Review of the Primary National
Ambient Air Quality Standards for
Nitrogen Dioxide:
Risk and Exposure Assessment
Planning Document

-------
This page left intentionally blank

-------
                                 EPA-452/D-15-001
                                     May 2015
  Review of the Primary National
Ambient Air Quality Standards for
          Nitrogen Dioxide:
  Risk and Exposure Assessment
         Planning Document
           U. S. Environmental Protection Agency
         Office of Air Quality Planning and Standards
          Health and Environmental Impacts Division
         Research Triangle Park, North Carolina 27711

-------
                                    DISCLAIMER
This document has been prepared by staff in the Health and Environmental Impacts Division,
Office of Air Quality Planning and Standards, U.S. Environmental Protection Agency (EPA).
Any findings and conclusions are those of the authors and do not necessarily reflect the views of
the Agency. This document is being circulated to facilitate discussion with the Clean Air
Scientific Advisory Committee and for public comment to inform the EPA's consideration of the
nitrogen dioxide primary National Ambient Air Quality Standards. This information is
distributed for the purposes of pre-dissemination peer review under applicable information
quality guidelines. It does not represent and should not be construed to represent any Agency
determination or policy.

Questions or comments related to this document should be addressed to Dr. Stephen Graham,
U.S. Environmental Protection Agency, Office  of Air Quality Planning and Standards, C539-07,
Research Triangle Park, North Carolina 27711 (email: graham.stephen@epa.gov) and Dr. Scott
Jenkins, U.S. Environmental Protection Agency, Office of Air Quality  Planning and Standards,
C539-06, Research Triangle Park, North Carolina 27711 (email: jenkins.scott@epa.gov).

-------
Table of Contents
1   INTRODUCTION	1-1
    1.1    Hi story of the Primary NO2NAAQS	1-3
    1.2    Approaches to Characterizing Risks	1-4
    1.3    Key Considerations in the Current Review	1-7
    1.4    Organization of This Document	1-9
2   AIR QUALITY AND HEALTH BENCHMARK COMPARISONS	2-1
    2.1    Overview of the Assessment from the Last Review	2-2
            2.1.1     Summary of Results	2-4
            2.1.2     Uncertainties and Limitations	2-5
    2.2    Overview of Information Available in This Review	2-6
            2.2.1     Characterizing Ambient NO2 Concentrations	2-6
            2.2.2     Evidence Informing Health Effect Benchmarks	2-13
            2.2.3     Preliminary Conclusions	2-16
    2.3    Proposed Analytical Approach and Illustrative Example	2-17
            2.3.1     Overview of Proposed Approach	2-17
            2.3.2     Illustrative Example: Characterizing Air Quality and Calculating
                     Benchmark Exceedances in an Example Urban Study Area
                     (Philadelphia)	2-44
3   HUMAN EXPOSURE ASSESSMENT	3-1
    3.1    Overview of Exposure Assessment in the Last Review	3-1
            3.1.1     Key Results	3-3
            3.1.2     Uncertainties and Limitations	3-3
    3.2    Consideration of Newly Available Information	3-7
            3.2.1     Emissions Inventory	3-7
            3.2.2     Air Quality Modeling	3-9
            3.2.3     Exposure Modeling	3-9
    3.3    Summary and Conclusions	3-11
4   HUMAN HEALTH RISK ASSESSMENT	4-1
    4.1    Risk Assessment Based on Information from Controlled Human Exposure Studies.
            	4-1
    4.2    Risk Assessment Based on Information from Epidemiology Studies	4-3
            4.2.1     Overview of the Assessment in the Last Review	4-3
            4.2.2     Consideration of Newly Available Information	4-5

-------
SUMMARY OF CONCLUSIONS AND NEXT STEPS	5-1
5.1    Summary of Preliminary Conclusions	5-1
5.2    Next Steps	5-2
REFERENCES	6-1
                                    in

-------
                                     List of Figures
Figure 1-1.  Risk characterization models employed inNAAQS Reviews	1-5
Figure 1-2.  Conceptual model for risk characterization in the last review of the primary NCh
            NAAQS	1-6
Figure 1-3.  Key considerations for updated quantitative analyses	1-7
Figure 2-1.  Locations of potential study areas to analyze in the air quality assessment
            categorized by availability of near-road monitor data and selection criteria ranking
            scheme	2-29
Figure 2-2.  Distribution of DM1H NOi concentrations (0 - 100th percentile) in the New York
            CBSA for a high-concentration year (1984) versus a low-concentration year (2007)
            adapted from Rizzo (2008) (left panel) and updated comparison with a recent low-
            concentration year (2011) (right panel)	2-32
Figure 2-3.  Predicted and observed NCh concentrations for winds from the west using based on
            data from a Las Vegas NV near-road measurement study. Predicted median (solid),
            predicted 98th and 2nd percentile (dotted), observed median (circles), and observed
            98th and 2nd percentiles (error bars) are shown	2-38
Figure 2-4.  AERMOD modeled maximum 1-hour NOi concentrations with increasing distance
            from a major road in Ft. Lauderdale, FL	2-40
Figure 2-5.  The maximum (left panel) and mean (right panel) number of days per year where
            DM1H NCh concentration was > 100 ppb (top panel) and > 200 ppb and associated
            with 3-year average 98th percentile DM1H NCh concentrations, using 1980-2014
            ambient monitor data	2-44
Figure 2-6.  Locations of the eight active and seven  inactive ambient monitors in the
            Philadelphia CBS A	2-46
Figure 2-7.  Principal components (PC) monitor scores plotted by year along with PC loadings
            plotted by DM1H percentile concentration value using the Philadelphia CBSA NOi
            ambient concentrations (1980-2013)	2-54
Figure 2-8.  Principal components scores plotted by  loadings for each the second (PC2) and
            third (PC3) components derived using the Philadelphia CBSA NOi ambient
            concentrations (1980-2013)	2-55
Figure 2-9.  Distribution of unadjusted (as is) 2011 ambient NCh concentrations, that adjusted
            using a proportional factor alone (all proportional), and that adjusted using a
            combined proportional factor and ratio approach (proportional to 98th percentile,
            non-linear above) in the Philadelphia CBSA at monitor ID 421010004	2-57
Figure 3-1.  Total annual NOx emissions (top) and annual emissions stratified by top four
            sectors (bottom),  2002-2013	3-8
                                              iv

-------
                                        List of Tables
Table 2-1.  Near-road NCh monitoring sites - active as of September 2014	2-9
Table 2-2.  Number of NOi monitors and 2010-2013 design values for CBSAs that have or are
           expected to have ambient concentration data available from newly sited near-road
           monitors	2-24
Table 2-3.  Preliminary near-road and area-wide CBSA data analysis using all available and
           concurrent NCh ambient monitor measurement data (2011-2014)	2-25
Table 2-4.  Number of NCh monitors and 2010-2013 design values for CBSAs that are not
           expected to have ambient concentration data available from newly sited near-road
           monitors	2-26
Table 2-5.  Population estimates for CBSA having at least 800,000 residents in 2013, ordered
           by descending population	2-27
Table 2-6.  Percent increase of on-road compared to near-road using NOi concentrations in
           three studies having both on-road and immediately near-road measurements	2-36
Table 2-7.  Potential factors that could be used to simulate on-road NOi concentrations from
           near-road monitors sited at 10 or 20 meters from a major road, stratified by
           concentration quintiles and meteorological conditions, based on analysis  of Las
           Vegas, NV near-road measurement study data (see Appendix A)	2-39
Table 2-8.  Attributes of active ambient monitors in the Philadelphia CBSA, as of 2014	2-47
Table 2-9.  Attributes of inactive ambient monitors in the Philadelphia CBSA and used in
           analysis of historical NOi concentration trends	2-48
Table 2-10. Proportional adjustment factors calculated for the Philadelphia CBSA, 2011-2014.
           	2-51
Table 2-11. Slope and intercept parameter estimates regressing DM1H concentrations (0-98th
           percentile) from the monitor having the highest design value (ID 421010004) on
           DM1H concentrations measured at five Philadelphia CBSA area-wide monitors....
           	2-52
Table 2-12. Individual monitor-based factors calculated to adjust DM1H ambient NOi
           concentrations above the 98th percentile DM1H in the Philadelphia CBSA	2-56
Table 2-13. Mean and maximum number of days per year ambient monitor NOi concentrations
           (area-wide, near-road, simulated on-road) are at or above selected  1-hour
           benchmark levels in Philadelphia CBSA, unadjusted air quality	2-59
Table 2-14. Mean and maximum number of days per year ambient monitor NOi concentrations
           (area-wide, near-road, simulated on-road) are at or above selected  1-hour
           benchmark levels in Philadelphia CBSA, air quality adjusted to just meet the
           existing standard	2-59
                                              v

-------
Table 2-15. Mean and upper percentile number of days per year ambient monitor NOi
          concentrations (area-wide and on-road) are at or above selected 1-hour benchmark
          levels in the 2008 REA Philadelphia CBSA, 2001-2003 air quality adjusted to just
          meet the existing standard	2-60
Table 3-1.  Summary of 2008 REA qualitative uncertainty analysis for the exposure
          assessment	3-5
Table 5-1.  Tentative schedule for next steps in the review of the primary NCh NAAQS	5-3
                               List of Appendices
APPENDIX A: ANALYSIS OF LAS VEGAS NEAR-ROAD NO2 MEASUREMENT DATA
          AND LOGIT MODEL DEVELOPED TO SIMULATE NO2 ON-ROAD
          CONCENTRATIONS	15 pages
APPENDIX B: CONCENTRATION PLOTS AND MAPS OF SELECTED PHILADELPHIA
          CBSA MONITORS	6 pages
                                         VI

-------
      LIST OF ACRONYMS/ABBREVIATIONS
AADT       Annual average daily traffic
AERMOD    EPA's Air Dispersion Model
AHR        Airway hyperresponsiveness
ANPR       Advanced notice of proposed rulemaking
APEX       EPA's Air Pollutants Exposure model
AQI         Air Quality Index
AQS         EPA's Air Quality System
CAA        Clean Air Act
CAPS        Cavity attenuated phase shift
CASAC      Clean Air Scientific Advisory Committee
CBS A       Core-based statistical area
CFR         Code of Federal Regulations
CO          Carbon monoxide
C-R         Concentration-response
EPA         Environmental Protection Agency
E-R         Exposure-response
FEM        Federal Equivalent Method
FEVi        Forced expiratory volume in one second, volume of air exhaled in first second of
             exhalation
FR          Federal Register
FRM        Federal Reference Method
HA          Hospital admission
HERO       Health and Environmental Research Online
HONO       Nitrous acid
HNOs        Nitric acid
IRP          Integrated Review Plan
ISA         Integrated Science Assessment
|ig/m3        micrograms per cubic meter
m           Meters
ME          Microenvironmental
MSA        Metropolitan statistical area
NAAQS      National ambient air quality standards
NCEA       National Center for Environmental Assessment
NCore       National Core Monitoring Network
NO          Nitric oxide
                                            vii

-------
NCh         Nitrogen dioxide
NOs"         Nitrate
NOx         NO+NO2
NOv         Total oxides of nitrogen (NOx + NOz)
NOz         Reactive oxides of nitrogen (e.g., HNOs, HONO, PAN, particulate nitrates)
Os           Ozone
OAQPS      Office of Air Quality Planning and Standards
OAR         Office of Air and Radiation
OMB        Office of Management and Budget
OR          Odds ratio
ORD         Office of Research and Development
PA          Policy Assessment
PAN         Peroxyacetyl nitrate
PCA         Principal Components Analysis
PM          Particulate matter
PM2.5        In general terms, particulate matter with an aerodynamic diameter less than or
             equal to a nominal 2.5 microns (|im); a measurement of fine particles
ppb          Parts per billion
ppm         Parts per million
QA          Quality assurance
QMP         Quality Management Plan
RE A         Risk and Exposure Assessment
RIA         Regulatory Impact Analysis
RR          Relative risk
RTF         Research Triangle Park
SES         Socioeconomic status
SLAMS      State and local monitoring stations
SO2         Sulfur dioxide
TBD         To be determined
                                            Vlll

-------
This page left intentionally blank
                  IX

-------
 i    1   INTRODUCTION
 2           The U.S. Environmental Protection Agency (EPA) is conducting a review of the air
 3    quality criteria and the primary (health-based) national ambient air quality standards (NAAQS)
 4    for nitrogen dioxide (NCh).1 The establishment and periodic review of NAAQS are governed
 5    primarily by sections 108 and 109 of the Clean Air Act (Act). The NAAQS are established for
 6    pollutants that may reasonably be anticipated to endanger public health and welfare, and whose
 7    presence in the ambient air results from numerous or diverse mobile or stationary sources. The
 8    Act requires that NAAQS are to be based on air quality criteria, which are to accurately reflect
 9    the latest scientific knowledge useful in indicating the kind and extent of identifiable effects on
10    public health or welfare that may be expected from the presence of the pollutant in ambient air.
11    Based on periodic reviews of the air quality criteria and standards, the Administrator is to make
12    revisions in the criteria and standards, and promulgate any new standards, as may be appropriate.
13    The Act also requires that an independent scientific review committee advise the Administrator
14    as part of this NAAQS review process, a function now performed by the Clean Air Scientific
15    Advi sory Committee (CASAC).
16           The overall plan for this review is presented in the Integrated Review Plan for the
17    Primary National Ambient Air Quality Standards for Nitrogen Dioxide (IRP) (U.S. EPA, 2014a).
18    The IRP summarizes the Clean Air Act  (CAA) requirements related to the establishment and
19    review of the NAAQS; the history of the primary NCh NAAQS, including the key science and
20    policy issues considered in the last review;  the anticipated process and schedule for the current
21    review of the primary NCh NAAQS; and the anticipated scope and organization of key
22    assessment documents in the current review, including the Integrated Science Assessment (ISA),
23    the Risk and Exposure Assessment (REA), if warranted, and the Policy Assessment (PA). The
24    IRP also lays out the key policy-relevant issues to be addressed in this review as a series of
25    questions that will frame our2 approach  to reaching conclusions on the degree to which the
26    available evidence and information could support retaining or revising the current primary NCh
27    NAAQS.
      1 The EPA is separately reviewing the welfare effects associated with oxides of nitrogen and the protection provided by the
      secondary NCh standard, in conjunction with a review of the secondary standard for sulfur dioxide (SCh) (U.S. EPA, 2014a,
      section 1.4).
      2 In this document, the terms "we" and "our" refer to staff in the EPA's Office of Air Quality Planning and Standards (OAQPS).
                                                  1-1

-------
 1          As a further step in planning for the current review, this document is intended to facilitate
 2    CASAC advice and public input to the EPA on potential support for updated quantitative
 3    analyses of NCh exposures and/or health risks. To facilitate such advice and input, the EPA staff
 4    has considered the degree to which important uncertainties identified in quantitative analyses
 5    from previous reviews have been addressed by newly available scientific evidence, tools, or
 6    information. Based on these considerations, this document presents our preliminary conclusions
 7    on the extent to which updated quantitative analyses of exposures and/or health risks are
 8    warranted in the current review. For updated analyses that are supported, this planning document
 9    also presents our anticipated approaches to conducting such analyses and, where appropriate,
10    preliminary results based on illustrative examples.
11          Staffs considerations and preliminary conclusions in  this planning document draw from
12    the scientific evidence assessed in the second draft of the Integrated Science Assessment for
13    Oxides of Nitrogen - Health Criteria (ISA) (U.S. EPA, 2015), the discussions of key issues in
14    the IRP (U.S. EPA, 2014a), the NO 2 Risk and Exposure Assessment Report from the last review
15    of the primary Mh NAAQS (U.S. EPA, 2008a), advances in modeling tools and techniques, and
16    new air quality data that have become available  since the last review. This document is being
17    submitted for review by the Clean Air Scientific Advisory Committee (CASAC) and made
18    available for public comment. The EPA staff will consider advice from CASAC and input from
19    the public in reaching conclusions regarding updated quantitative analyses in the current review.
20    These staff conclusions will be reflected in future documents3 generated as part of this review of
21    the primary NCh NAAQS, as described in Chapter 5 below.
22          The remainder of this chapter provides overviews of the history of the primary NCh
23    NAAQS (section 1.1); potential approaches to characterizing risks with quantitative analyses
24    (section  1.2); staffs  key considerations in evaluating the degree to which updated quantitative
25    analyses are supported in the current review (section 1.3); and the organization of the remainder
26    of this planning document (section 1.4).
      3 Future documents (i.e. Risk and Exposure Assessment, Policy Assessment) will also be reviewed by CASAC and made
      available for public comment.
                                                 1-2

-------
 1    1.1   HISTORY OF THE PRIMARY NO2 NAAQS
 2           On April 30, 1971, EPA promulgated NAAQS for NCh under section 109 of the CAA.
 3    The primary standard was set at 0.053 parts per million (ppm) (53 ppb), annual average (36 FR
 4    8186).4 The EPA completed reviews of the air quality criteria and NCh NAAQS in 1985 and
 5    1996, with decisions to retain the annual standard without revision (50 FR 25532, June 19, 1985;
 6    61 FR 52852, October 8, 1996).
 7           In the last review of the primary NO2 NAAQS, completed in 2010 (75 FR 6474,
 8    February 9, 2010), the EPA determined that the annual standard alone was not requisite5 to
 9    protect the public from respiratory effects that could result from short-term exposures to ambient
10    NO2.  To provide increased public health protection, including for at-risk populations such as
11    people with asthma, the EPA added a new short-term NO2 standard with a level of 100 ppb,
12    based on the 3-year average of the 98th percentile of the annual distribution of daily maximum  1-
13    hour NO2 concentrations. The EPA also retained the existing annual NO2 standard, with a level
14    of 53  ppb, to continue to provide  protection for effects potentially associated with long-term
15    exposures.6
16           The Administrator's final decisions on the standard placed primary emphasis on the
17    scientific evidence for respiratory effects attributable to short-term NO2 exposures. She viewed
18    the results of quantitative exposure and risk analyses as providing information in support of her
19    decision (75 FR 6498, February 9, 2010).7 The approaches employed in the last review to
      4 The secondary standard for NCh was set identical to the primary standard.
      5 In setting primary standards that are requisite to protect public health, as provided in section 109(b) of the Clean Air Act, the
      EPA's task is to establish standards that are neither more nor less stringent than necessary for these purposes.
      6 The existing primary NO2 NAAQS are specified at 40 CFR 50.11.
      7 The decisions made in the last review of the primary NCh standard were informed by the extensive body of scientific evidence
      published through early 2008 and assessed in the Integrated Science Assessment for Oxides of Nitrogen - Health Criteria (2008
      ISA, U.S. EPA, 2008b); the quantitative exposure and risk analyses in the Risk and Exposure Assessment to Support the Review
      of the NO2 Primary National Ambient Air Quality Standard (2008 REA, U.S. EPA, 2008a); the policy-relevant evidence- and
      exposure-/risk-based considerations related to the primary NCh NAAQS; the advice and recommendations of EPA's Clean Air
      Scientific Advisory Committee (CASAC, Henderson, 2008; Samet, 2008a,b, 2009); and public comments (75 FR 6474).
                                                     1-2

-------
 1   estimate NCh exposures and health risks are summarized briefly in section 1.2 below, and are
 2   discussed in more detail in subsequent chapters of this planning document.
 3          In conjunction with the revised primary NCh NAAQS, the EPA established a two-tiered
 4   monitoring network comprised of: (1) near-road monitors to be placed in locations of expected
 5   maximum 1-hour NCh concentrations near heavily trafficked roads in urban areas and (2)
 6   monitors located to characterize areas with the highest expected NCh concentrations at the
 7   neighborhood and larger spatial scales (referred to as area-wide monitors) (75 FR 6505 to 6506,
 8   February 9, 2010). Some near-road NCh monitors are currently in operation, with the  remainder
 9   of the anticipated monitors scheduled to become operational by January 1, 2017.8

10   1.2  APPROACHES TO CHARACTERIZING RISKS
11          In each NAAQS review, selection of the appropriate model for the characterization of
12   risks is influenced by the nature and strength of the evidence for the subject pollutant. Depending
13   on the type of evidence available, analyses may include quantitative risk assessments  based on
14   dose-response, exposure-response, or ambient concentration-response relationships. Analyses
15   may also be based on comparisons of health effect benchmark concentrations, drawn from
16   controlled human exposure studies, with modeled exposure estimates or ambient air quality
17   concentrations (i.e., as surrogates for potential ambient exposures). The variety of approaches
18   that  have been employed in NAAQS reviews is summarized in Figure 1-1.
     8 Subsequent to the 2010 rulemaking, the EPA revised the deadlines by which the near-road monitors are to be operational in
     order to implement a phased deployment approach (78 FR 16184, March 14, 2013).
                                                 1-4

-------
1
2

3
4
5
6
           Air Quality
           Monitoring/
            Modeling
           (Estimates of
             ambient
          concentrations)
                               Exposure Modeling
                            (Estimates of Inhalation [and, as
                             relevant, other route] exposure
                                   concentrations)
  Dosimetry
   Modeling
  (Estimates of
internal biomarker
  concentration)
                                  Exposure-response and/or
                                   health effect benchmarks
                                     (e.g., 03l S02l N02)
                                                            Internal
                                                     concentration-response
                                                         (e.g., CO, Pb)
                Ambient
          concentration-response
       and/or health effect benchmarks
          e.g., 03. N02, S02 PM
                                    Risk Assessment/
                                    Characterization
Figure 1-1. Risk characterization models employed in NAAQS Reviews.

        The conceptual model for the NCh health risk characterization conducted in the last
review is summarized below in Figure 1-2. This model was based on the available scientific
evidence assessed in the 2008 ISA, recognizing that the strongest evidence was for respiratory
effects attributable to short-term NCh exposures (U.S. EPA, 2008b, section 5.3).9
     9 As indicated in Figure 1-2, the 2008 REA focused on exposures to ambient NCh, though indoor sources of NCh and indoor
     exposures were also evaluated (U.S. EPA, 2008a, Chapter 8).
                                                    1-5

-------
                I  Key Sources of NO;,                        i
                  •Mobile source emissions to ambient air
                  •Stationary source emissions to ambient afr

                I  Exposure Pathways                        >
                  •Air - Outdoors
                  •Air - Indoors

                  Routes of Exposure
                  •Inhalation

                j  Key At-Risk Populations identified             ,
                  •People with respiratory disease, focus on asthma (children,
                  adults)

                i  Endpoints                               j
                  •Airway responsiveness, based on information from controlled
                  human exposure studies
                  •Emergency department visits for respiratory causes, based on
                  information from epidemiologic study

                I  Risk Metrics                              >
                  •Number of days per year with ambient !MO2 concentrations at or
                  above health effect benchmarks
                  •Number/percent of people with asthma estimated to experience
                  NO, exposure concentrations at or above health effect
                  benchmarks
                  •Estimated NO2-associated emergency department visits for
                  respiratory causes
 1
 2    Figure 1-2. Conceptual model for risk characterization in the last review of the primary
 3    NO2 NAAQS
 4           Based on the conceptual model summarized in Figure 1-2, the risk characterization in the
 5    last review employed three approaches to quantify NCh exposures and health risks (U.S. EPA,
 6    2008a):
 7        1)  Health effect benchmarks were identified based on information from controlled human
 8           exposure studies of NCh-induced increases in airway responsiveness. Ambient NCh
 9           concentrations were compared to these health effect benchmarks. In urban areas across
10           the U.S., such comparisons were made for ambient NCh concentrations at locations of
11           NCh monitoring sites and simulated concentrations on/near roadways (U.S. EPA, 2008a,
12           Chapter 7).
13        2)  Modeled estimates of personal NCh exposures were compared to health effect
14           benchmarks in a single urban area (Atlanta, GA). Exposures were characterized for
15           children with asthma and for people of all ages with asthma (U.S. EPA, 2008a, Chapter
16           8).


                                                    1-6

-------
 1
 2
 3
 4
 5
 6
 7
10
11
12
13
14
15
16
17
18
19
20
21
22
    3) Concentration-response relationships from an epidemiologic study were used to estimate
       NCh-associated emergency department visits for respiratory causes in Atlanta, GA (U.S.
       EPA, 2008a, Chapter 9).
       Exposures and risks were estimated for multiple NCh air quality scenarios, including for
ambient concentrations adjusted to just meet the existing annual NCh standard (i.e., the existing
NCh standard at the time of the last review) and for concentrations adjusted to just meet potential
alternative 1-hour standards with levels from 50 to 200 ppb. The quantitative analyses conducted
in the last review are discussed in more detail in the subsequent chapters of this planning
document.

1.3  KEY CONSIDERATIONS IN  THE CURRENT REVIEW
       In the current review, preliminary conclusions regarding the extent to which the newly
available evidence and information address important uncertainties and support updated
quantitative analyses are based on our consideration of a variety of factors. These include the
available health evidence; the available technical  information, tools, and methods; and judgments
as to the likelihood that particular quantitative analyses will add substantially to our
understanding of NCh  exposures or health risks beyond the insights gained from the analyses
conducted in the last review. These key considerations are summarized in Figure 1-3, below.
        Is appropriate
        scientific and technical
        information available
        to support
        quantitative
        assessments?
                     Yes
          Health evidence
          Air quality information
          Modeling approaches
          and tools
Is scientific and/or technical information
substantially different from the
information used in previous quantitative
assessments for the pollutant of interest?
             -and-
Does the new information appreciably
reduce the uncertainties or limitationsof
the last assessment?
                                                           Yes
Are results of quantitative
assessments likely to add
substantially to our
understanding of exposures or
pollutant-attributable health
risks, beyond the insights gained
from existing assessments?
                            Reduced uncertainty regarding health
                            outcomes, at-risk populations
                            Updated air quality information
                            Improved modeling approaches, tools
                                            -I
                                       Quantitative assessments
                                       conducted in previous reviews
                                       Results of preliminary or
                                       screening-level analyses
                                                                          Yes
              Quantitative
           assessments are not
               supported
                              Updated quantitative analyses
                              are not likely to substantially
                                  imp rove the utility of
                              exposure and risk estimates in
                                   the current review
                                    Updated quantitative analyses
                                      are likely to substantially
                                        improve the utility of
                                    exposure and risk estimates in
                                         the current review
Figure 1-3. Key considerations for updated quantitative analyses.
        An initial consideration is the available health effects evidence, and the foundation it may
provide for updated quantitative analyses. Our evaluation of the scientific evidence in this
                                                     1-7

-------
  1    planning document is based on the assessment of that evidence in the 2nd draft ISA (U.S. EPA,
 2    2015).10 In particular, we focus on information newly available in this review that addresses
 3    uncertainties identified in the last review and/or that may change major conclusions of the last
 4    review, such as causality determinations for NCh-associated health effects and conclusions
 5    regarding at-risk populations and lifestages (U.S. EPA, 2015).u
 6            Consistent with prior reviews, in considering the evidence with regard to support for
 7    quantitative analyses, we give primary consideration to health endpoints for which the ISA
 8    concludes the evidence supports a "causal" relationship or indicates that there is "likely to be a
 9    causal" relationship. In the current review, the 2nd draft ISA (U.S. EPA, 2015) reaches the
10    following conclusions in this regard:
11            •   The evidence supports "a causal relationship between short-term NCh exposure and
12               respiratory effects"  and the "strongest evidence is for effects on asthma exacerbation"
13               (U.S. EPA, 2015, Table 1-1, pp. 1-19).12 Key supporting evidence for these
14               conclusions comes from controlled human exposure studies of airway responsiveness
15               and from epidemiologic studies of asthma-related hospital admissions, emergency
16               department visits, and respiratory symptoms (U.S. EPA, 2015, section 1.5.1).

17            •   The evidence "indicates there is likely to be a causal relationship between long-term
18               NCh exposure and respiratory effects"  (U.S.  EPA, 2015, section 1.5.1, pp. 1-21 and 1-
19               21) and the "strongest evidence is for effects on asthma development" (U.S. EPA,
20               2015, Table 1-1).13 Key supporting evidence comes from epidemiologic cohort studies
      10  Staff will further consider the preliminary conclusions presented in this planning document in light of the assessment of the
      evidence in the in the final ISA.

      11 Conclusions in the 2nd draft ISA are based on a thorough evaluation of the available scientific evidence, taking into account
      factors such as the consistency and coherence of the evidence within and across disciplines (e.g., epidemiology, controlled human
      exposure, and toxicology), biological plausibility, and strength and specificity of effects (U.S. EPA, 2015, Preamble, section 5).
      With regard to health effects, the 2nd draft ISA uses a five-level hierarchy to classify the overall weight-of-evidence into one of
      the following categories: causal relationship, likely to be a causal relationship, suggestive but not sufficient to infer a causal
      relationship, inadequate to infer a causal relationship, and not likely to be a causal relationship (U.S. EPA, 2015, section 1.2).
      With regard to potential at-risk populations, the 2nd draft ISA classifies the evidence into one of the following categories:
      "adequate evidence," "suggestive evidence," "inadequate evidence," and "evidence of no effect" (U.S. EPA, 2015, section 7.2).

      12 The ISA in the last review of the Primary NChNAAQS concluded the available evidence indicated that there was "likely to be
      a causal relationship" between short-term NCh exposure and respiratory  effects.

      13 The ISA in the last review of the Primary NCh NAAQS concluded the available evidence was "suggestive, but not sufficient"
      to infer a causal relationship between long-term NO2 exposure and respiratory effects.

-------
 1              reporting associations between long-term ambient NCh concentrations (i.e., averaged
 2              over 1-10 years) and asthma incidence in children.  Support for the biological
 3              plausibility of effects attributable to long-term exposures is provided by "a small body
 4              of experimental studies" (U.S. EPA, 2015, Table 1-1).
 5          •   For all other health endpoints evaluated, the evidence is either "suggestive, but not
 6              sufficient, to infer a causal relationship" or "inadequate to infer a causal relationship"
 7              (U.S. EPA, 2015, section  1.5.2).14
 8          Evaluation of the support for quantitative analyses additionally gives primary
 9    consideration to populations and lifestages for which the ISA judges there is "adequate"
10    evidence for increased risk.15 In the current review, the 2nd draft ISA concludes that "there is
11    adequate evidence that people with asthma, children, and older adults are at increased risk for
12    NO2-related health effects" (U.S. EPA, 2015, Table 7-26). The second draft ISA concludes that
13    there is greater uncertainty for other  at-risk populations because the evidence is inconsistent
14    and/or because the evidence is for  effects that "are not  clearly related to NCh exposure" (U.S.
15    EPA, 2015, section 1.6.5, pp. 1-45 to 1-46).
16           Given these conclusions with regard to health endpoints and at-risk populations, our
17    consideration of potential updated  quantitative analyses in this document is focused on health
18    outcomes related  to asthma exacerbation in children and adults (short-term NCh exposures) and
19    the development of asthma in children (long-term NCh exposures). We focus particularly on the
20    key health studies that informed the 2nd  draft ISA's causality determinations.

21    1.4  ORGANIZATION OF THIS DOCUMENT
22           The remainder of this planning document presents our evaluations and preliminary
23    conclusions regarding the degree to which the newly available evidence and information
24    addresses important uncertainties and the support for updated quantitative analyses in the current
25    review. Chapters  2 through 4 present our considerations and preliminary conclusions regarding
26    the following:
27       •  Analyses comparing ambient  NCh concentrations with health effect benchmarks derived
28           from controlled human exposure studies (Chapter 2).
      14 Health outcomes for which the evidence is judged "suggestive, but not sufficient, to infer a causal relationship" can be
      considered as part of the overall consideration of the health evidence in the Policy Assessment.
      15 This is consistent with the approach adopted in the ongoing review of the ozone (Os) NAAQS (U.S. EPA, 2014d). The ISA
      framework for drawing conclusions about the role of various factors in modifying risks of air pollution exposures has been
      developed since the last review of the primary NCh NAAQS (U.S. EPA, 2015, section 1.6.5).
                                                   1-9

-------
 1       •  Assessment of human exposures based on modeled estimates in people with asthma, and
 2          comparing modeled 1-hour average exposures to 1-hour health effect benchmarks
 3          (Chapter 3).
 4       •  Health risk assessment (Chapter 4).
 5   Chapter 5 summarizes the conclusions from chapters 2 through 4, and discusses the next steps in
 6   the current review of the primary NCh NAAQS.
 7
 9
10
11
12
                                               1-10

-------
 i    2  AIR QUALITY AND HEALTH BENCHMARK

 2        COMPARISONS


 3           A key part of the body of scientific evidence identified in the 2nd draft ISA as supporting
 4    "a causal relationship between short-term NCh exposure and respiratory effects" (U.S. EPA,
 5    2015, Table 1-1) comes from controlled human exposure studies of airway responsiveness in
 6    people with asthma (U.S. EPA, 2015, section 1.5.1). In the last review, the REA used
 7    information from such studies to identify NCh health effect benchmarks. The REA compared
 8    these benchmarks with the ambient NCh concentrations estimated to occur under various air
 9    quality scenarios of interest (just meeting the existing and potential alternative standards). In
10    these analyses, ambient NCh concentrations served as surrogates for potential exposure
11    concentrations.16
12           This chapter presents the considerations leading to staffs preliminary conclusion that
13    new information available in the current review is expected to add substantially to our
14    understanding of the potential for population exposures to ambient NCh concentrations at or
15    above health effect benchmarks. Updated analyses that incorporate this new information would
16    be expected to provide additional perspective, beyond the analyses from the last review, on the
17    extent to which NCh exposures allowed by the current standard (and potential alternative
18    standards, as appropriate) could have important implications for public health.
19           Section 2.1 below provides an overview of the NCh air quality benchmark comparison
20    from the last review. Section 2.2 provides an overview of the information that is available in the
21    current review to inform updated analyses and presents staffs preliminary conclusion that
22    updated analyses should be considered. Section 2.3 provides an overview of staffs proposed
      16 If ambient concentrations are properly characterized (i.e., they appropriately capture temporal and spatial variability in
      concentrations across the selected study area), they would serve as a conservative estimate of ambient-related exposures. This is
      because ambient NCh concentrations are attenuated within indoor microenvironments where people commonly spend substantial
      time throughout their day.
                                                  2-1

-------
 1    analytical approach for updated analyses and presents preliminary results for a single illustrative
 2    urban study area.

 3    2.1   OVERVIEW OF THE ASSESSMENT FROM THE LAST REVIEW
 4           In the last review, the 2008 REA included analyses comparing ambient NCh
 5    concentrations at monitoring sites and on/near roadways to health effect benchmarks ranging
 6    from 100 to 300 ppb (U.S. EPA, 2008a, Chapter 7). Health effect benchmarks reflected the range
 7    of NCh concentrations that had been reported to increase airway responsiveness in the majority
 8    of people with asthma, based on a meta-analysis of individual study data presented in the 2008
 9    ISA (U.S. EPA, 2008b, Table 3.1-3).17 These comparisons of ambient NCh concentrations to
10    health effect benchmarks provided perspective on the extent to which, under various air quality
11    scenarios, populations could potentially experience 1-hour exposures to NCh concentrations that
12    could be of concern, particularly for people with asthma (U.S. EPA, 2008a, Chapter 7).
13           The 2008 REA's air quality assessment was based on NCh concentrations measured at
14    available U.S. monitoring sites,18 with a particular focus on 18 Core Based Statistical Areas
15    (CBSAs).19 The 2008 REA examined the potential for ambient NCh concentrations to be greater
16    than or equal to health effect benchmarks when air quality in the CBSAs was adjusted to just
17    meet the then-existing standard (i.e., annual standard with a level of 53 ppb) or potential
18    alternative 1-hour standards with levels ranging from 50 to 200 ppb (and 98th or 99th percentile
      17 Health effect benchmarks are discussed in more detail in section 2.2.2 below.
      18 Air quality data was separated into two six-year time periods, 1995 to 2000 (representing historical air quality) and 2001 to
      2006 (representing recent air quality) (U.S. EPA, 2008a, section 7.2.2). After applying a 75% data completeness criterion the
      final analytical data base included 627 monitors collecting ambient concentrations for 4,177 site-years of data (a valid monitoring
      day had >18 hourly measurements; monitors included in the analysis had >75% valid monitoring days in a year). Note, current
      validity criteria use calendar quarters (75% valid monitoring days in a quarter, having all four quarters complete) to ascertain a
      complete year.
      19 At the time the assessment was conducted, we used the terms CMSA/MSA to describe the monitors associated with
      metropolitan statistical areas. We replaced that terminology here with CBSA to reflect current terminology. First, using the
      complete set of ambient monitor data, we identified whether or not monitors belonged to a CBSA.. Then, CBSA-named study
      areas were identified as those having annual mean NCh concentrations occurring at a minimum of one monitor in the CBSA at or
      above 25.7 ppb (i.e., the 90th percentile concentrations across all study areas and site-years) and/or had at least one reported 1-
      hour NO2 concentration greater than or equal to 200 ppb. All remaining sites not included in this collection of CBSA-named
      study areas were aggregated into either one of two groups: all other CBSA or all other non-CBSA.
                                                     2-2

-------
 1    forms).20 For the air quality scenarios evaluated, the 2008 REA highlighted the number of days in
 2    each CBS A and each year with ambient NCh concentrations at or above one or more of the
 3    health effect benchmarks.
 4            At the time of the last review, we also focused portions of the air quality analyses on
 5    characterizing NCh concentrations occurring on roads and in near-road environments. Mobile
 6    sources are the largest contributors to total annual NOx emissions in the U.S. (U.S. EPA, 2008b,
 7    section 2.2.1) and monitor-based research studies had demonstrated large gradients in ambient
 8    NCh concentrations around major roadways, with higher concentrations occurring closer to  roads
 9    and lower concentrations occurring farther away from those roads (U.S. EPA, 2008b, sections
10    2.5.3.2, 2.5.4). Because the ambient monitoring network present at the time of the last review
11    was not designed to systematically measure NCh concentrations near the most heavily trafficked
12    roadways, the 2008 REA simulated ambient NCh concentrations on-/near-roads using
13    information from monitoring studies published in the scientific literature.21 Specifically, to
14    estimate on-/near-road concentrations the 2008 REA categorized ambient NCh monitors based
      20 Because annual average ambient NCh concentrations were below the level of the annual standard (i.e., 53 ppb) and most of the
      potential alternative 1-hour standards evaluated, ambient concentrations were primarily adjusted upwards to simulate just
      meeting the then-existing and potential alternative standards (U.S. EPA, 2008a, section 6.3.1). For the adjusted air quality
      standard scenarios, a proportional adjustment approach was used. This approach was supported by within-monitor comparisons
      of low and high NCh concentration years that largely demonstrated characteristics of a proportional relationship. Specifically,
      linear regressions were performed using the distributions of daily maximum 1-hour concentrations for a low-concentration year
      and a high-concentration year, measured at the same ambient monitor. Statistically significant linear regression slopes and model
      R2 values strongly supported features of linearity. However, in a few instances this analysis identified the presence of statistically
      significant regression intercepts and deviations from linearity at upper percentile concentrations, introducing uncertainty into the
      conclusion that a proportional relationship existed at all monitors (Rizzo, 2008; U.S. EPA, 2008a, section 7.4.5).
      21 At the time of the last review, based on the available evidence, there was uncertainty regarding the locations of maximum NCh
      concentrations with respect to roadway emissions and transformation of NO to NCh. Therefore, we characterized these simulated
      concentrations as on-/near-road.

-------
 1    on their distance from a road22 and applied literature-derived factors to concentrations at
 2    monitoring sites > 100 meters (m) from a road (U.S. EPA, 2008a, section 7.2).23
 3            For each CBSA and monitor year, the air quality at monitor locations was first adjusted to
 4    just meet the existing annual standard or potential alternative 1-hour standards. In cases where
 5    monitors were sited > 100 meters (m) from a road, simulated on-road concentrations were
 6    obtained by applying on-road simulation factors after the air quality adjustment. The 2008 REA
 7    presented the number of days per year with simulated 1-hour NCh concentrations on-, near-, and
 8    away-from-roads at or  above the health effect benchmarks (U.S. EPA, 2008a, section 7.2.4).

 9    2.1.1  Summary of Results
10            The 2008 REA presented a number of results from the analyses of NCh air quality and
11    health effect benchmarks (U.S. EPA, 2008a, sections 7.3.3 and 7.3.4; Appendix A), including the
12    following:
13        1.   On average, simulated NCh concentrations on/near roads were about 80% higher than
14            measured ambient concentrations in the same CBSA at monitoring sites > 100 m from a
15            road (U.S. EPA, 2008a, section 7.3.4).

16       2.   When air quality was adjusted to just meet the existing annual NCh standard, most
17            CBS As were estimated to have between 100 and 300 days per year with simulated on-
18            /near-road 1-hour NCh concentrations > 100 ppb; between 25 and 100 days per year with
19            simulated NCh  concentrations > 200 ppb (U.S. EPA, 2008a, Figure 7-6); and between 1
20            and 20 days per year with simulated NCh concentrations > 300 ppb (U.S. EPA, 2008a,
21            Appendix A, Table A-122). There were fewer days per year with such NCh
22            concentrations at the locations of the ambient monitors (U.S. EPA, 2008a, Figure 7-3).

23       3.   Compared to just meeting the existing annual standard:
      22 In this assessment, road distances to each monitor were generally determined using a Tele-Atlas roads database in a GIS
      application. The road types used to identify near-road monitors were those defined as: l=primary limited access or interstate,
      2=primary US and State highways, 3=Secondary State and County, 4=freeway ramp, 5=other ramps. Note only the monitors
      falling within the 18 identified study areas had estimated distances to these identified roads types, all other monitors (either
      characterized as 'other CMSA/MSA' or 'all other non-CMSA/MSA') were not used to simulate on-road concentrations.

      23 The 2008 REA derived a distribution of factors using data from eleven published studies that reported NCh concentrations on-
      roads (5 studies) and/or near-roads (6 studies) and NCh concentrations within and/or beyond 100 meters (m) from a road and
      assuming an exponential model for fitting the data. The 2008 REA then probabilistically applied these factors to ambient NCh
      concentrations reported at ambient monitor sites > 100 m from a road (assumed in the 2008 REA to represent background NCh
      concentrations, not influenced by roads). Major road types were defined in the 2008 REA as primary limited access or interstate,
      primary US and State highways, Secondary State and County, freeway ramp, and other ramps (2008 REA Appendix A, Table A-
      7). See Table 7-10 of the 2008 REA for the specific values of distributions that were used and Appendix A, sections for the
      studies used and the derivation methodology  (U.S. EPA, 2008a).
                                                     2-4

-------
 1              a.  When air quality was adjusted to just meet alternative 1-hour standards with
 2                 levels of either 50 or 100 ppb, fewer days per year had simulated 1-hour NCh
 3                 concentrations at or above health effect benchmarks (U.S. EPA, 2008a, Table 7-
 4                 29).

 5              b.  When air quality was adjusted to just meet an alternative 1-hour standard with a
 6                 level of 150 ppb, similar numbers of days per year had simulated 1-hour NCh
 7                 concentrations at or above health effect benchmarks (U.S. EPA, 2008a, compare
 8                 estimates in Figure 7-6 to those in Figures 7-7 and 7-8).

 9              c.  When air quality was adjusted to just meet an alternative 1-hour standard with a
10                 level of 200 ppb, generally larger numbers of days per year had simulated 1-hour
11                 NCh concentrations at or above health effect benchmarks (U.S. EPA, 2008a,
12                 compare estimates in Figure 7-6 to those in Figures 7-7 and 7-8).

13    2.1.2 Uncertainties and Limitations
14          The 2008 REA identified several sources of uncertainty associated with these analyses of
15    ambient air quality (U.S. EPA, 2008a, section 7.4, Table 7-31). Key sources of uncertainty are
16    summarized below.
17       1.  Spatial representativeness of assessment: The 2008 REA noted that, relative to the area
18          encompassed by the CBS As that comprised the urban study locations, there are a
19          relatively small number of ambient monitors in each location. To the extent there are
20          locations where ambient NCh concentrations exceed those measured by ambient
21          monitors, the occurrence of NCh concentrations at or above health effect benchmarks
22          could be underestimated.  To address this uncertainty in part, the 2008 REA developed the
23          approach to estimate on-road NCh concentrations, though it is possible that other local
24          sources exist, perhaps differing in emissions from mobile sources, and are not accounted
25          for by the existing monitoring network (U.S. EPA, 2008a, section 7.4.4).
26       2.  Simulated on-/near-road concentrations: The statistical model developed in the 2008 REA
27          to simulate on-/near-road NCh concentrations was based on measurement data reported in
28          a limited number of peer-reviewed studies. Most  of these studies used averaging times
29          much longer than the 1-hour concentrations relevant for the health benchmarks (i.e., 7-14
30          days or longer). The relationships between the study-related longer-term averaging times
31          and our use of short-term averaging times (1-hour) was not known at that time. In
32          addition, the derived factors were applied to concentrations at the away-from-road sites
33          (>100 m from roads) without considering the potential relationship between the derived
34          factors and ambient concentrations. The 2008 REA noted that if there is a concentration
35          dependence in the relationship between NCh  on/near roads and NCh away from roads, the
36          approach used would bias the simulated concentrations, though the direction of such
37          potential bias was not known. Other uncertainties related to the appropriateness of
38          applying the literature-derived factors to specific  U.S. urban study areas include; not
39          accounting for in-vehicle  penetration and decay of NCh that would likely be associated
40          with actual on-road exposures; the potential for emissions from non-road sources to
41          influence the monitors > 100 m from the road affecting their representativeness of
                                                2-5

-------
 1          background concentrations; and the selection of an exponential decay model (U.S. EPA,
 2          2008a, section 7.4.6) to define the concentration decline with distance from the roadway.
 3       3.  Adjusting ambient concentrations to just meet air quality standards: The 2008 REA noted
 4          that there is uncertainty in the approach used to adjust air quality to just meet the existing
 5          annual standard and potential alternative 1-hour NCh standards. This reflects the
 6          uncertainty in the true relationship between the adjusted concentrations meant to simulate
 7          a hypothetical future scenario and the historical unadjusted air quality. The adjustment
 8          factors used to simulate just meet the existing annual and alternative 1-hour standards
 9          assumed that all hourly concentrations would change proportionately at each ambient
10          monitoring site. The 2008 REA's discussion of uncertainty with the air quality
11          adjustment focused on two areas: (1) uncertainty in the appropriateness of using a
12          proportional adjustment approach and (2) uncertainty in applying the same approach to
13          all ambient monitors within each urban study location (U.S. EPA, 2008a, section 7.4.5).
14       4.  Health effect benchmarks: The health effect benchmarks used were based on a meta-
15          analysis of individual data from controlled human exposure studies presented in the 2008
16          ISA (U.S. EPA, 2008b). The 2008 ISA meta-analysis evaluated the direction of the
17          change in airway responsiveness, though it did not evaluate the magnitude of this change.
18          Therefore, there was uncertainty in the magnitude and severity of effects that occur
19          following exposures to NCh concentrations at or above health effect benchmarks (U.S.
20          EPA, 2008a,  section 4.2.5). In addition, the 2008 REA highlighted uncertainties related to
21          the use of benchmarks based on studies using a variety of exposure periods (generally 30
22          minutes to 2 hours) and subjects with asthma whose disease status was characterized as
23          mild, as opposed to those more severely affected (U.S. EPA, 2008a, section 1 A.I}.

24    2.2  OVERVIEW OF INFORMATION AVAILABLE IN THIS REVIEW
25          The following sections provide an overview of the information available in the current
26    review that would be expected to reduce uncertainties  from the last review and to inform the
27    design and interpretation of updated analyses. Section  2.2.1 discusses the data available in the
28    current review to inform the characterization of ambient NCh concentrations, including
29    concentrations on and near roadways. Section 2.2.2 provides an overview of the health
30    information assessed in the 2nd draft ISA (U.S. EPA, 2015) that could inform the identification of
31    NCh health effect benchmarks in the current review. Section 2.2.3 presents staffs preliminary
32    conclusion that an updated analysis comparing ambient NCh concentrations to health effect
33    benchmarks is supported in the current review.

34    2.2.1  Characterizing Ambient NOi Concentrations
35          Given the importance of roadway-associated NCh  concentrations in the last review,  a
36    critical consideration in the current review is the extent to which new information could better
37    inform our understanding of ambient NCh concentrations  on and near major roadways. When
38    evaluating the information available in this review to inform the characterization of ambient NCh
39    concentrations, we consider the available ambient NCh measurement data (section 2.2.1.1),
                                                2-6

-------
 1    information on important sources of NOx emissions (section 2.2.1.2), and information from
 2    modeling analyses of ambient NCh concentrations (section 2.2.1.3).

 3    2.2.1.1 Ambient measurement data
 4          This section discusses the ambient measurement data available in the current review that
 5    could provide the air quality basis for updated analyses. This includes data available from the
 6    existing NCh ambient monitoring network (section 2.2.1.1.1), including the recently deployed
 7    near-road monitors and data available from research studies that have characterized ambient NCh
 8    concentrations (section 2.2.1.1.2).

 9    2.2.1.1.1  NO2 ambient monitoring network
10          The existing NCh ambient monitoring network in the U.S. includes over 400 monitors.
11    Ambient NCh monitors are sited to represent various spatial scales, including microscale (in
12    close proximity, up to 100 m from a source), middle scale (several city blocks, 100 to 500 m),
13    neighborhood scale (0.5 to 4 km), and urban scale (4 to 50 km) (40 CFR Part 58, Appendix D).24
14    In the last review of the primary NCh NAAQS, EPA promulgated new monitoring requirements
15    mandating that state and local air monitoring agencies install near-road NCh monitoring stations
16    in large urban areas. Under these new requirements, state and local air agencies will operate one
17    near-road NCh monitor in any CBSA with a population of 500,000 or more and two near-road
18    NCh monitors in CBSAs with populations of 2,500,000 or more or in any CBSA with a
19    population of at least  500,000 and with roadway segments carrying traffic volumes of at least
20    250,000 vehicles per day.  These monitors are intended to measure ambient NCh concentrations
21    in the near-road environment where evidence indicates that peak ambient NCh concentrations
22    due to on-road mobile source activity can occur. The network is developing over time; the first
23    of three phases became operational in January of 2014 and the second phase in January of 2015.
24    In the current review, these near-road monitors will provide a key source of new information on
25    NCh concentrations around major roadways.
26          Table 2-1, below, lists the CBSAs with near-road monitors currently in operation. All
27    near-road monitors are required to be within 50 m of the target roadway, though the majority are
28    within 30 m. Any updated air quality analyses conducted in the current review will consider
      24 Criteria for siting ambient NCh monitors are given in the State and Local Air Monitoring Stations/National Air Monitoring
      Stations/Photochemical Monitoring Stations (SLAMS/NAMS/PAMS) Network Review Guidance (U.S. EPA, 1998).
                                                 2-7

-------
 1   information from these monitors, as well as updated information on ambient NCh concentrations
 2   from the entire monitoring network, as it becomes available.
 3          In addition to the newly available hourly NCh concentrations from the near-road
 4   monitors, updated air quality information is also available in the current review from the broader
 5   NCh ambient monitoring network. Based on these monitors, Figures 2-11 and 2-12 of the 2nd
 6   draft ISA (U.S. EPA, 2015) summarize the 98th percentiles of daily maximum 1-hour NCh
 7   concentrations and annual average NCh concentrations, respectively. From 2011 to 2013, all
 8   areas of the U.S. met the existing primary NCh NAAQS (U.S. EPA, 2015, Figures 2-11 and 2-
 9   12, Tables 2-3 and 2-4). For the NCh air quality assessment in this review, we will consider
10   further updated information from these monitors as it becomes available.

11
                                                2-8

-------
1   Table 2-1. Near-road NOi monitoring sites - active as of September 2014.
CBSA Name
Atlanta-Sandy Springs-Roswell, GA
Austin-Round Rock, TX
Baltimore-Columbia-Towson, MD
Birmingham-Hoover, AL
Boise, ID
Boston-Cambridge-Newton, MA-NH
Buffalo-Cheektowaga-Niagara Falls, NY
Charlotte-Concord-Gastonia, NC-SC
Cincinnati, OH-KY-IN
Cleveland-Elyria, OH
Columbus, OH
Dallas-Fort Worth-Arlington, TX
Denver-Aurora-Lakewood, CO
Des Moines-West Des Moines, IA
Detroit-Warren-Dearborn, Ml
Hartford-West Hartford-East Hartford, CT
Houston-The Woodlands-Sugar Land, TX
Indianapolis-Carmel-Anderson, IN
Jacksonville, FL
Kansas City, MO-KS
Los Angeles-Long Beach-Anaheim, CA
Louisville/Jefferson County, KY-IN
Memphis, TN-MS-AR
Milwaukee-Waukesha-West Allis, Wl
Minneapolis-St. Paul-Bloomington, MN-WI
Nashville-Davidson-Murfreesboro-Franklin, TN
New Orleans-Metairie, LA
New York-Newark-Jersey City, NY-NJ-PA
Philadelphia-Camden-Wilmington, PA-NJ-DE-MD
Phoenix-Mesa-Scottsdale, AZ
Pittsburgh, PA
Portland-Vancouver-Hillsboro, OR-WA
Providence-Warwick, RI-MA
Raleigh, NC
Richmond, VA
Riverside-San Bernardino-Ontario, CA
San Antonio-New Braunfels, TX
San Francisco-Oakland-Hayward, CA
San Jose-Sunnyvale-Santa Clara, CA
Seattle-Tacoma-Bellevue, WA
St. Louis, MO-IL
Tampa-St. Petersburg-Clearwater, FL
Target
Road
-85
-35
-95
-20
-84
-93
-90
-77
-75
-271
-270
-635
-25
-235
-96
-84
-69/US 59
-70
-95
-70
-5
-264
-40
-94
-94/I-35W
-40/1-24
-610
-95/US 1
-95
-10
-376
-5
-95
-40
-95
-10
-35
-880
US 101
-5
-64
-275
Annual
Average
Daily Traffic
284,920
188,150
186,750
141,190
103,000
193,000
131,019
153,000
163,000
153,660
142,361
235,790
249,000
110,000
140,500
159,900
324,119
189,760
139,000
114,495
272,000
163,000
140,850
133,000
277,000
144,204
68,015
311,234
124,610
320,138
87,534
156,000
186,300
141,000
151,000
245,300
201,840
216,000
191,000
237,000
159,326
190,500
Distance
to Road
(m)
2
27
16.15
23.2
32
10
20
30
8
-
32
24
8.7
13
8.5
17.7
24
24.5
20
20
9
32
23.75
14
32.5
30
28.5
20
12
12
18
25
5
20
20
50
20
20
32
4.5
25
20
Probe
Height
(m)
4.5
4
4
5.5
4.6
4
4
4.5
4.7
-
5.3
4
5
3
5.2
3.6
4
4
4.6
3
4.5
4.7
4.3
3.5
4.9
4.5
4.2
4.6
5
5.1
3
3
3.9
4.3
3.3
4.5
4
6.4
6.4
3
3
5
Start Date
6/15/14
4/16/14
4/1/14
1/1/14
4/1/12
6/1/13
3/24/14
6/22/14
1/1/14
9/1/2014
1/1/14
4/2/14
6/1/13
1/1/13
7/27/11
4/1/13
1/22/14
2/7/14
1/1/14
1/1/14
1/1/14
1/1/14
7/1/14
1/1/14
4/1/13
1/1/14
3/18/14
6/26/14
1/1/14
2/13/14
7/29/14
4/21/14
4/1/14
1/1/14
10/17/13
8/1/14
1/8/14
2/1/14
8/21/14
3/24/14
1/1/13
3/1/14
2
3
                                             2-9

-------
 1    2.2.1.1.2  NO2 measurement research studies
 2          As noted above (section 2.1), the 2008 REA simulated on-road NCh concentrations by
 3    applying a distribution of factors to NCh concentrations at monitor locations > 100 m from the
 4    road. In the current review, in addition to the data from recently sited near-road NCh monitors,
 5    our characterization of roadway-associated NCh concentrations will be informed by available
 6    research studies that have evaluated ambient NCh concentrations on-road (either in-traffic or
 7    curb side) and near roadways.
 8          The 2nd draft ISA identifies a number of studies that have characterized ambient NCh
 9    concentrations around roadways (U.S. EPA, 2015, section 2.5.3). Most studies that were
10    available in the last review used passive samplers requiring sampling periods on the order of a
11    week or longer (U.S. EPA, 2008a, section 7.2).25 This was  identified as one of the important
12    uncertainties in estimating on-road concentrations (i.e., simulating 1-hour on-road NCh
13    concentrations based on studies that used longer averaging times). Thus, at the time of the last
14    review we had a limited understanding of short-term (1-hour) NCh concentrations on/near-roads
15    relative to concentrations measured away from roads. Several recent studies have used sampling
16    methods that allow for improved temporal resolution (U.S. EPA, 2015, section 2.5.3.1, Table 2-
17    6, Figures 2-17 and 2-18) and, as described  below (section 2.3), information from such studies
18    could inform an updated characterization of on- and near-road NCh concentrations in the current
19    review.
20          There are two recent near-road transect measurement studies conducted in the U.S. where
21    temporally and spatially refined NCh concentration data will be available for this review. The
22    first study was conducted in Las Vegas, Nevada from December 2008 to 2009 (Kimbrough et al.,
23    2013). This study used continuous gas analyzers to collect  5-minute measurements of NCh and
24    NOx concentrations at both upwind (100 m) and downwind (20, 100, and 300 m) sites from a
25    major roadway (Interstate  15). The four monitoring locations used in the study were selected
26    along an east-west transect, approximately perpendicular to the roadway and occurring along a
27    railroad spur right-of-way. The second study was conducted in Detroit, Michigan from
28    September 2010 to June 2011 (Batterman et al., 2014). This study also collected 5-minute NCh
      25 The exception was the study by Rodes and Holland (1981), which evaluated 1-hour NCh concentrations in Los Angeles, CA.
      Authors reported that hourly NO2 concentrations were about 80 to 200% higher near a major highway (8 m from road) than
      concentrations away from the highway (400 to 500 m from the road). Because this is an older study, the 2nd draft ISA notes that
      "the vehicle fleet was not strictly regulated for NOx emissions" and "[a]s a result, the concentrations observed may not be
      relevant to current conditions" (U.S. EPA, 2015, p. 2-55).
                                                  2-10

-------
 1    concentrations. While the Detroit site did not favor a perpendicular transect (to the freeway)
 2    similar to the one used in Las Vegas, four stations were deployed for this study—three
 3    downwind and one upwind. Near-road measurement data from these two near-road transect
 4    studies are expected to provide information regarding the overall relationship between NCh
 5    concentrations and distance from the roadway. Data from these studies could inform or be used
 6    to evaluate an updated mathematical/statistical approach to use in simulating on-road
 7    concentrations using concentrations at a distance from the road and/or could be used to evaluate
 8    similar on- and near-road concentrations predicted using air quality.
 9          Further, there are recent air pollution roadway studies that, in addition to having near-road
10    measurements, also collected on-road NCh concentrations. As part of EPA's Geospatial
11    Measurement of Air Pollution (GMAP) program, mobile and stationary measurements of NCh
12    concentrations were collected in five study areas, two of which may have temporally and
13    spatially informative NCh data available for this review to support the approach proposed to
14    simulate on-road concentrations. The first study was conducted in Research Triangle Park, North
15    Carolina, during morning-hour commutes (7:00 AM-11:30 AM) from August to October 2012
16    (Mukerjee et al., 2015). This study used an electric vehicle instrumented with a cavity attenuated
17    phase shift spectrometry-based monitor to measure in-traffic NCh concentrations at  1-second
18    intervals on an interstate, major arterials, and collector roadways. Fixed site measurements
19    included meteorological parameters only (wind speed and direction) and not NCh concentrations,
20    an important study limitation to directly informing on-road to near-road concentration
21    relationships in this assessment. Study results however, could provide insight into the spatial
22    distribution of on-road NCh concentrations in an urban study area and the effect of important
23    influential factors.
24          A second GMAP study that could provide more complete data for informing  the on-road
25    to away-from-road concentration relationship was conducted in Phoenix, Arizona from October
26    to November 2013 (Baldauf et al., 2015). This study also used a mobile and fixed-site
27    measurement approach along two segments of Interstate 17. In-traffic NO2 measurements were
28    made at 1-second intervals during morning (9:00 AM-12:00 PM) and afternoon hours (2:00-5:00
29    PM)  on interstate,  arterial, collector, and residential roadways giving on-major-road (i.e., the
30    interstate) and away-from-major-road (i.e., arterial and residential) NO2 measurements. We
                                                 2-11

-------
 1    expect that the measurement data from both of these studies could provide further support to
 2    developing a factor(s) to simulate on-road concentrations from existing near-road monitor data.26
 3    2.2.1.2 Emissions information
 4          If updated air quality analyses are conducted, information on NOx emissions27 will inform
 5    our characterization of the important sources contributing to monitored NCh concentrations. This
 6    section provides an overview of the information available on NOx emissions at a national level
 7    based on recent updates to EPA's National Emissions Inventory (NET) and incorporated into the
 8    2011NEI.28
 9          At a national level, anthropogenic sources account for more than 90% of NOx emissions
10    in the 2011 NEI. Vehicles are the largest source, with highway and off-highway vehicles
11    contributing almost 60% of the total NOx  emissions nationally. Other important sources include
12    fuel combustion-utilities (14% of total), fuel combustion-other (11% of total), and biogenics and
13    wildfires (8% of total) (U.S. EPA, 2015, section 2.3.1, Figure 2-3). Compared to the national
14    averages, urban areas have greater contributions to total NOx emissions from both highway
15    vehicles and off-highway vehicles and smaller contributions from  other sources (U.S. EPA,
16    2015, Figure 2-4, Table 2-1). For example, in the 21 largest CBSAs in the U.S., more than half
17    of the urban NOx emissions are from highway vehicles. Together, highway vehicles and off-
18    highway vehicles and engines account for more than three quarters of total emissions in these
19    large CBSAs (U.S. EPA, 2015, section 2.3.2).
20          While an emissions summary at a national level is useful, important emissions sources can
21    vary across locations. As discussed below (section 2.3.2), NOx emissions sources, including
22    mobile sources and important stationary sources, will be characterized in more detail in specific
23    urban study areas selected for any updated air quality analyses.
      26 The three additional study areas identified in the GMAP program include Detroit, Michigan; San Francisco, California; and
      Charleston, South Carolina.
      27 Oxidized nitrogen compounds are emitted to the atmosphere primarily as NO, with NO converting to NO2 following its
      reaction with O3. Collectively, NO and NO2 are referred to as NOx (U.S. EPA, 2008b, section 2.2).
      28 The NEI is a national compilation of emissions estimates from all source sectors, collected from state, local, and tribal air
      agencies as well as those developed by EPA. The NEI is developed on a tri-annual basis, with 2011 being the most recent base
      year currently available and referred to as 2011 NEI.  The next NEI base year will be 2014 and will be available in 2016.  For
      information on the NEI, see http://www.epa.gov/ttn/chief/eiinformation.html.
                                                    2-12

-------
 1    2.2.1.3 NOi modeling research studies
 2          Two new modeling analyses could also inform our characterization of ambient NCh
 3    concentrations around roadways in the current review. Model estimated NCh concentrations at
 4    road-side receptors and at varying distances from major roadways could become available in this
 5    review to further inform an updated approach to simulate on-road NCh concentrations. The
 6    intended purpose of these modeled concentrations is to provide support for estimating on-road
 7    concentrations based on either using the new near-road monitor concentrations or using other
 8    away-from-road concentrations. Having modeled concentrations at varying distances from a road
 9    affords great flexibility in developing potential on-road simulation factors to be used, particularly
10    in knowing the influential factors that could affect the relationship between road-side and away-
11    from-road concentrations (e.g. wind speed/direction, mixing heights, presence of nearby
12    stationary sources)
13          The first modeling analysis used hourly link-based emissions varied by day type
14    (weekday vs. weekend) and hour of day developed for 17 road segments of interstate 95 in
15    Broward County, Florida (including Pompano, Ft. Lauderdale, and Dania beaches) using EPA's
16    AERMOD dispersion model (Thurman et al., 2013). The model was used to predict hourly NCh
17    concentrations at road-side receptors generally spaced 5, 10, 20, 25, 50, 75, and 100 m from the
18    road, as well as to predict roadway median concentrations over a five-year period (2006-10).
19    While there are six ambient monitors available that measured NCh concentrations to evaluate
20    model predictions, their time-averaging (weekly-average) and siting far from the roads evaluated
21    are an important limitation to this analysis.
22          The second modeling analysis is planned to technically correspond (e.g., same site
23    characteristics, meteorology, years) to the 2008-09 Las Vegas, NV measurement study
24    (Kimbrough et al., 2013) described above in section 2.2.1.1.2. The planned approach is to be
25    similar to that used for the Broward County modeling analysis,  though differing by using the
26    latest version of AERMOD, incorporating the most recent emissions data, and in having a robust
27    measurement data  set available (i.e., to better inform and evaluate existing model parameter
28    settings and assumptions). Once complete,  it is expected that these modeled data could provide
29    further support to developing a simulation factor(s) or other approach to simulate on-road
30    concentrations from existing near-road monitoring  data or from existing area-wide monitoring
31    data.

32    2.2.2 Evidence Informing Health Effect Benchmarks
33          The primary goal of any updated NO2 air quality analyses will be to inform conclusions
34    regarding the likelihood that the existing or potential alternative standards would allow for
35    exposures to ambient NO2 concentrations that could be of concern for public health. One way to
                                                2-13

-------
 1    accomplish this, as was done in the 2008 REA , is to compare ambient NCh concentrations
 2    adjusted to just meet the existing (and alternative, if appropriate) NCh standards with health
 3    effect benchmarks.
 4          In the last review, 1-hour NCh health effect benchmarks from 100 to 300 ppb were
 5    evaluated. These benchmarks were based on the results of an ISA meta-analysis of individual-
 6    level data from controlled human exposure studies of non-specific airway responsiveness in
 7    people with asthma.29 The results of this meta-analysis indicated that the majority of study
 8    volunteers experienced increased airway responsiveness following exposures to NCh
 9    concentrations of 100 to 300 ppb (or higher) for 30 minutes to 2 hours (U.S. EPA, 2008b, Tables
10    3.1-2 and 3.1-3). At the time of the last review, airway responsiveness was the only health effect
11    endpoint for which controlled human exposure studies had reported effects following exposures
12    to NO2 at or near ambient concentrations (U.S. EPA, 2008b, section 3.4).30
13          In the current review, the evidence to inform the identification of NCh health effect
14    benchmarks has not changed substantially from that used in the 2008 REA, though the 2nd draft
15    ISA includes expanded analyses of that evidence (U.S. EPA,  2015, section 5.2.2.1). Specifically,
16    the 2nd draft ISA includes an expanded evaluation of the magnitude and potential clinical
17    relevance of reported changes in airway responsiveness31 and a discussion of the limitations
18    impacting characterization of the exposure-response relationship between NCh  and airway
19    responsiveness (U.S. EPA, 2015, section 5.2.2.1). The updated meta-analysis in the 2nd draft ISA
20    also presents results based on a broader range of studies, including studies of both non-specific
21    airway responsiveness and of specific airway responsiveness  following allergen challenge (U.S.
22    EPA, 2015, Tables 5-5 and 5-6).  While the available evidence has not changed  substantially, as
23    summarized below these expanded analyses and discussions provide additional information to
24    better inform our understanding of the potential public health implications of exposures at or
25    above NCh health effect benchmarks of 100 ppb and higher.
      29 Increased airway responsiveness indicates the potential for worsened control of asthma symptoms. Non-specific airway
      responsiveness in these studies was assessed using stimuli such as carbachol, methacholine, histamine, cold air, or SO2 (U.S.
      EPA,2008b,Table3.1-2).
      30 The lowest NCh exposure concentration for which airway responsiveness has been evaluated is 100 ppb.
      31 Analyses of the individual data from a subset of studies indicates that a statistically significant fraction of study participants
      exposed to NCh concentrations from 100 to  530 ppb experienced clinically-relevant increases in airway responsiveness, indicated
      by a halving of the provocative dose (U.S. EPA, 2015, Figure 5-1). Evidence for clinically relevant increases was stronger for
      NO2 exposure concentrations < 250 ppb (80%, p=0.035) than for concentrations >250 ppb (73%, p=0.052).
                                                   2-14

-------
 1           As was the case in the last review, 100 ppb NCh is the lowest exposure concentration that
 2    has been evaluated for its impact on airway responsiveness. Of the five studies assessed in the
 3    ISA that evaluated exposures to 100 ppb NCh (Orehek et al.,  1976; Ahmed et al., 1983a; Ahmed
 4    et al. 1983b; Hazucha et al.,  1983; Tunnicliffe et al.,  1994), one (Orehek et al., 1976) was later
 5    reported (Dawson and Schenker, 1979) to show statistically significant increases in airway
 6    responsiveness. When individual data from across studies was analyzed together in a meta-
 7    analysis, 66% of study participants (a statistically significant  percentage; p < 0.05) were
 8    estimated to experience a NO2-induced (i.e., relative to filter air) increase in non-specific airway
 9    responsiveness following 1-hour exposures to 100 ppb NO2 (U.S. EPA, 2015, Table 5-4). When
10    this meta-analysis was updated to also include information  from studies evaluating  specific
11    airway responsiveness following allergen challenge,  61% of study participants experienced
12    increased airway responsiveness following exposures to 100 ppb NO2 (a marginally statistically
13    significant percentage; p = 0.08) (U.S. EPA, 2015, Table 5-6).
14           Controlled human exposure studies have also  evaluated airway responsiveness following
15    exposures to NCh concentrations greater than 100 ppb (U.S. EPA, 2015, Table 5-2).32 These
16    include one study that did not report a statistically significant increase in airway responsiveness
17    following resting exposures to 140 ppb NO2 (Bylin et al., 1988); five studies that evaluated
18    resting exposures to NO2 concentrations from 200 to 270 ppb (Orehek et al., 1976; Torres et al.,
19    1990; Bylin et al., 1988; Strand  et al., 1997;  1998; Barck et al., 2002), three of which reported
20    statistically significant NO2-induced increases in airway responsiveness (i.e., at 250 to 270 ppb);
21    and four studies that evaluated resting exposures to NO2 concentrations from 400 to 530 ppb
22    (Bylin et al., 1985; Mohsenin et al., 1987; Bylin et al., 1988; Tunnicliffe et al., 1994), three of
23    which  reported statistically significant NO2-induced increases in airway responsiveness (i.e., 400
24    to 500 ppb).
25           When individual-level data were combined across subsets of these studies, the meta-
26    analyses in the 2nd draft ISA indicate that statistically significant majorities of study participants
27    experienced increased airway responsiveness following exposures to NO2 concentrations from
28    100 to 200 ppb (labeled as "100 < [NO2] < 200" in U.S. EPA, 2015, Tables 5-4 and 5-6); 200 to
      32 These include studies where participants were at rest during exposure periods and studies where participants engaged in
      exercise during exposures. As noted in the 2nd draft ISA, "the literature on airway responsiveness supports the development of a
      refractory period following bouts of exercise" (U.S. EPA, 2015, p. 5-35). Consistent with the possibility that exercise may lead to
      a period of reduced airway responsiveness, the 2nd draft ISA notes larger increases in airway responsiveness in studies with
      participants at rest than in studies with participants engaged in exercise. In identifying NCh health effect benchmarks, we focus
      on studies that evaluated participants while at rest.
                                                   2-15

-------
 1    270 ppb (labeled as "200 < [NO2] < 300" in U.S. EPA, 2015, Tables 5-4 and 5-6), and 400 to
 2    530 ppb (labeled as "[NCh] > 300" in U.S. EPA, 2015, Tables 5-4 and 5-6). These percentages
 3    were statistically significant in analyses that included data from studies of non-specific airway
 4    responsiveness and in analyses that combined data from studies of non-specific and specific
 5    airway responsiveness.
 6          With regard to the health benchmarks appropriate for evaluation in this review, 100 ppb is
 7    the lowest NCh exposure concentration for which the evidence indicates the potential for NCh-
 8    induced increases in airway responsiveness. Given this, we reach the preliminary conclusion that
 9    100 ppb is an appropriate health effect benchmark to evaluate. However, we also recognize the
10    important uncertainties associated with the evidence for increased airway responsiveness
11    following exposures to 100 ppb NCh. These include the general lack of statistically significant
12    results in individual studies at 100 ppb and the lack of an exposure-response relationship based
13    on available studies. Such uncertainties will be taken into consideration when interpreting the
14    potential public health implications of NCh air quality concentrations that equal or exceed the
15    100 ppb health effect benchmark.
16          With respect to exposures to higher NCh concentrations, meta-analyses of pooled data
17    consistently indicate that statistically significant majorities of study participants experienced
18    increased airway responsiveness. In addition, individual studies have reported NCh-induced
19    increases in airway responsiveness with greater consistency and statistical precision. Thus, as
20    NCh exposure concentrations increase, we have increasing confidence in estimates of the
21    percentage of individuals with asthma who could experience increased airway responsiveness.
22    Given the evidence and the results of meta-analyses presented in the 2nd draft ISA, we reach the
23    preliminary conclusion that it is appropriate in the current review to evaluate NCh health effect
24    benchmarks as high as 400 ppb.

25    2.2.3 Preliminary Conclusions
26          As indicated in section 2.1.1 above, an important uncertainly identified in the 2008 REA
27    was the characterization of 1-hour NCh concentrations  on and around roadways, given the
28    limited information available in the last review. Based on the information discussed above, we
29    have a substantially improved body of information available in the current review to inform an
30    updated characterization of 1-hour NCh concentrations around roadways (section 2.2.1). In
31    particular, data from recently deployed NCh monitors near major roads, combined with new
32    information from monitoring and modeling  studies of NCh concentration gradients around roads,
33    will substantially improve our understanding of ambient NCh concentrations in the on-road and
34    near-road environments. This new information is expected to provide important perspective,
35    beyond what is available from the last review, on the extent to which NCh exposures on and near
                                                 2-16

-------
 1    roads could have potentially important implications for public health. Therefore, we reach the
 2    preliminary conclusion that an updated analysis comparing ambient NCh concentrations to health
 3    effect benchmarks is supported in the current review, with a particular focus on updating
 4    analyses of concentrations on and near major roadways.

 5    2.3  PROPOSED ANALYTICAL APPROACH AND ILLUSTRATIVE
 6         EXAMPLE
 7          Given the preliminary conclusion that updated analyses comparing ambient NCh
 8    concentrations to health effect benchmarks are supported in the current review, this section
 9    describes our proposed technical approach to conducting such analyses (section 2.3.1) and
10    preliminary results for an example urban study area (section 2.3.2).

11    2.3.1  Overview of Proposed Approach
12          Conducting an air quality assessment requires health effect benchmark concentrations of
13    concern for the general population or sensitive study group of interest, identification of a study
14    area(s) of interest,  and characterization of respective air quality (including measured,  adjusted,
15    and simulated, depending on the air quality scenario and concentration type). Each of these
16    components and an overview of the output data metrics are described in the following sections.

17    2.3.1.1 Identification of health effect benchmark levels
18          An evaluation of the controlled human exposure-based literature in the 2nd draft ISA (U.S.
19    EPA, 2015) and summarized in  section 2.2.2 above has identified a range of 1-hour
20    concentrations to consider in this air quality assessment. Because there is no apparent dose-
21    response relationship (see section 4.1 below), a range of concentrations of concern (100-400
22    ppb) will be evaluated in 100 ppb increments, yielding benchmark levels of 100, 200, 300, and
23    400 ppb. Instances when ambient concentrations in selected study areas  are at or above these
24    levels will be counted and summarized using the approach described in the following sections
25    (2.3.1.2 and 2.3.1.3).

26    2.3.1.2 Initial selection of study areas
27          While all of the existing ambient monitoring data are considered in this assessment, a few
28    of the air quality scenarios (e.g., air quality adjusted to just meet the existing standards) and
29    microenvironmental evaluations (e.g., on-road concentrations) warrant the defining of a specific
30    geographic domain. The following  are the proposed criteria for identifying study areas to
31    evaluate in the air quality assessment, followed with a list of candidate study areas when
32    applying the criteria to recent (2010-2013) ambient concentrations.
33
                                                2-17

-------
 1           1.   One of the most important attributes of a study area is the ability to use the
 2               monitoring data available to characterize the NCh area-wide and microenvironmental
 3               concentrations (both the highest annual and daily maximum 1-hour, DM1H) within
 4               the CBSA (e.g., near roadway, area influenced by significant stationary source(s), and
 5               to a lesser extent, background concentrations). Monitors sited within each CBSA will
 6               be identified using delineation files available through the U.S. Census Bureau.33
 7               Then, the distribution of hourly concentrations for each monitor year is screened to
 8               assess whether or not  standard completeness criteria are met. Ascertaining a valid
 9               year of monitoring data is a multi-step process. First, valid days are defined as those
10               having at least 18 hours of measurements. Next, a valid quarter is identified as having
11               at least 75% of valid days within a three-month calendar period (-68-70 days).
12               Finally, where all four quarters in a calendar year are valid, the year of monitoring
13               data is considered complete. Ambient monitor data will be grouped into the following
14               four categories:
15                  a.  Area-wide concentrations. CBS As having the maximum number of monitors
16                     and monitoring years of data will be given selection preference, considering
17                     the availability of both recent and historical air quality data.
18                  b.  Near-road concentrations. CBS As having near-road monitoring data will be
19                     given preference for selection, particularly if they meet completeness criteria.
20                     However, given that most of these monitors just began collecting
21                     concentrations in 2014, near-road data will not be excluded for having an
22                     incomplete year. Planned analyses would consider the relationship of near-
23                     road to area-wide concentrations, particularly when simultaneous
24                     measurements are collected. These analyses can remain informative if using
      33 The counties comprising each CBSA are listed at http://www.census.gov/population/metro/data/def.htmL as originally defined
      by the Office of Management and Budget (OMB) February 2013 Bulletin 13-01 (available at
      http: //www.whitehouse. gov/sites/default/files/omb/bulletins/2013/b-13-01 .pdf). Rather than using partial counties (if any were
      identified as such for a given CBSA), a monitor was included as part of the CBSA if it were situated anywhere within the listed
      county.
                                                   2-18

-------
 1                      less than a complete year of data, particularly when understanding seasonal
 2                      variation. CBS As having the maximum number of hours in a year
 3                      simultaneously monitored at near-roads and area wide monitors would be
 4                      given selection preference. CBS As not having a new near-road monitor would
 5                      still be considered as a potential study area,34 though still would require a
 6                      minimum number of area-wide monitors along with having other potential
 7                      near-source monitoring data (subsection c. and d. immediately below).
 8                  c.  Background concentrations.  Of lesser importance than criteria a. and b., the
 9                      selected study area should have valid hourly  concentrations measured at either
10                      a background or low-concentration monitor to some provide context for better
11                      understanding spatial variability in concentrations across the study area (i.e.,
12                      concentrations relative to those in likely high-concentration environments)
13                      and/or possibly for use in estimating on-road NCh concentrations in a
14                      generally similar manner as was done in the 2008 REA.
15                  d.  Other high NO2 concentration environments. Of lesser importance than
16                      criteria a., b., and c., CBSAs having valid monitoring data that can
17                      characterize potential highly influential emission sources other than on-road
18                      (e-g-, stationary sources, airports) will be given preference for selection.
19           2.  CBSAs having the highest annual and/or DM1H concentrations in the U.S. will be
20               given preference for selection as a study area. Justification for this criterion would
21               include, 1) monitors having the highest concentrations would  require  the smallest
22               adjustment upwards to just meet the existing standard, possibly limiting uncertainty in
23               generated results and 2) the risk associated with highest concentrations  (even
24               considering unadjusted concentrations) is by definition of greatest importance when
25               performing an assessment that uses health effect benchmark concentrations. Existing
26               design  values would be used to inform this selection criterion.35
27           3.  The list of selected CBSAs should capture areas where large portions of the U.S.
28               population reside, as this is a better representation of potential risks to populations at
      34 There are a few CBSAs that have an existing monitor sited in close proximity to a roadway not necessarily meeting the current
      near-road monitor requirements. For example, Chicago monitor ID 170313103 and El Paso monitor ID 481410044 were
      estimated to be about 20 m and 38 m, respectively from a major road in the 2008 REA (Appendix A, Table A-7).
      35 Monitor design values or the annual and 1-hour primary standards (annual average and 98th percentile DM1H averaged across
      3-years, respectively) for the period extending from 2002-2013 are available at http://www.epa.gov/airtrends/values.html.
                                                   2-19

-------
 1               an local, urban, and national scales as well as increasing the likelihood for
 2               appropriately representing important study groups (e.g.,  children with asthma). In
 3               addition, study area selection will be further guided by overall geographical location
 4               (e.g., climatic regions of the U.S.) to adequately represent areas across the U.S.
 5               having seasonal, atmospheric, or other influential factors that contribute to variability
 6               in concentrations.
 7            4.  Following the initial screening of potential study areas described above, additional
 8               information and data analyses are required to inform the decision to select study
 9               areas, but more so to retain an initially selected study area for analysis in the air
10               quality assessment. The purpose is to provide additional support to the approach used
11               to adjust NO2 concentrations to just meet the existing standards, not just at the highest
12               design value monitors alone, but also considering all monitors in a CBSA (area-wide,
13               near-road, other source-oriented, where available). Further, the availability of this
14               information and data analyses could also inform decisions made in simulating other
15               high NO2 concentrations not captured by the existing ambient measurements. This
16               criterion is mostly applied during the early stages of the  air quality assessment, with
17               some preliminary concentration analyses provided for study area selection here.36 The
18               available information and data analyses needed would include:
19                   a.  Ambient monitor meta-data. Preference for retaining an initially identified
20                       study area would be given to CBS As having information readily available that
21                       characterizes important monitor site attributes (e.g., geographic coordinates,
22                       local land use, monitor type, etc.) to indicate potentially important emission
23                       sources, such as roads  and stationary emission sources.
24                   b.  Local NOx source emissions data. Preference  for retaining an initially
25                       identified study area would be given to CBS As having proximally located and
26                       detailed source emissions information available to characterize potential
27                       individual monitor near-source influences.
      36 In general, most CBSAs selected using the first three criteria would generally meet this fourth selection criterion when
      considering the extent of available metadata used to describe the attributes of each monitor. However, some elements of the
      proposed evaluation, such as the historical concentrations, number of active monitors, or the degree of specificity regarding
      proximal emission source types will not necessarily inform the selection of a study area but rather inform the characterization of
      uncertainties associated with concentrations adjusted to just meet air quality standards, the simulated high-concentration
      environments (if any), and the estimated number of benchmark exceedances. We provide a preliminary application of the fourth
      criterion below, largely as an assessment of inter-monitor concentration ranges and correlations using the most recent data (2011-
      2014, See Table 2-3). A detailed application of the fourth criterion is given in the illustrative example that follows in section 2.3.
                                                      2-20

-------
 1                 c. Historical ambient monitoring concentrations. Preference for retaining an
 2                    initially identified study area would be given to CBS As where there are
 3                    substantial data or information available to characterize trends in NCh
 4                    concentrations that reflect changes in emissions over a period of time. Needed
 5                    are a collection of monitors having both current (2010-2014) and historical
 6                    (1980's - 2000's) concentrations. Of interest are intra- and inter-monitor
 7                    concentration ranges and correlations, particularly at the upper percentiles of
 8                    DM1H concentration distribution.
 9          We first evaluated the recent ambient monitoring data using the first three selection
10    criteria above, with results summarized in four tables that follow here.
11          Table 2-2 indicates the CBS As having a newly sited near-road monitor (and expected to
12    have near-road data available for this review). In using the most recent design values available
13    for the two 3-year averaging periods from 2010-2013, the CBSAs are listed in order first by the
14    maximum number of monitors available in a given year or over a 3-year averaging period and
15    then by 2010-2012 design values. Three categories ("strong", "moderate", or "limited") are used
16    to characterize the strength of information to support the  selection of a CBS A as a study area for
17    the air quality analysis. Regarding the number of monitors available, having a minimum of three
18    area-wide monitors indicated moderate support given that this number can form the simplest 2-
19    dimensional geometric shape (and to a limited extent, approximate an air quality  surface), with
20    CBSAs having greater than 3 monitors indicating strong  support. The thresholds used for
21    categorizing the design values were based on the overall  distribution of concentrations, selecting
22    for where concentrations were greatest (and thus requiring the least adjustment for just meeting
23    the existing standard). For annual average concentrations, the general range of design values
24    extended from a few to 30 ppb,  thus the upper portions of this range (i.e. >15 ppb) indicated
25    areas having moderate support,  with those >20 ppb indicated areas having strong support.  For the
26    DM1H design value, having a value of at least 50 ppb (or a factor of 2 less than the standard
27    level of 100 ppb) indicated areas having moderate support, with those >60 ppb indicated areas
28    having strong support. And finally, CBSAs identified as a strong candidates for the air quality
29    analyses had at a minimum, a strong rating for the number of monitors available and at least a
30    moderate rating when considering design values, while CBSAs identified as a possible
31    candidates had moderate ratings for the number of monitors available. Eleven CBSAs are
32    identified as strong candidates using these criteria, with an additional six CBSAs indicated as
33    possible candidates to include in our air quality assessment.
34          Table 2-3 summarizes preliminary analyses of the ambient concentration data for CBSAs
35    having newly sited near-road monitors, considering all hours where the newly sited near-road
36    monitors have measured NCh concentrations. Similar to the application of the selection criteria

                                                 2-21

-------
 1    described above though considered here are availability of concurrent measurements at the area-
 2    wide monitors and near-road monitors. "Moderate" or "strong" support was indicated by the
 3    number of available area-wide monitors (at least 3 or >3, respectively), the range in 98th
 4    percentile DM1H concentrations (at least 10-20 ppb or >20 ppb, respectively), the number of
 5    years of on-road monitoring (>1 year, indicated strong support only) and number of hours per
 6    year (>4,000 and >6000 hours, or about 50% or 75% or the year, respectively).37 In this
 7    evaluation, CBS As identified as strong candidates was largely based on the having a strong
 8    characterization assigned for the number of available monitors as well having a strong
 9    characterization regarding the range in concentrations. CBS As identified as possible candidates
10    had at least a moderate rating assigned for the number of available monitors. In reviewing this
11    table, one CBSA (i.e., St. Louis) is now elevated from being a possible to strong candidate,
12    bringing the total number of CBSAs strongly considered to twelve. In addition, two new CBSAs
13    are now considered as possible candidates to include (i.e., Denver, Providence), when
14    considering the number of near-road 1-hour measurements, NCh concentration ranges and
15    correlations of upper percentile concentrations with area-wide monitors.
16          Finally, Table 2-4 provides a summary of 2010-2013 design value information similar as
17    that in Table 2-2, though for CBSAs for which we do not anticipate having data available from
18    newly sited near-road monitors. We categorized the  supporting information regarding the
19    number of area-wide monitors available and respective design values identical to that used for
20    Table 2-2. Based on the selection criteria, four additional CBSAs are indicated as strong
21    candidates, along with two additional CBSAs indicated as possible candidates to evaluate. Thus
22    to summarize the application of the first two study area selection criteria, 16 CBSAs are
23    identified a strong candidates, while 10 CBSAs are identified as possible candidates.
24          Regarding the third study area selection criterion, Table 2-5 lists CBSAs having at least
25    800,000 residents estimated for 2013, ordered by descending population. When considering the
26    16 CBSAs indicated above as strong candidates, these CBSAs collectively would include just
27    over 97 million people (or approximately 31% of total U.S. population). When considering the
28    10 CBSAs identified as possible candidates, an additional 27 million people could be included to
29    the group of study areas, thus comprising approximately 39% of the total U.S. population. The
      37 Not entirely used for selecting study areas though informative to understanding relationship of near-road monitor
      concentrations with area-wide monitors were correlations of the upper percentile concentrations and the presence or not of a low-
      concentration monitor. A preliminary assessment of these particular attributes is provided in Table 2-2, however these criteria
      would be evaluated in greater detail when the air quality assessment is performed for a selected study area.
                                                  2-22

-------
1    locations of these 26 study area candidates are illustrated in Figure 2-1. Included in the figure
2    also are the 10 most populated CBS As identified as a limited study area candidate using the
3    selection criteria, most of which do or may have near-road monitoring data available for analysis
4    in this current review.
5
                                                 2-23

-------
1   Table 2-2. Number of NOi monitors and 2010-2013 design values for CBSAs that have or
2   are expected to have ambient concentration data available from newly sited near-road
3   monitors.
CBSA Name/Selection Indicator
(Strong candidate)
(possible candidate)
Los Angeles-Long Beach-Anaheim, CA
Houston-The Woodlands-Sugar Land, TX
Riverside-San Bernardino-Ontario, CA
Dallas-Fort Worth-Arlington, TX
San Francisco-Oakland-Hayward, CA
San Diego-Carlsbad, CA
New York-Newark-Jersey City, NY-NJ-PA
Phoenix-Mesa-Scottsdale, AZ
Philadelphia-Camden-Wilmington, PA-NJ-
DE-MD
Pittsburgh, PA
Boston-Cambridge-Newton, MA-NH
Miami-Fort Lauderdale-West Palm Beach, FL
Atlanta-Sandy Springs-Roswell, GA
Kansas City, MO-KS
Richmond, VA
Minneapolis-St. Paul-Bloomington, MN-WI
St. Louis, MO-IL
Baltimore-Columbia-Towson,
MD
Oklahoma City, OK
Tampa-St. Petersburg-Clearwater, FL
Cincinnati, OH-KY-IN
Buffalo-Cheektowaga-Niagara Falls, NY
Denver-Aurora-Lakewood, CO
San Antonio-New Braunfels, TX
Cleveland-Elyria, OH
San Jose-Sunnyvale-Santa Clara, CA
Milwaukee-Waukesha-West Allis, Wl
Detroit-Warren-Dearborn, Ml
New Orleans-Metairie, LA
Hartford-West Hartford-East Hartford, CT
Nashville-Davidson-Murfreesboro-Franklin,
TN
Des Moines-West Des Moines, IA
Providence-Warwick
RI-MA

Jacksonville, FL
Orlando-Kissimmee-Sanford, FL
Portland-Vancouver-Hillsboro
, OR-WA
Austin-Round Rock, TX
Boise City, ID
Charlotte-Concord-Gastonia, NC-SC
Indianapolis-Carmel-Anderson, IN
Louisville/Jefferson County, KY-IN
Memphis, TN-MS-AR
Birmingham-Hoover,
AL

Columbus, OH
Raleigh, NC
Seattle-Tacoma-Bellevue, WA
Support Color Scheme:
strong
moderate
Number of monitors per year
2010
16
16
12
11
8
8
7
6
5
5
5
5
3
3
3
3
3
2
2
2
2
2
2
2
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
2011
14
16
9
14
10
8
7
5
4
5
5
5
3
3
3
3
3
2
2
2
2
1
2
1
1
2
1
1
1
1
1
1
1
1
1
1
1
0
1
1
1
1
0
0
0
0
2012
9
16
11
11
11
8
8
5
5
5
6
3
3
3
3
3
2
2
2
1
1
2
2
1
1
2
1
2
1
1
1
1
2
1
1
1
0
0
1
1
1
1
0
0
0
0
2013
13
16
11
11
10
8
8
5
6
6
5
4
3
3
2
3
3
2
2
2
2
0
2
2
1
2
1
2
1
1
1
1
2
1
1
1
1
0
1
2
1
1
0
0
0
0
2010-
2012
4
9
6
9
8
6
5
4
3
4
5
2
3
3
3
2
0
2
2
1
1
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
2011-
2013
5
9
3
9
8
5
6
4
3
4
4
2
3
1
2
3
2
2
2
1
1
0
1
0
1
2
1
1
1
0
1
1
1
1
1
1
0
0
0
0
1
0
0
0
0
0
> 3 monitors
at least 3 monitors
Annual Average NOz (ppb)
2010
max
26
15
23
13
16
21
22
25
23
15
19
10
14
15
12
10
13
18
9
6
15
13
28
4
16
14
13
12
8
10
13
9
10
9
6
9
3
10
12
13
12
10




2011
max
25
14
21
13
16
20
25
25
20
16
20
8
13
15
10
9
13
18
10
5
13
8
24
4
15
15
11
12
8
11
12
8
11
8
5
9
3

10
11
11
10




2012
max
21
15
22
12
15
20
22
26
18
14
19
8
12
14
10
11
14
16
9
5
4
10
25
4
14
13
12
13
8
9
12
8
10
8
5
9


9
10
11
9




2013
max
23
13
21
12
17
19
22
25
17
11
18
8
9
13
8
9
11
15
9
5
12

24
5
13
15
10
12
6
8
10
7
10
8
5
10
5

8
12
11
8




Annual avg NOz > 20 ppb
Annual avg NOz > 15 ppb
DM1H 98th pet 3-
year avg NOz (ppb)
2010-12
max
67
60
72
56
74
73
70
66
65
53
51
47
56
53
52
46

57
54
35
32



52
50
49
48
48
46
43
42
42
40
36
34










2011-13
max
64
59
62
53
68
73
67
64
61
49
50
46
51
52
47
44
53
52
54
34
30

62

50
51
49
44
48

42
39
43
38
34
34




44





DM1H NOz > 60 ppb
DM1H NOz > 50 ppb
                                            2-24

-------
1   Table 2-3. Preliminary near-road and area-wide CBSA data analysis using all available
2   and concurrent NOi ambient monitor measurement data (2011-2014).
CBSA Name/Selection Indicator
(Strong candidate)
(possible candidate)
Houston-The Woodlands-Sugar Land, TX
Los Angeles-Long Beach-Anaheim, CA
Dallas-Fort Worth-Arlington, TX
San Francisco-Oakland-Hayward, CA
New York-Newark-Jersey City, NY-NJ-PA
Boston-Cambridge-Newton, MA-NH
Philadelphia-Camden-Wilmington, PA-NJ-
DE-MD
Phoenix-Mesa-Scottsdale, AZ
Pittsburgh, PA
St. Louis, MO-IL
Atlanta-Sandy Springs-Roswell, GA
Denver-Aurora-Lakewood, CO
Kansas City, MO-KS
Minneapolis-St. Paul-Bloomington, MN-WI
Providence-Warwick, RI-MA
Baltimore-Columbia-Towson, MD
Cincinnati, OH-KY-IN
Detroit-Warren-Dearborn, Ml
Indianapolis-Carmel-Anderson, IN
Richmond, VA
San Antonio-New Braunfels, TX
Tampa-St. Petersburg-Clearwater, FL
Austin-Round Rock, TX
Birmingham-Hoover
AL
Buffalo-Cheektowaga-Niagara Falls, NY
Charlotte-Concord-Gastonia, NC-SC
Cleveland-Elyria, OH
Columbus, OH
Des Moines-West Des Moines, IA
Hartford-West Hartford-East Hartford, CT
Jacksonville, FL
Louisville/Jefferson County, KY-IN
Memphis, TN-MS-AR
Milwaukee-Waukesha-West Allis, Wl
New Orleans-Metairie, LA
Portland-Vancouver-Hillsboro, OR-WA
Raleigh, NC
San Jose-Sunnyvale-Santa Clara, CA
Seattle-Tacoma-Bellevue, WA
Boise City, ID
Miami-Fort Lauderdale-West Palm Beach, FL
Nashville-Davidson-Murfreesboro-Franklin,
TN
Oklahoma City, OK
Orlando-Kissimmee-Sanford, FL
Riverside-San Bernardino-Ontario, CA
San Diego-Carlsbad,
Support Cole
CA
strong
r bcheme:
moderate
Area wide
monitors with
concurrent
measurements (n)
16
16
11
11
9
7
7
5
4
4
3
3
3
3
3
2
2
2
2
2
2
2
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
Near-road to area-
wide 98th pet
concentration range
Oto33ppb
-17 to 28 ppb
-9to21ppb
1 to 38 ppb

-6 to 27 ppb
-4 to 22 ppb
-4 to 22 ppb
1 to 26 ppb
2 to 25 ppb
7 to 39 ppb
-12 to 6 ppb
-7 to 33 ppb
-6 to 21 ppb
14 to 43 ppb
2 to 12 ppb
14 to 29 ppb
Oto 16 ppb
11 to 13 ppb
4 to 13 ppb
10 to 19 ppb
17 to 21 ppb
24 ppb
12 ppb
Oppb
-2 ppb
2 ppb
-4 ppb
-3toO ppb
6 to 15 ppb
8 ppb
19 ppb
14 ppb
14 ppb
6 ppb
5 ppb
-Ippb
Ippb
21 ppb
Monitor
years
(n)
1
1
1
1
1
2
1
1
1
2
1
2
2
2
1
1
1
4
1
2
1
1
1
1
1
1
1
1
2
2
1
1
1
1
1
1
1
1
1
Simultaneous
Hours/year
(n)
5500
4000
4000
5500-7500
2000
4000-6500
4500-6000
5000-7500
1000-2500
6000-8500
3000
3000-7000
3000-6000
6000
5000
4000
6500-8000
2000-8500
4000-6000
2000-5000
6500
4000
4000
7000
5500
4000
2000
7000
8000
5000-6000
4000
500
1000
5000
6500
3000
3500
4000
4000
Correlation of near-road
and area-wide >90th pet
concentrations
small -moderate
small -moderate
moderate
small -moderate

small -strong
small -moderate
small -moderate
small -moderate
moderate -strong
small
small -moderate
small -moderate
small -moderate
small -moderate
small
small
small -strong
moderate
small -moderate
small -moderate
moderate
moderate
small
moderate
moderate
moderate
moderate
moderate
small -moderate
small
small
small
moderate
moderate
moderate
moderate
moderate
moderate
Low-
concentration
monitor?
likely
uncertain
likely
likely
uncertain
yes
likely
likely
likely
yes
likely
no
no
no
likely
no
likely
likely
no
no
uncertain
likely
uncertain
no
no
no
no
no
no
no
no
uncertain
uncertain
uncertain
no
no
no
no
uncertain
No Area-wide Monitor Data Available At Time of Analysis
No Near-Road Monitor Data Available At Time of Analysis
> 3 monitors
3 monitors
Max range > +20 ppb
Max range +10-20 ppb
>1 year >6000 hrs
na >4000 hrs
Strong: r > 0.66
Moderate: r =0.33-0.66
Small: r<0.33
yes
likely
                                           2-25

-------
1   Table 2-4. Number of NOi monitors and 2010-2013 design values for CBSAs that are not
2   expected to have ambient concentration data available from newly sited near-road
3   monitors.
CBSA Name/Selection Indicator

(Strong candidate)
(possible candidate)
Santa Maria-Santa Barbara, CA
Baton Rouge, LA
Sacramento-Roseville-Arden-Arcade, CA
Chicago-Naperville-Elgin, IL-IN-WI
Washington-Arlington-Alexandria, DC-VA-
MD-WV
Farmington, NM
Beaumont-Port Arthur, TX
El Paso, TX
Bakersfield, CA
Springfield, MA
San Luis Obispo-Paso
Grande, CA
Robles-Arroyo
Gillette, WY
Durango, CO
Wichita, KS
El Centre, CA
Las Vegas-Henderson-Paradise, NV
Stockton-Lodi, CA
Ogden-Clearfield, UT
Tucson, AZ
Oxnard-Thousand Oaks-Ventura, CA
Bismarck, ND
Sioux City, IA-NE-SD
Vernal, UT
Harrisburg-Carlisle, PA
Urban Honolulu, HI
Fresno, CA
Las Cruces, NM
Riverton, WY
Provo-Orem, UT
New Haven-Milford, CT
Reno, NV
Visalia-Porterville, CA
Worcester, MA-CT
Albuquerque, NM
Little Rock-North Little Rock-Conway, AR
Lexington-Fayette, KY
York-Hanover, PA
Kingsport-Bristol-Bristol, TN-VA
Yuba City, CA
Portland-South Portland, ME
Support Color Sch
strong
moderate
Number of monitors per year
2010
11
8
8
6
6
4
4
3
3
3
3
3
3
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
1
1
1
1
1
1
1
1
1
1
1
1
2011
11
8
8
5
6
3
4
2
3
3
3
3
3
2
2
2
2
2
2
2
2
2
3
2
1
3
2
3
1
1
1
1
1
1
1
1
1
1
1
1
2012
11
8
8
6
6
3
4
2
3
3
3
4
3
2
2
2
2
2
2
2
2
2
4
1
1
4
2
3
1
1
1
1
1
1
1
1
1
1
1
1
2013
11
8
6
5
7
3
4
3
4
3
3
4
2
2
2
3
2
2
2
2
2
2
2
1
1
4
2
2
1
1
1
1
1
1
1
1
1
1
1
1
2010-
2012
11
7
7
2
5
2
4
1
3
3
3
1
0
1
1
2
1
1
1
2
2
1
2
1
1
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
2011-
2013
10
7
4
0
4
3
4
1
2
3
3
1
0
1
1
0
1
2
1
2
2
1
1
1
1
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
> 3 monitors
At least 3 monitors
Annual Average NOz (ppb)
2010
max
9
13
12
25
18
12
8
17
14
15
6
7
6
8
14
13
14
17
12
10
6
3
5
11
3
13
9
1
15
14
16
13
14
12
10
9
13
10
8
10
2011
max
9
12
13
23
16
13
7
17
15
16
6
6
7
12
14
13
15
16
12
9
5
3
5
12
3
13
8
1
17
15
17
12
17
13
10
8
13
10
8
9
2012
max
10
11
12
22
17
13
6
16
15
14
7
8
6
10
14
14
14
16
13
10
5
6
7
3
3
13
7
2
17
13
14
12
13
14
11
8
12
11
10
10
2013
max
10
10
10
21
13
12
6
14
14
14
7
9
6
9
13
14
16
17
11
9
5
4
11
3
3
14
7
1
19
14
16
13
12
12
10
7
10
11
10
8
Annual avg NOz > 20 ppb
Annual avg NOz > 15 ppb
DM1H 98th pet 3-
year avg NOz (ppb)
2010-12
max
43
54
51
62
55
38
37
61
58
47
38
32

64
62
54
51
49
47
38
36
30
30
22
22



58
57
55
53
52
51
51
49
49
46
45
43
2011-13
max
36
52
50

51
41
35
59
46
46
38
32

65
64

53
55
46
37
35
37
34
21
21

41
5
66
55
56
52
52
48
50
45
44
52
47
44
DM1H NOz > 60 ppb
DM1H NOz > 50 ppb
4
5
                                           2-26

-------
1   Table 2-5. Population estimates for CBSA having at least 800,000 residents in 2013,
2   ordered by descending population.
CBSA Name
(strong candidate) or (possible candidate) based on Tables
2-2 through 2-4.
New York-Newark-Jersey City, NY-NJ-PA
Los Angeles-Long Beach-Anaheim, CA
Chicago-Naperville-Elgin, IL-IN-WI
Dallas-Fort Worth-Arlington, TX
Houston-The Woodlands-Sugar Land, TX
Philadelphia-Camden-Wilmington, PA-NJ-DE-MD
Washington-Arlington-Alexandria, DC-VA-MD-WV
Miami-Fort Lauderdale-West Palm Beach, FL
Atlanta-Sandy Springs-Roswell, GA
Boston-Cambridge-Newton, MA-NH
San Francisco-Oakland-Hayward, CA
Phoenix-Mesa-Scottsdale, AZ
Riverside-San Bernardino-Ontario, CA
Detroit-Warren-Dearborn, Ml
Seattle-Tacoma-Bellevue, WA
Minneapolis-St. Paul-Bloomington, MN-WI
San Diego-Carlsbad, CA
Tampa-St. Petersburg-Clearwater, FL
St. Louis, MO-IL
Baltimore-Columbia-Towson, MD
Denver-Aurora-Lakewood, CO
Pittsburgh, PA
Charlotte-Concord-Gastonia, NC-SC
Portland-Vancouver-Hillsboro, OR-WA
San Antonio-New Braunfels, TX
Orlando-Kissimmee-Sanford, FL
Sacramento-Roseville-Arden-Arcade, CA
Cincinnati, OH-KY-IN
Cleveland-Elyria, OH
Kansas City, MO-KS
Las Vegas-Henderson-Paradise, NV
Columbus, OH
Indianapolis-Carmel-Anderson, IN
San Jose-Sunnyvale-Santa Clara, CA
Austin-Round Rock, TX
Nashville-Davidson-Murfreesboro-Franklin, TN
Virginia Beach-Norfolk-Newport News, VA-NC
Providence-Warwick, RI-MA
2013 Population
19,949,502
13,131,431
9,537,289
6,810,913
6,313,158
6,034,678
5,949,859
5,828,191
5,522,942
4,684,299
4,516,276
4,398,762
4,380,878
4,294,983
3,610,105
3,459,146
3,211,252
2,870,569
2,801,056
2,770,738
2,697,476
2,360,867
2,335,358
2,314,554
2,277,550
2,267,846
2,215,770
2,137,406
2,064,725
2,054,473
2,027,868
1,967,066
1,953,961
1,919,641
1,883,051
1,757,912
1,707,369
1,604,291
Near Road
Monitor?
Yes
Yes
No
Yes
Yes
Yes
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
Yes
Yes
Yes
No
Yes
Yes
Yes
Yes
Yes
No
Yes
                                            2-27

-------
CBSA Name
(strong candidate) or (possible candidate) based on Tables
2-2 through 2-4.
Milwaukee-Waukesha-West Allis, Wl
Jacksonville, FL
Memphis, TN-MS-AR
Oklahoma City, OK
Louisville/Jefferson County, KY-IN
Richmond, VA
New Orleans-Metairie, LA
Hartford-West Hartford-East Hartford, CT
Raleigh, NC
Salt Lake City, UT
Birmingham-Hoover, AL
Buffalo-Cheektowaga-Niagara Falls, NY
Rochester, NY
Grand Rapids-Wyoming, Ml
Tucson, AZ
Urban Honolulu, HI
Tulsa, OK
Fresno, CA
Bridgeport-Stamford-Norwalk, CT
Worcester, MA-CT
Albuquerque, NM
Omaha-Council Bluffs, NE-IA
Albany-Schenectady-Troy, NY
Bakersfield, CA
New Haven-Milford, CT
Knoxville, TN
Greenville-Anderson-Mauldin, SC
Oxnard-Thousand Oaks-Ventura, CA
El Paso, TX
Allentown-Bethlehem-Easton, PA-NJ
Baton Rouge, LA
McAllen-Edinburg-Mission, TX
Dayton, OH
2013 Population
1,569,659
1,394,624
1,341,746
1,319,677
1,262,261
1,245,764
1,240,977
1,215,211
1,214,516
1,140,483
1,140,300
1,134,115
1,083,278
1,016,603
996,554
983,429
961,561
955,272
939,904
926,710
902,797
895,151
877,905
864,124
862,287
852,715
850,965
839,620
831,036
827,048
820,159
815,996
802,489
Near Road
Monitor?
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
Yes
Yes
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
1
2
                                             2-28

-------
           Candidate Areas
           No Near Road
               Limited (1)
            Q  Moderate (2)
            •  Strong (4)
           Near Road
            PI  Limited (9)
               Moderate (8)
            •  Strong (12)
USGS The National Map National Boundaries Oatasat. Nalicmai Elevation Dalas«1, Geopaph -; Nancs
Inlormaiiofi Sfsiem Niidoruil Hydrography Dntas
-------
 1    2.3.1.3.1  Adjusting air quality to just meet the existing standards
 2           Unadjusted air quality, termed "as is" in the prior review, represent ambient conditions as
 3    they are at the time of measurement. While unadjusted air quality presents perspective regarding
 4    existing conditions, it does not provide the specific effect that just meeting a particular standard
 5    has on ambient concentrations, exposures, and health risk. To evaluate the ability of a specific air
 6    quality standard to protect public health, ambient NCh concentrations need to be adjusted such
 7    that they simulate levels of NCh that would just meet the existing standards (i.e., 100 ppb, 98th
 8    percentile DM1H averaged across 3-years;  53 ppb, annual average) or potential alternative
 9    standards. Such adjustments allow comparisons of the level of public health protection that could
10    be associated with just meeting the existing and potential alternative standards.
11           All areas of the United States currently have ambient NCh levels below the existing
12    standards, albeit to varying degrees.38 Therefore, to simulate just meeting the existing standards,
13    NCh air quality levels in all study areas must be adjusted upward. There are a few ways this can
14    be performed, although any method must consider a few factors including representativeness,
15    applicability, and degree of complexity. The method used to adjust ambient concentrations to
16    just meet air quality standards in the 2008 REA was based on analysis of historical ambient air
17    quality occurring in six study areas (Rizzo, 2008),  using concentrations that reflected actual
18    high-concentration conditions (i.e., generally supporting a representativeness criterion). Based on
19    the largely consistent patterns observed across study areas and relative simplicity (requiring only
20    the calculation of design values), a proportional approach was selected to adjust ambient
21    concentrations upwards in all study  areas (i.e., supporting applicability), assuming that the
22    overall concentration distribution that exists with recent ambient conditions is identical to that
23    which would exist with higher concentrations that just meet the existing standard(s).
24           As mentioned in section 2.1, there were instances in the 2008 REA NCh air quality
25    characterization where the recent distribution of DM1H concentrations was not entirely
26    proportional to the historical air quality distribution. For example, Figure 2-2 shows the 0 to
27    100th percentile DM1H ambient concentrations measured at a single monitor (ID: 340273001) in
28    the New York CBS A, for years 1984 (high-concentration year) and 2007 (low-concentration
      38 In general, only one of the two existing standards (1-hour or annual) would be the controlling standard in a particular area, and
      is identified by the monitor design value that is closest to the particular standard. Given the form and level of the existing 1-hour
      standard and overall degree of variability in hourly ambient concentrations, it is likely that the 1-hour standard is the controlling
      standard in most areas, where such a 3-year average value can be calculated. Preliminary analyses shown below indicate this is
      always the case for the study areas considered in the air quality assessment.
                                                  2-30

-------
 1    year) (Figure 2-2, left panel). Across this 28 year time frame, DM1H concentrations across the
 2    entire distribution have decreased proportionally by about the same amount (about 40%) except
 3    for the single DM1H concentration which has decreased by a greater amount (about 50%) over
 4    that same time period.  This observed non-linearity at upper percentile concentrations (i.e., the
 5    rate of decrease of upper percentile concentrations over time was greater than that of other
 6    percentiles of the distribution) was common for most of the monitors evaluated by Rizzo  (2008)
 7    from the six selected study areas. More specifically, for the years where the higher ambient
 8    concentrations were measured, the upper percentile concentrations tended to be greater than what
 9    would be expected when assuming an entirely proportional relationship with more recent low
10    ambient concentrations. There can be some variability in this relationship of course, as illustrated
11    by the most recent low-concentration year data available for the same NY monitor (Figure 2-2,
12    right panel), whereas a few of the upper percentile concentrations appear to deviate from
13    linearity in the opposite direction, though the maximum still expresses the greatest deviation
14    from linearity.  Thus, when using a proportional approach to adjust recent ambient conditions
15    upwards to reflect a higher concentration air quality scenario, the estimated upper-most
16    percentile concentrations would tend to be less than those measured in an actual, and similarly
17    high, concentration year (i.e., the historical ambient concentrations).
18          Variable changes in selected percentiles of the concentration distribution occurring over
19    time could be driven by a number of influential factors. For example, particular emissions
20    reductions and control strategies that target sporadic NOx emission release events could lead to
21    reduction in sporadic high-concentration events observed at ambient monitors. It is also possible
22    that NCh-favorable atmospheric conditions and other important precursor emissions (e.g., NOx)
23    were present at the time of the historical measurements and contributed to the occurrence of the
24    observed sporadic, high NCh concentration events. With these and other potential influential
25    factors in mind, this suggests that certain conditions existing at an historical time may be
26    reasonably assumed to exist within our hypothetical air quality scenario of just meeting the
27    existing standard(s), possibly requiring additional complexity to the approach beyond a simple
28    proportional adjustment.
29          The selection of the adjustment approach should  also consider particular elements of the
30    NCh standard (i.e., form and averaging time) and the risk output generated for this. In reviewing
31    Table 2-2 and Table 2-4, currently the 1-hour standard is the controlling standard in all potential
32    study areas for the 3-years where that 1-hour standard can be calculated. In this air quality
33    assessment, we are interested in calculating the  number of times 1-hour concentrations are at or
34    above health effect benchmarks, the lowest of which is 100 ppb. Therefore, the ambient
35    concentrations that are at or above the 98th percentile DM1H (i.e., 100 ppb at the highest/design
36    monitor) are the most important feature of the concentration distribution to characterize well at

                                                2-31

-------
1
2
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
      any monitor when adjusting concentrations to just meet the existing 1-hour standard (more so at
      the highest design value monitor, though also of relative importance at the other monitors in the
      CBSA).
                     New_York(340273001)
                  High Year: 1984, Low Year: 2007
                                                                  New_York(340273001)
                                                               High Year: 1984, Low Year: 2011
                  50     100     150
                     High Year Concentration (ppb}
                                                               50      100      150      200
                                                                 High Year Concentration (ppb}
     Figure 2-2. Distribution of DM1H NOi concentrations (0 - 100th percentile) in the New
     York CBSA for a high-concentration year (1984) versus a low-concentration year (2007)
     adapted from Rizzo (2008) (left panel) and updated comparison with a recent low-
     concentration year (2011) (right panel).

            With all  of this considered, a two-step adjustment approach is proposed in this
     assessment to adjust the recent ambient concentrations in a study area to just meet the existing
     standards. For this two-step adjustment approach, a proportional approach  is used, as was done
     in the 2008 REA, though here only  applied to concentrations up to and including the 98th
     percentile DM1H (adjustment step  1). An additional modification to address the observed
     deviations from linearity at the upper percentile concentrations is also proposed, particularly at
     and above the 98th percentile (adjustment step 2). In this way, this two-step approach utilizes the
     simplicity of the proportional approach used in the 2008 REA but addresses more fully, the
     observed changes in peak concentration distributions to better capture the distribution of high
     NCh concentrations when adjusting air quality to meet the existing standards. As noted above, it
     is important to characterize well the adjusted concentrations that are at or above the 98th
     percentile DM1H for the analyses we are proposing. The general details in the approach are as
     follows and proposed to be applied  to the selected study areas.
            1.  As was done in the 2008 REA for all selected study areas, compare historical high-
               concentration year data (selected based on having  concentrations close to or at
               existing DM1H standard) with recent low-concentration year data within monitors
                                                2-32

-------
 1               (for all possible monitors in a CBS A and considering similar years)39, using the
 2               distribution of DM1H concentrations.
 3                  a.  Plot and evaluate for proportionality using visual inspection and linear
 4                      regression coefficient estimates and fit statistics, using both the full
 5                      distribution and those concentrations at or below the 98th percentile DM1H.
 6                  b.  Regress all CBSA monitors against the CBSA design monitor DM1H also
 7                      across a several year period, noting variation (if any) in regression coefficients
 8                      and fit statistics.
 9                  c.  Qualitatively discuss the potential impact to adjusted concentrations using the
10                      above information and considering ambient monitor site attributes (i.e.,
11                      section 2.3.1.2, distance from roads, the number/type/amount of emission
12                      sources in close proximity to a monitor, concentration correlations).
13           2.  Calculate design values for each monitor having recent (2010-2013) air quality,
14               considering both the 1-hour and annual standards, and identify the controlling
15               standard and design monitor.40 Calculate the proportional adjustment factor needed to
16               just meet the existing standard by dividing the standard level by the design value.
17           3.  Adjustment step 1: Adjust all DM1H concentrations proportionally, up to and
18               including the 98th percentile DM1H, at all monitors and for each single year in each
19               study area using the adjustment factor derived from the design monitor (and
20               considering the 3-year averaging period).
21           4.  Adjustment step 2: For DM1H concentrations above the 98th percentile DM1H, rather
22               than use proportionally adjusted concentrations, calculate a set of ratios for each
23               monitor by dividing the DM1H concentrations that are above the 98th percentile
24               DM1H by the 98th percentile DM1H.41 This set of ratios is proposed to be based on
25               the individual monitor and derived using a range of historical and recent ambient
26               monitor concentrations (e.g., 2000-2013) that meet completeness criteria. A mean or
27               median value across years is proposed as a reasonable approximation to use in
      39 An important limitation is that the monitors in operation can vary year-to-year, even more variable when considering the
      decades that exists between high-concentration and low-concentration years of interest, leading to fewer monitors available to
      evaluate in an area.
      40 See http://www.epa.gov/airtrends/values.html. To date, 2013 is most recent complete year for most monitor sites. The period
      2010-2013 would include four annual design values and two 3-year averaged hourly design values.
      41 For a full year of ambient data, there would be 365 or 366 DM1H concentrations. Therefore, upwards to seven unique ratios
      could be calculated using these seven days having DM1H concentrations above the 98th percentile DM1H.
                                                   2-33

-------
 1              adjusting the upper percentile concentrations at each monitor. Where adjustment
 2              factors cannot be calculated (i.e., the monitor is newly sited), adjustment factors from
 3              the design monitor would be used.
 4
 5          One remaining issue in this adjustment approach regards that of the role of the near-road
 6    data. To date, there are no near-road monitors that reported three years of complete data; most
 7    CBS As have only a single year of near-road monitor data, many of which are not considered as
 8    having a complete year. Therefore, design values cannot necessarily be calculated using these
 9    near-road monitors. We propose that, where simultaneous measurements are available at the
10    monitor having the highest design value for the year and hours that the near-road monitor active,
11    and where the near-road monitor is reporting greater concentrations than the monitor having the
12    highest design value, the proportional adjustment factor would then be calculated similar to what
13    is outlined above in step 2 (98th percentile DM1H) but would be based on the available near-road
14    monitor data. This proportional adjustment factor would then be applied to all years of air quality
15    evaluated in the particular CBS A. Otherwise, concentrations measured at the monitor having the
16    highest design value will be used to calculate the adjustment factor.

17    2.3.1.3.2  Simulating air quality to represent on-road concentrations
18          In the discussion that follows, we first briefly describe the approach used in the 2008
19    REA to simulate on-road NO2 concentrations. This is followed by a consideration of the
20    information available in the current review that could further inform our understanding of factors
21    that contribute to variability in near-road and away-from-road NO2 concentrations and their
22    relationships. This information includes the scientific evidence summarized in the 2nd draft ISA;
23    on-road, near-road and away-from-road measurement data and related analyses; and outputs
24    from recent and planned near-road modeling studies. Based  on this available information, several
25    options are available in the current review to estimate on-road NO2 concentrations.  One options
26    would be to assume recent near-road ambient monitored NO2 concentrations are a reasonable
27    approximation of on-road NO2 concentrations, particularly for instances where monitors are sited
28    in close proximity to a major road (e.g., at or within 10 m or other proximal distance). In
29    addition, we could develop a set of on-road simulation factors using information from one or
30    more of the following sources: 1) ratios of on-road to away-from road concentrations, based on
31    available measurement data from research studies; 2) a statistical/fitted model using available
32    measurement-based near-road transect study data, or 3) air quality model-based on- and near-
33    road transect study concentrations.
34          As described earlier, the 2008 REA derived factors from exponential models that were
35    individually fitted to data obtained from eleven published studies having measured  either on-road
                                                2-34

-------
 1    (5 studies) or near-road (6 studies) NCh concentrations, along with having measured a number of
 2    corresponding away-from-road NCh concentrations. All of the studies reported time-averaged
 3    concentrations sampled over at least 1-2 week periods. Using the ratio of on-road and away-
 4    from-road42 concentrations estimated from the fitted exponential models, two empirical discrete
 5    distributions of factors were generated and distinguished by one of either two seasons ("summer"
 6    and "not summer") based on when the original study data were collected. To simulate hourly on-
 7    road NCh concentrations in selected study areas analyzed in the 2008 REA air quality
 8    characterization, the distributions of these factors were randomly sampled and applied to ambient
 9    NCh concentrations measured at monitors sited >100 m from a major road and considering the
10    appropriate season. Using this approach, simulated on-road NCh concentrations in the 2008 REA
11    were, on-average, 80% higher than respective ambient levels at distances >100 m from a major
12    road (2008 REA, section 7.3.2).
13           The 2nd draft ISA (U. S. EPA, 2015) identifies a few studies that fit an
14    exponential/logarithmic function to the near-road and away-from-road NCh concentrations
15    (section 2.5.3.1), providing some support for using an exponential model in quantifying the
16    pattern of decreasing concentrations with increasing distance from a road, much like that used in
17    the 2008 REA (e.g., Cape et al., 2004;  Gilbert et al., 2003). In summarizing the published
18    literature for where on-/near-road and away-from-road concentrations were measured (including
19    the same longer-term time-averaged study data used for the 2008 REA),  the 2nd draft ISA states
20    "NCh concentrations measured from 0 to 20 m from the road range up to 20 ppb higher, or up to
21    100% higher than concentrations measured between 80 and 500 m from a road" (U.S. EPA,
22    2015). This value (up to 100%) is also generally consistent with the factors used to estimate on-
23    road concentrations in the 2008 REA. The 2nd draft ISA also evaluated studies having a more
24    intensive sampling regimen and, when restricted to measurements closest to the near-road
25    environments, data from these studies "suggest concentrations at 15 to 20 m [from a road]
26    average 20-40% higher than concentrations 80 m from the road." Further, the 2nd draft ISA
27    highlights greater differences in the near-road/away-from-road concentrations during daylight
28    hours, during summer hours, as well as noting the existence of an inverse relationship between
29    the difference and the concentration level.
      42 In the exponential model used in the 2008 REA, "background" concentrations are approximated and assumed to be not directly
      influenced by road emissions. These estimated concentrations served as the away-from-road concentrations in calculating the
      distribution of factors used to simulate on-road concentrations.
                                                2-35

-------
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
            Absent from the 2nd draft ISA analysis however, is an assessment of the quantitative
      relationship between on-road (i.e., at or within 1 m from a road) and immediately near-road (e.g.,
      10 to 20 m away-from-road) measurement concentrations. For example, using data available
      from three studies reporting concentrations measured <1 m from a road and at other immediately
      near-road sample sites, the percent increase in on-road concentrations was calculated here (Table
      2-6). The calculated percent increase was wide ranging, however, when considering urban roads
      having the highest on-road concentrations, on average, the on-road concentrations could be about
      15% to 35% higher than concentrations sited at about 10 to 20 meters away-from-roads. It is
      worth noting that each of these three measurement studies reported concentrations that were
      time-averaged over at least a one-week period.
            We propose to include in this first proposed approach, data from any newly identified
      research studies (e.g., the GMAP studies mentioned above) and additional studies identified in
      the ISA to further characterize a quantitative relationship between on-road measurements and
      concentrations measured immediately away-from-roads. In doing so, we will be cognizant of the
      study time-averaging in summarizing the data, noting those having shorter-term average (e.g.,
      hourly) and on-road measurement in particular, where data are available.
      Table 2-6. Percent increase of on-road compared to near-road using NCh concentrations in
      three studies having both on-road and immediately near-road measurements.
Study
Author
Bell and
Ashenden
(1997)
Cape et al.
(2004)
Monn et al.
(1997)
Road
Description/
Season
Rural/Summer
Rural/
Not Summer
Two-way
Trunk/Annual
Single-way
Trunk/Annual
Non-Trunk/
Annual
Urban/
Unknown
On-
Road/Near-
Road Site
Distance (m)
<1/20
<1/20
1/10
1/10
1/10
0/20
Mean On-
Road/Near-Road
Concentration
(ppb)
16/9
15/11
19/15
7/5
7/6
43/38
Percent increase of on-road to
nearest-road
Mean
88%
48%
35%
44%
15%
15%
Median
90%
34%
26%
46%
21%
19%
Range
(min to max)
27% to 183%
-12% to 143%
21% to 65%
31% to 51%
-7% to 30%
-3% to 24%
Comments
Summer data:
May-September
Site distances are
variable based
on actual road
shoulder width

19
20
21
22
23
24
25
            A second proposed approach involves the development of a statistical model similar to
     that developed in the 2008 REA, though using newly available near-road monitoring data
     collected in Las Vegas, NV. Details of the statistical model development and on-road simulation
     factor output is fully described in Appendix A, while information regarding the measurement
     data collection are found in Kimbrough et al. (2013). Briefly, near-road measurements of air
     quality, traffic, and meteorology were collected at a study area located adjacent to Interstate-15
                                               2-36

-------
 1    (1-15) in Las Vegas NV during Dec. 2008 to Jan. 2010. Downwind sampling sites were located
 2    approximately 20 m, 100 m, and 300 m east of the interstate. These measured 1-hour average
 3    NCh concentrations were fitted to statistical models to generate factors for use in estimating on-
 4    road NCh concentrations from concentrations measured at the newly sited near-road monitors.
 5           Several statistical models were developed from this Las Vegas near-road monitoring data
 6    set and evaluated to describe the pattern in the concentration decrease observed with increasing
 7    distance from a major road. Based on model fits (R2) and overall form, a logit-ln function was
 8    determined as most appropriate, though from a practical perspective, the model fit is similar to
 9    an exponential decay (e.g., see Figure 2-3). Considered also was the influence of local
10    meteorological conditions (e.g., wind direction and approximate mixing heights). The logit-ln
11    functions were then used to estimate on-road NCh concentrations and concentrations predicted at
12    varying distances from the road (i.e., 10, 20, 40, 50,  100, and 300 meters).
13           Using each hourly prediction, statistically modeled on-road NCh concentrations were
14    compared to the modeled near-road (e.g., 10 m) concentrations to calculate the percent increase
15    in on-road concentrations. The distributions of percent increases were then stratified by the near-
16    road concentration distribution quintiles and averaged for each; included also was the overall
17    average percent increase that was calculated using the entire distribution for each statistical
18    model developed. Given that for many hours of the day, meteorological data were available that
19    corresponded to concentration measurements, we evaluated five model scenarios: 1) all wind and
20    atmospheric  stability conditions combined, 2) winds from the west (210°-330°, where the
21    monitors were downwind  of the highway), 3) winds from the east (30°-150°, where the monitors
22    were upwind of the highway), 4) inversion conditions (convective mixing height less than 300
23    m),  and 5) non-inversion conditions (convective mixing height greater than 300 m).
24
                                                2-37

-------
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
          60
                                      100
                                                  150
                                           Distance from road (nn)
                                                               200
                                                                            250
                                                                                         300
Figure 2-3. Predicted and observed NCh concentrations for winds from the west using
based on data from a Las Vegas NV near-road measurement study. Predicted median
(solid), predicted 98th and 2nd percentile (dotted), observed median (circles), and observed
98th and 2nd percentiles (error bars) are shown.

       For this REA plan, we generated a set of factors to use in simulating on-road NCh
concentrations, designed specifically for the measured concentrations at the new near-road
monitors sited at or around 10m and 20 m from a road.43 These factors are provided in Table 2-7
and are stratified by the away-from-road concentration distribution quintiles and considering
varying meteorological conditions. To a limited extent (and as discussed in the 2nd draft ISA),
concentration level affects the value of the adjustment factor; in general, lower concentration
quintiles have greater percent differences between on-road and away-from-road concentrations
      43 It is possible to generate on-road simulation factors to use for any monitor distance (e.g., 80 m, 200 m from the road), though
      the principal objective here was to use the new near-road monitor data.
                                                 2-38

-------
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
      than higher concentration quintiles. Meteorological conditions also affect the value of the
      adjustment factor; conditions where winds were predominantly from the west (downwind) or
      having a greater mixing height have greater percent difference between on-road and away-from-
      road concentrations than when winds were from the east or atmospheric inversions were present.
      In considering these factors and their influential variables, and seeking a generally conservative
      though simple approach using these results to simulate on-road NCh concentrations, we propose
      an increase in NCh concentrations by 15% and 20% could be applied where we have near-road
      concentrations at monitors sited within 10 and 20 m from a major road, respectively. These
      selected values are at the upper range of values provided in Table 2-7, particularly when
      considering the highest concentrations values and associated high-concentration meteorological
      conditions (and hence when benchmark exceedances would be expected to occur).
      Table 2-7. Potential factors that could be used to simulate on-road NOi concentrations
      from near-road monitors sited at 10 or 20 meters from a major road, stratified by
      concentration quintiles and meteorological conditions, based on analysis of Las Vegas, NV
      near-road measurement study data (see Appendix A).
Near-road
distance
10 meters
20 meters
Concentration
distribution
quintile
1
2
3
4
5
overall
1
2
3
4
5
overall
Average upwards adjustment from near-road concentrations given
meteorological condition
All
13%
19%
14%
10%
11%
13%
19%
25%
17%
11%
12%
17%
Westerly
winds
25%
15%
8%
8%
10%
13%
34%
22%
11%
10%
11%
17%
Easterly
winds
10%
12%
10%
9%
9%
10%
15%
17%
11%
12%
11%
13%
Atmospheric
inversion
10%
14%
13%
12%
12%
12%
13%
19%
17%
15%
14%
16%
Non
inversion
7%
20%
22%
21%
15%
17%
10%
27%
29%
26%
17%
22%
16
17
18
19
20
21
22
23
            For comparison, these statistical model developed factors provided in Table 2-7 are
     higher than those that could be approximated from a meta-analysis performed by Karner et al.
     (2010). By using a similar collection of on- and near-road study data used by the 2008 REA,
     Karner et al. (2010) applied a locally-weighted regression approach (LOESS) to roadway-edge
     normalized concentrations and approximated the overall pattern in decreasing concentrations
     with increasing distance from a road. Based on Figure 3 of that study (see the 2nd draft ISA
     Figure 3-2) and approximating both the on-road concentrations and concentrations at a distance
                                               2-39

-------
 1
 2
 3
 4
 5
 9
10
11
12
13
14
15
16
17
18
19
20
of 10 m, 20 m, and 100 m from a road, we estimate the factors would be about 6%, 11%, and
33%, respectively. Note that many of the on-road NCh concentrations used in the Karner et al.
(2010) study were modeled using an exponential equation and, where measured concentrations
were available, were from longer-term sampling durations (i.e., time-averaged over at least a
week or more).
       And finally, newly available data from air quality modeling analyses could also be used
to inform the development of an approach or factors to use in simulating on-road concentrations
from either the near-road monitor data, or from monitors sited at further distances from a road.
As an example, Figure 2-4 illustrates the modeled NCh concentration decline with distance from
a road in Ft. Lauderdale FL (Thurman, 2013). Hourly concentrations modeled at on-road
receptors could be used in a manner similar to that done above using the statistical model output,
and also using any away-from-road receptor distance of interest.  For instance, maximum NCh
concentration are about 325 and 295 ug/m3 for the southbound and northbound traffic lanes,
respectively, modeled at a distance of about 5 m from the source emission release (Figure 2-4).
Assuming this receptor is representative of on-road concentrations, these on-road concentrations
represent an increase of approximately 35% that of the corresponding concentrations estimated at
the cross-road distance of 25 m, and considering either traffic direction.
             350 r
              -100
                              -50
                                      -25   -10     10   25
                                       Cross-road distance (m)
                                                              50
                                                                      75
                                                                              100
Figure 2-4. AERMOD modeled maximum 1-hour NOi concentrations with increasing
distance from a major road in Ft. Lauderdale, FL.
                                               2-40

-------
 1          In summary, the newly deployed near-road monitors will better-characterize NCh
 2    concentrations occurring around roadways, compared to the monitoring information available in
 3    the last review. Based on this newly available information, as well as information from the types
 4    of monitoring and modeling analyses described above, staff will consider the extent to which it is
 5    appropriate to apply on-road simulation factors to ambient NCh concentrations at sites of near-
 6    road monitors. The appropriateness of applying such a factor to estimate on-road concentrations
 7    in a given location will depend on the characteristics of the near-road site under evaluation,
 8    including the proximity of the monitor to the road and the degree to which the near-road monitor
 9    could also reflect NOx emissions from nearby stationary sources.
10          To the extent it is judged appropriate to apply an on-road simulation factor, staff will
11    consider several potential approaches to doing so. As discussed above, studies indicate higher
12    on-road or curb-side NCh concentrations compared with concentrations measured at short
13    distances from the road along a transect (e.g., Bell et al., 1997; Cape et al., 2004; Monn et al.,
14    1997). Statistical models developed using near-road measurement data also indicate a pattern of
15    increasing concentrations in close proximity to the roadway, with maximum NCh concentrations
16    estimated to occur on-roads. The concentration change with distance to a road has been
17    described using an exponential/logarithmic decay equation (Cape et al., 2004; Gilbert et al.,
18    2003; U.S. EPA., 2008b; U.S. EPA, 2015) and more recently using LOESS smoothing (Karner et
19    al., 2010) and a logit-ln function (Appendix A). A recent emissions/dispersion modeling analysis
20    also indicates  a similar pattern in decreasing concentration with increasing distance from road
21    (Thurman, 2013). Based on all of these analyses, on-road NCh concentrations could be simulated
22    by increasing NCh concentrations measured at near-road monitoring sites (i.e., within 10-20 m
23    from a major road) by about 6%  to 35%, depending on near-road road distance, the approach
24    selected (i.e., factors-based using on-road/near-road measurement concentrations, statistical
25    model-based largely using near-road measurement data, or air quality model-based using
26    emissions and meteorological data), and the characterization of other important emissions
27    sources in the vicinity of near-road monitors that could influence measured concentrations.

28    2.3.1.3.3  Characterization of uncertainty in estimated concentrations
29          As was done in all of our recent REAs (U.S. EPA, 2008; U.S. EPA, 2009; U.S. EPA, 2010;
30    U.S. EPA, 2014b) a systematic approach adapted from WHO (2008) will be used here to succinctly
31    characterize uncertainties for each particular component comprising the assessment. First, staff
32    will identify, incorporate, and qualitatively describe any observed variability in input data sets,
33    influential attributes, overall composition of the knowledge-base, and estimated parameters within
34    the analyses performed to re-characterize (if needed) previously identified uncertainties in the 2008
35    REA (see Table 7-31 in U.S. EPA, 2008) and to identify additional uncertainties that were not
                                                2-41

-------
 1    previously evaluated. In addition, and where possible, sensitivities of important variables anticipated
 2    to significantly influence estimated concentrations (e.g., air quality adjusted to just meet the existing
 3    standards, simulated on-road concentrations) will be evaluated to characterize the potential
 4    magnitude and direction of influence the identified uncertainties may have on the estimated
 5    concentrations.44
 6    2.3.1.4 Calculating benchmark exceedances
 7          As was discussed above in section 2.3.1.1, we have identified 1-hour concentrations
 8    ranging from 100 to 400 ppb (in 100 ppb increments) as the benchmark levels to consider in the
 9    air quality assessment. The complete set of DM1H concentrations will be used to calculate the
10    number of days per year the benchmark levels are exceeded at each monitor, for each year of air
11    quality, and within each study area. Each of the air quality scenarios considered in this
12    assessment (as is, existing standard, potential alternative standards (if any)) and adjusted high-
13    concentration environments (e.g., on-road) will be summarized using descriptive statistics such
14    as the mean and maximum number of exceedances per year for each individual CBS A.
15          Staff performed an analysis of the historical air quality to provide context for this
16    benchmark analysis, focusing here on instances where concentrations have been at or just around
17    the level  of the existing  1-hour standard. It is worth noting that the historical  data are
18    representative of real air quality scenarios that existed at the time of monitoring and that changes
19    in emissions control and atmospheric conditions that have  occurred since that time would
20    preclude  us from drawing complete conclusions about the number of exceedances associated
21    with a given 98th percentile DM1H concentration if attempting to use such information as a
22    prediction for future air  quality. Nevertheless, using these unadjusted data are considered
23    informative given general consistencies in the overall concentration distribution over  time at
24    each monitor and what would be expected regarding the number of exceedances given the form
25    of the existing 1-hour standard (i.e., for a complete year of data, there would be about 8 days
26    having concentrations at or above the 98th percentile DM1H value of 100 ppb).
27          We first calculated all rolling 3-year average 98th percentile DM1H values for  all
28    individual monitors in operation from 1980-2014 that met the completeness criteria described
29    above (section 2.3.1.2).  Staff then counted the number of days having exceedances of the 1 -hour
30    benchmark levels (i.e., 100, 200, 300,  and 400 ppb, if any) for each individual year within that 3-
      44 A qualitative characterization of low, moderate, and high is assigned to the magnitude of influence and knowledge-base
      uncertainty descriptors, using quantitative observations relating to understanding the uncertainty, where possible.
                                                 2-42

-------
 1    year period and identified the maximum number of exceedances (thus, the highest observed
 2    number of exceedances at that monitor for a single year) and calculated the mean number of
 3    exceedances (thus, the average of the observed number of exceedances at that monitor across the
 4    3-year averaging period) given the 3-year average 98th percentile DM1H for that monitor.
 5    Results of this analysis are presented in Figure 2-5.
 6          Based on the analysis of all available historical air quality and considering the form and
 7    level of the existing 1-hour standard, the maximum number of days in  a single year that the
 8    DM1H was > 100 ppb ranged from about 10 to 20 days (Figure 2-5, top left panel), while on
 9    average across a 3-year period, the number of days/year having similar benchmark exceedances
10    ranged from about 6 to 13 days (Figure 2-5, top right panel). This mean number of days per year,
11    on average, corresponds well with general expectations described above (i.e., on average there
12    could be about 8 days/year with exceedances given the form of the standard). Furthermore, and
13    according to the analysis of all  available historical air quality, exceeding a 1-hour benchmark
14    level of 200 ppb is a rare occurrence when considering the form and level of the existing 1-hour
15    standard. For example, of the 23 times a monitor had a 3-year average  98th percentile DM1H of
16    100 ppb, there were no exceedances of the 200 ppb benchmark on 19 of these occasions (Figure
17    2-5, bottom left panel). When averaging across the 3-year period, the mean number of days per
18    year having a similar benchmark  exceedance drops to 1  or less,  again with most monitors
19    recording no  exceedances of the 200 ppb 1-hour benchmark. It should be noted that monitors in
20    California CBSAs  (Los Angeles,  San Francisco, etc.) constitute the bulk of the data where 3-year
21    average 98th percentile DM1H concentrations were at or above 100 ppb, though the results of
22    this analysis when  excluding these areas are similar (data not shown), albeit with a tighter range
23    of values than when including the monitoring data in California (e.g., the mean number of
24    days/year having DM1H > 100 ppb ranged from about 6 to 9).
25
                                               2-43

-------
               Maximum Number of Exceed antes of 100 ppb
                                                            Mean Number of Exceedances of 100 ppb
                                      •  V.-.H-
                                      '...:•: '..••
                25     50      75    100     125
                3-Vear Avg DM1H 98lh Percent!le)MOz Concentration (ppb)
25     50     75    100     125
3-year Avg DM1H 98'" Percentile NO, Concentration (ppb)

8
a 20
i
5

•s
I
0
Ma




xi mum Nun



• • •
iberof Exce



.
edances of



• •
•«•» *
100 ppb


',
• •
* —



'
• ». ••»•••

i
8
« 20 -
1
X
T-H

D
1 ,
|
0
n




(lean Number of Exceedances of 200 ppb



•




..,;;
                25     50      75     100     125
                3-year Avg OM1H 98" Percentile NO, Concentration (ppbI
 25     50     75     100    125
3-year Avg DM1H 98th Percentile NO2 Concentration (ppb)
                                                                                           150
 3   Figure 2-5. The maximum (left panel) and mean (right panel) number of days per year
 4   where DM1H NOi concentration was > 100 ppb (top panel) and > 200 ppb and associated
 5   with 3-year average 98th percentile DM1H NOi concentrations, using 1980-2014 ambient
 6   monitor data.
 8   2.3.2 Illustrative Example: Characterizing Air Quality and Calculating
 9          Benchmark Exceedances in an Example Urban Study Area
10          (Philadelphia)
11          This section presents results associated with use of the above proposed approaches to
12   characterize air quality in an example study area, namely the Philadelphia CBS A. The
13   Philadelphia CBSA was identified above as a strong candidate for selection in this assessment
14   and was a study areas evaluated in the 2008 REA air quality characterization, thus providing
15   comparative benchmark exceedance results for the two analysis periods. Further, there are a
16   variety of monitoring locations included in the Philadelphia CBSA including industrial,
17   residential (urban-core and suburban), and agricultural areas, as well as having a newly sited
18   near-road monitor. In the first section describes the general attributes of the study area, focusing
19   primarily the ambient monitoring attributes and NOx emissions information. Then, details are
                                                2-44

-------
 1    provided regarding the adjustments made to simulate air quality that just meets the existing
 2    standards and for estimating on-road concentrations from near-road concentrations. Finally,
 3    exceedances of benchmark levels are calculated for two air quality scenarios (unadjusted ambient
 4    concentrations and adjusted to just meeting the existing standard) and for the two distinct
 5    concentration types (measured area-wide/near-road and simulated on-road) using the most
 6    recently available monitoring data (2011-2014).

 7    2.3.2.1 Ambient monitor attributes
 8          Staff considered the complete set of all available monitors in the Philadelphia CBS A from
 9    1980-2013 (Figure 2-6) to best characterize air quality when applying the analytical  approaches
10    described above. Note that over this period of interest, monitors may either begin or cease
11    operation, and in any given year, monitoring data may not meet data completeness criteria. Thus
12    when considering the continuity of NCh measurements across the study area, on occasion, the
13    overall analytical data set may be fragmented. Regardless, the goal in this assessment for each
14    analytical approach is to take the necessary steps to ensure appropriate use of the data in an effort
15    to better understand their representativeness in describing the study area's air quality.
16          The siting of the ambient monitors is of particular importance, recognizing that the
17    purpose of the monitoring could have an influence on the measured NCh concentrations and
18    subsequent interpretation in the air quality characterization.  Specific monitoring site attributes
19    available in EPA's Air Quality System (AQS) were  summarized, including the monitoring
20    objective, measurement scale, and predominant land-use. Additional features such as monitor
21    proximity to NOx stationary emission sources within 5 km and emission sources having at least
22    10 tons per year (tpy) were identified using each monitoring site and emission source geographic
23    coordinates and NOx emissions estimates from the 2011 National Emissions Inventory (NEI).
24    Each of these attributes is summarized in Table 2-8 (active monitors) and Table 2-9  (inactive
25    monitors) to provide perspective on the representativeness of the ambient NCh monitoring
26    network in the Philadelphia CB SA.
27          The land-use field indicates the prevalent land use within Vi mile of the monitoring site.
28    Most of the Philadelphia CBS A monitors are characterized as within residential and commercial
29    areas, with a few falling within industrial and agricultural areas. The measurement scale
30    represents the air volumes associated with the monitoring area dimensions.  Most of the
31    Philadelphia CBSA monitors have measurement scales of neighborhood (500 m to 4 km), though
32    one was identified as urban scale (4 to 50 km) along with one microscale monitor (in close
33    proximity, up to 100 m from a source). The monitor objective describes the monitor in terms of
34    its attempt to generally characterize health effects, photochemical activity, transport, or welfare
35    effects. Monitors that would be useful for evaluating public  health would be characterized as
                                                2-45

-------
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
     having a monitor objective of population exposure and/or highest concentration, clearly the
     intent of most of the monitors in the Philadelphia CBS A. The monitor indicated as having
     source-oriented objective is the near-road monitor (monitor ID 421010075).
            Mobile sources (e.g., automobiles) are the most significant contributors to NOx emissions
     in the U. S. (2nd draft ISA, section 2.3.1). Except for the new near-road monitor, distances of each
     ambient monitor to major roads were obtained from the 2008 REA. The estimated distances of
     the monitors to major roads ranged from a few meters to several hundred meters, although, on
     average, most of the ambient monitors are placed at a distance of 100 meters or greater from a
     major road (Table 2-8  and Table 2-9).
            Stationary sources (e.g., power generating utilities combusting fossil fuels) can also
     contribute significantly to NOx emissions in the U.S  (2nd draft ISA, section 2.3.1). When
     considering the active  monitors in the Philadelphia CBS A and the most recent NOx emissions
     categorized by source type, electric power generation, petroleum refineries, and municipal waste
     combustion are important stationary sources potentially influencing NO2 concentrations at a few
     of the ambient monitors.
16
17
18
                              v .     '••!!•
                                 ,.,.-.,/,
                              >; ^  < f
                                 ' f£'ff.      S
           • Active
           if Active - Design Value
             Active - Near Road
             Inactiv
                                                                     UUU^     / hi aril.
                                                                     ''•••" \-.l.li..«l"**
                                              4^04i)UOO^-  '   \^fr     ' •-! ".1 . . l.rlH-lon
                                       1000330of       ,.,,,,,'f^""1	
                                         If,, ...    X^rf"»~                 I 1II.I.-II\M.I.I  .
                                       -  " '       -         ;:n., yr'-.rv"—---"^
                                       ;
                                               ^    jr                       11.11
                                                                          ;>-.."
     Figure 2-6. Locations of the eight active and seven inactive ambient monitors in the
     Philadelphia CBSA.
                                                2-46

-------
1     Table 2-8. Attributes of active ambient monitors in the Philadelphia CBSA, as of 2014.
Attribute
Latitude
Longitude
Elevation (m)
Start year
End year
Land use
Measurement scale3
Monitor objective 1 b
Monitor objective 2
Distance from roadway (m) c
Operational Monitor ID
100031010
39.817222
-75.563889
0
2013
2014
agricultural
-
-
-
na
100032004
39.739444
-75.558056
0
2001
2014
commercial
-
pop expos
high cone
82
340070002
39.934446
-75.125291
4
2012
2014
industrial
neighbor
pop expos
-
na
420170012
40.107222
-74.882222
12
1974
2014
residential
neighbor
pop expos
.
393
420450002
39.835556
-75.3725
3
1974
2014
industrial
neighbor
pop expos
.
413
421010004
(design)
40.008889
-75.09778
22
1977
2014
residential
urban
pop expos
.
45
421010047
39.944651
-75.165206
21
1982
2014
residential
neighbor
pop expos
.
66
421010075
(near-road)
40.054128
-74.984802
9
2014
2014
commercial
microscale
high cone
source
12
NOX Emissions by Stationary Source Type (in total tons per year (tpy), summed for all sources within 5 Km of monitor that emit 10 or more tpy) d
Electricity generation (combustion)
Petroleum refinery
Municipal waste combustor
Chemical plant
Pulp and paper plant
Wastewater treatment facility
Institutional (school, hosp., prison)
Automobile/truck/parts plant
Steam heating facility
Military base
Petroleum storage facility
Source type not characterized
-
-
-
-
-
-
-
-
-
-
-
180
948
.
.
13
-
.
.
11
-
.
.
248
26
.
297
14
-
14
.
-
24
.
.
13
107
.
-
-
-
-
.
-
-
.
.
26
950
2146
1260
275
240
56
.
-
-
.
.
11
-
-
-
-
92
-
30
-
-
12
16
183
379
1315
.
14
-
14
66
-
24
-
-
-
-
-
.
-
-
-
11
-
-
.
.
-
2
3
4
5
6
1 Measurement scales: microscale (close proximity, up to 100 m from a source), neighborhood scale (0.5-4 km), and urban scale (4-50 km).
b Monitor objectives: population exposure, highest concentration, and source-oriented.
c All sites except for the 2014 near-road monitor used information generated from the 2008 REA (see footnote 22). Na is for where monitor not evaluated in 2008 REA.
d 2011 NOx emissions data were obtained from EPA's Emission Inventory System (EIS) Gateway located at https://eis.epa.gov/eis-svstem-web (2011 NEI version 2). Associated
documentation is available at http://www.epa.gov/ttn/chief/net/2011inventorv.htmltfinventorvdoc.
                                                         2-47

-------
1
2
Table 2-9. Attributes of inactive ambient monitors in the Philadelphia CBSA and used in analysis of historical NOi
concentration trends.
4
5
6
7
Attribute
Latitude
Longitude
Elevation (m)
Start year
End year
Land use
Measurement scale a
Monitor objective 1 b
Monitor objective 2
Distance from roadway (m) c
Historical Monitor ID
100031003
39.761111
-75.491944
65
1992
2000
residential
-
pop expos
-
189
100031007
39.551111
-75.730833
20
1992
1999
agricultural
-
-
-
144
100032002
39.757778
-75.546389
46
1978
1992
commercial
neighbor
high cone
-
na
100033001
39.812222
-75.455556
30
1978
1992
residential
neighbor
high cone
-
na
340070003
39.923042
-75.097617
7.6
1979
2008
residential
neighbor
pop expos
-
405
420910013
40.112222
-75.309167
53
1974
2008
residential
neighbor
pop expos
-
630
421010029
39.957222
-75.173056
25
1975
2005
commercial
neighbor
high cone
-
103
NOx Emissions by Source Type (summed tons per year (tpy), sources within 5 Km emitting at least 10 tpy) d
Electricity generation
Petroleum refinery
Chemical plant
Automobile/truck/parts plant
Municipal waste combustor
Steel mill
Steam heating facility
Wastewater treatment facility
Hot mix asphalt plant
Pharmaceutical manufacturing
Institutional
Type not characterized
948
-
27
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
948
-
27
-
-
-
-
-
-
-
-
199
963
1490
-
-
-
166
-
-
-
-
-
54
26
-
14
-
297
-
-
14
-
-
-
13
-
-
-
-
735
85
-
-
16
26
-
53
353
-
-
-
-
-
24
-
-
-
66
-
a Measurement scales: microscale (close proximity, up to 100 m from a source), neighborhood scale (0.5-4 km), and urban scale (4-50 km).
b Monitor objectives: population exposure, highest concentration, and source-oriented.
c All sites except for the 2014 near-road monitor used information generated from the 2008 REA (see footnote 23). Na is for where monitor not evaluated in 2008 REA.
d 2011 NOx emissions data were obtained from EPA's Emission Inventory System (EIS) Gateway located at https://eis.epa.gov/eis-svstem-web (2011 NEI version 2). Associated
documentation is available at http://www.epa.gov/ttn/chief/net/2011inventorv.htmltfinventorvdoc. 2011 NEI may not represent actual emissions when monitor was in operation.
                                                         2-48

-------
 1    2.3.2.2 Air quality standard scenario adjustments
 2          As described above in section 2.3.1.3.1, there are two steps to the approach used to adjust
 3    ambient concentrations such that they just meet the existing standard(s). First, a proportional
 4    factor is needed to adjust the ambient monitoring concentration up to and including the DM1H
 5    98th percentile concentration. This proportional adjustment factor can be derived from the
 6    currently available monitor design values. Because there are two standards (annual and hourly)
 7    and two types of monitoring data considered (area-wide and near-road), there is additional
 8    explanation needed. For the annual standard and each separate year of air quality, the result of
 9    dividing the standard level (i.e., 53  ppb) by the monitor design value is used to estimate an
10    annual adjustment factor. For example, Table 2-2 indicates the maximum annual average
11    concentration in the Philadelphia CBSA for year 2011 is 20 ppb. Thus, 53/20 yields the
12    minimum proportional adjustment factor of 2.65 based on all available annual average
13    concentrations calculated for the Philadelphia monitors operating in 2011. When considering the
14    hourly standard, the maximum design value for the 3-year average DM1H 98th percentile for
15    2011-2013  is  61 ppb (Table 2-2). Dividing the DM1H standard level of 100  ppb by 61 ppb yields
16    a proportional adjustment factor of 1.64 for  this three-year averaging period  (2011-2013). For all
17    available area-wide monitors, design values were used in this manner to calculate all possible
18    proportional adjustment factors, with the minimum factors for each  year or period summarized
19    below in Table 2-10.45
20          While the currently available near-road data do not constitute a complete year for most
21    monitoring sites and years (including the Philadelphia CBSA), the near-road data can  still be
22    used to inform the proportional adjustment factor. Many  near-road monitors will likely have a
23    complete year of data  for 2014, though it is  highly unlikely that any monitor would have a
24    complete set of three-year continuous monitoring available for this review. Therefore, the true
25    design values for the current DM1H standard cannot be calculated using the near-road monitor
      45 Plots of the high-concentration and low-concentration year data for the Philadelphia CBSA (both using the same low- and
      high-concentration years as Rizzo (2008) and updated to include the most recent low-concentration year) are provided in
      Appendix B.
                                                 2-49

-------
 1    data. However, we feel it is informative to our estimates to calculate the similar standard
 2    averaging time metrics for each type of monitor (near-road and area-wide) using year 2014
 3    alone,46 as well as for any other comparable years where a comparable statistic can be calculated
 4    using the near-road data. In Philadelphia, while the near-road monitor (ID 421010075) collected
 5    measurement data in 2014, the full year of data has not yet been uploaded to EPAs' Air Quality
 6    System (AQS). Thus when using the available concentration data from the near-road monitor
 7    (currently, the first three quarters of the year), an estimate of the annual average and single year
 8    DM1H 98th percentile are used here to calculate the two potential  adjustment factors. Similarly
 9    and using the area-wide monitor concentration data available for 2014, two potential adjustment
10    factors are also calculated. Both the near-road and area-wide adjustment factors calculated using
11    the incomplete year data set are provided in Table 2-10.
12          As described above in screening the ambient monitoring data for areas having high
13    ambient concentrations, the hourly standard is the controlling standard for each of the years
14    considered in this analysis. When compared to the annual metric,  the hourly metric has the
15    lowest estimated proportional adjustment factors in all comparable instances (or conversely, the
16    greatest relative monitor design value). In the Philadelphia CBSA, the monitor having the
17    highest design value for 2011-2013 (ID 421010004) yields the minimum adjustment factor of
18    1.64, while another area-wide monitor (ID 421010047) yields the minimum adjustment factor of
19    1.66 for 2014. Note that in the Philadelphia CBSA, the area-wide  monitor adjustment factor is
20    less than the estimated adjustment factor of 1.78 when using the available 2014 near-road data
21    (Table 2-10). Therefore in this CBSA, the near-road monitor concentrations will not directly
22    inform the value of the proportional adjustment factor.
23          For all of the proposed study areas  selected in this air quality assessment, a single
24    proportional factor, derived from one monitor in a CBSA, is used to adjust concentrations (up to
25    and including the DM1H 98th percentile) measured at all of the monitors in that CBSA. Thus, the
26    monitor having the highest design value will have adjusted concentrations that just meet the
27    existing hourly standard (a 3-year average DM1H 98th percentile of 100 ppb), while all other
28    monitors will have hourly design values less than that value. This assumption in applying the
29    single factor derived from one monitor to other monitors is reasonably justified by the following
30    analysis, where we compared DM1H concentrations at the monitor having the highest design
      46 The set of 3-year average DM1H 98th percentile design values for 2012-2014 will be calculated for the existing area-wide
      monitors when these data become available, however still, metrics calculated for the near-road data are most appropriately
      compared with area-wide monitor data on an individual year basis.
                                                 2-50

-------
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
value (421010004) with concentrations measured at the other monitors in the CBS A having
measurement data for the same years.
Table 2-10. Proportional adjustment factors calculated for the Philadelphia CBSA, 2011-
2014.
Monitor
Data Set
Area-Wide
Near-Road
Standard
Averaging
Time
Annual
DM1H
Annual
DM1H
Year
2011
2012
2013
2014
2011-2013
2014a
2014
2014a
Proportional
Adjustment
Factor
2.65
2.94
3.12
2.99
1.64
1.66
3.50
1.78
Bold font indicates the factor to be applied to proportionally adjust hourly ambient concentrations from the minimum through the
QS'percentileDMlH.
a A single year is used for the DM1H for 2014 alone to calculate an adjustment factor because neither the near-road or area-wide
data set contained a complete year.

       In considering the set of valid years of recent (2001-2013) ambient monitor data available,
we selected a number of years during which simultaneous measurements were made at the
monitor having the highest current design value and at the five Philadelphia CBSA monitors
having more than one year of monitor data over this period. We used a simple linear regression
of the DM1H concentrations (0 through 98th percentiles) for each year to generate regression
parameter estimates and standard errors, tests of significance (students-t), and linear model fit
statistics (R2) (Table 2-11). In general, the monitor having the highest design value (for 2011-
2013) has an overall higher concentration distribution than the other monitors, even considering
the earliest years of data included in this analysis. More specifically, there is consistency in the
estimated regression slopes for each monitor over time, generally ranging from about 0.85-0.95,
and most regression intercepts are negative and generally range from -3 to -5 ppb. Model R2 are
0.98 or better, indicating a strong linear fit for all regressions.
       These linear regression results suggest similar changes in the overall ambient air quality
distribution have occurred with each monitor compared to the monitor having the highest design
value over this period of time, providing support to the use of a proportional adjustment factor
developed this single monitor and applied to the other monitors in the Philadelphia CBSA. The
consistent presence of small, though  statistically significant negative intercepts could indicate the
contribution of either a constant source emission or a transformation process exclusively
influencing concentrations  at the monitor having the highest design value. Note that when
adjusting this monitor to just meet the hourly standard,  this factor (the value of the regression
                                                 2-51

-------
1
2
3
4
5
6
7
 9
10
11
12
13
14
15
16
17
     intercept, or this potential exclusive source contribution/transformation process) would also
     increase proportionally. However, the relationship between the other monitors and the monitor
     with the highest design value after adjusting hourly concentrations remains preserved (i.e., the
     concentration adjustment would not change the regression slope).
     Table 2-11. Slope and intercept parameter estimates regressing DM1H concentrations (0-
     98th percentile) from the monitor having the highest design value (ID 421010004) on DM1H
     concentrations measured at five Philadelphia CBSA area-wide monitors.

Monitor ID
420450002
340070003
420170012
420910013
421010047
421010004
Year
2001
2004
2009
2013
2001
2004
2007
2001
2004
2009
2013
2001
2004
2007
2001
2004
2009
2013
Intercept
Estimate
-0.20
-3.47
-3.18
-5.19
-7.77
-3.86
1.13
-4.79
-1.73
-5.79
-5.06
-5.84
-6.05
-1.94
4.34
6.90
2.98
-0.19
SE
0.45
0.39
0.15
0.19
0.50
0.33
0.29
0.37
0.36
0.21
0.32
0.42
0.21
0.24
0.29
0.24
0.35
0.27
ProbT
0.66
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
0.49
Slope
Estimate
0.86
1.00
0.92
0.89
1.11
1.01
0.90
0.97
0.85
0.91
0.83
0.92
0.88
0.87
0.98
0.85
0.91
0.96
SE
0.01
0.01
0.00
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.01
ProbT
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
R2
0.9852
0.9899
0.9975
0.9951
0.9893
0.9929
0.9924
0.9921
0.9885
0.9953
0.9841
0.9887
0.9964
0.9944
0.9954
0.9949
0.9870
0.9913
           The second step of the approach used to adjust ambient concentrations such that monitors
    in the CBSA just meet the existing standard(s) involves the use of individual monitor-based
    adjustment factors. This second set of adjustment factors is used to simulate the observed
    deviations in linearity when comparing low-concentration to high-concentration years using the
    same monitor. To derive this set of additional adjustment factors, we first evaluated changes (if
    any) in the overall distribution of air quality that could have resulted from ambient air quality
    rule-driven changes in emissions, with a particular focus on the upper percentile concentrations.
    Staff used the complete historical ambient monitor concentration data set (1980-2013) and first
    calculated daily maximum concentrations at each monitor having some measurement data across
                                               2-52

-------
 1    the three-decade period,47 then generated percentiles of the concentration distribution in 1
 2    percent increments, and finally normalized this distribution to reflect a mean of zero and having
 3    a standard deviation of one. Normalized DM1H concentrations for the upper 50th percentile (i.e.,
 4    51-100 percentiles) were analyzed using principal components analysis (PCA)48 to determine
 5    whether there were multivariable features of the concentration distribution expressing
 6    correlations with important explanatory variables such as monitor year or  site ID.
 7           The first three components output from the PC A comprised nearly  63% of the variance
 8    within the data set, effectively reducing the original 50-variable data set to three new, though
 9    consolidated variables. Results for the first principal component (PCI, explaining nearly 40% of
10    the variance) are provided in Figure 2-7. Much of the concentration distribution  (upwards to
11    about the 90th percentile) is positively correlated with the first component  (i.e., the PCI
12    loadings). The remaining upper percentile DM1H variables gradually shift to being  negatively
13    correlated with the first component, beginning with the 95th percentile concentrations and
14    progressing to having maximum negative correlations at the  99th and 100th percentiles. The PCI
15    scores are calculated for each individual monitor site-year and are a linear combination of the
16    PCI  loadings and the normalized concentrations. When the PCI  scores are plotted against the
17    year the monitor data were collected, there is a general trend of increasing score with increasing
18    monitor year. The historical (1980s) monitoring data have mostly negative PCI scores, then a
19    transition from negative to positive scores  occurs at or around the mid to late 1990s, with recent
20    (2000s) monitoring data having mostly positive scores for the first principal component.
21    These observed changes in the overall NCh concentration distribution over time  and occurring
22    across all monitors in the Philadelphia CBS A could correspond to the response to a  final EPA
23    rule (the 1998 NOx SIP Call). The rule required 22 states (including Pennsylvania) and the
24    District of Columbia to submit State implementation plans (SIPs) for reducing emissions of NOx
25    with such emission reduction measures to be in place by May 1, 2003. Thus, when considering
26    the observed relationship between PCI scores  and PCI loadings  and the consistent relative
27    reductions in the upper percentile concentrations (95th and above) compared with other portions
      47 There were a total of six monitors in the Philadelphia CBSA having data spanning the 1980-2013 monitoring period: IDs
      420450002, 420910013, 340070003, 420170012, 421010004 and 421010047.
      48 Principal components analysis can be thought of as developing a progressive series of individual linear correlations through a
      multidimensional variable space (i.e., in this illustrative example, 50 dimensions), with the first correlation/component explaining
      the greatest amount of variance existing across that multidimensional space (via a single variable comprised of some linear
      combination of the original 50 variables). Subsequent components explain progressively less variance than prior components,
      though are generated to again have maximum explanatory power given the unexplained variance that remains.
                                                  2-53

-------
2
3
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
     of the concentration distribution, and its correspondence with the timing of the NOx emissions
     reduction rule, we elected to use the recent ambient concentration data to best simulate deviation
     from linearity at upper percentile concentrations that may exist when considering our
     hypothetical high-concentration adjusted air quality scenario.
                Philadelphia Ambient Concentrations: PCI (38.9%)
                     Daily Maximum 1-hour Percentile(by loadings)
                        91          81           71          61
                                                                          51
           15
          10 --

           5 -:
        2
       o   n
       i/i   u
       T-\
       *,
          -10 --
          -15
                                        A
                         AAA

                                                   I.
   •          :      .!»    ji1.,'««'.    v*
     ii-".    -   .:»•:             '    '      .-.
       * *      • I * *       •   *
                   t**
                                                                 0.15

                                                               -  0.1
                                                               i
                                                               -  0.05
                                                                             0.2
                                                                                   §
  -0.05 U

- -0.1

- -0.15

  -0.2
1980     1985    1990    1995     2000    2005
                         Year (by scores)
                                                                 2010
                                                                         2015
    Figure 2-7. Principal components (PC) monitor scores plotted by year along with PC
    loadings plotted by DM1H percentile concentration value using the Philadelphia CBSA
    NOi ambient concentrations (1980-2013). The scores and loadings plots are an overlay and
    not intended to directly relate the values of the primary (year) and secondary (DM1H
    percentile) X-axes.
           The second and third principal components explained 15.4% and 8.5% of the remaining
    variance, respectively (Figure 2-8). The pattern in PC scores indicates a limited degree of
    monitor-specific individuality for a few of the sites (IDs 42101004, 421010047, 420450002) and
    site-years (most in the 1990s), as distinguished from scores at other monitor site years. Again,
    upper percentile variables (largely the 90th and above) are strongly correlated with both PC2
    (positive) and PC3 (negative) and largely discordant with the lowest percentile variables (50th's,
    PC2) and mid percentile variables (65-70th s). These results could indicate nuanced changes in
    the NCh concentration distribution resulted from a localized emission change during that time
                                            2-54

-------
1 period, supporting the development of individual monitor-based adjustments to simulate upper
2 percentile concentrations.























3
Philadelphia Ambient Concentrations: PC2 & PCS
PC2 Loadings
0.28 0.21 0.14
8 -






"J 2
Q l

in
£ 0 -
u
4/1 -2
U
n
-A -
^r
p.
D
-8 -

i
-

1 1 1 1 1
p9S A

i i i i ii

0.07 0 -0.07
'
1 1 | ! 1 1 1
t 4iiOiflSM- laSB,
!
I 1 1 1 | 1 1
• 421010004 1993
* f>97 421010O47 19SS

-
-
-
-
_
-
-

-
-
-


"
1

-15

A I
A 59*
p93
A P92 j

A


• scores
* loadings



• 42045QQQ2

A p31
96
42101034


p9Q

p89
A pSS
A pS7
• 4201


• 4210100W 1996 • 421010004
P& i*» A»S»p79

A p7i
APS1
P7S
^^Sfo"

^dSoDia. •>&* 3^£ffij
4-u4MJUj«^M^ay|
T m j»_ iS^8»
"•4f$Sl
420450002 • 1994




-0.14 -0.21

ii i i i i
_


1994 A fcff
p5S p57
Ap59
• 42017C012 19S6 A P60

JLjUUXBMiiftml

^•BS
liM^fp-itsfift. -
•^Y)A± " loan
i&j^l-iSi $K
pS^Wia!^ 2013
SJ^S^^^5 "
Ajpipf • 42O4EOO02 2O09 *
420450002 199]



i i i i

• 42101COCH* 4&&-'3:>"
£ 1935* fl-n f\*!-f\in -i
421010004 19SO


A p72
A pE
1 1 1 1

AA*t^10*°^10*
i i i i

B$# ?9Sj
• :

- 0.27



r 0.18

:56
- 0.09 *
Quo
w


'~O
- o g
ro
U
- -0.09 °-


- -0.18
l- -0.27
-10 -5 0 5 10
PC2 Scores (15.4%)






4 Figure 2-8. Principal components scores plotted by loadings for each the second (PC2) and
5 third (PCS) components
6 (1980-2013).
derived using
the Philadelphia CBSA NOi ambient concentrations
The scores and loadings plots are an overlay and not
intended to directly
7 relate the values of the primary and secondary axes.
8
9 Based





on this analysis and the potential impact emission reductions have on
10 concentrations measured across study areas and at local
monitoring sites, we calculated the
1 1 second set of adjustment factors using the 2003-2013 ambient monitoring data in the
12 Philadelphia CBSA. As described above, there are up to seven DM1H
13 98th percentile DM1H for
a full year of ambient monitor data. At each
concentrations above the
monitor and for each year
14 of data, the seven highest DM1H concentrations were divided by their site-year's 98th percentile
15 DM1H concentration to generate a ratio.
16 calculated by
An adjustment factor for each monitor and DM1H was
averaging across the, at most, 1 1 ratio values (years) and are summarized in Table
17 2-12. There is general consistency in the
1 8 4th highest maximum, but
adjustment factor across monitors upwards to about the
beyond this point up to the maximum DM1H adjustment factor there is
2-55

-------
1
2
3
4
5
6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
      increasing variability in the factor across the monitors. The maximum adjustment factor
      exhibited the widest range of values; the minimum adjustment factor at monitor ID 340070002 is
      about 21%, while at monitor ID 421010002, the adjustment factor to be applied to the 98th
      DM1H to estimate the maximum DM1H is 56%.
      Table 2-12. Individual monitor-based factors calculated to adjust DM1H ambient NCh
      concentrations above the 98th percentile DM1H in the Philadelphia CBSA.
Site ID
4210100753
10003 10 10b
100032004
340070002°
420170012
420450002
421010004
421010047
Adjustment factor derived from ratio of DM1H concentrations to 98th percentile DM1H,
averaged across years 2003-2013
Max
DM1H
1.38
1.44
1.53
1.21
1.32
1.56
1.38
1.55
2nd DM1H
1.21
1.22
1.22
1.11
1.18
1.35
1.21
1.35
3rd DM1H
1.15
1.17
1.15
1.08
1.10
1.15
1.15
1.19
4th DM1H
1.09
1.10
1.12
1.05
1.08
1.10
1.09
1.14
5th DM1H
1.06
1.07
1.08
1.04
1.04
1.05
1.06
1.09
6th DM1H
1.04
1.05
1.02
1.04
1.03
1.03
1.04
1.04
7th DM1H
1.01
1.02
-
1.01
1.01
1.02
1.01
1.01
     a The near-road monitor (421010075) uses the ratios derived from the monitor having the highest design value (42101004).
     b Monitor ID 100031010 is newly sited (2013) and outside urban core of Philadelphia (outside Wilmington, DE). Data from a
     similar monitor (420910013) located outside urban core of Philadelphia (Montgomery County PA operating 2003-2008) was
     used to calculate ratios.
     c Monitor ID 340070003 (operating during 2003-2008) is sited in close proximity to newly sited monitor ID 340070002
     (operating during 2012-2013). The data from both monitors were combined to calculate ratios.

           To estimate ambient concentrations that just meet the existing standard for years
     evaluated in this illustrative example, for each year of ambient monitoring data and at all
     monitors, the appropriate year proportional factor described above and found in Table 2-10 is
     applied to all DM1H concentrations up to and including the 98th percentile DM1H. Then, the
     remaining upper percentile concentrations above the DM1H 98th percentile are adjusted for each
     of the eight monitors in the Philadelphia CBSA (the seven area-wide and one near-road monitor)
     using the individual monitor-based adjustment factors provided in Table 2-12.
            The results of applying these adjustments are illustrated in Figure 2-9 for one year of
     concentrations (2011) measured at the monitor (ID 421010004) having the highest design value.
     Plotted in this figure are the unadjusted DM1H concentrations, concentrations adjusted to just
     meet the  existing hourly standard using a proportional factor alone (and used in the 2008 REA),
     and concentrations adjusted to just meet the existing hourly standard using a proportional factor
     and additional factors for concentrations above the 98th percentile DM1H. Concentrations at or
     above the DM1H 80th percentile are plotted to highlight this portion of the distribution. Using the
     proportional adjustment alone for this year (an increase of 64%) appropriately increases the 80th
     and 100th DM1H unadjusted concentrations of 44 and 88 ppb (gray line, Figure 2-9) to 72 and
                                                 2-56

-------
1
2
3
4
5
6
 9
10
11
12
13
14
15
16
17
      144 ppb (red line, Figure 2-9). When addressing deviations from linearity above the 98th
      percentile, upper percentile concentrations extend to somewhat higher concentrations (blue line,
      Figure 2-9) when compared with that using a proportional factor alone to adjust all
      concentrations. For example and by design, the proportionally adjusted 98th percentile DM1H
      concentration of 125 ppb is used with the maximum DM1H adjustment factor of 1.38 to estimate
      a maximum DM1H concentration of about 173 ppb.
                                       PHIL421010004-2011
                     -As Is
                               -All Proportional
-Proportional to 98pct, Non-Linear Above
          100 --
           95 4-
       o
       a.
       $
       V=
       —
       •3
       £
           90 --
                                                                • Adjustment addressing non-linearity
           85 --
                         . .«P«, ..„,,„, ™~,.~.   ^
                      • (from design monitor)   - *
                         60
                              70   80    90   100   110  120   130  140
                                Daily Maximum 1-hr NO2 Concentration (ppb)
                       150   160   170
    Figure 2-9. Distribution of unadjusted (as is) 2011 ambient NCh concentrations, that
    adjusted using a proportional factor alone (all proportional), and that adjusted using a
    combined proportional factor and ratio approach (proportional to 98th percentile, non-
    linear above) in the Philadelphia CBSA at monitor ID 421010004.
    2.3.2.3 Simulated on-road concentrations
           In this example calculation, we applied results from one of the proposed approaches
    discussed in section 2.3.1.3.2 to simulate on-road concentrations for two air quality scenarios;
    the first scenario using the unadjusted near-road concentrations (monitor ID 421010075) and the
    second scenario following an adjustment to just meet the existing standard at this same near-road
    monitor. A concentration increase of 15% developed from the Las Vegas study was universally
                                               2-57

-------
 1    applied to the available 2014 near-road monitor DM1H NCh concentrations49 to estimate the
 2    distribution of on-road NCh concentrations (see section 2.3.1.3.2 and Table 2-7).
 3    2.3.2.4 Number of benchmark exceedances
 4          A total of two air quality scenarios were evaluated in the Philadelphia CBSA (i.e.,
 5    unadjusted concentrations, those adjusted to just meet the existing hourly standard), and for both
 6    of these air quality scenarios, we simulated on-road NCh concentrations based on the available
 7    2014 near-road monitoring data. Staff counted the number of days per monitor site-year the
 8    DM1H ambient concentrations exceeded the 1-hour benchmark levels of 100,  200, 300, and 400
 9    ppb, separately considering the area-wide, near-road and on-road concentrations. Presented are
10    the mean (averaged across all monitors for the area-wide) and maximum (maximum at a single
11    monitor) number of benchmark exceedances for each year.
12          Benchmark exceedance results using the unadjusted 2011-2014 air quality are presented
13    in Table 2-13. There were no exceedances of any benchmark when considering unadjusted air
14    quality measurements made at either the area-wide or near-road monitors. When simulating on-
15    road concentrations using the unadjusted 2014 near-road concentrations, there were no
16    exceedances  of the 100 ppb 1-hour benchmark level per year (and hence no exceedances of any
17    of the higher benchmark levels). As a reminder, a complete year of data is not yet available for
18    2014, thus it  is still possible that there could be an occurrence of a benchmark exceedance for
19    this year.
20          Benchmark exceedance results using the unadjusted 2011-2014 air quality adjusted to just
21    meet the existing 1-hour standard are presented in Table 2-14. On average, there are a handful of
22    exceedances  per year of the 100 ppb 1-hour benchmark level considering the area-wide monitors,
23    though the maximum estimated number of days with  an exceedance of that same benchmark in a
24    single year could be just over 20. Note that because the 2014 monitoring data were not yet fully
25    complete, it is possible there could be additional days having benchmark exceedances. However,
26    the actual estimated number of benchmark exceedances when the monitoring set is complete is
27    likely to fall within the range already estimated using the other years of air quality. When
28    simulating on-road NCh concentrations using the 2014 near-road monitor data adjusted to just
29    meet the existing 1-hour standard, there is only one additional day having an exceedance of the
30    100 ppb benchmark level compared with that estimated at that near-road monitor. If
      49 The Philadelphia near-road monitor 421010075 is sited 12m from the target roadway.

                                               2-58

-------
1
2
3
4
5
6
7
      extrapolating this data simply to represent a full year, it is possible that there could be as many as
      8 days per year where the 100 ppb 1-hour benchmark level is exceeded. For all study area
      locations (area-wide, near-road, and on-road) there were no exceedances of 1-hour benchmark
      levels at or above 200 ppb.
      Table 2-13. Mean and maximum number of days per year ambient monitor NOi
      concentrations (area-wide, near-road, simulated on-road) are at or above selected 1-hour
      benchmark levels in Philadelphia CBSA, unadjusted air quality.
Study Area
Location
Area-Wide
Near-Roadb
On-Roadb
Year
2011
2012
2013
2014a
2014a
2014a
DM1H > 1C
Mean
0
0
0
0
-
-
)0 ppb
Max
0
0
0
0
0
0
DM1H>
Mean
0
0
0
0
-
-
200 ppb
Max
0
0
0
0
0
0
DM1H>
Mean
0
0
0
0
-
-
300 ppb
Max
0
0
0
0
0
0
DM1H>^
Mean
0
0
0
0
-
-
00 ppb
Max
0
0
0
0
0
0
 9
10
11
12
13
14
     a The monitoring data for available for 2014 are not for a full year (e.g., the near-road monitor has data for quarters 1 through 3).
     b There is only one near-road monitor in the Philadelphia CBSA, therefore means are not calculated.

     Table 2-14. Mean and maximum number of days per year ambient monitor NOi
     concentrations (area-wide, near-road, simulated on-road) are at or above selected 1-hour
     benchmark levels in Philadelphia CBSA, air quality adjusted to just meet the existing
     standard.
Study Area
Location
Area-Wide
Near-Road"
On-Roadb
Year
2011
2012
2013
2014 a
2014 a
2014 a
DM1H>
Mean
6
2
1
3
-
-
100 ppb
Max
23
4
2
6
5
6
DM1H>2
Mean
0
0
0
0
-
-
00 ppb
Max
0
0
0
0
0
0
DM1H>3
Mean
0
0
0
0
-
-
00 ppb
Max
0
0
0
0
0
0
DM1H>4
Mean
0
0
0
0
-
-
00 ppb
Max
0
0
0
0
0
0
15
16
17
18
19
20
21
22
23
24
25
26
     a The monitoring data for available for 2014 are not for a full year (e.g., the near-road monitor has data for quarters 1 through 3).
     b There is only one near-road monitor in the Philadelphia CBSA, therefore means are not calculated.

           For comparison, the 2008 REA calculated the number of exceedances for two of the
     benchmarks (i.e., 100 and 200 ppb) using 2001-2003 ambient air quality adjusted to just meet the
     now existing standard, as well as using a probabilistic adjustment factor to simulate on-road NCh
     concentrations (Table 2-15). Results for the 2008 REA were summarized by site-year rather than
     by year alone and the available  area-wide monitors were separated by two distance from road
     categories; nevertheless the overall comparison to the current  analysis remains meaningful to a
     certain extent. There is consistency between the 2008 REA estimates and the current calculations
     when considering the area-wide monitors adjusted to just meet the existing standards, as both
     indicate on average approximately 3-6 days per year where the 100 ppb 1-hour benchmark is
                                                2-59

-------
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
      exceeded. Both analyses also predict, on average, no exceedances of the 200 ppb 1-hour
      benchmark at the area-wide monitors with air quality adjusted to just meet the existing standards.
            There are however large differences in the number of days per year having exceedances
      when considering the simulated on-road concentrations. In the 2008 REA, on average, over 100
      days per year were estimated to have an exceedance of the 100 ppb 1-hour benchmark, along
      with an upper percentile estimate of nearly 300 days per year. This is likely the result of the
      selected monitor data used in the 2008 REA, whereby monitors sited at least 100 m from a major
      road were used to simulate the on-road NCh concentrations. The on-road simulation approach
      assumed monitors sited 100 m or greater from a road was a reasonable distance to not have a
      direct influence from road emissions, an important uncertainty identified in that assessment
      (2008 REA, section 7.4.6). In reviewing the current monitor attribute data provided in Table 2-8
      and Table 2-9, while it is possible that at that distance from a road these monitors would have
      limited direct contribution from roadway emissions, there remains the potential for other source
      emissions to substantially influence concentrations measured at the monitor that were not
      accounted for before simulating the on-road concentrations. For example, NCh concentrations at
      monitor ID 420450002 (located 413m from a major road) could be largely influenced by several
      stationary sources, including NOx emissions from electricity generation (via combustion),
      petroleum refineries, and municipal waste combustion (Table 2-8). Using this generally high-
      concentration monitor (see 2008 REA Appendix A, Table A-54 and Figure A-54) and other
      monitors having similar stationary source emissions in close proximity (e.g., monitor
      421010029) to simulate the on-road NCh concentrations would likely overestimate the number of
      benchmark exceedances.
      Table 2-15. Mean and upper percentile number of days per year ambient monitor NOi
      concentrations (area-wide and on-road) are at or above selected 1-hour benchmark levels
      in the 2008 REA Philadelphia CBSA, 2001-2003 air quality adjusted to just meet the
      existing standard.
Area-wide Monitor
distance from road
>20m-<100mb
>100mc
On-Roadd
Site-years a
(n)
7
14
1400
DMlh
Mean
5
3
116
\ > 100 ppb
P99
11
15
294
DMlh
Mean
0
0
7
> 200 ppb
P99
0
1
68
      1 For area-wide monitors, this is the number of measurement monitor site years available (e.g., two monitors for two years with 3
      monitors for one year equals 7 site-years of data). The on-road site years are from 1,000 simulations using the 14 site-years of
      data from monitors sited >100m from a major road).
      b From Table 7-23 (U.S. EPA, 2008a).
      c From Table 7-24 (U.S. EPA, 2008a).
      d From Table 7-28 (U.S. EPA, 2008a).
                                                2-60

-------
 i    3   HUMAN EXPOSURE ASSESSMENT
 2         In the last review, in addition to analyses of NCh air quality, the EPA used an exposure
 3    model to generate estimates of 1-hour personal NCh exposures in an urban study area. These
 4    modeled 1-hour personal exposures were compared to 1-hour health effect benchmarks ranging
 5    from 100 to 300 ppb.50 In the current review, staff will use results of the updated comparison of
 6    NCh air quality with health effect benchmarks (Chapter 2, above) to consider the potential utility
 7    of performing an updated assessment of personal NCh exposures. To the extent the air quality
 8    assessment indicates little potential for the occurrence of ambient NCh concentrations at or above
 9    the various 1-hour health effect benchmark levels (i.e., indicating little potential for NCh
10    exposures of public health concern), the added value of more refined estimates of personal NCh
11    exposures would be limited. Alternatively, to the extent the air quality assessment indicates the
12    potential for NCh exposures of public health concern, more refined estimates of NCh  exposures
13    will be considered in the current review.
14         In this latter scenario, conclusions on whether to conduct an updated NCh exposure
15    assessment will also be informed by 1) the extent to which important uncertainties identified in
16    the last review have been addressed by newly available information, approaches, and tools, and
17    2) judgments as to the likelihood that an updated quantitative assessment would substantially add
18    to our understanding of NCh exposures, beyond the insights gained from the exposure
19    assessment conducted in the last review. This chapter provides overviews of the exposure
20    assessment conducted in the last review (section 3.1) and the new information, approaches and
21    tools available in the current review that could potentially inform an updated exposure
22    assessment, should one be supported (section 3.2).

23    3.1  OVERVIEW OF EXPOSURE ASSESSMENT IN THE LAST REVIEW
24          In the 2008 REA, population exposures to ambient NCh were simulated using  the Air
25    Pollutants Exposure (APEX) model (U.S. EPA, 2008a, Chapter 8).  The APEX model simulates
26    the movement of individuals through time and space across a user-defined modeling domain and
27    estimates their exposures to pollutants within indoor, outdoor, and in-vehicle
      50 As described above for the air quality assessment (section 2.1), 1-hour health effect benchmark concentrations were based on a
      meta-analysis of individual-level data from controlled human exposure studies of NCh-induced airway responsiveness (U.S.
      EPA,2008b, Table 3.1-3).


                                               3-1

-------
 1    microenvironments. APEX takes into account the most critical factors that contribute to total
 2    human exposure to ambient NCh, including the temporal and spatial distributions of people and
 3    the NCh concentrations they experience as they travel throughout an urban study area (including
 4    on- and near-roads), and the variation of NCh concentrations within various microenvironments
 5    (U.S. EPA, 2008a, section 8.2).
 6          The ambient concentrations for the exposure assessment in the last review were generated
 7    using the EPA's air dispersion model, AERMOD. AERMOD was used to simulate hourly NCh
 8    concentrations at a census block-level and at link-based roadway receptors, considering
 9    emissions from stationary, area-wide, and on-road mobile sources (U.S. EPA 2008a, section 8.4).
10    Multiple air quality scenarios were evaluated, including air quality adjusted to just meet the
11    existing annual standard and air quality adjusted to just meet alternative 1-hour standards with
12    levels ranging from 50 to 200 ppb (U.S. EPA, 2008a, section 8.9.4).51 Given the  resource-
13    intensive nature of this approach, only one study  area was selected for analysis (four counties
14    comprising the core urban area in Atlanta, GA). The exposure assessment served to complement
15    the results of the broader, less resource-intensive NCh air quality characterization described
16    above that largely relies on existing ambient monitor concentrations (section 2.1).
17          Exposure estimates focused on people with asthma, based on the evidence for potentially
18    increased health risks following NCh exposures (U.S. EPA, 2008b). Exposures were estimated
19    for all people with asthma (0 to 99 years of age) and in school-age children with  asthma (5 to 17
20    years of age). The 2008 REA noted that children spend more time engaged in outdoor activities,
21    possibly increasing their NCh exposures.  The 2008 REA compared the estimated 1-hour personal
22    exposures to health effect benchmarks ranging from 100 to 300 ppb; these benchmarks  were
23    based on the 2008 ISA's assessment of controlled human exposure studies that evaluated  airway
24    responsiveness following NCh exposures (see above, sections 2.1 and 2.2.2) (U.S. EPA, 2008a,
25    section 8.9.4).
      51 Adjusted air quality was based on the years 2001 to 2003. This three-year period was selected to encompass the most recent
      year of NOx emissions data available (i.e., 2002) at the time the exposure assessment was conducted (U.S. EPA, 2008a).
                                                 5-2

-------
 1    3.1.1 Key Results
 2          The 2008 REA presented a number of results from the exposure assessment in Atlanta
 3    (U.S. EPA, 2008a, section 8.9.4 and Appendix B), including the following:
 4       •  Roadway-related exposures accounted for more than 99% of exposures to NCh
 5          concentrations at or above 1-hour health effect benchmarks in Atlanta (U.S. EPA, 2008a,
 6          Figures 8-17 and 8-18). Of these roadway-related exposures, approximately 70% were
 7          estimated to occur in vehicles, with the remainder estimated to occur outdoors near roads.
 8       •  When air quality was adjusted to just meet the existing annual standard in Atlanta, almost
 9          all people with asthma (i.e., 88% to 99%, depending on the year) were estimated to
10          experience 1-hour exposures to NCh concentrations at or above 300 ppb at least six times
11          per year.52 For 1-hour health effect benchmarks of 100 or 200 ppb, all people with asthma
12          were estimated to experience at least six exposures to these concentrations per year (U.S.
13          EPA, 2008a, section 8.9.4.3; Appendix B, Tables B-43 to B-54).
14       •  Compared to the annual standard, when air quality was adjusted to just meet 1-hour
15          standards with levels of 100 or 50 ppb, there were substantial reductions in the number of
16          people estimated to experience six or more exposures per year to 1-hour NCh
17          concentrations at or above 300 ppb. Reductions were more modest for lower health effect
18          benchmarks (i.e., 100 to 200 ppb) and for smaller numbers of occurrences (i.e., one or
19          more, two or more, etc.). Reductions were  also more modest for 1-hour standards with
20          levels of 150 or 200 ppb (U.S. EPA, 2008a, Appendix B, Tables B-43 to B-54).

21    3.1.2 Uncertainties and Limitations
22          Important uncertainties identified for the 2008 REA NCh exposure estimates included
23    some of the same uncertainties identified for the air quality assessment (section 2.1.2, above).
24    For example, uncertainties in the approach used to adjust air quality and uncertainties in health
25    effect benchmarks were also important for the 2008 exposure assessment. The 2008 REA also
26    identified uncertainties specifically associated with the exposure model inputs, approach, and
      52 Six or more days per year was the largest number of occurrences that was specifically reported in the 2008 REA. Similar
      results were obtained for all people with asthma and children with asthma, though estimated exposures were somewhat higher for
      children (U.S. EPA, 2008a, section 8.9.4.3; Appendix B, Tables B-43 to B-54).
                                                 5-3

-------
 1    estimated benchmark exceedances. These uncertainties, and their potential implications, are
 2    discussed in detail in the 2008 REA (U.S. EPA, 2008a, section 8.12).
 3           Table 3-1 (below) provides the qualitative summary of the key uncertainties related to the
 4    exposure modeling and evaluated in the 2008 REA (U.S. EPA, 2008a, Table 8-17). While our
 5    approach to evaluating uncertainties has evolved since the time of the 2008 REA,53 the 2008
 6    REA characterization remains a reasonable starting point for the discussion that follows,
 7    focusing on the key uncertainties identified in the 2008 REA and any newly identified potential
 8    elements here. As such, this current evaluation is not intended to serve as a re-characterization of
 9    all previously identified uncertainties nor does it serve to fully characterize uncertainty in any
10    newly identified elements. If a new exposure assessment is performed using a similar modeling
11    approach, each element listed here (and any newly identified elements) would be newly
12    evaluated and characterized. The following discussion focusses on a few of the most important
13    uncertainties identified  and evaluated in the 2008 REA54 along with expanded context for
14    particular elements not  discussed at that time.
15           Regarding AERMOD inputs and algorithms, one important uncertainty identified as
16    specific to the 2008 REA exposure assessment was the AERMOD estimated concentrations used
17    to represent the air  quality surface across the Atlanta study area (U.S. EPA 2008b, section 8.12).
18    A performance evaluation using  limited ambient measurement data suggested a potential bias
19    towards overestimating ambient  concentrations, potentially attributable to uncertainty in mobile
20    source emissions and/or diurnal profiles used as inputs (among other  sources of uncertainty)
21    (U.S. EPA, 2008a,  section 8.4.8). Given the few monitors available for the evaluation, and the
22    overall strong confidence in the AERMOD system and other input data, the 2008 REA did not
23    adjust all estimated concentrations across the entire 4-county modeling domain based on the
      53 Included in the most recent uncertainty characterization approach used for the Os REA are implicit evaluations of the
      magnitude and direction of influence the uncertain element has on exposure results as well as an evaluation of the degree of
      uncertainty in the knowledge base used to inform the characterization. The 2008 REA approach only characterized the magnitude
      and direction of influence.
      54 The three categorizations used in the 2008 REA to characterize uncertainty (low, medium, and high) were intended to indicate
      the impact [magnitude] the type [element] of uncertainty potentially has on the estimated exposures and the potential direction of
      bias [influence] (under- or over-estimate, or unknown). A range in the magnitude is possible when considering multiple
      influential components exist within each of the identified elements. For example, regarding the 'low - high' range reported for
      the CHAD database, it was judged that there is a small (low) impact on exposure estimates when using CHAD diary data from
      selected studies rather than using CHAD as a whole. Also, the CHAD diaries used in the exposure modeling (generally nationally
      representative data as a whole) does not directly account for longer drive times observed for Atlanta commuters, potentially
      leading to large (high) exposure underestimations (see 2008 REA. Section 8.12.2.2). Details as to how particular judgments were
      made regarding each element in Table 3-1 here are provided in section 8.12 of the 2008 REA.
                                                     5-4

-------
1
2
3
4
5
6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
      differing concentrations observed at the few monitor locations. Nevertheless, ensuring the proper
      characterization of the hourly ambient concentrations input to APEX in estimating exposure is
      important and, in the absence of having spatially and temporally robust ambient measurement
      data, this could remain as a key uncertainty.
      Table 3-1. Summary of 2008 REA qualitative uncertainty analysis for the exposure
      assessment.
Source
AERMOD
Inputs and
Algorithms
APEX
Inputs and
Algorithms
Type [Element]
AERMOD formulations for
mobile sources
On-road emissions
C-3 monitoring data
Use of unadjusted NO2
concentrations
Meteorological data
Population data base
Commuting data base
CHAD data base
Meteorological data
Air exchange rates
A/C prevalence
Indoor sources not modeled
Indoor decay distribution
Indoor concentration
distribution
Longitudinal profile
Concentration/
Exceedance
Bias [Influence]
Direction
unknown
over
over
unknown
unknown
both
both
under
both
unknown
none
under
under
under
both
Characterization
of Uncertainty
[Magnitude]
Low
Low- Medium
Low
Low- Medium
Low- Medium
Low
Low- Medium
Low- High
Low
Medium
Low
Medium
Low - Medium
Medium- High
Low
     Brackets [ ] in column heading indicate current/evolved terms used in characterizing uncertainty. Our current approach would
     consider uncertainty in the knowledge-base, characterized using three categorical ratings.
           Regarding APEX inputs and algorithms, the commuting database was identified as a
     potentially important influential source of uncertainty. While there is limited uncertainty in the
     U.S. Census-derived database itself per se, the potentially influential elements of uncertainty
     considered here are the overall use of it in our exposure modeling approach and what may not be
     properly accounted for in moving simulated individuals across the exposure modeling domain.
     For example, the commuting option is only applied to employed individuals; when a work event
     occurs, the individual travels to a probabilistically determined work tract and then returns to the
     home tract when a  home event occurs. Ambient concentrations used for estimating
     microenvironmental concentrations associated with other travel events (e.g., trips to school, a
     park, or grocery store) are generally limited to either the domain average of all ambient
                                                  5-5

-------
 1    concentrations (which was the approach used in the 2008 REA) or a random selection of a user-
 2    defined number of receptors.
 3          The Consolidated Human Activity Database (CHAD) serves as a fundamental component
 4    in the exposure model that connects simulated individuals with the estimated
 5    microenvironmental concentrations. There were several components of CHAD that were
 6    considered in the 2008 REA as potentially influential and contributing to uncertainty in exposure
 7    estimates including a qualitative evaluation of survey year and methodology used, however
 8    given new consideration here, a few additional components can be identified. For example and
 9    related to the commuting issue identified above, an important element was mentioned therein
10    regarding the representativeness of CHAD diary drive durations (largely from national surveys)
11    for the people simulated in the Atlanta model domain. In addition, the 2008 REA also considered
12    the potential impact on exposures from using diaries dating back to 1984 and whether these
13    could appropriately capture commute times for the simulated years (2001-2003). Not discussed
14    in the 2008 REA uncertainty characterization however, though related to both commuting and
15    the CHAD diaries used in the exposure assessment, is the lack of linking the Consolidated
16    Human Activity Database (CHAD) diary driving event durations used by APEX with each
17    simulated individual's commute distance. Also not directly identified in the 2008 REA as an
18    uncertainty, though relevant to an evaluation of the CHAD database, were the limited number of
19    diaries available for use by APEX (approximately 15,000 total), particularly those from school -
20    age children (approximately 3,000 diary days).
21          In addition, though not specifically identified as an exposure uncertainty in the 2008 REA,
22    there was uncertainty associated with the factors approach used to adjust 1-hour AERMOD
23    ambient concentrations to predict on- and near-roadway concentrations. While AERMOD
24    predicted 1-hour NCh concentrations occurring at roadway link-based receptors, these estimated
25    on-road NCh concentrations could not be used directly as an input to APEX based on its existing
26    configuration. Thus, a distribution of factors were developed from the AERMOD predicted on-
27    road and census tract level concentrations (U.S. EPA, 2008a, section 8.7.2.5).  These differently
28    derived, though related, on-road factors used by APEX, along with the number of estimated on-
29    road peak concentrations, were compared with those used for the air quality characterization
30    (U.S. EPA, 2008a, section 8.4.8.3). The two similar, though independently developed,
31    distributions of factors used to simulate on-road concentrations were found to be comparable
32    across a wide range of estimated values, though they diverged at upper percentiles of the two
33    concentration distributions.
34          A few remaining important uncertainties identified in the 2008 REA were related to
35    estimation of indoor exposures (i.e., air exchange rates developed from a North Carolina research
36    study and applied to Atlanta), including the added exposure concentration contribution

-------
 1    originating from a single indoor source (i.e., gas stoves). Indoor source emissions will increase
 2    the frequency of peak NCh exposure concentrations, particularly given their short-term event
 3    durations (e.g., cooking with a gas stove). However, substantial uncertainty likely remains in
 4    generating reasonable emission rates for all indoor sources (e.g., gas stoves, heaters, fireplaces)
 5    and in appropriately simulating these types of events as they occur throughout the day. These
 6    general uncertainties regarding accurately estimating the source contribution to indoor
 7    concentrations, coupled with their limited relevance to improving our understanding of the
 8    relationship between ambient air quality and ambient-related exposures, and recognizing the
 9    aforementioned important uncertainties in simulating movement of individuals across the
10    modeling domain and estimating certain microenvironmental concentrations, likely preclude the
11    need for a focused consideration in this assessment.
12          Finally, broad context was added to the Atlanta study area by evaluating influential
13    attributes that may affect the estimated exposures (i.e., population density near roads, air
14    exchange rates/air conditioning prevalence, roads per capita, and daily vehicle miles travelled-
15    DVMT). We compared the values determined for Atlanta with those in other potential study
16    areas (e.g., Boston, Los Angeles, etc.) to evaluate the overall representativeness of the Atlanta
17    study area.  For a few elements Atlanta was similar to other urban areas (e.g., population per
18    total roadway miles), potentially indicating some degree of representativeness of the Atlanta
19    study area exposure results for other urban areas. For other elements Atlanta had lower values
20    (e.g., roads per capita, DVMT),  indicating a lack of representativeness. Regardless of the
21    outcome, having one study area with refined exposure  estimates in the 2008 REA presents an
22    issue regarding representativeness and is an uncertainty.

23    3.2  CONSIDERATION OF NEWLY AVAILABLE INFORMATION
24    3.2.1 Emissions Inventory
25          The 2008 REA document relied on emissions from the 2002 NEI, which was the most
26    current inventory available at that time. The NEI is compiled, in detail for all sectors every three
27    years, such that since then there are data for NEI years 2005, 2008, and the most current, 2011.
28    The NEFs NOx emissions trend has continued to show reductions, due both to actual decreases
29    and to methodological updates that improve estimation accuracy. Notable methods changes
30    include the evolution of EPA's on-road emissions models from the Mobile model, used in the
31    2002 NEI, to todays' Motor Vehicle Emission Simulator (MOVES). The more recent model
32    results for on-road, the  sector with the largest NOx emissions at about 35%, estimates slightly
33    lower overall on-road emissions and a redistribution among motor vehicle types. The following
34    graphs show NOx trends from all sectors and from the largest contributing sectors (mobile and
35    stationary fuel combustion, which account for approximately 75% of emissions), respectively.
                                                5-7

-------
3

4

5
                            Total U.S. NOX Emissions by Year
        25,000
                  NJ  NJ   NJ
                  O  O   O
                  O  O   O
                  NJ  UJ   -t»
                                          Year
                  Annual U.S. NOX Emissions by Major Sector & Year
        25,000
     ^ 20,000
     o
     O
     o

     IH 15,000
     *     '
     (O
     C
                                                     HIGHWAY VEHICLES


                                                     OFF-HIGHWAY
                                                 FUEL COMB. ELEC. UTIL.


                                                 FUEL COMB. INDUSTRIAL
                                                                 NJ
                                                                     (JJ
                                          Year
Figure 3-1. Total annual NOx emissions (top) and annual emissions stratified by top four

sectors (bottom), 2002-2013.

-------
 1         For mobile source modeling, the 2008 REA used traffic demand modeling conducted by
 2   the Atlanta Regional Commission, which resulted in four periods of a temporal profile,
 3   "morning, afternoon, evening, and nighttime." However, the most up to date version of the
 4   mobile model (currently MOVES 2014) includes emission factors that have temperature
 5   sensitivity as well as the ability to model with more specific temporal profiles. As part of the
 6   recent NEI process (2011 NEIv2), EPA has collected regional specific temporal profiles either
 7   from individual states submissions or by analyzing traffic count data from the Vehicle Travel
 8   Information System (VTRIS). These temporal profiles vary by geography as well as having
 9   distinct profiles by vehicle type and road type for day of the week and diurnal profiles (weekday
10   vs Saturday vs Sunday). These resulting profiles are expected to be more temporally
11   representative than those used previously. In addition to updating the profiles, the impact
12   temperature has on emissions could be captured by using hourly and region specific meteorology
13   (e.g., from Weather Research and Forecast modeling or from detailed ambient monitor data).
14   Temporal variability in meteorology could either increase or decrease emissions at any specific
15   hour by accounting for events such as increased emissions from air conditioning use or from cold
16   starts.

17   3.2.2  Air Quality Modeling
18          The AERMOD modeling system, including the AERMOD dispersion model and its
19          meteorological preprocessor, AERMET, have had 8 major update cycles since the last
20          NCh REA was conducted. These updates include major revisions to the NCh chemistry
21          options used to estimate NO/NCh partitioning (i.e., the Ozone Limiting Method and
22          Plume Volume Molar Ratio Method), which have resulted in more accurate estimates of
23          NO2 concentrations from stationary and mobile sources. Newer versions of AERMOD
24          have also incorporated new options for applying background data (NO2 and ozone for
25          NO/NO2 conversion) into the modeling scenario based on the wind direction and the
26          location of available monitoring data. Additionally, AERMET has been updated to
27          incorporate high-resolution meteorological data in order to provide more representative
28          and accurate inputs to the AERMOD model.

29   3.2.3  Exposure Modeling
30          There have been a number of updates made to the APEX model and many of the input
31   data sets used. Table 3-1 above highlights several  important elements of the model and its inputs
32   and serves to inform a basic structure to the discussion that follows.
                                               5-9

-------
 1    3.2.3.1 US population and commuting database
 2         APEX currently uses the 2010 census information, including an updated commuting
 3    database. This update does not necessarily confer an automatic reduction to the already limited
 4    degree of uncertainty, however it is a notable model improvement. APEX now links the CHAD
 5    diary drive times with the US census derived commute distances, a significant improvement in
 6    better estimating interpersonal variability in exposures occurring while inside a motor vehicle.

 7    3.2.3.2 CHAD activity pattern database
 8         The 2008 REA used a CHAD database containing approximately 23,000 diary days. The
 9    most recent version of CHAD master now has well over 50,000 diary days, including a
10    significantly greater number of diaries for children (nearly 18,000 diary days) and the majority of
11    the data are from year 2000 and beyond. Further, additional evaluations performed on the diary
12    data from CHAD and other activity pattern survey data, in particular time spent outdoors and the
13    frequency of outdoor event participation for both healthy individuals and those having asthma,
14    have improved our understanding of the representativeness of the database in capturing the
15    activities that people perform and the locations they visit. See U.S. EPA (2014b), Chapter 5.

16    3.2.3.3 Air exchange rates
17         Air exchange rate (AER) data used for estimating indoor residential and other building
18    exposures in the 2008 REA were re-evaluated and updated to a limited degree using recently
19    available residential AER data from the Detroit Exposure and Aerosol Research Study (DEARS)
20    (Williams et al.,  2008) and from a field study of 37 small and medium commercial buildings
21    throughout California conducted in 2008 to 2010 (Bennett et al., 2011). See U.S. EPA (2014b),
22    Appendix 5E.

23    3.2.3.4 Microenvironmental concentrations
24         The APEX model has additional options available to allow for estimation of more
25    spatially variable microenvironmental concentrations. First, microenvironment location codes are
26    now able to be linked with spatially varying ambient concentration locations  (e.g., census tracts,
27    grid points) in the modeling domain. This mapping indicates which set of ambient concentrations
                                               3-10

-------
 1    are to be used by APEX in calculating microenvironmental concentrations.55 As part of this
 2    significant model improvement, the road concentrations can now be input as a separate ambient
 3    air quality file rather than estimated using the ambient concentrations and a distribution of
 4    factors to simulate the on-road concentrations.

 5    3.2.3.5 Asthma prevalence
 6          Asthma prevalence has been updated to reflect recently collected data (2006-2010) which
 7    are now stratified by U.S. Census tracts using the most recent Census data. First, prevalence data
 8    were obtained from the Center for Disease Control (CDC) and Prevention's National Health
 9    Interview Survey (NHIS). Briefly, years 2006-2010 NHIS survey data were combined to
10    calculate asthma prevalence, defined as the probability of a "Yes" response to the question "do
11    you still have asthma?" among those that responded "Yes" to the question "has a doctor ever
12    diagnosed you with asthma?" The asthma prevalence was first stratified by NHIS defined
13    regions (Midwest, Northeast, South, and West), sex, age (single years for ages 0-17) or age
14    groups (ages > 18), and a family income/poverty ratio. These new asthma prevalence estimates
15    were then linked to 2010 US Census tract-level poverty ratio probabilities, also stratified by age
16    and age groups, to generate a final database consisting of U.S. Census tract-level asthma
17    prevalence for the entire U.S. See U.S.  EPA (2014b) Appendix 5C.

18    3.3   SUMMARY AND CONCLUSIONS
19          As described in this chapter, the decision of whether to conduct an updated model-based
20    exposure assessment will be  informed by consideration of several factors. The first of these
21    factors will be the results of updated analyses comparing NCh air quality with health effect
22    benchmarks (Chapter 2, above). To the extent these analyses indicate little potential for the
23    occurrence of ambient NCh concentrations at or above the various  1-hour health effect
24    benchmarks (i.e., indicating little potential for NCh exposures of public health concern), there
25    would be limited value to having more  refined estimates of personal NCh exposures.
      ss Seven locations are used; "Home" (H), "Work" (W), "Other" (O), "Roadway" (R), "Road near Work" (RW), "Near Home"
      (NH), and "Near Work" (NW). An 'either' location, "Last" (L), draws from either Near Home or Near Work, depending on the
      last location the individual was in. A person who is not employed has identical work and home locations. The H and W
      concentrations are calculated from the air quality data in a person's home and work sectors, respectively. The concentrations in
      the O location are calculated from a composite of set of air districts. By default, APEX uses the city-average air concentration to
      calculate O concentrations. If the user specifies roadway air quality districts, then APEX will use these AQ data to determine
      microenvironmental concentrations for R and RW locations. R is drawn from road concentrations near the Home location, while
      RW is drawn from road concentrations near the work location. NH is randomly sampled from a tract within a given distance from
      the H location, while NW is sampled near the work location.
                                                  3-11

-------
 1   Alternatively, to the extent the air quality assessment indicates the potential for NCh exposures
 2   of public health concern, generating more refined estimates of NCh exposures will be considered
 3   in the current review. In this latter scenario, conclusions on whether to conduct an updated NCh
 4   exposure assessment will also be informed by 1) the extent to which important uncertainties
 5   identified in the last review have been addressed by newly available information, approaches,
 6   and tools, and 2) judgments as to the likelihood that an updated quantitative assessment would
 7   substantially add to our understanding of NCh exposures beyond the insights gained from the
 8   exposure assessment conducted in the last review.
 9         As discussed in sections 3.2.1 to 3.2.3 above, there have been a number of improvements
10   to key exposure modeling inputs and approaches since the last review (sections 3.2.1 to 3.2.3,
11   above). This new information is substantially different from that used in the 2008 REA and
12   would be likely to appreciably reduce the uncertainties and limitations of the last assessment.
13   Therefore, to the extent health effect benchmark comparisons indicate the potential for the
14   current NAAQS to allow NCh exposures of public health concern, we reach the preliminary
15   conclusion that more refined model-based estimates of NCh exposures would be supported in the
16   current review.
                                               3-12

-------
 i    4   HUMAN HEALTH RISK ASSESSMENT
 2          For some pollutants and health endpoints, there is sufficient scientific evidence and
 3    information available to support the development of quantitative estimates of pollutant-related
 4    health risks. Depending on the evidence and information available, health risk assessments can
 5    be based on information from controlled human exposure studies or on information from
 6    epidemiologic studies. In the last review, the 2008 REA conducted a NCh human health risk
 7    assessment based on information from an epidemiology study.
 8          This chapter presents staffs considerations and preliminary conclusions regarding the
 9    information that could inform a potential updated human health risk assessment in the current
10    review. Section 4.1 discusses the extent to which the available evidence and information could
11    support a quantitative risk assessment based on information from controlled human exposure
12    studies. Section 4.2 discusses the extent to which the available evidence and information could
13    support an updated quantitative risk assessment based on information from epidemiology studies.

14    4.1   RISK ASSESSMENT BASED ON INFORMATION FROM
15         CONTROLLED HUMAN EXPOSURE  STUDIES
16         In some cases, population-level health risks can be estimated using information on
17         exposure-response relationships from controlled human exposure studies, combined with
18         modeled or monitored estimates of personal  exposures. This type of risk assessment
19         requires a robust evidence base comprised of controlled human exposure studies that are
20         similar in design (e.g., exposure methods, health effect measurements, study subject
21         characteristics56)  and that allow for quantification of an exposure-response function (e.g.,
22         see REA for ozone, U.S. EPA, 2014).
23          In the last review of the primary NCh NAAQS, a meta-analysis of information from
24    available controlled human exposure studies indicated that exposures to NCh concentrations
25    from 100 to 300 ppb could increase airway responsiveness in people with asthma, although the
26    magnitude of that increase was not quantified. As discussed in the 2008 ISA (U.S. EPA, 2008b,
27    section 3.1.3.2), there was considerable variability in methods and results across these studies,
28    and they did not provide a basis for deriving an exposure-response function. A human health risk
      56 Generalizability of results from controlled human exposure studies to at-risk populations can be limited because the most
      sensitive individuals (e.g., children, people with severe asthma) are often excluded.
                                               4-1

-------
 1    assessment based on information from these controlled human exposure studies was not
 2    conducted in the 2008 REA (U. S. EPA, 2008a).
 3          As discussed above (section 2.2.2), the evidence from controlled human exposure studies
 4    has not changed substantially since the last review. The 2nd draft ISA (U.S. EPA, 2015) includes
 5    updated meta-analyses of individual-level data from controlled human exposure studies that
 6    evaluated the occurrence of NCh-induced increases in airway responsiveness in people with
 7    asthma. In addition to assessing the direction of NCh-induced changes in airway responsiveness
 8    as was done during the last review, the updated meta-analyses also showed that about a quarter
 9    of the individuals with asthma exposed at rest to NCh experienced a clinically relevant increase
10    in airway responsiveness. These meta-analyses provide evidence supporting the occurrence of
11    increased airway responsiveness in people with asthma following exposures to NCh
12    concentrations at or above 100 ppb.
13          However, considerable variability in methods and results across these studies precludes
14    their use in deriving an exposure-response function for NCh-induced changes in airway
15    responsiveness. Specifically, these studies varied in their exposure protocols (e.g., exercise
16    versus rest, exposure durations ranged from 20 minutes to 3 hours), in the approaches used to
17    measure airway responsiveness (e.g., 20% reduction in forced expiratory volume in 1  second
18    (FEVi), 100% increase in specific airway resistance (sRaw)), in the time of measurement post-
19    exposure (e.g., immediately up to several hours), in their methods for administering
20    bronchoconstricting agents, and in the types of challenges used to induce airway responsiveness
21    (i.e., specific versus non-specific challenge) (U.S. EPA, 2015, section  5.2.2.1). Even within
22    studies that used either specific or non-specific challenge agents, there was considerable
23    variability. Non-specific airway responsiveness was evaluated using carbachol, methacholine,
24    histamine, SCh, or cold air as challenge agents. Specific airway responsiveness was evaluated
25    using ragweed, house dust mite, birch, timothy, or cat allergen as challenge agents (U.S. EPA,
26    2015, Tables 5-2 and 5-3).
27          Results are highly variable across these studies. The available information does not
28    demonstrate an exposure-dependent response and, therefore, this information is not sufficient to
29    support the derivation of an exposure-response function for use in quantitative estimates of NCh
30    health risks. Goodman et al. (2009) reached a similar conclusion, based on meta-analyses and
31    meta-regressions of information from studies of NCh-induced specific and non-specific airway
32    responsiveness. In addition, there is not strong evidence of an exposure-response relationship in
33    individual studies that evaluated exposures to multiple NCh concentrations (Bylin et al.,  1988;
34    Orehek et al., 1976). Therefore, while the available information is sufficient to support the
35    identification of health effect benchmarks for NCh, as described above (section 2.2.2), we reach
36    the preliminary conclusion that a quantitative risk assessment based on information from

                                                4-2

-------
 1    controlled human exposure studies is not supported by the evidence available in the current
 2    review.

 3    4.2  RISK ASSESSMENT BASED ON INFORMATION FROM
 4         EPIDEMIOLOGY STUDIES
 5          Risk estimates based on epidemiologic studies have the potential to provide perspective
 6    on the most serious pollutant-associated public health risks (e.g., hospital admissions, emergency
 7    department (ED) visits, premature mortality) in populations that often include at-risk groups.
 8    However, the amount of emphasis given to such quantitative risk estimates depends on the extent
 9    to which the underlying epidemiologic studies address key uncertainties, including the potential
10    for confounding by co-occurring pollutants. This section describes the epidemiology-based risk
11    assessment conducted in the 2008 REA (section 4.2.1) and staff s consideration of the relevant
12    evidence and information that is available in the current review (section 4.2.2).

13    4.2.1 Overview of the Assessment in the Last Review
14          In the last review of the primary NCh NAAQS, respiratory-related ED visits in the
15    Atlanta MSA were estimated as a function of short-term ambient NCh concentrations, based on
16    concentration-response relationships from an epidemiologic study by Tolbert et al. (2007) (U.S.
17    EPA, 2008a, Chapter 9).57 Specifically, the 2008 REA modeled respiratory-related ED visits
18    (including asthma, chronic obstructive pulmonary disease (COPD), upper respiratory illness,
19    pneumonia and bronchiolitis) for individuals of all ages based on a 3-day moving average of the
20    daily maximum 1-hour NCh concentrations measured at a single central-site monitor.58 The
21    selection of the Tolbert et al.  (2007) study as the basis for risk modeling reflected an emphasis on
22    (1) studies conducted within the U. S.; (2) studies of ambient NCh exposure (rather than indoor
23    exposure); (3) studies of respiratory-related ED visits or hospital admissions (given the clear
24    public health significance of this endpoint compared to symptoms and the degree of supporting
      57 As discussed above (section 1.2), the strongest evidence in the last review was for respiratory effects attributable to short-term
      NCh exposures. The study by Tolbert et al. (2007) was a key study supporting the relationship between NCh and respiratory
      effects (U.S. EPA, 2008b, Chapter 5).
      58 The monitor used in generating risk estimates (monitor id 131210048) matches that used in the Tolbert et al. (2007) study.
                                                 4-3

-------
 1    evidence presented in the 2008 ISA); and (4) studies that provided both single- and co-pollutant
 2    concentration-response functions.59
 3           The REA presented incidence estimates associated with NCh concentrations adjusted to
 4    just meet the annual NCh standard with its level of 53 ppb, NCh concentrations adjusted to just
 5    meet potential alternative 1-hour NCh standards with levels ranging from 50 to 200 ppb (98th and
 6    99th percentile forms),60 and unadjusted NCh concentrations. As discussed above (section 2.1),
 7    air quality adjustments were based on a proportional roll up of monitored NCh concentrations.

 8    4.2.1.1 Summary of the 2008 REA epidemiology-based risk assessment results
 9          The 2008 REA presented risk estimates based on single-pollutant models and co-pollutant
10    models (U.S. EPA, 2008a, Chapter 9). For adjusted air quality, the 2008 REA noted the
11    following:
12    •  When air quality was adjusted to simulate just meeting the existing annual standard, about 8
13       to 9% of respiratory-related ED visits in the Atlanta MSA were estimated to be attributable to
14       short-term NCh exposures, based on a single-pollutant model. Risk estimates based on co-
15       pollutant models remained positive, though they were smaller and confidence intervals were
16       wider than estimates  based  on the single pollutant model. Co-pollutant models that included
17       another roadway-associated pollutant (i.e., CO) resulted in modest reductions in NCh risk
18       estimates (i.e., about  7 to 8% of respiratory ED visits estimated to be associated with NCh).
19       The smallest risks were estimated with a co-pollutant model that also included PMio (i.e.,
20       about 3% of respiratory ED visits estimated to be associated with NCh) and with a multi-
21       pollutant model that included both PMio and Os (U.S.  EPA, 2008a, Tables 9-3 and 9-4).

22    •  When air quality was adjusted to simulate just meeting potential alternative standards with 1-
23       hour averaging times, standards with levels of 50, 100, and 150 ppb reduced estimated NCh-
24       associated risks compared to the annual standard alone. When air quality was adjusted to
25       simulate just meeting a potential alternative standard with a 1-hour averaging time and a
26       level of 200 ppb, estimated risks were similar to those estimated for the annual standard (U.S.
27       EPA, 2008a, Tables 9-3 and 9-4).
      59 The 2008 REA reflected the ISA conclusion that an important uncertainty in the NCh epidemiologic evidence is the extent to
      which NO2 is independently associated with [short-term] respiratory effects or if NCh is a marker for the effects of another
      traffic-related pollutant or mix of pollutants (U.S. EPA, 2008b, section 4.2.8, p. 37). This highlights the importance of including
      both single- and co-pollutant functions in modeling emergency department visits for NCh.

      60 These were the same air quality  scenarios evaluated for the air quality and exposure assessments (sections 2.1 and 3.1, above).


                                                  4-4

-------
 1    4.2.1.2 Uncertainties and limitations
 2          In the last review, the 2008 REA noted that a number of key uncertainties should be
 3    considered when interpreting these results with regard to decisions on the standard. These
 4    included the following (U.S. EPA, 2008a, section 9.6):
 5    •   Uncertainties in the estimates of NCh coefficients in concentration-response functions used in
 6       the assessment.
 7    •   Uncertainties concerning the specification of the concentration-response model (including
 8       the shape of the relationships) and whether or not a population threshold exists within the
 9       range of concentrations examined in the studies.
10    •   Uncertainty concerning possible confounding by co-occurring pollutants.
11    •   Uncertainty in the adjustment of air quality distributions to simulate just meeting various
12       standards. The REA recognized this as an important uncertainty, especially for scenarios
13       where considerable upward adjustment was required to simulate just meeting some of the
14       standards.
15    4.2.2 Consideration of Newly Available Information
16          The decision whether to conduct an updated epidemiology-based risk assessment in the
17    current review will be informed by staffs conclusions regarding the extent to which such an
18    updated assessment would reduce uncertainties and substantially improve our understanding of
19    NCh-attributable health risks, beyond the insights gained from the risk assessment conducted in
20    the last review. As discussed above (section 1.3), the strongest evidence in the current review,
21    based on the weight-of-evidence determinations in the 2nd draft ISA, is for respiratory effects
22    attributable to either short- or long-term NCh exposures (U.S. EPA, 2015, section 1.5.1). In this
23    section, staff considers the evidence that could inform an assessment of effects attributable to
24    short-term (4.2.2.1) or long-term (4.2.2.2) NCh exposures.

25    4.2.2.1 Health effects associated  with short-term exposures
26          For short-term NCh exposures, we first consider the overall strength of the scientific
27    evidence for NCh-attributable health endpoints and at-risk populations compared to the evidence
28    in the last review,  as presented in the 2nd draft ISA (U.S. EPA, 2015). As discussed above
29    (section 1.3), the 2nd draft ISA concludes that the evidence supports "a causal relationship
30    between short-term NCh exposure and respiratory effects" and that the "strongest evidence is for
31    effects on asthma exacerbation" (U.S. EPA, 2015, Table 1-1). These conclusions are based on a
32    number of epidemiologic, controlled human exposure, and animal toxicological studies, which
33    together describe a "coherent and biologically plausible pathway by which NCh exposure can
34    trigger an asthma exacerbation" (U.S. EPA, 2015, p. 1-17).
35          In  comparing the evidence for NCh-attributable respiratory effects available in the current
36    review to that available in the last review, the ISA notes the following (U.S. EPA, 2015, p. 1-19):
                                                4-5

-------
 1           Much of the evidence from epidemiologic and experimental studies was available
 2           in the 2008 ISA. However, compared to the 2008 ISA, this ISA more explicitly
 3           evaluates the coherence and biological plausibility for specific respiratory
 4           outcome groups. Rather than new evidence, the integration of epidemiologic and
 5           experimental evidence for asthma exacerbation—uptake of NCh in the respiratory
 6           tract and reactions to form reactive oxidation products, allergic inflammation,
 7           airway responsiveness, asthma symptoms, and hospital admissions and ED visits
 8           for asthma, associations with NCh measured in people's locations, which may
 9           better represent exposure, associations with adjustment for another traffic-related
10           pollutant—describes a coherent, biologically plausible pathway to support a
11           causal relationship between short-term NCh exposure and respiratory effects.
12    Thus, the change in the causality determination61 for respiratory effects attributable to short-term
13    NCh exposures is largely due to the evolution of the ISA's approach to assessing the evidence,
14    rather than the availability of substantially different evidence in the current review.
15           The evidence that has become available since the last review has not substantially
16    changed our understanding of health effects attributable to short-term NCh exposures or of the
17    populations potentially at increased risk of such effects.62 Updated risk estimates based on
18    information from epidemiology studies in the current review would be subject to the same
19    uncertainties identified in the 2008 REA.
20           In particular, recent studies do not provide an improved basis, compared to the last
21    review, for quantifying NCh-attributable risks independent of other roadway-associated
22    pollutants (e.g., carbon monoxide,  particulate matter, elemental carbon, and volatile organic
23    compounds). Table 1-1 of the 2nd draft ISA (U.S. EPA, 2015) concludes that an important
24    uncertainty in the current review continues to be the "[s]trength of inference from co-pollutant
25    models about independent associations of NCh, especially with pollutants measured at central
26    site monitors" (U.S. EPA, 2015). In particular, of the key studies supporting the causal
27    relationship with respiratory effects (U.S. EPA, 2015, Table 5-45), two U.S. studies evaluating
28    asthma-related hospital admissions or emergency department visits have become available since
      61 In the previous review, the ISA concluded that the evidence was "sufficient to infer a likely causal relationship between short-
      termNCh exposure and adverse effects on the respiratory system" (U.S. EPA, 2008b, section 5.3.2.1).
      62 Though the ISA's framework for identifying at-risk populations (U.S. EPA, 2015, section 1.6.5) was developed since the last
      review of the NO2 NAAQS, the 2008 ISA identified people with asthma, children, and older adults as populations potentially at
      increased risk for NCh-related effects (U.S. EPA, 2008b, section 4.3.1 and 4.3.2). As discussed above (section 1.3), in the current
      review the 2nd draft ISA concludes that "there is adequate evidence that people with asthma, children, and older adults are at
      increased risk for NO2-related health effects" (U.S. EPA, 2015, Table 7-26; section 1.6.5).
                                                   4-6

-------
 1    the last review (Strickland et al., 2010; Li et al., 2012). Neither of these studies reported NCh
 2    health effect associations in co-pollutant models that included other roadway-related pollutants.
 3          Based on the above considerations, an updated  epidemiology-based risk assessment
 4    estimating respiratory-related endpoints attributable to short-term NCh exposures would be
 5    subject to uncertainties that are essentially the same as those identified in the 2008 REA (U.S.
 6    EPA, 2008). We reach the preliminary conclusion that such an updated epidemiology-based risk
 7    assessment in the current review would not appreciably reduce uncertainties and limitations from
 8    the assessment conducting in the last review and would be unlikely to substantially improve our
 9    understanding of NCh-attributable health risks or increase our confidence in risk estimates
10    beyond the assessment from the last review.

11    4.2.2.2 Health effects associated with long-term exposures
12           As discussed above (section 1.3), the 2nd draft ISA concludes that the evidence "indicates
13    there is likely to be a causal relationship between long-term NCh exposure and respiratory
14    effects" (U.S. EPA, 2015, section 1.5.1, pp. 1-19 and 1-21) and the "strongest evidence is for
15    effects on asthma development" (U.S. EPA, 2015, Table 1-1). This contrasts with the conclusion
16    from the 2008 ISA that the evidence at that time was "suggestive but not sufficient to infer a
17    causal relationship" between long-term NCh exposures and respiratory effects. Key evidence
18    supporting the change to this causal determination comes from recent epidemiologic cohort
19    studies reporting associations between long-term ambient NCh concentrations (i.e., averaged
20    over 1-10 years) and development of asthma in children. There is some support for the
21    biological plausibility of effects attributable to long-term exposures provided by "a small body of
22    experimental studies" (U.S. EPA, 2015, Table 1-1).
23           As for short-term NCh exposures, an important issue in considering a potential
24    quantitative risk assessment  is the extent to which available epidemiologic studies report health
25    effect associations with long-term NCh in co-pollutant models, specifically for traffic-related
26    pollutants. This is an even more important issue for long-term NCh exposures, given the higher
27    correlations between long-term NCh concentrations and other pollutants reported in many
28    epidemiologic studies (U.S. EPA, 2015, Table 6-1).63
      63 In studies of NCh-associated asthma development, correlations with co-occurring pollutants were most often reported for PM.
      NO2 and PM were often highly correlated in these studies, particularly in studies that used land use regression or dispersion
      modeling to estimate long-term NCh exposures (e.g., rvalues were greater than 0.9 in several studies) (U.S. EPA, 2015, Table 6-
      1).
                                                  4-7

-------
 1          Of the key studies evaluating associations between long-term NCh and the development
 2   of asthma (U.S. EPA, 2015, Table 6-5), none evaluated associations in co-pollutant models for
 3   traffic-related pollutants. Table 6-5 of the 2nd draft ISA notes that an important remaining
 4   uncertainty is that "regarding potential confounding by traffic-related co-pollutants" (U.S. EPA,
 5   2015). In particular, the 2nd draft ISA notes that "[w]hen reported, correlations with PM2.5 and
 6   EC often were high (r = 0.7-0.96)" and that "[n]o co-pollutant models [were] analyzed" (U.S.
 7   EPA, 2015, Table 6-5).
 8          Although there are new epidemiologic studies available in the current review supporting
 9   a relationship between long-term NCh exposures and development of asthma in children, a
10   quantitative risk assessment based on information from such studies would be subject to
11   considerable  uncertainty due to the inability to distinguish the contributions of NCh from the
12   contributions of other highly correlated pollutants. Given these limitations, we reach the
13   preliminary conclusion that such a risk assessment would not substantially add to our
14   understanding of NCh-attributable health risks and would therefore be of limited value in
15   informing decisions in the current review.
                                                4-8

-------
 i    5   SUMMARY OF CONCLUSIONS AND NEXT STEPS
 2          The preliminary conclusions presented in chapters 2 through 4 of this planning document
 3    reflect the EPA staffs preliminary assessment of the degree to which updated quantitative
 4    analyses in the current review of the primary NCh NAAQS are likely to substantially add to our
 5    understanding of NCh exposures or health risks. In developing these preliminary conclusions,
 6    staff considered a variety of factors (Figure 1-3) including the availability of appropriate health
 7    evidence; the availability of technical information, tools, and methods; and judgments as to the
 8    potential for particular quantitative assessments to provide important insights into exposures or
 9    health risks, beyond the insights gained from previous assessments. This chapter summarizes
10    staffs preliminary conclusions (section 5.1) and discusses the next steps in the review of the
11    primary NCh NAAQS (section 5.2).

12    5.1  SUMMARY OF PRELIMINARY CONCLUSIONS
13          Summaries of staffs preliminary conclusions are as follows:
14       •  Air quality comparison to health benchmarks: New information from the NCh monitoring
15          network and from available research studies has the potential to substantially improve our
16          understanding of NCh concentrations around major roads in the current review. Staff
17          concludes that updated analyses comparing ambient NCh concentrations (i.e., as
18          surrogates for potential exposure concentrations) to health effect benchmarks would
19          better characterize a key uncertainty from the last review (i.e., uncertainty in ambient
20          NCh concentrations on- or near-roads).

21       •  Exposure assessment: While modeling tools have been updated since the last review,
22          staff reaches the preliminary conclusion that an updated exposure assessment would be
23          warranted only if the air quality assessment discussed in Chapter 2 indicates the potential
24          for NCh exposures that could be of public health concern (i.e., based on comparisons of
25          ambient Mh concentrations with health effect benchmarks). If the air quality assessment
26          indicates little potential for such exposures, including on or near major roads, staff
27          reaches the preliminary conclusion that an updated assessment of population exposures
28          would  be of limited use in informing decisions in the current review.

29       •  Risk assessment based on information from controlled human exposure studies:  Based on
30          the  evidence assessed in the 2nd draft ISA, staff reaches the preliminary conclusion that
31          available studies do not provide information to support the identification  of an NCh
32          exposure-response relationship with relevant health endpoints and at relevant NCh
33          concentrations. Therefore, as in the last review, staff reaches the preliminary conclusion
34          that the available evidence in the current review is not sufficient to support a risk
35          assessment based on exposure-response information from controlled human exposure
36          studies.

37       •  Risk assessment based on information from epidemiologic studies of health effects
38          associated with  short-term NCh exposure: Evidence that has become available since the


                                               5-1

-------
 1          last review has not substantially changed our understanding of the health effects
 2          attributable to short-term NCh exposures or our understanding of the populations at
 3          increased risk from such exposures. Recent U.S. epidemiology studies of asthma-related
 4          hospital admissions or emergency department visits also have not provided information
 5          on NO2 effects, independent of other traffic-related pollutants. Therefore, an updated
 6          epidemiology-based risk assessment estimating respiratory-related endpoints attributable
 7          to short-term NCh exposures would be subject to uncertainties that are essentially the
 8          same as those identified in the 2008 REA. We reach the preliminary conclusion that such
 9          an updated epidemiology-based risk assessment in the current review would be unlikely
10          to substantially improve our understanding of NCh-attributable health risks or increase
11          our confidence in risk estimates, beyond the assessment from the last review.
12       •  Risk assessment based on information from epidemiologic studies of health effects
13          associated with long-term NCh exposure: Key U.S. epidemiology studies of long-term
14          NCh and asthma incidence do not present analyses with co-pollutant models that include
15          highly correlated traffic-related pollutants.  A risk assessment quantifying the
16          development of asthma attributable to long-term NCh exposures would be subject to
17          considerable uncertainty due to the inability to distinguish the contributions of NCh from
18          the contributions of other pollutants. Therefore, we reach the preliminary conclusion that
19          such a risk assessment would be of limited value in informing decisions in the current
20          review.
21    5.2  NEXT STEPS
22         Given the preliminary conclusions summarized above, the next step in the current review
23    will be for staff to conduct the analyses comparing NCh air quality concentrations to health effect
24    benchmarks, as described in Chapter 2 of this planning document. Based on the results of these
25    analyses, staff will consider the potential for populations in U.S. urban areas just meeting the
26    existing  primary standards to  experience NCh exposures that may be of concern  in people with
27    asthma. In doing so, staffs considerations will focus on the frequency with which ambient NCh
28    concentrations in the U.S. could be at or above various health effect benchmarks.
29         The results of these analyses will  inform the subsequent steps appropriate for the current
30    review. If results indicate that the existing standards allow the potential for exposures to NCh
31    concentrations that may be of concern for public health, more refined quantitative analyses of
32    personal NCh exposures will be considered. If results indicate limited potential for such NCh
33    exposures, there will be less support for a more refined quantitative assessment of personal NCh
34    exposures.
35         To the extent analyses are limited to comparisons between ambient NCh concentrations
36    and health effect benchmarks, we anticipate that these analyses, as well as the evidence and
37    rationale supporting a conclusion not to conduct more refined analyses of personal NCh
38    exposures, will be incorporated into the first draft of the Policy Assessment (PA). Under this
39    potential scenario, a separate REA will  not be generated and CAS AC will review the analyses
40    comparing ambient NCh concentrations to health effect benchmarks, including staffs

                                                5-2

-------
1   interpretation of the results of these analyses, as part of its review of the PA. Based on this
2   potential scenario, Table 5-1 presents a tentative schedule for the documents remaining to be
3   produced for this  review. To the extent the tentative schedule presented in Table 5-1 requires
4   modification (e.g., to accommodate an exposure assessment), a revised schedule will be
5   communicated to CASAC and the public.
6   Table 5-1. Tentative schedule for next steps in the review of the primary NOi NAAQS.
Stage of Review
Integrated Review Plan (IRP)
Integrated Science Assessment (ISA)
Risk/Exposure Assessment (REA)
Policy Assessment (PA) including
quantitative analyses comparing
ambient NO2 concentrations to health
effect benchmarks
Major Milestone
Final IRP
1st draft ISA
CASAC public meeting for review of
the 1st draft ISA
2nd draft ISA
CASAC review of the 2nd draft ISA
Final ISA
REA Planning Document
CASAC review of REA Planning
Document
1st draft PA
Target Date
June 2014
November 201 3
March 12-13,2014
January 2015
June 2-3, 2015
Fall 201 5
May 4, 2015
June 2-3, 2015
Spring/Summer 2016
                                               5-3

-------
6   REFERENCES
Ahmed T., Dougherty R., Sackner M.A. (1983a). Effect of 0.1 ppm NO2 on pulmonary functions and non-specific
        bronchial reactivity of normals and asthmatics [final report]. (CR-83/11/BI). Warren, MI: General Motors
        Research Laboratories.
Ahmed T., Dougherty R., Sackner M.A. (1983b). Effect of NO2 exposure on specific bronchial reactivity in subjects
        with allergic bronchial asthma [final report]. (CR-83/07/BI). Warren, MI: General Motors Research
        Laboratories.
Baldauf R., Isakov V., Deshmukh P., and Venkatram A. (2015). Influence of solid noise barriers on near-road and
        on-road air quality. In preparation.
Barck C., Sandstrom T., Lundahl I, Hallden Gl, Svartengren M., Strand V., Rak S., Bylin, G. (2002). Ambient level
        of NO2 augments the inflammatory response to inhaled allergen in asthmatics. Respir Med 96: 907-917.
Batterman S., Burke I, Isakov V., Lewis T., Mukherjee B., and T. Robins. (2014). A comparison of exposure
        metrics for traffic-related air pollutants: application to epidemiology studies in Detroit, Michigan. Int J
        Environ Res Public Health.  11(9), 9553-9577.
Bell S. and T. W. Ashenden. (1997). Spatial and temporal variation in nitrogen dioxide pollution adjacent to rural
        roads. Water Air Soil Pollut. 95:87-98.
Bennett D.H, Fisk W., Apte M.G., Wu X., Trout A., Faulkner D., and D. Sulivan. (2011).  Ventilation, Temperature,
        and HVAC Characteristics in Small and Medium Commercial Buildings (SMCBs) in California.  Journal
        submission.
Bylin G. (1993). Health risk evaluation of nitrogen oxide: Controlled studies on humans. ScandJ Work Environ
        Health 19: 37-43.
Bylin G., Hedenstierna G., Lindvall T., Sundin B. (1988). Ambient nitrogen dioxide concentrations increase
        bronchial responsiveness in subjects with mild asthma. Eur Respir J I: 606-612.
Cape J.N., Tang Y.S., van Dijk N., Love L., Sutton M. A., and  S.C.F. Palmer. (2004). Concentrations of ammonia
        and nitrogen dioxide at roadside verges, and their contribution to nitrogen deposition. Environ Pollut.
        132:469-478.
Frey H.C. (2014). Letter from Dr. H. Christopher Frey, Chair,  Clean Air Scientific Advisory Committee to EPA
        Administrator Gina McCarthy. CASAC Review of the EPA's Integrated Review Plan for the Primary
        National Ambient Air Quality Standards for Nitrogen Dioxide (External Review  Draft). EPA-CASAC-14-
        001. May 10, 2014. Available at:
        http://vosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/89989229944F36B0852
        57CF300692E2A/$File/EPA-CASAC-14-001+unsigned.pdf.
Gilbert N.L., Woodhouse S., Stieb, D.M., Brook J.R. (2003). Ambient nitrogen dioxide and distance from a major
        highway. Sci Total Environ. 312(l-3):43-46.
                                                 6-1

-------
Hazucha, MJ; Ginsberg, JF; McDonnell, WF; Haak, ED, Jr; Pimmel, RL; Salaam, SA; House, DE; Bromberg, PA.
        (1983). Effects of 0.1 ppm nitrogen dioxide on airways of normal and asthmatic subjects. J Appl Physiol
        54: 730-739.
Henderson R. (2008). Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee to EPA
        Administrator Stephen Johnson. EPA-CASAC-08-015. Clean Air Scientific Advisory Committee's
        (CASAC) Peer Review of EPA's Integrated Science Assessment (ISA) for Oxides of Nitrogen - Health
        Criteria (Second External Review Draft). June 25, 2008. Available at:
        http://vosemite.epa. gov/sab/sabproduct.nsf/89BD55F21 Al AOAF88525747400685146/$File/EPA-CASAC-
        08-015-unsigned.pdf.
Jorres R., Magnussen H. (1990). Airways response of asthmatics after a 30 min exposure, at resting ventilation, to
        0.25 ppmNO2 or 0.5 ppm SO2. Eur Respir J 3: 132-137.
Karner A.A., Eisinger D.S., and D.A. Niemeier. (2010). Near-roadway air quality: Synthesizing the findings from
        real-world data. Environ Sci Technol. 44: 5334-5344. http://dx.doi.org/10.1021/esl00008x
Kimbrough E.S., Bauldauf R.W., and N. Watkins. (2013). Seasonal and diurnal analysis of NCh concentrations from
        a long-duration study conducted in Las Vegas, Nevada. JAWMA. 63(8)934-942.
Mohsenin V. (1987a). Airway responses to nitrogen dioxide in asthmatic subjects. J Toxicol Environ Health 22:
        371-380.
Monn Ch., Carabias V., Junker M, Waeber R., Karrer M, and H.U Wanner. (1997). Small-scale spatial variability
        of paniculate matter <10 um (PM10) and nitrogen dioxide. Atmos Environ. 31(15)2243-2247.
Mukerjee S., Smith L.,  Brantley H., Stallings C., Neas L., Kimbrough S., and R. Williams. (2015). Comparison of
        modeled traffic exposure zones using on-road air pollution measurements. Atmospheric Pollution Research.
        6:82-87.
Orehek J., Massari J.P., Gayrard P., Grimaud C., Charpin J.  (1976). Effect of short-term, low-level nitrogen dioxide
        exposure on bronchial sensitivity of asthmatic patients. J Clin Invest 51': 301-307.
Rizzo M. (2008). Investigation of how distributions of hourly nitrogen dioxide concentrations have changed over
        time in six cities. Nitrogen Dioxide NAAQS Review Docket (EPA-HQ-OAR-2006-0922). Available at:
        http://www.epa.gOv/ttn/naaqs/standards/nox/s noxcrrea.html.
Samet J. (2008a). Letter from Dr. Jonathan M. Samet, Chair, Clean Air Scientific Advisory Committee to EPA
        Administrator Stephen Johnson. Clean Air Scientific Advisory Committee's (CASAC) Peer Review of
        Draft Chapter 8 of EPA's Risk and Exposure Assessment to Support the Review of the NCh Primary
        National Ambient Air Quality Standard. EPA-CASAC-09-001. October 28, 2008. Available at:
        http://vosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/87D38275673D66B885
        2574F00069D45E/$File/EP A-CASAC-09-001-unsigned.pdf.
                                                 6-2

-------
Samet J. (2008b). Letter from Dr. Jonathan M. Samet, Chair, Clean Air Scientific Advisory Committee to EPA
        Administrator Stephen Johnson. Clean Air Scientific Advisory Committee's (CASAC) Review comments
        on EPA's Risk and Exposure Assessment to Support the Review of the NO2 Primary National Ambient Air
        Quality Standard. EPA-CASAC-09-003. December 16, 2008. Available at:
        http://vosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/9C4A540D86BFB67A8
        52575210074A7AE/$File/EP A-CASAC-09-003-unsigned.pdf.
Samet J. (2009). Letter from Dr. Jonathan M. Samet,  Chair, Clean Air Scientific Advisory Committee to EPA
        Administrator Lisa P. Jackson. Comments and Recommendations Concerning EPA's Proposed Rule for the
        Revision of the National Ambient Air Quality Standards (NAAQS) for Nitrogen Dioxide. EPA-CASAC-
        09-014. September 9, 2009. Available at:
        http://vosemite.epa.gov/sab/sabproduct.nsf/264cbl227d55e02c85257402007446a4/0067573718EDA17F85
        25762C0074059E/$File/EP A-CASAC-09-014-unsigned.pdf.
Strand V., Rak S., Svartengren M., Bylin G. (1997). Nitrogen dioxide exposure enhances asthmatic reaction to
        inhaled allergen in subjects with asthma. Am JRespir Crit Care Med 155: 881-887.
Strand V., Svartengren M., Rak S., Barck C., Bylin G. (1998). Repeated exposure to an ambient level of NO2
        enhances asthmatic response  to a nonsymptomatic allergen dose. Eur Respir J12: 6-12.
Thurman J., Bailey C., Watkins N., Baldauf R., and R. Erode. (2013). Use of AERMOD for NO2 Near-road
        Monitoring Implementation, presented at A&WMA Guideline on Air Quality Models: The Path Forward.
        March 19, 2013, Control # 44.
Tolbert P.E., Klein M., Peel J.L., Sarnat S.E., Sarnat J.A.. (2007). Multipollutant modeling issues in a study of
        ambient  air quality and emergency department visits in Atlanta.  J Expos Sci Environ Epidemiol.
        17(S2):S29-35.
Tunnicliffe W.S., BurgeP.S., Ayres J.G. (1994). Effect of domestic concentrations of nitrogen dioxide on airway
        responses to inhaled allergen in asthmatic patients. Lancet 344:  1733-1736.
U.S. EPA. (2008a). Risk and Exposure Assessment to Support the Review of the NO2 Primary National Ambient
        Air Quality Standard. U.S. EPA, Office of Air Quality Planning and Standards. Research Triangle Park,
        NC. EPA 452/R-08-008a/b. November 2008. Available at:
        http://www.epa.gOv/ttn/naaqs/standards/nox/s  noxcrrea.html.
U.S. EPA. (2008b). Integrated Science Assessment for Oxides of Nitrogen - Health Criteria. U.S. EPA, National
        Center for Environmental Assessment and Office, Research Triangle Park, NC. EPA/600/R-08/071. July
        2008. Available at: http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid= 194645.
U.S. EPA. (2009). Risk and Exposure  Assessment to  Support the Review of the SO2 Primary National Ambient Air
        Quality Standard. . U.S. EPA, Office of Air Quality Planning and Standards. Research Triangle Park, NC.
        July 2009. EPA- EPA-452/R-09-007. Available at:
        http://www.epa.gov/ttn/naaqs/standards/so2/data/200908SO2REAFinalReport.pdf.
                                                 6-3

-------
U.S. EPA. (2010). Quantitative Health Risk Assessment for Paniculate Matter. U.S. EPA, Office of Air Quality
        Planning and Standards. Research Triangle Park, NC. EPA-452/R-10-005. June 2010. Available at:
        http://www.epa.gov/ttn/naaqs/standards/pm/data/PM RA FINAL June  2010.pdf.
U.S. EPA. (2014a). Integrated Review Plan for the Primary National Ambient Air Quality Standards for Nitrogen
        Dioxide. U.S. EPA, National Center for Environmental Assessment and Office of Air Quality Planning and
        Standards, Research Triangle Park, NC. EPA-452/R-14-003. June 2014.  Available at:
        http://www.epa.gov/ttn/naaqs/standards/nox/data/201406finalirpprimaryno2.pdf.
U.S. EPA. (2014b). Health Risk and Exposure Assessment for Ozone, Final Report. U.S. EPA, Office of Air Quality
        Planning and Standards. Research Triangle Park, NC. EPA-452/R-14-004a. August 2014. Available at:
        http://www.epa.gov/ttn/naaqs/standards/ozone/data/20140829healthrea.pdf.
U.S. EPA. (2015). Integrated Science Assessment for Oxides of Nitrogen - Health Criteria (Second External Review
        Draft). U.S. EPA, National Center for Environmental Assessment and Office, Research Triangle Park.
        EPA/600/R-14/006. January 2015. Available at:
        http://cfpub.epa.gov/ncea/isa/recordisplav.cfm?deid=288043.
Williams R., Rea A., Vette A., Croghan C., WhitakerD., Stevens C., McDow S., FortmannR., Sheldon L., Wilson
        H., Thornburg J., Phillips M., Lawless P., Rodes C., and H. Daughtrey H. (2008). The design and field
        implementation of the Detroit Exposure and Aerosol Research Study. J Expos Sci Environ Epidem. 19:
        643-659.
                                                  6-4

-------
United States                      Office of Air Quality Planning and Standards     Publication No. EPA-452/D-15-001
Environmental Protection            Health and Environmental Impacts Division                            May 2015
Agency                           Research Triangle Park, North Carolina 27711

-------
 i   APPENDIX A: ANALYSIS OF LAS VEGAS NEAR-ROAD
 2   NO2 MEASUREMENT DATA AND LOGIT MODEL
 3   DEVELOPED TO SIMULATE NO2 ON-ROAD
 4   CONCENTRATIONS
 5          This appendix provides the memorandum from Jennifer Richmond-Bryant to Stephen Graham
 6   regarding an approach to use in simulating on-road NC>2 concentrations from near-road NC>2
 7   concentrations.

 8                                     Table of Contents
 9   A-l.    Overview	A-l
10   A-2.    Methods	A-l
11   A-3.    Results and Discussion	A-4
12   A-4.    References	A-7
13

14                                       List of Tables

15   Table A-l.   Concentration vs. distance-from-road model formulations	A-4
16   Table A-2.   Summary statistics for observed and predicted concentrations for all wind and stability
17              conditions combined	A-8
18   Table A-3.   Summary statistics for observed and predicted concentrations, westerly winds (210°-330°).
19              	A-8
20   Table A-4.   Summary statistics for observed and predicted concentrations, easterly winds (30°-150°)...
21              	A-8
22   Table A-5.   Summary statistics for observed and predicted concentrations for inversion conditions
23              (convective mixing height less than 300 m)	A-9
24   Table A-6.   Summary statistics for observed and predicted concentrations for non-inversion conditions
25              (convective mixing height greater than 300 m)	A-9
26   Table A-7.   Summary statistics for percent change in NC>2 concentration from modeled on-road to
27              concentrations at distances of 10 m and 20 m away from roads for all atmospheric
28              conditions combined, by quintile of NC>2 predicted concentrations at 10 m and 20 m away
29              from the roads	A-9
30   Table A-8.   Summary statistics for percent change in NC>2 concentration from modeled on-road to
31              concentrations at distances of 10 m and 20 m away from roads for winds from the west
32              (210°-330°), by quintile of NCh predicted concentrations at 10 m and 20 m away from
33              roads	A-10
34   Table A-9.   Summary statistics for percent change in NC>2 concentration from modeled on-road to
35              concentrations at distances of 10 m and 20 m away from roads for winds from the east (30°-
36              150°), by quintile of NC>2 predicted concentrations at 10 m and 20 m away from roads	
37              	A-10

                                             A-i

-------
 1    Table A-10.  Summary statistics for percent change in NC>2 concentration from modeled on-road to
 2                 concentrations at distances of 10 m and 20 m away from roads for inversion conditions
 3                 (mixing height < 300 m), by quintile of NC>2 predicted concentrations at 10  m and 20 m
 4                 away from roads	A-10
 5    Table A-l 1.  Summary statistics for percent change in NC>2 concentration from modeled on-road to
 6                 concentrations at distances of 10 m and 20 m away from roads for non-inversion conditions
 7                 (mixing height > 300 m), by quintile of NC>2 predicted concentrations at 10  m and 20 m
 8                 away from roads	A-ll
 9

10                                              List of Figures

11    Figure A-l.  The Las Vegas study area. Interstate-15 runs north-south, and the monitoring sites follow
12                 along a northwest-to-southeast transect along a road crossing	A-2
13    Figure A-2.  Predicted and observed NO2 concentrations for all wind and stability conditions combined.
14                 Predicted median (solid), predicted 98th and 2nd percentile (dotted), observed median
15                 (circles), and observed 98th and 2ndpercentiles (errorbars) are shown	A-ll
16    Figure A-3.  Predicted and observed NO2 concentrations for winds from the west. Predicted median
17                 (solid), predicted 98th and 2nd percentile (dotted), observed median (circles), and observed
18                 98th and 2nd percentiles (error bars) are shown	A-12
19    Figure A-4.  Predicted and observed NO2 concentrations for winds from the east. Predicted median
20                 (solid), predicted 98th and 2nd percentile (dotted), observed median (circles), and observed
21                 98th and 2nd percentiles (error bars) are shown	A-13
22    Figure A-5.  Predicted and observed NC>2 concentrations for inversion conditions (convective mixing
23                 height less than 300 m).  Predicted median (solid), predicted 98th and 2nd percentile (dotted),
24                 observed median (circles), and observed 98th and 2nd percentiles  (error bars) are shown....
25                 	A-14
26    Figure A-6.  Predicted and observed NO2 concentrations for non-inversion conditions (convective
27                 mixing height greater than 300 m). Predicted median (solid), predicted 98th  and 2nd
28                 percentile (dotted), observed median (circles), and observed 98th and 2nd percentiles (error
29                 bars) are shown	A-15
30
                                                     A-ii

-------
 i
 2
 3
 4
 5
I
5
V
           JPIl
                     uj
                     0
              p*
                     United States Environmental Protection Agency
                       National Center for Environmental Assessment
                        Research Triangle Park, North Carolina 27711
                                MEMORANDUM
      To:         Stephen Graham - Physical Scientist, Risk and Benefits Group, OAQPS
      From:      Jennifer Richmond-Bryant - Senior Physical Scientist, National Center
                  for Environmental Assessment, ORD
      Date:       May 4, 2015
      Subject:    Analysis of Las Vegas Near-Road N02 Measurement Data and Logit
                  Model Developed Simulate N02 On-Road Concentrations.
 8
 9
10
11
12
13
14

15

16
17

18
19
20
21
22
A-1.  Overview
This memo summarizes analyses performed on recently collected near-road measurement data in Las
Vegas, NV. Several statistical models were evaluated to describe the pattern in the concentration
reduction observed with increasing distance from a major road. Based on model fits (R2) and overall
form, a logit function was determined most appropriate. Proposed is an approach to use in estimating
on-road NO2 concentrations when having NO2 concentration measurements at a distance from a road,
particularly at locations in close proximity to a road (i.e., 10 m and 20 m).
A-2.  Methods
Study Area
Near-road measurements of air quality, traffic, and meteorology were obtained at sites in Las Vegas, NV.
The near road monitoring study area was chosen where 1) the Annual Average Daily Traffic (AADT)
exceeded 150,000 vehicles per day, 2) airflow downwind of the highway was not restricted by natural or
human-made structures, and 3) state and local governments permitted sampling sites within 300 meters
(m) of the road to be established. The study area and sampling sites are shown in Figure A-1.
                                            A-1

-------
                                !1/
                          Las Vegas Monitoring Sit
                               •II
   ';
   r 100 Meter
    Upwind Site (4i
—^
                                    -  7^20 Meter
                                      V- Road Sid* Sit* (1)

                                           ~ 100 Meter
                                             Downwind Site (2) '
                                                            300 Meter
                                                            Downwind Site (3)
             -Jr. ,  «
 1
 2    Figure A-l. The Las Vegas study area, lnterstate-15 runs north-south, and the monitoring sites follow
 3    along a northwest-to-southeast transect along a road crossing.

 4    The Las Vegas study area was located adjacent to lnterstate-15 (1-15). Along this segment of the road,
 5    AADT is approximately 206,000 vehicles per day, with 10% of those characterized as heavy-duty diesel
 6    trucks (Nevada DOT, 2006). At this location, 1-15 runs in the north-south direction, and the highway sits
 7    below grade with walls sloping upwards from the road at 20° angles. The terrain above the embankment
 8    is flat within a 10 km radius of the road. Downwind sampling sites were located approximately 20 m,
 9    100 m, and 300 m east of the highway, and an upwind site was placed approximately 100 m west of the
10    road. Meteorology at this study area is generally characterized as arid, with hot summers and sunshine
11    throughout the year. With  mountain ranges surrounding the Las Vegas metropolitan area, the area is
12    subject to atmospheric inversions. A detailed description of this study area is provided in Kimbrough et
13    al. (2013).

14    Data Collection

15    Nitric oxide (NO) and total  oxides of nitrogen (NOX) were monitored continuously by chemiluminescence
16    with a trace oxides of nitrogen analyzer (Ecotech, Model EC 9841 B, Knoxfield, VIC, Australia) with
17    measurements averaged every five minutes, and NO2 was estimated via differencing. Multipoint
18    calibration was performed  at the beginning of the study, and zero and span checks were performed
                                                  A-2

-------
 1    nightly for each of the gaseous monitors. Inlets for each of these monitors were placed approximately 3
 2    m above ground, and air pollutant concentrations were measured at the 20 m, 100 m, and 300 m
 3    downwind sites.

 4    Surface meteorological parameters monitored included wind speed, wind direction, air temperature,
 5    relative humidity, precipitation, and solar radiation. All meteorological parameters were measured at
 6    the Las Vegas airport (LAS), as part of the standard meteorological measurements made by the National
 7    Weather Service (NWS) at most major airports. The LAS meteorological station, which is approximately
 8    1.5 km from the Las Vegas near-road site, is part of the Automated Surface Observing Systems (ASOS)
 9    and has a 1-minute temporal resolution for wind speed and direction. Upper air data from the Universal
10    Rawinsonde Observation (RAOB) station in Mercury/Desert Rock, NV (KDRA, elevation 1006 m) was
11    used  as the primary upper air station. 1-Minute  ASOS wind data were processed for input to AERMET
12    using AERMINUTE for the period 12/1/2008 through 2/28/2010.  Upper air and surface data were
13    processed through AERMET to  obtain hourly averages of surface and upper air meteorological
14    parameters.

15    Data  were collected at the Las Vegas site between December 12, 2008 and January 21, 2010. There were
16    8,466 complete hours (81.8%) of data (based on the presence of NO2 data at all three downwind sites)
17    available for that time period and used in the analyses presented here.

18    On-Road Concentration Estimation

19    In order to estimate the on-road concentration of NO2 at each hour where  measurement data were
20    complete, a statistical distribution was fit to the concentrations measured at the three downwind
21    monitoring locations. All models were of the basic form:

22                f(C) = m*g(x) + b                                           (equation 1)

23    where f(C) = a statistical distribution fit to the concentration data across the three monitoring sites, g(x)
24    = a statistical distribution fit to  the location of the monitors across the three sites, m = the model
25    estimated slope, and b = the model estimated intercept. The slope for any given assumption is the first
26    derivative of the concentration function with respect to space, m = df(C)/dg(x). Linear, In-ln, In-linear,
27    and logit-ln distributions were fit to the  concentration data across the three monitoring sites. Table A-l
28    presents the forms of f(C) and g(x) for different assumptions about the model distribution. The
29    dependent and independent terms increase  in complexity. A lognormal distribution was fit to the
30    distance data in some of the models to linearize that term, since  the monitor sites were not evenly
31    spaced relative to one another. Likewise, a lognormal distribution was fit to the concentrations in some
32    of the models to linearize that term for the same reason. If a In-ln model were used, then the model
33    could not be solved for an on-road concentration (at a distance of x = 0), because the declining on-road
34    concentration would produce a negative slope, leading to  a solution of infinite concentration when
35    integrating the derivative to solve for f(C) at x =  0. A  logit function was fit to the concentrations for the
36    logit-ln model to test if the concentration distribution approximated an S-shaped curve.  Use of the logit-
37    In model was most physically sensible, because turbulent mixing related to traffic in some instances
38    could cause concentration levels to plateau near the road  and then gradually drop off in a manner

                                                  A-3

-------
 1    similar to a Gaussian distribution (centerline to lateral extrema). Note that the reference concentration
 2    was C(300), because the reference point had to be along the data distribution for the functional fit to
 3    apply. The reference was considered a point downstream where concentration returned to background
 4    levels. Using the 100 m upstream site as the reference point would have produced a mathematical
 5    instability.

 6    Table A-l. Concentration vs. distance-from-road model formulations.
MODEL FORM
linear
In-linear
In-ln
logit-ln
f(C)
c
ln(C)
ln(C)
gC(xy[eC(x)+eC(ref)]
g(x)
X
X
MX)
MX)
C(x)
b
exp(b)*exp(m*x)
exp(b)*xm
C(ref)+ln{b+mln(x)}-ln{l-[b+mln(x)]}
 8
 9
10
11
12
13

14

15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

30
31
When evaluating the performance of the selected models, the logit-ln formulation had the highest
median R2 of all model types (R2 = 94.7% compared with 84.6%, 93.7%, and 88.2% for the linear, In-ln,
and In-linear models, respectively). The median, average, and percentile statistics were calculated for R2
across the data, since there were 8,466 curve fits corresponding to each complete time period. These
results support use of the logit-ln model. In conjunction with the physical rationale described above, the
logit-ln model was used to estimate on-road concentrations.
A-3.   Results and Discussion
The on-road NO2 concentrations were estimated for five scenarios: 1) all wind and stability conditions
combined, 2) winds from the west (210°-330°, where the monitors were downwind of the highway), 3)
winds from the east (30°-150°, where the monitors were upwind of the highway), 4) inversion
conditions (convective mixing height less than 300 m), and 5) non-inversion conditions (convective
mixing height greater than 300 m). Summary statistics for each scenario are provided in Table A-2
through Table A-6, and predicted and observed data distributions are displayed in Figure A-2 through
Figure A-6.  In addition, limited results for NOX concentrations are included in the text for added context
and discussion clarity. For each scenario, estimated NO2 concentrations derived from the logit-ln model
are presented for the on-road, 10 m, 20 m, 30 m, 40 m, 50 m, 100 m, and 300 m sites. Observation data
are also presented so that estimates for the 20 m, 100 m, and 300 m  sites could be compared with
observations for validation. Model estimates were within 1% of observations. Good agreement between
observations and predictions is also indicated when examining the figures for each scenario, because
the median and range of the observations (given by the second to ninety-eighth percentile of the data)
coincide with the median and range of the model predictions.

For all study area conditions combined, the average and maximum on-road NO2 concentrations were 31
ppb and 104 ppb, respectively (Table A-2). Despite the use of the logit function to represent the
                                                   A-4

-------
 1    concentration variable, Figure A-2 (as well as the other scenarios, Figure A-3 through Figure A-6)
 2    illustrates that the concentration plateau with decreasing distance to the road is very limited in size.
 3    Overall, the concentration trend with increasing distance from the roads is consistent with
 4    exponential/logarithmic decay functions described in other similar measurement studies (section
 5    2.5.3.1, US EPA, 2015). NO2 concentrations declined from on-road to the 10 m site by 13%, on average,
 6    and from on-road to the 20 m site by an average of 17% for (Table A-7). A reduction in NOX at the 20 m
 7    site on average was 25% (not shown). When stratifying the percent change in concentration by quintile
 8    of concentration at the 10 m or 20 m location, the highest average difference between on-road and the
 9    10 m or 20 m site is 19% and 25%, respectively (Table A-7), and occurs in the second quintile (i.e. at low
10    NO2 concentrations). Some predictions in the first quintile are negative at the 10 m site or produce
11    higher values at the 10 m  or 20 m sites  compared with the on-road value (perhaps driven by inclusion of
12    concentration data for where predominant winds were from the east). This effectively reduced the
13    average difference between on-road and away from road concentrations considering the lowest
14    concentration quintile. At higher NO2 concentrations (i.e., quintile 4 and 5), the percent change
15    decreased to 10-11% for the 10 m site and 11-12% for the 20 m site, on average (Table A-7).

16    When segregating the study area data set by wind direction, on-road concentration estimates were
17    generally higher for winds from the west (Table A-3, average NO2: 36 ppb, max NO2: 74 ppb, average
18    NOX: 86 ppb, max NOX: 264 ppb) compared with concentrations when winds were from the  east (Table
19    A-4, average NO2: 23 ppb, max NO2: 75 ppb, average NOX: 38 ppb, max NOX: 215 ppb). The estimated
20    near-road gradient was slightly sharper for winds from the west, with an average on-road to 20 m
21    difference of 17% for NO2 (Table A-8) and 28% for NOX (not shown) for winds from the west and an
22    average difference of 13% for NO2 (Table A-9) and 14% for NOX (not shown) for winds from the east.
23    Interestingly, the average estimated NOX gradient (not shown) was twice as large for winds from the
24    west compared with winds from the east, but the difference in gradient was much smaller for NO2. In
25    general, the NO gradient tends to be sharper than the NO2 gradient (Karner et al., 2010). This finding
26    suggests that when the winds from the west disperse the NO away from the roadway, photochemistry
27    occurs over shorter time and length scales such that the NO disappears faster than the NO2  compared
28    with the case where the monitors are on the upwind side of the road. At lower NO2 concentrations, the
29    difference between on-road and the 10 m or 20 m sites was much larger when winds were from the
30    west compared with the difference for  winds from the east. For example, at 10 m, the difference was
31    10% when winds were from the east and 25% when winds were from the west (see Tables A-8 and A-9).
32    For the three highest concentration quintiles the percent difference was comparable regardless of wind
33    direction (8-10% at the 10 m site, 10-12% at the 20 m site).

34    When segregating the study area data set by stability conditions, concentrations were on average higher
35    during an inversion (Table A-5, average NO2: 38 ppb, max NO2: 79 ppb, average NOX:  86 ppb, max NOX:
36    264 ppb) compared with non-inversion conditions (Table A-6, average NO2: 23 ppb, max NO2: 82 ppb,
37    average NOX: 102 ppb, max NOX: 225 ppb). The estimated near-road gradient was slightly sharper for
38    non-inversion conditions, with an average on-road to 20 m difference of 22% for NO2 (Table A-ll) and
39    26% for NOX (not shown) for non-inversion conditions and an average difference of 16% for NO2 (Table
40    A-10) and 17% for NOX (not shown) for  inversion conditions. Across the second to fifth quintile for

                                                  A-5

-------
 1    concentration, the differences were roughly 20-30% higher for the non-inversion conditions compared
 2    with the inversion conditions when considering the on-road to 20 m comparison. This makes sense given
 3    that inversion quells convective mixing in the atmosphere. At the lowest concentrations, the
 4    relationship between differences and inversion conditions seems to reverse.

 5    This analysis is limited by the assumption that the logit model is appropriate for every atmospheric
 6    condition included in the data set. Model performance, as measured by R2, varies somewhat by
 7    atmospheric condition but is reasonably high. For example, median R2 decreases slightly from 96% to
 8    92% when stratifying by winds from the west vs. winds from the east. Median R2 declines from 93% to
 9    83% when stratifying by non-inversion vs. inversion conditions. Furthermore, model evaluations
10    (predicted vs.  observed concentrations) did not vary by atmospheric condition. Overall, the logit model
11    still appears to be a reasonable choice to fit the data. Additionally, discrepancies observed between the
12    summary statistics overall and when considering  particular atmospheric conditions might be attributed
13    to an absence of meteorological data at some times of day when concentrations were measured. For
14    example, only 907 hours of concentration data were used in developing the model for inversion
15    conditions, and 4,444 hours of concentration data were used in developing the model for non-inversion
16    conditions compared with 8,466 hours of data overall.  More specifically in this instance, the maximum
17    on-road NO2 estimate of 104 ppb in the overall dataset is not found in any of the stratifications. Another
18    limitation of this work is that this analysis was performed only for the Las Vegas study area. This study
19    area had limited influence from sources other than those originating from the roadway. This relationship
20    would not necessarily be representative for many urban sites with multiple sources including, for
21    example, emissions from additional arterial roads or combustion-related power plants.  However, this
22    work provides important insight about NO2 concentration changes from a  single highway.

23    From a practical perspective, this analysis can shed light on how well the existing near-road monitoring
24    network may be useful in understanding on-road NO2 concentrations. Greater similarity between the
25    percent differences during inversion conditions and differences across all wind directions suggests that
26    inversions are a prevalent feature of the meteorology in Las Vegas. Hence, estimates of 10-15%
27    reductions in NO2 concentration at the 10 m site and 10-20% at the 20 m site might be reasonable in
28    many cases for regions where inversions tend to occur. Where away from  road concentrations are low
29    (i.e., lower percentiles of the overall  concentration  distribution), reductions of 15-25% at the 10 m site
30    and 20-35% at the 20 m site might be more reasonable. In areas where inversions do not tend to occur,
31    the non-inversion conditions may be more typical, with differences of 15-20% at the 10 m site and
32    differences of 20-30% at the 20 m site, though keep in  mind, NO2 concentrations are generally lower for
33    this scenario. The selection of a higher gradient and applied equally to all possible atmospheric
34    conditions or concentration levels would tend to  produce a more conservative estimate of on-road NO2
35    concentrations. However, since the maximum estimated on-road NO2 concentration was 104 ppb, if
36    assuming a steeper gradient produced NO2 concentrations above this value, then the modeler would
37    have to question the validity of the gradient assumption.

38
                                                   A-6

-------
 i    A-4.   References
 2    Nevada DOT (2006) 2006 Annual Traffic Report. Carson City, Nevada. Available at:
 3       https://www.nevadadot.com/About NDOT/NDOT Divisions/Planning/Traffic/2006 Annual Traffic
 4       Report.aspx.
 5    Karner A.A., Eisinger D.S., Niemeier D.A. (2010). Near-roadway air quality: synthesizing the findings from
 6       real-world data. Environ Sci Technol. 44(14):5334-5344.
 7    Kimbrough E.S., Bauldauf R.W., and N. Watkins. (2013). Seasonal and diurnal analysis of NO2
 8       concentrations from a long-duration study conducted in Las Vegas, Nevada. JAWMA. 63(8)934-
 9       942.U.S. EPA. (2015). Integrated Science Assessment for Oxides of Nitrogen - Health Criteria
10       (Second External Review Draft). U.S. EPA, National Center for Environmental Assessment and Office,
11       Research Triangle Park. EPA/600/R-14/006. January 2015. Available at:
12       http://cfpub.epa.gov/ncea/isa/recordisplav.cfm?deid=288043.
                                                   A-7

-------
Table A-l. Summary statistics for observed and predicted concentrations for all wind and stability conditions
combined.

Distance
N02(ppb)
Avg
Stdev
5%
25%
50%
75%
95%
98%
OBSERVATIONS
20m
100m
300m

24.86
12.27
7.20
14.29
23.74
34.85
44.94
49.20
22.65
12.22
5.98
11.46
21.48
32.95
42.46
46.08
19.91
12.35
4.44
8.40
17.46
30.94
40.10
44.11
MODEL PREDICTED
On-road
10m
20m
30m
40m
50m
100m
300m

31.08
13.98
8.78
20.12
31.39
41.36
53.71
59.04
26.66
12.48
7.78
16.18
26.43
36.46
47.10
51.73
25.33
12.24
7.44
14.80
24.54
35.24
45.33
49.57
24.55
12.16
7.19
14.00
23.44
34.59
44.47
48.45
24.00
12.12
6.93
13.34
22.81
34.16
43.90
47.77
23.57
12.11
6.76
12.86
22.22
33.80
43.46
47.27
22.24
12.14
6.06
11.26
20.57
32.72
42.17
46.00
20.14
12.42
4.48
8.49
17.94
31.23
40.40
44.26
Table A-2. Summary statistics for observed and predicted concentrations, westerly winds (210°-330°).

Distance
N02(ppb)
Avg
Stdev
5%
25%
50%
75%
95%
98%
OBSERVATIONS
20m
100m
300m

29.93
11.40
10.88
20.62
31.82
38.64
46.70
50.49
27.16
11.73
7.95
16.12
29.76
36.37
43.49
46.75
24.65
12.54
5.11
12.05
27.90
34.72
42.04
45.28
MODEL PREDICTED
On-road
10m
20m
30m
40m
50m
100m
300m

35.72
11.91
15.93
27.83
36.37
43.85
54.48
59.31
31.28
11.32
12.28
22.55
32.65
39.70
48.46
52.35
29.94
11.37
10.96
20.64
31.77
38.61
46.95
50.76
29.16
11.45
10.19
19.36
31.08
37.90
46.03
49.55
28.61
11.53
9.63
18.33
30.68
37.45
45.48
48.75
28.18
11.60
9.15
17.61
30.41
37.15
44.98
48.18
26.84
11.89
7.69
15.40
29.46
36.27
43.71
46.58
24.73
12.54
5.14
12.13
27.94
34.84
42.26
45.41
Table A-3. Summary statistics for observed and predicted concentrations, easterly winds (30°-150°).

Distance
N02(ppb)
Avg
Stdev
5%
25%
50%
75%
95%
98%
OBSERVATIONS
20m
100m
300m

19.31
11.52
5.66
9.96
16.50
26.49
41.33
47.61
18.11
11.17
4.90
8.83
15.64
25.22
39.40
44.75
16.59
10.78
4.03
7.59
13.97
23.66
36.64
42.65
MODEL PREDICTED
On-road
10m
20m
30m
40m
50m
100m
300m

22.95
13.24
6.63
12.44
19.88
31.04
47.77
55.33
20.49
12.01
6.11
10.77
17.57
27.98
43.34
49.92
19.74
11.72
5.93
10.15
16.93
27.08
41.70
48.58
19.31
11.57
5.75
9.72
16.45
26.57
41.13
47.97
19.00
11.47
5.66
9.48
16.18
26.30
40.61
47.31
18.77
11.40
5.54
9.37
16.03
25.92
40.34
46.80
18.02
11.21
5.11
8.69
15.29
25.16
39.07
45.47
16.85
11.01
4.09
7.66
14.18
24.14
37.26
43.35
                                                 A-8

-------
Table A-4. Summary statistics for observed and predicted concentrations for inversion conditions (convective
mixing height less than 300 m).

Distance
N02(ppb)
Avg
Stdev
5%
25%
50%
75%
95%
98%
OBSERVATIONS
20m
100m
300m

31.72
12.14
11.85
22.79
32.58
40.89
50.90
53.85
28.14
11.54
9.65
18.83
28.50
37.07
46.09
49.89
25.96
11.16
8.50
16.95
26.53
33.99
43.95
46.89
MODEL
On-road
10m
20m
30m
40m
50m
100m
300m

38.41
14.42
13.70
27.96
39.44
49.30
60.18
65.58
33.33
12.60
12.41
23.71
33.97
42.91
53.02
56.92
31.80
12.19
11.90
22.73
32.77
41.03
50.95
53.78
30.91
11.98
11.45
21.78
31.84
39.96
49.79
52.70
30.28
11.84
11.19
21.19
31.10
39.28
48.77
52.08
29.78
11.75
10.95
20.67
30.52
38.61
48.12
51.48
28.26
11.52
10.00
19.04
28.67
36.89
46.48
49.36
25.84
11.32
8.41
16.47
26.05
34.12
43.94
47.04
Table A-5. Summary statistics for observed and predicted concentrations for non-inversion conditions
(convective mixing height greater than 300 m).

Distance
N02(ppb)
Avg
Stdev
5%
25%
50%
75%
95%
98%
OBSERVATIONS
20m
100m
300m

16.85
8.95
5.95
10.30
15.04
21.24
34.92
40.15
13.89
8.55
4.87
7.97
10.79
16.09
31.04
35.89
11.56
8.47
3.64
5.70
7.98
13.71
28.89
34.13
MODEL
On-road 10m 20m 30m 40m 50m 100m 300m

22.81 18.25 16.88 16.07 15.50 15.06 13.69 11.51
11.77 9.42 8.94 8.72 8.60 8.53 8.40 8.54
6.88 6.42 6.08 5.94 5.81 5.70 5.16 3.63
13.97 11.23 10.32 9.75 9.36 9.02 7.95 5.63
21.32 16.53 15.03 14.13 13.38 12.85 11.02 8.17
30.08 23.16 21.21 19.99 19.17 18.50 16.82 14.90
43.79 36.71 34.94 34.10 33.49 33.07 31.41 29.62
50.39 42.40 40.27 39.12 38.55 38.38 36.84 35.10
Table A-6. Summary statistics for percent change in NO2 concentration from modeled on-road to
concentrations at distances of 10 m and 20 m away from roads for all atmospheric conditions combined, by
quintile of NO2 predicted concentrations at 10 m and 20 m away from the roads.

Quintile
1
2
3
4
5
overall
10 m to on-road comparison
Cone range (ppb)
-1.27 to 15. 16
15. 17 to 23. 78
23. 78 to 33. 24
33. 24 to 42.55
42.55 to 69.01
-1.27 to 69.01
Average
% change
13%
19%
14%
10%
11%
13%
20 m to on-road comparison
Cone range (ppb)
1.32 to 13.79
13. 80 to 22. 21
22.21 to 31.95
31.95 to 41. 12
41.13 to 68.60
1.32 to 68.60
Average
% change
19%
25%
17%
11%
12%
17%
                                                A-9

-------
Table A-7. Summary statistics for percent change in NO2 concentration from modeled on-road to
concentrations at distances of 10 m and 20 m away from roads for winds from the west (210°-330°), by
quintile of NO2 predicted concentrations at 10 m and 20 m away from roads.

Quintile
1
2
3
4
5
overall
10 m to on-road comparison
Cone range (ppb)
0.50 to 20.69
20.70 to 30.15
30.15 to 36.48
36.49 to 42.85
42.89 to 64.84
0.50 to 64.84
Average
% change
25%
15%
8%
8%
10%
13%
20 m to on-road comparison
Cone range (ppb)
1.32 to 18.73
18.77 to 28.94
28.94 to 35. 50
35. 50 to 41.59
41.60 to 62.56
1.32 to 62.56
Average
% change
34%
22%
11%
10%
11%
17%
Table A-8. Summary statistics for percent change in NO2 concentration from modeled on-road to
concentrations at distances of 10 m and 20 m away from roads for winds from the east (30°-150°), by quintile
of NO2 predicted concentrations at 10 m and 20 m away from roads.

Quintile
1
2
3
4
5
overall
10 m to on-road comparison
Cone range (ppb)
-1.27 to 10. 23
10.23 to 16.01
16.02 to 24.25
24.26 to 38.08
38.10 to 69.01
-1.27 to 69.01
Average
% change
10%
12%
10%
9%
9%
10%
20 m to on-road comparison
Cone range (ppb)
2.31 to 9.70
9.70 to 15. 26
15. 26 to 23.50
23. 50 to 36.95
36.96 to 68.60
2.31 to 68.60
Average
% change
15%
17%
11%
12%
11%
13%
Table A-9. Summary statistics for percent change in NO2 concentration from modeled on-road to
concentrations at distances of 10 m and 20 m away from roads for inversion conditions (mixing height < 300
m), by quintile of NO2 predicted concentrations at 10 m and 20 m away from roads.

Quintile
1
2
3
4
5
overall
10 m to on-road comparison
Cone range
(ppb)
4.60 to 22.07
22.08 to 30.35
30.45 to 38.40
38.44 to 46.29
46.37 to 69.01
4.60 to 69.01
Average
% change
10%
14%
13%
12%
12%
12%
20 m to on-road comparison
Cone range (ppb)
5. 17 to 20.55
20.61 to 28.66
28.69 to 36.72
36.84 to 44.57
44.61 to 68.60
5. 17 to 68.60
Average
% change
13%
19%
17%
15%
14%
16%
                                               A-10

-------
Table A-10. Summary statistics for percent change in N02 concentration from modeled on-road to
concentrations at distances of 10 m and 20 m away from roads for non-inversion conditions (mixing height >
300 m), by quintile of N02 predicted concentrations at 10 m and 20 m away from roads.

Quintile
1
2
3
4
5
overall
10 m to on-road comparison
Cone range (ppb)
-1.27 to 10.36
10.36 to 15.01
15.01 to 19.78
19.79 to 27.27
27.29 to 67.78
-1.27 to 67.78
Average
% change
7%
20%
22%
21%
15%
17%
20 m to on-road comparison
Cone range (ppb)
2.31 to 9.63
9.64 to 13.53
13.54 to 17.94
17.94 to 25. 14
25.15 to 66.94
2.31 to 66.94
Average
% change
10%
27%
29%
26%
17%
22%
      70
      60
      50
    CL
    O.

    B40
      30
      20
      10
                       50
                                     100            150
                                           Distance from road (m)
                                                                  200
                                                                                250
                                                                                              300
Figure A-l. Predicted and observed N02 concentrations for all wind and stability conditions combined.
Predicted median (solid), predicted 98th and 2nd percentile (dotted), observed median (circles), and observed
98th and 2nd percentiles (error bars) are shown.
                                                A-ll

-------
                                     100            150
                                           Distance from road (m)
                                                                  200
                                                                                 250
                                                                                                300

Figure A-2. Predicted and observed N02 concentrations for winds from the west. Predicted median (solid),
predicted 98th and 2nd percentile (dotted), observed median (circles), and observed 98th and 2nd percentiles
(error bars) are shown.
                                                 A-12

-------
     70
     60
     50
   §40
   o 30
   (J
     10
                      50
                                     100
                                                    150
                                           Distance from road (m)
                                                                   200
                                                                                  250
                                                                                                300
Figure A-3. Predicted and observed N02 concentrations for winds from the east. Predicted median (solid),
predicted 98th and 2nd percentile (dotted), observed median (circles), and observed 98th and 2nd percentiles
(error bars) are shown.
                                                 A-13

-------
     70
     60
     50
   §40
     30
     20
     10  -
                                     100             150
                                           Distance from road (m)
                                                                   200
                                                                                  2SO
                                                                                                 300
Figure A-4. Predicted and observed N02 concentrations for inversion conditions (convective mixing height less
than 300 m). Predicted median (solid), predicted 98th and 2nd percentile (dotted), observed median (circles),
and observed 98th and 2nd percentiles (error bars) are shown.
                                                 A-14

-------
     70
     60
     50
     40
   o 30
     20
     10
                                     100             150
                                           Distance from road (m)
                                                                   200
                                                                                  250
                                                                                                300
Figure A-5. Predicted and observed N02 concentrations for non-inversion conditions (convective mixing height
greater than 300 m). Predicted median (solid), predicted 98th and 2nd percentile (dotted), observed median
(circles), and observed 98th and 2nd percentiles (error bars) are shown.
                                                 A-15

-------
This page left intentionally blank
             A-16

-------
APPENDIX B: CONCENTRATION PLOTS AND MAPS
OF SELECTED PHILADELPHIA CBSA MONITORS
       This appendix provides plots of low- and high-concentration year ambient monitoring data and
maps showing the local built environment and natural features surrounding monitors. All high-
concentration year data used were from 1984. Two sets of low concentration year data were used; the first
was 2007, the same year used in the 2008 REA (Rizzo, 2008), the second low concentration year was
from 2013 where available, otherwise, the most recent year of air quality data available.

                                         List of Figures
Figure B-l.   Comparison of low- (2007, y-axis top left panel) and high- (1984, x-axes top panels)
            concentration years for Philadelphia CBSA monitor ID 340070003. Map indicating the
            monitor location within CBSA (middle panel) and expanded view (bottom panel) to show
            the local built-environment and natural features proximal to the monitor	B-l
Figure B-2.   Comparison of low- (2007, y-axis top left panel; 2013, y-axis top right panel) and high-
            (1984, x-axes top panels) concentration years for Philadelphia CBSA monitor ID
            420170012. Map indicating the monitor location within CBSA (middle panel) and
            expanded view (bottom panel) to show the local built-environment and natural features
            proximal to the monitor	B-2
Figure B-3.   Comparison of low- (2007, y-axis top left panel; 2013, y-axis top right panel) and high-
            (1984, x-axes top panels) concentration years for Philadelphia CBSA monitor ID
            420450002. Map indicating the monitor location within CBSA (middle panel) and
            expanded view (bottom panel) to show the local built-environment and natural features
            proximal to the monitor	B-3
Figure B-4.   Comparison of low- (2007, y-axis top left panel; 2008, y-axis top right panel) and high-
            (1984, x-axes top panels) concentration years for Philadelphia CBSA monitor ID
            420910013. Map indicating the monitor location within CBSA (middle panel) and
            expanded view (bottom panel) to show the local built-environment and natural features
            proximal to the monitor	B-4
Figure B-5.   Comparison of low- (2007, y-axis top left panel; 2013, y-axis top right panel) and high-
            (1984, x-axes top panels) concentration years for Philadelphia CBSA monitor ID
            421010004. Map indicating the monitor location within CBSA (middle panel) and
            expanded view (bottom panel) to show the local built-environment and natural features
            proximal to the monitor	B-5
Figure B-6.   Comparison of low- (2007, y-axis top left panel; 2013, y-axis top right panel) and high-
            (1984, x-axes top panels) concentration years for Philadelphia CBSA monitor ID
            421010047. Map indicating the monitor location within CBSA (middle panel) and
            expanded view (bottom panel) to show the local built-environment and natural features
            proximal to the monitor	B-6

Reference
Rizzo M. (2008). Investigation of how distributions of hourly nitrogen dioxide concentrations have
       changed over time in six cities. Nitrogen Dioxide NAAQS Review Docket (EPA-HQ-OAR-
       2006-0922). Available at: http://www.cpa.gov/ttn/naaqs/standards/nox/s_nox_cr_rc
                                           B-i

-------
This page left intentionally blank
              B-ii

-------
30
70
.g.50
— 50
0
Z
T 40 -
T-t
13°-
O
"20
10
0 <
1
Philadelphia Monitor ID 340070003


V - 0.4B41X * S.WS6
B> • 0.9399





.



J



x
/
.X





s
s






..•••





.'
•






*





•




) 20 40 60 80 100 120 140 160
19S4DM1HNQ, (ppb)
Figure B-l. Comparison of low- (2007, y-axis top left panel) and high- (1984, x-axis top panel) concentration
years for Philadelphia CBSA monitor ID 340070003. Map indicating the monitor location within CBSA
(middle panel) and expanded view (bottom panel) to show the local built-environment and natural features
proximal to the monitor.
                                               B-l

-------

60

307DM1HN
1 0 i
CM «"
0
I
Philadelphia Monitor ID 420170012







)
R« = 0.88S



/




20 40



---




^




•"'










•





•











60 80 100 120 140 160 180
1984 DM1H NO, (ppb)

60

5 30




Philadelphia Monitor ID 420170012

B





0
4039X- 3





, \*f^

?S6







20 40





^




. • • •'











'




•









60 SO 100 120 140 160 130
1984 DM1HN02 (ppb)
Figure B-2. Comparison of low- (2007, y-axis top left panel; 2013, y-axis top right panel) and high- (1984, x-
axes top panels) concentration years for Philadelphia CBSA monitor ID 420170012. Map indicating the
monitor location within CBSA (middle panel) and expanded view (bottom panel) to show the local built-
environment and natural features proximal to the monitor.
                                               B-2

-------

70

I 40 -
|







x
s


20 40







, • '
-"•















*





60 80 100 120 140
1984DMlHNO,(ppb]




o
z
I 40 -

10 •



Philadelphia Monitor ID 420450002

v ublltii nsij.'.'
R' • 0.983





0




S
~^fr





m
x
J

20 40




,-="•





•
.












•






60 80 100 120 140
19S4DMlHNO,(ppb)
Figure B-3. Comparison of low- (2007, y-axis top left panel; 2013, y-axis top right panel) and high- (1984,
axes top panels) concentration years for Philadelphia CBSA monitor ID 420450002. Map indicating the
monitor location within CBSA (middle panel) and expanded view (bottom panel) to show the local built-
environment and natural features proximal to the monitor.
                                               B-3

-------

90

?70 .
o" 60
T 50
o 30 -
20
10
Philadelphia Monitor ID 420910013






0

V - O.AilSs > 6.5W
H' • O.B438


1




y


J-

y
/




•
,••





































.







25 50 75 100 125 150 175 200
1984 DMIHNOj (ppb)

90 i

?70
"
0™ 60
•T 50
I 40
0 30 .


Philadelphia Monitor ID 420910013






0
y = 0.<897l« J.200




7









^xj
,,.>*





,"







































25 50 75 100 125 150 175 200
19B4DM1HNO, (ppb)
Figure B-4. Comparison of low- (2007, y-axis top left panel; 2008, y-axis top right panel) and high- (1984,
axes top panels) concentration years for Philadelphia CBSA monitor ID 420910013. Map indicating the
monitor location within CBSA (middle panel) and expanded view (bottom panel) to show the local built-
environment and natural features proximal to the monitor.
                                               B-4

-------
Philadelphia Monitor ID 421010004


?70 .
o" 60
T 50
1 40
o 30
fM
JO
10


R* = Q.M

.7643
1







1
,'
I1
! '
*

0 25 50














;
*
























•







*








75 1QO 125 150 175 200 225
19S4DM1HNQ, (ppb)



?70
"
0™ 60
•T 50
1 10
en
a 30 i
N
10
0


Philadelphia Monitor ID 421010004







]




3483* » 6 8379
K- n.wi,,.i





.!'






'

25 50






. 1
1'








•

























•












75 10O 125 150 175 2OO 225
1984DMlHN02(ppb)
Figure B-5. Comparison of low- (2007, y-axis top left panel; 2013, y-axis top right panel) and high- (1984, x-
axes top panels) concentration years for Philadelphia CBSA monitor ID 421010004. Map indicating the
monitor location within CBSA (middle panel) and expanded view (bottom panel) to show the local built-
environment and natural features proximal to the monitor.
                                               B-5

-------
Philadelphia Monitor ID 4Z1Q1Q047
140
finn
| so
° 40
20
0
(






1





R' " 0.963












,.''
25 5O




.'




*

,



'
























75 1OQ 125 150 175 200 225
19S4DMlHMQ](ppb)
Philadelphia Monitor ID 421010047


firm
1 so -
I 60
m
20






y*0.419Ix 0,3996
R1 • 0.8504





0














i
,-'
25 50










'
























75 10O 125 150 175 200 225
19S4DM1HNO, (ppb}
Figure B-6. Comparison of low- (2007, y-axis top left panel; 2013, y-axis top right panel) and high- (1984, x-
axes top panels) concentration years for Philadelphia CBSA monitor ID 421010047. Map indicating the
monitor location within CBSA (middle panel) and expanded view (bottom panel) to show the local built-
environment and natural features proximal to the monitor.
                                               B-6

-------
This page left intentionally blank
             B-7

-------
United States                         Office of Air Quality Planning and Standards        Publication No. EPA-452/D-15-001
Environmental Protection              Health and Environmental Impacts Division                                May 2015
Agency                             Research Triangle Park, North Carolina 27711

-------