United States
Environmental Protection
Agency
Office Of Air Quality
Planning And Standards
Research Triangle Park, NC 27711
EPA-453/R-01-003
January 2001
Air
vvEPA
National-Scale Air Toxics
Assessment for 1996
NOTE: THIS DOCUMENT IS A PRELIMINARY DRAFT.
It has not been formally released by the U.S. Environmental
Protection Agency and should not at this stage be construed to
represent Agency policy or factual conclusions. This
document is being provided now for review to EPA's Science
Advisory Board. It should not be cited or referred to as EPA's
final National-Scale Air Toxics Assessment.
-------
EPA-453/R-01-003
National-Scale Air Toxics Assessment for 1996
EPA Office of Air Quality
Planning and Standards
DRAFT for EPA Science Advisory Board Review: January 18, 2001
-------
Table of Contents
Table of Contents i
List of Tables iv
List of Figures v
List of Appendices vi
1 Introduction 1
1.1 EPA's National Air Toxics Program 3
1.1.1 Background on the Air Toxics Program 3
1.1.2 Air Toxics Program Goals 4
1.2 Overview of National Air Toxics Assessment (NATA) Activities 5
1.2.1 Scope of NATA Activities 5
1.2.2 Purpose of NATA Activities 6
1.2.3 Links between NATA Activities and Other Program Elements 6
1.3 The National-Scale Assessments 7
2 The Initial National-Scale Assessment 8
2.1 Uses and Limitations 8
2.2 The EPA Risk Assessment Paradigm 9
2.3 Application of the EPA Risk Assessment Paradigm 10
2.4 Conceptual Model 10
2.4.1 Scope and Resolution 10
2.4.2 Time-Frame for Exposure 11
2.4.3 Rationale for Aggregating Air Toxics and Sources 11
2.4.4 Details of the Conceptual Model 11
2.4.4.1 Sources 12
2.4.4.2 Stressors 12
2.4.4.3 Pathways/Media 12
2.4.4.4 Exposure Routes 12
2.4.4.5 Subpopulations 13
2.4.4.6 Non-Human Receptors 13
2.5 Stakeholder Involvement 13
2.6 Peer Review Activities for the Initial National-Scale Assessment 14
2.6.1 Past Reviews of National-Scale Assessment Components 14
2.6.
2.6.
2.6.
2.6.
2.6.
2.6.
2.6.
.1 List of Urban Air Toxics 14
.2 1996 National Toxics Inventory 15
.3 ASPEN National Dispersion Model 15
.4 Hazardous Air Pollutant Exposure Model (HAPEM) 15
.5 EPA Risk Assessment Guidelines 15
.6 Microenvironment Factors 16
.7 ASPEN Results 16
2.6.2 Review of Planning and Scoping Document for the Initial National-Scale Assessment 16
Methods 17
3.1 Introduction 17
3.2 Exposure Assessment 20
3.2.1 Source Characterization: Emission Inventories 20
3.2.1.1 Approach 21
3.2.1.2 Scope 25
3.2.1.3 Sources of Data 27
3.2.1.3.1 Point Source Emissions 27
3.2.1.3.2 Non-Point Emissions 29
3.2.1.3.3 Mobile Sources 29
3.2.2 Environmental Fate and Transport Characterization 30
3.2.2.1 Overview of the ASPEN Dispersion Model 31
3.2.2.2 Application of ASPEN for the Initial National-Scale Assessment 32
3.2.3 Estimating Population Exposure 33
-------
3.2.3.1 Overview of HAPEM4 33
3.2.3.2 Application of HAPEM4 for the Initial National Scale Assessment 36
3.3 Dose-Response Assessment 37
3.3.1 Introduction 37
3.3.2 Cancer 38
3.3.2.1 Hazard Identification 38
3.3.2.2 Dose-Response Assessment for Carcinogens 39
3.3.3 Effects other than cancer 40
3.3.3.1 Hazard Identification 40
3.3.3.2 Dose-Response Assessment for Non-Carcinogens 41
3.4 Risk Characterization 41
3.4.1 Introduction 41
3.4.2 Cancer 43
3.4.3 Effects Other Than Cancer 44
3.4.4 Discussion of Uncertainties in the Dose-Response Assessment 49
3.4.4.1 Uncertainties in the Unit Risk Estimate 49
3.4.4.2 Uncertainties inReference Concentrations 50
4 Results and Discussion of Exposure Assessment 52
4.1 Introduction 52
4.2 Source Characterization: Emission Inventories 52
4.2.1 Summary and Discussion of 1996 Emission Inventory Results 53
4.2.2 Discussion of Inventory Uncertainties 56
4.2.2.1 Uncertainties in Completeness of Point Source Universe 57
4.2.2.2 Uncertainties Due to the Dynamic Nature of Emission Inventories 60
4.2.2.3 Uncertainties in Emission Locations 60
4.2.2.4 Uncertainty Due to Stack Parameter Defaults 63
4.2.2.5 Variations in Reported Emissions 64
4.2.2.6 Uncertainties Due to Pollutant Groupings 64
4.2.2.7 Particle Size and Reactivity Assignments 64
4.3 Environmental Fate and Transport Characterization 65
4.3.1 Summary of ASPEN Modeling Results 65
4.3.2 Discussion of Results 68
4.3.3 Comparison of ASPEN Modeling System to Monitoring Data 69
4.3.4 Discussion of ASPEN Dispersion Modeling Uncertainties 74
4.3.4.1 Emission Characterization Uncertainties 75
4.3.4.2 Meteorological Characterization Uncertainties 76
4.3.4.3 Model Formulation and Methodology Uncertainties 77
4.3.4.4 Uncertainty Due to Background 78
4.3.4.5 Summary of Model Uncertainty Investigations 79
4.4 Estimating Population Exposure 79
4.4.1 HAPEM4 Census Tract Level Exposure Estimates 80
4.4.2 National-Scale Assessment Exposure Estimates 82
4.4.3 Comparison of HAPEM4 Exposure Concentrations to ASPEN Ambient Concentrations... 85
4.4.4 Discussion of HAPEM4 Limitations and Uncertainties 87
5 Risk Characterization 90
5.1 Introduction 90
5.2 Cancer Risks 93
5.2.1 Pollutant-Specific Cancer Risks to Individuals 93
5.2.2 Pollutant-Specific Cancer Risks to Populations 95
5.2.3 Aggregate Cancer Risks of Multiple Pollutants to Individuals 97
5.2.4 Aggregate Cancer Risks of Multiple Pollutants to Populations 97
5.3 Non-Cancer Hazard 98
5.3.1 Pollutant-Specific Hazard Quotient for Individuals 98
5.3.2 Pollutant-Specific Hazard Quotient for Populations 99
5.3.3 Aggregate Target Organ Specific Hazard Index of Multiple Pollutants to Individuals 99
5.3.4 Aggregate TOSHI of Multiple Pollutants to Populations 100
-------
5.4 Discussion of the Risk of Diesel Exhaust 101
5.5 Uncertainty and Variability Analysis for the NATA National- Scale Assessment 103
5.5.1 Introduction 103
5.5.2 Source Characterization 104
5.5.2.1 Data Sources 104
5.5.2.2 Emission Locations 105
5.5.2.2.1 Point Sources 105
5.5.2.2.2 County-Level Emissions Sources 105
5.5.2.3 Stack Parameter Defaults 105
5.5.2.4 Particle Size and Reactivity Assignments 105
5.5.2.5 Chemical Speciation Data 105
5.5.3 Ambient Concentration Estimation 106
5.5.3.1 Temporal Resolution of Emissions 106
5.5.3.2 Simplifying Assumptions 106
5.5.3.3 Meteorological Characterization Uncertainties 106
5.5.3.4 Model Formulation and Methodology Uncertainties 106
5.5.3.4.1 Deposition and Dispersion Algorithms 106
5.5.3.4.2 Atmospheric Transformation Algorithms 107
5.5.3.4.3 Interpolation Between Census Tract Centroids 107
5.5.3.4.4 Summary 107
5.5.3.5 Illustration of Uncertainty and Variability Associated With Ambient Concentration
Estimates 108
5.5.4 Personal Exposure Assessment 110
5.5.4.1 Microenvironment Factors 110
5.5.4.1.1 Population Cohorts 110
5.5.4.2 Activity Pattern Sequence 110
5.5.4.3 Illustration of Uncertainty and Variability Associated with Exposure Estimates Ill
5.5.5 Illustration of Uncertainty and Variability for Dose-Response Assessment 112
5.5.5.1 Unit Risk Estimates (UREs) 112
5.5.5.2 Reference Concentrations (RfCs) 113
5.5.6 Illustration of Propagation of Uncertainty and Variability 114
5.5.7 Aggregation of Risk Across Pollutants 116
5.5.8 Recommendations for Further Characterization of Uncertainty 117
5.5.8.1 Role of Uncertainty Analysis 117
5.5.8.2 Technical Issues in Further Characterization of Uncertainty 118
5.5.8.3 Future Plans 120
6 Summary and Recommendations 121
6.1 Perspective on the National-Scale Assessment for 1996 121
6.2 Summary of Initial Results of National Scale Assessment 122
6.3 Recommendations 124
6.3.1 Identifying Air Toxics of Greatest Concern 124
6.3.2 Prioritizing Efforts to Reduce Emissions 125
6.3.3 Characterizing Contributions of Sources 125
6.3.4 Tracking Trends and Progress 126
6.3.5 Setting Data Collection and Research Priorities 126
7 References 130
111
-------
List of Tables
Table 3-1. Pollutant set for the initial national-scale assessment 18
Table 3-2. Urban air toxics believed to present risks from multipathway exposure 20
Table 3-3. Spatial Allocation Factors (SAF) Developed for EMS-HAP 24
Table 3-4. Non-Point Source Stationary Categories with EPA-Derived Emissions 29
Table 3 -5. Hazard identification and dose-response information for carcinogenic effects 45
Table 3-6. Non-Cancer Dose-Response Information 47
Table 3-7. Grouping of compounds by target organ and uncertainty factor for aggregation of effects other
than cancer 49
Table 4-1. Summary of Inventories Used as Input to EMS-HAP 53
Table 4-2. Comparison of EMS-HAP and NTI emission totals 55
Table 4-3. Summary of Data Sources to the NTI 56
Table 4-4. 1990 CEP and 1996 NTI National Emission Totals 58
Table 4-5. Facility Count Summary 59
Table 4-6. Summary of Defaulted Sources and Emissions for Lead, Chromium, and Cadmium 61
Table 4-7. Stack Parameter Default Statistics 63
Table 4-8. Pollutants grouped by the dominant source sector affecting their national average
concentrations 66
Table 4-9. Number of Counties of Each Dominant Source Sector for the 10 Highest County Median
Concentrations 66
Table 4-10. Comparison of the Measurement Data to Modeled Concentration 71
Table 4-11. Maximum Modeled Concentration Compared to Monitored Value 72
Table 4-12. Source location uncertainties for point source inventory for three metals 76
Table 4-13. HAPEM4 to ASPEN Average Ratio1 86
Table 5-1. Illustration: Parameters for lognormal distributions fitted to monitor-to-model ratios for seven
pollutants 109
Table 5-2. Illustration: Calculated percentiles for monitormodel ratio distribution 109
Table 5-3. Illustration: Percentiles for uncertainty and variability in the personal: ambient ratio
distribution 112
Table 5-4. Illustration: Percentiles for variability in the benzene URE 113
Table 5-5. Illustration: Combined uncertainty and variability, in terms of the risk ratio (i.e., the ratio of
"true" risk to estimated risk) 116
Table 5-6. Sources of uncertainty for the National-Scale Assessment 119
IV
-------
List of Figures
1-1. Overview of National Air Toxics Assessment (NATA) Activities 134
1-2. Links Between NATA Activities and Other Program Elements 135
2-1. NAS Risk Assessment/Risk Management Paradigm 136
2-2. National-Scale Air Toxics Health Assessment: Conceptual Model 137
2-3. Diagram of the HAPEM4 40 Cohort Groups (2x4x5=40) 138
3-1. 1996 NTI State and Local Agency Data Summary 139
3-2. Example Demographic Groups, Microenvironment, and Activities 140
3-3. Example of a Daily Exposure Scenario for a Cohort 141
4-1. Summary of 1996 NTI Emissions for 3 3 Air Toxics by Source Sector 142
4-2. Summary of 1996 NTI Emissions of 33 Air Toxics by Urban and Rural Designations 142
4-3. 1996 NTI-Aldehydes Emission Densities 143
4-4. 1996 NTI-Metals Emission Densities 143
4-5. 1996 NTI-Halides Emission Densities 144
4-6. 1996 NTI-POM and Hydrocarbons Emission Densities 144
4-7. Estimated Annual Average Concentrations (ug/m3) for Benzene 145
4-8. Percent Contribution to the Statewide Annual Average Ambient
Benzene Concentration Estimates 146
4-9. Comparison of Annual Average Model Concentrations for 10 Pollutants 147
4-10. Annual Average Concentrations for Urban and Rural Census Tracts 148
4-11. Relative contribution of major, area and mobile sources 149
4-12. Model-to-monitor scatter plot for benzene 150
4-13. Ratio box plot showing distribution of model/monitor ratios for each pollutant 151
4-14. POM Exposure Concentration Distribution Among Cohorts in an Urban NY Census 152
4-15. Exposure Results - Summary Table Example (Partial Table) 153
4-16. 1996 Modeled Exposure Concentrations for Acetaldehyde - Statewide Concentration Distribution
Estimates 154
4-17. 1996 Modeled Exposure Concentrations for Acetaldehyde - Statewide Source Sector Contribution
Estimates 155
4-18. 1996 Modeled Median Exposure Concentrations for North Carolina 156
4-19. Example State Exposure Concentration Map 157
4-20. Benzene Exposure Variability within a County 158
5-1. 1996 Risk Characterization - Distribution of lifetime cancer risk for the US population, based on
1996 exposure to all source sectors and background combined 159
5-2. 1996 Risk Characterization - Population whose 1996 exposure exceeded set cancer risk levels
based on all source sectors and background 160
5-3. 1996 Risk Characterization - Distribution of lifetime cancer risk for the US population, based on
1996 exposure to multiple carcinogens 161
5-4. 1996 Risk Characterization - Population whose 1996 exposure* exceeded set risk levels of risk for
carcinogens combined 162
5-5. 1996 Risk Characterization - Distribution of non-cancer hazard quotient for the US population,
based on 1996 exposure* to all source sectors and background combined 163
5-6. 1996 Risk Characterization - Adult population whose 1996 exposure* exceeded set non-cancer
hazard quotient levels based on all source sectors and background combined 164
5-7. 1996 Risk Characterization - Children population whose 1996 exposure* exceeded set non-cancer
hazard quotient levels based on all source sectors and background combined 165
5-8. 1996 Risk Characterization - Distribution of non-cancer target organ-specific hazard index
(TOSHI) for effects to the respiratory system, based on 1996 multiple-pollutant exposure to
adults in the US population 166
5-9. Illustration: Distribution of monitor-to-model ratios for stable gases, developed from ratios for
benzene and perchloroethylene 167
5-10. Illustration: Distribution of monitor-to-model ratios for reactive gases, developed from ratios for
formaldehyde and acetaldehyde 167
-------
List of Figures
5-11. Illustration: Distribution of monitor-to-model ratios for paniculate species, developed from ratios
for lead, chromium, and cadmium 167
5-12. Illustration: Distribution of ambient-to-personal concentration ratios for ozone, assumed to apply
for "typical" gaseous pollutants 168
5-13. Illustration: Distribution of ambient-to-personal concentration ratios for paniculate matter,
assumed to apply for "typical" paniculate pollutants 168
5-14. Illustration: Uncertainty and variability surrounding the URE for benzene, in terms of the ratio
between the estimated URE and the "true" URE 169
5-15. Illustration: Uncertainty and variability surrounding a typical RfC, in terms of the ratio between
the estimated RfC and the "true" RfC 169
5-16. Cancer - risk ratio for stable gas 170
5-17. Cancer - risk ratio for reactive gas 170
5-18. Cancer - risk ratio for paniculate 170
5-19. Noncancer - risk ratio for stable gas 171
5-20. Noncancer - risk ratio for reactive gas 171
5-21. Noncancer - risk ratio for paniculate 171
List of Appendices
A. Summary of July 2000 Peer Review of the Draft Document "Planning and Scoping the Initial
National-Scale Assessment: An Element of the EPA National Air Toxics Program"
B. HAPEM4 User's Guide, with Microenvironment Factor Report
C. EMS-HAP User's Guide
D. Development of the Emission Inventory
E. ASPEN User's Guide
F. Estimation of Background Concentrations for Diesel Paniculate Matter
G. Health Effects Information Used In Cancer and Noncancer Risk Characterization for the NATA 1996
National-Scale Assessment
H. Estimating Carcinogenic Potency for Mixtures of Fob/cyclic Organic Matter (POM) for the 1996
National-Scale Assessment
I. Model-to-Monitor Comparison Methods
J. Comparison of ASPEN Results to Monitored Concentrations
K. HAPEM4 Results
L. Risk Characterization Results
VI
-------
1 Introduction
This document describes the EPA's National-Scale Air Toxics Assessment, based on
emissions data for 1996 (called the "national-scale assessment"). The national-scale
assessment is a nationwide study of potential inhalation exposures and health risks
associated with 32 hazardous air pollutants (hereinafter called air toxics) and diesel
particulate matter (diesel PM), based on 1996 data, because 1996 emission inventories
were the most complete and up-to-date available. The initial national-scale assessment is
one component of the National Air Toxics Assessment (NATA), the technical support
component of EPA's National Air Toxics Program Specifically, NAT A includes
activities such as expanding air toxics monitoring, improving and periodically updating
emissions inventories, periodically conducting national- and local-scale air quality, multi-
media and exposure modeling, characterizing risks associated with air toxics exposures,
and continuing research on health and environmental effects and exposures to both
ambient and indoor sources.
The initial national-scale assessment is the first in an anticipated series of national-scale
assessments. These national-scale assessments may be repeated every 3 years, to track
trends in the reduction of emissions of air toxics, as well as progress in reducing risks
from air toxics exposure. The purpose of the national-scale assessment is to gain a better
understanding of the air toxics problem. It was not designed, and is not appropriate
specifically, for identifying local- or regional-scale air toxics "hot spots," nor is it
appropriate for identifying localized risks or individual risks from air toxics. Further
analyses on a national scale, and additional assessments on other scales (e.g., urban air
toxics assessments and residual risk assessments) are being performed in order to fully
characterize risks, especially disproportionate and cumulative risks.
Given the uncertainties and limitations associated with performing this national-scale
assessment, the EPA is seeking EPA Science Advisory Board (SAB) review on the
appropriateness of the methodologies and tools, as applied in this assessment, and
guidance on ways to improve future national-scale assessments. This document was
prepared for that SAB review.
The goals of this initial national-scale assessment are to assist in:
• Identifying air toxics of greatest potential concern, in terms of contribution to
population risk;
• Characterizing the relative contributions to air toxics concentrations and
population exposures from different types of air toxics emission sources;
• Setting priorities for the collection of additional air toxics data (e.g., emission
data, ambient monitoring data, data from personal exposure monitoring) for use in
local-scale and multipathway modeling and assessments, and for future research
to improve estimates of air toxics concentrations and their potential public health
impacts;
1
-------
• Establishing a baseline for tracking trends over time in modeled ambient
concentrations of air toxics; and,
• Establishing a baseline for measuring progress toward meeting goals for
inhalation risk reduction from ambient air toxics.
The results of the initial national-scale assessment will provide information to guide EPA
in developing and implementing various aspects of the national air toxics program.
However, the results will not be used to make specific regulatory decisions for air toxics.
While regulatory priority-setting will be informed by this and any future national-scale
assessments, risk-based regulations will be based on more refined and source-specific
data and assessments.
There are several other important limitations in the scope of the national-scale
assessment, and they are as follows:
• It is based on 1996 data and does not reflect significant reductions in air toxics
emissions that have occurred since that time
• It focuses only on 33 selected air toxics (32 air toxics of greatest concern in urban
areas and diesel particulate matter), and does not address the other 156 air toxics
listed in section 112(b) of the Clean Air Act
• It does not include risks due to non-inhalation exposure pathways (e.g.,
ingestion), which have been shown to be significant for some air toxics (e.g.,
mercury, dioxins)
• It does not include exposure from indoor sources of air toxics
• It focuses on average population risks, rather than individual extremes (i.e., does
not identify "hot spots")
• It does not reliably capture localized impacts and risks
• It currently includes estimates of background levels that are only approximate;
more research is needed to treat background contributions more explicitly
• Current model evaluation efforts indicate a tendency to under-predict ambient
concentrations, which may contribute toward an underestimation of risk
For these reasons, the initial national-scale assessment provides only a partial indication
of the total risk due to all air toxics. Its results should always be interpreted keeping this
perspective in mind.
-------
Section 1 of this document provides background information by describing EPA's
national air toxics program and by explaining the goals of NATA and the national-scale
assessments. Section 2 describes the purpose and goals of the initial national-scale
assessment, while Section 3 explains the methodologies used for the initial national-scale
assessment. Section 4 provides a summary of the results of that initial assessment.
Section 5 of the document presents the risk characterization, which includes an
uncertainty and variability analysis, and Section 6 provides a summary of results and
recommendations for future actions.
1.1 EPA's National Air Toxics Program
1.1.1 Background on the Air Toxics Program
The air toxics program was authorized under the 1970 Clean Air Act (CAA), and
redesigned and reauthorized under the 1990 CAA Amendments. The program is
designed to characterize, prioritize, and address, in an equitable manner (i.e., across
racial, cultural, and economic groups), the serious impacts of air toxics (also known as
hazardous air pollutants, or air toxics) on public health and the environment through a
strategic combination of regulatory approaches, voluntary partnerships, ongoing research
and assessments, and education and outreach.
Since 1990, EPA has made considerable progress in reducing emissions of air toxics
through regulatory, voluntary and other programs. To date, the overall air toxics program
has focused on reducing emissions of air toxics from major stationary sources through the
implementation of technology-based emissions standards, as required in section 112(d) of
the CAA. These actions have resulted in, or are projected to result in, substantial
reductions in air toxics emissions. Additionally, actions to address mobile sources under
other CAA programs have achieved significant reductions in air toxics emissions (e.g.,
the phase-out of lead from gasoline). Many motor vehicle and fuel emission control
programs of the past have reduced air toxics. Several current EPA programs further
reduce air toxics emissions from a wide variety of mobile sources. These include the
reformulated gasoline (RFG) program, the national low emission vehicle (NLEV)
program, Tier 2 motor vehicle emission standards and gasoline sulfur control
requirements, and recently finalized heavy-duty engine and vehicle standards and onroad
diesel fuel sulfur control requirements. In addition, certain other mobile source control
programs have been specifically aimed at reducing toxics emissions. These actions are
projected to reduce emissions substantially.
EPA expects, however, that the emission reductions that will result from these actions
may only be part of what is necessary to protect public health and the environment from
air toxics. In addition, section 112(f) of the CAA specifically directs EPA to assess the
risk remaining after implementation of technology-based standards (i.e., the residual risk)
in order to evaluate the need for additional stationary source standards to protect public
health and the environment. Under section 112(k), EPA will be performing
comprehensive local-scale assessments for several urban areas. In identifying
appropriate additional steps, EPA will use a risk-based focus to develop, implement and
facilitate additional federal, state and local regulatory and voluntary measures, if
necessary. In considering additional steps toward protecting human health and the
-------
environment, EPA will identify and focus on issues of highest priority.
1.1.2 Air Toxics Program Goals
EPA has ten long-range strategic goals [7] which establish the focus for the Agency's
work in the years ahead. One of these goals, EPA's Clean Air Goal, states that the air in
every American community will be safe and healthy to breathe. In particular, children,
the elderly, and people with respiratory ailments will be protected from health risks of
breathing polluted air. Reducing air pollution will also protect the environment, resulting
in many benefits, such as restoring life in damaged ecosystems and reducing health risks
to those whose subsistence depends directly on those ecosystems. The specific air toxics
objective under this goal is, by 2020, to eliminate unacceptable risks of cancer and other
significant health problems from air toxic emissions for at least 95% of the population,
with particular attention to children and other sensitive sub-populations, and substantially
reduce or eliminate adverse effects on our natural environment. Further, by 2010, the
tribes and EPA will have the information and tools to characterize and assess trends in air
toxics in Indian country.
EPA will progress toward meeting its air toxics program goals through a combination of
statutory authorities, regulatory activities and voluntary initiatives. EPA's overall
approach to reducing air toxics consists of the following four key components, which are
discussed in greater detail in the July 19, 1999 Federal Register notice for the National
Air Toxics Program: [2]:
1. Source-specific standards and sector-based standards. Section 112 of the CAA
specifies emission standards (technology-based standards) and residual risk
standards (risk-based standards), as well as area source standards identified by the
Integrated Urban Air Toxics Strategy. Additionally, section 129 of the CAA
requires standards for solid waste incineration and section 202(1) requires EPA to
promulgate reasonable requirements to control air toxics from motor vehicles and
their fuels.
2. National, regional, and community-based initiatives to focus on multi-media and
cumulative risks. Section 112(k)(4) of the CAA requires EPA to "encourage and
support area wide strategies developed by the state or local air pollution control
agencies." EPA's risk initiatives will include state, local and tribal program
activities consistent with the Integrated Urban Air Toxics Strategy on the local
level, as well as federal and regional activities associated with the multimedia
aspects of air toxics (e.g., the Great Waters program [Section 112(m)] and
Agency initiatives concerning mercury and other persistent and bioaccumulative
toxics [PBTs]). These Agency initiatives include collaboration between the air
and water programs on the impact of air deposition on water quality (e.g., by
accounting for the contribution of air deposition to the total maximum daily load
(TMDL) of pollutants to a water body), and collaboration between offices within
EPA's air program to assess the risks from exposures to air toxics from indoor
sources.
3. National air toxics assessment (NATA) activities. NATA activities will help EPA
-------
identify areas of concern, characterize risks, and track progress toward meeting
the overall air toxics program goals, as well as the risk-based goals of the various
activities and initiatives within the program. The NATA activities include
expansion of air toxics monitoring, improving and periodically updating
emissions inventories, national- and local-scale air quality, multi-media and
exposure modeling (including modeling which considers stationary and mobile
sources), continued research on health effects and exposures to both ambient and
indoor air, and use and improvement of exposure and risk assessment tools.
These activities will provide EPA with improved characterizations of air toxics
risk and risk reductions resulting from emissions control standards and initiatives
for both stationary and mobile source programs.
4. Education and outreach. In light of the scientific complexity inherent in air
toxics issues, EPA recognizes that the success of the overall air toxics program
depends, in part, on the Agency's ability to communicate effectively with the
public about air toxics risks and activities necessary to reduce those risks. This
includes education and outreach efforts on air toxics in the ambient as well as
indoor environments.
Under the CAA, EPA provides leadership and technical and financial assistance for the
development of cooperative federal, state, local, and tribal programs to prevent and
control air pollution. A strong partnership with and among the different governments that
each play a key role in air quality protection is critical to achieving clean air, because it is
the sum of these collective efforts that constitutes the national air quality program.
The Agency is committed to working with cities, communities, state, local and tribal
agencies, and other groups and organizations that can help implement activities to reduce
air toxics emissions. For example, EPA has been working with its regulatory partners
and interested stakeholders on activities related to NATA. In addition, the Agency will
continue to work with its regulatory partners and stakeholders on regulation development.
EPA also expects to involve local communities and industries in the development of local
risk initiatives such as the urban community-based pilot projects (i.e., local-scale
assessments).
1.2 Overview of National Air Toxics Assessment (NA TA)
Activities
As described in section 1.1, the overall air toxics program consists of four central
elements: (1) source-specific and sector-based standards, (2) national, regional and
community-based initiatives, (3) NATA activities, and (4) education and outreach. This
section focuses on the third of these elements, the NATA activities.
1.2.1 Scope of NATA Activities
EPA includes, within the scope of the NATA activities, all data gathering, analyses,
assessments, characterizations, and related research needed to support the air toxics
program, as shown in Figure 1-1. The figure depicts two levels of assessments.
National-scale assessments are shown centrally in boldface, and refined local-scale
assessments are depicted underneath. Other NATA activities that support each type of
-------
assessment (e.g., emission characterization, meteorological data collection) are shown in
boxes at the top of the figure, with arrows depicting the points in the assessment where
the information is used. The information from the NATA activities flows to the right side
of the figure, where the most important uses for the outputs are shown.
1.2.2 Purpose of NATA Activities
EPA envisions the NATA activities as an ongoing, permanent part of the air toxics
program. EPA intends to use NATA products to inform the entire air toxics program
with a body of coordinated information that expands and evolves to fit the needs of each
part of that program.
EPA's ongoing data collection and research will allow the Agency to continue to improve
its understanding of air toxics emissions, ambient concentrations (with a multi-media
focus, where appropriate), outdoor and indoor exposures (by inhalation and other
pathways, where appropriate), and health and environmental effects. These efforts will
also help to improve the tools and methods for assessing and characterizing public health,
environmental hazards, and cumulative risks associated with exposures to air toxics.
EPA is currently working to ensure stakeholder involvement in the planning and conduct
of these activities, and to ensure appropriate peer review of the underlying science and
assessment methods.
As shown in Figure 1-1, these NATA activities will serve several purposes. The
information and assessments developed by NATA activities will help EPA:
1. Determine priorities for regulatory programs as well as for national, regional, and
community-based initiatives;
2. Assess progress toward CAA goals and EPA's long-range strategic goals;
3. Inform state, local, and tribal programs and support public right-to-know
initiatives regarding risks associated with exposure to air toxics; and,
4. Support prospective assessments of the benefits estimated to result from
implementation of statutory air toxics mandates (as required by section 812 of the
CAA).
1.2.3 Links between NATA Activities and Other Program Elements
Figure 1-2 displays important examples of regulatory programs, risk-based initiatives,
and special studies that will utilize information from current and future NATA activities.
National-scale information produced by these NATA activities will specifically support
setting priorities and estimating progress by each of these elements of the air toxics
program. In turn, it is anticipated that the regulatory programs, initiatives, and studies
shown on Figure 1-2 will result not only in continued reductions in air toxics emissions
and risk, but will also lead to enhanced knowledge and tools to improve EPA's ability to
characterize air toxics risk. The risk reduction progress that is made, and the improved
risk characterization information and tools that become available, will then be reflected in
future NATA activities.
-------
1.3 The National-Scale Assessments
The purpose of the national-scale assessments is to gain a better understanding of the air
toxics problem on a national scale by compiling and analyzing existing air toxics data.
The goal of the national-scale assessments is to assist in: 1) identifying air toxics of
greatest potential concern in terms of contribution to population risk; 2) characterizing
the relative contributions of various types of emission sources to air toxics concentrations
and population exposures; 3) setting priorities for collection of additional air toxics data
and research to improve estimates of air toxics concentrations and their potential public
health impacts; 4) tracking trends in modeled ambient air toxics concentrations over time;
and, 5) measuring progress toward meeting goals for inhalation risk reduction from
ambient air toxics. To accomplish these goals, the national-scale assessment will be
repeated every three years, with the next national-scale assessment being performed in
2003, with 1999 air toxics data.
EPA particularly notes that, while the initial national-scale assessment (i.e., an
assessment which is national in scale, has low resolution, and represents only one part of
the overall NATA activities) can help set general programmatic priorities and provide
direction for the design of local-scale initiatives and more refined special studies, it is not
intended to serve as the basis for setting standards or addressing specific local concerns
or community environmental justice issues. EPA will use more refined and local-scale
information and assessment tools developed within NATA activities as the basis for risk-
based standard setting (e.g., residual risk standards) and for local initiatives.
-------
2 The Initial National-Scale Assessment
2.1 Uses and Limitations
The results of the initial national-scale assessment presented in this document provide
important information to help EPA continue to develop and implement various aspects of
the national air toxics program. However, it is important to note that these results will
not be used directly to regulate sources of air toxics emissions. Although the national-
scale assessments (both the initial assessment and future assessments) will inform the
regulatory priority-setting process, risk-based regulations will be supported by more
refined and source-specific data and assessment tools.
More specifically, the national-scale assessment results will assist in:
• Identifying air toxics of greatest potential concern, in terms of contribution to
population risk;
• Characterizing the relative contributions by different types of air toxics emission
sources to air toxics concentrations and population exposures (i.e., major, area
and other, on-road and non-road mobile, and background sources);
• Setting priorities for the collection of additional air toxics data (e.g., emission
data, ambient monitoring data, data from personal exposure monitoring) and
research to improve estimates of air toxics concentrations and their potential
public health impacts;
• Establishing a baseline for tracking trends over time in modeled ambient
concentrations of air toxics; and,
• Establishing a baseline for measuring progress toward meeting goals for
inhalation risk reduction from ambient air toxics.
The results of the initial national-scale assessment will help identify areas of the country
and pollutants where additional investigation is needed, and will help target locations
where more refined, regional- and local-scale analyses should be done.
There are several other important limitations in the scope of the national-scale
assessment, and they are as follows:
• It is based on 1996 data and does not reflect significant reductions in air toxics
emissions that have occurred since that time
• It focuses only on 33 selected air toxics (32 air toxics of greatest concern in urban
areas and diesel paniculate matter), and does not address the other 156 air toxics
listed in section 112(b) of the Clean Air Act
-------
• It does not include risks due to non-inhalation exposure pathways (e.g.,
ingestion), which have been shown to be significant for some air toxics (e.g.,
mercury, dioxins)
• It does not include exposure from indoor sources of air toxics
• It focuses on average population risks, rather than individual extremes (i.e., does
not identify "hot spots")
• It does not reliably capture localized impacts and risks
• It currently includes estimates of background levels that are only approximate;
more research is needed to treat background contributions more explicitly
• Current model evaluation efforts indicate a tendency to under-predict ambient
concentrations, which may contribute toward an underestimation of risk
For these reasons, the initial 1996 national-scale assessment should be considered to
provide only a partial indication of the total risk due to all air toxics. Its results should
always be interpreted with this perspective in mind.
The initial national-scale assessment is the first step in an iterative and evolving process
to assess and characterize risks from exposures to air toxics, measure progress in meeting
goals, and inform future directions for EPA's national air toxics program. While there
continue to be significant uncertainties and gaps in methods, models, and data that limit
EPA's ability to assess risks to public health and the environment associated with
exposures to air toxics, continued research will enable future assessment activities, both
at the national level and at more refined levels, to yield improved assessments of
cumulative air toxics risks. An important component of future NATA activities will be to
repeat the national-scale assessment every three years, with the next national-scale
assessment being performed in 2003 with 1999 air toxics data.
2.2 The EPA Risk Assessment Paradigm
Because cancer and noncancer health impacts generally cannot be directly isolated and
measured with respect to environmental exposures, EPA and others have spent more than
two decades developing an extensive set of risk assessment methods, tools, and data that
serve the purpose of estimating health risks for many Agency programs. Although
significant uncertainties remain, EPA's risk assessment science has been extensively
peer-reviewed, is widely used and understood by the scientific community, and continues
to expand and evolve as scientific knowledge advances. All NATA risk assessments will
be based on the most current and appropriate risk estimation methods.
EPA's framework for assessing and managing risks reflects the risk assessment and risk
management paradigm set forth by the National Academy of Sciences (NAS) in 1983 [3],
shown in Figure 2-1. This figure identifies research, risk assessment, and risk
management as three separate but connected elements. The NAS concluded that risk
-------
assessment and risk management are "two distinct elements" between which agencies
should maintain a clear conceptual distinction. The 1983 NAS report identified four
steps integral to any risk assessment: 1) hazard identification, 2) dose-response
assessment, 3) exposure assessment, and 4) risk characterization. As described in the
next section, the NAS paradigm for risk assessment serves as the basis for the NATA
national-scale assessment.
2.3 Application of the EPA Risk Assessment Paradigm
As described in more detail in section 3, the initial national-scale assessment includes
four major components:
(1) compiling a 1996 national emissions inventory of air toxics emissions
from outdoor sources;
(2) estimating 1996 air toxics ambient air concentrations for 33 air toxics (32
urban air toxics and diesel particulate matter (diesel PM)) nationwide;
(3) estimating 1996 population inhalation exposures to these air toxics; and,
(4) characterizing potential public health risks associated with these
exposures, including both cancer and noncancer effects.
The following sections describe the elements of the initial national-scale assessment.
2.4 Conceptual Model
2.4.1 Scope and Resolution
The initial national-scale assessment was national in scope, covering the contiguous
United States, Puerto Rico, and the Virgin Islands. The assessment excluded Alaska,
Hawaii, and U.S. territories other than Puerto Rico and the Virgin Islands because the
data needed to support the models (e.g., census tract and meteorological information)
were not readily available for these areas.
The pollutants identified and peer-reviewed for inclusion in this initial national-scale
assessment were the air toxics identified as priority pollutants in EPA's Integrated Urban
Air Toxics Strategy (IUATS)[2]. These 33 air toxics are a subset of EPA's list of 188 air
toxics, under CAA section 112(b), and have been identified as those pollutants that
present the greatest threat to public health in the largest number of urban areas. This
assessment also included diesel particulate matter, an indicator of diesel exhaust. EPA
has recently listed this likely human carcinogen as a mobile source air toxic and is
addressing this pollutant in several regulatory actions.1 Because the current dioxin
exposure and human health reassessment is undergoing review, the initial national-scale
assessment did not include the class of compounds known as dioxins. Since the most
significant exposure route for dioxin is ingestion rather than inhalation, dioxin's relative
1 In the Mobile Source Air Toxics rule (December 2000), under section 202(1)(2) of the CAA, EPA listed
diesel particulate matter plus diesel exhaust organic gases, collectively, as a mobile source air toxic.
10
-------
contribution to this study's inhalation risk estimates would likely not have been great. It
is expected that dioxins will be included in future national-scale assessments where
ingestion and inhalation exposures are assessed. In summary, 32 of the priority
pollutants in EPA's Integrated Urban Air Toxics Strategy, as well as diesel PM, make up
the 33 air toxics that were included in the initial national-scale assessment. These 33 air
toxics will hereinafter be referred to as "the pollutant set."
The national-scale assessment included a risk characterization based on estimates of
inhalation exposure concentrations determined at the census-tract level. Because of
uncertainties associated with the accuracy of the results at the census-tract level, the
results of the risk characterization have been aggregated and presented at the county level
or higher. EPA strongly cautions that census-tract level estimates are not reliable, and
results for individual census tracts are not presented. EPA chose the county level of
resolution for the assessment for three reasons. First, the inventory data for some
pollutants and source sectors was only available at a county level resolution. Second,
uncertainties inherent in other model input data (e.g., terrain, meteorology) and the
simplifying assumptions made in the models (e.g., transformation chemistry) themselves
rendered estimates at the census-tract level highly uncertain. Third, census-tract level
exposure estimates included additional variability (e.g., microenvironment data, human
activity pattern data) that is introduced only at local levels. For these reasons, EPA's
confidence in the accuracy of estimates for any given census tract is low.
2.4.2 Time-Frame for Exposure
The initial national-scale assessment focused on average yearly exposures for effects
other than cancer, and assumed lifetime exposures based on annual averages for
carcinogenic effects. Subchronic and acute exposures were not included in the initial
national-scale assessment. The national-scale assessment excluded acute and subchronic
exposures because of the nature of the emissions data, which contained only yearly total
emissions. If the emission inventories are later expanded to cover short-term (e.g.,
hourly, daily) emission rates, EPA may incorporate shorter exposure times into future
national-scale assessments.
2.4.3 Rationale for Aggregating Air Toxics and Sources
Exposure to air toxics from all sources is determined by a multiplicity of interactions
among complex factors, including the location and nature of the emissions, the existence
of multiple sources, local climate, location of receptor populations, and the specific
behaviors and physiology of those populations. Risks associated with these exposures
are influenced by the particular combination of air toxics that people actually inhale, and
the chemical and biological interactions among those air toxics.
Because of this high level of complexity, the magnitude of risks associated with
inhalation of air toxics can be usefully depicted by aggregating risk across both
substances and sources. Given the goals of this assessment, and the purposes for which
EPA intends to use it, currently available risk assessment data, tools, and guidance were
judged sufficient.
2.4.4 Details of the Conceptual Model
11
-------
The following subsections include summary descriptions of the risk dimensions and
elements of the national-scale assessment, as recommended by EPA's Cumulative Risk
Assessment Guidance [4]. The conceptual model for the national-scale assessment
appears in Figure 4.
2.4.4.1 Sources
The dispersion modeling, from which the exposure assessment and risk characterization
were developed, included stationary and mobile source emissions for the contiguous US,
Puerto Rico, and the Virgin Islands.
By limiting the exposure assessment and risk characterization of the initial national-scale
assessment to sources having data within the available emission inventories, EPA
excluded all non-inventoried sources. This limitation effectively excluded releases (1)
from many natural processes, (2) from indoor sources (e.g., paints, carpets, etc.), and (3)
from surface water, groundwater, or soil. While EPA takes these releases and their
potential to cause adverse health effects seriously, EPA lacked adequate model inputs
(i.e., data on substance identities and release rates) needed to quantify them in the
assessment.
2.4.4.2 Stressors
The initial national-scale assessment encompassed the pollutant set described in section
2.5.1. Later national-scale assessments may expand to cover additional Clean Air Act air
toxics, to the extent that available emission and toxicity data allow doing so. EPA chose
to limit the initial national-scale assessment to this pollutant set for two reasons. First,
these air toxics, in aggregate, appeared highly likely to encompass most of the total air
toxics-related risk to human populations [2]. Second, the initial national-scale
assessment, in combination with similar future assessments, will serve as an important
vehicle to fulfill EPA's assessment commitments under the Integrated Urban Air Toxics
Strategy, which is focused specifically on these air toxics.
2.4.4.3 Pathways/Media
The dispersion modeling step of the assessment included evaluation of the transport of
particles and gases through the air to receptors within 50km of sources. Atmospheric
transformation and losses from the air by deposition were included in the modeling,
where data permit. For 13 pollutants with available ambient monitoring data, background
concentration data due to sources located more than 50 km away were included. These
background estimates for each pollutant were assumed constant across the US. The
assessment excluded accretion in water, soil, or food associated with deposition from air,
and bioaccumulation of air toxics in tissues. Although EPA takes potential transport of
air toxics into other media very seriously, refined tools to model multipathway
concentrations on the national scale are not yet readily available for use for many
pollutants. Future local- or urban-scale assessments will include multipathway
calculations, which may be added to national-scale assessments when adequate models
and input data become available.
2.4.4.4 Exposure Routes
12
-------
The national-scale assessment focused on exposures due to inhalation of ambient air.
Human receptors were modeled as they moved within 37 separate microenvironments
such as residences, offices, schools, outdoor work sites, and automobiles. The exposure
assessment estimated air concentrations of each substance within each microenvironment,
using the outdoor concentration, proximity of the microenvironment to sources, time of
day, air exchange rate, and other. Human activities (e.g., exercising, sleeping) were
reflected in the assessment by the amount of time individuals spend in each
microenvironment.
The national-scale assessment excluded human exposures via ingestion or dermal
contact. This was a consequence of the lack of multipathway models suitable for
calculations at the national scale. As modeling tools become available to estimate
transfers of substances from air to other media on a national scale, future national-scale
assessments may include dermal and ingestion exposures.
2.4.4.5 Subpopulations
The national-scale assessment characterized risks to 40 distinct human subpopulations,
divided into five life stage cohorts, two genders, and four racial/ethnic cohorts. Figure 2-
3 shows the 40 possible cohorts chosen for this assessment. These cohorts were selected
to mirror tract-level demographic census data, to support selection of a representative
model population for each census tract. Life stages that were separately assessed
included children aged 5 or less, children aged 6-11, children aged 12-17, adults aged 18-
65, and adults aged 65 or greater. Racial/ethnic groups included African American,
Caucasian, Hispanic, and Asian and other. Exposures and risks are estimated separately
for each of the 40 cohorts in each census tract. Within each census tract, the proportion
of the total population from each cohort was tailored to demographic census data for that
particular tract and presented as the "average cohort" for that tract. Graphs of estimated
exposure and risk provide various percentiles only at the county level or higher, for the
general population.
2.4.4.6 Non-Human Receptors
The initial national-scale assessment excluded non-human receptors (e.g., wildlife and
native plants). This limitation resulted from the extreme complexity of considering
potential adverse ecological impacts to the multiplicity of ecosystems that exist within
such a large area. Future local- and urban-scale assessments may be expanded to include
non-human receptors, contingent on the availability of necessary resources, data, and
methodologies. However, EPA does not envision including non-human receptors in
future national-scale assessments unless greatly improved models and tools become
available.
2.5 Stakeholder Involvement
EPA has worked with the following groups as stakeholders for NATA activities in
general, and for the initial national-scale assessment in particular: state, local, and tribal
governments; industry; small businesses; and public interest groups.
As part of the outreach efforts on NATA and the national-scale assessment, EPA has held
informal discussions with several stakeholder groups, including representatives from the
13
-------
state and Territorial Air Pollution Program Administrators and the Association of Local
Air Pollution Control Officials (STAPPA/ALAPCO), the National Environmental Justice
Advisory Council (NEJAC), the Clean Air Act Advisory Committee (CAAAC), the U.S.
Conference of Mayors, Congressional representatives, tribal air contacts, and industry
groups.
In addition, on October 18 and 19, 1999, EPA held two public meetings in Washington
D.C. which were attended by representatives of regulatory, public interest, and business
and industry groups and associations. EPA sought input from interested stakeholders as
to the most appropriate and effective ways to present the results of the initial air toxics
assessment. Participants discussed the components of EPA's current air toxics program
provided input on approaches to presenting initial results from the national-scale air
toxics assessment activities. The Agency emphasized its desire to clearly communicate
to the public and its regulatory partners both what can be learned from such national-
scale assessments, as well as the limitations and uncertainties of such information.
Through comments received at the October meetings, and in follow-up comment letters
received after the meeting, stakeholders provided ideas, cautions, criticisms, and other
substantive input. EPA has factored this input into plans for the presentation of national-
scale assessment results. During April and May 2000, EPA conducted a six-week
preview of the results of the ambient concentration modeling step of the national-scale
assessment with its regulatory partners at the state, local and tribal level. The purpose of
this preview was to enable EPA's regulatory partners to provide feedback on the
appropriate interpretation and communication of the results of the initial national-scale
assessment and to provide a quality assurance assessment of the results.
EPA is currently seeking input from these same stakeholder groups on interpretation and
communication of the exposure assessment and risk characterization results, and on
quality assurance of these results.
2.6 Peer Review Activities for the Initial National-Scale
Assessment
Many of the national-scale assessment components described above have undergone
peer-review and public review, either on their own or as critical elements of other
analyses. EPA also instituted a peer-review for: (1) the planning and scoping document
for the initial national-scale assessment, and (2) for this draft NATA national-scale
assessment report, using procedures recommended by the EPA Science Policy Council
[5].
2.6.1 Past Reviews of National-Scale Assessment Components
2.6.1.1 List of Urban Air Toxics
In 1997, EPA developed an initial list of potential candidate urban air toxics using a risk-
based ranking methodology, and conducted a public review of the national emissions
14
-------
inventory for those candidate air toxics. During January 1998, a panel of technical
experts from outside the U.S. EPA reviewed the air toxics ranking methodology and
analysis. The final methodology used to select the 33 urban air toxics identified in the
Integrated Urban Air Toxics Strategy incorporated revisions based on comments raised in
the 1998 peer review. EPA also received public comments on the draft list of urban air
toxics [6], which led to further modifications of the identification methodology and the
underlying data inputs.
2.6.1.2 1996 National Toxics Inventory
EPA made the draft 1996 National Toxics Inventory (NTI) and documentation available
for review and comment between April and August 1999, and received extensive
comments and revisions from industry, state and local agencies, and others. Before
incorporation into the final NTI, revisions were subjected to a rigorous review process to
ensure internal consistency. Further details of the review process are described in section
3.2.1. In addition to review and comment on the 1996 NTI, methods and assumptions
used to develop toxic emission estimates for benzene, 1,3-butadiene, formaldehyde, and
acetaldehyde from on-highway mobile sources were described in an EPA technical report
[7] and peer reviewed, in addition to undergoing review by industry groups, state
agencies, and local governments (peer review comments are available at:
http://www.epa.gov/otaq/toxics.htm).
2.6.1.3 ASPEN National Dispersion Model
In 1996, the EPA Science Advisory Board (SAB) reviewed the Cumulative Exposure
Project methodology [5], including the underlying scientific basis for the project and
specific details of the modeling methodology. This review included the use of the
ASPEN dispersion model in developing estimates of ambient concentrations of air toxics.
The SAB review found that the overall conceptual framework and underlying scientific
foundation was sound but stressed the importance of comparing the ASPEN predictions
with measured ambient concentrations.
2.6.1.4 Hazardous Air Pollutant Exposure Model (HAPEM)
The EPA Office of Transportation and Air Quality (formerly the Office of Mobile
Sources) has used earlier versions of HAPEM, the exposure model used for this
assessment, in analyses of exposure to carbon monoxide and vehicle-related air toxics.
Two studies of air toxics exposure associated with vehicle emissions in 1993 and 1999
[9], were peer-reviewed by a panel of independent experts. Details on this version of
HAPEM and the peer-review comments can be found at:
http://www.epa.gov/oms/toxics.htm.
2.6.1.5 EPA Risk Assessment Guidelines
Since cancer and non-cancer health impacts cannot be directly isolated and measured,
EPA and others have spent more than two decades developing an extensive set of risk
assessment methods, tools and data that serve the purpose of estimating health risks for
many Agency programs. EPA develops and publishes its risk assessment methods in the
15
-------
form of risk assessment guidelines that have been extensively peer-reviewed, are widely
used and understood by the scientific community, and continue to expand and evolve as
scientific knowledge advances.
2.6.1.6 Microenvironment Factors
The microenvironment (ME) factors (described in section 3.2.3) represent the
relationship between the ambient concentration and the microenvironment of interest.
The ME factor approach for determining microenvironment concentrations is the only
computationally feasible way (in lieu of indoor/outdoor mass balance models) for
predicting exposures on a national scale. To help assure the credibility of the factors, and
their appropriateness for use in the national-scale assessment, the ME factors underwent a
peer review consistent with EPA's peer-review guidance [5].
2.6.1.7 ASPEN Results
Ambient concentration estimates for the pollutant set, calculated by the initial
Assessment System for Population Exposure Nationwide (ASPEN) model runs, were
reviewed by state, local, and tribal authorities to find and correct potential input errors,
and to obtain further input in developing useful presentation formats. The ASPEN model
was subsequently re-run to incorporate the corrections.
2.6.2 Review of Planning and Scoping Document for the Initial
National-Scale Assessment
In July 2000, six technical experts from outside the U.S. EPA completed a peer review of
the draft planning and scoping document. The reviewers were asked to consider the
appropriateness of approaches used to (1) process the state-derived National Toxics
Inventory for dispersion modeling, (2) estimate ambient concentrations using the
Assessment System for Population Exposure Nationwide (ASPEN) model, (3) estimate
human inhalation exposures using the Hazardous Air Pollutant Exposure Model version 4
(HAPEM4), and (4) estimate, aggregate, and interpret associated cancer and non-cancer
risks. The detailed charge to the reviewers, a summary of the reviewers' comments, and
the EPA's response to comments are provided in Appendix A.
16
-------
3 Methods
3.1 Introduction
EPA conducted the initial national-scale assessment to demonstrate an approach for
characterizing air toxics risks nationwide. This assessment expanded on the approach
first used by the EPA in the 1998 Cumulative Exposure Project (CEP) [8], in which
annual ambient concentrations of 155 air toxics were estimated for approximately 61,000
census tracts in the contiguous US. In the CEP, estimated concentrations, which are
based on 1990 emission estimates, are assumed to equal human exposure concentrations
for all persons living in each census tract, and are evaluated, by comparing to currently
available "health benchmarks," to assess potential health risks. In this initial national-
scale assessment, EPA augmented the CEP approach by: (1) using a 1996 emission
inventory compiled by a more rigorous method; (2) incorporating human demographics
and behavior into the development of exposure estimates; (3) developing a
comprehensive risk characterization, including estimating cumulative risk associated with
multiple air toxics; and, (4) expanding geographic coverage to include Puerto Rico and
the Virgin Islands.
The four major components of the national-scale assessment were as follows:
1. Compiling a national emissions inventory for 1996 of the pollutant set emissions
from outdoor sources. The types of emissions sources in the inventory include
major stationary sources (e.g., large waste incinerators and factories), area sources
(e.g., dry cleaners, small manufacturers, consumer products), and both onroad and
nonroad mobile sources (e.g., cars, trucks, and boats).
2. Estimating 1996 air toxics ambient concentrations. EPA used the ASPEN air
dispersion model and the 1996 inventory to estimate annual average ambient
concentrations for the pollutant set (see pollutant names and corresponding CAS
numbers listed in Table 3-1) across the contiguous United States, Puerto Rico, and
the Virgin Islands. As part of this modeling exercise, estimated concentrations
were compared to available ambient monitoring data as a partial evaluation of
model performance. Details of this comparison are provided in Appendices I & J.
3. Estimating 1996population exposures. Rather than assume that all people within
each census tract were exposed to a concentration that was equivalent to the
ASPEN prediction for that census tract, EPA used an inhalation exposure model
and the ambient concentrations from item 2 above to estimate human exposure to
the pollutant set across the contiguous United States (and Puerto Rico and the
Virgin Islands). Exposure modeling was an important step in this assessment
because it provided more realistic estimates of population exposures to air toxics
from outdoor emission sources by accounting for the time people spend indoors
and in other microenvironments (e.g., in vehicles), patterns of movement (e.g.,
commuting between home and work locations), and activity levels. Detailed
information about the structure, function, and use of the HAPEM4 exposure
model is provided in the user's guide (Appendix B).
17
-------
4. Characterizing potential public health risks. EPA characterized potential
population health risks associated with inhalation of air toxics, including both
cancer and noncancer effects, using available information on the health effects of
each pollutant, current Agency risk assessment and risk characterization
guidelines, and estimated population exposures. This characterization quantified,
to the extent possible, potential cumulative risks to public health due to inhalation
of air toxics from outdoor emission sources, discussed the uncertainties and
limitations of the assessment, and identified other potential risks to public health
from exposures that are beyond the scope of this quantitative assessment.
The approach outlined for the national-scale assessment was fundamentally based on
national-scale modeling techniques to estimate ambient pollutant concentrations and
Table 3-1. Pollutant set for the initial national-scale assessment.2
Pollutant CAS #
Acetaldehyde
Acrolein
Acrylonitrile
Arsenic compounds
Benzene
Beryllium compounds
1,3 -Butadiene
Cadmium compounds
Carbon tetrachloride
Chloroform
Chromium compounds
Coke Oven Emissions
1 , 3 -Dichloropropene
Diesel paniculate matter
Ethylene dibromide (1,2-dibromoethane )
Ethylene dichloride (1,2-dichloroethane)
Ethylene oxide
Formaldehyde
Hexachlorobenzene
Hydrazine, hydrazine sulfate
Lead compounds
75070
107028
107131
71432
106990
56235
67663
8007452
542756
106934
107062
75218
50000
118741
302012
This list is the list of 33 urban air toxics, except that dioxins have been removed from the list and diesel
paniculate matter has been added.
18
-------
Pollutant CAS #
Manganese compounds
Mercury compounds
Methylene chloride
Nickel compounds
Poly chlorinated biphenyls (PCBs)
Polycyclic Organic Matter (POM) (including PAHs)
Propylene dichloride (1,2-dichloropropane)
Quinoline
1, 1,2,2-Tetrachloroethane
Tetrachloroethylene (perchloroethylene)
Trichloroethylene (TCE)
Vinyl chloride
75092
1336363
78875
91225
79345
127184
79016
75014
population exposures nationwide. While such computer models necessarily required
simplifying assumptions and introduced significant uncertainties, they were needed to
conduct such a large-scale assessment since direct measurements of ambient air toxics
concentrations are limited, and direct personal exposure measurements are relatively rare.
Such measurements are available for only a subset of air toxics in relatively few locations
and for small study populations. Those ambient air toxics data used in the model to
monitor comparison were taken from the air toxics archive. The air toxics archive is a
collection of ambient air toxics data that is compiled directly from State and local
agencies and which is supplemented with any other ambient air toxics data on AIRS. The
specific data used in the model to monitor comparison can be found at:
www.epa.gov/ttn/uatw/nata/mtom_pre.html#data. Although EPA is working to expand
the number and locations of ambient air toxics monitors and the study of personal
exposures, direct measurement of air toxics concentrations or exposures is not currently
practical for all air toxics of interest across all areas of the country. As such
measurement data become available over time, they can and will be used to evaluate the
models so as to better understand some of the uncertainties in such assessments and to
improve modeling tools.
In describing what this assessment included, it is also important for the reader to
recognize potentially important sources and pathways of risks to public health that were
considered beyond the scope of this assessment, and were therefore not included. First,
while EPA recognizes that indoor sources of air toxics emissions likely contribute
substantially to the total exposures that people experience for a number of these air
toxics, additional work is needed to better develop our tools for assessing these indoor
sources of exposure on a national scale, that was not a part of this assessment. Further,
for a subset of air toxics that persist or bioaccumulate in the environment, dietary intake
(e.g., from eating contaminated fish) likely contributes much more to total exposure than
does the inhalation pathway that was be addressed in this assessment. Table 3-2 lists
19
-------
those urban air toxics believed most likely to present important multipathway risks.
These and other important aspects of total population exposures to air toxics will be
addressed more fully over time as part of future NATA assessment activities, as more
comprehensive data and assessment tools become available. As a result of excluding
these potentially-important contributors to total risk from this assessment, the results of
this assessment should always be viewed with its limitations in mind, and the risks
interpreted as only a portion of the total risks which may be associated with these air
toxics.
Table 3-2. Urban air toxics believed to present risks from multipathway exposure.
Hexachlorobenzene
Lead compounds
Mercury compounds
Poly chlorinated bipheny Is (PCBs)
Polycyclic Organic Matter (including 7-PAH)
2,3,7,8-TCDD (dioxin)
3.2 Exposure Assessment
EPA's 1992 guidelines for exposure assessment [10] establish a broad framework for
exposure assessments by describing the general concepts of exposure assessment,
including definitions and associated measurement units, and by providing general
guidance on the planning and conduct of an exposure assessment. The guidelines also
provide information on presenting the results of the exposure assessment and
characterizing uncertainty.
The guidelines define human exposure as contact with a chemical or agent at the visible
external boundary of a person, including skin and openings into the body such as mouth
and nostrils (but not necessarily contact with exchange boundaries where absorption may
take place, such as skin, lung, and gastrointestinal tract). Therefore, an exposure
assessment is the quantitative or qualitative evaluation of contact, and includes such
characteristics as intensity, frequency, and duration of contact. Often, an assessment also
will evaluate the rate and route at which a chemical crosses the external boundary (dose)
and the amount absorbed (internal dose). The numerical output of an exposure
assessment may be either exposure or dose, depending on the purpose of the evaluation
and availability of appropriate data.
An exposure assessment has three major components: source characterization,
environmental fate and transport characterization, and characterization of exposure.
These components are discussed individually below.
3.2.1 Source Characterization: Emission Inventories
In the first step of an exposure assessment for air toxics, the specific air toxics emitted
20
-------
and the sources of their airborne emissions are determined. Data are collected on the
emission rates of the pollutants and the release parameters of the source. Knowledge of
the emission rate and release characteristics enables the pollutant fate and transport to be
estimated.
Ideally, the emission estimates are from direct measurements of representative source
emissions. Although such measurements are likely to provide the most accurate data for
an emission source, these data are typically not available because such sampling is often
too time- and resource-intensive. When specific emission measurements are not feasible
or available, other emission estimation methods, including material balances and
emission factors, are sometimes used as an alternate method. Emission factors indicate
the quantity of a pollutant typically released to the atmosphere for a particular source
operation, and are usually considered to be representative of an industry or emission type
as a whole. Each approach to estimating emissions, including use of direct measurement
data, has an inherent level of uncertainty, which adds to the overall uncertainty of a risk
analysis.
Depending on the analysis, source and emissions data can be derived from broad-scale
emission inventories, specific data collection efforts with particular industries, or
information from regional, state, or local agencies. Other information, such as
geographic location of release points, temporal pattern of emissions (e.g., periodic
"puffs" vs. constant emission rates), and release height may be necessary, depending on
the level of detail needed or types of exposure examined in the assessment.
3.2.1.1 Approach
The majority of emissions information used in the national-scale assessment was
extracted from EP A's National Toxics Inventory (NTI), which contains estimates of the
emissions of the 188 air toxics listed in section 112(b) of the CAA. In addition to the
NTI, EPA's National Emission Trends (NET) inventory was also used as a source of
emissions data for air toxics precursors in order to estimate air toxics generated through
secondary formation in the atmosphere and for diesel paniculate estimates for Puerto
Rico and the Virgin Islands. Diesel paniculate emission estimates for the contiguous
U.S. were extracted from an inventory developed separately by EPA (see further details
on this in section 3.2.1.2).
Before the emissions data from the NTI, NET, and diesel particulate inventories could be
used as input to dispersion modeling, the emissions data required significant preparation.
Some of this preparation work occurred during the compilation of the inventories and
some occurred in the Emissions Modeling System for Hazardous Air Pollutants (EMS-
HAP) which is a series of computer programs that process emission inventory data for
subsequent air quality modeling. Appendix C provides the EMS-HAP User's Guide and
the methodologies used to process the emission data for this assessment. The necessary
inventory preparation steps are described below:
• Compiling detailed air toxics, air toxics precursors, and diesel PM emissions
inputs for all known stationary and mobile sources. The sources of this data are
described in section 3.2.1.3.
21
-------
• Performing quality assurance of the point source inventory location and stack
parameter data and incorporating defaults for missing or erroneous data where
possible. In some cases, missing or erroneous information can be found using the
Toxics Release Inventory database. EMS-HAP defaults missing point source
locations information to the centroid of the zip code (if zip code information is
available) or to a census tract within the county. Where possible, stack
parameters are defaulted based on the source category classification code or
standard industrial classification code. Where not possible (lack of code in the
inventory or lack of data for a particular SCC or SIC code, stack parameters) are
set to conservative defaults (10 meter stack height, 1 meter stack diameter, 1
meter/second velocity, and a temperature of 295 Kelvin).
• Grouping individual pollutant species into compound groups. The NTI contains
approximately 400 different species representing the 188 air toxics listed in
section 1 12(b) of the CAA. Many of the species belong to compound classes.
Grouping of these species is necessary for many reasons. One reason is that the
individual chemical species belonging to groups are not geographically
representative. For example, "lead oxide" may have been reported in just a few
counties, whereas other counties aggregated their lead oxide emissions into "lead
compounds." Grouping allows for pollutants with similar characteristics to be
modeled together for purposes of efficiency. For example, specific lead species
and compounds reported as the broad group "lead compound" are grouped to be
subsequently modeled as "lead compounds-fine" and "lead compounds-coarse."
These groups allow ASPEN to distinguish between the different deposition rates
for fine and coarse lead particulates.
• Assigning each pollutant to a reactivity class (high, medium, low, etc.) or
particulate size class (fine or coarse particulate) to allow for the ASPEN model to
perform decay and deposition calculations. These reactivity classes are based on
the rate of reaction of the pollutant with OH and NO3 radicals and had previously
been established for the gaseous air toxics in the Cumulative Exposure Project.
• Temporally allocating emission values to eight annual 3-hours emission rates.
Emissions are temporally allocated based on the type of source using a database
of temporal profiles by source classification code. The majority of the profiles are
from a database originally developed for regional emission modeling studies
under the National Acid Precipitation program. The factors for the National -Scale
Air Toxics Assessment consolidate the seasonal and day-of-week classes in order
to reflect hourly activity for an annually averaged day.
• Grouping all source categories into the following four source sectors: 1) major, 2)
area and other, 3) on-road, 4) non-road. These sectors and the methodology for
grouping are discussed below.
All of the raw inventory inputs for this assessment exist as estimates for point
22
-------
sources, non-point stationary sources, and mobile sources. "Point" sources
provide emissions data at the facility and sub-facility level and include location
coordinates (e.g., latitude and longitude). "Non-point" stationary source and
"mobile" source data exist as emissions estimates for an entire source category
aggregated to the county level. Inventory data files for these different types of
sources are generally maintained separately and include different data elements.
For the purpose of aggregating air toxics emission sources in the national-scale
assessment, all emissions inventory inputs were grouped into four sectors:
"major," "area and other," "onroad," and "nonroad." Dispersion modeling was
performed for each sector separately, so that concentration estimates of each air
toxic could be attributed to each sector. Each sector is further defined as follows:
Major sources are large stationary sources that emit more than 10 tons per year of
any listed air toxic (CAA, section 112(b)) or a combination of listed air toxics of
25 tons per year or more. Typical examples of major sources include electric
utility plants, chemical plants, steel mills, oil refineries, and large hazardous waste
incinerators. These sources may release air toxics from equipment leaks, when
materials are transferred from one location to another, or during discharge through
emissions stacks or vents.
Area and Other sources are smaller stationary sources that emit less than 10 tons
per year of a single air pollutant or less than 25 tons per year of a combination of
air toxics. The emission inventory includes facility data for some area sources
and aggregated emission estimates at the county level for the remaining area
sources. Typical examples of area sources include neighborhood dry cleaners and
gas stations. Though emissions from individual area sources are often relatively
small, collectively their emissions can be of concern particularly where large
numbers of sources are located in heavily populated areas. "Other" stationary
sources are sources that may be more appropriately addressed by other programs
rather than through regulations developed under certain air toxics provisions
(sections 112 or 129) in the Clean Air Act. Examples of other stationary sources
include wildfires and prescribed burning, which have emissions that are being
addressed through the burning policy agreed to by the EPA and the USDA. For
this assessment, the "area" and "other" sectors have been combined in the
calculations and presentation of the current national-scale assessment
Onroad mobile sources comprise vehicles used on roads and highways (e.g., cars,
trucks, buses).
Nonroad mobile sources are all remaining mobile sources (e.g., trains,
lawnmowers, construction vehicles, farm machinery).
Major and area source facilities are drawn from the "point" source inventory files,
meaning those with known geographic locations (i.e., latitude and longitude).
Area and other source categories that are aggregated as county-level emissions are
drawn from the "non-point" source inventory files, meaning those stationary
sources that do not have location coordinates but instead exist as county-wide
23
-------
total emissions by source category. Onroad and nonroad sources exist as distinct
sectors in the "mobile" source inventories and are also aggregated to the county
level.
Spatially allocating county-level emissions to the census-tract level using
surrogate data, such as population, industrial land or roadway miles, which are
available at the census-tract level. A description of the available surrogates for
use with EMS-HAP is shown in Table 3-3. Tract-level emissions for a source
category are computed based on the percentage of the matching surrogate in the
tract for that county. For example, the consumer products usage source category
is matched to population. If 10 percent of the population of the county is in tract
A, then tract A gets 10 percent of the county's consumer products usage
emissions.
Table 3-3. S
Code for
set of
SAFs
SAFl
SAF2
SAF3
SAF4
SAF6
SAF7
SAF8
SAF9
SAF10
SAF12
SAF13
SAF14
Surrogate
Residential land
Commercial land
Industrial land
Utility land
Sum of commercial
land and industrial
land
Farm land
Orchard land
Confined feeding
Farm land & confined
feeding
Rangeland
Forest land
Rangeland & forest
land
patial Allocation Factors (SAP) Developed for EMS-HAP
Definition
USGS land use categories: Residential, plus
one-third of mixed urban and built-up land plus
one-third of other urban and built-up land
USGS land use categories: Commercial and
services, plus one-half of industrial and
commercial complexes, plus one-third of mixed
urban and built-up land plus one-third of other
urban and built-up land
USGS land use categories: industrial, plus
one-half of industrial and commercial complexes,
plus one-third of mixed urban and built-up land,
plus one-third of other urban and built-up land
USGS land use category: "transportation,
communications, and utilities"
Sum of commercial land and industrial land, as
defined above
USGS land use category: "cropland and pasture"
USGS land use category: "orchards, groves,
vineyards, nurseries, and ornamental horticultural
areas"
USGS land use category "confined feeding"
USGS land use categories "cropland and pasture"
plus "confined feeding"
USGS land use categories: "herbaceous
rangeland" plus "scrub and brush" plus "mixed
rangeland"
USGS land use categories: "deciduous forest"
plus "evergreen forest" plus "mixed forest land"
Sum of rangeland and forest land, as defined
above
Origin of Data
mid-70's to 80's
mid-70's to 80's
mid-70's to 80's
mid-70's to 80's
mid-70's to 80's
mid-70's to 80's
mid-70's to 80's
mid-70's to 80's
mid-70's to 80's
mid-70's to 80's
mid-70's to 80's
mid-70's to 80's
How EPA developed
the set of SAFs
from CEPa'b
from CEPa'b
from CEPa-b
from CEPa'b
land use data from
developers of CEPa'b, SAF
recomputed
from CEPa'b
from CEPa-b
from CEPa'b
from CEPa'b
from CEPa'b
from CEPa-b
from CEPa-b
24
-------
Code for
set of
SAFs
SAP 15
SAF17
SAP 18
SAF19
SAF20
SAF21
SAF22
SAF24
SAF25
SAF26
SAF27
SAF28
SAF29
Surrogate
Water
Mining & quarry land
I/population density
I/population density
Population
Railway miles
Roadway miles
50% Population &
50% roadway miles
25% Population &
75% roadway miles
Tract area
Urban: Inverse
population density
Rural:farmland
Urban: population
Rural: tract area
Sum of farmland and
orchard land
Definition
US Census category: water area
USGS land use category: "strip mines, quarries,
and gravel pits"
Inverse of: census tract population (defined
above) divided by census tract area. Tracts with
zero population assigned spatial factors of zero.
Inverse of: census tract population (as defined
above) divided by census tract land area. Tracts
with zero population assigned tract population of
one.
U.S. Census category: 1990 residential
population
Total railway miles, as reported in TIGER/Line
Total miles of all roadway types in each census
tract, as reported in TIGER/Line
Surrogate based equally on population fraction
and on roadway mile fractions for each of four
roadway types
Surrogate based on population fraction and
roadway mile fractions, respectively weighted by
25% and 75%, for each of four roadway types
The area of census tracts (including land and
water)
Inverse population density (18) for urban0
counties; farmland (7) for rural0 counties
Population (20) for urban0 counties; tract area for
(26) rural0 counties
Sum of farmland and orchard land, as defined
above
Origin of Data
1990
mid-70's to 80's
1990
1990
1990
1993
1993
1990-93
1990-93
1990
1990, mid-70's to 80's
1990
mid-70's to 80's
How EPA developed
the set of SAFs
from CEPa-b
from CEPa-b
from CEPa'b
population and land area
data from CEPb, SAF
recomputed.
from CEPa-b
from CEPa-b
from CEPa'b
0.5*SAF20 + 0.5*SAF22
0.25*SAF20 + 0.75*SAF22
tract areas computed from
CEP tract radiib data SAF
recomputed.
SAF 18 from CEP,
SAF 7 from CEP,
urban/rural county
designations from 1990 and
1996 census data
SAF 20 from CEP,
SAF 26 from CEP,
urban/rural county
designations from 1990 and
1996 census data
land use data from
developers of CEPa'b, SAF
recomputed.
1 except that changes were made to SAFs in Halifax and South Boston in Virginia
b except for census tracts in the Virgin Islands and Puerto Rico which were not modeled in the CEP
0 the designation of urban and rural counties is based on data from the U.S. Census Bureau as follows: a county is considered "urban"
if either: 1) it includes a metropolitan statistical area with a population greater than 250,000; or, 2) the U.S. Census Bureau designates
more than fifty percent of the population as "urban."
3.2.1.2 Scope
The NTI was the source of the majority of emissions inputs to the national-scale
assessment. It contains estimates of air toxics emitted from many anthropogenic source
categories for 188 hazardous air pollutants, for the 50 U.S. states, District of Columbia,
Puerto Rico and Virgin Islands. With a few exceptions (e.g., wildfires), the NTI does not
include emissions of air toxics from natural sources, indoor sources, or accidental
25
-------
releases. Sources which are not included in the NTI were also excluded from the initial
national-scale assessment.
For the purposes of the national-scale assessment, the pollutants were limited to 32 of the
33 air toxics included in EPA's Integrated Urban Air Toxics Strategy [2] (dioxins were
not included) and diesel particulate matter. The geographic domain was limited to the
contiguous 48 states, District of Columbia, Puerto Rico and Virgin Islands. Background
documentation on the development of the 1996 NTI are included in Appendix D or can
be accessed on the EPA web site at http://www.epa.gov/ttn/chief/nti. The detailed 1996
NTI data files are available at ftp://ftp.epa.gov/Emislnventory/nti_96.
The NET provides EPA's latest estimates of national emissions for criteria air pollutants:
carbon monoxide (CO), nitrogen oxides (NOx), volatile organic compounds (VOCs
[excludes certain non-reactive organic compounds]), sulfur dioxide (SO2 ), particulate
matter less than 10 microns (PM10), particulate matter less than 2.5 microns (PM2.5),
and lead (Pb). [72]
To account for secondary formation of volatile air toxics species, point, non-point
stationary and mobile source emission data were input to the ASPEN model for non-air
toxics VOC species resulting from a speciation of the 1996 NET (Version 3) inventory.
These data were included to account for secondary formation of volatile air toxics
species. In addition, NET data were used to estimate diesel particulate emissions for
Puerto Rico and the Virgin Islands. The raw NET data are available at
ftp://ftp. epa. gov/Emi slnventory/net 96.
The diesel PM emissions data used for the 1996 national-scale assessment were extracted
from an inventory developed as part of the rulemaking on Heavy-Duty Engine and
Vehicle Standards and Highway Diesel Fuel Sulfur Control Requirements [13}. This
inventory is based on Federal Highway Administration estimates of truck operation,
estimates of the distribution of fuel type and weight classes of truck from the EPA Office
of Transportation Air Quality (OTAQ), and emission factors provided by OTAQ. The
nonroad emissions in this inventory were derived from OTAQ's draft June 2000
NONROAD model run for 1996. Both of these PM inventories reflect changes in
methods and data sources since the release of versions used for the 1996 NET and 1996
NTI. Time did not allow for estimates of other pollutants from diesel vehicles and
equipment to be revised accordingly, but an exploratory analysis indicated that the effect
on estimates of other air toxics would not have been large. The diesel particulate
emissions are all found in the onroad and nonroad source sectors. More details on the
diesel PM inventories can be found at http://www.epa.gov/otaq/hdmodels.htm.
The NET, diesel particulate, and early versions of the 1996 NTI inventories all contain
emissions estimates for the 50 states. The geographic domain of the national-scale
assessment included the contiguous United States, Puerto Rico, and the Virgin Islands.
In all cases, emissions for the territories included in the assessment were derived in part
or in total via extrapolation of emissions estimates from surrogate U.S. locations.
26
-------
3.2.1.3 Sources of Data
EPA prepared the 1996 NTI using various sources of data. The five primary sources of
1996 NTI data are: (1) state and local air toxics inventories developed by state and local
air pollution control agencies, (2) existing databases related to EPA's Maximum
Achievable Control Technology (MACT) program which requires emission standards
under Section 112(d) of the CAA. (www.epa.gov/ttn/uatw/eparules.html) (3) Toxics
Release Inventory (TRI) data (www.epa.gov/tri/), (4) emissions estimated by using
mobile source methodologies developed by experts in EPA's Office of Transportation
and Air Quality, and (5) emission estimates for 30 of 500 non-point source categories
generated using emission factors and activity data. Much of the state/local, TRI, and
MACT emissions data may have been generated by the sources themselves. The 1996
NTI is the first national modeling emission inventory constructed using state and local air
toxic inventory data and containing stationary and mobile source data.
To improve the quality of the draft 1996 NTI, EPA requested comments on the inventory
from state and local agencies, industry and others. The draft 1996 NTI and
documentation were available to all parties for review and comment from April 30, 1999
- August 13, 1999. A state-only preview of the ASPEN concentration results from March
31, 2000 to May 12, 2000 resulted in additional state/local revisions to the NTI. The
EPA received extensive comments and revisions from industry, state and local agencies,
and others located in 42 states, Puerto Rico, and the Virgin Islands. Figure 3-1 shows the
states and local agencies that ultimately provided inventory data, comments, and
revisions to the EPA. Forty-one states provided point source revisions, 27 states
provided non-point stationary source revisions and 18 states provided mobile source
revisions. All requested revisions were evaluated and incorporated as appropriate into
the NTI prior to final ASPEN modeling. Documentation for all emissions estimates in
the 1996 NTI is available in Appendix D.
3.2.1.3.1 Point Source Emissions
In compiling point source emissions information for the NTI, preference was given to
state- and locally-generated information where available except for utility mercury
emissions and municipal waste combustion emissions. While agencies are neither
required to compile air toxics emissions data nor to submit such data to EPA, most states
did respond to EPA's invitation to do so. EPA did not apply many quality checks on
state/local data before incorporating it into the NTI, but in the course of three rounds of
state/local review of the draft NTI and national-scale assessment results, many anomalies
were noted by EPA or the state/local agencies and resolved by corrections or deletions.
Where state or local agency data were not provided, existing emissions data from EPA's
regulatory development (MACT) databases were utilized. If these databases differed in
pollutant coverage, MACT data were used to add any missing emissions in order to be as
comprehensive in pollutant coverage as possible. If neither of state/local or MACT
sources contained information for a known stationary source, the NTI used estimates
based on information from EPA's TRI.
Most state agencies participated in the development of the 1996 NTI. Forty-seven of the
50 states, Puerto Rico, the Virgin Islands, and the District of Columbia participated in the
development of the 1996 NTI either by providing emissions data or by reviewing the
27
-------
draft inventory. Thirty-six states initially provided draft 1996 air toxics emissions
inventories. The data collected from the state or local agencies varied significantly in
terms of completeness, coverage, format, and quality. The majority of the 36 states
provided data for point sources (primarily large industrial sources which would be
defined as major sources). The number of air toxics included in the state or local agency
inventories varied. Some state and local agencies compile emission inventories for fewer
than the 188 air toxics (e.g., the RAPIDS inventory of Great Lakes states), while other
states such as California and Louisiana compile emission inventories for more than the
188 air toxics.
In addition to the 14 states that did not provide draft air toxics inventory data, several
data gaps were identified in the state databases provided. Data gaps in the state
inventories included: (1) emissions from entire counties missing from the state databases;
(2) missing emission sources; (3) lack of stack parameters; and (4) lack of facility
location data (latitude/longitude or Universal Transverse Mercator (UTM) coordinates).
In order to find missing facilities for entire states, counties, or individual facilities, EPA
compared existing state air toxics inventories first to MACT data, then to TRI data, and
then to NET data. For most facilities that appeared in both the state database and in
either the MACT or TRI database, EPA assumed that the state databases were more
accurate and, thus, no further revisions were made. This assumption of the data quality
hierarchy was necessary due to resource limitations that prohibited EPA from comparing
various emission estimates for a given facility, obtaining documentation for the disparate
estimates, and choosing the most appropriate information across data bases. In at least
one case, this hierarchy resulted in the omission of significant emissions in the national-
scale assessment. A major source of lead, a lead smelter in Missouri, does not show any
lead emissions in the version of the 1996 NTI modeled for this assessment. This resulted
because the state agency's emissions estimates for this facility were used but erroneously
did not include the lead emissions. Since the facility was already present in the NTI from
the state's inventory, TRI estimates were not added. Later, after ASPEN modeling, it
was discovered that TRI contained nearly 100 tons of lead emissions for this facility. It is
possible that other facilities and emissions could have been omitted from this assessment.
Similarly, it is possible that a facility could have been double counted if it appeared in
two or more inventories, but with different facility names. These and other potential
sources of error are discussed in subsequent sections of this report.
Because the ASPEN model requires a model-ready inventory, the association of stack
parameters and location data to each facility's emission estimate was required. If this
information was found to be missing, incorrect, or out of range, it was corrected with
defaults. Defaulting schemes were performed in both the NTI development and in the
subsequent processing in EMS-HAP. Default stack parameters associated with Standard
Industrial Classification (SIC) codes were used for emissions reported at the SIC code
level. Also default stack parameters associated with source classification codes (SCCs)
were employed for emissions reported at the SCC level. Where default parameters were
available by either SCC or SIC, the SCC took precedence.
In order to develop appropriate defaults for missing location data, first comparisons were
28
-------
made to other inventories (e.g., TRI, NET, or the Ozone Transport Assessment Group
(OTAG) inventory). Where latitude/longitude coordinates were found they were added
to the NTI. Where the facilities could not be matched to other databases, defaulting
schemes were employed to place facilities within known zip codes or counties. These
schemes are explained in the NTI documentation for major sources (Appendix D) and in
the EMS-HAP User's Guide (Appendix C). EPA assigned locations to all facilities
except for 87 facilities that were defined as "portable" in Colorado and Idaho. These
were not included in the assessment.
3.2.1.3.2 Non-Point Emissions
Where possible, EPA compiled the
non-point stationary source emissions
(which are included in the "area and
other" source sector) from the 1996
state emission data sets. The majority
of the 36 states that initially supplied
draft 1996 air toxics emission
inventories did not provide non-point
source data. EPA evaluated and
supplemented the state data sets with
non-point source data gathered during
the development of MACT standards
and with TRI data. EPA then
generated non-point emission
estimates for 30 remaining source
categories (listed in Table 3-4) by
using activity data and emission
factors and then allocating the
estimates from the national, state, or
regional level to the individual
counties. EPA was careful to avoid
duplicating emissions (e.g., emissions
from large dry cleaners included in
point source files would be subtracted
from non-point source calculations).
For example, if the non-point source
estimates were based on raw material
usage, the point source fraction
already accounted for in this fashion
was subtracted prior to calculating the
non-point source emissions.
Table 3-4. Non-Point Source Stationary
Categories with EPA-Derived Emissions.
Categories
Animal Cremation
Asphalt Paving: Cutback Asphalt
Auto body Refinishing Paint Application
Aviation Gasoline Distribution: Stage I & II
Consumer Products Usage
Dental Preparation and Use
Drum and Barrel Reclamation
Fluorescent Lamp Recycling
Food and Agricultural Products: Cotton Ginning
Gasoline Distribution Stage II
General Laboratory Activities
Geothermal Power
Hospital Sterilizers
Human Cremation
Lamp Breakage
Miscellaneous Organic Chemical Processes
Open Burning: Forest and Wildfires
Open Burning: Prescribed Burnings
Open Burning: Scrap Tires
Pesticide Application
Residential Fuel Use: Anthracite Coal
Residential Fuel Use: Bituminous and Lignite Coal
Residential Fuel Use: Distillate Oil
Residential Fuel Use: Natural Gas
Residential Fuel Use: Wood/Wood Residue
Softwood Drying Kilns
Structure Fires
Surface Coating Operations: Architectural
Surface Coating Operations: Traffic Markings
Surface Coating Operations: Industrial Maintenance
3.2.1.3.3 Mobile Sources
The EPA's Office of Transportation
and Air Quality (OTAQ) provided
direction and advice on which
emission factors and speciation profiles should be used in the development of air toxics
29
-------
emission estimates for mobile sources. The mobile source estimates were determined
using a combination of methods. For highway mobile sources, emission factors for
benzene, 1,3-butadiene, formaldehyde and acetaldehyde were modeled for 10 urban areas
and 16 geographic regions, using OTAQ's MOBTOXSb model [14]. The urban areas
and geographic regions modeled were selected to encompass a broad range of inspection
and maintenance (I/M) programs, fuel parameters, and temperature regimes. The intent
of the selection process was to best characterize the different combinations needed to
perform accurate nationwide toxic emissions estimates. Every county in the U.S. was
then "mapped" to one of these modeled areas or regions (i.e., the emission factor for the
modeled area was also used for the area "mapped" to it). Mapping was done based on a
combination of geographic proximity, I/M program, and fuel control programs. The
resulting county level emission factors were then multiplied by county-level VMT
estimates. To estimate emissions of other air toxics from highway mobile sources, data
from speciation profiles were applied to VOC emissions for gaseous air toxics and POM,
and to PM emissions estimates for metal emissions and dioxins (although dioxin
emissions were subsequently removed from this study). Where emissions of gaseous
compounds were impacted by the use of reformulated gasoline or winter oxygenated
gasoline, these impacts were accounted for at the county level.
The 1996 NTI contains nonroad mobile emission estimates for 2- and 4-stroke gasoline-
powered engines, diesel engines, aircraft, locomotives and commercial marine vessels.
For gasoline-powered engines and diesel engines, data from speciation profiles were
applied to county level VOC and PM estimates generated by the April, 1999 draft version
of the EPA NONROAD model. Again, impacts of gasoline fuel control programs on
emissions were accounted for at the county level. Emissions of aircraft, locomotives, and
commercial marine vessels were estimated by applying data from speciation profiles to
national VOC and PM estimates, then using activity data to allocate nationwide
emissions to the county level.
3.2.2 Environmental Fate and Transport Characterization
After the pollutants of interest and their sources and emission rates are defined, the
exposure assessment process continues with estimation of pollutant fate and transport in
the atmosphere. This step describes how the pollutant is transported, dispersed, and
transformed over the area of interest. Initially, the fate of the emitted pollutants is largely
determined by the source release characteristics. After pollutants are released to the
atmosphere, their transport, dispersion, and transformation are governed by
meteorological principles, terrain characteristics, wet and dry deposition rates, and certain
chemical properties of the air toxics (e.g., aqueous solubility, vapor pressure, molecular
diffusivity, melting point, and adsorption characteristics). For a limited subset of air
toxics, it may be important to consider deposition from air to soil, vegetation, or water
bodies. For others, such deposition is not important.
Various mathematical models (e.g., Gaussian puff models, Gaussian plume models) [75],
each with specific data needs, have been developed or are under development to describe
the transport and fate of pollutants released to the atmosphere. The model chosen must
be appropriate for the intended application, which may range from estimates of short-
term peak concentrations immediately adjacent to a facility, to long-term concentrations
30
-------
over a citywide area or deposition over thousands of miles. The reactivity and
persistence of each air pollutant will also influence its fate, and these factors can be
important in estimating exposure for certain pollutants. Additionally, secondary
transformation products of some air toxics may need to be identified for consideration in
risk assessment. Any available air toxics monitoring data can be used either to check the
validity of modeled concentration estimates or as a primary or supplemental source of
information for the exposure assessment itself.
For a limited subset of air toxics, greater exposures occur through non-inhalation
exposures than through inhalation exposures. These air toxics typically are persistent in
the environment and have a strong tendency to bioaccumulate. Exposure assessments
can consider exposures that occur through routes other than inhalation by using
multipathway models. The simplest multipathway exposure assessments require
chemical-specific data (e.g., octanol-water partition coefficient) to model the partitioning
of the chemical in the environment, and uptake rates (e.g., consumption rate of drinking
water) to predict intakes. Combining this information yields general predictions of non-
inhalation exposure. EPA's current national-scale exposure models do not have the
capability to quantify non-inhalation exposures, so they were not included in this initial
assessment, although this may become possible in the future.
3.2.2. 1 Overview of the ASPEN Dispersion Model
To develop national-scale estimates of annual average ambient concentrations of air
toxics, EPA used the Assessment System for Population Exposure Nationwide (ASPEN)
model that was developed and used in EPA's Cumulative Exposure Project (CEP)[7<5].
In general, ASPEN uses a Gaussian model formulation and climatological data to
estimate annual average pollutant concentrations at the centroid of each census tract
within the modeling domain. Specifically, for each source, the model calculates sets of
eight 3-hour ground-level concentrations as a function of radial distance and direction
from the source at a set of receptors laid out in a radial grid pattern. These concentrations
represent the steady-state concentrations that would occur with constant emissions and
meteorological parameters. For each grid receptor, concentrations are calculated for each
of a standard set of stability class/wind speed/wind direction combinations. These
concentrations are averaged together using the annual frequency of occurrence of each
combination (i.e., the climatology) as weightings.
These meteorological frequency distributions are typically prepared for the entire
simulation period, usually one or more years. For ASPEN, however, meteorological data
are stratified by time of day into eight 3-hour time blocks, to preserve any characteristic
diurnal patterns that might be important in subsequent estimation of population exposure.
In addition to the climatology, other inputs to ASPEN that are specified by time block
include emission rate, mixing height, and reactive decay rates. The resulting output of
ASPEN is a grid of annual average concentration estimates for each source/pollutant
combination by time block.
The ASPEN model takes into account important determinants of pollutant concentrations,
such as: rate of release, location of release, the height from which the pollutants are
31
-------
released, wind speeds and directions from the meteorological stations nearest to the
release, breakdown of the pollutants in the atmosphere after being released (i.e., reactive
decay) settling of pollutants out of the atmosphere (i.e., deposition), transformation of
one pollutant into another (i.e., secondary formation).
For all pollutants outdoor concentrations include a "background" component.
Background is an essential part of the total air quality concentrations. Background
includes concentrations due to natural sources, sources not in the emissions inventory,
and long-range transport. In this study, except for diesel PM, background concentrations
are represented by concentration values measured at "clean air locations" where
available. Non-zero background values for 13 pollutants, identified from the technical
literature and reported in the CEP study [16] were used to sum with the ASPEN-
estimated concentrations in each census tract. Except for diesel PM, the background
value is assumed to be constant for all census tracts due to insufficient data for
assessment of geographic variations. Where background concentration values were not
identified in the technical literature and reported in the CEP study for other air toxics,
their background concentrations are assumed to be zero. This result may be an
underestimate of outdoor concentrations in some cases.
Annual average concentration estimates for grid receptors surrounding each emission
source are spatially interpolated to the census tract centroids within the 50 kilometers
impact zone, and contributions from all modeled sources are summed to give estimates of
cumulative ambient concentration increments in each census tract [77]. By accounting
for all identified source categories (including background concentrations), the sum of the
concentration increments should yield an estimate of the overall concentration of each air
toxic within each census tract. These estimates are designed to represent population-
weighted concentration averages (each census tract generally represents between 2,500
and 8,000 people). More detailed information on ASPEN is provided in the ASPEN
User's Guide (Appendix E).
3.2.2.2 Application of ASPEN for the Initial National-Scale Assessment
For this assessment, the ASPEN model was run in the same manner (e.g., reactivity class
assignments, wet and dry deposition for particulates, secondary formation) as was done in
the CEP, with some exceptions:
1. The input inventories were different in data year, approach and some
emissions processing techniques. For example, location and stack parameter
defaulting techniques were different. Changes were also made to spatial
allocation and temporal allocation factors for some emission sources. See section
3.2.1 for a description of inventories and how they were prepared for the ASPEN
model.
2. In 1996, there was 75 percent completeness of meteorological data at 357
National Weather Service surface stations versus 214 in 1990. Data from the 357
stations were used for this assessment. Due to the use of an increased number of
stations, the average distance between the emission source and the meteorological
station improved (decreased) in the 1996 ASPEN run.
32
-------
3. The modeling domain was extended to Puerto Rico and the Virgin Islands.
4. For diesel PM only, instead of using monitored air quality data to establish
background concentrations, a modeling-based approach was developed to provide
a rough approximation of concentrations due to transport from sources located
between 50 km and 300 km from the receptor. See Appendix F for a more
complete discussion of this approach.
To evaluate the quality of ASPEN outputs as a check on input and model execution
accuracy, predicted concentrations of toxic pollutants were examined graphically to
establish a framework from which "unusual" data values might be identified. The most
effective inspection method proved to be the preparation of matrix scatter plots in which
predicted concentrations of each pollutant were plotted against the concentrations of all
other pollutants at that location. Generally, these plots showed high correlation among
pollutants that had common sources (e.g. benzene, toluene, ethyl benzene) and low
correlation between pollutants that did not generally share common sources (e.g. benzene
and vinyl chloride). Plots were developed separately for each state using the total
predicted concentration from ASPEN, but also for each major source component
including mobile onroad, mobile nonroad, point and non-point source contributors.
Using these plots along with sorted rankings of pollutant concentrations, predicted values
were identified that either appeared very large and/or inconsistent with other pollutant
levels. These suspect values were then subjected to verification and/or correction
through discussion with emission inventory and modeling experts. From this process, a
significant number of errors were corrected before the final ASPEN modeling was
performed.
3.2.3 Estimating Population Exposure
In the third step of the initial national-scale assessment, ambient air toxics concentrations
derived from ASPEN modeling were used to estimate human exposures to air toxics.
Inhalation was the only exposure route considered for this initial assessment. For
characterization of personal exposure, a model was needed that would allow for
consideration of inhalation exposures to various population groups, who may have
different levels of exposure as a result of differences in proximity to sources of exposure
(due to location of residence, occupational setting, etc.).
An inhalation exposure model, the Hazardous Air Pollutant Exposure Model, Version 4
(HAPEM4) was selected for use to estimate personal exposure and account for
differences in exposures among the population.
3.2.3.1 Overview of HAPEM4
The HAPEM4 is an exposure model that is capable of assessing average long-term
inhalation exposures of the general population, or a specific sub-population, over spatial
scales ranging from urban to national. HAPEM4 utilizes a relatively transparent set of
exposure assumptions and approximations (Appendix B). HAPEM4 uses the general
exposure modeling approach of tracking representatives of specified demographic groups
as they move among indoor and outdoor microenvironments and between geographic
33
-------
locations performing various activities. Figure 3-2 shows example demographic groups,
microenvironments, and activities. The estimated pollutant concentrations in each
microenvironment visited are combined into a time-weighted average exposure
concentration, which is assigned to members of the demographic group.
HAPEM4 uses four primary sources of information: ambient air concentration data,
population data, population activity data, and microenvironmental data. HAPEM4 also
contains several special features to improve its exposure prediction capabilities. These
data sources and features are described in more detail below.
Ambient Air Concentration Data
The HAPEM4 requires annual-averaged, diurnally-distributed air quality data. In
addition, HAPEM4 can also evaluate the contributions of sub-sets of the air quality data
(e.g., air concentration values for specific source sectors such as point source, area
source, mobile source). While the air concentration data for HAPEM4 must be in a
specific format (e.g., annual average and diurnally distributed), the source of the data
could be either from an air dispersion model or an ambient monitor. The most common
form of ambient air concentration data for HAPEM4 is output data in the ISCLT format.
Population Data
The U.S. Census Bureau is the primary source of most population demographic data.
The U.S. Census Bureau collects information on where people live, their demographic
makeup (e.g., age, gender, ethnic group), and employment. The HAPEM4 model uses
1990 U.S. Census data reported at the spatial resolution of census tracts, which are small,
relatively permanent statistical subdivisions of a county. Census tracts usually contain
between 2,500 and 8,000 residents.
Population Activity Data
HAPEM4 uses two types of population activity data: activity pattern data and
commuting pattern data. An activity pattern is a series of discrete events of varying time
intervals describing an individual's lifestyle and routine. An activity pattern typically
includes the amount of time spent in each of a set of microenvironments (e.g., at home, at
work, in an automobile, etc.), and a description of what the individual was doing in each
microenvironment (e.g., sleeping, eating, exercising, etc.). EPA's Consolidated Human
Activity Database (CHAD) [75], containing more than 22,000 person-days of activity
pattern records from 12 studies, is incorporated into HAPEM4.
Because activity data are not available at a high enough resolution to estimate the
exposure of each individual in the population, HAPEM4 groups activity pattern data
together for people with similar demographic characteristics that are expected to
influence exposure to air pollutants (e.g., age, gender, race), and HAPEM4 estimates
exposures for these demographic groups.
Annual average activity pattern sequences are built by randomly selecting (with
replacement) 365 daily diary entries. Day of the week, as well as season type, are
considered in this selection process. It is important to note that construction of an annual
34
-------
average activity pattern in this manner results in the loss of day-to-day correlations in
activity patterns.
The commuting data contained in HAPEM4 have been derived from a special 1990 US
Census database that specifies for each US Census tract the number of residents that work
in each US Census tract, i.e., the population associated with each home tract/work tract
pair.
Microenvironment Data
In order to calculate the exposure concentration for each demographic group, an estimate
is required of the concentration in each microenvironment (ME) specified by the activity
pattern. In HAPEM4, these ME concentration estimates are derived from the ambient
concentration estimate for the census tract (obtained from ASPEN) and a set of 3 ME
factors: PEN, PROX, and ADD. The penetration factor, PEN, is an estimate of the ratio
of ME concentration to the concurrent outdoor concentration in the immediate vicinity of
the ME. These pollutant-specific estimates are derived from reported measurement
studies. The proximity factor, PROX, is an estimate of the ratio of the outdoor
concentration in the immediate vicinity of the ME to the outdoor concentration
represented by the air concentration data. ADD is an additive factor that accounts for
emission sources within or near a particular microenvironment, i.e., indoor emission
sources.
In HAPEM4, these ME concentration estimates are derived from the ambient
concentration estimate for the census tract (obtained from ASPEN) and a set of 3 ME
factors as follows:
C(i,k,t) = [ASPEN]i x PEN x PROX + [ADD]
Where:
C(j kt) = concentration predicted within exposure district i and microenvironment k
in time step t
[ASPEN]j = ambient concentration estimated from ASPEN in district i
PEN = penetration factor
PROX = proximity factor
ADD = additive factor accounting for sources within the microenvironment
Features of HAPEM4
It is important to note that the HAPEM4 model has been designed to predict inhalation
exposures for population groups, not individuals within these groups. However, the
HAPEM4 model contains a stochastic feature to allow the exposure modeler to try and
capture some of the variability in activity patterns within these groups. The stochastic
feature predicts the annual exposure concentrations for a randomly selected set of 30
estimates for each demographic group.
The HAPEM4 contains a commuting feature that allows the movement of people
between home and work locations. In general, the user can specify whether a specific
demographic group is a "commuting" group or a "non-commuting" group. For work
35
-------
related activities, the "commuting" group is then placed in a different census tract
location. A default file, developed from 1990 US Census data, that depicts tract-to-tract
commuting patterns is included with the HAPEM4 model. The model can be run either
with or without the commuting feature.
Details on both of these features can be found in the HAPEM4 User's Guide (Appendix
B).
3.2.3.2 Application of HAPEM4 for the Initial National Scale Assessment
For this assessment, the model was applied on the national scale. HAPEM4 estimated
personal inhalation exposures from ambient air toxics concentrations by defining the
exposed population in terms of geographic and demographic distributions and then
accounting for the various microenvironments to which people may be exposed. These
microenvironments were addressed by considering the time people may spend in each of
these microenvironments and the air toxics concentration in those microenvironments
relative to the ambient air toxics concentrations predicted by ASPEN.
Ambient Air Concentration Data
For the national-scale assessment, annual average ambient concentrations for each US
Census tract were estimated with the ASPEN model. In order to preserve any
characteristic diurnal patterns in ambient concentrations that might be important in the
estimation of population exposure, ASPEN annual average concentration estimates are
stratified by time of day, with an annual average for each of the (8) 3-hour time blocks
(e.g., midnight to 3am, 3am to 6am). ASPEN air quality files were also provided by each
of the 4 major source sectors (i.e., major, area, mobile onroad, mobile nonroad). Thus,
the results of HAPEM4 can be summarized for each of the 4 groups or a combination of
them.
Population Data
For this application, HAPEM4 divided the population into 40 demographic groups, based
on combinations of age (5 categories), race (4 categories), and gender. Figure 2-3 depicts
the 40 demographic groups identified and used for the initial assessment.
Population Activity Data
For each demographic group, 365 activity patterns were selected randomly (with
replacement) and then combined to develop the average fraction of time in each of the 37
microenvironments for each of the (8) 3-hour time blocks. One hundred such annual
activity patterns were constructed for each demographic group. Then, for each US Census
tract, 30 of the 100 annual patterns were randomly selected (with replacement) to
represent typical annual time allocations for group members in that tract. The result was
a set of 30 annual exposure concentrations estimates for each demographic group in each
census tract. Figure 3-3 shows an example of a daily inhalation exposure scenario and
calculations performed to derive a daily exposure value. Because each component of the
national-scale assessment has been designed to predict more "typical" or population-
based estimates rather than extremes or individuals, the HAPEM4 inhalation
concentrations presented for this study, as well as that employed for the subsequent risk
36
-------
characterization, were derived and aggregated using the median exposure concentration
from this group of 30 concentrations.
Microenvironment Data
For most MEs, HAPEM4 uses a PROX value of 1.0. However, when assessing exposure
to motor vehicle emissions, for MEs near roadways (e.g., in-vehicle), the pollutant
concentration contribution in the immediate vicinity of the ME is generally higher than
the average pollutant concentration contribution over the census tract. Thus, PROX
values of greater than 1.0 are used. For this application, which addresses only exposure
resulting from outdoor emission sources, ADD was uniformly set equal to zero. A
complete listing of the ME factors employed in the national scale assessment for each
pollutant are presented in the report "Development of Microenvironmental Factors for the
HAPEM4 in Support of the National Air Toxics Assessment (NATA) - External Review
Draft Report" [19]. A summary of the peer review comments on this report and how they
were addressed in this assessment is included in Appendix B.
3.3 Dose-Response Assessment
3.3.1 Introduction
Within EPA's paradigm for risk assessment (Figure 2-1, outer circle), the dose-response
assessment phase of a risk assessment is based on two sequential analyses. The first
analysis is the hazard identification, which identifies contaminants that may pose health
hazards at environmentally relevant concentrations and qualitatively describes the effects
that may occur in humans. The second analysis is the human health dose-response
assessment, which characterizes the relationship between the concentration, exposure, or
dose of a pollutant and the resultant health effects.
The types of effects relevant to each chemical (e.g., cancer, or effects other than cancer)
are determined as part of the hazard identification. Factors such as the experimental
route of exposure, the type and quality of the effects, the biological plausibility of
findings, and the consistency of findings across studies, all contribute to the nature of the
hazard identification statement.
The nature of quantitative dose-response assessment typically varies among pollutants.
Sufficient data often exist for criteria air pollutants, such as ozone or carbon monoxide,
so that relatively complete dose-response relationships can be characterized. In such
cases, there is no need for extrapolation to lower doses because adequate human health
effects data are available at environmentally relevant levels. However, this has not been
the case for most air toxics. Epidemiologic and toxicologic data on air toxics have
typically resulted from exposure levels that were high relative to environmental levels.
Generally, dose-response assessment methods for air toxics consist of two parts. First is
the evaluation of data in the observable range, and second is the extrapolation from the
observable range to low doses/risks. Recent terminology refers to the result of analysis in
the observable range as the "point of departure" from which extrapolation begins. The
approaches used for evaluation in the observable range are similar for all types of effects,
37
-------
but EPA's current extrapolation methods differ considerably for cancer and noncancer
effects.
3.3.2 Cancer
3.3.2. 1 Hazard Identification
The EPA's 1986 Guidelines for Carcinogen Risk Assessment [20] provide guidance on
hazard identification for carcinogens in this assessment. This approach recognizes three
broad categories of data: (1) human data (primarily epidemiological); (2) results of long-
term experimental animal bioassays; and (3) supporting data, including a variety of short-
term tests for genotoxicity and other relevant properties, pharmacokinetic and metabolic
studies, physio-chemical properties, and structure-activity relationships. In hazard
identification of carcinogens under the 1986 guidelines, human data, animal data, and
supporting evidence are combined to characterize the weight-of-evidence (WOE)
regarding the agent's potential as a human carcinogen into one of several hierarchic
categories:
• Group A - Carcinogenic to Humans: Agents with adequate human data to
demonstrate the causal association of the agent with human cancer (typically
epidemiologic data).
• Group B - Probably Carcinogenic to Humans: Agents with sufficient evidence
(i.e., indicative of a causal relationship) from animal bioassay data, but either
limited (i.e., indicative of a possible causal relationship, but not exclusive of
alternative explanations) human evidence (Group Bl), or with little or no human
data (Group B2).
• Group C - Possibly Carcinogenic to Humans: Agents with limited animal
evidence and little or no human data.
• Group D - Not Classifiable as to Human Carcinogenicity: Agents without
adequate data either to suggest or refute the suggestion of human carcinogenicity.
• Group E - Evidence of Non-carcinogenicity for Humans: Agents that show no
evidence for carcinogenicity in at least two adequate animal tests in different
species or in both adequate epidemiologic and animal studies [20].
It is important to note that the WOE categories under the 1986 cancer guidelines express
only a relative level of certainty that these agents may cause cancer in humans. The
categories specifically do not connote relative level of hazard, or degree of conservatism
applied in developing a dose-response assessment. For example, a substance in group C
(possible human carcinogen) may very well impart a greater cancer risk to more people
than another substance in group A (known human carcinogen). EPA has classified
substances as "possible" carcinogens only because the amount and quality of evidence
was insufficient to place them in a higher group, not because EPA believes they
necessarily present less risk.
38
-------
EPA's 1986 carcinogen risk assessment guidelines were the product of nearly two
decades of experience and scientific consensus building. EPA has since gained
considerable experience in applying cancer risk assessment approaches. Likewise, the
science of risk assessment and toxicological testing has continued to evolve while EPA
has had to address situations not explicitly discussed in the 1986 guidelines, e.g.,
children's risk assessment. The revision of EPA's carcinogen risk assessment guidelines
that is currently underway will consolidate the Agency's experience, provide more
comprehensive and transparent guidance on topics not fully developed in the original
guidelines, and provide flexibility to accommodate anticipated advances in the science.
In a 1996 Federal Register notice (61 FRNo. 123, 32799- 32801) EPA announced that,
pending publication of the final revised guidelines, the principles and procedures of the
draft revised guidelines would be applied in part or in whole on a case-by-case basis for
new assessments. Further, the 1996 proposed guidelines represent the evolution of risk
assessment methods rather than a "sea change" in those methods. Application of these
approaches is felt to be reflective of EPA's accumulated experience and in keeping with
advancing knowledge on cancer assessment and, therefore, provides the Agency with
more experience to draw upon in finalizing the guidelines. Accordingly, substances in
the current assessment that have EPA hazard identification and dose-response
information developed since 1996 reflect the proposed cancer guidelines, using the most
recent 1999 guidelines draft [27].
3.3.2.2 Dose-Response Assessment for Carcinogens
EPA's 1986 cancer risk assessment guidelines [20] adopted a default assumption that
chemical carcinogens would exhibit risks at any dose. Extrapolation of cancer risk using
the linearized multistage model, which results in a linear extrapolation of risk in the low
dose region, was proposed as a reasonable upper-bound on risk (i.e., the true value of the
risk is unknown, and may be as low as zero), and this approach has been used for most
chemicals with adequate data since then. Extrapolation of cancer risk using other linear
extrapolation models, such as have been used with human data available for some known
human carcinogens (e.g., benzene, hexavalent chromium) results in estimates which,
although conceivably surrounded by less uncertainty, are still characterized by the
Agency as plausible upper bound estimates (i.e., the risk is likely to be lower but may be
greater). Although the 1986 guidelines also supported the use of non-linear low-dose
extrapolations (given adequate mechanistic data), until recently the low-dose linearity
assumption has been used without exception in estimating carcinogenic potency.
Since the publication of EPA's original cancer guidelines, considerable new knowledge
has been developed regarding the processes of chemical carcinogenesis and the
evaluation of human cancer risk. The revision of the cancer guidelines currently in
progress [21] departs substantially from the original guidelines by distinguishing between
linear and nonlinear modes of action. The cancer data in the observable range are to be
analyzed using a dose-response model similar to the models used for noncancer effects.
The method of extrapolation to lower doses from the point of departure may vary
depending on whether the assessment of the available data on the mode of action of the
chemical indicates a linear or nonlinear mode of action.
39
-------
Under the proposed guidelines, a linear extrapolation will remain appropriate when the
evidence supports a mode of action of gene mutation due to direct deoxyribonucleic acid
(DNA) reactivity or another mode of action that is thought to be linear in the low dose
region. A linear mode of action also will serve as a default when available evidence is
not sufficient to support a nonlinear extrapolation procedure, even if there is no evidence
for DNA reactivity. The linear extrapolation method has also been revised and simplified
from that employed under the 1986 guidelines. Nonlinear methods are to be used if there
is sufficient evidence to support a nonlinear mode of action.
EPA's process of estimating cancer risk is based on the unit risk estimate (URE). A URE
represents an estimate of the increased cancer risk from a lifetime (generally assumed to
be 70 years) exposure to a concentration of one unit of exposure. The URE for inhalation
exposures is typically expressed as risk per microgram of pollutant per cubic meter of air.
The URE is a plausible upper-bound estimate of the risk (i.e., the risk is likely to be
lower, but may be greater). EPA defines an upper bound as a plausible upper limit to the
true value of a quantity. Because UREs reflect unquantifiable assumptions about effects
at low doses, their upper bounds are usually not true statistical confidence limits.
Available data were insufficient to support assumptions of threshold or sublinear dose-
response for the substances in this assessment, so estimates of cancer risk were developed
by linear extrapolation of the URE (i.e., by multiplying the estimated lifetime average
daily exposure in micrograms per cubic meter by the URE). UREs used in this
assessment were developed by EPA and by the California EPA, and selected for use by a
priority system described in Appendix G.
3.3.3 Effects other than cancer
3.3.3. •/ Hazard Identification
Due to the wide variety of endpoints, hazard identification procedures for effects other
than cancer have not been described as completely in EPA guidance as procedures for the
identification of carcinogens. However, the EPA has published guidelines for assessing
several specific types of noncancer effects, including mutagenicity [22], developmental
toxicity [23], neurotoxicity [2¥], and reproductive toxicity [25].
Under these guidelines for identification of long-term (chronic) hazards other than
cancer, scientists from EPA (and from other agencies that assess dose-response
relationships) review the health effects literature and characterize its strengths and
weaknesses, using a narrative approach rather than a formal classification scheme.
Available data on different endpoints are arrayed and discussed, and the effects (and their
attendant dose/exposure levels) described. Particular attention is given to the critical
effect (defined as the first adverse effect, or its known precursor, that occurs to the most
sensitive species as the dose rate of an agent increases) in well-designed studies.
Information is presented in a narrative description that discusses factors such as the
methodological strengths and weaknesses of individual studies (as well as the overall
database), the length of time over which the studies were conducted, routes of exposure,
and possible biological modes of action. Assessors consider the severity of effects,
which may range from severe frank effects that can cause incapacitation or death to
40
-------
subtle effects that may occur at the cellular level but are early indicators of toxic effects.
Not all effects observed in laboratory studies are judged to be adverse. The distinction
between adverse and non-adverse effects is not always clear-cut, and considerable
professional judgment is applied to identify adverse effects. All of these observations are
integrated into a presentation that gives a concise profile of the toxicological properties of
the pollutant.
3.3.3.2 Dose-Response Assessment for Non-Carcinogens
The inhalation reference concentration (RfC) is the primary Agency consensus
quantitative toxicity metric for use in noncancer risk assessments for chronic inhalation
exposure. The RfC is an estimate (with uncertainty spanning perhaps an order of
magnitude) of a continuous inhalation exposure to the human population (including
sensitive subgroups) that is likely to be without an appreciable risk of deleterious effects
during a lifetime. The RfC is derived after a thorough review of the health effects data
base for an individual chemical, and identification of the most sensitive and relevant
endpoint (i.e., the critical effect) and the principal study(ies) demonstrating that endpoint.
Inhalation RfCs are derived according to the Agency's Methods for Derivation of
Inhalation Reference Concentrations and Application of Inhalation Dosimetry [26]. The
evaluation of and choice of data on which to base the RfC derivation are critical aspects
of the assessment and require scientific judgment.
Derivation of the RfC typically begins with identification of the critical effect from the
available valid human and animal study data, followed by identification of a lowest-
observed-adverse-effect level (LOAEL) or, preferably, a no-observed-adverse-effect
level (NOAEL). Some assessments model the dose-response relationship to interpolate a
benchmark dose (BMD), usually the dose at which 5 percent of the organisms are
predicted to respond. BMDs are used interchangeably with NOAELs. The LOAELs or
NOAELs from animal studies are converted to human equivalent concentrations (HECs)
using dosimetric methods [26]. The NOAEL[HEC] or LOAEL[HEC] from one or a few
well-conducted studies are the key values gleaned from evaluation of the dose-response
data. The RfC is then derived by consistent application of uncertainty factors (UFs) to
account for recognized uncertainties in the extrapolation from the experimental data and
exposure conditions to an estimate (the RfC) appropriate to the assumed human lifetime
exposure scenario [26]. RfCs (and similar dose-response values) used in this assessment
were developed by EPA, the US Agency for Toxic Substances and Disease Registry, and
the California EPA. They were selected for use by a priority system described in
Appendix G.
3.4 Risk Characterization
3.4.1 Introduction
Under EPA's risk assessment paradigm, the final product in the risk assessment process
is the risk characterization, in which the information from the previous steps is integrated
and overall conclusions about risk are synthesized and presented in a way that is
appropriate and informative for decision-makers. In general, the nature of a risk
characterization will depend on the information available, the intended use of the risk
information, and the resources (including time) available. In all cases, however, major
41
-------
issues associated with determining the nature and extent of the risk should be identified
and discussed. Further, the EPA Administrator's March 1995 Policy for Risk
Characterization [27] specifies that a risk characterization "be prepared in a manner that
is clear, transparent, reasonable, and consistent with other risk characterizations of similar
scope prepared across programs in the Agency." The 1995 Guidance for Risk
Characterization [28] lists several guiding principles for defining risk characterization in
the context of risk assessment. The three principles with respect to the information
content and uncertainty aspects of risk characterization are as follows:
1. The risk characterization integrates the information from the exposure and dose-
response assessments, using a combination of qualitative information, quantitative
information, and information regarding uncertainties. A good characterization
should include different kinds of information from all portions of the foregoing
assessment, carefully selected for reliability and relevance.
2. The risk characterization includes a discussion of uncertainty and variability. The
risk assessor must distinguish between variability (arising from true
heterogeneity) and uncertainty (resulting from a lack of knowledge).
3. Well-balanced risk characterizations present risk conclusions and information
regarding the strengths and limitations of the assessment for other risk assessors,
EPA decision-makers, and the public. "Truth in advertising" is an integral part of
the characterization, discussing all noteworthy limitations while taking care not to
become mired in analyzing factors that are not significant.
Risk assessments are intended to address or provide descriptions of risk to: (1)
individuals exposed at average levels and those in the high-end portions of the risk
distribution; (2) the exposed population as a whole; and (3) important subgroups of the
population such as highly susceptible demographic groups or life stages, if known.
Because cancer and noncancer dose-response assessment methods are currently quite
different, risk characterizations also differ and are discussed separately.
Given the goals of the initial national-scale assessment, the risk characterization's
purposes are best served by an analysis of broad geographic scale, recognizing several
significant limitations. First, the resolution is coarse. Quantitative estimates of cancer
risk and non-cancer hazard are calculated at the census-tract level, but presented only
using statistics which summarize their distributions at the county level or higher. Second,
the risk characterization limits itself to potential human exposures and health effects,
including limited information on variations in exposure and risk among specific
subpopulations. Third, the risk characterization includes only inhalation exposure, and
excludes estimates of air toxics uptake by ingestion and dermal contact. Fourth, the risk
characterization includes only inventoried major, area, and mobile sources of the study
pollutants. For these reasons, the results of this assessment represent only a portion of
the true risks associated with these air toxics. Interpretation and use of these risk results
should thus bear in mind these limitations and focus on their relative aspects rather than
their absolute magnitudes.
42
-------
3.4.2 Cancer
In this assessment, cancer risk is defined as the probability of contracting cancer
following exposure to a pollutant over a 70-year period (assumed human lifespan) at the
estimated exposure concentration. This estimate of risk focuses on the additional lifetime
risk of cancer predicted from the exposure being analyzed, beyond that due to any other
factors, and utilizes cancer potency factors which the Agency considers to be plausible
upper-bounds (i.e., the true risk potencies are likely to be lower, but may be greater). It is
noted that in this assessment, the estimated exposure concentrations are not considered to
be upper-bound. Rather, they represent central tendency estimates of exposure
concentrations for each demographic group at the geographic unit of analysis (e.g.,
census tract, county, etc.). Estimates of risk are expressed as a probability, usually
represented in scientific notation as a negative exponent of 10. For example, an
additional lifetime risk of contracting cancer of 1 chance in 10,000 (or one additional
person in 10,000) is written as IxlO"4 or le-4.
The distribution of individual exposures and risks within a given population can also be
presented, providing an estimate of the number of people exposed to various predicted
levels of risk. The Agency's risk characterization guidelines recommend that risk
assessments describe individual risk, population risk, and risk to important subgroups of
the population such as highly exposed or highly susceptible groups [25]. Quantitative
individual cancer risks are calculated by multiplying the corresponding exposure estimate
by the URE.
People are typically exposed to multiple chemicals rather than a single chemical. In rare
cases where WOE classifications and UREs are available for the chemical mixture of
concern or for a similar mixture, risk characterization can be conducted on the mixture
using the same procedures used for a single compound. However, cancer dose-response
assessments and UREs are usually available only for individual compounds within a
mixture. Consequently, in assessments of carcinogens for which there is an assumption
of a linear dose-response, the cancer risks predicted for individual chemicals are typically
added to estimate cumulative risk associated with groups of chemicals, as recommended
by EPA's guidelines for assessment of mixtures [29].
For the NATA national-scale assessment the risk estimates for cancer have been
expressed in terms of the probability of contracting cancer from a lifetime of exposure.
For substances for which UREs have been developed by linear extrapolation to low doses
(including all the carcinogenic air toxics in this assessment), probabilities were calculated
by multiplying the URE by the estimated lifetime average daily exposure.
Lifetime cancer risks were calculated and aggregated as follows, in order to focus the
assessment on those air toxics that drive the assessment:
1. Air toxics-specific cancer risks for each substance having a URE in Table 3-5
were calculated for the median exposure estimates within each census tract. Plots
were prepared showing the frequency distribution of risk for each air toxic across
all census tracts, and population sizes living in tracts where the median cancer risk
estimates exceeded fixed levels.
43
-------
2. Air toxics for which estimated cancer risk exceeded le-6 (1 in 1 million) in the
99th percentile census tract were grouped by WOE (as per the 1986 cancer
guidelines), using information shown in Table 3-5. The development of the UREs
shown in Table 3-5 for total particulate organic matter (POM) and 7-PAH
(polycyclic aromatic hydrocarbon (PAH)) is described in Appendix H. In the
spirit of the 1996 proposed classification of "known" and "likely" carcinogens,
risks of different air toxics were combined only within two groups: the category A
pollutants, and across all category B and C pollutants.
3. Air toxics for which estimated cancer risk did not exceed le-6 (1 in 1 million) in
the 99* percentile census tract were not included in aggregate risk estimates of
multiple air toxics because their relatively small contribution to the risk sum
would be within rounding error.
4. In combining risks across multiple carcinogens, this assessment did not consider
information supporting non-additive aggregation (as recommended in EPA's draft
mixtures assessment guidelines), because such information was not available.
Accordingly, the assessment used the mixtures guidelines default assumption of
additivity of risks, and combined risks in the manner described above by
summing them, using the independence formula in the mixtures guidelines.
3.4.3 Effects Other Than Cancer
Unlike linear dose-response assessments for cancer, in most cases, noncancer risks
generally are not expressed as a probability of an individual suffering an adverse
effect. Instead, "risk" for noncancer effects typically is quantified by comparing the
exposure to the reference level as a ratio. The "hazard quotient" (HQ) is the exposure
divided by the reference level (e.g., the RfC or other similar value). For a given air
toxic, exposures below the reference level (HQ<1) are not likely to be associated with
adverse health effects. With exposures increasingly greater than the reference level
(i.e., HQs increasingly greater than 1), the potential for adverse effects increases. The
HQ, however, should not be interpreted as a probability of adverse effects.
While some risk assessments may involve significant exposure to only a single
compound, exposure to a mixture of compounds that may produce similar or
dissimilar health effects more accurately reflects "real" conditions. In a few cases,
reference levels may be available for a chemical mixture of concern or for a similar
mixture. In such cases, risk characterization can be conducted on the mixture using
the same procedures used for a single compound. However, noncancer health effects
data are usually available only for individual compounds within a mixture. In
screening assessments for such cases, a conservative "hazard index" (HI) approach, in
which all the HQs for individual contaminants are summed, is sometimes used. This
approach assumes that even when individual pollutant levels are lower than the
corresponding reference levels,
44
-------
Table 3-5. Hazard identification and dose-response information for carcinogenic effects.
Weight of Evidence Unit Risk
Air Toxics Estimate Source
EPA IARC (per Ug/m')
Acetaldehyde
Acrylonitrile
Arsenic compounds
Benzene
Beryllium compounds
1,3 -Butadiene
Cadmium compounds
Carbon tetrachloride
Chloroform
Chromium compounds
Coke Oven Emissions
1 , 3 -Dichloropropene
Ethylene dibromide (1,2-dibromoethane)
Ethylene dichloride (1,2-dichloroethane)
Ethylene oxide
Formaldehyde
Hexachlorobenzene
Hydrazine, hydrazine sulfate
Lead compounds
Methylene chloride
Nickel compounds
Poly chlorinated bipheny Is (PCBs)
Polycyclic Organic Matter (total)
Polycyclic Organic Matter (7 -PAH)
Propylene dichloride (1,2-dichloropropane)
Quinoline
1, 1,2,2-Tetrachloroethane
Tetrachloroethylene (perchloroethylene)
Trichloroethylene (TCE)
Vinyl chloride
B2
Bl
A
A
Bl
B2
Bl
B2
B2
A
A
B2
B2
B2
Bl
Bl
B2
B2
B2
B2
A
B2
9
B2
B2
C
C
B2-C
B2-C
A
2B
2A
1
1
1
2A
1
2B
2B
1
.
2B
2A
2B
1
2A
2B
2B
2B
2B
2B
2A
9
9
-
-
3
2A
2A
1
2.2E-06
6.8E-05
4.3E-03
7.8E-06
2.4E-03
1E-05
1.8E-03
1.5E-05
2.3E-05
4.1E-03
6.2E-04
4.0E-06
2.2E-04
2.6E-05
8.8E-05
1.3E-05
4.6E-04
4.9E-03
1.2E-05
4.7E-07
1.2E-04
1.1E-04
5.5E-05
2.0E-04
1.9E-05
3.4E-03
5.8E-05
5.9E-06
2.0E-06
8.8E-06
IRIS3
IRIS3
IRIS4
IRIS4'5'6
IRIS3'6
EPA NCEA4'6'7
IRIS3
IRIS3
IRIS3
IRIS4'6'8
IRIS3
IRIS3'6
IRIS3
IRIS3
CALEPA
IRIS3
IRIS3
IRIS3
CALEPA
IRIS3
3,8
IRIS3
10
10
CONV ORAL3
CONV ORAL3
IRIS3
CALEPA
CALEPA
EPA
NCEA3'6'7'11
3 Upper confidence limit URE; (assessments that did not specify method were assumed to use the UCL).
4 Maximum likelihood URE.
5 Higher of 2 recommended UREs was selected.
6 Assessment consistent with 1996 proposed cancer guidelines.
7 Advanced draft of IRIS assessment, expected to be finalized shortly.
8 Value includes assumptions on speciation of emissions; details will be provided in report text.
9 WOE varies among individual compounds.
10 Note that the California EPA estimates for various polycyclic organic compounds are based on a toxic equivalency approach, where
the potency of individual compounds is estimated based on relative activity rather than individual assessments of bioassay data.
45
-------
some pollutants may work together such that their potential for harm is additive and the
combined exposure to the group of chemicals poses greater likelihood of harm. This
assumption of dose additivity is most appropriate to compounds that induce the same
effect by similar modes of action [29]. As with the HQ, the HI should not be interpreted
as a probability of adverse effects, or as strict delineation of "safe" and "unsafe" levels
[29]. Rather, the HI is a rough measure of potential for risk and needs to be interpreted
carefully.
Although the HI approach encompassing all chemicals in a mixture may be appropriate
for a screening-level study, it is important to note that application of the HI equation to
compounds that may produce different effects, or that act by different mechanisms, could
overestimate the potential for effects. Consequently, EPA generally prefers a more
refined approach of calculating a separate HI for each noncancer endpoint of concern for
which modes of action are known to be similar [29].
For the NATA national-scale assessment, the risk characterization for effects other than
cancer has been expressed in terms of the hazard quotient (HQ) for inhalation. As
discussed in section 3.3.3.2 above, many RfCs incorporate protective assumptions in the
face of uncertain data, so that an HQ greater than one does not necessarily indicate a
likelihood of adverse effects. The HQ cannot be translated to a probability that adverse
effects will occur, and it is unlikely to be proportional to risk.
Different pollutants may cause completely different adverse health effects or act via
completely different modes of action, so it is often inappropriate to aggregate HQs
associated with different substances. EPA has drafted revisions to Agency guidelines on
assessing the impact of mixtures [30], which recognize combining effects of different
substances in specific and limited ways. The national-scale assessment has aggregated
non-cancer HQs of air toxics that act by similar toxic modes of action, or (where this
information is not incorporated in the dose-response assessment) that affect the same
target organ. Aggregation in this way produced a "target-organ-specific hazard index"
(TOSHI), defined as the sum of hazard quotients for individual air toxics that affect the
same organ or organ system.
Non-cancer HQs were calculated and aggregated as follows, in order to focus the
assessment on those non-carcinogenic air toxics that drive the assessment:
1. The HQ for each air toxic having an RfC or similar value in Table 3-6 was
calculated for the median exposure within each census tract. Plots were prepared
showing the frequency distribution of HQ for each air toxic across all census
tracts, and population sizes living in tracts where the median HQ exceeded fixed
levels.
2. Air toxics for which estimated HQ exceeded 0.01 in the 99th percentile census
tract were grouped by target organ, as shown in Table 3-7. Information on target
The development of UREs for total particulate organic matter (POM) and 7-PAH (polycyclic aromatic hydrocarbon (PAH)) are
described in Appendix H.
11 URE based on whole life exposure was selected over a URE based on adult exposure only.
46
-------
organs for each pollutant was obtained from dose-response assessments and from
the scientific literature. To avoid aggregating HQs with widely divergent levels
of uncertainty, HQs for different air toxics were combined only within two
groups: those with "high certainty" RfCs (i.e., whose total uncertainty factor was
100 or less) and those with "low certainty" RfCs (whose total uncertainty factor
was greater than 100).
3. Air toxics for which HQ did not exceed 0.01 in the 99* percentile census tract
were not included in aggregate TOSHIs for multiple air toxics because their
relatively small contribution to the TOSHI sum would be within rounding error.
4. For each of the target organs shown in Table 3-7, the HQ for each air toxic was
summed to create the TOSHI (data permitting).
Table 3-6. Non-Cancer Dose-Response Information.
Urban Air Toxics
CAS # (or Equivalent)12 UFxMF13
(mg/m3)
Target Organ
for Chronic
Critical Effect1
Target Organs
for Other
Chronic
Effects
Source
Acetaldehyde
Acrolein
Acrylonitrile
Arsenic compounds
Benzene
Beryllium compounds
1,3 -Butadiene
Cadmium compounds
Carbon tetrachloride
Chloroform
Chromium compounds
1 ,3-Dichloropropene
75070
107028
107131
71432
106990
56235
67663
542756
9.0E-03
2.0E-05
2.0E-03
3.0E-05
6.0E-02
2.0E-05
8.0E-03
2.0E-05
4.0E-02
9.8E-02
l.OE-04
2.0E-02
1000
1000
100/10
1000
10
10
300
30
300
100
90
30
Nasal epithelium
Nasal epithelium
Nasal epithelium,
brain
Skeleton (fetal
malformation)
Blood, bone
marrow
Lung
Reproductive
system
Kidney
Liver
Liver, kidney
Respiratory tract
(necrosis)
Nasal epithelium
Growth rate,
blood, and kidney
Mucous
membranes
(irritation)
Central nervous
system
(depression)
Skin and mucous
membranes
(irritation)
Central nervous
system
(depression)
Immune system
Cardiovascular
system, blood
Lung
Kidney
Central nervous
system
(depression)
Liver, kidney, GI
tract, immune
system
Urinary bladder
IRIS
IRIS
IRIS
CAL EPA
CAL EPA
IRIS
CAL EPA
CAL EPA
CAL EPA
ATSDR
IRIS
IRIS
Includes EPA reference concentrations (RfCs), California EPA reference exposure levels (RELs),
ATSDR minimum risk levels (MRLs), and HEAST inhalation reference doses (RfDs) converted to
concentrations in air.
13 Modifying factors of 1 are not shown.
14 Critical effect listed is the adverse effect upon which the RfC or equivalent health-based value is based.
47
-------
Urban Air Toxics
CAS # (or Equivalent)12 UFxMF13
(mg/m3)
Target Organ
for Chronic
Critical Effect1
Target Organs
for Other
Chronic
Effects
Source
Ethylene dibromide (1,2-dibromoethane)
Ethylene dichloride (1,2-dichloroethane)
Ethylene oxide
Formaldehyde
Hexachlorobenzene
Hydrazine, hydrazine sulfate
Lead compounds15
Manganese compounds
Mercury compounds16
Methylene chloride
Nickel compounds
Propylene dichloride (1,2-
dichloropropane)
Tetrachloroethylene(perchloroethylene)
Trichloroethylene (TCE)
Vinyl chloride
106934
107062
75218
50000
118741
302012
PB CMPDS
MN CMPDS
HG CMPDS
75092
NI CMPDS
78875
127184
79016
75014
8.0E-04
2.4E+00
3.0E-02
9.8E-03
3.0E-03
2.0E-04
1.5E-03
5.0E-05
3.0E-04
l.OE+00
2.0E-04
4.0E-03
2.7E-01
6.0E-01
l.OE-01
100
90
100
30
100
300
1
1000
30
30
30
300
100
100
300
Reproductive
system
Kidney
Blood
Respiratory
epithelium
Liver
(developmental)
Liver, thyroid
Central nervous
system
(neurobehavioral
effects)
Central nervous
system
(neurobehavioral
effects)
Central nervous
system
Liver
Respiratory
system, immune
system
Nasal epithelium
Central nervous
system
(depression)
Central nervous
system
(depression)
Liver
Liver, kidney,
testes
Liver
Eyes, mucous
membranes,
central nervous
system
Immune system
(sensitization)
Immune system,
kidney, blood
Respiratory
system, spleen
Blood,
cardiovascular
system, kidney
Respiratory
system
Kidney,
cardiovascular
system
Blood
Heart, liver,
kidney
Liver, kidney
Kidney, central
nervous system
(depression)
CAL EPA
ATSDR
CAL EPA
ATSDR
CAL EPA
CAL EPA
NAAQS
IRIS
IRIS
ATSDR
ATSDR
IRIS
ATSDR
IRIS
CAL EPA
EPA has not developed an RfC for lead. The National-Scale Assessment uses the National Ambient Air
Quality Standard for lead, which was developed using the EPA Integrated Exposure, Uptake, Biokinetic
Model, and did not use the UF/MF method. Because sensitive human subpopulations were modeled, the
effective UF is 1.
16 Hazard calculations for mercury compounds were based on the RfC for elemental mercury.
48
-------
Table 3-7. Grouping of compounds by target organ and uncertainty factor for aggregation
of effects other than cancer.
Target Organs/Systems for Chronic TT_ ,„ „ ,.,,,-,.
fe fe Ffi UF x MF Contaminant Groupings
Respiratory system (including nasal
epithelium, mucous membranes, and
lung)
Blood (including bone marrow and
spleen)
Central nervous system (including
neurobehavioral effects and CNS
depression)
Liver and kidney17
Cardiovascular system
Immune system (including sensitization)
1-100
101- 1000
1-100
101- 1000
1-100
101- 1000
1-100
101- 1000
1-100
101- 1000
1-100
101- 1000
beryllium, cadmium, chromium VI, 1,3-dichloropropene,
formaldehyde, nickel
Acetaldehyde, acrolein, acrylonitrile, arsenic, propylene dichloride,
sthylene oxide, hydrazine, manganese
benzene, ethylene oxide, hexachlorobenzene, lead
Arsine, 1,3 -butadiene, propylene dichloride, hydrazine
Benzene, chloroform, ethylene oxide, lead, manganese, mercury,
etrachloroethylene, trichloroethylene
Acrylonitrile, vinyl chloride
Cadmium, chloroform, chromium VI, ethylene dibromide, ethylene
dichloride, hexachlorobenzene, lead, methylene chloride,
etrachloroethylene, trichloroethylene,
Acetaldehyde, carbon tetrachloride, hydrazine, vinyl chloride
xad, methylene chloride, tetrachloroethylene
1,3 -Butadiene
beryllium, chromium VI, formaldehyde, hexachlorobenzene,
nickel
-
3.4.4 Discussion of Uncertainties in the Dose-Response Assessment
3.4.4.1 Uncertainties in the Unit Risk Estimate
The process of Unit Risk Estimate (URE) development includes the following important
sources of uncertainty:
1. Many of the air toxics included in this assessment were classified as probable
carcinogens, which means that data were not sufficient to prove these substances
definitely cause cancer in humans. It is possible that some are not human
carcinogens at environmentally relevant doses, and that true risk associated with
these air toxics is zero.
2. All UREs used in this assessment were based on linear extrapolation from high to
low doses. To the extent that true dose-response relationships for some air toxics
are less than linear, this assumption may result in significant overestimates of risk.
3. UREs for most of these substances were developed from animal data using
conservative methods to extrapolate between species. Actual human responses
may differ from the predicted ones.
Liver and kidney effects were combined because most HAPs that affect either, affect both.
49
-------
4. Most UREs used in this assessment (typically those from assessments based on
animal data) were based on the statistical upper confidence limit (UCL) of the
fitted dose-response curve, but a few (typically those from assessments based on
human data) were based on the statistical best fit ("maximum likelihood
estimate," or MLE). The reader should be aware that MLE estimates for some
known carcinogens are somewhat less conservative than UCL estimates.
Nevertheless, because of the combination of assumptions used in the face of all four
sources of uncertainly described above, EPA considers all its UREs to be upper-bound
estimates. True risk would probably be less, but could be greater.
For the NATA national-scale assessment, hazard identification and dose-response
assessments for carcinogenic effects (Table 3-5) were obtained from peer-reviewed
sources and prioritized according to (1) applicability, (2) conceptual consistency with
EPA risk assessment guidance, and (3) level of review received. A discussion of sources
and details of the prioritization process are presented in Appendix G.
3.4.4.2 Uncertainties in Reference Concentrations
In the development of reference concentrations (RfCs), uncertainty factors (UFs) are
applied as appropriate for the following extrapolations or areas of uncertainty:
• Laboratory animal data to humans;
• Average healthy humans to sensitive humans;
• Subchronic to chronic exposure duration;
• LOAEL to NOAEL; and
• Incomplete database.
In addition to UFs (which may be 10, 3 or 1), modifying factors (MFs) may also be
applied. A modifying factor is a factor used in the derivation of a reference dose or
reference concentration. The magnitude of the MF reflects the scientific uncertainties of
the study and database not explicitly treated with standard uncertainty factors (e.g., the
completeness of the overall database). A MF is greater than zero and less than or equal to
10, and the default value for the MF is 1.
The composite UF depends on the number of extrapolations required. RfCs have been
derived for various substances using composite UFs that range from 10 to 3,000, with
most RfCs using factors of 100 to 1,000. The use of order-of-magnitude uncertainty
factors for RfCs and the definition of the RfC as having "uncertainty spanning perhaps an
order of magnitude" are indications of the general lack of precision in the estimates.
It is important to note that the composite UF expresses only our relative certainty about
the dose at which these agents may not cause adverse non-cancer effects in humans and
not the absolute hazard when the RfC is exceeded. This is because the magnitude of the
UF provides no information on the relative level of hazard or degree of conservatism
applied in developing a dose-response assessment. For example, one substance's RfC
that is based on a composite UF of 3000 may be no further beneath the true threshold for
effects than the RfC for a different substance that is based on a UF of 10. The higher UF
50
-------
was assigned to the first substance only because the amount and quality of evidence was
insufficient to support a lower UF. Thus, RfCs developed using high UFs should not be
assumed to be more conservative than those using low UFs.
It should also be noted that exposures above an RfC do not necessarily imply
unacceptable risk or that adverse health effects are expected. Because of the inherent
conservatism of the RfC methodology, the significance of exceedances must be evaluated
on a case-by-case basis, considering such factors as the confidence level of the
assessment, the size of UFs used, the slope of the dose-response curve, the magnitude of
the exceedance, and the number or types of people exposed at various levels above the
RfC.
For the NATA national-scale assessment, hazard identification and dose-response
information for non-cancer health effects (3- 5) were obtained from peer-reviewed
sources and prioritized according to (1) applicability, (2) conceptual consistency with
EPA risk assessment guidance, and (3) level of review received. A discussion of sources
and details of the prioritization process are presented in Appendix G.
51
-------
4 Results and Discussion of Exposure Assessment
4.1 Introduction
It is important to note that the ambient air quality predictions, and subsequent exposure
and risk assessment results for this initial national scale assessment are being derived at a
broad geographical scale. Because of limitations associated with the emission inventory,
dispersion modeling and exposure modeling approaches, local scale "hot spots" (i.e.,
fence line impacts, complex terrain issues, persistent flow situations, fumigation events,
etc.) are not predicted with this assessment. Further, the assessment includes only a
single exposure media (i.e., air inhalation). Thus, extreme care should be taken in
interpreting the results of this assessment. The results are best used in combination with
other media assessments, as well as local scale assessments, to better evaluate the full
potential health implications of a particular pollutant. Finally, it is equally important to
note that this assessment is still undergoing scientific peer review and the results should
be considered draft until that process is complete.
The discussion of results follows the sequence of the four major components of the
assessment: emission inventory development, fate and transport characterization,
population exposure estimation, and risk characterization. The first three components,
the exposure assessment components, are discussed in this section, and the last
component, risk characterization, is discussed in section 5. The discussion of each
component includes a presentation of important results and a discussion of the major
sources of uncertainty associated with that component. The fate and transport section
includes a comparison of model performance which utilizes a comparison between
modeled and monitored ambient concentrations. In each section, the results presented
and discussed represent only a portion of the full results for that component; remaining
results for emission inventory and dispersion modeling portions of the assessment are
available on the NATA website (http://www.epa.gov/ttn/uatw/nata), and results for the
exposure modeling portion of the assessment are provided in Appendix K.
This assessment is the first national-scale assessment, and the results should be
considered to be preliminary. The national-scale assessments may be repeated every 3
years, and these future national-scale assessments will expand in scope and will make use
of improved data and assessment tools.
4.2 Source Characterization: Emission Inventories
This section provides an overview of the emission estimates in the NTI, which provide
the majority of the emissions data inputs to the national-scale assessment. The NATA
website contains a complete set of tables, maps and graphics that portray the emission
inputs to the national-scale assessment (http://www.epa.gov/ttn/uatw/nata/natsal.html).
EPA prepared emission inventory ASPEN-input files for 1) direct emissions of air toxics,
2) direct emissions of diesel particulate matter (PM), and 3) pollutants that can transform
in the atmosphere to produce air toxics. These inventories are described in section
3.2.1.2. The majority of the emissions data used in the national-scale assessment were
52
-------
from EPA's 1996 NTI. In addition to the NTI, emissions were used from a diesel PM
inventory from mobile sources, and VOC air toxics precursor emissions were derived
from the NET criteria database. Table 4-1 summarizes the inventories used as input to
EMS-HAP to prepare ASPEN inputs.
Table 4-1. Summary of Inventories Used as Input to EMS-HAP
Directly emitted air toxics from the 1 996 NTI
Air toxics precursors from the 1996 NTI and the 1996
NET speciated for particular VOC's
Diesel PM from a mobile source diesel PM inventory
(48 states) as discussed in 3.2.1.2 and the 1996 NET
(Puerto Rico and the Virgin Islands) for PM-10
Stationary
Source
Inventory
X
X
Mobile
Source
Inventory
X
X
X
The NET, diesel PM, and early versions of the 1996 NTI inventories all contained
emissions estimates for the 50 states. The geographic domain of the national-scale
assessment included the 48 contiguous states, District of Columbia, Puerto Rico, and
Virgin Islands. In all of the inventories, emissions for the territories were derived in part
or in total via extrapolation of emissions estimates from surrogate U.S. locations. More
information regarding how emissions from all inventories were prepared for ASPEN
dispersion model input can be found in Appendix C.
4.2.1 Summary and Discussion of 1996 Emission Inventory Results
There are a number of ways that the 1996 NTI data can be summarized. The NATA
website (http://www.epa.gov/ttn/uatw/nata/natsal .html) provides various summaries of
the emissions data used in the national scale assessment. Available on the site are county
emissions totals by source sector and pollutant, emissions density maps, and pie and bar
charts. Figure 4-1 shows that the national total of the 33 urban air toxics (this includes
dioxins and excludes diesel PM) emitted from the four source sectors is 1.1 million tons
(for 1996). This pie chart illustrates that approximately half of the air toxics in this
assessment come from mobile sources and the other half from stationary sources. The
majority of these stationary source emissions are from the area and other source sector.
This would be expected since these pollutants were selected, in part, based on their
presence in urban areas and from area sources.
Another way the data can be examined is by their distribution between urban and rural
counties. Counties are defined as urban or rural based on population data provided by the
Bureau of the Census. For purposes of developing EPA's Integrated Urban Air Toxics
Strategy, a county is considered "urban" if either: 1) it includes a metropolitan statistical
area with a population greater than 250,000; or, 2) the U.S. Census Bureau designates
more than fifty percent of the population as "urban." The Integrated Urban Air Toxics
Strategy (online at www.epa.gov/ttn/uatw/urban/urbanpg.html) is an important part of
EPA's national air toxics program. Please note that the definition of "urban" does not
necessarily apply for regulatory or implementation purposes.
Figure 4-2 shows a summary of NTI emissions of 33 urban air toxics by urban and rural
53
-------
designations. As in the previous pie chart (Figure 4-1), the "area and other" category
dominates the total emissions, as would be expected for this pollutant set.
Figures 4-3 to 4-6 are maps that show emissions densities by groups of air toxics (metals,
semi-volatile organic compounds (SVOCs), etc.) across the U.S., based on EMS-HAP
processed emissions. In some cases, sharp contrasts are obvious along state or county
lines. This is likely primarily an artifact of variations in state and local agency-reported
data. The maps illustrate that the greatest emission densities tend to be in metropolitan
areas and in the eastern U.S. This is to be expected since these pollutants were identified
based on their presence in urbanized locations. This result is similar across the pollutant
groups shown. A complete set of emission density maps for the all the pollutants is
available at http://www.epa.gov/ttn/uatw/nata/natsal.htmltfemission.
For the purposes of the national-scale assessment, several assumptions were made to
accommodate using the NTI data in computer models. For this reason, the emissions
input to the ASPEN model, which represents the emissions output from EMS-HAP, is not
identical to the NTI data from which it originated. The county-level emission summaries
on the NATA website reflect the EMS-HAP processed inventory and therefore, will not
match exactly to similar county-level summaries directly from the NTI. National level
summaries of the EMS-HAP emissions and NTI emissions are shown in Table 4-2, and
the principle ways in which the two differ are described below:
Differences due to pollutant groupings. Proper grouping is essential for assuring that
the most accurate deposition and decay characteristics are assigned to air toxics provided
in the emission inventory. The grouping decisions made for the national-scale assessment
reflect "downstream" data needs, such as making the resultant concentration estimates
reflect compounds for which health benchmark information exists. These decisions made
for dispersion and exposure modeling are not always the same as those used to
summarize pollutant groupings in the NTI.
For the NTI emission summaries, emissions of particular metal compound species (e.g.
lead oxide) belonging to groups were summed with no adjustment. For ASPEN
modeling, particular metal compound species were adjusted to account only for the moles
of the metal in the compound (e.g., emissions of the oxide fraction of lead oxide were
subtracted from the emissions of lead oxide). This resulted in apparently lower metal
compound emissions in the EMS-HAP summaries than the NTI summaries, but actually
this effect is merely the result of two different definitions of "metal compounds."
Differences due to different processing of emissions records that were reported in
pounds per hour. For the NTI emission summaries, these records were not included in
the summary because the operating schedule was unknown. In EMS-HAP, these records
were converted to tons/year assuming a maximum operating schedule (8760 hours per
year). This resulted in higher emissions in the EMS-HAP summaries than the NTI
summaries.
Differences due to geographic domain. The NTI national summaries include Alaska
and Hawaii. EMS-HAP national summaries do not.
54
-------
Table 4-2. Comparison of EMS-HAP and NTI emission totals.
Pollutant
1,1,2,2-Tetra-
chloroethane
1,3-Butadiene
1,3-Dichloro-
propene
7-PAH
Acetaldehyde -
Primary
Acrolein -
Primary
Acrylonitrile
Arsenic
Compounds
Benzene
Beryllium
Compounds
Cadmium
Compounds
Carbon
Tetrachloride
Chloroform
Chromium
Compounds
Coke Oven
Emissions
Diesel
Particulate
Matter
Ethylene
Dibromide
Ethylene
Dichloride
Ethylene Oxide
Formaldehyde -
Primary
Hexachloro-
A26benzene
Hydrazine
Manganese
Compounds
Mercury
Compounds
Methylene
Chloride
Nickel
Compounds
Perchloro-
ethylene
Polychlorinated
Biphenyls
Propylene
Dichloride
Quinoline
Trichloro-
ethylene
Vinyl Chloride
Major
EMS-
HAP
Emiss.
(tons/yr)
9
2748
4
105
8815
233
1001
316
13388
33
68
380
2752
676
1433
N/A
8
683
307
15959
1
25
1567
106
32117
600
6403
0
81
10
10354
851
NTI
Emiss.
(tons/yr)
9
2743
4
105
8799
234
1001
317
13487
33
68
372
2693
680
1433
N/A
8
684
279
15961
1
25
1567
106
30460
601
6411
0
81
10
10354
851
Diff.
0%
0%
0%
0%
0%
0%
0%
0%
-1%
0%
0%
2%
2%
-1%
0%
0%
0%
9%
0%
0%
0%
0%
0%
5%
0%
0%
0%
0%
0%
0%
0%
Area and Other
EMS-
HAP
Emiss.
(tons/yr)
116
16304
21350
707
18826
16457
258
76
60905
7
86
105
626
423
4
N/A
4
94
1114
142589
0
2
1116
51
52889
500
37510
0
15
4
15361
384
NTI
Emiss.
(tons/yr)
116
19148
21178
828
21691
19447
258
77
68940
7
84
104
615
429
4
N/A
4
87
1112
160622
0
2
1123
50
52477
517
37092
0
15
4
15077
384
Diff.
0%
-17%
1%
-17%
-15%
-18%
0%
-2%
-13%
0%
2%
1%
2%
-2%
0%
-4%
8%
0%
-13%
-1%
0%
-1%
1%
1%
-3%
1%
0%
0%
0%
2%
0%
Onroad Mobile
EMS-
HAP
Emiss.
(tons/yr)
0
23596
0
42
28790
4996
0
0
169060
0
0
0
0
14
0
133556
0
0
0
83354
0
0
6
0
0
11
0
0
0
0
0
55 o
NTI
Emiss.
(tons/yr)
0
23488
0
42
28698
4960
0
0
168212
0
0
0
0
13
0
N/A
0
0
0
83006
0
0
6
0
0
11
0
0
0
0
0
0
Diff.
0%
0%
0%
1%
0%
1%
0%
1%
1%
0%
0%
0%
0%
1%
0%
0%
0%
0%
0%
0%
0%
1%
1%
0%
1%
0%
0%
0%
0%
0%
0%
Nonroad Mobile
EMS-
HAP
Emiss.
(tons/yr)
0
9460
0
17
40764
7342
0
2
93623
0
0
0
0
34
0
341241
0
0
0
86374
0
0
35
7
0
89
0
0
0
0
0
0
NTI
Emiss.
(tons/yr)
0
9864
0
19
40828
7376
0
2
98703
0
0
0
0
35
0
N/A
0
0
0
86440
0
0
36
7
0
93
0
0
0
0
0
0
Diff.
0%
-4%
0%
-8%
0%
0%
0%
-5%
-5%
-6%
-6%
0%
0%
-1%
0%
0%
0%
0%
0%
0%
0%
0%
1%
0%
-5%
0%
0%
0%
0%
0%
0%
Total
EMS-
HAP
Emiss.
(tons/yr)
124
52108
21354
872
97195
29029
1259
395
336976
40
154
485
3378
1147
1437
474797
13
777
1421
328276
1
28
2724
163
85006
1199
43914
0
96
15
25715
1235
NTI
Emiss.
(tons/yr)
125
55243
21181
994
100016
32017
1259
396
349342
40
153
476
3308
1157
1437
N/A
13
771
1391
346028
1
28
2732
163
82937
1221
43503
0
96
15
25431
1235
Diff.
0%
-6%
1%
-14%
-3%
-10%
0%
0%
-4%
0%
1%
2%
2%
-1%
0%
-1%
1%
2%
-5%
0%
0%
0%
0%
2%
-2%
1%
0%
0%
0%
1%
0%
*Only HAPs for which concentration results have been posted are shown; NTI totals include Alaska and Hawaii, EMS-HAP totals do not
-------
Differences due to facilities with missing or incorrect location data. As mentioned
previously, locations for some facilities were corrected in EMS-HAP. In addition, 87
facilities in the NTI could not be associated with county coordinates. Forty-three of these
sources emit the pollutants targeted in the national-scale assessment, but could not be
included because of this missing location information. These differences would account
for variations among county totals where location default schemes placed facilities in
EMS-HAP in different counties than in the NTI, and would make emissions lower in the
EMS-HAP inventory where these facilities are missing (Colorado and Idaho).
4.2.2 Discussion of Inventory Uncertainties
Because the 1996 NTI is a composite of emissions estimates generated by state and local
regulatory agencies, EPA, and industry; and because emission estimation techniques vary
with the agency providing data, the pollutants and the source categories, it is
understandable that the uncertainties should vary widely among emissions estimates. In
some cases, an estimate may be derived from few or only one emissions measurement at
a similar source. The NTI estimates originated from a variety of sources and estimation
methods, as well as for differing purposes, and they will in turn vary in quality, number
of pollutants included, level of detail, and geographic coverage. EPA has not attempted to
verify estimation methodology from the primary sources submitting data or from other
EPA databases (e.g., state inventories, TRI). EPA has not undertaken a full QA/QC
evaluation of the NTI because (1) most estimates submitted by state and local agencies
did not include supporting documentation, (2) EPA was aware of the uneven quality and
planned to view the national-scale assessment results in light of that awareness, (3) this
national-scale assessment was planned as a trial effort, not a definitive result, and (4)
EPA did not have sufficient time or resources available. Nevertheless, EPA recognizes
that the lack of such an evaluation represents an important source of uncertainty in the
assessment. Table 4-3 summarizes the origins of the stationary source emission estimates
in the NTI.
Table 4-3. Summary of Data Sources to the NTI.
All Stationary Sources
All Point Sources
From state or local agencies
From MACT
From TRI
All Non-point stationary
sources
EPA-generated (via emission
factors/activity data)
From state or local agencies
and TRI
From MACT
Emissions
of 188 Air
Toxics
2,301,700
1,174,700
401,300
619,000
153,500
1,128,610
787,100
182,900
158,600
Percent
Emissions
34% of all point
source emissions
53% of all point
source emissions
13% of all point
source emissions
70% of all non-
point emissions
16% of all non-
point emissions
14% of all non-
point emissions
Number of
Facilities or
Categories
55,411
facilities
4,310
facilities
1,847
facilities
30 categories
405
categories
65 categories
Percent of
Facilities or
Categories
90%
7%
3%
6%
81 %
13%
56
-------
It should also be noted that toxic emissions data for nonroad equipment are much more
limited than data for onroad mobile sources. While EPA has basic emission factors for
VOC and PM for most of the nonroad categories, there is very little VOC speciation data
for the given categories that would allow EPA to develop good estimates of toxic
emission rates. Given the large variety of nonroad engine sizes, types and uses, as well
as the likelihood that this variety will result in some differences in VOC composition, it is
important that EPA obtain or develop speciated VOC data specific to each nonroad
category in order to more accurately project nonroad mobile source air toxics (MSAT)
inventories. In its section 202(1)(2) rulemaking for mobile source air toxics, the Office of
Transportation and Air Quality outlined a strategy to obtain and evaluate these data.
Some comparisons to other inventory databases and discussion of the primary sources of
uncertainty in the NTI are discussed in Section 4.2.1.2.1.
4.2.2. •/ Uncertainties in Completeness of Point Source Universe
Although there are not any other national-level air toxics inventories that include point
and non-point stationary and mobile sources available for comparison to the NTI, there
are other emissions data sets that have been used in other dispersion modeling exercises.
The only two nationwide air toxics data sets are the one used in the 1990 Cumulative
Exposure Project (CEP) study and the Toxics Release Inventory. In addition to these air
toxics inventories, the NET is a nationwide criteria pollutant inventory that includes
emissions from all source sectors mentioned above, but with different source category
definitions (e.g., major sources in the NET have 100 tons of criteria pollutant actual
emissions or other limits selected by the states submitting the data).
Cumulative Exposure Project (CEP)
The 1990 CEP included 1990 TRI data, but also relied heavily on VOC and PM emission
estimates from the interim 1990 NET inventory. These criteria pollutant emissions were
converted to individual pollutant emissions via speciation profiles. In general, speciation
profiles are industry-specific conversion factors that are used to estimate individual air
toxics emission rates from criteria pollutant emissions. This method creates an air toxics
emission inventory with minimal resources but produces an uncertain inventory,
particularly for stationary sources where industry-specific speciation profile information
is very limited. In addition much of the NET inventory is "grown" from prior years'
estimates and is thought to be less accurate than the year-specific emissions data
compiled in the NTI. Therefore, the NTI would be deemed to be of superior quality to
the CEP emissions estimates. A comparison of 1990 CEP inventory to 1996 NTI
emission totals for selected air toxics is presented in Table 4-4.
57
-------
Table 4-4. 1990 CEP and 1996 NTI National Emission Totals.
1990 CEP
POLLUTANT NAME
Benzene
Formaldehyde
Tetrachloroethylene
Methylene Chloride
Acetaldehyde
1,3 -Butadiene
Trichloroethylene
POM as 7-PAH
Acrolein
Coke Oven Emissions
1 , 3 -Dichloropropene
Chloroform
Ethylene Bichloride
Vinyl Chloride
Acrylonitrile
Manganese Compounds
Nickel Compounds
Carbon Tetrachloride
Ethylene Oxide
Arsenic Compounds
Chromium Compounds
1,2-Dichloropropane
Cadmium Compounds
Mercury Compounds
Ethylene Dibromide
1, 1,2,2-Tetrachloroethane
Beryllium Compounds
PCBs
Hydrazine
Quinoline
Hexachlorobenzene
Total
587285
412450
116435
115340
112639
87600
82344
76504
72628
29127
20732
17520
11570
6351
5110
4030
4015
3978
1836
1095
996
735
299
267
223
35.1
33.8
29.2
15.4
13.8
9.96
1996 NTI
Total
349342
346028
43503
82937
100016
55243
25431
994
32017
1437
21181
3308
771
1235
1259
2732
1221
476
1391
396
1157
96.1
153
163
12.8
125
40.2
0.21
27.5
14.8
1.00
Toxics Release Inventory (TRI)
The TRI differs from the NTI in several fundamental ways. Most significantly is the fact
that the TRI universe of sources is a subset of the NTI universe, although the TRI
contains many pollutants other than the 188 air toxics. The TRI contains data from only
industrial sources that are large enough (in terms of chemicals usage, emissions, number
of employees, etc.) to meet certain reporting requirements. In addition to these TRI
sources, the NTI also contains emission estimates for smaller sources, sources not
associated with industry (e.g., wildfires, consumer product usage) and mobile sources.
The two inventories also differ in the level of detail. For instance, TRI sources report
"stack" and "fugitive" emissions, while the NTI contains point source process parameter
58
-------
details necessary for computer modeling (e.g., stack height, flow rate, temperatures, etc.)
that may include many emission points (stacks) at a given facility. EPA uses information
from the TRI and other EPA programs, as well as information reported voluntarily by
state and local agencies, to monitor emissions and emission trends across a broad
spectrum of source categories.
National Emission Trends (NET)
The NET contains criteria pollutant emissions for point, non-point stationary, and mobile
sources, compiled by EPA in conjunction with state and local criteria pollutant inventory
submittals. The NET and NTI inventories are designed for different purposes and contain
emissions estimates for different pollutants (criteria pollutants in the NET, air toxics in
the NTI). However, there may be significant overlap of included point sources,
particularly among large point facilities. Therefore, the NET provides a good resource
for checking the NTI's completeness and for gleaning missing location or facility details.
Table 4-5 compares the number of major source NTI facilities with major source
facilities in the TRI and unique facilities in the NET. Note that for states that did not
provide air toxics inventories to EPA, the 1996 NET includes approximately four times
the number of facilities as the NTI. This fact suggests that large point source facilities
may be missing from the NTI in these states.
Table 4-5. Facility Count Summary.
State
AK*
AL
AR
AZ
CA
CO
CT
DC*
DE
FL
GA*
HI*
IA*
ID
IL
IN
KS
KY
LA
MA
MD
ME
Ml*
MN
MO
MS
Unique
Facilities in
NTI
16
240
301
244
7,416
3,424
68
1
48
359
240
16
138
56
8,813
1,587
150
250
301
311
730
144
341
210
791
150
Major
Source
Facilities in
TRI
3
151
99
24
176
20
49
0
18
102
158
3
101
6
220
287
75
98
89
55
35
20
168
103
116
104
Unique
Facilities in
NET
28
811
114
309
18,870
4,645
660
14
87
509
417
155
63
21
9,713
1,321
1,963
370
664
495
439
223
1,966
696
758
131
States that did not provide point source data files to draft NTI
State
MT*
NC
ND
NE
NH
NJ
NM
NV*
NY
OH*
OK*
OR
PA
Rl
SC
SD
TN
TX
UT
VA
VT
WA
Wl
WV
WY
TOTAL
Unique
Facilities in
NTI
28
2,369
56
290
53
148
34
12
3,528
421
94
187
586
505
406
24
502
2,426
137
1,460
86
269
744
172
115
40,997
Major
Source
Facilities in
TRI
10
230
11
40
13
62
6
4
103
275
71
61
206
15
116
16
193
288
24
137
3
66
147
39
8
4,424
Unique
Facilities in
NET
217
925
65
704
184
864
299
109
1,584
1,900
373
395
977
110
489
22
588
1,203
333
2,299
126
276
1,585
229
270
61,568
59
-------
4.2.2.2 Uncertainties Due to the Dynamic Nature of Emission Inventories
Like other air pollution inventories, the NTI data set for 1996 is dynamic, and it will
likely change, as resources permit, with periodic updates as new, more reliable data
become available. For this reason, future assessments based on the 1996 NTI may
produce somewhat different results. As the results of the ASPEN and subsequent
HAPEM models are evaluated, modifications to the emission inventory input will
continue to surface such as the lead emissions that are known to be missing from the
Missouri lead smelter discussed in section 3.2.1.3.1. Also, it is likely that some facilities
were double counted under different names, and that some plants that had closed have not
been omitted from the 1996 data set.
The county-level mobile, area and other sources may be based on VMT, activity data,
and/or emission factors, and these components are updated routinely. For example, the
nonroad mobile source air toxics emissions are based on VOC and PM estimates from the
June 2000 version of the NONROAD model. When this model run was complete, the
VOC and PM nonroad estimates from individual equipment categories were consolidated
into gasoline 2-stroke and 4-stroke and diesel PM source sectors and used to prepare the
NTI air toxics estimates. However, subsequently a decision was made to revise inventory
estimates for recreational equipment. When the revised 1996 NET was developed, these
estimates were replaced with significantly lower recreational equipment estimates
originally developed for the 1991 EPA Nonroad Engine and Vehicle Emission Study.
The change affects both 2 and 4-stroke emissions, but it is most noticeable with 2-strokes,
since that is what makes up most of the recreation equipment VOC inventory. Additional
revisions to inventory estimates for recreational equipment and other equipment
categories are anticipated before a final version of NONROAD is released. (More
background on the draft NONROAD model, its history and recent developments are
available at http://www.epa.gov/otaq/nonrdmdl.htm)
4.2.2.3 Uncertainties in Emission Locations
The location of emissions is important for all types of sources, whether expressed at a
given latitude/longitude or at the county level. Many locations are unknown or
incomplete, yet for dispersion modeling each facility point source must have an exact
location. If the location coordinates were missing from point sources, they were placed
via the default mechanisms in EMS-HAP. How well this default location reflects reality
will affect the model results and the uncertainty of the overall analysis. To illustrate this
point, Table 4-6 shows, by state, the percent of point sources, and the corresponding
percent of emissions of lead, chromium, and cadmium emissions, that were located via
default algorithms. The table shows the percentages based on sources with default
locations determined by the facilities' zip codes or counties, as well as the percentage of
facilities were not defaulted because their coordinates were within the correct county or
within reasonable bounds outside the county. Although this comparison does not provide
a quantitative assessment of uncertainty, the magnitude of sources and emissions
defaulted in a given state suggests the uncertainty in the results due to potential location
errors.
60
-------
Table 4-6. Summary of Defaulted Sources and Emissions for Lead, Chromium, and Cadmium.
State
All
All
All
All
Alabama
Arizona
Arkansas
California
Colorado
Connecticut
Delaware
District of
Columbia
Florida
Georgia
Idaho
Illinois
Indiana
Iowa
Kansas
Kentucky
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
Nevada
New Hampshire
New Jersey
Metal
All
Lead
Cadmium
Chromium
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
Percent of Sources
County
Defaulted
10.7
10.8
10.3
10.9
23.3
1.3
9.7
19.1
6.4
37.5
0.0
0.0
11.3
17.6
13.0
1.7
14.3
3.4
2.0
6.6
22.5
5.8
19.5
3.2
12.7
22.7
16.3
14.2
13.6
34.9
12.5
0.0
1.5
Zip Code
Defaulted
0.6
0.5
0.4
0.5
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
1.7
0.0
0.0
0.0
1.9
0.8
0.0
1.6
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Not
Defaulted,
Outside
County
4.4
4.1
5.2
4.7
3.8
9.0
2.9
0.7
0.0
4.2
0.0
0.0
4.0
1.6
4.3
0.4
40.3
0.0
0.0
0.0
2.2
5.8
2.3
0.0
0.4
0.0
2.3
3.1
0.0
0.0
0.0
0.0
3.1
Not
Defaulted,
Inside
County
84.4
84.6
84.2
83.8
72.9
89.7
87.4
80.1
93.6
58.3
100.0
100.0
84.8
80.8
82.6
97.8
45.0
94.9
98.0
93.4
75.3
86.5
77.4
96.8
85.3
77.3
81.4
82.7
86.4
65.1
87.5
100.0
95.4
Percent of Emissions
County
Defaulted
10.5
12.8
25.0
3.5
58.6
2.0
10.5
9.6
1.1
26.4
0.0
0.0
6.7
46.0
0.9
0.4
21.2
0.1
0.0
7.0
25.8
0.5
6.1
0.3
16.7
35.7
40.8
4.5
8.9
3.7
0.3
0.0
0.0
Zip Code
Defaulted
0.2
0.3
0.2
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
0.2
1.3
0.0
0.0
0.0
0.2
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Not
Defaulted,
Outside
County
3.7
3.8
6.4
3.0
4.1
29.3
5.3
0.3
0.0
0.3
0.0
0.0
0.6
1.4
0.0
0.0
32.4
0.0
0.0
0.0
0.1
0.2
4.1
0.0
0.1
0.0
1.0
0.4
0.0
0.0
0.0
0.0
0.1
Not
Defaulted,
Inside
County
85.6
83.1
68.4
93.4
37.3
68.7
84.2
90.1
98.9
73.3
100.0
100.0
92.8
52.7
99.1
99.6
46.2
98.6
100.0
93.0
74.0
99.1
89.8
99.7
83.2
64.3
58.2
95.1
91.1
96.3
99.7
100.0
99.9
61
-------
Table 4-6. Summary of Defaulted Sources and Emissions for Lead, Chromium, and Cadmium.
State
New Mexico
New York
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
Pennsylvania
Puerto Rico
PJiode Island
South Carolina
South Dakota
Tennessee
Texas
Utah
Vermont
Virginia
Washington
West Virginia
Wisconsin
Wyoming
Metal
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
All
Percent of Sources
County
Defaulted
0.0
1.2
2.6
16.7
3.1
8.3
48.8
10.2
20.0
36.4
4.8
12.5
61.1
10.2
12.5
10.3
17.7
24.1
1.9
5.4
0.0
Zip Code
Defaulted
0.0
0.1
0.1
0.0
0.4
0.0
3.1
0.4
0.0
59.1
0.0
0.0
16.6
0.1
0.0
0.0
0.0
0.0
0.0
1.5
0.0
Not
Defaulted,
Outside
County
0.0
0.7
1.9
0.0
1.3
5.6
1.6
1.3
80.0
0.0
5.7
0.0
0.6
1.6
0.0
6.9
1.0
1.7
0.0
1.5
0.0
Not
Defaulted,
Inside
County
100.0
97.9
95.3
83.3
95.1
86.1
46.5
88.1
0.0
4.5
89.6
87.5
21.7
88.1
87.5
82.8
81.3
74.1
98.1
91.7
100.0
Percent of Emissions
County
Defaulted
0.0
13.7
1.5
2.0
21.3
15.1
28.3
1.9
2.8
10.7
34.8
5.5
16.4
24.9
0.8
1.2
40.8
39.2
0.8
0.5
0.0
Zip Code
Defaulted
0.0
0.0
0.0
0.0
0.1
0.0
0.7
0.0
0.0
88.0
0.0
0.0
6.2
0.1
0.0
0.0
0.0
0.0
0.0
0.0
0.0
Not
Defaulted,
Outside
County
0.0
0.0
0.1
0.0
0.8
2.9
0.1
5.1
97.2
0.0
0.1
0.0
0.0
0.8
0.0
10.6
0.6
0.5
0.0
0.4
0.0
Not
Defaulted,
Inside
County
100.0
86.3
98.4
98.0
77.7
82.0
70.9
93.0
0.0
1.3
65.1
94.5
77.4
74.3
99.2
88.2
58.7
60.3
99.2
99.1
100.0
Nationwide, for primary emissions of pollutants that were modeled (i.e., no reactive
precursors to secondary formation), the following statistics apply to default locations:
Out of the 48,657 sites (there can be multiple sites at a single facility) from the point
source inventory simulated via the ASPEN model (i.e., those that contained the 33 urban
air toxics and were not dropped due to lack of information on locations or zero
emissions), 4212 were defaulted by county-level defaults (census tract centroids) and 460
were defaulted by zip-code (zip code centroids). There were also 43 sites that were
dropped from the EMS-HAP input prior to modeling due to the fact that their locations
were identified as "portable" and therefore could not be defaulted.
In addition to point sources that required exact location coordinates, all county-level
emissions needed to be spatially allocated to the census-tract level prior to ASPEN
modeling. Each source category's emissions were placed based on an appropriate
surrogate such as population or land use. In some cases, the initial emission estimate
may have been made at the national or state level and allocated to counties by the same
62
-------
surrogate. This type of spatial placement may lead to emissions being too dispersed or
too concentrated at a given location. For example, landfill emissions were first placed in
the appropriate counties based on information collected via MACT standard
development. Subsequently, within EMS-HAP, these emissions were further allocated to
census tracts based on reciprocal population density. This assumption is based on the
concept that landfills are placed away from populated areas. In reality, landfills are
located at specific geographic coordinates rather than spread throughout census tracts
within a county. Since these exact location data were not available, the surrogate
approach was used as a best approximation.
The spatial allocation of county emissions, which takes place within EMS-HAP is
described above in 3.2.1.1. This necessary approximation of emission location adds
uncertainty to the overall analysis although it represents the best method available. As
shown in Table 3-3, some of the allocation surrogate information used in this process is
very old (e.g., 1970's).
Uncertainty due to spatial allocation schemes is also particularly pronounced for mobile
source categories, especially within the nonroad sector where, in order to simplify
computation, the county-level emissions from almost 100 source categories (e.g.,
recreational marine vessels, lawn mowers, construction equipment, etc.) were
consolidated into three source category groups (2-stroke gas, 4-stroke gas, and diesel
engines) prior to spatial allocation.
The spatial allocation methods used in this study were revised for some source categories
as ASPEN concentration results were reviewed in order to improve the overall analysis.
4.2.2.4 Uncertainty Due to Stack Parameter Defaults
A facility in the NTI can contain many emissions release points. Each release point
requires several process parameters in order to characterize it sufficiently for dispersion
modeling. When these parameters were missing from the original inventory data or were
out of reasonable range, they were replaced with defaults either within the NTI or EMS-
HAP. Table 4-7 describes the number of vertical emission release points that had
defaulted process parameters. Of the 97,365 unique vertical stacks, 63,292 contained at
least one defaulted stack parameter.
Table 4-7. Stack Parameter Default Statistics.
Parameter
NTI Stack Height
NTI Stack Diameter
NTI Exit Velocity
NTI Exit Temperature
Number of Defaults
in NTI
28,550
32,176
40,361
28,890
in EMS-HAP
16,048
20,722
20,816
12,982
Percent of Total
Vertical Stacks
45.8
54.3
62.8
43.0
63
-------
4.2.2.5 Variations in Reported Emissions
Since no standardized requirements for collecting or submitting air toxics inventories to
EPA currently exist to make these databases consistent, and because resources among the
agencies are variable, the quality, coverage and number of air toxics in these inventories
vary significantly. For instance, the emissions of poly chlorinated biphenyls (PCB)
appear in only 121 counties in the NTI out of a national total of 3,145 counties. It is
unclear whether the reported emissions are correct and additional emissions are missing
from the other counties, or if PCB, a banned substance, is no longer emitted from
anthropogenic sources.
4.2.2.6 Uncertainties Due to Pollutant Groupings
As discussed previously, many of the air toxics from the NTI were reported either as
groups of compounds (e.g., lead compounds) or individual chemical species (e.g., lead
oxide). For subsequent modeling, these pollutants had to be grouped together in order to
create consistency across the model's geographic domain. This creates uncertainties
because the groupings made for the national-scale assessment do not necessarily account
for the difference in toxicological characteristics among individual species in the group.
In addition, numerous assumptions were made to establish the pollutant groups and the
resulting pollutant group characteristics (e.g., deposition).
Another related uncertainty in grouping metal species stems from the adjustments made
to the mass of metal species emissions to allow for modeling only the metal portion of the
compound. First there is an assumption that when emissions for a particular species are
reported, they are for the entire compound and not just the metal portion. Second, for
metals reported as broad compound groups or subgroups (e.g., lead & compounds,
alkylated lead), no adjustments were made since the particular species was not known.
4.2.2.7 Particle Size and Reactivity Assignments
All of the emissions input for ASPEN modeling must be identified by particle size
fractions (coarse versus fine) and reactivity. This tells the dispersion model how to
account for deposition and subsequent chemical reactions. None of the emission
inventories used in this assessment contained this information, and so it had to be
assigned by EMS-HAP. Since ASPEN uses different deposition rates for fine and coarse
particulates, each metal particulate must be classified by its particle size. Except for
mercury, the percentage of a species assigned to coarse/fine particulate (the coarse/fine
split) was estimated as the percentage of coarse/fine particulate obtained from the
inventory used for the Cumulative Exposure Project. These percentages vary among the
onroad, nonroad and stationary source categories, but do not vary within those broad
groups. For example, the coarse/fine split for chromium compounds emitted from a
chromium electroplating operation is the same as for chromium emitted from an
incinerator. For mercury, it was assumed that all mercury species reported are gaseous
except for mercury emissions from mobile sources and mercuric chloride. Mercuric
chloride was assumed to be 100 percent fine particulate. These assignments add to the
overall uncertainty.
64
-------
4.3 Environmental Fate and Transport Characterization
4.3.1 Summary of ASPEN Modeling Results
The ASPEN model was used to estimate ambient concentrations at each census tract in
the contiguous United states, plus Puerto Rico and the Virgin Islands. Based on these
estimates, annual average modeled concentrations (in ng/m3) were calculated for each
state and territory, and the statistical distributions described by the 25th, median, and 75th
percentiles across all census tracts were calculated and displayed graphically for each of
the 33 air toxics; national averages were similarly calculated. These "statewide estimate"
charts can be viewed at www.epa.gov/ttn/uatw/nata/chartconc.html. A value for
background concentration, except for diesel PM, was also displayed. In addition, the
percent contribution to the statewide annual average ambient concentration estimates
were calculated and displayed. These "percent contribution" charts can be viewed at
www.epa.gov/ttn/uatw/nata/chartconc.html. Figures 4-7 and 4-8 provide an example of
each type of display for benzene. As can be seen in Figure 4-7, benzene is fairly
ubiquitous, is prevalent in all states, and as expected, the median value is generally higher
in the more populated or industrial states. As can be seen in Figure 4-8, the percent
contribution for benzene is higher for mobile sources (both onroad and nonroad) than
stationary sources. The contribution from background is also appreciable.
When examining the source contribution plots for all of the modeled air toxics (at
http://www.epa.gov/ttn/uatw/nata/chartconc.html), it can be seen that no single source
sector is the main contributor to the estimated concentrations. Table 4-8 summarizes the
dominant contributor for each of the modeled air toxics. The results show that, on a
national level, about half of the pollutants have "area and other sources" as the dominant
contributing source sector. As seen in the table, the dominant source sector may be
different when the source contributions are examined at the state level. For example,
except for coke oven emissions, all of the pollutants in which major is the dominant
source sector at the National level show the area and other category to be dominant for a
number of states. For those pollutants where background is dominant, the next highest
contributing source sector is the "area and other" sector.
To further explore the geographic variability of the source sector on the ambient
concentrations, we examined the dominant source sector for the counties with the 10
highest median concentrations. Table 4-9 shows the number (out of 10) of these
concentrations dominated by the various source types. The asterisked pollutants in the
table are those in which the dominant source type differed from the dominant types at the
national level and state levels as shown in Table 4-8. Tables 4-8 and 4-9 show that most
pollutants vary geographically in their dominant source sector.
65
-------
Table 4-8. Pollutants grouped by the dominant source
sector affecting their national average concentrations.
Area and Other
ArsenicM
BerylliumM
CadmiumM
ChromiumM
1 , 3 -dichloropropene
Ethylene oxideM
ManganeseM
Methylene chloride
Nickel
Perchloroethylene
POMM
1,1,2,2-
Tetrachloroethane
Vinyl chlorideM
Major
AcrylonitrileA
Coke oven emissions
HydrazineA
Propylene dichlorideA
QuinolineA
Mobile
(onroad and nonroad
combined)
Acetaldehyde
AcroleinA
Benzene
1,3 -Butadiene
Formaldehyde
Background
Carbon tetrachloridea
Chloroform3
Ethylene dibromidea
Ethylene dichloridea
Hexachlorobenzenea
Mercury3
Polychlorinated biphenyls
(PCB) a
TrichloroethyleneA' a
M = for several states, the dominant sector is major;
A= for several states, the dominant sector is area and other
A= next highest sector after background is area and other
Table 4-9. Number of Counties of Each Dominant Source Sector
for the 10 Highest County Median Concentrations.
Pollutant
Acetaldehyde
Acrolein
Acrylonitrile
Arsenic Compounds
Benzene
Beryllium Compounds
1,3 -Butadiene
Cadmium Compounds
Carbon Tetrachloride
Chloroform
Chromium Compounds
Coke Oven Emissions
1 , 3 -Di chl oropropene
Ethylene Dibromide
Ethylene Bichloride
Ethylene Oxide
Major
2
3
7
3
8
7
10
Area and
Other
3
10
7
3
7
7
3
10
10
Mobile
10
5
10
3
Background
10
2
10
10
66
-------
Pollutant
Formaldehyde
Hexachlorobenzene
Hydrazine
Manganese Compounds
Mercury Compounds
Methylene Chloride
Nickel Compounds
Perchl oroethy 1 ene
Polychlorinated Biphenyls
Polycyclic Organic Matter
Propylene Bichloride
Quinoline
1 , 1 ,2,2-Tetrachloroethane
Tri chl oroethy 1 ene
Vinyl Chloride
Major
7
7
8
6
1
7
1
2
9
Area and
Other
6
O
O
O
6
2
4
9
1
10
10
O
9
8
1
Mobile
4
Background
7
4
9
A comparison of estimated annual average concentration distribution for 10 selected
pollutants is shown in Figure 4-9. These pollutants were selected to include air toxics
that are relatively inert (e.g., benzene), air toxics that are subject to chemical
transformation in the atmosphere (e.g., formaldehyde), and metal particles that are
subject to deposition. As can be seen in this figure, the distribution of the annual average
concentrations among pollutants spans several orders of magnitude. Benzene and
formaldehyde have much higher annual average concentrations than metals. One to two
orders of magnitude variability exists for each air toxic.
Figure 4-10 provides a comparison of estimated annual average concentration for the
same 10 selected pollutants with each census tract categorized as urban or rural. Each
census tract is designated as either urban or rural as part of the dispersion modeling
process, since dispersion parameters differ for these two types of locations. In general,
census tracts with population density greater than 750 people/km are designated as
urban, while other census tracts are designated as rural [15]. This results in
approximately an even split of census tracts into the urban and rural designations. As can
be seen in this figure, the annual average concentrations are generally higher in the urban
census tracts than rural tracts; this result is likely due both to higher emission densities
for many pollutants in urban areas and to a closer proximity of many emission sources to
the modeled receptors (i.e., census tract centroids).
Figure 4-11 shows, for each State, the relative contributions of pollutants, broken out into
three categories: mobile source dominated pollutants, metals, and stationary source non-
metals. In each bar of this figure, 49 boxes representing 48 states and the District of
Columbia are displayed. Concentrations are scaled by a maximum value and sorted from
minimum to maximum. Thus, each box represents the magnitude of relative average
concentration for the state. As can be seen from this figure, some pollutants (e.g.,
67
-------
propylene dichloride) have an impact in only one or two states while other pollutants
(e.g., benzene) appear more uniformly distributed across the country.
4.3.2 Discussion of Results
From an examination of the ASPEN dispersion modeling results, the following general
observations can be made:
• Concentration estimates are a complex function of a number of factors, including
emissions density (number of sources in a particular area), meteorology and
source characteristics, rather than just total emissions.
• Of the four main source sectors (area and other, major, onroad, nonroad), no one
type is a dominant contributor to the estimated concentrations. On a national
level, about half of the pollutants have "area and other sources" as the dominant
contributing source sector.
• There is considerable variability among the national, state and the county levels in
terms of contributions of certain pollutants by source type.
• Both emissions and estimated concentrations are generally higher in urban than in
rural areas.
• Because different types of sources contribute to emissions in different areas of the
country, the highest ambient average concentrations of the individual pollutants
occur in different states (i.e., no one state has the highest concentrations of all the
pollutants).
• Some pollutants (e.g., benzene) are more evenly distributed around the country
while other pollutants (e.g., vinyl chloride) are more related to isolated areas of
industrial activity.
• Most of the stationary non-metal pollutants have higher concentrations in only a
few states whereas the concentrations from metals and mobile source-dominated
pollutants are more evenly distributed across the country. Among the 20 modeled
non-metal pollutants, twelve have impacts on one to three states, and only four are
widely distributed across states. For metals, two pollutants out of eight have a
major impact on one or two states, while the other six pollutants have an impact in
more than ten states. Mobile source-dominated pollutants are generally more
uniformly distributed across the country.
• Background concentrations are an essential part of the total air quality
concentrations and include pollutant concentrations due to natural sources,
sources not in the emissions inventory, and long-range transport. Based on the
CEP study, non-zero estimates of background concentrations for only 13
pollutants are available. For seven of these (PCBs, ethylene dibromide, carbon
68
-------
tetrachloride, hexachlorobenzene, ethylene dichloride, chloroform, and mercury),
the background dominates the total estimated average concentration.
4.3.3 Comparison of ASPEN Modeling System to Monitoring Data
The ASPEN system as designed and applied in the initial national scale assessment
provides broad-scale air quality impacts (census tract resolution). The assessment
system is not designed to capture more localized areas of high concentration or "hot-spot"
impacts. The system evaluation presented below (model-to-monitoring comparison
study) was conducted to evaluate the ASPEN system's ability to predict an air quality
concentration at a given point in space (i.e., an air quality
monitor). No detailed evaluation of the ASPEN system's
inability to predict "hot-spot" impacts was conducted. It is
anticipated that, following completion of more detailed local
scale assessments that are to be conducted as part of future impacts. Such impacts could
^ be orders of magnitude
The initial National-Scale
Assessment does not capture
local-scale or "hot spot"
greater than those predicted
at the census tract resolution
provided by the ASPEN
modeling system.
NATA activities, such an evaluation can be made. However,
EPA conducted a few crude comparisons to assess the
potential for the ASPEN system to underestimate "hot spots",
as predicted by local-scale modeling assessments. One such
comparison, for a major source in an urban environment
found local fence line impacts approximately 30 times higher then those predicted at the
census tract centroid by ASPEN in the initial national scale assessment. A similar
comparison in a more rural setting found maximum predicted fence line impacts greater
than 2 orders of magnitude above those predicted by ASPEN at the census tract centroid.
While these two comparisons are not a complete evaluation of expected differences
between the ASPEN initial national scale assessment predictions and those from local
scale "hot spot" assessments, they are simple examples of why other scale assessments
must be considered to fully understand the air toxic issues in a particular location.
The remainder of this section compares the modeled air quality estimates with currently
available, but geographically limited, ambient air monitoring data. A representative
subset of seven air toxics (benzene, perchloroethylene, formaldehyde, acetaldehyde,
cadmium, chromium, and lead) was selected for this evaluation. These pollutants were
selected because they represent impacts by different combinations of mobile, county-
level and point sources; include reactive and non-reactive compounds; and include those
with both primary emissions and secondary formation in the atmosphere. They also
include those air toxics with the largest number of monitoring sites.
For each monitor-pollutant combination, a comparison was made between the monitored
concentration and the concentration estimated by the ASPEN model at the monitor
location to get point-to-point comparisons. Steps in this comparison, for each pollutant at
each monitor, were as follows:
The ASPEN model was used to estimate concentrations at the exact locations of the
monitors. Using the monitor latitude and longitude coordinates, the monitor location is
assumed to represent a census tract with 100 m radius. Concentrations are interpolated to
69
-------
this "census tract" using the log-linear interpolation scheme used in ASPEN to get the
annual average estimate.
For monitored concentrations, an annual average estimate was calculated by averaging all
of the measured values obtained by a monitor across the course of a year. In calculating
the average monitored concentration, any individual samples where the concentration
was found to be below the detection limit were assigned a value of one-half the detection
limit. Further details are available in the full report (Appendix J, Section Il.b.ii.)
Model-monitor ratios are calculated by dividing the modeled concentration by the
monitored concentration. A ratio of 1.0 means that the modeled concentration equals the
monitored concentration, while a ratio less than 1.0 means that the model underestimates
the monitored concentration, and a ratio greater than 1.0 means that the model
overestimates the monitored concentration. A detailed explanation of the results are
presented in the full report (Appendix J), Section V.
In general, the modeled estimates for most of the pollutants examined are typically lower
than the measured ambient annual average concentrations when evaluated at the exact
location of the monitors. However, when the maximum modeled estimate for distances
up to 10-20 km from the monitoring location are compared to the measured
concentrations, the modeled estimates are closer to monitored concentrations. This result
can be attributed, in part, to spatial uncertainty of the underlying emission and
meteorological data, and the tendency of current air toxics monitoring networks to
measure the higher, if not highest, local air pollution impact areas. It also shows that the
model estimates are more uncertain at the census-tract level but are more reliable for
larger geographic scales, like county or state. Nevertheless, there are many locations for
several of the studied air toxics (including the aldehydes and metals) for which the model
estimates are still significantly lower than the measured concentrations even at distances
up to 50 km. For these instances, the difference between modeled and monitored
concentrations might be attributed to underestimated or missing emissions data in the
model as well as uncertainty in chemical transformation for the aldehydes. The
limitations of modeled concentrations resulting from isolated point sources using a
geographically sparse ASPEN receptor network in rural areas may be another
contributing factor. Yet another reason for the discrepancies may be attributed to
monitors being sited to find peak concentrations. Often, the ambient concentration
decays quickly around the peak area. Even under the scenario of a "perfect" model and
"perfect" monitors, if the monitor is sited exactly at the peak and the emissions (or
meteorological inputs) inputs are even slightly inaccurate, the model will tend to
underestimate results. This would be especially likely for pollutants dominated by point
sources with elevated releases, because any errors in release height, exit velocity, and/or
emissions location will likely cause the model to find a peak concentration different from
the true peak level. A more detailed discussion of the uncertainties of ASPEN and the
monitoring data can be found in Appendix J of this document. A description/basis of the
methodology used for the model-to-monitor comparison is given in Appendix I, and a
detailed discussion of the model to monitor comparison results is provided in Appendix J.
70
-------
A summary of the results from the model-to-model comparison are presented below.
Table 4-10 summarizes the comparisons on a point-to-point basis. The best agreement is
observed for benzene. The results are within a factor of two for 89 percent of the cases.
The median ratio of model to monitor comparisons is 0.92. The lack of agreement for the
other air toxics on a point to point basis can also be seen in Table 4-8, which shows the
median ratios varying between 0.65 for formaldehyde to 0.17 for lead. The number of
points with agreement within a factor of 2 (or within 30 percent) are also correspondingly
lower. These results can also be observed using the ratio box plot graphs, shown in
Figure 4.12.
Table 4-10. Comparison of the Measurement Data to Modeled Concentration.
Pollutant
Benzene
Perchl oroethy 1 ene
Formaldehyde
Acetaldehyde
Lead
Cadmium
Chromium
Number
of Sites
87
44
32
32
242
20
36
Median of
Ratios*
0.92
0.52
0.65
0.60
0.17
0.18
0.15
Percentage
within factor
of two**
89
55
53
59
18
15
28
Percentage
within factor
of 30%***
59
32
28
22
10
5
19
Percentage
Under-
estimated
59
86
88
91
91
85
83
Ratio represents (Ambient Measurement Concentration/Modeled Concentration).
" This represents the percentage of sites for which the model estimate is somewhere between half and
double the monitor average.
*** This represents the percentage of sites for which the model/monitor ratio is between 0.7 and 1.3.
While all the air toxics except benzene show relatively poor agreement on a point-to-
point basis (with the modeled estimates being systematically lower than the monitor
averages), they compare more favorably when the maximum model-estimated
concentration is examined within 30 km of the monitoring site as shown below in Table
4-11.
This technique is referred to as "MAXTOMON" and compares the maximum model
estimate within x kilometers of the monitor TO the monitor average. All model estimates
are considered (both estimates at monitor sites as well as the estimates at census tract
centroids) in computing the maximum values. This is an example of a point-to-range
tool. This tool is used here to test whether the frequent underestimation by the model at
monitoring sites was due to location uncertainties or due to systematic underestimation.
The reader is referred to Appendix J for further details on this technique.
71
-------
Table 4-11. Maximum Modeled Concentration Compared to Monitored Value.
Pollutant
Benzene
Perchloroethylene
Formaldehyde
Acetaldehyde
Lead
Cadmium
Chromium
Number of
Sites
87
44
32
32
242
20
36
Percent Missing Low at a Radius of:
0 km (Exact
Monitor Site)
59
86
88
91
91
85
83
10km
25
43
56
56
65
60
39
20km
20
23
31
38
51
35
28
30km
11
9
31
34
40
25
25
The improved comparisons, using the MAXTOMON technique, shown in Table 4-12 can
be attributed to three reasons:
• Many emission sources are not precisely located. The EMS-HAP model defaults
locations when they are not provided or when total emissions exist for the county.
Note, however, that this could contribute to either under or over predicting and is
a "non-bias" source of uncertainty.
• Many of the monitors were likely sited to find peak concentrations. For the point
source situations with elevated emission releases, the monitors frequently
represent hot spot locations where the ambient concentration falls off quickly
around the peak concentration area.
• Since the emissions inventory is likely missing sources, the modeling assessment
is under predicting air concentrations.
In the following sections, the model-to-monitor comparisons are discussed for benzene,
other gases, and for the metals.
Benzene
The relationship between model estimates and monitored values for benzene can be
described by the scatter plot shown in Figure 4-12. Figure 4-12 shows the point to point
comparison of modeled and monitored annual average concentrations. As also shown in
Figure 4-13 and Table 4-10, most of the points in the scatter plot fall between the 2:1 and
1:2 lines, showing good agreement between model predictions and monitor
measurements.
72
-------
This result is not surprising given the availability of good monitoring and emissions data
for this ubiquitous pollutant.
Perchloroethylene, Formaldehyde, and Acetaldehyde
The model-to-monitor relationship on a point-to-point basis is similar for the other three
gases investigated (perchloroethylene, formaldehyde, and acetaldehyde). In the ratio box
plots in Figure 4-13, however, one can see that the model's estimates tend to be lower
than the monitor averages. The typical values, however, agree well with the median
ratios all within a factor of 2. Nevertheless, a large percentage of the modeled estimates
are less than the monitored concentration for these gases on a point-to-point basis (see
Table 4-10 and Figure 4-13). This can be attributed, in part, to spatial uncertainty in the
underlying emissions for these pollutants; missing source emissions data; and/or
underestimated emissions.
To examine spatial uncertainty in the modeling system for the gases, the monitored
concentration is compared to the maximum modeled estimate in its vicinity (the
"MAXTOMON" technique described above). The results for these gases are presented in
Table 4-11. Table 4-11 shows nearby modeled concentrations which are greater than the
measured average concentration for many of the monitors.
This is especially true for perchloroethylene. In close vicinity (10 km or less) of most
monitors, higher modeled concentrations are observed. This result for perchloroethylene
suggests that uncertainties in the magnitude and location of the nearby area sources may
at least be partly responsible for the underestimation on a point-to-point basis. In other
words, the model may be systematically underestimating ambient concentrations or it
may just be finding the peak concentration in the wrong place.
For the two aldehydes, many monitors also have nearby modeled concentrations which
are greater than their measured values. However, a large fraction of the aldehyde
monitors cannot be associated with larger modeled values, even within 50 km. This
suggests systematic underestimation by the modeling system for the aldehydes, at least
for some of the areas. This may be attributed, in part, to the nature and treatment of these
air toxics. It may also be due to underestimated emissions. The two aldehydes are
mobile-source dominated, but a large fraction of their ambient concentrations are
secondarily formed. The chemical reactions resulting in their formation are simulated in
ASPEN. This adds an additional source of uncertainty to the modeling system and
distinguishes the aldehydes from the other air toxics in this comparison.
Metals
For the metals, the monitored concentrations are typically much higher than the modeled
concentrations when compared at the same location. The difference is most dramatic for
source-oriented monitors. Based on the median ratio, the source-oriented lead monitors
are typically underestimated by a factor of 7.5, and the others are underestimated by a
factor of 4.9. Only 17 percent of the source-oriented monitors and 18 percent of the other
monitors are estimated within a factor of two at the exact location of the model estimate.
73
-------
A combination of several factors may be responsible for these discrepancies:
• Missing emissions from the inventory (e.g., missing point sources, lack of
treatment for possible re-entrainment effects)
• Spatial uncertainty in emission locations due to defaulted locations for point
sources.
• Spatial uncertainty of nearby impacts from elevated point sources (i.e., narrow
plume impact) together with a small number of receptors.
• High coarse particle deposition velocities.
The effects of missing source location data and defaulted emission locations were
explored. This analysis focused on 30 of the 42 monitors which were underestimated by
a factor of 10 or greater. This analysis demonstrated that, for the included monitoring
locations, several nearby lead sources are missing location information (have uncertain
locations). The effects for spatial uncertainty of lead emissions were also examined using
the MAXTOMON technique. Because this analysis did not reveal higher predicted
concentrations within a relatively large region surrounding the monitor, it also suggests
that emission sources are likely missing or underestimated. However, this analysis may
be less definitive for point source pollutants in rural areas. Such monitors tend to
represent small (less than 0.5 km) areas. The surrounding regions also have few census
tracts and the small number of ASPEN receptors limit the opportunity to find other peak
concentrations. Regarding deposition velocities, it was estimated that ASPEN has a bias
to under predict average lead concentrations by 20-30 percent, because of high coarse
particle deposition velocities.
It appears that the current modeling system is underestimating lead for a large percentage
of the monitors used in this evaluation. It should be noted that the monitors do not
represent a random sampling of all census tracts. To attain better modeled results in the
vicinity of isolated point sources, emissions and source locations should be more
accurately characterized. In addition, a denser receptor network may be required.
The results for cadmium and chromium also show many locations with low modeled
concentrations. However, the amount of disagreement between the modeled estimates
and monitored concentrations appear to be dependent on the different source regions
represented. This suggests possible differences in the state inventories; however,
generalizations are difficult because of the limited number of monitoring locations
included.
4.3.4 Discussion of ASPEN Dispersion Modeling Uncertainties
In this discussion we will summarize the studies and conclusions reached regarding
uncertainties in the ASPEN modeling results. The main sources for uncertainty in the
ASPEN dispersion modeling results are 1) emission characterization uncertainties (e.g.,
specification of source location, emission rates and release characteristics); 2)
74
-------
meteorological characterization uncertainties (e.g., representativeness); 3) model
formulation and methodology uncertainties (e.g., characterization of dispersion, plume
rise, deposition, and uncertainties associated with using a "net" of receptors for
characterizing concentration impacts at particular locations); and 4) uncertainty of
background concentrations.
4.3.4.1 Emission Characterization Uncertainties
When performing comparisons between modeled and monitored concentration values, it
is of key importance to properly locate sources relative to monitoring locations and
properly characterize whether the emissions are low-level (i.e., those closer to the ground
such as fugitive emissions) or elevated stack emissions. Inspection of the Gaussian
plume formulations reveals that the simulated impact at the surface generally decreases as
1/(distance)3, where a is on the order of 1 to 2. The low-level emissions affect a receptor
by a factor of 5 to 8 times more than if the emissions were from an elevated stack (as the
elevated emissions will be diluted during transport and dispersion to the surface) [31, 32].
The National Toxics Inventory (NTI) provides the "raw" emissions data that are
processed and made "model-ready" for ASPEN by the Emissions Modeling System for
Hazardous Air Pollutants (EMS-HAP). Annual emission rates are provided for point
sources. Emissions not allocated to point sources are in the form of "non-point" county-
level summaries. The ASPEN model requires higher resolution both temporally and
spatially. Temporally, EMS-HAP allocates the annual total emissions equally into days
and then allocates daily emissions into eight three-hour periods using temporal profiles
based on the type of emission source. The county-wide area and mobile emissions are
allocated by EMS-HAP to a specific census tract centroid within the county, based on
surrogates such as population density, land use, roadway miles, etc., depending on the
source category. For point source emissions, EMS-HAP processing is required whenever
latitude/longitude coordinates or source emission characteristics (e.g., stack parameters)
are missing. When the location coordinates are missing or suspect, EMS-HAP assigns a
location using the zip code's centroid (or if this is missing to a census tract centroid
chosen at random within the source's county).
To assess the extent of point source emissions that were assigned coordinates by EMS-
HAP, the point source inventory for three metals (lead, cadmium, chromium) was
investigated. Table 4-12 shows the percentage of sources and the corresponding
percentage of mass for which point source coordinates were deemed suspect. It would
appear that, although the percentage of sources with suspect locations was approximately
the same (about 15 percent) for all three metals across the United States, the amount of
mass differed greatly, ranging from 7 to 32 percent.
In part to check the accuracy of the stack release heights in the inventory, members of the
EPA project team visited a lead smelter in Herculaneum, Missouri. According to data
files available to the modeling team (Richard Daye, EPA Region 7, personnel
communication), 89.91 tons per year of lead are emitted from the 550-foot stack in the
center of the facility, and 7.7 tons per year are released as "fugitive" emissions, that is,
escaping from the facility through open doors, windows, etc. However, the emissions for
75
-------
an ongoing 2-month study at the facility suggest that the "fugitive" emissions are of order
50 tons rather than 7.7 tons. EPA cannot generalize to all other sources from one site
visit, but this does reveal the possibility of the types of emission characterization
uncertainties that can occur. All other factors being equal, an increase from 7.7 to 50
tons in low-level emissions would likely increase the modeled annual average
concentration at the monitoring site by a factor of 3.
Table 4-12. Source location uncertainties for point source inventory for three metals.
Percent of sources with suspect
coordinates
Percent of mass for which sources
were assigned coordinates
Lead
15.4
16.9
Cadmium
15.8
31.6
Chromium
16.2
6.6
EPA concluded that since surrogates (such as population) are used to allocate county-
wide emissions to specific locations (rather than placing emissions at their true locations),
and since almost 15 percent of the point source inventory have uncertain coordinates,
EPA can anticipate that the ASPEN modeling system will likely not provide good
correlation with observed concentration values at specific locations. If the modeling
results are in accord with observations, it will be for those instances where point source
emissions (location and release height) are well characterized, and where use of a county-
level inventoried source category for the characterization of the other emissions is
appropriate (i.e., the emissions are ubiquitous with respect to the monitoring location).
4.3.4.2 Meteorological Characterization Uncertainties
Meteorological data are a critical input for ISC-based Gaussian air quality models like
ASPEN. EPA analyzed the effects of using 1990 Hourly United States Weather
Observations (HUSWO) and 1996 International Surface Weather Observations (INSWO)
databases on the annual average concentrations predictions. The 1990 and 1996
meteorological inputs differed in three respects: the methods in which the data were
obtained, the number of stations employed, and differences in meteorological conditions
between 1990 and 1996. Because the same emissions input and fixed receptor locations
were used for both model simulations, differences in annual average concentration
estimates between the model simulations are entirely attributable to the differences
between the meteorological inputs for these two years.
The number of surface stations in most states increased in going from the 1990 HUSWO
data set to the 1996 INSWO data set. In the 1990 model simulation, the mean source-to-
meteorological station separation distance was approximately 70 km; this mean
separation distance was reduced to under 50 km in the 1996 simulation by the inclusion
of more meteorological observation locations. The increased density of surface stations
in 1996 is anticipated to result in a better representation of the overall climate at most
emission sources. Due to differences in reported variables, number and location of valid
76
-------
stations, and meteorology itself, statewide annual-averaged concentration estimates were
found to vary from 16.9 percent lower to 84.4 percent higher in 1996, as compared to
1990. Greater variability in modeled concentrations was found when comparisons were
made at the receptor level demonstrating the importance of meteorology in dispersion-
based models.
4.3.4.3 Model Formulation and Methodology Uncertainties
The ASPEN air quality model employs a Gaussian plume model for the characterization
of the transport and dispersion. The particular algorithms employed within the model
were extracted from version 2 of the Industrial Source Complex Long-Term model
(ISCLT2). ISCLT2 has a history of development that dates back to the 1960's, so there is
a history of usage of this model that allows EPA to anticipate the uncertainty in its
estimates (given well-characterized emissions). In general, field study comparisons have
shown that approximately 90 percent of the estimated annual averages are within a factor
of two of those observed [33, 34, 35, 36, 37, 38]. These comparisons were for sulfur
dioxide concentrations at receptors that were generally within 20 km of the sources and
so transformation and deposition effects would be considered negligible.
To verify that the model formulation algorithms were performing as anticipated, several
investigations were conducted. In the first, EPA investigated whether the deposition
algorithms were working properly. In the second, EPA jointly tested whether the
dispersion characterizations were operating as expected, and whether the use of a "net" of
receptors might lead to a bias in the modeling results.
The ASPEN model simulates the effect of dry deposition of particulate by adding an
additional decay term to the emission rate in calculation of ambient concentrations. The
decay term is a function of the deposition velocity, downwind distance from the source,
and plume dimensions with respect to the mixing height. Deposition velocity is also a
function of the particle size, wind speed, and the land-use type. The ASPEN model
allows different deposition options for fine and coarse particulate and urban/rural
environments. In order to analyze the effect of these options on the modeled ambient
concentrations, EPA performed test case simulations using lead emissions from mobile
nonroad sources in Colorado. Different compositions of fine/coarse fractions were used
while the total emission rate was held constant. Five different scenarios were used for
this test case: 10 percent fine and 90 percent coarse, 25 percent fine and 75 percent
coarse, 50 percent fine and 50 percent coarse, 75 percent fine and 25 percent coarse, and
90 percent fine and 10 percent coarse. Emissions from 17 pseudo-point sources of 10m
height, Im/s exit velocity, and T = 295 K were considered. For fine particles, the
ASPEN deposition velocities are generally similar to those estimated by ISC and
scattered around the 1:1 ratio line. The deposition velocities for coarse particles are
much higher for ASPEN than for ISCST3. The effects of these differences were
extrapolated to the national scale by estimating the fraction of lead emissions that were
assigned as coarse particles and as fine particles. For the entire U.S., the total lead
emissions were 66.5 g/s and the percent contribution from different source categories was
as follows: 49 percent of all lead emissions were accounted for by major sources, 28
percent by area sources, less than 0.01 percent by mobile onroad, and 23 percent by
77
-------
mobile nonroad sources. For the ASPEN simulations this means that about 50 percent of
all lead emission sources were treated as point sources and about 50 percent - as pseudo
point sources. By this means, it was estimated that ASPEN would predict average lead
concentrations in the air 20-30 percent lower than would typically be predicted by
ISCST3 because of differences in coarse particle deposition velocities between the two
models.
Gaussian dispersion models are designed to work with inert (non-reactive) pollutants.
Although they can be made to handle linear production or removal effects, nonlinear
chemistry effects are typically not treated. To address this limitation, a simple
mechanism was built to estimate the concentration from those species that could be
formed or destroyed in the atmosphere through secondary chemical reactions (secondary
formation), and within the ASPEN modeling system, this secondary formed
concentration is added to the concentration attributable to primary emissions. Analysis of
ASPEN modeled nationwide mean values for formaldehyde and acrolein suggest that 23
percent and 44 percent (respectively) of the total modeled concentration is attributable to
secondary formation. Results from a recent study using the research version of the
EPA's Ozone Isopleth Plotting Program (OZIPR, a photochemical grid model), suggest
that on a national scale, secondary formation for formaldehyde and acrolein would
account for 90 percent and 85 percent (respectively) of the total modeled concentration
[39]. However, there has not been any direct comparison of ASPEN and OZIPR modeled
concentrations with equivalent emissions and meteorology, accounting for the fact that
ASPEN provides values at specific locations while OZIPR provides grid-averaged
values. Therefore, generalizing current results is problematic. However, these results do
suggest that the ASPEN modeling system for reactive species may be underestimating (or
overestimating) the total concentration, because ASPEN may be underestimating the
component of the concentration that is produced (or destroyed) through nonlinear
chemical reactions. At this point, EPA cautions that concentration estimates for reactive
species should be considered more uncertain than for non-reactive species, all other
factors being equal.
EPA specifically investigated whether the interpolation scheme used within ASPEN
might be underestimating the modeled impacts. This concern arose because a "net" of
receptors is employed by ASPEN, and then concentrations at specific points are
estimated by interpolating within the "net". EPA wondered whether ASPEN might
underestimate peak ambient concentrations because it might "average out" the peak
values by combining them with lower concentrations nearby. To do this, EPA simulated
three types of emissions sources, and compared the ASPEN estimates downwind from
each source to the estimates derived from a more recent dispersion model, the Industrial
Source Complex Long-Term Model Version 3 (ISCLT3). The simulations were run
under a variety of wind speed conditions. The ASPEN estimates are lower than ISCLT3's
by about 10 percent in the near distances, with the underestimation increasing to about 25
percent at 30 km downwind.
4.3.4.4 Uncertainty Due to Background
For all air toxics except for diesel PM, a uniform "background" concentration was
78
-------
applied across all geographic areas to account for long-range transport. (The ASPEN
model does not simulate the transport of any pollutant beyond 50 km from its original
emission point or for emissions from natural sources.) The use of a uniform background
concentration is a simplifying assumption because such extrinsic concentrations vary
geographically for many of these pollutants.
4.3.4.5 Summary of Model Uncertainty Investigations
Past studies of the performance of long-term air quality models (for air toxics at more
localized scales) suggest that 90 percent of the estimated concentrations should be within
a factor of 2 of those observed, if the emissions are well characterized and the
meteorological data are representative. If differences between observed and estimated
concentration values differ greatly from what has historically been seen, one questions
whether the model formulations are in error, whether the meteorological data is
unrepresentative, or whether the emission characterization is in error. The investigations
of the model formulation and methodology characterization uncertainties suggest that the
differences in the dispersion algorithms between standard long-term algorithms and those
within ASPEN are minor, contributing to differences of less than 10 percent in the near
field. Significant differences were seen in the ASPEN deposition velocities for coarse
particles, which were estimated to result in a bias to underestimate lead concentrations by
roughly 30 percent. This bias is of importance only for those toxics simulated as having a
significant fraction of coarse particles, as there was no bias seen in ASPEN deposition
velocities for fine particles. Preliminary investigation of uncertainties associated with
chemical reaction effects suggests that we should consider concentration estimates for
reactive species as more uncertain than for non-reactive species. To improve the
representativeness of the meteorological data, the INSWO surface weather observations
data archive was employed. This reduced the average separation distance between source
locations and observation meteorological sites, and is anticipated to aid in providing more
representative meteorological data. The spot checks on location uncertainties for the
three metals suggest that 6 to 30 percent of these emissions were assigned to default
locations, rather than specified at their true locations. A site visit to one lead smelter
suggests that close inspection of facilities in the vicinity of monitors might reveal other
emissions not reported. These emission characterization uncertainties could have a
greater impact on the model-to-monitor comparison results (resulting in differences of a
factor of 3 or more) than uncertainties seen elsewhere in the modeling algorithms or the
meteorological characterizations (which appear to result in differences of 30 to 80
percent).
4.4 Estimating Population Exposure
In this phase of the national-scale assessment, characterization of population exposure,
ambient air toxics concentrations derived from ASPEN modeling were used to estimate
human exposures to the 33 pollutants included in the ambient concentration analysis.
The exposure assessment was conducted with the HAPEM4 model, which predicts
annual average inhalation exposure estimates. Under the national-scale assessment, the
exposure results serve two purposes. First, the results of the exposure assessment are
directly presented on a pollutant-by-pollutant or geographic area-specific basis, for
comparative purposes. Second, the inhalation exposure estimates serve as input to the
79
-------
risk characterization (see section 4.3).
As applied in the national-scale assessment, the HAPEM4 model predicted a series of
inhalation exposure concentrations for each pollutant, for 40 demographic groups, at over
61,000 census tract locations in the contiguous US, as well as Puerto Rico and the Virgin
Islands. In all, the HAPEM4 model predicts over 74 million exposure estimates as part
of the national-scale assessment. Therefore, to be useful, these results must be
aggregated into the geographic areas and exposure demographic groups of concern. The
aggregation of the 74 million exposure estimates occurs at two levels: at the census-tract
level and across census tracts. At the census-tract level, aggregation is across the
predicted exposures within and across different demographic groups. Across census
tracts, aggregation of results to a larger geographic area (i.e., counties, states) is made to
match the reliability of exposure estimates with that of the emissions inventory and air
quality phases of the assessment.
For presentation and comparison of the exposure assessment results over a broad
geographic area (i.e., counties, states), the study focuses on exposures derived for the
"general population." For the risk characterization, an "age-related" breakdown of the
exposure results is important in defining a lifetime (i.e., 70 years) exposure and because
health responses to pollutants vary with age. Details on the aggregation of the HAPEM4
model results for these two purposes, in the national-scale assessment, are presented
below.
4.4.1 HAPEM4 Census Tract Level Exposure Estimates
The HAPEM4 model is designed to predict inhalation exposure concentrations at the
census-tract level. As applied in the national-scale assessment, the HAPEM4 model
predicted a set of 30 annual exposure concentration estimates (i.e., 30 different activity
pattern scenarios) for each of 40 demographic groups at each census tract, or a total of
1200 exposure concentration estimates per census tract.
It is important to note that, although these results from the HAPEM4 assessment are
derived at the census-tract level, the results are this level are not reliable, due to
recognized limitations in the emissions inventory and dispersion model results.
However, to minimize data aggregation errors, the census tract results serve as the basis
for determining pollutant exposure and risk distributions over larger geographic areas.
In the exposure assessment, certain demographic groups include commuting to other
census tracts. As applied in the national-scale assessment, the exposure concentrations
from these groups (for time spent in the home or work tracts) are recorded in their
"home" census tract. Thus, if a person spends 8-hours in his "work tract" location, the
exposure concentration predicted during this 8-hour period would be included in the
person's "home tract" exposure concentration. In urban settings, the commuting feature
can tend to "spread" the predicted exposure concentrations from central business districts
(which generally have higher ambient concentrations) to the suburban areas (which
generally have lower ambient concentrations).
80
-------
Finally, for census tracts with "zero population," the HAPEM4 model will predict an
exposure estimate of "zero," even though the tract may have a significant ambient
concentration.
Variability Within a Census Tract
It is important to note that the predicted range of exposure estimates for each census tract
(i.e.,1200 exposure concentrations) results from variability in demographic makeup,
variations in activity patterns, and variations in commuting patterns and are derived from
an ambient air quality estimate from a single location (at the census tract centroid as
predicted by the ASPEN model). The modeling exercise did not attempt to quantify the
variability in exposures across individuals within demographic groups. Thus, the
HAPEM4 model, as applied in the national-scale assessment, does not predict a
"complete" distribution of inhalation exposure concentrations across a census tract, only
the distribution that results from "exposure-related" variations. This portion of the
distribution is expected to be small compared to the actual variations in air quality across
the tract. An air quality model applied at a more localized scale would be required to
predict the "more complete" distribution of exposure concentrations across a census tract.
Local scale modeling efforts have shown that ambient concentration gradients can easily
span several orders of magnitude across a small geographic area, such as a census tract.
The distribution resulting from "exposure-related" variations alone is expected to be far
less than an order of magnitude at most locations across the study domain.
Figure 4-14 shows an example of this "exposure- related" variation across an example
urban census tract. In this example, predicted POM exposure concentrations vary by less
than 5 percent between different demographic groups and less than 1 percent within a
specific demographic group. In general the youngest and oldest demographic groups
considered (i.e., 0-5 year olds and 65+ years old) are predicted to have the lowest
exposure levels. These groups are expected to spend more time indoors, where POM
concentration levels are about 30 percent lower than outdoor levels (as defined by the
HAPEM4 model). Thus, EPA can expect the annual average exposure levels for these
groups to be lower than demographic groups which spend more time outdoors. This
relatively small variation in exposure concentrations resulting from "exposure-related"
variations can be found for most of the pollutants and locations included in the
assessment.
Aggregation of Exposure Estimates Within a Census Tract
As noted above, the HAPEM4 model predicts 1200 exposure estimates at each census
tract. These exposure estimates have been aggregated in two different ways. For
presentation and comparison of the exposure assessment results over a broad geographic
area, an assessment of exposure estimates to "the general population" is the most feasible
approach. For the risk characterization, an "age-related" breakdown of the exposure
results is important. These approaches are discussed further below:
Aggregation of Exposures for County, State, and National Level Estimates
Since the goal of the national-scale assessment is to assess exposure and risk levels across
81
-------
a wide population distribution, and the variability of exposure concentrations resulting
from "exposure-related" factors is small, for presentation and area-to-area comparison of
exposure results (i.e., in the tables and charts discussed below), the 1200 exposure
concentrations were characterized by a single "median exposure concentration" at the
census-tract level. This was done by population-weighting the median exposure for each
of the 40 demographic groups. This population-weighted "median exposure
concentration" represents a "best estimate" of the population exposure for a given census
tract. For the national-scale assessment, census-tract level "median exposure
concentrations" were estimated for the contiguous United States, plus Puerto Rico and
the Virgin Islands. This population exposure approach should be carefully considered
when examining and interpreting the results of this national scale assessment. The
national-scale assessment cannot be used to predict local scale "hot spot" exposures or to
capture the range of exposures experienced by individuals, including the maximum
exposed individuals.
Aggregation of Exposures for the Risk Characterization
Although the variation across different demographic groups was small (about 5 percent or
less), for defining carcinogenic risks (see section 4.3), the variation across different age
groups was retained to build a 70-year lifetime exposure period for a "typical" person at
each census tract. This was done by aggregating the predicted median exposures for each
of 5 age groups (0 - 5, 6 - 11, 12 - 17, 18 - 64, and 65+ years). For accessing non-cancer
risks (see section 4.3), population weighted exposures were determined for 2 age groups,
children (<12 years) and adults (12+ years).
4.4.2 National-Scale Assessment Exposure Estimates
As previously noted, the exposure results under the national scale assessment serve two
purposes. First, the results of the exposure assessment are directly presented on a
pollutant-by-pollutant or geographic area specific basis for comparative purposes. Under
this assessment, the median census tract exposure concentrations (as described above)
were used as the basis to derive the statistical distributions at the county, state, and
national levels. Results of these distributions have been prepared in both a tabular and
graphical format and are discussed further below. Second, the inhalation exposure
estimates serve as input to the risk characterization. Results of these exposure estimates
are presented as part of the risk characterization in section 4-3.
Presentation of County, State, and National Exposure Estimates
Aggregated exposure information, at the county, state, and national level is available in
both tabular and graphical formats. The tabular information is summarized on a
pollutant-by-pollutant basis for the entire study domain. Figure 4-15 shows an example
of part of the tabular distribution tables (actual tables contain over 3300 lines).
Graphically, three types of chart presentations are available, "Statewide Exposure
Concentration Estimates," "Percent Contribution," and "Individual State Exposure
Concentration Estimates." The "Statewide Exposure Concentration Estimates" and
"Percent Contribution" charts have been prepared on a pollutant-by-pollutant basis. The
"Individual State Exposure Concentration Estimates" are presented on a state-by-state
pollutant basis. In addition, state and national maps of the exposure estimates on a
82
-------
pollutant-by-pollutant basis have also been prepared. Examples of each chart, and a map,
are presented in Figures 4-16 through 4-19, respectively. Also attached to each exposure
assessment chart/map will be an additional text page (called "page 2") that contains a set
of caveats that the user is encouraged to review. A copy of "page 2," along with a
complete set of charts, is provided in Appendix K. The contents of each presentation
type are described in further detail below.
Exposure Distribution Tables:
The tables contain exposure concentration distributions (i.e., as
percentiles), in addition to a breakdown of exposure concentrations into
major, area and other, onroad mobile, and nonroad mobile sources, as well
as background concentrations for each pollutant at the county, state, and
national levels. Results are also segregated by urban and rural county
designation.
Statewide Exposure Concentration Estimates Charts:
These charts display a horizontal bar for each state, to represent the central
part of the distribution of the exposure concentrations in that state (the
25th to 75th percentile), as well as the median value for the selected
pollutant. If the air pollutant is a carcinogen, the bar chart also displays a
line at the exposure concentrations representing the " 1 in a Million Cancer
Risk", the "10 in a Million Cancer Risk" and the "100 in a Million Cancer
Risk"18. For pollutants with noncancer risks, the bar chart displays a line at
the exposure concentrations representing a " Hazard Quotient = 1.0" (for
reference purposes, some charts may contain a line representing a Hazard
Quotient = 0.1). An example is provided in Figure 4-16.
Percent Contribution Charts:
These charts contain horizontal bars representing the percent contribution
of four source types (major, area and other, onroad mobile, and nonroad
mobile) as well as the contribution from the estimated background to the
statewide average exposure concentration estimate for the selected
pollutant (see Figure 4-17).
Individual State Exposure Concentration Estimates Charts:
These charts depict the median exposure concentrations for each pollutant
in the NATA analysis on a single chart. A chart is provided for each state.
Pollutant exposure concentration symbols are plotted to give the
relationship of the state's median exposure concentrations to the cancer or
noncancer risk level, where applicable. An example is provided in Figure
4-18.
18 Concentrations at various levels of cancer risk and non-cancer hazard were determined using procedures
described in section 3.3 and assume a continuous lifetime exposure to the concentrations predicted by this
assessment. It is important to note that none of these risk-based concentrations represent regulatory
standards.
83
-------
National and State Maps of Modeled Exposure Concentrations:
These maps illustrate the modeled human exposure concentration to air
toxics by county in 1996. Map colors indicate categories of exposure risk /
hazard, and the corresponding ranges of inhalation exposure
concentration. The exposure concentration value displayed in maps is the
county median. Pollutant exposure concentration is expressed in
micrograms per cubic meter of air (i g/m3). Figure 4-19 provides an
example of a state map with estimated exposure concentrations.
Interpretation of County, State, and National Exposure Estimates
It is important to note that the exposure distributions presented in this part of the national
scale assessment represent the range of census tract median exposure concentrations
across a geographic area (i.e., county, state, and national levels). As previously noted,
the available information about the range of exposures across individuals or, spatially,
within census tracts, is not reliable. This information is best suited to estimate "trends"
in exposure or general population level exposures. The results of the exposure
assessment presented at this level are best used to define broad geographic areas and
pollutants of initial concern. A relatively low predicted exposure value does not
necessarily mean that a particular area or pollutant is exempt from health-related air toxic
issues. In certain situations, a more refined or local scale assessment may be needed to
assess whether there is an air toxics problem. In areas with a relatively high predicted
exposure value, the reliability of the estimate, and the data used to derive that estimate
must be further examined before concluding that concern is warranted.
The results of the exposure assessment are only meaningful when examined at the
individual county level or above. Comparison across geographic areas can be made by
examining the graphical and tabular exposure results. However, it is important to note
that the confidence in the predictions may be better in some geographic areas than others
or may be better for certain pollutants. The variability in the quality of the predictions
can vary from not only state-to-state but county-to-county within the same state. Thus,
while the presentation materials contain multiple geographic areas (i.e., states compared
on a single chart), the data are presented in this fashion for convenience and to minimize
the number of charts/graphs that is required to summarize the assessment.
For reference, cancer and noncancer health criteria lines have been included on the
graphical presentations. These have been placed only to compare the presented exposure
estimates in a general context to the health criteria. It is strongly advised that the risk
characterization section, as well as the assessment's limitations, be reviewed before
assessing a particular health concern in a specific geographic area. A discussion of
exposure results on a pollutant-by-pollutant basis is made in the risk characterization
results section.
Care should be taken when trying to interpret the geographic distribution of exposure
concentrations across a county. Figure 4-20 presents the range of exposure
concentrations across tracts within a county. This shows that, for benzene exposure
estimates, the range of exposures within a county (the tract-to-tract variations) appears to
84
-------
be directly related to the population of that county. Thus, at least for benzene, the most
populated counties are predicted to have the largest range in exposure concentrations, and
thus the greatest uncertainty in the geographic distribution of these estimates. It is not
clear whether this is a true relationship or an artifact of the assessment, in that a denser
modeling (both ambient and exposure) grid is utilized in more populated areas (more
census tracts).
4.4.3 Comparison of HAPEM4 Exposure Concentrations to ASPEN
Ambient Concentrations
To illustrate the contribution of the HAPEM4 model to the assessment, census tract
median exposure concentrations were compared to corresponding census tract median
ambient concentrations as predicted by ASPEN. Table 4-13 shows the average ratio of
HAPEM4 to ASPEN for each of the study pollutants. This ratio is presented for total
levels as well as for each of the five source sectors considered in the assessment (i.e.,
major, area and other, onroad mobile, nonroad mobile, and background). In general the
HAPEM4 exposure predictions are 5-40 percent lower than the corresponding predicted
air quality values. This reduction likely results from the inability of many pollutants to
penetrate efficiently into an indoor environment (see section 3.2.3). Exposures resulting
from emissions originating from major, area, nonroad mobile, and background source
sectors are seen to be about 20-30 percent lower than the corresponding predicted air
quality values for most gaseous pollutants, with some metals approaching a 40 percent
reduction from air quality levels. Exposures resulting from emissions from onroad
mobile sources range from 40 percent lower than the corresponding predicted air quality
values for chromium to 24 percent greater than the predicted air quality values for
benzene. Benzene exposure levels for this source sector are most likely higher than their
corresponding ambient levels because of the relatively high "proximity term" (see section
3.2.3) assigned to the "in-vehicle" microenvironment. This "proximity term" is required
to adjust the ASPEN predicted ambient level, which is assumed representative of the
census tract centroid, to that which EPA would expect immediately outside of the
microenvironment. For most microenvironments this term was set to unity. However,
for the transportation related microenvironments it is presumed that the ambient
concentrations immediately outside the vehicle (i.e., very close to the pollutant source)
are considerably higher than the census tract centroid value predicted by ASPEN. Thus
for pollutants where emissions are significantly dominated by onroad mobile sources an
appropriate proximity term was developed and applied specific for this assessment. For
example, for benzene, a proximity term of 6.6 was applied to the ASPEN census tract
centroid levels to predict the expected ambient concentration immediately outside a
vehicle.
85
-------
Table 4-13. HAPEM4 to ASPEN Average Ratio1.
Pollutant
Acetaldehyde
Acrolein
Acrylonitrile
Arsenic
Benzene
Beryllium
1,3 -Butadiene
Cadmium
Carbon Tetrachloride
Chloroform
Chromium
Coke Oven Emissions
1 , 3 -Dichloropropene
Diesel Paniculate Matter
Ethylene Dibromide
Ethylene Bichloride
Ethylene Oxide
Formaldehyde
Hexachlorobenzene
Hydrazine
Lead
Manganese
Mercury
Methylene Chloride
Nickel
Poly chlorinated biphenyls (PCBs)
Perchloroethylene
Polycyclic Organic Matter (7 -PAH)
Polycyclic Organic Matter (Total)
Propylene Bichloride
Quinoline
1, 1,2,2-Tetrachloroethane
Trichloroethylene
Vinyl Chloride
Source Sector
Major
0.79
0.82
0.75
0.79
0.85
0.79
0.83
0.79
0.71
0.82
0.64
0.75
0.82
0.79
0.87
0.79
0.69
0.81
0.83
0.84
0.74
0.79
0.80
0.79
0.79
0.75
0.75
0.76
0.79
0.75
0.79
0.84
0.75
Area and
Other
0.79
0.80
0.75
0.79
0.85
0.79
0.83
0.79
0.71
0.82
0.63
0.75
0.83
0.79
0.87
0.79
0.69
0.81
0.83
0.84
0.73
0.79
0.81
0.79
0.79
0.76
0.75
0.76
0.79
0.75
0.79
0.84
0.75
Onroad
Mobile
0.93
1.06
0.80
1.24
0.93
0.66
0.72
0.89
0.84
0.75
0.80
0.80
0.76
0.76
Nonroad
Mobile
0.80
0.82
0.79
0.85
0.79
0.84
0.79
0.66
0.70
0.72
0.84
0.75
0.80
0.79
0.78
0.77
Estimated
Background
0.85
0.72
0.82
0.79
0.87
0.69
0.81
0.79
0.80
0.79
0.75
0.84
Total
0.87
0.92
0.75
0.79
0.98
0.79
0.90
0.79
0.72
0.82
0.64
0.75
0.83
0.71
0.79
0.87
0.79
0.75
0.81
0.83
0.84
0.73
0.79
0.80
0.79
0.79
0.75
0.75
0.76
0.79
0.75
0.79
0.84
0.75
1 Average ratio developed by comparing (HAPEM4 Concentration)/(ASPEN Concentration) at each census tract
86
-------
4.4.4 Discussion of HAPEM4 Limitations and Uncertainties
Like any model, especially when applied on a very broad geographical scale, HAPEM4
relies on a number of assumptions and approximations in estimating inhalation
exposures. In general, the accuracy of the results is primarily limited by:
1. The emissions inventory and dispersion modeling results;
2. Reliability of ME factors for various pollutants;
3. Activity pattern information (e.g., patterns of time spent in various
microenvironments for the populations in the geographic areas modeled); and
4. The ability of the chosen population cohorts to adequately represent the true
demographics of every census tract.
More specifically, limitations and uncertainties include:
1. Uncertainty with the emissions inventory and the ASPEN dispersion modeling,
which was discussed in the previous sections.
2. Indoor emission sources - The exposure estimates do not include exposures
related to emissions of air toxics from indoor sources (e.g., off-gassing from
building or consumer products, smoking, internal combustion sources, etc.).
Indoor sources of some air toxics are likely to be important in assessing total
inhalation exposures, and will be addressed in the future as additional data are
obtained and new analyses are conducted.
3. Spatial resolution - Because of the spatial uncertainties associated with air toxics
emission inventory data and the associated ASPEN model ambient concentration
estimates, results of the HAPEM4 inhalation exposure model are compiled and
presented at county level or larger spatial scales.
4. Exposure routes - The HAPEM4 model only estimates inhalation exposures, since
it is based on ambient air concentration data as input. Thus, it does not address
exposures through other pathways (e.g., ingestion). This is especially important
for toxic pollutants that are persistent and bioaccumulate, such as mercury,
dioxins, and PCBs. Emissions of these pollutants disperse through the
atmosphere and eventually deposit to land or water bodies. Once deposited, they
can bioaccumulate up the food chain. For example, mercury bioaccumulates most
efficiently in the aquatic food chain (e.g., fish). Fish consumption is the primary
route of exposure to mercury [40]. For dioxins, (which bioaccumulate mainly in
animal fat tissue) population exposures are primarily due to ingestion of dairy
products, fish, beef, pork, and poultry [41]. Multimedia exposure models are
needed to address such multipathway exposures. These will be included in future
national-scale assessments as appropriate tools and data become available.
87
-------
5. Population exposure estimates - The exposure estimates represent midrange
estimates of population exposures. Due to a number of factors, some individuals
may have substantially higher or lower exposures. It is important to note that the
exposure model, as applied on the national scale, is not designed to quantify these
extreme values of individual exposures.
6. Representativeness of the ME factors - Exposure models must predict the
relationship between the ambient air quality (outside) and that in a
microenvironment (inside or outside). When applied on a local scale, exposure
models can employ detailed mass balance equations to predict this relationship.
However, on a national scale, the development of such a detailed relationship is
not feasible. Thus, the HAPEM4 model, applied on a national scale, relies on
generalized ME factors, which do not account for variability (e.g., penetration
affected by air exchange rate, which is a function of ambient temperature,
heating/cooling system, and open windows). These factors can be represented as
simple first-order relationships between the outdoor and the indoor air quality.
For some pollutants and microenvironments this relationship is well-documented
and well-understood. However, for many pollutants and microenvironments this
relationship is either not clear or has not been measured. As part of this
assessment, EPA conducted a detailed study of exposure literature to develop ME
factors for each air toxic in the study [19]. This ME study has undergone a
separate technical peer review. The technical reviewers agreed that there is a
great degree of uncertainty associated with these numbers. They have suggested
that this uncertainty be assessed either qualitatively or quantitatively as part of
this study. In future versions of this assessment EPA plans to define these ME
factors as distributions rather than fixed "best estimate" values as was employed
for this initial assessment.
7. Representativeness of the population cohorts - The assessment assumed that the
40 cohort groups selected can represent the activity patterns of the general
population in all areas of the country. The groups were selected to represent
variability in population activity patterns while at the same time maintaining the
ability to present the exposure assessment results in a manner that will allow for
an adequate lifetime exposure aggregation. It is also possible that members of the
same cohort may have significantly different activity patterns in different census
tracts. The usage of the 40 cohort groups reflects the finest resolution that EPA
believes is possible with currently-available models.
8. Representativeness of the activity pattern sequence - When selecting multiple 24-
hour activity patterns to construct an annual average pattern, patterns are
combined that pertain to different individuals, so that day-to-day correlations in
activities are not preserved. For example, for day 1 the pattern may specify a
house with an attached garage, and for day 2 a house without an attached garage.
In this situation, the HAPEM4 model would underestimate the annual average
exposure for a person residing in a house with an attached garage, and
-------
overestimate the exposure of the person in the house without an attached garage,
and overestimate the exposure of the person in the house without an attached
garage. As a result the aggregated activity pattern is more representative of a
population average pattern for the demographic group, than any individual
pattern. Thus, the distribution of exposure concentrations for the group estimated
by HAPEM4 represents the uncertainty in the population average exposure
concentration, rather than the variability of the individual exposure concentrations
among members of the group. Uncertainty and variability of input data other than
activity data are not considered, so that the resulting uncertainty information
provided by the prediction distributions is an underestimate of the overall
uncertainty.
89
-------
5 Risk Characterization
5.1 Introduction
As described in Section 2, this assessment was based on EPA's paradigm for risk
assessment, a framework to assess and manage risks developed by the National Academy
of Sciences in 1983. The paradigm divides the risk assessment and risk management
process into different general phases (as shown in Figure 2-1). The phases that comprise
risk assessment are (1) exposure assessment (which describes how humans come into
contact with pollutants), (2) dose-response assessment (which describes the adverse
health effects the pollutants may produce and at what doses these effects may occur), and
(3) risk characterization (which combines the exposure and dose-response information to
draw qualitative or quantitative inferences about risk). The risk characterization section
presents an interpretation and discussion of the results of the NATA national-scale
assessment, including the uncertainties associated with each element, and makes specific
recommendations for future directions of air toxics assessments on the basis of this
information. Complete results for the risk characterization portion of the assessment are
provided in Appendix L.
This risk characterization examines inhalation risks for 32 hazardous air pollutants
associated with 1996 emissions in urban and rural areas nationwide. This assessment was
conducted on a coarse spatial resolution (i.e., census tracts) and used many simplifying
assumptions. Therefore, it should be noted that individuals within a census tract may
have substantially higher or lower exposures (and concomitant risks) than estimated here.
The characterization is intended to help identify pollutants of greatest potential concern
(among these 32), prioritize efforts to reduce emissions, provide a baseline for measuring
future trends, and to help set research priorities. EPA plans to update this assessment
every three years. The next assessment will focus on 1999 emissions, concentrations and
risks.
As described in Section 3, several risk presentation formats have been used. Cancer risks
to individuals are presented as lifetime (e.g., 70-year) individual risks. These risks are
expressed in terms of the estimated "upper-bound" (i.e., likely actually to be lower, but
may be higher) probability that a person with the median exposure in a census tract will
contract cancer. Non-cancer hazard to individuals is expressed in terms of the hazard
quotient, defined as the ratio between the estimated median exposure in an individual's
census tract and the reference concentration (or similar value). The reference
concentration is an exposure that is believed to be without significant risk of adverse non-
cancer health effects in a chronically-exposed population, including sensitive individuals.
Cancer risks to individuals exposed to multiple pollutants were combined by summing,
with known human carcinogens kept separate from probable and possible carcinogens.
Non-cancer hazards for multiple pollutants were combined by summing hazard quotients
for pollutants affecting the same target organs to create a target organ-specific hazard
index (TOSHI) for non-carcinogenic effects. Non-carcinogenic pollutants having "high
certainty" RfCs were summed separately from those with "low certainty" RfCs. Due to
90
-------
the broad scale of the assessment, the risk characterization focused on results at the
national level, which is the level at which EPA believes the results are most meaningful.
Estimated census tract median cancer risks and non-cancer hazards to individuals have
been shown as box plots. The plots include a range extending from the 5* to the 99
percentile census tracts. Cancer risks and non-cancer hazards to populations have been
expressed as total numbers of people (i.e., not truncated at the 5th and 99th percentiles)
who reside in census tracts where the estimated median risk (or hazard quotient) exceeds
fixed levels within the contiguous US, Puerto Rico, and the Virgin Islands.
For a complete understanding of the risk characterization, it is important to remember
that the scope and methods of this assessment have imposed the following limitations and
uncertainties:
1. Risk and hazard quotient levels are not regulatory levels. The determination of
what is an acceptable or unacceptable risk depends on additional factors and more
refined information.
2. The risk estimates presented are based on the assumption that pollutant exposures
would remain at 1996 levels over a lifetime. They did not take into account
significant reductions that have taken effect since 1996, or future reductions
expected from: 1) mobile sources; 2) major industrial sources; 3) State or industry
initiatives; and 4) facility closures. For example, EPA expects exposures to some
gaseous air toxics from onroad mobile sources to be reduced about 50% by 2007
and 60% by 2020. While such a "snapshot" type of assessment may not be able
to provide direct estimates of absolute risk levels, it nonetheless provides a useful
yardstick against which one may evaluate the risk-reduction potential of
hypothetical risk reduction scenarios, and which may be used to assess the
relative contributions of various pollutants and source groupings to national-scale
risks.
3. The risk estimates for chromium, nickel, and polycyclic organic matter were
based on conservative assumptions of speciation that were applied uniformly to
all areas. Actual risks associated with these pollutants may be lower than
estimated in some areas.
4. The risk estimates represent risks associated with midrange estimates of
population exposures within each census tract. Some individuals may have had
substantially higher or lower exposures and risks. It is important to understand
that the dispersion and exposure models used by the assessment were not
designed to quantify inter-individual variability in exposures.
5. All cancer risk estimates should be considered as conservative, but not worst-
case. Because they represent a composite of upper bound UREs and median
population exposures (which may be underestimated), the true risks would
91
-------
probably be less, but could be greater. For more highly exposed segments of the
population, these cancer risk estimates would be correspondingly higher.
6. EPA is currently reassessing the carcinogenic effects of 19 of the air pollutants
included in this study. Cancer unit risk estimates could change substantially as a
result of these reassessments.
7. For pollutants that have more than one unit risk estimate (e.g., benzene and vinyl
chloride), this characterization uses the highest available unit risk. For more
details about these unit risks, see Appendix G.
8. This characterization presents non-cancer hazard in terms of the hazard quotient,
which is the ratio of a given exposure level to the reference concentration (RfC)
or similar value for a pollutant. The RfC is an estimate of the continuous lifetime
inhalation exposure that the EPA believes is likely to have no appreciable risk of
deleterious non-cancer effects. Although hazard quotients below 1.0 (i.e.,
exposures below the RfC) are believed safe, hazard quotients above 1.0 are not
necessarily harmful. Nevertheless, as the hazard quotient increases above 1,
potential for adverse effects also increases.
9. EPA combined non-cancer hazard estimates for multiple pollutants using the
target-organ-specific hazard index (TOSHI), defined as the sum of hazard
quotients for individual toxic air pollutants that affect the same organ or organ
system. This method is a simplified approximation of the potential aggregate
effect because different substances may affect the same organ in different and
non-additive ways. As with the hazard quotient, aggregate exposures below a
hazard index of 1.0 will likely not result in adverse non- cancer health effects over
a lifetime of exposure. However, a hazard index greater than 1.0 does not
necessarily suggest a likelihood of adverse effects.
10. Estimates of the number of people at various cancer risk or non-cancer hazard
levels were based on census tract population estimates from 1990, the most recent
available. The total population of the contiguous US, Puerto Rico, and the Virgin
Islands in 1990 was 251 million, including 207 million adults and 44 million
children.
11. Exposure pathways other than inhalation, as well as indoor sources of air toxics,
may contribute substantially to total risks for some pollutants. This assessment
does not address oral or dermal exposures, or inhalation exposure resulting from
indoor sources.
12. The simplifying assumptions necessary for national-scale modeling have
introduced significant uncertainties into each component of the assessment.
Because of these uncertainties, EPA will not use the results of this assessment to
determine source-specific contributions or to set regulatory requirements.
92
-------
These uncertainties are discussed in greater detail in Section 5.4. This section includes
qualitative descriptions of all major sources of uncertainty and variability in the national-
scale assessment, provides a simple illustration of one method by which the magnitude of
variability and uncertainty can be estimated, and provides recommendations for future
efforts for a more complete quantification of variability and uncertainty in the future.
5.2 Cancer Risks
5.2.1 Pollutant-Specific Cancer Risks to Individuals
Figure 5-1 shows the distribution of estimated cancer risks for each of the pollutants
nationwide, from the 5th to the 99th percentile. Risks were based on the median exposure
within each of approximately 61,000 census tracts nationwide. Because census tracts are
intended to include more or less similar population sizes, the distribution of tract median
risks should be generally representative of the distribution of risks for "typical"
individuals in the US. It should be noted that this distribution does not represent risks to
all individuals in the US. Some individuals within a census tract may have higher or
lower exposures (and concomitant risks) that those shown. (For example, the 5*
percentile risk level on the figure indicates that approximately 5% of the population lives
in census tracts where the median risks are at that level or lower. The 99* percentile risk
level indicates that 99% of the population lives in census tracts where the median risks
are at that level or lower.)
The extremes at either end of the distributions (i.e., less than the 5th and greater than the
99th percentiles) were not shown because EPA believes that, given the broad scope of this
assessment, these risk estimates were less reliable than the information shown on the
graphs. That does not mean these risks are unimportant, however. For example,
approximately 2.5 million people resided in census tracts where the median estimated
risks were higher than the 99th percentile risk shown. Because this assessment was
designed to evaluate average exposure and risk at the national scale, more refined local-
scale assessments will be needed to adequately characterize exposures and risks at the
upper end of the national range.
In general, narrow distributions on Figure 5-1 (e.g., for PCBs or carbon tetrachloride)
suggest that background sources were dominant. Risks associated with these
background-dominated pollutants were similar in all tracts. Broad risk distributions (e.g.,
for coke oven emissions) suggest dominance by major sources that can strongly affect
limited areas. Risks associated with these pollutants varied with location by many orders
of magnitude. Distributions intermediate between narrow and broad (e.g., for benzene or
perchloroethylene) suggest dominance by mobile or area (and other) sources whose
impacts were more widespread than major sources but less widespread than background.
Detailed risk distributions for major, area, mobile, and background sources are shown in
Appendix L (Figures 1-5).
Based on an examination of Figure 5-1, pollutants can be grouped into scale-related
categories as follows:
93
-------
National risk drivers: Figure 5-1 shows that, for at least 50% of the US population,
estimated inhalation cancer risks associated with three pollutants - benzene, carbon
tetrachloride, and formaldehyde - approached or exceeded 10 in 1 million. Benzene and
formaldehyde risks were associated primarily with mobile and background sources,
whereas carbon tetrachloride risks were virtually all from background.
Regional risk drivers: Four more pollutants - chromium, ethylene oxide, polycyclic
organic matter, and coke oven emissions - showed estimated cancer risks exceeding 10 in
1 million for more than 1% of the US population. The first three of these pollutants
originated primarily from area (and other) sources; the fourth was associated entirely with
major sources.
Important national contributors: Estimated risks to 50% or more of the US population
for five more pollutants- acetaldehyde, 1,3-butadiene, ethylene dibromide, ethylene
dichloride, and perchloroethylene - were near or above 1 in 1 million. Mobile sources
were most important for emissions of acetaldehyde and 1,3-butadiene, background
sources for ethylene dibromide and ethylene dichloride, and area sources for
perchloroethylene.
Important regional contributors:
Nine more pollutants - arsenic,
nickel, acrylonitrile, cadmium,
chloroform, 1,3-dichloropropene,
hydrazine, quinoline, and
trichloroethylene - showed estimated
cancer risks exceeding 1 in 1 million
for some of the US population. Major
sources contributed important
amounts of hydrazine and quinoline,
although area (and other) sources
were also important for these
pollutants. Risks for the remaining
seven pollutants were dominated by
area (and other) sources. (Note that
one pollutant that met this condition, 7-
is redundant with total POM. Because
focuses on it.)
Cancer risk to individuals:
• National risk drivers: benzene, carbon
tetrachloride, formaldehyde
• Regional risk drivers: chromium, ethylene
oxide, polycyclic organic matter, coke oven
emissions
• Important national contributors:
acetaldehyde, 1,3-butadiene, ethylene
dibromide, ethylene dichloride,
perchloroethylene
• Important regional contributors: arsenic,
nickel, acrylonitrile, cadmium, chloroform,
1,3-dichloropropene, hydrazine, quinoline,
trichloroethylene
-PAH, was omitted from this discussion because it
total POM is a more inclusive group, discussion
For the remaining carcinogenic pollutants in the national-scale assessment - vinyl
chloride, beryllium, hexachlorobenzene, lead, methylene chloride, PCBs, propylene
dichloride, and 1,1,2,2-tetrachloroethane - 99% of the US population lived in census
tracts where the median cancer risk did not exceed 1 in 1 million. It is important to note,
however, that this result does not rule out potential concerns for these pollutants. First,
this national-scale assessment lacks the resolution to evaluate all local inhalation risk hot
spots. Therefore, there may be locations where these substances pose significant risks
due to concentrations in air, but this assessment lacked the resolution to detect them.
94
-------
Additionally, there is the potential for the omission of important sources in the inventory,
especially since this is the first time such a comprehensive inventory has been attempted.
Furthermore, beryllium, hexachlorobenzene, lead, and PCBs all tend to accumulate in
soil and are persistent in the environment. Because this assessment was limited to
inhalation exposure, these substances may pose important risks via the ingestion pathway
that were beyond the scope of this assessment.
In general, the most important national-scale "risk drivers" originated predominantly
from mobile (i.e., benzene, formaldehyde) or background (i.e., carbon tetrachloride,
formaldehyde) sources. Most of the "second-tier" pollutants originated mostly from area
or background sources. Pollutants associated with major sources were generally not seen
to be as important nationally, but were very important in some regions. Given the
resolution of this national-scale assessment, this result is not very surprising; the
assessment tools are simply inadequate to quantify localized impacts around isolated
stationary sources. More rigorous localized assessments will be needed to quantify risks
associated with such individual sources; this is one of the goals of the assessments being
carried out as a part of EPA's residual risk assessment efforts. Finally, it is important to
realize that "background" in this assessment was really a miscellaneous category,
attempting to capture sources that were not inventoried or sources beyond 50 km that
were inventoried. The category included natural background emissions, historic
emissions of persistent substances, international transport, and interstate transport. The
first three of these subcategories are not controllable under the Clean Air Act, but
elements of the fourth - interstate transport - are. Thus, some part of the background
sources component of risk for each substance may be addressable by the air toxics
program. This addressable portion likely varies among different substances, and the
current state of EPA's dispersion models and databases is insufficient for quantitative
estimation.
5.2.2 Pollutant-Specific Cancer Risks to Populations
Figure 5-2 shows numbers of people residing in census tracts for which the median
exposure corresponds with estimated lifetime cancer risks exceeding three fixed levels
(100, 10, and 1 in one million) for emissions from all sources combined. The same
information for each source sector (major, area, mobile, and background sources) is
provided in Appendix L (Figures 13-17). Unlike the individual risk figures (which
exclude the extreme ends of the distribution), these population figures include the entire
assessed population in order to increase sensitivity to those potential urban-scale hot
spots that affect the median exposure estimate at the census-tract level.
Based on an examination of Figure 5-2, pollutants can be grouped into scale-related
categories as follows:
National risk drivers: Figure 5-2 shows that three pollutants - benzene, formaldehyde,
and chromium - presented estimated risks of 10 in 1 million or higher to more than 10
million people. The first two pollutants posed such risks for more than 100 million
people. Contributions of individual source sectors (shown in Appendix L, Figures 13-17)
suggest that benzene originated predominantly from a mixture of onroad and nonroad
95
-------
mobile sources, but formaldehyde was emitted predominantly by nonroad mobile
sources. Both of these pollutants had an important background component (i.e., greater
than 1 in 1 million risk to the entire population) as well. In contrast, chromium emissions
originated predominantly from a mixture of major and area sources.
Regional risk drivers: Coke oven emissions posed estimated risks of 100 in 1 million or
higher to more than 100,000 people, and acrylonitrile, arsenic, and hydrazine presented
such risks to more than 10,000
people. Arsenic is emitted by both
major and mobile sources; the other
three pollutants are emitted
predominantly or exclusively by
Cancer risk to populations:
• National risk drivers: benzene,
formaldehyde, chromium
major sources. These substances * Regional risk drivers: arsenic, coke oven
may be good candidates for more emissions, acrylomtnle, hydrazme
r- i 7 , , i • • Important national contnbutors:
refined, local-scale analyses in areas —* ,, , , r—~—TT~- , ., ,
, . . J acetaldenyde, carbon tetracnloride, etnylene
where emissions sources exist. dibromide, ethylene dichlonde, polycyclic
organic matter, perchloroethylene
Important regional contributors: 1,3-
butadiene, cadmium, 1,3-dichloropropene,
ethylene oxide
Important national contributors:
Six more pollutants - acetaldehyde,
carbon tetrachloride, ethylene
dibromide, ethylene dichloride,
polycyclic organic matter, and perchloroethylene - presented risks of 1 in 1 million or
higher to more than 100 million people. Mobile sources were most important for
emissions of acetaldehyde, background sources for carbon tetrachloride, ethylene
dibromide, and ethylene dichloride, and area (and other) sources for polycyclic organic
matter and perchloroethylene.
Important regional contributors: Four more pollutants -1,3-butadiene, cadmium, 1,3-
dichloropropene, and ethylene oxide - posed estimated cancer risks of 1 in 1 million or
higher to more than 10 million people. Cadmium, 1,3-dichloropropene and ethylene
oxide originated mostly from area sources, and 1,3-butadiene from mobile sources
(mostly onroad).
The remaining carcinogenic pollutants in the national-scale assessment - nickel, vinyl
chloride, beryllium, chloroform, hexachlorobenzene, lead, methylene chloride, PCBs,
propylene dichloride, quinoline, 1,1,2,2-tetrachloroethane, and trichloroethylene - were
estimated to exceed risks of 1 in 1 million for fewer than 10 million people and to exceed
risks of 10 in 1 million for fewer than 1 million people. As noted in the discussion of
individual risk, this result does not exonerate these pollutants. In fact, Figure 5-2 shows
that beryllium, chloroform, and quinoline contribute estimated inhalation risks exceeding
10 in 1 million for more than 100,000 people, making them potentially important urban-
scale concerns. Also, as noted previously, potentially important ingestion exposures for
beryllium, hexachlorobenzene, lead, and PCBs were not included in the assessment.
96
-------
5.2.3 Aggregate Cancer Risks of Multiple Pollutants to Individuals
Figure 5-3 shows the distribution of estimated cancer risks for multiple pollutants
combined, from the 5* to the 99* percentile in the US population. The risks are based on
median exposures within each census tract nationwide. Known human carcinogens were
aggregated separately from probable and possible carcinogens, and the figure shows risks
by individual source sector, and for all source sectors combined.
The median aggregate risk estimate for all known carcinogens and source sectors
combined was 10 in 1 million and the 99th percentile risk was 60 in 1 million. The
median aggregate risk estimate for all probable and possible carcinogens and source
sectors combined was 20 in 1 million and the 99* percentile risk was 100 in 1 million.
While EPA has not combined risk estimates for known and probable/possible
carcinogens because of their different magnitudes of relative uncertainty, it appears
reasonable to assume that total risk associated will all carcinogens in this assessment
could approach 30 in 1 million risk to the median US receptor.
Background sources alone were estimated to provide an aggregate risk of approximately
3 per million for known carcinogens and 4 per million for probable/possible carcinogens.
Because background exposures were assumed to be ubiquitous, they effectively
established the "floor" for the distributions of total risk. Major sources contributed little
to the central part of the total risk distributions, but made dominant contributions to the
upper end of the distribution for known carcinogens. Mobile sources were the most
important contributors to total risk for known carcinogens; area and mobile sources were
of roughly equal importance in determining total risk for probable and possible
carcinogens.
5.2.4 Aggregate Cancer Risks of Multiple Pollutants to Populations
Figure 5-4 shows numbers of people residing in census tracts for which the median
exposures to all carcinogens combined corresponded with estimated lifetime cancer risks
exceeding three fixed levels - 100, 10, and 1 in one million. As before, known
carcinogens were aggregated separately from probable and possible carcinogens. The
figure shows affected populations by source sector, and for all source sectors combined.
Based on Figure 5-4, approximately 130 million people resided in census tracts where
cancer risk exceeded 10 in 1 million, and 800,000 lived in tracts where risk exceeded 100
in 1 million from known carcinogens. For probable and possible carcinogens,
approximately 200 million people lived in tracts where risks exceeded 10 in 1 million and
2 million lived in tracts where risks were above 100 in 1 million.
As already noted, the entire assessed population experienced estimated risks exceeding 1
in 1 million from background sources alone. For known carcinogens, major sources were
responsible for most risks exceeding 100 in 1 million. For probable/possible carcinogens,
risks above 100 in 1 million were associated more or less equally with major, area, and
nonroad mobile sources. The greatest contributors to "intermediate risks" (i.e., between
10 and 100 in a million) were onroad and nonroad mobile sources.
97
-------
5.3 Non-Cancer Hazard
5.3.1 Pollutant-Specific Hazard Quotient for Individuals
Figure 5-5 shows the distribution of non-cancer hazard quotient (HQ) for 27 pollutants
nationwide, from the 5* to the 99* percentile. As with the cancer risk figures, HQs were
based on the median exposure within each census tract nationwide and are therefore
likely to be similar to the distribution of risks for individuals in the US. The extremes at
either end of the distributions were truncated at the 5* and 99* percentiles, but readers
should keep in mind that approximately 2.5 million people reside in census tracts where
the median HQs are higher than the upper end of the distribution on the graphs.
In general, a very narrow distribution on Figure 5-5 (e.g., chloroform, ethylene
dibromide) suggests that background sources were dominant. HQs associated with these
pollutants were similar in all tracts. A very broad risk distribution (e.g., lead) suggests
dominance by major sources or by a few large area sources. Hazards associated with
these pollutants varied with location by several orders of magnitude. Distributions in
between narrow and broad (e.g., benzene, manganese) suggest that dominant sources
were either mobile or area sources that impacted more areas than major sources but were
less ubiquitous than background. Detailed risk distributions for major, area, mobile, and
background sources are shown in Appendix L (Figures 7-11).
National hazard drivers: Figure 5-5 shows that for at least 50% of the US population, the
inhalation HQ associated with a single pollutant - acrolein - was approximately 4. The
HQ for the most exposed 1% of the population was approximately 20. No other
pollutants approached within an order of magnitude of acrolein's HQ distribution. Most
of the acrolein exposure was associated with mobile sources, with area (and other)
sources also contributing.
It is important to note that many reference concentrations incorporate protective
assumptions in the face of uncertain data. For this reason, an HQ greater than 1.0 does
not necessarily suggest a likelihood of adverse effects, whereas an HQ less than 1.0 does
suggest that adverse effects are unlikely. Furthermore, the HQ cannot be translated to a
probability that adverse effects will occur, and is not likely to be proportional to risk. An
HQ greater than one can be best described as indicating that a potential may exist for
adverse health effects, warranting further investigation.
Three more pollutants - formaldehyde, acetaldehyde, and manganese - showed HQs
exceeding 0.1 for some of the US population. While an HQ of 0.1 does not suggest a
present potential for adverse health effects, the fact that these HQs are within an order of
magnitude of 1.0 suggests that some potential may exist for adverse health effects if
emissions increase in the future. Also, these substances could potentially combine with
other pollutants to present an aggregate threat, or pose local threats. The remaining
pollutants in the national-scale assessment were found not to contribute HQs exceeding
0.1 for 99% of the US population. As with carcinogens, this result does not exonerate
these pollutants because of the coarse resolution of the assessment, potential omissions in
the inventory, and the fact that ingestion exposure was not assessed.
98
-------
5.3.2 Pollutant-Specific Hazard Quotient for Populations
Figures 5-6 and 5-7 show numbers of adults and children, respectively, residing in census
tracts where median exposures exceeded three fixed HQ levels (0.1, 1.0, and 10) for
emissions from all sources combined. The same information is provided in Appendix L
(Figures 19-23 and 25-29) for each source sector (major, area, mobile, and background
sources). Unlike the individual HQ figures (which exclude the extreme ends of the
distribution), these population figures include the entire assessed population, with the
goal of increasing sensitivity to identify potential urban-scale areas of concern. (As
already noted, the national-scale assessment was designed to assess exposure and risk
levels across large populations. Therefore, this use to identify urban-scale areas may
produce false negatives.)
Pollutants can be grouped in categories of decreasing importance in the assessment as
follows:
National hazard drivers: Figures 5-6 and 5-7 show that one pollutant, acrolein, presented
an HQ exceeding 10 to more than 20 million adults and 4 million children. Virtually all
adults and children in the US population lived in census tracts where the median HQ
exceeded 1.0. Similar results were obtained by considering mobile sources alone
(Appendix L, Figures 21, 22, 27, and 28).
Regional hazard drivers: Three more pollutants - formaldehyde, acetaldehyde, and
manganese - presented HQs greater than 1.0 to more than 100,000 adults and 20,000
children. Most of the formaldehyde and acetaldehyde exposures were attributable to
nonroad mobile sources, while manganese exposures originated primarily from area
sources.
Important regional contributors: Two more pollutants - cadmium and arsenic -
presented HQs greater than 1.0 to more than 10,000 people. Arsenic contributions
originated mostly from a mixture of major and area sources, and cadmium came mostly
from area sources alone.
5.3.3 Aggregate Target Organ Specific Hazard Index of Multiple
Pollutants to Individuals
Aggregating hazards of multiple pollutants that have the potential to cause adverse health
effects other than cancer is inherently more complex than assessing cancer risk because
(1) it is necessary to consider different toxic effects and mechanisms that may not be
additive, and (2) it is necessary to consider adults and children separately.
Regarding the first issue, the most effective way to aggregate hazards for multiple
pollutants would be to combine hazard quotients for pollutants that cause the same
adverse effects by the same toxic mechanism. However, because detailed information on
toxic mechanisms was not available for most of the substances in this assessment, EPA
used a simpler and more conservative method, the target organ-specific hazard index
(TOSHI). The TOSHI is the sum of hazard quotients for pollutants that affect the same
organ or organ system. This assessment calculated TOSHIs for the respiratory system,
99
-------
blood and blood-forming tissues, central nervous system, liver and kidney, cardiovascular
system, and immune system.
Exposures to children and adults were estimated separately to respect children's
potentially greater susceptibility to some toxic effects. This susceptibility may be due to
physiological differences from adults, different activity patterns that may lead to higher
exposures, or both. However, dose-response assessments for non-cancer effects
developed by EPA and other agencies do not currently include separate reference
concentrations (or similar values) for adults and children. Therefore, adult and child
hazard quotients all had the same denominators. Also, the HAPEM4 exposure model
takes into account the different activity patterns of children but does not consider their
higher mass-specific inhalation rates. Because adults tend to commute to areas of higher
concentrations more frequently than children do, HAPEM4 systematically produced
higher inhalation exposure estimates, and higher TOSHIs, for adults than for children.
For this reason, most of the discussion of individual non-cancer hazards in this section
will focus on adults.
To avoid mixing hazard quotients for well-understood pollutants with those from
pollutants for which data are sparse, separate TOSHIs were calculated for two groups -
pollutants for which the total RfC uncertainty factor ranged from 1 to 100, and those for
which the uncertainty factor exceeded 100.
Figure 5-8 shows distributions of individual TOSHIs for the respiratory system for adults.
Distributions of individual TOSHIs for six organs or organ systems, for adults and for
children are included in Appendix L (Figures 33-38 for adults and Figures 45-50 for
children). Of the six TOSHIs computed, only the respiratory TOSHI for low-certainty
pollutants exceeded 1.0 for more than 1% of the US population. This result was
dominated by a single substance, acrolein. The median of this TOSHI distribution was
approximately 4, and its 99* percentile was approximately 20. More than three quarters
of the population had a respiratory TOSHI greater than 1.0.
Much of the respiratory TOSHI was contributed by onroad mobile sources, with nonroad
mobile and area (and other) sources contributing lesser but still important amounts.
Major sources did not add much to the respiratory TOSHI nationally, and none of the
pollutants in this TOSHI had an estimated background contribution.
5.3.4 Aggregate TOSHI of Multiple Pollutants to Populations
Figures in Appendix L (Figures 39-44 and 51-56) show numbers of adults and children,
respectively, residing in census tracts for which the median exposures to multiple
pollutants produced TOSHIs exceeding three fixed levels (0.1, 1.0, and 10) for emissions
from all sources combined. Unlike the individual TOSHI figures (which exclude the
extreme ends of the distribution), these population figures include the entire assessed
population in order to increase sensitivity to potential urban-scale hot spots.
Pollutants can be grouped in categories of decreasing importance in the assessment as
follows:
100
-------
National hazard drivers: Results show that nearly the entire US population resided in
census tracts where the median TOSHI for respiratory effects exceeded 1.0, and that
about 25 million adults and 4 million children resided in tracts where the respiratory
TOSHI exceeded 10 (see Appendix L, Figures 39 and 51). These results are for the low-
certainty group of pollutants alone, and were dominated by acrolein.
Regional hazard drivers: TOSHIs for four more organs or organ systems - blood and
blood-forming tissues, central nervous system, liver or kidney, and immune system -
exceeded 1.0 for more than 100,000 adults and 20,000 children. Results for the immune
system were for high-certainty pollutants; results for the other three target organs were
for low-certainty pollutants. Of these, the central nervous system TOSHI potentially
affected the largest populations. Approximately 600,000 adults and 100,000 children
resided in tracts where the CNS TOSHI exceeded 1.0, and 10,000 adults lived in tracts
where the CNS TOSHI exceeded 10.
The sixth TOSHI, for the cardiovascular system, did not exceed 1.0 for more than 10,000
people nationwide.
5.4 Discussion of the Risk of Diesel Exhaust
Although EPA is providing concentration exposure information on diesel particulate
matter as a surrogate for diesel exhaust, the Agency is unable to provide the same
quantitative information in this risk characterization as is provided for the other 32 air
toxics. At the national level, EPA believes that diesel exhaust is one of the air toxics that
poses the greatest risks to the public based on its potential carcinogenic effects and other
health effects related to diesel exhaust, especially since diesel engine emissions provide a
substantial contribution to fine particle emissions.
EPA=s Clean Air Scientific Advisory Committee (CASAC) recently approved
conclusions that EPA has reached regarding the lung cancer hazard and risk of diesel
exhaust [42]. In EPA=s draft Health Assessment Document for Diesel Exhaust (HAD),
the Agency concluded that diesel exhaust is likely to be carcinogenic to humans at
environmental exposure levels that the public faces (classifying it as a "probable" human
carcinogen in the scheme used in this NATA report)[¥3]. However, as stated in the
"Heavy-Duty Engine and Vehicle Standards and Highway Diesel Fuel Sulfur Control
Rule" [44], EPA has concluded that the available data are not sufficient to develop a
confident estimate of cancer unit risk. The Agency concluded in developing its
perspective on risk in the HAD that there is a reasonable potential that environmental
lifetime cancer risks ^environmental risk=) from diesel exhaust may exceed one in a
hundred thousand and could be as high as one in a thousand. The environmental risk
estimates included in the Agency=s risk perspective are meant only to gauge the possible
magnitude of risk to provide a means to understand the potential significance of the lung
cancer hazard. The estimates are not to be construed as cancer unit risk estimates and are
not suitable for use in analyses which would estimate possible lung cancer cases in the
exposed populations.
101
-------
EPA recognizes that, as in all such risk assessments, there are uncertainties in this
assessment of the environmental risk range including limitations in exposure data,
uncertainty with respect to the most accurate characterization of the risk increases
observed in the epidemiological studies, chemical changes in diesel exhaust over time,
and extrapolation of the risk from occupational to ambient environmental exposures. As
with any such risk assessment for a carcinogen, despite EPA=s thorough examination of
the available epidemiologic evidence and exposure information, at this time EPA can not
rule out the possibility that the lower end of the risk range includes zero.19 However, it
is the Agency=s best scientific judgment that the assumptions and other elements of this
analysis are reasonable and appropriate for identifying the risk potential based on the
scientific information currently available.
Even the lower end of the risk range (presented in the risk perspectives section of the
Diesel Exhaust HAD) is above the level that has historically warranted regulatory
concern at EPA for air toxics. The Agency believes that areas of the U.S. that have
relatively higher annual exposure levels for diesel exhaust, certainly those counties and
States with annual exposure average levels above 2 micrograms per cubic meter, should
consider the scientific judgments that the Agency has made in the risk perspectives
section of the HAD while considering the important limitations in their efforts to
compare air toxic risks and set priorities for their programs. At the higher exposure
levels found in a number of urban areas in NAT A, there is an overlap between what the
occupational levels were in the epidemiological studies that EPA considered and
environmentally equivalent exposures.
There is substantial evidence that diesel exhaust alone and as part of mixture of fine
particles is associated with harmful respiratory and cardiovascular health effects
including an association with premature mortality. In addition to the direct emissions of
diesel engines of fine particulate, the NOx, SO2, and VOC emissions from these sources
are transformed into substantial concentrations of fine particles in the atmosphere (e.g.
nitrates and sulfates). The Agency provided an assessment of the seriousness of the
health effects associated with human exposure to fine particles in a Criteria Document
(CD) in 1996 [45]. Recent major reanalysis of two of the most critical studies regarding
the health effects of long-term exposure to fine particles examined in that CD confirmed
the findings of associations between long-term fine particle exposure and mortality [46].
19EPA's scientific judgment (which CAS AC has supported) is that diesel exhaust is likely to be
carcinogenic to humans. Notably, similar scientific judgments about the carcinogenicity of diesel exhaust
have been recently made by the National Toxicology Program for the Department of Health and Human
Services, NIOSH, WHO, and OEHA of the State of California. In the risk perspective discussed above,
EPA recognizes the possibility that the lower end of the environmental risk range includes zero. The risks
could be zero because (1) some individuals within the population may have a high tolerance level to
exposure from diesel exhaust and therefore are not susceptible to the cancer risks from environmental
exposure and (2) although EPA has not seen evidence of this, there could be a threshold of exposure below
which there is no cancer risk.
102
-------
5.5 Uncertainty and Variability Analysis for the NA TA National-
Scale Assessment
5.5.1 Introduction
EPA's guidelines for risk characterization recommend that estimates of health risk be
presented in the context of uncertainties and limitations in the data and methodology.
The degrees to which different types of uncertainty and variability need to be quantified,
and the amount of uncertainty that is considered acceptable, may vary with the scope and
purpose of the assessment. Because the national-scale assessment is generally intended
for prioritization, tracking national trends and progress, and setting the research agenda
for the air toxics program, EPA can accept a higher degree of uncertainty than if the
assessment were intended for supporting regulatory actions or as a final assessment of
risk at the local scale. Instead, the national-scale assessment is only a part of the analyses
that EPA intends to conduct to guide and inform the Air Toxics Program. It is inevitable
that local details in a nationwide study will be more uncertain than in an optimized local
study, and that the uncertainty of the nationwide study will be described in more general
terms.
Here, we refer to "uncertainty" as imperfect knowledge regarding the values of specific
parameters included in the assessment, and "variability" as real differences in the values
of specific parameters among places or individuals included in the assessment.
Generally, uncertainty can be reduced by gathering better data, whereas variability cannot
be reduced but can be characterized through more refined information or model
resolution.
EPA had hoped to undertake a quantitative analysis of uncertainty and variability within
each component of the national-scale assessment, using a "bottom-up" approach. The
intent was to methodically estimate the range of possible values (and use frequency
distributions where supported by more complete data) for each parameter used in the risk
calculations. These ranges and distributions would then be used as input for separate
calculations of the propagation of uncertainty and variability for all variables combined.
However, the EPA technical experts who contributed the various components of this
assessment could not with confidence place quantitative estimates, or even semi-
quantitative order-of-magnitude estimates, on uncertainty and variability for many of the
input parameters. It appears unlikely that a complete "bottom-up" approach to
characterizing uncertainty will be feasible without significant additional work.
Nevertheless, in order to obtain a minimal estimate of how much higher or lower the risks
calculated by this assessment are likely to be, we have provided an illustration of a "top-
down" approach to estimating some of the uncertainty and variability associated with (1)
modeling of ambient concentrations, (2) estimation of personal exposure, and (3) dose-
response assessment. This illustration also shows how this uncertainty and variability
might propagate into the final estimates of risks.
The illustration has several important limitations that readers should keep in mind. First,
it was not possible to fully separate variability from uncertainty with this "top-down"
103
-------
approach. Therefore, the propagated characterization of uncertainty incorporates both.
Second, the quantitative uncertainty estimates did not include all sources of uncertainty,
and the combined estimates of uncertainty seem likely to be underestimates. True
uncertainty and variability may be greater, but are not likely to be less.
The uncertainty and variability section ends with recommendations to develop a plan for
a more rigorous and comprehensive assessment of uncertainty and variability in this and
future national-scale assessments the future, based on more refined information for
individual components of the assessment.
5.5.2 Source Characterization
5.5.2.1 Data Sources
As described in Section 4.2.1.3, the 1996 NTI is a composite of emissions estimates
generated by State and local regulatory agencies and EPA using emission estimation
techniques determined by agency, pollutant, and source category. These emissions
estimates differed in quality, number of pollutants included, level of detail, and
geographic coverage. Furthermore, EPA did not attempt to verify the methods by which
emissions were estimated or undertake a full quality assurance and quality control
evaluation of the NTI.
As discussed in Section 4.2.1.2.1, EPA compared the information in the NTI to the
emissions in the CEP, and the point sources included in the TRI, and NET data sets,
which have been used in other dispersion modeling exercises. However, each of these
comparison data sets was developed for different purposes and had different facility and
pollutant coverage. Also, their data were compiled by methods likely to be surrounded
by at least as much uncertainty as the NTI. In general, the NTI compares favorably to
these data sets considering its inclusion of all anthropogenic source sectors and the level
of emissions detail required for this study (e.g., location coordinates, stack parameters,
etc.).
The quality and associated uncertainty in the NTI emissions data varies according to
source sector and individual source categories and pollutants. For point sources,
including those categorized as "major sources," emissions estimates and the associated
facility/stack details are more certain where more effort has been concentrated on
improving estimates (e.g., under certain MACT standards development efforts that
included gathering individual facility emissions data or within states that have gathered
emissions data for certain facilities for permitting or planning purposes). By their nature,
nonpoint emissions (area, other, or mobile), that are compiled in the NTI as county-wide
emissions estimates, rather than at known location coordinates, are less certain than point
sources. Many of these nonpoint emissions are estimated by using emission factors, that
may be out dated, and surrogate information (e.g., industry sales, population. Uncertainty
due to such surrogate allocation schemes are particularly pronounced for mobile source
categories, especially within the nonroad sector where, in order to simplify computation,
the county-level emissions from hundreds of nonroad sources (e.g., recreational marine
vessels, lawn mowers, construction equipment, etc.) were consolidated into three source
104
-------
category groups (2-stroke gas, 4-stroke gas, and diesel engines) prior to spatial allocation.
It is important to note that the 1996 NTI is not static. EPA intends to periodically update
it, as resources allow and as more reliable data (e.g., addition of missing information,
removal of double-counting, improved emission factors) become available for that year.
Thus future assessments based on the 1996 NTI may produce different results. These
factors represent important sources of uncertainty in the assessment.
5.5.2.2 Emission Locations
5.5.2.2.1 Point Sources
Locations for many point sources in the NTI were unknown, and had to be placed via
default mechanisms (e.g., using zip codes or counties). EPA prepared a State-by-State
analysis of proportions of emissions of three metals that were located by default. This
analysis suggested that "area and other" sources dominated the total emissions, as
expected for these pollutants. Although this comparison did not provide a quantitative
assessment of uncertainty, the magnitude of sources and emissions defaulted in a given
State suggests that there is uncertainty in the results due to potential location errors.
Nationwide, fewer than 10% of point source sites of primary modeled emissions of the 32
pollutants in this assessment were based on default locations.
5.5.2.2.2 County-Level Emissions Sources
All county-level emissions (i.e., mobile sources and many area and other sources) were
spatially allocated to the census-tract level prior to ASPEN modeling, using appropriate
surrogates such as population or land use using methods described in Appendix 3. In
addition to uncertainties introduced by the use of surrogates for allocation, the surrogate
information itself was subject to substantial uncertainty due to age or variable quality.
5.5.2.3 Stack Parameter Defaults
A facility in the NTI may have contained many emissions release points, and each release
point required several process parameters for dispersion modeling. When these
parameters were missing or were out of reasonable range they were replaced by defaults.
EPA compiled a table of the percentage of facilities for which defaults were used for
eight individual parameters, and found that on average, about one-quarter of the facility
data had to be augmented by one or more defaults.
5.5.2.4 Particle Size and Reactivity Assignments
All emissions input for ASPEN modeling had to be categorized according to particle size
fraction (coarse versus fine) and reactivity to support the model's accounting for
deposition and subsequent chemical reactions. None of the emissions inventories used in
this assessment contained this information; all had to be assigned by EMS-HAP. These
assignments added to the overall uncertainty.
5.5.2.5 Chemical Speciation Data
The NTI did not include uniform speciation information for HAP groups (e.g., polycyclic
organic matter, chromium and compounds) uniformly. In the process of preparing the
105
-------
NTI data for ASPEN modeling, EPA made important assumptions about chemical
speciation within these groups, creating additional uncertainty in the modeling results.
5.5.3 Ambient Concentration Estimation
5.5.3.1 Temporal Resolution of Emissions
The NTI provided the raw emissions data, which were processed and made "model-
ready" for ASPEN by EMS-HAP. In addition to the spatial allocation described in
section 1.2.1 above, the ASPEN model also required higher temporal resolution. To
support this, EMS-HAP allocated annual emissions into days and daily emissions into
three-hour periods.
5.5.3.2 Simplifying Assumptions
The ASPEN model does not simulate local terrain effects and the local meteorological
conditions associated with terrain effects, which may have caused local-scale
inaccuracies in the predicted concentrations. Also, ASPEN does not model transport of
any pollutant beyond 50 km from its original emission point. For some air toxics in this
assessment (e.g., reactive volatile organic compounds), this assumption was reasonable,
while for others (e.g., persistent metal compounds), this assumption represented a
significant source of uncertainty. Finally, the ambient concentration estimates included
uniform "background" concentrations for 13 pollutants, applied across all geographic
areas. This was an important simplifying assumption, because such extrinsic
concentrations may vary geographically for many of these pollutants. These three
simplifying assumptions create currently unquantifiable but important uncertainties in the
ASPEN outputs.
5.5.3.3 Meteorological Characterization Uncertainties
Meteorological data are a critical input for the ASPEN model. EPA's analysis of the
influence of two different sources of input data showed that ambient concentration
estimates varied within a range of minus 17% to plus 84%. EPA used the 1996 database,
which had a lower mean source-to-meteorological station separation distance that was
more representative, for this assessment.
5.5.3.4 Model Formulation and Methodology Uncertainties
5.5.3.4.1 Deposition and Dispersion Algorithms
The ASPEN air quality model employs a Gaussian plume model to characterize transport
and dispersion, using algorithms extracted from version 2 of the Industrial Source
Complex Long-Term model (ISCLT2). This model has a history of usage and
development that dates back to the 1960's, suggesting that we can anticipate its level of
uncertainty. As discussed in Section 4.2.2.4, comparisons have suggested that
approximately 90% of the model's estimates are within a factor of two of those observed.
As described in Section 4.2.2.4, EPA verified that the model formulation algorithms were
performing as anticipated by using lead emissions to test both the deposition and
dispersion algorithms were working properly, and whether the use of a "net" of receptors
106
-------
might lead to a bias in the modeling results. The results suggested that ASPEN may
predict average ambient lead concentrations 20 to 30% lower than a more refined model,
because ASPEN uses coarse particle deposition velocities that are higher than typically
used in local-scale modeling.
5.5.3.4.2 Atmospheric Transformation Algorithms
Although Gaussian dispersion models can be modified to handle zero-order (linear)
production or removal, they typically do not treat nonlinear chemistry effects. To address
this limitation EPA developed a mechanism to separately estimate exponential formation
and added (or subtracted) the secondary formed concentration to that attributable to
primary emissions. Using this mechanism, ASPEN results for formaldehyde and acrolein
were 23% and 44% attributable secondary formation. However, results from the more
refined OZIPR model suggest that secondary formation for formaldehyde and acrolein
would account for 90% and 85% (respectively) of the total. These results should be
considered only approximate, because the two models were not directly compared (i.e.,
using equivalent emissions and meteorology), but they do suggest that ASPEN may be
underestimating changes in concentration due to reactivity.
5.5.3.4.3 Interpolation Between Census Tract Centroids
EPA's comparison of ASPEN and ISCLT3 outputs (to determine if ASPEN's
interpolation scheme might be underestimating actual impacts) found that ASPEN
estimates were lower than ISCLT3's by about 10% in the near distances, with the
underestimation increasing to about 25% at 30 km downwind.
5.5.3.4.4 Summary
As discussed above, the use of ASPEN to estimate ambient concentrations added
important elements of uncertainty to the national-scale assessment. The simplifying
assumptions that made the a national-scale dispersion modeling possible at all - lack of
terrain effects, a 50-km radius for effects, and the application of uniform national
background concentrations for some pollutants - each have uncertainties that would be
difficult to quantify but that may be important.
However, it was possible to make some rough estimates of the performance of the entire
modeling system, and of some of its parts. Past studies with air toxics at the local scale
have suggested that about 90% of estimated concentrations should be within a factor of 2
of those observed, assuming well-characterized emissions and representative
meteorological data. EPA's comparisons of the formulation and methodology used by
ASPEN and other dispersion models suggested that differences in dispersion algorithms
were minor, likely leading to performance differences of less than 10% in the near field.
ASPEN's deposition velocities for coarse particles showed larger differences, and were
estimated to bias concentrations roughly 30% low, a bias that affects only pollutants
simulated as having coarse particles. Investigations of uncertainties associated with
chemical reaction effects suggested that concentration estimates for reactive species
should be considered more uncertain than for non-reactive species.
107
-------
Spot checks on location uncertainties for three metals suggested that 6 to 30% of these
emissions were assigned to locations rather than modeled at their true locations. A site
visit to one lead smelter found fugitive emissions that appeared greater than those
reported in the NTI, suggesting that close inspection of facilities might reveal other
under-reported non-point emissions. These emission characterization uncertainties could
have a greater impact on the model-to-monitor comparison results (resulting in
differences of a factor of 3 or more) than uncertainties seen elsewhere in the modeling
algorithms or the meteorological characterizations (which appear to result in differences
of 30 to 80%).
5.5.3.5 Illustration of Uncertainty and Variability Associated With Ambient
Concentration Estimates
EPA's statistical comparison of ASPEN model estimates with monitored ambient
concentrations at the same locations (Section 4.2.2.3) provides an opportunity to estimate
the uncertainty in the model's performance for seven substances (benzene,
perchloroethylene, formaldehyde, acetaldehyde, cadmium, chromium and lead)
representing different source types, deposition characteristics, and atmospheric reactivity.
The model-to-monitor comparison showed that modeled estimates for most of the
pollutants examined were on average lower than measured concentrations. The degree to
which the model underestimated the measured concentration was greater for reactive
gases than for stable ones, and greater still for particulates. In addition, the variance of
the ratios generally appeared to increase with the degree of underestimation.
If there were sufficient monitored values, and these values represented truth, this
information would be a reasonable approximation of total uncertainty contributed by the
National Toxics Inventory, all ASPEN model inputs, and dispersion model error.
Although there is uncertainty regarding the monitored values, the following illustration
shows how a model-to-monitor comparison approach can be used to estimate model
system uncertainty. The illustration also provides, for a limited set of pollutants,
calculated values based on this approach.
As described in Section 4.2.2.3, the tendency of the model to underestimate measured
levels decreased when the comparison was made with the highest modeled concentrations
within 10 and 20 km of monitor locations. This result suggested that some part of the
underestimation could be attributed to spatial uncertainty of the underlying emission and
meteorological data and the tendency of air monitoring networks to select sites in high-
impact areas. Nevertheless, there were many locations for which maximum model
estimates were still lower than measured concentrations, even at distances up to 50 km.
These underpredictions were judged most likely due to underestimated (or missing)
emissions data or uncertainty in chemical transformation assumptions.
In this illustration, total uncertainty surrounding the modeled estimates of ambient
concentrations was estimated quantitatively, as follows: (1) The model-to-monitor ratios
from the original analysis were inverted, to provide an estimate of the factor by which the
ASPEN model under-predicted each monitored concentration. (2) The populations of
resulting monitor-to-model ratios for seven pollutants were fitted to lognormal
108
-------
distributions using Crystal Ball 4.0 software. The parameters of the fitted distributions
are shown in Table 5-1. Variability could not be estimated because only the 1996 NTI
was modeled. Therefore, comparisons to monitored data for other years were not
possible.
In addition to expressing the uncertainty of the ASPEN estimates, the fact that the mean
monitor-to-model ratio for all
seven pollutants exceeded 1.0 was
consistent with the reported
tendency of ASPEN to under-
predict ambient concentrations.
Furthermore, both the degree of
bias and the total range of
uncertainty varied between classes
of pollutants. Accordingly, to
better support extrapolation of
these results to pollutants that were
not monitored in 1996, EPA
developed composite monitor-to-
model frequency distributions for
three classes of pollutants: stable gases; reactive gases (i.e., subject to atmospheric
transformation), and particulates using standard Monte Carlo simulation techniques21.
Stable gases were represented by the geometric mean of randomly-selected ratios for
benzene and perchloroethylene; reactive gases were represented by the geometric mean
of ratios for
formaldehyde and
acetaldehyde.
Particulates were
represented by the
geometric mean of
ratios for lead,
cadmium, and
Table 5-1. Illustration: Parameters for lognormal
distributions fitted to monitor-to-model ratios for
seven pollutants.
Pollutant
Benzene
Perchl oroethy 1 ene
Formaldehyde
Acetaldehyde
Lead
Cadmium
Chromium
Mean
1.19
2.26
2.28
2.69
15.37
12.53
6.05
Standard
Deviation
0.51
1.56
1.60
2.25
37.23
24.40
9.51
Table 5-2. Illustration: Calculated percentiles for
monitormodel ratio distribution.
Monitor Model
Ratio for:
Stable gas
Reactive gas
Particulate
2.5%
0.69
0.76
1.2
5%
0.78
0.88
1.4
50%
1.4
2.0
4.9
95%
2.6
4.3
16
97.5%
2.9
5.0
20
chromium. Results are shown in Table 5-2 and Figures 5-9 to 5-11.
A monitor-to-model ratio of 1.0 indicates perfect agreement. For stable gases, 95% of the
ratios fell between 0.69 and 2.9. For gases subject to transformation, 95% of the ratios
fell between 0.76 and 5.0. For particulates, 95% of the ratios fell between 1.2 and 20.
Because these ranges clearly show different amounts of both bias and variance, these
three groups of pollutants have been considered separately in the analysis of aggregate
uncertainty.
20 The lognormal distribution was the continuous distribution of best fit for five of the pollutants, and
provided second-best fit for the other two.
21 Simulation was performed with Crystal Ball 4.0 software, using 10,000 iterations with Monte Carlo
sampling and an initial random number seed of 0.
109
-------
This characterization of uncertainties in the emissions inventory and dispersion modeling
portions inherently assumed that monitored ambient concentrations were measured
without error, and that ASPEN's divergence from monitored levels was the result of
either model error or errors in inputs to the model. Of course, ambient concentrations
cannot be measured without error, and further errors may have occurred when these data
were compiled and reported. Nevertheless, EPA believes that this approach has the
potential to provide a fair approximation of the total uncertainty associated with the use
of ASPEN for this assessment.
5.5.4 Personal Exposure Assessment
5.5.4. •/ Microenvironment Factors
To predict personal exposure, HAPEM4 modeled the relationship between ambient air
quality and concentrations in microenvironments occupied by human receptor
populations. Because its use on a national-scale precluded detailed mass balance
equations, HAPEM4's microenvironment modeling relied on microenvironment factors
(simple first-order relationships between outdoor and indoor air quality). However, the
outdoor-indoor relationship is not well documented or understood for many pollutants
and technical reviewers of this approach agreed there is a great degree of uncertainty
associated with it, suggesting that this uncertainty be assessed qualitatively or
quantitatively.
5.5.4.1.1 Population Cohorts
The assessment assumed that information from EPA's CHAD database for 40 cohort
groups adequately represented the activity patterns of the general population in all areas
of the country. Groups were selected to represent variability by age, race, and gender, in
order to support comparisons of various demographic groups while still allowing
aggregation of exposure across ages. However, there is no way to measure whether these
data actually captured the full range of human activity in each tract. For example, it is
possible and even probable that members of the same cohort behave differently in
different census tracts. It is also possible that selecting different types of cohorts (e.g.,
economic classes, inside vs. outside workers, etc.) may have encompassed more
variability.
5.5.4.2 Activity Pattern Sequence
HAPEM4 constructed annual average activity patterns from multiple 24-hour activity
patterns by combining patterns from different individuals, so that day-to-day correlations
in activity were not preserved. These aggregated activity patterns were therefore more
representative of population averages for that demographic group, rather than individual
patterns. For this reason, exposure distributions for the each group represented
uncertainty in the population average rather than variability in individual exposures.
Uncertainty and variability of input data other than activity pattern were not considered,
so that the resulting uncertainty information provided by the prediction distributions is an
underestimate of the overall uncertainty.
110
-------
5.5.4.3 Illustration of Uncertainty and Variability Associated with Exposure
Estimates
A comparison of HAPEM4 exposure estimates against personal monitoring data
(analogous to the ASPEN model-to-monitor comparison) would be valuable to quantify
the aggregate uncertainty of the exposure estimation methods. However, EPA is not
aware of any body of personal monitoring data that has not already been used in
developing HAPEM4, making any such comparison invalid. Therefore, it is not possible
to determine if the HAPEM4 exposure estimates were biased relative to measured
exposures. Nonetheless, it is possible as an illustration to compare ratios of ambient
concentrations to personal exposures from modeling and monitoring studies for other
pollutants to determine if HAPEM4 has captured an appropriate range of variability. For
this illustration, EPA utilized data collected for two criteria pollutants, ozone and
particulate matter, to develop frequency distributions for the exposure modeling portion
of the risk calculation.
Several simple statistics confirmed that HAPEM4 did not capture inter-individual
variability in exposure, in that it produced exposure estimates that varied too little among
people living in the same tract. For example, in no case (considering all tracts and
pollutants) did the 90th percentile exposure exceed the 50th percentile by more than 2%.
An analysis of the full range of exposures within several tracts showed that exposures for
the least and most exposed individuals differed by less than 20%. Additionally, a tract-
by-tract correlation analysis of ASPEN and HAPEM4 results produced correlation
coefficients equal to or greater than 0.998. These results contrasted sharply with real-
world measurements of personal exposure, which tend to have much greater interpersonal
variability.
Databases of matched ambient and personal monitoring data sufficient to support
estimating uncertainty and variability in personal exposure among people living in the
same area were not available for the pollutants in this assessment. As a surrogate for this
illustration, EPA used Spearman correlation coefficients for ozone and particulate matter
as examples of "typical" gases and particulates. These data may not be directly
applicable because (1) they were developed from monitoring data at a coarser resolution
than used in the national-scale assessment, and (2) pollutant characteristics may differ.
However, they are used here to illustrate data needs for "top down" estimates of
uncertainty and variability for air toxics.
The correlation coefficients (seasonal averages of 0.49 for ozone and 0.13 for PM, from
correlation analyses of ambient and personal measurements reported in EPA's draft
exposure assessment for particulate matter were input into a Monte Carlo simulation,
which developed a set of estimated personal-to-ambient ratios by sampling two correlated
uniform distributions22.
22 Populations of personal-to-ambient ratios were calculated using 10,000 iterations with Crystal Ball 4.0,
by dividing an personal exposure selected from a uniform distribution between 0.1 and 10 by an ambient
concentration linked to the exposure level by the appropriate correlation coefficient.
Ill
-------
The resulting distri
5-12 and 5-13.
For "typical"
gases, 95% of
simulated
ambient-to-
personal ratios
fpll hptwppn 0 OQ
)utions of ambient-to-personal ratios shown in Table 5-3 and Figures
Table 5-3. Illustration: Percentiles for uncertainty and variability in
the personal: ambient ratio distribution.
Personal: ambient ratio
for:
Gas
Particulate
2.5%
0.09
0.13
5%
0.14
0.21
50%
1.0
1.0
95%
7.6
4.5
97.5%
13
7.1
and 13.0. For particulates 95% of the ratios fell between 0.13 and 7.08. The relatively
weak observed correlation between ambient and personal concentrations for ozone and
PM results from elements of both uncertainty (e.g., errors in measurement of ambient and
personal concentrations) and variability (e.g., varying characteristics of different
microenvironments). Although data were too sparse to support separating variability and
uncertainty, it appears likely that variability among microenvironments is the dominant
factor.
While these distributions are obviously limited in their applicability to the use of
HAPEM4 to predict individual exposures, EPA currently lacks the information on
correlated ambient and personal data that would be needed to do better. Furthermore, this
analysis focuses only on the variations in ambient-to-personal ratios, and assumes that
HAPEM4 contributed no bias. EPA intends to seek better data in the hope of extending
and improving this part of the uncertainty analysis for future national-scale assessments.
5.5.5 Illustration of Uncertainty and Variability for Dose-Response
Assessment
Uncertainty in dose-response assessments is amenable at least to partial quantification.
Specifically, assessments that use statistical methods to determine benchmark doses and
to fit dose-response relationships to toxicological data often provide confidence intervals
for the results. This information is available for many RfCs and UREs. In addition, EPA
genetically considers RfCs for effects other than cancer as being surrounded by an
uncertainty "spanning perhaps one order of magnitude." These uncertainty
characterizations clearly incorporate both uncertainty and variability in a manner which
does not allow them to be separated. Additionally, it is important to realize that the
confidence intervals surrounding RfCs and UREs include only the statistical uncertainty
in interpreting the data. Uncertainties inherent in the choice of models to extrapolate
from animals to humans and from high to low doses are potentially far larger, and cannot
be quantified.
5.5.5.1 Unit Risk Estimates (UREs)
The UREs used in the national-scale assessment are subject to four major areas of
variability and uncertainty. First, many of the pollutants were classified as probable
carcinogens because data were not sufficient to prove causality in humans. It is possible
that some of these pollutants do not cause cancer at environmentally relevant doses, and
that true risk associated with these air toxics is zero. Second, all UREs in this study were
based on linear extrapolation from high to low doses. It is possible that the true dose-
response relationships for some pollutants may be less than linear, resulting in an
112
-------
overestimate of risk. Third, most UREs in this study were developed from animal data
using conservative methods to extrapolate between species. Human responses may differ
from the predicted ones. The first three elements are comprised entirely of uncertainty.
Fourth, most UREs in this study were based on statistical upper confidence limits, though
some were based on statistical best fits. (While this does not affect overall uncertainty,
UREs based on best fits should be unbiased, while those based on upper confidence
limits should be biased high.) This fourth element represents a combination of variability
(i.e., based on variation responses of different people or animals) and uncertainty (i.e.,
potential errors in the measurement of exposure and response). Because of the aggregate
treatment all four sources of variability and uncertainty described above, EPA considers
all its UREs to be upper-bound estimates.
Of these four areas of uncertainty and variability, only the variability element of the
fourth is amenable to quantitative analysis for this illustration. Some dose-response
assessments that determine points-of-departure or fit dose-response curves to data using
statistical methods also include enough information to support fitting a frequency
distribution to the URE. To illustrate this approach, EPA developed a frequency
distribution for benzene (Table 5-4 and Figure 5-14), using the lognormal distribution
and confidence
interval reported
in the recent
IRIS
reassessment.
To be consistent
with the use of ratios to describe uncertainty in the exposure assessment, the distribution
was assigned a mean of 1.0.
It is important to remember that this illustration is based on a single, well-understood
substance. While aggregated uncertainty illustration at the end of this section assumes
this distribution is typical of other carcinogens, it is possible that other substances (many
of which lack distributional information) may have greater statistical error terms. More
importantly, this frequency distribution represents only the statistical error term in the
dose-response assessment and does not address the other three important sources of
uncertainty in dose-response assessment. Information sufficient to quantify these other
uncertainties does not currently exist, and this estimate of uncertainty and variability
should be considered a minimum for the dose-response assessment as a whole. The true
aggregate of uncertainty and variability is likely to be much greater.
5.5.5.2 Reference Concentrations (RfCs)
EPA and other agencies express uncertainty in reference concentrations for effects other
than cancer using a series of uncertainty and modifying factors (UFs and MFs). UFs are
assigned for extrapolation (1) between species, (2) to sensitive individuals within a
species, (3) from subchronic to chronic exposure duration, (4) to estimate no-effect levels
from lowest-effect levels, and (5) to account for incomplete data. MFs may also be
assigned to consider other issues not covered by the standard UF categories. The
aggregate UF/MF depends on the number of extrapolations required, and is best viewed
Table 5-4. Illustration: Percentiles for variability in the benzene
URE.
Ratio of "true" URE to
the estimated URE
Benzene
2.5%
0.14
5%
0.19
50%
1.0
95%
5.3
97.5%
7.2
113
-------
as an expression of the possible range within which the RfC could change when more
complete data become available.
Quantitative estimates of the uncertainty associated with each of these factors are not
available. In general, if sufficient data were available to support a quantitative
assessment of these factors for any substance, the factors would not be needed in the first
place. For this reason, it is not possible to develop total uncertainty ranges for RfCs by
aggregating uncertainties of each UF or MF. However, EPA's definition for the RfC2
specifically describes it as having "uncertainty spanning perhaps one order of
magnitude." Because the variability among test organisms is likely to be only a small
part of this order-of-magnitude range, this component can be considered to be dominated
by uncertainty. Although the actual uncertainty in the RfC may vary among substances,
this order-of-magnitude generic range is probably the most reasonable characterization
for the purposes of the national-scale assessment.
Thus, for illustrating the uncertainty and variability surrounding RfCs and similar values,
EPA assigned the RfC a uniform distribution with upper and lower bounds of 3.0 and 0.3,
respectively, as shown in Figure 5-15.
5.5.6 Illustration of Propagation of Uncertainty and Variability
As discussed above, EPA has developed an illustration of a "top-down" uncertainty
assessment that quantitatively estimates a portion of the variability and uncertainty
associated with three major components of this assessment - the estimation of ambient
concentrations, the estimation of personal exposures associated with the ambient
concentrations, and the assessment of dose-response.
The illustration for ambient concentrations inherently includes all the uncertainties in the
NTI, the other inputs to ASPEN, and modeling error within ASPEN itself. Although the
relative contributions of each element cannot be distinguished, the method used has the
potential to fairly portray both bias and uncertainty for the aggregate NTI-ASPEN
component of the assessment. Because the annual average concentration for a specific
time and location has no variation, variability should not be an important component in
this part of the illustration.
The uncertainty and variability analysis for personal exposure characterizes the
"expected" variation between ambient and personal exposures noted for ozone and
particulates in other studies. In contrast, HAPEM4 produced essentially perfect
correlations between ambient levels and personal exposures in the national-scale
assessment. Therefore, applying an interpersonal variation term developed elsewhere
may better describe the variability that should be present in the HAPEM4 personal
exposure estimates, but was not. However, this illustration does not attempt to estimate
potential bias in the HAPEM4 personal exposure estimates, which would require personal
monitoring data not already used in the development of HAPEM4. As described, this
23 An estimate (with uncertainty spanning perhaps an order of magnitude) of a continuous inhalation
exposure to the human population (including sensitive subgroups) that is likely to be without an appreciable
risk of deleterious effects during a lifetime.
114
-------
element of the uncertainty/variability illustration should be dominated by variability,
most likely among microenvironments.
The uncertainty/variability analysis for dose-response should be considered a generic
illustration that would need to be repeated on a pollutant-specific basis. For carcinogens,
the variability surrounding the URE has been fitted to a distribution that should be
considered a lower estimate of the URE's plausible variability and uncertainty. However,
readers should keep in mind that most carcinogens are far less studied than benzene and
may have correspondingly greater uncertainty. Also, the analysis for benzene includes
only the statistical error term; other important sources of uncertainty that may be far
larger are outside the scope of the current analysis.
For noncarcinogens the uncertainty/variability analysis is based on the stated order-of-
magnitude of uncertainty in EPA's definition for the RfC (which should be dominated by
uncertainty rather than variability). All values within a 10-fold range were considered
equally likely. While obviously represents an oversimplification of the uncertainty
associated with the RfC, it nonetheless provides a quantitative illustration of the stated
uncertainties associated with the RfC derivation process. Better characterization of the
distribution of possible RfC values would require significant additional research.
The last step in the "top-down" illustration of a quantitative variability/uncertainty
assessment approach was the aggregation of uncertainty and variability associated with
ASPEN, HAPEM4, and the dose-response assessment process were combined by Monte
Carlo simulation24, as follows:
(Mon2Mod}(Pers2Amb}(DR) = RR
Where:
Mon2Mod= ratio of monitored to modeled concentration
Pers2Amb = ratio of ambient concentration to personal exposure concentration
DR = of "true" RfC or URE to deterministic estimate
RR = ratio of "true" risk to deterministic estimate
The calculation was repeated 10,000 times, with each of the three input distributions
sampled randomly and without correlation to the others. The risk ratio was estimated
separately for carcinogens and noncarcinogens, and for stable gases, transformed gases,
and particulates, for a total of six aggregate uncertainty estimates (Table 5-5 and Figures
5-16 to 5-21 below). As discussed, uncertainty and variability could not be separated for
this demonstration exercise because they were inextricably mixed in some elements.
24 Simulations were conducted with Crystal Ball 4.0 software, using 10,000 iterations, Monte Carlo
sampling and an initial seed value of 0.
115
-------
Table 5-5. Illustration: Combined uncertainty and variability, in
terms of the risk ratio (i.e., the ratio of "true" risk to estimated risk).
Risk Ratio for:
Cancer: stable gas
Cancer: reactive gas
Cancer: particulate
Noncancer: stable gas
Noncancer: reactive gas
Noncancer: particulate
2.5%
0.06
0.08
0.23
0.13
0.16
0.48
5%
0.11
0.14
0.41
0.22
0.27
0.76
50%
1.4
2.0
4.7
2.1
2.9
7.0
95%
20
29
61
19
29
57
97.5%
36
51
100
33
48
92
Because the propagation of variability and uncertainty did not capture important
uncertainties in the dose-response assessment process, and did not include potential bias
in the personal
exposure or
dose-response
elements, these
percentiles
should be
considered a low
estimate of the
total plausible
uncertainty and
variability surrounding the risk characterization.
For stable gases, 95% of likely cancer risk and noncancer hazard quotient values lay
within a range from approximately one order of magnitude below and 1.5 orders of
magnitude above the deterministic risk estimate. For reactive gases, 95% of likely values
fell between approximately one order of magnitude below and 1.7 orders of magnitude
above the deterministic value. For particulates, 95% of likely values fell between
approximately half an order of magnitude below and 2 orders of magnitude above the
point values. In general, stable gases had the smallest plausible risk range, followed by
reactive gases, and then by particulates.
This illustration of an approach to estimating combined variability and uncertainty from
the individual components of the assessment is a relatively crude estimate that is
nevertheless somewhat useful for putting the risk results for individual census tracts into
perspective. For example, a typical census tract median risk estimate for benzene was
about 10 in 1 million. Since benzene is a stable gas, one can use the table to provide a
minimal 95% confidence interval for benzene-related risks in that tract extending from
0.6 to 360 in 1 million. Similarly, a typical census tract median cancer risk for chromium
was about 2 in a million. The illustration provides a 90% confidence interval on this tract
estimate ranging from 0.8 in a million to about 120 in a million. While these confidence
interval illustrations are only approximations, EPA believes that they provide an
appropriate sense of the imprecision of risk estimates that deterministic risk estimates do
not give.
5.5.7 Aggregation of Risk Across Pollutants
Because the above illustration does not characterize the variability and uncertainty
associated with all 32 pollutants, it was not logical to finish it with a quantitative
propagation of uncertainty associated with aggregating risk across pollutants. However,
some insights about this can be gained from procedures for propagation of statistical
error. When aggregating risks across multiple pollutants, it is apparent from Table 5-5
that differences among classes of pollutants will lead to differences in aggregate
uncertainty/variability. Also, by simple error propagation theory, we can show that if:
116
-------
where RA is aggregate risk and n is the risk associated with pollutant i, the uncertainty
associated with the aggregate risk, UR, can be shown to be:
UR =
where Ur; is the uncertainty of the risk associated with pollutant i, and the uncertainties
associated with each pollutant risk are assumed to be independent. If we simplify this
equation by dividing each side by RA, we obtain:
UR ^-i _ , /_ 2
RA
A
where the left-hand side of the equation now corresponds to the fractional or relative
uncertainty in the aggregate risk. Examination of the right-hand side of the equation
shows that (1) pollutants with larger absolute uncertainties in their risks will contribute
more to the aggregate uncertainty, and (2) when individual pollutant risk uncertainties are
comparable, pollutants that dominate or drive the aggregate risk will also tend to
dominate the uncertainty in that result.
The implications of the partial quantification of uncertainties in this case study are
multiple. First, the uncertainties were large, limiting the use of the absolute values of the
risk results in any decision-making process. Second, the magnitudes of the uncertainties
tended to be similar across pollutants, reinforcing the notion that the relative
interpretation of risks can be useful in terms of setting priorities for further analysis or
data gathering. Third, while the uncertainties associated with the dose-response
component of the assessment were large, there appeared to be equally large uncertainties
associated with the emissions and dispersion modeling component of the assessment,
especially for particulates and reactive gases. Finally, mathematical procedures for
propagation of uncertainties associated with the aggregation of risks across pollutants
shows that pollutants which dominate or drive aggregate risk levels will also contribute
the most to the uncertainties in those aggregate risks.
5.5.8 Recommendations for Further Characterization of Uncertainty
5.5.8.1 Role of Uncertainty Analysis
Although the quantitative estimates of uncertainty described above are the best that can
currently be developed, they have important limitations. They are based on available
indicators of uncertainty and variability, which resulted in all the sources of uncertainty
being aggregated into only three steps and in the use results from one or a few pollutants
being applied to many other pollutants. For example, the aggregate uncertainty and
117
-------
variability in the emissions estimates and the dispersion and transformation modeling are
characterized by the monitor-to-model concentration comparisons, which in turn are
limited to the relatively few pollutants and locations with monitoring data. Thus, the
estimates described above likely do not properly capture all the important sources of
uncertainty in the population risk estimates. To the extent that this or future national
scale assessments are used in decision making, whether to set priorities among pollutants,
to decide on control actions, or to judge the progress of the control program, better
understanding of uncertainties will assist in that decision making process. Also, because
of the aggregation approach, the uncertainty estimates described above do not indicate
how the uncertainty in the risk estimates would be reduced if uncertainty in specific
components of the assessment were reduced. That is, they do not tell us much about how
to get the best "bang for the buck" in reducing the uncertainty in future assessments. For
both reasons, we believe that it should be a goal to improve the characterization of the
uncertainties in the national-scale risk assessment process.
EPA plans to repeat the national scale assessment in 2002-2003, for calendar year 1999
emissions and meteorological conditions, and we believe that steps already taken or
underway will lead to improved inputs. For example, we expect more states to provide
emission inventory data, and both EPA and states will have more experience with such
inventories than in this assessment for 1996. Because the 1999 national-scale assessment
will give us risk estimates for a second point in time, and incorporate improvements in
several important areas, reducing and better understanding the uncertainties the 1999
assessment will be arguably more important than better understanding the uncertainties in
this 1996 assessment. Nevertheless, EPA will be pursuing better approaches to
characterizing the uncertainties in this 1996 assessment, both to help focus the efforts to
improve the 1999 assessment and to test and refine techniques that may ultimately be
applied to the 1999 assessment.
5.5.8.2 Technical Issues in Further Characterization of Uncertainty
In a formal bottom-up uncertainty analysis, many individual components or input values
are considered to have uncertainty or variability, and that uncertainty or variability is
represented by a frequency distribution. Variability is then propagated upwards through
the assessment calculations using Monte Carlo methods, similar to the general approach
used in the case study. Given the broad geographic, multi-source, and multi-pollutant
scope of the national scale assessment, there are very many components or input values
that affect the final risk estimates, and hence that contribute to the uncertainty of the risk
estimates. Table 5-6 below is a structured list of components. The list provides a
framework for considering how a more complete, "bottom-up" uncertainty assessment
might be conducted. It should be noted that for most of the components listed there are
actually many individual quantitative input values that could be further distinguished and
subject to a formal uncertainty analysis. Moreover, many listed components of the
national scale assessment have "structural" uncertainties that are not directly addressable
by a Monte Carlo approach. Also, the chain of modeling steps involved in producing one
"run" from one set of inputs is time consuming. Repeating it hundreds or thousands of
time to create a distribution of risk estimates would be a demanding approach, and less
demanding alternatives should be considered. It may be sufficient, for example, to
118
-------
characterize the uncertainty of each main modeling step separately, and then consider
how these uncertainties affect the aggregate uncertainty.
Table 5-6. Sources of uncertainty for the national-scale assessment.
Emissions Data
Stationary source emission data sources:
a HAP inventories developed by State and local air agencies
a Databases related to EPA's MACT program
a Toxics Release Inventory data
a Emission factors and activity data
Mobile source emission data sources:
a On-road sources:
a Non-road sources
Preparing Emissions Data for Dispersion Modeling
Use of compound classes to group pollutant species
Assumptions used for other input data
a Use of default physical release characteristics (i.e., temporal pattern, spatial
pattern, release height, etc.) by SIC of SCC code
a Estimation of chemical characteristics (e.g., vapor-particle ratio, secondary
formation, reactivity class, particle size class)
a Use of surrogate data to allocate county emissions to census tracts
ASPEN Dispersion Modeling
Model error
a Limitations of Gaussian models (e.g., 50-km limit)
a Use of a single background assumption for selected HAPs
a Use of an assumption of flat terrain
a Use of meteorological data from nearest airport
HAPEM4 Exposure Modeling
a Use of ME factors to extrapolate from census tract to microenvironment
concentrations
a ME factors assumed independent of geography
a Use of CHAD activity data to represent behavior of entire demographic cohorts
a Annualized behavior of cohorts assembled from daily diaries
Dose-Response Assessment
Carcinogens
a Probable and possible human carcinogens assumed carcinogenic
a UREs based on linear extrapolation from high to low doses
119
-------
a Most UREs developed from animal data extrapolated to humans
a Most UREs based on the statistical upper confidence limit of dose-response
curve: some based on maximum likelihood estimate
a Grouping of aggregate cancer risk by weight-of-evidence
Non-Carcinogens
a Laboratory animal data to humans;
a Average healthy humans to sensitive humans;
a Subchronic to chronic exposure duration;
a LOAELtoNOAEL
a Incomplete database.
Risk Characterization
Cancer
a Use of assumption that cancer risks for different pollutants are additive
a Degree to which aggregation of risks based on upper-bound UREs may propagate
overestimates
Non-Cancer
a Assumption that effects to the same organ or organ system are additive
a Use of the RfC for the critical effect (i.e., the adverse effect appearing at the
lowest dose) to all effects (i.e., that may appear at substantially higher doses)
a Grouping of aggregate noncancer hazard by total uncertainty and modifying
factor
5.5.8.3 Future Plans
EPA intends to consider over the next year how to formulate an uncertainty analysis that
is appropriate to the particular nature of the national scale assessment process, with its
very large number of numerical inputs, particular structural approaches, and multiple
modeling steps. We will have to consider where to treat critical inputs individually and
where and how to use a more aggregated approach. We will also have to estimate the
bias, uncertainty, distributional shape, etc, of the inputs that are treated through frequency
distributions. This may involve the use of panels of experts in each of the following
areas: (1) emission inventories, (2) dispersion modeling, (3) exposure modeling, (4) dose-
response assessment, and (5) risk characterization. Additional data collection efforts
might need to be initiated to provide needed quantitative information to support the
process. Although it may not be possible to develop true frequency distributions to
characterize uncertainties surrounding some of the items in Table 5-6 without additional
data, these panels might be charged with providing bounding estimates at the least and
more, if possible.
120
-------
6 Summary and Recommendations
6.1 Perspective on the National-Scale Assessment for 1996
The purpose of the national-scale assessment is to gain a better understanding of the air
toxics problem. The national-scale assessment was not designed, and is not appropriate
specifically, for identifying local- or regional-scale air toxics "hot spots," nor is it
appropriate for identifying localized risks or individual risks from air toxics. Further
analyses on a national scale, and additional assessments on other scales (e.g., urban air
toxics assessments and residual risk assessments) are being performed in order to fully
characterize risks, especially disproportionate and cumulative risks. This means that this
initial national-scale assessment is not a complete characterization of the exposures and
risks associated with air toxics. Therefore, before considering the results, it is important
to understand the context of the assessment.
This initial application of the national-
scale assessment should be viewed
more as a step in building the analytical
framework necessary to estimate risks
of air toxics on a national scale - not as
producing a definitive estimate of these
risks. That is, the individual emissions,
dispersion/fate, exposure and risk
characterization tools have been
integrated for the first time and an
initial evaluation of the limitations and
uncertainty has been completed. In
addition, much of the data collection,
while representing a significant effort
on the part of many contributors, was
performed for the first time in this
initial national-scale assessment. Also,
some of the pre-existing tools were
used in their existing form in the
interest of time, even though
approximations and simplifications
could be improved with more effort.
Accordingly, EPA is continuing to
develop the process and an understanding of the uncertainties and limitations of the
national-scale assessment. Nevertheless, this application is complete enough to warrant
a scientific peer review that will provide a basis for making improvements to any future
national-scale assessment, through improved data collection, revisions to the tools and
appropriate research.
It is critical to understand that the national-scale assessment does not include other
relevant exposure scenarios (or scales). For example, the national-scale assessment does
Key Limitations of this National-
Scale Assessment:
• It is based on 1996 data
• It only includes 33 air toxics
• It only addresses inhalation
exposures and risks (not
ingestion, a significant
exposure pathway for some
air toxics (e.g., mercury))
• It does not capture localized
impacts and risks
• It focuses on average
population risks, rather than
individual extremes, for
ambient outdoor exposure
only
121
-------
not estimate exposures and risks at a local-scale. At best, such high exposures are partly
reflected in average exposure concentrations calculated in each census tract. The
national-scale assessment also does not include risks associated with emissions of air
toxics indoors. In addition, the national-scale assessment does not include exposures and
risks from ingestion and dermal exposure that may be posed by these pollutants. These
non-inhalation risks are an important component of the air toxics program and should not
be ignored in an overall characterization of the risks associated with air toxics. This
means that this initial national-scale assessment is not a complete characterization of the
exposures and risks associated with air toxics. Therefore, it is not appropriate to
determine that a particular air toxic is not significant based solely on this analysis.
As a consequence, the results should neither be viewed as final nor considered definitive
or in isolation. The assessment does, however represent our best effort to develop a
process for risk characterization for air toxics on a national scale. Additionally, the
results provide a picture from which to review and evaluate the technical aspects of a
national-scale assessment, to gain insight into the areas of greatest uncertainty, to aid in
identifying research priorities, and to provide an initial sense of relative priorities among
these first 33 air toxics for the air toxics program.
6.2 Summary of Initial Results of National Scale Assessment
In the risk characterization presented in Section 5, EPA grouped pollutants into four
categories based on the magnitude of the risk or hazard estimates and the number of
people potentially affected. Magnitude of risk was expressed by classifying a substance
as a "driver" (i.e., contributing a relatively large share of the total) or an "important
contributor" (i.e., contributing a smaller but still important share of the total). The
number of people affected was expressed by assigning a substance national scope (i.e.,
with potential impacts to millions of people) or regional scope (i.e., with potential
impacts to tens or hundreds of thousands of people). This categorization scheme
produced four groupings: national drivers, regional drivers, important national
contributors, and important regional contributors. Twenty-three of the 32 pollutants were
placed in one of these groups. One pollutant - poly cyclic organic matter - was grouped
both with regional drivers and important national contributors.
National drivers included acrolein, benzene, carbon tetrachloride, chromium, and
formaldehyde. Regional drivers included acrylonitrile, arsenic, coke oven emissions,
ethylene oxide, hydrazine, manganese, and polycyclic organic matter. Important national
contributors were acetaldehyde, 1,3-butadiene, ethylene dibromide, ethylene dichloride,
perchloroethylene, and polycyclic organic matter. Important regional contributors were
cadmium, chloroform, 1,3-dichloropropene, nickel, quinoline, and trichloroethylene.
In addition, as explained in Section 5.4, EPA believes that diesel exhaust is also one of
the air toxics that poses the greatest risks to the public based on its potential carcinogenic
effects and other health effects related to diesel exhaust, especially since diesel engine
emissions provide a substantial contribution to fine particle emissions. For the nine air
toxics not found to be important contributors to inhalation risks on a national or regional
scale, this result does not necessarily mean these pollutants are not important. It could
122
-------
indicate that their main impacts may be limited to the local or neighborhood scales at
which we expect the national-scale assessment methodology to under-predict individual
risks. These pollutants would therefore be better investigated with local-scale data and
assessment tools. It may also be that the initial national-scale assessment underestimated
ambient concentrations, and therefore exposures and risks, as appears to be the case with
many of the metals.
Mobile sources air toxics showed a strong association with national-scale risks, but the
remaining mobile source pollutants appeared to have limited potential for national- or
regional-scale risks. Major sources, in contrast, showed a strong association with
regional risks rather than national risks. Area sources appeared to produce important
risks on both the national and regional scales. Background sources were associated
exclusively with nationwide risks, as expected. Because background was assumed to be
the same in all tracts, exposure to background pollutants varied only with different human
activity.
In summary, the results of the national-scale assessment suggest that 23 of the 32 air
toxics have the potential to present some noteworthy risk to some group of people in the
U.S. It is important to remember the previously stated limitations when interpreting the
results of this assessment.
• Non-inhalation exposures such as ingestion and dermal, are not included. A complete
picture of risk would include additional pathways for exposure. This is especially
important for pollutants that persist in the environment and bioaccumulate, such as
mercury, dioxins, andPCBs.
• The highest localized exposures and risks are not captured by the national-scale
approach. As mentioned earlier in this report, two comparisons between the results of
the 1996 national-scale assessment and results from local-scale refined assessments
indicate that this limitation can lead to significant underestimation of risks in the
vicinity of individual point sources. These two comparisons showed an under
prediction of local-scale risks by a factor of 30 in an urban area and by a factor
greater than 100 in a rural setting.
• In a direct comparison with ambient monitoring data, the ASPEN model was found to
consistently underpredict annual average concentrations. While the best agreement
between the ASPEN model and ambient data was found for benzene (which ASPEN
underpredicted by about 10%), concentration estimates of some metal s were found to
be underpredicted by more than a factor of 5.
• Information on dioxins is still under review, and therefore, this pollutant has not been
included in the risk characterization. Since dioxins are considered to be an important
potential health threat, efforts will be made to include them in future assessments.
• Indoor sources of pollution are not included. While these are considered outside the
scope of the current study, it is important to recognize that, for certain hazardous air
123
-------
pollutants, total long-term human exposures can be significantly influenced and
sometimes dominated by exposures due to indoor sources.
• Sources, emission estimates, and exposure factors have a high degree of uncertainty.
6.3 Recommendations
As stated in section 2.1, the results of the initial national-scale assessment (in addition to
other assessments) are intended to help inform EPA as it continues to develop and
implement various aspects of the national air toxics program. The initial national-scale
assessment will assist by:
• Identifying air toxics of greatest potential concern, in terms of contribution to
population risk;
• Characterizing the relative contributions to air toxics concentrations and
population exposures from different types of air toxics emission sources;
• Setting priorities for the collection of additional air toxics data (e.g., emission
data, ambient monitoring data, data from personal exposure monitoring) for use in
local-scale and multipathway modeling and assessments, and for future research
to improve estimates of air toxics concentrations and their potential public health
impacts;
• Establishing a baseline for tracking trends over time in modeled ambient
concentrations of air toxics; and,
• Establishing a baseline for measuring progress toward meeting goals for
inhalation risk reduction from ambient air toxics.
The recommendations have been categorized below according to the assessment goal that
they support.
6.3.1 Identifying Air Toxics of Greatest Concern
Given the limitations and uncertainties of this initial assessment, it is not possible, at this
time, to identify definitively the air toxics of greatest concern from among these 33.
Nevertheless, it is possible to identify air toxics that appear to be important; that is, those
air toxics posing estimated risks at or above levels typically addressed by EPA.
However, it is not possible to eliminate other air toxics from consideration.
Of the 33 hazardous air pollutants assessed, those that appear to pose the greatest health
threats to individuals (from inhalation exposure) in all parts of the U.S. are chromium,
acrolein, benzene, formaldehyde, and carbon tetrachloride. Pollutants having the
greatest potential to pose health threats to individuals in some regions of the US include
acrylonitrile, coke oven emissions, hydrazine, ethylene oxide, manganese, andpolycyclic
organic matter. In addition, as explained in Section 5.4, EPA believes that diesel exhaust
is also one of the air toxics that poses the greatest risks to the public based on its potential
124
-------
carcinogenic effects and other health effects related to diesel exhaust, especially since
diesel engine emissions provide a substantial contribution to fine particle emissions.
Although other pollutants besides these twelve substances may also be important
contributors of health risk in some areas, these eleven pollutants account for most of the
total estimated national air toxics-related health risk in this assessment.
6.3.2 Prioritizing Efforts to Reduce Emissions
As discussed above, the results of the initial national-scale assessment should be used
cautiously when planning efforts to reduce air toxics emissions at the national level. The
national-scale assessment results may be used, in conjunction with other information, for
planning program activities, assessment activities and emission reduction efforts. For
example, the program decisions related to the urban air toxics program and priorities
associated with national technology-based standards for area and mobile sources can be
informed by results from the initial national-scale assessment. However, these decisions
should not be based solely on the results of the initial national-scale assessment. Because
other pollutants included in the initial national-scale assessment may have been
inadequately characterized or may be important contributors of health risk at a local scale
or health risk due to non-inhalation exposure (e.g., ingestion), they should not be
excluded from future consideration or assessments based on results from the initial
national-scale assessment alone.
Given the limitations and uncertainties, state and local agencies need to be particularly
careful in interpreting the initial national-scale assessment results. While these results
can be helpful in planning assessments within a state or local area, they should not be
used without confirming information to reduce the pollutant scope of an assessment, or to
decide on major control steps. The results alone should not be used to dismiss local
complaints about air toxics pollutants.
6.3.3 Characterizing Contributions of Sources
Given the limitations and uncertainties of this initial assessment, it is not possible, at this
time, to definitively characterize the relative contribution of sources. In particular, the
emissions for metals (with their corresponding low ambient and exposure concentrations
and risk estimates) may not be adequately characterized by the models in the assessment,
particularly at the national scale. Thus, at this time, it is difficult to describe the relative
risk contribution from different sources.
Accordingly, characterizing the relative contribution of sources based on the results of
the initial national-scale assessment can only be done in general terms. The results
appear to show that mobile and area sources of air toxics are responsible for a majority of
the health risk concerns that can be identified on a national scale. Nevertheless, certain
air toxics from major sources may present significantly greater levels of individual risk to
smaller portions of the population than the risks posed to a broader portion of the
population from mobile or area sources. Local-scale assessments (such as those being
performed as part of EPA's residual risk program and those being anticipated as part of
future NATA activities under the urban strategy) will be needed to more accurately
characterize the exposures and risks associated with major sources before a complete
125
-------
characterization of the relative contribution of all sources can be made.
As discussed in Section 5.4, the Agency is unable to provide the same quantitative
information for diesel PM in this risk characterization as is provided for the other 32 air
toxics. At the national level, EPA believes that diesel exhaust is also one of the air toxics
that poses the greatest risks to the public based on its potential carcinogenic effects and
other health effects related to diesel exhaust, especially since diesel engine emissions
provide a substantial contribution to fine particle emissions. Diesel PM emissions result
mainly from mobile sources and in the national-scale assessment EPA focused on
estimating the ambient concentrations and exposure from only these sources. EPA spent
considerable effort estimating the 1966 emissions inventories from onroad sources
(primarily diesel trucks) and nonroad sources (such as construction and farm equipment)
for this study. The national-scale assessment results suggest that throughout the country
most of the population's exposure to diesel PM emissions results from nonroad sources.
There is a relatively wide range of exposures concentrations that occurs in various
counties throughout the country with the highest exposures levels occurring as expected
in heavily populated urban areas.
6.3.4 Tracking Trends and Progress
EPA plans to update this assessment every three years and the next assessment will occur
in 2002-2003. The next assessment will focus on 1999 emissions, concentrations and
risks. The assessment of 1999 data will provide results that can be used to assess relative
reductions in air toxics emissions, ambient levels, exposures, and risks since 1996 (by
comparing with results of this assessment). It will also be important to determine how
effective this tool is in accomplishing that objective, to address the limitations and
uncertainties in the initial national-scale assessment, as well as the limitations and
uncertainties in the second national-scale assessment, to provide an adequate basis for
comparison. EPA will attempt to compare on a national-scale the relative impact of
reductions of the 33 air toxics since 1996. As previously mentioned, EPA plans to
develop a more complete approach for assessing uncertainties and variability of national-
scale assessment results over the course of the next year.
However, EPA will also begin to analyze its pilot monitoring results, residual risk
assessments, other localized efforts in an attempt to measure progress and the
effectiveness of the air toxics program.
EPA can also utilize the national-scale assessment approach in a predictive mode to
evaluate potential changes in population-based risks for candidate emission-reduction
scenarios. Results from such assessments can be used to inform discussions on future
voluntary or regulatory actions to reduce emissions of air toxics and their associated
risks.
6.3.5 Setting Data Collection and Research Priorities
The evaluation of the NATA national-scale assessment results is an iterative process.
The current evaluation has demonstrated the need for better information that, in turn, will
permit an improved evaluation in the future. As a consequence, EPA's Office of
126
-------
Research and Development has drafted an air toxics research strategy. This draft strategy
is expected to undergo peer review by the EPA's Science Advisory Board in the spring of
2001.
The initial national-scale assessment should assist EPA in understanding our risk
assessment tools (the National Toxics Inventory, dispersion and exposure modeling, and
dose-response information) the proper uses and limitations of each, and how to most
effectively improve them. The results of the assessment have shown that the following
steps should be taken to improve the quality of the next such assessment:
Improve the quality of emission data. EPA has already requested that State, local, and
tribal authorities that submit emission information to the National Toxics Inventory, for
the assessment year 1999, make specific improvements in the way they speciate
polycyclic organic matter, chromium, and nickel. Other important improvements include
(1) entering location coordinates and stack heights for all point source emissions, (2)
entering the specific method by which emissions were quantified, and (3) better
characterizing the spatial nature of mobile source emissions. In addition, it would be
desirable to conduct a study on a small sample of sources, to see if the emissions are
accurately located and that their rates are accurately estimated.
Improve support for urban-scale modeling. EPA plans to evaluate the initial national-
scale assessment by assessing air toxics on urban and local scales using more refined air
quality modeling tools that factor in specific local information such as terrain and local
weather patterns. The results of national-, urban-, and local-scale modeling would be
compared to provide a more complete context for the characterization of air toxics. State,
local, and tribal authorities have been encouraged to revisit and revise inventories in areas
of relatively high risk where urban-scale modeling would be valuable. EPA should focus
on improving spatial allocation methods for area and mobile sources and reducing the
need to rely on these methods by better characterizing the spatial resolution of its
emission inventory.
Improve the characterization of background. Background sources of five substances —
benzene, carbon tetrachloride, formaldehyde, ethylene dibromide, and ethylene dichloride
- were found to contribute substantially to estimated air toxics-associated health risks.
However, the assumption of a ubiquitous background concentration for each of these
represents an important simplification that is unlikely to hold true. With the exception of
diesel PM (see Appendix F), background was defined as that part of ambient
concentrations not caused by modeled sources within 50 kilometers of the receptor
population. This definition encompasses natural sources, international transport,
intranational transport, and persistent historic emissions, all of which may vary with
location. In addition, one component of background - intranational transport - most
likely includes sources that EPA has the authority to control and is therefore not properly
"background" at all. The uncertainties surrounding the treatment of background should
be reduced by (1) establishing a remote monitoring network for these substances, and (2)
improving the capabilities of EPA's national dispersion modeling approach to extend
beyond the current 50-km range.
127
-------
Provide support for future model-to-monitor comparisons for ambient air toxics
concentrations. Data from existing State and local air monitoring programs have already
been compiled to summarize EPA's current knowledge about ambient air toxics, and to
serve as a "reality check" on modeling results. As a result of Congressional direction and
SAB review of EPA's air toxics monitoring concept paper, an improved, expanded and
more representative air toxics monitoring network will be available in the near future to
better support model evaluation. To assist with the development of this network and to
provide better model evaluation information in the short-term, pilot monitoring studies
have been initiated which will have multiple monitors in four urban areas (Seattle,
Tampa, Providence, and Detroit) and also provide information in six smaller
communities across the country. Information from this assessment will also be used to
inform the monitor siting process. Limited information on background concentrations
will also be provided. These new data will provide a wider range of information than the
current monitoring data set.
Provide support for future model-to-monitor comparisons for exposure. Personal
exposure monitoring studies are needed to evaluate the ability of the exposure modeling
approach embodied in HAPEM4 to capture the average exposure levels for air toxics in
individual census tracts as well as to describe the complete distribution of that can be
expected across each census tract. Without such information, exposure models cannot be
improved to capture the full range of risks experienced by the exposed population.
Improve dose-response information. Within the scope of this assessment, most of the
cancer risk and virtually the entire non-cancer hazard are associated with substances in
the "lower-certainty" groups. For example, formaldehyde and carbon tetrachloride are
probable (i.e., less certain) carcinogens, and acrolein has a reference concentration that
has a total uncertainty factor of more than 100. EPA already has new assessments in
progress for these and other substances that are important sources of risk in this
assessment, but may want to consider expediting these assessments and subjecting them
to review by the EPA's SAB in addition to regular internal and external peer review.
Other areas in which dose-response information can be improved include development of
(1) organ-specific RfC (to reduce conservatism associated with applying critical-effect
RfCs to all target organs when combining non-cancer hazard across pollutants) and (2)
UREs based on maximum-likelihood rather than upper-bound estimates (to reduce
conservatism associated with combining cancer risks across pollutants).
Extend EPA risk assessment guidelines to be more inclusive of children and other
vulnerable subpopulations. EPA's current IRIS assessments for effects other than cancer
are severely limited by their use of the average inhaled concentration as the exposure
metric. In this way the IRIS assessments fail to support consideration of the different
inhalation rates and body weights of children. This prevented the initial national-scale
assessment from fully differentiating between adults and children. For the next such
assessment, it will be desirable to either adjust estimated inhalation concentrations to
children's equivalence, or to develop separate reference concentrations for children.
128
-------
Improve modeling to include multipathway exposures. One major limitation of this
assessment is that it does not fully account for exposure pathways, such as ingestion,
which are important for many persistent, bioaccumulative toxics. There are a variety of
approaches for accommodating other than inhalation exposures to air pollutants, each
with attendant strengths and limitations. In the Mercury Study Report to Congress [40],
non-inhalation mercury exposures (from fish ingestion) were estimated using monitoring
data (fish tissue), with a recognition that a portion of these exposures result from air
emissions. For local scale assessments, EPA has been developing an improved multi-
media fate and transport model for air pollutants (TRUVI.FaTE) which will be getting
some initial applications in 2001-2002. Additionally, development of a more refined
ingestion exposure model for air pollutants (ingestion component of TREVI.Expo) is
planned for 2002. These tools are expected to contribute to our multipathway exposure
modeling capabilities.
129
-------
7 References
7 U.S. EPA. 2000. Strategic Plan. EPA Document #EPA 190-R-00-002.
2 U.S. EPA. 1999. Integrated Urban Air Toxics Strategy. 64 FR 38705. Available on-line
at http ://www. epa.gov/ttn/uatw/urban/urbanpg.html
3 National Research Council. 1983. Risk assessment in the federal government. Managing
the process. National Academy Press, Washington, DC.
4 U.S. EPA. 1997. Cumulative Risk Assessment Guidance - Phase I: Planning and Scoping.
Memo from Administrator Browner, July 3, 1997. Available on-line at
http ://www. epa. gov/ordntrnt/ORD/spc/cumulrsk.htm.
5 U.S. EPA. 1998. Peer Review. Science Policy Council Handbook. Office of Research and
Development, Washington, DC. EPA 100-B-98-001. Available on-line at
http://www.epa.gov/ordntrnt/ORD/spc/prhandbk.pdf
6 U.S. EPA. 1998. Draft Integrated Urban Air Toxics Strategy to Comply with Section
112(k), 112(c)(3) and section 202(1) of the Clean Air Act. 63 FR 49240.
7 U.S. EPA. 1999. Analysis of the Impacts of Control Programs on Motor Vehicle Toxic
Emissions and Exposure Nationwide. Prepared for U. S. EPA, Office of Transportation
and Air Quality, by Sierra Research, Inc., and Radian International Corporation/Eastern
Research Group, Report No. EPA420-R-99-029/030.
8 Woodruff, T.J., D.A. Axelrad, J. Caldwell, R. Morello-Frosch, and A. Rosenbaum. 1998.
Public health implications of 1990 air toxics concentrations across the United States. Env.
Health. Pers. 106(5): 245-251. Available on-line at
www. epa. gov/CumulativeExposure/CEPpapers/paperW ACMR.html.
9 U.S. EPA. 1999. Final Assessment of Motor Vehicle Toxic Emissions and Exposure in
Urban Areas and Nationwide. Office of Transportation and Air Quality. Available on-line at
http://www.epa.gov/orcdizux/toxics.htm.
10 U.S. EPA. 1992. Guidelines for Exposure Assessment. National Center for
Environmental Assessment, Washington, DC. EPA/600Z-92/001, or FR 57: 22888 - 22938.
Available on-line at http://www.epa.gov/nceawwwl/pdfs/guidline.pdf
11 Rosenbaum, A.S.; Ligocki, M.P.; Wei, Y.H. 1998. Pages 5-9 in "Modeling Cumulative
Outdoor Concentrations of Hazardous Air Pollutants, Volume 1: Text"; SYSAPP-99-96/33r2,
Prepared for U.S. Environmental Protection Agency, Office of Policy, Planning and
Evaluation, by Systems Applications International, Inc., San Rafael, CA. 1998.
72 U.S. EPA. 1998. "National Air Pollutant Emission Trends Procedures Document, 1900-
1996." Office of Air Quality Planning and Standards. May 1998. EPA-454/R-98-008.
13 U.S. EPA. 2000. Control of Air Pollution From New Motor Vehicles: Heavy-Duty
130
-------
Engine and Vehicle Standards; Highway Diesel Fuel Sulfur Control Requirements; Proposed
Rules. 65 FR 35429-35478.
14 U.S. EPA. 1999. Analysis of the Impacts of Control Programs on Motor Vehicle Toxics
Emissions and Exposure in Urban Areas and Nationwide. Prepared for U. S. EPA, Office of
Transportation and Air Quality, by Sierra Research, Inc., and Radian International
Corporation/Eastern Research Group. Report No. EPA 420 BR-99-029/030
15 Code of Federal Regulations, Title 40 (Protection of the Environment) Appendix W to
Part 51-Guideline on Air Quality Models.
7<5Rosenbaum, A.S., D.A. Axelrad, T.J. Woodruff, Y.H. Wei, M.P. Ligocki, and J.P.
Cohen, 1999. National estimates of outdoor air toxics concentrations. Journal of the Air
& Waste Management Association, Vol.49, pp.1138-1152.
77U.S. EPA, 2000. User's Guide for the Assessment System for Population Exposure
Nationwide (ASPEN, Version 1.1) Model. EPA-454/R-00-017. U.S. Environmental
Protection Agency, Research Triangle Park, NC, 108pp.
18 McCurdy, T., Glen, G, Smith, L., and Lakkadi, Y. 2000. The National Exposure
Research Laboratory's Consolidated Human Activity Database. J. Exp. Anal. Environ.
Epidem. (2000) 10, 566-578.
19 U.S. EPA. 2000. Development of Microenvironmental Factors for the HAPEM4 in
Support of the National Air Toxics Assessment (NAT A), External Review Draft. Prepared
for the U.S. EPA, Office of Air Quality Planning and Standards, by ICF Consulting, Inc., with
TRJ Environmental, Inc.
20 U.S. EPA. 1986. Guidelines for Carcinogen Risk Assessment. 51 FR 33992-34003.
27 US EPA. 1999. Proposed Guidelines for Carcinogen Risk Assessment. NCEA-F-0644,
available on-line at http://www.epa.gov/ncea/raf/car2sab/preamble.pdf
22 U.S. EPA. 1986. Guidelines for Mutagenicity Risk Assessment. 51 FR 34006-34012.
23 U.S. EPA. 1991. Guidelines for Developmental Toxicity Risk Assessment. 56 FR 63798-
63826.
24 U.S. EPA. 1998. Guidelines for Neurotoxicity Risk Assessment; Notice 60. 63 FR
26926-26954. Available on-line at http://www.epa.gov/ncea/nurotox.htm.
25 U.S. EPA. 1996. Guidelines for Reproductive Toxicity Risk Assessment. National
Center for Environmental Assessment. EPA/630/R-96/009. Available on-line at
http://www.epa.gov/ORD/WebPubs/repro/.
26 U.S. EPA. 1994. Methods for Derivation of Inhalation Reference Concentrations and
Application of Inhalation Dosimetry. Office of Research and Development, Washington, DC.
131
-------
EPA/600/8-90/066F.
27 U.S. EPA. 1995. Policy for Risk Characterization at the U.S. Environmental Protection
Agency. Memo from Carol Browner, available on-line at
http://www.epa.gov/ordntrnt/ORD/spc/rcpolicy.htm.
28 U.S. EPA. 1995. Guidance for Risk Characterization . Science Policy Council. Available
on-line at http://www.epa.gov/ordntrnt/ORD/spc/rcguide.htm.
29 U.S. EPA. 1986. Guidelines for the Health Risk Assessment of Chemical Mixtures. 52 FR
34014-34025.
30 U.S. EPA. 1999. Guidance for Conducting Health Risk Assessment of Chemical
Mixtures. External Scientific Peer Review Draft, September 1999,NCEA-C-0148. Available
on-line at http://www.epa.gov/ncea/mixtures.htm.
31 Gifford, F.A. and Hanna, S.R. 1973. Modeling urban air pollution. Atmospheric
Environment. (7): 131-136.
32 Hanna, S.A., Briggs, G.A., and Hosker, R.P. 1982. Handbook on Atmospheric
Diffusion. Available as DE82002045 (DOE/TIC-11223) from the National Technical
Information Service, U.S. Department of Commerce, Springfield, VA, 22161, 108 pages.
33 Calder, K.L., (1971): A climatological model for multiple source urban air pollution.
Paper presented at First Meeting of the NATO/CCMS Panel on Modeling. Paper
published in Appendix D of User's Guide for the Climatological Dispersion Model.
EPA-R4-73-024. Office of Research and Development, U.S. Environmental Protection
Agency, Research Triangle Park, NC 27711, pages 73-105.
34 Martin, D.O., (1971): An urban diffusion model for estimating long term average
values of air quality. J. of the Air Pollution Control Association. Vol 21(1): 16-19.
35 Irwin, J.S. and Brown, T.M., (1985): A sensitivity analysis of the treatment of area
sources by the Climatological Dispersion Model. Journal of the Air Pollution Control
Association. Vol 35(4):39-364.
36 Pooler, F., (1961): A prediction model of mean urban pollution for use with standard wind
roses. International Journal of Air and Water Pollution. Vol. 4(3/4): 199-211.
37 Turner, D.B. and Irwin, J.S., (1983): Comparison of sulfur dioxide estimates from the
model RAM with St. Louis measurements. Air Pollution Modeling and Its Application
II, (Edited by C. De Wispelaere), Plenum Press, pages 695-707.
38 Turner, D.B., Zimmerman, J.R., and Busse, A.D., (1971): An evaluation of some
climatological models. Paper presented at Third Meeting of the NATO/CCMS Panel on
Modeling. Paper published in Appendix E of User's Guide for the Climatological Dispersion
Model. EPA-R4-73-024. Office of Research and Development, U.S. Environmental
132
-------
Protection Agency, Research Triangle Park, NC 27711, pages 107-131.
39 U.S. EPA, (1999): A simplified approach for estimating secondary production of
hazardous air pollutants (HAPS) using the OZIPR model. EPA-454R-99-054. U.S.
Environmental Protection Agency, Research Triangle Park, NC 27711, 86 pages.
40 U.S. EPA. 1997. Mercury Study Report to Congress. Office of Air Quality Planning and
Standards and Office of Research and Development. U.S. Environmental Protection Agency.
Washington, D.C. December 1997. EPA-452/R-97-007.
41 ATSDR. 1998. Toxicological Profile for Chlorinated Dibenzo-p-Dioxins (Update). U.S.
Department of Health and Human Services. Agency for Toxic Substances and Disease
Registry. December 1998.
42 U.S. EPA. 2000. Review of EPA's Health Assessment Document for Diesel Exhaust.
Science Advisory Board. Clean Air Scientific Advisory Committee. December 2000. EPA
600/8-90/057E
43 U.S. EPA. 2000. Draft Health Assessment Document for Diesel Exhaust. July 2000.
44 Federal Register, 2001. USEPA; 40 CFR Parts 69, 80, and 86; [AMS-FRL-6923-7] RIN
2060-AI69; "Control of Air Pollution from New Motor Vehicles: Heavy-Duty Engine and
Vehicle Standards and Highway Diesel Fuel Sulfur Control Requirements";
http ://www. epa. gov/fedrgstr/EP A-AIR/2001 /January/Day-18/aO 1 a. htm
45 U.S. EPA. 1996. Air Quality Criteria for Paniculate Matter. U.S. Environmental
Protection Agency, Office of Research and Development, Washington, D.C. 20460.
EPA/600/P-95/001a-c.
¥<5Krewski D, Burnett RT, Goldbert MS, Hoover K. Siemiatycki J, Jerrett M,
Abrahamowicz M, and White WH. 2000. Reanalysis of the Harvard Six Cities Study and
the American Cancer Society Study of Parti culate Air Pollution and Mortality. Special
Report to the Health Effects Institute, Cambridge MA, July 2000.
133
-------
Figures
-------
Figure 1 -1. Overview of National Air Toxics Assessment (NATA) Activities.
Data Inputs
and Research
Areas
Screening-level,
National-scale
Assessments
(Initial Assessment)
Refined Local-scale
Assessments,
Evaluation of Initial
Assessment, and
Related Activities
- State/local/industry
emissions data
- Emission Factors
- Emission
Characterization
- Meteorological and
Topographical Data
- "Background" Air Quality
- Atmospheric Chemistry
and Physics
- Human Activity Data
(CHAD)
- Microenvironment
Factors
- Census Data
- Hazard ID and Dose-
Response Assessments
- Ecological
Assessments
- Agency Risk Guidelines
1
^
Emission
Inventories
(1996 NTI)
1
i
i
Air Dispersion
Modeling
— p
(ASPEN and
EMS-HAP)
t '
Priority Air
Toxics
s \
Refined Local-
Scale
Inventories
V J
1
s \
Expanded Air
"* *" Quality Monitoring
V J
i
Inhal
Expo
(HAP
i
i
'
ation
sure
sling
EM4)
k
r
' Indoor, ^
Microenvironmenta
1, and Personal
Exposure
^ Monitoring ^
r
Risk Assessment/
Characterization
^
(Nationwide and
Urban areas)
i
f ^\
Refined & Local-Scale Modeling
Multimedia Models
Exposure Models
Air Quality Models
V J
Source
Measurement
Programs
Priority Setting
(e.g., Federal and
State/I ocal/Tribal
Regulatory Programs,
Initiatives)
Progress Toward
Statutory Risk
Reduction Goals
(e.g., 75% cancer
incidence reduction in
urban areas)
Progress Toward
Future Risk-Based
Goals
Public Right to
Know
(e.g., Trends Report)
CAA Section 812
Prospective
Benefits
Assessment
-------
Figure 1-2. Links Between NATA Activities and Other Program Elements.
Regulatory
Programs
Priorities/
Progress from
Current NATA
Activities
Risk-based
Initiatives and
Special
Studies
MACT/GACT
Standards
Urban Area
Source
Standards
Incineration
Standards
c
Mobile Source
Studies/
1 Standards
V
otner •>
Regulatory
Programs (e.g.,
Ozone, PM
NAAQS, non-air
Urban Strategy:
Local Studies/
Initiatives
Great Waters
Studies
Mercury Studies
PBT Initiatives
Indoor
Environments
Program
Residual Risk
Assessments/
Standards
Electric Utilities
Study and
Regulatory
Determination
Ongoing
Reductions in Air
Toxics Emissions
and Risk
Ongoing Input
to Future
NATA
Activities
Improved Risk
Characterizations
135
-------
Figure 2-1. MAS Risk Assessment/Risk Management Paradigm
Research
Laboratory and
field observations
Information on
extrapolation methods
Field measurements,
characterization of
populations
Risk
Assessment
Toxicity Assessment
Hazard Identification and
Dose-Response
Assessment
Research needs identified
risk assessment process
Exposure Assessment
Emissions Characterization
Risk
Characterization
Risk
Management
Development of
regulatory options
Evaluation of public
health, economic, social,
political consequences of
regulatory options
Agency decisions
and actions
Developed from: MAS, 1983
136
-------
Figure 2-2. National-Scale Air Toxics Health Assessment: Conceptual Model.
Heavy lines indicate dimensions/elements included in the Initial National Scale Assessment;
Light lines indicate dimensions/elements that may be included in future assessments
Sources
Stressors
Major
Industrial
Small "area"
sources
3.
Mobile (on-
and off-road)
±
Extrinsic
"background" in air
I
Indoor air
sources
Extrinsic "background"
in other media
33 Priority Air Toxics
Subset of PBTs
Other Air Toxics
Subset of PBTs
Pathways/
Media
Routes
Subpopulations
Outdoor a i
I
1
r
Indoor air Water Food SoN
1
1
Inhalation Ingestion Dermal
1
Endpoints
(Specific non-
cancer target organ
endpoints shown for
example purposes)
Measures
Pollutant-specific and
cumulative (e.g., by
cancer type, weight of
evidence; by target-
organ-specific hazard
index), by State by county
(also all counties andall
urban counties)
1 1 1
....... African
>anic White .
American
General
Population
Asian American
1 1 1
1
I
Young
Children
1 1
Cancers R
(leukemia, lung, others) ^ "
Blood (including
marrow & spleen)
Adoles-
cents
:
.
CMS
Adults
1.
.
Liver &
kidney
Elderly
Cardio-
vascular
Other health
effects
Cancers
(leukemia, lung, others)
Respiratory
1
Distribution of
high-end cancer pop
risk estimates
Blood (including
marrow & spleen)
1
Possible Carcinogens
Probable Carcinogens
Known Carcinogens
Estimated percent of
>ulation within specified
cancer risk ranges
Estimated
number of
cancer cases
-
Cardiovascular Hazard Index
Liver and Kidney Hazard Index
CMS Hazard Index
Blood Hazard Index
Respiratory System Hazard Index
Distribution of
estimated
index values
Estimated percent of
population within specified
ranges of index values
-------
Figure 2-3. Diagram of the HAPEM4 40 Cohort Groups (2x4x5=40).
Gender (2):
Race (4):
Male
Age (5):
African
American
0-5 yrs
65+ yrs
Example Cohort: Female; Hispanic; 18-64yrs
138
-------
Figure 3-1.1996 NTI State and Local Agency Data Summary.
Gray - states who submitted air toxics inventory data
/• - states who submitted revisions
- local agencies who submitted revisions
-------
Figure 3-2. Example Demographic Groups, Microenvironment, and Activities.
[e.g., Groups by age,
gender.ethnic background]
[e.g., home, school, car, park]
[e.g., bicycling, sleeping, eating]
140
-------
Figure 3-3. Example of a Daily Exposure Scenario for a Cohort.
Cohort: Black Male 18-64
Daytype: Weekday
Daily Exposure Concentration:
XDay=(Xi +X2+X3+X4+X5+X6+X7)/24hrs)
Home Census Tract Concentration XH
Work Census Tract Concentration Xw
Activity/Location
nite 3am 6am 9am Noon 3
1.5
1.5
1.2
1.1
Sleeping @ Home
1.1
1.2
1.2
1.8
2.5
3.8
Dm 6pm 9
2.8
4.5
1.7
2.8
Jog in Park DJ^ef° Work in Office P.me
a Work Home
— "v — ~^-\ i-^ — ~~~-^^~~ — ~^- -^
Eat In
resturant
,
Dm Midr
1.5
1.8
Sleeping @ Home
Xi=(XH(0-3)* 3hrs+ XH(3.6)* 1 .5hrs) * MEHome
1 -5nrs+ XH(&9)* O.Shrs) * MEpark
X4=(XW(6.9)*1 -Shrs-1- \o-i2)* 3hrs)+
Xw(l2-3)* 3hrS+ \(3-e* 2hrS)) * MEOffice)
X5=(Xw(^6)*1hr+Xw(&9)*1hr)*MECar
X6=(XH(6.9)*2hrs)*MERest
X7=(XH(9.12,* 3hrs) * MEHome
-------
Figure 4-1. Summary of 1996 NTI Emissions for 33 Air Toxics by Source Sector.
29%
9%
40%
D Major
• Area & Other
D Non-Road
a On-Road
22%
Figure 4-2. Summary of 1996 NTI Emissions of
33 Air Toxics by Urban and Rural Designations.
Emissions
(Million
tons/year)
Major
Area &
Other
On-
Road
Non-
Road
142
-------
Figure 4-3. 1996 NTI - Aldehydes Emission Densities.
Tons/Year/Sq Mile
0.00032 - 0.04731
0.04731-0.10288
0.10288-0.17952
0.17952-44.41798
Figure 4-4. 1996 NTI - Metals Emission Densities.
Tons/Year/Sq Mile
,0-0.00007
0.00007-0.00034
0.00034-0.00237
0.00237-0.70105
143
-------
Figure 4-5. 1996 NTI -Halides Emission Densities.
Tons/Year/Sq Mile
0-0.01
0.01 - 0.02
0.02 - 0.06
0.06 - 9.28
Figure 4-6. 1996 NTI -POM and Hydrocarbons Emission Densities.
Tons/Year/Sq Mle
0-0.05
0.05 - 0.09
0.09-0.17
0.17-9.1
144
-------
Figure 4-7. Estimated Annual Average Concentrations (ug/m3) for Benzene.
Arizona -
Arkansas -
California -
Colorado -
Connecticut -
Delaware-
Washington DC-
Florida -
Georgia -
Idaho -
Illinois -
Indiana -
lowa-
Kansas -
Kentucky -
Lousiana -
Maine -
Maryland -
Massachusetts -
Michigan -
Minnesota -
Mississippi -
Missouri -
Montana -
Nebraska -
Nevada -
New Hampshire -
New Jersey -
New Mexico -
New York
North Carolina -
North Dakota -
Ohio-
Oklahoma -
Oregon -
Pennsylvania -
Rhode Island -
South Carolina -
South Dakota -
Tennessee -
Texas -
Utah-
Vermont-
Virginia-
Washington-
West Virginia-
Wisconsin -
Wyoming -
Puerto Rico -
Virgin Islands -
National -
1 1 ! v™
\ ' \ \ • KC>
III: : 25th 1 | 1 75th
_ ' Median
1 1 1 L- !
Ill
1 1 1
1 1 1
III:
II 1
III
1 1 1
III
III
III
II 1
III
III :
1 1 1
II 1 :
1 1 1
II I
1 1 1 :
II 1 :
ill
III :
III
III
1 1 1
II 1 :
1 1 1
l|l
ill
III :
1 1 1
II 1
III :
III :
ill:
III
III:
III
III
ill
ill
III
^111: :
cfel 1 1 1 :
°-| 1 1 1
Estimated Annual Average Ambient Concentrations
145
-------
Rgure 4-8. Percent Contribution to the Statewide Annual Average Ambient Benzene Concentration Estimates
Alabama
Arizona
Arkansas
California
Colorado
Connecticut
Delaware
Washington DC
Florida
Georgia -
Idaho -
Illinois -
Indiana -
Iowa-
Kansas -\
Kentucky -\
Lousiana -\
Maine -
Maryland -
Massachusetts -
Michigan -
Minnesota -
Mississippi -
Missouri -
Montana -
Nebraska -
Nevada -
New Hampshire
New Jersey -
New Mexico -
New York
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
Pennsylvania
Rhode Island
South Carolina
South Dakota
Tennessee
Texas
Utah
Vermont
Virginia
Washington -
West Virginia-
Wisconsin-
Wyoming -
Puerto Rico -
Virgin Islands -
National -
1
50 100 0 50 100 0 50 100 0 50 100 0 50 100
Major Area and Other Onroad Nonroad Background
Percent Contribution to the Statewide Annual Average Ambient Concentration Estimates
146
-------
Figure 4-9. Comparison of Annual Average Model Concentrations for 10 Pollutants.
Distribution of Modeled Concentration from ASPEN, 1996
10 -
Arsenic Benzene Butadiene Formaldehyde Lead Mercury Meth-Chl. Nickel 7-PAH Vinyl-Chl.
-------
Figure 4-10. Annual Average Concentrations for Urban and Rural Census Tracts.
Distributions of Modeled Concentrations from ASPEN, 1996 (rural white, urban black)
10°-
Arsenic Benzene Butadiene Formaldehyde Lead Mercury Meth-Chl. Nickel 7-PAH Vinyl-Chl.
-------
Figure 4-11. Relative contribution of major, area and mobile sources.
Concentrations scaled by C (HAP): min - white, max - black
Ill 1111 "•'
-------
Figure 4-12. Model-to-monitor scatter plot for benzene.
Model Estimate
(ug/m3)
9:
B-.
7:
6:
5:
4:
3:
2-.
1:
jar"
1 2 5
Monitor Average (ug/rn3)
Note: Most points fall within the factor of two wedge, and none are far outside the wedge.
150
-------
Figure 4-13. Ratio box plot showing distribution of model/monitor
ratios for each pollutant.
Model/Monitor
Ratio
2-
1-
1/2-
1/4-
1/8-
1/16-
1/32-
n
0
jijfa ^r
* /
r-.
_
_
i
>JB^
/ ,
p
I
j— ,
' — '
I
i?d
_
i
\
? iflr *zr
X
?
-
i i
fj >^^
^
Note: The bottom of each box is the 25th percentile, the top is the 75th percentile, and the horizontal
line in the middle is the median.
151
-------
Figure 4-14. POM (Total) Exposure Concentration Distribution Among Cohorts in an Urban NY Census
Cohort (Tract Pop)
0.0215
0.0220 0.0225
POM Exposure Concentration (ug/m3)
0.0230
0.0235
-------
Figure 4-15. Exposure Results - Summary Table Example (Partial Table)
K^
L*J
1996 Modeled Exposure Concentrations for Acetaldehyde (CAS#75070)
EPA strongly cautions that these modeling results should not be used to draw conclusions about local concentrations or risk. The results are most meaningful when viewed at the state or national level; for smaller areas, the modeling becomes
less certain. In addition, these results represent conditions in 7996 rather than current conditions.
• The exposure estimates presented below are representative of midrange estimates of population exposures. Due to a number of factors, some individuals may have substantially higher or lower exposures. It is important to note that the model, as
applied on the national scale, is not designed to quantify these extreme values of individual exposures.
• Note that for certain chemicals, exposure pathways other than inhalation as well as indoor sources of air toxics may contribute substantially to total exposures of concern. This assessment does not address these other routes of exposure (i.e.,ingestion
or dermal) or inhalation exposure resulting from indoor sources.
• The emissions used in this assessment do not reflect potentially significant emission reductions that have taken effect since 1996, including those from: 1) mobile source regulations which are being phased in over time; 2) many of the air toxics
regulations EPA has issued for major industrial sources; 3) State or industry initiatives; and 4) any facility closures.
• Simplified modeling assumptions may introduce significant uncertainties into each component of the assessment. See the full discussion of these limitations.
• Because of these uncertainties, EPA will not use the results of this assessment to determine source-specific contributions or to set regulatory requirements. However, EPA expects to use these results to inform decisions about the priorities of the air
toxics program as well as to guide the collection of additional data that could lead to regulatory decisions.
Draft for Scientific Peer Review
State
National
National
National
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
County
All
All Rural Counties
All Urban Counties
State Total
State Rural Counties
State Urban Counties
Autauga County
Baldwin County
Barbour County
Bibb County
Blount County
Bullock County
Butler County
Calhoun County
Chambers County
Cherokee County
Chilton County
Choctaw County
Clarke County
Clay County
Cleburne County
Coffee County
Colbert County
Conecuh County
Coosa County
Covington County
Crenshaw County
FIPS
99999
99998
99998
01000
01000
01000
01001
01003
01005
01007
01009
01011
01013
01015
01017
01019
01021
01023
01025
01027
01029
01031
01033
01035
01037
01039
01041
Urban
or
Rural
-
R
U
-
R
U
U
U
U
R
U
R
R
R
U
R
R
R
R
R
R
U
U
R
R
R
R
Estimated Annual Exposure Concentrations (^g/m3) for Acetaldehyde (CAS#75070)
Percentile Distribution of Exposure Concentrations Across Census Tracts
5th
3.09E-02
1.15E-02
6.69E-02
7.37E-02
5.98E-02
1.43E-01
1.26E-01
1 .54E-01
5.67E-02
1.46E-01
1 .58E-01
8.40E-02
5.08E-02
2.17E-01
1.10E-01
1.62E-01
1.10E-01
7.88E-02
5.93E-02
7.78E-02
1.05E-01
8.71 E-02
1 .50E-01
4.85E-02
9.49E-02
5.76E-02
7.19E-02
10th
6.62E-02
2.72E-02
1.37E-01
9.38E-02
6.92E-02
1.87E-01
1 .39E-01
1 .68E-01
6.65E-02
1.51E-01
1 .64E-01
8.94E-02
5.30E-02
2.28E-01
1.15E-01
1.62E-01
1.11E-01
7.93E-02
6.06E-02
7.84E-02
1.10E-01
9.89E-02
1.51E-01
5.00E-02
9.53E-02
5.99E-02
7.76E-02
25th
1.67E-01
5.87E-02
3.07E-01
1.59E-01
9.16E-02
2.79E-01
1.62E-01
1 .86E-01
8.22E-02
1.66E-01
1 .86E-01
1 .06E-01
6.60E-02
2.65E-01
1.23E-01
1.62E-01
1.18E-01
8.07E-02
6.34E-02
8.00E-02
1.23E-01
1.18E-01
1.79E-01
5.41 E-02
9.67E-02
6.58E-02
9.28E-02
Median
4.43E-01
1.03E-01
5.91 E-01
2.97E-01
1.31 E-01
4.56E-01
2.90E-01
2.16E-01
1.01 E-01
2.28E-01
2.32E-01
1.19E-01
6.85E-02
3.31 E-01
1.78E-01
1.70E-01
1.39E-01
8.31 E-02
6.75E-02
8.31 E-02
1.31 E-01
1.31 E-01
2.95E-01
5.89E-02
9.89E-02
8.21 E-02
1.17E-01
Average
6.56E-01
1.58E-01
7.66E-01
3.93E-01
1.59E-01
5.04E-01
2.47E-01
2.26E-01
1.19E-01
2.40E-01
2.20E-01
1.19E-01
1.02E-01
3.05E-01
1.78E-01
1.76E-01
1.44E-01
8.41 E-02
6.98E-02
8.85E-02
1.28E-01
1.51 E-01
2.42E-01
6.13E-02
1.04E-01
8.68E-02
1.14E-01
75th
8.44E-01
1.76E-01
9.56E-01
5.32E-01
1.94E-01
6.25E-01
3.09E-01
2.66E-01
1.80E-01
3.04E-01
2.48E-01
1 .34E-01
1.39E-01
3.73E-01
2.02E-01
1.87E-01
1.60E-01
8.66E-02
7.52E-02
9.14E-02
1.35E-01
1.94E-01
3.55E-01
6.45E-02
1.09E-01
1.16E-01
1.31 E-01
90th
1.34E+00
2.73E-01
1 .48E+00
8.96E-01
2.59E-01
9.81 E-01
3.73E-01
2.99E-01
1.85E-01
3.43E-01
2.49E-01
1.52E-01
1.66E-01
4.09E-01
2.49E-01
2.03E-01
1.70E-01
8.98E-02
9.12E-02
1.02E-01
1.42E-01
2.00E-01
3.72E-01
7.61 E-02
1.15E-01
1.28E-01
1.47E-01
95th
1.82E+00
3.43E-01
2.04E+00
1 .02E+00
3.31 E-01
1 .07E+00
3.74E-01
3.00E-01
1.89E-01
3.56E-01
2.50E-01
1 .58E-01
2.00E-01
4.15E-01
2.60E-01
2.08E-01
1.72E-01
9.09E-02
9.84E-02
1 .06E-01
1.44E-01
2.04E-01
3.74E-01
8. 11 E-02
1.17E-01
1.31 E-01
1 .54E-01
Contribution to Average from ...
Major
5.23E-03
6.02E-03
5.05E-03
4.51 E-03
2.28E-03
5.57E-03
1 .66E-02
3.36E-03
1.38E-04
1.43E-07
7.51 E-06
4.75E-05
O.OOE+00
2.27E-06
4.39E-07
9.18E-04
7.54E-04
6.39E-04
4.67E-03
2.87E-04
2.16E-05
5.50E-06
1 .38E-03
6.87E-05
7.62E-04
O.OOE+00
6.52E-05
Area and
Other
5.39E-02
3.97E-02
5.70E-02
4.53E-02
3.24E-02
5.15E-02
4.52E-02
3.60E-02
4.97E-02
4.71 E-02
2.71 E-02
2.85E-02
4.66E-02
3.39E-02
6.00E-02
3.36E-02
3.61 E-02
4.41 E-02
3.15E-02
2.93E-02
3.16E-02
3.35E-02
2.74E-02
3.22E-02
3.67E-02
3.32E-02
2.65E-02
Onroad
Mobile
3.89E-01
8.37E-02
4.56E-01
2.89E-01
1.01 E-01
3.78E-01
1 .48E-01
1.18E-01
4.21 E-02
1.66E-01
1 .68E-01
3.49E-02
4.45E-02
2.28E-01
9.89E-02
1.24E-01
8.80E-02
3.33E-02
2.61 E-02
5.07E-02
8.32E-02
7.45E-02
1.87E-01
2.44E-02
5.56E-02
4.08E-02
4.44E-02
Nonroad
Mobile
2.08E-01
2.86E-02
2.48E-01
5.42E-02
2.40E-02
6.86E-02
3.68E-02
6.88E-02
2.71 E-02
2.70E-02
2.53E-02
5.51 E-02
1.07E-02
4.24E-02
1 .94E-02
1 .72E-02
1.94E-02
6.07E-03
7.60E-03
8.21 E-03
1.31 E-02
4.27E-02
2.60E-02
4.68E-03
1.04E-02
1.28E-02
4.25E-02
Estimated
Background
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
O.OOE+00
Page 1 of 1
Release Version: August, 2000
-------
Figure 4-16. 1996 Modeled Exposure Concentrations
Acetaldehyde - Statewide Concentration Distribution Estimates
National
Alabama
Arizona
Arkansas
California
Colorado
Connecticut
Delaware
District Of Columbia
Florida
Georgia
Idaho
Illinois
Indiana
Iowa
Kansas
Kentucky
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
Nevada
New Hampshire
New Jersey
New Mexico
New York
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
Pennsylvania
Puerto Rico
Rhode Island
South Carolina
South Dakota
Tennessee
Texas
Utah
Vermont
Virgin Islands
Virginia
Washington
West Virginia
Wisconsin
Wyoming
O.C
— i — i 1_
1 1
25th median 751
h
* Upper bound cancer risk assuming lifetime exposure
s^_
^
^
r~~
^~
(
r
C
\
\
\
^
x
/
Xs
<
x
^
^
xj
^
r
5
V,
X
V
/
\
^
1
^
x
^
s
(
-.
'v
^^
"-
^v
1X1 x,
\
1
1
r
(
c
<^
^
. ^^
• ^7
;2
,_)
•
c
C
^7
•~ —
\
^^^
3
H
2
>
1
1
1
1
1
1 A
1
>
E
\
^
1
t.
^
\
. — '
01 0.010
c
^
* —
"7
\
s
^
/r~
\ •-
^^^
/ 1 v
^Ji
—_-•''
i
i
i
1
1
i
i i
i
i
t
X
i i
i
i i
i.
1
1
f
c.
'
^*
^
N
H~~i
i^ k <. i /
Cg£ ^t-O *-
•^ i t
^
^~
^
--^>O^^'f
;^o
s^ ^
' Vx^
;.. )
-..%
s
^
~
1
1
1
1
-J \
i
i
i
i
H
:
i i
i i
i
i
\ \
r^ri
i
i i
L"
i
1
TTT
T^
I
1
i
'v
Mr
^
\
\
\
\
$
i
i
i
I*'
i
i
A
v
L
\ 1
' """x"
«T) v
^<
_
^ \^
^\~^
-' l^ \
-; — s
^ )
• -^
T
-^
i
^
^^
<^S
— • —
_^
V
— ^
— — .
^
7
1
.2 In
, ro ~"
d c
••" O
i
(
^
^
• — .
/j
^^~
• — .
• — i
7
r
^
^
^
/
•^«.
s
I
•^
•^
~~
s
s
s
-^.
/
V
1X1
/
2
ro
O
c
o
5
rc
c
T—
s
>
5
'
X,
^
^
. X.
V
l
ro
CT
LU
'E
m
O
a
ro
ro
:
0.100 1.000 10.000
"w
UL
o
c
O
c
0
2
ro
c
o
o
100
Inhalation Exposure Concentration (jiig/m )
154
-------
Figure 4-17. 1996 Modeled Exposure Concentrations
Acetaldehyde - Statewide Source Sector Contribution Estimates
National
Alabama
Arizona
Arkansas
California
Colorado
Connecticut
Delaware
District Of Columbia
Florida
Georgia
Idaho
Illinois
Indiana
Iowa
Kansas
Kentucky
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
Nevada
New Hampshire
New Jersey
New Mexico
New York
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
Pennsylvania
Puerto Rico
Rhode Island
South Carolina
South Dakota
Tennessee
Texas
Utah
Vermont
Virgin Islands
Virginia
Washington
West Virginia
Wisconsin
Wyoming
P
i
a
i 1
a
^
/ ^
\^^
i \^ x
%
/**
C
>
N/v
( cX *•
\ g
/^~xx?
( (7 \x
\ i ^) &
xN\,
VA—
/
\/
a
a
a
i
a
CP
a
a
a
a
a
a
a
a
a
i
a
a
a
a
a
a
i i
a
i
a
a
a
a
a
a
i — i
j i /~-
a (c
a x
^^
a C ~— '
r~^X
p _
a^
o
a
a
a
p
i
a
i
£L
\X
T^S
> W ^
^^x_y
^4
Cl/
N
s/
1
1
1
1
1
1
1
i
i
1
1
1
1
1
1
1 1
i
(
5ZS
"n\>> )
\ \J;
— -^ \^Ms\
/\ \ I
r.'Xx^
i V v^1
Xyv"1^1^
"* r\ \
VI \
XX
\ \
Vv'l
1
o ) |
§
1
p
1
i
1
1
i
i
i
i
i
i
i
i
i
i
i
i
i
i 1
EP
i
EP
i 1
CP
i 1
CZI
i
CZI
CP /
^=1^^
1 \
C=II\
c=KX
SK^)
C2H
r^=^i
^=1
—f — i
j 1
CP
i — 1
CZI
CZI
1
i — 1
1
EP
CP
i 1
i 1
CP
i
i 1
i
i 1
a
CZI
CP
1=1
CZI
j '
'
j '
j — '
i
j — '
a
/N '
X/^
A\
^^
AP^X^
XX/o")
) ^\r^
^x^O7
\J>
>
]
/\
/^\^
,Q//\^
\x /
XX
\
\
0 100 0 100 0 100 0 100 0 10C
Major Area & Other Onroad Nonroad Background
Percent Contribution to Statewide Annual Average Exposure Concentration Estimates
155
-------
Figure 4-18. 1996 Modeled Median Exposure Concentrations for North Carolina
10
(Jl
ON
o
0)
u
o
o
0)
(A
X
LU
jo
TO
0.1
0.01
0.001
0.0001
0.00001
0.000001
0.0000001
0.00000001
QJ
|
^^^^
s
^-s^
\ \
s
\
•
\^
\ \
\ \
v)
\s
-A-
/:
t
\
\ \
V
-f— \
s
>
-^
z
Al
s
fe
,-r
\\
>s
s
\ V
fc^
s~\
c
w
?
s
\
-A-
"H~
V \\
\ \
V
\
-4-
2
V \
v SJ
\
>"
— t
— \
r\
*)
\\
\ \
} ^
S
-v^
\
\
<•
\
f—£
M-
•N \
\\
^^
^ --
^
^7L
a
! 1
|d
,£!S
j:
n
<
t<= Mnt
r
\ \
V \\
\ \
-\\'
>
Q -
_Q
'ro
U)
0
— -
<^\
S
<" ,
\ vJ
s\
>
-A—
— p
\
- \
\\
^\y
T\ '
1 \
)
"
A Median Exposure Concentration Greater than 1 in a Million Cancer Risk*
• Median Exposure Concentration Greater than 10 in a Million Cancer Ris
O Median Exposure Concentration Greater than RfC (Non-Cancer)
• Median Exposure Concentration Less than Shown Health Criteria
or No Health Criteria Considered
sk*
A
^\
2
r >
VL '
v\
A\
S
^>
y
\ \^
\ \
s \
\\
\v
-%-
^\
4^
, \
\ v
\\
^'
-9-
\ \
\ ^
\
~\
> \
)
«=
\ /
\ \
) \
)
\
\s
— ^s
r>
L^
\<
•
•
h^
w
\\
\\
|
cf
1 \\
2J
' r~\
^
\
\ <
\
\
\
s
\\
\s>
\
\
"> S
\)
st,
\
\
f
n
Vt
5
V Xx
>
•-
-
5
\ \
Vv 5
NS
> V
<"r
•*
A
\
3
i
•
s
rM
\N
— _
, f-^
- — >
-%-
f^\
\ \ >
\\
3
^
\ \
\ ]
V
=P
5? £ £ £
, -
ff 8
cf
c
v
cf
/ / / /
* « z
S if
* Upper bound cancer risk assuming lifetime exposure
Release Version: September 2000
-------
Figure 4-19. Example State Exposure Concentration Map
1996 Estimated County Median Exposure Concentration
Aeetaldehyde - NORTH CAROLINA Counties
Distribution of U.S. Inhalation Exposure Concentration
Cancer
Risk
> 1OO In a million
30-100 in a million
1 0—30 In a million
3—10 in a million
1 —3 in a mi! fen
< 1 in a mil Ron
County Median Exposure Concentration
( micrograms / cubic meter )
Source: U.S. EPA / GAQPS
NATA Natlona I—Sea IB Air Toxics Assessment
157
-------
(Jl
oo
1.0E+02
•5 1.0E+01
ns
O
o
1.0E-03
1 .OE-04
1,000
Figure 4-20. Benzene Exposure Variability within a County.
•* . ' • .
.-..: / „•;• :> .; ..B .."
***5>^*f-*-''% *•*•*.-: ••
"\/-v^;;yV.^''!••":
• .;• *«rf.r"*
. t
10,000
100,000
County Population
1,000,000
10,000,000
-------
Figure 5-1.1996 Risk Characterization
Distribution of lifetime cancer risk for the US population, based on 1996
exposure* to all source sectors and background combined.
Upper-Bound Lifetime Cancer Risk per Million
0.001 0.01 0.1 1 10
100
Coke Oven Emissions
1,1,2,2-Tetrachloroethane
* Results are based on inhalation exposure to outdoor sources only. Although these results assume continuous exposure to
1996 levels of air toxics over a lifetime, current and planned control programs are expected to substantially reduce these
exposures and associated cancer risk for some pollutants. See additional information on the following page.
159
-------
Figure 5-2.1996 Risk Characterization
Population whose 1996 exposure* exceeded set
cancer risk levels based on all source sectors and background.
0.01
0.1
Millions of People
1 10
100
1000
Arsenic
§ Benzene
0)
o
•§ Chromium
ra
^ Coke Oven Emissions
i Nickel
Vinyl Chloride
Ace (aldehyde
Acrylonitrile
Beryllium
1 ,3-Butadiene
Cadmium
Carbon Tetrachloride
Chloroform
1 ,3-Dichloropropene
Ethylene Dibromide
c
a) Ethylene Dichloride
|5 Ethylene Oxide
o
JJ Formaldehyde
'8
o Hexachlorobenzene
° Hydrazine
5 Lead
CL
Methylene Chloride
PCBs
POM (7-PAH)
Propylene Dichloride
Quinoline
1 ,1 ,2,2-Tetrachloroethane
Perchloroethylene
Trichloroethylene
III
| | t t-t-111- f---t-1-t--t1-ttt
^ hl - — t t
| > T 1'" h~~t-1-t~t~
'
I
-iUH^.U
"tttt \ \-rr
"rrrrf
t
-i-
:tttt:::::t:::;
.............. _j___j_^__
-tt-mi- —
.... h...t....t...f..t..m.f
i r 1
i tr
.:.:.:~~L.:.:
/
|
INI
-L-^
...
3
< -
X;
\
•~^
•^.
_h
tl'ZlJ
_L,
=
v
^
Vy
j
....
1
N^;
g
w
^^J
sx.^>
,-. .H^^.-,-. ........ -,-^f-, .-,-,
C%^J\JX^
^^g^
r5];;h^> i
!^
'
^
3
*
r>A
^^t
So^
•f -J-
s
s;
«
_y
s7
...
......
^C
\
f--j
\^
:^>
*i
-t
...t....l....
s
(
*^
~^,
-Ji
s
)
....
t
ft ^
X
^
c
-~^,
o
\
-— »,
^j..
X
^
X;
X
*^»,
•^^,
-/-
-^k.
•^,
>-~l
>
X
\^
_.J5i
***.
^
->
7
^
^
^
f
A
^sX
;^^
W:
r
*j — i
I l>100
ii-i-i-i-i-i;i;i;i-i-i-i-i-n>iuir
j
^
i
t
c
i
[
D
3
3
3—
0
J—
0
in 1 Million
i 1 Million
1 Million
» Results are based on inhalation exposure to outdoor sources only. Although these results assume continuous exposure to 1 996
levels of air toxics over a lifetime, current and planned control programs are expected to substantially reduce these exposures and
associated cancer risk for some pollutants. See additional information on the following page.
160
-------
Figure 5-3. 1996 Risk Characterization
Distribution of lifetime cancer risk for the US population,
based on 1996 exposure* to multiple carcinogens.
0.001
0.01
Upper-Bound Lifetime Cancer Risk per Million
0.1 1
10
100
Population Percentile
25 50 75 95 99
* Results are based on inhalation exposure to outdoor sources only. Although these results assume continuous exposure to
1996 levels of air toxics over a lifetime, current and planned control programs are expected to substantially reduce these
exposures and associated cancer risk for some pollutants. See additional information on the following page.
161
-------
Figure 5-4. 1996 Risk Characterization
Population whose 1996 exposure* exceeded
set risk levels of risk for carcinogens combined.
0.01
0.1
Millions of People
1 10
1000
> 100 in 1 Million
>10 in 1 Million
>1 in 1 Million
* Results are based on inhalation exposure to outdoor sources only. Although these results assume continuous exposure to 1996 levels
of air toxics over a lifetime, current and planned control programs are expected to substantially reduce these exposures and associated
cancer risk for some pollutants. See additional information on the following page.
162
-------
Figure 5-5. 1996 Risk Characterization
Distribution of non-cancer hazard quotient for the US population, based on
1996 exposure* to all source sectors and background combined.
0.00001
0.0001
Hazard Quotient (Exposure/RfC)
0.001 0.01 0.1
10
100
"c Ethylene Dichloride
Population Percentile
25 50 75 95 99
* Results are based on inhalation exposure to outdoor sources only. Although these results assume continuous exposure to
1996 levels of air toxics over a lifetime, current and planned control programs are expected to substantially reduce these
exposures and associated non-cancer risk for some pollutants. See additional information on the following page.
163
-------
Figure 5-6. 1996 Risk Characterization
Adult population whose 1996 exposure* exceeded set non-cancer
hazard quotient levels based on all source sectors and background combined.
0.01
0.1
Millions of Adults
1 10
100
1000
Benzene
Beryllium
Cadmium
Chloroform
Chromium
o~ 1,3-Dichloropropene
^ Ethylene Dibromide
.S1 Ethylene Dichloride
1
Carbon Tetrachloride
Hydrazine
Manganese
Propylene Dichloride
Vinyl Chloride
r-^Li
_
T
(
s
•*»*
&
ys^
c
FS
£7
:_y
-/
X
j'
OA;
^
^—
1
y
s
7^
_^
/
^
T
\
\
^
/
/o
X
fff
,^/
J
(
\
%
X
)
<
c
^
"• —
•• 1-
;
\ .XV--4 |
r^l
•
l:.:.:.:.:.:.:.:.:'!':.:!^.:.!^-.
D
S X
^
yjf 6
^—
1
'
^
.7
1
Jv
?
1
^
^
-^
J
-1
•s.
\
w
)
f(
5
\
>
rz
/^,
A
•"~«.
2
s^_
NJ
I
—
-J-
"cS
3
O
CL
W
"to
•5
1—
:HI
1 1 H
a > 10
a >1
a >o.
* Results are based on inhalation exposure to outdoor sources only. Although these results assume continuous exposure to 1996
levels of air toxics over a lifetime, current and planned control programs are expected to substantially reduce these exposures and
associated non-cancer risk for some pollutants. See additional information on the following page.
164
-------
Figure 5-7. 1996 Risk Characterization
Children population whose 1996 exposure* exceeded set non-cancer
hazard quotient levels based on all source sectors and background combined.
0.01
0.1
Millions of Children
1 10
100
1000
Benzene
Beryllium
Cadmium
Chloroform
Chromium
o
o
V 1,3-Dichloropropene
ID Ethylene Dibromide
1= Ethylene Dichloride
03
-e
g Ethylene Oxide
•^ Formaldehyde
Hexachlorobenzene
Lead
Mercury
Methylene Chloride
Nickel
Perchloroethylene
Trichloroethylene
Acetaldehyde
5- Acrolein
o
/v Acrylonitrile
^ Arsenic
|j 1,3-Butadiene
> Carbon Tetrachloride
Hydrazine
Manganese
Propylene Dichloride
Vinyl Chloride
L
D
*-
'H
"^
"7]
*"S
>
f
\
/
1
^
X
f
^F-
X;
J
V
\
/
';
(
t**^J
S
X
*•
^
W1
^
6
n<^
- -N,
'.,
X_ „
s
I7
\
rv~
-,
^
^_
X
x7
— >
^_
S
4
S'
S
J.
-3
^s
^
V
?,
c
%
)
/
\
^v
''^
*^_
V
J
zv
4 (
o
s
r*i Jj*
s *— *
"*s^^S
0
-re
|
V,
)
"t
^x.
X.
\
•. p-
s
J
/•
/
X
f
v
7
J
iffi
•x
s
Vi
;
X
X
?"
--
^
•s,
>
-
>
--
/
f'
•
f-*l
^^
1$^
iV^
^xS~
s
)
?v
i)
f
<
-~^
\
>
3
J
e
Q
o
T
c
y
E
73
2
o
\
s
••»
V
^
^
s,
^
r-
-s
^^J
s
IH
|...........................| H
I |H
Q > 10
Q >1
Q > 0.1
* Results are based on inhalation exposure to outdoor sources only. Although these results assume continuous exposure to 1996
levels of air toxics over a lifetime, current and planned control programs are expected to substantially reduce these exposures and
associated non-cancer risk for some pollutants. See additional information on the following page.
165
-------
Figure 5-8. 1996 Risk Characterization
Distribution of non-cancer target organ-specific hazard index (TOSHI) for
effects to the respiratory system, based on 1996 multiple-pollutant exposure*
to adults in the US population.
0.001
0.01
Hazard Quotient (Exposure/RfC)
0.1 1
10
100
- Mobile Non-Road
Population Percentile
5 25 50 75 95 99
* Results are based on inhalation exposure to outdoor sources only. Although these results assume continuous exposure to
1996 levels of air toxics over a lifetime, current and planned control programs are expected to substantially reduce these
exposures and associated non-cancer risk for some pollutants. See additional information on the following page.
166
-------
Figure 5-9. Illustration: Distribution of monitor-to-model ratios for stable gases,
developed from ratios for benzene and perchloroethylene.
1 0,000 Tr
.031 -
.023 -
£
la .015 -
a
.a
o
£ .008 -
Forecast: Stable Gas - Mon:Mod Ratio
ials Frequency Chart 82 Oi
,,,llll
I
nj
III
P
0.00 0.88 1.75
llllllllllll.ll,l,, ,
itliers
- 309
- 231.7
-T|
n
- 154.5 -0
3
- 77.25 -5
-------
Figure 5-12. Illustration: Distribution of ambient-to-personal concentration ratios for ozone, assumed to
apply for "typical" gaseous pollutants.
1 0,000 T
.087
.065
£
'•n .043
B
JD
O
(L .022
nnn
rials
Forecast: Gas - Pers:Amb Ratio
Frequency Chart 195
.UUU k
0.00
lllSlnllllln..,,,!...... ,
Outliers
866
649.5
T1
3
433 -g
n
=1
216.5 $
n
< U
3.75 7.50 11.25 15.00
Figure 5-13. Illustration: Distribution of ambient-to-personal concentration ratios for paniculate matter,
assumed to apply for "typical" paniculate pollutants.
1 0,000 Tr
.067 -
ncn
Probability
3 b b c
3 ->• CO C
3 -vl CO C
Forecast: Participate - Pers:Amb Ratio
ials Frequency Chart 209
I
IliBlliiBiliiLiiin.i.o..,
Outliers
- 668
cn-i
TI
3
11,1 .0
ro
3
167 &
n
^ , . ^
0.00 2.00 4.00 6.00 8.00
168
-------
Figure 5-14. Illustration: Uncertainty and variability surrounding the URE for benzene,
in terms of the ratio between the estimated URE and the "true" URE.
1 0,000 Tr
.056 -
.042 -
£
la .028 -
ra
J3
a
Ql -014 -
.000 -
als Frequency Chart 199Ou
SlIlBllllIlllllllIlllllllllll,.,,,,,....!,,....,
tilers
~ 555
- 416.2
-n
^
n
°77 ^ .O
CB
3
- 138.7 ^
n
fy «
0.00 2.00 4.00 6.00 8.00
Figure 5-15. Illustration: Uncertainty and variability surrounding a typical RfC, in terms
of the ratio between the estimated RfC and the "true" RfC.
Forecast: Noncarcinogen - Pers:Risk Ratio
10,000 Trials
.014 H
.010
JD
O
.007 -
.003 -
.000
0.25
0.94
1.63
2.31
0 Outliers
- 135
- 101.2
- 67.5
- 33.75
- 0
CD
3.00
169
-------
Figure 5-16. Cancer - risk ratio for stable gas.
10,000 Tria
.281 -
£
5
- 0
Figure 5-17. Cancer - risk ratio for reactive gas
10,000 Tria
.278 -
£
fl
.0
o
n -u'u
Forecast: Cancer - Risk Ratio, Reactive Gas
Is Frequency Chart 1
III
> 4
0.00 20.00 40.00 60.00 80.00
36 Outliers
- 2783
-n
5
j=
c
n
3
- 695.7 ^
- 0
Figure5-18. Cancer - risk ratio for particulate
10,000 Tria
.353 -
£
a
A
D
n -UBB •
000 -
Forecast: Cancer - Risk Ratio, Particulate
Is Frequency Chart
IllllM
t <
0.00 68.75 137.50 206.25 275.00
64 Outliers
- 3533
Tl
n
J3
C
n
a
- 883.2 «5
- 0
170
-------
Figure 5-19. Noncancer - risk ratio for stable gas
10,000 Tria
.134 -
JD
O
Q_ -034
Forecast: Noncancer - Risk Ratio, Stable Gas
Is Frequency Chart
0.00
I 111
11.25 22.50 33.75 45.00
60 Outliers
- 1340
- 670 JQ
CD
=1
- 335 $
- 0
Figure 5-20. Noncancer - risk ratio for reactive gas.
10,000 Tria
.126 -
£
fl
JQ
o
n -UJ/
Forecast: Noncancer - Risk Ratio, Reactive Gas
Is 1
|||
> 4
0.00 15.00 30.00 45.00 60.00
79 Outliers
- 1262
- 946.5
-n
5
- 631 JO
n
- 315.5 ^
- 0
Figure 5-21. Noncancer - risk ratio for particulate.
10,000 Tria
.110 -
Probability
Forecast: Noncancer - Risk Ratio, Particulate
Is 1
0.00
lllllllLlh
30.00 60.00 90.00 120.00
65 Outliers
- 1104
- 828
T1
n
- 552 .0
CD
3
- 276 ^
- 0
171
-------
Appendix A
Summary of July 2000 Peer Review of the Draft Document
'Planning and Scoping the Initial National-Scale Assessment:
An Element of the EPA National Air Toxics Program"
(Provided in electronic format on Appendices CD)
-------
Appendix A - Summary
Summary of July 2000 Peer Review of the Draft Document "Planning and Scoping the
Initial National-Scale Assessment: An Element of the EPA National Air Toxics Program"
In July 2000, six non-U.S. EPA scientists completed a peer review of the "Planning and Scoping
the Initial National-Scale Assessment: An Element of the EPA National Air Toxics Program'".
The reviewers were asked to focus their review on the main body of the planning and scoping
document as well as the supporting technical information. Reviewers were asked to consider the
appropriateness of approaches used to (1) process the State-derived National Toxics Inventory
for dispersion modeling, (2) estimate ambient concentrations using the Assessment System for
Population Exposure Nationwide (ASPEN) model, (3) estimate human inhalation exposures
using the Hazardous Air Pollutant Exposure Model version 4 (HAPEM4), and (4) estimate,
aggregate, and interpret associated cancer and non-cancer risks. Specific charge questions, the
reviewer's comments as well as the EPA's response to these comments are contained in this
appendix
-------
Appendix B
HAPEM4 Documentation
(Provided in electronic format on Appendices CD)
-------
Appendix B - Summary
HAPEM4 Documentation
This appendix contains detailed documentation for the HAPEM4 exposure model. Two items
are include "The HAPEM4 User's Guide" and the "Development of Microenvironmental Factors
for the HAPEM4 in Support of the National Air Toxics Assessment (NATA)"
The HAPEM4 User's Guide contains:
Chapter 1 - Provides a brief introduction to HAPEM4 modeling fundamentals including
a brief history of the development of HAPEM4.
Chapter 2 - Provides an overview of the various components of HAPEM4 and basic i
information needed to run the model.
Chapter 3 - Provides a description of the format, data, and options for each of HAPEM4
input files.
Chapter 4 - Provides a description of the format and data associated with each of
HAPEM4 output files.
Chapter 5 - Provides a description of the purpose and operations, inputs, and outputs,
including a brief description of the computer code, for each of HAPEM4 computer
programs.
Chapter 6 - References.
The Microenvironmental Factors Report contains detailed information on:
• Defining the HAPEM4 ME Factors
• Calculating HAPEM4 ME Concentrations and Estimation of Proximity Factors
• Estimating Ambient HAP Concentrations Using ASPEN
• Literature Search
• Grouping HAPs and Microenvironments
• A complete listing of the ME factors for the Urban 33 pollutants
• ME Factors for Diesel PM (Provided as an addendum to the Report)
• The Use of a Linear Model for the Initial NATA Assessment
-------
Appendix C
EMS-HAP User's Guide
(Provided in electronic format on Appendices CD)
-------
Appendix C - Summary
The Emissions Modeling System for Hazardous Air Pollutants (EMS-HAP) User Guide
Synopsis
1.1 What is EMS-HAP?
The Emissions Modeling System for Hazardous Air Pollutants (EMS-HAP) is a series of
computer programs that process emission inventory data for subsequent air quality modeling.
EMS-HAP accomplishes two goals.
1. It processes an emission inventory, such as the 1996 National Toxics Inventory, for use in
the Assessment System for Population Exposure Nationwide (ASPEN) dispersion model.
2. It allows you to estimate future emissions resulting from user-designed emission
reduction scenarios and growth.
To accomplish the first goal, EMS-HAP:
• quality assures point source inventory location and stack parameter data and
defaults missing or erroneous data where possible,
• groups individual pollutant species (e.g. lead oxide, lead chromate into lead
compounds),
• facilitates the selection of pollutants and pollutant groups for modeling,
• spatially allocates area and mobile source emissions from the county level to the
census tract level using surrogates such as industrial land or roadway miles,
• allocates aircraft emissions to airport locations,
• temporally allocates annual emission values to average hourly values based on the
type of source, and,
• produces emission files formatted for direct input into the ASPEN model.
To accomplish the second goal, EMS-HAP adjusts point, area and mobile emissions to account
for growth and emission reductions resulting from user-designed scenarios such as the
implementation of the Maximum Achievable Control Technology (MACT) standards.
The U.S. Environmental Protection Agency's Office of Air Quality Planning and Standards
(EPA/OAQPS), developed EMS-HAP to facilitate multiple runs of ASPEN and to analyze
emission reduction scenarios. ASPEN can be used to estimate annual average ambient air
quality concentrations of multiple pollutants emitted from a large number of sources at a large
scale (i.e. nationwide) as part of a national air toxics assessment.
-------
Although we tailored EMS-HAP to process the 1996 National Toxics Inventory (NTI),
you can use it for any emission inventory following the instructions in this guide. The
1996 NTI is the first comprehensive model-ready national inventory of toxics, containing
facility-specific estimates of hazardous air pollutants (HAPs).
While other emission models, such as EMS-95 and EPS 2.0, are available, they do not
address the details of the 1996 NTI or the input requirements of the ASPEN model.
1.2 How do I use the EMS-HAP User's Guide?
This guide describes the programs that comprise EMS-HAP, and gives instructions on
how to use them to create ASPEN emission input files for base year or projected year
inventories of your choice. This manual is not specific to any one input inventory. For
example, you are not limited to using the 1996 NTI to run EMS-HAP. You need only
make sure your input inventory meets the requirements described within each program.
This guide also provides information on how we used EMS-HAP to process 1996
emissions data for a national screening study.
We present the programs in the order you may choose to use them. Chapter 2 describes
the aircraft emissions processing program. Chapters 3 through 7 describe the point
source processing programs. Chapters 8 through 10 describe the programs for area and
mobile source processing. Each chapter describes the function of the program, how to
run the program, all required ancillary input files and emission inventory data
requirements, and how to evaluate the output to determine if the data were processed
successfully. In this guide, all ancillary SAS data files are named without their
extension, since SAS files extension names vary with system and engine type. All
programs are also named without their extension.
Appendix A presents the file formats of the ancillary input files. Appendix B contains
sample batch files for running the EMS-HAP programs. Appendix C discusses
preparation of the point source component of the 1996 NTI for input into EMS-HAP.
Appendix D presents the methodologies used to prepare emission input files for the
ASPEN model for a national air toxics assessment. Appendix D also discusses how we
developed the key ancillary input files, such as the spatial allocation factor files, provided
with EMS-HAP. The ancillary files provided with EMS-HAP are those we used to
produce the 1996 ASPEN modeling inventory.
A separate user's guide is available for the ASPEN model. Users familiar with ASPEN
model input requirements will have a better understanding of EMS-HAP.
-------
Appendix D
Development of the Emission Inventory
(Provided in electronic format on Appendices CD)
-------
Appendix D - Summary
Development of the Emission Inventory
This appendix includes a technical paper describing the development of the 1996
National Toxic Inventory (NTI). The following is the abstract for this paper:
The 1990 amendments to the Clean Air Act (CAA) established the need for a
comprehensive hazardous air pollutant (HAP) emissions inventory effort that can be used
to track progress by the Environmental Protection Agency (EPA) over time in reducing
HAPs in ambient air. To estimate risk and HAP emission reductions, the EPA compiled
the 1996 National Toxics Inventory (NTI) to provide a model-ready emissions inventory.
The 1996 NTI contains estimates of facility-specific HAP emissions and their
source-specific parameters necessary for modeling such as location and facility
characteristics (stack height, exit velocity, temperature, etc.). Complete source category
coverage is needed for modeling, and the NTI contains estimates of emissions from
major, area, and mobile source categories. Compiling this huge amount of data presents
a significant challenge to EPA. To compile the data, the EPA first solicited HAP
emissions data from states, and 36 states, Puerto Rico and the Virgin Islands delivered
HAP emissions inventories to the EPA. These state data varied in completeness, format,
and quality. The EPA evaluated the state data and supplemented it with data gathered
while developing Maximum Achievable Control Technology (MACT) standards and
with Toxic Release Inventory (TRI) data. Then the EPA estimated emissions for other
states and for sources not included in the state data to produce a complete model-ready
national 1996 inventory. The EPA released the draft 1996 NTI for external comment and
received revisions from 42 states, industry, and other organizations. The EPA released
the final 1996 NTI in June 2000. This paper discusses the compilation of the 1996 NTI
in order to evaluate the success of EPA's national air toxics program and presents
summary emissions data from the 1996 NTI.
In addition to the paper include in Appendix D, the following are reference files that
provide further detail on the inventory development. These documents are avaiable at
http://www.epa.gov/ttn/chi ef/ei_guide.html#airtoxics
• Documentation for the 1996 Base Year National Toxics Inventory for Aircraft Sources
• Documentation for the 1996 Base Year National Toxics Inventory for Area Sources
• Documentation for the 1996 Base Year National Toxics Inventory for Commercial
Marine Vessel and Locomotive Mobile Sources
• Documentation for the 1996 Base Year National Toxics Inventory for Nonroad Vehicle
and Equipment Mobile Sources
• Documentation for the 1996 Base Year National Toxics Inventory for Onroad Sources
• Documentation for the 1996 Base Year National Toxics Inventory for Point Sources
-------
Appendix E
ASPEN User's Guide
(Provided in electronic format on Appendices CD)
-------
Appendix E - Summary
This appendix contains the " User's Guide for the Assessment System for Population
Exposure Nationwide Model (ASPEN, Version 1.1"
This user's guide provides documentation for the Assessment System for Population
Exposure Nationwide (ASPEN, Version 1.1), referred to hereafter as ASPEN. It includes
a technical description of the ASPEN algorithms, user instructions for running the model
and a tutorial for getting started. The ASPEN model consists of a dispersion and mapping
module. The dispersion module is a Guassian formulation for estimating ambient annual
average concentrations at a set of fixed receptors within the vicinity of the emission
source. The mapping module produces a concentration at each census tract. Input data
needed are emissions data, meteorological data and census tract data.
-------
Appendix F
Estimation of Background Concentrations for Diesel Particulate Matter
(Provided in electronic format on Appendices CD)
-------
Appendix F - Summary
Estimation of Background Concentrations for Diesel Particulate Matter
This appendix contains the calculations utilized to determine the background
concentrations for Diesel Particulate Matter. Background concentrations are an
essential part of the total air quality concentration to be considered in determining source
impacts. Background air quality includes pollutant concentrations due to: 1) natural
sources; 2) nearby sources that are unidentified in the inventory; and 3) long range
transport into the modeling domain. Typically, monitored air quality data should be used
to establish background concentrations.
-------
Appendix G
Health Effects Information Used in Cancer and Noncancer Risk Characterization
for the NATA 1996 National-Scale Assessment
(Provided in electronic format on Appendices CD)
-------
Appendix G - Summary
Health Effects Information Used In Cancer and Noncancer Risk Characterization
for the NATA 1996 National-Scale Assessment
This appendix contains the hazard identification and dose-response assessment
information for the NATA national-scale assessment. The criteria for selection of the
information was obtained from various sources and prioritized according to (1)
applicability, (2) conceptual consistency with EPA risk assessment guidelines, and (3)
level of review received. The prioritization process was aimed at incorporating into our
assessment the best-available science with respect to dose-response information. The
following sources were used:
• US Environmental Protection Agency (EPA)
• Agency for Toxic Substances and Disease Registry (ATSDR)
• California Environmental Protection Agency (CalEPA)
• International Agency for Research on Cancer (IARC)
-------
Appendix H
Estimating Carcinogenic Potency for Mixtures of Polycyclic Organic Matter (POM)
for the 1996 National-Scale Assessment
(Provided in electronic format on Appendices CD)
-------
Appendix H - Summary
Estimating Carcinogenic Potency for Mixtures of Polycyclic Organic Matter (POM)
for the 1996 National-Scale Assessment
The polycyclic organic matter (POM) category within the Clean Air Act's section 112(b)
list of hazardous air pollutants comprises a broad spectrum of compounds having widely
varying toxic potencies. Because all these compounds have been listed as a single
category under the Act, the 1996 National Toxics Inventory (NTI) also records them only
as a group for the great majority of sources, usually in terms of total polynuclear aromatic
hydrocarbons (PAH - one type of POM) or total POM. Most of these entries do not
include information on the method used to estimate the emission rate.
For this reason, the NTI data could not support modeling of individual POM compounds
for the initial national-scale assessment. The alternative-modeling modeling POM as a
group - was a significant simplifying step because different types of emission sources
may be expected to produce different characteristic mixtures of POM compounds. These
different mixtures have the potential to vary substantially in toxic potency per unit mass.
The method to aggregate these pollutants for the national scale assessment is presented in
this appendix.
-------
Appendix I
Model-to-Monitor Comparison Methods
(Provided in electronic format on Appendices CD)
-------
Appendix I - Summary
Protocol for Model-to-Monitor Comparisons for the
Initial National Scale Assessment
This appendix provides the protocol for "Model-to-Monitor" comparisons for national
scale Assessment (Protocol) that was reviewed by the EPA Science Advisory Board
(SAB) on August 18, 2000. The SAB review of the Protocol is available on the internet
at: http://www.epa.gov/sab/ec0015.pdf
Not all of the methods described in the Protocol were used in the model-to-monitor
comparison included in the NATA Report; some of the methods that were used were
modified; and, some new methods were introduced. The decision to diverge somewhat
from the methods described in the Protocol was based on the recommendations of the
SAB, as well as our own judgment.
-------
Appendix J
Comparison of ASPEN Results to Monitored Concentrations
(Provided in electronic format on Appendices CD)
-------
Appendix J - Summary
Initial National-Scale Assessment
Comparison of ASPEN Modeling System Results to Monitored Concentrations
This appendix describes the results of a model-to-monitor comparison we conducted for
subset of the 33 urban HAPs. We view this comparison as a evaluation of ASPEN and he
inputs that go into ASPEN, including: emissions data from various sources including the
National Toxics Inventory (NTI), the Emissions Modeling System for Hazardous Air
pollutants (EMS-HAP), and meteorological data. For most of the pollutants evaluated, we
found that ASPEN estimates tended to be lower than the monitor averages at the exact
locations of the monitors. In general it appears that ASPEN is underestimating monitor-
based HAP concentrations.
Possible reasons for ASPEN to underestimate HAP concentrations include:
1) The National Toxics Inventory (NTI) missing specific emissions sources (for many of
the sources in the NTI some of the emissions parameters are defaulted or missing)
2) Emission rates being underestimated in many locations. We believe the ASPEN model
itself is contributing in only a minor way to the underestimation. This is mainly
due to output from the antecedents of the ASPEN model comparing favorably to
monitoring data in cases where the emissions and meteorology were accurately
characterized and the monitors took more frequent readings. In simulations we
ran, the ASPEN s estimates compared favorably to the estimates derived from a
more meticulous model.
3) Monitor siting may have also contributed to the underestimation. Sites are normally
situated to find peak pollutant concentrations, which implies that errors in the
characterization of sources would tend to make the model underestimate the
monitor values.
4) Finally, we are not sure of the accuracy of the monitor averages, which have their own
sources of uncertainty.
Our results suggest that the model estimates are uncertain on a local scale (i.e., at the
census tract level). We believe that the model estimates are more reliably interpreted as
being a value likely to be found within 30 km of the census tract location.
-------
Appendix K
HAPEM4 Results
(Provided in electronic format on Appendices CD)
-------
Appendix K - Summary
HAPEM4 Result Charts
This appendix presents the graphical results of the HAPEM4 exposure modeling
Included are:
• Pollutant specific exposure concentration charts
• Pollutant specific percent contribution charts
• State summary exposure concentration estimate charts
In addition, the "Page 2" referred to on each chart has been provided at end of the
appendix.
-------
Appendix L
Risk Characterization Results
(Provided in electronic format on Appendices CD)
-------
Appendix L - Summary
Risk Characterization Charts
This appendix contains the risk and population summary charts for carcinogens as well as
noncarcinogens. Charts are also summarized by source sector (i.e., major, area, mobile
onroad, mobile off-road, background). A set of the Key Risk Assumptions and
Limitations is also provided.
-------
TECHNICAL REPORT DATA
(Please read Instructions on reverse before completing)
1. REPORT NO.
EPA-453/R-01-003
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
5. REPORT DATE
National-Scale Air Toxics Assessment for 1996
January 2001
. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Roy L. Smith and Ted Palma
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Research Triangle Park, NC 27711
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
12. SPONSORING AGENCY NAME AND ADDRESS
13. TYPE OF REPORT AND PERIOD COVERED
Director
Office of Air Quality Planning and Standards
Office of Air and Radiation
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
14. SPONSORING AGENCY CODE
EPA/200/04
15. SUPPLEMENTARY NOTES
16. ABSTRACT
The document describes the EPA's National-Scale Air Toxics Assessment, based on emissions data for
1996. The national-scale assessment is a nationwide study of potential inhalation exposures and health
risks associated with 32 hazardous air pollutants (i.e., air toxics) and diesel paniculate matter, based on
1996 data. This initial national-scale assessment is one component of the National Air Toxics
Assessment (NATA), the technical support component of EPA's National Air Toxics Program. The
purpose of the national-scale assessment if to gain a better understanding of the air toxics problem in
the U.S. The document explains the methodologies used for the initial national-scale assessment and
provides a summary of the results. The document also presents the risk characterization, which
includes an uncertainty and variability analysis, and provides an overall summary of results and
recommendations for future actions.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b. IDENTIFIERS/OPEN ENDED TERMS
c. COSATT Field/Group
Air toxics, NAT A, risk assessment, exposure
18. DISTRIBUTION STATEMENT
Release Unlimited
19. SECURITY CLASS (Report)
Unclassified
21. NO. OF PAGES
238
20. SECURITY CLASS (Page)
Unclassified
22. PRICE
EPA Form 2220-1 (Rev. 4-77) PREVIOUS EDITION IS OBSOLETE
------- |