80� 63 6 316 309 286 246 206 170
2nd day 89� 63 10 198 286 277 251 206 182
3rd day 87� 62 6 284 263 264 245 208 170
6th day 89� 62 10 263 233 250 236 219 170
12th day 78� 60 10 233 202 227 221 202 146
30th day 101� 74 38 263 104 255 247 222 181
1998 data from Leandra, Mpumalanga, South Africa, courtesy of Eskom.
Fig. 6. Graphical representation of daily PM10 continuous
ambient measurements from Leandra, South Africa, taken for
the 1998 calendar year. The distribution closely resembles the
lognormal distribution.
970 J.C. Chow et al. / Chemosphere 49 (2002) 961–978
�xx� za=2rffiffiffi
np < l < �xxþ za=2
rffiffiffi
np ð2Þ
where �xx is the sample average for n observations, r2 isthe population variance, and za=2 is the value of thestandard deviation leaving an a=2 area at the tail ends ofthe normal distribution.
In the case of the Leandra samples, and for a known
PM10 population average and standard deviation of
80� 63 lg/m3, the 95% confidence interval is calculatedfor a sample size of 30 as between 57 and 103 lg/m3. Ifthe population standard deviation is not known, it can
be approximated by the sample standard deviation. A
longer sampling interval is required to estimate the pop-
ulation mean with a higher confidence interval (Walpole
and Meyers, 1978). The ideal situation would be to
sample everyday so there would be no uncertainty about
the population mean.
The confidence interval also provides an estimate of
the precision of the sample mean (Walpole and Meyers,
1978), i.e., the statistical uncertainty or sample error
which is generally expressed as the difference between
the population and the sample means. This is estimated
based on the population standard deviation and the
sample size:
e ¼ za=2rffiffiffi
np ð3Þ
Table 3 presents the errors for various sample sizes,
given the known standard deviation of 63 lg/m3. Thisemphasizes the importance of estimating the population
variance from a reasonably large (n > 30) randomsample before setting the sample size for a full-scale
study.
Table 4 compares the statistics of samples collected
on certain days of the week and every sixth day with a
random sample selected for its similarity to the popu-
lation (Walpole and Meyers, 1978). The sample means
are tested for their similarity to the random sample ac-
cording to the null hypothesis. All calculated t-statistics
fall within the tabulated critical range of �1.98. Thetabulated critical value at a 0.05 level of significance
(Walpole and Meyers, 1978) implies that sampling on
any weekday is acceptable. However, the every-sixth-day
sampling schedule reports a t-statistic close to zero and
shows the greatest similarity to the random sample and
is the preferred sampling frequency. This is due to the
fact that every-sixth-day sampling gives an equal num-
ber of samples on each day of the week. The Sunday
sample varies most from the random sample and sub-
sequently introduces the greatest bias. This is explained
by the fact that a large proportion of the Leandra PM
originates from residential coal combustion in neigh-
boring townships, and that cooking activities on Sun-
days differ from those on normal working days.
7. Designing an outdoor exposure network
Watson et al. (1997) outline the following steps that
can be followed when designing an outdoor exposure
network:
1. Identify political boundaries of populated areas: Plot
populated entities (Census Bureau Metropolitan Sta-
tistical Areas, counties, zip code areas, census tracts,
or census blocks). Identify where the majority of the
Table 3
The variation of random error with sample size for the 1998 data from Leandra, South Africa
Sample size (n)
365 182 122 61 30 12 6
Sampling frequency, days
(approximate)
1 2 3 6 12 30 60
Error (lg/m3) 6 9 11 16 23 36 50
Interval (lg/m3) �xxa� 6 �xx� 9 �xx� 11 �xx� 16 �xx� 23 �xx� 36 �xx� 50a�xx represents the population mean. At 95% interval, za=2 ¼ 1:96 given the known standard deviation of 63 lg/m3 in Eq. (3).
Fig. 7. Graphical representation of 100 random sample means
of 30 observations each, from Leandra, South Africa, calcu-
lated from continuous daily measurements for the 1998 calen-
dar year. The distribution is approximately Gaussian.
J.C. Chow et al. / Chemosphere 49 (2002) 961–978 971
people live. Identify a grouping of populated entities
that define a contiguous area and designate this as the
planning area.
2. Identify natural air basins: Compare outer boundaries
of the study area on a topographic map showing ter-
rain that might engender trapping, channeling, or
separation of source emissions from populated areas.
When terrain features are near the planning area
boundary, add or subtract population entities to cor-
respond as closely as possible to the terrain features.
When terrain features are significant within the plan-
ning area boundary, identify potential community
monitoring zones that are separated by ridges, lakes,
or valleys, or that are bounded on one edge by a sea-
coast.
3. Locate existing air quality monitoring sites: Plot the
locations of existing monitoring sites, if any, within
the planning area. Examine the extent to which these
correspond to populated areas. Identify large dis-
tances between existing sites, and identify sites that
appear to represent the same sizes of populated areas.
Evaluate the justification for excluding existing sites
outside of the initial planning area boundaries. If
these are community-oriented sites, extend the initial
monitoring boundaries with populated entities to in-
clude these sites. Alternatively, evaluate these sites for
potential as special monitoring, transport, or back-
ground sites.
4. Locate emissions sources and population: Plot major
land use within the populated entities within the cat-
egories of commercial, residential, industrial, or agri-
cultural. Plot the major roadways. Plot emissions
from major point sources for primary PM, sulfur di-
oxide, and oxides of nitrogen. Use a gridded emis-
sions inventory or maps of source type and density,
if available. Each monitoring site in a monitoring
zone should be affected by similar emission sources.
Determine which populated areas coincide with or
are in close proximity to areas of high source density
and which are in areas of low source density. When
evaluating community exposures to emissions, con-
sider populations at work and leisure activities, as
well as at home. Population density is important both
for determining exposure and for estimating emis-
sions from vehicles, cooking, wood-burning, etc.
5. Identify meteorological patterns: Plot wind directions
and speeds, vertical temperature structure, and fre-
quencies of fogs by season. Determine how these vary
within and around the initial area to be represented
by monitoring. Extend the dimensions of community
monitoring zones that include large source emissions
in the downwind direction, using terrain as a guide
for potential channeling.
6. Compare PM concentrations: Estimate the spatial
uniformity of average and maximum concentrations
from previous measurements or model calculations
within the potential community monitoring zones
for annual, seasonal, and maximum PM concentra-
tions. Combine zones surrounding neighborhood-
scale monitors for which these concentrations are
similar.
7. Locate sites:Where existing sites are within each mon-
itoring zone, give them first priority of PM2:5 monitor-
ing when they meet siting criteria. Where monitoring
zones do not contain existing sites, select new sites.
8. Particulate monitor siting criteria
There are many local sources of suspended particles,
and sampling sites may be affected by these sources if
they are too close. Internal requirements are those for
Table 4
Introduction of bias into results by selecting specific weekdays on which to sample
Daily average
PM10 (lg/m3)Popu-
lation
53 Randomly
selected days
Days of the week Every
6th daySun Mon Tue Wed Thu Fri Sat
Mean 80 79 66 83 93 82 80 84 75 78
Std dev. 63 63 47 65 62 66 77 61 63 67
Min 6 7 9 13 23 7 9 9 6 7
Max 316 233 202 316 248 286 309 238 263 263
90th percentile 170 180 124 169 202 153 191 168 167 176
95th percentile 206 201 139 189 210 232 261 192 209 212
98th percentile 246 215 192 217 227 246 299 200 222 233
99th percentile 286 224 197 267 238 266 304 219 242 247
Sample pair Ran53-
Ran53
Ran53-
Sun
Ran53-
Mon
Ran53-
Tue
Ran53-
Wed
Ran53-
Thu
Ran53-
Fri
Ran53-
Sat
Ran53-
6th Day
t-statistic 0 1.1446 �0.3308 �1.0391 �0.1825 �0.0540 �0.3623 0.3100 0.0515
Weekday means are compared to that of a random sample of 53 days, at a 0.05 level of significance. Samples from a daily PM10 data set
(lg/m3) acquired with beta attenuation monitors. 1998 data from Leandra, Mpumalanga, South Africa, courtesy of Eskom.
972 J.C. Chow et al. / Chemosphere 49 (2002) 961–978
operating the needed instruments, while external criteria
address site surroundings to achieve specific monitoring
purposes. Internal criteria include:
• Long-term site commitment: Sites are meant to mea-
sure trends as well as compliance, and a long-term
commitment from the property owner for continued
monitoring is required. Public buildings such as
schools, fire stations, police stations, recreation halls,
and hospitals often have more stability and a motive
for public service than do private or commercial
buildings.
• Sufficient operating space: A large, flat space, elevated
at least 1 m but no more than 14 m above ground
level, is needed to place monitors and monitoring
probes. The space available for samplers should be
at least 5 m distant and upwind (most common wind
direction) from building exhausts and intakes and at
least 2 m from other monitoring instruments� inletsand exhausts, as well as away from walls, parapets,
or penthouses that might influence air flow. Buildings
housing large emitters, such as coal-, waste-, or oil-
burning boilers, furnaces or incinerators, should be
avoided.
• Access and security: Access to the sampling platform
should be controlled by fencing or elevation above
ground level. Sampler inlets should be sufficiently dis-
tant (>10 m) from public access to preclude purpose-ful contamination from reaching them in sufficient
quantities to bias samples. Access should be con-
trolled by a locked door, gate, or ladder with docu-
mentation of site visitations and the purposes of
those visits.
• Safety: Wiring, access steps, sampler spacing, and
platform railings should comply with all relevant
codes and workplace regulations, as well as common
sense, to minimize potential for injury to personnel or
equipment.
• Power: Power should be sufficient for the samplers to
be operated on a long-term basis, as well as for spe-
cial study and audit samplers to be located at a site.
Where possible, a separate circuit breaker should be
provided for each instrument to prevent an electrical
malfunction in one monitor from shutting off power
to the other monitors at the site.
• Environmental control: Environments surrounding
monitoring instruments should be maintained within
the manufacturers specifications for proper instru-
ment function. Most filter-based PM samplers are
designed to operate under a wide range of environ-
mental conditions and can be located outdoors in
most types of weather. Several continuous moni-
toring methods may require environmental shelters
with temperature and humidity controls to protect
their electronic sensing and data acquisition mecha-
nisms.
These criteria may be tightened or relaxed for special-
purpose, transport, and background monitors. For ex-
ample, battery-powered portable monitors (Baldauf
et al., 2001) may be located on utility poles at various
elevations to assess zones of influence and zones of
representation for sources and receptors (Chow et al.,
1999, 2002b).
External siting criteria refer to the environs sur-
rounding a measurement location, and these differ de-
pending on the zone of representation intended for a
specific monitoring site:
• Exposure: Large nearby buildings and trees extend-
ing above the height of the monitor may present bar-
riers or deposition surfaces for PM. Certain trees
may also be sources of PM in the form of detritus,
pollen, or insect parts. Interferences can be mini-
mized by locating samplers > 20 m from nearby treesand twice the difference in elevation from nearby
buildings or other obstacles.
• Distance from nearby emitters: The monitor should
be outside the zone of influence of sources located
within the designated zone of representation for the
monitoring site. Neighborhood and urban zones of
representation are needed for community-oriented
compliance monitors. These should generally be at
least 1 km from very large, visibly identifiable source
areas occupied by major industries such as cement
and steel production or ore processing. A minimum
distance of �50 m from busy paved highways is usu-ally outside the road�s immediate zone of influencefor a rooftop monitor. These siting criteria were
established for PM10 monitoring siting (US Environ-
mental Protection Agency, 1987), and they have pro-
ven their validity in PM10 network design. For larger
than middle-scale monitoring, no unpaved roads with
significant traffic or residential wood-burning appli-
ances should be located within 100 m of the monitor-
ing location. Background monitoring sites should be
located >100 km from large population centers, and>100 m from roads and wood-burning (burning is
common, though often intermittent, in camping, for-
ested, and agricultural areas).
• Proximity to other measurements: Other air quality
and meteorological measurements can aid in the in-
terpretation of high PM levels, and with all other
considerations being equal, PM2:5 sites should give
preference to existing sites that make other measure-
ments. For example, high local wind gusts may ex-
plain high PM readings as caused by wind blown
dust. These gusts are often localized, and would not
be detected on a more distant monitor. Similarly, a
strong correspondence between hourly carbon mon-
oxide and PM readings would indicate that locally
emitted vehicle exhaust is a large contributor at that
site. This conclusion would be more tenuous if the
J.C. Chow et al. / Chemosphere 49 (2002) 961–978 973
carbon monoxide measurements were not collocated.
In particular, collocating PM2:5 and PM10 monitors
will provide information on both fine and coarse
fractions of suspended particles.
9. Evaluating zones of representation
A site originally selected to represent community
exposure (generally on a neighborhood or urban scale)
may have its zone of representation change owing to
long-term changes in land use or short-term events that
affect that particular site. The following approach can be
taken to evaluate individual site representation:
• Annual site surveys: The land use and sources around
a monitoring site may change from year to year, espe-
cially in high-growth areas. Maps should be updated
as part of the annual measurement network sum-
mary, and the setbacks from emissions sources and
obstructions should be re-evaluated to ascertain that
they are still met.
• Records of intermittent events: High PM2:5 or PM10
concentrations may have corresponded to a specific
event, such as construction or a fire, occurring near
the measurement location. Visual events should be
recorded as part of the measurement network main-
tenance, and these should be summarized at the end
of each year for inclusion in the annual network eval-
uation report.
• Satellite monitoring studies: To evaluate the zones of
representation of sites, many monitors may be lo-
cated at different distances around and between
monitors in a monitoring zone. Spatial uniformity
measures can be determined for these temporary lo-
cations and compared to those from the sites within
the monitoring zone to evaluate how well the long-
term sites represent population exposures to PM.
10. Conclusions
Designing an effective PM sampling network to rep-
resent outdoor human exposure requires more than
simply selecting measurement locations. For a given
level of resources, a balance is needed between the
number of sites, the sampling period to be represented,
sample duration and frequency, and the variables mea-
sured. These are different for different network objec-
tives. Sampling designs that determine compliance with
standards are the most common air quality networks,
and relatively consistent guidance is available for these
networks. These criteria are not sufficient for other
monitoring objectives, and more specific guidance is
needed for each of these other objectives.
Acknowledgement
This paper was supported in part by EPA. However,
it has not undergone technical review by EPA. There-
fore, the views are those of the authors and do not
represent EPA policy. Mention of trade names or
commercial products does not constitute EPA endorse-
ment or recommendation.
References
Albert, R., Horwitz, W., 1996. Coping with sampling variability
in biota: Percentiles and other strategies. In: Keith, L.H.
(Ed.), Principles of Environmental Sampling, 2nd ed.
American Chemical Society, Washington, DC, pp. 653–
669.
Allen, G.A., Sioutas, C., Koutrakis, P., Reiss, R., Lurmann,
F.W., Roberts, P.T., 1997. Evaluation of the TEOM�
method for measurement of ambient particulate mass in
urban areas. J. Air Waste Mgmt. Assoc. 47 (6), 682–689.
Andricevic, R., 1990. Cost-effective network design for ground-
water flow monitoring. Stoch. Hydrol. Hydraul. 4 (1), 27–
41.
Babich, P., Wang, P.Y., Allen, G., Sioutas, C., Koutrakis, P.,
2000. Development and evaluation of a continuous ambient
PM2:5 mass monitor. Aerosol Sci. Technol. 32 (4), 309–324.
Baldauf, R.W., Lane, D.D., Marotz, G.A., Wiener, R.W., 2001.
Performance evaluation of the portable Minivol particulate
matter sampler. Atmos. Environ. 35, 6087–6091.
Baron, P.A., Willeke, K. (Eds.), 2001. Aerosol Measurement:
Principles, Techniques and Applications, 2nd ed. John
Wiley & Sons, New York, NY.
Bettinelli, M., Baroni, U., Pastorelli, N., 1989. Microwave oven
sample dissolution for the analysis of environmental and
biological materials. Anal. Chim. Acta 225, 159–174.
Birch, M.E., Cary, R.A., 1996. Elemental carbon-based method
for monitoring occupational exposures to particulate diesel
exhaust. Aerosol Sci. Technol. 25 (3), 221–241.
Bond, T.C., Charlson, R.J., Heintzenberg, J., 1998. Quantifying
the emission of light-absorbing particles: Measurements
tailored to climate studies. Geophys. Res. Lett. 25 (3), 337–
340.
Borgman, L.E., Gerow, K., Flatman, G.T., 1996. Cost-effective
sampling for spatially distributed phenomena. In: Keith,
L.H. (Ed.), Principles of Environmental Sampling, 2nd ed.
American Chemical Society, Washington, DC, pp. 753–
778.
Bowne, N.E., Londergan, R.J., 1983. Overview, results and
conclusions for the EPRI plume model validation and
development project: Plains site. Report No. EA-3074
PROJECT 1616. Electric Power and Research Institute,
Palo Alto, CA.
Box, G.E.P., Hunter, W.G., Hunter, J.S., 1978. Statistics for
Experimenters: An Introduction to Design, Data Analysis,
and Model Building. John Wiley & Sons, New York, NY.
Cachier, H., Bremond, M.P., Buat-M�eenard, P., 1989. Determi-
nation of atmospheric soot carbon with a simple thermal
method. Tellus 41B, 379–390.
Camisani-Calzolari, F.A.G.M., 1984. Geostatistical appraisal
of a tabular uranium deposit in South Africa. In: Verly, G.,
974 J.C. Chow et al. / Chemosphere 49 (2002) 961–978
David, M., Journel, A.G., Marechal, A. (Eds.), Geostatis-
tics for Natural Resources Characterization, Part 2. D.
Reidel Publishing Co., Bordrecht, pp. 935–949.
Chao, C.Y.H., Tung, T.C., 2001. An empirical model for
outdoor contaminant transmission into residential buildings
and experimental verification. Atmos. Environ. 35 (9),
1585–1596.
Chin, C.-J., Wang, C.-F., Jeng, S.-L., 1999. Multi-element
analysis of airborne particulate matter collected on PTFE-
membrane filters by laser ablation inductively coupled
plasma mass spectrometry. J. Anal. Atom. Spectrom. 14,
663–668.
Chow, J.C., Liu, C.S., Cassmassi, J.C., Watson, J.G., Lu, Z.,
Pritchett, L.C., 1992. A neighborhood-scale study of PM10
source contributions in Rubidoux, California. Atmos.
Environ. 26 (4), 693–706.
Chow, J.C., Watson, J.G., Pritchett, L.C., Pierson, W.R.,
Frazier, C.A., Purcell, R.G., 1993. The DRI Thermal/
Optical Reflectance carbon analysis system: Description,
evaluation and applications in US air quality studies.
Atmos. Environ. 27A (8), 1185–1201.
Chow, J.C., 1995. Critical review: Measurement methods to
determine compliance with ambient air quality standards
for suspended particles. J. Air Waste Mgmt. Assoc. 45 (5),
320–382.
Chow, J.C., Watson, J.G., Lu, Z., Lowenthal, D.H., Frazier,
C.A., Solomon, P.A., Thuillier, R.H., Magliano, K.L.,
1996. Descriptive analysis of PM2:5 and PM10 at regionally
representative locations during SJVAQS/AUSPEX. Atmos.
Environ. 30 (12), 2079–2112.
Chow, J.C., Watson, J.G., 1997. Fugitive dust and other source
contributions to PM10 in Nevada�s Las Vegas Valley.Report No. 4039.1F. Prepared for Clark County Depart-
ment of Comprehensive Planning, Las Vegas, NV, by
Desert Research Institute, Reno, NV.
Chow, J.C., Watson, J.G., 1999. Ion chromatography in
elemental analysis of airborne particles. In: Landsberger,
S., Creatchman, M. (Eds.), Elemental Analysis of Airborne
Particles, vol. 1. Gordon and Breach Science, Amsterdam,
pp. 97–137.
Chow, J.C., Watson, J.G., Green, M.C., Lowenthal, D.H.,
DuBois, D.W., Kohl, S.D., Egami, R.T., Gillies, J.A.,
Rogers, C.F., Frazier, C.A., Cates, W., 1999. Middle- and
neighborhood-scale variations of PM10 source contributions
in Las Vegas, Nevada. J. Air Waste Mgmt. Assoc. 49 (6),
641–654.
Chow, J.C., Watson, J.G., Crow, D., Lowenthal, D.H.,
Merrifield, T., 2001. Comparison of IMPROVE and
NIOSH carbon measurements. Aerosol Sci. Technol. 34
(1), 23–34.
Chow, J.C., Watson, J.G., 2001. Zones of representation for
PM10 measurements along the US/Mexico border. Sci. Total
Environ. 276 (1–3), 49–68.
Chow, J.C., Watson, J.G., Edgerton, S.A., Vega, E., 2002a.
Chemical composition of PM10 and PM2:5 in Mexico City
during winter 1997. Sci. Total Environ. 287 (3), 177–
201.
Chow, J.C., Watson, J.G., Edgerton, S.A., Vega, E., 2002b.
Spatial differences in outdoor PM10 mass and aerosol
composition in Mexico City. J. Air Waste Mgmt. Assoc.
52 (4), 423–434.
Chow, J.C., Watson, J.G., Hackett, E.I., Stone, R.H., Hins-
vark, B.A., 2002c. Method No. 822: General atomic absorp-
tion procedure for trace metals in airborne material
collected on filters. In: Lodge, J.P. (Ed.). Methods of Air
Sampling and Analysis, 4th ed. Air and Waste Management
Association, Pittsburgh, PA, accepted for publication.
Cohen, B.S., McCammon, C.S.J. (Eds.), 2001. Air Sampling
Instruments for Evaluation of Atmospheric Contaminants,
9th ed. American Conference of Governmental Industrial
Hygienists, Cincinnati, OH.
Cooper, D.W., 1975. Statistical errors in beta absorption
measurements of particulate mass concentration. J. Air
Pollut. Control Assoc. 25, 1154–1155.
Cooper, D.W., 1976. Significant relationships concerning
exponential transmission or penetration. J. Air Pollut.
Control Assoc. 26, 366–367.
Courtney, W.J., Shaw, R.W., Dzubay, T.G., 1982. Precision
and accuracy of a beta-gauge for aerosol mass determina-
tion. Environ. Sci. Technol. 16 (4), 236–239.
Darby, W.P., Ossenbruggen, P.J., Gregory, C., 1974. Placement
of samplers in an air monitoring network. In: Proceedings––
Institute of Environmental Sciences, 1974.
de Marsily, G., Lavedan, G., Boucher, M., Fasanimo, G., 1984.
Interpretation of interference tests in a well field using
geostatistical techniques to fit the permeability distribution
in a reservoir model. In: Verly, G. (Ed.), Geostatistics for
Natural Resources Characterization. D. Reidel Publishing,
Bordrecht, pp. 831–849.
de Nevers, N., Lee, K.W., Frank, N.H., 1977. Extreme values
in TSP distribution functions. J. Air Pollut. Control Assoc.
27, 995–1000.
de Nevers, N., Lee, K.W., Frank, N.H., 1979. Patterns in TSP
distribution functions. J. Air Pollut. Control Assoc. 29 (1),
32–37.
Eldred, R.A., Cahill, T.A., Flocchini, R.G., 1997. Composition
of PM2:5 and PM10 aerosols in the IMPROVE Network.
J. Air Waste Mgmt. Assoc. 47 (2), 194–203.
Elsom, D.M., 1978. Spatial correlation analysis of air pollution
data in an urban area. Atmos. Environ. 12, 1103–1107.
Engelbrecht, J.P., Swanepoel, L., 1998. Aerosol monitoring and
source apportionment in Qalabotjha, Free State, South
Africa. Report MC5. Mintek, Randburg, South Africa.
Engelbrecht, J.P., Reddy, V.S., Swanepoel, L., Mostert, J.C.,
1998. Aerosol monitoring and source apportionment in the
Vaal Triangle. Special publication No. 17. Mintek, Rand-
burg, South Africa.
Fung, K.K., 1990. Particulate carbon speciation by MnO2oxidation. Aerosol Sci. Technol. 12 (1), 122–127.
Gandin, L.S., 1970. The planning of meteorological station
networks. Report No. 111. World Meteorological Organi-
zation, Geneva, Switzerland.
Gillies, J.A., Watson, J.G., Rogers, C.F., DuBois, D.W., Chow,
J.C., Langston, R., Sweet, J., 1999. Long term efficiencies of
dust suppressants to reduce PM10 emissions from unpaved
roads. J. Air Waste Mgmt. Assoc. 49 (1), 3–16.
Goldstein, I.F., Landovitz, L., Block, G., 1974. Air pollution
patterns in New York City. J. Air Pollut. Control Assoc.
24, 148–152.
Goldstein, I.F., Landovitz, L., 1977a. Analysis of air pollution
patterns in New York City––I. Can one station represent
the large metropolitan area? Atmos. Environ. 11, 47–52.
J.C. Chow et al. / Chemosphere 49 (2002) 961–978 975
Goldstein, I.F., Landovitz, L., 1977b. Analysis of air pollution
patterns in New York City––II. Can one aerometric station
represent the area surrounding it? Atmos. Environ. 11, 53–
57.
Gordon, T., Nadziejko, C., Schlesinger, R.B., Chen, L.C., 1998.
Pulmonary and cardiovascular effects of acute exposure to
concentrated ambient particulate matter in rats. Toxicol.
Lett. 96–97 (1–3), 285–288.
Green, A.E., Singhal, R.P., Venkateswar, R., 1980. Analytic
extensions of the Gaussian plume model. J. Air Pollut.
Control Assoc. 30 (7), 773–776.
Handscombe, C.M., Elsom, D.M., 1982. Rationalisation of the
national survey of air pollution monitoring network of the
United Kingdom using spatial correlation analysis: A case-
study of the greater London area. Atmos. Environ. 16 (5),
1061–1070.
Hansen, A.D.A., Rosen, H., Novakov, T., 1984. The aethalo-
meter––An instrument for the real-time measurement of
optical absorption by aerosol particles. Sci. Total Environ.
36, 191–196.
Hansen, A.D.A., Bodhaine, B.A., Dutton, E.G., Schnell, R.C.,
1988. Aerosol black carbon measurements at the South
Pole: Initial results, 1986–1987. Geophys. Res. Lett. 15 (11),
1193–1196.
Hansen, A.D.A., Conway, T.J., Steele, L.P., Bodhaine, B.A.,
Thoning, K.W., Tans, P., Novakov, T., 1989. Correlations
among combustion effluent species at Barrow, Alaska:
Aerosol black carbon, carbon dioxide, and methane. J.
Atmos. Chem. 9 (1–3), 283–299.
Hansen, A.D.A., Rosen, H., 1990. Individual measurements of
the emission factor of aerosol black carbon in automobile
plumes. J. Air Waste Mgmt. Assoc. 40 (12), 1654–1657.
Harman, J.N., 1989. Inductively coupled plasma emission
spectroscopy. In: Lodge, J.P. (Ed.), Methods of Air
Sampling and Analysis, 3rd ed. Lewis Publishers, Chelsea,
MI, pp. 88–92.
Holmes, J., Samberg, T., McInnes, L., Zieman, J., Zoller, W.H.,
Harris, J., 1997. Long-term aerosol and trace acidic gas
collection at Mauna Loa Observatory 1979–1991. J. Geo-
phys. Res. 102 (D15), 19007–19019.
Holton, J.R., 1982. An Introduction to Dynamic Meteorology,
3rd ed. Academic Press, New York, NY.
Hougland, E.S., 1977. Air pollutant monitor network design
using mathematical programming. Ph.D. Dissertation, Vir-
ginia Polytechnic Institute and State University.
Husar, R.B., 1974. Atmospheric particulate mass monitor-
ing with a b radiation detector. Atmos. Environ. 8, 183–188.
Jaklevic, J.M., Gatti, R.C., Goulding, F.S., Loo, B.W., 1981. A
beta-gauge method applied to aerosol samples. Environ. Sci.
Technol. 15 (6), 680–686.
Johnson, N.L., Kotz, S., 1970. Continuous Univariate Distri-
butions––1. John Wiley & Sons, New York, NY.
Journel, A.G., 1980. The lognormal approach to predicting
local distributions of selective mining unit grades. J. Math.
Geol. 12 (4), 285–303.
Kassim, A.H.M., Kottegoda, N.T., 1991. Rainfall network
design through comparative kriging methods. Hydrolog.
Sci. J. 36 (3), 223–240.
Kidwell, C.B., Ondov, J.M., 2001. Development and evaluation
of a prototype system for collection of sub-hourly ambient
aerosol for chemical analysis. Aerosol Sci. Technol. 35 (1),
596–601.
Klein, F., Ranty, C., Sowa, L., 1984. New examinations of the
validity of the principle of beta radiation absorption for
determinations of ambient air dust concentrations. J.
Aerosol Sci. 15, 391–394.
Koponen, I.K., Asmi, A., Keronen, P., Puhto, K., Kulmala,
M., 2001. Indoor air measurement campaign in Helsinki,
Finland 1999––The effect of outdoor air pollution on indoor
air. Atmos. Environ. 35 (8), 1465–1477.
Kotchmar, D.J., McMullen, T.B., Hasselblad, V., 1987. Ade-
quacy of a single monitoring site for defining mean outdoor
concentrations of fine particles in a demarcated residential
community. J. Air Pollut. Control Assoc. 37 (4), 377–381.
Krige, D.G., 1978. Lognormal-de Wijsian geostatistics for ore
evaluation. South African Institute of Mining and Metal-
lurgy, South Africa.
Lefohn, A.S., Knudsen, H.P., Logan, J.A., Simpson, J.,
Bhumralkar, C.M., 1987. An evaluation of the Kriging
method to predict 7-h seasonal mean ozone concentrations
for estimating crop losses. J. Air Pollut. Control Assoc. 37
(5), 595–602.
Lilienfeld, P., Dulchinos, J., 1972. Portable instantaneous mass
monitoring for coal mine dust. J. Am. Ind. Hyg. Assoc. 33,
136–145.
Lilienfeld, P., 1975. Design and operation of dust measuring
instrumentation based on the beta-radiation method. Staub-
Reinhaltl. Luft 35, 458.
Liu, D.L., Nazaroff, W.W., 2001. Modeling pollutant penetra-
tion across building envelopes. Atmos. Environ. 35 (26),
4451–4462.
Macias, E.S., Husar, R.B., 1976. Atmospheric particulate mass
measurement with beta attenuation mass monitor. Environ.
Sci. Technol. 10, 904–907.
Mazzera, D.M., Lowenthal, D.H., Chow, J.C., Watson, J.G.,
Grubisic, V., 2001. PM10 measurements at McMurdo
Station, Antarctica. Atmos. Environ. 35 (10), 1891–
1902.
McMurry, P.H., 2000. A review of atmospheric aerosol mea-
surements. Atmos. Environ. 34 (12–14), 1959–1999.
Meyer, M.B., Lijek, J., Ono, D.M., 1992. Continuous PM10
measurements in a woodsmoke environment. In: Chow,
J.C., Ono, D.M. (Eds.), Transactions: PM10 Standards and
Nontraditional Particulate Source Controls. Air & Waste
Mgmt. Association, Pittsburgh, PA, pp. 24–38.
Meyer, P.D., Valocchi, A.J., Eheart, J.W., 1994. Monitoring
network design to provide initial detection of groundwater
contamination. Water Resour. Res. 30 (9), 2647–2659.
Moschandreas, D.J., Watson, J.G., D�Abreton, P., Scire, J.S.,Zhu, T., Klein, W., Saksena, S., Wilson, W.E., 2002.
Methodology of exposure modeling. Chemosphere, this
issue.
Munn, R.E., 1975. Suspended particulate concentrations:
Spatial correlations in the Detroit-Windsor area. Tellus
XXVII (4), 397–405.
Munn, R.E., 1981. The estimation and interpretation of air
quality trends, including some implications for network
design. Environ. Monit. Assess. 1, 49–58.
Nader, J.S., Allen, R., 1960. A mass loading and radioactivity
analyser for atmospheric particulates. J. Am. Ind. Hyg.
Assoc. 21 (August), 300–307.
976 J.C. Chow et al. / Chemosphere 49 (2002) 961–978
Nesbitt, K.J., Carter, K.R., 1996. Integration of immunoassay
field analytical techniques into sampling plans. In:
Keith, L.H. (Ed.), Principles of Environmental Sampling,
2nd ed. American Chemical Society, Washington, DC,
pp. 727–735.
Noll, K.E., Miller, T.L., 1977. Air Monitoring Survey Design.
Ann Arbor Science, Ann Arbor, MI.
Ondov, J.M., Kidwell, C.B., 2002a. Sub-hourly analysis of
elements in ambient aerosol by atomic spectroscopy after
dynamic preconcentration. Presented at the PM2000:
Particulate Matter and Health––The Scientific Basis for
Regulatory Decision Making Conference, Charleston, SC
(1-24-2000). Air and Waste Management Association,
Pittsburgh, PA.
Ondov, J.M., Kidwell, C.B., 2002b. Improved system for sub-
hourly measurement of elemental constituents of ambient
aerosol particles. Presented at the A&WMA Measurement
of Toxic and Related Air Pollutants Conference, Research
Triangle Park, NC (9-12-2000). Air and Waste Management
Association, Pittsburgh, PA.
Ott, W.R., 1995. Environmental Statistics and Data Analysis.
Lewis Publishers, Boca Raton, FL.
Patashnick, H., Hemenway, C.L., 1969. Oscillating fiber
microbalance. Rev. Sci. Instrum. 40 (8), 1008–1011.
Patashnick, H., 1987. On-line, real-time instrumentation for
diesel particulate testing. Diesel Prog. N. Am. 53, 43–44.
Patashnick, H., Rupprecht, E.G., 1991. Continuous PM10
measurements using the tapered element oscillating micro-
balance. J. Air Waste Mgmt. Assoc. 41 (8), 1079–1083.
Peterson, J.T., 1970. Distribution of sulfur dioxide over
metropolitan St. Louis as described by empirical orthogonal
functions, and its relation to meteorological parameters.
Atmos. Environ. 4, 501–518.
Polissar, A.V., Hopke, P.K., Paatero, P., Kaufmann, Y.J., Hall,
D.K., Bodhaine, B.A., Dutton, E.G., Harris, J.M., 1999.
The aerosol at Barrow, Alaska: Long-term trends and
source locations. Atmos. Environ. 33 (16), 2441–2458.
Quinn, P.K., Coffman, D.J., Kapustin, V.N., Bates, T.S.,
Covert, D.S., 1998. Aerosol optical properties in the marine
boundary layer during the First Aerosol Characterization
Experiment (ACE 1) and the underlying chemical and
physical aerosol properties. J. Geophys. Res. 103 (D13),
16547–16564.
Rouhani, S., Ebrahimpour, M.R., Yaqub, I., Gianella, E.,
1992a. Multivariate geostatistical trend detection and net-
work evaluation of space–time acid deposition data––I.
Methodology. Atmos. Environ. 26A (14), 2603–2614.
Rouhani, S., Ebrahimpour, M.R., Yaqub, I., Gianella, E.,
1992b. Multivariate geostatistical trend detection and net-
work evaluation of space–time acid deposition data––II.
Application to NADP/NTN data. Atmos. Environ. 26A
(14), 2615–2626.
Rupprecht, E.G., Meyer, M.B., Patashnick, H., 1992. The
Tapered Element Oscillating Microbalance as a tool for
measuring ambient particulate concentrations in real time.
J. Aerosol Sci. 23, S635–S638.
Rupprecht, E.G., Patashnick, H., Beeson, D.E., Green, R.E.,
Meyer, M.B., 1995. A new automated monitor for the
measurement of particulate carbon in the atmosphere. In:
Cooper, J.A., Grant, L.D. (Eds.), Proceedings, Particulate
Matter: Health and Regulatory Issues. Air and Waste
Management Association, Pittsburgh, PA, pp. 262–
267.
Russo, D., 1984. Design of an optimal sampling network for
estimating the variogram. Soil Sci. Soc. Am. J. 48, 708–716.
Sabaton, C., 1976. Etude d�optimisisation d�un reseau de
surveillance de la pollution atmospherique dans la region
Parissienne. In: Benarie, M.M. (Ed.), Atmospheric Pollu-
tion. Elsevier, Amsterdam, pp. 159–171.
Samet, J.M., Spengler, J.D. (Eds.), 1991. Indoor Air Pollu-
tion––A Health Perspective. Johns Hopkins University
Press, Baltimore, MD.
Schmid, H.P., Laskus, L., Abraham, H.J., Baltensperger, U.,
Lavanchy, V.M.H., Bizjak, M., Burba, P., Cachier, H.,
Crow, D.J., Chow, J.C., Gnauk, T., Even, A., ten Brink,
H.M., Giesen, K.P., Hitzenberger, R., et al., 2001. Results
of the ‘‘Carbon Conference’’ international aerosol carbon
round robin test: Stage 1. Atmos. Environ. 35 (12), 2111–
2121.
Seigneur, C., Pai, P., Lourensz, R.S., Hopke, P.K., Grosjean,
D., 1997. Review of air quality models for particulate
matter. Report No. CP015-97-1a. Prepared for American
Petroleum Institute, Washington, DC, by Atmospheric &
Environmental Research, Inc., San Ramon, CA.
Seigneur, C., 2001. Current status of air quality models for
particulate matter. J. Air Waste Mgmt. Assoc. 51 (11),
1508–1521.
Sem, G.J., Borgos, J.W., 1975. Experimental investigation of
the exponential attenuation of beta radiation for dust
measurements. Staub-Reinhaltl. Luft 35 (1), 5–9.
Siegel, S., 1956. Nonparametric Statistics for the Behavioral
Sciences. McGraw-Hill, New York, NY.
Speer, R.E., Barnes, H.M., Brown, R., 1997. An instrument for
measuring the liquid water content of aerosols. Aerosol Sci.
Technol. 27 (1), 50–61.
Spurny, K.R., Kubie, G., 1961. Analytische methoden zur
bestimmung von aerosolen unter verwendung von
membranultrafiltern: v. herstellung und anwendung von
mit radioaktivem 63nickel-isotop markierten membranultra-
filtern. Coll. Czechoslovak Chem. Commun. 26, 1991–1998.
Stalker, W.W., Dickerson, R.C., 1962a. Sampling station and
time requirements for urban air pollution surveys Part II:
Suspended particulate matter and soiling index. J. Air
Pollut. Control Assoc. 12 (3), 111–128.
Stalker, W.W., Dickerson, R.C., 1962b. Sampling station and
time requirements for urban air pollution surveys Part III:
Two- and four-hour soiling index. J. Air Pollut. Control
Assoc. 12 (4), 170–178.
Stalker, W.W., Dickerson, R.C., Kramer, G.D., 1962. Sampling
station and time requirements for urban air pollution
surveys Part IV: 2- and 24-hour sulfur dioxide and summary
of other pollutants. J. Air Pollut. Control Assoc. 12 (8),
361–375.
Stolzenburg, M.R., Hering, S.V., 2000. Method for the auto-
mated measurement of fine particle nitrate in the atmo-
sphere. Environ. Sci. Technol. 34 (5), 907–914.
Stull, R.B., 1988. An Introduction to Boundary Layer Mete-
orology. Kluwer Academic Publishers, Dordrecht.
Tanaka, S., Yasushi, N., Sato, N., Fukasawa, T., Santosa, S.J.,
Yamanaka, K., Ootoshi, T., 1998. Rapid and simultaneous
multi-element analysis of atmospheric particulate matter
using inductively coupled plasma mass spectrometry with
J.C. Chow et al. / Chemosphere 49 (2002) 961–978 977
laser ablation sample introduction. J. Anal. Atom. Spec-
trom. 13, 135–140.
Trujillo-Ventura, A., Ellis, J.H., 1991. Multiobjective air
pollution monitoring network design. Atmos. Environ.
25A (2), 469–479.
Tsai, C.J., Cheng, Y.H., 1996. Comparison of two ambient beta
gauge PM10 samplers. J. Air Waste Mgmt. Assoc. 46 (2),
142–147.
US Environmental Protection Agency, 1987. Revisions to the
national ambient air quality standards for particulate
matter: 40 CFR parts 51 and 52. Federal Register 52 (July
1), 24634–24669.
US Environmental Protection Agency, 1997. National ambient
air quality standards for particulate matter: Final rule.
Federal Register 62(138), 38651–38701.
US Environmental Protection Agency, 2002. Air quality criteria
for particulate matter (April 2002)––Volumes I and II––
third external review draft. Report No. EPA/600/P-99/002a.
US Environmental Protection Agency, Office of Research
and Development, Washington, DC.
Vedal, S., 1997. Critical review––Ambient particles and health:
Lines that divide. J. Air Waste Mgmt. Assoc. 47 (5), 551–
581.
Venkatram, A., 1988. On the use of kriging in the spatial
analysis of acid precipitation data. Atmos. Environ. 22 (9),
1963–1975.
Volpi, G., Gambolati, G., 1978. On the use of a main trend for
the kriging technique in hydrology. Adv. Water Res. 1, 345–
349.
Walpole, R.E., Meyers, R.H., 1978. Probability and Statistics
for Engineers and Scientists, 2nd ed. Collier Macmillan,
London.
Watson, J.G., Chow, J.C., DuBois, D.W., Green, M.C., Frank,
N.H., Pitchford, M.L., 1997. Guidance for network design
and optimal site exposure for PM2:5 and PM10. Report No.
EPA-454/R-99-022. US Environmental Protection Agency,
Research Triangle Park, NC.
Watson, J.G., Chow, J.C., Moosm€uuller, H., Green, M.C.,
Frank, N.H., Pitchford, M.L., 1998. Guidance for using
continuous monitors in PM2:5 monitoring networks. Report
No. EPA-454/R-98-012. US Environmental Protection
Agency, Research Triangle Park, NC.
Watson, J.G., Chow, J.C., Frazier, C.A., 1999. X-ray fluores-
cence analysis of ambient air samples. In: Landsberger, S.,
Creatchman, M. (Eds.), Elemental Analysis of Airborne
Particles, vol. 1. Gordon and Breach Science, Amsterdam,
pp. 67–96.
Watson, J.G., Chow, J.C., 2001a. Ambient air sampling. In:
Baron, P., Willeke, K. (Eds.), Aerosol Measurement:
Principles, Techniques, and Applications, 2nd ed. John
Wiley & Sons, New York, pp. 821–844.
Watson, J.G., Chow, J.C., 2001b. Estimating middle-, neigh-
borhood-, and urban-scale contributions to elemental car-
bon in Mexico City with a rapid response aethalometer.
J. Air Waste Mgmt. Assoc. 51 (11), 1522–1528.
Watson, J.G., 2002. Critical review–Visibility: Science and
regulation. J. Air Waste Mgmt. Assoc. 52 (6), 628–713.
Wedding, J.B., Weigand, M.A., 1993. An automatic particle
sampler with beta gauging. J. Air Waste Mgmt. Assoc. 43
(4), 475–479.
Wilkins, E.M., 1971. Variational principle applied to numerical
objective analysis of urban air pollution distributions.
J. Appl. Meteorol. 10, 974–981.
Williams, W.H., 1978. A Sampler on Sampling. John Wiley &
Sons, New York, NY.
Wilson, R., Spengler, J.D. (Eds.), 1996. Particles in Our Air:
Concentrations and Health Effects. Harvard University
Press, Cambridge, MA.
Wilson, W.E., Suh, H.H., 1997. Fine particles and coarse
particles: Concentration relationships relevant to epidemio-
logic studies. J. Air Waste Mgmt. Assoc. 47 (12), 1238–
1249.
Woldt, W., Bogardi, I., 1992. Ground water monitoring
network design using multiple criteria decision making
and geostatistics. Water Res. Bull. 28 (1), 45–62.
Yocom, J.E., McCarthy, S.M., 1991. Measuring Indoor Air
Quality––A Practical Guide. John Wiley & Sons, Chiches-
ter, West Sussex, England.
978 J.C. Chow et al. / Chemosphere 49 (2002) 961–978