&EPA
United States
Envirofunmlal Protection
Agnncy
Health Risk and Exposure Assessment
for Ozone
Final Report
Chapter 4 Appendices
-------
This page left intentionally blank
-------
EPA-452/R-14-004b
August 2014
Health Risk and Exposure Assessment for Ozone
Final Report
Chapter 4 Appendices
U.S. Environmental Protection Agency
Office of Air and Radiation
Office of Air Quality Planning and Standards
Health and Environmental Impacts Division
Risk and Benefits Group
Research Triangle Park, North Carolina 27711
-------
DISCLAIMER
This final document has been prepared by staff from the Risk and Benefits Group, Health
and Environmental Impacts Division, Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency. Any findings and conclusions are those of the authors and do
not necessarily reflect the views of the Agency.
Questions related to this document should be addressed to Dr. Bryan Hubbell, U.S.
Environmental Protection Agency, Office of Air Quality Planning and Standards, C539-07,
Research Triangle Park, North Carolina 27711 (email: hubbell.bryan@epa.gov).
-------
CHAPTER 4 APPENDICES
Title
APPENDIX 4A: Ambient Air Quality Monitoring Data
APPENDIX 4B: Modeling Technical Support Document for the
2013 Ozone Risk and Exposure Assessment
APPENDIX 4C: Air Quality Spatial Fields for the National
Mortality Risk Burden Assessment
APPENDIX 4D: Model-based Air Quality Adjustment Using the
Higher-order Decoupled Direct Method (HDDM)
APPENDIX 4E: Evaluation of Seattle Air Quality
Pages
(4A-1 to 4A-37)
(4B-1 to 4B-73)
(4C-lto4C-18)
(4D-lto4D-179)
(4E-1 to 4E-8)
-------
This page left intentionally blank
-------
APPENDIX 4A
Ambient Air Quality Monitoring Data
Table of Contents
4A-1. AMBIENT OZONE MONITORING AND AIR QUALITY DATA 4A-1
4A-1.1 Overview of Ambient Ozone Monitoring 4A-1
4A-1.2 Ambient Ozone Concentration Data 4A-3
4A-1.3 Data Handling 4A-5
4A-2. AIR QUALITY INPUTS FOR THE EXPOSURE AND CLINICAL-BASED RISK
ASSESSMENTS 4A-9
4A-2.1 Urban Case Study Areas 4A-9
4A-2.2 Air Quality Inputs to the Air Pollutants Exposure (APEX) Model 4A-10
4A-2.3 Evaluation of Air Quality Spatial Field Techniques 4A-27
4A-3. AIR QUALITY INPUTS FOR THE EPIDEMIOLOGY-BASED RISK ASSESSMENT
4 A-31
4A-3.1 Urban Case Study Areas 4A-31
4A-3.2 Air Quality Inputs to the Benefits Mapping and Analysis Program
(BenMAP) 4A-33
4A-4. REFERENCES 4A-37
4A-i
-------
List of Tables
Table 4A-1. Information on the 15 Urban Case Study Areas in the Exposure and Clinical Risk
Assessments 4A-10
Table 4A-2. Information on the 12 Urban Case Study Areas in the Epidemiology Based Risk
Assessment 4A-32
Table 4A-3. Summary of the air quality inputs toBenMAP 4A-36
List of Figures
Figure 4A-1. Map of U.S. ambient Os monitoring sites in operation during the 2006-2010
period 4A-3
Figure 4A-2. Map of monitored 8-hour Os design values for the 2006-2008 period 4A-8
Figure 4 A-3. Map of monitored 8-hour Os design values for the 2008-2010 period 4 A-9
Figure 4A-4. Numerical example of the Voronoi Neighbor Averaging (VNA) technique.. 4A-11
Figure 4A-5. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Atlanta 4A-13
Figure 4A-6. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Baltimore 4A-14
Figure 4A-7. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Boston 4A-15
Figure 4A-8. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Chicago 4A-16
Figure 4A-9. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Cleveland 4A-17
Figure 4A-10. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Dallas 4A-18
Figure 4A-11. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Denver 4A-19
Figure 4A-12. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Detroit 4A-20
Figure 4A-13. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Houston 4A-21
Figure 4A-14. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Los Angeles 4A-22
Figure 4A-15. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for New York 4A-23
Figure 4A-16. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Philadelphia 4A-24
Figure 4A-17. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Sacramento 4A-25
4A-ii
-------
Figure 4A-18. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for St. Louis 4A-26
Figure 4A-19. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Washington, D.C 4A-27
Figure 4A-20. Maps of monitored values (left), nearest neighbor spatial fields (center), and VNA
spatial fields (right) for selected hours in Atlanta (top), and Philadelphia
(bottom) 4A-28
Figure 4A-21. Density scatter plots and performance statistics for the cross-validation analysis of
nearest neighbor (NN; left) and Voronoi Neighbor Averaging (VNA; right) spatial
field techniques applied to monitored hourly Os concentrations in Atlanta,
2005 4A-29
Figure 4A-22. Density scatter plots and performance statistics for the cross-validation analysis of
nearest neighbor (NN; left) and Voronoi Neighbor Averaging (VNA; right) spatial
field techniques applied to monitored hourly Os concentrations in Detroit,
2005 4A-30
Figure 4A-23. Density scatter plots and performance statistics for the cross-validation analysis of
nearest neighbor (NN; left) and Voronoi Neighbor Averaging (VNA; right) spatial
field techniques applied to monitored hourly Os concentrations in Philadelphia,
2005 4A-30
-------
4A-1. AMBIENT OZONE MONITORING AND AIR QUALITY DATA
This section provides a brief overview of ambient Os monitoring in the U.S. (Section 4A-
1.1), the ambient Os concentration data extracted for use in the risk and exposure assessments
(Section 4A-1.2), and the data handling procedures used for determining compliance with the
existing and potential alternative standards as well as some other relevant air quality metrics
(Section 4A-1.3).
4A-1.1 Overview of Ambient Ozone Monitoring
The Clean Air Act establishes air quality monitoring requirements to provide information
on ambient concentrations for six criteria pollutants, including Os, and makes provisions for the
collection of other ambient air quality measurements, such as Os precursors. The federal
regulations for ambient air quality monitoring, including establishment and periodic assessment
of local monitoring networks, approved monitoring methods, operating schedules, and protocols
for data reporting, quality assurance, and certification, are in Part 58 of the Code of Federal
Regulations.
There were over 1,300 ambient Os monitoring sites actively operating in the U.S. in
2010. These monitoring sites are operated by over 100 federal, state, local, and tribal agencies,
and can be grouped into one of the following networks:
1) State and Local Air Monitoring Stations (SLAMS): Monitoring sites operated by state,
local, and tribal governments for the purposes of determining compliance with the
National Ambient Air Quality Standards (NAAQS), and providing ambient air quality
information to help state and local public health agencies evaluate and implement air
quality control programs. There were over 1,100 SLAMS Os monitoring sites operating
in 2010, making up over 80% of the U.S. ambient Os monitoring network. There are two
important subcategories of SLAMS monitors:
a. National Multi-pollutant Monitoring Network (NCore): Approximately 80 monitoring
sites (60 urban and 20 rural) operated by state and local agencies. These sites monitor
six criteria pollutants (CO, NO2, Os, SO2, PMio, and PM2.s) and other important
parameters for the purposes of assessing multi-pollutant impacts on public health, and
supporting air quality forecasting.
b. Photochemical Assessment Monitoring Stations (PAMS): Approximately 80
monitoring sites operated by state and local agencies with EPA funding. These sites
monitor Os and its precursors, including NO, NO2, total NOx, total reactive nitrogen
(NOy), and over 60 volatile organic compounds (VOCs) for the purposes of
understanding Os chemistry and transport, aiding photochemical modeling, and
4A-1
-------
evaluating Os precursor emissions control strategies in areas designated
nonattainment for the Os NAAQS. Some PAMS monitoring sites are co-located with
NCore monitoring sites.
2) Clean Air Status and Trends Network (CASTNET): Approximately 80 Os monitoring
sites operated year-round by EPA and the National Park Service (NPS) for the purpose of
determining Os levels in national parks and other rural areas.
3) Special Purpose Monitoring Stations (SPMS): These monitoring sites are used to support
various air quality monitoring objectives, such as specific public health and welfare
impacts studies, model evaluation, or monitoring network assessments. These
monitoring sites are often operated on a temporary basis (up to 24 months), and are
generally not used to determine compliance with the NAAQS. Some of these monitoring
sites may be operated by local industry or other private interest groups.
SLAMS monitoring sites are required to monitor for Os only during the required Os
monitoring season, which is defined for each state in Table D-3 of 40 CFR Part 58. Many states
also operate their Os monitors outside of the required Os monitoring season. States that are
required to operate some or all of their Os monitors on a year-round basis include Arizona,
California, Hawaii, Louisiana, New Mexico, and Texas. EPA regional offices may approve
waivers effectively shortening the length of the required Os monitoring season for some
individual monitoring sites (e.g. rural monitoring sites which may be inaccessible during the
winter months). CASTNET and NCore monitoring sites are typically operated on a year-round
basis.
The Federal Reference Method (FRM) for Os measurement is the Chemiluminescence
Method (CLM). The first ultraviolet (UV) absorption photometric analyzers were approved as
Federal Equivalent Methods (FEMs) in 1977 and gained rapid acceptance for NAAQS
compliance purposes due to ease of operation, relatively low cost, and reliability. All SLAMS
and CASTNET Os monitoring sites in the U.S. have been operating UV analyzers since 2005.
Figure 4A-1 shows the locations of the ambient Os monitoring sites used in the risk and
exposure assessments. Gray dots represent SLAMS monitoring sites, green dots represent
CASTNET sites, blue dots represent NCore and/or PAMS monitoring sites, and black dots
represent SPMS and other monitoring sites for which data were available.
4A-2
-------
Figure 4A-1. Map of U.S. ambient
period.
* SLAMS
• CASTNET
• NCORE/PAMS
• SPMS/OTHER
monitoring sites in operation during the 2006-2010
4A-1.2 Ambient Ozone Concentration Data
EPA's Air Quality System (AQS) database is a national repository for many types of air
quality and related monitoring data. AQS contains monitoring data for the six criteria pollutants
dating back to the 1970's, as well as more recent additions such as PIVfo.s speciation, air toxics,
and meteorology data. As of 2010, over 100 federal, state, local, and tribal agencies submitted
hourly Os concentration data collected from over 1,300 ambient monitoring sites to AQS.
Air quality monitoring data from 1,468 ambient Os monitoring sites in the U.S. were
extracted for use in the risk and exposure assessments. The initial dataset consisted of hourly Os
concentrations in ppb collected from these monitors between 1/1/2006 and 12/31/2010. Data for
nearly 1,400 of these monitors were extracted from the AQS database in October 2012, and the
remaining data were extracted from the CASTNET database at the same time. CASTNET
monitors operated by the National Park Service were included in the data extracted from AQS,
but the CASTNET monitors operated by EPA did not begin reporting data to AQS until 2011.
Data collected from these EPA operated CASTNET monitors prior to 2011 did not meet EPA's
quality assurance requirements, but the data were subject to quality assurance criteria, and it is
4A-3
-------
generally agreed that data collected from CASTNET monitors prior to 2011 is of comparable
quality to the regulatory data stored in AQS.
There were a number of subtle, yet noteworthy differences between the air quality data
used in the 1st draft REA and the air quality data used in this draft.
1. In the 1st draft REA, multiple AQS data extractions were used for the air quality inputs to
various parts of the risk and exposure assessments. In this draft, all air quality inputs
were derived from the data extraction described above.
2. In the 1st draft REA, data collected from EPA operated CASTNET sites and other non-
regulatory monitoring sites were not included in the air quality inputs to the risk and
exposure assessments. In this draft, these monitors were included, but were not used to
make determinations of meeting the existing standard or the potential alternative
standards.
3. In the 1st draft REA, data collected with Os analyzers not using federal reference or
equivalent methods were not included in the air quality inputs to the risk and exposure
assessments. In this draft, these monitors were included, but were not used to make
determinations of meeting the existing standard or the potential alternative standards.
4. In the 1st draft REA, reported hourly Os concentrations lower than the minimum detection
limit (MDL, 5 ppb for most Os analyzers) were replaced with a value of /^ MDL. This is
called the "standard sample value" for criteria pollutant data extracted from AQS. In this
draft, the actual reported sample values were used, effectively allowing concentrations
down to 0 ppb.
5. In the 1st draft REA, hourly Os concentrations flagged by the monitoring agencies and
concurred by EPA as having been affected by exceptional events were removed from the
air quality inputs to the risk and exposure assessments. In this draft, these data were
included, but were not used to make determinations of meeting the existing standard or
the potential alternative standards, which is consistent with EPA's exceptional events
policy.
6. In this draft, missing data gaps of 1 or 2 hours in length were filled in using linear
interpolation. These short gaps often occur at regular intervals in the ambient data due to
an EPA requirement for monitoring agencies to perform routine quality control checks on
their Os monitors. Quality control checks are typically performed between midnight and
6:00 AM when Os concentrations are low. Missing data gaps of 3 hours or more in
length were not replaced, and interpolated values were not used to make determinations
of meeting the existing standard or the potential alternative standards.
7. In this draft, hourly Os concentrations from 7 monitoring sites where multiple Os
analyzers were operated simultaneously in the same physical location were combined to
4A-4
-------
create a single hourly site record using the highest reported hourly concentration in each
hour.
8. In some instances, EPA regional offices may approve the data combinations for pairs of
nearby Os monitoring sites affected by short distance site relocations for the purpose of
making NAAQS determinations. In this draft, hourly Os concentrations from 12 such
pairs of monitoring sites were combined to create a single hourly site record for each pair.
4A-1.3 Data Handling
To determine whether or not the NAAQS have been met at an ambient air quality
monitoring site, a statistic commonly referred to as a "design value" must be calculated based on
3 consecutive years of data collected from that site. The form of the existing Os NAAQS design
value statistic is the 3-year average of the annual 4th highest daily maximum 8-hour Os
concentration in ppb, with all decimal digits truncated. The existing Os NAAQS are met at an
ambient monitoring site when the design value is less than or equal to 75 ppb. The data handling
protocols for calculating design values for the existing Os NAAQS are in 40 CFR Part 50,
Appendix P. In counties or other geographic areas with multiple monitors, the area-wide design
value is defined as the design value at the highest individual monitoring site, and the area is said
to have met the NAAQS if all monitors in the area are meeting the NAAQS. The initial hourly
Os concentration dataset was split into two design value periods, 2006-2008 and 2008-2010, and
subsequent analyses were conducted independently for these two periods. The following daily
summary statistics were calculated from the hourly Os concentrations:
1. Daily Maximum 8-hour Average (MDA8) Concentration: There are 24 consecutive 8-
hour periods in each day (midnight - 8:00 AM, 1:00 AM - 9:00 AM, ...,11:00 PM - 7:00
AM). Rolling 8-hour averages were calculated for each period with 6, 7, or 8 hours of
data available, using 6, 7, or 8 as the divisor, respectively, and 8-hour periods with fewer
than 6 hours of data available were not used. The 8-hour average values were stored in
the 1st, or start, hour of the 8-hour period. The MDA8 value is the highest of the 8-hour
average values for each day, and the MDA8 values for two consecutive days may have
some hours in common. MDA8 values were considered to be valid if there were
sufficient data available to calculate at least 18 of 24 possible 8-hour averages, or, if used
for design value calculations, if the MDA8 value is greater than the level of the standard.
This is the daily metric used in design values and the Smith et al. (2009) short-term
mortality study.
2. Daily 10:00 AM - 6:00 PM Mean Concentration: This is the rolling 8-hour average value
as defined above for the 8-hour period starting at 10:00 AM. This is the daily metric used
in the Zanobetti and Schwartz, 2008 short-term mortality study.
4A-5
-------
3. Daily Maximum 1-hour Concentration: This is the highest hourly Os concentration
reported during a given day. Daily Maximum 1-hour values were considered to be valid
if there were at least 18 hourly Os concentrations reported in a given day. This is the
daily metric used in the Jerrett et al. (2009) long-term mortality study.
4. Daily 24-hour Average Concentration: This is the simple arithmetic average of the
hourly Os concentrations reported during a given day. Daily 24-hour average values
were considered to be valid if there were at least 18 hourly Ch concentrations reported in
a given day. This is the daily metric used in the Bell et al. (2004) short-term mortality
study.
The daily summary statistics described above were then used to calculate the following
annual summary statistics:
1. Annual 4th Highest MDA8 Concentration: This is the 4th highest valid MDA8 value
measured at a given monitoring site in a given year. The 4th highest MDA8 values were
considered to be valid if there were valid MDA8 values available for at least 75% of the
days in the required Os monitoring season, or, if used for design value calculations, if the
4th highest MDA8 value was greater than the level of the standard. Data collected outside
of the required Os monitoring season were used in determining the 4th highest MDA8
value, but were not used in determining validity.
2. May - September Average MDA8: This is the average of all available valid MDA8
values at a given monitoring site during the May - September period (153 days). The
May - September average MDA8 values were considered to be valid if valid MDA8
values were available for at least 114 days (75%) in May - September of a given year.
Three-year averages of these values were calculated for the 2006-2008 period and used to
create the May - September average MDA8 national fused air quality surface described
in Appendix 4C. This surface was then used in the national-scale risk assessment based
on Smith et al. (2009) described in the Health Risk and Exposure Assessment (HREA)
Chapter 8.
3. June - August Average Daily 10:00 AM - 6:00 PM Mean: This is the average of all
available daily 10:00 AM - 6:00 PM mean values at a given monitoring site during the
June - August period (92 days). The June - August average daily mean values were
considered to be valid if daily 10:00 AM - 6:00 PM mean values were available for at
least 70 days (75%) in June - August of a given year. Three-year averages of these
values were calculated for the 2006-2008 period and used to create the June - August
average daily 10:00 AM - 6:00 PM mean national fused air quality surface described in
4A-6
-------
Appendix 4C. This surface was then used in the national-scale risk assessment based on
Zanobetti and Schwartz (2008) described in HREA Chapter 8.
4. April - September Average Daily Maximum 1-hour Concentration: This is the average
of all available valid daily maximum 1-hour values at a given monitoring site during the
April - September period. The April - September average daily maximum 1-hour values
were considered to be valid if valid daily maximum 1-hour values were available for at
least 137 days (75%) in April - September of a given year. Three-year averages of these
values were calculated for the 2006-2008 period and used to create the April - September
average daily maximum 1-hour national fused air quality surface described in Appendix
4C. This surface was then used in the national-scale risk assessment based on Jerrett et
al. (2009) described in HREA Chapter 8.
The design value statistic for the existing Os standard is the 3-year average of the annual
4th highest MDA8 values described above. Design values greater than the level of the existing
standard (76 ppb or higher) are automatically valid. Design values less than or equal to the level
of the existing standard must be based on 3 valid 4th highest MDA8 values, with the additional
requirement that the 3-year average of the annual data completeness statistics (percent of valid
MDA8 values within the required Os monitoring season) must be at least 90%. These same
criteria were chosen to determine design values for the potential alternative standards. The
implications of this choice are that in some cases, a monitoring site may have different design
values based on the level of the standard. This may occur for one of two reasons:
1. A monitoring site with insufficient data to determine a design value at higher standard
level may have sufficient data to show a violation at a lower standard level. For example,
if a monitoring site has a 3-year average annual 4th highest MDA8 value of 72 ppb, but
does not meet the data completeness criteria described above, then the site has a design
value of 72 ppb for the potential alternative standards of 70 ppb, 65 ppb, and 60 ppb, but
does not have a design value for the existing standard of 75 ppb.
2. EPA's current practice is to use only "valid" MDA8 values when determining the annual
4th highest MDA8 value. A valid MDA8 value is either based on at least 18 of 24
possible 8-hour average values, or it is greater than the level of the standard. This can
cause the design value to change based on the level of the standard, due to whether
certain days are considered valid or not. For example, suppose the five highest MDA8
values at a particular monitoring site for a given year are 78 ppb, 76 ppb, 75 ppb, 74 ppb,
and 70 ppb, but the 4th highest value is based on only 12 valid 8-hour averages. Then for
the existing standard, the 4th highest MDA8 value is 70 ppb, but for the 70 ppb alternative
standard, the 4th highest MDA8 value is 74 ppb.
4A-7
-------
Figure 4A-2 and Figure 4A-3 show the design values for the existing Os NAAQS for all
regulatory monitoring sites in the U.S. for the 2006-2008 and 2008-2010 periods, respectively.
In general, Os design values were lower in 2008-2010 than in 2006-2008, especially in the
Eastern U.S. There were 518 Os monitors in the U.S. with design values above the existing
standard in 2006-2008, compared to only 179 in 2008-2010.
8-Hour Ozone Design Values, 2006-2008
• 33-60 ppb (49 Sites)
61 - 65 ppb (65 Sites)
O 66-70 ppb (140 Sites)
O 71 - 75 ppb (279 Sites)
• 76-120 ppb (518 Sites)
PUERTO RICO
Figure 4A-2. Map of monitored 8-hour Os design values for the 2006-2008 period.
4A-8
-------
A a 8-Hour Ozone Design Values, 2008-2010
• 33-60 ppb (79 Sites)
• 61-65 ppb (165 Sites)
O 66-70 ppb (305 Sites)
• 71-75 ppb (300 Sites)
• 76-120 ppb (179 Sites)
PUERTO RICO
Figure 4A-3. Map of monitored 8-hour Os design values for the 2008-2010 period.
4A-2, AIR QUALITY INPUTS FOR THE EXPOSURE AND CLINICAL-BASED RISK
ASSESSMENTS
This section describes the 15 urban case study areas used in the exposure and clinical risk
assessments described in the HREA Chapters 5 and 6 (Section 4A-2.1), the hourly Cb
concentration spatial fields used as inputs to the Air Pollutants Exposure Model (APEX; Section
4A-2.2), and some methods evaluations supporting the change from nearest neighbor to the
Voronoi Neighbor Averaging (VNA) technique for generating the spatial fields (Section 4A-2.3).
4A-2.1 Urban Case Study Areas
The 15 urban case study areas in the exposure (HREA Chapter 5) and clinical risk
(HREA Chapter 6) assessments covered a large spatial extent, with boundaries generally similar
to those covered by the respective Combined Statistical Areas (CSA) defined by the U.S. Census
Bureau. Table 4A-1 gives some basic information about the 15 urban case study areas in the
4A-9
-------
exposure and clinical risk assessments, including the number of counties, number of ambient Os
monitoring sites, the required Os monitoring season, and the 2006-2008 and 2008-2010 design
values. All 15 urban case study areas had 8-hour Os design values above the existing standard in
2006-2008, while 13 areas had 8-hour Os design values above the existing standard in 2008-
2010. Chicago (74 ppb) and Detroit (75 ppb) had design values meeting the existing standard
during the 2008-2010 period. The design values in the 15 urban areas decreased by an average
of 6 ppb between 2006-2008 and 2008-2010, ranging from no change in Sacramento to a
decrease of 15 ppb in Atlanta.
Table 4A-1. Information on the 15 Urban Case Study Areas in the Exposure and Clinical
Risk Assessments.
Area Name
Atlanta
Baltimore
Boston
Chicago
Cleveland
Dallas
Denver
Detroit
Houston
Los Angeles
New York
Philadelphia
Sacramento
St. Louis
Washington
#of
Counties
33
7
10
16
8
11
13
9
10
5
27
15
7
17
26
#of03
Monitors
13
7
14
26
13
20
26
12
22
54
31
19
26
17
22
Population
(2010)
5,618,431
2,710,489
5,723,468
9,686,021
2,881,937
6,366,542
3,390,504
5,218,852
5,946,800
17,877,006
21,056,173
7,070,622
2,755,972
2,837,592
5,838,518
Required Os
Monitoring Season
March - October
April - October
April - September
April - October
April - October
January - December
March - September
April - September
January - December
January - December
April - October
April - October
January- December
April - October
April - October
2006-2008
DV (ppb)
95
91
83
78
82
89
86
81
91
119
90
92
102
85
87
2008-2010
DV (ppb)
80
89
77
74
77
86
77
75
85
112
84
83
102
77
81
4A-2.2 Air Quality Inputs to the Air Pollutants Exposure (APEX) Model
The Air Pollutants Exposure Model (APEX) described in HREA Chapter 5 requires
spatial fields of air quality concentrations with no missing values as inputs. In the 1st draft REA,
the spatial fields were generated using hourly Os concentrations from the nearest available
monitoring site for each census tract in four of the 15 urban case study areas (Atlanta, Denver,
Los Angeles, and New York). Here, the spatial fields were generated with hourly Os
concentrations interpolated using the Voronoi Neighbor Averaging (VNA; Gold, 1997; Chen et
al., 2004) technique described below for each census tract in the 15 urban case study areas. The
following paragraphs provide a numerical example of VNA used to estimate an Os concentration
value for census tract "E" in Figure 4A-4 below.
4 A-10
-------
A
D
Monitor: *
80 ppb
10 miles
G
*
B
Monitor:
90 ppb ^
15 miles f
7
. // .
/
/ «
*
Monitor:
100 ppb
20 miles
C
*
F
Monitor:
60 ppb
15 miles
I
*
#= Cent
*
= AirP
= AirP
Figure 4A-4. Numerical example of the Voronoi Neighbor Averaging (VNA) technique.
The first step in the VNA technique is to identify the set of nearest monitors for each
census tract. The left-hand panel of Figure 4A-4 presents a numerical example with nine census
tracts (squares) and seven monitoring sites (stars), with the focus on identifying the set of nearest
neighboring sites to census tract "E" in the center of the panel. The Delaunay triangulation
algorithm identifies the set of nearest neighboring monitors by drawing a set of polygons called
the "Voronoi diagram" around the census tract "E" centroid and each of the monitoring sites.
Voronoi diagrams have the special property that each edge of each of the polygons are the same
distance from the two closest points, as shown in the right-hand panel of Figure 4A-4.
The VNA technique then chooses the monitoring sites whose polygons share a boundary
with the census tract "E" centroid. These monitors are the "Voronoi neighbors", which are used
to estimate the concentration value for census tract "E". The VNA estimate of the concentration
value in census tract "E" is the inverse distance squared weighted average of the four monitored
concentrations. The further the monitor is from the center of census tract "E", the smaller the
weight. For example, the weight for the monitor in census tract "D" 10 miles from the census
tract "E" centroid is calculated as follows:
i/io2
1/102+1/152+1/152+1/202
= 0.4675
Equation (4A-1)
The weights for the other monitors are calculated in a similar fashion. The final VNA
estimate for census tract "E" is calculated as follows:
4 A-11
-------
VNA(E} = 0.4675 * 80 + 0.2078 * 90 + 0.2078 * 60 + 0.1169 * 100 = 80.3 ppb
Equation (4A-2)
The observed hourly Os concentrations in the 15 urban case study areas were used to
calculate VNA estimates for approximately 24,935 census tracts * 43,824 hours -1.1 billion
values. The actual number of values was lower than this, because values were not calculated for
hours outside the required monitoring season. However, the same VNA procedure was also used
to create hourly spatial surfaces based on air quality adjusted to meet the existing standard of 75
ppb, and air quality adjusted to meet the potential alternative standards of 70 ppb, 65 ppb, 60
ppb, and 55 ppb, which effectively increased the total number of VNA estimates by roughly a
factor of 5. The computations were executed using the R statistical computing program (R,
2012), with the Delaunay triangulation algorithm implemented in the "deldir" package (Turner,
2012).
Figure 4A-5 through Figure 4A-19 show maps of the 2006-2008 and 2008-2010 design
values and May-September "seasonal" average MDA8 values in the 15 urban case study areas
based on the VNA spatial fields. The top panels in each figure show the design values, while the
bottom panels show the seasonal average values. The left-hand panels in each figure show the
2006-2008 values, while the right-hand panels show the 2008-2010 values. The colored circles
in each panel represent census tract centroids, and the colored squares represent monitoring sites.
In each panel, counties colored pink indicate the study area boundaries used in the Zanobetti and
Schwartz (2008) and/or Smith et al. (2009b) epidemiology studies1, where applicable. Counties
colored gray indicate additional counties within the CBS A boundaries, and counties colored
peach indicate any additional counties included in the exposure and lung function risk
assessments. Note that the maps show some monitors outside of the study area boundaries. This
is because the VNA surfaces were generated using data from monitors within a 50 km radius of
the study area boundaries, in addition to the monitors inside the study areas.
1 The Zanobetti and Schwartz (2008) and Smith et al. (2009) study area boundaries were identical for 6 of the 12
urban case study areas, and had at least one county in common for all 12 urban case study areas. The counties
colored pink in Figures 4A-5 through 4A-19 refer to counties included in either of those two studies.
4 A-12
-------
50 60 70 80 90 100 50
2006 - 2008 Design Value (ppb)
60 70 80 90 100
2008 - 2010 Design Value (ppb)
30 40 50 60 70 30 40 50 60 70
2006 - 2008 Seasonal Average (ppb) 2008 - 2010 Seasonal Average (ppb)
Figure 4A-5. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Atlanta.
4 A-13
-------
50 60 70 80 90 100 50 60 70 80 90 100
2006 - 2008 Design Value (ppb) 2008 - 2010 Design Value (ppb)
30 40 50 60 70 30 40 50 60 70
2006 - 2008 Seasonal Average (ppb) 2008 - 2010 Seasonal Average (ppb)
Figure 4A-6. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Baltimore.
4 A-14
-------
50 60 70 80 90 100 50 60 70 80 90 100
2006 - 2008 Design Value (ppb) 2008 - 2010 Design Value (ppb)
30 40 50 60 70 30 40 50 60 70
2006 - 2008 Seasonal Average (ppb) 2008 - 2010 Seasonal Average (ppb)
Figure 4A-7. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Boston.
4 A-15
-------
50 60 70 80 90 100
2006 - 2008 Design Value (ppb)
50 60 70 80 90 100
2008 - 2010 Design Value (ppb)
30 40 50 60 70 30 40 50 60 70
2006 - 2008 Seasonal Average (ppb) 2008 - 2010 Seasonal Average (ppb)
Figure 4A-8. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Chicago.
4 A-16
-------
50 60 70 80 90 100 50 60 70 80 90 100
2006 - 2008 Design Value (ppb) 2008 - 2010 Design Value (ppb)
30 40 50 60 70 30 40 50 60 70
2006 - 2008 Seasonal Average (ppb) 2008 - 2010 Seasonal Average (ppb)
Figure 4A-9. Maps of design values and May - September average MDA8 values based on
VNA spatial fields for Cleveland.
4 A-17
-------
50 60 70 80 90 100 50 60 70 80 90 100
2006 - 2008 Design Value (ppb) 2008 - 2010 Design Value (ppb)
30 40 50 60 70 30 40 50 60 70
2006 - 2008 Seasonal Average (ppb) 2008 - 2010 Seasonal Average (ppb)
Figure 4A-10. Maps of design values and May - September average MDA8 values based
on VNA spatial fields for Dallas.
4 A-18
-------
50 60 70 80 90 100 50 60 70 80 90 100
2006 - 2008 Design Value (ppb) 2008 - 2010 Design Value (ppb)
30 40 50 60 70 30 40 50 60 70
2006 - 2008 Seasonal Average (ppb) 2008 - 2010 Seasonal Average (ppb)
Figure 4A-11. Maps of design values and May - September average MDA8 values based
on VNA spatial fields for Denver.
4 A-19
-------
50 60 70 80 90 100 50 60 70 80 90 100
2006 - 2008 Design Value (ppb) 2008 - 2010 Design Value (ppb)
30 40 50 60 70 30 40 50 60 70
2006 - 2008 Seasonal Average (ppb) 2008 - 2010 Seasonal Average (ppb)
Figure 4A-12. Maps of design values and May - September average MDA8 values based
on VNA spatial fields for Detroit.
4A-20
-------
50 60 70 80 90 100 50 60 70 80 90 100
2006 - 2008 Design Value (ppb) 2008 - 2010 Design Value (ppb)
30 40 50 60 70 30 40 50 60 70
2006 - 2008 Seasonal Average (ppb) 2008 - 2010 Seasonal Average (ppb)
Figure 4A-13. Maps of design values and May - September average MDA8 values based
on VNA spatial fields for Houston.
4A-21
-------
50 60 70 80 90 100 50 60 70 80 90 100
2006 - 2008 Design Value (ppb) 2008 - 2010 Design Value (ppb)
30 40 50 60 70 30 40 50 60 70
2006 - 2008 Seasonal Average (ppb) 2008 - 2010 Seasonal Average (ppb)
Figure 4A-14. Maps of design values and May - September average MDA8 values based
on VNA spatial fields for Los Angeles.
4A-22
-------
50 60 70 80 90 100
2006 - 2008 Design Value (ppb)
50 60 70 80 90 100
2008 - 2010 Design Value (ppb)
30 40 50 60 70 30 40 50 60 70
2006 - 2008 Seasonal Average (ppb) 2008 - 2010 Seasonal Average (ppb)
Figure 4A-15. Maps of design values and May - September average MDA8 values based
on VNA spatial fields for New York.
4A-23
-------
50 60 70 80 90 100 50 60 70 80 90 100
2006 - 2008 Design Value (ppb) 2008 - 2010 Design Value (ppb)
30 40 50 60 70 30 40 50 60 70
2006 - 2008 Seasonal Average (ppb) 2008 - 2010 Seasonal Average (ppb)
Figure 4A-16. Maps of design values and May - September average MDA8 values based
on VNA spatial fields for Philadelphia.
4A-24
-------
50 60 70 80 90 100 50 60 70 80 90 100
2006 - 2008 Design Value (ppb) 2008 - 2010 Design Value (ppb)
30 40 50 60 70 30 40 50 60 70
2006 - 2008 Seasonal Average (ppb) 2008 - 2010 Seasonal Average (ppb)
Figure 4A-17. Maps of design values and May - September average MDA8 values based
on VNA spatial fields for Sacramento.
4A-25
-------
50 60 70 80 90 100 50 60 70 80 90 100
2006 - 2008 Design Value (ppb) 2008 - 2010 Design Value (ppb)
30 40 50 60 70 30 40 50 60 70
2006 - 2008 Seasonal Average (ppb) 2008 - 2010 Seasonal Average (ppb)
Figure 4A-18. Maps of design values and May - September average MDA8 values based
on VNA spatial fields for St. Louis.
4A-26
-------
50 60 70 80 90 100 50 60 70 80 90 100
2006 - 2008 Design Value (ppb) 2008 - 2010 Design Value (ppb)
30 40 50 60 70 30 40 50 60 70
2006 - 2008 Seasonal Average (ppb) 2008 - 2010 Seasonal Average (ppb)
Figure 4A-19. Maps of design values and May - September average MDA8 values based
on VNA spatial fields for Washington, D.C.
4A-2.3 Evaluation of Air Quality Spatial Field Techniques
As mentioned previously, in the 1st draft REA the air quality spatial fields used as inputs
to APEX were based on hourly Os concentrations from the nearest monitoring site, while in this
4A-27
-------
draft the spatial fields were based on the VNA technique described in the previous section. This
section presents an evaluation comparing the relative accuracy of the nearest neighbor (NN) and
VNA techniques for generating spatial fields of hourly Os concentrations.
The evaluations were conducted over 4km gridded domains in Atlanta, Detroit, and
Philadelphia using monitored hourly Cb concentrations from 2005. Due to potential
discrepancies in the availability of data, only data collected within the required Os monitoring
seasons for each area (March - October for Atlanta; April - September for Detroit; April -
October for Philadelphia) were considered. Figure 4A-20 below shows maps of the monitored
values (AQS, left), NN values (center), and VNA values (right) in Atlanta (top) and Philadelphia
(bottom) for a single selected hour in each area. These maps show an extreme example of the
differences in the NN and VNA spatial fields. The NN fields (center column) have very sharp
breaks between air quality monitors, while the VNA fields (right column) tend to produce much
smoother surfaces. The NN fields also have a tendency to spread concentrations out over a large
area if the monitoring network is sparse. For example, the highest concentration at the monitor
in southern Atlanta is spread out over a large area in the NN fields, but this effect is somewhat
mitigated in the VNA fields.
AQS (07/26(2005 14:00:00) NN (07(26/2005 14:00:00) VNA (07/26/2005 14:00:00)
20 40 60 8'0 TOO120 140 0
AQS (06(26/2005 14:00:00!
20 4'0 6'0 80 100 120 140 0
NN (06/26/2005 14:00:00)
20 4'0 60 8'0 100 120 140
VNA (06(26/2005 14:00:00)
020 40 60 80 100 120 140
100120140
Figure 4A-20. Maps of monitored values (left), nearest neighbor spatial fields (center), and
VNA spatial fields (right) for selected hours in Atlanta (top), and Philadelphia (bottom).
4A-28
-------
A cross-validation analysis was performed to evaluate the relative accuracy of the
estimates from these two methods. For each hour in the required Os season, the concentrations
from each monitor in the 4 km gridded domain were withheld (one at a time) from the input
dataset, and the concentrations from the remaining monitors were used to estimate the
concentration at the monitoring site that was withheld using NN and VNA. Additional
monitoring sites within a 50 km radius of the 4 km gridded domain were used in the estimates,
but were not included in the set of monitors that were withheld from the analysis. The estimated
values were then compared to the respective monitored concentrations that were withheld, and
the relative accuracy of NN and VNA were assessed using three summary statistics: the
coefficient of variation (R2), the mean bias, and the root mean squared error (RMSE).
Figure 4A-21 (Atlanta), Figure 4A-22 (Detroit), and Figure 4A-23 (Philadelphia) show
density scatter plots and performance statistics for the NN and VNA spatial fields based on the
cross-validation analysis described above for the three urban areas. Each figure shows the
monitored hourly Os concentrations in ppb (AQS; x-axis) from the monitors which were
withheld from the analysis, versus the respective values estimated using the NN (left panel) and
VNA (right panel) techniques. In each panel, the plot region was split into 2 ppb x 2 ppb
squares, with the colors indicating the number of points falling into each square. The diagonal
line in each panel is a one-to-one reference line, and performance statistics for each method are
included in the upper left-hand corner. High R2 values, low mean bias values, and low RMSE
values are indicative of good performance.
AQS vs. NN
R-square = 0.77
Mean Bias = -0.(
RMSE = 10.78
200
150
ioo;
o
<
AQS vs. VNA
R-square = 0.839
Mean Bias = -0.52
RMSE = 8.63
200
150
100
50
50
AQS 100
150
50
AQS 100
150
Figure 4A-21. Density scatter plots and performance statistics for the cross-validation
analysis of nearest neighbor (NN; left) and Voronoi Neighbor Averaging (VNA; right)
spatial field techniques applied to monitored hourly Os concentrations in Atlanta, 2005.
4A-29
-------
AQSvs. NN
R-square = 0.811
Mean Bias = -0.6
RMSE = 8.85
200
150
AQS vs. VNA
100
50
<
z
R-square - 0.853
Mean Bias = -0.42
RMSE = 7.56
200
150
100
50
50
AQS 100
150
50
AQS 100
150
Figure 4A-22. Density scatter plots and performance statistics for the cross-validation
analysis of nearest neighbor (NN; left) and Voronoi Neighbor Averaging (VNA; right)
spatial field techniques applied to monitored hourly Os concentrations in Detroit, 2005.
AQSvs. NN
R-square = 0.798
Mean Bias = 0.08
RMSE = 9.48
200
150
AQS vs. VNA
ioo:
50
R-square = 0.868
Mean Bias = -0.01
RMSE = 7.5
200
150
100
50
50
AQS 100
150
50
AQS 100
150
Figure 4A-23. Density scatter plots and performance statistics for the cross-validation
analysis of nearest neighbor (NN; left) and Voronoi Neighbor Averaging (VNA; right)
spatial field techniques applied to monitored hourly Os concentrations in Philadelphia,
2005.
Both techniques had relatively low mean bias statistics in all three urban areas (< 1 ppb in
all cases). However, the VNA estimates had higher R2 statistics and lower RMSE statistics than
the NN estimates in all three urban areas, both of which indicate that the VNA technique
consistently produces more accurate estimates than NN. This is also reflected by the fact that
there is generally less scatter in the density plots based on the VNA estimates than those based
4A-30
-------
on the NN estimates. From a physical perspective, the VNA technique is also more appealing
since it does not produce sharp breakpoints in the estimated values between adjacent monitors
when the reported concentrations are different.
4A-3. AIR QUALITY INPUTS FOR THE EPIDEMIOLOGY-BASED RISK
ASSESSMENT
This section describes the 12 urban case study areas for the epidemiology-based risk
assessment described in HREA Chapter 7 (Section 7.4A-3.1), and the spatially averaged
"composite monitor" values used as inputs to the Benefits Mapping and Analysis Program
(BenMAP; Section 4A-3.2).
4A-3.1 Urban Case Study Areas
Three distinct sets of boundaries were considered for the 12 urban case study areas in the
epidemiology-based risk assessment:
1. Core Based Statistical Areas2 ("CBSA" boundaries)
2. Area boundaries defined in the Zanobetti and Schwartz (2008) study ("Z & S"
boundaries)
3. Area boundaries defined in Smith et al. (2009) study ("Smith" boundaries)
In the 1st draft REA, the short-term mortality risk estimates were based on the
concentration response functions estimated in the Zanobetti and Schwartz (2008) study using the
Z & S boundaries. Here, the primary set of short-term and long-term mortality risk estimates
described in HREA Chapter 7 are based on the concentration response functions estimated in the
Smith et al, 2009 study using the CBSA boundaries. The other two sets of boundaries are used
in sensitivity analyses. The first sensitivity analysis uses the Z & S boundaries to assess the
impact of changing from the quadratic rollback method used to adjust air quality in the 1st draft
REA to the HDDM adjustment method described in HREA Chapter 4 and Appendix 4D. The
second sensitivity analysis uses the Smith boundaries to assess the impact of pairing air quality
information based on the CBSA boundaries with the concentration-response functions which
were derived from air quality information based on the Smith boundaries.
Table 4A-2 gives some basic information about the 12 urban case study areas in the
epidemiology-based risk assessment for each set of boundaries. In general, the Z & S and Smith
areas were generally smaller and more focused on the urban population centers than the CBS As.
: Core Based Statistical Areas (CBSAs) are used by the Office of Management and Budget (OMB) to group U.S.
counties into urbanized areas. These groupings are updated by OMB every 5 years. The CBSAs used in the
epidemiology based risk assessment are based on the OMB delineations from 2008. For more information see:
http://www.whitehouse.gov/sites/default/files/omb/assets/bulletins/blO-02.pdf
4A-31
-------
The Z & S and Smith areas were identical in 6 of the 12 urban case study areas, and had at least
one county in common for all 12 study areas. The CBSAs were typically smaller than the
respective study areas described in Section 4A-2.1 for the exposure and clinical risk assessments,
with the exceptions of Baltimore, Dallas, and Houston, where the areas were identical. The final
two columns of Table 4A-2 show the annual 4th highest MDA8 values in ppb based on monitors
within the three sets of boundaries for 2007 and 2009, the two years upon which the risk
estimates in HREA Chapter 7 are focused.
Table 4A-2. Information on the 12 Urban Case Study Areas in the Epidemiology Based
Risk Assessment.
Area Name
Atlanta
Baltimore
Boston
Cleveland
Denver
Detroit
Houston
Los Angeles
New York
Philadelphia
Sacramento
St. Louis
Boundary
Definition
CBSA
Z&S
Smith
CBSA
Z&S
Smith
CBSA
Z&S
Smith
CBSA
Z&S
Smith
CBSA
Z&S
Smith
CBSA
Z&S
Smith
CBSA
Z&S
Smith
CBSA
Z&S
Smith
CBSA
Z&S
Smith
CBSA
Z&S
Smith
CBSA
Z&S
Smith
CBSA
Z&S
Smith
#of
Counties
28
4
2
7
2
1
7
3
1
5
1
1
10
1
3
6
1
1
10
1
1
2
1
1
23
5
6
11
1
1
4
1
1
16
2
1
#of03
Monitors
13
5
3
7
3
1
11
5
2
10
4
4
16
3
6
8
4
4
22
17
17
21
17
17
22
6
7
15
4
4
17
8
8
17
5
2
Population
(2010)
5,268,860
3,105,873
1,612,474
2,710,489
1,425,990
620,961
4,552,402
2,895,958
722,023
2,077,240
1,280,122
1,280,122
2,543,482
600,158
1,613,764
4,296,250
1,820,584
1,820,584
5,946,800
4,092,459
4,092,459
12,828,837
9,818,605
9,818,605
18,897,109
8,175,133
9,124,246
5,965,343
1,526,006
1,526,006
2,149,127
1,418,788
1,418,788
2,812,896
1,318,248
319,294
2007 4th
high (ppb)
102
98
98
92
83
73
89
88
72
83
83
83
97
76
76
93
92
92
90
90
90
105
105
105
94
83
94
102
95
95
93
90
90
94
94
91
2009 4th
high (ppb)
77
77
77
83
71
66
75
75
75
72
71
71
79
63
72
73
73
73
91
86
86
108
108
108
81
78
78
74
72
72
96
96
96
74
70
65
4A-32
-------
Since Os is not emitted directly but formed indirectly through photochemical reactions,
precursor emissions may continue to react and form Os downwind of emissions sources, and thus
the highest Os concentrations are often measured downwind of the highest concentration of
precursor emissions in the urban population center. This phenomenon is reflected in Table 4A-2,
where the highest monitored value in the CBS A occurs outside of the Smith boundaries in 9 of
12 areas in 2007, and in 7 of 12 areas in 2009. In addition, there were some instances where the
highest monitored Os concentrations occurred outside of the CBSAs, but within the respective
study areas used in the exposure and clinical risk assessments, which were designed to always
include the monitor associated with the highest area-wide design value. For example, in Los
Angeles, the CBS A includes Los Angeles County, CA and Orange County, CA, but the highest
Os concentrations are typically measured further downwind in Riverside County, CA and San
Bernardino County, CA.
4A-3.2 Air Quality Inputs to the Benefits Mapping and Analysis Program (BenMAP)
The air quality monitoring data used as inputs to the Benefits Mapping and Analysis
Program (BenMAP) were daily time-series of spatially averaged "composite monitor" values for
the 12 urban case study areas based on monitored air quality data. These composite monitor
values were calculated by taking hour-by-hour spatial averages of the hourly Os concentrations
for all monitors within a given study area, then calculating the four daily metrics described in
Section 4A-1.3:
1. Daily Maximum 8-hour Average (MDA8) Concentration
2. Daily 10:00 AM - 6:00 PM Mean Concentration
3. Daily Maximum 1-hour Concentration
4. Daily 24-hour Average Concentration
These four daily metrics were calculated based on each of the three sets of urban case
study area boundaries listed in the previous section, for a total of 12 daily metric/area boundary
combinations. Although these 12 values were provided based on available monitoring data for
each day in the 2006-2010 period, only a subset of these data were used for the air quality inputs
to BenMAP. Since the BenMAP software is designed to run for only one year at a time, we
chose to focus on air quality data from 2007 and 2009. In most cases, the 2007 data was meant
to represent a year with high Os levels, while 2009 was meant to represent a year with low Os
levels. In some areas, the data were also subset to a particular "Os season", either to match the
period used in the respective epidemiology study, or to avoid the potential disparity in data
availability which could arise if the air quality data included months outside of the required Os
monitoring season. Appendix 4D contains "box and whisker" plots showing the distribution of
4A-33
-------
composite monitor values in each area for current air quality, air quality adjusted to meet the
existing Os standard, and air quality adjusted to meet the potential alternative standards of 70
ppb, 65 ppb, and 60 ppb. These plots are stratified to show the effects of varying spatial extents
(CBSAs vs. Z & S areas), Os season lengths (June - August vs. April - October), and years
(2007 vs. 2009).
4A-34
-------
Table 4A-3 gives a list of the air quality inputs to the various BenMAP outputs discussed
in HREA Chapter 7, including the health endpoints modeled, the epidemiology studies from
which the concentration-response functions were derived, which urban case study areas were
modeled, which air quality metrics were used, which Os season was modeled, and which set of
boundaries were used in calculating the composite monitor values. Rows shaded pink represent
air quality inputs used in primary risk estimates, while rows shaded blue represent air quality
inputs for risk estimates included as sensitivity analyses.
4A-35
-------
Table 4A-3. Summary of the air quality inputs to BenMAP.
Health Endpoint
Emergency Room
Visits, Respiratory
Asthma Exacerbation,
Wheeze
Asthma Exacerbation,
Wheeze
Emergency Room
Visits, Asthma
Mortality, Long-Term
(Total and Respiratory)
Hospital Admissions,
All Respiratory
Hospital Admissions,
All Respiratory
Hospital Admissions,
Chronic Lung Disease
Hospital Admissions,
Chronic Lung Disease
Hospital Admissions,
Asthma
Mortality, Non-
Accidental
Mortality, Non-
Accidental
Mortality, Non-
Accidental
Mortality, Non-
Accidental
Emergency Room
Visits, Respiratory
Emergency Room
Visits, Respiratory
Mortality, All Cause
Mortality, All Cause
Epidemiology
Study
Darrow et al.
(2011)
Gent et al. (2003)
Gent et al. (2004)
Ito et al. (2007)
Jerrett et al. (2009)
Katsouyanni et al.
(2009)
Lin et al. (2000)
Lin et al. (2008)
Medina-Ramon et
al. (2006)
Silverman and Ito
(2010)
Smith et al. (2009)
Smith et al. (2009)
Smith et al. (2009)
Smith et al. (2009)
Strickland et al.
(2010)
Tolbert et al.
(2007)
Zanobetti and
Schwartz (2008)
Zanobetti and
Schwartz (2008)
Urban Case
Study Area(s)
Atlanta, GA
Boston, MA
Boston, MA
New York, NY
all 12 areas
Detroit, Ml
Los Angeles,
CA
New York, NY
all 12 areas
New York, NY
all 12 areas
all 12 areas
all 12 areas
all 12 areas
Atlanta, GA
Atlanta, GA
all 12 areas
all 12 areas
Air Quality
Metric
MDA8
1-Hour Max
MDA8
MDA8
Seasonal
Average3
1-Hour Max
24-Hour
Average
1-Hour Max
MDA8
MDA8
MDA8
MDA8
MDA8
MDA8
MDA8
MDA8
10AM-6PM
Mean
10AM-6PM
Mean
Os Season
March -
October
April -
September
April -
September
April -
September
April -
September
June -
August
June -
August
April -
October
May -
September
April -
August
Required Os
season
April -
October
June -
August
Required Os
season
May -
October
March -
October
June -
August
June -
August
Study Area
Boundaries
CBSA
CBSA
CBSA
CBSA
CBSA
CBSA
CBSA
CBSA
CBSA
CBSA
CBSA
CBSA
CBSA
Smith
CBSA
CBSA
CBSA
Z&S
3 For the Jerrett et al. (2009) long-term mortality study, the air quality inputs were based on an annual metric instead
of a daily metric. The annual metric was the April - September average of the daily maximum 1-hour
concentrations.
4A-36
-------
4A-4. REFERENCES
Abt Associates, Inc. (2010). Environmental Benefits and Mapping Program (Version 4.0).
Bethesda, MD. Prepared for U.S. Environmental Protection Agency, Office of Air
Quality Planning and Standards, Research Triangle Park, NC. Available on the Internet
at: http://www.epa.gov/air/benmap.
Bell, M.L., McDermott, A., Zeger S.L., Samet, J.M., Dominici, F. (2004). Ozone and Short-
Term Mortality in 95 U.S. Urban Communities, 1987-2000. Journal of the Air and Waste
Management Association, 292:2372-2378.
Chen, J., Zhao, R., Li, Z. (2004). Voronoi-based k-order Neighbor Relations for Spatial
Analysis. Journal of Photogrammetry and Remote Sensing, 59(1-2), 60-72.
Gold, C. (1997). Voronoi methods in GIS. In: Algorithmic Foundation of Geographic
Information Systems (va Kereveld M., Nievergelt, J., Roos, T., Widmayer, P., eds.).
Lecture Notes in Computer Science, Vol 1340. Berlin: Springer-Verlag, 21-35.
Jerrett, M., Burnett, R.T., Pope III C.A., Ito, K., Thurston, G., Krewski, D., Shi, Y., Calle, E.,
Thun, M. (2009). Long-Term Os Exposure and Mortality. New England Journal of
Medicine, 360:1085-1095.
R Core Team (2012). R: A language and environment for statistical computing. R Foundation
for Statistical Computing, Vienna, Austria, http ://www.R-project.org/.
Smith, R.L.; Xu, B., Switzer, P. (2009). Reassessing the Relationship between Ozone and Short-
Term Mortality in 95 U.S. Urban Communities. Inhalation Toxicology, 21: 37-61.
Turner, R. (2012). deldir: Delaunay Triangulation and Dirichlet (Voronoi) Tessellation. R
package version 0.0-19. http://CRAN.R-project.org/package=deldir
U.S. Environmental Protection Agency. (2012a). Integrated Science Assessment for Ozone and
Related Photochemical Oxidants: Final Report. U.S. Environmental Protection Agency,
Research Triangle Park, NC. EPA/600/R-10/076C. Available on the Internet at:
http://www.epa.gOv/ttn/naaqs/standards/ozone/s o3 2008 isa.html
U.S. Environmental Protection Agency. (2012b). Health Risk and Exposure Assessment for
Ozone, First External Review Draft. U.S. Environmental Protection Agency, Research
Triangle Park, NC. EPA/600/R-10/076C. Available on the Internet at:
http://www.epa.gov/ttn/naaqs/standards/ozone/s_o3_2008_rea.html
U.S. Environmental Protection Agency. (2012c). Total Risk Integrated Methodology (TRIM) -
Air Pollutants Exposure Model (APEX) Documentation (TRIM-Expo / APEX, Version
4.4). Available on the Internet at: http://www.epa.gov/ttn/fera/human_apex.html
Wells, B., Wesson, K., Jenkins, S. (2012). Analysis of Recent U.S. Ozone Air Quality Data to
Support the Os NAAQS Review and Quadratic Rollback Simulations to Support the First
Draft of the Risk and Exposure Assessment. Available on the Internet at:
http://www.epa.gOv/ttn/naaqs/standards/ozone/s o3td.html
Zanobetti, A., and J. Schwartz (2008). Mortality Displacement in the Association of Ozone with
Mortality: An Analysis of 48 Cities in the United States. American Journal of
Respiratory and Critical Care Medicine, 177:184-189.
4A-37
-------
APPENDIX 4B
Modeling Technical Support Document for the 2013 Ozone Risk
and Exposure Assessment
Table of Contents
4B-1. MODEL SET-UP AND SIMULATION 4B-1
4B-1.1 Model Domain 4B-1
4B-1.2 Model Time Period 4B-2
4B-1.3 Model Inputs: Meteorology 4B-3
4B-1.4 Model Inputs: Emissions 4B-4
4B-1.5 Model Inputs: Boundary and Initial conditions 4B-6
4B-2. EVALUATION OF MODELED OZONE CONCENTRATIONS 4B-7
4B-2.1 Operational Evaluation in the Northeast U.S 4B-9
4B-2.2 Operational Evaluation in the Southeast U.S 4B-27
4B-2.3 Operational Evaluation in the Midwest 4B-34
4B-2.4 Operational Evaluation in the Central U.S 4B-45
4B-2.5 Operational Evaluation in the Western U.S 4B-57
4B-3. REFERENCES 4B-71
4B-i
-------
List of Tables
Table 4B-1. Geographic elements of domain used in the CMAQ/HDDM modeling 4B-2
Table 4B-2. Vertical layer structure for 2007 WRF and CMAQ simulations 4B-4
Table 4B-3. Summary of emissions totals by sector for the 12km Eastern U.S. domain... 4B-6
Table 4B-4. Summary of CMAQ model performance at AQS monitoring sites in the
Northeastern U.S 4B-10
Table 4B-5. Summary of CMAQ model performance at AQS monitoring sites in the Boston
area 4B-15
Table 4B-6. Summary of CMAQ model performance at AQS monitoring sites in the New
York area 4B-17
Table 4B-7. Summary of CMAQ model performance at AQS monitoring sites in the
Philadelphia area 4B-20
Table 4B-8. Summary of CMAQ model performance at AQS monitoring sites in the
Baltimore area 4B-22
Table 4B-9. Summary of CMAQ model performance at AQS monitoring sites in the
Washington D.C. area 4B-25
Table 4B-10. Summary of CMAQ model performance at AQS monitoring sites in the
Southeastern U.S 4B-28
Table 4B-11. Summary of CMAQ model performance at AQS monitoring sites in the Atlanta
area 4B-32
Table 4B-12. Summary of CMAQ model performance at AQS monitoring sites in the Midwest.
4B-35
Table 4B-13. Summary of CMAQ model performance at AQS monitoring sites in the Chicago
area 4B-39
Table 4B-14. Summary of CMAQ model performance at AQS monitoring sites in the
Cleveland area 4B-41
Table 4B-15: Summary of CMAQ model performance at AQS monitoring sites in the Detroit
area 4B-43
Table 4B-16. Summary of CMAQ model performance at AQS monitoring sites in the central
U.S 4B-46
Table 4B-17. Summary of CMAQ model performance at AQS monitoring sites in the Saint
Louis area 4B-50
Table 4B-18. Summary of CMAQ model performance at AQS monitoring sites in the Dallas
area 4B-52
Table 4B-19. Summary of CMAQ model performance at AQS monitoring sites in the Houston
area 4B-55
Table 4B-20. Summary of CMAQ model performance at AQS monitoring sites in the western
U.S 4B-58
Table 4B-21. Summary of CMAQ model performance at AQS monitoring sites in the Denver
area 4B-63
Table 4B-22. Summary of CMAQ model performance at AQS monitoring sites in the
Sacramento area 4B-65
Table 4B-23. Summary of CMAQ model performance at AQS monitoring sites in the Los
Angeles area 4B-67
4B-ii
-------
List of Figures
Figure 4B-1. Map of the CMAQ modeling domain 4B-2
Figure 4B-2. Map of normalized mean bias for MDA8 Os concentrations in the Northeastern
U.S. for winter months in 2007 4B-11
Figure 4B-3. Map of normalized mean bias for MDA8 Os concentrations in the Northeastern
U.S. for spring months in 2007 4B-12
Figure 4B-4. Map of normalized mean bias for MDA8 Os concentrations in the Northeastern
U.S. for summer months in 2007 4B-13
Figure 4B-5. Map of normalized mean bias for MDA8 Os concentrations in the Northeastern
U.S. for fall months in 2007 4B-14
Figure 4B-6. Time series of 8-hr daily maximum Os concentrations at Boston monitoring sites
for April-October 2007. Observed values shown in black and modeled values
shown in red 4B-15
Figure 4B-7. Time series of hourly Os concentrations at Boston monitoring sites for January
2007. Observed values shown in black and modeled values shown in red. 4B-16
Figure 4B-8. Time series of hourly Os concentrations at Boston monitoring sites for April
2007. Observed values shown in black and modeled values shown in red. 4B-16
Figure 4B-9. Time series of hourly Os concentrations at Boston monitoring sites for July 2007.
Observed values shown in black and modeled values shown in red 4B-16
Figure 4B-10. Time series of hourly Os concentrations at Boston monitoring sites for October
2007. Observed values shown in black and modeled values shown in red. 4B-17
Figure 4B-11. Time series of 8-hr daily maximum Os concentrations at New York monitoring
sites for April-October 2007. Observed values shown in black and modeled
values shown in red 4B-18
Figure 4B-12. Time series of hourly Os concentrations at New York monitoring sites for January
2007. Observed values shown in black and modeled values shown in red. 4B-18
Figure 4B-13. Time series of hourly Os concentrations at New York monitoring sites for April
2007. Observed values shown in black and modeled values shown in red. 4B-18
Figure 4B-14. Time series of hourly Os concentrations at New York monitoring sites for July
2007. Observed values shown in black and modeled values shown in red. 4B-19
Figure 4B-15. Time series of hourly Os concentrations at New York monitoring sites for October
2007. Observed values shown in black and modeled values shown in red. 4B-19
Figure 4B-16. Time series of 8-hr daily maximum Os concentrations at Philadelphia monitoring
sites for April-October 2007. Observed values shown in black and modeled
values shown in red 4B-20
Figure 4B-17. Time series of hourly Os concentrations at Philadelphia monitoring sites for
January 2007. Observed values shown in black and modeled values shown in red.
4B-21
Figure 4B-18. Time series of hourly Os concentrations at Philadelphia monitoring sites for April
2007. Observed values shown in black and modeled values shown in red. 4B-21
Figure 4B-19. Time series of hourly Os concentrations at Philadelphia monitoring sites for July
2007. Observed values shown in black and modeled values shown in red. 4B-21
Figure 4B-20. Time series of hourly Os concentrations at Philadelphia monitoring sites for
October 2007. Observed values shown in black and modeled values shown in
red 4B-22
-------
Figure 4B-21. Time series of 8-hr daily maximum Os concentrations at Baltimore monitoring
sites for April-October 2007. Observed values shown in black and modeled
values shown in red 4B-23
Figure 4B-22. Time series of hourly Os concentrations at Baltimore monitoring sites for January
2007. Observed values shown in black and modeled values shown in red. 4B-23
Figure 4B-23. Time series of hourly Os concentrations at Baltimore monitoring sites for April
2007. Observed values shown in black and modeled values shown in red. 4B-23
Figure 4B-24. Time series of hourly Os concentrations at Baltimore monitoring sites for July
2007. Observed values shown in black and modeled values shown in red. 4B-24
Figure 4B-25. Time series of hourly Os concentrations at Baltimore monitoring sites for October
2007. Observed values shown in black and modeled values shown in red. 4B-24
Figure 4B-26. Time series of 8-hr daily maximum Os concentrations at Washington D.C.
monitoring sites for April-October 2007. Observed values shown in black and
modeled values shown in red 4B-25
Figure 4B-27. Time series of hourly Os concentrations at Washington D.C. monitoring sites for
January 2007. Observed values shown in black and modeled values shown in red.
4B-26
Figure 4B-28. Time series of hourly Os concentrations at Washington D.C. monitoring sites for
April 2007. Observed values shown in black and modeled values shown in red.
4B-26
Figure 4B-29. Time series of hourly Os concentrations at Washington D.C. monitoring sites for
July 2007. Observed values shown in black and modeled values shown in red.4B-
26
Figure 4B-30. Time series of hourly Os concentrations at Washington D.C. monitoring sites for
October 2007. Observed values shown in black and modeled values shown in
red 4B-27
Figure 4B-31. Map of normalized mean bias for MDA8 Os concentrations in the Southeastern
U.S. for winter months in 2007 4B-28
Figure 4B-32. Map of normalized mean bias for MDA8 Os concentrations in the Southeastern
U.S. for spring months in 2007 4B-29
Figure 4B-33. Map of normalized mean bias for MDA8 Os concentrations in the Southeastern
U.S. for summer months in 2007 4B-30
Figure 4B-34. Map of normalized mean bias for MDA8 Os concentrations in the Southeastern
U.S. for fall months in 2007 4B-31
Figure 4B-35. Time series of 8-hr daily maximum Os concentrations at Atlanta monitoring sites
for April-October 2007. Observed values shown in black and modeled values
shown in red 4B-32
Figure 4B-36. Time series of hourly Os concentrations at Atlanta monitoring sites for April
2007. Observed values shown in black and modeled values shown in red. 4B-33
Figure 4B-37. Time series of hourly Os concentrations at Atlanta monitoring sites for July 2007.
Observed values shown in black and modeled values shown in red 4B-33
Figure 4B-38. Time series of hourly Os concentrations at Atlanta monitoring sites for October
2007. Observed values shown in black and modeled values shown in red. 4B-33
Figure 4B-39. Map of normalized mean bias for MDA8 Os concentrations in the Midwest for
winter months in 2007 4B-35
4B-iv
-------
Figure 4B-40. Map of normalized mean bias for MDA8 Os concentrations in the Midwest for
spring months in 2007 4B-36
Figure 4B-41. Map of normalized mean bias for MDA8 Os concentrations in the Midwest for
summer months in 2007 4B-37
Figure 4B-42. Map of normalized mean bias for MDA8 Os concentrations in the Midwest for
fall months in 2007 4B-38
Figure 4B-43. Time series of 8-hr daily maximum Os concentrations at Chicago monitoring sites
for April-October 2007. Observed values shown in black and modeled values
shown in red 4B-39
Figure 4B-44. Time series of hourly Os concentrations at Chicago monitoring sites for January
2007. Observed values shown in black and modeled values shown in red. 4B-40
Figure 4B-45. Time series of hourly Os concentrations at Chicago monitoring sites for April
2007. Observed values shown in black and modeled values shown in red. 4B-40
Figure 4B-46. Time series of hourly Os concentrations at Chicago monitoring sites for July
2007. Observed values shown in black and modeled values shown in red. 4B-40
Figure 4B-47. Time series of hourly Os concentrations at Chicago monitoring sites for October
2007. Observed values shown in black and modeled values shown in red. 4B-41
Figure 4B-48. Time series of 8-hr daily maximum Os concentrations at Cleveland monitoring
sites for April-October 2007. Observed values shown in black and modeled
values shown in red 4B-42
Figure 4B-49. Time series of hourly Os concentrations at Cleveland monitoring sites for April
2007. Observed values shown in black and modeled values shown in red. 4B-42
Figure 4B-50. Time series of hourly Os concentrations at Cleveland monitoring sites for July
2007. Observed values shown in black and modeled values shown in red. 4B-42
Figure 4B-51. Time series of hourly Os concentrations at Cleveland monitoring sites for October
2007. Observed values shown in black and modeled values shown in red. 4B-43
Figure 4B-52. Time series of 8-hr daily maximum Os concentrations at Detroit monitoring sites
for April-October 2007. Observed values shown in black and modeled values
shown in red 4B-44
Figure 4B-53. Time series of hourly Os concentrations at Detroit monitoring sites for April
2007. Observed values shown in black and modeled values shown in red. 4B-44
Figure 4B-54. Time series of hourly Os concentrations at Detroit monitoring sites for July 2007.
Observed values shown in black and modeled values shown in red 4B-44
Figure 4B-55. Map of normalized mean bias for MDA8 Os concentrations in the Central U.S. for
winter months in 2007 4B-46
Figure 4B-56. Map of normalized mean bias for MDA8 Os concentrations in the Central U.S. for
spring months in 2007 4B-47
Figure 4B-57. Map of normalized mean bias for MDA8 Os concentrations in the Central U.S. for
summer months in 2007 4B-48
Figure 4B-58. Map of normalized mean bias for MDA8 Os concentrations in the Central U.S. for
fall months in 2007 4B-49
Figure 4B-59. Time series of 8-hr daily maximum Os concentrations at Saint Louis monitoring
sites for April-October 2007. Observed values shown in black and modeled
values shown in red 4B-50
4B-v
-------
Figure 4B-60. Time series of hourly (^concentrations at Saint Louis monitoring sites for
January 2007. Observed values shown in black and modeled values shown in red.
4B-51
Figure 4B-61. Time series of hourly Cb concentrations at Saint Louis monitoring sites for April
2007. Observed values shown in black and modeled values shown in red. 4B-51
Figure 4B-62. Time series of hourly Os concentrations at Saint Louis monitoring sites for July
2007. Observed values shown in black and modeled values shown in red. 4B-51
Figure 4B-63. Time series of hourly Os concentrations at Saint Louis monitoring sites for
October 2007. Observed values shown in black and modeled values shown in
red 4B-52
Figure 4B-64. Time series of 8-hr daily maximum Os concentrations at Dallas monitoring sites
for April-October 2007. Observed values shown in black and modeled values
shown in red 4B-53
Figure 4B-65. Time series of hourly Os concentrations at Dallas monitoring sites for January
2007. Observed values shown in black and modeled values shown in red. 4B-53
Figure 4B-66. Time series of hourly Os concentrations at Dallas monitoring sites for April 2007.
Observed values shown in black and modeled values shown in red 4B-53
Figure 4B-67. Time series of hourly Os concentrations at Dallas monitoring sites for July 2007.
Observed values shown in black and modeled values shown in red 4B-54
Figure 4B-68. Time series of hourly Os concentrations at Dallas monitoring sites for October
2007. Observed values shown in black and modeled values shown in red. 4B-54
Figure 4B-69. Time series of 8-hr daily maximum Os concentrations at Houston monitoring sites
for April-October 2007. Observed values shown in black and modeled values
shown in red 4B-55
Figure 4B-70. Time series of hourly Os concentrations at Houston monitoring sites for January
2007. Observed values shown in black and modeled values shown in red. 4B-56
Figure 4B-71. Time series of hourly Os concentrations at Houston monitoring sites for April
2007. Observed values shown in black and modeled values shown in red. 4B-56
Figure 4B-72. Time series of hourly Os concentrations at Houston monitoring sites for July
2007. Observed values shown in black and modeled values shown in red. 4B-56
Figure 4B-73. Time series of hourly Os concentrations at Houston monitoring sites for October
2007. Observed values shown in black and modeled values shown in red. 4B-57
Figure 4B-74. Map of normalized mean bias for MDA8 Os concentrations in the Western U.S.
for winter months in 2007 4B-59
Figure 4B-75. Map of normalized mean bias for MDA8 Os concentrations in the Western U.S.
for spring months in 2007 4B-60
Figure 4B-76. Map of normalized mean bias for MDA8 Os concentrations in the Western U.S.
for summer months in 2007 4B-61
Figure 4B-77. Map of normalized mean bias for MDA8 Os concentrations in the Western U.S.
for fall months in 2007 4B-62
Figure 4B-78. Time series of 8-hr daily maximum Os concentrations at Denver monitoring sites
for April-October 2007. Observed values shown in black and modeled.... values
shown in red 4B-63
Figure 4B-79. Time series of hourly Os concentrations at Denver monitoring sites for April
2007. Observed values shown in black and modeled values shown in red. 4B-64
4B-vi
-------
Figure 4B-80. Time series of hourly Cb concentrations at Denver monitoring sites for July 2007.
Observed values shown in black and modeled values shown in red 4B-64
Figure 4B-81. Time series of hourly Cb concentrations at Denver monitoring sites for October
2007. Observed values shown in black and modeled values shown in red. 4B-64
Figure 4B-82. Time series of 8-hr daily maximum Os concentrations at Sacramento monitoring
sites for April-October 2007. Observed values shown in black and modeled
values shown in red 4B-65
Figure 4B-83. Time series of hourly Osconcentrations at Sacramento monitoring sites for
January 2007. Observed values shown in black and modeled values shown in red.
4B-66
Figure 4B-84. Time series of hourly Os concentrations at Sacramento monitoring sites for April
2007. Observed values shown in black and modeled values shown in red. 4B-66
Figure 4B-85. Time series of hourly Os concentrations at Sacramento monitoring sites for July
2007. Observed values shown in black and modeled values shown in red. 4B-66
Figure 4B-86. Time series of hourly Os concentrations at Sacramento monitoring sites for
October 2007. Observed values shown in black and modeled values shown in
red 4B-67
Figure 4B-87. Time series of 8-hr daily maximum Os concentrations at Los Angeles monitoring
sites for April-October 2007. Observed values shown in black and modeled
values shown in red 4B-68
Figure 4B-88. Time series of hourly Os concentrations at Los Angeles monitoring sites for
January 2007. Observed values shown in black and modeled values shown in red.
4B-68
Figure 4B-89. Time series of hourly Os concentrations at Los Angeles monitoring sites for April
2007. Observed values shown in black and modeled values shown in red. 4B-68
Figure 4B-90. Time series of hourly Os concentrations at Los Angeles monitoring sites for July
2007. Observed values shown in black and modeled values shown in red. 4B-69
Figure 4B-91. Time series of hourly Os concentrations at Los Angeles monitoring sites for
October 2007. Observed values shown in black and modeled values shown in
red 4B-69
Figure 4B-92. Map of mean observed MDA8 Os concentrations at Los Angeles monitoring sites
for summer months (June, July, Aug) 2007 4B-70
Figure 4B-93. Map of normalized mean bias for MDA8 Os concentrations at Los Angeles
monitoring sites for summer months (June, July, Aug) 2007 4B-70
4B-vii
-------
4B-1. MODEL SET-UP AND SIMULATION
The air quality modeling underlying the HDDM adjustment methodology described in
the Health Risk and Exposure Assessment (HREA) Chapter 4 and Appendix 4D was performed
using CMAQv4.7.1 with HDDM for ozone (Os) (www.cmaq-model.org). A modified version of
CMAQ-HDDM-3D was used that tracked species concentrations through all modeled processes,
but tracked Os sensitivities through the chemistry, transport, and dry deposition subroutines only
and did not account for the effects of aerosol and cloud processing due to the uncertainties in
these model processes with respect to Os sensitivity as well as to conserve computational costs.
CMAQ was run using the carbon bond 2005 (CB05) gas-phase chemical mechanism (Gery et al.,
1989; Yarwood et al., 2005) and the AERO5 aerosol module which includes ISORROPIA for
gas-particle partitioning of inorganic species (Nenes et al., 1998) and secondary organic aerosol
treatment as described in Carlton et al. (2010).
4B-1.1 Model Domain
For this analysis, all CMAQ/HDDM runs were performed for a domain that covers the 48
contiguous states included portions of southern Canada and Northern Mexico with a 12 x 12 km
resolution (Figure 4B-1). The CMAQ simulations were performed with 24 vertical layers with a
top at about 17,600 meters, or 50 millibars (mb). Table 4B-1 and Table 4B-2 provides some
basic geographic information regarding the CMAQ domain and vertical layer structure,
respectively. Results from the lowest layer of the model were used for analyses to support the Os
HREA.
4B-1
-------
12US2 domain
x.y origin: -241
col: 396 row:246
Figure 4B-1. Map of the CMAQ modeling domain.
Table 4B-1. Geographic elements of domain used in the CMAQ/HDDM modeling.
Element
Map Projection
Grid Resolution
True Latitudes
Dimensions
Vertical extent
CMAQ Modeling Configuration: National Grid
Lambert Conformal Projection
12km
33 deg N and 45 deg N
396 x 246 x 24
24 Layers: Surface to 50 millibar level (Table 4B-2)
4B-1.2 Model Time Period
The CMAQ/HDDM modeling was performed for January and April-October of 2007.
The simulations included 10 day "ramp-up" periods from December 22-31, 2006 and from
March 22-31 2007 to minimize the effects of initial conditions. The ramp-up days were not
considered in the analysis for the HDDM results.
4B-2
-------
4B-1.3 Model Inputs: Meteorology
CMAQ model simulations require inputs of meteorological fields, emissions, and initial
and boundary conditions. The gridded meteorological data for the entire year of 2007 at the 12
km continental United States scale domain was derived from version 3.1 of the Weather
Research and Forecasting Model (WRF), Advanced Research WRF (ARW) core (Skamarock et
al., 2008). The WRF Model is a next-generation mesoscale numerical weather prediction system
developed for both operational forecasting and atmospheric research applications (http://wrf-
model.org). The 2007 WRF simulation included the physics options of the Pleim-Xiu land
surface model (LSM), Asymmetric Convective Model version 2 planetary boundary layer (PEL)
scheme, Morrison double moment microphysics, Kain-Fritsch cumulus parameterization scheme
and the RRTMG long-wave radiation (LWR) scheme (Gilliam and Pleim, 2010).
The WRF meteorological outputs were processed to create model-ready inputs for
CMAQ using the Meteorology-Chemistry Interface Processor (MCIP) package (Otte et al.,
2010), version 3.6, to derive the specific inputs to CMAQ: horizontal wind components (i.e.,
speed and direction), temperature, moisture, vertical diffusion rates, and rainfall rates for each
grid cell in each vertical layer. The WRF simulation used the same CMAQ map projection, a
lambert conformal projection centered at (-97, 40) with true latitudes at 33 and 45 degrees north.
The 12 km WRF domain consisted of 459 by 299 grid cells. The WRF simulation utilized 34
vertical layers with a surface layer of approximately 38 meters. Table 4B-2 shows the vertical
layer structure used in WRF and the layer collapsing approach to generate the CMAQ
meteorological inputs. CMAQ resolved the vertical atmosphere with 24 layers, preserving
greater resolution in the PEL.
In terms of the 2007 WRF meteorological model performance evaluation, an approach
which included a combination of qualitative and quantitative analyses was used to assess the
adequacy of the WRF simulated fields (U.S. EPA, 2011). The qualitative aspects involved
comparisons of the model-estimated synoptic patterns against observed patterns from historical
weather chart archives. Additionally, the evaluations compared spatial patterns of monthly
average rainfall and monthly maximum planetary boundary layer (PEL) heights. The statistical
portion of the evaluation examined the model bias and error for temperature, water vapor mixing
ratio, solar radiation, and wind fields. These statistical values were calculated on a monthly
basis.
4B-:
-------
Table 4B-2. Vertical layer structure for 2007 WRF and CMAQ simulations.
Layer Top Height (m)
17,145
14,490
12,593
11,094
9,844
8,766
7,815
6,962
6,188
5,477
4,820
4,208
3,635
3,095
2,586
2,198
1,917
1,644
1,466
1,292
1,121
952
787
705
624
544
465
386
307
230
153
114
76
38
Pressure (mb)
50
95
140
185
230
275
320
365
410
455
500
545
590
635
680
716
743
770
788
806
824
842
860
869
878
887
896
905
914
923
932
937
941
946
WRF
34
33
32
31
30
29
28
27
26
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
Depth (m)
2,655
1,896
1,499
1,250
1,078
951
853
775
711
657
612
573
539
509
388
281
273
178
174
171
168
165
82
81
80
80
79
78
78
77
38
38
38
38
CMAQ
24
23
22
21
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
Depth (m)
4,552
2,749
2,029
1,627
1,368
1,185
539
509
388
281
273
178
174
171
168
165
163
160
157
78
77
76
38
38
4B-1.4 Model Inputs: Emissions
The emissions data used are based on the 2007 Version 5 emissions modeling platform
developed for the Paniculate Matter (PM) NAAQS rule (U.S. EPA 2012a, U.S. EPA 2012b).
Some small updates to the 2007v5 platform are enumerated below. First we give a general
summary of the emissions processing performed for the PM NAAQS 2007 modeling emissions
inputs (more details are available in U.S. EPA 2012a, 2012b, and in the PM NAAQS section of
http://www.epa.gov/ttn/chief/emch/index.html). The 2008 National Emissions Inventory,
4B-4
-------
Version 2 (http://www.epa.gov/ttn/chief/net/2008inventory.html) was the starting point for these
emissions, with updates to specific source categories made where necessary to better represent
the year 2007. Emissions were processed to photochemical model inputs with the SMOKE
modeling system version 3.1 (Houyoux et al., 2000). For this analysis, emissions from wildfires
and prescribed burns are estimated based on a multi-year average of data from 2003 through
2010. Electric generating utilities (EGUs) emissions for 2007 are temporalized based on average
temporal profiles from 3 years of data. In addition, U.S. emissions are included from other point
sources, area sources, agricultural sources (ammonia only), anthropogenic fugitive dust sources,
nonroad mobile sources, onroad mobile sources, and biogenic sources. Onroad mobile sources
were created using EPA's MOVES 2010b model (www.epa.gov/otaq/models/moves), except that
California emissions were adjusted to match the county total emissions obtained directly from
the California Air Resources Board. Biogenic emissions were estimated using the Biogenic
Emissions Inventory System version 3.14 (BEISv3.14) (Pierce et al., 1998). Other North
American emissions are based on a 2006 Canadian inventory and 2008 Mexican inventory.
Emissions totals within the 12 km Eastern domain are summarized in Table 4B-3 for CO, NH3,
NOx, PMio, PM2.5, SO2, and VOC.
There are a few differences between the emissions data used for this analysis and the data
documented in the PM NAAQS 2007v5 platform technical support document (TSD). First, the
years used to compute the average fires were 2003-2010, versus 2003-2009 for the PM NAAQS.
Second, point source emissions for South Dakota were updated with more recent data. Finally, a
correction was made the spatial surrogates used for oil and gas emissions in the Western
Regional Air Partnership states and updated spatial surrogates for gas stations and dry cleaners
were used.
4B-5
-------
Table 4B-3. Summary of emissions totals by sector for the 12km Eastern U.S. domain.
Sector
Name
afdust
ag
c1c2rail
avefire
nonpt
nonroad
on road
othar
othon
othpt
ptipm
ptnonipm
cSmarine
US
cSmarine
non US
be is
total US
anthro
total
Sector description
Anthropogenic fugitive dust
Agricultural sources
Locomotive and marine
mobile sources (except C3
marine)
Average year fire emissions
Area sources
Off road equipment
Onroad mobile vehicles
Canada and Mexico area
sources
Canada and Mexico onroad
mobile sources
Canada and Mexico point
sources
Point sources: electric
generation units
Point sources other than
electric generating units
C3 marine vessels within 4
miles of the U.S. coast
C3 marine vessels more
than 4 miles off the U.S.
coast
Biogenic emissions
Total U.S. anthropogenic
emissions used in HDDM
(NOx and VOC only)
Domain-wide total
Emissions (1000 tons/year)
CO
219
15,598
4,335
17,834
36,757
4,225
5,173
1,331
704
2,934
13
86
8,211
97,420
NH3
3,595
0.6
256
155
1.9
145
671
25
21
25
68
4,964
NOx
1,338
216
1,229
1,878
7,561
918
631
1,280
3,357
2,077
138
1,047
1,931
17,578
23,602
PMio
5,854
44
1,589
768
188
363
1,510
23
241
437
583
12
87
11,699
PM2.5
825
41
1,347
676
178
277
451
18
159
330
409
11
80
4,802
SO2
49
118
402
101
40
154
11
2,504
9,136
1,589
105
646
14,854
VOC
60
2,797
6,671
2,781
3,186
1,815
405
626
43
1,074
5.1
38
48,616
13,819
68,117
4B-1.5 Model Inputs: Boundary and Initial conditions
The lateral boundary concentrations for the 12km US2 domain are provided by a three-
dimensional global atmospheric chemistry model, the GEOS-CHEM (Yantosca, 2004) model
(standard version 8-03-02 with version 8-02-03 chemistry). The global GEOS-CHEM model
simulates atmospheric chemical and physical processes driven by assimilated meteorological
observations from the NASA's Goddard Earth Observing System (GEOS-5). This model was
4B-6
-------
run for 2007 with a grid resolution of 2.0 degree x 2.5 degree (latitude-longitude) and 46 vertical
layers up to 0.01 hPa. The predictions were processed using the GEOS-2-CMAQ tool (Akhtar et
al., 2012, Henderson et al., 2013) and used to provide one-way dynamic boundary conditions at
one-hour intervals. The Os from these GEOS-Chem runs was evaluated by comparing to
satellite vertical profiles and ground-based measurements and found acceptable model
performance (Akhtar et al., 2012; Henderson et al., 2013).
Initial conditions were extracted from a slightly older model simulation using GEOS-
CHEM (Yantosca, 2004) version 8-02-03. The model simulation from which the initial
conditions were extracted was also run with a grid resolution of 2.0 of 2.0 degree x 2.5 degree
(latitude-longitude) and 47 vertical layers. A GEOS-Chem evaluation was conducted for the
purpose of validating the 2007 GEOS-Chem simulation outputs for their use as inputs to the
CMAQ modeling system. This evaluation included reproducing GEOS-Chem evaluation plots
reported in the literature for previous versions of the model (Lam, 2010).
4B-2. EVALUATION OF MODELED OZONE CONCENTRATIONS
CMAQ is a peer-reviewed, community air quality model that simulates the formation and
fate of photochemical oxidants, aerosol concentrations, acid deposition, and air toxics, over
multiple scales for given input sets of meteorological conditions and emissions. In order to
assure that CMAQ is an appropriate tool to estimate the AQ changes expected to result from a
given set of emissions reductions, the model is typically evaluated for each new
application. This evaluation consists of assessments of the model itself and an assessment of this
particular application of the model.
An independent panel consisting of academic and government experts assessed the
science within the CMAQ model version 4.7.1 in September 2011
(http://www.epa.gov/AMD/Reviews/201 l_CMAO_Review_FinalReport.pdf). Among the
conclusions of this peer-review report was the finding that the CMAQ science and evaluation
efforts was of "very high quality" and have provided a "foundation for the more reliable use of
the CMAQ modeling system".
As CMAQ model version 4.7.1 was being developed, a series of incremental diagnostic
tests were performed to assess how model performance varied across a variety of model
improvements. This analysis is summarized in Foley et al. (2010) as well as Godowitch et al.,
(2011). While time and resource intensive, this systematic incremental testing shows "the effect
of each scientific improvement on the simulated fields." This evaluation allowed for a clear
comparison with previous model versions and provided assurance that CMAQ v4.7.1 yielded
equivalent or improved performance relative to previous CMAQ versions.
4B-7
-------
Numerous dynamic evaluations of CMAQ's ability to simulate the change in air quality
resulting from emissions reductions have been conducted and summarized in the peer-reviewed
literature. For instance, Napelenok et al. (2011) concluded that the CMAQ model "is able to
reproduce the observed change in daily maximum 8-hour Os levels" at the majority of locations
when emissions uncertainty is considered. Other dynamic evaluations (Zhou et al., 2013,
Godowitch et al., 2010, Gilliland et al., 2008, Godowitch et al., 2007) have suggested that
CMAQ may be a conservative estimate of the air quality improvements resulting from emissions
reductions.
This TSD summarizes the ability of the model to reproduce 2007 conditions simulated
using specific emissions, meteorological, initial conditions and boundary conditions inputs
described above. This operational evaluation shows that the CMAQ model predictions for 2007
are equivalent or better than typical regional modeling simulations as summarized in Simon et al.
(2012).
In the following sections we present general model performance statistics and plots for
five regions of the U.S. We compare model predictions of maximum daily 8-hr average
(MDA8) Os concentrations to measurements reported in EPA's Air Quality System (AQS) which
is a repository of air pollution measurements made by EPA, state, local, and tribal agencies. For
the 2007 model performance evaluation, we ran CMAQ for 2007 using the same emissions as
those used for the 2007 HDDM runs, except that we included actual wild fires instead of average
fires and we used CEM data to for hourly EGU emissions (U.S. EPA, 2012b) instead of the
multi-year average temporal method used for the HDDM runs.
The model statistics presented here include mean bias, mean error, normalized mean bias,
and normalized mean error as calculated in Simon et al. (2012). Our analysis focuses on regional
model evaluation statistics from five U.S. regions as well as evaluations of the 15 urban areas
included in the Risk and Exposure Assessment. The five regions are defined as follows:
Northeast (Connecticut, Delaware, District of Columbia, Maine, Maryland, Massachusetts, New
Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont), Southeast
(Alabama, Florida, Georgia, Kentucky, Mississippi, North Carolina, South Carolina, Tennessee,
Virginia, West Virginia), Midwest (Illinois, Indiana, Michigan, Ohio, Wisconsin), Central
(Arkansas, Iowa, Kansas, Louisiana, Minnesota, Missouri, Nebraska, Oklahoma, Texas), and
West (Arizona, California, Colorado, Idaho, Nevada, New Mexico, Oregon, Utah, Wyoming).
Statistics for model performance in these regions and urban areas are shown by season in Table
4B-4 through Table 4B-23 for observed days with MDA8 Os values >= 60 ppb, observed days
with MDA8 Os < 60 ppb, and for all observed days. Plots are provided to show regional maps of
Normalized Mean Bias by season and time series of modeled and measured Os concentrations in
each urban area. Time series are provided for MDA8 Os from April-October 2007 and for
4B-8
-------
hourly Os from one month from each season in 2007 (January, April, July, October) where
monitoring data is available. Note that time series show average concentrations across all
monitors within each urban area and the number of monitors included in this average sometimes
changes by season since different monitors within each area take measurements over different
periods of the year.
4B-2.1 Operational Evaluation in the Northeast U.S.
Table 4B-4 shows that in the Northeastern U.S., model mean bias was generally less than
6 ppb and normalized mean bias was less than 15% in most cases. High Os days and
summertime days were more likely to be under-estimated by the model while low Os days and
wintertime days were more likely to be over-estimated by the model. Performance was best in
the spring and fall and had the largest errors in the winter. Five of the 15 urban areas evaluated
were in the Northeast: Boston, New York, Philadelphia, Baltimore, and Washington D.C.
Model performance at Boston area monitoring sites (Table 4B-5) was similar to that at
other Northeastern U.S. sites. The time series plots show that the model has skill at reproducing
measured day-to-day variability in MDA8 Cb concentrations (Figure 4B-6). Hourly daytime and
nighttime Os concentrations are also well modeled in all seasons with exception of a few 2-3 day
periods in July in which the model over-estimates daytime and nighttime Cb with some over-
estimates of daytime peaks in October (Figure 4B-7 through Figure 4B-10).1
Bulk performance statistics for MDA8 model estimates in New York (Table 4B-6) also
look equivalent to both the Boston statistics and those of the Northeast as a whole. Again, the 7-
month time series of MDA8 Os at NY area sites shows that the model captures synoptic
variations in Os concentrations (Figure 4B-11). Hourly New York area Os in January is generally
well captured by the model, although CMAQ does somewhat under-estimate the daytime peaks
(Figure 4B-12). The April time series (Figure 4B-13) shows that the model captures the range of
measured daytime and nighttime Os values but has a 3-day period of over-estimates in early
April and a week-long period of under-estimates later in the month. July and October Os is well
captured by the model on both low and high Os days and at night (Figure 4B-14 and Figure 4B-
15).
Bulk model performance statistics for Philadelphia (Table 4B-7) are equivalent to those
for the Northeast as a whole. The time series plots show that variations of MDA8 are well
captured in Philadelphia (Figure 4B-16) and that the model does a reasonable job of estimating
day to night Os changes in all seasons but has some under-estimates of daytime Os in January
(Figure 4B-17) and some over-estimates of daytime and nighttime Os in October (Figure 4B-20).
1 Note that the Y-axis scale for the various time series are not consistent
4B-9
-------
Baltimore MDA8 Os performance (Table 4B-8) is generally similar to Os performance in
the rest of the Northeast except for summertime values which are somewhat more overestimated
in Baltimore (8 ppb MB in Baltimore versus 4 ppb MB in the Northeast). Still, these bias and
error statistics are well within the range of state-of-the-science model performance described by
Simon et al. (2012). As shown by the time series in Figure 4B-21, most model overestimates in
this area occur from late July to late August. Outside of that time period, the MDA8 variations
from April-October are well captured by the model. The hourly time series plots for Baltimore
generally show reasonable agreement between the observed and modeled values although the
period of over-estimated daytime Os concentrations in July is apparent in Figure 4B-24.
Bulk statistics for Washington D.C. sites (Table 4B-9) are similar to statistics for the rest
of the Northeast region. The period of overestimated MDA8 values in late July through August
which was seen in nearby Baltimore is less pronounced in Washington D.C. (Figure 4B-26).
Hourly time series generally show reasonable performance with some overestimates of nighttime
Os in October (Figure 4B-30).
Table 4B-4. Summary of CMAQ model performance at AQS monitoring sites in the
Northeastern U.S.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
5085
0
5085
10420
1856
12276
11859
4114
15973
10036
1030
11066
MB (ppb)
-4.65
NA
-4.65
-0.69
-6.34
-1.47
5.58
-0.51
4.01
2.86
-4.21
2.21
NMB (%)
-17.0
NA
-17.0
-1.4
-9.1
-3.1
13.4
-0.7
8.2
8.1
-6.1
5.8
ME (ppb)
5.83
NA
5.83
5.63
8.25
6.03
7.99
7.78
7.93
6.23
8.06
6.40
NME (%)
21.2
NA
21.2
13.2
11.8
12.9
19.2
10.9
16.1
17.7
11.7
16.7
4B-10
-------
O3_8hrmax NMB (%) lor run 2007ee_ORDBC_v5_07c_v471_12US2 lor Winter for MANE-VU [No Cutoff]
units « %
coverage lirr,1 . 75%
l>50
140
30
120
10
Jo
1-10
-20
1-30
-40
<-50
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-2. Map of normalized mean bias for MDA8 Os concentrations in the
Northeastern U.S. for winter months in 2007.
4B-11
-------
O3_8hrmax NMB (%) for run 2Q07ee_ORDBC_v5_07c_v471_12US2 for Spring for MANE-VU [No Cutoff]
unrts = %
coverage limit = 75%
>50
40
30
20
10
0
-10
-20
-30
-40
<-50
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-3. Map of normalized mean bias for MDA8 Os concentrations in the
Northeastern U.S. for spring months in 2007.
4B-12
-------
O3_8hrmax NMB (%) for run 2007ee_ORDBC_v5_07c_v471_12US2 for Summer for MANE-VU [No Cutoff]
units = %
coverage limit = 75%
>50
40
30
20
10
0
-10
-20
-30
-40
<-50
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-4. Map of normalized mean bias for MDA8 Os concentrations in the
Northeastern U.S. for summer months in 2007.
4B-13
-------
O3_8hrmax NMB (%) for run 2007ee_ORDBC_v5_07c_v471_12US2 for Winter for MANE-VU [No Cutoff]
unrts = %
coverage limit = 75%
>50
40
30
20
10
0
-10
-20
-30
-40
<-50
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-5. Map of normalized mean bias for MDA8 Os concentrations in the
Northeastern U.S. for fall months in 2007.
4B-14
-------
Table 4B-5. Summary of CMAQ model performance at AQS monitoring sites in the
Boston area.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
265
0
265
680
87
767
784
210
994
433
62
495
MB (ppb)
-3.91
NA
-3.91
-1.32
-4.95
-1.73
6.71
-2.88
4.69
3.69
-6.48
2.42
NMB (%)
-15.1
NA
-15.1
-3.1
-7.0
-3.8
16.5
-4.0
9.9
11.3
-9.1
6.4
ME (ppb)
5.40
NA
5.40
5.73
8.46
6.04
8.47
8.76
8.53
6.22
9.92
6.68
NME (%)
20.9
NA
20.9
13.5
11.9
13.2
20.8
12.1
18.0
19.0
13.9
17.8
2007ee_ORDBC_v5_07c_v471_12US2 O3_8hrmax for AQSJJally Site: Load_File
100 -
80 -
60 -
40 -
20 -
AQS_Daily
2007ee ORDBC v5 07c V471 12US2
# of Sites: 12
Sile:Load Fife
iiiiiu imiimiii miiiiMiiiimiJiiiiii iimiiiiiu
AprOI AprlS May05 May 23 Jun 09 Jun 26 Jul 12 Jul 27 Aug 12 Aug 29 Sep 15 Oct02 Oct19
Date
Figure 4B-6. Time series of 8-hr daily maximum Os concentrations at Boston monitoring
sites for April-October 2007. Observed values shown in black and modeled values shown
in red.
4B-15
-------
2007ee ORDBC _v507c v471 12US2 O3 for AQS Hourly Site: Load File
I
s
50 -
40 -
30 -
2°-
10 -
o -
AQS Hourly
2007eejDRDBC_vb_07c_v471 _12US2
# of Sites: 3
Site: Load File
Jan 01 Jan 03 Jan 05 Jan OB Jan 10 Jan 12 Jan 15 Jan 17 Jan 20 Jan 22 Jan 24 Jan 27 Jan 29 Jan 31
Date
Figure 4B-7. Time series of hourly Os concentrations at Boston monitoring sites for
January 2007. Observed values shown in black and modeled values shown in red.
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_File
100 -
80 -
60 H
40 -
20 -
AQS_Hourly
2007ee ORDBC v5 07c v471 12US2
# of Sites: 12
Site: Load File
Apr 01 Apr 03 Apr 05 Apr 07 Apr 10 Apr 12 Apr 14 Apr 17 Apr 19 Apr 21 Apr 23 Apr 26 Apr 28 Apr 30
Date
Figure 4B-8. Time series of hourly Os concentrations at Boston monitoring sites for April
2007. Observed values shown in black and modeled values shown in red.
2007ee_ORDBC_v5_07c_v47l_12US2 O3 for AQS_Hourly Site: Load_Flle
100 -
80 -
60 -
40-
20 -
AQSJHourly
2007ee ORDBC v5 07e V471 12US2
# of Sites: 11
Site: Load File
JulOI Jul03 JulOS Jui07 Jul09 JuM 1 Jul 14 Jut 16 Jul 18 Jul 20 Jul 22 Jul 24 Jul 27 Jul 29 Jul 31
Date
Figure 4B-9. Time series of hourly Os concentrations at Boston monitoring sites for July
2007. Observed values shown in black and modeled values shown in red.
4B-16
-------
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_File
80 -
60 -
.o
& 4Q -
8
20 -
0 -
AQSJHourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: 3
Site: Load File
OctOI Ocl03 Oct 05 OctOS Oct 10 Oct 12 Oct 15 Ocl17 Oct 20 Oct 22 Oct 24 Ocl 27 Oct 29 Oct 31
Dale
Figure 4B-10. Time series of hourly Os concentrations at Boston monitoring sites for
October 2007. Observed values shown in black and modeled values shown in red.
Table 4B-6. Summary of CMAQ model performance at AQS monitoring sites in the New
York area.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
916
0
916
1547
200
1747
1469
635
2104
1443
108
1551
MB (ppb)
-4.14
NA
-4.14
-1.20
-6.05
-1.75
5.16
1.48
4.05
2.01
-5.60
1.48
NMB (%)
-17.8
NA
-17.8
-3.0
-8.5
-4.0
12.3
2.0
7.9
6,2
-8.1
4.2
ME (ppb)
5.34
NA
5.34
6.04
9.45
6.43
8.53
9.25
8.75
6.40
8.90
6.58
NME (%)
22.9
NA
22.9
15.0
13.2
14.7
20.4
12.7
17.1
19.7
12.8
18.7
4B-17
-------
2007ee ORDBC _v5 07c v471 12US2 O3 Shrmax for AQS Dally Site: Load File
AQS Daily
2007ee ORDBC v5 07c v471 12US2
20 -
Apr 01 Apr 18 May 05 May 23 Jun 09 Jun 26 Jul 12 Jul 27 Aug 12 Aug 29 Sep 15 Oct02 Oct19
Date
Figure 4B-11. Time series of 8-hr daily maximum Os concentrations at New York
monitoring sites for April-October 2007. Observed values shown in black and modeled
values shown in red.
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_File
40 -
30 -
I
- 20 -
n
O
10 -
o -
AQS_Hourly
2007ee ORDBC v5 07c v471 12US2
« of Sites: 11
Site: Load File
Jan 01 Jan 03 Jan 05 Jan 08 Jan 10 Jan 12 Jan 15 Jan 17 Jan 20 Jan 22 Jan 24 Jan 27 Jan 29 Jan 31
Dale
Figure 4B-12. Time series of hourly Os concentrations at New York monitoring sites for
January 2007. Observed values shown in black and modeled values shown in red.
2007ee_ORDBC_v5_07c_v47l_12US2 O3 for AQS_Hourly Site: Load_Flle
80 -
— 60 -
|
8 40 ~
20 -
AQS_Hourly
2007ee ORDBC v5 07c V471 12US2
it of Sites: 17
Site: Load File
Apr 01 Apr 03 Apr 05 Apr 07 Apr 10 Apr 12 Apr 14 Apr 17 Apr 19 Apr 21 Apr 23 Apr 26 Apr 28 Apr 30
Date
Figure 4B-13. Time series of hourly Os concentrations at New York monitoring sites for
April 2007. Observed values shown in black and modeled values shown in red.
4B-18
-------
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_File
120
100
» 60
o
40
20
AQS_Hourly
2007ee ORDBC v5 07c v471 12US2
# of Sites: 17
Site: Load File
Jul01 Jul03 Jul05 Jul07 Jul09 JuM1 JuM4 Jul 16 Jul 18 Jul 20 Jul 22 Jul 24 Jul 27 Jul 29 Jul 31
Date
Figure 4B-14. Time series of hourly Os concentrations at New York monitoring sites for
July 2007. Observed values shown in black and modeled values shown in red.
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_Flle
80 -
60 -
40 -
20 -
0 -1
AQS_Hourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: 17
Site: Load File
OctOI OctOS Oct05 OctOS OctIO Oct12 Oct15 Ocl17 Oct 20 Oct 22 Oct 24 Oct 27 Oct 29 Oct 31
Date
Figure 4B-15. Time series of hourly Os concentrations at New York monitoring sites for
October 2007. Observed values shown in black and modeled values shown in red.
4B-19
-------
Table 4B-7. Summary of CMAQ model performance at AQS monitoring sites in the
Philadelphia area.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
684
0
684
942
204
1146
778
507
1285
1005
103
1108
MB (ppb)
-3.48
NA
-3.48
-0.80
-7.01
-1.91
6.25
1.23
4.27
2.69
-4.51
2.02
NMB (%)
-14.4
NA
-14.4
-1.9
-9.9
-4.0
9.44
1.7
7.7
7.8
-6.7
5.4
ME (ppb)
5.15
NA
5.15
5.93
8.6
6.40
14.1
7.80
8.79
6.01
7.68
6.16
NME (%)
21.3
NA
21.3
14.1
12.1
13.5
21.3
10.8
15.9
17.4
11.4
16.4
2007ee ORDBC v5 07c v471 12US2 O3 Shrmax lor AQS Dally Site: Load_Flle
120 -
100 -
80 ~
60 -
40-
20 -
— AQS_Daily
2007ee ORDBC v5 07c V471 12US2
# of Sites: 15
Site: Load File
Apr 01 Apr 18 May 05 May 23 Jun 09 Jun 26 Jul 12 Jul 27 Aug 12 Aug 29 Sep 15 Oct 02 Oct 19
Dale
Figure 4B-16. Time series of 8-hr daily maximum Os concentrations at Philadelphia
monitoring sites for April-October 2007. Observed values shown in black and modeled
values shown in red.
4B-20
-------
2007ee ORDBC v5 07c v471 12US2 O3 for AQS Hourly Site: Load File
40 -
30 -
20 -
AQS Hourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: B
Site: Load File
JanOI Jan03 Jan 05 Jan 08 Jan 10 Jan 12 Jan 15 Jan17 Jan 20 Jan 22 Jan 24 Jan 27 Jan 29 Jan 31
Dale
Figure 4B-17. Time series of hourly Os concentrations at Philadelphia monitoring sites for
January 2007. Observed values shown in black and modeled values shown in red.
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_Flle
100 -
80 -
60 -
8 40 -I
20 -
0 -
AQS Hourly
2007ee ORDBC v507cv471 12US2
it of Sites: 15
Site: Load File
Apr 01 Apr 03 Apr 05 Apr 07 Apr 10 Apr 12 Apr 14 Apr 17 Apr 19 Apr 21 Apr 23 Apr 26 Apr 28 Apr 30
Date
Figure 4B-18. Time series of hourly Os concentrations at Philadelphia monitoring sites for
April 2007. Observed values shown in black and modeled values shown in red.
2007ee ORDBC v5 07c V471 12US2 O3 for AQS Hourly Site: Load File
140 -
120 -
100 -
| 80 -
g 60 -
40 -
20 -
0 -
AQS Hourly
2007ee ORDBC v5 07c v471 12US2
# of Sites: 15
Site: Load File
JulOI Jul03 JulOS Jill 07 Jul09 JuM 1 JuM4 Jul 16 Jul 18 Jul 20 Jul 22 Jul 24 Jill 27 Jul 29 Jul 31
Date
Figure 4B-19. Time series of hourly Os concentrations at Philadelphia monitoring sites for
July 2007. Observed values shown in black and modeled values shown in red.
4B-21
-------
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_Flle
80 -
60 -
.0
i
40 -
20 -
0 -
AQSJHourly
2007ee ORDBC v5 07c V471 12US2
# of Sites; 15
Site: Load File
OctOI Ocl03 Oct 05 Oct 08 OctIO Oct 12 Ocl 15 Ocl 17 Oct 20 Oct 22 Oct 24 Ocl 27 Ocl 29 Oct 31
Date
Figure 4B-20. Time series of hourly Os concentrations at Philadelphia monitoring sites for
October 2007. Observed values shown in black and modeled values shown in red.
Table 4B-8. Summary of CMAQ model performance at AQS monitoring sites in the
Baltimore area.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
78
0
78
358
92
450
382
255
637
389
51
440
MB (ppb)
-1.50
NA
-1.50
4.62
-0.86
3.50
10.40
4.18
7.90
6.63
-1.07
5.74
NMB (%)
-7.0
NA
-7.0
11.1
-1.2
7.4
23.2
5.7
14.1
17.9
-1.5
14.1
ME (ppb)
4.63
NA
4.63
6.81
5.92
6.62
11.60
8.95
10.50
8.35
7.67
8.27
NME (%)
21.5
NA
21.5
16.3
8.5
14.0
25.8
12.3
18.8
22.5
11.0
20.2
4B-22
-------
2007ee_ORDBC_v5_07c_v471_12US2 O3_8hrmax for AQS_Dally Site: Load_Flle
120 -
4 100~
a.
'x' 80 -
1 60 -
8' ..
20 -
AQS_Daily
2007ee ORDBC v5 07c V471 12US2
# of Sites: 7
Site: Load File
Apr 01 Apr 18 May 05 May 23 Jun 09 Jun 26 Jul 12 Jul 27 Aug 12 Aug 29 Sep15 Oct 02 Oct19
Date
Figure 4B-21. Time series of 8-hr daily maximum Os concentrations at Baltimore
monitoring sites for April-October 2007. Observed values shown in black and modeled
values shown in red.
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_Flle
50
40 -
30 -
g 20 H
AQS_Hourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: 1
Site: Load Fits
Jan 01 Jan 03 Jan 05 Jan 08 Jan 10 Jan 12 Jan 15 Jan 17 Jan 20 Jan 22 Jan 24 Jan 27 Jan 29 Jan 31
Date
Figure 4B-22. Time series of hourly Os concentrations at Baltimore monitoring sites for
January 2007. Observed values shown in black and modeled values shown in red.
2007ee ORDBC v5 07c v471 12US2 O3 for AQS Hourly Site: Load File
100 -
80 -
i. 60 -
E
o
O 40 -
20 -
0 -
AQS Hourly
2007ee ORDBC v5 07c v471 12US2
* of Sites: 7
Site: Load File
Apr 01 Apr 03 Apr 05 Apr 07 Apr 10 Apr 12 Apr 14 Apr 17 Apr 19 Apr 21 Apr 23 Apr 26 Apr 28 Apr 30
Date
Figure 4B-23. Time series of hourly Os concentrations at Baltimore monitoring sites for
April 2007. Observed values shown in black and modeled values shown in red.
4B-23
-------
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_FIIe
140
120 -
100 -
n 60 -
O
40 -
20 -
0 -
AQSJHourly
2007ee ORDBC v5 07c V471 12US2
it of Sites: 7
Site: Load File
JulOI Jul03 Jul05 Jul07 Jul 09 Jul 11 JuM4 Jul 16 Jul 18 Jul 20 Jul 22 Jul 24 Jul 27 Jul 29 Jul 31
Date
Figure 4B-24. Time series of hourly Os concentrations at Baltimore monitoring sites for
July 2007. Observed values shown in black and modeled values shown in red.
2007ee ORDBC _v507c v471 12US2 O3 for AQS Hourly Site: Load File
100 -
80 -
60 -
8 40 -\
20 -
0 -
AQS Hourly
2007ee ORDBC v5 07c v471 12US2
#olSiles:7
Site: Load_File
OctOI Oct03 Oct 05 Oct 08 OctIO Oct 12 Oct 15 Oct17 Oct 20 Oct 22 Oct 24 Oct 27 Oct 29 Oct 31
Dale
Figure 4B-25. Time series of hourly Os concentrations at Baltimore monitoring sites for
October 2007. Observed values shown in black and modeled values shown in red.
4B-24
-------
Table 4B-9. Summary of CMAQ model performance at AQS monitoring sites in the
Washington D.C. area.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
565
0
565
1120
334
1454
1066
819
1885
1188
106
1394
MB (ppb)
-4.97
NA
-4.97
-0.95
-4.98
-1.88
6.55
2.36
4.73
2.07
-2.03
1.46
NMB (%)
-20.5
NA
-20.5
-2.2
-7.3
-3.8
13.9
3.3
8.3
5.4
-3.0
3.5
ME (ppb)
6.12
NA
6.12
6.11
7.00
6.31
8.24
7.13
7.76
7.22
8.37
7.39
NME (%)
25.2
NA
25.2
14.2
10.2
12.9
17.5
10.2
13.6
19.0
12.4
17.4
2007ee_ORDBC_v5_07c_v471_12US2 O3_8hrmax for AQS_Dally Site: Load_Flle
100 -
eo -
8 40 H
20 -
AQS_Daily
2007ee ORDBC v5 07c v471 12US2
# of Sites: 21
Site: Load File
lllllllllllllllllilllMlil
Api-01 Apr18 May 05 May 23 Jun 09 Jun 26 Jul 12 Jul 27 Aug 12 Aug 29 Sep 15 Oct 02 Oc(19
Date
Figure 4B-26. Time series of 8-hr daily maximum Os concentrations at Washington D.C.
monitoring sites for April-October 2007. Observed values shown in black and modeled
values shown in red.
4B-25
-------
2007ee ORDBC v507cv47112US2 O3 for AQS Hourly Site: Load File
CJ
O
40 -
30 -
20 -
10 -
0 -
AQS Hourly
2007ee ORDBC v5 07c v471 12US2
# of Sites: 8
Site: Load File
Jan 01 Jan 03 Jan 05 Jan 08 Jan 10 Jan 12 Jan 15 Jan 17 Jan 20 Jan 22 Jan 24 Jan 27 Jan 29 Jan 31
Date
Figure 4B-27. Time series of hourly Os concentrations at Washington D.C. monitoring
sites for January 2007. Observed values shown in black and modeled values shown in red.
2007ee ORDBC v5 07c V471 12US2 O3 for AQS Hourly Site: Load File
80 -
60 -
g 40 H
20 -
0 -
AQS Hourly
2007ee ORDBC v5 07c v471 12US2
# of Sites: 20
Site: Load_File
Apr 01 Apr 03 Apr 05 Apr 07 Apr 10 Apr 12 Apr 14 Apr 17 Apr 19 Apr 21 Apr 23 Apr 26 Apr 28 Apr 30
Date
Figure 4B-28. Time series of hourly Os concentrations at Washington D.C. monitoring
sites for April 2007. Observed values shown in black and modeled values shown in red.
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_Flle
120 -
100 -
80 -
CL
60 -
40 -
20 -
0 -
CO
AQS Hourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: 21
Site: Load File
JulOI Jul03 JUI05 JUI07 JUI09 Jul 11 Jul 14 Jul 16 Jul 18 Jul 20 Jul 22 Jul 24 Jul 27 Jul 29 Jul 31
Date
Figure 4B-29. Time series of hourly Os concentrations at Washington D.C. monitoring
sites for July 2007. Observed values shown in black and modeled values shown in red.
4B-26
-------
2007ee ORDBC v5 07c v471 12US2 O3 for AQS Hourly Site: Load File
100 -
I "
0 40 -I
AQS Hourly
200/ee ORDBC V5 07c V471 12US2
# of Sites: 21
Site: Load File
OctOI Oct03 Oct 05 OctOS OctIO Oct 12 Oct 15 Oct17 Ocl 20 Oct22 Oct 24 Oct 27 Oct 29 Oct 31
Date
Figure 4B-30. Time series of hourly Os concentrations at Washington D.C. monitoring
sites for October 2007. Observed values shown in black and modeled values shown in red.
4B-2.2 Operational Evaluation in the Southeast U.S.
Model performance in the Southeastern U.S. was in the range of reported performance
for state of the science models (Simon et al., 2012). Mean bias for MDA8 Os was less than 5
ppb at most sites in the winter and spring and less than 10 ppb at most sites in the summer and
fall seasons. Normalized mean bias was generally less than 10% at most sites in the winter and
spring and less than 25% at most sites in the summer and fall. Also, there were very few days
above 60 ppb in the winter season (4 days at two sites and 0,1, or 2 days at all other sites), but
the model generally had trouble capturing Os on those days and had mean biases in the range of -
5 to -15 ppb on those days. The higher biases in the summer and fall are most pronounced along
the Gulf coast and at sites in Florida. Atlanta was the only one of the 15 urban study areas from
the HREA which was located in the Southeast region.
There were no Os measurements in the Atlanta area during the winter season. Mean bias
and normalized mean bias at Atlanta sites for the spring, summer, and fall months were typical of
performance throughout the Southeast region. The April to June MDA8 Os time series (Figure
4B-35) shows that the model does a good job of capturing the variability between high and low
Os days. The hourly time series plots for April (Figure 4B-36), July (Figure 4B-37), and October
(Figure 4B-38) show reasonable model performance during daytime hours but some persistent
overestimates of nighttime Os, especially in April and July.
4B-27
-------
Table 4B-10. Summary of CMAQ model performance at AQS monitoring sites in the
Southeastern U.S.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
6378
53
6431
13326
5190
18516
14129
6336
20465
14105
1743
15848
MB (ppb)
0.13
-10.40
0.04
1.10
-4.94
-0.60
10.3
0.3
7.20
5.43
-1.44
4.67
NMB (%)
0.4
-16.1
0.1
2.3
-7.4
-1.1
23.2
0.5
13.8
13.9
-2.1
11.1
ME (ppb)
5.32
11.10
5.37
5.22
6.45
5.57
11.2
6.7
9.82
7.67
6.40
7.53
NME (%)
15.7
17.2
15.7
10.8
9.6
10.4
25.3
9.6
18.8
19.7
9.5
17.9
O3_8hrmax NMB (%) for run 2007ee_ORDBC_v5_07c_v471_12US2 for Winter for VISTAS [No Cutoff]
units = %
coverage limit = 75%
A*
y"
r* A
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-31. Map of normalized mean bias for MDA8 Os concentrations in the
Southeastern U.S. for winter months in 2007.
4B-28
-------
03_8hrmax NMB (%) for run 2007ee_ORDBC_v5_07c_v471_12US2 for Spring tor VISTAS [No Cutoft]
units = %
coverage limit = 75%
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-32. Map of normalized mean bias for MDA8 Os concentrations in the
Southeastern U.S. for spring months in 2007.
4B-29
-------
O3_8hrmax NMB (%) for run 2007ee_ORDBC_v5_07c_v471_12US2 for Summer for VISTAS [No Cutoff]
units = %
coverage limit = 75%
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-33. Map of normalized mean bias for MDA8 Os concentrations in the
Southeastern U.S. for summer months in 2007.
4B-30
-------
O3_8hrmax NMB (%) for run 2007ee_ORDBC_v5_07c_v471_12US2 for Fall for VISTAS [No Cutoff]
units = %
coverage limit = 75%
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-34. Map of normalized mean bias for MDA8 Os concentrations in the
Southeastern U.S. for fall months in 2007.
4B-31
-------
Table 4B-11. Summary of CMAQ model performance at AQS monitoring sites in the
Atlanta area.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
0
0
0
703
295
998
510
467
977
575
87
662
MB (ppb)
NA
NA
NA
0.52
-3.08
-0.54
10.00
3.77
7.04
5.45
4.01
5.26
NMB (%)
NA
NA
NA
1.1
-4.6
-1.0
21.7
5.2
11.9
13.7
5.9
12.1
ME (ppb)
NA
NA
NA
5.99
6.03
6.00
11.50
8.46
10.00
7.72
6.27
7.53
NME (%)
NA
NA
NA
12.5
9.0
11.2
24.8
11.5
17.0
19.4
9.2
17.3
2007ee_ORDBC_v5_07c_v471_12US2 O3_8hrmax for AQS_Dally Site: Load_Flle
120 -
S' 100 -
CL
CL
^ 80 -
h»
a 60 -
8
40 -
20 -
AQS_Daily
2007ee ORDBC v5 07c v471 12US2
# of Sites: 11
Site: Load File
lllllllllllllllllilllMlil
Api-01 Apr18 May 05 May 23 Jun 09 Jun 26 Jul 12 Jul 27 Aug 12 Aug 29 Sep15 Oct 02 Oc(19
Date
Figure 4B-35. Time series of 8-hr daily maximum Os concentrations at Atlanta monitoring
sites for April-October 2007. Observed values shown in black and modeled values shown
in red.
4B-32
-------
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_Flle
100 -
80 -
CL 60 -
in
40-
20 -
0 -
AQS Hourly
2007ee ORDBC v5 07c V471 12US2
* of Sites: 11
Site: Load File
Apr 01 Apr 03 Apr 05 Apr 07 Apr 10 Apr 12 Apr 14 Apr 17 Apr 19 Apr 21 Apr 23 Apr 26 Apr 28 Apr 30
Dale
Figure 4B-36. Time series of hourly Os concentrations at Atlanta monitoring sites for April
2007. Observed values shown in black and modeled values shown in red.
2007ee ORDBC v5 07c v471 12US2 O3 for AQS Hourly Site: Load File
120 -
100 -
_ ao -
! ,-
8
40 -
20 -
0 -
AQS Hourly
2007ee ORDBC v5 07c v471 12US2
# of Sites: 11
Site: Load_File
Jul 01 Jul 03 Jul 05 Jul 07 Jill 09 Jul11 JuM4 Jul 16 Jul 18 Jul 20 Jul 22 Jul 24 Jul 27 Jul 29 Jul 31
Date
Figure 4B-37. Time series of hourly Os concentrations at Atlanta monitoring sites for July
2007. Observed values shown in black and modeled values shown in red.
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load File
100 -
80 -
!o 60 -
Q.
a.
O 40 -
20 -
0 -
AQS Hourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: 11
Site: Load File
OctOI Oct03 Oct 05 Oct 08 OctlO Oct 12 Oct 15 Oct 17 Oct 20 Oct 22 Oct 24 Oct 27 Oct 29 Oct 31
Date
Figure 4B-38. Time series of hourly Os concentrations at Atlanta monitoring sites for
October 2007. Observed values shown in black and modeled values shown in red.
4B-33
-------
4B-2.3 Operational Evaluation in the Midwest
The model performed well compared to observed Os concentrations in the Midwest.
Mean bias for MDA8 Cb was around 5 ppb or less at most sites in the winter, spring, and fall and
less than 7 ppb at most sites in the summer. Normalized mean bias for MDA8 Os was less than
15% at most sites except in the winter when it was somewhat higher (less than 20% at most sites)
due to lower observed Os concentrations. The model was more likely to be biased high on low
Os days and biased low on higher Cb days. No distinct spatial trends are apparent from the maps
of normalized mean bias (Figure 4B-39 through Figure 4B-42). Three urban areas in the
Midwest were examined more closely for this evaluation: Chicago, Cleveland, and Detroit.
Chicago performance statistics for MDA8 Os were similar to those of the rest of the
region. The time series of MDA8 Os from Apr-Oct 2007 (Figure 4B-43) as well as the hourly
time series for January, April, July, and October (Figure 4B-44 through Figure 4B-47) all show
that the model does a good job of capturing synoptic variations in Os observed in Chicago and
reasonably captures both day and nighttime measured concentrations.
No measurements were made during the winter season in Cleveland. MB for MDA8 Os
on high days (> 60 ppb) was similar to MB in the rest of the region, but MB on low Os days was
somewhat higher than was typically seen in the Midwest (4 ppb vs. 1 ppb in the spring, 11 ppb
vs. 6 ppb in the summer, and 4 ppb vs. 3 ppb in the fall). The time series plots for Cleveland
sites show good correlation with observed Os, although a period of moderate model over-
estimates in daytime Os is depicted in July (Figure 4B-48 and Figure 4B-50). Nighttime Os
concentrations are often overestimated by the model except in the time series shown for April,
July, and October the latter portion of October, 2007.
Detroit area sites did not report any Os measurements during the winter of 2007. Detroit
performance statistics for MDA8 Os were similar to those from the rest of the Midwest; however
under-estimates on high Os days were more pronounced in Detroit than in the rest of the region.
The time series shows that the model accurately estimates both day and nighttime hourly Os in
Detroit in April and July and generally captures the variations in MDA8 Os across the April-
October time period.
4B-34
-------
Table 4B-12.
Midwest.
Summary of CMAQ model performance at AQS monitoring sites in the
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
2824
0
2824
9447
2403
11850
12478
4114
16592
8207
1382
9589
MB (ppb)
-3.76
NA
-3.76
1.11
-6.19
-0.37
6.31
-1.19
4.45
2.94
-4.48
1.87
NMB (%)
-15.1
NA
-15.1
2.6
-8.9
-0.8
14.2
-1.7
8.8
8.2
-6.5
4.6
ME (ppb)
5.36
NA
5.36
5.55
7.38
5.92
8.39
7.56
8.18
6.22
7.72
6.44
NME (%)
21.5
NA
21.5
12.8
10.6
12.2
18.8
10.8
16.1
17.3
11.1
15.8
O3_8hrmax NMB {%) for run 2007ee_ORDBC_v5_07c_v471_12US2 for Winter for MWRPO [No Cutoff]
units = %
coverage limit = 75%
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-39. Map of normalized mean bias for MDA8 Os concentrations in the Midwest
for winter months in 2007.
4B-35
-------
O3_8hrmax NMB (%) for run 2007ee_ORDBC_v5_07c_v471_12US2 tor Spring for MWRPO [No Cutoff]
units = %
coverage limit» 75%
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-40. Map of normalized mean bias for MDA8 Os concentrations in the Midwest
for spring months in 2007.
4B-36
-------
O3_8hrmax NMB (%) for run 2007ee_ORDBC_v5_07c_v471_12US2 for Summer for MWRPO [No Cutoff]
units = %
coverage limit» 75%
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-41. Map of normalized mean bias for MDA8 Os concentrations in the Midwest
for summer months in 2007.
4B-37
-------
O3_8hrmax NMB (%) tor run 2007ee_ORDBC_v5_07c_v471_12US2 for Fall for MWRPO [No Cutoff]
units = %
coverage limit» 75%
^ r -
•J /* •
.1 / • *
\a L— —•* 1 <
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-42. Map of normalized mean bias for MDA8 Os concentrations in the Midwest
for fall months in 2007.
4B-38
-------
Table 4B-13. Summary of CMAQ model performance at AQS monitoring sites in the
Chicago area.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
963
0
963
1510
218
1728
1700
448
2148
1375
136
1511
MB (ppb)
-3.93
NA
-3.93
0.70
-5.55
-0.09
8.10
1.70
6.76
1.93
-4.61
1.34
NMB (%)
-17.5
NA
-17.5
1.7
-8.1
-0.2
18.8
2.5
13.9
6.1
-6.7
3.8
ME (ppb)
5.13
NA
5.13
6.02
6.62
6.1
10.0
8.65
9.74
5.62
9.05
5.93
NME (%)
22.9
NA
22.9
15.1
9.7
14.0
23.2
12.5
20.0
17.8
13.2
17.0
2007ee_ORDBC_v5_07c_v471_12US2 O3_8hrmax for AQS_Dally Site: Load_Flle
120 -
100 -
80 -
40 -
20 -
AQS Daily
2007ee ORDBC v5 07c V471 12US2
# of Sites: 24
Site: Load File
Apr 01 Apr 18 May 05 May 23 Jun 09 Jun 26 Jul 12 Jul 27 Aug 12 Aug 29 Sep15 Oct 02 Oct19
Date
Figure 4B-43. Time series of 8-hr daily maximum Os concentrations at Chicago
monitoring sites for April-October 2007. Observed values shown in black and modeled
values shown in red.
4B-39
-------
2007ee ORDBC v5 07c v471 12US2 O3 for AQS Hourly Site: Load File
40 -
30 -
~ 20 H
o
10 -
o -
AQS Hourly
200/ee ORDBC V5 07c V471 12US2
# of Sites: 10
Site: Load File
Jan 01 Jan03 Jan 05 Jan 08 JanlO Jan 12 Jan 15 Jan17 Jan 20 Jan 22 Jan 24 Jan 27 Jan 29 Jan 31
Date
Figure 4B-44. Time series of hourly Os concentrations at Chicago monitoring sites for
January 2007. Observed values shown in black and modeled values shown in red.
2007ee ORDBC v5 07c v471 12US2 O3 for AQS Hourly Site: Load File
80 -
— 60 -
|
g 40 H
20 -
0 -
AQS Hourly
2007ee ORDBC v5 07c v471 12US2
# of Sites: 24
Site: Load File
Apr 01 Apr 03 Apr 05 Apr 07 Apr 10 Apr 12 Apr 14 Apr 17 Apr 19 Apr 21 Apr 23 Apr 26 Apr 28 Apr 30
Date
Figure 4B-45. Time series of hourly Os concentrations at Chicago monitoring sites for
April 2007. Observed values shown in black and modeled values shown in red.
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_File
AQS Hourly
2007ee ORDBC_v5_07c v471_12US2
9 of Sites: 24
Site: Load File
JulOI JUI03 JUI05 JUI07 JUI09 Jul11 Jul 14 Jul 16 Jul 18 Jul 20 Jul 22 Jul 24 Jul 27 Jul 29 Jul 31
Date
Figure 4B-46. Time series of hourly Os concentrations at Chicago monitoring sites for July
2007. Observed values shown in black and modeled values shown in red.
4B-40
-------
60 -
20 -
0 -
2007ee ORDBC v507cv47112US2 O3 for AQS Hourly Site: Load File
AQS Hourly
2007ee ORDBC v5 07c v471 12US2
# of Sites: 17
Site: Load File
OctOI Oct03 OctOS OctOB OctIO Oct 12 Oct 15 Oct17 Oct 20 Oct 22 Oct 24 Ocl27 Oct 29 Oet31
Date
Figure 4B-47. Time series of hourly Os concentrations at Chicago monitoring sites for
October 2007. Observed values shown in black and modeled values shown in red.
Table 4B-14. Summary of CMAQ model performance at AQS monitoring sites in the
Cleveland area.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
0
0
0
491
113
604
714
187
901
543
60
603
MB (ppb)
NA
NA
NA
4.35
-4.80
2.64
11.00
2.25
9.16
4.41
-4.48
3.52
NMB (%)
NA
NA
NA
10.5
-6.7
5,8
25.8
3.2
19.0
11.6
-6.5
8.6
ME (ppb)
NA
NA
NA
7.03
8.46
7.30
12.90
9.48
12.20
7.38
8.97
7.54
NME (%)
NA
NA
NA
17.0
11.7
15.5
30.3
13.6
25.3
19.4
13.1
18.4
4B-41
-------
2007ee ORDBC V507CV47112US2 O3 Bhrmax for AQS Daily Site: Load File
100 -
I 80 -
x
9
E 60 -
5
O 40 -
20 -
AQS Daily
2007ee ORDBC v5 07c v471 12US2
#a\ Sites: 10
Site: Load File
niiiiiiiimiiiiiimmm
Apr 01 Apr 18 May 05 May 23 Jun 09 Jun 26 Jul 12 Jul 27 Aug 12 Aug 29 Sep15 Oct 02 Oct19
Date
Figure 4B-48. Time series of 8-hr daily maximum Os concentrations at Cleveland
monitoring sites for April-October 2007. Observed values shown in black and modeled
values shown in red.
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_Flle
80 -
60 -
8 40 H
20 -
AQS Hourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: 10
Site: Load File
Apr 01 Apr 03 Apr 05 Apr 07 Apr 10 Apr 12 Apr 14 Apr 17 Apr 19 Apr 21 Apr 23 Apr 26 Apr 28 Apr 30
Date
Figure 4B-49. Time series of hourly Os concentrations at Cleveland monitoring sites for
April 2007. Observed values shown in black and modeled values shown in red.
2007ee_ORDBC v5 07c V471 12US2 O3 for AQS Hourly Site: Load_Flle
100
8
80 -
60 -
20 -
0
AQS Hourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: 10
Site: Load File
Jul 01 Jul 03 Jul 05 Jul 07 Jul 09 JuM1 Jul 14 Jul 16 Jul 18 Jul 20 Jul 22 Jul 24 Jul 27 Jul 29 Jul 31
Date
Figure 4B-50. Time series of hourly Os concentrations at Cleveland monitoring sites for
July 2007. Observed values shown in black and modeled values shown in red.
4B-42
-------
2007ee ORDBC v507cv47112US2 O3 for AQS Hourly Site: Load File
80 -
_ 60 -
« 40 -
O
20 -
0 -4
AQS Hourly
2007ee ORDBC v5 07c v471 12US2
# of Sites: 10
Site: Load File
OctOI Oct03 OctOS OctOB OctIO Oct 12 Oct 15 Oct17 Oct 20 Oct 22 Oct 24 Ocl27 Oct 29 Oet31
Date
Figure 4B-51. Time series of hourly Os concentrations at Cleveland monitoring sites for
October 2007. Observed values shown in black and modeled values shown in red.
Table 4B-15: Summary of CMAQ model performance at AQS monitoring sites in the
Detroit area.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
0
0
0
443
90
533
614
191
805
220
46
166
MB (ppb)
NA
NA
NA
-1.04
-8.61
-2.32
1.65
-7.16
-0.44
-1.01
-14.70
-3.38
NMB (%)
NA
NA
NA
-2.5
-11.7
-4.9
3,9
-9.9
-0.8
-2.6
-21.2
-7.7
ME (ppb)
NA
NA
NA
4.93
9.15
5.64
7.19
10.9
8.07
6.65
15.6
8.2
NME (%)
NA
MA
NA
11.7
12.4
11.9
16.7
15.1
16.2
17.1
22.6
18.6
4B-43
-------
2007ee_ORDBC_v5_07c_v471_12US2 O3_8hrmax for AQS_Dally Site: Load_Flle
100 -
"I 80 -
!= 60 -
!
O 40 -
20 -
AQS_Daily
2007ee ORDBC v5 07c v471 12US2
# of Sites: 9
Site: Load File
Apr 01 Apr 15 Apr 29 May 14 May 29 Jun 13 Jun 27 Julll Jul 24 Aug 07 Aug 22 Sep 06 Sep 21
Date
Figure 4B-52. Time series of 8-hr daily maximum Os concentrations at Detroit monitoring
sites for April-October 2007. Observed values shown in black and modeled values shown
in red.
2007ee ORDBC v5 07c v471 12US2 O3 for AQS Hourly Site: Load File
O
80 -
60 -
20 -
AQS_Hourly
2007ee ORDBC v5 07c v471 12US2
tt of Sites: 9
Sile: Lond File
AprOI Apr03 Apr05 Apr 07 ApMO Apr12 Apr14 Apr17 Apr 19 Apr21 Apr23 Apr 26 Apr 28 Apr30
Date
Figure 4B-53. Time series of hourly Os concentrations at Detroit monitoring sites for April
2007. Observed values shown in black and modeled values shown in red.
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_File
100 -
80 -
a. 60 -
8 40-
20 -
0 -
AQS Hourly
2007ee_ORDBC_v5_07c_v471 _12US2
# of Sites: 9
Site: Load File
Jul 01 Jul 03 Jul 05 Jul 07 Jul 09 Jul 11 Jul 14 Jul 16 Jul 18 Jul 20 Jul 22 Jul 24 Jul 27 Jul 29 Jul 31
Date
Figure 4B-54. Time series of hourly Os concentrations at Detroit monitoring sites for July
2007. Observed values shown in black and modeled values shown in red.
4B-44
-------
4B-2.4 Operational Evaluation in the Central U.S.
Mean bias for MDA8 Cb concentrations were in the range of 4 ppb or less at Central U.S.
monitoring sites during the winter and spring and less than 13 ppb and 6 ppb in the summer and
fall respectively. Normalized mean bias was less than 15% at most sites in winter, spring, and
fall and less than 35% at most sites in the summer. Summertime model overestimates were
primarily located along the Gulf coast of Texas and Louisiana, mostly occurring on days with
MDA8 Os concentrations less than 60 ppb. This is similar to summertime overestimates
reported for the Southern U.S. which also were more pronounced along the Gulf Coast.
Houston, Dallas, and Saint Louis were the three cities from the 15 urban study areas which were
located in the Central U.S. region.
Saint Louis mean bias for MDA8 was moderately better than mean bias in the rest of the
region for winter and spring periods and substantially better than the mean bias in the rest of the
Central region in the summer and fall. The time series plots show that the model does a good job
of replicating the observed variations in Os (both MDA8 and hourly). Nighttime Os is generally
well simulated with the exception of a few several day periods in July when nighttime
concentrations were overestimated by the model.
Dallas performance statistics for MDA8 Os were very similar to those presented for the
Central region as a whole with the exception of comparisons made for the fall season which
showed that Dallas performed substantially better than average for the region. The time series
plot of MDA8 Os modeled and measured values (Figure 4B-64) demonstrate consistent high bias
on very low Os days in May through early September but not in April or October. Hourly Os is
fairly well replicated in January (Figure 4B-65), April (Figure 4B-66), and October (Figure 4B-
68) but is overestimated during July both on low Os days and at night (Figure 4B-67).
As mentioned above, the largest summertime overestimates in MDA8 Os in the Central
U.S. occur along the Texas Gulf coast. This is demonstrated by the Houston model performance
statistics. Houston model performance is relatively good and is equivalent to that in the rest of
the Central U.S. during the winter, spring, and fall seasons and on high days during the summer.
However summertime MDA8 Os is overestimated by 15 ppb and 48% on summer days with
observed levels less than 60 ppb in the Houston area. This summertime overestimate of low Os
days is demonstrated in the time series plots in Figure 4B-69 and Figure 4B-72.
4B-45
-------
Table 4B-16.
central U.S.
Summary of CMAQ model performance at AQS monitoring sites in the
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
11190
33
11223
13077
2137
15214
14142
2477
16619
13209
1359
14568
MB (ppb)
-0.42
-13.9
-0.46
2.88
-6.68
1.54
11.4
0.60
9.82
4.39
-3.33
3.67
NMB (%)
-1.4
-21.5
-1.5
6.7
-10.0
3.3
29.7
0.9
22.8
11.8
-4.9
9.2
ME (ppb)
5.61
14.5
5.64
6.35
8.26
6.62
12.6
111
11.9
7.54
7.04
7.49
NME (%)
18.7
22.4
18.7
14.7
12.4
14.2
32.8
11.1
27.5
20.3
10.4
18.7
O3_8hrmax NMB (%) for run 2007ee_ORDBC_v5_07c_v471_12US2 for Winter for CENRAP [No Cutoff]
•f raje limit . 75%
>50
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-55. Map of normalized mean bias for MDA8 Os concentrations in the Central
U.S. for winter months in 2007.
4B-46
-------
O3_8hrmax NMB (%) for run 2007ee_ORDBC_v5_07c_v471_12US2 for Spring for CENRAP [No Cutoff]
« s
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-56. Map of normalized mean bias for MDA8 Os concentrations in the Central
U.S. for spring months in 2007.
4B-47
-------
O3_8hrmax NMB (%) for run 2007ee_ORDBC_v5_07c_v471_12US2 for Summer for CENRAP [No Cutoff]
- .
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-57. Map of normalized mean bias for MDA8 Os concentrations in the Central
U.S. for summer months in 2007.
4B-48
-------
O3_8hrmax NMB (%) for run 2007ee_ORDBC_v5_07c_v471_12US2 for Fall for CENRAP [No Cutoff]
•
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-58. Map of normalized mean bias for MDA8 Os concentrations in the Central
U.S. for fall months in 2007.
4B-49
-------
Table 4B-17. Summary of CMAQ model performance at AQS monitoring sites in the Saint
Louis area.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
266
0
266
812
139
951
834
447
1281
829
113
942
MB (ppb)
-0.16
NA
-0.16
1.09
-6.93
-0.08
8.71
0.66
5.90
2.38
-6.19
1.35
NMB (%)
-0.7
NA
-0.7
2.5
-10.6
-0.2
18.9
0.9
10.6
6.6
-8.8
3.4
ME (ppb)
4.10
NA
4.10
5.31
7.85
5.68
9.83
9.34
9.66
6.34
8.46
6.60
NME (%)
18.1
NA
18.1
11.9
12.0
11.9
21.4
12.7
17.4
17.6
12.1
16.4
2007ee ORDBC_v5_07c_v471_12US2 O3 Bhrmax lor AQS Daily Site: Load_File
120 -
100 -
| 80-
X
I 60 -
0 40 H
20 -
AQS Daily
2007ee ORDBC v5 07c v471 12US2
# of Sites: 14
Site: Load File
iDiiiiiiiini iiniiiiini uuimiiii iiiiiiiMiiiiiiiii miiiimiiiiniir [iiniiiiiiii
AprOI Apr18 May 05 May 23 Jun 09 Jun 26 Jul 12 Jul 27 Aug 12 Aug 29 Sep15 Oct 02 Oct19
Date
Figure 4B-59. Time series of 8-hr daily maximum Os concentrations at Saint Louis
monitoring sites for April-October 2007. Observed values shown in black and modeled
values shown in red.
4B-50
-------
2007ee ORDBC v5 07c v471 12US2 O3 for AQS Hourly Site: Load File
»
o
30 -
10 -
o -
AQS Hourly
200/ee ORDBC V5 07c V471 12US2
# of Sites: 3
Site: Load File
Jan 01 Jan03 Jan 05 Jan 08 JanlO Jan 12 Jan 15 Jan17 Jan 20 Jan 22 Jan 24 Jan 27 Jan 29 Jan 31
Date
Figure 4B-60. Time series of hourly Os concentrations at Saint Louis monitoring sites for
January 2007. Observed values shown in black and modeled values shown in red.
2007ee ORDBC v5 07c v471 12US2 O3 for AQS Hourly Site: Load File
.0
80 -
60 ~
40 H
20 -
AQS Hourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: 14
Site: Load File
Apr 01 Apr 03 Apr 05 Apr 07 Apr 10 Apr 12 Apr 14 Apr 17 Apr 19 Apr 21 Apr 23 Apr 26 Apr 28 Apr 30
Date
Figure 4B-61. Time series of hourly Os concentrations at Saint Louis monitoring sites for
April 2007. Observed values shown in black and modeled values shown in red.
2007ee_ORDBC_v5_07c_v471 12US2 O3 for AQS Hourly Site: Load_File
120 -
100 -
m
O
60 -
40 -
20 -
AQS^ Hourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: 14
Site: Load File
Jul 01 Jul 03 Jul 05 Jul 07 JulO9 JuM1 JuM4 Jul 16 Jul18 Jul 20 Jul 22 Jul 24 Jul 27 Jill 29 Jul 31
Date
Figure 4B-62. Time series of hourly Os concentrations at Saint Louis monitoring sites for
July 2007. Observed values shown in black and modeled values shown in red.
4B-51
-------
2007ee ORDBC v5 07c v471 12US2 O3 for AQS Hourly Site: Load File
80 -
60 -
20 -
0 -\
AQS Hourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: 14
Site: Load File
OctOI Ocl03 Ocl 05 Oct 08 OctIO Oct 12 Oct 15 Oct 17 Oct 20 Oct 22 Oct 24 Ocl 27 Ocl 29 Oct 31
Date
Figure 4B-63. Time series of hourly Os concentrations at Saint Louis monitoring sites for
October 2007. Observed values shown in black and modeled values shown in red.
Table 4B-18. Summary of CMAQ model performance at AQS monitoring sites in the
Dallas area.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
1483
0
1483
1360
168
1528
1289
260
1549
1306
223
1529
MB (ppb)
-2.12
NA
-2.12
2.92
-8.10
1.71
11.10
3.35
9.77
1.67
-2.77
1.02
NMB (%)
-7.4
NA
-7.4
7.0
-12.3
3.8
28.3
4.7
21.9
4,3
-4.1
2.4
ME (ppb)
5.30
NA
5.30
6.77
9.38
7.05
12.30
8.59
11.70
6.89
6.69
6.86
NME (%)
18.6
NA
18.6
16.2
14.2
15.8
31.5
12.0
26.3
17.8
9.8
15.9
4B-52
-------
2007ee ORDBC v507cv47112US2 O3 Bhrmax for AQS Daily Site: Load File
120
100 -
I 80 H
40 -
20 ->
AQS Daily
2007ee ORDBC v5 07c v471 12US2
# of Sites: 17
Site: Load File
iiiiiiimimiiiiiiiiiimiiillllllill
AprOI Apr18 May 05 May23 Jun 09 Jun 26 Jul 12 Jul 27 Aug 12 Aug 29 Sep15 Oct 02 Oct19
Date
Figure 4B-64. Time series of 8-hr daily maximum Os concentrations at Dallas monitoring
sites for April-October 2007. Observed values shown in black and modeled values shown
in red.
2007ee ORDBC v5 07c v471 12US2 O3 for AQS Hourly Site: Load File
ft
60 -
50 -
40 -
30 -
20 -
10 -
0
AQS_Hourly
2007ee_ORDBC_v5_07c_v471_1 2US2
* of Sites: 17
Site: Load File
Jan 01 Jan 03 Jan 05 Jan 08 Jan 10 Jan 12 Jan 15 Jan 17 Jan 20 Jan 22 Jan 24 Jan 27 Jan 29 Jan 31
Date
Figure 4B-65. Time series of hourly Os concentrations at Dallas monitoring sites for
January 2007. Observed values shown in black and modeled values shown in red.
2007ee ORDBC v5 07c v471 12US2 O3 for AQS Hourly Site: Load File
80 -
_ 60 -
20 -
AQS Hourly
2007ee ORDBC v5 07c v471 12US2
# of Sites: 17
Site: Load_File
AprOI AprOS AprOS Apr 07 ApMO Apr 12 Apr14 Apr17 Apr 19 Apr 21 Apr23 Apr 26 Apr 28 Apr30
Date
Figure 4B-66. Time series of hourly Os concentrations at Dallas monitoring sites for April
2007. Observed values shown in black and modeled values shown in red.
4B-53
-------
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_Flle
100 -
ao -
60 -
40 -
20 -
0 -
AQS Hourly
2007ee ORDBC v5 07c V471 12US2
* of Sites: 17
Site: Load File
JulOI JUI03 Jul05 Jul 07 Jul09 Jul11 Jul 14 Jul 16 Jul 18 Jul 20 Jill 22 Jul 24 Jul 27 Jul 29 Jul 31
Dale
Figure 4B-67. Time series of hourly Os concentrations at Dallas monitoring sites for July
2007. Observed values shown in black and modeled values shown in red.
2007ee ORDBC v5 07c v471 12US2 O3 for AQS Hourly Site: Load File
100 -
80 -
I '
8 4Q
AQS Hourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: 17
Site: Load File
OctOI Oct03 Oct05 OctOB Oct 10 Oct 12 Oct15 Oct17 Oct 20 Oct 22 Oct 24 Oct 27 Oct29 Oct 31
Date
Figure 4B-68. Time series of hourly Os concentrations at Dallas monitoring sites for
October 2007. Observed values shown in black and modeled values shown in red.
4B-54
-------
Table 4B-19. Summary of CMAQ model performance at AQS monitoring sites in the
Houston area.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
1772
0
1772
1569
276
1845
1706
165
1871
1608
259
1867
MB (ppb)
0.48
NA
0.48
5.82
-5.23
4.17
15.00
1.71
13.80
5.41
-3.38
4.19
NMB (%)
1.8
NA
1.8
14.5
-7.7
9.4
48.4
2.5
40.2
14.9
-4.8
10.2
ME (ppb)
5.97
NA
5.97
9.05
10.40
9.25
16.8
11.00
16.30
8.73
9.05
8.78
NME (%)
22.2
NA
22.2
22.5
15.2
20.8
54.2
15.8
47.3
24.1
12.8
21.4
2007ee ORDBC v5 07c v471 12US2 O3 Shrmax lor AQS Dally Site: Load_Flle
120 -
100 -
| 80 -
X,
I 60 -
g(
8 40 -
20 -
AQS_.Daily
2007ee ORDBC v5 07c V471 12US2
# of Sites: 21
Site: Load File
Apr 01 Apr 18 May 05 May 23 Jun 09 Jun 26 Jul 12 Jul 27 Aug 12 Aug 29 Sep 15 Oct 02 Oct 19
Dale
Figure 4B-69. Time series of 8-hr daily maximum Os concentrations at Houston
monitoring sites for April-October 2007. Observed values shown in black and modeled
values shown in red.
4B-55
-------
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_File
AQS_Hourly
2007ee ORDBC v5 07c V471 12US2
tt of Sites: 20
Site: Load File
Jan 01 Jan03 Jan 05 Jan 08 JanIO Jan 12 Jan 15 Jan17 Jan 20 Jan 22 Jan 24 Jan 27 Jar 29 Jan 31
Date
Figure 4B-70. Time series of hourly Os concentrations at Houston monitoring sites for
January 2007. Observed values shown in black and modeled values shown in red.
2007ee ORDBC v5 07c V471 12US2 O3 for AQS Hourly Site: Load File
80 -
ft
S 40
20 -
Q A
AQS^Hourly
2007ee_ORDBC_v5_07c_v471 _12US2
# ol Sites: 21
Site: Load_File
AprOI Apr03 Apr 05 Apr 07 ApMO Apr 12 Apr 14 Apr 17 Apr 19 Apr 21 Apr23 Apr 26 Apr 28 Apr30
Dale
Figure 4B-71. Time series of hourly Os concentrations at Houston monitoring sites for
April 2007. Observed values shown in black and modeled values shown in red.
2007ee_ORDBC_v5_07c_v471 12US2 O3 for AQS_Hourly Site: Load File
100 -
80 -
eo -
40-
20 -
0 -
AQS Hourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: 21
Site: Load File
Jill 01 Jul03 Jul 05 Jul07 Jul09 JuM 1 Jul 14 Jul 16 Jul 18 Jul 20 Jul 22 Jul 24 Jul 27 Jul 29 Jul 31
Date
Figure 4B-72. Time series of hourly Os concentrations at Houston monitoring sites for July
2007. Observed values shown in black and modeled values shown in red.
4B-56
-------
2007ee ORDBC v507cv47112US2 O3 for AQS Hourly Site: Load File
100 -
80 -
"I 60 -
AQS Hourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: 21
Site; Load File
OctOI Oct03 Oct05 Oct 08 Oct 10 Oct 12 Oct 15 Oct 17 Oct 20 Oct 22 Oct 24 Oct 27 Oct 29 Oct 31
Date
Figure 4B-73. Time series of hourly Os concentrations at Houston monitoring sites for
October 2007. Observed values shown in black and modeled values shown in red.
4B-2.5 Operational Evaluation in the Western U.S.
Again model statistics for MDA8 Os in the Western U.S. are in the range of what has
been reported for state of the science model performance in the literature (Simon et al., 2012).
Mean bias at most sites was less than 6 ppb in the winter, spring, and fall and less than 10 ppb in
the summer. Normalized mean bias at most monitoring locations was less than 10% in the
winter and spring, less than 20% in the summer and less than 15% in the fall. Only 3 sites in the
West recorded Os concentrations equal to or above 60 ppb in winter (a Riverside California site
reported 13 days and the other two sites in Sacramento, CA and Sublette county WY reported
only 2 and 3 days above 60 ppb). These high wintertime observations were substantially
underestimated by the model with an average MB of-34.5 ppb but likely for different reasons.
The high days in Riverside California are probably due to traditionally understood Os formation
that occurs on warm sunny days. The high Os concentrations in Wyoming are an example of
wintertime Os formation that occurs during cold pool meteorology events which have substantial
snow cover and extreme temperature inversions and are still an active area of research. Some
spatial trends in normalized mean bias are apparent in the winter (Figure 4B-74) and in the
summer (Figure 4B-76). Wintertime Os is more likely to be overestimated on the West Coast
and more likely to be underestimated in the Intermountain West. Summertime model
overestimates are greatest along the Southern Coast of California in locations that generally have
low observed Os concentrations. Three urban study areas from the HREA are located in the
Western U.S. and are evaluated in this section: Denver, Sacramento, and Los Angeles.
Denver area model performance was generally comparable to model performance in the
rest of the Western U.S. although summertime overestimates of MDA8 were greater in Denver.
These summertime overestimates can be seen on many days in Figure 4B-78 and Figure 4B-80.
4B-57
-------
Figure 4B-79 shows that the model generally captures measured hourly Cb concentrations in the
Denver area on mid to high Os days but often overestimates very low Os days in the spring. Fall
hourly Os estimates from the model are reasonably well captured in Figure 4B-81.
Sacramento area model performance for MDA8 Os values was reasonably good in all
seasons except for the 2 days with high observed Os in the winter. The model underestimated
those high wintertime concentrations by 32 ppb. Other than those days, model bias was
generally less than 10% in the Sacramento area. The time series figures of MDA8 and hourly Os
concentrations show that the model does well at capturing the day to day and day to night Os
variations in all seasons.
Los Angeles generally had reasonable model performance with low normalized mean
bias number (0-8%) with two exceptions. The 13 high wintertime Os days measured in central
Riverside County were not captured by the model which had a mean bias for those days at that
site of-36 ppb and -45%. Also, the model tended to overestimate Os on summertime days with
observed concentrations below 60 ppb (15 ppb mean bias). These low summer Cb
concentrations generally occurred along the coast (Figure 4B-92). A map of summertime
normalized mean bias (Figure 4B-93) shows that the low Os sites are the locations with the
largest model bias. Monitors away from the coast generally had fairly low bias (0-20%) with the
exception of two sites in western Riverside County. The summer Os overestimates shown in the
MDA8 time series (Figure 4B-87) are therefore due to performance at those coastal sites. The
hourly time series for January (Figure 4B-88), April (Figure 4B-89), and October (Figure 4B-90)
generally show good hourly Os performance during the day but model overestimates at night.
Table 4B-20. Summary of CMAQ model performance at AQS monitoring sites in the
western U.S.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
23890
18
23908
22670
5101
27771
21098
9708
30806
26055
1691
27746
MB (ppb)
1.31
-34.5
1.29
0.61
-4.84
-0.40
7.50
1.29
5.74
3.40
-3.32
2.99
NMB (%)
4.1
-45.8
4.0
1.3
-7.2
-0.8
17.0
2.8
11.0
8.9
-4.9
7.4
ME (ppb)
5.79
34.5
5.81
5.61
7.30
5.92
9.96
8.94
9.64
7.08
9.81
7.25
NME (%)
17.9
45.8
17.9
12.1
10.9
11.8
22.6
12.8
18.5
18.4
14.4
18.0
4B-58
-------
O3_8hrmax NMB (%) tor run 2007ee_ORDBC_v5_07c_v471_12US2 tor Winter tor WRAP [No Cutoff]
coverage limit = 75%
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-74. Map of normalized mean bias for MDA8 Os concentrations in the Western
U.S. for winter months in 2007.
4B-59
-------
O3_8hrmax NMB (%) for run 2007ee_ORDBC_v5_07c_v471_12US2 for Spring for WRAP [No Cutoff]
coverage limit = 75%
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-75. Map of normalized mean bias for MDA8 Os concentrations in the Western
U.S. for spring months in 2007.
4B-60
-------
O3_8hrmax NMB (%) for run 2007ee_ORDBC_v5_07c_v471_12US2 for Summer for WRAP [No Cutoff]
coverage timit ~ 75%
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-76. Map of normalized mean bias for MDA8 Os concentrations in the Western
U.S. for summer months in 2007.
4B-61
-------
O3_8hrmax NMB (%) for run 2007ee_ORDBC_v5_07c_v471_12US2 for Fall lor WRAP [No Cutoff]
units - %
coverage limit =* 75%
CIRCLE=AQS_Daily;CIRCLE=CASTNET_Daily;
Figure 4B-77. Map of normalized mean bias for MDA8 Os concentrations in the Western
U.S. for fall months in 2007.
4B-62
-------
Table 4B-21. Summary of CMAQ model performance at AQS monitoring sites in the
Denver area.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
1006
0
1006
893
182
1075
427
653
1080
993
53
1046
MB (ppb)
-3.84
NA
-3.84
3.36
-2.42
2.38
10.2
4.94
7.02
3.19
-1.59
2.95
NMB (%)
-11.5
NA
-11.5
7.3
-3.7
4.9
19.4
7.2
11.3
8.4
-2.5
7.5
ME (ppb)
7.23
NA
7.23
6.24
6.06
6.21
11.5
8.18
9.49
6.47
6.38
6.46
NMB (%)
21.6
NA
21.6
13.7
9.4
12.7
21.9
11.9
15.2
17.0
9.8
16.4
100 -
S 80 -
X
60 -
;
o
40 -
20 -4
2007ee ORDBC V5_07c V471 12US2 O3_8hrmax for AQS Dally Site: Load File
AQS Daily
2007ee ORDBC v5 07c V471 12US2
# of Sites: 12
Site: Load File
IMIMIIIUIIIIIIIIMMI mil limilimHmMmiinillllFllllllllllllllllMMIIIMIimillllMllllllUIMIIIIIimiimillimilllll Mill M
AprOI AprlS May 05 May 23 Jun 09 Jun 26 Jul 12 Jul 27 Aug 12 Aug 29 Sep 15 Oct 02 Oct 19
Date
Figure 4B-78. Time series of 8-hr daily maximum Os concentrations at Denver monitoring
sites for April-October 2007. Observed values shown in black and modeled values shown
in red.
4B-63
-------
2007ee ORDBC v507cv47112US2 O3 for AQS Hourly Site: Load File
80 -
60 -
o 40 H
20 -
0
AQS Hourly
2007ee ORDBC v5 07c v471 12US2
# of Sites: 12
Sile: Lond File
Apr 01 Apr 03 Apr 05 Apr 07 Apr 10 Apr 12 Apr 14 Apr 17 Apr 19 Apr 21 Apr 23 Apr 26 Apr 28 Apr 30
Date
Figure 4B-79. Time series of hourly Os concentrations at Denver monitoring sites for April
2007. Observed values shown in black and modeled values shown in red.
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_Flle
I
120 -
100 -
80 -
8 60 H
40 -
20 -
AQS_Hourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: 12
SHE: Load File
Jill 01 Jul 03 JUI05 JUI07 Jul09 JuM1 Jul14 Jul 16 Jul 18 Jul 20 Jul 22 Jul 24 Jul 27 Jul 29 Jul 31
Date
Figure 4B-80. Time series of hourly Os concentrations at Denver monitoring sites for July
2007. Observed values shown in black and modeled values shown in red.
2007ee ORDBC v5 07c v471 12US2 O3 for AQS Hourly Site: Load File
60 -
20 -
0 -\
AQS_Hourly
2007ee ORDBC v5 07c v471 12US2
# of Sites: 12
Site: Load File
OctOI Oct03 Oct 05 Oct 08 OcMO Oct 12 Oct 15 Oct17 Oct 20 Oct22 Oct 24 Oct 27 Oct 29 Oct31
Date
Figure 4B-81. Time series of hourly Os concentrations at Denver monitoring sites for
October 2007. Observed values shown in black and modeled values shown in red.
4B-64
-------
Table 4B-22. Summary of CMAQ model performance at AQS monitoring sites in the
Sacramento area.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
1374
2
1376
1516
239
1755
1443
619
2062
1809
150
1959
MB (ppb)
1.04
-31.80
0.99
-0.89
-4.83
-1.42
4.10
-1.39
2.45
1.80
-7.82
1.06
NMB (%)
3.5
-49.8
3.4
-2.0
-7.1
-3.0
9.1
-2.0
4.7
4.8
-11.1
2.6
ME (ppb)
5.41
31.80
5.45
5.28
6.37
5.43
6.58
7.81
6.95
6.30
10.10
6.59
NME (%)
18.4
49.8
18.5
11.8
9.3
11.3
14.6
11.2
13.2
16.7
14.3
16.4
100 -
I
80 -
60 -
40 -
20
2007ee_ORDBC_v5_07c_v471_12US2 O3_8hrmax for AQS_Dally Site: Load_Flle
AQS_ Daily
2007ee ORDBC v5 07c V471 12US2
tt of Sites: 24
Site: Load_File
iimiiiimiiiiiiiimiiimiiiiiiiiiiiiiiiiiiiiiimiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiHiiiiiiiiiiiiiiiiiiiiiMiiiiiiim
Apr 01 Apr 18 May 05 May 23 Jun 09 Jun 26 Jul 12 Jul 27 Aug 12 Aug 29 Sep 15 Oct 02 Oct19
Dale
Figure 4B-82. Time series of 8-hr daily maximum Os concentrations at Sacramento
monitoring sites for April-October 2007. Observed values shown in black and modeled
values shown in red.
4B-65
-------
2007ee ORDBCv5 07c v471 12US2 O3 for AQS Hourly Site: Load File
50 -
40 -
20 -
10 -
AQS Hourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: 16
Site. Load_File
Jan 01 Jan 03 Jan 05 Jan OS Jan 10 Jan 12 Jan 15 Jan17 Jan 20 Jan 22 Jan 24 Jan 27 Jan 29 Jan 31
Dale
Figure 4B-83. Time series of hourly Os concentrations at Sacramento monitoring sites for
January 2007. Observed values shown in black and modeled values shown in red.
20Q7ee_ORDBC_v5_07c v471 12US2 O3 for AQS Hourly Site: Load_Flle
100 -
80 -
-g. 60 -
a.
-------
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_File
70 -
60 -
50 -
40 -
to
O 30 -
20 -
10 -
I
AQSJHourly
2007ee ORDBC v5 07c v471 12US2
# of Sites: 24
Site: Load File
OctOI Get 03 Ocl05 Oct 08 OctIO Oct 12 Oct15 Oct17 Oct 20 Oct 22 Oct 24 Oct 27 Oct 29 Oct 31
Date
Figure 4B-86. Time series of hourly Os concentrations at Sacramento monitoring sites for
October 2007. Observed values shown in black and modeled values shown in red.
Table 4B-23. Summary of CMAQ model performance at AQS monitoring sites in the Los
Angeles area.
Season
Winter
Spring
Summer
Fall
MDA8
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
< 60 ppb
> 60 ppb
All Days
No. of Days
3710
13
3723
2911
968
3879
1972
1896
3868
3461
335
3796
MB (ppb)
1.40
-35.60
1.27
1.55
-3.76
0.22
15.0
3.35
9.27
2.95
-2.66
2.46
NMB (%)
4.4
-44.9
4.0
3.3
-5.4
0.4
32.7
4.4
15.4
7.7
-3.8
5.9
ME (ppb)
5.39
35.6
5.49
6.12
7.20
6.39
16.2
10.4
13.3
8.30
10.60
8.51
NME (%)
16.9
44.9
17.2
13.0
10.4
12.2
35.3
13.7
22.1
21.5
15.1
20.5
4B-67
-------
2007ee ORDBC v507cv47112US2 O3 Bhrmax for AQS Daily Site: Load File
120 -
„ 100 -
I
* 80
Rt
«5 60
s
40
AQS Daily
2007ee ORDBC v5 07c v471 12US2
# of Sites: 44
Site: Load File
iiiiiiimimiiiiiMiiiiiiimiiiimi
AprOI Apr18 May 05 May23 Jun 09 Jun 26 Jul 12 Jul 27 Aug 12 Aug 29 Sep15 Oct 02 Oct19
Date
Figure 4B-87. Time series of 8-hr daily maximum Os concentrations at Los Angeles
monitoring sites for April-October 2007. Observed values shown in black and modeled
values shown in red.
2007ee_ORDBC_v5_07c_v471_12US2 O3 for AQS_Hourly Site: Load_Flle
50 -
40 -
_o
8: 30 H
CO
O
20 -
10 -
AQS Hourly
2007ee ORDBC v5 07c_v471 12US2
# of Sites: 43
Site: Load_File
4
u
Jan 01 Jan 03 Jan 05 Jan 08 Jan 10 Jan 12 Jan 15 Jan 17 Jan 20 Jan 22 Jan 24 Jan 27 Jan 29 Jan 31
Date
Figure 4B-88. Time series of hourly Os concentrations at Los Angeles monitoring sites for
January 2007. Observed values shown in black and modeled values shown in red.
2007ee ORDBC v5 07c v471 12US2 O3 for AQS Hourly Site: Load File
100 -
80 -
60 -
40-
20 -
AQS_Hourly
2007ee J3RDBC_vb_07c_v471 _12US2
# of Sites: 43
Site: Load File
Apr 01 Apr 03 Apr 05 Apr 07 Apr 10 Apr 12 Apr 14 Apr 17 Apr 19 Apr 21 Apr 23 Apr 26 Apr 28 Apr 30
Date
Figure 4B-89. Time series of hourly Os concentrations at Los Angeles monitoring sites for
April 2007. Observed values shown in black and modeled values shown in red.
4B-68
-------
2007ee ORDBC _v507c _v471 12US2 O3 for AQS Hourly Site: Load File
120 -
100 -
-R 80 -
K
g 60 H
40 -
20 -
AQS Hourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: 44
Site: Load File
JulOI Jul03 Jul05 Jul07 Jul 09 Jul 11 JuM4 Jul 16 Jul 18 Jul 20 Jul 22 Jul 24 Jul 27 Jul 29 Jul 31
Date
Figure 4B-90. Time series of hourly Os concentrations at Los Angeles monitoring sites for
July 2007. Observed values shown in black and modeled values shown in red.
2007ee ORDBC w5 07c V471 12US2 O3 for AQS Hourly Site: Load File
80 -
60 -
I
o
20 -
AQS_Hourly
2007ee ORDBC v5 07c V471 12US2
# of Sites: 44
Site: Load File
H
OctOt Oct03 OctOS OctOS OcflO Oct 12 Oct 15 Oct 17 Oct 20 Oct 22 Oct 24 Oct 27 Ocl 29 Oct 31
Date
Figure 4B-91. Time series of hourly Os concentrations at Los Angeles monitoring sites for
October 2007. Observed values shown in black and modeled values shown in red.
4B-69
-------
units - ppb
coverage limit - 75%
CIRCLE=AQS_Daily;
Figure 4B-92. Map of mean observed MDA8 Os concentrations at Los Angeles monitoring
sites for summer months (June, July, Aug) 2007.
units = %
coverage limit • 76%
CIRCLE=AQS_Daily;
Figure 4B-93. Map of normalized mean bias for MDA8 Os concentrations at Los Angeles
monitoring sites for summer months (June, July, Aug) 2007.
4B-70
-------
4B-3. REFERENCES
Akhtar, F., Henderson, B., Appel, W., Napelenok, S., Hutzell, B., Pye, H., Foley, K. (2012).
Multiyear Boundary Conditions for CMAQ 5.0 from GEOS-Chem with Secondary
Organic Aerosol Extensions, 11th annual Community Modeling and Analysis System
conference, Chapel Hill, NC, October 2012.
Carlton, A.G., Bhave, P.V., Napelenok, S.L., Edney, E.D., Sarwar, G., Finder, R.W., Pouliot,
G.A., Houyoux, M. (2010). Model Representation of Secondary Organic Aerosol in
CMAQv4.7. Environmental Science & Technology, 44: 8553-8560.
Foley, K.M., Roselle, S.J., Appel, K.W., Bhave, P.V., Pleim, I.E., Otte, T.L., Mathur, R., Sarwar,
G., Young, J.O., Gilliam, R.C., Nolte, C.G., Kelly, J.T., Gilliland, A.B., Bash, J.O.
(2010). Incremental testing of the Community Multiscale Air Quality (CMAQ) modeling
system version 4.7, Geoscientific Model Development, 3: 205-226.
Gery, M.W., Whitten, G.Z., Killus, J.P., Dodge, M.C. (1989). A photochemical kinetics
mechanism for urban and regional scale computer modeling. Journal of Geophysical
Research-Atmospheres, 94: 12925-12956.
Gilliam, R.C., Pleim, I.E. (2010). Performance Assessment of New Land Surface and Planetary
Boundary Layer Physics in the WRF-ARW. Journal of Applied Meteorology and
Climatology, 49: 760-774.
Gilliland, A.B., Hogrefe, C., Finder, R.W., Godowitch, J.M., Foley, K.L., Rao, S.T. (2008).
Dynamic evaluation of regional air quality models: assessing changes in O3 stemming
from changes in emission and meteorology. Atmospheric Environment, 42: 5110-5123.
Godowitch, J.M, Gilliam, R.C., Rao, S.T. (2011). Diagnostic evaluation of ozone production and
horizontal transport in a regional photochemical air quality modeling system.
Atmospheric Environment, 45: 3977-3987.
Godowitch, J.M., Pouliot, G.A., Rao, S.T. (2010). Assessing multi-year changes in modeled and
observed urban NOx concentrations from a dynamic model evaluation perspective.
Atmospheric Environment, 44: 2894-2901.
Godowitch, J.M., Hogrefe, C., Rao, S.T. (2008). Diagnostic analysis of a regional air quality
model: changes in modeled processes affecting ozone and chemical-transport indicators
form NOx point source emissions reductions. Journal of Geophysical Research-
Atmospheres, 113(D19):D19303,DOI: 10.1029/2007JD009537.
Grell, G. A., Dudhia, A. J., and Stauffer, D. R. (1994). A description of the Fifth-Generation
Penn State/NCAR Mesoscale Model (MM5). NCAR Technical Note NCAR/TN-
398+STR. Available at http://www.mmm.ucar.edu/mm5/docl.html.
Henderson, B.H., Akhtar, F., Pye, H.O.T., Napelenok, S.L., Hutzell, W.T. (2013) A database and
tool for boundary conditions for regional air quality modeling: description and
evaluation, Geoscientific Model Development Discussions, 6, 4665-4704.
Henze, O.K., J.H. Seinfeld, N.L. Ng, J.H. Kroll, T-M. Fu, DJ. Jacob, C.L. Heald (2008). Global
modeling of secondary organic aerosol formation from aromatic hydrocarbons: High- vs.
low-yield pathways. Atmos. Chem. Phys., 8: 2405-2420.
Lam, Y.F., Fu, J.S., Jacob, D.J., Jang, C., Dolwick, P., 2010 2006-2008 GEOS-Chem for CMAQ
Initial and Boundary Conditions. 9th Annual CMAS Conference, October 11-13, 2010,
Chapel Hill, NC.
4B-71
-------
Nenes, A., Pandis, S.N., Pilinis, C. (1998). ISORROPIA: A new thermodynamic equilibrium
model for multiphase multicomponent inorganic aerosols. Aquatic Geochemistry, 4: 123-
152.
Napelenok, S.L., Foley, K.M, Kang. D.W., Mathur, R., Pierce, T., Rao, S.T. (2011). Dynamic
evaluation of regional air quality model's response to emission reduction in the presence
of uncertain emission inventories. Atmospheric Environment, 45: 4091-4098.
Otte T.L., Pleim, I.E. (2010). The Meteorology-Chemistry Interface Processor (MCIP) for the
CMAQ modeling system: updates through v3.4.1. Geoscientific Model Development, 3:
243-256.
Simon, H., Baker, K.R., Phillips, S. (2012). Compilation and interpretation of photochemical
model performance statistics published between 2006 and 2012. Atmospheric
Environment, 61:124-139.
Skamarock, W.C., Klemp, J.B., Dudhia, 1, Gill, D.O., Barker, D.M., Dudia, M.G., Huang, X.,
Wang, W., Powers, J.G. (2008). A Description of the Advanced Research WRF Version
O
U.S. EPA (2011). Meteorological Model Performance for Annual 2007 Simulations, Office of
Air Quality Planning and Standards, Research Triangle Park, NC., 27711, EPA-454/R-
11-007.
U.S. EPA (2012a), Air Quality Modeling Technical Support Document for the Regulatory
Impact Analysis for the Revisions of the National Ambient Air Quality Standard for
Particulate Matter. Office of Air Quality Planning and Standards, Research Triangle
Park, NC, December 2012. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/data/201212aqm.pdf
U.S. EPA (2012b). Technical Support Document: Preparation of Emissions Inventories of the
Version 5, 2007-based Platform. Office of Air Quality Planning and Standards, Research
Triangle Park, NC. December 2012.
Yantosca, B. (2004). GEOS-CHEMv7-01-02 User's Guide, Atmospheric Chemistry Modeling
Group, Harvard University, Cambridge, MA, October 15, 2004.
Yarwood, G., Rao, S., Yocke, M., Whitten, G.Z. (2005). Updates to the Carbon Bond chemical
mechanism: CB05. Final Report to the US EPA, RT-0400675, December 8, 2005:
http://www.camx.com/publ/pdfs/CB05_Final_Report_120805.pdf.
Zhou, W., Cohan, D.S., Napelenok, S.L. (2013). Reconciling NOx emissions reductions and
ozone trends in the U.S., 2002-2006, Atmospheric Environment, 70: 236-244.
4B-72
-------
This page left intentionally blank
4B-73
-------
APPENDIX 4C
Air Quality Spatial Fields for the National Mortality Risk Burden
Assessment
Table of Contents
4C-1. OVERVIEW 4C-1
4C-2. AIR QUALITY SPATIAL FIELD TECHNIQUES 4C-2
4C-2.1 Voronoi Neighbor Averaging (VNA) 4C-2
4C-2.2 Community Multi-scale Air Quality (CMAQ) Model 4C-3
4C-2.3 Enhanced Voronoi Neighbor Averaging (eVNA) 4C-4
4C-2.4 Downscaler (DS) 4C-5
4C-3. EVALUATION OF AIR QUALITY SPATIAL FIELD TECHNIQUES 4C-7
4C-3.1 Data 4C-7
4C-3.2 Methods 4C-7
4C-3.3 Results 4C-9
4C-4. AIR QUALITY INPUTS TO THE NATIONAL-SCALE MORTALITY RISK
BURDENT ASSESSMENT 4C-12
4C-5. REFERENCES 4C-16
4C-i
-------
List of Tables
Table 4C-1. Summary of the cross-validation performance metrics for the four air quality
spatial field techniques 4C-12
List of Figures
Figure 4C-1. Numerical example of the Voronoi Neighbor Averaging (VNA) technique applied
to a model grid domain 4C-3
Figure 4C-2. Numerical example of the Enhanced Voronoi Neighbor Averaging (eVNA)
technique applied to a model grid domain 4C-5
Figure 4C-3. Example of the "4-fold" cross-validation scheme used in the evaluation of the air
quality spatial field techniques for the southern Lake Michigan area 4C-9
Figure 4C-4. Cross-validation results for the 2007 annual 4th highest daily maximum Cb
concentrations 4C-10
Figure 4C-5. Cross-validation results for the 2007 May-September mean of the daily maximum
8-hour Os concentrations 4C-11
Figure 4C-6. May-September average daily maximum 8-hour Os concentrations in ppb, based
on a Downscaler fusion of 2006-2008 average monitored values with a 2007
CMAQ model simulation 4C-13
Figure 4C-7. June-August average daily 10am-6pm mean Cb concentrations in ppb, based on a
Downscaler fusion of 2006-2008 average monitored values with a 2007 CMAQ
model simulation 4C-14
Figure 4C-8. April-September average daily maximum 1-hour Os concentrations in ppb, based
on a Downscaler fusion of 2006-2008 average monitored values with a 2007
CMAQ model simulation 4C-15
4C-ii
-------
4C-1. OVERVIEW
The need for greater spatial and temporal coverage of air quality concentration estimates
has grown in recent years as epidemiology and exposure studies that link air quality to health
effects have become more robust and as regulatory needs have increased. These health studies
have historically relied upon direct measurements of air quality concentrations, but prohibitive
logistics and costs limit the spatial coverage and temporal resolution of available ambient
monitoring networks. Numerical methods of interpolation, which predict unknown values from
data observed at known locations, have historically been used by researchers to extend the spatial
coverage of these monitoring networks with a high degree of confidence to inform exposure
studies. However, simple kriging approaches such as Voronoi Neighbor Averaging (VNA) do
not take advantage of the greater availability of model predictions of air quality concentrations
that can enhance the predictive capabilities of numerical methods. Such "data fusion" methods
employ both ambient air quality monitoring data and air quality modeling simulation data as
inputs, and therefore take advantage of the measurement data's accuracy at specific locations and
the air quality model's spatial coverage to generate more robust spatial predictions. For
regulatory purposes, the enhanced Voronoi Neighbor Averaging (eVNA) has been the preferred
method used by EPA to make spatial predictions of ozone (Cb) as part of conducting health
benefit assessments (U.S. EPA, 2010). Given the interest and value of these methods, research
and development efforts have focused on improving their predictive capabilities. This includes
the Office of Research and Development's development of the Downscaler (DS) model that can
potentially provide predictions that are improved over those of eVNA (Berrocal et al., 2011).
This Appendix describes the methods, evaluation, and results of four different techniques
for predicting air quality concentrations. These four techniques are:
1) VNA (interpolating the monitoring data),
2) CMAQ (using the absolute modeled air quality concentrations),
3) eVNA, and
4) Downscaler
EPA used the method with the best performance based on the evaluation to generate
national air quality spatial fields of seasonally averaged Os concentrations as inputs to the
national mortality risk burden assessment in Chapter 8 of the Os Health Risk and Exposure
Assessment (HREA).
Air quality spatial fields are also used in two other applications for the risk and exposure
assessments. In Appendix 4A of the HREA, we describe the methodology used to create urban-
scale spatial fields of hourly Os concentrations for use in the exposure modeling. In Appendix
4A to the Welfare Risk and Exposure Assessment (WREA), we evaluate these same four spatial
4C-1
-------
field techniques for use in creating national-scale air quality spatial fields for the W126 exposure
metric.
4C-2. AIR QUALITY SPATIAL FIELD TECHNIQUES
This section briefly describes the methodology of each of the four techniques considered
for generating air quality spatial fields as inputs to the national risk mortality burden analyses
described in the HREA Chapter 8.
4C-2.1 Voronoi Neighbor Averaging (VNA)
The Voronoi Neighbor Averaging (VNA; Gold, 1997; Chen et al., 2004) interpolation
technique uses inverse distance squared weighted averages of the ambient concentrations from a
set of nearest neighboring monitors to estimate the concentration at a specified location (in this
case CMAQ grid cell centers). VNA identifies the nearest neighboring monitors for the center of
each grid cell using a Delaunay triangulation algorithm, then takes the inverse distance squared
weighted average of the hourly Os concentrations from each neighboring monitor to estimate an
hourly Os concentration value for the center of the grid cell. The following paragraphs provide a
numerical example of the VNA technique applied to a model grid domain.
The first step in VNA is to identify the set of nearest monitors for each grid cell in the
domain. The left-hand panel of Figure 4C-1 below presents a numerical example with nine
model grid cells and seven monitoring sites, with the focus on identifying the set of nearest
neighboring sites to the center of grid cell "E", the center cell. The Delaunay triangulation
algorithm identifies the set of nearest neighboring monitors by drawing a set of polygons called
the "Voronoi diagram" around the center of grid cell "E" and each of the monitoring sites.
Voronoi diagrams have the special property that the each edge of the polygons are the same
distance from the two closest points, as shown in the right-hand panel below.
4C-2
-------
A
D
Monitor: *
80 ppb
10 miles
G
*
B
Monitor:
90 ppb ^
15 miles f
7
. // .
/
/ «
*
Monitor:
100 ppb
20 miles
C
*
F
Monitor:
60 ppb
15 miles
I
*
#= Center Grid-Cell "E':
*
= Air Pollution Monitor
# = Center Grid-Cell "E"
*
= Air Pollution Monitor
Figure 4C-1. Numerical example of the Voronoi Neighbor Averaging (VNA) technique
applied to a model grid domain.
VNA then chooses the monitoring sites that share a boundary with the center of grid cell
"E". These are the nearest neighboring sites, which are used to estimate the concentration value
for grid cell "E". The VNA estimate of the concentration value in grid cell "E" is the inverse
distance squared weighted average of the four monitored concentrations. The further the monitor
is from grid cell "E", the smaller the weight.
For example, the weight for the monitor in grid cell "D" 10 miles from the center of grid
cell "E" is calculated as follows:
1/102
1/102 + 1/152 + 1/152 + 1/202
= 0.4675
Equation(4C-l)
The weights for the other monitors are calculated in a similar fashion. The final VNA
estimate for grid cell "E" is calculated as follows:
VNA(E} = 0.4675 * 80 + 0.2078 * 90 + 0.2078 * 60 + 0.1169 * 100 = 80.3 ppb
Equation (4C-2)
4C-2.2 Community Multi-scale Air Quality (CMAQ) Model
For more than a decade, the EPA's Community Multi-scale Air Quality (CMAQ; Foley et
al., 2010) model has been a valuable computational tool used by EPA and states to inform air
quality management programs. The CMAQ system simultaneously models multiple air
4C-2
-------
pollutants, including Os, parti culate matter, and a variety of air toxics to help regulators
determine the best air quality management scenarios for their communities, states, and countries.
CMAQ is also used by states to assess implementation actions needed to attain National Ambient
Air Quality Standards.
The CMAQ system includes emissions, meteorology, and photochemical modeling
components. Research continues in all of these areas to reduce biases and uncertainties in model
simulations. CMAQ is a multi-scale system that has been applied over continental, national,
regional, and urban modeling domains with progressively finer resolution in a series of nested
grids. The CMAQ modeling community includes researchers, regulators, and forecasters in
academia, government, and the private sector with thousands of users worldwide.
Modeled air quality concentrations from CMAQ simulations have a twofold purpose in
the generation and analysis of air quality spatial fields. First, the modeled concentrations are
"fused" with the ambient measurement data using the eVNA and DS techniques. Second, the
original modeled concentrations are evaluated against the resulting concentration estimates from
the other spatial field techniques, to ensure that those techniques successfully reduce biases in
the modeled air quality fields.
4C-2.3 Enhanced Voronoi Neighbor Averaging (eVNA)
Enhanced Voronoi Neighbor Averaging (eVNA; Timin et al., 2010) is a direct extension
of VNA used to combine monitored and modeled air quality concentration data. Continuing
from the previous numerical example for VNA, suppose the model grid cells containing monitors
are associated with modeled concentrations as shown in Figure 4C-2 below. The modeled
concentrations are used to weight the VNA estimates relative to the modeled concentration
gradient:
eVNA(E} = £?-i Weight; * Monitor/ * f Equation (4C-3)
"i-i u L L
where Monitor* represents the monitored concentration for a nearest neighboring monitor,
Weightt represents the inverse distance squared weight for Monitor t,
Models represents the modeled concentration for grid cell "E", and
Modeli represents the modeled concentration in the grid cell containing Monitor*.
4C-4
-------
A
Model: D
95 ppb
Monitor: *
80 ppb
10 miles
G
*
Model: g
100 ppb
Monitor:
90 ppb
15 miles
Model: TE
85 ppb /
^ -u ^
/
Model: H
120 ppb
*
Monitor:
100 ppb
20 miles
C
*
Model: F
80 ppb
Monitor:
60 ppb
15 miles
I
*
#- Center Grid-Cell "E"
*
= Air Pollution Monitor
Figure 4C-2. Numerical example of the Enhanced Voronoi Neighbor Averaging (eVNA)
technique applied to a model grid domain
Based on the values shown in Figure 4C-2, the eVNA estimate for grid cell "E" is
calculated as follows:
eVNA(E} = (0.4675 * 80 * —) + (0.2078 * 90 * —) + (0.2078 * 60 * —) + (0.1169 * 100 * —) = 70.9 ppb
^ J \ 95/ V 100/ V 80/ \ 120/ FF
Equation (4C-4)
In this example, eVNA adjusts the modeled concentration in grid cell "E" downward to
reflect the tendency for the model to over-predict the monitored concentrations. In general, the
eVNA method attempts to use the monitored concentrations to adjust for model biases, while
preserving local gradients in the modeled concentration fields. The computations for VNA and
eVNA were executed using the R statistical computing program (R, 2012), with the Delaunay
triangulation algorithm implemented in the "deldir" package (Turner, 2012).
4C-2.4 Downscaler (DS)
The Downscaler (DS) model is EPA's most recently developed "data fusion" method for
spatially predicting air pollution concentrations. Downscaler essentially operates by calibrating
CMAQ data to the observational data, and then uses the resulting relationship to predict
"observed" concentrations at new spatial points in the domain. Although similar in principle to a
linear regression, spatial modeling aspects have been incorporated for improving the model fit,
4C-5
-------
and a Bayesian1 approaching to fitting is used to generate an uncertainty value associated with
each concentration prediction. The uncertainties that DS produces are a major distinguishing
feature from earlier fusion methods previously used by EPA such as the "Hierarchical Bayesian"
(HB) model (McMillan et al., 2009). The term "downscaler" refers to the fact that DS takes grid-
averaged predictions from an air quality model such as CMAQ for input and produces point-
based estimates, thus "scaling down" the area of data representation. Although this allows air
pollution concentration estimates to be made at point locations where no observations exist,
caution is needed when interpreting any within-grid cell spatial gradients generated by DS since
they may not exist in the input datasets. The theory, development, and initial evaluation of DS
can be found in the earlier papers of Berrocal, Gelfand, and Holland (2009, 2010, and 2011).
Downscaler develops a relationship between observed and modeled concentrations, and
then uses that relationship to spatially predict what measurements would be at new locations in
the spatial domain based on the input data. This process is separately applied for each time step
(daily in this work) of data, and for each of the pollutants under study (Os and fine particulate
matter, PM2.s). In its most general form, the model can be expressed in an equation similar to
that of linear regression:
Y(s, t) = ~/J0(s, t) + &(t) * ~x(s, t) + e(s, t) Equation (4C-5)
where Y(s,t) is the observed concentration at point s and time t,
~x(s,t) is the CMAQ concentration at point s and time t, (This value is a weighted average
of both the grid cell containing the monitor and neighboring grid cells.)
~fio(s,t) is the intercept, composed of global and local components,
fti(t) is the global slope, (Local components of the slope are contained in the ~x(s,t) term),
e(s,t) is the model residual error.
Downscaler has additional properties that differentiate it from linear regression:
1) Rather than just finding a single optimal solution to Equation (4C-5), DS uses a
Bayesian approach so that uncertainties can be generated along with each concentration
prediction. This involves drawing random samples of model parameters from built-in "prior"
distributions and assessing their fit on the data on the order of thousands of times. After each
iteration, properties of the prior distributions are adjusted to try to improve the fit of the next
iteration. The resulting collection of ~/?o and fii values at each space-time point are the
1 Bayesian statistical modeling refers to methods that are based on Bayes' theorem, and model the world in
terms of probabilities based on previously acquired knowledge.
4C-6
-------
"posterior" distributions, and the means and standard distributions of these are used to predict
concentrations and associated uncertainties at new spatial points.
2) The model is "hierarchical" in structure, meaning that the top level parameters in
Equation (4C-5) (i.e., ~fio(s,t), fti(t), ~x(s,t)) are actually defined in terms of further parameters
and sub-parameters in the DS code. For example, the overall slope and intercept is defined to be
the sum of a global (one value for the entire spatial domain) and local (values specific to each
spatial point) component. This gives more flexibility in fitting a model to the data to optimize
the fit (i.e. minimize e(s,t)).
EPA has recently used DS in other applications, such as providing spatial predictions of
national daily 8-hour Os and PM2.5 concentrations at a census tract resolution to the Centers for
Disease Control (CDC) for the Public Health Air Surveillance Evaluation (PHASE) project as
part of their Environmental Public Health Tracking (EPHT) program.
4C-3. EVALUATION OF AIR QUALITY SPATIAL FIELD TECHNIQUES
This section describes the data, methods, and results of an evaluation that was performed
in order to assess the relative accuracy of the predictions generated by the four air quality spatial
field techniques described in the previous section.
4C-3.1 Data
The evaluation was designed to assess the relative ability of each spatial field technique
to reproduce monitored concentrations for two annual air quality metrics: 1) the 4th highest daily
maximum 8-hour Os concentration (henceforth referred to as the "4th max"), and 2) the May -
September mean of the daily maximum 8-hour Os concentrations (henceforth referred to as the
"seasonal mean"). For the ambient monitoring data, these two metrics were calculated for all
monitors in the contiguous U.S. with complete data for 2007 based on the initial dataset and the
data completeness criteria described in HREA Appendix 4A. For the air quality modeling data,
the two metrics were calculated from hourly Os concentrations based on a CMAQ simulation
with a 12 km gridded domain covering the contiguous U.S., and 2007 emissions and
meteorology inputs (EPA, 2012b).
4C-3.2 Methods
Cross-validation is a method commonly used to evaluate the ability of statistical models
to make accurate predictions. In a cross-validation analysis, the data are split into two subsets,
the "calibration" subset, and the "validation" subset. The calibration subset is used to "fit" the
model, usually by estimating parameters which establish a relationship between the variable of
interest and one or more dependent variables. The resulting model fit is then applied to the
4C-7
-------
dependent variable(s) in the validation subset, and the predictive ability of the model is assessed
by how accurately it is able to reproduce the variable of interest in the validation subset.
The evaluation used a systematic "4-fold" cross-validation scheme based on the CMAQ
modeling domain (e.g., 12 km x 12 km grid covering the continental U.S.). The CMAQ
modeling domain was divided into four groups, or "folds", so that each 2x2 block of 12 km grid
cells had one member in each fold. Figure 4C-3 shows an example of the resulting four folds
with Os monitor locations for the area surrounding southern Lake Michigan. Four cross-
validations were performed using VNA, eVNA, and DS, for both the 4th max and seasonal mean
metrics. The calibration subset in the first cross-validation consisted of the monitors in folds 2,
3, and 4 as shown in Figure 4C-3 (blue dots), while the validation subset consisted of the
monitors in fold 1 (red dots). The remaining cross-validations were performed in a similar
manner, with three of the four folds used as the calibration subset and the final fold used as the
validation subset. Thus, this method resulted in four cross-validation partitions, each with
approximately 75% of the monitoring data used in the calibration subset, and the remaining 25%
used in the validation subset. Each monitor was included in the validation subset exactly once,
resulting in a validation dataset of observed 4th max and seasonal mean values paired with VNA,
eVNA, and DS predictions of those values at monitor locations. The CMAQ predictions were
simply the modeled 4th max and seasonal mean values for the 12 km grid cells containing Os
monitors.
4C-8
-------
4 Fold Validation
• Fold 1 for validation
• Folds 2: 3; 4 for calibaration
Figure 4C-3. Example of the "4-fold" cross-validation scheme used in the evaluation of the
air quality spatial field techniques for the southern Lake Michigan area.
4C-3.3 Results
The cross-validation predictions based on the four air quality spatial field techniques
were compared with the observed 4th max and seasonal mean values based on the ambient data.
The comparison focused on three performance metrics: 1) the root mean squared error (RMSE),
2) the coefficient of variation (R2), and 3) the mean bias (MB). The results of these comparisons
are shown in Figure 4C-4 (4th max) and Figure 4C-5 (seasonal mean), and Table 4C-1 contains a
summary of the three performance metrics for each technique.
The cross-validation results clearly showed that VNA, eVNA, and DS more accurately
predict monitored 4th max and seasonal mean concentrations than the CMAQ model. The scatter
plots and the mean bias statistics indicated that both eVNA and DS were effective at reducing the
amount of bias present in the modeled concentrations. The differences between VNA, eVNA,
and DS were much smaller, but the performance statistics consistently indicated that overall DS
was able to most accurately reproduce the observed values. DS had the highest R2 values, and
the lowest RMSE and MB values of the four techniques for both air quality metrics. In contrast
4C-9
-------
to eVNA, DS seemed to be an improvement over VNA, which did not make use of the modeled
concentration data. Based on these results, DS was deemed the most appropriate technique for
generating spatial fields of seasonal mean Os concentrations for the national risk mortality
burden analyses.
Validation Results - 2007 4th Highest Daily Maximum 8-hour O3 (ppb)
o
Csl
O
o
- RMSE = 8.36
RA2 = 0.501
MB = 3.22
O
CMAQ
V I
o
Obs
eVNA
- RMSE = 6.13
RA2 = 0.645
MB = 0.09
40
60
o
CN
O
o
RMSE = 5.34
RA2 = 0.702
MB = 0.32
80
100
120
Figure 4C-4. Cross-validation results for the 2007 annual 4th highest daily maximum
concentrations.
4C-10
-------
Validation Results - 2007 May - September Average Daily Maximum 8-hour O3 (ppb)
CMAQ
eVNA
RMSE = 3.56
RA2 = 0.
MB = 0.08
- RMSE = 3.39
RA2 = 0.826
MB = -0.02
20
30
40 50
60
70
80
Figure 4C-5. Cross-validation results for the 2007 May-September mean of the daily
maximum 8-hour Os concentrations.
4C-11
-------
Table 4C-1. Summary of the cross-validation performance metrics for the four air quality
spatial field techniques.
Performance
Metric
RMSE
RMSE
R2
R2
MB
MB
Os Metric
4th max
seasonal mean
4th max
seasonal mean
4th max
seasonal mean
VNA
5.34
3.56
0.702
0.808
0.32
0.08
CMAQ
8.36
6.83
0.501
0.634
3.22
4.71
eVNA
6.13
3.85
0.645
0.779
0.09
0.06
DS
5.01
3.39
0.736
0.826
0.03
-0.02
4C-4, AIR QUALITY INPUTS TO THE NATIONAL-SCALE MORTALITY RISK
BURDENT ASSESSMENT
Three air quality spatial fields were created using DS as inputs to the national mortality
risk burden analyses. These fields were based on three seasonal mean Os metrics:
1) May-September average daily maximum 8-hour Os concentration (consistent with the
metric used by Smith et al., 2009);
2) June-August average daily 10am-6pm mean Os concentration (consistent with the metric
used by Zanobetti and Schwartz, 2008); and
3) April-September average daily maximum 1-hour Os concentration (consistent with the
metric used by Jerrett et al., 2009).
For the ambient monitoring data, the 2006-2008 average of these annual metrics were
calculated for all monitors in the contiguous U.S. with complete data for 2006-2008 based on the
initial dataset and the data completeness criteria described in HREA Appendix 4A. For the air
quality modeling data, these metrics were calculated from hourly Os concentrations based on a
CMAQ simulation with a 12 km gridded domain covering the contiguous U.S., and 2007
emissions and meteorology inputs. This simulation differed from the one used for the
evaluations in that the wildfire and power plant emissions inputs were based on multi-year
averages instead of year-specific estimates. Appendix 4B contains a complete description of this
CMAQ simulation and provides relevant model evaluation results.
Figure 4C-6, Figure 4C-7, and Figure 4C-8 show maps of the DS air quality spatial fields
for the May-September average daily maximum 8-hour Os concentrations, the June-August
average daily 10am-6pm mean Os concentrations, and the April-September average daily
maximum 1-hour Os concentrations, respectively. The overall spatial pattern seen in these three
fields is very similar, with the highest concentration values for each metric occurring in Southern
California.
4C-12
-------
30
40
50
60
70
80
Figure 4C-6. May-September average daily maximum 8-hour Os concentrations in ppb, based on a Downscaler fusion of
2006-2008 average monitored values with a 2007 CMAQ model simulation.
4C-13
-------
30
40
50
60
70
80
Figure 4C-7. June-August average daily 10am-6pm mean Os concentrations in ppb, based on a Downscaler fusion of 2006-
2008 average monitored values with a 2007 CMAQ model simulation.
4C-14
-------
20
30
40
60
70
80
90
Figure 4C-8. April-September average daily maximum 1-hour Os concentrations in ppb, based on a Downscaler fusion of
2006-2008 average monitored values with a 2007 CMAQ model simulation.
4C-15
-------
4C-5. REFERENCES
Abt Associates, Inc. (2010). Environmental Benefits and Mapping Program (Version 4.0).
Bethesda, MD. Prepared for U.S. Environmental Protection Agency Office of Air Quality
Planning and Standards. Research Triangle Park, NC. Available on the Internet at
.
Berrocal, V.J., Gelfand, A.E., Holland, D.M. (2009). A Spatio-Temporal Downscaler for
Output from Numerical Models. Journal of Agricultural, Biological, and Environmental
Statistics, 15(2), 176-197.
Berrocal, V.J., Gelfand, A.E., Holland, D.M. (2010). ABivariate Space-Time Downscaler
Under Space and Time Misalignment. Annals of Applied Statistics, 4(4), 1942-1975.
Berrocal, V.J., Gelfand, A.E., Holland, D.M. (2011). Space-Time Data Fusion Under Error in
Computer Model Output: An Application to Modeling Air Quality. Biometrics, 68(3),
837-848.
Chen, J., Zhao, R., Li, Z. (2004). Voronoi-based k-order neighbor relations for spatial analysis.
ISPRS J Photogrammetry Remote Sensing, 59(1-2), 60-72.
Foley, K.M., Roselle, S.J., Appel, K.W., Pleim, I.E., Otte, T.L., Mathur, R., Sarwar, G., Young,
J.O., Gilliam, R.C., Nolte, C.G, Kelly, J.T., Gilliland, A.B., Bash, J.O. (2010).
Incremental testing of the Community Multiscale Air Quality (CMAQ) modeling system
version 4.7. Geoscientific Model Development, 3, 205-226.
Gold, C. (1997). Voronoi methods in GIS. In: Algorithmic Foundation of Geographic
Information Systems (va Kereveld M., Nievergelt, J., Roos, T., Widmayer, P., eds.).
Lecture Notes in Computer Science, Vol 1340. Berlin: Springer-Verlag, 21-35.
Hall, E., Eyth, A., Phillips, S. (2012). Hierarchical Bayesian Model (HBM)-Derived Estimates
of Air Quality for 2007: Annual Report. EPA/600/R-12/538. Available on the Internet
at: http://www.epa.gov/heasd/sources/projects/CDC/AnnualReports/2007 HBM.pdf
Jerrett, M., R.T. Burnett, C.A. Pope III, K. Ito, G. Thurston, D. Krewski, Y. Shi, E. Calle, M.
Thun. (2009). Long-term O3 exposure and mortality. N. Eng. J. Med., 360:1085-1095.
McMillan, N.J., Holland, D.M., Morara, M., Feng, J. (2009). Combining Numerical Model
Output and Particulate Data using Bayesian Space-Time Modeling. Environmetrics, Vol.
21,48-65.
R Core Team. (2012). R: A language and environment for statistical computing. R Foundation
for Statistical Computing, Vienna, Austria. http://www.R-project.org/.
Smith, RL; Xu, B; Switzer, P. (2009). Reassessing the relationship between ozone and short-
term mortality in U.S. urban communities. Inhal Toxicol 21: 37-61.
Timin B, Wesson K, Thurman J. (2010). Application of Model and Ambient Data Fusion
Techniques to Predict Current and Future Year PM2.5 Concentrations in Unmonitored
Areas. Pp. 175-179 in SteynDG, Rao St (eds.). Air Pollution Modeling and Its
Application XX. Netherlands: Springer.
Turner, R. (2012). deldir: Delaunay Triangulation and Dirichlet (Voronoi) Tessellation. R
package version 0.0-19. http://CRAN.R-project.org/package=deldir
4C-16
-------
U.S. EPA. (2012a). Health Risk and Exposure Assessment for Ozone, First External Review
Draft. Available on the Internet at:
http://www.epa.gov/ttn/naaqs/standards/ozone/s_o3_2008_rea.html
U.S. EPA. (2012b). Air Quality Modeling Technical Support Document for the Regulatory
Impact Analysis for the Revisions to the National Ambient Air Quality Standards for
Particulate Matter. Available on the Internet at:
http://www.epa.gov/ttn/naaqs/standards/pm/data/201212aqm. pdf
Wells, B., Wesson, K., Jenkins, S. (2012). Analysis of Recent U.S. Ozone Air Quality Data to
Support the O3 NAAQS Review and Quadratic Rollback Simulations to Support the First
Draft of the Risk and Exposure Assessment. Available on the Internet at:
http://www.epa.gov/ttn/naaqs/standards/ozone/s_o3_td.html
Zanobetti, A., and J. Schwartz. (2008). Mortality Displacement in the Association of Ozone
with Mortality: An Analysis of 48 Cities in the United States. American Journal of
Respiratory and Critical Care Medicine, 177, 184-189.
4C-17
-------
This page left intentionally blank
4C-18
-------
APPENDIX 4D
Model-based Air Quality Adjustment Using the Higher-order
Decoupled Direct Method (HDDM)
Table of Contents
4D-1, MOTIVATION FOR AN NEW TECHNIQUE TO SIMULATE OZONE
CONCENTRATIONS UNDER ALTERNATIVE STANDARDS 4D-1
4D-2, HIGHER-ORDER DECOUPLED DIRECT METHOD (HDDM) 4D-2
4D-2.1 Capabilities 4D-2
4D-2.2 Limitations 4D-4
4D-3. APPLYING HDDM/CMAQ TO ADJUST OZONE TO JUST MEET EXISTING AND
ALTERNATIVE STANDARDS: METHODOLOGY 4D-5
4D-3.1 Conceptual Framework 4D-5
4D-3.2 Application to Measured Os Concentrations in 15 Urban Areas 4D-6
4D-3.2.1 Multi-step Application of HDDM Sensitivities 4D-7
4D-3.2.2 Relationships between HDDM Sensitivities and Modeled Os Concentrations
4D-16
4D-3.2.3 Application of Sensitivity Regressions to Ambient Data 4D-24
4D-3.2.4 Alternate Methodologies 4D-25
4D-4, APPLYING HDDM/CMAQ TO ADJUST OZONE TO JUST MEET EXISTING AND
ALTERNATIVE STANDARDS: RESULTS 4D-26
4D-4.1 Emission Reductions Applied to Meet Alternative Standards 4D-26
4D-4.2 Design Values 4D-28
4D-4.3 Distribution of Hourly Os Concentrations 4D-40
4D-4.4 Standard Errors for Predicted Hourly and Daily 8-hr Maximum Os Concentrations
4D-71
4D-i
-------
4D-4.5 Air Quality Inputs for the Epidemiology-Based Risk Assessment 4D-90
4D-4.6 Air Quality Inputs for the Exposure and Clinical Risk Assessment 4D-99
4D-4.7 Comparing Air Quality Adjustments Based on NOx Reductions Only to Air
Quality Adjustments Based on NOx and VOC Reductions 4D-162
4D-5. REFERENCES 4D-177
4D-ii
-------
List of Tables
Table 4D-1. X and Y outpoints used in Equations (4D-4) and (4D-8). Note: the NOx/VOC
sensitivity case was only performed in two cities 4D-12
Table 4D-2. Percent emissions reductions used for each urban area to obtain each standard. 27
Table 4D-3. Design values for the Atlanta area regulatory monitors from observed data and for
adjustments to meet the existing and potential alternative standards in 2006-2008
and 2008-2010 4D-29
Table 4D-4. Design values for the Baltimore area regulatory monitors from observed data and
for adjustments to meet the existing and potential alternative standards in 2006-
2008 and 2008-2010 4D-29
Table 4D-5. Design values for the Boston area regulatory monitors from observed data and for
adjustments to meet the existing and potential alternative standards in 2006-2008
and 2008-2010 4D-30
Table 4D-6. Design values for the Chicago area regulatory monitors from observed data and
for adjustments to meet the existing and potential alternative standards using NOX
and VOC emissions reductions in 2006-2008 and 2008-2010 4D-30
Table 4D-7. Design values for the Cleveland area regulatory monitors from observed data and
for adjustments to meet the existing and potential alternative standards in 2006-
2008 and 2008-2010 4D-31
Table 4D-8. Design values for the Dallas area regulatory monitors from observed data and for
adjustments to meet the existing and potential alternative standards in 2006-2008
and 2008-2010 4D-32
Table 4D-9. Design values for the Denver area regulatory monitors from observed data and for
adjustments to meet the existing and potential alternative standards using NOX and
VOC emissions reductions in 2006-2008 and 2008-2010 4D-33
Table 4D-10. Design values for the Detroit area regulatory monitors from observed data and for
adjustments to meet the existing and potential alternative standards in 2006-2008
and 2008-2010 4D-33
Table 4D-11. Design values for the Houston area regulatory monitors from observed data and
for adjustments to meet the existing and potential alternative standards in 2006-
2008 and 2008-2010 4D-34
Table 4D-12. Design values for the Los Angeles area regulatory monitors from observed data
and for adjustments to meet the existing and potential alternative standards using
the lower bound of the 95th percent confidence interval of estimated hourly Os in
2006-2008 and 2008-2010 4D-35
Table 4D-13. Design values for the New York area regulatory monitors from observed data and
for adjustments to meet the existing and potential alternative standards using the
lower bound of the 95th percent confidence interval of estimated hourly Os in
2006-2008 and 2008-2010 4D-36
Table 4D-14. Design values for the Philadelphia area regulatory monitors from observed data
and for adjustments to meet the existing and potential alternative standards in
2006-2008 and 2008-2010 4D-37
Table 4D-15. Design values for the Sacramento area regulatory monitors from observed data
and for adjustments to meet the existing and potential alternative standards in
2006-2008 and 2008-2010 4D-38
-------
Table 4D
Table 4D
Table 4D
Table 4D
Table 4D
-16. Design values for the Saint Louis area regulatory monitors from observed data
and for adjustments to meet the existing and potential alternative standards in
2006-2008 and 2008-2010 4D-39
Design values for the Washington D.C. area regulatory monitors from observed
data and for adjustments to meet the existing and potential alternative standards in
2006-2008 and 2008-2010 4D-39
Mean standard error (ppb) in adjusted hourly Cb concentration in each urban area
for each standard 4D-74
95th percentile standard error (ppb) in adjusted hourly Os concentration in each
urban area for each standard 4D-75
-20. Comparison of NOx-only and NOx/VOC emission reductions applied in
sensitivity analyses for nine urban areas 4D-163
-17.
-18.
-19.
List of Figures
Figure 4D-1. Flow diagram demonstrating DDM model-based Cb adjustment approach... 4D-6
Figure 4D-2. Conceptual picture of 3-step application of HDDM sensitivities 4D-9
Figure 4D-3. Comparison of brute force and 3-step HDDM Os estimates for 50% NOx cut
conditions: Atlanta, Baltimore, Boston, Chicago, Cleveland, Dallas, Denver, and
Detroit 4D-13
Figure 4D-4. Comparison of brute force and 3-step HDDM Os estimates for 50% NOx cut
conditions: Houston, Los Angeles, New York, Philadelphia, Sacramento, St.
Louis, and Washington D.C 4D-14
Figure 4D-5. Comparison of brute force and 3-step HDDM Os estimates for 90% NOx cut
conditions: Atlanta, Baltimore, Boston, Chicago, Cleveland, Dallas, Denver, and
Detroit 4D-15
Figure 4D-6. Comparison of brute force and 3-step HDDM Os estimates for 90% NOx cut
conditions: Houston, Los Angeles, New York, Philadelphia, Sacramento, St.
Louis, and Washington D.C 4D-16
Figure 4D-7. Relationship between SNOX and hourly Os at a NOx-limited site downwind of
Atlanta (summer). Relationships are shown for one nighttime hour, one morning
rush-hour hour, one daytime hour, and one evening rush-hour hour. The solid
blue line show the linear regression for these points and the dotted blue line shows
the floor value used for SNOX based on the 5th percentile modeled value 4D-19
Figure 4D-8. Relationship between SNOX and hourly Os at a NOx-saturated site in Queens
County, NY (autumn). Relationships are shown for one nighttime hour, one
morning rush-hour hour, one daytime hour, and one evening rush-hour hour. The
solid blue line show the linear regression for these points and the dotted blue line
shows the floor value used for SNOX based on the 5th percentile modeled value.
4D-20
Figure 4D-9. Relationship between SNOX and hourly Os at a NOx-saturated site in Suffolk
County, NY on Long Island (spring). Relationships are shown for one nighttime
hour, one morning rush-hour hour, one daytime hour, and one evening rush-hour
hour. The solid blue line show the linear regression for these points and the
4D-iv
-------
dotted blue line shows the floor value used for SNOX based on the 5th percentile
modeled value 4D-21
Figure 4D-10. Relationship between S2NOx and SNOX at a NOx-limited site downwind of Atlanta
(summer). Relationships are shown for one nighttime hour, one morning rush-
hour hour, one daytime hour, and one evening rush-hour hour. The solid blue line
shows the linear regression for these points 4D-22
Figure 4D-11. Relationship between S2NOx and SNOX at a NOx-saturated site in Queens County,
NY (autumn). Relationships are shown for one nighttime hour, one morning
rush-hour hour, one daytime hour, and one evening rush-hour hour. The solid
blue line shows the linear regression for these points 4D-23
Figure 4D-12. Relationship between S2NOx and SNOX at a NOx-saturated site in Suffolk County,
NY on Long Island (spring). Relationships are shown for one nighttime hour, one
morning rush-hour hour, one daytime hour, and one evening rush-hour hour. The
solid blue line shows the linear regression for these points 4D-24
Figure 4D-13. Hourly Os distributions at Atlanta area regulatory monitoring sites for observed
air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations,
red boxes (pink whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 75 ppb adjustment scenarios and blue boxes (light blue
whiskers/dots) show the predicted distribution of hourly Os concentrations for the
65 ppb adjustment scenario. Boxes show the interquartile range, whiskers extend
to 1.5 x the interquartile range and dots depict outlier values 4D-42
Figure 4D-14. Hourly Os distributions at Baltimore area regulatory monitoring sites for observed
air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations,
red boxes (pink whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 75 ppb adjustment scenarios and blue boxes (light blue
whiskers/dots) show the predicted distribution of hourly Os concentrations for the
65 ppb adjustment scenario. Boxes show the interquartile range, whiskers extend
to 1.5 x the interquartile range and dots depict outlier values 4D-43
Figure 4D-15. Hourly Os distributions at Boston area regulatory monitoring sites for observed
air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations,
red boxes (pink whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 75 ppb adjustment scenarios and blue boxes (light blue
whiskers/dots) show the predicted distribution of hourly Os concentrations for the
65 ppb adjustment scenario. Boxes show the interquartile range, whiskers extend
to 1.5 x the interquartile range and dots depict outlier values 4D-44
Figure 4D-16. Hourly Os distributions at Chicago area regulatory monitoring sites for observed
air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations,
red boxes (pink whiskers/dots) show the predicted distribution of hourly Os
4D-v
-------
concentrations for the 75 ppb adjustment scenarios and blue boxes (light blue
whiskers/dots) show the predicted distribution of hourly Os concentrations for the
65 ppb adjustment scenario. Boxes show the interquartile range, whiskers extend
to 1.5 x the interquartile range and dots depict outlier values 4D-45
Figure 4D-17. Hourly Os distributions at Cleveland area regulatory monitoring sites for observed
air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations,
red boxes (pink whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 75 ppb adjustment scenarios and blue boxes (light blue
whiskers/dots) show the predicted distribution of hourly Os concentrations for the
65 ppb adjustment scenario. Boxes show the interquartile range, whiskers extend
to 1.5 x the interquartile range and dots depict outlier values 4D-46
Figure 4D-18. Hourly Os distributions at Dallas area regulatory monitoring sites for observed air
quality, and air quality adjusted to meet the existing (75 ppb) and alternative (65
ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red
boxes (pink whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 75 ppb adjustment scenarios and blue boxes (light blue
whiskers/dots) show the predicted distribution of hourly Os concentrations for the
65 ppb adjustment scenario. Boxes show the interquartile range, whiskers extend
to 1.5 x the interquartile range and dots depict outlier values 4D-47
Figure 4D-19. Hourly Os distributions at Denver area regulatory monitoring sites for observed
air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations,
red boxes (pink whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 75 ppb adjustment scenarios and blue boxes (light blue
whiskers/dots) show the predicted distribution of hourly Os concentrations for the
65 ppb adjustment scenario. Boxes show the interquartile range, whiskers extend
to 1.5 x the interquartile range and dots depict outlier values 4D-48
Figure 4D-20. Hourly Os distributions at Detroit area regulatory monitoring sites for observed
air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations,
red boxes (pink whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 75 ppb adjustment scenarios and blue boxes (light blue
whiskers/dots) show the predicted distribution of hourly Os concentrations for the
65 ppb adjustment scenario. Boxes show the interquartile range, whiskers extend
to 1.5 x the interquartile range and dots depict outlier values 4D-49
Figure 4D-21. Hourly Os distributions at Houston area regulatory monitoring sites for observed
air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations,
red boxes (pink whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 75 ppb adjustment scenarios and blue boxes (light blue
4D-vi
-------
whiskers/dots) show the predicted distribution of hourly Os concentrations for the
65 ppb adjustment scenario. Boxes show the interquartile range, whiskers extend
to 1.5 x the interquartile range and dots depict outlier values 4D-50
Figure 4D-22. Hourly Os distributions at Los Angeles area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White
boxes (black whiskers/dots) show the observed distribution of hourly Os
concentrations, red boxes (pink whiskers/dots) show the predicted distribution of
hourly Os concentrations for the 75 ppb adjustment scenarios and blue boxes
(light blue whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 65 ppb adjustment scenario. Boxes show the interquartile
range, whiskers extend to 1.5 x the interquartile range and dots depict outlier
values 4D-51
Figure 4D-23. Hourly Os distributions at New York area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White
boxes (black whiskers/dots) show the observed distribution of hourly Os
concentrations, red boxes (pink whiskers/dots) show the predicted distribution of
hourly Os concentrations for the 75 ppb adjustment scenarios and blue boxes
(light blue whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 65 ppb adjustment scenario. Boxes show the interquartile
range, whiskers extend to 1.5 x the interquartile range and dots depict outlier
values 4D-52
Figure 4D-24. Hourly Os distributions at Philadelphia area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White
boxes (black whiskers/dots) show the observed distribution of hourly Os
concentrations, red boxes (pink whiskers/dots) show the predicted distribution of
hourly Os concentrations for the 75 ppb adjustment scenarios and blue boxes
(light blue whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 65 ppb adjustment scenario. Boxes show the interquartile
range, whiskers extend to 1.5 x the interquartile range and dots depict outlier
values 4D-53
Figure 4D-25. Hourly Os distributions at Sacramento area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White
boxes (black whiskers/dots) show the observed distribution of hourly Os
concentrations, red boxes (pink whiskers/dots) show the predicted distribution of
hourly Os concentrations for the 75 ppb adjustment scenarios and blue boxes
(light blue whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 65 ppb adjustment scenario. Boxes show the interquartile
range, whiskers extend to 1.5 x the interquartile range and dots depict outlier
values 4D-54
Figure 4D-26. Hourly Os distributions at St. Louis area regulatory monitoring sites for observed
air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
4D-vii
-------
(black whiskers/dots) show the observed distribution of hourly Os concentrations,
red boxes (pink whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 75 ppb adjustment scenarios and blue boxes (light blue
whiskers/dots) show the predicted distribution of hourly Os concentrations for the
65 ppb adjustment scenario. Boxes show the interquartile range, whiskers extend
to 1.5 x the interquartile range and dots depict outlier values 4D-55
Figure 4D-27. Hourly Os distributions at Washington, D.C. area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White
boxes (black whiskers/dots) show the observed distribution of hourly Os
concentrations, red boxes (pink whiskers/dots) show the predicted distribution of
hourly Os concentrations for the 75 ppb adjustment scenarios and blue boxes
(light blue whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 65 ppb adjustment scenario. Boxes show the interquartile
range, whiskers extend to 1.5 x the interquartile range and dots depict outlier
values 4D-56
Figure 4D-28. Monthly Os distributions at Atlanta area regulatory monitoring sites for observed
air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations,
red boxes (pink whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 75 ppb adjustment scenarios and blue boxes (light blue
whiskers/dots) show the predicted distribution of hourly Os concentrations for the
65 ppb adjustment scenario. Boxes show the interquartile range, whiskers extend
to 1.5 x the interquartile range and dots depict outlier values 4D-57
Figure 4D-29. Monthly Os distributions at Baltimore area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White
boxes (black whiskers/dots) show the observed distribution of hourly Os
concentrations, red boxes (pink whiskers/dots) show the predicted distribution of
hourly Os concentrations for the 75 ppb adjustment scenarios and blue boxes
(light blue whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 65 ppb adjustment scenario. Boxes show the interquartile
range, whiskers extend to 1.5 x the interquartile range and dots depict outlier
values 4D-58
Figure 4D-30. Monthly Os distributions at Boston area regulatory monitoring sites for observed
air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations,
red boxes (pink whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 75 ppb adjustment scenarios and blue boxes (light blue
whiskers/dots) show the predicted distribution of hourly Os concentrations for the
65 ppb adjustment scenario. Boxes show the interquartile range, whiskers extend
to 1.5 x the interquartile range and dots depict outlier values 4D-59
Figure 4D-31. Monthly Os distributions at Chicago area regulatory monitoring sites for observed
air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
4D-viii
-------
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations,
red boxes (pink whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 75 ppb adjustment scenarios and blue boxes (light blue
whiskers/dots) show the predicted distribution of hourly Os concentrations for the
65 ppb adjustment scenario. Boxes show the interquartile range, whiskers extend
to 1.5 x the interquartile range and dots depict outlier values 4D-60
Figure 4D-32. Monthly Os distributions at Cleveland area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White
boxes (black whiskers/dots) show the observed distribution of hourly Os
concentrations, red boxes (pink whiskers/dots) show the predicted distribution of
hourly Os concentrations for the 75 ppb adjustment scenarios and blue boxes
(light blue whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 65 ppb adjustment scenario. Boxes show the interquartile
range, whiskers extend to 1.5 x the interquartile range and dots depict outlier
values 4D-61
Figure 4D-33. Monthly Os distributions at Dallas area regulatory monitoring sites for observed
air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations,
red boxes (pink whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 75 ppb adjustment scenarios and blue boxes (light blue
whiskers/dots) show the predicted distribution of hourly Os concentrations for the
65 ppb adjustment scenario. Boxes show the interquartile range, whiskers extend
to 1.5 x the interquartile range and dots depict outlier values 4D-62
Figure 4D-34. Monthly Os distributions at Denver area regulatory monitoring sites for observed
air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations,
red boxes (pink whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 75 ppb adjustment scenarios and blue boxes (light blue
whiskers/dots) show the predicted distribution of hourly Os concentrations for the
65 ppb adjustment scenario. Boxes show the interquartile range, whiskers extend
to 1.5 x the interquartile range and dots depict outlier values 4D-63
Figure 4D-35. Monthly Os distributions at Detroit area regulatory monitoring sites for observed
air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations,
red boxes (pink whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 75 ppb adjustment scenarios and blue boxes (light blue
whiskers/dots) show the predicted distribution of hourly Os concentrations for the
65 ppb adjustment scenario. Boxes show the interquartile range, whiskers extend
to 1.5 x the interquartile range and dots depict outlier values 4D-64
Figure 4D-36. Monthly Os distributions at Houston area regulatory monitoring sites for observed
air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
4D-ix
-------
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations,
red boxes (pink whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 75 ppb adjustment scenarios and blue boxes (light blue
whiskers/dots) show the predicted distribution of hourly Os concentrations for the
65 ppb adjustment scenario. Boxes show the interquartile range, whiskers extend
to 1.5 x the interquartile range and dots depict outlier values 4D-65
Figure 4D-37. Monthly Os distributions at Los Angeles area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White
boxes (black whiskers/dots) show the observed distribution of hourly Os
concentrations, red boxes (pink whiskers/dots) show the predicted distribution of
hourly Os concentrations for the 75 ppb adjustment scenarios and blue boxes
(light blue whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 65 ppb adjustment scenario. Boxes show the interquartile
range, whiskers extend to 1.5 x the interquartile range and dots depict outlier
values 4D-66
Figure 4D-38. Monthly Os distributions at New York area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White
boxes (black whiskers/dots) show the observed distribution of hourly Os
concentrations, red boxes (pink whiskers/dots) show the predicted distribution of
hourly Os concentrations for the 75 ppb adjustment scenarios and blue boxes
(light blue whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 65 ppb adjustment scenario. Boxes show the interquartile
range, whiskers extend to 1.5 x the interquartile range and dots depict outlier
values 4D-67
Figure 4D-39. Monthly Os distributions at Philadelphia area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White
boxes (black whiskers/dots) show the observed distribution of hourly Os
concentrations, red boxes (pink whiskers/dots) show the predicted distribution of
hourly Os concentrations for the 75 ppb adjustment scenarios and blue boxes
(light blue whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 65 ppb adjustment scenario. Boxes show the interquartile
range, whiskers extend to 1.5 x the interquartile range and dots depict outlier
values 4D-68
Figure 4D-40. Monthly Os distributions at Sacramento area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White
boxes (black whiskers/dots) show the observed distribution of hourly Os
concentrations, red boxes (pink whiskers/dots) show the predicted distribution of
hourly Os concentrations for the 75 ppb adjustment scenarios and blue boxes
(light blue whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 65 ppb adjustment scenario. Boxes show the interquartile
4D-x
-------
range, whiskers extend to 1.5 x the interquartile range and dots depict outlier
values 4D-69
Figure 4D-41. Monthly Os distributions at St. Louis area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White
boxes (black whiskers/dots) show the observed distribution of hourly Os
concentrations, red boxes (pink whiskers/dots) show the predicted distribution of
hourly Os concentrations for the 75 ppb adjustment scenarios and blue boxes
(light blue whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 65 ppb adjustment scenario. Boxes show the interquartile
range, whiskers extend to 1.5 x the interquartile range and dots depict outlier
values 4D-70
Figure 4D-42. Monthly Os distributions at Washington, D.C. area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White
boxes (black whiskers/dots) show the observed distribution of hourly Os
concentrations, red boxes (pink whiskers/dots) show the predicted distribution of
hourly Os concentrations for the 75 ppb adjustment scenarios and blue boxes
(light blue whiskers/dots) show the predicted distribution of hourly Os
concentrations for the 65 ppb adjustment scenario. Boxes show the interquartile
range, whiskers extend to 1.5 x the interquartile range and dots depict outlier
values 4D-71
Figure 4D-43. Mean standard error in maximum daily 8-hr Os concentrations at each monitoring
location in Atlanta for the 75 ppb adjustment scenario (top row) and the 60 ppb
adjustment scenario (bottom row) for the 2006-2008 time period (left column)
and the 2008-2010 time period (right column) 4D-76
Figure 4D-44. Mean standard error in maximum daily 8-hr Os concentrations at each monitoring
location in Baltimore for the 75 ppb adjustment scenario (top row) and the 60 ppb
adjustment scenario (bottom row) for the 2006-2008 time period (left column)
and the 2008-2010 time period (right column) 4D-77
Figure 4D-45. Mean standard error in maximum daily 8-hr Os concentrations at each monitoring
location in Boston for the 75 ppb adjustment scenario (top row) and the 60 ppb
adjustment scenario (bottom row) for the 2006-2008 time period (left column)
and the 2008-2010 time period (right column) 4D-78
Figure 4D-46. Mean standard error in maximum daily 8-hr Os concentrations at each monitoring
location in Chicago for the 75 ppb adjustment scenario (top row) and the 60 ppb
adjustment scenario (bottom row) for the 2006-2008 time period (left column)
and the 2008-2010 time period (right column) 4D-79
Figure 4D-47. Mean standard error in maximum daily 8-hr Os concentrations at each monitoring
location in Cleveland for the 75 ppb adjustment scenario (top row) and the 60 ppb
adjustment scenario (bottom row) for the 2006-2008 time period (left column)
and the 2008-2010 time period (right column) 4D-80
Figure 4D-48. Mean standard error in maximum daily 8-hr Os concentrations at each monitoring
location in Dallas for the 75 ppb adjustment scenario (top row) and the 60 ppb
adjustment scenario (bottom row) for the 2006-2008 time period (left column)
and the 2008-2010 time period (right column) 4D-81
4D-xi
-------
Figure 4D-49. Mean standard error in maximum daily 8-hr Os concentrations at each monitoring
location in Washington D.C. for the 75 ppb adjustment scenario (top row) and the
60 ppb adjustment scenario (bottom row) for the 2006-2008 time period (left
column) and the 2008-2010 time period (right column) 4D-82
Figure 4D-50. Mean standard error in maximum daily 8-hr Os concentrations at each monitoring
location in Denver for the 75 ppb adjustment scenario (top row) and the 60 ppb
adjustment scenario (bottom row) for the 2006-2008 time period (left column)
and the 2008-2010 time period (right column) 4D-83
Figure 4D-51. Mean standard error in maximum daily 8-hr Os concentrations at each monitoring
location in Detroit for the 75 ppb adjustment scenario (top row) and the 60 ppb
adjustment scenario (bottom row) for the 2006-2008 time period (left column)
and the 2008-2010 time period (right column) 4D-84
Figure 4D-52. Mean standard error in maximum daily 8-hr Os concentrations at each monitoring
location in Houston for the 75 ppb adjustment scenario (top row) and the 60 ppb
adjustment scenario (bottom row) for the 2006-2008 time period (left column)
and the 2008-2010 time period (right column) 4D-85
Figure 4D-53. Mean standard error in maximum daily 8-hr Os concentrations at each monitoring
location in Los Angele for the 75 ppb adjustment scenario (top row) and the 60
ppb adjustment scenario (bottom row) for the 2006-2008 time period (left
column) and the 2008-2010 time period (right column) 4D-86
Figure 4D-54. Mean standard error in maximum daily 8-hr Os concentrations at each monitoring
location in New York for the 75 ppb adjustment scenario (top row) and the 60 ppb
adjustment scenario (bottom row) for the 2006-2008 time period (left column)
and the 2008-2010 time period (right column) 4D-87
Figure 4D-55. Mean standard error in maximum daily 8-hr Os concentrations at each monitoring
location in Philadelphia for the 75 ppb adjustment scenario (top row) and the 60
ppb adjustment scenario (bottom row) for the 2006-2008 time period (left
column) and the 2008-2010 time period (right column) 4D-88
Figure 4D-56. Mean standard error in maximum daily 8-hr Os concentrations at each monitoring
location in Sacramento for the 75 ppb adjustment scenario (top row) and the 60
ppb adjustment scenario (bottom row) for the 2006-2008 time period (left
column) and the 2008-2010 time period (right column) 4D-89
Figure 4D-57. Mean standard error in maximum daily 8-hr Os concentrations at each monitoring
location in St. Louis for the 75 ppb adjustment scenario (top row) and the 60 ppb
adjustment scenario (bottom row) for the 2006-2008 time period (left column)
and the 2008-2010 time period (right column) 4D-90
Figure 4D-58. Composite monitor daily maximum 8-hour Os values for Atlanta based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's
represent mean values, whiskers extend up to 1.5x the inter-quartile range from
the boxes, and circles represent outliers 4D-93
Figure 4D-59. Composite monitor daily maximum 8-hour Os values for Baltimore based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's
represent mean values, whiskers extend up to 1.5x the inter-quartile range from
the boxes, and circles represent outliers 4D-93
Figure 4D-60. Composite monitor daily maximum 8-hour Os values for Boston based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's
4D-xii
-------
represent mean values, whiskers extend up to 1.5x the inter-quartile range from
the boxes, and circles represent outliers 4D-94
Figure 4D-61. Composite monitor daily maximum 8-hour Os values for Cleveland based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's
represent mean values, whiskers extend up to 1.5x the inter-quartile range from
the boxes, and circles represent outliers 4D-94
Figure 4D-62. Composite monitor daily maximum 8-hour Os values for Denver based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's
represent mean values, whiskers extend up to 1.5x the inter-quartile range from
the boxes, and circles represent outliers 4D-95
Figure 4D-63. Composite monitor daily maximum 8-hour Os values for Detroit based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's
represent mean values, whiskers extend up to 1.5x the inter-quartile range from
the boxes, and circles represent outliers 4D-95
Figure 4D-64. Composite monitor daily maximum 8-hour Os values for Houston based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's
represent mean values, whiskers extend up to 1.5x the inter-quartile range from
the boxes, and circles represent outliers 4D-96
Figure 4D-65. Composite monitor daily maximum 8-hour Os values for Los Angeles based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's
represent mean values, whiskers extend up to 1.5x the inter-quartile range from
the boxes, and circles represent outliers 4D-96
Figure 4D-66. Composite monitor daily maximum 8-hour Os values for New York based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's
represent mean values, whiskers extend up to 1.5x the inter-quartile range from
the boxes, and circles represent outliers 4D-97
Figure 4D-67. Composite monitor daily maximum 8-hour Os values for Philadelphia based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's
represent mean values, whiskers extend up to 1.5x the inter-quartile range from
the boxes, and circles represent outliers 4D-97
Figure 4D-68. Composite monitor daily maximum 8-hour Os values for Sacramento based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's
represent mean values, whiskers extend up to 1.5x the inter-quartile range from
the boxes, and circles represent outliers 4D-98
Figure 4D-69. Composite monitor daily maximum 8-hour Os values for St. Louis based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's
represent mean values, whiskers extend up to 1.5x the inter-quartile range from
the boxes, and circles represent outliers 4D-98
Figure 4D-70. Change in VNA estimates of the daily maximum 8-hour average (MDA8) Os
concentrations based on HDDM adjustments in Atlanta 4D-102
Figure 4D-71. Change in VNA estimates of the daily maximum 8-hour average (MDA8) Os
concentrations based on FtDDM adjustments in Baltimore 4D-103
Figure 4D-72. Change in VNA estimates of the daily maximum 8-hour average (MDA8) Os
concentrations based on FtDDM adjustments in Boston 4D-104
Figure 4D-73. Change in VNA estimates of the daily maximum 8-hour average (MDA8) Os
concentrations based on FtDDM adjustments in Chicago 4D-105
4D-xiii
-------
Figure 4D-74. Change in VNA estimates of the daily maximum 8-hour average (MDA8) Os
concentrations based on HDDM adjustments in Cleveland 4D-106
Figure 4D-75. Change in VNA estimates of the daily maximum 8-hour average (MDA8) Os
concentrations based on HDDM adjustments in Dallas 4D-107
Figure 4D-76. Change in VNA estimates of the daily maximum 8-hour average (MDA8) Os
concentrations based on HDDM adjustments in Denver 4D-108
Figure 4D-77. Change in VNA estimates of the daily maximum 8-hour average (MDA8) Os
concentrations based on HDDM adjustments in Detroit 109
Figure 4D-78. Change in VNA estimates of the daily maximum 8-hour average (MDA8) Os
concentrations based on HDDM adjustments in Houston 4D-110
Figure 4D-79. Change in VNA estimates of the daily maximum 8-hour average (MDA8) Os
concentrations based on HDDM adjustments in Los Angeles 4D-111
Figure 4D-80. Change in VNA estimates of the daily maximum 8-hour average (MDA8) Os
concentrations based on HDDM adjustments in New York 4D-112
Figure 4D-81. Change in VNA estimates of the daily maximum 8-hour average (MDA8) Os
concentrations based on HDDM adjustments in Philadelphia 4D-113
Figure 4D-82. Change in VNA estimates of the daily maximum 8-hour average (MDA8) Os
concentrations based on HDDM adjustments in Sacramento 4D-114
Figure 4D-83. Change in VNA estimates of the daily maximum 8-hour average (MDA8) Os
concentrations based on HDDM adjustments in St. Louis 4D-115
Figure 4D-84. Change in VNA estimates of the daily maximum 8-hour average (MDA8) O3
concentrations based on HDDM adjustments in Washington, D.C 4D-116
Figure 4D-85. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Atlanta, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-117
Figure 4D-86. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Atlanta, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-118
Figure 4D-87. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Baltimore, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-119
Figure 4D-88. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Baltimore, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-120
Figure 4D-89. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Boston, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-121
Figure 4D-90. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Boston, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-122
4D-xiv
-------
Figure 4D-91. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Chicago, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-123
Figure 4D-92. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Chicago, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-124
Figure 4D-93. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Cleveland, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-125
Figure 4D-94. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Cleveland, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-126
Figure 4D-95. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Dallas, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-127
Figure 4D-96. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Dallas, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-128
Figure 4D-97. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Denver, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-129
Figure 4D-98. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Denver, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-130
Figure 4D-99. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Detroit, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-131
Figure 4D-100. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Detroit, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-132
Figure 4D-101. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Houston, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-133
Figure 4D-102. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Houston, 2008-2010. The points are colored
4D-xv
-------
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-134
Figure 4D-103. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Los Angeles, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-135
Figure 4D-104. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Los Angeles, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-136
Figure 4D-105. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for New York, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-137
Figure 4D-106. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for New York, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-138
Figure 4D-107. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Philadelphia, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-139
Figure 4D-108. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Philadelphia, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-140
Figure 4D-109. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Sacramento, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-141
Figure 4D-110. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Sacramento, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-142
Figure 4D-111. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for St. Louis, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-143
Figure 4D-112. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for St. Louis, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color
bar were set to the nearest value within the color bar 4D-144
Figure 4D-113. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Washington, D.C., 2006-2008. The points are
colored according to the change in ppb, and values falling outside the range in the
color bar were set to the nearest value within the color bar 4D-145
4D-xvi
-------
Figure 4D-114. Changes in annual 4th highest MDA8 and May-September average MDA8 values
based on HDDM adjustments for Washington D.C., 2008-2010. The points are
colored according to the change in ppb, and values falling outside the range in the
color bar were set to the nearest value within the color bar 4D-146
Figure 4D-115. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Atlanta versus population and
population density 4D-147
Figure 4D-116. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Baltimore versus population and
population density 4D-148
Figure 4D-117. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Boston versus population and
population density 4D-149
Figure 4D-118. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Chicago versus population and
population density 4D-150
Figure 4D-119. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Cleveland versus population
and population density 4D-151
Figure 4D-120. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Dallas versus population and
population density 4D-152
Figure 4D-121. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Denver versus population and
population density 4D-153
Figure 4D-122. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Detroit versus population and
population density 4D-154
Figure 4D-123. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Houston versus population and
population density 4D-155
Figure 4D-124. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Los Angeles versus population
and population density 4D-156
Figure 4D-125. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for New York versus population
and population density 4D-157
Figure 4D-126. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Philadelphia versus population
and population density 4D-158
Figure 4D-127. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Sacramento versus population
and population density 4D-159
Figure 4D-128. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for St. Louis versus population and
population density 4D-160
4D-xvii
-------
Figure 4D-129. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Washington, D.C. versus
population and population density 4D-161
Figure 4D-130. Composite monitor daily maximum 8-hour Os values for Denver based on
observed and adjusted air quality for the NOx-only and NOx/VOC scenarios.
Boxes represent the median and quartiles, x's represent mean values, whiskers
extend up to 1.5x the inter-quartile range from the boxes, and circles represent
outliers 4D-165
Figure 4D-131. Composite monitor daily maximum 8-hour Os values for Detroit based on
observed and adjusted air quality for the NOx-only and NOx/VOC scenarios.
Boxes represent the median and quartiles, x's represent mean values, whiskers
extend up to 1.5x the inter-quartile range from the boxes, and circles represent
outliers 4D-165
Figure 4D-132. Composite monitor daily maximum 8-hour Os values for Houston based on
observed and adjusted air quality for the NOx-only and NOx/VOC scenarios.
Boxes represent the median and quartiles, x's represent mean values, whiskers
extend up to 1.5x the inter-quartile range from the boxes, and circles represent
outliers 4D-166
Figure 4D-133. Composite monitor daily maximum 8-hour Os values for Los Angeles based on
observed and adjusted air quality for the NOx-only and NOx/VOC scenarios.
Boxes represent the median and quartiles, x's represent mean values, whiskers
extend up to 1.5x the inter-quartile range from the boxes, and circles represent
outliers 4D-166
Figure 4D-134. Composite monitor daily maximum 8-hour Os values for New York based on
observed and adjusted air quality for the NOx-only and NOx/VOC scenarios.
Boxes represent the median and quartiles, x's represent mean values, whiskers
extend up to 1.5x the inter-quartile range from the boxes, and circles represent
outliers 4D-167
Figure 4D-135. Composite monitor daily maximum 8-hour Os values for Philadelphia based on
observed and adjusted air quality for the NOx-only and NOx/VOC scenarios.
Boxes represent the median and quartiles, x's represent mean values, whiskers
extend up to 1.5x the inter-quartile range from the boxes, and circles represent
outliers 4D-167
Figure 4D-136. Composite monitor daily maximum 8-hour Os values for Sacramento based on
observed and adjusted air quality for the NOx-only and NOx/VOC scenarios.
Boxes represent the median and quartiles, x's represent mean values, whiskers
extend up to 1.5x the inter-quartile range from the boxes, and circles represent
outliers 4D-168
Figure 4D-137. April-October seasonal average of daily maximum 8-hour Os values at Denver
area monitor locations for observed 2006-2008 conditions (left panel), 75 ppb
adjustment scenarios (middle panels), and 65 ppb adjustment scenarios (right
panels). NOx-only adjustments are shown in top panels, NOx/VOC adjustments
are shown in bottom panels 4D-170
Figure 4D-138. April-October seasonal average of daily maximum 8-hour Os values at Detroit
area monitor locations for observed 2006-2008 conditions (left panel), 75 ppb
adjustment scenarios (middle panels), and 65 ppb adjustment scenarios (right
4D-xviii
-------
panels). NOx-only adjustments are shown in top panels, NOx/VOC adjustments
are shown in bottom panels 4D-171
Figure 4D-139. April-October seasonal average of daily maximum 8-hour Os values at Houston
area monitor locations for observed 2006-2008 conditions (left panel), 75 ppb
adjustment scenarios (middle panels), and 65 ppb adjustment scenarios (right
panels). NOx-only adjustments are shown in top panels, NOx/VOC adjustments
are shown in bottom panels 4D-172
Figure 4D-140. April-October seasonal average of daily maximum 8-hour Os values at Los
Angeles area monitor locations for observed 2006-2008 conditions (left panel), 75
ppb adjustment scenarios (middle panels), and 65 ppb adjustment scenarios (right
panels). NOx-only adjustments are shown in top panels, NOx/VOC adjustments
are shown in bottom panels 4D-173
Figure 4D-141. April-October seasonal average of daily maximum 8-hour Os values at New
York area monitor locations for observed 2006-2008 conditions (left panel), 75
ppb adjustment scenarios (middle panels), and 65 ppb adjustment scenarios (right
panels). NOx-only adjustments are shown in top panels, NOx/VOC adjustments
are shown in bottom panels 4D-174
Figure 4D-142. April-October seasonal average of daily maximum 8-hour Os values at
Philadelphia area monitor locations for observed 2006-2008 conditions (left
panel), 75 ppb adjustment scenarios (middle panels), and 65 ppb adjustment
scenarios (right panels). NOx-only adjustments are shown in top panels,
NOx/VOC adjustments are shown in bottom panels 4D-175
Figure 4D-143. April-October seasonal average of daily maximum 8-hour Os values at
Sacramento area monitor locations for observed 2006-2008 conditions (left
panel), 75 ppb adjustment scenarios (middle panels), and 65 ppb adjustment
scenarios (right panels). NOx-only adjustments are shown in top panels,
NOx/VOC adjustments are shown in bottom panels 4D-176
4D-xix
-------
4D-1. MOTIVATION FOR AN NEW TECHNIQUE TO SIMULATE OZONE
CONCENTRATIONS UNDER ALTERNATIVE STANDARDS
As part of the reviews of the National Ambient Air Quality Standards (NAAQS) for
ozone (Os), EPA estimates health risks after Os has been adjusted to just meet the existing
standard and potential alternative standards. The first draft documents for this review rely upon
the quadratic rollback method used in previous reviews to adjust or "roll back" hourly Os
concentrations in urban areas. Although the quadratic rollback method simulates historical
patterns of air quality changes more closely than some alternative methods (e.g. simply shaving
peak concentrations off at the NAAQS level), its implementation requires some assumptions that
may not always hold true. Specifically, the quadratic rollback method assumes that all monitors
in an urban area will have the same response to emissions changes not allowing for temporally
varying response depending on time of day. In addition, it assumes that O3 concentrations never
increase in response to emissions reductions. However, during NOx-saturated (i.e. VOC limited)
conditions, NOx reductions can result in Os increases (Seinfeld and Pandis, 1998). Finally, since
the quadratic rollback method is purely a mathematical technique and does not account for
physical and chemical atmospheric processes or the sources of emissions precursors that lead to
Os formation, a backstop or "floor" must be used to ensure that predicted Os is not reduced
below "background" concentrations1.
EPA has received comments during past Os NAAQS reviews and during the January 9-
10, 2012 and Sep 11-13, 2013 Clean Air Scientific Advisory Committee (CASAC) meetings for
this Os NAAQS review which encourage the use of alternate methods to quadratic rollback. In
addition, the National Research Council of the National Academies (NRC, 2008) recommended
that EPA explore how emissions reductions might affect temporal and spatial variations in Os
concentrations, and that EPA include information on how NOx versus VOC control strategies
might affect risk and exposure to Os.
Photochemical modeling can simulate the Os response to emission reductions while
avoiding the limitations presented by the quadratic rollback method. While there are
uncertainties inherent in any modeling exercise due to uncertainties in inputs and model
parameters, photochemical modeling provides a more representative characterization of the
spatial and temporal responses of Os to emissions reductions. In this document we present a
1 Background Os has been characterized in previous reviews of the Os NAAQS as "policy relevant background" or
PRB, defined as Os concentrations that would exist in the absence of North American anthropogenic emissions.
In the current review, we have refined the concept of background Os to recognize that there are several possible
definitions of background O3, reflecting both the geographic source of emissions, e.g. U.S., North American,
Global non-U.S., and whether emissions are anthropogenic or natural in origin. In the cases described in this
document, "background" refers to O3 that would exist in absence of U.S. anthropogenic emissions.
4D-1
-------
model-based Os adjustment methodology that allows for adjustments to Os concentrations. This
new approach is firmly rooted in the latest atmospheric modeling science and was peer reviewed
in Simon et al. (2012). This analysis uses EPA's Community Multi-scale Air Quality (CMAQ)
photochemical model (www.cmaq-model.org) instrumented with the Higher order Direct
Decoupled Method (HDDM) - a tool that generates modeled sensitivities of Os to emissions
changes - to estimate the distribution of Os concentrations associated with achievement of the
existing Os standard and alternative standards for multiple urban areas. The HDDM sensitivities
are applied to ambient measurements of Os to estimate how Os concentrations would respond to
changes in U.S. anthropogenic emissions. We use this methodology to estimate Os
concentrations meeting the existing and potential alternative standards in the health risk and
exposure assessment (HREA) and the policy assessment (PA). The photochemical modeling
incorporates emissions from non-anthropogenic sources and anthropogenic emissions from
sources in the U.S and in portions of Canada and Mexico. Pollution from sources in other
locations within and outside of North America is included as transport into the boundary of the
modeling domain. Because the application of the model-based approach focuses on reductions
in U.S. anthropogenic emissions while holding constant those emissions that influence U.S.
background, all changes in Cb will be relative to U.S. background. This does not mean that
background Os concentrations will be constant between recent ambient Os conditions and after
just meeting the existing or alternative standard levels, because of nonlinearities in the formation
ofOs.
4D-2. HIGHER-ORDER DECOUPLED DIRECT METHOD (HDDM)
4D-2.1 Capabilities
Chemical transport models, such as CMAQ, simulate the effects of physical and chemical
processes in the atmosphere to predict 3-D gridded pollutant concentrations (Foley et al., 2010,
Appel et al., 2008, Appel et al., 2007, Byun and Schere, 2006). These models account for the
impacts of emissions, transport, chemistry, and deposition on spatially and temporally varying
pollutant concentrations. Required model inputs include time-varying emissions and
meteorology fields, time varying concentrations of pollutants at the boundaries of the model
domain (i.e. boundary conditions), and a characterization of the 3-D field of chemical
concentrations to initialize the model (i.e. initial conditions).
Beyond modeling the concentrations of ambient Os, chemical transport models can be
used to estimate the response of ambient Os concentrations to changes in emissions. One
technique to simulate the response of Os to emissions changes, the brute force method, requires
the modeler to explicitly model this response by directly altering the emissions inputs in the
model simulation. This technique provides an estimate of the Os concentration at the altered
4D-2
-------
emission level, but often does not provide accurate information regarding the response of Os to
other levels of emissions since the chemistry for Os formation is nonlinear. Therefore, when
using only brute force techniques a new model simulation would need to be performed for every
emissions scenario under consideration.
Other analytical techniques have been developed to estimate the Os response to emissions
changes without performing multiple simulations. One such method is termed the Decoupled
Direct Method (DDM) (Dunker, 1984). DDM, solves for sensitivity coefficients which are
defined as the partial derivative of the atmospheric diffusion equations that underly the model
calculations, Equations (4D-1) and (4D-2).
Equation(4D-l)
_
Equation (4D-2)
Here, Sij(t), the sensitivity, gives the change in model concentration, Ci, (for instance Os
concentration) with an incremental change in any input parameter, pj (in this case emissions).
Equation (4D-2) allows us to normalize the sensitivity coefficient, Sij(t), so that it shows
response in relative terms for the input rather than in absolute units. Therefore, Pj (x,t) is the
normalized input and Sj is a scaling variable (Yang et al., 1997). In general terms, the sensitivity
coefficient tells us how a model output (Os concentration) will change if a model input
(emissions of NOx or VOC) is varied. This first order sensitivity coefficient, Sij(t) is quite
suitable for small changes, but gives a linear response which is unlikely to represent the results
of large changes in very nonlinear chemical environments. Second (and third) order derivatives
can be calculated to give higher order sensitivity coefficients (Hakami et al., 2003). Higher order
sensitivity coefficients give the curvature and inflection points for the response curve and can
capture the nonlinearities in the response of Os to emissions changes. Using Higher order DDM
(HDDM) allows for the sensitivities to be more appropriately applied over larger emissions
changes. Hakami et al. (2003) report that for an application in California, HDDM gave
reasonable approximations of Os changes compared to brute force results for emission changes
up to 50% using the first three terms of the Taylor series expansion, Equation (4D-3).
4D-3
-------
= c(o).
Equation (4D-3)
Here Ae represents the relative change in emissions (for instance Ae = -0.2 would be
equivalent to reducing emissions by 20%), Sn(0) is the nth order sensitivity coefficient, C(0) is the
concentration under baseline conditions (no change in emissions) and Rn+i is a remainder term.
A variant of DDM called DDM-3D has been implemented into several chemical transport
models, including CMAQ, for both Os and particulate matter (PM) predictions (Dunker, 1984;
Yang et al., 1997; Hakami et al., 2003; Cohan et al., 2005; Napelenok et al., 2006; Koo et al.,
2010; Zhang et al., 2012). These implementations allow the modeler to define the parameters for
which first and higher order sensitivities will be calculated. For instance, the sensitivity can be
calculated for emissions from a specific source type, for emissions in a specific geographic
region, and for emissions of a single Os precursor or for multiple Cb precursors. In addition,
sensitivities can be calculated to boundary conditions, initial conditions, and various other model
inputs. Sensitivities to different sets of parameters can be calculated in a single model simulation
but computation time increases as the number of sensitivities increases. Outputs from an HDDM
simulation consist of time varying 3-D fields of first and second order sensitivities.
For the purposes of the Os NAAQS analysis, HDDM provides an improved approach
compared to existing quadratic rollback techniques for several reasons. First, it captures non-
linearity of Os response to emissions changes, representing both increases and decreases in Os
concentrations resulting from emissions reductions. Second, HDDM characterizes different Os
response at different locations (downtown urban versus downwind suburban) and at different
times of day allowing us to incorporate temporal and spatial variations in response into the Os
adjustment methodology. Finally, HDDM eliminates the need to use "background" Os as a floor
for rollback since predicted sensitivities are based on model formulations that explicitly account
for background sources.
4D-2.2 Limitations
In addition to the many potential benefits of using HDDM to understand and characterize
Os response to emissions changes, there are several limitations. First, HDDM encompasses all
of the uncertainties of the base photochemical model formulation and inputs. So uncertainties in
how the physical and chemical processes are treated in the model and in the model inputs
propagate to the HDDM results. Also, HDDM can capture response to larger emissions changes
than DDM but it is still most accurate for small changes. The larger the relative change in
emissions, the less likely that the HDDM sensitivities will properly capture the change in Os that
4D-4
-------
would be predicted by a brute force model simulation. Several studies have reported reasonable
performance of HDDM for Os up to 50% emission changes (Hakami et al., 2003; Cohan et al.,
2005; Hakami et al., 2004), but the magnitude of change over which HDDM will give accurate
estimates will depend on the specific modeling episode, size of the model domain, emissions and
meteorological inputs, and the size of the emissions source to which the sensitivity is being
calculated. In this work, we applied sensitivities derived from model simulations done under
varying NOx levels (see Section 4D-4D-3.2.1) and found that using this technique we were able
to replicate brute force estimates using HDDM sensitivities for up to 90% NOx reductions with a
mean bias of less than 3 ppb and a mean error of less than 4 ppb.
4D-3, APPLYING HDDM/CMAQ TO ADJUST OZONE TO JUST MEET EXISTING
AND ALTERNATIVE STANDARDS: METHODOLOGY
4D-3.1 Conceptual Framework
This section outlines the methodology from Simon et al. (2012) in which we apply
CMAQ/HDDM to estimate hourly Os concentrations that might result from meeting the existing
and potential alternative standards. As part of the methodology, photochemical modeling results
are not used in an absolute sense, but instead are applied to modulate ambient measurements,
thus tying estimated Os distributions to measured values. The basic steps are outlined below and
in Figure 4D-1 Details are given in section 4D-4D-3.2. The inputs, set-up, and evaluation of the
modeling system are described in Appendix 4B.
• Step 1: Run CMAQ simulation with HDDM to determine hourly Ch sensitivities to NOx
emissions and VOC emissions for the grid cells containing monitoring sites in an urban
area.
Step 2: For each monitoring site, season, and hour of the day use linear regression to
relate first order sensitivities of NOx and VOC (SNOX and Svoc) to modeled Os and second
order sensitivities to NOx and VOC (S2NOXand S2voc) to the first order sensitivities.
Step 3: For each measured hourly Os value, calculate the first and second order
sensitivities based on monitoring site-, season-, and hour-specific functions calculated in
Step 2.
• Step 4: Adjust measured hourly 2006-2010 Os concentrations for incrementally
increasing levels of emissions reductions using assigned sensitivities and then recalculate
2006-2008 and 2008-2010 design values until all monitors in an urban area just meet the
existing and potential alternative levels of the standard.
4D-5
-------
I "
Natural
J
I
Anthropogenic
Canada and |
Mexico |
I
O3 and O3 Precursor Emissions i
Recent Monitored O3
(2006-2010)
Step 3:
Use Regressions and
Observed Ozone to
Predict Sensitivities
f Hourly Ozone \
/ Observations Pai red with
\ Sensitivities for 2006-2010 J
\AtAII Monitor Locations J
)
f Unique L
/ Relationships
Sensitivities a
Location fo
\ Season
\ Hour-of-th
f
Meteorology
nitial and Boundary conditions
Other Model Inputs
CMAQ: /^Gridded hourly O3 A
HDDM Modeling \ Concentrations and I— »
(Jan, Apr-Oct 2007) \ Sensltlvltles J
near \
between \
nd Hourly \
^^ Create Regressions
e-Day ^/
\
/Select Emissions^ Step 4:
^/ Reductions to ^v . Adjust Hourly Ozone
N. which Sensitivities / to Meet Alternate
\Will Be Applied^/ Standards
Steplb:
Extract Output
E
f Ozone Concentrations ^
/ & Sensitivities at \
Locations of
Monitoring Sitesfor 1
VEach Modeled Hour /
(Jan, Apr-Oct 2007) ^/
/Zdi
f Val
V Sf
VA
usted Hourly Ozone\
jes for 2006-201 Oat \
h Monitor Location to
owAttainmentwith J
ternate Standards _/
Figure 4D-1. Flow diagram demonstrating DDM model-based Os adjustment approach.
4D-3.2 Application to Measured Os Concentrations in 15 Urban Areas
In this analysis, we apply the model-based adjustment approach described above for just
meeting the existing standard and three potential alternative standards. The analysis covers the
15 urban study areas listed in HREA Chapter 4 using photochemical modeling for January, and
April-October of 2007 and ambient data for the years 2006-2010. When running CMAQ with
HDDM, additional information is required to designate model inputs for calculating sensitivities.
In this analysis, HDDM was set up to calculate the sensitivity of Os concentrations to U.S.
anthropogenic NOX and VOC emissions.2 US anthropogenic emissions were defined as all
emissions in the following sectors: commercial marine and rail, onroad mobile, offroad mobile,
Electric Generating Unit (power plant) point sources, non-EGU point, and non-point area. These
sectors accounted for 17.6 million of the total domain-wide 23.6 million tons per year of NOx
emissions and 13.8 million of the total domain-wide 68.1 million tons per year of VOC
emissions. Sensitivities were not determined for biogenic, fire, Canadian, or Mexican emissions.
In addition, sensitivities were not calculated for any emissions originating from outside the
domain (i.e. entering through the use of boundary concentrations).
2 Sensitivities only tracked U.S. emissions in the contiguous 48 states.
4D-6
-------
4D-3.2.1 Multi-step Application of HDDM Sensitivities
As discussed in Section 4D-4D-2.2 of this appendix, HDDM has been reported to
reasonably replicate brute force emissions reductions up to a 50% change in emissions. For this
analysis, it was desirable to have confidence that the HDDM sensitivities could replicate the
entire range of emissions reductions. Evaluations of the HDDM estimates compared to brute
force emissions reduction model runs confirm that the HDDM estimates of Os response to NOx
reductions are fairly comparable for a 50% change. However, HDDM and brute force estimates
begin to diverge in comparisons under larger emissions changes (90%). Consequently four
additional CMAQ/HDDM runs were performed under different levels of NOx and VOC
emissions reductions in order to characterize Os sensitivities to NOx reductions over a larger
range of emissions changes. One CMAQ/HDDM simulation was performed with U.S.
anthropogenic NOx cut by 50%. A second additional simulation was performed with a 90% NOx
reduction. Another set of CMAQ/HDDM simulations were performed for simultaneous NOx and
VOC cuts (50% NOx and 50% VOC; 90% NOx and 90% VOC). Emissions of other species were
not modified from the base case in these four simulations. These additional HDDM simulations
give Os sensitivities to NOx and VOC under conditions with lower NOx (or NOx and VOC)
emissions in the U.S. These sensitivities are used in a multistep adjustment approach.
Figure 4D-2 gives a conceptual picture of the multistep adjustment procedure using first-
order sensitivities. Sensitivities from the base run are used to adjust Os concentrations for NOx
emissions reductions up to X%. Additional emission reductions beyond X% use sensitivities
from the 50% NOx cut run until reductions exceed (X+Y)%. Finally, sensitivities from the 90%
NOx cut run are applied for any emission reductions beyond (X+Y)%. In order to more closely
approximate the non-linear Os response to any level of emissions reductions, 2nd order terms are
added to the multistep approximation method in Equations (4D-4) to (4D-7). P represents the
percentage NOx cut for which the AOs values are being calculated, S and S2 are the first and
second order Os sensitivities to U.S. NOx emissions, and X and Y are described above.
Alternately, Equation (4D-8) can be used with Equations (4D-5) to (4D-7) if simultaneous NOx
and VOCc cuts are being simulated. Note that the 50% and 90% cut sensitivities used in
Equation (4D-8) come from model runs which cut both VOC and NOx by these percentages in
contrast to the NOx cut only sensitivities that are used in Equation 4D-4. Consequently,
Equation (4D-8) works only for equal percentage cuts in both NOx and VOC emissions. In
equations (4D-5) to (4D-8), we cap A03, so that Os never drops below zero due to emissions
changes.
4D-7
-------
- -o x $NoXjHat + — x S.vojrfraf, - 6 x Sv0jcso%cur + — x
9
c*
—CX 5>Vx%cur + ~T x S.vojreo Heat
Equation (4D-4)
a = •
— forP-
:::
-a x Svojrbait + — x 5jWfraj([ - a X
h^ fc^
-fe X 5ivojr%u +~ x ^.VjtHur — 6 X 5l»O!. + X
I :
— c X J,varM%cut "*" 2 X ^°j:io^''r ~ c x ^Wft»*enl "*" 5 X ^oc£o'*«'>
4D-8
Equation (4D-5)
Equation (4D-6)
0 for P < (X + V)
V)
Equation (4D-7)
Equation (4D-8)
-------
I NOx first-order approximation from base DDM simulation
I NOx first-order approximation from 50% NOx cut DDM simulation
I NOx s first-order approximation from 90% NOx cut DDM simulation
Theoretical ozone response curve to NOx emission reductions
( CMAQ DDM simulations
Estimated
concentration!
after P%
reduction in
baseline NOx
emissions
C
o
4-J
CO
•M
C
OJ
u
C
o
u
OJ
C
o
M
o
100 P 90 X+Y 50 X
% reduction in baseline NOx emissions
Figure 4D-2. Conceptual picture of 3-step application of HDDM sensitivities.
Starting concentration
The ideal value for equation transition points, X and Y, are determined by minimizing the
least square mean error between the adjusted concentrations using the multistep approach and
modeled concentrations from brute force NOx cut runs. We first determined the value of X
which gave the lowest error compared to brute forces estimates at 50% NOx cuts. Then holding
X constant, we determined the value of Y which gave the lowest error compared to brute force
estimates at 90% NOx cuts. This process was performed independently for each of the 15 urban
areas in this analysis.
Error in HDDM estimates of hourly Os is defined here as the difference between HDDM
estimates and brute force Os. Based on Equations (4D-4) to (4D-7), this can be calculated from
Equations (4D-9) and (4D-10) for 50% NOx cuts:
f =
SF so
Equation (4D-9)
4D-9
-------
-X
100 X 5-w*>«* 2X1QQ1
2(50-1")
(2 X (SO -*))" »
2X1001 X SJ
Equation(4D-10)
Equation (4D-10) can be rearranged to appear in the form: AX2 + BX + C:
/5ig^t -*X^'QJM«yur\ : f-Sxox*.. SX-S.va^Hfut ___ 400 X 5|0r^fcmr\
£ 2X100* 2X100- / \ 100 100 2x100*
2xi00
„ ,. ^n,
Equation(4D-ll)
Equation(4D-12)
Equation(4D-13)
Equation(4D-14)
Next, the error is squared, summed over all points (error can be calculated for each
hourly Os value at each monitoring location), and the derivative is set to 0 to determine X which
gives the least squares error (Equations (4D-15), (4D-16), and (4D-17)).
(2AC + S:) J: + 2.BCX + C:
=0
Equation(4D-15)
Equation(4D-16)
Equation(4D-17)
The value of X that gives the least squares error will occur at one of the 3 roots of the
trinomial in Equation (4D-17) or at 0 or 50. All real roots, 0, and 50 were input into Equation
(4D-16) and X was set to the value which resulted in the lowest error in each city. An analogous
procedure was followed to determine Y using the 90% NOX cut brute force simulation and
Equations (4D-18) through (4D-24).
4D-10
-------
-X
100
X"
2 X
2F
100
10(90- (AT +y))
2 = A2Y* + 2ABYJ+ (2AC + B:)K: + 2BCY + C2
.
SI —
_ _ /-^J.va*»w« 10xJ.ve»»tfi« _ 1&0>;i:9g-x)s«te»-«ut\
~ I, ioo 100 :xioss /
Equation(4D-18)
Equation(4D-19)
Equation (4D-20)
Equation(4D-21)
oHcut - bOzmssF.w
Equation (4D-22)
S2)F:
Equation (4D-23)
i'*S + (3 E245)F: + (2 I2.4C + S*)F + (E2SC) = 0
Equation (4D-24)
This methodology can also be adjusted to calculate least squares error cut points for the
combined NOX and VOC emissions reduction case. X and Y cut points which have the least
square error in each urban area are shown in Table 4D-1. This 3-step adjustment methodology
was shown to be a robust method for minimizing error in the HDDM applications for larger
percentage changes in emissions by Simon et al. (2012). Figure 4D-3 through Figure 4D-6 are
density scatter plots that compare hourly Os estimates from brute force with hourly Os estimates
from the 3-step HDDM adjustments at all monitor locations in each of the 15 urban areas
evaluated in this study. The colors in these plots depict the percentage of points falling at any
one location. Mean error for the 50% and 90% 3-step HDDM adjustment NOX cut cases
compared to brute force results are less than 1 ppb and 4 ppb respectively in all 15 case study
areas.
4D-11
-------
Table 4D-1. X and Y cut points used in Equations (4D-4) and (4D-8). Note: the NOx/VOC
sensitivity case was only performed in two cities.
Urban Area
Atlanta
Baltimore
Boston
Chicago
Cleveland
Dallas
Denver
Detroit
Houston
Los Angeles
New York
Philadelphia
Sacramento
Saint Louis
Washington D.C.
NO, cuts only
X
40
40
39
38
42
37
39
38
37
38
37
40
38
39
40
Y
46
46
45
45
45
46
45
45
44
46
45
45
45
45
46
NOX and VOC
combined cuts
X
40
43
Y
46
45
4D-12
-------
Atlanta Area Sites
Baltimore Area Sites
Boston Area Sites
* s-
Y = 0.24 +0.97 *X
MB = -0,9 ppb
ME = 1 ppb
0 20 40 60 80 100 140
O3 at 50% NOx cut - Brute Force
Chicago Area Sites
- Y = 0.43 + 0.98 * X
Q
S
-1-
MB =-0,2 ppb
ME = 1 ppb
! ! I ! ! i !
' 20 40 60 80 100 140
O3 at 50% NOx cut - Brute Force
Denver Area Sites
°-l Y = 0.24+ 0.99" X
§-
MB = -0,4 ppb
ME = 0,3 ppb
\ \ \ \ \ : \
0 20 40 60 80 100 140
O3 at 50% NOx cut - Brute Force
5
4
3
- 2
- 1
- 5
- 4
3
2
- 6
- 5
- 3
- 2
- 1
Is-
Y = 0.77 + 0.97 * X
MB = -0.5 ppb
ME = 1 ppb
0 20 40 60 80 100 140
O3 at 50% NOx cut - Brute Force
Cleveland Area Sites
s,
X <
o
z
a? SH
Y = 0.82 + 0.97 ' X
MB = -0.4 ppb
ME = 1 ppb
! I ; ! !
0 20 40 60 80 100 140
O3 at 50% NOx cut - Brute Force
Detroit Area Sites
- Y = 0.62 + 0.97 " X
g-
B-.
8? S-
MB = -0.3 ppb
ME = 1 ppta
1 I : i I i
0 20 40 60 80 100 140
OS at 50% NOx cut - Brute Force
5
4
3
2
- 1
-s s_
Y = 0.93 +0.97'X
MB = -0,Zppb
ME = 0.7 ppb
0 20 40 60 80 100 140
03 at 50% NOx cut - Brute Force
Dallas Area Sites
Y = 0.27 +0.97 *X
2 S-
Q "~
a °
-1- i-i.
* s-
-------
Houston Area Sites
Los Angeles Area Sites
New York Area Sites
i- Y = 0.87+ 0.97* X
Q "~
g o
-1- I—I ,
MB = -0.3 ppb
ME = 1 ppb
0 20 40 60 80 100 140
O3 at 50% NOx cut - Brute Force
Philadelphia Area Sites
°- Y = 0.86+ 0.97* X
S? g-
MB- -0.3 ppb
ME = 1 ppb
n i i i i i r
0 20 40 60 80 100 140
OS at 50% NOx cut - Brute Force
Washington D.C. Area Sites
^- Jrf
Q "~
??-
8j
X c
o
Y = 0.72 + 0.97 ' X
MB = -0.5 ppb
ME = 1 ppb
\ I I I I i
0 20 40 60 80 100 140
OS at 50% NOx cut - Brute Force
- 4
3
2
•\
5
A
3
Z
- 1
8-.
Y = 0.49 + 0.99 * X
I
140
0 20 40 60 80 100
OS at 50% NOx cut - Brute Force
Sacramento Area Sites
1-1 Y = 0.32 + 0.99 * X
MB = 0.02 ppb
ME = P.4 ppb
1 I I I I I
0 20 40 60 80 100 140
OS at 50% NOx cut - Brute Force
MB = -0,1 ppb
ME = 1 ppb
0 20 40 60 80 100 140
OS at 50% NOx cut - Brute Force
Saint Louis Area Sites
1-1 Y = 0.22 + 0.98 * X
MB - -0.5 ppb
ME = 1 ppb
I I I I I I T
0 20 40 60 80 100 140
OS at 50% NOx cut - Brute Force
Figure 4D-4. Comparison of brute force and 3-step HDDM Os estimates for 50% NOx cut
conditions: Houston, Los Angeles, New York, Philadelphia, Sacramento, St. Louis, and
Washington D.C.
4D-14
-------
Atlanta Area Sites
Baltimore Area Sites
Boston Area Sites
Y = -2.3 + 0.97 * X
MB = -3 ppb
= 4ppb
0 20 40 60 80 100 140
OS at 90% NOx cut - Brute Force
Chicago Area Sites
s? g-
Y = 2.9 +0.87 "X
MB = -2 ppb
ME = 3 ppb
1 I I I I i I
0 20 40 60 80 100 140
OS at 90% NOx cut - Brute Force
Denver Area Sites
Y = 1.1+0.94*X
Q "~
9 =
-1- r-i .
3
X
O
MB = -1 ppb
ME = z ppb
1 I I I I i
0 20 40 60 80 100 140
OS at 90% NOx cut - Brute Force
- 4
- 3
- 2
- 1
- 5
5
4
- 3
- 2
- 1
Q "~
e °
-1- i-i.
«•-•
Y = 1.5-f 0.9* X
MB = -2 ppb
ME = 3 ppb
0 20 40 60 80 100 140
OS at 90% NOx cut - Brute Force
Cleveland Area Sites
Y = 0.82 + 0.93 * X
s-
MB = -2 ppb
ME = 3 ppb
1 I I I I I
0 20 40 60 80 100 140
OS at 90% NOx cut - Brute Force
Detroit Area Sites
2 SH
Q "~
« s-l
CO
O
Y = 1.7 +0.89 *X
MB = -2 ppb
ME = 3 ppb
I I I I I i i
20 40 60 80 100 140
OS at 90% NOx cut - Brute Force
« 8-1
CO
O
MB = -1 ppb
ME = 2 ppb
0 20 40 60 80 100 140
OS at 90% NOx cut - Brute Force
Dallas Area Sites
Y = 1.1+0.87* X
M0 = -3 ppb
ME= 4 ppb
i i i i i i r
0 20 40 60 80 100 140
O3 at 90% NOx cut - Brute Force
- 4
- 3
2
1
Figure 4D-5. Comparison of brute force and 3-step HDDM Os estimates for 90% NOx cut
conditions: Atlanta, Baltimore, Boston, Chicago, Cleveland, Dallas, Denver, and Detroit.
4D-15
-------
Houston Area Sites
Los Angeles Area Sites
New York Area Sites
Q "~
9 =
-1- f-< .
15 -l
to
o
Y = 0.99 + 0.91 *X
MB =-2 ppb
ME = 4 ppb
T I I I I I
0 20 40 60 80 100 140
O3 at 90% NOx cut - Brute Force
Philadelphia Area Sites
2 ?H
Q "~
V = 1.9 +0.89 "X
-2ppb
ME = 3 ppb
1I I I I I
0 20 40 60 80 100 140
O3 at 90% NOx cut - Brute Force
Washington D.C. Area Sites
°-| V = 0.81+0.91'X
Q "~
9 °
-1- i—i.
MB = -2 ppb
ME = 3 ppb
III
0 20 40 60 80 100 140
O3 at 90% NOx cut - Brute Force
- 3
- 2
- 1
2
Q
9
gs-
01
IS o
MB = -D.9 ppb
ME = 2 ppb
~1IIIIIT
0 20 40 60 80 100 140
O3 at 90% NOx cut - Brute Force
Sacramento Area Sites
Y = 1.1 +0.96'X
D.3ppb
ME = 1 ppb
I I I I I I I
0 20 40 60 80 100 140
O3 at 90% NOx cut - Brute Force
- 6
5 1^
4 5
X
3 g
-2 g
- 1 o
-2
Y = 3.3 +0.85 *X
MB - -2 ppb
ME = 4 ppb
\ I I I I I
0 20 40 60 80 100 140
O3 at 90% NOx cut - Brute Force
Saint Louis Area Sites
I I I I I
0 20 40 60 80 100 140
O3 at 90% NOx cut - Brute Force
Figure 4D-6. Comparison of brute force and 3-step HDDM Os estimates for 90% NOx cut
conditions: Houston, Los Angeles, New York, Philadelphia, Sacramento, St. Louis, and
Washington D.C.
4D-3.2.2 Relationships between HDDM Sensitivities and Modeled Os
Concentrations
First and second order hourly Os sensitivities to VOC and NOx were extracted from the
HDDM simulation for model grid cells that contained the Os monitors in the 15 urban areas.
Extracted data included modeled sensitivities at monitor locations for all modeled hours in 2007.
These sensitivities cannot be applied directly to observed values for two reasons: 1) high
4D-16
-------
modeled Os days/hours do not always occur concurrently with high observed Os days/hours and
2) the modeling time period includes 8 months in 2007 but the time period we are analyzing in
this HREA includes five full years of ambient data, 2006-2010. As to the first point,
photochemical models are generally used in a relative sense for purposes of projecting design
values to assess just meeting the NAAQS standard. In this manner, model predictions are
"anchored" to measured ambient values. In general, the average response on high modeled days
is used for this purpose. This allows for more confidence in calculated results when "less than
ideal model performance [occurs] on individual days" (U.S. EPA, 2007). Similarly, for this
analysis we believe it is appropriate to account for the fact the model does not always perfectly
agree with measurements and that sensitivities from a low Os modeled day would not be
appropriate to apply to a high Os measured day (and vice-versa) even if they occur on the same
calendar day. For the second point, due to current resource and time constraints we were only
able to model 8 months in 2007. However, the Os exposure analysis evaluates the effects of Os
decreases for two 3-year periods in 2006-2008 and 2008-2010 most of which is outside the
modeled time period. For both of these reasons, a method was developed to generalize the
modeled site-, season-, and hour-specific sensitivities so that they could be applied to ambient
data during 2006-2010.3
Simon et al. (2012) describe how first order sensitivities are generally well correlated
with hourly modeled Os concentrations and second order sensitivities are well correlated to first
order sensitivities. Based on their analysis, we create a separate linear regression for SNOX and
Svoc as functions of hourly Os (i.e. SNOX = mxCb + b) for every site, season4, and hour-of-the day
examined in this analysis. For instance, for summer 8-am hours at Detroit Site 260990009, SNOX
and Os values from all 8 am hours in June-August 2007 are used to fit this relationship. Since
only 8 months were modeled (Jan, April-October) regressions for summer season relationships
include more data points (92) than those for winter (31), spring (61), and fall (61). Similarly,
S2NOx and S2voc were calculated as a function of SNOX and Svoc respectively and S2NOx,voc was
calculated as a linear combination of both SNOX and Svoc. Figure 4D-7, Figure 4D-8, and Figure
4D-9 show examples of these regressions for first order NOx sensitivities for a NOx-limited site
(summertime, downwind of Atlanta), a NOx-saturated site (autumn, Queens NY), and a
transitional site which switches between chemical regimes (spring, Long Island NY). Example
3 The 8 months modeled covered a variety of conditions such that we can use the results from this modeled time
period in conjunction with the ambient data from the longer 5-year period for estimating response and applying
adjustments
4 Here seasons are defined as follows:
For ambient data, Winter = December, January, February; Spring = March, April, May;
Summer = June, July, August; Fall = September, October, November.
For modeled data, Winter = January; Spring = April, May; Summer = June, July, August;
Fall = September, October
4D-17
-------
relationships are shown for four different times of day with different Cb response behavior:
nighttime (1:00 LST), mid-afternoon (15:00 LST), morning rush-hour (8:00 LST), and evening
rush-hour (18:00 LST). The Atlanta area NOx-limited site (Figure 4D-7) generally showed
positive sensitivities which would lead to decreasing Os with decreasing emissions. However
some limited Os increases with NOx reductions from NOx (negative sensitivities) occurred at this
site during nighttime and rush-hour times. The Queens site had negative sensitivities at these
four hours on almost all days indicating strongly NOx saturated conditions. The slopes from the
regressions at the Queens site were negative indicating larger Os increases are high
concentrations. Finally, the Long Island site had both positive and negative sensitivities but the
slope was positive resulting in Os increases for low basecase concentrations but Os decreases at
higher basecase concentrations. Correlations were strongest at the NOx limited site and during
rush-hour and nighttime periods for the other sites. Figure 4D-10, Figure 4D-11, and Figure
4D-12 show these regressions from the same sites and time-periods for second order sensitivities.
Comparisons between brute force and HDDM Os estimates shown in Figure 4D-5 and
Figure 4D-6, demonstrate that for the vast majority of data points, HDDM replicates brute force
with minimal errors. However, these figures show a small number of instances in which HDDM
predicts very high hourly Os (> 100 ppb) while the brute force emissions simulations show much
lower Os (< 40 ppb). In such cases, base modeled Os is low due to NOx titration and increases
occur with reductions of NOx. The HDDM sensitivities for these few points appear to be too
high to be applied over large emissions changes because of strongly nonlinear chemistry. While
the brute force simulations predict modest increases in Os at these hours, the HDDM method
predicts very large increase in Os at these hours. To avoid outlier sensitivities from biasing this
analysis, we set a floor value for negative SNOX values (Os increases) in each regression so that
SNOX never drops below the 5th percentile modeled SNOX (for each site, hour, season grouping)
when negative SNOX values are simulated (see dotted blue lines in Figure 4D-7, Figure 4D-8, and
Figure 4D-9). In addition, SNOX is prevented from dropping below 0 when all modeled SNOX
values were positive (i.e. no modeled increases in Os). These floor values have the most
influence when regressions with highly NOx-saturated conditions, such as those shown for
Queens NY, are applied to ambient Os values that are higher than the modeled range. For
instance in Figure 4D-9, if there were any measured autumn 18:00 Os concentrations above 30
ppb at that site in 2006-2010, the floor would prevent the extrapolation of Os increases from the
modeled time period to conditions that do not apply. This floor is necessary for Os increases
which, if extrapolated, could unrealistically increase Os without a bound. In contrast, the Os
decreases can never exceed the total measured Os so positive NOx sensitivities have a built-in
upper bound.
4D-18
-------
For the 50% and 90% emissions cut CMAQ/HDDM simulation, regressions were
performed for first order NOx and VOC sensitivities with modeled Os from the base HDDM
simulation. The regression technique was performed for the first and second order NOx and VOC
sensitivities from the base run and the 50% emissions cut and 90% emissions cut simulations.
The sensitivities from the emissions cut runs were fitted to hourly Os concentrations in the base
simulation. Simon et al. (2012) found that correlation coefficients using for sensitivities from
NOx cut simulations to base case Os concentrations were similar to those with Os concentrations
from the NOx cut runs.
Atlanta Site 131510002, summer, 1:00
Atlanta Site 131510002, summer, 15:00
o
CO
0-0
5
z o
o o
cor = 0.8
0 10 20 30 40 50 60
hourly 05 (ppb)
Atlanta Site 131510002, summer, 8:00
50. _ ,100
16)
hourly O5 (ppb)
150
Atlanta Site 131510002, summer, 18:00
I I I
0 10 20
u__, 30 , 40 50 60
hourly 05 (ppb)
i i i i i i r
0 20 40 60 80 120
hourly O5 (ppb)
Figure 4D-7. Relationship between SNOX and hourly Os at a NOx-limited site downwind of
Atlanta (summer). Relationships are shown for one nighttime hour, one morning rush-
hour hour, one daytime hour, and one evening rush-hour hour. The solid blue line show
the linear regression for these points and the dotted blue line shows the floor value used for
SNOX based on the 5th percentile modeled value.
4D-19
-------
NV Site 360810098, autumn, 1:00
NV Site 360810098, autumn, 15:00
I I I ! I I I
0 5 10 15 20 25 30
hourly 03 (ppb)
NV Site 360810098, autumn, 8:00
i
[I 20 40 , .,60 80
TiourlyO,
! I !
I 40 6
ourly Q3 (ppb)
NV Site 360810098, autumn, 18:00
'' 1
-------
NV Site 361030004, spring, 1:00
NV Site 361030004, spring, 15:00
1=1
1^
.Q
Q-
Q-
W
O
^r
i i i i i
0 10 20 30 40 50 60 70
nourfy 03 (ppb)
NV Site 361030004, spring, 8:00
Q-
Q-
0 10 u 20, 30, 40 50
hourly Q3 (ppb)
cor = 0.5
i i i
) 20 40 6L,
hourly 03 (ppb)
NV Site 361030004, spring, 18:00
~\
4.0 _ 60 ..80
»(PI
100
:c ">
o
hourly O3 (ppb)
80
Figure 4D-9. Relationship between SNOX and hourly Os at a NOx-saturated site in Suffolk
County, NY on Long Island (spring). Relationships are shown for one nighttime hour, one
morning rush-hour hour, one daytime hour, and one evening rush-hour hour. The solid
blue line show the linear regression for these points and the dotted blue line shows the floor
value used for SNOX based on the 5th percentile modeled value.
4D-21
-------
Atlanta Site 131510002; summer, 1:00
Atlanta Site 131510002, summer, 15:00
Q-
o.
UD
CD -
cor = 0.6
o
G
cor = 0.3
-10
SNOK (PP&)
i I
-40 -20
i
SNOK
i
40
i
60
Atlanta Site 131510002, summer, 8:00
Atlanta Site 131510002, summer, 18:00
o
o -
OJ
.Q CD
Q_ CD -
Q. i—
:•:
O
M 2
CO o
o
o _
i I
-30 -20 -10,
SNOK
i
10
I
20
co- = 0.75
-40 -20
i
0
NOX
(PP
20
PD)
i
40
Figure 4D-10. Relationship between »V2Arox and SNOX at a NOx-limited site downwind of
Atlanta (summer). Relationships are shown for one nighttime hour, one morning rush-
hour hour, one daytime hour, and one evening rush-hour hour. The solid blue line shows
the linear regression for these points.
4D-22
-------
NV Site 360810098 autumn, 1:00
NV Site 360810098 autumn, 15:00
0
S
NV Site 360810098 autumn, 8:00
NV Site 360810098 autumn, 18:00
I I I I I
-100 -80 -60 -40 -20
SNOH (PPD)
Figure 4D-11. Relationship between S2NOx and SNOX at a NOx-saturated site in Queens
County, NY (autumn). Relationships are shown for one nighttime hour, one morning rush-
hour hour, one daytime hour, and one evening rush-hour hour. The solid blue line shows
the linear regression for these points.
4D-23
-------
NV Site 361030004 spring, 1:00
NV Site 361030004spring, 15:00
Q.
Q-
NV Site 361030004 spring, 8:00
NV Site 361030004spring, 18:00
.C:
Q-
O
'XI
Figure 4D-12. Relationship between S2NOx and »SWo* at a NOx-saturated site in Suffolk
County, NY on Long Island (spring). Relationships are shown for one nighttime hour, one
morning rush-hour hour, one daytime hour, and one evening rush-hour hour. The solid
blue line shows the linear regression for these points.
4D-3.2.3 Application of Sensitivity Regressions to Ambient Data
To apply the HDDM adjustments to observed data, sensitivities must be determined for
each hour from 2006-2010 at each site based on the linear relationship from the modeled data
and the observed Os concentration. The linear regression model also allows us to quantify the
standard error of each predicted sensitivity value at each hour and site.
Observed hourly Os from 2006-2010 at each monitor location was adjusted by applying
incrementally increasing emissions reductions using Equations (4D-4) to ( 4D-8) and
recalculating MDA8 values for incrementally increasing emissions reductions until an emissions
level is reach for which all monitors in an urban area achieved design values at the standard level
being evaluated (75, 70, 65, or 60 ppb). Therefore all monitors within an urban area were treated
4D-24
-------
as responding to the same percentage reduction in NOx or NOx/VOC emissions. The final
emissions reductions that were applied in each urban study area are given in Table 4D-2.
The standard error associated with each predicted sensitivity from the linear model can be
propagated through Equations (4D-4) and (4D-8) to quantify the standard error in the final
predicted Os concentration. This gives a measure of uncertainty in the predicted mean sensitivity
at any given Os concentration and allows us to quantify how much our predicted Os could
change given that uncertainty. Further details on the standard error and uncertainty from the
linear fits used in each of the urban study areas are provided in section 4D-4.44D-4.4.
4D-3.2.4 Alternate Methodologies
The methodology used for most of the 15 urban areas is described above. In most cases
the Os response was calculated using Equation (4D-4) which represents a NOx-only control
strategy. In addition, sensitivities applied to observed data were based on the predicted
sensitivity for that hourly Os from the site, season, and hour-specific regression. However, two
alternate scenarios were used for a few cities as described below.
A combined NOx/VOC control strategy was applied to two cities for which controlling
monitors were sensitive to changes in VOC emissions: Chicago and Denver. These were the
only urban study areas in which the addition of VOC reductions allowed adjusted air quality to
just meet the targeted standard levels with smaller reductions in NOx emissions than would be
required if VOC emissions were held constant. In these cases, Equation (4D-8) was applied in
place of Equation (4D-4) and the X,Y cut points described in Table 4D-1 were derived for the
NOx/VOC control case based on brute force runs that reduced the two Os precursors by equal
percentages.
In two cities, New York and Los Angeles, results were affected by aberrant behavior at a
few monitors which occurred during rush-hour and nighttime hours. At a few highly urbanized
monitors during some seasons, the Os increases during highly titrated rush-hour and nighttime
hours appear to be overestimated, leading to predictions that when NOx was reduced Os would
peak during rush-hour periods on the very highest Os days (compare outlier dots for 65 ppb
standard at 9:00-17:00 to outlier dots for 65 ppb standard at 18:00-8:00 in Figure 4D-23). Even
at these monitors, the majority of days did not have this problem as shown by the more normal
diurnal pattern for the boxes (interquartile range) and whiskers (1.5 x interquartile range) in
Figure 4D-23. As mentioned above, for each predicted hourly Os concentration we calculated a
standard error based on the uncertainty in the fitted regressions representing the variability in
modeled sensitivities which is not explained by hourly Os levels alone for each site, season, and
hour. So, in addition to the predicted Os, we can create alternate hourly Os datasets based on the
standard error in these values. To address the aberrant behavior on the very highest Os days in
4D-25
-------
the adjustment scenarios in New York and Los Angeles, we took the 95th percentile confidence
interval for each hourly Os prediction and used the lower bound value to determine the NOx
reductions required to meet the existing and potential alternative standards and to calculate the
hourly Os that would occur under that reduced emission scenario. During most hours and at
most monitors, the 95th percentile range was small (less than +/- 2.7 ppb) so the use of the lower-
bound of the 95th percentile range made little difference in predicted Os concentrations (Table
4D-18 and Table 4D-19). However, using the lower bound of the 95th percent confidence interval
allowed us to dampen the effect of over predicted Os increases during rush-hour times. Since the
Os concentrations for each standard level were created using consistent methodology, these Os
datasets can be used to compare between standards in these cities. In addition, at a single
standard level (for instance 75 ppb) these Os values can be compared against those obtained
using the base methodology in order to quantify uncertainty due to variability in modeled
sensitivities. However, Os values for one standard obtained using the 95th percentile confidence
interval lower bound are not contrasted against Os concentrations for a potential alternative
standard which were obtained using the base methodology since these datasets are not directly
comparable.
4D-4, APPLYING HDDM/CMAQ TO ADJUST OZONE TO JUST MEET EXISTING
AND ALTERNATIVE STANDARDS: RESULTS
4D-4.1 Emission Reductions Applied to Meet Alternative Standards
Table 4D-2 reports the percentage reduction in domain-wide emissions that were used to reach
the existing and potential alternative standard levels in each urban area. Percentages in Chicago
and Denver represent reductions in anthropogenic NOx and VOC. Percentages in all other cities
represent reductions only in anthropogenic NOx emissions. Percentages in New York and in Los
Angeles were calculated based on air quality estimates at the lower end of the 95th percentile
confidence interval as discussed above in section 4D-3.2.4. Please note that these reductions and
broad nationwide emission cuts are not intended to represent recommended control scenarios
since they would not be the most efficient method for achieving the standard in many localized
areas.
4D-26
-------
Table 4D-2. Percent emissions reductions used for each urban area to obtain each
standard.
Urban Area
Atlanta
Baltimore
Boston
Chicago
Cleveland
Dallas
Denver
Detroit
Houston
Los Angeles
New York
Philadelphia
Sacramento
Saint Louis
Washington
D.C.
Years
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
Standard Level*
75 ppb
50%
23%
46%
44%
40%
13%
19%
N/A
48%
50%
50%
50%
51%
15%
59%
N/A
62%
42%
87.1%
87%
64%
52%
54%
42%
63%
64%
45%
10%
53%
31%
70 ppb
58%
43%
54%
52%
49%
40%
52%
27%
61%
64%
57%
58%
65%
46%
69%
54%
68%
53%
89.3%
89%
74%
67%
61%
52%
70%
71%
56%
34%
60%
50%
65 ppb
64%
54%
61%
60%
61%
53%
66%
55%
73%
77%
65%
64%
78%
64%
76%
66%
74%
63%
91 .2%
91%
92%
89%
68%
61%
76%
77%
66%
50%
67%
60%
60 ppb
71%
62%
69%
67%
70%
65%
80%
70%
88%
88%
72%
71%
87%
87%
84%
78%
82%
75%
93.2%
93%
N/A
N/A
74%
68%
84%5
84%
75%
63%
74%
71%
* N/A values for the 75 ppb standard level mean that a particular urban area did not have any design values above 75
for that 3-year period so no controls were needed. N/A values for the 60 ppb standard level mean that this
adjustment methodology was not able to bring design values down to 60 for that particular city and 3-year period.
5 An error was discovered in the Sacramento adjustment to 60 ppb for the 2006-2008 design value period in which
84% NOX cuts were applied instead of 85% NOX cuts. The 84% NOX cuts thus simulate meeting a standard of 61
ppb for Sacramento for those years. Results for the 85% NOX cut case are not expected to be substantially
different from the 84% NOX cut case that was analyzed.
4D-27
-------
4D-4.2 Design Values
Table 4D-3 through Table 4D-17 report the design values for sites in each of the urban
areas for recent years (2006-2008 and 2008-2010) and for adjusted Os levels representing just
meeting standards of 75, 70, 65, and 60 ppb. In each table, the highest design value for each
scenario is displayed in bold text. These tables demonstrate that in some urban areas high Os
values at monitors in different locations have different magnitudes of response to reductions in
NOx (and VOC) emissions. Atlanta monitor 132470001, a downwind monitor, had the highest
design value for 2006-2008. With the NOx reduction scenarios to meet various standard levels,
the two downtown monitors (131210055 and 130890002) switch to being the highest monitors in
the area. These downtown locations which have large amounts of spatially concentrated NOx
emissions are expected to be more VOC limited and therefore less responsive to NOx emission
reductions than surrounding rural and suburban areas. This phenomenon occurs to varying
degrees in each of the 15 urban areas included in this analysis. For instance, the difference in
response between monitors on high days is very pronounced in Cleveland and in New York. In
Cleveland, a downwind monitor located east of the city along Lake Erie was the high monitor for
both 2006-2008 and 2008-2010 (39071001). That monitor responded more dramatically to NOx
cuts than another monitor (390850003) which was closer to the city. As a result of adjusting air
quality to just meet a 60 ppb standard level at monitor 390850003, the design value at monitor
390071001 was lowered to 45 ppb for the 2006-2008 data and 44 ppb for the 2008-2010 data.
Similarly in the New York area, the highest measured Os values occurred at downwind sites in
Connecticut and on Long Island. Two highly urbanized sites (360050110 in the Bronx and
360610135 in Manhattan) have lower observed design values but become the controlling
monitors in the adjustment scenarios for 75, 70, and 65 ppb. In contrast, several cities do not
show this change in location of the highest monitor after adjustments to just meet potential
alternative standard levels. For example, the highest monitor for the 2008-2010 design value
period in the Dallas area (484393009, located on the North side of Fort Worth) remained the
highest monitor through all adjustment scenarios for that time period. Similarly, the controlling
monitor in Sacramento (060670012, a downwind suburban site) remained the highest or second
highest site for all adjustment scenarios. However, in Sacramento, another downwind monitor
(060610006) which did not have as high of an observed design value, responded less to
emissions changes and essentially "caught up" to the controlling monitor in the 65 and 60 ppb
scenarios.
It is important to note that a model-based adjustment technique is uniquely capable of
capturing this type of spatial heterogeneity in response to emissions reductions. The quadratic
rollback technique used in the first draft HREA forces Os at all monitoring sites in an urban area
to respond identically.
4D-28
-------
Table 4D-3. Design values for the Atlanta area regulatory monitors from observed data
and for adjustments to meet the existing and potential alternative standards in 2006-2008
and 2008-2010.
Monitor
132470001
132230003
131510002
131350002
131210055
131130001
130970004
130890002
130850001
130770002
130670003
2006-2008
obs
95
80
94
88
91
86
87
93
77
84
85
75 ppb
72
62
72
71
75
67
68
75
59
65
68
70 ppb
66
58
65
67
70
61
63
70
55
60
63
65 ppb
61
55
61
63
65
57
59
65
53
56
59
60 ppb
55
51
55
58
59
53
54
60
49
52
53
2008-2010
obs
78
70
79
74
80
N/A
75
79
71
68
76
75 ppb
71
64
73
71
75
N/A
69
75
66
63
72
70 ppb
64
59
66
66
70
N/A
64
70
60
58
66
65 ppb
59
55
60
61
65
N/A
59
64
56
54
62
60 ppb
55
52
55
58
60
N/A
55
60
52
51
57
The highest DV for each scenario is shown in bold. N/A values indicate that there was not enough ambient data to
compute a design value for that monitor during a specific 3-year period.
Table 4D-4. Design values for the Baltimore area regulatory monitors from observed data
and for adjustments to meet the existing and potential alternative standards in 2006-2008
and 2008-2010.
Monitor
240030014
240051007
240053001
240130001
240251001
240259001
245100054
2006-2008
obs
87
80
85
83
91
89
66
75 ppb
72
68
74
68
75
75
61
70 ppb
66
63
69
63
70
69
58
65 ppb
61
59
65
59
65
65
56
60 ppb
56
53
60
53
59
59
52
2008-2010
obs
79
77
78
76
89
78
67
75 ppb
67
68
69
64
75
68
62
70 ppb
63
64
65
60
70
64
59
65 ppb
58
60
62
56
65
60
55
60 ppb
54
56
58
53
60
56
53
The highest DV for each scenario is shown in bold.
4D-29
-------
Table 4D-5. Design values for the Boston area regulatory monitors from observed data
and for adjustments to meet the existing and potential alternative standards in 2006-2008
and 2008-2010.
Monitor
250010002
250051002
250070001
250092006
250094004
250095005
250170009
250171102
250213003
250250041
250250042
250270015
2006-2008
obs
79
80
83
81
78
79
75
78
82
74
67
82
75 ppb
70
70
74
73
72
72
68
70
73
72
66
75
70 ppb
67
67
70
69
68
68
64
66
68
69
64
70
65 ppb
62
62
65
62
63
62
58
59
62
63
60
63
60 ppb
58
57
60
57
58
56
52
53
56
58
56
58
2008-2010
obs
73
75
78
74
N/A
71
68
71
73
72
62
76
75 ppb
72
73
75
72
N/A
69
66
69
72
72
63
74
70 ppb
68
67
69
68
N/A
66
62
64
68
70
63
70
65 ppb
63
62
65
63
N/A
62
58
58
62
65
60
65
60 ppb
58
57
60
56
N/A
56
53
53
57
60
57
58
The highest DV for each scenario is shown in bold. N/A values indicate that there was not enough ambient data to
compute a design value for that monitor during a specific 3-year period.
Table 4D-6. Design values for the Chicago area regulatory monitors from observed data
and for adjustments to meet the existing and potential alternative standards using NOX and
VOC emissions reductions in 2006-2008 and 2008-2010.
Monitor
170310001
170310032
170310064
170310076
170311003
170311601
170314002
170314007
170314201
170317002
170436001
170890005
170971002
170971007
171110001
2006-2008
obs
76
74
71
73
73
75
62
66
71
70
63
66
71
72
64
75 ppb
73
74
70
72
73
72
63
66
70
68
63
63
69
71
61
70 ppb
65
70
66
67
70
66
61
62
66
63
59
57
63
64
55
65 ppb
59
65
62
62
65
60
58
58
62
58
54
53
58
59
51
60 ppb
53
58
56
56
60
53
54
53
56
53
49
47
54
54
47
2008-2010
obs
69
68
64
67
66
70
65
59
69
63
60
66
64
73
64
75 ppb
69
68
64
67
66
70
65
59
69
63
60
66
64
73
64
70 ppb
66
69
65
65
66
67
65
59
68
62
58
62
61
70
61
65 ppb
60
65
61
60
63
60
62
57
63
58
54
56
56
64
55
60 ppb
54
60
57
56
60
55
58
53
59
53
51
51
52
59
51
4D-30
-------
171971011
180890022
180890030
180892008
180910005
180910010
181270024
181270026
550590019
66
73
77
73
69
70
74
70
78
63
70
75
71
66
67
72
67
75
55
64
70
66
58
58
65
59
68
50
59
65
62
53
53
60
54
62
46
53
60
56
47
47
56
50
56
62
61
64
67
65
65
67
62
74
62
61
64
67
65
65
67
62
74
59
59
63
65
61
61
64
58
70
52
54
60
62
55
54
59
53
64
49
50
56
57
51
49
55
49
59
The highest DV for each scenario is shown in bold.
Table 4D-7. Design values for the Cleveland area regulatory monitors from observed data
and for adjustments to meet the existing and potential alternative standards in 2006-2008
and 2008-2010.
Monitor
390071001
390350034
390350064
390355002
390550004
390850003
390850007
390930018
391030003
391331001
391530020
2006-2008
obs
84
78
74
81
73
78
76
74
72
73
82
75 ppb
69
74
67
74
61
75
72
65
61
62
69
70 ppb
62
69
61
68
55
70
67
59
55
56
62
65 ppb
55
64
56
61
49
65
62
53
49
50
54
60 ppb
45
57
47
50
42
60
55
44
41
42
44
2008-2010
obs
77
75
68
75
77
76
72
70
70
67
75
75 ppb
64
72
62
68
63
75
70
62
58
57
64
70 ppb
58
67
57
62
55
70
66
56
52
51
57
65 ppb
50
62
51
55
48
65
60
49
47
46
50
60 ppb
44
57
45
48
42
60
56
43
41
41
44
The highest DV for each scenario is shown in bold.
4D-31
-------
Table 4D-8. Design values for the Dallas area regulatory monitors from observed data and
for adjustments to meet the existing and potential alternative standards in 2006-2008 and
2008-2010.
Monitor
480850005
481130069
481130075
481130087
481210034
481211032
481390016
481391044
482311006
482510003
482570005
483670081
483970001
484390075
484391002
484392003
484393009
484393011
2006-2008
obs
83
74
80
82
91
81
75
N/A
70
83
73
84
75
89
83
87
87
79
75 ppb
70
67
71
65
73
65
62
N/A
56
69
59
66
61
75
71
75
73
64
70 ppb
66
63
68
63
68
62
59
N/A
54
66
57
62
57
70
68
70
68
61
65 ppb
61
59
63
59
62
57
55
N/A
51
61
53
57
53
64
64
65
63
58
60 ppb
57
56
59
55
56
53
51
N/A
47
56
50
52
49
58
60
60
57
54
2008-2010
obs
77
67
78
78
80
78
72
68
64
80
67
75
74
85
79
86
82
79
75 ppb
66
63
71
65
67
65
61
56
54
69
56
61
62
74
70
75
71
66
70 ppb
62
59
67
61
62
60
57
53
50
65
53
57
57
69
66
70
66
61
65 ppb
57
56
63
58
57
55
54
50
47
59
50
54
53
63
62
65
61
58
60 ppb
54
54
58
54
53
51
50
47
45
55
47
50
49
58
58
60
55
54
The highest DV for each scenario is shown in bold. N/A
compute a design value for that monitor during a specific
values indicate that there was not enough ambient data to
3-year period.
4D-32
-------
Table 4D-9. Design values for the Denver area regulatory monitors from observed data
and for adjustments to meet the existing and potential alternative standards using NOX and
VOC emissions reductions in 2006-2008 and 2008-2010.
Monitor
080013001
080050002
080130011
080310014
080310025
080350004
080590002
080590005
080590006
080590011
080690011
080691004
081230009
2006-2008
obs
71
71
81
73
N/A
82
78
78
86
81
82
71
76
75 ppb
69
65
70
70
N/A
72
73
70
75
75
70
61
66
70 ppb
66
61
65
67
N/A
67
69
65
70
70
65
57
62
65 ppb
65
58
60
64
N/A
61
64
61
64
65
59
53
57
60 ppb
60
54
55
60
N/A
56
58
56
58
59
55
50
53
2008-2010
obs
70
67
73
68
65
76
73
72
77
72
74
65
71
75 ppb
70
66
71
68
64
74
72
71
75
72
72
63
69
70 ppb
69
63
66
67
62
69
70
67
70
70
65
57
63
65 ppb
65
59
61
64
59
64
65
61
65
65
60
53
59
60 ppb
60
52
53
60
56
54
56
54
55
57
52
48
52
The highest DV for each scenario is shown in bold. N/A values indicate that there was not enough ambient data
compute a design value for that monitor during a specific 3-year period.
to
Table 4D-10. Design values for the Detroit area regulatory monitors from observed data
and for adjustments to meet the existing and potential alternative standards in 2006-2008
and 2008-2010.
Monitor
260490021
260492001
260990009
260991003
261250001
261470005
261610008
261630001
261630019
2006-2008
obs
74
76
81
80
76
78
74
71
82
75 ppb
61
58
69
75
70
63
61
64
74
70 ppb
56
54
64
70
65
58
56
59
69
65 ppb
52
50
61
65
60
55
53
56
64
60 ppb
49
46
57
60
55
51
49
51
59
2008-2010
obs
68
68
74
73
72
71
66
66
75
75 ppb
68
68
74
73
72
71
66
66
75
70 ppb
59
56
66
70
69
60
58
62
70
65 ppb
54
51
61
65
64
56
54
58
65
60 ppb
49
47
57
60
57
51
49
53
59
The highest DV for each scenario is shown in bold.
4D-33
-------
Table 4D-11. Design values for the Houston area regulatory monitors from observed data
and for adjustments to meet the existing and potential alternative standards in 2006-2008
and 2008-2010.
Monitor
480391004
480391016
481671034
482010024
482010026
482010029
482010046
482010047
482010051
482010055
482010062
482010066
482010070
482010075
482010416
482011015
482011034
482011035
482011039
482011050
483390078
2006-2008
obs
85
76
N/A
83
80
85
75
76
80
91
81
89
74
76
89
74
80
73
87
80
80
75 ppb
63
56
N/A
64
61
63
60
61
63
70
62
75
66
66
70
57
63
59
63
61
56
70 ppb
59
53
N/A
60
57
58
57
58
59
65
58
70
64
64
65
55
60
56
59
58
52
65 ppb
55
51
N/A
56
54
54
54
54
56
60
55
65
61
62
62
52
57
53
56
55
49
60 ppb
50
47
N/A
51
50
47
49
49
51
53
50
60
59
59
56
48
53
49
51
51
45
2008-2010
obs
85
74
75
83
78
81
72
76
77
82
72
75
73
74
77
73
76
75
81
75
71
75 ppb
73
63
65
75
68
72
66
73
71
75
65
75
71
73
72
65
71
69
68
67
61
70 ppb
67
60
61
69
63
66
61
68
66
69
62
70
68
69
67
61
66
64
65
62
56
65 ppb
61
55
57
63
58
59
56
62
61
64
57
65
64
65
62
57
62
59
59
58
50
60 ppb
54
50
52
55
52
51
51
55
55
57
52
59
60
60
57
52
55
53
53
53
45
The highest DV for each scenario is shown in bold. N/A
compute a design value for that monitor during a specific
values indicate that there was not enough ambient data to
3-year period.
4D-34
-------
Table 4D-12. Design values for the Los Angeles area regulatory monitors from observed
data and for adjustments to meet the existing and potential alternative standards using the
lower bound of the 95th percent confidence interval of estimated hourly Os in 2006-2008
and 2008-2010.
Monitor
060370002
060370016
060370113
060371002
060371103
060371201
060371301
060371602
060371701
060372005
060374002
060375005
060376012
060379033
060590007
060591003
060592022
060595001
060650004
060650008
060650012
060651010
060651016
060651999
060652002
060655001
060656001
060658001
060658005
060659001
060659003
060710001
060710005
060710012
060710306
2006-2008
obs
96
107
69
92
73
97
58
78
103
92
59
64
105
94
73
66
87
83
N/A
79
102
N/A
105
N/A
86
97
107
107
N/A
102
63
86
119
96
89
75 ppb
63
64
54
69
68
54
56
71
63
75
54
56
57
55
58
52
55
61
N/A
55
58
N/A
60
N/A
55
56
57
63
N/A
55
48
57
70
60
60
70 ppb
60
60
52
64
64
52
54
67
59
70
52
54
54
52
56
51
52
57
N/A
53
55
N/A
57
N/A
54
54
54
59
N/A
52
47
56
65
57
57
65 ppb
56
57
50
59
61
50
52
63
55
65
52
53
51
49
54
50
50
54
N/A
52
52
N/A
54
N/A
52
52
51
55
N/A
49
46
54
61
56
56
60 ppb
53
53
48
54
56
47
50
59
52
60
50
51
47
47
51
48
48
51
N/A
51
49
N/A
51
N/A
51
50
47
51
N/A
46
45
53
55
54
54
2008-2010
obs
89
103
72
84
70
91
N/A
69
90
87
61
61
97
91
68
66
81
74
97
81
102
78
102
77
85
95
102
97
93
96
63
80
112
99
87
75 ppb
63
65
55
69
68
53
N/A
68
62
75
55
56
57
54
58
53
54
60
60
55
58
52
59
52
55
55
55
60
58
53
48
54
68
60
58
70 ppb
60
61
53
64
64
51
N/A
64
58
70
54
54
54
52
55
52
52
57
56
54
55
51
56
51
53
54
52
56
55
51
47
52
64
57
56
65 ppb
56
57
50
59
60
49
N/A
61
55
65
53
52
50
50
53
51
50
54
53
53
52
49
53
49
52
52
50
53
51
48
47
51
59
55
53
60 ppb
52
53
48
54
56
47
N/A
57
51
60
52
50
47
48
51
49
47
51
49
52
49
48
50
48
50
49
47
49
48
46
46
50
55
54
52
4D-35
-------
060711004
060711234
060712002
060714001
060714003
060719004
061110007
061110009
061111004
061112002
061112003
061113001
110
80
112
96
116
116
75
80
83
88
64
61
67
56
68
60
68
67
47
47
49
49
48
47
62
55
63
56
63
62
46
46
48
48
47
46
57
54
58
53
59
58
44
44
46
46
46
46
52
53
53
50
54
53
43
43
45
44
46
45
100
75
101
96
103
102
78
79
79
86
63
63
65
54
65
60
65
64
48
47
47
49
47
47
61
53
61
57
61
60
46
46
46
48
46
46
56
52
56
53
56
55
45
45
45
46
46
45
52
51
51
50
51
50
44
43
44
44
45
44
The highest DV for each scenario is shown in bold. N/A values indicate that there was not enough ambient data to
compute a design value for that monitor during a specific 3-year period.
Table 4D-13. Design values for the New York area regulatory monitors from observed
data and for adjustments to meet the existing and potential alternative standards using the
lower bound of the 95th percent confidence interval of estimated hourly Os in 2006-2008
and 2008-2010.
Monitor
090010017
090011123
090013007
090019003
090070007
090090027
090093002
340030006
340170006
340190001
340210005
340230011
340250005
340273001
340315001
360050110
360050133
360610135
360715001
360790005
2006-2008
obs
99
102
102
106
104
90
108
N/A
92
94
101
100
94
94
86
89
87
N/A
97
91
75 ppb
75
60
70
66
62
61
64
N/A
69
54
59
59
58
51
53
74
70
N/A
49
53
70 ppb
67
50
61
58
53
54
56
N/A
64
45
50
50
51
42
45
70
66
N/A
41
46
65 ppb
39
36
40
39
37
37
40
N/A
60
38
36
37
37
38
37
65
54
N/A
35
35
2008-2010
obs
91
91
89
91
86
77
88
84
91
88
87
86
87
87
83
76
82
80
88
87
75 ppb
74
64
68
70
63
61
64
71
71
59
63
64
65
56
60
70
72
75
54
59
70 ppb
65
54
60
61
54
53
55
63
64
49
53
53
55
46
50
66
65
70
45
50
65 ppb
43
37
42
41
40
39
41
46
60
39
39
39
39
40
37
60
54
65
36
37
4D-36
-------
360810124
360850067
361030002
361030004
361030009
361192004
88
95
98
102
104
104
65
61
65
61
65
64
58
55
58
53
58
56
44
45
40
40
41
37
85
90
90
84
92
87
70
66
71
61
70
67
61
59
61
54
61
57
46
47
43
42
44
38
The highest DV for each scenario is shown in bold. N/A values indicate that there was not enough ambient data to
compute a design value for that monitor during a specific 3-year period.
Table 4D-14. Design values for the Philadelphia area regulatory monitors from observed
data and for adjustments to meet the existing and potential alternative standards in 2006-
2008 and 2008-2010.
Monitor
100031007
100031010
100031013
240150003
340010006
340070003
340071001
340110007
340150002
340290006
420170012
420290100
420450002
420910013
421010004
421010024
2006-2008
obs
80
83
78
90
N/A
87
86
81
87
87
92
82
83
84
67
89
75 ppb
60
65
64
68
N/A
75
65
61
70
67
75
63
68
67
59
74
70 ppb
57
60
60
63
N/A
70
60
57
65
62
70
59
64
62
56
69
65 ppb
53
56
56
58
N/A
65
55
52
60
57
64
54
59
57
53
63
60 ppb
50
51
53
52
N/A
60
52
49
55
53
58
50
55
53
50
58
2008-2010
obs
75
76
75
80
74
N/A
80
76
81
81
83
76
74
78
66
82
75 ppb
64
67
67
68
64
N/A
69
63
72
69
74
65
68
71
62
75
70 ppb
59
62
63
63
60
N/A
63
59
67
64
69
60
64
66
59
70
65 ppb
55
57
59
58
56
N/A
59
54
62
59
64
56
59
61
55
65
60 ppb
51
53
56
54
52
N/A
54
51
57
55
60
52
55
56
52
60
The highest DV for each scenario is shown in bold. N/A
compute a design value for that monitor during a specific
values indicate that there was not enough ambient data to
3-year period.
4D-37
-------
Table 4D-15. Design values for the Sacramento area regulatory monitors from observed
data and for adjustments to meet the existing and potential alternative standards in 2006-
2008 and 2008-2010.
Monitor
060170010
060170012
060170013
060170020
060570005
060570007
060571001
060610002
060610004
060610006
060670002
060670006
060670010
060670011
060670012
060670013
060670014
060675003
060950004
060950005
060953003
061010003
061010004
061130004
061131003
2006-2008
obs
96
76
70
98
91
87
70
90
89
90
78
87
79
82
99
78
N/A
95
60
68
75
72
85
76
76
75 ppb
71
66
61
74
69
66
61
69
67
73
64
71
65
65
75
64
N/A
72
57
57
59
59
69
59
59
70 ppb
66
64
59
69
66
63
59
65
63
68
61
66
62
61
70
61
N/A
68
56
55
55
56
66
55
56
65 ppb
63
63
58
64
62
60
58
61
60
64
58
62
59
58
65
59
N/A
64
55
52
53
53
64
52
53
60 ppb
56
60
57
58
57
56
57
56
55
59
54
56
55
53
59
54
N/A
58
53
49
49
48
61
48
49
2008-2010
obs
90
71
N/A
89
84
81
N/A
87
78
90
75
85
75
77
99
N/A
57
92
63
69
70
66
76
72
72
75 ppb
67
61
N/A
68
63
63
N/A
66
59
73
62
69
62
61
75
N/A
51
70
59
58
56
54
62
56
56
70 ppb
63
60
N/A
64
60
59
N/A
62
56
69
58
65
59
57
70
N/A
49
65
58
55
53
51
60
53
53
65 ppb
59
59
N/A
60
57
57
N/A
58
53
65
55
61
56
54
65
N/A
47
61
56
53
51
49
58
50
50
60 ppb
54
57
N/A
55
53
54
N/A
54
49
60
52
56
53
51
59
N/A
45
55
54
50
49
46
56
47
47
The highest DV for each scenario is shown in bold. N/A
compute a design value for that monitor during a specific
values indicate that there was not enough ambient data to
3-year period.
4D-38
-------
Table 4D-16. Design values for the Saint Louis area regulatory monitors from observed
data and for adjustments to meet the existing and potential alternative standards in 2006-
2008 and 2008-2010.
Monitor
170831001
171170002
171190008
171191009
171193007
171630010
290990019
291130003
291831002
291831004
291890004
291890005
291890014
295100085
295100086
2006-2008
obs
73
70
76
78
77
72
N/A
81
N/A
N/A
78
76
82
N/A
81
75 ppb
61
58
67
67
68
65
N/A
67
N/A
N/A
71
65
73
N/A
75
70 ppb
57
54
62
62
63
60
N/A
62
N/A
N/A
66
60
68
N/A
70
65 ppb
52
50
57
57
57
55
N/A
56
N/A
N/A
60
55
62
N/A
65
60 ppb
48
46
52
52
52
51
N/A
51
N/A
N/A
55
50
57
N/A
60
2008-2010
obs
69
66
71
72
68
68
72
72
77
74
N/A
65
71
68
N/A
75 ppb
67
64
70
70
67
67
70
70
75
72
N/A
64
70
68
N/A
70 ppb
62
60
67
66
63
64
67
65
70
66
N/A
59
66
69
N/A
65 ppb
57
55
61
61
59
59
62
60
65
61
N/A
55
62
65
N/A
60 ppb
52
51
56
56
53
55
56
55
59
54
N/A
51
57
60
N/A
The highest DV for each scenario is shown in bold. N/A
compute a design value for that monitor during a specific
values indicate that there was not enough ambient data to
3-year period.
Table 4D-17. Design values for the Washington D.C. area regulatory monitors from
observed data and for adjustments to meet the existing and potential alternative standards
in 2006-2008 and 2008-2010.
Monitor
110010025
110010041
110010043
240090011
240170010
240210037
240313001
240330030
240338003
510130020
510590005
510590018
2006-2008
obs
80
86
87
79
82
82
84
83
87
85
79
86
75 ppb
62
73
75
61
63
65
68
68
65
70
61
68
70 ppb
58
69
70
57
59
61
63
64
60
66
58
64
65 ppb
53
64
65
54
55
56
59
60
56
62
54
60
60 ppb
48
59
60
49
50
52
54
54
52
56
50
56
2008-2010
obs
75
77
79
77
75
75
74
79
77
79
67
73
75 ppb
68
74
75
69
68
69
68
74
70
74
62
68
70 ppb
60
69
70
61
61
63
63
67
62
68
57
62
65 ppb
55
65
65
56
56
58
59
62
57
63
53
58
60 ppb
49
60
60
49
50
52
53
55
51
57
50
52
The highest DV for each scenario is shown in bold.
4D-39
-------
510590030
510591005
510595001
510610002
510690010
511071005
511530009
511790001
515100009
85
83
83
70
72
82
78
81
81
69
68
66
54
55
62
58
60
65
64
64
61
51
52
58
54
56
61
60
59
57
48
49
53
49
52
56
55
54
52
45
46
49
46
48
53
81
68
66
65
68
75
70
70
74
75
65
62
59
61
68
61
63
68
68
60
58
53
55
60
55
57
61
63
57
54
50
51
56
51
53
56
55
54
51
45
46
50
46
48
51
4D-4.3 Distribution of Hourly Os Concentrations
Figure 4D-13 through Figure 4D-27 display diurnal boxplots of hourly Cb concentrations
at monitor locations in each urban area for observed air quality, air quality adjusted to meet the
existing standard, and an example of air quality adjusted to meet a potential alternative standard
(65 ppb), for 2006-2008 and 2008-2010. Note that these plots include data from multiple
monitoring sites within each urban area so they generally encompass the overall distribution of
Os at both the urban core sites and the downwind suburban sites. The hourly plots show similar
patterns in most cities for which Os concentrations during daytime hours decrease from observed
air quality (black) to air quality adjusted to meet the existing standard (red), and decrease further
when adjusted to meet a potential alternative standard of 65 ppb (blue). These daytime decreases
are generally seen most on high Os days represented by outlier dots. In some cities the mid-
range Os days, represented by the 25th - 75th percentile boxes, remained fairly constant (Boston)
and in other cities, mid-range Os days decreased (Atlanta). Although daytime Os decreases,
concentrations during morning rush-hour generally increase. These increases are associated with
VOC-limited and NOx titration conditions near NOx sources during rush-hour periods. Reducing
NOx under those conditions results in less Os titration and thus increases Os concentrations.
Nighttime increases in Os are often seen to a lesser extent than morning rush-hour increases.
These phenomena generally lead to a flattening of the diurnal Os pattern with smaller differences
between daytime and nighttime concentrations as NOx emissions are reduced. Cases that
required more substantial NOx cuts to reach 75 and 65 ppb standards generally have more
pronounced patterns of decreases in daytime Os and increases in nighttime Os leading to a flatter
diurnal Os pattern (e.g. Los Angeles in Figure 4D-22). Two cities, Houston and New York, do
not follow this general pattern. In Houston (Figure 4D-21), mid-range Os values increase during
daytime hours as the highest Os concentrations decrease. This pattern is consistent with NOx-
limited conditions on high Os days but VOC-limited conditions on mid-range Os days or
potentially NOx-limited conditions at high Os sites and VOC limited conditions at mid-range Os
sites. Therefore NOx reductions, which are applied on all days, increase Os on mid-range Os
4D-40
-------
days. Note that mid-range Os days in Houston start out quite low (20-30 ppb) and increase
modestly. The changes in Os from observed air quality to the existing standard in New York
look similar to the trends shown for other cities, but the diurnal Os pattern for the 65 ppb
potential alternative standard has some unrealistic features which were previously discussed in
section 4D-4D-3.2.4 of this appendix. The steep NOX cuts (-90%) applied to meet a 65 ppb
standard in New York result in a very flat diurnal pattern for mid-range values (see boxes from
Figure 4D-23) but the outliers represented by dots in Figure 4D-23 show an inverse diurnal
pattern with lower Os concentrations during the day and higher values at night. This is a result
of the modeled sensitivities showing that nighttime Os does not respond to NOx emissions
changes on high Os nights while daytime Os does respond to NOx emission changes during those
same days. Thus the daytime values drop substantially while the nighttime values show no
change. These results are not seen when the sensitivities are applied for the more modest cuts
(50-65%) to reach 75 ppb. The diurnal pattern in New York for the 65 ppb case for the highest
days is clearly unrealistic as there is no chemical process expected to result in daytime values
lower than values at night, rather it appears to be an artifact of applying this model-based
methodology to extremely large emissions reductions in a location where relationships derived
under 3 discrete conditions (base, 50% NOx cut, 90% NOx cut) cannot fully characterize
sensitivities over the entire range of possible reductions. As discussed previously in section 4D-
4D-3.2.4, the use of the lower bound from the 95th percentile confidence interval can mitigate
some of this behavior, but in the case of New York some unrealistic outliers still remain. It
should be noted, however that this unrealistic diurnal pattern is only seen in the outlier points,
and is not seen the box and whiskers which represent 1.5 times the interquartile range (most of
the data).
Figure 4D-28 through Figure 4D-42 display the same information as Figure 4D-13
through Figure 4D-27 but for monthly rather than hourly distributions. Note that missing months
in these plots indicate that there was no monitor data available for those months and years.
Similar to the diurnal plots, the seasonal distributions become flatter when meeting 75 and 65
ppb standard levels especially for the highest Os days. This is due to more Os decreases during
summer months and more Os increases in winter months. The Os increases in the winter are
consistent with the understanding that solar insolation rates are lower in the winter reducing total
photochemical activity and shifting the net effect of NOx emissions on Os which can both create
Os through photochemical pathways and destroy Os through titration. In addition, the decreases
on the highest Os days and increases on the lowest Os days show a visible compression of the Os
distribution in these plots similar to what was seen in the diurnal plots. The changes for mid-
range Os days also show a pattern of shifting higher mid-range Os earlier in the year. While in
most cities, the highest interquartile Os concentrations in the recent conditions occur in the
4D-41
-------
summer months (June-August), in many areas the highest interquartile Os concentrations shift to
spring months (April-May) for the adjustment scenarios of meeting 75 and 65 ppb standard
levels. This pattern can be seen most dramatically in Atlanta, Baltimore, Boston, Denver, Los
Angeles, New York, Philadelphia, Sacramento, and Washington D.C. This pattern is consistent
with higher contribution from non-U.S. anthropogenic sources at lower standard levels than
under recent observed conditions. Many of these non-U.S. anthropogenic sources such as
stratospheric intrusions and international transport have been shown to peak during spring
months as discussed in the Integrated Science Assessment (EPA, 2013).
atlanta sites: 2006-2008 atlanta sites: 2008-2010
8-
_
CL
8
s-
observed
75 ppb standard
Q W
II
aB
O2o0
!T???88°8 -
nL
«" ???^
• nnrirh^ • ' '
,.,F™"lr
_ li.n • LJ
flBbi-ai?;.
1111 nflOiiiigiggii.
8
observed
75 ppb standard
65 ppb standard
'*?•»!!.,'
0 2 4 6 8 10 13 16 19 22
hour
0 2 4 6 8 10 13 16 19 22
hour
Figure 4D-13. Hourly Os distributions at Atlanta area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-42
-------
Baltimore sites: 2006-2008
Baltimore sites: 2008-2010
8-
f—
O Q
a.
0§
§-
8-
observed
75 ppb standard
55 ppb standard
i i§
Ilillii
9?-
t
I
: nnTQ-ln :
0 2 4 6 8 10 13 16 19 22
hour
observed
75 ppb standard
65 ppb standard
0246
10 13 16 19 22
hour
Figure 4D-14. Hourly Os distributions at Baltimore area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-43
-------
Boston sites: 2006-2008
Boston sites: 2008-2010
g-
.a o
a.00
a.
i-
observed
75 ppb standard
55 ppb standard
o
O O
oflSiU8o
oi|8i °
! !
0246
10 13 16 19 22
hour
observed
75 ppb stiindttrd
55 ppb standard
kill 1
If ! , =: :
0246
10 13 16 19 22
hour
Figure 4D-15. Hourly Os distributions at Boston area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-44
-------
Chicago sites: 2006-2008
Chicago sites: 2008-2010
Q.
0-5
a-
observed
75 ppb standard - NOx/VOC scenario
65 ppb standard - NOx/VOC scenario
8-
_Q
CL
8
observed
70 ppb standard - NOx/VOC scenario
65 ppb standard - NOx/VOC scenario
0 2 4 6 8 10 13 16 19 22
hour
0246
10 13 16 19 22
hour
Figure 4D-16. Hourly Os distributions at Chicago area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-45
-------
Cleveland sites: 2006-2008
Cleveland sites: 2008-2010
observed
7& ppb standard
65 ppb standard
0 2 4 6 8 10 13 16 19 22
observed
75 ppb standard
65 ppb standard
0 2 4 6 8 10 13 16 19 22
Figure 4D-17. Hourly Os distributions at Cleveland area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-46
-------
Dallas sites: 2006-2008
Dallas sites: 2008-2010
observed
75 ppb standard
55 ppb standard
0 2 4 6 8 10 13 16 19 22
_Q
0-0
Q_(a
8
8-
observed
75 ppb standard
55 ppb standard
!l"
II
iS,,8?=
r : i: JM
Iflnrfp : aj
3i§§
:*i8:
.? .;-
' Si
J
iaSSiflli
0 2 4 6 8 10 13 16 19 22
hour
Figure 4D-18. Hourly Os distributions at Dallas area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-47
-------
Denver sites: 2006-2008
Denver sites: 2008-2010
3-
s-
observed
75 ppb standard - NOx/VOC scenario
65 ppb standard - NOx/VOC scenario
8
o .
observed
75 ppb standard - NOx/VOC scenario
65 ppb standard - NOx/VOC scenario
0246
10 13 16 19 22
hour
0 2 4 6 8 10 13 16 19 22
hour
Figure 4D-19. Hourly Os distributions at Denver area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-48
-------
Detroit sites: 2006-2008
Detroit sites: 2008-2010
observed
75 ppb standard
55 ppb standard
0 2 4 6 8 10 13 16 19 22
observed
70 ppb standard
65 ppb standard
0246
10 13 16 19 22
hour
Figure 4D-20. Hourly Os distributions at Detroit area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-49
-------
Houston sites: 2006-2008
Houston sites: 2008-2010
8-
.0
Q.
Q.
observed
75 ppb standard
55 ppb standard
Wa^J
RfflflflpB0 UdfflF
. _l 'IJUJ j. j. j. j. j. i j j x i ^ . j. i »
8-
8
observed
75 ppb standard
55 ppb standard
o§ o °
0008°^,
6-0
888 "??1
* : *i- : 'wir i;:''
'itiulaiPi^
X J. J. J. i
0 2 4 6 8 10 13 16 19 22
hour
0 2 4 6 8 10 13 16 19 22
hour
Figure 4D-21. Hourly Os distributions at Houston area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-50
-------
Los Angeles sites: 2006-2008
Los Angeles sites: 2008-2010
observed
75 ppb standard
55 ppb standard
' ' •
,: afJUUv
observed
75 ppb standard
65 ppb standard
0 2 4 6 8 10 13 16 19 22
0 2 4 6 8 10 13 16 19 22
hour
Figure 4D-22. Hourly Os distributions at Los Angeles area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-51
-------
New York sites: 2006-2008
New York sites: 2008-2010
observed
75 ppb standard
55 ppb standard
0 2 4 6 8 10 13 16 19 22
observed
75 ppb standard
55 ppb standard
0 2 4 6 8 10 13 16 19 22
Figure 4D-23. Hourly Os distributions at New York area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-52
-------
Philadelphia sites: 2006-2008
Philadelphia sites: 2008-2010
8-
o cr>
a.00
0.
a-
observed
7& ppb i"t?'nri;iirl
65 ppb standard
jllHUHIlHi!
*: innl
observed
75 ppb standard
65 ppb standard
0 2 4 6 8 10 13 16 19 22
0 2 4 6 8 10 13 16 19 22
hour
Figure 4D-24. Hourly Os distributions at Philadelphia area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-53
-------
Sacramento sites: 2006-2008
-8-
g-
observed
7& ppb standard
65 ppb standard
O O
1 0 o
i a o
i! i Hi
llll! I ! LoQOdn Hll
o 6 u J-
Q.
CL
8
Sacramento sites: 2008-2010
observed
75 ppb standard
65 ppb standard
•: •:"' : : " irtiiBfe: =
0 2
8 10 13
hour
16 19 22
0 2 4 6 8 10 13
hour
16 19 22
Figure 4D-25. Hourly Os distributions at Sacramento area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-54
-------
St Louis sites: 2006-2008
St Louis sites: 2008-2010
observed
75 ppb standard
55 ppb standard
0 2 4 6 8 10 13 16 19 22
observed
75 ppb standard
55 ppb standard
»!??,. -|l jl "!|
\ Illlf ft™
•••UP™!!!!!
0 2 4 6 8 10 13 16 19 22
Figure 4D-26. Hourly Os distributions at St. Louis area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-55
-------
Washington D.C. sites: 2006-2008
Washington D.C, sites: 2008-2010
8-
observed
75 ppb standard
55 ppb standard
I
O o
'Ǥ
Ml
^ H
MttUift
observed
75 ppb standard
65 ppb standard
0 2 4 6 8 10 13 16 19 22
0 2 4 6 8 10 13 16 19 22
hour
Figure 4D-27. Hourly Os distributions at Washington, D.C. area regulatory monitoring
sites for observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations, red
boxes (pink whiskers/dots) show the predicted distribution of hourly Os concentrations for
the 75 ppb adjustment scenarios and blue boxes (light blue whiskers/dots) show the
predicted distribution of hourly Os concentrations for the 65 ppb adjustment scenario.
Boxes show the interquartile range, whiskers extend to 1.5 x the interquartile range and
dots depict outlier values.
4D-56
-------
atlanta sites: 2006-2008
atlanta sites: 2008-2010
observed
75 ppb standard
55 ppb standard
3456789 10 12
observed
75 ppb standard
55 ppb standard
23456789 10 12
Figure 4D-28. Monthly Os distributions at Atlanta area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-57
-------
g-
i-
Baltimore sites: 2006-2008
observed
75 ppb standard
55 ppb standard
f 1!
123456789 10 12
month
Baltimore sites: 2008-2010
observed
75 ppb standard
65 ppb standard
1234567
month
9 10 12
Figure 4D-29. Monthly Os distributions at Baltimore area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-58
-------
Boston sites: 2006-2008
Boston sites: 2008-2010
observed
75 ppb standard
55 ppb standard
8
o _
observed
75 ppb standard
55 ppb standard
I t
,
•nlHlHHlHI • *
1 23456789 10
month
123456789 10
month
12
Figure 4D-30. Monthly Os distributions at Boston area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-59
-------
Q.
0-5
9-
R-
Chicago sites: 2006-2008
observed
75 ppb standard - NOx/VOC scenario
65 ppb standard - NOx/VOC scenario
Chicago sites: 2008-2010
observed
75 ppb standard - NOx/VOC scenario
65 ppb standard - NOx/VOC scenario
123456789 10
month
12
123456789 10
month
12
Figure 4D-31. Monthly Os distributions at Chicago area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-60
-------
Cleveland sites: 2006-2008
Cleveland sites: 2008-2010
.
CLC
Q.'
observed
75 ppb standard
55 ppb standard
i
I I (
|IiB|
E
I i
4 5 6 7 8 9 10 12
month
observed
75 ppb standard
55 ppb standard
Figure 4D-32. Monthly Os distributions at Cleveland area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-61
-------
Dallas sites: 2006-2008
Dallas sites: 2008-2010
observed
75 ppb standard
55 ppb standard
23456789 10
month
12
observed
75 ppb standard
55 ppb standard
1234567
month
9 10 12
Figure 4D-33. Monthly Os distributions at Dallas area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-62
-------
9-
s-
Denver sites: 2006-2008
observed
7& ppb standard - NOx/VOC scenario
65 ppb standard - NOx/VOC scenario
i lii
8
o .
8-
1 23456789 10
month
12
Denver sites: 2008-2010
observed
75 ppb standard - NOx/VOC scenario
65 ppb standard - NOx/VOC scenario
123456789 10
month
12
Figure 4D-34. Monthly Os distributions at Denver area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-63
-------
8-
Q.
Q.
Detroit sites: 2006-2008
observed
75 ppb standard
55 ppb standard
IBI1
HJ
I
H
678
month
12
Detroit sites: 2008-2010
observed
75 ppb standard
55 ppb standard
23456789 10 12
Figure 4D-35. Monthly Os distributions at Detroit area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-64
-------
Houston sites: 2006-2008
Houston sites: 2008-2010
observed
75 ppb standard
55 ppb standard
1 23456789 10
month
observed
75 ppb standard
55 ppb standard
1234567
month
9 10 12
Figure 4D-36. Monthly Os distributions at Houston area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-65
-------
Qu
Q.
Los Angeles sites: 2006-2008
observed
75 ppb standard
55 ppb standard
i
H..':-;;- ..•-•"
- i = | T ' a -
« c. c o u e a
..
8
23456789 10
month
12
Los Angeles sites: 2008-2010
observed
75 ppb standard
HB(B
Itttii ill
123456789 10
month
12
Figure 4D-37. Monthly Os distributions at Los Angeles area regulatory monitoring sites
for observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations, red
boxes (pink whiskers/dots) show the predicted distribution of hourly Os concentrations for
the 75 ppb adjustment scenarios and blue boxes (light blue whiskers/dots) show the
predicted distribution of hourly Os concentrations for the 65 ppb adjustment scenario.
Boxes show the interquartile range, whiskers extend to 1.5 x the interquartile range and
dots depict outlier values.
4D-66
-------
New York sites: 2006-2008
New York sites: 2008-2010
8-
observed
75 ppb standard
55 ppb standard
i.'.
-
8
_a
CL
CL
8
5?-
observed
75 ppb standard
55 ppb standard
jiAttiiii^
I,
123456789 10 12
month
1 23456789 10
month
12
Figure 4D-38. Monthly Os distributions at New York area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-67
-------
g-
i-
Philadelphia sites: 2006-2008
observed
75 ppb standard
55 ppb standard
O 0
III,
8-
Philadelphia sites: 2008-2010
observed
75 ppb standard
55 ppb standard
III 11,
123456789 10 12
month
1234567
month
9 10 12
Figure 4D-39. Monthly Os distributions at Philadelphia area regulatory monitoring sites
for observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations, red
boxes (pink whiskers/dots) show the predicted distribution of hourly Os concentrations for
the 75 ppb adjustment scenarios and blue boxes (light blue whiskers/dots) show the
predicted distribution of hourly Os concentrations for the 65 ppb adjustment scenario.
Boxes show the interquartile range, whiskers extend to 1.5 x the interquartile range and
dots depict outlier values.
4D-68
-------
Sacramento sites: 2006-2008
Sacramento sites: 2008-2010
observed
75 ppb standard
55 ppb standard
i 23456789 10 12
observed
75 ppb standard
55 ppb standard
23456789 10 12
8
Figure 4D-40. Monthly Os distributions at Sacramento area regulatory monitoring sites
for observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations, red
boxes (pink whiskers/dots) show the predicted distribution of hourly Os concentrations for
the 75 ppb adjustment scenarios and blue boxes (light blue whiskers/dots) show the
predicted distribution of hourly Os concentrations for the 65 ppb adjustment scenario.
Boxes show the interquartile range, whiskers extend to 1.5 x the interquartile range and
dots depict outlier values.
4D-69
-------
St Louis sites: 2006-2008
St Louis sites: 2008-2010
observed
75 ppb standard
55 ppb standard
i 23456789 10 12
observed
75 ppb standard
55 ppb standard
23456789 10
Figure 4D-41. Monthly Os distributions at St. Louis area regulatory monitoring sites for
observed air quality, and air quality adjusted to meet the existing (75 ppb) and alternative
(65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes (black
whiskers/dots) show the observed distribution of hourly Os concentrations, red boxes (pink
whiskers/dots) show the predicted distribution of hourly Os concentrations for the 75 ppb
adjustment scenarios and blue boxes (light blue whiskers/dots) show the predicted
distribution of hourly Os concentrations for the 65 ppb adjustment scenario. Boxes show
the interquartile range, whiskers extend to 1.5 x the interquartile range and dots depict
outlier values.
4D-70
-------
Washington D.C. sites: 2006-2008
observed
75 ppb standard
55 ppb standard
i 23456789 10 12
Washington D.C, sites: 2008-2010
observed
75 ppb standard
55 ppb standard
123456789 10
month
Figure 4D-42. Monthly Os distributions at Washington, D.C. area regulatory monitoring
sites for observed air quality, and air quality adjusted to meet the existing (75 ppb) and
alternative (65 ppb) standards for 2006-2008 (left) and 2008-2010 (right). White boxes
(black whiskers/dots) show the observed distribution of hourly Os concentrations, red
boxes (pink whiskers/dots) show the predicted distribution of hourly Os concentrations for
the 75 ppb adjustment scenarios and blue boxes (light blue whiskers/dots) show the
predicted distribution of hourly Os concentrations for the 65 ppb adjustment scenario.
Boxes show the interquartile range, whiskers extend to 1.5 x the interquartile range and
dots depict outlier values.
4D-4.4 Standard Errors for Predicted Hourly and Daily 8-hr Maximum Os
Concentrations
Standard error values, s2, for the predicted sensitivity are calculated using Equations (4D-
25) and (4D-26) as described by Wilks (2006):
MSE =
n-2
Equation (4D-25)
s2{Yh} = MSE
i (xh-xy
Equation (4D-26)
In Equation (4D-25) mean standard error (MSE) is calculated as the sum of squared residual
values from the linear model divided by n-2 to account for two degrees of freedom in the fit
4D-71
-------
(slope and intersect). In Equation (4D-26), the first term characterizes the uncertainty in
estimating the true mean given a fixed sample size of n and the second term characterizes the
uncertainty in the slope. Taken together, the standard error for each predicted sensitivity
coefficient is impacted by the goodness of fit for the linear model (MSB), the sample size (n),
and the distance of the x value from the bulk of the data. These standard error values represent
the uncertainty in the predicted central tendency from the linear model.
The standard errors were propagated through Equations (4D-4) to (4D-8) to derive a
standard error in hourly Cb concentrations. These hourly Os standard errors are thus impacted
both by the standard error in the sensitivity and by the amount of emissions reduction applied
(i.e. standard error increases with larger emissions changes and thus at lower Cb standard levels).
These hourly Os standard errors represent uncertainty in predicted Os concentrations due to
uncertainty in predicting the central tendency with the linear regression. Standard error values for
predicted hourly Os were generally small relative to total predicted hourly Os concentrations.
This indicates that even when regression fits had somewhat lower correlation coefficients, the
resulting uncertainty from the use of the regression line does not substantially affect predicted Os
concentrations.
4D-72
-------
Table 4D-18 and Table 4D-19 show the mean and 95th percentile standard error for all
hourly Os values (2006-2010) in each urban area for the existing standard of 75 ppb and the
potential alternative standards of 70 ppb, 65 ppb, and 60 ppb.
To further put this uncertainty into a context relevant to the form of the Os standard and
the metric used in health analyses, we propagated standard errors for hourly Os to uncertainties
in 8-hr daily maximum Os concentrations using Equation (4D-27):
SE(Y) =
64 8
Equation (4D-27)
Boxplots showing the range of 8-hr daily maximum Os standard errors calculated for each cases
study area are shown in the chapter 4 of the HREA. Maps of mean standard error in 8-hr daily
maximum Os concentration at each monitoring location are shown in Figure 4D-43 through
Figure 4D-57. These figures generally show that 8-hr daily maximum Os mean standard error is
small (less than 0.4 ppb) with slightly higher values up to 1.3 ppb in the most NOx limited areas
at the lower standard levels.
4D-73
-------
Table 4D-18. Mean standard error (ppb) in adjusted hourly
urban area for each standard.
concentration in each
Urban Area
Atlanta
Baltimore
Boston
Chicago*
Cleveland
Dallas
Denver*
Detroit
Houston
Los
Angeles*
New York*
Philadelphia
Sacramento
Saint Louis
Washington
D.C.
Years
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
Standard Level
75 ppbt
0.65
0.37
0.62
0.62
0.51
0.16
0.29
N/A
0.60
0.60
0.59
0.57
0.51
0.17
0.60
N/A
0.73
0.52
1.15
1.14
0.84
0.69
0.66
0.62
0.49
0.49
0.57
0.13
0.67
0.49
70 ppb
0.69
0.68
0.65
0.64
0.53
0.51
0.66
0.41
0.65
0.66
0.64
0.63
0.62
0.48
0.70
0.56
0.83
0.55
1.17
1.15
1.06
0.92
0.72
0.65
0.55
0.56
0.61
0.48
0.71
0.66
65 ppb
0.74
0.71
0.70
0.69
0.59
0.54
0.77
0.68
0.76
0.80
0.74
0.72
0.77
0.58
0.80
0.65
0.95
0.62
1.18
1.17
1.39
1.34
0.80
0.71
0.61
0.62
0.70
0.58
0.78
0.72
60 ppb1
0.81
0.77
0.78
0.75
0.69
0.63
0.99
0.83
0.96
0.95
0.86
0.84
0.85
0.82
0.92
0.81
1.11
0.76
1.21
1.19
N/A
N/A
0.90
0.79
0.81
0.70
0.81
0.66
0.87
0.84
* Values are from the standard NOx reduction scenario for all cities except Chicago and Denver, for which value are
from the standard NOx/VOC reduction scenario and New York and Los Angeles, for which values based on
reductions required when using the lower bound of the 95th percentile confidence interval.
•f N/A values for the 75 ppb standard level mean that a particular urban area did not have any design values above 75
for that 3-year period so adjustments were made to ambient data.
{N/A values for the 60 ppb standard level mean that this adjustment methodology was not able to bring design
values down to 60 for that particular city and 3-year period.
4D-74
-------
Table 4D-19. 95th percentile standard error (ppb) in adjusted hourly Os concentration in
each urban area for each standard.
Urban Area
Atlanta
Baltimore
Boston
Chicago*
Cleveland
Dallas
Denver*
Detroit
Houston
Los
Angeles*
New York*
Philadelphia
Sacramento
Saint Louis
Washington
D.C.
Years
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
Standard Level
75 ppbt
1.28
0.72
1.22
1.21
1.02
0.33
0.54
N/A
1.13
1.13
1.16
1.12
1.14
0.38
1.12
N/A
1.45
0.98
2.84
2.78
1.64
1.34
1.29
1.25
1.10
1.11
1.07
0.26
1.29
0.94
70 ppb
1.34
1.33
1.26
1.26
1.05
1.03
1.22
0.78
1.19
1.22
1.26
1.24
1.35
1.07
1.28
1.05
1.66
1.04
2.86
2.80
2.15
1.87
1.39
1.29
1.22
1.24
1.14
0.90
1.35
1.27
65 ppb
1.42
1.38
1.34
1.33
1.19
1.09
1.43
1.27
1.35
1.42
1.48
1.45
1.70
1.28
1.46
1.20
1.96
1.15
2.89
2.83
2.87
2.98
1.56
1.41
1.35
1.38
1.29
1.07
1.46
1.36
60 ppb1
1.56
1.47
1.49
1.46
1.41
1.29
1.89
1.56
1.69
1.69
1.79
1.75
1.89
1.84
0.69
1.48
2.41
1.39
2.93
2.86
N/A
N/A
1.78
1.58
1.54
1.55
1.51
1.21
1.63
1.57
* Values are from the standard NOx reduction scenario for all cities except Chicago and Denver, for which value are
from the standard NOx/VOC reduction scenario and New York and Los Angeles, for which values based on
reductions required when using the lower bound of the 95th percentile confidence interval.
•f N/A values for the 75 ppb standard level mean that a particular urban area did not have any design values above 75
for that 3-year period so adjustments were made to ambient data.
{N/A values for the 60 ppb standard level mean that this adjustment methodology was not able to bring design
values down to 60 for that particular city and 3-year period.
4D-75
-------
mean standard error (ppb) at Atlanta sites
75 ppb adjustment case (2006-2008)
mean standard error I ppb) at Atlanta sites
75 ppb adjustment case (2008-2010)
V
O.D
0.4
D.8
1.2
mean standard error {ppb) at Atlanta sites
60 ppb adjustment case (2006-2008)
0.0
i
D.4
i
D.8
1J2
mean standard error {ppb) at Atlanta sites
60 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
0.0
i
0.4
i
D.8
1.2
Figure 4D-43. Mean standard error in maximum daily 8-hr Os concentrations at each
monitoring location in Atlanta for the 75 ppb adjustment scenario (top row) and the 60 ppb
adjustment scenario (bottom row) for the 2006-2008 time period (left column) and the
2008-2010 time period (right column).
4D-76
-------
mean standard error (ppb) at Baltimore sites
75 ppb adjustment case (2006-20*8)
mean standard error (ppb) at Baltimore sites
75 ppb adjustment case (2008-2010)
O.D 0.4 D.8 1.2
mean standard error (ppb) at Baltimore sites
60 ppb adjustment case (2006-2008)
V ' ' "! ! i
0.0 0.4 D.8 1J2
mean standard error (ppb) at Baltimore sites
60 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
0.0
i
0.4
i
D.8
1.2
Figure 4D-44. Mean standard error in maximum daily 8-hr Os concentrations at each
monitoring location in Baltimore for the 75 ppb adjustment scenario (top row) and the 60
ppb adjustment scenario (bottom row) for the 2006-2008 time period (left column) and the
2008-2010 time period (right column).
4D-77
-------
mean standard error (ppb) at Boston sites
75 ppb adjustment case (2006-20*8)
mean standard error (ppb) at Boston sites
75 ppb adjustment case (2008-2010)
0.0
0.4
D.8
1.2
mean standard error (ppb) at Boston sites
60 ppb adjustment case (2006-2008)
v • •"! r i
0.0 0.4 D.8 1J2
mean standard error (ppb) at Boston sites
60 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
0.0
i
0.4
i
D.8
1.2
Figure 4D-45. Mean standard error in maximum daily 8-hr Os concentrations at each
monitoring location in Boston for the 75 ppb adjustment scenario (top row) and the 60 ppb
adjustment scenario (bottom row) for the 2006-2008 time period (left column) and the
2008-2010 time period (right column).
4D-78
-------
mean standard error (ppb) at Chicago sites
75 ppb adjustment case (2006-2009)
mean standard error (ppb) at Chicago sites
75 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
mean standard error (ppb) at Chicago sites
60 ppb adjustment case (2006-2008)
V ' ' "! ! i
0.0 0.4 D.8 1J2
mean standard error (ppb) at Chicago sites
60 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
0.0
i
0.4
i
D.8
1.2
Figure 4D-46. Mean standard error in maximum daily 8-hr Os concentrations at each
monitoring location in Chicago for the 75 ppb adjustment scenario (top row) and the 60
ppb adjustment scenario (bottom row) for the 2006-2008 time period (left column) and the
2008-2010 time period (right column).
4D-79
-------
mean standard error (ppb) at Cleveland sites
75 ppb adjustment case (2006-2008}
mean standard error (ppb) at Cleveland sites
75 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
0.0
1
0.4
I
D.8
1.2
mean standard error (ppb) at Cleveland sites
60 ppb adjustment case (2006-2008}
mean standard error (ppb) at Cleveland sites
60 ppb adjustment case (2008-2010}
O.D
0.4
D.8
1.2
O.D
I
0.4
I
D.8
1.2
Figure 4D-47. Mean standard error in maximum daily 8-hr Os concentrations at each
monitoring location in Cleveland for the 75 ppb adjustment scenario (top row) and the 60
ppb adjustment scenario (bottom row) for the 2006-2008 time period (left column) and the
2008-2010 time period (right column).
4D-80
-------
mean standard error (ppb) at Dallas sites
75 ppb adjustment case (2Q06-2008|
mean standard error (ppb) at Dallas sites
75 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
0.0
1
0.4
I
D.8
1.2
mean standard error (ppb) at Dallas sites
60 ppb adjustment case (2006-2008)
mean standard error (ppb) at Dallas sites
60 ppb adjustment case (2QOB-2010f
O.D
0.4
D.8
1.2
O.D
I
0.4
I
D.8
1.2
Figure 4D-48. Mean standard error in maximum daily 8-hr Os concentrations at each
monitoring location in Dallas for the 75 ppb adjustment scenario (top row) and the 60 ppb
adjustment scenario (bottom row) for the 2006-2008 time period (left column) and the
2008-2010 time period (right column).
4D-81
-------
mean standard error (ppb) at D.C. sites
75 ppb adjustment case (2006-2008)
mean standard error (ppb) at D.C. sites
75 ppb adjustment case (2008-2010)
•-
O.D
0.4
D.8
1.2
mean standard error (ppb) at D.C. sites
60 ppb adjustment case (2006-2008)
0.0
I
D.4
i
D.8
1J2
mean standard error (ppb) at D.C. sites
60 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
0.0
i
0.4
i
D.8
1.2
Figure 4D-49. Mean standard error in maximum daily 8-hr Os concentrations at each
monitoring location in Washington D.C. for the 75 ppb adjustment scenario (top row) and
the 60 ppb adjustment scenario (bottom row) for the 2006-2008 time period (left column)
and the 2008-2010 time period (right column).
4D-82
-------
mean standard error (ppb) at Denver sites
75 ppb adjustment case (2Q06-2fl08|
mean standard error (ppb) at Denver sites
75 ppb adjustment case (2008-2010)
r
O.D 0.4
D.8
1.2
mean standard error (ppb) at Denver sites
60 ppb adjustment case (200G-200B)
r
0.0
i r
0.4
i r
D.8
1.2
mean standard error (ppb) at Denver sites
60 ppb adjustment case (2008-2010)
r
i-V-
*
J •
i
tt
« *
\ *] •
I
O.D
0.4
D.8
1.2
O.D
I
0.4
I
D.8
1.2
Figure 4D-50. Mean standard error in maximum daily 8-hr Os concentrations at each
monitoring location in Denver for the 75 ppb adjustment scenario (top row) and the 60 ppb
adjustment scenario (bottom row) for the 2006-2008 time period (left column) and the
2008-2010 time period (right column).
4D-83
-------
mean standard eiror (ppb) at Detroit sites
75 ppb adjustment case (2Q06-2008|
mean standard eiror (ppb) at Detroit sites
75 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
0.0
1
0.4
I
D.8
1.2
mean standard error (ppb) at Detroit sites
60 ppb adjustment case (2006-2008}
mean standard eiror (ppb) at Detroit sites
60 ppb adjustment case (2QOB-2010f
O.D
0.4
D.8
1.2
O.D
I
0.4
I
D.8
1.2
Figure 4D-51. Mean standard error in maximum daily 8-hr Os concentrations at each
monitoring location in Detroit for the 75 ppb adjustment scenario (top row) and the 60 ppb
adjustment scenario (bottom row) for the 2006-2008 time period (left column) and the
2008-2010 time period (right column).
4D-84
-------
mean standard error (ppb) at Houston sites
75 ppb adjustment case (2Q06-2008|
mean standard error (ppb) at Houston sites
75 ppb adjustment case (2008-2010)
'A )
O.D
0.4
D.8
1.2
mean standard error (ppb) at Houston sites
60 ppb adjustment case (2006-2008}
0.0
1
0.4
I
D.8
1.2
mean standard error (ppb) at Houston sites
60 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
O.D
I
0.4
I
D.8
1.2
Figure 4D-52. Mean standard error in maximum daily 8-hr Os concentrations at each
monitoring location in Houston for the 75 ppb adjustment scenario (top row) and the 60
ppb adjustment scenario (bottom row) for the 2006-2008 time period (left column) and the
2008-2010 time period (right column).
4D-85
-------
mean standard error (ppb) at Los Angeles sites
75 ppb adjustment case (2006-2049)
standard error (ppb) at Los Angeles sites
75 ppb adjustment case (2008-2010)
0.0
0.4
D.8
1.2
0.0
i
D.4
i
D.8
1J2
mean standard error (ppb) at Los Angeles sites
60 ppb adjustment case (2006-2008)
(iv^s-*—i—i
"*k ••* /
mean standard error (ppb) at Los Angeles sites
60 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
0.0
i
0.4
i
D.8
1.2
Figure 4D-53. Mean standard error in maximum daily 8-hr Os concentrations at each
monitoring location in Los Angele for the 75 ppb adjustment scenario (top row) and the 60
ppb adjustment scenario (bottom row) for the 2006-2008 time period (left column) and the
2008-2010 time period (right column).
4D-86
-------
mean standard error (ppb) at New York sites
75 ppb adjustment case (2006-20*8)
mean standard error (ppb] at New York sites
75 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
mean standard error (ppb) at New York sites
65 ppb adjustment case (2006-2008)
0.0
i
D.4
i
D.8
1J2
mean standard error (ppb) at New York sites
65 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
0.0
i
0.4
i
D.8
1.2
Figure 4D-54. Mean standard error in maximum daily 8-hr Os concentrations at each
monitoring location in New York for the 75 ppb adjustment scenario (top row) and the 60
ppb adjustment scenario (bottom row) for the 2006-2008 time period (left column) and the
2008-2010 time period (right column).
4D-87
-------
mean standard error (ppb) at Philadelphia sites
75 ppb adjustment case (2006-2049)
mean standard error (ppb) at Philadelphia sites
75 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
mean standard error (ppb) at Philadelphia sites
60 ppb adjustment case (2006-2008)
v • • "i r i
0.0 0.4 D.8 1J2
mean standard error (ppb) at Philadelphia sites
60 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
0.0
i
0.4
i
D.8
1.2
Figure 4D-55. Mean standard error in maximum daily 8-hr Os concentrations at each
monitoring location in Philadelphia for the 75 ppb adjustment scenario (top row) and the
60 ppb adjustment scenario (bottom row) for the 2006-2008 time period (left column) and
the 2008-2010 time period (right column).
4D-88
-------
mean standard error (ppbf at Sacramento sites
75 ppb adjustment case (2006-20*8)
mean standard error (ppb) at Sacramento sites
75 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
mean standard error (ppb) at Sacramento sites
60 ppb adjustment case (2006-2008)
! ! !
0.0 0.4 D.8 1J2
mean standard error (ppb) at Sacramento sites
60 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
0.0
i
0.4
i
D.8
1.2
Figure 4D-56. Mean standard error in maximum daily 8-hr Os concentrations at each
monitoring location in Sacramento for the 75 ppb adjustment scenario (top row) and the 60
ppb adjustment scenario (bottom row) for the 2006-2008 time period (left column) and the
2008-2010 time period (right column).
4D-89
-------
mean standard error (ppb) at St Louis sites
75 ppb adjustment case {2006-2008}
mean standard error (ppb) at St Louis sites
75 ppb adjustment case (2008-2010)
O.D 0.4
D.8
1.2
mean standard error (ppb) at St Louis sites
60 ppb adjustment case (2006-2008)
\ II\
O.D 0.4 D.8 1.2
mean standard error (ppb) at St Louis sites
60 ppb adjustment case (2008-2010)
O.D
0.4
D.8
1.2
O.D
I
0.4
I
D.8
1.2
Figure 4D-57. Mean standard error in maximum daily 8-hr Os concentrations at each
monitoring location in St. Louis for the 75 ppb adjustment scenario (top row) and the 60
ppb adjustment scenario (bottom row) for the 2006-2008 time period (left column) and the
2008-2010 time period (right column).
4D-4.5 Air Quality Inputs for the Epidemiology-Based Risk Assessment
The air quality inputs to the epidemiology-based risk assessment discussed in HREA
Chapter 7 were spatially averaged "composite monitor6" values for 12 of the 15 urban study
areas. The procedure for calculating these values is described in Appendix 4A. Figure 4D-58
through Figure 4D-69 show boxplots of the composite monitor daily maximum 8-hour values for
observed air quality, air quality adjusted to meet the existing standard, and the potential
alternative standards of 70 ppb, 65 ppb, and 60 ppb in the 12 urban areas in the epidemiology-
6 Composite monitor values are calculated as the average of all monitors within each CBSA. See Appendix 4A for
more details.
4D-90
-------
based risk assessment. There are eight panels in each figure. The panels are designed to show
contrasts based on three factors:
1. Spatial extent of the urban study area: the top row of panels in each figure are based on
the smaller areas from the Zanobetti and Schwartz (2008) study (Z & S),, while the
bottom row of panels are based on the larger Core Based Statistical Areas (CBSAs).
Length of the Ch season: the 1st and 3rd columns of panels in each figure are based on a
shorter June - August Os season, which was used in Z & S, while the 2nd and 4th columns
of panels are based on a longer April - October Os season. The Smith et al. (2009) study
was based on the required Os monitoring season, which varied by area but often
encompassed the April-October period.
2. Year: the two left-hand columns of panels in each figure are based on 2007, while the
two right-hand columns of panels in each figure are based on 2009. The epidemiology-
based risk assessment focused on these two years: 2007, to represent a year with higher
Os concentrations, and 2009, to represent a year with lower Os concentrations.
There are a few properties common to nearly all of the figures. First, the highest
composite monitor values represented by the top whiskers decrease when air quality is adjusted
to meet the existing standard, and continue to decrease as air quality is further adjusted to meet
lower alternative standard levels. By contrast, the lowest composite monitor values represented
by the bottom whiskers increase when air quality is adjusted to meet the existing standard, and
continue to increase as air quality is further adjusted to meet lower alternative standard levels.
The behavior of the 25th percentile, median, mean, and 75th percentile values making up the
boxes varies by urban area and across the three contrasting factors.
The spatial extent contrasts show that the base composite monitor values based on the
smaller Z & S study areas tend to be slightly lower than the base composite monitor values based
on the CBSAs. This is because the highest Os concentrations are often located downwind of the
urban area's population center, and the smaller Z & S areas often do not capture the highest
observed concentrations. Two notable exceptions to this tendency are Atlanta and Sacramento,
where the highest monitored concentrations are located near the population center. The spatial
extent contrasts also show that when air quality is adjusted, the highest composite monitor values
tend to decrease more quickly for the CBSAs than for the Z & S areas, and conversely the lowest
composite monitor values tend to increase more slowly for the CBSAs than for the Z & S areas.
This is consistent with observed air quality patterns, which show that as NOx emissions decrease,
Os concentrations decrease more quickly downwind of the urban population center, and may
increase near the population center due to less titration. This phenomenon also affects the center
of the distribution, where we see that the 25th percentile, median, mean, and 75th percentile of the
composite monitor values decrease more quickly for the CBSAs than for the Z & S areas. In
4D-91
-------
cases where we see increases in these values, the values tend to increase more quickly for the Z
& S areas than for the CBS As.
The seasonal contrasts show that the composite monitor values based on observed air
quality tend to be lower in the spring and fall months than in the summer months. The highest
composite monitor values tend to occur in the June-August period, so the upper tail of the
distribution often does not change, but the center and lower tail values are usually lower for the
April-October period. One notable exception is Houston, where the highest Os concentrations
are often observed in the spring and fall months. When Os concentrations are adjusted to meet
the existing standard, there are many cases where net decreases in the 25th percentile, median,
and mean composite monitor values for the June-August period turn into net increases in these
values for the April-October period. When air quality is further adjusted to meet the potential
alternative standards, initial increases in these values for the April-October period often persist,
and in a few cases the increases grow larger as the level of the standard decreases. By contrast,
initial increases in the 25th percentile, median, and mean composite monitor values are often
reversed when air quality is further adjusted to meet the potential alternative standards.
Finally, the year contrasts are meant to show the effects of changes in emissions and
meteorology on the composite monitor values. Precursor emissions were generally higher in
2007 than in 2009, and meteorological conditions were generally more favorable to Os formation
in 2007 than in 2009. Thus, the composite monitor values based on observed air quality are
generally higher in 2007 than in 2009, with the exception of Houston, where the climate regime
was the reverse of most of the rest of the U.S. during those two years. When air quality are
adjusted to meet the existing and potential alternative standards, the decreases in the 2007
composite monitor values tend to be larger than in 2009, and there tend to be fewer increases.
However, it is worth noting that since the emissions and meteorological inputs used in the
CMAQ/HDDM modeling for the air quality adjustments were based on 2007 data, we tend to
have more confidence in the adjusted values for 2007 than for 2009.
4D-92
-------
Z & S, June-August, 2007
Z & S, April-October, 2007
Z & S, June-August, 2009
Z & S, April-October, 2009
base 75 70 65 60
CBSA, June-August, 2007
base 75 70 65 60
CBSA, April-October, 2007
base 75 70 65 i
CBSA, June-August, 2009
base 75 70 65 E
CBSA, April-October, 2009
base 75 70 65
60
base 75
70 65
60
base 75 70 65
base 75
70
65
Figure 4D-58. Composite monitor daily maximum 8-hour Os values for Atlanta based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's represent
mean values, whiskers extend up to 1.5x the inter-quartile range from the boxes, and
circles represent outliers.
Z & S, June-August, 2007
Z & S, April-October, 2007
Z & S, June-August, 2009
Z & S, April-October, 2009
base 75 70 65 60
CBSA, June-August, 2007
base 75 70 65 60
CBSA, April-October, 2007
base 75 70 65 60
CBSA, June-August, 2009
base 75 70 65 60
CBSA, April-October, 2009
base 75 70 65
60
base 75
70 65
60
base 75 70 65
base 75
70
65
Figure 4D-59. Composite monitor daily maximum 8-hour Os values for Baltimore based
on observed and adjusted air quality. Boxes represent the median and quartiles, x's
represent mean values, whiskers extend up to 1.5x the inter-quartile range from the boxes,
and circles represent outliers.
4D-93
-------
Z & S, June-August, 2007
Z & S, April-October, 2007
Z & S, June-August, 2009
Z & S, April-October, 2009
base 75 70 65 60
CBSA, June-August, 2007
36 75 70 65 60
CBSA, April-October, 2007
base 75 70 65 i
CBSA, June-August, 2009
base 75 70 65 E
CBSA, April-October, 2009
base 75 70 65
60
base 75
70 65
60
base 75 70 65
base 75
70
65
Figure 4D-60. Composite monitor daily maximum 8-hour Os values for Boston based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's represent
mean values, whiskers extend up to 1.5x the inter-quartile range from the boxes, and
circles represent outliers.
Z & S, June-August, 2007
Z & S, April-October, 2007
Z & S, June-August, 2009
Z & S, April-October, 2009
base 75 70 65 60
CBSA, June-August, 2007
base 75 70 65 60
CBSA, April-October, 2007
base 75 70 65 60
CBSA, June-August, 2009
base 75 70 65 60
CBSA, April-October, 2009
base 75 70 65
60
75 70 65
60
base 75 70
65
base 75 70
65
Figure 4D-61. Composite monitor daily maximum 8-hour Os values for Cleveland based
on observed and adjusted air quality. Boxes represent the median and quartiles, x's
represent mean values, whiskers extend up to 1.5x the inter-quartile range from the boxes,
and circles represent outliers.
4D-94
-------
Z & S, June-August, 2007
Z & S, April-October, 2007
Z & S, June-August, 2009
Z & S, April-October, 2009
base 75 70 65 60
CBSA, June-August, 2007
36 75 70 65 60
CBSA, April-October, 2007
base 75 70 65 i
CBSA, June-August, 2009
base 75 70 65 E
CBSA, April-October, 2009
base 75 70 65
60
base 75
70 65
60
base 75 70 65
base 75
70
65
Figure 4D-62. Composite monitor daily maximum 8-hour Os values for Denver based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's represent
mean values, whiskers extend up to 1.5x the inter-quartile range from the boxes, and
circles represent outliers.
Z & S, June-August, 2007
Z & S, April-October, 2007
Z & S, June-August, 2009
Z & S, April-October, 2009
base 75 70 65 60
CBSA, June-August, 2007
base 75 70 65 60
CBSA, April-October, 2007
base 75 70 65 60
CBSA, June-August, 2009
base 75 70 65 60
CBSA, April-October, 2009
base 75 70 65
60
75 70 65
60
base 75 70
65
base 75 70 65
Figure 4D-63. Composite monitor daily maximum 8-hour Os values for Detroit based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's represent
mean values, whiskers extend up to 1.5x the inter-quartile range from the boxes, and
circles represent outliers.
4D-95
-------
Z & S, June-August, 2007
Z & S, April-October, 2007
Z & S, June-August, 2009
Z & S, April-October, 2009
base 75 70 65 60
CBSA, June-August, 2007
base 75 70 65 60
CBSA, April-October, 2007
base 75 70 65 i
CBSA, June-August, 2009
base 75 70 65 E
CBSA, April-October, 2009
base 75 70 65
60
base 75
70 65
60
base 75 70 65
base 75
70
65
Figure 4D-64. Composite monitor daily maximum 8-hour Os values for Houston based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's represent
mean values, whiskers extend up to 1.5x the inter-quartile range from the boxes, and
circles represent outliers.
Z & S, June-August, 2007
Z & S, April-October, 2007
Z & S, June-August, 2009
Z & S, April-October, 2009
base 75 70 65 60
CBSK,, June-August, 2007
base 75 70 65 60
CBSA, April-October, 2007
base 75 70 65 60
CBSA, June-August, 2009
base 75 70 65 60
CBSA, April-October, 2009
base 75 70 65
60
base 75
70 65
60
base 75 70 65
base 75
70
65
Figure 4D-65. Composite monitor daily maximum 8-hour Os values for Los Angeles based
on observed and adjusted air quality. Boxes represent the median and quartiles, x's
represent mean values, whiskers extend up to 1.5x the inter-quartile range from the boxes,
and circles represent outliers.
4D-96
-------
Z & S, June-August, 2007
Z & S, April-October, 2007
Z & S, June-August, 2009
Z & S, April-October, 2009
base 75 70 65 60
CBSA, June-August, 2007
base 75 70 65 60
CBSA, April-October, 2007
base 75 70 65 i
CBSA, June-August, 2009
base 75 70 65 E
CBSA, April-October, 2009
base 75 70 65
60
base 75
70 65
60
base 75 70 65
base 75
70
65
Figure 4D-66. Composite monitor daily maximum 8-hour Os values for New York based
on observed and adjusted air quality. Boxes represent the median and quartiles, x's
represent mean values, whiskers extend up to 1.5x the inter-quartile range from the boxes,
and circles represent outliers.
Z & S, June-August, 2007
Z & S, April-October, 2007
Z & S, June-August, 2009
Z & S, April-October, 2009
base 75 70 65 60
CBSA, June-August, 2007
base 75 70 65 60
CBSA, April-October, 2007
base 75 70 65 60
CBSA, June-August, 2009
base 75 70 65 60
CBSA, April-October, 2009
base 75 70 65
60
base 75
70 65
60
base 75 70 65
base 75
70
65
Figure 4D-67. Composite monitor daily maximum 8-hour Os values for Philadelphia based
on observed and adjusted air quality. Boxes represent the median and quartiles, x's
represent mean values, whiskers extend up to 1.5x the inter-quartile range from the boxes,
and circles represent outliers.
4D-97
-------
Z & S, June-August, 2007
Z & S, April-October, 2007
Z & S, June-August, 2009
Z & S, April-October, 2009
base 75 70 65 60
CBSA, June-August, 2007
36 75 70 65 60
CBSA, April-October, 2007
base 75 70 65 i
CBSA, June-August, 2009
base 75 70 65 E
CBSA, April-October, 2009
base 75 70 65
60
base 75
70 65
60
base 75 70 65
base 75
70
65
Figure 4D-68. Composite monitor daily maximum 8-hour Os values for Sacramento based
on observed and adjusted air quality. Boxes represent the median and quartiles, x's
represent mean values, whiskers extend up to 1.5x the inter-quartile range from the boxes,
and circles represent outliers.
Z & S, June-August, 2007
Z & S, April-October, 2007
Z & S, June-August, 2009
Z & S, April-October, 2009
base 75 70 65 60
CBSA, June-August, 2007
base 75 70 65 60
CBSA, April-October, 2007
base 75 70 65 60
CBSA, June-August, 2009
base 75 70 65 60
CBSA, April-October, 2009
base 75 70 65
60
base 75
70 65
60
base 75 70 65
base 75
70
65
Figure 4D-69. Composite monitor daily maximum 8-hour Os values for St. Louis based on
observed and adjusted air quality. Boxes represent the median and quartiles, x's represent
mean values, whiskers extend up to 1.5x the inter-quartile range from the boxes, and
circles represent outliers.
4D-98
-------
4D-4.6 Air Quality Inputs for the Exposure and Clinical Risk Assessment
The air quality inputs for the exposure and clinical-based risk assessments discussed in
HREA Chapters 5 and 6 include spatial surfaces of hourly Os concentrations estimated for each
census tract in the 15 urban study areas using Voronoi Neighbor Averaging (VNA). The VNA
technique and its application to ambient concentrations is described in Appendix 4A. In this
section, we present three types of figures that summarize the data from the hourly VNA surfaces
(1) observed air quality, (2) air quality adjusted to meet the existing standard,7 and (3) air quality
adjusted to meet the potential alternative standard of 65 ppb.
Figure 4D-70 through Figure 4D-84 show density scatter plots of the change in daily
maximum 8-hour average (MDA8) Os concentrations versus the observed concentration based
on the hourly VNA estimates in each area. In each of these figures, the left-hand panels show the
observed MDA8 values (x-axis) versus the change in those values that occur when air quality is
adjusted to meet the existing standard (y-axis). The right-hand panels show the MDA8 values for
air quality adjusted to meet the existing standard7 (x-axis) versus the additional change in those
values that occur when air quality is further adjusted to meet the alternative standard of 65 ppb
(y-axis). The top panels show values based on 2006-2008, while the bottom panels show values
based on 2008-2010. Within each panel, the x and y values are rounded to the nearest integer and
colored to show the relative frequency of each 1 ppb x 1 ppb square within the plot region.
Values falling outside of the plot region were set to the nearest value within the plot region, and
frequencies above the range in the color bar were set to the highest value within the color bar.
Figure 4D-85 through Figure 4D-114 show maps of the changes in design values (3-year
average of the annual 4th highest MDA8 values) and May - September average MDA8 values
based on the ambient data and the hourly VNA surfaces. There are two figures for each of the 15
urban study areas, one based on 2006-2008 and one based on 2008-2010 air quality. In each
figure, the panels on the left show the changes in these values that occur when air quality is
adjusted to meet the existing standard,7 and the panels on the right show the additional changes
in these values that occur when air quality is further adjusted to meet the alternative standard of
65 ppb. The top panels show the changes in the design values, while the bottom panels show the
changes in the May - September average MDA8 values. Within each panel, squares show values
based on observed data at ambient Os monitoring sites while circles show values based on VNA
estimates at census tract centroids. Regions shaded pink indicate counties in either the Z & S or
Smith study areas (for the 12 urban study areas in the epidemiology-based risk assessment). The
Z & S and Smith study areas had a few counties in common in all 12 areas, and were identical in
7 Chicago and Detroit were already meeting the existing standard in 2008-2010. The 2008-2010 figures for those
areas are based on air quality adjusted to meet the 70 ppb alternative standard instead of the existing standard.
4D-99
-------
6 of the areas. Regions shaded gray indicate additional counties in the CBSA, and regions shaded
peach indicate any additional counties in the study areas for the exposure and clinical-based risk
assessment. The maps also show monitors that are located outside of the exposure study areas.
These monitors were adjusted along with the other monitors and used in the VNA estimates, but
were not used when determining the emissions reductions necessary to meet standard levels.
Figure 4D-115 through Figure 4D-129 show changes in design values (3-year average of
the annual 4th highest MDA8 values) and May - September average MDA8 values in the 15
urban study areas versus population and population density. The population counts and
population density information for each census tract were obtained from the U.S. Census Bureau
based on the 2010 U.S. Census. Within each figure, the top panels show histograms of the total
population stratified by the change in design value or seasonal average, while the bottom panels
show scatter plots of population density (x-axis) versus change in design value or seasonal
average (y-axis). The left-hand panels are based on 2006-2008, while the right-hand panels are
based on 2008-2010. The first and third rows of panels show changes in design values, while the
second and fourth rows of panels show changes in May-September average values. Finally, the
first and third columns show the changes that occur when air quality is adjusted to meet the
existing standard,7 while the second and fourth columns show the additional changes that occur
when air quality is further adjusted to meet the alternative standard of 65 ppb. Within each panel,
values associated with census tracts falling within the epidemiology study areas as defined
previously are colored pink, values associated with additional census tracts falling within the
CBSA are colored gray, and values associated with any additional census tracts falling within the
exposure study areas are colored peach. Population density is shown on a logarithmic scale, with
values falling outside of the plot region set to the nearest values within the plot region.
In general, the density scatter plots show that the UDDM adjustment procedure predicts
increases in MDA8 Os at low ambient concentrations, and decreases in MDA8 Os at high
concentrations. The vast majority of the increases in MDA8 Os occur at ambient concentrations
below 50 ppb. The relationship between the starting concentrations and the changes in these
values based on the HDDM adjustments is fairly linear with strong negative correlation in all 15
urban areas. In some areas, such as Baltimore and Philadelphia, there is a bimodal pattern near
the center of the distribution that could indicate a differing behavior in the urban population
center versus the surrounding suburban areas.
The maps reveal several trends in the spatial pattern of changes in Os in the urban study
areas. The design values decreased almost universally when air quality was adjusted to meet the
existing standard with a few exceptions,8 and continued to decrease when air quality was further
: All design values from the VNA surfaces decreased when adjusted from recent conditions to the 75 ppb level
except for the following: small areas in Boston where design values increased by up to 1 ppb (2008-2010),
4D-100
-------
adjusted to meet the 65 ppb alternative standard. The design values also tended to decrease more
quickly in suburban and rural areas than in the urban population centers. The May-September
"seasonal" average MDA8 values also followed this trend to some extent. However, the
decreases in the seasonal average values were nearly universal in the suburban and rural areas,
while the behavior in the urban population centers varied amongst the cities. The seasonal
average values in the urban population centers followed one of three distinct patterns:
1. The seasonal average values decreased when air quality was adjusted to meet the existing
standard, and continued to decrease when air quality was further adjusted to meet the 65
ppb alternative standard. (Atlanta, Sacramento, Washington D.C.)
2. The seasonal average values increased or remained constant when air quality was
adjusted to meet the existing standard, then decreased when air quality was adjusted to
meet the 65 ppb alternative standard. (Baltimore, Cleveland, Dallas, Detroit, Los
Angeles, New York, Philadelphia, St. Louis)
3. The seasonal average values increased when air quality was adjusted to meet the existing
standard, and continued to increase or remained constant when air quality was further
adjusted to meet the 65 ppb alternative standard. (Boston, Chicago, Houston, Denver)
The population plots show a clear and consistent trend regarding the population living in
areas associated with various changes in design values and seasonal average concentrations. In
almost every figure, there is a positive correlation between population density and change in
design value or seasonal average concentration. This suggests that in almost every scenario,
when NOx emissions were reduced, the suburban areas surrounding the urban population center
experience more Os reductions than in the urban population center. For the 12 urban areas
examined in the epidemiology-based risk assessment, the majority of the areas associated with
increases in the seasonal average concentration occurred within the epidemiology study area,
which tended to be focused on the urban population center.
These figures show that using the HDDM adjustment methodology, peak Os
concentrations are reduced in urban areas with large domain-wide reductions in U.S.
anthropogenic NOx emissions. In most cases, seasonal average Os concentrations also decreased
with large domain-wide reductions in U.S. anthropogenic NOx emissions. However, there were a
few cases, such as Chicago and Houston, where current NOx emissions were high enough to
cause titration in the urban population centers so that seasonal average Os concentrations were
below regional background levels.
Chicago where design values increased by up to 2 ppb (2006-2008) and 4 ppb (2008-2010) and New York where
design values increased by up to 3 ppb (2008-2010). All areas where design values increases started with very low
design values based on recent air quality conditions.
4D-101
-------
Atlanta 2006 - 2008
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0,6
% of points
0.8
1,0
0.0
0,2
0.4 0.6
% of points
0.8
1,0
Atlanta 2008-2010
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
o
20
40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0.4 0.6
% of points
0.8
1.0
0.0
0.2
0,4 0.6
% of points
0.8
1.0
Figure 4D-70. Change in VNA estimates of the daily maximum 8-hour average (MDA8)
concentrations based on HDDM adjustments in Atlanta.
4D-102
-------
Baltimore 2006 - 2008
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0,6
% of points
0.8
1,0
0.0
0,2
0.4 0.6
% of points
0.8
1,0
Baltimore 2008 - 2010
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0.6
% of points
0.8
1.0
0.0
0.2
0.4 0.6
% of points
0.8
1.0
Figure 4D-71. Change in VNA estimates of the daily maximum 8-hour average (MDA8)
concentrations based on HDDM adjustments in Baltimore.
4D-103
-------
Boston 2006 - 2008
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0,6
% of points
0.8
1,0
0.0
0,2
0.4 0.6
% of points
0.8
1,0
Boston 2008 - 2010
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
o
20 40 60 80 100
Observed MDA8 O3 (ppb)
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0.6
% of points
0.8
1.0
0.0
0.2
0.4 0.6
% of points
0.8
1.0
Figure 4D-72. Change in VNA estimates of the daily maximum 8-hour average (MDA8)
concentrations based on HDDM adjustments in Boston.
4D-104
-------
Chicago 2006 - 2008
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
o
20 40 60 80 100
Observed MDA8 O3 (ppb)
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0,6
% of points
0.8
1,0
0.0
0,2
0.4 0.6
% of points
0.8
1,0
Chicago 2008 - 2010
Change in MDA8 03 from Observed to 70 ppb Change in MDA8 03 from 70 ppb to 65 ppb
o
20 40 60 80 100
Observed MDA8 O3 (ppb)
20 40 60 80 100
70 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0.6
% of points
0.8
1.0
0.0
0.2
0.4 0.6
% of points
0.8
1.0
Figure 4D-73. Change in VNA estimates of the daily maximum 8-hour average (MDA8)
concentrations based on HDDM adjustments in Chicago.
4D-105
-------
Cleveland 2006 - 2008
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0,6
% of points
0.8
1,0
0.0
0,2
0.4 0.6
% of points
0.8
1,0
Cleveland 2008-2010
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0.6
% of points
0.8
1.0
0.0
0.2
0.4 0.6
% of points
0.8
1.0
Figure 4D-74. Change in VNA estimates of the daily maximum 8-hour average (MDA8)
concentrations based on HDDM adjustments in Cleveland.
4D-106
-------
Dallas 2006 - 2008
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0,6
% of points
0.8
1,0
0.0
0,2
0.4 0.6
% of points
0.8
1,0
Dallas 2008-2010
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0.6
% of points
0.8
1.0
0.0
0.2
0.4 0.6
% of points
0.8
1.0
Figure 4D-75. Change in VNA estimates of the daily maximum 8-hour average (MDA8)
concentrations based on HDDM adjustments in Dallas.
4D-107
-------
Denver 2006 - 2008
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0,6
% of points
0.8
1,0
0.0
0,2
0.4 0.6
% of points
0.8
1,0
Denver 2008 - 2010
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
o
20 40 60 80 100
Observed MDA8 O3 (ppb)
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0.6
% of points
0.8
1.0
0.0
0.2
0.4 0.6
% of points
0.8
1.0
Figure 4D-76. Change in VNA estimates of the daily maximum 8-hour average (MDA8)
concentrations based on HDDM adjustments in Denver.
4D-108
-------
Detroit 2006 - 2008
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0,6
% of points
0.8
1,0
0.0
0,2
0.4 0.6
% of points
0.8
1,0
Detroit 2008-2010
Change in MDA8 03 from Observed to 70 ppb Change in MDA8 03 from 70 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
70 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0.6
% of points
0.8
1.0
0.0
0.2
0.4 0.6
% of points
0.8
1.0
Figure 4D-77. Change in VNA estimates of the daily maximum 8-hour average (MDA8)
concentrations based on HDDM adjustments in Detroit.
4D-109
-------
Houston 2006 - 2008
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0,6
% of points
0.8
1,0
0.0
0,2
0.4 0.6
% of points
0.8
1,0
Houston 2008-2010
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0.6
% of points
0.8
1.0
0.0
0.2
0.4 0.6
% of points
0.8
1.0
Figure 4D-78. Change in VNA estimates of the daily maximum 8-hour average (MDA8)
concentrations based on HDDM adjustments in Houston.
4D-110
-------
Los Angeles 2006 - 2008
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0,6
% of points
0.8
1,0
0.0
0,2
0.4 0.6
% of points
0.8
1,0
Los Angeles 2008 - 2010
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0.6
% of points
0.8
1.0
0.0
0.2
0.4 0.6
% of points
0.8
1.0
Figure 4D-79. Change in VNA estimates of the daily maximum 8-hour average (MDA8)
concentrations based on HDDM adjustments in Los Angeles.
4D-111
-------
New York 2006 - 2008
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0,6
% of points
0.8
1,0
0.0
0,2
0.4 0.6
% of points
0.8
1,0
New York 2008-2010
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0.6
% of points
0.8
1.0
0.0
0.2
0.4 0.6
% of points
0.8
1.0
Figure 4D-80. Change in VNA estimates of the daily maximum 8-hour average (MDA8)
concentrations based on HDDM adjustments in New York.
4D-112
-------
Philadelphia 2006 - 2008
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0,6
% of points
0.8
1,0
0.0
0,2
0.4 0.6
% of points
0.8
1,0
Philadelphia 2008-2010
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0.6
% of points
0.8
1.0
0.0
0.2
0.4 0.6
% of points
0.8
1.0
Figure 4D-81. Change in VNA estimates of the daily maximum 8-hour average (MDA8)
concentrations based on HDDM adjustments in Philadelphia.
4D-113
-------
Sacramento 2006 - 2008
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0,6
% of points
0.8
1,0
0.0
0,2
0.4 0.6
% of points
0.8
1,0
Sacramento 2008 - 2010
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0.6
% of points
0.8
1.0
0.0
0.2
0.4 0.6
% of points
0.8
1.0
Figure 4D-82. Change in VNA estimates of the daily maximum 8-hour average (MDA8)
concentrations based on HDDM adjustments in Sacramento.
4D-114
-------
St. Louis 2006 - 2008
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0,6
% of points
0.8
1,0
0.0
0,2
0.4 0.6
% of points
0.8
1,0
St. Louis 2008-2010
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0.6
% of points
0.8
1.0
0.0
0.2
0.4 0.6
% of points
0.8
1.0
Figure 4D-83. Change in VNA estimates of the daily maximum 8-hour average (MDA8)
concentrations based on HDDM adjustments in St. Louis.
4D-115
-------
Washington 2008-2010
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0,6
% of points
0.8
1,0
0.0
0,2
0.4 0.6
% of points
0.8
1,0
Washington 2006 - 2008
Change in MDA8 03 from Observed to 75 ppb Change in MDA8 03 from 75 ppb to 65 ppb
20 40 60 80 100
Observed MDA8 O3 (ppb)
120
20 40 60 80 100
75 ppb Adjusted MDA8 O3 (ppb)
120
0.0
0.2
0,4 0.6
% of points
0.8
1.0
0.0
0.2
0.4 0.6
% of points
0.8
1.0
Figure 4D-84. Change in VNA estimates of the daily maximum 8-hour average (MDA8)
O3 concentrations based on HDDM adjustments in Washington, D.C.
4D-116
-------
Atlanta 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-85. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Atlanta, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-117
-------
Atlanta 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5 0
Change in O3 (ppb)
10
Figure 4D-86. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Atlanta, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-118
-------
Baltimore 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-87. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Baltimore, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-119
-------
Baltimore 2008 - 2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-88. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Baltimore, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-120
-------
Boston 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-89. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Boston, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-121
-------
Boston 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-90. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Boston, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-122
-------
Chicago 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-91. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Chicago, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-123
-------
Chicago 2008 - 2010
Change from Observed to 70 ppb Change from 70 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-92. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Chicago, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-124
-------
Cleveland 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
OD
<
Q
I -
a
Z!
CD
Q
2
-------
Cleveland 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-94. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Cleveland, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-126
-------
Dallas 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-95. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Dallas, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-127
-------
Dallas 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-96. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Dallas, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-128
-------
Denver 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-97. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Denver, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-129
-------
Denver 2008 - 2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-98. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Denver, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-130
-------
Detroit 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-99. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Detroit, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-131
-------
Detroit 2008- 2010
Change from Observed to 70 ppb Change from 70 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-100. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Detroit, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-132
-------
Houston 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
oo
<
Q
CO
Q
2
IB
D)
-
".
-------
Houston 2008- 2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
oo
<
Q
CO
Q
2
IB
D)
CD
J3
E
-^
"5.
-------
Los Angeles 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5 0
Change in O3 (ppb)
10
Figure 4D-103. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Los Angeles, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-135
-------
Los Angeles 2008 - 2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-104. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Los Angeles, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-136
-------
New York 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-105. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for New York, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-137
-------
New York 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-106. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for New York, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-138
-------
Philadelphia 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-107. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Philadelphia, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-139
-------
Philadelphia 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
oo
<
Q
E?
if
oo
Q
2
-------
Sacramento 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
OD
<
Q
a
ZJ
oo
Q
2
(D
CT
_
(D
_Q
E
(D
-^^
Q.
(D
t/1
-20
-10 -5 0
Change in O3 (ppb)
10
Figure 4D-109. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Sacramento, 2006-2008. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-141
-------
Sacramento 2008 - 2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-110. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Sacramento, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-142
-------
St. Louis 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
00
<
Q
E?
if
a
Z!
CO
Q
5
-------
St. Louis 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5
Change in O3 (ppb)
0
10
Figure 4D-112. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for St. Louis, 2008-2010. The points are colored
according to the change in ppb, and values falling outside the range in the color bar were
set to the nearest value within the color bar.
4D-144
-------
Washington 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5 0
Change in O3 (ppb)
10
Figure 4D-113. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Washington, D.C., 2006-2008. The points are
colored according to the change in ppb, and values falling outside the range in the color bar
were set to the nearest value within the color bar.
4D-145
-------
Washington 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20
-10 -5 0
Change in O3 (ppb)
10
Figure 4D-114. Changes in annual 4th highest MDA8 and May-September average MDA8
values based on HDDM adjustments for Washington D.C., 2008-2010. The points are
colored according to the change in ppb, and values falling outside the range in the color bar
were set to the nearest value within the color bar.
4D-146
-------
Atlanta 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Atlanta 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-25 -20 -15 -10
Change in O3 (ppb)
II.
-25 -20 -15 -10 -5 0 -25 -20 -15 -10 -5 0
Change in O3 (ppb) Change in O3 (ppb)
-10 -5
Change inO3 (ppb)
D Epi Study Area O C8SA
Exposure Area
D Epi Study Area D CBSA
-10 -5
Change in O3 (ppb)
Exposure Area
Atlanta 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Atlanta 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
8-
10 100 1,000 10,000
Population Density (people.'km"2)
10 100 1,000 10,000
Population Density (people;'kiTT'2)
10 100 1,000 10,000
Population Density (people/kmA2)
10 100 1,000 10,000
Population Density {people.>'krn"2)
I
8-
CO O
<
Q
Q. O
QJ
ta
10 100 1,000 10,000
Population Density (people/km^)
•» Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/km*2)
Exposure Area
10 100 1,000 10,000
Population Density (people/kiri*2)
Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/'kny'2)
Exposure Area
Figure 4D-115. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Atlanta versus population and population
density.
4D-147
-------
Baltimore 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Baltimore 2008- 2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
8 I
I
-20 -15 -10
Change in O3 (ppb)
-20 -15 -10 -5
Change inO3 (ppb)
-15 -10 -5 0
Change in O3 (ppb)
-15 -10 -5 0
Change in OS (ppb)
I
CO
o> ---
2 =
-20 -15 -10
Change in O3 (ppb)
D Epi Study Area H GBSA
-20 -15 -10 -5 0
Change in OS (ppb)
D Exposure Area
-15 -10 -5 0
Change inO3 (ppb)
D Epi Study Area P CBSA
-15 -10 -5 0
Change in OS (ppb)
n Exposure Area
Baltimore 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Baltimore 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
f SB?-I
8-
t-r*
I..'
10 100 1,000 10,000
Population Density (people:'km"2)
10 100 1,000 10,000
Population Density (people;'kiTT'2)
10 100 1,000 10,000
Population Density (people/kmA2)
10 100 1,000 10,000
Population Density {people:'km"2)
I!
CL. O _.
-------
Boston 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Boston 2008- 2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
£ I
I
=5 ^_
El
<
-15 -10 -5 0
Change in O3 (ppb)
-15 -10 -5 0
Change inO3 (ppb)
-15 -10 -5 0
Change in O3 (ppb)
-15 -10 -5 0
Change in O3 (ppb)
-15 -10 -5 0
Change in O3 (ppb)
-15 -10 -5 0
Change inO3 (ppb)
-15 -10 -5 0
Change inO3 (ppb)
D Epi Study Area H GBSA
Exposure Area
D Epi Study Area P CBSA
-15 -10 -5 0
Change in OS (ppb)
n Exposure Area
Boston 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Boston 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
10 100 1,000 10,000
Population Density (people/km"2)
10 100 1,000 10,000
Population Density (people;'km'2)
10 100 1,000 10,000
Population Density (people/kmA2)
10 100 1,000 10,000
Population Density {people:'km"2)
-
10 100 1,000 10,000
Population Density (people/km^)
•» Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/km*2)
Exposure Area
o 100 1,000 10,000
Population Density (people/kin'2)
Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/'kny'2)
Exposure Area
Figure 4D-117. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Boston versus population and population
density.
4D-149
-------
Chicago 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Chicago 2008 - 2010
Change from Observed to 70 ppb Change from 70 ppb to 65 ppb
£ I
I
=5 ^_
El
<
tfl O c
(D =
-g, 1
I 5
5 a
if'
-15 -10 -5 0
Change in O3 (ppb)
-15 -10 -5 0
Change inO3 (ppb)
Change in OS (ppb)
Change in OS (ppb)
-15 -10 -5 0
Change in O3 (ppb)
-15 -10 -5 0
Change inO3 (ppb)
D Epi Study Area H GBSA
Exposure Area
Change inO3 (ppb)
D Epi Study Area • CBSA
Change in OS (ppb)
Exposure Area
Chicago 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Chicago 2008- 2010
Change from Observed to 70 ppb Change from 70 ppb to 65 ppb
10 100 1,000 10,000
Population Density (people/km"2)
10 100 1,000 10,000
Population Density (people;'km'2)
10 100 1,000 10,000
Population Density (people/knr'2)
10 100 1,000 10,000
Population Density (people.>'krn"2)
I!
CX O o
(D v
ro
10 100 1,000 10,000
Population Density (people/km^)
•» Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/km*2)
Exposure Area
10 100 1,000 10,000
Population Density (people/km*2)
Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/'kny'2)
Exposure Area
Figure 4D-118. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Chicago versus population and population
density.
4D-150
-------
Cleveland 2006-2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Cleveland 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-15 -10 -5
Change in O3 (ppb)
-15 -10 -5
Change inO3 (ppb)
-15 -10 -5 0
Change inO3 (ppb)
D Epi Study Area H CBSA
Exposure Area
D Epi Study Area P CBSA
-15 -10 -5 0
Change in OS (ppb)
n Exposure Area
Cleveland 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Cleveland 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
10 100 1,000 10,000 10 100 1,000 10,000
Population Density (people/km*2) Population Density (people;'km'2)
10 100 1,000 10,000
Population Density (people/kmA2)
10 100 1,000 10,000
Population Density (people.>'krn"2)
.T-.
10 100 1,000 10,000
Population Density (people/knifl2)
« Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/km*2)
Exposure Area
10 100 1,000 10,000
Population Density (people/km*2)
Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/'kny'2)
Exposure Area
Figure 4D-119. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Cleveland versus population and population
density.
4D-151
-------
Dallas 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Dallas 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-20 -15 -10 -505
Change in O3 (ppb)
D Epi Study Area H CBSA
-20 -15 -10 -5 0
Change inO3 (ppb)
15 -10 -5 0
Change inO3 (ppb)
Exposure Area
D Epi Study Area P CBSA
-15 -10 -5
Change in OS (ppb)
n Exposure Area
Dallas 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Dallas 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
a
l
8-
10 100 1,000 10,000
Population Density (people/km"2)
10 100 1,000 10,000
Population Density (people;'km'2)
10 100 1,000 10,000
Population Density (people/kmA2)
10 100 1,000 10,000
Population Density {people:'km"2)
10 100 1,000 10,000
Population Density (people/km^)
•» Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/km"-2)
Exposure Area
10 100 1,000 10,000
Population Density (people/kin'2)
Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/'kny'2)
Exposure Area
Figure 4D-120. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Dallas versus population and population density.
4D-152
-------
Denver 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Denver 2008 - 2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-15 -10
-15 -10
Change in O3 (ppb)
Change inO3 (ppb)
Change in O3 (ppb)
Change in OS (ppb)
I
CO
-15 -10 -5 0
Change in O3 (ppb)
-15 -10 -5 0
Change inO3 (ppb)
-15 -10 -5 0
Change inO3 (ppb)
-15 -10 -5 0
Change in O3 (ppbj
D Epi Study Area H GBSA
Exposure Area
Denver 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
P Epi Study Area • CBSA n Exposure Area
Denver 2008 -2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
10 100 1,000 10,000
Population Density (people/km"2)
10 100 1,000 10,000
Population Density (people/km'2)
10 100 1,000 10,000
Population Density (people/kmA2)
10 100 1,000 10,000
Population Density {people:'km"2)
Q. O o
O v
tn
10 100 1,000 10,000
Population Density (people/km^)
•» Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/km*2)
Exposure Area
10 100 1,000 10,000
Population Density (people/kin'2)
Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/'kny'2)
Exposure Area
Figure 4D-121. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Denver versus population and population
density.
4D-153
-------
Detroit 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Detroit 2008-2010
Change from Observed to 70 ppb Change from 70 ppb to 65 ppb
If-
---
tf) O
(D =
-g, 1
I 5'
Change in O3 (ppb)
-20 -15 -10 -5
Change inO3 (ppbj
-15 -10
Change in O3 (ppb)
Change in OS (ppb)
II.
-20 -15 -10 -505
Change in O3 (ppb)
D Epi Study Area H GBSA
-20 -15 -10 -5 0
Change inO3 (ppb)
-15 -10 -5 0
Change inO3 (ppb)
Exposure Area
D Epi Study Area P CBSA
-15 -10 -5 0
Change in OS (ppb)
n Exposure Area
Detroit 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Detroit 2008-2010
Change from Observed to 70 ppb Change from 70 ppb to 65 ppb
Q c
H,
-^ R '
CT 0
^? a *"
"rt o
rs ,,
c ^
^
<
8-
10 100 1,000 10,000
Population Density (people/km"2)
10 100 1,000 10,000
Population Density (people/km'2)
10 100 1,000 10,000
Population Density (people/knr'2)
10 100 1,000 10,000
Population Density (people.>'krn"2)
*•
10 100 1,000 10,000
Population Density (people/km^)
•» Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/km"-2)
Exposure Area
10 100 1,000 10,000
Population Density (people/kiri*2)
Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/'kny'2)
Exposure Area
Figure 4D-122. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Detroit versus population and population
density.
4D-154
-------
Houston 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Houston 2008- 2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
£ I
I
=5 ^_
El
<
-g, 1
I 5
5 a
if'
-25 -20 -15 -10 -5 0 5 10 -25 -20 -15 -10 -5 0 5 10
Change in O3 (ppb) Change in Q3 (ppb)
Change in O3 (ppb)
Change in O3 (ppb)
-25-20-15-10-5 0 5 10 -25 -20 -15 -10 -5 0 5 10
Change in O3 (ppb) Change inO3 (ppb)
-15 -10 -5 0
Change in O3 (ppb)
D Epi Study Area H GBSA
Exposure Area
D Epi Study Area • CBSA
-15 -10 -5 0
Change in OS (ppb)
n Exposure Area
Houston 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Houston 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
%**m
a
10 100 1,000 10,000
Population Density (people/kmn2)
10 100 1,000 10,000
Population Density (people;'km'2)
10 100 1,000 10,000
Population Density (people/kmA2)
10 100 1,000 10,000
Population Density {people:'km"2)
'
10 100 1,000 10,000
Population Density (people/knifl2)
« Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/km*2)
Exposure Area
10 100 1,000 10,000
Population Density (people/kin'2)
Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/'kny'2)
Exposure Area
Figure 4D-123. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Houston versus population and population
density.
4D-155
-------
Los Angeles 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
£ I
I
=5 ^ -
t=
C
<
-50 -40 -30 -20 -10 0 10
Change in OS (ppb)
-50 -40 -30 -20 -10 0
Change in O3 (ppb)
-50 -40 -30 -20 -10 0 10
Change in O3 (ppb)
D Epi Study Area H CBSA
-50 -40 -30 -20 -10 0 10
Change inO3 (ppb)
Exposure Area
Los Angeles 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
I.:
Is"
10 O'
I '
»•••£** •:'•
S-
10 100 1,000 10,000
Population Density (people:'km"2)
10 100 1,000 10,000
Population Density (people;'km'2)
s-
10 100 1,000 10,000
Population Density (people/knifl2)
« Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/km*2)
Exposure Area
Los Angeles 2008 - 2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-50 -40 -30 -20 -10 0 10
Change inO3 (ppb)
-50 -40 -30 -20-10 0 10
Change in O3 (ppb)
-50 -40 -30 -20 -10 0 10
Change in O3 (ppb)
D Epi Study Area • CBSA
-50 -40 -30 -20 -10 0 10
Change in OS (ppb)
n Exposure Area
Los Angeles 2008 - 2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
10 100 1,000 10,000
Population Density (people/kmA2)
10 100 1,000 10,000
Population Density (people.>'krn"2)
Q. o
QJ
-------
New York 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
New York 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Cfl O *
(D =:
I
-40 -30 -20 -10 0 10 -40 -30 -20 -10 0 10
Change in O3 (ppb) Change in O3 (ppb)
to -
-20 -10 0
Change in O3 (ppb)
-30 -20 -10 C
Change in O3 (ppb)
I
-40 -30 -20 -10 0 10 -40 -30 -20 -10 0 10
Change in O3 (ppb) Change inO3 (ppb)
D Epi Study Area H GBSA
Exposure Area
-30 -20 -10 0 10
Change inO3 (ppb)
D Epi Study Area • CBSA
-30 -20 -10 0
Change in O3 (ppbj
n Exposure Area
New York 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
New York 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
10 100 1,000 10,000
Population Density (people/km"2)
10 100 1,000 10,000
Population Density (people;'km'2)
10 100 1,000 10,000
Population Density (people/kmA2)
10 100 1,000 10,000
Population Density (people:'km*2)
10 100 1,000 10,000
Population Density (people/km^)
•» Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/km*2)
Exposure Area
0 100 1,000 10,000
Population Density (people/kin'2)
Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/'kny'2)
Exposure Area
Figure 4D-125. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for New York versus population and population
density.
4D-157
-------
Philadelphia 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Philadelphia 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Cfl O
'krn"2)
*
10 100 1,000 10,000
Population Density (people/km^)
•» Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/km"-2)
Exposure Area
10 100 1,000 10,000
Population Density (people/kmA2)
Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/'kny'2)
Exposure Area
Figure 4D-126. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Philadelphia versus population and population
density.
4D-158
-------
Sacramento 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Sacramento 2008 - 2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
S I
I
O -1
•25 -20 -15 -10 -5 0 5 -25 -20 -15 -10 -5 0 5
Change in O3 (ppb) Change in Q3 (ppb)
I g~-
I
-30 -25 -20 -15 -10 -505
Change inO3 (ppb)
-30 -25 -20 -15 -10 -505
Change in O3 (ppb)
I
i)
8-
^C'JkirtSHttr.'
10 100 1,000 10,000
Population Density (people/kmn2)
10 100 1,000 10,000
Population Density (people;'km'2)
10 100 1,000 10,000
Population Density (people/kmA2)
10 100 1,000 10,000
Population Density (people:'km"2)
8-
Q. O :
QJ
CO
8?-
10 100 1,000 10,000
Population Density (people/knifl2)
« Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/km*2)
Exposure Area
10 100 1,000 10,000
Population Density (people/kiri*2)
Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/'kny'2)
Exposure Area
Figure 4D-127. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Sacramento versus population and population
density.
4D-159
-------
St. Louis 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
St. Lou is 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
S I
I
-15 -10
-15 -10
-15 -10
-15 -10
Change in O3 (ppb)
Change lnO3 (ppb)
Change in O3 (ppb)
Change in OS (ppb)
I
CO
-15 -10 -5 0
Change in O3 (ppb)
Epi Study Area • CBSA
-15 -10 -5 0
Change inO3 (ppb)
D Exposure Area
-15 -10 -5 0
Change inO3 (ppb)
P Epi Study Area • CBSA
-15 -10 -5 0
Change in O3 (ppbj
n Exposure Area
St. Louis 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
St. Louis 2008 -2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
10 100 1,000 10,000
Population Density (people/kmn2)
10 100 1,000 10,000
Population Density (people;'km'2)
10 100 1,000 10,000
Population Density (people/kmA2)
10 100 1,000 10,000
Population Density (people:'km"2)
10 100 1,000 10,000
Population Density (people/km^)
•» Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/km*2)
Exposure Area
10 100 1,000 10,000
Population Density (people/kin'2)
Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/'kny'2)
Exposure Area
Figure 4D-128. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for St. Louis versus population and population
density.
4D-160
-------
Washington 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Washington 2008 - 2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
-25 -20 -15 -10 -5 0
Change in O3 (ppb)
D Epi Study Area H CBSA
-25 -20 -15 -10 -5 0
Change inO3 (ppb)
D Exposure Area
-15 -10 -5
Change inO3 (ppb)
-15 -10
D Epi Study Area P CBSA
Change in OS (ppb)
Exposure Area
Washington 2006 - 2008
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
Washington 2008-2010
Change from Observed to 75 ppb Change from 75 ppb to 65 ppb
10 100 1,000 10,000
Population Density (people.'km"2)
10 100 1,000 10,000
Population Density (people;'kiTr'2)
10 100 1,000 10,000
Population Density (people.''knr*2)
10 100 1,000 10,000
Population Density (people.>'krn"2)
Q. O O
O v
tn
8,
|
O O
10 100 1,000 10,000
Population Density (people/km^)
•» Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/km*2)
Exposure Area
10 100 1,000 10,000
Population Density (people/km*2)
Epi Study Area • CBSA
10 100 1,000 10,000
Population Density (people/'kny'2)
Exposure Area
Figure 4D-129. Changes in VNA estimates of annual 4th highest MDA8 and May - September
average MDA8 based on HDDM adjustments for Washington, D.C. versus population and
population density.
4D-161
-------
4D-4.7 Comparing Air Quality Adjustments Based on NOX Reductions Only to Air
Quality Adjustments Based on NOX and VOC Reductions
As mentioned in section 4D-4D-3, HDDM-adjustment scenarios could be carried out
either by applying across-the-board reductions in U.S. anthropogenic NOx emissions or by
applying across-the-board reduction in U.S. anthropogenic NOx and VOC emissions (with equal
percentage reductions for the two precursors). The core analysis applied the NOx-only
reductions to all urban study areas except Chicago and Denver for which combined NOx and
VOC reductions were applied. However, in order to address the question of how the choice of
NOx-only emissions reductions affects estimated air quality distributions for adjustments to the
existing and alternative standard levels we performed sensitivity analyses for seven cities:
Denver, Detroit, Houston, Los Angeles, New York, Philadelphia, and Sacramento. For these
seven cities, we performed both NOx-only and N/VOC emissions reductions scenarios for all
standard levels and compared the resulting air quality results. Although in six of these cities the
additional VOC reductions did not result in needing lower percentage NOx reductions, we
recognize that some VOC reductions are likely to occur in future years due to on-the-books
mobile source rules. However, the large VOC cuts applied here are larger than expected
reductions from these rules. Several caveats need to be noted for this analysis. First, this is
meant as a sensitivity test only and not as a potential realistic alternative control scenario.
Second, because the methodology described in Section 4D-3 restricts the calculations to equal
percentage cuts in NOx and VOC, this sensitivity does not necessarily identify the optimal
combination of NOx and VOC reduction levels. Table 4D-20 shows the percentage reductions
that were applied in each scenario for each of the seven cities.
4D-162
-------
Table 4D-20. Comparison of NOx-only and NOx/VOC emission reductions applied in
sensitivity analyses for nine urban areas.
City
Chicago
Dallas
Denver
Detroit
Houston
Los Angeles
(95% LB)
New York
(95% LB)
Philadelphia
Sacramento
Years
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
2006-2008
2008-2010
NOx-only
75
ppb
41%
N/A
50%
50%
54%
24%
59%
N/A
62%
42%
87%
87%
64%
52%
54%
42%
63%
64%
70
ppb
54%
42%
57%
58%
67%
53%
69%
54%
68%
53%
89%
89%
74%
67%
61%
52%
70%
71%
65
ppb
65%
56%
65%
64%
76%
70%
76%
66%
74%
63%
91%
91%
92%
89%
68%
61%
76%
77%
60
ppb
75%
67%
72%
71%
89%
89%
84%
78%
82%
75%
93%
93%
N/A
N/A
74%
68%
84%
84%
NOx/VOC
75
ppb
19%
N/A
51%
50%
51%
15%
60%
N/A
65%
40%
95%
93%
60%
41%
57%
37%
65%
65%
70
ppb
52%
27%
60%
59%
65%
46%
73%
53%
73%
52%
96%
95%
71%
55%
65%
52%
73%
74%
65
ppb
66%
55%
69%
68%
78%
64%
85%
69%
81%
65%
98%
97%
89%
86%
71%
62%
80%
81%
60
ppb
80%
70%
77%
76%
87%
87%
90%
85%
87%
85%
99%
98%
N/A
N/A
79%
72%
88%
88%
4D-163
-------
Figure 4D-130 through Figure 4D-136 show boxplots of composite monitor daily
maximum 8-hr Os values for recent conditions (base) and each of the eight adjustment scenarios
(NOx-only and NOx/VOC for 75, 70, 65, and 60 ppb standard levels) in the seven cities
evaluated. A range of results can be seen in different urban areas. Denver, Houston, Los
Angeles, and New York showed the largest difference between Os concentrations in NOx-only
versus NOx/VOC scenarios while Detroit, Philadelphia and Sacramento had relatively less
difference in the Os distributions estimated in two types of scenarios. In all cities, the NOx-only
and NOx/VOC scenarios had very similar Os concentrations at the upper end of the distribution
(top whiskers and outlier dots in the boxplots). This is not surprising since the adjustment
scenarios were implemented to obtain identical 4th high Os concentrations. Mid-range Os
concentrations (25th-75th percentiles) were generally lower in the NOx/VOC adjustment scenarios
than the NOx-only adjustment scenarios for the same standard level with the exception of New
York.
The reduction of mid-range Os concentrations in the NOx/VOC scenarios compared to the
NOx-only scenarios tended to be larger in 2007 than in 2009 and tended to be larger at lower
alternate standard levels. In most urban study areas the reduction in mid-level Os in the
NOx/VOC scenario compared to the NOx-only scenario was modest but in Los Angeles it was
significant. The change in mid-range Os concentrations between the two sets of scenarios was so
dramatic in Los Angeles that in many cases the 75th percentile concentration in the NOx/VOC
scenario was lower than the 25th percentile concentration in the comparable NOx-only scenario.
The most dramatic differences between the NOx-only and the NOx/VOC scenarios occurred at
the low end of the Os distribution. In all urban study areas, there were smaller increases in Os at
low Os concentrations in the NOx/VOC scenario when compared to the NOx-only scenario for
the same standard level. This is especially evident for extreme low concentrations (bottom
whiskers for blue NOx-only scenarios are much higher than bottom whiskers for red NOx/VOC
scenarios in the boxplots) but can also be seen in the 25th percentile Os values which are
represented by the bottom of the boxes. The reductions were most apparent for Denver,
Houston, Los Angeles, New York, and Philadelphia but were more modest in Detroit and
Sacramento.
4D-164
-------
Z & S, June-August, 2007
Z & S, April-October, 2007
NOxonly
NOx/VOC
base 75 70 65 60
MSA, June-August, 2007
Z & S, June-August, 2009
Z & S, April-October, 2009
NOxonly
NOxWOC
NOxonly
NOxA/OC
base 75 70 65 60
MSA, April-October, 2007
base 75 70 65
MSA, June-August, 2009
base 75 70 65 60
MSA, April-October, 2009
base 75
70
65
NOxonly
NOx/VOC
NOx only
NOx/VOC
70
ffi
60
base 75
70
65
base 75
70
65
60
Figure 4D-130. Composite monitor daily maximum 8-hour Os values for Denver based on
observed and adjusted air quality for the NOx-only and NOx/VOC scenarios. Boxes
represent the median and quartiles, x's represent mean values, whiskers extend up to 1.5x
the inter-quartile range from the boxes, and circles represent outliers.
Z & S, June-August, 2007
Z & S, April-October, 2007
base 75 70 65 60
MSA, June-August, 2007
Z & S, June-August, 2009
Z & S, April-October, 2009
NOx only
NOx/VOC
NOxonly
NOxA/OC
base 75 70 65 60
MSA, April-October, 2007
base 75 70 65
MSA, June-August, 2009
base 75 70 65 60
MSA, April-October, 2009
base 75
70
65
. 0
NOxonly
NOx/VOC
NOxonly
NOx/VOC
70
6!
60
base 75
70
base 75
70
Figure 4D-131. Composite monitor daily maximum 8-hour Os values for Detroit based on
observed and adjusted air quality for the NOx-only and NOX/VOC scenarios. Boxes
represent the median and quartiles, x's represent mean values, whiskers extend up to 1.5x
the inter-quartile range from the boxes, and circles represent outliers.
4D-165
-------
Z & S, June-August, 2007
Z & S, April-October, 2007
Z & S, June-August, 2009
NOxonly
NOx/VOC
base 75 70 65 60
MSA, June-August, 2007
Z & S, April-October, 2009
NOxonly
NOxA/OC
base 75 70 65 60
MSA, April-October, 2007
base 75 70 65
MSA, June-August, 2009
base 75 70 65 60
MSA, April-October, 2009
NOxonly
NOxWOC
base 75
70
65
base 75
70
65
60
Figure 4D-132. Composite monitor daily maximum 8-hour Os values for Houston based on
observed and adjusted air quality for the NOx-only and NOX/VOC scenarios. Boxes
represent the median and quartiles, x's represent mean values, whiskers extend up to 1.5x
the inter-quartile range from the boxes, and circles represent outliers.
Z & S, June-August, 2007
Z & S, April-October, 2007
Z & S, June-August, 2009
Z & S, April-October, 2009
base 75 70 65 60
MSA, June-August, 2007
base 75 70 65 60
MSA, April-October, 2007
base 75 70 65 60
MSA, June-August, 2009
base 75 70 65 60
MSA, April-October, 2009
NOxonly
NOx/VOC
NOxonly
NOxA/OC
75
70
65
base 75
70
65
h II
base 75
base 75
70
6fi
Figure 4D-133. Composite monitor daily maximum 8-hour Os values for Los Angeles
based on observed and adjusted air quality for the NOx-only and NOX/VOC scenarios.
Boxes represent the median and quartiles, x's represent mean values, whiskers extend up
to 1.5x the inter-quartile range from the boxes, and circles represent outliers.
4D-166
-------
Z & S, June-August, 2007
Z & S, April-October, 2007
Z & S, June-August, 2009
Z & S, April-October, 2009
NOxonly
NCMVOC
NOxonly
NOxWOC
NOxonly
NOxA/OC
base 75 70 65 60
MSA, June-August, 2007
base 75 70 65 60
MSA, April-October, 2007
base 75 70 65
MSA, June-August, 2009
base 75 70 65 60
MSA, April-October, 2009
NOxonly
NOx/VOC
NOx only
NOxA/OC
base 75
70
65
base 75
70
ffi
60
base 75
70
65
base 75
70
65
60
Figure 4D-134. Composite monitor daily maximum 8-hour Os values for New York based
on observed and adjusted air quality for the NOx-only and NOX/VOC scenarios. Boxes
represent the median and quartiles, x's represent mean values, whiskers extend up to 1.5x
the inter-quartile range from the boxes, and circles represent outliers.
Z & S, June-August, 2007
Z & S, April-October, 2007
Z & S, June-August, 2009
base 75 70 65 60
MSA, June-August, 2007
Z & S, April-October, 2009
NOxonly
NOx/VOC
base 75 70 65 60
MSA, April-October, 2007
75
70
65
base 75 70 65 60
MSA, June-August, 2009
base 75 70 65 60
MSA, April-October, 2009
NOxonly
NOxWOC
NOxonly
NOxA/OC
70
65
h II
base 75
base 75
70
65
6fi
Figure 4D-135. Composite monitor daily maximum 8-hour Os values for Philadelphia
based on observed and adjusted air quality for the NOx-only and NOX/VOC scenarios.
Boxes represent the median and quartiles, x's represent mean values, whiskers extend up
to 1.5x the inter-quartile range from the boxes, and circles represent outliers.
4D-167
-------
Z & S, June-August, 2007
Z & S, April-October, 2007
Z & S, June-August, 2009
base 75 70 65 60
MSA, June-August, 2007
base 75 70 65 60
MSA, April-October, 2007
Z & S, April-October, 2009
base 75 70 65
MSA, June-August, 2009
base 75 70 65 60
MSA, April-October, 2009
base 75
70
65
base 75
70
ffi
60
base
base 75
70
65
60
Figure 4D-136. Composite monitor daily maximum 8-hour Os values for Sacramento
based on observed and adjusted air quality for the NOx-only and NOX/VOC scenarios.
Boxes represent the median and quartiles, x's represent mean values, whiskers extend up
to 1.5x the inter-quartile range from the boxes, and circles represent outliers.
Figure 4D-137 through Figure 4D-143 show maps of the 2006-2008 observed and
adjusted April-October seasonal mean of the daily maximum 8-hr Os value at monitor locations
in each of the urban study areas. Adjusted values are shown for the 75 ppb and 65 ppb NOx-only
and NOx/VOC adjustment scenarios. As described earlier in this appendix and in FtREA Chapter
4, these figures show that for observed 2006-2008 Os values, the seasonal mean of the daily
maximum 8-hr Os concentration was suppressed in the highly urbanized areas compared to
surrounding locations. In these figures, it is clear that Os concentrations are lower in the center
city areas for all cities except Sacramento. The NOx-only adjustment scenarios resulted in three
types of behavior as demonstrated by the top panels on these maps. First, in Denver, Houston,
Los Angeles, and New York the NOx-only adjustment cases resulted in increasing seasonal mean
concentrations in center city locations where Os was suppressed in observations and decreasing
seasonal mean concentrations in outlying areas where Os was higher in the observations. This
trend is so dramatic in Los Angeles and New York that the spatial gradient becomes inverted in
the adjustment cases compared to the observed values (i.e. highest concentrations in urban core
areas and lower concentrations in surrounding areas). The second type of response occurred in
Detroit where there was little change at the monitor in the center of Wayne County but seasonal
mean Os at the outlying monitors decreased in the NOx-only adjustment cases. Finally, in
Sacramento and Philadelphia, the NOx-only adjustment cases cause relatively small decreases in
4D-168
-------
seasonal mean Os throughout the area. When comparing the NOx-only to the NOx/VOC
adjustment scenarios three types of patterns emerge. In Denver and Los Angeles, the NOx/VOC
adjustment scenarios lead to lower seasonal mean Os concentrations than the equivalent NOx-
only adjustment scenarios in the urban core areas but looked similar to the NOx-only adjustment
scenario in the outlying areas. In Houston, Detroit, Philadelphia, and Sacramento, seasonal mean
Os concentrations in equivalent NOx-only and NOx/VOC adjustment scenarios were very similar
at monitors throughout the areas although the NOx/VOC adjustment scenarios lead to slightly
lower concentrations. Finally, in New York the 75 ppb adjustment scenarios were similar for the
NOx/VOC and NOx-only cases but for the 65 ppb adjustment scenarios the NOx/VOC case
actually lead to higher seasonal mean Os concentration in the urban core area than the NOx-only
case.
An evaluation of the composite monitor and spatial plot maps leads to several general
conclusions for this sensitivity analysis. First, the NOx/VOC reduction scenarios tended to
mitigate increases that occurred in the NOx-only scenario at the lower end of the Os distribution.
Second the effect on the NOx/VOC scenario versus the NOx-only scenario was less dramatic for
mid-range Os concentrations and varied from city to city. The NOx/VOC scenarios lead to lower
mid-range Os concentrations than the NOx-only scenarios except in the case of New York.
Third, the high end Os concentrations at various standard levels were similar in the NOx-only and
the NOx/VOC scenarios. Finally, the VOC reductions tended to have more impact in urban core
areas and relatively little impact in outlying areas. The effects of these air quality sensitivity
analyses on risk are evaluated and discussed in HREA Chapter 7.
4D-169
-------
75 ppb rollback- NOx only 65 ppb rollback- NOx only
April-October Average MDA8
Observed 2006 - 2008
_L
75 ppb rollback- NOx/VOC 65 ppb rollback- NOx/VOC
Figure 4D-137. April-October seasonal average of daily maximum 8-hour Os values at
Denver area monitor locations for observed 2006-2008 conditions (left panel), 75 ppb
adjustment scenarios (middle panels), and 65 ppb adjustment scenarios (right panels).
NOx-only adjustments are shown in top panels, NOx/VOC adjustments are shown in
bottom panels.
4D-170
-------
75 ppb rollback- NOx only 65 ppb rollback- NOx only
tf
June-August Average MDA8
Observed 2006 - 2008
\
75 ppb rollback- NOx/VOC 65 ppb rollback- NOx/VOC
.
Figure 4D-138. April-October seasonal average of daily maximum 8-hour Os values at
Detroit area monitor locations for observed 2006-2008 conditions (left panel), 75 ppb
adjustment scenarios (middle panels), and 65 ppb adjustment scenarios (right panels).
NOx-only adjustments are shown in top panels, NOX/VOC adjustments are shown in
bottom panels.
4D-171
-------
75 ppb rollback- NOx only 65 ppb rollback- NOx only
April- October Average MDA8
Observed 2006 - 2008
75 ppb rollback- NOx/VOC 65 ppb rollback- NOx/VOC
Figure 4D-139. April-October seasonal average of daily maximum 8-hour Os values at
Houston area monitor locations for observed 2006-2008 conditions (left panel), 75 ppb
adjustment scenarios (middle panels), and 65 ppb adjustment scenarios (right panels).
NOx-only adjustments are shown in top panels, NOX/VOC adjustments are shown in
bottom panels.
4D-172
-------
75 ppb rollback- NOx only 65 ppb rollback- NOx only
April-October Average MDA8
Observed 2006 - 2008
£
75 ppb rollback- NOx/VOC 65 ppb rollback- NOx/VOC
Figure 4D-140. April-October seasonal average of daily maximum 8-hour Os values at Los
Angeles area monitor locations for observed 2006-2008 conditions (left panel), 75 ppb
adjustment scenarios (middle panels), and 65 ppb adjustment scenarios (right panels).
NOx-only adjustments are shown in top panels, NOX/VOC adjustments are shown in
bottom panels.
4D-173
-------
75 ppb rollback- NOx only 65 ppb rollback- NOx only
'—r~~i l-'— —~ "—
June-August Average MDA8
Observed 2006 - 2008
75 ppb rollback-NOx/VOC
- v —~~~~ ' ~~
65 ppb rollback-NOx/VOC
Figure 4D-141. April-October seasonal average of daily maximum 8-hour Os values at
New York area monitor locations for observed 2006-2008 conditions (left panel), 75 ppb
adjustment scenarios (middle panels), and 65 ppb adjustment scenarios (right panels).
NOx-only adjustments are shown in top panels, NOX/VOC adjustments are shown in
bottom panels.
4D-174
-------
75 ppb rollback- NOx onl
65 ppb rollback- NOx only
'
June-August Average MDA8
Observed 2006 - 2008
75 ppb rollback- NOx/VOC 65 ppb rollback- NOx/VOC
Figure 4D-142. April-October seasonal average of daily maximum 8-hour Os values at
Philadelphia area monitor locations for observed 2006-2008 conditions (left panel), 75 ppb
adjustment scenarios (middle panels), and 65 ppb adjustment scenarios (right panels).
NOx-only adjustments are shown in top panels, NOX/VOC adjustments are shown in
bottom panels.
4D-175
-------
75 ppb rollback- NOx only 65 ppb rollback- NOx only
April-October Average MDA8
Observed 2006 - 2008
75 ppb rollback- NOx/VOC 65 ppb rollback- NOx/VOC
Figure 4D-143. April-October seasonal average of daily maximum 8-hour Os values at
Sacramento area monitor locations for observed 2006-2008 conditions (left panel), 75 ppb
adjustment scenarios (middle panels), and 65 ppb adjustment scenarios (right panels).
NOx-only adjustments are shown in top panels, NOX/VOC adjustments are shown in
bottom panels.
4D-176
-------
4D-5, REFERENCES
Appel, K.W., Gilliland, A.B., Sarwar, G., Gilliam, R.C. (2007). Evaluation of the Community
Multiscale Air Quality (CMAQ) model version 4.5: Sensitivities impacting model
performance Part I - Ozone. Atmospheric Environment, 41: 9603-9615.
Appel, K.W., Bhave, P.V., Gilliland, A.B., Sarwar, G., Roselle, S.J. (2008). Evaluation of the
community Multiscale air quality (CMAQ) model version 4.5: Sensitivities impacting
model performance; Part II - particulate matter. Atmospheric Environment, 42: 6057-
6066.
Baker, K. and P. Dolwick. Meteorological Modeling Performance Evaluation for the Annual
2005 Eastern U.S. 12-km Domain Simulation, USEPA/OAQPS, February 2, 2009.
Baker, K. and P. Dolwick. Meteorological Modeling Performance Evaluation for the Annual
2005 Continental U.S. 36-km Domain Simulation, USEPA/OAQPS, February 2, 2009.
Byun, D.W., and Ching, J.K.S., eds., 1999. Science algorithms of EPAModels-3 Community
Multiscale Air Quality (CMAQ modeling system, EPA/600/R-99/030, Office of Research
and Development).
Byun, D., Schere, K.L. (2006). Review of the governing equations, computational algorithms,
and other components of the models-3 Community Multiscale Air Quality (CMAQ)
modeling system. Applied Mechanics Reviews, 59: 51-77'.
Cohan, D.S., Hakami, A., Hu, Y., Russell, A.G. (2005). Nonlinear response of ozone to
emissions: Source apportionment and sensitivity analysis. Environmental Science &
Technology, 39: 6739-6748.
Cohan, D.S., Napelenok, S.L. (2011). Air quality response modeling for decision support.
Atmosphere, 2: 407-425; doi:10.3390/atmos2030407.
Dunker, A.M. (1984). The decoupled direct method for calculating sensitivity coefficients in
chemical kinetics. Journal of Chemical Physics, 81; 2385-2393.
Foley, K.M., Roselle, S.J., Appel, K.W., Bhave, P.V., Pleim, I.E., Otte, T.L., Mathur, R., Sarwar,
G, Young, J.O., Gilliam, R.C., Nolte, C.G, Kelly, J.T., Gilliland, A.B., Bash, J.O.
(2010). Incremental testing of the Community Multiscale Air Quality (CMAQ) modeling
system version 4.7, Geoscientific Model Development, 3: 205-226.
Grell, G., J. Dudhia, and D. Stauffer (1994). A Description of the Fifth-Generation Penn
State/NCAR Mesoscale Model (MM5), NCAR/TN-398+STR., 138 pp, National Center
for Atmospheric Research, Boulder CO.
Hakami, A., Odman, M.T., Russell, A.G. (2003). High-order direct sensitivity analysis of
multidimensional air quality models. Environmental Science & Technology, 37: 2442-
2452.
Hakami, A., Odman, M.T., Russell, A.G. (2004). Nonlinearity in atmospheric response: A direct
sensitivity analysis approach. Journal of Geophysical Research, 109: D15303;
doi: 10.1029/2003 JD004502.
Houyoux, M.R., Vukovich, J.M., Coats, C.J., Wheeler, N.J.M., Kasibhatla, P.S. (2000). Emission
inventory development and processing for the Seasonal Model for Regional Air Quality
(SMRAQ) project. Journal of Geophysical Research-Atmospheres, 105: 9079-9090.
4D-177
-------
Koo, B., Dunker, A.M., Yarwood, G. (2007). Implementing the decoupled direct method for
sensitivity analysis in a particulate matter air quality model. Environmental Science &
Technology, 41: 2847-2854.
Napelenok, S.L., Cohan, D.S., Hu, Y., Russell, A.G. (2006). Decoupled direct 3D sensitivity
analysis for particulate matter (DDM-3D/PM). Atmospheric Environment, 40: 6112-
6121.
National Research Council of the National Academies (2008). Estimating Mortality Risk
Reduction and Economic Benefits from Controlling Ozone Air Pollution. The National
Academies Press, Washington, D.C
Pierce, T.; Geron, C.; Bender, L.; Dennis, R.; Tonnesen, G. Guenther, A. (1998). Influence of
increased isoprene emissions on regional ozone modeling. J. Geophys. Res., 103: 25611-
25629.
Seinfeld, J.H. and S.N. Pandis. (1998). Atmospheric chemistry and physics from air pollution to
climate change. John Wiley & Sons, Inc, New York, 1998.
Simon, H., Baker, K.B., Akhtar, F., Napelenok, S.L., Possiel, N., Wells, B., Timin, B. (2013) A
direct sensitivity approach to predict hourly ozone resulting from compliance with the
national ambient air quality standard, Environmental Science and Technology, 47, 2304-
2313.
U.S. EPA, 2007. Guidance on the Use of Models and Other Analyses for Demonstrating
Attainment of Air Quality Goals for Ozone, PM2.5, and Regional Haze, Research
Triangle Park. EPA-454/B-07-002: http://www.epa.gov/ttn/scram/guidance/guide/final-
03-pm-rh-guidance.pdf.
U.S. EPA, Emissions Modeling for the Final Mercury and Air Toxics Standards Technical
Support Document, 201 la, EPA-45A/R-11-011:
http://epa.gov/ttn/chi ef/emch/toxics/MATS_Final_Emissions_Modeling_TSD_9Dec2011
.pdf.
U.S. EPA, Air Quality Modeling Technical Support Document: Final EGU NESHAP, 201 Ib,
EPA-45A/R-11-009.
U.S. EPA, Health Risk and Exposure Assessment for Ozone: First External Review Draft, 2012,
EPA 452/P-12-001.
U.S. EPA Integrated Science Assessment for Ozone and Related Photochemical Oxidants, 2013,
EPA600/R-10/076F
Wilks, D.S. Statistical Methods in the Atmospheric Sciences, Second Edition, Elsevier,
Burlington, MA
Yang, Y.-J., Wilkinson, J.G., Russell, A.G. (1997). Fast, direct sensitivity analysis of
multidimensional photochemical models. Environmental Science & Technology, 31:
2859-2868.
Zhang, W., Capps, S.L., Hu, Y., Nenes, A., Napelenok, S.L., Russell, A.G. (2012). Development
of the high-order decoupled direct method in three dimensions for particulate matter:
Enabling advanced sensitivity analysis in air quality models. Geoscientific Model
Development, 5: 355-368.
4D-178
-------
This page left intentionally blank
4D-179
-------
APPENDIX 4E
Evaluation of Seattle Air Quality
List of Figures
Figure 4E-1. All-site maximum hourly Os (ppb) in Seattle by year and month. This figure
combines all hourly Os data from 11 sites in the Seattle area and plots the highest-
site value at each hour for Mar-Sep. Boxes represent the median and quartiles,
whiskers extend up to 1.5x the inter-quartile range from the boxes, and circles
represent outliers 4E-1
Figure 4E-2. Map of Seattle area monitoring sites used to create the ozone surfaces for the
exposure assessment. Dark brown site is the most urban location (in a residential
area of downtown Seattle), orange sites are suburban, sites in the far northern and
southern portions of the area are plotted in blue and green respectively, dark green
site is a high elevation site 4E-2
Figure 4E-3. Map with aerial photography overlay showing where monitors are located with
respect to highly urbanized (gray) areas versus suburban and rural areas (green).
Downtown Seattle site is circled in orange 4E-3
Figure 4E-4. Example of a day with large spatial ozone gradients in the Seattle area.
Maximum 8-hr daily maximum ozone value for July 11, 2007 4E-4
Figure 4E-5. Example of a day with large spatial ozone gradients in the Seattle area.
Maximum 8-hr daily maximum ozone value for July 26, 2007 4E-5
Figure 4E-6. Example of a day with large spatial ozone gradients in the Seattle area.
Maximum 8-hr daily maximum ozone value for August 29, 2007 4E-5
Figure 4E-7. Example of a day with large spatial ozone gradients in the Seattle area.
Maximum 8-hr daily maximum ozone value for September 11, 2007 4E-6
Figure 4E-8. Cumulative distribution of population living within set distances of an ozone
monitor in each urban study area 4E-7
4E-i
-------
Investigation of the monitoring network and ambient data trends in Seattle raised serious
concerns about the appropriateness of using 2006 data for the purpose of an exposure analysis.
Seattle experienced some high ozone events in 2006 that were substantially higher than any
measured ozone that occurred in the modeled time period (Jan, April-October 2007) (Figure 4E-
1). The model generally predicted ozone disbenefits to NOX reductions in Seattle on the low
ozone days that were predominant in 2007. Because these conditions are not likely to occur on
high ozone days, we had serious concerns about extrapolating 2007 sensitivities to 2006 high
ozone days.
100-
50-
0-
2006
2007
2008
yearfac
2009
2010
monthfac
Figure 4E-1. All-site maximum hourly Os (ppb) in Seattle by year and month. This figure
combines all hourly Os data from 11 sites in the Seattle area and plots the highest-site value
at each hour for Mar-Sep. Boxes represent the median and quartiles, whiskers extend up
to 1.5x the inter-quartile range from the boxes, and circles represent outliers.
4E-1
-------
Further concerns come from the lack of measured urban ozone concentrations in all
years, but especially in 2006. The Seattle monitoring network is mainly focused on downwind
and background areas. There is a single urban monitor in a residential area of downtown Seattle
(Figure 4E-2 and Figure 4E-3). The urban monitor was not in operation from January-
September of 2006, so no urban measurements were made during most of the 2006 ozone
season. Performing a census-tract level exposure analysis for the Seattle and Tacoma urban
areas would not be appropriate based solely on ozone values in outlying rural and suburban
areas.
Figure 4E-2. Map of Seattle area monitoring sites used to create the ozone surfaces for the
exposure assessment. Dark brown site is the most urban location (in a residential area of
downtown Seattle), orange sites are suburban, sites in the far northern and southern
portions of the area are plotted in blue and green respectively, dark green site is a high
elevation site.
4E-2
-------
Figure 4E-3. Map with aerial photography overlay showing where monitors are located
with respect to highly urbanized (gray) areas versus suburban and rural areas (green).
Downtown Seattle site is circled in orange.
4E-2
-------
Additional concerns were raised about the potential insufficient density of the Seattle
ozone monitoring network and subsequently its adequacy for use in a census-tract level exposure
analysis beyond the concerns above for 2006. The Seattle area ozone monitoring network was
designed primarily to address NAAQS compliance. For this purpose it is appropriate to focus on
downwind suburban areas where ozone concentrations are expected to be highest. Modeling
suggests that there are often large spatial gradients of ozone within the Seattle/Tacoma area.
Figure 4E-4 through Figure 4E-7 show 8-hr daily maximum ozone concentration fields for the
Seattle area from the 2007 CMAQ modeling described in Appendix 4A. Spatial fields for
exposure modeling are created using interpolation of monitor data and do not capture any
modeled spatial gradients. Therefore it is expected that at many times, interpolation of the single
Seattle urban monitor with downwind suburban monitors will not adequately represent the ozone
values in highly populated urban areas such a Tacoma. In addition, unlike most other urban
areas, the Seattle monitoring network for the 2006-2010 time period did not include any upwind
monitors (on the West side of Puget Sound).
n
ppbV
40.0
30.0
20.0
218
July 11,2007 0:00:00
Min= 24.0 at (44,243), Max= 80.3 at (45,223)
Figure 4E-4. Example of a day with large spatial ozone gradients in the Seattle area.
Maximum 8-hr daily maximum ozone value for July 11, 2007.
4E-4
-------
20.0
ppbV
July 26,2007 0:00:00
Min= 14.6at(21,242),Max= 62.6 at (65,218)
Figure 4E-5. Example of a day with large spatial ozone gradients in the Seattle area.
Maximum 8-hr daily maximum ozone value for July 26, 2007.
60.0
ppbV
50.0
10.0
30.0
20.0
218
21
65
August 29,2007 0:00:00
M i n = 14.1 at (44,228 ), M ax= 57.4 at (40,225)
Figure 4E-6. Example of a day with large spatial ozone gradients in the Seattle area.
Maximum 8-hr daily maximum ozone value for August 29, 2007.
4E-5
-------
ppbV
40.0
30.0
20.0
218
21
65
September 11,2007 0:00:00
Min= S.4 at (44,228), Max= 67.0 at (45,222)
Figure 4E-7. Example of a day with large spatial ozone gradients in the Seattle area.
Maximum 8-hr daily maximum ozone value for September 11, 2007.
To further analyze the appropriateness of the Seattle monitoring network for the purpose
of performing an exposure assessment, the representativeness of the ozone monitoring network
to total population in Seattle was compared to all other urban study areas (Figure 4E-8). This
analysis showed that the Seattle monitoring network was much less representative of ozone
concentrations where people live than the ozone networks in the other 15 urban study areas. For
instance, while more than 75 % of the area-wide population lived within 20 km of an ozone
monitor in each of the other 15 urban areas, in Seattle only about 55% of the population lives
within 20 km of an ozone monitor. Furthermore, in 13 urban study areas, more than 50% of the
population lives within 10 km of an ozone monitor, in 2 urban study areas more than 50% of the
population lives within 13 km of an ozone monitor, yet in Seattle, more than 50% of the
population lives within 18 km of an ozone monitor. Based on both the sharp spatial gradients
predicted by the model and the much sparser monitor coverage in highly populated areas in
Seattle compared to other urban study areas, it was determined that the monitoring network in
Seattle likely does not capture ozone values representing levels to which large portions of the
population are exposed and therefore it would not be appropriate to rely upon the currently
available Seattle ozone monitors for the purpose of the exposure analysis which requires accurate
fine scale estimates of ozone concentrations where people are exposed.
4E-6
-------
10 15 20 10
Radius Around Monitors (km)
Figure 4E-8. Cumulative distribution of population living within set distances of an ozone
monitor in each urban study area.
4E-7
-------
This page left intentionally blank
4E-8
-------
United States Office of Air Quality Planning and Standards Publication No. EPA-452/R-14-004b
Environmental Protection Health and Environmental Impacts Division August 2014
Agency Research Triangle Park, NC
------- |