*«< PRC*4' Risk and Exposure Assessment to Support the Review of the NC>2 Primary National Ambient Air Quality Standard: First Draft ------- EPA-452/P-08-001 April 2008 Risk and Exposure Assessment to Support the Review of the NC>2 Primary National Ambient Air Quality Standard: First Draft U.S. Environmental Protection Agency Office of Air Quality Planning and Standards Research Triangle Park, North Carolina ------- Disclaimer This draft document has been prepared by staff from the Ambient Standards Group, 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. This document is being circulated to obtain review and comment from the Clean Air Scientific Advisory Committee (CASAC) and the general public. Comments on this draft document should be addressed to Scott Jenkins, U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards, C504-06, Research Triangle Park, North Carolina 27711 (email: Jenkins.scott@epa.gov). ------- Table of Contents List of Tables v List of Figures vii 1. INTRODUCTION 1 1.1 OVERVIEW 1 1.2 HISTORY 4 1.2.1 History of the NO2NAAQS 4 1.2.2 Health Evidence from Previous Review 5 1.2.3 Assessment from Previous Review 5 1.3 SCOPE OF THE RISK AND EXPOSURE ASSESSMENT FOR THE CURRENT REVIEW 6 1.3.1 Species of Nitrogen Oxides Included in Analyses 6 1.3.2 Scenarios Addressed in First Draft Assessment 6 2. SOURCES, AMBIENT LEVELS, AND EXPOSURES 8 2.1 SOURCES OF NO2 8 2.2 AMBIENT LEVELS OF NO2 8 2.3 EXPOSURE TO NO2 9 3. AT RISK POPULATIONS 12 3.1 OVERVIEW 12 3.2 DISEASE AND ILLNESS 12 3.3 AGE 12 3.4 PROXIMITY TO ROADWAYS 13 4. HEALTH EFFECTS 14 4.1 INTRODUCTION 14 4.2 ADVERSE RESPIRATORY EFFECTS FOLLOWING SHORT-TERM EXPOSURES 14 4.2.1 Overview 14 4.2.2 Effects Based on Controlled Human Exposure Studies 15 4.2.2.1 Overview 15 4.2.2.2 Airways Responsiveness 16 4.2.2.3 Conclusions 18 4.2.3 Epidemiology Literature 20 4.2.3.1 Hospital Admissions and Emergency Department Visits 20 4.2.3.2 Respiratory Illness and Symptoms 21 4.2.3.3 Conclusions Regarding the Epidemiology Literature 23 4.2.4 Toxicology Literature 24 4.3 OTHER AD VERSE EFFECTS FOLLOWING SHORT-TERM EXPOSURES 24 4.4 ADVERSE EFFECTS FOLLOWING LONG-TERM EXPOSURES 25 5. OVERVIEW OF RISK AND EXPOSURE ASSESSMENT 28 April 2008 Draft ------- 5.1 INTRODUCTION 28 5.2 GOALS 30 5.3 GENERAL APPROACH 30 5.4 ADDITIONAL CONSIDERATIONS 31 5.4.1 Adjustment of Ambient Air Quality 32 5.4.2 Adjustment of Potential Health Effect Benchmark Levels 33 6. AMBIENT AIR QUALITY AND HEALTH RISK CHARACTERIZATION 34 6.1 OVERVIEW 34 6.2 APPROACH 35 6.2.1 Air Quality Data Screen 36 6.2.2 Selection of Locations for Air Quality Analysis 37 6.2.3 Estimation of On-Road Concentrations using Ambient Concentrations 38 6.3 AIR QUALITY AND HEALTH RISK CHARACTERIZATION RESULTS 41 6.3.1 Ambient Air Quality (As Is) 41 6.3.2 Ambient Air Quality Adjusted to Just Meet the Current Standard 46 6.3.3 On-Road Concentrations Derived From Ambient Air Quality (As Is) 47 6.3.4 On-Road Concentrations Derived From Ambient Air Quality Adjusted to Just Meet the Current Standard 54 6.4 UNCERTAINTY AND VARIABILITY 58 6.4.1 Air Quality Data 59 6.4.2 Measurement Technique for Ambient NO2 60 6.4.3 Temporal Representation 60 6.4.4 Spatial Representation 61 6.4.5 Air Quality Adjustment Procedure 61 6.4.6 On-Road Concentration Simulation 62 6.4.7 Health Benchmark 64 7. EXPOSURE ASSESSMENT AND HEALTH RISK CHARACTERIZATION 66 7.1 OVERVIEW 66 7.2 OVERVIEW OF HUMAN EXPOSURE MODELING USING APEX 66 7.3 CHARACTERIZATION OF STUDY AREAS 69 7.3.1 Study Area Selection 69 7.3.2 Study Area Descriptions 70 7.4 CHARACTERIZATION OF AMBIENT HOURLY AIR QUALITY DATA USING AERMOD 71 7.4.1 Overview 71 7.4.2 General Model Inputs 72 7.4.2.1 Meteorological Data 72 7.4.2.2 Surface Characteristics and Land Use Analysis 73 7.4.2.3 Additional AERMOD Input Specifications 76 7.4.3 Emissions Estimates 77 7.4.3.1 On-Road Emissions Preparation 77 7.4.3.2 Stationary Sources Emissions Preparation 82 7.4.3.3 Fugitive and Airport Emissions Preparation 84 7.4.4 Receptor Locations 88 7.4.5 Estimate Air Quality Concentrations 90 7.5 POPULATION MODELED 91 April 2008 Draft iii ------- 7.5.1 Simulated Individuals 92 7.5.2 Employment Probabilities 92 7.5.3 Commuting Patterns 92 7.6 CONSTRUCTION OF LONGITUDINAL ACTIVITY SEQUENCES 93 7.7 CALCULATING MICROENVIRONMENTAL CONCENTRATIONS 96 7.7.1 Microenvironments Modeled 97 7.7.2 Microenvironment Descriptions 98 7.7.2.1 Microenvironment 1: Indoor-Residence 98 7.7.2.2 Microenvironments 2-7: All Other Indoor Microenvironments 103 7.7.2.3 Microenvironments 8 and 9: Outdoor Microenvironments 104 7.7.2.4 Microenvironment 10: Outdoors-General 104 7.7.2.5 Microenvironments 11 and 12: In Vehicle- Cars and Trucks, and Mass Transit 104 7.8 EXPOSURE AND HEALTH RISK CALCULATIONS 105 7.9 EXPOSURE MODELING AND HEALTH RISK CHARACTERIZATION RESULTS 107 7.9.1 Overview 107 7.9.2 Annual Average Exposure Concentrations (as is) 108 7.9.3 One-Hour Exposures (as is) Ill 7.9.3.1 Maximum Estimated Exposure Concentrations 112 7.9.3.2 Number of Estimated Exposures above Selected Levels 112 7.9.3.3 Number of Repeated Exposures Above Selected Levels 119 7.9.4 One-Hour Exposures Associated with Just Meeting the Current Standard 121 7.9.4.1 Number of Estimated Exposures above Selected Levels 121 7.9.4.2 Number of Repeated Exposures Above Selected Levels 123 7.10 VARIABILITY AND UNCERTAINTY 124 7.10.1 Introduction 124 7.10.2 Input Data Evaluation 126 7.10.3 Meteorological Data 126 7.10.4 Air Quality Data 126 7.10.5 Population and Commuting Data 127 7.10.6 Activity Pattern Data 127 7.10.7 Air Exchange Rates 128 7.10.7.1 Extrapolation among cities 128 7.10.7.2 Within CSA uncertainty 129 7.10.7.3 Variation in measurement averaging times 130 7.10.8 Air Conditioning Prevalence 130 7.10.9 Indoor Source Estimation 131 8. REFERENCES 133 April 2008 Draft iv ------- List of Tables Number age Table 1. Summary of Key Controlled Human Exposure Studies of Airways Responsiveness.... 19 Table 2. Counts of complete site-years of NO2 monitoring data 37 Table 3. Locations selected for Tier INO2 Air Quality Characterization, associated abbreviations, and values of selection criteria 39 Table 4. Monitoring site-years and annual average NO2 concentrations for two monitoring periods, historic and recent air quality data (as is) 43 Table 5. Number of exceedances of short-term (1-hour) potential health effect benchmark levels in a year, 1995-2000 historic NO2 air quality (as is) 44 Table 6. Number of exceedances of short-term (1-hour) potential health effect benchmark levels in ayear, 2001-2006 recent NO2 air quality (as is) 45 Table 7. Estimated annual average NO2 concentrations for two monitoring periods, historic and recent air quality data adjusted to just meet the current standard (0.053 ppm annual average) 48 Table 8. Estimated number of exceedances of short-term (1-hour) potential health effect benchmark levels in a year, 1995-2000 NO2 air quality adjusted to just meet the current standard (0.053 ppm annual average) 49 Table 9. Estimated number of exceedances of short-term (1-hour) health effect benchmark levels in a year, 2001-2006 NO2 air quality adjusted to just meet the current standard (0.053 ppm annual average) 50 Table 10. Estimated annual average on-road concentrations for two monitoring periods, historic and recent ambient air quality (as is) 51 Table 11. Estimated number of exceedances of short-term (1-hour) potential health effect benchmark levels in a year on-roads, 1995-2000 historic NO2 air quality (as is) 52 Table 12. Estimated number of exceedances of short-term (1-hour) potential health effect benchmark levels in a year on-roads, 2001-2006 historic NO2 air quality (as is) 53 Table 13. Estimated annual average on-road concentrations for two monitoring periods, air quality data adjusted to just meet the current standard (0.053 ppm annual average).. 55 Table 14. Estimated number of exceedances of short-term (1-hour) potential health effect benchmark levels in a year on-roads, 1995-2000 historic NO2 air quality adjusted to just meet the current standard (0.053 ppm annual average) 56 Table 15. Estimated number of exceedances of short-term (1-hour) potential health effect benchmark levels in a year on-roads, 2001-2006 recent NO2 air quality adjusted to just meet the current standard (0.053 ppm annual average) 57 Table 16. Summary of qualitative uncertainty analysis for the air quality and health risk characterization 65 Table 17. Seasonal specifications by selected case-study locations 74 Table 18. Characterization of monthly precipitation levels in selected case-study locations compared to NCDC 30-year climatic normals, 2001-2003 76 Table 19. Statistical summary of AADT volumes (one direction) for Philadelphia County AERMOD simulations 78 Table 20. Average calculated speed by link type for Philadelphia County 80 Table 21. On-road area source sizes for Philadelphia County 81 April 2008 Draft v ------- Table 22. Stationary NOx emission sources modeled in Philadelphia County 84 Table 23. Emission parameters for the three Philadelphia County fugitive NOx area emission sources 85 Table 24. Philadelphia International airport (PHL)NOx emissions 87 Table 25. Philadelphia CMSA NOx monitors 88 Table 26. Comparison of ambient monitoring and AERMOD predicted NO2 concentrations... 91 Table 27. Studies in CHAD used for the exposure analysis 94 Table 28. List of microenvironments modeled and calculation methods used 98 Table 29. Air conditioning (A/C) prevalence estimates with 95% confidence intervals 99 Table 30. Geometric means (GM) and standard deviations (GSD) for air exchange rates by city, A/C type, and temperature range 100 Table 31. Probability of gas stove cooking by hour of the day 103 Table 32. Adjustment factors and potential health effect benchmark levels used by APEX to simulate just meeting the current standard 107 April 2008 Draft VI ------- List of Figures Number Page Figure 1. Example of Light- and heavy-duty vehicle NOX emissions grams/mile (g/mi) for arterial and freeway functional classes, Philadelphia 2001 80 Figure 2. Locations of the four ancillary area sources. Also shown are centroid receptor locations 86 Figure 3. Centroid locations within fixed distances to major point and mobile sources 89 Figure 4. Frequency distribution of distance between each Census receptor and its nearest road- centered receptor 90 Figure 5. Distribution of AERMOD predicted annual average NO2 concentrations at each of the 16,857 receptors in Philadelphia County for years 2001-2003 109 Figure 6. Estimated annual average total NO2 exposure concentrations for all simulated persons in Philadelphia County, using modeled 2001-2003 air quality (as is), with modeled indoor sources 110 Figure 7. Comparison of AERMOD predicted and ambient monitoring annual average NO2 concentrations (as is) and APEX exposure concentrations (with and without modeled indoor sources) in Philadelphia County for year 2002 Ill Figure 8. Estimated maximum NO2 exposure concentration for all simulated persons in Philadelphia County, using modeled 2001-2003 air quality (as is), with and without modeled indoor sources. Values above the 99th percentile are not shown 113 Figure 9. Estimated number of all simulated asthmatics in Philadelphia County with at least one NO2 exposure at or above the potential health effect benchmark levels, using modeled 2001-2003 air quality (as is), with modeled indoor sources 114 Figure 10. Estimated number of simulated asthmatic children in Philadelphia County with at least one NO2 exposure at or above the potential health effect benchmark levels, using modeled 2001-2003 air quality (as is), with modeled indoor sources 114 Figure 11. Comparison of the estimated number of all simulated asthmatics in Philadelphia County with at least one NO2 exposure at or above potential health effect benchmark levels, using modeled 2002 air quality (as is) , with and without modeled indoor sources 115 Figure 12. Fraction of time all simulated persons in Philadelphia County spend in the twelve microenvironments associated with the potential NO2 health effect benchmark levels, a) > 200 ppb, b) > 250 ppb, and c) > 300 ppb, year 2002 simulation with indoor sources 117 Figure 13. Fraction of time all simulated persons in Philadelphia County spend in the twelve microenvironments associated with the potential NO2 health effect benchmark levels, a) > 200 ppb, b) > 250 ppb, and c) > 300 ppb, year 2002 simulation without indoor sources 118 Figure 14. Estimated percent of all asthmatics in Philadelphia County with repeated NO2 exposures above potential health effect benchmark levels, using 2003 modeled air quality (as is), with modeled indoor sources 120 Figure 15. Estimated percent of all asthmatics in Philadelphia County with repeated NO2 exposures above potential health effect benchmark levels, using modeled 2002 air quality (as is), with and without indoor sources 121 April 2008 Draft vii ------- Figure 16. Estimated percent of all asthmatics in Philadelphia with at least one exposure at or above the potential health effect benchmark level, using modeled 2001-2003 air quality just meeting the current standard, with modeled indoor sources 122 Figure 17. Estimated number of all asthmatics in Philadelphia with at least one exposure at or above the potential health effect benchmark level, using modeled 2002 air quality just meeting the current standard, with and without modeled indoor sources 123 Figure 18. Estimated percent of asthmatics in Philadelphia County with repeated exposures above health effect benchmark levels, using modeled 2002 air quality just meeting the current standard, with and without modeled indoor sources 124 Figure 19. Geometric mean and standard deviation of air exchange rate bootstrapped for Los Angeles residences with A/C, temperature range from 20-25 degrees centigrade (from EPA, 2007g) 130 April 2008 Draft viii ------- 1. INTRODUCTION 1.1 OVERVIEW 3 The U.S. Environmental Protection Agency (EPA) is conducting a review of the national 4 ambient air quality standards (NAAQS) for nitrogen dioxide (NO2). Sections 108 and 109 of the 5 Clean Air Act (The Act) govern the establishment and periodic review of the air quality criteria 6 and the NAAQS. These standards are established for pollutants that may reasonably be 7 anticipated to endanger public health and welfare, and whose presence in the ambient air results 8 from numerous or diverse mobile or stationary sources. The NAAQS are based on air quality 9 criteria, which reflect the latest scientific knowledge useful in indicating the kind and extent of 10 identifiable effects on public health or welfare that may be expected from the presence of the 11 pollutant in ambient air. The EPA Administrator promulgates and periodically reviews primary 12 (health-based) and secondary (welfare-based) NAAQS for such pollutants. Based on periodic 13 reviews of the air quality criteria and standards, the Administrator makes revisions in the criteria 14 and standards and promulgates any new standards as may be appropriate. The Act also requires 15 that an independent scientific review committee advise the Administrator as part of this NAAQS 16 review process, a function now performed by the Clean Air Scientific Advisory Committee 17 (CAS AC). 18 The Agency has recently made a number of changes to the process for reviewing the 19 NAAQS (described at http://www.epa.gov/ttn/naaqs/). In making these changes, the Agency 20 consulted with CAS AC. This new process, which is being applied to the current review of the 21 NC>2 NAAQS, contains four major components. Each of these components, as they relate to the 22 review of the NC>2 primary NAAQS, is described below. 23 The first of these components is an integrated review plan. This plan presents the 24 schedule for the review, the process for conducting the review, and the key policy-relevant 25 science issues that guide the review. The integrated review plan for this review of the NC>2 26 primary NAAQS is presented in the Integrated Review Plan for the Primary National Ambient April 2008 Draft 1 ------- 1 Air Quality Standard for Nitrogen Dioxide (EPA, 2007a). The policy-relevant questions 2 identified in this document to guide the review are: 3 • Has new information altered the scientific support for the occurrence of health effects 4 following short- and/or long-term exposure to levels of NOX found in the ambient air? 5 • What do recent studies focused on the near-roadway environment tell us about health 6 effects of NOX? 7 • At what levels of NOX exposure do health effects of concern occur? 8 • Has new information altered conclusions from previous reviews regarding the plausibility 9 of adverse health effects caused by NOX exposure? 10 • To what extent have important uncertainties identified in the last review been reduced 11 and/or have new uncertainties emerged? 12 • What are the air quality relationships between short-term and long-term exposures 13 toNOx? 14 Additional questions will become relevant if the evidence suggests that revision of the current 15 standard might be appropriate. These questions are: 16 • Is there evidence for the occurrence of adverse health effects at levels of NOX lower than 17 those observed previously? If so, at what levels and what are the important uncertainties 18 associated with that evidence? 19 • Do exposure estimates suggest that exposures of concern for NOx-induced health effects 20 will occur with current ambient levels of NC>2 or with levels that just meet current, or 21 potential alternative, standards? If so, are these exposures of sufficient magnitude such 22 that the health effects might reasonably be judged to be important from a public health 23 perspective? What are the important uncertainties associated with these exposure 24 estimates? 25 • Do the evidence, the air quality assessment, and the risk/exposure assessment provide 26 support for considering different standard indicators or averaging times? 27 • What range of levels is supported by the evidence, the air quality assessment, and the 28 risk/exposure assessments? What are the uncertainties and limitations in the evidence 29 and the assessments? April 2008 Draft ------- 1 • What is the range of forms supported by the evidence, the air quality assessment, and the 2 exposure/risk assessments? What are the uncertainties and limitations in the evidence 3 and the assessments? 4 The second component of the review process is a science assessment. A concise 5 synthesis of the most policy-relevant science has been compiled into a draft Integrated Science 6 Assessment (draft ISA). The draft ISA is supported by a series of annexes that contain more 7 detailed information about the scientific literature. The current draft of the ISA to support this 8 review of the NO2 primary NAAQS is presented in the Integrated Science Assessment for 9 Oxides of Nitrogen - Health Criteria (Second External Review Draft), henceforth referred to as 10 the draft ISA (EPA, 2008a). 11 The third component of the review process is a risk and exposure assessment, the first 12 draft of which is described in this document. The purpose of this draft document is to 13 communicate EPA's assessment of exposures and risks associated with ambient NO2. It is 14 supported by a more detailed technical support document, henceforth referred to as the draft 15 TSD. This first draft of the risk and exposure assessment develops estimates of human 16 exposures and risks associated with current ambient levels of NO2 and with levels that just meet 17 the current standard. The second draft of this document will also consider levels of NO2 that just 18 meet any potential alternative standards that are identified for consideration. The results of the 19 risk and exposure assessment will be considered alongside the health evidence, as evaluated in 20 the final ISA, to inform the policy assessment and rulemaking process (see below). The draft 21 plan for conducting the risk and exposure assessment to support the NO2 primary NAAQS is 22 presented in the Nitrogen Dioxide Health Assessment Plan: Scope and Methods for Exposure 23 and Risk Assessment, henceforth referred to as the Health Assessment Plan (EPA, 2007b). 24 The fourth component of the process is the policy assessment and rulemaking. The 25 Agency's views on policy options will be published in the Federal Register as an advance notice 26 of proposed rulemaking (ANPR). This policy assessment will address the adequacy of the 27 current standard and of any potential alternative standards, which will be defined in terms of 28 indicator, averaging time, form,1 and level. To accomplish this, the policy assessment will 29 consider the results of the final risk and exposure assessment as well as the scientific evidence 1 The "form" of a standard defines the air quality statistic that is to be compared to the level of the standard in determining whether an area attains the standard. April 2008 Draft 3 ------- 1 (including evidence from the epidemiological, controlled human exposure, and animal 2 toxicological literatures) evaluated in the final ISA. Taking into consideration CASAC advice 3 and recommendations as well as public comment on the ANPR, the Agency will publish a 4 proposed rule, to be followed by a public comment period. Taking into account comments 5 received on the proposed rule, the Agency will issue a final rule to complete the rulemaking 6 process. 7 1.2 HISTORY 8 1.2.1 History of the NO2 NAAQS 9 On April 30, 1971, EPA promulgated identical primary and secondary NAAQS for NO2 10 under section 109 of the Act. The standards were set at 0.053 parts per million (ppm), annual 11 average (36 FR 8186). In 1982, EPA published Air Quality Criteria for Oxides of Nitrogen 12 (EPA, 1982), which updated the scientific criteria upon which the initial NO2 standards were 13 based. On February 23, 1984, EPA proposed to retain these standards (49 FR 6866). After 14 taking into account public comments, EPA published the final decision to retain these standards 15 on June 19, 1985 (50 FR 25532). 16 On July 22, 1987, EPA announced that it was undertaking plans to revise the 1982 air 17 quality criteria (52 FR 27580). In November 1991, EPA released an updated draft air quality 18 criteria document for CASAC and public review and comment (56 FR 59285). The draft 19 document provided a comprehensive assessment of the available scientific and technical 20 information on health and welfare effects associated with NO2 and other oxides of nitrogen. The 21 CASAC reviewed the draft document at a meeting held on July 1, 1993 and concluded in a 22 closure letter to the Administrator that the document "provides a scientifically balanced and 23 defensible summary of current knowledge of the effects of this pollutant and provides an 24 adequate basis for EPA to make a decision as to the appropriate NAAQS for NO2" (Wolff, 25 1993). The Air Quality Criteria Document for the Oxides of Nitrogen was then finalized (EPA, 26 1993). 27 The EPA also prepared a Staff Paper that summarized an air quality assessment for NO2 28 conducted by the Agency (McCurdy, 1994), summarized and integrated the key studies and 29 scientific evidence contained in the revised air quality criteria document, and identified the 30 critical elements to be considered in the review of the NO2 NAAQS. The CASAC reviewed two April 2008 Draft 4 ------- 1 drafts of the Staff Paper and concluded in a closure letter to the Administrator (Wolff, 1995) that 2 the document provided a "scientifically adequate basis for regulatory decisions on nitrogen 3 dioxide." In September of 1995, EPA finalized the Staff Paper entitled, "Review of the National 4 Ambient Air Quality Standards for Nitrogen Dioxide: Assessment of Scientific and Technical 5 Information" (EPA, 1995). 6 In October 1995, the Administrator announced her proposed decision not to revise either 7 the primary or secondary NAAQS for NO2 (60 FR 52874; October 11, 1995). A year later, the 8 Administrator made a final determination not to revise the NAAQS for NO2 after careful 9 evaluation of the comments received on the proposal (61 FR 52852, October 8, 1996). The level 10 for both the existing primary and secondary NAAQS for NO2 is 0.053 parts per million (ppm) 11 (100 micrograms per cubic meter of air [|j,g/m3]), annual arithmetic average, calculated as the 12 arithmetic mean of the l-hourNO2 concentrations. 13 1.2.2 Health Evidence from Previous Review 14 The prior Air Quality Criteria Document (AQCD) for Oxides of Nitrogen (EPA, 1993) 15 concluded that there were two key health effects of greatest concern at ambient or near-ambient 16 levels of NO2, increased airways responsiveness in asthmatic individuals after short-term 17 exposures and increased occurrence of respiratory illness in children with longer-term exposures. 18 Evidence also was found for increased risk of emphysema, but this was of maj or concern only 19 with exposures to levels of NO2 much higher than then-current ambient levels. The evidence 20 regarding airways responsiveness was drawn largely from controlled human exposure studies. 21 The evidence for respiratory illness was drawn from epidemiological studies that reported 22 associations between respiratory symptoms and indoor exposures to NO2 in people living in 23 homes with gas stoves. The biological plausibility of the epidemiological results was supported 24 by toxicological studies that detected changes in lung host defenses following NO2 exposure. 25 Subpopulations considered potentially more susceptible to the effects of NO2 included 26 individuals with preexisting respiratory disease, children, and the elderly. 27 1.2.3 Assessment from Previous Review 28 In the previous review of the NO2 NAAQS, risks were assessed by comparing ambient 29 monitoring data, which was used as a surrogate for exposure, with health benchmark levels 30 identified from controlled human exposure studies. At the time of the review, a few studies April 2008 Draft 5 ------- 1 indicated the possibility for adverse health effects due to short-term (e.g., 1-hour) exposures 2 between 0.20 ppm and 0.30 ppm NC>2. Therefore, the focus of the assessment was on the 3 potential for short-term (i.e., 1-hour) exposures to NO2 levels above potential health benchmarks 4 in this range. The assessment used monitoring data from the years 1988-1992 and screened for 5 sites with one or more hourly exceedances of potential short-term health effect benchmarks. 6 Predictive models were then constructed to relate the frequency of hourly concentrations above 7 short-term health effect benchmarks to a range of annual average concentrations, including the 8 current standard. Based on the results of this analysis, both CAS AC (Wolff, 1995) and the 9 Administrator (60 FR 52874) concluded that the minimal occurrence of short-term peak 10 concentrations at or above a potential health effect benchmark of 0.20 ppm (1-hr average) 11 indicated that the existing annual standard would provide adequate health protection against 12 short-term exposures. This conclusion was instrumental in providing the rationale for the 13 decision in the last review to retain the existing annual standard. 14 1.3 SCOPE OF THE RISK AND EXPOSURE ASSESSMENT FOR THE 15 CURRENT REVIEW 16 1.3.1 Species of Nitrogen Oxides Included in Analyses 17 The nitrogen oxides (NOX) include multiple gaseous (e.g., NO2, NO, HONO) and 18 particulate (e.g., nitrate) species. As discussed in the integrated review plan (2007a), the current 19 review of the NO2 NAAQS will focus on the gaseous species of NOX and will not consider health 20 effects directly associated with particulate species of NOX. Of the gaseous species, EPA has 21 historically determined it appropriate to specify the indicator of the standard in terms of NO2 22 because the majority of the information regarding health effects and exposures is for NO2. The 23 current draft ISA has found this to be the case and, therefore, NO2 will be used as the indicator 24 for the gaseous NOX in the risk and exposure assessment described in this document. 25 1.3.2 Scenarios Addressed in First Draft Assessment 26 The first draft of the risk and exposure assessment, described in this document, details the 27 assessment of risks and exposures associated with recent ambient levels of NO2 and with levels 28 associated with just meeting the current standard. The second draft of this document will also 29 describe the assessment of risks and exposures associated with just meeting potential alternative April 2008 Draft 6 ------- 1 standards. Completion of the second draft of the risk and exposure assessment will follow the 2 completion of the final ISA, thereby allowing the choice of potential alternative standards to be 3 informed by the information in the final ISA. April 2008 Draft ------- i 2. SOURCES, AMBIENT LEVELS, AND EXPOSURES 2 2.1 SOURCES OF NO2 3 Ambient levels of NC>2 are the product of both direct NC>2 emissions and emissions of 4 other NOX (e.g, NO) which can then be converted to NC>2 (for a more detailed discussion see the 5 draft ISA, section 2.2). Nationally, anthropogenic sources account for approximately 87% of 6 total NOX emissions. Mobile sources (both on-road and off-road) account for about 60% of total 7 anthropogenic emissions of NOX, while stationary sources (e.g., electrical utilities and industry) 8 account for the remainder (annex Table 2.6-1). Highway vehicles represent the major mobile source 9 component. In the United States, approximately half the mobile source emissions are contributed by 10 diesel engines and half are emitted by gasoline-fueled vehicles and other sources (annex section 11 2.6.2 and Table 2.6-1). Apart from these anthropogenic sources, there are also natural sources of 12 NOX including microbial activity in soils, lightning, and wildfires (draft ISA, section 2.2.1 and 13 annex section 2.6.2). 14 2.2 AMBIENT LEVELS OF NO2 15 According to monitoring data, nationwide levels of ambient NC>2 (annual average) 16 decreased 41% between 1980 and 2006 (draft ISA, Figure 2.4-4). Between 2003 and 2005, 17 national mean concentrations of NC>2 were about 15 ppb for averaging periods ranging from a 18 day to a year. The average daily maximum hourly NC>2 concentrations were approximately 30 19 ppb. These values are about twice as high as the 24-h averages. The highest maximum hourly 20 concentrations (-200 ppb) between 2003 and 2005 are more than a factor often higher than the 21 mean hourly or 24-h concentrations (draft ISA, Figure 2.4-2). The highest levels of NC>2 in the 22 United States can be found in and around Los Angeles, in the Midwest, and in the Northeast. 23 Nitrogen dioxide is monitored mainly in large urban areas and, therefore, data from the 24 NC>2 monitoring network is generally more representative of urban areas than rural areas. Levels 25 in non-urban areas can be estimated with modeling. Model-based estimates indicate that NC>2 26 levels in many non-urban areas of the United States are less than 1 ppb. Levels in these areas 27 can approach policy-relevant background concentrations, which are those concentrations that 28 would occur in the United States in the absence of anthropogenic emissions in continental North 29 America (defined here as the United States, Canada, and Mexico). For NC>2, policy-relevant April 2008 Draft 8 ------- 1 background concentrations are estimated to range from 0.1 ppb to 0.3 ppb (draft ISA, section 2 2.4.6.1). 3 Ambient levels of NO2 exhibit both seasonal and diurnal variation. In southern cities, 4 such as Atlanta, higher concentrations are found during winter, consistent with the lowest mixing 5 layer heights being found during that time of the year. Lower concentrations are found during 6 summer, consistent with higher mixing layer heights and increased rates of photochemical 7 oxidation of NC>2. For cities in the Midwest and Northeast, such as Chicago and New York City, 8 higher levels tend to be found from late winter to early spring with lower levels occurring from 9 summer through the fall. In Los Angeles the highest levels tend to occur from autumn through 10 early winter and the lowest levels from spring through early summer. Mean and peak 11 concentrations in winter can be up to a factor of two larger than in the summer at sites in Los 12 Angeles. In terms of daily variability, NC>2 levels typically peak during the morning rush hours. 13 Monitor siting plays a key role in evaluating diurnal variability as monitors located further away 14 from traffic will show cycles that are less pronounced over the course of a day than monitors 15 located closer to traffic. 16 2.3 EXPOSURE TO NO2 17 Human exposure to an airborne pollutant is defined as contact between a person and the 18 pollutant at a specific concentration for a specified period of time (draft ISA, section 2.5.1). The 19 integrated exposure of a person to a given pollutant is the sum of the exposures over all time 20 intervals for all microenvironments in which the individual spends time. Microenvironments in 21 which people are exposed to air pollutants such as NC>2 typically include residential indoor 22 environments and other indoor locations, near-traffic outdoor environments and other outdoor 23 locations, and in vehicles (draft ISA, Figure 2.5-1). 24 There is a large amount of variability in the time that individuals spend in different 25 microenvironments, but on average people spend the majority of their time (about 87%) indoors. 26 Most of this time is spent at home with less time spent in an office/workplace or other indoor 27 locations (draft ISA, Figure 2.5-1). On average, people spend about 8% of their time outdoors 28 and 6% of their time in vehicles. Significant variability surrounds each of these broad estimates, 29 particularly when considering influential personal attributes such as age or gender; when 30 accounting for daily, weekly, or seasonal factors influencing personal behavior; or even when April 2008 Draft 9 ------- 1 characterizing individual variability in time spent in various locations (McCurdy and Graham, 2 2003; Graham and McCurdy, 2004). Typically, the time spent outdoors or in vehicles could vary 3 by 100% or more depending on which of these influential factors are considered. One potential 4 consequence of this is that exposure misclassification can result when total human exposure is 5 not disaggregated between relevant microenvironments and the variability in time spent in these 6 locations is not taken into account. 7 Such misclassification, which can occur in epidemiological studies that rely on ambient 8 pollutant levels as a surrogate for exposure, may obscure the true relationship between ambient 9 air pollutant exposures and health outcomes. Thus, use of ambient pollutant levels as a surrogate 10 for exposures can introduce uncertainty that should be considered when interpreting the 11 epidemiological literature. This uncertainty in exposure estimates can result from differences 12 between ambient levels and actual exposures as well as from the NO2 monitoring approach itself. 13 Results have been mixed regarding the ability of ambient levels of NO2 to act as a 14 surrogate for personal exposures to NC>2. Studies examining the association between ambient 15 NC>2 and personal exposure to NC>2 have generated mixed results due to 1) the prevalence of 16 indoor sources of NC>2; 2) the spatial heterogeneity of NC>2 in study areas; 3) the seasonal and 17 geographic variability in the infiltration of ambient NC>2; 4) differences in the time spent in 18 different microenvironments; and 5) differences in study design (draft ISA, section 2.5.6.2). As 19 a result, some researchers have concluded that ambient NC>2 may be a reasonable proxy for 20 personal exposure, while others have noted that caution must be exercised. Overall, the body of 21 evidence examined in the draft ISA demonstrates that ambient NO2 concentrations are associated 22 with personal exposures; however, the strength of that association varies considerably. 23 The current approach to monitoring ambient NO2 can also introduce uncertainty into 24 exposure estimates. For example, the method for estimating NO2 levels (i.e., subtraction of NO 25 from a measure of total NOX) is subject to interference by NOX oxidation products. Limited 26 evidence suggests that these compounds result in an overestimate of NO2 levels by roughly 20 to 27 25% at typical ambient levels. Smaller relative errors are estimated to occur in measurements 28 taken near strong NOX sources since most of the mass emitted as NO or NO2 would not yet have 29 been further oxidized. Relatively larger errors appear in locations more distant from strong local 30 NOX sources. Additionally, many NO2 monitors are elevated above ground level in the cores of 31 large cities. Because most sources of NO2 are near ground level, this produces a gradient of NO2 April 2008 Draft 10 ------- 1 with higher levels near ground level and lower levels being detected at the elevated monitor. 2 One comparison has found an average of a 2.5-fold increase in NCh concentration measured at 4 3 meters above the ground compared to 15 meters above the ground. Levels are likely even higher 4 at elevations below 4 meters (draft ISA, section 2.5.3.3). Another source of uncertainty in 5 exposure estimates can result from monitor location. NC>2 monitors are sited for compliance 6 with air quality standards rather than for capturing small-scale variability in NC>2 concentrations 7 near sources such as roadway traffic. Significant gradients in NC>2 concentrations near roadways 8 have been observed in several studies, and NC>2 concentrations have been found to be correlated 9 with distance from roadway and traffic volume (draft ISA, section 2.5.3.2). 10 April 2008 Draft 11 ------- 3. AT RISK POPULATIONS 2 3.1 OVERVIEW 3 Specific subpopulations are at increased risk for suffering NCVrelated health effects. 4 This could occur because they are affected by lower levels of NC>2 than the general population 5 (susceptibility), because they experience a larger health impact than the general population to a 6 given level of exposure (susceptibility), and/or because they are exposed to higher levels of NC>2 7 than the general population (vulnerability). In discussions of susceptibility, the draft ISA focuses 8 on disease-mediated (e.g., asthma, cardiovascular disease) and age-mediated susceptibility (i.e., 9 children and elderly) (draft ISA, sections 4.3.1 and 4.3.2). In discussions of vulnerability, the 10 draft ISA focuses on age-mediated vulnerability (i.e., children and elderly) and vulnerability in 11 individuals who spend a large amount of time on or near roadways due to the location of their 12 residence, their occupation, or the fact that they spend time commuting in traffic (draft ISA, 13 section 4.3.5). These groups are discussed in more detail below. 14 3.2 DISEASE AND ILLNESS 15 Recent evidence strengthens the conclusion, drawn in the 1993 Criteria Document, that 16 asthmatics are likely more susceptible than the general population to the effects of NC>2 17 exposure. In addition, recent evidence broadens this likely susceptible population to include 18 those with other pulmonary conditions and individuals with upper respiratory viral infections 19 (draft ISA, section 4.3.1). These conclusions are based on an array of both short- and long-term 20 studies reporting associations between NC>2 and respiratory and cardiac health effects. The most 21 extensive supporting evidence is available for asthmatics. In addition to the large number of 22 epidemiological studies that have reported associations between NC>2 exposure and health effects 23 in asthmatics, human clinical studies demonstrate that airways hyperresponsiveness in asthmatics 24 is the most sensitive clinical indicator of response to NC>2 (draft ISA, section 4.3.1). 25 3.3 AGE 26 The draft ISA identifies both children (i.e., <18 years of age) and older adults (i.e., >65 27 years of age) as groups that are potentially more susceptible than the general population to the 28 health effects associated with NC>2 exposure (draft ISA, section 4.3.2). In children, the April 2008 Draft 12 ------- 1 developing lung is highly susceptible to damage from exposure to environmental toxicants 2 (Dietert et al., 2000) likely because eighty percent of alveoli are formed postnatally and changes 3 in the lung continue through adolescence (draft ISA, section 4.3.2). The basis for the increased 4 susceptibility in the elderly is not known, but one hypothesis is that it may be related to changes 5 in antioxidant defenses in the fluid lining the respiratory tract (draft ISA, section 4.3.2). In 6 addition, the generally declining health status of many elderly individuals may increase their 7 risks for pollution-mediated effects (draft ISA, section 4.3.2). 8 3.4 PROXIMITY TO ROADWAYS 9 The draft ISA also includes discussion of vulnerable populations that experience 10 increased NC>2 exposures on or near roadways (draft ISA, section 4.3.5). Many studies find that 11 indoor, personal, and outdoor NC>2 levels are strongly associated with proximity to traffic or to 12 traffic density (draft ISA, section 2.5.4). Due to high air exchange rates, NC>2 levels inside a 13 vehicle could rapidly approach levels outside the vehicle during commuting (draft ISA, section 14 4.3.5). Mean in-vehicle NC>2 levels are between 2 and 3 times ambient levels measured at fixed 15 sites nearby (draft ISA, sections 2.5.4 and 4.3.5). Therefore, individuals with occupations that 16 require them to be in traffic or close to traffic (e.g., bus and taxi drivers, highway patrol officers, 17 toll collectors) and those who spend time commuting in traffic could be exposed to relatively 18 high levels of NC>2 compared to ambient levels. Due to the high peak exposures while driving, 19 total personal exposure could be underestimated if exposures while commuting are not 20 considered. In some cases, exposure in traffic can dominate personal exposure to NC>2 (Lee et 21 al., 2000; Son et al., 2004) (draft ISA, section 2.5.4). 22 April 2008 Draft 13 ------- 4. HEALTH EFFECTS 2 4.1 INTRODUCTION 3 The draft ISA, along with its associated annexes, provides a comprehensive review and 4 assessment of the scientific evidence related to the health effects associated with NC>2 exposures. 5 For these health effects, the draft ISA characterizes judgments about causality with a hierarchy (for 6 discussion see draft ISA, section 1.6) that contains the following five levels. 7 • Sufficient to infer a causal relationship 8 • Sufficient to infer a likely causal relationship (i.e., more likely than not) 9 • Suggestive but not sufficient to infer a causal relationship 10 • Inadequate to infer the presence or absence of a causal relationship 11 • Suggestive of no causal relationship 12 Judgments about causality are informed by a series of decisive factors that are based on those set 13 forth by Sir Austin Bradford Hill in 1965 (draft ISA, Table 1.6-1). These decisive factors 14 include strength of the observed association, availability of experimental evidence, consistency 15 of the observed association, biological plausibility, coherence of the evidence, temporal 16 relationship of the observed association, and the presence of an exposure-response relationship. 17 For purposes of the characterization of NC>2 health risks, staff have judged it appropriate to focus 18 on endpoints for which the draft ISA concludes that the available evidence is sufficient to infer 19 either a causal or a likely causal relationship. 20 4.2 ADVERSE RESPIRATORY EFFECTS FOLLOWING SHORT-TERM 21 EXPOSURES 22 4.2.1 Overview 23 The draft ISA concludes that, when taken together, recent studies provide scientific 24 evidence that NC>2 is associated with a range of respiratory effects and are sufficient to infer a 25 likely causal relationship between short-term NC>2 exposure and adverse effects on the 26 respiratory system (draft ISA, section 5.3.2.1). This finding is supported by a large body of 27 epidemiologic evidence, in combination with findings from human and animal experimental 28 studies. The epidemiologic evidence for respiratory effects can be characterized as consistent, in April 2008 Draft 14 ------- 1 that associations are reported in studies conducted in numerous locations with a variety of 2 methodological approaches. Considering this large body of epidemiologic studies alone, the 3 findings are coherent in the sense that the studies report associations with respiratory health 4 outcomes that are logically linked together. A number of these epidemiologic studies have been 5 conducted in locations where the ambient NC>2 levels are well below the level of the current 6 NAAQS. Health effects associations have been observed in epidemiologic studies reporting 7 maximum ambient concentrations as high as 100 to 300 ppb, concentrations within the range of 8 the controlled animal and human exposures used in current toxicological and clinical studies 9 reporting respiratory effects (see draft ISA, Tables 5.3-2 and 5.3-3). This evidence is discussed 10 in more detail below. 11 4.2.2 Effects Based on Controlled Human Exposure Studies 12 4.2.2.1 Overview 13 Controlled human exposure studies have addressed the consequences of short-term (e.g., 14 15-minutes to several hours) NC>2 exposures for a number of health endpoints including airways 15 responsiveness, host defense and immunity, inflammation, and lung function (draft ISA, section 16 3.1). The draft ISA concludes that in asthmatics, NC>2 may increase the allergen-induced airways 17 inflammatory response at exposures as low as 0.26-ppm for 30 min (draft ISA, Figure 3.1-2) and 18 NC>2 exposures between 0.2 and 0.3 ppm for 30 minutes can result in small but significant 19 increases in non-specific airways responsiveness (draft ISA, section 5.3.2.1). In contrast, the 20 draft ISA concludes that 1) limited evidence indicates that NC>2 may increase susceptibility to 21 injury by subsequent viral challenge at exposures as low as 0.6 ppm for 3 hours; 2) evidence 22 exists for increased airways inflammation at NC>2 concentrations less than 2.0 ppm; and 3) the 23 direct effects of NC>2 on lung function in asthmatics have been inconsistent at exposure 24 concentrations below 1 ppm (draft ISA, section 5.3.2.1). As a result, although studies on all of 25 these endpoints provide qualitative support for the ability of NC>2 to cause adverse effects on 26 respiratory health, the focus for purposes of characterizing risks associated with ambient NO2 is 27 airways responsiveness (see below). April 2008 Draft 15 ------- 1 4.2.2.2 Airways Responsiveness 2 Inhaled pollutants such as NC>2 may have direct effects on lung function, or they may 3 enhance the inherent responsiveness of the airways to a challenge with a bronchoconstricting 4 agent (draft ISA, section 3.1.3). Asthmatics are generally much more sensitive to nonspecific 5 bronchoconstricting agents (e.g., cholinergic drugs, cold air, histamine, etc.) than non-asthmatics, 6 and airways challenge testing is used as a diagnostic test in asthma. An increase in airways 7 responsiveness in asthmatics is one indicator of increased severity of disease and worsened 8 asthma control while effective treatment often reduces airways responsiveness. Aerosolized 9 allergens can also be used in controlled airways challenge testing in the laboratory. The degree 10 of responsiveness to allergens is a function of the concentration of inhaled allergen, the degree of 11 sensitization to the allergen, and the degree of nonspecific airways responsiveness. Following 12 inhalation of a non-specific bronchoconstricting agent or an allergen, asthmatics may experience 13 both an "early" response, with a decline in lung function within minutes after the challenge, and 14 a "late" response, with a decline in lung function hours after the exposure. The early response 15 primarily reflects release of histamine and other inflammatory mediators by airways mast cells 16 while the late response reflects enhanced airways inflammation and mucous production. 17 Airways responsiveness can be measured by assessing changes in pulmonary function (e.g., 18 decline in FEVi) or changes in the inflammatory response (e.g., using markers in 19 bronchoalveolar lavage (BAL) fluid or induced sputum) (draft ISA, section 3.1.3.1). 20 Folinsbee (1992) conducted a meta-analysis using individual level data from 19 clinical 21 NC>2 exposure studies measuring airways responsiveness in asthmatics (draft ISA, section 22 3.1.3.2). These studies included NC>2 exposure levels between 0.1 ppm and 1.0 ppm and most of 23 them used non-specific bronchoconstricting agents such as methacholine, carbachol, histamine, 24 or cold air. The largest effects were observed for subjects at rest. Among subjects exposed at 25 rest, 76% experienced increased airways responsiveness following exposure to NC>2 levels 26 between 0.2 and 0.3 ppm. Because this meta-analysis evaluated only the direction of the change 27 in airways responsiveness, it is not possible to discern the magnitude of the change from these 28 data. However, the results do suggest that short-term exposures to NC>2 at near-ambient levels 29 (<0.3 ppm) can alter airways responsiveness in people with mild asthma (draft ISA, section 30 3.1.3.2). April 2008 Draft 16 ------- 1 Several studies published since the last review address the question of whether low-level 2 exposures to NO2 enhance the response to specific allergen challenge in mild asthmatics (draft 3 ISA, section 3.1.3.1). These recent studies suggest that NO2 may enhance the sensitivity to 4 allergen-induced decrements in lung function, and increase the allergen-induced airways 5 inflammatory response. Strand et al. (1997) demonstrated that single 30-minute exposures to 6 0.26-ppm NCh increased the late phase response to allergen challenge 4 hours after exposure, as 7 measured by changes in lung function. In a separate study (Strand et al., 1998), 4 daily repeated 8 exposures to 0.26-ppm NO2for 30 minutes increased both the early and late-phase responses to 9 allergen, as measured by changes in lung function. Barck et al. (2002) used the same exposure 10 and challenge protocol in the earlier Strand study (0.26 ppm for 30 min, with allergen challenge 11 4-h after exposure), and performed BAL 19 hours after the allergen challenge to determine NO2 12 effects on the allergen-induced inflammatory response. Compared with air followed by allergen, 13 NO2 followed by allergen caused an increase in the BAL recovery of polymorphonuclear (PMN) 14 cells and eosinophil cationic protein (ECP) as well as a reduction in total BAL fluid volume and 15 cell viability. ECP is released by degranulating eosinophils, is toxic to respiratory epithelial 16 cells, and is thought to play a role in the pathogenesis of airways injury in asthma. Subsequently, 17 Barck et al. (2005) exposed 18 mild asthmatics to air or 0.26 ppm NO2 for 15 minutes on day 1, 18 followed by two 15 minute exposures separated by 1 hour on day 2, with allergen challenge after 19 exposures on both days 1 and 2. Sputum was induced before exposure on day 1 and after 20 exposures (morning of day 3). Compared to air plus allergen, NO2 plus allergen resulted in 21 increased levels of ECP in both sputum and blood and increased myeloperoxidase levels in 22 blood. All exposures in these studies (Barck et al., 2002, 2005; Strand et al., 1997, 1998) used 23 subjects at rest. They used an adequate number of subjects, included air control exposures, 24 randomized exposure order, and separated exposures by at least 2 weeks. Together, they indicate 25 the possibility for effects on allergen responsiveness in some asthmatics following brief 26 exposures to 0.26 ppm NO2. However, other recent studies have failed to find effects using 27 similar, but not identical, approaches (draft ISA, section 3.1.3.1). The differing findings may 28 relate in part to differences in timing of the allergen challenge, the use of multiple versus single- 29 dose allergen challenge, the use of BAL versus sputum induction, exercise versus rest during 30 exposure, and differences in subject susceptibility (draft ISA, section 3.1.3.1). Table 1 (below) April 2008 Draft 17 ------- 1 provides summary information on the key controlled human exposure studies identified in the 2 draft ISA that evaluated airways responsiveness. 3 4.2.2.3 Conclusions 4 Based on the draft ISA's evaluation of controlled human exposure studies, staff have 5 judged that the strongest basis for the characterization of NC>2 risks is airway responsiveness in 6 asthmatics. Asthmatic volunteers have been exposed to NC>2 in the absence of other pollutants 7 that often confound associations in the epidemiology literature. Therefore, these studies provide 8 evidence for a direct relationship between exposure to NC>2 and this respiratory health effect. 9 However, because many of the studies of airways responsiveness evaluate only a single level of 10 NC>2 and because of methodological differences between the studies, staff have judged that the 11 data are not sufficient to derive an exposure-response relationship in the range of interest. 12 Therefore, the most appropriate approach to characterizing risks based on the controlled human 13 exposure studies evidence for airways responsiveness is to compare estimated NC>2 air quality 14 and exposure levels with potential health effect benchmark levels. Estimates of hourly peak air 15 quality concentrations and personal exposures to ambient NC>2 concentrations at and above 16 specified potential health effect benchmark levels provides some perspective on the public health 17 impacts of health effects that we cannot currently evaluate in quantitative risk assessments. Staff 18 recognizes that there is high inter-individual variability in responsiveness such that only a subset 19 of asthmatic individuals exposed at and above a given benchmark level would actually be 20 expected to experience any such potential adverse health effects. 21 To identify these potential health effect benchmarks, staff have relied on the draft ISA's 22 evaluation of the NC>2 human exposures studies. Controlled human exposure studies involving 23 allergen challenge in asthmatics suggest that NC>2 exposure may enhance the sensitivity to 24 allergen-induced decrements in lung function and increase the allergen-induced airways 25 inflammatory response at exposures as low as 0.26-ppm NC>2for 30 min (draft ISA, Figure 3.1-2 26 and section 5.3.2.1). Exposure to NO2 also has been found to enhance the inherent 27 responsiveness of the airways to subsequent non-specific challenges (draft ISA, sections 3.1.4.2 28 and 5.3.2.1). In general, small but significant increases in non-specific airways responsiveness April 2008 Draft 18 ------- Table 1. Summary of Key Controlled Human Exposure Studies of Airways Responsiveness Study Tunnicliffe, 1994 Devalia, 1994 Strand, 1997 Strand, 1998 Barck, 2005 Barck, 2005 Barck, 2002 Bylin, 1985 Mohsenin, 1987 Strand, 1996 Torres, 1990 Rubenstein, 1990 Torres, 1991 Witten, 2005 Torres, 1991 Jenkins, 1999 Jenkins, 1999 Witten, 2005 Roger, 1990 NO2 Exposure Level (ppm) 0.4 0.4 0.26 0.26 0.26 0.26 0.26 0.3 0.5 0.26 0.25 0.3 0.25 0.4 0.25 0.4 0.2 0.4 0.15-0.6 Exposure Duration 1-hour 6-hours 30-minutes 30-minutes (4x per day) 15 -minutes (3x over 2 days) 15 -minutes (3x over 2 days) 30-minutes 20-minutes 1-hour 30-minutes 30-minutes 30-minutes 30-minutes 3 -hours 30-minutes 3 -hours 6-hours 3 -hours 75-minutes Study Population Mild asthmatics Mild asthmatics Mild asthmatics Mild to Moderate asthmatics Mild asthmatics Mild asthmatics Mild asthmatics Mild asthmatics Asthmatics Mild asthmatics Mild asthmatics Asthmatics Mild asthmatics Mild asthmatics Mild asthmatics Mild asthmatics Mild asthmatics Mild asthmatics Mild asthmatics Allergen versus non- specific Allergen Allergen Allergen Allergen Allergen Allergen Allergen Non- specific Non- specific Non- specific Non- specific Non- specific Non- specific Allergen Non- specific Allergen Allergen Allergen Non- specific Metric Used Lung function Lung function Lung function Lung function Lung function Inflammatory Markers (sputum, blood) Inflammatory Markers (BAL) Lung function Lung function Lung function Lung function Lung function Lung function Inflammatory Markers (sputum) Lung Function Lung function Lung function Lung function Lung function Number of Subjects 8 8 18 16 18 18 13 8 8 19 14 9 11 15 11 11 11 15 21 Exercise No No No No No No No No No No No Yes Yes Yes Yes Yes Yes Yes Yes Statistically Significant X X X X X X X X X X Statistically Non Significant X X X X X X X X X April 2008 Draft 19 ------- 1 have been observed in the range of 0.2 to 0.3 ppm NO2 for 30 minute exposures in asthmatics. 2 Therefore, in the risk characterization described in Chapters 5-7 of this document, staff judge 3 that 1-hour NO2 levels in this range are appropriate to consider as potential health benchmarks 4 for comparison to air quality levels and exposure estimates. To characterize health risks with 5 respect to this range, potential health effect benchmark values of 0.20 ppm (200 ppb), 0.25 ppm 6 (250 ppb), and 0.30 ppm (300 ppb) have been employed to reflect the lower- middle- and upper- 7 end of the range identified in the draft ISA as the lowest levels at which controlled human 8 exposure studies have provided sufficient evidence for the occurrence of NO2-related airway 9 responsiveness. 10 4.2.3 Epidemiology Literature 11 4.2.3.1 Hospital Admissions and Emergency Department Visits 12 Epidemiologic evidence exists for positive associations between short-term ambient NO2 13 concentrations below the current NAAQS and increased numbers of emergency department 14 visits and hospital admissions for respiratory causes, especially asthma (draft ISA, section 15 5.3.2.1). Total respiratory causes for emergency department visits and hospitalizations typically 16 include asthma, bronchitis and emphysema (collectively referred to as COPD), pneumonia, upper 17 and lower respiratory infections, and other minor categories. Temporal associations between 18 emergency department visits or hospital admissions for respiratory diseases and ambient levels 19 of NO2 have been the subject of over 50 peer-reviewed research publications since the last 20 review. These studies have examined morbidity in different age groups and have often utilized 21 multi-pollutant models to evaluate potential confounding effects of co-pollutants. 22 Of the emergency department visit and hospital admission studies reviewed in the NOx 23 draft ISA, 6 were conducted in the United States (draft ISA, Table 5.3-4). Of these 6 studies, 24 only 3 evaluated associations with NO2 using multi-pollutant models (Peel et al., 2005 and 25 Tolbert et al., 2007 in Atlanta; Ito et al., 2007 in New York City). In the study by Peel and 26 colleagues, investigators evaluated emergency department visits among all ages in Atlanta, GA 27 during the period of 1993 to 2000. Using single pollutant models, the authors reported a 2.4% 28 (95% CI: 0.9, 4.1) increase in respiratory emergency department visits associated with a 30-ppb 29 increase in 1-h max NO2 levels. For asthma visits, a 4.1% (95% CI: 0.8%, 7.6%) increase was 30 detected only in individuals 2 to 18 years of age. Tolbert and colleagues reanalyzed these data April 2008 Draft 20 ------- 1 with 4 additional years of information and found essentially similar results in single pollutant 2 models (2.0% increase, 95% CI: 0.5, 3.3). This same study found that the associations were 3 positive, but not statistically-significant, in multi-pollutant models that included PMi0 or ozone 4 (Os). In the study by Ito and colleagues, investigators evaluated emergency department visits for 5 asthma in New York City during the years 1999 to 2002. The authors found a 12 % (95% CI: 6 7%, 15%) increase in risk per 20 ppb increase in 24-hour ambient NO2. Risk estimates were 7 robust and remained statistically significant in multi-pollutant models that included PM2 5, Os, 8 CO, and SO2. 9 4.2.3.2 Respiratory Illness and Symptoms 10 Studies of Ambient NO 2 11 Epidemiologic studies using community ambient monitors have found associations 12 between ambient NO2 concentrations and respiratory symptoms (draft ISA, sections 3.1.4.2 and 13 5.3.2.1, Figure 3.1-6) in cities where NO2 concentrations were within the range of 24-hour 14 average concentrations observed in recent years. Several studies have been published since the 15 last review of the NO2 NAAQS including 3 multi-city studies in urban areas covering the 16 continental United States and southern Ontario. These are the Harvard Six Cities Study (Six 17 Cities; Schwartz et al., 1994), the National Cooperative Inner-City Asthma Study (NCICAS; 18 Mortimer et al., 2002), and the Childhood Asthma Management Program (CAMP; Schildcrout et 19 al., 2006). 20 Schwartz el at (1994) studied 1,844 schoolchildren, followed for 1 year, as part of the Six 21 Cities Study that included the cities of Watertown, MA, Baltimore, MD, Kingston-Harriman, 22 TN, Steubenville, OH, Topeka, KS, and Portage, WI. Respiratory symptoms were recorded 23 daily. The authors reported a significant association between 4-day mean NO2 levels and 24 incidence of cough among all children in single-pollutant models, with an odds ratio (OR) of 25 1.61 (95% CI: 1.08, 2.43) standardized to a 20-ppb increase in NO2. The incidence of cough 26 increased up to approximately mean NO2 levels (-13 ppb) (p = 0.01), after which no further 27 increase was observed. The significant association between cough and 4-day mean NO2 level 28 remained unchanged in models that included Os, but was attenuated and lost statistical 29 significance in two-pollutant models that included PMio (OR = 1.37 [95% CI: 0.88, 2.13]) or 30 SO2 (OR = 1.42 [95% CI: 0.90, 2.28]). April 2008 Draft 21 ------- 1 Mortimer et al. (2002) studied the risk of asthma symptoms among 864 asthmatic 2 children in the eight cities that were part of the NCICAS. The eight study locations included 3 New York City, NY, Baltimore, MD, Washington, DC, Cleveland, OH, Detroit, MI, St Louis, 4 MO, and Chicago, IL. Subjects were followed daily for four 2-week periods over the course of 5 nine months with morning and evening asthma symptoms and peak flow recorded. The greatest 6 effect was observed for morning symptoms using a 6-day moving average, with a reported OR of 7 1.48 (95% CI: 1.02, 2.16). Although effects were generally robust in multi-pollutant models that 8 included O3 (OR for 20-ppb increase in NO2 = 1.40 [95% CI: 0.93, 2.09]), O3 and SO2 (OR for 9 NO2 =1.31 [95% CI: 0.87, 2.09]), or O3, SO2, and PMio (OR for NO2 = 1.45 [95% CI: 0.63, 10 3.34]), they were not statistically-significant. 11 Schildcrout et al. (2006) investigated the association between ambient NO2 and 12 respiratory symptoms and rescue inhaler use as part of the CAMP study. The study reported on 13 990 asthmatic children living within 50 miles of an NO2 monitor in Boston, MA, Baltimore, MD, 14 Toronto, ON, St. Louis, MO, Denver, CO, Albuquerque, NM, or San Diego, CA. Symptoms and 15 use of rescue medication were recorded daily, resulting in each subject having an approximate 16 average of two months of data. The authors reported the strongest association between NO2 and 17 increased risk of cough for a 2-day lag, with an OR of 1.09 (95% CI: 1.03, 1.15) for each 20-ppb 18 increase in NO2 occurring 2 days before measurement. Multi-pollutant models that included CO, 19 PMio, or SO2 produced similar results (see Figure 3.1-5, panel A of the draft ISA). Additionally, 20 increased NO2 exposure was associated with increased use of rescue medication, with the 21 strongest association for a 2-day lag, both for single- and multi-pollutant models (e.g., for an 22 increase of 20-ppb NO2 in the single-pollutant model, the RR for increased inhaler usage was 23 1.05 (95% CI: 1.01, 1.09). 24 Studies of Indoor NO2 25 Evidence supporting increased respiratory morbidity following NO2 exposures is also 26 found in studies of indoor NO2 (draft ISA, section 3.1.4.1). For example, in a randomized 27 intervention study in Australia (Pilotto et al., 2004), students attending schools that switched out 28 unvented gas heaters, a major source of indoor NO2, experienced a decrease in both levels of 29 NO2 and in respiratory symptoms (e.g., difficulty breathing, chest tightness, and asthma attacks) 30 compared to students in schools that did not switch out unvented gas heaters (levels were 47.0 April 2008 Draft 22 ------- 1 ppb in control schools and 15.5 ppb in intervention schools) (draft ISA, section 2.7). An earlier 2 indoor study by Pilotto and colleagues (1997) found that students in classrooms with higher 3 levels of NO2 also had higher rates of respiratory symptoms (e.g., sore throat, cold) and 4 absenteeism than students in classrooms with lower levels of NC>2. This study detected a 5 significant concentration-response relationship, strengthening the argument that NC>2 is causally 6 related to respiratory morbidity. A number of other indoor studies conducted in homes have also 7 detected significant associations between indoor NC>2 and respiratory symptoms (draft ISA, 8 section 3.1.4.1). 9 4.2.3.3 Conclusions Regarding the Epidemiology Literature 10 As mentioned above (see section 1.1), the NC>2 epidemiological literature will be 11 considered during the policy assessment and rulemaking stage of the NAAQS review process as 12 part of an evidence-based approach to assessing the adequacy of potential alternative standards. 13 This use of the epidemiological literature will be reflected in Agency rulemaking documents 14 (i.e., ANPR, proposed rulemaking, and final rulemaking). However, the appropriateness of the 15 epidemiological literature for use as the basis of a quantitative risk assessment is a separate issue 16 and is discussed below. 17 The preferred approach for conducting a risk assessment based on concentration-response 18 relationships from the epidemiological literature would be to rely on studies of ambient NC>2 19 conducted in multiple locations throughout the United States that employ both single-pollutant 20 and multi-pollutant models. This approach would provide a range of concentration-response 21 functions that are relevant to specific cities in the United States. However, the relatively small 22 number of NC>2 epidemiological studies conducted in the United States and the difficulty in 23 separating direct effects of NC>2 from those associated with a traffic-related pollutant mixture that 24 includes NC>2 (draft ISA, section 5.4) would increase the quantitative uncertainty associated with 25 a risk assessment based on the epidemiological literature. These factors make it particularly 26 difficult to quantify with confidence the unique contribution of NO2 to respiratory health effects. 27 Therefore, staff judge it unlikely that a quantitative risk assessment based on the available NC>2 28 epidemiological literature would meaningfully inform a decision to retain or revise the standard. 29 This judgment, along with consideration of the resource requirements associated with conducting 30 such an assessment, have led staff to conclude that it is not appropriate to conduct a quantitative April 2008 Draft 23 ------- 1 assessment of NC>2 risks based on the epidemiological literature to support this review of the 2 NO2 NAAQS. 3 4.2.4 Toxicology Literature 4 Although the animal toxicology literature is not used as a quantitative basis for evaluating 5 NC>2 risks in this assessment, toxicology studies are important for their ability to provide 6 mechanistic insights into health effects that have been observed in humans and because they can 7 support the plausibility of associations observed in the epidemiological literature. For example, 8 animal studies provide evidence that NC>2 can impair the respiratory host defense system 9 sufficiently to render animals more susceptible to respiratory infections. Mortality rates 10 following infection with a respiratory virus have been evaluated in the presence and absence of 11 NC>2. Susceptibility to bacterial and viral pulmonary infections, as measured by this approach, 12 increases with NO2 exposures as low as 0.5 ppm (draft ISA, sections.1.1 and5.3.2.1). In 13 addition, increased airways responsiveness has been detected in animals exposed to NC>2 levels 14 between 1 and 4 ppm (draft ISA, section 5.3.2.1 and Table 5.3-3). Six-week exposures to 4.0 15 ppm NC>2 or longer exposures (e.g., 12 week) to lower levels (e.g., 1 ppm) of NC>2 have caused 16 airways hyperresponsiveness to histamine in guinea pigs (draft ISA, section 3.1.3.2). 17 Toxicologic studies have also detected indications of increased inflammation following NO2 18 exposures < 1.0 ppm in vitamin C-deficient guinea pigs (draft ISA, section 3.1.2). Thus, the 19 toxicology literature provides qualitative support for the NC>2 findings reported in humans. 20 4.3 OTHER ADVERSE EFFECTS FOLLOWING SHORT-TERM 21 EXPOSURES 22 The epidemiologic evidence is suggestive but not sufficient to infer a casual relationship 23 between short-term exposure to NC>2 and nonaccidental and cardiopulmonary-related mortality. 24 Results from several large U.S. and European multi-city studies and a meta-analysis study 25 indicated positive associations between ambient NC>2 concentrations and the risk of all-cause 26 (nonaccidental) mortality, with effect estimates ranging from 0.5 to 3.6% excess risk in mortality 27 per standardized increment (draft ISA, section 3.3.1, Figure 3.3-2, section 5.3.2.3). In general, 28 the NC>2 effect estimates were robust to adjustment for co-pollutants. Both cardiovascular and 29 respiratory mortality have been associated with increased NO2 concentrations in epidemiologic April 2008 Draft 24 ------- 1 studies (draft ISA, Figure 3.3-3); however, similar associations were observed for other 2 pollutants, including PM and SCh. The range of risk estimates for mortality excess was 3 generally smaller than that for other pollutants such as PM. In addition, while NO2 exposure, 4 alone or in conjunction with other pollutants, may contribute to increased mortality, evaluation 5 of the specificity of this effect is difficult. Clinical studies showing hematologic effects and 6 animal toxicological studies showing biochemical, lung host defense, permeability, and 7 inflammation changes with short-term exposures to NO2 provide limited evidence of plausible 8 pathways by which risks of morbidity and, potentially, mortality may be increased, but no 9 coherent picture is evident at this time (draft ISA, section 5.3.2.3). 10 The available evidence on the effects of short-term exposure to NO2 on cardiovascular 11 health effects is inadequate to infer the presence or absence of a causal relationship at this time. 12 Evidence from epidemiologic studies of heart rate variability, repolarization changes, and cardiac 13 rhythm disorders among heart patients with ischemic cardiac disease are inconsistent (draft ISA, 14 sections 3.2.1 and 5.3.2.2). In most studies, associations with PM were found to be similar or 15 stronger than associations with NO2. Generally positive associations between ambient NO2 16 concentrations and hospital admissions or emergency department visits for cardiovascular 17 disease have been reported in single-pollutant models (draft ISA, section 3.2.2); however, most 18 of these effect estimate values were diminished in multi-pollutant models that also contained CO 19 and PM indices (draft ISA, section 5.3.2.2). Mechanistic evidence of a role for NO2 in the 20 development of cardiovascular diseases from studies of biomarkers of inflammation, cell 21 adhesion, coagulation, and thrombosis is lacking (draft ISA, sections 3.2.1.4 and 5.3.2.2). 22 Furthermore, the effects of NO2 on various hematological parameters in animals are inconsistent 23 and, thus, provide little biological plausibility for effects of NO2 on the cardiovascular system. 24 4.4 ADVERSE EFFECTS FOLLOWING LONG-TERM EXPOSURES 25 The epidemiologic and toxicological evidence examining the effect of long-term 26 exposure to NO2 on respiratory morbidity is suggestive but not sufficient to infer a casual 27 relationship at this time. A number of epidemiologic studies examined the effects of long-term 28 exposure to NO2 and reported positive associations with decrements in lung function and 29 partially irreversible decrements in lung function growth (draft ISA, section 3.4.1, Figures 3.4-1 30 and 3.4-2, section 5.3.2.4). However, similar associations have also been found for PM, Cb, and April 2008 Draft 25 ------- 1 proximity to traffic (<500 m) and the high correlation among traffic-related pollutants made it 2 difficult to accurately estimate the independent effects in these long-term exposure studies. 3 Results from the available epidemiologic evidence investigating the association between long- 4 term exposure to NC>2 and increases in asthma prevalence and incidence are suggestive but not 5 always consistent (draft ISA, sections 3.4.2 and 5.3.2.4). Epidemiologic studies conducted in 6 both the United States and Europe also have produced inconsistent results regarding an 7 association between long-term exposure to NC>2 and respiratory symptoms (draft ISA, sections 8 3.4.3 and 5.3.2.4). While some positive associations were noted, a large number of symptom 9 outcomes were examined and the results across specific outcomes were inconsistent. Animal 10 toxicological studies demonstrated that NC>2 exposure resulted in morphological changes in the 11 centriacinar region of the lung and in bronchi olar epithelial proliferation (draft ISA, section 12 3.4.4), which may provide some biological plausibility for the observed epidemiologic 13 associations between long-term exposure to NC>2 and respiratory morbidity. Susceptibility to 14 these morphological effects was found to be influenced by many factors, such as age, 15 compromised lung function, and acute infections. 16 The available epidemiologic and toxicological evidence is inadequate to infer the 17 presence or absence of a causal relationship for carcinogenic, cardiovascular, and reproductive 18 and developmental effects related to long-term NCh exposure. Epidemiologic studies conducted 19 in Europe have shown an association between long-term NC>2 exposure and increased incidence 20 of cancer (draft ISA, section 5.3.2.5). However, the animal toxicological studies have provided 21 no clear evidence that NC>2 acts as a carcinogen (draft ISA, sections 3.5.1 and 5.3.2.5). The very 22 limited epidemiologic and toxicological evidence does not suggest that long-term exposure to 23 NO2has cardiovascular effects (draft ISA, sections 3.5.2 and 5.3.2.5). The epidemiologic 24 evidence is not consistent for associations between NC>2 exposure and growth retardation; 25 however, some evidence is accumulating for effects on preterm delivery (draft ISA, sections 26 3.5.3 and 5.3.2.5). Scant animal evidence supports a weak association between NC>2 exposure 27 and adverse birth outcomes and provides little mechanistic information or biological plausibility 28 for the epidemiologic findings. 29 The epidemiologic evidence is inadequate to infer the presence or absence of a causal 30 relationship between long-term exposure to NC>2 and mortality (draft ISA, section 5.3.2.6). In the 31 United States and European cohort studies examining the relationship between long-term April 2008 Draft 26 ------- 1 exposure to NC>2 and mortality, results were generally inconsistent (draft ISA, section 3.6, Figure 2 3.6-2, and section 5.3.2.6). Further, when associations were suggested, they were not specific to 3 NCh, but also implicated PM and other traffic indicators. The relatively high correlations 4 reported between NC>2 and PM indices make it difficult to interpret these observed associations at 5 this time (draft ISA, section 5.3.2.6). 6 April 2008 Draft 27 ------- i 5. OVERVIEW OF RISK AND EXPOSURE ASSESSMENT 2 5.1 INTRODUCTION 3 Human exposure, regardless of the pollutant, depends on where an individual is located 4 and what they are doing at a given moment of time. The magnitude of the exposure can depend 5 on a variety of factors, such as personal attributes (e.g., age or gender), emission sources (e.g., 6 automobile exhaust, indoor gas stoves), and physical-chemical properties of the pollutant (e.g., 7 atmospheric chemistry). The risk of an adverse health effect following exposure to a pollutant is 8 also dependent on a number of factors, such as the individual's personal attributes (age, gender, 9 preexisting health conditions) and the toxic properties of the pollutant (e.g., as indicated by dose- 10 or concentration-response relationships). An important feature of a combined exposure 11 assessment and health risk characterization is to maintain their expected degree of correlation, 12 considering common influential factors and the variability that occurs in personal behavior and 13 exposure concentrations across time and space. 14 One method to assess exposure to air pollutants is through analysis of air quality 15 concentrations. Ambient monitoring can serve as an indicator of potential exposures that a 16 population residing in an area might have. Depending on the spatial density of the monitoring 17 network and the frequency of sample collection, the measured concentrations can provide a 18 useful record of ambient concentrations that vary over time and across a geographic area. 19 Ambient NC>2 concentrations have been linked with adverse health responses and thus are 20 considered a reasonable surrogate for exposure. However, the actual exposures that individuals 21 experience might be influenced by other sources not measured by the ambient monitor (e.g., 22 indoor sources). In addition, while temporal variability can be well represented with continuous 23 ambient monitoring (e.g., hourly measures throughout the year), the spatial and temporal 24 variability in human activities is not considered, further adding to exposure error. 25 Another method for determining people's exposure to a substance is through personal 26 measurements of the pollutant(s). Personal exposures can provide a reasonable estimate of an 27 individual's total exposure since it accounts for different concentrations an individual encounters 28 over time, including high concentrations that may result from outdoor and indoor source 29 emissions. As described in section 2.5 of the NOX ISA however, the availability of personal April 2008 Draft 28 ------- 1 exposure measurements for NC>2 is limited to only a few studies performed in U.S., each 2 containing a limited number of study subjects. The measurement of personal NC>2 exposure is 3 further restricted by the sampling device detection capabilities, resulting in measurement periods 4 of days to weeks when measured. This time-averaging of personal exposure concentrations may 5 provide some information most relevant to health effects associated with long-term exposures to 6 NC>2, but is less informative for evaluating health effects that result from hourly or daily (or even 7 multiple peak) exposures. 8 Inhalation exposure models are useful in realistically estimating personal exposures, 9 particularly those exposure models that can simulate human activity patterns over variable 10 periods of time. The value of these advanced models is further supported by recognizing that 11 exposure measurements cannot be performed for a large population and/or cannot be used to 12 evaluate alternative exposure scenarios such as simulating just meeting the current or alternative 13 standards. Exposure models are capable of performing any number of simulations (e.g., an entire 14 population, selected individuals) and in any location (e.g., urban area, CMSA, census block), the 15 scope of which depends on the availability of relevant input data. Inhalation exposure models 16 are typically driven by estimates of ambient outdoor concentrations of the pollutants, since the 17 contribution of ambient conditions to total exposure is of primary interest. These outdoor 18 concentrations, which can vary by time of day as well as by location, may be provided by 19 measurements, by air quality models, or by a combination of these. In addition, exposure models 20 can estimate concentrations associated with indoor source emissions to provide perspective on 21 the relative contribution such sources have on total exposure. Thus, the complexity of modeling 22 exposure and the usefulness of the results generated is driven by the temporal and spatial 23 variability in ambient and other concentrations persons may be exposed to, the ability to capture 24 variability (both inter- and intra-personal) in human activities, and whether the most important 25 sources contributing to total exposure are represented. 26 Each of these elements of exposure and risk have been considered in the development of 27 the approach for conducting these assessments in the draft document entitled Nitrogen Dioxide 28 Health Assessment Plan: Scope and Methods for Exposure and Risk Assessment (EPA, 2007b). 29 That draft document was reviewed by CASAC and the public at a public meeting on October 24- 30 25, 2007. Comments received at that meeting informed the approach adopted by staff for 31 conducting the risk and exposure assessment presented herein. April 2008 Draft 29 ------- 1 This draft assessment summarizes the results of a risk characterization and exposure 2 assessment associated with recent ambient levels of NO2 and with ambient levels of NO2 3 simulated to just meeting the current NO2 standard of 0.053 ppm annual average. The second 4 draft assessment and the final assessment also will evaluate exposures and health risks associated 5 with any potential alternative standards that are identified for consideration (also see section 6 1.3.2 of this document). Additional details are available in the Exposure and Risk technical 7 support document (draft TSD) (EPA, 2008b) that supports this assessment. 8 5.2 GOALS 9 The goals of this draft NO2 risk and exposure assessment, for both recent ambient air 10 quality conditions and for where ambient concentrations just meet the current standard, are to 1) 11 estimate short-term exposures and potential human health risks associated with ambient NO2; 2) 12 evaluate the quantitative relationship between long-term average NO2 air quality and short-term 13 levels of NO2 that exceed health effect benchmark levels; 3) determine factors contributing to 14 persons estimated to be most frequently exposed to concentration at or above selected 1-hour 15 concentrations; and 4) identify important assumptions and uncertainties associated with the 16 estimates of exposure and the risk characterization. 17 18 19 5.3 GENERAL APPROACH Exposures were assessed in a two-step process. In the first step, scenario-driven air quality analyses were performed using ambient NO2 concentrations for years 1995 through 2006. 20 This air quality data, as well as other NO2 concentrations derived from ambient levels, were used 21 as a surrogate to estimate potential human exposure. All U.S. monitoring sites where NO2 data 22 have been collected are represented by this analysis and, as such, the results generated are 23 considered a broad characterization of national air quality and human exposures that might be 24 associated with these concentrations. 25 In the second step, detailed modeling of population exposures was conducted. For this 26 exposure analysis, a probabilistic approach was used to model individual exposures considering 27 the time people spend in different microenvironments and variable NO2 concentrations that occur 28 within these microenvironments across time, space, and microenvironment type. This approach 29 to assessing exposures was more resource intensive than using ambient levels as a surrogate for April 2008 Draft 30 ------- 1 exposure, therefore staff included only a few specific locations in the U.S. for potential inclusion 2 in this part of the assessment. Although the geographic scope of this analysis was restricted, the 3 approach used provides realistic estimates of NO2 exposures, particularly those exposures 4 associated with important emission sources of NOX and NC>2, and serves to complement to the 5 broad air quality characterization. 6 For the characterization of risks in both the air quality analysis and the exposure 7 modeling, staff used the range of health short-term potential health effect benchmark values 8 based on the draft ISA (i.e., 1-hr NC>2 levels ranging from 200 to 300 ppb). To assess potential 9 health risks, benchmark values of 200, 250, and 300 ppb were selected and compared to both 10 NC>2 air quality levels and estimates of NC>2 exposure. When NC>2 air quality was used as a 11 surrogate for exposure, the output of the analysis were estimates of the number of times per year 12 specific locations experience 1-hr levels of NC>2 that have been shown to potentially cause 13 adverse health effects in susceptible individuals. When personal exposures were simulated, the 14 output of the analysis were estimates of the number of individuals at risk for experiencing daily 15 maximum 1 -hr levels of exposure to NC>2 of ambient origin that have been shown to potentially 16 cause adverse health effects in susceptible individuals. The rationale and details for each of the 17 approaches used and the range of potential health effect benchmarks identified is described 18 below in Chapter 6 (Air Quality Characterization and Associated Health Risk) and Chapter 7 19 (Exposure Assessment and Associated Health Risk). 20 5.4 ADDITIONAL CONSIDERATIONS 21 A primary goal of this draft of the risk and exposure assessments is to evaluate the ability 22 of the current NC>2 standard of 0.053 ppm annual average to protect public health. All areas of 23 the United States have annual average levels below the current standard. Therefore, in order to 24 evaluate the ability of the current standard to protect public health, NO2 concentrations need to 25 be adjusted such that they simulate levels of NC>2 that just meet the current annual standard. 26 Two different adjustment procedures, although mathematically equivalent, were used for 27 the two different approaches to estimate NC>2 exposures. For the air quality characterization, a 28 proportional roll-up of air quality concentrations was performed. The exposure modeling used a 29 proportional roll-down of the potential health effect benchmark levels. Each of these is briefly 30 described below. April 2008 Draft 31 ------- 1 These procedures were necessary to provide insights into the degree of exposure and risk 2 which would be associated with an increase in ambient NO2 levels such that the levels were just 3 at or near the current standard in the urban areas analyzed. Staff recognizes that it is extremely 4 unlikely that NO2 concentrations in these urban areas would rise to meet the current NAAQS and 5 that there is considerable uncertainty with the simulation of conditions that just meet the current 6 annual standard. 7 5.4.1 Adjustment of Ambient Air Quality 8 Based on the form of the standard and observed trends in ambient monitoring, ambient 9 NO2 concentrations were proportionally rolled-up at each location using the maximum annual 10 average concentration that occurred in each year. While annual average concentrations have 11 declined significantly over the time period of analysis, the variability in the concentrations, both 12 the annual average and 1-hour concentrations, have remained relatively constant (see section 2.5 13 in the draft TSD). Therefore, proportional adjustment factors F for each location (/') and year (/) 14 were derived by the following: 15 16 ^=53/Cmax,y eq(l) 17 18 where, 19 20 Fy = Adjustment factor (unitless) 21 CmaXjij = Maximum annual average NO2 concentration at a monitor in a location / and 22 yeary (ppb) 23 24 In these cases where staff simulated a proportional roll-up in ambient NO2 concentrations 25 using eq (1), it is assumed that the current temporal and spatial distribution of air concentrations 26 (as characterized by the current air quality data) is maintained and increased NOX emissions 27 contribute to increased NO2 concentrations, with the highest monitor (in terms of annual 28 averages) being adjusted so that it just meets the current 0.053 ppm annual average standard. 29 Values for each air quality adjustment factor used for each location evaluated in the air quality 30 and risk characterization are given in the draft TSD (section 2.5). For each location and calendar 31 year, all the hourly concentrations in a location were multiplied by the same constant value F to 32 make the highest annual mean equal to 53 ppb for that location and year. For example, of 33 several monitors measuring NO2 in Boston for year 1995, the maximum annual mean April 2008 Draft 32 ------- 1 concentration was 30.5 ppb, giving an adjustment factor of F = 53/30.5 = 1.74 for that year. All 2 hourly concentrations measured at all monitoring sites in that location would then be multiplied 3 by 1.74, resulting in an upward scaling of hourly NO2 concentrations for that year. Therefore, 4 one monitoring site in Boston for year 1995 would have an annual average concentration of 5 0.053 ppm, while all other monitoring sites would have an annual average concentration below 6 that value, although still proportionally scaled up by 1.74. Then, using the adjusted hourly 7 concentrations to simulate just meeting the current standard, the metrics of interest (e.g., annual 8 mean NO2 concentration, the number of potential health effect benchmark exceedances) were 9 estimated for each site-year. 10 5.4.2 Adjustment of Potential Health Effect Benchmark Levels 11 Rather than proportionally modify the air quality concentrations used for input to the 12 exposure model, a proportional roll-down of the potential health effect benchmark level was 13 performed. This was done to reduce the processing time associated with the exposure modeling 14 simulations since there were tens of thousands of receptors modeled in each location. In 15 addition, because the adjustment is proportional, the application of a roll-down of the selected 16 benchmark level is mathematically equivalent to a proportional roll-up of the air quality 17 concentrations. The same approach used in the air quality adjustment described above was used 18 in the exposure modeling to scale the benchmark levels downward to simulate just meeting the 19 current standard. For example, an adjustment factor of 1.59 was determined for Philadelphia for 20 year 2001, based on a maximum predicted annual average NO2 concentration of 33 ppb for a 21 modeled receptor placed at an ambient monitoring location. Therefore, the 1-hour potential 22 health effect benchmark levels of 200, 250, and 300 ppb were proportionally rolled-down to 126, 23 157, and 189 ppb, respectively for year 2001. This procedure was applied for each year within 24 each location where an exposure modeling was performed to simulate just meeting the current 25 standard. April 2008 Draft 33 ------- i 6. AMBIENT AIR QUALITY AND HEALTH RISK 2 CHARACTERIZATION 3 6.1 OVERVIEW 4 Ambient monitoring data for each of the years 1995 through 2006 were used in this 5 analysis to characterize NO2 air quality across the U.S. This air quality data, as well as other 6 NO2 concentrations derived from ambient levels, were used as a surrogate to estimate potential 7 human exposure. Because the current standard is based on annual average levels of NO2 while 8 the most definitive health effects evidence is associated with short-term (i.e., 30-minute to 1- 9 hour, or one to several day) exposures, the air quality analysis required the development of a 10 model that relates annual average and short-term levels of NO2. To characterize this relationship 11 and to estimate the number of exceedances of the potential health effect benchmarks in specific 12 locations, several possible models were explored (i.e., exponential regression, logistic regression, 13 a regression assuming a Poisson distribution, and an empirical model). An empirical model, 14 employing the annual average and hourly concentrations, was chosen to avoid some of the 15 difficulties in extrapolating outside the range of the data. A detailed discussion justifying the 16 selection of this approach is provided in Appendix D of the draft TSD. 17 A total of four air quality scenarios were evaluated using the empirical model for each of 18 two distinct ambient monitoring periods, resulting in a total of eight separate analyses. The 19 available NO2 air quality were divided into two groups; one contained data from years 1995- 20 2000, representing an historical data set; the other contained the monitoring years 2001-2006, 21 representing recent ambient monitoring. Each of these monitoring year-groups were evaluated 22 considering the NO2 concentrations as they were reported and representing the conditions at that 23 time (termed in this assessment "as is"). This served as the first air quality scenario. The second 24 scenario considered the ambient NO2 concentrations simulated to just meeting the current 25 standard of 0.053 ppm annual average. The 3rd and 4th scenarios followed in similar fashion, 26 however these scenarios used the ambient monitoring data to estimate NO2 concentrations that 27 might occur on roadways to generate on-road concentrations for as is air quality and for ambient 28 concentrations just meeting the current standard. Again, each of these four scenarios was 29 evaluated using both the historical and recent data air quality data sets. April 2008 Draft 34 ------- 1 Since all of the NC>2 ambient monitoring sites are represented by this analysis, the 2 generated results are considered a broad characterization of national air quality and human 3 exposures that might be associated with these concentrations. The output of this air quality 4 characterization was used to estimate the number of times per year specific locations experience 5 levels of NC>2 that could cause adverse health effects in susceptible individuals. Each location 6 that was evaluated contained one to several monitors operating for a few to several years, 7 generating a number of site-years of data. The number of site-years in a location were used to 8 generate a distribution of two exposure and risk characterization metrics; the annual average 9 concentrations and the numbers of exceedances that did (observed data) or could occur 10 (simulated data) in a year for that location. The mean and median values were reported to 11 represent the central tendency of each metric for the four scenarios in each air quality year- 12 group, while the minimum value served to represent the lower bound. Since there were either 13 multiple site-years or numerous simulations performed at each location using all available site- 14 years of data, results for the upper percentiles included the 95th, 98th and 99th percentiles of the 15 distribution. 16 6.2 APPROACH 17 There were three broad steps to allow for the characterization of the air quality. The first 18 step involved collecting, compiling, and screening the ambient air quality data collected since the 19 prior review in 1995. A screening of the data followed to ensure consistency with the NC>2 20 NAAQS requirements. Then, criteria based on the current standard and the potential health 21 effect benchmark levels were used to identify specific locations for analysis using descriptive 22 statistical analysis of the screened data set. All other monitoring data not identified by the 23 selected criteria were grouped into one of two non-specific categories. These locations (both the 24 specific and non-specific) served as the geographic centers of the analysis, where application of 25 the empirical model was done to estimate concentrations and exceedances of potential health 26 effect benchmark levels. In addition to use of the ambient concentrations (as is), and ambient 27 concentrations just meeting the current standard, on-road concentrations were estimated in this 28 air quality characterization to approximate the potential exposure and risk metrics associated 29 with these concentrations. April 2008 Draft 35 ------- 1 6.2.1 Air Quality Data Screen 2 NC>2 air quality data and associated documentation from the years 1995 through 2006 3 were downloaded from EPA's Air Quality System (AQS) for this purpose (EPA, 2007c, d). A 4 site was defined by the state, county, site code, and parameter occurrence code (POC), which 5 gives a 10-digit monitor ID code. As required by the NO2 NAAQS, a valid year of monitoring 6 data is needed to calculate the annual average concentration. A valid year at a monitoring site 7 was comprised of 75% of valid days in a year, with at least 18 hourly measurements for a valid 8 day (thus at least 274 or 275 valid days depending on presence of a leap year and a minimum of 9 4,932 or 4,950 hours). This served as the screening criterion for data used in the analysis. 10 Site-years of data are the total numbers of years the collective monitors in a location were 11 in operation. Of a total of 5,243 site-years of data in the entire NO2 1-hour concentration 12 database, 1,039 site-years did not meet the above criterion and were excluded from any further 13 analyses. In addition, since shorter term average concentrations are of interest, the remaining 14 site-years of data were further screened for 75% completeness on hourly measures in a year (i.e., 15 containing a minimum of 6,570 or 6,588, depending on presence of a leap year). Twenty-seven 16 additional site-years were excluded, resulting in 4,177 complete site-years in the analytical 17 database. Table 2 provides a summary of the site-years included in the analysis, relative to those 18 excluded, by location and by two site-year groups.2 The air quality data from AQS were 19 separated into these two groups, one representing historic data (1995-2000) and the other 20 representing more recent data (2001-2006) to represent temporal variability in NC>2 21 concentrations within each location. The selection of locations was a companion analysis to the 22 screening, however, it is discussed in a separate section. 214 of 18 named locations and the 2 grouped locations contained enough data to be considered valid for year 2006. April 2008 Draft 36 ------- 2 3 4 Table 2. Counts of complete site-years of NO2 monitoring data. Location Boston Chicago Cleveland Denver Detroit Los Angeles Miami New York Philadelphia Washington Atlanta Colorado Springs El Paso Jacksonville Las Vegas Phoenix Provo St. Louis Other CMSA Not MSA Total Com 1995-2000 58 47 11 26 12 193 24 93 46 69 24 26 14 6 16 22 6 56 1135 200 Number of Dlete 2001-2006 47 36 11 10 12 177 20 81 39 66 29 0 30 4 35 27 6 43 1177 243 4177 Site-Years Incorr 1995-2000 16 20 2 10 4 16 1 12 6 21 5 4 11 0 4 8 0 3 249 112 iplete 2001-2006 34 22 2 4 1 19 4 24 8 18 1 4 0 2 9 25 0 9 235 141 1066 Site-^ % Cor 1995-2000 78% 70% 85% 72% 75% 92% 96% 89% 88% 77% 83% 87% 56% 100% 80% 73% 100% 95% 82% 64% fears iplete 2001-2006 58% 62% 85% 71% 92% 90% 83% 77% 83% 79% 97% 0% 100% 67% 80% 52% 100% 83% 83% 63% 80% 5 6 7 8 9 10 11 12 13 14 15 16 6.2.2 Selection of Locations for Air Quality Analysis Criteria were established for selecting sites with high annual means and/or frequent exceedances of potential health effect benchmarks. Selected locations were those that had a maximum annual mean NC>2 level at a particular monitor greater than or equal to 25.7 ppb, which represents the 90th percentile across all locations and site-years, and/or had at least one reported 1-hour NC>2 level greater than or equal to 200 ppb, the lowest level of the potential health effect benchmarks. A location in this context would include a geographic area that encompasses more than a single air quality monitor (e.g., particular city, metropolitan statistical area (MSA), or consolidated metropolitan statistical area or CMSA). First, all monitors were identified as either belonging to a CMSA, a MSA, or neither. Then, locations of interest were identified through statistical analysis of the ambient NC>2 air quality data for each site within a location. April 2008 Draft 37 ------- 1 Fourteen locations met both selection criteria and an additional four met at least one of 2 the criteria (see Table 3).3 In addition to these 18 specific locations, the remaining sites were 3 grouped into two broad location groupings. The Other CMSA location contains all the other sites 4 that are in MS As or CMS As but are not in any of the 18 specified locations. The Not MSA 5 location contains all the sites that are not in an MSA or CMSA. The final database for analysis 6 included air quality data from a total of 205 monitors within the named locations, 331 monitors 7 in the Other CMSA group, and 92 monitors in the Not MSA group. 8 6.2.3 Estimation of On-Road Concentrations using Ambient Concentrations 9 Since mobile sources can account for a large part of personal exposures to ambient NO2 10 in some individuals, the potential impact of roadway levels of NO2 was evaluated. A strong 11 relationship has been reported between NO2 levels measured on roadways and NO2 measured at 12 increasing distance from the road. This relationship has been described previously (e.g., Cape et 13 al., 2004) using an exponential decay equation of the form: 14 15 Cx=Cb + Cve-* eq(2) 16 where, 17 18 Cx = NO2 concentration at a given distance (x) from a roadway (ppb) 19 Cb = NO2 concentration (ppb) at a distance from a roadway, not directly influenced 20 by road or non-road source emissions. 21 Cv = NO2 concentration contribution from vehicles on a roadway (ppb) 22 k = Rate constant describing NO2 combined formation/decay with perpendicular 23 distance from roadway (meters'1) 24 x = Distance from roadway (meters) 3 New Haven, CT, while meeting both criteria, did not have any recent exceedances of 200 ppb and contained one of the lowest maximum concentration-to-mean ratios, therefore was not separated out as a specific location for analysis. April 2008 Draft 38 ------- 1 2 Table 3. Locations selected for Tier INO2 Air Quality Characterization, associated abbreviations, and values of selection criteria. Location Type1 Code Description Abbreviation CMSA* CMSA CMSA* CMSA* CMSA* CMSA* CMSA CMSA* CMSA* CMSA* MSA* MSA* MSA* MSA MSA* MSA* MSA MSA* MSA/CMSA - 1122 1602 1692 2082 2162 4472 4992 5602 6162 8872 0520 1720 2320 3600 4120 6200 6520 7040 - - Boston-Worcester-Lawrence, MA-NH-ME-CT Chicago-Gary-Kenosha, IL-IN- Wl Cleveland-Akron, OH Denver-Boulder-Greeley, CO Detroit-Ann Arbor-Flint, Ml Los Angeles-Riverside-Orange County, CA Miami-Fort Lauderdale, FL New York-Northern New Jersey-Long Island, NY-NJ-CT- PA Philadelphia-Wilmington- Atlantic City, PA-NJ-DE-MD Washington-Baltimore, DC-MD- VA-WV Atlanta, GA Colorado Springs, CO El Paso,TX Jacksonville, FL Las Vegas,NV-AZ Phoenix-Mesa,AZ Provo-Orem,UT St, Louis, MO-IL Other MSA/CMSA Other Not MSA Boston Chicago Cleveland Denver Detroit Los Angeles Miami New York Philadelphia Washington DC Atlanta Colorado Springs El Paso Jacksonville Las Vegas Phoenix Provo St. Louis Other CMSA Not MSA Maximum # of Exceedances of 200 ppb 1 0 1 2 12 5 3 3 3 2 1 69 2 2 11 37 0 8 10 2 Maximum Annual Mean (ppb) 31.1 33.6 28.1 36.8 25.9 50.6 16.8 42.2 34.0 27.2 26.6 34.8 35.1 15.9 27.1 40.5 28.9 27.2 31.9 19.7 1 CMSA is consolidated metropolitan statistical area; MSA is metropolitan statistical area according to the 1 999 Office of Management and Budget definitions (January 28, 2002 revision). * Indicates locations that satisfied both the annual average and exceedance criteria. 5 6 9 10 11 Much of the decline in NC>2 concentrations with distance from the road has been shown to occur within the first few meters (approximately 90% within 10 meter distance), returning to near ambient levels between 200 to 500 meters (Rodes and Holland, 1981; Bell and Ashenden, 1997; Gilbert et al., 2003; Pleijel et al., 2004). Theoretically, NC>2 concentrations can increase at a distance from the road due to chemical interaction of NOX with Os, the magnitude of which can be driven by certain meteorological conditions (e.g., wind direction). However, this relationship April 2008 Draft 39 ------- 1 developed from NC>2 measurement studies was used to estimate NCh concentrations that occur 2 on the roadway and not used to estimate NC>2 concentrations at a distance from the road. At a 3 distance of 0 meters, referred to here as on-mad., the equation reduces to the sum of the non- 4 source influenced NC>2 concentration and the concentration contribution expected from vehicle 5 emissions on the roadway using 6 7 Cr=Ca(\+m) eq(2) 8 where, 9 10 Cr = 1-hour on-roadNO2 concentration (ppb) 11 Ca = 1 -hour ambient monitoring NC>2 concentration (ppb) either as is or modified 12 to just meet the current standard 13 m = Modification factor derived from estimates of Cv/Cb (from eq (1)) 14 15 and assuming that Ca = Ct,.4 16 17 To estimate on-roadway NC>2 levels as a function of the level recorded at ambient 18 monitors and the distance of those monitors from a roadway, empirical data from the scientific 19 literature have been used. A literature review was conducted to identify published studies 20 containing NO2 concentrations on roadways and at varying distances from roadways (see section 21 2.6.1 of the draft TSD for more detail). Ratios identified from this literature review were used to 22 estimate m empirically (draft TSD, section 2.6.2). To estimate NC>2 levels on roadways, each 23 monitor site was randomly assigned one on-road factor (m) for summer months and one for non- 24 summer months from the derived empirical distribution. On-road factors were assigned 25 randomly because we expect the empirical relationship between Cv and Cb to vary from place to 26 place and we do not have sufficient information to match specific ratios with specific locations. 27 Hourly NC>2 levels were estimated for each site-year of data in a location using eq (2) and the 28 randomly assigned on-road factors. The process was simulated 100 times for each site-year of 29 hourly data. For example, the Boston CMSA location had 210 random selections from the on- 30 road distributions applied independently to the total site-years of data (105). Following 100 4 Note that Ca differs from Cb since Ca may include the influence of on-road as well as non-road sources. However, it is expected that for most monitors the influence of on-road emissions is minimal so that Ca = Cfr April 2008 Draft 40 ------- 1 simulations, a total of 10,500 site-years of data were generated using this procedure (along with 2 21,000 randomly assigned on-road values selected from the appropriate empirical distribution). 3 Simulated on-road NO2 concentrations were then used to generate concentration 4 distributions for the annual average concentrations and distributions for the number of 5 exceedances of short-term health potential health effect benchmark levels. Mean and median 6 values are reported to represent the central tendency of each parameter estimate. Since there 7 were multiple site-years and numerous simulations performed at each location using all valid 8 site-years of data, results for the upper percentiles were expanded to the 95th, 98th and 99th 9 percentiles of the distribution. In using the Boston CMSA data as an example, 4700 site years of 10 on-road concentration hourly data were simulated, and both the annual average concentration 11 and numbers of exceedances of potential health effect benchmark levels were calculated. The 12 95th, 98th and 99th percentiles were the 4465th, the 4606th, and the 4653th highest values, 13 respectively, of the 4700 calculated and ranked values. Roadways with high vehicle densities are 14 likely better represented by on-road concentration estimates at the upper tails of the distribution. 15 6.3 AIR QUALITY AND HEALTH RISK CHARACTERIZATION 16 RESULTS 17 6.3.1 Ambient Air Quality (As Is) 18 As described earlier, the air quality data obtained from AQS were separated into two 19 groups, one representing historic data (1995-2000) and the other representing more recent data 20 (2001-2006). A summary of the descriptive statistics for ambient NO2 concentrations at each 21 selected location is provided in Table 4. Detailed descriptive statistics regarding concentration 22 distributions for particular locations and specific monitoring years are provided in the draft TSD 23 (section 2.4 and Appendices B and C). None of the locations contained an exceedance of the 24 current standard of 0.053 ppm. The highest observed annual average concentrations were 25 measured in Los Angeles, New York, and Phoenix during the historic monitoring period, 26 however as with most of the locations, recent concentrations are lower across all percentiles of 27 the distribution. 28 The estimated number of exceedances of the three potential health effect benchmark 29 levels (200, 250, and 300 ppb NO2 for 1-hr) is shown in Tables 5 and 6. The exceedances of 30 each benchmark were totaled for the year at each monitor; a monitor value of 10 could represent April 2008 Draft 41 ------- 1 ten 1-hr exceedances that occurred in one day, 10 exceedances in 10 days, or some combination 2 of multiple hours or days that totaled 10 exceedances for the year. In general, the number of 3 benchmark exceedances was low across all locations. The average number of exceedances of the 4 lowest potential health effect benchmark level across each location was typically none or one. 5 Considering that there are 8760 hours in a year, this amounts to small fraction of the year 6 (0.01%) containing an exceedance. For locations predicted to have a larger number of yearly 7 average exceedances, estimates were primarily driven by a single site-year of data. For example, 8 the Colorado Springs mean estimate is 3 exceedances per year for the years 1995-2000; however, 9 this mean was driven by a single site-year that contained 69 exceedances of 200 ppb. That 10 particular monitor (ID 0804160181) does not appear to have any unusual attributes (e.g., the 11 closest major road is beyond a distance of 160 meters and the closest stationary source emitting > 12 5 tons per year (tpy) is over 4 km away) except that a power generating utility (NAICS code 13 221112) located 7.2 km from the monitor has estimated emissions of 4205 tpy. It is not known 14 at this time whether this particular facility is influencing the observed concentration exceedances 15 at this specific monitoring site. Similarly, in Phoenix a single year from one monitor (ID 16 0401330031) was responsible for all observed exceedances of 200 ppb. This monitor is located 17 78 m from the roadway and 9 of 10 stationary sources located within 10 km of this monitor 18 emitted less than 60 tpy (one emitted 272 tpy). It is not known if observed exceedances of 200 19 ppb at this monitor are a result of proximity of major roads or stationary sources. Detroit 20 contained the largest number of excedances of 200 ppb (a maximum of 12) for air quality data 21 from years 2001-2006 (Table 6). Again, all of those exceedances occurred at one monitor (ID 22 2616300192) during one year (2002). The number of exceedances of higher potential 23 benchmark concentration levels was less than for 200 ppb. Most locations had no exceedances 24 of 250 or 300 ppb, with higher numbers confined to the same aforementioned cities where 25 exceedances of 200 ppb was observed. April 2008 Draft 42 ------- 1 2 Table 4. Monitoring site-years and annual average NO2 concentrations for two monitoring periods, historic and recent air quality data (as is). Location Boston Chicago Cleveland Denver Detroit Los Angeles Miami New York Philadelphia Washington Atlanta Colorado Springs2 El Paso Jacksonville Las Vegas Phoenix Provo St. Louis Other CMSA Not MSA 1995-2000 Site- Years 58 47 11 26 12 193 24 93 46 69 24 26 14 6 16 22 6 56 1135 200 Annual Mean (ppb) 1 mean min med p95 p98 p99 18 24 23 16 19 26 10 27 23 20 14 16 23 15 14 30 24 18 14 8 5 9 17 6 12 4 6 11 15 9 5 7 14 14 3 24 23 5 1 0 19 24 23 9 19 26 9 27 21 22 15 17 23 15 8 30 24 19 14 7 31 32 28 35 26 45 17 41 33 26 25 24 35 16 27 36 24 26 24 16 31 34 28 35 26 46 17 42 34 27 27 35 35 16 27 40 24 26 26 19 31 34 28 35 26 46 17 42 34 27 27 35 35 16 27 40 24 27 28 19 2001-2006 Site- Years 47 36 11 10 12 177 20 81 39 66 29 0 30 4 35 27 6 43 1177 243 Annual Mean (ppb) 1 mean min med p95 p98 p99 15 24 19 26 19 22 9 23 20 18 12 - 15 14 11 25 24 15 12 7 5 16 14 18 14 4 6 10 14 7 3 - 8 13 1 11 21 8 1 1 13 23 19 27 19 22 8 24 19 19 14 - 16 14 9 24 23 15 12 6 25 32 24 37 23 36 15 36 29 25 19 - 21 15 22 35 29 22 20 14 30 32 24 37 23 37 16 40 30 26 23 - 22 15 23 37 29 25 22 16 30 32 24 37 23 40 16 40 30 26 23 - 22 15 23 37 29 25 24 16 1 The mean is the sum of the annual means for each monitor in a particular location divided by the number of site-years across the monitoring period. The min, med, p95, p98, p99 represent the minimum, median, 95* , 98th, and 99th percentiles of the distribution for the annual mean. 2 Colorado Springs monitoring data were collected as part of short-term study completed in September 2001 , therefore there are no 2001-2006 data. 4 5 6 7 April 2008 Draft 43 ------- 1 2 Table 5. Number of exceedances of short-term (1-hour) potential health effect benchmark levels in a year, 1995-2000 historic NOi air quality (as is). Location Boston Chicago Cleveland Denver Detroit Los Angeles Miami New York Philadelphia Washington Atlanta Colorado Springs El Paso Jacksonville Las Vegas Phoenix Provo St. Louis Other CMSA Not MSA Exceedances of 200 ppb 1-hour 1 mean min med p95 p98 p99 0 0 0 0 0 0 0 0 0 0 0 3 0 0 1 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 3 1 0 0 0 0 0 3 2 0 11 0 0 0 0 0 0 0 1 2 3 2 1 1 3 1 1 69 2 0 11 37 0 0 0 0 1 0 1 2 3 4 1 3 3 2 1 69 2 0 11 37 0 8 0 1 Exceedances of 250 ppb 1-hour 1 mean min med p95 p98 p99 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 1 1 1 0 0 0 0 1 0 23 0 0 3 3 0 0 0 0 0 0 1 1 1 2 0 0 0 1 0 23 0 0 3 3 0 4 0 0 Exceedances of 300 ppb 1 -hour 1 mean min med p95 p98 p99 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 4 0 0 3 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 4 0 0 3 0 0 0 0 0 1 The mean number of exceedances represents the number of exceedances occurring at all monitors in a particular location divided by the number of site- years across the monitoring period. The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for the number of exceedances in any one year within the monitoring period. 4 5 April 2008 Draft 44 ------- 1 2 Table 6. Number of exceedances of short-term (1-hour) potential health effect benchmark levels in a year, 2001-2006 recent NOi air quality (as is). Location Boston Chicago Cleveland Denver Detroit Los Angeles Miami New York Philadelphia Washington Atlanta El Paso Jacksonville Las Vegas Phoenix Provo St. Louis Other CMSA Not MSA Exceedances of 200 ppb 1-hour 1 mean min med p95 p98 p99 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 12 0 2 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 12 0 3 0 1 0 0 0 2 0 0 0 0 0 0 0 0 0 0 12 1 3 0 1 0 0 0 2 0 0 0 0 0 1 Exceedances of 250 ppb 1-hour 1 mean min med p95 p98 p99 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 8 0 2 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 8 0 3 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 8 1 3 0 1 0 0 0 1 0 0 0 0 0 1 Exceedances of 300 ppb 1 -hour 1 mean min med p95 p98 p99 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 0 3 0 0 0 0 0 0 0 0 0 0 0 0 1 The mean number of exceedances represents the number of exceedances occurring at all monitors in a particular location divided by the number of site- years across the monitoring period. The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for the number of exceedances in any one year within the monitoring period. 4 5 April 2008 Draft 45 ------- 1 6.3.2 Ambient Air Quality Adjusted to Just Meet the Current Standard 2 Table 7 presents descriptive statistics for the ambient NC>2 levels in each location after 3 applying an air quality adjustment procedure that rolls-up NC>2 concentrations to simulate just 4 meeting the current annual standard. Note that the 99th percentile annual average level for all 5 locations is 53 ppb for these simulations, except for the other CMSA location. This is a direct 6 consequence of the air quality adjustment procedure that sets the highest monitor in a location to 7 the current standard with the other monitors adjusted proportionally, and the number of site-years 8 available for each location. Mean and median values are similar when comparing the historic 9 annual average concentrations with the more recent estimates following the air quality 10 adjustment procedure with one exception, Denver. This is probably because the mean of 11 Denver's annual average ambient concentrations (as is) was also higher for the more recent air 12 quality analysis period (26 ppb) versus the historic data set (16 ppb, see Table 4). This suggests 13 that the air quality adjustment procedure affected the two sets of data comparably. 14 As expected, the number of estimated potential benchmark exceedances is greater when 15 air quality is modified to just meet the current standard than air quality levels as is (compare 16 Tables 8 and 9 to Tables 5 and 6). The cities with the largest estimated number of potential 17 benchmark exceedances are the same as those predicted to have largest number of exceedances 18 in the "as-is" scenario (i.e., Colorado Springs, Detroit, Phoenix). The rationale explaining these 19 results is also the same. That is, the results are due to the influence of a single monitor within 20 their respective monitoring network. Miami and Jacksonville are also predicted to have a 21 relatively large number of exceedances. This result is most likely due to the small network size 22 in these locations (n=l for Jacksonville, n=5 for Miami). Having few ambient monitors in a 23 given location could bias the mean estimate in either direction, but most likely biases estimates 24 high in these locations because of the unusually large number of peak concentrations in one or 25 more years (see draft TSD section 2.4 and Appendices B and C). In addition, Miami contained 26 some of the lowest annual average concentrations which results in high air quality simulation 27 factors across all years of data. That factor, coupled with a high coefficient of variance (COV) 28 (-13 0%) for hourly concentrations at two of the monitors in Miami (IDs 1201180021, 29 1208600271) clearly played a significant role in the higher estimated number of exceedances 30 (see draft TSD section 2.4 and Appendices B and C). Denver also contained a high COV April 2008 Draft 46 ------- 1 (-110%) for hourly concentrations using the earlier air quality period (1995-2000), likely 2 associated with the higher estimate of exceedances at this location (99th percentile of 141) 3 following the air quality adjustment procedure compared with only 2 observed exceedances 4 when considering the "as-is" air quality. Both the mean and maximum estimate of exceedances 5 for Provo (ID 4904900021) during 2001-2006 were also likely influenced by the small network 6 size (n=l) in this location and one particular year (2006) that contained numerous concentrations 7 above 150 ppb prior to the concentration roll-up. 8 6.3.3 On-Road Concentrations Derived From Ambient Air Quality (As Is) 9 Descriptive statistics for estimated on-road NO2 levels are presented in Table 10. These 10 estimated on-road levels were generated using the simulation procedure described above (section 11 5.2.3). The simulated on-road annual average concentrations are, on average, a factor of 1.8 12 higher than their respective ambient levels. This falls within the range of ratios reported in the 13 draft ISA (about 2-fold higher concentrations on roads) (draft ISA, section 2.5.4). Los Angeles, 14 New York, Phoenix, and Denver (recent data only for this location) are predicted to have the 15 highest on-road NO2 levels. This is a direct result of these locations already containing the 16 highest "as-is " levels prior to the on-road simulation. 17 The median of the simulated concentration estimates for Los Angeles were compared 18 with NO2 measurements provided by Westerdahl et al. (2005) for arterial roads and freeways in 19 the same general location during spring 2003. Although the averaging time is not the same, 20 comparison of the medians is judged to be appropriate.5 The estimated median on-road level for 21 2001-2006 is 41 ppb which falls within the range of 31 ppb to 55 ppb identified by Westerdahl et 22 al. (2005). 23 On average, most locations are predicted to have fewer than 10 exceedances per year for 24 the 200 ppb potential health effect benchmark while the median frequency of exceedances in 25 most locations is estimated to be 1 or less per year (Tables 11 and 12). There are generally fewer 5Table 10 considers annual average of hourly measurements while Westerdahl et al. (2005) reported between 2 to 4 hour average concentrations. Over time, the mean of 2-4 hour averages will be similar to the mean of hourly concentrations, with the main difference being in the variability (and hence the various percentiles of the distribution outside the central tendency). April 2008 Draft 47 ------- 1 Table 7. Estimated annual average NO2 concentrations for two monitoring periods, historic and recent air quality data 2 adjusted to just meet the current standard (0.053 ppm annual average). Location Boston Chicago Cleveland Denver Detroit Los Angeles Miami New York Philadelphia Washington Atlanta Colorado Springs2 El Paso Jacksonville Las Vegas Phoenix Provo St. Louis Other CMSA Not MSA 1995-2000 Site- Years 58 47 11 26 12 193 24 93 46 69 24 26 14 6 16 22 6 56 1135 200 Annual Mean (ppb) 1 mean min med p95 p98 p99 32 39 47 29 45 31 34 35 39 42 32 38 43 53 29 45 53 37 26 22 10 15 37 10 26 4 19 14 25 20 11 14 30 53 7 36 53 11 1 1 33 40 53 29 51 32 31 35 35 45 31 45 40 53 17 44 53 39 26 20 53 53 53 53 53 52 53 53 53 53 53 53 53 53 53 53 53 53 43 51 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 48 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 53 50 53 2001-2006 Site- Years 47 36 11 10 12 177 20 81 39 66 29 0 30 4 35 27 6 43 1177 243 Annual Mean (ppb) 1 mean min med p95 p98 p99 32 41 48 47 49 33 35 35 41 40 34 - 42 53 28 40 53 38 25 22 11 27 41 33 42 5 19 15 26 19 9 - 24 53 4 19 53 19 1 3 28 39 53 53 50 33 32 35 40 44 40 - 43 53 21 40 53 38 26 20 53 53 53 53 53 53 53 53 53 53 53 - 53 53 53 53 53 53 43 46 53 53 53 53 53 53 53 53 53 53 53 - 53 53 53 53 53 53 48 53 53 53 53 53 53 53 53 53 53 53 53 - 53 53 53 53 53 53 51 53 1 The mean is the sum of the annual means for each monitor in a particular location divided by the number of site-years across the monitoring period. The min, med, p95, p98, p99 represent the minimum, median, 95* , 98th, and 99th percentiles of the distribution for the annual mean. 2 Colorado Springs monitoring data were collected as part of short-term study completed in September 2001 , therefore there are no 2001-2006 data. April 2008 Draft 48 ------- 1 2 3 4 Table 8. Estimated number of exceedances of short-term (1-hour) potential health effect benchmark levels in a year, 1995- 2000 NOi air quality adjusted to just meet the current standard (0.053 ppm annual average). Location Boston Chicago Cleveland Denver Detroit Los Angeles Miami New York Philadelphia Washington Atlanta Colorado Springs El Paso Jacksonville Las Vegas Phoenix Provo St. Louis Other CMSA Not MSA Exceedances of 200 ppb 1-hour1 mean min med p95 p98 p99 0 0 3 8 13 1 10 0 0 1 4 30 4 12 3 12 1 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 13 0 8 0 0 0 0 0 1 15 0 0 0 0 0 0 1 0 24 19 25 5 27 1 1 4 19 180 14 20 28 57 5 1 1 18 1 1 24 141 25 8 34 2 12 9 21 241 14 20 28 198 5 1 3 53 2 1 24 141 25 9 34 3 12 17 21 241 14 20 28 198 5 15 6 87 Exceedances of 250 ppb 1-hour1 mean min med p95 p98 p99 0 0 1 2 4 0 2 0 0 0 0 15 1 2 1 4 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 10 5 15 0 6 0 0 1 2 123 6 7 13 4 0 0 0 4 1 0 10 28 15 2 15 1 9 3 3 135 6 7 13 92 0 0 1 15 1 0 10 28 15 2 15 3 9 3 3 135 6 7 13 92 0 14 1 42 Exceedances of 300 ppb 1-hour1 mean min med p95 p98 p99 0 0 0 1 2 0 1 0 0 0 0 8 0 0 1 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 4 10 0 2 0 0 1 1 72 2 1 11 0 0 0 0 1 0 0 3 9 10 0 8 0 5 2 1 83 2 1 11 31 0 0 0 8 1 0 3 9 10 2 8 1 5 2 1 83 2 1 11 31 0 13 1 21 1 The mean number of exceedances represents the number of exceedances occurring at all monitors in a particular location divided by the number of site- years across the monitoring period. The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for the number of exceedances in any one year within the monitoring period. 5 6 April 2008 Draft 49 ------- 2 Table 9. Estimated number of exceedances of short-term (1-hour) health effect benchmark levels in a year, 2001-2006 3 air quality adjusted to just meet the current standard (0.053 ppm annual average). Location Boston Chicago Cleveland Denver Detroit Los Angeles Miami New York Philadelphia Washington Atlanta El Paso Jacksonville Las Vegas Phoenix Provo St. Louis Other CMSA Not MSA Exceedances of 200 ppb 1-hour1 mean min med p95 p98 p99 0 1 1 2 8 0 17 0 1 0 8 7 31 1 0 88 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 7 0 0 0 0 0 0 0 0 1 1 1 0 11 0 0 0 0 6 22 0 0 0 0 0 0 1 2 4 7 45 1 66 1 2 2 48 24 72 3 0 526 2 1 17 5 15 4 7 45 5 69 2 25 5 56 27 72 12 1 526 5 3 44 5 15 4 7 45 6 69 5 25 6 56 27 72 12 1 526 5 5 57 Exceedances of 250 ppb 1-hour1 mean min med p95 p98 p99 0 0 0 0 4 0 3 0 0 0 1 1 15 0 0 34 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 7 0 0 0 0 0 0 0 0 1 2 34 0 18 0 1 1 9 3 46 0 0 205 1 0 4 1 0 1 2 34 0 23 1 7 1 10 6 46 2 0 205 1 1 14 1 0 1 2 34 1 23 1 7 2 10 6 46 2 0 205 1 2 20 Exceedances of 300 ppb 1-hour1 mean min med p95 p98 p99 0 0 0 0 3 0 1 0 0 0 0 0 7 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 1 28 0 11 0 0 0 2 0 25 0 0 1 0 0 2 0 0 1 1 28 0 19 0 1 1 5 1 25 0 0 1 1 0 8 0 0 1 1 28 1 19 0 1 1 5 1 25 0 0 1 1 1 9 1 The mean number of exceedances represents the number of exceedances occurring at all monitors in a particular location divided by the number of site- years across the monitoring period. The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for the number of exceedances in any one year within the monitoring period. April 2008 Draft 50 ------- 1 Table 10. Estimated annual average on-road concentrations for two monitoring periods, historic and recent ambient air 2 quality (as is). Location Boston Chicago Cleveland Denver Detroit Los Angeles Miami New York Philadelphia Washington Atlanta Colorado Springs El Paso Jacksonville Las Vegas Phoenix Provo St. Louis Other CMSA Not MSA 1995-2000 Site- Years 5800 4700 1100 2600 1200 19300 2400 9300 4600 6900 2400 2600 1400 600 1600 2200 600 5600 113500 20000 Annual Mean NO2 (ppb)1 mean min med p95 p98 p99 33 44 42 29 35 48 19 50 43 37 26 30 42 28 26 54 43 33 26 14 7 11 22 8 15 5 7 14 19 12 6 9 17 18 4 30 29 7 1 0 33 44 41 19 34 47 17 49 40 38 25 30 40 27 16 52 42 33 25 12 59 68 61 67 52 87 33 81 68 56 49 51 67 37 56 76 58 51 47 31 67 75 65 78 57 97 38 91 76 61 57 64 75 39 62 83 62 58 53 35 71 79 67 81 59 104 39 96 80 64 60 73 82 41 63 88 64 61 57 39 2001-2006 Site- Years 4700 3600 1100 1000 1200 17700 2000 8100 3900 6600 2900 - 3000 400 3500 2700 600 4300 117700 24300 Annual Mean NO2 (ppb)1 mean min med p95 p98 p99 27 43 36 48 34 41 17 43 37 33 21 - 27 25 20 45 43 27 21 12 7 20 18 23 18 5 7 12 18 9 4 - 10 17 2 14 26 10 1 1 25 42 35 46 34 40 15 41 34 33 23 - 27 25 15 43 41 27 21 11 51 66 51 74 47 71 30 70 57 52 40 - 42 34 45 70 61 44 39 27 57 72 54 83 52 80 33 79 63 57 43 - 45 36 50 79 69 49 45 31 60 76 58 87 54 85 36 85 68 61 47 - 48 37 53 84 70 52 48 33 1 The mean is the sum of the annual means for each monitor in a particular location divided by the number of site-years across the monitoring period. The min, med, p95, p98, p99 represent the minimum, median, 95* , 98th, and 99th percentiles of the distribution for the annual mean. April 2008 Draft 51 ------- 1 Table 11. Estimated number of exceedances of short-term (1-hour) potential health effect benchmark levels in a year on- 2 roads, 1995-2000 historic NO2 air quality (as is). Location Boston Chicago Cleveland Denver Detroit Los Angeles Miami New York Philadelphia Washington Atlanta Colorado Springs El Paso Jacksonville Las Vegas Phoenix Provo St. Louis Other CMSA Not MSA Exceedances of 200 ppb 1-hour1 mean min med p95 p98 p99 3 12 10 7 10 45 0 20 5 4 4 20 7 0 6 36 2 2 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 4 0 1 0 0 0 0 2 0 0 3 0 0 0 0 14 79 74 41 48 236 4 109 31 23 31 170 33 1 37 256 9 14 6 2 37 142 108 94 72 417 6 230 60 43 57 264 58 2 66 319 33 25 18 7 54 183 129 102 86 550 8 384 84 58 87 320 76 4 97 390 34 35 32 14 Exceedances of 250 ppb 1-hour1 mean min med p95 p98 p99 1 2 2 2 4 13 0 5 1 0 1 11 2 0 1 14 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 15 12 9 21 71 1 28 4 3 3 106 9 0 11 107 0 1 1 1 10 31 30 17 34 146 4 65 11 7 11 181 19 1 15 200 1 8 3 2 15 53 49 33 35 211 6 129 15 11 21 216 30 1 19 280 4 12 6 4 Exceedances of 300 ppb 1-hour1 mean min med p95 p98 p99 0 0 1 1 2 4 0 1 0 0 0 6 1 0 1 7 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 1 4 14 21 0 5 1 1 1 47 5 0 6 26 0 0 0 0 1 6 10 6 21 48 3 14 4 2 1 119 7 0 11 103 0 4 1 1 3 10 17 7 26 78 4 31 7 2 2 159 11 0 11 181 0 10 2 2 1 The mean number of exceedances represents the number of exceedances occurring at all monitors in a particular location divided by the number of site- years across the monitoring period. The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for the number of exceedances in any one year within the monitoring period. April 2008 Draft 52 ------- 2 Table 12. Estimated number of exceedances of short-term (1-hour) potential health effect benchmark levels in a year on- 3 roads, 2001-2006 historic NO2 air quality (as is). Location Boston Chicago Cleveland Denver Detroit Los Angeles Miami New York Philadelphia Washington Atlanta El Paso Jacksonville Las Vegas Phoenix Provo St. Louis Other CMSA Not MSA Exceedances of 200 ppb 1-hour1 mean min med p95 p98 p99 1 10 3 8 5 11 0 9 1 1 1 1 3 1 3 70 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 2 50 21 39 29 70 3 48 6 6 8 6 15 6 21 547 2 1 1 8 142 36 69 44 131 7 90 14 14 16 9 23 15 44 662 7 5 4 17 188 42 82 45 183 13 143 29 21 25 15 24 23 61 662 14 10 8 Exceedances of 250 ppb 1-hour1 mean min med p95 p98 p99 0 2 1 2 2 2 0 2 0 0 0 0 2 0 0 33 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 11 4 8 16 13 2 8 1 0 1 1 8 0 2 234 0 0 0 1 29 7 15 22 29 5 19 1 1 3 1 15 1 5 606 1 1 2 4 44 9 20 28 48 5 25 2 2 6 2 15 3 7 612 2 1 3 Exceedances of 300 ppb 1-hour1 mean min med p95 p98 p99 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 13 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 13 2 2 1 0 0 0 0 5 0 0 3 0 0 0 0 6 3 7 14 7 4 3 1 0 1 0 8 0 0 423 0 0 1 0 8 3 7 21 13 5 6 1 0 2 0 8 0 0 435 1 0 2 1 The mean number of exceedances represents the number of exceedances occurring at all monitors in a particular location divided by the number of site- years across the monitoring period. The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for the number of exceedances in any one year within the monitoring period. April 2008 Draft 53 ------- 1 predicted exceedances of the potential health effect benchmark levels when considering recent 2 air quality compared with the historic air quality. Areas with a relatively high number of 3 exceedances (e.g., Provo) are likely influenced by the presence of a small number of monitors 4 and one or a few exceptional site-years where levels reached the upper percentiles. 5 The number of predicted benchmark exceedances across large urban areas may be used to 6 broadly represent particular locations within those types of areas. For example, Chicago, New 7 York, and Los Angeles are large CMSAs, have several monitoring sites, and have a large number 8 of roadways. Each of these locations was estimated to have, on average, about 10 exceedances 9 of 200 ppb per year on-roads. Assuming that the on-road exceedances distribution is 10 proportionally representing the distribution of roadways within each location, about one-half of 11 the roads in these areas would not have any on-road concentrations in excess of 200 ppb. This is 12 because the median value for exceedances of 200 ppb in most locations was estimated as zero. 13 However, Tables 11 and 12 indicate that there is also a possibility of tens to just over a hundred 14 exceedances in a year as an upper bound estimate on certain roads/sites. 15 6.3.4 On-Road Concentrations Derived From Ambient Air Quality Adjusted to Just 16 Meet the Current Standard 17 Table 13 presents descriptive statistics for estimated on-road NO2 concentrations 18 assuming each location just meets the current 0.053 ppm annual standard. These on-road 19 concentrations were generated using the simulation procedure described above (see section 20 5.2.1.3) applied to air quality data that has been modified to simulate each location just meeting 21 the annual standard. On average, these simulated on-road annual average concentrations are 22 about 1.8 times higher than the accompanying ambient concentrations (Table 7). Tables 14 and 23 15 present estimates for the number of exceedances of the three selected potential health effect 24 benchmark levels (i.e., 200, 250, and 300 ppb NO2 1-hr). 25 The mean number of estimated exceedances of 200 ppb ranges from tens to several 26 hundreds (Tables 14 and 15), sharply increased from the previous on-road estimates using the air 27 quality (as is). Some of the highest exceedance estimates occurred in the locations described 28 previously as being influenced by a few concentrations at the upper percentiles of their 29 distributions in a small number of years and/or monitoring sites (e.g., Miami, Colorado Springs, 30 Provo). Compared to the means, median estimated exceedances of 200 ppb are lower, on April 2008 Draft 54 ------- 1 Table 13. Estimated annual average on-road concentrations for two monitoring periods, air quality data adjusted to just meet 2 the current standard (0.053 ppm annual average). Location Boston Chicago Cleveland Denver Detroit Los Angeles Miami New York Philadelphia Washington Atlanta Colorado Springs El Paso Jacksonville Las Vegas Phoenix Provo St. Louis Other CMSA Not MSA 1995-2000 Site- Years 5800 4700 1100 2600 1200 19300 2400 9300 4600 6900 2400 2600 1400 600 1600 2200 600 5600 113500 20000 Annual Mean (ppb)1 mean min med p95 p98 p99 58 72 86 53 81 56 62 64 71 77 57 69 77 96 53 82 96 68 46 39 13 18 47 12 33 6 24 18 31 26 14 18 38 67 8 46 67 14 1 1 57 72 84 49 83 55 56 62 67 77 55 73 74 95 34 78 95 68 46 35 103 112 123 112 124 102 111 104 111 116 111 118 122 128 113 115 129 106 84 90 117 123 128 124 129 114 124 117 123 124 126 127 129 131 125 127 139 118 95 104 125 130 136 129 133 122 128 123 128 130 129 131 138 144 130 129 144 124 103 115 2001-2006 Site- Years 4700 3600 1100 1000 1200 17700 2000 8100 3900 6600 2900 - 3000 400 3500 2700 600 4300 117700 24300 Annual Mean (ppb)1 mean min med p95 p98 p99 58 74 88 85 90 61 63 63 74 73 61 - 75 96 50 72 95 69 46 39 14 35 53 42 54 7 25 18 33 23 12 - 30 67 5 24 67 25 1 3 53 72 86 85 87 60 57 61 71 74 66 - 74 94 36 71 93 67 45 35 105 113 123 124 123 105 112 103 111 114 111 - 112 129 112 110 128 106 84 89 120 124 130 130 129 116 126 119 125 124 126 - 124 139 124 125 131 118 95 101 126 130 146 141 134 123 129 125 128 128 129 - 128 145 129 127 138 126 102 109 1 The mean is the sum of the annual means for each monitor in a particular location divided by the number of site-years across the monitoring period. The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for the annual mean. April 2008 Draft 55 ------- 1 Table 14. Estimated number of exceedances of short-term (1-hour) potential health effect benchmark levels in a year on- 2 roads, 1995-2000 historic NOi air quality adjusted to just meet the current standard (0.053 ppm annual average). Location Boston Chicago Cleveland Denver Detroit Los Angeles Miami New York Philadelphia Washington Atlanta Colorado Springs El Paso Jacksonville Las Vegas Phoenix Provo St. Louis Other CMSA Not MSA Exceedances of 200 ppb 1-hour1 mean min med p95 p98 p99 78 172 321 214 405 100 363 77 114 219 251 304 178 610 238 250 443 148 52 95 0 0 1 0 2 0 1 0 0 0 0 0 0 40 0 0 1 0 0 0 13 61 195 23 284 18 260 11 27 101 42 77 82 549 26 105 230 48 6 7 411 727 1045 1261 1227 489 1045 412 570 852 1094 1320 692 1426 1107 953 1643 620 268 549 677 1001 1221 1921 1439 791 1334 693 797 1070 1472 1756 951 1515 1674 1326 1871 871 444 928 790 1170 1439 2215 1589 927 1427 930 942 1185 1640 1879 1105 1801 1882 1435 2058 966 592 1203 Exceedances of 250 ppb 1-hour1 mean min med p95 p98 p99 23 59 124 97 175 33 162 23 32 73 106 120 57 263 89 83 135 46 15 39 0 0 0 0 2 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 1 7 38 5 97 2 93 1 4 18 7 11 24 195 5 17 32 6 0 1 131 303 566 511 576 173 579 127 181 351 535 565 215 773 574 379 543 259 84 221 257 512 663 1142 776 318 737 258 308 457 843 769 347 839 688 466 697 356 156 438 334 643 761 1574 872 432 791 420 364 525 947 930 447 1002 860 563 817 432 231 635 Exceedances of 300 ppb 1-hour1 mean min med p95 p98 p99 8 22 51 45 80 12 72 8 9 27 45 60 21 114 36 33 43 16 5 17 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 1 40 0 32 0 0 2 1 1 8 66 1 3 2 0 0 0 43 137 304 228 317 62 316 40 52 158 277 294 78 407 280 181 208 99 25 91 106 230 380 582 424 127 396 91 104 220 435 371 162 443 369 296 303 163 57 198 131 322 392 908 482 184 430 171 138 270 514 416 200 470 422 364 339 200 90 318 1 The mean number of exceedances represents the number of exceedances occurring at all monitors in a particular location divided by the number of site- years across the monitoring period. The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for the number of exceedances in any one year within the monitoring period. April 2008 Draft 56 ------- 2 Table 15. Estimated number of exceedances of short-term (1-hour) potential health effect benchmark levels in a year on- 3 roads, 2001-2006 recent NOi air quality adjusted to just meet the current standard (0.053 ppm annual average). Location Boston Chicago Cleveland Denver Detroit Los Angeles Miami New York Philadelphia Washington Atlanta El Paso Jacksonville Las Vegas Phoenix Provo St. Louis Other CMSA Not MSA Exceedances of 200 ppb 1-hour1 mean min med p95 p98 p99 87 176 387 277 440 106 406 84 174 208 335 389 607 278 149 516 182 64 101 0 0 14 0 17 0 3 0 0 0 0 4 56 0 0 1 0 0 0 12 61 268 113 309 23 306 14 60 83 135 257 542 43 19 345 69 6 7 458 805 1117 964 1214 533 1173 458 726 874 1293 1251 1385 1319 758 1664 762 333 569 753 1022 1322 1233 1444 788 1345 709 973 1171 1647 1604 1642 1929 1172 1966 1100 569 874 990 1139 1735 1560 1628 893 1416 872 1184 1310 1755 1737 1743 2196 1352 2115 1216 740 1095 Exceedances of 250 ppb 1-hour1 mean min med p95 p98 p99 23 59 149 87 166 31 193 25 51 63 143 144 273 101 33 228 59 19 39 0 0 0 0 0 0 0 0 0 0 0 0 5 0 0 0 0 0 0 1 7 65 22 90 2 113 1 7 10 21 66 202 6 1 72 8 0 1 137 335 573 337 513 186 669 149 239 327 687 530 789 680 203 729 302 105 232 263 560 676 430 689 290 855 295 383 426 973 858 924 828 303 818 468 207 419 330 620 846 557 744 363 923 413 521 558 1093 971 1027 1045 370 847 576 300 569 Exceedances of 300 ppb 1-hour1 mean min med p95 p98 p99 7 23 62 28 67 10 88 8 16 21 61 54 125 42 7 134 20 6 16 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 15 5 25 0 35 0 1 1 4 20 74 0 0 5 1 0 0 38 128 326 125 265 59 367 49 77 127 339 221 436 354 48 643 127 31 95 93 295 407 203 322 115 542 110 153 181 510 350 490 502 70 693 211 72 184 132 354 428 283 385 150 588 177 227 224 656 441 557 565 95 694 260 120 264 1 The mean number of exceedances represents the number of exceedances occurring at all monitors in a particular location divided by the number of site- years across the monitoring period. The min, med, p95, p98, and p99 represent the minimum, median, 95th, 98th, and 99th percentiles of the distribution for the number of exceedances in any one year within the monitoring period. April 2008 Draft 57 ------- 1 average by about 60%, indicating the presence of highly influential data at the upper percentiles 2 of the distribution at each location. This is evident when considering the 95th - 99th percentiles, 3 where several hundred to around two thousand exceedances of 200 ppb were estimated. 4 However, the estimated number of exceedances is lower for locations containing more site-years 5 of data than for the locations with the fewest site-years. This trend is consistent with those 6 described earlier, whereas estimates of exceedances in the simulated data for the large urban 7 areas are stabilized by greater sample size (both the number of monitors and 1-hour values). The 8 median number of exceedances of 200 ppb at the locations containing a larger monitoring 9 network (i.e. at least 40 site-years per year-group) was estimated to be between 10 and 100 per 10 year. Upper bounds for the locations with the greatest number of monitoring sites approach 11 around 1,000 estimated on-road exceedances per year upon just meeting the current standard. 12 It should be noted that the estimated on-road concentrations and number exceedances for 13 many of the locations were higher for the 2001-2006 rolled-up data when compared with the 14 1995-2000 rolled-up data. To obtain generally comparable results across the two time periods, 15 the assumption for the concentration roll-up was that a similar level of variability be maintained 16 from year-to-year (or year-group to year-group). As described in section 2.5 of the draft TSD, a 17 slight increase in hourly COV occurred from 1995-2006 (-10% for all locations). The effect 18 may have finally emerged in this combined simulation by generating a greater number of 19 concentrations above the potential health effect benchmarks that may have previously been just 20 below the threshold in the earlier on-road simulations considering the as is ambient 21 concentrations. 22 6.4 UNCERTAINTY AND VARIABILITY 23 This uncertainty analysis first identifies the sources of the assessment that do or do not 24 contribute to uncertainty, and provide a rationale for why this is the case. A qualitative 25 evaluation follows for the types and components of uncertainty, resulting in a matrix describing, 26 for each source of uncertainty, both the direction and magnitude of influence has on exposure 27 estimates. The bias direction indicates how the source of uncertainty is judged to influence 28 estimated concentrations, either the concentrations are likely "over-" or "under-estimated". In 29 the instance where two types or components of uncertainty result in offsetting direction of April 2008 Draft 58 ------- 1 influence, the uncertainty was judged as "both". The magnitude indicates an estimated size of 2 influence the uncertainty has on estimated concentrations. "Minimal" uncertainty was noted O 4 where quantitative evidence indicates the influence is either conditional and/or limited to few 5 components in type. A characterization of "moderate" was assigned where multiple components 6 of uncertainty existed within a given type and act in similar direction, however the presence of 7 all at once may be dependent on certain conditions. "Major" uncertainty was used where 8 multiple components of uncertainty exist within a given type, the components have few limiting 9 conditions, and the components consistently act in similar bias direction. "Unknown" was 10 assigned where there was no evidence reviewed to judge the uncertainty associated with the 11 source. Table 16 provides a summary of the sources of uncertainty identified in the air quality 12 characterization and the judged bias and magnitude of each. 13 6.4.1 Air Quality Data 14 One basic assumption is that the AQS NC>2 air quality data used are quality assured 15 already. Reported concentrations contain only valid measures, since values with quality 16 limitations are either removed or flagged. There is likely no selective bias in retention of data 17 that is not of reasonable quality, it is assumed that selection of high concentration poor quality 18 data would be just as likely as low concentration data of poor quality. Given the numbers of 19 measurements used for this analysis, it is likely that even if a few low quality data are present in 20 the data set, they would not have any significant effect on the results presented here. Therefore, 21 the air quality data and database used likely contributes minimally to uncertainty. Temporally, 22 the data are hourly measurements and appropriately account for variability in concentrations that 23 are commonly observed for NC>2 and by definition are representative of an entire year. In 24 addition, having more than one monitor does account for some of the spatial variability in a 25 particular location. However, the degree of representativeness of the monitoring data used in this 26 analysis can be evaluated from several perspectives, one of which is how well the temporal and 27 spatial variability are represented. In particular, missing hourly measurements at a monitor may 28 introduce bias (if different periods within a year or different years have different numbers of 29 measured values) and increase the uncertainty. Furthermore, the spatial representativeness will 30 be poor if the monitoring network is not dense enough to resolve the spatial variability (causing April 2008 Draft 59 ------- 1 increased uncertainty) or if the monitors are not evenly distributed (causing a bias). Additional 2 uncertainty regarding temporal and spatial representation by the monitors is expanded below. 3 6.4.2 Measurement Technique for Ambient NOi 4 One source of uncertainty for NC>2 air quality data is due to interference with other 5 oxidized nitrogen compounds. The ISA points out positive interference, commonly from HNOs, 6 of up to 50%, particularly during the afternoon hours, resulting in overestimation of 7 concentrations. Also, negative vertical gradients exist for monitors (2.5 times higher at 4 meter 8 vs. 15 meter vertical siting (draft ISA, section 2.5.3.3), thus monitors positioned on rooftops may 9 underestimate exposures. Only 7 of the 1776 monitors in the named locations contained 10 monitoring heights of 15 meters or greater, with nearly 60% at 4 meters or less height, and 80% 11 at 5 meters or less in height. Not accounting for this potential vertical gradient in NO2 12 concentrations may generate underestimates of exceedances for some site-years, however the 13 overall impact of inferences made for the locations included in this assessment is likely minimal 14 since most monitors sited at less than 4-5 meters in vertical height. 15 6.4.3 Temporal Representation 16 Data are valid hourly measures and are of similar temporal scale as identified health 17 effect benchmark concentrations. There are frequent missing values within a given valid year 18 which contribute to the uncertainty as well as introducing a possible bias if some seasons, day 19 types (e.g., weekday/weekend), or time of the day (e.g., night or day) are not equally represented. 20 Since a 75 percent daily and hourly completeness rule was applied, some of these uncertainties 21 and biases were reduced in these analyses. Data were not interpolated in the analysis. Similarly, 22 there may be bias and uncertainty if the years monitored vary significantly between locations. 23 Although monitoring locations within a region do change over time, the NO2 network has been 24 reasonably stable over the 1995-2006 period, particularly at locations with larger monitoring 25 networks, so the impact to uncertainty is expected to be minimal regarding both bias direction 26 and magnitude. It should also be noted that use of the older data in some of the analyses here 27 carries the assumption that the sources present at that time are the same as current sources, 28 adding uncertainty to results if this is not the case. Separating the data into two 5 year groups 6 28 monitors did not have height reported (therefore, 177 + 28 = 205 total number of monitors in named locations) April 2008 Draft 60 ------- 1 (historic and recent) before analysis reduces the potential impact from changes in national- or 2 location-specific source influences and is judged to have a minimal magnitude. 3 6.4.4 Spatial Representation 4 Relative to the physical area, there are only a small number of monitors in each location. 5 Since most locations have sparse siting, the monitoring data are assumed to be spatially 6 representative of the locations analyzed here. This includes areas between the ambient monitors 7 that may or may not be influenced by similar local sources of NC>2. For these reasons the 8 uncertainty and bias due to the spatial network may be moderate, although the monitoring 9 network design should have addressed these issues within the available resources and other 10 monitoring constraints. This air quality characterization used all monitors meeting the 75 11 percent completeness criteria, without taking into account the monitoring objectives or land use 12 for the monitors. Thus, there will be some lack of spatial representation and likely moderate 13 uncertainty due to the inclusion/exclusion of some monitors that are very near local sources 14 (including mobile sources). 15 6.4.5 Air Quality Adjustment Procedure 16 The primary uncertainty of the empirical method used to estimate exceedances under the 17 current-standard scenario is due to the uncertainty of the true relationship between the annual 18 mean concentrations and the number of exceedances. The empirical method assumes that if the 19 annual means change then all the hourly concentrations will change proportionately. However, 20 different sources have different temporal emission profiles, so that applied changes to the annual 21 mean concentrations at monitors may not correspond well to all parts of the concentration 22 distribution equally. Similarly, emissions changes that affect the concentrations at the site with 23 the highest annual mean concentration will not necessarily impact lower concentration sites 24 proportionately. This could result in overestimations in the number of exceedances at lower 25 concentration sites within a location, however it is likely to be minimal given that the highest 26 concentrations typically were measured at the monitoring sites with the highest annual average 27 concentrations within the location (draft TSD, Appendix C). This minimal bias would apply to 28 areas that contain several monitors, such as Boston, New York, or Los Angeles. Universal 29 application of the proportional simulation approach at each of the locations was done for 30 consistency and was designed to preserve the inherent variability in the concentration profile. A April 2008 Draft 61 ------- 1 few locations were noted that may have an exceptional number of exceedances as a result of the 2 air quality adjustment approach, particularly those locations with few monitoring sites that 3 contained very low annual average concentrations and/or atypical variability in hourly 4 concentrations. These locations (e.g., Miami, Jacksonville, Provo) could contain moderate 5 overestimations at the upper tails of the concentration distribution, leading to bias in number of 6 estimated exceedances at both the upper percentiles and the mean for the scenarios using the air 7 quality simulated to just meet the current standard. 8 6.4.6 On-Road Concentration Simulation 9 On-road and ambient monitoring NC>2 concentrations have been shown to be correlated 10 significantly on a temporal basis (e.g., Cape et al., 2004) and motor vehicles are a significant 11 emission source of NOX, providing support for estimating on-road concentrations using ambient 12 monitoring data. The relationship used in this analysis to estimate on-road NC>2 concentrations 13 was derived from data collected in measurement studies containing mostly long-term averaging 14 times, typically 14-days or greater in duration (e.g., Roorda-Knape, 1998; Pleijel et al., 2004; 15 Cape et al, 2004), although one study was conducted over a one-hour time averaging period 16 (Rodes and Holland, 1981). This is considered appropriate in this analysis to estimate on-road 17 hourly concentrations from hourly ambient measures, assuming a direct relationship exists 18 between the short-term peaks to time-averaged concentrations (e.g., hourly on-road NC>2 19 concentrations are correlated with 24-hour averages). While this should not impact the overall 20 contribution relationship between vehicles and ambient concentrations on roads, the decay 21 constant k will differ for shorter averaging times. The on-road concentration estimation also 22 assumes that concentration changes that occur on-road and at the monitor are simultaneous (i.e., 23 within the hour time period of estimation). Since time-activity patterns of individuals are not 24 considered in this analysis, there is no bias in the number of estimated exceedances. The long- 25 term data used to develop the model were likely collected over variable meteorological 26 conditions (e.g., shifting wind direction) and other influential attributes (e.g., rate of 27 transformation of NO to NC>2 during the daytime versus nighttime hours) than would be observed 28 across shorter time periods. This could result in either over- or under-estimations of 29 concentrations, depending on the time of day. The variability in NC>2 concentration within an 30 hour is also not considered in this analysis, that is, the on-road concentration at a given site will April 2008 Draft 62 ------- 1 likely vary during the 1-hour time period. If considering personal exposures to individuals 2 within vehicles that are traveling on a road, it is likely that their exposure concentrations would 3 also vary due to differing roadway concentrations. This could also result in either over- or 4 under-estimations of concentrations, depending on the duration of travel and type of road 5 traveled on. 6 On-road concentrations were not modified in this analysis to account for in-vehicle 7 penetration and decay. This indicates that in-vehicle concentrations would be overestimated if 8 using the on-road concentrations as a surrogate, given that reactive pollutants (e.g., PM2.5) tend 9 to have a lower indoor/outdoor (I/O) concentration ratio (Rodes et al., 1998). Chan and Chung 10 (2003) report mean (I/O) ratios of NO2 for a few roadways and driving conditions in Hong Kong. 11 On highways and urban streets, the value is centered about 0.6 to 1.0, indicating decay of NO2 as 12 it enters the vehicle. 13 At locations where traffic counts are very low (e.g., on the order of hundreds/day) the on- 14 road contribution has been shown to be negligible (Bell and Ashenden, 1997; Cape et al., 2004), 15 therefore any rural areas just meeting the standard with minimal traffic volumes would likely 16 have resulted in small overestimations of NO2 concentrations using eq (2). For any monitor that 17 is sited in close proximity of the roadway (14 monitors were sited at <10 m from a major road), 18 on-road concentrations may have been overestimated using eq (2), since the assumption is that 19 the ambient concentration is equivalent to the non-source impacted concentration. In some 20 locations (i.e., Boston, Chicago, Denver, Los Angeles, Miami, St. Louis, and Washington DC), 21 at least half of the monitors used in this analysis are sited < 100 m from a major road (see draft 22 TSD, Table 5, section 2.3.3), a distance noted by some researchers a possibly receiving notable 23 impact from vehicle emissions (e.g., Beckerman et al., 2008). In addition, NOX is primarily 24 emitted as NO (e.g., Heeb et al., 2008; Shorter et al., 2005), with substantial secondary formation 25 due predominantly to NO + Os -> NO2 + O2. Numerous studies have demonstrated the Os 26 reduction that occurs near major roads, reflecting the transfer of odd oxygen to NO to form NO2, 27 a process that can impact NO2 concentrations both on- and downwind of the road. Some studies 28 report NO2 concentrations increasing just downwind of roadways and that are inversely 29 correlated with Os (e.g., Beckerman et al., 2008), suggesting that peak concentration of NO2 may 30 not always occur on the road, but at a distance downwind. Uncertainty regarding where the peak 31 concentration occurs (on-road or at a distance from the road) in combination with the form of the April 2008 Draft 63 ------- 1 exponential model used to estimate the on-road concentrations (the highest concentration occurs 2 at zero distance from road) could also lead to overestimation. However, the interpretation of the 3 estimate is what may be most uncertain, that is whether the exceedances are occurring on the 4 road or nearby. 5 Another source of uncertainty is the extent to which the near-road study locations 6 represent the locations studied in these analyses. The on-road and near-road data were collected 7 in a few locations, most of them outside of the United States. The source mixes (i.e., the vehicle 8 fleet) in study locations may not be representative of the U.S. fleet. Without detailed information 9 characterizing the emissions patterns for the on-road study areas, there was no attempt to match 10 the air quality characterization locations to specific on-road study areas, which might have 11 improved the precision of the estimates. However, since concentration ratios were selected 12 randomly from all the near-road studies and applied to each monitor individually, and since we 13 estimated overall minimum and upper bounds using multiple simulations, the analysis provides a 14 reasonable lower and upper bound estimate of the uncertainly. 15 6.4.7 Health Benchmark 16 The choice of potential health effect benchmarks, and the use of those benchmarks to 17 assess risks, can introduce uncertainty into the risk assessment. For example, the potential health 18 effect benchmarks used were based on studies where volunteers were exposed to NC>2 for 19 varying lengths of time. Typically, the NC>2 exposure durations were between 30 minutes and 2 20 hours. This introduces some uncertainty into the characterization of risk, which compared the 21 potential health effect benchmarks to estimates of exposure over a 1-hour time period. Use of a 22 1-hour averaging time could over- or under-estimate risks. In addition, the human exposure 23 studies evaluated airways responsiveness in mild asthmatics. For ethical reasons, more severely 24 affected asthmatics and asthmatic children were not included in these studies. Severe asthmatics 25 and/or asthmatic children may be more susceptible than mildly asthmatic adults to the effects of 26 NO2 exposure. Therefore, the potential health effect benchmarks based on these studies could 27 underestimate risks in populations with greater susceptibility. April 2008 Draft 64 ------- 1 2 3 Table 16. Summary of qualitative uncertainty analysis for the air quality and health risk characterization. Source Air Quality Data Ambient Measurement Temporal Representation Spatial Representation Air Quality Adjustment On-Road Simulation Health Benchmarks Type Database quality Interference Vertical siting Scale Missing data Years monitored Source changes Scale Monitor objectives Temporal scale Spatial scale Temporal scale Decay Spatial scale Model used Non US studies used Averaging time Susceptibility Bias Direction both over under none both both over both both over over both over over over unknown unknown under Magnitude minimal moderate minimal none minimal minimal minimal moderate moderate moderate moderate minimal minimal moderate minimal unknown minimal moderate Notes: Bias Direction: indicates the direction the source of uncertainty is judged to influence either the concentration or risk estimates. Magnitude: indicates the estimated size of influence. minimal - influence is either conditional and/or limited to few components in type moderate - multiple components of uncertainty existed within a given type and act in similar direction, however the presence of all at once may be dependent on certain conditions. major - multiple components of uncertainty exist within a given type, the components have few limiting conditions, and the components consistently act in similar bias direction. April 2008 Draft 65 ------- i 7. EXPOSURE ASSESSMENT AND HEALTH RISK 2 CHARACTERIZATION 3 7.1 OVERVIEW 4 This section documents the methodology and data used in the inhalation exposure 5 assessment and associated health risk characterization for NC>2 conducted in support of the 6 current review of the NC>2 primary NAAQS. Two important components of the analysis include 7 the approach for estimating temporally and spatially variable NC>2 concentrations and simulating 8 human contact with these pollutant concentrations. Both air quality and exposure modeling 9 approaches have been used to generate estimates of 1-hour NC>2 exposures within selected urban 10 areas of the U.S. across a 3-year period (2001-2003). Exposures and risk were characterized 11 considering recent air quality conditions (as is) and for air quality adjusted to just meet the 12 current NC>2 standard (0.053 ppm annual average). Details on the approaches used are provided 13 below and in Chapter 3 in the draft TSD. Briefly, the discussion includes the following: 14 • Description of the inhalation exposure model and associated input data 15 • Evaluation of estimated NO2 exposures 16 • Assessment of the quality and limitations of the input data for supporting the goals of 17 the NC>2 NAAQS exposure and risk characterization. 18 The combined dispersion and exposure modeling approach was both time and labor 19 intensive. To date, only the exposure and risk results for the Philadelphia case-study are 20 complete and are presented in this draft document. Location-specific input data for Philadelphia 21 and the other selected case-study areas are presented where collected (mainly meteorological 22 data) to provide information on the relative variability of the input data to be used. 23 7.2 OVERVIEW OF HUMAN EXPOSURE MODELING USING APEX 24 The purpose of this exposure analysis is to allow comparisons of population exposures to 25 ambient NC>2 among and within selected locations, and to characterize risks associated with 26 current air quality levels and with just meeting the current 0.053 ppm annual average standard. 27 This section provides a brief overview of the model used by EPA to estimate NC>2 population April 2008 Draft 66 ------- 1 exposure. Details about the application of the model to estimate NC>2 population exposure are 2 provided in the following sections and in Chapter 3 of the draft TSD. 3 The EPA has developed the Air Pollutants Exposure Model (APEX) model for estimating 4 human population exposure to criteria and air toxic pollutants. APEX serves as the human 5 inhalation exposure model within the Total Risk Integrated Methodology (TRIM) framework 6 (EPA 2006a; 2006b) and was recently used to estimate population exposures in 12 urban areas 7 for the O3 NAAQS review (EPA, 2007g; 2007h). 8 APEX is a probabilistic model designed to account for sources of variability that affect 9 people's exposures. APEX simulates the movement of individuals through time and space and 10 estimates their exposure to a given pollutant in indoor, outdoor, and in-vehicle 11 microenvironments. The model stochastically generates a sample of simulated individuals using 12 census-derived probability distributions for demographic characteristics. The population 13 demographics are drawn from the year 2000 Census at the tract, block-group, or block level, and 14 a national commuting database based on 2000 census data provides home-to-work commuting 15 flows. Any number of simulated individuals can be modeled, and collectively they approximate 16 a random sampling of people residing in a particular study area. 17 Daily activity patterns for individuals in a study area, an input to APEX, are obtained 18 from detailed diaries that are compiled in the Consolidated Human Activity Database (CHAD) 19 (McCurdy et al., 2000; EPA, 2002). The diaries are used to construct a sequence of activity 20 events for simulated individuals consistent with their demographic characteristics, day type, and 21 season of the year, as defined by ambient temperature regimes (Graham and McCurdy, 2004). 22 The time-location-activity diaries input to APEX contain information regarding an individuals' 23 age, gender, race, employment status, occupation, day-of-week, daily maximum hourly average 24 temperature, the location, start time, duration, and type of each activity performed. Much of this 25 information is used to best match the activity diary with the generated personal profile, using 26 age, gender, employment status, day of week, and temperature as first-order characteristics. The 27 approach is designed to capture the important attributes contributing to an individuals' behavior, 28 and of likely importance in this assessment (i.e., time spent outdoors) (Graham and McCurdy, 29 2004). Furthermore, these diary selection criteria give credence to the use of the variable data 30 that comprise CHAD (e.g., data collected were from different seasons, different states of origin, 31 etc.). April 2008 Draft 67 ------- 1 APEX has a flexible approach for modeling microenvironmental concentrations, where 2 the user can define the microenvironments to be modeled and their characteristics. Typical 3 indoor microenvironments include residences, schools, and offices. Outdoor microenvironments 4 include for example near roadways, at bus stops, and playgrounds. Inside cars, trucks, and mass 5 transit vehicles are microenvironments which are classified separately from indoors and 6 outdoors. APEX probabilistically calculates the concentration in the microenvironment 7 associated with each event in an individual's activity pattern and sums the event-specific 8 exposures within each hour to obtain a continuous series of hourly exposures spanning the time 9 period of interest. The estimated pollutant concentrations account for the effects of ambient 10 (outdoor) pollutant concentration, penetration factors, air exchange rates, decay/deposit! on rates, 11 proximity to important outdoor sources, and indoor source emissions, each depending on the 12 microenvironment, available data, and estimation method selected by the user. And, since the 13 modeled individuals represent a random sample of the population of interest, the distribution of 14 modeled individual exposures can be extrapolated to the larger population. 15 The model simulation can be summarized in the following five steps: 16 1. Characterize the study area. APEX selects census blocks within a study area - and 17 thus identifies the potentially exposed population - based on user-defined criteria and 18 availability of air quality and meteorological data for the area. 19 2. Generate simulated individuals. APEX stochastically generates a sample of 20 hypothetical individuals based on the census data for the study area and human 21 profile distribution data 22 3. Construct a sequence of activity events. APEX constructs an exposure event 23 sequence spanning the period of the simulation for each of the simulated individuals 24 and based on the activity pattern data. 25 4. Calculate hourly concentrations in microenvironments. APEX users define 26 microenvironments that people in the study area would visit by assigning location 27 codes in the activity pattern to the user-specified microenvironments. The model then 28 calculates hourly concentrations of a pollutant in each of these microenvironments for 29 the period of simulation, based on the user-provided microenvironment descriptions, 30 the hourly air quality data, and for some of the indoor microenvironments, indoor April 2008 Draft 68 ------- 1 sources of NC>2. Microenvironmental concentrations are calculated for each of the 2 simulated individuals. 3 5. Estimate exposures. APEX estimates a concentration for each exposure event based 4 on the microenvironment occupied during the event. These values can be averaged 5 by clock hour to produce a sequence of hourly average exposures spanning the 6 specified exposure period. These hourly values may be further aggregated to produce 7 daily, monthly, and annual average exposure values. 8 7.3 CHARACTERIZATION OF STUDY AREAS 9 7.3.1 Study Area Selection 10 The selection of areas to include in the exposure analysis takes into consideration the 11 location of field and epidemiological studies, the availability of ambient monitoring and other 12 input data, the desire to represent a range of geographic areas, population demographics, general 13 climatology, and results of the ambient air quality characterization. 14 Locations of interest were initially identified through a similar statistical analysis of the 15 ambient NC>2 air quality data described above for each site within a location. Criteria were 16 established for selecting sites with high annual means and/or high numbers of exceedances of 17 health effect benchmark concentrations. The analysis considered all ambient monitoring data 18 combined (1995-2006), as well as the more recent air quality data (2001-2006) separately. 19 The 90th percentile served as the point of reference for the annual means, and across all 20 complete site-years for recent ambient monitoring (2001 through 2006), this value was 23.52 21 ppb. Seventeen locations contained one or more site-years with an annual average concentration 22 at or above the 90th percentile. When combined with the number of 1-hour NC>2 concentrations 23 at or above 200 ppb, only two locations fit these criteria, Philadelphia and Los Angeles. 24 Considering the short-term criterion alone, Detroit contained the greatest number of exceedances 25 of 200 ppb (numbering 12 for years 2001-2006). Two additional locations were selected by 26 considering geographic/climatologic representation and also their historic ambient 27 concentrations. Atlanta (1 exceedance of 200 ppb and a maximum annual average concentration 28 of 26.63 ppb for years 1995-2006) and Phoenix (maximum annual mean concentration of 37.09 29 ppb for 2001-2006 and 37 exceedances of 200 ppb for years 1995-2006) were selected to April 2008 Draft 69 ------- 1 represent the southern and western region of the US from the pool of remaining locations with 2 either exceedances of the 90th percentile annual mean concentration or 200 ppb 1-hour. 3 4 To summarize, the following 5 urban areas were selected for a detailed exposure analysis: 5 • Philadelphia, PA 6 • Atlanta, GA 7 • Detroit, MI 8 • Los Angeles, CA 9 • Phoenix, AZ 10 The exposure periods modeled were 2001 through 2003 to envelop the most recent year 11 of travel demand modeling (TDM) data available for the respective study locations (i.e., 2002) 12 and to include a 3 years of meteorological data to achieve stability in the dispersion/exposure 13 model estimates. 14 7.3.2 Study Area Descriptions 15 The APEX study area has traditionally been on the scale of a city or slightly larger 16 metropolitan area, although it is now possible to model larger areas such as combined statistical 17 areas (CSAs). In this analysis the study area is defined by a single or few counties. The 18 demographic data used by the model to create personal profiles is provided at the census block 19 level. For each block the model requires demographic information representing the distribution 20 of age, gender, race, and work status within the study population. Each block has a location 21 specified by latitude and longitude for some representative point (e.g., geographic center). The 22 current release of APEX includes input files that already contain this demographic and location 23 data for all census tracts, block groups, and blocks in the 50 United States, based on the 2000 24 Census. 25 The first area study area selected for a detailed exposure analysis was Philadelphia 26 County since the TDM data were readily available and was one of two locations where both 27 selection criteria were met using recent air quality (the other being Los Angeles). Philadelphia 28 County is a large part of the Philadelphia-Wilmington-Atlantic City CMS A, comprised of 16,857 29 blocks and containing a total population of 1,475,651 persons (representing approximately 97% 30 of the county population). April 2008 Draft 70 ------- 1 7.4 CHARACTERIZATION OF AMBIENT HOURLY AIR QUALITY 2 DATA USING AERMOD 3 7.4.1 Overview 4 Air quality data used for input to APEX were generated using AERMOD, a steady-state, 5 Gaussian plume model (EPA, 2004). For each identified case-study location, the following steps 6 were performed 7 1. Collect and analyze general input parameters. Meteorological data, processing 8 methodologies used to derive input meteorological fields (e.g., temperature, wind 9 speed, precipitation), and information on surface characteristics and land use are 10 needed to help determine pollutant dispersion characteristics, atmospheric 11 stability and mixing heights. 12 2. Estimate emissions. The emission sources modeled included, major stationary 13 emission sources, on-road emissions that occur on major roadways, and fugitive 14 emissions. 15 3. Define receptor locations. Three sets of receptors were identified for the 16 dispersion modeling, including ambient monitoring locations, census block 17 centroids, and links along major roadways. 18 4. Estimate concentrations at receptors. Hourly concentrations were estimated for 19 each year of the simulation (years 2001 through 2003) by combining 20 concentration contributions from each of the emission sources and accounting for 21 sources not modeled. 22 The AERMOD hourly concentrations were then used as input to the APEX model to 23 estimate population exposure concentrations. Details regarding both modeling approaches and 24 input data used are provided below and in Chapter 3 of the draft TSD. Hourly NO2 25 concentrations were estimated for each of 3 years (2001-2003) at each of the defined receptor 26 locations (census blocks and roadway links) using hourly NOX emission estimates and dispersion 27 modeling. Relevant input data collected for Philadelphia as well some of the data collected as 28 part of the other selected case-study locations to be evaluated in the second draft risk and 29 exposure assessment are presented below. April 2008 Draft 71 ------- 1 7.4.2 General Model Inputs 2 7.4.2.1 Meteorological Data 3 All meteorological data used for the AERMOD dispersion model simulations were 4 processed with the AERMET meteorological preprocessor, version 06341. Raw surface 5 meteorological data for the 2001 to 2003 period were obtained from the Integrated Surface 6 Hourly (ISH) Database,7 maintained by the National Climatic Data Center (NCDC). The ISH 7 data used for this study consists of typical hourly surface parameters (including air and dew point 8 temperature, atmospheric pressure, wind speed and direction, precipitation amount, and cloud 9 cover) from hourly Automated Surface Observing System (ASOS) stations. No on-site 10 observations were used. 11 Surface meteorological stations for this analysis were those at the major airports of each 12 of the five cities in the study: 13 • Philadelphia: Philadelphia International (KPHL) 14 • Atlanta: Atlanta Hartsfield International (KATL) 15 • Detroit: Detroit Metropolitan (KDTW) 16 • Los Angeles: Los Angeles International (KLAX) 17 • Phoenix: Phoenix Sky Harbor International (KPHX). 18 The selection of surface meteorological stations for each city minimized the distance 19 from the station to city center, minimized missing data, and maximized land-use 20 representativeness of the station site compared to the city center. The total number of surface 21 observations per station were compiled and the percentage of those observations accepted by 22 AERMET (i.e., those observations that were both not missing and within the expected ranges of 23 values) were typically >99%. 24 Mandatory and significant levels of upper-air data were obtained from the NOAA 25 Radiosonde Database.8 Upper air observations show less spatial variation than do surface 26 observations; thus they are both representative of larger areas and measured with less spatial 27 frequency than are surface observations. The selection of upper-air station locations for each 7 http://wwwl.ncdc.noaa.gov/pub/data/techrpts/tr200101/tr2001-01 .pdf 8 http://raob.fsl.noaa.gov/ April 2008 Draft 72 ------- 1 city minimized both the proximity of the station to city center and the amount of missing data in 2 the records. The selected stations were: 3 • Philadelphia: Washington Dulles Airport (KIAD) 4 • Atlanta: Peachtree City (KFFC) 5 • Detroit: Detroit/Pontiac (KDTX) 6 • Los Angeles: Miramar Naval Air Station near San Diego (KNKX) 7 • Phoenix: Tucson (KTWC). 8 The total number of upper-air observations per station per height interval, and the 9 percentage of those observations accepted by AERMET were typically >99% for the pressure, 10 height, and temperature parameters however, dewpoint temperature, wind direction, and wind 11 speed parameters had lower acceptance rates (sometimes <75%), particularly when considering 12 greater atmospheric heights. 13 7.4.2.2 Surface Characteristics and Land Use Analysis 14 In addition to the standard meteorological observations of wind, temperature, and cloud 15 cover, AERMET analyzes three principal variables to help determine atmospheric stability and 16 mixing heights: the Bowen ratio9, surface albedo10 as a function of the solar angle, and surface 17 roughness11. 18 The January 2008 version of AERSURFACE was used to estimate land-use patterns and 19 calculate the Bowen ratio, surface albedo, and surface roughness as part of the AERMET 20 processing. AERSURFACE uses the US Geological Survey (USGS) National Land Cover Data 21 1992 archives (NLCD92)12. Three to four land-use sectors were manually identified around the 22 surface meteorological stations using this land-use data. These land-use sectors are used to 23 identify the Bowen ratio and surface albedo, which are assumed to represent an area around the 24 station of radius 10 km, and to calculate surface roughness by wind direction. 9 For any moist surface, the Bowen Ratio is the ratio of heat energy used for sensible heating (conduction and convection) to the heat energy used for latent heating (evaporation of water or sublimation of snow). The Bowen ratio ranges from about 0.1 for the ocean surface to more than 2.0 for deserts. Bowen ratio values tend to decrease with increasing surface moisture for most land-use types. 10 The ratio of the amount of electromagnetic radiation reflected by the earth's surface to the amount incident upon it. Value varies with surface composition. For example, snow and ice vary from 80% to 85% and bare ground from 10% to 20%. 11 The presence of buildings, trees, and other irregular land topography that is associated with its efficiency as a momentum sink for turbulent air flow, due to the generation of drag forces and increased vertical wind shear. 12 http://seamless.usgs.gov/ April 2008 Draft 73 ------- 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 A monthly temporal resolution was used for the Bowen ratio, albedo, and surface roughness for all five meteorological sites. Because the five sites were located at airports, a lower surface roughness was calculated for the 'Commercial/Industrial/Transportation' land-use type to reflect the dominance of transportation land cover rather than commercial buildings. Los Angeles and Phoenix are arid regions, which increases the calculated albedo and Bowen ratio values and decreases the surface roughness values assigned to the 'Shrubland' and 'Bare Rock/Sand/Clay' land-use types to reflect a more desert-like area. Philadelphia and Detroit each have at least one winter month of continuous snow cover, which tends to increase albedo, decrease Bowen ratio, and decrease surface roughness for most land-use types during the winter months compared to snow-free areas. Seasons were assigned for each site based on 1971-2000 NCDC 30-year climatic normals and on input from the respective state climatologists. Table 17 provides the seasonal definitions for each city. Further discussion of the land use and surface analysis, as well as a discussion of the difference in results from employing the new AERSUKFACE tool is given in Appendix E of the draft TSD. Table 17. Seasonal specifications by selected case-study locations. Location Philadelphia Atlanta Detroit Los Angeles Phoenix Winter (continuous snow) Dec, Jan, Feb Dec, Jan, Feb, Mar Winter (no snow) Dec, Jan, Feb Spring Mar, Apr, May Mar, Apr, May Apr, May Apr, May, Jun Apr, May, Jun Summer Jun, Jul, Aug Jun, Jul, Aug Jun, Jul, Aug Jul, Aug, Sep Jul, Aug, Sep Fall Sep, Oct, Nov Sep, Oct, Nov Sep, Oct, Nov Oct, Nov, Dec, Jan, Feb, Mar Oct, Nov, Dec, Jan, Feb, Mar Season definitions provided by the AERSURFACE manual: Winter (continuous snow): Winter with continuous snow on ground Winter (no snow): Late autumn after frost and harvest, or winter with no snow Spring: Transitional spring with partial green coverage or short annuals Summer: Midsummer with lush vegetation Fall: Autumn with unharvested cropland 19 20 April 2008 Draft 74 ------- 1 2 Meteorological Data A nalysis 3 The AERMET application location and elevation were taken as the center of each 4 modeled city, estimated using Google Earth version 4.2.0198.2451 (beta). They are as follows: 5 • Philadelphia: 39.952 °N, 75.164 °W, 12m 6 • Atlanta: 33.755 °N, 84.391 °W, 306 m 7 • Detroit: 42.332 °N, 83.048 °W, 181m 8 • Los Angeles: 34.053 °N, 118.245 °W, 91 m 9 • Phoenix: 33.448 °N, 112.076 °W, 330m 10 For each site in this study, the 2001-2003 AERSUKFACE processing was run three times 11 - once assuming the entire period was drier than normal, once assuming the entire period was 12 wetter than normal, and once assuming the entire period was of average precipitation 13 accumulation. These precipitation assumptions influence the Bowen ratio, as discussed above. 14 To create meteorological input records that best represent the given city for each of the 15 three years, the resulting surface output files for each site were then pieced together on a month- 16 by-month basis, with selection based on the relative amount of precipitation in each month. Any 17 month where the actual precipitation amount received was at least twice the 1971-2000 NCDC 18 30-year climatic normal monthly precipitation amount was considered wetter than normal, while 19 any month that received less than half the normal amount of precipitation amount was considered 20 drier than normal; all other months were considered to have average surface moisture conditions. 21 Surface moisture conditions were variable when considering the month-location 22 combinations to 30-year climatic normals, with much of the precipitation across the 3-year 23 period reflective of typical to dry conditions (Table 18). April 2008 Draft 75 ------- 2 3 4 Table 18. Characterization of monthly precipitation levels in selected case-study locations compared to NCDC 30-year climatic normals, 2001-2003. 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Location Philadelphia Atlanta Detroit Los Angeles Phoenix Number of Months with Precipitation Level1 Avg 27 28 26 12 11 Dry 7 7 9 15 22 Wet 2 1 1 9 3 1Precipitation Level Definitions Avg: <2 times the normal precipitation level and >1/2 the normal amount. Dry: <1/2 the normal monthly precipitation amount. Wet: >2 times the normal precipitation level. 7.4.2.3 Additional AERMOD Input Specifications Since each of the case-study locations were MS A/CMS As, all emission sources were characterized as urban. The AERMOD toxics enhancements were also employed to speed calculations from area sources. NOX chemistry was applied to all sources to determine NC>2 concentrations. For the each of the roadway, fugitive, and airport emission sources, the ozone limiting method (OLM) was used, with plumes considered ungrouped. Because an initial NC>2 fraction of NOX is anticipated to be about 10% or less (Finlayson-Pitts and Pitts, 2000; Yao et al., 2005), a conservative value of 10% for all sources was selected. For all point source simulations the Plume Volume Molar Ratio Method (PVMRM) was used to estimate the conversion of NOX to NC>2, with the following settings: 1. Hourly series of Os concentrations were taken from EPA's AQS database13. The complete national hourly record of monitored 63 concentrations were filtered for the four monitors within Philadelphia County (stations 421010004, 421010014, 421010024, and 421010136). The hourly records of these stations were then averaged together to provide an average Philadelphia County concentrations of O3 for each hour of 2001-2003. 13 http://www.epa.gov/ttn/airs/airsaqs/detaildata/downloadaqsdata.htm April 2008 Draft 76 ------- 1 2. The equilibrium value for the NO2:NOX ratio was taken as 75%, the national average 2 ambient ratio.14 3 3. The initial NO2 fraction of NOX is anticipated to be about 10% or less. A default 4 value of 10% was used for all stacks (Finlayson-Pitts and Pitts, 2000). 5 7.4.3 Emissions Estimates 6 7.4.3.1 On-Road Emissions Preparation 1 Information on traffic data in the Philadelphia area was obtained from the Delaware 8 Valley Regional Planning Council (DVRPC15) via their most recent, baseline travel demand 9 modeling (TDM) simulation - that is, the most recent simulation calibrated to match observed 10 traffic data. 11 Emission Sources and Locations 12 The TDM simulation's shapefile outputs include annual average daily traffic (AADT) 13 volumes and a description of the loaded highway network. The description of the network 14 consists of a series of nodes joining individual model links (i.e., roadway segments) to which the 15 traffic volumes are assigned, and the characteristics of those links, such as endpoint location, 16 number of lanes, link distance, and TDM-defined link daily capacity.16 17 The full set of links in the DVRPC network was filtered to include only those roadway 18 types considered major (i.e., freeway, parkway, major arterial, ramp), and that had AADT values 19 greater than 15,000 vehicles per day (one direction). Then, link locations from the TDM were 20 modified to represent the best known locations of the actual roadways, since there was not 21 always a direct correlation between the two. The correction of link locations was done based on 22 the locations of the nodes that define the end points of links with a GIS analysis, as follows. 23 A procedure was developed to relocate TDM nodes to more realistic locations. The 24 nodes in the TDM represent the endpoints of links in the transportation planning network and are 25 specified in model coordinates. The model coordinate system is a Transverse Mercator 14 Appendix W to CFR 51, page 466. http://www.epa.gov/scram001/guidance/guide/appw 03.pdf. 15 http://www.dvrpc.org/ 16 The TDM capacity specifications are not the same as those defined by the Highway Capacity Manual (HCM). Following consultation with DVRPC, the HCM definition of capacity was used in later calculations discussed below. April 2008 Draft 77 ------- 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 projection of the TranPlan Coordinate System with a false easting of 31068.5, false northing of- 200000.0, central meridian: -75.00000000, origin latitude of 0.0, scale factor of 99.96, and in units of miles. The procedure moved the node locations to the true road locations and translated to dispersion model coordinates. The Pennsylvania Department of Transportation (PA DOT) road network database17 was used as the specification of the true road locations. The nodes were moved to coincide with the nearest major road of the corresponding roadway type using a built- in function of ArcGIS. Once the nodes had been placed in the corrected locations, a line was drawn connecting each node pair to represent a link of the adjusted planning network. To determine hourly traffic on each link, the AADT volumes were converted to hourly values by applying DVRPC's seasonal and hourly scaling factors. The heavy-duty vehicle fraction - which is assumed by DVRPC to be about 6% in all locations and times - was also applied18. Another important variable, the number of traffic signals occurring on a given link, was obtained from the TDM link-description information. Table 19 summarizes the AADT volumes used in the simulations for each road type. Table 19. Statistical summary of AADT volumes (one direction) for Philadelphia County AERMOD simulations. Statistic Count Minimum AADT Maximum AADT Average AADT Road Type Arterial Freeway Ramp Arterial Freeway Ramp Arterial Freeway Ramp Arterial Freeway Ramp CBD 186 11 0 15088 15100 44986 39025 21063 25897 Fringe 58 10 4 15282 18259 16796 44020 56013 40538 21196 40168 24468 Suburban 210 107 3 15010 15102 15679 48401 68661 24743 20736 33979 18814 Urban 580 98 1 15003 15100 16337 44749 68661 16337 22368 31294 16337 http://www.pasda.psu.edu/ 18 As shown by Figure 1, NOX emissions from HDVs tend to be higher than their LDV counterparts by about a factor of 10. However, the HDV fraction is less than 10% of the total VMT in most circumstances, mitigating their influence on composite emission factors, although this mitigating effect is less pronounced at some times than others. For example, nighttimes on freeways tend to show a smaller reduction in HDV volume than in total volume, and thus an increased HDV fraction. This effect is not captured in most TDMs or emission postprocessors and - both to maintain consistency with the local MPO's vehicle characterizations and emissions modeling and due to lack of other relevant data - was also not included here. The net result of this is likely to be slightly underestimated emissions from major freeways during late-night times. April 2008 Draft 78 ------- 1 Emission Source Strength 2 On-road mobile emission factors were derived from the MOBILE6.2 emissions model as 3 follows. The DVRPC-provided external data files describing the VMT distribution by speed, 4 functional class, and hour, as well as the registration distribution and Post-1994 Light Duty 5 Gasoline Implementation for Philadelphia County were all used in the model runs without 6 modification. To further maintain consistency with the recent DVRPC inventory simulations 7 and maximize temporal resolution, the DVRPC's seasonal particulate matter (PM) MOBILE6 8 input control files were also used.19 These files include county-specific data describing the 9 vehicle emissions inspection and maintenance (I/M) Programs, on-board diagnostics (OBD) start 10 dates, vehicle miles traveled (VMT) mix, vehicle age distributions, default diesel fractions, and 11 representative minimum and maximum temperatures, humidity, and fuel parameters. The 12 simulations are designed to calculate average running NOX emission factors. 13 These input files were modified for the current project to produce running NOX emissions 14 in grams per mile for a specific functional class (Freeway, Arterial, or Ramp) and speed. 15 Iterative MOBILE6.2 simulations were conducted to create tables of average Philadelphia 16 County emission factors resolved by speed (2.5 to 65 mph), functional class, season, and year 17 (2001, 2002, or 2003) for each of eight combined MOBILE vehicle classes. 20 The resulting 18 tables were then consolidated into speed, functional class, and seasonal values for combined 19 light- and heavy-duty vehicles. Figure 1 shows an example of the calculated emission factors for 20 Autumn, 2001. 19 The present emissions model input files were based on MPO-provided PM, rather than NOX input files for a few reasons. First, MPO-provided PM files were used because they contain quarterly rather than annual or biannual information. In all cases the output species were modified to produce gaseous emissions. Further, many of the specified input parameters do not affect PM emissions, but were included by the local MPO to best represent local conditions, which were preserved in the present calculations of NOX emissions. This usage is consistent with the overall approach of preserving local information wherever possible. 20 HDDV - Heavy-Duty Diesel Vehicle, HDGV - Heavy-Duty Gasoline Vehicle, LDDT - Light-Duty Diesel Truck, LDDV - Light-Duty Diesel Vehicle, LDGT12 - Light-Duty Gasoline Truck with gross vehicle weight rating < 6,000 Ibs and a loaded vehicle weight of < 5,750 Ibs, LDGT 34 - Light-Duty Gasoline Truck with gross vehicle weight rating between 6,001 - 8,500 and a loaded vehicle weight of < 5,750 Ibs, LDGV - Light-Duty Gasoline Vehicle, MC - Motorcycles. April 2008 Draft 79 ------- 1 2 3 4 5 6 9 10 11 12 13 14 15 16 11 E 5 in c g 'm (A E LU X O z a> O) TO 0) 20 18 16 14 12 10 8 6 4 2 0 v •Fall Arterial LDV Fall Freeway LDV -Fall Arterial HDV Fall Freeway HDV 10 20 30 40 Average Speed (mph) 50 60 70 Figure 1. Example of Light- and heavy-duty vehicle NOX emissions grams/mile (g/mi) for arterial and freeway functional classes, Philadelphia 2001. To determine the emission strengths for each link for each hour of the year, the Philadelphia County average MOBILE6.2 speed-resolved emissions factor tables were merged with the TDM link data, which had been processed to determine time-resolved speeds. The spatial-mean speed of each link at each time was calculated following the methodology of the Highway Capacity Manual.21 Table 20 shows the resulting average speed for each functional class within each TDM region. Table 20. Average calculated speed by link type for Philadelphia County. Ramp Arterial Freeway Averag CBD N/A 34 51 Fringe 35 31 62 e Speed (mph) Suburban 35 44 66 Urban 35 32 62 Notes: N/A not available The resulting emission factors were then coupled with the TDM-based activity estimates to calculate emissions from each of the 1,268 major roadway links. However, many of the links 21 As defined in Chapter 9 of Recommended Procedure for Long-Range Transportation Planning and Sketch Planning. NCHRP Report 387, National Academy Press, 1997. 151 pp., ISBN No: 0-309-060-58-3. April 2008 Draft 80 ------- 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 were two sides of the same roadway segment. To speed model execution time, those links that could be combined into a single emission source were merged together. This was done for the 628 links (314 pairs) where opposing links were spatially paired and exhibited similar activity levels within 20% of each other. Other Emission Parameters Each roadway link is characterized as a rectangular area source with the width given by the number of lanes and an assumed universal lane width of 12 ft (3.66 m). The length and orientation of each link is determined as the distance and angle between end nodes from the adjusted TDM locations. In cases where the distance is such that the aspect ratio is greater than 100:1, the links were disaggregated into sequential links, each with a ratio less than that threshold. There were 27 links that exceeded this ratio and were converted to 55 segmented sources. Thus, the total number of area sources included in the dispersion simulations is 982. Table 21 shows the distribution of on-road area source sizes. Note that there are some road segments whose length was zero after GIS adjustment of node location. This is assumed to be compensated by adjacent links whose length will have been expanded by a corresponding amount. Table 21. On-road area source sizes for Philadelphia County. Minimum Median Average 1-a Deviation Maximum Segment Width (m) 3.7 11.0 13.7 7.7 43.9 Lanes 1.0 3.0 3.8 2.1 12.0 Segment Length (m) 0.0 220.6 300.2 259.5 1340.2 Resulting daily emission estimates were temporally allocated to hour of the day and season using MOBILE6.2 emission factors, coupled with calculated hourly speeds from the postprocessed TDM and allocated into SEASHR emission profiles for the AERMOD dispersion model. That is, 96 emissions factors are attributed to each roadway link to describe the emission strengths for 24 hours of each day of each of four seasons and written to the AERMOD input control file. April 2008 Draft 81 ------- 1 The release height of each source was determined as the average of the light- and heavy- 2 duty vehicle fractions, with an assumed light- and heavy-duty emission release heights of 1.0 ft 3 (0.3048 m) and 13.1 ft (4.0 m), respectively.22 Because AERMOD only accepts a single release 4 height for each source, the 24-hour average of the composite release heights is used in the 5 modeling. 6 Since surface-based mobile emissions are anticipated to be terrain following, no elevated 7 or complex terrain was included in the modeling. That is, all sources are assumed to lie in a flat 8 plane. 9 7.4.3.2 Stationary Sources Emissions Preparation 10 Data for the parameterization of major point sources in Philadelphia comes primarily from 11 two sources: the 2002 National Emissions Inventory (NEI; EPA, 2007e) and Clean Air Markets 12 Division (CAMD) Unit Level Emissions Database (EPA, 2007f). The NEI database contains 13 stack locations, emissions release parameters (i.e., height, diameter, exit temperature, exit 14 velocity), and annual NOX emissions. The CAMD database has information on hourly NOX 15 emission rates for all the units in the US, where the units are the boilers or equivalent, each of 16 which can have multiple stacks. These two databases generally contain complimentary 17 information, and were first evaluated for matching facility data. Then, annual emissions data 18 from the NEI were used to scale the hourly CAMD data where discrepancies existed between 19 estimated annual emissions. 20 Data Source Alignment and Scaling 21 To align the information between the two emission databases and extract the most useful 22 portion of each for dispersion modeling, the following methodology was used. 23 1. Collect data for all stacks within Philadelphia County. 24 2. Combine individual stacks that have identical stack physical parameters and were co- 25 located within about 10 m, to be simulated as a single stack with their emissions 26 summed. 27 3. Retain facilities with total emissions from all stacks exceeding 100 tpy NOX. 28 4. Remove fugitive releases, to be analyzed as a separate source group. 22 4.0 m includes plume rise from truck exhaust stacks. See Diesel Paniculate Matter Exposure Assessment Study for the Ports of Los Angeles and Long Beach. State of California Air Resources Board, Final Report, April 2006. April 2008 Draft 82 ------- 1 This resulted in 19 distinct, combined stacks from the NEI (Table 22). Then, the CAMD 2 database was queried for facilities that matched the facilities identified from the NEI database. 3 Facility matching was done on the facility name, Office of Regulatory Information Systems 4 (ORIS) identification code (when provided) and facility total emissions to ensure a best match 5 between the facilities. Once facilities were paired, individual units and stacks in the data bases 6 were paired based on annual emission totals. Most facilities contained similar total annual NOX 7 emissions when comparing the two databases, although at one facility, nearly half of the NEI 8 emissions (without fugitives) do not appear in the CAMD database, while another identified in 9 the NEI was not included at all in the CAMD. The reason for this is unknown and no 10 information was readily available on the relative accuracy of the two databases. 11 Hourly emissions in the CAMD database were scaled using a factor to match the NEI 12 annual total emissions based on each of the matched stacks/units. This includes accounting for 13 where emissions were reduced or absent from the CAMD database. Then for consistency, the 14 2001 and 2003 hourly emission profiles were also determined using the same scaling factors, but 15 applied to the respective CAMD emission profile. Details for source parameters and scaling 16 factors are provided in Chapter 3 of the draft TSD. 17 April 2008 Draft 83 ------- 1 2 Table 22. Stationary NOx emission sources modeled in Philadelphia County. Stack No. 817 818 819 820 821 855 856 857 858 859 860 861 862 863 864 865 866 867 868 NEI Site ID NEIPA2218 NEIPA2218 NEI40720 NEI40720 NEI40720 NEI40723 NEI40723 NEI40723 NEI40723 NEI40723 NEI7330 NEI7330 NEI7330 NEI7330 NEIPA101353 NEIPA101353 NEIPA101356 NEIPA101356 NEIPA2222 Facility Name EXELON GENERATION CO - DELAWARE STATION EXELON GENERATION CO - DELAWARE STATION JEFFERSON SMURFIT CORPORATION (U S) JEFFERSON SMURFIT CORPORATION (U S) JEFFERSON SMURFIT CORPORATION (U S) Sunoco Inc. - Philadelphia Sunoco Inc. - Philadelphia Sunoco Inc. - Philadelphia Sunoco Inc. - Philadelphia Sunoco Inc. - Philadelphia SUNOCO CHEMICALS (FORMER ALLIED SIGNAL) SUNOCO CHEMICALS (FORMER ALLIED SIGNAL) SUNOCO CHEMICALS (FORMER ALLIED SIGNAL) SUNOCO CHEMICALS (FORMER ALLIED SIGNAL) TRIGEN-SCHUYLKILL TRIGEN-SCHUYLKILL GRAYS FERRY COGENERATION PARTNERS GRAYS FERRY COGENERATION PARTNERS TRIGEN- EDISON SIC Code 4911 4911 2631 2631 2631 2911 2911 2911 2911 2911 2869 2869 2869 2869 4961 4961 4911 4911 4961 NAICS Code 221112 221112 32213 32213 32213 32411 32411 32411 32411 32411 325998 325998 325998 325998 22 22 22 22 62 ORIS Code 3160 3160 54785 54785 Stack Emiss. (tpy) 4.82 287.8 0.148 113.8 114.46 26.2 1.3 1.4 19.3 1032.8 0.033 49.1 34.6 77.2 128.6 61.5 143.2 90.3 130.5 Stack X (deg) -75.1358 -75.1358 -75.2391 -75.2391 -75.2391 -75.2027 -75.2003 -75.203 -75.2027 -75.2124 -75.0715 -75.0715 -75.0715 -75.0715 -75.1873 -75.1873 -75.1873 -75.1873 -75.1569 Stack Y (deg) 39.96769 39.96769 40.03329 40.03329 40.03329 39.92535 39.91379 39.92539 39.92535 39.90239 40.00649 40.00649 40.00649 40.00649 39.94239 39.94239 39.94239 39.94239 39.94604 4 7.4.3.3 Fugitive and A irport Emissions Preparation 5 Fugitive emission releases, as totaled in the NEI database, were modeled as area sources 6 with the profile of these releases determined by the overall facility profile of emissions. In 7 addition, emissions associated with the Philadelphia International Airport were estimated. 8 Fugitive Releases 9 Thirty five combined stacks were identified during the point source analysis (see previous 10 section) that were associated with facilities considered major emitters, but where the emissions 11 from the stacks are labeled Fugitive in the NEI. These stacks have zero stack diameter, zero April 2008 Draft 84 ------- 1 2 3 4 5 6 7 8 9 10 11 12 13 emission velocity, and exit temperature equal to average ambient conditions (295 K). Thus, we determined it was not appropriate to include these in the point source group simulation. These 35 stacks occur at only two facilities in the County: Exelon Generation Co - Delaware Station (NEI Site ID: NEIPA2218) and Sunoco Inc. - Philadelphia (NEI Site ID: NEI40723). Consequently, they were grouped by facility. The Sunoco emissions were grouped into two distinct categories based on release heights. Thus, to accommodate all of these sources most efficiently, a total of three area source groups were created: one for Sunoco emissions at 3.0 m, one for Sunoco emissions greater than 23.0 m, and one for Exelon. Their combined area source parameters are given in Table 23. Table 23. Emission parameters for the three Philadelphia County fugitive NOx area emission sources. No. 1 2 3 NEI Site ID NEI PA221 8 NEI 40723 NEI 40723 Facility Name Exelon Generation Co- Delaware Station Sunoco Inc. - Philadelphia Sunoco Inc. - Philadelphia Combined #of Stacks 2 26 7 NEI 2002 Emissions (tpy) 5.2 1,680.4 350.8 Average Stack Height (m) 6.5 3 26.7 Stacks Used for Profile ** 817+818 855+856 + 857+ 858+ 859 855+856 +857+ 858+ 859 Scaled Emissions (tpy) * 2001 4.8 1,873.8 391.2 2002 5.2 1,681.4 351.0 2003 6.4 2,202.4 459.8 14 15 16 17 18 19 20 21 22 In the case of the Sunoco emissions, the vertices of the area sources were determined by a convex hull encapsulating all the points. In the case of Excelon, only two points are provided, which is insufficient information to form a closed polygon. Instead, the boundary of the facility was digitized into a 20-sided polygon. Figure 2 shows the locations of these polygons. Emission profiles for the fugitive releases were determined from the CAMD hourly emission database in a method similar to that for the point sources. We determined scaling factors based on the ratio of the 2002 fugitive releases described by the NEI to the total, non- fugitive point source releases from the same facility. All stacks within that facility were April 2008 Draft 85 ------- 1 2 3 4 5 10 11 12 13 14 combined on an hourly basis for each year and the fugitive to non-fugitive scaling factor applied, ensuring that the same temporal emission profile was used for fugitives as for other releases from the facility, since the origins of the emissions should be parallel. We created external hourly emissions files for each of the three fugitive area sources with appropriate units (grams per second per square meter). Sunoco (ReleaseHght = 3m) \ ^ Sunoco (ReleaseHght = 23+ m) KPHL Airport Baggage Handling Area 6 7 8 Figure 2. Locations of the four ancillary area sources. Also shown are centroid receptor 9 locations. Philadelphia International Airport Another significant source of NOx emissions in Philadelphia County not captured in the earlier simulations is from operation of the Philadelphia International Airport (PHL). PHL is the only major commercial airport in the County and is the largest airport in the Delaware Valley. April 2008 Draft 86 ------- 1 2 3 4 5 6 7 8 9 10 The majority of NOx emissions in the NEI23 database attributable to airports in Philadelphia County are from non-road mobile sources, specifically ground support equipment. There is another airport in the County: Northeast Philadelphia Airport. However, because it serves general aviation, is generally much smaller in operations than PHL, and has little ground support equipment activity - which is associated primarily with commercial aviation - all airport emissions in the County were attributed to PHL. The PHL emissions were taken from the non- road section of the 2002 NEI, and are shown by Table 24. Table 24. Philadelphia International airport (PHL) NOx emissions. State and County Philadelphia, PA PHL Total scc 2265008005 2267008005 2270008005 2275020000 2275050000 NOx (tpy) 4.6 5.1 196.2 0.01 2.5 208.4 SCC Level 1 Description Mobile Sources Mobile Sources Mobile Sources Mobile Sources Mobile Sources SCC Level 3 Description Off-highway Vehicle Gasoline, 4- Stroke LPG Off-highway Vehicle Diesel Aircraft Aircraft SCC Level 6 Description Airport Ground Support Equipment Airport Ground Support Equipment Airport Ground Support Equipment Commercial Aircraft General Aviation SCC Level 8 Description Airport Ground Support Equipment Airport Ground Support Equipment Airport Ground Support Equipment Total: All Types Total 11 12 13 14 15 16 17 18 19 As with the fugitive sources discussed above, the airport emissions are best parameterized as area sources. The boundary of the area source was taken as the region of operation of baggage handling equipment, including the terminal building and the region surrounding the gates. This region was digitized into an 18-sided polygon of size 1,326,000 m2, and included in the AERMOD input control file. The activity profile for PHL was taken to have seasonal and hourly variation (SEASHR), based on values from the EMS-HAP model.24 These factors are disaggregated in the EMS-HAP model database based on source classification codes (SCCs), which were linked to those from 23 http://www.epa.gov/ttn/chief/net/2002inventory.html 24 EPA 2004, User's Guide for the Emissions Modeling System for Hazardous Air Pollutants (EMS-HAP) Version 3.0, EPA-454/B-03-006. April 2008 Draft 87 ------- 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 the NEI database. The EMS-HAP values provide hourly activity factors by season, day type, and hour; to compress to simple SEASHR modeling, the hourly values from the three individual day types were averaged together. The total emissions for each SCC were then disaggregated into seasonal and hourly components and the resulting components summed to create total PHL emissions for each hour of the four annual seasons. These parameterized emissions were then normalized to the total cargo handling operational area, to produce emission factors in units of grams per second per square meter and included in the AERMOD input file. Figure 2 also shows the location of the PHL area source. 7.4.4 Receptor Locations Three sets of receptors were chosen to represent the locations of interest. First, all NOX monitor locations, shown by Table 25, within the Philadelphia CMSA were included as receptor locations. Although all receptors are assumed to be on a flat plane, they are placed at the standard breathing height of 5.9 ft (1.8 m). Table 25. Philadelphia CMSA NOx monitors. CMSA Philadelphia- Wilmington- Atl antic City, PA-NJ-DE-MD Site ID 100031003 100031007 100032004 340070003 420170012 420450002 420910013 421010004 421010029 421010047 Latitude 39.7611 39.5511 39.7394 39.923 40.1072 39.8356 40.1122 40.0089 39.9572 39.9447 Longitude -75.4919 -75.7308 -75.5581 -75.0976 -74.8822 -75.3725 -75.3092 -75.0978 -75.1731 -75.1661 The second receptor locations were selected to represent the locations of census block centroids near major NOX sources. GIS analysis was used to determine all block centroids in Philadelphia County that lie within a 0.25 mile (400 m) of the roadway segments and also all block centroids that lie within 6.2 miles (10 km) of any major point source. 12,982 block centroids were selected due to their proximity to major roadways; 16,298 centroids were selected due to their proximity to major sources. The union of these sets produced 16,857 unique block April 2008 Draft ------- 1 centroid receptor locations, each of which was assigned a height of 5.9 ft (1.8 m). The location 2 of centroids that met either distance criteria and included in the modeling is shown by Figure 3. 3 4 6 7 8 9 10 11 12 13 14 V c«i*in«d Point sack ' Census Block centroldc within 400m of Major Roadway Edges OR wtthln 10hm of stacks Figure 3. Centroid locations within fixed distances to major point and mobile sources. The third set of receptors was chosen to represent the on-road microenvironment. For this set, one receptor was placed at the center of each of the 982 sources. The distance relationship between the road segments and block centroids can be estimated by looking at the distance between the road-centered and the block centroid receptors. Figure 4 shows the histogram of the shortest distance between each centroid receptor and its nearest roadway-centered receptor. April 2008 Draft 89 ------- OLOOLOOLOOLOOLOOLOOLOOLOOLO I Distance (m) 2 Figure 4. Frequency distribution of distance between each Census receptor and its nearest 3 road-centered receptor. 4 5 The centroids selected were those within 10 km of any major point source or 400 m from 6 any receptor edge, so the distances to the nearest major road segment can be significantly greater 7 than 400 m. The mode of the distribution is about 150 m and the median distance to the closest 8 roadway segment center is about 450 m. However, these values represent the distances of the 9 block centroids to road centers instead of road edges, so that they overestimate the actual 10 distances to the zone most influenced by roadway by an average of 14 m and a range of 4 m to 11 44 m (see Table 21 above). 12 7.4.5 Estimate Air Quality Concentrations 13 The hourly concentrations estimated from each of the three source categories were 14 combined at each receptor. Then a local concentration, reflecting the concentration contribution 15 from emission sources not included in the simulation, was added to the sum of the concentration 16 contributions from each of these sources at each receptor. The local concentration was estimated 17 from the difference between the model predictions at the local NC>2 monitors and the observed 18 values. It should be noted that this local concentration may also include any model error present 19 in estimating concentration at the local monitoring sites. Table 26 presents a summary of the 20 estimated local concentration added to the AERMOD hourly concentration data. 21 April 2008 Draft 90 ------- 1 Table 26. Comparison of ambient monitoring and AERMOD predicted NO2 2 concentrations. Year and Monitor ID Annual Average NO2 concentration (ppb) Monitor AERMOD Inititial Difference1 AERMOD Final2 2001 4210100043 4210100292 4210100471 mean 26 28 30 7 22 20 18 6 10 11 19 33 32 2002 4210100043 4210100292 4210100471 mean 24 28 29 7 21 19 17 7 10 11 18 32 31 2003 4210100043 4210100292 4210100471* mean 24 25 25 7 22 26 17 3 -1 6 13 28 32 1 the difference represents concentrations attributed to sources not modeled by AERMOD and model error. 2 the mean difference between measured and modeled was added uniformly at each receptor hourly concentration to generate the AERMOD final concentrations. * monitor did not meet completeness criteria used in the air quality characterization. 5 7.5 POPULATION MODELED 6 A detailed consideration of the population residing in each modeled area was included 7 where the exposure modeling was performed. The assessment included the general population 8 residing in each modeled area and susceptible subpopulations identified in the ISA. These 9 include population subgroups defined from either an exposure or health perspective. The 10 population subgroups identified by the ISA and that were modeled in the exposure assessment 11 include asthmatics of all ages and asthmatic children (ages 5-18). In addition to these population 12 subgroups, activities for those susceptible to potentially greater exposure to NO2 were 13 considered, including those commuting on roadways and persons residing near major roadways. 14 While the total population exposure was estimated, the focus of the analysis was on the 15 susceptible individuals. April 2008 Draft 91 ------- 1 7.5.1 Simulated Individuals 2 APEX takes population characteristics into account to develop accurate representations of 3 study area demographics. Population counts and employment probabilities by age and gender 4 are used to develop representative profiles of hypothetical individuals for the simulation. Block- 5 level population counts by age in one-year increments, from birth to 99 years, come from the 6 2000 Census of Population and Housing Summary File 1 (SF-1). This file contains the 100- 7 percent data, which is the information compiled from the questions asked of all people and about 8 every housing unit. The total population considered in this analysis was approximately 1.48 9 million persons, of which there a total simulated population of 163,000 asthmatics. The model 10 simulated approximately 281,000 children, of which there were about 48,000 asthmatics. Due to 11 random sampling, the actual number of specific subpopulations modeled will vary slightly by 12 year. 13 7.5.2 Employment Probabilities 14 Employment data from the 2000 Census provide employment probabilities for each 15 gender and specific age groups for every Census tract. The employment age groupings were: 16- 16 19, 20-21, 22-24, 25-29, 30-34, 35-44, 45-54, 55-59, 60-61, 62-64, 65-69, 70-74, and >75 years 17 of age. Children under the age of 16 are assigned employment probabilities of zero. 18 7.5.3 Commuting Patterns 19 To ensure that individuals' daily activities are accurately represented within APEX, it is 20 important to integrate working patterns into the assessment. Commuting data were originally 21 derived from the 2000 Census and were collected as part of the Census Transportation Planning 22 Package (CTPP) (US DOT, 2007). CTPP contains tabulations by place of residence, place of 23 work, and the flows between the residence and work. 24 It is assumed that all persons with home-to-work distances up to 120 km are daily 25 commuters, and that persons who travel further than 120 km do not commute daily. Therefore 26 the list of commuting destinations for each home tract is restricted to only those work tracts that 27 are within 120 km of the home tract. 28 APEX allows the user to specify how to handle individuals who commute to destinations 29 outside the study area. One option is to drop them from the simulation. If they are included, the 30 user specifies values for two additional parameters, called LM and LA (Multiplicative and April 2008 Draft 92 ------- 1 Additive factors for commuters who Leave the area). While a commuter is at work, if the 2 workplace is outside the study area, then the ambient concentration cannot be determined from 3 any air district (since districts are inside the study area). Instead, it is assumed to be related to 4 the average concentration CAVE® over all air districts at the time in question. The ambient 5 concentration outside the study area at time t, Courft), is estimated as: 6 7 C'OUT (t) = LM * CAVE ft) + LA 8 9 The microenvironmental concentration (for example, in an office outside the study area) 10 is determined from this ambient concentration by the same model (mass balance or factor) as 11 applied inside the study area. The parameters LM and LA were both set to zero for this modeling 12 analysis; thus, exposures to individuals are set to zero when they are outside of the study area. 13 Although this tends to underestimate exposures, it is a small effect and this was done since we 14 have not estimated ambient concentrations of NC>2 in counties outside of the modeled areas. 15 School age children did not have commuting to and from school. This results in the 16 implicit assumption that children attend a school with ambient NC>2 concentrations similar to 17 concentrations near their residence. 18 7.6 CONSTRUCTION OF LONGITUDINAL ACTIVITY SEQUENCES 19 Exposure models use human activity pattern data to predict and estimate exposure to 20 pollutants. Different human activities, such as spending time outdoors, indoors, or driving, will 21 result in varying pollutant exposure concentrations. To accurately model individuals and their 22 exposure to pollutants, it is critical to understand their daily activities. 23 The Consolidated Human Activity Database (CHAD) provides data for where people 24 spend time and the activities performed. CHAD was designed to provide a basis for conducting 25 multi-route, multi-media exposure assessments (McCurdy et al., 2000; EPA, 2002). Table 27 26 summarizes the studies in CHAD used in this modeling analysis, providing nearly 16,000 diary - 27 days of activity data (3,075 diary-days for ages 5-18) collected between 1982 and 1998. April 2008 Draft 93 ------- 1 2 Table 27. Studies in CHAD used for the exposure analysis. Study name Baltimore California Adolescents (GARB) California Adults (GARB) California Children (GARB) Cincinnati (EPRI) Denver (EPA) Los Angeles: Elementary School Los Angeles: High School National: NHAPS-Air National: NHAPS- Water Washington, D.C. (EPA) Total diary days Geographic coverage One building in Baltimore California California California Cincinnati metro, area Denver metro, area Los Angeles Los Angeles National National Wash., D.C. metro, area Study time period 01/1997-02/1997, 07/1998-08/1998 10/1987-09/1988 10/1987-09/1988 04/1989-02/1990 03/1985-04/1985, 08/1 985 11/1982-02/1983 10/1989 09/1990-10/1990 09/1992-10/1994 09/1992-10/1994 11/1982-02/1983 Subject ages 72-93 12- 17 18-94 <1 -11 <1 -86 18-70 10- 12 13-17 <1 -93 <1 -93 18-98 Diary- days 292 181 1,552 1,200 2,587 791 51 42 4,326 4,332 639 15,993 Diary-days (ages 5-1 8) 0 181 36 683 740 7 51 42 634 691 10 3,075 Diary type and study design Diary Recall; Random Recall; Random Recall; Random Diary; Random Diary; Random Diary Diary Recall; Random Recall; Random Diary; Random Reference Williams et al. (2000) Robinson et al. (1989), Wiley etal. (1991 a) Robinson etal. (1989), Wiley etal. (1991 a) Wiley etal. (1991b) Johnson (1989) Johnson (1984) Aklandetal. (1985) Spier etal. (1992) Spier etal. (1992) Klepeis et al. (1996), Tsang and Klepeis (1996) Klepeis et al. (1996), Tsang and Klepeis (1996) Hartwell etal. (1984), Aklandetal. (1985) April 2008 Draft 94 ------- 1 Typical time-activity pattern data available for inhalation exposure modeling consist of a 2 sequence of location/activity combinations spanning 24-hours, with 1 to 3 diary-days for any 3 single individual. Exposure modeling typically requires information on activity patterns over 4 longer periods of time, e.g., a full year. For example, even for pollutant health effects with short 5 averaging times (e.g., NC>2 1-hour average concentration) it may be desirable to know the 6 frequency of exceedances of a concentration over a long period of time (e.g., the annual number 7 of exceedances of a 1-hour average NC>2 concentration of 200 ppb for each simulated individual). 8 Long-term multi-day activity patterns can be estimated from single days by combining 9 the daily records in various ways, and the method used for combining them will influence the 10 variability of the long-term activity patterns across the simulated population. This in turn will 11 influence the ability of the model to accurately represent either long-term average high-end 12 exposures, or the number of individuals exposed multiple times to short-term high-end 13 concentrations. 14 A new algorithm has been developed and incorporated into APEX to represent the day- 15 to-day correlation of activities for individuals. The algorithms first use cluster analysis to divide 16 the daily activity pattern records into groups that are similar, and then select a single daily record 17 from each group. This limited number of daily patterns is then used to construct a long-term 18 sequence for a simulated individual, based on empirically-derived transition probabilities. This 19 approach is intermediate between the assumption of no day-to-day correlation (i.e., re-selection 20 of diaries for each time period) and perfect correlation (i.e., selection of a single daily record to 21 represent all days). 22 The steps in the algorithm are as follows: 23 1. For each demographic group (age, gender, employment status), temperature range, 24 and day-of-week combination, the associated time-activity records are partitioned into 25 3 groups using cluster analysis. The clustering criterion is a vector of 5 values: the 26 time spent in each of 5 microenvironment categories (indoors - residence; indoors - 27 other building; outdoors - near road; outdoors - away from road; in vehicle). 28 2. For each simulated individual, a single time-activity record is randomly selected from 29 each cluster. 30 3. A Markov process determines the probability of a given time-activity pattern 31 occurring on a given day based on the time-activity pattern of the previous day and April 2008 Draft 95 ------- 1 cluster-to-cluster transition probabilities. The cluster-to-cluster transition 2 probabilities are estimated from the available multi-day time-activity records. If 3 insufficient multi-day time-activity records are available for a demographic group, 4 season, day-of-week combination, then the cluster-to-cluster transition probabilities 5 are estimated from the frequency of time-activity records in each cluster in the CHAD 6 database. 7 Further details regarding the Cluster-Markov algorithm and supporting evaluations are 8 provided in Appendix F of the draft TSD. 9 7.7 CALCULATING MICROENVIRONMENTAL CONCENTRATIONS 10 Probabilistic algorithms are used to estimate the pollutant concentration associated with 11 each exposure event. The estimated pollutant concentrations account for temporal and spatial 12 variability in ambient (outdoor) pollutant concentration and factors affecting indoor 13 microenvironment, such as a penetration, air exchange rate, and pollutant decay or deposition 14 rate. APEX calculates air concentrations in the various microenvironments visited by the 15 simulated person by using the ambient air data estimated for the relevant blocks/receptors, the 16 user-specified algorithm, and input parameters specific to each microenvironment. The method 17 used by APEX to estimate the microenvironment depends on the microenvironment, the data 18 available for input to the algorithm, and the estimation method selected by the user. At this time, 19 APEX calculates hourly concentrations in all the microenvironments at each hour of the 20 simulation for each of the simulated individuals using one of two methods: by mass balance or a 21 transfer factors method. 22 The mass balance method simulates an enclosed microenvironment as a well-mixed 23 volume in which the air concentration is spatially uniform at any specific time. The 24 concentration of an air pollutant in such a microenvironment is estimated using the following 25 processes: 26 • Inflow of air into the microenvironment 27 • Outflow of air from the microenvironment 28 • Removal of a pollutant from the microenvironment due to deposition, filtration, and 29 chemical degradation 30 • Emissions from sources of a pollutant inside the microenvironment. April 2008 Draft 96 ------- 1 A transfer factors approach is simpler than the mass balance model, however, most 2 parameters are derived from distributions rather than single values to account for observed 3 variability. It does not calculate concentration in a microenvironment from the concentration in 4 the previous hour as is done by the mass balance method, and it has only two parameters. A 5 proximity factor is used to account for proximity of the microenvironment to sources or sinks of 6 pollution, or other systematic differences between concentrations just outside the 7 microenvironment and the ambient concentrations (at the measurements site or modeled 8 receptor). The second, a penetration factor, quantifies the amount of outdoor pollutant penetrates 9 into the microenvironment. 10 7.7.1 Microenvironments Modeled 11 In APEX, microenvironments represent the exposure locations for simulated individuals. 12 For exposures to be estimated accurately, it is important to have realistic microenvironments that 13 match closely to the locations where actual people spend time on a daily basis. As discussed 14 above, the two methods available in APEX for calculating pollutant levels within 15 microenvironments are: 1) factors and 2) mass balance. A list of microenvironments used in this 16 study, the calculation method used, and the type of parameters used to calculate the 17 microenvironment concentrations can be found in Table 26. April 2008 Draft 97 ------- 1 2 3 Table 28. List of microenvironments modeled and calculation methods used. 4 5 6 7 10 11 12 13 14 15 16 Microenvironment Indoors - Residence Indoors - Bars and restaurants Indoors - Schools Indoors - Day-care centers Indoors - Office Indoors- Shopping Indoors - Other Outdoors - Near road Outdoors - Public garage - parking lot Outdoors - Other In-vehicle - Cars and Trucks In-vehicle - Mass Transit (bus, subway, train) Calculation Method Mass balance Mass balance Mass balance Mass balance Mass balance Mass balance Mass balance Factors Factors Factors Factors Factors Parameter Types used1 AER and DE AER and DE AER and DE AER and DE AER and DE AER and DE AER and DE PR PR None PE and PR PE and PR 1 AER=air exchange rate, DE=decay-deposition rate, PR=proximity factor, PE=penetration factor 7.7.2 Microenvironment Descriptions 7.7.2.1 Microenvironment 1: Indoor-Residence The Indoor-Residence microenvironment uses several variables that affect NC>2 exposure: whether or not air conditioning is present, the average outdoor temperature, the NC>2 removal rate, and an indoor concentration source. Air conditioning prevalence rates Since the selection of an air exchange rate distribution is conditioned on the presence or absence of an air-conditioner, for each modeled area the air conditioning status of the residential microenvironments is simulated randomly using the probability that a residence has an air conditioner. For this study, location-specific air conditioning prevalence was calculated using the data and survey weights from the American Housing Survey of 2003 (AHS, 2003a; 2003b). Table 29 contains the values for air conditioning prevalence used for each modeled location. April 2008 Draft 98 ------- 1 2 Table 29. Air conditioning (A/C) prevalence estimates with 95% confidence intervals. AHS Survey Philadelphia Atlanta Detroit Los Angeles Phoenix Housing Units 1,943,492 797,687 1,877,178 3,296,819 - A/C Prevalence (%) 90.6 97.0 81.4 55.1 - se 1.3 1.2 1.8 1.7 - L95 88.1 94.7 78.0 51.7 - U95 93.2 99.3 84.9 58.4 - Notes: se - Standard error L95 - Lower limit on 95th confidence interval U95 - Upper limit on 95th confidence interval 4 5 Air exchange rates 6 Air exchange rate data for the indoor residential microenvironment were obtained from 7 EPA (2007g). Briefly, data were reviewed, compiled and evaluated from the extant literature to 8 generate location-specific AER distributions categorized by influential factors, namely 9 temperature and presence of air conditioning. In general, lognormal distributions provided the 10 best fit, and are defined by a geometric mean (GM) and standard deviation (GSD). To avoid 11 unusually extreme simulated AER values, bounds of 0.1 and 10 were selected for minimum and 12 maximum AER, respectively. 13 Fitted distributions were available for one of the cities modeled in this assessment, Los 14 Angeles. For the other four of the locations to be modeled, a distribution was selected from one 15 of the other locations thought to have similar characteristics to the city to be modeled, 16 qualitatively considering factors that might influence AERs. These factors include the age 17 composition of housing stock, construction methods, and other meteorological variables not 18 explicitly treated in the analysis, such as humidity and wind speed patterns. The distributions 19 used for each of the modeled locations are provided in Table 30. April 2008 Draft 99 ------- 1 2 3 4 Table 30. Geometric means (GM) and standard deviations (GSD) for air exchange rates by city, A/C type, and temperature range. Area Modeled Los Angeles Philadelphia and Detroit Atlanta (No A/C) Atlanta (A/C) Study City Houston Inland California Los Angeles New York City Outside California Research Triangle Park, NC A/C Type Central or Room A/C No A/C Central or Room A/C No A/C Central or Room A/C No A/C Central or Room A/C No A/C Central or Room A/C No A/C Central or Room A/C Temp (°C) <=20 20-25 25-30 >30 <=10 10-20 >20 <=25 >25 <=10 10-20 20-25 >25 <=20 20-25 25-30 >30 <=10 10-20 20-25 >25 <=10 10-25 >25 <=10 10-20 >20 <=10 10-20 20-25 25-30 >30 <=10 10-20 >20 <=10 10-20 20-25 >25 N 15 20 65 14 13 28 12 226 83 17 52 13 14 721 273 102 12 18 390 148 25 20 42 19 48 59 32 179 338 253 219 24 61 87 44 157 320 196 145 GM 0.4075 0.4675 0.4221 0.4989 0.6557 0.6254 0.9161 0.5033 0.8299 0.5256 0.6649 1.0536 0.8271 0.5894 1.1003 0.8128 0.2664 0.5427 0.7470 1.3718 0.9884 0.7108 1.1392 1 .2435 1.0165 0.7909 1 .6062 0.9185 0.5636 0.4676 0.4235 0.5667 0.9258 0.7333 1.3782 0.9617 0.5624 0.3970 0.3803 GSD 2.1135 1.9381 2.2579 1.7174 1 .6794 2.9162 2.4512 1.9210 2.3534 3.1920 2.1743 1.7110 2.2646 1.8948 2.3648 2.4151 2.7899 3.0872 2.0852 2.2828 1 .9666 2.0184 2.6773 2.1768 2.1382 2.0417 2.1189 1.8589 1.9396 2.2011 2.0373 1 .9447 2.0836 2.3299 2.2757 1.8094 1.9058 1.8887 1.7092 5 6 7 8 7V02 removal rate For this analysis, the same NC>2 removal rate distribution was used for all microenvironments that use the mass balance method. This removal rate is based on data April 2008 Draft 100 ------- 1 provided by Spicer et al. (1993). A total of 6 experiments, under variable source emission 2 characteristics including operation of gas stove, were conducted in an unoccupied test house. A 3 statistical distribution could not be described with the limited data, therefore a uniform 4 distribution was approximated by the bounds of the 6 values, a minimum of 1.02 and a maximum 5 ofl.4511'1. 6 Indoor source contributions 1 A number of studies, as described in section 2.5.5 of the NOX ISA, have noted the 8 importance of gas cooking appliances as sources of NCh emissions. An indoor emission source 9 term was included in the APEX simulations to estimate NO2 exposure to gas cooking (hereafter 10 referred to as "indoor sources"). Three types of data were used to implement this factor: 11 • The fraction of households in the Philadelphia MSA that use gas for cooking fuel 12 • The range of contributions to indoor NO2 concentrations that occur from cooking 13 with gas 14 • The diurnal pattern of cooking in households. 15 Households using gas for cooking fuel. The fraction of households in Philadelphia 16 County that use gas cooking fuel (i.e., 55%) was taken from the US Census Bureau's American 17 Housing Survey for the Philadelphia Metropolitan Area: 2003. 18 Concentration Contributions. Data used for estimating the contribution to indoor NO2 19 concentrations that occur during cooking with gas fuel were derived from a study sponsored by 20 the California Air Resources Board (CARB, 2001). For this study a test house was set up for 21 continuous measurements of NO2 indoors and outdoors, among several other parameters, and 22 conducted under several different cooking procedures and stove operating conditions. 23 A uniform distribution of concentration contributions for input to APEX was estimated as 24 follows. 25 • The concurrent outdoor NO2 concentration measurement was subtracted from each 26 indoor concentration measurement, to yield net indoor concentrations 27 • Net indoor concentrations for duplicate cooking tests (same food cooked the same 28 way) were averaged for each indoor room, to yield average net indoor concentrations 29 • The minimum and maximum average net indoor concentrations for any test in any 30 room were used as the lower and upper bounds of a uniform distribution April 2008 Draft 101 ------- 1 This resulted in a minimum average net indoor concentration of 4 ppb and a maximum 2 net average indoor concentration of 188 ppb. 3 Diurnal Pattern of Cooking Events. An analysis by Johnson et al (1999) of survey data 4 on gas stove usage collected by Koontz et al (1992) showed an average number of meals 5 prepared each day with a gas stove of 1.4. The diurnal allocation of these cooking events was 6 estimated as follows. 7 • Food preparation time obtained from CHAD diaries was stratified by hour of the day, 8 and summed for each hour, and summed for total preparation time. 9 • The fraction of food preparation occurring in each hour of the day was calculated as 10 the total number of minutes for that hour divided by the overall total preparation time. 11 The result was a measure of the probability of food preparation taking place during 12 any hour, given one food preparation event per day. 13 • Each hourly fraction was multiplied by 1.4, to normalize the expected value of daily 14 food preparation events to 1.4. 15 The estimated probabilities of cooking by hour of the day are presented in Table 31. 16 For this analysis it was assumed that the probability that food preparation would include stove 17 usage was the same for each hour of the day, so that the diurnal allocation of food preparation 18 events would be the same as the diurnal allocation of gas stove usage. It was also assumed that 19 each cooking event lasts for exactly 1 hour, implying that the average total daily gas stove usage 20 is 1.4 hours. April 2008 Draft 102 ------- 2 3 Table 31. Probability of gas stove cooking by hour of the day. 4 5 6 7 8 9 10 11 12 13 14 Hour of Day 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Probability of Cooking (%)1 0 0 0 0 0 5 10 10 10 5 5 5 10 5 5 5 15 20 15 10 5 5 0 0 Values rounded to the nearest 5%. Data sum to 1 45% due to rounding and scaling to 1 .4 cooking events/day. 7.7.2.2 Microenvironments 2-7: All Other Indoor Microenvironments The remaining five indoor microenvironments, which represent Bars and Restaurants, Schools, Day Care Centers, Office, Shopping, and Other environments, were all modeled using the same data and functions. As with the Indoor-Residence microenvironment, these microenvironments use both AER and removal rates to calculate exposures within the microenvironment. The air exchange rate distribution (GM = 1.109, GSD = 3.015, Min = 0.07, Max = 13.8) was developed based on an indoor air quality study (Persily et al, 2005; see EPA, 2007g for details in derivation). The decay rate is the same as used in the Indoor-Residence microenvironment discussed previously. The Bars and Restaurants microenvironment included an estimated contribution from indoor sources as was described for the Indoor-Residence, only April 2008 Draft 103 ------- 1 there was an assumed 100% prevalence rate and the cooking with the gas appliance occurred at 2 any hour of the day. 3 7.7.2.3 Microenvironments 8 and 9: Outdoor Microenvironments 4 Two outdoor microenvironments, the Near Road and Public Garage/Parking Lot, used the 5 transfer factors method to calculate pollutant exposure. Penetration factors are not applicable to 6 outdoor environments (effectively, PEN=1). The distribution for proximity factors were 7 developed from the dispersion model estimated concentrations, using the relationship between 8 the on-road to receptor estimated concentrations. 9 7.7.2.4Microenvironment 10: Outdoors-General 10 The general outdoor environment concentrations are well represented by the modeled 11 concentrations. Therefore, both the penetration factor and proximity factor for this 12 microenvironment were set to 1. 13 7.7.2.5 Microenvironments 11 and 12: In Vehicle- Cars and Trucks, and Mass Transit 14 Penetration factors were developed from data provided in Chan and Chung (2003). 15 Inside-vehicle and outdoor NC>2 concentrations were measured with for three ventilation 16 conditions, air-recirculation, fresh air intake, and with windows opened. Since major roads were 17 the focus of this assessment, reported indoor/outdoor ratios for highway and urban streets were 18 used here. Mean values range from about 0.6 to just over 1.0, with higher values associated with 19 increased ventilation (i.e., window open). A uniform distribution was selected for the 20 penetration factor for Inside-Cars/Trucks (ranging from 0.6 to 1.0) due to the limited data 21 available to describe a more formal distribution and the lack of data available to reasonably 22 assign potentially influential characteristics such as use of vehicle ventilation systems for each 23 location. Mass transit systems, due to the frequent opening and closing of doors, was assigned a 24 point estimate of 1.0 based on the reported mean values for open windows ranging from 0.96 and 25 1.0. Proximity factors were developed from the dispersion model estimated concentrations, 26 using the relationship between the on-road to receptor estimated concentrations. April 2008 Draft 104 ------- 1 7.8 EXPOSURE AND HEALTH RISK CALCULATIONS 2 APEX calculates the time series of exposure concentrations that a simulated individual 3 experiences during the simulation period. APEX determines the exposure using hourly ambient 4 air concentrations, calculated concentrations in each microenvironment based on these ambient 5 air concentrations (and indoor sources if present), and the minutes spent in a sequence of 6 microenvironments visited according to the composite diary. The hourly exposure concentration 7 at any clock hour during the simulation period is determined using the following equation: N Z^~< hourly-mean , ^•ME(j) T (j) 9 where, 10 d = Hourly exposure concentration at clock hour / of the simulation period 11 (ppb) 12 N = Number of events (i.e., microenvironments visited) in clock hour /' of 13 the simulation period. 14 C^j™"" = Hourly mean concentration in microenvironmenty (ppm) 15 t(j) = Time spent in microenvironmenty (minutes) 16 T =60 minutes 17 18 From the hourly exposures, APEX calculates time series of 1-hour average exposure 19 concentrations that a simulated individual would experience during the simulation period. 20 APEX then statistically summarizes and tabulates the hourly (or daily, annual average) 21 exposures. In this analysis, the exposure indicator is 1-hr exposures above selected health effect 22 benchmark levels. From this, APEX can calculate two general types of exposure estimates: 23 counts of the estimated number of people exposed to a specified NC>2 concentration level and the 24 number of times per year that they are so exposed; the latter metric is in terms of person- 25 occurrences or person-days. The former highlights the number of individuals exposed at least 26 one or more times per modeling period to the health effect benchmark level of interest. APEX 27 can also report counts of individuals with multiple exposures. This person-occurrences measure April 2008 Draft 105 ------- 1 estimates the number of times per season that individuals are exposed to the exposure indicator 2 of interest and then accumulates these estimates for the entire population residing in an area. 3 APEX tabulates and displays the two measures for exposures above levels ranging from 4 200 to 300 ppb by 50 ppb increments for 1-hour average exposures. These results are tabulated 5 for the population and subpopulations of interest. 6 To simulate just meeting the current standard, dispersion modeled concentration were not 7 rolled-up as done in the air quality characterization. A proportional approach was used as done 8 in the Air Quality Characterization, but to reduce processing time, the potential health effect 9 benchmark levels were proportionally reduced by the similar factors described for each specific 10 location and simulated year. Since it is a proportional adjustment, the end effect of adjusting 11 concentrations upwards versus adjusting benchmark levels downward within the model is the 12 same. The difference in the exposure and risk modeling was that the modeled air quality 13 concentrations were used to generate the adjustment factors. Table 32 provides the adjustment 14 factors used and the adjusted potential health effect benchmark concentrations to simulate just 15 meeting the current standard. When modeling indoor sources, the indoor concentration 16 contributions needed to be scaled downward by the same proportions. April 2008 Draft 106 ------- 2 3 4 Table 32. Adjustment factors and potential health effect benchmark levels used by APEX to simulate just meeting the current standard. Simulated Year (factor) 2001 (1.59) 2002 (1.63) 2003 (1.64) Potential Health Effect Benchmark Level (ppb) Actual 150 200 250 300 150 200 250 300 150 200 250 300 Adjusted 94 126 157 189 92 122 153 184 91 122 152 183 6 7.9 EXPOSURE MODELING AND HEALTH RISK 7 CHARACTERIZATION RESULTS 8 7.9.1 Overview 9 The results of the exposure and risk characterization are presented here for Philadelphia 10 County. Several scenarios were considered for the exposure assessment, including two 11 averaging time for NC>2 concentrations (annual and 1-hour), inclusion of indoor sources, and for 12 evaluating just meeting the current standard. To date, year 2002 served as the base year for all 13 scenarios, years 2001 and 2003 were only evaluated for a limited number of scenarios. 14 Exposures were simulated for four groups; children and all persons, and the asthmatic population 15 within each of these. 16 The exposure results summarized below focus on the population group where exposure 17 estimations are of greatest interest, namely asthmatic individuals. However, due to certain 18 limitations in the data summaries output from APEX, some exposure data could only be output 19 for the entire population modeled (i.e., all persons - includes asthmatics and healthy persons of April 2008 Draft 107 ------- 1 all ages). The summary data for the entire population (e.g., annual average exposure 2 concentrations, time spent in microenvironments at or above a potential health effect benchmark 3 level) can be representative of the asthmatic population since the asthmatic population does not 4 have its microenvironmental concentrations and activities estimated any differently from those of 5 the total population. 6 7.9.2 Annual Average Exposure Concentrations (as is) 7 Since the current NO2 standard is 0.053 ppm annual average, the predicted air quality 8 concentrations, the measured ambient monitoring concentrations, and the estimated exposures 9 were summarized by annual average concentration. The distribution for the AERMOD predicted 10 NO2 concentrations at each of the 16,857 receptors for years 2001 through 2003 are illustrated in 11 Figure 5. Variable concentrations were estimated by the dispersion model over the three year 12 period (2001-2003). The NO2 concentration distribution was similar for years 2001 and 2002, 13 with mean annual average concentrations of about 21 ppb and a COV of just over 30%. On 14 average, NO2 annual average concentrations were lowest during simulated year 2003 (mean 15 annual average concentration was about 16 ppb), largely a result of the comparably lower local 16 concentration added (Table 26). While the mean annual average concentrations were lower than 17 those estimated for 2001 and 2002, a greater number of annual average concentrations were 18 estimated above 53 ppb for year 2003. In addition, year 2003 also contained greater variability 19 in annual average concentrations as indicated by a COV of 53%. April 2008 Draft 108 ------- 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 200 "o. 180 o o 140 - 120 - 0) 0) 100 - ro | 80 c •5 1! Q O OL UJ 40 - 20 - -5-4-3-2-1012345 Normal Quantile Figure 5. Distribution of AERMOD predicted annual average NOi concentrations at each of the 16,857 receptors in Philadelphia County for years 2001-2003. The hourly concentrations output from AERMOD were input into the exposure model, providing a range of estimated exposures output by APEX. Figure 6 illustrates the annual average exposure concentrations for the entire simulated population (both asthmatics and healthy individual of all ages), for each of the years analyzed and where indoor sources were modeled. While years 2001 and 2002 contained very similar population exposure concentration distributions, the modeled year 2003 contained about 20% lower annual average concentrations. The lower exposure concentrations for year 2003 are similar to what was observed for the predicted air quality (Figure 5), however, all persons were estimated to contain exposures below an annual average concentration of 53 ppb, even considering indoor source concentration contributions. Again, while the figure summarizes the entire population, the data are representative of what would be observed for the population of asthmatics or asthmatic children. April 2008 Draft 109 ------- 100 90 - 1 2 3 4 5 6 9 10 11 12 13 14 15 16 17 2001 with indoor sources 2002 with indoor sources 2003 with indoor sources 15 20 25 30 Annual Average NO2 Exposure (ppb) Figure 6. Estimated annual average total NOi exposure concentrations for all simulated persons in Philadelphia County, using modeled 2001-2003 air quality (as is), with modeled indoor sources. The AERMOD predicted air quality and the estimated exposures for year 2002 were compared using their respective annual average NCh concentrations (Figure 7). As a point of reference, the annual average concentration for 2002 ambient monitors ranged from 24 ppb to 29 ppb. Many of the AERMOD predicted annual average concentrations were below that of the lowest ambient monitoring concentration of 24 ppb, although a few of the receptors contained concentrations above the highest measured annual average concentration . Estimated exposure concentrations were below that of both the modeled and measured air quality. For example, exposure concentrations were about 5 ppb less than the modeled air quality when the exposure estimation included indoor sources, and about 10 ppb less for when exposures were estimated without indoor sources. In comparing the estimated exposures with and without indoor sources, indoor sources were estimated to contribute between 1 and 5 ppb to the total annual average exposures. April 2008 Draft 110 ------- 100 90 - 80 - 70 - 60 - 01 Q_ 50 - 40 - 30 - 20 - 10 - AERMOD Predicted 2002 air quality (as is) APEX Exposure 2002 no indoor sources APEX Exposure 2002 with indoor sources 10 15 20 25 30 Annual Average NO2 Cocentrations (ppb) 35 40 45 2 Figure 7. Comparison of AERMOD predicted and ambient monitoring annual average 3 NOi concentrations (as is) and APEX exposure concentrations (with and without 4 modeled indoor sources) in Philadelphia County for year 2002. 5 7.9.3 One-Hour Exposures (as is) 6 Since there is interest in short-term exposures, a few analyses were performed using the 7 APEX estimated exposure concentrations. As part of the standard analysis, APEX reports the 8 maximum exposure concentration for each simulated individual in the simulated population. 9 This can provide insight into the proportion of the population experiencing any NC>2 exposure 10 concentration level of interest. In addition, exposures are estimated for each of the selected 11 potential health effect benchmark levels (200, 250, and 300 ppb, 1-hour average). An 12 exceedance was recorded when the maximum exposure concentration observed for the individual 13 was above the selected level in a day (therefore, the maximum number of exceedances is 365 for 14 a single person). Estimates of repeated exposures are also recorded, that is where 1-hour 15 exposure concentrations were above a selected level in a day added together across multiple days 16 (therefore, the maximum number of multiple exceedances is also 365). Persons of interest in this 17 exposure analysis are those with particular susceptibility to NC>2 exposure, namely individuals 18 with asthma. The health effect benchmark levels are appropriate for estimating the potential risk April 2008 Draft 111 ------- 1 of adverse health effects for asthmatics. The majority of the results presented in this section are 2 for the simulated asthmatic population. However, the exposure analysis was performed for the 3 total population to assess numbers of persons exposed to these levels and to provide additional 4 information relevant to the asthmatic population (such as time spent in particular 5 microenvironments). 6 7.9.3.1 Maximum Estimated Exposure Concentrations 1 A greater variability was observed in maximum exposure concentrations for the 2003 8 year simulation compared with years 2001 and 2002 (Figure 8). While annual average exposure 9 concentrations for the total population were the lowest of the 3-year simulation, year 2003 10 contained a greater number of individual maximum exposures at and above the lowest potential 11 health effect benchmark level. When indoor sources are not modeled however, over 90% of the 12 simulated persons do not have an occurrence of a 1-hour exposure above 200 ppb in a year. 13 7.9.3.2 Number of Estimated Exposures above Selected Levels 14 When considering the total asthmatic population simulated in Philadelphia County and using 15 current air quality of 2001-2003, nearly 50,000 persons were estimated to be exposed at least one 16 time to a one-hour concentration of 200 ppb in a year (Figure 9). These exposures include both 17 the NC>2 of ambient origin and that contributed by indoor sources. The number of asthmatics 18 exposed to greater concentrations (e.g., 250 or 300 ppb) drops dramatically and is estimated to be 19 somewhere between 1,000 - 15,000 depending on the 1-hour concentration level and the year of 20 air quality data used. Exposures simulated for year 2003 contained the greatest number of 21 asthmatics exposed in a year consistently for all potential health effect benchmark levels, while 22 year 2002 contained the lowest number of asthmatics. Similar trends across the benchmark 23 levels and the simulation years were observed for asthmatic children, albeit with lower numbers 24 of asthmatic children with exposures at or above the potential health effect benchmark levels. 25 For example, nearly 12,000 were estimated to be exposed to at least a one-hour NC>2 26 concentration of 200 ppb in a year (Figure 10). Additional exposure estimates were April 2008 Draft 112 ------- 1 2 3 4 5 6 7 100 90 80 70 g 60 8. 50 g. 40 8. 30 20 10 * 2001 with indoor sources o 2002 with indoor sources A 2003 with indoor sources x 2002 no indoor sources 50 100 150 200 250 300 350 Maximum 1-hour Exposure (ppb) in a Year 400 450 500 Figure 8. Estimated maximum NOi exposure concentration for all simulated persons in Philadelphia County, using modeled 2001-2003 air quality (as is), with and without modeled indoor sources. Values above the 99th percentile are not shown. April 2008 Draft 113 ------- 6.0E+4 200 Potential Health Effect Benchmark Level (ppb) 300 2003 AQ (as is) - with indoor souces 2002 AQ (as is) - with indoor sources 2001 AQ (as is) - with indoor sources Simulated Year - Scenario 1 2 3 4 5 Figure 9. Estimated number of all simulated asthmatics in Philadelphia County with at least one NOi exposure at or above the potential health effect benchmark levels, using modeled 2001-2003 air quality (as is), with modeled indoor sources. g 8 1.2E+4- d Number of Asthmatic Child to Selected 1-hour Level In a ' i. 0) 03 r1 P ? ? | - " ? t . I! m ,x 2.0E+3 O.OE+0 ^ ,,--' ;' lx k ^^ ^ — • ' — — — _ -^ L * — , — ?i — - _ 6 -^i " = ---- ? 200 250 300 Potential Health Effect Benchmark Level (ppb) 2003 AQ (as is) - with indoor souces 2002 AQ (as is) - with indoor sources 2001 AQ (as is) - with indoor sources Simulated Year - Scenario 6 1 Figure 10. Estimated number of simulated asthmatic children in Philadelphia County with at least one NOi exposure at or above the potential health effect benchmark levels, using modeled 2001-2003 air quality (as is), with modeled indoor sources. April 2008 Draft 114 ------- 5.0E+4- 200 Potential Health Effect Benchmark Level (ppb) 300 2002 AQ (as \s) - with indoor sources 2002 AQ (as is) - no indoor sources Simulated Year - Scenario 1 2 3 4 5 6 9 10 11 12 13 14 15 16 17 18 19 20 Figure 11. Comparison of the estimated number of all simulated asthmatics in Philadelphia County with at least one NOi exposure at or above potential health effect benchmark levels, using modeled 2002 air quality (as is) , with and without modeled indoor sources. generated using the modeled 2002 air quality (as is) and where the contribution from indoor sources was not included in the exposure concentrations. APEX allows for the same persons to be simulated, i.e., demographics of the population were conserved, as well as using the same individual time-location-activity profiles generated for each person. Figure 11 compares the estimated number of asthmatics experiencing exposures above the potential health effect benchmarks, both with indoor sources and without indoor sources included in the model runs. The number of asthmatics at or above the selected concentrations is reduced by between 50-80%, depending on benchmark level, when not including indoor source (i.e., gas cooking) concentration contributions. An evaluation of the time spent in the 12 microenvironments was performed to estimate where simulated individuals are exposed to concentrations above the potential health effect benchmark levels. Currently, the output generated by APEX is limited to compiling the microenvironmental time for the total population (includes both asthmatic individuals and healthy persons) and is summarized to the total time spent above the selected potential health April 2008 Draft 115 ------- 1 effect benchmark levels. As mentioned above, the data still provide a reasonable approximation 2 for each of the population subgroups (e.g., asthmatics or asthmatic children) since their 3 microenvironmental concentrations and activities are not estimated any differently from those of 4 the total population by APEX. 5 As an example, Figure 12 (a, b, c) summarizes the percent of total time spent in each 6 microenvironment for simulation year 2002 that was associated with estimated exposure 7 concentrations at or above 200, 250, and 300 ppb (results for years 2001 and 2003 were similar). 8 Estimated exposures included the contribution from one major category of indoor sources (i.e., 9 gas cooking). The time spent in the indoor residence and bars/restaurants were the most 10 important for concentrations >200 ppb, contributing to approximately 75% of the time persons 11 were exposed (Figure 12a). This is likely a result of the indoor source concentration contribution 12 to each individual's exposure concentrations. The importance of the particular 13 microenvironment however changes with differing potential health effect benchmark levels. 14 This is evident when considering the in-vehicle and outdoor near-road microenvironments, 15 progressing from about 19% of the time exposures were at the lowest potential health effect 16 benchmark level (200 ppb) to a high of 64% of the time exposures were at the highest 17 benchmark level (300 ppb, Figure 12c). 18 The microenvironments where higher exposure concentrations occur were also evaluated 19 for the exposure estimates generated without indoor source contributions. Figure 13 illustrates 20 that the time spent in the indoor microenvironments contributes little to the estimated exposures 21 above the selected benchmark levels. The contribution of these microenvironments varied only 22 slightly with increasing benchmark concentration, ranging from about 2-5%. Most of the time 23 associated with high exposures was associated with the transportation microenvironments (In- 24 Vehicle or In-Public Transport) or outdoors (Out-Near Road, Out-Parking Lot, Out-Other). The 25 April 2008 Draft 116 ------- In-Public Trans \Other Out-Other Out-Parking Lot Out-Near Road In-Other In-Shopping—7 In-Office-/ In-Day Care—' In-School—' In-Bar & Restaurant In-Residence a) > 200 ppb In-Public Trans k Other In-Residence Out-Other Out-Parking Lot In-Bar & Restaurant b) > 250 ppb In-Bar & Restaurant In-Public Trans |n.Residence J Hn-School Out-Other Out-Near Road Out-Parking Lot c) > 300 ppb 2 Figure 12. Fraction of time all simulated persons in Philadelphia County spend in the 3 twelve microenvironments associated with the potential NO2 health effect 4 benchmark levels, a) > 200 ppb, b) > 250 ppb, and c) > 300 ppb, year 2002 5 simulation with indoor sources. April 2008 Draft 117 ------- 1 2 3 4 5 In-Bar& Restaurant In-Residence- Other In-Public Trans— In-Vehicle Out-Near Road Out-Parking Lot Out-Other a) > 200 ppb In-Bar& Restaurant /—In-School In-Vehicle Out-Near Road Out-Other Out-Parking Lot b) > 250 ppb In-Bar& Restaurant,-In-School In-Residence^k //~tn-°al9.are In-Public In-Vehicle Out-Near Road Out-Other Out-Parking Lot c) > 300 ppb Figure 13. Fraction of time all simulated persons in Philadelphia County spend in the twelve microenvironments associated with the potential NOi health effect benchmark levels, a) > 200 ppb, b) > 250 ppb, and c) > 300 ppb, year 2002 simulation without indoor sources. April 2008 Draft 118 ------- 1 importance of time spent outdoors near roadways exhibited the greatest change in contribution 2 with increased health benchmark level, increasing from around 30 to 44% of time associated 3 with concentrations of 200 and 300 ppb, respectively. 4 7.9.3.3 Number of Repeated Exposures Above Selected Levels 5 In the analysis of persons exposed, the results show the number or percent of those with 6 at least one exposure at or above the selected potential health effect benchmark level. Given that 7 the benchmark is for a small averaging time (i.e., one-hour) it may be possible that individuals 8 are exposed to concentrations at or above the potential health effect benchmark levels more than 9 once in a given year. Since APEX simulates the longitudinal diary profile for each individual, 10 the number of times above a selected level is retained for each person. Figure 14 presents such 11 an analysis for the year 2003, the year containing the greatest number of exposure concentrations 12 at or above the selected benchmarks. Estimated exposures include both those resulting from 13 exposures to NC>2 of ambient origin and those resulting from indoor source NC>2 contributions. 14 While a large fraction of individuals experience at least one exposure to 200 ppb or greater over 15 a 1-hour time period in a year (about 32 percent), only around 14 percent were estimated to 16 contain at least 2 exposures. Multiple exposures at or above the selected benchmarks greater 17 than or equal to 3 or more times per year are even less frequent, with around 5 percent or less of 18 asthmatics exposed to 1-hour concentrations greater than or equal to 200 ppb 3 or more times in 19 a year. 20 Exposure estimates for year 2002 are presented to provide an additional perspective, 21 including a lower bound of repeated exposures for this population subgroup and for exposure 22 estimates generated with and without modeled indoor sources (Figure 15). Most asthmatics 23 exposed to a 200 ppb concentration are exposed once per year and only around 11 percent would 24 experience 2 or more exposures at or above 200 ppb when including indoor source contributions. 25 The percent of asthmatics experiencing multiple exposures a and abovet 250 and 300 ppb is 26 much lower, typically less than 1 percent of all asthmatics are exposed at the higher potential 27 benchmark levels. Also provided in Figure 14 are the percent of asthmatics exposed to selected 28 levels in the absence of indoor sources. Again, without the indoor source contribution, there are 29 reduced occurrences of multiple exposures at all of the potential health effect benchmark levels 30 compared with when indoor sources were modeled. April 2008 Draft 119 ------- 1 2 Estimated Number of Repeated Exposures in a Year 3 4 5 6 7 Potential Health Effect Benchmark Level (ppb) Figure 14. Estimated percent of all asthmatics in Philadelphia County with repeated NO2 exposures above potential health effect benchmark levels, using 2003 modeled air quality (as is), with modeled indoor sources. Estimated Number of Repeated Exposures in a Year Health Effect Benchmark Level (ppb) April 2008 Draft 120 ------- 1 Figure 15. Estimated percent of all asthmatics in Philadelphia County with repeated NO2 2 exposures above potential health effect benchmark levels, using modeled 2002 air quality (as is), with and without indoor sources. 4 7.9.4 One-Hour Exposures Associated with Just Meeting the Current Standard 5 To simulate just meeting the current NC>2 standard, the potential health effect 6 benchmark level was adjusted in the exposure model, rather than adjusting all of the hourly 7 concentrations for each receptor and year simulated (see section 5.4.2 and section 7.8 above). 8 Similar estimates of short-term exposures (i.e., 1-hour) were generated for the total population 9 and population subgroups of interest (i.e., asthmatics and asthmatic children). 10 7.9.4.1 Number of Estimated Exposures above Selected Levels 11 In considering exposures estimated to occur associated with air quality simulated to just 12 meet the current annual average NC>2 standard, the number of persons experiencing 13 concentrations at or above the potential health effect benchmarks increased. To allow for 14 reasonable comparison, the number of persons affected considering each scenario is expressed as 15 the percent of the subpopulation of interest. Figure 16 illustrates the percent of asthmatics 16 estimated to experience at least one exposure at or above the selected potential health effect 17 benchmark concentrations, with just meeting the current standard and including indoor source 1 8 contributions. While it was estimated that about 30% percent of asthmatics would be exposed to 19 200 ppb (1-hour average) at least once in a year for as is air quality, it was estimated that around 20 80 percent of asthmatics would experience at least one concentration above the lowest potential 21 health effect benchmark level in a year representing just meeting the current standard. Again, 22 estimates for asthmatic children exhibited a similar trend, with between 75 to 80 percent exposed 23 to a concentration at or above the lowest potential health effect benchmark level at least once per 24 year for a year just meeting the current standard (data not shown). The percent of all asthmatics 25 experiencing the higher benchmark levels is reduced to between 3 1 and 45 percent for the 250 26 ppb, 1-hour benchmark, and between 10 and 24 percent for the 300 ppb, 1-hour benchmark level 27 associated with air quality representing just meeting the current annual average standard. 28 April 2008 Draft 121 ------- 90 1 2 3 4 5 6 7 8 9 10 11 12 13 200 Potential Health Effect Benchmark Level (ppb) 2003 AQ (std) - with indoor soucrces 2002 AQ (std) - with indoor soucrces 2001 AQ (std) - with indoor soucrces Simulated Year - Scenario Figure 16. Estimated percent of all asthmatics in Philadelphia with at least one exposure at or above the potential health effect benchmark level, using modeled 2001-2003 air quality just meeting the current standard, with modeled indoor sources. In evaluating the influence of indoor source contribution for the scenario just meeting the current standard, the numbers of individuals exposed at selected levels are reduced without indoor sources, ranging from about 26 percent lower for the 200 ppb level to around 11 percent for the 300 ppb level when compared with exposure estimates that accounted for indoor sources (Figure 17). April 2008 Draft 122 ------- I.E! 300 Potential Health Effect Benchmark Level (ppb) 2002 AQ (std) - with indoor sources 2002 AQ (std) - no indoor sources Simulated Year - Scenario 1 2 3 4 5 6 1 9 10 11 12 13 14 15 16 Figure 17. Estimated number of all asthmatics in Philadelphia with at least one exposure at or above the potential health effect benchmark level, using modeled 2002 air quality just meeting the current standard, with and without modeled indoor sources. 7.9.4.2 Number of Repeated Exposures Above Selected Levels For air quality simulated to just meet the current standard, repeated exposures at the selected potential health effect benchmarks are more frequent than that estimated for the modeled as is air quality. Figure 19 illustrates this using the simulated asthmatic population for year 2002 data as an example. Many asthmatics that are exposed at or above the selected levels are exposed more than one time. Repeated exposures above the potential health effect benchmark levels are reduced however, when not including the contribution from indoor sources. The percent of asthmatics exposed drops with increasing benchmark level, with progressively fewer persons experiencing multiple exposures for each benchmark level. April 2008 Draft 123 ------- 90 1 2 3 4 5 6 7 Estimated Number of Repeated Exposures in a Year 200 - with indoor soucrces 250 - with indoor sources 300 - with indoor sources r 200 - no indoor sources fc=b? 250 - no indoor sources CD 300 - no indoor sources Health Effect Benchmark Level (ppb) Figure 18. Estimated percent of asthmatics in Philadelphia County with repeated exposures above health effect benchmark levels, using modeled 2002 air quality just meeting the current standard, with and without modeled indoor sources. 9 7.10 VARIABILITY AND UNCERTAINTY 10 7.10.1 Introduction 11 The methods and the model used in this assessment conform to the most contemporary 12 modeling methodologies available. APEX is a powerful and flexible model that allows for the 13 realistic estimation of air pollutant exposure to individuals. Since it is based on human activity 14 diaries and accounts for the most important variables known to affect exposure, it has the ability 15 to effectively approximate actual conditions. In addition, the input data selected were the best 16 available data to generate the exposure results. However, there are constraints and uncertainties 17 with the modeling approach and the input data that limit the realism and accuracy of the model 18 results. April 2008 Draft 124 ------- 1 All models have limitations that require the use of assumptions. Limitations of APEX lie 2 primarily in the uncertainties associated with data distributions input to the model. Broad 3 uncertainties and assumptions associated with these model inputs, utilization, and application 4 include the following, with more detailed analysis summarized below and presented previously 5 (see EPA, 2007g; Langstaff, 2007). In addition, while at this time only the analyses for 6 Philadelphia County were complete, uncertainties are discussed regarding some of the location- 7 specific input gathered to date. Identified uncertainties include: 8 9 • The CHAD activity data used in APEX are compiled from a number of studies in 10 different areas, and for different seasons and years. Therefore, the combined data set 11 may not constitute a representative sample for a particular study scenario. 12 • Commuting pattern data were derived from the 2000 U.S. Census. The commuting 13 data address only home-to-work travel. The population not employed outside the 14 home is assumed to always remain in the residential census tract. Furthermore, 15 although several of the APEX microenvironments account for time spent in travel, the 16 travel is assumed to always occur in basically a composite of the home and work 17 block. No other provision is made for the possibility of passing through other blocks 18 during travel. 19 • APEX creates seasonal or annual sequences of daily activities for a simulated 20 individual by sampling human activity data from more than one subject. Each 21 simulated person essentially becomes a composite of several actual people in the 22 underlying activity data. 23 • The APEX model currently does not capture certain correlations among human 24 activities that can impact microenvironmental concentrations (for example, cigarette 25 smoking leading to an individual opening a window, which in turn affects the amount 26 of outdoor air penetrating the microenvironment). 27 • Certain aspects of the personal profiles are held constant, though in reality they 28 change as individuals age. This is only important for simulations with long 29 timeframes, particularly when simulating young children (e.g., over a year or more). April 2008 Draft 125 ------- 1 7.10.2 Input Data Evaluation 2 Modeling results are heavily dependent on the quality of the data that are input to the 3 system. As described above, several studies were reviewed, and data from these studies were 4 used to develop the parameters and factors that were used to build the microenvironments in this 5 assessment. A constraint on this effort is that there are a limited number of NC>2 exposure studies 6 to use for evaluation. 7 The input data used in this assessment were selected to best simulate actual conditions 8 that affect human exposure. Using well characterized data as inputs to the model lessens the 9 degree of uncertainty in exposure estimates. Still, the limitations and uncertainties of each of the 10 data streams affect the overall quality of the model output. These issues and how they 11 specifically affect each data stream are discussed in this section. 12 7.10.3 Meteorological Data 13 Meteorological data are taken directly from monitoring stations in the assessment areas. 14 One strength of these data is that it is relatively easy to see significant errors if they appear in the 15 data. Because general climactic conditions are known for each area simulation, it would have 16 been apparent upon review if there were outliers in the dataset. However, there are limitations in 17 the use of these data. Because APEX only uses one temperature value per day, the model does 18 not represent hour-to-hour variations in meteorological conditions throughout the day that may 19 affect both NC>2 formation and exposure estimates within microenvironments. 20 7.10.4 Air Quality Data 21 Air quality data used in the exposure modeling was determined through use of EPA's 22 recommended regulatory air dispersion model, AERMOD (version 07026 (EPA, 2004)), with 23 meteorological data discussed above and emissions data based on the EPA's National Emissions 24 Inventory for 2002 (EPA, 2007b) and the CAMD Emissions Database (EPA, 2007c) for 25 stationary sources and mobile sources determined from local travel demand modeling and EPA's 26 MOBILE6.2 emission factor model. All of these are high quality data sources. Parameterization 27 of meteorology and emissions in the model were made in as accurate a manner as possible to 28 ensure best representation of air quality for exposure modeling. Further, minor sources not 29 included in the dispersion modeling were captured and any remaining long-term errors in the 30 results corrected through use of local concentrations derived from monitor observations. Thus, April 2008 Draft 126 ------- 1 the resulting air quality values are free of systematic errors to the best approximation available 2 through application of modeled data. 3 7.10.5 Population and Commuting Data 4 The population and commuting data are drawn from U.S. Census data from the year 5 2000. This is a high quality data source for nationwide population data in the U.S. However, the 6 data do have limitations. The Census used random sampling techniques instead of attempting to 7 reach all households in the U.S., as it has in the past. While the sampling techniques are well 8 established and trusted, they introduce some uncertainty to the system. The Census has a quality 9 section (http://www.census.gov/quality/) that discusses these and other issues with Census data. 10 In addition to these data quality issues, certain simplifying assumptions were made in 11 order to better match reality or to make the data match APEX input specifications. For example, 12 the APEX dataset does not differentiate people that work at home from those that commute 13 within their home tract, and individuals that commute over 120 km a day were assumed to not 14 commute daily. In addition to emphasizing some of the limitations of the input data, these 15 assumptions introduce uncertainty to the results. 16 Furthermore, the estimation of block-to-block commuter flows relied on the assumption 17 that the frequency of commuting to a workplace block within a tract is proportional to the 18 amount of commercial and industrial land in the block. This assumption introduces additional 19 uncertainty. 20 7.10.6 Activity Pattern Data 21 It is probable that the CHAD data used in the system is the most subject to limitations 22 and uncertainty of all the data used in the system. Much of the data used to generate the daily 23 diaries are over 20 years old. Table 44 indicates the ages of the CHAD diaries used in this 24 modeling analysis. While the specifics of people's daily activities may not have changed much 25 over the years, it is certainly possible that some differences do exist. In addition, the CHAD data 26 are taken from numerous surveys that were performed for different purposes. Some of these 27 surveys collected only a single diary-day while others went on for several days. Some of the 28 studies were designed to not be representative of the U.S. population, although a large portion of 29 the data are from National surveys. Furthermore, study collection periods occur at different 30 times of the year, possibly resulting in seasonal differences. A few of these limitations are April 2008 Draft 127 ------- 1 corrected by the approaches used in the exposure modeling (e.g., weighting by US population 2 demographics for a particular location, adjusting for effects of temperature on human activities). 3 A sensitivity analysis was performed to evaluate the impact of the activity pattern 4 database on APEX model results for O3 (see Langstaff (2007) and EPA (2007g)). Briefly, 5 exposure results were generated using APEX with all of the CHAD diaries and compared with 6 results generated from running APEX using only the CHAD diaries from the National Human 7 Activity Pattern Study (NHAPS), a nationally representative study in CHAD. There was very 8 good agreement between the APEX results for the 12 cities evaluated, whether all of CHAD or 9 only the NHAPS component of CHAD is used. The absolute difference in percent of persons 10 above a particular concentration level ranged from -1% to about 4%, indicating that the exposure 11 model results are not being overly influenced by any single study in CHAD. It is likely that 12 similar results would be obtained here for NC>2 exposures, although remains uncertain due to 13 different averaging times (1-hour vs. 8-hour average). 14 7.10.7 Air Exchange Rates 15 There are several components of uncertainty in the residential air exchange rate 16 distributions used for this analysis. EPA (2007d) details an analysis of uncertainty due to 17 extrapolation of air exchange rate distributions between-CMSAs and within-CMSA uncertainty 18 due to sampling variation. In addition, the uncertainty associated with estimating daily air 19 exchange rate distributions from air exchange rate measurements with varying averaging times is 20 discussed. The results of those investigations are briefly summarized here. 21 7.10.7.1 Extrapolation among cities 22 Location-specific distributions were assigned in the APEX model, as detailed in the 23 indoors-residential microenvironment. Since specific data for all of the locations targeted in this 24 analysis were not available, data from another location were used based on similar influential 25 characteristics. Such factors include age composition of housing stock, construction methods, 26 and other meteorological variables not explicitly treated in the analysis, such as humidity and 27 wind speed patterns. In order to assess the uncertainty associated with this extrapolation, 28 between-CSA uncertainty was evaluated by examining the variation of the geometric means and 29 standard deviations across cities and studies. April 2008 Draft 128 ------- 1 The analysis showed a relatively wide variation across different cities in the air exchange 2 rate geometric mean and standard deviation, stratified by air-conditioning status and temperature 3 range. This implies that the air exchange rate modeling results would be very different if the 4 matching of modeled locations to study locations was changed. For example, the NC>2 exposure 5 estimates may be sensitive to the assumption that the Philadelphia air exchange rate distributions 6 can be represented by the New York City air exchange rate data. 7 7.10.7.2 Within CSA uncertainty 8 There is also variation within studies for the same location (e.g., Los Angeles), but this is 9 much smaller than the variation across CMSAs. This finding tends to support the approach of 10 combining different studies for a CMS A. In addition, within-city uncertainty was assessed by 11 using a bootstrap distribution to estimate the effects of sampling variation on the fitted geometric 12 means and standard deviations for each CMSA. The bootstrap distributions assess the 13 uncertainty due to random sampling variation but do not address uncertainties due to the lack of 14 representativeness of the available study data or the variation in the lengths of the AER 15 monitoring periods. 16 1,000 bootstrap samples were randomly generated for each AER subset (of size N), 17 producing a set of 1,000 geometric mean and geometric standard deviation pairs. The analysis 18 indicated that the geometric standard deviation uncertainty for a given CSA/air-conditioning- 19 status/temperature-range combination tended to have a range of at most from fitted GSD-1.0 hf1 20 to fitted GSD+1.0 hf1', but the intervals based on larger AER sample sizes were frequently much 21 narrower. The ranges for the geometric means tended to be approximately from fitted GM-0.5 22 hf1 to fitted GM+0.5 hf1', but in some cases were much smaller. Figure 12 illustrates such 23 results for Los Angeles as an example. 24 April 2008 Draft 129 ------- 4.0— 3.5 — > 5 3.0— 3 55 .° 2 5 — ' 0.5 1.0 1.5 Geometric Mean 2.5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 •Bootstrapped Data +++Original Data Figure 19. Geometric mean and standard deviation of air exchange rate bootstrapped for Los Angeles residences with A/C, temperature range from 20-25 degrees centigrade (from EPA, 2007g). 7.10.7.3 Variation in measurement averaging times Although the averaging periods for the air exchange rates in the study data varied from one day to seven days, the analyses did not take the measurement duration into account and treated the data as if they were a set of statistically independent daily averages. To investigate the uncertainty of this assumption, correlations between consecutive 24-hour air exchange rates measured at the same house were investigated using data from the Research Triangle Park Panel Study (EPA, 2007g). The results showed extremely strong correlations, providing support for the simplified approach of treating multi-day averaging periods as if they were 24-hour averages. 7.10.8 Air Conditioning Prevalence Because the selection of an air exchange rate distribution is conditioned on the presence or absence of an air-conditioner, for each modeled area, the air conditioning status of the residential microenvironments was simulated randomly using the probability that a residence has an air conditioner, i.e., the residential air conditioner prevalence rate. For this study we used location-specific data from the American Housing Survey of 2003. EPA (2007g) details the specification of uncertainty estimates in the form of confidence intervals for the air conditioner prevalence rate, and compares these with prevalence rates and confidence intervals developed from the Energy Information Administration's Residential Energy Consumption Survey (RECS) April 2008 Draft 130 ------- 1 of 2001 for more aggregate geographic subdivision (e.g., states, multi-state Census divisions and 2 regions). 3 Air conditioning prevalence rates for the 5 locations from the American Housing Survey 4 (Table 48) ranged from 55% for Los Angeles to 97% for Atlanta. Reported standard errors were 5 relatively small, ranging from less than 1.2% for Atlanta to 1.8% for Detroit. The corresponding 6 95% confidence intervals are also small and range from approximately 4.6% to 6.9%. The 7 RECS prevalence estimates and confidence intervals compared with the similar locations using 8 AHS data were mixed. Good agreements between the AHS and RECS confidence intervals was 9 found for Atlanta and Detroit. Poor agreement with the AHS for either the Census Region or 10 Census Division estimates was shown for Los Angeles and Philadelphia, with estimates of those 11 owning A/C lower when considering the RECS data. However, since the AHS survey results are 12 city-specific and were based on a more recent survey, the AHS prevalence estimates were used 13 for the APEX modeling. 14 Furthermore, some residences use evaporative coolers, also known as "swamp coolers," 15 for cooling. The estimation of air exchange rate distributions from measurement data used here 16 did not take into account the presence or absence of an evaporative cooler. Based on statistical 17 comparison tests (i.e., F-test, Kruskal-Wallis, Mood) for where information was available to 18 generate AER distributions with and without swamp cooler ownership, it was determined that 19 presence or absence of such data did not alter the statistical air exchange model (EPA, 2007d). 20 7.10.9 Indoor Source Estimation 21 Other indoor NC>2 emission sources, such as gas pilot lights, gas heating, or gas clothes 22 drying were not included in this analysis, due to lack of data for characterization. 23 The data used to estimate the average number of daily food preparation events is 24 somewhat dated (1992) and may therefore be unrepresentative of current conditions, and may 25 lead to under- or over-estimates of exposure to exceedances of threshold concentrations of 26 concern. For example, if the population of Philadelphia County in 2003 prepares food at home 27 less frequently than the 1992 survey population, then the number of such exposures may be over- 28 estimated. 29 As noted above, it was assumed that the probability that a food preparation event 30 included stove use was the same no matter what hour of the day the food preparation event April 2008 Draft 131 ------- 1 occurred. If such probabilities differ, then the diurnal allocation of cooking events may differ 2 from the actual pattern. To the extent that the gas stove usage patterns may correlate with 3 ambient concentration patterns, the number of exposures to exceedances of threshold 4 concentrations of concern may be under- or over-estimated. For example, if gas stove usage and 5 ambient concentrations are positively correlated (e.g., if cooking tends to occur during evening 6 rush hour) and the diurnal allocation assumed here results in a lower correlation (e.g., if the 7 diurnal allocation understates the probability of gas stove usage at times of high ambient 8 concentrations) then the number of such exposures may be under-estimated. Or, for another 9 example, if the diurnal pattern allocation assumed here understates the probability of gas stove 10 usage at times when simulated subjects are assumed to be at home, then the number of such 11 exposures may be under-estimated. 12 The durations of the CARB cooking tests ranged from 21 minutes to 3 hours with an 13 average of about 70 minutes. But for implementation in APEX it was assumed that each cooking 14 event lasts exactly an hour. That is, the randomly selected net concentration contribution was 15 added to hourly average indoor concentration for the hour it was selected to occur. Because the 16 mass balance algorithm leads to carryover from one hour to the next, some of the indoor cooking 17 impact will influence subsequent hours. However, the impact of the cooking event may be 18 overstated or understated for cooking events longer or shorter than 1 hour. 19 April 2008 Draft 132 ------- i 8. References 2 3 AHS. (2003a). American Housing Survey for 2003. Available at: 4 http ://www. census. gov/hhes/www/housing/ahs/ahs. html. 5 6 AHS. (2003b). Source and Accuracy Statement for the 2003 AHS-N Data Chart. Available at: 7 http: //www .census. gov/hhes/www/housing/ahs/03 dtchrt/source .html 9 Barck, C, Sandstrom T, Lundahl J, Hallden G, Svartengren M, Strand V, Rak S, Bylin G. 10 (2002). Ambient level of NC>2 augments the inflammatory response to inhaled allergen in 11 asthmatics. RespirMed. 96:901'-917. 12 13 Barck, C, Lundahl J, Hallden G, Bylin G. (2005). Brief exposures to NC>2 augment the allergic 14 inflammation in asthmatics. Environ Res. 97:58-66. 15 16 Beckerman, B, Jerrett M, Brook JR, Verma DK, Arain MA, Finkelstein MM. (2008). 17 Correlation of nitrogen dioxide with other traffic pollutants near a major expressway. 18 AtmosEnviron. 42:275-290. 19 20 Bell, S and Ashenden TW. (1997). Spatial and temporal variation in nitrogen dioxide pollution 21 adjacent to rural roads. Water Air SoilPollut. 95:87-98. 22 23 Bylin, G, Lindvall T, Rehn T, Sundin B. (1985) Effects of short-term exposure to ambient 24 nitrogen dioxide concentrations on human bronchial reactivity and lung function. Eur. J. 25 Respir. Dis. 66:205-217. 26 27 Cape, JN, Tang YS, van Dijk N, Love L, Sutton MA, Palmer SCF. (2004). Concentrations of 28 ammonia and nitrogen dioxide at roadside verges, and their contribution to nitrogen 29 deposition. Environ Pollut. 132:469-478. 30 31 CARB. (2001). Indoor air quality: residential cooking exposures. Final report. California Air 32 Resources Board, Sacramento, California. Available at: 33 http ://www.arb. ca. gov/research/indoor/cooking/cooking.htm. 34 35 Chan, AT and Chung MW. (2003). Indoor-outdoor air quality relationships in vehicle: effect of 36 driving environment and ventilation modes. Atmos Environ. 37:3795-3808. 37 38 Devalia, JL, Rusznak C, Herdman MJ, Trigg CJ, Tarraf H, Davies RJ. (1994) Effect of nitrogen 39 dioxide and sulphur dioxide on airway response of mild asthmatic patients to allergen 40 inhalation. Lancet 344:1668-1671. April 2008 Draft 133 ------- 1 2 Dietert, RR, Etzel RA, Chen D, Halonen M, Holladay SD, Jarabek AM, Landreth K, Peden DB, 3 Pinkerton K, Smialowicz RJ, Zoetis T. (2000). Workshop to identify critical window of 4 exposure for children's health: immune and respiratory systems work group summary. 5 Environ Health Perspect. Suppl. 108(3):483-490. 6 7 EPA. (1982). Air Quality Criteria for Oxides of Nitrogen. National Center for Environmental 8 Assessment, Research Triangle Park, NC. EPA-600/8-82-026. Available at 9 http ://cfpub .epa. gov/ncea/cfm/recordi splay. cfm?deid=4443 6. 10 11 EPA. (1993). Air Quality Criteria Document for the Oxides of Nitrogen. National Center for 12 Environmental Assessment, Research Triangle Park, NC. EPA-600/8-91/049F. 13 Available at: http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=40179. 14 15 EPA. (1995). Review of the National Ambient Air Quality Standards for Nitrogen Dioxide: 16 Assessment of Scientific and Technical Information. OAQPS Staff Paper. Office of Air 17 Quality Planning and Standards, Research Triangle Park, NC. EPA-452/R-95-005. 18 September 1995. 19 20 EPA. (2002). Consolidated Human Activities Database (CHAD) Users Guide. Database and 21 documentation available at: http://www.epa.gov/chadnetl/. 22 23 EPA. (2004). AERMOD: Description of Model Formulation. Office of Air Quality Planning 24 and Standards, Research Triangle Park, NC. EPA-454/R-03-004. Available at: 25 http://www.epa.gov/scram001/7thconf/aermod/aermod_mfd.pdf 26 27 EPA. (2006a). Total Risk Integrated Methodology (TRIM) - Air Pollutants Exposure Model 28 Documentation (TRIM.Expo / APEX, Version 4) Volume I: User's Guide. Office of Air 29 Quality Planning and Standards, Research Triangle Park, NC. June 2006. Available at: 30 http://www.epa.gov/ttn/fera/human apex.html. 31 32 EPA. (2006b). Total Risk Integrated Methodology (TRIM) - Air Pollutants Exposure Model 33 Documentation (TRIM.Expo / APEX, Version 4) Volume II: Technical Support 34 Document. Office of Air Quality Planning and Standards, Research Triangle Park, NC. 35 Available at: http://www.epa.gov/ttn/fera/human_apex.html. 36 37 EPA. (2007a). Plan for Review of the Primary National Ambient Air Quality Standard for 38 Nitrogen Dioxide. Available at: 39 http://www.epa.gov/ttn/naaqs/standards/nox/s_nox_cr_pd.html. 40 April 2008 Draft 134 ------- 1 EPA. (2007b). Nitrogen Dioxide Health Assessment Plan: Scope and Methods for Exposure 2 and Risk Assessment. Office of Air Quality Planning and Standards, Research Triangle 3 Park, NC. Available at: http://www.epa.gov/ttn/naaqs/standards/nox/s_nox_cr_pd.html. 4 5 EPA. (2007c). US EPA Air Quality System (AQS). Download Detailed AQS Data. Available 6 at: http://www.epa.gov/ttn/airs/airsaqs/detaildata/downloadaqsdata.htm. 7 8 EPA(2007d). Field Guide to Air Quality Data (vl.0.0). February 28, 2007. Available at: 9 http://www.epa.gov/ttn/airs/aqsdatamart/documentation/index.htm. 10 11 EPA. (2007e). 2002 National Emissions Inventory Data & Documentation. Office of Air 12 Quality Planning and Standards, Research Triangle Park, NC. Available at: 13 http://www.epa.gov/ttn/chief/net/2002inventory. html. 14 15 EPA. (2007f). Clean Air Markets - Data and Maps. Office of Atmospheric Programs, 16 Washington, DC. Emissions Prepackaged Data Sets. Available at: 17 http://camddataandmaps.epa.gov/gdm/index.cfm?fuseaction=emissions. wizard. 18 19 EPA. (2007g). Ozone Population Exposure Analysis for Selected Urban Areas (July 2007). 20 Office of Air Quality Planning and Standards, Research Triangle Park, NC. EPA-452/R- 21 07-010. Available at: http://epa.gov/ttn/naaqs/standards/ozone/s_o3_cr_td.html. 22 23 EPA. (2007h). Review of the National Ambient Air Quality Standards for Ozone: Assessment 24 of Scientific and Technical Information. OAQPS Staff paper. Office of Air Quality 25 Planning and Standards, Research Triangle Park, NC. EPA-452/R-07-007a. Available 26 at: http://epa.gov/ttn/naaqs/standards/ozone/s_o3_cr_sp.html. 27 28 EPA. (2008a). ISA for Oxides of Nitrogen - Health Criteria (Second External Review Draft). 29 National Center for Environmental Assessment, Research Triangle Park, NC. Available 30 at: http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=l89147. 31 32 EPA. (2008b). Risk and Exposure Assessment to Support the Review of the NO2 Primary 33 National Ambient Air Quality Standard: Draft Technical Support Document (TSD). 34 Office of Air Quality Planning and Standards, Research Triangle Park, NC. 35 36 Finlayson-Pitts, BJ and Pitts JN. (2000). Chemistry of the Upper and Lower Atmosphere. 37 Academic Press, San Diego CA. Page 17. 38 39 Folinsbee, LJ. (1992). Does nitrogen dioxide exposure increase airways responsiveness? 40 ToxicollndHealth. 8:273-283. April 2008 Draft 135 ------- 1 2 Gilbert, NL, Woodhouse S, Stieb DM, Brook JR. (2003). Ambient nitrogen dioxide and 3 distance from a major highway. Sci Total Environ. 312:43-46. 4 5 Graham, SE and McCurdy T. (2004). Developing meaningful cohorts for human exposure 6 models. J Expos Anal Environ Epidemiol. 14(1)23-43. 7 8 Heeb, NV, Saxer CJ, Forss AM, Bruhlmann S. (2008). Trends of NO-, NO2-, and NH3- 9 emissions from gasoline-fueled Euro-3- to Euro-4-passenger cars. Atmos Environ. 10 42:2543-2554 11 12 Ito, K. (2007). Association between doarse particles and asthma emergency department (ED) 13 visits in New York City. Presented at: American Thoracic Society international 14 conference; San Francisco, CA. 15 16 Jenkins, HS, Devalia JL, Mister RL, Bevan AM, Rusznak C, Davies RJ. (1999). The effect of 17 exposure to ozone and nitrogen dioxide on the airway response of atopic asthmatics to 18 inhaled allergen: dose- and time-dependent effects. Am. J. Respir. Crit. CareMed. 19 160:33-39. 20 21 Johnson, T, Mihlan G, LaPointe J, Fletcher K, Capel J. (1999). Estimation Of Carbon 22 Monoxide Exposures and Associated Carboxyhemoglobin Levels In Denver Residents 23 Using pNEM/CO (version 2.0). Prepared by ICF Kaiser Consulting Group for the U.S. 24 Environmental Protection Agency under Contract No. 68-D6-0064, March 1999. 25 26 Jorres, R, Magnussen H. (1990). Airways response of asthmatics after a 30 min exposure, at 27 resting ventilation, to 0.25 ppm NO2 or 0.5 ppm SO2. Eur. Respir. J. 3:132-137. 28 29 Jorres, R, Magnussen H. (1991). Effect of 0.25 ppm nitrogen dioxide on the airway response to 30 methacholine in asymptomatic asthmatic patients. Lung 169:77-85. 31 32 Jorres, R, Nowak D, Grimminger F, Seeger W, Oldigs M, Magnussen H. (1995). The effect of 1 33 ppm nitrogen dioxide on bronchoalveolar lavage cells and inflammatory mediators in 34 normal and asthmatic subjects. Eur Respir J. 8:416-424. 35 36 Koontz, MD, Mehegan LL, and Nagda NL. (1992). Distribution and Use of Cooking Appliances 37 That Can Affect Indoor Air Quality, Report No. GRI-93/0013. Gas Research Institute, 38 Chicago, IL. 39 April 2008 Draft 136 ------- 1 Langstaff, JE. (2007). OAQPS Staff Memorandum to Ozone NAAQS Review Docket (OAR- 2 2005-0172). Subject: Analysis of Uncertainty in Ozone Population Exposure Modeling. 3 [January 31,2007]. Available at: 4 http://www.epa.gOv/ttn/naaqs/standards/ozone/s o3crtd.html. 5 6 Lee, K, Yang W, Bofinger ND. (2000). Impact of microenvironmental nitrogen dioxide 7 concentrations on personal exposures in Australia. J Air Waste Manage Assoc. 50:739- 8 1744. 10 McCurdy, TR. (1994). Analysis of high 1 hour NO2 values and associated annual averages 11 using 1988-1992 data. Office of Air Quality Planning and Standards, Research Triangle 12 Park, NC. Available in Docket A-93-06. 13 14 McCurdy, T, Glen G, Smith L, Lakkadi Y. (2000). The National Exposure Research 15 Laboratory' s Consolidated Human Activity Database. JExp Anal Environ Epidemiol. 16 10: 566-578. 17 18 McCurdy T and Graham SE. (2003). Using human activity data in exposure models: analysis of 19 discriminating factors. J Expos Anal Environ Epidemiol. 13:294-317. 20 21 Mohsenin, V. (1987) Airway responses to nitrogen dioxide in asthmatic subjects. J. Toxicol. 22 Environ. Health 22:371-380. 23 24 Mortimer, KM, Neas LM, Dockery DW, Redline S, Tager IB. (2002). The effect of air 25 pollution on inner-city children with asthma. EurRespirJ. 19:699-705. 26 27 National Cooperative Highway Research Program (NCHRP) (1997). Recommended Procedure 28 for Long-Range Transportation Planning and Sketch Planning, NCHRP Report 387, National 29 Academy Press. 151 pp., ISBN No: 0-309-060-58-3. 30 31 Peel, JL, Tolbert PE, Klein M, Metzger KB, Flanders WD, Knox T, Mulholland JA, Ryan PB, 32 Frumkin H. (2005). Ambient air pollution and respiratory emergency department visits. 33 Epidemiology. 16:164-174. 34 35 Persily, A, Gorfain J, Brunner G. (2005). Ventilation design and performance in U.S. office 36 buildings. ASHRAE Journal. April 2005, 30-35. 37 38 Pilotto, LS, Douglas RM, Attewell RG, Wilson SR (1997) Respiratory effects associated with 39 indoor nitrogen dioxide exposure in children. Int. J. Epidemiol. 26: 788-796. April 2008 Draft 137 ------- 2 Pilotto, LS, Nitschke M, Smith BJ, Pisaniello D, Ruffm RE, McElroy HJ, Martin J, Killer JE. 3 (2004). Randomized controlled trial of unflued gas heater replacement on respiratory 4 health of asthmatic schoolchildren. IntJ Epidemiol. 33:208-214. 6 Pleijel, H, Karlsson GP, Gerdin EB. (2004). On the logarithmic relationship between 7 concentration and the distance from a highroad. Sci Total Environ. 332:261-264. 9 Rodes, C, Sheldon L, Whitaker D, Clayton A, Fitzgerald K, Flanagan J, DiGenova F, Hering S, 10 Frazier C. (1998). Measuring Concentrations of Selected Air Pollutants Inside 11 California Vehicles. California Environmental Protection Agency, Air Resources Board. 12 Final Report, December 1998. 13 14 Rodes, CE and Holland DM. (1981). Variations of NO, NO2 and O3 concentrations downwind 15 of a Los Angeles freeway. Atmos Environ. 15:243-250. 16 17 Roger, LJ, Horstman DH, McDonnell W, Kehrl H, Ives PJ, Seal E, Chapman R, Massaro E. 18 (1990) Pulmonary function, airway responsiveness, and respiratory symptoms in 19 asthmatics following exercise inNCh. Toxicol. Ind. Health 6:155-171. 20 21 Roorda-Knape, MC, Janssen NAH, De Hartog JJ, Van Vliet PHN, Harssema H, Brunekreef B. 22 (1998). Air Pollution from traffic in city districts near major roadways. Atmos Environ. 23 32(11)1921-1930. 24 25 Rubenstein, I, Bigby BG, Reiss TF, Boushey HA Jr. (1990) Short-term exposure to 0.3 ppm 26 nitrogen dioxide does not potentiate airway responsiveness to sulfur dioxide in asthmatic 27 subjects. Am. Rev. Respir. Dis. 141:381-385. 28 29 Schildcrout, JS, Sheppard L, Lumley T, Slaughter JC, Koenig JQ, Shapiro GG. (2006). 30 Ambient air pollution and asthma exacerbations in children: an eight-city analysis. Am J 31 Epidemiol. 164:505-517. 32 33 Schwartz, J, Dockery DW, Neas LM, Wypij D, Ware JH, Spengler JD, Koutrakis P, Speizer FE, 34 Ferris BG, Jr. (1994). Acute effects of summer air pollution on respiratory symptom 35 reporting in children. Am JRespir Crit Care Med. 150:1234-1242. 36 37 Shorter, JH, Herndon S, Zahniser MS, Nelson DD, Wormhoudt J, Demerjian KL, Kolb CE. 38 (2005). Real-Time Measurements of Nitrogen Oxide Emissions from In-Use New York 39 City Transit Buses Using a Chase Vehicle. Environ Sci Technol. 39:7991-8000. April 2008 Draft 138 ------- 2 Son, B, Yang W, Breysse P, Chung T, Lee Y. (2004). Estimation of occupational and 3 nonoccupational nitrogen dioxide exposure for Korean taxi drivers using a 4 microenvironmental model. Environ Res. 94:291-296. 5 6 Spicer, CW, Kenny DV, Ward GF, Billick IH. (1993). Transformations, lifetimes, and sources 7 of NO2, HONO, and HNO3 in indoor environments. JAWMA. 43(11): 1479-1485. 8 9 Strand, V, Salomonsson P, Lundahl J, Bylin G. (1996). Immediate and delayed effects of 10 nitrogen dioxide exposure at an ambient level on bronchial responsiveness to histamine in 11 subj ects with asthma. Eur Respir J. 9:73 3 -740. 12 13 Strand, V, Rak S, Svartengren M, Bylin G. (1997). Nitrogen dioxide exposure enhances 14 asthmatic reaction to inhaled allergen in subjects with asthma. Am J Respir Crit Care 15 Med. 155:881-887. 16 17 Strand, V, Svartengren M, Rak S, Barck C, Bylin G. (1998). Repeated exposure to an ambient 18 level of NO2 enhances asthmatic response to nonsymptomatic allergen dose. Eur Respir 19 J. 12:6-12. 20 21 Tolbert, PE, Klein M, Peel JL, Sarnat SE, Sarnat JA. (2007). Multipollutant modeling issues in 22 a study of ambient air quality and emergency department visits in Atlanta. J Expos Sci 23 Environ Epidemiol. 17S2:S29-35. 24 25 Tunnicliffe, WS, Burge PS, Ayres JG. (1994) Effect of domestic concentrations of nitrogen 26 dioxide on airway responses to inhaled allergen in asthmatic patients. Lancet 344:1733- 27 1736. 28 29 US DOT. (2007). Part 3-The Journey To Work files. Bureau of Transportation Statistics 30 (BTS). Available at: http://transtats.bts.gov/. 31 32 Westerdahl, D, Fruin S, Sax T, Fine PM, Sioutas C. (2005). Mobile platform measurements of 33 ultrafme particles and associated pollutant concentrations on freeways and residential 34 streets in Los Angeles. Atmos Environ. 39:3597-3610. 35 36 Witten, A, Solomon C, Abbritti E, Arjomandi M, Zhai W, Kleinman M, Balmes J. (2005). 37 Effects of nitrogen dioxide on allergic airway responses in subjects with asthma. J. 38 Occup. Environ. Med. 47:1250-1259. 39 April 2008 Draft 139 ------- 1 Wolff, GT. (1993). Letter to EPA Administrator Carol Browner: "CASAC Closure on the Air 2 Quality Criteria Document for Oxides of Nitrogen." EPA-SAB-CASAC-LTR-93-015, 3 September 30. 4 5 Wolff, GT. (1995). Letter to EPA Administrator Carol Browner: "CAS AC Review of the Staff 6 Paper for the Review of the National Ambient Air Quality Standards for Nitrogen 7 Dioxide: Assessment of Scientific and Technical Information." EPA-SAB-CASAC- 8 LTR-95-004, August 22. 9 10 Yao, X, Lau NT, Chan CK, Fang M. (2005). The use of tunnel concentration profile data to 11 determine the ratio of NO2/NOX directly emitted from vehicles. Atmos Chem Phys Discuss. 12 5:12723-12740. Available at: http://www.atmos-chem-phys-discuss.net/5/12723/2005/acpd- 13 5-12723-2005.pdf. April 2008 Draft 140 ------- United States Office of Air Quality Planning and Standards EPA-452/P-08-001 Environmental Protection Air Quality Strategies and Standards Division April 2008 Agency Research Triangle Park, NC ------- |