/
/
/
/ j - v \
h V jr; 1%
EPA/600/R-93/050 ~
RESEARCH TRIANGLE INSTITUTE March 1992
RTl/4948/108-02F
PARTICLE TOTAL EXPOSURE ASSESSMENT METHODOLOGY (PTEAM):
RIVERSIDE CALIFORNIA PILOT STUDY
FINAL REPORT
VOLUME I
by
E.D, Pelizzari, K.W. Thomas, CA Clayton, R.W. Whlmore
R.C. Shores, H.S. Zelon, and R.L. Perritt
RTI Work Assignment Leaden E.D. Pelizzari
Research Triangle Institute
Post Office Box 12194
Research Triangle Park, NC 27709-2194
Contract Number: 68-02-4544
Work Assignment Number: fV-108
Project Officer: David O. Hinton
Atmospheric Research and Exposure Assessment Laboratory
Exposure Assessment Division
Environmental Monitoring Branch
Task Manager: Lance A. Wallace
Atmospheric Research and Exposure Assessment Laboratory
Human Exposure & Field Research Division
Human Exposure Research Branch
PREPARED FOR
United States Environmental Protection Agency
Research Triangle Park, NC 27711
POST OFFICE BOX 12194 RESEARCH TRIANGLE PARK, NORTH CAROLINA 27709
-------
TECHNICAL RETORT DATA
(rteese read Instructions on the reverse before compter'
1. REPORT NO.
EPA/600/R-93/0b0
2.
3.
PB93-166957
4. TITLE AND SUBTITLE
Particle Total Exposure Assessment Methodology
(PTEAM): Riverside, CA Pilot Study Volume I
8. REPORT DATE
March 1992
6. PERFORMING ORGANIZATION CODE
7. AUTHOR1SI
Pellizzari, E.D.. Thomas, K.W., Clayton, C.A.,
Whitmore, R.W., Shores, R.C. Zelon, H.S. Perrltt, R
S. PERFORMING ORGANIZATION REPORT NO,
L> RTI/4948/108-02F
9. PERFORMING ORGANIZATION NAME AND AOORESS
Research Triangle INstitute
Research Triangle Park, NC 27709
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO,
68-02-4544
12. SPONSORING AGENCY NAME AND AOORESS
USEPA/ORD/AREAL
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
1*. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
The goal of this study was to estimate the frequency distribution of exposure of at
urban population to inhalable particles (less than 10 micrometers in diameter). A
probability sampling design was used to select 178 nonsmoking residents aged 10 or
above in Riverside, CA. Each person was monitored for two consecutive 12—h periods
during the fall of 1990, Concurrent samples were collected in the home and
immediately outside the home. The indoor-outdoor samples included both inhalable
particles ( lOu) and the fine fraction ( 2.5u), A central site operated for all
48 days of the study, collecting 96 12-hour samples using reference samplers
(dichotomous and hi—vol) side by side with the personal and indoor—outdoor monitors
The findings include 1) precision of tbe personal and indoor samplers was excellent
2) agreement of the samplers with the reference samplers was acceptable (10-20%);
3) personal exposure in the daytime exceeded both indoor and outdoor concentrations
by 50%; 4) personal exposure overnight was comparable to outdoor and slightly higher
than indoor concentrations; 5) 14 of the 15 prevalent elements were also elevated
by about 50% in the daytime personal samples; 6) persons carrying out household
cleaning activities or exposed to environmental tobacco smoke had significantly
higher exposures; 7) persons going to work had significantly lower exposures.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b. IDENTIFIERS/OPEN ENOED TERMS
c. COS at i Field, Group
exposure
inhalable particles (PM-10)
fine particles (FM-2.5)
personal monitors
18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (Thts Report!
UNCLASSIFIED
21. NO. OF PAGES
425
20 SECURITY CLASS (This page:
UNCLASSIFIED
22. PRICE
EPA f orin 2220—7 (R«*» 4—77) previous edition is obsolete
-------
EPA/600/R-
March 1992
RTI/4948/108-02F
PARTICLE TOTAL EXPOSURE ASSESSMENT METHODOLOGY (PTEAM):
RIVERSIDE, CALIFORNIA PILOT STUDY
E.D. Pellizzari, K.W. Thomas, C.A. Clayton, R.W. Whitmore
R.C. Shores, H.S. Zelon, and R.L Perritt
RTI Work Assignment Leaden E.D. Pellizzari
Research Triangle Institute
Post Office Box 12194
Research Triangle Park, NC 27709-2194
Contract Number 68-02-4544
Work Assignment Number: IV-108
Project Officer David O. Hinton
Task Manager: Lance A. Wallace
U.S. Environmental Protection Agency
Research Triangle Park, NC 27709
FINAL REPORT
by
Submitted by:
Approved by:
E.D. Pellizzari
T.D. Hartwell
Work Assignment Leader
Deputy Project Director
-------
PREFACE
This final report describes the results of a field monitoring program on
exposures to inhalable particles of residents of the City of Riverside,
California. Here we describe and discuss the survey design, survey
operations, field monitoring and methods, quality assurance, and statistical
analysis of inhalable particles (PM10) and fine particles (PM25) and
associated elements. Two additional reports will also be prepared. The
second report will present chemical and statistical analysis results for
phthalates and polynuclear aromatic hydrocarbons, an effort sponsored by the
California Air Resources Board. The third report will be authored by the
School of Public Health at Harvard University and will discuss the modeling of
indoor/outdoor/personal air relationships for particles and elements.
i
-------
ACKNOWLEDGEMENTS
Numerous dedicated scientists at several institutions and within the
U.S. EPA and California Air Resources Board provided invaluable guidance that
led to the successful implementation of PTEAM. The authors especially thank
personnel at Harvard and Acurex, our collaborators in this study, who provided
superb skills and expertise necessary to conduct the field monitoring.
We are most indebted to the hundreds of citizens in Riverside,
California who conscientiously wore monitors, kept diaries, and answered
questions about their activities.
ii
-------
ABSTRACT
The primary objective of EPA's Particle Total Exposure Assessment
Methodology (PTEAM) Study was to estimate the frequency distribution of
exposure of an urban population to inhalable particles (less than 10
micrometers in aerodynamic diameter). A probability sampling design was used
to select 178 nonsmoking residents aged 10 or older representing a population
of 139,000 non-smoking residents of the City of Riverside, California. Each
of the 178 selected persons and their homes were monitored for two consecutive
12-hour periods during the fall of 1990. During each monitoring period a
personal PM10, indoor PM10 and PM2 5, and outdoor PM10 and PM25 sample was
collected. The personal samplers and the indoor/outdoor samplers were
specially designed for this study. These samplers were also operated side-by-
side with dichotomous and Wedding high-volurae reference method samplers at one
ambient air monitoring station set up in Riverside to provide a comparison of
the sampling methods.
Population-weighted mean daytime personal PM10 concentrations were 150
jug/m3, more than 50% higher than the mean daytime values measured in the homes
or outdoors at the homes and the central monitoring site. This suggests that
ambient air monitoring may underestimate the population's actual exposure to
PM10. Overnight exposures also exceeded overnight indoor concentrations.
This suggests that models of exposure relying on time-weighted averages of
indoor and outdoor concentrations may also underestimate actual exposures.
Approximately 25% of the population's 24-hour exposure to PMl0 exceeded the
150 pg/m3 National Ambient Air Quality Standard while over 90% of the
population's exposure exceeded the 50 ^g/m3 California Ambient Air Quality
Standard.
The causes of the higher exposures are not completely determined, but
probably include various personal activities. For example, persons who
engaged in housework (vacuuming, dusting, cooking, etc.) had significantly
higher exposures. Also, homes with smokers had significantly higher levels of
both PM10 and PM2 5 compared to homes without smokers. Persons who went to
work had significantly lower exposures than persons who did not work.
Overnight personal PM10 concentrations were lower than those measured in the
daytime. Since this includes the sleeping period, this is further evidence
iii
-------
for the importance of personal activities in contributing to the increased
personal exposures observed.
The personal and microenvironraental monitors developed for this study
showed excellent precision (median relative standard deviations of about 5%).
As had been indicated in a previous pilot study, the new samplers collected
10-15% more mass than the dichotomous samplers, which in turn collected about
7% more mass than the Wedding samplers. These differences between monitoring
methods are all statistically significant, but do not affect the general
conclusions.
All filters were analyzed for selected elements by x-ray fluorescence
(XRF); 15 elements were found at measurable concentrations in a majority of
the samples. Host elemental concentration distributions were similar to those
observed for PM10. Elemental daytime concentrations were higher than
overnight concentrations for all sample types. Daytime personal elemental
concentrations were much higher than indoor and outdoor levels for 14 of the
15 elements. This increase in personal exposures to the elements suggests
that the increase in personal exposures to the total mass of the particles was
not due to organic particles from the body (skin flakes) or clothing; had this
been the case, no increase in the elemental concentrations would have been
expected. As was the case with the total mass, overnight personal elemental
concentrations were typically higher than indoor levels but lower than outdoor
levels. Distributions of sulfur and chlorine were not similar to those of the
other elements or PM10.
iv
-------
TABLE OF CONTENTS
Page
PREFACE i
ACKNOWLEDGEMENTS ii
ABSTRACT iii
LIST OF FIGURES viii
LIST OF TABLES x
ABBREVIATIONS xiv
SECTION
1 INTRODUCTION 1-1
Background 1-1
Preliminary Research and Nine-Home Study 1-4
PTEAM Program Objectives 1-5
2 CONCLUSIONS 2-1
Survey Design and Operations 2-1
Field Monitoring and Methods 2-2
Particle and Elemental Concentration Data 2-3
3 RECOMMENDATIONS 3-1
Sampling Bias 3-1
Further Data Analysis 3-1
Field Monitoring and Methods 3-2
Quality Assurance 3-2
Survey Operations 3-3
Response Rate Improvement 3-3
4 SURVEY DESIGN 4-1
Study Objectives and Target Population 4-1
Statistical Sampling Design 4-5
v
-------
TABLE OF CONTENTS (continued)
Paoe
5 SURVEY OPERATIONS 5-1
Introduction and Overview 5-1
6 FIELD MONITORING AND METHODS 6-1
Introduction 6-1
Technical Approach 6-3
Results and Discussion 6-29
7 QUALITY ASSURANCE 7-1
Introduction 7-1
Quality Assurance Project Plan 7-1
Performance and Technical Systems Audit Results 7-1
Audit and Data Quality Results 7-12
Summary of Quality Assurance Objective Results 7-21
Effectiveness of QA/QC Plan 7-24
8 SAMPLE WEIGHTING AND NONRESPONSIVE ADJUSTMENTS 8-1
Weights Based on the Sampling Design 8-3
Weight Adjustments for Nonresponse 8-5
Response Rate improvement 8-21
Compensating for Nonresponse Bias 8-27
Temporal Randomization 8-29
9 STATISTICAL ANALYSIS OF PARTICLE DATA 9-1
Introduction 9-1
Data Availability and Quality Control 9-2
Temporal-Site Aerosol Concentrations 9-9
Personal and Residential Data Analysis 9-27
10 STATISTICAL ANALYSIS OF ELEMENTAL DATA 10-1
Introduction 10-1
Preliminary Analyses 10-1
Elemental Analysis Quality Control 10-5
Elemental Analysis Results 10-39
11 REFERENCES 11-1
vi
-------
TABLE OF CONTENTS (continued)
APPENDIX
A Table of Contents for OMB Supporting Statement
B Participant Consent Form
C Supplemental Consent Form for Guardians of Minors
D Refusal Documentation and Conversion Form
E Lead Letter and Informational Brochure Sent to Potential Participants
F Table of Contents for Field Interviewer Instruction Manual
G Household Enumeration Questionnaire
H Study Questionnaire
I Time/Activity Survey
J Memoranda to the Record
K Temporal-Site 12-Hour Aerosol Concentration Data
L Cascade Impactor (Temporal-Site) Aerosol Concentration Data
M Outdoor (SAM), Indoor (SIM), and Personal (PEM) Aerosol Concentration Data
N Tabulations of Questionnaire Responses
O Clustering of Smokers
P Plots of XRF Analysis Bias Over Time
Q Plots of Duplicate Analysis Elemental Concentration Results
R Bar Plots of Mean Element Concentrations for PEM, SIM, SAM 2.5 and 10 urn
Samples
S Quality Control Review Memorandum for XRF Filter Analysis Data
T Memoranda Describing PTEAM Database and Performance Audits
vii
-------
LIST OF FIGURES
Figure Page
2-1 24-Hour PM^q Distributions 2-4
4-1 First-Stage Sample Selection Process 4-11
4-2 Second- and Third-Stage Sample Selection Process 4-19
8-1 Illustration of the Necessity of Sampling Weights
for Unbiased Population Inferences 8-2
8-2 Nonresponse Adjustment Process for Household Screening Weights 8-10
8-3 Nonresponse Adjustment Process for Person-Level Weights 8-17
8-4 Nonresponse Adjustment Process for Household-Level Weights for Homes
Selected for Particulate Monitoring 8-19
8-5 Nonresponse Adjustment Process for Households Selected for
Phthalate and PAH Monitoring 8-20
9-1 Distributions of Daytime Temporal Site PM^ and PM2 5
Concentrations (ng/m3) " 9-13
9-2 Distributions of Nighttime Temporal Site PM^q and PM2 5
Concentrations (jig/m3) ' 9-14
9-3 24-Hour PM10 Distributions 9-30
9-4 Estimated PM^q Distributions (Weighted) 9-33
10-1 Percent Bias for Calcium in SRM 1832 by XRF Analysis Over Time.
Associated Analysis Batch Designated by A, B, or C 10-16
10-2 Percent Bias for Lead in SRM 1833 by XRF Analysis Over Time.
Associated Analysis Batch Designated by A, B, or C 10-17
10-3 Concentration of Calcium (ng/m^) in Samples Analyzed with
Batch A or B Versus Duplicate Analysis in Batch C (some observations
are hidden) 10-33
10-4 Concentration of Chlorine (ng/m^) in Samples Analyzed with Batch A or B
Versus Duplicate Analysis in Batch C (some observations are hidden). 10-34
10-5 Mean Silicon Concentrations 10-56
viii
-------
LIST OF FIGURES (continued)
Figure Page
10-6 Mean Aluminum Concentrations 10-57
10-7 Mean Iron Concentrations 10-58
10-8 Mean Sulfur Concentrations 10-59
10-9 Mean Chlorine Concentrations 10-60
10-10 Mean PM2 5 and PM^q Concentrations 10-61
10-11 Distribution of Silicon Concentrations for Residence PM^ q Samples 10-62
10-12 Distribution of Lead Concentrations for Residence PM^ q Samples 10-63
10-13 Distribution of Chlorine Concentrations for Residence PM^q Samples 10-64
10-14 Distribution of Sulfur Concentrations for Residence PM10 Samples 10-65
ix
-------
LIST OF TABLES
Table Page
4-1 Census Tracts Defining Geographic Strata 4-8
4-2 Relative Advantages and Disadvantages of Selecting Participants During
the Household Screening Interviews and After all Screening
Interviews Have Been Completed 4-14
6-1 PTEAM 1990 Pilot Study Training and Rehearsal Schedule 6-4
6-2 Total Number of Samples Scheduled for Collection in Pilot PTEAM Study .... 6-8
6-3 PTEAM Field Monitoring Schedule 6-9
6-4 Typical Appointments and Activities at One Participant's Home 6-13
6-5 Periods of Collection and Particle Sizes for the Cascade Impactor 6-19
6-6 Typical Dally Schedule for the Field Monitoring Staff 6-26
6-7 Sample Collection Status for PTEAM 1990 Pilot Study 6-31
7-1 Summary of Flow Rate Audit Results 7-3
7-2 Summary of Audit Results from Balance Audit 7-9
7-3 Temperature/Pressure Correction Calculations Compared to
Auditor Calculations 7-18
7-4 Method Detection Limit and Field Blank Data for Particle Sampling 7-25
8-1 Household Enumeration Sample Results 8-7
8-2 Monitoring Sample Results 8-11
8-3 Summary of Pilot PTEAM Response Rates 8-12
8-4 Monitoring Phase Response Rates by Age of Selected Participant 8-14
8-5 Monitoring Phase Response Rate by Education of the Head of Household .. 8-15
8-6 Response Rates Achieved in Comparable Exposure Monitoring Studies .... 8-23
8-7 Analysis of Screening Phase Response Rates for Selected Exposure
Monitoring Studies 8-25
x
-------
Table
LIST OF TABLES (continued)
Page
9-1 Summary of Availability and Quality of 12-Hour Aerosol Samples 9-3
9-2 Summary of Distributions of Net Weights (fig) for Blank Filters 9-5
9-3 Summary of Distributions of Relative Standard Deviations (%)
Calculated from Collocated Aerosol Concentration Data 9-7
9-4 Collocated PEM Aerosol Data from Samples Worn by Project Staff 9-8
9-5 Distributions of all 12-hour Temporal-Site Aerosol Concentrations (ng/m3) ... 9-10
9-6 Distributions of Daytime Temporal-Site Aerosol Concentrations (ng/m3) 9-11
9-7 Distributions of Nighttime Temporal-Site Aerosol
Concentrations (ng/m3) 9-12
9-8 Pearson Correlations of Temporal-Site Aerosol Concentrations and
Log (Concentrations) 9-15
9-9 Temporal-Site Relationships Between SAM, PEM, and Wedding
Aerosol Concentration Measurements and Those of Dichots 9-17
9-10 Temporal-Site Relationships Between SAM, PEM, and Wedding
Aerosol Concentration Measurements 9-19
9-11 Temporal-Site Relationships Between SAM, PEM, and Wedding PM^q
Aerosol Concentrations and the Dichot Fine And Coarse Fractions 9-21
9-12 Summary of Distributions of Meteorological Data 9-23
9-13 Correlations of Meteorological Data 9-24
9-14 Pearson Correlations of Meteorological and Temporal-Site Data 9-25
9-15 Temporal-Site SAM2 5 and SAM10 Mean Concentrations (fig/m3), by
Wind Direction and Speed 9-26
9-16 Distributions of 24-Hour PM.,q Concentrations (ng/m3) 9-29
9-17 Weighted Distributions of Personal, Indoor, and Outdoor Particle
Concentrations (jig/m3) 9-32
9-18 Weighted Distributions of Ratios of Particle Concentrations 9-34
xi
-------
LIST OF TABLES (continued)
Table Page
9-19 Weighted Distributions of PM2 5/PM10 Concentration Ratio 9-37
9-20 Correlations of Persona! and Residence Particle Concentrations 9-38
9-21 Correlations of Temporal-Site, Personal, and Residence
Aerosol Concentrations 9-40
9-22 Correlations of Meteorological Data with Personal and
Residence Data 9-41
9-23 Weighted Distributions of Percent of Time Spent in Various Environments,
Based on the Nighttime Diary Responses 9-44
9-24 Weighted Distributions of Percent of Time Spent in Various Environments,
Based on Responses to the Daytime Diary 9-45
9-25 Effects of Activities of Mean PM2 5 and PM^q Indoor (SIM)
Concentrations (wj/m3) ....' 9-46
9-26 Effects of Particle-Generating Activities on Mean PM^q Personal
Exposures (ug/m3) 9-48
10-1 Comparison of Percent Measurables for 10 M.m Samples Between Energy-
Dispersive and Wavelength Dispersive XRF Analyses 10-3
10-2 Weighted Estimates of Percent Measurable for individual (PEM) and Households
(SIM and SAM) Populations, by Time of Day; Primary Elements 10-9
10-3 Weighted Estimates of Percent Measurable for Individual (PEM) and Households
(SIM and SAM) Populations, by Time of Day: Secondary Elements ... 10-10
10-4 Estimates of Percent Measurable for Temporal-Site Samples, by Time
of Day: Primary Elements 10-11
10-5 Estimates of Percent Measurable for Temporal-Site Samples, by Time
of Day: Secondary Elements 10-12
10-6 Summary of Results for Standard Reference Material Blanks 10-14
10-7 Summary of Results for Standard Reference Material Samples Having
Non-Zero Levels 10-15
xii
-------
LIST OF TABLES (continued)
Table Page
10-8 Summary of Laboratory, Field, and Special Blank Elemental
Analysis Results 10-19
10-9 Summary of Dichot Blank Elemental Analysis Results 10-21
10-10 Summary of Element Relative Standard Deviations for Collocated Samples . 10-24
10-11 Summary of Distributions of Percentage Differences Between Elemental
Concentrations from Two Collocated Dichot Samplers 10-29
10-12 Summary of Relative Standard Deviations for Duplicate XRF Analyses
Conducted by EPA on PEM/SIM/SAM Filters 10-30
10-13 Summary of Relative Standard Deviations and Percentage Differences
Between XRF Analyses Conducted by EPA on PEM/SIM/SAM Filters,
by Analysis Batch 10-31
10-14 Summary of Distributions of Differences and Percentage Differences
Between Elemental Concentrations as Determined by EPA and LBL
Analyses on the Same PEM, SIM, or SAM Filter 10-36
10-15 Median Differences and Percentage Differences Between Elemental
Concentrations as Determined by EPA and LBL Analyses on the Same
Dichot Filters 10-37
10-16 Summary of Distributions of Differences and Percentage Differences
Between Elemental Concentrations as Determined by EPA and DRI
Analyses on the Same Filter 10-38
10-17 Mean Elemental Concentrations (ng/m3) for Temporal-Site Samples 10-42
10-18 Weighted Mean Elemental Concentrations (ng/m3) for Residence Data .... 10-43
10-19 Relative Standard Errors (%) for Mean Elemental Concentrations 10-44
10-20 Weighted Mean Elemental/Particle Mass Ratios (%) for Residence Data .. . 10-45
10-21 Relative Standard Errors (%) for Mean Mass Ratios 10-46
10-22 Weighted Distributions of Element Concentrations (ng/m3) for Residence Data10-47
10-23 Correlations Between Outdoor, Indoor and Personal Element Concentrations 10-55
xiii
-------
ABBREVIATIONS
AER
air exchange rate
ACH
air changes per hour
ADP
automated data processing
AMQ
Activity Modification Questionnaire
BNL
Brookhaven National Laboratories
CAA
Clean Air Act of 1977
CAAQS
California Ambient Air Quality Standard
CARB
California Air Resources Board
CATS
capillary adsorption tube sampler
DICHOT
dicotomous sampler
EPA
U.S. Environmental Protection Agency
ETS
environmental tobacco smoke
FP
fine particle
FSUs
first-stage sampling units
GC/ECD
gas chromatography with an electron capture detector
GC/MS
gas ch romatog raphy/mass spectrometry
HEQ
Household Enumeration Questionnaire
HPF
House Plan Form
ICR
Information Collection Request
ID
identification
I/O
indoor/outdoor
IP
inhalable particulates
IRB
Institutional Review Board
LBL
Lawrence Berkeley Laboratory
LOD
limits of detection
MDL
method detection limits
MEMs
microenvironmentai monitors
MET
meteorological
NAAQS
National Ambient Air Quality Standards
NAMS
National Air Monitoring Stations
NCC
National Computer Center (U.S. EPA)
xiv
-------
-------
ABBREVIATIONS (continued)
NIST
National Institute of Standards and Technology
OMB
Office of Management and Budget
PAHs
poly cyclic aromatic hydrocarbons
PCF
permission consent form
PEM
personal exposure monitor
PFT
perfluorocarbon tracer
PM
particulate matter
PMR
probability minimum replacement
PPS
probability proportional to size
PTEAM
Particle Total Exposure Assessment Methodology
QA
quality assurance
QAPP
Quality Assurance Project Plan
QC
quality control
PMCH
perfluorinated methylcydohexane
RSD
relative standard deviation
RSP
respirable suspended particulates
SAM
Stationary Ambient Monitor
SIM
Stationary Indoor Monitor
SLAMS
State and Local Monitoring Stations
SMS As
Standard Metropolitan Statistical Areas
SQ
Study Questionnaire
SSI
size selective inlet
SSUs
second-stage sampling units
TAD
Time Activity Diary
TAQ
Time and Activity Questionnaire
TEAM
Total Exposure Assessment Methodology
TOD
time of day
TOW
time of week
TSP
total suspended particulates
XRF
X-ray fluorescence
xv
-------
-------
SECTION 1
INTRODUCTION
BACKGROUND
The goal of existing environmental programs is the protection of public health and
welfare. One conceptual approach to achieving this goal is to determine the degree of risk
that environmental pollution poses to public health. A risk model using this approach requires
knowledge of five fundamental components: (1) the sources of pollutants, (2) the fate and
transport of these pollutants from sources to humans, (3) the exposure of humans to these
pollutants, (4) the actual doses received by exposed people, and (5) the health effects
resulting from the doses. These five components, linked in a chain from source to effect,
comprise the full risk model (Ott et al., 1986). The output of each component serves as the
input to the next component, and the absence of valid information on any component may
impair the model's ability to accurately assess public health risks. Human exposure data
constitute one link that provides a scientific basis for making regulatory policies designed to
protect public health (Ott et al., 1986).
Of the five components in the risk model, the least information is available for
exposure and dose (Ott et al., 1986). In considering the human component of environmental
protection, one must consider both the concentrations and the manner by which the pollutant
actually reaches the people. Then a number of fundamental questions may be asked about
estimating the risk of pollutants to the general public (Newill, 1987). For example, how many
people are exposed to the pollutant, what are the sources of these exposures, and finally,
what are the effects of these exposures?
Often, we do not really know to what degree people are actually exposed to
contaminants released by pollutant sources (Newill, 1987). Even if the effects of a pollutant at
a given concentration are well known, it is not possible to determine the risk to public health if
we do not know the number of people exposed or the concentrations to which they were
exposed. Thus, a lack of adequate exposure data prevents the completion of the risk
equation (Ott et al., 1986).
The measurement of exposure serves as a critical parameter in environmental
protection. The data from exposure measurements help to evaluate progress in the efforts to
control environmental pollution and provide guidance for modifying approaches to make them
1-1
-------
more effective (Newill, 1987). Ultimately, the effectiveness of the regulatory process on
reducing pollutant levels may be judged by the exposure measurements.
In the early 1980's, the U.S. Environmental Protection Agency (EPA) began to
undertake programs to find the missing exposure data needed to complete the risk equation
and answer a number of the above questions (Newill, 1987). Personal monitors were used as
an effective means of measuring the exposures of individuals throughout their daily activities.
Besides measuring an individual's actual exposure to pollutants, the idea of estimating the
exposures of entire populations was formulated. Because it is prohibitively expensive to
measure the exposure of everyone in a large metropolitan area, the simpler approach of
combining probability sampling with environmental monitoring was chosen. In this manner,
the exposures of a representative probability sample of the population can be used to make
inferences about exposures of the entire population.
This concept was initially applied in exposure studies involving volatile organic
chemicals (VOCs). A number of these studies have now been completed, including VOC and
carbon monoxide personal exposure monitoring (Wallace, 1986; Pelizzari et al., 1986a,
1986b, 1988a, 1988b, 1988c; Hartwell et al., 1986). In these studies, personal monitors for
VOCs were developed and validated by researchers, and used by a selected probability
sample of participants to distinguish between microenvironments such as commuting, home,
and work. By combining data from a personal monitor with activity log information, the
environments responsible for exposure could be resolved. The VOC and carbon monoxide
exposure field studies were the beginning of an emerging field of total human exposure
assessment. The four basic ingredients of a Total Exposure Assessment Methodology
(TEAM) Study are (1) a representative probability sample of the population of concern, (2)
direct measurement of the pollutant concentrations reaching people through relevant media,
(3) direct measurement of body burden, and (4) direct recording of each person's daily
activities. The TEAM VOC studies have collected data on the exposures of populations to
important toxic pollutants in 10 cities in 6 states representing over 1 million people (Wallace,
1986; Pellizzari et al., 1986a, 1986b, 1988a, 1988b, 1988c; Hartwell et al., 1987).
The need to consider size characteristics and chemical composition in the control of
airborne aerosols has been a matter of continuing concern to EPA since the establishment of
the total suspended particle ambient air quality standards in 1971. Driven by the U.S. EPA's
requirement under the amended Clean Air Act (CAA) of 1977, research has been under way
1-2
-------
to develop information for size-specific standards for inhalable particulates (IP). IPs are
defined as airborne particles of less than or equal to 15 jim aerodynamic equivalent diameter
(Miller et al., 1979). As a result of this research, a PM-jq standard was established (Federal
Register, 1987).
PM.|0 (aerodynamic diameter = 10 nm) is the promulgated National Ambient Air
Quality Standard (NAAQS) representing that portion of particulate matter that can deposit in
the tracheobronchial and alveolar regions of the lung (Federal Register, 1987; Phalen et al.,
1986). The 24-hour average standard is 150 jig/m^. The State of California has set a 24-
hour PM.j q ambient air quality standard of 50 jig/m3. Several case studies have reported on
personal indoor/outdoor relationships for 10 jim size particles. Direct and indirect contribution
of outdoor PM10 to indoor air and personal exposures has been reported (Kamens et al.,
1991; Lioy, 1989; El-Shoboksky and Hussein, 1988; Kulmala et al., 1987; Sexton et al., 1984,
1985). PM-jq includes the respirable particle fraction (RSP, <3.5 nm) and fine particles
(<2.5 jim) (U.S. EPA, 1982). Earlier studies examined these relationships for respirable
suspended particulates (RSP) (Sega et al., 1986; National Academy of Sciences, 1985;
Spenglerand Socozek, 1984; Tosteson et al., 1982; Dockery and Spengler, 1981; National
Research Council, 1981) and for total suspended particles (Alzona et al., 1979; Thompson
et al., 1973; Yocom, 1982).
A second particle size range of less than or equal to 2.5 jim diameter (PMg 5) is
important because the predominant species of this particle size is responsible for entering into
the gas exchange region (alveolar sacs) of the respiratory tract. Thus, data collected in this
size range could be useful in conjunction with epidemiological health parameters in
understanding health effects.
Other than the cited case studies, no large-scale study has been conducted which
examines the frequency distribution of PM^q or PM2 5 and personal exposures for a major
population. Therefore in 1986 the Congress mandated that the EPA undertake a TEAM Study
of exposure to particles.
The major objective of the EPA's Particle Total Exposure Assessment Methodology
(PTEAM) Study is to estimate the frequency distribution of PM10 in a target population of
individuals. The frequency distribution for the whole target population can be estimated by
acquiring exposure data from a statistically valid sample of the target population. With
knowledge of such an estimated distribution of concentrations of particles of a given size
1-3
-------
(PM10) or smaller (PMg 5), it is possible to report an estimated proportion of the target
population that is exposed to PM^ levels exceeding, for example, a level of 150 fig/m^, the
24-hr NAAQS. Such data could be used in a regulatory program designed to maintain the
exposure of a finite percentage of the population below some target concentration (Newill,
1987). Data of this nature have important policy implications because they demonstrate that
reduction of exposures by strategies previously not considered could be more effective in
reducing risks than the more conventional regulatory strategies (Newill, 1987). Also, TEAM
studies have shown the importance of indoor air and human activities in the exposure
equation.
PRELIMINARY RESEARCH AND NINE-HOME STUDY
To achieve the primaiy objective of the PTEAM Study on a large-scale, two preliminary
phases of research had to be undertaken. First, we needed to develop methods of gathering
the necessary information for the PTEAM program, and second we needed to field test these
methods in a nine-home PTEAM Study to uncover problem areas for subsequent corrective
action, if necessary. The results of the preliminary research have been reported (RTI and
Harvard, 1990) regarding methods and instrument performance.
In designing a large-scale study, a number of issues were addressed. For example:
• What are the benefits of 12-hour versus 24-hour monitoring periods?
• Is the particular day or day of the week when an individual is monitored an
important consideration?
• What are the pros and cons of stratifying days by weekend/weekday or non-
workday/workday?
• Should one individual or several individuals per home be monitored?
• For indoor (microenvironmental) monitoring, how many sampling locations per home
should be used?
• How extensive does the outdoor monitoring portion of the population study need to
be?
• Do the survey questionnaires adequately define the individual's activities, locations,
and potential exposure to sources?
Such issues need to be considered from several perspectives. For instance, what are the
relative costs of one approach versus another? What are the burdens placed on the
1-4
-------
participants and their households? What benefits, in terms of increased precision or
increased representativeness are likely to be achieved by adopting one approach rather than
another?
Insight into such issues can often be gained by conducting a property designed
preliminary study. As a preliminary to the PTE AM population study, a nine-home study was
conducted in March 1989 in the San Gabriel Valley area of Southern California, which
includes cities such as Arcadia, Temple City, EI Monte, Monrovia, Azusa, Covina, and
Glendora. This study involved personal aerosol monitoring (12-hour measurements over a
period of several days) for two participants in each of nine volunteer households, using
personal exposure monitors (PEMs) and microenvironmental monitors, which monitor particles
in and near their homes.
The nine-home study's primary puipose was to develop a methodology for personal
exposure monitoring of particulates that could be applied later in the large-scale population
study. The nine-home study was also designed to produce data that could be subjected to
statistical analyses in an attempt to understand the extent of spatial and temporal variation of
the particulate concentrations. This information was expected to aid in addressing the
population study design issues. The results of the nine-home study have been previously
reported (RTI and Harvard, 1990).
PTEAM PROGRAM OBJECTIVES
The primary objective of the PTEAM program is to estimate the frequency distribution
of exposure of an urban population to inhalable and respirable particles. Measurements of
human exposure to inhalable and respirable particles are important for making health risk
assessments. Besides determining particle exposure levels, it is important to determine the
sources of particles. The source profile and the particle mass measurement data can then be
used to develop source apportionment and source receptor models for human exposure to
particles. This information, along with available toxicological data for human exposure to
particles, is useful in making future health risk analyses.
Several specific aims, which are important elements of the primary objective of
PTEAM, are:
(1) to provide the probability-based monitoring information for determining the
frequency distribution of exposure,
1-5
-------
(2) to measure human exposure to inhalable and respirable particles for health risk
assessments,
(3) to define the relationship between personal exposure,
microenvironmental/ambient air levels, and outdoor ambient air standards,
(4) to determine the sources of particles,
(5) to develop a source apportionment or source receptor model for exposure to
particles using source profile and mass measurement data.
Personal exposure monitoring provides the means of measuring actual human
exposure, whereas the fixed indoor and outdoor monitors help to determine the relative
contribution of indoor and outdoor sources to personal exposure. Two microenvironments,
indoor-home and outdoor-home, were addressed in this program. Furthermore, ambient
monitoring at a central site provided macroenvironmental indices of exposure. Questionnaires
were used to assess the frequency of exposure to sources. Differences between various
measurement methods were assessed. To investigate relationships between indoor exposure
and ambient air standards, it was important to make PM^q comparisons.
It was necessary to combine both mass and chemical measurements to provide the
information for determining the frequency distribution of exposure. Mass concentration
measurements were made gravimetricaliy to provide an index of particle exposure. Filter
samples used for mass measurement were also subjected to elemental analysis. Elemental
analysis is useful for both source receptor modeling and direct measuring of the level of
exposure to assist in understanding the observed particle and chemical exposure information.
Questionnaires were used to help identify activity-source occurrences leading to potential
exposure to elements and other chemical species being measured. To assess the
contribution of environmental tobacco smoke (ETS) to a person's exposure to inhalable and
respirable particles, nicotine was measured as a tracer of ETS.
Since the PTEAM Study was a cooperative venture through an independently funded
research effort supported by the California Air Resources Board (CARB), the study site was in
California. In addition to mass and elements, the particle-bound organics and semivolatile
organics were measured as indicators of potential mutagenicity and carcinogenicity. Thus, an
additional objective was to measure semivolatile organics in subgroups of the Southern
California test population. CARB sponsored the measurement and analyses of polyaromatic
hydrocarbons (PAHs) and phthalates in both indoor and outdoor microenvironmental samples.
1-6
-------
Results of measurements for semi-volatiles and nicotine are not included in this report; these
data will be described in separate reports to follow.
This report presents the results for the PTEAM Study conducted in Riverside,
California representing a target population of 139,000+16,000 (S.E.) nonsmoking residents
aged 10 and above. The city of Riverside was selected because it is known to have highly
variable outdoor PM^g concentrations and because the socioeconomic characteristics of the
community appeared to provide a reasonably representative microcosm of the Southern
California population. A wide range of outdoor levels offers the best chance of determining
the contribution of outdoor levels to indoor levels and personal exposure. The fall season was
selected to capture the wide range of meteorological conditions that have strong effects on the
outdoor levels of particles.
1-7
-------
-------
SECTION 2
CONCLUSIONS
SURVEY DESIGN AND OPERATIONS
Riverside, California was selected as the urban area most suitable for the PTEAM
study because of the temporal variability of inhalable particulate concentrations and the
absence of a single dominant stationary source of pollutants. A probability sampling design
was used to select participants in Riverside to provide the basis for inferences to the target
population's distribution of personal exposures to inhalable particles (PM^q). Population units
sampled for data collection were person-days because the observed sampling outcome
depended upon the time during which the measurements were performed. The sample was
designed to yield data for 175 nonsmoking residents aged 10 or older representing 139,000 +
16,000 non-smoking residents of Riverside. People working at least 30 hours per week
outside their home and people exposed to passive smoking within their homes were slightly
oversampled to support separate statistical analyses. The survey design resulted in a sample
of people for which defensible inferences can be made to the target population.
An effective effort was made to improve participant response through press coverage
and a lead mailing using the EPA logo with a letter signed by several officials, including the
mayor. Because potential participants remembered the press coverage, the interviewers had
more credibility when approaching the participant during household screening interviews.
Screening interviews at the households selected by probability sampling and participant
selection and recruitment were conducted during one interviewer visit. As a result, 178
persons were successfully recruited into the study, resulting in an overall response rate of
49%, which was similar to response rates for many previous chemical monitoring studies that
required cooperation during several visits to the participant's home as well as the burden of
wearing a chemical monitor for 24 hours. Questionnaires were carefully designed to elicit
information about the participant's locations, activities, and potential exposure to particle
sources during each 12-hour monitoring period. These questionnaires were successfully
administered to all study participants.
2-1
-------
FIELD MONITORING AND METHODS
The PTEAM field monitoring program was designed to collect sufficient samples to
characterize the exposure of a representative population in Riverside, California to inhalable
particulates. A total of 178 persons were selected for monitoring, each was monitored for two
consecutive 12-hour periods. During each monitoring period one personal PM-jg, one indoor
PMi0, one indoor PM2 5, one outdoor PM10, and one outdoor PM2 5 sample were scheduled
for collection. Approximately 4 persons were monitored each day from September 22 through
November 9, 1990. Additional samples were collected to measure nicotine levels and air
exchange rates in each study home to provide source characterization data. Under direction
from the California Air Resources Board, PAH/phtha!ate samples were collected in a subset of
125 homes. A monitoring site was set-up at one location (referred to as the temporal site in
this study since 12-hour samples were collected continuously throughout the course of the
study) in Riverside to monitor ambient levels of particulates. Reference method PM10
samplers and dichotomous samplers were operated at this temporal site alongside personal
and stationary monitors identical to those used for participant monitoring. A mobile laboratory
was set-up in Rivefside for performing all particulate filter weighing operations.
Over 95% of the personal and stationary particulate samples, nicotine samples, and air
exchange samples were successfully collected as scheduled. Over 90% of the PAH/phthalate
samples were successfully collected. All filters were weighed on-site in the weight laboratory
under controlled conditions of temperature and humidity. All sample collection and weight
data were entered directly into computers in the field to reduce data transcription errors and
facilitate construction of the database. The only significant portion of data not collected as
intended during the study was measurement of meteorological conditions at the temporal site.
Meteorological data was obtained from three airports in the vicinity of Riverside to supplement
the data from the temporal site.
The particle sample collection and weighing methods performed very well. Sample
collection and weighing precision was better than 5% relative standard deviation for most
collocated particle samplers. The median particle filter background for persona! exposure
(PEM), stationary indoor (SIM), and stationary ambient (SAM) monitors was 9 jig as measured
on field blank filters. The median background of 9 ng is equivalent to a concentration of
approximately 3 jxg/m3 for a 12-hour sample volume of 2.98 m3. Over 85% of the filters that
were re-weighed for quality control purposes passed the maximum ± 4 jig tolerance on the
2-2
-------
first pass. All of the filters passed the tolerance criteria before the weight data were accepted.
Careful preparation and operation of the weigh facility greatly improved weighing operations
compared to the prepilot study. The procedures implemented to streamline filter handling and
reduce filter contamination were successful.
PARTICLE AND ELEMENTAL CONCENTRATION DATA
Particle Concentration Distributions
A primary goal of this research was to measure the distribution of exposure of an
urban population with regard to the 150 jig/m3 PM^ National Ambient Air Quality Standard
(NAAQS) and the 50 ng/m3 California Ambient Air Quality Standard (CAAQS). Total 24-hour
personal exposure concentrations, measured as person-days, were calculated as a time-
weighted average across two consecutive measurement periods of approximately 12 hours
each. Personal exposure monitors (PEM) were used to measure particle concentrations in the
participant's breathing zone while stationary indoor (SIM) and stationary ambient monitors
(SAM) were used to measure particle concentrations inside and outside the home.
Approximately 25% of the person-days exceeded the 150 p.g/m3 level and over 90% of the
person-days exceeded the 50 jig/m3 level during this study (Figure 2-1). The median 24-hour
concentration for personal exposure to PM1 q was 102 jj.g/m3 with a range from 34 to 287
jig/m3. Personal exposure concentrations were greater than those measured both inside and
outside of residences using SIM and SAM microenvironmental monitors. They were also
higher than concentrations measured at the ambient air monitoring site using several different
methods. These data indicate that a significant portion of the population was exposed to
particle levels higher the NAAQS and most of the population's exposure exceeded the
q
CAAQS. Outdoor air concentrations of PM^q at the residences were greater than 150 ng/m
on approximately 15% of the household-days. Indoor and outdoor 24-hour PM^q levels were
above 50 |igftn3 during over 70% of the household-days during this study.
Particle concentration levels were also examined on a daytime/nighttime basis.
Daytime personal exposure PM^q concentrations were much higher, as a whole, than daytime
indoor or outdoor PM^ q levels. Over 40% of the population was exposed to concentrations
above 150 ng/m3 during the daytime, and 10% of the population was exposed to levels above
260 ng/m3. Daytime personal exposure concentrations ranged from 35 to 455 ^g/m3 with a
q
median of 130 ng/m . In contrast, nighttime personal exposure concentrations were lower
2-3
-------
TEMPORAL SITE
HOUSEHOLD DAYS
(WEIGHTED)
PERSON DAYS
(WEIGHTED)
90th
75th
Mean
50th
25th
10th
###
###
###
M
###
90th
75tF
Mean
50th
25th
10th
DICHOT WEDDING SAM PEM
(n=47) (n=45) (n=43)(n=39)
OUTDOOR INDOOR PERSONAL
SAM SIM PEM
(n=153) (n=157) (n=161)
Figure 2-1. 24-Hour PMio Distributions
2-4
-------
than those measured outdoors and only slightly higher than indoor levels of PM1Q. Nighttime
personal exposure concentrations ranged from 19 to 278 ng/m3 with a median of 66 |ig/m3
Only 10% of the population was exposed to PM^q levels above 150 pg/m3 overnight.
Approximately 75% of the population was exposed to PM^q levels above 50 ng/m3 overnight.
Daytime indoor PM10 concentrations ranged from 17 to 513 p.g/m3 with a median of 82
ng/m3. Daytime outdoor levels ranged from 16 to 507 jig/m3 with a median of 84 n-g/m3.
Nighttime indoor PM10 concentrations ranged from 14 to 180 ^g/m3 with a median of 52
^g/m3. Nighttime outdoor levels ranged from 14 to 223 |ig/m3 with a median of 74 n.g/m3.
Ambient air concentrations of PM1 q and PMg 5 were measured throughout the course
of the study at one fixed outdoor monitoring site in Riverside. A side-by-side trial, using the
same personal and microenvironmental sampling systems that were used at homes, along
with dichotomous and Wedding reference sampling methods, was conducted at this ambient
air monitoring site. The ambient air monitoring site was designated as the "temporal site" in
this study because air samples were collected by all four samplers for 96 consecutive 12-hour
monitoring periods. The PEM (personal) sampling system mounted at the temporal site
resulted in PM^q concentrations 8% higher than the dichotomous samplers and 13% to 24%
higher than the Wedding samplers. These differences were statistically significant at the 0.05
level. The personal exposures measured during this study would not have been predicted
from ambient air or microenvironmental monitoring. The population's actual 24-hour PM10
exposure as measured by the personal monitors would have been underestimated by 30 to
40% based on ambient air levels measured by identical monitors located at the temporal site.
The population's daytime median exposure would have been underestimated by 50% using
ambient air measurements at the temporal site.
Activity questionnaires and diaries were used to elicit information about activities or
locations that could affect a person's exposure to aerosols during the periods in which they
were monitored. Reported activities were examined for statistically significant correlations with
personal and indoor air particle concentrations. Persons reporting daytime house work (indoor
cleaning, cooking, etc.) were exposed to significantly higher PM10 concentrations than those
reporting no house work. Daytime indoor PM10 and PM2 5 levels were also significantly
higher in homes with reported house work. Persons that did not go to work were exposed to
significantly higher PM1Q concentrations than persons reporting that they went to work. Many
of the persons who did not go to work reported house work as an activity. The presence of
2-5
-------
tobacco smoke was associated with significantly higher daytime and nighttime indoor PM-jq
and PM2 5 concentrations; personal exposure PM^ levels were significantly higher for those
persons reporting exposure to tobacco smoke during the nighttime period only. No significant
indoor or personal exposure concentration increases were associated with reported spraying
of consumer products. Daytime personal and indoor particle concentrations were significantly
higher in homes that did not report vehicle engines operated in attached garages. This
surprising result may be related to other factors associated with owning a garage and
suggests the need for analyses that attempt to deal simultaneously with several variables. No
other activities or house conditions were significantly correlated with personal or indoor
particle concentrations.
No strong correlations were observed between meteorological factors (temperature,
dew point, wind speed) and particle concentrations. Ambient air PM^q and PM2 5
concentrations did appear to have some dependence on wind speed and direction. Strong
Santa Ana winds with gust above 40 knots were present during two time periods and were
associated with high PM^q and low PM2 5 levels. Visibility was improved with northerly winds
(generally from the desert) in the 3 to 8 knot range when compared to westerly winds (from
the Los Angeles basin) at 3 to 8 knots. Northerly winds at 3 to 8 knots also were associated
with lower particle concentrations than were observed with westerly winds at 3 to 8 knots for
both mean PM^q (42 vs 89 ng/m^) and mean PM2 5 (20 vs 50 ng/m3) levels. Additional
source characterization will be performed by Harvard University based on both particle and
elemental data. Results of this work will be reported in a separate document.
Elemental Concentration Distributions
Concentrations of selected elements were measured by X-ray fluorescence (XRF)
analysis of all personal, indoor, outdoor, and dichotomous sample filters. Of 13 primary
elements of interest, eight (Si, Al, Co, Fe, K, Mn, Br, Pb) were present at measurable levels
on sufficient samples for further data analysis. Seven elements of secondary interest (S, Zn,
CI, Ti, Cu, Sr, P) were also measurable on at least 50% of the personal sample filters. Three
elements of primary interest (V, Se, Ni) were measurable on fewer than 20% of the personal,
indoor, and outdoor filters. Two other elements (Cd, As) were also rarely measurable on
dichot filters, as well as on the personal, indoor, and outdoor filters.
2-6
-------
Relative elemental concentration distributions in personal, indoor, and
outdoor air were similar to those observed for PM10 in most cases. Personal
daytime concentrations of most elements were much higher than those measured
indoors or outdoors. Personal nighttime concentrations were often higher than
Indoor levels but lower than outdoor levels. Sulfur was the most notable
exception; it was primarily associated with particles smaller than 2.5 jura and the
concentration distributions for personal, indoor, and outdoor locations were much
more similar than those of any other element. Elements associated with soils
(Si, Al, Ga) were present at the highest concentrations, usually over 1000 ng/m3
for PM10 samples. These three elements were usually found at much lower
concentrations on the PM2 5 samples. Potassium and sulfur concentrations were
also often above 1000 ng/m3. Lead was measured in a high percentage of samples,
with mean concentrations ranging from 17 to 40 ng/m3. Mean chlorine
concentrations were observed above 200 ng/m3, but these may not accurately
reflect actual air concentrations since other data suggest losses due to
volatility during analysis or storage.
Mean element/particle mass ratios were calculated for the personal, indoor,
and outdoor samples. Ratios for daytime personal samples were similar to those
for indoor and outdoor samples. These data suggest that organic particles from
people and their clothes were not responsible for most of the increase observed
for personal exposure FM10 concentrations. Further efforts are needed to
determine whether the elevated daytime personal exposures are a result of a bias
due to the design or operation of the personal sampling system or are a
reflection of truly higher exposures to particles. If people are exposed to
concentrations of particles higher than those measured with stationary monitors,
then additional research is necessary to understand the mechanism that causes the
elevated exposure and to determine the sources of the particles.
2-7
-------
-------
SECTION 3
RECOMMENDATIONS
SAMPLING BIAS
Further work should be conducted to determine if elevated daytime personal exposures
observed during PTEAM prepilot and pilot studies reflect true increases in PM^ q exposures or
are the result of bias associated with the monitoring equipment. Specific experiments should
be conducted to explore the possibility that the impactor unit is oversampling particles larger
than 10 |im from particle bounce or dislodgement due to orientation and movement. Such
experiments should consider all of the actions to which a person might subject the monitor.
Personal sample filters should continue to be examined by SEM to determine if particle sizes
or origins are skewed when compared to microenvironmental particle distributions. Additional
experiments should also be conducted to address the Issue of the personal particle cloud
(elevated particle concentrations around a person). Such experiments might be difficult to
perform but might lead to a better understanding of the mechanism of a person's exposure to
airborne particles.
The small but statistically significant bias between the SAM and PEM samplers and the
dichotomous and Wedding samplers at the temporal site also bears further investigation.
PM^q concentrations measured with the personal and microenvironmental monitors were
consistently higher than those measured with the reference monitors. It is important to
understand whether the PEM and SAM are oversampling particles larger than 10 |im or if the
differences are a result of other operational parameters. An understanding of the physical
factors leading to sampling biases in studies such as this is essential to fully interpret the
data.
FURTHER DATA ANALYSIS
Many additional analyses of the PTEAM data are possible and should be performed.
Source-receptor modeling based on elemental, particle, and air exchange data should be
performed. Associations between elemental concentrations and the questionnaire and diary
information could be investigated. Activities of persons in the upper percentiles of particle
exposure should be examined to learn if the causes for high exposure (i.e., occupation, house
characteristics, etc.) can be discerned. The relationship between particle concentration,
3-1
-------
reported smoking, and nicotine levels should be examined. An evaluation of the utility of
nicotine monitoring in similar studies would be useful. Further analysis of the meteorological
data over smaller time increments with SIM and SAM data might reveal associations with wind
direction and speed that were not observed previously. The association between
meteorological conditions and elemental concentrations has not yet been examined.
FIELD MONITORING AND METHODS
The field monitoring portion of the PTEAM pilot study was very successful. Based on
this experience, several recommendations are suggested to improve similar studies in the
future. Weight increases observed for the field blanks were low but not insignificant, ranging
from -29 to 34 fig, with a median of 9 fig. These background weights should not affect data
analysis for samples from locations with generally high particulate levels. However, these
background levels may significantly affect samples collected in locations with lower particle
concentrations. Efforts should be made to isolate the source of the background, and eliminate
it if possible, before conducting a similar study in environments exhibiting low aerosol levels.
The temperature and humidity control system in the weigh trailer should be improved
to handle extreme ambient temperature and humidity fluctuations observed in locations like
Riverside. Filter accounting procedures instituted in the field should be re-evaluated and
strengthened to ensure that filters are only used once for sample collection. Meteorological
data collection systems should be periodically tested during use to ensure proper operation.
Back-up equipment should be available for immediate substitution if a problem develops.
QUALITY ASSURANCE
Overall, despite the diversity and complexity of the field study, the operation was
smooth and successful. Several recommendations can be made to improve data quality in
similar large scale studies. Improvements can be made in the PEM samplers, making them
less obtrusive to the participant's everyday activities. Operations at the temporal site should
be better integrated with other field study operations. Temporal site samplers and
meteorological systems should be rigorously tested and audited early in the study to uncover
potential problems allowing for early corrective actions. Finally, software used to merge
weighing and sample collection data should be improved to allow field personnel to evaluate
data quality as the study progresses.
3-2
-------
SURVEY OPERATIONS
Preliminary survey documents and questionnaires were prepared only in English.
Future studies conducted in locations with significant populations of primarily Spanish-
speaking households should include survey documents prepared in Spanish. Spanish-
speaking personnel or hired translators must be present when activity questionnaires are
administered in the home.
RESPONSE RATE IMPROVEMENT
Future exposure monitoring studies must continue to try innovative survey procedures
that may produce higher response rates. A high response rate is the only guarantee that
n on response bias will be small. Protocols used in previous large-scale environmental
monitoring studies, and their results in achieving desired response rates, should be examined
to design future protocols using the most successful strategies. In order to reduce potential
systematic bias and capture important population, data the screening-phase response rate
must be kept as high as possible. Screening-phase response rates may be improved if they
are conducted without providing advance notification or any details about the monitoring
phase of the study. Pre-survey publicity is recommended, but it should not be mentioned by
the interviewer until the household screening interview has been completed and someone has
been selected for monitoring. Materials that present the objectives of the study should be
clear, succinct, and professional In appearance. Enhancing participant incentives may be
another way to improve response rates. An attractive alternative to large fixed cash incentives
might be a lottery. Each participant would be given the chance of winning a much larger sum
of money while still receiving a small cash incentive (e.g. $20).
We also strongly recommend that the survey work be scheduled so that all selected
housing units can be worked to completion. Sufficient interviewers must be hired to schedule
most monitoring appointments a week or two in advance with enough time to reach homes
that were difficult to contact. Participant monitoring days should be randomly selected
because exposure measurements can very greatly from one day to the next. One strategy for
implementing random-day sampling would be to select a target week for the participation of
each household. Within the target week each sample member would be assigned to either a
work day or non-work day. The participant could select any day of the designated type from
the sample week for participation. Procedures for randomly selecting participation days can
3-3
-------
add logistic difficulties and costs to the monitoring program. The need for random day
selection must be balanced against logistical considerations and cost efficiency of data
collection.
3-4
-------
SECTION 4
SURVEY DESIGN
A probability sampling design was used to select the sample units to be monitored in
the Pilot PTEAM Study. This design was chosen because probability sampling methods
provide the most defensible basis for inferences from a sample to the target population from
which the sample was selected (Hansen et al., 1983; Williams et al., 1983). Probability
sampling also provides a defensible basis for calculating the standard errors of survey
statistics.
A multistage area household sampling design was used to support placement of
exposure monitoring instruments at the homes. Area household samples are efficient for field
monitoring studies because the area clustering reduces field travel costs relative to more
geographically disperse samples of households.
The target population and statistical sampling design are presented in this section in
relation to the study objectives.
STUDY OBJECTIVES AND TARGET POPULATION
The primary objective of the Pilot PTEAM Study was to estimate the population
distribution of personal exposures to inhalable particles (less than 10 microns) for the
nonsmoking residents of Riverside, California. The study's primary data collection instrument
was a personal exposure monitor to be worn by or kept in the vicinity of each study participant
for 24 hours. The PTEAM personal monitor characterized a person's exposure for a 24-hour
time period. The measurements cannot be interpreted as representing any longer period of
time (e.g., all weekdays or all workdays).
In contrast, surveys of public opinion usually assume that the population characteristics
being measured (e.g., a person's opinion of the President's performance) are constant
throughout the period of data collection. For environmental surveys like the PTEAM study,
this is an incorrect assumption. When the observed outcomes depend on the time periods
during which the measurements are performed, the measurement process characterizes a
population unit defined in both time and space, and the sampling design should select
sampling units defined both temporally and spatially. To make the most defensible inferences
4-1
-------
from such units to a population, each temporal and space population unit should have a
known, positive probability of selection (see Gilbert, 1987).
Therefore, the population units to be sampled for data collection in the PTEAM study
are person-days. The persons included in the PTEAM target population were the permanent
residents of the city of Riverside, California (excluding the area around March Air Force Base,
south of Allessandro Boulevard) who were at least 10 years of age at the time of the survey
and who were not currently tobacco smokers. The days comprising the target population
included all days in the field data collection period, September 22, 1990 through November 9,
1990. Justification for the choice of the target area, the target time frame (Fall 1990), the
target age range, and the exclusion of smokers is presented in the subsections that follow.
Target Area
Technical and logistical criteria considered in selecting the Particle-TEAM Pilot Study
site included the following:
• Ambient aerosol levels should be somewhat heterogeneous and variable over the
study area to make it possible to discover correlates of the variability in ambient
levels.
• Ambient air pollution should not be dominated by a particular stationary source (i.e.,
steel mills, smelters, etc.).
• Size-fractioned particle data (e.g., PM2 5, PM10, IP15, or total suspended
particulates [TSP]) that could provide a basis for determining relative source
contributions must be easily available.
• Housing stock (apartment, multifamily, single-family housing) in the communities
should be heterogeneous.
• The site must be readily accessible by personnel conducting the study.
Standard Metropolitan Statistical Areas (SMSAs) that would yield communities that
might meet many of the selection criteria included Los Angeles, Philadelphia, Baltimore, St.
Louis, and Steubenville. Historic PM2 5, IP15, PM^, and TSP data were analyzed and the
top 20 communities/SMSAs were ranked according to the highest levels of PM2 5 and PM^q.
Based on this analysis, and the collaborative sponsorship between the U.S. EPA and the
CARB, the Particle-TEAM study site was chosen in California.
4-2
-------
Based on analysis of the U.S. EPA IP network monitoring data and on recent air
quality data reported by the CARB, the Riverside area was found to be most suitable for
conducting the pilot study. Riverside also met the selection criteria listed above, and
appeared to provide the greatest potential for observing a wide range of ambient
concentrations of inhalable particles (IPs). A wide range of outdoor levels in the sample
makes it easier to estimate a relationship between outdoor measurements and indoor or
personal measurements.
The area around March Air Force Base south of Allessandro Boulevard was excluded
from the target population because it was expected to have different IP levels from the rest of
Riverside. The IP levels in that area are affected by the Air Force base and by separation
from the rest of Riverside by foothill mountains. Comparisons of IP measurements from this
area to a temporal fixed-site station in Riverside would not be meaningful.
The city of Riverside was operationally defined for selecting sample areas in terms of
1980 Census geographic units. The current (1990) city limits were approximated by including
1980 Census blocks that had been annexed since 1980. The block records added to the
1980 Riverside blocks to construct the sampling frame were:
• Tract 301, blocks 114 and 902
• Tract 414.02, blocks 104 and 105
• Tract 422.01, blocks 904, 905, and 906
• Tract 422.04, blocks 101, 103, 104, and 106
• Tract 425.02, blocks 101 through 104.
No extra effort was necessary to exclude the area around March Air Force Base south of
Allessandro Boulevard because this area was not included in the 1980 city limits of Riverside.
Target Time Period
Data collection was scheduled for the Fall of 1990 because temporal fixed-site
monitoring data suggested that the ambient air in Riverside has the greatest range of IP levels
during this season. Monitoring during a time period with highly variable outdoor air levels of
IPs facilitates estimation of relationships between outdoor measurements and indoor or
personal measurements.
4-3
-------
For the same reasons, longer data collection periods were considered (e.g., a full
year). However, the available resources were not sufficient to extend the data collection to
much more than 6 to 8 weeks.
Target Aoe Range
Including a broad age range in the study makes it easier to estimate relationships
between age and personal exposures. However, very young children cannot cany the
personal exposure monitor. For children who can carry the monitor, the air route of exposure
might not be sufficient; ingestion may be an important route of exposure to particulates for
children. Even children who are physically able to carry the monitor would be less likely than
adults to wear it because it would interfere too much with their normal activities. Another
potential problem is that young children might not be able to provide reliable time-activity data.
Because of their activity patterns, children are likely to have higher exposures to IPs than
adults (as was suggested by the Prepilot PTEAM data). Given all these considerations, age
10 was selected as the minimum age for inclusion in the target population. Children aged 10
or older were not expected to have difficulty carrying the monitor.
Excluding Smokers
If smokers had been monitored in the PTEAM study, their personal exposure
measurements would have underestimated their actual exposures to IPs because the monitors
cannot account for the particles inhaled directly with tobacco smoke. Most analyses would
have been restricted to the nonsmoking subpopulation, for whom the personal exposure
measurements would more accurately reflect actual exposures. Budget limitations limited the
PTEAM sample size to about 175 people, which was minimal for the primary study objective
of estimating a population distribution of exposures. Because the sample size was small and
because smokers would have to be excluded from most analyses, we decided to simply
exclude smokers from the target population. Consequently, we also excluded homes in which
all household members aged 10 or older were smokers. However, nonsmokers living in
homes with other smokers were included in the target population.
4-4
-------
STATISTICAL SAMPLING DESIGN
Probability sampling methods were used to select people to participate in the Pilot
PTEAM Study because of the need to extend inferences from the sample to the target
population from which the sample was selected. Probability sampling refers to selecting units
into the sample in such a way that every member of the target population is assigned a
known, positive probability of selection. The sample was designed to yield complete data for
approximately 175 nonsmokers aged 10 or older. Environmental samples and questionnaire
data were actually collected for 178 population members.
Probability sampling methods were used to select a sample of person-days with known
probabilities so that inferences can be extended reliably from the sample to the target
population. A stratified, multistage area household sampling design was used because the
environmental samples must be collected in-person and because face-to-face interviewing is
necessary to obtain the highest response rates. Moreover, field data collection is more
efficient for a clustered sample of households than for a more geographically disperse sample
of households.
Geographic areas were selected at the first stage of sampling. At the second stage,
brief interviews were conducted for a sample of housing units within the selected areas to
identify the household members eligible for the PTEAM study. These data were used by the
interviewer to determine which individual, if any, to select from the household for personal
exposure monitoring. People were selected so that overall person-level probabilities of
selection were approximately equal, except for people who worked at least 30 hours per week
outside the home and people who were exposed to passive smoking in their homes. The
latter subpopulations were slightly oversampled to obtain sufficient sample sizes to support
separate statistical analyses.
At the last stage of sampling, the day of participation was to be selected at random for
each sample subject. Person-days with known probabilities should be selected for monitoring
with known probabilities because the population units observed are person-days. To make
the most defensible inferences from such population units, every population unit (defined in
time and space) should have a known, positive probability of selection (see Gilbert, 1987).
However, random assignment of sample persons to monitoring days was abandoned on
September 21 (the day before environmental sampling was to begin) because it was resulting
4-5
-------
in unfilled monitoring appointment slots that would have lengthened the field data collection
period and increased field data collection costs.
Details of the sample selection process are provided in the subsections that follow.
First-Stage Sampling of Geographic Areas
The first-stage sampling units (FSUs) were defined in terms of 1980 Census blocks
because they were the smallest Census-defined areas that provided complete coverage of the
target area. The estimated number of occupied residences in each area was used as a size
measure for the first stage of sampling. Because the 1980 Census counts were considerably
out of date, we began by updating these counts.
One goal of the sampling design was to obtain an equal probability sample of housing
units at the second stage of sampling. To accomplish this goal we required a reasonably
accurate estimate of the number of housing units in each block when the first-stage sample
was selected.
Because the 1980 Census data were out of date, we purchased a data tape from a
marketing firm (Donnelley Marketing Company) that contained their estimate of the number of
housing units in each block group (group of blocks with the same leading digit) in Riverside,
California. Their counts were based on telephone directory listings, vehicle registrations, etc.,
and were more up to date.
We compared the 1980 Census counts for block groups to Donnelley Marketing's
counts. Block groups that had grown considerably since 1980 (over 100 additional housing
units and over 50 percent increase) were identified for updating the block counts. Eight block
groups containing 98 Census blocks were identified for updating.
Maps of the identified block groups were prepared. An interviewer experienced in field
counting and listing of housing units for area surveys was recruited. This person provided
updated counts of housing units for all 98 blocks.
Blocks were combined, as necessary, to form FSUs expected to contain a sufficient
number of occupied housing units for the second-stage screening interviews (i.e., 30 or more
housing units). A stratified sample of 36 FSUs (area segments) was then selected with
probabilities proportional to size using a probability minimum replacement (pmr) sequential
sampling algorithm (Chromy, 1979).
4-6
-------
Because the ambient levels of IPs were expected to be fairly uniform in Riverside, the
sampling frame was stratified primarily to ensure that the sample areas were representative of
the target population. Therefore, approximately the same sampling rates were used for all
strata.
The sampling frame was stratified by (a) geographic area, (b) housing unit values, and
(c) prevalence of detached, single-family dwellings. This stratification increases the precision
of survey statistics if IP measurements are correlated with the strata.
Four geographic strata were defined using as approximate boundaries the Riverside
Freeway (which divides the target area roughly in half from northeast to southwest) and a
north-south line approximately through the middle of the city. The Census tracts defining the
four geographic strata within the city of Riverside are shown in Table 4-1.
The average appraised value of the housing units in each FSU computed by combining
the 1980 Census data for owned and rented housing units. For rented dwellings, the
appraised value was estimated as 100 times the monthly rent. The distribution of housing unit
values was examined separately for each geographic stratum to define high and low
socioeconomic strata within each geographic stratum.
Likewise, the distribution of the proportion of detached single-family dwellings was
examined for each socioeconomic stratum within geographic strata to define strata with high
and low proportions of single-family dwellings.
The sample of 36 FSUs was proportionally allocated to the four geographic strata.
Within each geographic stratum, the FSUs were selected with probabilities proportional to size
(pps) using a sequential probability minimum replacement (pmr) sampling algorithm (Chromy,
1979). The size measure for each area was the 1980 Census count of occupied housing
units for the area or, if available, the recently updated field count of housing units. The
sampling frame for each geographic stratum was sorted in a serpentine manner by the
following variables:
(1) socioeconomic status (high or low),
(2) proportion of detached, single-family dwellings (high or low), and
(3) the size measure.
Sequential selection from the serpentine-sorted sampling frame ensured proportional
representation of the strata formed by the first two variables, as discussed by Williams and
4-7
-------
TABLE 4-1. CENSUS TRACTS DEFINING GEOGRAPHIC STRATA
Stratum3 Census Tracts
Northeast
CO
o
302,
303, 307, 308, 31
1, 402, 404,
423
Southeast
304,
305,
306, 312, 422.01,
422.02, 422.03,
422.04, 425.01
Southwest
313,
317,
414.01,
414.02
Northwest
309,
310,
314.01,
314.02, 315.01, 315.02,
316,
401,
408.02, 409,
410, 411,
412, 413
a Geographical subdivision of Riverside, CA.
4-8
-------
Chromy (1980). The final sorting by the size measure helped ensure that both large and
small FSUs would be represented in the sample. Proportional allocation to all strata and
ppsselection of FSUs at the second stage of sampling using updated size measures,
facilitated the selection of housing units with approximately equal probabilities.
Stratification by proximity to major highways was also considered as a potential
way to stratify with regard to ambient air sources of aerosols. However, we concluded
that highways would only act as sources for homes within a few hundred yards. Such
geographically small strata would be difficult to construct and would receive rather small
sample sizes, unless they were oversampled. Oversampling the strata would adversely
affect precision for overall population estimates. Therefore, we decided not to stratify by
proximity to highways.
The expected frequency of selection, %(r,i), of the lib FSU in stratum rcan be
represented as
S(r,i) I S(r,+), (4-1)
where n^r) is the number of FSUs to be selected from stratum r, S(r,i) is the size
measure for the Ah FSU in stratum r, and S(r,+) is the total of the size measures for
stratum r. The expected frequencies of selection were all less than one and, hence, can
be considered probabilities of selection.
A sample of 36 FSUs was considered appropriate, given the cost and precision
constraints of the survey. A larger first-stage sample was not recommended because of
the travel costs associated with setting up, calibrating, and collecting the environmental
monitoring instruments in the sample homes. A smaller first-stage sample size was not
recommended because of the potential for loss of precision with larger cluster sizes.
Selecting 36 FSUs resulted in an average of about 30 homes screened and five persons
monitored per sample area. Based on past experiences with personal exposure
monitoring studies, these cluster sizes were considered satisfactory.
The FSUs selected into the first-stage sample are referred to as sample
segments. RTI prepared a field packet for each sample segment. Each field packet
contained a map showing the boundaries of the segment and an estimate of the number
of housing units in the segment.
4-9
-------
During June and July of 1990, approximately three months prior to the beginning
of PTEAM field data collection, field staff were sent to the sample segments to list the
current housing units for use as the second-stage sampling frame. Their first task in
each area was to make a quick count of the total housing units in each sample segment.
If any quick count differed by 30% or more from what was expected (based on the 1980
Census counts and updates), the field staff checked with the RTi sampling staff to verify
that they were working the correct area. If they were working the correct area and there
were too many housing units present to efficiently list all of them (approximately 200 or
more), the sample segment was divided into subsegments, a quick count of the current
housing units was obtained for each subsegment, and one subsegment was selected
with probability proportional to the quick count. A subsegment was selected for six
sample segments. In these cases, only the housing units in the sample subsegment
were listed (instead of all housing units in the sample FSU) for the second stage of
sampling.
The procedure for calculating the conditional probability of selecting a
subsegment is defined here. Letting ^,2 N^(r,i) index the subsegments created for
the /th FSU in stratum "i", M^(r,i,j) was defined to be the quick count of housing units in
the th subsegment. Then, the conditional probability of selecting the (r,i,j)-th
subsegment, given that the ith segment was selected for the first-stage sample, Sv is
W1 (r,i) M-\ (r,i,j) if the (r,i)-th segment
E was subsegmented
*
(4-2)
1 if the (r,/)-th segment was not subsegmented.
The unconditional probability of selection for the (r,/,»-th subsegment is then given by the
product of (4-1) and (4-2).
The first-stage sample selection process is summarized in Figure 4-1.
4-10
-------
36 FSUs (area segments) selected
using sequential sampling with
probabilities proportional to size
1980 Census housing unit counts
updated based on current field
counts for 98 Census blocks
Subsegment selected with probability
proportional to the field count for
6 area segments
First stage sampling units (FSUs)
defined from 1980 Census blocks
covering the target portion of
Riverside, California
Frame stratified by:
(a) geographic area (NE, SE, SW, NW);
(b) housing unit values (high/low);
(c) proportion detached single-family (high/low).
Figure 4-1. First-stage Sample Selection Process
4-11
-------
Second-Stage Sample of Homes for Rosterino
Having located the sample segment or subsegment, the field staff listed all
potential housing units (referred to as lines) in each selected area. Given the number of
lines (potential housing units) listed for each segment, a sample of 780 lines was
allocated to the segments for enumeration interviews to achieve approximately equal
probabilities of selection for all housing units in the target population. This resulted in
larger sample allocations to the segments that had experienced the most unanticipated
growth relative to the size measures used for sample selection.
The sample size was designed to yield 175 completed personal exposure
monitorings with one person selected from each of 175 homes. An additional "hold"
sample of 60 sample lines (potential housing units) was selected and held in reserve for
use if the response from the segments worked first should suggest that the 780 sample
lines would not be sufficient. None of the "hold" sample lines were ever fielded. In fact,
100 of the original 780 sample lines were never fielded. Thus, the total number of
sample lines fielded was 680.
The procedure for calculating the conditional probability of selecting a housing
unit is defined here. Letting k=\,2,...,N2(r,i,j) index the lines (potential housing units)
listed for the th sample segment (or subsegment), define to be the sample
allocation to the (r,/,/)-th segment for the primary sample of 780 lines. Given the sample
allocation, m2(r,i,j), an equal probability sample of housing units was selected from those
listed for each segment using a sequential probability minimum replacement (pmr)
selection algorithm (Chromy, 1979). Therefore, the conditional probability of selecting the
(f,/,/,*Mh listed housing unit, given that the (r,/,y)-th segment was selected into the
first-stage sample, S.(, is
P2(rMk\jeSi) = m2(r,ij) I N2{r,i,j). (4-3)
The area segments selected for each geographic stratum were randomly
allocated to three "waves" for data collection. The waves were used to distribute home
monitoring evenly throughout the field study effort. The randomization was restricted to
allocate approximately the same number of segments from each stratum to each wave
and to result in exactly twelve total segments allocated to each wave. When the sample
yield from the first two waves (in terms of completed monitoring appointments) was
4-12
-------
greater than expected, 100 sample lines were randomly deleted from the third wave by
selecting a systematic subsample from the 262 sample lines originally assigned to
Wave 3.
Because all segments had the same probability of being assigned to the third
wave, the subsampling affects the probabilities of selection for all sample lines. In
particular, after selecting 162 of the 262 Wave-3 lines to be retained, the conditional
probability of selecting the (r,/,/fc)-th listed housing unit, given that the th segment
was selected Into the second-stage sample, is
P*2(rUk\'£S{) = (2/3)P2 + (1/3)(162/262) P2 , (4~4)
where Pg refers to the conditional probability for the original sample of 780 lines given by
(4-3). The factor of "1/3" results from randomly assigning one-third of the segments to
Wave 3, and the factor of "162/262" results from retaining 162 of the original 262 sample
lines in Wave 3 with equal probabilities.
A "missed housing unit procedure* was employed when sample housing units
were identified in the field to ensure that all housing units that could be identified at the
time of field data collection had a positive probability of being included in the sample.
This procedure included in the sample not only the housing units listed on the selection
sample lines, but also
• any non-listed housing units located within the selected sample housing units,
and
• any non-listed housing units located between a selected sample housing unit
and the next listed housing unit.
The probability of selection for each "added housing unit" is the same as that for the
listed sample housing unit that resulted in its inclusion in the sample.
Third-Staoe Sampling of People and Homes for Monitoring
Two basic approaches for sampling participants were considered: (1) selecting
the participants during the household screening interviews and (2) selecting them after all
screening interviews had been completed and compiled into an automated data base.
Table 4-2 summarizes our assessment of the relative advantages of these approaches
4-13
-------
TABLE 4-2. RELATIVE ADVANTAGES AND DISADVANTAGES OF SELECTING
PARTICIPANTS DURING THE HOUSEHOLD SCREENING INTERVIEWS AND
AFTER ALL SCREENING INTERVIEWS HAVE BEEN COMPLETED
Sample
Characteristic
Sample During Screening
Interviews
Sample After All Homes
Screened
No. sample persons
No. participants
Time lag between
screening and
monitoring
Opportunity for
stratification
Person response
rate
random variable
depends on household screening
response rate, person response
rate, distribution of household size,
and distribution of smokers by
household size
1 day to 5 weeks
must be kept simple
should be 2% to 5% higher
fixed
depends on person response
rate
2-4 months
can make full use of
screening data
may be 2% to 5% lower
Cost
lower cost
higher cost
4-14
-------
prior to implementation. Based on the considerations outlined (primarily the shorter time lag
between screening and monitoring), we decided to select sample persons in the field during
the household screening interviews.
A preprinted label affixed to the household roster was used to identify which
household member, if any, was selected for personal exposure monitoring. The selection
algorithm was designed so that the overall probabilities of selection were approximately equal
for all members of the target population. However, people who worked at least 30 hours per
week outside the home and people exposed to passive smoking in their homes were selected
into the sample with slightly higher probabilities than the remainder of the population to slightly
increase the sample sizes for analyses of these subpopulations. People residing in
households with more than three survey-eligible household members (nonsmokers aged 10 or
older) had slightly lower probabilities of selection than the remainder of the population. The
additional screening interviews that would have been necessary to equalize the selection
probabilities for members of larger households were not warranted because of the small
number of households of that size (less than 5% of the population).
IP concentrations were monitored indoors in the "primary living area" and outdoors
whenever a household member was selected for personal exposure monitoring. Because the
Prepilot PTEAM data indicated that 12-hour indoor IP concentrations were generally
comparable in all rooms of a home, we chose not to randomly select sample rooms in the
homes monitored (Clayton et al., in press).
The probability that a home was selected for monitoring is the probability that at least
one of its household members was selected for personal exposure monitoring. Because the
person-level probabilities of selection were approximately equal, the probability of selecting a
home for monitoring was approximately proportional to the number of survey-eligible
household members. Thus, the probabilities of selection for homes have greater variability
than those for the individuals monitored within the homes, which results in some variance
inflation and loss of precision for household-level analyses relative to person-level analyses.
To achieve approximately equal probabilities of selection, people were selected for
monitoring in screened households using a multistage process. Households containing only
one or two survey-eligible household members (nonsmokers age 10 or older) were selected
for monitoring with probabilities proportional to the number of survey-eligible household
members. All households containing three or more survey-eligible household members were
4-15
-------
selected for monitoring. After a home was selected for monitoring, a single survey-eligible
household member was selected at random for personal exposure monitoring. The
conditional probability of selecting Person p from the (r.ij.k)-th housing unit, given that the
housing unit was selected for the second-stage sample, S2, is
'l/3 if A/3(r,/,;,*) * 3 (4.5)
1/A/3(r,/,/,*) if N3(r,i,j,K) > 3
WV.MteS2) =
where A/3(r,/,/,/c) is the number of nonsmoking household members aged 10 or older in the
(r,i,j,k)-th housing unit.
Oversampling of people who worked at least 30 hours per week outside the home and
of people who were passive smokers in their homes was achieved by rejecting 30% of the
sample members who were not so employed or lived in a home with no smokers. When an
individual was rejected, the home was not monitored either. This mechanism was used for
the oversampling because it could be implemented easily with sample selection occurring in
the field during the screening interviews. Thus, the conditional probability of retaining the p-th
household member selected from the th sample household, given that the p-th member
was initially selected into the third-stage sample, S3, is given by
P4(r,i,j,k,p\£S3) =
r
1.0 if Person "p" works at least 30 hours per
week outside the home or (4-6)
any household member is a smoker
0.7 otherwise
Therefore, the unconditional probability of selecting Person p from the (r,i,j,lc)-th housing unit is
the product of (4-1) (4-2), (4-4), (4-5), and (4-6).
Identifying the sample members in the field was facilitated by pre-selecting the sample
and printing the selections on labels affixed to the household enumeration questionnaires.
Each label identified, by household size, the roster line number of the person selected for
monitoring. Equal initial probabilities of selection were achieved for people in homes
containing only one or two survey-eligible members by sampling members from such homes
as if they all contained exactly three eligible people and selecting each home only when one
of its eligible members was selected. For example, if a household contained two eligible
members and Person 2 was pre-selected for monitoring, then Person 2 would be monitored.
4-16
-------
However, if the home had only one eligible member and Person 2 was pre-selected (as
indicated by the preprinted label), then no one in the home would be monitored. In addition,
the label indicated whether or not the selected person would be rejected if that person was not
employed at least 30 hours per week outside the home or lived in a home with no smokers.
This process resulted in identification of 257 sample persons.
Since indoor and outdoor particulate monitors were placed at a home whenever any
member of the household was selected for personal monitoring, the conditional probability of
selecting a home for indoor and outdoor particulate monitoring is the sum of the probabilities
of selection for personal monitoring for all the household members. Therefore, the conditional
probability of selecting the (r,/,/,/r)-th housing unit for monitoring, given that the fcth housing unit
was selected into the second-stage screening sample, S^ is given by
NsirMQ
Pm{r,iJ,k\kES2) = £ P3< r, p j /ceS2) P4(r, /./,/c.p | peS3). (4"7>
p=1
The unconditional probability of selecting the (r,/,/,*)-th housing unit for monitoring is given by
the product of (4-1), (4-2), (4-4), and (4-7).
A subsample of the housing units selected for monitoring particulates was also
selected for monitoring phthalates and polycyclic aromatic hydrocarbons (PAHs). The sample
selection label for each sample housing unit was preprinted with an indicator variable that had
the following values regarding the subsample for phthalates and PAHs:
I - Indoor air only
B - Both indoor and outdoor air
N - Neither indoor nor outdoor air.
These indicators were randomly assigned to the sample housing units at rates of 60/175 for
both "I" and "B" and 55/175 for "N" with the goal of collecting indoor phthalate and PAH
samples in 120 homes and outdoor phthalate and PAH samples in 60 homes selected at
random from the targeted 175 homes with particulate monitoring.
When the sample yield from the first two waves was analyzed, we found that the
subsample for phthalates and PAHs was yielding fewer monitored homes than desired.
Hence, after the Wave-3 sample had been randomly subsampled from 262 sample lines down
to 162 sample lines, a systematic random sample of 46 Ns were changed to Bs. Therefore,
4-17
-------
the rate of selection for both indoor and outdoor phthalate and PAH monitoring was increased
by 46/162 for Wave-3 sample lines. Since ail area segments had a one-third chance of being
randomly assigned to Wave 3, the conditional probability of selecting the housing
unit for indoor monitoring of phthalates and PAHs, given that the housing unit was selected
into the second-stage sample, S2, is given by
Pl(r,i,j,k\kES2) = (2/3)(120/175) + (1/3)[(120/175) + (46/162)]. (4-8)
Likewise, the probability of selection for outdoor monitoring of phthalates and PAHs,
given that the (r,/,/*)-th housing unit was selected into the second-stage sample, S2, is given
by
P0(r,iJ,k\kES2) = (2/3)(60/175) + (1/3)[(6G/175) + (46/162)]. (4-9)
The process of selecting persons and homes for monitoring is summarized in
Figure 4-2.
Fourth-Staae Temporal Sampling
The final stage in the statistical sampling design was to randomly select a day for
participation for each sample subject. However, as previously discussed, random assignment
of sample persons to monitoring days was abandoned on September 21 (the day before
environmental monitoring began) because it was resulting in unfilled monitoring appointment
slots. Approximately half of the appointment slots in the first few days of monitoring had not
been filled by September 21.
This section describes the temporal randomization procedure that was implemented
during the two weeks prior to September 21, our experience with that process, and the
procedure used to assign people to monitoring days for the remainder of the PTEAM data
collection.
The initial PTEAM random day selection was performed in two stages. First, the 36
area segments were randomly ordered for field data collection within each of six geographic
strata constructed to correspond with interviewer assignments. Second, each interviewer
selected one day at random for each participant from the first seven days available to the
4-18
-------
680 sample lines selected
from all housing units listed
for the 36 area segments
632 eligible sample housing
units identified; 443 screening
interviews completed
257 sample persons selected;
subsample of 181 housing units selected
for Indoor phthalate and PAH monitoring;
subsample of 98 units selected for
outdoor phthalate and PAH monitoring
Figure 4-2. Second- and Third-stage Sample Selection Process
4-19
-------
interviewer for appointment scheduling. Each interviewer made appointments on either the
even or the odd dates. Thus, the randomly selected days were selected from approximately a
two-week period. The available days became fewer as days were filled because the
interviewers were not allowed to make appointments further than two weeks in advance.
Randomly assigning the day of monitoring was regarded as an experimental
component of the PTEAM data collection program. Concerned that response rates might be
unacceptabiy low if sample subjects were required to participate on the day randomly selected
for them, we allowed participants to choose any available appointment slot for participation if
they could not participate on the selected day.
Up to the point where random day selection was abandoned, almost 90 percent of the
participants had agreed to be monitored on their randomly selected day. One problem that
resulted in open appointment slots in the first few days of the schedule (prior to abandoning
the random day procedure) was that one interviewer had difficulty implementing the PTEAM
procedures; she conducted very few interviews and resigned after about two weeks. Most of
the dates that had been assigned to her were unfilled when environmental sampling was
scheduled to begin. These problems may have been alleviated had more interviewers been
hired or participant recruitment had begun earlier.
After the random day selection was abandoned, appointments for environmental
monitoring were scheduled for the first available appointment slot whenever possible. This
maximized the chance that all available appointment slots would be filled, making optimum
use of the field staff's time. This was also considered useful from the standpoint of reducing
the potential for bias that could result from letting each participant choose their day of
participation. Thus, the field staff were instructed to schedule the participants for the first
available monitoring appointment and to emphasize the importance of participating on that
date exactly as if it were the randomly selected date.
If the sample subject indicated that there would be a problem with participating on the
assigned day (either the randomly selected day or the first available day), the interviewers
were instructed to determine why that day seemed to be a problem and attempt to alleviate
the person's concerns. If the problem was related to the activities that the person would be
doing that day, the importance of representing all types of activities was to be explained, arid
the interviewers were instructed to suggest ways to keep the monitor in close proximity while
4-20
-------
doing various types of activities. Ultimately, 84% of the participants agreed to be monitored
on the assigned day.
4-21
-------
-------
SECTION 5
SURVEY OPERATIONS
INTRODUCTION AND OVERVIEW
Survey research staff were involved in the conduct of several aspects of the field
operations of this study including the design of the study and the preparation of the data
collection instruments. Survey staff supervised the interviewers who accomplished the various
field activities, including the count and list necessary for creating the sample, the screening
and respondent selection, and the interviewing of the selected respondents. Survey staff
worked with the monitoring staff to develop schedules for collection of samples and to
coordinate the flow of sampling time information and respondent data. Procedures used to
process data from the various data collection instruments were developed and implemented.
Each of these activities is described in detail in this section.
Survey Preparations
Before initiating field efforts, survey staff undertook several activities required to
prepare for or to support the field data collection process. These included questionnaire and
study design development, preliminary sampling activities and material preparation. Most of
these activities were accomplished concurrently.
Study Design and Questionnaire Development ~
Survey project staff served on the committee established by EPA to design the field
component of this study. The committee met numerous times and developed study plans that
were presented to the full project staff. The committee used the evolving study plans to
formulate the questionnaires needed to collect the data required to satisfy the analytical needs
of the study. As the study design changed, so did the questionnaires. Multiple drafts were
prepared and circulated for review within the committee and among the project staff. After
internal consensus was approached, the draft documents were sent to the two sponsoring
organizations, EPA and CARB. Reviewer comments were included as new drafts were
prepared and circulated. The development of the documents continued through many
iterations until the design of the study and the determination of data needs was completed.
5-1
-------
After the study design was finalized and final data collection instruments were
prepared, survey staff prepared requests for study approval from the Federal Office of
Management and Budget (OMB) and from the RTI Institutional Review Board (IRB). The
OMB supporting statement was prepared using sections provided by other RTI staff and EPA
project personnel. The draft statement was circulated among the RTI project staff and
revised. The revised version was sent to EPA project staff for initial review. Comments were
received and incorporated. The next draft was sent to EPA for formal review prior to
submission to OMB. Final comments were incorporated and a final version was prepared and
submitted to EPA. The supporting statement contained all of the mandated sections as
shown in Appendix A, Table of Contents for OMB Supporting Statement.
We prepared a project protocol and submitted it to the RTI IRB. This committee
reviews protocols along with the data collection instruments and other documents used In the
field to assure compliance with all of the regulations concerning the use of human subjects in
research. Following a verbal presentation to the committee, we responded to questions and
received approval to field the study as designed.
Preliminary Sampling Activities -
At the request of the sampling staff, survey staff hired personnel in Riverside to go to
designated sample areas and conduct quick counts of housing units. This information was
used to update Census data and aid the process of selecting sample segments for inclusion in
the study.
Using the data provided from the quick counts, the sampling staff selected the
segments from which the monitoring sample was selected. The first step in the sampling plan
was the "count and list*. We recruited and hired interviewers who were experienced in this
procedure. The interviewers were trained by an on-site field supervisor and were given
assignments.
Following survey standard procedures, the Interviewers located each assigned
segment and confirmed its boundaries. A quick count of the number of housing units was
completed and used to determine if subsegmenting was required. Finally, each housing unit
was listed with unique identifiers and was located on the segment sketch.
The initial segment listing completed by each interviewer was reviewed completely by
the field supervisor, as was a sample of remaining work. Immediate feedback was provided if
5-2
-------
errors were found. As listings were returned to RTl, they underwent additional quality checks
in the sampling department, and interviewers were contacted to resolve problems. Completed
segment listings were keyed and the data used to select the sample of housing units to be
contacted and screened.
Field Material Preparation ~
As the study design and data collection instrumentation needs reached closure, we
were able to develop the various supporting material and procedures needed to field the
study. Documents developed included the participant consent form (Appendix B), a
supplemental consent form for guardians of minor respondents (Appendix C), a form for
documenting refusals and subsequent conversion efforts (Appendix D), and a lead letter and
informational brochure (Appendix E). We also developed or obtained various interviewer
supplies, such as 'Sorry I Missed You' cards, appointment calendars, interview scheduling
forms and appointment reminder cards. We developed a Field Interviewer Instruction Manual.
which was used by the interviewers during training and as a resource during the field data
collection period. The Table of Contents of this document appears in Appendix F.
We developed procedures for scheduling appointments on randomly selected days,
procedures for creating unique participant identification numbers in the field, and procedures
for scheduling appointments. We worked with the sponsoring agencies and the city of
Riverside to develop a public relations strategy, including contacts with the media and a press
release.
As documented previously, we tested most of the study materials in a series of focus
group meetings held in Riverside. We were particularly concerned about the material
developed for use in the initial contacts with selected households. During the focus group
meetings, we displayed and discussed versions of the mailing envelopes, lead letter, and
informational brochures to be used in lead mailings to the addresses of the selected sample
households. We also discussed incentive payments and interviewer scripts to be used during
contacts. We asked respondents what would make them likely to participate and what might
make them refuse. Almost all of the focus group participants expressed an interest in being
part of the study. We used the information gathered during the meetings to revise the
material and presentations used during the field effort.
5-3
-------
Recruit. Hire, and Train Field Staff
An experienced staff of field interviewers led by an experienced supervisor is vital to
the successful completion of an effort like PTEAM. We began by recruiting a field supervisor
from the South Coast area. RTI's Regional Supervisor from Los Angeles recommended an
experienced supervisor whose time commitments where ideal for the effort required for this
prefect. We contacted her, explained the study and described the effort required, and she
accepted the assignment. She immediately began the process of recruiting the interviewing
staff. With the assistance of the Regional Supervisor in Los Angeles, we determined the
number of interviewers needed and the number of hours each would be required to work
during the field period. Initial attempts were made to recruit from the area in and immediately
adjacent to Riverside. When it became apparent that there were not enough qualified staff in
the limited area, we expanded the recruiting area. As the size of the effort and the amount of
work required from each interviewer became more apparent, we decided to increase the
number of interviewers and to recruit at least one bilingual interviewer.
After we identified a pool of applicants, we contacted each one and provided details
about the study and the commitment required. We reviewed references for each applicant
interested in participating in the study. The final list of applicants was prepared and job offers
sent to each, along with the necessary paperwork to complete the application process.
A multi-day training session was held in the first week of September. Each interviewer
received a copy of the Field Interviewer Instruction Manual and was given an allocation of
home study time before the training session. Classroom training consisted of a review of the
study and the requirements for successful completion. We provided detailed instructions for
each step of the process. We reviewed each data collection instrument on a
question-by-question basis. Trainees participated in mock interviews as a group and
individual practice interviews in pairs. Each interviewer practiced and demonstrated the
introduction at the household and the motivational material used when talking to respondents.
Each step of the data collection process was reviewed and practiced. All assignment
materials were prepared as a group and checked before the interviewers went to the field.
We assigned each interviewer an allotment of days on which they could set the initial
monitoring appointments, and then practiced the process of determining the first random day
on which each appointment would be offered.
5-4
-------
Each interviewer was required to demonstrate adequate performance in all
components of the study before leaving the training site and beginning to contact potential
respondents. Interviewers received assignments and began work immediately after
completing training. Project staff remained in Riverside for several days after training to assist
the interviewers if problems arose, to assist the supervisor as she established the required
flow of information, and to provide moral support to the field interviewing staff during the first
few days of the effort. In addition, the survey staff completed some local logistical
arrangements for the monitoring teams prior to their arrival.
Screening and Sample Selection
Each interviewer received an assignment consisting of several sampling areas,
grouped geographically to minimize travel. The materials used for each segment of the
assignment consisted of the segment sketch and listing forms created during the initial count
and list process, and a set of labeled Household Enumeration Questionnaires (HEQs)
(Appendix G). Each questionnaire was uniquely labeled with address information and a line
number corresponding to the segment listing form. Each questionnaire thus represented one
selected housing unit that was contacted and worked to completion.
After planning their work loads, the interviewers began to contact the housing units in
their assignments. Choosing times carefully to maximize the probability of a successful
contact on the first attempt, the interviewer approached the housing unit and attempted to
complete the HEQ. Information on the HEQ was obtained only from residents of the housing
unit who were at least 16 years of age. The HEQ created a roster of the residents of the
household, excluded the non-eligible residents, and guided the interviewer through a sample
member selection process.
Once a sample member was selected, the interviewer made an immediate effort to
contact that person. If the person was at home, the interviewer attempted to begin the
recruiting process right away. If the respondent was not at home, or was unavailable at the
time, the interviewer made a specific appointment to return to complete the enrollment and
interview process.
Each housing unit assigned to an interviewer was worked until a final result code could
be assigned and documented by circling a number in Section C on the first page of the HEQ.
Code 09 - Screening Complete - was the desired code. Some of the sample housing units
5-5
-------
were ineligible because they were vacant, or not housing stock, or were temporary homes.
Other units could not be screened because of refusals to provide the information, language
barriers, or the fact that no one was at home, or no one was eligible to provide the data, even
after repeated return visits to the housing unit. Interviewers were required to visit each
housing unit multiple times, on different days, and at different times of the day. Each visit
was documented in the Record of Calls, Section B of the HEQ. The Record of Calls was
reviewed by the supervisor before the interviewer could terminate work on the case.
Interviewers reported progress on their assignments to the field supervisor during
scheduled phone reports or routine meetings. The supervisor compiled the data on the
screening effort and reported it to survey staff electronically and during routine phone calls.
The progress reports were used to determine estimates of response rates and overall
progress judged against the time available to complete the effort. This information was used
to estimate the number of supplemental cases that had to be fielded and the need for
additional interviewing staff.
Completing the Interview and Setting Appointments
As soon as was practical after completing the screener and selecting the sample
respondent, the interviewer met with the potential respondent and began the enrollment
process. The interviewer explained the study in detail, stressing the requirements of complete
participation. Photographs of the monitoring equipment and typical placements were used as
part of the explanation. After answering any questions about the study, the interviewer gave
the participant time to read the consent form. If necessary, the interviewer read the consent
form to the participant. The bilingual interviewer prepared a Spanish version of the consent
form and other introductory material, to be used as a guide, but not as a replacement for the
original material.
After completing the informed consent process, the interviewer completed the Study
Questionnaire (Appendix H) with the respondent. The final step in the process was setting
appointments for the monitoring team to come to the housing unit, set up the equipment, and
subsequently retrieve it. Initial attempts were made to establish monitoring appointments in a
random fashion. Each interviewer had a list of unique dates available to them. Using a
random number assignment from the sampling label on the HEQ, the interviewer determined
which day from their list of available days was selected. This was the day that was initially
5-6
-------
offered to the respondent. If the respondent was unable to participate on the selected date,
the interviewer was allowed to offer any of the available days in the two week time window.
Shortly after the beginning of data collection, it became apparent that the scheduling
algorithm was not generating an appointment schedule that maximized the number on
appointments scheduled for each day. This would have created problems for the monitoring
teams by requiring them to remain in the field for a longer period than was originally
scheduled and budgeted. This became an unacceptable constraint aid the random
assignment process was dropped. The interviewers were told to schedule appointments in
the first available slot in their assigned days. This process succeeded in filling appointment
slots as quickly as possible.
When the monitoring team arrived at a housing unit, they checked for the signed
consent form, and provided details about the equipment being placed in the home. They told
the respondents what to do if there were problems with the equipment. The monitoring team
also provided additional details about the information that would be collected from the
respondent at the end of the monitoring period. Respondents were asked about their
activities during each 12-hour monitoring period. They were asked where they spent their
time, what they did, and what potential exposures to the material of interest they may have
had. This information was recorded on the Time Activity Diary and its supplement
(Appendix I). At the final visit, the monitoring team answered any final questions and paid an
incentive to the respondent before leaving. A final document completed by the respondent
was the Activity Modification Questionnaire, on which the respondent was asked to report any
changes from his or her normal routine caused by participation in the study. These
questionnaires were collected in an anonymous fashion and not linked in any way to the
respondents or their data.
All documents collected by the field staff or the monitoring team were returned to RTI
at the end of the field period. Consent forms and incentive receipts were separated from the
other documents, sorted by ID number and filed in secured storage. Since these forms
contain names and linkages to other data, they will be kept separate from other information to
assure the confidentiality promised to the respondents. The remaining data collection
instruments were batched for processing, which included editing for completeness and
legibility and internal consistency. After editing, documents were sent for keying and
5-7
-------
subsequently returned for storage. After machine edits, the keyed data was made available
for statistical analysis.
5-8
-------
SECTION 6
FIELD MONITORING AND METHODS
INTRODUCTION
Overview
The PTEAM field monitoring program was designed to collect sufficient samples to
characterize the exposure of a representative population in Riverside, California to inhalable
particulates. A total of 178 persons were monitored over two consecutive 12-hour periods.
During each monitoring period one personal PM^q, one indoor PM^, one indoor PM2 5, one
outdoor PM.|q, and one outdoor PMg 5 sample was scheduled for collection. Approximately
four persons were monitored each day from September 22 through November 9, 1990.
Additional samples were collected to measure nicotine levels and air exchange rates in each
study home to provide source characterization data. Under direction from the California Air
Resources Board, PAH/phthalate samples were collected at a subset of 125 homes and
outdoor PAH/phthalate samples were also collected at 65 of the 125 homes with indoor
samples. An ambient monitoring site was set up at one location in Riverside to monitor
ambient levels of particulates during each 12-hour period throughout the course of the study.
Reference method PM^q and dichotomous samplers were operated at this temporal site
alongside personal and stationary monitors identical to those used for participant monitoring.
A mobile weigh laboratory was set up in Riverside to perform all particulate filter weighing
operations. Protocols for sample collection were prepared and included in Volume II of the
PTEAM Pilot Study Work Plan (RTI, 1990a).
The schedule and procedures used to collect particle, PAH/phthalate, air exchange,
and nicotine samples are described in this section of the report because their collection is
interrelated. Data analysis of particle and elemental concentrations are included in Sections 9
and 10 of this report. Results and data for the other sample types will be prepared and issued
under separately; RTI will prepare a report to CARB describing the PAH/phthalate portion of
this study and Harvard will prepare a report for EPA describing source modeling based in part
on air exchange and nicotine data.
6-1
-------
Training and Rehearsal
Personnel from RTI, Harvard, and Acurex conducted a 3-day training and rehearsal
session at Research Triangle Institute during the third week of August, 1990. The purpose of
the session was threefold. First, all persons involved in the field effort were introduced to the
overall study design and received training in both logistical and technical aspects of sample
collection operations. Second, a dress rehearsal was conducted combining all aspects of
operations as they were to be carried out in the field. In this dress rehearsal two participants
were monitored for two 12-hour periods, the temporal site equipment was operated, and the
weigh trailer was operated. The rehearsal provided both realistic training and an opportunity
to evaluate sampling protocols, sampling logistics, and equipment function with sufficient time
for modification before the monitoring was begun in Riverside. The third purpose of the
training session was to allow personnel from the three organizations to work together and
coordinate activities.
Field Operations
Several types of air samples were collected during the field study including particles,
polycyclic aromatic hydrocarbons (PAHs), and phthalates. Air exchange samples and nicotine
samples were also collected for source characterization purposes. A personal exposure
monitor (PEM) was used to collect particles up to 10 nm aerodynamic diameter and nicotine in
the participant's breathing zone. The PEM sample was also collected at the temporal site to
assess comparability to reference method samplers. The other monitors used for particle
monitoring were the stationary indoor monitor (SIM) and the stationary ambient monitor (SAM)
for outdoors. The SIM and SAM were used to collect two particle fractions, up to 2.5 jim and
up to 10 nm. The SIM 10 jim monitor was also equipped to collect nicotine. Both SIM and
SAM systems were deployed at each participant's home and the SAMs were deployed at the
temporal site. The collection of PAHs/phthalates also took place inside and outside of a
subset of participant homes. Air exchange measurements were carried out at each home by
releasing a perfluorocarbon tracer at a constant rate and using a capillary adsorbent tube
sampler (CATS) to collect the tracer gas over the 12-hour monitoring period.
Particles were collected at the temporal site using a reference method PM^q sampler
(Wedding) that sampled for particles up to 10 nm at a high sampling rate. A dichotomous
particle sampler (dichot), which split the particulates into 2 fractions, was also used. One filter
6-2
-------
collected particles under 2.5 jim (fine) and another filter collected primarily particles from
2.5 jim to 10 urn (coarse). A cascade impactor was also used to measure the particle size
distribution over four time periods during the field study.
In support of the sampling a small meteorology station was operated at the temporal
site. This station did not operate well and little useful meteorological data was collected at the
temporal site. Supplemental local airport meteorological data were obtained.
A weigh trailer, specifically designed for weighing filters in a constant tempera-
ture/humidity environment, was employed for all filter weighing. This on-site facility was
necessary to process the nearly 3000 filters used during the study, each of which had to be
weighed before and after sample collection.
TECHNICAL APPROACH
Training and Rehearsal
A training and dress rehearsal session was conducted at RTI during the third week of
August, 1990. A total of 33 people participated in the training session including 14 from RTI,
10 from Acurex, and 9 from Harvard. Training sessions were conducted over the first day aid
a half, as shown in the schedule in Table 6-1. Training topics included field team organization
and logistics, sample collection methodology, sample and data tracking procedures, and
hands-on practice with sampling equipment and data manipulation. Each training topic was
covered by the organization and supervisor responsible for the activity.
Research Triangle Institute, as the prime contractor, was responsible for the overall
conduct of the field monitoring effort In Riverside. Two subcontractors, Harvard University
School of Public Health and Acurex Corporation, played major roles in the PTEAM particulate
monitoring field effort Each organization brought specific expertise to the effort and contrib-
uted jointly in the collection of samples. Harvard supplied the methodology and equipment for
personal and stationary monitoring at the participants* homes. Harvard was also responsible
for the collection of nicotine samples and measurement of air exchange rates in each house.
Acurex was responsible for deploying and operating the weigh trailer laboratory and the
monitoring equipment at the temporal site. RTI was responsible for directing field operations,
organizing field study logistics, recruiting and scheduling participation of the study population,
administration of all survey instruments, and collection of PAH/phthalate samples for the
CARB component of the study.
6-3
-------
TABLE 6-1. PTEAM 1990 PILOT STUDY
TRAINING AND REHEARSAL SCHEDULE
TRAINING
Day 1, Wednesday August 22, 1990
1. Introduction and overview of the study (Dr. Pellizzari) 8:45 am.
2. Participant selection (H. Zelon) 9:00 a.m.
3. Field organization (K. Thomas) 9:20 a.m.
a. Overview
b. General sampling schedule
c. Appointment schedules
d. Activities at each appointment
e. Personnel responsibilities
4. Sample and data collection methodology
a. Temporal site/weigh trailer (R. Clayton) 9:40 a.m.
b. Particulate collection (S. Froehlich) 10:00 a.m.
BREAK 10:30 a.m.
c. Air exchange measurement (S. Froehlich) 10:45 a.m.
d. PAH and phthalate collection (D. Whitaker) 11:00 a.m.
e. Questionnaire administration (M. Hoffman) 11:20 a.m.
5. Field logistics (K. Thomas) 11:40 a.m.
a. Travel
b. Lodging
c. Decorum
LUNCH 12:00 p.m.
6. Equipment and data collection training/practice 1:30 p.m.
a. Particulate collection (S. Froehlich)
-Harvard personnel, Acurex temporal site personnel
b. PAH and phthalate collection (D. Whitaker)
--RTI personnel
c. Questionnaire administration (M. Hoffman)
—Acurex personnel
BREAK 3:30 p.m.
7. Interteam equipment/procedure familiarization 3:45 p.m.
a. Personnel will rotate to observe equipment and
questionnaire, procedures will be demonstrated but
not practiced
(continued)
6-4
-------
TABLE 6-1 (continued)
TRAINING (continued)
8. Air exchange deployment
a. RTI field coordinators and Harvard supervisors will
go to the two rehearsal homes to set up air exchange
devices and prepare house plans
b. Arrive at Home 998
c. Arrive at Home 999
6:30 p.m.
7:45 p.m.
Day 2, Thursday August 23, 1990
9. Sample and data tracking
a. PAH/phthalate (D. Whitaker)
b. Particulate and air exchange (P. Reading)
--Bar codes
-Computer software
-Written forms (S. Froehlich & D. Whitaker)
c. Computer tracking practice
BREAK
DRESS REHEARSAL
1. Groups 1 and 2 prepare equipment/filters
a. Load filters into impactors
-Group 1 personnel loads
-Group 2 personnel loads
b. Prepare sampling equipment
-Pumps, computers, stands, carrycases, powercords,
paperwork, etc.
2. Group 1 initial sample visit at House 998
3. Group 2 initial sample visit at House 999
4. Temporal site monitoring begins
5. Field teams assemble at workroom
6. Group 1 second visit to House 998
7. Group 2 second visit to House 999
8. Temporal site monitoring filter changes
9. Group 1 unloads filters
8:45 a.m.
9:05 a.m.
9:30 a.m.
10:30 a.m.
10:45 a.m.
6:00 p.m.
6:00 p.m.
6:00 p.m.
Day 3, Friday August 24, 1990
6:00 a.m.
7:15 am.
7:15 a.m.
7:15 a.m.
10:00 a.m.
(continued)
6-5
-------
TABLE 6-1 (continued)
DRESS REHEARSAL (continued)
10. Group 2 unloads filters
11. Groups 1 and 2 clean impactors
12. Weigh personnel weigh used filters
13. Group 1 loads filters
14. Group 2 loads filters
15. Groups 1 and 2 prepare for final visits
16. Group 1 final visit to House 998
17. Group 2 final visit to House 999
18. Temporal site monitoring completed
19. Sampling equipment/data organization
a. Unload used filters
b. Data organization/backup/transfer
c. Equipment cleanup and packing
d. Field supervisors meet to discuss results
20. Final filters weighed
21. Databases merged
22. Particulate concentration results evaluated
23. Data processing evaluated
10:00 a.m.
11:00 a.m.
11:00 a.m.
2:00 p.m.
2:00 p.m.
3:00 p.m.
5:45 p.m.
5:45 p.m.
5:45 p.m.
7:30 p.m.
Monday August 27, 1990
6-6
-------
The dress rehearsal began the first evening when the field supervisors visited the two
dress rehearsal participant homes to deploy air exchange emitters and determine monitor
placement locations. On the second day, personnel from RTI, Harvard, and Acurex were
divided into two groups (essentially the same groups that were to be deployed in Riverside
during the field monitoring). Each group prepared monitoring equipment and made the initial
visit to one participant home on the evening of the second day. During this visit monitoring
equipment was deployed and sample collection was begun. The group returned to the homes
the next morning to collect samples and replace filters and made a final visit that evening to
complete sample collection. Temporal site samplers were operated during the two 12-hour
monitoring periods. Weighing operations were conducted in the weigh trailer, which was set
up on the RTI campus.
Evaluations of the dress rehearsal were made by the organizational supervisors for
RTI, Harvard, and Acurex. The results were used to fine-tune procedures and equipment
before proceeding with the field monitoring effort in Riverside.
Field Monitoring
Sample Collection Schedule-
Field monitoring was scheduled to begin in Riverside, California during the third week
of September, 1990. The study design called for 175 persons living in Riverside to be moni-
tored for two consecutive periods of approximately 12 hours each. The number and type of
samples scheduled for collection is presented in Table 6-2. During each monitoring period
one personal PM10 sample was collected. Indoor PM^ and PM2 5 samples were collected
in the main living area at the home during each period. Outdoor PM^q and PMg 5 samples
were also collected each period, at a specified distance from the home's main entrance when
possible. The citrate coated backing filters from the personal and indoor PM^q sample
impactors were retained for later nicotine analysis. A perfluorotracer was released into the
home at a constant rate using permeation tubes in heater blocks distributed in several rooms
of the home. Integrated tracer gas samples were collected in three rooms using capillary tube
samplers during each monitoring period. Polycydic aromatic hydnocarton (PAH)/phthalate
samples were scheduled for collection indoors at a subset of 120 homes and outdoors at a
subset of 60 of the homes with indoor samples.
6-7
-------
TABLE 6-2. TOTAL NUMBER OF SAMPLES SCHEDULED FOR
COLLECTION IN PILOT PTEAM STUDY
Sample Type Location Samples Duplicates Blanks Total
PEM (10 urn)
Participant
350
-
18
368
Technician
20
20
-
40
Temporal Site
88
4
-
92
SIM/SAM (2.5 ^m)
Indoor
350
18
18
386
Outdoor
350
18
-
368
Temporal Site
88
4
-
92
SIM/SAM (10 |im)
Indoor
350
18
18
386
Outdoor
350
18
-
368
Temporal Site
88
4
-
92
CATs (Air Exchange)
Indoor
1050
50
50
1150
Nicotine
Participant
350
-
18
368
Indoor
350
18
18
386
PAH/Phthalate
Indoor
240
12
12
264
Outdoor
120
6
6
132
Dichots (2)
Temporal Site
352a
-
-
352
Wedding PM10 (2)
Temporal Site
176
-
-
176
Cascade Impactor (1)
Temporal Site
30
15
4
49
a Includes both 176 coarse and 176 fine fractions.
6-8
-------
TABLE 6-3. PTEAM FIELD MONITORING SCHEDULE
Month Date Day Sampling Sampling Monitoring Begun on Each
Group 1 Group 2 Date for the Listed Participant
Activity Activity IDs
SEPT
19
20
21
22
23
24
25
26
27
28
29
30
WED
THU
FRI
SAT
SUN
MON
TUE
WED
THU
FRI
SAT
SUN
Travel
Setup
Setup
Begin
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
Sample
001 *,002"
003,004*,005**,006**
007,008,009*,010**
011 *,012*
013,014**,015"
016**,017,018*,019*
020*,021 *,022,023*
024**,025**,026**,027
028,029*,030**,031*
OCT
1
MON
Sample
032,033*,034,035*
2
TUE
Sample
036,038,039
3
WED
Sample
040,041 *,042,043**
4
THU
Sample
044*,045**,046,047*
5
FRI
Sample
Travel
048**,049*,050~,051
6
SAT
Sample/
Overlap
Sample/
Overlap
052*,053**,054,055
7
SUN
Travel
Sample
057*,058**,059**
8
MON
Sample
060**,061**
9
TUE
Sample
062*,064,065**,066
10
WED
Sample
067*,068,069*,070
11
THU
Sample
071,073,074**,075
12
FRI
Sample
076*,077**,079
13
SAT
Sample
080,082,083**
(continued)
6-9
-------
TABLE 6-3. (continued)
Sampling Sampling Monitoring Begun on Each
Group 1 Group 2 Date for the Listed Participant
Month Date Day Activity Activity IDs
14
SUN
Sample
084,085*,086",087"
15
MON
Sample
088*,089,090,091*
16
TUE
Sample
092",056",093*,094
17
WED
Sample
095*,096",097",098
18
THU
Sample
099", 100", 101*,
102"
19
FRI
Sample
103,104**,105,106
20
SAT
Sample
107",108,109*,078*
21
SUN
Sample
110*,111,112",113**
22
MON
Sample
037,114,115*,116"
23
TUE
Sample
117**118-119"
120"
24
WED
Sample
121*,122", 123",
124"
25
THU
Travel
Sample
125,126",127*,128"
26
FR!
Sample/
Overlap
Sample
Overlap
129", 130",131 ",132
27
SAT
Sample
Travel
133,134*,135,136
28
SUN
Sample
137*,138*,139*,140"
29
MON
Sample
141,142*,143**,144"
30
TUE
Sample
145",146M48
31
WED
Sample
149*,151 *,152*
NOV
1
THU
Sample
153,154*, 155,156*
2
FRI
Sample
157**, 158*. 159*, 160**
3
SAT
Sample
161 **,162*, 163**,
(continued)
6-10
-------
TABLE 6-3. (continued)
Sampling
Sampling
Monitoring Begun on Each
Group 1
Group 2
Date for the Listed Participant
Month
Date
Day
Activity
Activity
IDs
4
SUN
Sample
166,167**, 168*
5
MON
Sample
169*,170*,171**,172*
6
TUE
Sample
173*M74**,175*,176
7
WED
Sample
177**,178**, 179*,
180**
8
THU
Sample
181 *,182*,183,184
9
FRI
End
Sample
10
SAT
Pack/
Weigh
NOV
11
SUN
Pack/
Weigh
12
MON
Pack/
Ship
13
TUE
Pack/
Ship
14
WED
Travel
•Denotes indoor PAH/phthalate samples scheduled.
"Denotes indoor and outdoor PAH/phthalates scheduled.
6-11
-------
A total of 178 participants actually was included in the particle monitoring in Riverside.
A subset of 125 of these participants was scheduled for PAH/phthalate monitoring. Table 6-3
lists the date monitoring was begun for each participant and shows whether indoor and
outdoor PAH/phthalate samples were scheduled at the home.
Four visits were scheduled at each home over a 3-day period to begin and complete
all sample collection activities as described in Table 6-4. The first visit was used to place air
exchange tracer gas emitters in the home, measure the home's volume, and familiarize the
participant with the monitoring procedures. Particle and PAH/phthalate sample collection was
begun during the second visit, typically scheduled the next day at 7:00 or 8:30 pm. Indoor
and outdoor monitors were set up and started during this visit and the participant was given
the personal monitoring system to wear. Tracer gas samplers (CATS) were deployed and the
participant was informed that questions about activities on the monitoring day would be asked
at the next two visits. The third visit was made the next day between 4:00 and 8:30 am
(depending on the participant's work or school schedule). At this visit the filters deployed the
previous evening were retrieved, new filters were deployed, and an activity recall
questionnaire was administered. All sample collection devices and filters were removed from
the home during the fourth visit, typically scheduled for 5:00 or 6:00 pm, and another activity
recall questionnaire was administered. Finally, the participant completed an activity
modification questionnaire and was paid the $100 incentive by the field staff.
Filters were changed in the temporal site monitoring equipment twice a day, typically at
7:00 am and 6:00 pm to coincide with sample collection times at the participant homes.
Weighing operations were conducted in the weigh trailer from 7:00 or 8:00 am until at least
6:00 pm, and later when necessary due to weighing backlogs.
Personal and Residential Sample Collection--
Personal Particle Monitoring (PEM^--
A Personal Exposure Monitor (PEM) was used for collecting respirable particles to
which a participant was exposed. The PEM is an electronically flow-controlled, battery
operated Casella pump used to sample air through a portable impactor while being worn by a
participant. The impactor contained a 37-mm diameter Teflon filter having a 2-fim pore size.
The impactor collected particles having an aerodynamic diameter up to 10 jim. A constant
flow rate of 4 L/min was used to sample over periods of approximately 12 hours.
6-12
-------
TABLE 6-4. TYPICAL APPOINTMENTS AND ACTIVITIES AT
ONE PARTICIPANT'S HOME
INITIAL VISIT
- Harvard supervisor conducts this visit
- Air exchange emitters deployed
- House measured and house plan drawn
- Sample collection locations determined and recorded
- Appointment schedule confirmed with participant
SET-UP VISIT
- Three-person sampling team conducts this visit
- Set up and start indoor and outdoor particulate monitors
- Set up and start personal particulate monitor
- Deploy CATS for air exchange measurements
- Set up and start indoor and outdoor PAH/phthalate
monitors
(NOTE: PAH/phthalate samples not collected in every
home)
- Discuss activity questionnaire with participant
- Confirm remaining appointment times
CHANGE-OUT VISIT
- Three-person sampling team conducts this visit
- Remove used particulate filters, replace with new filters
- Remove used CATS, replace with new CATS
- Remove used PAH/phthalate filters, replace with new
filters
- Administer activity questionnaire
- Confirm final appointment time
FINAL VISIT
- Three-person sampling team conducts this visit
- Remove used particulate filters
- Remove used CATS
- Remove used PAH/phthalate filters
- Administer activity questionnaire
- Administer activity modification questionnaire
- Pay participant $100 incentive, get signed receipt
- Remove equipment from home
THURSDAY 6:00 PM
FRIDAY 8:30 PM
SATURDAY 6:30 AM
SATURDAY 6:00 PM
6-13
-------
The PEM impactor was worn on the participant's lapel (in the participant's breathing
zone) while being shielded from clothing fibers and dander which could otherwise be
collected. The pump was worn in a fanny pack that was designed to slide around the waist as
desired for sitting, walking, etc. The participant was encouraged to continue normal activities
while wearing the monitor. However, during certain activities such as sleeping or showering,
the PEM was kept as near to the person as practical.
Only technical staff involved in observing field activities wore duplicate PEMs. One
PEM impactor was worn approximately 10 cm under the normally situated PEM impactor for
duplicate sample collection.
Stationary Particle Monitoring (SIM and SAM)--
For collecting respirable particles from indoor and outdoor environments at the
participants' homes, the Stationary Indoor Monitor (SIM) and the Stationary Ambient Monitor
(SAM) were used. The SIM and SAM monitors were identical, each consisting of an
electronically flow-controlled Medo pump used to draw air through stationary impactors.
Impactors with particle size cuts of 10 ^im and 2.5 jim were deployed at all indoor and outdoor
locations. Each impactor used a 37-mm Teflon filter having a 2-pin pore size. Each impactor
collected particles at a constant air flow rate of 4 Umin over a period of approximately 12
hours.
The SIM sampling heads (2.5 and 10 p.m) were oriented sideways to avoid particle
deposition due to gravity. These sampling heads were attached to a stand approximately
1 meter above the floor, which is the breathing level while seated. Further details concerning
siting of the SIMs is discussed in a subsequent section.
The SAM sampling heads (2.5 and 10 >im) were also oriented sideways and, due to its
placement outdoors, also had an inverted bowl-type rain shield approximately 30 cm in
diameter. The sampling heads were at least 1.5 m above ground level and were generally
located on the side of the home facing the street Further details concerning siting of the
SAMs are discussed in a subsequent section.
Nicotine Monitorina-
Nicotine samples were collected to evaluate the contribution of particles due to
environmental tobacco smoke. All PEM, SIM and SAM sampling heads contained a glass
6-14
-------
fiber backing filter (40 mm) to support the Teflon filter. These backing filters were
impregnated with citric add to capture nicotine after particles were trapped out of the air
stream being sampled. As such, the nicotine was collected using the same equipment and
procedures as for the particles (PEM, SIM, SAM). However, only the filters for the PEM and
10-^m SIM sampling heads were saved for analysis.
Polvcvclic Aromatic Hydrocarbon (PAHVPhthalate Monitoring"
Polycyclic aromatic hydrocarbons (PAHs) and phthalates were collected indoors, or
both indoors and outdoors at designated participants' homes. The monitor for this task
consisted of a sampling pump and a cartridge containing a sorbent bed of XAD-2 (- 5 g)
preceded by a quartz fiber filter. The XAD bed was 22 mm in diameter and was inside a
glass sampling head connected to a sampling box containing four Medo pumps (operated in
opposing phases for noise dampening). The pump sampled air at a constant flow rate of
approximately 18 L/min over periods of approximately 12 hours.
The sampling head was placed in a box for protection in both indoor and outdoor
locations. The sampling head was located approximately 1 to 1.5 m above ground or floor
level. Further details concerning siting are discussed in a subsequent section.
Air Exchange Monitoring-
Air exchange rates were measured at participant homes to aid particle source
modeling. The measurement involved the constant release of perfiuoromethylcyclohexane, a
"perfluorotracer" (PFT), while simultaneously sampling for the tracer with a sorbent. Three
sources of PFT were put into each home 24 hours prior to the first sample collection visit.
Each PFT source was heated to 40°C and was located 0.5 to 1.5 m above the floor. At the
initial visit when PFT sources were deployed, the interior dimensions of each home were
measured and recorded. Interior walls were essentially assumed nonexistent for this
approach. At each of the subsequent two visits capillary adsorption tubes (CATS) were
placed in the home at three sites and were situated 1 to 1.5 m above the floor. Two CATS
were placed in the main living area, one in the bedroom, and one near the center of the
home. The CATS passively collected the PFT onto their specially activated charcoal-type
substrate for periods of approximately 12 hours, in conjunction with the other sampling
activities.
6-15
-------
Sampler Placement Strateav-
All sampler locations were determined during the initial visit to each home,
approximately 24 hours before sampling was to begin, while the PFT sources were placed in
the home. Locations were selected after carefully determining the layout of the home.
Locations were selected after the living habits of the participant, the type of dwelling (home,
apartment, etc.), and the outside layout of the yard or grounds.
Once the layout of the home was determined, the main living area for the participant
(the room in which the participant spends most time while awake) was ascertained. Particles,
nicotine, and PAH/phthalates were collected in the main living area and two CATS and one
PFT source were deployed. The participant's bedroom was the second location for a CATS
sampler and the PFT source was placed in the general sleeping area of the home. The third
PFT source was located in the geometric center of the dwelling.
Additional criteria for sample placement indoors included keeping the source/samplers
at least 2 meters away from exterior doors, windows, and ventilation registers. The PFT
sources and CATS were kept at least 2 meters apart. The PAH/phthalate sampling pumps
were placed no closer than 1.5 meters to the sampling head when possible. These ideal
criteria were followed as closely as practical after considering layout and size.
For outside sampling there were two criteria. For the SAM, the samples were ideally
collected 1.5 m from the wall facing the street outside of the main living area of the home.
Additionally, the SAM was not located within 1 m of trees and bushes or within 5 m of any
type of air vent For second floor apartments a "yardarm" was deployed from a window or
balcony to support the impactor heads. If a yardarm was not possible it was acceptable to
collect samples at ground level for second floor apartments. Non-ideal situations required
some reasonable compromises. All outdoor monitors were protected by a rain shield that
allowed free air flow to the filters.
For the PAH/phthalate sampling outdoors, the criteria were to place the sampler on the
side of the house opposite to the roadway, and to place it a distance of two building-heights
away from the home, or if not possible, no less than 2.5 m from the home. For apartments,
balconies were acceptable but samplers were not to be placed within 5 m of a parking area.
6-16
-------
Temporal Monitoring-
The temporal sampling site was selected based on several criteria. It must provide
representative samples for the area, meet accepted siting requirements for reference method
PM.j q monitoring, have power and security, and have convenient access. A good, though not
perfect, choice used for this site was the California School for the Deaf. It was representative
in that it was as close to the geographic center of Riverside as possible. Other aspects to
note was that it bordered a segment (neighborhood) used for sampling homes. It was only
about 1/2 miles from the freeway, but this is similar to many residential neighborhoods in
Riverside. The temporal site was at a slightly higher elevation than much of the city.
The samplers were placed on a platform approximately 2 m above ground so that the
sampling height was 3.4 m. The platform was located approximately 1.5 building-heights
away from a school building and was in a grassy area at least 15 m from any trees. A 50-
amp power line was run to the platform, a security fence was erected, and access to the
platform was available 24 hours per day.
PEM and SAM Monitorina--
All particulate monitoring using the PEM and SAM sampling heads and pumps was
carried out similarly to the sampling conducted in participants' homes. The difference is that
the samplers were located on the temporal site platform. The periods of collection were
approximately 12 hours and roughly coincided with the sampling periods used for the homes.
Dichotomous Sample Collection-
Two dichotomous samplers were operated at the temporal site. The Sierra
dichotomous sampler consisted of a tripod holding a sampling head connected to an air
sampling pump. The air sampling head fractionated the particulate so that particles up to
2.5 jim in aerodynamic diameter were collected by one filter (37 mm) and particles up to
10 nm in aerodynamic diameter were collected by the other filter (37 mm). The sampler
contained a bug screen and rain deflector. Filters were replaced at the end of each
monitoring period of approximately 12 hours.
6-17
-------
High Volume PM1Q Sample Collection-
Two Wedding reference method high-volume PM^ q monitors were operated at the
temporal site. The Wedding sampler consisted of a large air sampling pump and a sampling
head that trapped particles up to 10 jim in aerodynamic diameter onto an 8" x 10" glass fiber
filter with a 2-^m pore size. The air sampling system used a volumetric flow control system
and maintained a constant flow rate of 1.13 m3/min. The sampling head used a cyclonic
fractionation technique for collection of the particles of interest. As with the other samplers,
the period of collection was approximately 12 hours.
Cascade Impactor Sample Collection-
The determination of the size profiles of particulates in the Riverside area was carried
out by using a cascade impactor. The Anderson cascade impactor consists of an air sampling
pump plus eight stages for particle collection by size. Each stage contains a number of jets of
a specific diameter that cause particles of a certain size or larger to impact on a filter medium.
Air passing through one stage enters another stage with smaller jets allowing collection of
successively smaller particles. This is typically carried out for particle size cuts from 10 jim to
0.5 jim going from the first to last stage. The total flow rate of 1.0 ft ^/min was passed into all
stages throughout the collection period. The impaction media used were 3" circular glass fiber
filters with a 2-p.m pore size. The period of collection varied from 3.5 to 10 days in order to
collect enough matter for accurate weighing. This sampling was carried out four times during
the sampling trip. Table 6-5 show® the periods of collection and the particle size cuts
collected during each period.
Meteorological Data Collection-
Attempts were made to collect meteorological data at the temporal site. The measure-
ments taken included wind speed, wind direction, temperature and relative humidity through
the use of an Odessa meteorological system. The collection of meteorological data at the
temporal site was incomplete due to data logging problems. Supplemental meteorological
data were obtained from three airports in and near Riverside.
6-18
-------
TABLE 6-5. PERIODS OF COLLECTION AND PARTICLE SIZE CUTS BY STAGE FOR THE CASCADE IMPACTOR
Start End
Collection Collection Stage: 0 1 2 3 4 5 6 7
Aerodynamic Particle Size Cut at Each Filter Stage
Run No. Date Time Date Time (microns)
1
9/28
20:15
10/3
07:32
10.7
6.9
4.7
3.2
2.1
1.1
0.70
0.50
2
10/6
19:42
10/11
18:27
10.7
6.9
4.7
3.2
2.1
1.1
0.70
0.50
3
10/24
19:30
10/28
07:05
9.8
6.3
4.3
2.9
1.9
1.0
0.60
0.50
4
10/30
18:34
11/9
18:06
9.8
6.3
4.3
2.9
1.9
1.0
0.60
0.50
-------
Weigh Trailer Operations-
The weigh trailer was a mobile laboratory situated at a fixed location. It was designed
for weighing filters in a stable environment (temperature and humidity) with minimal vibration
at the balance. The balances were placed on tables with legs extending through the trailer
floor to the ground. The trailer interior was maintained at 21 +3°C and relative humidity was
maintained at 50 +2 percent. A steam humidifier was used to help maintain the humidity level
as part of the HVAC system. The floor, shelves and other surfaces were cleaned with a damp
mop or rag to reduce dust. Anti-static measures were used to prevent erroneous weighings.
All filters were subjected to at least 24-hour exposure to the trailer atmosphere prior to tare
and final weighings.
PEM, SAM. SIM and Dichotomous Sampler Filters—
The PEM, SAM, SIM and dichotomous samplers all used 37-mm Teflon filters, and
their weighing procedures were identical. The balance used was a Cahn Model 30
microbalance with an accuracy of 1.0 jig. It was connected to a microcomputer with weighing
software developed for this program. Once tared, all filters were inspected for holes or other
imperfections prior to use and were kept in a barcode-labelled Petri dish. Filters were
equilibrated in the weigh trailer for 24 hours prior to being weighed (e.g. taring, reweighing or
at any other weighing).
A set of 10 filters was weighed using the following steps. First the balance was zeroed
and the calibration checked using NIST traceable masses. Then each filter was weighed and
the weight recorded once the computer recognized a stable reading (1-2 min). A manual
reading using a 90-sec period was used for a subset of the filters due to a computer program
problem. After each set of 10 filters was weighed, the zero was checked to within +0.004 mg
and a 200 mg weight to within +0.002 mg. If either zero or the 200 mg weighing failed their
test then the zero/calibration was repeated and the previous set of filters was reweighed.
Otherwise weights for the 10 filters were accepted.
High-Volume Filters and Cascade lmpactor Filters-
The high-volume filters and the cascade impactor fitters were larger and were weighed
in a somewhat different manner, using a Mettler H15 balance with an accuracy of+0.1 mg
and an upper mass limit of 200 g. A set of up to ten filters was weighed using the following
6-20
-------
steps. First the balance was zeroed and the calibration checked with an NI ST traceable 5 g
weight. Then each filter was weighed, allowing settling of the balance for a set period, and
the weight was manually recorded. After each set of filters was weighed, the zero was
checked. If the zero was outside of a+0.3 mg limit, then the balance was re-zeroed, the
calibration weight was reweighed and the previous set of filters was reweighed.
Quality Control Procedures-
To assure high quality data, measures were carried out to check weight data. The
primary quality control procedure was mentioned in the previous two sections, which was to
bracket any set of 10 (or fewer) filters with zeroes and mass readings which were within
specifications. The other procedure used was the reweighing of filters by a different operator.
One filter was selected from each set of 10 to be reweighed. This applied to all filter types.
AH procedures used were identical to original weighing and were carried out for tare and final
weighings. For the 37-mm Teflon filters, the reweighing had to agree to within +4 jig of the
original weight to be accepted. If any filter fell outside of this limit, the set of 10 filters from
which it was taken was reweighed by the primary weighing person. No data were accepted
until this reweighing procedure was carried out.
Changes in Wortolan Protocols
The sampling protocols used for this effort are found in the Work Plan (RTI, 1990a).
Additional information concerning procedures for the Wedding and dichot samplers is provided
in more detail elsewhere (U.S. EPA, 1990). Based on laboratory testing and trials during the
dress rehearsal, some protocols were modified before field monitoring operations began.
These changes from protocols described in the Work Plan are noted below.
• The nomenclature of MEM (microenvironmental monitor), used during the prepilot
study has been changed to SIM for indoor sampling and SAM for outdoor
sampling.
• Flowrates used for the SIM and SAM were 4 L7min instead of 10 L/min. The PEM
flowrate of 4 L/min was unchanged. The same sampling impactor used on the
PEMs was used with the SIM and SAMs.
• No passive nicotine collection was carried out. All active nicotine collection utilized
the citrate treated AP-40 glass fiber filters as part of the sampling head. Nicotine
samples were collected along with only PEM and SIM 10 urn samples, although
6-21
-------
the treated AP-40 filters were placed in all sampling heads to ensure identical
conditions for particulate sampling.
• Some cleaning steps for the particulate sampling impactors used for the PEM, SIM
and SAM sampling were modified. The cleaning procedure for the impactor plates
was strictly separate from the other parts, even though the steps were similar. All
parts were washed using Alconox and water by scrubbing with a brush; for
impactor plates an initial removal of grease using Kimwipes was carried out. All
parts except exterior body parts were sonicated once for 15 min in Alconox and
water. Subsequent rinsing was carried out with deionized/charcoal filtered water.
An ethanol dip was carried out to expedite drying, except that only the non-
greased surfaces of the impactor plates were dipped in the ethanol. Drying was
expedited by convection and took 8-12 hours.
The protocol for weighing particulate filters was modified as follows:
• The nominal relative humidity inside the weigh trailer was set at 50% instead of
40%.
• The threshold for quality control check filter weighings (1 of 10 filters) was ±4 ng
instead of +10 fig.
The protocol for air exchange rate measurement was modified as follows:
• All sampling with CATS was carried out over 12-hour periods corresponding to the
particulate monitoring periods. Equations concerning LOD and maximum PFT
concentration should correspond to this change.
• House dimension measurements were made with a measuring wheel and not the
electronic rangefinder.
• A pair of CATS was placed in the main living area only. The other two areas only
had one CATS each.
Modifications of the PAH/phthalate protocol are as follows:
• The nominal flowrate expected through the sampling train was 18 limin instead of
20 L/min.
• The sonication method was chosen for extraction of the XAD.
The protocols for the Hi-Vol and Dichot sampling operations had the following
modifications:
• The weighing procedures used were those described in the weighing protocol and
not those in the protocol for fixed site operations.
6-22
-------
Data Acquisition and Sample Tracking
Sample and Participant Identification-
Unique series of identification numbers were assigned to each type of sample
collection device (particle filters, nicotine filters, PFT collection tubes, etc.) and for certain
pieces of data collection equipment (pumps, sampling heads, etc.). The identification system
ensured that the type of device could be determined from the first few digits of the ID number.
Bar code labels with the appropriate identification numbers were printed for attachment to the
sample collection devices and sampling equipment For the particle filters, which required
weighing before and after sampling, the bar code identifier (attached to the plastic containers)
was scanned as input to the weighing software.
An additional unique series of identifiers were assigned at the participant level. Each
questionnaire or document assigned to the participant was labeled with the bar code identifier.
The ID of the participant filling out the questionnaires served as the master ID for each
household. Sets of pre-labeled questionnaires were packaged together for ease of use by the
field teams.
Preparations for Monitoring--
Prior to monitoring, the particle filters were weighed. The weight data were collected
via a direct electronic link between a personal computer and a Cahn balance. A bar code
reader attached to the system allowed the technician to scan in filter identification numbers.
After weighing, the filters were placed into sampling bases by the field teams. Since
IDs could not be attached directly to filters, a program operating on a lap-top computer
allowed the team to scan IDs of the filters and the sampling device they were placed in, to
permit tracking of the filters.
In-Home Operations--
When field technicians entered a home to set up monitoring, they took with them the
lap-top microcomputer equipped with a bar-code reading wand. They will also took the
package of pre-labeled questionnaires, the pre-assembled sampling devices (sampling bases)
and PFT collection tubes. First, the operator ID and the participant ID for the home were
entered. On the first visit, the technician was prompted to enter the participant's name and
6-23
-------
address. On subsequent visits, the name and address corresponding to the participant ID
were displayed as a check.
Next, the technician entered information ttiat applied to all the monitors in the home,
such as calibration equipment IDs and temperature data. Then, the technician entered data
(pump IDs, pump location) for each personal and microenvironmental monitor. When a
monitoring period started, the sampling device was assembled with a particular pump. At this
point, by scanning the device ID, the filter ID was linked with a pump, a location and a
participant. When the twelve hours ended, the device was scanned again and final
information (calibration, elapsed timer, etc.) was entered. Set-up and removal times for the
PFT collection tubes was also entered into the computer. Dates, times and operator initials
were collected automatically for installation and removal of each filter. The lap-top program
allowed input of more than one collection device of a given type, if required for quality control
purposes. The program also checked the ID numbers to insure that the proper type of device
was scanned. This procedure established the linkage between various sample collection
devices and the participants. Once this link was established, either the participant ID or the
sample collection device ID alone was sufficient identification for data records throughout the
remainder of the study. All of the data described previously were also written on forms to
provide a back-up to the data stored on diskette.
Post-Monitoring Operations~
When the collection devices were disassembled, the lap-top program was used again
to ensure that filters were replaced in the correct storage containers. Technicians also had
the opportunity to check the data, add missing information, and make corrections or
comments. When processing was complete, the data on the finished filters were written to a
floppy diskette. Filter weighing after monitoring was done as in the pre-monitoring step.
Field Operations Loaistics-
Collecting samples at 178 homes in Riverside was a very large undertaking. Five
major activities were conducted simultaneously and were interdependent. Participant
recruitment was ongoing throughout the course of the study with the goal of beginning to
monitor four new participants each day. Particle, nicotine, and air exchange monitoring were
conducted at each participant home. The CARB component of the study required
6-24
-------
PAH/phthalate sample collection at a majority of the homes. A temporal site monitoring
location with access to power was obtained and a platform was erected to support the
monitoring equipment. Ambient air monitoring with eight sample collection devices was
conducted at the temporal site during every 12-hour period throughout the study. And a
mobile weigh laboratory was shipped to the apartment complex where the workroom was
located and was operated continuously to weigh all of the participant filters collected during
the field study.
At any given time 15 to 18 people were involved in the field operations. All field
activities occurred seven days a week over a nine-week period without stopping. Five locally
hired staff members were responsible for recruiting participation in the study population and
scheduling monitoring visits. Two eleven-person groups composed of RTI, Harvard, and
Acurex personnel completed all of the sample collection, activity questionnaire administration,
and filter weighing tasks. (The first 11-person group worked the first and last three weeks in
the field while the second 11-person group performed operations during the middle three
weeks).
The site administrator from RTI was responsible for coordinating all field activities and
ensuring that all procedures were being correctly followed. The site administrator was also
responsible for the CARB component of the sample collection. A supervisor from Harvard
was responsible for all PEM, SIM, SAM, nicotine, and air exchange sample preparation and
collection. This person also made the initial visit to each home to set up air exchange devices
and to discuss the study with the participant. A supervisor from Acurex was responsible for all
temporal site monitoring activities and the operation and maintenance of the weigh laboratory.
Two three-person teams prepared sample collection materials and conducted all sample
collection and questionnaire administration at the participant homes. Another staff member
was responsible for making all tare and final weighings of the filters used for particulate
collection. The final staff member assisted during the initial home visit and maintained air
exchange measurement equipment.
Most staff members performed very well working long hours for 19 to 23 straight days.
A typical daily schedule, shown in Table 6-6, would begin at 5:00 to 6:00 am preparing
sampling materials (Table 6-5). Each team visited two homes in the morning. After returning,
the filters from the previous 12 hours were removed from impactors, the impactor systems
were cleaned, and then were reassembled with new filters for the next two sampling periods.
6-25
-------
TABLE 6-6. TYPICAL DAILY SCHEDULE FOR THE FIELD MONITORING STAFF
MORNING
- TEAM A (first three-person sampling team)
6:30 a.m.
- Change-out visit to participant home
7:30 a.m.
- Change-out visit to participant home
- TEAM B (second three-person sampling team)
6:00 a.m.
- Change-out visit to participant home
8:00 am.
- Change-out visit to participant home
7:00 a.m.
- Temporal site monitors serviced
7:00 a.m.
MIDDLE OF THE DAY
• Filters weighed in weigh trailer
- Sampling teams unload used filters, clean impactors, and
load new filters in impactors
- Equipment repairs performed as necessary
- All data processing and paperwork completed
LATE AFTERNOON AND EVENING
- Harvard supervisor visits 4 new homes
- Air exchange appointment
5:00 p.m.
- Air exchange appointment
6:00 p.m.
- Air exchange appointment
7:00 p.m.
- Air exchange appointment
8:00 p.m.
-TEAM A
- Final visit to participant home
5:00 p.m.
- Final visit to participant home
6:00 p.m.
- Set-up visit to participant home
7:00 p.m.
- Set-up visit to participant home
8:30 p.m.
- TEAM B
- Final visit to participant home
5:00 p.m.
- Final visit to participant home
6:00 p.m.
- Set-up visit to participant home
7:00 p.m.
- Set-up visit to participant home
8:30 p.m.
- Temporal site monitors serviced
6:00 p.m.
NIGHT 9:30 -
- Sampling teams unload used filters and clean impactors 11:30 p.m.
- Equipment repairs performed as necessary
6-26
-------
Final visits were made in two homes in the late afternoon by each team, and then monitoring
at two new homes was begun in the evening. Filter removal and impactor handling
procedures were repeated after the evening appointments. The teams typically worked from
eariy morning to 10:00 or 11:00 at night, with a couple of hours off in the middle of the day.
Sixteen home monitoring visits were scheduled and had to be coordinated each day, along
with two temporal site visits and many other participant recruiting visits.
Several apartments were rented as living quarters for the field staff. One apartment
was used solely as a workroom and was the base for all field operations. Workroom spaces
were divided into office, equipment preparation and repair, equipment cleaning, and impactor
assembly areas. The impactor assembly required a very clean space to avoid contamination
of sample filters during insertion or removal from the impactors. One room was designated for
filter and impactor handling. There was no way to turn the space into a true clean-room, but
efforts were made to minimize dust. A portable electrostatic air cleaner was used to filter
room air, the carpet was covered, and dust producing materials were kept out of the room.
Arrangements were made with the apartment management to park the weigh trailer in a
walled compound within the complex and to obtain power from an existing electrical service.
Having the weigh facility, workroom, and living quarters at one location greatly simplified all
facets of field monitoring logistics.
Post Field-Studv Activities
Filter Inventory/Archival-
All PEM, SIM, SAM, dichot, and Wedding filters were hand carried from Riverside to
RTI to minimize the loss of particle from the filters that could occur during commercial
shipping. Each sample was logged-in upon receipt at RTI and its condition noted. All PEM,
SIM, SAM, and dichot filters were sorted by numerical filter number order. Filters were kept in
their original plastic slides, bundled in batches of 10, and stored in boxes. The boxes were
then placed in a locked refrigerator set at 4.5°C. A hard copy inventory of all sample filters
returned from Riverside was set up and maintained in the laboratory to record the status of
each filter.
6-27
-------
Database Additions-
Much of the sample collection data associated with particle, nicotine, air exchange, and
PAH/phthalate samples entered in the field directly onto computer magnetic storage media.
These magnetic records were then transferred to a central microcomputer at RTI where the
total PTEAM database was accumulated. Several additions were made to the database after
returning from the field. Dichot and Wedding field blank filter numbers and weight differences
were added separately to the database. Meteorological data was obtained from the NOAA
National Climatic Data Center in Asheville, NC for three airports around the study area.
Hourly wind speed and direction were obtained for Riverside Municipal Airport (6:00 am to
7:00 pm only). Hourly ambient temperature, dew point temperature, wind speed, and wind
direction data were obtained for Ontario International Airport and March Air Force Base on a
24-hour basis. A subjective evaluation of the cleanliness of each participant home was also
entered into the database. This was called a "dirt factor", which ranged from 0 to 3.0, where 0
was extremely clean and 3.0 was extremely dirty.
Database Construction-
Floppy diskettes containing all the field information, plus the weighing data, were
delivered to RTI and were copied into the host PC system. Additional data from the central
site monitoring operations were obtained in machine-readable form from Acurex.
Meteorological data from three local airports were obtained and keyed. For ease and speed
of processing, the database was maintained as several smaller files in logical groups: one or
more files of gravimetric data, a file of elemental analysis data, a file of nicotine data, etc.
The study questionnaires were edited by RTI's Data Preparation department according
to procedures developed by project staff, and passed to RTI's Data Entry group for keying.
Data entry programs were prepared for each form, using Easy Entry software on the VAX
cluster. Screener and questionnaire data were keyed into the VAX, and passed to the
statistical staff. The data were identified by the sampling ID as well as the monitoring ID.
All coded or numeric data keyed by RTI's Data Entry staff were 100% verified. That is,
after data were keyed once, a second person rekeyed the entire form. All differences
between the two keyings were automatically detected and resolved before the data were
accepted by the entry system.
6-28
-------
Following entry of the data, all data files on the PC were converted to dBASE III
format. Documentation of the file formats, including variable positions, lengths, types and
meanings was developed. All gravimetric and air exchange data were printed and reviewed
against hard copy by the chemistry staff. Following corrections, gravimetric values were
computed both with and without correction to standard temperature and pressure. All data
were then delivered in machine-readable form to the QA staff for review and approval prior to
statistical analysis.
Database Verification-
All PEM, SIM, SAM, nicotine, air exchange, and PAH/phthalate data in the database
were 100% verified by the three RTI field coordinators upon the completion of the field study.
After the original database was compiled from field computer records, the records were printed
for examination. The field coordinators examined every entry and compared the computer
record to the written records made by each sampling team as they collected the samples in
the field. All discrepancies were investigated and corrected when possible. Records with
unusable or suspect data were flagged and returned to the database with the accompanying
flag. The database coordinator also scanned the database to discover specific problems,
including out-of-range pump flows or particulate catches, duplicate uses of filter codes,
missing data, etc. These problems were flagged and investigated by the field coordinators.
At this point the data records were sent to quality assurance personnel for further examina-
tion. All dichotomous, Wedding, cascade impactor, and meteorological data from the temporal
site were sent to RTI quality assurance personnel for examination. The filter weighing records
were first examined by field personnel and then forwarded to quality assurance personnel.
RESULTS AND DISCUSSION
Overall Evaluation
Overall, the field monitoring program was very successful. Monitoring was completed
on schedule and 178 participants were monitored instead of the scheduled 175. No major
technical problems were encountered during the monitoring; in fact, the equipment and
procedures performed well for extended periods of continuous use. Sampling completion
rates and minor technical problems will be discussed below. Field personnel worked very
6-29
-------
hard and the integration of people from RTI, Harvard, and Acurex into efficient sampling teams
was successful.
Particle monitoring proceeded very well. The personal monitor pumping systems
performed exceedingly well; very few pump failures occurred and consistent flow rates were
maintained throughout the study. Stationary monitoring of particles also had few problems
and almost all samples were collected as scheduled. Preparing 50 to 60 filter and impactor
assemblies consistently each day while maintaining low particle background levels was
difficult. Background levels below 20 pig/filter were achieved for most field blanks under less
than clean-room conditions. The median background level was 9 ng/filter.
Weighing operations went very well during this pilot study. Over 200 separate
weighings were made on most days. Vibration and static electricity problems were largely
eliminated by careful preparation of the facility and equilibration of the filters. Temperature
and humidity control was not perfect, but the number of quality assurance weighing failures
(differences between weighings of the same filter greater than +4 ^g) was less than 15%. All
of the filters passed the tolerance criteria before the weigh data were accepted.
Sample Completion Rates
Overall sample completion rates exceeded 95% for particle, nicotine, and air exchange
samples. Completion rates for PAH/phthalates samples were somewhat lower but still above
90%. A breakdown of sample collection data is presented in Table 6-7. The number of
collections that were attempted was higher than the number originally scheduled for two
reasons. First, more participants were actually monitored than called for on the original
schedules. Second, four monitoring days were added to allow completion of the study,
therefore temporal site monitoring was conducted for four more days than originally planned.
Field Monitoring Problem Areas
A total of 99 PEM, SIM, and SAM particle samples was not successfully collected.
Thirty-six samples were lost because filters were used twice to collect samples. Forty-six
samples were lost due to pump or power problems. Six filters were reported as missing and
seven filters had incorrectly reported weighings. One sample was collected with the wrong
cut-size impactor and the final filter had data entry problems that could not be resolved. The
6-30
-------
TABLE 6-7. SAMPLE COLLECTION STATUS FOR PTEAM 1990 PILOT STUDY
Scheduled (Work Plan)/Attempted/Collected
Sample type
Location
Samples
Duplicates
Blanks
Total
PEM (10 jim)
Participant
350/356/347
0/0/0
18/17/17
368/373/364
Technician
20/17/16
20/17/17
0/0/0
40/34/33
Temporal Site
88/96/88
4/3/3
0/1/1
92/100/92
SIM/SAM
Indoor
350/356/341
18/17/15
18/17/16
386/390/372
(2.5 nm)
Outdoor
350/350/334
18/17/17
0/0/0
368/367/351
Temporal Site
88/96/92
4/3/3
0/1/1
92/100/96
SIM/SAM
Indoor
350/355/333
18/17/17
18/17/16
386/389/366
(10 jim)
Outdoor
350/350/336
18/17/17
0/0/0
368/367/352
Temporal Site
88/96/91
4/4/4
0/1/1
92/101/96
PAH/Phthalates
Indoor
240/250/236
12/16/12
12/12/12
264/278/260®
Outdoor
120/130/115
6/8/7
6/6/6
132/144/128b
Air Exchange
Indoor
1050/1060/1046
50/356/350
40/94/87
1140/1510/1483
Nicotine
Participant
350/356/347
0/0/0
18/18/16
368/374/363
Indoor
350/355/336
18/17/16
18/17/16
386/389/366
Technician
20/17/16
20/17/17
0/0/0
40/34/33
Dichots
Fine
Temporal Site
88/96/95
88/96/95
0/21/19
176/213/209
Coarse
Temporal Site
88/96/95
88/96/94
0/21/18
176/213/207
SSI (PM10)
Temporal Site
88/96/92
88/96/88
0/44/43
176/236/223
Cascade0
Temporal Site
30/40/32
15/0/0
4/0/0
49/40/32
® Eighteen of these samples not collected over the entire 12-h period
Twenty of these samples not collected over the entire 12-h period.
c Five cascade runs attempted, four successful, 8 filter stages each.
-------
unsuccessful samples represent only 4.4% of all attempted PEM, SIM, and SAM sample
collections. Another 125 samples were reported with comments (observations at the collec-
tion, weighing, or log-in stages) and were flagged in the database. These comments were
later reviewed to determine if the reported observation would affect sample quality.
Pumps used to collect PAH/phthalate samples were susceptible to fluctuations in line
voltages encountered in Riverside. The pumps were operated for hundreds of hours at RTI
with no failures. However, voltage drops and power outages were very common in Riverside
throughout the study period. Outages were particularly common during periods of high winds,
but voltage drops occurred regularly. The PAH/phthalate pumps were designed with an
automatic shut-off system to protect the pumps in case flow was reduced below a set point.
Voltage drops and power outages often tripped the shut-off, turning the pumps off and ending
sample collection. If the voltage drop was very short then the sampling time was retained on
a LED display. If the drop lasted longer, then the LED display was lost and the sample
collection time was unknown. Thirty-eight of the 380 scheduled PAH/phthalate samples were
collected over a known duration that was less than the scheduled 12 hours. Another 29
samples were lost due to power failures that resulted in sample collection for an unknown
duration.
Several temporal site dichotomous and SSI samples were lost due to power and
equipment failures. Much of the meteorological data collected at the temporal site were
unusable due to equipment failure and difficulties retrieving the data from the unit's storage
media The temperature and humidity control systems in the weigh trailer worked well most of
the time although several problems were encountered. First, power failures in the Riverside
power grid damaged some of the trailer's control components. Second, large swings in
outdoor temperature and relative humidity from day to night and over several weeks taxed the
system's control capabilities to the fullest extent. None of the problems described above
should affect the use of data collected during this study in describing personal exposure to
particulates in Riverside.
6-32
-------
SECTION 7
QUALITY ASSURANCE
INTRODUCTION
Quality assurance activities for the PTEAM study were conducted in four phases with
results summarized in this section. First, a Quality Assurance Project Plan was prepared for
the PTEAM Pilot Study. Second, systems and performance audits were conducted during the
field study in Riverside. Third, audits of data quality were conducted at RTI as the PTEAM
database was constructed. The audits of data quality were concerned with documenting
details involving the collection, completeness, and acceptability of data generated during the
study. Finally, the original quality assurance objectives for the PTEAM study were
reexamined after statistical analyses of the data were completed. These objectives included
completeness of sample acquisition, precision, accuracy, and detection limits. Complete
discussions of the data quality parameters are included in Sections 6, 9, and 10 but will be
summarized in this Section as they relate to the quality assurance objectives.
QUALITY ASSURANCE PROJECT PLAN
Prior to beginning the study a Quality Assurance Project Plan (QAPP) was prepared by
RTI and approved by EPA (RTI, 1990b). Described in the QAPP were project organization
and responsibility guidelines, quality assurance objectives, sampling and calibration
procedures, sample custody, and data reduction and reporting guidelines. The QAPP was an
important part of the overall project quality assurance in the planning, implementation, and
reporting phases of the study. Results of the technical and data audits, and a summary of the
quality assurance objective results are described below. Based on these results, the
effectiveness of the QAPP is summarized in the last part of this Section.
PERFORMANCE AND TECHNICAL SYSTEMS AUDIT RESULTS
Two performance and technical systems audits were conducted, one at the beginning
and the other near the end of field monitoring activities. Study personnel were rotated three
times during the study. Each rotation included two sampling teams, a balance operator,
temporal site operator, and supervisory and support personnel. Auditing activities were
undertaken to evaluate data collection while the first and third groups of study personnel were
7-1
-------
on-site. The first audit was intended to ensure that samplers were operated properly with
appropriate data collection procedures. The second audit was to ensure that corrective
actions were incorporated and that data quality remained acceptable throughout the
monitoring activities.
Systems Performance
The performance of each type of particulate sampler used in the study was evaluated
by measuring the flow rate with an audit device. This flow rate was compared against the
reported flow rate and/or design set point. The results are shown in Table 7-1. The balances
used in filter weighing were checked using NIST-traceable Class S weights. The results of the
balance audit are shown in Table 7-2.
The technical systems audit evaluated all components of the program, including in-
home/personal monitoring, temporal site, weighing facility, and workroom operations. The
Pre-Pilot Study (RTI and Harvard, 1990) provided valuable experience and guidance on how
to improve the quality of data generated. One element intended to minimize poor quality data
was a software improvement to merge all the data files and calculate concentrations on a
real-time basis in the field. This and other monitoring improvements implemented as a result
of the Pre-Pilot Study experiences were also evaluated during the technical systems audit.
More specifically, the technical systems audit evaluated the following:
• Field operations at numerous homes during sampling,
• Field operations at the temporal ambient monitoring facility,
• Filter loading and sample handling at the staging area, and
• Filter-weighing procedures and conditions at the weighing room facility.
Observations at Homes
Auditors accompanied study personnel on visits to 15 homes and observed all phases
of sample and data collection. Several observations were made pertinent to data quality.
Location of outdoor ambient air monitors at the homes was challenging in many cases due to
small yards, limited distances to roads and parking, and varied structures including multi-floor
apartments. The sampler's placement may not always have yielded a representative sample
of the ambient air but in most cases did represent air impacting directly on the participant's
7-2
-------
TABLE 7-1. SUMMARY OF FLOW RATE AUDIT RESULTS
Performance
Sampler Type Date of Sampler Audit Flow Acceptable Within Outside
Flow units Audit Flow Flow Difference Range Tolerance Tolerance
Hi-VOL PM10
Sampler "D'1
(rrr/min)
9/28/90a
1.13
1.07
0.06
+ 0.11
X
Hi-VOL PM10
(m3/min)
Sampler *DN
9/28/90b
1.13
1.07
0.06
+ 0.11
X
Hi-yOLPM10
(m /min)
Sampler "A"
9/28/90a
1.13
1.08
0.05
±0.11
X
Hi-yOL PM10
(m /min)
Sampler "A"
9/28/90b
1.13
1.08
0.05
+ 0.11
X
Dichotomous
Sampler "B"
(L/min)
Total
Fine
Coarse
10/29/90a
16.67
15.0
1.67
15.60
13.95
1.56
1.07
1.05
0.11
+ 1.7
+ 1.5
+ 0.17
X
X
X
-------
TABLE 7-1. (continued)
Performance
Sampler Type Date of Sampler Audit Flow Acceptable Within Outside
Flow units Audit Flow Flow Difference Range Tolerance Tolerance
Dichotomous
Sampler HC"
(L/min)
Total
Fine
Coarse
10/29/92°
16.67
15.0
1.67
15.75
14.27
1.56
0.92
0.73
0.11
+ 1.7
+ 1.5
+ 0.17
X
X
X
Dichotomous
Sampler "C"
(L/min)
Total
Fine
Coarse
9/28/903
16.67
15.0
1.67
15.25
13.47
1.62
1.42
1.53
0.05
+ 1.7
+ 1.5
±0.17
X
X
Dichotomous
Sampler "B"
(L/min)
Total
Fine
Coarse
9/28/90
16.67
15.0
1.67
15.34
13.57
1.57
1.33
1.43
0.10
+ 1.7
+ 1.5
+ 0.17
X
X
PEM PM 302
(L/min)
9/30/90
4.0
3.92
0.08
+ 0.4
X
PEM Cassella P.
(L/min)
9/30/90
4.0
3.96
0.04
+ 0.4
X
PEM B10017
(L/min)
9/30/90
4.0
4.07
-0.07
+ 0.4
X
PEM P-335
(L/min)
10/29/90
4.0
3.84
0.16
+ 0.4
X
PEM FM035
(L/min)
10/29/90
4.0
3.87
0.13
+ 0.4
X
-------
TABLE 7-1. (continued)
Sampler Type
Flow units
Date of Sampler Audit Flow Acceptable
Audit Flow Flow Difference Range
Performance
Within
Tolerance
Outside
Tolerance
PEM FM073
(L/min)
PEM FM061
(L/min)
PEM FM077
(L/min)
PEM PM316
(L/min)
SAM B044
(L/min)
SAM P-2100
(L/min)
SAM FM014
(L/min)
SAM FM002
(L/min)
SAM
(L/min)
10/29/90
10/29/90
10/29/90
10/29/90
9/30/90
9/30/90
9/28/90
9/28/90
9/30/90
4.0
4.0
4.0
4.0
4.0
4.0
4.0
4.0
4.0
3.63
3.98
3.88
3.93
3.97
3.97
4.28
4.30
4.25
0.37
0.02
0.12
0.07
0.03
0.03
0.28
0.30
0.25
±0.4
+ 0.4
±0.4
+ 0.4
±0.4
+ 0.4
+ 0.4
±0.4
+ 0.4
-------
TABLE 7-1. (continued)
Sampler Type
Flow units
Date of Sampler Audit Flow Acceptable
Audit Flow Flow Difference Range
Performance
Within
Tolerance
Outside
Tolerance
SAM
(L/min)
SAM FM049
(L/min)
SAM FM057
(L/min)
SAM FM038
(L/min)
SAM FM001
(L/min)
SAM FM049
(L/min)
SAM P-2152
(L/min)
SAM P-3079
(L/min)
SIM P-3040
(L/min)
9/30/90
9/30/90
9/30/90
10/29/90
10/30/90
10/30/90
10/29/90
10/29/90
9/30/90
4.0
4.0
4.0
4.0
4.0
4.0
4.0
4.0
4.0
4.17 -0.17
4.20 - 0.20
4.08 - 0.08
4.08 - 0.08
3.95
3.87
0.05
0.13
4.38 - 0.38
4.23 - 0.23
4.06 - 0.06
±0.4
+ 0.4
+ 0.4
+ 0.4
+ 0.4
±0.4
+ 0.4
+ 0.4
+ 0.4
-------
TABLE 7-1. (continued)
Sampler Type
Flow units
Performance
Date of Sampler Audit Flow Acceptable Within Outside
Audit Flow Flow Difference Range Tolerance Tolerance
SIM P-2141
(L/min)
SIM P-2141
(L/min)
SIM P-3063
(L/min)
SIM P-3063
(Umin)
SIM P-3070
(L/min)
SIM FM053
(L/min)
SIM P-2179
(L/min)
SIM FM029
(L/min)
9/30/90a 4.0
mom1
a
SIM ---
(L/min)
9/30/90
9/30/90
9/30/90
9/30/90
9/30/90
10/29/90
10/29/90
4.0
4.0
4.0
4.0
4.0
4.0
4.0
4.0
4.01
4.00
3.98
4.04
3.73
4.05
3.97
3.88
3.83
-0.01
0.00
0.02
0.04
0.27
0.05
0.03
0.12
0.17
+ 0.4
+ 0.4
+ 0.4
+ 0.4
+ 0.4
+ 0.4
+ 0.4
+ 0.4
+ 0.4
-------
TABLE 7-1. (continued)
Sampler Type
Flow units
Date of Sampler Audit Flow Acceptable
Audit Flow Flow Difference Range
Performance
Within
Tolerance
Outside
Tolerance
SIM FM077
(L/min)
SIM FM053
(L/min)
Cascade Sampler
(L/min)
Cascade Sampler
(L/min)
10/29/90
10/29/90
9/28/90
4.0
4.0
10/29/90 28.3
4.02
4.05
76.4° 28.4
34.3
-0.02
0.05
48.0
-6.0
+ 0.4
+ 0.4
+ 2.8
+ 2.8
j* Flow rate of sampler audited prior to sampling.
Flow rate of sampler audited after sample collection.
c Flow rate indicated on cascade dry gas meter, obviously a leak. Operator adjusted sampler flow rate until auditor dry gas meter
equaled 28.316 L/min so that a sample run could begin. Operator was to then check sampler flow rate after sampling and repair
the leak.
d Sampler being operated within the manufacturer's design criteria (+ 10% flow rate) however, sampler calibration different from
the auditors indicated flow by greater than + 10 percent.
-------
TABLE 7-2. SUMMARY OF AUDIT RESULTS FROM BALANCE AUDIT
Balance Performancea
Instrument Date of Indicated Audit Weight Acceptable" Within Outside
Type and Units Audit Weight Weight Diff. Range Tolerance Tolerance
Cahn Model C-31
Microbalance
9/30/90
50.012
50.008
0.004
+ .050
X
(units are
70.028
70.023
0.005
X
milligrams)
80.043
80.037
0.006
X
99.993
99.988
0.005
X
110.007
110.002
0.005
X
120.009
120.003
0.006
X
170.021
170.011
0.010
X
200.005
199.993
0.012
X
Analytical
9/30/90
5.0000
5.0000
0.0000
±.005
X
Balance
6.9995
7.0000
- 0.0005
X
Metier,
9.0000
9.0001
- 0.0001
X
Model H15
4.0000
4.0001
-0.0001
X
(units are grams)
4.4999
4.5001
- 0.0002
X
2.4999
2.5000
- 0.0001
X
-------
TABLE 7-2. (continued)
Balance Performance3
Instalment Date of Indicated Audit Weight Acceptable" Within Outside
Type and Units Audit Weight Weight Diff. Range Tolerance Tolerance
Cahn, Model C-31
10/27/90
50.012
50.008
0.004
+ .050 x
micro balance
70.031
70.023
0.008
X
(units are
80.047
80.037
0.010
X
milligrams)
99.997
99.988
0.009
X
110.002
110.002
0.012
X
120.013
120.003
0.010
X
170.028
170.011
0.017
X
199.990
199.970
0.020
X
Metier Model H15
10/27/90
2.497
2.500
0.003
+ 0.005 x
Analytical
4.000
4.000
0.000
X
balance
4.500
4.499
0.001
X
(units are
5.000
4.998
0.002
X
in grams)
7.000
6.998
0.002
X
Performance of each balance based on the average performance, both of these balances are considered to be within
tolerance.
Absolute difference not as important as is repeatability of the balance because both the initial and tare weights are
determined on same balance.
-------
residence. PEM samplers were bulky and somewhat awkward, and could not remain attached
during vigorous exercise, perhaps resulting in activity modification by the participant For
example, at least one participant stayed home rather than going to work while wearing the
PEM. Finally, project documents were provided only in English and Spanish speaking
interviewers and sampling team members were not always available to visit homes at which
Spanish was the primary language. These language barriers may have influenced
participation rates and could have resulted in inaccurate answers on the survey documents.
Additional efforts must be made to remove language barriers in future studies. Participant
appointment scheduling was originally established on a random-day assignment procedure.
With this strategy five appointments went unfilled in the first three days of sample collection.
The random-day selection procedure was dropped in order to achieve the necessary data
recovery on schedule.
Observations at the Temporal Site
Operations at the temporal site in some cases were not well-integrated into the overall
study operations. This was indicated by the fact that provisions were not made in the data
management system to include the high volume or dichotomous samplers. The operator used
a Lotus 1 -2-3 spreadsheet to manage data; these spreadsheets were submitted to RTI at the
completion of the study. Initial auditing activities found that the operator was not completely
prepared to collect data. The dichotomous samplers were not being operated close to the
sampler's design flow rate, the cascade impactor was uncalibrated and leaking, and the
meteorological system was not recording data. In addition, the temporal site was being
affected by voltage surges that tripped the circuit breakers several times and shut down the
entire site. Many of these problems were corrected as a result of the initial audit.
Filter Weighing and Handling
The weighing room facility was designed and built by Acurex personnel. Previous
PTEAM studies have experienced problems with the weighing facilities and substantial
resources were invested during this study to ensure an acceptable quality of data. The
operator did experience problems with the micro-balance that required a service
representative. No weight data were lost due to balance malfunction. To satisfy the QC
requirements, some batches of filters had to be re-weighed, sometimes two and three times.
7-11
-------
Only 15% of the original weighings failed the +4 ng limit, a much better rate than was
achieved with a +10 ng limit during the pre-pilot study. Multiple weighings of the same filters
were not recognized by the data management software, resulting in incorrect calculation of
concentrations. The software used the first weight value recorded rather than the weight that
satisfied the QC requirements. This problem was resolved by Acurex and RTI personnel.
QA procedures for the weight room included the use of one reference weight. Field
experience, however, showed the need for more than one reference weight, so this procedural
change was incorporated in the filter weighing routine. Because QC software would not
accommodate the entry of two QC weights, the QC acceptance criteria were evaluated without
the use of the computer.
One room in the Riverside study office was designated as the filter loading and sample
handling area. After observing sane incidences of filter contamination, an air purifier was
purchased and operated in this area, access to the room was limited, and cleaning
procedures were instituted. These efforts lowered the incidence of filter contamination. A
group of sample impactors was used twice without removing or replacing the filters, causing
two samples to be lost for each sample head. After this experience, filter handling procedures
were evaluated and some procedural steps altered to prevent a recurrence.
AUDIT OF DATA QUALITY RESULTS
An audit of data quality was completed on the field study databases. The field study
data files available for review are as follows:
Temporal Site and Ambient Data -
CASCADE - data from cascade impactor runs at temporal site. 05/08/91. 40 records,
16 variables. DISK 2.
CASCALB - cut size calibration data for cascade impactor. 05/08/91. 104 records,
8 variables, DISK 2.
TEMPORAL - gravimetric data from SAMs and PEMs operated at the temporal site, and
data from the co-located PEMs worn by PTEAM staff. 05/08/91. 332 records, 60
variables, DISK 2.
DICHOT - data from the two dichotomous samplers operated at the temporal site, as
obtained from Acurex. 05/08/91. 192 records, 52 variables, DISK 2.
7-12
-------
WEDDING - data from the two Wedding samplers operated at the temporal site, as
obtained from Acurex. 05/08/91. 192 records, 35 variables, DISK 2.
DICBLK - weight data on blank filters for the dichotomous samplers operated at the
temporal site. 05/08/91. 41 records, 3 variables, DISK 2.
WEDBLK - weight data on blank filters for the Wedding samplers operated at the
temporal site. 05/08/91. 44 records, 3 variables, DISK 2.
METDATA - meteorological data from three nearby airports during the time of PTEAM
data collection. 05/08/91. 1176 records, 14 variables, DISK 2.
AVMETDAT - meteorological data from three nearby airports summarized into 12 hour
periods Airing the time of PTEAM data collection. 05/08/91. 96 records, 34 variables,
DISK 2.
Gravimetric Analysis Data --
BLANKS - data from various types of blank particle filters. 05/08/9! 61 records,
36 variables, DISK 2.
QAWTS - data from QA weighings of particle filters. Filters may have been weighed
twice before or after sampling or at both points. 05/08/91. 640 records, 9 variables,
DISK 2.
FIELD - gravimetric data from SIM and SAMs operated in participants' homes and PEMs
worn by participants. 05/08/9! 1836 records, 60 variables, DISK 3.
ROTAMETER - calibration data for rotameters used for flow checks in field. 05/08/91.
6 records, 7 variables. DISK 2.
MHINFO - information on sampled households such as room considered 'main living
area (SIM site)', single or multi-family structure, 'dirt level' in home, and indicator of PAH
sampling. 05/08/91, 184 records, 10 variables, DISK 2.
Questionnaire Data --
QUEST1 - data from the household questionnaires administered to the participants.
This file contains data for Period 1 of monitoring. 05/08/9! 191 records, 125 variables,
DISK !
QUEST2 - additional data from the household questionnaires administered to the
participants. This file contains data for Period 2 of monitoring. 05/08/9! 191 records,
125 variables, DISK !
7-13
-------
QUEST3 - additional data from the household questionnaires administered to the
participants. This file contains data not specific to a particular monitoring period.
05/08/91. 191 records, 76 variables, DISK 1.
SCREENER - data from the household screening forms administered to all the
households. 05/08/91. 680 records, 72 variables, DISK 1.
DIARY - data from the time and activity diaries filled out by the participants. 05/08/91.
3536 records, 32 variables, DISK 2.
SAMPWTS - calculated weights and other sampling data for the sample housing units
for which a person was selected for monitoring. See memo of February 26, 1991 for
notes on use. 05/08/91. 257 records, 14 variables, DISK 1.
Elemental Analysis Data --
XDIC_SRM - elemental data from XRF analysis of SRM samples during analysis of
dichot filters. 16 records.
XDIC_BK1 - elemental data from XRF analysis of laboratory blank samples during
analysis of dichot filters. 16 records.
XDIC_BK2 - elemental data from XRF analysis of dichot field blank filters. 8 records.
XFDC_FM - elemental data from XRF analysis of fine dichot filters. 195 records.
XCDC_FM - elemental data from XRF analysis of coarse dichot filters. 195 records.
XPM_SRM - elemental data from XRF analysis of SRM samples during analysis of
SIM/SAM/PEM filters. 72 records.
XPM_BK1 - elemental data from XRF analysis of laboratory blank samples during
analysis of SIM/SAM/PEM filters. 65 records.
XPM_BK2 - elemental data from XRF analysis of SIM/SAM/PEM field blank filters.
60 records.
XPM_TEM - elemental data from XRF analysis of SIM/SAM/PEM filters from temporal
site samplers. 325 records.
XPM_HOM - elemental data from XRF analysis of SIM/SAM/PEM filters from samplers
in homes. 1835 records.
Sampler operation and data calculation procedures were referenced to the Pilot Study
Work Plan, Volume II (protocols). For the EPA-approved PM-jq samplers operating at the
7-14
-------
temporal site, EPA guidelines were also referenced. These guidelines are the "Reference
Methods for Determination of Particulate Matter as PM^q in the Atmosphere," Sections 2.10
and 2.11 of the Quality Assurance Handbook for Air Pollution Measurement Systems.
Volume II. for the dichotomous and Wedding samples, respectively.
Since the concentrations are all calculated by a computer, the data were examined in
three different perspectives: (1) correctness of calculations/formulas that were used to
generate the concentrations; (2) data input, specifically weights and volumes, completeness of
records, acceptable limits, and accuracy of data entries; and (3) representativeness; i.e., do
the data reflect expected concentrations, and are the numbers within reasonable ranges?
Temporal Site and Ambient Data
PM-jq temporal site sampler concentrations were averaged for each respective run,
including total concentrations from dichotomous samplers B and C, Wedding samplers A
and D, and pump FM014 (SAM). Then each respective sampler concentration was subtracted
from the average. The resulting differences were used to identify outlier data. These data
were evaluated for the correctness of calculations, data input, and representativeness.
Outlier data evaluated included the following:
• For run 32, dichotomous sampler C had a concentration of 257.4|ig/nn^. This high
concentration can be attributed to a catch of 521,8jxg (fine + coarse) and a run
time of 267 minutes, or 4.45 hours. This was appropriately flagged in the
database.
• Runs 37 and 38 had excessively high catches on both Wedding samplers. These
data were appropriately flagged in the database.
• Run 40 indicated a 10(ig catch for dichotomous sampler B (the coarse final weight
was omitted for this run). This was corrected in the database.
• For run 70, dichotomous sampler C had a concentration of 243.4jig/m3. This high
concentration can be attributed to a coarse catch of 2079.Ojig; however, looking at
the coarse filter tare weight (106.81) and a coarse final weight (107.37), the total
coarse catch should have been 560.0|i.g, not the reported 2079.0p.g. This error
was corrected within the database.
• Run 92 yielded a range of concentrations from 145.4ng/m3 to 241.7 ng/m3 across
the samplers. High speed Santa Ana winds were blowing during Run 92 causing
a great deal of blowing dust and sand at ground level.
7-15
-------
PM2 g temporal site sampler concentrations were averaged together for each run.
Then, each respective sampler concentration was subtracted from the mean concentration.
The resulting differences were used to recognize outlier data. Once again, these data were
evaluated for correctness of calculations, data input, and representativeness. All outlier data
identified were appropriately corrected or flagged within the database.
Additional data validation and OA support activities were conducted by the QA/QC
team during the evaluation of the temporal site and ambient data to verify the accuracy of the
databases, as follows:
• Concentrations were verified by recalculation using the rotameter indicated
readings, rotameter calibration equation, and time on/time off. The rotameter
calibration equation is traceable to NIST, although the calibration dates are as
much as 1.5 years old.
• Calculations for each rotameter were performed to verify the data set.
• Weights for this database were verified by a visual comparison with the weighing
facility notebook. Ten percent of the entries were checked. Reweighs and QC
pass/fail checks were examined for any deviation from protocol.
• Field monitoring coordinators and database management personnel have
performed QC checks on the data set for accuracy of start/stop times, filter
identification numbers, and base/filter head identification numbers.
• The database needed to incorporate ambient temperature and pressure, and
calibration equation data so that it would be all-inclusive and any calculations
could be made without requiring additional data. These data were added by RTI
personnel to the database. Calculations were conducted to assure that the
appropriate temperature/pressure corrections were applied to the data. Table 7-3
contains the comparisons.
• Wedding sampler concentrations were standardized using the seasonal averages
for the Riverside area, and not using ambient conditions. Wedding concentrations
were later standardized at RTI using ambient conditions.
• Wedding sampler field data sheets were compared to the data file. Some
discrepancies were noted and corrected in the database.
• The Wedding samplers had 17 days of flow rates that exceeded the design flow
rates. There were six days of no flows, and eight entries of P1 pressure readings
were transcribed incorrectly.
• One of the two dichotomous samplers at the temporal site was operated with the
coarse flow rate slightly below the specified ± 10% tolerance in the design flow
7-16
-------
rate. This was a result of the operator not conducting QC checks on both the total
and coarse flow rates even after a reminder of this procedure during the initial field
audit. An effort was made to evaluate the impact of this sampler's operation by
comparing it's results with the second collocated dichotomous sampler. One
potential impact would have been a biasing of the fine data high. The median
% RSD for 93 collocated fine sample pairs was 3.2%, and the mean was 5.6%
suggesting any bias was minimal. Neither dichotomous sampler produced fine
concentrations that were consistently higher than the other, in terms of total
PM.jq, the median %RSD for 90 collocated samples was 4.5% with a mean of
5.5%, once again, a very acceptable level of precision. In this case the
dichotomous sampler with the lower coarse flow rate resulted in higher total PM10
concentrations than the other sampler in many, but not all, sampling periods. In
all comparisons with other sampling methods, data for the two dichotomous
samplers were averaged further reducing any small potential bias.
• Dichot sampler concentration data were standardized using the seasonal averages
for the Riverside area, and not ambient conditions. Dichot concentrations were
later standardized at RTI using ambient conditions.
• Dichotomous filter weights were verified against the notebook values. Filter
weighings were entered in the notebook using a different colored pen. This aided
in tracking filters.
• Dichot sampler rotameter calibrations were conducted according to the latest EPA
guidance documentation, i.e., by adjusting the concentrations for that portion of the
fine particles that end up on the coarse filter.
The SAM samplers (FM014 - PM10 and FM002 - PM2 5) operating at the temporal site
indicated higher concentrations than the other PM^ Q and PM2 5 samplers operating at the
same site. This difference between the samplers' indicated concentrations is discussed in
greater detail in the statistical analysis section (Section 9) of this report.
Cascade impactor samples were collected only for the purpose of evaluating particle
size distributions in the ambient air at the Riverside temporal site. Each cascade impactor
sample was collected over multiple (8 to 21) 12-hour monitoring periods in order to collect
sufficient particle masses at each stage for weighing. The cascade data should be used only
for particle size distributions and not as a measure of absolute concentrations. No precision
estimates are available for the size distributions since only one cascade impactor was
operated. A second cascade impactor should be included in the future to allow
measurements for precision.
7-17
i
-------
TABLE 7-3. TEMPERATURE/PRESSURE CORRECTION CALCULATIONS
COMPARED TO AUDITOR CALCULATIONS
Study Data
Actual Flow
Max/Min
Max/Min
Mean
Flow
Audit Flow
Liters/min
°F
°C
°C
Liters/min
Liters/min
Black Box
(Qa)
Pa/Pstd
(Ta)
^std
(Qstd)
4577
2.597
.9729
88/66
31.3/18.9
25.0
2.525
2.527
4578
2.034
.9476
86/68
30.0/20.0
25.0
1.987
1.982
4580
3.225
.9713
78/73
25.6/22.8
24.2
3.144
3.141
4582
2.396
.9743
84/64
28.9/17.8
23.3
2.343
2.348
Standard conditions are defined for these data as 25°C and 760 mmHg
p*
Qstd ~ Qa
a Tstd
pstd
-------
At the onset of the Pilot Study, the meteorological data were to be collected using a
Climatronics meteorological system. Soon after the beginning of the study, however, the
system malfunctioned and did not record any data. Auditing activities identified that the
meteorological system was not operating and the site operator was instructed to make the
necessary repairs. After the initial audit, the Climatronics system was repaired. Unfortunately,
data logging malfunctions occurred and the system did not operate properly for most of the
Pilot Study. To correlate the PM^q concentrations with meteorological events, the
meteorological data from surrounding weather stations were obtained from the National
Oceanic and Atmospheric Administration's National Climatic Data Center in Asheville, North
Carolina. The data are considered to be of acceptable quality, but have not been scrutinized
by the QA team.
Gravimetric Analysis Data
The following are notes and details of QA support activities regarding the gravimetric
databases:
• The weights for the 37-mm filters were scrutinized by checking approximately 10%
of the entries. This was done by comparing the weights data printout and the
laboratory notebook. Specific items looked for were transcription errors, initial and
final reweighs, tares initial and final, and quality control reweighs for pass/fail.
• The database contains tare weights for some filters, sometimes re-tares, but not
finals. Not all tared filters were used for sample collection.
• The database contains only the re-tare and not both tare and re-tare, indicating
that the filter weights used were the weights that satisfied the QC criteria.
• The database contains only the acceptable final weights used by the program.
• 10% or one out of every ten filter weights in the database was verified against the
notebook.
• There were no discrepancies between the notebook and the database with the
exception of last digit rounding.
• The final weight of filter F5526 was less than the tare weight (final @ 106.645 and
tare @ 106.674). This data record was flagged in the database.
• The indicated catch within the database was verified by the catch calculated using
the notebook entries. There were no discrepancies.
7-19
-------
• The notebook entries were not always included in the data on the same day
indicated by the notebook. This may be an indication of the balance operator
either working past midnight or beginning data entry the next day,
• Six rotameter calibrations were conducted using NIST-traceable volumetric
standards.
• Six rotameters were used during the field study to establish the flow rates of the
PEM, SIM, and SAM samplers. All were calibrated between 2/20/89 to 9/4/90,
This time period may be considered excessive, but the performance audits
indicated acceptable agreement.
• The database identified rotameter serial numbers and indicated starting and
stopping divisions. The average sample flow was calculated using the appropriate
rotameter calibration.
• Calculated concentrations were validated using checked filter weights from the
weight notebook and the rotameter calibrations.
Elemental Analysis Data
All PEM, SIM, SAM, and dichotomous sample filters were analyzed by wavelength-
dispersive X-ray fluorescence at the EPA facility in Research Triangle Park. Over 99% of the
acceptable filters with particle samples from Riverside were successfully analyzed by XRF. A
subset of 107 filters was analyzed by Lawrence Berkeley Laboratories as a QC check for the
EPA analysis. A second subset of 26 filters was analyzed by a different XRF laboratory using
an energy-dispersive methd prior to any EPA analyses in order to evaluate whether the
elements of interest would be measurable on sufficient filters for further data analysis. A
detailed discussion of QC data including the analysis order selection process, sampling and
analysis precision, instrument bias over time, and interlaboratory results appears in Section
10. A quality assurance review was performed on the elemental data reported by the three
laboratories. A memorandum describing the results of the review is included in Appendix S.
Procedures used to convert blank and QC sample data from ng/cm2 to ng/m3 using an
assumed flow rate of 4 L/min for 720 minutes are described. These conversions were used to
make all XRF data comparable. Calculations for applying particle size correction factors to
the LBL data, and comparisons of data for three elements on several filters are also included.
The calculations used to implement the corrections were manually verified before the LBL
elemental data were accepted into the PTEAM database.
7-20
-------
SUMMARY OF QUALITY ASSURANCE OBJECTIVE RESULTS
Quality assurance objectives were outlined in the Quality Assurance Project Plan for
the PTEAM Pilot Study. Detailed analyses of completeness, sample collection and analysis
precision, and accuracy are reported in Sections 6, 9, and 10. Results of these data analyses
will be summarized here with reference to the original quality assurance objectives (QAOs).
Completeness
The original QAO objective for completeness was data collection for at least 95% of
175 participants who agreed to participate in the study. In terms of achieving this objective,
sampling was completed for 178 participants in Riverside. The QAO for completeness was
met for all categories of sample collection except cascade impaction monitoring. Individual
measures of completeness for each sample type are listed in Section 6, Table 6-7. For all
types of particle samples, the number of successfully collected samples was equal to or above
95% of the originally scheduled number of samples for 175 participants. In terms of the
number of particle sample collections actually attempted, completeness exceeded 92% in all
cases and in most cases was above 95%. One exception was cascade monitoring where the
completion rates were 65% for those originally scheduled and 80% of those attempted. The
target number of participants for PAH/phthalate collection was 120 indoors and 60 outdoors.
Because of operational difficulites with the pumps in Riverside the number of participants
scheduled for PAH/phthalate monitoring was increased to 125 indoors and 65 outdoors.
Sample completion rates were kept above 95% of the originally scheduled samples by
increasing the number of participants monitored. Sample completion rates exceeded 94% for
all categories of air exchange and nicotine samples. Over 99% of the samples submitted for
X-ray fluorescence analysis of elements were successfully analyzed.
Precision
Particle Sample Collection and Weighing-
Original QAO objectives for precision were 10% mean relative standard deviation for
PEM, SIM, and SAM methods and 7% for dichotomous and Wedding (hi-vol) methods. All
particle collection systems met the QAO for sample collection and analysis precision.
Distributions of %RSDs for paired collocated samples of each type are reported in Section 9,
Table 9-3. PEM, SIM, and SAM median RSDs were lower than 5.4% and the mean RSDs
7-21
-------
were all lower than 7.5%. Wedding and dichotomous median RSDs were less than 4.6% and
mean RSDs were less than 5.7%. No collocated samples were collected for the cascade
impactor.
Elemental Analysis-
The QAO for elemental analysis precision was 5% mean relative standard deviation.
Precision was measured in several ways during the elemental analysis. First, sample
collection and analysis precision were evaluated by analyzing collocated samples. Results
are reported for the 15 elements most often found at measurable concentrations in Section
10, Table 10-10. Median and mean RSDs exceeded 5% for most elements and most sample
types. The range of median RSDs was from 0% to 37%, with most median RSDs less than
10%. Mean RSDs for most elements were less than 14%. Second, analysis precision over
time was evaluated by blind reanalysis of selected filters at an analysis time one to three
months after the original analysis. Results are reported in Section 10, Table 10-12 for
duplicate analyses for the same PEM, SIM, and SAM filters. Median RSDs range from 1.4%
to 14% for fifteen elements, with most having RSDs less than 10%. These results are also
explored from the aspect of analysis date in Table 10-13. Small but significant differences
were observed for some elements over the analysis time, suggesting a possible small change
in instrument response. Results for the analyses of standard reference materials, discussed
In Section 10, also indicate small (2-5%) changes in instrument performance over the several
months in which samples were analyzed. The largest differences were for chlorine, possibly
indicating losses due to volatility during storage or analysis. The QAO of 5% RSD was not
met for many elements. This QAO was based on the results from XRF analyses made during
Pre-Pilot Study using a different XRF technique from the one employed at the EPA facility in
this study. This data should be considered before setting future QAOs for elemental analysis.
Also, if the expected precision for particle collection is 10% mean RSD, it may be unrealistic
to expect elemental precision to be any better than 10% mean RSD.
Accuracy
Particle Sample Collection and Weighing--
The QAO for sample collection accuracy was based on a 10% tolerance in the
sampler's flow rate. Audit results in Table 7-1 indicate that PEM, SIM, SAM, and Wedding
7-22
-------
samplers met this QAO. One of the two dichotomous samplers was operated outside of the
10% tolerance range. (Concentration data from this sampler was discussed earlier in this
section). The other dichotomous sampler met the tolerance criteria. Formal QAOs were not
set for balance accuracy, but both balances used in this study passed performance audits as
described in Table 7-2.
Elemental Analysis-
Accuracy QAOs for the elements ranged from 5% to 20% bias based on the analysis
of NIST standard reference materials (SRMs), with the QAO for each element depending on
the specified elemental uncertainty for the SRM. All elemental QAOs for accuracy were met,
based on multiple analyses of the SRMs. The SRM results are summarized in Section 10,
Table 10-7. All median % biases were equal to or smaller than +5.5% except for calcium at
7.1%. Maximum % biases for a single element in a single SRM analysis never exceeded
±10.4%.
Detection Limits and Blanks
Particles-
The QAO for particle method detection limits (MDLs) for PEM sampling was 8 |xg/m^
and those for the SIM and SAM were 4 jig/m^. The lower MDLs for the SIM and SAM were
based on the original expectation that their flow rates would be 10 L/min as compared to the 4
L/min used by the PEM. A final decision was made before the study began to use a 4 L/min
flow and the same impactors for PEM, SIM, and SAM for better comparability. The MDLs for
dichotomous and Wedding samplers were 10 and 5 ng/m^ respectively. All MDLs were
calculated from the standard deviations in the field blank data using the following formula:
MDL = (Field Blank Std. Dev.) x (one-tailed t-statistics for N-1 degrees of freedom)
where N is the number of field blanks collected and weighed. All 51 PEM, SIM, and SAM field
blanks were grouped together to calculate an MDL since their sample collection parameters
were identical. The mean mass increase for these blanks was 9.5 |xg, the standard deviation
was 9.0 ^g, and the t-statistic at the 0.99 level for 50 degrees of freedom is 2.403. The
calculated MDL mass increase was 22 jxg. A concentration based MDL was then calculated
assuming a sampler flow of 4 L/min for 12 hours resulting in a total sample volume of
7-23
-------
2.88 m3. The calculated MDL for PEM, SIM, and SAM samples was 7.5 jig/m3, which meets
the QAO. Similar calculations were performed for the dichotomous and Wedding samplers
and the MDL results are reported in Table 7-4. All particle sampling methods achieved the
QAOs.
Elemental Analysis-
Detection limits were not directly calculated for the analysis of elements by the XRF
method used at EPA. Uncertainty limits were used instead, and these were based on several
factors including the element of interest, the concentration of the element in the sample, and
the propagated uncertainty calculated for sampling and analysis parameters. The uncertainty
limit was calculated as three times the propagated uncertainty. Uncertainty limits changed
from sample to sample for each element. Tables 10-8 and 10-9 in Section 10 describe mean
uncertainty limits and their standard deviations for lab and field blanks. No specific QAOs
were set for element detection limits. The percent of samples containing measurable (above
the uncertainty limit) concentrations of each element is described in Section 10, Tables 10-2
through 10-5.
EFFECTIVENESS OF QA/QC PLAN
Sampling Procedures
The sampling procedures effectively supported the QA objectives by providing
estimates for precision and accuracy for the sampling and analytical methods documented.
Prior to conducting the field study, a dress rehearsal was used to train field personnel and as
a method to test and refine the sampling procedures. Some procedural changes occurred
during and after the rehearsal, but none after the initiation of the Pilot Study.
Sample Custody
Sample custody procedures outlined in the QAPP were effective, and only minimal
custody problems occurred. RTI personnel were responsible for sample custody during field
operations and the follow-up sample analysis. Bar codes were assigned to each sample to
facilitate tracking. These bar codes were then scanned into the computerized database at
7-24
-------
TABLE 7-4. METHOD DETECTION LIMIT AND FIELD BLANK DATA FOR PARTICLE SAMPLING
Field Blank Type
Number
Collected
Median ng
Mass
Increase
Mean |ig
Mass
Increase
Standard
Deviation
t-statistics
(0.99 level)
Calculated Method
12-Hour Sang>le Detection £imit,
Volume m"
lig/nT
PEM, SIM, SAM 51
Dichotomous 37
9
4
9.5
5.4
9.9
3.B
2.403 2.88
2.432 1.15 (coarse)
11.5 (fine)
7.5
8.0
0.8
Wedding
43
100
200
500
2.416
815
1.5
-------
several stages of data collection. Tracking the sample through these stages and maintaining
hard-copy data forms provided a thorough record of sample custody. No sample custody
changes are anticipated for future studies.
Calibration
The calibration procedures effectively defined the samplers' response and generated
data acceptable for the study.
Data Reduction. Validation, and Reporting
As indicated earlier, the temporal site needed to be better integrated into the study.
This could best be accomplished by including all temporal samplers into the data management
system. This would improve communication between the field supervisor and temporal site
operator and also the timeliness of data being reported.
Internal Quality Control Checks
Weighing room facilities functioned satisfactorily and the QC checks were sufficient to
generate data of acceptable quality. Internal QC was not deficient, but the way QC data were
used to validate data needs improvement. The data management software needs to be
revised and tested. Testing of this software should ensure that the most current weight is
used to calculate concentrations.
Performance and Systems Audit
In conclusion, the performance and systems auditing activities were sufficient to
ensure data of acceptable quality.
Two audits were conducted during this study, one at the beginning and one near the
end. The effectiveness can be measured by the auditor's familiarity with the study procedures
and the operational details identified during the two audits. The two audits provided the
auditors with the data to make conclusive statements about the temporal site and to ensure
that operational details were reviewed and discussed with study personnel.
7-26
-------
Data Precision. Accuracy, and Completeness
Data precision, accuracy, and completeness have been documented for all the field
measurements used during the Pilot Study.
QA Reports
Preliminary results of the audit were discussed with the field study personnel at the
conclusion of each audit. Deficiencies identified during the audit were discussed and
recommendations provided.
Performance and systems audits and audits of databases are reported within RTI in
the form of memoranda. The database memoranda are included with the study database and
are available with the appropriate data. Copies of pertinent performance and system audit
and database memoranda are included in Appendix T.
7-27
-------
-------
SECTION 8
SAMPLE WEIGHTING AND NONRESPONSE ADJUSTMENTS
The Pilot PTEAM Study was based on a probability sample selected from a
well-defined population, as discussed in Section 4. Therefore, proper analysis of the PTEAM
data, when appropriate, must be based on the probabilities of selection and the other features
of the sampling design, such as stratification and multistage sampling.
Sampling weights, defined as the reciprocals of the probabilities of selection, enable
design-unbiased estimation of linear population parameters, such as population totals (see
Figure 8-1). In addition, the sampling weights are adjusted to partially compensate for the
potential bias due to survey nonresponse.
Sampling weights based on the probabilities of selection were computed for each stage
of sampling. Weight adjustments to compensate for potential nonresponse bias were
computed for each level of nonresponse. Analysis weights were then computed to support
statistical analyses for the following sets of PTEAM observations:
• housing units selected for household enumeration
• people selected for particulate monitoring
• homes selected for particulate monitoring
• homes selected for indoor phthalate and PAH monitoring
• homes selected or outdoor phthalate and PAH monitoring.
Estimation of sampling variances and standard errors for statistics calculated from
probability sampling data is properly based on the randomization distribution induced by the
sampling design. The analyses must account for all features of the sampling design, such as
stratification and multistage sampling. Such analyses are robust because they make no
assumptions regarding the distribution of occurrence (e.g., normality) of the survey items.
Hence, analyses based on the design-induced distribution provide the most defensible basis
for inference from the sample to the target population (see Hansen et al., 1983; Williams
etal., 1983).
The classical approach to estimating standard errors for nonlinear statistics, such as
means and proportions, based on complex probability sampling designs is a first-order Taylor
Series linearization method (Tepping, 1968). Alternative variance estimation techniques for
complex survey designs include jackknifing and balanced repeated replication (Skinner et al.,
8-1
-------
Sample Sampling
Stratum No. Pop. Members Stratum Value* Size Weight
1 50,000 150 jig/m3 100 500
2 150,000 50 ng/m3 100 500
'Assume (for simplicity) that everyone in the stratum has the same PM10 exposure
Sample Average = (1Q°)(15°) * 000)(50) _ 100
100 + 100
Population Mean - (50.000)(150) + (150,000)(50) = 75 ^
50,000 * 150,000
Weighted Sample Mean - [(1
-------
1989). Standard statistical software packages (e.g., SAS, SPSS, BMDP, IMSL, etc.) do not
typically include any of these algorithms.
Therefore, RTI has developed special-purpose software for analysis of complex survey
data (Shah et al., 1989). Standard errors are estimated using the classical Taylor Series
method because such estimates are both computationally and statistically efficient. The
software includes procedures for survey-based estimation of standard errors of population
totals, means, proportions, and ratios as well as linear and logistic regression relationships.
RTI software for analysis of complex sample survey data has been reviewed by several
non-RTI researchers and generally found to be the most efficient software currently available
(Cohen et al., 1986; Francis and Sedransk, 1979). RTI used its special-purpose software to
compute sampling variances for survey statistics computed for the Pilot PTEAM study.
Computation of the statistical sampling weights, defined as the reciprocals of the
probabilities of selection, is discussed in the first section below. Statistical adjustments
designed to partially compensate for the potential bias due to survey nonresponse and the
resulting final analysis weights are discussed in the second section.
WEIGHTS BASED ON THE SAMPLING DESIGN
Because a three-stage sampling design was used to select housing units and people
for monitoring, the sampling weights have three weight components, one for each stage of
sampling. At each stage of sampling, the weight component is the reciprocal of the
conditional probability of selection at that stage of sampling.
First-stage sampling units (FSUs) were selected with probabilities proportional to size.
Thus, the weight factors for the first stage of sampling are the reciprocals of the expected
frequencies of selection given by Equation 4-1. Six of the 36 sample FSUs, or area
segments, were subsegmented, and the second factor in the sampling weight for these FSUs
is given by the reciprocals of the conditional probabilities of selection given by Equation 4-2.1
Segment 30 was subsampled twice. The initial segment was estimated to contain
approximately 1,200 housing units. A subarea expected to contain approximately 450 housing
units was first selected, then a second subarea, for which 141 housing units were listed, was
selected from within the first subarea. The product of two weight factors for subsegmenting was
used for this segment.
8-3
-------
Sample housing units were selected for household enumeration interviews at the
second stage of sampling. Sampling weight components were computed as the reciprocals of
the conditional probabilities of selection given by Equation 4-3 for all 780 primary sample
lines. In addition, subsampling to 680 sample lines was incorporated into the sampling
weights by the following weight factor
[{2/3) + (1/3)(162/262)]~1
based on the conditional probability of selection given by Equation 4-4.
Therefore, the sampling weights for the housing units selected for enumeration
interviews were the products of four factors corresponding to: (1) selecting area segments,
Equation 4-1; (2) subsegmenting, Equation 4-2; (3) selecting 780 primary sample lines,
Equation 4-3; and (4) subsampling to 680 sample lines, Equation 4-4. Because the size
measures for the area segments were updated prior to sample selection and the sample
allocation to each segment was based on the number of lines actually listed, the sampling
weights (and unconditional probabilities of selection) were approximately equal for all housing
units selected for household enumeration.
Members of the households that completed the enumeration interview were selected
for particulate monitoring using a two-stage sampling process. Persons were first selected
with the approximately equal conditional probabilities of selection given by Equation 4-5.
Then, 30 percent of the people who did not work at least 30 hours per week outside the home
and lived in a home with no smokers were randomly deleted (a simple procedure for
oversampling the complement in the field). Therefore, the sampling weights for the people
selected for particulate monitoring were the products of the sampling weights for household
screening and the reciprocals of the conditional probabilities of selection given by Equations
4-5 and 4-6.
Because indoor and outdoor particulate monitors were placed at a home if, and only
if, a member of the household was selected for personal monitoring, the conditional probability
of selecting a home for particulate monitoring was computed as the sum of the conditional
probabilities of selection for ail survey-eligible household members (nonsmokers aged 10 or
older), as shown by Equation 4-7. Because the person-level probabilities of selection were
not always equal for all members of a household, the probabilities of selection had to be
computed for all household members and summed to obtain the probability of selecting the
8-4
-------
housing unit for monitoring. Therefore, the sampling weight for each housing unit selected for
particulate monitoring was the product of the sampling weight for household screening and the
reciprocal of the conditional probability of selecting someone from the household for
monitoring, given by Equation 4-7.
Subsamples of the homes selected for particulate monitoring were selected for indoor
monitoring and for outdoor monitoring of phthalates and PAHs. The conditional probabilities
of selection for these subsamples, given only that the homes were included in the screening
sample, are given by Equations 4-8 and 4-9. Weights were computed based on these
probabilities of selection, but they were found to be inefficient; their sum was a poor estimate
of the total number of households in the target population. The problem was that they did not
reflect the fact that the subsamples for phthalate and PAH monitoring were selected from the
homes selected for particulate monitoring, not directly from all homes selected for screening
interviews.
Therefore, we chose to weight the homes selected for phthalate and PAH monitoring
conditionally, given knowledge of which homes were selected for particulate monitoring.
Hence, the conditional weight components for the indoor and outdoor samples for phthalate
and PAH monitoring were calculated as the ratios of the total number of homes selected for
particulate monitoring (irrespective of whether or not they participated) divided by the number
of those homes that were selected for indoor and outdoor samples, respectively, for phthalate
and PAH monitoring. The products of these conditional weight components and the sampling
weights for households selected for particulate monitoring are the sampling weights for
households selected for indoor and outdoor monitoring for phthalates and PAHs.
WEIGHT ADJUSTMENTS FOR NONRESPONSE
The data collected in a probability-based sample survey, like PTEAM, provide the
basis for design-unbiased estimation of linear population statistics, such as population totals,
for the population of people or households that would have responded to a census of the
entire population using the same survey methods. Statistical weight adjustment procedures
are generally used to partially compensate for the potential bias that results from differences
in survey characteristics for respondents and nonrespondents. A complete discussion of
methodology for dealing with incomplete data in sample surveys is found in Madow et al.
(1983).
8-5
-------
Nonresponse occurred in the Pilot PTEAM Study at two stages of sampling:
households selected for screening (enumeration interviews) and households selected for
monitoring. Weight adjustment procedures were used to partially compensate for the potential
bias due to nonresponse. The weight adjustment procedures partition the respondents and
nonrespondents into weighting classes. The sampling weight of each respondent is simply
multiplied by the ratio of a control total for each weighting class divided by the sum of the
sampling weights of the respondents that belong to the weighting class. The adjusted weights
of the respondents then sum to the control total. If the respondents and nonrespondents are
more alike within classes than between classes with respect to their survey responses and/or
their propensity to respond, then nonresponse bias is likely to be reduced. However,
weighting classes are generally required to contain at least 20 to 30 respondents to avoid loss
of precision due to unequal weighting.
Household Screening Nonresponse
The results of the household screening sample for the PTEAM study are presented in
Table 8-1. We see that 632 of the 680 sample lines were occupied permanent residences
and, therefore, eligible for the survey. Of these 632 eligible residences, a completed
screening interview (household enumeration) was obtained for 443 residences. Therefore, the
response rate for the screening phase of the study was 70.1%.
Among the 632 eligible sample households, the three most prevalent reasons for not
completing the household enumeration interview were (1) refusal - 12.0 percent; (2) unable to
contact any household member to be interviewed (mostly no one found at home after
repeated attempts) - 11.7%; and (3) unable to gain access into a controlled-access apartment
complex - 4.7%. The latter two situations in which no one could be contacted to conduct an
interview account for 16.4% of the eligible sample households. This non-interview rate is
unusually high for an area household survey. This high rate occurs partly because
controlled-access apartment complexes were included in the sample for two area segments.
We contacted the apartment managers and attempted to gain access, but were unable to do
so within the time allowed for field data collection. Another contributing factor seemed to be
that many Riverside residents worked in Los Angeles. These people tended to leave home
early in the morning and return home late in the evening.
8-6
-------
TABLE 8-1. HOUSEHOLD ENUMERATION SAMPLE RESULTS
Survey Result Frequency Percentage
Eligible
632
92.9
Completed Enumeration Interview
443
65.1
Refused
76
11.2
No One Home
70
10.3
No Eligible Respondent Home
4
0.6
Language Barrier
5
0.7
Other*
34
5.0
Ineligible
48
7.1
Vacant
41
6.0
Not a Housing Unit
7
1.0
TOTAL
680
100.0
a Could not gain entry into 30 units in controlled-access apartment complexes.
8-7
-------
Finally, the survey schedule itself contributed to the high occurrence of finding no one
at home. Because of pressure to complete at least 175 monitoring appointments with an
absolute minimum of gaps in the appointment schedule, we fielded more sample lines than
were absolutely necessary. Interviewers naturally contacted sample housing units with few
previous attempts before recontacting those for which several attempts had already been
made. This survey effect is supported by the wave-specific response rates: 76% for Wave 1;
77% for Wave 2 (excluding Segment 20, which was entirely controlled-access apartments);
and 61 % for Wave 3.
The variables considered to define weighting classes to adjust for screening
nonresponse must be known for both the screening respondents and the nonrespondents.
Therefore, sampling design variables were the primary candidate variables. The variables
considered for the PTEAM sample were the three first-stage stratification variables: (1) the
four geographic strata; (2) the high and low housing value strata; and (3) the high and low
proportion single-family strata. One disadvantage of the latter two variables is that they were
based on 1980 Census data, which were considerably out-of-date. The response rates for the
low and high percent single-family strata were 64% and 73%, respectively, which at least
partly reflects the inability to gain entry into controlled-access apartment complexes. The
response rates were essentially equal for the two housing unit value strata, approximately
70%. When these two variables were crossed, the response rate effects were inconsistent in
direction, and we decided that they were not appropriate for defining the weighting classes.
Therefore, we decided to use the four geographic strata as weighting classes. The
screening response rates for these weighting classes were as follows:
No.
No. Eligible Responding Response
Geographic Stratum Households Households Rate
Northeast 109 77 70.6%
Southeast 186 140 75.3%
Southwest 77 53 68.8%
Northwest 260 173 66.5%
These strata are appropriate weighting classes because of the differences in response rates
among them, and because they contain reasonably large numbers of respondents. The grand
total of the adjusted, analysis weights for the screening sample, 68,896, estimates the total
number of housing units eligible for screening (permanent residences in the target portion of
8-8
-------
Riverside). Figure 8-2 summarizes the nonresponse adjustment process for household
screening interviews.
Monitoring Phase Nonresponse
Survey results for households selected for personal exposure monitoring are
presented in Table 8-2. The total number of households in which someone was selected for
monitoring was 257. Of these, the number for which particulate monitoring and the study
questionnaire were completed is 178. Therefore, the response rate for the monitoring phase
of the study was 69.3%. Combining this response rate with the 70.1% response rate for the
screening phase of the study produces an overall response rate of 48.5% for the Pilot PTEAM
Study. A response rate this low allows considerable potential for nonresponse bias.
However, it is consistent with the results of comparable exposure monitoring studies, as was
shown in Table 3-1.
Among the 257 households selected for exposure monitoring, Table 8-2 shows that
the three most prevalent reasons for not being monitored were (1) refusal ~ 16.7%; (2) unable
to contact selected participant -- 4.7%; and (3) agreed but never scheduled - 4.3%. Another
3.5 percent of the nonresponse resulted from cancelled or missed appointments that were not
rescheduled. These results strongly suggest that additional sample households could have
been monitored if we had been able (from the cost standpoint) to extend the period of field
data collection. We chose to end field data collection on the day that the targeted number of
households, 175, had been monitored.
The above response rate discussion focuses on sample subjects who completed both
the study questionnaire and personal exposure monitoring. In fact, there are several
additional endpoints that should be considered with regard to response rates. Indoor and
outdoor particulate monitors were placed at homes in which a household member was
monitored. Indoor and outdoor phthalate and PAH monitors were also placed at a subsample
of these homes. Table 8-3 presents the response rates for all phases of the study. The study
questionnaire was completed by 191 sample members for an overall response rate of 52.1%.
Although all these individuals agreed to be monitored, 13 were not monitored which yields the
178 exposure monitoring participants and 48.5% overall response rate previously discussed.
Indoor and outdoor particulate samples were not successfully collected
8-9
-------
Sampling weight based on
probability of selection for
632 eligible housing units
'
Nonresponse adjustment for
443 responding households
based on 4 geographic strata
Figure 8-2. Nonresponse Adjustment Process for
Household Screening Weights
8-10
-------
TABLE 8-2. MONITORING SAMPLE RESULTS
Survey Result Frequency Percentage
Household Selected for Monitoring
257
58.0
Completed Monitoring on Date Assigned®
150
33.9
Completed Monitoring on Another Date®
28
6.3
Canceled Appointment; Not Rescheduled
4
0.9
Missed Appointment; Not Rescheduled
5
1.1
Agreed to be Monitored; Never Scheduled
11
2.5
Unable to Contact Selected Participant
12
2.7
Dropped from the Sampleb
4
0.9
Refused
43
9.7
Household Not Selected for Monitoring
186
42.0
TOTAL
443
100.0
8 Completed study questionnaire and personal exposure monitoring.
b After a member of the community threatened the field sampling team with a shotgun, no
further attempts were made to conduct monitoring in Segment 20.
8-11
-------
TABLE 8-3. SUMMARY OF PILOT PTEAM RESPONSE RATES
Conditional
Overall
Survey Result
Frequency
Rate
Rate
Selected for enumeration
632
Completed enumeration interview
443
70.1%
Selected for monitoring
257
Completed questionnaire
191
74.3%
52.1%
Completed personal particulates
178
69.3%
48.5%
Completed indoor particulates
177
68.9%
48.3%
Completed outdoor particulates
175
68.1%
47.7%
Selected for indoor phthalates and PAHs
181
Completed indoor phthalates and PAHs
125
69.1%
48.4%
Selected for outdoor phthalates and PAHs
98
Completed outdoor phthalates and PAHs
65
66.3%
46.5%
8-12
-------
and analyzed for a few of these homes, which resulted in overall response rates of 48.3% and
47.7% for indoor and outdoor particulate monitoring, respectively. The overall response rate
for indoor monitoring of phthalates and PAHs was comparable, 48.4%, but for outdoor
monitoring of phthalates and PAHs the rate was slightly lower, 46.5%, for two reasons. First,
the outdoor phthalate and PAH monitors could not be placed at some apartments because
these monitors could not be hung outside a window like the particulate monitors. Second, the
phthalate and PAH monitors were designed to be sensitive to flow fluctuations and thus
unacceptable flow rates resulted in shut-down during electrical brownouts.
To define weighting classes for the monitoring phase of the Pilot PTEAM study, we
considered the following variables:
(1) the geographic strata;
(2) household size;
(3) age of the selected participant; and
(4) education of the head of the household.
The latter two variables were found to be the most promising with respect to creating
categories with distinctly different response rates. As discussed earlier, relatively
homogeneous weighting classes with major differences between classes are most likely to
result in reductions in nonresponse bias.
When we examined the monitoring phase response rates by the age of the selected
participant, as shown in Table 8-4, we concluded that two weighting classes based on age of
participant should be defined as follows:
(1) Age 10 through 25; and
(2) Age 26 or older or age not given.
The monitoring phase response rates were 75.0% for the younger age group and 65.8 percent
for the older age group. This considerable difference in response rates suggests that there is
considerable potential for reducing nonresponse bias by using the age of participant weighting
classes for nonresponse weight adjustments.
Likewise, we examined the distribution of monitoring phase response rates by the
number of years of schooling completed by the head of the household, as shown in Table 8-5.
We concluded that two weighting classes should be defined, as follows:
8-13
-------
TABLE 8-4. MONITORING PHASE RESPONSE RATES BY
AGE OF SELECTED PARTICIPANT
Number of
Age of Participant Number in Sample Respondents Response Rate8
Missing
2
0
0.0%
10-18
47
35
74.5%
19-25
49
37
75.5%
26-44
109
72
66.1%
45-59
22
15
68.2%
60 or Older
28
19
67.9%
Total
257
178
69.3%
8 Both study questionnaire and personal exposure monitoring required for response.
8-14
-------
TABLE 8-5. MONITORING PHASE RESPONSE RATE BY EDUCATION
OF THE HEAD OF HOUSEHOLD
Years of School
Completed by Head of
Household
Number in
Sample
Number of
Respondents
Response Rate"
Missing
26
14
53.8%
11 or Less
51
33
64.7%
12-15
122
88
72.1%
16 or More
58
43
74.1%
Total
257
178
69.3%
8 Both study questionnaire and personal exposure monitoring required for response.
8-15
-------
(1) Less than 12 or unknown years of schooling; and
(2) 12 or more years of schooling.
The education of head-of-household rates were 61.0% and 72.8% for the households headed
by people with lower and higher levels of education, respectively. This large difference in
response rates again suggests considerable potential for reducing nonresponse bias by using
these weighting classes for nonresponse weight adjustments.
Two sets of adjusted person-level analysis weights were defined:
(1) weights for all participants who completed the study questionnaire; and
(2) weights for all participants who completed both the study questionnaire and
personal exposure monitoring.
The education of head weighting classes were used to perform an initial weighting-class
nonresponse adjustment. Within each weighting class, the weights for the respondents were
ratio-adjusted to sum to the same total as all the sample members belonged to the same
weighting class. A final weight adjustment was then implemented using the age-of-participant
weighting classes. Within each weighting class, the weights for the respondents were
ratio-adjusted to sum to the estimated number of population members belonging to the
weighting class based on the entire screening sample.
The monitoring phase sample sizes were not sufficiently large to use both weighting
class variables simultaneously. The adjustment by the age of the selected participant was
implemented at the last stage so that the final adjustment for the person-level weights would
be to estimated person-level totals. The entire screening sample was used as the basis for
the final adjustment because of the larger size of the screening sample. The total estimated
number of population members to which the person-level weights were adjusted at the final
stage was 138,948 nonsmokers aged 10 or older residing in the target portion of Riverside,
California, at the time of the Pilot PTEAM study. The nonresponse adjustment process for
person-level weights is summarized in Figure 8-3.
Three sets of analysis weights were computed for households selected for particulate
monitoring:
(1) weights for households that completed the study questionnaire;
(2) weights for households that completed both the study questionnaire and indoor
particulate monitoring; and
8-16
-------
Sampling weight based on probability
of selection for 257 persons
Screening nonresponse adjustment factor
based on 4 geographic strata
Adjustment to the total
number of people in the
target population for 2
age classes based on all
screening interviews
Adjustment to the total
number of people in the
target population for 2
age classes based on all
screening interviews
Nonresponse adjustment for
191 people who completed the
study questionnaire based on
2 levels of education of
the head of household
Nonresponse adjustment for
178 people who completed
both the study questionnaire
and personal exposure
monitoring based on 2 levels of
education of the head
of household
Figure 8-3. Nonresponse Adjustment Process
for Person-level Weights
8-17
-------
(3) weights for households that completed both the study questionnaire and outdoor
particulate monitoring.
The age-of-partieipant weighting classes were used to perform an initial weighting-class
nonresponse adjustment. Within each weighting class, the weights for the responding
households were ratio-adjusted to sum to the same total as all sample households belonging
to the same weighting class. The number of years of schooling completed by the head of the
household was then used for a final weight adjustment. Within each weighting class, the
weights for the responding households were ratio-adjusted to sum to the estimated total
number of households belonging to the weighting class based on the entire screening sample.
The adjustment by education of the head of the household was performed at the last
stage so that the final adjustment for household-level weights would be to estimated
household-level totals. The total estimated number of households to which the
household-level weights were adjusted was 61,520 households containing at least one
nonsmoker aged 10 or older residing in the target portion of Riverside, California, at the time
of the Pilot PTEAM study. The process of adjusting household-level weights for homes
selected for particulate monitoring is summarized in Figure 8-4.
Two sets of analysis weights were computed for households selected for ohthalate
and PAH monitoring:
(1) weights for households that completed both the study questionnaire and indoor
phthalate and PAH monitoring; and
(2) weights for households that completed both the study questionnaire and outdoor
phthalate and PAH monitoring.
The sizes of the samples selected for phthalate and PAH monitoring were not sufficient to use
either the age-of-participant or education weighting classes; the smallest weighting class
would have contained too few respondents. Therefore, a single, overall ratio adjustment to
the estimated 61,520 households in the target population was used to produce the analysis
weights for phthalate and PAH monitoring. This adjustment reduces nonresponse bias for
estimated population totals, but not for means and proportions. The weight adjustment
process for households selected for phthalate and PAH monitoring is summarized in
Figure 8-5.
8-18
-------
Sampling weight based on probability
of selection for 257 households
Screening response adjustment factor
based on 4 geographic strata
Nonresponse adjustment
for 19jL households that
completed the study
questionnaire based on
2 levels of the age of
the selected participant
Adjustment to the total
number of households in
the target population
for 2 levels of education
of the head of household
based on all screening
interviews
Adjustment to the total
number of households in
the target population
for 2 levels of education
of the head of household
based on all screening
interviews
Adjustment to the total
number of households in
the target population
for 2 levels of education
of the head of household
based on all screening
interviews
Nonresponse adjustment
for 177 households that
completed both the study
questionnaire and indoor
based on 2 levels of the
age of the selected
participant
particulate monitoring
Nonresponse adjustment
for 175 households that
completed both the study
questionnaire and outdoor
based on 2 levels of the
age of the selected
participant
particulate monitoring
Figure 8-4. Nonresponse Adjustment Process for
Household-level Weights for Homes Selected
for Particulate Modeling
8-19
-------
Screening response
adjustment factor based
on 4 geographic strata
Screening nonresponse
adjustment factor based
on 4 geographic strata
Sampling weight based on
probability of selection
for indoor monitoring
for 181 households
Sampling weight based on
probability of selection
for outdoor monitoring
for 98 households
For 125 households that
completed indoor monitoring,
a single, overall adjustment
to the total number of
households in the target
population based on ail
screening interviews
For 65 households that
completed outdoor monitoring,
a single, overall adjustment
to the total number of
households in the target
population based on all
screening interviews
Figure 8-5. Nonresponse Adjustment Process for
Households Selected for Phthalate
and PAH Monitoring
8-20
-------
Memoranda to "the record" were produced that documented the final Pilot PTEAM
weight files and briefly discussed their use. Those memoranda are reproduced as
Appendix J.
As discussed in Section 4, households and persons therein could not both be
selected for monitoring with equal probabilities. Believing that person-level analyses would be
somewhat more important than household-level analyses, the sample selection procedure was
intended to produce analysis weights with less variability, and, hence, less loss of precision
due to unequal weighting, for person-level analyses than for household-level analyses.
Therefore, the survey design effect attributable to unequal weighting was computed
for each set of final analysis weights as follows'.
DEFFwt = n E WT2 / (E WT)2 (Ecl- 8"1)
where the summation, E, is over all units in the sample, n is the number of responding units,
and WT is the final analysis weight. This design effect due to unequal weighting is 1.00 if all
the sampling weights are equal and is a variance inflation factor, exceeding 1.00, otherwise.
For the two sets of person-level analysis weights, the unequal weighting design effect was
approximately 1.16. Each set of household-level analysis weights had a larger unequal
weighting design effect. The unequal weighting design effect for households with indoor
monitoring of particulates was 1.22. Therefore, the sample selection strategy achieved the
desired goal of producing comparable levels of unequal weighting effects with slightly less loss
of precision for person-level analyses.
RESPONSE RATE IMPROVEMENT
The household screening phase of the Pilot PTEAM study achieved a response rate
of 70.1%, and the personal exposure monitoring phase of the study achieved a response rate
of 69.3% . Therefore, the overall study response rate was 48.5%. This low response rate
allows considerable potential for nonresponse bias. However, it is consistent with the results
of comparable exposure monitoring studies, as shown in Table 8-6.
Several special activities were undertaken for the Pilot PTEAM study in an attempt to
improve the response rate relative to previous exposure monitoring studies. An advance
notification letter and glossy brochure describing the study were mailed to sample residences
8-21
-------
with mailable addresses prior to contacting them for the household screening interview. The
letter was signed by the Assistant Administrator of EPA's Office of Research and
Development, the Chief of the Research Division of the California Air Resources Board, and
the Mayor of the City of Riverside. Pre-survey publicity resulted in a front-page article in the
Riverside Press-Enterprise newspaper on Thursday, September 6, immediately prior to the
beginning of field data collection. The interviewers carried all these materials (advance letter,
brochure, and newspaper article) for explaining the study and recruiting participants. If the
sample households did not remember receiving the advance mailing materials, these materials
were provided at the door. The purpose and scope of the study were explained to all
participants before beginning the household screening interviews.
Participants were selected in the field during the household screening interviews. We
expected that this procedure would improve overall study response rates because it would
eliminate the need for a second call to most sample households to recruit the selected
participants.
Participants were selected in the field during the household screening interviews for
three of the exposure monitoring studies listed in Table 8-6: (1) the Baltimore, Maryland
TEAM study; (2) the Woodland, California study of exposures to organic compounds; and (3)
the Pilot PTEAM Study.
Table 8-6 shows that the Baltimore study achieved considerably higher response
rates both at the screening phase and overall. Table 8-7 presents a more detailed analysis of
the screening phase response rate for these three studies and the earlier large-scale TEAM
studies. For each study listed in Table 8-7, the screening interview was essentially the
rostering of household members, although the amount of data collected regarding each
household member varied somewhat. Understanding the differences in study protocol and
how they relate to differences in response rates may help us design future exposure
monitoring studies that achieve higher response rates by using the most successful strategies
from the earlier studies.
The Baltimore TEAM study contacted sample households with no advance notification
or pre-survey publicity, so-called "cold contacts." In addition, the interviewers for the Baltimore
TEAM study were instructed to begin the household screening interview immediately following
a minimal introduction. Moreover, the household screening interview was just a simple roster
8-22
-------
TABLE 8-6. RESPONSE RATES ACHIEVED IN COMPARABLE EXPOSURE MONITORING STUDIES
Description of Study
Location
Season
Screening
Response
Rate
Number
Monitored
Monitoring
Response
Rate
Overall
Response
Rate
Total Exposure Assessment for VOCsa
Bayonne, NJ
Fall '81
87%
154
55%
48%
Total Exposure Assessment for VOCsa
Elizabeth, NJ
Fall '81
84%
201
51%
43%
Total Exposure Assessment for VOCsa
Greensboro, NC
Spring '82
95%
24
80%
76%
Total Exposure Assessment for VOCsa
Devils Lake, ND
Fall "82
96%
24
67%
64%
Total Exposure Assessment for VOCsa
Los Angeles, CA
Winter '84
87%
117
64%
56%
Total Exposure Assessment for VOCsa
Antioch/Pittsburg, CA
Summer '84
89%
71
64%
57%
Total Exposure Assessment for VOCsa
Baltimore, MD
Spring '87
95%
155
62%
59%
CO Exposure Monitoring Study'3
Washington, DC
Winter '83
70%
1,161
58%
41%
CO Exposure Monitoring Study''
Denver, CO
Winter '83
76%
485
43%
33%
Nonoccupational Pesticide Exposure Study0
Jacksonville, FL
Summer '86
74%
65
54%
40%
Nonoccupational Pesticide Exposure Study0
Jacksonville, FL
Spring *87
66%
53
73%
48%
Nonoccupational Pesticide Exposure Study0
Jacksonville, FL
Winter '88
81%
55
61%
49%
Nonoccupational Pesticide Exposure Study0
Springfield/Chicopee, MA
Spring '87
70%
49
55%
39%
Nonoccupational Pesticide Exposure Study0
Springfield/Chicopee, MA
Winter '88
84%
37
51%
43%
VOC, WOC, and SVOC Exposure Studyd
Woodland, CA
Spring '90
69%
128
74%
51%
Pilot PTEAM Study
Riverside, CA
Fall "90
70%
178
69%
49%
Legend: CO * Carbon monoxide; SVOC = Semivolatile organic compound; VOC = Volatile organic compound; WOC = Very volatile organic compound.
a Wallace (1987).
Whitmore et al. (1984).
0 Immerman and Schaum (1990).
Final report in preparation.
-------
of the household members for the Baltimore study. Thus, the interview was short (usually 5
minutes or less) and was begun as quickly as possible after contacting a potential respondent.
The Woodland study also used "cold contact" screening interviews, but the
interviewers were Instructed to explain the purpose and scope of the study before beginning
the household screening interview. This longer introduction may have resulted in refusals at
the screening stage from people who had decided, based on the introduction, that they did not
want to participate in the monitoring phase. The screening interview itself was slightly longer
than that used for the Baltimore study, which may also have contributed to lower screening
phase response rates.
Of course, the PTEAM study did not use "cold contact" interviews. Sample
households were sent advance packets describing the study in detail, the study was
discussed in a headline article in the local newspaper, and the interviewers were trained to
fully explain the study to households that did not recall receiving the advance mailing and to
present the newspaper article to all potential respondents. Again, this procedure probably
resulted in refusals at the screening stage from people who had simply decided that they did
not want to participate in the monitoring phase of the study. The length of the PTEAM
screening interview was comparable to, but slightly shorter than, that used for the Woodland
study. Again, this could partially account for lower screening-phase response rates relative to
the Baltimore study.
All the earlier TEAM studies listed in Tables 8-6 and 8-7 used "cold contact"
screening interviews that were conducted several weeks before participants were recruited for
second phase monitoring appointments. The interviewers provided minimal introduction to the
study prior to the screening interviews for these earlier studies because lengthy explanations
were not necessary for conducting the short screening interviews.
Whenever participants are recruited in the field, as in the Baltimore, Woodland, and
PTEAM studies, there is a danger that the urgent need to schedule monitoring appointments
may distract the interviewer from diligent completion of screening interviews for households
that are difficult to contact. Having a sufficiently large field staff to keep appointments
scheduled well in advance may be critical for achieving the best possible response rates for
these studies. Table 8-7 shows that we were not able to contact an eligible screening
respondent for about 12 percent of the eligible sample households for the PTEAM study. We
believe that a major reason for this outcome is the fact that many Riverside residents make a
8-24
-------
TABLE 8-7. ANALYSIS OF SCREENING PHASE RESPONSE RATES FOR SELECTED EXPOSURE MONITORING STUDIES
Description of Study
Location
Date
Screening
Response
Rate
Number
Eligible
Sample
Homes
Percent
Eligible Not
at Home
Percent Refusals
Among Eligibles
at Home
Total Exposure Assessment for VOCsa
Bayonne, NJ
Fall '81
87%
2,063
6.5%
4.5%
Total Exposure Assessment for VOCsa
Elizabeth, NJ
Fall '81
84%
3,145
6.6%
5.9%
Total Exposure Assessment for VOCsa
Los Angeles, CA
Winter '84
87%
1,219
3.6%
8.5%
Total Exposure Assessment for VOCsa
Antioch/Pittsburg, CA
Summer '84
89%
561
3.0%
7.5%
Total Exposure Assessment for VOCsa
Baltimore, MD
Spring "87
95%
577
1.4%
3.2%
VOC, WOC, and SVOC Exposure Studyb
Woodland, CA
Spring '90
69%
285
7.0%
24.9%
Riot PTEAM Study
Riverside, CA
Fall '90
70%
632
11.7%
13.6%
Legend: CO = carbon monoxide; SVOC = Semivolatile organic compound; VOC = Volatile organic compound; WOC = Very Volatile organic compound.
® Wallace (1987).
b Final report in preparation.
-------
long commute to Los Angeles for work and therefore, are hard to find at home. However, the
urgency of scheduling monitoring appointments is also considered to be a contributing factor
for the low screening phase response rates for PTE AM.
Table 8-7 also shows that refusals to the household screening interview were major
contributors to the low screening-phase response rates for the Woodland and PTEAM studies.
Since these were the only two studies under consideration that provided full explanations of
the study to the potential respondents prior to the screening interview, we expect that the
presentation of more information than was essential for completion of the screening interviews
contributed to the lower screening-phase response rates for these two studies.
Low screening-phase response rates might not be a major problem if (1) that
nonresponse did not result an any systematic bias, (2) no important population data were
collected in the screening interviews, and (3) the overall-study response rates were not
adversely affected. However, none of these conditions are likely to be met in practice.
Nonrespondents are almost always different from respondents in some systematic way. As
discussed later, the data collected from screening-phase respondents who refuse to
participate in the monitoring phase can be very important for reducing nonresponse bias.
And, Table 8-6 shows that the studies that achieve the lowest screening-phase response rates
generally also achieve the lowest overall-study response rates.
Therefore, we recommend that the interviewers immediately begin the short
household screening interviews for future exposure monitoring studies without providing
details about the monitoring phase of the study. Likewise, we recommend against using
advance notification mailings prior to the screening interviews. Pre-survey publicity is
recommended, but it should not be mentioned by the interviewer until the household screening
interview has been completed and someone has been selected for monitoring. When the
screening and monitoring phases occur several weeks apart, advance notification mailings to
the people selected for the monitoring phase of the study are recommended.
After someone has been selected for monitoring, the interviewers should have access
to materials that present the objectives of the study, explain what participation involves, and
discuss incentives for participation. The materials should be as clear, succinct, and
professional in appearance as possible to have the maximum positive impact. Such materials
may include the following:
8-26
-------
• a letter of endorsement from recognized public officials;
• a glossy study brochure;
• photographs of the monitoring equipment in use, replicas of the equipment, or the
actual equipment itself; and
• a video-taped recruitment script, preferably featuring a recognized public official.
A professionally produced video tape can be quite effective for recruitment, but it is also rather
expensive to produce. Also, each interviewer needs to have access to a playback unit or
some other alternative for households that do not have their own video tape player. RTI used
video tapes quite successfully for participant recruitment for studies of the prevalence of HIV
infection (RTI, 1990).
In addition to recommending "cold-contact" household screening interviews and use of
professional looking recruitment aids for monitoring phase recruitment, we strongly
recommend that the survey work be scheduled so that all sample housing units can be
worked to completion. One implication is that we must hire a sufficient number of interviewers
to schedule most monitoring appointments a week or two in advance. When the first week of
monitoring begins, the process of establishing the first week's appointments should be
completed. During the fast week or two of monitoring, the interviewers should be working
primarily on contacts with households that have been difficult to contact or have previously
refused.
Enhancing participant incentives may be another way to increase response rates.
Each participant in the PTEAM study was paid $100, given a copy of the EPA publication
"The Inside Story," and told that they would receive their monitoring results. An more
attractive alternative might be a lottery. Each participant would then have a small chance of
winning a much larger sum of money. The lottery could be combined with a small cash
incentive (e.g., $20) for each participant.
COMPENSATING FOR NONRESPONSE BIAS
Future exposure monitoring studies must continue to try innovative survey procedures
that may produce higher response rates. A high response rate is the only guarantee that
nonresponse bias will be small. However, some additional gains may be possible using
analysis procedures with greater potential for reducing nonresponse bias. The PTEAM and
earlier exposure monitoring studies conducted by RTI (Table 8-6) have all used weighting
8-27
-------
class nonresponse weight adjustment procedures to partially compensate for the potential bias
due to nonresponse (Madow, et at., 1983).
An alternative procedure that has the potential for achieving greater reductions in
nonresponse bias is to weight each observation Inversely to the predicted probability of
obtaining a response (lannacehione et al., 1991). This methodology requires that sufficient
data be collected regarding both respondents and nonrespondents to fit a model that
produces reasonably accurate predictions regarding which sample members will respond. In
order to apply this methodology, data that will be useful for predicting the probability of
participating in the monitoring phase must be collected for both respondents and
nonrespondents during the household screening interviews.
Data items that are potentially useful for predicting the probability of participating in the
monitoring phase include the following:
• age of the selected respondent
• sex of the selected respondent
• education of the selected respondent
• type of work (occupation)
• place of employment (industry)
• attitude toward surveys
• concern for environmental issues
• attitude toward the EPA or other sponsoring agencies
• general degree of skepticism
• education of the head of the household
• household income
• type of dwelling (apartment, condominium, detached residence, etc.)
• location of residence (urban, suburban, rural)
• dwelling tenure (owned or rented).
Many other variables could be added to this list. In practice, however, we would choose a
relatively small number of items for at least two reasons. First, a large number of items
increases the respondent burden and may reduce the screening phase response rate, thus
compromising the goal of reducing overall nonresponse bias. Second, the number of
independent variables that can be included in the response model is directly proportional to
the sample size.
8-28
-------
A large sample is more likely to result in a useful model for the probability of
responding and has the potential for including a larger number of independent variables.
Therefore, we recommend this nonresponse compensation procedure primarily for large-scale
exposure monitoring studies.
TEMPORAL RANDOMIZATION
An experiment to randomly assign sample subjects to days for interviewing was
included in the Pilot PTEAM study. On the day before environmental data collection was to
begin, we quit randomly assigning people to days and assigned each sample member to the
first available monitoring appointment because the monitoring appointments were not being
filled sufficiently. For the entire study, we recorded whether the subject participated on the
chosen day (randomly selected or first available day) or on another day. The interviewers
recorded that 84% (150/178) of the sample members participated on their selected day. This
high rate of willingness to be monitored on the selected day suggests that future attempts to
select a random day for participation may be successful.
The days for monitoring should be randomly selected because we have found that the
exposure measurements can vary greatly from one day to the next (Clayton et al., in press).
In this situation, each exposure measurement is representative of a person-day and cannot be
extrapolated beyond that. Therefore, the units of observation are person-days, and random
selection of monitoring days is of comparable importance to random selection of people to be
monitored (Gilbert, 1987).
The plan originally being tested in the Pilot PTEAM study was to randomly select a day
from those currently available to the interviewer at any point in time (usually about a 2-week
period), and to allow a participant who could not participate on that day to choose any of the
available days to keep the response rate as high as possible.
From the standpoint of reducing nonresponse bias, the following procedure may be
preferable. First, randomly select a target week for participation for each sample household.
Begin attempting the screening interviews sufficiently far in advance that most interviews can
be scheduled Airing the target week. Allow a substitute week to be selected when the
sample subject cannot be contacted prior to the target week. Then, within the target week,
randomly assign each sample member to either a work day or a non-work day in proportion to
the number of days that the person expects to be working during the designated week. The
8-29
-------
person could then select any day of the designated type (work or non-work) from the sample
week for participation. Monitor the person on the selected day whether or not the person
actually went to work on that day.
A person who was selected to be monitored on a work day and refused would not be
allowed to substitute a non-work day. Although we might lose a few people to nonresponse
who would otherwise have participated, we would have a greater potential for reducing
nonresponse bias. The type of day (work or non-work) for which a person had been selected
might be a useful correlate with participation that could be used for defining nonresponse
weight adjustment classes or used as a predictor variable in a model for predicting the
probability of responding.
Procedures for randomly selecting a day for participation necessarily add logistic
difficulties and costs to the monitoring program. For example, if sample households are
randomly assigned to weeks for data collection prior to the screening interviews, then the
number of appointments scheduled in any given week will be a random variable.
Consequently, the environmental sampling protocol will have to allow for varying numbers of
appointments per day. In practice, the need for random selection of monitoring days must be
balanced against logistical considerations and cost efficiency of data collection.
8-30
-------
SECTION 9
STATISTICAL ANALYSIS OF PARTICLE DATA
INTRODUCTION
The various types of data available for statistical analysis can be broadly categorized
as follows:
1. Particle data
a. Blank filter
b. Temporal-site
c. Residence
2. Meteorological data
3. Questionnaire and activity diary data
This chapter provides statistical summaries and analysis results that characterize these types
of data and some of the associations among them. The emphasis of the chapter is on the
particle data, especially types 1b and 1c. The temporal site data furnish extensive
methodological information regarding particle measurement technologies. The residence data
support the analyses relating to the primary study objectives - characterizing the personal,
indoor, and outdoor particle concentration distributions for the target populations of
households and individuals in Riverside in the fall of 1990.
The first part of this report section describes the types of data collected and the
information available for evaluating data quality. The extent of the particle sample data
collected and those samples deemed suitable for analysis are described first to provide an
overview of the data set. Quality control measurements included the use of blank filters to
measure background contamination of the filters during handling and collocated samples to
measure sampling and analysis precision. Multiple samplers, including dichotomous,
Wedding, PEM and SAM were used side-by-side at one location throughout the course of the
study. This location was called the temporal site, and comparisons of data collected at this
location were carried out to help evaluate the performance of the PEM, SIM and SAM
monitors as compared to reference ambient air monitoring methods. Meteorological data were
gathered from three nearby weather stations and included hourly measurements of
temperature, dew point, wind speed, and wind direction. Associations between the
meteorological data and ambient air particle concentrations were investigated.
9-1
-------
Results of the statistical analysis of the personal, indoor, and outdoor particulate
concentrations are described in the second part of this section. A primary objective of this
study was to evaluate 24-hour personal exposure concentration distributions with regard to the
150 ng/m3 National Ambient Air Quality Standard (NAAQS) and the 50 ng/m3 California
Ambient Air Quality Standard (CAAQS). These data are presented as well as the distributions
over 12-hour daytime and overnight time periods. Associations between the residence
(personal, indoor, outdoor) data and temporal-site, meteorological, questionnaire, and diary
data are presented and discussed.
DATA AVAILABILITY AND QUALITY CONTROL
Availability and Quality of Filter Data
The availability and quality of the aerosol data are summarized in Table 9-1. Rows in
the table correspond to the various sample types for which aerosol data were acquired.
Within each row, counts are given that indicate the number of filters processed during the
study, the number usable and unusable, and the percentage usable. The usable and
unusable categories are each further broken down into subcategories that identify the status
of the samples. Codes 01 through 04, as defined at the bottom of Table 9-1, were assigned
during field operations, while codes 05 through 08 were assigned after completion of the data
collection activities. The latter included: (a) reassignment of some of the original 03 codes
(to codes 05 or 06), based upon review of the data and associated comments recorded by the
field staff in the database, and (b) reassignment of (01 or 03) codes based upon statistical
outlier considerations. Codes 07 and 08 indicate potential outliers. AH statistical analysis
results and data summaries described in this chapter consider samples with data Quality
codes fdataflao) equal to 01. 03. 06. and 08 to be usable. For each of the sample type
classifications identified in Table 9-1, over 90% of the samples were judged usable. For the
samples collected at the residences, about 94% of the samples were deemed usable.
Blanks
Blank filters were used during this study to evaluate potential problems of filter
contamination during handling. Two types of blanks were used; field blanks and special
blanks. Field blank filters were pre-weighed and assembled into impactors with greased
impactor plates, just as was done for the sample filters. These impactor assemblies were
9-2
-------
TABLE 9-1. SUMMARY OF AVAILABILITY AND QUALITY OF 12-HR AEROSOL SAMPLES3
Number of Samples, Classified by Dataflag
Usable Samples
Unusable Samples
Percent
Sample Type
01
03
06
08
Total
02
04
05
07
Total
Total
Usable
a.
342
7
3
2
354
12
6
2
0
20
374
94.7
SAM
331
11
3
0
345
11
5
3
2
21
366
94.3
SAM (temporal site)
83
5
2
2
92
2
2
3
0
7
99
92.9
PMi?M
337
5
3
4
349
13
9
1
0
23
372
93.8
SAM
332
7
3
3
345
8
6
8
1
23
368
93.8
PEM
320
14
5
0
339
4
5
8
0
17
356
95.2
SAM (temporal site)
85
7
2
0
94
3
2
1
0
6
100
94.0
PEM (temporal site)
88
2
0
0
90
7
1
1
0
9
99
90.9
PEM (project staff)
30
2
0
0
32
2
0
0
0
2
34
94.1
Wedding^
171
6
0
1
178
12
0
2
0
14
192
92.7
Dichot0
Fine
181
6
0
1
188
2
0
2
0
4
192
97.9
Coarse
177
6
0
2
185
3
0
3
1
7
192
96.4
Blanks
Field
Special
(ungreased
impaclors)
50
9
1
0
0
0
0
0
51
9
1
0
0
0
0
0
0
0
1
0
52
9
98.1
100.0
a Samples are not exactly of 12-hr duration, but are referred to as 12-hr samples for convenience.
b RTI data analysis uses data with flags 01, 03, 06 and 08. Definitions of data flag codes are as follows:
01 = no problem identified 05 = originally 03, data deemed unusable
02 = severe problem identified in field, unusable data 06 = originally 03, data deemed unusable
03 = some problem identified in field, questionable data 07 = outlier, data deemed unusable
04 = ID problem, unusable data 08 = outlier, data deemed usable
c All wedding and dichot samples were taken at the temporal site.
-------
carried from the workroom to the field monitoring sites along with the samples. The impactors
were then returned to the workroom and the filters were removed and reweighed along with
sample filters. During the course of the field monitoring it was observed that increases in field
blank filter masses were typically in the 0 to 15 pig range, up to a maximum of 34 jig. One
potential contaminant was the grease applied to impactor plates. Nine special blanks were
prepared that were identical to the field blanks except that no grease was applied to the
impactor plates.
Gravimetric results for the blank filters (field blanks and special blanks representing
ungreased impactor plates) are given in Table 9-2. Blank filters, used for PEM, SIM, and
SAM measurements, showed small increases in mass with a median of 9 ng. This median
value was used to adjust the particle catches of all PEM, SIM, and SAM sample filters.
Special blanks had similar increases with a median of 5 ng.
The lower value for the special blanks suggests that grease contamination may have
been responsible for some, but not all, of the filter contamination. Other potential contaminant
sources include airborne dust, dander from persons handling the filters, and small metal
particles from the impactor assemblies. Over 90% of the PM2 5 samples and 98% of the
PM-jq samples had particle catches greater than 40 jig, so the impact of the filter
contamination on the results of this study are small. However, the level of background
contamination may be important if similar monitoring is conducted with small air volumes
q
(approximately 3 m ) in locations with lower particle concentrations than were observed in
Riverside. In general, filter contamination was very small, especially considering that a true
clean-room environment could not be maintained for filter handling operations during the field
study.
Blank dichot and Wedding filters showed a median increase in mass of 4 fig and 100
pig, respectively. These median increases were all judged to be small, in comparison both to
the tare filter weight (about 100 mg and 4 g small and large filters, respectively) and to the
observed net magnitudes and ranges of aerosol masses found for exposed filters. Dichot and
Wedding sample aerosol catches were not adjusted for the observed background.
Collocated Samples
In order to characterize the precision of the aerosol measurements, collocated samples
were collected periodically for some sample types and routinely for other sample types.
9-4
-------
TABLE 9-2. SUMMARY OF DISTRIBUTIONS OF NET WEIGHTS (jig) FOR BLANK FILTERS
Field Blanks
PEM, SIM,
SAM
Wedding
Dichot
Special
Blanks8
No. of Blank Filters
51
44
41
9
Distribution of Blank Filter Catches (jjtg):
Minimum
-29
-1000
0
1
Maximum
34
2400
15
19
Mean
9.5
170
4.6
6.8
Std. Deviation
9.0
534
4.0
5.2
Percentiles
10th
2
-300
0
-
25th
5
-200
1
-
50th (median)
9
100
4
5
75th
15
400
7
-
90th
19
800
10
-
Approximate Filter Weight (milligrams):
100
4400
100
100
Approximate Range of Catches for
Exposed Filters (jig):
Minimum
30
8000
40
-
Maximum
1500
172000
2000
-
8 Blank filters associated with ungreased impactor plates.
9-5
-------
Relative standard deviations (RSDs) for the particulate concentrations were calculated for
each pair.
RSD = 100% x (standard deviation of paired particle concentrations) /
{mean of paired particle concentrations) (Eq. 9-1)
= 100% x 1.414 x |Xj - X2\/ (Xi + X2),
where and denote the two observed concentrations. The distributions of these RSDs
are characterized by the results presented in Table 9-3, which shows the number of pairs for
which RSDs were calculated, along with the mean, median, maximum, and standard deviation
of the RSDs. The results are given by particle size (2.5 or 10 jim) and sample type. The
results imply excellent agreement (high precision) between the collocated samples, with very
few exceptions. For instance, the median relative standard deviations varied from about 2%
to 5%, depending on the sample type.
As indicated in Table 9-3, duplicate SIM and SAM samples were obtained for a subset
of the participants' homes; however, due to the awkwardness of wearing two PEM monitors,
the study participants were not asked to provide collocated PEM samples. Rather, such
collocated data were periodically acquired by having field project staff wear two PEM
samplers, one mounted above the other. The resultant PEM concentration data are listed in
Table 9-4. The correlation between the paired concentrations was high (over 0.96). In 12 of
the 14 cases, the lower-mounted PEM exhibited higher PM10 concentration levels than the
top-mounted PEM, though the differences were usually not large and the overall mean
difference was not statistically significant at the 0.05 level (t-test and signed-rank test). One of
the two exceptions ~ and the one displaying the largest percentage discrepancy (29.6%)
between the dual-mounted samplers -- occurred for period 92 (ID 195), which was
characterized by a severe wind storm. The staff member wearing these PEMs spent a large
portion of this period driving a car with the window down; hence the the top sampler, on a
level with the open window, was essentially unprotected from the outside, while the bottom
sampler was protected by the car door. The lower portion of the table gives the mean and
median of the differences and of the percentage differences - for all 14 pairs of samples, and
for the first 13 pairs. Overall, the lower-mounted PEM concentrations tended to exceed those
of the other PEM by about 4%. However, tests of the 13 paired again indicated no statistically
significant difference in the means.
9-6
-------
TABLE 9-3. SUMMARY OF DISTRIBUTIONS OF RELATIVE STANDARD DEVIATIONS (%)
CALCULATED FROM COLLOCATED AEROSOL CONCENTRATION DATA
Percentiles
Particle
Size
Sample
Type
No.
of
Pairs
Mean
RSD
10
25
50
75
90
Maximum
Std.
Dev.
2.5
SIM
15
7.4
_c
0.7
3.5
11.4
-
32.7
9.0
SAM
17
7.8
-
3.3
5.3
8.0
-
36.3
9.3
SAM"
3
2.4
-
-
2.0
-
-
3.7
1.2
Dichot®
93
5.6
0.5
1.2
3.2
5.0
8.0
78.1
10.7
10 nm
SIM
17
7.4
-
0.8
3.3
7.1
-
37.7
10.8
SAM
18
2.4
-
1.8
2.5
2.9
-
6.5
1.7
SAM8
4
3.8
-
-
3.7
-
-
5.4
1.5
PEMa
3
3.8
-
-
4.4
-
-
5.3
1.8
PEMb
15
6.0
-
2.1
4.6
8.8
-
20.9
5.5
Dichot0
90
5.5
1.0
2.2
4.5
6.6
9.6
53.0
6.5
Wedding3
87
2.7
0.3
0.8
1.7
4.0
6.4
14.9
2.7
a Located at temporal site.
b Worn by field project staff.
c Not estimated due to small sample size.
-------
TABLE 9-4. COLLOCATED PEM AEROSOL DATA FROM SAMPLERS WORN BY PROJECT STAFF
Person ID
Period
PEM Cone, (fig/m3)
Top Bottom
Difference
(Top-Bottom)
Percentage
Difference
199
17
53.5
60.2
-6.7
-11.8
200
19
57.8
60.1
-2.4
-4.0
198
37
32.9
35.6
-2.8
-8.1
198
38
84.8
87.1
-2.4
-2.8
198
39
52.1
48.8
3.3
6.5
198
40
99.7
103.3
-3.6
-3.6
198
41
73.1
77.6
-4.5
-6.0
199
74
106.9
94.3
12.5
12.4
196
82
145.5
166.1
-20.6
-13.2
196
83
35.6
43.1
-7.4
-18.9
195
89
21.7
22.4
-0.7
-3.0
195
90
162.5
163.6
-1.0
-0.6
195
91
22.5
24.1
-1.6
-7.1
195
92
259.6
192.7
66.9
29.6
Median
-2.4
-3.8
Mean
2.1
-2.2
Excluding last observation8:
Median
-2.4
-4.0
Mean
-2.9
-4.6
Note: Correlation of top and bottom PM10 concentrations is 0.963.
3 Data for the last observation occurred under unusual circumstances, as described in the text.
9-8
-------
TEMPORAL-SITE AEROSOL CONCENTRATIONS
Throughout the study, ambient or aerosol concentration data were generated from filter
samples at the temporal site during the same period of time when data was also obtained at
the residences, using SAMs (PMg 5 and PM^), PEMs (PM^q), a pair of Dichot samplers
(PM2 5 and PM10), and a pair of Wedding samplers (PM^q). Time periods for these samples
coincided approximately with the data collected at the homes (i.e., two time periods per day
beginning and ending in the morning and the evening). These concentration data are listed in
Appendix K. For purposes of statistical analysis, whenever data from a pair of samplers of
the same type were available, the concentrations were averaged. This occurred routinely for
the dichot and Wedding samplers, and occasionally for the other types.
Detailed results summarizing the distributions of the temporal-site concentrations are
presented in Table 9-5 (daytime and nighttime samples), Table 9-6 (daytime samples), and
Table 9-7 (nighttime samples). Essential features of the distributions are discernible from the
boxplots given in Figures 9-1 and 9-2, which show results for the daytime and nighttime
samples, respectively.
Several conclusions can be drawn from the results of these tables and figures. PM2 5
concentration levels as measured by the SAMs tended to be higher than the corresponding
dichot levels (by about 30%). Very little difference between the daytime and nighttime PM2 5
distributions was evident. For the PM-jq samples, daytime concentration levels tended to be
higher than nighttime levels. On average, the SAM levels were highest, followed by the PEM
levels, the Dichot levels, and the Wedding levels. For example, medians for the various
methods, expressed as percentage differences relative to the dichot median, were as follows:
Median PM^q Concentration Percent Difference Relative
(jig/m^) to Dichot
Time of Day
PEM
SAM
Wedding
Dichot
PEM
SAM
Wedding
Daytime
75.8
75.8
66.7
70.3
+8%
+8%
-5%
Nighttime
55.8
60.6
43.0
52.0
+7%
+17%
-17%
Associations between the various temporal-site samplers were examined using
Pearson correlations (see Table 9-8), both overall and by time of day. The correlations were
calculated for both the concentrations and the logarithms of the concentrations. The former
are shown in the upper right triangle (above the diagonal) while the latter are given in the
9-9
-------
TABLE 9-5. DISTRIBUTIONS OF ALL 12-HR® TEMPORAL
SITE AEROSOL CONCENTRATIONS (^ig/m3)
PMp.s PMi0
SAM
Dichot3
PEM
SAM
Dichot3
We
-------
TABLE 9-6. DISTRIBUTIONS OF DAYTIME TEMPORAL SITE
AEROSOL CONCENTRATIONS (ng/m3)
PM25 PMio
SAM Dichot8 PEM SAM Dichot* Wedding3
Sample Size
45
47
42
45
47
45
Minimum
1.4
3.6
19.3
18.2
14.7
15.7
Maximum
151.4
118.5
188.7
2212
226.6
167.5
Mean
46.7
35.3
88.9
91.0
78.3
76.5
(Std. Error)
(5.3)
(4.0)
(6.7)
(7.1)
(6.0)
(5.8)
Geometric Mean
34.1
26.0
78.3
79.1
68.5
67.0
(Std. Error)
(4.5)
(3.2)
(6.4)
(6.5)
(5.4)
(5.4)
Std. Deviation
35.4
272
43.6
48.0
41.3
38.6
Geometric Std.
2.40
2.28
1.69
1.73
1.70
1.71
Deviation6
Percentiles
10th
14.1
8.9
36.5
37.4
32.9
33.3
25th
20.9
14.2
61.2
57.8
48.6
44.0
50th (median)
33.1
25.3
75.8
75.8
70.3
66.7
75th
72.1
50.1
120.2
124.8
113.2
96.0
90th
106.0
76.2
148.8
157.9
126.9
139.6
a Average of two samplers (if both available).
b In contrast to the other statistics, the gsd is a unitless quantity.
9-11
-------
TABLE 9-7. DISTRIBUTIONS OF NIGHTTIME TEMPORAL SITE
AEROSOL CONCENTRATIONS (^g/m3)
PM2.
5
PM10
SAM
Dichot8
PEM
SAM
Dichot3
Wedding3
Sample Size
44
48
45
45
48
46
Minimum
6.5
4.4
16.0
16.7
12.6
15.7
Maximum
177.4
139.3
196.7
211.0
177.2
198.2
Mean
47.7
37.4
69.9
76.7
64.0
58.8
(Std. Error)
(6.6)
(4.8)
(6.5)
(7.1)
(5.5)
(6.7)
Geometric Mean
32.9
26.1
59.1
65.1
54.7
47.1
(Std. Error)
(4.4)
(3.3)
(5.1)
(5.6)
(4.5)
(4.5)
Std. Deviation
43.5
33.5
43.8
47.9
38.4
45.2
Geometric Std.
2.41
2.41
1.78
1.76
1.76
1.89
Deviation15
Percentiles
10th
9.6
7.2
27.0
29.8
28.2
22.6
25th
20.7
15.4
42.2
48.9
39.1
29.5
50th (median)
33.2
25.7
55.8
60.6
52.0
43.0
75th
58.4
41.4
83.6
91.5
75.1
61.0
90th
113.9
86.5
151.6
155.9
133.1
135.4
a Average of two samplers (if both available).
b In contrast to the other statistics, the gsd is a unitless quantity.
9-12
-------
I*g/m3
160 ¦
150
140
130 -|
120
110
100
90
80
70
60
50
40
30
20
10
0
###
90th
75th
Mean
50th
25th
10th
###
SAM
2.5ft
DICHOT
2.5 ft,
PEM
10 p
SAM
10(1
DICHOT
10/i
WEDDING
10(i
Figure 9-1. Distributions of Daytime Temporal Site PM10
and PM25 Concentrations (p.g/m3).
9-13
-------
—
—
90th
—
###
###
###
75th
###
—
###
Mean
50th
###
—
25th
—
10th
SAM DICHOT PEM SAM DICHOT WEDDING
2.5ji 2.5ji 10ft 10ft 10/t 10jt
Figure 9-2. Distributions of Nighttime Temporal Site
PM10 and PMg6 Concentrations (ng/m3).
9-14
-------
TABLE 9-8. PEARSON CORRELATIONS OF TEMPORAL SITE AEROSOL
CONCENTRATIONS AND LOG (CONCENTRATIONS)3
OVERALL CORRELATIONS (Sample sizes range from 89 to 95)
PM?s
SAM Dichotb
PEM
PM
10
SAM Dichotb Weddingb
SAM
-
0.97
0.93
0.93
0.88
0.94
Dichotb
0.93
-
0.87
0.85
0.81
0.89
PEM
0.90
0.83
.
0.99
0.98
0.97
SAM
0.91
0.83
0.99
-
0.98
0.96
Dichotb
0.89
0.82
0.98
0.98
-
0.95
Weddingb
0.89
0.84
0.96
0.96
0.96
-
DAYTIME CORRELATIONS (Sample sizes range from 42 to 47)
PM.
'2.5
PM
11L
SAM
Dichot"
PEM
SAM
Dichotb
Weddingb
SAM
0.96
0.91
0.89
0.82
0.94
Dichot5
0.91
-
0.82
0.76
0.68
0.84
PEM
0.89
0.79
-
0.98
0.96
0.97
SAM
0.89
0.75
0.99
-
0.98
0.95
Dichotb
0.83
0.74
0.98
0.98
-
0.92
Wedding6
0.93
0.84
0.97
0.97
0.96
-
PM
'2.5
PM
10
NIGHTTIME CORRELATIONS (Sample sizes range from 44 to 48)
PMoc PM,
SAM
Dichotb
PEM
SAM
Dichotb
Weddingb
PM2.5
SAM
_
0.99
0.98
0.99
0.98
0.98
Dichotb
0.96
-
0.96
0.97
0.97
0.98
PM10
PEM
0.93
0.91
.
0.99+
0.99
0.97
SAM
0.94
0.92
0.99
-
0.99
0.98
Dichotb
0.92
0.93
0.99
0.92
-
0.98
Weddingb
0.92
0.91
0.96
0.96
0.97
a Correlations shown above the diagonals are for concentration-scale data; those below the
diagonals are for logarithms of concentrations.
b Average of two samplers.
9-15
-------
lower left portion. Correlations of PM2 5 and PM-jq measurements were high at night (0.96 or
above) for both the Dichot and SAM samplers. During the day, the SAM PM2 5 correlations
with the PM.|o measurements ranged from 0.82 to 0.94, while the corresponding Dichot
correlations of PM 2 5 with PM^ ranged from 0.68 to 0.84. Among methods measuring
PM^q, the nighttime correlations tended to be higher than the daytime. The highest
correlations occurred between the PEM and the SAM measurements (0.98 to 0.99+), while
the poorest correlations of these types occurred between the Dichot and the Wedding
samplers (0.92 to 0.98), despite the fact that the data for both these types of methods
consisted of averages over a pair of instruments.
Regressions applied to the temporal site data were used to further examine
relationships between types of samplers. The SAM, PEM, and Wedding levels were modeled
using dichot levels as independent variables. The following models were used:
ln(V) = ln(A + BXi +e
where Y is the measured concentration via SAM, PEM, or Wedding; X is the measured
concentration from dichot; A, B, and C are parameters to be estimated; and e is the random
error term, assumed to be distributed with mean zero and constant variance. Model (2) is a
special case of Model (1), with A=0. Thus Model (2) is preferred so long as it furnishes an
adequate fit The models imply a linear relationship among concentrations measured by a
given method and the dichot, but assume that the measurement errors in the Y variable are
proportional to the expected value of Y. This error structure indicates that estimation of the
mode! parameters should be carried out in the logarithmic scale. The estimates of A and B
were obtained by nonlinear least squares, while the estimate of C was obtained as the
geometric mean of the Y:X ratios (i.e., by exponentiating the least squares estimate of ln(C)).
Table 9-9 presents the model estimation results; it gives the R2 statistics (proportion of
the total variance explained by the model), estimates of the model parameters (denoted by A),
and estimated measures of precision for the estimates. For the first model, the precision
measures are reported in terms of asymptotic standard errors, which can be used to form
9-16
-------
Dep.
Var.
Y
TABLE 9-9. TEMPORAL SITE RELATIONSHIPS BETWEEN SAM, PEM, AND WEDDING
AEROSOL CONCENTRATION MEASUREMENTS AND THOSE OF DICHOTSa
Model Y = A + BX Model Y - CX
Indep.
Var.
X
Time of
Day
n
R2
A
ase(A)
6
ase(6)
R2
C
ase(C)
95% CI for Cb
Low High
dic25
All
89
0.867
0.90
0.88
1.209
0.062
0.865
1.257
1.035
1.173
1.346
Day
45
0.817
0.05
1.37
1.289
0.105
0.817
1.292
1.058
1.154
1.447
Night
44
0.921
1.51
1.00
1.141
0.067
0.916
1.221
1.039
1.130
1.320
DIC10
All
90
0.972
-0.87
1.02
1.165
0.022
0.972
1.149
1.010
1.126
1.172
Day
45
0.969
0.99
1.58
1.134
0.032
0.968
1.151
1.015
1.118
1.185
Night
45
0.976
-2.62*
1.33
1.200
0.032
0.974
1.147
1.014
1.116
1.179
DIC10
All
87
0.970
1.26
0.94
1.084
0.021
0.970
1.108
1.011
1.084
1.132
Day
42
0.965
3.38*
1.66
1.063
0.032
0.961
1.120
1.016
1.084
1.157
Night
45
0.974
0.44
1.14
1.087
0.029
0.973
1.097
1.014
1.066
1.128
DIC10
All
91
0.918
-0.09
1.36
0.929
0.032
0.918
0.928
1.019
0.894
0.963
Day
45
0.916
1.11
2.22
0.976
0.044
0.916
0.995
1.024
0.950
1.043
Night
46
0.931
0.54
1.59
0.854
0.040
0.931
0.866
1.025
0.824
0.910
SAM
'2.5
SAM10
PEM10
WED10
a Averages over duplicates are used as source data whenever valid duplicate measurements are available for a method (almost all cases for
the Dichot and Wedding measurements, but only a few cases for the SAM and PEM measurements). For both model forms, the Rz statistic
is defined as 1 -(residual sum of squares)/(total corrected sum of squares),
b The difference between geometric means for the two methods is statistically significant (0.05 level) when the CI range does not include 1.00.
* Test of significance for the A parameter rejected the hypothesis of a zero intercept, at the 0.05 level of significance.
-------
approximate confidence intervals for the parameters or to perform hypothesis tests. To test
that A is zero, for instance, the ratio of the estimate to its asymptotic standard error (ase) was
computed and compared to percentage points of the standard normal distribution. At the 5%
level of significance (critical value of 1.96), the hypothesis that A=0 was not rejected except
for two borderline cases (SAM nighttime and PEM daytime). Model (2) thus seems to apply,
except possibly in those cases. For the second model, the precision of the estimate of C is
reported as a geometric standard error. Approximate confidence intervals for C can be
constructed by multiplying and dividing the estimated C by G*. where G is the geometric
standard error (gse) and t is the appropriate percentage point of the t distribution with n-1
degrees of freedom. The approximate 95% confidence limits shown in the last two columns of
the table were computed in this manner with t=2. The fact that the confidence intervals for C
do not include 1.00 (with one exception) indicates that there is a statistically significant (0.05
level) difference between the geometric means of the two methods represented by X and Y.
The results indicate that the SAM PM2 5 concentrations exceed those of the dichot by about
26%, while the SAM and PEM PM^q concentrations exceed the dichot by about 15% and
11%, respectively. For those three cases, the daytime estimates of C were slightly higher
than those for the night; for the Wedding, however, the day/night differences were large:
daytime levels were equal for the two methods, but the Wedding levels for the nighttime
samples were about 13% lower than those of the dichot. The estimate of C for the Wedding
also exhibited poorer precision than the PEM and SAM parameters, even though the
estimated relationship for the Wedding was based on averages of two measurements.
Additional relationships of the same form were estimated for the SAM, PEM, and
Wedding PM^q samples. These results are given in Table 9-10. The PEM and SAM show a
very high association (R^>0.98), with the PEM undersampling by about 5% relative to the
SAM. While the PEM and SAM, on average, produce higher PM^q concentrations than the
Wedding, this relation can not be expressed as a single percentage, since the assumption of
zero intercepts for the models relating the PEM and SAM concentrations to the Wedding
concentrations does not appear to be valid.
The results of Table 9-9 suggest the possibility that the over- or undersampling of one
method relative to the dichot may not be uniform for the coarse and fine fractions. For
9-18
-------
TABLE 9-10. TEMPORAL-SITE RELATIONSHIPS BETWEEN SAM, PEM, AND WEDDING
AEROSOL CONCENTRATION MEASUREMENTS3
Model Y = A + BX Model Y = CX
Dep. Indep. 95o/o Cl for cb
Var. Var. Time of
Y X Day n R2 A ase(A) © ase(6) R2 C ase(C) Low High
PEM
10
PEM
10
SAM
10
SAM
WED,
WED.,
10
All
85
0.983
1.37
0.77
0.931
0.015
0.982
0.953
1.008
0.938
0.969
Day
41
0.978
2.54
1.36
0.921
0.023
0.976
0.957
1.013
0.932
0.982
Night
44
0.986
0.75
0.91
0.936
0.019
0.986
0.950
1.010
0.931
0.969
All
84
0.926
5.57*
1.64
1.084
0.039
0.915
1.202
1.018
1.159
1.247
Day
40
0.948
4.27*
2.09
1.058
0.041
0.943
1.131
1.020
1.087
1.178
Night
44
0.911
4.22
2.49
1.163
0.069
0.905
1.270
1.028
1.202
1.341
All
87
0.911
6.32*
1.92
1.117
0.044
0.900
1.247
1.020
1.200
1.297
Day
43
0.933
2.10
2.42
1.126
0.049
0.932
1.163
1.022
1.113
1.215
Night
44
0.907
6.62*
2.79
1.178
0.073
0.895
1.336
1.029
1.262
1.414
0 Averages over duplicates are used as source data whenever valid duplicate measurements are available for a method (almost all cases for
the Dichot and Wedding measurements, but only a few cases for the SAM and PEM measurements). For both model forms, the R statistic
is defined as 1-(residual sum of squares)/(total corrected sum of squares).
The difference between geometric means for the two methods is statistically significant (0.05 level) when the CI range does not include 1.00.
* Test of significance for the A parameter rejected the hypothesis of a zero intercept, at the 0.05 level of significance.
-------
example, the SAM:dichot ratio for PM2 5 seems higher than the corresponding ratio for PM^.
By definition,
ctichot PM-)o = 100[dichot fine] + 1.00[dichot coarse]. (Eq. 9-4)
For another PM10 method (say, SAM), an analogous relationship can be expressed as
SAM PM<\q = a[dichot fine] + ^[dichot coarse]. (Eq. 9-5)
Equal over- or undersampling of the fine and coarse fractions would be indicated if a = 6.
However, if a * 6, then a nonuniform over- or undersampling of the fine and coarse fractions
would be implied. To explore this possibility, the coarse fraction concentration for the dichot
was determined (by subtracting the PM2 5 concentration from the PM^q concentration).
Estimates of a and P for the SAM relationship were obtained in two ways. One was to apply
nonlinear least squares to the model
In (V) = In a Xf + pXc (Eq. 9-6)
where
Xp = dichot fine-fraction concentration (PM2 5)
Xc = dichot coarse-fraction concentration (PM1Q- PM2 5)
Y = SAM PM-jq concentration.
The second way of estimating a and p involved direct use of model (9-5), which was fit by
ordinary regression. The first estimation approach is preferable if the observations exhibit a
homogeneous multiplicative error structure, while the latter is preferable if an additive structure
occurs. Similar models for the PEM and Wedding data were also fit.
The results, presented in Table 9-11, do seem to indicate that all three methods tend
to sample the fine fraction at a higher rate than the coarse fraction (or, alternatively, that the
dichot tends to undersample the fine fraction relative to these methods). This discrepancy
seems more pronounced for the Wedding than for the SAM or PEM. The coefficient on the
fine fraction for the SAM is 1.222 (±0.035), which compares favorably with the estimated C of
1.257 for the SAM PM2 5 versus the dichot PM2 5 (see Table 9-9).
In addition to the 12-hour monitors, a Cascade impactor operated periodically at the
temporal site produced four multi-period, multi-cut observations. These data, and their
associations with the 12-hour monitoring data, are given in Appendix L.
9-20
-------
TABLE 9-11. TEMPORAL SITE RELATIONSHIPS BETWEEN SAM, PEM, AND WEDDING
PM10 AEROSOL CONCENTRATIONS AND THE DICHOT FINE AND COARSE FRACTIONS
InfY) = Infa(Fine) + B(Coarse)l Y = ot(Fine) + B(Coarse)
Dep.
Time of
n
R2
a
&
R2
a
0
Var. Y
Day
(s.e.)
(s.e.)
(s.e.)
(s.e.)
SAM10
All
90
0.974
1.222
1.083
0.979
1.280
1.023
(0.035)
(0.032)
(0.021)
(0.024)
Day
45
0.969
1.210
1.106
0.973
1.317
1.000
(0.058)
(0.045)
(0.035)
(0.050)
Night
45
0.977
1.242
1.047
0.985
1.234
1.091
(0.045)
(0.046)
(0.028)
(0.05)
PEM10
All
87
0.970
1.158
1.064
0.970
1.249
0.916
(0.037)
(0.033)
(0.023)
(0.026)
Day
42
0.964
1.226
1.042
0.961
1.335
0.861
(0.060)
(0.045)
(0.039)
(0.033)
Night
45
0.974
1.120
1.073
0.988
1.127
1.104
(0.048)
(0.049)
(0.024)
(0.041)
WED10
All
91
0.928
1.095
0.783
0.948
1.216
0.692
(0.052)
(0.043)
(0.029)
(0.033)
Day
45
0.949
1.327
0.760
0.932
1.325
0.687
(0.065)
(0.043)
(0.047)
(0.038)
Night
46
0.940
1.012
0.717
0.978
1.229
0.493
(0.062)
(0.060)
(0.033)
(0.058)
NOTE: The standard error (s.e.) of each estimate is given in parentheses.
Estimates for the logarithmic-scale model were obtained via nonlinear least squares; the concentration-scale model,
via ordinary least squares. The R2 values In both cases were determined as
1 - (residual sum of squares)/(total corrected sum of squares).
9-21
-------
Meteorological Data
Most of the meteorological data collected at the temporal site were not usable due to
equipment failure. Therefore, data were obtained from three nearby weather stations (Ontario
Airport, Riverside Airport, March Air Force Base). These hourly data (temperature, dew point,
wind speed and direction) were averaged over time to produce a series of 96 12-hour
averages corresponding approximately with the temporal site monitoring periods (daytime
consisted of all measurements taken between 6:50 AM and 5:50 PM, inclusive). These
averages were then averaged over the stations; since Riverside Airport furnished data only for
daytime periods, the nighttime averages were computed only over the other two stations. The
meteorological data are summarized overall and by the time of day in Table 9-12. The mean
daytime temperature was 76°F, and the mean nighttime temperature was 62°F. The dew
point showed little variation, with a daytime mean of 43.5° and a nighttime mean of 45.6°.
The wind speed tended to be higher during the day than at night, with means of 7.5 and 3.3
knots, respectively. Pearson correlations for the 12-hour average temperatures, dew points,
and wind speeds are presented in Table 9-13. Overall, there were no strong correlations
between them. However, correlations between temperature and the other two meteorological
measurements showed different daytime and nighttime patterns.
Associations Between Meteorological Data and Temporal-Site Aerosol Concentrations
Correlations of the meteorological data with the temporal-site aerosol concentrations
are displayed in Table 9-14. No strong correlations were evident. However, the strongest
occurred for nighttime concentrations with dew point and wind speed, with dew point positively
correlated and wind speed negatively so. Table 9-15 provides further insight into the
association between winds and the temporal-site particle concentrations. The table partitions
the temporal-site data into six meteorological categories: low, moderate, and high winds from
the north and from the west, where the north category included northeast, north, and
northwest winds (after rounding the direction given in degrees to the nearest of the eight
compass points), and the west category included west and southwest winds (similarly
rounded). For each category, the table gives the mean SAM PM2 5 and PM^q concentrations
and the ratio of the means. The results indicate a dear pattern in the ratios for the north
winds, with a predominance of fine particles in the low wind case (<3 knots) and a
predominance of coarse particles in the high wind case (>8 knots). A similar but much less
9-22
-------
TABLE 9-12. SUMMARY OF DISTRIBUTIONS OF METEOROLOGICAL DATA
Percentiles
Mean S.E. Min 25th 50th 75th Max.
Overall8
Average Temperature (°F)
69.0
0.9
52.2
62.0
66.6
76.0
88.5
Average Dew Point (°F)
44.0
1.2
17.0
35.1
47.0
54.7
58.8
Average Wind Speed (Knots)
5.4
0.4
0.5
2.6
5.1
6.7
18.9
Davtimeb
Average Temperature (°F)
75.9
0.9
63.9
69.9
76.0
80.8
88.4
Average Dew Point (°F)
43.5
1.7
19.0
32.5
44.5
54.4
58.8
Average Wind Speed (Knots)
7.5
0.5
3.9
5.7
6.6
7.4
18.3
Niahttime0
Average Temperature (°F)
62.2
0.6
52.2
60.5
62.0
65.4
69.7
Average Dew Point (°F)
44.6
1.8
17.0
36.4
49.3
55.0
58.8
Average Wind Speed (Knots)
3.3
0.5
0.5
1.9
2.6
3.4
18.9
a Overall Observation = simple average of daytime and nighttime.
b Daytime Observation = average over Ontario, Riverside, and March AFB stations and over daytime
periods (0650-1750)
0 Nighttime Observation = average over Ontario and March AFB stations and over nighttime periods
(1850-0550)
9-23
-------
TABLE 9-13. CORRELATIONS OF METEOROLOGICAL DATA8
Pearson Correlations
Quantities Correlated All Periods Daytime Periods Nighttime Periods
Temperature, Dew Point 0.08 0.04 0.43*
Temperature, Wind Speed 0.33* -0.36* 0.00
Dew Point, Wind Speed -0.37* -0.40* -0.44*
a Observations are as defined in Table 12. Sample sizes are 48 for the daytime and nighttime cases
and 96 for the "all periods" case.
* Significantly differerent from zero at the 0.05 level.
9-24
-------
TABLE 9-14. PEARSON CORRELATIONS OF METEOROLOGICAL AND TEMPORAL-SITE DATA3
OVERALL CORRELATIONS (Sample sizes range from 87 to 95)
Temp. Dew Point Wind Speed
PM25 SAM 0.03 0.26* -0.25*
Didior -0.02 0.31* -0.32*
PM10 PEM 0.17 0.22* 0.03
SAM 0.11 0.15 0.03
Dichoth 0.10 0.11 0.10
Wedding 0.21* 0.15 -0.02
DAYTIME CORRELATIONS (Sample sizes range from 42 to 47)
Temp. Dew Point Wind Speed
PMo c SAM 0.07 0.28 -0.20
DBioT 0.03 0.41* -0.37*
PM10 PEM -0.03 0.12 0.10
SAM -0.06 0.01 0.18
Dichotb -0.13 -0.04 0.29*
Wedding 0.13 0.12 -0.00
NIGHTTIME CORRELATIONS (Sample sizes range from 44 to 48)
Temp. Dew Point Wind Speed
PMpc SAM
Dicriotb
0.05
0.24
-0.36'
0.00
0.23
-0.35
PM10 PEM
SAM
0.04
0.33*
-0.29
0.06
0.30*
-0.30'
Dichot
0.02
0.29*
-0.30'
Wedding
0.03
0.21
-0.29'
a Meteorological observations are as defined in Table 9-12.
k Average of two samplers (if both available).
* Significantly different from zero at the 0.05 level.
9-25
-------
TABLE 9-15. TEMPORAL-SITE SAM2 5 AND SAM10 MEAN CONCENTRATIONS (ng/m3),
BY WIND DIRECTION AND SPEED
Wind
Direction®
Wind Speed
(Knots)
S*M2.5
Cone.
SAM10
Cone.
N
Mean
S.E.
N
Mean
S.E.
2.5/10
North
Less than 3
14
69.1
15.0
15
94.0
15.B
0.74
3 to 8
8
20.2
8.8
9
42.4
9.6
0.48
8 or more
9
24.5
7.3
9
90.9
19.7
0.27
West
Less than 3
15
51.3
9.8
15
84.5
10.9
0.61
3 to 8
40
49.6
5.6
39
88.6
7.0
0.56
8 or more
3
32.4
11.6
3
70.5
19.8
0.46
a Wind directions were based on the 12-hr wind vector, averaged over stations. Directions were then rounded to the nearest compass
point (N, NW, W, etc.). The "north" category includes cases where this rounded wind direction was north, northeast, or northwest. The
"west" category includes the west and southwest directions.
-------
pronounced pattern in the ratios was evident for the west wind case. The mean PM^q
concentrations for the north wind case seemed to be lower for the moderate wind speed
category than for the low or the high wind categories; this difference was not apparent for the
west wind case.
Winds were predominantly from west and southwest, with about two-thirds of the
samples collected under these wind conditions. Westerly winds rose above 8 knots (12-hour
average) during only three monitoring periods and the typical wind speed was 3 to 8 knots.
Winds from the west were generally on-shore winds that traveled across the greater Los
Angeles basin before arriving in Riverside. These winds probably carried some particles
generated in the LA. basin. Winds originated from the northerly directions during the
remaining third of the study monitoring periods. Northerly winds were more likely to be either
less than 3 knots (15 monitoring periods) or greater than 8 knots (9 monitoring periods).
These winds generally came off of the desert and over the San Bemadino area before arriving
in Riverside. Strong northeasterly winds were typically Santa Ana winds with low humidity that
carried coarse particles at ground level. Very strong Santa Ana winds with gusts above 40
knots were observed November 2-3, and November 6-7, with high PM^ and low PMg 5
concentrations during these time periods. General visibility was usually much greater when
q
northerly winds were from 3 to 8 knots, and mean PM10 (42 vs 89 jig/nr) and PM2 5 (20 vs
50 n-g/m3) concentrations were much lower than with westerly winds at 3 to 8 knots.
Few fires were reported in Southern California during the course of the field study.
One local fire was reported in a very large manure pile at a fertilizer plant between Ontario
and Riverside. This fire began during Santa Ana winds on November 3 and was still
smoldering through November 8. The odor from this fire was observed in the southwestern
part of Riverside on November 4-5, but there was no visible smoke. Another fire burned in
Los Angeles during Santa Ana winds on November 6, but this fire was downwind and over
thirty miles away. Fires were not a major particle source during this study.
PERSONAL AND RESIDENTIAL DATA ANALYSIS
Data Collection and Treatment
Each of the 178 persons who participated in the study was asked to wear a PEM for
two consecutive periods (nominally 12 hours each). Concurrent PM^ and PM2 5 samples
were collected indoors by a SIM in the main living area and outdoors by an identical (SAM)
9-27
-------
monitor at each home. This resulted in 10 samples per household (day and night samples for
PM2 5 SIM and SAM and PM^q SIM, SAM, and PEM). Concentrations from duplicate
samples, where used, were averaged to obtain one observation. The participant and
residence data are listed in Appendix M.
Descriptive statistics for the participant and residence data were calculated using
sampling weights to adjust for the disproportionality of sample selection, as described in
Section 8. SAM and SIM data were weighted to reflect the target population of household
days, while the PEM data were weighted to reflect a target population of person-days. The
statistics were produced using SUDAAN (SUrvey DAta ANalysis) software. SUDAAN includes
statistical analysis procedures developed at RTI to analyze data collected using complex
survey designs.
Distribution of 24-Hour PM^q Concentrations
One primary goal of this study was to examine the distribution of exposure of an urban
population to airborne particles with regard to the 150 *ig/m3 National Ambient Air Quality
Standard and the California Ambient Air Quality Standard (CAAQS) of 50 p.g/m3. Of particular
interest was the measurement of personal exposure levels as compared to microenviron-
mental levels inside and outside the home and an ambient air monitoring site such as our
temporal site. The 24-hour concentrations were calculated as time-weighted averages across
the two approximately 12-hour measurement periods. Distributions of 24-hour PM^
concentrations are reported for each sample type in Table 9-16. Data were included only
when usable samples were obtained during both 12-hour monitoring periods. Bar plots of
these data are presented in Figure 9-3.
Personal exposure concentrations, measured as person-days, were higher than all
other fixed location concentrations. Approximately 25% of the person-days exceeded the
24-hour 150 jig/m3 level during this study. Over 90% of the population was exposed to 24-
hour PM-jq levels above 50 ng/m3. The 150 ng/m3 level was also exceeded outdoors at
home for approximately 15% of the population. This concentration was reached indoors at
home for less than 10% of the population and for all monitor types at the temporal site.
These data indicate that a significant portion of the study population is exposed to particle
levels higher than the NAAQS and CAAQS, and that ambient air and microenvironmental
9-28
-------
TABLE 9-16. DISTRIBUTIONS OF 24-HOUR PM10 CONCENTRATIONS (ng/m3)
Temporal Site Data Residence Data3
Dichot Weddin SAM PEM SAM SIM PEM
9
Sample Size
47
45
43
39
153
157
161
Minimum
15.1
16.2
17.5
17.6
14.8
19.9
33.6
Maximum
147.3
163.5
181.0
169.5
280.4
324.8
286.9
Mean
70.9
67.3
83.2
79.8
91.2
79.0
112.5
(Std. Error)
(4.9)
(5.5)
(6.3)
(5.9)
(4.1)
(4.1)
(6.0)
Geometric Mean
63.6
58.7
73.6
71.6
81.4
68.5
100.7
(Std. Error)
(4.5)
(4.6)
(5.7)
(5.6)
(3.7)
(3.8)
(5.6)
Percentiles
10th
37.0
32.2
43.3
36.0
47.2
32.9
52.1
25th
46.1
42.3
53.7
52.3
60.9
45.1
70.1
50th (median)
60.2
51.8
67.9
71.0
78.3
65.3
102.5
75th
94.1
86.9
114.0
107.4
112.5
106.4
148.2
90th
125.4
134.6
147.6
138.9
159.7
143.6
183.5
Std. Errors of
Percentiles
10th
3.3
4.0
3.0
25th
2.7
3.3
6.0
50th (median)
3.3
3.7
6.1
75th
12.5
9.3
7.0
90th
6.4
6.0
9.4
a Statistics other than the sample size, minimum, and maximum are calculated using weighted data;
they provide estimates for the target population of person-days (PEM) or of household days (SIM,
SAM).
9-29
-------
/ig/m3
TEMPORAL
SITE
HOUSEHOLD DAYS
(WEIGHTED)
PERSON DAYS
(WEIGHTED)
190 -
90th
170 -
24-hr
150 -
NAAQS
¦ —
75tK
130 -
90th
110 -
###
Mean
SOth
90 -
75th
###
—
###
###
###
70 -
Mean
###
###
25th
—
50th
24-hr
50 -
CAAQS
25th
10th
10th
30
DICHOT WEDDING SAM PEM OUTDOOR INDOOR PERSONAL
SAM SIM PEM
(n-47) (n=45) (n=43)(n=39) (n=153) (n-157) (n=161)
Figure 9-3. 24-Hour PMio Distributions
9-30
-------
monitoring do not necessarily represent people's actual exposures. It should be remembered
that the PEM sampling system operated at the temporal site resulted in PM10 concentrations
8% higher than the dichotomous samplers and 13% to 24% higher than the Wedding sampler.
These differences were small but statistically significant at the 0.05 level. It is clear though
that a portion of the population is exposed to high particulate levels. It is also clear from
examining PEM data collected at the temporal site, as compared to the PEM data collected on
individuals, that monitoring PM^q levels in ambient air at this one site underestimated the
population's actual exposure by 35% to 40%.
Distribution of 12-Hour PM1Q and PM^ ^ Concentrations
The daytime and nighttime SAM, SIM, and PEM aerosol concentrations are
summarized in Table 9-17. For each of the ten types of samples (columns of the table), the
table gives a series of summary statistics. Except for the first three rows (sample size, and
minimum and maximum concentrations), all of the statistics pertain to the indicated target
populations. These statistics include estimates of the population mean and population
geometric mean, and estimates of selected population percentiles. Approximate standard
errors for each of these reported statistics are also given. (It should be noted that the
method employed by SUDAAN for determining percentile estimates is based upon
interpolations from histogram-type estimates; consequently, the estimates are not invariant to
scale, and when small sample sizes and/or small relative frequencies occur, instability in the
estimated percentiles and/or in their estimated standard errors can result.) The table also
gives estimated population standard deviations and population geometric standard deviations
(gsd's). Figure 9-4 provides a graphical summary of the PM^ q concentration distributions that
highlights some of the similarities and differences among the household outdoor, household
indoor, and personal exposure distributions. Table 9-18 provides statistics that summarize the
distributions of concentration ratios from the different matrices: indoor versus outdoor
(SIM/SAM), personal versus indoor (PEM/SIM), and personal versus outdoor (PEM/SAM).
Tables 9-17 and 18, and Figure 9-4, reveal the following:
• SIM versus SAM: Distributions of the daytime particulate levels for the indoor and
outdoor samplers differed only slightly, for both PM2 5 and PM1Q. Medians of the
SIM/SAM ratios were 0.94 and 0.96, respectively, for the two size cuts. The
9-31
-------
TABLE 9-17. WEIGHTED DISTRIBUTIONS OF PERSONAL, INDOOR, AND OUTDOOR8 PARTICLE CONCENTRATIONS (ng/m3)
DAYTIME NIGHTIME
PM25
PM1p
PM
>2.5
PM10
SAM
SIM
SAM
SIM
PEM
SAM
SIM
SAM
SIM
PEM
Sample Size
167
173
165
169
171
161
166
162
163
168
Minimum
7.4
2.8
16.2
16.6
35.1
3.4
2.9
13.6
14.1
19.1
Maximum
187.8
238.3
506.6
512.8
454.8
164.2
133.3
222.9
180.3
278.3
Mean
48.9
48.2
94.9
94.7
149.8
50.5
36.2
86.3
62.7
76.8
(Std. Error)
(3.5)
(4.1)
(5.5)
(5.7)
(9.2)
(3.7)
(2.2)
(4.4)
(3.2)
(35)
Geometric Mean
37.7
35.0
82.7
78.2
128.7
37.2
26.7
74.5
53.1
67.9
(Std. En-or)
(2.5)
(3.3)
(4.1)
(5.0)
(8.5)
(3.1)
(1.9)
(4.0)
(3.1)
(3.1)
Std. Deviation
38.1
41.3
59.6
63.1
88.5
41.0
29.9
51.3
40.3
42.4
Geometric Std.
Deviation**
2.84
2.57
3.09
2.96
3.30
3.52
3.07
3.56
3.80
2.96
Percentiles
10th
14.9
11.5
42.8
30.9
59.9
14.5
10.0
39.3
25.2
36.6
25th
23.4
19.3
56.9
49.5
86.1
23.0
14.8
53.6
33.5
48.1
50th (median)
35.5
33.5
84.1
81.7
129.7
35.0
25.9
74.1
51.6
66.2
75th
60.1
61.5
110.8
127.2
189.1
64.9
48.9
103.7
84.8
98.8
90th
102.2
101.0
157.2
180.7
263.1
120.7
82.7
167.8
116.9
135.0
Std. Errors of
Percentiles
10th
1.6
3.4
2.3
3.4
4.0
2.1
0.9
7.4
1.5
1.5
25th
2.1
1.4
4.5
4.3
9.4
2.7
1.3
3.4
2.4
3.1
50th
4.0
4.5
4.7
8.3
7.5
2.4
2.4
4.8
3.5
4.3
75th
3.9
3.3
4.0
9.4
10.8
4.6
5.3
5.1
4.7
8.2
90th
4.6
6.7
7.2
11.0
12.0
5.8
5.8
4.3
5.3
10.1
a Statistics other than the sample size, minimum, and maximum are calculated using weighted data; they provide estimates for the target
population of person-days (PEM) or of household-days (SIM, SAM).
In contrast to the other statistics, the gsd is a unitless quantity.
-------
^g/m3
280
240
200
160
120
80
40
0
OUTDOOR INDOOR PERSONAL OUTDOOR INDOOR PERSONAL
(SAM) (SIM) (PEM) (SAM) (SIM) (PEM)
Figure 9-4. Estimated PMio Distributions (Weighted)
9-33
DAYTIME
NIGHTTIME
90th
75th
###
Mean
50th
###
###
25th
10th
m
###
###
-------
TABLE 9-18. WEIGHTED DISTRIBUTIONS® OF RATIOS OF PARTICLE CONCENTRATIONS
PM
Z5_
SIM/SAM
DAYTIME
NIGHTIME
PM
10
PM,
SIM/SAM PEM/SIM PEM/SAM
Z5_
PM
SIM/SAM
IIL
SIM/SAM PEM/SIM PEM/SAM
Sample Size
Geometric Mean
(Std. Error)
Percentiles
10th
25th
50th (median)
75th
90th
Std. Errors of
Percentiles
162
0.919
(0.070)
0.405
0.693
0.942
1.323
2.050
159
0.928
(0.056)
163
1.559
(0.062
158
1.521
(0.089)
153
0.733
(0.039)
151
158
0.704
(0.035)
1.240
(0.044)
155
0.891
(0.036)
0.394
0.863
0.623
0.379
0.385
0.798
0.433
0.634
1.145
0.992
0.505
0.501
0.967
0.619
0.964
1.521
1.484
0.718
0.690
1.184
0.878
1.272
1.942
2.262
0.951
0.915
1.470
1.167
1.729
2.831
3.292
1.428
1.305
2.097
1.675
10th
0.080
0.038
0.081
0.063
0.021
0.021
0.027
0.024
25th
0.041
0.060
0.067
0.072
0.025
0.029
0.034
0.040
50th
0.070
0.059
0.054
0.138
0.045
0.031
0.035
0.043
75th
0.107
0.148
0.063
0.120
0.037
0.049
0.077
0.115
90th
0.682
0.070
0.259
0.479
0.205
0.397
0.190
0.127
a Statistics other than the sample size are calculated using weighted data; they provide estiamtes for the target population of person-days
(PEM) or of household-days (SIM, SAM).
-------
distribution of nighttime outdoor concentrations was fairly comparable to the
daytime distribution; however, the nighttime indoor levels were considerably lower
than the outdoor levels, on average, with estimated population medians of the
SIM/SAM ratios equal to 0.72 and 0.69, respectively, for PM2 5 and PM^q.
• PEM versus SIM and SAM: Nighttime PEM PM1Q concentrations (median = 66.2
p.g/m3) tended to fall between the outdoor and indoor PM^q concentrations
(medians = 74.5 and 53.1 ng/m3, respectively). Daytime PEM PM1Q
concentrations, on the other hand, tended to substantially exceed both the SAM
and the SIM levels (SAM, SIM, and PEM PM-jq medians were 84.1, 81.7, and
Q
129.7 ng/m , respectively). Daytime personal exposures to PM10 for the target
population (non-smoking residents of Riverside over age 9) were roughly twice as
high as the nighttime exposures (median daytime concentration = 129.7 jxg/m3,
Q
median nighttime concentration = 66.2 ng/n-r).
Daytime personal exposure PM10 concentrations exceeded indoor and outdoor
microenvironmental concentrations. Over 40% of the population was exposed to PM^ q levels
above 150 iig/m3 during the daytime. Ten percent of the population was exposed to very
high PM10 levels, above 263 jig/m3, for integrated exposure durations of approximately 12
hours. These high exposure levels would not have been predicted from either
microenvironmental or ambient air monitoring. It is likely that a person's movements and
activities (i.e., vacuuming, yard work, occupational) generate a higher particle concentration In
his or her vicinity, and that a person may get closer to localized particle sources than
microenvironmental or ambient monitors would be placed. It is significant that personal
exposure PM10 levels dropped by 50% from the daytime to overnight time periods. The
sampling monitors could not be worn to bed and thus become indoor microenvironmental
monitors while individuals slept. Overnight personal PM10 levels were still higher than indoor
levels, reflecting the individual's activities while awake. It is possible that the personal
sampling filters were being contaminated with particles from the body, thereby overstating
actual exposures to airborne PM^q levels. Elemental analysis of the samples (Section 10)
appears to refute this premise; element concentration ratios to total particle mass were similar
between personal, indoor, and outdoor samples with the exception of sulfur. If the personal
samples had been contaminated with organic debris then the elemental concentration ratios to
total particle mass should have been lower. It is also possible that, due to movement, the
9-35
-------
personal monitor is oversampling particles larger than 10 (im that are not adhering to the
impactor plate. Further work will be necessary to evaluate this potential bias.
Table 9-19 summarizes distributions of PM2 5/PM10 concentrat'on ratios for outdoor
and indoor samplers, by time of day. The table gives sample sizes (number of pairs with
usable data), and estimated population means, geometric means, and percentiles (10th, 25th,
50th, 75th, and 90th) for the population of eligible households. The distributions were similar
for the four cases, but generally showed a higher fine fraction for indoor samples and for
nighttime samples. The four medians, expressed as percentages, for instance, were as
follows:
Daytime Nighttime
SAM 46.9 51.5
SIM 49.8 55.6
To examine the degree of association between the various types of field
concentrations, Pearson correlations were calculated overall and by time of day. These are
shown in Table 9-20. (Weighted correlations were calculated for a few cases and were found
to be quite comparable to the unweighted correlations given in the table.) There are basically
three types of correlations in the table, as depicted below:
Sample Types
Different Matrix and Cut Size:
2.5(1
SAM,
10(1
PEM
2.5n
SAM,
10(1
SIM
2.5(i
SIM,
10(1
PEM
2.5 (i
SIM,
10(1
SAM
Different Matrix, Same Cut Size:
2.5(1
SAM,
2.5(1
SIM
10(i
SAM,
10(1
PEM
10(i
SAM,
10(1
SIM
10ji
SIM,
10(1
PEM
Same Matrix, Different Cut Size:
2.5(i
SAM,
10(1
SAM
2.5(1
SIM,
10(1
SIM
The strongest correlations occurred in the last group ~ that is, between the paired PM2 5 and
PM^q samples. For all three groups, the nighttime correlations were uniformly higher than the
daytime correlations of the same type.
9-36
-------
TABLE 9-19. WEIGHTED DISTRIBUTIONS® OF PM2 5/PM10 CONCENTRATION RATIO
DAYTIME
NIGHTTIME
Outdoor
Indoor
Outdoor
Indoor
Sample Size
Mean
(Std. Error)
Geometric Mean
(Std. Error)
Percentiles
10th
25th
50th (median)
75th
90th
Std. Errors of
Percentiles
10th
25th
50th
75th
90th
160
0.470
(0.016)
0.444
(0.017)
0.274
0.371
0.469
0.571
0.671
0.018
0.018
0.015
0.019
0.012
167
0.492
(0.021)
0.455
(0.022)
0.250
0.347
0.498
0.607
0.735
0.030
0.046
0.020
0.024
0.028
154
0.522
(0.017)
0.497
(0.019)
0.308
0.406
0.515
0.646
0.731
0.023
0.028
0.022
0.027
0.016
160
0.550
(0.014)
0.517
(0.016)
0.301
0.440
0.556
0.694
0.771
0.023
0.017
0.015
0.023
0.012
a Statistics other than sample size are calculated using weighted data; they provide estimates for the
target population of household-days.
9-37
-------
TABLE 9-20. CORRELATIONS OF PERSONAL AND RESIDENCE PARTICLE CONCENTRATIONS
OVERALL CORRELATIONS (Sample sizes range from 308 to 339)
PMas
PM10
SAM SIM
PEM
SAM
SIM
PM„, SAM - 0.67
SIM
0.37
0.59
0.91
0.57
0.52
0.88
PM10 PEM
SAM
SIM
0.41
0.70
0.52
DAYTIME CORRELATIONS (Sample sizes range from 158 to 173)
PM2.5
PM10
SAM SIM
PEM
SAM
SIM
PM,SAM - 0.65
25 SIM
0.38
0.54
0.89
0.46
0.55
0.88
PM,0 PEM
SAM
SIM
0.35
0.63
0.46
NIGHTTIME CORRELATIONS (Sample sizes range from 150 to 168)
PM2.5
PM10
SAM SIM
PEM
SAM
SIM
PM, c SAM - 0.74
25 SIM
0.58
0.77
0.97
0.75
0.61
0.93
PM10 PEM
SAM
SIM
-
0.62
0.80
0.65
9-38
-------
Associations of Residence Data with Temporal-Site Data and Meteorological Data
The residence data of those households/individuals having data in a given time period
(one of the 96 data collection periods) were matched with the temporal-site data from the
same period. Correlations of the residence aerosol concentrations with the temporal-site
concentrations were then generated. These unweighted, Pearson correlations are given in
Table 9-21 for all 96 time periods, for the 48 daytime periods, and for the 48 nighttime
periods. The indicated sample sizes refer to the number of usable residence samples that
could be matched with the temporal-site data. As might be expected, the highest correlations
were between the outdoor residence (SAM) samples and the temporal-site samples; nighttime
correlations were consistently higher than the corresponding daytime correlations. For each
resident sample type (row of the table), the correlations with the concentrations from various
types of temporal-site methods show little variation, particularly at night. For example, the
correlations between the nighttime SAM PM2 5 samples and the temporal-site samples (both
2.5n and 10h) ranged from 0.95 to 0.96.
Correlations were least strong between daytime temporal-site and personal PM^g
concentations. It is apparent that the population's actual PM^g exposure may not be well
predicted using ambient air monitoring stations. Personal activities and localized processes
appear to lead to particle concentrations that are higher in the vicinity of individuals than In
ambient air or larger microenvironments. Exposure to localized particle sources varies
considerably among individuals. Some individuals may be highly exposed because of their
occupation and commuting to jobs. Others may be exposed to sources at home including
cooking, cleaning, and yard work. Still others may be exposed to environmental tobacco
smoke. These exposures will vary not only between individuals, but over time for a given
individual. Occupationally exposed persons may be exposed to lower PM1 q levels on days
off if there are particle sources at their job, or conversely may see higher PM^g levels from
housework if they are exposed to lower concentrations at work. The representativeness of
using PM^ q concentrations measured at ambient air monitoring stations to evaluate a
population's exposure and possible health effects should be reconsidered based on these
data. The participant's daytime exposures may be underestimated while nighttime exposures
may be overestimated based on ambient monitoring data.
The average 12-hour meteorological data (see Section 9.5) were matched by time period
with the resident concentration data. Table 9-22 gives the correlations between the resident
9-39
-------
TABLE 9-21. CORRELATIONS OF TEMPORAL-SITE, PERSONAL, AND RESIDENCE
AEROSOL CONCENTRATIONS
OVERALL CORRELATIONS (Sample sizes range from 295 to 335)
Temporal-Site Data
PMp.S PM10
Residence Data SAM Dichot3 PEM SAM Dichot3 Wedding
a
SAM
0.90
0.89
0.85
0.84
0.79
0.84
SIM
0.68
0.66
0.68
0.65
0.63
0.66
PEM
0.36
0.33
0.43
0.42
0.41
0.44
SAM
0.75
0.72
0.78
0.78
0.76
0.75
SIM
0.49
0.47
0.55
0.53
0.53
0.53
DAYTIME CORRELATIONS (Sample sizes range from 144 to 169)
Temporal-Site Data
PM?S PM1Q
Residence Data SAM Dichot8 PEM SAM Dichot3 Wedding
a
SAM
0.83
0.82
0.75
0.73
0.65
0.75
SIM
0.70
0.69
0.64
0.60
0.57
0.62
PEM
0.39
0.37
0.37
0.37
0.34
0.38
SAM
0.62
0.57
0.67
0.66
0.62
0.61
SIM
0.53
0.51
0.52
0.51
0.50
0.51
NIGHTTIME CORRELATIONS (Sample sizes range from 149 to 168)
Temporal- Site Data
H!u EStt.
Residence Data SAM Dichot3 PEM SAM Dichot3 Wedding
a
PM, 5 SAM 0.96
' SIM 0.72
PM10 PEM 0.55
SAM 0.92
SIM 0.60
0.95 0.96
0.71 0.72
0.54 0.53
0.91 0.93
0.58 0.59
0.96
0.96
0.95
0.71
0.71
0.71
0.54
0.54
0.55
0.93
0.93
0.92
0.59
0.59
0.59
8 Average of two samplers.
9-40
-------
TABLE 9-22. CORRELATIONS OF METEOROLOGICAL DATA WITH PERSONAL AND RESIDENCE DATA
OVERALL DAYTIME NIGHTTIME
Type of Dew Wind Dew Wind Dew Wind
Field Sample Temp. Point Speed Temp. Point Speed Temp. Point Speed
PM25 SAM -0.02 0.26 -0.21 -0.02 0.26 -0.10 0.05 0.25 -0.35
SIM 0.16 0.20 -0.10 0.03 0.25 -0.20 0.08 0.17 -0.25
PM10 PEM 0.41 0.02 0.15 0.08 0.07 -0.11 0.09 0.03 -0.18
SAM 0.02 0.15 -0.05 -0.12 0.07 0.10 0.03 0.25 -0.36
SIM 0.22 0.04 0.05 -0.05 0.06 -0.09 0.06 0.07 -0.20
Sample Size Range 327-339
158-173
155-168
-------
data and the meteorological data. No strong correlations were found. For the nighttime
sampling periods, the wind speed appeared to be negatively correlated with the concentration
levels.
Questionnaire and Diary Data
Questionnaires were administered to the participants at the end of each 12-hour
monitoring period. Weighted tabulations of the responses are presented in Appendix N. If the
particular item was associated with the home, then household population estimates were
produced using the household-level sampling weights. If the item referred to individuals (e.g.,
types of activities engaged in), then person-level sampling weights were used so as to
produce statistics pertaining to the population of eligible participants.
For each of the two monitoring periods, the participants were asked (in Part B of the
questional re) to record their activities, including a description of each activity, its start and
stop time, its location (indoor or outdoor, at home or away, or in transit), and how much of the
time was in the presence of smokers. The information collected in these diaries was first
summarized for each individual and time period by determining the percentage distribution of
time spent by the individual in the different microenvironments. These percentage
distributions were then summarized over individuals by computing population estimates
(means, 25th, 50th, and 75th percentiles) for each microenvironment. Tables 9-23 and 9-24
provide the summary results for nighttime and daytime, respectively. During the nighttime,
most members of the population were estimated to spend over 90% of their time indoors at
home. During the day, however, there was more variability in environments, with an average
of 51% of the time spent indoors at home, 26% spent indoors away from home, 12% spent
outdoors, and 9% spent in transit. The percentage of time in the presence of smokers was
generally very small for the target population. On average, less than 2% of the time was
spent indoors or in transit in the presence of smokers. This was true for both daytime and
nighttime monitoring periods. (Recall that smokers were not eligible for the study.
Appendix O presents an analysts, based on responses to the Household Enumeration
Questionnaire, of how smokers tend to cluster within homes.)
9-42
-------
Associations of Field Data with Questionnaire and Diary Data
A portion of the 12-Hour Time/Activity Survey was aimed at collecting information on
activities that may increase exposure to particies. This included Question 2 of part C, which
sought data on spraying activities, on indoor activities such as vacuuming and dusting, and on
outdoor activities such as burning leaves. The requested information was obtained in the
following form (illustrated for "dusting"):
Done Nearby
Done or At Home
Activity Nearby At home Away At home Away Resp. Away
a) dusting Y/N (1) (2) (3) (4) (5)
The respondent was first asked to indicate if the activity took place; if so, he or she was asked
to fill in the length of time in minutes under the appropriate heading. For example, if the
respondent was at home while someone else was dusting for 15 minutes, then he or she
would answer 'Y' to "Done or Nearby" and insert "15" under "Nearby, At home."
A person (participant) was defined as exposed if the respondent answered T to "Done
or Nearby" and the activity was listed in categories (1), (2), (3), or (4). A household was
defined as having the activity present during the monitoring period if the respondent answered
'Y' to "Done or Nearby" and the activity was listed in categories (1), (3), or (5). Certain
activities were then combined to form analysis variables (variable names denoted by
uppercase):
HOUSE WORK consisted of vacuuming, dusting, carpet cleaning, indoor cooking, and
using a clothes dryer.
SPRAYING included either pump or propeilant type spraying of paints, cleaners,
disinfectants, air fresheners, hair care products, perfumes/colognes, deodorants,
cooking products, lubricants, insecticides, repellents, and plant care products.
OUTDOOR ACTIVITIES included lawn mowing, gardening, burning leaves or rubbish,
outdoor cooking, and outdoor recreation.
Too few participants answered 'Y' to make any analysis meaningful for the outdoor activities.
In addition to the above, the questionnaire was used to identify participants who went
to WORK (Question 1a, Part C), to identify participants or other household members
9-43
-------
TABLE 9-23. WEIGHTED DISTRIBUTIONS OF PERCENT OF TIME SPENT IN VARIOUS
ENVIRONMENTS, BASED ON THE NIGHTTIME DIARY RESPONSES3
Percentiles
Environment
Mean
S.E.
Min.
25th
50th
75th
Max.
Indoors at Home
92.6
1.3
7.2
92.0
96.0
98.0
100
Indoors away from Home
2.6
1.0
0.0
1.2
2.5
3.7
88.6
Outdoors near Home
0.9
0.2
0.0
0.4
0.8
1.2
27.2
Outdoors away from Home
0.6
0.4
0.0
0.8
1.6
2.4
60.9
In Transit
0.9
0.2
0.0
0.2
0.4
0.6
13.9
a Statistics other than the minimum and maximum are calculated using weighted data; they
provide estimates for the target population of person-days.
9-44
-------
TABLE 9-24. WEIGHTED DISTRIBUTIONS OF PERCENT OF TIME SPENT IN VARIOUS
ENVIRONMENTS, BASED ON RESPONSES TO THE DAYTIME DIARY0
Percentiles
Environment
Mean
S.E.
Min.
25th
50th
75th
Max.
Indoors at Home
51.4
2.0
0.0
23.0
53.1
80.4
100
Indoors away from Home
25.7
1.6
0.0
3.5
14.6
49.8
90.3
Outdoors near Home
4.6
0.6
0.0
0.8
1.6
5.0
37.1
Outdoors away from Home
7.8
1.2
0.0
1.5
3.1
9.6
74.9
In Transit
9.4
0.7
0.0
2.6
6.2
12.3
55.8
a Statistics other than the minimum and maximum are calculated using weighted data; they
provide estimates for the target population of person-days.
9-45
-------
TABLE 9-25. EFFECTS OF ACTIVITIES ON MEAN
PM2 5 AND PM10 INDOOR (SIM) CONCENTRATIONS (fig/m3)
Homes With Activity8
Homes Without Activity8
Geom
Activity and
Geom.
Sample Type
N
Mean
S.E.
Mean
S.E.
N
Mean
S.E.
Mean
S.E.
HOUSE WORK6
PM2S Day
114
53.9*
5.3
39.9*
4.4
59
36.9
4.4
27.1
3.1
Night
65
34.6
3.6
25.3
2.9
101
37.2
2.9
27.5
2.0
PM10 Day
111
106.4*
7.8
91.3*
7.3
58
71.1
7.6
57.1
6.3
Night
66
61.9
5.3
52.2
5.0
97
63.2
3.9
53.7
3.1
SPRAYING0
PM2.s Day
72
43.4
4.7
32.1
3.6
101
51.3
5.4
37.0
4.2
Night
56
37.4
4.3
26.8
3.0
110
35.6
3.0
26.6
2.6
PM10 Day
70
92.6
8.4
76.6
7.3
99
96.1
6.7
79.2
6.0
Night
55
65.3
6.2
54.7
5.1
108
61.4
4.3
52.3
3.9
SMOKE*
PM25 Day
28
71.1*
6.4
60.0*
5.6
143
43.7
4.5
31.4
3.3
Night
31
61.2*
7.3
51.0*
7.5
133
29.5
2.0
22.5
1.6
PM10 Day
28
125.6*
9.2
114.0*
9.7
139
87.8
6.2
72.0
5.0
Night
30
92.9*
8.8
83.6*
9.0
131
54.6
3.2
47.1
3.0
EXHAUST*
PM^ Day
31
28.2
5.5
20.6
3.2
121
50.2*
4.1
39.1*
3.8
Night
8
-
-
-
-
138
36.7
2.4
26.7
1.9
PM10 Day
30
62.3
7.4
50.0
5.7
118
101.2*
6.7
87.3*
6.0
Night
7
-
-
-
-
135
63.1
3.8
52.9
3.3
a Asterisks indicate that the means or geometric means for homes with and without the activity are significantly
different (significant level ^ 0.05). The asterisk is shown beside the higher member of the pair. Means,
geometric means, and their standard errors were calculated using weighted data: they apply to subsets of the
population of household-days.
b Based on responses to questions C2a, b, c, h, and i of the Questionnaire.
c Based on responses to question C2, part 1 of the Questionnaire.
d Based on responses to question C4 of the Questionnaire.
e Based on responses to question C7 of the Questionnaire.
9-46
-------
potentially exposed to EXHAUST from vehicle engines running in an attached garage
(Question 7a, Part C), and to identify households with exposure to tobacco SMOKE (one or
more cigarettes, cigars, or pipefuls indicated in Question 4, Part C). Participants' exposures
to tobacco SMOKE were based on whether or not exposure was reported on the Recall Diary
(Part B of the Questionnaire).
The effects of HOUSE WORK, SPRAYING, SMOKE, and EXHAUST on indoor (SIM)
PM2 5 and PM^g concentrations is summarized in Table 9-25, which gives
estimatedpopulation means and geometric means, and their standard errors, for the exposed
and unexposed domains. The effects of activities on personal PM10 levels is summarized
similarly in Table 9-26. During the daytime monitoring period, those homes and persons
associated with HOUSE WORK had significantly higher mean aerosol levels (significance level
of 0.05) than those not indicating HOUSE WORK. This same effect was also apparent for the
WORK variable in Table 9-26, which showed lower average levels for those going to work
(i.e., many persons reporting not going to work also reported house work activities). The
personal and indoor mean concentration levels for persons reporting SPRAYING were not
significantly different from the corresponding means of those not reporting SPRAYING. The
reported presense of tobacco SMOKE was associated with higher SIM aerosol levels (both
PM2 5 and PM-jq in daytime and nighttime periods), and with nighttime personal levels (PEM
PM^g). The daytime personal means were not significantly different. Daytime personal and
indoor aerosol levels were higher for homes in which no vehicles were run in attached
garages during the monitoring period. This counterintuitive result is likely due to other factors
that may be correlated with participants' having a garage, and suggests the need for analyses
that attempt to deal simultaneously with several explanatory variables. Unfortunately, small
sample sizes will severely limit the meaningfulness of results from such analyses.
9-47
-------
TABLE 9-26. EFFECTS OF PARTICLE-GENERATING ACTIVITIES ON
MEAN PM10 PERSONAL EXPOSURES (ng/m3)
Persons Exposed to Activity® Persons Not Exposed to Activity
Activity and Geom. Geom.
Sample Type N Mean S.E. Mean S.E. N Mean S.E. Mean S.E.
WORK13
Day 59 126.8 11.6 107.6 10.7 111 162.1* 9.9 141.6* 8.5
HOUSE WORK5
Day 110 161.5* 11.0 142.2* 9.6 61 125.0 11.4 104.4 10.6
Night 64 73.0 5.6 65.9 4.9 104 79.5 3.2 69.4 2.8
SPRAYING**
Day 70 158.3 12.0 138.5 10.6 101 143.1 10.5 121.6 8.9
Night 55 81.5 5.9 69.7 5.5 113 74.6 4.1 67.1 3.5
SMOKE6
Day 61 155.2 15.3 131.4 15.9 110 146.8 7.7 127.3 6.3
Night 29 104.2* 8.0 96.6* 6.7 139 71.4 3.3 63.3 2.7
EXHAUSTf
Day 31 96.5 8.7 85.5 7.0 121 166.3* 9.0 144.8* 8.1
Night 7 - - - - 140 77.5 3.8 68.2 3.1
a Asterisks indicate that the means or geometric means for persons with and without the activity are
significantly different (significantc level ^ 0.05). The asterisk is shown beside the higher member of the
pair. Means, geometric means, and their standard errors were calculated using weighted data: they apply
to subsets of the population of person-days.
Based on responses to question C1 a of the Questionnaire.
0 Based on responses to questions C2a, b, c, h, and i of the Questionnaire.
d Based on responses to question C2, part 1 of the Questionnaire.
® Based on responses to Time and Activity Diary.
Based on responses to question C7 of the Questionnaire.
9-48
-------
SECTION 10
STATISTICAL ANALYSIS OF ELEMENTAL DATA
INTRODUCTION
All PEM, SIM, SAM and diehotomous filters used for obtaining the aerosol
concentrations, via gravimetric analysis, were also subjected to elemental analysis (EA) by
X-ray fluorescence (XRF) to determine the concentration of selected elements in the collected
particles. Elemental data are useful for both source-receptor modeling and directly measuring
the level of exposure to potentially toxic metals. Important toxic metals, such as Pb and As,
and marker elements for environmental tobacco smoke (Cd), oil combustion (V), automobile
emissions (Br), crustals/soils (Si, Al, Ca), woodburning (K), coal combustion (Se), and
industrial and other sources (Fe, Mn, Ni) are of primary interest as targets in the collected
particle samples. The suitability of XRF was first determined by evaluating the range of
measurable concentrations for each element on a subset of filters. After the suitability of XRF
as the analysis method was confirmed, all sample filters were put into a strict order for
analysis to randomize the distribution of each filter type across the analysis time of several
months. All of the approximately 2500 filters were analyzed once by XRF at the EPA facility
in the Research Triangle Park, A subset of filters was subjected to a second analysis to
evaluate analysis precision and the potential for bias over analysis time. A second subset of
filters was analyzed at the Lawerence Berkeley Laboratories (LBL) for quality assurance
purposes. A description of these activities and analysis results are presented in this section.
PRELIMINARY ANALYSES
Method Selection
X-ray fluorescence was thought to be the best method for measuring the elemental
concentrations of primary interest for the large number of filters in a reasonable time and at a
reasonable cost. However, very small air sample sizes were collected during this study and it
was not clear that measurable concentrations of the primary elements would be present on
enough samples to allow further source modeling and statistical analyses of exposures. The
suitability of the XRF method was examined before launching into a full-scale analysis of all
sample filters.
10-1
-------
A small subset of sample filters was selected to evaluate the range of measurable
elemental concentrations by XRF analysis at two laboratories using different XRF techniques.
All PEM, SIM, and SAM samples were first ranked by the total mass of particulate collected
on the sample filter. These rankings were then divided into percentiles of collected mass.
Three filters were selected from each of the 10th, 20th, 30th, 40th, 50th, 60th, 75th, and 90th
percentiles of collected mass. An attempt was made to distribute the filters across sample
type, date collected, and collection time by night or day. An additional collocated sample was
selected for samples chosen at the 10th and 50th percentiles. This selection scheme was
designed to elicit information about the distribution of measurable elemental concentrations
across all of the samples collected during the study based on the mass of collected particles,
and the potential impact of non-measurable values on future modeling and statistical analysis.
The filters were first analyzed at an independent laboratory by wavelength-dispersive
XRF (WD-XRF) using a protocol that was most sensitive for the widest range of elements.
This analysis technique is very time intensive, requiring several hours for each filter analysis.
Each of the filters analyzed by WD-XRF was then analyzed by energy-dispersive XRF (ED-
XRF) at the EPA facility in the Research Triangle Park. This analysis technique is much
faster than the wavelength-dispersive method but is also somewhat less sensitive. Elemental
concentrations were considered measurable if the reported concentrations were greater than
three times the reported uncertainty in the measurement. Percent measurable results for both
laboratories are presented in Table 10-1 ranked by the percentile of the total mass of
particulate collected on the filter. Ten of the thirteen elements of interest had percent
measurable values above 60% at both laboratories. Arsenic and cadmium were not detected
on any filters by either laboratory. Vanadium was not detected by ED-XRF, but with the more
sensitive WD-XRF method it was detected on 81 % of the filters.
The WD-XRF analysis method yielded significantly more measurable values for
vanadium, and to a lesser extent, lead. The improved information content had to be balanced
against the much higher cost and time required for analysis using the WD-XRF method.
These constraints made it impractical for all 2500 sample filters to be analyzed by the
WD-XRF method. Other elemental analysis methods sensitive enough to measure cadmium
and arsenic would likewise be prohibitively expensive and allow only a fraction of the total
samples to be analyzed. Energy-dispersive XRF analysis at the EPA facility was chosen to
provide the most elemental data for the largest number of samples. Additional analyses were
10-2
-------
a'tilc
atcl
9
9
10
11
19
19
19
30
30
31
38
40
41
48
49
51
52
60
60
62
74
74
76
90
90
91
%
TABLE 10-1. COMPARISON OF PERCENT MEASURABLES FOR 10 SAMPLES BETWEEN ENERGY-DISPERSIVE AND
WAVELENGTH-DISPERSIVE XRF ANALYSES
A1 Si S K
ED- WD- ED- WD- ED- WD- ED- WD-
XRF XRF XRF XRF XRF XRF XRF XRF
Ca V Mn Fe
ED- WD- ED- WD- ED* WD- ED- WD-
XRF XRF XRF XRF XRF XRF XRF XRF
Zn As Br Cd Pb
ED- WD- ED- WD- ED- WD- ED- WD- ED- WD-
XRF XRF XRF XRF XRF XRF XRF XRF XRF XRF
+
+
+
+
+
4-
+
+
+
+
+
+
+
+
+
4-
4-
4-
+
+
4-
+
4-
4-
+
+
4-
•fr
+
+
+
+
+
¦f
4-
+
+
4-
4-
4-
4-
4-
~
4-
4-
4-
4-
4-
+
4-
+
4-
+
+
+
¦f
4-
+
4-
+
4-
4-
+
+
+
+
+
4-
+
4-
4-
+
+
+
4*
4-
+
4-
4-
+
+
4-
+
+
+
+
4-
4-
+
+
•f
4-
4-
4-
+
+
+
+
+
+
4-
+
4-
4-
4-
+
+
4-
4-
4-
+
+
4*
4-
+
4-
+
+
4-
4-
4-
4-
4-
4-
4-
4-
4-
4*
4-
4-
+
4-
4-
4-
4-
4-
4-
+
4»
+
4-
4-
4-
4-
4-
81
100 100 100 100 100 100 100 100 100
81
96 100 100 100 100 100
88 100
62 96
-------
conducted to verify the percent measurable results at EPA's facility using subsets of SIM and
SAM 2.5 filters fromthe homes and PEM, SAM, and dichot filters from the temporal site.
All remaining sample filters were then ordered for ED-XRF analysis at the EPA facility.
Collocated pairs of dichotomous and PEM/SAM 10 urn filters were included in the ED-
XRF analyses described above. Correction factors were applied to several of the elements to
compensate for the size and distribution of particles on the sample filters. These correction
factors have been previously determined for dichot samples at the EPA facility. The analysis
results for the collocated filters were used to calculate particle size attenuation correction
factors for the PEM, SIM, and SAM 10 nm samples collected at all locations.
Selection of Sample Analysis Order
All PEM, SIM, and SAM sample filters were arranged in an order for XRF analysis to
ensure that analysis bias over time, if it were to exist, would be distributed evenly across
sample types. Several layers of ordering were imposed on the sample filters, subdividing
each type into groups with specific attributes. The initial ordering is described as follows:
1. Sample Type
PEM, SIM, SAM 10 nm and SIM, SAM 2.5 |im.
2. Week of Collection
Collected during weeks 1 through 7 of field monitoring.
3. Time of Day
Collected during overnight or daytime monitoring period.
4. Catch Mass
High or low particle mass collected on the filter.
This ordering resulted in 140 groups of filters containing 7 to 19 individual sample filters in
each group. Each sample filter was then assigned a random number. The filters within each
group were subdivided into three smaller subgroups, numbered one, two or three. AH of the
number one subgroups were then merged into one large analysis batch; likewise the number
two and number three subgroups were merged together into two more analysis batches. The
filters within each analysis batch were then placed in numerical order based on their random
number assignment. Each analysis batch was then sent to EPA for XRF analysis. Analysis of
10-4
-------
one analysis batch was completed before the next batch was sent. This convoluted filter
ordering ensured that sample filter types would be evenly distributed across analysis time with
a random order of analysis for filters within each batch. Any bias in analysis over time would,
therefore, be distinguishable from other possible trends among the samples (i.e. collection
date, particle mass collected, etc.).
Dichotomous sample filters were analyzed as coarse/fine pairs. The dichot filters were
randomly ordered as pairs and then divided into three analysis batches. These dichot
analysis batches were analyzed separately from the PEM, SIM, SAM batches. Analysis of all
dichot filters in one batch was completed at the EPA facility before the next batch was sent for
analysis.
ELEMENTAL ANALYSES QUALITY CONTROL
Quality Control Sample Description
Quality control samples were collected and analyzed to evaluate the background and
precision of sample collection and elemental analysis. These QC samples included:
1. Filter Blanks
Unused filters, from the same lot(s) as sample filters.
2. Field Blanks
Filters placed into the impactor systems (with greased
impactor plates) and carried to and from field monitoring
sites along with the samples during the field study.
3. Special Field Blanks
Filters placed into impactor systems ( with ungreased
impactor plates) and carried to and from field monitoring
sites along with the samples during the field study.
4. Collocated Samples
Samples collected side-by-side at one location.
5. Duplicate Analyses
Sample filters analyzed a second time at the same facility.
Some filter blanks were analyzed by XRF prior to beginning field monitoring to ensure that
background elemental concentrations were low and that the filters were acceptable. One filter
blank was then included with each set of 35 sample filters analyzed by XRF at the EPA
10-5
-------
facility, for a total of 62 blanks for PEM/SIM/SAMs and 12 blanks for dichots. Fifty-one field
blanks were deployed during the study for PEM/SIM/SAM samples and all were analyzed by
XRF. Eight dichot field blanks were analyzed by XRF. Nine special blanks, similar to field
blanks except that no grease was applied to the impactor plates, were deployed during the
study and these nine filters were also analyzed by XRF. Numerous collocated samples were
collected for all sample types; all were analyzed by XRF. Forty-five PEM/SIM/SAM filters
analyzed in each of the first two XRF analysis batches were analyzed a second time in the
third analysis batch.
Quality control samples were also used to evaluate the accuracy of the elemental
analysis. Two types of quality control samples were used during this study to evaluate
accuracy:
1. NISTSRM
NIST Standard Reference Materials (SRMs) nos. 1833 and
1832, containing known elemental concentrations, were
analyzed along with each set of 35 sample filters.
2. Interlaboratory Duplicate Analyses
Sample filters analyzed first at the EPA facility and then
at Lawrence Berkeley Laboratories (LBL).
The SRM analyses provided a measure of analytical accuracy over time with each group of
samples analyzed. Interlaboratory analyses were conducted by energy-dispersive XRF at LBL
on approximately 100 PEM/SIM/SAM filters and 20 dichot filters. Particle size attenuation
corrections calculated by EPA were applied to the analysis results from LBL. Approximately
26 filters were analyzed by both ED-XRF and WD-XRF, and these interlaboratory results can
also be reviewed, however, different attenuation corrections were applied at the laboratory
using the WD-XRF protocol. The differences in attenuation factors will be most significant for
Al, Si, Ca, K, and CI.
Elemental Analysis Results
This section "provides a descriptive summary of elemental concentrations measured on
residence and temporal site sample filters and the associated quality control (QC) data. The
sample data consist of the PEM, SIM, and SAM measurements associated with participants
and PEM, SAM, and dichot temporal-site measurements. The QC data included EA of
10-6
-------
standard reference material (SRM) samples with "known" amounts of specific elements, as
well as EA of laboratory and field blanks, collocated field samples, duplicate analyses by the
same laboratory and method, and duplicate analyses by different laboratories. In summarizing
temporal-site or residence concentration data, averages of duplicate XRF analyses are
computed (whenever they occur), then averages of collocated samples are taken (whenever
they occur) and treated as a single observation. As an example, suppose that a given
participant's indoor air for a daytime sample was designated to receive collocated SIMs, and
that one of these samples was subjected to XRF twice. This would produce observed
concentrations X1 and Xg for a given element. The SIM concentration for that time period
and household would thus be determined as
[0.5 (Xi * X2) * Y\ / 2 , (Eq. 10-1)
where V denotes the elemental concentration for the collocated sample.
The XRF analyses provided data for 42 specific elements. These 42 elements
consisted of 13 primary elements designated as of a priori interest due to knowledge
concerning their most significant sources:
Primary Sources Elements
Crustals/soils: Si, Al, Ca
Industrial: Fe, Mn, Ni
Combustion -
Wood: K
Coal: Se
Oil: V
Vehicle Emissions Br
Tobacco: CD
Toxic Metals: Pb, As
Percentage Measurable
Eight of the 13 primary elements were generally found in measurable quantities in the
residence samples (PEM/SIM/SAM samples), and, in particular, were measurable in over 50%
of the daytime or nighttime personal (PEM) PM^q samples. Of the 29 secondary elements,
seven met this same criterion. The classification of the elements is depicted below:
10-7
-------
Percent Measurable
High
Primary
Secondary
Si, Al, Ca, Fe,
Mn, K, Br, Pb
S, Zn, CI, Ti,
Cu, Sr, P
Low
Ni, SE, V, Cd,
As
Cr, Ba, Rb, Sn, Zr,
La, Co, Ga, Y, Zu, Te,
Hg, W, Sb, Ag, Ge, I,
Cs, Mo, Rh, Pd, Sc
"High percent measurable" elements are defined as those that exhibited over 50% measurable
for either the daytime or nighttime PEM PM10 samples, where the measurability threshold is
defined as three times the reported uncertainty in the XRF measurement. This threshold level
is referred to as an uncertainty limit (UL). Uncertainty limits were associated with each
reported element concentration and were dependent on the magnitude of the concentration.
Specific percentages of the samples in which the elemental concentrations are above
the UL are given in Tables 10-2 through 10-5:
Table 10-2: Primary elements - residence data
Table 10-2: Secondary elements ~ residence data
Table 10-4: Primary elements - temporal site data
Table 10-5: Secondary elements - temporal site data.
Results are reported by time of day and by type of sample in each of these tables. The
residence results are weighted to reflect the populations of individuals (PEM samples) or
households (SIM or SAM samples).
For the most part, results presented in subsequent subsections are confined to either
the primary elements, the hiah-percent-measurable elements, or the combination of these two
groups.
For the 10 nm samples, seven elements were almost always present in measurable
amounts: Si, Ca, Fe, K, S, Zn, and CI. The following 21 elements were almost never present
in measurable amounts (less than or equal to 10 percent measurable for all ten cases in
Tables 10-2 or 10-3): Se, V, Cd, As, La, Co, Ga, Y, Au, Te, Hg, W, Sb, Ag, GE, I, Cs, Mo,
Rh, Pd, and Sc. As expected (due to the higher flow rate), the dichot samplers generally
produced higher percentage measurable values than the corresponding PEMs and SAMs at
the temporal site. The major exception to this was the 2.5 ^im potassium values.
10-8
-------
TABLE 10-2. WEIGHTED ESTIMATES OF PERCENT MEASURABLE FOR INDIVIDUAL (PEM) AND HOUSEHOLD
(SIM AND SAM) POPULATIONS, BY TIME OF DAY: PRIMARY ELEMENTS
Daytime Nighttime
PM25 PM10 PM2.5 ^10
Element SAM SIM SAM SIM PEM SAM SIM SAM SIM PEM
Si
95.7
92.1
100.0
100.0
100.0
83.0
88.1
100.0
100.0
98.2
Al
7.0
5.9
79.3
53.8
77.4
1.3
0.0
62.4
30.2
30.1
Ca
99.5
100.0
100.0
100.0
100.0
99.2
100.0
100.0
100.0
98.2
Fe
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
98.2
Mn
55.3
36.0
99.5
90.6
98.8
50.8
33.6
98.9
81.4
89.2
Ni
0.3
0.9
2.1
4.3
20.2
3.4
0.8
5.7
6.0
7.5
K
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
98.2
Se
1.0
0.8
2.6
0.9
0.4
10.0
3.6
9.7
2.5
1.5
V
0.6
0.5
2.3
2.5
6.3
0.5
1.5
6.5
0.6
1.7
Br
51.6
53.2
61.1
72.0
86.8
65.1
52.6
79.1
75.9
85.3
Cd
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Pb
46.3
35.6
74.7
63.7
75.5
61.8
47.7
80.7
65.9
59.2
As
0.0
0.0
0.0
0.3
2.2
0.7
0.0
0.0
0.0
0.0
-------
TABLE 10-3. WEIGHTED ESTIMATES OF PERCENT MEASURABLE FOR INDIVIDUAL (PEM) AND HOUSEHOLD
(SIM AND SAM) POPULATIONS, BY TIME OF DAY: SECONDARY ELEMENTS
Daytime
Nighttime
Element
pm25
SAM SIM
SAM
PM10
SIM
PEM
PM2.5
SAM SIM
SAM
PM10
SIM
PEM
s
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
98.2
Zn
97.9
98.5
100.0
100.0
100.0
96.0
96.6
99.4
100.0
98.2
CI
48.3
63.9
92.5
99.1
100.0
70.3
65.5
97.8
98.6
96.4
Ti
8.5
13.8
97.5
87.1
98.3
3.2
3.7
92.6
79.2
85.7
Cu
21.9
38.8
71.7
79.3
94.3
33.1
35.2
78.7
70.8
78.2
Sr
18.5
15.5
92.5
78.5
91.6
13.8
11.7
91.0
65.4
73.0
P
0.5
0.5
7.6
27.1
55.6
0.9
1.6
12.4
16.1
23.7
Cr
2.4
5.2
14.2
22.0
38.7
0.0
2.1
10.1
12.5
16.4
Ba
4.9
9.4
18.9
20.3
36.0
7.9
1.8
20.3
12.5
11.2
Rb
0.9
0.0
18.0
17.1
37.7
0.6
0.4
16.8
6.9
3.8
Sn
3.6
11.4
4.6
7.8
9.3
10.7
6.9
11.8
4.4
12.8
Zr
0.0
0.0
0.0
1.7
15.1
0.0
0.5
0.4
2.5
4.7
La
1.3
0.6
1.1
0.0
1.0
0.0
0.4
1.6
1.7
1.2
Co
0.0
0.0
0.0
0.0
1.6
0.0
0.0
0.0
0.0
0.5
Ga
0.6
0.0
1.8
0.8
1.1
0.0
0.6
0.0
0.5
0.5
Y
0.0
0.6
0.0
0.0
0.0
0.5
0.0
0.0
0.0
1.5
Au
0.0
0.4
0.0
0.4
0.9
0.0
0.0
2.4
0.6
0.6
Te
0.0
0.0
0.0
0.4
1.3
0.0
0.0
0.0
0.0
0.0
Hg
0.0
0.0
0.0
0.0
0.5
0.0
0.0
0.0
0.0
0.6
W
0.0
0.0
0.0
0.0
1.0
0.0
0.0
0.0
0.0
0.0
Sb
0.5
0.5
1.2
0.0
0.4
0.0
0.0
1.1
1.5
0.5
Ag
0.0
0.0
0.0
0.0
0.6
0.0
0.5
0.0
0.0
0.0
Ge
2.1
0.5
0.0
1.2
0.0
0.0
0.0
0.5
2.0
0.6
I
0.0
1.0
0.4
0.5
0.4
0.5
2.5
0.5
0.6
0.0
Cs
0.0
0.0
0.8
0.0
0.4
0.5
0.0
0.3
0.6
0.0
Mo
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Rh
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Pd
0.0
0.0
0.7
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Sc
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
-------
TABLE 10-4. ESTIMATES OF PERCENT MEASURABLE FOR TEMPORAL SITE SAMPLES,
BY TIME OF DAY: PRIMARY ELEMENTS
Daytime Nighttime
s PM10 EM10_
Element SAM Dichot PEM SAM Dichot SAM Dichot PEM SAM Dichot
Si
93.3
100.0
100.0
100.0
100.0
65.9
100.0
100.0
100.0
100.0
Al
8.9
100.0
73.8
73.3
100.0
2.3
87.5
35.6
37.8
93.7
Ca
100.0
100.0
100.0
100.0
100.0
97.7
100.0
100.0
100.0
100.0
Fe
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
Mn
57.8
93.5
100.0
97.8
100.0
34.1
93.7
93.3
95.6
100.0
Ni
0.0
0.0
7.1
2.2
26.1
6.8
0.0
8.9
4.4
45.8
K
97.8
8.7
100.0
100.0
100.0
100.0
4.2
100.0
100.0
100.0
Se
4.4
0.0
0.0
2.2
43.5
6.8
0.0
4.4
4.4
54.2
V
0.0
0.0
14.3
11.1
34.8
0.0
0.0
6.7
2.2
54.2
Br
53.3
95.7
61.9
68.9
93.5
54.5
100.0
71.1
71.1
95.8
Cd
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Pb
55.6
100.0
78.6
73.3
97.8
47.7
100.0
57.8
60.0
97.9
As
2.2
6.5
2.4
0.0
6.5
0.0
0.0
0.0
0.0
0.0
-------
TABLE 10-5. ESTIMATES OF PERCENT MEASURABLE FOR TEMPORAL
SiTE SAMPLES, BY TIME OF DAY: SECONDARY ELEMENTS
Daytime Nighttime
*5 PM10 __£Ma.5 EM10.
Element SAM Dichot PEM SAM Dichot SAM Dichot PEM SAM Dichot
s
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
Zn
91.1
100.0
100.0
97.8
100.0
97.7
100.0
100.0
100.0
100.0
CI
53.3
93.5
100.0
91.1
100.0
70.5
100.0
97.8
97.8
100.0
Ti
17.8
45.7
95.2
93.3
97.8
6.8
33.3
82.2
88.9
100.0
Cu
93.3
100.0
100.0
100.0
100.0
97.7
100.0
97.8
100.0
100.0
Sr
24.4
52.2
88.1
82.2
97.8
11.4
45.8
84.4
80.0
100.0
P
0.0
4.3
7.1
2.2
4.3
0.0
0.0
4.4
0.0
0.0
Cr
0.0
10.9
7.1
17.8
63.0
0.0
22.9
4.4
2.2
66.7
Ba
11.1
19.6
26.2
26.7
78.3
13.6
16.7
17.8
11.1
64.6
Rb
0.0
6.5
14.3
17.8
84.8
0.0
0.0
0.0
4.4
62.5
Sn
2.2
19.6
2.4
2.2
37.0
2.3
20.8
13.3
2.2
27.1
Is
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
La
0.0
0.0
2.4
0.0
4.3
0.0
0.0
2.2
0.0
0.0
Co
0.0
8.7
0.0
0.0
8.7
0.0
4.2
0.0
0.0
4.2
Ga
0.0
0.0
4.8
2.2
10.9
0.0
0.0
0.0
0.0
0.0
Y
0.0
0.0
0.0
0.0
0.0
2.3
0.0
0.0
0.0
0.0
Au
0.0
0.0
0.0
0.0
0.0
0.0
2.1
0,0
0.0
2.1
Te
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Hg
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
W
0.0
0.0
0.0
0.0
0.0
0,0
0.0
0.0
0.0
0.0
Sb
0.0
2.2
0.0
2.2
8.7
0.0
2.1
0.0
0.0
4.2
Ag
0.0
2.2
0.0
0.0
2.2
0.0
0.0
0.0
0.0
0.0
Ge
2.2
2.2
0.0
0.0
2.2
0.0
4.2
0.0
0.0
6.2
I
2.2
2.2
0.0
0.0
4.3
0.0
2.1
0.0
0.0
2.1
Cs
2.2
0.0
0.0
0.0
2.2
0.0
0.0
2.2
0.0
0.0
Mo
0.0
0.0
0.0
0.0
0.0
0.0
0.0
2.2
0.0
0.0
Rh
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Pd
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Sc
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0,0
-------
Standard Reference Material Data
Tables 10-6 and 10-7 summarize the XRF results for the SRM samples. Table 10-6
indicates the results for those cases in which the "known" concentration level of the element
was zero on the SRM (SRM blanks), while Table 10-7 presents results for those cases in
which the "known" concentration was positive. For most of the SRM blank analyses, the
observed concentrations for Fe, Mn, K, V, Zn, Cu, and Co exceeded their ULs (i.e., would
have been declared measurable). This was evident from plots of the concentration data and
the associated ULs, and is also apparent from the results in Table 10-5. It is important to
note that only Fe typically exceeded the ULs for lab and field blanks. With the possible
exception of zinc, there did not appear to be major trends over analysis time in the observed
concentrations for these blanks; for zinc, a possible increasing trend in levels was observed.
For the spiked SRMS (Table 10-7), there was generally a small range in the percentage bias,
which was computed as
100% x (observed - true) / true. {Eq. 10-2)
Plots of % bias quantities over times of analysis, however, revealed a systematic trend
for many of the elements; in particular, negative trends were apparent for Ca, Fe, Mn, V, Pb,
Zn, and Cu (and possibly for cobalt). These trends are illustrated in Figures 10-1 and 10-2 for
Ca and Pb. Plots for the other 11 elements are presented in Appendix P. Percent biases
observed for calcium during analysis of sample batches A and B were generally in the range
of 7% to 8%. The bias during analysis of batch C samples was lower at 5% to 6.5%. A
similar trend was observed for lead, but the overall biases were closer to 0%. The trends
suggest that instrument performance changed over time and elemental concentrations
reported for samples analyzed in batch C may be slightly lower than those determined for the
two earlier batches, A and B. However, the magnitude of the differences are small and almost
always less than the uncertainty calculated for each measurement. Overall, the XRF
analytical system accuracy appeared to be very good. All median percent biases were 7% or
less with a range from -8% to 10%.
Blanks
Laboratory and field blanks prepared by RTI were subjected to XRF analyses along
with the other samples. The results of the elemental analyses for these samples are
10-13
-------
TABLE 10-6. SUMMARY OF RESULTS FOR STANDARD
REFERENCE MATERIAL BLANKS
Observed Concentrations" Range of ULs"
Type of
Sample
Element
Minimum
Median
Maximum
Minimum
Maximum
PEM/SIM/S
Al
-220
970
900
1200
1300
AM (n=36)
Ca
-10
32
72
56
70
Fe
160
240
310
87
110
Mn
29
53
65
18
22
K
34
97
150
54
66
V
100
160
210
78
92
Pb
-15
12
30
22
26
Zn
25
43
61
20
26
Ti
-100
-22
8.1
28
67
Cu
77
96
110
30
38
Co
73
120
190
80
95
Dichot
Al
79
190
460
260
290
(n=8)
Ca
-0.6
9.1
17
12
16
Fe
42
53
66
20
24
Mn
8.9
11
13
4.1
4.6
K
12
17
24
12
14
V
28
34
40
18
19
Pb
-0.3
1.2
3.9
4.7
5.2
Zn
3.3
11
13
4.1
5.6
Ti
-19
-9.9
-5.5
5.9
14
Cu
18
21
24
6.8
8.2
Co
13
21
28
17
19
a All results are reported in ng/m3, based on a nominal flow rate and sample duration.
10-14
-------
TABLE 10-7. SUMMARY OF RESULTS FOR STANDARD REFERENCE MATERIAL
SAMPLES HAVING NON-ZERO LEVELS
3\a
Type of Sample Element True Cone, (ng/nr)
% Bias in Observed Concentrations
Minimum
Median
Maximum
PEM/SIM/SAM
(n= 36)
Dichot
(n=8)
Sib
76000
+
2600
-5.3
-2.1
0.8
Si0
77000
±
4900
-1.4
0.9
2.6
At
32000
2100
-6.0
3.2
9.9
Ca
47000
±
3100
5.1
7.0
8,6
Fe
35000
±
1000
-4.6
-1.8
2.2
Mn
11000
±
1200
1.1
2.8
8.0
K
43000
+
4200
-0.9
0.3
1.5
V
12000
+
1200
3.2
5.3
7.5
Pb
34000
+
1800
-1.7
0.8
5.2
Zn
8000
+,
760
-8.4
-5.5
-1.0
Ti
35000
±
4700
1.5
3.2
4.3
Cu
5500
±
360
-7.0
-4.1
1.1
Co
2400
+
160
-2.3
1.7
10.4
Sib
16000
±
570
-4.4
-2.0
0.1
Si°
17000
+
1100
-2.3
0.5
3.6
AI
7000
+
450
-4.2
3.6
5.6
Ca
10000
±
680
5.9
7.1
8.6
Fe
7600
±
230
-3.4
-1.0
1.4
Mn
2400
+
260
1.5
3.0
6.2
K
9500
+
910
-0.3
0.3
1.2
V
2500
+
260
3.7
5.4
7.1
Pb
7300
+
400
-1.3
1.1
3.8
Zn
1700
+
160
-7.7
-4.4
-2.7
Ti
7500
+
1000
2.3
3.1
4.5
Cu
1200
+
79
-6.1
-3.0
1.1
Co
520
+
34
-1.7
3.5
6.6
a True and observed concentrations (and the uncertainties) were developed from the loaded amounts
by multiplying by 2.604 for the PEM/SIM/SAM samples and by 0.566 for the dichot samples. These
reflect nominal flow rates and sample durations.
b Silicon concentration in SRM 1832.
c Silicon concentration in SRM 1833.
10-15
-------
6.0 +
5.5 +
5.0 +
A
PCDIF
9.0 +
8.5 +
8.0 +
7.5 + A B
A A
A B
B B
B
7.0 + B B B
A
A A
6.5 +
C
C C
C
C
I
4 -~—»
01APR1991 01MAY1991 01JUN1991 01JU11991 01AUG1991
XRFOATE
Figure 10-1. Percent Bias for Calcium in SRM 1832
by XRF Analysis Over Time. Associated
Analysis Batch Designated by A, B, or C
10-16
-------
PCDIF |
5 +
4 +
3
2
B B
C C
1 +
0 +
-1 +
A
AA
A
A A
A
-2
1 .+
01APR1991
B B
B B
8
01JUN1991
XRFDATE
01JUL1991 01AU61991
01MAY1991
Figure 10-2. Percent Bias for Lead in SRM 1833 by XRF
Analysis Over Time. Associated Analysis Batch
Designated by A, B, or C
10-17
-------
summarized in Table 10-8 for the 13 elements of primary interest and 7 other elements with
high percentage measurable values in the samples. Means and standard deviations of both
the observed concentrations and the associated ULs are presented. Using an approximate t-
test, field blank concentrations were significantly higher than lab blank concentrations for Si,
S, and Zn. In each of these cases the observed means were below their respective mean
ULs. This was also true for the other elements, except for iron. Special blanks were
equivalent to field blanks except the impactor plates with which they were associated were not
greased. Table 10-9 gives similar results for dichot samples undergoing XRF. Due to the
small sample sizes, no tests of significance for differences between dichot lab and field blanks
were performed.
In general, elemental background concentrations were very low for the 13 elements of
primary interest and 7 secondary elements with high percentage measurable values on the
samples. Only mean iron concentrations were greater than the mean uncertainty limit. In the
case of iron, lab blank concentrations were higher than either the field blank or special blank
concentrations, suggesting that contamination did not occur during filter handling and may
have been lowered by the tapping procedure prior to weighing (the lab blank filters were not
weighed). Iron concentrations on the field blanks were 4 to 100 times lower than the mean
sample iron concentrations.
Collocated Samples
Numbers of collocated sample pairs, by particle size and type, are given below:
Locations of Sampling
Size
Type
Sample Homes
Temporal Project
Site Staff
Cut (urn)
2.5
SIM
SAM
Dichot
SIM
SAM
PEM
Dichot
15
17
3
90
10
17
18
4
3
90
15
10-18
-------
TABLE 10-8. SUMMARY OF LABORATORY, FIELD, AND SPECIAL BLANK
ELEMENTAL ANALYSIS RESULTS
Element Type of Blank n Cone. Mean® Cone. Std. Dev. UL Mean UL Std. Dev.
SI
LAB
62
-5.5
52
130
3.8
FIELD
51
22*
85
190
84
SPECIAL
9
2.3
100
170
37
AL
LAB
62
-0.57
130
370
12
FIELD
51
-2.4
230
590
290
SPECIAL
9
-5.1
190
530
120
CA
LAB
62
-7.4
14
32
0.87
FIELD
51
-6.8
18
32
2.1
SPECIAL
9
-12
5.1
32
0.57
FE
LAB
62
40
100
24
23
FIELD
51
28
47
21
13
SPECIAL
9
7.6
16
16
2.3
MN
LAB
62
0.77
2.9
7.5
0.41
FIELD
51
0.26
3.4
8.0
2.9
SPECIAL
9
-0.03
3.1
7.3
0.39
Nt
LAB
62
-0.82
2.5
7.5
0.41
FIELD
51
-0.83
2.6
8.0
2.9
SPECIAL
9
-2.1
3.9
7.3
0.55
K
LAB
62
2.1
11
29
0.96
FIELD
51
4.3
15
31
3.6
SPECIAL
9
6.4
7.4
30
0.78
SE
LAB
62
-0.27
1.6
4.3
0.39
FIELD
51
0.01
1.8
4.6
1.7
SPECIAL
9
-0.26
1.2
4.3
0.41
V
LAB
62
-1.7
4.1
17
0.49
FIELD
51
-1.2
6.2
18
2.3
SPECIAL
9
0.00
2.0
17
0.41
BR
LAB
62
0.29
1.9
5.5
0.23
FIELD
51
0.54
1.9
6.0
2.1
SPECIAL
9
-0.26
1.6
5.5
0.00
CD
LAB
62
-3.7
11
35
1.2
FIELD
51
0.34
13
36
7.3
SPECIAL
9
-8.1
13
34
1.2
PB
LAB
62
0.80
4.6
14
0.45
FIELD
51
0.73
5.6
15
5.2
SPECIAL
9
3.6
4.1
15
0.57
10-19
(continued)
-------
TABLE 10-8. (continued)
lement
Type of Blank
n
Cone. Mean8
Cone. Std. Dev.
UL Mean
UL Std. Dev.
AS
LAB
62
-0.74
2.7
7.8
0.26
FIELD
51
-1.1
2.7
8.4
2.9
SPECIAL
9
-2.3
2.0
7.6
0.34
S
LAB
62
'7-2 k
15
49
1.2
FIELD
51
-0.52
18
51
13
SPECIAL
9
-5.2
15
48
1.6
ZN
LAB
62
0.09
3.0
6.0
0.45
FIELD
51
4#
5.6
7.0
2.4
SPECIAL
9
-0.72
2.5
5.8
0.57
CL
LAB
62
1.4
11
36
1.04
FIELD
51
2.7
19
38
13
SPECIAL
9
6.9
9.5
35
0.94
Tl
LAB
62
-0.67
10
51
1.4
FIELD
51
-1.3
14
50
4.5
SPECIAL
9
0.64
2.6
50
0.96
CU
LAB
62
3.9
6.1
7.9
0.93
FIELD
51
2.3
5.2
8.2
2.8
SPECIAL
9
4.0
3.9
7.8
0.55
SR
LAB
62
1.9
2.1
6.4
0.31
FIELD
51
2.0
2.5
6.9
2.4
SPECIAL
9
2.5
2.8
6.5
0.39
P
LAB
62
-13
27
78
2.6
FIELD
51
-16
33
84
30
SPECIAL
9
-14
31
76
1.8
a Concentrations and uncertainty limits (Uls) are expressed in ng/m3 units, based on a nominal flow
rate and sample duration.
b Field blank significantly different from the lab blank.
10-20
-------
TABLE 10-9. SUMMARY OF DICHOT BLANK ELEMENTAL ANALYSIS RESULTS
Cone. Std.
Element Type of Blank® Cone. Meanb Dev.b UL Meanb UL Std. Dev.b
SI
LAB
-2.2
12
38
10
FIELD
-11
8.9
38
12
AL
LAB
3.8
31
110
32
FIELD
-34
40
110
37
CA
LAB
-1.4
2.9
7.0
0.23
FIELD
-2.0
1.9
6.9
0.21
FE
LAB
4.6
8.4
4.4
1.5
FIELD
1.4
3.2
3.7
0.41
MN
LAB
0.38
0.62
1.7
0.12
FIELD
0.14
0.53
1.7
0.14
Nl
LAB
-0.34
0.74
1.6
0.13
FIELD
-0.20
0.33
1.7
0.14
K
LAB
0.64
2.1
6.6
0.24
FIELD
-0.18
2.2
6.7
0.45
SE
LAB
-0.05
0.28
0.93
0.09
FIELD
-0.11
0.36
0.93
0.10
V
LAB
0.04
0.78
3.8
0.12
FIELD
-0.53
0.50
3.8
0.15
BR
LAB
0.05
0.41
1.2
0.07
FIELD
-0.19
0.57
1.2
0.00
CD
LAB
0.07
2.6
7.5
0.39
FIELD
1.1
2.9
7.7
0.19
PB
LAB
0.20
0.94
3.0
0.15
FIELD
-0.25
1.2
3.0
0.09
AS
LAB
-0.45
0.60
1.7
0.10
FIELD
-0.02
0.57
1.7
0.00
S
LAB
-1.4
4.9
11
0.75
FIELD
-1.5
3.6
11
1.0
ZN
LAB
-0.11
0.39
1.3
0.10
FIELD
-0.06
0.38
1.3
0.05
CL
LAB
1.3
3.6
8.5
0.71
FIELD
0.13
2.2
8.4
0.94
Tl
LAB
0.36
0.76
11
0.42
FIELD
-0.93
2.6
11
0.23
10-21
(continued)
-------
TABLE 10-9. (continued)
Cone. Std.
Element Type of Blank3 Cone. Mean® Dev. UL Mean" UL Std. Dev.
cu
LAB
0.61
0.39
1.7
0.12
FIELD
0.96
0.91
1.8
0.12
SR
LAB
0.17
0.60
1.4
0.10
FIELD
0.40
0.50
1.4
0.08
P
LAB
-0.88
6.9
22
4.9
FIELD
-4.8
8.0
22
5.2
® Sample sizes are12 and 8 respectively for lab and field blanks
Units are in ng/m , based on a nominal flow rate and sample duration, (UL = uncertainty limits).
10-22
-------
Table 10-10 provides summary statistics characterizing the relative standard deviations
for each of the above ten types of collocated samples. For those pairs with both values as
measurables, the table gives the number of such pairs, the mean RSD, the median RSD, the
maximum RSD, and the standard deviations of the RSDs. The results are presented for those
15 elements with high percentage measurable values. For most types of samples and most
elements in which reasonable sample sizes were attained, the median RSDs were less than
15% and in many cases (especially for the 10 \im samples), they were less than 5%. The
major departure from this occurred for copper in the dichot 10 urn samples, where a median
RSD of 26% was noted, and for the SAM 2.5 |im samples.
Since members of each pair of dichot samples are distinguishable by sampler (B or C),
the direction of concentration differences between the two can be Investigated to determine if
one sampler consistently or usually yielded larger values than the other. Table 10-11 provides
a summary of the distribution of percentage differences between the B and C samplers.
Minimum, median, and maximum percentage differences are given separately for the 2.5 urn
and 10 nm size cuts. The results indicate that the median percentage difference for the
10 Jim samples was less than 10% for 14 of the 15 elements and was generally positive;
copper had a median percent difference of -33.9%, however. Copper concentrations showed
a median percent difference of -40.8% for the 2.5 jim samples (i.e., generally a lower Cu
concentration for Sampler C than Sampler B). No plausible explanation was found for the
differing copper concentrations measured in the two dichot samplers. It has been surmised
that a source of copper contamination, possibly from wear of brass fittings, was present in
Sampler B.
Duplicate Analyses
Three basic types of duplicate elemental analyses were carried out for QC purposes.
The first type involved EPA's reanalysis of the same filter by XRF. Prior to EA, filters were
first assigned to one of three batches: A, B, or C. Filters in Batch A were analyzed first by
XRF, then those in Batch B, and finally those in Batch C. A subset of 90 of the filters in the
first two batches was randomly re-assigned to Batch C (44 from Batch A and 46 from
Batch B). Tables 10-12 and 10-13 summarize the resultant data. Table 10-12 shows
statistics that summarize the RSDs between the paired analyses. These provide an indication
of the degree to which discrepant concentrations occurred, but do not indicate if there is a
10-23
-------
TABLE 10-10. SUMMARY OF ELEMENT RELATIVE STANDARD
DEVIATIONS FOR COLLOCATED SAMPLES
Sample Cut Size and Type
2.5 fim SIM
2.5 SAM
No. of Pairs
Element (Both Meas.) Mean
Si
13
11.2
Al
0
Ca
15
7.6
Fe
15
9.0
Mn
5
4.3
K
15
8.9
Br
6
17.3
Pb
5
9.0
S
15
4.5
Zn
14
21.5
CI
3
222
Ti
0
Cu
4
33.4
Sr
0
P
0
Si
15
20.1
Al
1
45.0
Ca
16
13.3
Fe
17
19.2
Mn
10
17.9
K
17
13.4
Br
12
17.9
Pb
9
9.8
S
17
4.1
Zn
17
17.4
CI
10
24.5
Ti
1
9.6
Cu
2
20.9
Sr
1
4.7
P
0
RSD Distribution (%)
Median Maximum Std. Dev.
6.7
59.5
15.3
3.3
49.2
12.3
7.8
22.6
6.6
4.6
5.9
1.3
5.5
34.9
9.5
12.1
41.4
14.4
7.8
17.0
5.8
2.7
12.8
4.0
16.2
67.4
19.0
15.5
49.6
24.8
27.7
70.3
26.9
20.7
48.4
15.0
45.0
45.0
6.8
39.1
132
12.5
47.2
16.2
16.9
40.0
12.5
9.3
45.1
13.1
18.9
36.4
10.4
9.0
24.2
5.9
3.2
13.4
3.6
12.8
94.3
21.9
24.3
58.6
20.6
9.6
9.6
20.9
24.3
4.7
4.7
4.7
10-24
(continued)
-------
TABLE 10-10. (continued)
RSD Distribution (%)
Sample Cut Size and Type
Element
no. or rairs -
(Both Meas.)
Mean
Median
Maximum
Std. Dev.
2.5 |i.m SAM (Temporal Site)
Si
3
22.5
10.9
47.0
21.3
Al
0
Ca
3
7.5
6.1
13.9
5.8
Fe
3
5.4
3.7
11.8
5.8
Mn
2
17.0
17.0
17.5
0.8
K
3
5.2
5.9
7.8
3.1
Br
2
4.9
4.9
6.5
2.1
Pb
1
4.7
4.7
4.7
S
3
2.8
3.6
3.6
1.4
Zn
3
9.2
1.6
25.8
14.4
CI
2
36.9
36.9
53.7
23.8
Ti
0
Cu
3
16.9
15.2
29.7
11.9
Sr
0
P
0
2.5 jim Dichot (Temporal Site)
Si
79
12.8
8.3
120.2
15.9
Al
9
24.6
12.1
112.5
34.0
Ca
89
9.2
3.9
114.3
16.2
Fe
90
14.8
9.2
103.8
18.4
Mn
79
12.7
9.8
64.5
12.2
K
90
7.4
3.5
106.8
13.2
Br
80
11.1
8.7
53.8
9.9
Pb
78
11.4
7.6
46.9
10.2
S
90
5.2
3.5
68.0
8.8
Zn
90
12.6
6.0
105.6
16.8
CI
84
19.2
18.5
50.9
12.6
Ti
14
13.2
7.4
59.5
16.5
Cu
89
32.6
29.8
113.5
23.4
Sr
13
9.0
8.3
20.7
5.6
P
0
10-25
(continued)
-------
TABLE 10-10. (continued)
RSD Distribution (%)
No. of Pairs
Sample Cut Size and Type Element (Both Meas.) Mean Median Maximum Std. Dev.
10 jim SIM
10 nm SAM
Si
17
6.0
3.2
37.8
9.0
Al
8
13.5
9.7
39.7
12.3
Ca
17
4.2
2.9
23.0
5.2
Fe
17
7.6
4.2
43.7
10.5
Mn
15
11.3
6.0
62.5
15.2
K
17
5.4
3.8
30.2
6.9
Br
12
13.5
11.7
27.8
9.1
Pb
11
18.5
12.4
50.9
15.5
S
17
7.9
2.5
64.2
15.1
Zn
17
13.7
6.4
59.1
16.2
CL
17
23.5
16.9
70.6
19.3
Tl
12
9.8
10.1
20.3
7.2
Cu
12
20.1
16.7
84.3
22.0
Sr
9
12.3
8.8
29.2
11.1
P
4
12.8
12.1
18.8
5.1
Si
18
4.5
3.5
12.9
3.9
Al
9
9.7
6.3
30.4
8.6
Ca
18
2.6
2.7
6.5
1.7
Fe
18
3.1
2.8
10.5
2.5
Mn
18
9.6
4.9
39.7
10.8
K
18
3.9
2.8
9.7
2.6
Br
11
14.6
13.0
40.4
11.3
Pb
12
12.9
11.8
23.3
7.4
S
18
4.5
3.7
18.9
4.1
Zn
18
9.3
6.9
32.3
9.1
CI
18
23.3
15.1
85.9
24.2
Ti
17
6.4
5.2
30.3
7.9
Cu
10
9.3
4.7
49.8
14.7
Sr
17
13.7
12.7
26.9
7.2
P
0
10-26
(continued)
-------
TABLE 10-10. (continued)
No. of Pairs
RSD Distribution (%)
Sample Cut Size and Type Element (Both Meas.) Mean Median Maximum Std. Dev.
10 (am SAM (Temporal Site)
10 |a.m PEM (Temporal Site)
Si
4
3.4
2.0
8.9
3.8
Al
3
10.4
5.9
20.3
8.5
Ca
4
2.5
2.4
4.1
1.4
Fe
4
2.9
3.1
5.1
2.1
Mn
4
5.6
4.5
12.8
5.8
K
4
3.5
3.5
4.3
0.9
Br
2
26.7
26.7
36.8
14.4
Pb
2
9.5
9.5
18.9
13.2
S
4
2.3
1.1
6.1
2.5
Zn
4
11.8
13.7
19.2
8.1
CI
4
9.9
10.1
16.0
6.0
Ti
4
9.1
6.8
18.6
6.6
Cu
4
16.5
17.2
31.2
13.8
Sr
3
7.5
8.1
12.7
5.6
P
0
Si
3
4.8
2.8
10.6
5.1
Al
3
18.3
20.3
20.7
3.8
Ca
3
4.6
3.0
8.8
3.7
Fe
3
8.6
8.0
17.2
8.3
Mn
3
7.5
9.0
10.1
3.5
K
3
2.8
0.2
8.2
4.6
Br
1
2.6
2.6
2.6
Pb
2
10.2
10.2
12.8
3.7
S
3
3.9
3.3
6.8
2.6
Zn
3
11.0
10.3
12.7
1.4
CI
3
4.8
4.9
6.6
1.8
Ti
3
10.8
2.3
28.7
15.6
Cu
3
15.6
3.7
39.6
20.8
Sr
3
9.9
5.8
22.3
11.0
P
0
10-27
(continued)
-------
TABLE 10-10. (continued)
RSD Distribution (%)
No. of Pairs
Sample Cut Size and Type Element (Both Meas.) Mean Median Maximum Std. Dev.
10 jxm PEM (Project Staff)
10 pim Dichot (Temporal Site)
Si
15
6.1
5.3
17.2
5.1
Al
10
10.6
8.5
25.5
7.8
Ca
15
4.1
2.2
16.5
4.4
Fe
15
6.9
6.6
21.1
5.5
Mn
15
15.5
13.2
43.2
13.5
K
15
4.7
4.1
19.6
4.7
Br
10
14.1
12.3
30.9
8.7
Pb
9
8.1
6.2
27.3
8.3
S
15
4.7
3.9
11.9
3.3
Zn
15
11.0
10.6
26.3
6.7
CI
15
15.1
10.6
44.3
12.1
Ti
15
10.7
9.8
24.8
8.0
Cu
13
9.9
6.4
34.7
10.1
Sr
12
9.2
9.4
27.3
8.3
P
3
3.5
3.8
6.6
3.2
Si
90
9.4
6.7
74.0
12.3
Al
80
12.4
9.9
65.0
11.5
Ca
90
8.5
6.4
63.9
10.4
Fe
90
9,7
6.9
60.5
10.6
Mn
90
12.1
7.8
96.3
14.3
K
90
8.2
5.7
63.8
10.7
Br
80
14.2
10.4
85.6
13.9
Pb
83
17.2
11.2
73.9
16.3
S
90
7.6
5.1
88.3
12.6
Zn
90
10.7
6.0
80.7
15.5
CI
90
12.9
9.1
58.7
12.8
Ti
88
11.1
8.7
68.7
11.5
Cu
90
27.8
26.0
104.8
21.9
Sr
89
15.1
12.0
73.1
13.7
10-28
-------
TABLE 10-11. SUMMARY OF DISTRIBUTIONS OF PERCENTAGE DIFFERENCES
BETWEEN ELEMENTAL CONCENTRATIONS FROM TWO COLLOCATED DICHOT SAMPLERS3
PM25 Distribution of % Differences PM^ Distribution of % Differences
No. of Pairs Median % No. of Pairs Median %
Element (Both Meas.) Min. % Diff. Diff. Max. % Diff (Both Meas.) Min % Diff. Diff. Max % Diff.
Si 79 -53.1
Al 9 -40.7
Ca 89 -103.8
Fe 90 -85.3
Mn 79 -63.2
K 90 -48.8
Br 80 -76.1
Pb 78 -66.4
S 90 -64.5
Zn 90 -149.3
CI 84 -71.9
Ti 14 -84.1
Cu 89 -160.5
Sr 13 -29.3
P 0
-0.5 170.0 90
-13.9 158.5 80
0.1 161.6 90
8.5 146.7 90
0.0 91.1 90
0.6 151.0 90
3.9 59.8 80
2.1 54.2 83
1.0 96.1 90
-1.9 104.8 90
-4.2 55.9 90
-8.5 12.5 88
-40.8 84.2 90
8.0 20.7 89
0
-104.7 6.2 89.0
-91.9 -2.5 46.7
-90.3 7.6 79.9
-85.5 9.0 82.2
-136.2 7.8 80.1
-90.2 7.0 78.7
-121.0 7.2 71.0
-104.5 5.8 92.5
-124.9 5.6 79.2
-114.1 3.7 110.9
-70.7 1.7 83.1
-83.5 6.3 97.1
-148.2 -33.9 66.8
-103.4 8.6 83.3
a Percentage Difference = 100% (Sampler B Cone. - Sampler C. Cone.)
/(Average of Sampler B and Sampler C Cone.)
-------
TABLE 10-12. SUMMARY OF RELATIVE STANDARD DEVIATIONS FOR DUPLICATE XRF
ANALYSES CONDUCTED BY EPA ON PEM/SIM/SAM FILTERS
Element
RSD Distribution (%)
No. of Pairs3
Mean
Median
Maximum
Std. Dev.
Si
90
4.3
2.8
43.7
5.9
Al
45
11.4
8.8
43.8
9.8
Ca
90
2.4
1.4
38.2
4.2
Fe
90
3.5
2.4
45.0
5.3
Mn
82
9.3
7.4
45.6
8.1
K
90
2.8
1.8
48.0
5.4
Br
55
12.6
9.6
42.0
9.8
Pb
58
12.4
10.4
41.3
9.6
S
90
4.4
3.4
26.5
4.7
Zn
90
5.7
4.4
25.1
4.9
CI
87
19.6
13.5
81.4
18.0
Ti
75
10.1
4.8
49.6
12.1
Cu
64
11.4
8.1
43.0
10.9
Sr
68
15.4
13.8
52.4
13.6
P
13
16.1
11.8
36.8
13.4
8 The number of pairs is the number of paired duplicate analyses in which both members of the pair
were deemed measurable.
Note: Duplicate analyses were performed on the PM10 samples only. Members of duplicate pairs
were assigned to different XRF batches.
10-30
-------
TABLE 10-13. SUMMARY OF RELATIVE STANDARD DEVIATIONS AND PERCENTAGE DIFFERENCES BETWEEN
XRF ANALYSES CONDUCTED BY EPA ON PEM/SIM/SAM FILTERS, BY ANALYSIS BATCH
Batch C Versus Batch Aa Batch C Versus Batch Bb
Element No. of Pairs Median RSD Median % Mean % Diff. No. of Pairs Median RSD Median % Mean % Diff.
(Both Meas.) Diff. (Both Meas.) Diff.
Si 44 2.2
Al 23 11.4
Ca 44 1.4
Fe 44 1.5
Mn 43 7.2
K 44 1.8
Br 32 8.2
Pb 30 11.5
S 44 3.4
Zn 44 4.4
CI 44 12.8
Ti 38 4.3
Cu 31 6.1
Sr 35 7.6
P 8 26.3
-1.9 -3.7* 46
-7.2 -7.7 22
-0.3 -2.4 46
-1.9 -3.8* 46
-5.2 -42 39
-1.8 -3.3* 46
-8.9 -5.2 23
-5.4 -2.5 28
-2.7 -3.8** 46
-5.1 -5.4** 46
-10.8 -12.8** 43
-4.8 -9.9** 37
-0.3 -3.3 33
-1.1 -1.5 33
-8.7 -10.0 5
3.1 -2.8 -3.4**
6.8 -6.5 -5.0
1.4 -1.3 -2.4**
3.4 -1.5 -2.0
7.9 3.5 1.9
2.1 -1.0 -2.6**
11.8 -3.6 -6.3
8.9 5.1 1.7
3.3 -3.9 -4.2**
4.2 0.2 -0.5
15.9 -18.3 -25.9**
4.9 -2.4 -2.6
8.2 0.7 3.3
15.4 -2.8 -1.6
3.6 3.8 -0.5
a Percentage Difference = 100% (Batch C Cone. - Batch A Cone.)
/(Average of Batch A and Batch C Cone.)
b Percentage Difference = 100% (Batch C Cone. - Batch B Cone.)
/(Average of Batch B and Batch C Cone.)
* Significantly different from zero at 0.05 level, based on t-test.
" Significantly different from zero at 0.01 level, based on t-test.
-------
specific temporal trend (i.e., whether Batch C is consistently showing lower (or higher)
concentrations than the other batches). Such tests for trend are given in Table 10-13, which
shows the following Batch C versus Batch A results (with similar results for Batch versus
Batch B): the number of paired analyses with both members measurable, the median RSD,
the media percent difference (Batch C - Batch A), and the mean percent difference. An
approximate t-test was conducted to assess the significance of true mean percentage
differences between batches.
While the median RSDs are less than 14% for all of the 15 key elements, the results in
Table 10-13 suggest that there are systematic trends in the analysis. A decreasing trend in
median and mean concentrations was estimated for all 15 elements when comparing Batch A
with C, and for 10 elements when comparing B with C. The largest percentage decrease
occurred for chlorine. Statistically significant differences in mean percent differences between
batches were found (0.05 level) for the following:
Both comparisons: Si, K, S, CI
Batch C - A Only: Fe, Zn, Ti
Batch C - B Only: Ca
Scatter plots of the Batch A and B concentrations versus their counterparts in Batch C are
given in Figures 10-3 and 10-4 for Ca and CI to illustrate the trends. Plots for 13 other
elements are displayed in Appendix Q. The difference in mean calcium concentrations
between filters analyzed in batches B and C was -2.4%. Although the change was statistically
significant the magnitude was small and not readily discernible in Figure 10-3, most values lie
close to the line of equal concentrations. For chlorine the percent difference between filters
analyzed in batches A and B, then reanalyzed in batch C, was -13% and -25% respectively.
These relatively large differences are observed in Figure 10-4 where most observations fall
above the line of equal concentrations.
In general, both SRM and duplicate analyses indicate a small downward trend in
reported concentrations of several elements over analysis time. It appears that a change in
instrument response may have resulted in elemental concentrations for batch C samples that
were 2% to 5% lower than would have been reported if they had been analyzed in batches A
or B. It is important to note that the magnitude of the differences over analysis time is in most
cases less than the uncertainties associated with reported concentrations for all elements
except chlorine. Chlorine is an interesting case; reported concentrations decreased
10-32
-------
ACONC |
14000 +
12000 +
10000 +
8000 +
6000 +
4000 +
2000 +
+
0 2000 4000 6000 8000 10000 12000 14000
CCONC
Figure 10-3. Concentration of Calcium (ng/m3) in Samples
Analyzed with Batch A or B Versus Duplicate
Analysis in Batch C (some observations
are hidden)
10-33
-------
ACONC
3500 +
3000 +
2500 ~
2000 +
1500 +
1000 +
500 +
B BB
¦+.
¦+.
¦+.
+-
0 500 1000 1500 2000 2500 3000 3500
CCONC
Figure 10-4. Concentration of Chlorine (ng/m3) in Samples
Analyzed with Batch A or B Versus Duplicate
Analysis in Batch C (some observations are hidden)
10-34
-------
throughout the course of the study. Twenty-six filters were first analyzed by WD-XRF; the
chlorine concentration's median difference was -120% when the same filters were analyzed at
EPA several weeks later. Chlorine concentrations of different samples decreased 13% to 26%
between duplicate analyses at EPA over the course of about three months. These data are
suggestive of losses due to volatilization from the sample filter during analysis or over time.
However, median chlorine concentrations analyzed at LBL were 27% higher than
concentrations measured at an earlier time at EPA. Reported chlorine concentrations may not
accurately reflect actual concentrations in the air at the time of sample collection. It is
possible that small decreases in other elemental concentrations over analysis time could be
due to particle loss from the filters during handling.
The second type of duplicate analysis, summarized in Tables 10-14 and 10-15,
involved an interlaboratory comparison of some of the filters by LBL The filters were first
analyzed by EPA and then were sent as blind samples to LBL. Results for the analysis of
approximately 100 PEM, SIM, and SAM sample filters are reported in Table 10-14 when
analyses at both labs resulted in concentrations above the uncertainty limits. Median
differences ranged from 76% for manganese to -11 % for aluminum. Except for chlorine and
manganese the absolute percent differences were 21% or less. For all cases except
aluminum, median elemental concentrations measured at LBL were higher than those
measured at EPA even though the samples were analyzed at LBL at a later time. Standard
reference materials 1832 and 1833 were not analyzed at LBL so no direct comparison of
potential instrument bias is possible. Ten sets of dichot filter sets (coarse/fine pairs) were also
analyzed at LBL after an initial analysis at EPA. Results are reported in Table 10-15 for fine
and coarse analysis pairs with measurements at both labs above the ULs. Total elemental
concentrations {fine plus coarse fractions) are reported for filter sets when both the fine and
coarse concentrations at both labs were above the ULs. Median percent differences were
generally less than 22%. High manganese and titanium concentrations were measured in the
fine (PM2 5) fraction at LBL, but there was only one measurable pair in each case. Chlorine
was consistently lower at LBL for the fine filters but higher than EPA on the coarse filters.
The third type of duplicate analysis involved a second interlaboratory comparison. A
subset of 26 of the filters analyzed by EPA using ED-XRF were also analyzed by WD-XRF at
another facility. Table 10-16 presents the minimum, median, and maximum differences and
10-35
-------
TABLE 10-14. SUMMARY OF DISTRIBUTIONS OF DIFFERENCES AND PERCENTAGE DIFFERENCES BETWEEN
ELEMENTAL CONCENTRATIONS AS DETERMINED BY EPA AND LBL ANALYSES ON THE SAME PEM, SIM, OR SAM FILTER
Distribution of Differences8
Distribution of % Differences**
Element
No. of Pairs
(Both Meas.)
Min. Diff.
Median Diff.
Max Diff.
Min % Diff.
Median %
Diff.
Max % Diff.
Si
Al
Ca
Fe
Mn
K
Br
Pb
94
35
98
96
47
98
41
57
-5200
-3200
-370
-130
6.1
-71
-14
-27
320
-330
35
46
27
61
0.95
9.6
2400
680
290
480
71
550
18
72
-42
-70
-11
-71
11
-43
52
-39
14
-11
5.8
8.1
76
13
3.2
21
50
32
109
27
346
95
162
253
S
Zn
CI
Ti
Cu
Sr
98
89
84
45
35
48
-43
-32
-1400
-170
-32
-9.9
140
11
38
32
8.6
2.4
670
64
370
250
58
25
-6.3
-62
-80
-37
-28
-40
12
20
14
21
27
17
38
135
333
149
217
183
a Difference = LBL concentration - EPA concentration (ng/m3)
b Percentage Difference = 100% (Difference)/(EPA Concentration)
-------
TABLE 10-15. MEDIAN DIFFERENCES AND PERCENTAGE DIFFERENCES BETWEEN
ELEMENTAL CONCENTRATIONS AS DETERMINED BY EPA AND LBL ANALYSES ON THE SAME DICHOT FILTERS
Element
No. of Pairs
Median Differences3
Median % Differences0
Finec
Coarsec
Total0
Fine
Coarse
Total
Fine
Coarse
Total
Si •
9
0
0
47
26
.
Al
0
0
0
-
•
-
-
-
-
Ca
10
10
10
6.1
19
20
5.7
3.4
3.3
F©
10
10
10
9.4
75
77
11
7.3
8.0
Mn
1
10
1
6.9
2.0
6.3
150
13
37
K
10
10
10
6.6
48
57
14
11
11
Br
10
6
6
1.0
0.9
2.3
15
37
18
Pb
8
10
8
2.8
1.2
5.0
6.3
16
10
S
10
10
10
81
67
155
6.5
20
8.6
Zn
10
10
10
3.7
3.5
6.8
11
19
14
CI
10
10
10
-78
41
-22
-45
22
6.7
Ti
1
10
1
14
14
26
82
13
2
Cu
10
10
10
2.3
0.8
3.2
8.8
1.8
6.0
Sr
0
10
0
.
0.6
-
-
8.9
-
a Concentration differences in ng/m3, LBL-EPA.
b Percentage Difference = 100% (Difference/(EPA Concentration).
c Both EPA and LBL concentrations measurable.
d Fine and coarse concentrations measurable for both EPA and LBL filter.
-------
TABLE 10-16. SUMMARY OF DISTRIBUTIONS OF DIFFERENCES AND PERCENTAGE
DIFFERENCES BETWEEN ELEMENTAL CONCENTRATIONS AS DETERMINED
BY ED-XRF AND WD-XRF ANALYSES ON THE SAME FILTER
Distribution of Differences3
Distribution of % Differences'1
No. of Pairs
Min.
Median
Max.
Min. %
Median
Max. %
Element
(Both Meas.)
Diff.
Diff.
Diff.
Diff.
Diff.
Diff.
Si
26
-5900
-1400
-480
-50.3
-28.1
-19.8
Al
18
-2900
-1100
-180
-98.8
-34.3
-10.5
Ca
26
-500
-150
-60
-21.0
-9.4
-3.0
Fe
26
61
169
400
4.1
10.3
23.6
Mn
24
-17
-2.5
6.9
-79.4
-6.2
11.5
K
26
-420
-160
-54
-31.1
-17.7
-10.4
Br
21
-5.1
1.0
5.7
-60.7
6.5
25.3
Pb
16
-5.0
0.2
15
-23.0
0.5
23.8
S
26
-27
140
650
-6.3
12.5
27.7
Zn
26
-12
-4.2
6.7
-66.2
-5.8
8.4
CI
23
-1500
-350
-130
-611.3
-118.1
-41.4
Ti
26
-43
-0.9
21
-47.4
0.7
12.6
Cu
22
-12
-0.3
6.7
-97.1
-0.5
22.7
Sr
22
-2.4
2.2
8.9
-12.9
14.6
53.5
P
0
^Difference = ED-XRF Concentration - WD-XRF Concentration (ng/m3)
Percentage Difference = 100% (Difference)/(ED-XRF Concentration)
-------
percent differences for the two labs/methods. Median percent differences greater than +25%
were observed for Si, Al, and CI. The use of different particle size attenuation correction
factors may have been responsible for the differences in Si and Al, while chemical reaction or
volatility losses of CI may have occurred between the initial analysis by WD-XRF and the later
analysis by ED-XRF.
ELEMENTAL ANALYSIS RESULTS
Temporal-Site Data
Temporal-site mean concentrations for the 20 principal elements, based on the XRF
analyses, are reported in Table 10-17. The means are given by time of day, size cut, and
sampler type. Sample sizes are approximately 45 for each mean (see Tables 9-6 and 9-7).
For virtually all elements, for both daytime and nighttime, and for both 2.5 and 10 jim
samples, the dichot levels were lower than those of the SAM or PEM. The only notable
exceptions were chlorine and copper. The PEM and SAM mean concentrations were quite
comparable throughout. For most elements, higher mean concentrations were found in the
daytime samples than in the nighttime samples. Selenium, vandium, and nickel were
measurable on many more dichot PM^ q samples than PEM or SIM samples due to the larger
dichot particle mass catches.
Residence Data
Estimated mean elemental concentrations for the SAM, SIM, and PEM residence data
are presented in Table 10-18. Sample sizes upon which these means are based range from
161 to 173 (see Table 9-16). The relative standard errors of these estimates, expressed as a
percent of the mean, are shown in Table 10-19. The estimates for the SAM and SIM samples
are weighted to produce means applicable to the target population of household-days, while
the PEM estimates are weighted to apply to the target population of person-days.
Table 10-20 shows estimated mean mass ratios for the principal elements, i.e., the weighted
average of:
100% [element concentration] I [particulate concentration]
These estimates (expressed as a percentage) are given for the same ten categories as those
used in Table 10-18. Table 10-21 gives the relative standard errors of the mean mass ratios.
10-39
-------
For many of the elements, the patterns of mean concentration levels over time of day
and over sample type were very similar to those noted for the gravimetric results. This was
the case for Si, Al, and Fe, for example, whose concentration means (weighted) are shown in
the form of bar charts in Figures 10-5,10-6, and 10-7, respectively. (Note that the
concentration scales in these figures are not the same.) The basic patterns are:
. PEM 10 jim daytime concentrations much higher than either SIM or SAM
PEM 10 nm nighttime concentrations falling between those of SIM and SAM
Higher daytime than nighttime concentrations (for all sample types, and for 2.5 nm
samples, as well as 10 nm samples)
SAM 10 urn concentrations somewhat higher than SIM concentrations.
These results parallel the gravimetric results for aerosol concentrations reported in Section 9.
Some of the elements do not follow these general trends. Sulfur, for example, exhibited
concentration means that did not vary greatly by time of day or by sample type; also, a much
higher proportion of the sulfur occurred in the fine fraction (see Figure 10-8). Chlorine, for the
most part, did follow the basic pattern; however, the SAM daytime mean concentration was
lower than the nighttime SAM mean and lower than the daytime SIM mean (see Figure 10-9).
Mean particle concentrations for the same sample media are shown for comparison in
Figure 10-10. Bar charts are shown for ten other elements with high percent measurable
values in Appendix R.
Elements associated with soils were present at the highest concentrations, generally
greater than 1000 ng/m^ in the PM^q samples. These elements were found at much lower
concentrations on the PM2 5 samples, indicating that soil/crustal particles were typically
greater than 2.5 \im in aerodynamic diameter. Lead was present at higher proportions on the
PM2 5 samples than the PM10 samples, but the differences were small. Cadmium was not
measured above the UL in any sample, and arsenic was measurable in only a few samples.
Other, more sensitive, analytical methods will be required to measure concentrations of these
elements because of the small particle mass collected during this study. Nickel was
measurable in 20% of the PEM PM^q samples, which also had the highest particle
concentrations. This element was measurable on less than 10% of all other sample types.
Much work is needed to apply these elemental data in source/receptor models using
meteorological, air exchange, and personal activity information collected during the course of
the PTEAM Pilot Study.
10-40
1
-------
Elemental Distributions
Estimated percentiles for the 15 elements with high percent measurable values are
shown for the PEM, SEM, and SAM residence data, by day and night, in Table 10-22.
Distribution patterns that were observed for particle concentrations (Figure 9-4) are very
similar to those observed for the elements Si, Al, Fe, Ca, Mn, K, Ti, Sr, Zn, Cu, Br, and Pb.
Examples of these weighted distributions are shown for Si in Figure 10-11 and Pb in Figure
10-12. The fact that the concentrations of elements of PEM daytime samples had
distributions similar to the particle distributions suggests that the observed elevated personal
exposure levels were not due to organic particles from the body or clothing. Had this been
the case, ratios of elemental mass to total particle mass would have been lower for the PEM
samples as compared to the SIM or SAM samples.
Somewhat different distributional patterns were observed for CI and S, as shown in
Figures 10-13 and 10-14. Chlorine daytime outdoor concentrations were lower than indoor
levels; no other element exhibited this behavior. Sulfur was unique, showing nearly equal
distributions across all sample types, both during the daytime and overnight. For particle
concentrations and most other elements, the daytime PEM concentrations were much higher
than SAM or SIM, but this was not true for sulfur. Sulfur was predominantly associated with
particles smaller than 2.5 iim (Figure 10-8). The distributions and correlations (Table 10-23)
suggest an outdoor source. These data strongly suggest that increases observed in daytime
PEM concentrations were due to particles larger than 2.5 This would be compatible with
the idea of a person increasing their exposures by resuspending larger particles from surfaces
on or near themselves as they move around. It also does not eliminate the possibility of bias
due to oversampling particles larger than 10 urn due to movement of the impactor. Further
investigation is warranted to determine the reason for high daytime personal exposure
concentrations.
10-41
-------
TABLE 10-17. MEAN ELEMENTAL CONCENTRATIONS (ng/m3) FOR TEMPORAL SITE SAMPLES
DAYTIME NIGHTTIME
ELEMENT
PM25
PM10
pm2,5
PM10
SAM
Dichot
PEM
SAM
Dichot
SAM
Dichot
PEM
SAM
Dichot
Si
660
350
7300
7200
5800
260
150
3300
3500
2800
Al
ja
160
2900
2900
2400
-
51
1200
1200
1100
Ca
330
170
2500
2500
1900
150
81
1200
1300
1000
Fe
380
220
2300
2200
1700
170
120
1100
1100
860
Mn
13
8.5
48
49
40
6.6
5.6
23
24
21
Ni
-
-
-
-
-
-
-
-
-
2.6
K
190
110
1000
990
780
115
75
550
570
450
Se
-
-
-
-
1.2
-
-
-
-
1.6
V
-
-
-
-
5.2
-
-
-
-
5.8
Br
9.7
7.5
10
12
9.1
10
8.9
12
13
11
Pb
19
16
31
31
25
17
14
22
25
22
S
1500
1100
1700
1800
1500
1600
1200
1700
1900
1500
Zn
39
27
63
65
50
32
18
44
47
32
CI
82
86
220
230
300
82
140
310
330
470
Ti
26
14
190
190
170
10
7.2
88
99
82
Cu
27
41
44
48
69
27
55
37
38
86
Sr
-
2.0
17
17
13
3.2
1.4
11
12
7.9
aFewer than 30% of samples with concentrations greater than the uncertainty limit.
-------
TABLE 10-18. WEIGHTED MEAN ELEMENTAL CONCENTRATIONS (rig/m3) FOR RESIDENCE DATA®
DAYTIME NIGHTTIME
PMpfi PMjjq PMg,B PM10
ELEMENT SAM SIM SAM SIM PEM SAM SIM SAM SIM PEM
Si
740
700
7700
6300
12000
380
360
5000
3300
4200
Al
_b
-
3100
2300
4700
-
-
2000
1200
1400
Ca
330
380
2300
2300
4300
170
200
1500
1200
1700
Fe
400
340
2300
1800
3400
260
200
1700
980
1200
Mn
12
9.8
51
38
69
9.9
7.5
37
22
24
K
220
260
1100
1100
1900
150
200
800
650
800
Br
8.8
9.1
10
13
25
11
8.6
13
11
14
Pb
20
17
30
27
40
23
20
32
27
26
S
1500
1300
1800
1700
1800
1600
1300
1900
1500
1500
Zn
41
42
65
86
150
38
34
56
60
67
CI
83
130
230
410
840
170
100
500
290
440
Ti
-
-
210
190
390
-
-
140
100
130
Cu
-
11
15
22
41
9.6
8.9
17
15
19
Sr
.
.
18
15
25
.
-
14
9.8
11
P - 230
® Results are weighted to reflect the target population of person-days (PEM) or household-days (SIM and SAM).
Fewer than 30% of samples with concentrations greater than the uncertainty limit.
-------
TABLE 10-19. RELATIVE STANDARD ERRORS (%) FOR MEAN ELEMENTAL CONCENTRATIONS3
DAYTIME NIGHTTIME
PMg.5 PMio PM25 PMjo
ELEMENT SAM SIM SAM SIM PEM SAM SIM SAM SIM PEM
Si
10.8
8.3
7.2
8.2
7.3
Al
-
-
8.9
9.6
8.5
Ca
9.4
8.7
7.2
8.7
6.3
Fe
8.4
8.3
7.1
8.8
8.1
Mn
7.4
8.1
7.9
9.0
7.6
K
6.8
8.4
6.0
7.6
6.6
Br
9.7
8.4
8.7
8.2
13.0
Pb
7.5
7.1
7.1
8.1
9.7
S
10.9
10.6
10.6
8.5
8.7
Zn
7.0
5.8
5.1
5.7
6.7
CI
15.0
13.9
11.9
8.9
8.7
Ti
-
-
7.0
8.0
8.0
Cu
-
8.0
4.8
5.8
5.7
Sr
-
-
5.2
5.7
6.2
P
.
.
.
_
8.5
7.3
5.4
5.0
6.9
9.2
-
-
5.7
7.7
11.1
9.2
5.2
6.8
6.9
8.4
6.5
5.2
5.4
7.3
9.5
7.0
7.8
5.7
7.7
9.4
5.3
9.8
4.3
7.3
7.6
6.8
5.9
6.4
5.6
10.0
7.5
16.5
5.3
11.3
11.3
8.4
8.3
8.0
7.3
7.0
7.1
6.0
6.3
5.0
5.7
11.9
12.7
10.0
9.7
8.6
-
-
4.9
7.1
7.9
10.9
11.3
8.2
5.7
10.9
.
4.1
5.6
6.6
a The relative standard errors apply to the estimated means given in Table 10-18.
-------
TABLE 10-20. WEIGHTED MEAN ELEMENTAL/PARTICLE MASS RATIOS (%) FOR RESIDENCE DATA3
DAYTIME
NIGHTTIME
WENT
PM
2.5
PMi0
PM2.5
PM10
SAM
SIM
SAM
SIM
PEM
SAM
SIM
SAM
SIM
PEM
Si
2.16
2.37
9.04
7.15
7.57
1.39
1.74
7.02
5.97
5.72
A!
3.63
2.56
2.88
2.78
2.08
1.87
Ca
0.94
1.19
2.70
2.57
2.85
0.58
0.90
2.25
1.71
1.56
Fe
1.17
1.09
2.72
1.99
2.14
0.83
0.85
2.25
1.71
1.56
Mn
0.04
0.03
0.06
0.04
0.04
0.03
0.03
0.05
0.04
0.03
K
0.60
0.69
1.27
1.18
1.22
0.47
0.72
1.09
1.11
1.09
S
3.30
3.22
1.92
2.01
1.36
3.62
4.28
2.23
2.72
2.16
Zn
0.13
0.13
0.08
0.11
0.11
0.12
0.15
0.07
0.11
0.10
CI
0.16
0.27
0.28
0.44
0.58
0.40
0.34
0.68
0.52
0.61
Ti
0.25
0.21
0.25
0.19
0.18
0.18
Cu
0.04
0.02
0.03
0.03
0.03
0.04
0.02
0.03
0.03
Sr
0.02
0.02
0.02
0.02
0.02
0.02
P
0.15
a Results are weighted to reflect the target population of person-days (PEM) or household-days (SIM and SAM). Estimated means < 0 are
reported as 0.
-------
TABLE 10-21. RELATIVE STANDARD ERRORS (%) FOR MEAN MASS RATIOS3
DAYTIME NIGHTTIME
PM25 PM10 PM2.s PMio
ELEMENT SAM SIM SAM SIM PEM SAM SIM SAM SIM PEM
Si
8.2
12.8
4.6
5.4
3.3
13.1
10.9
6.4
4.8
5.3
Al
-
-
6.1
7.1
4.2
-
-
8.2
5.8
7.4
Ca
7.8
13.0
5.0
6.0
4.0
12.4
9.0
5.9
5.0
4.3
Fe
6.4
11.7
4.1
5.3
3.5
9.4
7.1
5.4
4.2
5.2
Mn
7.9
9.5
4.3
5.9
3.4
9.9
8.7
5.4
4.3
5.1
K
4.9
7.9
3.3
4.2
2.2
9.6
7.2
4.3
3.5
3.9
Br
5.9
5.8
4.5
8.1
14.1
5.4
6.3
3.6
6.8
17.9
Pb
7.0
7.3
5.8
6.0
9.4
6.8
16.5
4.8
7.5
9.2
S
6.9
8.1
7.6
7.8
5.5
6.2
7.0
5.9
7.0
6.7
Zn
12.9
8.9
4.5
5.9
5.0
13.3
14.6
8.7
5.7
5.2
CI
7.2
8.3
8.8
5.7
4.6
23.5
12.8
10.6
8.1
6.5
Ti
-
-
4.2
5.5
3.9
-
-
5.7
5.4
4.2
Cu
-
13.0
6.5
7.7
7.4
13.1
16.3
7.6
6.4
11.2
Sr
-
.
3.5
5.8
3.1
13.3
9.9
4.5
4.8
3.1
a The relative standard errors apply to the estimated means given in Table 10-20.
-------
TABLE 10-22. WEIGHTED DISTRIBUTIONS OF ELEMENT CONCENTRATIONS (ng/m3) FOR RESIDENCE DATA3
DAYTIME
NIGHTTIME
PM
Element/Percentile SAM
23-
SIM
SAM
PM
UL
SIM
PEM
PM
SAM
2£.
SIM
SAM
PM
m.
SIM
PEM
Si
10th
290
220
4000
2000
3400
140
150
2500
1400
1500
25th
410
330
5200
3000
5200
210
220
3000
1900
2300
50th (median)
560
540
6800
4900
9200
340
320
4600
2700
3300
75th
760
820
8600
8100
16000
480
480
6300
4200
5300
90th
1200
1300
12000
12000
22000
650
580
8500
5700
7800
95th
1700
1800
15000
15000
28000
780
840
9900
6700
10000
99th
4400
3700
38000
31000
59000
1200
1100
13000
11000
14000
Al
10th
-D - 1300
450
1000
750
240
400
25th
1900
950
1900
1200
610
700
50th (median)
2500
1900
3400
1700
990
1000
75th
3600
3000
6000
2600
1600
1800
90th
5200
4700
8900
3500
2100
3200
95th
7000
6000
11000
4200
2800
4200
99th
17000
15000
27000
4700
4300
4900
(Continued)
-------
TABLE 10-22. (continued)
DAYTIME NIGHTTIME
PM?,5 PM^ PMgs PM10
Element/Percentile SAM SIM SAM SIM PEM SAM SIM SAM SIM PEM
Ca
10th
130
120
1200
690
1200
63
90
790
510
690
25th
190
180
1500
1100
2000
85
120
1000
670
920
50th (median)
270
260
2100
1700
3400
140
170
1300
1000
1300
75th
370
420
2700
2700
5500
210
240
1900
1400
2000
90th
520
700
3800
4200
8600
310
320
2400
2000
3300
95th
670
1100
4500
5200
11000
360
440
3100
2900
4500
99th
1400
1600
8600
13000
16000
580
600
4600
3700
5700
Fe
10th
170
110
1200
520
930
92
83
870
400
440
25th
230
170
1500
890
1400
140
110
1100
520
640
50th (median)
340
270
2100
1400
2600
230
170
1430
840
960
75th
490
440
2800
2000
4400
350
280
2200
1200
1400
90th
690
630
3700
3300
6800
490
380
2800
1700
2300
95th
750
820
4200
4700
8700
560
450
3400
2200
3000
99th
2000
1300
11000
9100
21000
700
550
4100
3600
4000
-------
TABLE 10-22. (continued)
DAYTIME NIGHTTIME
^ Zha.
Element/Percentile SAM SIM SAM SIM PEM SAM SIM SAM SIM PEM
Mn
10th
4.2
1.0
26
10
19
2.2
0.5
17
6.6
8.4
25th
6.5
5.0
32
19
28
5.1
3.4
23
10
13
50th (median)
11
8.2
46
30
49
8.6
6.3
30
20
20
75th
16
12
61
47
93
14
10
48
28
31
90th
20
19
79
69
140
19
17
65
38
49
95th
22
24
96
87
170
23
20
74
47
58
99th
38
45
230
190
390
32
26
100
74
72
K
10th
100
90
580
340
640
70
71
470
280
340
25th
150
120
800
550
910
93
89
610
370
460
50th (median)
180
200
1000
880
1500
120
140
720
540
680
75th
270
310
1300
1400
2400
190
230
980
770
1000
90th
360
460
1600
2000
3700
240
360
1300
1100
1400
95th
510
740
1800
2500
4300
290
590
1300
1500
1800
99th
780
1300
4500
4200
8600
510
1500
1600
2400
2100
-------
TABLE 10-22. (continued)
DAYTIME NIGHTTIME
pm25 pm10 pm2.5 pm10
Element/Percentile SAM SIM SAM SIM PEM SAM SIM SAM SIM PEM
Br
10th
1.3
1.6
1.4
2.8
7.3
3.0
2.8
3.9
4.2
5.5
25th
4.4
4.1
4.9
7.2
12
5.3
4.5
7.9
6.6
8.7
50th (median)
7.8
7.9
10
11
18
8.9
7.1
12
10
12
75th
12
13
15
17
27
14
12
18
15
17
90th
17
17
19
24
38
22
15
24
21
21
95th
20
25
23
30
55
27
19
31
24
25
99th
31
30
28
40
210
32
25
35
35
47
Pb
10th
6.5
3.1
11
8.6
14
7.4
4.0
11
6.6
6.4
25th
11
8.1
18
15
19
12
7.7
17
12
12
50th (median)
18
15
27
23
32
18
15
25
22
19
75th
25
23
38
33
46
27
22
40
34
29
90th
34
30
47
47
74
40
34
60
46
43
95th
40
41
57
63
100
50
40
68
54
51
99th
82
68
120
110
150
64
48
80
92
390
(Continued)
-------
TABLE 10-22. (continued)
DAYTIME NIGHTTIME
PM2.5 PM10 PM2S PM^
Element/Percentile SAM SIM SAM SIM PEM SAM SIM SAM SIM PEM
10th
25th
50th (median)
75th
90th
95th
99th
40
94
190
300
440
550
980
10th
200
190
310
360
530
260
220
380
430
470
25th
350
320
530
670
890
760
570
1000
770
720
50th (median)
1400
1200
1600
1600
1600
1400
1100
1700
1300
1300
75th
2200
1900
2500
2200
2400
2200
1800
2500
2000
2100
90th
2800
2600
3100
3100
3200
3100
2500
3400
2800
2700
95th
3600
3500
4300
4300
4300
3500
3100
4000
3400
3200
99th
6600
6000
7600
6400
5400
5200
4700
6300
5200
5000
(Continued)
-------
TABLE 10-22. (continued)
DAYTIME NIGHTTIME
PMgS PM10 PM25 PM
10
Element/Percentile SAM SIM SAM SIM PEM SAM SIM SAM SIM PEM
CI
10th
12
23
68
120
220
18
10
100
77
130
25th
27
38
92
160
410
38
30
180
130
200
50th (median)
53
77
160
280
700
84
60
360
230
370
75th
83
140
250
510
1100
170
120
600
350
580
90th
143
250
550
860
1600
370
230
1200
580
910
95th
190
520
810
1100
1800
540
300
1500
860
1100
99th
790
820
1200
1600
3400
1400
740
2000
1300
1400
Ti
10th
110
53
120
64
40
51
25th
140
92
180
94
59
80
50th (median)
180
150
260
120
83
110
75th
240
250
500
170
130
160
90th
330
360
750
240
180
260
95th
380
490
940
270
240
320
99th
970
800
1400
360
330
370
(Continued)
-------
TABLE 10-22. (continued)
DAYTIME NIGHTTIME
pm2.5 pm10 pm2.5 pm10
Element/Percentile SAM SIM SAM SIM PEM SAM SIM SAM SIM PEM
Zn
10th
13
14
28
38
62
13
11
22
27
29
25th
21
23
39
48
83
19
16
31
36
38
50th (median)
36
36
63
68
120
29
26
46
50
61
75th
50
55
80
110
192
48
43
70
74
80
90th
73
81
110
140
260
78
68
110
97
110
95th
88
89
130
180
340
100
78
130
130
150
99th
120
170
160
360
490
150
130
170
180
200
Cu
10th
1.0
6.6
7.3
15
0.7
0.7
5.6
4.4
6.3
25th
3.8
10
11
22
3.6
2.9
9.3
7.8
9.9
50th (median)
8.5
14
16
34
6.7
6.3
14
12
15
75th
13
18
25
49
11
11
20
19
21
90th
22
25
42
72
16
17
28
28
31
95th
29
27
56
79
32
24
37
34
50
99th
69
35
100
180
63
67
110 •
71
76
(Continued)
-------
TABLE 10-22. (continued)
PMps PMm PMPfi PM10
Element/Percentile SAM SIM SAM SIM PEM SAM ' SIM SAM SIM PEM
Sr
10th - - 9.5 5.2 9.3 - - 7.7 5.1 5.6
25th - - 12 8.5 13 - - 10 6.7 7.0
50th (median) - 17 13 22 - 13 9.4 10
75th - 21 18 32 - 17 12 14
90th - - 27 25 46 - - 21 16 20
95th - - 30 33 60 - 25 18 23
99th - - 59 46 86 - - 33 23 27
Results are weighted to reflect the target population of person-days (PEM) or household-days (SIM and SAM). Estimates of the higher
percentiles may be subject to substantial imprecision.
Fewer than 30% of the samples with concentrations greater than the uncertainty limit.
-------
TABLE 10-23. CORRELATIONS BETWEEN OUTDOOR, INDOOR
AND PERSONAL ELEMENT CONCENTRATIONS
Element
Outdoor vs Indoor
Indoor vs Personal
S
0.93
0.93
Pb
0.73
0.58
Br
0.63
0.48
Al
0.56
0.47
Mn
0.55
0.55
Si
0.54
0.53
Fe
0.52
0.49
K
0.47
0.55
Ti
0.47
0.46
Zn
0.45
0.56
Sr
0.42
0.58
Cu
0.32
0.30
Ca
0.27
0.56
CI
0.16
0.61
P
0.13
0.56
10-55
-------
ng/m^
12800
11200
9600
8000
6400
4800
3200
1600
0
DAYTIME
NIGHTTIME
lOjJm
2.
SAM SIM PEM SAM SIM PEM
Figure 10-5. Mean Silicon Concentrations
10-56
-------
, 3
ng/m
4800
4200
3600
3000
2400
1800
1200
600
0
DAYTIME
NIGHTTIME
lO^im
2 . 5pm I^ZZj l__
SAM SIM PEM SAM SIM PEM
Figure 10-6. Mean Aluminum Concentrations
10-57
-------
ng/nT*
4000
3500
3000
2500
2000
1500
1000
500
0
DAYTIME
NIGHTTIME
lOjXm
2 . 5pm
SAM SIM PEM SAM SIM PEM
Figure 10-7.
Mean Iron Concentrations
-------
g/m3
2 000
1750
1500
1250
1000
750
500
250
0
DAYTIME
NIGHTTIME
lO^lm
7 . 5nm
SAM SIM PEM SAM SIM PEM
Figure 10-8. Mean Sulfur Concentrations
10-59
-------
/m3
960
840
720
600
480
360
240
120
0
DAYTIME
NIGHTTIME
lOpiD
2 . 5|4iti
SAM SIM PEM SAM SIM PEM
Figure 10-9. Mean Chlorine Concentrations
10-60
-------
^g/m3
160
140
120
100
80
60
40
20
0
SAM SIM PEM SAM SIM PEM
Figure 10-10. Mean PM^ and PM10 Concentrations
10-61
0 A Y T I M E
NIGHTTIME
10/im
2.5/tm
-------
ng/m^
24000
21000
18000
15000
12000
9000
6000
3000
0
DAYTIME
NIGHTTIME
90th
75th
50th
25th
10th
OUTDOOR INDOOR PERSONAL OUTDOOR INDOOR PERSONAL
Figure 10-11. Distribution of Silicon Concentrations
for Residence PMt0 Samples
10-62
-------
m3
80
70
60
50
40
30
20
10
0
DAYTIME
NIGHTTIME
90th
75th
50th
25th
10th
OUTDOOR INDOOR PERSONAL OUTDOOR INDOOR PERSONAL
Figure 10-12. Distribution of Lead Concentrations
for Residence PM10 Samples
10-63
-------
ng/m^
2000
1750
1500
1250
1000
750
500
250
0
DAYTIME
NIGHTTIME
90th
75th
50th
25th
110th
OUTDOOR INDOOR PERSONAL OUTDOOR INDOOR PERSONAL
Figure 10-13. Distribution of Chlorine Concentrations
for Residence PM10 Samples
10-64
-------
ng/m3
4000
3500
3000
2500
2000
1500
1000
500
0
DAYTIME
NIGHTTIME
90th
75th
50th
25th
10th
OUTDOOR INDOOR PERSONAL OUTDOOR INDOOR PERSONAL
Figure 10-14. Distribution of Sulfur Concentrations
for Residence PM10 Samples
10-65
-------
\
\
\
\
-------
SECTION 11
REFERENCES
Alzona, J., Cohen, B. L, Rudolph, H., Jow, H. N., and Frohliger, J. O. Indoor-Outdoor
Relationships for Airborne Particulate Matter of Outdoor Origin, Atmos. Environ.. 13:1-6, 1979.
Chromy, J. R.. Sequential Sample Selection Methods. In: Proceedings of the American
Statistical Association Section on Survey Research Methods, pp. 401-406, 1979.
Clayton, C. A., Pellizzari, E. D., and Weiner, R. W. Use of a Pilot Study for Designing a
Large-Scale Probability Study of Personal Exposure to Aerosols, Journal of Exposure Analysis
and Environmental Epidemiology, in press.
Cohen, S. B., Burt, V. L., and Jones, G. K. Efficiencies in Variance Estimation for Complex
Survey Data, The American Statistician, 40{2):157-163, 1986.
Dockery, D. W., and Spengler, J. D. Personal Exposure to Respirable Particulates and Sulfates,
J. Air Pollution Control Assoc.. 31:153-159, 1981.
El-Shoboksky, M. S., and Hussein, F. M. Correlation Between Indoor-Outdoor Inhalable
Particulate Concentrations and Meteorological Variables, Atmos. Environ., 22:2667-2674, 1988.
Federal Register. Regulation for Implementation of Revised Particulate Standards, July 1,1987,
Volume 52, No. 126, pp. 24676-24715. U.S. Government Printing Office, Washington, DC, 1987.
Francis, I. and Sedransk, J. A Comparison of Software for Processing and Analyzing Survey
Data, Bulletin of the International Statistical Institute. 48:1-31, 1979.
Gilbert, R. O. Statistical Methods for Environmental Pollution Monitoring. Van Norstrand
Reinhold, New York, NY, 1987.
Hansen, M. H., Madow, W. G., and Tepping, B. J. An Evaluation of Model-Dependent and
Probability-Sampling Inferences in Sample Surveys, Journal of the American Statistical
Association. 78(384):776-793, 1983.
Hartwell, T. D., Clayton, C. A., Michie, R. W., Whitmore, R. W., Zelon, H. S., Jones, S. M., and
Whitehurst, D. A. Study of Carbon Monoxide Exposure of Residents of Washington. DC and
Denver. CO. EPA-600/S4-84-031, PB84-183516, Environmental Monitoring Systems Laboratory,
U. S. Environmental Protection Agency, Research Triangle Park, NC, 1986.
Hartwell, T.D., Pellizzari, E., Perritt, R„ Whitmore, R., Zelon, H., Sheldon, L., Sparacino, C., and
Wallace, L. Results from the Total Exposure Assessment Methodology (TEAM) Study in Selected
Communities in Northern and Southern California, Atmos. Environ.. 21:1995-2004, 1987.
11-1
-------
lannacchione, V. G., Milne, J. G., and Folsom, R. E. Response Probability Weight Adjustment
Using Logistic Regression. In: Proceedings of the American Statistical Association Section on
Survey Research Methods, 1991.
Immerman, F. W. and Schaum, J. L. Nonoccupational Pesticide Exposure Study: Final Report,
EPA/600-3-90-003. U.S. Environmental Protection Agency, Washington, DC, 1990.
Kamens, R., Lee, C. T., Weiner, R., and Leith, D. A Study to Characterize Indoor Particles in
Three Non-Smoking Homes, Atmos. Environ.. 25:939-948, 1991.
Kulmala, M. V., Scari, H., Raunema, T., and Olin, M. Indoor Air Aerosols: Outdoor Air Influence
on Indoor Air, In: Proceedings of the 4th International Conference on Indoor Air Quality and
Climate. Volume 1. B. Seifert et a!., eds., pp. 564-568. Oraniendruck GmbH Publishers, Berlin,
1987.
Lioy, P. J. Waldman, J. M., Buckley, T., Butler, J., and Pietarinen, C. The Personal, Indoor and
Outdoor Concentrations of PM-10 Measured in an Industrial Community During the Winter.
Atmos. Environ.. 24:57-66, 1989.
Uoy, P. J. Wainman, T., and Marple, V. An Intercomparison of the Indoor Air Sampling Impactor
and the Dicotomous Samples for a 10 jim Cut Size, J. Air Pollution Control Assoc.. 38:668-670,
1988.
Madow, W. G., Olkin, I., and Rubin, D. B.p eds. Incomplete Data in Sample Surveys. Volume 2:
Theory and Bibliographies. Academic Press, New York, NY, 1983.
Miller, F. J., et al. Size Considerations for Establishing a Standard for Inhalable Particles, J. Air
Poll. Cont. Assoc.. 29:610, 1979.
National Academy of Sciences. Epidemiology of Air Pollution, pp. 1-224. National Academy
Press, Washington, DC, 1985.
National Research Council. Committee on Indoor Air Pollutants. In: Indoor Air Pollutants, pp.
1-537. National Academy Press, Washington, DC, 1981.
Newill, V. A. The Role of Total Exposure Measurement in Risk Management, In: Proceedings of
1987 EPA/APCA Symposium on Measurement of Toxic and Related Air Pollutants. APCA Pub.
VIP-8, EPA Report No. 600/9-87-010, pp. 1, 1987.
Ott, W., et al. The Environmental Protection Agency's Research Program on Total Human
Exposure, Environment International. 12:475, 1986.
Pellizzari, E. D., Perritt, K., Hartwell, T. D., Michael, L. C., Sparacino, C. M., Sheldon, L. S.,
Whitmore, R., Uninger, C., Zelon, H., Handy, R. W., and Smith D. Total Exposure Assessment
Methodology (TEAM) Study: Elizabeth and Bavonne, New Jersey, Devils Lake, North Dakota,
and Greensboro, North Carolina. Volume II. Final Report, EPA Contract No. 68-02-3679,1986a.
11-2
-------
Pellizzari, E. D., Perritt, K., Hartwell, T. D., Michael, L. C., Whitmore, R., Handy, R. W., Smith D.,
and Zelon H. Total Exposure Assessment Methodology (TEAM) Study: Selected Communities
in Northern and Southern California, Volume III, Final Report. EPA Contract No. 68-02-3679,
1986b.
Pellizzari, E. D., Hartwell, T. D., Zelon, H., Perritt, R., Sebestik, J., Williams, W., Smith D. J.,
Keever, J., Decker, C. E., Jayanty, R. K. M., Thomas, K., Whitaker, D. A., and Michael, L C.
Baltimore Total Exposure Assessment Methodology (TEAM) Study. Final Report. EPA Contract
No. 68-02-4406, 1988.
Pellizzari, E. D., Michael, L. C., Perritt, K., Smith D. J., Hartwell, T. D., and Sebestik, J.
Comparison of Indoor Toxic Air Pollutant Levels in Several Southern California Communities.
Draft Final Report. EPA Contract No. 68-02-4544, 1988.
Pellizzari, E. D., Thomas, K. W., Smith D. J., Perritt, K., and Morgan, M. Total Exposure
Assessment Methodology (TEAM): 1987 Study in New Jersey. Draft Final Report. EPA Contract
No. 68-02-4544, 1988.
Phalen, R. F., Hinds, W. C., John, W. S., Lioy, P. J., Lippmann, M., McCawley, M. A., Rabbee,
O. G., Sodertiolm, S. C., and Stuart, B. O. Rationale and Recommendations for Particle Size
Selective Sampling in the Workplace, Appl. Ind. Hva.. 1:3-14, 1986.
Research Triangle Institute and Harvard School of Public Health. Particle Total Exposure
Assessment Methodology (PTEAM): Pilot Study. Volume 2. Protocols for Environmental Sampling
and Analysis. Workplan for EPA Contract No. 68-02-4544, EPA Work Assignment 67, CARB
Agreement No. A833-060, U. S. Environmental Protection Agency, Research Triangle Park, NC,
1990a.
Research Triangle Institute and Harvard School of Public Health. Particle Total Exposure
Assessment Methodology (PTEAM): Pilot Study Volume III: Quality Assurance Project Plan.
Workplan for EPA Contract No. 68-02-4544, EPA Work Assignment 67 and CARB Agreement No.
A833-060. Environmental Protection Agency, Research Triangle Park, NC, 1990b.
Research Triangle Institute and Harvard School of Public Health. Nine-Home Particle TEAM
Study. Final Report for EPA Contract No. 68-02-4544, Work Assignment II-66. U. S.
Environmental Protection Agency, Research Triangle Park, NC, 1990.
Research Triangle Insisute. National Household Seroprevalence Survey Feasibility Study Final
Report. Research Triangle Institute, Research Triangle Park, NC, 1990.
Sega, K, Fugas, M., Kalima, N., and Sisovic, A. Indoor-Outdoor Relationships for Responsible
Particles, Total Suspended Particulate Matter and Smoke Concentrations in Modern Office
Buildings. Environ. Int.. 12:71-74,1986.
Sexton, K., Liu, K. S., Hayward, S. B., and Spengler, J. D. Characterization and Source
Apportionment of Wintertime Aerosol in a Wood-Burning Community, Atmos. Environ.. 19:1225-
1236, 1985.
11-3
-------
Sexton, K., Spengier, J. D., and Treitman, R. D. Personal Exposure to Respirable Particles: A
Case Study in Waterbury, Vermont, Atoms. Environ., 18:1385-1398, 1984.
Shah, B. V., LaVange, L. M., Barnwell, B. G., Killinger, J. E., and Wheeless, S. C. SUDAAN:
Procedures for Descriptive Statistics. User's Guide. Research Triangle Institute, Research
Triangle Park, NC, 1989.
Skinner, C. J., Holt, D„ and Smith, T. M. F., eds. Analysis of Complex Surveys. Wiley,
Chichester, England, 1989.
Spengier, J. D., and Socozek, M. L. Evidence for Improved Ambient Air Quality and the Need
for Personal Exposure Research, Environ. Sci. Technol. 18:268a-380a, 1984.
Tepping, B. J. Variance Estimation in Complex Surveys. In: Proceedings of the American
Statistical Association Section on Social Statistics, pp. 11-18, 1968.
Thompson, C. R., Hensel, E. G., and Katz, G. Outdoor-Indoor Levels of Six Air Pollutants, J. Air
Pollution Control Assoc.. 23:10-18, 1973.
Tosteson, T., Spengier, J. D., and Weber, R. S. Aluminum, Iron and Lead Content of Respirable
Particulate Samples from a Personal Monitoring Study, Environ. Int. 2:265-268, 1982.
U. S. EPA. Air Quality Criteria Document for Particulate Matter and Sulfur Dioxides. Volume 2.
ECAO, RTP, NC. EPA 600/8-82-029b, 1982.
U. S. EPA. Quality Assurance Handbook for Air Pollution Measurement Systems. Volume 2.
(Ambient Air Specific Methods). Sections 2.10 and 2.11, U. S. Environmental Protection Agency,
Research Triangle Park, NC, 1977.
Wallace, L. A. The Total Exposure Assessment Methodology (TEAM) Study: Summary and
Analysis. Volume I. Final Report. EPA Contract No. 68-02-3679, 1986.
Wallace, L. A. The Total Exposure Assessment Methodology (TEAM) Study: Summary and
Analysis. Volume I. EPA/600/6-87/002a., Office of Research and Development, U.S.
Environmental Protection Agency, Washington, DC, 1987.
Whitmore, R. W., Jones, S. M., and Rosenzweig, M.S. Final Sampling Report for the Study or
Personal CO Exposure. NTIS PB84-181957. Research Triangle Institute, Research Triangle Park,
NC, 1984.
Williams, R. L., Folsom, R. E., and LaVange, L. M. The Implication of Sample Design on Survey
Data Analysis. In: Statistical Methods and the Improvement of Data Quality. Tommy Wright, ed..
Academic Press, New York, NY, 1983.
Williams, R. L., and Chromy, J. R. SAS Sample Selection MACROS. In: Proceedings of the Fifth
Annual SAS Users Group International Conference, pp. 392-396, 1980.
Yocom, J. E. Indoor-Outdoor Air Quality Relationships, J. Air Pollution Control Assoc. 32:500-
520, 1982.
11-4
-------
f
APPENDIX A
TABLE OF CONTENTS FOR OMB SUPPORTING STATEMENT
-------
-------
TABLE OF CONTENTS
Paoe
A. JUSTIFICATION FOR THE STUDY A-l
A.l Need for the Information Collection A-l
A.2 Description and Practical Utility of the
Information Collection Activity A-3
A.2.a The Information Collection Activity A-3
A.2.b Uses of the Data A-9
A.3 Minimizing Burden A-10
A.4 Nonduplication A-ll
A.5 Consideration of Alternatives A-12
A.6 Minimizing of Burden for Small Businesses A-12
A.7 Consideration of Collection A-12
A.8 Paperwork Reduction Act Guidelines A-12
A.9 Consultations A-12
A.10 Confidentiality A-15
A.11 Sensitive Questions A-15
A. 12 Cost to the Government and to the Respondents A-15
A. 12.a Cost to the Federal Government A-15
A.12.b Cost to the Respondents A-15
A. 13 Estimation of Respondent Burden A-16
A.14 Reasons for Change in Burden A-16
A.15 Scheduling A-17
A. 16 Standard Industrial Classifications A-17
B. STATISTICAL METHODOLOGIES B-l
B.l Description of Activity B-l
B.2 Procedures Used for the Collection of Information B-2
B.2.a Stratification and Sample Selection B-3
B.2.b Estimation Procedures B-15
B.2.C Expected Precision of Parameter Estimates B-17
B.3 Nonresponse, Maximization of Response Rates, Accuracy
and Reliability B-22
B.4 Test of Procedures B-22
B.5 Other B-23
B.5.a Tabulation, Analysis, and Publication Plans B-23
B.5.b Individuals Consulted on Statistical Plans B-23
B.5.C Contractor B-24
C. REFERENCES C-l
APPENDIX A Questionnaires and Participant Consent Forms A-l
APPENDIX B Motivation for P-TEAM Target Population Decisions B-l
APPENDIX C Data Analysis C-l
A-l
-------
A
-------
APPENDIX B
PARTICIPANT CONSENT FORM
-------
A
-------
UMB wo.: *UttU-UU,}/
Expires: June, 1992
THE PERSONAL EXPOSURE STUDY OF AIRBORNE PARTICLES IN RIVERSIDE
Participant Consent Fore
I understand that the Research Triangle Institute, (RTI) under contract from the
United States Environmental Protection Agency (EPA), is engaged in a study of the
potential exposure to certain substances by residents of Southern California living in
areas having varying levels of these substances in the environment. I understand that
this study is being conducted in order to help measure the levels of exposure to the
selected substances in populations environmentally exposed, and is limited to the
purpose stated. I further understand that the survey is b^ng conducted in
cooperation with and under co-sponsorship of the California ir Resources Board
(CARB).
I do hereby freely consent to participate in this study of potential exposure to
selected chemical compounds and substances and understand that my participation will
consist of providing some or all. of the following data: (1) answers to questions
related to environmental exposure and work and living conditions, (2) a record of my
activities and locations during the time that 2 am being monitored, (3) responses to
supplementary questions about activities of interest that I have undertaken, (4)
samples of the air that I breathe collected through the use of a personal exposure
monitor (PEM), and (5) samples of the air inside and outside my home collected through
the use of a fixed location, micro environmental monitor (MEM).
I understand that an agent of the Research Triangle Institute will administer the
questionnaire in my home, and at the same time make arrangements regarding collection
of the environmental samples. I understand that I will receive an incentive payment
of one hundred dollars for my complete participation. 1 understand that a small
number of households will be selected for the collection of duplicate samples (to be
collected at the same time as the original samples) but that such selection would not
entitle me to further compensation.
I understand that my name will not be voluntarily disclosed, and that my name will not
be referred to in anyway when compiling and evaluating the results of the study. I
understand that participation in this study may result in no direct benefits to me,
other than the results of my sample analyses which I will receive upon written
request, and that I am free to withdraw at anytime. It has been explained to me that
there are no significant risks to me from participation in this study. I further
understand that while participating in this study I will be free to ask any questions
concerning the study. If I have any further questions about the project, I know that
I am free to contact:
Harvey Zelon, RTI
Michele Hoffman, RTI
| Telephone
Lance Wallace, USEPA Telephone (703) 349-8970. or
Liz Ota, CARB Telephone (916) 323-1503.
Participant Name:
(Print)
(Signature)
Participant ID:
Date: / /
Chemistry ID: (PLACE LABEL HERE)
IIHnacc •
Guardian:
-------
-------
APPENDIX C
SUPPLEMENTAL CONSENT FORM FOR GUARDIANS OF MINORS
-------
-------
THE PERSONAL EXPOSURE STUDY OF AIRBORNE PARTICLES IN RIVERSIDE
SUPPLEMENTAL PERMISSION FORM
FOR GUARDIANS OF MINOR
I understand that my minor child has been selected for participation in this
study. I have read and listened to all explanatory material and have had the
opportunity to ask question about my child's participation. All aspects of
the study have been fully explained to my child and me.
I do hereby freely consent to have my child participate in this study.
Child's Name:
Participant ID:
Guardian Name:
(Print Name) (Signature)
Date: / / Witness:
-------
-------
APPENDIX D
REFUSAL DOCUMENTATION AND CONVERSION FORM
-------
-------
REFUSAL DOCUMENTATION AND CONVERSION FORM
A. Case Identification:
1. Respondent's ID: -
2. Respondent's Name:
3. Respondent's Address:
4. Telephone Number:
B. Initial Refusal Information:
1. Date of Contact: / /_
2. Time of Contact:
3. Reason for Initial Refusal:
4. Strength of initial refusal:
Strong Moderate, but firm Moderate, not firm Mild
5. Estimate of probability of success of conversion attempt:
High Medium Low
6. Things to try during recontact:
7. Things to avoid during recontact:
8. Disposition (FS USE ONLY):
Recontact by new FI: Assigned to: Date:
Recontact by FS
Recontact by Home Office: Date called to RTI:
No further contact
n-1
-------
-------
APPENDIX E
LEAD LETTER AND INFORMATIONAL BROCHURE SENT TO
POTENTIAL PARTICIPANTS
-------
-------
„1t° tf»,
i -Sm* I UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, O.C. 204«0
September,1990
Dear Riverside Resident,
OFFICE Of
RESEARCH AND DEVilOPMENT
As you may know, smog and haze In Riverside and other areas in the Los
Angeles basin are harmful to the health and well-being of our citizens. He
have recently become aware that Indoor air pollution can also cause health
problems. The U.S. Environmental Protection Agency (EPA) and the
California Air Resources Board (ARB) have joined forces to carry out a
Bajor study of air pollution in Riverside. This study has the full support
of the Riverside City Council.
We are writing to ask for your cooperation in the study. After a brief
Initial interview, some persons will be invited to take part In a second
phase of the study. If you are invited and choose to participate in the
second phase, you will receive one hundred dollars ($100) fonyour help.
Your family has been chosen as one of a small group of families to
participate in the first part of the study. Each household in Riverside
was given the same chance of being selected. Because you will be
representing other families like your own, your participation is vital to
the success of the study.
In the next week or two, an interviewer from Research Triangle
Institute, which is carrying out the study for us, will visit your home and
ask for your cooperation. The interviewer will be wearing an
identification badge. This first interview will normally take less than 15
minutes. All of the information you provide will be kept confidential.
The interviewer will be glad to answer all your questions.
For the second part of the study, one member of some of the sampled
families will be asked to wear a small monitor for 24 hours to measure his
or her exposure to air pollution, and will be asked to provide information
on his or her activities during that time. In addition, these families
will be asked to allow monitors to be placed in their homes to measure
pollution levels. The monitors are completely harmless, and all the
information obtained will be kept strictly confidential.
We urge you to take part in this important scientific study. Further
details of tne study are given in the enclosed brochure. If you have any
questions, please call one of the numbers given at the end of the brochure.
Erich Bretthauer, Ph.D. John R. Holmes, Ph.D. Terry*trizzel
Assistant Administrator Chief, Research Division Mayor
Office of Research California Air City of Riverside
and Development Resources Board
IIS Environmental Ptwm4on**cyti*i>»
-------
-------
VHP 13 SK/NSOR1NG THE STUDY?
The U.S. EPA and the California ARB contracted
¦nil Research TYiangle Institute (KIT) and Harvsrf
University School of Public Health to carry out the
study. Itnbtm-foriNofiireaeaichinstiiiMeln
North Carolina that has previously conducted studies
of people's exposures is chemicals. The Harvard
Univwity School of Public Health is wortd-
rcnownedft*Miiworic on indoor air pollution caused
by panicles.
WILL MY DATA BE KEPT
CONFIDENTIAL?
Yes. Hie eoimcwrtwfll keep aO the information
yon provide completely confidential. Oidyyonwit
receive data about yourself and your home. No one
el*, not even (he sponsoring government agencies,
wiB ever know who participated In die study.
WHO DO 1 CONTACT IP I WANT
MORE INFORMATION?
If yon would Hke more informuion ahout any pan of
the study, or have any questions about the study and
your participation. pleMe feel free 10 call:
Mr. Harvey Zeion
Project Manager
Research Triangle Institute
Telephone-Toll free: (800)334-8571
Dr. Lance Wallace
EPA Project Coordinator r
UJS. Hnvinjnnteatal ProtectkM Agency *
Telephone-(703) 349-8970
MiLfaOta
ARB Project Coordinasor
California Air Resources Board
Telephone -
-------
-------
APPENDIX F
TABLE OF CONTENTS FOR FIELD INTERVIEWER
INSTRUCTION MANUAL
-------
-------
TABLE OF CONTENTS
PAGE
1. OVERVIEW 1-1
1.1 Introduction to the Study 1-1
1.2 The Role of the Interviewer 1-2
1.3 Outline of the Manual 1-2
2. PROJECT INFORMATION 2-1
3. ADMINISTRATIVE PROCEDURES 3-1
3.1 Introduction 3-1
3.2 Employment 3-1
3.3 Equipment and Supplies 3-1
3.4 Production, Time, and Expense Reporting 3-2
3.4.1 Reporting Production 3-2
3.4.2 Reporting Time 3-2
3.4.3 Reporting Expenses 3-4
3.4.4 Allowable Time and Expense Charges 3-4
3.4.5 Other Time and Expense Considerations 3-5
3.4.6 Procedures for Paying Interviewers 3-6
3.4.7 Expense Advances 3-7
3.5 Interviewers' Responsibilities -. 3-7
3.5.1 Assignments 3-7
3.5.2 Scheduling Work 3-7
3.5.3 Maintaining Contact with Supervisors 3-8
3.5.4 Verification of Your Work 3-8
3.6 Professional Ethics 3-9
3.7 Confidentiality 3-9
3.8 Forms' Handling 3-10
4. HOUSEHOLD SCREENING 4-1
4.1 Overview 4-1
4.2 Sample Assignments 4-1
4.3 Contacting the Sample Housing Units 4-3
4.4 Completing the Household Enumeration Questionnaire (HEQ) 4-11
4.5 Selecting a Study Participant 4-13
4.6 Recruiting the Selected Participant 4-14
4.7 Appointment Schedules 4-19
4.8 Refusals 4-26
4.9 Validations 4-29
5. ADMINISTERING THE STUDY QUESTIONNAIRE 5-1
5.1 Overview 5-1
5.2 Question by Question Specifications 5-1
5.3 Final Steps 5-3
-------
-------
/HAT 13 Tht3 STUDY?
ftiisisamajorsiady of human exposure IB environ-
nental pollution. lis goaf ki to cotleet Information
m people** actual apnum to selected pollutants
10 that Ik* IIS. Environmental Protection Agency
[EPA) tad the California Afar Resources Boar*
(ARB) on develop Ibt best «r»le|in for
prelecting publk health. Similar audio have been
ooducted in cities across the United Stales, includ-
ing jewil m California. This is the first large-scale
study to measure the amount and kind of particles
ihat people are breathing as they go about their
roniul activities.
WHAT PARTICLES ARE BEING
MEASURED AND WHY?
T1* study wffl measure the tiny particles which are
typically proem in die air and which can be breathed
deep Into the tangs. These include dust, smoke, soot,
ash, poUen, fungi, and fibers, they may contain
varous chemicals and metals, and they may cause
difficulty breathing, especially for people with
asthma, alltagies, and other respiratory diseases.
WHY SHOULD I PAn flCIPATE?
Because you have been selected to represent many of
your fellow Riverside residents, your participation Is
very important to the success of the study. You will
have the satisfaction of knowing you have contrib-
uted to an important scientific study—very few
people ever have such a chance. Many of our
participants in previous studies have said they
enjoyed perticipeiing and learning about (heir sources
of exposure.
We realize thai you wiD need lo spend some time
answering our questions and will be carrying a
monitor around for 24 hours—therefore, we win
give yam $100.00 Immediately after jrou complete
the mowltoring period. You wilt also receive a copy
of an EPA publication. The Inside Story, which
provide* many valuable hints about how to improve
the quality of the air in your home. After all the
samples collected in the study have been analyzed,
we will send you the information about the exposures ,
measured for you and your home.
WOULD I BE ASKED TO b .AY HON
FROM WORK OR TO CHANGE MY
ACTIVITIES?
Quite the contrary! It is important tons thai yon
behave exactly as normal on tfie day you panicipa
You can wear the monitor or keep it nearby tfurin]
your wort, social, and lecreatkml activities.
WHY DO PEOPLE NEED TO BE
MONITORED?
To determine health risks, we must first understan
people's exposures to pollutants. Because people
mobile, spending lime st home, at wort, in their c
shopping, etc., we can obtain an accwate estimate
people's exposure* only by having them wear
personal monitors as they move about during die i
In addition, studies have shown that the greatest
exposure to many pollutants comes front indoor
sourco and activioes, such as smoking, cooking,
©VK1 inl>RJBB£ flflfl 1 iiCtnCuOu^o ^^5
measure the air inside and outside your home lo
compare it to your personal air samples.
WHAT WILL I BE ASKED TO DO?
PHASE1
ESffty IfeCMKMIO fVCBB¥in| 011 UfuCINTO will PC
visited by an interviewer from Research IViangle
Instituu. a scientific research organization. The
initial interview will nonmtily tike less than IS
gmntiApg, The Interviewer may (hen ask on*
person in ytm household to be a participant in the
second phase of tbe study.
PHASE2
If yon w a member of your household Is invited to
participate in die second phase of die study, you
wiUbeaskedto wear or cany a small personal
monitor for 24 houn, as will 174 other residents of
Riverside. Similar monitofs will be placed in
yotar living room and outdoors for the came lime
period You will also be asked questions about
your activities while you were wearing the
monitor. One or two days before the monitors are
placed in yotr home, a technician will place a
small tube containing a harmless ami odorless
material in your home to measure the amount of
air moving into and out of your home.
WHAT ARE THE MONITORS LIKE?
The persona) ntoir » • a small, quiet, battery-
operated pump with a small fiher attached. Youc
wear it in the hip pack or small backpack thai we
provide. It weighs only a few pounds and is perfc
Safe. Tha fawVw g«d tiuMnnr mwiiirn ffg mill
boxes with pumps »d filters. Hiey wil be pbcet
out-of-the-way places in your tivint area and yart
Because they are made of harmless materials, no
damage will result to you or your home under anj
circumstances. You are not liable if the monitor si
working or is damaged.
-------
-------
HAT IS THkS STUDY?
1u» is a major study of human exposure to environ-
ttntai pollution. Its goal Is to collect Information
n people's actual exposures to selected pollutants
»that the UJS. Environmental Protection Agency
EPA) and the California Air Resources Board
ARB) en develop the test strategies for
rotectiag public health. Similar studies tern been
onducted in cities across the Untied States, includ-
ig several in California. This is the first large-scale
tudy to measure die amount and kind of particles
iat people we breathing as tfiey go about their
crmal activities.
VHAT PARTICLES ARE BEING
MEASURED ANP WHY?
lie study wiO measure the tiny panicles which are
ftrically present in the air and which can he breathed
-------
-------
APPENDIX G
HOUSEHOLD ENUMERATION QUESTIONNAIRE
-------
-------
urus no.:
Expires: June, 1992
Research Triangle Institute
THE PERSONAL EXPOSURE STUDY OF AIRBORNE PARTICLES IN RIVERSIDE
HOUSEHOLD ENUMERATION QUESTIONNAIRE
A. Household Identification
(PLACE LABEL HERE)
Segment No: 1
D-D-D
Street Address
rm
SHU No:
City"
State
Zip
Sample Person No.: | j |
County
Telephone No.:
B. Record of Calls
Day
Date
Time
Result of Call
Code
FI ID Number
atn/pm
am/pm
am/pm
am/pm
C. Final Screening Result
D. Informant ID
FS USE ONL
Ineligible HU:
(Circle One)
Vacant 01
Not an HU 02 Section D
Temporary/Vacation Htm 03
Screening Not Completed:
Refusal 04
(Provide Documentation)
No one at home 05 FS
(after repeated visits) Approval
No eligible respondent 06
(after repeated visits)
Language Barrier 07
Other (SPECIFY) 08
Name:
Address:
FS
APPROVAL:
City State Zip
111111
Relationship/title:
Telephone Number:
No phone 00
Refused 01
Comments:
Verified ?
Yes ...01
No ....02
Date
of
Verificatioi
_/ /
Screening complete
W
NOTES:
-------
-------
E. Household Roster
BE SURE YOU ARE TALKING TO AN ELIGIBLE ENUMERATION INFORMANT; i.e. A FULL-TIME RESIDENT
OF THE HOUSEHOLD AT LEAST 16 YEARS OF AGE.
1. How many people currently live in this home? people
2. What are the names of all the people who currently live here? Let's list them in
order of age, starting with the oldest.
CHECKPOINT: DOES THE NUMBER OF NAMES IN THE ROSTER EQUAL THE ENTRY FOR QUESTION 1?
| | Yes - CONTINUE WITH QUESTION 3
~No - RESOLVE WITH RESPONDENT, CORRECT QUESTION OR ROSTER AS
NECESSARY, THEN CONTINUE.
3. For each person in the roster, including yourself, I need to know the following:
a. age (in years at last birthday),
b. the type of work of each person currently enployed outside of the home for a
least 30 hours a week or aore, and
c. current smoking status (Yes/No). (Yes ¦ at least 1 cigarette, cigar, or
pipeful per day)
INDICATE THE ENUMERATION RESPONDENT BY PLACING "R" IN THE MARGIN BESIDE THE NAME
Household Member Name
(a)
Age
(b)
Type of Work Outside The
Hone 30 Hrs a Keek Or More
(c)
Smoker(Y/N),
(d)
Consecutive #
for Non-Smoker
|
G-2
-------
4. Is there anyone else, such as a relative or roomer, who 4s currently living here, who
you did not list above?
Yes - ADD TO LIST AND COMPLETE ALL INFORMATION
No - CONTINUE
5. How many years of school has the head of the household completed? years
F. Sample Respondent Selection
(PLACE SAMPLING LABEL HERE)
INTERVIEWER: USE THE FOLLOWING SAMPLING INSTRUCTIONS TO DETERMINE WHICH, IF ANY, OF THE
ROSTERED MEMBERS SHOULD BE INCLUDED IN THE STUDY.
1. Consecutively number the non-smoking household residents 10 years of age or older
in column (d) of the roster.
2. Circle the number of non-smokers aged 10 or more on line 1 of the sampling label.
3. Circle the Y or N on line 2 i anted i ate ly below the circled number on line 1.
4. If circled letter is N, thank respondent for cooperation. Answer any questions.
Leave household. Circle code 20 in Section H.
5. If circled letter is Y, circle the person number on line 3 immediately below the
circled number on line 1.
6. Circle the selected number from line 3 in column (d) of the roster. This is your
selected respondent. At the same time, circle the respondent's name in column
(a) and enter the same number in the "sample person number" blanks on the label
on the top of the front page of this document. Enter selected respondents name
on line 1, Section 6.
7. If the entry 1n column (b) for the selected respondent indicates that he/she is
employed or if anyone in the household Is a smoker, continue with the contacting
and recruiting process. Otherwise, 1f the selected respondent is not employed
and no one smokes, use the Inclusion indicator on line 4 of the sample label to
determine if the respondent is to be included. If the Indicator is "Y", continue
with the contacting and recruiting process. If the indicator is "N", thank the
respondent, answer any questions, and leave. Circle code 20 in Section H.
8. Use the random day number found on line 5 of the sample label to determine the
selected monitoring day.
INSTRUCTIONS FOR THE MONITORING DAY SELECTION PROCESS ARE IN YOUR FI MANUAL. ENTER
SELECTED DAY IN SECTION g, LINE $, BELOW.
G-3
-------
6. Participant Recruitment
1. Selected Participant's Name:
2. Recruitment effort completed during enumeration visit.
Yes • GO TO 5
No - CONTINUE
3. Appointment to return to complete recruitment of the selected respondent.
Day: __ Oate: Time:
4. Appointment made by:
Selected participant Enumeration respondent
5. Selected Day for Monitoring
6. Monitoring Appointment:
Day: Date: Time:
H. Results of Recruitment Effort
10 - Participant agrees to participate on selected date
11 - Participant agrees to participate on other date
12 - Unable to schedule monitoring appointment on available dates
13 - No contact with selected participant after multiple attempts
14 - Refusal
20 - Household not selected
If refusal, indicate main reason:
6-4
-------
APPENDIX H
STUDY QUESTIONNAIRE
-------
OMB No.: 2080-0037
Expires: June, 1992
THE PERSONAL EXPOSURE STUDY OF AIRBORNE PARTICLES IN RIVERSIDE
Sponsored by:
Conducted by:
U.S. Environmental Protection Agency
Environmental Monitoring systems Laboratory
Research Triangle Park, N.C.
and
California Air Resources Board
Sacremento, California
Research Triangle Institute
Research "'angle Park, NC 27709
and
Harvard School of Public Health
Harvard University
Boston, MA 02115
QUESTIONNAIRE
The Research Triangle Institute and the Harvard School of Public Health
are undertaking a research study for the U.S. Environmental Protection
Agency and the California Air Resources Board to assess levels of human
exposure to particulate matter during normal daily activities. The infor-
mation recorded in this questionnaire will be held in strict confidence and
will be used solely for research into the effects of environmental factors
on public health. All results will be summarized for groups of people; no
information about individual persons will be released without the consent
of the individual. While you are not required to respond, your cooperation
is needed to make the results of this survey comprehensive, accurate, and
timely.
NOTIFICATION TO RESPONDENT OF ESTIMATED BURDEN
Public reporting burden for this collection of information is estimated to
vary fro® 3.5 to 4.0 hours per response, with an average of 3.75 hours per
response, including time for reviewing instructions and completing the
survey and log. Send comments regarding the burden estimate or any other
aspect of this collection of information, Including suggestions for reducing
this burden, to Chief, Information Policy Branch, PM-223, U.S. Environmental
Protection Agency, 401 M St., S.W., Washington, DC 20460; and to the Office
of Information and Regulatory Affairs, Office of Management and Budget,
Washington, DC 20503, marked "Attention: Desk Officer for EPA."
Participant ID #
Seg # SHU # CHK MHU Per ID #
Chemistry ID # (PLACE LABEL HERE!
H-l
-------
-------
The purpose of this questionnaire Is to obtain information about you, your
residence, your occupation and the environment in which you Mork. Ne are
asking the sane questions of each respondent in the study.
First, I would like to ask soae general questions about you.
1. Sex? (by observation) 1) HALE 2) FEMALE
2. What is your year of birth? 19
3. What is the last year of school which you completed? (CIRCLE ONE)
(IF CURRENTLY IN SCHOOL, INDICATE CURRENT YEAR}
Elementary 12 3 4 5 6
Jr/Sr. High 7 8 9 10 11 12
College (Tech School) 13 14 15 16 17 17+
4. To what ethnic group (race) do you belong?
(REC0R0 BY OBSERVATION, IF POSSIBLE)
1) White, Non-Hispanic Q
2) Hispanic
3) Black
4) Asian
5) Other (SPECIFY)
0
5. About when was your home originally built? Please consider when it was
originally built, not when it was remodeled, added to, or converted.
(CHECK ONE BOX)
1) 1985 to Present ^
2) 1980-1984
3) 1975-1979
4) 1970-1974
5) 1960-1969
6) 1950-1959
7) 1949 or Earlier
8) Don't Know
-------
6. Do you have:
a. an unvented clothes dryer located in the house
or an attached structure, such as a garage?
b. an unvented kerosene heater in the house or an
attached structure?
c. a fireplace or wood stove in the house or an
attached structure?
d. a whole-house or attic fan?
Yes
1
1
No
0 0
7. How many rooms do you have in your home? (Do not count bathrooms,
porches, balconies, foyers, or halls.)
rooms
8. Please list each room in your home and estimate the percentage of usable
floor space covered by rugs or carpets.
Room
% Of Floor Covered
bv Rug or Carpet
(Other areas not identified as a unique room.)
9. Do you have any pets, such as dogs or cats or other furry animals, which
usually spend some time each day in your home?
Yes
No
H-3
-------
10. How many people regularly smoke INSIDE YOUR HOME?
0°
01
02
3 or more
Finally, I have i few questions about your occupation and work environment.
11. Do you have a paid job outside of the hone?
11 Yes (CONTINUE)
2) No, self-employed in the home. (CONTINUE)
3) No, a full-time student. (CONTINUE)
4) No, full-time homemaker. (GO TO 20)
5) No, out of work just now, but usually employed. (60 TO 20)
6) No, retired or disabled (GO TO 20)
7) No, other (SPECIFY) (GO TO 20)
12. At the present time, how many hours per day and days per week do you
normally work at your primary job or attend classes?
1) hours/day 2) days/week
13. What are your primary job duties?
14. a. Which of the following describes your primary work or school setting?
jTj Indoors (CONTINUE)
jl] Outdoors (GO TO 15)
jTj In a vehicle (GO TO 16)
b. Which of the following best describes your indoor work or school
setting?
0 office, educational facility, medical facility
J2J warehouse, factory, plant
[i] retail, sales
m ...
-------
15. What 1s tht address of the specific location where your primary work
station is located? (Include the building name or number number, if
appropriate.) If student, indicate name and location of school.
ZIP
16. Do you have another job, or if employed, do you go to school part-time?
1) Yes (CONTINUE) 2) No (GO TO 20)
(IF RESPONDENT INDICATES BOTH SECOND JOB AND SCHOOL, ANSWER 17-19 FOR SECOND
JOB ONLY)
17. How many hours per day and days per week do you work during a normal week
at your second job or are you in school?
1) hours/day 2) days/week
18. What are your primary duties at this job?
19. a. Which of the following describes the work or school setting?
[Tj Indoors (CONTINUE)
JTj Outdoors (GO TO 20)
|7j In a vehicle (GO TO 20)
b. Which of the following best describes this indoor work or school
setting?
jT office, educational facility, medical facility
|Fj warehouse, factory, plant
jij retail, sales
Other, (SPECIFY)j
H-5
-------
20. Do you have a part-tine job or work as a volunteer?
1) Yes (CONTINUE) 2) No (GO TO 24)
21. How many hours per week do you work part-time or as a volunteer?
hours
22. What are your job duties at this job?
23. a.
Which of the following describe the work setting?
[T] Indoors (CONTINUE)
Outdoors (GO TO 24)
In a vehicle (GO TO 24)
Which of the following best describe this indoor work setting?
1 office, educational facility, medical facility
warehouse, factory, plant
retail, sales
Other, (SPECIFY):
0
24. INTERVIEWER: IF RESPONDENT REFUSES TO RESPOND OR DEMONSTRATES
RELUCTANCE, DO NOT ATTEMPT TO FORCE A RESPONSE.
As I indicated before, all your responses are strictly confidential, and
you may refuse to answer any question. In order to make comparisons of
groups of people, information about approximate household income is
important. Please estimate the total gross income of all members of the
household. Which of the following categories contains your estimate?
1. less than $10,000 Q
2. $10,000 to $29,999
3. $30,000 to $49,999 ^
[94] re
4. $50,000 to $79,999 £
5. $80,000 or greater
97
DK
H-6
-------
This completes the interview. Are there any questions which you have that 1
can answer? (ANSWER ANY QUESTIONS AND CONTINUE) Thank you very auch for your
cooperation.
Interviewer #
~~~~~~
Date of Interview |"| [| [
Comments:
H-7
-------
RECORD ANSWERS TO QUESTIONS A, 6, AND C BY OBSERVATION.
A. Which best describes this building?
1. A MOBILE HOME OR TRAILER
2. A ONE-FAMILY HOUSE DETACHED FROM ANY OTHER HOUSE
3. A ONE-FAMILY HOUSE ATTACHED TO ONE OR MORE HOUSES
4. A BUILDING FOR 2 FAMILIES
5. A BUILDING FOR 3 OR 4 FAMILIES
6. A BUILDING FOR 5 TO 9 FAMILIES
7. A BUILDING FOR 10 TO 19 FAMILIES
8. A BUILDING FOR 20 OR MORE FAMILIES
9. A BOAT, TENT, VAN, ETC.
10. OTHER, PLEASE SPECIFY;
B. Is the house located within 100 yards of a busy roadway?
Yes
No
C. Are any of the following sources of dirt located within 100 yards of this
house?
Yes No
1) Dirt drive r~l n
0 0
2) Other, (SPECIFY) [l] [2]
u o
-------
-------
12 HOUR TIME/ACTIVITY SURVEY
A. Identification
Participant QQ-Q^Q-Q-O-QQ
B. Monitoring Period 1
1. Date: / /
2. PEM Monitoring start time : pm
3. PEM Monitoring stop time : a*
C. Monitoring Period 2
1. Date: / /
2. PEM Monitoring start time : am
3. PEM Monitoring stop time : pm
1-1
-------
-------
B. Recall Diary
PERIOD NO.: I 2
TIME ACTIVITY
INSIDE
OUTSIDE
TRAVEL ON
ROADWAY*
TIME IN PRESENCE
OF SMOKING
DESCRIPTION
START
END
HOME
OTHER
NEAR HOME
OTHER
1
2
3
4
5
00 X
hr/mln
1
2
3
4
5
00 %
•
•
Sr/nHn
1
2
3
4
5
00 X
~Rr/jnTn"
1
2
3
4
5
00 X
-HF/itF*
1
2
3
4
5
00 X
•
hr/rain
1
2
3
4
5
• 00 X
hr/m1 n
1
2
3
4
5
00 X
Kr/nHn
Include:
iT
i'2'
(3)
time walking to and from vehicle
tine walking, biking, etc. along a roadway
tine in a car, bus, truck, etc. on a roadway
-------
C. Potential Exposure Influence During the Monitoring Period
1. a. Did you go to work during this monitoring period?
{TJ Yes (GO TO b)
[2] No (GO TO 2)
1. b. Of the time you spent at work during this monitoring period, how long
did you spend in the following locations?
INTERVIEWER: ENTER RESPONSES IN i-iii IN HOURS/MINUTES. ACCOUNT FOR
ENTIRE WORK DAY.
i) At your usual indoor work location /
ii) At your usual outdoor work location /
iii) At another work location /
2. I am going to read you a list of activities. For each activity, please
tell me if you did this activity yourself, were near where this activity
was done, or if this activity was done at your home while you were away.
INTERVIEWER: CIRCLE Y OR N BY EACH ACTIVITY AND FOR EACH ACTIVITY WITH A
Y CIRCLED, ENTER THE MINUTES THE RESPONDENT ENGAGED IN OR WAS IN THE SAME
AREA FOR EACH ACTIVITY AT HOME OR AWAY. ENTER -0- WHERE NECESSARY. NEAR
THE ACTIVITY IS 6ENERALLY WITHIN 100 YARDS OR CLOSE ENOUGH TO SMELL ODORS
FROM THE ACTIVITY.
Activity
a) vacuuming
b) dusting
c) carpet cleaning
d) lawn mowing
e) gardening
f) burning leaves
or rubbish
g) outdoor cooking-
grilling, frying,
barbecuing
Done or DONE NEARBY At Home,
Nearby At home Awav At home Awav Re so. Away
Y/N
Y/N
Y/N
Y/N
Y/N
Y/N
Y/N
1-3
-------
Activity
h) indoor cooking -
grilling, frying
Done or
Wearbv
Y/N
DONE
At home Away
NEARBY
At home Away
At Home,
Resp. Away
1) using clothes dryer y/N
j) woodworking
k) metal working,
welding
1) spray painting
m) other painting
n) outdoor recreation
o) other activities
near areas with
dust, smoke, or pollen
(SPECIFY);
Y/N
Y/N
Y/N
Y/N
Y/N
Y/N
3. Which of the following heating, ventilating, or air cleaning devices were used
in your home during this period? Please Indicate the duration of use, in hours,
for each Item used.
a) natural ventilation
(open doors or windows)
b) central air conditioning or heating
(Estimate actual time the device
was running.)
c) whole-house or attic fan
d) ultrasonic or cool mist humidifiers
e) filtration systems (including
filters, ionizers and electrostatic
precipitators)
f) unvented kerosene heaters
g) fire place
h) wood-burning stove
Yes
0
No
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
Duration
1-4
-------
4. How many cigarettes, pipefuls, and/or cigars were smoked in your house during
this period?
# cigarettes # pipes and/or cigars
5. Did you smoke any tobacco products (cigarettes, cigars, pipes) during this
period?
El N°
6. Were there any other sources of smoke present in the home during this period,
such as burnt food or candles?
[0 YeS
0No
7. a. Was a vehicle started or run in a garage attached to your home during this
period?
JT] NO (60 TO END)
[2] Yes (CONTINUE)
b. For how long? ninute(s)
c. Was it diesel powered?
1-5
-------
APPENDIX J
MEMORANDA TO THE RECORD
-------
-------
/m
RESEARCH TRIANGLE INSTITUTE ' 1 * 1
Center for Research in Statistics
April 23, 1991
MEMORANDUM
TO: The Record
FROM: Roy Vhitmore
@uo
SUBJECT: The final weight file for P-TEAM screeners
The final weight file for the P-TEAM screeners Is a SAS file on the RTI
VAX system. The file name 1s DISK5:[BLJPAT.SCREENDAT3]SHUDATA2.SASEB$DATA.
The SAS data set name 1s SHUDATA2. The file contains one record for each
of the 680 sample lines fielded for household screening. The variables
needed for record Identification and proper statistical analysis of the P-
TEAM screening data are the following:
Variable Description
STRATUM GEOGRAPHIC STRATUM (SW,NV,NEfSE)
SEGNUM SEGMENT NUMBER (01-36)
HID HOUSEHOLD ID
ASCRNVT SCREENING VT ADJUSTED FOR NONSCREENED ELIGIBLE HOUSING UNITS
Sampling variances should be computed treating the sample PSUs
Identified by SEGNUM as being selected with replacement within the explicit
sampling strata Identified by STRATUM.
Adjustments for Item nonresponse In the screener data base can be made
as ratio adjustments within the screener weighting classes, which were
simply the geographic strata Identified by the variable STRATUM. For all
analyses, overall adjustments that essentially delete the units with
missing data are satisfactory for low levels of Item nonresponse when
estimating means or proportions.
Figure J-l. Documentation of Final Weight File for PTEAM Screeners
Office Box 12194 Research Triangle Park, Nofh Carolina 27709-2194
telephone 919 541-5890 Fax 919 541-5965
J-l
-------
-------
RESEARCH TRIANGLE INSTITUTE
March 8, 1991
MEMORANDUM
TO: The Record
FROM: Roy Wtiitmore
Huu
SUBJECT: The P-TEAM final •eight file and Its use
The final P-TEAK weight file Is a SAS file on the RTI VAX system. The
file name is D1SK5:[BLJPAT.SCREENDAT3]SAMPHHS.SA$EB$DATA. The SAS data set
name is SAMPHHS. The file contains one record for each of the 257 sample
housing units for which a sample person was selected for Monitoring. The
variables needed for record Identification and for proper statistical
analysis of the P-TEAM data are the following:
Variable Description
STRATUM GEOGRAPHIC STRATUM (SV.NV.NE.SE)
SEGNUM SEGMENT NUMBER (01-36)
HID HOUSEHOLD ID
RID RESPONDENT ID
HWT3 Q HOUSEHOLD-LEVEL QUESTIONNAIRE ANALYSIS WEIGHT
PWT3~Q PERSON-LEVEL QUESTIONNAIRE ANALYSIS WEIGHT
HWT3"IM HOUSEHOLD-LEVEL INDOOR PARTICULATE MONITORING WEIGHT
HVT3~0M HOUSEHOLD-LEVEL OUTDOOR PARTICULATE MONITORING WEIGHT
PWT3"QM PERSON-LEVEL PARTICULATE MONITORING WEIGHT
CIWT? QM HOUSEHOLD-LEVEL INDOOR CARB SAMPLE WEIGHT
C0WT3~QM HOUSEHOLD-LEVEL OUTDOOR CARB SAMPLE WEIGHT
HEADE5CL HOUSEHOLD HEAD EDUCATION CLASS (1 - LT 12 OR . ; 2 ¦ 6E 12)
AGECL SAMPLE PERSON AGE CLASS (1 ¦ IE 25 \ 2 » 6T 25 OR . )
Sampling variances should be computed treating the sample PSUs
Identified by SEGNUM as being selected with replacement within the explicit
sampling strata Identified by STRATUM.
Adjustments for Item nonresponse for household-level questionnaire and
particulate monitoring data (weighted by HWT3_Q, HWT3_IM, or HWT3 OM) can
be made as ratio adjustments within the educatlon-of-head welghtlng'classes
(HEADEDCL). Likewise, adjustments Item nonresponse for person-level
questionnaire and particulate monitoring data (weighted by PVT3 Q or
PWT3 QM) can be made as ratio adjustments within the age-of-participant
weighting classes (AGECL). Adjustments for Item nonresponse for CARB
samples (weighted by CIWT3 QM or C0WT3_QM) can be made as overall ratio
adjustments* which will aTfect estimates of population totals, but not
estimates of means or proportions. For all analyses, overall adjustments
that essentially delete the units with missing data are satisfactory for
low levels of Item nonresponse when estimating means or proportions.
Posi Office Box 12194 Research "fciangle Park, North Carolina 27709-2194 Telephone 919-541-6000
Figure J-2. Documentation of Final PTEAM Weight File for Monitored Homes
1 O
-------
-------
APPENDIX K
TEMPORAL-SITE 12-HOUR AEROSOL CONCENTRATION DATA
-------
-------
Dup
Dup
T tiw
SAM
SAM
DIC/B
DIC/C
PEM
PEM
SAM
of
PM2.5
PM2.5
PM2.5
PM2.5
PMIO
PMIO
PMIO
Panod
Day
Cone
Cone
CONC
CONC
Cone
Cone
Cone
1
N
9.9
14.0
14.0
27.0
28.4
2
D
20.9
17.5
17.4
43.0
42.0
3
N
21.6
17.6
19.5
31 .9
39. 7
4
D
23. 1
24.0
24.6
47.7
5
N
22.1
21 .2
22.5
45 .4
48.9
6
D
29.9
39. 7
41.7
66.7
7
N
25.2
23.0
22.9
42.2
49.4
0
D
44.0
30.3
31.0
77.7
0 7.2
9
N
20.4
23.5
23.2
60.7
62.2
10
0
14.1
22.4
24. 1
42.0
11
N
11.5
17.5
41.7
47.5
12
D
32 .5
25.0
64 . 1
70.0
13
N
39.9
35.7
33.3
75.7
67.2
14
0
15
N
10.3
9.6
9.0
51 . 7
50. 2
16
D
23. 3
24.0
61.1
1 7
N
41.4
33 . 2
31.3
68. 1
71.6
10
D
24.0
22.5
21 .0
66 . 7
70.0
19
N
40.2
37 . 7
34.2
05. 7
91 .5
20
D
54. 0
51.9
40. 2
93. 9
96. 9
21
N
30.0
36.7
77.7
79.0
22
0
05.9
67. 1
6?. 6
145.9
135.4
130.9
23
N
120. 1
09.2
03.9
160. 7
179.1
24
0
45.5
28. 1
20.3
01.6
04 .0
25
N
60.5
61 .7
49.0
44.0
93.3
93.4
28
0
35.6
24.9
22.9
65. 5
27
N
20.3
16.4
14.7
49.4
49.2
20
D
36.0
39.0
24.0
69 .4
74.0
29
N
31 .7
25.0
24. 3
55.0
61.9
30
D
22.2
19.3
10.2
45.0
40.0
31
N
26.0
10.9
17.8
49.6
50. 7
32
0
46.4
11.3
39.2
109. 1
132.4
33
N
21 .0
6.5
6.3
20.9
29.0
34
D
16. 1
0.6
10.7
42.9
41.2
35
N
4.6
5.0
10.0
36
D
13.9
13T3
37
N
34.7
33.3
33.9
57.4
60.6
30
0
33. 1
27.4
27.5
73.9
75.2
39
N
39. 2
27.6
39.3
66 . 5
66. 7
40
O
•O. 6
66. 1
64.9
137.8
163.6
41
N
90.0
74.9
69.8
119.2
130.5
42
0
103. 1
04. 7
01 .6
158.0
145.5
43
N
55.6
42.2
40. 1
03 . 6
09.3
44
O
94.5
75.4
76.3
1 SO. 3
126. 7
45
N
37. 1
36.8
46
0
127.0
95.4
96. 1
171.7
176.6
47
N
109.0
73.6
73. 1
151.6
155.9
48
0
151.4
117.7
119.3
108. 7
207.6
49
N
70. 3
40. 1
42.6
97 . 7
102.9
Dup
SAM WED/A WEO/O D1C/B OIC/C SAM PEM SAM
WED DIC/B DIC/C
PMIO FINE FINE
PM10 PMIO
Cone CONC
28.
42
30.
45
37.
04
7
4
4
7
0
3
42.0
01 .2
40.0
47.0
20.2
59. 7
51.1
43.9
63.5
57.6
01.4
57.2
00. 0
54.9
130.4 147.4
121.2
93.0
77.8
72.4
40.6
66. 1
40. 7
39.4
21.2
95 . 5
23.B
33.6
17.8
70. 7
PMIO
CONC
26.0
39.9
27.9
42.4
34.4
06. 1
42.4
00.6
40. 7
47.6
26.6
61 .0
51.6
42.9
59.9
57 .5
62.4
50.0
03. 1
53.0
146.2
149.7
92.2
79.5
72.2
39.6
64. 1
39.9
40.0
24 . 0
95. 1
21.5
32.9
16.6
66. 7
54 .0
106 . 7
56.1 57.4
132.2 136.B
97.0 100.7
131.1 132.0
PMIO
CONC
32.0
39.0
35. 1
49.4
40.9
75.3
47.6
71.3
52. 1
47.5
44 . 5
60.3
64.0
50. 7
49. 3
62.6
62.5
76. 1
76. 9
67.4
120. 7
146.8
72
82
57
46
77
53
43
39,
127
25.6
35.0
7.9
74. 3
55.5
60.9
51.2
50.3
105.0
122.7 130.4
160.0 166.3
64.1 57.5
114.3
130.9
76.9
1 16.O
77
153
132
171
83
PMIO
CONC
31.0
30. 5
33 . 7
49 . 0
4 1.9
76. 1
44.6
71.4
49.0
50.2
PM2.5 PMIO PMIO A/0 CORS
Flag Flag Flag FLAG FLAG
CORS
FLAG
59.4
46.9
46.2
56.4
56.7
66.0
70.0
62.2
112.7
131.0
60 . 7
74.4
54.0
36.7
63.2
47 .3
40. I
36.9
23. 7
32.0
17.3
60. 7
54.0
66.9
76. 1
1 14
1 10
122
71
110
66
145 . 1
121.7
163.5
79.6
33
-------
Dup
Dup
T
SAM
SAM
OIC/B
DIC/C
PEM
PEM
SAM
Of
PM2.5
PM2.5
PM2.5
PM2 . 5
PMIO
PMIO
PMIO
P«riod
Day
Cone
Cone
CONC
CONC
Cone
Cone
Cone
50
0
82.4
77.7
74.7
126. 3
124 .8
51
N
35.4
25.3
24.9
52. 7
59.8
52
D
55.4
42.8
43.0
96.4
105.0
53
N
27.7
24.6
24.4
49.5
51.9
54
D
17.9
14.3
14.0
33.4
36.4
55
N
22.3
17.1
16.7
56
0
12.3
4.9
6.0
52.9
57.8
57
N
6.9
4.3
4.5
16.0
16 . 7
58
0
1 .4
3.4
3.9
19.3
18.2
59
N
7.2
5.7
8.7
24.9
23.9
60
0
25.2
19. 1
18.8
71.3
67.0
61
N
42. 1
35.5
34.9
58.5
61.1
62
0
22.8
17.2
16.6
65. 1
59.6
63
N
79.8
70. 7
67. 1
100.7
108.3
64
0
14.4
13.3
12.5
57.8
46 . 7
65
N
46.8
43.0
39.7
64.4
69.5
66
0
40. 5
32.3
32.8
90.8
85.3
91.0
67
N
76.6
75.5
69.4
102. 1
105.5
68
0
91 .0
71.0
71.6
145.1
157.9
69
N
111 .0
116.9
103.0
97.2
146.2
70
0
106.0
78.6
76. 1
148.8
160.3
71
N
170. 1
142.2
136.4
186.2
203.6
72
0
76.0
56.9
55.6
90.5
109. 1
73
N
177.4
139.5
137.2
1 96. 7
211 .0
74
0
93.0
73.9
72. 1
120.2
137.8
75
N
145. 7
127.4
117.0
167.5
182. 2
76
0
59.0
46.4
49.1
100.6
108. 2
77
N
15.0
15.4
40.9
45. 7
78
0
32. 1
27.8
27.4
72.8
77.5
79
N
26.8
23.5
23.2
49.0
50.3
80
0
8.5
9.4
9.8
34.0
37.4
81
N
13.5
8.9
8.8
63. 7
53.9
82
0
8.9
5.2
5.3
61.2
66.6
83
N
9.6
8.0
8.2
39.4
42.9
84
0
30.3
8.7
9.2
104.4
118.7
85
N
12.9
13. 7
14.4
35.4
36.0
86
0
16.8
12.2
12.7
36.5
35.0
87
N
37.9
35. 1
36.5
54.9
56.9
88
0
38.9
32.3
35.5
73.4
75.8
89
N
27.5
26.3
26.3
52.8
57.3
90
0
17.6
12.2
10.5
96.6
94. 7
91
N
6.9
4.6
4.8
42.4
44.9
92
D
72. 1
17.5
19.5
182.4
221 .2
93
N
6.5
8.0
8.5
26.0
26. 7
25. 7
94
0
29.0
29.9
22.9
50. 7
69. 5
73.1
95
N
53.0
48.3
49.5
81.0
86.4
96
0
15.7
6.4
6.6
28. 2
32. 7
Dup
SAM WEO/A WED/D DIC/B DIC/C SAM PEM SAM
PM10 PMIO
Cone CONC
55.4
88.2
26. 8
123.6
30.5
27.3
27.7
39. 6
17.1
16.9
16.3
62.5
48. 7
51.4
91.5
42 . 4
61.3
77.3
89.7
139.8
136.5
142.3
195. 1
97.7
198.6
125.5
168.6
96.6
34. 7
66.0
29.8
33.4
43.8
40.9
29.5
71.5
47.8
60. 7
36.9
65.0
29.4
150. 7
21.8
57.6
69. 1
25.6
PMIO
CONC
117.8
31.9
81 .9
29. 7
30.8
26.6
39. 7
16.4
14.5
15.0
56.5
49.8
48. 5
89. 7
40. 7
60.8
74. 7
90.0
139.4
137.2
139.2
192.3
98. 7
197.7
124.2
168. 1
95.5
33.7
67.5
32.6
33.2
43.6
43.4
29.5
72.5
31.3
30.4
48.5
67.0
35.7
65.9
28.3
155.8
20.9
59.4
70.0
24 . 8
PMIO
CONC
121.0
45.9
88 . 5
45.8
30.3
36
41
16
14
20
61.8
55.0
50.0
97.0
44 . 3
64.2
76 . 8
100.2
124.0
137.0
124.5
152.3
87.0
183.0
118.1
162.6
91 . 1
39.6
64.8
42.6
32.2
56.9
56.2
37.0
99.2
34.2
32.6
53.7
64. 7
51.2
88.6
40.9
203. 1
26.2
62. 1
75.2
28 . 3
PMIO
CONC
112.8
47.6
87.2
43.5
28. 1
33.8
40. 7
14 . 7
14.9
23.2
56.2
51 .2
47.3
87 . 7
38. 7
59.0
75.5
90.9
117.1
129.2
165.9
82.3
171.5
111.9
146.3
93.4
39.4
64.7
41.7
30.8
56.0
67.0
38.2
107. 1
37.6
33.2
53.0
70.5
49.3
90.8
36.6
250.0
30.2
66. 7
76.3
28. 9
MED DIC/B OIC/C
PMIO FINE FINE
PM2.5 PM10 PMIO A/D CORS CORS
Flag Mag Flag FLAG FLAG FLAG
11
8
-------
APPENDIX L
CASCADE 1MPACT0R (TEMPORAL-SITE) AEROSOL CONCENTRATION DATA
-------
-------
APPENDIX L
CASCADE IMPACTOR (TEMPORAL-SITE) AEROSOL CONCENTRATION DATA
Data from the Cascade impactor collected at the temporal site
consisted of four runs, as depicted 1n Table 8B-1. The concentrations
for the various cut sizes are shown. The Cascade Impactor results are
compared to the results for the other temporal site methods in Table 8B-
2. Since the Cascade cut sizes were not equivalent to the 2.5 and 10/im
sizes used for the other methods, these levels were estimated using
linear Interpolation. The Cascade levels varied greatly from the
levels of the other methods. For runs 1 and 4 and the lO^m levels of
run 2, the Cascade levels were higher. For run 3 and the 2.5/im level
for run 2, the Cascade levels were lower. For run 3, the Cascade level
was lower than the minima for the other methods, while for the 10^m
levels for run 1, the Cascade level was higher than the maxima for the
other methods.
i _i
-------
TABLE L-l. CASCADE IMPACTOR DATA
Run
1
Run
2
Run
3
Run
4
Sep 28
20:15
Oct 06
19:42
Oct 24
19:30
Oct 30
18:34
to
to
to
to
Oct 03
07:32
Oct 11
18:27
Oct 28
07:05
Nov 09
18:06
Cut Size
Cone.
Cut Size
Cone.
Cut Size
Cone.
Cut Size
Cone.
03)
(pm)
0*g/m3)
10.7
39.1
10.7
38.0
9.8
11.1
9.8
25.0
6.9
15.8
6.9
19.9
6.3
2.9
6.3
28.4
4.7
11.4
4.7
19.5
4.3
4.7
4.3
13.6
3.2
6.5
3.2
3.4
2.9
8.1
2.9
17.1
2.1
22.3
2.1
5.8
1.9
2.9
1.9
15.2
1.1
10.9
1.1
6.3
1.0
9.3
1.0
10.8
0.7
6.0
0.7
2.9
0.6
5.2
0.6
6.9
0.5
1.6
0.5
0.5
0.5
2.3
0.5
0.0
<0.5
0.0
<0.5
0.0
<0.5
0.0
<0.5
0.0
L-2
-------
TABLE L-2. COMPARISON OF CASCADE IMPACTOR RESULTS WITH OTHER METHODS
No. 12-Hr Cut Cascade Temporal Site Concentrations (tig/in3)
Run Periods Size Cone, (/ig/m3) Method N Mean Std Dev Minimum Maximum
9 2.5 /im
10 pm
10 2.5 /im
10 fim
7 2.5 ^m
10 ftm
20
10 pm
43.2
SAM
6
37.8
14.0
18.3
54.8
Dlchot
8
30.6
12.3
9.3
50.0
106.3
PEM
7
74.2
13.8
51.7
93.9
SAM
8
73.5
15.4
50.2
96.9
Dlchot
8
61.5
10.3
47.8
77.5
Wedding
8
58.7
11.1
43.4
82.0
16.7
SAM
8
28.9
9.6
16.1
46.4
Dlchot
10
18.3
9.6
4.8
33.6
89.3
PEM
9
53.4
26.4
18.0
109.1
SAM
8
63.6
31.1
29.8
132.4
Dlchot
10
52.2
32.0
12.6
127.2
Wedding
8
42.5
26.7
17.2
95.3
24.6
SAM
7
92.1
44.2
40.5
170.1
Dlchot
7
76.3
35.9
32.5
139.3
47.2
PEM
6
122.5
45.2
64.4
186.2
SAM
7
133.2
46.9
69.5
203.6
Dlchot
7
110.1
34.1
61.6
159.1
Wedding
7
119.7
46.3
61.0
193.7
43.2
SAM
19
24.5
17.3
6.5
72.1
Dlchot
20
18.2
12.7
4.7
48.9
118.4
PEM
20
62.2
35.8
26.3
182.4
SAM
20
66.6
43.2
26.3
221.2
Dlchot
20
61.4
44.0
28.2
226.6
Wedding
20
49.2
29.5
21.3
153.2
-------
-------
APPENDIX M
OUTDOOR (SAM), INDOOR (SIM), AND PERSONAL (PEM)
AEROSOL CONCENTRATION DATA
-------
-------
I . *
< N C
"fn
Ift
X •
n N t
o 01
555
(ft 0. U.
O 9
X - •
•ft a u.
O w
X - *
uf X **"
a cl u.
o x • u
3 < C
o
r> *>
m n
* »
«- l«-
*«•
CM
© to
<0 ©
» rj
tf) *>
<0
tn —•
ift
ox • o
3mn C
6«flr *
'*8
rf>
5«^
WJ5
OX O Q
3 « - C
OlA| 0
X O 0
« — c
•ft x o
o. o
)-0«( ^ ^ ® cd »ft *••
n «
K 9
*» *
a
6
•
r>
01 K
- ffl
r>
tf> —
f*
r>
*»
est
*
* n
M 91
) tO • C
40 X o «M
a o (si
252*
aS5
N«*N060r-'Ktf)0 citf>0»-*»«***f«hi*©r>PieD®*»%#f,l>*-«ft»f>©€>fM**"'©»-«f>© —
^«^AO«fnn^non9>oON
w *ti on Airt0NN«»ift009)^n
«ti*)iflir)Ci90iA
»- •- fs* .
i *• *
¦ °s
o
zozozozozozozozozozozozorozozozozozozozozozozozoz
*j
o
a
«»
C
c
o
« o
^ M
+>
u,
M-l
-------
2-W
Q)Nft<49Ui9V(a^9^^UlkUlkuliuri-N-*l«^fi»^Ol0OtfOOO(0Q»NO^a^9^autfl)^«
¦o
1
a
szazozozozozozozozozozozozozozozozozozozozozozozo ¦o-
X"I
-j •una«**kwu(OW'»iu*iuuw>iu<'»«i
X ¦>«»fikU««OWOk«»Ult<**atflHWIiMUI»','Q9*'N«w«kKi«*it»saNON-0«k«->U''«ui*(iii uoau--iau n OI
?3*
i|JU4ftOOUlW9U(*l
oao^^^uowoi'jiA-uutaktf
m *> o a> -• ¦*
musg l»fo lk
U W • — • fc O
AMmtf OA Oul<<^0
05
0 Z VI
3 - M
ft © Z
9
•sj
<*
*
<0
0 ? u> o
3 - C
OOIO
O9isa»oj«i»-4u)0«
U«^»4rjh)OONlO
«OmO
mOsg
><099 CBvi*-m*a>*0l0OloOi0v9CPl0i0OOD ^OaA\l>*4Ut9U
• —oo*j®o ——
)-*n4 uo&O-OsfcNUu^raa^^Auute^tuo^w
21*
3 — »
ft O Z
* a
-i 01
*M C* - - W
* u> O -• —
(«» W * W * t* <
-• CD — V OO-
o
<4
NjOO*w,^OOfc^7'--0(wO»^>N-» ^fary}raMN) C
OIO
o lot
3 N> m
ft • z
oi
O "0
0 z ©> o
3NnC
ft • z o
01
tg rv> rg *0
dim N «6
»w(^y •»<
(pi;vuou<
— CM
* ^
— ©-»v*r\a-»cr®'^w
•HiO^UaXICJ M • (A*»k^U)U
ft • Z
w»
a
o 3 w o
3 > C
ft' IO
ut
313
• - Z
40 O
n TJ l/J
II5
?1S
» - z
»
O ' z
I*
-------
tf)
* . p>
< X -
a u.
m
x • a>
*1 (V 0
"IE
O 0>
I — •
< z
a u.
o »
i - «
v) a u.
«•
o c
i - «
ui T 0-
o. £ u.
m
CM
in
o
a x
• u
•
•
•
3 <
CM C
in
n
o
2 0
n
cn
m
a o
r- in CM * tn — t© * cm ^ o m
® o
-ON CIN
#«- N«n-n
Ci^jeo <">4O<0«-tiO
r> M CM
*
^ ID
X
• u
•
• •
> « • •
•
¦ •
•
• • •
«•• •*«••
• • •
«
• •
<
M C
~ m <7> O •- O in 91 0 o ^
V)
X 0
cm cm "-i cm cm cm
«¦> ^
CM CO CI Hf>
i m t ^ ^ a> ^ jo
* N o
o
03
•
•
3 ~
C
©
O W
X o
N
n
a u
U*)
(DCDKlD^ttnOlOM^^
— m
— 03
— « <0 tf>
htfin9ie06(DN000N**
o
n n
X
• o
• •
• «
• « • •
•
• « • •
•
• • •
¦ «•••«•••
• • « . «
~
• •
**
C
r»NNnrjNfflpnow»-
O «H
® CM
n ^
CM
co^m
X o
(N f I W N 1 ^ •- •- PJ N r- n
a> «-
•»
cj *- «¦»
r>
•» CM w •-
O O <0
(On •¦
M m ffi N O N *
+ m •-
*
(0 (0 9)
« m •
m
«•
w"
•" ""
**"
a z
o u
ft
3 »-»
- c
•
•
«
O 1/5
X 0
*n
O
m
a u
CM
®
X
o o
<7> <0
N O O
" m
9i CO ^
6
9
1
e
2
3
0
1
3
o «- O r- c
CM «
- c
• *
• •
• * »
• • •
• * •
~ •••••••#
• «it*
»
• •
tfl
x o
•" ISi
e* o
N W ff»
in o
c ts <" no
«D
f- m
a. u
a9)n6tf)9) tmO
— cm
* « *
o en ^
lAoi^ocn
*¦*
•- CM —
^ «-
X
o o
e
f 1 o «•) r- n ^
* Cft
*
CM •— O
Nt«M ^ n N •-
o *
Ul
- c
*¦•«•••«•«•*»
•
• •
• • • • •
•
• • •
¦
• « •
• ••• •«••
< t « «
•
• •
a
X 0
iA(0"nO^iAON«O9)«-
CD
«- f-
* <© o «-
«
© f. -
r-
in o
or» «> io — tn <0 a> o
o w - o
e>
— 0)
a o
CM * N
•*<*)©« o
Ifl f) K "t
CM O
~> «¦» cm •• ••
•»
»•
CM
« *- •» « N
CM —
CM
«- #-
t <•- >.
-o* ZOZOZOZOZOZOZOZOOZOOOZOZOZOZOZOZOZOZOZOZOZOZOZOZO
t- o
¦>- r.eDi-«n09iof>o«io«io<-Nri-N4r-a'>ai0(iiO>'N<-M
i_ MNNNiMnNPiNnNntinnnnnnonnnnnnnnnnrinnnonnrnpun^ni****
~>
:c
0
a
o o
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
-------
in
X • a>
< ft m »- *-
i/> r. — " — — - ......_ n
a u.
i/»
x • o>
— X -
Q. u.
o o>
X - e «- — - —
X-B „-...- _ .... I.. - - - - *- .. - - - - .- .. ,- .. - - - - a - Z - ,. - - «- - - _____.-^
tniu
o w
X - •
a. a u.
rft •• h
Q X • O
3 < *
a u
>««i««iin«)OM()-«i-i(io#«o
< t\i c <•> no©»-r-»-m*'0>in©«n<7><»>*»- • b>*»o*»<**3* *> — «opii»o**",9,'"®0'<«o®*,#-'"*"«»«
O)J0 »-CFI«00>(|)OlOP4>0 «•!«> — *> — *> O « —«>«•>«**» Str^OK-^lfinmfX^inMNWIt'-CMNpi
in
a X • o
3 - w C
Q 40 J 0
ft
•O
in
X • u
"Ifl t
M X 0
a q
c^naj^c-oo®'* ^)«-p> — o«ffi <¦~» o
Nffl^©NC®O^(^PlJt'*'rt'-n'vi/lOC0ONC«-N*^
O1t»'>«0«iO<'-»<80K«CNCC«in4NMOt>'CNN<-O«l
^ •• CM
ftntoOscn^toMrv-oioip^eo^ot^flht
{Vir^aOO0>^OA^tn
O
©
10
X o rt
So -
Q I O U
3 ~ - C
uIAZ O
£ u
in x O A«9i«otw) *-9iO^NncnNOHMn^"CiAMC'tiOtf)«n«^NNic^9»ir>oin9)
CJ
X © O
« - C
V) X o
a a
iO0^0OtNntt9»^4p^tAf)^f)(D«f*ttO^«-NiA^ N • O N Ol Ifl •¦
NO>tNO«>«OO«'>®«#«"l®©JO*-«>**««l>*«><*»r-»l«H,»Wr»«,>rp>
0«o«O)f^aO9i«>o»«*)o^*mr4«0«Ar-^inr>9)^N««^^flOtt
-------
< (M «
U> X -
a u.
ift
X
•
9
CM
c
(/}
a
u.
o o>
I - « - -
X - e - - ^_ -
®
©
©
<0
ift
*
*
o c*
(0 r»
01
•• lA (M OD
CO -
40
iA^0N«
a* * *
*» +
trt
© r- ©
«r>
©
®
O
w
©
r*.
**•
®
CM P»
r*
CD
in
* © ©
l/J
X 0
a a
r> •-
CM •""
fi
CM •-
^ ci
— — *
o
0> <•> rw ^ —
cm r>
rj
i
3
4
5
4
2
0
3
+ o
—
1© r>
CM
o
O
«>
Ot <<) m
CD tp
CM *
ID
CM
r>
T"
P-
«•
<0
o -
m
<0
*-
CO © CO
(ft
X 0
a u
CM —
CM
""
•>•
*-
N (O »- f
CM * «
CM CM O
® if>
v"J •-
CD
10
CM
fM
(M
**
n n
•0
N
WON
O X o o
3 < «- c
Oti)l 0
©
©
X
o u
CM O
N
M
—
O*
e- CM
8
3
7
O *
m
— — ift «0 n
M> 01 ^
r- cm
m
_
«->
ri
CM
«0
C|
©
—
m
at o
<0
—
<
- c
• •
• •
•
•
•
•
•
• •
• •
« ¦ • • »
• • •
• ¦
•
•
•
•
•
•
•
*
«
«
•
•
•
« •
•
•
•
•
en
X 0
^ n
CM ^
V*
*
CM
CM ^
<0 (0 fl
* *
9*
^ CM rt CD CM
On A
o ®
m
*
iA
O
0>
©
Ift
CF>
&
<0
rt %n
ift
<0
a o
O tf)
m
*
tf)
CM
CM
fM ^
" <
iff «
<0 ft
T
tr»
n «0
O
«n
o
a x © u
3 «-« - C
O t/> £ 0
*
•ft
X
o u
tft o
P--
CM
^ o
*
CO CD s) 9> (0
in
ro «
O «D
<0 tt Ift
fM O
•»
NflN
«ft
®
•n
m
in
n
CM
«0
fj
©
tft
10
^ 01 CM
M
- c
» «
• •
¦
• ¦
«
•
• ••••*
•
•
• »
• »
• • •
• •
«
• • •
•
•
«
•
V
«
•
•
«
• .
•
«
*
, t •
to
X 0
o CO
CM CO
n
9v O
<0
CD
0) CD (0 CM
i0
*
® -
* ©
o 0
01
O h-
k
®
0i
*•
©
CM
o
< CO
CO
o eo *
a u
* «
a en
®
CM *
r>
ift
« « rj r) ^ ^
CM
f«.
10 «
*> 9
© <¦> r>
®
*
ift
©
®
c
0t O
®
o
0»
01 Ift CM
O V)
CO CO
CO
®
Ift «
«
CD •" •• ©
CM
o ^
0»
rs»
r*>
®
0»
CO
©
CO
C5
«> 0i
®
0»
10
© <0 f-
— C
t *
• •
•
• •
•
•
(
• ~ • • 4
• •
•
• • • • •
• • •
• «
%
• •
•
•
•
»
•
•
«
•
¦
• •
*
•
•
• • «
x o
® cn
co co
<0
fM •-
CM
o
^ 0> <0 o o
9 ®
CO
® — o ® —
CM —
a n
CM
o -
Ift
*
*
®
CD
••
O
**• ©
0»
®
CO
— ^ «
a v
^ CO
o ®
*
P0 CO
®
CO
c
co — o 0V O
CO O
•-
® ift ift © rj
o o ©
ift
CD «
©
01
(0
h»
0»
Ift
0» ®
®
r-
0»
miON
•-
CM
•»
w
— +*
••
- <•
#• •"
•-
CM
•-
••
«-
•-
rj —
o «
K Q
¦o
O
i.
«
a
OZOZOZOZOZOZOZOZOZOZOZOZOZOZOZOZOZ3ZOZOZOZOZOZOZO
M)iAtf)«OiAiAtf)iOiAiAiniAtf)iD4DiA«in«iOCefiCOO
Nn^n^ntnt^6«A«tn«iDc^ttr-(DKco
occtD«o«ccoc6«cec«ioovc
©
©
•"
w
CM
(M
CO
CO
hO
Ift
©
© f- f-
•«- 1-4
o©
o o
o
©©
©
o
o
o © © •- •-
tp» ^
W"
^ ^ ^ ^
^ ^*
9*
^ •— ^
M
CM
CM
CM
CM
Cm
CM
CM
CM CM
CM
CM
CM
CM CM CM
4J
•- w
•- •-
•-
•P. ^ ^ ^ •-
•" ««*
^ V ^
**
^
••
••
*-
••
*-
•"
•-
w ~-
~*•
-------
67
08
09
70
69
70
69
70
69
70
71
72
71
72
71
72
71
72
73
74
73
74
73
74
73
74
75
76
75
76
75
76
75
76
77
70
77
78
77
78
79
80
79
80
79
80
81
82
81
Dud
Dud
Oup
Duo
Tim#
PEM
SIM
SIM
SAM
SAM
SIM
SIM
SAM
SAM
of
PM10
PM10
PM10
PM10
PM 10
PMZ.5
PM2.5
PM2.5
PM2.!
Day
Cone
Cone
Cone
Cone
Cone
Cone
Cone
Cone
Cone
N
188.0
146. 1
171.1
99.7
117.8
0
248.9
144.8
168.9
75.4
102.0
N
136.6
147.1
167.9
1 10.3
117.9
0
251.0
221 .6
139.5
150.4
90.2
N
90.3
24.9
168.9
7 .9
6.5
120.7
D
264.8
97.0
167.4
88.9
101 .8
35.7
N
54.5
43.4
164.8
166. 1
31.9
135.6
D
85.0
34. 3
146. 1
20. 1
95.4
103.
N
211.5
139.5
165. 1
111.8
130.7
0
136.2
172.5
158.2
92.0
126.8
N
159.8
156.1
155.9
116.7
120.0
0
247.0
143.0
62.2
N
157.0
167.2
186.5
133.3
149. 1
D
137.8
128.6
112.7
81 .5
75.8
*
N
119.8
40.6
188.9
26.3
141 .3
D
94 .2
66.2
121.9
34.6
74.3
N
137.9
1 13.5
204.7
97 .8
149.0
D
100.0
144.7
145.6
80.6
98.5
N
121.3
95.6
200.3
163.6
0
178.9
99.6
148.0
54.8
99.4
N
81 .2
192.0
191.1
155.5
0
176.6
165.0
146. 1
94. 1
99.6
92.
N
145.7
106.5
222.9
69.2
69.8
151 .9
0
70.2
102.0
102.5
155.0
53.3
97.3
N
115.9
72.8
197.6
51.6
157.2
0
314.8
120.3
130. 1
63. 1
82.4
N
1 <0.6
122. 1
209.3
99.0
149.3
0
130.5
93. 2
45. 5
N
205.0
160.5
220.2
126.0
164.2
D
124.1
114.8
100.4
62.9
57.2
N
64.8
147.1
168.2
99.0
133.8
D
166.5
190.9
99. 1
142.1
58.3
N
97.9
106.9
1 18.5
77.0
92.5
0
103.9
73.9
69.5
36.3
32. 2
N
37.0
31.7
67.8
13.5
28.8
0
110.4
75.5
103.6
21.8
4 1.4
N
65.2
30.2
36.7
76.4
16.9
23. 1
D
83.4
99.0
86.8
33.2
34.9
36.6
N
67.3
93.6
64. 1
25.5
20.8
22.
0
309.5
207.9
83.8
80.4
46.4
30.2
N
42.0
71.3
34.5
0
80.2
45.9
11.5
N
24.6
21.7
55.8
14.0
30. 1
0
129.7
56,5
34.0
18.9
7.6
N
62.7
57.7
63.3
32.3
31.8
D
43. 1
22.3
4 1.2
12.1
15.1
N
32.2
33.5
56.9
14.0
0
61.6
30.9
45.2
18.1
8.4
N
51.4
19.3
70.9
2.9
25.3
PEN
PM2.5 PM2.5 PM10 PM10 PM10 PM2.5 PH2.5
f lag
SAM
SIM
SAM
ag Flag Flag Flag
88
3
i
-------
T lute
PEM
ticioant
of
PMIO
ID
Par1od
Day
Cone
154A
82
0
104.6
155A
81
N
129.1
155A
82
D
45.7
1 56A
81
N
49.6
156A
82
0
65.8
157A
83
N
97.9
157A
84
D
248.8
158A
83
N
56. 1
158A
84
0
180.5
159A
83
N
57.2
159A
84
0
52.9
160A
83
N
34.3
160A
84
D
90.3
161A
85
N
78.5
161A
86
0
137.4
162A
85
N
54. 1
162A
86
0
80.2
163A
85
N
41 .8
163A
86
D
143.8
164 A
85
N
81.2
164A
86
D
118.8
166A
87
N
45.9
166A
88
D
35. 1
167A
87
N
164.3
167A
88
0
173.1
168A
87
N
66.2
168A
88
0
115.5
169A
89
N
67.4
169A
90
D
150.0
170A
89
N
37.4
1 7 OA
90
0
179.3
17 1 A
89
H
36. 7
1 7 1A
90
0
152.4
1 72A
89
N
30.1
1 72A
90
0
58.0
173A •
91
N
78.3
173A
92
0
1 7 4A
91
N
47.8
174A
92
D
216.5
1 75A
91
N
65.1
175A
92
D
281 .4
1 76A
91
N
20.4
176A
92
0
46.8
177A
93
N
47.8
177A
94
0
170A
93
N
38.3
178A
94
0
130.0
179A
93
N
42.9
179A
94
D
42.7
Dud
Dup
SIM
SIM
SAM
SAM
PM10
PMIO
PM10
PMIO
Cone
Cone
Cone
Cone
62.2
91.5
79. 1
85. 3
62.2
60.5
65.0
16.7
15.0
66.6
41.8
63.2
1 10.8
67.0
198.4
381 .4
25.2
70. 8
117.7
503.6
26.2
38.2
22. 1
51.6
53.0
109.9
77.5
57. 1
58.7
120.3
55.9
40.4
38.4
42.6
37.0
30.9
46. 1
98.8
101 .0
40.4
96.3
58.9
161 .0
52.8
28.8
40.2
19.0
60.5
81.5
123.8
73.3
57.7
91.0
69.9
52.6
96.5
89. 7
94. 7
140.0
42.9
6? .0
131.7
160. 1
53.4
63.0
65.8
54 . 8
81 . 1
24.2
62. 1
48.4
90.2
85.9
29.9
78. 1
97.8
42.6
22.6
130.9
84. 1
47.7
29.4
344.4
78.6
26. 7
32. 3
20.8
41 .5
72.8
35.5
78.8
78.5
14.1
24.9
48.9
Dud Duo
SIH SIM SAM SAM PEN SIM SAM SIM SAM
PM2.5 PM2.5 PM2.5 PM2.5 PMIO PMIO PM10 PM2.5 PM2 .5
Cone Cone Cone Cone Flag Flag Flag Flag Flag
.2
.5
. 4
.4
.9
. 5
2
.6
.5
10.2
55. 1
41.6
2.9
6
70
55
6
30
7
22
7
8
16.6
20.3
6.7
10. 1
7.2
36.8
65.6
35.2
14.1
6.2
53.7
63. 3
23. 7
24.6
19.8
20. 7
17.6
33.4
21.9
10.8
18.3
7.5
35. 1
27.2
19.5
37.9
6.0
169.4
3.9
8.7
29.9
6.9
15.8
6.3
7 .6
7.3
9.5
19.7
26.2
24. 1
10.2
14.5
11.2
19.8
163.9
19,7
184 . 7
14.3
32.8
17.6
8.9
9.5
12.2
14.2
16.9
20.4
10.3
21 .2
27.4
35.3
42.6
33 . 1
45.4
28. 5
35.3
46.3
39.0
35.4
31.9
24.4
8.6
25.5
8.8
27 . 1
7.4
58.2
16.6
19.7
12.1
26.0
31.6
4. 1
21.6
25.5
8.4
20.9
8
1
8
3
1
3
6 -
-------
-------
APPENDIX N
TABULATIONS OF QUESTIONNAIRE RESPONSES
-------
-------
TABULATIONS OF QUESTIONNAIRE RESPONSES
GENERAL QUESTIONS
Ql. Sex?
1)
2)
Hale
Female
Q2. Year of birth?
1900-1925
1926-1950
1951-1960
1961-1970
1971-1980
Q3. What 1s the last year of school
which you completed?
Elementary (1-6)
Jr./Sr. High (7-12)
College or Tech School
Q4. To what ethnic group (race) do you belong?
White, Non-Hispanic
Hispanic
Black
Asian
Other
Q5. About when was your home originally
built? Please consider when 1t was
originally built, not when it was
remodeled, added to, or converted.
1985 to present
1980-1984
1975-1979
1970-1974
1960-1969
1950-1959
1949 or Earlier
Don't Know
Sample
Frequency
74
103
14
37
49
50
27
14
87
74
109
45
11
9
4
15
7
23
12
23
35
39
22
Weighted
Percentage
42.5
57.5
9.1
19.3
29.7
27.1
14.8
9.1
50.5
40.4
58.8
28.8
5.2
5.4
1.8
11.0
3.1
13.2
4.9
12.8
19.5
23.3
12.3
M_1
-------
Sample
Frequency
Weighted
Percentage
Q6. Do you have:
b.
an uninvented clothes dryer Yes 10
located tn the house No 165
or an attached structure,
such as a garage?
an unvented kerosene heater Yes 1
in the house or an attached No 174
structure?
7.2
92.8
0.3
99.7
c. a fireplace or wood stove Yes 81
1n the house or an No 93
attached structure?
47.0
53.0
Q9.
Q10.
a whole-house or
attic fan?
Q7. How many rooms do you have
in your home?
Do you have any pets, such
as dogs or cats or other furry
animals, which usually spend
some time each day in your
home?
How many people regularly
smoke INSIDE YOUR HOME?
Yes
No
3
4
5
6
7
8
9
Yes
No
0
1
2
34
140
14
32
31
35
19
11
1
78
96
123
30
20
19.4
80.6
8.9
23.2
24.9
22.8
12.2
7.6
0.5
45.1
54.9
3 OR MORE 4
69.0
18.8
10.1
2.2
Qll. Do you have a paid job
outside of the home?
Yes
No,
No,
No,
No,
No,
No,
self-employed in the home
a full-time student
full-time homemaker
out of work just now,
but usually employed,
retired or disabled
other
104
4
29
14
10
12
4
54.3
2.9
16.6
8.9
7.4
8.2
1.6
-------
Sample
Frequency
Weighted
Percentage
Q12. At the present time, how many
hours per day and days per week
do you normally work at your
primary job or attend classes?
[WORKERS OR STUDENTS ONLY (Qll)]
Hours/day 1-7 46 34.1
8 64 45.7
>8 26 20.2
Days/Week 1-4 21 17.1
5 100 71.5
>5 15 11.3
Q14a. Which of the following describes
your primary work or school
setting?
[WORKERS OR STUDENTS ONLY (Qll)]
Indoors 112 82.0
Outdoors 14 9.0
In a vehicle 5 3.9
Don't Know 6 5.1
Q14b. Which of the following best
describes your indoor work or
school setting?
[INDOOR SETTING ONLY (Q14a)]
Office, educational
Facility, medical facility
Warehouse, factor, plant
Retail, sales
Other
Q16. Do you have another job, or
1f employed, do you go to
school part-time?
[WORKERS OR STUDENTS ONLY (Qll)]
Q17. How many hours per day and
days per week do you work
during a normal week
at your second job or
are you in school?
[WORKERS OR STUDENTS ONLY (Q16)]
Hours/day
Days/week
81 66.1
18 16.1
8 6.6
14 11.2
Yes 19 11.2
No 112 88.8
1-4 14 73.9
*5 5 26.1
1-4 13 65.9
£5 6 34.1
-------
Sample
Frequency
Weighted
Percentage
Q19a. Which of the following
describes the work or
school setting?
[WORKERS OR STUDENTS ONLY (Q16)]
Indoors
In a vehicle
Q19b. Which of the following best
describes this indoor work
or school setting?
[INDOOR SETTING ONLY (Q19a)]
Office, educational
facility, medical
facility
Warehouse, factory, plant
Retail, sales, repair
Other
Q20. Do you have a part-time job
or work as a volunteer?
Q21. How many hours per week do you
work part-time or as a
volunteer?
[ONLY YES RESPONSES TO Q20]
Q23a. Which of the following
describe the work setting?
[ONLY YES RESPONSES TO Q20]
Indoors
Outdoors
Don't know
Q23b. Which of the following best
describe this indoor work
setting?
[INDOOR SETTING ONLY (Q23a)]
Office, educational
facility, medical facility
Retail, sales, repair
Other
19 96.2
1 3.8
14 71.4
1 6.5
2 10.5
2 11.5
Yes 21 13.0
No 150 87.0
1-5 9 48.2
6-10 4 17.0
>10 8 34.8
17 78.6
3 17.6
1 3.8
9 51.0
1 5.3
8 43.8
N-d
-------
Sample
Frequency
Weighted
Percentage
Q24. Please estimate the total gross
Income of all members of the
household.
Less than $10,000 17 9.4
$10,000 to $29,999 58 35.2
$30,000 to $49,999 53 29.2
$50,000 to $79,999 38 20.9
$80,000 or greater 4 2.1
Don't know 6 3.1
-------
INTERVICWER OBSERVATIONS
QA.
A mobile home or trailer
A one-family house detached
from any other house
A one-family house attached
to one or more houses
A building for 2 families
A building for 3 or 4
fam111es
A building for 5 to 9
fami11es
A building for 10 to
19 families
A building for 20 or
more families
QB. Is the house located within
100 yards of a busy roadway?
Yes
No
QC. Are any of the following sources
of dirt located within 100
yards of this house?
1. Dirt drive
2. Other
Weighted
Percentage
0 0
120 78.9
10 7.8
1 0.6
8 6.0
2 1.0
4 2.3
7 3.4
59 38.8
94 61.2
Yes 9 7.3
No 136 92.7
Yes 21 18.5
No 95 81.5
Sample
Frequency
Which best describes this building?
N-6
-------
QUESTIONS PERTAINING TO THE NIGHTTIME MONITORING PERIOD
Sample Weighted
Frequency Percentage
Ql. Did you go to work during
this monitoring period?
Yes 5 2.3
No 172 97.7
Q2. For each activity, please tell
me 1f you did this activity
yourself, were near where this
activity was done, or If this
activity was done at your home
while you were away.
Personal Exposure Household Exposure
Done or Sample Weighted Sample Weighted
Activity Nearby Frequency Percentage Frequency Percentage
a.
vacuuming
Yes
4
2.4
3
1.3
No
174
97.6
174
98.7
b.
dusting
Yes
8
6.2
8
4.1
No
170
93.8
169
95.9
c.
carpet cleaning
Yes
0
0
0
0
No
178
100
177
100
d.
lawn mowing
Yes
1
0.5
No
177
99.5
e.
gardening
Yes
1
0.5
No
177
99.5
f.
burning leaves
Yes
0
0
or rubbish
No
178
100
g.
outdoor cooking-
Yes
1
0.4
grilling, frying
No
177
99.6
barbecuing
h. indoor cooking -
Yes
48
29.1
48
26.1
grilling, frying
No
130
70.9
129
73.9
1. using clothes
Yes
20
11.2
20
10.6
dryer
No
158
88.8
157
89.4
N-7
-------
NIGHTTIME MONITORING PERIOD (continued)
Personal Exposure Household Exposure
Done or Sample Weighted Sample Weighted
Activity Nearby Frequency Percentage Frequency Percentage
J.
woodworking
Yes
2
0.9
2
0.9
No
176
99.1
175
99.1
k.
metal working,
Yes
0
0
0
0
welding
No
178
100
177
100
1.
spraying
Yes
59
32.8
60
36.1
No
119
67.2
117
63.9
a.
other painting
Yes
1
0.4
1
0.5
No
177
99.6
176
99.5
n.
outdoor recreation
Yes
9
4.7
No
169
95.3
0.
other activities
Yes
6
3.4
3
1.7
near areas with
No
172
96.6
174
98.3
dust, smoke, or
pollen
N-8
-------
NIGHTTIME HONITORING PERIOD (continued)
Sample Weighted
Frequency Percentage
Q3. Which of the following
heating, ventilating, or
air cleaning devices
were used 1n your home
during this period?
Please indicate the
duration of use, 1n hours,
for each item used.
a. natural ventilation Yes 117 68.6
(open doors or windows) No 59 31.4
b. central air conditioning Yes 26 13.0
or heating (Estimate No 150 87.0
actual time the device
was running.)
c. whole-house or attic fan Yes 3 2.8
No 173 97.2
d. ultrasonic or cool mist Yes 4 1.8
humidifiers No 171 98.2
e. filtration systems (in- Yes 1 1.0
eluding filters, ionizers No 175 99.0
and electrostatic
precipitators)
f. unvented kerosene heaters Yes 0 0
No 176 100
g. fire place Yes 1 0.4
No 175 99.6
h. wood-burning stove Yes 0 0
No 176 100
N-9
-------
NIGHTTIME MONITORING PERIOD (continued)
Sample Weighted
Frequency Percentage
Q4. How many cigarettes, plpefuls,
and/or cigars were smoked
in your house during this
period?
# Cigarettes
Q5. Did you smoke any tobacco
products (cigarettes,
cigars, pipes) during
this period?
Q6. Were there any other
sources of smoke present
in the home during this
period, such as burnt
food or candles?
Q7a. Was a vehicle started or
run In a garage attached
to your home during this
period?
Q7b. If yes, for how long?
Q7c. Was it dlesel powered?
0 142 80.6
1-10 30 17.2
>10 3 2.2
Pipes and/or
Cigarettes
0 113 92.7
1 1 0.6
Don't Know 11 6.7
Yes 1 1.3
No 175 98.7
Yes 14 7.1
No 162 92.9
No 147 92.5
Yes 8 5.8
Don't know 3 1.7
minutes
1 6 81.3
2 2 18.7
Yes 0 0
No 9 100.0
N-10
-------
QUESTIONS PERTAINING TO THE DAYTIME MONITORING PERIOD
QIa. Did you go to work
this monitoring period?
Yes
No
Q2. For each activity, please tell
me if you did this activity
yourself, were near where this
activity was done, or if this
activity was done at your home
while you were away.
Sample
Frequency
62
115
Weighted
Percentage
34.3
65.7
Personal Exposure Household Exposure
Activity
Done or
Nearby
Sample
Frequency
Weighted
Percentage
Sample Weighted
Frequency Percentage
a.
vacuuming
Yes
44
26.8
44
28.8
No
134
73.2
133
71.2
b.
dusting
Yes
41
24.0
42
2 5.1
No
137
76.0
135
74.9
c.
carpet cleaning
Yes
4
2.4
5
3.9
No
174
97.6
173
96.1
d.
lawn mowing
Yes
5
3.4
No
173
96.6
e.
gardening
Yes
11
7.7
No
167
92.3
f.
burning leaves
Yes
0
0
or rubbish
No
178
100
g-
outdoor cooking-
Yes
6
4.6
grilling, frying
No
172
95.4
barbecuing
h.
indoor cooking -
Yes
75
45.6
75
42.8
grilling, frying
No
103
54.4
102
57.2
1.
using clothes
Yes
41
23.7
39
20.6
dryer
No
137
76.3
138
79.4
j.
woodworking
Yes
2
0.9
1
0.4
No
176
99.1
176
99.6
N-ll
-------
DAYTIME MONITORING PERIOD (continued)
Personal Exposure Household Exposure
Done or Sample Weighted Sample Weighted
Activity Nearby Frequency Percentage Frequency Percentage
k. metal working,
Yes
3
1.7
2
0.8
welding
No
175
98.3
175
99.2
1. spraying
Yes
73
44.0
72
38.9
No
105
56.0
105
61.1
m. other painting
Yes
4
1.7
4
2.7
No
174
98.3
173
97.3
n. outdoor recreation
Yes
37
23.1
No
141
76.9
o. other activities
Yes
27
15.1
11
5.1
near areas with
No
151
84.9
166
94.9
dust, smoke, or
pollen
N-12
-------
DAYTIME MONITORING PERIOD (continued)
Sample Weighted
Frequency Percentage
Q3. Which of the following
heating, ventilating, or
air cleaning devices
were used in your home
during this period?
Please Indicate the
duration of use, in hours,
for each Item used.
a. natural ventilation Yes 137 78.9
(open doors or windows) No 38 21.1
b. central air conditioning Yes 28 13.5
or heating (Estimate No 148 86.5
actual time the device
was running.)
c. whole-house or attic fan Yes 3 1.8
No 173 98.2
d. ultrasonic or cool mist Yes 6 3.8
humidifiers No 170 96.2
e. filtration systems (in- Yes 0 0
eluding filters, Ionizers No 176 100
and electrostatic
precipitators)
f. unvented kerosene heaters Yes 0 0
No 176 100
g. fire place Yes 1 0.4
No 175 99.6
h. wood-burning stove Yes 0 0
No 176 100
-------
DAYTIME MONITORING PERIOD (continued)
Q4.
Q5.
Q6.
Q7a.
Sample
Frequency
•MMMMMMMMMdMMtaMMMMMMMHMHMLh
How many cigarettes, pipefuls,
and/or cigars were smoked
1n your house during this
period?
Did you smoke any tobacco
products (cigarettes,
cigars, pipes) during
this period?
Were there any other
sources of smoke present
1n the home during this
period, such as burnt
food or candles?
Was a vehicle started or
run 1n a garage attached
to your home during this
period?
Q7b. If yes, for how long?
Q7c. Was it diesel powered?
# Cigarettes
0
1-10
>10
# Pipes and/or
Cigarettes
0
1
>1
Don't Know
Yes
No
Yes
No
No
Yes
Don't know
minutes
1
2-5
>5
Yes
No
Don't Know
147
21
6
110
1
1
10
1
174
9
166
125
31
5
12
15
3
1
26
1
Weighted
Percentage
84.0
13.2
2.8
91.6
0.6
0.6
6.9
1.3
98.7
4.6
95.4
74.4
22.8
2.7
41.4
50.1
8.5
2.5
94.8
2.7
N-14
-------
APPENDIX 0
CLUSTERING OF SMOKERS
-------
-------
APPENDIX 0
CLUSTERING OF SMOKERS
A question arose during analysis of the PTEAM data regarding the
clustering of smoking habits within households In the PTEAM population.
The question was whether smoking status tended to cluster within households
more than would be expected if people independently decided whether or not
to smoke. This question was examined using the data from the Household
Enumeration Questionnaire. Smoking status was determined for every person
listed on the household roster for all 443 households that completed the
enumeration (screening) questionnaire.
One way to analyze this question is to consider the distribution of
smoking status among the people of smoking age 1n the target population.
From Table X-l, we see that the percentage of members of the PTEAM
population aged 18 or older who are smokers is 26.0 percent (39,094 /
150,118). If the members of each household who were 18 years old or older
independently decided to be smokers or non-smokers, then the distribution
of the number of smokers for each household size would be the binomial
probability distribution (n = number of household members 18 years old or
older, p = 0.260).
Table X-2 presents the estimated distribution of the number of smokers
by household size, the corresponding binomial distribution, and the Chi -
square goodness-of-fit statistic for the hypothesis that the sample 1s a
realization of the binomial distribution for each household size. The null
hypothesis cannot be rejected for single-person households, which suggests
that people 18 years old or older who live alone independently decide
whether or not to be smokers. However, the null hypothesis is rejected at
the 5 percent level of significance for households containing 2, 3, or 4
household members 18 years old or older. Therefore, the data support the
alternative hypothesis that the household members aged 18 or older do not
independently decide whether or not to smoke when two or more live
together.
Another way to analyze the clustering of smokers within PTEAM
households 1s compare the percentage of households in which either no one
or everyone smokes with the percentage of households that contain both
0-1
-------
smokers and non-smokers. Table X-3 shows the estimated population
proportions for these two events, the proportions that would be expected If
the household members 18 years old or older Independently decided to be
smokers, and the ch1-square goodness-of-f1t statistic for households
containing either 2 or 3 members aged 18 or older. (The sample size was
not sufficient to test this hypothesis for households containing 4 members
aged 18 or older.) In both cases the null hypothesis that the sample is a
realization of the specified binomial probability distribution 1s rejected
at the 5 percent level of significance. Therefore, this analysis also
supports the alternative hypothesis that smoking status 1s more clustered
within households in the PTEAM population than would be expected by chance.
0-2
-------
Table 0-1. Estimated Total and Mean Number of Household
Members by Age and Smoking Status
Characteristic
Statistic
Smoking
Household
Nonsmoking
Household
Total
Number of Nonsmokers
Aged 10 or Older
Total
Std. Err.
41,674
5,217
97,274
13,035
138,948
16,019
Mean
Std. Err.
1.53
0.10
2.34
0.08
2.02
0.08
Number of Household
Members Aged 10 or
Older
Total
Std. Err.
80,768
9,935
97,274
13,035
178,042
19,399
Mean
Std. Err.
2.96
0.12
2.34
0.08
2.58
0.08
Number of Household
Members Aged 18 or
Older
Total
Std. Err.
67,549
8,545
82,569
10,603
150,118
16,264
Mean
Std. Err.
2.48
0.08
1.98
0.05
2.18
0.05
Number of Households
Observed
175
268
443
0-3
-------
Table 0-2. Distribution of Number of Smokers by Number
of Household Members Aged 18 or Older
No. Household
Members Aged
18 or Older
0
Nunfcer of Smokers
1 2 3+
Estimated Populatl
on Distribution
1
0.738
0.262
2
0.653
0.233
0.115
3
0.450
0.322
0.147 0.081
4
0.234
0.325
0.308 0.134
Binomial
(n = No.
18+; p - 0.260)
1
0.740
0.260
2
0.548
0.385
0.068
3
0.405
0.427
0.150 0.018
4
0.300
0.421
0.222 0.057
Chi-Square
Goodness-
-of-FIt
95-th
Statistic
Percentile
1
<0.01
3.84
2
11.34
5.99
3
25.98
7.61
4
17.56
9.49
0-4
-------
Table 0-3. Clustering of Smoking Status by Number
of Household Members Aged 18 or Older
Clustering of Smokers
No. Household Mixed
Members Aged All or Smokers and
18 or Older None Non-Smokers
Estimated Populatl
on Distribution
2
0.768
0.233
3
0.531
0.469
Binomial (n « No.
O
to
CM
•
o
II
a.
• m
+
00
n
2
0.615
0.385
3
0.423
0.577
Ch1-Square
Goodness-of-Fit
Statistic
95-th
Percentile
2
9.78
3.84
3
4.80
3.84
0-5
-------
-------
APPENDIX P
PLOTS OF XRF ANALYSIS BIAS OVER TIME
-------
-------
PCDIF |
1 +
C
0 ~
-4 ~
A
AA
A
B B
B
B
B B
-1 + B
AA B
B
C
C
-2 ~ C
C
.3 +
C
C
-5 | A
01APR1991 01MAY1991 01JUN1991 01JUL1991 01AUG1991
XRFDATE
Figure P-l. Percent bias for silicon in SRM 1832 by XRF analysis
over time. Associated analysis batch designated by
A, B, or C.
-------
PCDIF I
3.0 ~
2.5 ~ C
A
C
8 C
2.0 ~ C C
1.5 ~ B
1.0 +
-0.5 +
•1.0 ~
-1.5 +
A
B B
B
B C
A A B 8
0.5 + C
A
0.0 * A A
B
A B
01APR1991 01NAY1991 0MUN1991 01JUL1991 01AUG1991
XRFDATE
Figure P-2. Percent bias for silicon in SRM 1833 by XRF analysis
over time. Associated analysis batch designated by
A, B, or C.
P-2
-------
PCDIF |
10 +
8 +
6 +
4 4
2 +
0 ~
AA
B B
AA
A
-2 |
01APR1991 01MAY1991 01JUH1991 01JUL1991 01AUG1991
XRFDATE
Figure P-3. Percent bias for aluminum in SRM 1832 by XRF analys
over time. Associated analysis batch designated by
A, B, or C.
-------
PCDIF j
3 ~
2 ~
1 +
0 ~
-1 +
-2 ~
-3 ~
-4 +
-5 +
—+
01APR19S1
A A
A A
AA
+
01MAY1991
B B
4-.-...
01JUL1991
C C
CC
C
01JUN1991
XRFDATE
01AUG1991
Figure P-4. Percent bias for iron in SRH 1833 by XRF analysis
over time. Associated analysis batch designated
by A, B, or C.
P-4
-------
PCD IF (
8 ~
A
7
6 +
A
S ~ A
A
4 +
8
6
B
A B
A
3 ~ B 8 B
C
A
B C
A B C
2 B C
C
c
A C
€
1 ~
L. .
01APR1991 01MAY1991 01JUN1991 01JUL1991 01AUG1991
XRFDATE
Figure P-5. Percent bias for manganese in SRM 1832 by XRF analysis
over time. Associated analysis batch designated by A,
B, or C.
-------
PCDIF |
1.5 ~ B
AA
1.0 ~ B
B
A
0.5 + A B
A
A
0.0 ~ A A
-0.5 +
-1.0 +
B B C
B B
B
C
C C
C
c
...+ 4
01APR1991 01MAY1991 01JUH1991 01JUL1991 01AUG1991
XRFDATE
Figure P-6. Percent bias for potassium in SRM 1833 by XRF analysis
over time. Associated analysis batch designated by A,
B, or C.
P-6
-------
PCD IF
7.5 +
7.0 ~
6.S ~
6.0 ~
5.5 +
5.0 +
4.5 +
4.0 +
3.5 +
3.0 ~
I
AA
A
B
B
B B B
CC
C
01APRI991
01MAY1991 01JUN1991
01JUU991
01AJG1991
XRFDATE
Figure P-7. Percent bias for vanadium in SRM 1832 by XRF analysis
over time. Associated analysis batch designated by
A, B, or C.
P-7
-------
PCOIF
-1 +
-2 ~
-3 +
-4 ~
-5 +
-6 +
-7 *
-8 ~
-9 +
AA
A
b Be
B B
B
B
B
B
01MAY1991 01JUN1991
C
C
c c
. — —...
01JUL1991 01AUG1991
01APR1991
XRFDATE
Figure P-8. Percent bias for zinc in SRM 1833 by XRF analysis
over time. Associated analysis batch designated
by A, B, or C.
P-8
-------
PCDIF |
4,5 ~
4.0 *
3.5 * B
0
A B B
A B
A A
3.0 +
2.0 ~
l.S +
A
k
C
A
2.5 + C
C
C
01APR1991 Q1MAY1991 01JUN1991 01JUL1991 01A0G1991
XRFDATE
Figure P-9. Percent bias for titanium in SRM 1833 by XRF analysis
over time. Associated analysis batch designated by
A, B, or C.
Pv A
-------
PCD IF
Z +
I +
0 ~
-1 +
-2 +
-3 ~
-4 +
-5 +
-6 *
-7 ~
B B
+
01JUN1991
XRFDATE
C
C C
01APR1991 01MAY1991
01JUL1991 01AUG1991
~re P-10. Percent bias for copper in SRM 1832 by XRF analysis
over time. Associated analysis batch designated by
A, B, or C.
d_i n
-------
PCDIF j
12 +
10 ~
8 ~
6 ~
4 ~
Z +
o +
-2 +
.4 *
A
A A
A
A
AA
6 B
B
€
C C
C
C
c c
01AUG1991
0IAPR1991 01MAY1991 01JUN1991 01JUL1991
XRFDATE
Figure P-ll. Percent bias for cobalt in SRM 1832 by XRF analysis
over time. Associated analysis batch designated by
A, B, or C.
P-ll
-------
-------
APPENDIX Q
PLOTS OF DUPLICATE ANALYSIS ELEMENTAL CONCENTRATION RESULTS
-------
-------
ACONC
80000 ~
70000 ~
60000 +
50000 +
40000 +
30000 ~
20000 +
10000 +
+
•+.
0 20000 40000 60000 80000
CCONC
Figure Q-l. Concentration of silicon (ng/m3) in samples analyzed
with batch A or B versus duplicate analysis in batch C
(observations may be hidden).
-------
ACONC
40000 +
35000 +
30000 ~
25000 ~
20000 «¦
J5000 +
10000 +
5000 ~
—...4-.—. + ~ + + + ~
0 5000 10000 15000 20000 25000 30000 35000 40000
CCONC
Figure Q-2. Concentration of aluminum (ng/m^) in samples analyzed
with batch A or B versus duplicate analysis in batch C
(observations may be hidden).
Q-2
-------
ACONC
lOOOO ~
8000 +
6000 +
4000 +
2000 +
+
I
0 2000 4000 6000 8000 10000
CCONC
Figure Q-3. Concentration of potassium (ng/m3) in samples analyzed
with batch A or B versus duplicate analysis in batch C
(observations may be hidden).
0-3
-------
ACONC
25000 ~
20000 ~
15000 +
10000 «•
5000 +
+
I
.4.
0 5000 10000 15000 20000 25000
CCONC
Figure Q-4. Concentration of iron (ng/m3) in samples analyzed
with batch A or B versus duplicate analysis in batch C
(observations may be hidden).
Q-4
-------
ACONC
500 +
400 +
300 +
200 +
100 ~
0 100 200 300 400 500
CCONC
Figure Q-5. Concentration of manganese (ng/m3) in samples analyzed
with batch A or B versus duplicate analysis in batch C
(observations may be hidden).
-------
ACONC
250 +
200 +
150 ~
100 ~
50 *
+
.+¦
.+¦
•4.
0 50 100 150 200 250
CCONC
Figure Q-6. Concentration of lead (ng/m^) in samples analyzed
with batch A or B versus duplicate analysis in batch C
(observations may be hidden).
Q-6
-------
ACONC
6000
SOOO
4000
3000
2000
1000
0 1000 2000 3000 4000 5000 6000
CCOHC
Figure Q-7.
Concentration of sulfur (ng/m3) in samples analyzed
with batch A or B versus duplicate analysis in batch C
(observations may be hidden).
-------
ACONC
80 *
70 +
60 ~
50 ~
40 +
30 ~
20 +
AB Al
BA AA
10 +
.BBB
0 ~ '
I
0
70
-+
30
SO
60
80
10
20
CCONC
Figure Q-8. Concentration of bromine (ng/m3) in samples analyzed
with batch A or B versus duplicate analysis in batch C
(observations may be hidden).
r\
-------
ACONC
500 +
400 ~
300 +
200 +
100 +
ABA
0 ~
-+-
0
100
200
300
400
500
CCONC
figure Q-9. Concentration of zinc (ng/m3) in samples analyzed
with batch A or B versus duplicate analysis in batch t
(observations may be hidden).
n_o
-------
ACONC
2500 *
2000 +
1500 +
1000 ~
500 +
+
,+ .......—
0 500 1000 1500 2000 2500
CCONC
Figure Q-10- Concentration of titanium (ng/m^) in samples analyzed
with batch A or B versus duplicate analysis in batch C
(observations may be hidden).
Q-10
-------
ACONC
90 +
80 ~
70 ~
60 +
50 ~
40 *
30 +
20 +
10 +
0 +
0
B BA
BBA
AAABA*' A
BA AAM B
BA /AB
AAMBBBA
A A
—+-
20
40
—+-
60
80
..4-.
100
CCONC
Figure Q-ll. Concentration of strontium (ng/m3) in samples analyzed
with batch A or B versus duplicate analysis in batch C
(observations may be hidden).
Q-ll
-------
ACONC |
120 +
100 ~
80 ~
60 +
40 ~
20 +
0 +'
0
—+-
20
40
60
CCONC
80
-.4..
100
—+
120
Figure Q-12.
Concentration of copper (ng/m3) in samples analyzed
with batch A or B versus duplicate analysis in batch C
(observations may be hidden).
Q-12
-------
ACONC |
800 +
700 +
600 +
500 +
400 +
300 +
200 ~
100 +
_+ +......—4—— ~—— +— +•— .f
100 200 300 400 500 600 700 800 900
CCONC
Figure 0-13. Concentration of phosphorus {ng/m3) in samples analyzed
with batch A or B versus duplicate analysis in batch C
(observations may be hidden).
Q-13
-------
-------
APPENDIX R
BAR PLOTS OF MEAN ELEMENT CONCENTRATIONS FOR PEM, SIM,
SAM 2.5 AND 10 Jim SAMPLES
-------
-------
ng/Ki
4800
DAYTIME
NIGHTTIME
4200
3600
3000
2400
1800
1200
600
10|ira
2. 5|Un
SAM SIM PEM SAM SIM PEM
Figure R-l. Mean Calcium Concentrations.
R-l
-------
ng/m
80 -
70 -
60 -
50 -
40 -
30 -
20 -
10 -
0
-------
ng/m3
2000
1750
1500
1250
1000
750
500
250
0
SAM SIM PEM SAM SIM PEM
Figure R-3. Mean Potassium Concentrations.
0 A Y T I *4 i
NIGHTTIME
10/im
2.5j»m
-------
ng/»
40
DAYTIME
NIGHTTIME
35 -
30 -
25
20 -
15 -
10
5 -
lOftm
2.5/im
SAM SIM PEM SAM SIM PEM
Figure R-4. Mean Bromine Concentrations.
R-4
-------
ng/m3
48
42
36
30
24
18
12
6
0
SAM SIM PEM SAM SIM PEM
Figure R-5. Mean Lead Concentrations.
DAYTIME
NIGHTTIME
10^m
2.5 /*m
-------
ng/m^
160
140
120
100
80
60
40
20
0
SAM SIM PEM SAM SIM PEM
Figure R-6. Mean Zinc Concentrations.
DAYTIME
NIGHTTIME
10^m
2.5/tm
-------
ng/m^
400
350
300
250
200
150
100
50
0
SAM SIM PEM SAM SIM PEM
Figure R-7. Mean Titanium Concentrations.
0 A Y T I M E
NIGHTTIME
10/
-------
1
ng/nr
48
42
36
30
24
18
12
6
0
Figure R-8. Mean Copper Concentrations.
NI6HTT1HE
R-8
A
-------
ng/ii
32
DAYTIME
NIGHTTIME
28 -
24 -
20
16
12
8
10/tm
4 -
2.5/i m
SAM SIM PEM SAM SIM PEM
Figure R-9. Mean Strontium Concentrations.
-------
ng/m^
240
210
ISO
150
120
90
60
30
0
SAM SIN PEM SAM Sit! PEM
Figure R-10. Mean Phosphorus Concentrations.
R-10
DAYTIME
NIGHTTIME
10/im
2.5gm
-------
APPENDIX S
QUALITY CONTROL REVIEW
MEMORANDUM FOR XRF FILTER ANALYSIS DATA
-------
-------
RESEARCH TRIANGLE INSTITUTE
Center for Environmental Measurements and Quality Assurance
MEMORANDUM January 10, 1992
TO: Dr. E. D. Pellizzari and Dr. S. V. Kulkami
VV A
FROM: Richard C. Shores and Dr. Karen w. Gold
SUB J: Internal quality control review of PTE AM XRF filter analysis data from Riverside,
California.
The purpose of this memorandum is to document a quality control review of the XRF data
being reported to RH from EPA, DRI and LBL. Filters were analyzed by DR1 to determine if
the XRF analytical technique was going to provide sufficient resolution and detection of the
elements of interest to support the PTEAM objectives. The EPA (ManTech personnel) analyzed
all of the particulate samples and these results are considered to be die primary PTEAMs data.
Filters analyzed by LBL were considered to be a QC check on the EPA data.
The EPA and DRI data were corrected for attenuation or self-absorption effects due to
particle size and elemental composition of the samples before correcting for line overlap. RTI
personnel applied no collection factors to the EPA or DRI data.
LBL provided RTI with XRF analysis of three types of particulate filters: 1) dichotomous
fine, SIM and SAM (2.5 pm), 2) PEM, SIM and SAM (10 pm), and 3) dichotomous coarse
(10 pm > 2.5 pm). The attenuation correction factors discussed in this memorandum were
applied to the LBL data by RTI personnel. These correction factors should have been used by
LBL before correcting for line overlap. According to the enclosed memorandum from
Dr. R. Kellogg, this error should be less than die uncertainty reported with each elemental
analysis. Dr. R. Kellogg of ManTech conducted the EPA's XRF analysis of the PTEAM filters.
Sample analysis results for three elements (Al, Br and V) from a total of 10 filters were
reviewed, (five filters analyzed by both EPA and DRI ami five filters analyzed by both EPA ami
LBL). There were no sets of filters analyzed by both DRI and LBL.
XRF analysis of filter blanks were provided to RTI in units of ng/cm2. These data were
converted to concentration units of ng/m5 as described in this memo.
Dichotomous Fine, SIM and SAM (PMjj)
The dichotomous fine, SIM and SAM (2.5 pm) filter data were corrected for self-
absorption due to elemental composition of the samples as follows.
True Concentration »Indicated concentration/Correction factor
Post Office Box 12194 Research Triangle Park, North Carolina 27709-2194
Telephone 919 541-6914 Fax:919 541-5929
c. *
-------
Page 2 of 6
where:
Correction factor = (1 - exp (- MU * MASS/AREA))/(MU*MASS/AREA)
MU * mass absorption coefficient, given below (calculated assuming a
standard or typical elemental composition for a fine particulate
ambient air sample)
MASS = sample mass deposited on filter (pg)
AREA = area of filter (PEM, SIM and SAM @ 7.5 cm2 or Dichotomous @
6.8 cm1)
Correction
Element
MU fcmVue)
Factor Error
A1
0.00140
+/-
0.110
Si
0.00098
+/-
0.100
P
0.00077
+,t-
0.080
S
0.00056
+/-
0.030
ci
0.00077
+/-
0.030
At
0.00000
+/-
0.030
K
0.00047
+/-
0.030
Ca
0.00043
+/-
0.030
Elements not listed above received no attenuation correction.
PEM, SIM, and SAM (PM*)
PEM, SIM, and SAM (10 pm) particulate filter data were corrected using the following
equation and factors.
True concentration = Indicated concentration/Correction factor
where:
Element Correction Factor
A1 0.61 +/- 0.15
Si 0.67 +/- 0.13
K 0.93 +/- 0.10
Ca 0.94 +/- 0.05
S-2
-------
Page 3 of 6
The dichotomous fine (2.5 pm) particle correction factors v re applied to P, S, and CI.
All other elements had no correction factor applied to the data.
The PEM, SIM, and SAM (PMl0) correction factors were derived by EPA from a
comparison of dichotomous filter data (total, coarse, and fine, corrected for attenuation) and
PEM, SIM, and SAM filter data (not corrected for attenuation) collected over the same time
period at the same location (collocated data).
Dichotomous Coarse (PMW > PMjj)
Dichotomous coarse filters were corrected for attenuation by both the coarse and fine
particles using the following equation.
C, ¦= C, + F (C„)
where
ec
c.
cv =
Q =
AF *
K =
(F) CP
Corrected dichotomous coarse concentration (ng/m3)
Uncorrected dichotomous coarse concentration (ng/m3)
Flow fraction of the dichotomous sampler calculated as coarse flow/total flow.
This was a constant for each of the two dichotomous samplers (0.08 and 0.10)
operated.
Uncorrected dichotomous fine concentration (ng/m3)
Corrected dichotomous fine concentration (ng/m3)
Attenuation correction factor for the dichotomous fine particles (PM^j)
described above.
Attenuation correction factor for coarse particles as shown below.
Element
Correction Factor +/- error
A1
S
P
S
a
K
Ca
Sc
Ti
0.56 +/- 0.15
0.57 +/- 0.13
0.65 +/- 0.20
0.82 +/- 0.13
0.82 +/- 0.10
0.85 +/- 0.10
0.83 +/- 0.05
0.85 +/- 0.08
0.81 +/- 0.08
S-3
-------
Page 4 of 6
V
0.87 +/- 0.08
Cr
0.87 +/- 0.08
Mn
0.88 +/- 0.08
Fe
0.95 +/- 0.04
Co
0.93 +/- 0.05
Ni
0.93 +/- 0.05
Cu
0.93 +/- 0.05
Zn
0.93 +/- 0.05
These correction factors were derived using procedures developed by T. G. Dzubay
(Advances X-Ray Anal. 18:619. 1974) for 20 pm coarse particulates. Elements not listed above
received no attenuation correction.
Data Comparison
Twenty six SIM and SAM filters were analyzed by both the EPA facility and the DRI
facility. A limited check on interlaboratory comparability was made by comparing the reported
results for three elements (Ai, Br, and V) from five of these filters, as shown below.
Filter
Al
Br
V
EPA DRI
%A
EPA
DRI
%A
EPA
DRI
%A
F6208
2358.0 3382.6
43.5
15.1
16.7
10.6
24.8
26.7
7.7
F5081
6539.9 8853.7
35.4
21.1
21.1
0.0
41.4
39.3
-5.1
F6809
8826.6 11759.4
33.2
26.5
32.2
21.5
56.6
18.6
-67.1
F6881
1579.9 2466.3
56.1
17.7
16.3
-7.9
7.6
12.3
61.8
F5264
2660.1 3168.7
19.1
8.2
8.3
1.2
8.5
16.1
89.4
One hundred and seven dichotomous (coarse/fine), PEM, SIM and SAM filters were
analyzed by both the EPA facility and the LBL facility. A limited check on interlaboratory
comparability was made by comparing the reported results for three elements (Al, Br, and V)
from five of these filters, as shown below.
Al
Br
y
Filter
EPA
LBL %A
EPA
LBL
%A
EPA
LBL
%A
F4112
3236.9
2375.00 -26.6
2.2
3.39
54.1
2.8
17.20
514.3
F4449
713.7
833.93 16.9
2.8
2.61
-6.8
5.2
23.23
346.7
S-4
-------
Page 5 of 6
F4095
744.0
767.86
3.2
2.7
3.98
47.4
4.3
7.51
F4147
968.3
898.21
-7.2
1.4
3.96
182.9
6.7
0.0
F4055
953.0
862.50
-9.5
2.4
3.05
27.1
4.7
7.97
Filter Blanks
Elemental analyses of laboratory and field blanks data were provided to RTI in units of
ng/cm2. To make the particulate data comparable to the blank data, the blank data have been
converted to concentration units of ng/m3. The conversions were calculated as follows:
PEM, SIM, SAM, Blanks;
I 1
Field Blank Data (ng/m3) = I Field Blank Data n6 f x 2.604 cr^
\ cm2 / m3
where:
area of particulate filter on which sample was collected = 7.5 cm2
sampler flow rate = 4 L/min
sample collection time = 720 min (12 hour sample period)
Conversion Factor = 7-5 cm2 x sample x min x 1000L _ 2.604 cm*
sample 720 min 4L m3 m3
Dichotomous Blanks;
Field Blank Data (ng/m3) = I Field Blank Data
nS x 0.566 0111
cm
m
where:
area of particulate filter on which sample was collected = 6.8 cm2
sampler flow rate = 16.74 min
sample collection time = 720 min (12 hour sample period)
c* c
-------
Page 6 of 6
Conversion Factor = 6-8 cm2 x sample x x 1000L = 0.566 ctT^
sample 720 min 16.7L "m3 m3
In conclusion, the review of the EPA, LBL and DRI data on PTEAMS XRF indicates
that these data have been appropriately corrected for absorption effects. These data have been
reviewed and discussed in greater detail within the PTEAMs final report
RCS/mtd
Attachment
cc: Pam Reading
•Kent Thomas
File: 4948-0803980
S-6
-------
Date.' September 11, 1991
To: Kent Thomas
Researoh Triangle Institute
From: Bob Kellogg
ttaMech Environmental
Subject: Attenuation Corrections for XRF Data on PTEAM Samples
Since LBL applied no attenuation corrections to their XRF data
from the PTEAM samples, I am reporting in this memo the values
Mhich we use for coarse dichot and PhlO samples. These values
should be used to correct XRF data before correcting for overlap.
This is of course impossible on the LBL data set so some errors may
still be present for elements where strong overlap exists. Me can
pursue this matter at a later date if it should be necessary.
Element
PM10
Dichot Coarse
A1
0.61
4-
0.15
0.56
4- 0.15
Si
0.67
4-
0.13
0.57
4- 0.13
P
0.65
4- 0.20
S
0.82
4- 0.13
CI
0.82
4- 0.10
K
0.93
4-
0.10
0.85
~- 0.10
Ca
0.94
4-
o.os
0.83
4- 0.05
so
0.8S
~- 0.08
Ti
0.81
4- 0.08
V
0.87
4- 0.08
Cr
0.87
4- 0.08
Mn
0.88
4- 0.08
Fe
0.95
4- 0.O4
Co
0.93
4- 0.05
Ni
0.93
4- 0.05
Cu
0.93
4- 0.05
Zn
0.93
4- 0.05
Note that for PM10 only 4 elements have attenuation factors. The
elements P, S, and CI are treated as PH2.5 samples and all other
elements receive no attenuation correotion. For ooarse dichot
samples elements which do not appear in the above list receive no
attenuation corrections.
Example calculation using hypothetical LBL data for A1 in PM10
sample:
LBL reports 230 +- 53 for A1 in a PM10 sample.
Correoted data would be: 230/.61 = 377
-------
Corrected uncertainty would be: SQR(53~2+(.15*230/.61)""2) = 77.S
Attenuation corrections for fine (PM2.5) samples are
calculated for each element (A1 thru Ca only) for each sample using
data in the following table:
nrt
MlVHAU/MU
AT1H
*1, CH2/V0
tun
\
.00000
.000
13 Al
.00140
.110
14 11
.00090
.100
If '
.00077
.oto
1< t
.0005*
.010
If «.
.OOOfT
.0)0
II M
.00000
.010
It K
,00047
.010
19 CA
,0004)
.030
For example, to calculate the attenuation for S in a PM2.5 sample
whose deposit mass is 822 micrograms over an area of 7.00 cm2 do
the following:
attn=(l-exp(-.00056*§22/7>)/(.00056*822/7) = .9678
Tne uncertainty in the attenuation from the table is .03.
This attenuation and uncertainty is used in the same manner as in
the example for the PMtO sample above.
If 1 oan be of any additional help please call.
S-8
-------
APPENDIX T
MEMORANDA DESCRIBING PTEAM DATABASE AND PERFORMANCE AUDITS
-------
-------
RESEARCH TRIANGLE INSTITUTE
Center for Environmental Measurements and Quality Assurance
April 11,1991
MEMORANDUM
TO: E.D. Pelizzari, Project Director
S.V. Kulkarni, Quality Assurance Officer
FROM: R.C. Shores
C.O. Whitaker (_ o ^
SUB J: Review and Assessment of Riverside, CA, PTEAM Data
This memorandum documents the completion of an extensive review of PTEAM data.
Review comments by Ken Thomas, Pam Reading, and Andy Clayton have been incorporated.
The measured concentrations from the various samplers operating at the Temporal site were
compared to determine if any one sampler was reporting significantly different values than the
others. Because concentrations were calculated using the same techniques, solutions to any
problems indicated at the Temporal site could be applied across the PTEAM network. Using this
logic, each PMjq sampler operating at the Temporal site was compared to the other PMjq
samplers operating at that site. This review technique was also used for PM^j samplers
operating at the site. Significant differences indicated by any one sampler or sample were
investigated further and appropriately flagged in the data files.
The PTEAM samplers were compared to EPA-designated PMjq and PM^ g samplers.
Results indicated that the PMjq PTEAM samplers measured concentrations averaging 8 percent
higher than the EPA-designated samplers, and that the PM^j samplers measured concentrations
averaging 23 percent higher than the EPA-designated samples. This analysis is being reviewed
in greater detail by Andy Clayton.
Attached is a discussion that explains this review technique in greater detail and the results
of die PMjq and PM2 5 comparisons. Also attached are the review and discussion of the
following data bases:
Post Office Box 12194 Research Triangle Park, North Carolina 27709-2194
•telephone 919 541-6914 Fax: 919 541-5929
T-1
-------
Memorandum - Page 2
April II, 1991
Allsoit
Wedding
Dichots
Weights
Rotameter
Cascade
Meteorological
Containing all 37-mm filters. Stationary Indoor Monitors,
Stationary Ambient Monitors, and personnel; no
dichotomous filters
Containing all Wedding sampler information
Containing all dichotomous sampler information
Containing 37-mm filter information; no dichotomous filters
Calibrations including serial numbers, slope, and intercept
Containing all sampler information
Meteorological data from three airport locations
Based on (his review and evaluation (he dad are judged (o be acceptable for project
use.
/dmh
Attachment
File: 4948-08(73656
1-2
-------
DISCUSSION
Since the concentrations are calculated by computer, the data weir examined from three
different perspectives. The first involved validating the formulas that were used to calculate the
concentrations. The second perspective involve! verifying raw data inputs, most specifically
weights and volumes, completeness of records, reasonableness of values, and accuracy of data
entries. The final perspective involved evaluating the representativeness; i.e., did the measured
concentrations from individual samplers reflect expected concentrations, and were the measured
concentrations within reasonable ranges?
Sampler operation and data calculation procedures were referenced to the Pilot Study
Work Plan, Volume II (protocols). For the EPA-approved PMjq samplers operating at the
Temporal site, EPA guidelines were also referenced. These guidelines were taken from the
"Reference Method for Determination of Particulate Matter as PMjq in the Atmosphere,"
Sections 2.10 and 2,11 of the Quality Assurance Handbook for Air Pollution Measurement
Systems. Volume n. for the Dichotomous and Wedding samplers, respectively.
These procedures were used to verify that the measured concentrations were calculate!
correctly. These measured concentrations were compared to the data summaries and adjusted to
standard temperature and pressure (at 25°C and 760 mmfig).
The following subsection discusses the review of PMjq and PMjj data and how these
data compare across sampler type operated at the Temporal site. The Temporal site data quality
is representative of the overall performance of all the samplers in the field. This quality
assurance check at this site will indicate the representativeness of the data quality collected
during the study. The Wedding/Dichotomous EPA-designated reference samplers were
compared to PTEAM samplers to assess die representativeness of the PTEAM data quality.
Concentrations that deviated significantly from the mean measured concentration for all sampler
types were examined closely to determine if they were valid measurements. A potential outlier
was defined as differing by a factor of 2 or more from the average.
T«3
-------
PM |Q DATA
The actual concentrations are presented in Table I; each run is plotted and presented in
Figure 1. From this data set, concentration values (PM jq) were averaged for each respective run
across the PMjq samplers (dichotomous samplers B and C, Wedding A and D, and the
Stationary Ambient Monitor [SAM]). Then each concentration was subtracted from Ac average
run concentration. These resulting differences are presented in Figure 2. From these figures,
suspect deviations could be recognized.
Suspect data were investigated and appropriately flagged in the data file. This analysis was
conducted to evaluate data quality.
The SAM indicated higher concentrations than the other PM jq (dichotomous and
Wedding) samplers. The SAM-indicated concentrations were regressed onto the indicated
average Dichotomous and Wedding sampler concentrations. The resultant slope showed that
SAM-indicated concentration values averaged 8 percent greater than the indicated dichotomous
and Wedding concentrations. This comparison is presented in Figure 3.
pmZ5 DATA
The actual PM2 5 measured concentrations are presented in Table 2; each run is plotted
and presented in Figure 4. From this data set, measured concentrations (PM^j) were averaged
for each run across the PM2 5 samplers (dichotomous samplers B and C and SAM PM2 5).
Then each concentration was subtracted from the average run concentration. These resulting
differences are presented in Figure 5.
Suspect data were investigated and appropriately flagged in die data file. This analysis was
conducted to evaluate data quality.
The SAM indicated higher concentrations than the dichotomous samplers. These higher
concentrations attribute to most of the significant deviations (see Figure 4). The SAM-indicated
concentrations were regressed onto the indicated average dichotomous concentrations. The
resultant slope showed that the SAM-indicated concentrations averaged 23 percent greater than
the indicated dichotomous concentrations. This comparison is presented in Figure 6.
T-4
-------
TABLE 1. TEMPORAL SITE PMie CONCENTRATIONS
DICOT
WEDDING
SAM
PMl0pg/m3
PM10pg/mJ
PM10pg/m-
DATE
RUN #
B
C
A
D
9/22/90PM
I
31.6
31.4
28.0
26.8
28.4
9/23/90AM
2
37.6
37.0
40.5
38.9
41.2
9/23/90PM
3
34.7
33.3
29.7
27.9
39.8
9/24/90AM
4
47.4
47.4
43.6
41.3
47.5
9/24#)PM
5
40.1
41.1
36.2
34.5
49.2
9/25/90AM
6
72.0
72.7
81.9
85.9
65.7
9/25flOPM
7
46.5
43.6
41.7
42.4
49.5
9/26/90 AM
8
67.5
67.6
77.1
78.4
86.0
9/26/90PM
9
50.9
50.9
39.1
40.8
62.4
9/27/90AM
10
46.7
46.7
45.8
46.7
41.6
9/27/90PM
11
43.4
27.4
26.5
48.2
9/28/90AM
12
57.7
57.1
59.6
69.9
9/28/90PM
13
62.2
57.9
51.6
51.6
67.9
9/29/90AM
14
85.4
87.4
44.3
44.1
9.2
9/29/90PM
15
49.3
45.7
42.7
41.9
50.6
9/30/90AM
16
49.2
46.4
59.2
57.0
59.6
9/30/90PM
17
60.3
54.4
55.4
56.5
72.2
10/1/90 AM
18
59.5
54.0
58.4
60.6
70.2
10/1/90PM
19
73.6
65.9
55.5
57.5
92,5
10/2*90 AM
20
73.8
74.9
77.7
81.7
97.9
1Q/2/90PM
21
65.4
60.4
53.5
53.6
80.6
10/3/90AM
22
112.9
105.4
138.3
133.2
140.9
10/3/90PM
23
141.7
126.5
117.9
148.9
180.5
10/4/90AM
24
67.0
64.0
87.1
87.3
82.6
10/4/90PM
25
79.0
71.8
75.2
78.5
94.2
10/5/90 AM
26
53.7
50.7
67.8
69.1
1Q/5/90PM
27
45.8
35.9
39.5
39.5
49.4
1(V6^0AM
28
73.9
60.2
62.7
62.0
74.3
1Q/6/90PM
29
51.7
46.3
39.7
39.8
62.4
10/7/90 AM
30
41.5
38.8
37.8
39.2
49.2
10/7/90PM
31
39.1
36.6
21.0
25.2
59.0
10/8/90 AM
32
119.4
257.4
90.4
92.2
130.6
10/8/90PM
33
25.4
23.6
23.2
21.5
29.8
10/9/90 AM
34
33.2
31.2
31.6
31.6
40.7
KV9/90PM
35
8.3
17.4
17.4
16.6
(Continued)
T-5
-------
TABLE I. TEMPORAL SITE PMl0 CONCENTRATIONS
(Continued)
DICOT
WEDDING
SAM
PM)Qpg/ra'
PMjojig/m3
PM1Qpg/m'
DATE
RUN #
B
C
A
D
10/10/90 AM
36
69.4
64.2
66.3
64.1
102.4
10/1
-------
TABLE 1. TEMPORAL SITE PM10 CONCENTRATIONS
(Continued)
DICOT
WEDDING
SAM
PM^^g/m*
PM10pg/mJ
PMjoiig/m*
DATE
RUN#
B
C
A
D
1Q/27/90PM
71
1483
161.7
192.8
195.2
200.5
10/2 8/90 AM
72
81.6
77.2
92.0
95.2
106.9
10/28/90PM
73
1782
167.1
196.3
200.7
209.3
1Q/29/90AM
74
11U
105.4
119.4
121.1
134.9
1Q/29/90PM
75
158.4
142.6
166.6
170.5
180.7
1Q/30/90 AM
76
86.2
88.4
92.2
93.5
106.0
10/3(WOPM
77
39.2
39.0
34.5
34.4
45.3
10/31/90 AM
78
62.5
62.4
64.1
67.3
76.7
10G1/90PM
79
42.1
41.2
29.6
33.2
50.0
11/1/90AM
80
31.6
30.2
32.7
33.4
37.1
11/1/90PM
81
56.3
55.4
43.7
44.6
53.8
11/2/90AM
82
54.8
65.1
40.1
43.7
66.8
11/2/90PM
83
36.7
37.8
29.3
30.2
42.0
11/3/90AM
84
95.8
103.4
69.9
72.9
115.7
11/3/90PM
85
34.7
38.1
0.0
32.6
35.5
11/4/90 AM
86
31.9
32.5
0.0
30.7
34.1
11/4/90PM
87
53.8
53.0
48.4
50.6
56.5
11/5/90AM
88
62.2
67.7
59.0
67.1
74.2
11/5/90PM
89
50.8
48.9
48.9
37.0
56.3
11/6/90AM
90
85.7
87.8
62.6
65.2
93.8
11/6/90PM
91
40.5
36.3
28.7
28.5
44.6
11/7/90AM
92
196.5
241.7
145.4
154J
219.1
11/7WM
93
26.6
30.5
22.0
21.7
25.6
11/8/90AM
94
60.2
64.6
55J
58.9
71.9
11/8/90PM
95
74.9
75.9
69.5
72.4
85.8
11/9/90AM
96
27.3
27.9
24.5
24.3
31.8
T-7
-------
280
260
240
220
200
180
160
140
120
100
80
60
40
20
0
Figure 1.
PMiq concentrations indicated by the dichotomous, wedding, ami pump
samplers operating at the Temporal site
-------
7*
CO
120
100
BO
60
40
20
0
-20
-40
-60
-80
100
120
140
fftiflMtJ ll uNfttllJj U
Lit
i jljul (III ll k. (IklliJAk I
llljlaalliklfe 1.1 illllR.LL-
i if 1] I! ^|m| ipm
11)
[ ifr Mr i^ipMi I
[riTinn r ^W'u1
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
run
Figure 2. PM j0 indicated concentration differences between (he Temporal site samplers
b*%«4 fK* mm-im mt+mmw
-------
240
220
200
160
160
140
120
100
BO
60
40
20
0
y «mx ~ b
y-Predicted SAM PM|0eo«ocntrMiM
¦ •Slep*a ljOTft
t • Didiol and Wadttig PMjg eonOMCrttiM
b-bta«cpi-9J12
' 0 I «—0-
J.
J.
40 80 120 160 200
AVBRAOE DiCIKJT AND WBDOINd FM,0 CONC.
240
Linear regression comparison of PTEAM sampler data to
samplers.
-------
TABLE 2. TEMPORAL SITE PMu DATA
DICOT SAM
DATE RUN # (PMi^Jg/m*) (PM^g/m3)
B C
9/22^0PM
I
13.9
14.6
9.9
9/23/90AM
2
16.9
16.8
20.5
9/23WPM
3
17.4
19.2
21.6
9/24/90AM
4
23.9
23.7
23.0
9/24/90PM
5
20.8
22.1
22.3
9/25/90AM
6
38.0
39.9
29.5
9/25/90PM
7
23.2
22.3
25.2
9/26/90AM
8
28.8
29.4
44.2
9/26/90PM
9
23.0
22.8
28.4
9/27/90AM
10
22.2
23.8
14.0
9/27/90PM
U
17.2
0.0
11.6
9/28/90AM
12
24.0
0.0
32.4
9/28/90PM
13
34.7
32.5
40.2
9/29/90AM
14
45.7
47.4
8.4
9/29/90PM
15
9.6
9.0
18.5
9/30/90AM
16
23.5
24.2
9/30/90PM
17
32.0
30.1
41.8
10/1,90AM
18
21.6
20.1
24.1
1CV1/90PM
19
36.5
33.2
48.8
10/2/90 AM
20
49.7
46.2
55.4
1Q/2/90PM
21
37.7
35.6
10/3/90AM
22
62.8
58.5
87.2
10/3/90PM
23
86.1
81.0
129.1
10/4/90AM
24
26.3
26.4
44.8
10/4/90PM
25
47.3
43.0
61.0
10/5/90AM
26
23.5
21.6
35.0
10/5/90PM
27
16.2
14.5
20.4
10/6/90AM
28
37.8
22.9
36.6
10/6/90PM
29
25.2
23.8
32.0
1Q/7/90AM
30
18.6
17.6
22.3
10/7/90PM
31
18.7
17.7
26.1
10/8/90AM
32
10.9
38.9
45.8
10/8/90PM
33
6.7
6.5
21.0
10/9/90AM
34
8.4
10.3
15.9
(Continued)
T-11
-------
TABLE 2. TEMPORAL SITE PM^ DATA
(Continued)
DICOT SAM
DATE RUN # (FMaJigAn') (PM^^g/m')
B C
10/9/90PM
35
4.8
52
KV1090AM
36
13.2
12.7
63.2
10/1090PM
37
32,6
33.2
35.0
10/11/90AM
38
25.9
26.0
32.3
10/11/90PM
39
27.2
38.6
39.6
10/l2y90AM
420
60.9
60.8
90.8
1Q/12/90PM
41
73.0
68.0
90.3
10/13^0 AM
42
80.0
77.0
101.7
10/13/90PM
43
41.3
39.3
56.0
1W14/90AM
44
71.6
72.4
93.6
1Q/14/90PM
45
36.3
36.0
IO/15/90AM
46
91.0
91.6
127.8
1Q/15#0PM
47
71.5
71.0
111.0
10/16*90AM
48
112.2
113.8
152.4
1Q/16/90PM
49
39.3
41.8
70.6
10/17/90 AM
50
73.6
70.8
82.8
IQ/17/90PM
51
24.9
24.4
35.4
10/18/90AM
52
40.9
41.1
55.2
10/18y90PM
53
24.1
23.8
27.6
10/19,90 AM
54
14.0
13.8
18.1
10/19/90PM
55
17.0
16.6
22.3
10/20/90AM
56
5.0
6.1
12.1
1Q/20/90PM
57
43
4.7
6.8
10/21/90 AM
58
3.6
4.0
1.3
1Q/21/90PM
59
5.9
8.9
72
10/22/90 AM
60
18.2
18.0
24.9
10/22/90PM
61
34.8
34.2
41.8
10/23/90AM
62
16.4
15.8
22.9
10/23/90PM
63
69.0
65.5
79.4
10/24/90AM
64
12.7
12.0
14.3
10/24/90PM
65
41.9
38.8
46.5
10/25/90AM
66
30.3
30.8
40.1
10/25/90PM
67
73.8
67.9
76.2
10/26/90 AM
68
66.5
67.0
89.5
(Continued)
T-12
-------
TABLE 2. TEMPORAL SITE PMjj DATA
(Continued)
DICOT SAM
DATE RUN # (PMuPg/m1) (PMajig/to1)
B C
1(V26/90PM
69
100.3
94.7
111.9
10/27/90AM
70
73.4
71.1
104.2
10/27/90PM
71
138.4
132.8
167.5
10/2 8/90AM
72
53.3
52.1
74.4
10/28/90PM
73
135.7
133.5
175.9
I0/29/90AM
74
69.5
67.8
91.1
10/29/90PM
75
124.0
113.9
144.5
1OG0/90AM
76
43.9
46.5
57*
10/3Q/90PM
77
15.0
15.4
10/31/90 AM
78
26.8
26.5
31.8
10/31/90PM
79
23.2
22.9
26.7
11/1/90 AM
80
9.4
9.8
8.5
11/WM
81
9.0
9.0
13.5
11/2/90AM
82
5.3
55
8.9
ll/2v90PM
83
8.1
8.4
9.4
11/3/90 AM
84
8.7
9.2
30.1
11/3/90PM
85
14.0
14.7
12.8
11/4/90 AM
86
12.1
12.6
16.4
11/4/90PM
87
35.1
36.5
37.6
11/5/90 AM
88
31.1
34.1
38.1
11/5/90PM
89
26.0
26.1
27.0
11/6/90 AM
90
12.1
10.5
17.4
11/6/90PM
91
4.8
5.0
6.8
11/7/90 AM
92
17.3
19.2
71.4
11/7£0PM
93
8.3
8.8
6.4
11/8/90 AM
94
22.3
49.0
28.6
11/8/90PM
95
48.0
49.2
52.6
11/9/90AM
96
6.4
6.6
15.3
T-13
-------
190
180
170
160
150
140
130
120
110
100
90
ao
70
60
50
40
30
20
10
0
9
20 40 60 80 100
RUN
~ OtCOT B ~ DfCOT C O SAM
Figure 4. PM2 5 concentrations indicated by the Dichotomous and pump samplers
operating at the Temporal site
-------
-I
*
1
m
3
U
tt
UJ
£
o
o
oc
o
<
oc
u
z
o
o
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
RUN
Figure S. PMj 5 indicated concentration differences between the Temporal site
samplers and the run mean concentrations
-------
y-«i*b
yhoBcWSAMPMj ^conowibitinB
"¦Slept- 1133
* - DidxX PMjj aoooMfrabon
b - Intercept - 3 J«0
O-J 1 1 1 1 I I I i i i ¦
JL
0 20 40 60 80 100 120 140 t60
AVERAGE DfCHOTPMjj CONC.
180
Figure 6. Linear regression comparison of PTEAM sampler data to EPA designated
samplers.
-------
ALLSORTDATA
These data contain information recorded by the sampling teams in the field, including
sampler/filter ID, temperatures, times, and rotameter readings. Concentrations were calculated in
the field to assess the data quickly and to allow for corrective action should any problems be
indicated Also included are concentrations calculated by RH puject personnel.
• In the allsort data set, measured concentrations were verified by recalculation using the
rotameter readings, rotameter calibration equation, and time on/time off. The
rotameter calibration standard was traceable to NIST, although the calibration dates
are as much as 1.5 years old
• Since data are entered on computer, only a few calculations for each rotameter were
performed to verify the data set
• Weight values in the allsort data base were verified by visual comparison with entries
in the weigh trailer notebook. Ten percent of the entries were checked against the
laboratory notebook. Reweighs and QA passtfail checks were examined for any
deviation from protocol.
• Field monitoring coordinators and data base merging and management personnel have
performed QC checks on the data set for accuracy of start/stop times, filter
identification numbers, and base/filter head identification numbers.
• Since the data set contains all the data from the field operations (i.e., pump numbers
filter type, nicotine, 2.5^,, or 10^), a more descriptive name should be given to the file
for ease of identification for persons exposed to PTEAM data for the first time. Such a
name as "field_data" might be employed to follow a more descriptive overture as the
self-explanatory "weights," "dicots," or "wedding" data sets.
• Data for Stationary Indoor Monitor (SIM), SAM, and Personal Environmental Monitor
(PEM) should be separated into columns for PMjq and PM2 5 and duplicates. As the
data file is now, both PM jq and PM2 5 are in the same column and may be combined
during analysis.
• The data set needs to incorporate ambient temperature and barometric pressure and
calibration equations so that the data base will be considered all-inclusive. Thus, any
calculations could be made within the data base and not require additional data.
Calculations were performed to assure that proper tempera ture/pressure adjustments had
been applied to the data. The comparisons are as follows:
T-17
-------
Temperature/Pressure Corrections
RT1 Audit
Adjusted
Black
Actual
Max/Min
Max/Min/Mean
STD Value
Volume
Difference
Box
Volume
Pa/P-.^
•F
•c
("3>
(m3)
4577
2.60
.9729
88/66
31.1/1SJ9/25j0
2J3
2.53
0.00
4578
2.03
.9476
86/68
30.Q/20.CV25.0
1.98
1.98
0.00
4580
3.23
.9713
78/73
25.6/22.8/24.2
3.14
3.14
0.00
4582
2.40
.9743
84/64
28.9/17.8/23.3
2.35
2.34
-0.0
Where:
Qstcj = Sampler flow rate adjusted to EPA standard conditions of 25°C and 760 mm Hg, units
Qa = Actual volumetric flow rate at ambient temperature and barometric pressure, units
Pa * Ambient barometric pressure, mm Hg
P5t(} = Standard barometric pressure, mm Hg
Tstd = Standard temperature, 298K
Tg * Current ambient temperature, K(K=°C + 273)
Pa Tstd
Qstd = Q4 x
®*std
WEDDING DATA
These data contain information recorded by the sampling team in the field, including
sampler/filter ID, temperatures, pressures, times, and weights. Also included are concentrations
calculated by the field personnel. These concentrations were calculated in the field to assess the
data quickly and to allow for timely corrective actions should any problems be indicated.
Specific observations follow:
• Acurex personnel have reported the Wedding sampler concentrations adjusted to
standard conditions (25°C and 760 mm Hg) using the seasonal averages (temperature
and pressure) for the Riverside area, and not the actual conditions. The data are
considered comparable with the other PTEAM data corrected to standard conditions
using actual pressure and temperatures.
T-18
-------
• RTI personnel compared the information on Wedding field data sheets to the data
file. Some discrepancies were noted, but have been resolved.
• Examination of the Wedding data base showed that seventeen days of flow rates
exceeded the design flow rates for the samplers, six days had no flows, and eight
entries of PI pressure readings were transcribed incorrectly.
DICHOT DATA BASE
These data contain information recorded by the sampling team in the field, including
sampler/filter ID, temperatures, pressures, run time, course/fine flows, filter weight, and weight
of particulate catch. Also included are concentrations calculated by the field personnel. These
concentrations were calculated in the field to assess the data quickly to allow for timely
corrective actions should any problems be indicated Specific observations follow:
• The dichotomous samplers were operated without proper attention to the coarse flow
rates. One hundred six samples were collected with coarse flow rates in excess of the
acceptable tolerance of ±10% of the design flow rate. This will impact the data by
biasing the fine data "high"; however, the total PMjg concentrations would remain
relatively unaffected This problem is the result of the operator not conducting QC
checks on both the total and coarse flow rates. These QC checks were included in all
referenced procedures and the operator was reminded of the need for these checks
during auditing activities.
• Acurex personnel adjusted the concentration data to standard conditions using
seasonal averages for the Riverside area.
• RTI personnel verified filter weights from the electronic data base against the
notebook values. Dichot filter weighings were entered into the notebook by Acurex
personnel using a differently colored pen. This aided in tracking filters.
• Rotameter calibration procedures used by Acurex personnel were according to the
latest EPA guidance documentation. This means that the concentrations have been
calculated by adjusting the concentrations for that portion of the fine particles that
end up on the coarse filter.
• The dichotomous sampler concentrations were adjusted to EPA standard conditions
(25°C and 760 mm Hg) by Acurex personnel from season averages representing the
Riverside area.
T-19
-------
WEIGHTS DATA
These data contain weight information recorded by the field sampling team, including
filter ID, weight of clean and exposed filters, filter type, date, time, and operator. This data set
also includes quality control checks for the filters by indicating reweights for tare and final
weighing. Specific audit checks follow:
• RTI personnel scrutinized the weights for the 37-mm filters by checking
approximately 10 percent of the entries. This was done by comparing the weights
data printout and the laboratory notebook. Specific items reviewed were
transcription errors, initial and final reweighs, initial and final tares, and quality
control reweighs for pass/faiL
• RTI personnel discovered that the data base contained tare weights for some filters,
but no final weights. These filters could be unused and need to be deleted or marked
unused.
• The RTI auditors found that the data base contains only it-tare weights, indicating
that the filter weights, used were the weights that satisfied the QC criteria.
• The RH auditors verified that the data base contained only the re-final weights,
indicating that the program was using the proper filter weights.
• RTI auditors checked 10% of the filter weights in the data base and verified them
against the notebook. Auditors were confident that this was representative of the
whole population of weights.
• With the exception of last digit rounding, RTI auditors were satisfied that there were
no discrepancies between the notebook and the data base.
• RTI auditors found that for filter number F5526, the final weight (106.645) is less
than the tare weight (106.674).
• RTI auditors verified the indicated catch within the data base by recalculating the
catch using the notebook entries. There were no discrepancies.
• Notebook entries were not always included in the data base on the same day
indicated by the notebook. This may be an indication of the balance operator either
working past midnight or the operator beginning data entry the next day.
T-20
-------
¦ Acurex personnel recorded the dichotomous filter weights in the notebook in a
different color of ink. This assisted in filter searches, although some of the notebook
data entries were difficult to read
ROTAMETER DATA
These data contain rotameter ID, slope, intercept, calibration, and date.
• Harvard personnel calibrated six rotameters using NIST-traceable volumetric
standards.
• Rotameters used during the field study were calibrated between 2/20/89 and 9/4/90.
This time period may be considered excessive; however, these performance audits
indicated acceptable agreement
• RTI data management personnel at RTP formatted the data base to identify rotameter
serial numbers and indicate starting and stopping readings. RH auditors verified that
the average sample flow was calculated using the appropriate rotameter calibration.
CASCADE IMPACTOR DATA
These data contain information recorded by the sampling team in the field, including run
number, date, time, aerodynamic particle size cut at each point, and respective concentrations.
Cascade impactor data were evaluated by comparing them with the indicated Wedding
sampler concentrations. Because the cascade impactor operated from 8 to 21 12-hour Wedding
sampling periods, the Wedding concentrations were averaged over the cascade impactor
sampling period. The ratios indicated by Wedding concentrations divided by cascade
concentrations were 0.5,0.4,2.3, and 0.4. These data indicate that the data of the third run of the
cascade impactor probably should not be considered valid. This also indicates that the cascade
data should be considered only to indicate size distribution and not absolute concentrations.
• The data base needs to be edited to include the start and stop times for each of the
cascade impactor samples, as follows:
T-21
-------
Cascade
Run #
Start
End
1
2
3
4
Date Time
9/28/90 20:15
10/6/90 19:42
10/24/90 19:30
10/30/90 18:34
Date Time
10/3/90 7:32
10/11/90 18:27
10/28/90 7:05
11/9/90 18:06
• There are no precision estimates available for these data. To provide a good
indicator of sampler precision, a second collocated cascade impactor should have
also been operated.
• Because each cascade impactor sample was a composite of numerous (8 to 21)
sample periods, the cascade impactor results may be of limited use.
METEOROLOGICAL DATA
These data contain information recorded by the sampling team in the field, including
wind speed and direction, temperature, dew point, date, and time.
At the onset of the pilot study, the meteorological data were being collected by a
Climatronics meteorological system. Soon after the study began, however, the system
malfunctioned and was no longer collecting data. The complete system was not available for the
first or second audit of the Temporal site, due to system failure. The audit team, however, was
able to assess the wind direction and wind speed, along with temperature and relative humidity,
on the first audit During the remaining days of the study, the Climatronics system was being
repaired. Unfortunately, the system was not repaired before the pilot study was concluded.
Due to the malfunctions of the meteorological collection system, the Temporal site
meteorological data have been judged to be unsatisfactory.
To correlate the PMjq concentrations with meteorological events, meteorological data
from surrounding weather stations were obtained from the National Oceanic and Atmospheric
Administration's National Climatic Data Center in Asheville, North Carolina.
The data are considered to be of acceptable quality, but have not been scrutinized by the
PTEAM's QA staff.
T-22
A
-------
RESEARCH TRIANGLE INSTITUTE
_EI
Center for Environmental Measurements and Quality Assurance
May 13,1991
MEMORANDUM
TO: E.O. Pelizzari
S.V. Kulkami
FROM: R.C. Shores
C.O. Whi taker
SUB J: Interim Assessment of PTE AM Data
The attached list of data files has been reviewed and the data files are considered
satisfactory to release for further analysis. The following details were reviewed more
closely:
The paniculate data have been corrected to standard conditions using an average
temperature from the local airports and an average barometric pressure from the Temporal site.
This correction was applied to all outdoor particulate samples.
Per earlier discussions with Pam Reading and Kent Thomas, the Wedding sample
calibrations were recalculated to follow Section 2.11 of EPA's QA Handbook for Air Pollution
Measurement Systems.
Files of greatest concern were those directly associated with the calculation of particulate
concentrations. Questionnaire data were not reviewed.
Attachment
RCS:dmh
cc: P, Reading
K. Thomas
File: 4948-08C/3782 _ _
Post Office Box 12194 Research Triangle Park, North Carolina 27709-2194
Telephone 919 541-6914 Fax: 919 541-5929
T-23
-------
DATA FILES IN PARTICLE TEAM PILOT DATA BASE
By Data Type
Page 1 of 3
RELEASE #1, MAY 8, 1991
TEMPORAL SITE AND AMBIENT DATA
CASCADE - data from cascade Impactor runs at temporal site. 05/08/91.
40 records, 16 variables. DISK 2.
CASCALB - cut size calibration data for cascade Impactor. 05/08/91.
104 records, 8 variables, DISK 2.
TEMPORAL - gravimetric data from MEMs and PEMs operated at the central
site, and data from the co-located PEMs worn by PTEAM staff. 05/08/91.
332 records, 60 variables, DISK 2.
DICHOT - data from the two dlchotomous samplers operated at the central
site, as obtained from Accurex. 05/08/91. 192 records, 52 variables,
DISK 2.
WEDDING - data from the two Wedding samplers operated at the central
site, as obtained from Accurex. 05/08/91. 192 records, 35 variables,
DISK 2.
DICBLK - weight data on blank filters for the dlchotomous samplers
operated at the central site. 05/08/91. 41 records, 3 variables, DISK
2.
WEDBLK - weight data on blank filters for the Wedding samplers operated
at the central site. 05/08/91. 44 records, 3 variables, DISK 2.
METDATA - meteorological data from three nearby airports during the
time of PTEAM data collection. 05/08/91. 1176 records, 14 variables,
DISK 2.
AVMETDAT - meteorological data from three nearby airports sumnarized
Into 12 hour periods during the time of PTEAM data collection.
05/08/91. 96 records, 34 variables, DISK 2.
T-24
-------
DATA FILES IN PARTICLE TEAM PILOT DATA BASE
By Data Type
Page 2 of 3
RELEASE #1, KAY 8, 1991
GRAVIMETRIC ANALYSIS
BLANKS - data from various types of blank particle filters. 05/08/91.
61 records, 36 variables, DISK 2.
QAWTS - data from QA weighings of particle filters. Filters may have
been weighed twice before or after sampling or at both points.
05/08/91. 640 records, 9 variables, DISK 2.
FIELD - gravimetric data from MEMs operated in participants homes and
workplaces and PEMs worn by participants. 05/08/91. 1836 records, 60
variables, DISK 3.
ROTAMETR - calibration data for rotameters used for flow checks 1n
field. 05/08/91. 6 records, 7 variables. DISK 2.
MHINFO - Information on sampled households such as room considered
•main living area (SIM site)', single or multi-family structure, 'dirt
level' In home, and Indicator of CARB sampling. 05/08/91, 184 records,
10 variables, DISK 2.
T-25
-------
DATA FILES IN PARTICLE TEAF PILOT DATA BASE
By Data Type
Page 3 of 3
RELEASE #1, KAY 8, 1991
QUESTIONNAIRE DATA
QUEST1 - data from the household questionnaires administered to the
participants. This file contains data for Period 1 of monitoring.
05/08/91. 191 records, 125 variables, DISK 1.
QUEST2 - additional data from the household questionnaires administered
to the participants. This file contains data for Period 2 of
monitoring. 05/08/91. 191 records, 125 variables, DISK 1.
QUEST3 - additional data from the household questionnaires administered
to the participants. This file contains data not specific to a
particular monitoring period. 05/08/91. 191 records, 76 variables,
DISK 1.
SCREENER - data from the household screening forms administered to the
all households. 05/08/91. 680 records, 72 variables, DISK 1.
DIARY - data from the time and activity diaries filled out by the
participants. 05/08/91. 3536 records, 32 variables, DISK 2.
SAMPWTS - calculated weights and other sampling data for the sample
housing units for which a person was selected for monitoring. See memo
of February 26, 1991 for notes on use. 05/08/91. 257 records, 14
variables, DISK 1.
T-26
-------
RESEARCH TRIANGLE INSTITUTE
ZB3
Center for Environmental Quality Assurance
' October 8, 1990
MEMORANDUM
TO: S.V. Kulkarni, Quality Assurance Officer
FROM: R.C. Shores
SUBJ: Preliminary Results from PTEAM Audit
From September 27 through October 2, 1990. (including the weekend), Craig
Whitaker and I conducted an audit of the PTEAMs study In Riverside,
California. The auditors worked on the weekend, anticipating that the
participants would not have to go to work and would not Bind delays. The
auditors accompanied the saapllng teams to 10 hoaes and observed all phases of
the sampling. All components of the prograa were audited including temporal
site, weighing facility, and work room operations. The merge program was run
and concentrations were calculated for the operating samplers. The following
concerns were identified and should be addressed:
1. The cascade lmpactor was not operating and the audit results Indicat-
ed a significant leak in the system.
2. PEM data collected at the temporal site were recorded in terms of
hours instead of minutes. The merge program revealed the problem.
The problem was Identified and the operator was notified.
3. The merge program should use the moat current weight. The most
current weight is the weight that has satisfied QC requirements. The
program is using the first weight, which may contain significant
error.
4. Random selection of appointment times did not work. Five partici-
pants were missed in the first 3 days of saapllng because they had
not been scheduled. Interviewers did not understand the importance
of maintaining a full schedule for the saapllng teams. Additionally
there was confusion about the times that should be scheduled for the
sampling teams. These problems indicate poor training and prepara-
tion of the interviewers. These problems were resolved after the
field manager met with the interviewers to prioritize the full-
schedule and the proper schedule times. Since that meeting, only one
participant was missed.
Post Office Box 12194 Research Triangle Park. North Carolina 27709-2194
Telephone 919 541-6922 Fax:919 541-7215
-------
MEMORANDUM - Kulkarni
October 8, 1990
Page 2
5. Dlchotonous sampler flow rates should be calibrated and operated more
closely to the sampler's design flow rate. The performance audit
results are satisfactory. During the audit one of the samplers
needed to be calibrated and set points determined.
Additional details that should be addressed In the final report are as
follows:
1. A significant portion of the Riverside population speaks only in
Spanish; this may or may not have an impact on the data collected.
Some of the interviewers spoke Spanish, but none of the sampling
teams did. In addition, none of the mailing documents provided to
the potential participants or actual participants were provided In
Spanish. It appears that there is the potential to bias the selec-
tion of participants. Out of the 10 homes visited by the auditors,
persons In 2 of the homes spoke little English and one person expres-
sed frustration with the fact that the sampling team did not speak
Spanish.
2. The pollution plume present in Riverside Is the result of air moving
from Los Angeles to Riverside and passing through the mountains that
surround Riverside. The pollution Inversion was estimated to be less
than 1,000 feet throughout the audit. Depending upon which way the
wind was blowing and which mountain pass the Los Angeles plume was
passing through, there were times when parts of Riverside were
(visibly) pollution-free. The temporal site is located at a higher
elevation than the majority of Riverside and on the down side slope
of the Hawarden Hills. The final report should put into perspective
the placement of the temporal site and its representativeness in
characterizing the ambient conditions existing in Riverside.
One additional detail that needs to be mentioned is that the Hi-Vols and
the two dichotomous samplers were not set-up In the merge program. The
temporal site operator calculates the air concentration using a Lotus program
with all the data entered manually. All temporal site samplers should have
been included in the merge program with data entered using the bar code
reader.
T-28
-------
MEMORANDUM - Kulkarni
October 8, 1990
Page 3
In conclusion, the entire operation appears to be running smoothly and
field personnel should be conplenented.
/dmh
cc: E. Pellizzari
C. Whitaker
D. Whitaker
K. Thomas
S. Cooper
File: 4657-98E/3495
T-29
-------
RESEARCH TRIANGLE INSTITUTE
/RTI
Center for Environmental Quality Assurance
November 13, 1990
MEMORANDUM
TO: S.V. Kulkarni, QAO
E.D. Pellizzarl, Project Director L £o*S
FROM: Craig 0. Whitaker and Richard C. Shores
SUBJ: Preliminary Results from the Second PTEAM Audit
From October 27 through October 31, 1990, Mr. Craig Whitaker conducted an
audit of the PTEAM study in Riverside, California. Mr. Whitaker accompanied
the sampling teams to five homes and observed all phases of sampling.
Components of the program that were audited included temporal site, weighing
facility, and work room operations. The following concerns were identified
and should be addressed:
1. The data collected for the study are not being processed in Riverside;
I.e., the raw numbers, filter weights, flows, etc., are collected, but
are not reduced to concentrations. Instead, the data are sent back to
RTI for data processing. Consequently, the study has allowed the
possibility for problems to go unnoticed for several days. If the
concentrations are known relatively soon, data quality is insured to a
higher degree by inference and trends. If concentrations are
Incorrect, corrective actions can be made immediately. Reviewing the
concentrations relatively close in time to collection Is a quality
control check. By removing this check, the quality of data is
diminished.
2. Due to voltage surges, there were audible changes in the drone of the
high volume sampler motors. According to Gene Stevenson of Acurex,
this phenomenon occurred most often in the morning. He attributed the
drop in line voltage to mini-brownouts. I am concerned that the
phenomenon will affect the velocity and cutpoint of the sampler.
3. The outdoor California Air Resources Board (CARB) pumps are cutting
off from 1-6 hours into the sampling period. During period 1, of 80
Indoor pumps, 68 completed the 12-hour run, 4 were partial runs, and 8
pumps did not have enough data/run time. Consequently, they must be
dropped. During period 2, 72 of 80 pumps completed the 12-hour run
time and 8 pumps ran part of the 12-hour period. For the outdoor CARB
st Office Box 12194 Research Triangle Park, North Carolina 27709-2194
ephone 919 541-6922 Fax:919 541-7215
T-30
A
-------
Kulkarni/Pel1lzzari
November 13, 1990
Page 2
pumps, during period l, 37 of 44 ran for 12 hours; 7 pumps did not run
enough to retrieve partial run time data and have been dropped.
During period 2, 30 of 44 ran 12 hours, 12 ran partial runs, and 2
were dropped for insufficient run times. The following tables
summarize the indoor and outdoor pump run tines:
Indoor
Period 1 Period 2
Total 80 80
Completed 12-hr. run 68 72
Partial run 4 8
Dropped 8 0
Outdoor
Period 1 Period 2
Total 44 44
Coapleted 12-hr. run 37 30
Partial run 0 12
Dropped 7 2
Puap interruptions Bight also be attributed to voltage dropouts.
There also seems to be a correlation between higher daytime
temperatures and pumps shutting down. Of course higher daytiae
temperatures will result in higher electricity consumption and
brownouts. According to the QAPjP, the completion claimed is 240
Indoor samples and 120 outdoor samples.
4. For future work, the CARB data sheets need an entry blank for
rotameter ID numbers for traceability.
5. In reviewing the dBASE files, I noted a few discrepancies in the data
which could possibly be typographical errors. For example, filter
number F4910 on October 2 had a starting rotameter of 106.0 (3.944
liters/minute) and an ending rotameter reading of 139.0 (4.858
liters/iinute), 23X above the design flow rate. The merge program
hopefully will flag these errors for review by data analysis
personnel.
T-31
-------
#
Kulkarnl/Pel 1izzari
November 13, 1990
Page 3
6. The operators are concerned with the variation In temperature In the
weigh trailer. They believe that the expansion and contraction of the
metal components on the balances have an effect on the indicated
weight. The need for two quality control reference weights Manifested
when the operators noted a discrepancy in the indicated weight and
believed the balance was broken, It was the reference weight,
however, that had been changed. The cause was not known, but nay be
due to mishandling.
The QA weighing program Is not currently working. The pass/fail
calculations are now perforated by hand.
?. Due to the architectural diversity of hones being sampled, placement
of FLAMs within the 5-foot offset criterion has become very difficult.
For example, wings off of the main living area of the buildings,
(i.e., garages, bedrooms extensions, etc.) create a niche. This niche
leaves only one open area for prevailing winds.
Due to small lot sizes, FLAMs are being located near parked cars,
street sources, and street activities such as traffic and
pedistrlans.
8, The FLIM placements are difficult due to small interior dimensions.
The typical house is 1,000 to 1,600 square feet of heated floor space.
Guidelines for placement of FLIMs are on an interior wall away from
windows, doors, vents, and air conditioning units with the head
placement at least 2 feet from the wall and 2 feet from internal
doors.
These criteria, along with the placement of existing furniture make
placement of FLIMs extremely difficult.
In conclusion, the entire operation appears to be running smoothly and
field personnel should be complemented. Please contact me at 541-5988, if you
have any questions or comments concerning the audit.
/win
CC; D. Whitaker
K. Thomas
S. Cooper
File: 4657-98E/3549
T-32
-------
Reproduced by NTIS
National Technical Information Service
U.S. Department of Commerce
Springfield, VA 22161
u
U O
° 1
2,2
(J Cm
o ^
C/5 *3
S 8
•H Q,
*4-1
o u
a tz
u £3
^ >
a;
d 'O
-------
-------
-------
------- |