EPA/600/A-97/054
Particulate Data from the First Year of Monitoring in Phoenix:
Part I. Fine and Coarse Mass
Jack Suggs and Jack Shreffler
U.S. Environmental Protection Agency
NERL, ORD
Research Triangle Park NC 27711	~
Abstract
As a result of recent findings of statistical relationships between ambient particulate matter
concentrations and mortality, the EPA's Office of Research and Development has begun to establish
monitoring sites to collect data which have more detail (in terms of size cuts and composition) than
those used in previous epidemiological studies. The Phoenix site was designed to evaluate various
particulate samplers in an area that is expected to be heavily influenced by a component of windblown
dust.
This paper examines one year (February 1995-January 1996) of particulate mass data from that
site. The data include 24-hr fine (PM2.5) and coarse (PM2.5 to PM10) mass fractions from three
different samplers. In addition, 1-hr PM2.5 and PM10 measurements are available from continuous
methods and are supported by 1-hr meteorological data. This paper also addresses method
comparisons, precision and accuracy of the instruments, and characterization of the relation between
fine and coarse mass. Seasonal and meteorological influences on the distributions of mass are
explored. Special studies on the reliability of the continuous methods are examined.
Introduction
In February 1995 the National Exposure Research Laboratory (NERL) at Research Triangle
Park, N.C. began the collecting particulate matter (PM) data with regard to aerosol size and
composition in Phoenix, AZ. This study is part of an ongoing project identified as the National PM
Research Monitoring Network which was initiated in response to the EPA's Office of Research and
Development's (ORD) research strategy to examine PM associations with morbidity and mortality in
various neighborhoods in different U.S. metropolitan areas with widely differing PM size and
composition characteristics 1. The Phoenix study primarily focused on the operational aspects of
collecting and characterizing daily PM size and composition data. Along with meteorological data,
daily fine (0-2.5 jum), PM10 (0-10 /xm) and coarse (2.5-10 /xm) particle mass (^g/m3) including metals
and organic carbons were collected. Particles were measured with four different makes of samplers: 1)
Tapered Element Oscillating Microbalance (TEOM) (Rupprecht & Patashnick, Inc. Series 1400a), 2)
Dual Fine Particle Sequential Sampler (DFPSS), 3) Versatile Air Pollutant Sampler(VAPS), and 4)
Anderson Model SA 241 Dichotomous Sampler (Dichot). Three different TEOM instruments were
used to provide integrated 1-hour mass measurements for particles. The first two were used to collect
PM2.5 and PM10 mass. The PM2.5 TEOM was fitted with a University Research Glassware (URG)
cyclone inlet and the PM10 TEOM was fitted with an Anderson Model 246B PM10 inlet. The third
TEOM was used to conduct three special studies: to determine background noise using blank filters

-------
(Feb. 7-Feb. 28), to collect PM2.5 with case temperature at 40 C instead of the prescribed 30 C
(May 18-Aug. 24), and to collect PM1 (0-1.0 /tm) mass (Feb. 28-May 17 and Aug. 8 -Jan. 22). A
URG PM2.5 or PM1 cyclone inlet was used in the?'* studies. TEOM coarse mass was determined by
subtracting TEOM PM2.5 from TEOM PM10 mass. It should be noted that under certain conditions
of rapidly fluctuating humidity, negative measurements are possible at low concentrations for the
TEOM 2. Negative hourly values are considered valid and are used to calculate 24-hour averages.
(Note: Due to low humidity in Phoenix, TEOMS were operated at 30 C rather than the recommended
50 C.) The DFPSS, which also employs a URG cyclone inlet identical to the PM2.5 TEOM,
provided 24-hour integrated PM2.5 mass on both quartz and Teflon filters (quartz for carbon analysis,
Teflon for mass and X-ray analysis). The VAPS is a dichotomous sampler employing a virtual
impactor to provide a cut point of 2.5 /xm. It provides two separate channels to collect the PM2.5
samples- one employing quartz filters for carbon data and the other employing Teflon filters for mass
and X-ray determinations. Coarse particles were collected on the third channel using polycarbonate
filters. During this study, the VAPS operated on a 24-hour sampling interval at 33 LPM using an
Anderson Model 246B PM10 inlet (twice the flow rate for which the PM10 inlet was designed). The
Dichot employs the Anderson PM10 inlet along with a virtual impactor and operates on a 24-hour
interval at a flow rate of 16.7 LPM collecting particles on two channels- one for fine and the other for
coarse particles. For both dichotomous samplers, PM-10 mass is determined by adding the mass from
the coarse and fine channels.
Method Comparisons
Since the DFPSS, VAPS and Dichot all measure on a 24-hour interval, the TEOM hourly mass
measurements (along with accompanying meteorological data) were averaged over a 24-hour period
beginning at 7:00 a.m. to correspond to the changing of filters on the other instruments. At least 20
hourly measurements were required for a daily average. The frequency of sampling varied with
methods. The TEOM and DFPSS sampled daily while the VAPS and Dichot sampled every third day
with the exception of some special study periods: between February 26 and March 4 the VAPS was
operated for seven consecutive days but the Dichot was not operated; between March 11 and March
17 both the VAPS and Dichot were operated for six consecutive days; and between September 14 and
September 18 the VAPS and Dichot were operated for five consecutive days, and Teflon filters were
used on the VAPS coarse channel.
A time series plot of 24-hour fine particle mass (^g/m3) for each of the four methods is shown
in Figure 1 along with wind speed (m/sec) plotted as a solid line. This is a scatter plot showing
relative daily concentrations and sampling frequencies over a year-long period. The most noteworthy
thing about this plot is that beginning in October the concentrations of fine particles begin to increase
to the highest levels of the year. Corresponding to this increase in concentration is a decrease in wind
speed. It should be noted that the drop in wind speed is accompanied by a shift to a predominately
easterly direction during this period. This increase in concentration is also characteristic of coarse and
PM10 particles plotted over time shown in Figures 2 and 3. Detailed information on the distribution of
particle mass is given in Table 1.
Box plots in Figure 4 display the distributional characteristics of measurements for each method

-------
and" for each particle size. The shaded areas contain 50% of the measurements between the upper and
lower quartiles (25th and 75th percentile). About 99% of the measurements are contained between the lo-
bars that extend from ^ach end of the box. The horizontal lines represent potential outliers. The middle
of the notch is the median or 50th percentile and the edge of the notches represent upper and lower
confidence intervals for the medians. Any two notches that do not overlap indicate that the two methods
being compared are significantly different. Statistical comparisons were based on the difference (in logs)
between daily pairs of measurements for two methods. The TEOM and DFPSS were determined to be
statistically the same (within 1%) for fine mass but both measured significantly higher (15% to 17%
higher) than either the VAPS and Dichot which were within 2% of each other for fine particles. Box
plots for coarse and PM10 are also shown in Figure 4. None of the methods were equivalent for coarse
or PM10 mass based on percent differences that ranged in absolute value between 26% for TEOM-
versus- Dichot on PM10 to 90% for TEOM-versus-VAPS on coarse mass. The VAPS median in Figure
4 is noticeably lower for coarse mass than the other methods. This is probably the result of a lower cut
point for the VAPS coarse fraction caused by operating at a higher flow rate (33 LPM) compared to the
usual flow rate of 16.7 LPM normally used for a PM10 inlet3. Particle stability on polycarbonate filters
was reported to be a problem due to static charges developed during sampling. Coarse particle mass is a
factor of 2 to 3 times higher than the fine mass for TEOM and Dichot samplers. This is the reciprocal of
coarse/fine ratios typically found at sites not located in desert regions 4,5. On the other hand, the VAPS
collected about equal amounts of coarse and fine mass with a ratio of .97 which is significantly lower
than the other two samplers. Here again this may be attributed to the high rate of flow. However, this
ratio changed to a factor of 2 for the 5-day special study where Teflon filters replaced polycarbonate
filters on the coarse channel.
Particles with a size fraction less than or equal to 1.0 [im (PM1) were collected as part of a
special study on a third TEOM. The distributional scatter of daily mass is plotted over time in Figure 5
along with PM2.5 mass. As with other particle sizes there is also a rise in PM1 mass during the fall
months compared to measurements made in the spring and late summer. Figure 6 is a scatter plot of
differences between logs of daily pairs of PM2.5 and PM1 plotted against their mean. Although there is
more scatter at the lower concentrations, a fitted line shows that on the average PM2.5 mass is about
40% higher than PM1 over a range of from zero to approximately 40.0 jxg/m3.
Instrument Precision and Accuracy
The error of a measurement is simply the difference between the measurement and the actual (or
true) value as expressed in the following linear model:
Measurement = True value + Measurement Error.
The average of the measurement errors for repeated measurements at a given true value is a measure of
the instrument bias or systematic error. The smaller the bias, the more accurate the instrument is
considered to be. Although absolute bias cannot be estimated without knowing the true value, it can
usually be controlled by proper calibration and therefore is not considered a serious problem. On the
other hand, measurement error fluctuates around the bias in a random manner that is more serious and

-------
difficult to control (or at least minimize) since this fluctuation can result from the combination of several
different external influences and environmental conditions (e.g. weather, operators, handling, etc.) . The
standard deviation of the measurement error is a measure of the precision of the instrument. Precision (or
imprecision) is simply a measure of how well an instrument can repeat itself under the same conditions.
For ambient instruments this is a question of how well two or more instruments of the same type -
operating simultaneously in close proximity - can replicate each other over several days of sampling. The
total uncertainty associated with a single daily measurement is a combination of the day-to-day variation
between true values plus instrument precision. Therefore, it is important that the instrument precision be
many times smaller (an order of magnitude as a rule of thumb) than the day-to-day variation in the true
concentrations in order to have high efficiency in the measurement process 6. Precision also determines
the sensitivity of an instrument by controlling how quickly it responds to changes in daily concentration
levels. None of the instruments used in this study were duplicated for the purpose of estimating
precision. Nevertheless, estimates of precision for each individual instrument were determined using
multivariate techniques on the available data \ For two instruments, this basically involved subtracting
the eovariance between paired measurements (which is an estimate of the variance of the true values)
from the variance of the daily measurements for each separate instrument and taking the square root. For
more than two instruments, calculations were slightly more involved 7. The underlying assumption is that
different instruments do replicate each other since they simultaneously measure the same true unknown
daily concentration but with possibly different bias as well as different precision. This possibility exists
even for duplicate instruments, but usually only the bias is hypothesized to be different while the
precision is assumed to be same for both instruments. The average difference between daily pairs of
measurements for any two instruments is an estimate of the difference in bias. The percent differences
discussed in the previous section are a measure of bias between any two instruments relative to their
average. The more precise the instruments are, the better we are able to detect small differences in bias
between the instruments. Table 2 gives an estimate of precision for each different instrument and for
each particle size. The original measurements were transformed into logarithms prior to analysis.
Therefore, the estimates of precision are expressed as a percent of the concentration in the original units
(usually referred to as the percent coefficient of variation). For example, the precision of the TEOM for
a single 24-hour measurement was estimated to be +28%. This implies that about one third of the true
values are measured with error greater than 28% of the concentration level. It also implies that if two
TEOMs with the same precision were operated side-by-side then about half of the daily pairs would
differ by more than .95(28%)= 27% and, about one time out of 20, pairs would differ by as much as
2.77(28%)= 78% due to random error even if they had the same bias. Precision for the other
instruments can be summarized in a similar way.
Background noise for the TEOM was estimated using blank filters to be + .5 ^g/m3 (1-sigma) for
24-hour measurements. Since there is such a small probability of background noise contributing more
than 1.5 fig/m3 (i.e. 3-sigma) to a measurement, anything above this value must be a real quantity.
According to the ACSCEI9, 1.5 /xg/m3 would constitute a limit of detection (LOD) for TEOM 24-hour
averages.
The standard deviation of the true values for this study was estimated to be 48% (of the mean
concentration level) for fine particles, and 40% for coarse and PM10 particles. Therefore, in the

-------
absence of measurement error, nearly all 24-hour fine particle concentrations would fall between zero
and 30 ng/m3 assuming that the overall mean is about 14 ng/m3. For the TEOM, instrument imprecision
accounts for about 25% (i.e. .282/(.482-f ,282)) of the total variation of a single 24-hour measurement.
Taking into account instrument precision widens the interval to (-5, 33) (ig/m3 allowing for negative
TEOM measurements. This agrees with the percentiles of the actual measurements shown in Table 1 for
the TEOM. This increase in the upper limit is an example of how imprecision can lead to overestimation
of peak values which may result in falsely claiming noncompliance with a standard when the true value is
actually below the standard10,11. One way to reduce the effects of instrument error is through the diligent
use of quality control procedures. Another way would be to deploy several instruments of the same type
at a given site and use the average of the measurements. For example, the instrument precision for the
VAPS (6%) contributes only 2% to the overall variation in a single daily PM2.5 measurement. On the
other hand, the imprecision of the TEOM ( 28%) contributes about 25% to the overall variation. The
ratio of the precision estimates of two different instruments is a measure of their relative efficiency.
Based on the estimates of precision from this study, the average of twenty-two TEOMs would be
required in order to be as efficient as the VAPS. This is impractical as well as expensive but it does
illustrate the importance of instrument efficiency in achieving high quality data.
As previously mentioned, when duplicate instruments are used, the precision is usually assumed
to be the same for both instruments. In this case, the error variance of a single instrument is estimated as
half the variance of the difference between paired values. This may not always be a safe assumption even
when the instruments are the same type since one instrument may have a significantly higher precision
than the other. For example, except for the temperature levels at which the instruments were operated,
the TEOM at 40 C may be considered a duplicate for the TEOM at 30 C for 24-hour averages. The
precision for the TEOM (30) was estimated to be 24%, and, for the TEOM (40) , the precision was
so low as to be statistically indistinguishable from zero. Assuming they have the same precision, the
estimate for each instrument is 17%.
A final note regarding precision as it relates to linear regression: one of the necessary
assumptions for regression is that the dependent variable be measured without error. .Not accounting for
error in the measurement will cause underestimations of the regression coefficient onto the true
concentrations often referred to as the structural regression coefficient12. The magnitude of the
underestimation depends on the imprecision of the measuring instrument being used. In this study, a
precision of less than 10% reduces the regression coefficient by 4% for fine particles, but an imprecision
of 28% reduces the estimate by as much as 1/4. This could seriously affect the estimates of mortality as
it relates to particle loadings: the greater the imprecision, the lower the estimate of mortality for a given
true concentration. Adjustments in regression coefficients due to error would help to normalize results
for comparisons across sites where instruments measure with different precision.
Hourly TEOM Measurements
One of the unique things about the TEOM is the ability to collect particle data on a continuous
basis which makes it possible to examine diurnal patterns. Hourly values of TEOM fine particles are
plotted as a time series in Figure 7. The specific humidity - a measure of the water content in the air in
units of grams II ,0 per kilograms of moist air-is also plotted. During the time between Julian dates of 225

-------
to 235 corresponding to calendar dates of mid to late August there is a noticeable scatter in the hourly
fine fraction. This corresponds to the time when there is relatively high water content (>7.0 g/kg) in the
air suggesting 'hat the TEOM is unstable under such conditions. The operating temperature of the TEOM
was 30 C. During this period of time a special study was being conducted with another TEOM
sampling with a temperature of 40 C. Values from both instruments are plotted in Figure 8, Although
there are spikes in the readings of both instruments corresponding to high values of specific humidity
during this time period, the TEOM at 40 C appears more stable with spikes about 1/5 as great as the
TEOM at 30 C .
The relative instability of the TEOMs is even more noticeable in Figure 9 where the differences
in hourly fine particle mass between the TEOMs are plotted against the specific humidity. These
differences are clustered about a fitted line (solid) at zero (broken line). The scatter about this line is
extremely tight up to about 7 g/kg. About 15% of all hourly measurements were taken when the specific
humidity was below 7 g/kg. In this range both monitors have an average concentration of 10.2 /xg/m}
with approximately 99% of the measurements for each monitor between limits (-9, 27) /ig/m3. Above 7
g/kg specific humidity this spread in hourly fine mass stays about the same for the TEOM at 40 C but
the width of the interval increases to (-25, 47) fxg/m3 for the TEOM at 30 C. Above 12 g/kg specific
humidity this spread increases to (-49, 64) fig/m3 even though the average for both instruments remains
approximately 10 /ig/m3. It appears that the higher temperature reduces the scatter in hourly
measurements by a factor of 3 when the specific humidity is above 12 g/kg. Even with this
nonhomogeneity in scatter, the averages remain comparable across the entire range of concentrations
regardless of the water content. This comparability between averages was also evident for 24-hour values
involving the TEOM and DFPSS where the difference was within 1 %.
As was demonstrated above, diurnal patterns can be highly variable from hour to hour and across
days. By averaging over all days for a given hour we are able to smooth the diurnal plots. These patterns
for all particle sizes along with wind speed are shown in Figure 10. Peak levels occur around 7:00 to
8:00 a.m. and 10:00 to 11:00 p.m. with lowest levels generally occurring about 4:00 p.m.. Wind speeds
are maximum in the afternoon causing dilution in the mass concentrations. When these plots are done by
season, the diurnal pattern does not change; however, the vertical scales change to reflect a slowdown in
wind speed during the fall and winter months and an increase in mass concentrations during this same
period.
Conclusions
The results of this pilot study have great value in providing direction and expectations for future
work in the National PM Research Monitoring Network. Comparisons between methods reported in this
study form a baseline for future comparisons under similar conditions especially for fine particle mass.
Monitors equipped with cyclones measured significantly higher concentrations than the dichotomous
samplers for the same particle size. It should be pointed out that differences may be due in part to
conditions under which the samplers were operated. The dichotomous samplers were entirely outside the
sampling trailer while the TEOMs and DFPSS only had the inlets outside. Therefore it is possible that
ambient temperature may have affected the observed differences in mass rather than the type of inlets.
Measurement precision was estimated without benefit of duplicate samplers by using a model that takes

-------
advantage of simultaneous readings from all instruments for each particle size. Verification of this
procedure would require that duplicate samplers be operated under similar conditions; however, there is
no guarantee that instruments of the same type have the same precision. For fine particles the VAPS and
DFPSS had the best precision followed by the Dichot and TEOM, If the degree of imprecision relative to
the total variability in a measurement is not taken into account, false conclusions concerning compliance
testing or the statistical association between particle mass and mortality in urban areas could result. The
fluctuations in TEOM hourly measurements caused by water in the atmosphere is effectively canceled by
daily averaging. Raising the operating temperature in the housing to 40 C also reduces the instability in
hourly measurements due to water (>7 g/kg) in the air. Peaks and valleys in daily concentration levels
for all particle sizes appear to be seasonal and are inversely related to seasonal wind speed patterns. This
is examined in more detail in Part II13. Diurnal patterns for all particle sizes show a similar behavior
with respect to hourly wind speed with peak concentrations occurring in the morning and late evening
regardless of the time of year.
Acknowledgments
The authors acknowledge the collaboration of all other scientists and engineers participating in the US
EPA's National PM Research Monitoring Network and especially R. Zweidinger, L. Purdue, J. Bowser,
R. Stevens, and D. Gemmill.
References
1.	Zweidinger, R.B., and L.J. Purdue, "Particle Research Program Status Report", USEPA Memo,
June 10, 1996.
2.	Allen, G.A. and R.F. Burton, "Comparison of The TEOM Continuous Mass Analyzer with
Gravimetric Respirable Mass," Proceedings of the 1991 U.S. EPA/AWMA International Symposium
on Measurement of Toxic and Related Air Pollutants, VIP-21, Air and Waste Management
Association, Pittsburgh, 1991, 166-171.
3.	Lane, D.D., S.J. Randtke, R.E. Carter, R.K. Stevens, "Particle and Gas Transmission
Characteristics of a VAPS System and Three Different Inlets," in Proceedings of the 1993 U.S.
EPA/AWMA International Symposium on Measurement of Toxic and Related Air Pollutants, VIP-34,
Air and Waste Management Association, Pittsburgh, 1993, 457-463.
4.	Swartz, J., Dockery, D.W. and Neas, L.M., "Is Daily Mortality Associated Specifically with Fine
Particles?", J.Air&Waste Manage...Assoc., 1996 46 927-939.
5.	Burton, R.M., Suh, H.H and Koutrakis, P., "Spatial Variation in Particulate Concentrations within
Metropolitan Philadelphia," Envir. Sci Tech, 1996 3D 400-407.
6.	Grubbs, F.E., "Errors of measurement, Precision, Accuracy and the Statistical Comparison of
Measuring Instruments," Technometrics 1973 15 53-66.

-------
7.	Christensen, R. and L.G. Blackwood, "Tests for Precision and Accuracy of Multiple Measuring
Devices," Technometrics, 1993 35 411-420.
8.	Hawkins, D.M., "The Detection of Errors in Multivariate Data Using Principal Components,"
Journal of the American StatisticaLA.ssociation 1974 69 340-344
9.	American Chemical Society Committee on Environmental Improvement(ACSCEI), Guidelines for
data acquisition and data quality evaluation in environmental chemistry, Anal. Chem. 1980 52 2242-
2249.
10.	Suggs, J.C., and T.C. Curran, "An Empirical Bayes Method For Comparing Air Pollution Data To
Air Quality Standards," Atmospheric Environment 1983 11 837-841.
11.	Curran, T.C. and J.C. Suggs, "Effects of Measurement Uncertainty on Air Quality Summary
Statistics," Atmospheric Environment 1985
12.	Snedecor, G.W., and W.G. Cochran, Statistical Methods, The Iowa State University Press:
Ames, 1967; pp 164-165.
13.	Shreffler, J.H., and J.C. Suggs, "Particulate Data from the First Year of Monitoring in Phoenix:
Part II," in Proceedings of the 1997 U.S. EPA/AWMA Symposium on Measurement of Toxic and
Related Air Pollutants, Air and Waste Management Association, Pittsburgh, 1997.
Disclaimer
This paper has been reviewed in accordance with the US EPA's peer review policies for
scientific papers and approved for publication. The contents are not intended to represent agency
policy. Mention of trade names or commercial products does not constitute endorsement or
recommendation for use.

-------
Table 1. Distributional statistics for 24-hr particle mass (pcg/m3).


N
Min
5%
25%
50%
75%
95%
Max
Mean
Std Dev

TEOM
335
-5.9
3.4
8.0
11.6
18.0
29.6
40.2
13.6
8.1

DFPSS
284
2.1
5.1
8.0
10.6
16.0
27.5
37.4
12.9
7.0
Fine
VAPS
99
3.4
4.7
6.5
8.2
12.5
23.8
26.2
10.2
5.5

Dichot
81
.6
4.4
6.8
8.6
12.3
24.8
27.3
10.9
6.4

TEOM
326
-4.0
12.4
21.0
28.1
38.8
60.8
103.7
31.3
15.5
Coarse
VAPS
101
-0.2
3.8
5.4
8.0
11.9
26.1
43.4
10.2
7.7

Dichot
81
3.9
9.1
16.6
23.3
28.8
48.9
81.4
24.7
13.1

TEOM
327
-1.8
18.0
28.7
39.9
55.8
86.7
129.1
44.6
22.0
PM10
VAPS
99
6.9
9.0
12.3
16.3
23.1
52.2
67.2
20.4
12.5

Dichot
81
11.3
15.0
22.9
32.0
40.8
70.8
104.6
35.8
18.5
Table 2. Precision of individual 24-hr measurements.

TEOM
DFPSS
VAPS
Dichot
Fine
 28%
8%
6%
17%
Coarse
18%

 44%
21%
PM10
17%

 26%
+14%

-------
o
T
O
CO
O
CM
TEOM
DFPSS
VAPS
D1CH0T
o o o
~o
A f? D
TD
m
o
Q.
CO
x>
c
02/08/95
06/30/95
11/19/95
Time In days
Figure 1. Time series of fine particle mass
D
TEOM

VAPS
ft
OICHOT
cP
'<#Jp
 ate"  a
. F ^ 6	&
J! *'	<5?
o B
O Oo0 b
f
D
TGM
O
VAPS
*
DiCHOT
cP.-P
,d^ D RP
. 3 q*
11/19/95
Time in days
Figure 2. Time series of coarse parties
Time in days
Figure 3. Time Series of PM-10 particles

-------
TEOM DFPSS VAPS DICHOT TEOM VAPS DICHOT TEOM VAPS DICHOT
Fine Fine Fine Fine Coarse Coarse Coarse PM10 PM10 PM10
Figure 4. Box plots of particle mass

-------
a
o
PM2.5
+
PM1.0
o
~
~
o
~
D
~
~
rP ~
r-) HiJ O '"O ^ tSP
^SD	~ tP
h
o ~
TT- ^Sp.
02/07/95
06/29/95
11/18/95
Time in days
Figure 5. Time series of TEOM fine particles
Mean (ug/m3)
Figure 6. Paired Differences in logs between PM25 and PM1 plotted against means

-------
100
200
Julian Date
300
400
100
200
300
400
Julian Date
Figure 7, Hourly values of TEOM fine particles 
-------
5
10
15
Specific humidity (g/kg)
Figure 9. Plot of fine particle differences between TEOMS operated at 30 and 40 degrees C
o-o
~
o
PM10
o
PM2.5

PMCOARSE

PM1
o
Wind speed
X
~ .
TJ
/ X
/
~
 \
/
o ~
~
/
~
8C
o
~
o -
N
~ .
& - A

o
O-o.
\
o
o
O-o jo
~ -D/
\
~
\
 o.
 o
o,
/ o
o N<> - ^ - o
'O.
O-o.
/
'O.
'o.
-o-0o/
-o'
10
I
15
I
20
Hour
Figure 10. TEOM diurnal plots

-------
j TECHNICAL REPORT DATA
1. REPORT NO, 2.
EPA/600/A-97/054
.
4. TITLE AND SUBTITLE
Particulate Data from the First Year of Monitoring
in Phoenix: Part I. Fine and Coarse Mass
5.REPORT DATE
6.PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Jack Sugfts and Jack Shreffler
8.PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
National Exposure Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park NC 27711
10.PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
12. SPONSORING AGENCY NAME AND ADDRESS
National Exposure Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park NC 27711
13.TYPE OF REPORT AND PERIOD COVERED
Symposium Proceedings
14. SPONSORING AGENCY CODE
EPA/600/09
15. SUPPLEMENTARY NOTES
16. ABSTRACT
As a result of recent findings of statistical relationships between ambient particulate matter
concentrations and mortality, the EPA's Office of Research and Development has begun to establish
monitoring sites to collect data which have more detail (in terms of size cuts and composition) than
those used in previous epidemiological studies. The Phoenix site was designed to evaluate various
particulate samplers in an area that is expected to be heavily influenced by a component of windblown
dust.
This paper examines one year (February 1995-January 1996) of particulate mass data from that
site. The data include 24-hr fine(PM2.5) and coarse(PM2.5 to PM10) mass fractions from three
different samplers. In addition, 1-hr PM2.5 and PM10 measurements are available from continuous
methods and are supported by 1-hr meteorological data. This paper also addresses method
comparisons, precision and accuracy of the instruments, and characterization of the relation between
fine and coarse mass. Seasonal and meteorological influences on the distributions of mass are
explored. Special studies on the reliability of the continuous methods are examined.
17. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
b.IDENTIFIERS/ OPEN ENDED
TERMS
c.COSATI



18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (This
Report)
UNCLASSIFIED
21.NO. OF PAGES

20. SECURITY CLASS (This
Page)
UNCLASSIFIED
22. PRICE

-------