United States
Environmental Protection
Agency
Multi-Site Evaluations of
Candidate Methodologies for
Determining Coarse
Particulate Matter (PM10.2.5)
Concentrations:
August 2005 Updated Report
Regarding Second-Generation
and New PM10-2.5 Samplers
RESEARCH AND DEVELOPMENT
-------
-------
September 2006
E P A/600/R-06/093
Multi-Site Evaluations of Candidate
Methodologies for Determining
Coarse Particulate Matter (PM10.2.5)
Concentrations:
August 2005 Updated Report Regarding Second-
Generation and New PM10_2 5 Samplers
By
Robert Vanderpool, Timothy Hanley, Fred Dimmick, and Elizabeth Hunike
United States Environmental Protection Agency, 109 T.W. Alexander Drive
Research Triangle Park, NC 27711
Paul Solomon
United States Environmental Protection Agency, P.O. Box 93478
Las Vegas, NV 89193
Frank McElroy, Robert Murdoch, and Sanjay Natarajan
RTI International, 3040 Cornwallis Road
Research Triangle Park, NC 27709
National Exposure Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC, 27711
-------
Notice
The U.S. Environmental Protection Agency through its Office of Research and Development partially
funded and collaborated in the research described here under contract number 68-D-00-206 (Alion Science and
Technology). It has been subjected to the Agency's peer and administrative review, and it has been approved
for publication as an EPA document. Mention of trade names or commercial products does not constitute
endorsement or recommendation for use.
ii
-------
Abstract
Field studies were conducted to evaluate the performance of sampling methods for measuring the
coarse fraction of PMi0 (PM10-2.5) in ambient air. In the first stage of this evaluation, both time-integrated filter-
based and direct continuous methods were evaluated. As the primary basis of comparison upon which to
evaluate the candidate coarse fraction sampling methods, the PMi 0.2 5 concentration was estimated by taking the
difference between concentrations measured with a PMi0 FRM sampler and a PM2 5 FRM sampler. Sampling
sites in Gary, IN, Phoenix, AZ, and Riverside, CA were selected to provide diverse challenges to the samplers
with respect to aerosol concentration, size distribution, and composition. The first performance evaluations
were conducted during 2003 and 2004. Instrument manufacturers were provided the results of these studies and
encouraged to revise their PMi 0-2.5 samplers to address the measurement uncertainties identified. EPA
conducted a follow-on field study of these second-generation samplers in Phoenix, AZ during April and May,
2005. The 2005 field study also evaluated four new prototype samplers designed to measure ambient PMi 0-2.5
aerosols. Two of these newer samplers were time-integrated, filter-based designs, and two were continuous-
measurement designs. Despite operational problems encountered with some of the instruments during this
phase of the testing, overall measurement results were encouraging and the instrument manufacturers are
continuing to improve the reliability of the coarse mode samplers.
iii
-------
Contents
Notice ii
Abstract iii
Figures vi
Tables viii
Acknowledgments ix
Chapter 1: Introduction 1
Chapter 2: Description of PMi 0-2.5 Samplers 3
Collocated PM2 5 and PMi0 FRM Samplers 3
R&P Model 2025 Sequential Dichotomous Sampler 5
Kimoto Inc. Model SPM-613D Dichotomous Beta Gauge 6
R&P Continuous Coarse TEOM Monitor 6
TSI Inc. Model 3321 Aerodynamic Particle Sizer (APS) 7
R&P Single-Event Dichotomous Sampler 8
Sierra-Andersen Model 241 Dichotomous Sampler 8
BGI frmOMNI Ambient Air Sampler (Filter Reference Method) 9
Grimm EnviroCheck Model 1.107 Sampler 9
R&P Dichotomous TEOM Sampler 9
Chapter 3: Site Setup and Operating Procedures 11
Chapter 4: Site Characteristics 13
Gary, IN (March - April, 2003) 13
Phoenix, AZ (May - June, 2003) 14
Riverside, CA (July - August, 2003) 14
Phoenix, AZ (January 2004) 15
Phoenix, AZ (April - May, 2005) 15
Chapter 5: Test Results 18
Collocated PM2 5 and PM10 FRM Samplers 18
Year 2003, 2004, and 2005 Test Results 18
R&P Model 2025 Sequential Dichotomous Sampler 18
Year 2003 and 2004 Test Results 18
Sampler Design Modifications 22
Year 2005 Phoenix Test Results 22
Single-Event, Manual Dichotomous Sampler 22
Multi-Event, Manual Dichotomous Sampler 22
iv
-------
R&P Coarse TEOM Monitor 24
Year 2003 and 2004 Test Results 24
Sampler Design Modifications 26
Year 2005 Phoenix Test Results 26
Kimoto Inc. Model SPM-613D Dichotomous Beta Gauge 27
Year 2003 and 2004 Test Results 27
Sampler Design Modifications 29
Year 2005 Phoenix Test Results 29
Runs 1-15 29
Runs 16-30 30
TSI Inc. Model 3321 Aerodynamic Particle Sizer (APS) 30
Year 2003 and 2004 Test Results 30
Sampler Design Modifications 31
Year 2005 Phoenix Test Results 33
BGI frmOMNI Ambient Air Sampler (Filter Reference Method) 33
Year 2005 Phoenix Test Results 33
Grimm EnviroCheck Model 1.107 Sampler 35
Year 2005 Phoenix Test Results 35
R&P Dichotomous TEOM Sampler 36
Year 2005 Phoenix Test Results 36
Chapter 6: Summary 38
Chapter 7: References 41
v
-------
Figures
Number Page
2-1 Schematic diagram of the FRM samplers used in the PMi 0-2.5 difference method 3
2-2 Schematic diagram of flow system and sample exchange mechanism of the
R&P Model 2025 sequential dichotomous sampler 5
2-3 Photograph of the Kimoto Model SPM-613D beta gauge 6
2-4 Schematic diagram of the Kimoto Model SPM-613D beta gauge 6
2-5 Photograph of the R&P PMi 0-2.5 TEOM 7
2-6 Photograph and measurement schematic of the TSI Aerodynamic Particle Sizer 7
2-7 Photograph of the R&P single-event dichotomous sampler 8
2-8 Photograph of the Sierra-Andersen Model 241 dichotomous sampler 8
2-9 Photograph of the BGI Omni saturation sampler 9
2-10 Photograph of the Grimm Model 1.107 sampler 10
2-11 Diagram of the R&P dichotomous TEOM sampler 10
3-1 Photograph of the PMi 0-2.5 sampler evaluation platform at the Gary, IN site 11
3-2 Photograph of the cassette storage canisters and temperature-controlled shipping cooler 12
4-1 Timeline of Gary, IN PMi 0-2.5 concentration showing level of agreement between site weighing
and RTP, NC weighings 14
4-2 Site versus RTP, NC weighing of PM2 5, PM10.2.5, and PM10 concentrations at the 2003
Phoenix, AZ site 14
5-1 Performance of the R&P dichotomous samplers in Gary, IN versus the collocated FRM samplers 20
5-2 Timeline of R&P sequential dichot versus FRM PMi0 concentrations in Phoenix, AZ 20
5-3 Timeline of R&P sequential and single-event dichots versus collocated FRMs
in Phoenix, AZ (2005) 22
5-4 R&P coarse TEOM versus FRM PMi0-2.5 concentrations in Gary, IN 24
5-5 Regression of coarse TEOM versus PMi0-2.5 FRMs during the 2005 Phoenix tests 27
5-6 Timeline of Kimoto SPM-613D versus FRM PM2.5 concentrations in
Phoenix, AZ (2003) 27
5-7 Timeline of FRM and Kimoto PM2 5 concentrations during the 2005 Phoenix tests 29
5-8 Timeline of FRM and Kimoto PMi 0-2.5 concentrations during the 2005 Phoenix tests 30
5-9 Regression of Kimoto-2 versus the FRM showing the influence of Run 30 data on the regression
outcome 30
5-10 Timeline of mean APS and FRM PMi0-2.5 concentrations during the 15-day
Phoenix 2004 field tests 31
5-11 Regressions of PMi0-2.5 concentrations estimated by the Model 3321 APS versus those measured by
the collocated FRM samplers 31
5-12 Regression of the BGI Omni PM2 5 concentrations versus those of the collocated PM2 5
FRM samplers 34
5-13 Regression of Omni PMi0 concentrations versus those of the collocated PMi0 FRM samplers 34
5-14 Timeline of Grimm Model 1.107 PM2 5 concentrations versus those of the
collocated PM25 FRMs 35
5-15 Timeline of Grimm Model 1.107 PMi0-2.5 concentrations versus those of the
collocated FRMs 35
vi
-------
5-16 Timeline of R&P dichotomous TEOM PM25 concentrations versus those of the
collocated PM25 FRMs 36
5-17 Timeline of R&P dichotomous TEOM PM10.2.5 concentrations versus those of the
collocated FRMs 37
vii
-------
Tables
Number Page
2-1 Inventory of Samplers Used in the 2003 and 2004 Multi-Site Performance Evaluations 4
4-1 Inventory of Samplers Used in the 2005 Phoenix Performance Evaluation 16
5-1 Inter-Manufacturer Precision of the Collocated FRM Samplers as a Function of Sampling Site 19
5-2 Performance of the R&P 2025 Sequential Dichot versus the FRM during 2003 19
5-3 Comparison of Sequential versus Manual Operation of the R&P 2025 Dichots
and Sierra-Andersen Dichot in Phoenix, AZ (2004 Tests) 21
5-4 Performance of the R&P Sequential Dichots and Single-Event Dichots versus the FRMs
during the 2005 Phoenix Tests 23
5-5 Performance of the R&P Coarse TEOM versus the FRM 25
5-6 Performance of the Kimoto SPM-613D Beta Gauge Dichot versus the FRM 28
5-7 Performance of the TSI APS Model 3321 versus the FRM 32
viii
-------
Acknowledgments
The authors gratefully acknowledge the cooperation of the following instrument manufacturers
involved in these field studies: BGI, Inc., Grimm Aerosol Technik BmbH & Co., Kimoto Electric Co.,
Thermo Electron Corporation, Tisch Environmental, Inc., and TSI, Inc.
ix
-------
-------
Chapter 1
Introduction
In response to increasing evidence of the adverse health effects
associated with exposure to ambient fine particles, the United
States Environmental Protection Agency (EPA) promulgated in
1997 a national ambient air quality standard (NAAQS) forPM25
(U.S. EPA, 1997). Accompanying the standard were strict
design and performance requirements that candidate PM25
samplers must meet in order to be approved by EPA for use in
making compliance measurements (Noble et al., 2001). The
1997 regulations retained the existing annual PMi0 standard and
made only slight modifications to the statistical basis upon which
to assess compliance with the 24-hour PMi0 standard.
Based on subsequent litigation, the U. S. Court of Appeals for the
District of Columbia reviewed the 1997 regulations and upheld
EPA's promulgation of the PM25 standard. While
acknowledging the need to regulate coarse particles, the Court
vacated the 1997 PMi0 standard after concluding that PM10 is a
"poorly matched indicator for coarse particulate pollution"
because PMi0 includes the PM25 fraction. EPA did not appeal
this ruling and now intends to promulgate a new NAAQS for
PMio_2.5 (i.e. the coarse fraction of PM10).
Inherent to any new NAAQS is the need for sampling and
analysis methods capable of measuring the new indicator with
known quality. In support of this goal, the purpose of these field
studies was to conduct a survey of available instrumentation
designed to measure the coarse fraction of PMi0, and to conduct
a multi-site performance evaluation of these instruments.
Sampling sites were selected in order to evaluate the instruments
under a wide variety of environmental conditions, particle
concentrations, particle size distributions, and particle
compositions. At three separate cities (Gary, IN, Phoenix, AZ,
and Riverside, CA) thirty daily, 22-hour tests were conducted.
Following the Riverside study, an additional fifteen days of 22-
hour tests were conducted at the Phoenix, AZ sampling site. In
addition to filter-based samplers that provide integrated test
results, near real-time PMi0.25 monitors were evaluated that
possess time resolutions of one hour or less. Multiple monitors
of each type were used in order to determine the inherent
precision of each sampler's design.
A July 2004 report provided a description of the instruments
evaluated in the 2003 and the 2004 field studies, outlined the
sampling and analysis procedures used to conduct the
performance evaluations, described the characteristics of each of
the three sampling sites, and provided a summary of test results.
Since the time that report was prepared, additional validation of
the data has been conducted, and this August 2005 report
provides an updated presentation of those results. In addition,
shape factors have been incorporated into the data obtained with
the TSI aerodynamic particle sizerthat appreciably improved its
agreement with the collocated FRM samplers.
Following the completion of the 2003 and 2004 field tests, EPA
has been working with the instrument manufacturers to improve
the measurement performance of their respective PMi0.25
samplers. This report details the modifications that were made to
each of the samplers as a result of that collaborative effort. In
April and May, 2005, these "second generation" samplers were
evaluated during a 30-day field tests in Phoenix, AZ. In addition
to these samplers, the field tests incorporated four new PMi0.2 5
samplers designs. Two of these designs were integrated, filter-
based units (BGI Incorporated frmOmni saturation sampler and
the Rupprecht & Patashnick (R&P) single-event dichotomous
sampler) and two of the designs (Grimm EnviroCheck Model
1.107 and the R&P dichotomous TEOM sampler) were designed
to provide PM2 5 and PMi0.2 5 measurements on a near real-time
basis. This report describes these samplers in detail and presents
a summary of the 2005 Phoenix field test results for all samplers
involved in the study.
To challenge the same suite of candidate samplers to different
aerosol size distributions, aerosol composition, and
environmental conditions, preparations are currently underway to
1
-------
conduct an additional sampler evaluation study in Birmingham,
AL during fall 2005.
2
-------
Chapter 2
Description of PM10.2.5 Samplers
Selection of the samplers to be involved in the field comparison
study was based on the following criteria. First, all samplers had
to be designed to provide a measurement of the mass
concentration of PMi 0-2.5 aerosols based on aerodynamic
diameter. Selected filter-based samplers had to be capable of
providing integrated samples at least every 24 hours and use the
PM2.5 FRM's standard cassette and Teflon after-filter. Selected
continuous and semi-continuous instruments had to be capable of
providing PMi 0-2.5 mass measurements at least every one hour.
All samplers had to be capable of automated operation over a
period of 24 hours, with active control of flow rates. Last, all
selected samplers had to be either commercially available or in
the final prototype stage of their design.
Based on these criteria, five separate PMi 0-2.5 measurement
approaches were selected for evaluation in the 2003 and 2004
field studies. Table 2-1 lists each sampler used in this study, its
manufacturer, its inlet type, its inlet flow rate in actual liters per
minute (1pm), and the number of replicate samplers used at each
sampling site. For the filter-based samplers, the filter
composition is listed along with the species to be determined
during the filter's post-sampling gravimetric and/or chemical
analysis. Due to funding constraints, not all the collected filters
could be chemically analyzed. Instead, a representative subset of
archived filters from each site was selected for chemical analysis
based on the review of the comparative mass concentration
results.
The voluntary participation and involvement of the PMi 0-2.5
sampler manufacturers during these field studies was a critical
component of studies' success. With the exception of the PM2 5
and PM10 FRM samplers that were supplied by EPA, all field
samplers in this study were supplied by their respective
manufacturers. The supplied samplers all represented the latest
models of each design and were equipped with the most current
hardware, firmware, and software. All manufacturers supervised
installation and calibration of their respective samplers during
the initial shakedown tests conducted in Research Triangle Park
(RTP), NC and provided technical reviews of SOPs written for
the instrument's setup, calibration, and operation. Each
manufacturer was also provided the opportunity to visit each
field site during site setup in order to verify the working
condition of their samplers. At the completion of sampling at
each field site, each manufacturer was supplied their respective
field data in order to ensure that their sampler data was being
properly retrieved from the instrument, correctly analyzed, and
correctly interpreted.
Collocated PM2.5 and PM10 FRM Samplers
In the first PM10-2.5 measurement approach, commonly referred to
as the difference method, a designated PM2 5 FRM sampler is
collocated with a designated PMKI FRM sampler. For accurate
determination of PM10-2.5 concentrations, the PM10 sampler is
simply a designated PM25 FRM with its WINS fractionator
replaced by a straight downtube (Figure 2-1). Both samplers are
LJ
i
PM1
PM1
pm2.
r
PMC fraction
removed in WINS
PMn = PM10-PM25
Figure 2-1. Schematic diagram of the FRM samplers used in
the PM-i0-2.5 difference method.
3
-------
Table 2-1. Inventory of Samplers Used in the 2003 and 2004 Multi-Site Performance Evaluations
Measurement Method
PM
Metric
Sampler
Manufacturer(s)
Inlet Type
Inlet Flow
Rate (alpm)
# Used
Filter
Composition
Species Analyzed
Integrated FRM
PM10
BGI, R&P, AND
Std. PM10
16.7
3
Teflon
Mass, sulfate,
nitrate, metals
Integrated FRM
PM10
BGI
Std. PM10
16.7
1
Quartz
EC, OC
Integrated FRM
PM2.5
BGI, R&P, AND
Std. PM10
16.7
3
Teflon
Mass, sulfate,
nitrate, metals
Integrated FRM
PM2.5
AND
Std. PM10
16.7
1
Quartz
EC, OC
Integrated Dichot,
sequential
PM2.5,
PM10-2.5
R&P
Std. PM10
16.7
3
Teflon
Mass, sulfate,
nitrate, metals
Integrated Dichot,
manual (Phoenix 2004
only)
PM2.5,
PM10-2.5
Sierra-Andersen
Std. PM10
(non-
louvered)
16.7
2
Teflon
Mass, sulfate,
nitrate, metals
Integrated Dichot,
sequential
PM2.5,
PM10-2.5
R&P
Std. PM10
16.7
1
Quartz
EC, OC
Coarse TEOM
PM10-2.5
R&P
Std. PM10
(modified
for 50 Ipm)
50.0
3
Glass fiber
none
Beta Attenuation
PM2.5,
PM10-2.5
Kimoto
Std. PM10
16.7
3
Polyfon
none
Time of Flight (APS)
PM10-2.5
TSI
Std. PM10
16.7
2
Total = 22
none
none
4
-------
installed, calibrated, operated, and analyzed using standard PM2 5
protocols. The two samplers thus have identical inlet aspiration
characteristics, produce identical PMKi fractions, and collect
aerosol at the same face velocity through the same filter media.
At the completion of concurrent sampling periods, the PMi 0-2.5
concentration is calculated as the numerical difference between
the measured PMKI concentration and the measured PM25
concentration. Due to its fundamental measurement principle,
the difference method was used as the basis of comparison upon
which to evaluate the performance of the other PMi 0-2.5 samplers
in the study. For purposes of this report, data collected using this
method is termed PM10-2.5 FRM data.
In this study, a designated PMi0-PM2 5 FRM pair was used from
each of three separate sampler manufacturers: Thenno-Andersen
(AND), BGI, and Rupprecht and Patashnick (R&P). Each of
these six FRM samplers were operated with preweighed Teflon
filters for subsequent gravimetric and ion chromatography (IC)
or x-ray fluorescence (XRF) analysis. A fourth set of PM2 5 and
PM10 FRM samplers was used and both samplers were equipped
with a quartz filter to enable subsequent thermal optical
measurement of the aerosol's elemental carbon (EC) and organic
carbon (OC) constituents. In this study, the prefired quartz filters
were not analyzed gravimetrically but were archived under cold
conditions for subsequent EC/OC analysis.
R&P Model 2025 Sequential Dichotomous
Sampler
The Model 2025 dichotomous sampler (dichot) was designed to
provide integrated measurement of both fine and coarse fractions
of a PM10 aerosol. As depicted in Figure 2-2a, the sampler
actively provides volumetric flow control through a standard
16.71pm PM10 inlet. Following the aspirated aerosol's
fractionation in the inlet's internal fractionator, the resulting
PM10 aerosol enters a virtual impactor where the aerosol is then
split into major and minor flow streams. Ideally, the major flow
(maintained at 15 1pm) is intended to collect only the PM2 5
fraction of the PMi 0 aerosol while the minor flow (maintained at
1.7 1pm) is intended to collect only the PMi 0-2.5 fraction of the
PM10 aerosol. In practice, however, this size fractionation is
never ideal, and 10% of the PM2 5 mass theoretically deposits
onto the PMi 0-2.5 filter. The presence of these fine particles is
numerically accounted for during subsequent calculation of the
PMi0-2.5 concentration. Assuming that particle losses within the
instrument are negligible, the sum of the measured PM2 5 and
PM10-2 5 concentrations provide a measure of the ambient
aerosol's PMKI concentration.
The Model 2025 sequential dichot allows unattended, multi-day
operation through the use of a filter exchange mechanism (Figure
2-2b) for transferring filter cassettes from a supply tube to the
sampling position, then conducting a post-sampling transfer of
the cassettes to a storage tube. During this study, however, the
multi-day capability of the Model 2025 was not used; supply
magazines were manually loaded with only one cassette shortly
before each test, and the post-sampling cassette was manually
retrieved from the storage magazine shortly after each test.
Procedures for gravimetric and chemical analysis of the Model
2025's filters were identical to those of the FRM's filters.
Four separate R&P sequential dichotomous samplers were used
during the 2003 and 2004 field tests, three of which were
equipped with Teflon filters while the fourth was equipped with
prefired quartz filters to enable determination of elemental and
organic carbon components of the ambient aerosol.
Kimoto Electric, Model SPM-613D
PM-10 inlet
(16.7 lAnin)
1.7 Wmin
Coarse EE
(a)
Dichotomous
Splitter
I Filter E
Cassettes
IS Irtnin
3 Fins
5 l/min
20 l/min
Flow
Flow
Controller
Controller
*8
Pump
Top
View
tin
(b)
Front
Viev/
Supfiiy
Storann
Tube
Figure 2-2. Schematic diagram of flow system (a) and sample
exchange mechanism (b) of the R&P model 2025 sequential
dichotomous sampler.
5
-------
Dichotomous Beta Gauge
Manufactured by Kimoto Electric Co., LTD., the Kimoto SPM-
613D dichotomous beta gauge (Figure 2-3) is designed to
provide near real-time measurement of both the fine and coarse
fractions of the PMn, aerosol. Similar to the R&P Model 2025
dichot, the SPM-613D aspirates the ambient aerosol through a
standard 16.7 1pm inlet and introduces the fractionated PMio
aerosol into a custom-designed virtual impactor. The virtual
impactorinthe SPM-613D lias different dimensions than that of
the R&P design and operates its major and minor flow rates at
slightly different flow rates. 15.4 1pm and 1.3 1pm, respectively.
Flow control in the two SPM-613D channels is monitored using
separate mass flow sensors. The system's flow control system,
however, is designed to maintain the calibrated mass flow rate
and thus does not maintain true volumetric flow rates through the
inlet at actual ambient temperature and pressure conditions. By
conducting flow rate calibrations at the sampler's inlet under
actual temperature and pressure conditions, however, the effect
of this lack of volumetric flow control is minimal if ambient
conditions do not differ substantially from those existing during
the flow calibration.
03-08 2003
Figure 2-3.
gauge.
Photograph of the Kimoto model SPM-613D beta
collected on a paper roll composed of low hygroscopicity
polyfon. Following each hour of aerosol collection, the
attenuation of11 Pm beta rays by each channel' s aerosol deposit
is quantified using two separate sets of beta sources and
detectors. Based on previous span calibrations performed by the
user, the theoretical relationship between beta attenuation and
collected aerosol mass is used to estimate the mass of each
separate aerosol deposit. Because beta rays are also attenuated
by condensed water, an external heater is located downstream of
the sampler's inlet and maintains the temperature of the aspirated
airstream above 25 °C. As in the R&P 2025 dichot, numerical
corrections must be made to account for the theoretical mass of
fine particulates contained within the SPM-613D's coarse
channel filter. In the Kimoto design, these mathematical
corrections are made within the system's software. Three
identical SPM-613D beta gauges were used during the course of
the study at all three sampling sites.
Q or—
Figure 2-4.
gauge.
Schematic diagram of the Kimoto model SPM-613D beta
Downstream of the SPM-613D's virtual impactor (Figure 2-4).
the separate fine and coarse flow streams are continuously
R&P Continuous Coarse TEOM Monitor
As designed by Misra, et al. (2001) and licensed to R&P, the
coarse TEOM (Figure 2-5) was designed to provide a near real-
time measurement of PMjpgj concentrations. The instrument
aspirates ambient aerosol at 50 1pm through a custom size-
selective inlet, which was made by modifying a standard 16.7
lpffl size-selective PMiQ inlet by adjusting the internal
dimensions in an effort to provide a 10 um cutpoint at 50 1pm.
Downstream of the inlet, the PMio fraction then enters a custom
6
-------
AhkuJ
r S ¦: i
Dolidtoh
•: pTK-r and
photodetectors
~ -i J]j
n- ' l D:
Id flow control
i^diptgnp
Paricis-t-^
r^csnt- a r
La-aer 1
i, I
TSI Inc. Mode! 3321 Aerodynamic Particle
Sizer (APS)
The final measurement approach used in the 2003 and 2004 field
studies involved the TSI Inc. Model 3321 APS (Figure 2-6a) to
estimate the mass of ambient coarse particles based on their
aerodynamic properties in an accelerating flow stream. To adapt
the 5 1pm APS to field use, a standard 16.7 1pm PMj» inlet was
used in conjunction with a custom designed flow splitter located
downstream of the inlet. In the splitter, a sharp-edged, isokinetic
nozzle extracts a representative sample of the PMKI aerosol for
measurement in the APS. The remaining 11.7 1pm portion of the
PMio aerosol was drawn through a total filter using a
volumetrically controlled vacuum source. The mass of the
aerosol collected on the total filter was not quantified.
Figure 2-5. Photograph of the R&P PM^sTEOM.
virtual impactor whose major and minor flow rates are 48 1pm
and 2 1pm, respectively. In this design, the fine fraction (major
flow) is collected in a replaceable total filter and the collected
fine aerosol mass is not subsequently quantified. Downstream of
the virtual impactor, coarse aerosols in the minor flow stream are
first heated to 50 °C to minimize interferences from particle
bound water and are then deposited in a standard R&P 1400a
Tapered Element Oscillating Microbalance (TEOM). The mass
of the deposited aerosol is then estimated based on the observed
change in vibrational frequency of the TEOM filter during the
collection period. Due to the high flow rate ratio between the
total and minor flows (25 to 1), no correction is made for the
mass of fine particles on the coarse filter in this design. The
PM103,5 mass concentration is then calculated as the measured
coarse mass divided by the volume of ambient air aspirated
during the sampling event. Three replicate R&P coarse TEOMs
were used during this field study in order to determine the
inherent measurement precision of the samplers.
The 5 1pm representative aerosol sample is introduced into the
APS, and the aerodynamic diameter of individual particles is
estimated using time of flight technology, as depicted in
Figure 2-6b. The mass of each particle is then calculated based
on its measured aerodynamic diameter and a particle density
specified by the user. For purposes of this field study, a particle
density of 2 g/cm3 was assumed as representative for the coarse
fraction of PMi0 aerosols. The mass concentration of PMiD.2 5
7
-------
aerosols is then calculated as the sum of the mass of all particles
penetrating the PMjo inlet whose aerodynamic diameters were
greater than 2.5 micrometers. Because the APS is only capable
of resolving particles larger than approximately 0.7 micrometers
aerodynamic diameter, the system is not applicable for
measurement of either PM2 5 or PM10 ambient concentrations
because particulate mass less than 0.7 micrometers is not
quantified.
It should be noted that the primary purpose of incorporating the
two APS units into the field study was to provide ambient
aerosol size distribution information at each site. However,
because the APS's measurement method has legitimate potential
for providing continuous PM^a concentration measurements, it
was evaluated in this study in the same manner as the other
PMkj-3 5 samplers.
R&P Single-Event Dichotomous Sampler
In terms of aerosol aspiration, fractionation, and aerosol
col lection, the R&P single-event dichot sampler is identical in
design and fractionation performance to that of the Model 2025
sequential sampler. Designed in 2004 (Figure 2-7), the single-
event sampler was developed as a smaller, lower-cost unit for
users who do not require unattended, multi-day sampling
capabilities. As in the case of the sequential dichot, the ambient
aerosol is aspirated through a standard PM10 low-vol inlet at a
volumetric flow rate of 16.71pm. The PM10 aerosol that exits the
inlet is then separated into fine and coarse mode fractions in a
virtual impactor at flow rates of 15 1pm and 1.71pm.
respectively. Samples are collected on standard 47 mm filter
sampler remain stationary. The potential for post-sampling loss
of large particles, therefore, is thus minimized in the single-event
dichot design.
Because the single-event dichot was developed in 2004, the
design was not evaluated until the 2005 Phoenix field tests. For
evaluation purposes, the manufacturer provided EPA with two
prototype single-event dichot units.
Sierra-Andersen Model 241 Dichotomous
Sampler
At the request of the State of Arizona Department of
Environmental Regulations, limited tests were conducted with
two Sierra-Andersen Model 241 dichotomous samplers during
the January 2004 Phoenix field tests. Being the first designated
PMjo sampler, the Model 241 (Figure 2-8) has been widely
deployed for routine monitoring of both PM2 5 and PM10.g5. In
addition, the size selective performance of the sampler's 246b
inlet has been fully wind tunnel evaluated and served as the basis
for the 1997 PM2 j FRM's inlet design. The 246b inlet was
designed to efficiently aspirate ambient aerosols and to provide
an internal cutpoint of 10 micrometers aerodynamic diameter at a
volumetric flow rate of 16.7 1pm. In the 241 dichot, P\1
aerosols exiting the inlet are then size fractionated in a virtual
impactor whose design was based on research conducted by Loo
Figure 2-7, Photograph of the R&P single event dichotomous
sampler.
cassettes and are retrieved manually following each sampling
event. Unlike the design of the Model 2025 sampler where post-
sampling cassettes are transported to a storage canister for
subsequent retrieval by the user, the cassettes in the single-event
Figure 2-8. Photograph of the Sierra-Andersen model 241
dichotomous sampler.
and Cork (1988). The separate fine and coarse aerosols (at flow
rates of 15 1pm and 1.7 1pm, respectively) are then collected on
separate 37 mm filters for subsequent retrieval and analysis.
8
-------
Because the Model 241 does not provide active volumetric flow
control, the flow rate of each unit was measured and recorded at
the end of each test. If necessary, adjustments were made
immediately prior to the next test to provide the correct channel
flow rates at actual sampling conditions. The 37 mm Teflon
filters used in the Model 241 were equilibrated, handled, and
analyzed using the same procedures as the 47 mm filters used
during these tests.
BGI frmOMNI Ambient Air Sampler (Filter
Reference Method)
The BGI Omni sampler was designed for monitoring agencies
interested in conducting short-term, saturation sampling at a
relatively low cost. To allow flexible setup and operation, the
Omni system was designed to operate continuously on AC power
or through use of a solar power supply. In conjunction with the
low power requirements of the sampling pump, the 5 1pm flow
rate enables operation up to 48 hours using its built-in storage
battery. A constant volumetric flow rate of 5 1pm is maintained
at the system's inlet using calibrated ambient temperature
sensors, ambient pressure sensors, and flow sensors.
A photograph of the Omni unit is shown in Figure 2-9. Its inlet
was designed by geometrically scaling down the P.M; FRM's
16.71pm inlet, and its aspiration performance has been evaluated
in the laboratory at wind speeds up to 11 km/hr. Unlike the
design of the PM25 FRM's inlet which has a fixed internal
cutpoint of 10 micrometers, the cutpoint of the Omni inlet was
designed to be user-selectable at 1,2.5, or 10 micrometers. This
is accomplished through the use of interchangeable, single-jet
impaction nozzles whose jet diameters were designed to provide
the desired cutpoints. An ungreased impaction plate is designed
to remove particles above the stage's cutpoint. Particles
penetrating the impaction stage collect on a standard 47 mm filter
for subsequent gravimetric and/or chemical analysis. To provide
a measure of an airshed's TSP (total suspended particulate
matter), the system may also be configured without an internal
impaction jet.
Prototype Omni units became available for evaluation onlv after
, the
-two
that were configured as PM10 samplers and two configured as
PM2 5 samplers. Because preliminary tests showed that excessive
bounce of large particles could occur from the ungreased PM2 5
impaction plate, the plate of the PM2$ Omni units was greased
daily during the 2005 Phoenix tests using a high vacuum silicone
grease.
Grimm EnviroCheck Model 1.107 Sampler
The Grimm EnviroCheck Model 1.107 sampler was designed to
measure the size distribution of airborne particles in ambient air
or in occupational settings. The system aspirates the surrounding
aerosols at 1.2 1pm through a small, omni-directional inlet
(Model 170M) equipped with a bug screen. The aspirated
aerosol is not heated and thus is designed to retain the aerosol's
volatile and semi-volatile components. In the instrument's
sensing region, individual particles pass through a flat laser
beam, and the scattered signal is detected at an angle of
90 degrees. Based on the intensity of the scattered signal, each
particle is classified into one of 31 different size channels. Using
signal to mass concentration algorithms developed by the
manufacturer, the system can report ambient aerosol
concentrations as PM, I){ PM2 5, PMia_2,5, or PMlt> All analyzed
particles are collected on a PTFE filter, which can be
subsequently removed and analyzed gravimelrically or
chemically.
All sensitive components of the Grimm 1.107 are housed within
a weatherproof housing (Figure 2-10) whose internal temperature
is maintained at 22 °C. The Grimm 1.107 units first became
available to EPA for the 2005 Phoenix tests and represented the
only continuous instruments that were mounted directly on the
roof of the motor home. All other continuous samplers involved
in the field campaigns were not inherently weatherproof and thus
had to be mounted inside the motor home.
Figure 2-9. Photograph of BGI Omni saturation samjble!Pscs
ufactur
9
Diagram of the
R&P Dichotomous TEO
Tli
by Figure 2-10. Photograph of the G i
fin
TH
sta
ex
a -
ids
inn
co
-------
goal is to measure both the non-volatile and volatile components
of ambient aerosol. Every six minutes, the aerosol of interest is
collected by a standard TEOM mass sensing unit and the mass
concentration quantified. Because loss of volatile and semi-
volatile aerosol components is inherent to the TEOM's use of a
heater to remove particle-bound water, this estimated mass
concentration often underestimates the actual concentration of
the aerosol. To correct for this inadvertent loss of aerosol mass,
a switching valve in the FDMS unit periodically diverts the
aerosol flow stream into a purge filter conditioning section,
which purges the airstream at a regulated temperature of 4 °C.
The resulting purged airstream then passes through the TEOM
mass sensor, and the change in mass is noted during the 6 minute
time period. Any decrease in mass concentration measured
during this time period is numerically added to that mass
concentration measured for the non-purged aerosol sample. In
the dichotomous TEOM, two separate TEOM sensing units are
used to quantify both the fine particle concentration and the
coarse particle concentration. As in the case of the sequential
dichot and the manual dichot, the theoretical presence of fine
aerosols within the coarse aerosol fraction must be numerically
accounted for during calculation of PM25 and PM10.2.5
concentrations.
Two prototype dichot TEOM units were made available to EPA
by R&P for the Phoenix 2005 tests. Similar to the coarse TEOM
units, the dichot TEOM units were installed inside the motor
home with their downtubes extending through the roof up to their
external sampling inlets.
10
-------
Chapter 3
Site Setup and Operating Procedures
All field and laboratory activities associated with this study were
conducted by RTI International under EPA contract 68-D-00-
206. Prior to conducting the study. RTI International developed
a Quality Assurance Project Plan (QAPP) that encompassed all
aspects of the study's field and laboratory activities. The QAPP
was subsequently reviewed and approved by QA personnel
within EPA's Office of Research and Development (ORD) prior
to initiation of the study. All field and laboratory operations of
the study were also reviewed and approved during a
comprehensive Systems Audit conducted by ORD QA personnel
prior to the field tests.
The multi-site performance evaluations of the 20 separate field
samplers involved in the 2003 and 2004 studies presented a
unique logistical challenge. With the exception of the FRM
samplers and the R&P dichots. none of the other samplers have
weather enclosures and must thus be protected from the elements
during sampling. To enable efficient transportation of all field
equipment and to house the field samplers, a 25 foot long motor
home was adapted for use in this study. The twelve FRM and
R&P dichot samplers were installed either on the roof of the
motor home or on a 10' by 10' auxiliary platform positioned
immediately adjacent to the motor home. The remaining eight
PMi 0-2.5 samplers were installed inside the motor home with their
downtubes extending through the roof of the motor home and
attached to their respective inlets. Within + 0.2 m. all inlets were
at the same 5 m elevation above ground level.
The motor home' s environmental controls maintained the interior
temperature at 23 °C + 3 °C during all field tests. Per compliance
testing requirements, the inlets of all samplers were installed 2 m
above the sampling platform and all samplers were spaced
horizontally at least 1 m apart from each other. At each site, the
motor home and auxiliary platform were free of nearby
obstructions that might adversely influence the spatial uniformity
of PM10_2 5 concentrations. Figure 3-1 is a photograph of the
sampling setup at the Gary, IN sampling site.
Figure 3-1. Photograph of the PM s sampler evaluation
platform at the Gary, IN site.
Before each field test, all samplers were cleaned and leak-
checked. Each sampler was then calibrated for volumetric flow
rate, ambient temperature, and ambient pressure measurement
using a calibrated transfer standard (BGI DeltaCal). For
calibration of the 50 1pm flow rate of the R&P coarse TEOM, a
BGI TriCal was equipped with a 55 1pm capacity flow module
that had been specifically designed and fabricated for this
purpose. Following the calibration of each instrument, a
performance audit was conducted using a separate audit device,
and any necessary adjustments were made to the instruments. In
addition to the initial audit conducted at each field site,
performance audits were also conducted following Run 15 and
Run 30. Field blank tests of the filter-based samplers were
conducted at the same frequency as that of the performance
audits. Field blank values were not used to correct the measured
PM concentrations but served as an indicator of filter mass
changes during all phases of filter handling and transport. Per
the field study's QAPP, the occurrence of excessive field blank
values would have triggered a systems audit and appropriate
corrective actions would have been taken.
At each sampling site, 30 daily, 22-hour tests were conducted
from 11 am (local time) to 9 am of the following morning. The
two-hour interval between successive tests enabled the site
11
-------
operator sufficient time for sample changeover, data recording,
and minor maintenance while still allowing for daily sampling.
Typically, a 45-day test period was required to complete site
setup, 30 days of sampling, and site shutdown.
Gravimetric analysis of the filter-based samplers' Teflon filters
was conducted both in the EPA weighing facility in RTP, NC
and at each sampling site. At each site, one room of a two-room
hotel suite was set up and used for all filter conditioning,
weighing, and handling operations. The proper temperature and
relative humidity within the room was maintained through
appropriate use of the room's AC unit in conjunction with use of
in-room humidifier and dehumidifier systems. In RTP,
presampling filters were equilibrated and preweighed in an
environmentally controlled chamber whose temperature and
relative humidity setpoints were 22 °C and 35%. respectively.
All filter weighings were conducted using a Calm C-44
microbalance that had a readability of 1 f ig and a capacity of 5 g.
The analytical balance was tared and calibrated prior to each
weighing session, and Class 1 calibration weights were used
during each session to verify the balance's internal calibration.
In order to increase the confidence in the gravimetric analysis,
100% replicate weighings (with a 5 fig reweigh threshold) were
used for each filter during all preweighing and postweigliing
operations. Quality control also included the use of three
laboratory' blank filters during each weighing session. At the
completion of the preweighing in RTP, the filters were loaded in
sampling cassettes and the cassettes were sealed with metal
endcaps, and the sealed cassettes placed in Thermo-Andersen
cassette canisters (Figure 3-2). The canisters were then shipped
to the field site in coolers designed to maintain post-sampling
filters at temperatures below 2 °C.
facility setup within the hotel room. Through careful adjustment
of the room's environmental controls, site personnel were able to
maintain the room's weighing conditions within allowable
temperature and relative humidity limits. Presampling and
postsampling site weighings were conducted using a Sartorius
MC5 microbalance with the same capability as the Calm
microbalance used for the RTP weighings. Identical weighing
protocols were used at all field sites and at the RTP weighing
facility. Once postsampling filters were weighed at the site, they
would be shipped to RTP under cold conditions for final
postweigliing and subsequent archiving. Conducting filter
weighing at the field site enabled faster determination of test
results than could be obtained if samples were shipped back to
RTP. Conducting filter weighing at the site and at RTP also
enabled measurement of particle losses that might occur during
shipping. Lastly, site weighing provided valid test results in the
event that a cooler might be inadvertently damaged or lost during
its shipment back to the RTP weigliing facility. As will be
presented, property observing sampling handling, shipping, and
analysis protocols typically resulted in excellent agreement
between gravimetric results obtained at each field site's weigliing
laboratory versus results obtained at the RTP weigliing
laboratory.
Figure 3-2. Photographs of the cassette storage canisters and
temperature-controlled coolers.
Upon receipt of the preweighed filters from RTP, field personnel
would then unpack and equilibrate the filters in the weigliing
12
-------
Chapter 4
Site Characteristics
Following the initial installation and evaluation of the PMi 0-2.5
samplers in RTP, NC to verify the proper operating condition of
the samplers and to finalize operating protocols, successive field
tests were conducted in Gary, IN (2003), Phoenix, AZ (2003,
2004, and 2005), and Riverside, CA (2003). The following
section will provide a description of these three sites along with
the meteorological conditions and aerosol characteristics
encountered during each site's field tests.
Gary, IN (March - April, 2003)
The Gary, IN site was selected as representing a midwest
industrial city where primary PM10-2.5 aerosols are predominantly
generated by industrial activity rather than by wind blown soils.
Selection and setup of the Gary, IN sampling site was made in
cooperation with personnel from the Indiana Department of
Environmental Management. This State and Local Air
Monitoring site (AIRS # 18-089-0022, N41° 36.391', W87°
18.308') is located approximately 2 km south of Lake Michigan
and is immediately adjacent to the property line of a steel mill.
Nearby sources of emissions include the steel mill, which was
located approximately 0.7 km northwest of the site, anda0.5 km
long open coal pile that was located approximately 0.5 km
northeast of the site.
The 30 days of testing at the Gary site were conducted from
March 6 to April 7, 2003. Weather at the site was typically
cloudy, windy, and cold, and only one rain event occurred during
the study period. Temperatures at the site ranged from -15.1 °C
to 27.8 °C and a mean daily site temperature of 4.6 °C was
recorded.
As measured by the three collocated FRM samplers, PM25
concentrations measured at the Gary site during the tests ranged
from 10.3 (ig/m3 to 46.9 (ig/m3, with a measured mean of
22.8 (ig/m3. Excellent inter-manufacturer agreement was
observed among the filter-based PM25 FRM samplers, as
expressed by the coefficient of variation (CV) of 1.5%. It should
be noted that the inter-manufacturer precision calculated for all
FRM data throughout this report is data representing three
different manufacturer's reference method samplers. Because
each manufacturer has different fabrication facilities, uses
different components, and uses different flow control strategies,
it is reasonable to expect that intra-manufacturer precision of the
reference method samplers may be even better than the reported
inter-manufacturer precision for the PM25, PM10.2.5, and PM10
metrics.
As expressed by a coefficient of variation of 2.4%, excellent
inter-manufacturer agreement was also observed for the PMi0
FRM measurements. PMi0 concentrations measured during the
tests ranged from 22.6 (ig/m3 to 85.0 (ig/m3, with a measured
mean of 42.6 (ig/m3. PMi0.2.5 concentrations (expressed as the
numerical difference between collocated PMi0 and PM2 5 FRM
measurements), ranged from 4.5 |ig/m3 to 58.1 (ig/m3 with a
measured mean of 19.9 (ig/m3. Inter-manufacturer precision of
PMio-25 concentrations was determined to be 5.7% CV. As
indicated by a mean PM25 /PM10 ratio of 0.55 during the 30
sampling events, slightly more than one-half of the site's PMi0
aerosol was associated with PM2 5 aerosols. PM2 5 /PMi0 ratios
ranged from 0.32 to 0.83 during the 30 days of testing, indicating
that the size distribution of ambient aerosols was quite variable
during the month-long field tests. Predominant winds from the
direction of the nearby steel mill typically contributed to PM2 5
concentrations at the site, while winds predominating from the
direction of the open coal piles resulted in measurement of high
PMio-25 site concentrations.
Filter weighing at the Gary site began with Run 5 filters. As
indicated in Figure 4-1, excellent agreement was observed
between PMi0-25 concentrations based on the site weighings
versus the RTP weighings during Runs 5 through 30. The filter
shipping and handling protocols designed for the study,
therefore, appeared to result in negligible PM2 5 or PM, 0 particle
loss from the FRM filters during their transport from the field
site to the RTP weighing facility.
13
-------
M1025 FRM MEASUREMENTS - GARY vs RTP WEIGHING
Gary, IN (March - April, 2003)
RTP/Gary = 1.00
Figure 4-1. Timeline of Gary, IN PM10-2.5 concentration showing level
of agreement between site weighing and RTP, NC weighings.
Phoenix, AZ (May - June, 2003)
The first set of tests conducted in Phoenix, AZ occurred during
early summer of 2003. Phoenix represents a large arid
southwestern urban area with natural and anthropogenic sources
of particulate pollution. Phoenix is a city of approximately 1.3
million people, with a total population of 3.3 million in the
greater metropolitan area. The sampling site was selected
primarily to challenge the coarse particle samplers with high
concentrations of dry, wind blown crustal materials. Through
cooperation with personnel at the Air Quality Division of the
Maricopa County Enviromnental Services Department, the
county-operated Durango Complex sampling site (AIRS # 04-
013-9812, N33° 25.589', W112° 7.085') in the southwestern
portion of Phoenix was selected as an appropriate field site. The
site is located at a county facility complex, which includes a
prison, offices, and general storage. The site location is
impacted by commercial districts to the north and two main
interstate highways located to the east and northeast of the site.
With the predominant wind direction being from the west and
southwest, however, the site is primarily impacted by large
windblown soils originating from non-vegetated, open fields and
from earthmoving activities during nearby commercial
construction.
The month-long field tests at the Phoenix sampling site were
conducted from May 14 to June 15, 2003. Weather at the site
was typically clear, windy, and very hot, and no rain events
occurred during the 30-day study period. Temperatures at the
sampling site ranged from 17.1 °C to 43.5 °C and a mean daily
site temperature of 32.3 °C was recorded.
PM2 5 concentrations measured during the 2003 Phoenix tests
ranged from 6.4 (ig/m3 to 22.0 (ig/m3, with a measured mean of
11.0 (ig/m3. As observed during the Gary tests, excellent inter-
manufacturer agreement was achieved among the filter-based
FRM samplers. As expressed by the coefficient of variation,
mean inter-manufacturer precision for PM2 5 was determined to
be 3.4%. As expressed by a coefficient of variation of 3.1%,
excellent inter-manufacturer agreement was also observed for the
PMio FRM measurements. PMKi concentrations measured
during the tests ranged from 37.1 (ig/m3 to 230.9 (ig/m3 with a
measured mean of 66.7 (ig/m3. PM10_2.5 concentrations
(expressed as the numerical difference between collocated PMKi
and PM25 FRM measurements), ranged from 26.5 (ig/m3 to
209.0 (ig/m3, with a measured mean of 55.7 (ig/m3. Inter-
manufacturer precision of PMi0-2.5 concentrations measured by
the three FRM pairs was determined to be 3.3% CV. As
indicated by the mean PM25/PM10 ratio of 0.18 during the
30 sampling events, a large fraction of the site's PMKI
concentration was associated with PMi 0-2 5 aerosols. PM2 5/PM10
ratios ranged from 0.10 to 0.28, which indicated that coarse
particle mass dominated the PMKI concentrations during each
day of the Phoenix tests. Figure 4-2 depicts the daily dominance
of the coarse particles in the Phoenix area and also shows the
strong agreement observed between the FRM filter weighings
conducted at the sampling site versus those conducted at the RTP
weighing facility.
Site versus RTP FRM Weighing
Phoenix, AZ (May - June 2003)
PM2.5 (RTP)
PM10 (RTP)
PMc (RTP)
PM2.5 (Site)
PM10 (Site)
PMc (Site)
Sample Day
Figure 4-2. Site versus RTP, NC weighing of PM2.5, PM^-2.5, and
PM10 concentrations at the 2003 Phoenix, AZ site.
Riverside, CA (July - August, 2003)
The Riverside, CA sampling site was selected as a West Coast
site where significant secondary fine mode aerosols might be
present in conjunction with primary coarse aerosols. Riverside is
a city of approximately 260,000 people, and industrial,
agricultural, transportation soil, and marine sources all
contribute to airborne particulates. Selection and setup of the
Riverside site was made through cooperation with the University
of California-Riverside (UCR). The monitoring site is located on
the edge of a 420-acre Agricultural Operations Center of UCR
and is operated by the South Coast Air Quality Management
District (California Air Resources Board Site # 33162, N33°
57.679', W117° 20.017'). The agricultural test facility contains
large areas of citrus trees and is adjacent to interstate highway I-
215, approximately 60 km northeast of the Pacific Ocean. Local
sources of ambient aerosols include photochemical pollution
from mobile sources, resuspended road dust, dairy and
14
-------
agricultural farms, wind-blown soils, and sea salt particles from
the ocean.
Field tests were conducted at the Riverside sampling site from
July 23 to August 24,2003. Weather at the sampling site during
the 30 daily tests was typically warm with clear or partly cloudy
skies. No rain events occurred during the Riverside field tests
although morning fog was occasionally observed at the site.
Temperatures at the site ranged from 15.4 °C to 40.4 °C, and a
mean daily site temperature of 25.9 °C was recorded.
As had been experienced during the Gary and Phoenix sites,
excellent inter-manufacturer agreement was observed among the
filter-based FRM samplers. As expressed by the coefficient of
variation, mean inter-manufacturer precision for PM25 was
determined to be 3.1%. Daily PM25 concentrations measured
during the tests ranged from 9.9 (ig/m3 to 32.7 (ig/m3, with a
measured mean of 17.7 (ig/m3. As expressed by a coefficient of
variation of 2.9%, excellent inter-manufacturer agreement was
also observed for the PMi0FRM samplers. PMi0 concentrations
measured during the tests ranged from 27.0 |ig/m3 to 69.3 |ig/m\
with a measured mean of 48.0 (ig/m3. PM10.2.5 concentrations
(expressed as the numerical difference between collocated PMi0
and PM25 FRM measurements), ranged from 16.2 (ig/m3 to
46.1 (ig/m3 with a measured mean of 30.4 (ig/m3. Inter-
manufacturer precision of PMi0.25 FRM measurements was
determined to be 4.1% CV. As indicated by the mean
PM2.5/PM10 ratio of 0.37 during the 30 sampling events,
approximately two-thirds of the site's PMi0 concentration was
associated with PMi 0-2 5 aerosols. PM2 5/PM10 ratios ranged from
0.25 to 0.50 during the 30 days of testing at the Riverside site,
indicating that coarse particles dominated the PMi0 aerosol
during all tests.
Phoenix, AZ (January 2004)
Following the completion of the Riverside, CA field tests, a
second set of tests was conducted at the Phoenix, AZ sampling
site in order to investigate issues raised during the prior Phoenix
field tests. In particular, tests were conducted to investigate
potential loss of large particles within the R&P sequential
dichotomous samplers. Tests were also conducted with a
prototype coarse TEOM constructed by USC to compare its
performance versus two commercial coarse TEOM units
manufactured by R&P.
These follow-up field tests in Phoenix were conducted from
January 10 to January 25,2004. PM2 5 concentrations measured
during the 2004 Phoenix tests ranged from 6.0 (ig/m3 to
22.4 (ig/m3, with a measured mean of 13.2 (ig/m3. As observed
during the previous field campaigns, strong inter-manufacturer
agreement was achieved among the filter-based FRM samplers.
As expressed by the coefficient of variation, mean inter-
manufacturer precision for PM2 5 measurements was determined
to be 2.2%. As expressed by a coefficient of variation of 3.6%,
excellent inter-manufacturer agreement was also observed for the
PMio FRM measurements. PMi0 concentrations measured
during the tests ranged from 14.8 (ig/m3 to 177.5 (ig/m3 with a
measured mean of 52.8 (ig/m3. PMi0.25 concentrations
(expressed as the numerical difference between collocated PMi0
and PM2 5 FRM measurements), ranged from 7.7 (ig/m3 to 95.1
(ig/m3 with a measured mean of 39.5 |ig/m\ Inter-manufacturer
precision of PMi0.2 5 concentrations measured by the three FRM
pairs was determined to be 4.8% CV. As indicated by the mean
PM2 5/PM10 ratio of 0.30 during the 30 sampling events, a large
fraction of the site's PMi0 concentration was associated with
PMio-25 aerosols. PM25/PMi0 ratios ranged from 0.13 to 0.57,
which indicated that coarse particle mass dominated the PMi0
concentrations during each day of the Phoenix tests.
Weather at the site during the first 10 days of testing was
typically warm with clear or partly cloudy skies. Movement of a
cold front into the Phoenix area during Runs 10-11 dramatically
altered the weather and the nature of the aerosol at the sampling
site. Prior to this time, the weather at the sampling site was
warm and the relative humidity was typically in the 30% to 40%
range. During Runs 12-15, however, rain events were common
and the relative humidity at the site was typically in excess of
60%. These rain events resulted in significantly lower PMi 0-2.5
concentrations. As measured by the FRM pairs, the mean
PMio-2 5 concentration during Runs 1-11 was approximately 50
(ig/m3, while the PMi0.2 5 concentration during Runs 12-15 was
only 10 |ig/m3. The size distribution of the ambient aerosol was
also appreciably different during these two time periods. For
Runs 1-11, the PM2 5/PMi0 ratio averaged only 0.24 (with a low
of 0.13) but averaged 0.49 (with a high of 0.57) during Runs 12-
15.
As was observed during previous field tests, the gravimetric
samplers (PM25 FRMs, PM10 FRMs, and R&P dichots) all
provided precise test results as evidenced by coefficient of
variations typically on the order of a few percent. The overall
data capture rate of the filter-based samplers during the 15 day
Phoenix study was 98%.
Phoenix, AZ (April - May, 2005)
Following design modifications to the PMi0.25 samplers
evaluated in the 2003 and 2004 field campaigns, another series
of tests was conducted at the Phoenix sampling site. The
purpose of these tests was to evaluate the effectiveness of the
design modifications to the samplers since the time of the
January 2004 Phoenix tests and to evaluate four newly available
PMio-25 sampler designs. Addition of these new samplers
increased the total sampler count to 28 units. Table 4-1 provides
a description of the various samplers used during the 2005
15
-------
Table 4-1. Inventory of Samplers Used in the 2005 Phoenix Performance Evaluation
Measurement Method
PM
Metric
Sampler
Manufacturer(s)
Inlet Type
Inlet Flow
Rate (alpm)
# Used
Filter
Composition
Species Analyzed
Integrated FRM
PM10
BGI, R&P, AND
Std. PM10
16.7
3
Teflon
Mass, sulfate,
nitrate, metals
Integrated FRM
PM10
BGI
Std. PM10
16.7
1
Quartz
EC, OC
Integrated FRM
PM2.5
BGI, R&P, AND
Std. PM10
16.7
3
Teflon
Mass, sulfate,
nitrate, metals
Integrated FRM
PM2.5
AND
Std. PM10
16.7
1
Quartz
EC, OC
Integrated Dichot,
sequential
PM2.5,
PM10-2.5
R&P
Std. PM10
16.7
2
Teflon
Mass, sulfate,
nitrate, metals
Integrated Dichot,
single-event
PM2.5,
PM10-2.5
R&P
Std. PM10
(non-
louvered)
16.7
2
Teflon
Mass, sulfate,
nitrate, metals
Integrated Saturation
Monitor
PM2.5,
PM10-2.5
BGI
Custom
"Total"
5.0
4
Quartz
EC, OC
Coarse TEOM
PM10-2.5
R&P
Std. PM10
(modified for
50 Ipm)
50.0
3
Glass fiber
none
Beta Attenuation
PM2.5,
PM10-2.5
Kimoto
Std. PM10
16.7
2
Polyfon
none
Time of Flight (APS)
PM10-2.5
TSI
Std. PM10
16.7
2
none
none
Dichotomous TEOM
PM2.5,
PM10-2.5
R&P
Std. PM10
16.7
2
Glass fiber
none
Light Scattering at 90°
Grimm
Custom
"Total"
1.2
3
Total = 28
Teflon
none
16
-------
Phoenix field tests.
Tests were conducted at the Phoenix site from April 27 to May
28, 2005. All site setup, operating procedures, and analysis
protocols were identical to the 2003 and 2004 field campaigns.
During the 30 daily sampling events, PM2 5 concentrations (as
measured by the R&P, BGI, and Andersen FRMs) at the site
averaged 9.9 (ig/m3 and ranged from 4.9 (ig/m3 to 16.6 (ig/m3.
PMi0-2.5 concentrations averaged 46.2 (ig/m3 and ranged from
23.4 (ig/m3 to 122.8 (ig/m3. PM10 concentrations averaged 56.0
(ig/m3 and ranged from 30.1 |ig/m3 to 134.6 (ig/m3. Similar to
FRM results obtained during previous field campaigns, the inter-
manufacturer precision for the PM25, PM10-2.5, andPMio metrics
was strong, as indicated by calculated coefficient of variations of
2.8%, 2.4%, and 1.9%, respectively. The mean PM25/PMi0
concentration ratio for the study was 0.18, indicating that a large
percentage of the PMi 0 aerosol was associated with coarse mode
aerosols. Incidentally, this mean ratio of 0.18 is identical to that
observed during the 2003 Phoenix tests conducted during a
similar time of year. The lowest PM2 5/PMi0 ratio occurred on
the final testing day during which the PMi 0-2.5 concentration was
measured to 134.6 (ig/m3. Thus, over 90% of the PM10 mass
concentration was associated with PMi0-2 5 aerosols during this
particular sampling event. Weather during the study was
generally hot and no rain events were recorded during the 30
days of sampling. In many respects, the ambient aerosol and
sampling conditions during this recent study were similar to
those of the May 2003 Phoenix study.
17
-------
Chapter 5
Test Results
Collocated PM2.5 and PM10 FRM Samplers
Year 2003, 2004, and 2005 Test Results
As previously described, field tests involved the use of four sets
of PM2 5 and PMi0 samplers from BGI, Andersen, and R&P.
Because there exist no absolute standards for ambient particulate
matter, the absolute accuracy of these devices cannot be
determined from these tests. However, the performance of the
three separate manufacturers' samplers with respect to each other
can be calculated. As summarized in Table 5-1, the inter-
manufacturer precision of the FRM samplers was considered to
be excellent for all three metrics (PM2 5, PM10_2 5, and PM10) at
all three sampling sites. Calculating the PMi0.2 5 concentration as
the numerical difference between collocated designated PMi0
and PM2 5 FRMs did not produce any zero or negative PMi 0-2.5
concentrations during the 13 5 sampling events conducted during
the field studies. Because three separate manufacturer's
samplers were used during these events, this represents a total of
405 separate difference method measurements during the five
separate field campaigns.
With the exception of a pump failure and a faulty ambient
temperature sensor connection, few functional problems were
experienced with the eight FRM samplers despite the wide range
of environmental conditions experienced during the study. The
three performance audits conducted at each sampling site
revealed that the FRMs generally maintained their flow rate,
temperature, and pressure calibrations within the required
specifications. Overall data capture rate for the FRM samplers
during the three-site, five field campaign studies was determined
to be 99%.
R&P Model 2025 Dichotomous Samplers
Year 2003 and 2004 Test Results
Only two operational problems were experienced with the four
R&P Model 2025 sequential dichotomous samplers during the
2003 and 2004 field studies. In Gary, a faulty cassette seal in
one of the dichot's coarse channels caused the majority of the
coarse aerosol to bypass the collection filter. As a result, the
coarse particle mass concentration measured by this instrument
was significantly less than that measured by the other collocated
dichots. The data for this sampler's coarse channel was thus
invalidated. The second problem experienced with the Model
2025 dichots occurred towards the latter half of the Phoenix
tests, where significantly low PM2 5 and PMi0 measurements
were obtained by one of the dichots. At the completion of the
Phoenix tests, this behavior was explained by the discovery of a
dense spider web in the dichot's size selective inlet. The PM2 5
and PM10 measurements for this instrument were thus invalidated
for 17 of the 30 sampling events. At all sites, invalid data were
not used to calculate daily aerosol mass concentrations nor used
to estimate intra-sampler precision. Discounting the invalid data
obtained due to the presence of the spider web, overall data
capture rate of the dichots during the study was 98%.
Performance audits of the Model 2025 dichots indicated that they
maintained their flow rate, temperature, and pressure calibrations
within the required specifications.
Table 5-2 summarizes the field performance of the Model 2025
dichots at all three Year 2003 sites in comparison to the
collocated FRM samplers. As the table indicates, excellent intra-
sampler precision was observed for the R&P dichots at all three
sites for all three metrics. As an example, the precision
(expressed as the coefficient of variation) in Gary for PM2 5,
PMi0-2.5, and PMi0 concentrations was determined to be 3.8%,
3.2%, and 1.7%, respectively. The largest coefficient of
variation (4.1%) was observed inPhoenix (May - June, 2003) for
measurement of PMi0-2.5 aerosols.
As Table 5-2 indicates, the PM2 5 concentrations measured by the
R&P dichots in Gary and Riverside agreed almost exactly with
concentrations measured by the collocated FRM samplers. For
18
-------
Table 5-1. Inter-Manufacturer Precision of the Collocated FRM Samplers as a Function of Sampling Site
Metric
Gary, IN
Phoenix, AZ
(May - June, 2003)
Riverside, CA
Phoenix, AZ
(January 2004)
Phoenix, AZ
(April - May, 2005)
PM2.5
PM10-2.5
PM10
1.5%
5.7%
2.4%
3.4%
3.6%
3.3%
3.1%
4.1%
2.9%
2.6 %
4.7%
3.6%
2.8%
2.4%
1.9%
Table 5-2. Performance of the R&P 2025 Sequential Dichot versus the FRM during 2003
Metric
Performance Criteria
Gary, IN
Phoenix, AZ
(May - June, 2003)
Riverside, CA
PM2
PM10-2.
PMn
Dichot CV
Regression Equation
(Dichot vs. FRM)
Coefficient of determination
(R2)
Mean Dichot/FRM Ratio
Dichot CV
Regression Equation
(Dichot vs. FRM)
Coefficient of determination
(R2)
Mean Dichot/FRM Ratio
Dichot CV
Regression Equation
(Dichot vs. FRM)
Coefficient of determination
(R2)
Mean Dichot/FRM Ratio
3.8%
Dichot = 1.01 *FRM
-0.10
0.991
1.00
3.2%
Dichot = 0.87*FRM
+ 0.39
0.968
0.90
1.7%
Dichot = 0.95*FRM
-0.33
0.982
0.94
2.3%
Dichot = 1,24*FRM
-1.6
0.974
1.08
4.1%
Dichot = 0.71 *FRM
+ 4.8
0.982
0.80
2.9%
Dichot = 0.76*FRM
+ 5.7
0.982
0.85
1.3%
Dichot = 1.00*FRM + 0.0
0.995
1.00
1.6%
Dichot = 0.95*FRM +
0.21
0.982
0.96
1.2%
Dichot = 1.00*FRM
- 1.21
0.992
0.97
19
-------
PM25 measurements in Gary, Figure 5-1 depicts the strong
agreement and high correlation between the R&P sequential
dichotomous samplers and the collocated PM2 5 FRMs.
post-sampling transport operations associated with the sequential
design. These modified units were evaluated in Phoenix, AZ in
January 2004 during 15 days of testing.
Dichot versus FRM PM2.5 Concentrations
Gary, IN (March - April, 2003)
Dichot PM25 = 1.01 *FRM PM2.5-O.IO
R2 = 0.991
Dichot PM10_25 = 0.87*FRM PM10.2.5 + 0.3
R2 = 0.968
Dichot PM10 = 0.95*FRM PM10 - 0.33
R2 = 0.982
5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0
FRM PM25 Concentrations (micrograms/m3)
Figure 5-1. Performance of the R&P dichotomous samplers in
Gary, IN versus the collocated FRM samplers.
DICHOT AND FRM TIMELINE (PM10)
Phoenix, AZ (May - June, 2003)
PM10 FRM
R&P DICHOTS
10 12 14 16 18 20 22 24 26 28 30
Sample Day
Figure 5-2. Timeline of R&P Sequential Dichot versus FRM PM1(
concentrations in Phoenix, AZ
In Phoenix, however, the dichots consistently over-predicted the
PM2 5 concentration by about 8%. This over-measurement is
hypothesized to be due to the inadvertent intrusion of coarse
mode aerosols into the fine channel, which lias been known to
occur in virtual impactors (Allen et al., 1999).
The Model 2025 dichots consistently under-measured PMi 0-2.5
concentrations at all three sites although results were highly
correlated (mean R2 equaled 0.977). A high coefficient of
determination at a site indicates that the sampler's performance,
with respect to the collocated FRMs, was very consistent during
the 30 days of testing. For PMi 0-2.5, mean sampler to FRM ratios
at Gary, Phoenix, and Riverside were determined to be 0.90,
0.80, and 0.96, respectively. Summing the dichots' measured
PM2.5 and PM] 0-2.5 concentrations to estimate the PMKI
concentrations, it was observed that mean sampler to FRM ratios
for PM10 in Gary, Phoenix, and Riverside were 0.94, 0.85, and
0.97, respectively. For Phoenix, therefore, 15% of the aspirated
PM10 aerosol mass cannot be accounted for when compared to
the collocated PMKI FRM samplers. The consistency of this
behavior in Phoenix is illustrated in Figure 5-2. Because
collected aerosol deposits in the R&P dichots are analyzed
gravimetrically using the same procedure as that of the FRM
samplers, it was hypothesized that the noted PMKI mass balance
problem in the R&P dichots was either due to internal particle
losses or loss of particles prior to their gravimetric analysis. In
particular, the question was raised whether large particles were
being lost during the mechanical transport of the PMi 0-2.5 filter
cassette from its sampling position to its post-sampling position
in the storage magazine. To address this issue, R&P modified
two of the four dichotomous samplers from automatic sequential
operation to manual operation. The sample cassettes in the
manual samplers, therefore, do not undergo the pre-sampling or
Results from these tests of the R&P dichotomous samplers are
presented in Table 5-3. While the nature of the aerosol during
these January 2004 tests did not produce PMKI measurement
discrepancies as experienced during the May - June, 2003 tests,
8% of the aspirated PMi 0 aerosol was still being unaccounted for
in the sequential dichotomous samplers. Mean dichot to FRM
ratios for PMi 0-2.5 and PMKI were determined to be 0.89 and0.92,
respectively. When the R&P samplers were operated in manual
mode, however, the ratio of the samplers to the collocated FRMs
was 0.99 and 1.00 for PMi0-2.5 and PMKI, respectively. Results
for the dichot performance versus that of the FRMs for these two
metrics were extremely highly correlated as indicated by R2
values of 0.998 and 0.999, respectively.
Due to the lack of active flow control in the Sierra-Andersen
dichot samplers, less confidence can be placed in the results of
these limited field tests. Overall, however, it appears that the
Sierra-Andersen dichot behaves somewhat similarly to the R&P
manual dichot. Mean Sierra-Andersen dichot to FRM ratios for
PM10-25 and PM10 were determined to be 0.95 and 0.97,
respectively. Similarto the virtual impactor in the R&P model, it
also appears that some intrusion of coarse particles into the fine
channel occurs in the Sierra-Andersen's virtual impactor. It may
be the nature of virtual impactor technology, therefore, that some
contamination of the fine particle mass by coarse particles is
unavoidable. Because the resulting bias in measured PM25
concentrations is dependent upon the size distribution of the
PM10 aerosol, significant measurement biases will occur only if
the coarse fraction of PMKI appreciably exceeds the PM2 5
fraction.
As a result of these follow-up tests in Phoenix, it was concluded
that large particles were not being lost in the R&P 2025 dichot
during collection or during post-sampling transport but lost
20
-------
Table 5-3. Comparison of Sequential versus Manual Operation of the R&P 2025 Dichots and Sierra-Andersen Dichot in Phoenix, AZ (2004 Tests)
R&P Sequential Dichot
Metric
R&P Sequential Dichot
Sierra-Andersen Dichot
(Manual Mode)
Slope = 1.10
Slope = 1.07
Slope = 1.13
Intercept = -0.64
Intercept = -0.39
Intercept = -1.17
PM2.5
R2 = 0.991
R2 = 0.990
R2 = 0.988
CV = 2.3%
CV = 3.4%
CV = 1.6%
Mean Dichot/FRM ratio = 1.05
Mean Dichot/FRM ratio = 1.04
Mean Dichot/FRM ratio = 1.03
Slope = 0.81
Slope = 0.96
Slope = 0.91
Intercept = 2.04
Intercept = 0.75
Intercept = 1.45
PM10-2.5
R2 = 0.979
R2 = 0.998
R2 = 0.995
CV = 3.5%
CV = 1.2%
CV = 1.8%
Mean Dichot/FRM ratio = 0.89
Mean Dichot/FRM ratio = 0.99
Mean Dichot/FRM ratio = 0.95
Slope = 0.86
Slope = 0.99
Slope = 0.95
Intercept = 2.35
Intercept = 0.53
Intercept = 0.90
PM10
R2 = 0.981
R2 = 0.999
R2 = 0.996
CV = 2.9%
CV = 1.0%
CV = 1.0%
Mean Dichot/FRM ratio = 0.92
Mean Dichot/FRM ratio = 1.00
Mean Dichot/FRM ratio = 0.97
21
-------
during their automated transport to the storage container. Since
the time that these field tests were conducted, R&P has
conducted additional tests of the samplers in Phoenix, AZ and
concluded that the majority of the particle loss is associated with
the cassette's post-sampling horizontal movement rather than its
subsequent vertical placement within the storage canister.
Sampler Design Modifications
As described in the previous section, the results of the Phoenix
2004 dichot tests, as well as subsequent laboratory testing by
R&P, indicated that loss of large particles was occurring in the
sequential dichot during the post-sampling transfer of the
cassette into its storage canister. To address this issue, the
cassette exchange mechanism of the sequential dichot was
redesigned by R&P to provide a much slower acceleration of the
cassette during its post-sampling movement. The effectiveness
of this design change was subsequently evaluated in the Phoenix
tests conducted in 2005. No other design changes to the
sequential dichot sampler were considered necessary at this time.
Year 2005 Phoenix Test Results
The 2005 Phoenix tests involved the use of two R&P sequential
dichots (with redesigned cassette exchange mechanisms) in
conjunction with two single-event samplers. For both PM2 5 and
PMio-2 5, Figure 5-3 indicates that good agreement was generally
observed between both dichotomous samplers designs and the
collocated FRM samplers. As observed during previous Phoenix
field campaigns, the timeline indicates that the PM25
concentrations during the 30-day study were relatively constant
but that wide variations in PMi 0-2.5 typically occurred. The
respective results obtained with the single-event dichot and
f
—*— FRM PM2.5
-¦-FRM PM10-2.5
—*— Manual Dichot PM2.5
—¦—Manual Dichot PM10-2.5
—*— Sequential Dichot PM2.5
—¦—Sequential DichotPM10-2.5
a/V/V A
Figure 5-3. Timeline of R&P sequential and single-event dichots
versus collocated FRMs in Phoenix, AZ (2005).
sequential dichot will be discussed separately.
Single-Event, Manual Dichotomous Sampler
Table 5-4 provides a summary of the results obtained with the
dichots versus the collocated FRM samplers. On average, the
manual dichots tended to over-measure PM2 5 concentrations by
approximately 10% when compared to the three collocated PM2 5
FRMs. This performance is quite consistent with the results
obtained during the previous two Phoenix campaigns and is
again hypothesized to result from the inadvertent intrusion of a
small fraction of coarse particles into the fine channel of the
virtual impactor. At these PM2 5 concentrations, however, the
over-sampling results in an over-measurement of only
approximately 1 (ig/m3. PM2 5 measurement results for the two
manual dichots versus the collocated FRMs were highly
correlated as evidenced by the calculated R squared value of
0.983. Intra-samplerPM25 measurement precision for the single-
event dichots was strong as indicated by the CV of 1.9%.
For the manual dichot's PMi0.2 5 and PMKI measurements, mean
Dichot/FRM concentration ratios were calculated to be 0.99 and
1.01, respectively. Correlation coefficients forthese two metrics
were calculated to be 0.995 and 0.995, respectively, indicating
extremely consistent measurement performance independent of
concentration. From an aerosol mass balance perspective, the
mean Dichot/FRM PMKI concentration ratios indicated that the
aspirated aerosol is being transported and collected efficiently
within the single-event sampler and that a negligible amount of
particle loss occurs during filter retrieval. These results are quite
consistent with the January 2004 Phoenix tests when the R&P
sequential dichotomous sampler was modified to operate in
manual, single-event mode.
The only operational problem experienced with the manual
dichot was a periodic leak check failure, which was diagnosed as
being associated with the virtual impactor assembly. In
particular, the clips on the impactor's housing could not be
adjusted to provide sufficient compression between the upper
and lower housing. Greasing of the virtual impactor's o-rings
reduced the leak rate but did not eliminate it entirely. Based on
the magnitude of the leak, however, it was judged that the
presence of the leak did not adversely affect the performance of
the virtual impactor nor degrade the quality of the measured data.
Since the end of the Phoenix study, R&P has addressed the leak
issue by replacing the virtual impactor's assembly clips with
multiple set-screws.
Multi-Event, Sequential Dichotomous Sampler
Similar to results obtained with the manual dichot, inspection of
Table 5-4 reveals that the sequential dichot also tended to
overestimate PM25 concentrations by approximately 10%.
Again, this behavior is consistent with the 2003 and 2004 field
studies and reflects the inherent behavior of the virtual impactor
rather than any particular aspect of the sequential design.
Results of the PM2 5 measurements were highly correlated as
22
-------
evidenced by the R squared value of 0.978. For the PMi0-2.5
channel, the Dichot/FRM concentration ratio is 0.93 indicating
that 7% of the aspirated coarse mode mass is being lost; probably
in the post-sampling phase of the sequence. This percentage
loss, however, is substantially reduced from the over 20%
PMi 0-2.5 mass loss
23
-------
Table 5-4. Performance of the R&P Sequential Dichots and Single-Event Dichots versus the FRMs during the 2005 Phoenix Tests
PM2.5
PM10-2.5
PM10
FRM (3 each)
9.8 |jg/m3
46.1 |jg/m3
56.0 |jg/m3
Precision (CV)
2.8%
2.4%
1.9%
Single-Event Dichot (2 each)
10.9 |jg/m3
45.5 |jg/m3
56.6 |jg/m3
Precision (CV)
1.9%
2.4%
2.4%
Mean Ratio to FRM
1.11
0.99
1.01
Slope
1.07
1.01
1.02
Intercept
0.21
-1.40
-0.79
R2
0.983
0.995
0.995
Sequential Dichot (2 each)
10.7 |jg/m3
42.7 |jg/m3
53.5 |jg/m3
Precision (CV)
5.4%
3.0%
2.6 %
Mean Ratio to FRM
1.09
0.93
0.96
Slope
1.01
0.90
0.92
Intercept
0.74
1.10
1.97
R2
0.978
0.997
0.998
Mean Sequential/Single-Event Concentration
Ratio
1.00
0.96
0.97
24
-------
measured during a similar time of year in 2003. From a mass
balance perspective, the PM10 loss within the sequential dichot
was only 4% during these recent tests versus the 15% loss
observed in 2003. It appears, therefore, that the redesigned
exchange mechanism in the sequential sampler is more effective
in preventing large particle loss than that of the original design.
It should also be noted that the ambient conditions in Phoenix
coupled with the aerosol's size distribution and morphology
would tend to maximize any large particle loss. For most
sampling locations, one might expect virtually identical mass
loss between the manual dichot and a sequential dichot equipped
with the new exchange mechanism. Correlations for the
sequential sampler's PMi0-2.5 andPMKI measurements were quite
high, as indicated by the R squared values of 0.997 and 0.998,
respectively.
Overall, the sequential dichot sampler (with modified exchange
mechanism) provided results similar to those of the single-event
dichot design. For PM2 5, PMKI-2.5, and PMKI metrics, the mean
of daily sequential/single-event ratios was determined to be 1.00,
0.96, and 0.97, respectively.
There was one operational issue that occurred regarding the
sequential dichot's modified exchange mechanism. On quite a
number of occasions during the 30-day study, the sample
cassettes would tend to bind within the new mechanism. This
problem occurred during both pre-sampling exchange operations
as well as post-sampling exchange operations. Fortunately, the
site operator was present during these events and could manually
move the cassette into the correct position. As a result, the data
capture rate with the sequential sampler was high. However, for
most routine sampling situations where the operator would not
be present during sample exchange operations, a large
percentage of the data would have been lost. Following the
study, these samplers were returned to the manufacturer, and
R&P has purportedly identified and addressed the cassette
transfer problem.
R&P Coarse TEOM Samplers
Year 2003 and 2004 Test Results
Few operational problems were experienced with the three R&P
coarse TEOM monitors during the three-site study. The
exception occurred during the Riverside testing, where the third
coarse TEOM monitor consistently measured about 17% higher
than the other two coarse TEOM units, which agreed extremely
well with each other. The exact reason for the consistent
difference between the third unit and the other two units is not
known but may have been an operational problem associated
with the TEOM control unit itself. In two successive tests,
exchanging the inlets and virtual impactors between units three
and two did not appear to correct the noted discrepancy.
Table 5-5 summarizes the field performance of the R&P coarse
TEOM monitors at all three sampling sites in comparison to the
collocated FRM samplers. Considering that these are automated
samplers that provide both sampling and mass analysis, strong
intra-manufacturer precision was observed for the three coarse
TEOM monitors during all four field campaigns, as indicated by
measured CVs of 4.4%, 6.6%, 9.4%, and 2.7%, respectively.
The higher CV value for the Riverside data is indicative of the
problem mentioned in the previous paragraph.
At the Gary, Riverside, and Phoenix (2004) field sites, the coarse
TEOMs produced PMi 0-2.5 values that were consistently lower
than those measured by the collocated FRMs. On average, the
coarse TEOMs provided PMKI-2.5 measurements that were 31%,
24%, and 21% lower than the FRMs in Gary, Riverside, Phoenix
(2004), respectively. This underestimation may be partly due to
the fact that the sampler's inlet reportedly provides an internal
cutpoint closer to 9 |im than its 10 |im design cutpoint (Misra et
al., 2003). Note from the table that the data is strongly correlated
for Gary and Phoenix (2004) and that near zero intercepts were
observed for regressions of the coarse TEOMs versus the
collocated FRMs during all field campaigns but Phoenix (2003).
The Gary, IN timeline presented in Figure 5-4 illustrates that the
coarse TEOM monitors track the FRMs well but consistently
provide anunder-measurement of PM10-25 concentrations. Based
on the high coefficient of determination in Gary of 0.983, this
behavior was very consistent as a function of concentration
during the 30-day test period.
R&P COARSE TEOM AND FRM TIMELINE (PM10-2.5)
Gary, IN (March - April, 2003)
PMc (PM10-PM2.5)
R&P Continuous Coarse
Coarse TEOM/FRM = 0.69
E
W
E
U)
o
E
c
o
O
CL
Sample Day
Figure 5-4. R&P coarse TEOM versus FRM PM10_2.5concentrations
in Gary, IN
Better agreement between the coarse TEOMs and the FRM was
observed during the May to June 2003 tests conducted in
Phoenix. For these tests, the coarse TEOMs provided PMKI-2.5
concentrations that averaged 5% higher than those measured by
the collocated FRM samplers. As depicted in Table 5-5,
however, the slope and intercept for the TEOM versus FRM
regression deviated significantly from one and zero, respectively.
The January 2004 Phoenix tests were designed to compare the
performance of the prototype coarse TEOM (constructed by
USC) versus two of the coarse TEOM units manufactured by
25
-------
Table 5-5. Performance of the R&P Coarse TEOM versus the FRM
Metric
Gary, IN
Phoenix, AZ
(May - June, 2003)
Riverside, CA
Phoenix, AZ (January
2004)
PMio-2.5
Slope = 0.68
Intercept = +0.18
R2 = 0.983
CV = 4.4%
Mean TEOM/FRM ratio =
0.69
Slope = 0.79
Intercept = +12.6
R2 = 0.953
CV = 6.6%
Mean TEOM/FRM ratio =
1.05
Slope = 0.77
Intercept = -0.50
R2 = 0.926
CV = 9.4%
Mean TEOM/FRM ratio :
0.76
Slope = 0.77
Intercept = +0.70
R2 = 0.999
CV = 2.7%
Mean TEOM/FRM ratio =
0.79
26
-------
R&P. The purpose of these tests was to ensure that the R&P
manufactured units met all the USC prototype design
specifications. If differences in performance between the USC
coarse TEOM and the R&P coarse TEOMs was observed, the
field tests provided a potential opportunity to identify and correct
any noted problems.
Results showed that the two R&P coarse TEOM samplers agreed
well with each other during Runs 1-11 as evidenced by a mean
CV of 2.0%. During the rain event days (Runs 12-15), however,
CVs averaged approximately 10%. Forthe entire 15-day study,
coarse R&P TEOM 1 measured an average PM10-2.5
concentration of 31.5 |ig/m3 while R&P TEOM 2 measured an
average PMi0.2 5 concentration of 31.0 |ig/m3 During this same
time period, the USC coarse TEOM measured an average
PMi0-2.5 of 30.2 (ig/m3. For the 15-day study, the R&P coarse
TEOMs provided PM10-2.5 concentrations that averaged 21% less
than the collocated FRM pairs. Similarly, the USC coarse
TEOM produced PMi 0-2.5 concentrations that averaged 22% less
than the collocated FRM pairs. This behavior was fairly
consistent throughout the 15 days of testing and did not change
during Runs 12-15. Based on the strong agreement between the
USC coarse TEOM and the two collocated R&P coarse TEOMs,
it was concluded that the R&P coarse TEOMs had been
accurately manufactured by R&P per the USC design
specifications.
Sampler Design Modifications
Subsequent to the 2003 and 2004 field tests, three design and
operating changes were made to the R&P coarse TEOM sampler.
As was discussed in the previous section, the inlet's 9 |im
cutpoint was suspected to be responsible for a significant portion
of the systematic negative bias observed during the 2003 and
2004 field tests. Analysis of the APS size distribution data
collected during the two years of field tests revealed that,
depending upon the size distribution of the aspirated aerosol,
approximately 10% of the PMi 0-2.5 negative measurement bias
could be attributed to the inlet's lower cutpoint. Based on
conventional impactor theory, the diameter of the inlet's internal
nozzle was thus increased from 1.7 cm to 1.9 cm. The effect of
the dimensional change on the inlet's internal cutpoint was then
evaluated by USC in the laboratory using poly disperse
calibration aerosols. Through use of an APS to measure the
aerosol's size distribution before and after the inlet, it was
concluded that the cutpoint of the modified inlet was
approximately 10 |im aerodynamic diameter.
To further validate the inlet's cutpoint, a coarse TEOM was
fitted with the new inlet design and the system was collocated
outdoors with a PM2 5 MetOne BAM and an R&P PMi0 FRM.
During each of five separate tests, the reference ambient PMi 0.2.5
concentration was estimated by subtracting the PM25
concentration measured by the BAM from the PMi0
concentration measured by the PMi0 FRM sampler. Over
PM10-2 5 concentrations ranging from approximately 10 |ig/m3 to
33 (ig/m3, the coarse TEOM to reference PMi0.2 5 concentration
ratios for the five days were 0.99,0.91,1.10,0.95, and 0.92; and
averaged 0.97. Plotting the coarse TEOM's response versus the
reference PM10-2.5 concentration resulted in a slope of 0.93, an
intercept of 0.80 |ig/m3. and an R2 value of 0.95. In conjunction
with the laboratory test results, these results were used to
conclude that the modified inlet exhibited acceptable size
selective performance.
In addition to this physical design change, two operational
changes were made to the coarse TEOM's design. In the
prototype coarse TEOMs, measured coarse mass on the TEOM
element had been divided by 25 to account for the 25:1 ratio
between the inlet's flow rate of 50 1pm versus the 21pm used by
the TEOM sensing unit. To account for theoretical particle
losses within the virtual impactor and associated transport
tubing, the designers decided to modify the conversion factor
from 25:1 to 23:1. This operational modification results in
measured PMi 0-2.5 concentrations approximately 9% higher than
in the prototype TEOMs. The final operational modification
made to the coarse TEOM was a reduction in the TEOM's
operating temperature from 50 °C to 40 °C. Although coarse
mode aerosols were not considered to be hygroscopic, concern
was expressed that a fraction of any volatile or semi-volatile
PM10-2 5 components might be inadvertently lost at the higher
operating temperature, thus accounting for a portion of the
negative measurement bias observed during the 2003 and 2004
field tests.
All three coarse TEOM units evaluated in the 2005 Phoenix tests
used the modified inlet, were operated at 40 °C, and used a 23:1
factor in estimating ambient PMi 0-2.5 concentrations.
Year 2005 Phoenix Test Results
No operational problems were noted with the three coarse
TEOM samplers during the entire 30-day study conducted in
2005. Pre-study, mid-study, and post-study performance audits
indicated that all three units were operating within the required
specifications for total flow rate, ambient temperature
measurement, ambient pressure measurement, and leak rate.
During all three audits, total flow rates were typically within 2-
3% of their 50 1pm design flow rate and primary flows were
typically within 2-3% of their design flow rate of 2.0 1pm. Per
the SOPs for operation of the coarse TEOM, the tapered element
filters on each unit were replaced after 15 days of sampling even
though filter capacity during this test series had only reached
approximately 40%. The data capture rate for the coarse
TEOMs during the 30-day study was 100%. Hourly
concentrations from each TEOM sampler were averaged over the
same 22-hour period during which the integrated FRM filter
sampling was conducted.
During the 30 days of testing, the meanPMio-2 5 concentrations
reported by the three coarse TEOM samplers were 47.4 (ig/m3,
48.1 (ig/m3, and 50.0 (ig/m3, respectively. The TEOM-3 unit
typically measured a higher PMi 0-2.5 concentration than the other
27
-------
two units, although the difference was typically less than 1.0
(ig/m3 (with the exception of Day 20). In general, the level of
agreement among the three coarse TEOM samplers was
generally excellent during each of the 30 test days. On average,
the coefficient of variation for the three coarse TEOM samplers
was determined to be 3.6%. By comparison, the intra-
manufacturer precision of the three PM2 5 FRM samplers was
determined to be 2.4% CV.
A regression of TEOM PMi0-2.5 concentrations versus those of
the collocated PM10-2.5 FRMs is presented in Figure 5-5. Using
the daily TEOM/FRM PMi 0-2.5 ratio as a measure of accuracy,
the minimum ratio measured was 0.94 (Day 5) and the maximum
ratio was 1.17 (Day 19). On average, the mean daily
TEOM/FRM PM10-2.5 ratio during these tests was determined to
be 1.04.
TEOM PM10.2.5 versus FRM PMc
Phoenix, AZ 4/27/05 - 5/28/05
140
130
120
— 110
o ®
35 to 100
5 E
CO 90
8 2
§ " 80
°s I 70
J I 60
s g1 60
O o
w 1 40
30
20
10
0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
FRM PM10.2s Concentration
:iqiim S-S- Regression nf nnarse TFOM versus PM,n„ FRMs
during the 2005 Phoenix tests.
It is worth noting that this 1.04 value compares very closely to
the mean ratio of 1.05 observed in May 2003, which was before
the design modifications. During the 2003 Phoenix tests,
however, values of the slope, intercept, and coefficient of
determination for the coarse TEOM units were determined to be
0.79,12.8 (ig/m3, and 0.95, respectively. Values of the slope and
intercept during the 2003 Phoenix tests, therefore, deviated
appreciably from one and zero, respectively. As indicated by the
regression presented in Figure 20, the values of the slope,
intercept, and coefficient of determination during the 2005
Phoenix tests were 1.09, -1.9 (ig/m3, and 0.982, respectively.
The response of the three coarse TEOMs during the 2005
Phoenix tests thus shows a improvement over results obtained
during the 2003 field campaign during times in which the nature
of the aerosol is expected to be similar. Considering that the
aerosol during the upcoming Birmingham tests is expected to be
more variable in composition and might contain a liigher volatile
content than the Phoenix aerosol, it will be interesting to
compare the Birmingham results to those of the recent Phoenix
tests.
Kimoto SPM-613D Dichotomous Beta-
Gauge Monitors
Year 2003 and 2004 Test Results
No significant operational problems were encountered during
field operation of the Kimoto SPM-613D dichotomous beta
gauge samplers at the three sampling sites. Overall data capture
rate during 2003 and 2004 was nearly 100% at all three sites.
Table 5-6 summarizes the performance of the three Kimoto units
in comparison with the collocated FRM samplers. Inspection of
the table reveals that precision of the samplers was generally
good for all three metrics at all three sampling sites. In general,
liigher intra-sampler CV values (i.e. less precision) were
observed for measurements of PMi 0-2.5 concentrations than for
measurement of PM2 5 concentrations.
At all three sites, the Kimoto SPM-613D units tended to
significantly over-estimate the PM25 concentrations when
compared to the collocated PM25 samplers. For PM25
measurements, the mean sampler to FRM ratio at Gary, Phoenix
(2003), Riverside, and Phoenix (2004) were calculated as 1.26,
1.70, and 1.65, and 1.43, respectively. As illustrated in Figure 5-
6 for Phoenix (2003), this over-estimation by the Kimoto units
was quite consistent versus those of the collocated PM2 5 FRM
/\ -m- Kimoto PM2.5
Vs
Figure 5-6. Timeline of Kimoto SPM-613D versus FRM PM2.5
concentrations in Phoenix, AZ (2003).
samplers. The consistency of the Kimoto' s performance at each
sampling site was high as evidenced by R squared values of
0.949,0.947,0.904, and 0.939, respectively. As was the case for
the R&P dichot, it is hypothesized that this over-estimation
might be due, in part, to the inadvertent intrusion of coarse mode
particles into the sampler's fine mode channel. This hypothesis
is supported by the fact that larger overestimations occur at sites
with the lowest mean PM2.5/PM10 ratios. The fact that the
Kimoto sampler typically provides PMKI concentrations liigher
28
-------
Table 5-6. Performance of the Kimoto SPM-613D Beta Gauge Dichot versus the FRM
Metric
Gary, IN
Phoenix, AZ
(May - June, 2003)
Riverside, CA
Phoenix, AZ (January
2004)
PM2.5
Slope = 1.17
Intercept = +0.16
R2 = 0.949
Slope = 2.03
Intercept = -3.4
R2 = 0.947
Slope = 2.07
Intercept = -6.9
R2 = 0.904
Slope = 1.43
Intercept = -0.11
R2 = 0.939
CV = 7.1 %
CV = 5.9%
CV = 4.1 %
CV = 5.2%
Mean Kimoto/FRM
Mean Kimoto/FRM
Mean Kimoto/FRM
Mean Kimoto/FRM
ratio = 1.26
ratio = 1.70
ratio = 1.65
ratio = 1.43
PM10-2.5
Slope = 0.885
Intercept = +0.34
R2 = 0.978
Slope = 0.920
Intercept = +5.9
R2 = 0.995
Slope = 1.17
Intercept = -2.7
R2 = 0.957
Slope = 0.99
Intercept = +1.66
R2 = 0.994
CV= 10.5%
CV = 9.5%
CV = 5.8%
CV = 9.9%
Mean Kimoto/FRM
Mean Kimoto/FRM
Mean Kimoto/FRM
Mean Kimoto/FRM
ratio = 0.91
ratio = 1.04
ratio = 1.08
ratio = 1.05
PM10
Slope = 1.02
Intercept = +2.5
R2 = 0.987
Slope = 1.02
Intercept = +7.8
R2 = 0.996
Slope = 1.53
Intercept = -10.6
R2 = 0.880
Slope = 1.07
Intercept = +2.9
R2 = 0.998
CV = 4.3%
CV = 7.4%
CV = 3.5%
CV = 7.3%
Mean Kimoto/FRM
Mean Kimoto/FRM
Mean Kimoto/FRM
Mean Kimoto/FRM
ratio = 1.09
ratio = 1.16
ratio = 1.29
ratio = 1.14
29
-------
than the collocated PMKi FRM samplers, however, may indicate
that other measurement uncertainties may be responsible for the
observed PM2 5 measurement bias.
The Kimoto SPM-613D units provide more accurate
measurements of ambient PMi 0-2.5 concentrations than PM2 5
concentrations. For PMi 0-2.5 measurements, the mean samplerto
FRM ratio at Gary, Phoenix (2003), Riverside, and Phoenix
(2004) was calculated as 0.91,1.04,1.08, and 1.05, respectively.
Consistency of this performance during the month-long
sampling at each site is demonstrated by the high coefficient of
determination (0.978, 0.995, 0.957, and 0.994, respectively)
obtained in sampler versus FRM PMi0-2.5 regressions.
Sampler Design Modifications
Following the Phoenix 2004 tests, several design modifications
were made to the Kimoto SPM 613-D dichotomousbeta gauge in
an attempt to improve its overall performance. First, design
changes were made to reduce the size of the PM10-2.5 aerosol
deposit on the paper tape roll. Reducing the diameter from 11
mm to 8 mm effectively reduced the deposition area by
approximately one-half, thus essentially doubling the PM10-2.5
measurement sensitivity. This modification also made it easier
for the user to visually recognize the PM2 5 deposit from the
PMi 0-2.5 deposit on the beta tape in the event that post-sampling
chemical analysis of the collected aerosol was desired.
Modifications to the Kimoto design were also made to convert
the flow control algorithm from mass flow control to volumetric
flow control. Although the 1.3 1pm coarse flow rate is still
controlled using a critical orifice, an in-line mass flow sensor (in
conjunction with the measured ambient temperature and
pressure) allows calculation of the actual coarse channel flow
rate. A separate flow control unit continuously adjusts the total
sampling flow rate to 16.71pm at the inlet's ambient conditions.
The measured channel flow rates, ambient temperature, and
ambient pressure are continuously recorded and available to the
user following each sampling event. Reported PM2 5 and PMi 0-2 5
concentrations are now reported at actual conditions.
Results from the previous four field campaigns indicated that the
SPM 613-D typically measured PMi 0-2.5 concentrations
accurately but produced PM2 5 measurements that were 26% to
70% higher than measured by the collocated PM2 5 FRM
samplers. Because the size selective performance of the
Kimoto's custom designed virtual impactor lias not been
rigorously determined in the laboratory, uncertainties existed
regarding the cutpoint and slope of its fractionation curve. To
address this uncertainty, a new virtual impactor was constructed
based on the Loo and Cork design. Because this design requires
coarse and fine channel flow rates to be maintained at 1.7 1pm
and 15.0 1pm, respectively, a 1.7 1pm critical orifice was
designed to replace the 1.3 1pm orifice.
During the first 15 sampling events of the Phoenix 2005 tests.
both Kimoto units were configured with the custom virtual
impactor used during the previous four field campaigns. For
Runs 16 through 30, the virtual impactor in one of the units was
replaced with the Loo and Cork design and system flow rates
were adjusted accordingly.
The influence of these design changes was evaluated during the
2005 Phoenix field tests.
Year 2005 Phoenix Test Results
Runs 1-15
As mentioned, the two Kimoto units were identically configured
during the first 15 sampling events in terms of their virtual
impactors and channel flow rates. During these 15 days of
testing, the mean PM2 5 concentrations reported by the two
samplers were 12.6 (ig/m3 and 11.9 (ig/m3, respectively. The
Kimoto-1 unit usually reported PM2 5 concentrations higher than
those reported by the Kimoto-2 unit. On average, the Kimoto-1
to Kimoto-2 PM25 concentration ratio was 1.06 for the 15
combined tests. Precision between the two Kimoto samplers
was fairly good during each of the 15 test days and the maximum
CV measured was 13.8%. On average, the coefficient of
variation for the 15 discrete PM2 5 measurements was 5.0%. By
comparison, the intra-manufacturer precision of the three PM2 5
FRM samplers was determined to be 2.8% CV.
Both Kimoto units provided PM2 5 mean measurement responses
that exceeded the collocated PM25 FRM samplers' mean
measurement of 7.7 |ig/m.3, For details, the timeline of Kimoto
and FRM PM2 5 concentrations is presented in Figure 5-7. The
Timeline of FRM and Kimoto PM25 Concentrations
Runs 1-30, Phoenix, AZ 4/27X35 -5/28/05
25.0
20.0
15.0
10.0
5.0
0.0
0
5
10
15
20
25
30
Figure 5-7. Timeline of FRM and Kimoto PM25 concentrations
during the 2005 Phoenix tests. The virtual impactor of Kimoto-2 was
replaced following Day 15.
extent of the over-estimation (as calculated by the mean Kimoto
to FRM concentration ratio) was variable and ranged from a
minimum of 1.29 on Day 12 to a maximum of 2.03 on Day 2.
For the two Kimoto units during the 15 sampling events, the
meanPM25 overestimation averaged 1.61 (i.e., 61% higher than
the collocated PM25 FRMs). The mean Kimoto/FRM ratios
during these 15 tests were calculated to be 1.66 and 1.57 for
30
-------
Kimoto-1 and Kimoto-2, respectively. The slope, intercept, and
coefficient of determination for the Kimoto-1 unit during these
tests were determined to be 1.42, 1.7 (ig/m3, and 0.74,
respectively. For the Kimoto-2 unit, these values were
determined to be 1.29, 2.0 (ig/m3, and 0.73, respectively.
For PMio-25 measurements, both Kimoto units provided
concentration measurements similar to each other, with the
exception of Run 9. For the first 15 days, the mean Kimoto-
l/Kimoto-2 ratio was 1.04. If one chooses to treat Day 9 as a
potential outlier, then the Kimoto-l/Kimoto-2 ratio reduces to
1.03. An inspection of the Runs 1-15 PM10-2 5 timelines in Figure
5-8 reveals that the two Kimoto units tracked the FRMs with the
exception of Day 4. On average, the extent of the Kimoto's
PMio-25 estimation averaged 1.11 and 1.13 for Kimoto-1 and
Kimoto-2, respectively. If Run 4 is considered an outlier,
however, the concentration ratios reduce to 1.10 and 1.05,
respectively. As in the case of the PM25 measurement, the
reason for the bias between the Kimoto units and the FRM units
for Run 4 is not clear and needs to be investigated for potential
influences of meteorology and aerosol chemistry.
Timeline of FRM and Kimoto PM10.25 Concentrations
Runs 1-30, Phoenix, AZ 4/27/05 - 5/27/05
-FRM
- Kimoto-1
Kimoto-2
Figure 5-8. Timeline of FRM and Kimoto PM10-2.5concentrations
during the 2005 Phoenix field tests. The virtual impactor in the
Kimoto-2 unit was replaced following Day 15.
Runs 16-30
Following Run 15, the manufacturer's virtual impactor was
removed from the Kimoto-2 unit and replaced with one based on
the Loo and Cork design. The flow control system within in the
Kimoto-2 unit was modified to provide 15 1pm and 1.67 1pm to
the impactor's fine and coarse channels, respectively. Runs 16
through 30 were then conducted with the Kimoto-2 unit in this
configuration. No physical or operational changes were made in
the Kimoto-1 unit.
If one compares the PM2 5 timelines inFigure 5-7, it appears that
the Kimoto-2 data moves somewhat towards the FRM data
following the virtual impactor modification. The extent of the
improvement is far from complete, however. The mean PM2 5
concentrations measured by the Kimoto units during Runs 16-30
were 16.6 (ig/m3 and 14.9 (ig/m3, respectively, and exceed the
FRMs' mean of 12.0 (ig/m3. For these 15 tests, the mean bias
ratio of the Kimoto-1 and Kimoto-2 units was 1.39 and 1.24,
respectively. Note that the Kimoto-l's overestimation of 1.39
during these tests is noticeably lower than the 1.66 value
observed during Runs 1-15 even though its configuration lias not
changed. Again, this is hypothesized to be due to the improved
response of the Kimoto units to high PM2 5 concentrations versus
lowerPM2 5 concentrations. The slope, intercept, and coefficient
of determination for the Kimoto-1 unit during these tests was
determined to be 1.10, 3.4 (ig/m3, and0.88, respectively. Forthe
Kimoto-2 unit, these values were determined to be 1.15, 1.1
(ig/m3, and 0.83, respectively. Compared to the Run 1-15 data,
these values generally indicate better agreement with the FRMs
during Runs 16-30 than was observed during Runs 1-15.
Figure 5-9 provides a comparison of the Kimoto-2 data
regression for Runs 16-29 versus Runs 16-30. As noted on the
figure, inclusion of the Run 30 data results in a slope, intercept,
and coefficient of determination of 0.80,11.7 (ig/m3, and 0.976,
respectively. For Runs 16-30, the mean Kimoto-2 to FRM
PMio-2 5 ratio was calculated to be 1.08. If one eliminates the
Run 30 data from the regression, the slope, intercept, and
coefficient of determination improve to values of 1.03, 0.2
(ig/m3, and 0.993, respectively. For Runs 16-29, the mean
Kimoto-2 to FRM PMi0-2.5 ratio was calculated to be 1.09. This
comparison illustrates that one data point can dramatically affect
the values of the regression coefficients even though the mean
sampler to FRM ratio may not change appreciably.
Influence of Run 30 Data on Kimoto-2 Regression
Phoenix, AZ 4/27/05 - 5/28/05
Runs 16-30 (Pod. CV =0.381
~ Runs 16-29
¦ Run 16-30
^—Linear (Runs 16-29)
¦ ¦ ¦ Linear (Run 16-30)
FRM PM10.2,5 Cone,
(micrograms per cubic meter)
Figure 5-9. Regression of Kimoto-2 versus the FRM showing the
influence of the Run 30 data on the regression outcome.
TSI Inc. Model 3321 APS
Year 2003 and 2004 Test Results
Few problems were experienced with the two TSI Model 3321
APS units during the course of the 2003 and 2004 field tests.
The exception occurred approximately halfway through the field
sampling in Phoenix (2003) when the response of APS Unit 2
began to deviate substantially from that of Unit 1. During the
31
-------
units' subsequent return to the manufacturer for cleaning, a
circuit board within Unit 2 was diagnosed as faulty and was
replaced. Data from this unit during the second half of the
Phoenix tests, therefore, were not used in comparing the
performance of the APS units to that of the collocated FRM
samplers. Overall data capture rate for the APS units during the
three-city study was 85%.
In the August 2004 report, it was noted that the APS units tended
to track the FRM's fairly well but tended to under-predict the
PMi 0-2.5 concentration by about a factor of two when compared
to measurements provided by the FRM samplers. Mean sampler
to FRM PMio.2 5 ratios for Gary, Phoenix (2003), Riverside, and
Phoenix (2004) were calculated in 2004 to be 0.42, 0.55, 0.58,
0.62, respectively. Since that time, however, it has been
recognized that the APS does not properly account for a non-
spherical particle's shape factor when estimating mass
concentration as a function of size. Because coarse mode
particles are typically generated by mechanical means and are
not as hygroscopic as fine mode aerosols, they tend to be non-
spherical in nature. As a result, mass concentrations of coarse
aerosols reported by the APS tend to be negatively biased.
effect of hygroscopic growth on APS measurement is far from
certain, one hypothesis is that these larger, wetter particles may
become more difficult to transport efficiently to the APS's
sensing zone and are thus not quantified. Laboratory tests
conducted by Volckens and Peters (2005) showed that counting
efficiencies of the Model 3321 were high for large, solid
particles but that efficiencies progressively declined from 75%
for 0.8 micrometer particles to only 25% for 10 micrometer
120
100
80
60
40
20
0
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14 15
16
Proper conversion of APS response to PMi 0-2.5 concentration
requires an estimate of an aerosol's specific gravity and shape
factor. Based on a review of the literature (Stein et al, 1988;
Noll et al., 1988; Lin et al., 1994), a specific gravity of 2.0 and a
shape factor of 1.4 were adopted for all APS calculations. Table
5-7 summarizes the influence of incorporating the shape factor of
1.4 into all APS field data. As opposed to the data reported in
the August 2004 report, mean APS to FRM PM10-2.5 ratios for the
four field campaigns are now calculated to be 0.76, 0.94, 1.00,
and 1.02. In comparing regression results to those reported in
the August 2004, it is observed that the primary influence of the
shape factor's use is movement of regression slopes significantly
closer to unity. Values of the regression's intercepts and
coefficients of determinations did not appreciably change.
Figure 5-10 provides a timeline of the APS's PMi 0-2.5 responses
versus those of the collocated FRMs during the Phoenix 2004
field campaign. Figure 5-11 provides APS versus FRM PMi0-2.5
(PMc) regressions for data collected during the 2003 and 2004
field campaigns. Noted regression coefficients provided in the
figure differ slightly from those in Table 5-7 due to differences
in treatment of apparent outliers.
Sampler Design Modifications
An inspection of Table 5-7 reveals that PMi0-2.5 concentrations
estimated by the APS tended to agree well at the 2003 Riverside
site and the 2003 and 2004 Phoenix sites but less agreement was
observed in Gary, IN. Although there naturally existed
differences is particle size distribution and composition among
these sites, the primary difference was that sampling conditions
in Gary were considerably cooler and more humid than during
the other field campaigns. These conditions in Gary could result
in particle growth due to the uptake of water vapor. While the
Figure 5-10. Timeline of mean APS and FRM PM^-2.5
concentrations during the 15-day Phoenix 2004 field tests.
(A) Gary, IN
y = 0 6G x + 1,6
r* = 0-52
20 40 50 80 100
FRM PM-C m3
_ (C) Phoenix. A2 -1
FRM PM-C. (jg m'
E
3 150 -
: J?
/r
y = 0 90 X - 0 52
r1' 1.00
j I 1 I i_
a. B0
fj
2
1—•—I—'—1—c
(D) Phoamx, AZ - 2
/
/
^ y = 0.99 K* 022
¦IT P «¦ 0.99
A 1 . 1 ¦ i ¦ 1 ¦
Figure 5-11. Regressions of PM10-25 concentrations estimated by
the Model 3321 APS versus those measured by the collocated FRM
samplers. The Phoenix, AZ-1 and Phoenix, AZ-2 designations refer
to data collected in Phoenix during 2003 and 2004, respectively.
—> 1 ' r
» IB) Riverside, CA
v = 1.00 X -2,5
r =0.64
32
-------
Table 5-7. Performance of the TSI APS Model 3321 versus the FRM
Metric
Gary, IN
Phoenix, AZ
(May - June, 2003)
Riverside, CA
Phoenix, AZ (January
2004)
Slope = 0.68
Intercept = 1.7
PM-io-2.5 R2 = 0.53
Mean APS/FRM ratio :
0.76
Slope = 0.92
Intercept = 0.97
R2 = 0.98
Mean APS/FRM ratio :
0.94
Slope = 1.05
Intercept = -2.6
R2 = 0.84
Mean APS/FRM ratio :
1.00
Slope = 1.00
Intercept = 0.27
R2 = 0.99
Mean APS/FRM ratio =
1.02
33
-------
liquid droplets. If this hypothesis is correct, then drying of the
aerosol prior to its introduction into the APS may be one
approach towards improving the PMi 0-2.5 measurement
performance of the APS.
To test this hypothesis, one of the two APS units was modified to
include a smart heater upstream of the APS inlet. The heating
element consisted of a silicone resistance heater tape wrapped
around the downtube's 1.25" external diameter. The moisture
content of the aspirated aerosol was continuously measured and
the heater activated if the flow stream's relative humidity
exceeded 45%. This design modification was made immediately
prior to the 2005 tests conducted in Phoenix, AZ. As was
expected for the Phoenix site at that time of year, however, the
aerosol heater was activated on only a few occasions during the
30-day test period.
Year 2005 Phoenix Test Results
Pre-study, mid-study, and post-study performance audits of the
two Model 3321 APS units indicated that they were both
operating within the required specifications for total sampling
flow rate. Audits of each system's 11.7 1pm auxiliary flow
control system also revealed that they were also routinely
performing within specification. Unlike the previous four field
campaigns, however, functionality problems with the two APS
units occurred during the 2005 Phoenix study. Periodic
inspection of the APS's real-time response often revealed that
the two units were not providing the same measurement results.
In particular, one of the APS units sometimes indicated PM
concentrations orders of magnitude higher than the other unit.
The frequency and magnitude of the measurement problem was
quite variable but affected both APS units. During the few days
when both APS units appeared to provide valid measurement
results, the precision between the two units was good, as
indicated by a coefficient of variation of 7.0%.
For comparing the performance of the two APS units to the
collocated FRM samplers, it was first necessary to eliminate data
that represents apparent outliers, then apply an assumed particle
shape factor and particle density to the remaining data.
Following an inspection of all the raw data provided by the two
units, data were eliminated for 14 of the 30 sampling days. For
the remaining 16 sampling events, a shape factor of 1.4 and a
particle specific gravity of 2.0 were used to estimate the mass
concentration of PMi0.2 5 aerosols. For this limited data, the two
APS units typically under-measured PMi 0-2.5 concentration by
approximately 14%. A regression of the resulting APS versus
FRM data reveals a slope of 0.84, an intercept of 0.55 (ig/m3, and
a coefficient of determination of 0.942. The valid data are thus
highly correlated and have a low intercept but a slope
significantly lower than unity. Because of the overall
functionality issues experienced with both APS units during the
2005 Phoenix tests, however, observations and conclusions made
regarding these limited test results should be considered with
caution.
Since the conclusion of the Phoenix tests, TSI has been actively
investigating the field data and conducting laboratory tests with
both APS units in an effort to identify and resolve the source of
the functionality problem. While the results of these efforts are
preliminary, it appears that incorrect voltage settings were made
to both APS units during their factory calibration immediately
prior to the Phoenix 2005 field tests. As a result, the counting
response of all size channels significantly exceeded calibrated
values when high ambient aerosol concentrations were
encountered. It is expected that the functionality problems with
the two units, however, will be identified and resolved in time
for the fall 2005 tests in Birmingham, AL. Unlike the situation
in Phoenix, the Birmingham tests should provide an opportunity
to evaluate the effect of drying of the aspirated aerosol on the
response of the APS.
BGI frmOMNI Ambient Air Sampler (Filter
Reference Method)
Year 2005 Phoenix Test Results
The Phoenix 2005 study design called for the use of two Omni
units configured to measure PM25 and two Omni units
configured to measure PMi0. However, only three Omni units
were initially available for evaluation due to delivery delays of
the fourth unit. Functionality problems were encountered with
the Omni control units, which reduced the data capture rate. As
a result, the two PM2 5 Omni units were both concurrently
operated on only 12 of the 30 sampling events, and the two PMi0
Omni units were concurrently operated on only 12 of the 30
sampling events. Because PMi 0.2.5 precision calculations require
that both PM25 Omni units and both PMi0 Omni units be
functional at the same time, precision of Omni PMi 0-2.5
measurements could be calculated on only 6 of the 30 sampling
events.
As mentioned earlier, the PM2 5 impaction plates of the Omni
were cleaned and then greased on a daily basis. None of the
Omni's PMi0 stages were greased during the study.
For the 30 days of sampling, the Omni units on average tended to
over-measure PM2 5 concentrations by approximately 7% when
compared to the three collocated PM2 5 FRMs. At these PM2 5
concentrations, however, the overmeasurement represents only
approximately 1 (ig/m3. The Omni'sPM2 5 measurements versus
those of the collocated PM2 5 FRMs are plotted in Figure 5-12.
The plotted data are somewhat scattered, as indicated by the
coefficient of determination of 0.808. Slope of the regression
line is 0.92 and the intercept is 1.46 |ig/m3. Excluding Day 30
(during which very highPMi0.2 5 concentrations were measured)
did not change the correlation appreciably.
The scatter of the data may be a reflection of the relatively low
mass collected by the Omni filters versus the FRM filters. Due
to the Omni's 51pm flow rate, it can be expected that the Omni's
34
Figure 5-12. Regression of the BGI Omni PM2 5 concentrations
versus those of the collocated PM2.5 FRM samplers.
-------
Omni PM25 versus FRM PM2.5
Phoenix, AZ 4/27/05 - 5/28/05
value. Systematic biases in measured PMi 0 concentrations could
also be caused by differences in the Omni's collection efficiency
curve versus that of the FRM's internal PMn , fractionator. In
particular, a cutpoint less than 10 micrometers and/or a sharper
efficiency curve in the Omni unit would tend to reduce
penetration of large particles to the Omni's afterfilter. Whatever
the cause of the bias, the magnitude of the under-measurement is
accentuated by the large particle distribution inherent to the
Phoenix airshed.
PM2 5 mass gain would be only approximately one third that of
the 16.7 1pm FRM sampler. At these relatively low
concentrations, uncertainties in gravimetric analysis could result
in greater concentration measurement uncertainty for the PM2 5
Omni than that of the PM2 5 FRM. For example, if one assumes
a gravimetric measurement uncertainty of 10 micrograms, this
equates to a concentration measurement uncertainty of only 0.5
(ig/m3 for the FRM sampler in a 22-hour sampling period. For
the same measurement uncertainty, however, this results in a
1.5 (ig/m3 uncertainty in PM25 concentrations measured by the
Omni units. As will be discussed, much higher correlations and
better precisions were observed with the PM2 5 Omni units where
appreciably higher mass gains occurred.
Precision (as expressed by the coefficient of variation) between
the two collocated Omni PM2 5 samplers was measured to be
8.8% for the 12 days during which both PM2 5 Omni units were
operating. This value is higher than the 3.2% precision of the
three FRMs during the same 12 sampling events and may be
another indication that low mass gains on the Omni filters can
result in greater measurement uncertainty. This conjecture is
supported by the fact that the PMn , precision for the Omni units
was 3.3% for the 12 days that both PMn, Omni units were
operating. This precision compares favorably with the FRM's
precision of 2.2% during the same 12 sampling events.
For all 30 sampling events, the PMn, Omni units provided PMn,
concentrations lower than the three collocated FRM samplers
(Figure 5-13). On average, the Omni's PMn, concentrations
were ll%lowerthantheFRMsamplers. The greatest difference
(19%) occurred during Run 30 in which the FRM PMn,
concentration was measured to be 134.7 |ig/m3. The slope,
intercept, and R squared values of the PMn, Omni versus the
PM10 FRM were determined to be 0.83, 3.59 (ig/m3, and 0.97,
respectively. If one excludes Run 30 from the regression, then
the slope, intercept, and R squared values are 0.93, -1.79 (ig/m3,
and 0.969, respectively. The Run 30 data point, therefore,
strongly influences the slope and intercept of the PMn, regression
curve.
Systematic biases in measured PMn, concentrations could be
attributed to incomplete aspiration of large particles by the low
flow rate Omni inlet. Since large particle aspiration efficiency
typically decreases with increasing wind speed, correlating Omni
PM10 concentrations with site meteorological data may be of
Because the PMi0-2.5 aerosol comprises such a large percentage
of the PM10 aerosol in the Phoenix area, the Omni units
consistently underestimated the PMi0.2.5 concentrations. On
average, the Omni units under-measured PMi0.2.5 concentrations
by approximately 15% when compared to the collocated FRMs.
The slope, intercept, and R squared values of the PMi0.2.5 Omni
versus the PMi0.2.5 FRM were determined to be 0.81,1.17 (ig/m3,
and 0.949, respectively. If one excludes the Run 30 data from
the regression then the slope, intercept, and R squared values are
0.95, - 4.16 (ig/m3, and 0.932, respectively. The Run 30 data
point, therefore, strongly influences the slope and intercept of the
PM10_2.5 regression curve similar to its influence on the PM10
regression curve.
CV's forthe Omni PMi0.2.5 measurements averaged 5.0%, which
compared favorably to the FRM's PMi0-2 5 CV of 3.5% measured
during the same test days. It should be reiterated, however, that
precision calculations can only be based on the 6 days in which
all four Omni units were operating during the 30 day study.
Since the completion of the 2005 Phoenix tests, BGI has
Omni PM10 versus FRM PM10
Phoenix, AZ 4/27/05 - 5/28/05
140
120
100
|
E
80
i i60
i S
o .2
1
40
20
0
0
20
40
60
80
100
120
140
identified that faulty relative humidity circuitry in the Omni units
was responsible for the periodic functionality problems
encountered in Phoenix. This problem lias subsequently been
Figure 5-13. Regression of Omni PM10 concentrations versus those
of the collocated PM10 FRM samplers.
addressed and no further problems have been encountered with
repaired units. In addition the single-stage PM2 5 impactor of the
prototype Omni units has been replaced with a 5 1pm version of
BGI's sharp-cut cyclone design. This modification will thus not
Omni Regression
35
-------
require the user to clean and prepare the PM2 5 impaction surface
daily as was previously required.
Grimm EnviroCheck Model 1.107
Year 2005 Phoenix Test Results
During the 30 days of testing, the three Grimm Model 1.107
samplers measured average PM25 concentrations of 13.0, 12.8,
and 13.2 (ig/m3, respectively. Precision among the three Grimm
samplers was excellent during each of the 30 test days and the
maximum CV measured was only 2.2%. On average, the
coefficient of variation for the PM2 5 measurements was 1.5%.
By comparison, the intra-manufacturer precision of the three
PM25 FRM samplers was determined to be 2.8% CV.
A timeline of measured Grimm versus FRM PM2 5 concentrations
is provided in Figure 5-14. Inspection of the timeline indicates
that the Grimm samplers tracked the FRMs well but
overestimated PM2 5 concentrations on each of the 30 test days.
The extent of the overestimation (as calculated by the Grimm to
FRM concentration ratio) was variable and ranged from a
minimum of 1.12 on Day 21 to a maximum of 1.86 on Day 10.
For the 30 sampling events, the Grimm's PM2 5 over-estimation
averaged 1.37 (i.e., 37% higher than the collocated PM2 5 FRMs).
Timeline of FRM and Grimm PM^ Concentrations
Phoenix, AZ 4/27/05 - 5/28/05
h\ I
K j\l
A /
/ V/
1 vX/
\ „ J.i
1 \ ;
v\/
A/ \ v m N
V
\/v
-¦-FRM PM2.5
Grimm PM2.5
Figure 5-14. Timeline of Grimm Model 1.107 PM25 concentrations
versus those of the collocated PM25 FRMs.
Regression of the Grimm's PM2 5 measurement response versus
those of the PM25 FRMs indicated that results were well
correlated (R squared = 0.908) but that the slope was 0.83 and
the intercept was 4.80 |ig/m\ The influence of the large
intercept on the PM25 response is particularly important
considering that the mean PM2 5 concentration during these tests
was less than 10 |ig/m\ The upcoming tests in Birmingham
should provide an opportunity to evaluate the Grimm's response
at higher PM2 5 concentrations than were encountered at the
Phoenix sampling site.
During the 30 days of testing, the three Grimm samplers
measured average PMi0-2 5 concentrations of 70.0,73.2, and 68.3
(ig/m3, respectively. With the exception of a few sampling
events, daily precision among the three Grimm samplers was
excellent and averaged 4.1% CV during the 30 day study. By
comparison the intra-manufacturer precision of the three PMi 0.2 5
FRM samplers was determined to be 2.4% CV.
Inspection of the timeline (Figure 5-15) of Grimm and FRM
PM10_2.5 responses indicates that the Grimms again tracked the
FRM fairly well but overestimated PMi0-2.5 concentrations on
each of the 30 test days. The extent of the measurement bias was
variable and ranged from a minimum of 1.08 on Day 13 to a
maximum of 1.84 on Day 4. For the 30 sampling events, the
Grimm'sPMio-2 5 overestimation averaged 1.53 (i.e., 53%higher
than the collocated PMi0.2.5 FRMs). A regression of the Grimm
versus FRM PMi0-2 5 concentrations resulted in slope, intercept,
and R squared values of 1.35,8.4 (ig/m3, and 0.847, respectively.
Unlike the response of the Grimm to PM25 aerosols, the PMi0.2 5
measurement bias is associated more with the regression's slope
than its intercept. Inspection of the Grimm's responses to
PM10-15 concentrations showed no real trend in instrument bias
versus PMi0-2.5 concentration.
The highest measured PMi0-2.5 concentration during the 2005
Timeline of FRM and Grimm PM10.2.5 Concentrations
Phoenix, AZ 4/27/05 - 5/27/05
Figure 5-15. Timeline of Grimm Model 1.107 PM10-2.5
concentrations versus those of the collocated FRMs.
Phoenix study occurred during Day 30. Fortius sampling event,
the three Grimm units reported concentrations of 133.4, 177.6,
and 139.2 (ig/m3, respectively. The response of Unit 2 for this
sampling event is thus significantly higher than that of the other
two Grimm units, though no operational problems were noted by
the site operator. If one chooses to exclude all data from this
sampling event, then the correlation between the Grimms and the
FRMs improves somewhat but results in dramatically different
slopes and intercepts than if the Day 30 data is included.
Specifically, using only Run 1-29 data results in slope, intercept,
and R squared values of 1.87, -13.7(ig/m3, and 0.887,
respectively. As had been illustrated in Figure 5-9 for the
Kimoto-2 unit, excluding the Day 30 data point from the 2005
Phoenix data has a dramatic influence on the regression
coefficients.
36
-------
Because the Phoenix PMKi aerosol is heavily dominated by
coarse mode particles, one would expect that the Grimm's PMKi
response would be more similar to its PM10-2.5 response than its
PM2.5 response. Inspection of the Grimm's actual data reveals
that this hypothesis is correct. During the 30 days of testing, the
three Grimm samplers measured mean PMn, concentrations of
83.0, 86.0, and 83.5 (ig/m3, respectively. As in the case of the
PMi 0-2.5 measurements, precision of the Grimm samplers for
PMi 0 measurement was excellent, as evidenced by the calculated
CV of 3.4%.
Inspection of the Grimm PMKi data indicated that the Grimms
tracked the FRM well but over-estimated PMKi concentrations on
each of the 30 test days. The extent of the measurement bias was
variable and ranged from a minimum of 1.18 on Day 13 to a
maximum of 1.71 on Day 4. For the 30 sampling events, the
Grimm's PMKi5 over-estimation averaged 1.49 (i.e., 49% higher
than the collocated PMKi FRMs). A regression of the Grimm
versus FRM PMKi concentrations resulted in slope, intercept, and
R squared values of 1.37,6.9 (ig/m3, and 0.900, respectively. As
is in the case of the PM10-2.5 data analysis, removing the Day 30
data from the regression resulted in dramatically different
correlation coefficients. Similar to observations of the PMi0-2.5
data, inspection of the Grimm's response to PMi 0 aerosols shows
no real trend in instrument bias versus PMKI concentration.
R&P Dichotomous TEOM Sampler
Year 2005 Phoenix Test Results
No operational problems were noted with the two dichot-TEOM
samplers during the entire 30-day study. Pre-study, mid-study,
and post-study performance audits indicated that both units were
operating within the required specifications for channel flow
rates, ambient temperature measurement, and ambient pressure
measurement. Per our SOPs for operation of the dichot TEOM,
the tapered element filters on each unit's channels were replaced
after 15 days of sampling even though only moderate increases
in filter capacity readings were noted during this time period. At
the completion of the first 15 days of testing, filter capacities of
the units' fine and coarse channels averaged only 38% and 21%,
respectively. At the completion of the subsequent two weeks of
sampling (i.e., following Run 30), filter capacities of the units'
fine and coarse channels averaged 44% and 28%, respectively.
Data was captured for all sampling events with the exception of
tests conducted on May 15th (Run 18). During this test, data
collected by dichot TEOM-1 could not be recovered. Data
capture rate for the dichot TEOMs during the 30-day study was
thus calculated to be 98%.
A timeline of the two dichot-TEOMs' responses versus mean
FRM PM25 is provided in Figure 5-16. Inspection of the
timeline indicates that the two dichot TEOMs generally tracked
the PM2 5 FRMs during the 30-day study. The notable exception
was during Runs 1-3 for dichot TEOM-2, during which times
daily PM2 5 concentration measurements were -7.1, -24.2, and -
7.5 (ig/m3, respectively. Similar measurement problems were
noted with the dichot TEOM-1 during Runs 1-3 although the
magnitude of the under-measurement was not as great as that of
dichot TEOM-2.
Timeline of FRM and Dichot TEOM PM2.s Concentrations
Runs 1-30, Phoenix, AZ 4/27/05 - 5/28/05
25.0
20.0
15.0
10.0
5.0
0.0
-5.0
10.0
15.0
20.0
25.0
30.0
0
5
10
15
20
25
30
Figure 5-16. Timeline of R&P dichotomous TEOM PM2 5
concentrations versus those of the collocated PM2 5 FRMs.
Inspection of the data from the two dichot-TEOMs revealed that
a large percentage of the reported hourly PM2 5 concentrations
were less than zero. Forthe dichot-TEOM-1 and dichot-TEOM-
2, the percentages of negative PM2 5 concentrations were 20%
and 38%, respectively. There appeared to be no discernable
pattern (e.g. time of day) during which negative PM25
concentration values were reported. As can be expected,
however, the preponderance of these negative values adversely
influences the level of agreement between the dichot-TEOMs
and the collocated PM2 5 FRM samplers. If one chooses to use
all 30 days of data, then the meanPM2 5 concentrations measured
by the two dichot-TEOMs were 7.9 and 3.5 (ig/m3, respectively,
compared to the FRMs' mean PM2 5 concentration of 9.9 (ig/m3.
Mean dichot TEOM to FRM ratios for the two units were 0.80
and 0.63, respectively. A regression of mean dichot TEOM
response versus mean PM25 FRM response results in slope,
intercept, and correlation coefficient values of 1.9, -13.1 (ig/m3,
and 0.765, respectively. On average, the dichot TEOM-
l/TEOM-2 ratio was 2.23 forthe 30 sampling events.
If one chooses to treat the Day 1-3 data as outliers, then the level
of agreement between the dichot-TEOMs and the collocated
PM2 5 FRMs improves somewhat. For the Run 4-30 data, the
meanPM2 5 concentrations measured by the two dichot-TEOMs
were 8.7 (ig/m3 and 5.4 (ig/m3, respectively, compared to the
FRMs' mean PM2 5 concentration of 10.2 (ig/m3. Mean dichot -
TEOM-to-FRM ratios for the two units were 0.85 and 0.42,
respectively. A regression of mean dichot-TEOM responses
versus meanPM25 FRM responses results in slope, intercept, and
correlation coefficient values of 1.58, -9.2 (ig/m3, and 0.770,
respectively. On average, the dichot-TEOM-l/TEOM-2 ratio
was 1.58 for the Run 4-30 sampling events.
As opposed to the PM2 5 measurement, where a large percentage
37
-------
of the measured concentrations were negative, negative PMi0.2 5 calculated slope, intercept, and R2 values. For the Run 4-30 data,
concentrations were reported for only 9 hourly sampling events the precision between the two dichot-TEOMs was 6.2% and the
during the 30 days of testing. As depicted in Figure 5-17, there mean TEOM-l/TEOM-2 ratio was 1.03.
also appeared to be better PMi 0.2 5 agreement between the dichot-
TEOMs and the FRMs than was observed for the PM25
measurements. The mean PMi 0-2.5 concentrations reported by
dichot-TEOM-1 and TEOM-2 were determined to be 3 8.1 (ig/m3
and 41.1 (ig/m3, respectively, compared to the FRMs' mean
PMi0-2.5 concentration of 46.2 (ig/m3. Mean dichot-TEOM to
FRM ratios for the two units were 0.85 and 0.89, respectively.
Expressed as the coefficient of variation, the level of precision
between the two dichot-TEOMs was determined to be 5.4% for
PMi0-2.5 measurements. On average, the dichot-TEOM-
l/TEOM-2 ratio was 0.95 for the 30 PMi0-2.5 sampling events.
A comparison between the dichot-TEOMs' mean response
versus the collocated FRMs indicated that results were very
highly correlated (R2 = 0.992) with a low intercept (0.73 (ig/m3)
but that the slope was only 0.85. Inspection of the data showed
that this measurement response was highly consistent with time
and was virtually independent of ambient PMi0.2 5 concentration.
Elimination of the Run 1-3 data from the regression did not
significantly alter the calculated slope, intercept, or R2 values.
For the Run 4-30 data, the precision between the two dichot-
TEOMs was 5.2% and the mean TEOM-l/TEOM-2 ratio was
0.98.
Forthe dichot-TEOMs, PMi0 concentrations canbe calculated as
the numerical sum of measured PM25 and PMi 0-2.5
concentrations. The mean PMi0 concentrations reported by the
dichot-TEOM-1 and TEOM-2 were determined to be 47.0 (ig/m3
and 44.6 (ig/m3, respectively, compared to the FRMs' meanPMio
concentration of 56.0 (ig/m3. Mean dichot-TEOM to FRM ratios
forthe two units were 0.84 and 0.80, respectively. Expressed as
the coefficient of variation, the level of precision between the
two dichot-TEOMs was determined to be 9.7% for PMi0
measurements. On average, the dichot-TEOM-l/TEOM-2 ratio
was 1.05 forthe 30 PMi0 sampling events.
A comparison between the dichot-TEOMs' meanPMio response
versus the collocated PMi0 FRMs indicated that results were
very highly correlated as indicated by the R2 value of 0.950.
Unlike the TEOM's PMi 0-2.5 response, however, the regression
between the TEOMs and the collocated FRMs resulted in a slope
close to unity (0.972) but the intercept was -8.7 (ig/m3. On
average the dichot-TEOMs' calculated PMi0 concentration was
82% of the value measured by the PMi0 FRMs. Elimination of
the Run 1-3 data from the regression only slightly altered the
Timeline of FRM and Dichot TEOM PM10.2.5 Concentrations
Runs 1-30, Phoenix, AZ 4/27/05 - 5/27/05
38
:igure 5-17. Timeline of R&P dichotomous TEOM PM^-2.5
:nnrpntratinns vprsi is thnsp nfthp rnllnratprl FRMs
-------
Chapter 6
Summary
Through coordination with state and local air monitoring
agencies, the Gary, IN, Phoenix, AZ, and Riverside, CA
sampling sites met the study's siting objectives well and
challenged the candidate samplers with a wide range of aerosol
size distributions, aerosol concentrations, and meteorological
conditions. Relatively few operational problems were
experienced with the sampling equipment and the overall data
capture rate during the five separate field campaigns exceeded
95%. Prestudy, midstudy, and poststudy performance audits
conducted at each sampling site revealed that the samplers
typically held their calibrations well during the month-long field
tests. The involvement and cooperation of the various sampler
manufacturers was a key factor in the study's ability to
successfully determine the inherent performance of the samplers.
The filter-based, integrated samplers involved in the study
provided precise test results at all three sampling sites during the
five field campaigns and their overall data capture rate was
approximately 99%. For the FRM samplers, the mean inter-
manufacturer coefficients of variation for PM2 5, PMi 0-2.5, and
PM10 were 2.7%, 4.1%, and 2.8%, respectively. As an example,
for three samplers that provide a mean concentration of
25 ng/m3, a 4% CV would equate to readings of 24, 25, and
26 |ig/nr\ Effective shipping protocols resulted in negligible
particle loss during transport of collected aerosol samples from
each sampling site to the RTP weighing facility. Concentrations
calculated using site weighing data versus the use of RTP
weighing data typically agreed within 1% of each other.
Independent of design (i.e., R&P sequential, R&P single-event,
or Sierra-Andersen single-event), the intra-manufacturer
precisions of the filter-based dichotomous samplers were
excellent for PM25, PMi 0-2.5, and PMi0 measurements. For
example, the coefficient of variations for the three R&P Model
2025 dichotomous samplers during all five field campaigns for
PM25, PMi0-2.5, andPMio measurements averaged 3.0%, 3.1%,
and 2.3%, respectively.
At the Gary, IN and the Riverside, CA sampling sites, the PM2 5
concentrations measured by the R&P dichotomous samplers
agreed almost exactly with those measured by the collocated
PM25 FRM samplers. During all three field campaigns in
Phoenix, however, the dichots typically measured PM25
concentrations that were approximately 10% higher than those of
the PM2 5 FRMs. It is hypothesized that this PM2 5 measurement
bias resulted from the inadvertent contamination of the fine
particle fraction with a small percentage of coarse mode
particles. Because this behavior was independent of dichot
sampler design (i.e., R&P sequential, R&P sequential converted
to manual mode, R&P single-event, or Sierra-Andersen single-
event), this contamination may be inherent to virtual impactor
size fractionation technology. However, because the resulting
bias in measured PM2 5 concentrations depends upon the size
distribution of the PMi0 aerosol, significant measurement biases
will occur only if the coarse fraction of PMi0 appreciably
exceeds the PM2 5 fraction.
During the Year 2003 field tests, the R&P dichotomous samplers
underestimated PMi0.2 5 concentrations at all sampling sites, and
under-measured PM10-2.5 by 20% at the Phoenix site. Mass
balance calculations revealed that 15% of the aspirated PMi0
mass during the Phoenix tests was not accounted for during
subsequent gravimetric measurement of fine and coarse channel
filters. During the 2004 follow-up tests in Phoenix and during
subsequent laboratory tests by the manufacturer, the loss of
coarse mode aerosols during the samplers' automated post-
sampling movement of the coarse particle cassette to the sample
storage position was identified as the source of the measurement
bias. As demonstrated during the 2005 Phoenix field tests, a
redesigned cassette exchange mechanism effectively reduced this
coarse particle loss from 20% to 7%. Because the dry, bouncy
nature of windblown coarse particles tends to maximize the
extent of the particle loss, it is expected that this modified
exchange mechanism will result in minimal coarse particle loss
in most sampling locations. In R&P's new single-event
dichotomous sampler, the sampling cassettes remaining
stationary during all phases of sampler operation. As a result,
39
-------
dichot to FRM ratios for PMi0-2.5 concentrations and PMi0
concentrations were measured to be 0.99 and 1.01, respectively,
during the 2005 Phoenix tests.
With the exception of the problem noted during the Riverside
tests, excellent inter-manufacturer precision of the R&P coarse
TEOM samplers was observed at all three sampling sites, and no
operational problems were encountered with the samplers.
During the January 2004 Phoenix tests, very close agreement
was observed between the USC prototype coarse TEOM versus ¦
the two coarse TEOMs manufactured by R&P, indicating that the
USC prototype's basic design had been faithfully duplicated by
R&P. However, with the exception of the 2003 Phoenix tests,
the coarse TEOM tended to underestimate the PM10-2.5
concentration by as much as 30%. The high correlation between
the coarse TEOMs' response versus the collocated FRMs
indicated that this performance was very consistent from one
sampling event to another. Following the 2004 Phoenix tests,
modifications were made to the coarse TEOM design including
increasing the inlet's cutpoint to approximately 10 |im
aerodynamic diameter. The 2005 Phoenix follow-up tests
showed excellent intra-sampler precision (CV = 2.4%) and
improved correlation coefficients when compared to those
obtained during the 2003 Phoenix field tests. However, because
the mean coarse TEOM to PMi 0-2.5 FRM ratio of 1.05 was
identical during the 2003 and 2005 Phoenix tests, it is not yet
certain if the 30% under-measurement bias (observed at the Gary
and Riverside sites) has been properly addressed by the coarse
TEOM's design modifications. The fall 2005 sampler evaluation
tests in Birmingham, AL should provide an additional
opportunity to assess the effectiveness of these design changes.
During the 2003 and 2004 field tests, the Kimoto SPM-613D
samplers provided precise, highly correlated test results at all
three sites for PM25, PM10-2.5, and PMi0 measurements.
Although performance varied by site, the Kimoto units generally
provided PM10-2.5 measurements within 10% of those of the
collocated FRM samplers. However, the SPM-613D units
consistently provided PM2 5 concentrations significantly higher
than the collocated PM2 5 FRM samplers. As an example, the
mean over-estimation in PM2 5 concentrations at the Phoenix site
was 70%. The magnitude of the Kimoto's PM2 5 bias suggested
that possible intrusion of coarse mode particles into the fine
channel only partly accounted for the bias. The fact that the
sampler produced PM10 concentrations higher than the collocated
PM10 FRM samplers also suggested that the problems were
associated with the aerosol's analysis, rather than with regard to
aerosol sampling and transport. Following the 2004 tests,
modifications were made to the Kimoto's design that included
reduction of the PM2 5 deposition area to increase measurement
sensitivity, and change in mass flow control to volumetric flow.
In addition, a Loo and Cork virtual impactor design was
evaluated as a replacement to the system's custom virtual
impactor. However, as indicated by the 2005 Phoenix results
presented in section 5.4.3, is appears that these design
modifications did not adequately address the Kimoto's PM2 5
measurement bias. Based on a review of the data, it is
recommended that future instrument development initiatives
focus on accurate calibration of the fine channel's beta gauge at
low PM2 5 concentrations. It also recommended that the system's
software be validated to ensure that the measured PM2 5 aerosol
mass during a sampling event is being accurately converted to
PM2 5 concentration.
With the exception of a single electronics failure, the two TSI
Model 3321 units appeared to function well and provided
acceptable levels of precision during the 2003 and 2004 field
tests. Following a review of the literature, a coarse aerosol
specific gravity and shape factor of 2.0 and 1.4, respectively,
were used to convert the APS' response to PMi0-2.5 mass
concentration. For the Riverside, 2003 Phoenix, and 2004
Phoenix field campaigns, results were highly correlated with the
collocated FRM samplers and provided similar PMi 0-2.5
concentrations. For the Gary, IN data, however, the APS results
were less correlated and under-measured PM10-2.5 by
approximately 30%. Based on the hypothesis that the negative
PM10-2 5 measurement bias might be attributable to transport
losses of large, hygroscopic particles in the humid Gary sampling
environment, a smart heater was designed and constructed to
heat the incoming aerosol if the ambient relative humidity
exceeded 45%. Unfortunately, the environmental conditions
during the 2005 Phoenix tests did not enable this modification to
be evaluated. In addition, operational problems were
encountered with both APS units during the 2005 Phoenix tests
that invalidated a large percentage of the collected data. While
the source of the operational problem is still under investigation,
it is believed to be associated with a calibration error that
occurred during the factory servicing of the two APS units
immediately prior to the 2005 Phoenix tests. The manufacturer
has indicated that this problem can be properly identified and
addressed prior to the start of the fall 2005 field tests in
Birmingham.
The battery-operated, 5 1pm BGI Omni samplers were designed
to provide a low-cost means of conducting saturation monitoring
studies. Four prototype Omni monitors first became available
for evaluation during the 2005 Phoenix tests. Two of these units
were configured to operate as PM2 5 samplers while two were
configured to operate as PMi0 samplers. While functional
problems with the Omni's relative humidity sensor limited the
overall data capture rate, the samplers were generally able to
maintain their flow rate, ambient temperature measurement, and
ambient pressure measurement calibrations. The Omni's intra-
sampler precision for PM2 5 was determined to be 8.8%. On
average, the Omni units provided PM25 concentrations that
exceeded the PM2 5 FRMs by approximately 7%. The relatively
large scatter of the Omni PM2 5 data versus the collocated PM2 5
FRMs may be due to uncertainties in gravimetric measurements
associated with the relatively small amount of mass collected at
the sampler's 5 1pm flow rate. The Omni's PMi0 CV of 3.3%
40
-------
compares favorably with the FRM's PM25 CV of 2.2% during
the same sampling events. On average, the Omni's PMi0
concentrations were 11% lower than those of the collocated
PMi o FRM samplers. Systematic biases in PMi 0 measurements
may reflect inadequate aspiration of large particles and/or a PMi 0
fractionation curve that does not sufficiently match that of the
PMioFRM inlet.
No operational problems were encountered with the three Grimm
EnviroTrack Model 1.107 samplers during the 30-day 2005
Phoenix study and the units tended to hold their calibrations
well. Intra-sampler precision among the three units was excellent
as evidenced by CVs of 2.2%, 4.1%, and 3.4% for PM25, PMi0.
2 5, and PM10 measurements, respectively. In comparison to the
collocated FRMs, the Grimm units over-measured PM2 5, PMi0.
2 5, and PMio concentrations during all 30 sampling events. On
average, the Grimm units over-predicted PM25 concentrations by
37% when compared to the collocated PM25 FRMs. Results
were well correlated (R2 = 908) for PM25 but the slopes and
intercepts were 0.83 and 4.8 (ig/m3, respectively. PMi0.25
results were also well correlated (R2 = 0.847) but slopes and
intercepts were 1.35 and 8.4 (ig/m3, respectively. As for all
samplers that became available only in time for the Phoenix 2005
field tests, these limited tests results are insufficient to make
strong conclusions regarding the Grimm's measurement
capabilities. The upcoming field tests in Birmingham will
provide additional comparative data upon which to make more
reliable observations and conclusions.
No functional problems were apparent with the two prototype
dichotomous TEOM samplers during the 2005 Phoenix tests.
Upon examination of the collected field data, however, it became
apparent that some operational problems existed with the units
during the 30-days of testing. In particular, negative PM25
concentrations were reported at a frequency of 20% and 3 8% for
the two TEOM units, respectively. Although results are
preliminary, the manufacturer reports that flow leaks have been
discovered in the purge filter conditioning section of the
prototype units. Leaks in this component would tend to occur
during the instrument's purge cycle and introduce ambient
aerosol into the flow stream where it would be subsequently
collected and analyzed. Since this measured concentration is
subsequently subtracted from the concentration measured during
the instrument's normal sampling cycle, these component leaks
would tend to produce negative PM25 calculations. The
manufacturer indicates that this problem can be adequately
addressed prior to start of the fall 2005 field tests in Birmingham.
As compared to the PM25 measurements, negative PMi0.25
concentrations were reported by the dichotomous TEOM during
only 9 hourly events. Mean dichot to FRM PMi0.2 5 ratios for the
two units were determined to be 0.85 and 0.89, respectively. The
mean PMi0.2.5 response of the two units was very highly
correlated with the collocated FRMs, as evidenced by a mean
correlation coefficient of 0.992.
41
-------
Chapter 7
References
Allen, G.A., Oh, J.A., Koutrakis, P., Sioutas, C. (1999).
Techniques for high-quality ambient coarse particle mass
measurements. J. of Air Waste Manage. Assoc. 49:133-141.
Lin, J., Noll, K.E., Holsen, T.M. (1994). Dry deposition
velocities as a function of particle size in the ambient
Atmosphere. Aerosol Sci. Technol. 20:239-252.
Loo, B., Cork, C. (1988). Development of high efficiency virtual
impactors. Aerosol Sci. Technol. 9:167-176.
Misra, C., Geller, M., Sioutas, C., Solomon, P. (2001).
Development and evaluation of a PMi0 impactor-inlet for a
continuous coarse particle monitor. Aerosol Sci. Technol.
37:271-281.
Misra, C., Geller, M., Shah, P., Sioutas, C., Solomon, P. (2003).
Development and evaluation of a continuous coarse (PMi0 -
PM2.5) particle monitor. J. of Air Waste Manage. Assoc.
51:1309-1317.
Noble, C.A., Vanderpool, R.W., Peters, T.M., McElroy, F.F.,
Gemmill, D.B., Wiener, R.W. (2001). Federal reference and
equivalent methods for measuring fine particulate matter.
Aerosol Sci. Technol. 34:457-464.
Noll, K.E., Fang, K.Y.P., Watkins, L.A. (1988). Characterization
of the deposition of particles from the atmosphere to a flat
plate. Atmospheric Environment 22:1461-1468.
Stein, F., Esmen, N.A., Corn, M. (1988). The shape of
atmospheric particles in Pittsburgh air. Atmospheric
Environment 22:1461-1468.
U.S. EPA (1997). Part 50 - National Primary and Secondary
Ambient Air Quality Standards, Federal Register, Vol. 62, No.
138, p38711. July 18, 1997.
Volckens, J. and Peters, T.M. (2005). Counting and particle
transmission efficiency of the aerodynamic particle sizer. J.
Aerosol Sci. 36:1400-1408.
42
-------
DISCLAIMER
This document has been reproduced from the best
copy furnished by the sponsoring agency.
-------
O CDA
S&%ZTr\
United States
Environmental Protection
Agency
Office of Research
and Development (8101R)
Washington, DC 20460
EPA 600/R-06/093
September 2006
www.epa.gov
------- |