Particulate Matter Urban-Focused
Visibility Assessment
Final Document
-------
DISCLAIMER
This document has been prepared by staff from the Office of Air Quality Planning
and Standards, U.S. Environmental Protection Agency (EPA). Any opinions, findings,
conclusions, or recommendations are those of the authors and do not necessarily reflect
the views of the EPA. Questions related to this document should be addressed to Ms.
Vicki Sandiford, U.S. Environmental Protection Agency, Office of Air Quality Planning
and Standards, C504-06, Research Triangle Park, North Carolina 27711 (email:
sandiford.vicki@epa.gov).
Elements of this report have been provided to the EPA by Abt Associates, Inc. and
Stratus Consulting Inc. in partial fulfillment of Contract No. EP-D-08-100, Work Assignments 0-
11 and 0-19. Any opinions, findings, conclusions, or recommendations are those of the authors
and do not necessarily reflect the views of the EPA or its contractors.
-------
EPA 452/R-10-004
July 2010
Particulate Matter
Urban-Focused Visibility Assessment
Final Document
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Health and Environmental Impacts Division
Research Triangle Park, NC
-------
This page left intentionally blank.
-------
TABLE OF CONTENTS
List of Tables iv
List of Figures v
List of Acronyms/Abbreviations vi
1 INTRODUCTION 1-1
1.1 PM NAAQS BACKGROUND 1-4
1.2 VISIBILITY EFFECTS SCIENCE OVERVIEW 1-7
1.3 GOALS AND APPROACH 1-11
1.4 SCOPE OF URBAN-FOCUSED VISIBILITY ASSESSMENT 1-12
1.4.1 Background 1-12
1.4.2 Selection of Alternative Scenarios for First Draft Assessments 1-14
1.4.3 Selection of Alternative Scenarios for Second Draft Assessments 1-15
1.5 ORGANIZATION OF DOCUMENT 1-18
2 URBAN VISIBILITY PREFERENCE STUDIES 2-1
2.1 METHODS USED IN VISIBILITY PREFERENCE STUDIES 2-2
2.2 DENVER, COLORADO 2-3
2.3 VANCOUVER, BRITISH COLUMBIA, CANADA 2-8
2.4 PHOENIX, ARIZONA 2-13
2.5 WASHINGTON, DC 2-15
2.5.1 Washington, DC 2001 2-16
2.5.2 Washington, DC, 2009 2-18
2.6 SUMMARY OF PREFERENCE STUDIES AND SELECTION OF CANDIDATE
PROTECTION LEVELS 2-27
3 ESTIMATION OF RECENT PM MASS AND SPECIES CONCENTRATIONS
AND PMio LIGHT EXTINCTION 3-1
3.1 SUMMARY OF PREVIOUS CHARACTERIZATIONS OF PM CONCENTRATIONS
AND LIGHT EXTINCTION 3-1
3.1.1 PM2.5 and PMio-2.5 3-1
3.1.2 PMio light extinction 3-4
3.2 OVERVIEW OF APPROACH AND DATA SOURCES FOR URBAN STUDY
ANALYSIS 3-9
3.2.1 Study Period, Study Areas, Monitoring Sites, and Sources of Ambient PM
Data 3-10
3.2.2 Use of CMAQ Model Validation Runs for 2004 to Augment Ambient Data... 3-16
3.2.3 Use of Original IMPROVE Algorithm to Estimate PMio light extinction 3-19
3.3 DETAILED STEPS 3-19
3.3.1 Hourly PMi.s Component Concentrations 3-19
3.3.2 Hourly PMio-2.5 Concentrations 3-28
-------
3.3.3 Hourly Relative Humidity Data 3-29
3.3.4 Calculation of Daylight 1-HourPMio Light Extinction 3-29
3.3.5 Exclusion of Hours with Relative Humidity Greater than 90 Percent from PMio
Light Extinction NAAQS Scenarios and Most Results 3-30
3.3.6 Calculation of Daily Maximum 1-Hour PMio Light Extinction 3-34
3.4 SUMMARY OF RESULTS FOR CURRENT CONDITIONS 3-34
3.4.1 Levels of Estimated PMi.s, PMi.s Components, PMio-2.5, and
Relative Humidity 3-34
3.4.2 Levels of Estimated PMio light extinction 3-37
3.4.3 Patterns of Relative Humidity and Relationship between Relative Humidity and
PMio light extinction 3-43
3.4.4 Tile Plots of Hourly PMio Light Extinction 3-46
3.4.5 Extinction Budgets for High PMio Light Extinction Conditions 3-63
3.5 POLICY RELEVANT BACKGROUND 3-85
4 PMio LIGHT EXTINCTION UNDER "WHAT IF" CONDITIONS OF JUST
MEETING SPECIFIC ALTERNATIVE SECONDARY NAAQS 4-1
4.1 ALTERNATIVE SECONDARY NAAQS BASED ON PMio LIGHT EXTINCTION AS
THE INDICATOR 4-1
4.1.1 Indicator and Monitoring Method 4-1
4.1.2 Alternative Secondary NAAQS Scenarios based on PMio light extinction 4-2
4.1.3 Monitoring Site Considerations for Alternative Secondary NAAQS Based on
Measured PMio light extinction 4-2
4.1.4 Approach to Modeling "What If Conditions for Alternative Secondary NAAQS
Based on Measured PMio Light Extinction 4-3
4.2 ALTERNATIVE SECONDARY PM2.5 NAAQS BASED ON ANNUAL AND 24-
HOUR PM2.5 MASS 4-10
4.2.1 Secondary NAAQS Scenarios Based on Annual and 24-hour PMi.sMass 4-10
4.2.2 Approach to Modeling Conditions If Secondary PMi.s NAAQS Based on Annual
and 24-hour PM2.s Mass Were Just Met 4-11
4.3 RESULTS FOR EACH "JUST MEET" ALTERNATIVE SECONDARY NAAQS
SCENARIO 4-12
5 SUMMARY 5-1
5.1 URBAN VISIBILITY PREFERENCE STUDIES REANALYSIS 5-1
5.2 CURRENT VISIBILITY CONDITIONS 5-2
5.3 VISIBILITY CONDITIONS FOR ALTERNATIVE SECONDARY PM NAAQS
SCENARIOS 5-4
6 REFERENCES 6-1
11
-------
APPENDICES
A. PM2.s Monitoring Sites and Monitors Providing 2005-2007 Data for the Analysis of
Light Extinction in the 15 Study Areas
B. Distributions of Estimated PMi.s and Other Components under Current Conditions
C. Development of PRB Estimates of PMi.s Components, PMio-i.s, and PMio Light
Extinction
D. Relationships between PM Mass Concentration and PMio Light Extinction under
Current Conditions
E. Differences in Daily Patterns of Relative Humidity and PMio Light Extinction between
Areas and Seasons
F. Distributions of Maximum Daily Daylight PMio Light Extinction under "Just Meets"
Conditions
G. Additional Information on the Exclusion of Daylight Hours with Relative Humidity
Greater than 90 Percent
H. Inter- Year Variability
I. Daylight Hours
J. Logit Memorandum
in
-------
LIST OF TABLES
Table 2-1. VAQ of Denver photos substantively misclassified by maj ority of
participants 2-6
Table 2-2. Summary of photographs used in British Columbia study 2-11
Table 2-3. Logit model estimated VAQ values corresponding to various percent
acceptablilityvalues for the four cities 2-29
Table 3-1. Annual Mean Reconstructed 24-hour Light Extinction Estimates by Region 3-7
Table 3-2. Urban Visibility Assessment Study Areas 3-11
Table 3-3. PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the Tacoma
Study Area 3-14
Table 3-4. Number of days per quarter in each study area 3-16
Table 3-5. Assumed daylight hours by season (Local Standard Time) 3-30
Table 3-6. Comparison of Meteorological Parameters for Daylight Hours with Relative
Humidity Greater than 90 Percent and Other Daylight Hours,
During 2005-2007 3-32
Table 3-7. Percentage of daily maximum hourly values and individual hourly values of
daylight PMio light extinction exceeding CPLs (excluding hours with relative
humidity greater than 90 percent) 3-41
Table 3-8. The numbers of common and unique days selected for each of the 15 urban areas
by the top 10% of daily maximum and the top 2% of all hours form 3-80
Table 4-1. Alternative secondary NAAQS Scenarios for PMio light extinction 4-2
Table 4-2. Current Conditions PMIO light extinction design values for the study areas 4-5
Table 4-3. Percentage reductions in non-PRB PMio light extinction required to "just meet"
the NAAQS scenarios based on measured light extinction (Mm"1) 4-9
Table 4-4. Percentage reductions required in non-PRB PM2.5 mass to "just meet" NAAQS
scenarios based on annual and 24-hour PM2.s mass 4-12
Table 4-5. PMio light extinction design values for "just meet" secondary NAAQS scenarios
based on measured PMio light extinction (excluding hours with relative humidity
greater than 90 percent) 4-16
Table 4-6. PMio light extinction design values for "just meet" secondary NAAQS scenarios
based on PM2.5 mass (excluding hours with relative humidity greater than 90
percent) 4-17
Table 4-7. Percentage of days with maximum 1-hour daylight PMio light extinction above
CPLs for each NAAQS scenario under "just meet" conditions across three years
(two in the case of Phoenix and Houston) 2-20
IV
-------
LIST OF FIGURES
Figure 1-1. Progression from PM characteristics to PM light extinction that shows the
modeling approach (shaded light green) as well as the use of direct measurements
(shaded blue) as alternative ways to estimate PM light extinction 1-9
Figure 1-2. Progression from PM light extinction to value of visual air quality (VAQ) 1-10
Figure 2-1. Percent of Denver participants who considered VAQ in each photograph
"acceptable." 2-5
Figure 2-2. Photograph time of day information 2-7
Figure 2-3. Denver photograph time of day results (9:00 a.m. photographs eliminated),
with the broader range (17.7 dv and 24.6 dv) of the 50% acceptability
criteria shown 2-8
Figure 2-4. Composite Chilliwack, BC photograph shows VAQ of 14.1 dv and 34 dv 2-9
Figure 2-5. Percent of BC participants who consider VAQ in each photograph
"acceptable." 2-12
Figure 2-6. Reproduction of image with the best VAQ (15 dv) used in the
Phoenix study 2-14
Figure 2-7. Percent of Phoenix participants who consider VAQ in each image
"acceptable." 2-15
Figure 2-8. Reproduction of the image with the best VAQ (8.8 dv) used in the
Washington, DC study 2-17
Figure 2-9. Percent of 2001 Washington participants who considered VAQ acceptable in
each image 2-18
Figure 2-10. Percent of 2009 Test 1 study participants who considered VAQ acceptable
in each image, showing the range of the lower and upper bound of 50%
acceptability criteria 2-20
Figure 2-11. Combined results of Washington, DC 2001 and 2009 Test 1 (showing 50%
acceptability criteria from 2009, Test 1) 2-21
Figure 2-12. Comparison of results from Test 1 and Test 2 (Smith and Howell, 2009) 2-22
Figure 2-13. Average visibility ratings for the Washington, DC WinHaze images by
participants in Tests 1-3 conducted by Smith and Howell 2009) 2-23
Figure 2-14. Comparison of results from the Smith and Howell (2009) Test 1 and
Test3 2-25
Figure 2-15. Composite results from Smith and Howell (2009) Tests 1 and 3, and Abt
(2001) Washington, DC pilot study 2-26
Figure 2-16. Summary of results of urban visibility studies in four cities, showing the
identified range of the 50% acceptance criteria 2-28
Figure 3-1. Annual average and 24-hour (98* percentile 24-hour concentrations)
PM2.s concentrations in ug/m3, 2007 3-2
Figure 3-2. Reconstructed 24-hour light extinction in U.S. urban areas in 2003 3-6
Figure 3-3. Isopleth map of annual total reconstructed PMio light extinction
based on 2000-2004 IMPROVE data 3-8
Figure 3-4. January and August monthly average diurnal profiles of PM2.s components
derived from the 2004 CMAQ modeling platform, for the Detroit study
area 3-18
-------
Figure 3-5. Sequence of steps used to estimate hourly PM2.5 components and PMio light
extinction 3-23
Figure 3-6. Example from Detroit study area 3-25
Figure 3-7. Distribution of PM parameters and relative humidity across the 2005-2007 period,
by study area 3-35
Figure 3-8. Distributions of estimated daylight 1-hour PMio light extinction and maximum
daily daylight 1-hour PMio light extinction (in Mm"1 units) across the 2005-2007
period, by study area (excluding hours with relative humidity greater than 90
percent) 3-39
Figure 3-9. Distributions of 1-hour PMio light extinction levels by daylight hour (x-axis)
across the 2005-2007 period, by study area (excluding hours with relative
humidity greater than 90 percent) 3-42
Figure 3-10. Distributions of 1-hour relative humidity levels by daylight hour across the 2005-
2007 period, by study area (excluding hours with relative humidity greater than
90 percent) 3-44
Figure 3-11. Scatter plot of daylight 1-hour relative humidity (percent) vs. reconstructed PMio
light extinction (Mm"1) across the 2005-2007 period, by study area (excluding
hours with relative humidity greater than 90 percent) 3-45
Figure 3-12. Tile Plots of Hourly PMio Light Extinction 3-48
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-hour PMio light Extinction and for the Top 2 Percent of Individual Daylight
Hours 3-65
Figure 3-14. Average PMio light extinction budgets for the 15 cities for hours in the (a) top
10 percent of the maximum daily PMio light extinction and (b) top 2 percent
of all daylight hours of PMiolight extinction 3-81
Figure 4-1. Comparison of daily max and all daylight hour design values for PMio
light extinction 4-6
Figure 4-2. Comparison of required percentage reductions in non-PRB PMio light extinction
needed to meetNAAQS scenarios 4-10
Figure 4-3. Distributions of daily maximum daylight 1 -hour PMio light extinction under two
"just meet" secondary NAAQS scenarios (excluding hours with relative humidity
greater than 90 percent) 4-14
VI
-------
LIST OF ACRONYMS/ABBREVIATIONS
AQS
BAM
BC
CAA
CAIR
CASAC
CBS A
CCN
CDPHE
CMAQ
CONUS
CPL
CRA
CSA
CSN
CTM
ORE
dv
EPA
FEM
FRM
GEOS
IMPROVE
ISA
Km
LCD
LOESS
Mm
MSA
N
NAAQS
NARSTO
NCEA
NOAA
EPA's Air Quality System
Beta Attenuation Mass Monitor
British Columbia
Clean Air Act
Clean Air Interstate Rule
Clean Air Scientific Advisory Committee
Consolidated Business Statistical Area
Cloud Condensation Nuclei
Colorado Department of Public Health and Environment
Community Multiscale Air Quality
CMAQ simulations covering continental US
Candidate Protection Level
Charles River Associates
Consolidated Statistical Area
Chemical Speciation Network
Chemical Transport Model
Direct Radiative Effects
Deciview
United States Environmental Protection Agency
Federal Equivalent Method
Federal Reference Method
Global Scale Air Circulation Model
Interagency Monitoring of Protected Visual Environment
Integrated Science Assessment
Kilometer
Liquid Crystal Display
Locally weighted Scatter Plot Smoothing
Megameter
Metropolitan Statistical Area
Nitrogen
National Ambient Air Quality Standards
North American Research Strategy for Tropospheric Ozone
National Center for Environmental Assessment
National Oceanic and Atmospheric Administration
vn
-------
NOx
NFS
NRC
NWS
OAQPS
OAR
OMB
ORD
PA
PM
PM2.5
PM10
PMiQ-2.5
PRB
REA
RF
RH
SANDWICH
SEARCH
SMOKE
S
S02
sox
STP
TEOM
UBC
UFVA
VAQ
Nitrogen oxides
National Park Service
National Research Council
National Weather Service
Office of Air Quality Planning and Standards
Office of Air and Radiation
Office of Management and Budget
Office of Research and Development
Policy Assessment
Particulate Matter
Particles with a 50% upper cut-point of 2.5 um aerodynamic
diameter and a penetration curve as specified in the Code of
Federal Regulations.
Particles with a 50% upper cut-point of 10± 0.5 um aerodynamic
diameter and a penetration curve as specified in the Code of
Federal Regulations.
Particles with a 50% upper cut-point of 10 um aerodynamic
diameter and a lower 50% cut-point of 2.5 um aerodynamic
diameter.
Policy Relevant Background
Risk and Exposure Assessment
Radiative Forcing
Relative Humidity
Skilfate, Adjusted Nitrate, Derived Water, Inferred Carbonaceous
mass approach
Southeastern Aerosol Research and Characterization Study
Sparse Matrix Operator Kernal Emissions
Sulfur
Sulfur Dioxide
Sulfur Oxides
Standard Temperature and Pressure
Tapered Element Oscillating Microbalance
University of British Columbia
Urban-Focused Visibility Impact Assessment
Visual Air Quality
Vlll
-------
1 INTRODUCTION
The U.S. Environmental Protection Agency (EPA) is presently conducting a review of
the 2006 national ambient air quality standards (NAAQS) for particulate matter (PM). Sections
108 and 109 of the Clean Air Act (Act) govern the establishment and periodic review of the
NAAQS. The NAAQS are to be based on air quality criteria, which are to accurately reflect the
latest scientific knowledge useful in indicating the kind and extent of identifiable effects on
public health or welfare that may be expected from the presence of the pollutant in ambient air.
The EPA Administrator is to promulgate and periodically review, at no later than five-year
intervals, "primary" (health-based) and "secondary" (welfare-based) NAAQS for such
pollutants. Based on periodic reviews of the air quality criteria and standards, the Administrator
is to make revisions in the air quality criteria and standards, and to promulgate any new
standards, as may be appropriate. The Act also requires that an independent scientific review
committee advise the Administrator as part of this NAAQS review process, a function performed
by the Clean Air Scientific Advisory Committee (CASAC).
The current suite of secondary standards for PM2.5 and PMio were set in 2006 to be
identical to the primary standards, on the basis that these standards would, in conjunction with
the Regional Haze Program1, provide appropriate protection to address PM-related welfare
effects, including visibility impairment, effects on vegetation and ecosystems, materials damage
and soiling, and effects on climate change (71 FR 61144, October 17, 2006). At that time, the
EPA revised the level of the 24-hour PM2.5 primary standard to 35 ug/m3 (calculated as a 3-year
average of the 98th percentile of 24-hour concentrations at each population-oriented monitor),
retained the level of the PM2.5 annual primary standard at 15 ug/m3 (calculated as the 3-year
average of the weighted annual mean PM2.5 concentrations from single or multiple community-
oriented monitors), and revised the form of the annual PM2 5 primary standard by narrowing the
constraints on the optional use of spatial averaging2. With regard to the primary standards for
PMio, EPA retained the 24-hour PMio standard at 150 ug/m3 (not to be exceeded more than once
per year on average over 3 years) and revoked the annual standard because available evidence
generally did not suggest a link between long-term exposure to current ambient levels of coarse
particles and health effects. The 2006 primary standards were based primarily on a large body of
1 See http://www.epa.gov/air/visibilitv/program.html for more information on EPA's Regional Haze
Program.
2 In the revisions to the PM NAAQS finalized in 2006, EPA tightened the constraints on the spatial
averaging option limiting the conditions under which some areas may average measurements from multiple
community-oriented monitors to determine compliance (see 71 FR 61165-61167, October 17, 2006).
1-1
-------
epidemiological evidence relating ambient PM concentrations to various adverse health
outcomes. (As noted below, portions of the 2006 decision were reversed and remanded by the
Court of Appeals for the District of Columbia Circuit.)
In the Integrated Review Plan for the National Ambient Air Quality Standards for
Particulate Matter, March 2008 (US EPA, 2008a), developed early in the current review of the
PM NAAQS,3 the EPA outlined the science policy questions that frame this review, outlined the
process and schedule for the review, and provided descriptions of the purpose, contents, and
approach for developing the key documents that will be developed in the review.4 EPA has
completed the process of assessing the latest available policy-relevant scientific information to
inform the review of the PM standards. The final assessment is contained in the final Integrated
Science Assessment for Particulate Matter (ISA, US EPA, 2009a) which was released in
December 2009. The final PM ISA includes a summary of the scientific evidence for the
relationship of PM to visibility effects, remote area and urban haze conditions, the PM
components responsible for visibility impacts, and studies of public preference with respect to
urban visibility conditions.
Building upon the visibility effects evidence presented in the PM ISA, as well as CASAC
advice (Samet, 2009a and b) and public comments on the plan for and first draft of the Urban-
Focused Visibility Assessment (UFVA) (US EPA, 2009b, c), EPA's Office of Air Quality
Planning and Standards (OAQPS) developed a second draft UFVA (US EPA, 2010a) which
described the quantitative assessments conducted by the Agency to support the review of the
secondary PM standards. This second draft document presented the methods, key results,
observations, and related uncertainties associated with the quantitative analyses performed and
was reviewed and discussed by CASAC and the public at a March 10-11 meeting. Based on
input received at the March 2010 meeting and in a subsequent letter (Samet, 2010a), this final
UFVA document includes the following changes: 1) inclusion of the complete logit memo in
Appendix J and streamlining of logit discussion in chapter 2 to reduce redundancy and reflect
this addition; 2) Figure 3-13 and associated text was modified to provide a more consistent
comparison of speciated PM mass for the top 10% and 2% of maximum daylight hours and all
daylight hours, respectively; 3) addition of footnotes and caveats in the text to acknowledge that
3 See http://www.epa.gov/ttn/naaqs/standards/pm/sjm index.html for more information on the current and
previous PM NAAQS reviews.
4 On November 30, 2007, EPA held a consultation with the Clean Air Scientific Advisory Committee
(CASAC) on the draft IRP (Henderson, 2008). Public comments were also requested on the draft plan and
presented at that CASAC teleconference. The final IRP incorporated comments received from CASAC and the
general public on the draft plan as well as input from senior Agency managers. CASAC is an independent scientific
advisory committee established to meet the requirements of section 109(d)(2) of the Clean Air Act See
http://vosemite.epa.gov/sab/sabpeople.nsf/WebCommittees/CASAC for more information, and, in particular,
information on the CASAC PM Review Panel activities.
1-2
-------
the St. Louis data is now considered unrealistically high and is not carried forward into the
second draft PA5; 4) modification of Table 3-6 to correctly omit non-daylight hours which were
inadvertently included in the second draft UFVA, report results for four study areas for which
results were missing in the second draft UFVA, and separate "mist" from "smoke/haze" to
reflect that "mist" is a natural condition while "smoke/haze" is not always a natural condition;
and 5) addition of an integrative summary (chapter 5).
In addition, a preliminary draft PA (US EPA, 2009d) was released in September 2009 for
informational purposes and to facilitate discussion with CASAC at the October 5-6, 2009
meeting on the overall structure, areas of focus, and level of detail to be included in the PA. This
preliminary draft PA was discussed in conjunction with CASAC review of and public comment
on the second draft ISA, first draft UFVA, and first draft health risk assessment documents
produced in support of this PM NAAQS rulemaking. CASAC comments on the preliminary
draft PA were considered in developing the first external review draft PA (US EPA, 201 Ob). The
first draft PA, which built upon the information presented in the final ISA and second draft
UFVA, was released for CASAC review and public comment in February of 2010 (US EPA,
201 Ob). EPA presented an overview of the first draft PA at the CASAC meeting on March 10,
2010. CASAC and public review of the first draft PA was discussed during public
teleconferences on April 8-9, 2010 (75 FR 8062, February 23, 2010) and May 7, 2010 (75 FR
19971, April 16, 2010). CASAC (Samet, 201 Ob) and public comments on the first draft PA were
considered in developing the second draft PA which will be reviewed by CASAC at an
upcoming meeting scheduled for July 26-27, 2010 (75 FR 32763), June 9, 2010).
The PA is intended to help "bridge the gap" between the Agency's scientific assessments,
presented in the ISA and UFVA, and the judgments required of the Administrator in determining
whether it is appropriate to retain or revise the secondary PM standards. The PA is intended to
provide a transparent staff analysis of the scientific basis for alternative policy options for
consideration by senior EPA management prior to rulemaking by integrating and interpreting
information from the ISA and the UFVA to frame policy options and to facilitate CAS AC's
advice to the Agency and recommendations on any new standards or revisions to existing
standards as may be appropriate, as provided for in the Clean Air Act. A second draft PA (US
EPA, 2010c) has been released in conjunction with this final UFVA document.
5 Comments concerning unrealistically high PM10.2 5 values for St. Louis are viewed as credible, but were
received too late in the review process to permit reanalysis using an alternate data set or to remove St. Louis from all
portions of this document. However, the text has been revised to caution readers with respect to the St. Louis
results, and they are not included in the visibility effects discussion in the second draft PM Policy Assessment
document. Some graphics have been updated to exclude St. Louis results in this final UFVA.
1-3
-------
1.1 PM NAAQS BACKGROUND
In the review of the secondary PM NAAQS completed in 2006, EPA took into account
that the Regional Haze Program, authorized under sections 169 A and 169B of the CAA, was
established to address all human-caused visibility impairment in federal Class I areas. The
national goal of this program is to prevent any future, and remedy any existing, impairment of
visibility in mandatory class I Federal areas (Class I areas) which results from manmade air
pollution. This program also mandates that states develop SIPs to ensure that reasonable
progress is made towards meeting those goals. Because Congress explicitly targeted Class I
areas for this pristine level of protection, it can be concluded that Congress did not envision such
a stringent goal in non-Class I areas. See American Trucking Ass'n v. Browner, 175 F. 3d 1027,
1056-57 (D. C. Cir. 2002) (upholding this position). However, Congress recognized that
visibility impairment can and often does occur in areas outside federal Class I areas, including
urban areas and judged that protection from visibility impairment was important in those areas as
well. In this regard, Congress included visibility effects in the definition of public welfare
effects that should be protected under the national ambient air quality standards (NAAQS)
program authorized in sections 108 and 109 of the CAA. As a result, EPA may establish
secondary standards addressing visibility impairment notwithstanding existence of the Regional
Haze Program. Under the NAAQS program, it is up to the Administrator to judge the requisite
level of public welfare visibility protection.
Recognizing that efforts were underway to provide increased protection to Class I areas
under the Regional Haze Program, EPA focused the 2006 PM NAAQS review on visibility
impairment in non-Class I areas. Because most of the available non-Class I PM data came from
PM monitoring sites located primarily in urban areas, the assessments took on an urban focus. In
addition, EPA considered available information on people's preferences for different levels of
visual air quality which came from studies conducted in urban areas and information regarding
existing urban visibility programs and goals.
In an effort to minimize the factors that historically had complicated efforts to address
visibility impairment nationally, (i.e., the substantial East/West differences in factors
contributing to impairment in Class I areas), EPA staff noted that with respect to fine particles,
East/West differences in urban areas are substantially smaller than in rural areas. Further,
relative humidity levels, though generally higher in eastern than western areas, are appreciably
lower in both regions during daylight, as compared to nighttime, hours. The PM2.5 data available
at that time in urban areas were obtained using a filter -based Federal Reference Method (FRM)
which captures ambient PM2.5 on a filter and then dries it to get the dry PM2.5 mass
concentration. By drying the sample, most water and to some extent other labile PM compounds
evaporate so that the original characteristics (e.g., particle size and composition) of the ambient
1-4
-------
PM are altered. Using PM and meteorological data from 161 cities, EPA staff assessed the
correlations between PM2.5 levels and reconstructed (RE) (i.e., calculated) light extinction during
daylight hours for different regions of the country. This assessment showed that the strongest
correlation in the relationship of ambient PM light extinction to dry PM2.5 mass concentration
was during afternoon periods when lower relative humidity conditions generally prevailed in all
regions of the country and ambient PM was drier (Staff Paper, US EPA, 2005). While EPA
recognized that the effect of ambient PM on visibility results from the ambient particle
characteristics of size, concentration, and composition (including associated water) present in the
air in the sight path of the observer, given the data availability at the time, EPA viewed the FRM
altered PM2.5 mass concentration as an acceptable indicator for addressing ambient PM-related
visibility effects at the national scale during afternoon hours. Thus, the 2005 Staff Paper chose
to address the issue of regional differences in terms of averaging time rather than indicator,
discussing the use of a sub-daily afternoon dry PM2.5 standard, because the generally lower
afternoon relative humidity tended to produce a more uniform relationship between light
extinction and dry PM2.s mass concentration throughout the country, therefore providing a more
uniform level of visibility protection nationwide. This more uniform level of visibility
protection, however, was limited to the afternoon hours of the day when relative humidity and
visibility impairment tend to be the lowest.
Based on the above, in the 2005 PM Staff Paper, EPA staff recommended a separate sub-
daily secondary standard to address visibility impairment using dried PM2.s mass concentration
as the indicator, a recommendation endorsed by CASAC. In the 2006 proposal notice, however,
EPA proposed to revise the secondary standards by making them identical to the suite of
proposed primary standards for fine and coarse particles, to provide protection against PM-
related public welfare effects including visibility impairment, effects on vegetation and
ecosystems, materials damage and soiling, and climate, while soliciting comment on adding a
new sub-daily PM2.s secondary standard to address visibility impairment primarily in urban areas
(71 FR 2620). CASAC provided additional advice to EPA in a letter to the Administrator
requesting reconsideration of CAS AC's recommendations for both the primary and secondary
PM2.5 standards as well as standards for thoracic coarse particles (Henderson, 2006). With
regard to the secondary standard, CASAC reaffirmed "... the recommendation of Agency staff
regarding a separate secondary fine particle standard to protect visibility.... the CASAC wishes
to emphasize that continuing to rely on primary standards to protect against all PM-related
adverse environmental and welfare effects assures neglect, and will allow substantial continued
degradation, of visual air quality over large areas of the country" (Henderson, 2006).
On September 21, 2006, EPA announced its final decisions to provide increased
protection of public welfare by making the secondary NAAQS identical to the revised primary
1-5
-------
standards (71 FR 61144, October 17, 2006). This suite of secondary standards was designed to
address both visibility and other non-visibility welfare related effects. Specifically, with regard
to the secondary welfare effect of visibility impairment, the Administrator believed that revising
both the 24-hour and annual PM2.5 secondary standards to be identical to the revised suite of
PM2.5 primary standards was a reasonable policy approach to address visibility impairment
primarily in urban areas. With regard to the other non-visibility PM-related welfare effects such
as vegetation and ecosystems, materials damage and soiling, and climate, the Administrator
concluded that it was appropriate to address these effects by revising the current suite of PM2.5
secondary standards, making them identical in all respects to the suite of primary PM2.5
standards, while retaining the current 24-hour PMio secondary standard and revoking the current
annual PMio secondary standard. In particular for coarse particles, EPA retained PMio as the
indicator for purposes of regulating the coarse fraction of PMio and retained the 24-hour
secondary PMio standard at 150 |ig/m3 and revoked the annual secondary PMio standard.
Several parties filed petitions for review following promulgation of the revised PM
NAAQS in 2006. These petitions addressed a number of issues, including the decision to set the
secondary PM2.5 standards identical to the primary standards. On judicial review the court
remanded the secondary PM2.5 NAAQS to EPA because the Agency failed to adequately explain
why setting the PM2.5 secondary standards equal to the primary PM2.5 standards provided the
required protection from visibility impairment. In particular, the Agency failed to identify a
target level of visibility impairment that would be requisite to protect the public welfare, and
improperly relied on a misleading comparison of the number of counties which would be in
nonattainment for the revised primary NAAQS compared to one alternative secondary standard
under consideration. Among other things, this equivalence analysis failed to address the issue of
regional differences in humidity-related effects on visibility. American Farm Bureau Federation
v. EPA, 559 F. 3d 512, 530-31 (D.C. Cir. 2009).
The analyses developed for and described in this document reflect consideration of the
issues raised by the court. In particular, a) the reanalysis of public preference studies (described
in Chapter 2) provides information useful for the selection of "target levels" for urban visibility
protection; b) the analyses of the factors contributing to visibility impairment for selected urban
areas, including PM species component contributions and variations in relative humidity, provide
information useful for better characterization of regional differences important for development
of a national standard (Chapter 3); and c) the analyses of alternative standards (Chapter 4) using
different combinations of indicator, averaging times, levels and forms provide information useful
in understanding the degree of visibility protection provided by alternative standards.
1-6
-------
1.2 VISIBILITY EFFECTS SCIENCE OVERVIEW
Light extinction is the loss of light per unit of distance and occurs when light is scattered
and/or absorbed. Particulate matter and gases can both scatter and absorb light. Light scattering
by gases (e.g., nitrogen, oxygen, etc.) that comprise the pollutant free or clean atmosphere (also
known as Rayleigh or clean-air scattering) is related to the density of the air, which is
sufficiently constant with elevation that it can be taken to be a time invariant constant that
depends principally on elevation above sea level. NC>2 is the only atmospheric pollutant gas that
absorbs light appreciably and its effects are generally small (i.e., less than 5%) compared to PM
light extinction. Hereinafter the phrase "PM light extinction" indicates that the Rayleigh
contribution to light extinction (nominally considered 10 Mm"1) has been subtracted out and the
NC>2 contribution is considered negligible or is simply excluded due to the measurement
approach used. By contrast, the term "light extinction" or "total light extinction" is meant to
include both the Rayleigh and NC>2 contributions.
Visual air quality (VAQ) is defined as the visibility effect caused solely by air quality
conditions and excluding those associated with meteorological conditions like fog and
precipitation. It is commonly measured as either light extinction (in terms of inverse
megameters, Mm"1) or the haziness index (in terms of deciview, dv) (Pitchford and Malm, 1993).
The haziness index measured in deciview units was developed for use in visibility perception
studies because it has a more linear relationship to perceived changes in haze compared with
light extinction. It is defined as ten times the natural logarithmic of one tenth of the light
extinction in inverse megameter units (Mm"1) (Pitchford and Malm, 1993). Light extinction and
haziness are physical measures of the amount of visibility impairment (e.g., the amount of
"haze"), with both increasing as the amount of haze increases. Visual range, defined as the
greatest distance that a large dark object can be seen, was developed for military and
transportation safety use. Under conditions that meet certain standard assumptions, visual range
is inversely related to light extinction (Pitchford and Malm, 1993).
PM is a heterogeneous mixture of particles of different sizes and chemical compositions.
While visibility impairment has been associated most often with PM2.5, larger particles such as
those found in PMi0 may be a significant contributor in some areas. Thus, the UFVA considers
the visibility impairment caused by all particles 10 microns or smaller. As stated above, the
degree of visibility impairment caused by a given mass of PM depends in large part on the size,
density and chemical composition of the PM. If the ambient PM has a large number of
hygroscopic particles (i.e., particles that readily absorb moisture from the atmosphere), and also
occurs when the relative humidity of the air is higher, those particles will grow larger in size and
have a larger haze effect than if those same particles occurred in ambient air with lower relative
humidity.
1-7
-------
Ambient PM light extinction is most accurately determined by direct measurements.
However, as shown in Figure 1-1, the ambient PM light extinction can also be estimated from
dry PM mass and composition data and relative humidity using an algorithm. One well
established algorithm, known as the IMPROVE algorithm6, accounts for water present in
hygroscopic PM components and uses assumed light extinction efficiencies for each of the major
PM species. Because there is limited ambient PM light extinction data available in urban areas,
the assessments below will principally use monitored and modeled dry PM mass and species
estimates, along with relative humidity measurements as input to the IMPROVE algorithm for
estimating ambient PM light extinction.
6 Malm, et al., 1994 and DeBell, 2006. (See also ISA, section 9.2.2.2, pgs. 9-7 and 9-8.)
1-8
-------
Figure 1-1. Progression from PM Characteristics to PM Light Extinction That Shows the
Modeling Approach (shaded light green) as well as the Use of Direct
Measurements (shaded blue) as Alternative Ways to Estimate PM Light
Extinction
Measured
Dry PM Mass
Concentration
Dry PM Composition
Ambient Relative
Humidity
Derived
Hygroscopic
Growth Model
Ambient PM
Characteristics
Light Extinction
Model
Ambient PM Light
Extinction
>,
Ambient PM Light
Extinction
The extent to which any amount of light extinction affects a person's ability to view a
scene depends on both scene and light characteristics. For example the appearance of a nearby
object (i.e., a building) is generally less sensitive to a change in light extinction than the
appearance of a similar object at a greater distance. For a scene with known characteristics, the
degradation in the scene associated with a change in light extinction can be determined and the
resulting appearance can be realistically displayed on a digital photograph of the scene using the
WinHaze system7. Figure 1-2 below shows the progression from PM light extinction to
perceived visual air quality impacts to the valuation of those perceived impacts.
7 Molenar, et al., 1994
1-9
-------
Survey studies have used sets of photographs or computer simulated images developed
from a base photo depicting a range of visibility conditions on urban scenes to assess the
individual's opinion on the acceptability of those VAQ conditions. For the specific scenes used
in such studies there is a known or predetermined one-to-one correspondence between the
amount of ambient or computer generated haze captured in the photographs or images,
respectively, and the associated amount of ambient PM light extinction. For visibility preference
studies, visibility levels are generally characterized using the haze index in units of deciview
(similar to the decibel scale for sound).
Figure 1-2. Progression from PM Light Extinction to Value of Improved Visual Air
Quality (VAQ)
Relationship Steps from PM
Light Extinction to VAQ
Ambient PM Light
Extinction
Perceived Visual Air
Quality (Images)
Value of Improved
VAQ
Non-Air Quality
Information
WinHaze
Modeling
Scene and Lighting
Characteristics
Valuation Studies ><
Public-Scene
Contextual
Information
1-10
-------
1.3 GOALS AND APPROACH
The goal of the UFVA is to characterize visibility impairment in 15 selected urban areas
under recent air quality conditions, "just meet" air quality scenarios for both current secondary
PM2.5 standards, and under scenarios using various alternative standards which utilize different
indicators, averaging times, forms and levels to identify those that better reflect the relationship
between ambient PM and visibility impairment. In particular, the UFVA focuses on the use of a
PMio light extinction-based indicator for a possible secondary PM NAAQS (see Figure 1-1 and
1-2). This is done by comparing estimates of hourly PMio light extinction in 15 major U.S.
urban areas over a three-year period (2005-2007) to a range of light extinction values, i.e.,
candidate protection levels (CPLs), beyond which half of the participants in assessed urban
visibility preference studies indicated the haze conditions were unacceptable (see discussion in
chapter 2 below, Stratus Consulting Inc., 2009 and Appendix J). In addition, the UFVA includes
additional characterizations of the effectiveness of the current and an alternative suite of PM2.5
secondary standards8.
The previous PM NAAQS review used the results of visibility preference survey studies
conducted in Denver (1990), Phoenix (2003), and British Columbia (1993) as the basis for
suggesting that a standard set to protect visibility conditions to a level within a visual range from
between about 40 km to about 60 km (corresponding to light extinction from -100 Mm"1 to -67
Mm"1) could represent an appropriate degree of welfare protection from PM9. With the
exception of a small pilot study conducted in Washington, DC in 2001 (9 participants; Abt
Associates Inc., 2001), and a replicate study also conducted for Washington, DC in 2009 (26
participants; Smith and Howell, 2009), there are no additional visibility preference survey studies
upon which to base the selection of CPLs.
The EPA staff, with contractor support, has conducted a more detailed, in-depth
assessment of the results from these studies, including the two Washington, DC studies. This
assessment includes an analysis that combines data from across all studies using graphical and
logit model analysis to examine the consistency of the results between the surveys (Stratus
Consulting Inc., 2009 and Appendix J). Based on the results of this analysis, we have been able
to refine the range of visibility conditions put forth in the 2006 review, that could represent an
appropriate degree of public welfare visibility protection, and to determine a central tendency
value for the CPLs. These analyses and results are described below in chapter 2.
EPA also included an assessment of the sub-daily PM25 mass concentration indicator, which was
explored in the 2005 PM staff paper and which was considered a viable option by EPA staff and CAS AC in the
2006 review. These latter assessments are summarized in Appendix D.
9 Light extinction is inversely related to visual range.
1-11
-------
In the previous PM NAAQS review, the characterization of urban visibility conditions
were based on IMPROVE algorithm estimates using the 2001 to 2003 PM2.5 mass and speciation
data from 161 urban areas by assuming a constant composition for every hour of the day equal to
the 24-hour measured composition and by using either actual or monthly average (10-year mean)
hour of the day relative humidity. Statistical relationships between hourly light extinction
estimates and concurrent hourly PM2.5 mass concentrations were used to show that daytime and
especially afternoon relationships are relatively strong with a similar linear relationship for both
eastern and western urban areas (i.e. R2>0.6, slope ~6 m2/g).
The current assessment of urban visibility conditions (as described in chapter 3) uses a
modeling approach to estimate hourly light extinction using PM2.5 mass and speciation data with
measured relative humidity. However, it differs by replacing the unrealistic assumption of
constant composition for PM2.5, with composition that is made to vary during the day using
urban-specific monthly mean diurnal variations of species concentrations determined from
regional air quality model results, while constraining the means of the hourly species
concentration for each day to closely match the 24-hour duration measured species
concentrations.
1.4 SCOPE OF URBAN-FOCUSED VISIBILITY ASSESSMENT
This section provides an overview of the scope and key design elements of the UFVA,
including the process that has been followed to design the analyses. Following initiation of this
PM NAAQS review in 2007, we began the design of the assessments in the UFVA by revisiting
the analyses completed during the previous PM NAAQS review (Abt Associates Inc., 2001; US
EPA, 2005, chapter 6) with an emphasis on considering key limitations and sources of
uncertainty recognized in that review.
1.4.1 Background
As an initial step in this review, EPA invited a wide range of external experts as well as
EPA staff, representing a variety of areas of expertise to participate in a workshop titled,
"Workshop to Discuss Policy-Relevant Science to Inform EPA's Integrated Plan for the Review
of the Secondary PM NAAQS" (72 FR 34005, June 20, 2007). This workshop provided an
opportunity for the participants to broadly discuss the key policy-relevant issues around which
EPA would structure the PM NAAQS review and to discuss the most meaningful new science
that would be available to inform our understanding of these issues. One session of this
workshop centered on issues related to visibility impacts associated with ambient PM.
Specifically, the discussions focused on the extent to which new research and/or improved
methodologies were available to inform how EPA evaluated visibility impairment in this review.
1-12
-------
Based in part on these workshop discussions, EPA developed a draft IRP outlining the
schedule, the process, and the key policy-relevant science issues that would guide the evaluation
of the air quality criteria for PM and the review of the primary and secondary PM NAAQS,
including initial thoughts for conducting quantitative assessments (US EPA, 2007, chapter 6).
On November 30, 2007, CAS AC held a teleconference with EPA to provide its comments on the
draft IRP (72 FR 63177, November 8, 2007). Public comments were also presented at that
teleconference. A final IRP incorporating comments received from CASAC and the general
public on the draft plan was issued in March 2008 (US EPA, 2008a).
In articulating a rationale for the urban focus of this assessment, we reviewed the
available information and found the following information compelling: 1) PM levels in urban
areas are often in excess of those of the surrounding region since urban haze typically includes
both regional and local contributions (US EPA, 2009a; sections 9.2.3.3 and 9.2.3.4), suggesting
the potential for higher levels of PM-induced visibility impairment in urban areas; 2) the
existence of numerous urban visibility protection programs and goals demonstrating that urban
VAQ is noticed and considered an important value to urban residents (US EPA, 2009a; section
9.2.4); and 3) the existence of large urban populations means that potentially more people are
routinely affected by poor VAQ than in rural areas. These features of urban areas have led EPA
staff to conclude that urban dwellers represent a susceptible population group for adverse PM-
related effects on visibility. However, this conclusion is not meant to imply that there are not
other susceptible populations or individuals living in other non-urban and non-Class I areas that
are currently adversely impacted by ambient PM-related visibility conditions. Unfortunately,
visibility preferences and PM levels in these areas have not been well characterized. Although
this visibility assessment focuses only on selected urban areas, a new secondary PM standard would
apply to all non-Class I areas of the country.
On October 6-8, 2008 the EPA sponsored an urban visibility workshop in Denver,
Colorado to identify and discuss methods and materials that could be used in "next step" projects
to develop additional information about people's preferences for reducing existing impairment of
urban visibility, and about the value of improving urban visibility. Invited individuals came
from a broad array of relevant technical and policy backgrounds, including visual air quality
science, sociology, psychology, survey research methods, economics, and EPA's process of
setting NAAQS. The 23 people who attended the workshop (including one via teleconference
line) came from EPA, the National Oceanic and Atmospheric Administration (NOAA), National
Park Service, academia, regional and state air pollution planning agencies, and consulting
1-13
-------
firms.10 The information discussed at this Workshop was useful in informing subsequent steps in
the process.
1.4.2 Selection of Alternative Scenarios for First Draft Assessments
In designing the quantitative assessments to include in the first draft UFVA, EPA staff
developed a planning document outlining the initial design for the PM NAAQS visibility
assessment - Paniculate Matter National Ambient Air Quality Standards: Scope and Methods
Plan for Urban Visibility Impact Assessment, henceforth Scope and Methods Plan (US EPA,
2009b). This planning document was released for CASAC consultation and public review in
February 2009. Based on consideration of CASAC and public comments on the Scope and
Methods Plan, along with ongoing review of the latest PM-related literature, several aspects of
the original scope of the urban visibility conditions assessment, as depicted in Figure 1-1 of
section 1.3 of the Scope and Methods document (US EPA, 2009b), were modified in the first
draft UFVA (US EPA, 2009c). Taking into account the nature of urban versus more remote area
PM composition, and input received at the April 2, 2009 CASAC meeting, EPA staff concluded
that it was unnecessary to develop a new urban-optimized algorithm at this time and that it
remained appropriate in the context of this assessment to use the original IMPROVE algorithm
to relate urban PM to local haze (PM light extinction). One of the primary reasons for initially
considering an urban-optimized algorithm was a concern that the organic components of PM in
urban areas, being generally nearer their emission sources, would have a lower ratio to the
measured organic carbon mass than the ratio of organic component mass to measured organic
carbon mass currently used for the more aged PM organic components found in remote areas.
As described below in chapter 3, this concern has been addressed by using the SANDWICH
mass balance approach to estimate the PM organic component mass, which negates the need to
estimate organic component mass from measured organic carbon mass.
With regard to the urban visual air quality preference assessment described in the Scope
and Methods document (US EPA, 2009b, section 1.3), more significant modifications occurred.
EPA staff decided to conduct a reanalysis of the urban visibility preference studies available at
the time of the 2006 PM NAAQS review, rather than conduct new public preference studies, as it
has become apparent that the results of these studies would be unlikely to be completed in time
to inform this review. Recognition that the initial plans described in the Scope and Methods
document were possibly overly ambitious was also shared by members of CASAC (see
individual member comments; Samet, 2009a). The analysis, therefore, relied on existing, rather
than new, urban visibility preference studies and was designed to explore the similarities and
10 To view the complete report from the October 2008 urban visibility workshop, see:
http://vista.cira.colostate.edu/improve/Publications/GrayLit/grav literature.htm
1-14
-------
differences (comparability) among these studies. Information drawn from these results informed
the selection of VAQ CPLs (described in chapter 2 below) to be used in subsequent impact
assessments. Further, information presented during the public comment phase of the April 2,
2009 CAS AC meeting and later provided to EPA staff, led to the inclusion of a recent study by
Smith and Howell (2009) for Washington, DC in the reanalysis.
1.4.3 Selection of Alternative Scenarios for Second Draft Assessments
The first draft UFVA was reviewed at an October 2009 CASAC meeting, and a CASAC
letter providing its advice and recommendations was submitted to the Administrator in
November 2009 (Samet, 2009b). In its letter, the CASAC indicated support for EPA staffs
approach to evaluating the nature and degree of PM-related visibility impairment, including
EPA's focus on non-Class I areas, including in particular, urban areas as an "effective
complement" to the Regional Haze Rule. In this regard, CASAC expressed support for
consideration of a new PM light extinction indicator, a one hour averaging time, and for the
range of selected candidate protection levels.
• Indicator: PMio Light Extinction
There are a number of different ways to measure ambient PM: particle counts, surface
area, volume, mass concentrations, and concentration of components. Each of these different
characteristic of ambient PM can be important in the context of different effects. For example,
particle count may be important from the perspective of cloud formation or to characterize the
abundance of ultrafine PM, which is of interest for health effects. In a similar way PM light
extinction measures the characteristic of ambient PM most relevant and directly related to the
effect of PM visibility impairment. Thus, as described in the Scope and Methods document (US
EPA, 2009b) and first and second drafts of the UFVA, EPA staff has continued to focus
assessments in this final document in terms of ambient PMio light extinction as the indicator for
PM visibility impairment, instead of the traditional PM2 5 mass concentration. Unlike the current
FRM measurement method for PM mass concentration, which generally changes the
composition and size of the particles by driving off most of the water, direct measurement of
ambient PM light extinction captures the PM-induced visibility impairment of the particles as
they exist in the atmosphere. PM light extinction, like conventional PM mass concentration, is a
measurable physical characteristic of atmospheric PM. PM light extinction can also be
calculated using a simple algorithm such as the IMPROVE algorithm11
Section 109 (b) (2) of the CAA states that "Any national secondary ambient air quality
standard prescribed under subsection (a) of this section shall specify a level of air quality the
11 Malm, et al, 1994 and DeBell, 2006. (See also ISA, section 9.2.2.2, pgs. 9-7 and 9-8.)
1-15
-------
attainment and maintenance of which ... is requisite to protect the public welfare from any
known or anticipated adverse effects associated with the presence of such air pollutant in the
ambient air...." (emphasis added). In addition, section 108 (a) (2) states that the air quality
criteria "for an air pollutant shall accurately reflect the latest scientific knowledge useful in
indicating the kind and extent of all identifiable effects on public health or welfare which may be
expected from the presence of such pollutant in the ambient air, in varying quantities. The
criteria ... shall include information on (A) those variable factors (including atmospheric
conditions) which of themselves or in combination with other factors may alter the effects on
public health or welfare of such air pollutant;..." (emphasis added). Thus, EPA staff believes
that the visibility effects of PM important to the public welfare are precisely the visibility effects
of PM occurring in the ambient air, which necessarily entails association with ambient
atmospheric conditions that affect the nature or magnitude of the PM visibility effect. These
ambient conditions lead to constant changes in the size and composition of particles as these
particles come in contact with other pollutants or natural components, become oxidized/age as
they are transported great distances, and shrink or grow in the absence or presence of water
vapor, or other atmospheric gases. The combined effect of all these interactions of ambient PM
with real time atmospheric conditions and chemistry on the public welfare effect of visibility
impairment depends on factors other than dry PM mass concentration alone. Use of PMio light
extinction as the indicator for a secondary PM NAAQS is thus a more direct measure of the
relationship between ambient PM and the public welfare effect of visibility impairment than any
dry PM mass concentration (either PM2.5 or any other dry mass fraction).
• Averaging times: Daylight Daily Max. 1 Hour or All Daylight Hours
It is necessary to also identify an averaging time to apply along with the CPLs in the
assessments described in chapters 3 and 4. Because the nature of visibility impairment and its
impact on the public welfare is sufficiently different and less well understood at night, this
assessment only considers visibility conditions that occur during daylight hours. Though not
directly supported by preference or other studies, EPA staff believes that a short averaging time
(e.g. an hour) may be more appropriate than longer time periods (e.g. multiple hours) since VAQ
impacts are instantaneously perceived. This is also consistent with staffs belief that most
individuals in an urban setting experience urban VAQ in relatively short-term incidental and
intermittent periods when they have the opportunity to be outdoors (e.g. during commutes to
work, school, shopping, etc.). Since this fraction of the public may experience poor VAQ during
a relatively small time period and not have the opportunity to see it improve later during the
same day, it seems appropriate to EPA staff to consider assessing the current and projected
conditions in chapters 3 and 4 by comparing the 1-hour daily maximum light extinction to each
1-16
-------
of the three CPLs supported by the preference studies. There is uncertainty associated with
predicting the duration of the effect associated with such brief periods of exposures, i.e., it is not
known how long the person remembers the poor VAQ conditions once he/she goes indoors and
is removed from the sight.
Alternately, a complementary fraction of the public may have multiple or continuing
opportunities to experience visibility throughout the day. People in this situation can experience
a variety of conditions ranging from improvement, maintenance, or diminished VAQ throughout
the day. For them, a day with several hours that exceed acceptable VAQ levels may represents a
greater impact on their wellbeing than on a day with only one such hour. To assess impacts
more related to this portion of the population, in which the degree of impact depends upon the
conditions present across multiple hours of exposure, EPA staff has also considered all daylight
hours which have light extinction levels beyond the three CPLs, as well as the 1-hour daily
maximum light extinction in the assessments described in chapters 3 and 4.
• Level: Candidate Protection Levels (CPLs)
In order to identify a range of light extinction levels associated with acceptable VAQ to
compare to current and projected conditions in the assessment in chapters 3 and 4 of this
document, CPLs have been selected in a range from 20 dv to 30 dv (74 Mm"1 to 201 Mm"1) based
on the composite results and the effective range of 50th percentile acceptability across the four
urban preference study areas shown in Figure 2-16. A midpoint of 25 dv (122 Mm"1) was also
selected for use in the assessment. These three values provide a low, middle, and high set of
light extinction conditions that are used in subsequent sections of the UFVA to define daylight
hours with urban haze conditions that have been judged unacceptable by the participants of these
preference studies. As discussed in greater detail in section 1.2 above, PM light extinction is
taken to be light extinction minus the Rayleigh scatter (i.e. light scattering by atmospheric gases
is about 10 Mm"1) and NC>2 contribution (assumed to be negligible), so the PM light extinction
levels that correspond to low, middle and high CPLs are about 64 Mm"1, 112 Mm"1 and
191 Mm"1, respectively.
• Forms: Percentiles and Relative Humidity Constraints
In considering an appropriate range of forms to consider in the analyses of alternative PM
light extinction visibility standards analyzed in chapter 4 of this final UFVA, staff considered
what frequency of conditions at or below the CPLs should be considered acceptable. Again,
none of the preference studies provided insight into this aspect of acceptability. Because the
nature of the public welfare effect is one of aesthetics and/or on feelings of wellbeing and not
directly related to a physical health outcome, EPA staff believes that it is not necessary to
1-17
-------
eliminate all such exposures and that some number of hours/days with poor VAQ can reasonably
be tolerated. In the first draft UFVA, staff selected the 90th and 95th percentiles to assess. In the
CAS AC letter following the review of the first draft UFVA, CAS AC recommended that other
percentiles be considered, up to and including the 98thpercentile used for the current 24-hour primary
and secondary standards. EPA staff has therefore considered the 90th, 95th and 98th percentiles per
year in the second and final iterations of this document. Due to inter-annual variability in
meteorology and other circumstances that affect air quality, EPA staff is recommending using a
three year average form of the standard for purposes of consistency and stability, as is the current
24-hour primary PM2.5 standard. By considering all of the combinations of the two hourly forms
(i.e. each daylight hour and daylight 1-hour daily maximum), the three CPLs and the three
frequencies, a total of 18 separate alternative secondary PM NAAQS scenarios were generated
for use in the assessments described below in chapters 3 and 4 (See table 4-1). An additional
CASAC recommendation, that the relative humidity (RH) limit be lowered from 95% to 90%
and used as a screen (i.e., hours above it should be discarded) rather than as a cap, to more
clearly exclude weather events like fog or precipitation and to minimize effects of measurement
error and spatial variability, was also incorporated into the second and final iterations of this
document.
1.5 ORGANIZATION OF DOCUMENT
The remainder of this document is organized as follows: Chapter 2 includes an analysis
of the urban visibility preference studies with a discussion of similarities and differences
regarding the approaches and methods used and results obtained for each study. This chapter
also includes a summary discussion of the results of a composite assessment of the combined
results from the four urban areas (Denver, Phoenix, British Columbia, and Washington, DC, an
accompanying logit (statistical) analysis, and use of these results in the selection of the
alternative levels evaluated in the remainder of the assessment. The complete description of the
logit analyses is found in Appendix J. Chapter 3 describes the analytical approach, methods, and
data used in conducting the assessment of recent urban visibility conditions, both in terms of PM
mass concentration and PM light extinction for the set of urban case studies included in this
analysis. Selected results are presented in chapter 3, with additional results found in the
Appendices. Chapter 4 presents estimates of PM mass concentration and PMio light extinction
conditions generated for the urban case studies for two alternative PM2.5 mass concentration
levels and for the three light extinction CPLs. Additional information regarding approaches,
results, method validation studies and uncertainty assessments for both chapters 3 and 4 are
presented in Appendices A-J).
1-18
-------
2 URBAN VISIBILITY PREFERENCE STUDIES
The purpose of this chapter is to describe the reanalysis of available urban visibility
preference studies conducted by EPA staff with contractor support. The reanalysis covered the
three completed urban visibility preference survey studies plus a pair of smaller focus studies
designed to explore and further develop urban visibility survey instruments. The three
completed survey studies (all in the west) included Denver, Colorado (Ely et al., 1991), one in
the lower Fraser River valley near Vancouver, British Columbia (BC), Canada (Pryor, 1996),
and one in Phoenix, Arizona (BBC Research & Consulting, 2003). The first pilot focus group
study was conducted in Washington, DC on behalf of EPA to inform the 2006 PM NAAQS
review (Abt Associates Inc., 2001). In response to an EPA request for public comment on the
Scope and Methods Plan (74 FR 11580, March 18, 2009) for the current review, Dr. Anne Smith
provided comments (Smith, 2009) about the results of a new Washington, DC focus group study
that had been conducted using methods and approaches similar to the method and approach
employed in the EPA pilot study (Smith and Howell, 2009). When taken together, these studies
from the four different urban areas included a total of 852 individuals, with each individual
responding to a series of questions answered while viewing a set of images of various urban
VAQ conditions. The apparent similarity in the methods used across the studies made it appear
initially that the studies were comparable.
However, in order to ensure that our basis for selecting an appropriate range of CPLs was
sound, we, along with contractor support, undertook a detailed reanalysis to determine the
robustness of the survey study results, the appropriateness of comparing each study's results to
the others, and the key uncertainties relevant to data interpretation. This reanalysis included a
statistical analysis using a logit regression model to assess the comparability of different datasets.
Limited discussion of logit model results occurs in the body of this chapter when pertinent to
informing staff judgments regarding comparability of study results. A detailed description of the
logit assessment is provided in the contractor memo included as Appendix J of this document.
The following sections (sections 2.1 to 2.5) examine in detail the study methods used and results
obtained from each of the available studies.
2-1
-------
2.1 METHODS USED IN VISIBILITY PREFERENCE STUDIES
In all but one1 of the visibility preference studies assessed in this document, participants
were shown a series of different VAQ conditions projected on a large screen using a slide
projector. In the earliest two studies (the Denver and lower Frazer River Valley British
Columbia studies) the range of VAQ conditions were presented by projecting photographs
(slides) of actual VAQ conditions. The photographs were taken on different days from the same
location, and presented the same scene. Photographs were selected to avoid depicting significant
weather events (e.g., rain, snow, or fog), and where measured extinction data were available
from the time the photograph was taken.
The Phoenix study and the Washington, DC projects used computer generated
photographic-quality images to present different VAQ conditions. Using an original near-
pristine base photograph, additional images representing a range of VAQ conditions were
generated using the WinHaze software program, which is based on a technique described in
Molenar et al. (1994). The Phoenix study and the 2001 Washington, DC project projected slides
of digital images prepared by WinHaze. The 2009 Washington, DC project presented images
directly from the desktop version of WinHaze using either a liquid crystal display (LCD)
projector or a computer monitor.
WinHaze analysis synthetically superimposes a uniform haze on a digitized, actual
photograph. The WinHaze computer algorithm calculates how a given extinction level would
impair the appearance of each individual portion of the photograph. A major advantage of
presenting WinHaze-generated images is that they provide viewers depictions of alternative
VAQ levels, with each image containing exactly the same scene, with identical light angle, time
of day properties, weather conditions, and specific scene content details (e.g., the amount of
traffic in a intersection). Additional details about WinHaze, and a discussion of the applicability
of WinHaze images for regulatory purposes, is in the 2004 PM Criteria Document (U.S. EPA,
2004). The desktop version of WinHaze is available online (Air Resources Specialists, 2003).
The first urban visibility preference study (Denver, CO; Ely et al., 1991) developed the
basic survey method used in all the subsequent studies. Although there are variations in specific
details in each study, all the studies use a similar overall approach (key variations are discussed
in the section on each study later in this chapter). This approach consisted of conducting a series
of group interview sessions, where the participants were shown a set of photographs or images of
alternative VAQ conditions and asked a series of questions.
1 Smith and Howell (2009) used digital projection technology not used by the other studies to present the
series of VAQ conditions. Some of the participants in the Smith and Howell study were shown images using a LCD
projector connected to a laptop computer. In other sessions, participants in the Smith and Howell study were shown
images on a computer monitor connected to the computer.
2-2
-------
The group interview sessions were conducted multiple times with different participants.
Ideally the participants will be a representative sample of the residents of the metropolitan area.
While all studies agree that this is the preferred approach, due to the high cost of organizing and
conducting a series of in-person group interviews with a large, statistically representative sample,
only the Phoenix study was able to fully meet this objective. During the group interview
sessions, the participants were instructed to consider whether the VAQ in each photograph or
image would meet an urban visibility standard, according to their own preferences and
considering three factors:
1. The standard would be for their own urban area, not a pristine national park area
where the standards might be stricter.
2. The level of an urban visibility standard violation should be set at a VAQ level
considered to be unreasonable, objectionable, and unacceptable visually.
3. Judgments of standards violations should be based on visibility only, not on
health effects.
The photographs (images) were not shown in order of ascending or descending VAQ
conditions; the VAQ conditions were shown in a randomized order (with the same order used in
each group interview session). In order to check on the consistency of each individual's
answers, the full set of photographs (images) shown during the group interview included
duplicates with the identical VAQ conditions.
The participants were initially given a set of "warm up" exercises to familiarize them with
how the scene in the photograph or image appears under different VAQ conditions. The
participants next were shown 25 randomly ordered photographs (images), and asked to rate each
one based on a scale of 1 (poor) to 7 (excellent). They were then shown the same photographs or
images again (in the same order), and asked to judge whether each of the photographs (images)
would violate what they would consider to be an appropriate urban visibility standard (i.e.,
whether the level of impairment was "acceptable" or "unacceptable"). While the studies all
asked the basic question, "What level of visibility degradation is acceptable?", the term
"acceptable" was not defined, so that each person's response was based on his/her own values
and preferences for VAQ.
2.2 DENVER, COLORADO
The Denver urban visibility preference study (Ely et al., 1991) was conducted on behalf
of the Colorado Department of Public Health and Environment (CDPHE). The study consisted
of a series of focus group sessions conducted in 1989 with participants from 16 civic
2-3
-------
associations, community groups, and employees of state and local government organizations.2
The participants were not selected to be a fully representative sample of the Denver metropolitan
population but were instead selected to take advantage of previously scheduled meetings.
During the 16 focus group sessions, a total of 214 individuals were asked to rate
photographs of varying visibility conditions in Denver. The photographs were taken November
1987 through January 1988 by a camera in Thornton, Colorado. Thornton is suburb of Denver,
located approximately six miles north of downtown Denver. The photographs were taken as part
of a CDPHE study of Denver's air quality. The scene in the photographs was toward the south
from Thornton and included a broad view of downtown Denver and the mountains to the south.
Each group was shown one of two sets of 20 randomly ordered unique photographs (13 of the
sessions included 5 duplicate slides, for a total of 25 photographs, to evaluate consistency of
responses). The two sets of different slides were used to investigate whether the responses
between the two sets of photographs were different (no differences were found). Approximately
100 participants viewed each photograph. Projected color slides were used to present the
photographs to focus group participants, and were projected on a large screen
The VAQ conditions in each Denver photograph were recorded when the photograph was
taken and measured by a transmissometer yielding hourly average light extinction, bext- The
transmissometer was located in downtown Denver, approximately eight miles from the camera
and in the middle of the camera's view path. Ely et al. (1991) provide the time of day and
measured extinction level for each photograph. The extinction levels presented in the Denver
photographs ranged from 30 to 596 Mm"1. This corresponds to 1 Idv to 41dv, approximating the
10th to 90th percentile of wintertime visibility conditions in Denver in the late 1980s.
The participants first rated the VAQ in each photograph on a 1 to 7 scale, and
subsequently were asked if each photograph would violate an urban visibility standard. The
individual's rating on the 1 to 7 scale and whether the photograph violated a visibility standard
were highly correlated (Pearson correlation coefficient greater than 80%).
The percent of participants who found a photograph acceptable to them (i.e., would meet
an appropriate urban visibility standard) was calculated for each photograph. Figure 2-1 shows
the results of the Denver participants' responses, with VAQ measured in deciviews.
! No preference data were collected at a 17 focus group session due to a slide projector malfunction.
2-4
-------
Figure 2-1. Percent of Denver Participants Who Considered VAQ in Each
Photograph "Acceptable"
100%
O)
=0,
II
n
ra =
Q.
50%
0%
/ 4» »
»* . * t **•
10
15
20
25 30
Deciview
35
40
45
• % acceptable Denver standard
Ely et al. (1991) introduce a "50% acceptability" criteria analysis of the Denver
preference study results. The 50% acceptability criteria is designed to identify the VAQ level
that best divides the photographs into two groups: those with a VAQ rated as acceptable by the
majority of the participants, and those rated not acceptable by the majority of participants. While
no single VAQ level creates a perfect separation between the two groups, the CDPHE identified
a VAQ of 20.3 dv as the point that best separates the Denver study responses into "acceptable"
and "not acceptable" groups. Based in part on the findings of the Denver visibility preference
study, the CDPHE established a Denver visibility standard at bext = 76 Mm"1 (dv = 20.3).
Using 20.3 dv as the 50% acceptability criteria led to six photographs being
inconsistently rated by the majority of the viewers. A photograph was inconsistently rated for
two possible reasons; either the photograph's VAQ was at least 1 dv better than the Denver
standard (i.e., dv < 19.3) but was judged to be "unacceptable" by a majority of the participants
rating that photograph, or the VAQ was at least 1 dv worse than the standard (> 21.3 dv) but
found to be acceptable by the majority of the participants. This definition of inconsistent rating
helps evaluate the robustness of the study results to support the selection of the Denver urban
visibility standard at 76 Mm"1 (20.3 dv) by identifying photographs with VAQ a minimum of 1
dv above or below the standard and ignoring "near misses" involving photographs within 1 dv of
the standard. A change of 1 or 2 dv in uniform haze under many viewing conditions will be seen
2-5
-------
as a small but noticeable change in the appearance of a scene, regardless of the initial haze
condition (U.S. EPA, 2004).
Table 2-1 presents information about the six photographs that were inconsistently rated.
All six of the inconsistently rated photographs were taken at 9:00 a.m. The five inconsistently
rated photographs with a VAQ better than the Denver standard have a VAQ at least 2 dv below
the standard. The VAQ in the only inconsistently rated photograph with air quality worse than
the standard (Photograph #6) is 1.1 dv above the standard. The study used 18 photographs from
9:00 a.m., so a third of the 9:00 a.m. photographs were inconsistently rated. Conversely, none of
the 32 photographs taken at noon or 3:00 p.m. were inconsistently rated.
Table 2-1. VAQ of Denver Photos Substantively Misclassified by Majority of Participants
Photograph #
1444
1854
1954
2055
2460
3685
VAQ in photograph
in extinction
(Mm1)
VAQ in
photograph (dv)
13.8
16.9
16.9
17.0
17.9
21.4
% of participants
who rated the photo
"acceptable"
43%
43%
31%
42%
13%
72%
Time of day of
photograph
9:00 a.m.
9:00 a.m.
9:00 a.m.
9:00 a.m.
9:00 a.m.
9:00 a.m.
Figure 2-2 shows the same data results about percent of participants who rated each
photograph acceptable as in Figure 2-1, but with the time of day of each photograph indicated by
different colors. The time of day colors clearly indicate how inconsistently participants rated
some of the 9:00 a.m. photographs.
Eliminating the 9:00 a.m. photographs creates a "hole" in the range of remaining
photographs; there are no photographs with a VAQ between 17.7 dv and 20.3 dv. As seen in
Figure 2-3, this is a critical range in evaluating the responses. All of the photographs with a
VAQ equal to or better (i.e., a lower dv value) than 17.7 dv are rated acceptable by the majority
of the participants, and all photographs with a VAQ at or above 20.3 dv are rated not acceptable.
After eliminating the 9:00 a.m. photographs, any VAQ level between 17.7dv and 20.3 dv would
completely divide the photographs into two groups with no inconsistent ratings.
2-6
-------
Figure 2-2. Photograph Time of Day Information
100%
O)
c
(0 :
TO
Q.
I §
(0 =
Q.
0%
10
15
20
25 30
Deciview
35
40
45
• 9:00 AM • 12:00 PM 3:00 PM—Denver standard
A modestly broader range of VAQ conditions provides an even more unambiguous
interpretation of the Denver study results. Every photograph with a VAQ of 17.7 dv or lower
was rated acceptable by 89% or more of participants, and every photograph with a VAQ of 24.6
or higher was rated not acceptable by 84% or more of the participants. The 17.7 dv to 24.6 dv
range separating the results is shown in Figure 2-3, which also eliminates the 9:00 a.m. results.
2-7
-------
Figure 2-3. Denver Photograph Time of Day Results (9:00 a.m. Photographs
Eliminated), with the Broader Range (17.7 dv and 24.6 dv) of the
50% Acceptability Criteria Shown
IULT/0
)
3
$
0)
• o
u
<
0%
• * V * ;
9
•
I
0
%
^
.
1 % •• •___.
i i i i i i
10 15 20 25 30 35 40 45
Deciview
• 12:00 PM 3:00
d
2.3 VANCOUVER, BRITISH COLUMBIA, CANADA
The BC urban visibility preference study (Pryor, 1996) was conducted on behalf of the
BC Ministry of Environment following the methods used in the Denver study. Participants were
students at the University of British Columbia, who were in one of four focus group sessions
with between 7 and 95 participants. A total of 180 participants completed the surveys (29 did
not complete the survey).
The BC study used photographs (projected as slides) depicting various VAQ conditions
in two cities (Chilliwack and Abbotsford) in the lower Fraser River valley in southwestern BC.
Abbotsford is located approximately 75 miles east of Vancouver, BC, and had a 2006 population
of 159,000 (Statistics Canada, 2009a). Abbotsford has a diverse and successful economy, with
approximately 25% of the labor force working in the Vancouver metropolitan area. Chilliwack
is adjacent to Abbotsford to the east. Both cities have experienced rapid population growth,
growing faster than the Vancouver metropolitan area, and are considered suburbs (or exurbs) of
Vancouver.
The survey was conducted at the University of British Columbia (UBC) in 1994. The
participants were 206 undergraduate and graduate students enrolled in classes in UBC's
2-8
-------
Department of Geography. Information about student demographics and where they lived prior
to enrolling at UBC (which potentially influences their knowledge of, and preferences for,
Vancouver area visibility) is not available.
The BC survey showed 20 unique photographs to the participants in random order. Ten
photographs were from Chilliwack, and 10 were from Abbotsford. The Chilliwack photographs
were taken at the Chilliwack Hospital, and the scene includes a complex foreground with
downtown buildings, with mountains in the background up to 40 miles away. Figure 2-4 is a
composite of two of the Chilliwack photographs used in the preference study, showing the scene
with a good visibility day (14.1 dv) in the middle and a significantly impaired day (34 dv) around
the border (Jacques Whitford AXYS, 2007). The Abbotsford photographs were taken at the
Abbotsford Airport. The Abbotsford scene includes fewer man-made objects in the foreground
and is primarily a more rural scene with the mountains in the background up to 36 miles away.
Figure 2-4. Composite Chilliwack, BC Photograph Shows VAQ of 14.1 dv and 34 dv
The photographs were taken in July and August 1993 as part of a VAQ and fine
parti culate monitoring project sponsored by the BC Ministry of Environment, Lands and Parks
(REVEAL, the Regional Visibility Experimental Assessment in the Lower Fraser Valley). All of
the photographs were taken at either 12:00 p.m. or 3:00 p.m. VAQ data were available for each
photograph from visibility monitors near the location of each camera. The types of VAQ
measurement data available from the two locations were not identical. The Chilliwack location
2-9
-------
measurement data available from the two locations were not identical. The Chilliwack location
used both an open-chamber nephelometer and a long path transmissometer and collected hourly
average data on both aerosol light scattering (bsp) and total extinction (bext)., respectively. The
visibility monitoring at the Abbotsford location had only a nephelometer and collected only bsp
data.
As explained in section 1.3, total light extinction is the sum of scattering by gases (bsg)
and particles (bsp) plus light absorption by gases (bag) and particles (bap). In order to present the
preference results from the BC study in comparable terms, bext for the Abbotsford photographs is
estimated by assuming that the average of the ratios of PM light extinction (i.e., bap+bsp) to PM
light scattering (bsp) for all ten of the Chilliwack photographs can be multiplied by the
Abbotsford nephelometer determined bsp values corresponding to each of its photographs to
estimate its PM light extinction value. By assuming that absorption by gases (bag) is zero, total
light extinction is equal to the PM light extinction (i.e., bap+ bsp) plus particle scattering by gases
(i.e., bsg that is approximately equal to 10 Mm"1). Table 2-2 presents the data from the
photographs used in the BC study, including the estimated b^ for the Abbotsford photographs.
There are two caveats to be noted about the extinction data for the photographs reported
in Pry or, 1996. First, in Table 2 of the original article, two of the Abbotsford photographs are
listed with the same date and time (12:00 p.m., 7/26/1993). There is no information provided for
a 3:00 p.m., 7/26/1993 Abbotsford photograph, although there is a Chilliwack photograph from
that time. The preference and VAQ data are presumed to be correct for both photographs and
one of the two identical date/time labels is assumed to be a typographic error. The second caveat
is that b^ levels from the same date and time can differ substantially between Abbotsford and
Chilliwack, and the relative levels can change rapidly, even though the two cities are only 25
miles apart. For example, at 12:00 p.m. on 8/19/1993, the bsp level in Chilliwack was about one-
third of the Abbotsford bsp level. By 3:00 p.m. the situation was reversed, with the Chilliwack
bsp level 50% higher than Abbotsford. In those three hours the Chilliwack bsp level had more
than doubled (from 46 Mm"1 to 105 Mm"1), and the Abbotsford level had fallen by over half
(from 145 Mm"1 to 67 Mm"1). Such substantial changes in measured bsp levels occurring across a
relatively short period of time and short distance, may reflect an inherent uncertainty introduced
by using a single measure of light extinction from a portion of visual scene (where the
nephelometer or transmissometer was operating) to assess visibility conditions throughout an
actual photographs of a complex scene. Spatial and temporal non-uniformity of visibility
conditions within a scene are an atmospheric condition known to occur on some days, and may
contribute to the variability in participant responses in preference studies utilizing actual
photographs.
2-10
-------
Table 2-2. Summary of Photographs Used in British Columbia Study
Date
Time
bsp
bext
Ratio
(bexrbsg)/bsp
Estimated
bext
Deciview
Chilliwack
7/26/93 12
7/26/93 3:(
7/27/93 12
7/27/93 3:(
8/2/93 12:
8/2/933:0
8/5/93 12:
8/5/933:0
8/19/93 12
8/19/93 3:(
:00 p.m.
)0 p.m.
:00 p.m.
)0 p.m.
00 p.m.
0 p.m.
00 p.m.
0 p.m.
:00 p.m.
)0 p.m.
Average
86
67
63
119
18
20
45
51
46
105
62
128
112
105
185
37
36
70
96
81
170
102
1.372
1.522
1.508
1.471
1.5
1.3
1.333
1.686
1.543
1.524
1.476
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
25.49
24.16
23.51
29.18
13.08
12.81
19.46
22.62
20.92
28.33
21.96
Abbotsford
7/26/93 12
7/26/93 12
7/27/93 12
7/27/93 3:(
8/2/93 12:
8/2/933:0
8/5/93 12:
8/5/933:0
8/19/93 12
8/19/93 3:(
:00 p.m.
:00 p.m.
:00 p.m.
)0 p.m.
00 p.m.
0 p.m.
00 p.m.
0 p.m.
:00 p.m.
)0 p.m.
Average
39
82
104
132
24
25
62
75
67 NA
145
76
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
68
131
205
164
45
47
121
102
224
109
122
19.17
25.73
30.20
27.97
15.04
15.48
24.93
23.22
31.09
23.89
23.67
Figure 2-5 presents the results of the BC study. The division corresponding to the
Denver "50% acceptable" criteria occurs between 22.6 dv and 23.2 dv. All of the photographs
with a VAQ better than 22.6 dv were rated acceptable by the majority of the participants with
one exception (47% of the participants judged the 19.2 dv photograph to be acceptable). All
photographs with a VAQ better than 19.2 dv were rated acceptable by over 90% of the
2-11
-------
participants. All photographs with a VAQ worse than 22.6 dv were rated not acceptable by the
majority of the participants, and all photographs with a VAQ worse than 28.3 dv were rated not
acceptable by over 90% of the participants.
Figure 2-5 also suggests that there may be some difference between the preferences
expressed for the Chilliwack scene and those for the Abbotsford scene. All photographs were
rated by the same individuals (students at UBC), but the summary of the responses indicate that
the participants may have rated as acceptable a worse level of impaired VAQ impairment (e.g.,
higher dv levels) in photographs showing more of a downtown area (Chilliwack) than in less
congested scenes (Abbotsford). The strongest evidence for this hypothesis, however, is the
preference for a single photograph (the 19.0 dv photograph from Abbotsford, rated as acceptable
by 47%), previously identified as an outlier observation.
Figure 2-5. Percent of BC Participants Who Consider VAQ in Each Photograph
"Acceptable"
i 100%
a
«
o
•
w 50%
ra
a.
0%
Results of British Columbia Visibility Study
* *
'".
10
15
20
25 30
Deciview
35
40
45
» Chilliwack • Abbotsford
The BC Ministry of the Environment is considering the BC urban visibility preference
study as part of establishing urban and wilderness visibility goals in BC.
2-12
-------
2.4 PHOENIX, ARIZONA
The Phoenix urban visibility preference study (BBC Research & Consulting, 2003),
which was conducted on behalf of the Arizona Department of Environmental Quality, used
group interviews based on the methods used in the Denver study, with two major exceptions: (1)
the focus group participants were selected as a representative sample of the Phoenix area
population, and (2) the pictures presented in the focus groups were computer-generated images
to depict specific uniform haze conditions.
The Phoenix study included 385 participants in 27 separate focus group sessions.
Participants were recruited using random digit dialing to obtain a sample group designed to be
demographically representative of the larger Phoenix population. During July 2002, group
interview sessions took place at six neighborhood locations throughout the metropolitan area to
improve the participation rate. Participants received $50 as an inducement to participate.
Three sessions were held in Spanish in one region of the city with a large Hispanic
population (25%), although the final overall participation of native Spanish speakers (18%) in
the study was below the targeted level. The age distribution of the participants corresponded
reasonably well to the overall age distribution in the 2000 U.S. Census for the Phoenix area
(BBC Research & Consulting, 2003). Participants slightly over-represented the middle-income
range ($50,000 to $74,999), compared with 2000 Census data, and slightly under-represented
very low-income ranges (under $24,999). The distribution of participant education levels was
fairly consistent with the education distribution in the 2000 Census.
Photographic-quality slides of the images were developed using the WinHaze software
(Molenar et al., 1994). The scene used in the Phoenix study images was taken at a water
treatment plant. The view is toward the southwest, including downtown Phoenix, with the Sierra
Estrella Mountains in the background at a distance of 25 miles. Figure 2-6 shows the image with
thebestVAQ(15dv).
2-13
-------
Figure 2-6 Reproduction of Image with the Best VAQ (15 dv) Used in the Phoenix Study
The study used a total of 21 unique WinHaze images. Four of the 21 unique images were
randomly selected and used twice to evaluate consistency so that participants viewed a total of
25 images. The 25 images were randomly ordered, with all participants viewing the images in
the same order. The WinHaze images used in the Phoenix study do not include layered haze, a
frequent and widely recognized form of visibility impairment in the Phoenix area.
The VAQ levels in the 21 unique images ranged from 15 dv to 35 dv (the extinction
coefficient bext ranged from 45 Mm"1 to 330 Mm"1). As in the Denver study, participants first
individually rated the randomly shown slides on the same VAQ scale of 1 to 7. Participants were
instructed to rate the photographs solely on visibility and to not base their decisions on either
health concerns or what it would cost to have better visibility. Next, the participants individually
rated the randomly ordered slides as "acceptable" or "not acceptable," defined as whether the
visibility in the slide is unreasonable or objectionable.
Figure 2-7 presents the percent acceptability results from the Phoenix study. The
combination of the use of WinHaze images and the larger number of participants than in the
Denver study may account for the "smoother" backwards S-shaped pattern of preferences.
2-14
-------
Figure 2-7. Percent of Phoenix Participants who Consider VAQ in Each Image
"Acceptable"
100%
10
15
20
25 30
Deciview
35
40
45
Ninety percent or more of the participants rated a VAQ of 20 dv or better as acceptable, and 70%
rated a VAQ of 22 dv or better as acceptable. The "50% acceptable criteria" was met at
approximately 24.3 dv (with 51.3% of the participants rating that image as acceptable). The
percent acceptability declines rapidly as VAQ worsens; only 27% of the participants rated a
26 DV image as acceptable, and fewer than 10% rated a 29 dv image as acceptable.
The Phoenix urban visibility study formed the basis of the decision of the Phoenix
Visibility Index Oversight Committee for a visibility index for the Phoenix metropolitan area
(Arizona Department of Environmental Quality, 2003). The Phoenix Visibility Index establishes
an indexed system with 5 categories of visibility conditions, ranging from "Excellent" (14 dv or
less, which was a better VAQ than any of the images used in the Phoenix study) to "Very Poor"
(29 dv or greater, which less than 10% of the study participants rated as acceptable). The
"Good" range is 15 dv to 20 dv (more than 90% of the participants rated images in this VAQ
range as acceptable). The environmental goal of the Phoenix urban visibility program is to
achieve continued progress through 2018 by moving the number of days in poorer quality
categories into better quality categories.
2.5 WASHINGTON, DC
One of the Washington, DC urban visibility pilot studies was conducted on behalf of
EPA (Abt Associates Inc., 2001). It was designed to be a pilot focus group study, an initial
2-15
-------
developmental trial run of a larger study. The intent of the pilot study was to refine both focus
group method design and potential survey questions. Due to funding limitations, only a single
focus group session took place, consisting of one extended session with nine participants. No
further urban visibility focus group sessions were held in Washington, DC on behalf of EPA.
In March 2009, Dr. Anne Smith conducted a separate study of Washington urban
visibility, using the same photographs and similar approach as the 2001 study (Smith and
Howell, 2009). On behalf of the Utility Air Regulatory Group, Dr. Smith presented comments
(Smith, 2009) to the CAS AC at a public meeting held on April 2, 2009 to review EPA's plan
(US EPA, 2009b) for conducting further urban visibility studies in support of PM NAAQS
reviews. Dr. Smith submitted the Smith and Howell (2009) report to the CAS AC as part of the
public comment process. The Smith and Howell study conducted three study variations of a
Washington, DC preference study, including one experiment involving 26 participants designed
to replicate the EPA pilot study (Abt Associates Inc., 2001). Both the Abt Associates Inc.
(2001) study results and the results of the Smith and Howell (2009) study are discussed below.
2.5.1 Washington, DC 2001
The EPA's Washington, DC study (Abt Associates Inc., 2001) adopted the general study
methods used in the Denver, BC, and Phoenix studies, modifying them appropriately to be
applicable in an eastern urban setting. Washington's (and the entire East's) current visibility
conditions are typically substantially worse than western cities and have different characteristics.
Washington's visibility impairment is primarily a uniform whitish haze dominated by sulfates,
and the relative humidity levels are higher compared with the western study areas. In addition,
the relatively low-lying terrain3 in Washington, DC provides substantially shorter maximum
sight distances.
The Washington, DC focus group session included questions on valuation, as well as on
preferences. The focus group content dealing with preferences for an urban visibility standard
was similar to the focus group sessions in the western studies.
A single scene of a panoramic photograph taken from Arlington National Cemetery in
Virginia was used, and included an iconic view of the Potomac River, the National Mall, and
downtown Washington, DC. All of the distinct buildings in the scene are less than four miles
from the camera, and the higher elevations in the background are less than 10 miles from the
camera. Figure 2-8 presents the photograph with the best VAQ used in the study.
3The maximum elevation in Washington, DC is 409 feet.
2-16
-------
Figure 2-8. Reproduction of the Image with the Best VAQ (8.8 dv) Used in the
Washington, DC Study
The Washington, DC study used 20 unique images generated by WinHaze, each prepared
from the same original photograph. Humidity and gaseous light scattering was held constant in
preparing the WinHaze images, as was the relative chemical mix of aerosol particulates in the
photos (i.e., only the aerosol concentrations were increased to create the images with worse
VAQ). Five of the images were repeated as a consistency check, so participants viewed a total
of 25 slides. The VAQ in the images ranged from 8.8 to 38.3 dv Figure 2-9 presents the percent
acceptability results from the 2001 Washington study. Because only nine participants were
involved in the study, the possible values of "percent acceptable" are limited to multiples of 1/9.
Figure 2-9 also shows an anomalous result involving one of the five repeated images. Three of
the repeat images had the same ranking each time they were presented (i.e., all nine participants
rated them acceptable or not acceptable both times they rated that slide). One of the images (the
image with 8.8 dv, the best VAQ image used in the study) was rated acceptable by all nine
participants the first time it was used, but the repeat of that slide was rated not acceptable by one
participant. Another image, however, had a substantially different result. The 30.9 dv image
was rated acceptable by five of the nine participants the first time it was presented, but the repeat
of the slide was only rated acceptable by one of the nine participants. The responses for all five
pairs of repeated images are shown in red on Figure 2-9, including the images which were
identically rated both times they were presented.
2-17
-------
Figure 2-9. Percent of 2001 Washington Participants Who Considered VAQ Acceptable
in Each Image
Re
100%
0)
c
•j=
(0 =
{£ 0)
^ 1
n a. 50%
0. 0>
•- 0
11
(0 •
a.
ss
0%
c
jsults of 2001 Washington DC Preference Study
) 10
* •
* * *
15 20 25 30 35
Deciview
* Unique Images • 5 Repeated Images
40 45
In the 2001 Washington, DC study, all images with a VAQ below 25.9 dv were rated
acceptable by the majority of the participants, and all images with a VAQ below 29.2 dv were
rated acceptable by at least four of the nine (44%) participants. All images with a VAQ above
30.9 dv were rated not acceptable. The "50% acceptability criteria" division occurs in the range
of 25.9 dv to 30.9 dv, with the anomalous result of the inconsistent responses to the repeated
image with 30.9 dv effectively broadening this range and adding uncertainty to identifying a
clear division.
2.5.2 Washington, DC 2009
The Smith and Howell (2009) study conducted additional focus group sessions based on
the methods and materials used in the 2001 Washington, DC study. Smith and Howell recreated
the WinHaze images used in the 2001 Washington, DC urban visibility preference study, using
the description in the report on the 2001 study (Abt Associates Inc., 2001), and created images
using currently available desktop computer version of WinHaze (Version 2.9.0). Smith and
Howell used a shortened version of the same question protocol as the 2001 study. The WinHaze
images were presented to a total of 64 participants who were all employees of Charles River
Associates (CRA International, Inc). (Smith and Howell also are CRA International employees).
The CRA employees were based at the firm's Washington, DC and Houston, Texas offices (44
2-18
-------
and 20 participants, respectively). The Houston participants were included to explore whether
familiarity with Washington, DC VAQ conditions developed from currently living in the
Washington region noticeably influenced the responses. As noted by Smith and Howell, the
participants were not a representative sample of either metropolitan area's population; all
participants were employed, and the participant group included a higher proportion of college
educated individuals and higher household incomes than the general population.
Eight of the Washington-based participants and all of the Houston participants viewed the
WinHaze images on a desktop computer monitor. The remaining Washington participants
viewed the images projected on a screen.
The stated purpose of the Smith and Howell study was to explore the robustness of the
2001 results. To investigate this issue, Smith and Howell conducted three different tests
concerning urban visibility preferences. Each participant was involved with only one test. The
three tests were:
* Test 1 - replicated the Abt Associates Inc. (2001) study
* Test 2 - reduced the upper end of the range of VAQ by eliminating the 11 images
used in Test 1 with a VAQ above 27.1 dv
* Test 3 - increased the upper end of the range of VAQ by including two new images
of worse VAQ; the two new images had a VAQ of 42 dv and 45 dv
Sixteen employees from the Washington, DC office and 10 participants from the Houston
office took Test 1 (a total of 26 participants). All the participants viewed the same unique 20
Washington, DC WinHaze images as the 2001 study (plus repeated images for a total of 25
images shown to participants). Images were presented in the same random order as in the 2001
study. Figure 2-10 presents the results of Test 1. The results for the 16 Washington participants
are indicated in blue and results for the 10 Houston participants in red. Although all images used
in the study were of Washington, DC, the results suggest that there is not a significant difference
in the preferences of participants based in the two offices. The scene in the images is an
immediately recognizable iconic view of the National Mall and downtown Washington, DC,
which may influence the similarity of responses by residents of the two cities.
2-19
-------
Figure 2-10. Percent of 2009 Test 1 Study Participants Who Considered VAQ
Acceptable in Each Image, Showing the Range of the Lower and Upper
Bound of 50% Acceptability Criteria
100%
60%
0%
A A
\
10
15 20 25
DttcMaw
30
35
40
• Washington
A
Using the combined Test 1 results from the two CRA offices (26 total participants), the
majority of participants in the 2009 study rated all VAQ images with 25.9 dv or less as
acceptable and all VAQ images with 29.2 dv or greater as not acceptable. The image of 27.1 dv
was rated as acceptable by 50% of the total participants (56% of the Washington-based and 40%
of the Houston-based participants). All images with a VAQ less than 22.9 dv were rated
acceptable by at least 90% of the participants, and all images with a VAQ greater than 32.3 dv
were rated not acceptable by 88% of the participants.
Figure 2-11 presents the Abt Associates Inc., 2001 study and Smith and Howell 2009
(Test 1) study results on a single graph, representing the results of 35 total participants of
preferences for urban visibility in Washington, DC. The results from the 2009 study on Figure
2-11 combine the Test 1 responses from the two CRA offices. Figure 2-11 also shows the 50%
acceptability criteria range (22.9 dv to 32.3 dv) from the 2009 Test 1. In comparison, the 2001
study 50% acceptability range was 25.9 dv to 30.9 dv. Inspection of the points in Figure 2-11
indicates that the results from the 2009 study (Test 1) are not appreciably different than the
results of the 2001 Washington study. This observation of similar results is confirmed by a logit
regression analysis of the 2001 and 2009 Test 1 data that includes estimates of the 50% criteria
2-20
-------
deciview values with confidence intervals and hypothesis testing of the similarity of the values,
as described in Appendix J (see Tables 6, 7 & 8 and Figure 3).
Figure 2-11. Combined Results of Washington, DC 2001 and 2009 Test 1 (showing 50%
Acceptability Criteria from 2009, Test 1)
.a
as
+•»
a.
.^i(
™ ™
10 15 20 25 30 35 40
Deciview
2001 Study
2009, Test 1 ----- Lower Bound ---- Upper Bound
In Test 2, Smith and Howell reduced the range of VAQ images to images with a VAQ of
27.1 dv or less. The 26 participants in the Test 2 study were different people than the Test 1
participants. Test 2 presented only the nine unique clearest WinHaze images from the full Test 1
set of 20 images, along with 3 duplicates for a total of 12 images. This constricted the VAQ
levels presented to the range that the majority of participants in the 2001 study rated as
acceptable and reduced the upper end of the VAQ range by 11.2 dv.
Figure 2-12 presents the Test 1 and Test 2 results. Test 2 found a substantial shift in the
responses regarding which VAQ levels are considered acceptable. The smaller number of
images used in Test 2 made identifying the range of the 50% acceptability criteria more difficult
than in Test 1. The lower bound of the range occurs between 15.6 and 18.7 dv, and the upper
bound occurs between 24.5 and 27.1 dv. Smith and Howell conclude that the shift in the
acceptability responses between Test 1 and Test 2 suggests that the VAQ levels identified as
acceptable in an urban visibility preference study conducted using the general approach
previously used in the all the studies may be influenced by the range of VAQ images presented.
2-21
-------
Figure 2-12. Comparison of Results from Test 1 and Test 2 (Smith and Howell, 2009)
Results of Smith and Howell Test 1 and Test 2
100%
£*
&
Q.
0)
(0
Q.
(0
Q.
10
15
20 25
Deciview
30
35
40
Test 2 »Test1
In order for the range of images shown to be able to influence the acceptability ratings,
participants would need to be aware of the upper and lower bounds of the range prior to the
judging acceptability. However, since they were shown images randomly with respect to the
VAQ levels, asked to rate each one before going to the next image, and were not given a chance
to revise their acceptability ratings, this was not possible during the acceptability exercise itself.
The only other opportunity participants could have to learn the VAQ range is during the VAQ
rating exercise done just prior to the acceptability rating. However, in the VAQ rating exercise
where the participants were asked to rate the quality of visibility for the shown images on a scale
from 1 to 7, the images were also shown in a random order, participants were not aware how
many photographs would be shown or the range of conditions, they were asked to rate each one
using a value from 1 to 7 before going on to the next image and they did not have the opportunity
to revise the ratings of earlier viewed images.
Figure 2-13 shows the average visibility rating on the 1 to 7 scale for each image used in
each of the three tests conducted by Smith and Howell (2009). The consistency observed in the
relationship between VAQ deciview levels and the average scores assigned across the three tests
demonstrates that the participants come to the survey with the capability to consistently rate the
haze levels shown in the images, regardless of the breadth of the range used or the order or
2-22
-------
number of slides shown, and that they are aware of a full range of conditions, even when, as was
the case in Test 2, they were not shown the worst haze images.
Figure 2-13. Average Visibility Ratings for the Washington, DC WinHaze Images
by Participants in Tests 1-3 Conducted by Smith and Howell (2009).
Test 1 {sample of 26)
6
x v ——Test 3 (sample of 12)
I 4
Test 2 (sample of 26)
8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44
Dectvrew
Why then did Test 2 participants in the subsequent part of the survey rate images of haze
levels as unacceptable that were rated acceptable by participants in the other tests and the earlier
Washington, DC pilot study? In a three sentence script4 that constituted the only instructions
read prior to the acceptability rating, the participants were told that they would see the same set
of slides that they had just rated (i.e., on the 1 to 7 scale), and they were asked to rate them
according to whether the VAQ depicted were acceptable or unacceptable to them. Apparently by
directing then to rate the same images for acceptability, the participants understood that their
choices of visibility conditions were restricted to a range of conditions shown in the 1 to 7
ratings that they had just completed. For participants in Test 2 this would mean that by their own
1 to 7 ratings the range was restricted to include no poor visibility conditions (i.e. only scenes
rated from 3 to 7).
Smith and Howell (2009) concluded that the effects of a changed range on the
acceptability ratings results demonstrates that VAQ preference studies results are not robust and
4 The complete script for the acceptability/unacceptability part of the study is as follows. "Now you will be
shown the same set of slides that you just rated. Again each image will illustrate the effects of a different level of
visibility. This time, rate the slides according to whether the visibility is acceptable or unacceptable to you. "
2-23
-------
do not reflect an enduring view on the "unacceptability" of different levels of VAQ degradation.
However, there is an alternative explanation. It seems more likely that the use of such a severely
truncated range of VAQ conditions in Test 2, which did not include any of the images of VAQ
that previous studies identified as unacceptable, in effect fundamentally changed the implied
instructions for the participants. Instead of conveying that they were to identify VAQ levels that
they found acceptable among a full range of VAQ conditions from very poor to very good, the
implied message was that they should identify the VAQ levels that they found acceptable among
a curtailed range of VAQ conditions that only included average to very good VAQ. By this
reasoning, it would be inappropriate to include Test 2 results with those of the other tests as a
measure of VAQ preference for Washington, DC.
In Test 3, Smith and Howell expanded the VAQ range of WinHaze images shown to the
participants, including two new images with a worse VAQ. The new images had a VAQ of 42
dv and 45 dv, raising the upper end of the VAQ range by 6.7 dv. Test 3 also reduced the total
number of images shown to participants to 19 images by eliminating the use of the five repeat
images in Test 1, and also eliminated three additional images in order to reduce the participants'
time burden. The three deleted images had a VAQ of 11.1, 15.6, and 24.5 dv. The best VAQ
image shown to Test 3 participants was 8.8 dv (same as the best VAQ image in Tests 1 and 2).
However, in Test 3 there were no images with VAQ between 8.8 dv and 18.7 dv, creating a
significant "hole" in the distribution of VAQ conditions presented to the Test 3 participants.
Test 3 was conducted with 12 participants from the CRA Washington office (none of whom
participated in Test 1 or Test 2). No Houston participants were involved with Test 3. Figure 2-
13 shows that the Test 3 average ratings from 1 to 7 during the VAQ rating exercise increased
the average participant rating by about 1 at the low end of the scale (very poor VAQ). The
results of Test 3 are shown in Figure 2-14, along with the results of Test 1.
Test 3 resulted in an overall increase in the percent of respondents rating as acceptable
the VAQ images used in both tests. In Test 3 all images with a VAQ below 22.9 dv were rated
acceptable by 100% of the participants (similar to the Test 1 results), implying there was no
general change in the acceptability of the images with good VAQ. However, for all VAQ
images (that were used in both studies) between 25.9 dv and 33.6 dv, a noticeably larger
percentage of the participants in Test 3 rated the image as acceptable than in Test 1. At VAQ
levels worse than 33.6 dv, the majority of the participants found the VAQ level not acceptable in
both tests.
2-24
-------
Figure 2-14. Comparison of Results from the Smith and Howell (2009) Test 1 and Test 3
100%
50% -
8.
o%
10
15
20
25
Deciview
30
35
40
45
I Test 3 » Test 1
Given that most of the same images of VAQ conditions were used in all of the tests,
composite acceptability ratings (i.e., from the original pilot study (2001) and from Tests 1 and 3)
of each image were initially developed to evaluate whether increasing the number of participant
ratings for each image influenced the 50% acceptability value. Figure 2-15 shows composite
results from the combination of these three groups (total of 47 participants). The 50%
acceptability criteria value for this composite dataset lies unambiguously between the 30.1 dv (at
51.1%) and the 30.9 dv points (at 46.3%),
2-25
-------
Figure 2-15. Composite Results from Smith and Howell (2009) Tests 1 and 3, and Abt
(2001) Washington, DC Pilot Study
100%
50%
o%
Washington DC Studies Combined
2001 and Smith Tests 1 & 3
10
20
30
40
50
Deciview
To determine whether it would be appropriate to combine all of the results from the
Washington, DC studies, in order to increase the number of data points for Washington, DC, we
considered several factors. First, while the range limitations identified with Test 2 (i.e. an overly
restrictive range) that resulted in its results being eliminated from consideration in the selection
of appropriate CPLs do not apply to Test 3 results due to its somewhat more complete coverage
of the 1 to 7 rating range in the VAQ rating exercise, the number of participants in Test 3 (i.e.,
12) is small enough that the statistical uncertainty of the results may be an issue if used alone.
Second, it was not clear whether the significant "hole" in the Test 3 VAQ distribution between
8.8 dv and 18.7 dv potentially had an effect on participant acceptability responses. Finally, the
logit regression analysis which was applied to each of the individual Smith and Howell (2009)
tests as well as subsets of some of these tests to investigate differences in the preference curves
and 50% criteria deciview levels (Appendix J) concluded in part that Test 2 and Test 3 response
curves and 50% criteria values are statistically dissimilar from those of the 2001 and Test 1
Washington, DC studies. In contrast, the logit analysis concluded that the 2001 and Test 1
results have 50% criteria values that are statistically indistinguishable. These findings provided
2-26
-------
support for combining the 2001 and Test 1 data sets and for excluding the dissimilar results of
Test 2 and Test 3.
2.6 SUMMARY OF PREFERENCE STUDIES AND SELECTION OF
CANDIDATE PROTECTION LEVELS
As described above, because each of the studies reviewed in this assessment investigates
a common question and use similar approaches that are all derived from the method first
developed for the Denver urban visibility study, we concluded that it is reasonable to compare
the results from all four urban areas to identify overall trends in the study findings and that this
comparison can usefully inform the selection of CPLs for use in further analyses. However,
because variations in the specific materials and methods used in each study introduce
uncertainties, direct comparison of the study results should take these factors into account. Key
differences between the studies include:
* Image presentation methods (e.g., projected slides of actual photos, projected images
generated using WinHaze (a significant technical advance in the method of presenting
VAQ conditions), use of computer monitor screen
* Number of participants in each study,
* Participant representativeness of the general population of the relevant metropolitan
area, and
* Specific wording used to frame the questions used in the group interview process.
Figure 2-16 presents a graphical summary of the results of the studies in the four
cities and draws on results previously presented in Figures 2-3, 2-5, 2-7 and 2-11. As
described in the separate discussions for each urban area above, the data and curves
depicted in Figure 2-16 include the following modifications: 1) the Denver results omit
the 9:00 a.m. photograph results; 2) the Chilliwack and Abbotsford results appear as a
single set of data for the BC study; 3) the results from 2001 and 2009 (Test 1) studies of
VAQ preferences in Washington, DC are presented as a single combined set of data; 4)
the results from the 2009 Washington, DC study Tests 2 and 3 are not included.
2-27
-------
Figure 2-16. Summary of Results of Urban Visibility Studies in Four Cities, Showing the
Identified Range of the 50% Acceptance Criteria
Mm-1
50 Mm-»
100 Mm * 2OO Mm-1 400 Mm"1 800 Mm"1
100%
0%
10
Denver « Phoenix * BC * Washington
Denver Logit PhoenixLogit BC Logil DC Logit
Figure 2-16 shows the results of a logistical regression analysis using a logit model of the
greater than 19,000 ratings of haze images as acceptable or unacceptable. The logit model is a
generalized linear model used for binomial regression analysis which fits explanatory data about
binary outcomes (in this case, a person rating a VAQ image as acceptable or not) to a logistic
function curve. A more complete description of the logit model application to these data is
contained in Appendix J. The results shown in Figure 2-16 are from the more generalized of the
two logit assessment models (i.e. model 2) in which both the shape and displacement of the
curves for the four cities are permitted to vary independently.
The logit analysis city intercept coefficients (Appendix J, Table 3) are all positive and
statistically significant, indicating that the response functions for different cities shifted right
relative to the function for Denver. However, only the Phoenix interaction term is insignificant,
5 Top scale shows light extinction in inverse megameter units; bottom scale in deciviews. Logit analysis
estimated response functions are shown as the color-coded curved lines for each of the four urban areas.
2-28
-------
indicating that the Phoenix response function has a different shape that is steeper than the other
three cities, as can be seen in Figure 2-16. Figure 2-16 also shows the Washington, DC function
is modestly less steep than the others, but the decrease in the slope is not statistically significant.
The model results can be used to estimate the VAQ deciview values where the estimated
response functions cross the 50% acceptability level, as well as any alternative criteria levels.
Selected examples of these are shown in Table 2-3. A t-test of these 50% acceptance deciview
values for the four cities shows each to be significantly different from the others (Appendix J,
Table 5).
Table 2-3. Logit Model Estimated VAQ Values Corresponding to Various Percent
Acceptability Values for the Four Cities
90% Acceptability criteria
75% Acceptability criteria
50% Acceptability criteria
25% Acceptability criteria
10% Acceptability criteria
Denver
14.21
17.05
19.90
22.74
25.59
British
Columbia
16.80
19.63
22.45
25.28
28.10
Phoenix
24.15
21.80
24.15
26.51
28.87
Washington,
DC
23.03
26.03
29.03
32.03
35.03
Figure 2-16 also contains lines at 20 dv and 30 dv that effectively and pragmatically
identify a range where the 50% acceptance criteria occur across all four of the urban preference
studies. Out of the 114 data points shown in Figure 2-16, only one photograph (or image) with a
VAQ below 20 dv was rated as acceptable by less than 50% of the participants who rated that
photograph.6 Similarly, only one image with a VAQ above 30 dv was rated acceptable by more
than 50% of the participants who viewed it.7 These upper and lower range values are also
supported by the logit model data which estimates 50th percentile acceptability values near 20 dv
for Denver and near 30 dv for Washington, DC (see Table 2-4).
There are several hypotheses that may explain why the VAQ acceptability response
curves for the four cities are different and why some study results have greater variability than
6 Only 47% of the BC participants rated a 19.2 dv photograph as acceptable.
7 In the 2001 Washington, D.C. study, a 30.9 dv image was used as a repeated slide. The first time it was
shown 56% of the participants rated it as acceptable, and 11% rated it as acceptable the second time it was shown.
The same VAQ level was rated as acceptable by 42% of the participants in the 2009 study (Test 1).
2-29
-------
others.8 First, as mentioned, the use of photographs (Denver and BC surveys) versus WinHaze-
generated images (Phoenix and Washington, DC surveys) may play a significant role in
preference studies, perhaps introducing bias (such as suggested by the responses to the 9:00 a.m.
Denver photographs) as well as variability. Further, the use of photographs from different days
and times of day that rely on associated ambient measurements of light extinction to characterize
their VAQ level can introduce two other types of uncertainty. The intrinsic appearance of the
scene can change due to the changing shadow pattern and cloud conditions, and spatial variations
in air quality can result in ambient light extinction measurements not being representative of the
sight-path-averaged light extinction. WinHaze has neither of these sources of uncertainty
because the same base photograph is used (i.e. no intrinsic change in scene appearance) and the
modeled haze that is displayed in the photograph is determined based on uniform light extinction
throughout the scene.
Second, variation in the degree of representativeness of the participants and the sizes of
the participant samples involved may also be important factors. The small sample size and fairly
uniform population of respondents is a plausible explanation for the noisiness of the combined
Washington, DC results (35 participants, including 26 from a single consulting firm and 10 of
those from a different city) compared with the larger and more representative population of
responders from Phoenix (385 participants, carefully selected to be representative of the Phoenix
population).
A third hypothesis put forward by Smith and Howell (2009) is that the range of VAQ
images presented in the survey may influence the results. As discussed above, a more plausible
explanation it that the range of haze images shown to participants in the VAQ 1 to 7 rating
exercise was interpreted by participants as a restriction on acceptability rating exercise to confine
their rating to the range VAQ conditions shown, which for Test 2 was curtailed to only average
to good VAQ conditions. When other evidence is taken into account, the Smith and Howell
hypothesis seems an even more unlikely explanation for the differences in results between the
four urban preference studies. For example the Denver study included photographs with the
haziest conditions among the four studies, but resulted in the lowest haze condition for the 50th
percentile preference ratings among the four, not the highest as might be expected if the range of
haze levels were a significant factor influencing the results of preference studies. Also,
inspection of the average VAQ 1 to 7 ratings for the Phoenix and Denver studies showed that
they spanned the full ratings range of values similar to those for the Smith and Howell Test 1 and
3, so the participants in those studies were not presented with a restricted range within which to
select acceptable VAQ conditions, suggesting that the range itself was not an important factor
8 Variability here refers to the degree of scatter of the average acceptability ratings for each image around
the logit curve for that city.
2-30
-------
influencing their results. Values for the British Columbia 1 to 7 VAQ rating exercise were not
readily available.
A fourth major hypothesis is that urban visibility preferences may differ by location, and
the differences may arise from inherent differences in the cityscape scene used in each city. The
key evidence to suggest this hypothesis is that the apparent differences between the Denver
results (which found the 50% acceptance criteria occurred in the best VAQ levels among the four
cities) and the Washington, DC results (which found the 50% acceptance criteria occurred at the
worst VAQ levels among the four cities). This hypothesis suggests that these results may occur
because the most prominent and picturesque feature of the cityscape of Denver is the clearly
visible snow-covered mountains in the distance, while the prominent and picturesque features of
the Washington, DC cityscape are buildings relatively nearby without prominent and/or valued
scenic features that are more distant.
Finally, and perhaps of significant importance is that the sensitivity of individual scenes
to perceived changes in VAQ under changing light extinction levels can be quite different. As in
the fourth hypothesis, this may in part explain why the Denver study scene, with its long distance
to the mountain backdrop, resulted a preference for the best VAQ level, with a 50% criteria value
of about 20 dv, while the Washington, DC study scene, with much shorter sight paths yielded a
50% criteria VAQ value at a substantially worse level of about 30 dv. The distinction between
the last two hypotheses are that the earlier one speaks to the desirability of seeing distant
mountains versus this hypothesis which concerns the ability to perceive changes in haze at lower
light extinction levels. Additional studies, including directly comparable studies using similar
methods in diverse cities, would be useful to gain further understanding of preferences for urban
visibility.
Based on the composite results and the effective range of 50th percentile acceptability
across the four urban preference studies shown in Figure 2-16, CPLs have been selected in a
range from 20 dv to 30 dv (74 Mm"1 to 201 Mm"1) for the purpose of comparing to current and
projected conditions in the assessment in chapters 3 and 4 of this document. A midpoint of 25
dv (122 Mm"1) was also selected for use in the assessment. These three values provide a low,
middle, and high set of light extinction conditions that are used in subsequent chapters of the
UFVA to provisionally define daylight hours with urban haze conditions that have been judged
unacceptable by the participants of these preference studies. As discussed earlier (section 1.2),
PM light extinction is taken to be light extinction minus the Rayleigh scatter (i.e. light scattering
by atmospheric gases is about 10 Mm"1), so the low, middle and high CPL levels correspond to
PM light extinction levels of about 64 Mm"1, 112 Mm"1 and 191 Mm"1.
2-31
-------
This page left intentionally blank.
-------
3 ESTIMATION OF RECENT PM MASS AND SPECIES
CONCENTRATIONS AND PM10 LIGHT EXTINCTION
This chapter characterizes hourly PM conditions in terms of both PM2.5 mass
concentration and PMio light extinction in a set of urban study areas during 2005-2007. This
characterization supports the following goals: (1) to improve understanding of the levels,
patterns, and causes of PM-related impairment of urban visibility during daylight hours; (2) to
create the basis for projections of PM2.5 mass and PMio light extinction levels under "what if
scenarios; and (3) to examine the correlation between PMio light extinction and potential
alternative indicator(s) based on PM2.5 mass concentration. These goals are addressed in
chapters 3, 4 and Appendix D, respectively. A number of other appendices address related topics
of particular interest in more detail.
3.1 SUMMARY OF PREVIOUS CHARACTERIZATIONS OF PM
CONCENTRATIONS AND LIGHT EXTINCTION
3.1.1 PM2.5 and PMi0-2.s
Chapter 2 of the 2005 Staff Paper (US EPA, 2005) from the previous review and chapters
3 (especially section 3.5) and 9 (especially section 9.2.3) and Annex A of the final ISA (US
EPA, 2009a) from the current review present extensive characterizations of the levels,
composition, and temporal and spatial patterns of PM2.5 in U.S. urban areas. Both documents
present data summaries based on the approximately 1000 PM2.5 monitoring sites in the U.S. The
characterizations in the 2005 Staff Paper were based on 2001-2003 data. The characterizations
in the ISA are based on 2005-2007 data, which is the same time period used in this visibility
assessment. While there generally have been reductions in the concentrations of PM2.5 in many
areas as a result of emission reductions of PM2.5 and its precursors, the general patterns, and the
diversity of patterns across areas, noted in the 2005 Staff Paper still prevailed in the 2005-2007
period.
Using 2005-2007 air quality data, 38 urban areas violated the annual PM2.5 NAAQS set at
a level of 15|ig/m3 in 1997 and retained in the last review completed in 2006. Seventy-six areas
violated the 2006 24-hour NAAQS level of 35|ig/m3. There is considerable but not complete
overlap in the areas not meeting the two NAAQS. It should be noted that in many parts of the
U.S., PM2 5 concentrations in 2005 were high relative to the next three years. Figure 3-1
illustrates PM2.5 air quality in 2007 by representing each monitor by a symbol whose color
reflects the annual mean of the concentration at that site or the 98th percentile 24-hour
concentration, in both cases in that one year.
5-1
-------
jth
Figure 3-1. Annual Average and 24-hour (98 Percentile 24-hour Concentrations) PM2.s
Concentrations in ug/m3, 2007.
Annual
Concentration Range (pg/m3)
• 3.4-12.0(418 Sites)
O 12.1 -15.0 (356 Sites)
O 15.1 -18.0 (86 Sites)
• 18.1 -22.5 (14 Sites)
Puerto Rico
24-hour
Concentration Range (|jg/mj)
• 7-15 (38 Sites)
O 16 - 35 <662 Sites)
O 36-55 (166 Sites)
• 66-73 (18 Sites)
Puerto Rico
$-2
-------
Each urban area exhibits its own detailed patterns of observed concentration levels,
temporal and spatial variation, and composition. These differences are due to differences in local
and transported emissions and in meteorology. Because of differences in the placement of PM2 5
monitoring sites in each urban area, the actual levels and spatial pattern of PM2.5 and PM2.5
species concentrations may not be consistently discernable in all areas. This variability and
limited monitoring network make it difficult to offer concise generalizations, although some
broad similarities can be drawn among areas.
Midwestern, southeastern, and eastern urban areas have much higher sulfate levels than
do more western areas, attributable to the much higher emissions of SO2 in and upwind of them.
Areas in the upper Midwest and to a lesser extent upper East have notable nitrate concentrations
in winter but not in summer, while southeastern areas generally have lower nitrate concentrations
even in winter. Many western urban areas have large nitrate concentrations year round. In all
areas, carbonaceous material is an important component of PM2 5 and is attributable to many
emission sources of organic material in PM form and of organic PM precursor gases. In some
areas with high local use of wood for residential heating carbonaceous material is dominant
during the heating season. PM2 5 derived from crustal sources is generally a small fraction of
total mass, except during local high wind events or due to brief periods of intercontinental
transport of dust from Africa or Asia.
Comparison of PM2 5 species concentrations within and outside urban areas leads to the
conclusion that, in the eastern areas with high sulfate concentrations, the large majority of the
sulfate affecting any given urban area originates outside that area. Inward transport and local
generation of nitrate and carbonaceous material are more evenly balanced in eastern areas, with
some differences among areas. In western areas, local sources dominate for carbonaceous
material and nitrate, with the origins of the small sulfate component being more balanced. See
Figure 9-24 of the final ISA (US EPA, 2009a).
Southeastern areas have their highest PM2.5 concentrations in the summer, when
conditions are most conducive to sulfate formation. More northern areas, being affected by a
more balanced mix of contributors, tend not to have such a strongly seasonal pattern. The
seasonal patterns in western areas are individual and varied, related to differences in local
sources and formation and dispersion conditions. In all areas, inversion conditions with low
wind speeds are conducive to high concentrations due to the trapping of emissions from local
sources. Some western areas, especially those with valley or bowl-like topography, are
especially affected.
There is at present no systematic monitoring network in place for PMi0.2.5, as states have
until January 1, 2011, to implement required monitoring sites for PMio-2.5. Consequently,
estimates of PMio-2.s must be developed using data from PM2.5 and PMio monitoring sites and
5-3
-------
equipment, which are not always collocated and consistent. The 2005 Staff Paper presented such
estimates in section 2.4.3. The final ISA presents such estimates in Figure 3-10 and Table 3-9 of
section 3.5.1.1. The 2005 Staff Paper used a data-inclusive approach in which the best available
data on PM2.5 and PMio concentrations - in some cases not very robust data - were used to
estimate 2001-2003 PMio-2.5 concentrations for 351 metropolitan area counties. For these
counties, the annual mean PMio-2.5 concentrations were generally estimated to be below 40
ug/m3, with one maximum value as high as 64 ug/m3 and a median of about 10-11 ug/m3. The
ISA used a much more data-restrictive approach based only on paired (collocated) low-volume
filter-based samplers for both PMio and PM2.5. The ISA reports that only 40 counties have such
paired samplers. Using these available co-located PM measurements from 2005-2007, the mean
24-hr PMio-2.5 concentration in these 40 counties was 13 ug/m3. This urban visibility assessment
has used a data-inclusive approach to estimating PMio-2.5 concentrations, similar to that used for
the 2005 Staff Paper, where needed to obtain hourly PMio-2.5 estimates for the 15 selected study
areas, which are reported below in section 3.3.2.
Additional detail on PM2.5, PMio, and PMio-2.5 concentrations, composition, and patterns
appears in section 3.5.1.1 of the ISA. Also, chapter 6 of the 2004 PM Assessment by NARSTO
contains more detailed characterizations of PM in different parts of the U.S.
3.1.2 PMio Light Extinction
While total light extinction is directly measurable using a transmissometer and PMio light
extinction can be measured with other instruments, there are very few regularly operating
monitoring sites measuring either form of light extinction in urban areas, and generally those that
do operate do not submit data to AQS.1 Consequently, any characterization of PMio light
extinction conditions based on actual measurements is necessarily less comprehensive than for
PM2.s and PMio-2.5- Many monitoring sites that employ nephelometers, which measure light
scattering, operate that equipment in a heated mode for purposes of tracking "dry" PM2.5 mass
concentrations, and actual light scattering due to ambient PM is not reportable. There are many
more filter-based Aethalometers® and similar instruments for measuring light absorption in
operation and reporting to AQS, but light absorption is typically a small fraction of total PM
light extinction, so these data alone are not a good indicator of overall PMio light extinction in
1 EPA is aware of routine, long-term direct measurement of light extinction using transmissometers only in
the Phoenix, AZ, Denver, CO, and Washington, DC urban areas, none of which submit data to AQS, although the
site in Washington submits data to the IMPROVE program data system. Also, there is a large network of "visual
range" monitors in operation at U.S. airports, aimed at providing information to determine landing and takeoff
safety. Due to their locations and to the lack of data resolution (values of visual range above the level needed for
unlimited airport operations are not individually reported) the data from these monitors are not suitable for use in
this assessment. The final PM ISA discusses these monitors in section 9.2.2.3.
5-4
-------
urban areas. Also, there are unresolved issues of data corrections and comparability for the light
absorption data from these instruments now residing in AQS.
PMio light extinction can be "reconstructed" from measurements of PM2 5 mass
components and PMio-2.5 concentrations, in combination with relative humidity values, using
either of two versions of the formula known as the IMPROVE algorithm but excluding its term
for Rayleigh scattering by gases in clean air. (Section 9.2.2.2 of the ISA gives an overview of
the algorithm and its basis. Section 3.2.3 of this document discusses the application of the
original version of the IMPROVE algorithm in this assessment. PM2.5 component measurements
are generally available only on a 24-hour average basis, so it generally is possible to estimate
only 24-hour average PMio light extinction, unless additional information on hourly patterns is
brought to bear.2 Because EPA's Regional Haze Rule (RHR) currently requires states to address
visibility problems in Class I visibility protection areas, which are nearly all rural and remote,
there is a large body of literature characterizing PMio light extinction in remote rural areas, based
on data from the IMPROVE network's 24-hour samplers and on special studies. Sections 9.2.3.2
and 9.2.3.4 of the ISA summarize this literature. Section 9.2.3.3 of the ISA contrasts
concentrations of PM2.5 and PM2.5 components between rural and urban areas using data from the
rural IMPROVE network and the urban Chemical Speciation Network (CSN) but does not
present estimates of PMio light extinction in urban areas.
The CSN network provides 24-hour PM2.s species measurements at about 200 urban
sites, from which mass components can be derived. These sites have a mix of daily, one day in
three, and one day in six sampling schedules. The 2005 Staff Paper (and its references) may be
the only readily available prior assessment to use these urban PM2.s speciation monitoring data,
along with estimates of PMio-2.5 concentrations and data on relative humidity, to reconstruct daily
24-hour average light extinction in urban areas, for the year 2003.3 One presentation of the
results was in the form of a scatter plot of daily 24-hour reconstructed light extinction versus 24-
hour PM2.s concentration. This graphic appears here as Figure 3-2. (For the immediate purpose
of this section, it is the distribution of the data points along the y-axis that is of interest, not the
relationship between light extinction and PM2.s concentrations; the latter subject is addressed in
When the IMPROVE algorithm is used to estimate 24-hour PM10 light extinction from 24-hour PM2 5
species and PM10.2 5 concentrations, an assumption is made that every hour has the same PM concentrations but its
own relative humidity value. Hourly estimates of PM10 light extinction, including the strongly non-linear effect of
relative humidity, are then averaged to get the 24-hour PM10 light extinction estimate.
3 Estimates of light extinction in the 2005 Staff Paper include Rayleigh scattering of 10 Mm"1 and thus
represent "total" light extinction (excluding NO2 absorption). Adjustment for consistency must be made before any
close comparisons to PM10 light extinction values in this document.
3-5
-------
Appendix D.) Generally, most days have light extinction below 200 inverse megameters (Mm"1),
-1 4
but a small percentage of values were as high as about 750 Mm" .
Figure 3-2. Reconstructed 24-hour Light Extinction in U.S. Urban Areas in 2003
Source: Schmidt et al., 2005
Output D.3
(Relationship RE & PM2.5; Diurnal RE; Timeframe)
2 of 30
Bscon,
944
709
coc
50C
IOC
30»
1M)C
IOC
Significant relationship (low p-LValue)
East (circles): RE = y - 8.1 * PM,, R-0.69
West (stare): n = yV8.-'PM,","P>D.75
1C
r.o
no
Relatiou^lup behifen reconstructed light eitinctioD (RE) aod !4-Loui average PM2 ^, 2003. I'^ing actual^RH)
In addition to this scatter plot, a table developed for the previous PM NAAQS review
presented the annual average of estimates of 24-hour reconstructed light extinction values,
averaged across 161 urban areas grouped into seven regions (Schmidt, et al., 2005). Table 3-1
reproduces these estimates. For regions excluding Southern California, annual average 24-hour
4 Unfortunately, the file of paired data used to create this scatter plot is no longer available, so the actual
distribution of light extinction values cannot be described more specifically.
j-c
-------
light extinction ranged from 73 to 118 Mm"1. The estimate of the annual average 24-hour light
extinction for Southern California was 168 Mm"1. These estimates were based on 10-year
average 1-hour relative humidity values and 2003 PM monitoring data.
Table 3-1. Annual Mean Reconstructed 24-hour Light Extinction Estimates
by Region (Mm *)
Region
Reconstructed 24-hour Light
Extinction in 2003
Northeast
108
Southeast
98
Industrial Midwest
118
Upper Midwest
80
Southwest
73
Northwest
76
Southern California
168
Source: Output D.3, Schmidt et al, 2005. We note these regions were used to summarize PM2 5 patterns for the PM
NAAQS review 1997 (US EPA, 1996b).
Figure 3-3 is a contour map of annual average reconstructed 24-hour PMi0 light
extinction based on IMPROVE monitoring sites in 2000-2004, nearly all of which are remote
and rural (the three urban sites in Phoenix, AZ, Washington, DC, and Puget Sound, WA are
indicated by square symbols). A comparison of the mean urban light extinction levels by region
5-7
-------
listed in Table 3-1, with this map of rural light extinction indicates that in most parts of the U.S.,
light extinction levels in urban areas are notably higher than in the surrounding remote rural area,
with the northeast and the southeast regions having the most similarity between rural and urban
light extinction levels. This is consistent with observations of an "urban excess" of PM2.5 and
Figure 3-3. Isopleth Map of Annual Total Reconstructed PMio Light Extinction Based
on 2000-2004 IMPROVE Data.
* IMPROVE Site
• IMP ROVE Urban Site
Puerto Rico /
Virgin Islands
(Source: Spatial and Seasonal Patterns and Temporal Variability of Haze and its Constituents in the
United States Report IV, DeBell 2006)
PMio-2.5 and with the known high regional concentrations of sulfate in these eastern areas.
One-hour PMio light extinction values of course vary above and below the 24-hour
average, due to diurnal variations in PM2.5 component concentrations, PMio-2.5 concentrations,
and relative humidity. Although PMio light extinction was formally reconstructed on an hourly
basis in the 2005 Staff Paper analysis for the last review cited above, the actual full strength of
the diurnal pattern could not be discerned in that analysis because component mix was assumed
not to vary from hour to hour. Under the unverified assumption of constant component mix and
using actual hourly relative humidity data, the daily maximum daylight 1-hour PMio light
extinction values were roughly 50 percent higher than the 24-hour average PMio light extinction
-------
values.5 The new analysis presented in this document includes a closer look at diurnal patterns,
for 15 study areas.
3.2 OVERVIEW OF APPROACH AND DATA SOURCES FOR URBAN STUDY
ANALYSIS
As explained above, there are limited data from direct measurements of PMi0 light
extinction in urban areas. Consequently, this assessment has reconstructed hourly PMio light
extinction levels for daylight hours from values of hourly PM2.5 components, PMio-2.5, and
relative humidity. Hourly monitoring data for PM2.5 components and PMio-2.5 are also generally
lacking, so the estimates of these parameters have necessarily been developed from a
combination of other available ambient monitoring data and air quality modeling results from a
chemical transport model (CTM) run. Specifically, the ambient monitoring data starting points
are 24-hour PM2.5 mass measured by filter-based Federal Reference Method (FRM) or Federal
Equivalent Method (FEM) monitors6, 24-hour PM2 5 components measured by the filter-based
monitors of the CSN, and hourly PM2.5 mass measured by continuous instruments such as the
Tapered Element Oscillating Microbalance (TEOM), beta attenuation monitors (BAMs), and
nephelometers, which were used at different sites. The CTM-based diurnal profiles for
individual components, in conjunction with hourly PM2 5 measurements, are used to adjust and
allocate the 24-hour PM2 5 components measurements to individual hours of each day, as
described in detail below. In addition, levels of hourly PMio-2.s mass are calculated from
separate measurements of hourly PMio and hourly PM2 5 if both are available, or by applying
PMio-2.s to PM25 ratios to hourly PM2 5 data if both types of hourly measurements are not
available. The ambient data are from 2005-2007 and were all obtained from AQS in the first half
of 2009.
The CTM run was the "actual emissions" or "validation" run of the 2004 CMAQ
modeling platform with boundary conditions provided by GEOS-Chem global scale CTM.7 The
CTM modeling is used as one element in the development of realistic diurnal variations for each
of the major PM2.5 components used to estimate PMio light extinction, anchored to site-specific,
day-specific measurements of 24-hour concentrations. That is, monthly averaged diurnal
profiles for the five major components were generated using the CTM results, which were then
5 These observations on diurnal patterns come from examination of "Output D.3 (Relationship RE & PM2 5;
Diurnal RE; Timeframe) 8 of 30" and "Output D.3 (Relationship RE & PM2 5; Diurnal RE; Timeframe) 17 of 30",
Analyses of Paniculate Matter (PM) Data for the PM NAAQS Review, Schmidt et al., 2005.
6 Filter-based Federal Reference Method samplers and filter-based Federal Equivalent Method samplers
will both be referred to as FRM samplers in the remainder of this document.
7 GEOS-Chem is the NASA Goddard Earth Observing System-CHEMistry (global 3-D CTM for
atmospheric composition). This modeling platform, with an appropriately different emissions scenario, is also the
basis for the estimates of policy relevant background concentrations of PM25 presented in section 3.6 of the ISA (US
EPA, 2009a).
3-9
-------
combined with hour-specific measurements of PM2.5 to generate hourly concentration variations
for each of the 24-hour CSN sample days during the 2005-2007 period.
3.2.1 Study Period, Study Areas, Monitoring Sites, and Sources of Ambient PM
Data
At the time this assessment began, the ambient monitoring data from 2005-2007, but not
from 2008, had been certified as accurate and complete by the state/local monitoring agencies
that collected them, and the data had been extensively summarized and presented in the first draft
ISA. The EPA staff aimed to develop estimates of daylight hours PMio light extinction for a
reasonably representative number of days in each year of 2005-2007, to allow the application of
statistical forms based on three years of data. However, as explained in more detail below, in
several study areas the limited availability of starting data for these estimates resulted in estimate
sets that do not cover all three years. Also, even in areas with some data in all three years, the
number of days with valid estimates differs by year and is in some cases not large by typical
standards of monitoring data completeness.
For efficiency in the analysis, this visibility assessment uses the same 15 urban study
areas selected for the health risk assessment. These areas are listed in Table 3-2, along with the
area-wide (maximum) FRM-based 2005-2007 annual and 24-hour PM2.5 design values for each
study area based on the highest-reading monitor in each area, and for the specific site used in this
assessment.8 (See below for an explanation of the "site-specific" columns in Table 3-2.)
8 2005-2007 PM2 5 design values were taken from the information posted at
http://www.epa.gov/airtrends/values.html. and are consistent with the design values used in the health risk
assessment to "roll back" current concentrations to represent achievement of alternative annual and 24-hour PM2 5
NAAQS. Except in Dallas and Fresno, the area-wide design values are the highest design values of any monitoring
site in the designated (1997 NAAQS) nonattainment area that has sufficiently complete data to allow the calculation
of a design value according to the provisions of 40 CFR 50 appendix N. For Dallas, the design values come from a
site with nearly complete data, and are somewhat higher than the highest values from a site with complete data (see
the PM Risk Assessment, US EPA, 20 lOe, section 3.2.3). For Fresno, the area-wide design value is for the Fresno-
Madera CSA, which is only a portion of the San Joaquin Valley nonattainment area. Also, note that there are three
cases in which the nonattainment area does not include certain areas sometimes thought of as being part of the area
named in Table 2; monitors in these non-included areas were not considered in this assessment. (1) The design value
shown for Pittsburgh is for the Pittsburgh-Beaver nonattainment area; the Liberty-Clairton nonattainment area is
within the Pittsburgh CBS A but is distinct for regulatory purposes, and was not considered in this assessment. (2)
Baltimore was treated separately, although part of a CSA with Washington DC. (3) Berks Co., PA is part of the
Philadelphia-Camden-Vineland CSA, but not part of the Philadelphia-Wilmington nonattainment area.
3-10
-------
Table 3-2. Urban Visibility Assessment Study Areas
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Area-wide
2005-2007
Annual
Design Value
(Hg/m3)
10.2
17.4
19.6
12.6
11.6
12.8
15.8
16.5
18.7
16.2
17.2
16.5
15.6
15.0
15.9
Area-wide
2005-2007
24-hour
Design Value
(Ug/m3)
43
63
55
32
55
26
31
39
44
35
43
43
37
38
42
Site-specific
2005-2007
Annual Design
Value
(Ug/m3)
Same
Same
Same
7.9
10.7
11.5
13.1
14.5
Same
15.7
Same
15.0
14.5
14.7
14.4
Site-specific
2005-2007
24-hour
Design Value
(Ug/m3)
Same
Same
Same
15
48
25
25
34
Same
33
Same
40
35
37
42
2005 Staff Paper
Region
(See map in Table
3-1)
Northwest
Southern
California*
Southern California
Southwest
Northwest
Southeast
Southeast
Midwest
Southeast
Southeast
Midwest
Industrial Midwest
Northeast
Northeast
Northeast
* While not generally considered to be part of Southern California as the term is commonly used, Fresno lies just
south of the line used in the 2005 Staff Paper (US EPA 2005) (based on earlier work by others) to separate the
Southern California region from the Northwest region.
For time reasons and because it was anticipated that some study areas would not contain
more than one suitable study site, EPA staff sought to identify the single best study site in each
area. In identifying the single best study site in each study area first consideration was given to
the availability of collocated 24-hour data on PM2.5 and its components, because the contribution
of PM2.5 components to PMi0 light extinction will typically dominate the contribution from
PMio-2.5. Ideally, within each study area the three types of PM2.5 data (FRM PM2.5, CSN PM2.5
components, continuous PM2.5) would be available at a common site, and that site would be
located in a manner consistent with reliance on it to characterize visibility as it would be
perceived by a large number of area residents and visitors. As can be seen in Table 3-2, in 10 of
the 15 study areas the site providing FRM data for this assessment is not the area-wide design
value site, because the area-wide design value site did not have collocated CSN and/or
continuous PM2 5 data.
Appendix A provides details on the site(s) identified and used in each study area,
including information on the type of monitoring equipment that provided the data and other
information that may help interpret the results of the analysis. A portion of this table for a single
site - Tacoma - is presented here as Table 3-3 as an example. When viewing this document
3-11
-------
electronically, the site IDs in these tables are active links and can be used to view the location of
the site via GoogleMaps.9
In 11 of the study areas, the three types of PM2 5 data were available at a common site. In
the remaining four areas, Phoenix, AZ, Pittsburgh, PA, Baltimore, MD, and St. Louis, MO-IL,
two types of data were available at one site, but the remaining type of data had to be taken from
another site and treated as being representative of the former site.
The monitoring agencies described all but one of these sites as neighborhood or urban
scale, indicating those agencies' opinion that the sites represent concentrations in an area at least
0.5 to 4 km across. An aerial view of the remaining site (in Phoenix) which did not have a scale
characterization recorded in AQS suggests that it may be middle or neighborhood scale. As
already stated, selected sites are not necessarily the locations of the maximum measured annual
or 24-hour PM2.5 levels in their urban area.
Site days which were missing 1-hour PM2.5 concentration data points for more than 25
percent of daylight hours were excluded from the analysis, because such data gaps were judged
to result in too much uncertainty in estimates of 1-hour PM2.5 components, 1-hour PMi0 light
extinction, and daily maximum PMio light extinction. Days with fewer missing 1-hour PM2.5
concentration data points were retained, but no estimate of PMio light extinction was made for
hours without 1-hour PM2.5 concentration data (see below for more explanation). Hourly
2.5 presented more varied challenges. In four areas (Birmingham, Detroit, Baltimore, and
Philadelphia) the site that provides the continuous PM2.5 data also hosts a continuous FEM
monitor, and hourly PMio-2.5 could be calculated by difference for most hours. In other areas,
this was not the case, and either (1) hourly instruments at two different sites were used in this
subtraction (Tacoma, Los Angeles-South Coast Air Basin, Phoenix, St. Louis, Atlanta, and New
York-N. New Jersey) or (2) a single regionally applicable PMio-2.5 to PM2.5 ratio calculated as
part of the last review based on 2001-2003 24-hour FRM/FEM PMio and PM2.5 samples was
applied to 2005-2007 hourly PM2.5 data to estimate hourly PMio-2.5 (Fresno, Salt Lake City,
Dallas, Houston, and Pittsburgh). In the case of Los Angeles-South Coast Air Basin, the
continuous PMio and PM2.5 sites were quite distant and separated by a range of hills, so the
estimates of PMio-2.5 and its contribution to PMio light extinction are more uncertain than if the
monitors were clearly within the same air mass. Comments on the second review draft of this
document from those familiar with the monitoring sites in St. Louis indicate that the site selected
to provide continuous PMio monitoring, though less than a mile from the site of the PM2.5 data is
9 Additional meta data on each monitoring site, and access to daily and annual data listings, can be
conveniently obtained using GoogleEarth and the PM25, PM10, and CSN monitoring network KML files that can be
downloaded from http://www.epa.gov/airexplorer/monitor kml.htm.
3-12
-------
not representative of the urban area and resulted in unrealistically large PMio-2.5 values.10
Obviously, for the five study areas for which 1-hour PMio-2.5 was estimated by application of
ratios, PMi0-2.5 estimates can only represent broad trends, not hour-specific conditions at the
particular site. More description of the methods used for estimating hourly PMio-2.5 appears in
section 3.3.2.
10 Comments concerning unrealistically high PM10_2 5 values for St. Louis are viewed as credible, but were
received too late in the review process to permit reanalysis using an alternate data set or to remove St. Louis from all
portions of this document. However, the text has been revised to caution readers with respect to the St. Louis
results, and they will not be included in the visibility effects discussion in the final PM Policy Assessment
document. Some graphics have been updated to exclude St. Louis results.
3-13
-------
Table 3-3. PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the Tacoma
Study Area
Study
Area
First PM2.5
Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
Data Source for PMi0-2.s
Tacoma
AQS ID 530530029
State: Washington
City: Tacoma
MSA: Tacoma, WA
Local Site Name: TACOMA -
L STREET
Address: 7802 SOUTH L
STREET, TACOMA
0.5 miles east of 1-5
2005-2007 annual DV = 10.2
2005-2007 24-hr DV = 43
This is the highest 24-hour
PM2.5 DV site in the Seattle-
Tacoma-Olympia, WA annual
PM2 5 nonattainment area
Neighborhood Scale
Parameters taken from this
site:
* 24-hour FRMPM2.5 mass
(AQS parameter 88101;
one-in-three sampling
schedule)
* PM2 5 speciation (one-in-
six sampling schedule)
* l-hourPM25mass (AQS
parameter 88502,
Acceptable PM25 AQI &
Speciation Mass)
Correlated Radiance
Research M903
Nephelometry
No continuous PM10
monitoring at this site, see
right hand column.
N/A
AQS ID 530530031
State: Washington
City: Tacoma
MSA: Tacoma, WA
Local Site Name: TACOMA -
ALEXANDER A VE
Address: 2301 ALEXANDER AVE,
TACOMA, WA
6.4 miles NNE of PM2.5 site
Neighborhood Scale
Parameters taken from this site:
* 1-hour PM10 STP mass (AQS parameter
81102)
* Sample Collection Method:
INSTRUMENTAL-R&P SA246B-
INLET
* Sample Analysis Method: TEOM-
GRAVIMETRIC
7% of PM10.2.5 values were determined using
regional average PM10.2.5:PM25 ratios from
2005 Staff Paper
Additional Explanation
In this Table, the 1-hour concentration parameter "88502, Acceptable PM2.5 AQI & Speciation Mass" is the same as the ISA refers to
as "FRM-like" PM2 5 mass. An entry of "88501, PM2.5 Raw Data" indicates that the monitoring agency makes no representation as to
the degree of correlation with FRM PM2.5 mass. The latter type of continuous PM2 5 data were used only when the former were
unavailable.
Where PM10 was reported in STP, it was converted to LC before PM10.2.5 was calculated.
For convenience, continuous PM2.5 data was obtained through the AirNow website rather than from AQS, as an initial exploration
indicated that not all the desired 1-hour data had been submitted to AQS.
3-14
-------
The sampling schedule for CSN PM2.5 speciation monitoring was one-in-six days for
Tacoma, Phoenix, Houston, Detroit, and Philadelphia, and one-in-three days for the other study
areas. Not every scheduled CSN site day in 2005-2007 had data for all three types of PM2 5 data,
due to missed or invalid samples. Also, for continuous PM2.5, values for a small number of hours
of an otherwise data-sufficient day were sometimes missing, due to equipment failure or
servicing. EPA staff retained only those days in which 75 percent or more of daylight hours had
measurements of PM2.5 (see section 3.3. for more details). If for isolated hours at a site (or site
pair) with collocated measurements, PMio-2.5 concentrations could not be estimated because of
gaps in the same-hour continuous PMio and/or PM2.5 data, EPA staff used the regional ratio
approach described above to estimate PMio-2.5 for those specific hours. Table 3-4 provides more
detailed information on the quarterly distribution of the successfully matched and sufficiently
complete data available for use. As described later, for some parts of this assessment EPA staff
substituted data for the single missing quarters of data in Phoenix and Houston, to achieve
seasonal balance. For some sites, two CSN samplers operated on some days for data quality
assessment purposes; when this was the case, the results from the two samplers were averaged.
In this assessment, we have not excluded PM concentration data that may have been
affected by exceptional events such as wildfires and wind storms. Under EPA's Exceptional
Events rule, for existing NAAQS states may request exclusion of such data from regulatory
determinations, and accordingly such data are not reflected in design values for existing NAAQS
once exclusion is approved by EPA. A similar arrangement presumably would apply to a new or
revised secondary PM NAAQS. Design values for PMio light extinction under current
conditions (Table 4-2) and percentage reductions to "just meet" alternative secondary NAAQS
based on PMio light extinction (Table 4-3), presented below in chapter 4, may thus be
overestimates. Overestimation is more likely for the western study sites than for the eastern
study sites. However, PM2.5 design values shown in Table 3-2, and associated estimates of the
reductions needed from 2005-2007 PM2.5 level to just meet alternative secondary NAAQS based
on PM2.s mass (Table 4-4) do reflect the exclusion of at least some data affected by exceptional
events.
3-15
-------
Table 3-4. Number of Days per Quarter in Each Study Area
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Total Number
of Days
109
324
300
86
306
274
149
292
350
285
141
281
186
145
227
2005
Qi
0
19
26
0
27
22
21
26
30
20
12
25
19
15
22
Q2
0
24
28
13
28
24
20
27
30
25
12
23
17
11
23
Q3
0
27
22
11
30
26
10
24
29
25
10
25
15
13
13
Q4
0
27
28
14
26
22
14
27
30
22
11
21
11
10
14
2006
Qi
13
30
26
12
20
23
14
28
29
26
12
22
15
9
23
Q2
15
29
26
13
28
23
12
19
29
27
13
25
16
13
19
Q3
15
29
27
11
31
24
8
27
30
26
11
24
18
10
18
Q4
14
27
22
12
20
24
12
28
30
24
15
26
18
13
21
2007
Qi
12
26
21
0
23
18
15
29
30
25
11
22
12
13
19
Q2
13
28
26
0
25
23
14
25
30
19
11
22
12
14
15
Q3
14
30
24
0
19
24
9
22
27
26
12
23
17
12
19
Q4
13
28
24
0
29
21
0
10
26
20
11
23
16
12
21
Note: Only days with matched and sufficiently complete data were retained in the assessment.
3.2.2 Use of CMAQ Model Validation Runs for 2004 to Augment Ambient Data
Because systematic monitoring data on hourly PM2.s component concentrations are not
available for most of the 15 study areas, EPA staff extracted and applied certain information
from the modeling platform for calendar year 2004 described in section 3.7.1.2 of the ISA, in
which the global-scale circulation model GEOS-Chem was paired with the regional scale air
quality model CMAQ.11 The main use of this platform in the ISA is to estimate policy-relevant
background concentrations of PM2.5. For the urban-focused visibility assessment described here,
however, we used results from the validation run of the platform, in which emissions for all
emission source types and countries are included, to develop realistic diurnal variations of the
majorPM2.5 components.
EPA staff identified the one or more 36 km-by-36 km CMAQ grid cells generally
corresponding to the urbanized area surrounding each study site, thus omitting grid cells
dominated by rural land uses.12 We then extracted from the detailed model output for these grid
11 Similar modeling was not available for 2005, 2006, or 2007.
12 Urbanized area here refers to a specific land area identified by the U.S. Census Bureau based on
population density and other factors. Shape files for these areas were compared to the CMAQ grid to identify the
grid cells to be used.
3-16
-------
cells the day/hour-specific concentrations of sulfate, nitrate, elemental carbon, organic carbon,
and "crustal/unspeciated" PM2.5 during 2004, and then we averaged across grid cells and then
across days within the month for each individual hour of the day.13 Thus, for each species, we
obtained 24 hour-of-day values for a month, for each of the 12 calendar months. We then
averaged the 24 hour-of-day values in each monthly set for each component to obtain the
corresponding 24-hour average concentration for the month. We then divided each hour-of-day
value by the 24-hour value, to obtain a normalized diurnal profile for the pollutant, which was
taken as the initial representation of all days in that month for 2005, 2006, and 2007 (but further
adjusted day-by-day in a later step). In total, this resulted in 5 (components) x 12 (months) x 15
(study areas) = 900 profiles. Visual examination of a number of these showed them to be
reasonably smooth and generally to show morning (and sometimes also late afternoon/evening)
peaks which are the anticipated effect of higher vehicle traffic and lower mixing heights. The
peaks were generally moderate, as would be expected in light of the averaging of predictions for
multiple large grid cells, the averaging across days, and the generally moderate diurnal profiles
for SMOKE pre-processing of emissions in the CMAQ modeling platform. (Note, however, that
as described below a later step in the estimation process reduces the smoothness in the diurnal
pattern of PM components.) Sulfate, as would be expected for a regionally transported pollutant,
generally had a flatter diurnal profile than for other components. Hourly nitrate concentrations
were low when expected: during warmer months and in warmer areas. Figure 3-4 shows
example diurnal profiles for the five PM2.5 components, for the Detroit study area for the months
of January and August. Diurnal profiles like these were applied to 24-hour CSN measurements
of component concentrations, as explained in detail below.
13 For several of the listed components that are not direct CMAQ outputs, concentrations were estimated by
post-processing to aggregate the appropriate CMAQ outputs. The "crustal/unspeciated" CMAQ output results from
non-reactive dispersion of that portion of the PM2 5 emission inputs not assigned during SMOKE processing to a
more specific CMAQ species, and is considered in most EPA analyses to represent the same material as the "soil"
component reported for IMPROVE sampling.
3-17
-------
Figure 3-4. January and August Monthly Average Diurnal Profiles of PM2.s Components
Derived From the 2004 CMAQ Modeling Platform, for the Detroit Study
Area.
CMAQ Relative Patterns (JANUARY)
-soil
-ec
no3
so4
-ocm
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
CMAQ Relative Patterns (AUGUST)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
3-18
-------
3.2.3 Use of Original IMPROVE Algorithm to Estimate PMio Light Extinction
The EPA staff used the original IMPROVE light extinction algorithm, rather than the
more recent revised version, because the original version is considered more representative of
urban situations, when emissions are still fresh rather than aged as at remote IMPROVE sites.14
To maintain consistency with the form of the candidate protection levels (CPLs) for PMio light
extinction identified in chapter 2, EPA staff excluded from the IMPROVE algorithm for total
light extinction the term for Rayleigh scattering by gases in clean air. The formula for PMio light
extinction using the traditional IMPROVE algorithm but without the Rayleigh scattering term is
shown below.
= 3 x f(RH) x [Sulfate]
+ 3 xf(RH)x [Nitrate]
+ 4 x [Organic Mass]
+10 x [Elemental Carbon]
+ 1 x (Fine Soil]
+ 0.6 x [Coarse Mass]
light extinction (bextPM) is in units of Mm"1, the mass concentrations of the
components indicated in brackets are in ug/m3, and f(RH) is the unitless water growth term that
depends on relative humidity. We refer to the first five terms in this algorithm as the five PM2.5
components. In this algorithm, the sulfate and nitrate components are to be expressed as fully
neutralized and as retained and measured in the IMPROVE sampling and laboratory methods.
Associated water is to be omitted from all bracketed terms since the water absorption effect is
reflected in the f(RH) term. The organic mass component is to include the mass of associated
elements in addition to carbon. As described below, we included steps in our development of
estimates of hourly component concentration to ensure consistency with these aspects of the
IMPROVE algorithm.
3.3 DETAILED STEPS
3.3.1 Hourly PM2.s Component Concentrations
The task of estimating hourly PM2.5 component concentrations is in a sense over-
determined, given the four types of available information: 24-hour PM2.5 mass by filter-based
FRM, 24-hour component concentrations by CSN, hourly PM2 5 mass by continuous instrument,
14 Other differences between the original and revised algorithms include estimates of sea salt contributions
which can be important for near-coastal locations, inclusion of site-elevation specific Rayleigh light scattering and
provision for calculating NO2 light absorption when NO2 data are available. Their exclusion in this assessment is
not expected to make any appreciable difference to the results or conclusions.
3-19
-------
and diurnal profiles of components from the 2004 CMAQ run. There are multiple ways in which
two or three of these four data sources could be used to estimate hourly PM2.5 component
concentrations, and the result generally can be expected to be at least somewhat inconsistent with
the information in the remaining data source(s). For example, each 24-hour PM2.5 component
mass from CSN sampling can be apportioned to hours based on the monthly average diurnal
profile developed from the 2004 CMAQ run, but then in general the hourly values of PM2.5 mass
determined by summing the components in an hour would not exactly match the data from the
continuous PM2.5 instrument. EPA staff therefore used a sequence of steps which achieves a
prioritized compromise among the data sources. In this sequence, we have given greater weight
to the 24-hour FRM, CSN, and continuous PM2.5 mass data because these are instrument-based
and location- and day-specific, than to the CMAQ-based profiles which are CTM-based,
averaged to the month, and extrapolated from 2004 to each of 2005, 2006, and 2007.
Because of differences in filter materials, sample collection, laboratory analysis, and data
reporting, there are differences between the contribution of some PM components to PM2.5 mass
as reported by a filter-based 24-hour FRM sampler, and the mass of the same components as
reported by CSN (or IMPROVE) sampling. The following summary of these differences may be
helpful in understanding the steps used to develop estimates of hourly PM2.5 components in this
analysis. In the IMPROVE algorithm for reconstructing PMio light extinction, the light
extinction contribution multipliers per unit of mass concentration of components are not all the
same for the five principal components. Consequently, care is required to estimate these
components as consistently as possible with the IMPROVE sampling and analytical methods so
that particle mass is correctly assigned to the right component.
• Nitrate: CSN (and IMPROVE) sampling uses a Nylon filter for purposes of nitrate ion
quantification, while FRM sampling uses a Teflon filter for PM2.5 mass as a whole. The
Nylon filter limits the loss of nitrate in the form of nitric acid vapor which could
otherwise occur if the filter temperature rises above the temperature at the time of
collection, compared to the Teflon filter. The fine particle nitrate ion collected on nylon
and Teflon filters are assumed to be associated with ammonium ions, and for this analysis
ammonium is assumed to evaporate at the same rate as nitrate on the FRM filters15.
Hence, the nitrate ion and calculated ammonium nitrate concentrations reported by CSN
(and IMPROVE) sampling typically will be higher than the nitrate contribution to FRM
PM2.5 mass, particularly under warm ambient conditions. On the other hand, FRM
sampling may result in some water that is associated with nitrate being included in the
reported PM2.5 mass, while the nitrate mass reported by CSN (or IMPROVE) sampling
excludes all water. Continuous PM2.5 samplers employ a variety of methods for
measuring PM2.5 mass, with correspondingly different behaviors regarding retention/loss
15 EPA staff recognizes that fine particle nitrate may be in the form of calcium or sodium nitrate, but like
the IMPROVE program treats nitrate as ammonium nitrate.
3-20
-------
of nitrate. In this assessment's approach to estimating actual ambient concentrations and
PMio light extinction, the FRM measurement of nitrate is used in the calculation of the
concentration of organic carbonaceous material, but not in estimating ambient
concentrations of nitrate or PMio light extinction. The CSN-reported nitrate ion
concentration and corresponding ammonium nitrate mass is used for the latter purposes.
Sulfate: Unlike nitrate, sulfate is not subject to loss once collected by a filter, so the
sulfate ion mass reported by a CSN (or IMPROVE) sampler will be about the same as the
contribution of sulfate ion to the mass reported by FRM sampling. In FRM sampling,
sulfate ion may not be fully neutralized. When IMPROVE data are used to estimate
PMio light extinction, it is assumed that sulfate ion is fully neutralized. Even more
important than nitrate, FRM sampling results in water that is associated with sulfate being
included in the reported PM2.5 mass. While the water associated with the measured
sulfate ion is used in the calculation of the concentration of organic carbonaceous
material, it is not used in estimating ambient concentrations of sulfate or PMio light
extinction.
Elemental and Organic Carbon: Only the mass of carbon atoms is included in the
reported elemental carbon and organic carbon for a CSN (or IMPROVE) sampler. In
addition, the assignment of carbon atoms between the reported elemental and organic
amounts is dependent on the specifics of the two different thermo-optical analytical
methods used in the CSN vs. the IMPROVE network.16 Also, the quartz filter used to
quantify carbonaceous material in CSN and IMPROVE sampling both absorbs and loses
organic vapors during sampling, while the Teflon filter in a FRM sampler does not
absorb organic vapors (although PM on the filter may do so). Therefore, some method
other than direct measurement must be used to estimate the total mass concentration of
organic carbonaceous material in ambient air. The IMPROVE program adjusts for
absorption of vapors by subtracting a monthly average backup filter value, and then
applies a standard adjustment factor (1.4 in the original IMPROVE method) to the
remaining organic carbon measurement to estimate organic carbonaceous material. In
contrast, the standard reports from CSN sampling submitted to AQS do not include these
two adjustments, but it is routine for EPA staff to apply adjustments for the same
purpose, after reporting of CSN data to AQS. The latter are based on network-wide filter
field blanks and are judged as very approximate. For this assessment, the SANDWICH
approach to such adjustments (Frank, 2006) is used to estimate the organic mass through
a material balance of components measured on the CSN and FRM samplers.
Hourly PM^: The continuous instruments used for measuring hourly PM2.5 mass were
different among sites (as listed in Appendix A). None of the instrument types that
provided hourly data for this assessment, when averaged over 24 hours, exactly matches
either the measurement of PM2.5 mass from a FRM sampler or the sum-of-components
reportable from CSN sampling. Differences can arise because of differences in water
capture and retention, inconsistent absorption and loss of organic vapors and nitric acid
16 While CSN carbon sampling and analysis methods have recently been harmonized with IMPROVE
methods at many CSN sites, it was not until mid-2007 that the first 57 sites were using the harmonized methods.
Consequently, most of the elemental and organic carbon data used in this assessment were obtained with the original
CSN methods.
3-21
-------
vapor, etc. Furthermore, comparability between hourly and 24-hour integrated
measurements can only be made on a daily average basis. Consequently, the continuous
instruments providing data to this assessment can be assumed to have a range of
correlation performance versus the FRM. In light of these consistency issues, the hourly
data from the continuous instruments were taken to be most indicative of the relative
concentrations of PM2 5 from hour-to-hour, with less reliance on the absolute accuracy of
the continuous instruments.17
Taking into consideration the above information, EPA staff combined the four types of
available PM2.5 data in each study area using the following steps. Figure 3-5 provides a flow
chart to assist in understanding these steps.
1. The SANDWICH method (Frank, 2006) was used to subdivide the 24-hour PM2.5 mass
reported by the FRM for each day and site into sulfate (including associated ammonium
and residual water during filter equilibration and weighing), nitrate (including associated
ammonium, but not necessarily enough to fully neutralize the sulfate ion, and residual
water during filter weighing), elemental carbon, organic carbonaceous mass, and fine
soil/crustal mass. This is done using information from the CSN measurements, physical
models, and day-specific temperatures. The primary purpose of this SANDWICH step is
to estimate organic carbonaceous mass. Significantly, in the SANDWICH method, the
component referred to as organic carbonaceous mass is actually a residual whose value is
determined as the difference between the PM2 5 mass determined from weighing the FRM
filter and the sum of the estimated masses of the other four mass components as listed
above. Therefore, it is not necessary to adjust for organic carbon sampling artifacts or to
apply the 1.4 factor commonly used to estimate organic carbonaceous material from
IMPROVE measurements of organic carbon. The SANDWICH procedure did not
consider sea salt in the material balance, since this is generally a very small mass
constituent for the urban areas considered in this analysis. For the same reason, sea salt
was also not considered in the aerosol based light extinction algorithm.18
17 In 2006, EPA developed and promulgated criteria for approval of continuous PM2 5 samplers as "federal
equivalent methods". These criteria assure a minimum level of correlation between approved continuous
instruments and the FPJVI method, when data from both are expressed as 24-hour average concentrations. However,
in 2005-2007 no commercially available instruments were yet approved under those criteria.
18 After completion of the second draft UFVA, an error in the execution of the SANDWICH method was
discovered. The error had the effect of reducing the estimate of nitrate on the FPJVI filter for some sample days,
generally by no more than 3 or 4 ug/m3, and increasing the estimate of organic carbon mass on the FPJVI filter by the
same amount. The effect on estimates for PM10 light extinction was that light extinction had been overestimated for
some hours in some cities. The error has been corrected in this final version. The effect of the correction is in most
cases very small and visually imperceptible in graphics such as box plots.
3-22
-------
Figure 3-5. Sequence of Steps Used to Estimate Hourly PM2.s Components and PM10
Light Extinction
Consistent with FRM
FRM Data:
CMAQ: Diurnal
Profiles of PM2 5
Components
24-hour PM
Components by SANDWICH
2.5
Steps 2 and 3
Tentative 1-hour PM2.5 Components
and Sum-of-Components
Steps 5 and 6
1-hour PM2.5
Components Adjusted to be
Consistent With measured
1-hour PM2.5
Steps 7 and
1-hour Relative
Humidity Data
\ Estimates of 1-hour
! PM10-2.5
IMPROVE Light Extinction Algorithm
2. The CMAQ-derived monthly diurnal profiles for the sulfate, nitrate, elemental carbon,
organic carbon and fine soil/crustal components, like the examples for Detroit in Figure
3-4, were multiplied by the day-specific SANDWICH-based estimates of the 24-hour
average concentrations of these five PM2 5 components, to get day-specific hourly
estimates of these five components (including ammonium and water associated with
sulfate and nitrate ion).
3. The hourly concentrations of these five components (including ammonium and water
associated with sulfate and nitrate ion when the filter is weighed) were added together, to
get a sum-of-components estimate of hourly PM2.s mass for the day of the FRM
sampling.
3-23
-------
4. The hourly data from the continuous PM2.5 instrument on the day of the FRM sampling
were normalized by their 24-hour average, to get a diurnal profile. (Recall that days were
not used in this assessment if hourly PM2.5 mass data were missing for more than 25
percent of daylight hours.) This profile was applied to the 24-hour PM2 5 mass reported
by the FRM sampler, to get a preliminary, FRM-consistent estimate of hourly PM2.5 mass
for the day of the FRM sampling. This is straightforward when all 24 values of 1-hour
PM2.5 mass were available for the day. However, for some (but not many) days, some
values for continuously measured hourly PM2 5 mass were missing. In such cases, EPA
staff used only the hours with valid 1-hour PM2.5 mass values to develop the diurnal
profile and then applied the profile to the FRM value as just described. This keeps the
average of the valid 1-hour PM2 5 values equal to the 24-hour value from the FRM
sampler.
5. The two estimates of hourly PM2 5 mass from steps 3 and 4 were compared, hour-by-
hour. By virtue of the way they were derived, the averages of these estimates across all
24 hours of the day will necessarily be the same (and will be equal to the 24-hour FRM
measurement). However, while the diurnal pattern of these two estimates of the same
physical parameter should also be generally similar, it can be expected (and it is
observed) that the hourly measurements from the continuous PM2 5 instruments (after
adjustment to be consistent with the FRM data) have more hour-to-hour variability.
Figure 3-6 gives an example of this comparison, for one day for the Detroit study area.
3-24
-------
Figure 3-6. Example from Detroit Study Area.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
FRM adjusted Hourly PM2.5
Sum of diurnalized SANDWICHED species data using CMAQ profiles
Example comparison from the Detroit study area of hourly PM2.s mass on March 24, 2006 as
estimated by applying CMAQ-based diurnal profiles to SANDWICH estimates of 24-hour
component concentrations versus applying a diurnal profile derived from continuous PM2 5
measurements to FRM PM2.5 mass.
3-25
-------
Given that the continuous instrument is reacting to hour-specific local conditions that can
vary from hour-to-hour due to real variations in local emissions and dispersion/transport
conditions, while the CMAQ-based estimates contain much less specific information, the
diurnal pattern of PM2.5 mass observed by the continuous instrument (adjusted to be
consistent with the FRM value for 24-hour average PM^.s) was taken as more reliable.
Within each hour, the estimates of all five components from step 2 were increased or
decreased by a common percentage (referred to below as A; where the subscript /'
indicates the hour) so that the sum of the five components after this adjustment was equal
to the estimate of the hourly PM2.5 mass from step 4. The adjustment percentage varied
from hour-to-hour. Necessarily, in some hours the adjustment is an increase in the
concentrations of all components, and in other hours it is a decrease. While this
adjustment preserves the consistency between the 24 values of hourly PM2.5 mass and the
24-hour FRM mass, it can disturb the consistency between the daily average of hourly
estimates of PM2.5 components and the SANDWICH-based estimates of 24-hour average
component concentrations. This disturbance was generally small, because the
adjustments necessarily go in one direction for some hours and the other direction for
other hours. For example, for the particular day in Detroit used for illustration purposes
in Figure 6, the effect of this step was to cause a discrepancy of 3 percent between the
SANDWICH-based values of 24-hour sulfate concentration and the average of the 24
estimates of 1-hour sulfate concentrations (the positive percent indicates a higher
concentration in the result of this step than the SANDWICH-based value). The
discrepancies were 1, 1,2, and 2 percent for nitrate, elemental carbon, organic carbon,
and fine soil/crustal, respectively.
Each hourly estimate of sulfate concentration from step 6 (which includes estimates of
associated ammonium and particle bound water) was adjusted so that it excludes water
and reflects full neutralization and therefore is consistent with the reporting practices of
the IMPROVE program and the IMPROVE algorithm. This was done via these sub-
steps:
a. The 24-hour CSN value for the dry mass of sulfate ion (not SANDWICHed, no
ammonium or water) was multiplied by 1.375 to reflect an assumption of full
neutralization of dry sulfate mass.19
b. The ratio of this fully neutralized 24-hour sulfate mass to the SANDWICH-based
24-hour sulfate value was calculated.
c. This ratio was applied to each individual hour's sulfate concentration from step 6.
As in Step 6, it is possible for the 24 final hourly sulfate estimates to no longer be
exactly consistent with the 24-hour CSN sulfate measurement, both reported as fully
neutralized sulfate ion.
19 While it would have been possible to develop a more realistic estimate of partially neutralized sulfate, the
assumption of full neutralization was used to maintain consistency with the basis for the f(RH) term in the
IMPROVE algorithm.
3-26
-------
8. A similar adjustment as in step 7 (for sulfate) was made to each hour's nitrate
concentration from step 6, so that the estimate of hourly nitrate would reflect actual
atmospheric conditions and be consistent with the IMPROVE algorithm. However, the
ratio approach used in step 7(b) for sulfate could not be applied for nitrate, so this
adjustment had to be more complicated. Because in warm weather the FRM Teflon filter
does not retain nitrate, the initial FRM-consistent nitrate estimate derived by applying the
SANDWICH method to the FRM and CSN data can be zero. Such a zero value makes it
impossible to use the ratio approach in 7(b). Instead, the adjustment was made as
follows:
a. The 24-hour CSN value for nitrate ion (not SANDWICHed, no ammonium or
water) was multiplied by 1.29 to reflect an assumption of full neutralization by
ammonia.
b. This 24-hour value was then diurnalized using the CMAQ-based profile, similar
to step 2.
c. Each resulting hourly value of nitrate was further multiplied by the A factor from
step 6.
d. This new estimate of hourly nitrate was used to replace the initial nitrate value
that had resulted from step 6.
For cooler areas and days in which the 24-hour SANDWICH results include some nitrate,
the effect of these steps for nitrate are exactly the same as the effects of step 7 for sulfate
(except for the 1.29 vs. 1.375 neutralization factor). For warmer areas and days in which
the 24-hour SANDWICH results did not include any nitrate even though nitrate was
measured on the CSN Nylon filter, the effect of these steps is to assign the CSN nitrate to
each hour using a combination of the information in the CMAQ-based profiles and the
information provided by the continuous PM2.5 sampler. As in Step 6, it is possible for the
24 final hourly nitrate estimates to no longer be exactly consistent with the 24-hour CSN
nitrate measurement.
The net effect of these steps is believed by EPA staff to result in hourly PMio light
extinction estimates with the following features with respect to some of the complicating aspects
of PM sampling:
• The 24-hour average of the hourly nitrate concentrations used to estimate hourly PMio
light extinction agrees closely but not exactly with the 24-hour value provided by the
CSN sampling, and generally is higher than the contribution of nitrate to the FRM
measure of PM2.5 mass. In some mid-day hours in some areas, estimated hourly nitrate is
zero which is a more realistic approach than applying a 24-hour species mix to each hour.
3-27
-------
• The 24-hour average of the hourly organic carbonaceous material concentrations used to
estimate hourly PMio light extinction achieves FRM mass balance closure, taking into
account also the difference in nitrate and the possibly partial neutralization of sulfate ion
on the FRM filter. Because the Teflon filter used in FRM sampling is less subject to
positive artifacts for organic material, this approach sidesteps an area of uncertainty in the
IMPROVE sampling method. By relying on mass closure as the driving principle for
estimating organic material, it is not necessary to choose a multiplier to relate organic
carbon to organic carbonaceous material.20
• The 24-hour average of the hourly elemental carbon concentrations used to estimate
hourly PMio light extinction agrees closely but not exactly with the 24-hour value
provided by the CSN sampling, and with the contribution of elemental carbon to the
FRM measure of PM2.5 mass. Elemental carbon is generally defined by the thermal
optical transmission method used in CSN, rather than the thermal optical reflectance
method used by the IMPROVE network.
3.3.2 Hourly PMi0-2.s Concentrations
Three different paths were used to estimate hourly PMio-2.5 concentrations depending on
data availability, in the following order of preference:
1. When hourly data from a collocated PMio instruments were available at the continuous
PM2.5 site in a study area, PM2.5 was subtracted hour-by-hour from PMio. Negative
values were reset to zero. This was the approach most often used in Birmingham,
Detroit, Baltimore, and Philadelphia. This method should result in reliable estimates of
actual PMio-2.5 at the study site. (How well the study site represents the study area
generally, or the most visibility-impacted portions of the study area, is a separate issue.)
2. When collocated continuous PMio data were not available at the continuous PM2.5 site in
a study area, but continuous PMio data were available at another site in or near the same
study area, PMi0-2.5 was estimated by subtraction, implicitly assuming that the latter site
was also representative of PMio at the former site. This was the approach most often
used in Los Angeles, Phoenix, St. Louis, Atlanta, and New York. As a result, estimates
of PMio-2.5 for these areas could be affected by site-to-site differences. In particular, the
two sites in Los Angeles were a good distance apart, and the PMio site in Victorville may
represent influences from agricultural operations rather than typical urban influences. In
St. Louis, the PMio site may also have been influenced by particular local sources. In
both cases, very high estimates of hourly PMio-2.5 may not represent reality at the PM2.s
site, although they may be reasonable estimates for the PMio site.
20 In other work, EPA staff has observed that when applied to urban sampling data together with CSN
network-wide field blanks applied to reported OC measured concentrations, the multipliers that can be back-
calculated from the results of the SANDWICH method tend to be nearer to 1.4 than to the higher value used in the
new IMPROVE algorithm.
3-28
-------
3. If neither of the first two methods was possible, a regional average ratio of PMio-2.5 to
PM2 5 determined from an analysis of 24-hour data for the 2005 Staff Paper was applied
to hourly PM2.5 from the continuous instrument associated with the study area. This was
the approach used for all hours in Tacoma, Fresno, Salt Lake City, Dallas, Houston, and
Pittsburgh. With this approach, it is not possible for there to be any particularly high
estimates of hourly PMio-2.5.
The estimation of PMi0-2.5 was further complicated because some types of data were
missing for isolated hours in the 2005-2007 period. As result, even for a single study area more
than one method sometimes had to be used to estimate hourly PMio-2.5- Appendix A gives more
specifics about the estimation of hourly PMio-2.5 in each study area.
The three-path approach described here is similar to that used for the visibility analysis
reported in the 2005 Staff Paper. While the second and third paths involve the use of data and
assumptions that are not robust compared to the use of paired, collocated, same-method
continuous instruments or compared to the use of paired low-volume filter-based samplers, in
most areas and periods the contribution to PMi0 light extinction from the resulting PMio-2.5
concentrations was not large compared to the PMio light extinction contribution from PM2.5
components.
3.3.3 Hourly Relative Humidity Data
Hourly relative humidity (RH) data for each study area's primary monitoring site were
obtained hour-by-hour from the closest available non-missing relative humidity measurement, as
reported by either an air monitoring station reporting such data to AQS or a National Weather
Service (NWS) station. For the AQS RH data, parameter 62201 values were utilized. RH data
from both sources are expressed as percentages.
3.3.4 Calculation of Daylight 1-Hour PMio Light Extinction
Because the interest in this analysis is on visibility during daylight hours, EPA staff
applied a scheme to denote those hours that would be considered daylight hours. For simplicity,
all the days within each "season" in all study areas were considered to have the same daylight
hours.21 Table 3-5 shows the dividing times used to denote daylight hours for the study areas.
Unless otherwise stated, all subsequent discussion of the results refers only to the values of
parameters during these daylight hours.
The original IMPROVE algorithm was applied hour-by-hour to estimate PMio light
extinction in each study area for each daylight hour. When doing so, we capped the value of the
21 This simple approach does not account for the effects of the actual date within a three-month season,
latitude, or east-west position within a time zone on the actual local hours that are entirely daylight. Appendix I
examines the possible impact of this simplification, concluding that it is unlikely to affect later answers to policy
relevant questions.
3-29
-------
humidity adjustment factor in the IMPROVE algorithm ("f(RH)") at the value of 7.4 that it has
for a relative humidity of 95 percent. The effect of measurement errors in relative humidity at
values above 95 percent on the value of f(RH) and thus on reconstructed PMi0 light extinction is
considerable because of the highly nonlinear form of the function in that range. This creates
uncertainty as to the representativeness of the extinction values calculated with high values of
relative humidity.22 .
Table 3-5. Assumed Daylight Hours by Season (Local Standard Time)
First hour that is
entirely daylight
Last hour that is
entirely daylight
Number of
daylight hours
November-
January
8:00-9:00 AM
3: 00-4: 00PM
8
February-April
7:00-8:00 AM
5:00-6:00 PM
11
May-July
5:00-6:00 AM
6:00-7:00 PM
14
August-October
6:00-7:00 AM
5:00-6:00 PM
12
3.3.5 Exclusion of Hours with Relative Humidity Greater than 90 Percent from
Light Extinction NAAQS Scenarios and Most Results
As advised by CAS AC as part of its comments on the first public review draft of this
assessment, EPA staff considered whether to structure the PMio light extinction NAAQS
scenarios so that ambient data obtained during daylight hours in which relative humidity was
greater than 90 percent would play no role in the form of the NAAQS, i.e., so that those data
would not enter into the calculation of the design value. EPA staff obtained hourly
meteorological parameters from NWS monitoring sites near the 15 study sites (usually a major
airport), for 2005 through 2007, for all days in this period including days for which PM
observations to support estimates of PMio light extinction are not available. Using these data,
EPA staff compared the occurrence of liquid precipitation, hail, other frozen precipitation, fog,
mist, and smoke/haze during daylight hours with humidity greater than 90 percent and during all
other daylight hours.23 The first five of these conditions are generally considered natural causes
The IMPROVE program also caps the value of f(RH) at its value for a relative humidity of 95% when
reporting visibility in deciviews.
23 The "smoke/haze" category is not an original NWS reporting category. It is a combination of two
original NWS weather categories: smoke and haze. The explanation of these categories in the NWS documentation
does not allow EPA staff to be confident that these terms have distinct and clear meanings that are uniformly applied
across observation sites, so they have been combined in this presentation. As best EPA staff can determine, the
combined category reflects some mix of smoke from burning biomass, smoke from industrial processes, dust from
3-30
-------
of reduced visibility. Table 3-6 presents this comparison.24 The percentages of hours with each
of these five conditions individually are shown for the two sets of daylight hours.
wind storms, volcanic ash, and general urban haze. Also, the reported conditions may be at some distance from the
observation site.
24 Compared to the version of this table in the second external review draft, this version correctly omits
non-daylight hours which were inadvertently included in the earlier version, reports results for four study areas for
which results were missing in the earlier version, and separates "mist" from "smoke/haze" to reflect that "mist" is a
natural condition while "smoke/haze" is not always a natural condition.
3-31
-------
Table 3-6. Comparison of Meteorological Parameters for Daylight Hours with Relative Humidity Greater than 90 Percent and
Other Daylight Hours, During 2005 -2007
Study Area
Tacoma
Fresno
Los
Angeles
Phoenix
Salt Lake
City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Average
Daylight Hours with Relative Humidity <= 90%
Number
of
Hours
10,326
11,758
11,419
12,123
11,810
11,827
11,525
11,590
11,590
11,337
11,484
10,603
11,321
11,125
11,799
11,442
Percentage of Hours with Weather or Other
Condition
Liquid
Precip.
12%
3%
2%
1%
4%
4%
6%
5%
5%
5%
5%
5%
4%
4%
7%
5%
Hail
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
Other
Frozen
Precip.
0%
0%
0%
0%
2%
0%
0%
1%
0%
0%
3%
3%
1%
1%
1%
1%
Fog
0%
1%
0%
0%
1%
1%
1%
1%
1%
1%
1%
1%
2%
1%
1%
1%
Mist
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
Smoke/
Haze
4%
15%
8%
0%
4%
5%
6%
10%
9%
10%
9%
9%
12%
8%
10%
8%
Any
14%
17%
9%
1%
8%
8%
9%
14%
11%
13%
14%
14%
14%
11%
14%
11%
Daylight Hours with Relative Humidity > 90%
Number
of
Hours
1,756
342
713
43
304
223
645
583
486
867
676
1,261
858
878
397
669
Percentage of Hours with Weather or Other Condition
Liquid
Precip.
36%
25%
25%
67%
28%
68%
42%
56%
56%
50%
51%
46%
53%
47%
66%
48%
Hail
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
Other
Frozen
Precip.
1%
1%
0%
0%
40%
2%
0%
8%
0%
0%
16%
9%
5%
3%
8%
6%
Fog
10%
60%
12%
30%
42%
20%
25%
48%
41%
45%
39%
12%
38%
33%
48%
34%
Mist
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
0%
Smoke/
Haze
43%
65%
52%
40%
69%
82%
64%
82%
79%
81%
76%
72%
80%
64%
86%
69%
Any
63%
93%
73%
74%
85%
91%
75%
91%
86%
88%
92%
85%
90%
84%
96%
84%
3-32
-------
NWS observations of these conditions are instantaneous, and are generally made about
50 minutes after the hour. The relative humidity observations are made at the same time. It
should be noted that this analysis of the co-occurrence of high relative humidity and these five
conditions uses data from NWS sites other than the AQS sites that provided the data used to
estimate PMio light extinction. AQS sites could not be used for this analysis because they
generally do not report similar weather condition data.
The comparison for the 15 sites shows that in the set of hours with relative humidity
above 90 percent, the frequencies of liquid precipitation (rain), other frozen precipitation (snow
and sleet), or fog ranged as high as 68 percent, and were considerably higher for the same
condition than in the set of hours with lower relative humidity. The frequencies of hail and mist
were all less than 0.5 percent and thus too low for meaningful comparisons. Moreover, except in
Tacoma, the frequency of rain or fog at the observation moments during the hours with relative
humidity less than or equal to 90 percent was less than 8 percent. Also, a separate analysis (not
shown) indicated that rainy hours with lower relative humidity experience considerably less
accumulation than rainy hours with higher relative humidity. Based on this assessment, the 90%
relative humidity cutoff criteria is effective in that on average less than 6 percent of the daylight
hours are removed from consideration, yet those hours have on average about 10 times the
likelihood of rain, 6 times the likelihood of snow/sleet, and 34 times the likelihood of fog
compared to hours with 90% or less relative humidity.
Rain, snow/sleet, and fog cause a natural reduction in visibility, independent of PM
concentrations. To reduce the likelihood that a design value for a secondary PM NAAQS could
be affected by measurements made under natural weather conditions that reduce visibility, for
this assessment EPA staff eliminated from the design value definition any contribution from
PMio light extinction values that come from any daylight hours with relative humidity above 90
percent.25 Also, because PMio light extinction during such hours is not as likely to be the
primary cause of adverse effects on the public, all figures and tables in the body of this document
and in Appendices that present PMio light extinction values or statistics exclude values for such
hours (unless explicitly stated to include them), so that the patterns of PMio light extinction
during the remaining daylight hours can be seen clearly. Figures and tables that present PM
component concentrations and relative humidity values are based on all daylight hours, however.
More information on this topic can be found in Appendix G, which reports by study area
the percentages of daylight hours that were excluded from design values, the distribution of the
excluded hours by time of day, and the percentage of days that had one or more daylight hours
25 Another consideration is that instruments used to measure light extinction could be adversely affected if
allowed to operate without heating or other protective method (such as diffusion drying of incoming air) when
relative humidity is very high. If protected, however, the measured light scattering would not reflect actual ambient
conditions.
3-33
-------
eliminated. Appendix G also contains box plots which contrast the distributions of daylight 1-
hour PMio light extinction values (and maximum daily daylight 1-hour PMio light extinction, see
section 3.3.6) before and after this elimination step. The tile plots in Figure 3-12 also present
additional detailed information on the specific hours that had relative humidity values above 90
percent, and on the PMio light extinction values during those and other daylight hours.
3.3.6 Calculation of Daily Maximum 1-Hour PMio Light Extinction
Daily maximum 1-hour PMio light extinction is a statistic of interest in this assessment,
as briefly discussed in section 1.4.3. The daylight hour with the maximum value of PMio light
extinction and the corresponding PMio light extinction value were identified for each day for
each study area. As mentioned in section 3.2.1, days which were missing 1-hour PM2.5 values
for more than 25 percent of daylight hours were not used in this analysis. No further
completeness requirement for 1-hour data during a day was applied when selecting the daylight
hour with the maximum value of PMio light extinction.
3.4 SUMMARY OF RESULTS FOR CURRENT CONDITIONS
3.4.1 Levels of Estimated PM2.s, PM2.s Components, PMi0-2.s, and Relative Humidity
Figure 3-7 presents box-and-whisker plots to illustrate the distributions in each study area
of the estimates of 1-hour PM2.5 (the diurnalized FRM value, resulting from step 4 in section
3.4.1), PMio-2.5, and relative humidity over the entire 2005-2007 study period. In the plot for
each parameter, areas are ordered by longitude, to make it easier to see East-versus-West
regional differences. For these three parameters, the distributions are given for all the daylight 1-
hour estimates, including hours with relative humidity greater than 90 percent. Similar plots of
the daily maximum daylight 1-hour values of PM2.5 and PMio-2.5 concentrations and relative
humidity are available in Appendix B, as are plots of all daylight 1-hour values for each of the
PM2.s component species.26
From these plots we see that the distributions of PM2.5 generally trend toward higher
concentrations from West to East except for the two California urban locations which have PM2.s
concentrations more typical of eastern areas. The lowest median PM2.s concentrations are in
Tacoma, WA, and Phoenix, AZ. Median PMio-2.5 concentrations are highest in St. Louis, MO,
and Phoenix, AZ, and lower elsewhere. The highest outlier PMio-2.5 concentrations are in St.
Louis, MO, and Los Angeles, CA. Relative humidity is lowest for the western urban areas
except for Tacoma, WA, which is similar to the northeastern urban locations with respect to
26 In all box-and-whisker plots in this document, the box represents the 25th to 75th percentile range and the
whiskers represent the 10th and 90th percentile points of the data; individual data points below the 10th percentile and
above the 90th percentile are graphed as small circles (which may not all be visible because they may lie on top of
one another as is the case for relative humidity in Figure 3-7(c) because relative humidity is reported as an integer).
3-34
-------
Figure 3-7. Distribution of PM Parameters and Relative Humidity Across the 2005-2007
Period, by Study Area
(a) Estimates of 1-Hour PMi.s Mass, Based on Applying Continuous Instrument-based
Diurnal Profiles to 24-hour FRM PM2.5 Mass
PM 2.5 hourly (Daylight Hours)
10BS 3611 3029 988 3357 3019 1494 3063 3759
1 1
I
ii
1533 2816 181
8
I
-r T
1463 22M
3-35
-------
Figure 3-7. Distribution of PM Parameters and Relative Humidity Across the 2005-2007
Period, by Study Area, continued
(b) Estimates of 1-Hour PMio-2.s
PM Coarse hourly (Daylight Hours)
I
c
.£
1
§ -
J*
J>
3-36
-------
Figure 3-7. Distribution of PM Parameters and Relative Humidity Across the
2005-2007 Period, by Study Area, continued
(c) 1-Hour Relative Humidity
Relative Humidity hourly (Daylight Hours)
«
B
0)
a
2303
I
0* J®
.
-£«
/
f /
humidity. These hourly daylight PM concentration and relative humidity box and whisker plots
are consistent with our expectations based on regional 24-hour PM concentration values and
humidity climatology
3.4.2 Levels of Estimated PMi0 Light Extinction
Figure 3-8 presents box-and-whisker plots to illustrate the distributions of the estimates
of daylight 1-hour reconstructed PMio light extinction levels in each area in each year (excluding
hours with relative humidity greater than 90 percent). The distribution of (a) the daily maximum
1-hour values and (b) the individual 1-hour values are both shown. The horizontal dashed lines
in the plots represent the low, middle, and high CPLs for PMio light extinction as discussed in
section 2.6. These benchmarks for PMio light extinction are 64, 112, and 191 Mm"1,
corresponding to the benchmark VAQ values of 20 dv, 25 dv and 30 dv. Table 3-7 provides (a)
the percentages of days (across all of 2005-2007, unweighted) in which the daily maximum
3-37
-------
daylight 1-hour PMio light extinction level was greater than each of the three CPLs (excluding
hours with relative humidity greater than 90 percent), and (b) the similar percentage based on all
daylight hours (with the same exclusion).
As was also seen in the comparable PM2.5 concentration box and whisker plots in Figure
3-7, the high percentile hourly PMio light extinction values in Figure 3-8 tend to be higher in the
eastern urban areas and lower in the non-California western urban areas. The distributions of
maximum daily PMio light extinction values are higher (Figure 3-8a), as expected, than for all
hours (Figure 3-8b). Both Figure 3-8 and Table 3-7 indicate that all 15 urban areas have daily
maximum hourly PMio light extinctions that exceed even the highest of the CPLs some of the
time. Again, the non-California western urban locations have the lowest frequency of maximum
hourly PMio light extinction with values in excess of the high CPL for 8 percent or fewer of the
days. Except for the two Texas and the non-California western urban areas, all of the other
urban areas exceed that high CPL from about 20 percent to over 60 percent of the days. Based
on these estimated maximum hourly PMio light extinction estimates, all 15 of the urban areas
exceed the low CPL for about 40 percent to over 90 percent of the days. As noted in section
3.2.1, in 10 of the 15 study areas the study site used in this assessment is not the site in the study
area with the highest concentrations of PM2.5. Thus, these estimates may not characterize
visibility in the worst-visibility portion of each study area.
In the last review of the secondary PM NAAQS, the pattern of light extinction during the
day was of particular interest. To illustrate the distributions of 1-hour PMio light extinction
levels in specific daylight hours, Figure 3-9 shows the distributions of 1-hour PMio light
extinction across the entire three-year study period, individually for the study areas (excluding
hours with relative humidity greater than 90 percent). (Appendix E provides additional graphics
related to temporal/spatial patterns of light extinction.) These plots show that high PMio light
extinction can occur during any of the daylight hours, though for most of these urban areas the
morning hours have somewhat higher PMio light extinction than in the afternoon.27 Urban areas
without a pronounced preference for morning high PMio light extinction include Phoenix, AZ;
Salt Lake City, UT; Tacoma, WA; Fresno, CA; and Philadelphia, PA.
27 If hours with relative humidity greater than 90 percent were not eliminated, the tendency for higher PM10
light extinction in the morning hours would be stronger.
3-38
-------
Figure 3-8. Distributions of Estimated Daylight 1-Hour PM10 Light Extinction and
Maximum Daily Daylight 1-Hour PM10 Light Extinction (in Mm"1 units)
Across the 2005-2007 Period, by Study Area (Excluding Hours
with Relative Humidity Greater Than 90 Percent).
(a) Maximum Daily Values
Daily Maximum Light Extinction (Daylight Hours)
I
0
o
u
UJ
o
o
o
o
o
CO
O
o
o
o
N
O -
141
0
*
V -s«
=v J 1"
/ / /
3-39
-------
Figure 3-8. Distributions of Estimated Daylight 1-Hour PM10 Light Extinction and
Maximum Daily Daylight 1-Hour PM10 Light Extinction (in Mm"1 units)
Across the 2005-2007 Period, by Study Area (Excluding Hours with Relative
Humidity Greater Than 90 Percent), continued.
(b) Individual 1-Hour Values
Hourfy Extinction (Daylight Hours) RH<=90%
0 -
3511 3029
3357 3019
3063 3759 2471 1533
1873 1463 22S6
t $ t
* *
/
/
3-40
-------
Table 3-7. Percentage of Daily Maximum Hourly Values and Individual Hourly Values
of Daylight PM10 Light Extinction Exceeding CPLs (Excluding Hours
with Relative Humidity Greater Than 90 Percent).
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Average
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Average
Number of Days with
Estimates
109
324
300
86
306
273
148
289
349
279
141
277
181
143
225
229
Number of Daylight Hours
with Estimates
1087
3533
3048
988
3366
3043
1504
3096
3763
2507
1547
2842
1873
1468
2296
2398
Candidate Protection Level
64Mm1
112 Mm -1
191 Mm -1
(a) Percentage of Daily Maximum Hourly Values
Exceeding CPL
52
76
90
42
44
80
79
98
89
91
87
85
80
86
83
77
22
52
83
7
17
41
45
78
64
75
68
57
50
63
59
52
4
29
61
1
8
10
11
40
34
31
43
26
23
31
28
25
(b) Percentage of Individual Daylight Hours
Exceeding CPL
14
41
68
11
17
33
35
66
57
60
62
53
55
55
53
45
4
20
42
1
7
10
8
36
25
28
36
25
24
28
28
21
1
10
19
0
3
2
1
11
8
5
14
7
7
9
9
7
3-41
-------
Figure 3-9. Distributions of 1-Hour PM10 Light Extinction Levels by Daylight Hour Across the 2005-2007 Period, by Study
Area (Excluding Hours with Relative Humidity Greater Than 90 Percent).
Pittsburgh. PA
Salt Lake City. UT
ac orna. WA
* J 1
* * W
4 I 4. I J
i 4 + » i -
Los Angeles. CA
New York. NY
Philadelphia. PA
tH
! ft i £ 1! i
it
Baltimore. MD
Birmingham, AL
ii
9;
$
05 06 07 OS 09 10 11 12 13 14 15 16 17 18 05 06 07 OS 09 10 11 12 13 14 15 16 17 13 05 06 07 03 09 10 11 12 13 14 15 16 17 18 05 06 07 OS 09 10 11 12 13 14 15 16 17 18 05 06 07 08 09 10 11 12 13 14 15 16 17 18
3-42
-------
3.4.3 Patterns of Relative Humidity and Relationship between Relative Humidity
and PMio Light Extinction
Figure 3-9 shows the distribution of relative humidity values at each daylight hour, for
each study area across 2005-2007 (excluding hours with relative humidity greater than 90
percent).28 As expected, in every area relative humidity is lowest in the early afternoon, typically
the warmest part of the day. Relative humidity is most similar across areas in the early
afternoon, as observed in the 2005 Staff Paper. However, even in this period there are notable
differences among areas. This variation was not as evident in the information presented in the
2005 Staff Paper because only regionally averaged information was presented. In all areas, there
is considerable variation in hour-specific relative humidity during the three-year period.
To allow closer inspection of the relationship between PMio light extinction values and
relative humidity values, Figure 3-10 is a scatter plot of actual 1-hour relative humidity and 1-
hour reconstructed PMio light extinction (excluding hours with relative humidity greater than 90
percent). Horizontal lines are included in each of the individual plots corresponding to the three
benchmarks for PMio light extinction and a vertical line in each for the 90 percent relative
humidity cutoff. There are many instances with PMio light extinction greater than the CPLs
when relative humidity is 90 percent or lower. Notice that in Figure 3-10 there also are plenty of
high humidity conditions for each urban area that correspond to low PMio light extinction values.
This is because humid air does not by itself contribute to light extinction. Particles composed of
material that absorbs water in high relative humidity conditions (e.g., sulfate and nitrate PM)
swell to larger solution droplets that scatter more light than their smaller dry particle counterparts
in a less humid environment. The magnitude of the relative humidity effect on light extinction
depends directly on the concentration of these hygroscopic PM components. (Figure 3-10
reveals skips in reported relative humidity values for some but not all the study areas. This is a
result of calculations of relative humidity from dry and wet bulb temperatures reported to the
nearest whole Celsius degree.)
Similar information on diurnal patterns but broken out by season is given in Appendix E.
3-43
-------
Figure 3-10. Distributions of 1-Hour Relative Humidity Levels by Daylight Hour (X-axis) Across the 2005-2007 Period, by
Study Area (Excluding Hours with Relative Humidity Greater Than 90 Percent).
Pittsburgh. PA
Salt Lake City, UT
New York. NY
Philadelphia. PA
<1>
i
cc
Atlanta. GA
Birmingham, AL
Detroit, Ml
05 06 07 08 09 10 11 12 13 14 15 16 17 13 05 06 07 08 09 10 11 12 13 14 15 16 17 18 05 06 07 08 09 10 11 12 13 14 15 16 17 18 05 06 07 08 09 10 11 12 13 14 15 16 17 18 05 06 07 08 09 10 11 12 13 14 15 16 17 18
3-44
-------
Figure 3-11. Scatter Plot of Daylight 1-Hour Relative Humidity (Percent) vs. Reconstructed PM10 Light Extinction (Mm *)
Across the 2005-2007 Period, by Study Area (Excluding Hours with Relative Humidity Greater Than 90 Percent).
20 40 GO SO
0 20 40 60
Phoenix, AZ
Pittsburgh. PA
Salt lake City. UT
St. Louis. It
Tacoma. VVA
Fresno, CA
Houston. TX
Los Angeles. CA
New York. NY
Philadelphia. PA
LU
£
Atlanta. GA
Baltimore. MD
Birmingham. AL
Dallas. TX
20 40 60 30
20 40 60 8
Percent Relative Humidity
0 20 40 60
3-45
-------
3.4.4 Tile Plots of Hourly PMio Light Extinction
Figure 3-12 consists of "tile plots" that show the estimated levels of 1-hour PMio light
extinction for each daylight hour for each study area. These plots assist in understanding the
times of the year and hours of the day in which high relative humidity and high PMio light
extinction occur, both separately and together.
Time runs horizontally with each row of tiles representing a single day from midnight
(left site) to midnight (right side), and vertically from January (top) to December (bottom). Each
tile represents one hour of the year for which data to estimate PMio light extinction were
sufficient. Sites with 1:3 speciation sampling have more (and smaller) tiles than sites with 1:6
speciation sampling. The tick marks on the vertical axis identify the first available sample day of
each month identified by its month number.
PMio light extinction is presented in terms of four ranges or bins defined by the two
intervals between the three CPLs, a bin above the high CPL, and a bin below the low CPL. For
the hours with relative humidity of 90 percent and below (referred to as "Low RH bext" in the
figure legend), shades of green are used to indicate the CPL range. Contrasting blue color scales
are used for the tiles representing hours with relative humidity greater than 90 percent (referred
to as "High RH bext" in the shading legend), so that the hours excluded from the PM NAAQS
scenarios (see section 3.3.5 and Chapter 4) can be distinguished. Hours with missing PM2.5 data
from the continuous instrument have no estimates of PMio light extinction and are white. Such
cases are rare, following the prior complete exclusion of days in which more than 25 percent of
daylight hours were missing such data.
Note that for Tacoma and Phoenix there are plots for only two years because the third
year did not have suitable data, and for Phoenix and Houston only 9 months are shown for one of
the available years because suitable data were not available for the remaining quarter (the
available 9 months of results are stretched over the same vertical distance as the 12 months in the
other cases).
One observation that can be made in looking at these tile plots is that in very many cases,
days which have one or more hours with high PMio light extinction excluded because of high
relative humidity have other hours with high PMio light extinction which are not excluded.
Although none of the PMio light extinction NAAQS scenarios considered in Chapter 4
are based on averaging periods longer than one hour, these tile plots can be used to get a rough
sense of whether hours with high PMio light extinction tend to be isolated, such that average
values over several hours would be considerably lower, or tend to occur together, such that a
longer averaging period would produce roughly the same design value. A number of the eastern
urban areas have numerous day-long haze episodes throughout the year (e.g. St. Louis, Detroit,
Pittsburgh, Philadelphia and New York) or seasonally (e.g. Fresno and Salt Lake City, in the
3-46
-------
winter, and Los Angeles and Atlanta in the summer). Some of the urban areas have morning
haze levels that diminish later in the day on a year-around basis (e.g. Dallas) or seasonally (e.g.
Los Angeles, Birmingham and Atlanta in winter and Tacoma, Fresno, and St. Louis in the
summer). This type of information may be useful in this regard during the subsequent
preparation of the final Policy Assessment Document.
3-47
-------
Figure 3-12. Tile Plots of Hourly PM10 Light Extinction
Tacoma, WA
2005
2006
2007
4 -
5 -
7 -
8 -
9 -
10 -
11 -
12 -
5 -
6 -
7 -
8 -
9 -
10 -
11 -
12 -
.
Low RH bext
n
Inf
190
111
64
High RH bext
Inf
130
111
64
3-48
-------
Figure 3-12. Tile Plots of Hourly PM10 Light Extinction, continued
Fresno, CA
2005
2006
2007
1 -1
2 -
3 -
8 -
9 -
10 -
11 -
12 -
12
Hour
1 n
2 -
3 -
4 -
5 -
8 -
9 -
10 -
11 -
12 -
12
Hour
2 -
3 -
4 -
5 -
9 -
10 -
11 -
12 -
12
Hour
!
18
~~l
24
Low RH bext
n
Inf
180
111
64
High RH bext
Inf
190
111
64
3-49
-------
Figure 3-12. Tile Plots of Hourly PM10 Light Extinction, continued
Los Angeles, CA
2005
2006
2007
1 -
2 -
3 -
6 -
7 -
8 -
9 -
10 -
11 -
12 -
3 -
4 -
5 -
10 -
11 -
12 -
7 12
Hour
2-
3 -
6 -
7 -
8 -
9 -
10 -
11 -
12 -
18 24
12
Hour
!
18
~~l
24
Low RH bext
n
Inf
180
111
64
High RH bext
Inf
190
111
64
3-50
-------
Figure 3-12. Tile Plots of Hourly PM10 Light Extinction, continued
Phoenix, AZ
4 -I
5 1
6 -
7-
9 -
10 -
11 -
12 -
2005
I
I
JJ
1 I
r,
•
L.
1 -
2 -
3 -
4 -
5-
6 -
7 -
8 -
9 -
10 -
11 -
12 -
2006
».
^^m m
*"
•
•
• •
.
«
1
2007
Low RH bext
n
Inf
190
111
64
0
High RH bext
7 12 18 24
Hour
7 12 18 24
Hour
3-51
-------
Figure 3-12. Tile Plots of Hourly PM10 Light Extinction, continued
Salt Lake City, UT
2005
2006
2007
1 -,
2 -
7 -
9 -
10 -
11 -
12 -
12
Hour
1
2
3 -
4 -
6 -
7 -
8-
10 -
11 -
12 -
12
Hour
1
2
3 -
4 -
5 -
6
8
10 -
11 -
12 -
12
Hour
!
18
Low RH bext
n
Inf
180
111
64
High RH bext
Inf
190
111
64
~~l
24
3-52
-------
Figure 3-12. Tile Plots of Hourly PM10 Light Extinction, continued
Dallas, TX
2005
2006
2007
1 n
5 -
6 -
7 -
9 -
10 -
11 -
12 -
r.
•
12
Hour
7 -
10 -
11 -
12 -
12
Hour
2 -
3 -
6 -
7 -
9 -
10-
11 -
12 -
Low RH bext
n
Inf
180
111
64
High RH bext
Inf
190
111
64
3-53
-------
Figure 3-12. Tile Plots of Hourly PM10 Light Extinction, continued
Houston, TX
2005
2006
2007
2 -
3 -
7 -
10 -
11 -
12 -
12
Hour
2 -
3-
5 -
10 -
11 -
12 -
18 24
12
Hour
2 -
3 -
4 -
5 -
6 -
7 -
.
Low RH bext
n
Inf
180
111
64
High RH bext
Inf
190
111
64
Hour
3-54
-------
Figure 3-12. Tile Plots of Hourly PM10 Light Extinction, continued
St. Louis, IL
2005
2006
2007
1 -i
6 -
7 -
8 -
9 -
10 -
11 -
12 -
12
Hour
1 -i
2 -
3 -
7 -
10 -
11 -
12 -
cr
12
Hour
2 -
3 -
4 -
7-
9 -
10 -
11 -
12 -
Low RH bext
n
Inf
180
111
64
High RH bext
Inf
190
111
64
3-55
-------
Figure 3-12. Tile Plots of Hourly PM10 Light Extinction, continued
Birmingham, AL
2005
2006
2007
2
3
4 -
5 -
6
7
9
10
11 -
12
2 -
3
4
5 -
6
7
8 -
9
10
11
12
Hour
12
Hour
4 -
5 -
6 -
7 -
8 -
9 -
10 -
11 -
12 -
12
Hour
!
18
~~l
24
Low RH bext
n
Inf
180
111
64
High RH bext
Inf
190
111
64
3-56
-------
Figure 3-12. Tile Plots of Hourly PM10 Light Extinction, continued
Atlanta, GA
2005
2006
2007
1 -
2 -
3 -
4 -
7 -
9 -
10 -
11 -
12 -
1 n
2 -
3 -
4 -
5 -
7 -
8 -
9 -
10 -
11 -
12 -
2
3
4 -
5
6
7 -
9 -
10 -
11 -
12 -
r
Low RH bext
Hour
12
Hour
12
Hour
!
18
n
Inf
180
111
64
High RH bext
Inf
190
111
64
3-57
-------
Figure 3-12. Tile Plots of Hourly PM10 Light Extinction, continued
Detroit,
2005
2006
2007
3 -
4 -
5 -
6 -
7 -
8 -
9 -
10 -
11 -
12 -
3 -
7 -
8 -
9 -
10 -
11 -
12 -
IL • •
^M • • •
-
1 -i
2
3
4 -
5
6
7 -
8
9
10-
11 -
12 -
t
Low RH bext
n
Inf
180
111
64
High RH bext
Inf
190
111
64
12
Hour
18 24
3-58
-------
Figure 3-12. Tile Plots of Hourly PM10 Light Extinction, continued
Pittsburgh, PA
2005
2006
2007
1 -i
7 -
8 -
9 -
10 -
11 -
12 -
3 -
4 -
5 -
8 -
9 -
10 -
11 -
12 -
12
Hour
2 -
3 -
4 -
5 -
6 -
7 -
9 -
10 -
11 -
12 -
12
Hour
!
18
~~l
24
Low RH bext
n
Inf
180
111
64
High RH bext
Inf
190
111
64
3-59
-------
Figure 3-12. Tile Plots of Hourly PM10 Light Extinction, continued
Baltimore, MD
2005
2006
2007
2 -
3 -
4 -
5 -
6 -
7 -
9 -
10 -
11 -
12 -
12
Hour
1
2
3
4
5 -
6 -
7 -
10 -
11 -
12 -
12
Hour
2
3
4
5
6 -
7 -
9 -
10 -
12 -
Low RH bext
n
Inf
180
111
64
High RH bext
Inf
190
111
64
3-60
-------
Figure 3-12. Tile Plots of Hourly PM10 Light Extinction, continued
Philadelphia, PA
1 -•
2 -
3 -
4 -
5
6
7
8
9 -
10 -
11
12
2005
2006
2007
12
Hour
1 -,
2 -
3 -
4 -
5 -
6 -
7 -
11 -
12 -
7 12
Hour
2 -
3 -
4 -
7 -
8-
9 -
10 -
11 -
12 -
^r
i
12
Hour
i
18
~~l
24
Low RH bext
n
Inf
180
111
64
High RH bext
Inf
190
111
64
3-61
-------
Figure 3-12. Tile Plots of Hourly PM10 Light Extinction, continued
New York, NY
2005
2006
2007
12 -
2 -
3 -
4 -
5 -
6 -
7 _
8 -
9 -
10 -
11 -
12 -
2 -
3 -
4 -
5 -
6 -
7 -
8 -
9 -
10 -
11 -
12 -
Low RH bext
n
Inf
190
111
64
High RH bext
Inf
190
111
64
3-62
-------
3.4.5 Extinction Budgets for High PMio Light Extinction Conditions
An extinction budget for a single period shows the contribution that each PM component
makes to PMio light extinction via the additive terms of the IMPROVE algorithm. It can be
expected that the pattern in the extinction budgets will vary by time of year and by study area.
Examination of extinction budgets allows initial insights into what pollutants cause poor urban
visibility and what emission reduction approach may be most effective in reducing PMio light
extinction.
Figure 3-13 presents (a) day-specific maximum daylight 1-hour PMio light extinction
budgets for the 10 percent of the days in each study area that have the highest daily maximum 1-
hour PMio light extinction levels (excluding hours with relative humidity greater than 90
percent), and (b) light extinction budgets for the greatest 2 percent of all individual daylight
hours with the same relative humidity restrictions. The day and hour of each hourly budget are
indicated on the horizontal axis, and the hours are arranged chronologically. Note that the
vertical scale differs from city-to-city, to accommodate the wide variation in PMio light
extinction values.
Since there is an annual average of about 10 daylight hours per day 29, there are
approximately twice the number of hours (i.e., bars in the plots) included in the top 2% of all
daylight hours form compared to the number of hours in the top 10% of the daily maximum 1-
hour form. The rationale for pairing the top 10% of the maximum daily 1-hour PMio light
extinction form with the top 2% of all hours PMio light extinction form was the similarity of the
design values for the 90th and 98th percentiles for each form, respectively,, as discussed in
Chapter 4 (see Table 4-2 and Figure 4-1). In each of these plots the height of the shortest bar is
the PMio light extinction design value associated with the selected form (e.g., the smallest value
in the top 2% of all hours corresponds to the design value of the 98th percentile of all hours
form). The largest light extinction value in each pair of plots is identical, representing the single
largest daylight hour value for the city. As a result, the ranges of bar heights for each city's pair
of plots are approximately the same.
The paired extinction budget plots in Figure 3-13 provide a means to examine the
similarities and differences between the PM components that contribute to the daylight hours
with the greatest PMio light extinction as identified by the two forms. Though for each city there
are twice the number of hours selected by the 2% of all hours form compared to the 10% of
maximum daily form, the relative component contributions are generally quite similar. Much of
the reason for this similarity of the extinction budgets between forms has to do with the
selections of the same hours by both forms and having the additional hours for the 2% form
Daylight hours are determined for this assessment as described in section 3.3.4 and Appendix I.
3-63
-------
coming as multiple (often consecutive) hours in some of the same days that contained the top
10% maximum daily values. These multiple hours generally have similar relative composition.
For example notice that among the Tacoma top 2% of all hours, there are four consecutive hours
on November 8, 2007, which includes the largest hourly daylight PMio light extinction in the
Tacoma dataset.
3-64
-------
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-Hour PMio Light Extinction and for the Top 2 Percent of Individual
Daylight Hours
Tacoma
(a) Top 10 Percent of Days for Maximum Daily 1-Hour PM10 Light Extinction
Top 10% Daylight Hour Daily Maxima (RH<=90%)- Taeoma. WA
(b) Top 2 Percent of Individual Daylight Hours
Top 2* OiykjM Hnn (RH<40%| TaoMa V#
4. %. *_ V %. % %. >_ %
3-65
-------
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-Hour PM10 Light Extinction and for the Top 2 Percent of Individual
Daylight Hours, continued
Fresno
(a) Top 10 Percent of Days for Maximum Daily 1-Hour PM10 Light Extinction
8 -
Top 10% DayllgM Hour Daily Maxima (RH<=90%): Fresno. CA
(b) Top 2 Percent of Individual Daylight Hours
fcp2*Oay«9MHoire(RH«90*| Fresno, CA
^'y'-^^','^'^^
• EC • N03 D S04 D OCM Q PMc
3-66
-------
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-Hour PM10 Light Extinction and for the Top 2 Percent of Individual
Daylight Hours, continued
Los Angeles
(a) Top 10 Percent of Days for Maximum Daily 1-Hour PM10 Light Extinction
Top 10% Daylight Hour Dally Maxima (RH<=90%) Los Angeles. CA
8 .
(b) Top 2 Percent of Individual Daylight Hours
Top 2% DaytgN Han |RH<=90%) Los AnjelM CA
3-67
-------
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-Hour PM10 Light Extinction and for the Top 2 Percent of Individual
Daylight Hours, continued
Phoenix
(a) Top 10 Percent of Days for Maximum Daily 1-Hour PM10 Light Extinction
Top 10% Daylighl Hour Daily Maxima (RH<=90%): Phoenix, AZ
(b) Top 2 Percent of Individual Daylight Hours
3-68
-------
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-Hour PM10 Light Extinction and for the Top 2 Percent of Individual
Daylight Hours, continued
Salt Lake City
(a) Top 10 Percent of Days for Maximum Daily 1-Hour PM10 Light Extinction
Top 10* Daylight Hour Daily Maxima (RH<==90%) Salt lake Cily UT
(b) Top 2 Percent of Individual Daylight Hours
Top 2% Da»V» Hun |RH<*90*| Sal Ute C(». UT
8
%,;v\v*w*"V^vv>»\^"<•'"' V'^*'>*''"' ^*'\ V' ^ *'*' '"••-^"'v' ^^"'"' *"' *«* * *, *
. j 4 V* *' '* 6 * * * '* ~'f '* % 4 VJ 'J 4 4 4. 4 v* 4. -i '•* * * * * -V. 4 * * * * * * * * * 4 % * 4 % % « > '>. 'A
*'* * '/ V4 V '/ * * 4'4*»* ft''/ * '» '.>%%% % »V» 6 'r 4 4 *»»»'»'* * 4 V * %'4'VV4'4'4'» 4 4 'v> 'j ', '4 '* '*''i 't 'r '4''. »'V<""
• sol • EC • K03 0 S04 B OCM G PMc
3-69
-------
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-Hour PM10 Light Extinction and for the Top 2 Percent of Individual
Daylight Hours, continued
Dallas
(a) Top 10 Percent of Days for Maximum Daily 1-Hour PM10 Light Extinction
Top 10% Daylight Hour Daily Maxima (RH^90%>: Dallas. TX
soil • EC • NO3 P SO4 a OCM P PMc
(b) Top 2 Percent of Individual Daylight Hours
Top2%D3y»9«H««(RH<=S
-------
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-Hour PM10 Light Extinction and for the Top 2 Percent of Individual
Daylight Hours, continued
Houston
(a) Top 10 Percent of Days for Maximum Daily 1-Hour PM10 Light Extinction
Top 10% Daylighl Hour Daily Maxima
-------
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-Hour PM10 Light Extinction and for the Top 2 Percent of Individual
Daylight Hours, continued
30
St. Louis
(a) Top 10 Percent of Days for Maximum Daily 1-Hour PM10 Light Extinction
Top 10% Daylight Hour Daily Maxima (RH«=90%): SI. Louis, IL
soil • EC • NO3 n SO4 a OCM a PMc
(b) Top 2 Percent of Individual Daylight Hours
tip 2% Daylight Hours (RH<=9(»| St Lous. IL
.
% 4
'° See footnote 10 above regarding concerns with respect to the St. Louis results.
3-72
-------
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-Hour PM10 Light Extinction and for the Top 2 Percent of Individual
Daylight Hours, continued
Birmingham
(a) Top 10 Percent of Days for Maximum Daily 1-Hour PM10 Light Extinction
Top 10% Oaylighl Hour Dally Maxima (RHe=90%> Birmingham. AL
• EC • M03 D S04 • OCM D PMc[
(b) Top 2 Percent of Individual Daylight Hours
Top 2% OaytgK Hours (»H<=80*| Brnnnslom AL
3-73
-------
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-Hour PM10 Light Extinction and for the Top 2 Percent of Individual
Daylight Hours, continued
Atlanta
(a) Top 10 Percent of Days for Maximum Daily 1-Hour PM10 Light Extinction
Top 10% Daylight Hour Daily Maxima (RH<=90*) Atlanta. GA
VX XX VXXXXX X
soil • EC • N03 Q S04 • OCM D PMc
(b) Top 2 Percent of Individual Daylight Hours
Top 2% Dsylgrc Hotn i RH<=SM) Atlanta, GA
3-74
-------
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-Hour PM10 Light Extinction and for the Top 2 Percent of Individual
Daylight Hours, continued
Detroit
(a) Top 10 Percent of Days for Maximum Daily 1-Hour PM10 Light Extinction
8
Top 10% Daylight Hour Daily Maxima (RH<=90%)' Detroit, Ml
(b) Top 2 Percent of Individual Daylight Hours
Top2% DayligN Hours (RH«smi DslroJ. Ml
3-75
-------
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-Hour PM10 Light Extinction and for the Top 2 Percent of Individual
Daylight Hours, continued
Baltimore
(a) Top 10 Percent of Days for Maximum Daily 1-Hour PM10 Light Extinction
Top 10% Daylight Hour Daily Maxima (RH<=90%): Baltimore. MO
(b) Top 2 Percent of Individual Daylight Hours
Top 2% OaytgM Houre (RH<*90%) BaMoie, MO
3-76
-------
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-Hour PM10 Light Extinction and for the Top 2 Percent of Individual
Daylight Hours, continued
Pittsburgh
(a) Top 10 Percent of Days for Maximum Daily 1-Hour PM10 Light Extinction
Top 101t Daylighl Hour Daily Maxima (RH*=90%): Pittsburgh. PA
toll • EC • N03 D S04 a OCM LJ PMc
S
(b) Top 2 Percent of Individual Daylight Hours
3-77
-------
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-Hour PM10 Light Extinction and for the Top 2 Percent of Individual
Daylight Hours, continued
Philadelphia
(a) Top 10 Percent of Days for Maximum Daily 1-Hour PM10 Light Extinction
Top 10% Daylight Hour Daily Maxima {RHs=90%}: Phrladelphia, PA
soil • EC • N03 n SO4 m OCM a PMC
(b) Top 2 Percent of Individual Daylight Hours
3-78
-------
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-Hour PM10 Light Extinction and for the Top 2 Percent of Individual
Daylight Hours, continued
New York
(a) Top 10 Percent of Days for Maximum Daily 1-Hour PM10 Light Extinction
Top 10% Daylighl Hour Dally Maxima (RH"90*>: New York. NY
soil • EC • NO3 n SO4 n OCM u PMc
(b) Top 2 Percent of Individual Daylight Hours
Tdp Vk 09«M HOUS(HH<=«I%) Nw 1at_ M
3-79
-------
Table 3-8 shows the numbers of common and unique days selected across the two forms
as well as the numbers of days selected by each form. All of the cities have more common than
unique day selected by the two forms. Salt Lake City has the highest fraction of unique days, so
it is likely to have greater differences in their PM component composition selected by the two
forms. For example, the 2% of all hours form for Salt Lake City selected only one hour of one of
the days when the PM carbonaceous components was the major contributor compared with the 4
hours with high PM carbonaceous components selected by the top 10% of maximum daily form
(see Figure 3-13 for Salt Lake City). By comparison, for Salt Lake City most of the multiple
hours in single days had high contributions from PM nitrate. The overall effects of these
differences for Salt Lake City are more easily seen by viewing the average extinction budgets
using the two forms by city as shown in Figure 3-14. The 2% of all daylight hours form has a
greater contribution to light extinction by PM nitrate and a larger average light extinction than
for the 10% of maximum daily 1-hour form. Differences between the average extinction budgets
for the other urban areas are much smaller than for Salt Lake City.
Table 3-8. The Numbers of Common and Unique Days Selected for Each of the 15
Urban Areas by the Top 10% of Daily Maximum and the Top 2% of all
Hours Form. (Also shown are the numbers of days selected for each form.)
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Baltimore
Pittsburgh
Philadelphia
New York City
10% vs. 2%
Common
11
25
29
9
16
25
14
29
35
27
13
13
28
14
22
Unique
3
7
9
6
15
2
11
4
15
6
1
5
3
0
1
Number of Days
10%
11
32
30
9
31
27
14
29
35
27
14
18
28
14
23
2%
14
25
38
14
16
25
23
33
47
33
13
12
31
14
23
3-80
-------
Figure 3-14. Average PMi0 Light Extinction Budgets for the 15 Cities for Hours in the (a)
ToplO Percent of the Maximum Daily PMio Light Extinction and (b) Top 2 Percent
of all Daylight Hours of PMio Light Extinction.
(a)
Average of the Top 10% Daily Maxima Light Extinction Vak
•oil • EC • N03 n S04 a OCM n PMC
(b)
Average of Ihe Top 2% Hours Light Extinction Values by Area
The patterns of results for individual selected hours as shown in Figure 3-13 and for city
averages as shown in Figure 3-14 are generally as expected in light of emissions and climate
differences among study areas. Except for the PM2.5 soil component, each of the components of
PMio light extinction is a major contributor to extreme light extinction events at some time and
location. In the West, carbonaceous PM2.5 (i.e., organic mass and elemental carbon), nitrate,
and/or coarse mass (especially in Phoenix) tend to be most responsible for these high haze hours.
In the East it tends to be sulfate, nitrate, and the carbonaceous PM2.5 components that are the
large contributors to PMio light extinction. From the sample period dates we can determine the
seasonal variations in major components. Nitrate and carbonaceous PM2.5 contribute more to the
3-81
-------
extreme PMio light extinction periods during winter, while sulfate contributes more in the
summer. In many of the more northerly eastern urban areas, a combination of sulfate and nitrate
contributes to high PMio light extinction year-round.
Looking at individual urban areas, the following are some highlights:
• Tacoma has its highest PMio light extinction hours in the colder months and
primarily due to carbonaceous PM2.5 components. Because coarse PM was
estimated by applying a regional factor to the local PM2.5 mass value, it would not
have been possible for the results to indicate a significant coarse PM contribution
to PMio light extinction even if one existed at this site. However, from what EPA
staff know of the area, it is unlikely that there is a significant contribution from
coarse PM.
• Extreme haze hours in the two California urban areas are primarily caused by
high nitrate PM2.5, though Los Angeles has two extreme hours associated with
coarse PM and several other hours with moderate contribution from coarse PM.
Recall that estimates of coarse PM in Los Angeles are based in part on hourly
PMio measurements in Victorville, and may not represent coarse PM at the PM2.5
mass and speciation site in Rubidoux or in the larger South Coast Basin. Also,
such high coarse PM values may indicate influence from exceptional winds in
Victorville. Figure B-l(b) in Appendix B shows that several other days with high
daily maximum PM coarse concentrations had concentrations only about 60
percent or less than on the two days appearing in Figure 3-13; the fact that these
other days do not appear among the top 10 percent indicates that other
contributors to PMio light extinction were low on those days. Whether or not the
PMio measurements in Victorville represent the PM2.5 mass and speciation site in
Rubidoux, it can be concluded that nitrate and to a lesser extent sulfate dominate
PMio light extinction on the days likely to be above the CPLs. Because coarse
PM for Fresno was estimated by applying a regional factor to the local PM2.5 mass
value, it would not have been possible for the results to indicate a significant
coarse PM contribution to PMio light extinction even if one existed at the Fresno
site. However, given the presence of agricultural operations and occasional high
winds in the San Joaquin Valley, the possibility of a significant contribution from
coarse PM in some hours cannot not be ruled out.
• Phoenix is unique among the 15 urban areas in having most of its extreme PMio
light extinction caused by coarse PM, though there are a few top-10-percent days
where the maximum hourly haze is dominated by carbonaceous, sulfate, and
nitrate PM2 5. Unlike for Los Angeles, this domination by coarse PM is no doubt
3-82
-------
correct. PMio measurements for Phoenix come from a site near the center of the
metro area, while the PM2.5 measurements are from a more peripheral site (see
Appendix A) and are probably underestimates of PM2 5 at the PMio measurement
site; this would have only a small effect on estimates of coarse PM. While it is
quite possible that the very highest coarse PM concentration (indicated in Figure
B-l(b) to be about 500 |ig/m3) reflects the effect of exceptional winds, and might
be excluded under the Exceptional Event rule, the next-highest values of PMio
light extinction almost certainly would also be dominated by coarse PM
concentrations in the range of 150 to 200 |ig/m3 and many might not be
excludable.
Salt Lake City has extreme haze hours caused mostly by nitrate in the winter with
some periods with carbonaceous PM2.5 being the major contributor. Because
coarse PM in Salt Lake City was estimated by applying a regional factor to the
local PM2.5 mass value, it would not have been possible for the results to indicate
a significant coarse PM contribution to PMio light extinction even if one existed
at this site. However, from what EPA staff know of the area, it is unlikely that
there is a frequent large contribution from coarse PM. The area typically has at
most a few days per year with measured 24-hour average PMio as high as 150-200
|ig/m3. If this were all coarse PM, the contribution to 24-hour average light
extinction would be 90-120 Mm"1, with the possibility of much higher hourly
contributions by coarse mass during these few days.
Dallas and Houston have high sulfate PM2.5 contributions to PMio light extinction,
but Dallas also has some winter hours with extreme PMio light extinction with
substantial contributions from nitrate and organic carbonaceous material, while
Houston seems to have less contribution by nitrate. Because coarse PM in both
Dallas and Houston was estimated by applying a regional factor to the local PM2.5
mass value, it would not have been possible for the results to indicate a significant
coarse PM contribution to PMio light extinction even if one existed at this site.
However, from what EPA staff know of the areas, it is unlikely that there is a
frequent large contribution from coarse PM. Houston typically has at most a few
days per year with measured 24-hour average PMio as high as 150-200 |ig/m3. If
this were all coarse PM, the contribution to 24-hour average light extinction
would be 90-120 Mm"1. Dallas typically does not have PMio as high as 150
|ig/m3.
Sulfate in the summer and nitrate in the fall and winter are responsible for most of
the extreme PMio light extinction at St. Louis, though there are several maximum
3-83
-------
hourly periods where coarse PM is a major component. Recall that estimates of
coarse PM in St. Louis may be affected by a very local source (see Appendix A),
and thus the instances of high PMi0 light extinction due to coarse PM may be
limited in geographic scope.31
Birmingham and Atlanta are similar in having sulfate year-round and winter
carbonaceous PM2.5 as major contributors to their extreme PMio light extinction
periods. Coarse PM for Birmingham was estimated using data from a single site,
and the estimates should be reasonably representative. Coarse PM for Atlanta
was estimated using data from two fairly close sites and the estimates should be
reasonably representative.
Detroit has frequent large light extinction contributions from nitrate PM2.s, mostly
in the winter, as well as some contributions from sulfate PM2.5 year-round and
several fall and winter days with high contributions from carbonaceous PM2.5.
Coarse PM makes a notable contribution on a few days. Coarse PM for Detroit
was estimated using data from a single site near an automobile plant, and the
estimates should be reasonably representative for that site.
The remaining four urban locations (Pittsburgh, Baltimore, Philadelphia, and New
York) are similar in that most of their extreme PMio light extinction is from year-
round combinations of sulfate and nitrate. New York also has some winter
elemental and organic carbonaceous contributions to its extreme PMio light
extinction. Recall that the PM2.5 site representing the New York area is actually in
Elizabeth, NJ; emissions from diesel trucks on nearby interstate highways and/or
diesel engines associated with port activities might explain the carbonaceous
contributions. Coarse PM for Baltimore and Philadelphia was estimated using
data from a single site in each area, and the estimates should be reasonably
representative. Coarse PM for New York was estimated using data from two
fairly distant sites and the estimates may not be representative of both sites.
Because coarse PM was estimated for Pittsburgh by applying a regional factor to
the local PM2 5 mass value, it would not have been possible for the results to
indicate a significant coarse PM contribution to PMio light extinction even if one
existed at this site. However, exceedances of the PMio NAAQS are rare in
31 Comments concerning unrealistically high PM10_2 5 values for St. Louis are viewed as credible, but were
received too late in the review process to permit reanalysis using an alternate data set or to remove St. Louis from all
portions of this document. However, the text has been revised to caution readers with respect to the St. Louis
results, and they will not be included in the visibility effects discussion in the final PM Policy Assessment
document. Some graphics have been updated to exclude St. Louis results
3-84
-------
Pittsburgh suggesting that coarse PM likely is not a frequent significant
contributor to PMio light extinction.
3.5 POLICY RELEVANT BACKGROUND
Policy relevant background levels of PMio light extinction have been estimated for this
assessment by relying on outputs for the 2004 CMAQ run in which anthropogenic emissions in
the U.S., Canada, and Mexico were omitted, as described in the ISA. Estimates of PRB for PMio
light extinction were calculated from modeled concentrations of PM2.5 components using the
IMPROVE algorithm. The necessary component concentrations were extracted from the CMAQ
output files, as they were not summarized in the final ISA. More detail is provided in Appendix
C.
It is also necessary to have estimates of PRB for PMio-2.s, as input to the IMPROVE
algorithm. The final ISA for this review does not present any new information on this subject.
The approach used in the two previous reviews was to present the historical range of annual
means of PMio-2.s concentrations from IMPROVE monitoring sites selected as being least
influenced by anthropogenic emissions (US EPA, 2004, Table 3E-1). For this assessment, EPA
staff estimated PRB for PMio-2.s using a contour map based on average 2000-2004 PMio-2.s
concentrations from all IMPROVE monitoring sites, found in a recent report from the
IMPROVE program (DeBell, 2006). More detail is provided in Appendix C.
The outcome of the procedures for estimating PRB consists of hour-specific estimates of
PRB for PM2.s components and annual average estimates for PRB for PMio-2.5- Thus, hour-
specific estimates of PMio light extinction are possible, using the same hour-specific relative
humidity values as for the estimate of current conditions PMio light extinction.
In addition to allowing confirmation of the obvious fact that current conditions PMio light
extinction values are generally well above PRB conditions, the PRB estimates only play a role in
this assessment in the estimation of "what if scenarios representing compliance with alternative
NAAQS scenarios based on PMio light extinction. This role is described in section 4.1.4.
3-85
-------
This page left intentionally blank.
-------
4 PM10 LIGHT EXTINCTION UNDER "WHAT IF" CONDITIONS
OF JUST MEETING SPECIFIC ALTERNATIVE SECONDARY
NAAQS
4.1 ALTERNATIVE SECONDARY NAAQS BASED ON PMio LIGHT
EXTINCTION AS THE INDICATOR
4.1.1 Indicator and Monitoring Method
The indicator considered for the UFVA is PMio light extinction, assumed to be measured
by a continuous instrument, or instrument pair, capable of reporting both light scattering and
light absorption. EPA staff prepared a White Paper on Particulate Matter (PM) Light Extinction
Measurements (US EPA, 2010d) for the purpose of soliciting comments on prospective
measurement methods from the public and the Ambient Air Monitoring and Methods
Subcommittee (AAMMS). In its review (Russell and Samet, 2010), the AAMMS made the
recommendation to EPA that direct measurements be limited to PM2.5 light extinction as this can
be accomplished by a number of commercially available instruments and because PM2 5 is
generally responsible for most of the PM visibility impairment in urban areas. They indicated
that it is technically more challenging at this time to accurately measure the PMio-2.s component
of light extinction. These recommendations were received subsequent to completion of the
assessments described here, so they did not influence the use of PMio light extinction as the
indicator.
4.1.2 Alternative Secondary NAAQS Scenarios Based on PMio Light Extinction
Eighteen alternative NAAQS scenarios presented in Table 4-1 are analyzed in this
section. Nine are based on daily maximum daylight 1-hour PMio light extinction and nine on
PMio light extinction in all hours without the restriction to daily maxima. Within each set of
nine, the scenarios are ordered from least to most stringent.
4-1
-------
Table 4-1. Alternative Secondary NAAQS Scenarios for PM10 Light Extinction
Level
Annual
Percentile
Form
Scenarios Based on Daily Maximum Daylight 1-Hour PMi0 Light Extinction
(a) 191 Mm'1
(b) 191 Mm'1
(c) 191 Mm'1
(d) 112 Mm'1
(e) 112 Mm'1
(f) 112 Mm'1
(g) 64 Mm'1
(h) 64 Mm'1
(i) 64 Mm'1
90
95
98
90
95
98
90
95
98
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
Scenarios Based on Daylight 1-Hour PMio Light Extinction (All Daylight Hours)
(j) 191 Mm'1
(k) 191 Mm'1
(1) 191 Mm'1
(m) 112 Mm'1
(n) 112 Mm'1
(o) 112 Mm'1
(p) 64 Mm'1
(q) 64 Mm'1
(r) 64 Mm'1
90
95
98
90
95
98
90
95
98
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
4.1.3 Monitoring Site Considerations for Alternative Secondary NAAQS Based on
Measured PMio Light Extinction
It is useful to think ahead tentatively to monitor siting aspects of NAAQS
implementation, so that the results presented in the remainder of this chapter based on the 15
specific study sites can be better interpreted in terms of how well they might represent later
findings if these (and other) areas were to deploy PMio light extinction measurement instruments
as part of implementing a secondary NAAQS.
In light of the recommendations of the AAMMS (Russell and Samet, 2010), it is most
likely that the instruments that would be used if directly measured PM2.5 light extinction were
selected as the indicator to implement a secondary NAAQS would be "closed path" instruments
that react only to air quality in their immediate vicinity. However, light paths that matter to
perceived visual air quality are likely to be several kilometers long. Therefore, a monitoring site
should be at least neighborhood in scale, i.e., its relationship to emission sources and transport
should be such that measurements made at the site reasonably reflect concentrations in an area
surrounding the site of at least about 0.5 to 4 kilometers in diameter. The AAMMS
4-2
-------
recommendations also include advice concerning network design, and probe and siting criteria
applicable to a program of directly measuring PM2.5 light extinction.1
With regard to the monitoring sites used in this assessment, all are reported to be, or
appear to be, neighborhood or larger scale, and all are in areas where people are present during
daylight hours. The sites in Detroit (Dearborn) and New York (Elizabeth, NJ) are, however,
rather close to an industrial source and a major interstate highway interchange/turnpike exit,
respectively. Significantly, most of the study sites are not the highest PM2.5 concentration site in
their urban area, so a "what if scenario that manipulates the "current conditions" at these sites to
"just meet" an alternative secondary NAAQS might implicitly leave other parts of their urban
areas with PM2.5 light extinction above the NAAQS.
4.1.4 Approach to Modeling "What If Conditions for Alternative Secondary
NAAQS Based on Measured PMio Light Extinction
Before modeling "what if conditions, EPA staff augmented the data set described in
Table 4 so that the sets of study days for Houston and Phoenix were seasonally balanced despite
the lack of actual monitoring data for one quarter in each city. For the first quarter of 2005 in
Phoenix, we substituted the available 12 days from the first quarter of 2006. For the fourth
quarter of 2007 in Houston, we substituted 13 randomly drawn days from the fourth quarters of
2005 and 2006.
Also, Tacoma (originally) and Phoenix (after this augmentation) each have only two
calendar years of suitable data, while the form of the alternative NAAQS scenarios requires the
averaging of the 90th, 95th, or 98th percentile values from three years. In Tacoma and Phoenix,
for every step in the analysis at which a design value is used as an input or reported as an output,
we averaged the percentile values from the only two available years.
We modeled all daylight and daily maximum daylight 1-hour PMio light extinction under
each of the "what if scenarios (in which each study area "just meets" one of the 18 alternative
secondary NAAQS listed in section 4.1.2) via the following steps. These steps are essentially the
same as the "proportional rollback" steps that have been used in the health risk assessment
modeling of "what if conditions in several previous NAAQS reviews for PM and other criteria
pollutants. The steps are described here for the nine scenarios based on daily maximum daylight
1-hour PMio light extinction; similar steps were followed for the nine scenarios based on
percentiles of all daylight 1-hour PMio light-extinction. The referenced tables present results for
both sets of scenarios.
1 In chapter 4 of the second review draft of the PM Policy Assessment (US EPA, 2010c), EPA staff
considers as an alternative to directly measuring PM2 5 light extinction the use of speciated PM2 5 mass-calculated
light extinction by a method similar in concept to but simpler than the method described in section 3.2. Much of the
monitoring infrastructure needed to implement this approach is already deployed by state and local air agencies.
4-3
-------
1. After excluding hours with relative humidity greater than 90 percent, identify the
appropriate percentile (90th, 95th, or 98th) daily maximum daylight 1-hour PMio light
extinction value in each year, noting the day and hour each occurred, and average
these values across years to calculate the PMio light extinction design value for each
site consistent with the percentile form of the NAAQS scenario.2 The three resulting
design values for each area (for the 90th, 95th, and 98th percentile forms) are shown in
Table 4-2. (Note that in a number of cases, which are identified by a footnote, the
study area meets one or more of the NAAQS scenarios under current conditions. In
these cases, the "current conditions" PMio light extinction values are not adjusted,
i.e., PMio light extinction values are never "rolled up.") Notice that the design values
for the 90th percentile maximum daily 1-hour for most cities are generally similar to
the design values for the 98th percentile of all daylight hours. On average there are
about ten hours defined as daylight per day, so if the PMio light extinction were
randomly distributed among the daylight hours and days, the 90th percentile
maximum daily 1-hour would correspond to the 99th percentile of all hours; the fact
that the point of rough equivalency is the 98th percentile indicates a tendency for
hours with higher PMio light extinction to cluster together in the same day. Figure 4-
1 presents two scatter plots that relate the design values based on daily maximum 1-
hour PMio light extinction values and the design values based on all daylight 1-hour
PMio light extinction values. In Panel A, design values for the daily maximum and all
hours forms are paired by the defining percentile, and colors are used to distinguish
the 90th, 95th, and 98th percentile statistical forms. It appears from Panel A that the
design values for the two approaches to defining the NAAQS scenarios are highly
correlated but with the all hours approach resulting in numerically lower design
values than the daily maximum approach. The correlation breaks down for the 98th
percentile form for the few study areas with the highest levels of PMio light
extinction. Panel B compares the 90th percentile design values based on daily
maximum PMio light extinction with the 90th, 95th, and 98th percentile design values
based on all daylight hours PMio light extinction. There is close agreement between
the 90th percentile design values based on daily maximum values and the 98th
percentile design value based on all daylight hours.
2 Annual percentile values were picked from the set of day-specific or hour-specific estimates according to
the same scheme as used for the current 24-hour secondary PM2 5 standard, as explained in section 4.5(a) of 40 CFR
50 Appendix N. For example, if there are 60 daily maximum values in a year, the second highest value is the 98th
percentile value. Note that this differs from the algorithm used by some spreadsheet and other statistical programs,
which may interpolate between sample values. Also, this is a different approach than that used in the Regional Haze
program, in which conditions in the best and worst 20 percent of days are averaged together, rather than focusing on
conditions on the specific day at the 80th and 20th percentile points.
4-4
-------
Table 4-2. Current Conditions PMio Light Extinction Design Values for the Study
Areas.
Study Area
Design Value for
90th Percentile
Form (Mm *)
Design Value for
95th Percentile
Form (Mm *)
Design Value for
98th Percentile
Form (Mm *)
Design Values Based on Daily Maximum Daylight 1-Hour PMio Light Extinction
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
140*
330
454
105*
163*
184*
194
306
357
249
308
278
246
285
306
157*
460
550
144*
252
239
234
380
483
288
471
313
286
334
354
211
530
611
185
409
301
291
467
562
331
644
364
326
374
451
Design Values Based on Daylight 1-Hour PMio Light Extinction (All Daylight Hours)
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
76*
188*
261
68*
93*
113*
105*
193
173*
166*
212
167*
171*
183*
186*
105*
261
341
79*
141*
143*
128*
235
227
195
251
209
225
222
243
136*
368
441
94*
225
188*
171*
290
309
238
315
264
262
278
299
* This design value meets one or more of the NAAQS scenarios based on PM10 light extinction.
4-5
-------
Figure 4-1. Comparison of Daily Max and All Daylight Hour Design Values for PMi0
Light Extinction
(A) Comparison of Design Values Matched by Percentile Form
vnn
^ finn
01
— i^nn
03 OUU
>
c
01 Ann
v, 4UU
at
a
"* ^nn
3
O
ac
<- onn
O9
>>
ro inn
Q IUU
<
//
/
A
/ A
/ m
/ A * *
-/ *i
/_ •** • * •
%A**»
Af-
/ m.&
/.
* 90th Percentile
• 95th Percentile
A 98th Percentile
0 100 200 300 400 500 600 700
Daily Max Daylight Hour Design Value (Mm-1)
(B) Comparison of 90th Percentile Daily Maximum Design Values and 90th, 95th, and 98th
Percentile All Daylight Hours Design Values
Comparison of 90th percentile Daily Max Form to All Hours
Forms
90th percentile
95th percentile
98th percentile
0 100 200 300 400 500
Design Value for 90th Percentile Daily Max (Mm-1)
4-6
-------
2. Using the same days and hours, find the three (or two, in the case of Phoenix and
Houston for which there were only two years of suitable data available)
corresponding values of PRB PMio light extinction, and average these values across
years to calculate the PRB portion of the design value.
3. Subtract the value from step 2 from the value from step 1, to determine the non-PRB
portion of the design value.
4. Calculate the percentage reduction required in non-PRB PMio light extinction in
order to reduce the design value to the PMio light extinction level that defines the
NAAQS scenario, using the following equation:
Percent reduction required = 1 - (NAAQS level - PRB portion of the design value)/
(non-PRB portion of the design value)
The percentage reductions determined in step 4 are shown in Table 4-3. Figure 4-2
presents them graphically in the form of a scatter plot, comparing the required
reductions for scenarios based on daily maximum 1-hour daylight PMio light
extinction values to scenarios with the same level and percentile form but based on all
daylight hours 1-hour PMio light extinction values. For the NAAQS scenarios
involving higher levels and lower percentile forms, there are some notable differences
in the percentage reductions required for some area to attain. As was the case for the
design values, notice in Table 4-3 that there are generally similar percentage
reductions for each city and level for the 90th percentile maximum daily and 98th
percentile of all daylight hours.
As already stated, if the study area is meeting a NAAQS scenario in the current
conditions case, no adjustments were made to represent the "just meeting" case. In
effect, negative values for the percent reduction required to meet the NAAQS
scenario calculated by the above equation were re-set to zero.
5. Turning to the entire set of day/hour-specific actual and PRB daylight PMio light
extinction values for the three (or two) year period, determine the non-PRB portion of
PMio light extinction in that hour, reduce it by the percentage determined in step 4,
and add back in the PRB PMio light extinction. The result is the "just meets" PMio
light extinction value for that day and hour.
Note that in these steps, it is not necessary to make any explicit or implicit assumption
about what PM components would be reduced to allow the area to meet the NAAQS scenario, as
the NAAQS scenario's target design value is itself in units of light extinction. One path to
meeting a NAAQS scenario would be to reduce each of the five PM2 5 components (and thus the
annual and 24-hour design values shown in Table 3-2) and PMio-2.5 by the calculated "percent
reduction required". However, a lesser reduction in one or more of the six PM concentrations
could be offset by a greater reduction in one or more of the remaining concentrations. Thus, it is
4-7
-------
not possible to associate unique values of annual average and 24-hour average PM2.5 with the
"just meeting" NAAQS scenarios reported in Table 4-3.
4-8
-------
Table 4-3. Percentage Reductions in Non-PRB PMi0 Light Extinction Required to "Just
-1\3
Meet" the NAAQS Scenarios Based on Measured Light Extinction (Mm )
Scenario
Level/Form
Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Scenario
Level/Form
Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
NAAQS Scenarios Based on Daily Maximum 1-Hour Daylight PMi0 Light Extinction
(a)
191/90"1
(b)
191/95"1
(c)
191/98"1
(d)
112/ 90th
(e)
112/95"1
(f)
112/98"1
fe)
64/90"1
(h)
64/95"1
(i)
64/98"1
Percentage Reduction Required in Non-PRB PMi0 Light Extinction
0
43
59
0
0
0
2
38
48
24
39
32
23
34
38
0
59
66
0
24
21
20
51
61
35
60
40
34
43
47
10
64
69
0
54
38
35
60
67
44
71
48
42
49
58
22
67
76
0
32
41
44
64
70
57
65
60
56
62
64
31
76
81
22
56
54
56
72
77
63
77
65
63
67
69
52
79
82
40
73
65
63
77
81
68
83
70
67
71
76
59
82
87
39
61
69
70
80
84
77
80
78
76
79
80
63
87
89
56
75
75
78
85
87
80
87
81
80
82
83
78
89
90
66
85
81
80
88
90
83
91
83
82
84
87
NAAQS Scenarios Based on 1-Hour Daylight PM10 Light Extinction
(i)
191/90"1
(k)
191/95"1
0)
191/98"1
(m)
112/90"1
(n)
112/95"1
(o)
112/98"1
(P)
64/90"1
(q)
64/95"1
(r)
64/98"1
Percentage Reduction Required in Non-PRB PM10 Light Extinction
0
0
27
0
0
0
0
1
0
0
10
0
0
0
0
0
27
45
0
0
0
0
19
16
2
24
9
15
14
22
0
49
57
0
15
0
0
35
39
20
40
28
28
32
36
0
41
58
0
0
1
0
43
36
33
48
34
35
39
40
0
58
68
0
21
23
13
53
52
44
56
47
51
51
55
21
70
75
0
51
42
37
62
65
55
65
58
58
61
63
18
67
77
5
32
44
40
68
65
63
71
63
64
66
67
44
77
82
19
55
58
51
74
74
70
75
70
73
73
75
64
84
86
32
72
68
67
79
81
75
81
77
77
78
79
As a result of a formula error, the intended values of PRB PM10 light extinction were not properly used in
calculating the entries in this table, generally resulting in the required reductions shown here to be slightly smaller
than they should be. For a more detailed explanation of this issue, see the 2010 Lorang memo, "Explanation of
Error in Table 4-3 (Percentage reductions in non-PRB PM10 light extinction required to 'just meet' the NAAQS
scenarios based on measured light extinction) of the final Urban Focused Visibility Assessment", July 23, 2010. As
discussed in that memo, the only other results presented in this document that were affected by this error were
Figure 4-2, Figure 4-3(a), and Panels (a) through (r) of Appendix F. The effect on those figures was judged too
negligible to warrant regenerating them for this final version.
4-9
-------
Figure 4-2. Comparison of Required Percentage Reductions in Non-PRB PMi0 Light
Extinction Needed to Meet NAAQS Scenarios
o 100
20 40 60 80
Based on Daily Maximum Daylight LE
100
• 191 /90th • 191 /95th A 191 /98th * 112/90th * 112/95th • 112/98th +64/90th
- 64/95th 64/98th
4.2 ALTERNATIVE SECONDARY PM2.5 NAAQS BASED ON ANNUAL AND
24-HOUR PM2.5 MASS
4.2.1 Secondary NAAQS Scenarios Based on Annual and 24-Hour PM2.sMass
In this final assessment, EPA staff have modeled two "what if scenarios using the same
indicators and averaging periods as defined in the current suite of PM2.5 NAAQS set in 2006.
The first scenario uses the current suite of PM2.5 NAAQS levels and the second a suite of lower
levels considered in the health risk assessment (US EPA, 2010e):
• 15 |ig/m3 weighted annual average PM2.5 concentration and 35 |ig/m3 24-hour average
PM2.5 concentration with a 98th percentile form, both averaged over three years.
• 12 |ig/m3 weighted annual average PM2.5 concentration and 25 |ig/m3 24-hour average
PM2.5 concentration with a 98th percentile form, both averaged over three years.
4-10
-------
4.2.2 Approach to Modeling Conditions If Secondary PM2.s NAAQS Based on
Annual and 24-Hour PMi.s Mass Were Just Met
Because these NAAQS scenarios are based on PM2.5 mass as the indicator, rather than
light extinction, the steps needed to model "what if conditions are somewhat different, and
involve explicit consideration of changes in PM2 5 components.
1. Apply proportional rollback to all the PM2.5 monitoring sites in each study area,
taking into account PRB PM2.5 mass, to "just meet" the NAAQS scenario for the area
as a whole, not just at the visibility assessment study site. The health risk assessment
document describes this procedure in detail. The degree of rollback is controlled by
the highest annual or 24-hour design value, which in most study areas is from a site
other than the site used in this visibility assessment. The relevant result from this
analysis is the percentage reduction in non-PRB PM2.5 mass need to "just meet" the
NAAQS scenario, for each study area. These percentage reductions are shown in
Table 4-4. Note that Phoenix and Dallas meet the 15/35 NAAQS scenario under
current conditions, and require no reduction. PM2.5 levels in these two cities were not
"rolled up."
2. For each day and hour for each PM2 5 component, subtract the PRB concentration
from the current conditions concentration, to determine the non-PRB portion of the
current conditions concentration.
3. Apply the percentage reduction from step 1 to the non-PRB portion of each of the
five PM2 5 components. Add back the PRB portion of the component.
4. Re-apply the IMPROVE algorithm (section 3.2.3), using the reduced PM2.5
component concentrations, the current conditions PMio-2.s concentration for the day
and hour, and relative humidity for the day and hour. Include the term for Rayleigh
scattering.
4-11
-------
Table 4-4. Percentage Reductions Required in Non-PRB PM2.s Mass to "Just Meet'
NAAQS Scenarios Based on Annual and 24-Hour PM2.s Mass
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Percentage Reduction Required
(S)
Annual PM2.5 NAAQS = 15 ug/m3
24-Hour PM2.5 NAAQS = 35 ug/m3
19
45
37
0*
37
0*
6
10
22
8
19
19
6
8
17
(t)
Annual PM25 NAAQS = 12 ug/m3
24-Hour PM2.5 NAAQS = 25 ug/m3
43
61
55
22
56
7
27
37
45
30
43
43
33
35
41
* These areas meet this NAAQS scenario under current conditions.
4.3 RESULTS FOR EACH "JUST MEET" ALTERNATIVE SECONDARY
NAAQS SCENARIO
The modeling described in sections 4.1 and 4.2 resulted in estimates of PMi0 light
extinction for each day and hour in each study area, for each NAAQS scenario. Four summaries
of these conditions are presented here. Figure 4-3 shows two box-and-whisker plots of daily
maximum daylight 1-hour PMio light extinction. The top panel (a) is for the single illustrative
scenario of a NAAQS based on daily maximum daylight 1-hour PMio light extinction with a
level of 112 Mm"1 and a 90th percentile form, which was chosen for this illustration because it is
approximately mid-way among the nine scenarios based on daily maximum PMio light extinction
in terms of stringency.4 The bottom panel (b) is for the scenario of meeting the current suite of
secondary PM2.s NAAQS standards: 15 |ig/m3 annual average and 35 |ig/m3 24-hour average
(98th percentile form). A notable feature of this comparison is that in the top panel, all the study
4 Plots of the distribution of daily maximum PM10 light extinction for all 18 NAAQS scenarios based on
daily maximum PM10 light extinction, and of individual hourly PM10 light extinction for all 18 NAAQS scenarios
based on individual daylight hours, are provided in Appendix F.
4-12
-------
areas have a similar distribution of the daily maximum daylight 1-hour PMio light extinction,
while in the bottom panel this is not the case. This is expected, since a NAAQS based on a
measured daily maximum PMio light extinction indicator will of course result in areas achieving
similar daily maximum PMio light extinction patterns once each area reaches a "just meets"
condition. In areas with generally higher relative humidity conditions, concentrations of PM2.5
components and/or PMio-2.5 would need to be lower to achieve the "just meet" condition. In
contrast, in the NAAQS scenario represented by the bottom panel, concentrations of PM2.5 mass
will be similar across areas, but concentrations of PM2.5 components may not be, and levels of
PMio light extinction will not be similar in areas with dissimilar levels of relative humidity. The
specific differences among areas in the bottom panel are generally as expected, with the drier
study areas having lower levels of PMio light extinction.
Tables 4-5 and 4-6 summarize the "just meet" conditions in the NAAQS scenarios in
terms of the PMio light extinction design values. Table 4-5 addresses the 18 scenarios of
NAAQS based on measured PMio light extinction. When an area just meets a NAAQS scenario,
its design value in principle should exactly equal the NAAQS level, so preparation of this table
serves as a check against calculation errors. Note that the design values in Table 4-5, resulting
from the rollback steps described in section 4.1.4, in some cases do not exactly equal the
assumed level of the NAAQS, although all are quite close. Closer investigation has revealed that
this is mostly a result of hours switching their ranking in the rollback process. Hours can switch
rank because the level of PRB PMio light extinction varies with each hour, so a uniform
percentage reduction in non-PRB light extinction (step 5) can result in non-uniform percentage
reductions in actual PMio light extinction; a lower ranking hour can thereby move up in the post-
rollback ranking. In principle, rollback could be iterated to exactly achieve a design value equal
to the level of the NAAQS for each scenario. However, the discrepancies indicated in Table 4-5
were judged too small to justify iterative rollback, given other uncertainties in the analysis.
Table 4-6 addresses the two scenarios of NAAQS based on PM2.s mass, with PMio light
extinction design values shown for the 90th, 95th , and 98th percentile forms.
4-13
-------
Figure 4-3. Distributions of Daily Maximum Daylight 1-Hour PMi0 Light Extinction Under Two "Just Meet"
Secondary NAAQS Scenarios (Excluding Hours with Relative Humidity Greater Than 90 Percent)
(a) Secondary NAAQS Based on Daily Maximum Daylight 1-Hour PMio Light Extinction with a Level of 112 Mm * and a 90th
Percentile Form
ExtRollbackDailyMaxNAAQS112Pctl90DVsFromdaily.max
_ 8 -
109 324 300 98 306 273 158 289 349 279 141 277 181 143 225
8
o
..
e*
o
o
------
-J--
4-14
-------
Figure 4-3. Distributions of Daily Maximum Daylight 1-Hour PMi0 Light Extinction Under Two "Just Meet" Secondary
NAAQS Scenarios (Excluding Hours with Relative Humidity Greater Than 90 Percent), continued
(b) Secondary NAAQS of 15 ug/m3 for the Annual Average and 35 ug/m3 for the 98th Percentile 24-Hour Average
PMRollbackDailyMaxCasel NAAQS
CD
CD —
CO
-------
Table 4-5. PM10 Light Extinction Design Values for "Just Meet" Secondary NAAQS
Scenarios Based on Measured PM10 Light Extinction (Excluding Hours with
Relative Humidity Greater Than 90 Percent)
Level (Mm ')
Percentile Form
Tacoma, WA
Fresno, CA
Los Angeles, CA
Phoenix, AZ
Salt Lake City, UT
Dallas, TX
Houston, TX
St. Louis, IL
Birmingham, AL
Atlanta, GA
Detroit, MI
Pittsburgh, PA
Baltimore, MD
Philadelphia, PA
New York, NY
Level (Mm ')
Percentile Form
Tacoma, WA
Fresno, CA
Los Angeles, CA
Phoenix, AZ
Salt Lake City, UT
Dallas, TX
Houston, TX
St. Louis, IL
Birmingham, AL
Atlanta, GA
Detroit, MI
Pittsburgh, PA
Baltimore, MD
Philadelphia, PA
New York, NY
Secondary NAAQS Scenarios Based on Daily Maximum
(a)
191
90th
(b)
191
95th
(c)
191
98th
(d)
112
90th
(e)
112
95th
(f)
112
98th
(g)
64
90th
(h)
64
95th
(i)
64
98th
PMio Light Extinction Design Value
(based on same percentile form as the NAAQS scenario)
140
191
191
105
163
184
191
191
191
191
191
191
191
191
191
157
191
191
144
191
191
191
191
191
191
191
191
191
191
191
191
191
191
185
191
191
191
191
191
191
191
191
191
191
191
112
113
113
105
112
112
114
112
113
112
112
112
111
112
112
115
112
112
112
112
113
111
112
113
112
112
112
112
112
112
108
112
112
112
112
112
112
112
112
112
112
112
112
112
112
66
65
65
64
65
64
67
65
64
64
64
64
63
64
64
74
64
64
64
64
65
61
64
66
64
64
64
64
64
64
58
64
64
64
64
65
67
64
64
65
65
64
64
64
64
Secondary NAAQS Scenarios Based on All Daylight Hours
(i)
191
90th
(k)
191
95th
(1)
191
98th
(m)
112
90th
(n)
112
95th
(o)
112
98th
(P)
64
90th
(q)
64
95th
(r)
64
98th
PMio Light Extinction Design Value
(based on same percentile form as the NAAQS scenario)
76
188
191
68
93
113
105
191
173
166
191
167
171
183
186
105
192
191
79
141
143
128
191
191
191
191
191
191
191
191
136
191
191
94
191
188
171
191
191
192
191
191
191
191
192
76
112
112
68
93
112
105
112
112
113
112
113
112
112
112
105
113
113
79
112
112
113
112
112
112
112
113
112
112
112
112
111
112
94
112
113
111
112
112
113
112
112
112
113
113
63
65
64
64
64
65
65
65
65
65
64
65
64
65
65
64
64
65
64
64
64
66
64
64
63
65
65
65
64
65
59
63
64
64
64
66
61
65
65
65
64
65
64
65
65
4-16
-------
Table 4-6. PM10 Light Extinction Design Values for "Just Meet" Secondary NAAQS
Scenarios Based on PM2.s Mass (Excluding Hours with Relative Humidity
Greater Than 90 Percent)
Annual/24-Hour
PM2S NAAQS
City Name
Tacoma, WA
Fresno, CA
Los Angeles, CA
Phoenix, AZ
Salt Lake City, UT
Dallas, TX
Houston, TX
St. Louis, IL
Birmingham, AL
Atlanta, GA
Detroit, MI
Pittsburgh, PA
Baltimore, MD
Philadelphia, PA
New York, NY
Tacoma, WA
Fresno, CA
Los Angeles, CA
Phoenix, AZ
Salt Lake City, UT
Dallas, TX
Houston, TX
St. Louis, IL
Birmingham, AL
Atlanta, GA
Detroit, MI
Pittsburgh, PA
Baltimore, MD
Philadelphia, PA
New York, NY
(s)
15ng/m3 / 35ng/m3
90th %tile
Design Value
(Mm1)
95th %tile
Design Value
(Mm1)
98th %tile
Design Value
(Mm1)
(t)
12ng/m3 / 25ng/m3
90th %tile
Design Value
(Mm1)
95th %tile
Design Value
(Mm1)
98th %tile
Design Value
(Mm1)
Design Values Based on Daily Maximum Daylight 1-Hour PMio Light Extinction
119
191
312
105*
110
183*
185
286
285
230
256
229
233
263
255
131
266
361
143*
167
239*
222
354
394
266
387
258
272
308
295
178
304
429
185*
269
301*
276
441
464
307
536
299
308
346
376
93
142
231
96
83
172
148
252
213
181
187
167
169
190
182
102
196
261
135
125
224
178
289
300
208
277
188
202
222
213
136
224
355
161
198
282
220
363
365
243
401
218
221
254
268
Design Values Based on Daylight 1-Hour PMio Light Extinction (All Daylight Hours)
65
110
173
68*
63
113*
99
180
140
154
175
138
163
169
155
88
152
228
79*
95
143*
122
221
183
180
208
173
213
205
203
113
214
294
94*
150
188*
163
271
247
220
257
218
248
258
249
52
83
129
60
49
106
81
147
105
123
130
102
121
123
113
70
113
169
70
72
134
99
183
138
144
155
127
155
149
148
84
160
220
86
112
176
131
237
186
174
188
159
184
187
178
* Phoenix and Dallas meet 15 ug/m3/35 ug/m3 under current conditions, so these entries are essentially the same as
for current conditions.
4-17
-------
Table 4-7 summarizes all 20 scenarios in terms of the percentage of days (across 2005 to
2007, but after rollback) in which the daily maximum daylight 1-hour PMio light extinction
under "just meeting" conditions exceeds each of the CPLs. Part A of the table applies to
NAAQS scenarios based on daily maximum 1-hour PMio light extinction values. Part B of the
table applies to the scenarios based on 1-hour PMio light extinction values during all daylight
hours. Note that the reported percentages in both Part A and Part B is the percentage of days in
which the daily maximum daylight 1-hour PMio light extinction under "just meet" conditions
exceeds each of the CPLs; this allows comparison of the "effectiveness" of the two NAAQS
approaches using a consistent metric. (The 15/35 and 12/25 NAAQS scenarios are the same in
Part A and Part B, and are repeated only for convenience in making comparisons.) Hours with
relative humidity above 90 percent have been excluded from consideration, consistent with the
definition of the NAAQS scenarios. Also shown at the bottom of the table in each column
representing a NAAQS scenario is the average of these percentages of time across the 15 study
areas (this is the simple column average, not weighted by the number of days available in each
area). Comparisons of these percentages allows a rough indication of how the two scenarios of a
NAAQS based on PM2.5 mass compare to the other 18 scenarios in terms of protecting visual air
quality. Notice that the most restrictive of the two NAAQS scenarios based on PM2.5 mass
would reduce the projected 1-hour maximum daily PMio light extinction above the least
restrictive CPL (IQIMm"1) to less than 10 percent of the time for most of the urban areas (only
L.A., St. Louis, and Birmingham have values above 10 percent).5 However at the current PM
NAAQS level (i.e., 15/35) all of the eastern urban areas and Los Angeles exceed the least
restrictive CPL more than 10% of the time. Comparison of Parts A and B of Figure 4-7 indicates
that basing a PMio light extinction NAAQS scenario on daily maximum 1-hour PMio light
extinction has a lower percentage in excess of the 1-hour daily maximum versus the NAAQS
scenario based on all daylight hours PMio light extinction for a given level and percentile form of
the NAAQS. This is consistent with the results presented in Table 4-2 and Figure 4-1, which
indicated that current conditions design values are generally lower for the all hours approach.
Again there is near equivalence between the 90th percentile daily maximum and 98th percentile
all daylight hours in terms of the percent of days exceeding the daily maximum CPL values in
Table 4-7.
5 Comments were received concerning unrealistically high PM10.2 5 values for St. Louis and to a lesser
extent for Los Angeles. High contributions by PM10.2.5 would help explain why a PM2 5 standard would be less
effective in reducing visibility impacts. EPA staff view the comments concerning unreliable PM10.2.5 values for St.
Louis as credible, but these comments were received too late in the review process to permit reanalysis using an
alternate data set or to remove St. Louis from this document. However, the text has been revised to caution readers
with respect to the St. Louis results, and they will not be forwarded to the visibility effects discussion in the PM
Policy Assessment document.
4-18
-------
Table 4-7. Percentage of Days with Maximum 1-Hour Daylight PM10 Light Extinction Above CPLs For Each NAAQS
Scenario Under "Just Meet" Conditions Across Three Years (or Two in the Case of Phoenix and Houston)
(A) NAAQS Scenarios Based on Daily Maximum 1-Hour PMio Light Extinction
Scenario
NAAQS
Level
Mm1
NAAQS
Percentile
Form
Annual/
24-Hour
Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake
City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Average
Days with Max Hour Above
64 Mm -1
a
191
90
b
191
95
C
191
98
(1
112
90
e
112
95
f
112
98
g
64
90
h
64
95
i
64
98
s
IS/
35
t
127
25
Percentage of days
52
54
74
44
44
80
77
81
63
85
74
70
67
72
63
67
52
40
64
44
27
66
65
71
49
80
52
63
61
66
59
57
48
32
58
44
13
50
55
54
40
76
41
54
54
60
40
48
40
31
43
44
24
48
47
46
33
62
46
40
41
41
33
41
34
19
32
27
11
25
30
34
19
51
22
31
30
31
27
28
16
13
27
10
5
14
18
20
14
34
6
23
25
27
18
18
12
10
12
10
10
10
13
11
10
9
11
9
11
8
9
10
10
5
6
5
5
5
3
6
6
5
4
6
4
6
6
6
2
3
3
2
1
1
3
2
3
1
3
2
2
3
2
2
43
54
85
44
24
81
75
97
84
90
80
79
78
85
76
72
28
40
79
40
15
77
64
89
70
85
74
63
64
74
62
62
Days with Max Hour Above
112 Mm -1
a
191
90
b
191
95
Percenta
22
29
40
6
17
41
43
45
30
59
45
37
39
38
32
35
22
16
31
6
11
22
28
30
18
47
17
29
29
31
24
24
C
191
98
(1
112
90
e
112
95
f
112
98
g
64
90
h
64
95
i
64
98
s
157
35
t
127
25
ge of days
14
11
26
6
5
12
16
18
12
30
6
21
23
24
16
16
11
9
11
6
10
10
12
12
10
11
11
9
11
8
9
10
7
5
6
5
5
5
4
6
5
5
4
6
5
5
6
5
2
3
3
2
1
1
3
2
3
1
3
1
2
3
2
2
1
2
1
2
4
1
1
2
1
0
4
0
1
0
1
1
1
0
0
1
1
0
0
1
0
0
1
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
11
29
69
6
9
41
41
73
55
71
61
48
47
61
46
45
5
17
52
6
6
37
23
57
38
54
49
28
31
38
30
31
Days with Max Hour Above
191 Mm -1
a
191
90
b
191
95
C
191
98
(1
112
90
e
112
95
f
112
98
g
64
90
h
64
95
i
64
98
s
157
35
t
127
25
Percentage of days
4
9
11
1
8
10
11
11
10
11
10
9
12
8
10
9
4
5
6
1
5
5
6
5
5
4
4
5
6
5
6
5
3
3
3
1
1
1
2
3
2
2
3
1
2
3
2
2
1
2
1
1
4
1
1
2
1
0
4
0
1
0
1
1
1
0
0
1
1
0
1
1
0
0
1
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
9
37
1
4
10
11
36
24
25
33
16
19
29
19
18
0
5
19
1
2
8
3
20
13
8
10
5
8
9
8
8
4-19
-------
(B) NAAQS Scenarios Based on PMio Light Extinction During All Daylight Hours*
Scenario
NAAQS
Level
Mm1
NAAQS
Percentil
e Form
Annual/
24-
Hour
Area
Tacoma
Fresno
Los
Angeles
Phoenix
Salt Lake
City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Dhiladelphia
New York
Average
Days with Max Hour Above
64 Mm -1
,j
191
90
k
191
95
1
191
98
m
112
90
n
112
95
0
112
98
P
64
90
q
64
95
r
64
98
$
157
35
t
127
25
Percentage of days
52
76
86
44
44
80
77
98
89
91
84
85
80
86
83
77
52
65
83
44
44
80
77
92
85
91
78
83
72
82
74
73
52
51
76
44
33
80
77
84
72
87
72
73
66
73
64
67
52
56
75
44
44
80
77
78
74
81
66
69
60
68
62
66
52
41
61
44
27
63
70
65
59
76
57
55
44
59
44
54
40
27
46
44
15
45
53
51
42
63
45
42
38
43
34
42
40
30
40
44
24
42
49
39
42
51
41
35
28
34
31
38
23
17
27
29
12
21
35
27
25
31
27
22
14
23
18
23
10
7
14
17
6
11
13
15
15
14
10
11
10
9
11
12
43
54
85
44
24
81
75
97
84
90
80
79
78
85
76
72
28
40
79
40
15
77
64
89
70
85
74
63
64
74
62
62
Days with Max Hour Above
112 Mm -1
,j
191
90
k
191
95
1
191
98
m
112
90
n
112
95
0
112
98
P
64
90
q
64
95
r| .
64
98
157
35
t
127
25
Percentage of days
22
52
73
6
17
41
44
76
64
75
63
57
50
63
59
51
22
37
58
6
17
41
44
62
55
73
55
53
43
58
42
44
22
25
42
6
14
41
44
48
39
62
45
40
34
41
32
36
22
30
41
6
17
41
44
39
42
48
41
34
28
33
30
33
22
17
27
6
12
21
33
27
25
30
26
22
14
24
18
22
11
7
14
6
5
10
13
15
14
12
9
11
9
8
10
10
12
9
10
6
9
9
13
9
14
5
6
8
4
6
8
9
4
4
3
5
5
4
8
4
8
1
4
1
2
1
3
4
1
1
1
3
2
1
1
2
3
0
4
0
1
0
1
1
11
29
69
6
9
41
41
73
55
71
61
48
47
61
46
45
5
17
52
6
6
37
23
57
38
54
49
28
31
38
30
31
Days with Max Hour Above
191 Mm -1
,j
191
90
k
191
95
1
191
98
m
112
90
n
112
95
Percenta
4
29
42
1
8
10
11
39
34
31
40
26
23
31
28
24
4
17
27
1
8
10
11
27
26
30
26
22
14
24
18
18
4
7
13
1
5
10
11
14
14
12
9
11
9
8
10
9
4
10
12
1
8
10
11
9
15
6
6
8
5
8
8
8
4
5
3
1
5
4
7
4
9
2
4
2
2
1
3
4
0
112
98
P
64
90
q
64
95
r
64
98
$
157
35
t
127
25
ge of days
1
1
1
1
2
1
2
2
3
0
4
0
1
0
1
1
1
2
1
1
4
0
1
1
3
0
3
0
0
0
0
1
0
0
0
1
1
0
1
0
1
0
1
0
0
0
0
0
0
0
0
1
1
0
0
0
0
0
1
0
0
0
0
0
1
9
37
1
4
10
11
36
24
25
33
16
19
29
19
18
0
5
19
1
2
8
3
20
13
8
10
5
8
9
8
8
* Note that the table reports results based on daily maximum daylight hour, while the NAAQS scenarios in Panel B are based on all daylight
hours (in both cases excluding hours with RH>90%).
4-20
-------
5 SUMMARY
This chapter integrates the key information on the purpose, approach, principal results,
and significant technical issues of the assessment efforts that are characterized in greater detail in
chapters 2, 3 and 4 and the appendices of this final Urban Focused Visibility Assessment
(UFVA). Earlier versions of the UFVA1 document the original assessment and its evolution in
response to Clean Air Science Advisory Committee (CASAC) and public comments. This
chapter is organized by the three separate assessments that are described in greater detail in
chapters 2 through 4.
5.1 Urban Visibility Preference Studies Reanalysis
Purpose: The overall purpose in conducting a reanalysis of urban preference studies is to
determine whether there is a credible range of acceptable visual air quality conditions above
which the national public welfare is adversely affected. Similar, though not identical, visibility
preference studies were conducted in four metropolitan areas including Denver, CO; Vancouver,
BC (Canada); Phoenix, AZ; and Washington, DC. These studies were performed separately to
support the development of local visibility protection efforts, except for Washington, DC which
was the subject of two separate pilot studies designed to better understand visibility preference
studies. The common feature in each of these studies was that study participants were asked to
individually rate the acceptability of scenic images shown to them one at a time in random order
that depicted visibility impairment over a range of conditions from nearly pristine to highly
impaired.
Approach: The methodology used in the reanalysis involved a critical review of data
generated for each study to identify issues of consistency and to identify whether and when it
might be appropriate to compare/combine the studies' results. The results were displayed as
points on plots of percent of participants that rated each visibility condition acceptable versus the
amount of visibility impairment as measured in deciview units (i.e., a logarithmic transformation
of light extinction). Logit regression analysis was used to develop best fit curves for each of the
four urban area study results and to determine whether they differed significantly from each
other.
Principal Results: Logit regression applied to the results for each of the four urban areas
defined statistically significant relationships for each that are similarly shaped but with different
visibility impairment threshold value, defined here as the 50th percentile acceptability criteria in
1 Available at http://www.epa.gov/ttn/naaqs/standards/pm/sjm 2007 risk.html
5-1
-------
each urban area study. The range of 50th percentile values is from -20 dv to -30 dv, which when
expressed in terms of PMio light extinction corresponds to 64 Mm"1 to 191 Mm"1. The upper and
lower bounds of the range and a mid-point value of 25 dv (corresponding to 112 Mm"1) were
selected as Candidate Protective Levels (CPLs) to compare with the distributions of current
conditions for 15 urban areas (chapter 3) and to define alternative standards to be included in the
assessment (chapter 4).
Technical Issues: Each of the four urban preference study areas produced well defined
though statistically significant different results with respect to what visibility impairment level
divides acceptable from unacceptable conditions. A number of hypotheses concerning why the
results differed for each area are discussed in chapter 2. However, additional research to better
understand why the response distributions differed by location could usefully inform future PM
reviews.
5.2 Current Visibility Conditions
Purpose: The goal of this assessment is to develop a daylight hourly averaged PMio light
extinction dataset for several large urban areas to characterize current visibility conditions and
compare them to the CPLs in order to determine the extent, frequency and causes of visibility
impairment in cases where the CPLs are exceeded.
Approach: A simple linear algorithm, the IMPROVE algorithm,2 was used to estimate
PMio light extinction for 15 urban areas for the period from 2005 to 2007 from hourly PM2 5 and
PMio-2.5 mass and PM2.5 component concentrations and relative humidity data (used to estimate
the water component of the PM under ambient conditions). While PM2.5 mass concentration and
relative humidity data are available from continuous instruments on an hourly averaged basis,
PM2.5 composition data available from the Chemical Speciation Network (CSN) is based on 24-
hour average filter samples which are only collected on a one-day-in-three or one-day-in-six
basis. The methodology used to estimate hourly daylight PM2.5 components involved use of
CMAQ regional air quality modeling to generate monthly averaged species-specific diurnal
patterns (ratio of each hour to the 24-hour mean) for each of the urban areas. These are used to
calculate the hourly relative mix of species that are then scaled so that the sum of the PM2.5
component concentrations equals the measured hourly PM2.s mass concentration. PMio
continuous monitoring data was available at a few of the urban areas permitting hourly PMi0-2.5
concentrations to be determined by subtracting the hourly PM2.5 concentration. Elsewhere PMio-
2.5 was estimated using ratios of PM2.s to PMio. The resulting estimates of hourly averaged PM2.s
component concentrations, PMio-2.5 mass concentration and the measured relative humidity were
2 Malm, et al., 1994 and DeBell, 2006. (See also ISA, section 9.2.2.2, pgs. 9-7 and 9-8.)
5-2
-------
used as input to the IMPROVE algorithm to estimate hourly PMio light extinction to be
calculated for daylight hours with relative humidity no greater than 90%. Resulting daylight 1-
hour averaged PMio light extinction values were compiled on the basis of all hours and
maximum daily values for relative humidity conditions less than or equal to 90% to examine the
frequency that they exceed CPLs by urban area. Tables and plots of the total and component
contributions to PMio light extinction were produced to characterize the nature and causes of
visibility impairment.
Principal Results: All of the results are for 1-hour PMio light extinction during daylight
hours when relative humidity is no greater than 90% for the 15 selected urban areas.
• The use of this relative humidity cap significantly reduced the occurrence of visibility
impairment caused by meteorological conditions like fog and precipitation.
• Maximum daily hourly PM light extinction values exceeded the low, middle and high
CPL 77%, 52% and 26% of the days, respectively, when averaged across the 15 urban
areas. Eastern and California urban areas have the highest frequencies and non-
California Western urban areas have the lowest frequencies above each CPL.
• All hours PMio light extinction values exceeded the low, middle and high CPLs 45%,
22% and 7% of the hours, respectively, averaged across the 15 urban areas, with the
Eastern and California urban areas having the highest frequencies and non-California
Western urban areas having the lowest frequencies above each CPL.
• The range of PMio light extinction values and relative contributions by PM2.s components
for the most impaired 10% of maximum daily 1-hour hours are similar to the most
impaired 2% of all hours, because they include hours from a large number of days in
common.
• During the most visibility impaired hours PM2 5 nitrate is the dominant light extinction
contributor for several western urban areas (Fresno, LA, and Salt Lake City) while sulfate
tends to be the largest contributor in the Eastern urban areas. Carbonaceous PM2.5 (i.e.,
organic mass plus elemental carbon) is a major contributor at Tacoma and a significant
contributor at several other urban areas. Phoenix has significant light extinction
contribution by PMio-2.5. Thus, regional differences in the dominant component
contributing to visibility impairment are apparent.
Technical Issues: The approach used to determine the hourly PMio-2.5 concentrations
varied among the sites. Four of the sites had collocated continuous PM2 5 and PMio
measurements, while six urban areas used data from separate sites to determine the PMio-2.5 by
difference, and the remaining five urban areas used regionally determined ratios of PMio-2.5 to
PM2.5 to infer the PMio-2.5- The quality of the PMio-2.5 data inferred from separate sites for St.
Louis have been called into question in review comments on an earlier version of the UFVA and
are no longer viewed as credible. Therefore, the St. Louis results were appropriately labeled in
5-3
-------
the subsequent analyses in the final UFVA. A similar concern was raised regarding the LA data,
but the PMio-2.5 values are low during most of the visibility impaired days so they are less of an
issue compared with that of St. Louis and thus, were retained in these and subsequent
assessments. PMio-2.5 is a small component of the light extinction for all of the other urban areas
except for Phoenix, where high values are considered more plausible. Collocated continuous
PMio and PM2.5 monitoring at a greater number of urban areas would better address this issue in
the future.
While yielding generally reasonable results, the process employed to develop the hourly
PMio component information used as input to the IMPROVE algorithm is complex and subject
to comments that there are alternate approaches that could have been employed. The use of
regional modeling to generate PM2 5 species specific monthly averaged diurnal patterns would be
unnecessary if continuous PM2.5 speciation monitoring were more readily available. Use of an
algorithm to estimate PMio light extinction would be unnecessary if direct measurements of
continuous PMio light extinction were commonly available.
5.3 Visibility Conditions for Alternative Secondary PM NAAQS Scenarios
Purpose: The goal is to evaluate the effectiveness of alternative secondary PM2.5
NAAQS, including 2 that use the PM2 5 mass concentration indicator (i.e., the current 15 ug/m3
annual, 35 ug/m3 24-hour PM2.5 NAAQS and a more restrictive 12 ug/m3/25 ug/m3 alternative),
and 18 that use a 1-hour daylight PMio light extinction indicator (i.e., all combinations of both a
maximum daily 1-hr indicator and a 1-hr indicator based on all daylight hours with 3 percentile
forms (90th, 95th and 98th) and 3 levels (CPLs) for the 15 urban areas).
Approach: The hourly averaged PMio light extinction data set developed to characterize
current conditions for the 15 urban areas was used as the starting point for a rollback adjustment
to simulate just meeting the various alternative PM NAAQS scenarios. Rollback is not applied
to the PRB portion of the PM2 5 as estimated using CMAQ modeling, though it is applied
uniformly to all other PM2 5 components. This process produces adjusted daylight hourly PMio
light extinction data sets for each urban area that would just meet each of the alternative NAAQS
scenarios that can be assessed with respect to their visibility protection effectiveness.
Principal Results: Each of the PM NAAQS scenarios that used PMio light extinction as
the indicator produced similar distributions of hourly PMio light extinction across the 15 urban
areas. This is not the case for the two scenarios that used PM2 5 mass concentration as the
indicator, where for example the 90th percentile PMio light extinction values vary among the 15
urban areas by as much as a factor of 2 to 3. The maximum daily 1-hour form of the alternative
NAAQS is more restrictive for any percentile than the all-hours form. In fact the 90th percentile
of the maximum daily form is nearly identical with respect to design values for current
5-4
-------
conditions and the percent reduction required to just meet it as the 98th percentile for the all-
hours form.
Technical Issues: The rollback approach implicitly assumes all non-PRB PM2 5
components will be uniformly reduced to meet any of the alternative standards. In practice,
emission control programs that would be developed to meet standards will not operate in this
manner, nor will they be so fine tuned that each urban area would just meet the PM NAAQS. In
that sense, the rollback assessment produces idealized results which for the PMi0 light extinction
based NAAQS scenarios are more uniform across urban areas than is likely should such a
standard be implemented. The use of this nationally uniform emissions rollback approach is
justified by it providing a common basis for assessing the variations in the magnitude of
emissions controls required to meet NAAQS scenarios for urban areas across the country.
5-5
-------
This page left intentionally blank.
-------
6 REFERENCES
Abt Associates Inc., 2001. Assessing Public Opinions on Visibility Impairment due to Air Pollution: Summary
Report. Prepared for EPA Office of Air Quality Planning and Standards; funded under EPA Contract No.
68-D-98-001. Abt Associates Inc., Bethesda, MD. January. Available:
http://www.epa.gov/ttncaaal/tl/reports/vis rptfinal.pdf. Accessed 9/16/2008.
Air Resource Specialists. 2003. 2003 WinHaze Air Quality Modeler, Version 2.9. Air Resource Specialists, Inc.
Available: http://www.air-resource.com/downloads.php. Accessed 9/16/2008.
Arizona Department of Environmental Quality. 2003. Recommendation for a Phoenix Area Visibility Index.
Visibility Index Oversight Committee. Available: http://www.phoenixvis.net/PDF/vis 031403final.pdf.
Accessed 9/16/2008. BBC Research & Consulting. 2002. Phoenix Area Visibility Survey. Draft Report.
Available: http://www.azdeq.gov/environ/air/download/vis 021903f.pdf. Accessed 9/16/2008.
BBC Research & Consulting. 2003. Phoenix Area Visibility Survey. Draft Report. Available:
http://www.azdeq.gov/environ/air/download/vis 021903f.pdf. Accessed 9/16/2008.
DeBell L. (2006). Spatial and seasonal patterns and temporal variability of haze and its constituents in the United States: Report
IV. 1 Cooperative Institute for Research in the Atmosphere. Colorado State University Fort Collins, CO. 156388
Ely, D.W., J.T. Leary, T.R. Stewart, and D.M. Ross. 1991. The Establishment of the Denver Visibility Standard.
Presented at the 84th Annual Meeting & Exhibition of the Air and Waste Management Association. June
16-21.
Frank NH. (2006). Retained nitrate, hydrated sulfates, and carbonaceous mass in federal reference method fine
paniculate matter for six eastern U.S. cities. J Air Waste Manag Assoc, 56: 500-11
Henderson, R. (2006) Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee to the
Honorable Stephen L. Johnson, Administrator, US EPA. Clean Air Scientific Advisory Committee
Recommendations Concerning the Proposed National Ambient Air Quality Standards for Paniculate
Matter. March 21, 2006. Available: http://www.epa.gov/sab/pdf/casac-ltr-06-002.pdf
Henderson, R. (2008) Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee to the
Honorable Stephen L. Johnson, Administrator, US EPA. Clean Air Scientific Advisory Committee
Consultation on EPA's Draft Integrated Review Plan for the National Ambient Air Quality Standards for
Paniculate Matter January 3, 2008.
Jacques Whitford AXYS. 2007. The View Ahead; Identifying Options for a Visibility Management Framework for
British Columbia. Report for the British Columbia Ministry of Environment. Available:
http://www.env.gov.bc.ca/air/airqualitv/pdfs/view ahead.pdf. Accessed 8/5/2009.
Malm WC; Trijonis J; Sisler J; Pitchford M; Dennis RL (1994). Assessing the effect of SO2 emission changes on visibility.
Atmos Environ, 28: 1023-1034.
Molenar, J. V., W.C. Malm, and C.E. Johnson. 1994. Visual air quality simulation techniques. Atmospheric
Environment 28(5):1055-1063.
NARSTO (2004) Paniculate Matter Science for Policy Makers: A NARSTO Assessment. P. McMurry, M.
Shepherd, and J. Vickery, eds. Cambridge University Press, Cambridge, England. ISBN 0 52 184287 5
Pitchford, M. and W. Malm. 1993. Development and applications of a standard visual index. Atmospheric
Environment 28(5): 1049-1054.
6-1
-------
Pryor, S.C. 1996. Assessing public perception of visibility for standard setting exercises. Atmospheric Environment
30(15):2705-2716.
Russell, A; Samet, J.M. (2010). Letter from the Clean Air Science Advisory Committee (CASAC) Ambient Air
Monitoring and Methods Subcommittee (AAMMS) to the Honorable Lisa P, Jackson, Administrator, US
EPA. Review of the White Paper on Particulate Matter (PM) Light Extinction Measurements. April 29,
2010, Available at:
http://vosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/92C9F5AA09A76A938
52577150004 A782/$File/EP A-CASAC-10-010-unsigned.pdf
RWDI AIR. 2008. Final Report: Establishing a Visibility Goal for Wilderness and Urban areas in British Columbia
and Canada. Report to the British Columbia Ministry of Environment. Available:
http://www.env.gov.bc.ca/air/airquality/pdfs/visibility_goal_report_final.pdf. Accessed 8/5/2009.
Samet, J. (2009a) Letter from Clean Air Scientific Advisory Committee (CASAC) to the Honorable Lisa P. Jackson,
Administrator, US. EPA. Consultation on EPA's Particulate Matter National Ambient Air Quality
Standards: Scope and Methods Plan for Urban Visibility Impact Assessment. May 21, 2009. Available:
http://vosemite.epa.gov/sab/sabpeople.nsf/WebCommittees/CASAC.
Samet, J. (2009b) Letter from Clean Air Scientific Advisory Committee (CASAC) to the Honorable Lisa P. Jackson,
Administrator, US. EPA. Review of the Particulate Matter Urban-Focused Visibility Assessment (First
External Review Draft); November 24, 2009. - Available:
http://vosemite.epa.gov/sab/sabpeople.nsf/WebCommittees/CASAC.
Samet, J. (2010a). Letter from Clean Air Scientific Advisory Committee (CASAC) to the Honorable Lisa P.
Jackson, Administrator, US EPA. Review of Particulate Matter Urban-Focused Visibility Assessment
(Second External Review Draft, January 2010). April 20, 2010. Available:
http://yosemite.epa.gov/sab/sabproduct.nsf/OD5CB76AFE7F A77C8525770D004EED55/$File/EPA-
CASAC-10-009-unsigned.pdf
Samet, J. (201 Ob). Letter from Clean Air Scientific Advisory Committee to the Honorable Lisa P. Jackson,
Administrator, US EPA. CASAC Review of Policy Assessment for the Review of the PM NAAQS (First
External Revew Draft, March 2010). May 17, 2010. Available:
http://yosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/E504EE3276D87A9E8
525772700647 AFB/$File/EPA-CASAC-10-011-unsigned.pdf/.
Schmidt. M; Frank, N.; Mintz, D.; Rao, T.; McCluney, L. (2005). Analyses of paniculate matter (PM) data for the
PM NAAQS review. Memorandum to PM NAAQS review docket EPA-HQ-OAR-2001-0017. June 30,
2005. Available: http://www.epa.gOv/ttn/naaqs/standards/pm/sjm_cr_td.html
Smith, A. 2009. Comments on the First External Review Draft of EPA's "Integrated Science Assessment for
Particulate Matter." CRA International, Washington, DC. March 30. Prepared for the Utility Air
Regulatory Group. Submitted as public comment to the public meeting EPA Clean Air Science Advisory
Council. April 2.
Smith, A.E. and S. Howell. 2009. An Assessment of the Robustness of Visual Air Quality Preference Study Results.
CRA International, Washington, DC. March 30. Prepared for the Utility Air Regulatory Group. Submitted
as supplemental material to presentation by Anne Smith to the public meeting of the EPA Clean Air
Science Advisory Council. April 2.
Statistics Canada. 2009a. Population and Dwelling Count Highlight Tables, 2006 Census. Available:
http://wwwl2.statcan.gc.ca/census-recensement/2006/dp-pd/hlt/97-550/Index.cfm7Page = INDX&LANG
= Eng. Accessed 7/13/2009.
6-2
-------
Statistics Canada. 2009b. Population: Chilliwack, British Columbia (Census Agglomeration). Available:
http://wwwl2.statcan.ca/census-recensement/2006/dp-pd/prof/92-591/details/page.cfm7Lang = E&Geol =
CMA&Codel = 930_&Geo2 = PR&Code2 = 59&Data = Count&SearchText = Chilliwack&SearchType
= Begins&SearchPR = 01&B1 = Population&Custom. Accessed 7/13/2009.
Stratus Consulting Inc., 2009. Review of Urban Visibility Public Preference Studies: Final Report. Prepared for EPA
Office of Air Quality Planning and Standards; funded under EPA Contract No. EP-D-08-100 with Abt
Associates Inc., Bethesda, MD. September. Available:
http://www.re gulations.gov/search/Regs/home.html#documentDetail?R=0900006480a296e2
Trijonis, J. C. 1990. Natural Background Conditions for Visibility/Aerosols. Washington (DC): National Acid
Precipitation Assessment Program. NAPAP State of Science and Technology Report no. 24.
http://www.epa.gov/ttncaaal/tl/memoranda/rh envcurhrgd.pdf
US EPA (1996a). Air Quality Criteria for Paniculate Matter. Research Triangle Park, NC: National Center for
Environmental Assessment-RTF Office; report no. EPA/600/P-95/001aF-cF. 3v. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/sjmjr cd.html.
US EPA (1996b). Review of the National Ambient Air Quality Standards for Paniculate Matter: Policy Assessment
of Scientific and Technical Information, O AQPS Staff Paper. Research Triangle Park, NC 27711: Office
of Air Quality Planning and Standards; report no. EPA-452/R-96-013. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/sjmjr_sp.html.
U.S. EPA. (2003) Guidance for Estimating Natural Visibility Conditions Under the Regional Haze Program. U.S.
Environmental Protection Agency, Office of Air Quality Planning and Standards, Research Triangle Park,
NC. EPA-454/B-03-005
US EPA (2004). Air Quality Criteria for Paniculate Matter. National Center for Environmental Assessment, Office
of Research and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC 27711;
report no. EPA/600/P-99/002aF and EPA/600/P-99/002bF. October 2004. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/sjm crcd.html
US EPA (2005). Review of the National Ambient Air Quality Standards for Paniculate Matter: Policy Assessment
of Scientific and Technical Information, O AQPS Staff Paper. Research Triangle Park, NC 27711: Office
of Air Quality Planning and Standards; report no. EPA EPA-452/R-05-005a. December 2005. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/sjm crsp.html
U.S. Environmental Protection Agency. (2007). Draft Integrated Review Plan for the National Ambient Air Quality
Standards for Paniculate Matter. October 2007. U.S. Environmental Protection Agency, Research
Triangle Park, NC, EPA 452/P-08-006. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/sjm 2007jd.html
US EPA (2008a). Integrated Review Plan for the National Ambient Air Quality Standards for Paniculate Matter.
National Center for Environmental Assessment and Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency, Research Triangle Park, NC. Report No. EPA 452/R-08-004. March
2008. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/data/2008 03 final integrated reviewjlaapdf
U.S. EPA (2008b). Integrated Science Assessment for Paniculate Matter: First External Review Draft. National
Center for Environmental Assessment-RTF Division, Office of Air Quality Planning and Standards,
Research Triangle Park, NC. EPA/600/R-08/139 and 139A. December 2008. Available:
http://www.epa.gOv/ttn/naaqs/standards/pm/s jm_2007_isa. html.
US EPA (2009a). Integrated Science Assessment for Paniculate Matter: Final Report. National Center for
Environmental Assessment-RTF Division, Office of Air Quality Planning and Standards, Research
-------
Triangle Park, NC. EPA/600/R-08/139F. December 2009. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_2007_isa.html.
US EPA (2009b). Paniculate Matter National Ambient Air Quality Standards: Scope and Methods Plan for Urban
Visibility Assessment. Office of Air Quality Planning and Standards, U.S. Environmental Protection
Agency, Research Triangle Park, NC. EPA-452/P-09-001. February 2009. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm 2007_pd.html.
US EPA (2009c). Particulate Matter Urban-Focused Visibility Assessment: First External Review Draft. Office of
Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC.
EPA-452/P-09-005. September 2009. Available:
http://www.epa. gov/ttn/naaqs/standards/pm/s_pm_2007_pd.html
US EPA (2009d). Policy Assessment for the Review of the Particulate Matter National Ambient Air Quality
Standards - Preliminary Draft. Office of Air Quality Planning and Standards, U.S. Environmental
Protection Agency, Research Triangle Park, NC. EPA-452/P-09-007. September 2009. Available:
http://www.epa.gOv/ttn/naaqs/standards/pm/s jm_2007_pa.html.
US EPA (20 lOa). Particulate Matter Urban-Focused Visibility Assessment - Second External Review Draft. Office
of Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park,
NC. EPA-452/P-10-002. January 2010. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/sjm 2007 risk.html.
US EPA (20 lOb). Policy Assessment for the Review of the Particulate Matter National Ambient Air Quality
Standards-First External Review Draft. Office of Air Quality Planning and Standards, U.S. Environmental
Protection Agency, Research Triangle Park, NC. EPA 452/P-10-003. March 2010. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_2007_pa.html.
US EPA (2010c). Particulate Matter Urban-1 Focused Visibility Assessment - Second External Review Draft.
Office of Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle
Park, NC. EPA-452/P-10-002. January 2010. Available:
http://www.epa.gOv/ttn/naaqs/standards/pm/s jm_2007_risk.html.
US EPA (20 lOd). White Paper on PM Light Extinction Measurements. Office of Air Quality Planning and
Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC. January 2010. Available
at:
http://yosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/823a6c8842610e76852
5764900659b22!OpenDocument
US EPA (20 lOe). Quantitative Risk Assessment for Particulate Matter - Final Report. Office of Air Quality
Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA-452/R-
10-005. June 2010. Available: http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_2007_risk.html.
6-4
-------
APPENDICES1
A. PM2.s Monitoring Sites and Monitors Providing 2005-2007 Data for the Analysis of
PMio Light Extinction in the 15 Study Areas
B. Distributions of Estimated PMi.s and Other Components under Current Conditions
C. Development of PRB Estimates of PM2.5 Components, PMi0-2.s, and PMio Light
Extinction
D. Relationships between PM Mass Concentration and PMio Light Extinction under
Current Conditions
E. Differences in Daily Patterns of Relative Humidity and PMio Light Extinction between
Areas and Seasons
F. Distributions of Maximum Daily Daylight PMio Light Extinction under "Just Meets"
Conditions
G. Additional Information on the Exclusion of Daylight Hours with Relative Humidity
Greater than 90 Percent
H. Inter-Year Variability
I. Daylight Hours
J. Logit Memorandum
1 When viewing St. Louis results throughout these appendices, the reader should keep in mind that credible
comments concerning unrealistically high PM10_2 5 values were received but too late in the review process to permit
reanalysis using an alternate data set or to remove St. Louis from all portions of this document. However, the text in
the body of this document has been revised to caution readers with respect to the St. Louis results, and they will not
be included in the visibility effects discussion in the final PM Policy Assessment document. Some graphics have
been updated to exclude St. Louis results.
-------
This page left intentionally blank.
-------
APPENDIX A
PM2 5 MONITORING SITES AND MONITORS PROVIDING
2005-2007 DATA FOR THE ANALYSIS OF PM10 LIGHT
EXTINCTION IN THE 15 STUDY AREAS
A-l
-------
PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PM10 Light Extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.S
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Tacoma
AQS ID 530530029
State: Washington
City: Tacoma
MSA: Tacoma, WA
Local Site Name: TACOMA - L
STREET
Address: 7802 SOUTH L
STREET, TACOMA
0.5 miles east of 1-5
2005-2007 annual DV = 10.2
2005-2007 24-hr DV = 43
This is the highest 24-hour PM2.5
DV site in the Seattle-Tacoma-
Olympia, WA annual PM2.5
nonattainment area
Neighborhood Scale
Parameters taken from this site:
. 24-hour FRM PM2.5 mass
(AQS parameter 88101; one-in-
three sampling schedule)
• PM2.5 speciation (one-in-six
sampling schedule)
. 1-hourPM2.5 mass (AQS
parameter 88502, Acceptable
PM2.s AQI & Speciation Mass)
Correlated Radiance Research
M903 Nephelometry
No continuous PM10 monitoring at
this site, see right hand column..
NA
AQS ID 530530031
State: Washington
City: Tacoma
MSA: Tacoma, WA
Local Site Name: TACOMA -ALEXANDER
AVE
Address: 2301 ALEXANDER AVE, TACOMA,
WA
6.4 miles NNE of PM2.5 site
Neighborhood Scale
Parameters taken from this site:
• 1 -hour PM10 STP mass (AQS parameter
81102)
• Sample Collection Method:
INSTRUMENTAL-R&P SA246B-INLET
• Sample Analysis Method: TEOM-
GRAVIMETRIC
7% of PMio-2.s values were determined using
regional average PMio-2.s: PM2.5 ratios from
2005 Staff Paper
A-2
-------
PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PMi0 Light Extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Fresno
AQS ID 060190008
State: California
City: Fresno
MSA: Fresno, CA
Local Site Name: None given
Address: 3425 N FIRST ST,
FRESNO
2.5 miles west of the airport, 3
miles NNE of central Fresno
2005-2007 annual DV= 17.4
2005-2007 24-hr DV = 63
This is not the highest annual or
24-hr PM2.5 DV site in the San
Joaquin nonattainment area.
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM2.5 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.5 speciation (one-in-
three sampling schedule)
• 1-hour PM2.5 mass (AQS
parameter 88501, PM2.5 Raw
Data) Met-One BAM
No continuous PM10 monitoring at
this site, see right hand column..
NA
PMio-2.5 values were determined using
regional average PM10-2.5: PM25 ratios from
2005 Staff Paper
A-3
-------
PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PMi0 Light Extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Los Angeles
AQS ID 060658001
State: California
City: Rubidoux (West Riverside)
MSA: Riverside-San Bernardino,
CA
Local Site Name: None given
Address: 5888 MISSION BLVD.,
RUBIDOUX
Eastern SCAB, 0.4 miles from
Pomona Freeway.
2005-2007 annual DV= 19.6
2005-2007 24-hr DV = 55
This site is not the highest DV site
in the LA-South Coast
nonattainment area.
Neighborhood scale.
Parameters taken from this site:
• 24-hour FRM PM2.5 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.s speciation (one-in-
three sampling schedule)
. 1-hour PM2.5 (AQS
parameter 88502, Acceptable
PM2s AQI & Speciation Mass)
R&P1400TEOM
No continuous PMio monitoring at
this site, see right hand column..
NA
AQS ID 060710306
State: California
City: Victorville
MSA: Riverside-San Bernardino, CA
Local Site Name: MOVED FROM 060710014
Address: 14306 PARKAVE., VICTORVILLE,
CA
36 miles north of PM25 site, on the other side
of a range of hills. 0.4 miles from 1-15
Measurement Scale not given in AQS, but
appears Neighborhood by aerial image.
Parameters taken from this site:
• 1 -hour PM10 STP mass (AQS parameter
81102)
• Sample Collection Method:
INSTRUMENTAL-R&P SA246B-INLET
• Sample Analysis Method: TEOM-
GRAVIMETRIC
6% of PM10-2.5 values were determined using
regional average PM10-2.5: PM2.5 ratios from
2005 Staff Paper
A-4
-------
PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PMi0 Light Extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Phoenix
AQS ID 040137020 (FRM
&CSN)
State: Arizona
City: Scottsdale
MSA: Phoenix-Mesa, AZ
Local Site Name:
Address: 10844 EAST OSBORN
ROAD SCOTTSDALE'AZ
Reporting Agency: Salt River
Pima-Maricopa Indian Community
of Salt River Reservation
Eastern edge of the metro area,
largely surrounded by agricultural
fields.
2005-2007 annual DV = 7.9
2005-2007 24-hr DV = 15
This site is not the highest DV site
in the Phoenix-Mesa CBSA.
Neighborhood Scale
Parameters taken from this site:
. 24-hour FRM PM2.5 mass
(AQS parameter 88101; one-in-
six sampling schedule)
• PM2.s speciation (one-in-
three sampling schedule)
No continuous PM10 monitoring at
this site, see right hand column.
AQS ID 040139998
(Continuous)
State: Arizona
City: Phoenix
MSA: Phoenix-Mesa, AZ
Local Site Name: Vehicle
Emissions Laboratory
Address: 600 N 40th St &
Fillmore St
Measurement Scale not
available; 0.75 miles from
intersection of two
freeways, 1 mile from
Phoenix airport.
Parameters taken from this
site:
• 1-hour PM2.5 mass.
Nephelometer.
AQS ID 040133002
State: Arizona
City: Phoenix
MSA: Phoenix-Mesa, AZ
Local Site Name: CENTRAL PHOENIX
Address: 1645 E ROOSEVELT ST-CENTRAL
PHOENIX STN
1.8 miles NE of central Phoenix
Neighborhood Scale
Parameters taken from this site:
• 1 -hour PMio STP mass (AQS parameter
81102)
Sample Collection Method: INSTRUMENTAL-
R&P SA246B-INLET
Sample Analysis Method: TEOM-
GRAVIMETRIC
2% of PMio-2.5 values were determined using
regional average PM10-2.s: PM2.s ratios from
2005 Staff Paper
A-5
-------
PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PMi0 Light Extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Salt Lake City
AQS ID490353006
State: Utah
City: Salt Lake City
MSA: Salt Lake City-Ogden, UT
Local Site Name: UTM
COORDINATES = PROBE
LOCATION
Address: 1675 SOUTH 600 EAST,
SALT LAKE CITY
2.5 miles SSE of central Salt Lake
City
2005-2007 annual DV= 10.7
2005-2007 24-hr DV = 48
This is not the highest DV site in
the Salt Lake City CSA.
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM2.5 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.5 speciation (one-in-
three sampling schedule)
• 1-hour PM2.5 mass (AQS
parameter 88501, PM2.5 Raw
Data) FDMS-Gravimetric
No continuous PMio monitoring at
this site, see right hand column.
NA
PMio-2.5 values were determined using
regional average PM10-2.5: PM25 ratios from
2005 Staff Paper
Dallas
AQS ID 481130069
State: Texas
City: Dallas
MSA: Dallas, TX
Local Site Name: DALLAS
HINTON
Address: 1415 HINTON STREET
4.5 miles NE of central Dallas
2005-2007 annual DV = 11.5
2005-2007 24-hr DV = 25
This is not the highest DV site in
the Dallas-Ft. Worth CSA.
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM2.5 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.5 speciation (one-in-
three sampling schedule)
• 1-hourPM2.5 mass (AQS
parameter 88502, Acceptable
PM25 AQI & Speciation Mass)
TEOM Gravimetric 50 deg C
No continuous PM10 monitoring at
this site, see right hand column..
NA
PMio-2.5 values were determined using
regional average PM10-2.s: PM25 ratios from
2005 Staff Paper
A-6
-------
PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PMi0 Light Extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Houston
AQS ID 482010024
State: Texas
City: Not in a city
MSA: Houston, TX
Local Site Name: HOUSTON
ALDINE
Address: 4510 1/2 ALDINE MAIL
RD
10 miles NNE of central Houston
2005-2007 annual DV= 13.1
2005-2007 24-hr DV = 25
This is not the highest DV site in
the 'Houston-Baytown-Huntsville,
TX CSA.
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM2.5 mass
(AQS parameter 88101; one-in-six
day sampling schedule)
• PM2.5 speciation (one-in-six
sampling schedule)
• 1-hourPM2.s mass (AQS
parameter 88502, Acceptable
PM2.s AQI & Speciation Mass)
TEOM Gravimetric 50 deg C
No continuous PMio monitoring at
this site, see right hand column.
NA
PMio-2.5 values were determined using
regional average PM10-2.5: PM25 ratios from
2005 Staff Paper
A-7
-------
PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PMio Light Extinction in the 15 Study Areas
Study Area
St. Louis
First PM2.5 Monitoring Site
AQS ID 2951 00085
State: Missouri
City: St. Louis
MSA: St, Louis, MO-IL
Local Site Name: BLAIR STREET
CATEGORY A CORE SLAM
PM2.5.
Address: BLAIR S
2 miles north of central St. Louis
2005-2007 annual DV = 14.5
2005-2007 24-hr DV = 34
This is not the highest DV site in
the St. Louis nonattainment area.
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM2.5 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.5 speciation (one-in-
three sampling schedule)
• 1-hourPM2.5 mass (AQS
parameter 88502, Acceptable
PM2.5 AQI & Speciation Mass)
TEOM Gravimetric 30 deg C
No continuous PM10 monitoring at
this site, see right hand column.
Second PM2.5
Monitoring Site (if
applicable)
NA
PM10 data source for PM10.2.s
AQS ID 295100092 (2005 and 2006
data)
State: Missouri
City: St. Louis
MSA: St, Louis, MO-IL
Local Site Name:
Address: 3 NORTH MARKET
0.7 miles ESE of PM25 site, across the street
from the eastern edge of what appears to be
a recycling/municipal works yard.
Middle Scale
Parameters taken from this site:
• 1 -hour PMio STP mass (AQS parameter
81102)
• Sample Collection Method:
INSTRUMENTAL-R&P SA246B-INLET
• Sample Analysis Method: TEOM-
GRAVIMETRIC Site was on the other
(western) side of the recycling/municipal
works yard as site 295100093, below.
295100093 (2007 data)
Ststs' Missouri
Citv St Louis
MSA: St, Louis, MO-IL
Local Site Name: None given
Address: Branch Street
0.6 miles ESE of PM25 site, across the street
from the western edge of what appears to be
a recycling/municipal works yard.
Middle Scale
Parameters taken from this site:
• 1 -hour PMio STP mass (AQS parameter
81102)
• Sample Collection Method:
INSTRUMENTAL-R&P SA246B-INLET
• Sample Analysis Method: TEOM-
GRAVIMETRIC
4% of PM10-2.5 values were determined using
regional average PM10-2.5: PM2.5 ratios from
2005 Staff Paper
A-8
-------
PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PMi0 Light Extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Birmingham
AQS ID 010730023
State: Alabama
City: Birmingham
MSA: Birmingham, AL
Local Site Name:
Address: NO. B'HAM.SOU R.R.,
3009 28TH ST. NO
2.3 miles north of central
Birmingham
2005-2007 annual DV= 18.7
2005-2007 24-hr DV = 44
This is the highest DV site in the
Birmingham nonattainment area
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM2.5 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.5 speciation (one-in-
three sampling schedule)
• 1-hour PM2.5 mass (AQS
parameter 88502, Acceptable
PM2.5 AQI & Speciation Mass)
TEOM Gravimetric 50 deg C
. 1 -hour PM10 STP mass (AQS
parameter 81102)
Sample Collection Method:
INSTRUMENTAL-R&P SA246B-
INLET
Sample Analysis Method: TEOM-
GRAVIMETRIC
NA
Same as PM25 site.
0.3% of PM10-2.5 values were determined
using regional average PM10-2.5: PM2.5 ratios
from 2005 Staff Paper
A-9
-------
PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PMi0 Light Extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Atlanta
AQS ID 130890002
State: Georgia
City: Decatur
MSA: Atlanta, GA
Local Site Name: 2390-B
WILDCAT ROAD, DECATUR, GA
Address: SOUTH DEKALB
About 7 miles SE of central Atlanta
2005-2007 annual DV = 15.7
2005-2007 24-hr DV = 33
This is not the highest DV site in
the Atlanta nonattainment area.
Neighborhood Scale
Parameters taken from this site:
. 24-hour FRM PM2.5 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.5 speciation (one-in-
three sampling schedule)
• 1-hour PM2.5 mass (AQS
parameter 88502, Acceptable
PM2.5 AQI & Speciation Mass)
TEOM Gravimetric 30 deg C
No continuous PM10 monitoring at
this site, see right hand column.
NA
AQS ID 131210048
State: Georgia
City: Atlanta
MSA: Atlanta, GA
Local Site Name: Georgia Tech, Ford
Environmental Science and Technology Bldg,
roof
Address: GA. TECH., Ford ES&T Bldg, 311
Ferst St NW, Atlanta GA
8.6 miles NW of PM2.5 site
Neighborhood Scale
Parameters taken from this site:
• 1 -hour PM10 STP mass (AQS parameter
81102)
Sample Collection Method: INSTRUMENT
MET ONE 4 MODELS
Sample Analysis Method: BETA
ATTENUATION
8% of PMio-2.5 values were determined using
regional average PMio-2.s: PM2.5 ratios from
2005 Staff Paper
A-10
-------
PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PMi0 Light Extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Detroit
AQS ID 261630033
State: Michigan
City: Dearborn
MSA: Detroit, Ml
Local Site Name: PROPERTY
OWNED BY DEARBORN PUBLIC
SCHOOLS
Address: 2842 WYOMING
About 0.2 miles from Ford River
Rouge auto plant
2005-2007 annual DV= 17.2
2005-2007 24-hr DV = 43
This is the highest annual and 24-
hr DV site in the Detroit
nonattainment area
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM2.5 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.s speciation (one-in-six
sampling schedule)
• 1-hourPM2.s mass (AQS
parameter 88501, PM2.5 Raw
Data) TEOM Gravimetric 50 deg C
• 1 -hour PM10 STP mass (AQS
parameter 81102)
Sample Collection Method:
INSTRUMENTAL-R&P SA246B-
INLET
Sample Analysis Method: TEOM-
GRAVIMETRIC
NA
Same as PM25 site.
2% of PM10-2.5 values were determined using
regional average PM10-2.5: PM2.5 ratios from
2005 Staff Paper
A-ll
-------
PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PMi0 Light Extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Pittsburgh
AQS ID 420030008
State: Pennsylvania
City: Pittsburgh
MSA: Pittsburgh, PA
Local Site Name: None given
Address: BAPC 301 39TH
STREET BLDG #7
3 miles NE of central Pittsburgh,
0.5 miles from Allegheny River
2005-2007 annual DV= 15.0
2005-2007 24-hr DV = 40
This site is not the highest DV site
in the Pittsburgh nonattainment
area.
Urban Scale
Parameters taken from this site:
• 24-hour FRM PM2.5 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.5 speciation (one-in-
three sampling schedule)
• 1-hour PM2.5 mass (AQS
parameter 88502, Acceptable
PM2.5 AQI & Speciation Mass)
TEOM Gravimetric 50 deg C
No continuous PMio monitoring at
this site, see right hand column.
NA
PMio-2.5 values were determined using
regional average PM10-2.5: PM25 ratios from
2005 Staff Paper
Same as PM25 site.
5% of PM10-2.5 values were determined using
regional average PM10-2.5: PM2.5 ratios from
2005 Staff Paper
Baltimore
AQS ID 240053001 (FRM
&CSN)
State: Maryland
City: Essex
MSA: Baltimore, MD
Local Site Name: Essex
Address: 600 Dorsey Avenue
7 miles east of central Baltimore
2005-2007 annual DV = 14.5
2005-2007 24-hr DV = 35
This is not the highest DV site in
the Baltimore nonattainment area.
Neighborhood Scale
Parameters taken from this site:
. 24-hour FRM PM2.5 mass
(AQS parameter 88101; every
day sampling schedule)
• PM25 speciation (one-in-
three sampling schedule)
• 1 -hour PM10 LC mass (AQS
parameter 85101
AQS ID 245100040
(Continuous)
State: Maryland
City: Baltimore
MSA: Baltimore, MD
Local Site Name: Oldtown
Address: Oldtown Fire
Station, 1100 Hillen Street
1 mile NNE of Inner Harbor
area
Middle Scale
Parameters taken from this
site:
• 1-hour PM2.5 mass
(AQS parameter 88502,
Acceptable PM2.5 AQI &
Speciation Mass) TEOM
Gravimetric 50 deg C
A-12
-------
PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PMi0 Light Extinction in the 15 Study Areas
Study Area
Philadelphia
New York
First PM2.5 Monitoring Site
AQSID1 00032004 (DE)
State: Delaware
City: Wilmington
MSA: Wilmington-Newark, DE-MD
Local Site Name: CORNER OF
MLK BLVD AND JUSTISON ST
2.5 miles NE of central
Wilmington, 0.25 miles from the
Delaware River, 22 miles SWfrom
central Philadelphia
2005-2007 annual DV= 14.7
2005-2007 24-hr DV = 37
This is not the highest DV site in
the Philadelphia nonattainment
area
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM2.5 mass
(AQS parameter 881 01; every
day sampling schedule)
• PM2.5 speciation (one-in-six
sampling schedule)
• 1-hour PM2.5 mass (AQS
parameter 88501 , PM2.5 Raw
Data) Beta Attenuation
• 1 -hour PM10 STP mass (AQS
parameter 81 102)
AQS ID 340390004 (NJ)
State: New Jersey
City: Elizabeth
MSA: Newark, NJ
Local Site Name: ELIZABETH
LAB
Address: NEW JERSEY
TURNPIKE INTERCHANGE 13
1 .75 miles south of Elizabeth, at
the I-95 interchange with I-278
2005-2007 annual DV = 14.4
2005-2007 24-hr DV = 42
This is not the highest DV site in
the New York nonattainment area
Neighborhood Scale
Parameters taken from this site:
. 24-hour FRM PM2.5 mass
(AQS parameter 881 01; every
day sampling schedule)
• PM2.5 speciation (one-in-
three sampling schedule)
• 1-hourPM2.5 mass (AQS
parameter 88502, Acceptable
PM25 AQI & Speciation Mass)
TEOM Gravimetric 30 deg C
No continuous PM10 monitoring at
this site, see right hand column.
Second PM2.5
Monitoring Site (if
applicable)
NA
NA
PM10 data source for PM10.2.s
Same as PM25 site.
3% of PM10-2.5 values were determined using
regional average PM10-2.5: PM2.5 ratios from
2005 Staff Paper
AQS ID 36061 01 25
State: New York
City: New York
MSA: New York, NY
Local Site Name: PARK ROW
Address: 1 PACE PLAZA
Near the on-ramp to the Brooklyn Bridge,
Manhattan end
Measurement scale not stated.
Parameters taken from this site:
• 1 -hour PM10 STP mass (AQS parameter
81102)
Sample Collection Method: INSTRUMENTAL-
R&P SA246B-INLET
Sample Analysis Method: TEOM-
GRAVIMETRIC
2% of PMio-2.s values were determined using
regional average PM10-2.s: PM25 ratios from
2005 Staff Paper
A-13
-------
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PMi0 Light Extinction in the 15 Study Areas
Notes:
• In this Table, the 1-hour concentration parameter "88502, Acceptable PM2 5 AQI & Speciation Mass" is the
same as the ISA refers to as "FRM-like" PM25mass. An entry of "88501, PM25Raw Data" indicates that
the monitoring agency makes no representation as to the degree of correlation with FRMPM2 5 mass. The
latter type of continuous PM2 5 data were used only when the former were unavailable.
• Where PM10 was reported in STP, it was converted to LC before PM10.2.5 was calculated.
• All continuous PM2 5 data were obtained through the AirNow data system rather than from AQS, as an
initial exploration indicated that not all the desired 1-hour data from all sites had been submitted to AQS.
Data are submitted to the AirNow system within hours of collection and may not be subject to as much data
validation review as is typical for data in AQS, despite the opportunity offered by the AirNow system for
monitoring agencies to correct data after initial submission.
A-14
-------
APPENDIX B
DISTRIBUTIONS OF ESTIMATED PM2 5 AND OTHER
COMPONENTS
B-l
-------
Figure B-l - Distribution of Daily Maximum PM2.5and PMi0-2.s Across the 2005-2007 Period, by Study Area
(a) Daily Maximum Daylight PM2.5
Daily Maximum PM2.5 (Daylight Hours)
109 304 288
300 263 143 274 346 276 131 268 187 142 219
-r ~r
-8-
T
T T
i
i
t
-r -r T
•A'
/
-------
(b) Daily Maximum Daylight PMio-2.5
Daily Maximum Coarse (Daylight Hours)
109 304 288 86 300 263 143 274 346 276 131 268
142 219
15> co
O
O
o
o
>*
O
O
(N
«,~
B-3
-------
Figure B-2 - Distribution of Hourly PM2.s Components Across the 2005-2007 Period, by Study Area
(a) 1-Hour Daylight Sulfate (dry, fully neutralized)
Sulfate hourly (Daylight Hours)
I
o>
o
c
o
0
1086 3511 M2'9 9S8 3357 3019 1494 3063 3759 2471
1533
O
2816 1873 1463 2286
e
8
T T -•- ^F
•c*"" ^P-
* /
xv ^
^ .-x*
B-4
-------
Figure B-2 - Distribution of Hourly PM2.s Components Across the 2005-2007 Period, by Study Area, continued
(b) 1-Hour Daylight Nitrate (dry, fully neutralized, CSN method consistent)
Nitrate hourly (Daylight Hours)
o
to
o
10
E o
m -*
3
'S
c
o
O
1056 3511 3029 93S 3357 3019 1494 3063 3759 2471 1533 2816 1S73 1463 228J
O
O
*.
B-5
-------
Figure B-2 - Distribution of Hourly PM2.s Components Across the 2005-2007 Period, by Study Area, continued
(c) 1-Hour Daylight Elemental Carbon
Elemental Carbon hourly (Daylight Hours)
I
I
o>
o
c
o
0
1086 3511 3029
3357 3019 1494 3063 3759 2471 1533 2816 1873 1463 2286
O
-*• .J*
-------
Figure B-2 - Distribution of Hourly PM2.s Components Across the 2005-2007 Period, by Study Area, continued
(d) 1-Hour Daylight Organic Carbonaceous Material (by SANDWICH method)
Organic Carbon hourly (Daylight Hours)
CO
E
"
o
CM -
O
o
o
CO
.2 o
13 <°
0)
o
c
o
O 0
o
CM
3759
O
B-7
-------
Figure B-2 - Distribution of Hourly PM2.s Components Across the 2005-2007 Period, by Study Area, continued
(e) 1-Hour Daylight Fine Soil
Soil hourly (Daylight Hours)
o
CM
I
0)
o
3
1086 3511 302'9
3357 3019
O
1494 3063 3759 2471 1533 2816 1873 1463 22S6
O O
-------
APPENDIX C
DEVELOPMENT OF PRB ESTIMATES OF PM2 5
COMPONENTS, PM10_2 5, AND PM10 LIGHT EXTINCTION
Policy relevant background levels of PMio light extinction have been estimated for
this assessment by relying on outputs for the 2004 CMAQ run in which anthropogenic
emission in the U.S., Canada, and Mexico were omitted, as described in the ISA (US EPA,
2009a). Estimates of PRB for PMio light extinction were calculated from modeled
concentrations of PM2.5 components using the IMPROVE algorithm. The necessary
component concentrations were extracted from the CMAQ output files, as they were not
summarized in the ISA (US EPA, 2009a).
More specifically, for each study area, EPA staff overlaid CMAQ grid cells over
shapes representing the Census-defined urbanized area for each study area, and visually
identified the CMAQ grid cells that had a substantial portion of their area coincident with the
urbanized area. For each such grid cell, for each of the 12 months of the year, we obtained
the 24 values of the hour-specific average concentrations of the five PM2.5 components. We
then averaged these across the selected grid cells. Thus, a given hour of the day has the same
PRB estimate for a component on all days within a month, but months and study areas differ.
We generally observed that PRB concentrations did not vary greatly across the several grid
cells overlaying the urbanized area of a given study area; this is reasonable given the
exclusion of local anthropogenic sources from this CMAQ model run. CMAQ estimates of
PRB for the five PM2 5 components averaged across grid cells and months were not adjusted
in any.2
There are too many values of PRB to present or illustrate them comprehensively in
this document. Table C-l presents annual average concentrations by study area to
summarize these PRB estimates for the PM2 5 components (including the specific form
assumed for sulfate, nitrate, and organic carbon). The right hand column of the table shows
the PM2 5 mass calculated from the CMAQ-estimated components, including factors to fully
neutralize sulfate and nitrate (but with no water mass added). One notable feature of the
annual average of the PRB estimates is the relatively high values for elemental and organic
carbon PRB for the Tacoma study area. This area is often affected by wildfires for extended
2 This approach to estimation of PRB for PM2 5 shares the same information source but is more
disaggregated than the approach used in the health risk assessment for this review of the PM NAAQS (US EPA,
2010e). In the health risk assessment, PRB estimates for PM25 mass concentration are taken from the same CMAQ
model run, but are averaged by calendar quarter and by region of the country.
C-l
-------
periods in the autumn months, and such fires were included in the 2004 emissions scenario
for the PRB CMAQ run. A cursory review of information on fire events in 2005-2007
confirmed that the fire situation in this part of the country in 2004 was not an anomaly.
Another notable feature of the PRB estimates is that the values for nitrate and fine
soil/crustal are low relative to previous estimates of natural background concentrations of
these fine PM components in Class I areas. These previous estimates by Trijonis (1990),
repeated in the 2003 EPA guidance document "Guidance for Estimating Natural Visibility
Conditions Under the Regional Haze Program" (US EPA, 2003), are 0.10 ug/m3 for
(neutralized) nitrate and 0.50 ug/m3 for fine soil. These estimates are based largely on data
from the earliest of the IMPROVE monitoring stations, and thus may include some influence
from non-PRB emissions. On the other hand, it is understandable that the unadjusted output
from the PRB CMAQ scenario would underestimate nitrate and fine soil. CMAQ is known
to underestimate actual nitrate in many situations when provided with a complete NOx
inventory, and the nonanthropogenic emission inventory for NOx itself has uncertainties.
The non-anthropogenic emission inventory for the PRB CMAQ run may also quite easily
underestimate nonanthropogenic emissions of fine soil. However, even if the estimates for
PRB nitrate and fine soil were increased to match the Trijonis (1990) estimates, the resulting
values for PRB PMio light extinction would increase only a little. Even at 90 percent relative
humidity, the contribution to PMio light extinction calculated from the Trijonis estimates is
1.7 Mm"1, versus the average of about 0.5 Mm"1 using the estimates in Table C-l. The
increment of 1.2 Mm"1 would be only about 10 to 20 percent of the PRB PMio light
extinction estimates shown in Table C-4, and would not significantly affect the calculation of
light extinction values under the "what if scenarios.
C-2
-------
Table C-l. Summary of PRB Estimates for the Five PM2.s Components: Average 1-Hour
Values Across 2005-2007
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit-Ann
Pittsburgh
Baltimore
Philadelphia
New York City
Average
Average 1-Hour PRB Concentration Across 2005-2007 (ug/m3)
Sulfate
(dry, no
ammonium)
0.45
0.4
0.36
0.31
0.25
0.27
0.3
0.31
0.29
0.3
0.34
0.3
0.34
0.34
0.36
0.33
Nitrate
(dry, no
ammonium)
0.026
0.00062
0.0037
0.000052
0.00028
0.0022
0.0055
0.0027
0.007
0.016
0.00062
0.00052
0.0016
0.00097
0.0038
0.00
Elemental
Carbon
0.15
0.08
0.028
0.02
0.025
0.055
0.091
0.047
0.099
0.1
0.024
0.029
0.039
0.03
0.026
0.06
Organic
Carbonaceous
Material
1.3
0.74
0.3
0.26
0.26
0.59
0.86
0.53
1.1
1.1
0.32
0.36
0.44
0.36
0.31
0.59
Fine
Soil/Crustal
0.31
0.19
0.036
0.015
0.034
0.092
0.17
0.07
0.19
0.19
0.018
0.034
0.054
0.032
0.022
0.10
Calculated
PM25
2.4
1.6
0.9
0.7
0.7
1.1
1.5
1.1
1.8
1.8
0.8
0.8
1.0
0.9
0.9
1.20
c-:
-------
It is also necessary to have estimates of PRB for PMio-2.5, to feed into the IMPROVE
algorithm. It is not EPA's practice to rely on coarse PM estimates from CMAQ modeling, so
other sources of PRB estimates were considered. The final ISA for this review does not
present any new information on this subject. The approach used in the previous two Criteria
Documents was to present the historical range of annual means of PMio-2.5 concentrations
from IMPROVE monitoring sites selected as being least influenced by anthropogenic
emissions. See Table 3E-1 of the 2004 Criteria Document (reproduced here as Table C-3).
For sites in the lower 48 states, these annual means ranged from a low of 1.8 ug/m3 to a high
of 10.8 ug/m3. No cross-year average or median values were provided that could be used as
the point estimates needed in this assessment. Therefore, for this assessment, EPA staff
estimated PRB for PMio-2.5 using a contour map based on average 2000-2004 PMio-2.5
concentrations from all IMPROVE monitoring sites, found in a recent report from the
IMPROVE program (DeBell, 2006). We located each study area's position on this map, and
assigned it the mid-point of the range of concentrations indicated by the contour band for that
location. The contour map is reproduced here as Figure C-l. Stars show locations of the 15
study areas. In this reproduction, the midpoints of the contour ranges have been added to the
legend.
The results for PRB for coarse PM are shown in Table C-2. Lacking any other
information, these PRB values are taken to apply to every hour of the year. The contour map
and thus these values are influenced by data from IMPROVE sites that were not considered
in the 2004 Criteria Document because they are not sufficiently isolated from the influence of
anthropogenic emissions, including three IMPROVE sites in urban areas which clearly are
influenced by anthropogenic emissions, and thus may be overestimates of PRB for coarse
PM. Nevertheless, these values are generally within the range of values presented in the
Criteria Document for the more isolated sites. These values for the more isolated sites are
reproduced here in Table C-3 for ease of comparison. Further, these PRB values are low
enough that their exact values have little effect on the results of "what if estimation of PMio
light extinction levels under possible secondary PM NAAQS.
Table C-4 presents the resulting 2005-2007 average PRB daylight PMio light
extinction by study area, determined by using each daylight hour's f(RH),3 the hour-specific
PRB PM2.s component estimates (summarized only as annual averages in Table C-l), the
PRB PMio-2.5 estimates in Table C-2, and the IMPROVE algorithm. The sulfate and nitrate
component values in Table C-l are multiplied by 1.375 and 1.29 to reflect full neutralization,
before being used in the IMPROVE algorithm. While for conciseness Table C-4 presents
3 Hour-specific relative humidity for PRB conditions was assumed to be the same as measured for current
conditions.
C-4
-------
only the annual average PRM for PMio light extinction for all daylight hours in 2005-2007 in
the rollback analysis of "what if conditions hour-specific PRB values are retained and used.
The values of PRB PMio light extinction in Table C-4 range between 5 and 11 Mm"1.
For comparison, the default estimates of natural visibility conditions in the 2003 EPA
guidance document for Class I areas range between about 15 and 20 Mm"1, including the
Rayleigh contribution of about 10 Mm"1. Thus, on an annual average basis the range of PRB
estimates for PMio light extinction used for this assessment is very consistent with the range
of total light extinction values recommended in the guidance document.
C-5
-------
Figure C-l. Selection of PRB Values for PM10.2.5 Based on Contoured IMPROVE
Monitoring Data
O
Puerto Rico /
Virgin Islands
C-6
-------
Table C-2. Policy Relevant Background Concentrations of PMi0-2.s Used in This
Assessment, Based on Measurements at IMPROVE Sites
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
PRB PM10_2.5 Mass (ug/m3)
4.5
5.5
4.5
5.5
4.5
8.5
5.5
7.5
5.5
5.5
9.5
3.5
3.5
6.5
3.5
Table C-3. Ranges of 1990-2002 Annual Mean PM Concentrations at IMPROVE
Monitoring Sites (ug/m3)
Site
Acadia National Park. ME
Big Bend National Park, XX
Boundary Waters Canoe Area. MN
Bryce Canyon National Park. UT
Bndger Wilderness. WY
C'anyoulands National Park. UT
Denali National Park. AK
Gila Wilderness, NM
Glacier National Park. MT
Lassen Volcanic National Park. CA
Lone Peak Wilderness. UT
Lye Brook Wilderness. VT
Redwood National Park. CA
Three Sisters Wilderness, OR
Voyageurs National Park 1, MN
Voyagetirs National Park 2. MN
Yellowstone National Park 1. WY
Yellowstone National Park 1, WY
PM
Nonsulfate
2.6-4.7
2.7-4.9
2.6-3.9
1.7-2.4
1.5-2.2
1.9-3.2
0.7-2.4
2.4-3.4
3.8-5.5
1.7-4.5
3.1-5.3
2.3-4.8
2.8-4.6
2.0-5.4
3.2-3.5
2.6-5.4
2.0-3.0
1.7-4.1
2.5
(Total)
(49-8.2)
(5.0-7.8)
(4.4-5.8)
(2.6-3.4)
(2.1-2.9)
(2.8-4.0)
(1.1-3.2)
(3.4-4.5)
(4.8-6.5)
(2.1-5.1)
(4.1-6.9)
(4.5-8.8)
(3.6-5.4)
(2.7-6.5)
(5.1-5.9)
(4.1-7.2)
(2.6-3.6)
(2.3-4.7)
Xousulfate
4.6-11.3
8.8-15.7
5.0-10.2
4.4-7.6
3.7-6.5
5.1-10.5
2.0-7.5
4.9-7.9
7.6-14.2
4.0-8.1
7.1-10.9
4.2-9.7
6.0-10.6
4.0-8.1
5.7-11.2
5.2-10.8
6.0-9.2
3.6-9.0
IK
(Total)
(7.3-15.0)
(11.3-18.6)
(7.0-12.0)
(5.3-8.5)
(4.3-7.3)
(6.3-11.7)
(2.4-8.3)
(6.0-9.2)
(8.5-15.2)
(4.6-8.5)
(8.1-12.5)
(7.0-13.6)
(7.2-11.7)
(4.6-9.1)
(8.1-13.1)
(7.0-12.5)
(6.6-9.9)
(4.2-9.6)
Cam-it PM
1 .8-6.0
5.6-10.8
2.3-7.3
2.5-5.6
1.9-4.7
3.2-8.0
1.1-5.6
2.5-5.0
3.7-9.6
1.8-6.4
3.7-6.0
1.6-4.8
3.3-6.5
1.9-4.4
2.8-7.8
2.6-5.3
3.8-7.0
1.9-5.0
Source: Table 3E-1 of the 2004 Air Quality Criteria Document for PM (US EPA, 2004)
C-7
-------
Table C-4. 2005-2007 Average Policy Relevant Background Daylight PM10 Light
Extinction
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
2005-2007 Average Policy Relevant
Background Daylight PMio Light Extinction,
Mm1
11
11
9
8
5
8
10
9
9
10
7
7
8
8
8
-------
APPENDIX D
RELATIONSHIPS BETWEEN PM MASS
CONCENTRATION AND PM10 LIGHT EXTINCTION
UNDER CURRENT CONDITIONS
In the last review, the 2005 Staff Paper (US EPA, 2005) examined the correlation
between PMi0 light extinction and PM2 5 mass concentrations, each defined for various
consistent time periods. The 2005 Staff Paper analysis assumed that the percentage mix of PM2.5
components was the same in all 24 hours of each day, equal to that indicated by 24-hour CSN
sampling. The modeling of 1-hour PMi0 light extinction in this new assessment allows these
correlations to be re-examined, with the more realistic treatment in which the mix of PM2.5
components is modeled to vary during the day, based in part on diurnal profiles from CMAQ
modeling (see section 3.2.2).
Five scatter plot figures relating PM2.5 mass concentrations and PMi0 light extinction are
presented here for the individual study areas, using different time periods for the two parameters;
these time periods are not always matched. In each figure, the solid red curve was estimated by
applying locally weighted scatter plot smoothing (LOESS) to the data. LOESS is a form of
locally weighted polynomial regression (see http://support.sas.com/rnd/app/papers/loesssugi.pdf)
and is a convenient way to visualize whether a dense data cloud in a scatter plot reflects a more
linear or more nonlinear relationship. The LOESS results in each case indicate a generally linear
relationship as a central tendency but with considerable variability around that central tendency.
Table D-l presents squared correlation coefficients between observed and LOESS model-
predicted values for all five figures. Because the LOESS regressions are generally linear,
comparisons among these correlation coefficients should lead to the same qualitative conclusions
as if coefficients from linear regressions were compared. All values of PMio light extinction
presented here are based on excluding daylight hours with relative humidity greater than 90
percent; hence, a nominally 4-hour period might have as few a one 1-hour PMio light extinction
value, although this is rare in this data set (see the tile plots in Figure 3-12). However, values of
PM2.5 mass concentration do not exclude any hours within the time period specified. Note that if
several study areas were grouped by region and combined into a single scatter plot and LOESS
fit, similar to the analysis of this topic in the 2005 Staff Paper, the correlations would be weaker
than observed here for individual study areas.
Figure D-l compares 24-hour PM2 5 mass (as measured by the FRM/FEM filter-based
sampler) to daily maximum daylight PMio light extinction. The scatter is due the variations in
D-l
-------
PM2.5 concentration, in the mix of PM2.5 components, and in relative humidity during the day
and across days. Variations in PMio-2.5 concentrations also contribute to the scatter, in all five
comparisons presented here, since very high levels of PMi0-2.5 substantially influence PMi0 light
extinction. This source of variability in the scatter plots is particularly important for Los
Angeles, Phoenix, and St. Louis which have many (Phoenix) or some (Los Angeles and St.
Louis) hours with high PMio-2.5.
Because of the large scatter and low correlation coefficients when using 24-hour PM2.5
mass concentration to predict daily maximum daylight PMio light extinction, it is natural to
investigate how much the correlation improves when the PM2.5 mass indicator is limited to
shorter periods of time. The next four figures investigate correlations during such shorter
periods, both matched and un-matched in time.
Figure D-2 compares hourly PM2.5 mass (as actually measured by the continuous
instruments) vs. same-hour daylight PMio light extinction. Lack of agreement due to mismatch
of time period is not a factor in this comparison. However, there is still considerable scatter due
to variations in the mix of PM2.5 components and in relative humidity across hours and days. In
addition, continuous PM2 5 mass instruments do not register the mass of each component
consistently with FRM/FEM and CSN samplers and lab analysis methods. This affects the
scatter in this figure because the estimates of hourly PMio light extinction are linked to the
FRM/FEM and CSN measurements more strongly than to the continuous PM2 5 measurements.
Note that the correlation values in Table D-l for this comparison are better than those for the 24-
hour comparison in most but not all study areas. An implication of this figure and the
information in Table D-l is that a wide range of PMio light extinction levels can prevail in hours
that have the same PM2.5 mass concentration, even at a single site. Additional variability no
doubt exists across areas.
Figure D-3 compares 12-4 pm average PM2.5 mass vs. 12-4 pm average PMio light
extinction. The 2005 Staff Paper observed that because this time period is generally the time of
lowest relative humidity, the relationship between PM2 5 mass and PMio light extinction (i.e., the
ratio of the two or the slope of the regression line) is more uniform across areas during this
period than the relationship for values of each averaged over all 24 hours in a day. In addition,
the longer averaging period might be expected to reduce the effect of variability in the
measurement of hourly PM2 5 mass. However, comparison of Figures D-2 (time-matched single
hours) and D-3 (time-matched 4 afternoon hours) and the corresponding columns of Table D-l
indicates that, after exclusion of hours with relatively humidity greater than 90 percent, the
scatter in Figure D-3 is about the same as in Figure D-2. This residual scatter is due to
composition differences from hour-to-hour, as well as to variations in relative humidity during
hours with relative humidity of 90 percent or less. It can also be observed by comparing Figures
D-2
-------
D-2 and D-3 that the period between 12 pm and 4 pm generally has lower levels of PMio light
extinction than for all daylight hours taken together, even after the exclusion of the hours with
the highest relative humidity. (Note the change in scale between these two figures.)
Figure D-4 compares 12-4 pm average PM2.5 mass vs. daily maximum daylight 1-hour
PMio light extinction. This time-unmatched comparison tests the usefulness of a 12-4 pm PM2.5
mass indicator as a predictor of the daily PMio light extinction metric of potentially greatest
interest. The scatter in Figure D-4 is typically more than in Figure D-3 (4 time-matched
afternoon hours), because daily maximum daylight 1-hour PMio light extinction often occurs
earlier in the day than the 12-4 pm period used to average the PM2.5 mass, and the time period
mismatch introduces prediction errors due to changes in PM2.5 concentration and composition
and relative humidity. An implication is that while a secondary NAAQS based on 12-4 pm
average PM2.5 mass might achieve a given level of protection across days and areas in avoiding
high levels of PMio light extinction between 12 and 4 pm, with some variation across areas due
to composition and relative humidity differences, there could be considerable additional variation
in the level of protection against PMio light extinction during the earlier hours of the day when
some areas often have their highest PMio light extinction levels.
Figure D-5 compares 8 am-12 pm average PM2.5 mass vs. daily maximum daylight 1-
hour PMio light extinction. This comparison is of interest because it may reduce the number of
instances of time mismatch, versus the comparison made in Figure D-4, if the daily maximum
PMio light extinction often occurs between 8 am and 12 pm. The scatter in Figure D-5 is
typically less than in Figure D-4 and the squared correlation coefficients larger, indicating that
this earlier averaging period for PM2 5 mass more often encompasses the period of maximum
PMio light extinction. However, the scatter in Figure D-5 is greater than that in Figure D-3 (4
time-matched afternoon hours).
Figure D-6 provides another perspective on the possible use of PM2.5 mass concentration
as an indicator for a secondary PM NAAQS aimed at protecting visual air quality. Figure D-6
shows in box-and-whisker plot form two versions of the ratios of PMio light extinction to PM2 5
mass concentration, allowing a comparison across the 15 study areas of the central tendencies
and the distributions of these ratios. The Panel A version corresponds to the comparison in
Figure D-l (24-hour averages of PM2 5 mass and PMio light extinction) and the Panel B version
corresponds to the comparison in Figure D-2 (time-matched single hour values). The data points
in Figure D-6 were prepared as follows. In each day for each study area, the value of the
indicated PMio light extinction (24-hour average or 1-hour value) was divided by the indicated
PM2.5 concentration metric (24-hour average or 1-hour value). Ratios that reflect PM2.5
concentrations less than 5 |ig/m3 or PMio light extinction less than 64 Mm"1 were eliminated
before plotting, as such data points represent days or hours that could not play any role in
D-3
-------
determining compliance with any of the NAAQS scenarios considered in this assessment; also,
some of these low-concentration/extinction data pairs produced extreme ratios that obscured the
pattern for data pairs of most policy interest. The maximum ratio value for the vertical scale in
these plots is set at 40 to allow closer examination of the portion of the plot representing the bulk
of the data; this prevents a very small number of daily maximum data points for a few study
areas from appearing in Panel A and a very small percentage of 1-hour data points for a few
study areas (Los Angeles and St. Louis in particular) from appearing in Panel B. The notable
variation in the vertical positions of the 25-75 percentile boxes and the 90 percentile whiskers
representing the ratios in the 15 areas illustrates the point that because of differences in PM
composition mix and relative humidity (even after excluding hours with relative humidity greater
than 90 percent) across study areas, a secondary NAAQS based on PM2 5 mass concentration
would not give equal protection in terms of PMio light extinction levels across cities, days, and
hours.
In the first public review draft of this assessment, it was notable that the correlation
values for St. Louis and Philadelphia were much lower than for other areas. In this version
(reflecting both corrections to relative humidity inputs and exclusion of hours with very high
relative humidity) the correlation value for Philadelphia is about that for other eastern areas. The
correlation values for St. Louis remain notably low relative to the average of all areas, for all five
scatter plots. This is likely due to the influence of the high estimated values for PMio-2.s. In
several other cases of notably low correlation, the small available range of PM2.5values relative
to other areas contributes to the lower correlation values, e.g., in Phoenix, Dallas, and Houston.
D-4
-------
Table D-l. Squared Correlation Coefficients between Observed and LOESS
Model-predicted Values of PMio Light Extinction
Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
AVERAGE
Figure D-l
24-Hour
PM2.s Mass
vs. Daily
Maximum
Daylight 1-
Hour PMio
Light
Extinction
0.48
0.76
0.57
0.22
0.88
0.45
0.46
0.40
0.61
0.54
0.62
0.73
0.78
0.61
0.69
0.59
Figure D-2
1-Hour
PM2.s Mass
vs. Same-
Hour PMio
Light
Extinction
0.80
0.83
0.63
0.67
0.89
0.59
0.61
0.43
0.81
0.72
0.55
0.63
0.69
0.61
0.77
0.68
Figure D-3
12-4 pm
Average
PM2.s Mass
vs. 12-4 pm
Average
PMio Light
Extinction
0.78
0.90
0.66
0.73
0.95
0.54
0.62
0.20
0.78
0.80
0.61
0.66
0.69
0.57
0.76
0.68
Figure D-4
12-4 pm
Average
PM2.s Mass
vs. Daily
Maximum
Daylight 1-
Hour PMio
Light
Extinction
0.29
0.69
0.52
0.18
0.80
0.20
0.20
0.18
0.34
0.40
0.11
0.52
0.58
0.39
0.51
0.39
Figure D-5
8 am-12pm
Average
PM2.5 Mass
vs. Daily
Maximum
Daylight 1-
Hour PMio
Light
Extinction
0.65
0.83
0.69
0.20
0.89
0.36
0.30
0.36
0.44
0.70
0.30
0.62
0.71
0.50
0.62
0.54
D-5
-------
Figure D-l. Relationship Between 24-Hour PMi.s Mass vs. Daily Maximum Daylight 1-Hour PMio Light Extinction.
0 20 40 GO 80
0 20 40 GO 80
Pittsburgh. PA
Salt Lake City. UT
Tacoma. WA
Fresno. CA
Houston. TX
Los Angeles, CA
New York. NY
Philadelphia, PA
§
"•§
Atlanta. GA
Baltimore. MD
Birmingham. AL
Dallas. TX
Detroit. Ml
0 20 40 60 SO
0 20 40 60
Pm2.5 Daily average
0 20 40 60 80
D-6
-------
Figure D-2. Relationship Between Daylight 1-Hour PM2.s Mass vs. Same-Hour PMi0 Light Extinction.
Phoenix, AZ
Pittsburgh. PA
Salt Lake City. UT
Tacoma, WA
Houston TX
Los Angeles. CA
NewYorfc, NY
Philadelphia, PA
•Vr
Baltimore, MD
Birmingham. AL
Dallas. TX
50 100 150 200
50 100 150
PM2.5 (ug/m3)
50 100 150 200
D-7
-------
Figure D-3. Relationship Between 12-4 pm Average PM2.s Mass vs. 12-4 pm Average PMi0 Light Extinction.
0 20 40 60 30 100 120
0 20 40 60 30 100 120
Phoenix. AZ
Pittsburgh. PA
Salt Lake City. UT
St Louis. IL
Tacoma. WA
• *•'.
t •
v!.
Fresno. CA
Los Angeles. CA
New York. NY
Philadelphia. PA
- 200
Atlanta, GA
Baltimore. MD
Birmingham. AL
Dallas, TX
Detroit, Ml
0 20 40 60 80 100 120
0 20 40 GO 30 100 120
Pm2.5 12-4 average
0 20 40 60 80 100 120
D-8
-------
Figure D-4. Relationship Between 12-4 pm Average PM2.s Mass vs. Daily Maximum Daylight 1-Hour PMi0 Light
Extinction.
0 20 40 GO 80 100 120
0 20 40 GO 30 100 120
Phoenix. AZ
Pittsburgh. PA
Salt Lake City. UT
St. Louis, IL
Tacoma. WA
Los Angeles. CA
New York. NY
Philadelphia. PA
I
a
- 600
400
Atlanta. GA
Baltimore. MD
Birmingham. AL
Detroit. Ml
0 20 40 60 80 100 120
0 20 40 60 80 100 120
Pm2.5 12-4 average
0 20 40 60 BO 100 120
D-9
-------
Figure D-5. Relationship Between 8 am-12 pm Average PM2.s Mass vs. Daily Maximum Daylight 1-Hour PMi0 Light
Extinction
Phoenix, AZ
Pittsburgh. PA
Salt Lake City. UT
St Louis. IL
Tacoma. WA
200
0
Los Angeles, CA
New York. NY
Philadelphia, PA
I
8
1000
-200
- 0
Atlanta. GA
Baltimore. MD
Birmingham. AL
Detroit. Ml
50 100
Pm2.5 8-12 average
D-10
-------
Figure D-6. Distribution of Ratios of 1-Hour PMi0 Light Extinction and PM2.5 Mass
Concentration.
A - Ratios of Daily Maximum Daylight 1-Hour PMio Light Extinction to 24-Hour Average
PM2.5 Concentration.
T
-4-
4-
T
4-
T
T
T
D-ll
-------
B - Ratios of Daylight 1-Hour PMi0 Light Extinction to Same-Hour PM2.s Concentration
Ratio bext (^64) (Daylight Hours) to pm25avg (=-=5)
i i
ill
inn
T T
T
T
D-12
-------
APPENDIX E
DIFFERENCES IN DAILY PATTERNS OF RELATIVE
HUMIDITY AND PM10 LIGHT EXTINCTION BETWEEN
AREAS AND SEASONS
In the last review of the secondary PM NAAQS, the pattern of PMio light extinction
during the day was of particular interest. It was noted, using estimates of hourly PMio light
extinction based on a simpler approach than described for this analysis, that both (1) mid-day
PMio light extinction and (2) the slope of the relationship between PMio light extinction and
PM2 5 concentration varied less among regions of the country that at other times of the day.
This was attributed to greater homogeneity of relative humidity across regions in the mid-day
period. This is in contrast to the situation in the morning and later afternoon hours, when
more eastern areas typically experience higher relative humidity levels than the more arid
western and southwestern areas. The current analysis allows these patterns to be re-
examined.
Figures E-l through E-4 show the diurnal pattern of season-average, hour-specific
PMio light extinction and relative humidity for the four "daylight seasons." These graphics
exclude hours with relative humidity greater than 90 percent. Light extinction and relative
humidity for a given clock hour are averaged across the days in the season, across all three
years. Daylight hours (per the simplified schedule of Table 3-5) are indicated by solid
circles. Average 1-hour PMio light extinction generally is highest in the morning,
corresponding to higher relatively humidity (mostly due to lower temperature), higher
vehicle traffic, and less dispersive conditions than later in the day. As was observed in the
last review, there is more variation in average 1-hour PMio light extinction among areas in
the morning than at mid-day, although the morning variation has been reduced (relative to
same information in the first public review draft of this assessment) by the exclusion of hours
with relative humidity greater than 90 percent.
E-l
-------
Figure E-l. Diurnal and Seasonal Patterns of Relative Humidity (percent) and PMi0 Light Extinction (Mm *) for 2005-2007
(a) November-January
Phoenix, AZ
Pittsburgh, PA
81
8 -
«
fc -
1
(
JJ~ ~« -_r
V ,A£-C
mv^f^S^ rftf^
) 6 12 17 2
Hour
r § 8 -i
. « a -
« ,
- s ^ -
3 C
-"^NX^^
6 12 17 2
Hour
O CO
- s t -
3 <
twxccv rfP015
*S^
I jpH
i^\ k -rv5^
CP
'""1 l""l
B 12 17 2
Hour
r § 8n
r R g -,
L S E J
0 0
3 <
•rrtfCcrx
"^X^^
w_*
ojTiH ^^^rf"-**3
6 12 17 2
Hour
O n
r« 8J
- s t J
3 (
f^^rf^,^^
to »
-rfX_ ^ ^^ '
0 ^V**^
6 12 17 2
Hour
r §
"- R
i/>
0
3
Fresno, CA
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
{
10 (
s -
R -
s -
a
P 0
,-gjnr^ o
w
" i I-
L fi ^ J
^ to t
^. ^
%%«»P
- s i-
- K? S -
i P" (|?5
- ° E J
>- s s -
otctP «v=° *
r S s -i
r ° E -
- s § -
^s^
°"S_
ffbff^XgiariW
_ a jrt -,
- s E -
JL irt « j
fN W
tf««V^»
^^^^^
J™ h-
- s i • RH
_ tf>
•Ext
0 6 12 17 23 0 6 12 17 23 0 S 12 17 23 0 6 12 17 23 0 6 12 17 23
Hour Hour Hour Hour Hour
Atlanta, GA
Baltimore, MD
Birmingham, AL
Dallas, TX
Detroit, Ml
g"
^ H
t
»#«a*% x^
«LO
. "
>^"^
_
6 12 17 2
Hour
r* 8-
1
- S E -
3 (
N»^
mt£P^_ sm
V^89
1 6 12 17 2
Hour
•« ^"
<
CNJ CO "j
3 (
"'^^Nw «&
^fffspy
tvP**^
e 12 17 2
Hour
- K
3
E-2
-------
Figure E-2. Diurnal and Seasonal Patterns of Relative Humidity (percent) and PMi0 Light Extinction (Mm *) for 2005-2007,
continued
(b) February-April
Phoenix, AZ
- £ -
Pittsburgh, PA
Salt Lake City, UT
St. Louis, IL
Tacoma, WA
SjJ"^ ^^- j*
CDxixK%SVr^
3 6 12 17 2
Hour
Fresno, CA
""X /
V^_%W o°
^^^« rP
) 6 12 17 2
Hour
Atlanta, GA
(£(X&^* ,-d
.«W>
3&&P w» ^°
X^*^
3 6 12 17 2
Hour
« &
r-- CM ~
I
-a £ -
L a 87
3 (
- 1 8-,
-B ^
j- S £ -
3 (
- 1 S-i
r K 8-
-8 £ -
>- a s i
3 (
«P**
^W3*'
gnffrf\, rfrft
S 12 17 2
Hour
Houston, TX
• c*^
%-rf>
%>r
* .
>fnnnI«V ^s
-V^P"
rdao*1^^^ ^_nKft
6 12 17 2
Hour
(
If) CO
~ fO »
3 £
I
r § §n
i
- ° £ -
3 C
I
o n
- s £ -
<
f- S S -
3 (
VI?
^w^X^Bd5^
6 12 17 2
Hour
.os Angeles, C/
«w\ C'*«^X rfi
^^^
6 12 17 2
Hour
Birmingham, AL
ixP^b
p^N*.*
• ,fp**
v*^^
6 12 17 2
Hour
- s £ -
l/j CD <
r M co ~
3 (
I
r 1 8-1
r« »-
10 £
3 <
r § si
- s £ -
' 1/5 CO _|
~ CM 03
3 (
^X^
^ta^
tcK >^%>JLI 1
6 12 17 2
Hour
New York, NY
^
*^^_
-^^. ^Y<
S 12 17 2
Hour
Dallas, TX
«^V
N%^
»W^\, _J«"
>»«^
( 6 12 17 2
Hour
- s 8 J
S ^
L S 8 -
3 I
- 1 8-i
-s 8-
~ ° £ -
3 (
O CO
- s 8-
- S £ -
1/1 CO !
~ CM «O ~
3 (
*"NS^
W^-jii*^
6 12 17 2
Hour
Philadelphia, Pf
^
\/
>**rftt^*l«*rKi><
6 12 17 2
Hour
Detroit, Ml
<**x ^
^mtfrtxp
6 12 17 2
Hour
u>
~ r-
s c
- 8
*
3
i
r -
- H
- S ri • RH
• Ext
3
r §
7K
-Si
L a
3
E-3
-------
Figure E-3. Diurnal and Seasonal Patterns of Relative Humidity (percent) and PMi0 Light Extinction (Mm *) for 2005-2007,
continued
(c) May-July
Phoenix, AZ
Pittsburgh, PA
Salt Lake City, UT
St. Louis, IL
Tacoma, WA
I-
^ R -J
S i
(
8 1
i-
3 K -
(
8J
^ £ -
i
'^N^^rf
tte^Ohad^
) G 12 17 2
Hour
Fresno, CA
^
V p°d
L ^ ^^
*"**"~-iii._j-
) 6 12 17 2
Hour
Atlanta, GA
-\ ,
VV. r*
) 6 12 17 2
Hour
i- rt
- s 8 -
SJ <
- s £ -
1
Lf) CO
r CM o3
3 (
r 1 S -i
-R »•:
>- S £ -
3 (
o n
- R % "
- s £ -
^ S3 -
3 (
"X x
,JftA " ^
\-*^
S 12 17 2
Hour
Houston, TX
w*. /
* *
%«W»
J»Y* ^^^^-XJn
6 12 17 2
Hour
Baltimore, MD
frrf^» rf*
"• - -^P
Vl^*"'
6 12 17 2
Hour
m co
CM ffl
4
3 I
I
r § §n
'-* Rn
10 £ ~?
3 C
I
O c*l
1 s £ -;
m eo
3 (
% iSfd
^^^
TO -mi agn. CP
8 12 17 2
Hour
.os Angeles, Cf
fV ^ • •
>v VO3
6 12 17 2
Hour
Birmingham, AL
***•• rf^
^ *• ^°
^I^M*
6 12 17 2
Hour
-8 8 -
- s fe -
^5 CO '
~ M CO ~
3 (
I
r I 8-1
r8 8-
r ° £ -
3 <
r § gi
r s 8-
i
h S £ -
!_ m co
3 (
**v
\ /
>ta^"
6 12 17 2
Hour
New York, NY
~\S*
f»
y&T •^ , , i-fy
S 12 17 2
Hour
Dallas, TX
£****•
•"w^3
"Jir-
J 6 12 17 2
Hour
S ^:
L & S -
3 I
- 1 8-1
rs ^
"~ to ^
3 (
o n
L s fe -
irt CO 1
3 (
"\y
%^
0*1*.
^lllMB^t^
6 12 17 2
Hour
Philadelphia, Pf
"*\ /
«r! JD
6 12 17 2
Hour
Detroit, Ml
o%X«X
[>" ^ ^«trm
^Sfl^fS
6 12 17 2
Hour
>- R
K az
- S
3
i
r -
- S ri • RH
i EM
• Ext
3
-R
-Si
3
E-4
-------
Figure E-4. Diurnal and Seasonal Patterns of Relative Humidity (percent) and PMi0 Light Extinction (Mm *) for 2005-2007,
continued
(d) August-October
Phoenix, AZ
Pittsburgh, PA
Salt Lake City, UT
St. Louis, IL
Tacoma, WA
g -
8-
(
s J
<
(
8~l
i-
R -
a i
•
^1
8J
fc J
i
s -
i
«S*\
^^^f^
) 6 12 17 2
Hour
Fresno, CA
^.
* O
%' P°
^x*^0
) 6 12 17 2
Hour
Atlanta, GA
'V/0
9? ^*"
) 6 12 17 2
Hour
C? C9
- s 8-
"• S fc -
- S 8 -
3 C
- § 3n
-R s-
" ° - ~
_ m GO
?^ CM 03 J
9 C
- § 8n
^ 8-
- s E -
irt so
CM tQ
3 <
"\X
tftrf** _^p.
^•ll^0^
6 12 17 2
Hour
Houston, TX
"*•. /
• 0
to Xw-^
6 12 17 2
Hour
Baltimore, MD
,^-s^^
coH^^^^^^
6 12 17 2
Hour
O CO
r- R 8-
L S t J
Ln oo
"CM » ~
$
3 {
I
r— O lO — i
*- rt
'- e s§-
- s E -
F- s s -
3 (
o «
-8 a -
^ R ^ J
- S S -;
irt cd
~ CM CO
3 (
*• <£***
^Sr
taaJ^*t+,^fK&
e 12 17 2
Hour
.os Angeles, Ct
***** ^
Pa/ %X* jD
*^W^"
'""I™"!""!1""
B 12 17 2
Hour
Birmingham, AL
""A, X
033* • ^*^ f>°°
^^•^^
6 12 17 2
Hour
O CO
- e 8-
•- s E J
- s s -
3 <
t
r § §n
r« 8.
- ° E -
-a § -
3 (
- 1 S3 -
|
, s ^
irt ca
CM OS i
3 (
**% x
frfp0 x
«^_ _Jpxu'
6 12 17 2
Hour
New York, NY
_Vj£^
(fit,
w5 ^^^
6 12 17 2
Hour
Dallas, TX
r^m
vv«^
ntf^»
Nw^0^
» 6 12 17 2
Hour
8 S?
. B a -
5 r^ oj
- s t J
-a s -
3 (
- § g "I
rs »-:
- s E -
>- s s -
3 (
- 1 s -
- e 8-
- s £ -
- s s -5
3 (
fOH\ 0
*•
-------
This page left intentionally blank.
-------
APPENDIX F
DISTRIBUTIONS OF MAXIMUM DAILY AND HOURLY
DAYLIGHT PM10 LIGHT EXTINCTION - UNDER "JUST
MEET" CONDITIONS
F-l
-------
(a) NAAQS Scenario
Daily Max
191 Mm -1
90th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackDailyMaxNAAQSI 91 PctBODVsFromdaily.max
109 324 300 9S 30G 273 153 2B9 349 27g
277 131
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackLowRHDayHoursNAAQS! 91 PctBODVsFromdaily.max
1086 3511 3029 1108 3357 3019 1576 3063 3759 2471 1533
1873 1463 2286
S
H
S
n$- -P
-------
(b) NAAQS Scenario
Daily Max
191 Mm -1
95th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackDailyMaxNAAQSI 91 PctBSDVsFromdaily max
109 324 300
306 273 158 289 349 279
277 181 143 225
g
is
g -
•r •
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackLowRHDayHoursNAAQS! 91 Pctl95DVsFromdaily.max
^
F-2
-------
(c) NAAQS Scenario
Daily Max
191 Mm -1
98th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackDailyMaxNAAQSI 91 PctBSDVsFromdaily max
109 324 300
306 273 158 289 349 279
277 181 143 225
g
is
g -
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackLowRHDayHoursNAAQS! 91 Pctl98DVsFromdaily.max
F-4
-------
(d) NAAQS Scenario
Daily Max
112 Mm -1
90th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExIRollbackDailyMaxNAAQS! 12Pctl90DVsFromdaily max
109 324 300
306 273 158 289 349 279
277 181 143 225
g
is
g -
....... p.....
J_ --,&_-—-J—t
—a b- - ffj—
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExIRollbackLowRHDayHoursNAAQS! 12Pctl90DVsFromdaily.max
-------
(e) NAAQS Scenario
Daily Max
112 Mm -1
95th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExIRollbackDailyMaxNAAQS! 12Pctl95DVsFromdaily max
109 324 300
306 273 158 289 349 279
277 181 143 225
g
is
g -
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackLowRHDayHoursNAAQS! 12Pctl95DVsFromdaily.max
-------
(f) NAAQS Scenario
Daily Max
112 Mm -1
98th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExIRollbackDailyMaxNAAQS! 12Pctl98DVsFromdaily max
109 324 300
306 273 158 289 349 279
277 181 143 225
g
is
g -
•r •
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackLowRHDayHoursNAAQS! 12Pctl98DVsFromdaily.max
F-7
-------
(g) NAAQS Scenario
Daily Max
64 Mm-1
90th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackAIIDayHoursNAAQS64Pctl90DVsFromdaily.max
123B 3643 33B3
3457 3106 1736 3273
3262 1567 3179 2095 161B 2515
g
is
g -
^ .v" V
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExlRollbackLowRHDayHoursNAAQS64Pctl90DVsFromdaily.max
F-8
-------
(h) NAAQS Scenario
Daily Max
64 Mm-1
95th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackAIIDayHoursNAAQS64Pctl95DVsFromdaily.max
g
IS
g -
123B 3643 33B3
3457 3106 1736 3273
3262 1567 3179 2095 161B 2515
•r •
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExlRollbackLowRHDayHoursNAAQS64Pctl95DVsFromdaily.max
-------
(i) NAAQS Scenario
Daily Max
64 Mm-1
98th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackAIIDayHoursNAAQS64Pctl98DVsFromdaily.max
123B 3643 33B3
3457 3106 1736 3273
3262 1567 3179 2095 161B 2515
g
is
g -
^ .v" V
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExlRollbackLowRHDayHoursNAAQS64Pctl9BDVsFromdaily.max
F-10
-------
(j) NAAQS Scenario
All hours
191 Mm -1
90th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackDailyMaxNAAQSI 91 Pctl90DVsFromall.hours
109 324 300
306 273 158 289 349 279
277 181 143 225
g
IS
g -
•r •
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackLowRHDayHoursNAAQS! 91 Pctl90DVsFromall.hours
r>*
-------
(k) NAAQS Scenario
All hours
191 Mm -1
95th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackDailyMaxNAAQSI 91 Pctl95DVsFromall.hours
109 324 300
306 273 158 289 349 279
277 181 143 225
g
IS
g -
T
•r •
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackLowRHDayHoursNAAQS! 91 Pctl95DVsFromall.hours
I 8
tl
*• ^ ^
1 ? ^ ^ ^f ^
x ^ ./- x ^
* ^
-------
(1) NAAQS Scenario
All hours
191 Mm -1
98th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackDailyMaxNAAQSI 91 Pctl98DVsFromall.hours
109 324 300
306 273 158 289 349 279
277 181 143 225
g
is
g -
i
•r •
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackLowRHDayHoursNAAQS! 91 Pctl98DVsFromall.hours
^
F-13
-------
(m) NAAQS Scenario
All hours
112 Mm -1
90th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackDailyMaxNAAQS112Pctl90DVsFromall.hours
109 324 300
306 273 158 289 349 279
277 181 143 225
g
is
g -
•r •
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackLowRHDayHoursNAAQS! 12Pctl90DVsFromall.hours
^
F-14
-------
(n) NAAQS Scenario
All hours
112 Mm -1
95th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackDailyMaxNAAQS112Pctl95DVsFromall.hours
306 273 158 289 349 279
277 181 143 225
20
00
ght Extincti
600
40
00
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackLowRHDayHoursNAAQS! 12Pctl95DVsFromall.hours
40
I
o o | 8 °
|±Ji±iitit|(i(|
«^
F-15
-------
(o) NAAQS Scenario
All hours
112 Mm -1
98th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackDailyMaxNAAQS112Pctl98DVsFromall.hours
109 324 300
306 273 158 289 349 279
277 181 143 225
g
IS
g -
•r •
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackLowRHDayHoursNAAQS! 12Pctl98DVsFromall.hours
^
F-16
-------
(p) NAAQS Scenario
All hours
64 Mm-1
90th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackDailyMaxNAAQS64Pctl90DVsFromall.hours
109 324 300
306 273 158 289 349 279
277 181 143 225
g
is
g -
•r •
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExlRollbackLowRHDayHoursNAAQS64Pc«90DVsFromall.hours
F-17
-------
(q) NAAQS Scenario
All hours
64 Mm-1
95th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackDailyMaxNAAQS64Pctl95DVsFromall.hours
109 324 300
306 273 158 289 349 279
277 181 143 225
g
is
g -
O O
•r •
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExlRollbackLowRHDayHoursNAAQS64Pc«95DVsFromall.hours
i::
*
F-18
-------
(r) NAAQS Scenario
All hours
64 Mm-1
98th percentile
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackDailyMaxNAAQS64Pctl98DVsFromall.hours
109 324 300
306 273 158 289 349 279
277 181 143 225
g
is
g -
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
ExtRollbackLowRHDayHoursNAAQS64Pc«98DVsFromall.hours
*• o/ ^
1 ? ^ ^ ^f ^
x ^ ./- x ^
* ^
-------
(s) NAAQS Scenario
15 ug/m3 annual
35 ug/m3 24-hour
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
PMHollbackDailyMaxCasel NAAQS
I i ! i
Displayed: Hourly Daylight PM10 Light Extinction (excluding hours >90% RH)
PMRollbackCasel NAAQS
g -
1086 3511 3029 110S 3357 3002 1576 3063 3759 2471 1533
1S73 1463 2286
,
F-20
-------
(t) NAAQS Scenario
12 ug/m3 annual
25 ug/m3 24-hour
Displayed: Daily Max Daylight PM10 Light Extinction (excluding hours >90% RH)
PMHollbackDailyMaxCase2NAAQS
^ 90% RH)
PMFtollbackCase2NAAQS
1086 3511 3029 1108 3357 3019 1576 3063 3759 2471 1533
1873 1463 2286
F-21
-------
This page left intentionally blank.
-------
APPENDIX G
ADDITIONAL INFORMATION ON THE EXCLUSION OF
DAYLIGHT HOURS WITH RELATIVE HUMIDITY
GREATER THAN 90 PERCENT
G-l
-------
This appendix provides detailed information related to the exclusion of daylight hours
with relative humidity greater than 90 percent from the design value formula for the secondary
NAAQS scenarios based on PMi0 light extinction, as discussed in section 3.3.5. As described in
that section, these hours have also been excluded from graphical displays of the distribution of
PMio light extinction under current conditions and the various NAAQS scenarios, and from the
denominator of percentages of day or hours (as in Table 4-7).4
Table G-l shows how many estimates of 1-hour daylight PMio light extinction were
excluded, based both on individual hours and on days that were affected by the exclusion of one
or more daylight hours. Phoenix was not affected at all. Among the other areas, Detroit was the
least affected. For all areas, comparison of the percentage of hours affected to the percentage of
days affected indicates that several hours with high relatively humidity tend to occur in the same
day, rather than being evenly distributed across all days. For example, in Atlanta 24 percent of
daylight hours have relative humidity greater than 90 percent, which corresponds to about 876
hours per year (assuming there were data for every day of the year and given that on average
there are about 10 fully daylight hours per day). However, only 80 percent of the days
(corresponding to 292 days, if there were data for every day of the year) are affected. Thus, on
average, an affected day in Atlanta has about 3 affected hours. The tile plots in Figure 3-12 also
illustrate the tendency for hours with high PMio light extinction to cluster in some days.
Figure G-l shows when during the daylight hours these hours with relative humidity
greater than 90 percent occurred, prior to the their exclusion. Some but not all areas have a
strong tendency for the affected hours to be in the morning. The counts in this figure are across
all the days in 2006-2008 that have estimates of PMio light extinction, not all the actual calendar
days in that three year period. Given the regularity of the monitoring schedules, these results
should represent year-round conditions reasonably well. However, the estimates of PMio light
extinction for Phoenix and Houston are not seasonally balanced due to one calendar quarter with
no data in each case (see Table 3-4), so the true year-round time-of-day distributions of excluded
hours for these two areas may be somewhat different than shown here.
Figure G-2 contrasts the distribution of daylight PMio light extinction estimates before
and after the exclusion, based on both daily maximum values and all daylight hourly values
individually. The differences observable in the figure are consistent with the information on the
percentages of hours and day affected in the study areas. In most cases, the highest values of
light extinction are notably lower after exclusion, on both a daily maximum basis and individual
4 This appendix was prepared prior to the discovery of the SANDWICH processing error noted in the
footnote on page 3-22 of the main document, and has not been updated to incorporate that correction. Values for
PM10 light extinction in Figure G-2 and Table G-2 are therefore slightly inconsistent with values presented in the
main report, but this should have a negligible effect on the comparisons presented.
G-2
-------
hour basis, indicating that PM concentrations in some of the excluded hours are fairly high. If
only low-PM hours were excluded by the relative humidity screen, the highest values of PMio
light extinction would not have been affected.
Finally, Table G-2 contrasts PMio light extinction design values before and after the
exclusion, for the 90th and 95th percentile forms based on daily maximum daylight 1-hour PMio
light extinction, for current conditions. (A similar comparison for the 98th percentile form was
not generated.) As expected, design values are notably lower after the exclusion. For both
percentile forms, the largest reduction is in Los Angeles (represented by the Rubidoux site in the
far eastern part of the South Coast Air Basin). Phoenix had no hours with relative humidity
greater than 90 percent, and accordingly Table G-2 shows that its PMio light extinction design
values are not affected by the exclusion. Similarly, Detroit and Dallas had only a few hours with
relative humidity greater than 90 percent, and their design values are affected very little by the
exclusion.
Table G-l. Percent of Daylight Hours and Days Affected by the Elimination of
Hours with Relative Humidity Greater Than 90 Percent
Study Areas
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Percent of Daylight Hours
Excluded
12.3
3.6
10.6
0.0
2.9
2.8
9.6
6.4
4.4
24.1
2.3
11.4
10.6
9.6
9.1
Percent of Days with at
Least One Daylight Hour
Excluded
49.1
15.7
49.7
0.0
13.7
12.8
40.9
21.1
19.1
80.7
7.1
41.2
33.2
31.7
22.4
-------
Figure G-l. Distribution by Time of Day of Eliminated Daylight Hours with Relative Humidity Greater Than 90 Percent.
.1
1
Fresno, CA
llll.-
Atlanta, GA
Illllllll.
llllllll -
Houston, TX
Illlli... .
Baltimore, MD
• Illll
.11 .
Los Anqeles, CA
llll,
Birmingham, AL
•••••••••••-•_
•llllllll. I. •_
New York, NY
.••Imiiiiii-
Dallas, TX
-•llll..__. _
Philadelphia, PA
Detroit, Ml
0506 07 08 09 10 11 12 13 14 15 16 17 18 05 0607 0809 10 11 12 13 14 15 16 17 18 0506 07 0809 10 11 12 13 14 15 16 17 18 05 06 070809 10 11 12 13 14 15 16 17 18 05 0607 0809 10 11 12 13 14 15 16 17 18
Hour of Day
G-4
-------
Figure G-2. Comparison of Distributions of Estimated Daylight 1-Hour PMi0 Light
Extinction and Maximum Daily Daylight 1-Hour PMio Light Extinction Across the 2005-
2007 Period for Current Conditions, by Study Area, Before and After Elimination of Hours
with Relative Humidity Greater Than 90 Percent.
(a) Maximum Daily Values:
Before Elimination
Daily Maximum Extinction (Daylight Hours)
324 302 36 306 274 149 294
-9-
t I
T--T-
JjF 0>
£• o* * ^ .»V
After Elimination
Daily Maximum Extinction (Daylight Hours)
324 300
306 £73 146 EB9 349
Uilii
T-7--*-
^ ^
G-5
-------
(b) Individual 1-Hour Values:
Before Elimination
Hourly Extinction (Daylight Hours)
1238 3S433@3
After Elimination
Hourly Extinction (Daylight Hours)
3357 3015 14S4 3053 3759 £471 1533 2815 1S73 1453 223
G-6
-------
Table G-2. Comparison of 90th and 95th Percentile PM10 Light Extinction Design Values
for the 2005-2007 Period for Current Conditions Based on Maximum Daily 1-Hour
Daylight PMi0 Light Extinction, Before and After Elimination of Hours with
Relative Humidity Greater Than 90 Percent
Study
Areas
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake
City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
PMio Light Extinction Design Values Based on Daily Maximum 1 -Hour Values
90th Percentile
Before
Exclusion
244
381
919
105
176
189
253
359
366
380
313
368
399
382
339
After
Exclusion
140
338
469
105
164
183
194
307
357
249
310
278
246
286
306
Reduction
Due to
Exclusion
104
43
450
0
12
5
59
52
9
131
3
90
153
96
33
95th Percentile
Before
Exclusion
371
533
114
0
144
266
239
279
423
496
462
473
500
446
449
415
After
Exclusion
157
463
554
144
252
239
234
381
483
288
473
313
286
339
355
Reduction
Due to
Exclusion
215
70
586
0
13
0
44
42
13
174
0
187
159
110
61
G-7
-------
This page left intentionally blank.
-------
APPENDIX H
INTER-YEAR VARIABILITY
One aspect of a NAAQS is whether it is based on the level of the selected indicator for a
single year, or the average of the level of that indicator over multiple years. The NAAQS
scenarios examined in this assessment are all based on a three-year average approach. That is,
design values are based on the average of specified percentile values of PMio light extinction
from 2005, 2006, and 2007. Table H-l presents more detailed information on the variability of
these percentiles across these three years.
Figure H-l presents some of the information in Table H-l in graphical form, specifically
for the 90th percentile form for both the daily maximum and all hour approaches.
H-l
-------
Table H-l. Year-specific Percentile Values of PM10 Light Extinction for 2005, 2006,
and 2007
Study
Areas
Tacoma
Fresno
Los
Angeles
Phoenix
Salt Lake
City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Tacoma
Fresno
Los
Angeles
Phoenix
Salt Lake
City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
90th Percentile Form
2005
2006
2007
2005-
2007
Average
95th Percentile Form
2005
2006
2007
2005-
2007
Average
98th Percentile Form
2005
2006
2007
2005-
2007
Average
Based on Daily Maximum 1-Hour Daylight PMi0 Light Extinction
(Excluding hours with relative humidity greater than 90%)
NA
298
424
100
216
184
217
350
438
235
300
284
303
257
333
121
293
461
110
112
170
204
329
307
255
312
257
227
325
265
158
398
477
NA
161
197
161
239
325
257
313
292
208
274
320
140
330
454
105
163
184
194
306
357
249
308
278
246
285
306
NA
549
507
156
309
252
269
432
547
283
347
347
362
331
405
141
363
523
131
142
223
238
405
410
295
401
272
258
352
272
173
467
619
NA
305
242
196
303
493
286
664
320
239
318
384
157
460
550
144
252
239
234
380
483
288
471
313
286
334
354
NA
653
582
182
341
312
306
483
608
305
391
360
415
426
559
214
398
594
187
191
321
319
572
513
338
490
350
302
375
353
207
540
658
NA
696
271
248
347
565
351
1051
382
260
320
441
211
530
611
185
409
301
291
467
562
331
644
364
326
374
451
Based on 1-Hour Daylight PM10 Light Extinction (All Hours)
(Excluding hours with relative humidity greater than 90%)
NA
181
258
67
116
114
116
227
191
166
226
173
203
163
203
73
172
267
68
67
100
98
195
162
164
212
153
161
203
169
78
211
257
NA
97
125
100
157
166
168
198
176
150
182
187
76
188
261
68
93
113
105
193
173
166
212
167
171
183
186
NA
255
314
79
193
145
143
276
251
188
268
217
290
209
264
101
254
353
78
83
126
122
240
204
194
252
193
190
234
222
109
273
357
NA
148
158
119
188
226
202
234
218
196
223
244
105
261
341
79
141
143
128
235
227
195
251
209
225
222
243
NA
391
393
92
255
184
191
334
340
233
320
284
342
279
313
120
326
451
96
116
176
174
309
267
233
312
237
225
298
267
151
387
478
NA
304
203
148
226
319
248
313
272
218
258
317
136
368
441
94
225
188
171
290
309
238
315
264
262
278
299
H-2
-------
Figure H-l. Inter-year Variability in 90 Percentile 1-Hour Daylight PMi0 Light
Extinction (excluding hours with relative humidity greater than 90 percent)
(a) Daily Maximum Approach
2007
• Salt Lake City
•Birmingham
Baltimore
Los Angeles
• Dallas
-Atlanta
Philadelphia
Fresno
• Houston
• Detroit
New York
Tacoma
-*- St. Louis
— Pittsburgh
-a— Phoenix
(b) All Daylight Hours Approach
2005
2006
Year
2007
• Salt Lake City
•Birmingham
Baltimore
Los Angeles
- Dallas
-Atlanta
Philadelphia
Fresno
- Houston
- Detroit
New York
Tacoma
-H- St. Louis
Pittsburgh
-a— Phoenix
-------
This page left intentionally blank.
-------
APPENDIX I
DAYLIGHT HOURS
Table 3-5 shows the simple scheme used in this analysis to denote hours as fully daylight
and thus eligible for consideration in the calculation of design values for the secondary NAAQS
scenarios based on PMio light extinction. This scheme also has been used to select which hours
to show in various graphics. The scheme is based on applying a fixed set of fully daylight hours
for each three-month season (November to January, etc.). In reality, the local time minutes of
daylight vary continuously during the year, with latitude, and with the east-west position of a city
within its time zone. The hours that are fully daylight will change in increments rather than
continuously. This appendix examines how well the simple scheme reflects actual conditions and
how disparities if any might affect the results presented and the answers to policy relevant
questions that may be addressed in the subsequent policy assessment document.
Six study areas were selected for this examination: Tacoma, Los Angeles, Phoenix,
Houston, Detroit, and New York. These areas cover the extremes with regard to latitude and to
east-west position within time zone. For each area, the times of sunrise (defined by the leading
or top edge of the sun appearing above the horizon) and of sunset (defined by the leading or
bottom edge of the sun disappearing below the horizon) were obtained for each day of the year.
It is several minutes after each of these times that the sun is fully visible in the morning and not
visible at all in the evening.
Figure 1-1 shows the relationship between these sunrise and sunset times and the simple
scheme used to denote hours as fully daylight. The vertical scale is in hours with zero
corresponding to local noon. The smooth curves represent the actual times of sunrise (top of
figure) and sunset (bottom of figure). The stepped lines represent the scheme used to select the
first and last hour denoted as fully daylight. Months are indicated on the horizontal axis. The
figure indicates that the simple scheme has the effect of treating some hours as daylight that in
fact contain minutes prior to sunrise or after sunset, and conversely treating some hours as not
daylight that include no such minutes. In particular:
• In February, the hours from 7 am to 8 am and from 5 pm to 6 pm are treated as
daylight but include non-daylight minutes in most of the example areas.
• In April, the hour from 6 am to 7 am is treated as non-daylight but in many areas
includes only minutes that are after sunrise.
• In most of June and most of July, for Detroit and Tacoma only, the hour of 7 pm to 8
pm is treated as non-daylight but in fact has no minutes after sunset.
1-1
-------
• In October, the hours of 6 am to 7 am and 5 pm to 6 pm are treated as daylight but
include non-daylight minutes in all of the example areas.
The tile plots in Figure 3-12 can be used to assess the significance of these disparities,
i.e., whether they are likely to significantly affect PMi0 light extinction design values. Table 1-1
contains observations for each of the 24 combinations of the four time periods listed above and
the six example areas. Taken together, these observations make it likely that refining the scheme
for designating hours as fully daylight would not significantly change conclusions that can be
drawn from this assessment as it has been performed. Changing the scheme would involve
considerable effort in updating virtually every table and graphic in the assessment, however.
1-2
-------
Figure 1-1. Comparison of Actual Sunrise and Sunset Times to this Assessment's Scheme to Denote Hours as Fully
Daylight
o^nn
Z'tUU
oonn
zzuu
onnn
zuuu
1 Rnn
I OUU
(1) -icnn -
_ 1 DUU
.— i4nn
, I H-UU
1 9nn
^_ I ZUU
Omnn
I UUU
i snn
^J oUU
RDD
DUU
4nn
tuu
onn
zuu
~
Ja
1
Sunrise/Sunset for Six Sites
_ — ^ —
i'i ^^ -^^fa^
t gj&^^i ~T- \ "* i' fcj|_^
^r — ^-j»»^^ i —•
^ '^vt,^ 1 ^^^aai^^^~~'
^^* ^^r jz^J
ii i i i i i i ii
n- Feb- Mar- Apr- May- Jun- Jul- Aug- Sep- Oct- Nov- De
0 10 10 10 10 10 10 10 10 10 10 1
Date
3C-
0
Assumed daylight start
Assumed daylight end
TAG Rise
TAG Set
-LA Rise
LA Set
PHX Rise
PHX Set
HOU Rise
l_l/~\l 1 Cr^t
— MUU oet
HFT Ricn
~ L/C 1 Ixloc
ftFT 9ot
-NY Rise
NY Set
1-3
-------
Table 1-1. Observations from Tile Plots for Hours with Questionable Daylight/Non-Daylight Status in Six Study Areas
Study Area
February (AM and PM)
April (AM)
June-July (PM)
October (AM and PM)
Tacoma
The morning hour with questionable
daylight status tends to have RH > 90%.
The evening hour in question tends to
either have low PM light extinction or to
have RH > 90%.
The tile plot does not show data for the
morning hour that may better be denoted
daylight, but the instances of high PM light
extinction that do appear typically last
multiple hours.
Very late afternoon typically is not a
period of high PM light extinction.
Instances of high light extinction involving
the questionable hours are multi-hour and/or
involve RH > 90%.
Los Angeles
Instances of high light extinction
involving the questionable hours are
multi-hour and/or involve RH > 90%.
The tile plot does not show data for the
morning hour that may better be denoted
daylight, but the instances of high PM light
extinction that do appear typically last
multiple hours.
NA
Instances of high light extinction involving
the questionable hours are multi-hour and/or
involve RH > 90%.
Phoenix
Instances of high light extinction
involving the questionable hours are
multi-hour.
The tile plot does not show data for the
morning hour that may better be denoted
daylight, but early morning in April typically
is not a time of high PM light extinction.
NA
PM light extinction is usually low in
October; on those days with moderate levels
in the questionable hours, another hour in the
central part of the day has a similar level.
Houston
Instances of high light extinction
involving the questionable hours are
multi-hour and/or involve RH > 90%.
The tile plot does not show data for the
morning hour that may better be denoted
daylight, but the instances of high PM light
extinction that do appear typically last
multiple hours and/or involve RH >90%.
NA
The amount of information is limited due to
missing data. On those days with moderate to
high PM light extinction during the
questionable hours, another hour has a
similar level, or RH >90% plays a role.
Detroit
Instances of high light extinction
involving the questionable hours are
multi-hour.
The tile plot does not show data for the
morning hour that may better be denoted
daylight, but the instances of high PM light
extinction that do appear typically last
multiple hours.
July generally is a time of high PM
light extinction for the hours currently
considered daylight. Adding one more
late afternoon hour likely would not
affect design values.
Instances of high light extinction involving
the questionable hours are multi-hour.
New York
All but one instance of high light
extinction involving the questionable
hours are multi-hour.
The tile plot does not show data for the
morning hour that may better be denoted
daylight, but the instances of high PM light
extinction that do appear typically last
multiple hours.
NA
Instances of high light extinction involving
the questionable hours are multi-hour and/or
involve RH > 90%.
1-4
-------
APPENDIX J
LOGIT MEMORANDUM
J-1
-------
Memorandum
To: Vicki Sandiford, Office of Air Quality Planning and Standards,
U.S. Environmental Protection Agency
From: Leland Deck and Megan Lawson, Stratus Consulting Inc.
Date: 2/3/2010
Subject: Statistical analysis of existing urban visibility preference studies
During the CASAC meeting on October 5-6, 2009, Dr. Bill Malm and other CASAC members
suggested that a limited dependent variable statistical analysis could be used to analyze the
acceptability criteria responses in the four cities for which there are existing urban visibility
preference studies. It was the view of those Panel members that successful statistical analyses of
the studies results would provide an estimate of a "best fit" central tendency function describing
the results of the preference studies, as well as confidence intervals around the estimated
functions. Such analyses would also make it possible to conduct hypothesis testing, such as
examining whether the estimated 50% criteria level in one study is statistically different than the
50% criteria level in another study.
On the basis of the CASAC comments and the information available in the previous Stratus
Report (Stratus Consulting, 2009), EPA concluded it was appropriate to conduct further
statistical analyses on the available urban visibility preference studies. Subsequently, EPA asked
Stratus Consulting to re-examine the data from these studies and identify several methods for
statistical analyses along the lines CASAC members suggested. This memorandum provides a
description of the statistical analyses we conducted, and summarizes the results.
Data
While we do not have complete original response data from each preference study, certain data
available in all four studies can be used to derive a set of data for an analysis comparing the
results from each of the four1 cities. This available data is the percentage of respondents that
rated each individual photograph (or image) as acceptable. We also know the total number of
individuals that rated each photograph, as well as the haziness level in each photograph,
measured in deciviews (dv). Using these pieces of information we were able to assemble a
master data set of 19,280 observations from the original data. Each observation is associated
In the initial set of analyses discussed in this memorandum we combine the results from the 2001
Washington, DC focus group study with all 26 participants in the "Test 1" analysis from Smith and Howell
(2009). "Test 1" was designed to replicate the 2001 focus group study, with a goal of making two sets of
results directly comparable. Additional analysis described later in this memorandum uses a different set of
statistical techniques to examine the Washington, DC studies in more detail.
SCI 1979
-------
Stratus Consulting Memorandum (2/3/2010)
with an individual binary "yes" or "no" acceptability answer, the dv level, and the city location
for a single photograph.
For example, in the Phoenix study 385 participants rated each of 21 different WinHaze images.
Hence the Phoenix study contributes 8,085 (385 x 21) observations, nearly 41.9% of the total set
of 19,280 observations in the master data set. The 32 photographs used in the Denver study
contribute 6,848 observations (35.5% of the total), the 20 photographs in the British Columbia
contribute 3,600 observations (18.7% of the total), and the combined Washington, DC studies
(combining data from the DC-2001 study with the Test 1 data from the DC-2009 study)
contribute 747 (3.9% of the total). The 19,280 observations are fairly evenly split, with 9,452
"yes" observations, and 9,828 "no" responses.
The participants in each study viewed a series of images with different dv levels. While the data
collected by the original researchers included information linking each individual with their
ratings on each picture, such detailed information is currently only available for the Washington,
DC study conducted in 2009. Access to this additional level of information in the 2009
Washington study allows us to conduct an additional type of analysis accounting for individual
heterogeneity of preferences regarding acceptable levels of visibility.
Statistical Analysis Models
All of the analyses described in this memorandum are logistic regressions using the logit model.
The logit model is a generalized linear model used for binomial regression analysis which fits
explanatory data about binary outcomes (in this case, a person rating a photograph acceptable or
not) to a logistic function curve.
In the context of the preference studies, the logit model estimates the function that best
approximates the percentage of respondents that will rate a photograph acceptable based on a set
of explanatory variables. The observations on the dependent variable have one of two discrete
values: 1 (the person rated the photograph acceptable) or 0 (unacceptable). In our context, the
logit model estimates the proportion of participants who will find any particular dv level
acceptable. In our analysis, there were two basic types of explanatory (independent) variables;
one continuous numerical variable (the photograph's haziness level in dv), and a set of discrete
variables that identify which city the observation is from. We estimate two variations of the logit
model, using the basic explanatory variables in different ways.
The fundamental form of a logistic function is:
1
probability^ yes") = f(z) =
i + e '
where the variable z, known as the logit, is the influence of all the explanatory variables:
Page 2
SCI 1979
-------
Stratus Consulting Memorandum (2/3/2010)
In our analysis the estimated logistic function f (z) is the estimated probability of the participants
in the study rating a photograph acceptable, given the dv value of the photograph and what city
the observation came from.
We conducted the logit analysis using two alternative forms of the logit model.
Model 1 is a simple form of the logit model, and includes the dv value and uses the city
information to create a set of categorical indicator variables. This analysis assumes that all
respondents have a similar shape to their response function (the probability function of
responding "yes" given the dv level of a photograph), but investigates whether the location of the
response function differs in the four cities.
The logit for Model 1 is:
z = Intercept + frdv + /32BC + j83DC + /34Phoenix + s .
The variables BC (British Columbia), DC (Washington), and Phoenix are the indicator (or
"dummy" variables. For example, the BC variable is set equal to one if the observation is from
the BC study, and set to zero if that observation is from a study in a different city study. Denver
is used as the omitted city indicator variable, allowing the estimated coefficients on the other
three city indicator variables to estimate if the response function is different in those cities than
in Denver. The term e represents the error with which the model was estimated, or the difference
between the actual and predicted values of z. The logit model assumes that s has a mean of zero.
The Model 1 form of the logit model estimates a single "slope" for the response function in all
cities as fti, the coefficient for haziness (dv). The other terms shift the intercept. The intercept
for Denver is simply the estimated parameter Intercept. The effective intercept for the other
cities becomes the sum of Intercept plus the coefficient on the city's indicator variable, for
example the intercept for Washington is Intercept + $3.
Model 1 creates one test of the hypothesis that the responses in each city are the same. If the
estimated coefficient on a particular city variable is statistically significant, the analysis would
imply that the city's response function is likely shifted relative to the Denver function, and that
city would have a different dv value for the 50% criteria. A positive and significant city
coefficient shifts that city's response function to the right, resulting in the dv level where 50%
criteria level in that particular city is higher than Denver's.
Model 2 is a more general model than Model 1, and relaxes the assumption in Model 1 that the
slope of the response function is the same in every city. Model 2 includes not only dv and the
Page 3
SCI 1979
-------
Stratus Consulting Memorandum (2/3/2010)
city indicator variables as in Model 1, but also a set of interaction terms, where each city dummy
variable is multiplied by the dv level. The logit for Model 2 is:
z = Intercept + B^v + B2BC + /?3 (dv x BC) + B4DC
+ /?5 (dv x DC) + B6Phoenix + B1 (dv x Phoenix) + s.
For example, in Model 2 the estimated total intercept for Washington becomes Intercept + $4,
and the estimated slope of the Washington function is ft'4 + ft j.
In the fully interacted Model 2 a statistically significant estimate of the city indicator variable
coefficients (fi'2, $4, or fig) has the same implication as in Model 1; the response function is likely
shifted relative to the Denver function. A statistically significant estimate of the interaction term
coefficient (fi'3, ft5, or f$7) for a particular city implies that the response function has a different
slope than the Denver function.
The fully interacted model produces the same results as conducting a separate logit analysis for
each of the four cities. The interacted model, however, makes it easier to conduct hypothesis
testing on the estimated mean response functions.
The predicted mean dv values at each of the acceptance criteria presented here are a function of
the coefficients on dv and the other explanatory variables, each of which have their mean and
standard deviation. Therefore, a confidence interval constructed around this predicted mean
must account for both the variance and covariance of the parameter estimates. Using a Monte
Carlo estimation approach, we made 1000 random draws from the joint distribution of the
coefficients using the mean vector and variance-covariance matrix of the parameter estimates for
the distribution parameters. For each of these draws we then calculated the predicted mean dv.
After removing the lower and upper 5% of the simulated values, the lower and upper end of the
range of predicted values represent the lower and upper range of the 95% confidence interval.
Confidence intervals calculated using this procedure are known as Krinsky-Robb confidence
intervals (Krinsky and Robb, 1986). Because estimating Krinsky-Robb confidence intervals
requires a separate Monte Carlo analysis for each acceptability criteria dv level, we only estimate
confidence intervals for five different acceptability levels: 90%, 75%, 50%, 25%, and 10%.
The Krinsky-Robb procedure assumes that the estimated parameters are normally distributed,
which may or may not be true. To explore the potential impact of this assumption, for one logit
analysis we also conducted an alternative procedure that does not assume a normal distribution.
This alternative procedure (Hole, 2007) uses a bootstrap method to estimate the confidence
intervals for the estimated mean 50% criteria. The confidence intervals using the bootstrap were
within 1% of the confidence intervals using the Krinsky-Robb procedure, indicating that the
multivariate normal assumption imposed by the Krinsky-Robb procedure is not unreasonable.
We also conducted hypothesis tests using the median dv values estimated using the
Page 4
SCI 1979
-------
Stratus Consulting Memorandum (2/3/2010)
bootstrapping procedure. The conclusions from these hypothesis tests were identical to the
conclusions from the other hypothesis tests.
Statistical Analysis Results, Inter-City analyses
We conducted all the logit analyses described in this document using STATA® Data Analysis
and Statistical Software (Release ES 10.1), using the LOGIT procedure. The Krinsky-Robb
analysis used STATA's "wtpcikr" module. The bootstrap method (Hole, 2007) was conducted
using STATA's "bootstrap" module.
Model 1 Results, Inter-City Analysis
Table 1 presents the parameter estimates from the logit analysis with city indicators (Model 1)
which effectively shift the intercept. The Washington, DC data in this analysis includes both
DC-2001 and DC-2009 (Test 1) data. The Denver study is the omitted indicator city in this
analysis, so the intercept term coefficient for Denver is equal to the Constant. The intercept for
the other cities is the sum of the constant plus the coefficient for the respective city. The
coefficient for variable dv is the estimated slope for all four cities.
Table 1. Model 1 logit analysis results
Variable
dv
British Columbia
Washington, DC
Phoenix
Constant
Coefficient
(P)
-0.4187
1.1164
3.8743
1.8021
8.3073
Standard
error
0.0059
0.0630
0.1325
0.0576
0.1186
z-statistic
-71.09
17.72
29.25
31.31
70.07
Pr|p|=0
< 0.001
< 0.001
< 0.001
< 0.001
< 0.001
5% confidence
estimate
-0.430
0.993
3.615
1.689
8.075
95% confidence
estimate
-0.407
1.240
4.134
1.915
8.540
9 9
McFadden's pseudo-R for the Model 1 estimate was 0.474.
2 While pseudo-R2 is, like traditional R2, bounded between zero and one, it does not have the same
interpretation. R can be interpreted as the percentage of the variation in the dependent variable explained by
variation in the independent variables. Pseudo-R2, on the other hand, is the percent improvement in log
likelihood from using the full set of explanatory variables, relative to a model that uses only a constant. It
offers a sense for how much better the model fits when the explanatory variables are added, but cannot tell us
the percentage of variation we are explaining. Pseudo R2, instead of traditional R2, must be used in evaluating
logit and other maximum likelihood estimation models. Similar to R2, a higher pseudo-R2 indicates a model
with a better fit.
Page 5
SCI 1979
-------
Stratus Consulting Memorandum (2/3/2010)
The Log likelihood chi test strongly rejects the null hypothesis there is no effect of explanatory
variable
0.000).
r\
variables on the probability that a respondent would find a photograph acceptable (Pr (chi ) = 0 <
The z-statistic (also known as the Wald z-statistic) in a logit analysis is analogous to the t-
statistic in a conventional linear regression. The z-statistic is simply the ratio of the estimated
coefficient to its standard error, and can be used to estimate the probability that the estimated
coefficient is equal to zero. The column in Table 1 labeled "Pr |P| = 0" is the 2-tailed p-value
used in testing the null hypothesis that the estimated parameter is zero. The Pr |P| values shown
in Table 1 are all less than 0.005 ("~0"), indicating that all of the estimated coefficients are very
statistically significant. Because the city dummy variables are significant, in Model 1 we reject
the hypothesis that the four studies have an identical response function.
Figure 1 shows the estimated response functions in each city for the logit analysis with city
indicators, as well as the underlying data as was shown in Figure 14 of the Stratus Consulting
final report (Stratus Consulting, 2009). While Model 1 estimates the shape of a response
function that is identical in each city, the positive and significant coefficients on the city
variables in Model 1 result in the response functions for the different cities to shift to the right of
the Denver function.
The logit analysis results also support estimating the dv value where the 50% acceptability
criteria are met in each city. The 50% acceptability criteria occur at the level of haziness where
half the survey participants said the visibility is acceptable, and half said it was not acceptable.
In Figure 1, the 50% criteria level is the dv value where the estimated response function crosses
the 50% response level on the y axis.
As a sensitivity analysis, it is also possible to calculate the dv levels that meet alternative
decision criteria. For example, one can calculate the estimated dv level at which 75% of the
participants said the visibility was acceptable. This 75% criterion would occur at better visibility
(i.e., lower dv values) than the 50% criteria. Similarly, one can also calculate the estimated the
dv level that any desired percentage of the participants said was acceptable. The Model 1
estimates of alternative acceptability criteria dv values for each city are shown in Table 2.
Page 6
SCI 1979
-------
Stratus Consulting
Memorandum (2/3/2010)
100%
-------
Stratus Consulting Memorandum (2/3/2010)
The range of the Model 1 estimates of the 50% acceptability criteria is very consistent with the
Candidate Protection Level (CPL) range of 20 dv to 30 dv identified in the U.S. EPA (2009)
report Particulate Matter Urban-Focused Visibility Assessment; External Review Draft (UFVA).
Model 2 Results, Inter-City Results
Table 3 presents the parameter estimates from the fully interacted logit analysis, which
investigates whether both slope and the intercept of the estimated response function differ
between cities. Denver was again used as the omitted city in the fully interacted model.
Table 3. Model 2 logit analysis results
Variable Coefficient (P)
dv
British Columbia
Washington, DC
Phoenix
BC*dv
Wash, x dv
Phoenix x dv
Constant
-0.3862
1.0496
2.9450
3.5682
-0.0029
0.0200
-0.0797
7.6844
Standard
error
0.0094
0.3589
0.8458
0.3015
0.0162
0.0293
0.0136
0.1830
z-statistic
-41.16
2.92
3.48
11.84
-0.18
0.68
-5.88
41.99
Pr|p| = 0
< 0.001
0.003
< 0.001
< 0.001
0.860
0.495
< 0.001
< 0.001
5% confidence
estimate
-0.4045
0.3463
1.2873
2.9773
-0.0345
-0.0374
-0.1063
7.3257
95% confidence
estimate
-0.3678
1.7530
4.6026
4.1591
0.0288
0.0774
-0.0531
8.0431
The pseudo-R2 for the Model 2 estimate was 0.4756 (very similar to the Model 1 results), and the
Model 2 log likelihood chi test also strongly rejects the null hypothesis there is no effect of the
explanatory variables on the probability that a respondent would find a photograph acceptable
(Pr (chi2) = 0 < 0.000).
The city indicator coefficients in this full interaction model are all positive and statistically
significant, as they were in Model 1, indicating that the response functions for different cities
shifted right (relative to Denver). However, of all the interactions only the Phoenix interaction
term is significant, indicating that the Phoenix response function has a different slope than the
other three cities.
Figure 2 shows the estimated response functions in each city for Model 2, as well as the
underlying data.
PageS
SCI 1979
-------
Stratus Consulting
Memorandum (2/3/2010)
100%
£1
CO
D)
£ 50%
I
ra
Q.
o
•c
ro
Q.
0%
10
15
20
25
Deciview
30
35
40
45
A Denver
Denver Logit
Phoenix
Phoenix Logit
BC
BC Logit
Washington
DC Logit
Figure 2. Estimated response functions for four cities using the fully interacted logit
analysis.
The significantly different slope of the Phoenix response function is clearly visible in Figure 2.
The negative estimated coefficient on the Phoenix interaction term results in the Phoenix
response function being steeper than the other cities' functions. In other words, Phoenix
respondents' acceptance probabilities were more sensitive to changes in dv levels. Figure 2 also
shows the Washington, DC function is modestly less steep than the others, but the decrease in the
slope is not statistically significant. Therefore, while Washington, DC respondents are more
likely to accept worse visibility overall, they are just as responsive to changes in dv as
respondents in Denver and British Columbia.
As with Model 1, it is possible to use the Model 2 results to estimate the dv values where the
estimated response functions cross the 50% acceptability level, as well as any alternative criteria
levels. The Model 2 estimates of alternative acceptability dv values for each city are shown in
Table 4.
Page 9
SCI 1979
-------
Stratus Consulting Memorandum (2/3/2010)
Table 4. Model 2 estimated haziness (dv) levels of alternative acceptability criteria
Denver British Columbia
90% acceptability criteria
75% acceptability criteria
50% acceptability criteria
25% acceptability criteria
10% acceptability criteria
14.21
17.05
19.90
22.74
25.59
16.80
19.63
22.45
25.28
28.10
Washington, DC
23.03
26.03
29.12
32.03
35.03
Phoenix
24.15
21.80
24.15
26.51
28.87
The Model 2 estimates of the 50% acceptability criteria are nearly identical to the Model 1
estimates; the biggest difference is a 0.07 dv decrease in the Washington, DC 50% acceptability
criteria. The essentially identical estimates of the 50% acceptability criteria in Models 1 (city
indicator only) and Model 2 (full interaction) indicates the choice of model form does not change
the conclusion that the logit results are consistent with the 20 to 30 dv CPL range identified in
the draft UFVA (EPA, 2009).
We also conducted hypothesis testing with the four city data used in this section to examine the
probability that the 50% acceptance criteria in the four different cities are the same. We used the
full interaction model results for the hypothesis testing. Our approach estimated the mean 50%
criteria dv levels and standard error (based on the Krinsky-Robb confidence intervals) for each of
the four cities. We then conducted a hypothesis testing using a t-test to estimate the probability
the mean 50% criteria dv levels are the same in each pair of cities. The null hypothesis in this
hypothesis test is that the means are the same. As shown in Table 5, the null hypothesis is
strongly rejected for all pairs of cities, indicating that the mean 50% criteria dv levels differ for
all four cities.
Table 5. Hypothesis testing on whether the full interaction model mean
50% criteria dv levels are the same
British Columbia Phoenix Washington, DC
Mean dv = 22.45 Mean = 24.15 Mean dv = 29.12
Denver t-stat= 16.89 t-stat = 35.15 t-stat = 30.21
Mean dv= 19.90 Pr(Den = BC) ~0 Pr(Den = Ph)~0 Pr(Den = DC)~0
British Columbia
Phoenix
t-stat =12.08
Pr(BC = Ph)~0
—
t-stat = 2 1.23
Pr(BC = DC) ~ 0
t-stat = 16.53
Pr(Ph = DC) ~ 0
Page 10
SCI 1979
-------
Stratus Consulting Memorandum (2/3/2010)
Analysis of Washington, DC Preference Studies
There are two related studies of visibility preferences in Washington, DC. In 2001, in a project
sponsored by the U.S. Environmental Protection Agency, Abt Associates conducted a pilot focus
group study (DC-2001) of urban visibility preferences in Washington, DC. In 2009, in a study
for the Utility Air Regulatory Group, Smith and Howell conducted a series of three tests of urban
visibility preferences in Washington, DC. In their first test (DC-Test 1), Smith and Howell used
all the images used in the DC-2001 study, trying to replicate the DC-2001 study. Their second
test (DC-Test 2) used fewer of the Washington images, restricting the study to the 12 images
with better visibility (images with visibility of 27.1 dv or better). In the third test (DC-Test 3),
they expanded the range of images to include two hazier images (adding a 42 and 45 dv images,
and deleting images at 11.1, 15.6, and 24.5 dv).
An important question is whether the participant responses obtained in the DC-2001 study are
similar to the responses in Test 1, which was designed to replicate the DC-2001 study. A related
question is whether the responses in Tests 2 and 3 are similar to Test 1. To investigate these
questions we estimated logit response functions using the data from the four different
Washington, DC data sets (DC-2001, DC-Test 1, DC-Test 2, and DC-Test 3), using the full
interaction logit model specification.
The estimated coefficients from a full interacted model are presented in Table 6. The DC-2001
test is used as the omitted interaction variable.
Table 6. Logit regression results with full interacted model of Washington, DC studies
Variable
dv
Test 1
Test 2
Test3
Test 1 x dv
Test 2 x dv
Test 3 x dv
Constant
Coefficient
(P)
-0.4035
-1.5425
-0.7431
3.4109
0.0616
-0.1043
-0.0607
11.5621
Standard
error
0.0567
1.8785
2.0737
2.6980
0.0632
0.0804
0.0868
1.6777
z-statistic
-7.12
-0.82
-0.36
1.26
0.97
-1.30
-0.70
6.89
Pr|p|=0
< 0.001
0.412
0.720
0.206
0.330
0.194
0.485
< 0.001
5% confidence
estimate
-0.5146
-5.2242
-4.8075
-1.8772
-0.0624
-0.2618
-0.2309
8.2739
95% confidence
estimate
-0.2925
2.1392
3.3212
8.6990
0.1855
0.0532
0.1095
14.8504
Figure 3 shows the estimated full interaction logit function for the separate Washington, DC Test
data, including the DC-2001 data.
Page 11
SCI 1979
-------
Stratus Consulting
Memorandum (2/3/2010)
100%
10
15
20
25
dv
30
35
40
DC-2001
Test
1
—Test
2
Test
3
Figure 3. Full interaction logit results for four separate DC data sets.
Figure 3 suggests that while the logit functions from DC-2001 and Test 1 appear to be similar,
Test 2 and Test 3 appear to be substantially different. Estimating the 50% criteria levels, along
with the Krinsky-Robb confidence intervals, confirms this observation. Table 7 presents the
estimated mean 50% criteria levels and the Krinsky-Robb confidence intervals.
Table 7. Mean 50% criteria levels, and Krinsky-Robb
confidence intervals
Test 1
Test 2
Test3
DC-2001
Mean dv
29.30
21.30
32.26
28.65
5% confidence
estimate
28.59
20.57
31.37
27.46
95% confidence
estimate
29.97
22.03
33.16
29.70
Page 12
SCI 1979
-------
Stratus Consulting Memorandum (2/3/2010)
Hypothesis testing using the predicted mean dv distribution from the Krinsky-Robb procedure
provides statistical support for the conclusion that the Test 1 and DC-2001 results are similar, but
Test 2 and Test 3 results are different. The hypothesis testing results are presented in Table 8.
Table 8. Hypothesis testing on the individual coefficients in the full interaction model is the
same for the four different Washington, DC experiments
Test 2 Test 3 DC-2001
Mean dv = 21.30 Mean = 32.26 Mean dv = 28.65
Test 1 Reject hypothesis that Test 1 Reject hypothesis that Cannot reject hypothesis that
Mean dv = 29.30 = Test 2 (pr< 0.001) Test 1 = Test 3 (pr< 0.001) Test 1 = DC-2001 (pr = 0.15)
Test 2 - Reject hypothesis that Reject hypothesis that Test 3 =
Test 2 = Test 3 (pr < 0.001) DC-2001 (pr < 0.001)
Test 3 - Reject hypothesis that Test 3 =
DC-2001 (pr< 0.001)
As shown in Table 8, we cannot reject (at the 5% confidence level) the hypothesis that the mean
50% criteria level in the DC-2001 data and the Test 1 data are the same. In other words, it is
likely that the mean dv in Test 1 is the same as the mean dv in DC-2001. Thus, this hypothesis
test supports combining those two data sets together, as we did in the four city analysis presented
above. The results in Table 8 reject the hypothesis that Test 2 and Test 3 are the same as either
the Test 1 or DC-2001 results.
Further Analysis of the Washington, DC Test 1 Data
Smith and Howell conducted Test 1 using three distinct groups of respondents. Four of the
respondents in Test 1 were Washington, DC area residents that were used in a pilot test of the
testing procedure. Twelve of the respondents were CRA International employees who live in the
Washington, DC area, and ten of the respondents were CRA International employees who live in
the Houston, Texas area. The Test 1 participants were all shown the same images of
Washington, DC haze levels as the DC-2001 participants, and were asked about their preferences
for urban visibility in Washington, DC.
We investigated heterogeneity among these three groups' responses by conducting a full
interaction logit analysis using information about which of the three groups (pilot, DC area or
Houston area) the respondents were in. We also included the DC-2001 respondents (who were
all DC area residents) in this analysis to conduct hypothesis tests on whether the Test 1 groups
were different than the DC-2001 respondents. We used the pilot test respondents as the omitted
group in a full interaction model analysis. The results of the logit analysis are presented in Table
9.
Page 13
SCI 1979
-------
Stratus Consulting Memorandum (2/3/2010)
Table 9. Logit regression results with full interacted model of the 3 Test 1 groups and the
DC-2001 participants
Variable
dv
Test I/DC
Test 11 Houston
DC-2001
Test I/DC x dv
Test 1 /Houston x dv
DC-2001 x dv
Constant
Coefficient
(P)
-0.5719
-0.8344
-4.8831
-2.4042
0.1439
0.2643
0.1684
13.9663
Standard
error
0.1310
3.7361
3.5486
3.7273
0.1420
0.1372
0.1428
3.3284
z-statistic
-4.36
-0.22
-1.38
-0.65
1.01
1.93
1.18
4.20
Pr|0| = 0
0.000
0.823
0.169
0.519
0.311
0.054
0.238
0.000
5% confidence
estimate
-0.8287
-8.1570
-11.8382
-9.7095
-0.1344
-0.0047
-0.1114
7.4428
95% confidence
estimate
-0.3151
6.4881
2.0719
4.9012
0.4222
0.5332
0.4482
20.4898
Using the estimated coefficients in Table 9, we calculated estimated 50% criteria levels for each
group, along with the Krinsky-Robb confidence intervals, which are shown in Table 10.
Table 10. Mean 50% criteria levels, and Krinsky-Robb intervals for
the Test 1 groups and the DC-2001 participants
Test I/DC
Test 1 /Houston
Test I/Pilot
DC-2001
Mean dv
30.68
29.52
24.42
28.65
5% confidence
level
29.79
28.30
22.37
27.46
95% confidence
level
31.51
30.66
25.97
29.70
Table 10 suggests that the mean 50% acceptance criteria level for the Washington, DC area
residents in the 2001 study are closest to the mean 50% criteria level for the Test 1 Houston area
residents, and differ to a greater degree from the mean 50% criteria level for the Test 1
Washington area residents. Hypothesis testing confirms this finding, as shown in Table 11.
Page 14
SCI 1979
-------
Stratus Consulting
Memorandum (2/3/2010)
Table 11. Hypothesis tests of the mean 50% acceptance criteria level for the three groups
in the Test 1 data and the DC-2001
Houston
Mean dv = 29.52
Pilot
Mean dv = 24.42
DC-2001
Mean dv = 28.65
Test I/DC area Reject hypothesis
Mean dv = 30.68 Houston = Test I/DC
(pr = 0.06)
Reject hypothesis Pilot = Reject Test I/DC = 2001-DC
Test I/DC (pr < 0.001) (pr < 2%)
Test I/Houston
area
Reject Houston = Pilot
(pr< 0.001)
Cannot reject Houston = DC-2001 at
5% confidence (pr = 14%)
Test 1/Pilot
Reject Pilot = DC-2001 (pr <0.001)
These hypothesis test results in Table 11 provide some insight into the hypothesis tests in Table
8, which found the 50% mean criteria level (mean = 29.30 dv) estimated using the combined
Test 1 data is similar to the 50% criteria level from the DC-2001 data (mean = 28.65 dv). The
Table 11 results suggest that the Table 8 results could be the result of the Test 1 pilot participants
(mean = 24.42 dv) offsetting the Test I/DC area participants (mean = 30.68 dv), giving us a
mean estimate for the combined sample closest to the Houston area participants (mean = 29.52
dv).
Individual Heterogeneity
Individual respondents will likely have different general attitudes regarding haze than other
respondents, reflecting their individual preferences about urban visibility. An individual's
preferences may affect how they rate the acceptability of different dv levels. In the Smith and
Howell (2009) Washington, DC study we can track an individual's responses over all dv levels.3
This enables us to account for individual heterogeneity in our estimation procedure using
individual-specific indicators. These are called fixed-effect models and control for unobserved
differences between respondents.
We conducted a logit analysis on Test 1 data using individuals as the indicator variable. We
included slope interaction terms for the Washington and Houston area residents (with the pilot
slope interaction term omitted). Each individual4 also has an indicator which becomes the
3 While this level of data was originally collected for the studies in Denver, Phoenix, British Columbia and the
2001 Washington, DC study, the original data is not available at this time.
4 Respondents 1 and 13 are dropped in the individual heterogeneity analysis because they had identical
responses, accepting every dv level. The form of the logit model used in this analysis cannot be estimated
when all the responses are identical.
Page 15
SCI 1979
-------
Stratus Consulting
Memorandum (2/3/2010)
intercept term for that individual. The terms for Respondents 2 through 12 are intercept shifters
for DC respondents. Respondents 14 through 22 were Houston respondents, and Respondents 23
through 25 were pilot respondents. The results from this model are presented in Table 12.
Table 12. Logit analysis results of individual heterogeneity analysis
Variable
dv
Houston x dv
DCxdv
Respondent 2 (DC)
Respondent 3 (DC)
Respondent 4 (DC)
Respondent 5 (DC)
Respondent 6 (DC)
Respondent 7 (DC)
Respondent 8 (DC)
Respondent 9 (DC)
Respondent 10 (DC)
Respondent 1 1 (DC)
Respondent 12 (DC)
Respondent 14 (H)
Respondent 15 (H)
Respondent 16 (H)
Respondent 17 (H)
Respondent 18 (H)
Respondent 19 (H)
Respondent 20 (H)
Respondent 21 (H)
Respondent 22 (H)
Respondent 23 (P)
Respondent 24 (P)
Respondent 25 (P)
Respondent 26 (P)
Coefficient
(P)
-0.7315
0.1207
-0.3588
35.2050
35.2050
29.9950
34.3924
32.0347
31.0845
25.7365
36.1200
34.3924
28.7572
35.2050
16.6104
15.9236
15.9236
18.3145
20.8740
13.3405
19.8140
16.6104
18.8166
16.0044
17.1746
19.9838
18.2301
Standard
error
0.1911
0.2139
0.2658
5.9847
5.9847
5.2578
5.8635
5.5755
5.4326
4.5956
6.1617
5.8635
5.0615
5.9847
2.8047
2.7170
2.7170
2.9999
3.3153
2.3443
3.1722
2.8047
3.0538
4.3526
4.6884
5.4132
4.9705
z-statistic
-3.83
0.56
-1.35
5.88
5.88
5.7
5.87
5.75
5.72
5.6
5.86
5.87
5.68
5.88
5.92
5.86
5.86
6.11
6.3
5.69
6.25
5.92
6.16
3.68
3.66
3.69
3.67
Pr|p|=0
0
0.573
0.177
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
5%
confidence
estimate
-1.1060
-0.2986
-0.8799
23.4752
23.4752
19.6900
22.9002
21.1070
20.4369
16.7293
24.0434
22.9002
18.8369
23.4752
11.1133
10.5984
10.5984
12.4348
14.3761
8.7457
13.5966
11.1133
12.8313
7.4736
7.9854
9.3742
8.4882
95%
confidence
estimate
-0.3569
0.5399
0.1622
46.9349
46.9349
40.3001
45.8846
42.9624
41.7322
34.7438
48.1966
45.8846
38.6775
46.9349
22.1075
21.2488
21.2488
24.1942
27.3719
17.9353
26.0315
22.1075
24.8019
24.5353
26.3637
30.5933
27.9720
As in the analyses previously described, we used the logit analysis coefficients in Table 12 to
estimate the mean value for the 50% acceptance criteria. We also estimated the Krinsky-Robb
confidence intervals for each data subset using the fixed effects model. Because three of the Test
Page 16
SCI 1979
-------
Stratus Consulting
Memorandum (2/3/2010)
1 participants were deleted in the individual heterogeneity analyses, for comparison purposes we
also re-estimated a model without accounting for individual heterogeneity using the same data
set (i.e., with the two individuals deleted). The results are presented in Table 13.
Table 13. Estimated mean 50% criteria levels, and Krinsky-Robb intervals
for the Test 1 data accounting for individual heterogeneity
Washington area residents
Houston area residents
Pilot (DC residents)
Mean dv estimates without
Washington area residents
Houston area residents
Pilot (DC area residents)
Mean dv
30.57
29.40
24.40
Lower bound 95% Upper
29.97
28.41
22.60
bound 95%
31.18
30.33
25.91
individual heterogeneity (using same data)
30.02
28.50
24.42
29.19
27.25
22.37
30.77
29.58
25.97
Table 13 shows that including individual heterogeneity in the model modestly increased the
estimated mean 50% criteria levels.
Table 14 shows the results of hypothesis testing on the individual heterogeneity results.
Modeling with individual heterogeneity leads to rejecting the hypothesis that the mean dv levels
are the same in any of the three respondent groups.
Table 14. Hypothesis tests of the mean 50% acceptance criteria level
for the three groups in the Test 1 data modeled with individual
heterogeneity
Houston area
Mean dv = 29.40
Pilot (DC area)
Mean dv = 24.40
DC area Reject hypothesis
Mean dv = 30.57 Houston = DC (pr = 0.02)
Reject hypothesis
Pilot = DC (pr< 0.001)
Houston area
Reject Houston = Pilot
(pr< 0.001)
Page 17
SCI 1979
-------
Stratus Consulting Memorandum (2/3/2010)
Summary
This memorandum describes a series of logit regression analyses that estimated the percentage of
respondents that rated a haze (dv) level acceptable in four different studies of urban visibility.
The first analysis in this report estimated a separate logit function for each of the four studies:
Denver, British Columbia, Phoenix and Washington, DC (combining the data from the DC-2001
study and all Test 1 data from the DC-2009 study). The estimated mean 50% criteria levels in
the four cities (Table 4) are different, with the mean estimate ranging from 19.90 dv (Denver) to
29.03 dv (Washington, DC). The hypothesis tests presented in Table 5 found that there is a
statistically different logit function in each city (rejecting the null hypothesis that there was a
single function that applies to more than one city). The range of mean estimates from the 4 city
logit analysis is similar to the Candidate Protection Level range of 20 dv to 30 dv described in
the draft UFVA (EPA, 2009).
The remainder of this memorandum examined in more detail the data from the two Washington,
DC studies. In the first analysis focusing on only the Washington, DC data, we compared the
estimated mean 50% criteria levels from the 2001 study to the mean estimates from each of the
three tests in the 2009 study. Figure 3 and Table 7 show the estimated mean levels in the 2001
(mean = 28.65 dv) and 2009, Test 1 (mean = 29.30 dv) studies were similar, while the Test 2
(21.30 dv) and Test 3 (32.26) mean levels were quite different. The hypothesis tests presented in
Table 8 support that overall observation. The only hypothesis not rejected was the hypothesis
that the DC-2001 and Test 1 are the same (i.e., we cannot reject the hypothesis that they have the
same mean 50% criteria level). This finding supports our approach of combining the DC-2001
and the DC-2009, Test 1 data in the four city analysis.
In the second analysis of the Washington, DC data, we investigated whether the study
participants who lived in the Washington, DC Metro area had the same mean 50% criteria levels
as the participants who lived in the Houston metro area. This analysis involved three groups of
Washington, DC residents (the DC-2001 participants, the pilot project participants in the DC-
2009 study, and participants 1 through 12 in Test 1 of DC-2009). The hypothesis tests results in
Table 11 show that the participants in the DC-2001 and the Houston area residents in the DC-
2009 study are similar (i.e., we cannot reject the hypothesis they have the same mean 50%
criteria level). Our hypothesis testing further found however, that the DC-2001 participants had
statistically significantly different mean 50% criteria levels than either of the two groups of
Washington, DC area residents included in the DC-2009, Test 1 results.
The third analysis of the Washington, DC data investigated the effect of individual heterogeneity
of preferences. This analysis was limited to the DC-2009 data because it required more complete
information on the responses of each participant. The individual heterogeneity analysis found
modestly higher mean 50% criteria levels than the second analysis of the Washington, DC area
Page 18
SCI 1979
-------
Stratus Consulting Memorandum (2/3/2010)
residents. The hypothesis testing in this analysis rejected the hypothesis that the mean dv levels
were the same for the three groups who participated in Test 1.
This apparent inconsistency with the two hypothesis tests of analyses of subsets the Washington,
DC studies with the results of the hypothesis tests comparing the DC-2001 data with all of the
DC-2009, Test Idata may be due to having subdivided the participants of Test 1 into subsets
with too few members to provide stable results. Combining the DC-2001 data with all the Test 1
data provides the largest sample size available to estimate the logit preference function for
Washington, DC.
References
Hole, A.R. 2007. A comparison of approaches to estimating confidence intervals for willingness
to pay measures. Health Economics 16:827-840.
Krinsky, I. andL. Robb. 1986. On approximating the statistical properties of elasticities. The
Review of Economics and Statistics 68(4):715-719.
Smith, A.E. and S. Howell. 2009. An Assessment of the Robustness of Visual Air Quality
Preference Study Results. CRA International, Washington, DC. March 30. Prepared for the
Utility Air Regulatory Group. Submitted as supplemental material to presentation by Anne Smith
to the public meeting of the EPA Clean Air Science Advisory Council. April 2.
Stratus Consulting. 2009. Review of Urban Visibility Preference Studies. Prepared for Vicki
Sandiford, U.S. Environmental Protection Agency, Office of Air Quality Planning and
Standards. Stratus Consulting Inc., Boulder, CO. September 21.
U.S. EPA. 2009. Particulate Matter Urban-Focused Visibility Assessment; External Review
Draft. U.S. Environmental Protection Agency. September.
Page 19
SCI 1979
-------
This page left intentionally blank.
-------
United States Office of Air Quality Planning and Standards Publication No. EPA 452/R-10-004
Environmental Protection Health and Environmental Impacts Division
Agency Research Triangle Park, NC July 2010
------- |