&EPA
United States
EnviroimnU Protection
Agency
Risk and Exposure Assessment to Support
the Review of the Carbon Monoxide Primary
National Ambient Air Quality Standards:


First External Review Draft

-------

-------
                                                      EPA-452/P-09-008
                                                          October 2009
Risk and Exposure Assessment to Support the Review of the
Carbon Monoxide Primary National Ambient Air Quality
Standards:
First External Review Draft
                    U.S. Environmental Protection Agency
                  Office of Air Quality Planning and Standards
                  Health and Environmental Impacts Division
                    Research Triangle Park, North Carolina

-------
                                   DISCLAIMER


       This document has been reviewed by the Office of Air Quality Planning and Standards,
U.S. Environmental Protection Agency (EPA), and approved for publication. This draft
document has been prepared by staff from the Office of Air Quality Planning and Standards,
U.S. Environmental Protection Agency. Any opinions, findings, conclusions, or
recommendations are those of the authors and do not necessarily reflect the views of the EPA
Mention of trade names or commercial products is not intended to constitute endorsement or
recommendation for use.  This document is being provided to the Clean Air Scientific Advisory
Committee for their review, and made available to the public for comment. Any questions or
comments concerning this document should be addressed to Souad Benromdhane, U.S.
Environmental Protection Agency, Office of Air Quality Planning and Standards, C504-06,
Research Triangle Park, North Carolina 27711 (email: benromdhane.souad@epa.gov).
Elements of this report have been provided to the U.S. Environmental Protection Agency (EPA)
by Abt Associates, Inc. in partial fulfillment of Contract No. EP-D-08-100, Work Assignment 0-
08.
                              ACKNOWLEDGEMENTS


       In addition to EPA staff, personnel from Abt Associates, Inc. contributed to the writing of
this document. Specific chapters and appendices where Abt Associates, Inc. made significant
contributions include: chapter 5 (Approach to Exposure and Dose Assessment for the Current
Review), chapter 6 (Exposure/dose assessment), and appendices B (Mass Balance Model in
APEX), and C (COHb Module for pNEM/CO).

-------
                            Table of Contents

     List of Tables	iv
     List of Figures	iv
     I   INTRODUCTION	1-1
        1.1   BACKGROUND	1-1
        1.2   ASSESSMENTS FROM PREVIOUS REVIEWS	1-3
        1.3   CURRENT ASSESSMENT	1-5
        1.4   REFERENCES	1-6
     2  OVERVIEW OF EXPOSURE AND DOSE ASSESSMENT CONCEPTUAL
        MODEL	2-1
        2.1   SOURCES OF CARBON MONOXIDE	2-1
        2.2   EXPOSURE PATHWAYS AND RELEVANT MICROENVIRONMENTS	
             	2-1
        2.3   AT-RISK-POPULATIONS	2-2
        2.4   EXPOSURE AND DOSE METRICS	2-4
        2.5   RISK CHARACTERIZATION METRIC	2-4
        2.6   REFERENCES	2-6
     3   AIR QUALITY CONSIDERATIONS	3-1
        3.1   AMBIENT CO MONITORING	3-1
           3.1.1  Monitoring Network	3-1
           3.1.2  Analytical Sensitivity	3-2
           3.1.3  Patterns of CO Concentrations	3-4
           3.1.4  Policy-Relevant Background Concentrations	3-11
        3.2   STUDY AREAS FOR CURRENT ASSESSMENTS	3-12
        3.3   KEY OBSERVATIONS	3-16
        3.4   REFERENCES	3-17
     4  APPROACH TO RISK CHARACTERIZATION FOR THE CURRENT
        REVIEW	4-1
        4.1   CARDIOVASSCULAR DISEASE RELATED EFFECTS	4-1
        4.2   HEALTH EFFECTS BENCHMARKS	4-2
        4.3   KEY OBSERVATIONS	4-4
        4.4   REFERENCES	4-5
October 2009                       i           Draft - Do Not Quote or Cite

-------
         APPROACH TO EXPOSURE AND DOSE ASSESSMENT FOR CURRENT
         REVIEW	5-1
         5.1   MODEL OVERVIEW	5-1
         5.2   MODEL HISTORY AND EVOLUTION	5-2
         5.3   MODEL SIMULATION PROCESS	5-3
            5.3.1  Characterize the Study Area	5-5
            5.3.2  Generate Simulated Individuals	5-5
            5.3.3  Construct Activity Sequences	5-5
            5.3.4  Calculate Microenvironmental Concentrations	5-7
            5.3.5  Estimate Energy Expenditure and Ventilation Rates	5-9
            5.3.6  Calculate Exposure	5-11
            5.3.7  Calculate Dose	5-11
            5.3.8  Model Output	5-13
            5.3.9  Model limitations	5-13
         5.4   PERSONAL EXPOSURE AND THE IN-VEHICLE MICROENVIRONMENT
              	5-15
            5.4.1  Personal Exposure Monitoring Studies	5-15
            5.4.2  In-Vehicle Concentrations	5-17
         5.5   STRATEGY FOR CO EXPOSURE/DOSE ASSESSMENT FOR THE
              CURRENT REVIEW	5-23
            5.5.1  Background for Current Assessment Strategy	5-23
            5.5.2  Selected Approach  for Current Review	5-24
         5.6   KEY OBSERVATIONS	5-30
         5.7   REFERENCES	5-31
         EXPOSURE/DOSE ASSESSMENT AND RISK CHARACTERIZATION	6-1
         6.1   APPLICATION OF APEX4.3 TO CARBON MONOXIDE	6-1
            6.1.1  Study Areas and Exposure Periods	6-2
            6.1.2  Exposure Scenarios	6-6
            6.1.3  Populations-at-Risk	6-6
            6.1.4  Air Quality and Meteorological Data	6-8
            6.1.5  Microenvironments	6-13
            6.1.6  Time/Activity Patterns	6-14
         6.2   EXPOSURE AND DOSE ESTIMATES AND RISK CHARACTERIZATION.
              	6-14
            6.2.1  Denver- Scenarios A andB	6-14
            6.2.2  Los Angeles - Scenarios A andB	6-21
October 2009                         ii            Draft - Do Not Quote or Cite

-------
             6.2.3  Comparison of Denver and Los Angeles Estimates for End-of-hour COHb
                   Levels	6-26

          6.3   COMPARISON OF COHB ESTIMATES OBTAINED FROM THE 2000
               PNEM/CO AND 2009 APEX/CO ASSESSMENTS	6-26
             6.3.1  Important Differences Between the 2000 pNEM/CO and 2009 APEX/CO
                   Exposure Assessments	6-27
             6.3.2  Comparison of Estimated COHb Levels in Adults with Coronary Heart
                   Disease using the 2000 pNEM/CO and 2009 APEX/CO Assessments	
                   	6-28

          6.4   VARIABILITY ANALYSIS AND UNCERTAINTY CHARACTERIZATION
               	6-34
             6.4.1  Analysis of Variability	6-34
             6.4.2  Characterization of Uncertainty	6-36

          6.5   KEY OBSERVATIONS	6-42
          6.6   REFERENCES	6-44
                                  List of Figures

Figure 3-1. Spatial and Temporal Trends in the 2nd Highest 1-hour (top) and 8-hour Average
          (bottom) CO Ambient Monitoring Concentrations in Denver, Colorado, Years 1993 -
          2008	3-5
Figure 3-2. Spatial and Temporal Trends in the 2nd Highest 1-hour (top) and 8-hour Average
          (bottom) CO Ambient Monitoring Concentrations in Los Angeles, California, Years
          1993-2008	3-6
Figure 3-3. Diurnal Distribution of 1-hour CO Concentrations in Denver (Monitor 08-031-0002)
          by Day-type (weekdays-left; weekends-right), Years 1995 (top) and 2006 (bottom)...
          	3-8
Figure 3-4. Diurnal distribution of 1-hour CO concentrations in Los Angeles (Monitor 06-037-
          1301) by day-type (weekdays-left; weekends-right), years 1997 (top) and 2006
          (bottom)	3-9
Figure 5-1. Figure 5-1	Locations of active ambient CO monitors meeting 75% completeness
          criterion in 2006 along with locations of inactive ambient CO monitors, within the
          metropolitan Denver (top) and metropolitan Los Angeles (bottom)	5-27
Figure 5-2. Figure 5-2.. Distribution of 1-hour (top) and 8-hour average daily maximum (bottom)
          CO concentrations at ambient CO monitors in Denver County, year 2006	5-27
Figure 5-3. Figure 5-3.. Distribution of 1-hour (top) and 8-hour average daily maximum (bottom)
          CO concentrations at ambient CO monitors in Los Angeles County, year 2006.... 5-29
Figure 6-1. Map  of the Denver Study Area Defined as a Circle with Radius = 20 km Centered on
          Fixed-Site Monitor ID 080310002	6-4
October 2009                          in            Draft - Do Not Quote or Cite

-------
Figure 6-2. Map of the Los Angeles Study Area Defined as a Circle with radius = 20 km
          Centered on Fixed-Site Monitor No. 060371301	6-5
Figure 6-3. Percentage of Los Angeles and Denver Adults with Coronary Heart Disease (CHD)
          Estimated to Experience a Daily Maximum End-of-hour COHb Level At or Above
          the Specified Percentage for Air Quality Adjusted to Just Meeting the Current
          Standard.  Data taken from Tables 6-22 and 6-23	6-33
                                    List of Tables

Table 3-1. Descriptive Statistics for CO Concentrations Measured at Selected Fixed-Site
          Monitors in the Denver Metropolitan Area for the Years 2005 - 2007	3-13
Table 3-2. Descriptive Statistics for CO Concentrations Measured at Selected Fixed-Site
          Monitors in the Los Angeles Metropolitan Area for the Years 2005 - 2007	3-14
Table 5-1. Parameters of the Factors Model	5-9
Table 5-2. Carbon Monoxide Concentrations Inside and Immediately Outside Vehicles, and
          Indoor/Outdoor Vehicle Ratios	5-21
Table 5-3. Descriptive Statistics for CO Concentrations Measured Inside Vehicles and at Fixed-
          Site Monitors (from Shikiya et al., 1989)	5-23
Table 6-1. National Prevalence Rates for Coronary Heart Disease by Age Range	6-7
Table 6-2. National Prevalence Rates for Coronary Heart Disease by Gender	6-7
Table 6-3. National Prevalence Rates for Coronary Heart Disease Used in APEX, Stratified by
          Age and Gender	6-8
Table 6-4. Site Characteristics of Fixed-site CO Monitors Selected to Represent the Denver and
          Los Angeles Study Areas	6-8
Table 6-5. Descriptive Statistics for 1-hour Carbon Monoxide Concentrations Reported by the
          Selected Denver  and Los Angeles Monitors Before and After Estimation of Missing
          Values	6-9
Table 6-6. Descriptive Statistics for 1-hour Carbon Monoxide Concentrations Reported by the
          Denver and Los Angeles Monitors Before and After Adjustment to Simulate Just
          Meeting the Current 8-Hour CO NAAQS	6-11
Table 6-7. Site Characteristics of Meteorological Monitoring Stations Selected to Represent the
          Denver and Los Angeles Study Areas	6-11
Table 6-8. Number of Person-days for Adults with Coronary Heart Disease (CHD) in the
          Denver Study Area Estimated to Experience a 1-hour Daily Maximum CO Exposure
          at or Above the Specified Concentration	6-17
Table 6-9. Number of Adults with Coronary Heart Disease (CHD) in the Denver Study Area
          Estimated to Experience a 1-hour Daily Maximum CO Exposure at or Above the
          Specified Concentration	6-17
Table 6-10.Number of Person-days for Adults with Coronary Heart Disease (CHD) in the
          Denver Study Area Estimated to Experience an 8-hour Daily Maximum CO
          Exposure at or Above the Specified Concentration	6-18
October 2009
                                       IV
Draft - Do Not Quote or Cite

-------
Table 6-11.Number of Adults with Coronary Heart Disease (CHD) in the Denver Study Area
          Estimated to Experience an 8-hour Daily Maximum CO Exposure at or Above the
          Specified Concentration	6-18
Table 6-12.Number of Person-days for Adults with Coronary Heart Disease (CHD) in the
          Denver Study Area Estimated to Experience a Daily Maximum End-of-hour COHb
          Level at or Above the Specified Concentration	6-19
Table 6-13.Number of Adults with Coronary Heart Disease (CHD) in the Denver Study Area
          Estimated to Experience a Daily Maximum End-of-hour COHb Level at or Above
          the Specified Concentration	6-19
Table 6-14.Estimated Average Number of Days with a Daily Maximum End-of-hour COHb
          Level At or Above the Specified Concentration Per Adult With Coronary Heart
          Disease (CHD) in the Denver Study Area	6-20
Table 6-15.Number of Person-Days for Adults with Coronary Heart Disease (CHD) in the Los
          Angeles Study Area Estimated to Experience a 1-hour Daily Maximum CO Exposure
          At or Above the Specified Concentration	6-22
Table 6-16.Number of Adults with Coronary Heart Disease (CHD) in the Los Angeles Study
          Area Estimated to Experience a 1-hour Daily Maximum CO Exposure At or Above
          the Specified Concentration	6-22
Table 6-17.Number Of Person-Days For Adults with Coronary Heart Disease (CHD) in the Los
          Angeles Study Area Estimated to Experience an 8-hour Daily Maximum CO
          Exposure At or Above the Specified Concentration	6-23
Table 6-18.Number of Adults with Coronary Heart Disease (CHD) in the Los Angeles Estimated
          to Experience an 8-hour Daily Maximum CO Exposure At or Above the Specified
          Concentration	6-24
Table 6-19.Number of Person-Days For Adults With Coronary Heart Disease (CHD) in the Los
          Angeles Study Area Estimated to Experience a Daily Maximum End-of-hour COHb
          Level At or Above the Specified Concentration	6-24
Table 6-20.Number of Adults with Coronary Heart Disease (CHD) in the Los Angeles Study
          Area Estimated to Experience a Daily Maximum End-of-hour COHb Level At or
          Above the Specified Concentration	6-25
Table 6-21. Estimated Average Number of Days with a Daily Maximum End-of-hour COHb
          Level At or Above the Specified Concentration Per Adult with Coronary Heart
          Disease (CHD) in the Los Angeles Study Area	6-25
Table 6-22. Percentage of Denver Adults with Coronary Heart Disease (CHD) Estimated to
          Experience a Daily Maximum End-o-hour COHb Level At or Above the Specified
          Percentage Under Specified Air Quality Conditions for 1995	6-31
Table 6-23.Percentage of Los Angeles Adults with Coronary Heart Disease (CHD) Estimated to
          Experience a Daily Maximum End-of-hour COHb Level At or Above the Specified
          Percentage Under "Just Meeting"Conditions for 1997	6-32
Table 6-24. Summary of How Variability Was Incorporated Into the Exposure Assessment ..6-35
Table 6-25. Characterization of Key Uncertainties in the Draft Assessment for Denver and Los
          Angeles Areas	6-38
October 2009                          v            Draft - Do Not Quote or Cite

-------
 1
 2                                    1.   INTRODUCTION

 3         1.1   BACKGROUND
 4          The U. S. Environmental Protection Agency (EPA) is presently conducting a review of
 5   the national ambient air quality standards (NAAQS) for carbon monoxide (CO). Sections 108
 6   and 109 of the Clean Air Act (Act) govern the establishment and periodic review of the NAAQS.
 7   These standards are established for pollutants that may reasonably be anticipated to endanger
 8   public health and welfare, and whose presence in the ambient air results from numerous or
 9   diverse mobile or stationary sources. The NAAQS are to be based on air quality criteria, which
10   are to accurately reflect the latest scientific knowledge useful in indicating the kind and extent of
11   identifiable effects on public health or welfare that may be expected from the presence of the
12   pollutant in ambient air.  The EPA Administrator is to promulgate and periodically review, at
13   five-year intervals, "primary" (health-based) and "secondary" (welfare-based) NAAQS for such
14   pollutants. Based on periodic reviews of the air quality criteria and standards, the Administrator
15   is to make revisions in the criteria and standards, and promulgate any new standards, as may be
16   appropriate.  The Act also requires that an independent scientific review committee advise the
17   Administrator as part of this NAAQS review process, a function performed by the Clean Air
18   Scientific Advisory Committee (CASAC).
19          The current NAAQS for CO includes two primary standards to provide protection for
20   exposures to carbon monoxide. In 1994, EPA retained the primary standards at  9 parts per
21   million (ppm), 8-hour average and 35 ppm, 1-hour average, neither to be exceeded more than
22   once  per year (59 FR 38906).  These standards were based primarily on the clinical evidence
23   relating carboxyhemoglobin (COHb) levels to various adverse health endpoints and exposure
24   modeling relating CO exposures to COHb levels. The review completed in 1994 also reaffirmed
25   an earlier decision  that the evidence did not support the need for a secondary standard for CO (59
26   FR38906).
27          A subsequent review of the CO NAAQS was initiated in 1997, which led to the
28   completion of the 2000 Air Quality  Criteria Document for Carbon Monoxide (US EPA,  2000)
29   and a draft exposure analysis methodology document (US EPA, 1999). EPA put on hold the
30   NAAQS review when Congress requested that the National Research Council (NRC) review the
31   impact of meteorology and topography on ambient CO concentrations in high altitude and

     October, 2009                              1 -1                Draft - Do Not Cite or Quote

-------
 1    extreme cold regions of the U.S.  In response, the NRC convened the Committee on Carbon
 2    Monoxide Episodes in Meteorological and Topographical Problem Areas, which focused on
 3    Fairbanks, Alaska as a case-study. A final report, "Managing Carbon Monoxide Pollution in
 4    Meteorological and Topographical Problem Areas" (NRC, 2003), offered a wide range of
 5    recommendations regarding management of CO air pollution, cold start emissions standards,
 6    oxygenated fuels, and CO monitoring. Following completion of this NRC report, EPA did not
 7    conduct rulemaking to complete the review.
 8          EPA initiated the current review of the NAAQS for CO on September 13, 2007, with a
 9    call for information from the public (72 FR 52369) requesting the submission of recent scientific
10    information on specified topics. A workshop was held on January 28-29, 2008 (73 FR 2490) to
11    discuss policy-relevant scientific and technical information to inform EPA's planning for the CO
12    NAAQS review.  Following the workshop, EPA outlined the science-policy questions that would
13    frame this review, outlined the process and schedule that the review would follow, and provided
14    more complete descriptions of the purpose, contents, and approach for developing the key
15    documents that would be developed in the review in a draft Integrated Review Plan for the
16    National Ambient Air Quality Standards for Carbon Monoxide (US EPA, 2008a).  After CAS AC
17    and public input on the draft plan, EPA made the final plan available in August 2008 (US EPA,
18    2008b).  EPA is currently completing the process of assessing the latest available policy -
19    relevant scientific information to inform the review of the CO standards. The latest draft of this
20    assessment is contained in the second external review draft of the Integrated Science Assessment
21    for Carbon Monoxide (hereafter,  "draft ISA") (US EPA,  2009c) which was released in
22    September 2009 for review by the CASAC and for public comments. The draft ISA includes an
23    evaluation of the scientific evidence on the health effects of CO, including information on
24    exposure, physiological mechanisms by which CO might adversely impact human health, an
25    evaluation of the clinical evidence for CO-related morbidity, and an evaluation of the
26    epidemiological evidence for CO-related morbidity and mortality associations.l
27          Building upon the health effects evidence presented in the draft ISA as well as CASAC
28    advice (Brain and Samet, 2009) and public comments on a scope and methods planning
29    document for the exposure/risk assessment (hereafter, "Scope and Methods Plan") (US EPA,
      1 The draft ISA also evaluates scientific evidence for the effects of CO on public welfare which EPA will consider in
      its review of the need for a secondary standard. EPA is not intending to do a quantitative risk assessment for the
      secondary standard review.
      October, 2009                               1 -2                Draft - Do Not Cite or Quote

-------
 1   2009a), EPA's Office of Air Quality Planning and Standards (OAQPS) has developed this first
 2   draft Risk/Exposure Assessment describing the initial quantitative assessments being conducted
 3   by the Agency to support the review of the primary CO standards. This draft document is a
 4   concise presentation of the methods, key results, observations, and related uncertainties
 5   associated with the quantitative analyses performed.  The final REA will draw upon the final
 6   ISA and will reflect consideration of CASAC and public comments on this draft REA.
 7          The schedule for completion of this review is governed by a court order that specifies that
 8   EPA sign for publication notices of proposed and final rulemaking concerning its review of the
 9   CO NAAQS no later than October 28, 2010 and May 13, 2011, respectively. The order also  sets
10   dates for the following interim milestones: release of a first draft ISA by March 14, 2009
11   (completed), a first draft risk/exposure assessment by October 29, 2009, a final ISA by January
12   29, 2010, and a final risk/exposure assessment by May 28, 2010.
13          The final ISA and final REA will inform the policy assessment and rulemaking steps that
14   will lead to final decisions  on the CO NAAQS. The policy assessment will be described in a
15   Policy Assessment (hereafter, "PA") document, which will include staff analysis of the scientific
16   basis for alternative policy options for consideration by senior EPA management prior to
17   rulemaking. The PA will integrate and interpret information from the ISA and the REA to frame
18   policy options for consideration by the Administrator. The PA is intended to help "bridge the
19   gap" between the Agency's scientific and technical assessments, presented in the ISA and REA
20   and the judgments required of the Administrator in determining whether it is appropriate to retain
21   or revise the standards.  The PA is also intended to facilitate CASAC's advice to the
22   Administrator on the adequacy of existing standards, and any new standards or revisions to
23   existing standards as may be appropriate. OAQPS currently plans to release a draft PA in late
24   February 2010 for review by CASAC, as well as for public comment, in conjunction with
25   CASAC review and public comment of the second draft REA (US EPA, 2009c).

26         1.2   ASSESSMENTS FROM PREVIOUS REVIEWS
27          Reviews of the CO NAAQS completed in 1985 and 1994  included analysis of exposure
28   to ambient CO and associated internal dose in terms of COHb levels which were used to
29   characterize risks for at-risk populations (50 FR 37484; 59 FR 38906). These prior risk
30   characterizations compared the numbers of at-risk individuals and percent of the at-risk
31   population exceeding several potential health effect benchmarks,  expressed in terms of COHb

     October, 2009                               1 -3               Draft - Do Not Cite or Quote

-------
 1    levels. This characterization was based on COHb levels observed in several controlled human
 2    exposure studies reporting aggravation of angina associated with short-term (< 8-hr) CO
 3    exposures and described in EPA's Air Quality Criteria Document (AQCD) (US EPA,  1979; US
 4    EPA, 1984; US EPA, 1991).
 5          In the review completed in 1994, this characterization was performed for the at-risk
 6    population in the city of Denver, Colorado (US EPA, 1992). That analysis indicated that if the
 7    current 8-hr standard were just met, the proportion of the nonsmoking population with
 8    cardiovascular disease experiencing exposures at or above 9 ppm for 8 hrs decreased by an order
 9    of magnitude or more as compared to the proportion under then-existing CO levels, down to less
10    than 1 percent of the total person-days in that population. Likewise, meeting the current 8-hr
11    standard reduced the proportion of the nonsmoking cardiovascular-disease population person
12    days at or above COHb levels of concern by an order of magnitude or more relative to then-
13    existing CO levels.  More specifically, upon meeting the 8-hr standard, EPA estimated that less
14    than 0.1% of the nonsmoking cardiovascular-disease population would experience a COHb level
15    of about 2.1%. A smaller percentage of the at-risk population was estimated to exceed higher
16    COHb percentages.  The analysis also took into account that certain indoor sources (e.g., passive
17    smoking, gas stove usage) contributed to total CO exposure but could not be effectively
18    mitigated by setting more stringent ambient air quality standards.
19          In the subsequent review, initiated in  1997, EPA consulted with CAS AC on a draft
20    exposure analysis methodology document, Estimation of Carbon Monoxide Exposures and
21    Associated Carboxyhemoglobin Levels in Denver Residents using pNEM/CO (Version 2.0)
22    (Johnson, 1999). Although the EPA did not complete the review initiated in 1997, OAQPS
23    continued work on the CO exposure assessment to further develop the exposure assessment
24    modeling component of the Total Risk Integrated Methodology (TRIM) system. A subsequent
25    draft technical report (Johnson et al., 2000) was produced documenting the application of the CO
26    exposure and dose modeling methodology (and version 2.1  of pNEM/CO) for two study areas
27    (Denver and Los Angeles).  This report was subjected to an external peer review by three
28    exposure modeling experts convened by Science Applications International Corporation (SAIC,
29    2001).
      October, 2009                              1 -4               Draft - Do Not Cite or Quote

-------
 1         1.3  CURRENT ASSESSMENT
 2          In preparing the Scope and Methods Plan for the current health risk/exposure assessment,
 3   we considered the scientific evidence presented in the first draft ISA (US EPA, 2009b) and the
 4   key science policy issues raised in the IRP (US EPA, 2008b). EPA held a consultation with
 5   CASAC to solicit comments on the Scope and Methods Plan during a May 2009 CASAC
 6   meeting at which CASAC also provided comments on the first draft of the ISA. Public
 7   comments were also requested (74 FR 15265).  CASAC and public comments were considered
 8   in advance of the conduct of the analyses and results presented in this draft REA.  The design of
 9   the current risk assessment builds upon information presented in the second draft ISA (US EPA,
10   2009c) with particular attention to conclusions  regarding the adequacy of the air quality data for
11   the purposes of exposure assessment.
12          In this draft assessment we are relying on generally similar methodology and focusing on
13   the same two urban  areas (Denver and Los Angeles) as that used in the assessment for the
14   previous review. Although improvements have been made to the exposure model since the time
15   of the last review, we recognize significant data limitations in the current review.  In CAS AC's
16   comments on the first draft ISA, the Committee stated that the "current ambient monitoring
17   network is not well designed to characterize spatial and temporal variability in ambient
18   concentrations" and that "it does not adequately support detailed assessments of human
19   exposure" (Brain and Samet, 2009).  As a result, the draft assessment that we describe in this
20   document has implemented a much-simplified, screening-level approach focused on a single
21   monitor and an exposure situation of particular interest for ambient CO (as described in detail in
22   chapters 5 and 6). Based on the concerns raised by CASAC regarding the adequacy of the
23   current monitoring data for this purpose, staff decided not to perform a detailed  analysis
24   involving multiple monitors and comprehensive estimation of exposure concentrations in
25   multiple microenvironments, as has been done  in the past. In presenting this draft, screening-
26   level  assessment, however, we recognize that the simplifications in this approach contribute to
27   limitations and uncertainties in the interpretation of the results.  One purpose of this draft
28   document is to seek CASAC views, and public comment, regarding our characterization of the
29   results in light of uncertainties associated with the assessment design and inputs, and CASAC's
30   advice on the role of this assessment in informing the  current review of the CO NAAQS.
      October, 2009                              1 -5                Draft - Do Not Cite or Quote

-------
  1           1.4   REFERENCES

 2    Brain JD and Samet JM (2009).  Letter to EPA Administrator Lisa Jackson: Clean Air Scientific Advisory
 3            Committee's (CASAC) Peer Review of the Agency's  1st Draft Carbon Monoxide Integrated Science
 4            Assessment. EPA-CASAC-09-011. June 24, 2009.

 5    Johnson T, Mihlan, G LaPointe, J, Fletcher K, Capel J. (1999). Estimation of Carbon Monoxide Exposures and
 6            Associated Carboxyhemoglobin levels in Denver Residents Using pNEM/CO (Version 2.0) prepared by
 7            ICF Kaiser Consulting Group for U.S. EPA, Office of Air Quality Planning and Standards, under Contract
 8            No. 68-D6-0064, WA Nos. 1-19 and 2-24.

 9    Johnson T, Mihlan G, LaPointe J, Fletcher K, Capel J. (2000). Estimation of Carbon Monoxide Exposures and
10            Associated Carboxyhemoglobin levels in Denver Residents Using pNEM/CO (Version 2.1), prepared by
11            ICF Kaiser Consulting Group for U.S. EPA, Office of Air Quality Planning and Standards, under Contract
12            No. 68-D6-0064, WA Nos. 1-19, 2-24, 2-30, and 3-3,  Docket EPA-HQ-OAR-2008-0015-DRAFT-0009.

13    SAIC. (2001).  Memo to Harvey Richmmond, EPA, Technical Peer Review (including reviewers comments) of
14            "Estimation of Carbon Monoxide Exposures and Associated Carboxyhemoglobin Levels for Residents of
15            Denver and Los  Angeles Using pNEM/CO (version 2.1)", Docket EPA-HQ-OAR-2008-0015.

16    National Research Council. (2003). Managing Carbon Monoxide Pollution in Meteorological and Topographical
17            Problem Areas. Washington, D.C. The National Academies Press.

18    US EPA. (1979). Air Quality Criteria for Carbon Monoxide. Research Triangle Park, NC: Office of Health and
19            Environmental Assessment, Environmental Criteria and Assessment Office, report no. EPA/600/8-79-022.

20    US EPA (1984). Review of the NAAQS for Carbon Monoxide: Reassessment of Scientific and Technical
21            Information. Office of Air Quality Planning and Standards, report no. EPA-450/584-904. Research
22            Triangle Park, NC.

23     US EPA. (1991). Air Quality Criteria for Carbon Monoxide. Research Triangle Park, NC: Office of Health and
24            Environmental Assessment, Environmental Criteria and Assessment Office, report no. EPA/600/8-90/045F.

25    US EPA (1992). Review of the National Ambient Air Quality  Standards for Carbon Monoxide:  Assessment of
26            Scientific and Technical Information. Office of Air Quality Planning and Standards Staff Paper, report no.
27            EPA/452/R-92-004.

28    US EPA (1999). Total Risk Integrated Methodology - TRIM.Expo Technical Support Document, External Review
29            Draft, November 1999. Office of Air Quality Planning and Standards, U.S. Environmental Protection
30            Agency, Research Triangle Park, NC, report no. EPA-453/D-99-001. Available at:
31            http://www.epa.gov/ttn/fera/trim_fate.html#1999historical.

32    US EPA (2000). Air Quality Criteria for Carbon Monoxide. National Center for Environmental Assessment, Office
33            of Research and Development, U.S.  Environmental Protection Agency, Research Triangle Park, NC, report
34            no. EPA/600/P-99/001F. June 2000. Available at:
35            http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=18163.

36    US EPA (2008a). Draft Integrated Review Plan for the National Ambient Air Quality Standards for Carbon
37            Monoxide. U.S. Environmental Protection Agency, Research Triangle Park, NC, report no. EPA/452D-08-
38            001. Available at: http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_pd.html.

39    US EPA (2008b). Integrated Review Plan for the National Ambient Air Quality Standards for Carbon Monoxide.
40            U.S. Environmental Protection Agency, Research Triangle Park, NC, report no. EPA/452R-08-005.
41            Available at: http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_pd.html.
      October, 2009                                   1 -6                  Draft - Do Not Cite or Quote

-------
 2    US EPA (2009a).  Carbon Monoxide National Ambient Air Quality Standards: Scope and Methods Plan for Health
 3            Risk and Exposure Assessment U. S. Environmental Protection Agency, Research Triangle Park, NC,
 4            report no. EPA-452/R-09-004.  Available at: http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_pd.html.

 5    US EPA (2009b). Integrated Science Assessment for Carbon Monoxide - First External Review Draft. U.S.
 6            Environmental Protection Agency, Research Triangle Park, NC, report no. EPA/600/R-09/019. Available
 7            at: http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_isa.html.

 8    US EPA (2009c). Integrated Science Assessment for Carbon Monoxide - Second External Review Draft. U.S.
 9            Environmental Protection Agency, Research Triangle Park, NC, report no. EPA/600/R-09/019B. Available
10            at: http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_isa.html.
      October, 2009                                    1 -7                  Draft - Do Not Cite or Quote

-------
 1            2.   OVERVIEW OF EXPOSURE AND DOSE ASSESSMENT
 2                                   CONCEPTUAL MODEL

 3          In order to help inform the discussion of the CO assessment presented in chapters 5 and
 4    6, staff has briefly summarized the conceptual model for the consideration of exposure to
 5    ambient CO and associated health risk, from key sources through the identification of at-risk
 6    population groups, dose metric, and the risk characterization approach.
 7       2.1. SOURCES OF CARBON MONOXIDE
 8          Carbon monoxide in ambient air is formed primarily by the incomplete combustion of
 9    carbon-containing fuels and photochemical reactions in the atmosphere.  The amount of CO
10    emitted from these reactions, relative to carbon dioxide (CO2), is sensitive to conditions  in the
11    combustion zone. CO production relative to CO2 generally decreases with any increase  in fuel
12    oxygen (©2) content, burn temperature, or mixing time in the combustion zone (draft ISA,
13    section 3.2).  As a result, CO emissions from large fossil-fueled power plants are typically very
14    low because of the boilers highly efficient combustion and optimized fuel consumption.  In
15    contrast, internal combustion engines used in many mobile sources have widely varying
16    operating conditions. Therefore, higher and more varying CO formation results from the
17    operation of these mobile sources (draft ISA, section 3.2). In 2002, CO emissions from  on-road
18    vehicles accounted for 63% of total emissions by individual source sectors in the U.S. (draft ISA,
19    Figure 3-1).l As with previous reviews, mobile sources continue to be a significant source sector
20    for CO in ambient air.
21          Sources of indoor CO include infiltration of ambient air indoors,  as well as, where
22    present, indoor (nonambient) sources such as gas stoves and environmental tobacco smoke.
23    (draft ISA, section 3.6.5.2).

24       2.2. EXPOSURE PATHWAYS AND  RELEVANT MICROENVIRONMENTS
25          Human exposure to CO involves the contact (via inhalation) between a person and the
26    pollutant in the various locations (or microenvironments) in which people spend their time.
27    Studies of personal exposure to ambient CO have shown that the largest  percentage of the time in
      1 This was the most recent publicly available data tracking CO emissions in the National Emissions Inventory (US
      EPA, 2006), which included data from various sources such as industries and state, tribal, and local air agencies
      (draft ISA, p. 3-2).

      October, 2009                              2-1                Draft - Do Not Cite or Quote

-------
 1    which an individual is exposed to ambient CO occurs indoors (draft ISA, section 2.3).  As a
 2    result of people spending a significant amount of their time indoors (whether at home,  school,
 3    workplace or elsewhere), CO concentrations in indoor microenvironments are an important
 4    determinant of an individual's CO exposures. Microenvironments that may influence CO
 5    exposures typically include residential indoor environments and other indoor locations, near-
 6    traffic outdoor environments and other outdoor locations, and inside vehicles. As is summarized
 7    further in section 5.4, the highest exposure concentrations to ambient CO are experienced by
 8    individuals in transit on or near roadways (draft ISA, section 2.3).  Ambient concentrations near
 9    roadways are generally influenced by vehicle traffic densities (draft ISA, section 3.5.2.2). As a
10    consequence, near-road and in-vehicle exposure to CO will be much higher during commuting
11    times. Thus, exposure to CO near roadway and in vehicle microenvironments are of concern in
12    this review and are a focus of this draft assessment.
13          Although not the focus of this review, indoor sources such as gas stoves and
14    environmental tobacco smoke can, where present,  also be important contributors to total
15    exposure.  For example, some assessments performed for previous reviews have included
16    modeling simulations both without and with indoor sources (gas stoves and environmental
17    tobacco smoke) to provide context for the assessment of ambient CO exposure and dose (e.g.,
18    USEPA, 1994; Johnson et al., 2000).2 As noted in section 5.5, this draft assessment does not
19    include a simulation with indoor sources on.

20       2.3. AT-RISK-POPULATIONS
21          In considering populations for inclusion in this exposure/risk assessment, we considered
22    the evidence regarding those with increased susceptibility or vulnerability. The term
23    'susceptibility' has been used to characterize populations that have a greater likelihood of
24    experiencing effects related to ambient CO exposure, and the term 'vulnerability' has been used
25    to identify  those periods during an individual's life when they are more susceptible to
26    environmental exposures (draft ISA, section 5.7).  In reviewing and setting NAAQS, EPA is
27    required to establish a primary standard that provides protection for population groups that may
28    be at greater risk due to increased susceptibility and/or increased vulnerability.
      2 As has been recognized in previous CO NAAQS reviews, such sources cannot be effectively mitigated by setting
      more stringent ambient air quality standards and are therefore not a focus of this assessment of ambient CO
      exposure and dose.

      October, 2009                             2-2                Draft - Do Not Cite or Quote

-------
 1          The draft ISA states that the strongest evidence regarding CO induced health effects
 2    relates to cardiovascular morbidity indicating that a causal relationship is likely to exist between
 3    relevant short-term CO exposures and cardiovascular morbidity, particularly in individuals with
 4    coronary artery disease (CAD), also referred to as coronary heart disease (CHD) (draft ISA,
 5    section 5.8).  This evidence comes from human exposure studies of individuals with CAD, along
 6    with coherent results from recent epidemiologic studies reporting associations between short-
 7    term CO exposure and increased risk of emergency department visits and hospital admissions for
 8    individuals affected with ischemic heart disease (IHD) and related outcomes (draft ISA, section
 9    5.7). Other subpopulations potentially at risk include individuals with diseases such as chronic
10    obstructive pulmonary disease (COPD), anemia, or diabetes, and individuals in very early or late
11    life  stages, such as older adults or the developing young (draft ISA, section 2.6.1). There is
12    limited evidence available from controlled human exposure, epidemiologic, or toxicological
13    studies characterizing the nature of specific health effects of CO in these subpopulations.
14          The ISA notes that the most compelling evidence of a CO-induced effect on the
15    cardiovascular system at COHb levels relevant to the current NAAQS comes from a series of
16    controlled  human exposure studies among individuals with CHD (draft ISA, section 2.5.1).  The
17    draft ISA indicates that these studies demonstrate consistent decreases in the time to onset of
18    exercise induced angina and ST-segment changes (as indicators of myocardial  ischemia)
19    following CO exposures resulting in COHb levels of 3-6%, with one multicenter study reporting
20    similar effects at COHb levels as low as 2.0-2.4%.  It also recognizes that no human clinical
21    studies have evaluated the effect of controlled exposures to CO resulting in COHb levels lower
22    than 2% (draft ISA, section 5.2.6). Furthermore, human clinical studies of individuals without
23    diagnosed  heart disease that were  conducted since the 2000 CO AQCD did not report an
24    association between CO and ST-segment changes or arrhythmia (draft ISA, section 2.5.1)
25          Therefore, the primary target population for the assessment described in this  document
26    will be adults with CHD (also known as ischemic heart disease (IHD) or CAD). This is the same
27    population group that was the focus of the exposure/dose assessments conducted for previous
28    CO  NAAQS reviews. Coronary heart disease includes those who have angina pectoris (cardiac
29    chest pain), as well as those who have experienced a heart attack.  Approximately 13.7 million
30    people were diagnosed with CHD in 2007, which represent a large population that may be more
      October, 2009                              2-3               Draft - Do Not Cite or Quote

-------
 1    susceptible to ambient CO exposure when compared to the general population (draft ISA, section
 2    5.7).

 3       2.4. EXPOSURE AND DOSE METRICS
 4          Upon inhalation, CO diffuses through the respiratory zone (alveoli) to the blood where it
 5    binds to a number of heme-containing molecules, mainly hemoglobin (Hb), forming
 6    carboxyhemoglobin (COHb).  Inhaled ambient CO elicits various health effects through this
 7    binding and associated alteration of the function of a number of heme-containing molecules,
 8    mainly Hb (draft ISA, section 4.1). The dosimetry and pharmacokinetics of CO are discussed in
 9    detail in chapter 4 of the draft ISA (US EPA, 2009).  The best characterized health effect
10    associated with CO levels of concern is hypoxia (reduced Q^ availability) induced by increased
11    COHb levels in blood (draft ISA, section 5.1.2). Thus, the dose metric used to characterize
12    health risks associated with exposure to ambient CO in this assessment is the level of COHb in
13    the blood. The Coburn-Forster-Kane (CFK) model (draft ISA, section 4.2.1) has been used to
14    estimate dose (blood levels of COHb) for the exposure/dose modeling in this assessment (see
15    section 5.3.7 of this document).

16       2.5. RISK CHARACTERIZATION METRIC
17          The category  of health endpoints on which we focused in  the risk and exposure
18    assessment are those associated with coronary heart disease (see chapter 4). Similar to the
19    approach used in prior CO NAAQS reviews, we have estimated CO exposures and resulting
20    doses (i.e., COHb levels) for the defined at-risk population (people with CHD) and characterized
21    the risk for this population in urban study areas associated with CO levels representing recent air
22    quality and air quality adjusted to simulate just meeting the current CO NAAQS.  In previous
23    reviews, the COHb estimates were compared to potential health benchmarks (see section 1.2
24    above). Although the draft ISA  has described epidemiologic findings from a group of studies,
25    many of which were  conducted since the 2000 CO AQCD, that observe associations between
26    short term ambient CO exposures and increases in emergency department visits and hospital
27    admissions for cardiovascular effects (draft ISA, section 5.2.1.9), a number of issues complicate
28    the use of these studies in a quantitative risk assessment (draft ISA, section 5.2.3).  In
29    consideration of these issues and CAS AC views on the Scope and Methods Plan (Brain and
      October, 2009                             2-4                Draft - Do Not Cite or Quote

-------
 1    Samet, 2009; US EPA, 2009), risk has been characterized in this assessment using the potential
 2    health effect benchmark level approach.3 More specifically, we have estimated the number of
 3    persons and percent of the at-risk population (i.e., individuals with CHD) that exceed potential
 4    health effect benchmark levels, derived from the evaluation of the controlled human exposure
 5    studies mentioned above and specified in terms of COHb levels, associated with various CO air
 6    quality scenarios.
 7          The range of potential health effects benchmarks that we have used extends to lower
 8    levels than the range where controlled human exposure studies reported CO-related health effects
 9    (i.e., 3-6% COHb with one multicenter study reporting effects at 2.0 to 2.4% COHb using gas
10    chromatography (GC)) to take  into consideration both the uncertainty about the actual COHb
11    levels experienced in the controlled human exposure studies due to the use of different
12    measurement methods and that these studies did not include individuals with more severe CHD
13    who may respond at lower COHb levels relative to the subjects tested. In addition, there were no
14    studies evaluating effects of CO below 2.0-2.4% COHb levels.  Based on these considerations,
15    staff has included 1.5-, 2.0-, 2.5-, and 3.0 % COHb as potential health effect benchmark levels in
16    the current CO risk characterization.
17          Two metrics in this CO  exposure/dose assessment are considered: (1) estimates of the
18    number of people, percentage of the at risk population, and total number of person days for CO
19    at-risk populations exposed to CO at concentrations that exceed selected benchmarks (1.5, 2.0,
20    2.5, and 3.0 % COHb) for various CO air quality scenarios; (2) estimates of the number of
21    people and the total number of occurrences for which potential health effect COHb benchmark
22    levels are exceeded for the various CO air quality scenarios.
      3 While not used for the purposes of this quantitative assessment, EPA will fully consider the health evidence,
      including the epidemiological studies, in the Policy Assessment Document, along with considerations based on the
      risk and exposure assessment findings.

      October, 2009                             2-5                Draft - Do Not Cite or Quote

-------
  1        2.6. REFERENCES

  2    Brain JD and Samet JM (2009). Letter to EPA Administrator Lisa Jackson: Clean Air Scientific Advisory
  3            Committee's (CASAC) Peer Review of the Agency's 1st Draft Carbon Monoxide Integrated Science
  4            Assessment. EPA-CASAC-09-011. June 24, 2009.

  5    CDC (2007). Summary Health Statistics for U.S. Adults: National Health Interview Survey, 2007, series 10,
  6            number 240, May 2009.

  7    Coburn R, Forster R, Kane P (1965). Consideration of the physiological variables that determine the blood
  8            carboxyhemoglobin concentrations in man. J. CLin. Invest. 44:1899-1910.

  9    Johnson T, Mihlan G, LaPointe J, Fletcher K,  Capel J. (2000). Estimation of Carbon Monoxide Exposures and
10            Associated Carboxyhemoglobin levels in Denver Residents Using pNEM/CO (Version 2.1), prepared by
11            ICF Kaiser Consulting Group for U.S. EPA, Office of Air Quality Planning and Standards, under Contract
12            No. 68-D6-0064, WA Nos. 1-19, 2-24, 2-30, and 3-3, Docket EPA-HQ-OAR-2008-0015-DRAFT-0009.

13    US EPA. (1991). Air quality criteria for carbon monoxide. Research Triangle Park, NC: Office of Health and
14            Environmental Assessment, Environmental Criteria and Assessment Office; report no. EPA/600/8-90/045F.

15    US EPA (1992).  Review of the National Ambient Air Quality Standards for Carbon Monoxide: Assessment of
16            Scientific and Technical Information. Office of Air Quality Planning and Standards Staff Paper, report no.
17            EPA/452/R-92-004.

18    US EPA. (2006). 2002 National Emissions Inventory Data and Documentation. U.S. Environmental Protection
19            Agency, Office of Air and Radiation, Office of Air Quality Planning and Standards, Research Triangle
20            Park, NC. http://www.epa.gov/ttn/chief/net/2002inventory.html.

21    US EPA (2000). Air Quality Criteria for Carbon Monoxide. National Center for Environmental Assessment, Office
22            of Research and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC, report
23            no.  EPA/600/P-99/001F. June 2000. Available at:
24            http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=18163.

25    US EPA. (2009). Draft Integrated Science Assessment for Carbon Monoxide - Health Criteria (Second External
26            Review Draft). U.S. Environmental Protection Agency, Research Triangle Park, NC, Report no 600/R-
27            09/019B. Available at: http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_isa.html.
      October, 2009                                  2-6                  Draft - Do Not Cite or Quote

-------
 1                           3.  AIR QUALITY CONSIDERATIONS

 2          Ambient air quality data can be used as an indicator of exposure or used in conjunction
 3    with other information to estimate exposure concentrations. How well the ambient air quality is
 4    represented in a particular location is dependent on the ambient monitoring network design
 5    relative to the spatial and temporal characteristics of the pollutant as well understanding the
 6    concentration contribution from important local source emissions.  This chapter summarizes
 7    findings about the current air quality conditions and their spatial distribution, with particular
 8    focus on aspects informative to the design and conduct of this assessment and including
 9    descriptions of CO measurement methods, monitor siting requirements, and monitor locations
10    (section 3.1).1 Section 3.2 then draws upon the information presented in sections 3.1, among
11    other data, to select ambient air quality/study locations most useful in meeting the objectives of
12    the REA. Finally, key observations of the chapter are presented in section 3.3.

13         3.1   AMBIENT CO MONITORING
14          In this section, a broad overview of the monitoring network is provided (section 3.1.1)
15    and is followed by a summary of analytical detection issues (section 3.1.2). Ambient CO
16    concentrations and their spatial and temporal variability are characterized in section 3.1.3.
17    Lastly, estimates of policy-relevant background (PRB) concentrations which are defined as those
18    ambient concentrations that would occur in the U.S. in the absence of anthropogenic emissions in
19    continental North America are presented in section 3.1.4 of this document.

20         3.1.1   Monitoring Network
21          Ambient CO concentrations are measured by monitoring networks that are operated by
22    state and local monitoring agencies in the U.S., and are funded in part by the EPA. The main
23    network providing ambient data for use in comparison to the NAAQS is the State and Local Air
24    Monitoring Stations (SLAMS) network. The subsections below provide specific information
25    regarding the methods used for obtaining ambient CO measurements and the requirements that
26    apply to states in the design of the CO network.
27          Minimum monitoring requirements for CO were revoked in the 2006 revisions to ambient
28    monitoring requirements (see 71 FR 61236,  October 17, 2006).  This action was made to allow
29    for reductions in measurements of some pollutants (CO, SO2, NO2, and Pb) where measured
30    levels were well below the applicable NAAQS and air quality problems were not expected. CO
      1 A complete description of spatial scales is listed in 40 CFR Part 58 Appendix D, section 1.2.  Ambient monitoring
      of other NAAQS pollutants such as NO2 and SO2 follow the same general spatial scales.

      October, 2009                             3-1               Draft - Do Not Cite or Quote

-------
 1    monitoring activities have been maintained at some SLAMS and these measurements of CO are
 2    required to continue until discontinuation is approved by the EPA Regional Administrator.
 3           CO monitors are typically sited to reflect one of the following spatial scales2:
 4         •  Microscale: Data represents concentrations within a 100 m radius of the monitor. For
 5            CO, microscale monitors are sited 2 - 10 m from a roadway.  Measurements are
 6            intended to represent the near-road or street canyon environment.
 7         •  Middle scale: Data represents concentrations averaged over areas defined by 100 - 500
 8            m radii.  Measurements are intended to represent several city blocks.
 9         •  Neighborhood scale: Data represents concentrations averaged over areas defined by
10            0.5 - 4.0 km radii. Measurements are intended to represent extended portions of a  city.
11           In addition to monitoring required for determining compliance with the NAAQS,  the
12    EPA is currently in the process of implementing plans for a new network of multi-pollutant
13    stations called NCore that is intended to meet multiple monitoring objectives.  A subset of the
14    SLAMS network, NCore stations are intended to address integrated air quality management
15    needs to support long-term trends analysis, model evaluation, health and ecosystem studies, as
16    well as the more traditional objectives of NAAQS compliance and Air Quality Index reporting.3
17    States were required to submit to EPA Annual Monitoring Network Plans (AMNP) describing
18    their candidate NCore stations by July 1, 2009. EPA is reviewing these plans and intends to
19    provide station approvals later in 2009.  The complete NCore network, required to be fully
20    implemented by January 1, 2011, will consist of approximately 63 urban and 20 rural stations
21    and will include some existing SLAMS  sites that have been modified to include additional
22    measurements.  Each state will contain at least one NCore station, and 46 of the states plus
23    Washington, D.C. will have at least one urban  station. CO will be measured using trace-level
24    monitors, as will SO2, NO, and NOy.4 The majority of NCore stations will be sited  to represent
25    neighborhood, urban, and regional scales, consistent with the NCore network design objective of
26    representing exposure expected across urban and rural areas in locations that are not dominated
27    by local sources.

28         3.1.2  Analytical Sensitivity
29           To promote uniform enforcement of the air quality standards set forth under the C AA,
30    EPA has established provisions in the Code of Federal Regulations (CFR) under which analytical
      3 (http://www.epa.gov/ttn/amtic/ncore/index.html).
      4 NCore sites must measure, at a minimum, PM2 5 particle mass using continuous and integrated/filter-based
      samplers, speciated PM25, PM10_2.5 particle mass, speciatedPM10_25, O3, SO2, CO, NO/NOY, wind speed, wind
      direction, relative humidity, and ambient temperature (http://www.epa.gov/ttn/amtic/ncore/index.html).

      October, 2009                              3 -2                Draft - Do Not Cite or Quote

-------
 1    methods can be designated as federal reference methods (FRMs) or federal equivalent methods
 2    (FEMs). Measurements for determinations of NAAQS compliance must be made with FRMs or
 3    FEMs.5  Specifications for CO monitoring are designed to help states utilize equipment that has
 4    met performance criteria utilized in the FRM or FEM approval process; operational parameters
 5    are documented in 40 CFR Part 53, Table B-l. Given the levels of the CO NAAQS (35 ppm, 1-
 6    hour; 9 ppm, 8-hour), a 1.0 ppm lower detectable limit (LDL) is well below the NAAQS levels
 7    and is therefore sufficient for demonstration of compliance. However, with ambient CO levels
 8    now routinely near 1 ppm, there is greater uncertainty in a larger portion of the distribution of
 9    monitoring data because a large percentage of these measurements are  below the LDL of
10    conventional monitors. For this reason, a new generation of ambient CO monitors has been
11    designed that provides trace-level measurements with improved sensitivity at or below the
12    typical ambient CO levels measured in most urban and all  rural locations. Additionally, trace-
13    level CO measurements are needed to support additional objectives such as validating the inputs
14    to chemical transport models and assessing the role of transport between urban and rural areas
15    because background CO concentrations on the order of 0.1 ppm are well below the LDL of
16    conventional monitors. Newer GFC instruments have been designed for automatic zeroing to
17    minimize drift (US EPA, 2000).
18          Currently, a total of 13 approved FRMs are in use in the  SLAMS network, based on a
19    retrieval of data reported between 2005 and 2009.  Among these methods, nine are "legacy"
20    monitors with a federal method detection limit (MDL) listed as 0.5 ppm according to records in
21    EPA's Air Quality System (AQS).6  As discussed in the draft ISA, many of the reported
22    concentrations in recent years are near or below these MDLs (draft ISA, p. 3-43). Four of these
23    methods are newer trace-level methods with a federal MDL of 0.02 ppm and a growing body of
24    ambient data from trace-level CO instruments is becoming available. Among newer GFC trace-
25    level instruments, manufacturer-declared LDLs range from 0.02 - 0.04 ppm, with 24-hour zero
26    drift varying between 0.5% within 1 ppm and 0.1 ppm, and precision varying from 0.5% to 0.1
27    ppm. EPA performed MDL testing on several trace-level CO  monitors in 2005 and 2006
28    following the 40 CFR Part 136 procedures. Those tests demonstrated MDLs of approximately
29    0.017 - 0.018 ppm (17-18 ppb), slightly below the stated LDL of 0.02 - 0.04 ppm.
      5 As of August 2009, twenty automated FRMs had been approved for CO measurement. All EPA FRMs for CO
      operate on the principle of non-dispersive infrared (NDIR) detection and can include the gas filter correlation (GFC)
      methodology. An extensive and comprehensive review of NDIR, GFC, and alternative, non-FRM techniques for
      CO detection was included in the 2000 CO AQCD (US EPA, 2000).
      6 Among several of the older instruments (Federal Reference Method codes 008, 012, 018, 033, 041, 050, 051, and
      054), performance testing has shown LDLs of 0.62 - 1.05 ppm, with 24-hour drift ranging from 0.044 - 0.25 ppm
      and precision ranging from 0.022 - 0.067 ppm at 20% of the upper range limit of the instrument (Mitchie et al.,
      1983).
      October, 2009                              3 -3               Draft - Do Not Cite or Quote

-------
 1          Based on a retrieval of data reported between 2005 and 2009 to AQS, a total of 36 trace-
 2    level CO monitors have reported data with the majority of these monitors currently active. The
 3    majority of these active monitors are associated with the implementation of the NCore network.
 4    The extent to which trace-level monitors become integrated into non-NCore SLAMS stations,
 5    however, will depend on the availability of funding for states to replace well-operating legacy
 6    CO monitors as well as the possibility that monitoring requirements for CO might either
 7    encourage  or require such technological improvements.

 8         3.1.3  Patterns of CO Concentrations
 9          As  discussed in the draft ISA, the spatial and temporal patterns of ambient CO
10    concentrations are heavily  influenced by the patterns associated with mobile source emissions
11    (draft ISA, chapter 3).  Based on the 2002 National Emissions Inventory (NEI), on-road mobile
12    sources comprise about half of the total anthropogenic CO emissions, though in metropolitan
13    areas of the U.S. the contribution can be as high a 75% of all CO emissions due to greater motor
14    vehicle density. For example, emissions in Denver county originating from on-road mobile
15    sources is about 71% of total CO emissions (draft ISA, section 3.2). When considering all
16    mobile sources (non-road and on-road combined), the contribution to total  CO emissions can be
17    over 80%.  Again using Denver County  as an example, all mobile sources contribute to about
18    98% of total CO emissions. Temporally, the national-scale anthropogenic  CO emissions have
19    decreased 35% between 1990 and 2002. Nearly all the national-level CO reductions since 1990
20    are the result of emission reductions in on-road vehicles (draft ISA, Figure 3 .-2).
21          Nearly 400 ambient monitoring stations report continuous hourly averages of CO
22    concentrations across the U.S.  Over the period 2005-2007, 291 out of 376 monitors met a 75%
23    completeness requirement, spread among 243 counties, cities, or municipalities (draft ISA,
24    section 3.4.2.2).  All CO concentrations  measured at these monitoring sites are well below the
25    current NAAQS.  For example, in 2007, none of the monitors reported a second-highest 1-hour
26    CO concentration above 35 ppm, the level of the current 1-hour NAAQS, while only two sites
27    reported a 2nd highest 1-hour CO concentrations between 15.1 and 35.0 ppm. Only five counties
28    reported 2nd highest 8-hour CO concentration 5.0 ppm or higher.
      October, 2009                             3 -4                Draft - Do Not Cite or Quote

-------
         20
      -. 18


      3 16
      O
      o 14
      0)
      O)

         12
         10
       o
      ^  8
       a)  6

       O)
      if  4
      es
          2


          0





         12
       I 10
      ^

      O
      o
       a)  8
       O)
       re
       3
       o
      SI
      CO
      +J
       (/)
       0)

       O)
      es
 \ '
                                                    080310002

                                                    080310013

                                                    080310014

                                                    080310019
CT>CT>CT>CT>CT>CT>CT>OOOOOOOOO
CT>CT>CT>CT>CT>CT>CT>OOOOOOOOO


                                 Year
                                                                          -080310002
                                                         	 060370113

                                                         	060371002

                                                         	 080310019
              CO
              CT>
              CT>
     CT)
     CT)
CT>
CT)
CD
CT)
CT)
CT)
CT)
CO
CT)
CT>
CT)
CT)
CT)
O    T-
O    O
O    O
CN    CN

 Year
CN
O
O
CN
CO
O
O
CN
O
O
CN
LO
O
O
CN
co    r^   oo
o    o   o
o    o   o
CN    CN   CN
3    Figure 3-1.   Spatial and Temporal Trends in the 2"  Highest 1-hour (top) and 8-hour
4                 Average (bottom) CO Ambient Monitoring Concentrations in Denver,
5                 Colorado, Years 1993 - 2008.
     October, 2009
                                  5-5
                                            Draft - Do Not Cite or Quote

-------
         24
       Q.
       Q.

       O 18
       O
       0)
       O) -I c
       (0 la
       0)
       0)
       .C
       O)
       if
                                                         A
                                                      	060370002

                                                      	 060370113

                                                      	060371002

                                                      	060371103

                                                      —^060371201

                                                      —I- 060371301

                                                      --+- 060371601

                                                      -+- 060371701

                                                      	060372005

                                                      	060374002

                                                      - -  - 060375001
              CO
              CT>
              CT>
CT)
CT)
              CT>
              CT)
CD
CT)
CT)
CT)
CT)
CO
CT)
CT>
CT)
CT)
CT)
O    T-
O    O
O    O
CN    CN

 Year
     CN
     O
     O
     CN
     CO
     O
     O
     CN
     O
     O
     CN
     LO
     O
     O
     CN
     CD   r^   oo
     o   o   o
     o   o   o
     CN   CN   CN
         18
         ic
         15
      O
      O
       
              CT>
CT)
CT)
              CT)
              CT>
CD
CT)
CT)
CT>
CT)
CO
CT)
CT)
CT>
CT)
CT)
O
O
O
CN
O
O
CN
CN
O
O
CN
CO
O
O
CN
O
O
CN
LO
O
O
CN
CD
o
o
CN
o
o
CN
oo
o
o
CN
                                                 Year
3    Figure 3-2.  Spatial and Temporal Trends in the 2nd Highest 1-hour (top) and 8-hour
4                 Average (bottom) CO Ambient Monitoring Concentrations in Los Angeles,
5                 California, Years 1993 - 2008.
     October, 2009
                                                 Draft - Do Not Cite or Quote

-------
 1          The current levels of ambient CO across the U. S. reflect the steady declines in ambient
 2    concentrations that have occurred over the past several years.  On average across the U.S. the
 3    decline has been on the order of 50% since the early 1990s (draft ISA, Figure 3-31). As an
 4    example, Figures 3-1 and 3-2 illustrate the trends observed in Denver and Los Angeles,
 5    respectively, for the period from  1993 through 2008.  Both the 2nd highest 1-hour and 8-hour
 6    concentrations are plotted for each year from all existing monitors in those metropolitan areas
 7    respectively. Note, these figures indicate both a significant decrease in the 2nd highest 1-hour
 8    and 8-hour average CO concentrations and a relative decrease in spatial variability in ambient
 9    CO concentrations since the last review.
10          Carbon monoxide also exhibits hourly variability within a day, with two distinct temporal
11    patterns  noted for weekdays and weekends (draft ISA, section 3.5.2.2). The diel variation is
12    inherently  linked to the typical commute times-of-day that occurs within urban locations.  In
13    general,  in recent years observed mean and median concentrations for all hours of the day and
14    across all monitors within urban areas demonstrated limited variability, however 90th and  95th
15    percentile hourly concentrations generally exhibit early-morning and late afternoon peak CO
16    concentrations during weekdays (draft ISA,  Figure 3-33). The weekend diel variation in ambient
17    CO concentrations was much lower than that occurring during weekdays as expected due to the
18    relative absence of commuter vehicle traffic during the morning and evening hours of the day.
19    Most urban areas have relatively stable concentrations throughout weekend days at each of the
20    selected  percentiles, though a few locations (e.g., Phoenix, Los Angeles, Seattle) did have a more
21    pronounced late afternoon peak (draft ISA, Figure 3-34).
22          Staff investigated local hourly variation at two separate CO monitors located in Denver
23    and Los  Angeles to illustrate similar trends.  Figure 3-3 indicates that on average, peak ambient
24    CO concentrations that occur during typical  commute times in Denver ranged from about 1 to 5
25    ppm during weekdays in 1995, while, currently, ambient CO concentrations during morning and
26    afternoon commutes range from about 1 to 2 ppm.  Weekends tend to exhibit less variability
27    throughout the day. On average, CO ambient concentrations generally ranged from 1 to 3 ppm
28    throughout the day in  1995, while current weekend concentrations are less than  1 ppm for most
29    hours of the day.  In Los Angeles, both the concentration levels and variability are greater than
30    when compared with similar years and times of day in Denver (Figure 3-4).  Peak ambient CO
31    concentrations are more prominent during morning commutes and generally ranged from 2 to 10
32    ppm in 1995, while currently (year 2006) most commuting times are associated  with
33    concentrations ranging from between 1 and 5 ppm.  The weekend profile exhibits some variation
34    when considering either year, with maximum concentration levels and variability exhibited
35    during the  overnight hours.
36
      October, 2009                              3 -7               Draft - Do Not Cite or Quote

-------
                            Weekdays in 1995
                                          Weekends in 1995
2
3
4
5
CN
Q_
a
o t-f) __
"cc
S o _
tz ^
0
O
O
o in -
0 -

° 0 °
* ° i ° *
° ° °ej8 1 " o ll'°S8°°

1 1 1 1 1 1 I 1 1 1 1 1 1 1 1 1 1 1 I 1 1 1 1 1
0123456789 11 13 15 17 19 21 23
CN
d
o t-f) _
1 ° -
0
O
O
o in -
0 -


illUUliuii*
i i i i i i i i i i i i i i
0123456789 11 13
Clock Hour Clock Hour
CD -
o
CO Concentrat
2 4
i i
0 -
Weekdays in 2006


3 00
9 § & 8 ° °

I I I I I i I I I I I I I I I I I i I I I I I I
0123456789 11 13 15 17 19 21 23


iiililit^i
1 1 1 1 1 1 1 1 1 1
15 17 19 21 23




Weekends in 2006
CD -
"E"
Q. CD -
o
OS
i •* -
CD
O
c
o
O
o " -
O
o -


g S § o f * 8
f **ii*!4il4J4i
i i i i i i i i i i i i i i
0123456789 11 13
Clock Hour Clock Hour


o 8 8
° ° ° ° ° I
lltiliiiei
i i i i i i i i i i
15 17 19 21 23





Figure 3-3. Diurnal Distribution of 1-hour CO Concentrations in Denver (Monitor 08-031-0002) by Day-type (weekdays-left;
weekends-right), Years 1995 (top) and 2006 (bottom). The box encompasses concentrations from the 25th to 75th
percentiles or Interquartile range (IQR), the line bisecting the box is the median, the solid dot within the box is the
mean, the whiskers represent 1.5 times the IQR, and concentrations outside the whiskers are indicated by open circles.
    October, 2009
3-8
Draft - Do Not Cite or Quote

-------
                          Weekdays in 1997
                                             Weekends in 1997
 o
 o
          i  i  i  i  i  i  i  i  I  i  i
          0123456789
                                11   13
                              Clock Hour
 I  I  I
15   17
                                                19
        I  I  I
        21   23
                    o
                    o
                                                                                                                  o    o
I  I   I I   I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I  I
0123456789   11   13  15  17  19  21   23
                    Clock Hour
                          Weekdays in 2006
                                             Weekends in 2006



0 _
'E co -
a.
£i
C
.E CD -
01
cr
CD
0 „• -
o
O
0 " ~
0 -

8
l
0 ° 0
1 8 8 ° T
0
o \
o g
T ;
T
3^
3
&
s
$
i *
° *
° §
° o 8 § 3
08 * i
§ * 1 1 i !
. | j i 1 1 1 : :
In^iBiiian

I I I I I I I I
01 234567
1
2
3
4
5
1 1 1 1 1
89 11
i i i i i i i i i i i
13 15 17 19 21 23
0 _
*E co -
Q_
d_
C
.2 CD -
OJ
c:
CD
0 , -
o
O
0 " -
0 -

8
, 8 o
III

I I I
0 1 2


: : i s
8 8 o
0 0

yyyy
1 1 1 1
3456
Clock Hour


* 0
^!|MM»*

i i i i i i i i i i
789 11 13 15
Clock Hour

0 «
*
0 8 . |
o 9 o S 8
o § ° S * *
H H fi B H Q o

i i i i i i i
17 19 21 23




Figure 3-4. Diurnal distribution of 1-hour CO concentrations in Los Angeles (Monitor 06-037-1301) by day-type (weekdays-
left; weekends-right), years 1997 (top) and 2006 (bottom). The box encompasses concentrations from the 25th to 75th
percentiles or IQR, the
represent 1.5
times the
line bisecting the box is the median, the solid
IQR, and concentrations outside the whiskers
dot within the box is the mean,
the whiskers
are indicated by open circles.
October, 2009
3-9
                                                           Draft - Do Not Cite or Quote

-------
 1           Ambient monitor siting characteristics can also influence ambient CO concentration
 2    observations. Microscale and middle scale monitors are commonly used to measure significant
 3    source impacts, while neighborhood and urban scale monitors are designated for population-
 4    oriented monitoring (40 CFR Part 58 Appendix D). As CO concentrations primarily originate
 5    from vehicle emissions, the microscale and middle scale data can be a useful indicator of near-
 6    road air quality. Such data analyzed in the draft ISA were concluded to be consistent with hourly
 7    concentrations reported in the literature for the near road environment in the U.S. (draft ISA, p.
 8    3-63). Further, when considering monitoring scale across ambient monitors in the U.S., the
 9    median hourly CO concentration measured at microscale monitors was about 25% higher than at
10    middle scale monitors and 67% higher than at neighborhood scale monitors (draft ISA, Table 3-
11    12).  In general, similar patterns were present in the 1-hour daily max, 1-hour daily average, and
12    8-hour daily max distributions (draft ISA, Table 3-12).  These patterns are also consistent with
13    findings presented by other researchers regarding the relative decrease in concentration with
14    increasing distance from roadways, though the magnitude of the relationship can vary.  Two
15    studies summarized in the draft ISA (Zhu et. al., 2002; Baldauf et.  al., 2008) indicate that near-
16    road CO concentrations (i.e., measured within 20 m of an interstate highway) can range from 2 -
17    10 times greater than CO concentrations measured as far as 300 m from a major road (draft ISA,
18    Figures 3-26 and 3-27).
19          While recognizing that monitoring site attributes are not available for all monitors in the
20    current network and that data for some attributes may not reflect current conditions, 7 the draft
21    ISA also analyzed the information available for network monitors on average annual daily traffic
22    (AADT). The ISA noted that only two microscale monitors and two middle scale monitors in
23    the existing network are sited at roads with > 100,000 AADT, although it is not uncommon for
24    roadways within CSAs to have several roads with AADT > 100,000. The AADT ranged from
25    160,000-178,000 for the near-road monitors used in the aforementioned study by Zhu et al.
26    (2002) where CO concentrations were up to 10 times greater than monitors sited at 300 m from a
27    major road. Existing microscale sites near roads having only moderate traffic count data
28    (< 100,000AADT) may record concentrations that are not substantially different from those
29    obtained from neighborhood scale measurements (draft ISA, section 3.5.1.3).
30          Within a specific urban area however, consideration of only monitor scale or other
31    attributes reported in AQS, such as AADT estimates may be of limited use in efforts to
32    characterize the monitoring data as to its representation of local near-road  CO concentrations.
33    For example, of the five monitors meeting a 75% completeness criterion in the Denver
      7Note that recorded AQS monitoring site attributes are not always available for each monitor or may not always
      reflect potential source influences. For example, of 24 CO monitors in the Los Angeles CSA, AQS had no
      information regarding monitoring scale for 16 (draft ISA, Figure 3-20).
      October, 2009                              3-10                  Draft -Do Not Cite or Quote

-------
 1    Consolidated Statistical Area (CSA), three were microscale and two were neighborhood scale
 2    (draft ISA, section 3.5.1.2). While one of the microscale monitors sited within downtown
 3    Denver measured the highest hourly ambient CO concentrations (ID 080310002), another
 4    microscale monitor located outside the urban core measured the lowest hourly ambient CO
 5    concentrations (draft ISA, Figure 3-18). Further, the AADT estimate for a major road near the
 6    microscale monitor within the urban core (ID 080310002, AADT= 17,200) was lower than that
 7    listed for the microscale monitor outside the urban core (ID 080130009, AADT=20,000) (draft
 8    ISA, Table A-2). And, a third microscale monitor located 1.3 km from monitor ID 080310002,
 9    within the urban core, and measuring somewhat lower CO concentrations (but not lower than the
10    monitor outside the urban core) had only 500 AADT listed for the nearest major road. It is likely
11    that the higher CO concentrations measured at the downtown monitor reflect influences of the
12    denser roadway network surrounding that monitor in the downtown Denver area (Figure 3-17).8
13          Thus, to better characterize the representation of near-road CO concentrations for many
14    of the existing ambient monitors, additional analyses, beyond consideration of AQS attributes
15    such as monitoring scale and traffic count, would likely need to be performed (e.g., using GIS to
16    determine monitor distance from roads,  the number and type of roads within close proximity of
17    the monitor, and obtaining current traffic count data for all roads).

18         3.1.4   Policy-Relevant Background Concentrations
19          EPA has generally conducted NAAQS risk assessments that focus on the risks associated
20    with ambient levels of a pollutant that are in excess of policy-relevant background (PRB).
21    Policy-relevant background levels are defined, for purposes of this document, as concentrations
22    of a pollutant that would occur in the U.S. in the absence of anthropogenic emissions in the U.S.,
23    Canada, and Mexico. Over the continental  U.S. (CONUS), the 3-year (2005- 2007) average CO
24    PRB concentration is estimated to range from 118 to 146 ppb (draft ISA, section 3.5.4). Outside
25    the CONUS, the 3-year average CO PRB in three Alaskan sites is estimated to range from 127 to
26    135 ppb, and from 95 to 103 ppb in two Hawaiian monitoring locations.  The estimated PRB
27    concentrations exhibit significant within-location seasonal variation, with minimum
28    concentrations observed in the summer and fall and maximum concentrations occurring in the
29    winter and spring.  For example, PRB in two California sites is estimated to range from about 85
30    to 170 ppb, and one site in Colorado, ranged from about 80 to 140 ppb (Figure 3-40 of the draft
31    ISA).
32          Given that ambient concentrations of interest in this REA are well above the estimated
33    PRB levels discussed above and, thus the contribution of PRB to overall ambient CO
      8 Staff also recognizes some uncertainty in how well the AQS AADT estimates reflect current conditions at this
      monitor site.

      October, 2009                              3-11                 Draft -Do Not Cite or Quote

-------
 1    concentrations is very small, EPA is characterizing risks associated with ambient CO levels
 2    without regard to estimated PRB levels.

 3         3.2   STUDY AREAS FOR CURRENT ASSESSMENT
 4          Staff identified several criteria to select the exposure assessment study areas drawing
 5    from information discussed in the earlier sections of this Chapter and additional scientific
 6    evidence in the draft ISA. We selected Denver and Los Angeles as areas to focus the current
 7    assessment because (1) both cities have been included in prior CO NAAQS exposure
 8    assessments and thus serve as an important connection with past assessments, (2) they have
 9    historically had the highest CO ambient concentrations among urban areas in the U.S., and (3)
10    Denver is at high altitude and represents an important risk scenario due to the increased
11    susceptibility of individuals at high altitude from exposure to  CO. In addition, of 10 urban areas
12    across the U.S. having monitors meeting a 75% completeness criteria, the two locations were
13    ranked 1st (Los Angeles) and 2nd (Denver) regarding percent of elderly population within 5, 10,
14    and 15 km of monitor locations, and ranked 1st (Los Angeles) and 5th (Denver) regarding number
15    of 1- and 8-hour daily maximum CO concentration measurements (draft ISA, section 3.5.1.1).
16          Maximum and 2nd highest 1-hour and 8-hour average CO concentrations are provided in
17    Table 3-1 for all monitors located within four Denver-area counties (i.e., Adams, Boulder,
18    Jefferson, and Denver) having at least one year of ambient monitoring data for years 2005
19    through 2007.9 Table 3-2 provides a similar concentration summary for the Los Angeles
20    monitors in four counties (i.e., Los Angeles, Orange, Riverside,  and San Bernardino) that
21    reported CO concentrations for at least one year between 2005 and 2007.  Additional discussion
22    regarding specific sites and monitoring data used in the exposure modeling is provided in
23    sections 5.5 and 6.1.2.
24          In order to investigate ambient CO concentrations in each study area,  EPA initially
25    considered the sites listed in Tables 3-1 and 3-2 that reported data for at least one year between
26    2005 and 2007 with 75% completeness. As shown in these tables, maximum concentrations as
27    well as 2nd highest 1-hour and 8-hour average CO concentrations generally are quite
28    homogeneous and do not exhibit great variability.  Focusing on Denver and Los Angeles
29    Counties, however, the sites show higher concentrations for the year 2006 in  specific sites within
30    each county.
31          Considering the spatial scale and location of monitoring  sites in these two areas, staff
32    recognizes limitations related to the coverage provided by the available sites, particularly in the
33    Denver area, to support development of a comprehensive population exposure assessment for
34    these urban areas in this time period.  This issue is discussed further in chapter 5.
      ' There were no CO monitoring data reported for Arapahoe and Douglas Counties.
      October, 2009                               3-12                  Draft - Do Not Cite or Quote

-------
2   Table 3-1. Descriptive Statistics for CO Concentrations Measured at Selected Fixed-Site
3             Monitors in the Denver Metropolitan Area for the Years 2005 - 2007.
County
Adams
Boulder
Denver
Jefferson
Site ID
(location)
080013001
(Welby)
080130009
(440 Main St)
080130010
(21 50 28th St.)
080310002
(CAMP)
080310013
(NJE-E)
080310014
(Carriage)
080310019
(Fi rehouse #6)
080590002
Year
2005
2006
2007
2005
2006
2007
2005
2005
2006
2007
2005
2006
2005
2006
2005
2006
2007
2005
2006
Number
of 1 -hour
values
8693
8633
8663
8509
8531
8588
2978
8680
8672
8676
8674
8635
8121
8557
8640
8569
8412
8461
8603
CO concentration (ppm)
1-hour
Maximum
3.4
3.8
3.1
5
3.9
3.8
3.6
4.6
6.4
6
5.3
4.4
3.9
3.9
5.6
9.3
4.2
4.1
3.6
2nd Highest
3.3
3.8
3
4.8
2.8
3.4
3.2
4.3
4.6
5.9
3.6
3.9
3.4
3.5
4.2
5.7
4.1
3.6
3.5
8-hour average
Maximum
2.5
2.6
2.3
2.5
2.2
2.3
2
2.9
3.4
3.2
2.5
2.9
2.3
3
2.4
3.1
2.5
2.1
2
2nd
Highest
2.2
2.5
2.1
2.4
1.8
1.9
1.9
2.5
3.1
2.8
2.4
2.5
2.1
3
2.3
2.6
2.4
2
2
    October, 2009
3-13
Draft - Do Not Cite or Quote

-------
1   Table 3-2. Descriptive Statistics for CO Concentrations Measured at Selected Fixed-Site
2             Monitors in the Los Angeles Metropolitan Area for the Years 2005 - 2007.
County
Los
Angeles
Orange
Site ID
(location)
060370002
(Azusa)
060370113
(West Los
Angeles)
060371002
(Burbank)
060371103
(Los Angeles)
060371201
(Reseda)
060371301
(Lynwood)
060371601
(Pico Rivera)
060371701
(Pomona)
060372005
(Pasadena)
060374002
(Long Beach)
060375005
(Los Angeles)
060376012
(Santa Clara)
060379033
(Lancaster)
060590007
Year
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
Number
of 1 -hour
values
8355
8368
8344
8350
8365
8267
8279
8345
8334
8298
8265
8148
8018
8375
7954
8331
8275
8284
2538
4698
8318
8350
8335
8293
8274
8258
8338
8340
8216
7769
8364
8356
8311
8248
8339
8339
8265
7710
8226
8307
CO concentration (ppm)
1-hour
Maximum
2.5
2.2
2.6
3.4
2.9
2.7
4.4
4.3
3.7
3.9
3.5
3.2
5.1
4.8
3.7
7.4
8.4
7.8
3.3
3.1
4.8
4.2
3.3
3.3
4.3
4.1
3.3
4.2
4.2
3.3
2.8
2.8
3.3
2.2
2.0
1.9
2.9
3.2
2.5
4.1
2nd
Highest
2.3
2.2
2.4
3.1
2.8
2.5
4.0
4.2
3.7
3.4
3.3
3.1
4.9
4.7
3.5
7.2
8.2
7.0
3.2
3.1
3.7
4.0
3.1
3.1
4.3
3.7
3.2
4.0
4.0
3.2
2.7
2.7
3.2
2.0
2.0
1.9
2.5
2.8
2.3
4
8-hour average
Maximum
1.7
1.7
1.8
2.1
2.0
2.0
3.4
3.4
2.8
3.1
2.7
2.2
3.5
3.5
2.8
5.9
6.2
5.3
2.4
2.7
2.9
2.5
2.2
2.0
2.8
2.8
2.3
3.5
3.4
2.6
2.1
2.3
2.4
1.3
1.3
1.2
1.5
1.6
1.3
3.3
2nd
Highest
1.6
1.6
1.7
2.0
1.9
1.6
3.2
3.4
2.7
2.7
2.5
2.1
3.4
3.4
2.7
5.6
5.6
4.9
2.4
2.7
2.8
2.4
2.2
2.0
2.8
2.7
2.2
3.1
3.3
2.4
2.1
2.1
2.1
1.2
1.1
1.2
1.5
1.6
1.2
3.1
    October, 2009
3-14
Draft - Do Not Cite or Quote

-------
County
Riverside
San
Bernardino
Site ID
(location)
(Anaheim)
060591003
(Costa Mesa)
060592022
(Mission Viejo)
060595001
(La Habra)
060651003
(Riverside)
060655001
(Palm Springs)
060658001
(Rubidoux)
060659001
(Lake Elsinore)
060710001
(Barstow)
060710306
(Victorville)
060711004
(Upland)
060719004
(San
Bernardino)
Year
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
2005
2006
2007
Number
of 1 -hour
values
8342
7681
8308
8358
8160
8265
8336
8296
8333
8227
8211
8190
8385
8376
8296
8357
8351
8216
8348
8280
8312
8256
8290
8106
7847
8217
8289
8225
8348
8314
8210
8309
8240
8340
8330
CO concentration (ppm)
1-hour
Maximum
4.5
3.6
4.7
3.5
4.5
2.2
1.9
2.9
6.8
6.0
6.3
4.0
3.8
3.7
2.1
2.3
1.5
3.4
2.7
3.8
1.7
1.4
1.6
3.3
3.5
1.4
2.5
2.2
2.1
2.5
2.7
2.4
3.8
2.8
3.7
2nd
Highest
4.3
3.6
4.1
3.3
4.4
2.2
1.9
2.7
6.7
6.0
6.3
3.7
3.8
3.4
2.0
1.8
1.5
3.3
2.7
3.6
1.5
1.4
1.5
2.2
2.6
1.3
2.1
2.2
2.0
2.4
2.6
2.2
3.3
2.8
3.1
8-hour average
Maximum
2.9
2.9
3.2
3.0
3.1
1.6
1.6
2.2
3.1
2.9
2.9
2.4
2.4
2.2
0.8
0.9
0.8
2.5
2.3
2.9
1.0
1.0
1.4
1.3
1.2
0.7
1.6
1.6
1.6
1.9
1.9
1.7
2.5
2.2
2.3
2nd
Highest
2.9
2.3
3.1
2.5
2.6
1.6
1.3
2.0
3.0
2.9
2.7
2.2
2.1
2.0
0.7
0.8
0.7
2.3
2.1
2.5
1.0
1.0
1.3
1.2
1.1
0.6
1.4
1.5
1.5
1.7
1.8
1.6
2.2
2.0
2.1
October, 2009
3-15
Draft - Do Not Cite or Quote

-------
 1         3.3    KEY OBSERVATIONS
 2          Presented below are key observations resulting from the air quality considerations.
 3

 4          •   Automobiles are the primary contributor to CO emissions, particularly in urban areas
 5              due to greater vehicle and roadway densities.

 6          •   Recent (2005-2007) ambient CO concentrations across the U.S. are lower than those
 7              reported in the previous CO NAAQS review and are also well below the current CO
 8              NAAQS levels. Further, a large proportion of the reported concentrations are below
 9              the conventional instrument detection limit of 1 ppm.

10          •   Ambient CO concentrations are highest at monitors sited closest to roadways (i.e.,
11              microscale and middle scale monitors) and exhibit a diel variation linked to the
12              typical commute times of day, with peaks generally observed during early morning
13              and late afternoon during weekdays.

14          •   The currently available information for CO monitors indicates that siting of
15              microscale and middle scale monitors in the current network is primarily limited to
16              roads where traffic density described for them is moderate (<100,000 AADT),
17              however, factors other than reported AADT (e.g.,  orientation with regard to dense
18              urban roadway networks) can contribute to sites reporting higher CO concentrations.

19          •   Due to the limited number of existing ambient monitors and the monitor site
20              characteristics, it is difficult to fully characterize the current spatial and temporal
21              variability in CO ambient concentrations  across the two urban areas that are the focus
22              for this assessment, Denver and Los Angeles.
23
      October, 2009                              3-16                 Draft - Do Not Cite or Quote

-------
 1          3.4   REFERENCES

 2    Baldauf R; Thoma E; Hays M; Shores R; Kinsey J; Gullett B; Kimbrough S; Isakov V; Long T; Snow R; Khlystov
 3            A; Weinstein J; Chen FL; Seila R; Olson D; Gilmour I; Cho SH; Watkins N; Rowley P; Bang J. (2008).
 4            Traffic and meteorological impacts on near-road air quality: Summary of methods and trends from the
 5            Raleigh near-road study. J Air Waste Manag Assoc, 58: 865-878.

 6    US EPA (2000). Air Quality Criteria for Carbon Monoxide. EPA 600/P-99/00IF. U.S. Environmental Protection
 7            Agency, Research Triangle Park, NC.

 8    US EPA (2009). Integrated Science Assessment for Carbon Monoxide -Second External Review Draft. U.S.
 9            Environmental Protection Agency, Research Triangle Park, NC, report no. EPA 600/R-09/019B. Available
10            at: http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_isa.html.

11    Zhu Y; Hinds WC; Kim S; Shen S; Sioutas C. (2002). Study of ultrafine particles near a major highway with heavy-
12            duty diesel traffic. Atmos Environ, 36: 4323-4335
      October, 2009                                  3-17                    Draft - Do Not Cite or Quote

-------
 1             4. APPROACH TO RISK CHARACTERIZATION FOR
 2                                   CURRENT REVIEW
 3          This section describes the health effects evidence, dose metric of interest and
 4    approach to characterization of risk in support of the current review of the CO primary
 5    NAAQS. Similar to the approach used in prior CO NAAQS reviews, the approach to risk
 6    characterization presented in this section is based on the estimation of CO exposures and
 7    resulting doses  (an internal biomarker) for a defined at-risk population within urban study
 8    areas associated with CO  levels representing recent air quality and air quality adjusted to
 9    simulate just meeting the  current CO NAAQS.
10          Carbon  monoxide can elicit a broad range of effects in multiple tissues and organ
11    systems that are dependent upon concentration and duration of exposure, and that may
12    involve multiple mechanisms including hypoxic stress and others such as free radical
13    production and  the initiation of cell signaling. However, binding of CO to reduced iron in
14    heme proteins with subsequent alteration of heme protein function is the common
15    mechanism underlying the biological responses to CO (draft ISA, section 5.1).  Similarly,
16    based on the health effects evidence summarized in the draft ISA, the best characterized
17    dose metric for estimating exposure to CO associated with adverse health effects is blood
18    carboxyhemoglobin (COHb) levels. As described in the draft ISA, the most compelling
19    evidence of a CO-induced effect on the cardiovascular system at COHb  levels relevant to
20    the current NAAQS comes from a series of controlled human exposure studies among
21    individuals with coronary heart disease (CHD) (draft ISA, section 5.2).  Specifically for
22    the current analysis, we characterize risk for the population of interest (CHD population)
23    by using a potential health effects benchmark level approach,  in combination with short-
24    term CO exposure and dose modeling, to estimate the number and percent of the
25    population with CHD  that would potentially exceed COHb levels of concern, upon just
26    meeting various CO air quality scenarios. Section 4.1 presents a brief summary of the
27    health effects evidence from controlled human exposure studies (draft ISA, section 5.2.4)
28    and section 4.2  describes the rationale for the selection of potential health effects
29    benchmarks and their use in the characterization of risk for adults with CHD.  Section 4.3
30    presents  key observations relevant to the approach for the risk characterization.

31       4.1. CARDIOVASCULAR DISEASE RELATED EFFECTS
32          Controlled human studies provide strong evidence for an association between
33    short-term exposure to CO and exacerbation of preexisting coronary heart disease.
34    Several controlled human exposure studies discussed in the 2000 CO AQCD (section

      October  2009                          4-1            Draft - Do Not Cite or Quote

-------
 1    6.2.2, US EPA, 2000) showed that short-term exposure to CO and subsequent elevation
 2    of COHb levels enhance exercise-induced myocardial ischemia.
 3          Among those studies the draft ISA places emphasis on the work of Allred et al., a
 4    large multi-laboratory study designed to evaluate myocardial ischemia, as documented by
 5    electrocardiogram ST-segment changes and time to onset of angina, during a standard
 6    treadmill test, at CO exposures targeted to result in COHb levels of 2% and 4%. As
 7    described in the draft ISA (draft ISA, section 5.2.4), other controlled human exposure
 8    studies (Adams et al. 1988, Anderson et al. 1973, Kleinman et al 1989, Kleinman et al.,
 9    1998) involving individuals with stable angina have also demonstrated the capacity of
10    CO to decrease the time to onset of angina, as well as to reduce the duration of exercise at
11    COHb concentrations between 3 and 6% (as measured by CO-oximeter). A single study
12    by Sheps et al. (1987) observed no change in time to onset of angina or maximal exercise
13    time following a 1 h exposure to 100 ppm CO (targeted COHb of 4%) among a group of
14    30 patients with CHD. In a subsequent study conducted by the same laboratory, a
15    significant increase in number of ventricular arrhythmias during exercise was observed
16    relative to room air among individuals with CHD following a 1-hr exposure to 200  ppm
17    CO (targeted COHb of 6%), but not following a 1-hr exposure to 100 ppm CO (targeted
18    COHb of 4%) (Sheps et al.,  1990). The draft ISA notes that although the subj ects
19    evaluated in the studies described above are not necessarily representative of the most
20    sensitive population, the level of disease in these individuals was relatively severe, with
21    the majority either having a history of myocardial infarction or having > 70% occlusion
22    of one or more of the coronary arteries.
23          The draft ISA (draft ISA, section 5.2.4) states that no new human clinical studies
24    involving controlled short-term CO exposures among subjects with coronary artery
25    disease have been published since the 2000 CO AQCD.  However, a number of new
26    studies have investigated the effects of CO in healthy adults. Adir et. al., (1999) showed
27    that short-term exposure to CO at concentrations targeted to produce 4-6% COHb,
28    followed by a treadmill test (at maximal exercise capacity), caused a decrease in the
29    duration of exercise and in the metabolic equivalent units (indicative of the oxygen
30    consumed by the body during exercise).  The draft ISA notes that these results  are in
31    agreement with the findings of several studies cited in the 2000 CO AQCD which
32    observed decreases in exercise duration and maximal aerobic capacity among healthy
33    adults at COHb levels > 3% (draft ISA, section 5.2.4), which provides coherence with the
34    observed effects of short-term exposure to CO on exercise-induced myocardial ischemia
35    among patients with CHD.
      October 2009                          4-2            Draft - Do Not Cite or Quote

-------
 1       4.2. HEALTH EFFECT BENCHMARKS
 2          As in the review completed in 1994 and in the CO exposure/dose assessment
 3    completed in 2000 (section 6.3.2), a health effect benchmark level approach is used in the
 4    current analysis to estimate the number and percent of the population with CHD that
 5    would potentially exceed COHb levels of concern for specific CO air quality scenarios.
 6    Since the ISA has not identified new studies that demonstrate CO effects at COHb levels
 7    lower than those described in the 2000 AQCD, we are relying in the same studies as we
 8    did on the review completed in 1994. As mentioned above, a number of studies,
 9    described in detail in the 2000 CO AQCD (section 6.2.2, US EPA, 2000), showed
10    statistically significant group mean responses, measured in terms of reduced time to onset
11    of exercise-induced angina, in the range of 3 to 6 %COHb (measured by CO-oximeter) in
12    subjects with coronary heart disease. We note that the lowest COHb level at which
13    reduced time to onset of angina was observed was in the range of 2.0 to 2.4% COHb
14    (measured by gas chromatography), in a multi-center CO exposure study (Allred et al.,
15    1989a,b, 1991; draft ISA, section 5.2.6).  This range (2.0-2.4%) is representative of the
16    two individual COHb level averages obtained post-exposure (2.4%) and post-exercise
17    test (2.0%).  However, there was  no clear pattern across the different studies with respect
18    to the magnitude of the decreased time to onset of angina versus dose level. In addition,
19    these studies do not address the fraction of the population experiencing a specified health
20    effect at various dose levels. Thus, based on information in the draft ISA, staff
21    concluded that at this time there is insufficient controlled human exposure data to support
22    the development of quantitative dose-response relationships which would be required in
23    order to conduct a quantitative risk assessment for this health endpoint.

24          Potential health effect benchmark values used in the risk characterization linked
25    to the exposure/dose analyses were derived solely based on the controlled human
26    exposure literature.  This is primarily because CO concentrations reported in controlled
27    human exposure studies represent actual  personal exposures rather than concentrations
28    measured at fixed site ambient monitors.  In addition, controlled human exposure studies
29    can examine the health effects of short-term exposure to CO in the absence of co-
30    pollutants that can confound results in epidemiologic analyses; thus, health effects
31    observed in controlled human exposure studies can confidently be attributed to a defined
32    COHb dose level associated  with ambient short-term CO exposures.
33          In identifying the potential health effect benchmark levels for the risk
34    characterization, staff considered a number of factors in drawing on the results of
35    controlled human exposure studies. As noted above, the lowest group mean COHb level

      October 2009                          4-3             Draft - Do Not Cite or Quote

-------
 1    at which reduced time to onset of angina was observed was in the range of 2.0 to 2.4%
 2    COHb (measured by gas chromatography(GC)) in a multi-center CO exposure study
 3    (Allred et al., 1989a,b,  1991; draft ISA, section 5.2.6 and 2.5.1). Similar effects have not
 4    been evaluated below this range.
 5           Staff identified  potential health effects benchmarks of 1.5%, 2.0%, 2.5% and 3%
 6    COHb levels based on the consideration of the studies reporting adverse effects at COHb
 7    levels as low as 2 to 2.4% (using GC) discussed above. These levels reflect comments
 8    from the CASAC CO panel on the draft Analysis Plan (Brain and Samet, 2009) and
 9    include the range of levels considered in the review completed in 1994 (US EPA, 1992).
10    The potential health effects benchmarks extend lower than the range where controlled
11    human exposure studies reported CO-related health effects to take into consideration both
12    the uncertainty about the actual COHb levels experienced in the controlled human
13    exposure studies due to the use of different measurement methods and that these studies
14    did not include individuals with more severe cardiovascular disease who may respond at
15    lower COHb levels relative to the subjects tested.

16
17       4.3. KEY OBSERVATIONS
18    Presented below are key observations relevant to the risk characterization approach in
19    support of the current CO NAAQS review.
20
21          •  An important at-risk population for short-term exposure to is the adult CHD
22             population.
23          •  The data from controlled human  studies do not support the development of a
24             quantitative risk assessment due to lack of sufficient information to
25             characterize the dose-response relationship within the range of interest.
26             Instead, risk will be characterized in the current assessment using a potential
27             health effect benchmark levels approach (as in previous assessments).
28          •  Evaluation of health effects evidence reported in controlled human exposure
29             studies of short-term CO exposure in the adult CHD population is the basis for
30             the selection of potential  health effects benchmarks of 1.5, 2.0, 2.5, and 3.0%
31             COHb on the consideration of studies reporting adverse effects at COHb
32             levels in the range of 2 to 2.4%.
      October 2009                          4-4             Draft - Do Not Cite or Quote

-------
  1        4.4. REFERENCES

  2
  3    Adams, K.F.; Koch, G.; Chatterjee, B.; Goldstein, G.M.; O'Neil, J.J.; Bromberg, P.A.; Sheps, D.S.;
  4            McAllister, S.; Price, C.J.; Bissette, J. (1988) Acute elevation of blood carboxyhemoglobin to 6%
  5            impairs exercise performance and aggravates symptoms in patients with ischemic heart disease. J.
  6            Am. Coll. Cardiol. 12: 900-909

  7    Adir Y, Merdler A, Ben Haim S, Front A, Harduf R & Bitterman H ( 1999) Effects of Exposure to Low
  8            Concentrations of Carbon Monoxide on Exercise Performance and Myocardial Perfusion in
  9            Young Healthy Men. Occup Environm Med, 56: 535-538.

10    Allred, E.N.; Bleecker, E.R.; Chaitman, B.R.; Dahms, T.E.; Gottlieb, S.O.; Hackney, J.D.; Pagano, M.;
11            Selvester, R.H.; Walden,  S.M.; Warren, J. (1989a) Short-term effects of carbon monoxide
12            exposure on the exercise performance of subjects with coronary artery disease. N. Engl.  J. Med.
13            321: 1426-1432.

14    Allred, E.N.; Bleecker, E.R.; Chaitman, B.R.; Dahms, T.E.; Gottlieb, S.O.; Hackney, J.D.; Hayes, D.;
15            Pagano, M.; Selvester, R.H.; Walden, S.M.; Warren, J. (1989b) Acute effects of carbon monoxide
16            exposure on individuals with coronary artery disease. Cambridge, MA: Health Effects Institute;
17            research report no. 25.

18    Anderson EW, Andelman RJ, Strauch JM, Fortuin NJ and Knelson, JH (1973). Effect of low level carbon
19            monoxide exposure on onset and duration of angina pectoris. Annals of Internal Medicine 79: 46-
20            50.

21    Johnson T, Mihlan G, LaPointe J, Fletcher K, Capel J. (2000). Estimation of Carbon Monoxide Exposures
22            and Associated Carboxyhemoglobin levels in Denver Residents Using pNEM/CO (Version 2.1),
23            prepared by ICF Kaiser Consulting Group for U.S. EPA, Office of Air Quality Planning and
24            Standards, under Contract No. 68-D6-0064, WA Nos. 1-19,  2-24,  2-30, and 3-3, Docket EPA-HQ-
25            OAR-2008-0015-DRAFT-0009.

26    US EPA. (1991). Air Quality Criteria for Carbon Monoxide. Research Triangle Park, NC: Office of Health
27            and Environmental Assessment, Environmental Criteria and Assessment Office; report no.
28            EPA/600/8-90/045F.

29    US EPA (2000). Air Quality Criteria for Carbon Monoxide. National Center for Environmental
30            Assessment, Office of Research and Development, U.S. Environmental Protection Agency,
31            Research Triangle Park, NC 27711; report no. EPA/600/P-99/00IF. June 2000. Available:
32            http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=18163.

33    Sheps, D.S.; Herbst, M.C.; Hinderliter, A.L.; Adams, K.F.; Ekelund, L.G.;  O'Neil, J.J.; Goldstein, G.M.;
34            Bromberg, P.A.; Dalton, J.L.; Ballenger, M.N.; Davis,  S.M.; Koch, G.G. (1990) Production of
3 5            arrythmias by elevated carboxyhemoglobin in patients with coronary artery disease.  Ann. Intern.
36            Med. 113: 343-351.

37    Kleinman M T, Davidson DM, Vandagriff RB, Caiozzo VJ, Whittenberger JL.(1989). Effects of short-term
38            exposure to carbon monoxide in subjects with coronary artery disease. Arch. Environ. Health 44:
39            361-369.

40    Kleinman MT, Leaf DA, Kelly E,  Caiozzo V, Osann K, O'Niell T. (1998).  Urban angina in the mountains:
41            effects of carbon monoxide and mild hypoxemia on subjects with chronic stable angina. Arch.
42            EnviroaHealth 53: 388-397.
       October 2009                              4-5              Draft - Do Not Cite or Quote

-------
1    US EPA (2009). Integrated Science Assessment for Carbon Monoxide - Second External Review Draft.
2            U.S. Environmental Protection Agency, Research Triangle Park, NC, report no. EPA/600/R-
3            09/019B. Available at: http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_isa.html.
     October 2009                              4-6              Draft - Do Not Cite or Quote

-------
 1            5.  APPROACH TO EXPOSURE AND DOSE ASSESSMENT FOR
 2                                        CURRENT REVIEW

 3          This chapter presents an overview and description of the overall approach to estimating
 4    human exposure and dose for past and recent assessments and presents a strategy for the current
 5    exposure and dose assessments for the CO NAAQS review. Section 5.1 provides a brief
 6    overview of the exposure model, followed by a short history that explains the evolution of the
 7    exposure and dose models used in NAAQS reviews in section 5.2.  Section 5.3 provides a
 8    description of the exposure and dose models that have been used by OAQPS to conduct prior
 9    exposure and dose assessments for CO NAAQS and which serve as the basic modeling tools to
10    be used for the current assessment. Section 5.4 briefly summarizes information about personal
11    exposure and key microenvironments for CO.  Section 5.5 discusses the current monitoring
12    network in the two study areas selected for the current assessment and the limitations of the
13    existing monitoring network for purposes of conducting an exposure and dose assessment in
14    these areas. This final section also presents the strategy for the  current exposure and dose
15    analysis.

16         5.1  MODEL OVERVIEW
17          The Air Pollutants Exposure model (APEX) is a personal computer (PC)-based program
18    designed to estimate human exposure to criteria and air toxic pollutants at the local, urban, and
19    consolidated metropolitan levels.   APEX, also known as TREVI.Expo, is the human inhalation
20    exposure module of EPA's Total Risk Integrated Methodology  (TRIM) model framework (US
21    EPA, 1999), a modeling system with multimedia capabilities for assessing human health and
22    ecological  risks from hazardous and criteria air pollutants.l
23          APEX estimates human exposure to criteria and
24    toxic air pollutants at the local, urban, or consolidated
25    metropolitan area levels using a stochastic,
„,    «  .       .        . ,„        i   T,,     i i     i   i        with an environmental pollutant takes place
26    microenvironmental  approach. The model randomly
                                   ,.,.,..,,„        and which can be treated as a well
27    selects data tor a sample or hypothetical individuals trom
                                                             characterized, relatively homogeneous
28    an actual population database and simulates each            ...    ...      ,,   .. ,  ,
              ^ ^                                            location with respect to pollutant
                                                             concentrations for a specified time period.
29    hypothetical individual's movements through time and
30    space (e.g., indoors at home, inside vehicles) to estimate
                                                             A microenvironment is a three-
                                                             dimensional space in which human contact
      1 Additional information on the TRIM modeling system, as well as downloads of the APEX Model, user guides
      (U.S. EPA 2008a, 2008b), and other supporting documentation, can be found at http://www.epa.gov/ttn/fera.
      October, 2009                          5-1            Draft - Do Not Cite or Quote

-------
 1   his or her exposure to a pollutant. APEX can account for travel to and from work locations (i.e.,
 2   commuting) and provide estimates of exposures at both home and work locations for individuals
 3   who work away from home.

 4         5.2   MODEL HISTORY AND EVOLUTION
 5          APEX was derived from the National Ambient Air Quality Standards (NAAQS)
 6   Exposure Model (NEM) series of models. The NEM series was developed to estimate
 7   population exposures to the criteria pollutants (e.g., CO, ozone). In 1988, OAQPS first
 8   incorporated probabilistic elements into the NEM methodology and used activity pattern data
 9   based on available human activity diary studies to create an early version of probabilistic NEM
10   for ozone (i.e., pNEM/O3). In 1991, a probabilistic version of NEM was developed for CO
11   (pNEM/CO) that included a one-compartment mass-balance model to estimate CO
12   concentrations in indoor microenvironments. The application of this model to Denver, Colorado
13   is summarized in Johnson et al. (1992).  Between 1999 and 2001, updated versions of pNEM/CO
14   (versions 2.0 and 2.1) were developed that rely on detailed activity diary data compiled in EPA's
15   Consolidated Human Activities Database (CHAD) (McCurdy et al., 2000; US EPA,  2002) and
16   enhanced algorithms for simulating gas stove usage, estimating alveolar ventilation rate (a
17   measure of human respiration), and modeling home-to-work commuting patterns. A draft report
18   by Johnson et al. (2000) describes the application of Version 2.1 of pNEM/CO to Denver and
19   Los Angeles.
20          The first version of APEX was essentially identical to pNEM/CO (version 2.0) except
21   that it ran on a PC instead of a mainframe. The next version, APEX2, was substantially
22   different, particularly in the use of a personal profile approach rather than a cohort simulation
23   approach. APEX3 introduced a number of new features including automatic site selection from
24   national databases, a series of new output tables providing summary exposure and dose statistics,
25   and a thoroughly reorganized method of describing microenvironments and their parameters.
26   Johnson and Capel (2003) describe a case study in which Version 3.1 of APEX was used to
27   estimate population exposure to CO in Denver and Los Angeles.
28          The current version of APEX (Version 4.3) (US EPA, 2008a; 2008b) was used to
29   estimate CO exposure and dose as described in chapter 6 of this document. This version was
30   also recently used to estimate ozone (O3) exposures in 12 urban areas for the O3 NAAQS review
31   (US EPA, 2007), in estimating population exposures to nitrogen dioxide (NO2) in Atlanta as part
32   of the NO2 NAAQS review (EPA, 2008c), and in estimating sulfur dioxide (SO2) exposures for
33   asthmatics and asthmatic children in two study areas in Missouri as part of the SO2 NAAQS
34   review (US EPA, 2009a). There have been several recent enhancements to APEX since the prior
35   1994 CO NAAQS review, including:

     October, 2009                         5-2            Draft - Do Not Cite or Quote

-------
 1       •  Algorithms for the assembly of multi-day (longitudinal) activity diaries that model intra-
 2          individual variance, inter-individual variance, and day-to-day autocorrelation in diary
 3          properties;
 4       •  Methods for adjusting diary-based energy expenditures for fatigue and excess post-
 5          exercise oxygen (EPOC) consumption;
 6       •  New equations for estimation of ventilation (i.e., breathing rate);
 7       •  The ability to model commuters leaving the study  area;
 8       •  The ability to model air quality and exposure for flexible time scales;
 9       •  New output files containing diary event-level, time-step level, and hourly-level exposure,
10          dose, and ventilation data, and hourly-level microenvironmental data;
11       •  The ability to model the prevalence of disease states such as asthma or coronary heart
12          disease (CHD);
13       •  New output exposure tables that report exposure statistics for population groups and
14          lifestages such as children and active people under different ventilation levels;
15       •  The inclusion of commuting data from the 2000 census; and
16       •  Expanded options for modeling microenvironments.
17          As discussed below in section 5.3, due to limitations in the CO ambient monitoring data,
18    the current exposure/dose assessment does not take advantage of a number of the recent
19    advances listed above.

20         5.3   MODEL SIMULATION PROCESS
21          APEX is designed to simulate population exposure to criteria and air toxic pollutants at
22    local, urban, and regional scales.  The user specifies the geographic area to be modeled and the
23    number of individuals  to be simulated to represent this population.  APEX then generates a
24    personal profile for each simulated person that specifies various parameter values required by the
25    model. The model next uses diary-derived time/activity data matched to each personal profile to
26    generate an exposure event sequence (also referred to as "activity pattern" or "composite diary")
27    for the modeled individual that spans a specified time period, such as a calendar year. Each
28    event in the sequence specifies a start time, exposure duration, a geographic location, a
29    microenvironment, and an activity. Probabilistic algorithms are used to estimate the pollutant
30    concentration and ventilation (respiration) rate associated with each exposure event. The
31    estimated pollutant concentrations account for the effects  of ambient (outdoor) pollutant
32    concentration, penetration factor, air exchange rate, decay/deposition rate, and proximity to
33    emission sources, depending on the microenvironment, available data, and the estimation method
34    selected by the user.  The ventilation rate is derived from an energy expenditure rate estimated
35    for each individual and specified activity  performed.  Because the modeled  individuals represent

      October, 2009                           5-3            Draft - Do Not Cite or Quote

-------
 1    a random sample of the population of interest, the distribution of modeled individual exposures
 2    can then be extrapolated to the larger population of interest.
 3          The model simulation generally includes up to seven steps.

 4         1.   Characterize the study area: APEX selects sectors (e.g., census tracts) within a
 5             study area—and thus identifies the potentially exposed population — usually based on
 6             the user-defined center and radius of the study area and availability of air quality and
 7             weather input data for the area (section 5.3.1).
 8         2.   Generate simulated individuals: APEX stochastically generates a sample of
 9             simulated individuals based on the census data for the study area and human profile
10             distribution data (such as age-specific employment probabilities), (section 5.3.2)
11         3.   Construct activity sequences:  APEX constructs an exposure event sequence (activity
12             pattern) spanning the simulation period for each of the simulated persons based on the
13             CHAD activity pattern data (section 5.3.3).
14         4.   Calculate microenvironmental concentrations: APEX enables the user to define
15             microenvironments that people in a study area would visit (e.g., by grouping location
16             codes included in the activity pattern database). The model then calculates time-
17             averaged concentrations (e.g., hourly) of each pollutant in each of the
18             microenvironments for each  simulated person for the period of simulation, based on the
19             user-provided ambient air quality data (section 5.3.4).
20         5.   Estimate energy expenditure and ventilation rates: APEX constructs a time-series
21             of energy expenditures for each profile based on the activity event sequence. These
22             expenditures are adjusted to  ensure that they are physiologically realistic, and then
23             used to estimate a number of ventilation metrics that are later used in estimating dose
24             (section 5.3.5).
25         6.   Calculate exposure: APEX assigns a concentration to each exposure event based on
26             the microenvironment occupied during the  event and the person's activity. These
27             values are time-averaged (e.g., hourly) to produce a sequence  of exposures spanning
28             the specified exposure period (typically one year).  The hourly values may be further
29             aggregated to produce 8-hour, daily, monthly, and annual average exposure values
30             (section 5.3.6).
31         7.   Calculate dose: APEX optionally calculates hourly, daily, monthly, and annual
32             average dose values for each of the simulated individuals.  For the application of
33             APEX to CO, a module within the model estimates the percent COHb level in the
34             blood at the end of each hour based on the time-series of CO concentrations and
35             alveolar ventilation rates experienced by the simulated person (section 5.3.7).
36          The model simulation continues until exposures (and associated COHb levels) are
37    calculated for the user-specified number of simulated individuals.  Figure A-l in Appendix A
38    presents a conceptual model and simplified data flow of APEX used in this assessment.   The
39    following sections provide additional details on the general procedures and algorithms used in
      October, 2009                          5-4             Draft - Do Not Cite or Quote

-------
 1    each of the simulation steps listed above.  The specific inputs and algorithms used in applying
 2    APEX to CO for the current assessment are further described in section 6.1

 3         5.3.1  Characterize the Study Area
 4           An initial study area in an APEX analysis consists of a set of basic geographic units
 5    called sectors, typically defined as census tracts. The user may provide the geographic center
 6    (latitude/longitude) and radius of the study area and then APEX calculates the distances to the
 7    center of the study area of all the sectors included in the sector location database, and finally
 8    selects the sectors within the radius of the study area.  APEX then maps the user-provided air
 9    quality and meteorological data for specified monitoring districts to the selected sectors.  The
10    sectors identified as having acceptable air quality and meteorological data within the radius of
11    the study area are selected to comprise a final study area for the APEX simulation analysis. This
12    final study area determines the population make-up of the simulated persons (profiles) to be
13    modeled.

14         5.3.2  Generate Simulated Individuals
15           APEX stochastically generates a user-specified number of simulated (hypothetical)
16    persons to represent the population in the study area.  Each simulated person is represented by a
17    "personal profile." APEX generates the simulated person or profile by probabilistically selecting
18    values for a set of profile variables. The profile variables include:
19       •   Demographic variables (e.g., age, gender, home sector, work sector) that are generated
20           based on the census data;
21       •   Residential variables (e.g., air conditioning prevalence) which are generated based on
22           sets of distribution data;
23       •   Physiological variables (e.g., blood volume, body mass, resting metabolic rate) that are
24           generated based on age- and gender-specific distribution data; and
25       •   Daily varying variables (e.g., daily work status) which are generated based on
26           distribution data that change daily during the simulation period.
27           APEX first selects and calculates demographic, residential,  and physiological variables
28    (except for daily values) for each of the  specified number of simulated individuals. APEX then
29    follows each simulated individual over time and calculates exposures (and optionally doses) for
30    the individual over the specified time period. The profile variables are listed and described in
31    detail in section 5 of US EPA (2008b).

32         5.3.3  Construct Activity Sequences
33           APEX probabilistically creates a composite diary for each of the simulated persons by
34    selecting a 24-hour diary record - or diary day - from an activity database for each day of the

      October, 2009                           5-5             Draft - Do Not Cite or Quote

-------
 1    simulation period. (CHAD data are supplied with APEX for this purpose.)  A composite diary is
 2    a sequence of events that simulates the movement of a modeled person through geographical
 3    locations and microenvironments during the simulation period.  Each event is defined by
 4    geographic location, start time, duration, microenvironment visited, and an activity performed.
 5          The activity database input to APEX contains the following information for each diary
 6    day: age, gender, employment status, occupation, day-of-week (day-type), and maximum hourly
 7    average temperature.  This information enables APEX to select data from the activity database
 8    that tend to match the characteristics of the simulated person, the study area, and the specified
 9    time period.  APEX develops a composite diary for each of the simulated individuals according
10    to the following steps.
11         1.  Divide diary days in the CHAD database into user-defined activity pools, based on
12            day-type and temperature.
13         2.  Assign an activity  pool number to each day of the simulation period, based on the user-
14            provided daily maximum/average temperature data.
15         3.  Calculate a selection probability for each of the diary days in each of the activity pools,
16            based on age/gender/employment similarity of a simulated person to a diary day.
17         4.  Probabilistically select a diary day from available diary days in the activity pool
18            assigned to each day of the simulation period.
19         5.  Estimate a metabolic value for each activity performed while in a  CHAD location,
20            based on the activity-specific metabolic distribution data. This value is used to
21            calculate a ventilation rate for the simulated person performing the activity.
22         6.  Map the CHAD locations in the selected diary to the user-defined modeled
23            microenvironments.
24         7.  Concatenate the selected diary days into a sequential longitudinal  diary for a simulated
25            individual covering all days  in the simulated period.
26          APEX provides an optional longitudinal diary assembly algorithm that enables the user to
27    create composite diaries that reflect the tendency of individuals to repeat activities on a day-to-
28    day basis.  The user specifies values for two statistical variables (i.e., D and A) that relate to a
29    key daily variable, typically  the time spent per day in a particular microenvironment (e.g., in a
30    motor vehicle).  The D statistic reflects the relative importance of within person variance and
31    between person variance in the key variable.  The A statistic quantifies the lag-one (day-to-day)
32    variable autocorrelation.  APEX then constructs composite diaries that exhibit the  statistical
33    properties defined by the specified values ofD and A. The longitudinal diary assembly
34    algorithm is described in greater detail in section  6.3 of US EPA (2008b).
      October, 2009                          5-6            Draft - Do Not Cite or Quote

-------
 1         5.3.4   Calculate Microenvironmental Concentrations
 2          Probabilistic algorithms are used to estimate the pollutant concentration associated with
 3    each exposure event.  The estimated pollutant concentrations account for the effects of ambient
 4    (outdoor) pollutant concentration, penetration factor, air exchange rate, decay/deposition rate,
 5    and proximity to emission sources,  depending on the microenvironment, available data, and the
 6    estimation method selected by the user.
 7          APEX calculates air concentrations in the various microenvironments visited by the
 8    simulated person by using the ambient air data for the relevant census tracts and the user-
 9    specified method and parameters that are specific to each microenvironment. In typical
10    applications, APEX calculates hourly concentrations in all the microenvironments at each hour
11    of the simulation for each of the simulated individuals, based on the hourly ambient air quality
12    data specific to the geographic locations visited by the individual. APEX provides two methods
13    for calculating microenvironmental concentrations: the mass balance method and the transfer
14    factors method (described briefly below).  The user is required to specify a calculation method
15    for each of the microenvironments; there are no restrictions on the method specified for each
16    microenvironment (e.g., some microenvironments can use the transfer factors method while the
17    others can use the mass balance method).  As discussed in section 5.4, the current draft
18    assessment employed a simplified approach relying on the factors model  approach with
19    particular focus on the in-vehicle microenvironment.
20          Mass Balance Model
21          The mass balance method models an enclosed microenvironment  as a well-mixed volume
22    in which the  air concentration is assumed to be spatially uniform at any specific time. The
23    concentration of an air pollutant in  such a microenvironment is estimated using the following
24    four processes:
25       •  Inflow of air into the microenvironment;
26       •  Outflow of air from the microenvironment;
27       •  Removal of a pollutant from the microenvironment due to deposition, filtration, and/or
28          chemical degradation; and
29       •  Emissions from sources of a pollutant inside the microenvironment (if indoor sources are
30          modeled).
31          The mass balance model feature of APEX (see Appendix B) has not been used in the
32    APEX application for CO described in this draft REA.
33
3 4          Factors Model
      October, 2009                           5-7             Draft - Do Not Cite or Quote

-------
 1           The factors model approach is conceptually simpler than the mass balance method and
 2    has fewer user-specified parameters. It estimates the concentration in a microenvironment as a
 3    linear function of ambient concentration of that hour, regardless of the concentration in the
 4    microenvironment during the preceding hour.  Table 5-1 lists the parameters required by the
 5    factors model approach to calculate concentrations in a microenvironment without emissions
 6    sources.

 7    Table 5-1.  Parameters  of the Factors Model
 9
10
11
12

13
14
15
16
17
18
19
20
21
22
23
24
25
26
Variable
J proximity
J penetration
Definition
Proximity factor
Penetration factor
Units
N/A
N/A
Value Range
J proximity ^
^ — J penetration — ^
       The factors model approach uses the following equation to calculate hourly mean
concentration in a microenvironment from the user-provided hourly air quality data:
                            /••f hourlymean
                              ! = C
                     xf
                      J i
                                  ambient   J proximity   J penetration
J n<
                                                                           (5-1)
       where:
       C
-r hourlymean
 ME
       ^ambient

      Jproximity

      Jpenetration
Hourly concentration in a microenvironment (ppm)
Hourly concentration in ambient environment (ppm)
Proximity factor (unitless)
Penetration factor (unitless)
       The proximity factor (/proximity) is used to account for differences in ambient
concentrations between the geographic location represented by the ambient air quality data (e.g.,
a regional fixed-site monitor) and the geographic location of the particular microenvironment.
For example, a residence might be located near a heavily-trafficked roadway, whereby the
ambient air outside the house would likely have elevated levels of mobile source pollutants such
as carbon monoxide. In this case, a value greater than one for the proximity factor would be
appropriate to represent the increase in concentrations outside the home relative to the ambient
monitor.  Additionally, for some pollutants the process of infiltration may remove a fraction of
the pollutant from the air.  The fraction that is retained in the indoor air is given by the
      October, 2009
                                                       Draft - Do Not Cite or Quote

-------
 1    penetration factor (/penetration) and is dependent on the particular pollutant's physical and chemical
 2    removal rates.  Typically, the value of the penetration factor ranges from 0 to 1.

 3         5.3.5  Estimate Energy Expenditure and Ventilation Rates
 4           APEX includes a module that estimates COHb levels in the blood as a function of
 5    alveolar ventilation rate, the CO concentration of the respired air, endogenous CO production
 6    rate, and various physiological variables such as blood volume and pulmonary CO diffusion rate.
 7    Alveolar ventilation rate is estimated as a function of oxygen uptake rate, which in turn is
 8    estimated as a function of energy expenditure rate. This section provides a brief summary of the
 9    algorithm used to estimate  alveolar ventilation rate. A detailed description of the algorithm,
10    together with the distributions and estimating equations used in determining the value of each
11    parameter in the algorithm, can be found in Appendix C.
12
13           Energy Expenditure
14           McCurdy (2000) has recommended that measures of human ventilation (respiration) rate
15    be estimated as functions of energy expenditure rate.  The energy expended by an individual
16    during  a particular activity  can be expressed as
17                                      EE = (METS) x (RMR)                    (5-2)
18           in which EE is the average energy expenditure rate (kcal min"1) during the activity and
19    RMR is the resting metabolic rate of the individual expressed in terms of number of energy units
20    expended per unit of time (kcal min"1).  METS (i.e., metabolic equivalent of work) is a ratio
21    specific to the activity and  is dimensionless.
22           The METS concept provides a means for  estimating the alveolar ventilation rate
23    associated with each activity. For convenience, let EE(i,j,k) indicate the energy expenditure rate
24    associated with the ith activity on dayy for person k. Equation 5-2 can now be expressed as
25                               EE(i,j,k) = [METS(/j,£)]  x [RMR(£)]              (5 3)
26           in which RMR(&) is the average value for resting metabolic rate specific to person k.
27    Note that METS(/j,&) is specific to a particular activity performed by person k.
28
29           Oxygen Requirements for Energy Expenditure
30           Energy expenditure requires oxygen which is supplied by  ventilation (respiration).
31    ECF(&) represents an  energy conversion factor defined as the volume of oxygen required to
32    produce one kilocalorie of  energy in person k. The oxygen uptake rate (VO2) associated with a
33    particular activity can be expressed as
34                               V02(i,j,k) = [ECF(£)] x [EE(iJ,k)l         (5 4)
      October, 2009                          5-9            Draft - Do Not Cite or Quote

-------
 1          in which VO2(i,j,k) has units of liters oxygen min"1, ECF(k) has units of liters oxygen
 2    kcal"1, and EE(iJ,k) has units of kcal min"1. The value ofVO2(i,j,k) can now be determined from
 3    MET(/j,&) by substituting Equation 5-3 into Equation 5-4 to produce the relationship
 4                        VO2(iJ,K) = \ECF(K)] x [METS(/j,£)] x [RMR(£)].(5-5)
 5
 6          Excess Post-Exercise Oxygen Consumption
 1          At the beginning of exercise, there is a lag between work expended and oxygen
 8    consumption. During this work/ventilation mismatch, an individual's energy needs are met by
 9    anaerobic processes. The magnitude of the mismatch between expenditure and consumption is
10    termed the oxygen deficit. During heavy exercise, further oxygen deficit (in addition to that
11    associated with the start of exercise) may be accumulated.  At some point, oxygen deficit reaches
12    a maximum value, and performance and energy expenditure deteriorate. After exercise ceases,
13    ventilation and oxygen consumption will remain elevated above baseline levels.  This increased
14    oxygen consumption was historically labeled the "oxygen debt" or "recovery oxygen
15    consumption." However, the term "excess post-exercise oxygen consumption" (EPOC) has been
16    adopted for this phenomenon.  APEX has an algorithm for adjusting  the MET values to account
17    for EPOC. This algorithm is described in detail in section 7.2 of US  EPA  (2008b).
18
19          Alveolar Ventilation Rate
20          Alveolar ventilation (VA) represents the portion of the minute ventilation that is involved
21    in gaseous exchange with the blood. VO2 is the oxygen uptake that occurs during this exchange.
22    The absolute value of VA is known to be affected by total lung volume, lung dead space,  and
23    respiration frequency - parameters which vary according to person and/or exercise rate.
24    However, it is reasonable to assume that the ratio of VA to VO2 is relatively constant regardless
25    of a person's physiological characteristics or energy expenditure rate. Consistent with this
26    assumption, APEX converts each estimate of VO2(/',y,&) to an estimate of VA(/J,£) by the
27    proportional relationship
28                              VA.(JJ,K) = (19.63) x [VO2(/j,£)]                  (5-6)
29          in which both VA and VO2 are expressed in units of liters min"1.  This relationship was
30    obtained from Joumard et al. (1981), who based it on research by Galetti (1959).  Equation 5-6
31    can also be expressed by the equivalent equation
32                 VA.(iJ,K) = (19.63) x [METS(/j,£)] x [ECF(£)] x  [RMR(£)].       (5-7)
33          If ECF and RMR are specified for an individual, then Equation 5-7 requires only  an
34    activity-specific estimate of METS to produce an estimate of the energy expenditure rate for a
35    given activity.  APEX processes time/activity data obtained from the CHAD to create a sequence
      October, 2009                          5-10            Draft - Do Not Cite or Quote

-------
 1    of activity-specific METS values for each simulated individual.  APEX estimates RMR as a
 2    function of body mass based on probabilistic equations specific to age and gender using
 3    equations reported by Schofield (1985). A value of ECF is selected for each individual from a
 4    uniform distribution (minimum = 0.20, maximum = 0.21) based on data provided by Esmail et
 5    al. (1995). Using Equation 5-7 and these inputs, APEX calculates a sequence of VA values for
 6    each simulated individual. These values are provided to the algorithm that estimates the percent
 7    COHb in the blood resulting from the simulated exposure (see section 5.3.7 and Appendix C).

 8         5.3.6  Calculate Exposure
 9          APEX calculates exposure as a time series of exposure concentrations that a simulated
10    individual experiences  during the simulation period.  APEX determines the exposure using
1 1    hourly ambient air concentrations, calculated concentrations in each microenvironment based on
12    these ambient air concentrations, and the minutes  spent in a sequence of microenvironments
13    visited according to the composite diary. The hourly exposure concentration at any clock hour
14    during the simulation period is determined using the following equation:
                          N
                             Z/-< hourly-mean   <
                             ^ME(j)
                   r^    7=1
15                 Ct = - - -              (5-8)

16          where:
17          C; •     =     Hourly exposure concentration at clock hour /' of the simulation period
18                        (ppm)
19          N     =     Number of events (i.e., varied microenvironments visited/activities
20                        performed) in clock hour / of the simulation period.
21          C^j^ean =   Hourly mean concentration in microenvironmenty (ppm)
22          i(j)     =     Time spent in microenvironmenty (minutes)
23          T     =     60 minutes
24          From the hourly  exposures, APEX calculates time series of 8-hour and daily average
25    exposure concentrations that a simulated individual would experience during the simulation
26    period. APEX then statistically summarizes and tabulates the hourly,  8-hour, and daily
27    exposures in a series of output tables.

28         5.3.7   Calculate Dose
29          Using time/activity data obtained from several diary studies, APEX constructs a
30    composite diary for each simulated person in the specified at-risk population.  The composite
3 1    diary consists of a sequence of events spanning the specified period of the exposure assessment
      October, 2009                          5-11           Draft - Do Not Cite or Quote

-------
 1    (typically one calendar year).  Each event is defined by a start time, duration, a geographic
 2    location, a microenvironment, and an activity.  Using the algorithms described above in sections
 3    5.3.5 and 5.3.6, APEX provides estimates of CO concentration and alveolar ventilation rate for
 4    each event in the composite diary, for each simulated individual. APEX then uses these data,
 5    together with estimates of various physiological parameters specific to the simulated individual,
 6    to estimate the percent COHb in the blood at the end of each event. The percent COHb
 7    calculation is based on the solution to the non-linear Coburn, Forster, Kane (CFK) equation, as
 8    detailed in Appendix C.
 9          Briefly, the CFK model describes the rate of change in COHb blood levels as  a function
10    of the following quantities:
11       •  Inspired CO pressure;
12       •  COHb level;
13       •  Oxyhemoglobin (O2Hb) level;
14       •  Hemoglobin (Hb) content of blood;
15       •  Blood volume;
16       •  Alveolar ventilation rate;
17       •  Endogenous CO production rate;
18       •  Mean pulmonary capillary oxygen pressure;
19       •  Pulmonary diffusion rate of CO;
20       •  Haldane coefficient (M);
21       •  Barometric pressure; and
22       •  Vapor pressure of water at body temperature (47 torr).
23          If all of the listed quantities except COHb level are constant over some time interval, the
24    CFK equation has a linear form over the interval and is readily integrated. The solution to the
25    linear form gives reasonably accurate results for lower levels of COHb. However, CO and
26    oxygen compete for the available hemoglobin and are, therefore, not independent of each other.
27    If this dependency is taken into account, the resulting differential equation is no longer linear.
28    Peterson and Stewart (1975) proposed a heuristic approach to account for this dependency which
29    assumed the linear form and then adjusted the O2Hb level iteratively based on the assumption of
30    a linear relationship between COHb and O2Hb.  This approach was used in the COHb module of
31    the original CO-NEM exposure model (Biller and Richmond, 1982, Johnson and Paul, 1983).
32          Alternatively, it is possible to determine COHb at any time by numerical integration of
33    the nonlinear CFK equation if one assumes a particular relationship between COHb and O2Hb.


      October, 2009                         5-12           Draft - Do Not Cite or Quote

-------
 1   Muller and Barton (1987) demonstrated that assuming a linear relationship between COHb and
 2   O2Hb leads to a form of the CFK equation equivalent to the Michaelis-Menten kinetic model
 3   which is analytically integrable. However, the analytical solution in this case cannot be solved
 4   explicitly for COHb. Muller and Barton (1987) demonstrated a binary search method for
 5   determining the COHb value.
 6          The COHb module used in pNEM/CO employed a linear relationship between COHb and
 7   O2Hb which was consistent with the basic assumptions of the CFK model. The approach
 8   differed from the linear forms used by other modelers in that the Muller and Barton (1987)
 9   solution was employed. However, instead of the simple binary search described in the Muller
10   and Barton paper, a combination of the binary search and Newton-Raphson root finding methods
11   was used to solve for COHb (Press et al., 1986).
12          As mentioned above,  the current COHb module included in APEX is based on the
13   solution to the non-linear CFK  equation using the assumption adopted by Muller and Barton
14   (1987) which employs a linear relationship between O2Hb and COHb. The CFK equation does
15   not have an explicit solution, so an iterative solution or approximation is needed to calculate
16   each percent COHb value.  APEX4.3 solves the CFK equation using a fourth-order Taylor's
17   series with subintervals. This method, first incorporated in APEX3, is described in detail by
18   Glen  (2002) and summarized in Appendix C.  The selected method (fourth order Taylor series
19   with subintervals) was chosen because of its simplicity, fast execution speed, and ability to
20   produce relatively accurate estimates of percent COHb at both low and high levels of CO
21   exposure.

22         5.3.8  Model Output
23          All of the output files written by APEX are ASCII text files.  Appendix D lists each of
24   the output data files written for these simulations and provides descriptions of their content.
25   Additional output files that can be produced by APEX are listed in Table 5-1 of the APEX
26   User's Guide (US EPA, 2008a). These include tabulations of hourly exposure, ventilation, and
27   energy expenditures. Detailed event-level information can also be output. Specific outputs
28   generated for the purposes of the current CO exposure and dose assessment are discussed in
29   section 6.1.

30         5.3.9  Model Limitations
31          APEX attempts to reasonably represent a sample of individuals who reside within a
32   specific geographic area, and estimates the contact with the air pollutant given the inherent
33   variability in peoples' locations and activities. This sample of individuals is a "virtual" sample,
34   created by the model according to the relative frequencies of various demographic variables and
      October, 2009                         5-13            Draft - Do Not Cite or Quote

-------
 1    census data, with the goal of obtaining a representative sample (to the extent possible) of the
 2    actual population of interest in the study area. The activity patterns of the sampled individuals
 3    (e.g., the specification of indoor and other microenvironments visited and the time spent in each)
 4    are assumed by the model to be comparable to individuals with similar demographic
 5    characteristics, and are represented by actual time-location-activity patterns compiled in CHAD
 6    (US EPA, 2002; McCurdy et al., 2000).  The air pollutant exposure concentrations are estimated
 7    by the model using a set of user-input ambient outdoor concentrations and information on the
 8    physical factors that relate ambient pollutant to concentrations expected in various
 9    microenvironments. Although this aspect of APEX is not fully employed in the current
10    simplified, screening-level assessment, the model structure would allow one to account for the
11    most significant factors contributing to inhalation exposure - the temporal and spatial
12    distribution of people and pollutant concentrations throughout the study area and among the
13    microenvironments, providing there is sufficient input data to characterize these distributions -
14    while also providing the flexibility to adjust some of these factors to meet the exposure
15    assessment objectives. This may include exposure scenarios where ambient air quality is
16    adjusted to simulate just meeting the current or alternative standards under consideration.
17          While APEX is designed to represent the most important personal attributes  and physical
18    factors that influence human exposure, all models have limitations and require the use of
19    assumptions.  Some of the general limitations of APEX are associated with the
20    representativeness of the data distributions input to the model (e.g., human activity patterns)  and
21    assumptions made within various model algorithms including the following.
22       •  The population activity pattern data used in APEX (i.e., CHAD) are compiled from
23          studies conducted in a variety of geographic areas and during time periods that differed
24          as to season and calendar year, though a large portion of CHAD is from studies of
25          national scope. Consequently, the data base may  not have data diaries available that fully
26          represent a particular study scenario.  However, to better match the activity pattern data
27          to the simulated population residing in a particular location, diary pools can be created by
28          APEX that are for specific seasons and temperature ranges.
29       •  Commuting pattern data were derived from the 2000 U.S. Census. The commuting data
30          address only home-to-work travel. The population not employed outside the home is
31          assumed to always remain in the residential census tract.  Furthermore, although several
32          of the APEX microenvironments account for time spent in travel, the travel activity is
33          typically assumed to occur in a composite of the home and work tract.  No provision is
34          made for the possibility of passing through other tracts during travel.
35       •  APEX creates seasonal or annual sequences of daily activities for a simulated individual
36          by sampling human activity data from more than one subject. While there are input
37          variables (e.g., time spent outdoors) used to simulate the correlation of day-to-day
      October, 2009                         5-14            Draft - Do Not Cite or Quote

-------
 1
 2
 3       •  The model currently does not capture certain correlations among human activities that
 4          can impact microenvironmental concentrations (for example, cigarette smoking leading
 5          to an individual opening a window, which in turn affects the rate that outdoor air enters
 6          the residence or vehicle).
 7       •  Certain aspects of the personal profiles (e.g., weight) are held constant, though in reality
 8          they change as an individual ages. This is generally only an issue for simulations with
 9          long timeframes.
10          These and other uncertainties in model inputs and algorithms, and how they may affect
11    the estimated exposures and dose, are discussed in section 6.4.

12         5.4  PERSONAL EXPOSURE AND THE IN-VEHICLE MICROENVIRONMENT
13          This section summarizes key findings from personal exposure studies with particular
14    attention to microenvironments of importance to ambient CO exposures.

15         5.4.1  Personal Exposure Monitoring Studies
16          This section summarizes some of the findings from personal exposure studies, in
17    particular, through identifying the important microenvironments in assessing CO exposure and
18    providing context for measured CO exposure levels relevant to this draft REA.  Details regarding
19    personal exposure measurement studies of target populations are discussed in section 8.2 of the
20    1991 CO AQCD (US EPA, 1991), chapter 4 of the 2000 CO AQCD (US EPA, 2000), and
21    section 3.6 of the draft ISA.
22          As ambient CO concentrations have decreased dramatically over time, so have personal
23    CO exposure levels.2 However, while CO concentrations have declined over the past few
24    decades, some general patterns in the relationship between ambient, microenvironmental, and
25    exposure concentration still remain.
26          First, as a result of the significant time people spend indoors - whether at home, at
27    school, workplace, or other indoor location (section 3.6.2,  draft ISA) - indoor CO concentrations
28    are an important determinant of an individual's CO exposures. Recent population exposure
29    studies conducted in Milan, Italy support this conclusion (Bruinen de Bruin et al., 2004),
30    indicating that over 80% of the population exposure to CO can occur in indoor
31    microenvironments (draft ISA, Table 3-13). Taking into account the infiltration of ambient CO
            2 Many recently-conducted personal exposure studies in the U.S. have not included CO as an analyte,
      possibly due to high detection limits of personal exposure monitoring devices relative to ambient concentrations.

      October, 2009                          5-15            Draft - Do Not Cite or Quote

-------
 1    indoors, indoor CO concentrations are similarly an important determinant in an individual's
 2    exposure to ambient CO.
 3           Second, there is variability in the relationship between personal exposure and ambient
 4    concentrations, particularly when considering microenvironmental exposures.  For example, the
 5    draft ISA summarized the relationship between personal CO exposures in five broadly defined
 6    microenvironments (i.e., indoor residence, indoor other, outdoor near road, outdoor other, and
 7    in-vehicle) and ambient CO concentrations in Baltimore, MD based on data provided in Chang
 8    et al. (2000). On average, the indoor-to-ambient and outdoor-to-ambient ratios were about one,
 9    though most of the ratios observed across this set of indoor and outdoor microenvironments were
10    less than one.  With the exception of those for the in-vehicle microenvironments, which as a
11    group were generally above one, few ratios were above unity (draft ISA, p. 3-102, Figure 3-43).
12    Given the expected stability of CO as it infiltrates indoor microenvironments from outdoor air
13    and the lack of significant removal mechanisms of CO in outdoor microenvironments, it is likely
14    that the variability in personal/microenvironmental-to-ambient and outdoor-to-ambient ratios is
15    the result of spatial and temporal variability in outdoor concentrations with respect to
16    simultaneously measured ambient concentrations at fixed-site monitors, and also reflects the
17    impact of lag time associated with attaining steady state relationships, as well as potential
18    presence of non-ambient sources.
19           Third, because motor vehicles remain important contributors to ambient CO
20    concentrations, both the time spent in motor vehicles and the presence of elevated on-road CO
21    concentrations continue to be important contributors to personal exposures. For example, in the
22    same study summarized by the draft ISA on personal exposures occurring within particular
23    microenvironments (i.e., Chang et al., 2000), in-vehicle exposures were, on average, a factor of 3
24    to 4 greater than ambient concentrations (distance of ambient monitor to roadway not specified),
25    with most in-vehicle exposure-to-ambient concentration ratios greater than one (median of
26    approximately 2.5). Given this relationship, it should not be surprising that while about 8% of a
27    person's time per day is spent in transit, 13-17% of their total daily exposure occurs within an in-
28    vehicle microenvironment (e.g., Bruinen de Bruin et al., 2004; Scotto di Marco et al., 2005).
29          And finally, as for CO population exposure studies conducted in the U.S., two pertinent
30    studies could be found: one conducted in Denver CO and the other in Washington, DC  during
31    the winter of 1982 and 1983 (Akland et al., 1985).  Both studies collected measurements and
32    activity pattern diaries from a random sample of the population,  defined as including non-
33    institutionalized, non-smoking residents, 18 to 70 years of age, who lived in each respective
34    city's metropolitan area.  In both cities, when comparing the distribution of measured CO
35    concentrations from the monitoring network to measured personal exposures, two common
      October, 2009                          5-16            Draft - Do Not Cite or Quote

-------
 1    phenomena were observed. At the lowest percentiles of each distribution, ambient CO
 2    concentrations were consistently greater than the personal exposures.  At the highest percentiles
 3    of each distribution, ambient concentrations were consistently lower than the personal exposures
 4    (US EPA, 2000). Again, ambient  concentrations may be a reasonable indicator of exposure for a
 5    portion of the population, but given spatial and temporal variability in ambient concentrations
 6    and exposures associated with high concentration microenvironments, there will likely be a
 7    combination of exposures that are  under- and over-estimated when considering ambient
 8    concentrations alone. As an example of the potential to underestimate exposure concentration
 9    when solely relying on ambient fixed-site concentrations as an indicator of exposure, over 10%
10    of the daily maximum 8-hour personal exposures in Denver exceeded the NAAQS of 9 ppm, and
11    about 4% did  so in Washington (Akland et al., 1985).  This is in contrast to simultaneous CO
12    measurements at ambient fixed-site monitors where CO concentrations exceeded 9 ppm about
13    3% of the time in Denver and never exceeded 9 ppm in Washington D.C. (Akland et al., 1985).
14          Consistent with the above discussion, the Denver and Washington studies determined
15    that the highest average CO concentrations occurred when subjects were in a mobile source
16    influenced microenvironment (e.g., inside parking garages, in-vehicles).  Commute time was
17    also a factor; those who commuted 6 hours or more per week had higher average exposures than
18    those who commuted fewer hours  per week. Furthermore, mean CO concentrations within in-
19    vehicle microenvironments (ranging from 7.0 to 9.8 ppm) were greater than common outdoor
20    locations (ranging from 1.4 to 3.2  ppm) (US EPA, 2000).  In considering the results from the
21    Denver and Washington personal exposure studies it is important to recognize that CO emissions
22    from motor vehicle sources have declined dramatically since the early 1980's when these studies
23    were conducted. Consequently, both ambient fixed-site CO concentrations and in-vehicle CO
24    concentrations have also been reduced significantly since that time period.

25         5.4.2   In-Vehicle CO Concentrations
26          Given the contribution of in-vehicle exposures to total  CO exposure and our focus on
27    exposure to ambient CO, consideration of the contribution of ambient CO to in-vehicle
28    concentrations is important to CO  exposure assessment. Information useful to this consideration
29    includes the relationship between CO concentrations within vehicles to concentrations
30    simultaneously measured outside of vehicles and also at nearby fixed-site monitors.  The utility
31    of such data that has been reported in the extant literature to the assessment conducted here can
32    be determined by broadly evaluating the fundamental study design and by considering potential
33    influential factors that might affect measured CO concentrations (e.g., fleet characteristics,
34    monitor siting).  Accordingly, staff evaluated data reported in  several U.S. and  non-U.S. studies
35    that measured CO concentrations inside vehicles, immediately outside vehicles, at roadside

      October, 2009                         5-17           Draft - Do Not Cite or Quote

-------
 1    locations, and at fixed-site monitors. Particular attention is given to data available within the
 2    published literature that may be most appropriate for the purposes of the current exposure
 3    assessment. The research findings from a few of the more recent studies (i.e., since 1991) are
 4    summarized below.  In addition, discussion regarding these and supporting information from
 5    studies conducted in the 1980's follows.

 6         5.4.2.1  Studies Comparing CO Concentrations Inside and Outside Motor Vehicles
 7           Table 3-1 summarizes four relevant studies selected by staff that provided data
 8    comparing CO concentrations inside a motor vehicle with concentrations immediately outside
 9    the vehicle. Two of the studies reviewed were conducted in the U.S. (Chan et al., 1991; Rodes et
10    al., 1998).  Given the low-reactivity of CO, it is expected that the ratio of the two concentrations
11    (inside-vehicle versus outside-vehicle or I/O) would be equivalent to one (in the absence of in-
12    vehicle sources).
13           Boulter and McCrae (2005) measured CO concentrations inside vehicles, immediately
14    outside the vehicles, and under a range of vehicle ventilation conditions within two tunnels: one
15    in Graz, Austria and the other in Liverpool, England.  On average the I/O ratio ranged from one
16    to slightly above unity.  Statistical analysis indicated that the air conditioning (AC), fan, and
17    window operating conditions did not have a statistically significant affect on the I/O ratios.
18           Chan et al. (1991)  measured inside-vehicle concentrations of volatile organic compounds
19    (VOCs) and three criteria  air pollutants (ozone, CO, and NO2) during the summer of 1988 in
20    Raleigh, North Carolina.  Two four-door sedans of different ages were used to evaluate
21    in-vehicle concentrations of these compounds under different driving conditions. The study
22    evaluated a variety  of factors that could influence driver exposure, including varying traffic
23    patterns, car models, vehicle ventilation conditions, and driving periods. The median I/O ratio
24    for these two vehicles operated under a variety of conditions was 1.1 (Table 3.1). Chan et al.
25    (1991) note that the slightly higher in-vehicle concentration may be a function of differences in
26    interior and exterior sampling locations, and engine running loss emissions that contributed CO
27    to the interior of the vehicle.
28           During September and October 1997, Rodes et al. (1998) collected 2-hour pollutant
29    concentration measurements inside two vehicles during scripted commutes in Sacramento and
30    Los Angeles.  Similar measurements were made simultaneously outside the vehicles,  along the
31    roadways, and at the nearest ambient monitoring stations.  A variety of scenarios were studied
32    based on variables such as roadway type, traffic congestion, ventilation setting, and vehicle type.
33    Two commutes, one in the morning and one in the afternoon, were typically conducted for each
      October, 2009                         5-18            Draft - Do Not Cite or Quote

-------
 1    scenario.3 On average, all I/O ratios were less than one in both locations, though within the
 2    range of other researchers reporting this ratio.
 3           Sharp and Tight (1997) measured inside- and outside-vehicle CO concentrations using a
 4    single automobile and considering four different ventilation settings. The averaging time for CO
 5    measurements was one minute, thus the authors calculated I/O ratios based on both average data
 6    and peak data (Table 3-1). The I/O ratios were consistent with those of similar studies, with
 7    in-vehicle CO levels being slightly higher than those measured directly outside the vehicle. Peak
 8    concentration (inside and outside of the vehicle) comparisons did span a wider range of values
 9    that included I/O ratios both below and above unity.  The choice of ventilation setting (i.e.,
10    window open or mechanical ventilation) had an effect on average and peak CO concentrations;
11    in general, conditions associated with the lowest air exchange rates (i.e., windows closed and no
12    mechanical ventilation) had the lowest I/O ratios in these studies.
13           The findings reported in each of the above four studies are supported by  a review by
14    Flachsbart (1999) regarding  other studies published between 1982 and 1992 that measured
15    interior and exterior CO concentrations simultaneously during motor vehicle trips.  The I/O ratio
16    was similar for two studies: Petersen and Allen (1982) reported a ratio of 0.92 for a study in Los
17    Angeles, California; and Koushi  et al.  (1992) reported a ratio of 0.84 for a study in Riyadh,
18    Saudi Arabia. Both of these research studies reported no affect from altering ventilation
19    conditions.
20           In contrast, one study reported indicated I/O ratios could exceed unity with the
21    ventilation set to recirculate vehicle air (Abi Esber and El-Fadel, 2008). It is possible that this
22    was the result of a gradual build-up of CO concentrations within the vehicle cabin (ISA, section
23    3.6.6.2).  In addition, Colwill and Hickman (1980) reported that internal CO  levels were about
24    30 - 80% of exterior concentrations for a study conducted in London.  However, the large
25    difference in these I/O ratios when compared with those reported by most other researchers
26    could be explained by the location of the exterior probe (i.e., at bumper height compared with
27    probes commonly placed higher  on the vehicle) (Flachsbart, 1999).
             3 The study design also included several other driving scenarios: (1) a California school bus following a
      student route in Sacramento, (2) comparison of a sedan traveling in a Los Angeles carpool lane versus one traveling
      in a congested right hand lane, and (3) a sedan encountering situations to maximize the in-vehicle pollutant
      concentration levels.  These data are not included in this summary.
      October, 2009                           5-19            Draft - Do Not Cite or Quote

-------
Table 5-2.  Carbon Monoxide Concentrations Inside and Immediately Outside Vehicles, and Indoor/Outdoor Vehicle Ratios.
Study
Boulter and
McCrae
(2005)
Chan et al.
(1991)
Rodesetal.
(1998)
Sharp and
Tight (1997)
CO concentration (ppm)
Inside
Vehicle
6.19 ±2.08
4.94 ±1.80
(mean ± std)
11.0
(median)
Sacramento:
1.4-3.5
Los Angeles:
3.5-5.4
(mean range)

Outside
Vehicle
7.13 ±1.92
4.88 ±1.86
(mean ± std)
10.0
(median)
Sacramento:
2.2-4.2
Los Angeles:
4.4-5.6
(mean range)

Indoor/Outdoor
(I/O) Ratio
0.99 ± 0.07
1.10 ±0.07
(mean ± std)
1.10
(median)
Sacramento:
0.73- 0.90
Los Angeles:
0.88- 0.96
(mean range)
1.19 to 1.43
(mean range)
0.65 to 1.38
(max range)
Drive Conditions
Plabutsch Tunnel in Graz, Austria
Kingsway Tunnel in Liverpool, England
Two sedans were driven on three road
types (urban, interstate, rural) in Raleigh,
NC during summer 1988
Two vehicles (lead and following) were
driven on various road types in Sacramento
and Los Angeles during September and
October 1997. Note, minimum quantitation
limit was 2 ppm for Draeger Model 190
monitors used.
Nine test runs were conducted in Leeds,
England on three road types under four
ventilation conditions. Lowest values
associated with windows closed and no
mechanical ventilation.
Averaging Times
Statistics are based on average
CO concentrations for trips
through tunnels.
Statistics are based on air
samples collected over one hour
periods.
Statistics are based on averages
of 120 one-minute values
measured during two-hour
morning and afternoon commute
periods.
Statistics are based on mean
and maximum (max) CO
concentrations recorded during
each minute of each test run.
October, 2009
20
Draft - Do Not Cite or Quote

-------
 1           In general, the above results suggest that the I/O ratio tends toward unity when there are
 2    no interior sources of CO, the automobile engine does not contribute directly to its own interior
 3    concentrations, and the measurement probes are properly installed on the vehicle.4 This
 4    conclusion is consistent with theoretical expectations for a non-reactive pollutant. For example,
 5    CO concentrations inside vehicles can be estimated as a function of outside CO concentration,
 6    air exchange rate, a penetration factor, and the emission rates of indoor sources (e.g., exhaust
 7    leaks, smoking). If one assumes that (1) steady-state ventilation conditions exist, (2) the indoor
 8    removal rate (K) is zero (i.e., no loss of CO as it moves from outside to inside the vehicle), and
 9    (3) there are zero emissions from interior sources, then the CO concentration inside a vehicle can
10    be simplified to a function of outside CO concentrations and the penetration rate (i.e., infiltration
11    is generally equivalent to penetration).5  Under these stated conditions, the I/O ratio would
12    ultimately converge to unity.

13         5.4.2.2  Studies Comparing CO Concentrations Inside Motor Vehicles to
14                  Concentrations at Fixed-Site Monitors and Roadside Locations
15           A report by Shikiya et al. (1989) describes an in-vehicle study conducted in the South
16    Coast Air Basin of California during the summer of 1987 and winter of 1988.  Participants were
17    randomly-selected home-to-work commuters from a non-industrial business park.
18    Measurements of hazardous air pollutants (HAPs), CO, and Pb were collected from within
19    vehicles and contrasted with measurements at existing fixed-site monitoring stations. A total of
20    192 CO measurements were made each representing the average concentration of the round-trip
21    commute.6
22           On average, CO concentrations were 8.6 ppm, and even though the maximum observed
23    CO concentration in a vehicle was as high as 46 ppm, only 3 percent of the in-vehicle
24    commuting concentration measurements were greater than 20 ppm (Table 3-2). Shikiya et al.
25    (1989) also investigated several potentially influential variables.  Statistical differences in
26    concentration were noted for season (p = 0.01), and age of vehicle (p = 0.05).  Mean
27    concentrations using several other classification variables did not differ significantly at the p =
28    0.05 level (e.g., ventilation status, vehicle speed, freeway density during commute).
             4 Interior sources of CO to in-vehicle concentrations may include self-pollution such as that associated with
      defective exhaust systems or inadequate internal ventilation (draft ISA, p. 3-105). While automobile technology has
      advanced with improvements in these areas (e.g., Flachsbart et al., 1999), interior sources may contribute in some
      instances (e.g., older school buses, draft ISA, p. 3-105).
             5 See section 3.6.2 of the draft ISA.
             6 The average one-way commute time was 33 minutes.
             October 2009                       5-21         Draft - Do Not Quote or Cite

-------
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

12
13
       Table 3-2 also presents descriptive statistics for averages of individual 4-hour integrated
samples collected during peak commuting hours at three fixed-site monitors that encompassed
the general routes of the commuters: Long Beach (a coastal location), Los Angeles (a central
location), and Pomona (an inland location).  On average, CO concentrations were 3.7 ppm,
though maximum CO concentrations were as high as 8.7 ppm.  Other monitoring sites located in
Los Angeles, Rubidoux, and Upland reported ambient 24-hour average CO concentrations of 2.2
ppm during the summer.  This group of monitors were designated by Shikiya et al. (1987) to be
less influenced by local roadway emissions compared with the three other fixed-site monitors.
Using the mean for all in-vehicle CO concentrations during round-trip commutes (8.6 ppm) and
dividing by the mean of the integrated fixed-site concentrations measured during peak
commuting hours (3.7 ppm) gives an in-vehicle to ambient monitor ratio of about 2.4.

Table 5-3.   Descriptive Statistics for CO Concentrations Measured Inside Vehicles and  at
            Fixed-Site Monitors (from Shikiya et al., 1989).
Measurement
Average in-vehicle CO
concentration during
round-trip commute


Integrated fixed-site
concentration during peak
commuting hours (Long
Beach, Los Angeles, and
Pomona stations)
Grouping
variable
All
Seasonb

Vehicle yearc

All




Category
-
Summer
Winter
1973-83
1984-88
-




Number of
samples
192
80
112


19




CO concentration (ppm)a
Mean
8.6
6.5
10.1
9.4
7.8
3.7




Std
5.0
2.2
5.8


2.1




Max
46.4
14.6
46.4


8.6




Notes:
a Mean, std and max are the arithmetic mean, arithmetic standard deviation, and maximum CO concentrations.
b Statistically significant at p = 0.01 level.0 Summer: May - October, Winter: November - March.
c Statistically significant at p = 0.05 level.
14
15
16
17
18
19
20
       Further, the draft ISA notes that studies summarized in the 2000 CO AQCD found that
in-vehicle CO concentrations were generally two to five times higher than ambient CO
concentrations obtained at fixed-site monitors within the cities studied. However, several of
these studies were conducted when CO vehicle emissions were much higher and/or under
situations that are less relevant to the two urban areas of the U.S. included in the current
exposure and dose assessment, discussed below in section 5.5.  As described above, the findings
            October 2009
                                         5-22
Draft - Do Not Quote or Cite

-------
 1    reported by Shikiya et al. (1989) in a study conducted in Los Angles supports a ratio of about
 2    two. We note, however, that based on several factors discussed above, such as the traffic
 3    characteristics of the roadway and the site characteristics of the fixed-site monitor, the
 4    relationship can vary (e.g., higher ratios would be obtained using more remotely sited monitors,
 5    and the size of relationship may vary with absolute magnitude of roadway concentrations).
 6

 7         5.5   STRATEGY FOR CO EXPOSURE/DOSE ASSESSMENT FOR THE
 8              CURRENT REVIEW
 9         5.5.1  Background for Current Assessment Strategy
10          The draft Scope and Methods Plan for the current review of the primary CO NAAQS (US
11    EPA, 2009b) described an approach based on the application of APEX to estimate human
12    exposures to CO and the resulting dose and to characterize the  potential health risks that are
13    associated with recent ambient levels of CO and with ambient levels that just meet the existing
14    standards in two urban study areas (Denver and Los Angeles).  The characterization of health
15    risks focused on development of estimates of COHb levels and the number of people and the
16    total number of occurrences for which potential COHb benchmark levels are exceeded.
17          In consideration of information  on current locations of CO monitors discussed in chapter
18    3 of this document and in the draft ISA (sections 3.5-3.7),  and CAS AC comments on both the
19    draft Scope and Methods Plan and the first draft ISA, however, staff notes significant limitations
20    of the currently available CO monitors  related to their use  in detailed population exposure
21    assessment.
22       •   The number of CO monitors in Denver and Los Angeles counties has decreased since the
23          previous review, from 9 to 3 or  4 monitors operating in Denver (depending on the year
24          considered) and from 21 to 12 in Los Angeles.
25       •   The current levels of ambient CO  concentrations are much lower than in the last review,
26          and a significant number of the  measurements are near or below detection limits.
27       •   Concentrations of ambient CO occurring in key microenvironments are not reflected by
28          ambient monitors. As stated by the CAS AC CO Panel, "Relying only on EPA's fixed
29          monitoring network CO measurements may underestimate CO exposures for specific
30          vulnerable populations such as individuals residing near heavily trafficked roads and who
31          commute to work on a daily basis" (Brain and Samet, 2009, p.2).
32       •   As discussed chapter 3 above, the currently available CO monitors pose significant
33          limitations in our ability to fully characterize the current spatial and temporal variability
34          in CO ambient concentrations across the two urban areas of focus for this assessment,
35          Denver  and Los Angeles.  These limitations affect  our ability to derive detailed
36          relationships about CO concentrations in ambient air across the study area from which
37          detailed microenvironmental concentrations can be estimated. More broadly, the
            October 2009                       5-23        Draft - Do Not Quote or Cite

-------
 1          CASAC CO Panel expressed the view that "the current ambient monitoring network is
 2          not well designed to characterize spatial and temporal variability in ambient
 3          concentrations. Thus, it does not adequately support detailed assessments of human
 4          exposure or air quality modeling such as for photochemical oxidants" (Brain and Samet,
 5          2009, p. 11).
 6          In light of these limitations in the air quality data, including CASAC's concerns about the
 7    adequacy of the monitoring network to perform detailed exposure analyses, and in light of the
 8    findings from the CO assessments completed for the 1994 review (USEPA, 1994) and
 9    subsequently (Johnson, et al., 2000), as well as the lack of new evidence in the draft ISA to
10    support a quantitative risk characterization approach different from past assessments, staff
11    decided not to perform a detailed analysis involving multiple monitors and comprehensive
12    estimation of exposure concentrations in multiple microenvironments, as has been done in the
13    past.  Rather than develop such a detailed analysis using the available air quality data that has
14    been recognized as limited by CASAC for such a purpose, we developed an alternative
15    simplified, screening-level approach.  We recognize that there are uncertainties associated with
16    the revised approach and, thus, its utility is primarily as a screening assessment to provide some
17    perspective on current ambient CO concentrations and associated CO exposure, dose and risk.
18    One purpose of this draft document is to seek CASAC views on the extent to which this
19    assessment design  provides information useful to this current CO NAAQS review.
20          The following section presents the approach used to develop the current CO exposures
21    and COHb estimates for the two urban study areas presented in this document.

22         5.5.2   Selected Approach for Current Review
23          As discussed in section 5.5.1 above, despite the capabilities of the APEX model, staff felt
24    that the lack of spatial and temporal variability in available ambient monitoring data precluded
25    the development of a credible broad-scale urban exposure assessment such as that conducted
26    recently for the O3 NAAQS (US EPA, 2007).  Therefore, staff decided to perform a limited
27    exposure analysis using APEX and ambient data at a single monitor each in Denver and Los
28    Angeles counties.
29          In developing the approach, staff evaluated the monitoring data available in the selected
30    locations (i.e., Denver County and Los Angeles County) for years 2005 through 2007 (see Table
31    3-1 for sites IDs, locations, and number of sample-hours). Staff noted that, following the
32    examination of the existing and active monitors in Denver and Los Angeles counties and using a
33    75% completeness criterion, only two of the Denver County monitors had data for 2007.  Three
34    of the six Denver sites used in the exposure assessment conducted in 2000 (Johnson et al. 2000)
35    have been removed from the monitoring network and are no longer reporting CO concentrations
36    in Denver (Figure 5-1). Based on these observations, 2006 was chosen to be the most recent
            October 2009                       5 -24        Draft - Do Not Quote or Cite

-------
 1    year to be analyzed.  As a result of this choice, there were four monitors in Denver County and
 2    twelve monitors in Los Angeles County (Figure 5-1) with complete air quality data available.
 3          The monitor siting characteristics and breadth of spatial coverage are also important
 4    features to consider in representing the air quality in an area. Most of the monitors in Denver
 5    County are clustered within the central portion of the Denver County, with two monitors (IDs
 6    080310002 and 080310019) within 1 mile of one another and generally having a similar
 7    concentration distribution when considering their 1-hour and 8-hour average CO concentrations
 8    (Figure 5-2). There is limited spatial variation among the twelve Los Angeles County monitors
 9    (Figure 5-3). Further, of the monitors available in Denver County to use in an exposure
10    assessment, one monitor appears to best represent the highest population density in Denver
11    County (draft ISA, Figure 3.4-3).  This monitor (ID 08-031-0002) is located at 2105 Broadway.
12    This particular monitor was included in the previous reviews and continues to report the some of
13    the highest concentrations in the area (Figure 5-2 and Table 3-1).  It  is described as a micro-scale
14    site, within 6 meters from a roadway having 17,200 vehicles/day traffic volume, 7 meters from a
15    road with 10,000 vehicles/day, and 16 meters from a road with 1,000 vehicles/day. Based on the
16    same criteria (i.e., to envelop a study area that captures the population centers and where ambient
17    CO levels tend to be high), a single monitor in Los Angeles County (ID 060371301) was
18    selected for use in the exposure assessment for the Los Angeles study area.  This monitor is
19    described as representing a middle scale, and it is near to an arterial road, but 350 m from a
20    major freeway (the 1-105) with a traffic count close to 35,000 vehicles/day.  Staff note, however,
21    that a study of ambient CO concentrations related to motor vehicle traffic in Los Angeles and
22    Sacramento (Rodes et al., 1998) observed little difference in CO concentrations between arterial
23    roads and freeways for Los Angeles (draft ISA, p. 3-65).
            October 2009                       5-25        Draft - Do Not Quote or Cite

-------
                         CO Monitors in Denver
                    Active in 2006 (75% complete data)
                    ACTIVE

                  O INACTIVE
                         CO Monitors in LA County
                     Active in 2006 (75% complete data)
3   Figure 5-1.  Locations of active ambient CO monitors meeting 75% completeness
4               criterion in 2006 along with locations of inactive ambient CO monitors,
5               within the metropolitan Denver (top) and metropolitan Los Angeles (bottom).
          October 2009
5-26
Draft - Do Not Quote or Cite

-------
                     10-
                     7.5-
                   E
                   Q.
                   a.
                     2.5-
                      0 -
                              080310002
                                             080310013
                                                           080310014
                                                                          080310019
                                                      Site ID
                I   6

                O
                O
                x
                ro
                ro
                •c
                o


                CO
                    4-
                    2-
                           080310002
                                          080310013
                                                        080310014
                                                                       080310019
                                                   Site ID
3    Figure 5-2.   Distribution of 1-hour (top) and 8-hour average daily maximum (bottom) CO

4                 concentrations at ambient CO monitors in Denver County, year 2006.
            October 2009
5-27
Draft - Do Not Quote or Cite

-------
              Q.
              a.
              O
              O
                 101
                 8-
                 6.
                 4-
                 2-
                                        060370002 = A
                                        060370113 = 6
                                        060371002 = C
                                        060371103 = 0
                                        060371201 =E
                                        060371301 =F
                 060371701 = G
                 060372005 = H
                 060374002 = I
                 060375005 = J
                 060376012 = K
                 060379033 = L
                                                F     G     H
                                                 Site ID
           Q.

           O
           O
           E
           X
           ro
           ro
           •c
           o
           CO
6-
              2-
              o-i
                                           T
060370002 = A
060370113 = B
060371002 = 0
060371103 = 0
060371201 =E
060371301 =F
060371701 = G
060372005 = H
060374002 = I
060375005 = J
060376012 = K
060379033 = L
                                   ±   ^   ±
                         E    F     G
                               Site ID
                                                        H
Figure 5-3.  Distribution of 1-hour (top) and 8-hour average daily maximum (bottom) CO
             concentrations at ambient CO monitors in Los Angeles County, year 2006.
       October 2009
                            5-28
    Draft - Do Not Quote or Cite

-------
 1          In each of the two study areas, two broadly defined air quality scenarios were
 2    investigated. In the first scenario ("Scenario A"), all microenvironmental concentrations are set
 3    equal to the ambient concentrations measured at the single fixed-site monitor selected to
 4    represent each exposure modeling domain.  Staff used APEX to estimate the COHb levels in
 5    blood of the at-risk population (i.e., adults with CHD) by assuming that the population is
 6    exposed to the ambient concentrations measured at the selected near-road monitor for each study
 7    area throughout the entire simulation period. This general assumption likely results in over-
 8    estimates of CO exposure and COHb levels for much of the population because CO peak hourly
 9    concentrations are typically somewhat lower indoors than outdoors due to consideration of air
10    exchange (in the absence of indoor sources of CO). This scenario, however, may underestimate
11    CO exposure for some small portion of the population that may live in close proximity to heavily
12    trafficked roadways and spend appreciable time in transit on such roadways.  These individuals,
13    based on the analysis of air quality relationships and personal exposure  measurements, would
14    likely have periods of higher exposures than represented by Scenario A since CO concentrations
15    in vehicles, and  exposure concentrations for individuals in transit on roadways, are typically
16    higher than the concentrations measured at a near-road monitor. The impact of such higher
17    exposure periods on an individual's COHb levels will vary depending on the magnitude and
18    pattern of exposures in the prior and subsequent hours.
19          In the second scenario ("Scenario B"), we assume that the concentration outside a motor
20    vehicle is twice the ambient concentration and that the concentration inside the vehicle is the
21    same as the concentration immediately outside the vehicle; that is, the in-vehicle
22    microenvironment is set equal to twice the ambient monitor concentrations. For Scenario B, all
23    other microenvironmental concentrations are set equal to the ambient concentration based on the
24    fixed-site monitor concentration, consistent with their treatment in Scenario A. The intent of this
25    scenario is to determine the magnitude of the change in exposure and COHb levels when
26    incorporating a rough estimate of the greater exposure concentrations occurring inside motor
27    vehicles.  Further details regarding the air quality scenarios and specific exposure modeling
28    input data used for the assessment are given in chapter 6.
29
            October 2009                      5-29        Draft - Do Not Quote or Cite

-------
 1         5.6  KEY OBSERVATIONS
 2          Presented below are key observations related to the approach for the population
 3    assessment of CO exposure and dose.

 4       •  APEX, an EPA human exposure and dose model, has a long history of use in estimating
 5          exposure and dose for many of the criteria pollutants including CO, O3, SO2, and NO2.
 6          Over time, staff have improved and developed new model algorithms, incorporated
 7          newer available input data and parameter distributions, as well as performed several
 8          model evaluations, sensitivity analyses, and uncertainty characterizations for the same
 9          pollutants. Based on this analysis, APEX was judged to be an appropriate model to use
10          for assessing CO exposure and dose.
11       •  Personal CO exposure studies indicate that in general, indoor exposures contribute the
12          greatest portion of an individual's total daily exposure, though variability in exposure
13          concentrations may be driven largely by exposure in certain microenvironments, such as,
14          with regard to ambient CO, inside motor vehicles or when outdoors near roadways.
15          Accordingly, in estimating CO exposures and associated COHb levels an approach is
16          needed to estimate the generally higher in-vehicle and in-transit exposure concentrations
17          compared to the generally lower ambient concentrations concurrently reported by fixed
18          site ambient monitors.
19       •  Given the limitations in the number of ambient monitors currently in operation, the
20          limited spatial and temporal representation of ambient concentrations provided by the
21          current monitoring network, and limited number of CO concentrations at or above the
22          instrument detection limit, the simplified, screening-level approach used in this exposure
23          assessment does not employ detailed microenvironmental concentration modeling and
24          uses a single fixed-site monitor in each study area.  The single monitoring site selected in
25          each location typically reported a higher range of CO concentrations when compared
26          with other monitors in each area, and thus, when used as an input to an exposure model,
27          is generally considered likely to generate conservative (i.e., higher) estimates of exposure
28          for the large majority of the population.
29
            October 2009                      5-30        Draft - Do Not Quote or Cite

-------
  1          5.7   REFERENCES

 2    Abi-EsberL; El-FadelM. (2008).  "In-vehicle CO ingression: Validation through field measurements and mass
 3            balance simulations." Sci. Total Environ. 394:75-89.

 4    Akland GG, Hartwell TD, Johnson TR, Whitmore RW. (1985). Measuring human exposure to carbon monoxide in
 5            Washington, DC, and Denver, Colorado, during the winter of 1982-1983. Environ Sci Technol. 19: 911-
 6            918.

 7    Boulter, P., andMcCrae, I. (2005).  "Carbon Monoxide Inside Vehicles: Implications for Road Tunnel
 8            Ventilation." In: Annual Research Review 2005.  TRL Academy.

 9    Brain JD and Samet JM (2009). Letter to EPA Administrator Lisa Jackson: Clean Air Scientific Advisory
10            Committee's (CAS AC) Peer Review of the Agency's 1st Draft Carbon Monoxide Integrated Science
11            Assessment. EPA-CASAC-09-011. June 24, 2009.

12    Biller WF and Richmond HM. (1982).  Sensitivity Analysis on Coburn Model Predictions of COHb Levels
13            Associated with Alternative CO Standards. Report to Strategies and Air Standards Division of the Office
14            of Air quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park,
15            NC. November, 1982.

16    Biller WF, Feagans TB, Johnson TR, Duggan GR, Paul RA, McCurdy T, and Thomas HC.  (1981).  A General
17            Model for Estimating Exposure Associated with Alternative NAAQS.  Paper No. 81-18.4.  Presented at the
18            74th  Annual Meeting of the Air Pollution Control Association, Dallas, Texas.

19    Chan, C., Ozkaynak, H., Spengler, J., and Sheldon, L. (1991).  "Driver Exposure to Volatile Organic Compounds,
20            CO, Ozone, and NO2 Under Different Driving Conditions."  Environ. Sci. Technol. Volume 25, No. 5, pp.
21            964-972.

22    Chang LT; Koutrakis P; Catalano PJ; Suh HH. (2000). Hourly personal exposures to fine particles and gaseous
23            pollutants - Results from Baltimore, Maryland.  J Air Waste Manage Assoc.  50: 1223-1235.

24    Colwill, D., and Hickman, A.  (1980). "Exposure of Drivers to Carbon Monoxide."  Journal of the Air Pollution
25            Control Association.  Volume 30, pp. 1316-1319.

26    Esmail S, Bhambhani Y, Brintnell S. (1995). Gender Differences in Work Performance on the Baltimore
27            Therapeutic Equipment Work Simulator. Amer J Occup Therapy. 49:405-411.

28    Flachsbart, P.  (1999). "Human Exposure to Carbon Monoxide from Mobile Sources." Chemosphere - Global
29            Change Science. Volume 1, pp. 301 - 329.

30    Flachsbart, P., Mack, G., Howes, J., and Rodes, C. (1987). "Carbon Monoxide Exposures of Washington
31            Commuters." Journal of the Air Pollution Control Association. Volume 37, pp. 135 - 142.

32    Galetti, P. M.  (1959). Respiratory Exchanges During Muscular Effort. Helv Physiol Acta.  17:34-61.

33    Glen, G.  (2002).  Programmer's Guide for the APEX3 Model.  Prepared by ManTech Environmental Technology,
34            Inc., for the U.S. Environmental Protection Agency, Research Triangle Park, NC. April, 2002.

35    Joumard R, Chiron M, Vidon R, Maurin M, and Rouzioux JM (1981). Mathematical Models of the Uptake of
36            Carbon Monoxide on Hemoglobin at Low Carbon Monoxide Levels. Environ Health Persp.  41:277-
37            289.

38

              October 2009                         5-31         Draft - Do Not Quote or Cite

-------
  1    Johnson T. (2003).  A Guide to Selected Algorithms, Distributions, and Databases Used in Exposure Models
  2            Developed by the Office of Air Quality Planning and Standards. Prepared by TRJ Environmental, Inc., for
  3            Office of Research and Development, U.S. Environmental Protection Agency, Research Triangle Park,
  4            North Carolina.

  5    Johnson T and Capel J.  (2003). Total Risk Integrated Methodology. TRIM Expolnhalation User's Document.
  6            Volume III: Air Pollutant Exposure Model Criteria Air Pollutants Case Study (Carbon Monoxide
  7            Exposures in Los Angeles). Report prepared by TRJ Environmental, Inc., for the Office of Air Quality
  8            Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, North Carolina.

  9    Johnson T and Paul R. (1983).  The NAAQS Exposure Model (NEM) Applied to Carbon Monoxide. EPA Report
10            No. 450/5-83-003. Prepared by PEDCo Environmental, Inc. forthe Office of Air Quality Planning and
11            Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC.

12    Johnson T, Capel J, Olaguer E, Wijnberg L.  (1992). Estimation of Carbon Monoxide Exposures and Associated
13            Carboxyhemoglobin Levels in Denver Residents Using a Probabilistic Version of NEM. Report prepared
14            for the Office of Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research
15            Triangle Park, North Carolina.

16    Johnson T, Mihlan G, LaPointe J, et al.  (2000). Estimation of Carbon Monoxide Exposures and Associated
17            Carboxyhemoglobin Levels in Residents of Denver and Los Angeles Using pNEM/CO (Version 2.1).
18            Prepared by TRJ Environmental, Inc., for the Office of Air Quality Planning and Standards, U.S.
19            Environmental Protection Agency, Research Triangle Park, NC.

20    Koushi, P., Al-Dhowalis, K., and Niazi, S. (1992). Vehicle Occupant Exposure to Carbon Monoxide.  Journal of
21            the Air and Waste Management Association. Volume 42, pp. 1603 - 1608.

22    McCurdy  T, Glen G, Smith L, and Lakkadi Y. (2000).  The National Exposure Research Laboratory's
23            Consolidated Human Activity Database.  J Exp Anals and Environ Epi. Volume 10, pp. 566-578.

24    Muller, K. E., and C. N. Barton. (1987).  A Nonlinear Version of the Coburn, Forster and Kane Model of Blood
25            Carboxyhemoglobin.  Atmos Environ. 21:1963-1967.

26    Peterson JE, and Stewart RD. (1975). Predicting the Carboxyhemoglobin levels resulting from carbon monoxide
27            exposures. J. Appl. Physiol. 39(4): 633-638.

28    Petersen, W., and Allen, R. (1982). Carbon Monoxide Exposures to Los Angeles Area Commuters.  Journal of the
29            Air Pollution Control Association. Volume 32, pp. 826 - 833.

30    Press, W.  H., B. P. Flannery, S. A. Teukolsky, and W. T. Vettering. (1986).  Numerical Recipes. Cambridge
31            University Press.

32    Rodes, C., Sheldon, L., Whitaker, D.,  Clayton, A., Fitzgerald, K., Flanagan, J., DiGenova, F., Hering, S., and
33            Frazier, C.  (1998). Measuring Concentrations of Selected Air Pollutants Inside  California Vehicles.  Final
34            Report.  Prepared by Research Triangle Institute under Contract No. 95-339. California Air Resources
3 5            Board.  Sacramento, California.

36    SchofieldWN. (1985). Predicting basal metabolic rate, new standards,  and review of previous work.  HumNutr
37            ClinNutr. 39C(S1): 5-41.

38    Scotto di Marco G, Kephalopoulos S, Ruuskanene J, and Jantunene M. (2005). Personal carbon monoxide exposure
39            in Helsinki, Finland. Atmos Environ. 39Z: 2697-2707.

40


              October 2009                           5-32         Draft - Do Not Quote or Cite

-------
  1     Sharp, D., and Tight, M.  (1997). "Vehicle Occupant Exposure to Air Pollution." In: Policy, Planning, and
  2            Sustainability: Proceedings of Seminars C and D Held at PTRC European Transport Forum, Brunei
  3            University.  Pages 481-492.

  4     Shikiya, D., Liu, C., Kahn, M., Juarros, I, and Barcikowski, W. (1989). In-Vehicle Air Toxics Characterization
  5            Study in the South Coast Air Basin. Office of Planning and Rules, South Coast Air Quality Management
  6            District. May.

  7     US EPA.  (1991).  Air Quality Criteria for Carbon Monoxide. Research Triangle Park, NC: Office of Health and
  8            Environmental Assessment, Environmental Criteria and Assessment Office; report no. EPA/600/8-90/045F

  9     US EPA.  (2000).  Air Quality Criteria for Carbon Monoxide. National Center for Environmental Assessment,
10            Office of Research and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC
11            27711; report no. EPA/600/P-99/001F. June 2000. Available:
12            http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=18163.

13     US EPA.  (2002).  EPA's Consolidated Human Activities Database. Available at: http://www.epa.gov/chadnetl/.

14     US EPA.  (2007).  Ozone Population Exposure Analysis for Selected Urban Areas (July 2007). Office of Air
15            Quality Planning and Standards, Research Triangle Park, NC. EPA-452/R-07-010.  Available at:
16            http://epa.gov/ttn/naaqs/standards/ozone/s_o3_cr_td.html.

17     US EPA.  (2008a). Total Risk Integrated Methodology (TRIM) Air Pollutants Exposure Model Documentation
18            (TRIM.Expo/APEX, Version 4.3).  Volume  1: Users Guide. Report No. EPA-452/B-08-00la.  Office of
19            Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC.

20     US EPA.  (2008b). Total Risk Integrated Methodology (TRIM) Air Pollutants Exposure Model Documentation
21            (TRIM.Expo/APEX, Version 4.3).  Volume 2: Technical Support Document. Report No. EPA-452/B-08-
22            00Ib. Office of Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research
23            Triangle Park, NC.

24     US EPA.  (2008c). Risk and Exposure Assessment to Support the Review of the NO2 Primary National Ambient
25            Air Quality  Standard. EPA-452/R-08-008a.  November 2008.  Available at:
26            http://www.epa.gov/ttn/naaqs/standards/nox/data/2008112 l_NO2_REA_final.pdf.

27     US EPA.  (2009a). Risk and Exposure Assessment to Support the Review of the SO2 Primary National Ambient
28            Air Quality  Standard. EPA-452/R-09-007. August 2009. Available at:
29            http://www.epa.gov/ttn/naaqs/standards/so2/data/200908SO2REAFinalReport.pdf.

30     US EPA.  (2009b). Carbon Monoxide National Ambient Air Quality Standards: Scope and Methods Plan for
31            Health Risk and Exposure Assessment.  EPA-452/R-09-004. April 2009.  Available at:
32            http://www.epa.gov/ttn/naaqs/standards/co/data/2009_04_COScopeandMethodsPlan.pdf.

33     US EPA (2009). Integrated Science Assessment for Carbon Monoxide -Second External Review Draft. U.S.
34            Environmental Protection Agency,  Research Triangle Park, NC, report no. EPA 600/R-09/019B. Available
35            at: http://www.epa.gov/ttn/naaqs/standards/co/s_co_cr_isa.html..

36
              October 2009                          5-33         Draft - Do Not Quote or Cite

-------
 1                     6   EXPOSURE/DOSE ASSESSMENT AND RISK
 2                                      CHARACTERIZATION

 3          This chapter describes the draft assessment of ambient CO exposure and dose and
 4    associated characterization of risk and uncertainty. Section 6.1 provides an overview of the
 5    application of the APEX model in this assessment. In section 6.2, we present the exposure and
 6    dose estimates and describe the estimates in the context of potential health benchmarks.  The
 7    current assessment and associated exposure and dose estimates are described in comparison to
 8    estimates from an earlier CO assessment (Johnson et al., 2000) in section 6.3. The final section
 9    presents  an analysis of how variability was considered by  staff in this assessment followed by a
10    qualitative characterization of uncertainty (section 6.4).

11         6.1  APPLICATION OF APEX4.3 TO CARBON MONOXIDE
12          The previous analysis of population exposure to carbon monoxide (CO) employed the
13    pNEM/CO model (summarized in section 5.2 above) and focused on Denver and Los Angeles
14    study areas, comprising the  majority of census tracts within those metropolitan areas (Johnson et
15    al., 2000). Air quality data were obtained from multiple fixed-site monitors within the study
16    areas, and the exposure assessment accounted for the effects of geographic location, a diverse set
17    of microenvironments, commuting within the study area, and selected indoor sources (e.g.,
18    environmental tobacco smoke, gas stoves).
19          In the specific application of APEX4.3 described in this report, a different approach has
20    been taken.  The simplified  approach presented is intended to emphasize the contribution of CO
21    exposures occurring in vehicles to overall CO exposures.  The contribution of in-vehicle
22    exposures is thought to be influential in producing the highest exposures to ambient CO and
23    resulting COHb levels (draft ISA, section 3.6.6.2).  The ambient  CO concentration throughout
24    the study areas and all microenvironmental concentrations are generally assumed to equal the
25    hourly-average concentrations measured by a single high CO concentration monitor in one of the
26    two scenarios (Scenario A).  In a second scenario, in-vehicle concentrations are assumed to equal
27    twice the ambient monitor concentrations (Scenario B).  Therefore, the locations visited by
28    simulated persons in the study area are represented by at most two microenvironments (in-
29    vehicle and all other).  The study areas and scenarios are briefly described in sections 6.1.1 and
30    6.1.2 below.  The study area population of interest was limited to adults with coronary heart
31    disease (CHD) that reside within a defined distance of a single downtown fixed-site monitor (as
32    discussed in sections 5.5.2 and 6.1.3). Given simplified assumptions in the approach, the
33    generated results would tend to represent population exposures experienced by persons
34    residing/travelling in high CO (of ambient origin) concentration microenvironments. However,
      October, 2009                              6-1                Draft - Do Not Cite or Quote

-------
 1    note that this assessment does not include the contribution of indoor sources to total CO
 2    exposure, which in prior assessments have been shown to be important contributors to total
 3    exposures (see US EPA, 1992; Johnson et al., 2000).
 4          A general description of APEX4.3 and its capabilities, as well as the history of the
 5    pNEM/APEX series of exposure models  can be found in chapter 5.  This section focuses on the
 6    development of the  specific input files and parameters used in the current application of
 7    APEX4.3 to CO in the Denver and Los Angeles study areas. In particular, this section (and
 8    associated appendices) describes the
 9       •  Geographic  areas and time periods defined for the exposure analyses,
10       •  Exposure scenarios under evaluation,
11       •  Populations-at-risk and the associated prevalence rates for CHD,
12       •  Air quality and meteorological data used for each study area,
13       •  Microenvironments defined for each exposure scenario, and
14       •  Methods used to construct a composite diary for each simulated individual.
15          In addition to the application-specific input data bases described in this section, we used a
16    number of default databases provided with APEX4.3  as inputs to the model.  These included
17    national data files obtained from the U.S. Census Bureau (i.e., the 2000 Census data) for the
18    following types of information (http://www.epa.gov/ttn/fera/apex_download.html#input):
19       •  Population data by race, gender, age, and census tract;
20       •  Employment probabilities by gender, age, and census tract;
21       •  Locations of census tracts (latitude and longitude); and
22       •  Commuting flows for combinations of home and work census tracts.
23          Another default input file provided tables of age- and gender-specific physiological
24    parameters (e.g., weight). The contents of these default files will not be described in this section;
25    they are described in detail in the APEX Users Guide (US EPA, 2008a) and the APEX Technical
26    Support Document (US EPA, 2008b).

27         6.1.1   Study Areas and Exposure Periods
            As discussed in section 3.2, EPA  selected areas within Denver, Colorado, and Los
      Angeles, California, for the exposure assessment.  Briefly, considerations in selection of these
      areas include: the prior analysis of these locations in CO NAAQS reviews, the areas having
      historically elevated CO concentrations, and the areas currently having some of the most
      complete ambient monitoring data available.  The actual study areas were defined as including
      all census tracts within 20 km of the following fixed-site monitors.

      October, 2009                            6-2               Draft - Do Not Cite or Quote

-------
 1          Denver:      Monitor No. 080310002; 2105 Broadway, Denver, CO (CAMP site).
 2          Los Angeles:  Monitor No. 060371301; 11220 Long Beach Blvd., Lynwood, CA.
 3          The Denver study area includes most of the urbanized area inside the arc defined by
 4   Highway 470 (Figure 6-1) within Denver County. The Los Angeles study area is centered at
 5   Lynwood, CA and includes large portions of Los Angeles and Long Beach within Los Angles
 6   County (Figure 6-2).
 7          EPA selected the following calendar years as the study periods for each area.
 8          Denver:      1995 and 2006
 9          Los Angeles:  1997 and 2006
10   The year 2006 was selected for both cities because it was the most recent year of monitoring data
11   that met the 75% completeness requirement for the fixed-site monitors listed above.  The CO
12   levels reported for 2006 were well below the 8-hour NAAQS  and were considered representative
13   of current conditions in each study area. The year 1995 for Denver and the year 1997 for Los
14   Angeles were selected as periods for which the monitoring data indicated higher CO conditions
15   near or exceeding, the 8-hour CO NAAQS (9 ppm).  As discussed in section 6.1.4.3, staff
16   applied an adjustment to the monitoring data reported for these years to simulate ambient CO
17   levels that would just meet the current 8-hour NAAQS.
     October, 2009                            6-3               Draft - Do Not Cite or Quote

-------
                                                  r I ''      ~
                                                  :" "r~pTnornton
                                                  onitor No: 08-031-0002,=^^ k_

                                                    r;ri=
                    k. \ . r .  Ken Caryl
2   Figure 6-1.  Map of the Denver Study Area Defined as a Circle with Radius = 20 km

3                Centered on Fixed-Site Monitor ID 080310002.
    October, 2009
6-4
Draft - Do Not Cite or Quote

-------
                                                                    Angles National Forest


                                                                      Sierra Mad re
                                               '•South Pasadena fr- ]
                              ^\4m-^-:'^J -
-------
 1         6.1.2   Exposure Scenarios
 2          Two exposure scenarios were evaluated for each study area, designated as Scenario A
 3    and Scenario B. The scenarios differed according to the modeling factors assumed for the two
 4    broadly defined microenvironments modeled (i.e., in-vehicle and all others).  See section 6.1.5
 5    for additional details regarding microenvironments. Both Scenario A and B were focused on the
 6    adult CHD population living within each study area and are described in more detail below.

 7         6.1.2.1  Scenario A
 8          In this scenario,  the sequence of concentrations in each microenvironment was set equal
 9    to the ambient concentrations derived from the fixed-site monitor selected to represent the study
10    area. Essentially, the microenvironment assignments in each diary entry for each simulated
11    individual did not affect exposure concentration levels (i.e., the individual was exposed to the
12    ambient concentrations  as measured at the fixed-site monitor for all hours in the exposure
13    period).
14          The time/activity database that was used as the input for this scenario included all adult
15    diaries in CHAD and thus was not necessarily specific to the study area. Note however that the
16    APEX model samples from this pool of diaries to reflect the actual population distribution based
17    on the specific age and gender residing in each census tract.  In addition, the sampling from the
18    broad diary pool is also  guided by several temperature ranges and applied to observed
19    temperatures for the specific geographic region.

20         6.1.2.2  Scenario B
21          Scenario B assumed that the concentrations outside a motor vehicle were greater than that
22    measured at the ambient fixed-site monitor and that the concentrations inside the vehicle were
23    the same as the concentrations immediately outside the vehicle. The CO concentrations in all
24    other microenvironments were set equal to the ambient concentrations as measured at the fixed-
25    site monitor, consistent  with their treatment in Scenario A.  As in Scenario A, the input
26    time/activity database included all diaries in CHAD.

27         6.1.3   Populations-at-Risk
28          Staff defined the population group at risk within each study area to include adults ages 18
29    or older with CHD.  Coronary heart disease is caused by inadequate circulation of the blood to
30    the heart muscle, which is a result of the coronary arteries being blocked by cholesterol deposits
31    (ISA, section 5.2.1.9). The focus on adults with CHD is consistent with the previous (2000)
32    review of the CO NAAQS. The current and previous assessments focused on adults as the
33    incidence of CHD in younger individuals is extremely small.

      October, 2009                              6-6                Draft - Do Not Cite or Quote

-------
 1          At the time of the review completed in 1994, the estimated number of individuals with
 2   CHD represented about 3% of the entire (all ages) U.S. population (US EPA, 1992). More
 3   recently however, the National Health Interview Survey for 2007 reported a prevalence rate for
 4   CHD of about 6 percent for the population above 18 years of age (CDC, 2009; draft ISA, section
 5   5.7.2.1).  The current exposure/dose assessment requires estimates (by age and gender) of the
 6   fraction of the population in the Denver and Los Angeles study areas who have CHD.  Given the
 7   general similarity in regional rates for CHD, staff decided to apply national prevalence rates for
 8   CHD to each of the two study areas.  Table 6-1 provides national prevalence data for CHD by
 9   age obtained from the National Health Interview Survey of 2007 (CDC, 2009).  Table  6-2
10   provides national prevalence rates for CHD by gender obtained from the same source.  These
11   data were used to estimate gender-specific adjustment factors of 1.31 = males/total = 0.080/0.061
12   for males and 0.74 = females/total = 0.045/0.061 for females. Table 6-3 provides estimated
13   national prevalence rates for CHD by age range adjusted for gender differences using these
14   values.

15   Table 6-1.  National Prevalence Rates for Coronary Heart Disease by Age Range.
Age range
18 to 44
45 to 64
65 to 74
75+
Prevalence rate (fraction)
for coronary heart disease3
0.009
0.067
0.187
0.236
aSource: Coronary heart disease statistics in Table 2, "Summary Health
Statistics for U.S. Adults: National Health Interview Survey, 2007," U.S.
Department of Health and Human Services, Center for Disease Control,
Hyattsville, MD, May 2009.
16   Table 6-2. National Prevalence Rates for Coronary Heart Disease by Gender.
Age range
18+
Prevalence rate (fraction)
for coronary heart disease3
Total
0.061
Males
0.080
Females
0.045
aSource: Coronary heart disease statistics in Table 2, "Summary Health Statistics
for U.S. Adults: National Health Interview Survey, 2007," U.S. Department of Health
and Human Services, Center for Disease Control, Hyattsville, MD, May 2009.
17
     October, 2009
6-7
Draft - Do Not Cite or Quote

-------
 1   Table 6-3.  National Prevalence Rates for Coronary Heart Disease Used in APEX,
 2               Stratified by Age and Gender.
Age range
18 to 44
45 to 64
65 to 74
75+
Prevalence rate (fraction) for
coronary heart disease3
Males
0.012
0.088
0.244
0.310
Females
0.007
0.050
0.138
0.175
aSource: Values listed in Table 6-2 were multiplied by 1 .31 (=
0.080/0.061) for males and 0.74 (= 0.045/0.061) for females.
 4         6.1.4  Air Quality and Meteorological Data
 5         6.1.4.1  Selection of Fixed-Site Monitors
 6         Based on considerations described in sections 3.2 and 5.5.2, staff selected the downtown
 7   "CAMP" monitor (ID 080310002) to represent ambient CO concentrations in the Denver, CO
 8   study area and monitor 060371301 in Lynwood, CA to represent ambient CO concentrations in
 9   the Los Angeles study area. Details regarding each monitor's site characteristics are given in
10   (Table 6-4). Note that the ambient monitor in Denver is a microscale monitor sited within the
11   urban core and generally records the highest hourly CO concentrations within the county (Table
12   3-1).  Similarly, the middle scale monitor in the Los Angeles study also reported the highest CO
13   concentration levels in the Los Angeles study area (Table 3-2).
     October, 2009
6-8
Draft - Do Not Cite or Quote

-------
 1
 2
Table 6-4.  Site Characteristics of Fixed-site  CO  Monitors  Selected  to  Represent  the
           Denver and Los Angeles Study Areas.
Site Characteristic
Site ID
Street address
Town
Local site name
Latitude
Longitude
Elevation (above sea level), meters
Monitor start date
Measurement scale
Sample collection frequency
Sample analysis method
Monitor type
Reporting agency
Denver
080310002
2105 Broadway
Denver, CO
CAMP
39.751184
-104.987625
1593
January 1, 1971
Microscale
1 hour
Non-dispersive infrared
SLAMS
Colorado Department of
Public health and
Environment
Los Angeles
060371301
11220 Long Beach Blvd.
Lynwood, CA

33.928990
-118.210710
27
January 1, 1973
Middle scale
1 hour
Non-dispersive infrared
SLAMS
South Coast Air Quality
Management District
 4
 5
 6
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
     6.1.4.2  Estimation of Missing Air Quality Values
       APEX requires that each site-year of monitoring data be complete (i.e., it is free of hourly
gaps in concentration levels). The missing values in each data set were estimated by the
sequential application of the following three methods.

     1)  If the data gap was less than six continuous missing values, the missing values were
         estimated by linear interpolation using the valid values at the ends of the gap.
     2)  Where possible, data gaps of at least 6 hours were estimated as linear functions of
         hourly values reported by other ambient CO monitors in the area. Linear regression
         was used to develop a set of models that were specific to a time-of-day and at each
         monitor. The model selected to estimate missing values for a particular time of day
         was the model that maximized the variance explained (R2) for that hour, subject to the
         constraints that regression model R2 was greater than 0.5 and the number of
         measurements used in constructing the model was at least 50.
     3)  In cases where method 2 (above) could not be used (i.e., no models were available for a
         particular time-of-day) and the gap was less than 9 hours, the missing values were
         estimated by linear interpolation between the valid values at the ends of the gap.
       Table 6-5 provides descriptive statistics for 1-hour CO concentrations in each data set,
before and after estimating missing values.  The agreement between these statistics indicates that
the addition of the estimated missing-value concentrations did not significantly affect the
distribution of the hourly CO data.
      October, 2009
                                          6-9
Draft - Do Not Cite or Quote

-------
 1
 2
Table 6-5.  Descriptive Statistics for 1-hour CO Concentrations Reported by the Selected
            Denver and Los Angeles Monitors Before and After Estimation of Missing
            Values.
Area
Denver3
Los
Angelesb
Year
1995
2006
1997
2006
Missing
value
substitution
No
Yes
No
Yes
No
Yes
No
Yes
Samples
(n)
8697
8760
8672
8760
8302
8760
8275
8760
1-hour CO concentrations (ppm)
Mean
1.5
1.5
0.6
0.6
2.4
2.3
1.0
1.0
Std
1.2
1.2
0.4
0.4
2.2
2.2
0.9
0.90
Percentile
50
1.2
1.2
0.5
0.5
1.7
1.7
0.7
0.7
90
2.7
2.7
1.0
1.0
4.9
4.9
2.0
2.0
95
3.4
3.4
1.3
1.3
6.8
6.7
2.9
2.9
99
6.1
6.1
2.2
2.1
11.2
11.2
4.7
4.6
99.9
13.1
13.1
4.1
4.1
17.2
17.2
6.8
6.8
2nd
highest
16.4
16.4
4.6
4.6
18.8
18.8
8.2
8.2
Max
24.5
24.5
6.4
6.4
19.2
19.2
8.4
8.4
aSite ID 080310002
"Site ID 06037 1301
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
      6.1.4.3  Air Quality Adjustment to Simulate Just Meeting NAAQS
       In addition to modeling exposures based on recent air quality, exposures and resulting
dose were estimated for air quality conditions that just meet the current 8-hour CO NAAQS of 9
ppm.1  Because CO concentrations in recent years were significantly lower than the current
NAAQS, staff first selected an earlier year for each city (1995 for Denver and 1997 for Los
Angeles) to represent air quality conditions that were near the 8-hour CO standard.  Consistent
with the data adjustment approach employed in the previous draft CO exposure assessment
(Johnson et al., 2000) and risk and exposure assessments for other pollutants conducted in
support of other recent NAAQS reviews (e.g., US EPA, 2008c), as discussed in section 3.1.4
staff concluded (1) that the policy-relevant background levels of CO were negligible in each area
and (2) that the fixed-site monitoring data could be adjusted to simulate just meeting the current
CO standards by use  of a simple proportional adjustment of all hourly values. Consequently, the
following adjustment equation was employed:
                    COadj(m,/7) = (NAAQS/DV) x CO(m,h).
(6-1)
CO(m,h) is the 1-hour CO concentration at hour h for monitor m. It follows that COadj(w,/7) is
the adjusted CO concentration for hour h at monitor m through the use of the specific design
value (DV) for monitor m. Although the current 8-hour NAAQS for CO specifies a maximum
      1 The 8-hour CO NAAQS of 9 ppm was selected for purposes of simulating just meeting the CO NAAQS because it
      is the controlling standard from a control strategy development viewpoint.
      October, 2009                             6-10               Draft - Do Not Cite or Quote

-------
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20

21
22
23
      concentration of 9 ppm, which is not to be exceeded more than 1 time in a year, the NAAQS
      term in Equation 6-1 is equivalent to 9.4 ppm due to the application of a standard data rounding
      convention used in calculating design values2 (DVs) for CO (Laxton, 1990).
             The DVs for Denver for the year 1995 and for Los Angeles for 1997 were 9.5 ppm and
      15 ppm, respectively.  The Denver DV is calculated as the second-highest 8-hour CO
      concentration reported by monitor ID 080310002 for 1995.  The adjustment factor (or
      NAAQS/DV) that was applied equally to all 8,760 hourly ambient CO concentrations at that
      monitor is thus 9.4/9.5, or 0.99.  In a similar manner, the DV used in Los Angeles is the second-
      highest 8-hour CO concentration reported at monitor 060371301 for 1997, giving an ambient
      concentration adjustment factor of 9.4/15, or 0.63 which was applied equally to all 8,760 hourly
      ambient CO concentrations from the Los Angeles monitor.
             Table 6-6 lists descriptive statistics for the Denver and Los Angeles 1-hour data sets
      before and after adjustment. As expected, the adjusted data set for Denver 1995 is very similar
      to the unadjusted data set given that the adjustment factor used was close to unity.  For example,
      the maximum concentration was reduced from 24.5 ppm to 24.2 ppm.  The change in CO
      concentrations was greater as a result of adjusting the Los Angeles ambient data.  For example,
      the maximum CO concentration was reduced from 19.2 ppm to 12.0 ppm. The adjusted data
      sets, representing air quality simulated to just meet the current 8-hour CO NAAQS,for Denver
      and Los Angeles, exhibit their greatest differences at the extreme upper percentiles of the
      distribution (i.e., the 99.9th percentile and above).

      Table 6-6.  Descriptive Statistics for 1-hour Carbon Monoxide Concentrations Reported
                  by the Denver and Los Angeles Monitors Before and After Adjustment to
                  Simulate Just Meeting the Current 8-Hour CO NAAQS.
Area
Denver3
Los
Angelesb
Year
1995
1997
Adjusted
to just
meeting
NAAQS
No
Yes
No
Yes
1-hour CO concentrations (ppm)
Mean
1.5
1.5
2.3
1.5
Std
1.2
1.2
2.2
1.4
Percentile
50
1.2
1.2
1.7
1.1
90
2.7
2.7
4.9
3.1
95
3.4
3.4
6.7
4.2
99
6.1
6.0
11.2
7.0
99.9
13.1
12.9
17.2
10.8
ond
highest
16.4
16.2
18.8
11.8
Max
24.5
24.2
19.2
12.0
aSite ID 08031 0002
bSite ID 060371 301
24
       A design value is a statistic that describes the air quality status of a given area or monitor relative to the level of the
      NAAQS. For the CO 8-hour standard, the design value is the second highest daily, non-overlapping, maximum 8-
      hour average concentration over a year. The design value for the 1-hour standard is the second highest daily
      maximum 1-hour average concentration over a year.  The latest update (2007-2008) on the CO design values can be
      found at: http://www.epa.gov/airtrends/pdfs/dv co 2006 2008.pdf
      October, 2009
                                                6-11
Draft - Do Not Cite or Quote

-------
 1         6.1.4.4  Meteorological Stations
 2          A few algorithms within APEX require meteorological data (primarily temperature) from
 3    stations located within the study area.  For the analyses described in this report, hourly
 4    temperature data were obtained from meteorological stations located at or near the fixed-site CO
 5    monitor specified for each study area.  Table 6-7 identifies the meteorological stations used and
 6    selected site characteristics.
 7    Table 6-7.
       Site  Characteristics of  Meteorological  Monitoring  Stations  Selected  to
       Represent the Denver and Los Angeles Study Areas.
Site Characteristic
Site ID
Street address
Town
Latitude
Longitude
Elevation (above sea
level), meters
Sample collection
frequency
Reporting agency
Denver
080310002
2105 Broadway
Denver
39.751184
-104.987625
1593
1 hour
Colorado Department
of Public Health and
Environment
Los Angeles Site 1
060374002
3648 N. Long Beach
Blvd.
Long Beach, CA
33.823760
-118.189210
6
1 hour
South Coast Air
Quality Management
District
Los Angeles Site 2

Daugherty Field
Long Beach, CA
33.81667
-118.15
9.4
1 hour

 9
10

11
12

13
14
15
16
17
18
19
20
21
22
23
24
25
26
The procedure used for generating a complete meteorological data set was as follows.

Staff first checked on the availability of hourly temperature data for the specified years at
each CO fixed-site monitor specified for each study area.

For Los Angeles, temperature data were not available for the specified CO monitoring
site (Site ID 060371301). Consequently, we evaluated two alternative sites: Site 1
(located at CO monitoring site 060374002) and site 2 (located at Daugherty Field), which
are approximately 12 km and  15 km, respectively, from site 060371301.  These sites
were separated from each other by a distance of only 3.6 km. Temperature data for the
two years considered for the exposure analysis (i.e., 1997 and 2006) were reported by
both sites. Because Site 1 had fewer missing values for 2006, it was selected as the
primary meteorological site to represent the Los Angeles area for that year. Temperature
data from Site 2 were used to fill the single missing value in the Site 1 data set for year
2006. However, the 1997 data contained 2,263 missing values for Site 1  and only 9
missing values for Site 2. Consequently, Site 2 was selected as the primary
meteorological site to represent the Los Angeles area for 1997. Two of the nine missing
values from Site 2 were available from Site 1; staff replaced these two missing values
with corresponding values from Site 1. A linear interpolation, using the values at the end
      October, 2009
                                   6-12
Draft - Do Not Cite or Quote

-------
 1          of existing gaps, was used to fill in the seven remaining missing values in the 1997 data
 2          set in Site 2.3
 3       •  For Denver, temperature data were available for the CO ambient monitoring site used
 4          (i.e., Site ID: 080310002) and both years considered in the exposure assessment (1995
 5          and 2006). Linear interpolation was used to fill 18 of 41 missing values considering the
 6          1995 data and 11 of 11 missing values considering the 2006 data.  In one instance, the
 7          gap in hourly temperature data was 23 continuous hours.  Staff considered a linear
 8          interpolation to be inappropriate in this situation because it would likely not produce
 9          reasonable estimates of the variability in temperature (particularly the daily maximum)
10          occurring during the 23-hour gap. An alternative approach was used in which the
11          temperature data for corresponding hours in the previous day were substituted for the
12          missing data.
13         6.1.5  Microenvironments
14          As mentioned earlier, two general microenvironments were defined for the exposure
15    analyses: in vehicle and all other. Each microenvironment was defined as an aggregation of the
16    location codes used in CHAD to specify where each exposure event occurred. Note that location
17    is interpreted here as referring to the microenvironmental characteristics of a place (e.g., indoors
18    at school), rather than the particular geographic location.  Appendix E provides the mapping of
19    the CHAD location codes to the two APEX modeled microenvironments.
20          The factors approach was used to estimate a CO concentration in the two
21    microenvironments for each hour of the specified study period (Equation  5-1,  section 5.3.4). The
22    penetration factor for all microenvironments was set equal to 1 for both scenarios (see draft ISA,
23    section 3.6.5.1 for all indoor microenvironments and section 5.4.2.1 for the in-vehicle
24    microenvironment).  The proximity factor was set equal to 1 for scenario A  in both
25    microenvironments modeled (i.e., in vehicle and all other) and equal to 2 for the in-vehicle
26    microenvironment and 1 for the all other microenvironment in scenario B. The values used in
27    representing in-vehicle concentrations for scenario B were based on staffs evaluation of
28    measurement  studies that simultaneously measured CO concentrations within  motor vehicles and
29    at nearby fixed-site monitors (see section 5.4.2.2).
30          Staff did not adjust  ambient concentrations to estimate near-road microenvironmental CO
31    concentrations.  This was because the ambient CO concentrations from the two monitors used in
32    this assessment had the highest hourly CO concentrations recorded in each study area, and based
33    on the AQS noted monitoring scale, were already designated to capture near road CO
34    concentrations (i.e., microscale and middle scale). While higher near-road CO concentrations
35    are possible, staff judged that these ambient data would already represent upper percentile
            3 We used PROC EXPAND along with the JOIN option in SAS. The JOIN option fits a continuous curve
      to the data by connecting successive straight line segments.
      October, 2009                             6-13               Draft - Do Not Cite or Quote

-------
 1    ambient CO concentrations experienced by most persons residing or spending time near
 2    roadways in each study area.

 3         6.1.6   Time/Activity Patterns
 4          APEX constructs a 365-day longitudinal diary for each simulated individual by selecting
 5    24-hour diaries from those available in CHAD. In performing the exposure assessments
 6    described in this report, all available diaries for persons above age 17 in the CHAD database
 7    were used regardless of particular commuting patterns.

 8         6.1.6.1  Construction of Longitudinal Diaries
 9          As discussed in section 5.3.3, APEX provides a longitudinal diary assembly algorithm
10    that enables the user to create composite diaries that reflect the tendency of individuals to repeat
11    day-to-day activities.  The user specifies values for two statistical variables (D and A) that relate
12    to a key daily variable, typically the time spent per day in a particular microenvironment (e.g., in
13    a motor vehicle). The D  statistic reflects the relative importance of intra- and inter-personal
14    variance within the selected key daily variable. The A variable quantifies the day-to-day
15    autocorrelation in the selected key daily variable.  APEX then constructs composite diaries that
16    exhibit the statistical properties defined by the specified values ofD and A.
17          In this exposure assessment, we used the longitudinal diary algorithm to construct year-
18    long activity patterns for each simulated individual to reflect the day-to-day correlation of time
19    spent inside motor vehicles. Each diary day in the CHAD database was tagged with the number
20    of minutes spent in the vehicle microenvironment. Parameter settings ofD = 0.31 and A = 0.19
21    were specified to control the day-to-day repetition of time  spent in motor vehicles in the
22    constructed composite diaries. These particular D and A values were obtained from Isaacs et al.
23    (2009) (see Appendix F).
24          In selecting particular diaries to represent the simulated population, the CHAD data are
25    categorized or separated by APEX into data pools. In Scenario A and B, the pools were defined
                                                                     o
26    by three ranges for the maximum temperature of the diary  day (< 55.0 F, between 55.0 and 83.9
      o            o
27    F, and >84.0 F) and two day-types (i.e., weekend and week day); thus, there were 3x2 = 6
28    diary pools.  The window for age was set at 15%. For example, diaries can be selected for a
29    simulated individual of age 60 from CHAD individuals ranging from ages 51 though 69.

30         6.2  EXPOSURE AND DOSE ESTIMATES AND RISK CHARACTERIZATION
31         6.2.1   Denver - Scenarios A and B
32          Output files for APEX4.3 runs were generated for various combinations of calendar year
33    (1995 and 2006), exposure  scenario (A or B), and air quality condition (as is or just meeting the
      October, 2009                            6-14                Draft - Do Not Cite or Quote

-------
 1    8-hour NAAQS) in the Denver study area. These results are summarized in a series of tables that
 2    follow (Tables 6-8 through Table 6-14).
 3          Table 6-8 presents estimates for the number of person-days during the calendar year in
 4    which members of the Denver population-at-risk experienced a 1-hour daily maximum CO
 5    exposure at or above each of the indicated CO concentrations.  Results are presented for
 6    Scenarios A and B for each of two air quality conditions: as is conditions represented by 2006
 7    monitoring data and just meeting conditions as represented by  1995 monitoring data adjusted to
 8    simulate just meeting the 8-hour NAAQS. The maximum possible value for person-days of
 9    exposure is about 23.4 million person-days - the product of the estimated population-at-risk
10    (about 64,000) and the number of days in the specified exposure period (365).
11          Using a format similar to Table 6-8, Table 6-9 presents estimates of the number of
12    persons in the population-at risk that experienced at least one 1-hour daily maximum CO
13    exposure at or above each of the indicated CO concentrations.  In this table, the maximum
14    possible value is about 64,000 people - the estimated number of people in the population-at-risk.
15    Thus, each person can be counted no more than once in determining the value in Table 6-9.
16          Table 6-10 and Table 6-11  are comparable to Table 6-8 and Table 6-9,  respectively
17    though they provide estimates for 8-hour daily maximum exposures rather than 1-hour daily
18    maximum exposures. Again, the maximum possible value for person-days of exposure in Table
19    6-10 is about 23.4 million person-days; the maximum possible value of persons exposed in Table
20    6-11 is about 64,000  people.
21          Table 6-12 and 6-13 are also analogous to the prior tables,  though they present estimates
22    of the number of person-days in which members of the Denver population-at-risk experienced a
23    daily maximum end-of-hour COHb level at or above each of the indicated levels. Again, the
24    maximum possible value for person-days is about 23.4 million for Table 6-12 and the maximum
25    number of persons experiencing a maximum COHb level in the year in the population-at-risk is
26    about 64,000 for Table 6-13.
27          Table 6-14 provides estimates for the mean number of days per person  in which the
28    person experienced a daily maximum end-of-hour COHb level at or above each of the indicated
29    levels. These values were calculated by dividing the values listed  in Table 6-12 (total number of
30    person-days) by the comparable values in Table 6-13 (number of people); hence the maximum
31    possible value is 365.
32          Table 6-9 through Table 6-15 exhibit general patterns that are consistent with the input
33    data and parameter settings specified for the associated model runs. The Scenario B values are
34    greater than the comparable Scenario A values because Scenario B specifically accounts for in-
35    vehicle microenvironmental CO concentrations, while Scenario A assumes in-vehicle CO
36    concentrations (and all other microenvironments) are equal to the ambient fixed-site monitor
      October, 2009                             6-15               Draft - Do Not Cite or Quote

-------
 1    concentrations. However, the effect of better accounting for in-vehicle exposures (Scenario B)
 2    to the overall estimated population exposures and doses is primarily limited to differences
 3    observed in the upper percentiles of the distribution. For example, about 4.8% of simulated
 4    individuals are estimated to experience an end-of-hour COHb concentration at or above 1.0%
 5    when considering Scenario A and the as is air quality (Table 6-13). When considering Scenario
 6    B and as is air quality, about 7.0% of the population experiences an end-of-hour COHb level at
 7    or above 1.0%. Note however that when considering either scenario A or B, these data also
 8    indicate that between 93 and 95% of the population experienced an end-of-hour COHb
 9    concentration below 1.0%, suggesting that only an additional 2% of the population was affected
10    by the addition of the in-vehicle microenvironment. The effect of accounting for in-vehicle
11    concentrations (i.e,  Scenario A results compared with  Scenario B) was greater when considering
12    the air quality adjusted to just meeting the current standard. For example, approximately 11% of
13    simulated individuals were estimated to experience at  least one end-of-hour COHb above 2.5%
14    when considering Scenario B compared with over an order of magnitude fewer persons when
15    considering Scenario A.
16           The estimated number of person-days and persons considering air quality just meeting the
17    current standard is greater than that estimated considering as is air quality at comparable target
18    concentrations. For example, the entire simulated population was estimated to experience at
19    least one end-of-hour COHb concentration at or above 1.5% when considering the air quality just
20    meeting the current standard and Scenario A. This same COHb level was only experienced by
21    approximately 0.2% of the population when considering the as is air quality and Scenario A
22    (Table 6-13). This of course is because the monitoring data used to represent ambient as is air
23    quality have significantly lower CO levels than the data used to represent just meeting the 8-hour
24    standard conditions in Denver (i.e., the adjusted 1995 air quality simulation).
25           As described in sections 2.6 and 4.2, our characterization of health risk for CO in this
26    assessment focuses  on several risk metrics involving comparison of estimated COHb levels in
27    the adult CHD population to potential health benchmarks (1.5-3.0% COHb). Assessment results
28    involving this comparison for the Denver study area are emphasized in bold type in Tables 6-12
29    through 6-14. Well below 1 percent of the at-risk population was estimated to reach COHb
30    levels at or above 1.5% under as is conditions in both  scenarios. Under air quality conditions
31    just meeting the current standard, substantially greater percentages of the population were
32    estimated to reach COHb levels at or above all of the potential health benchmark levels, with
33    100% of the at-risk  population estimated to reach COHb levels > 1.5% in Scenario B for these
34    conditions (Table 6-13).
      October, 2009                             6-16               Draft - Do Not Cite or Quote

-------
1
2
    Table 6-8. Number of Person-days for Adults with Coronary Heart Disease (CHD) in the
              Denver Study Area Estimated to Experience a 1-hour Daily Maximum CO
              Exposure at or Above the Specified Concentration.
CO
Concentration
(ppm)
80
60
40
30
20
15
12
9
6
3
0
Number of person-days
"As Is" Air Quality (2006)a
Scenario A
0
0
0
0
0
0
0
0
6.4E+04
9.6E+05
2.3E+07
Scenario B
0
0
0
0
0
0
2.0E+03
1.1E+04
1.4E+05
1.4E+06
2.3E+07
"Just Meeting" Air Quality (1995)b
Scenario A
0
0
0
0
6.4E+04
1.3E+05
3.2E+05
8.3E+05
2.3E+06
1.0E+07
2.3E+07
Scenario B
0
0
2.8E+03
1.5E+04
1.0E+05
2.4E+05
5.9E+05
1.1E+06
3.2E+06
1.3E+07
2.3E+07
3 "As Is" air quality data are for the year 2006.
b Air quality data for the year 1 995 adjusted downwards to just meet 9 ppm, 8-hour NAAQS.
Note: Total adult population with CHD in the Denver study area is estimated to be about 64,000.
4
5
6
    Table 6-9. Number of Adults with Coronary Heart Disease (CHD) in the Denver Study
              Area Estimated to Experience a 1-hour Daily Maximum CO Exposure at or
              Above the Specified Concentration.
CO
Concentration
(ppm)
80
60
40
30
20
15
12
9
6
3
0
Number of persons
"As Is" Air Quality (2006)a
Scenario A
0
0
0
0
0
0
0
0
64,000
64,000
64,000
Scenario B
0
0
0
0
0
0
2,000
9,700
64,000
64,000
64,000
"Just Meeting" Air Quality (1995)b
Scenario A
0
0
0
0
64,000
64,000
64,000
64,000
64,000
64,000
64,000
Scenario B
0
0
2,800
14,000
64,000
64,000
64,000
64,000
64,000
64,000
64,000
a "As Is" air quality data is for the year 2006.
b Air quality data for the year 1 995 adjusted downwards to just meet 9 ppm, 8-hour NAAQS.
Note: Total adult population with CHD in the Denver study area is estimated to be about 64,000.
    October, 2009
                                           6-17
Draft - Do Not Cite or Quote

-------
1
2
    Table 6-10. Number of Person-days for Adults with Coronary Heart Disease (CHD) in the
               Denver Study Area Estimated to Experience an 8-hour Daily Maximum CO
               Exposure at or Above the Specified Concentration.
CO
Concentration
(ppm)
80
60
40
30
20
15
12
9
6
3
0
Number of person-days
"As Is" Air Quality (2006)a
Scenario A
0
0
0
0
0
0
0
0
0
1.3E+05
2.3E+07
Scenario B
0
0
0
0
0
0
0
0
7.4E+01
2.1E+05
2.3E+07
"Just Meeting" Air Quality (1995)b
Scenario A
0
0
0
0
0
0
0
1.3E+05
4.5E+05
4.1E+06
2.3E+07
Scenario B
0
0
0
0
3.4E+02
2.4E+03
2.0E+04
1.5E+05
6.1E+05
5.0E+06
2.3E+07
3 "As Is" air quality data is for the year 2006.
b Air quality data for the year 1 995 adjusted downwards to just meet 9 ppm, 8-hour NAAQS.
Note: Total adult population with CHD in the Denver study area is estimated to be about 64,000.
4
5
6
    Table 6-11. Number of Adults with Coronary Heart Disease (CHD) in the Denver Study
               Area Estimated to Experience an 8-hour Daily Maximum CO Exposure at or
               Above the Specified Concentration.
CO
Concentration
(ppm)
80
60
40
30
20
15
12
9
6
3
0
Number of persons
"As Is" Air Quality (2006)a
Scenario A
0
0
0
0
0
0
0
0
0
64,000
64,000
Scenario B
0
0
0
0
0
0
0
0
74
64,000
64,000
"Just Meeting" Air Quality (1995)b
Scenario A
0
0
0
0
0
0
0
64,000
64,000
64,000
64,000
Scenario B
0
0
0
0
340
2,400
17,000
64,000
64,000
64,000
64,000
3 "As Is" air quality data is for the year 2006.
b Air quality data for the year 1 995 adjusted downwards to just meet 9 ppm, 8-hour NAAQS.
Note: Total adult population with CHD in the Denver study area is estimated to be about 64,000.
    October, 2009
                                           6-18
Draft - Do Not Cite or Quote

-------
1
2
    Table 6-12. Number of Person-days for Adults with Coronary Heart Disease (CHD) in the
               Denver Study Area Estimated to Experience a Daily Maximum End-of-hour
               COHb Level at or Above the Specified Concentration.
COHb concentration
(percent)
8.0
6.0
5.0
4.0
3.0
2.5
2.0
1.5
1.0
0
Number of person-days
"As Is" Air Quality (2006)a
Scenario A
0
0
0
0
0
1.9E+03
3.9E+03
9.2E+03
2.0E+05
2.3E+07
Scenario B
0
0
0
0
0
1.9E+03
3.9E+03
9.3E+03
2.1E+05
2.3E+07
"Just Meeting" Air Quality
(1995)b
Scenario A
0
0
0
0
4.9E+01
3.3E+03
2.6E+04
1.8E+05
1.4E+06
2.3E+07
Scenario B
0
0
0
2.5E+01
1.8E+03
1.1E+04
5.8E+04
2.4E+05
1.7E+06
2.3E+07
a "As Is" air quality data is for the year 2006.
b Air quality data for the year 1 995 adjusted downwards to just meet 9 ppm, 8-hour NAAQS.
Note: Total adult population with CHD in the Denver study area is estimated to be about 64,000.
4
5
6
    Table 6-13. Number (and Percent) of Adults with Coronary Heart Disease (CHD) in the
               Denver Study Area Estimated to Experience a Daily Maximum End-of-hour
               COHb Level at or Above the Specified Concentration.
COHb
concentration
(percent)
8.0
6.0
5.0
4.0
3.0
2.5
2.0
1.5
1.0
0
Number of persons (percent0)
"As Is" Air Quality (2006)a
Scenario A
0
0
0
0
0
12(<0.1)
12(<0.1)
160(0.2)
3,100(5)
64,000(100)
Scenario B
0
0
0
0
0
12(<0.1)
12(<0.1)
160(0.2)
4,500 (7)
64,000(100)
"Just Meeting" Air Quality (1995)b
Scenario A
0
0
0
0
12(<0.1)
250 (0.4)
16,000(26)
64,000(100)
64,000(100)
64,000(100)
Scenario B
0
0
0
25(<0.1)
1,700(3)
7,100(11)
36,000 (56)
64,000(100)
64,000 (100)
64,000 (100)
a "As Is" air quality data are for the year 2006.
b Air quality data for the year 1 995 adjusted downwards to just meet 9 ppm, 8-hour NAAQS.
c Percent of adult CHD population.
Note: Total adult population with CHD in the Denver study area is estimated to be about 64,000.
    October, 2009
                                           6-19
Draft - Do Not Cite or Quote

-------
1   Table 6-14. Estimated Average Number of Days with a Daily Maximum End-of-hour
2              COHb Level At or Above the Specified Concentration Per Adult With
3              Coronary Heart Disease (CHD) in the Denver Study Area.
COHb
concentration
(percent)
8.0
6.0
5.0
4.0
3.0
2.5
2.0
1.5
1.0
0
Average of person-days/person
"As Is" Air Quality (2006)a
Scenario A
0
0
0
0
0
<0.1
<0.1
0.1
3.1
365
Scenario B
0
0
0
0
0
<0.1
<0.1
0.1
3.2
365
"Just Meeting" Air Quality (1995)b
Scenario A
0
0
0
0
<0.1
<0.1
0.4
2.9
22
365
Scenario B
0
0
0
0
<0.1
0.2
0.9
3.8
27
365
3 "As Is" air quality data is for the year 2006.
b Air quality data for the year 1 995 adjusted downwards to just meet 9 ppm, 8-hour NAAQS.
Note: Total adult population with CHD in the Denver study area is estimated to be about 64,000.
    October, 2009
6-20
Draft - Do Not Cite or Quote

-------
 1         6.2.2   Los Angeles - Scenarios A and B
 2          Output files for APEX4.3 runs were generated for various combinations of calendar year
 3    (1997 and 2006), exposure scenario (A or B), and air quality condition (as is or just meeting the
 4    8-hour NAAQS) in the Los Angeles study area. These results are summarized in a series of
 5    tables that follow (Tables 6-15 through Table 6-21).
 6          The same general patterns identified above using the Denver results were observed in the
 7    Los Angles results when considering either modeling scenario and air quality condition. For
 8    example in Scenario A, approximately 3% of the simulated population experienced  an end-of-
 9    hour COHb level at or above 1.5% considering the as is air quality compared with about 67% of
10    the population having the same or greater COHb level when using air quality adjusted to just
11    meeting the current standard (Table 6-20). This is as expected given the fixed-site input data for
12    2006 used to represent  ambient as is conditions in Los Angeles having lower CO concentrations
13    than the data used to represent just meeting the current standard (Table 6-5 and 6-6). Similarly, a
14    greater number of persons and person-days are estimated when considering scenario B compared
15    with scenario A.  For example, when considering the as is air quality, approximately 3% of the
16    simulated population experienced an end-of-hour COHb level at or above 1.5% considering
17    Scenario A compared with about 8% of the simulated population when separately accounting for
18    in-vehicle exposures (Scenario B). This is also as expected because,  as in the case of Denver,
19    Scenario B assumes CO concentrations in the Los Angeles in-vehicle microenvironment were
20    twice the ambient concentrations, while Scenario A assumes the in-vehicle concentrations (and
21    all other microenvironments) are equal to the ambient concentrations measured at the fixed-site
22    monitor.
23          As described in sections 2.6 and 4.2, our characterization of health risk for CO in this
24    assessment focuses on several risk metrics involving comparison of COHb levels in adults with
25    CFID to potential health benchmarks (1.5-3.0% COHb).  Assessment results involving this
26    comparison for the Los Angeles study area are emphasized in bold type in Tables 6-19 - 6-21.
27    Fewer than 1 percent of the at-risk population was estimated to reach COHB levels at or above
28    2.0% under as is conditions when considering both scenarios, with 3% and 8% at or above 1.5%
29    COHb in Scenarios A and B, respectively. Under air quality conditions just meeting the current
30    standards, substantially greater percentages of the population were estimated to reach COHb
31    levels at or above all of the potential health benchmark levels, with 67% and 80% of the at-risk
32    population estimated reach COHb levels  > 1.5% in Scenarios A and B, respectively, for these
33    conditions (Table 6-20).
34
      October, 2009                             6-21               Draft - Do Not Cite or Quote

-------
1
2
    Table 6-15. Number of Person-Days for Adults with Coronary Heart Disease (CHD) in the
               Los Angeles Study Area Estimated to Experience a 1-hour Daily Maximum
               CO Exposure At or Above the Specified Concentration.
CO
Concentration
(ppm)
80
60
40
30
20
15
12
9
6
3
0
Number of person-days
"As Is" Air Quality (2006)a
Scenario A
0
0
0
0
0
0
0
0
1.9E+06
1.5E+07
5.7E+-7
Scenario B
0
0
0
0
0
9.2E+03
6.3E+04
3.3E+05
2.8E+06
1.7E+07
5.7E+07
"Just Meeting" Air Quality (1997)b
Scenario A
0
0
0
0
0
0
1.6E+05
2.4E+06
7.9E+06
2.4E+07
5.7E+07
Scenario B
0
0
0
0
1.3E+04
1.4E+05
6.1E+05
3.2E+06
9.2E+06
2.5E+07
5.7E+07
3 "As Is" air quality data are for the year 2006.
b Air quality data for the year 1 997 adjusted downwards to just meet 9 ppm, 8-hour NAAQS.
Note: Total adult population with CHD in the Los Angeles study area is estimated to be about 160,000.
4
5
6
    Table 6-16. Number of Adults with Coronary Heart Disease (CHD) in the Los Angeles
               Study Area Estimated to Experience a 1-hour Daily Maximum CO Exposure
               At or Above the Specified Concentration.
CO
Concentration
(ppm)
80
60
40
30
20
15
12
9
6
3
0
Number of persons
"As Is" Air Quality (2006)a
Scenario A
0
0
0
0
0
0
0
0
160,000
160,000
160,000
Scenario B
0
0
0
0
0
8,800
44,000
110,000
160,000
160,000
160,000
"Just Meeting" Air Quality (1997)b
Scenario A
0
0
0
0
0
0
160,000
160,000
160,000
160,000
160,000
Scenario B
0
0
0
0
12,000
75,000
160,000
160,000
160,000
160,000
160,000
3 "As Is" air quality data is for the year 2006.
b Air quality data for the year 1 997 adjusted downwards to just meet 9 ppm, 8-hour NAAQS.
Note: Total adult population with CHD in the Los Angeles study area is estimated to be 160,000.
    October, 2009
                                           6-22
Draft - Do Not Cite or Quote

-------
1   Table 6-17. Number Of Person-Days For Adults with Coronary Heart Disease (CHD) in
2              the Los Angeles Study Area Estimated to Experience an 8-hour Daily
3              Maximum CO Exposure At or Above the Specified Concentration.
CO
Concentration
(ppm)
80
60
40
30
20
15
12
9
6
3
0
Number of person-days
"As Is" Air Quality (2006)a
Scenario A
0
0
0
0
0
0
0
0
1.6E+05
6.3E+06
5.7E+07
Scenario B
0
0
0
0
0
0
3.2E+01
2.9E+03
2.3E+05
7.0E+06
5.7E+07
"Just Meeting" Air Quality (1997)b
Scenario A
0
0
0
0
0
0
0
3.1E+05
2.2E+06
1.6E+07
5.7E+07
Scenario B
0
0
0
0
0
6.8E+02
8.5E+03
3.7E+05
2.7E+06
1.6E+07
5.7E+07
3 "As Is" air quality data is for the year 2006.
b Air quality data for the year 1 997 adjusted downwards to just meet 9 ppm, 8-hour NAAQS.
Note: Total adult population with CHD in the Los Angeles study area is estimated to be about 160,000.
    October, 2009
6-23
Draft - Do Not Cite or Quote

-------
2
3
4
Table 6-18. Number of Adults with Coronary Heart Disease (CHD) in the Los Angeles
           Estimated to Experience an 8-hour Daily Maximum CO Exposure At or
           Above the Specified Concentration.
CO
Concentration
(ppm)
80
60
40
30
20
15
12
9
6
3
0
Number of persons
"As Is" Air Quality (2006)a
Scenario A
0
0
0
0
0
0
0
0
160,000
160,000
160,000
Scenario B
0
0
0
0
0
0
32
2,300
160,000
160,000
160,000
"Just Meeting" Air Quality (1997)b
Scenario A
0
0
0
0
0
0
0
160,000
160,000
160,000
160,000
Scenario B
0
0
0
0
0
610
6,900
160,000
160,000
160,000
160,000
3 "As Is" air quality data is for the year 2006.
b Air quality data for the year 1 997 adjusted downwards to just meet 9 ppm, 8-hour NAAQS.
Note: Total adult population with CHD in the Los Angeles study area is estimated to be about 160,000.
5
6
1
Table 6-19. Number of Person-Days For Adults With Coronary Heart Disease (CHD) in
           the Los Angeles Study Area Estimated to Experience a Daily Maximum End-
           of-hour COHb Level At or Above the Specified Concentration.
COHb
concentration
(percent)
8.0
6.0
5.0
4.0
3.0
2.5
2.0
1.5
1.0
0
Number of person-days
"As Is" Air Quality (2006)a
Scenario A
0
0
0
0
0
0
1.6E+02
2.8E+04
1.6E+06
5.7E+07
Scenario B
0
0
0
0
0
0
3.2E+02
4.1E+04
1.8E+06
5.7E+07
"Just Meeting" Air Quality (1997)b
Scenario A
0
0
0
0
0
6.5E+01
6.1E+03
3.2E+05
5.1E+06
5.7E+07
Scenario B
0
0
0
0
0
6.8E+02
1.6E+04
4.8E+05
5.9E+0
5.7E+07
3 "As Is" air quality data is for the year 2006.
b Air quality data for the year 1 997 adjusted downwards to just meet 9 ppm, 8-hour NAAQS.
Note: Total adult population with CHD in the Los Angeles study area is estimated to be about 160,000.
    October, 2009
                                      6-24
Draft - Do Not Cite or Quote

-------
1
2
    Table 6-20. Number (and Percent) of Adults with Coronary Heart Disease (CHD) in the
               Los Angeles Study Area Estimated to Experience a Daily Maximum End-of-
               hour COHb Level At or Above the Specified Concentration.
COHb
concentration
(percent)
8.0
6.0
5.0
4.0
3.0
2.5
2.0
1.5
1.0
0
Number of persons (percent)0
"As Is" Air Quality (2006)a
Scenario A
0
0
0
0
0
0
130(<0.1)
4,600 (3)
147,000 (93)
160,000(100)
Scenario B
0
0
0
0
0
0
290 (0.2)
13,000(8)
150,000(97)
160,000 (100)
"Just Meeting" Air Quality (1997)b
Scenario A
0
0
0
0
0
65(<0.1)
3,900 (2)
110,000(67)
160,000(100)
160,000(100)
Scenario B
0
0
0
0
0
680 (0.4)
10,000(7)
127,000(80)
160,000(100)
160,000(100)
a "As Is" air quality data are for the year 2006.
b Air quality data for the year 1997 adjusted downwards to just meet 9 ppm, 8-hour NAAQS.
0 Percent of adult CHD population.
Note: Total adult population with CHD in the Los Angeles study area is estimated to be about 160,000.
4
5
6
    Table 6-21  Estimated Average Number of Days with a Daily Maximum End-of-hour
               COHb Level At or Above the Specified Concentration Per Adult with
               Coronary Heart Disease (CHD) in the Los Angeles Study Area.
COHb
concentration
(percent)
8.0
6.0
5.0
4.0
3.0
2.5
2.0
1.5
1.0
0
Average of person-days/person
"As Is" Air Quality (2006)a
Scenario A
0
0
0
0
0
0
<0.1
0.2
9.9
365
Scenario B
0
0
0
0
0
0
<0.1
0.3
12
365
"Just Meeting" Air Quality (1997)b
Scenario A
0
0
0
0
0
0
<0.1
2.1
32
365
Scenario B
0
0
0
0
0
0
0.1
3.0
37
365
a "As Is" air quality data are for the year 2006.
b Air quality data for the year 1 997 adjusted downwards to just meet 9 ppm, 8-hour NAAQS.
    October, 2009
                                          6-25
Draft - Do Not Cite or Quote

-------
 1         6.2.3  Comparison of Denver and Los Angeles Estimates for End-of-hour COHb
 2               Levels
 3          We can best compare Denver and Los Angeles estimates using the results provided in
 4    Tables 6-13 and 6-20. These tables provide estimates for the percentage of people in the
 5    population-at-risk that are estimated to experience a daily maximum end-of-hour COHb level at
 6    or above the specified value. For either scenario, the percentage of people with COHb levels at
 7    or above 1.5 percent is greater for Los Angeles than for Denver when considering the as is air
 8    quality. For example, in Los Angeles it was estimated that 3 and 8 percent of persons
 9    experienced an end-of hour COHb level at or above this level for Scenarios A and B,
10    respectively, for the as is case.  In Denver, virtually no persons (0.2%) experienced an end-of-
11    hour COHb at or above 1.5% associated with the as is case. This pattern is consistent with the
12    fact that the monitoring data used to represent as is conditions in Los Angeles exhibited higher
13    CO levels than Denver (Table 6-5).
14          The pattern noted above is reversed when considering air quality that just meets the
15    current standard. For example, the estimated percent of persons having a COHb level at or
16    above 2 percent for Scenarios A and B are 26 percent and 56 percent, respectively for Denver
17    (Table 6-13); the comparable estimates for Los Angeles are about 2 percent for Scenario A and 7
18    percent for Scenario B (Table 6-20). This pattern is also expected, since the air quality data used
19    to represent just meeting the current standard in Denver has higher CO levels at the upper
20    percentiles of the distribution than the Los Angeles data (Table 6-6).

21         6.3   COMPARISON OF COHB ESTIMATES OBTAINED FROM THE 2000
22              PNEM/CO AND DRAFT 2009 APEX/CO ASSESSMENTS
23          As part of the review of the CO NAAQS initiated in 1997, a draft CO exposure
24    assessment was prepared (Johnson, et al., 1999).  Subsequent to the discontinuation of that CO
25    NAAQS review, a revised document was completed (Johnson et al., 2000).  The 2000 document
26    was subsequently subject to peer review by several exposure modeling experts (SAIC, 2001).
27    The 2000 CO population exposure assessment was conducted for Denver using air quality data
28    for 1995 and for Los  Angeles using air quality data for 1997. The exposure and dose estimates
29    were obtained by applying pNEM/CO, a predecessor to APEX, to adults with ischemic heart
30    disease residing in a defined study area within each  city (Johnson et al., 2000). As part of
31    current (2009) draft exposure assessment described in section 6.1, staff has again used APEX to
32    estimate CO exposures and resulting COHb levels in a portion of Denver using 1995 air quality
33    data and in a portion  of Los Angeles using 1997 air  quality data.  In this case,  the population-at-
34    risk was defined as adults with  CHD which is approximately equivalent to the ischemic heart
35    disease definition used in the prior review.
      October, 2009                         6-26            Draft - Do Not Cite or Quote

-------
 1          In comparing the earlier pNEM/CO results with the exposure and dose estimates obtained
 2    from the current draft APEX/CO assessment for the same cities and years, it is important to
 3    understand the differences between the methodologies employed in the two assessments. The
 4    methods and results associated with the 2000 pNEM/CO analysis are described in detail in a
 5    report by Johnson et al. (2000). The methods used in the current (2009) draft APEX exposure
 6    assessment are described above in section 6.1. Section 6.3.1 provides a brief discussion of the
 7    important differences between the two assessments that may account for some of the observed
 8    differences in  the exposure estimates. Section 6.3.2 presents estimates of COHb levels in adults
 9    with CHD obtained from the two assessments.

10         6.3.1  Important Differences Between the 2000 pNEM/CO and 2009 draft APEX/CO
11               Exposure/Dose Assessments
12          In the 2000 pNEM/CO assessment, the Denver study area was defined to include the
13    census tracts located within 10 km of each of six fixed-site monitors in the Denver metropolitan
14    area. Air quality data for 1995 reported by these fixed-site monitors were used to represent
15    "existing conditions" in the study area. Because the second non-overlapping 8-hour maximum
16    CO concentration (design value) equaled 9.5 ppm, the existing conditions in Denver for 1995
17    were considered to approximate just meeting the 8-hour standard in which the DV equals 9.4
18    ppm.
19          In a similar manner, the Los Angeles study  area was defined to include all census tracts
20    within 10 km often fixed-site monitors in the Los Angeles metropolitan area.  Air quality data
21    for 1997 reported by these fixed-site monitors were used to represent "existing conditions" in the
22    study area. Because the 1997 CO levels in Los Angeles exceeded the 8-hour NAAQS, the
23    concentrations at each monitoring site were adjusted downwards so that the concentrations
24    associated with the DV site exactly met the 8-hour NAAQS (i.e., the adjusted maximum CO
25    concentration  at the DV site equaled 9.4 ppm).
26          Note that the air quality data used in the pNEM/CO assessments for each city included
27    data from multiple sites (6 in Denver, 10 in Los Angeles) that represented areas of varying CO
28    levels. The monitoring data associated with the DV (highest CO) site were  only applied to those
29    people who resided in the circular area within 10 km of that particular monitor.4 Data from other
30    (lower CO) sites were applied to the people in the study area who resided within the 10 km
31    circular areas centered on those sites.
32          In the 2009 draft APEX/CO assessment, the Denver and Los Angeles study areas are
33    each defined to include all census tracts within 20 km of a single fixed-site monitor. This
            4 In Denver, there was one instance where two monitors in close proximity to each other were
      geographically combined to represent a single composite monitoring location.

      October, 2009                          6-27           Draft - Do Not Cite or Quote

-------
 1    monitor is the DV site for the specified year of the assessment (1995 for Denver, 1997 for Los
 2    Angeles). Consequently, monitoring data associated with the DV (highest CO) site are applied
 3    to all people within the surrounding study area. This focus on high concentration CO monitors
 4    in the current assessment would tend to produce a greater percentage of persons exposed to
 5    higher CO levels than would the approach used in the 2000 pNEM/CO assessment.
 6          In the 2000 exposure assessment, pNEM/CO was set up to account for 15 distinct
 7    microenvironments. Each of the 12 enclosed microenvironments (including 3 motor vehicle
 8    microenvironments) was modeled using a sophisticated mass balance model. This model used
 9    probabilistic techniques to account for outdoor (ambient) air quality, air exchange rate, and
10    indoor emissions. When applied to building microenvironments characterized by relatively low
11    air exchange rates, the mass balance model in pNEM/CO yields hourly-average CO
12    concentrations in the building that tend to have less variance than the corresponding hourly-
13    average ambient concentrations outside the building. Relative to the ambient concentrations
14    outside the building, the indoor concentrations have lower peak values and the peaks are delayed
15    in time. This effect is not significant for in-vehicle microenvironments that are characterized by
16    relatively high air exchange rates.  In addition, two indoor sources of CO were evaluated in the
17    2000 pNEM/CO assessment for residential microenvironments: gas stoves and passive smoking.
18    The model was set up so that these sources could be turned on and off within the model. The
19    estimated number of people with COHb levels above 2.5 percent was noticeably higher when
20    pNEM/CO accounted for the specified  indoor sources.
21          The 2009 draft APEX/CO assessment specifies only two microenvironments (i.e., in
22    vehicle and all other). The CO concentrations in these microenvironments are modeled using a
23    simple proportionality factor (the proximity factor) applied to the corresponding ambient air
24    quality concentrations based on the fixed-site monitor values. For the in-vehicle
25    microenvironment, the proximity factor equals 1 for Scenario A and 2 for Scenario B. The
26    proximity factor for all other microenvironments equals  1 for both scenarios. And finally, no
27    provision has been made to account for the effects of indoor  sources in this draft of the current
28    CO assessment given the simplified approach used.

29         6.3.2  Comparison of Estimated COHb Levels in Adults with Coronary Heart
30               Disease using the 2000 pNEM/CO and 2009 Draft APEX/CO Assessments
31          Table 6-22 presents estimates for the percentage  of Denver adults with CHD who would
32    experience a daily maximum end-of-hour COHb level at or above the specified level under the
33    specified air quality conditions for 1995.  Table 6-23 presents similar estimates for Los Angeles.
34    Each table provides two sets of estimates for the 2000 pNEM/CO assessment (indoor sources
35    "on" and "off') and two sets for the current (2009) draft APEX/CO assessment (Scenarios A and
      October, 2009                         6-28            Draft - Do Not Cite or Quote

-------
 1   B).  See section 6.3.1 above for a brief discussion of the modeling assumptions used in
 2   developing each set of estimates.
 3          As expected, the COHb levels estimated by the 2000 pNEM/CO assessment are higher
 4   when internal sources are turned on.  As stated above, it was also expected that estimated COHb
 5   levels would be higher for Scenario B than for Scenario A in the current assessment, since
 6   Scenario B uses a larger proximity factor for the in-vehicle microenvironment.
 7          Because of the significant differences in modeling approaches employed by the two
 8   assessments, it is difficult to make a direct comparison of the results for pNEM/CO 2000 with
 9   the results for current APEX/CO draft assessment. As discussed in section 6.3.1, the two
10   assessments differ according to:
11       •  Boundaries of the study area defined for each city,
12       •  Monitors used to represent ambient CO levels,
13       •  Defined microenvironments,
14       •  Microenvironmental modeling approach used, and
15       •  Treatment of indoor sources.
16          With these caveats in mind, we observe that the estimated percentage of Denver adults
17   with CHD that experience end-of-hour COHb levels at or above 2 percent is higher for the 2009
18   draft APEX/CO assessment (Scenario A or  Scenario B for the just meeting the current 8-hour
19   standard case) than for the 2000 pNEM/CO assessment (as is case which was very close to just
20   meeting the 9 ppm CO NAAQS), regardless of whether internal sources are turned on or off in
21   the pNEM assessment (Table 6-22).  The corresponding results for Los Angeles (Table 6-23) for
22   the just meeting the standard case show values for the 2009 draft APEX/CO  assessment
23   (Scenario A or Scenario B) that are lower than the 2000 pNEM/CO assessment for the just
24   meeting the standard case when internal sources are turned on and higher when the sources are
25   turned off. Note that these patterns are not consistent across all COHb levels. For example, the
26   estimate listed for either study area for COHb levels at or above 4 percent is  higher for 2000
27   pNEM/CO (sources on) than for the corresponding values for 2009 draft APEX/CO assessment
28   (Scenarios A and B).
29          In the current assessment, assuming the ambient concentrations contributing to all
30   microenvironments are equal to the concentrations reported at the fixed-site monitor (for
31   Scenario A) indicates that there would not be any spatial heterogeneity in CO concentrations
32   across the study area, that is, the single monitor used in each study area is assumed to represent
33   all outdoor CO concentrations.  However, there are other ambient monitors within the 20 km
34   study area having lower CO concentrations  and these were used in the previous 2000 assessment.
35   Therefore, the assumption of spatial homogeneity would tend to contribute to the greater CO

     October, 2009                         6-29           Draft - Do Not Cite or Quote

-------
 1   population exposures estimated in the current assessment compared to those estimated in the
 2   previous 2000 assessment, when holding all other factors constant. Staff also assumed in the
 3   current assessment that the penetration of CO into all microenvironments was equivalent to one.
 4   This assumption would lead to a lack of attenuation of peak outdoor ambient CO concentrations
 5   that is expected to occur in indoor microenvironments when not accounting for physical
 6   processes described above. Therefore, greater population exposures would be estimated for the
 7   current assessment when compared with the exposures estimated in the prior 2000 assessment,
 8   holding all other factors constant. The impact of these simplifying assumptions is best illustrated
 9   in Figure 6-3 using the data provided in Tables 6-22 and 6-23 for the situation where no indoor
10   sources were modeled. Clearly, in the current assessment nearly all of the  simulated population
11   reaches a higher estimated daily maximum COHb level (at least once per year) for both areas
12   when compared with the previous 2000 assessment.  Once accounting for this higher population
13   distribution at COHb levels up to about 1% COHb, the general shape of the population
14   distribution is very similar for both the locations when compared with results in the previous
15   assessment, particularly when including the contribution of the in-vehicle microenvironment
16   separately (Scenario B) in estimating exposures.
     October, 2009                          6-30           Draft - Do Not Cite or Quote

-------
1
2
    Table 6-22. Percentage of Denver Adults with Coronary Heart Disease (CHD) Estimated
               to Experience a Daily Maximum End-of-hour COHb Level At or Above the
               Specified Percentage Under Specified Air Quality Conditions for 1995.
COHb
concentration
(percent)
6.0
5.0
4.0
3.0
2.5
2.0
1.5
1.0
0
Percentage of adults with coronary heart disease3 estimated
to experience a daily maximum end-of-hour COHb level
at or above the specified percentage
2000 pNEM/CO assessment for
"existing" conditions'3
Internal sources
on
0.2
0.6
1.6
5.5
10.4
19.9
37.6
83.2
100
Internal sources
off
0
0
0
<0.1
0.2
0.5
6.7
65.0
100
2009 draft APEX/CO assessment
for "just meeting" conditions0
Internal sources off
Scenario A
0
0
0
<0.1
0.4
25.6
99.7
100
100
Scenario B
0
0
<0.1
2.7
11.2
56.5
99.8
100
100
3 Characterized as "ischemic heart disease" in the 2000 pNEM/CO exposure assessment.
b "Existing" conditions: Denver CO conditions during 1995 with no adjustment. Second non-overlapping
8-hour maximum CO concentration (design value) equals 9.5 ppm. These conditions approximate "just
meeting" conditions for Denver (i.e., design value equals 9.4 ppm).
c "Just meeting" conditions: 1995 CO levels in Denver adjusted to simulate conditions when the second
non-overlapping 8-hour maximum CO concentration at the design value site equals 9.4 ppm.
4
5
    October, 2009
                                        6-31
Draft - Do Not Cite or Quote

-------
1   Table 6-23. Percentage of Los Angeles Adults with Coronary Heart Disease (CHD)a
2              Estimated to Experience a Daily Maximum End-of-hour COHb Level At or
3              Above the Specified Percentage Under "Just Meeting"Conditionsb for 1997.
COHb
concentration
(percent)
6.0
5.0
4.0
3.0
2.5
2.0
1.5
1.0
0
Percentage of adults with coronary heart disease3 estimated
to experience a daily maximum end-of-hour COHb level
at or above the specified percentage
2000 pNEM/CO assessment
Internal sources
on
0.2
0.8
2.2
5.1
9.0
16.8
32.3
79.0
100
Internal sources
off
0
0
0
<0.1
<0.1
0.5
5.2
58.1
100
2009 draft APEX/CO assessment
Internal sources off
Scenario A
0
0
0
0
<0.1
2.4
67.0
100
100
Scenario B
0
0
0
0
0.4
6.6
80.5
100
100
3 Characterized as "ischemic heart disease" in the 2000 pNEM/CO exposure assessment.
b"Just meeting" conditions: 1997 CO levels in Los Angeles adjusted to simulate conditions when the
second non-overlapping 8-hour maximum CO concentration at the design value site equals 9.4 ppm.

    October, 2009
6-32
Draft - Do Not Cite or Quote

-------
                          Denver - Just Meeting the Current Standard
•o
ti
0
0
W
0
o
^2

j_
0

ra
C
0
5
o.
0
Q.






^
o
0)
0)
(0
0)
o
2
^
o
're
c
o

JS
3
Q.
O
Q.


1UU 1
90 -
sn
ou
70 -

3" 60 -
O^

'o 50 -
^
o
-J 40 -
30 -

20 -
10 -
n -
\
: \ \\
: \\ 	 2000 pN EM/CO assessment
\ \ \
'- ^ \\ — B— Scenario A - Current Assessment
: "> \ \ —m— Scenario B - Current Assessment
\ \ \
1 *
\ \ \
\
i 1 \
1 \
: \
; '' i\
: \ \
\ \
; ^ \k

01234567
COHb level (%)
Los Angeles - Just Meeting the Current Standard
100 1
90 -

sn
ou
70 -

^ 60 -

"33 50 -
>
-1 40 -


30 -


20 -


10 -
n
- \
v\
: \ \\
: \ \\
: "^ \ \ 	 2000 pNEM/CO assessment
; \ \ l — H — Scenario A - Current Assessment
: \ \\ —m— Scenario B - Current Assessment
\\
\\


1 \
: >
\\
- \ \
:
\ \\
; ' \
: '( \\

: 'i 1
'• 	 1 	 1 	 1 	 " ^i^^^a 	 • 	 1 	 • 	 1 	 • 	 1 	 • 	 1 	
                                         COHb level
3   Figure 6-3.  Percentage of Los Angeles and Denver Adults with Coronary Heart Disease
4               (CHD) Estimated to Experience a Daily Maximum End-of-hour COHb Level
5               At or Above the Specified Percentage for Air Quality Adjusted to Just
6               Meeting the Current Standard. Data taken from Tables 6-22 and 6-23.
    October, 2009
6-33
Draft - Do Not Cite or Quote

-------
 1         6.4   VARIABILITY ANALYSIS AND UNCERTAINTY CHARACTERIZATION
 2          An important issue associated with any population exposure or risk assessment is the
 3    characterization of variability and uncertainty. Variability refers to the inherent heterogeneity in
 4    a population or variable of interest (e.g., residential air exchange rates) and cannot be reduced
 5    through further research, only better characterized with additional measurement. Uncertainty
 6    refers to the lack of knowledge regarding the values of model input variables (i.e., parameter
 1    uncertainty), the physical systems or relationships used (i.e., use of input variables to estimate
 8    exposure or risk or model uncertainty), and in specifying the scenario that is consistent with
 9    purpose of the assessment (i.e., scenario uncertainty).  Uncertainty is, ideally, reduced to the
10    maximum extent possible through improved measurement of key parameters and iterative model
11    refinement. The approaches used to assess variability and to characterize uncertainty in this
12    REA are discussed in the following two sections. Each section also contains a concise summary
13    of the identified components contributing to uncertainly and how each source may affect the
14    estimated exposures.

15         6.4.1   Analysis of Variability
16          The purpose  for addressing variability in this REA is to ensure that the estimates of
17    exposure and  risk reflect the variability of ambient CO concentrations and associated CO
18    exposure and  health  risk across the study locations and population.  In this draft assessment,
19    there are several algorithms that account for variability of input data when generating the number
20    of estimated benchmark exceedances or health risk outputs.  For example, variability may arise
21    from differences in the population residing within census tracts (e.g., age distribution) and the
22    activities that may affect CO population exposure and dose (e.g., time spent  inside vehicles,
23    moderate or greater exertion outdoors). A complete range of potential exposure levels and
24    associated risk estimates can be generated when appropriately addressing variability in exposure
25    and risk assessments; note however that the range of values obtained would be within the
26    constraints of the input parameters,  algorithms, or modeling system used, not the complete range
27    of the true exposure  or risk values.
28          Where possible, staff identified and incorporated the observed variability in input data
29    sets and estimated parameters within the exposure and dose assessment performed rather than
30    employing standard  default assumptions and/or using point estimates to describe model inputs.
31    The details regarding any variability distributions used in data inputs are  described in section 6.1.
32    To the extent possible given the data available for the assessment, staff accounted for variability
33    within the exposure  and dose modeling. APEX has been designed to account for variability in
34    some of the input data, including the physiological variables that are important inputs to
35    determining ventilation rates and COHb dose levels. As a result, APEX addresses much of the


      October, 2009                          6-34           Draft - Do Not Cite or Quote

-------
Table 6-24. Summary of How Variability Was Incorporated Into the Exposure and Dose
           Assessment.
Component
Simulated
Individuals
Ambient Input
Physiological
Factors Relevant to
Ventilation Rate and
Estimation of COHb
Levels
Variability Source
Population data
Activity patterns
Coronary heart
disease (CHD)
prevalence
Measured ambient CO
concentrations
Meteorological data
Resting metabolic rate
Metabolic equivalents
by activity (METS)
Oxygen uptake per unit
of energy expended
Weight (body mass)
Height
Blood volume
Hemoglobin content of
the blood
Pulmonary CO
diffusion rate
Endogenous CO
production rate
Comment
Individuals are randomly sampled from U.S. census
tracts used in model domains, by age (single years) and
gender.
Data diaries are randomly selected from CHAD using
six diary pools stratified by two day-types (weekday,
weekend) and three temperature ranges (< 55.0 F,
between 55.0 and 83.9°F, and >84.0 °F).
CHD prevalence is stratified by four age groups (18-44,
45-64, 65-74, and 75+) and both genders.
Temporal: 1 -hour CO for an entire year predicted using
ambient monitoring data.
Spatial: Local surface NWS stations used.
Temporal: 1-hour NWS temperature data for each year.
Three age-group (18-29, 30-59, and 60+) by gender
specific regression equations were used with body
mass as the independent variable (Johnson et al.,
2000).
Values randomly sampled from distributions developed
for specific activities (some age-specific) (US EPA,
2002).
Values randomly sampled from a uniform distribution
(Johnson et al., 2000).
Randomly selected from population-weighted lognormal
distribution with geometric mean (GM) and geometric
standard deviation (GSD) distribution specific to age
and gender derived from data from the National Health
and Nutrition Examination Survey (NHANES), for the
years 1999-2004.
Values randomly sampled from distribution based on
equations developed for each gender developed from
analyses (Johnson, 1998) of height and weight data
(Brainard and Burmaster, 1992) (see Appendix C for
details)
Values determined according to gender using equations
based on work by Allen et al (1956) (see Appendix C for
details).
Values randomly selected from distributions developed
by gender and age categories based on NHANES study
(US DHHS, 1982) (see Appendix C for details).
Values selected according to gender, height, and age
based on equations adapted from Salorinne (1976) (see
Appendix C for details).
Values randomly selected from lognormal distributions
according to equations specific to age, gender, and
menstrual phase (see Appendix C for details).
October, 2009
6-35
Draft - Do Not Cite or Quote

-------
 1    variability in exposure and dose estimates given variability in factors that affect human exposure
 2    and dose.  The variability accounted for in this analysis is summarized in Table 6-24.
 3         6.4.2  Characterization of Uncertainty
 4           While it may be possible to capture a range of exposure or risk values by accounting for
 5    variability inherent to influential factors, the true exposure or risk for any given individual is
 6    largely unknown. To characterize health risks,  exposure and risk assessors commonly use an
 7    iterative process of gathering data, developing models, and estimating  exposures and risks, given
 8    the goals of the assessment, scale of the assessment performed, and the limitations of the input
 9    data available. However, significant uncertainty often remains and emphasis is then placed on
10    characterizing the nature of that uncertainty and its impact on exposure and risk estimates.
11           The characterization of uncertainty can include either qualitative  or quantitative
12    evaluations, or perhaps a combination of both.  The approach can also  be tiered, that is, the
13    analysis can begin with a simple qualitative uncertainty characterization  then progress to a
14    complex probabilistic uncertainty analysis. This may follow when a lower tier analysis indicates
15    there is a high degree of uncertainty  for certain  identified sources, the sources of uncertainty are
16    highly influential variables in estimating the exposure and risk, and sufficient information and
17    other resources are available to conduct a  quantitative uncertainty assessment.  This is not to
18    suggest that quantitative uncertainty analyses should always be performed in all exposure and
19    risk assessments.  The decision regarding  the type of uncertainty characterization performed is
20    also informed by the intended scope and purpose of the assessment, whether the selected analysis
21    will provide additional information to the  overall decision regarding health protection, whether
22    sufficient data are available to conduct a complex quantitative analysis, and whether time and
23    resources are available for higher tier characterizations (US EPA, 2004; WHO, 2008).
24           The primary purpose of the uncertainty characterization approach selected in this draft
25    REA is to identify and compare the relative impact important sources of uncertainty  may have on
26    the estimated potential health effect endpoints.  The approach used to evaluate uncertainty was
27    adapted from guidelines outlining how to  conduct a qualitative uncertainty characterization
28    (WHO, 2008) and applied in the most recent NO2 (US EPA, 2008c) and  SO2 NAAQS reviews
29    (US EPA, 2009).  While it may be considered ideal to follow a tiered approach in the REA to
30    quantitatively characterize all identified uncertainties, staff selected the mainly qualitative
31    approach given the extremely limited data available to inform probabilistic analyses.
32           The qualitative approach used in this REA varies from that of WHO (2008) in that a
33    greater focus of the characterization  performed  was placed on evaluating the direction and the
      October, 2009                             6-36                Draft - Do Not Cite or Quote

-------
 1    magnitude5 of the uncertainty; that is, qualitatively rating how the source of uncertainty, in the
 2    presence of alternative information, may affect the estimated exposures and health risk results.
 3    In addition and consistent with the WHO (2008) guidance, staff discuss the uncertainty in the
 4    knowledge-base (e.g., the accuracy of the data used, acknowledgement of data gaps) and
 5    decisions made where possible (e.g., selection of particular model forms), though qualitative
 6    ratings were assigned only to uncertainty regarding the knowledge-base.
 7           First, staff identified the key aspects of the assessment approach that may contribute to
 8    uncertainty in the exposure and risk estimates and provide the rationale for their inclusion. Then,
 9    staff characterized the magnitude and direction of the influence on the assessment results for
10    each of these identified sources of uncertainty.  Consistent with the WHO (2008) guidance, staff
11    subjectively scaled the overall impact of the uncertainty by considering the  degree of severity of
12    the uncertainty as implied by the relationship between the source of the uncertainty and the
13    output of the air quality characterization.  Where the magnitude of uncertainty was rated low, it
14    was judged that large changes within the source of uncertainty would have only a small effect on
15    the exposure results.  A designation of medium implies that a change within the source of
16    uncertainty would likely have a moderate (or proportional) effect on the results. A
17    characterization of high implies that a small change in the source would have a large effect on
18    results.  Staff also included the direction of influence, indicating how the source of uncertainty
19    was judged to affect estimated exposures or risk estimates; either the estimated values were
20    likely over- or under-estimated. In the instance where the component of uncertainty can affect
21    the assessment endpoint in either direction, the influence was judged as both. Staff characterized
22    the direction of influence as unknown when there was no evidence available to judge the
23    directional nature of uncertainty associated with the particular source. Staff also subjectively
24    scaled the  knowledge-base uncertainty associated with each identified source using a three level
25    scale:  low  indicated significant confidence in the data used and its applicability to the assessment
26    endpoints, medium implied that there were some limitations regarding consistency and
27    completeness of the data used or scientific evidence presented, and high indicated the
28    knowledge-base was  extremely limited.
29           The output of the uncertainty characterization was a summary describing, for each
30    identified source of uncertainty, the magnitude of the impact and the direction of influence the
31    uncertainty may have on the exposure and risk characterization results. There are several
32    sources of uncertainty associated with this simplified approach for modeling CO population
33    exposure/dose and associated potential health risk, each summarized and discussed in Table 6-
34    25.
             5 This is synonymous with the "level of uncertainty" discussed in WHO (2008), section 5.1.2.2.
      October, 2009                              6-37               Draft - Do Not Cite or Quote

-------
1

2
Table 6-25. Characterization of Key Uncertainties in the Draft Assessment for Denver and Los Angeles Areas.
          Sources of Uncertainty
         Category
                     Element
                                     Influence of Uncertainty
                                      on Exposure/Dose or
                                          Risk Estimates
                                       Direction
             Magnitude
            Knowledge-
                Base
            Uncertainty
                                                                                      Comments3
                        Database Quality
                                          Over
                  Low
                 Low
           INF: There may be a limited number of poor quality high concentration data within the analytical data sets, potentially
           influencing the number of benchmark dose level exceedances.
           KB: EPA's Air Quality System data used in the analyses are of high quality. There is no other source of monitoring
           data as comprehensive. Data are being used in a manner consistent with one of the defined objectives of ambient
           monitoring.
       Ambient CO
       Concentrations
                 Spatial and
                 Temporal
                 Representation
                                                Over
                Medium
                 High
           INF: Use of a single fixed-site ambient CO monitor likely does not adequately represent spatial temporal variability in
           ambient CO levels throughout each study area. Given that typical in-vehicle to ambient concentration ratios range from
           2 to 4 and on-road sources tend to dominate CO emissions, it is likely that the spatial variability in ambient
           concentrations across a region would be less than this value.  Given that the single monitor selected for use in the
           exposure assessment had generally greater concentrations than other monitors within the broader metropolitan area, it
           is likely that exposures are overestimated for most simulated individuals.
           KB: In the absence of 1) a monitoring network designed to measure spatial variability in CO concentrations, 2)
           performing air quality modeling to estimate spatial and temporal variability in CO concentrations and, 3) analysis of any
           existing and representative monitoring data that can potentially indicate spatial concentration gradients, staff judge the
           uncertainty in the knowledge-base as high.
                        Missing Data
                        Substitution
                                          Under
                  Low
                 Low
           INF: Assuming there is an equal probability of missing low and high concentration hourly values, and that substituted
           data are limited by the bounds of the algorithm (i.e., as defined by limits in the measurement data), there may be a few
           missing high concentration data that could lead to underestimation in exposure concentrations and doses. This
           assumes that the substitution of low-level concentration data with potentially higher concentrations (within the bounds
           of the algorithm) does not affect exposure results.
           KB: All available measurement data are quality assured. Very few data values were substituted with respect to the
           number of measured values available in each location.
       Adjustment of Air
       Quality to
       Simulate Just
       Meeting the
                 Historical Data Used
Unknown
Medium
High
INF & KB: Even though the historical data represent a real air quality condition that may be similar to concentrations
levels expected to just meet the 8-hour current standard, the condition simulated is hypothetical. It is largely unknown
how influential factors such as emission levels per vehicle, vehicular traffic, and meteorology compare between an
earlier period of time and the hypothetical condition of just meeting the current standard.
                                  October, 2009
                                                                                     6-38
                                                                      Draft - Do Not Cite or Quote

-------
Sources of Uncertainty
Category
Current 8-Hour
Standard
APEX Inputs and
Algorithms
Element
Proportional
Approach Used
Population Database
Activity Pattern
Database
Longitudinal Profile
Algorithm
Meteorological Data
Influence of Uncertainty
on Exposure/Dose or
Risk Estimates
Direction
Both
Both
Unknown
Both
Both
Magnitude
Low
Low
Low-Medium
Low -Medium
Low
Knowledge-
Base
Uncertainty
Low
Low
Medium
Medium
Low
Comments3
INF: The magnitude of the adjustment applied to historical ambient concentration data was minimal for Denver (i.e.,
0.99 was the adjustment applied), though greater for Los Angeles (0.63). In comparing recent and historical ambient
CO concentrations (Table 6-5), a linear relationship exists between the range of concentrations reported for the two
time periods in both locations. More importantly, a strong proportional relationship is present when comparing the
recent and historic CO concentrations measured at the Los Angeles monitor.
KB: A similar proportional approach was judged adequate in simulating air quality conditions just meeting the 8-hour
CO NAAQS in prior assessments (US EPA, 1992; Johnson et al., 2000). An analysis of the CO concentration
distributions comparing 1995 and 1997 CO air quality data in the Denver and Los Angeles study areas, respectively
with more recent CO air quality data (i.e., 2006) in these same two areas shows a roughly linear or proportional change
throughout the distribution.
INF & KB: Population data are from the US Census Bureau, a reliable and quality assured source. Data used are
specifically for census tracts modeled in Denver and Los Angeles. Staff assumed any remaining uncertainties in the
database would have negligible influence on exposure and dose results.
INF: Data are actual records of the time spent in specific locations while performing specific activities in particular
locations. While not specific to a particular area, the activity patterns of a population are generally well represented by
the mainly population-based and nationally-representative survey data (e.g., see Table G-1 in Appendix G regarding
the patterns of typical commuting in CHAD versus the urban locations modeled in this assessment).
KB: Data are from a reliable and quality assured source (CHAD) and are from surveys of real persons. Features of an
individual's activity pattern are well represented, adjustments are made to represent the population distribution in a
specific area (using age and gender), and temperature is used to link CHAD diaries with the simulated individuals
residing in a specific area. However, there are several assumptions made that contribute to uncertainty in its use. For
example, activity patterns of persons surveyed over 30 years ago are assumed to represent a current persons activity
patterns.
INF: The magnitude of potential influence would be mostly directed toward estimates of multi-day exposures, not the
number or percent of persons having at least one exposure or dose above a selected level.
KB: In developing the longitudinal method, the evaluation indicated that both the Dand A statistics are reasonably
reproduced for the population. In addition, the approach was compared to two other independent methods used for
constructing longitudinal activity patterns (see Appendix B, Attachment 5 of US EPA , 2009). Note however, long-term
diary profiles (i.e., monthly, annual) do not exist for a population.
INF & KB: Data are from the National Weather Service, a well-known and quality-assured source. Daily maximum
temperatures are only used when selecting appropriate diaries to simulate individuals.
October, 2009
6-39
Draft - Do Not Cite or Quote

-------
Sources of Uncertainty
Category

Potential Health
Effect
Benchmark
Levels
Element
Algorithm and Input
Data for In-Vehicle
CO Concentrations
Algorithm and Input
Data for All Other
Microenvironmental
CO Concentrations
Commuting
Algorithm Not Used
CHD Prevalence
Susceptible
Population
Influence of Uncertainty
on Exposure/Dose or
Risk Estimates
Direction
Both
Over
Over
Both
Unknown
Magnitude
Medium
Low-Medium
Medium
Low
Low
Knowledge-
Base
Uncertainty
Medium
Medium
Low
Low
Medium
Comments3
INF: Given that on-road and in-vehicle CO concentrations are typically higher than ambient CO concentrations,
Scenario A likely underestimates in-vehicle exposures. There is variability between in-vehicle and ambient CO
concentrations that is not accounted for by using a single value to represent the relationship (i.e., a factor of two) such
as traffic density, local meteorology, driving conditions, and differences in vehicle age, technology, design and time of
operation.
KB: While most studies reviewed indicate that, on average, there may be a factor of two difference between the
ambient and in-vehicle CO concentrations, there are a limited number of studies that measured these concentrations
and even fewer were located in the U.S. It is largely unknown how this and other identified influential factors might
influence the true relationship between in-vehicle and ambient CO concentrations.
INF & KB: Even though CO is considered relatively inert, it is likely that the ambient contribution to exposures in all
indoor microenvironments are overestimated. This is a result of not considering air exchange that would delay and limit
infiltration of outdoor CO concentrations, thereby reducing indoor microenvironmental peak concentrations due to CO
of ambient origin. It is also possible that the residential indoor or outdoor microenvironment concentrations of
simulated individuals residing in close proximity to major roads may be underestimated. This is because the current
ambient monitors are unlikely to reflect the higher CO concentrations expected to occur near all major roads (draft ISA,
section 3.6.6.2). The simplified approach used an ambient monitor that may be representative of outdoor near-road
CO concentrations (i.e., microscale and middle scale monitors) experienced by only a portion of the population in each
study area. However, given the larger portion of time spent in locations other than in vehicles and outdoors near-roads
and the limited difference in exposures estimated when comparing scenario A to scenario B, it is likely that exposures
and doses have an overall tendency to be overestimated for most of the simulated population.
INF & KB: In using the ambient monitor that has the greatest CO concentrations compared to other monitoring data in
an area, it is assumed that these concentrations would represent conservative estimates of air quality in the area. This
would lead to overestimates in exposures, particularly when not considering commuting to other lower ambient
concentration locations.
INF & KB: Data are from the Centers for Disease Control, a well-known and quality-assured source. Though
prevalence data are not specific for each region, the national prevalence data were stratified by selected age-groups
and gender.
INF & KB: Data from a well-conducted multi-center controlled human exposure study demonstrate cardiovascular
effects in subjects with moderate to severe coronary artery disease at COHb levels as low as 2.0-2.4%. No laboratory
study has evaluated the effect of exposure to CO resulting in COHb levels below 2.0%. There is no established no
adverse effect level and, thus there is greater uncertainty about the lowest benchmark level used (i.e., 1 .5%) and
uncertainty about whether individuals with the most severe CHD are adequately represented. Given that the evidence
supporting the choice of benchmark levels is based on controlled human exposure data, we judge the influence of this
uncertainty on the risk characterization as being low.
aINF refers to comments associated with the influence rating; KB refers to comments associated with the knowledge-base rating.
October, 2009
6-40
Draft - Do Not Cite or Quote

-------
            Based on the qualitative judgments made by staff for a range of sources of uncertainty
      and their characterization as to direction and magnitude of influence on exposures and doses, the
      exposure and dose estimates are possibly overestimated for a larger portion of the population the
      assessment is intended to represent (i.e., those residing/travelling in high CO concentration
      microenvironments).  This is because:
 1       •  Of the four sources of uncertainty associated with potential overestimation (i.e.,
 2          spatial/temporal representation of monitoring data, data base quality, and absence of
 3          movement among different air quality districts during commutes, algorithm and input
 4          data for all other microenvironmental CO concentrations), three were estimated as having
 5          medium magnitude of influence, while the remaining source (i.e., ambient monitoring
 6          database quality) was ranked as having a low or a negligible magnitude of influence.

 7       •  The one source of uncertainty associated with potential underestimation (i.e., missing
 8          data substitution) was judged to have a low magnitude of influence on estimated
 9          exposures and doses.

10       •  Of the remaining identified sources of uncertainty judged by staff to have either
11          bidirectional influence (six sources) or unknown (three sources) direction, five sources
12          were judged to have a low magnitude of influence on estimated exposures and doses.
            While there was a wide-ranging level of uncertainty in the knowledge-base for the
      identified sources, there is relatively less uncertainty in  staff judgments regarding the sources
      associated with potential overestimation of CO exposure and resulting COHb levels.
13       •  A high degree of uncertainty in the knowledge-base was  assigned to two sources: the
14          spatial/temporal representation of monitoring data (direction of influence characterized as
15          over, with a medium rated magnitude) and the use of historical data in representing air
16          quality that just meet the current standard (direction of influence characterized as
17          unknown, with magnitude rated as medium).

18       •  The knowledge-base uncertainty was low for three of the five sources identified above as
19          being associated with either under- or overestimating exposures (the rating for the
20          remaining two sources was medium and high).

21       •  The knowledge-base uncertainty for sources with unknown or bidirectional influence  was
22          low (five sources), medium (four sources), and high (one source).
23
      October, 2009                          6-41               Draft - Do Not Cite or Quote

-------
 1         6.5  KEY OBSERVATIONS
 2          Presented below are key observations resulting from the exposure and dose assessment
 3    for ambient CO.

 4       •  An important limitation in the assessment for this review is the lack of detailed spatial
 5          representation of the current ambient monitoring data, which creates challenges for
 6          estimating the spatial variability of CO concentrations across a study area. This
 7          limitation contributed in part to the reasoning for the development of the simplified
 8          approach used in this assessment.
 9              o  In this simplified approach, staff used a single monitor recording the highest CO
10                 concentrations to represent the ambient air quality in each area. This was done to
11                 accommodate the potentially greater CO exposures expected to occur to persons
12                 residing in areas with higher CO concentrations (i.e., those occurring on or
13                 immediately near major roadways).
14              o  Using a single monitor in each study area, however, still posed difficulties in
15                 characterizing the full range of microenvironmental CO concentrations, such as
16                 CO levels in vehicles or near major roadways, concentrations outside residences,
17                 as well as those occurring within indoor microenvironments for the simulated
18                 population. This is a result of the limited information across a broad geographic
19                 area regarding the relationships between specific ambient monitor concentrations
20                 and microenvironmental concentrations.

21       •  One-hour and 8-hour average daily maximum exposures and the daily maximum end-of-
22          hour COHb blood levels were estimated using a simplified exposure modeling approach
23          involving two scenarios in two study areas: urban areas of Denver and Los Angeles
24          counties.  In Scenario A, CO concentrations in all microenvironments were set equal to
25          the ambient monitor concentrations and in Scenario B, the CO concentration for the in-
26          vehicle microenvironment were increased over those of the ambient monitor and all other
27          microenvironments were set equal to the ambient monitor concentrations.  The two air
28          quality conditions investigated by staff included as is air quality, and air quality for
29          higher CO levels, adjusted to simulate just meeting the current 8-hour CO NAAQS
30          (section 6.2).
31              o  Fewer than 1% of the study population in each study area (< 0.2%) were
32                 estimated to experience a daily maximum end-of-hour COHB level at  or above
33                 2.0% under as is air quality conditions in either  scenario.
34              o  Results for the two study areas differed appreciably for air quality adjusted to just
35                 meet the current standard.  For these conditions, the estimates of percent of
36                 population experiencing a daily maximum end-of-hour COHB level at or above
37                 potential health benchmarks were substantially greater for the Denver  study area
38                 (e.g., differing by a factor of 8 or more for the 2%  COHb benchmark).

39       •  Results generated in the current assessment for the  air quality conditions just meeting the
40          current NAAQS were compared with  estimates from the assessment conducted in 2000
41          (Johnson et al., 2000) for similar conditions in the Denver and Los Angeles study areas
      October, 2009                          6-42              Draft - Do Not Cite or Quote

-------
 1          (section 6.3). While focused on similar air quality conditions, the two assessments
 2          employed different versions of the exposure model (APEX vs pNEM) and there were
 3          significant differences in the approach used in each assessment. For example, as
 4          compared to the current assessment, the 2000 assessment employed more monitors to
 5          represent ambient CO levels, differentially treated a much greater number of
 6          microenvironments, and encompassed larger study areas.
 7              o  The estimated percent of persons with daily maximum end-of-hour COHb blood
 8                 levels when using air quality adjusted to just meet the current standard in both
 9                 Denver and Los Angeles was substantially greater in the current assessment when
10                 compared to that estimated in the 2000 assessment (e.g., a difference of a factor of
11                 10 or more at the 2% COHb benchmark).

12       •  Based on an overall qualitative judgment of the identified sources of uncertainty in the
13          assessment approach, selections made regarding input data, and algorithms used, and
14          their characterization as to direction and magnitude of influence, the exposure and dose
15          estimates for much of the simulated population represented by either scenario in this
16          assessment are likely overestimated (section 6.4, Table 6-25). There may be a smaller
17          fraction of the simulated population (e.g., those residing in close proximity to major
18          roads, persons regularly commuting for extended periods of time) where some periods of
19          exposure are underestimated due to the simplified assumptions made in estimating in-
20          vehicle and near-road CO exposures, although likely less so and for a yet smaller portion
21          of the population in scenario B.  The impact of such potentially higher exposure periods
22          on the population COHb levels will vary depending on the overall pattern of exposures.

23       •  Given the considerations described above regarding the characterization of uncertainty
24          and the tendency of the assessment approach to overestimate exposure and dose, staff
25          finds the utility of this assessment for the purpose of considering the adequacy of the
26          current standards to be limited.
      October, 2009                          6-43              Draft - Do Not Cite or Quote

-------
  1          6.6    REFERENCES

 2    CDC. (2009).  Summary Health Statistics for U.S. Adults: National Health Interview Survey, 2007. Series 10,
 3            Number 240. U.S. Department of Health and Human Services, Hyattsville, MD, May 2009.

 4    Isaacs K., McCurdy, T ErricksonA, Forbes S, GlenG, Graham S, McCurdyL,  Nysewander, M, Smith L,
 5            Tulve N, and Vallero D. (2009). Statistical properties of longitudinal time-activity data for use in EPA
 6            exposure models.  Poster presented at the American Time Use Research Conference; College Park MD,
 7            June 26, 2009.

 8    Johnson T, Mihlan G, LaPointe J, Fletcher K, Capel J. (1999 Draft). Estimation of Carbon Monoxide Exposures
 9            and Associated Carboxyhemoglobin Levels in Denver Residents Using pNEM/CO (Version 2.0). Report
10            prepared by ICF Consulting and TRJ Environmental, Inc., under EPA Contract No. 68-D6-0064. U.S.
11            Environmental Protection Agency, Research Triangle Park, North Carolina. March 1999.

12    Johnson T, Mihlan G, LaPointe J, Fletcher K, Capel J. (2000).  Estimation of Carbon Monoxide Exposures and
13            Associated Carboxyhemoglobin Levels for Residents of Denver and Los Angeles Using pNEM/CO
14            (Version 2.1). Report prepared by ICF Consulting and TRJ Environmental, Inc., under EPA Contract No.
15            68-D6-0064. U.S. Environmental Protection Agency, Research Triangle Park, North Carolina.  Available
16            at: http://www.epa.gov/ttn/fera/human_related.html.. June 2000.

17    S AIC. 2001. Technical Peer Review of "Estimation of Carbon Monoxide Exposures and Associated
18            Carboxyhemoglobin Levels for Residents of Denver and Los Angeles Using pNEM/CO (version 2.1)"
19            Prepared by Science Applications International Corporation under EPA Contract No. 68-D-98-113.
20            Available at: http://www.epa.gov/ttn/fera/human related.html..

21    US Census Bureau. (2009). American Fact Finder.  Census Summary File 3 (SF3) - custom tables. Available at:
22            www.factfinder.census.gov.

23    US EPA (1992). Review of the National Ambient Air Quality Standards for Carbon Monoxide:  Assessment of
24            Scientific and Technical Information. Office of Air Quality Planning and Standards Staff Paper, report no.
25            EPA/452/R-92-004.

26    US EPA. (2004). An Examination of EPA Risk Assessment Principles and Practices. Staff Paper prepared by the
27            US EPA Risk Assessment Task Force.  EPA/ 100/B-04/001. Available at:
28            http://www.epa.gov/OSA/pdfs/ratf-final.pdf.

29    US EPA. (2008a). Total Risk Integrated Methodology (TRIM) Air Pollutants Exposure Model Documentation
30            (TRIM.Expo/APEX, Version 4.3). Volume 1: Users Guide. Report No. EPA-452/B-08-00la. Office of
31            Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC.

32    US EPA. (2008b). Total Risk Integrated Methodology (TRIM) Air Pollutants Exposure Model Documentation
33            (TRIM.Expo/APEX, Version 4.3). Volume 2: Technical Support DocumenLReport No. EPA-452/B-08-
34            00Ib. Office of Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research
35            Triangle Park, NC.

36    US EPA. (2008c). Risk and Exposure Assessment to Support the Review of the NO2 Primary National Ambient
37            Air Quality Standard. EPA-452/R-08-008a. November 2008. Available at:
38            http://www.epa.gov/ttn/naaqs/standards/nox/data/2008112 l_NO2_REA_final.pdf.

39    US EPA. (2009). Risk and Exposure Assessment to Support the Review of the SO2 Primary National Ambient Air
40            Quality Standard.  EPA-452/R-09-007. August 2009. Available
41            athttp://www.epa.gov/ttn/naaqs/standards/so2/data/200908SO2REAFinalReport.pdf.
      October, 2009                             6-44                Draft - Do Not Cite or Quote

-------
1    WHO. (2008). Harmonization Project Document No. 6. Part 1: Guidance document on characterizing and
2            communicating uncertainty in exposure assessment. Available at:
3            http://www.who.int/ipcs/metliods/harmonization/areas/exposure/en/.
     October, 2009                              6-45                 Draft - Do Not Cite or Quote

-------
                             Appendix A




          Conceptual Model and Simplified Data Flow of APEX.
October 2009                       A-1       Draft - Do Not Cite or Quote

-------
      Air Exchange Rates &
        Building Volumes
       Emission Rates and
         Use Patterns for
         Indoor Sources
       (e.g., gas appliances,
        passive smoking)
       Ambient Fixed-Site
         Concentrations
A
k
Air Quality Specification
Seasonal Considerations
     (Temperature)
  Mass balance Model
  or Factors Approach

for Indoor and in Vehicle
   Microenvironments
       Outdoor
   Microenvironment
    Concentrations
Human Activity and
 Exertion Patterns
  Population &
Commuting Data
              Exposure Algorithms
               Distribution of People and
               Occurrences of Exposures
              Linked with Breathing Rate
                                                          Dose Algorithm
                         Distribution of People
                          and Occurrences of
                          Dose (COHb) Levels
Appendix A.  Conceptual Model and Simplified Data Flow of APEX.
                        September 2009
                                     A-2
                         Draft - Do Not Cite or Quote

-------
                                        Appendix B

                              Mass Balance Model in APEX

       The mass balance method models an enclosed microenvironment as a well-mixed volume
in which the air concentration is spatially uniform at any specific time.  The concentration of an
air pollutant in such a microenvironment is estimated using the following four processes:
       .  Inflow of air into the microenvironment;
       .  Outflow of air from the microenvironment;
       .  Removal of a pollutant from the microenvironment due to deposition, filtration,
          and/or chemical degradation; and
       .  Emissions from sources of a pollutant inside the microenvironment.

       Table B-l lists the parameters required by the mass balance method to calculate
concentrations in a  microenvironment.  The proximity factor (fproximity) is used to account for
differences in ambient concentrations between the geographic location represented by the
ambient air quality  data (e.g., a regional fixed-site monitor) and the geographic location of the
microenvironment (e.g., near a roadway).  This factor could take a value either greater than  or
less than 1. Emission source (ES) represents the emission rate for the emission source, and
concentration source (CS) is the mean air concentration resulting from the source. Removal is
defined as the removal rate of a pollutant from a microenvironment due to deposition, filtration,
and chemical reaction. The air exchange rate (Rair exchange) is expressed in air changes per hour.
       October 2009                          B-l            Draft - Do Not Cite or Quote

-------
•  Table B-l. Parameters of the Mass Balance Model
Variable
J proximity
cs
ES
r>
•**• removal
r>
•**• a/r exchange
V
Definition
Proximity factor
Concentration source
Emission source
Removal rate due to
deposition, filtration, and
chemical reaction
Air exchange rate
Volume of
microenvironment
Units
N/A
ppm
ug/hr
1/hr
I/to-
rn3
Value Range
J proximity *-*
CS>0
ES>0
/? > 0
•**• removal — "
/? > 0
f^-air exchange — "
F>0
       The mass balance equation for a pollutant in a microenvironment is described by the
differential equation
                dt
                                                                        (B-l)
       where:

       dCME(t)       =     Change in concentration in a microenvironment at time t (ppm),

                    =     Rate of change in microenvironmental concentration due to influx
                          of air (ppm/hour),

                    =     Rate of change in microenvironmental concentration due to outflux
                          of air (ppm/hour),

                    =     Rate of change in microenvironmental concentration due to
                          removal processes (ppm/hour), and

                    =     Rate of change in microenvironmental concentration due to an
                          emission source inside the microenvironment (ppm/hour).

Within the time period of an hour each of the rates of change, AC!n, AC0!d, ACremova/, and
      e, is assumed to be constant.
         removal
       October 2009
                                            B-2
Draft - Do Not Cite or Quote

-------
       The change in microenvironmental concentration due to influx of air is represented by the
following equation:


                 in      j       ambient  J proximity  J penetration    air exchange           \   )

       where:

       Cambient       =     Ambient hourly outdoor concentration (ppm)
      /proximity       =     Proximity factor
      /penetration      =     Penetration factor
       Rair exchange    =     Air exchange rate (I/hour)

       The change in microenvironmental concentration due to outflux of air is described by:


              AC   =dCaa®=R       xC   (r)                            (B-3)
                 out     ^      cur exchange   ME

       The change in concentration due to deposition, filtration, and chemical degradation in a
microenvironment is simulated by the first-order equation:

   ,,        d\^removal(t)   .                             ,      . ,                . ,
   - removal =      ~7     = '^deposition + ^ filtration + ^chemical) ^ ME\ V = ^remova X ^ME\ V   (B-4)

       where:

                    =     Removal rate of a pollutant from a microenvironment due to
                           deposition (I/hour)
                    =     Removal rate of a pollutant from a microenvironment due to
                           filtration (I/hour)
                    =     Removal rate of a pollutant from a microenvironment due to
                           chemical degradation (I/hour)
                    =     Removal rate of a pollutant from a microenvironment due to
                           overall removal (I/hour)
       As discussed in Section 6.2, EPA decided not to model indoor emissions of CO in the
current exposure assessment; consequently, the optional term ACTOMrce was uniformly set equal to
0.0 for this study.

       Combining Equation B-l with Equations B-2, B-3, and B-4 yields

              *> =ACfa -Rairexchange xC^CO-*,™* xC^CO                 (B-5)
       October 2009                           B-3            Draft - Do Not Cite or Quote

-------
       The solution to this differential equation is


              CMS (0 =	~	h (CME (0)	~~~) exP(~^co»*medO                   (B-6)
                        combined             combined

       where:

       CME(O)        =      Concentration of a pollutant in a microenvironment at the
                            beginning of a hour (ppm)
       Cusft)        =      Concentration of a pollutant in a microenvironment at time t within
                            the time period of a hour (ppm)

       -^combined              t\air exchange  ' -^removal
       Based on Equation B-6, the following three hourly concentrations in a microenvironment
are calculated:

                            ^   ,.           (-
                                         n
                                          combined
  f hourly end 	/~< equil  . s/~<   ff\\  (~* equil
                                      _ ^ eqml ^     /_ n      \
                                        ^ME  ) CXP ^ ^combined)
                                              quil\^   CAp ^ ^combined)
                    J*
                    0

       where:
                                          n
                                           combined
end
           ,
                           Equilibrium concentration in a microenvironment (ppm)
                           Concentration in a microenvironment at the beginning of an hour
                           (ppm)
                           Concentration in a microenvironment at the end of an hour (ppm)
           rlymea"      =     Hourly mean concentration in a microenvironment (ppm)
       At each hour time step of the simulation period, APEX uses Equations B-7, B-8, and B-9
to calculate the hourly equilibrium, hourly ending, and hourly mean concentrations.  APEX
reports hourly mean concentration as hourly concentration for a specific hour. The calculation
continues to the next hour by using c^rlyend for the previous hour as CME(O).
       October 2009                           B-4            Draft - Do Not Cite or Quote

-------
                                    Appendix C

                           COHb Module for APEX4.3

       This appendix describes the probabilistic carboxyhemoglobin (COHb) module used in the
current APEX4.3 model. The approach described here is based primarily on the COHb module
originally described by Biller and Richmond in two reports (Johnson et al., 1992; Johnson et al.,
2000) and used in EPA probabilistic NAAQS exposure model for CO (pNEM/CO), a
predecessor of APEX4.3. This appendix also describes the principal changes made to the COHb
module when it was incorporated into APEX4.3, including a change in the method used to solve
the Coburn-Forster-Kane (CFK) equation (Coburn et al., 1965).

C.I    The Base Physiological Model for Computing COHb Levels

       Using time/activity data obtained from various diary studies, APEX constructs a
composite diary for each simulated person in the specified population at risk.  The composite
diary consists of a sequence of events spanning the specified period of the exposure assessment
(typically one calendar year). Each event is defined by a start time, a duration, a geographic
location, a microenvironment, and an activity. Using various algorithms described in Section 5
of this report, APEX4.3 provides estimates of CO concentration and alveolar ventilation rate for
each event in the composite diary. APEX4.3 then uses these data, together with estimates of
various physiological parameters specific to the simulated individual, to estimate the percent
COHb in the blood (%COHb) as an average %COHb value over the duration of each  exposure
event  and as an instantaneous %COHb level at the end of each event.

       The %COHb calculation is based on the solution to the non-linear CFK equation,
previously described in Appendix E of Johnson  et al. (2000). The CFK model describes the rate
of change of COHb blood levels as a function of the following quantities:

       1.  Inspired CO pressure
       2.  COHb level
       3.  Oxyhemoglobin (O2Hb) level
       4.  Hemoglobin (Hb) content of blood
       5.  Blood volume
       6.  Alveolar ventilation rate
       7.  Endogenous CO production rate
       8.  Mean pulmonary capillary oxygen pressure
       9.  Pulmonary diffusion rate of CO
       10. Haldane coefficient (M)
       11. Barometric pressure
       12. Vapor pressure of water at body temperature (i.e., 47 torr).

       If all  of the listed quantities except COHb level are constant over some time interval, the
CFK equation has a linear form over the interval and is readily integrated.  The  solution to the
linear form gives reasonably accurate results for lower levels of COHb. However, CO and
oxygen compete for the available hemoglobin and are, therefore,  not independent of each other.
          October 2009                   C-1                   Draft - Do Not Cite or Quote

-------
If this dependency is taken into account, the resulting differential equation is no longer linear.
Peterson and Stewart (1975) proposed a heuristic approach to account for this dependency which
assumed the linear form and then adjusted the ChHb level iteratively based on the assumption of
a linear relationship between COHb and O2Hb. This approach was used in the COHb module of
the original CO-NEM exposure model (Biller and Richmond, 1982, Johnson and Paul, 1983).

       Alternatively, it is possible to determine COHb at any time by numerical integration of
the nonlinear CFK equation (e.g., by use of the Runge-Kutta method) if one assumes a particular
relationship between COHb and O2Hb. Muller and Barton (1987) demonstrated that assuming a
linear relationship between COHb and O2Hb leads to a form of the CFK equation equivalent to
the Michaelis-Menton kinetic model which can be analytically integrated.  However, the
analytical solution in this case cannot be solved explicitly for COHb.  Muller and Barton (1987)
demonstrated a binary search method for determining the COHb value.

       The COHb module used in pNEM/CO employed a linear relationship between COHb and
O2Hb which was consistent with the basic assumptions of the CFK model. The approach
differed from the linear forms used by other modelers in that the Muller and Barton (1987)
solution was employed.  However, instead of the  simple binary search described by Muller and
Barton (1987), a combination of the binary search and Newton-Raphson root-finding methods
was used to solve for COHb (Press et al., 1986).  Using the Muller and Barton (1987) solution
increased computation time compared to the Peterson and Stewart (1975) method but was shown
to be faster than fourth-order Runge-Kutta numerical integration.

       APEX4.3 employs a different approach in which the CFK equation is solved using a
fourth-order Taylor's series expansion with subintervals.  This method, first incorporated in
Version 3 of APEX,  is described in Section C.2 of this  appendix.  A more  detailed description
can be found in the Programmer's Guide for the APEX3 model (Glen, 2002).
C.2    The CFK Model for Estimation of Carboxyhemoglobin

       Table C-l defines the variables which appear in the equations of this section. Coburn,
Forster, and Kane (1965) derived the following differential equation governing COHb levels in
the blood upon exposure to CO.
              d[COHb} = Vco  }  PIco   Pc2
                 dt      Vb    BVb   MBVb[O2Hb}                       l q'  " '
where,

                   1     P  -P
              B = —— +  B . H2°                                        (Eq. C-2)
                  n        TT~                                            \ ~i     /
                 DLco      VA
          October 2009                   C-2                   Draft - Do Not Cite or Quote

-------
Table C-l.  Definitions of CFK Model Variables.
Variable
t
[COHb]

[02Hb]
[RHb]

[COHb]0
[THb]0
%[COHb]
%[02Hb]
%[COHb]0
%[COHB]«
Definition
Time from start of an exposure event
Concentration of carboxyhemoglobin (COHb) in blood at
time t
Concentration of oxyhemoglobin (O2Hb) in blood at time t
Concentration of reduced hemoglobin in blood

[COHb] at t = 0
[RHb] + [COHb] + [02Hb]
[COHb] expressed as percent of [RHb]0
[O2Hb] expressed as percent of [RHb]0
[COHb] at t  = 0
[COHb] at t  = ~
Pressure of inspired CO in air saturated with water vapor
at body temperature
Mean pulmonary capillary CO pressure

Mean pulmonary capillary O2 pressure

Barometric pressure
Vapor pressure of water at body temperature, or 47

Alveolar ventilation  rate

Endogenous CO production rate

Pulmonary CO diffusion rate

Haldane coefficient
Equilibrium constant for reaction O2 + RHb = O2Hb
Blood volume
Total hemoglobin in blood
Methemoglobin as weight percent of Hb
 PB
 PHO

 VA

 Vco
 DL
   CO
M
k
Vb
Hb
%MetHb
Notes:
1 Standard Temperature Pressure, and Dry (STPD)
Units
minutes
ml CO per ml blood at STPD

ml O2 per ml blood at STPD
equivalent ml CO per ml of
blood at STPD
ml CO per ml blood at STPD
torr

torr
torr

torr
torr

ml/min STPD
ml/min STPD
ml/min/torr, STPD
ml
g/100ml
          October 2009
                                                    Draft - Do Not Cite or Quote

-------
       If the only quantity in equation (C-l) that can vary with time is [COHb], the CFK
equation is linear and can be readily integrated.  However, since oxygen (02) and CO compete
for the available Hb, [COHb] and [O2Hb] must be related. Increasing [COHb] will result in
decreasing [O2Hb].  Thus the CFK equation is not linear and requires the relationship between
the two quantities to be known if it is to be accurately integrated over a wide range of COHb
levels.

       Various linear relationships between [COHb] and [O2Hb] have been used (see Marcus,
1980; McCartney, 1990; Muller and Barton,  1987; and Tikuisis et al.,  1987). A relationship not
previously used follows directly from the basic assumptions of the CFK model.  The CFK model
employs the Haldane coefficient, which is the equilibrium constant associated with the following
reaction representing the replacement of O2 in O2Hb by CO:

             CO + O2Hb ^O2 + COHb                                    (Eq. C-3)

       The following equation, the Haldane relationship, applies approximately at equilibrium
conditions.

             Pc0 [COHb}
             ^^	 = M                                           (Eq. C-4)
              Pcco[02Hb]

       The Haldane coefficient, M, is the chemical equilibrium constant for reaction (C-3). The
above reaction can also be viewed as the difference between two competing chemical reactions:

             CO + RHb ^ COHb                                          (Eq. C-5)

             O2 + RHb ^ O2Hb                                           (Eq. C-6)


       Subtracting (C-6) from (C-5) yields (C-3).  If (C-3) is in equilibrium, then (C-5) and (C-
6) are in equilibrium. If A; represents the equilibrium constant for (C-6) then:


                        -k                                              (Eq.C-7)

       It is known that an individual breathing air free of CO for an extended period will have
about 97% of their reactive Hb bound with oxygen (O2Hb) and the remainder (3%) as the
reduced form (RHb). It is also known that at one atmosphere barometric pressure, the mean
pulmonary capillary oxygen pressure is approximately 100 torr.  Substituting into (C-7) yields
0.32 as the approximate value of & at body temperature. From mass balance considerations:

             [O2Hb] + [COHb] + [RHb] = [THb]0                             (Eq. C-8)

       Eliminating [RHb] between (C-7) and (C-8) and solving for [O2Hb] yields:


          October 2009                   C-4                  Draft - Do Not Cite or Quote

-------
                        kPc
             [02Hb] =      °2   ([THb]0 - [COHb}}                         (Eq. C-9)
                      \ + kPc
       This equation represents the aforementioned linear form of the CFK equation.  It has the
same form as a relationship given by McCartney (1990), but replaces the constant in the
McCartney equation by the term in (C-9) involving the mean pulmonary capillary oxygen
pressure and the equilibrium constant k.  Substituting (C-9) into (C-l) yields a CFK equation free
of [C^Hb] and fully consistent with Coburn, Forster, and Kane's original derivation.
              d[COHb] = Vco  | Plco       [COHb]       l + kPct
                 dt    ~ Vb + BVh   [THb]0-[COHb]X kMBV,
(Eq. C-10)
       In working with the CFK model it is convenient to express COHb as a percent of [RHb]o.
Multiplying (C-10) by 100 and dividing by [RHb]o yields the expression

       d%[COHb]_   100    Vco | PIc       %[COHb]      100(1 + ^C02)    E  cn)
           dt       [THb]0   Vh   BVh   WO-%[COHb]  k[RHb]0MBVb


       Equation (C-l 1) can be written in the form suggested by Muller and Barton (1987):
           dt              WO-%[COHb]

where,
             C0 =-— P^ +    )                                    (Eq. C-13)
                         yV    BV}
                = — -    _                                             c
                  k[THb}0MBVb


       Given values for the atmospheric pressure and the physiological variables in equations
(C-l 2) through (C-l 4), the value of %[COHb] at time t can be found by numerical integration
using such techniques as the fourth-order Runge-Kutta method (Press et al., 1986). Muller and
Barton (1987) demonstrated that an equation of the form of (C-12) is equivalent to a Michaelis-
Menton kinetics model which can be integrated. The integration yields:

                                                                     =0   (Eq.C-15)
                                                 %[COHb]x-%[COHb]

       The equation for %[COHb]oo is obtained by setting equation (C-12) equal to zero and
solving for %[COHb], which is now equal to %[COHb]oo:
          October 2009                   C-5                  Draft - Do Not Cite or Quote

-------
               %[COHb]m =  10°C°                                      (Eq. C-16)
       Equation (C-15) cannot be solved explicitly for %[COHb]. Muller and Barton (1987)
suggest the binary search method as one way to find the value of %[COHb].  Press et al. (1986)
contend a combination of the binary search and Newton-Raphson methods is faster on average.
Consequently, the pNEM/CO version of the COHb module used a combination of the binary
search and Newton-Raphson root finding methods to solve for COHb (Press et al., 1986). Using
the Muller and Barton (1987) solution increased the computation time when compared with the
Peterson and Stewart (1975) method, however it was still shown to be faster than the fourth-
order Runge-Kutta numerical integration.

       The current version of APEX (APEX4.3) employs an alternative approach in which the
CFK equation is solved using a fourth-order Taylor's series expansion with subintervals.  This
method, first incorporated in Version 3 of APEX, is described in detail in the Programmer's
Guide for the APEX3 Model by Glen (2002).  This reference also includes the results of various
tests conducted on 10 candidate methods for solving the CFK equation. The  selected method
(fourth-order Taylor series with subintervals) was chosen because of its simplicity, fast execution
speed, and ability to produce relatively accurate estimates of %COHb at both low and high levels
of CO exposure. Additional information concerning the %COHb calculation method and its
theoretical basis can be found in Section 10.2 of US EPA (2008).

       In developing the fourth-order Taylor Series expansion approach, Glen (2002) began by
defining N(t) as the %COFIb level in the blood at time t, a quantity that is mathematically
restricted to range between 0 and 100 (percent). N(t) satisfies the following differential equation:

             ]ST(t) = Co - Ci N(t) / (100 - N(t) )                              (Eq. C-17)

where Co and Ci are constants (at least over the duration  of one event) that depend on physical
and physiological parameters and on the CO concentration in the air. Equation (C-17) is
equivalent to (C-12) above, except that (C-12) uses the symbol %[COHb] instead of N(t).

       The task of expanding N(t) in a Taylor's series becomes simpler if the following new
variables are defined:

             D0=  1-N(0)/100                                            (Eq. C-18)

             A0=  C0/(C0 + Ci)                                          (Eq. C-19)

             Ai =  Ci / (Co + Ci)                                          (Eq. C-20)

             D =  Do-Ai                                                (Eq. C-21)

             z  =  (Co + Ci) t / (100*D0*D0)                                (Eq. C-22)

       The z variable is a re-scaled time variable that is dimensionless. It is used as the

          October 2009                    C-6                   Draft - Do Not Cite or Quote

-------
independent variable for the Taylor's series expansion.  In equations expressed as functions of z
rather than t, any primes will indicate the derivatives with respect to z.

       Expressing (C-17) as a function of z yields the expression

             1ST (z) = Do 2 Ao - Do 2 AI N(z) / ( 100 - N(z) )                   (Eq. C-23)

The Taylor's series about the origin (z = 0) for N(z) is given by

       N(z) = N(0) + 1ST(0) z + 1ST'(0) z2 / 2 + 1ST'(0) z3 / 6 + Niv(0) z4/ 24 + ...   (Eq. C-24)

       Through a series of algebraic substitutions, Glen (2002) shows that the Taylor series
expansion of N(z) truncated to the fourth order can be represented by
              T4(z) = T3(z) - 100 AI Do D (Ai2 - 8 D AI + 6 D 2) z4 / 24       (Eq. C-25)
where
     T3(z) = N(0) +100 Do D z -100 AI D0 D z2 / 2 + 100 AI D0 D (Ai - 2D) z3 / 6    (Eq. C-26)

       Tests showed that the fourth-order Taylor series expansion (C-25) provided greater
accuracy than the third-order expansion for z values close to one.  Glen (2002) found that z
values below one generally correspond to N(0) values below forty to fifty percent for one-hour
exposure events.

       The z value for a given event depends on the event duration, the initial %COHb level
N(0), and on the physiological parameters, and can be directly evaluated at the start of each
event. For events with a z value above some threshold,  it is possible to improve the performance
of (C-25) by dividing the event into smaller events ("subintervals"), each with a shorter duration
and hence smaller z value.  As the subinterval duration decreases, accuracy increases at the
expense of program execution time. APEX4.3 enables the  user to select a limit on z which in
turn determines the number of subintervals to be used in applying the fourth-order Taylor
expansion.  Glen (2002) recommends that the limit on z be  set at 0.4 or 0.5.

C.3    Application of the COHb Model  in APEX4.3

Description of APEX4.3 for CO

       APEX4.3 follows the daily activities over an extended period of a finite set of simulated
individuals residing within a given geographic area. The period may be a single season or a
calendar year.  Each simulated individual is defined by a set of general demographic
characteristics that includes age, gender, and body weight.  The values of these factors are used
to derive values for blood volume, menstrual phase, endogenous CO production rate, and other
factors required by the COHb module (see Section C.4). The exposure of each individual is
represented by a continuous sequence of exposure events which span the time period of interest.
Each exposure event represents a time interval of 60 minutes or less during which the individual
          October 2009                   C-7                    Draft - Do Not Cite or Quote

-------
resides in a single environment and engages in a single activity. To permit calculation of hourly
average exposures, exposure events are not permitted to fall in more than one clock hour.
Consequently, the passage from one exposure event to the next is indicated by a change in
microenvironment, activity, or clock hour. Algorithms within APEX4.3 calculate an average CO
concentration for each exposure event according to the time, district, and microenvironment
specified for the event.  As the exposure events for a simulated individual are contiguous, the
model can combine these concentrations to output distributions of one-hour and running eight-
hour exposures for each individual. The exposures calculated for the simulated individuals can
then be population-weighted to produce exposure distributions for population groups of
particular interest (e.g., people with coronary heart disease).

       APEX4.3 constructs a year-long time/activity pattern for each simulated individual by
sampling 24-hour activity patterns from the Consolidated Human Activity Data Base (CHAD),
which is described in Section 5.3.3 of this report. The sampling approach attempts to match the
24-hour activity patterns to the simulated individual and exposure period according to the
demographic characteristics of the individual and the season, day type (weekday/weekend), and
maximum temperature of each day in the specified exposure period.

The COHb Module

       The COHb module in APEX4.3 currently employs the version of the CFK model
represented by equations (C-12) through (C-14) to compute an average COHb value over the
duration of each exposure event and an instantaneous COHb level at the end of each event. To
perform these computations, the COHb module requires information on each of the quantities
listed in the section describing the CFK model. In addition, the COHb level at the beginning of
the exposure event must be known. This latter quantity is usually the COHb level computed at
the end of the previous contiguous exposure event. To obtain the initial COHb at the start of the
exposure period, the computation is started one day before the beginning of the period.  The
effect of the initial COHb value on the end value is negligible after about 15 hours. The program
stores the calculated COHb values for each exposure event and outputs distributions of COHb
levels by population group for averaging times ranging from one hour to one day.

Assignment of CFK Model Input Data for an Exposure Event

       Section C.4 describes the equations and procedures used by the APEX4.3 COHb module
to obtain the values of the input variables for equations (C-2) and (C-13) through (C-16). A brief
overview is given here.

       The actual inspired CO level can change significantly during an exposure event.  The
model supplies an average exposure concentration for the event, which is used as the CO input.
The time constant for the change in COHb is sufficiently large that the use of concentrations
based on averaging times up to one hour can be used in place of the instantaneous concentrations
over the averaging time period with little loss of accuracy in estimating the COHb level at the
end of the exposure event.  Furthermore, applying the average concentrations to a contiguous
sequence of exposure events does not cause an accumulation of error.
          October 2009                    C-8                   Draft - Do Not Cite or Quote

-------
       The COHb model presently used in APEX4.3 does not account for changing barometric
pressure. It uses a constant barometric pressure which is a function of the average elevation of
an area above sea level.  The pressure at sea level is taken to be 760 torr.

       The remaining input variables to the CFK model are all physiological parameters. While
the Haldane coefficient, the  equilibrium constant k, and average pulmonary capillary oxygen
pressure are treated as having the same constant values for all individuals, the remaining
physiological input variables will vary among individuals.  The next section describes the
methods used to generate the various physiological input variables for each combination of
individual and calendar day  processed by APEX.

C.4    Computation of Input Data for the COHb Module

       As discussed in the previous section and in Sections 5.3.5 and 5.3.7 of the main body of
this report, the algorithms used to estimate VE and COHb require values for various
physiological parameters such as body mass, blood volume, and pulmonary diffusion rate. Table
C-2 provides a list and description of the principal parameters; additional parameters are listed
and described in Chapter 5 of US EPA (2008). An algorithm within APEX4.3 probabilistically
generates a value for each parameter on the list (collectively referred to as a physiological
profile) for each simulated individual. Figure C-l is a flow diagram showing the process by
which each physiological profile is generated. Each of the generated physiological profiles is
internally consistent, in that the functional relationships among the various parameters are
maintained. For example, blood volume is determined as a function of weight and height, where
height is estimated as a function of weight. Weight in turn is selected from a distribution specific
to gender and age.

       For each simulated individual, APEX4.3 computes exposure for a contiguous sequence of
exposure events spanning the total time period of the computation. This multi-day sequence of
exposure events is determined by random sampling day-long event sequences from a set of pools
of 24-hour activity patterns.  An individual 24-hour pattern in one of these pools is referred to
here as a unit exposure sequence (UES). Each pool consists of a collection of UESs that are
specific to selected demographic characteristics of the individual (e.g., age and gender), season,
day type (weekday/weekend),  and maximum daily temperature.

       A UES is a contiguous set of exposure events spanning 24 hours. Each event is
characterized by start time, duration in minutes, home/work status, microenvironment, and
activity. All exposure events are constrained to occur entirely within a clock hour.

       The CFK model within the COHb module is called for each exposure event. For each
event it requires the following data.

       Time duration of event, min
       Inspired CO partial pressure averaged over the event, torr
       Percent COHb at the start of the event
       Alveolar ventilation  rate, ml/min STPD
       Average pulmonary  capillary oxygen pressure, torr
          October 2009                   C-9                   Draft - Do Not Cite or Quote

-------
Haldane Coefficient
Equilibrium constant for the reaction of Q^
Atmospheric pressure, torr
Blood volume, ml
Total potential reduced hemoglobin content of blood, ml CO/ml STPD
Pulmonary CO diffusion rate, ml/min/torr STPD
Endogenous CO production rate, ml/min STPD
    October 2009                   C-10                  Draft - Do Not Cite or Quote

-------
Table C-2. Principal Parameters Included in the Physiological Profile for Adults for Applications of APEX4.3.
    Parameter
Algorithm(s)
 Containing
 Parameter
 Other Parameters
   Required for
    Calculating
    Parameter
                 Method Used to Estimate Parameter Value
 Age
COHb
Ventilation
rate
Demographic group
Randomly selected from population-weighted distribution specific to demographic
group
 Gender
COHb
Ventilation
rate
Demographic group
Randomly selected from population-weighted distribution specific to demographic
group
 Body Weight
COHb
Ventilation
rate
Gender
Age
Randomly selected from population-weighted lognormal distribution with geometric
mean (GM) and geometric standard deviation (GSD) distribution specific to age and
gender derived from data from the  National Health and Nutrition Examination
Survey (NHANES), for the years 1999-2004 (Isaacs and Smith, 2005)
 Height
COHb
Weight
Gender
Estimated using equations developed by Johnson (1998) using height and weight
data provided by Brainard and Burmaster (1992).

  Males:    height = 34.43 inches + (6.67)[ln(weight)] + (2.38 inches)(z)
  Females:  height = 48.07 inches + (3.07)[ln(weight)] + (2.48 inches)(z)

The z term is randomly selected from a unit normal [N(0,1)] distribution. Units:
height (inches), weight (Ibs).
 Menstrual phase
COHb
Gender
Age
If gender = female, menstrual phase was randomly assigned in alternating 14-day
cycles according to the following age-specific probabilities.

  Age < 12 or >50: 100% premenstrual
  Age 12 through 50:  50% premenstrual, 50% postmenstrual.
                            October 2009
                                          C-ll
                                                    Draft - Do Not Cite or Quote

-------
   Parameter
Algorithm(s)
 Containing
 Parameter
 Other Parameters
   Required for
    Calculating
    Parameter
                Method Used to Estimate Parameter Value
Blood volume
COHb
Gender
Weight
Height
Blood volume (Vb) was determined according to gender by the following equations
which are based on work by Allen et al. (1956) which was modified to accept the
units used for height and weight.

  Males:    Vb = (20.4)(weight) + (0.00683)(H3) - 30
  Females: Vb = (14.6)(weight) + (0.00678)(H3) - 30

Units: blood volume (ml), weight (Ibs), height (inches).
Hemoglobin
content of the
blood, Hb
COHb
Gender
Age
Randomly selected from normal distribution with arithmetic mean (AM) and
arithmetic standard deviation (ASD) determined by gender and age based
obtained from data from the National Health and Nutrition Examination Survey
(NHANES), for the years 1999-2004. (Isaacs and Smith, 2005)Units: grams of Hb
per deciliter of blood
                           October 2009
                                          C-12
                                                   Draft - Do Not Cite or Quote

-------
   Parameter
Algorithm(s)
 Containing
 Parameter
 Other Parameters
   Required for
    Calculating
    Parameter
                 Method Used to Estimate Parameter Value
Pulmonary CO
diffusion rate,

DL
COHb
Gender
Height
Age
Pulmonary CO diffusion rate (DL) was determined according to gender, height, and
age according to the following equations obtained from a paper by Salorinne (1976)
and modified to conform to the units used in the COHb module.

   Males:
                                                          = (0.361)(height) - (0.232)(age) + 16.3 ml/min/torr


                                                      Females:

                                                    DLCQ = (0.556)(height) - (0.115)(age) - 5.97 ml/min/torr


                                                    Units:

                                                    DL  (ml/min/torr), height (inches), age (years).


                                                    Given the alveolar ventilation rate for the exposure event the associated adjusted
                                                    pulmonary diffusion rate is calculated as:

                                                    DL  (Adjusted) = DL (Base) + 0.000845F, - 5.7
                           October 2009
                                          C-13
                                                    Draft - Do Not Cite or Quote

-------
Parameter
Endogenous CO
production rate
Resting
metabolic rate
(RMR)
Energy
conversion factor
(ECF)
NV02max
VO2max
Algorithm(s)
Containing
Parameter
COHb
Ventilation
rate
Ventilation
rate
Ventilation
rate
Ventilation
rate
Other Parameters
Required for
Calculating
Parameter
Gender
Age
Menstrual phase
Gender
Age
Body Weight
Gender
Gender
Age
NV02max
Body Weight
Method Used to Estimate Parameter Value
Endogenous CO production rate was randomly selected from a lognormal
distribution with geometric mean (GM) and geometric standard deviation (GSD)
determined according to the following equations specific to age, gender, and
menstrual phase.
Males, 18+: GM = 0.473, GSD = 1.316
Females, 18+, premenstrual: GM = 0.497, GSD = 1.459
Females, 1 8+, postmenstrual: GM = 0.31 1 , GSD = 1 .459
Units: GM (ml/hr), GSD (dimensionless).
See Section 5.3.5 of this report and Chapter 5 of US EPA (2008).
See Section 5.3.5 of this report and Chapter 5 of US EPA (2008).
See Section 5.3.5 of this report and Chapter 5 of US EPA (2008).
See Section 5.3.5 of this report and Chapter 5 of US EPA (2008).
October 2009
C-14
Draft - Do Not Cite or Quote

-------
               Gender
                                           Demographic Profile of
                                            Simulated Individual
                                                 Weight
                                                 Height
                                              Menstrual Phase
                                              Endogenous CO
                                              Production Rate
                                               Blood Volume
                                             Total Hemoglobin
                                                 Content
                                               of the Blood
                                               Pulmonary CO
                                               Diffusion Rate
                                           Resting Metabolic Rate
                                                 (RMR)
                                          Energy Conversion Factor
                                                  (ECF)
                                            Maximum Normalized
                                            Oxygen Uptake Rate
                                                (NV02mJ
                                           Maximum Oxygen Uptake
                                               Rate(V02mJ
                                   Age
Figure C-l.  Flow Diagram for Physiological Profile Generator.  Input data is supplied at the start
of the APEX4.3 computation.
            October 2009
C-15
Draft - Do Not Cite or Quote

-------
       Given these data as inputs, the module computes the percent COHb at the end of the
exposure event. This value is used by the module as the initial percent COHb for the next
contiguous exposure event. The module also computes the average percent COHb value for each
exposure event. The main program retains these values and uses them to calculate percent
COHb values for averaging times ranging from one hour to one day.

       Some of the above data do not change during an APEX4.3 computer run and, therefore,
need to be supplied to the computer program only once at the start. Some of the data vary with
the individual and therefore need to be supplied at the beginning of each activity day.  Other data
tend to change with the exposure event and therefore need to be supplied for each new exposure
event.

Barometric Pressure

       A constant barometric pressure is assumed for the study area based on the average height
above sea level:

       PB = 760 x exp(-0.0000386 x Altitude)                                (Eq. C-27)


where altitude is the average height (in feet) of the study area above sea level (US EPA, 1978).
The altitude was set at 5,183 feet for Denver and 328 feet for Los Angeles.

Average Pulmonary Capillary Oxygen Pressure

       The equation employed is based on an approximation used by Peterson and Stewart
(1975) in which 49 torr is subtracted from the partial pressure of inspired oxygen.  This leads to
the following approximate relationship:

              Pc02 = 0.209(PB - 47) - 49                                   (Eq. C-28)


where 0.209  is the mole fraction of O2 in dry air and 47 is the vapor pressure of water at body
temperature.  This expression was used in an investigation of the CFK equation by Tikuisis et al.
(1987). Often times a value 100 torr is commonly used as Equation (C-28) generates this value
for a barometric pressure equivalent to 760 torr.

Haldane Coefficient

       The value of 218 was used for the Haldane coefficient. While measured values in the
range 210 to 270 have been reported in the extent literature, most researchers use values within
the range of 210 to 240. In the early 1980's, the Clean Air Scientific Advisory Committee
(CASAC) expressed the opinion to EPA (Friedlander,  1982) that the most careful work done in
this area was that by Rodkey (1969), who determined a value of 218. This value was selected for
use in the COHb module of the earlier CO-NEM exposure model.  Other researchers using
values in the range 218 to 220 include Peterson and Stewart, 1970; Marcus, 1980; Collier and
          October 2009                   C-16                 Draft - Do Not Cite or Quote

-------
Goldsmith, 1983; and Muller and Barton, 1987.  As the value 218 falls within the range currently
used by researchers, EPA analysts have elected to continue using this value in APEX4.3.

Equilibrium Constant for the Reaction of O? and RHb

       This quantity was estimated in Section C.2 to have the value 0.32 based on the
observation that %[RHb] is about 3% in individuals breathing air which is free of CO and a
value of 100 torr for Pco .

Total Reduced Hemoglobin in the Absence of O? and CO

       The quantity [THb]o is expressed as equivalent milliliters of O2 or CO at STPD per
milliliter of blood.  Total Hb blood levels are customarily expressed as grams per deciliter of
blood.  The total Hb level in the absence of COHb and O2Hb would consist principally of RHb
which can react with Q^ or CO and MetHb which cannot.  Total Hb blood levels also tend to be
higher in people living at higher altitudes. To relate [THb]o to Hb, it is therefore necessary to
correct for the MetHb present, adjust for the effect of altitude, and convert to equivalent
milliliters of CO at STPD.  The later conversion  is based on the observation that a gram of
reduced Hb can react with a maximum of 1 .39 ml of Q^ or CO at STPD. The application of
these three factors yields the equation:
       [RHb]0 = 1 .39 x #6(100 - %MetHb) x 1 +       j        (£q. C-29)


where HbAlt is the percent increase in Hb due to exposure to altitude and is given by (EPA
1978):
       HbAlt = 2.76e°-ooon49Altitude

Hb in equation (C-29) is a sea level value.  Hb level in a human population is normally
distributed with the mean Hb and standard deviation both dependent on gender and age class (see
entry in Table C-2 for the distributions of Hb by age and gender). Given the hemoglobin content
of the blood based on the distributions listed in Table C-2, [THb]o is calculated using equation
(29). The weight percent MetHB, %MetHB, is taken to be 0.5% of the weight of Hb (Muller and
Barton, 1987).
Determination of Weight

       Body mass or weight (in kg) was determined by fitting lognormal distributions to data
organized by age and gender from the National Health and Nutrition Examination Survey for the
years 1999-2004 (Isaacs and Smith, 2005). Distribution parameters were estimated for single-
year age cohorts for both genders for ages 0-85. As the NHANES 1999-2004 studies only
covered persons up to age 85, linear forecasts for the parameters were made for ages 86-100, as
based on the data for ages 60 and greater.
          October 2009                   C-17                   Draft - Do Not Cite or Quote

-------
Determination of Height

       The following equations were used to estimate height as a function of gender and weight.
Equations C-30 and C-31 were derived by Johnson (1998) using height and weight data provided
by Brainard and Burmaster (1992).

  males:   height = 34.43 inches + (6.67)[ln(weight)] + (2.38 inches)(z)          (Eq. C-30)

  females:  height = 48.07 inches + (3.07)[ln(weight)] + (2.48 inches)(z)         (Eq. C-31)

where the z term was randomly selected from a unit normal [N(0,l)] distribution.

Base Pulmonary Diffusion Rate of CO

       A base lung diffusivity of CO for the individual is calculated as follows:

              Men:  DLm = 0.361 x height - 0.232 x age + 16.3                 (Eq. C-32)


              Women: DLm = 0.556 x height - 0.115 x age - 5.97               (Eq. C-33)


where height is in inches and age is in years.

       The regression equations were obtained from a paper by Salorinne (1976) and modified
to conform to the units used in the COHb module.  The Salorinne data were obtained for non-
exercising individuals.  Tikuisis et al. (1992), working with eleven male subjects at various
exercise levels, showed significant increase in lung diffusivity of CO with increasing alveolar
ventilation rate.  Regression analyses of data provided by Tikuisis for the individual subjects in
the study showed the relationship to be linear.  From this relationship and the heights and ages of
the subjects in the Tikuisis et al. study, it was determined that the Salorinne equations for male
subjects correspond to an alveolar ventilation rate of 6.69 1/min STPD.  In the absence of other
data it is assumed that this same value applies to women. Thus, for each twenty-four hour period
equations C-32 and C-33 are used to compute lung diffusion rates of CO for a base case alveolar
ventilation rate of 6.69 1/min STPD. As will be seen, this value is adjusted to account for the
actual ventilation rate experienced by the simulated individual during each individual exposure
event.

Endogenous Rate of CO Production

       The endogenous CO production rates taken from  a number of sources show the rate to be
distributed lognormally in the population (see Table C-3  for data and sources). The distribution
is different for men and women. For a woman there is a  further difference depending on whether
she is in her premenstrual or postmenstrual phase.  Table C-2 presents these distributions
classified by class, gender, and menstrual phase.


          October 2009                   C-18                   Draft - Do Not Cite or Quote

-------
       For each male individual, APEX4.3 specifies a single value for endogenous CO
production rate and uses it for all days of the year. For each female individual between 18 and
64 years of age, APEX4.3 specifies one value of endogenous CO production rate to represent
premenstrual days and one value to represent postmenstrual days.  Female individuals under 12
years and older than 50 are assumed to be premenstrual; consequently, APEX4.3 specifies a
single value for endogenous CO production rate to be used for all days of the year. The specified
values are randomly selected from the appropriate distributions presented in Table C-2. A
random number, z, is sampled from the standardized normal distribution, N(0,l) to make each
selection.  The appropriate endogenous CO production rate is then obtained from:

             Vco = 0.01667 x (geom.mean) x (geom.S.D.y                   (Eq. C-34)

The constant term converts ml/hr to ml/min.

       A probabilistic algorithm within APEX4.3 assigns a menstrual phase to each day of the
year for female individuals aged 12 to 50 years.  The algorithm randomly assigns a number
between 1 and 28 to January 1. The number is increased by one for each successive day until
number 28 is reached. The next day is numbered 1 and the 28-day numbering cycle is repeated
until each day of the year has been assigned a number between 1 and 28. Days numbered 1
through 14 are identified as post-menstrual days; days numbered 15 through 28 are identified as
pre-menstrual days.

INPUT DATA SUPPLIED WITH EACH EXPOSURE EVENT

Duration of Exposure Event

       The duration of the exposure event in minutes is supplied by the main program to the
COHb module.

Partial Pressure of Inspired Carbon Monoxide

       The main program supplies the inspired CO concentration averaged over the duration of
the exposure expressed as ppm.  This quantity is converted to pressure via:

             Plco = (CO) x (Pb - 47) x 10-6                                 (Eq. C-35)


Initial Percent COHb Level at Start of Exposure Event

       The program retains the percent COHb computed at the end of the previous exposure
event and uses this value as the initial percent COHb for the present event. The starting COHb at
the beginning of an activity day is  the final COHb level at the end of the preceding activity day.
This latter procedure is used for the first activity day of the overall computation since the
program starts the day before the overall period covered by the APEX4.3 computation.
          October 2009                   C-19                  Draft - Do Not Cite or Quote

-------
Alveolar Ventilation Rate
       The main program supplies the COHb module with ventilation rate derived from the
algorithm discussed in Section 5.3.5 of this report.

Adjusted Pulmonary Diffusion Rate of CO

       Given the alveolar ventilation rate for the exposure event the associated adjusted
pulmonary diffusion rate can be calculated from:
       DLm (Adjusted) = DLm (Base} + 0.000845F,, - 5.7
                                (Eq. C-36)
Table C-3.    Literature Data Used to Derive Geometric Mean and Standard Deviation
             Lognormal Distribution of Endogenous CO Production Rate.
Study Author
Brouillard et al.
(1975)
Burke et al.
(1974)
Coburn et al.
(1963)
Delivoria-
Papadopoules
etal. (1974)
Luomanmaki
and Coburn
(1969)
Lynch and
Moede(1972)
Merke et al.
(1975)
Werner and
Lindahl(1980)
Values for Endogenous CO Production Rate
0.81
0.37
0.43
0.35
0.45
0.45
0.57
0.23
0.38
0.4
0.72
0.48
0.64
0.4
0.54
0.57
0.49
0.58
0.4

0.26
0.54
0.51
0.42
0.81
0.37
0.23
0.86
0.47
0.76
0.33
0.45
0.52
0.39

0.6
0.72
0.34
0.41
0.26
0.23
0.25
0.35
0.23
0.48
0.7
0.5
0.59
0.43

0.45
0.99
0.41
0.54
0.65
0.33
0.2
0.52
0.24
0.31
0.58
0.33
0.8
0.35

0.39
0.48
0.26
0.38
0.51
0.42
0.22
0.8
0.55
0.7
0.38
0.45
0.72
0.51

0.4
0.53
0.16

0.62
0.44
0.15
0.54
0.32
0.36
0.51
0.36
0.54
0.42


0.43
0.3

0.44
0.29
0.21
0.68
0.43
0.65
0.55


0.57






0.48

0.28
0.35

          October 2009
C-20
Draft - Do Not Cite or Quote

-------
C.5    References

Biller WF and Richmond HM.  1982.  Sensitivity Analysis on Coburn Model Predictions of
       COHb Levels Associated with Alternative CO Standards. Report to Strategies and Air
       Standards Division of the Office of Air quality Planning and Standards, U.S.
       Environmental Protection Agency, Research Triangle Park, NC, November, 1982.

Brouillard RP, Contrad ME, Bensinger TA.  1975.  Effect of blood in the gut on measurements
       of endogenous carbon monoxide production. Blood. 45:67-69.

Brainard J and Burmaster D.  1992. Bivariate distributions for height and weight of men and
       women in the United  States. Risk Analysis.  12(2):267-275.

Burke PD, Rodkey FL, Blaschke TF, Collison HA, Waggoner JG.  1974. Comparison of plasma
       bilirubin turnover and carbon monoxide production in man. J Lab ClinMed.  83:29-37.

Coburn RF, Forster RE, Kane RB.  1965.  Considerations of the physiology and variables that
       determine blood carboxyhemoglobin concentration in man.  J din Invest.  44:1899-1910.

Coburn RF, Blakemore WS, Forster RE. 1963. Endogenous carbon monoxide production in
       man. JClin Invest. 42:1172-1178.

Collier C and Goldsmith JR.  1983. Interactions of carbon monoxide and hemoglobin at high
       altitude.  Atmos Environ. 17:723-728.

Delivoria-Papadoppulos M, Coburn RF, Forster RE.  1974.  Cyclic variation of rate of carbon
       monoxide production in normal women.  J ApplPhysiol.  36:49-71.

Friedlander SK.  1982. Letter from the Clean Air Scientific Advisory Committee to US EPA
       Administrator. August 31,  1982.

Glen G. 2002. Programmer's Guide for the APEX3 Model  Prepared by ManTech
       Environmental Technology, Inc., for the U.S. Environmental Protection Agency,
       Research Triangle Park, NC.  April 30.

Johnson T and Paul RA. 1983. The NAAQS Model (NEM) Applied to Carbon Monoxide. EPA
       450/5-83-003, U.S. Environmental Protection Agency, Research Triangle Park, NC.

Johnson T.  1998. Memo No. 5: Equations for Converting Weight to Height Proposed for the
       1998 Version of pNEM/CO. Memorandum Submitted to U.S. Environmental Protection
       Agency.  TRJ Environmental, Inc., 713 Shadylawn Road, Chapel Hill, North Carolina
       27514.

Johnson T, Capel J, Olaguer E, Wijnberg L.  1992. Estimation of Carbon Monoxide Exposures
       and Associated Carboxyhemoglobin Levels in Denver Residents Using a Probabilistic
       Version of NEM. Prepared by TRJ Environmental, Inc., for the Office of Air Quality
          October 2009                  C-21                  Draft - Do Not Cite or Quote

-------
      Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park,
      NC.

Johnson T, Mihlan G, LaPointe J, et al. 2000. Estimation of Carbon Monoxide Exposures and
      Associated Carboxyhemoglobin Levels in Residents of Denver and Los Angeles Using
      pNEM/CO (Version 2.1).  Prepared by TRJ Environmental, Inc., for the Office of Air
      Quality Planning and Standards, U.S. Environmental Protection Agency, Research
      Triangle Park, NC.

Luomanmaki K and Coburn RF.  1969. Effects of metabolism and distribution of carbon
      monoxide on blood and body stores. Am JPhysiol. 217(2):354-362.

Lynch SR and Moede AL. 1972.  Variation in the rate of endogenous carbon monoxide
      production in normal human beings. J Lab ClinMed. 79:85-95.

Marcus AH.  1980. Mathematical models for carboxyhemoglobin. Atmos Environ.  14:841-844.

McCartney ML. 1990. Sensitivity analysis applied to Coburn-Forster-Kane models of
      carboxhemoglobin rormation. Am IndHyg Assoc J'.  51(3): 169-177.

Merke C, Cavallin-Stahl E, Lundh B.  1975.  Carbon monoxide production and reticulocyte
      count in normal women: effect of contraceptive drugs and smoking. Ada Med Scan.
      198:155-160.

Muller KE. and Barton CN.  1987. A nonlinear version of the  Coburn, Forster and Kane model
      of blood carboxyhemoglobin. Atmos Environ.  21:1963-1967.

Peterson JE and Stewart RD. 1970. Absorption and elimination of carbon monoxide by inactive
      young men. Arch Environ Health.  21:165-171.

Peterson JE and Stewart RD. 1975. Predicting the carboxhemoglobin levels resulting from
      carbon monoxide exposures. JApplPhysiol. 39:633-638.

Press WH, Flannery BP, Teukolsky SA, Vettering WT. 1986.  Numerical Recipes: The Art of
      Scientific Computing. 1st  edition. Cambridge University Press, Cambridge MA.

Rodkey FL, O'Neal JD, Collison HA.  1969. Oxygen and carbon monoxide equilibria of human
      adult hemoglobin at atmospheric and elevated pressure. Blood. 33:57-65.

Salorinne Y. 1976. Single-breath pulmonary diffusing capacity.  ScandJResp Diseases.
      Supplementum 96.

Tikuisis P, Madill HD,  Gill BJ, Lewis WF, Cox KM, Kane DM.  1987. A critical analysis of the
      use of the CFK equation in predicting COHb formation. Am Ind Hyg Assoc J.
      48(3):208-213.

Tikuisis P, Kane DM, McLellan TM, Buick F, Fairburn SM. 1992.  Rate of formation of

          October 2009                  C-22                  Draft - Do Not Cite or Quote

-------
       carboxyhemoglobin in exercising humans exposed to carbon monoxide. J ApplPhysiol.
       72(4):1311-1319.

US EPA. 1978. Altitude as a Factor in Air Pollution. EPA-600/9-78-015. U.S. Environmental
       Protection Agency, Research Triangle Park, NC.

US EPA. 2008. Total Risk Integrated Methodology (TRIM) Air Pollutants Exposure Model
       Documentation (TRIM.Expo/APEX, Version 4.3).  Volume 2: Technical Support
       Document._Report No. EPA-452/B-08-001b. Office of Air Quality Planning and
       Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC.

Werner B and Lindahl J.  1980. Endogenous carbon monoxide production after bicycle exercise
       in healthy subjects and in patients with hereditary spherocytosis.  Scan J Lab Invest.
       40:319-324.
          October 2009                  C-23                  Draft - Do Not Cite or Quote

-------
                            Appendix D




                         Apex Output Files
October, 2009                     D-1           Draft - Do Not Cite or Quote

-------
   Appendix D. Apex Output Files.
 Output File Type
                            Description
Log
The Log file contains the record of the APEX model simulation as it
progresses. If the simulation completes successfully, the log file
indicates the input files and parameter settings used for the simulation
and reports on a number of different factors. If the simulation ends
prematurely, the log file contains error messages describing the critical
errors that caused the simulation to end.
Profile Summary
The Profile Summary file provides a summary of each profile modeled in
the simulation. Each line lists the person's age, gender and race, in
addition to a number of other personal profile variables that the model
uses to simulate exposure.
Sites
The Sites file lists the sectors, air districts, and zones in the study area,
and identifies the mapping between them.
Hourly
The Hourly file provides an hour-by-hour time series of exposures,
doses, and other variables for each modeled profile.
Daily
The Daily file provides a day-by-day time series of exposures, doses, and
other variables for each modeled profile.
Events
The Events file contains event-level information (including MET,
exposure, ventilation, and dose) for individuals in the simulation.
Settings in the Control file allow the user to write this information for all
persons, every Nth person, or for a set of specified profile IDs.
Microenvironment
Summary
The Microenvironment Summary file provides a summary of the time and
exposure by microenvironment for each profile modeled in the
simulation.
Microenvironment
Results
The Microenvironment Results file provides an hour-by-hour time series
of microenvironment concentrations and parameters for a pollutant for
each modeled profile for each location ("Home", "Work", and "Other").
P± Microenvironment Results file is generated for each pollutant.
Output Tables
The Output Tables file contains a series of tables summarizing the
exposure (and dose, if calculated) results of the simulation for a pollutant.
The percentiles and exposure/dose cut-off points used in these tables are
defined in the Control file. A Tables file is generated for each pollutant.
   October, 2009
                      D-2
Draft - Do Not Cite or Quote

-------
                              Appendix E

    Mapping of CHAD Location Codes to Microenvironments Defined for
              Application of APEX4.3 to Carbon Monoxide.
October, 2009                        E-1       Draft - Do Not Cite or Quote

-------
Appendix E. Mapping Of Chad Location Codes To Microenvironments Defined For
Application Of Apex4.3 To Carbon Monoxide.
! Mapping of CHAD activity locations to two APEX microenvironments : in-vehicle (2) and other (1)
CHAD Loc. Description APEX
U
X
30000
30010
30020
30100
30120
30121
30122
30123
30124
30125
30126
30127
30128
30129
30130
30131
30132
30133
30134
30135
30136
30137
30138
30139
30200
30210
30211
30219
30220
30221
30229
30300
30310
30320
30330
30331
30332
30340
30341
30342
30400
31000
31100
31110
31120
31121
31122
31130
31140
31150
31160
31170
31171
31172
31900
31910
31200
31210
31220
31230
31300
Uncertain of correct code =
No data
Residence, general =
Your residence =
Other residence =
Residence, indoor =
Your residence, indoor =
. . . , kitchen =
. . . , living room or family room =
. . . , dining room =
. . . , bathroom =
. . . , bedroom =
. . . , study or office =
. . . , basement =
. . . , utility or laundry room =
. . . , other indoor =
Other residence, indoor =
. . . , kitchen =
. . . , living room or family room =
. . . , dining room =
. . . , bathroom =
. . . , bedroom =
. . . , study or office =
. . . , basement =
. . . , utility or laundry room =
. . . , other indoor =
Residence, outdoor =
Your residence, outdoor =
. . . , pool or spa =
. . . , other outdoor =
Other residence, outdoor =
. . . , pool or spa =
. . . , other outdoor =
Residential garage or carport =
. . . , indoor =
. . . , outdoor =
Your garage or carport =
. . . , indoor =
. . . , outdoor =
Other residential garage or carport =
. . . , indoor =
. . . , outdoor =
Residence, none of the above =
Travel, general
Motorized travel =
Car
Truck
Truck (pickup or van) =
Truck (not pickup or van) =
Motorcycle or moped =
Bus
Train or subway =
Airplane =
Boat
Boat, motorized =
Boat, other
Travel, other
. . . , other vehicle =
Non-motorized travel =
Walk
Bicycle or inline skates/skateboard =
In stroller or carried by adult =
Waiting for travel =
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
1
1
1
1
1
2
2
1
1
1
1
1
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
October, 2009
E-2
Draft - Do Not Cite or Quote

-------
31310
31320
31900
31910
32000
32100
32200
32300
32400
32500
32510
32520
32600
32610
32620
32700
32800
32810
32820
32900
32910
32920
33100
31200
31210
31220
31230
31300
31310
31320
31900
31910
32000
32100
32200
32300
32400
32500
32510
32520
32600
32610
32620
32700
32800
32810
32820
32900
32910
32920
33100
33200
33300
33400
33500
33600
33700
33800
33900
34100
34200
34300
35000
35100
35110
35200
35210
35220
35300
35400
35500
. . . , bus or train stop =
. . . , indoors =
Travel, other
. . . , other vehicle =
Non-residence indoor, general =
Office building/ bank/ post office =
Industrial/ factory/ warehouse =
Grocery store/ convenience store =
Shopping mall/ non-grocery store =
Bar/ night club/ bowling alley =
Bar or night club =
Bowling alley =
Repair shop =
Auto repair shop/ gas station =
Other repair shop =
Indoor gym /health club =
Childcare facility
. . . , house =
. . . , commercial =
Large public building =
Auditorium/ arena/ concert hall =
Library/ courtroom/ museum/ theater =
Laundromat =
Non-motorized travel =
Walk
Bicycle or inline skates/skateboard =
In stroller or carried by adult =
Waiting for travel =
. . . , bus or train stop =
. . . , indoors =
Travel, other
. . . , other vehicle =
Non-residence indoor, general =
Office building/ bank/ post office =
Industrial/ factory/ warehouse =
Grocery store/ convenience store =
Shopping mall/ non-grocery store =
Bar/ night club/ bowling alley =
Bar or night club =
Bowling alley =
Repair shop =
Auto repair shop/ gas station =
Other repair shop =
Indoor gym /health club =
Childcare facility =
. . . , house =
. . . , commercial =
Large public building =
Auditorium/ arena/ concert hall =
Library/ courtroom/ museum/ theater =
Laundromat =
Hospital/ medical care facility =
Barber/ hair dresser/ beauty parlor =
Indoors, moving among locations =
School
Restaurant =
Church
Hotel/ motel
Dry cleaners =
Indoor parking garage =
Laboratory =
Indoor, none of the above =
Non-residence outdoor, general =
Sidewalk, street
Within 10 yards of street =
Outdoor public parking lot /garage =
. . . , public garage =
. . . , parking lot =
Service station/ gas station =
Construction site =
Amusement park =
1
1
2
2
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
October, 2009
E-2
Draft - Do Not Cite or Quote

-------
           Playground                          =   1  H
           ...,  school grounds                 =   1  H
           ...,  public or park                 =   1  H
           Stadium or amphitheater             =   1  H
           Park/ golf course                   =   1  H
           Park                                    1  H
           Golf  course                             1  H
           Pool/ river/ lake                   =   1  H
           Outdoor restaurant/ picnic          =   1  H
           Farm                                    1  H
           Outdoor, none of the above          =   1  H
October, 2009                               E-4         Draft - Do Not Cite or Quote

-------
                                     Appendix F

 Isaacs et al. (2009) Reference Used in Developing D and A Statistics Input to
                                    APEX Model

The following presents a reformatted version of the Isaacs et al. (2009) presentation to allow for
easier reading. The presentation is included at the end of the Appendix in its entirety.
Statistical Properties of Longitudinal Time-Activity Data for Use in EPA Exposure Models
Kristin Isaacs1, Thomas McCurdy2, April Errickson3, Susan Forbes3, Graham Glen1, Stephen
Graham4, Lisa McCurdy5, Melissa Nysewander1, Luther Smith1, Nicolle Tulve2, and Daniel
Vallero2

1 Alion Science and Technology, Research Triangle Park, NC, 2Human Exposure and
Atmospheric Sciences Division, National Exposure Research Laboratory, US Environmental
Protection Agency, Research Triangle Park, NC, 3School of Information and Library Science,
University of North Carolina at Chapel Hill, Chapel Hill, NC, 4Office of Air Quality Planning
and Standards, US Environmental Protection Agency, Research Triangle Park, NC,
5Homemaker, Durham, NC.

ABSTRACT
Realistic simulation of longitudinal activity patterns is necessary for appropriately reproducing
the frequency and duration of pollutant exposures in human exposure models.  In EPA's
exposure models, longitudinal activity diaries for simulated persons are constructed from the 1-
day cross sectional  activity diaries in the Consolidated Human Activity Database (CHAD).
Recently, new algorithms have been developed to construct longitudinal diaries from CHAD
diaries based on realistic variance and autocorrelation properties of diary characteristics relevant
to pollutant exposure.  Characteristics of particular interest include time spent in particular
microenvironments and time spent in activities that produce high ventilation rates. However,
few multi-day data  are currently available for estimating accurate statistical properties for these
quantities. Results from a recent time-activity study of 10 adults and one newborn child are
presented here.  The participants recorded their personal location and activity for two-week
periods in each of four seasons in 2006 and 2007. The data were recorded 24 hours a day, in
increments as small as one minute. Additional recording periods for these same individuals are
expected in the future. The diaries for all subjects were assessed to calculate the between-person
variance, the within-person variance, and the autocorrelation for various lags in the time spent in
outdoor, residence,  indoor (non-residence), and vehicle microenvironments, as well as for time
spent performing high-METS activities.  The effectiveness of various day-type definitions (for
example, weekend versus weekday, or workday versus non-workday) for grouping similar diary
days is examined. Seasonal variation in activity patterns is analyzed. These data have the
potential to aid in the development of improved input variance and autocorrelation statistics for
longitudinal diary assembly algorithms in EPA's human exposure models.
   October 2009                          F-1            Draft - Do Not Cite or Quote

-------
INTRODUCTION
Recently, new methods of assembling multi-day diaries in human exposure models from cross-
sectional single-day diaries have been proposed that are based on the variance and
autocorrelation statistics of the simulated population (Glen et al. 2007).  Appropriately modeling
intra- and interindividual variability using such algorithms may be essential in producing
appropriate estimates of exposure. In addition, reproducing realistic autocorrelations in key diary
properties may be required for the modeling of episodic exposure patterns.
Previously, longitudinal time activity-location data collected in children in the Southern
California Chronic Ozone Exposure Study (Geyh et al. 1999) have been analyzed to obtain
estimates of appropriate measures of variance and autocorrelation for use in the longitudinal
algorithm. Data from a new study in adults are now presented.

BACKGROUND
Exposure models require construction of human activity diaries that cover the entire simulation
period of a model run. This period is often several months, a year, or even longer. In EPA's
models, human activity diaries are usually drawn from EPA's CHAD (Consolidated Human
Activity Database;  McCurdy et al., 2000; http://www.epa.gov/chadnetl), which typically
includes just one day (24 hours) of activities from each person. A "longitudinal" diary is one
that covers the same person over a long period of time.  While the SHEDS modeling period may
be of user-specified duration, it is assumed in this section to be one year, to provide a concrete
example.

Recently, a new longitudinal diary assembly algorithm has been developed (Glen et al. 2007)
based on the variance and autocorrelation properties of the modeled simulation. The new method
requires the user to:
      1) select the  diary property most relevant to exposure for the current application (such as
outdoor time or time spent in vehicles)
      2) specify the D statistic, which relates the within-person and between-person variances
for this diary property; and
      3) specify the 1-day lag autocorrelation in this diary property.

The new method is currently implemented in EPA's APEX and SHEDS-Air Toxics models.
The new method allows the modeler to apportion the total variance in the key diary property into
the within- and between-person variances  ow2 and  c\,2 by specifying the D statistic, defined to
D pertains to the population as a whole and is bounded by zero and one. A value of zero implies
all persons have the same average behavior, whereas a value of one implies the greatest possible
difference in mean behavior that is consistent with the total variance.
In addition to targeting the within-person and between-person variances through setting the D
statistic, the new diary assembly method optionally allows targeting of the day-to-day
autocorrelation.  This is a measure of the tendency for similar diaries to occur on consecutive
days. The lag-one autocorrelation in a variable y is for a person defined as
                                 N-1
   October 2009                           F-2            Draft - Do Not Cite or Quote

-------
The population autocorrelation A is the mean of the A values for all individuals.  Autocorrelation
could be of interest to the exposure modeler if the concentration time series were strongly
episodic, for example.  In the diary assembly, a positive autocorrelation indicates a tendency for
diaries with x-scores near each other to be used on consecutive days, while a negative
autocorrelation indicates a tendency for dissimilar x-scores to be used on consecutive days.
Some preliminary values of A have been derived  from the same data that were used to estimate
D (Glen et al., 2007).

METHODS
Activity Diary Study
Activity-location data were collected from 10 adults.  Nine of the adults were working
professionals; one was a stay-at-home parent. Nine of the adults recorded  their personal location
and activity for two-week periods in each of four seasons in 2006 and 2007. Additional data
were collected in one of the male subjects in 1999, another male (the 10th  adult) in 2002, and in
one of the females in 2008 (collected during maternity leave). The data were recorded 24 hours
a day, in increments  as small as one minute.  In this preliminary analysis, the time spent
outdoors, indoors, in travel, and performing hard  work each day were  calculated from the diaries.
"Hard work" was self-reported by each individual, as defined as activities requiring heavy
breathing and/or sweating.  Daily high temperatures and precipitation  amounts were acquired for
each day of the study.

Variance and Autocorrelation Statistics
Variance and lag-one autocorrelation statistics were calculated for the studied individuals.
Variance statistics were estimated for both the raw measured  variables (ie. time in minutes) and
the scaled ranks of the variable for each person on a given day. The ratio of the between-person
variance to the total variance (the sum of the between- and within-person variance) was
calculated for the population. This ratio, calculated using the  raw variables, is the intraclass
correlation coefficient (ICC), while the same ratio, calculated using the ranks, is D, the diversity
statistic. The autocorrelation A was also calculated using both the raw variables and the scaled
ranks of the variables on each day for each person in the study.

Analysis of Time Spent in Locations/Activities
The longitudinal data were assessed to support decisions on optimal diary pools for exposure
modeling.  Time spent in each of the  examined locations/activities were assessed as a function of
day of the week (weekday versus weekend), day type (workday versus nonworkday), season,
temperature, precipitation, and gender.  These analyses were undertaken to assess the utility of
different diary pool definitions.  Optimal definitions of diary  pools can adequately capture
temporal patterns in activities while maximizing the number of activity diaries available for
sampling on a given  day for a simulated individual.
Differences between groups were assessed with the Wilcoxon signed rank  test (for 2 groups) or
the Kruskal-Wallis test (for more than 2 groups).  The Wilcoxon rank sum (two-sample) test was
used to test differences between genders.
    October 2009                           F-3             Draft - Do Not Cite or Quote

-------
RESULTS AND DISCUSSION
Individual Variability
Figure 1 shows an example of the individual variability in time spent in different
locations/activities for a single male subject; a 367-day period from this subject is depicted.
Distributions of time for this subject are also shown.  These figures demonstrate the large amount
of intra-individual variability that can be seen in longitudinal activity studies. Distributions of
time spent in locations/activities for the population is shown in Figure 4[sic 2].
  Examples of Day-to-Day Variability For a Single Subject (M) "s
                 Time Spent Outdoors                  100
                                                S
                                                       sp.ni outdoors f« an individual
                 Time Spent in Travel
       100  200  300  400   500  600     Ct   100   200  300  403   500  400
Distribution ol Time Spent'in Travel lor an Individual Cumuli™ Diminution nlTl™ Sprnl in TIJV.I lor .in lndivldu..
                              400
                              350
              Time Spent Doing Haid Work
       100  300   100  401}  500
Distribution of Time Spent Doing Hard Woik for an indi-.-i du.

  300
  275
                                               175
                                               150
                                               125
                                               100
                                              '  75
                                                50
                                                25
                                                 0
                                                     100  200   KB  «0  500
                                                           MinuiK
Figure 1. Time series and distributions of time spent in locations/activities for 367 days of
data from a single male subject. Note high degree of interpersonal variability in behavior.
       Time Spiiil Doing Hid Woik
                                                         Time Spent Doing H*ilWort

I20C
1 •-«!'-
SCO
500
MO
-.;„-.


100
to/1
I
3 M
| 40
z M

Iflfl 3W IN

1 » '
s i
Jn
/ i
J J i

n n Hn Ha itm IMB tun
15
-^"^ ~~ » \
! j
! f
1 6
^ i
t
M •§ MB

;
/
/
/
y
^x.
M UD flM Itt MM m II
        Time S(«nl Outdira!
                                 Tkm Spent In Tl«il
                                                          Time Sp.nl Ouldoon
                                                                                   Tkne Spent In Travel
-UIJ
180
5 160
S no
3 120
i 100
I SC-
| 50
I 40
20
IW
ItJO
E i«
Q 120
o 100
V Ofl
v 1 n
\. 1 40
l*»-ww. »

)
rf
/ k
L..
Figure 2. Distributions of time spent in different activities for all days for all subjects.
    October 2009
   F-4
Draft - Do Not Cite or Quote

-------
Variance and Autocorrelation Statistics
D, ICC, and A values for the population for time spent in different locations/activities are given
in Table 1. Values of the ICC are lower than D; while A for the raw variables were higher than
A for the scaled ranks. These trends were also consistent with observed tends in the Southern
California data. Values were also calculated by gender (Table 2), temperature categories (Table
3), and daytypes (Table 4) where possible.

The D and ranked A values were compared to those calculated for children from the Southern
California Chronic Ozone Exposure Study (SCCOES). The diversity (D) for this group of adults
for outdoor time were higher than those calculated for the children (0.38 versus 0.19).  The D
values for travel time in the current study were also higher (0.18 in children versus 0.36 in this
study). These differences reflect the increased heterogeneity in these variable in the studied
adults versus the (relatively homogenous) studied children.  The A values calculated for outdoor
time in this study were virtually identical to those estimated using data from SCCOES.
In general, differences between D by temperature and daytypes were notable, even considering
the small number subjects in this study. There were gender differences observed in D; the
mechanism of these differences is unclear, but are likely  influenced by the activity patterns of the
female who was not a worker.

There were observed differences in A by temperature, but especially by daytype. This is not
unexpected, as it is reasonable that the behavior of working adults is more consistent day-to-day
on workdays. These trends should be confirmed by analysis of other longitudinal data. Note
however, that such differences in are only important when strongly episodic behavior or
exposure is of interest. In general, the values of D are much more relevant to exposure.

           Table 1. Variance and Autocorrelation Statistics: All Days/Subjects
Location/Activity
Indoors
Outdoors
Travel
Hard Work
ICC
0.26
0.16
0.14
0.18
D
0.33
0.38
0.31
0.22
A (Raw)
0.23
0.22
0.12
0.17
A (Ranks)
0.34
0.31
0.19
0.19
            Table 2. Variance and Autocorrelation Statistics: By Gender
Location/Activity
Males
Indoors
Outdoors
Travel
Hard Work
Females
Indoors
Outdoors
Travel
Hard Work
ICC

0.36
0.14
0.36
-0.01

0.08
0.07
0.05
0.15
D

0.54
0.22
0.46
0.15

0.09
0.27
0.16
0.24
A (Raw)

0.25
0.24
0.17
0.22

0.37
0.35
0.15
0.16
A (Ranks)

0.16
0.22
0.08
0.20

0.25
0.18
0.11
0.21
   October 2009
F-5
Draft - Do Not Cite or Quote

-------
             Table 3. Variance and Autocorrelation Statistics: By Temperature
Location/Activity
Days with max temp
less than 50 degrees
Indoors
Outdoors
Travel
Hard Work
Days with max temp
greater or equal to 50 degrees
Indoors
Outdoors
Travel
Hard Work
ICC


0,37
0,20
0,23
0,21


0.12
0,09
0,10
0.01
D


0,37
0,27
0,37
0.31


0,26
0.24
0.24
0.20
A (Raw)


0.23
0.33
0.20
0.14


0.45
0.39
0.34
0.35
A (Ranks)


0.19
0.18
0.09
0.14


0.23
0.20
0.09
0.14
               Table 4. Variance and Autocorrelation Statistics: By Daytype
Location/Activity
Workday
Indoors
Outdoors
Travel
Hard Work
NonWorkday
Indoors
Outdoors
Travel
Hard Work
ICC

0.37
0.19
0.45
0.20

0.12
0.11
0.09
0.06
D

0.47
0.31
0.47
0.25

0.21
0.14
0.24
0.07
A (Raw)

0.56
0.78
0.30
0.53

0.59
0.60
0.38
0.43
A (Ranks)

0.05
0.07
0.01
-0.12

0.24
0.19
0.08
0.18
Time Spent in Different Locations/Activities
The time spent in different locations/activities for different day types, seasons, temperature
categories are presented in Figures 3-6. The effects of gender and precipitation were also
studied. There were no significant differences for these categories, and thus plots are not shown.
The plotted data represent all days for all subjects. The medians are represented by the midline
of the boxes,  the first and third quartiles by the ends of the boxes,  and the means by the stars.
The whiskers extend to cover data that lies beyond the boxedbut within the quartiles plus 1.5
times the interquartile range. Points outside this range are plotted.

Results by  day of the week and day type are presented in Figure 3. Day type (workday versus
nonworkday) was at least as good as day of the week in categorizing time/activities. This trend is
similar to that seen in a recent analysis of the larger, cross-sectional database of diaries from The
National Human Activity Pattern Survey (NHAPS, data not shown). That analysis indicated that
a workday/nonworkday was a better discriminator of time spent outside than a weekday/weekend
split.  As such, further comparisons are also presented for both workdays and nonworkdays.
    October 2009
F-6
Draft - Do Not Cite or Quote

-------
       Time Spent Outdoors By Day of Irve Week
                                          Spent in Tiatfel By Day of tile Week
                                                                        e Spent Doing Haid Woik By Day of the Week
        Tims Spent Outdoors By Day Type
                                         Time Spent in Travel By Day Type
                                                                        Time Spent Damn Hard Wo* By D«y Type
Figure 3. Time spent in different locations/activities as a function of day of the week, and
daytype  (workday versus nonworkdays).
The effect of season on time spent in locations/activities is shown in Figure 4. Seasonal effects
were apparent for time spent outdoors on nonworkdays, and for time spent doing hard work.
Travel was also affected by season, likely due to the large number of work-related travel days in
the fall for this particular group of workers.
         Time Spent Outdoors By Season
                                         Time Spent in Travel By Season
                                                                        Time Spent in Doing Hard Wo*k By Season
      Time Spent Outdoors By Season and Day Type
                                      Time Spent in Travel 8y Season and Day Type
                                                                     Time Spent Doing Hard Work By Season and Day Type
Figure 4. Time spent in different locations/activities as a function of season and daytype.
    October 2009
F-7
Draft - Do Not Cite or Quote

-------
The effect of temperature category is shown in Figure 5.  The temperature category was defined
as warmer = maximum temperature greater than or equal to 75 degrees, colder= maximum
temperature less than 75 degrees. Temperature category was better than or as good as season in
discriminating behavior in time spent outdoors, even when daytype was considered.
     Time Spent Outdoors By Tempeiature Category
                                    Time Spent in Travel By Temperature Category
                                                                 Time Spent Doing Hard Work By Temperature Category
  Time Spent Outdoors By Temperature Category and Day Type
                                Time Spent in Travel By Temperature Category and Day Type    Time Spent Doing Hard Work By Temperature Category and Day Type
Figure 5. Time spent in different locations/activities as a function of temperature category
(colder:  max temp< 75 degrees, warmer: max temp> 75 degrees) and daytype.
CONCLUSIONS
    •    The diversity (D) and autocorrelation (A) for this group of adults for outdoor time were
       higher than those calculated for children in a previous study. Thus these data provide
       some justification for considering age when considering D and A input values for EPA's
       exposure models.
    •   While the current data suggest possible effects of temperature,  daytype and gender on
       diversity (D) and autocorrelation (A), more data from this and other studies are needed to
       confirm these findings.  Such results could aid in the fine-tuning of the longitudinal diary
       algorithm.
    •    The analysis of the time spent in locations was consistent with recent findings from from
       cross-sectional diary studies indicating that workdays/nonworkdays may be a better
       grouping for diary pools than weekdays/weekends.
       Temperature category was at least as good as season in discriminating behavior for this
       population for time sprnt outdoors, especially when daytype was considered.  Such
       breakdowns by temperature and daytype may eliminate the need for diary pools different
       seasons, providing larger pools for diary sampling on a given day. Further analysis with
       other time-activity data  can confirm this trend.
    October 2009
F-8
Draft - Do Not Cite or Quote

-------
FUTURE WORK
   •    We plan to repeat this type of study periodically.  Data will be compared to/combined
       with analyses of other available longitudinal time/location/activity studies.

DISCLAIMER
The information in this document has been funded wholly (or in part) by the U. S. Environmental
Protection Agency (EPA contract 68-D-00-206).  It has been subjected to review by the EPA and
approved for publication. Approval does not signify that the contents necessarily reflect the
views of the Agency, nor does mention of trade names or commercial products constitute
endorsement or recommendation for use.

REFERENCES
Geyh A.S., Xue J., Ozkaynak H., and Spengler J.D.  (2000).  The Harvard Southern California
Chronic Ozone Exposure study: assessing ozone exposure of gradeschool-age children in two
southern California communities. Environ Health Perspect.  108:265-270.

Glen G., Smith L., Isaacs K., McCurdy T., and Langstaff J. (2008). A new method of
longitudinal diary assembly for human exposure modeling. J Expo Sci Environ Epidemiol.
18(3):299-311.

Xue J., McCurdy T., Spengler J., and Ozkaynak H. (2004). Understanding variability in time
spent in selected locations for 7-12-year old children. J Expo Sci Environ Epdemiol. 14:222-233.

McCurdy T., Glen G., Smith L., and Lakkadi L. (2000).  The National Exposure Research
Laboratory's Consolidated Human Activity Database. J Expo Anal Environ Epidemiol. 10:566-
78.
   October 2009                          F-9            Draft - Do Not Cite or Quote

-------
Isaacs et al. (2009) in original format.
             Statistical Properties  of Longitudinal  Time-Activity Data for Use  in  EPA Exposure  Models
       Kristin Isaacs1, Thomas McCurdy2, April Errickson3, Susan Forbes3, Graham Glen1, Stephen Graham4, Lisa McCurdy5, Melissa Nysewander1, Luther Smith1, Nicolle Tulve2, and Daniel Vallero2
       1Alion Science and Technology, Research Triangle Park, NC, 2Human Exposure and Atmospheric Sciences Division, National Exposure Research Laboratory, US Environmental Protection Agency, Research Triangle Park, NC, 3School of Information and
                  Library Science, University of North Carolina at Chapel Hill, Chapel Hill, NC, 4Office of Air Quality Planning and Standards, US Environmental Protection Agency, Research Triangle Park, NC, 5Homemaker, Durham, NC.
 ABSTRACT
 INTRODUCTION

BACKGROUND
                                  METHODS
                                  Variance and Autocorrelation Statistics
                                                jsubject. Notehighdegre.
                                                                        T.M»2 V..UK.
                                                                                        **•*"	*«***«
                                                                               W     05*
                                                                               41     Oil    OK
                                                                               M     O.S7    ry»
                                                                               «     ft*
                                                                                    as*	n>
                                                                                    «M    4*t

                                                                                                       category (colder: maxtemp=75 degrees, warmer: max tempi 75 degrees) and
                                                                                                       daytype.
                                                                                                      RESULTS AND DISCUSSION
                                                                                                      IniliviilualVatiability
                                                                                                                                             H Spent in Different Localions/AclNilies
                                                                                                                                            CONCLUSIONS
                                                                                                                                            DISCLAIMER
                                                                      ek, and daytype (workday ui
                                    October 2009
F-10
C/te or Quote

-------
                                 Appendix G

          Analysis of CHAD Diaries for Time Spent in Vehicles.

       The U. S. Census Bureau (2009) provides an on-line facility for accessing the
detailed census data included in their Summary File 3 (SF3). Using this resource, we
obtained information on travel time to work for "workers 16 years and over" (US Census
Bureau, 2009, Table P31) specific to Denver County, Colorado and Los Angeles, CA.
The counts in Table P31 for trips to work places  other than home were converted into the
percentages listed in Columns 2 and 3 of Table G-l. Although the P31  statistics applied
to people 16 years or older, we assumed that the  statistics were generally applicable to
people 18 years or older.
       We next determined the number of 24-hour diaries in CHAD that met the
following criteria: the subject was 18+ years of age and the diary reported at least one
minute in a motor vehicle between 6 am and 9 am. The number of these diaries that had
in-vehicle times corresponding to the bins listed in Table G-l are listed in Column 4 in
and the values were converted to the percentage values listed in Column 5.
Table G-l.   Representation of Denver and LA Commuting Characteristics in
             CHAD Diaries.
Travel time
(minutes)
(1)
Ito9
10 to 19
20 to 29
30 to 39
40 to 59
60 to 89
90+
Total
Percent of
commuters according
to SF3 census data
for Denver County
(2)
10.28
31.96
24.15
18.60
9.29
3.80
1.73
100.00
Percent of commuters
according to SF3
census data for Los
Angeles County
(3)
7.75
25.92
21.04
21.37
13.57
6.99
3.35
100.00
24-hour diaries meeting
inclusion criteria3
Number in
CHAD
(4)
563
1,676
1,068
1,111
665
407
258
5,748
Percent in
CHAD
(5)
9.79
29.16
18.58
19.33
11.57
7.08
4.49
100.00
""Subjects are 18+ years of age. Diaries include 1+ minute in motor vehicle between 6 am and 9 am.

-------

-------
oo
o
o
o


Q

-------