&EPA
United States
Environmental Protection
Policy, Planning,
And Evaluation
(PM-221J
1992
Baltimore Integrated Environmental
Management Project: Phase II
Ambient Air Toxics Report
PENNSYLVANIA
Printed on Recycled Paper
-------
PREFACE
This report was prepared under the auspices of the Baltimore
Integrated Environmental Management Project (IEMP). The
Baltimore IEMP is a collaborative effort of the State of
Maryland, Anne Arundel and Baltimore Counties, the City of
Baltimore and the U.S. Environmental Protection Agency. The
Environmental Protection Agency (EPA) initiated the project as
part of its pursuit of new approaches to environmental management
and policy. The purpose of the IEMP is to use an integrated
approach to identify and assess environmental issues that concern
managers, to set priorities for action among these issues, and to
analyze appropriate approaches to manage these problems.
The Baltimore IEMP represents the second of four geographic
projects that EPA initiated across the country. The Baltimore
area was chosen, not because it has a significant toxics problem,
but because EPA and local officials wanted to explore better ways
to identify, assess, and manage the human health risks of
environmental pollutants in the area. Other lEMPs include
Philadelphia, Santa Clara County, and Denver.
The decision-making structure of the Baltimore IEMP
consisted of two committees, which also served as the means for
State and local participations the Management Committee and the
Technical Advisory Committee. The Management Committee, with
members representing Baltimore City, Baltimore County, Anne
Arundel County, and the State, managed the IEMP and set its
overall policy directions. The Technical Advisory Committee,
composed of technical managers from the City of Baltimore, the
two counties, the State, as well as representatives from the
Regional Planning Council and the academic community, recommended
issues to study, advised the Management Committee on the
technical and scientific aspects of the project, and oversaw and
commented on all EPA and consultant work, EPA provided
administrative, technical, and analytical support.
The Baltimore IEMP examined five environmental issues: air
toxics, .Baltimore Harbor, indoor air pollution, lead paint
abatement, and potential contamination of ground water from
underground tanks. For further information on these reports or
other IEMP studies contact the Regulatory Integration Division,
the Office of Policy Analysis (PM-220) in the Office of Policy,
Planning and Evaluation, U.S. Environmental Protection Agency,
Washington, D.C. 20460
-------
ACKNOWLEDGEMENTS
Numerous individuals contributed to the preparation of this
report. We wish to mention the following individuals:
of the Air ToxicB Workgroup
Tad Aburn, Section Head, Hazardous Pollutant Evaluation Section,
Air Management Administration, Maryland Department of the
Environment
Don Andrew, Administrator, Engineering & Enforcement Program, Air
Management Administration, Maryland Department of the Environment
Anthony S. Bonaccorsi, formerly Director, Environmental Services,
Eastern Stainless Steel Company
Daryl Braithwaite, formerly Program Coordinator, Clean Water
Project
David Filbert, Chief, Bureau of Air Quality Management, Baltimore
County Department of Environmental Protection and Resource
Management
Michael K. Hettleman, President, The Southern Galvanizing Company
Frank Hoot, Assistant Commissioner, Environmental Health,
Baltimore City Health Department (Retired)
Dr. Genevieve Matanoski, Professor, Department of Epidemiology,
Johns Hopkins School of Hygiene and Public Health
Darryl W. Palmer, Environmental Manager, Agricultural Chemical
Group, FMC Corporation
Susan Wierman, Acting Deputy Director, Air Management
Administration, Maryland Department of the Environment
Bruce Windsor, Program Coordinator, American Lung Association of
Maryland.
ptotnhftr-a of the Indoor Air Workgroup who Contributed to the Design
of the Joint Air Toxics Study
David Filbert, Baltimore County Department of Environmental
Protection
Joseph Abey, Anne Arundel County Health Department
Dr. Charles Billings, Johns Hopkins University
Elkins W. Dahle, Baltimore City Health Department
-------
Dr. Katharine Parrel I/ Maryland Department of the Environment
Allan Heaver, Heaver Properties
Jack Lodge/ Baltimore Gas and Electric Company
Arthur Nieberding, Mueller Associates/ Inc.
Velma Rector/ American Lung Association of Maryland
Tom King, Mueller Associates, Inc.
oi! BPArB Bcono111^ c Benefits Brunch who Contributed to the
Chapter on Risk M*T>*o*ffn*nt
Brett Snyder
Joel Scheraga
Itemhqrg of the Balt^^re IBMP Management CTW? ttee
J. James Dieter/ Special Assistant to the Director, Department of
Environmental Protection and Resource Management, Baltimore
County
Max Eisenberg, Assistant Secretary for Toxics, Environmental
Science and Health, Department of the Environment, State of
Maryland
Robert Perciasepe, (formerly) Assistant Director, Department of
Planning, City of Baltimore
Claude Vannoy, Assistant to the County Executive for Land Use,
Anne Arundel County
MamHqrB of the Bflltj^mftre TUMP Technical Advisory Cotl"B'*t'tee
Jared L. Cohon, Vice Provost for Research, and Professor of
Geography and Environmental Engineering, Johns Hopkins University
(Chairman, Technical Advisory Committee)
Don Andrew, Administrator, Engineering & Enforcement, Air
Management Administration, Department of the Environment, State
of Maryland
Philip Clayton, Manager, Cooperative Clean Water Program,
Regional Planning Council
Emery Cleaves, Principal Geologist, Maryland Geological Survey
(Chairman, Ground-Water Subcommittee)
-------
Ralph Cullison, Baltimore City Department of Public Works, City
of Baltimore
N. Singh Dhillon, Director, Environmental Health, Anne Arundel
County Department of Health
Tom Ervin, Environmental Planner, Anne Arundel County Office of
Planning and Zoning (Chairman, Ecological Subcommittee)
Katherine Farrell, M.D., M.P.H., Chief, Division of Environmental
Disease Control, Department of the Environment, State of Maryland
David Filbert, Chief, Bureau of Air Quality Management, Baltimore
County Department of Environmental Protection and Resource
Management
Frank Hoot, (formerly) Assistant Commissioner, Environmental
Health, Baltimore City Health Department (Chairman, Human Health
Subcommittee)
Sam Martin, Consultant, Vice Chairman of TAG (represented the
Regional Planning Council during Phase I)
Janice Outen, Director, Division of Environmental Management,
Baltimore County Department of Environmental Protection and
Resource Management
Colin Thacker, (formerly) Management Assistant, Baltimore County
Department of Environmental Protection and Resource Management
Bill Wolinski, Water Quality Coordinator, Department of Public
Health
Staff of the U.S. Environmental Protection Acrencv
Daniel Beardsley, (formerly) Director, Regulatory Integration
Division
Arthur Koines, Chief, Geographic Studies Branch
John Chamberlin, (formerly) Site Director, Baltimore IEMP
Andrew Manale, Senior Analyst, IEMP
Catherine Tunis, Policy Analyst, IEMP
Ellen Tohn, (formerly) Policy Analyst, IEMP
Roberta Grossman, (formerly) Policy Analyst, IEMP
Alan Jones, (formerly) Policy Analyst, IEMP
-------
SchoolB
Orterio Villa, Jr.i Director, Central Regional Laboratory, EPA
Region III
John Austin, Central Regional Laboratory, EPA Region III
Gerry Akland, Environmental Monitoring Systems Laboratory, Office
of Research & Development, EPA
Andy Bond, Environmental Monitoring Systems Laboratory, Office of
Research & Development, EPA
Tom Hartledge, Environmental Monitoring Systems Laboratory,
Office of Research & Development, EPA
William Nelson, Environmental Monitoring Systems Laboratory,
Office of Research & Development, EPA
Dr. Lance Wallace, Environmental Monitoring Systems Laboratory,
Office of Research & Development, EPA
Dr. Joelien Leutas, Chief Genetic Bioassay Laboratory, Office of
Research & Development, EPA
Ila Cote, Office of Air Quality Planning and Standards, EPA
Joseph Padgett, Associate Director for Intermedia and
Intergovernmental Programs, Office of Air Quality Planning and
Standards, EPA
Tim Barry, Science Policy Branch, Regulatory Integration
Division, Office of Policy Analysis, EPA
Robert Omelia, Air Management Administration, Department of the
Environment, State of Maryland
Robert Merrey, Bureau of Air Quality Management, Baltimore County
Department of Environmental Protection and Resource Management
Adon Phillips, Baltimore County Office of Central Services
Dr. L. E. Bills and the staff of the Southeastern Community
Mental Health Center, Baltimore County Department of Health
Merreen Kelly, Baltimore County Public Schools
Arthur Himlin, Principal, and staff of Parkville Middle School
(Baltimore County)
George P. Dausch III, Principal, and staff of Dundalk Senior High
School (Baltimore County)
-------
Individual* from Conaultino Firm
Craig Koralek, VERSAR
Dennis Hlinka, VERSAR
Josephina Castellanos, VERSAR
Kevin Jameson, VERSAR
David Sullivan, Sullivan Environmental Associates
Jay Wind, American Management Systems
Elaine Haemisegger, Temple, Barker, & Sloane, Inc.
-------
TABLE OF CONTENTS
EXECUTIVE SUMMARY
EPILOGUE
I. INTRODUCTION 1-1
1. INTEGRATED ENVIRONMENTAL MANAGEMENT PROJECTS ... 1
2. ORGANIZATION OF THE PROJECTS 2
3. THE BALTIMORE I BMP 2
4. THE RESULTS OF THE PHASE I PRIORITY-SETTING
PROCESS 4
5. RELATIONSHIP TO THE EPA AND AMA AIR TOXICS
STRATEGIES 4
6. OVERVIEW OF THE REPORT 6
II. OBJECTIVES AND DESIGN OF THE BALTIMORE AMBIENT AIR TOXICS
STUDY II-l
1. STUDY OBJECTIVES 1
2. STUDY DESIGN 2
a. Pollutant Selection 2
b. Risk Characterization of the Selected
Pollutants 3
c. Risk Management: A Demonstration 4
3. HOW THE RESULTS OF THE AIR TOXICS STUDY SHOULD BE
VIEWED 4
III. SUMMARY OF THE PRINCIPLES OF RISK ASSESSMENT AND RISK
MANAGEMENT AS USED IN THE AMBIENT AIR TOXICS STUDY . III-l
1. RISK ASSESSMENT AND RISK MANAGEMENT 1
2. STEPS IN RISK ASSESSMENT 2
a. Hazard Identification 2
b. Dose-Response Evaluation 3
i. Cancer 4
ii. Cancer Caused by Exposures to Complex
Mixtures 6
iii. Noncancer Health Effects 7
c. Human Exposure Evaluation 8
d. Risk Characterization 9
i. Cancer 9
ii. Noncancer 11
3. INTERPRETING RISK ASSESSMENT RESULTS IN THE
BALTIMORE AIR TOXICS STUDY 13
4. RISK MANAGEMENT 14
-------
TABLE OF CONTENTS (CONTD.)
IV. SELECTING POLLUTANTS. EVALUATING TOXICITY. AND
ESTIMATING EXPOSURE IV-1
1. POLLUTANT SELECTION METHODOLOGY AND RESULTS .... 1
a. Pollutant Selection for Modelling 2
b. Pollutant Selection for TEAM Monitoring .... 5
2. HAZARD IDENTIFICATION/DOSE-RESPONSE EVALUATION . . 7
a. Cancer Effects 7
i. Weight of Evidence 7
ii. Dose-Response Evaluation 11
b. Noncancer Effects 11
i. Weight of Evidence 11
ii. Dose-Response Evaluation 16
3. HUMAN EXPOSURE EVALUATION 17
a. Air Dispersion Modelling 17
i. Emissions Assessment 17
ii. Emissions Characterization for Key
Sources 28
iii. Dispersion Modelling 29
iv. Model Performance Evaluation 31
b. Ambient Air Monitoring 34
i. TEAM VOC Sampling 34
ii. TEAM Metals Monitoring 39
iii. Available Area-Wide Sampling Data ... 42
c. Exposed Population 42
V. RISK ASSESSMENT SCREENING RESULTS V-l
1. RANKING OF TARGET COMPOUNDS AND SOURCES BY CANCER
POTENCY-WEIGHTED AMBIENT CONCENTRATIONS 3
2. RISK ASSESSMENT SCREENING RESULTS 4
a. Cancer Risks 6
i. Area-Wide Risks 6
ii. "Hotspot" Risks 24
b. Noncancer Risks 37
i. Results Based on Dispersion Modelling . . 37
ii. Results Based on Monitoring Data .... 39
c. Limitations 41
i. Exposure Limitations Based on Dispersion
Modelling 41
ii. Exposure Limitations Based on
Monitoring 44
iii. Dose-Response Limitations 45
3. RISK RESULTS COMPARISONS 46
a. Comparison with IEMP and Other Urban-Scale
Analyses 46
b. Comparison of Alternative Approaches to
Estimating Risks from the Products of
Incomplete Combustion • 52
c. Comparison of Risks in Air and Drinking
Water 54
-------
TABLE OF CONTENTS (CONTD.)
VI. RISK MANAGEMENT STRATEGIES FOR REDUCING RISKS;
A DEMONSTRATION VI-1
1. SELECTION OF SOURCES 3
2. ESTIMATING CONTROL OPTIONS COSTS, EFFECTIVENESS,
AND BENEFITS 9
a. Feasible Control Options and their Cost-
Effectiveness 9
b. Estimating the Benefits Associated with the
Control Options 19
3. ANALYSIS OF THE COST-EFFECTIVENESS OF CONTROL
STRATEGIES 21
a. Cost-Effectiveness Analysis Overview 21
b. Cost-Effectiveness Analysis Results 25
i. Results: Annual Excess Cancer
Incidence 25
ii. Results: Maximum Increased Lifetime
Individual Cancer Risk 34
iii. Limitations 38
4. MAXIMIZING THE BENEFITS TO SOCIETY IN CONTROLLING
AIR TOXICS 39
a. Results of the Analysis of Strategies to
Maximize Benefits 39
b. Comparison of Benefit and Cost-Effectiveness
Analyses 39
APPENDIXES
-------
TABLES
IV-1 CHEMICALS SELECTED FOR AIR DISPERSION MODELLING IN THE
BALTIMORE IEMP AIR TOXICS STUDY IV-6
IV-2 CHEMICALS SELECTED FOR ORD'S TEAM AMBIENT AIR
MONITORING AS PART OF THE BALTIMORE IEMP AIR TOXICS
STUDY IV-8
IV-3 SUMMARY OF POLLUTANTS SELECTED FOR MODELLING AND
MONITORING IN THE BALTIMORE AMBIENT AIR TOXICS
STUDY IV-9
IV-4 UNIT CANCER RISK FACTORS AND THE WEIGHT OF EVIDENCE FOR
CARCINOGENICITY FOR THE BALTIMORE IEMP TARGET
COMPOUNDS IV-12
IV-5 NONCANCER HEALTH EFFECTS AND THRESHOLD VALUES FOR
TARGET COMPOUNDS IN THE BALTIMORE IEMP AIR
ANALYSIS IV-18
IV-6 COMPARISON OF MEASURED VS. PREDICTED CONCENTRATIONS FOR
BALTIMORE MONITORING SITES DURING THE PERIOD 11/20/83 -
2/16/84 FOR MODEL RUN URBAN MODE IV-32
IV-7 COMPARISON OF MEASURED VS. MODELLED DATA FOR THE
BALTIMORE AND PHILADELPHIA IEMP STUDIES IV-33
IV-8 BALTIMORE IEMP AIR TOXICS STUDY AVERAGE MEASURED
AMBIENT CONCENTRATIONS AT THE TWO FIXED MONITORING
SITES: TARGET VOLATILE ORGANICS IV-37
IV-9 COMPARISON OF 1987 AND 1983/84 MONITORING
RESULTS IV-40
IV-10 AVERAGE MEASURED AMBIENT CONCENTRATIONS: METALS AT
FIXED MONITORING SITES IV-41
IV-11 AVAILABLE AREA-WIDE SUMMARY MONITORING DATA FOR
BALTIMORE IV-43
V-l RELATIVE RANKING OF TARGET COMPOUNDS BY CANCER POTENCY-
WEIGHTED AMBIENT CONCENTRATIONS V-4
V-2 RELATIVE RANKING OF SOURCES BY CONTRIBUTION TO CANCER
POTENCY-WEIGHTED AMBIENT CONCENTRATIONS V-5
V-3 AVERAGE INCREASED LIFETIME INDIVIDUAL CANCER RISK USING
AVAILABLE MONITORING DATA V-10
V-4 BALTIMORE IEMP AIR TOXICS STUDY PHASE II RESULTS . V-ll
-------
TABLES (CON'T)
V-5 AVERAGE INCREASED LIFETIME INDIVIDUAL CANCER RISKS
BASED ON AVAILABLE MONITORING DATA: SITES WITH THE
HIGHEST AND LOWEST MEASURED VALUES V-15
V-6 ESTIMATED ANNUAL EXCESS CANCER INCIDENCE FOR SELECTED
POLLUTANTS MODELLED IN THE BALTIMORE IEMP AIR TOXICS
STUDY (TOTAL STUDY AREA, 5 KM GRID SYSTEM) . . . . V-19
V-7 ESTIMATED ANNUAL EXCESS CANCER INCIDENCE FOR SELECTED
POLLUTANTS MODELLED IN THE BALTIMORE IEMP AIR TOXICS
STUDY BY SOURCE V-21
V-8 BALTIMORE IEMP AIR TOXICS STUDY ESTIMATED ANNUAL EXCESS
CANCER INCIDENCE BASED ON MODELLING BY SOURCE AND
POLLUTANT (TOTAL STUDY AREA, 5 KM GRID SYSTEM) . . V-23
V-9 AREA-WIDE ANNUAL EXCESS CANCER INCIDENCE USING
AVAILABLE MONITORING DATA V-25
V-10 POLLUTANT CONTRIBUTIONS BASED ON MODELLING TO MAXIMUM
INCREASED LIFETIME INDIVIDUAL CANCER RISK AT THREE
HOTSPOT LOCATIONS V-28
V-ll TOTAL SOURCE AND POLLUTANT CONTRIBUTION BASED ON
MODELLING AT EACH HOTSPOT LOCATION V-30
V-12 BALTIMORE IEMP AIR TOXICS STUDY ESTIMATED ANNUAL EXCESS
CANCER INCIDENCE BASED ON MODELLING BY POLLUTANT IN THE
GRID CELL OF HIGHEST PREDICTED ANNUAL EXCESS CANCER
INCIDENCE V-34
V-13 BALTIMORE IEMP AIR TOXICS STUDY ESTIMATED MAXIMUM
ANNUAL EXCESS CANCER INCIDENCE BY POLLUTANT AND SOURCE
IN THE GRID CELL OF HIGHEST ESTIMATED ANNUAL EXCESS
CANCER INCIDENCE V-35
V-14 RECEPTOR LOCATIONS WARRANTING FURTHER INVESTIGATION FOR
NONCANCER EFFECTS: POLLUTANT-SPECIFIC V-38
V-15 RECEPTOR LOCATIONS WARRANTING FURTHER INVESTIGATION FOR
NONCANCER EFFECTS: COMPLEX POLLUTANT MIXTURES . . .V-40
V-16 BALTIMORE IEMP AIR TOXICS STUDY NONCANCER RATIO
CALCULATIONS FROM EXPOSURE TO TARGET COMPOUNDS AT THE
TWO FIXED EXPOSURE TO TARGET COMPOUNDS AT THE TWO FIXED
TEAM MONITORING SITES V-42
V-17 A COMPARISON OF PREDICTED AVERAGE LIFETIME INDIVIDUAL
CANCER RISK IN BALTIMORE WITH OTHER STUDIES . . . .V-47
-------
FIGURES
1-1 BALTIMORE IEMP STUDY AREA 1-3
IV-1 AREA SOURCE GRID SYSTEM FOR BALTIMORE STUDY . . . IV-27
IV-2 BALTIMORE INTEGRATED ENVIRONMENTAL MANAGEMENT PROJECT
MAP OF AMBIENT AIR MONITORING IN 1987 IV-35
IV-3 BALTIMORE STUDY AREA MAP OF BOUNDARIES OF THE MODELLING
DOMAIN FOR THE 5 KM (STANDARD) GRID SYSTEM . . . .IV-44
IV-4 BALTIMORE STUDY AREA MAP OF BOUNDARIES OF THE MODELLING
DOMAIN FOR THE 2.5 KM (REFINE) GRID SYSTEM . . . IV-45
ZV-5 BALTIMORE STUDY AREA MAP OF BOUNDARIES OF THE MODELLING
DOMAIN IV-47
V-l MODELLING DOMAIN BY UTM COORDINATES V-7
V-2 BALTIMORE IEMP AIR TOXICS POINT & AREA SOURCE
CONTRIBUTION TO AVERAGE LIFETIME INDIVIDUAL CANCER RISK
BY GRID CELL V-8
V-3 BALTIMORE IEMP AIR TOXICS ANNUAL EXCESS CANCER
INCIDENCE, AVERAGE LIFETIME CANCER RISK AND EXPOSED
POPULATION BY GRID CELL V-17
V-4 ANNUAL EXCESS CANCER INCIDENCE BY GRID CELL FOR THE
STUDY AREA V-18
V-5 BALTIMORE IEMP AIR TOXICS POINT & AREA SOURCE
CONTRIBUTION TO ANNUAL EXCESS CANCER INCIDENCE BY GRID
CELL V-22
V-6 LOCATION OF HOTSPOTS EVALUATED IN THE AIR TOXICS
STUDY V-27
VI-1 BALTIMORE IEMP AIR TOXICS STUDY ANALYSIS OF THE COST-
EFFECTIVENESS OF CONTROL OPTIONS VI-24
VI-2 TOTAL ANNUALIZED COST VERSUS REDUCTION IN ANNUAL EXCESS
CANCER INCIDENCE VI-31
VI-3 TOTAL ANNUALIZED COST VERSUS REDUCTION IN MAXIMUM
LIFETIME INDIVIDUAL CANCER RISK VI-37
VI-4 SUMMARY OF COSTS OF CONTROLLING AIR TOXICS IN BALTIMORE
COMPARISON OF BENEFIT METHOD AND COST-EFFECTIVE
SCENARIOS VI-45
-------
TABLES (CON'T)
V-18 A COMPARISON OF PREDICTED MAXIMUM LIFETIME INDIVIDUAL
CANCER RISK IN BALTIMORE WITH OTHER STUDIES . . . .V-49
V-19 COMPARISON OF THE RESULTS OF ALTERNATIVE APPROACHES TO
ESTIMATING THE RISKS FROM THE SEMI-VOLATILE COMPONENT
OF THE PRODUCTS OF INCOMPLETE COMBUSTION V-53
VI-1 BALTIMORE IEMP AIR TOXICS STUDY ESTIMATES OF EXCESS
ANNUAL CANCER INCIDENCE FOR SOURCES SELECTED FOR
ANALYSIS OF CONTROL OPTIONS VI-4
VI-2 SOURCES POSING GREATER THAN 5 X 10'6 INCREASED LIFETIME
INDIVIDUAL CANCER RISK VI-7
VI-3 SOURCES & POLLUTANTS INCLUDED IN THE CONTROL
ANALYSIS VI-8
VI-4 DESCRIPTION OF CONTROL OPTIONS FOR POINT SOURCES VI-10
VI-5 DESCRIPTION OF CONTROL OPTIONS FOR AREA SOURCES . VI-13
VI-6 CONTROL OPTIONS & ANNUALIZED COSTS VI-18
VI-7 HEALTH AND WELFARE EFFECTS QUANTIFIED VI-20
VI-8 CONTROL OPTIONS, ANNUALIZED BENEFITS, AND COSTS FOR TSP
AND VOC CONTROLS VI-22
VI-9 BALTIMORE IEMP AIR TOXICS STUDY SCHEDULE OF CONTROL
STRATEGIES FOR REDUCING CANCER INCIDENCE PHASE II VI-27
VI-10 BALTIMORE IEMP AIR TOXICS STUDY SCHEDULE OF CONTROL
STRATEGIES FOR REDUCING LIFETIME INDIVIDUAL CANCER RISK
AT SITE OF MAXIMUM INDIVIDUAL CANCER RISK .... VI-35
VI-11 CONTROL STRATEGIES BASED ON BENEFITS ANALYSIS . . VI-40
-------
FIGURE (COMTD.)
VI-5 SUMMARY OF BENEFITS OF CONTROLLING AIR TOXICS IN
BALTIMORE COMPARISON OF BENEFIT METHOD AND COST-
EFFECTIVE SCENARIOS VI-46
VI-6 SUMMARY OF NET ECONOMIC BENEFITS OF CONTROLLING AIR
TOXICS IN BALTIMORE COMPARISON OF BENEFITS METHOD AND
COST-EFFECTIVE SCENARIOS VI-47
-------
EXECUTIVE SUMMARY
The Baltimore IEMP Ambient Air Toxics Study
INTRODUCTION
This report describes the second phase of a two-phase study
of human health risks from ambient air toxics in the greater
Baltimore area.1 The results of the first phase of the study are
presented in Baltimore Integrated Environmental Management
Project; Phase I Report. The air toxics study was conducted as
part of the Baltimore Integrated Environmental Management Project
(IEMP). The Regulatory Integrated Division (RID), Office of
Policy Analysis, of the U.S. Environmental Protection Agency
(EPA). The Baltimore IEMP is one of four full-scale geographic
projects initiated by RID over the past six years. The other
lEMPs focused on the Philadelphia metropolitan area, Santa Clara
Valley, California, and Denver, Colorado.
This executive summary is divided into five sections:
• Integrated Environmental Management, which describes the
need for and the basic concepts of the IEMP
• Baltimore IEMP, which summarizes the structure of the
overall project
• Ambient Air Toxics Study Oblectives. which presents the
goals of the ambient air toxics study
• Ambient Air Toxics Study Design, which discusses how the
ambient air toxics study was completed
• Conclusiona. which summarizes the major findings of the
ambient air toxics study and insights on additional
analysis that is needed on this topic
'This study estimated the increase in cancer and noncancer
risks resulting from exposure to ambient (i.e., outdoor)
concentrations of air toxics. It did not consider exposures
resulting from (1) indoor air, (2) the workplace, and (3) other
pathways (e.g., ingestion).
-------
INTEGRATED ENVIRONMENTAL MANAGEMENT
RID adopted the concept of integrated environmental
management as one solution to the shortcomings of EPA's
traditional focus on individual industries, pollutants, and media
in developing pollution control strategies. Such a piecemeal
approach may result in environmental programs and regulations
that do not attack the greatest risks to human health first; it
may also lead to a combination of national standards that do not
address the unique environmental issues found at the local level.
The IEMP strives to set priorities by comparing similar risks
across all sources, pollutants, and exposure pathways. The IEMP,
through its focus on individual geographic areas, is also
designed to address site-specific situations that are often
inadequately addressed by national regulations.
The IEMP is grounded primarily in the concepts of risk
assessment and risk management. Rough estimates of risk—that
is, the probability of adverse health effects—are used as a
common measure for comparing and setting priorities among
environmental issues that involve different pollutants, sources,
and exposure pathways. Risk management is the process by which
decision makers balance programs to reduce human health risks
against the available resources to support those programs.
Chapter III and Appendix A provide more detail on risk assessment
and risk management as used in the IEMP.
The Baltimore IEMP is also built on the concept that these
projects should be managed at the state and local level, thus
ensuring that issues of greatest local concern or risk are
adequately addressed. The Baltimore IEMP was particularly
successful in this regard, as detailed below.
BALTIMORE IEMP
The Baltimore IEMP was a cooperative effort involving the
governments of the State of Maryland, the City of Baltimore,
Baltimore County, Anne Arundel County/and EPA. A map of the
study area is presented in Figure 1. The Baltimore area was
chosen, not because it has a significant toxics problem, but
because EPA and local officials wanted to explore better ways to
identify, assess, and manage the potential human health risks of
environmental pollutants in the area. In addition, Baltimore was
attractive because of the opportunity to establish a management
structure with the potential to continue after completion of the
project.
The decision-making structure of the IEMP consisted of two
committees, which also represented the vehicles for State and
local participation: the Management Committee (MC) and the
-------
Technical Advisory Committee (TAG). The MC, with members
representing Baltimore City, Baltimore County, Anne Arundel
County, and the State, managed and oversaw the IEMP and set its
overall policy direction. The TAG, composed of technical
managers from the City of Baltimore, the two counties, the State,
as well as representatives from the Maryland Regional Planning
Council and the academic community, advised the MC on the
technical and scientific aspects of the project, oversaw and
commented on all EPA and consultant work, and recommended and
suggested issues to study. EPA provided administrative,
technical, and analytical support. In the second phase of the
project, a Risk Assessment Review Panel (RARP), consisting of
scientists from John Hopkins University, advised the committees
on scientific and technical questions related to public health.
Also in Phase II, special workgroups with members from both the
TAG and representatives from industry, public interest groups,
government, and academia were organized around each priority
issue.
The major task in Phase I of the Baltimore IEMP was to
identify environmental issues of concern in the study area and to
set priorities among them for further study and development of
control strategies in Phase II. The Baltimore IEMP set
priorities on the basis of available information, supplemented by
data from a brief ambient air monitoring effort conducted by the
EPA. Five environmental topics were chosen for further
examination in Phase II of the Baltimore IEMP. In addition to
ambient air toxics, these issues were:
• Multi-media metals, which focused on developing cost-
effective techniques for lead paint removal and dust
abatement in Baltimore homes
• Underground storage tanks, which identified the ground-
water resources at greatest potential risk if underground
tanks leak
• Indoor air pollution, which sought to develop the
information necessary to support discussion of possible
programs to reduce exposure to indoor air pollution and
to support the expansion of local government capability
to respond to inquiries concerning indoor air pollution
• Baltimore Harbor, which defined current and future uses
of the Harbor's waters and identified actions, additional
research, and institutional arrangements necessary to
help environmental decision makers improve water quality
and habitat in the Harbor to achieve the desired uses
The results of these studies are summarized in separate reports
available from EPA's RID.
-------
AMBIENT AIR TOXICS STUDY OBJECTIVES
The air toxics study was designed to test a new approach for
using risk assessment to set research and control priorities
among sources and pollutants contributing to the complex mixture
of ambient air toxics (often referred to as urban soup) in the
Baltimore area. The Baltimore air toxics workgroup identified
four major project objectives:
• To compile the best possible air toxics emissions
inventory given available resources
• To estimate annual average ambient air concentrations of
selected air toxics resulting from emissions from both
point (e.g., industrial) and area (e.g., cars and dry
cleaners) sources; these compounds include volatile
organic compounds (VOCs), semi-volatile products of
incomplete combustion (polycyclic organic matter), and
metals
• To estimate increased risks to human health posed by
these compounds and by individual sources and source
categories
• To demonstrate an approach for establish priorities by
developing and analyzing control strategies to reduce the
adverse health and environmental effects from emissions
of selected air toxics
In addition, the workgroup saw the Baltimore IEMP ambient
air toxics study as a complement to the air toxics strategies
recently developed by EPA and the Maryland AMA. EPA's national
air toxics strategy was first announced in June 1985 and was
envisioned as a three-pronged effort for routine releases
consisting of:2
• An enhanced and refocused air-toxics control program for
problems of national scope, building on the existing
National Emissions Standards for Hazardous Air Pollutants
(NESHAPs) program
• A new federal program to build state capability to deal
with air toxics within their boundaries
• An expanded effort to devise strategies to reduce risks
from multi-media/multi-source pollutants in specific
localities where problems may exist
2Source: U.S. EPA, "A Strategy to Reduce Risks to Public
Health from Air Toxics," June 1985.
-------
The IBMP in its design and implementation was viewed as having
its largest impact on the third component of the strategy.
However, the extensive data collection activities that accompany
an IEMP would also provide State and local officials with the
tools they would need to start building their own programs.
The air toxics program developed by the AMA comprises three
activities and is consistent with the EPA air toxics strategy:
• Continued enforcement of federal programs, such as
NESHAPS regulations
• Recently adopted state regulations for industrial point
sources
• An urban air toxics initiative to address risks from
exposure to complex chemical mixtures (i.e., urban soup)
The workgroup saw the Baltimore ambient air toxics study
complementing the AMA's activities on its urban air toxics
initiative by providing insights on the relative risks for
different types of sources and pollutants, as well as the
relative costs, effectiveness (e.g., health risk reduction), and
societal benefits of alternative control strategies for reducing
these risks.
AMBIENT AIR TOXICS STUDY DESIGN
The Baltimore air toxics study consisted of three major
activities: (1) an initial pollutant selection designed to focus
the air toxics study on the most important pollutants; (2) a
screening investigation of the risks associated with exposure to
the selected air toxics, and (3) a demonstration of several
approaches for identifying and setting priorities among
alternative strategies to reduce the risks.
Pollutant Selection
The pollutant selection resulted in two sets of compounds:
one for air dispersion modelling and the other for ambient air
monitoring. The modelling was performed as part of the Baltimore
ambient air toxics study. Air monitoring was conducted by EPA's
Office of Research and Development (ORD) as part of its Baltimore
TEAM study. Although the Baltimore IEMP was not part of the
Baltimore TEAM study, we coordinated with ORD to take advantage
of the opportunity to obtain additional site-specific air
3Code of Maryland Regulations 26.11.15, Toxic Air
Pollutants.
-------
monitoring data. In addition to the TEAM monitoring data, we
gathered readily available monitoring data for the Baltimore
area. These data were drawn from sampling programs conducted by
the AMA over the past three years and the IEMP during Phase I of
the Baltimore IEMP (1983/1984).
We used the following criteria to select compounds for
modelling:
• Likelihood of exposure or presence in the ambient air
• Existence of CAG unit cancer risk factors or threshold
values for noncancer effects*
• Feasibility of control
The selection of pollutants for monitoring considered primarily
the goals of the TEAM study.3
A Screening Investigation of Air Toxics Risks
After completion of the pollutant selection, rough estimates
of risks to human health were developed using the monitoring and
modelled data, standard EPA exposure assumptions, available dose-
response determinations made by EPA, and information on exposed
populations from the U.S. Census. Several different measures of
risk were calculated in this study to provide insight on the
relative importance of the pollutants and sources examined.
First, at the recommendation of the Johns Hopkins Risk
Assessment Review Panel (RARP), we separately ranked the
pollutants and sources examined in the Baltimore IEMP by their
cancer potency-weighted ambient concentrations. The cancer
potency-weighted ambient concentrations were developed by
multiplying estimated average ambient air concentrations—either
by pollutant or source—by the appropriate unit risk factor(s).
This approach is advantageous because it (1) avoids relying on
4The unit cancer risk factor combines estimates of cancer
potency with EPA standard exposure assumptions. The unit cancer
risk factor represents the upper-bound estimate of the increase
in cancer risk associated with continuous lifetime (i.e., 70
years) exposure to 1 ug/m3 of a specific compound.
5Some of the goals of the Baltimore TEAM study, a study
designed to assess the significance of indoor versus outdoor
exposures to selected chemicals, included: sampling chemicals
examined in past TEAM studies; investigating chemicals that have
not been adequately studied in the indoor environment; and
contributing new information on indoor pollutant levels in light
of the extensive data already available.
-------
highly uncertain exposure assumptions used by EPA in its risk
assessments (e.g., continuous exposures for 70 years), and (2)
does not incorporate population weighting factors. The obvious
downside to this ranking scheme is that it cannot be used to set
priorities in populated versus unpopulated areas.
The second approach ranked the pollutants and sources
examined using rough quantitative estimates of risk. This latter
analysis considered both cancer and noncancer effects. For area-
wide cancer risks, the most important risk measures were average
increased lifetime individual cancer risk and annual excess
cancer incidence. For so-called "hotspot" areas, the two cancer
risk measures of interest were maximum increased lifetime
individual cancer risk and annual excess cancer incidence in the
grid cell of highest predicted annual excess cancer incidence.6'7
Noncancer risks were evaluated in two ways. First, we
calculated in the increased concern for several noncancer effects
(liver toxicity, kidney toxicity, reproductive, neurological,
fetal, and blood) by dividing the predicted or measured ambient
air concentration of each pollutant by the no-effect threshold(s)
relevant to that pollutant. If the resulting ratio exceeded one,
we identified these exposures as a concern for noncancer effects.
Second, we developed a "hazard index" that summed individual
pollutant ratios by effect category. The latter analysis was
aimed at examining the impact of exposure to complex chemical
mixtures in the ambient air. As the index approaches one,
concern for the potential hazard of the chemical mixture
increases. If the index exceeds one, the concern is the same as
if a no-effect threshold were exceeded by the same amount by an
individual pollutant.
We emphasize that the risk estimates presented in this
report are highly uncertain and were calculated only for
screening purposes. For example, the preliminary emissions
estimates used in the modelling effort were based on limited
information that was available in 1985. Since this time,
emission estimates have changed and important new sources have
been introduced. Because of these changes, the numbers and the
conclusions presented below may no longer be valid. Also,
because of the generally conservative bias in the underlying data
used in the risk assessment, it is highly unlikely that the true
6See Chapter III for detail on the different risk measures
and their estimation.
7The term "hotspot" refers to specific areas in which people
are exposed to higher than average concentrations of the target
pollutants.
8
-------
risks would be as high as the risk estimates, and they could be
considerably lower.
A Demonstration of Risk Management Strategies
Because of the considerable uncertainties underlying our
risk estimates, as well as the limited resources available for
developing appropriate control options and costs, we decided that
the risk management phase of the ambient air toxics study should
be conducted as a demonstration. To this end, we used cost-
effectiveness and benefit-cost analysis to evaluate individual
control options and control option strategies to reduce the risks
examined.
A set of risk-based criteria were established to identify
the sources included in the risk management demonstration:
• The source must contribute approximately 1 percent of the
estimated excess annual cancer incidence
• The source is projected to pose an increased lifetime
individual cancer risk of 5 x 10"6
• The source accounts for at least 1 percent of the
estimated annual excess cancer incidence at the grid
cell of highest incidence
For the selected sources, we identified feasible control
options and estimated preliminary costs and pollutant removal
efficiencies. The removal efficiency estimates considered
individual toxic compounds, as well as co-control of particulates
and volatile organic compounds (VOCs).8 Actual costs and removal
efficiencies for a control at a particular facility may differ
from the engineering estimates we generated. Thus, a detailed
site-specific analysis is required to confirm whether an
identified option is practically feasible.
In the cost-effectiveness analysis, pollutant-specific
removal efficiencies were used to quantify the reduction in risk
afforded by each control option. For the benefit-cost analysis,
we used available EPA studies to quantify the economic value of
reducing each ton of particulate and VOC associated with the
control options evaluated. A computer (mixed integer program)
model was used to identify the least-cost mix of control options
that would achieve either a specified level of risk reduction or
monetary benefits.
8Particulates and VOCs are called criteria pollutants
because they serve to define ambient air quality under the Clean
Air Act.
-------
CONCLUSIONS
Limitations
To properly interpret the results presented below, the
following major limitations must be considered:
• This study was based on rough emissions data developed at
the start of the project; budget and time constraints
prevented us from updating the risk estimates when new or
better emissions data became available.
• Because of conservative assumptions regarding the potency
of compounds studied and the length of exposure (i.e., an
assumed continuous lifetime (70 year) exposure), it is
highly unlikely that the true risks would be as high as
the estimates, and they could be significantly lower.
• We identified control options only for the purpose of
developing rough ranges of expected costs associated with
controls and the possible reductions in emissions that
they realize; additional analysis will be needed to
confirm whether these options are economically and
technically feasible.
Because of these uncertainties, the results of the Baltimore
ambient air toxics study are best used to indicate which
pollutants and sources deserve further investigation.
Emissions Inventory
• We were successful in compiling a rough emissions
inventory for the purpose of developing screening risk
estimates. The rough emissions inventory was built using
available data from the Maryland AMA and engineering
estimates generated by EPA's technical contractor. The
risk estimates based on this inventory allowed us to
develop a preliminary ranking of pollutants and sources
deserving closer scrutiny.
• further work is needed to develop an emissions inventory
that could support regulatory decisions. To this end,
several modifications are needed. New sources need to be
added to the inventory, and the emissions estimates at
existing sources must be changed to reflect current
controls and operating practices, especially at Point
Source A.
10
-------
Screening Risk Assessment Results; Cancer
Pollutants
We ranked the risk-potential of the pollutants examined in
the Baltimore IEMP using (1) the RARP approach and (2)
preliminary estimates of human health risk (both incidence and
individual risk) based on monitoring and modelling. The results
of this effort are shown in Table 1 and support the following
conclusions:
• POM, chromium VI. benzene, and perchloroethvlene rank
most consistently near the too. Looking across the
different ranking schemes and risk measures, POM,
chromium VI, and benzene are generally the top three
pollutants of concern.
• POM clearly ranks highest based on the modelled results,
while perchloroethvlene ranks highest based on the
monitoring data. It is interesting to note that
perchloroethylene ranks significantly lower (i.e., in the
bottom half) based on modelled data. POM was not
considered categorically in the monitoring study;
however, monitoring data were available for one of POM's
constituents, B(a)P.
• Based on the RARP ranking. POM ranks five times higher
than the next closest pollutant, chromium VI. The
modelled results also show the same trend.
• The available area-wide monitoring indicates that
potential perchloroethvlene risks are twice as high as
the next highest ranked pollutant, chromium VI.
Continuing down the rank, perchloroethylene risks are
about three times higher than benzene, about seven times
higher than chloroform, and over ten times higher than
all other pollutants included in the area-wide
monitoring. However, it should be noted that the
perchloroethylene monitoring data are highly uncertain.
• The monitoring data show that all pollutants individually
pose average increased lifetime individual cancer risks
greater than 1 x 10'". The risk results based on source
modelling indicate that at the "hotspots," only two-
thirds of the pollutants examined pose maximum increased
lifetime individual cancer risks greater than 1 x 10" .
11
-------
• Based on monitoring, three pollutants—perchloroethvlene,
hexavalent chromium, and benzene—account for
approximately 87 percent of the estimated area-wide
average increased lifetime individual cancer risk. The
relative contribution of each of these three compounds
is: perchloroethylene (48 percent), hexavalent chromium
(24 percent), and benzene (15 percent).
• The modelling results indicated that POM alone accounts
for about two-thirds of the area-wide average increased
lifetime individual cancer risks. Chromium VI
contributes an additional 15 percent, and a variety of
other pollutants account for 10 percent or less. It is
interesting to note that when looking at the maximum
increased lifetime individual cancer risks, chromium VI
poses the highest concern, followed by POM.
Sources
Similar to the risk ranking performed for individual
pollutants, we evaluated and ranked the potential risks posed by
selected sources and source categories in the Baltimore study
area. Two ranking schemes were established using (1) the RARP
risk-ranking approach and (2) preliminary estimates of
quantitative risk (both incidence and individual risk). Because
of the extreme difficulty in relating ambient air concentrations
from monitoring back to individual release points, the ranking of
sources was based solely on modelled data.
• Based on the RARP ranking approach, which indicates the
potential for a source category to pose a risk to human
health. Point Source A ranked approximately seven times
higher than the next highest source, road vehicles.
Point Source A also ranked over ten times higher than all
other point sources combined.
• The preliminary risk estimates show that point and area
sources account for similar percentages of the estimated
annual excess cancer incidence. Point and area sources
contribute approximately 45 percent and 55 percent of the
total, as shown in Figure 2. Most of the area source
fraction is attributable to road vehicles (33 percent).
• The rouohlv even contribution of point and area sources
to the estimated annual excess cancer incidence does not
hold for individual pollutants. This is easily seen when
the two key pollutants in the modelling analysis of
cancer incidence, POM and chromium VI, are examined more
closely. Whereas both point and area sources account for
similar portions of the estimated annual excess cancer
incidence associated with POM (approximately 44 percent
versus 56 percent, respectively), point source emissions
13
-------
FIGURE 2
SOURCE CATEGORIES CONTRIBUTING TO ANNUAL
EXCESS CANCER INCIDENCE IN THE BALTIMORE
AREA
100%
80%
60%
40%
20%
0%
5%
7%
1%
Point Sources Road Vehicles Heating Solvent Usage Other
Source Category
-------
account for almost all (99.6 percent) of the estimated
annual excess cancer incidence associated with hexavalent
chromium.
• Approximately 80 percent of the estimated annual excess
cancer incidence is attributable to road vehicles,
heating, and Point Source A. The other point sources
evaluated (Point Sources B, C, and D) account for an
additional 10 percent; solvent usage and other sources,
primarily minor point sources, account for the final 10
percent.
• As shown in Figure 3, point sources were the malor
contributors to the average increased lifetime individual
cancer risks. Point sources account for almost all of
the risk at those locations with the highest estimated
lifetime individual cancer risks.
• As shown in Figure 4. point sources account for roughly
92 percent of the maximum increased lifetime individual
cancer risks at the three "hotspots" evaluated in the
Baltimore IEMP. Road vehicles accounts for an additional
5 percent.
the estimated maximum increased lifetime individual
cancer risk at two of the three "hotsoots." At the third
hotspot, Point Sources A, B, and C together account for
roughly 83 percent of the increased lifetime individual
cancer risk. Road vehicles and heating account for the
remainder.
Preliminary Cancer Risk Estimates
Before summarizing our preliminary cancer risk results,
several limitations are noted:
• This study estimated the increase in cancer resulting
from exposure to ambient (i.e.. outdoor) concentrations
of air toxics. It did not consider exposures resulting
from (1) indoor air, (2) the workplace, and (3) other
pathways (e.g., ingestion).
• This study used conservative estimates of increased
cancer risk to establish priorities among pollutants and
sources. The risk estimates are calculated using
modelled and monitored concentrations and EPA unit cancer
risk factors. There is considerable uncertainty in the
estimated concentrations, which could either overstate or
understate the true concentrations (see Chapter IV).
Unit cancer risk factors combine CAG potency estimates
with EPA exposure assumptions. The CAG potency estimates
15
-------
FIGURE 3
Baltimore IEMP Air Toxics
Point & Area Source Contribution to Average
Lifetime Individual Cancer Risk by Grid Cell
Average Lifetime Individual Cancer Risk
Point Source Contribution to Average Lifetime
individual Cancer Risk
Area Source Contribution to Increased
Individual Cancer Risk
0.000
UTU COORDINATES
u: The risk scale for the area source contribution
i^ntficanui larger than that used for the avenge
ifeame individual cancer risk and point source contribution
jphs.
16
-------
FIGURE 4
SOURCE CATEGORIES CONTRIBUTING TO
MAXIMUM INCREASED LIFETIME INDIVIDUAL
CANCER RISKS IN THE BALTIMORE AREA
100%
80%
60%
40%
20%
0%
-
/
/
/
92%
5% 1% 1% 1%
/ A , ,
1 . V / . / / . / / . /
Point Sources Road Vehicles Heating Solvent Usage
Source Category
Other
-------
provide a plausible upper limit to the cancer risk of a
compound (see Appendix A); however, the true value of the
risk is unknown and may be as low as zero. The exposure
assumptions are extremely conservative in that they
assume continuous exposure to outdoor air for 70 years.
Because of the generally conservative bias in the
information/ it is highly unlikely that the true risks
would be as high as the estimates/ and they could be
considerably lower.
• Based on the model performance evaluation, the dispersion
model consistently underpredicted ambient air
concentrations for all pollutants considered, except
trichloroethvlene. Predicted ambient air concentrations
of benzene and benzene-ring compounds (e.g., xylene,
toluene, and ethyl benzene) are likely to be
underestimated by at least a factor of two. Estimated
concentrations of trichloroethylene may have overstated
by a factor of two. The predicted values for the other
compounds evaluated (e.g., chloroform, carbon
tetrachloride, lf2-dichloroethane, 1,2-dichloropropane)
are best seen as lower estimates of actual ambient air
concentrations. Consequently, our exposure estimates
based on modelling are understated; however, our
estimates of risk using these modelled values are still
more likely to overstate than understate true risk.
Bearing these important limitations in mind, the following
risk results are presented:
• Based on modelling, the estimated average increased
lifetime individual cancer risks in Baltimore are
generally in the same range as those predicted for other
urban study areas. The comparison was made using results
from: the Philadelphia IBMP; the Kanawha Valley, West
Virginia, IEMP; the Santa Clara Valley, California, IEMP;
and the South Coast Air Quality Management study of the
Los Angeles basin. The estimated maximum increased
lifetime individual cancer risks tended to be lower in
Baltimore than in these other areas.
• Based on modelling, the area-wide average increased
lifetime individual cancer risk was fairly uniform across
all sections of the study area, approximately 1.5 x 10' .
The heavily industrialized southeast section of the study
area had the highest average lifetime individual cancer
risk.
18
-------
0 The available area-wide ambient air monitoring indicated
an eatimated average increased lifetime individual cancer
risk of 3-Q x 10.This result is approximately a
factor of three higher than that predicted by the
modelling.
• The average increased lifetime individual cancer risks at
the two TEAM monitoring sites were eatimated to be
approximately 1.5 x 10'" and 1.2 x IP"1, respectively.
These results are very close to the estimated area-wide
averages based on modelling.
• The maximum increased lifetime individual cancer risks at
the three "hotsoots" evaluated were estimated to range
from 5.4 x IP"1 to 1.3 x 10"*. These results are
approximately four to nine times higher than the
estimated area-wide average individual risks.
• Based on modelling and area-wide monitoring, four and
twelve annual excess cancer cases were predicted,
respectively. The difference between the estimates can
be explained largely by higher monitoring-based exposure
estimates for perchloroethylene, hexavalent chromium, and
benzene.
• The modelling results indicate an excess of 0.15 annual
cancer cases in the arid cell of highest predicted cancer
incidence. This grid cell accounts for less than 1
percent of the study area and approximately 4 percent of
the area-wide estimated excess annual cancer incidence.
• The average increased individual cancer risk in the grid
cell of highest incidence, while twice the area-wide
average, is less than the average increased individual
risk estimates in other arid cells. Figure 5 shows that
the grids with the highest average increased lifetime
individual cancer are not necessarily the grids with the
highest annual excess cancer incidence. This information
indicates that population density can be a more important
factor than exposure levels in identifying high-incidence
areas.
Air Toxics Risk Perspective
• The increased cancer risks from ingestion of chloroform,
a trihalomethane. in drinking water are comparable to
those from air toxics. The annual excess cancer
incidence from ingestion of chloroform in drinking water
was estimated at approximately three cases; the average
increased lifetime individual cancer risk was computed
roughly at 1 x 10"*. More detail on this issue can be
found in Chapter V.
19
-------
FIGURE 5
Baltimore IEMP Air Toxics
Annual Excess Cancer Incidence, Average Lifetime
Cancer Risk and Exposed Population by Grid Ceil
Annual Excess Cancer Incidence
Average Lifetime Individual Cancer Risk
Exposed Population
20
-------
• In our comparison of the approach used to quantify POM
risks in the Baltimore IBMP and the BfalP surrogate
approach used in the Sixth Month Study, we found that the
B(alP surrogate approach resulted in greater estimates of
annual excess cancer incidence (6.8 versus 2.1\, but
lower estimates of the increaaed lifetime individual
cancer risks for the three hotaoot sites (Hotspot #1;
7.91 x 10"J va. 3.7 x 10'"; Hotsoot *2t 1.4 x 10"* va. 6.3 x
10'"-. and Hotspot »3t 1.3 x 1Q'J vs. 6.3 x IQ'M. The
reason for this seeming contradiction is that the B(a)P
surrogate approach uses only one unit cancer risk factor
for all emissions of particulates that result from
incomplete combustion of organic matter. In contrast,
the unit cancer risk factors we used for the semi-
volatile fraction of the product of incomplete combustion
in the Baltimore IBMP are specific to types of sources of
combustion products.
Screening Risk Assessment Results» Noncancer
As with our cancer risk assessment, our findings for
noncancer effects are preliminary and subject to several
important caveatss
• This study used preliminary no-effect threshold values to
estimate noncancer risks. No-effect threshold values for
six categories of non-cancer effects were developed by
EPA toxicologists by pollutant. These values are
extremely uncertain and have not all undergone extensive
peer review.
• The no-effect threshold value for blood effects is
especially uncertain. This threshold value had not yet
undergone peer review and is, thus, subject to change.
If this no-effect threshold value is found to be accurate
(2.45 ug/m3 or approximately 0.8 ppb), monitoring data
across the country indicate that the benzene no-effect
threshold value for blood effects is exceeded in almost
every urban area in the country.
Our major findings on the need to further explore noncancer
effects are summarized below:
• The modelled data indicate that there is a need to
further examine benzene and xvlene exposures in the study
area. The refined (2.5 km) grid analysis shows benzene
ambient concentrations exceed the no-effect threshold for
blood effects in the southeast quadrant of the study
area. The discrete receptor analysis supports the need
to investigate benzene. It also identifies a section of
21
-------
the study area in which xylene exposures may need to be
looked into more carefully because they exceed the no-
effect thresholds for liver, kidney, reproductive,
neurological, fetal, and blood effects.
• The available monitoring data and the TEAM data for the
two fixed sites both show that benzene exposures exceed
the threshold for blood effects. Further work is needed
to confirm our measured concentrations of benzene.
• The noncancer risk results were driven primarily bv
exposure to a single pollutant, and not mixtures.
Further work is needed to verify the noncancer threshold
values to confirm this finding.
Risk Management; A Demonstration
• we identified control options only for the purpose of
developing rough ranges of expected costs associated with
controls and the possible reductions in emissions that
thev realize. Additional analysis will be needed to
confirm whether these options are economically and
technically feasible.
• Preliminary annualized costs of individual controls
ranged from a savings of S220.000 for the recovery of
trichloroethvlene from decreasing operations to S150
million for converting residences currently burning oil
for heat to natural gas. Control options and their costs
were identified using available information and
engineering judgment. Additional site-specific analysis
would be required to confirm the feasibility of these
options.
• Eight control options analyzed in this project could
result, if implemented, in savings while reducing risks.
These control options are designed to recover solvents.
• Of the control options that did not result in a net
savings, control potion 4 at Point Source C achieved the
most incidence reduction per dollar spent. However, this
option would reduce only 0.2 excess cancer cases every 70
years.
• Implementing either control option 2 or control option 3
at Point source B would result in approximately 10 cancer
cases avoided over a 70-vear period. Of the control
options that will not result in a net savings, these
options are some of the most cost-effective. Option 3
for miscellaneous industrial use of methylene chloride is
22
-------
also fairly cost-effective, but will only reduce three
cancer cases over a 70-year period.
Relatively small reductions in the area-wide annual
excess cancer incidence were identified using coat-
effectiveness analysis. The control strategies
identified by the model to lower the estimated annual
excess cancer incidence ranged from a reduction of
approximately 7 percent at a cost savings (attributable
to solvent recovery at area sources) to a reduction of 24
percent at an annual cost of cost about $211 million.
Roughly one case every two years could be avoided at a
cost of approximately $1.4 million.
The control strategies identified would reduce the annual
excess cancer incidence at the location of the highest
estimated annual excess cancer incidence 10.15 excess
cases per year) between 39 and 43 percent. This area
accounts for less than 1 percent of the study area
evaluated.
We can achieve significant reductions in the maximum
increased lifetime individual cancer risk (approximately
46 percent^ at an annualized cost of S46 million. The
most important costs would be borne by a single facility.
Also, after control the resulting maximum increased
lifetime individual cancer risk, about 7 x 10**, would
still be significantly above the average area-wide
lifetime individual cancer risk of about 1 to 2 x 10"*.
The strategies for reducing annual excess cancer
incidence and the maximum increased lifetime individual
cancer risk would reduce benzene ambient levels below the
no-effect threshold for noncancer effects. However, the
strategies considered in this study would not reduce
xylene ambient air concentrations, which also caused
concern for noncancer health effects.
Controlling emissions of air toxics, and considering the
co-control for ozone precursors and particulates. can
lead to net health and welfare benefits of approximately
SB.5 million per year, while reducing annual excess
cancer incidence bv approximately 16 percent. Selection
of control strategies to achieve a risk reduction level
above 19 percent will result in benefits failing to keep
pace with the control cost increases.
23
-------
I. INTRODUCTION
1. INTEGRATED ENVIRONMENTAL MANAGEMENT PROJECTS
This report describes the second phase of a two-phase study
of human health risks from air toxics in the greater Baltimore
area. The results of the first phase of the study are presented
in Baltimore Integrated Environmental Management Project; Phase I
Report. The study was conducted as part of the Baltimore
Integrated Environmental Management Project (IEMP). The
Regulatory Integration Division (RID) of the Environmental
Protection Agency (EPA) initiated the project as part of its
pursuit of new approaches to environmental management and policy.
The purpose of the IEMP was to identify and assess the
significance of selected environmental issues, to set priorities
for action among these issues, and to analyze alternative
approaches to manage the risks that are assessed.
EPA adopted the concept of integrated environmental
management as a potential solution to the shortcomings of the
traditional approach for pollution control. The traditional
approach of focusing on one pollutant or class of pollutants
within each medium at a time may result in environmental programs
and regulations that do not use resources as efficiently as
possible. Grounded in the concepts of risk assessment and risk
management, the IEMP uses estimates of risk—that is, the
probability of adverse effects—as a common measure for comparing
and setting priorities among environmental issues that involve
different pollutants, sources, and exposure pathways and that may
affect human health, ecosystems, and resources. The need to set
priorities is prompted by the realization over the past ten years
that hundreds of chemicals present in our environment can pose
some risk of causing cancer or other adverse health effects.
Comparing risks helps to establish priorities. It allows
environmental managers to focus limited resources in a manner
that will achieve the greatest public benefit. The structure of
the lEMPs is also built on the concept that these projects should
be managed at the state and local level, thus ensuring that
issues of greatest local concern or risk are adequately
addressed. The Baltimore IEMP was particularly successful in
this regard.
I-l
-------
2. ORGANIZATION OF THE PROJECTS
Each IEMP is divided into two phases. In the first, project
managers establish the decision-making structure of the project,
identify key environmental issues, and set priorities among them.
Risk is but one of the criteria used in ranking issues; the
others include analytical feasibility, relevance to EPA, state
and local program objectives, and the potential for effective
mitigation. In the second, the IEMP studies the priority issues
in greater detail and develops and evaluates potential strategies
for their control or resolution.
3. THE BALTIMORE IEMP
The Baltimore IEMP was a cooperative effort involving the
governments of the State of Maryland, the City of Baltimore,
Baltimore County, Anne Arundel County, and EPA. A map of the
study area is presented in Figure 1-1. The Baltimore area was
chosen, not because it has a significant toxics problem, but
because EPA and local officials wanted to explore better ways to
identify, assess, and manage the potential human health risks of
environmental pollutants in the area. In addition, Baltimore was
attractive because of the opportunity to establish a management
structure with the potential to continue after completion of the
project. This project represents the second of four, full-scale
geographic projects initiated to date. The others are the
Philadelphia IEMP, the Santa Clara, California IEMP, and the
Denver IEMP.
The decision-making structure of the IEMP consisted of two
committees that also served as vehicles for State and local
participation! the Management Committee (MC) and the Technical
Advisory Committee (TAG). The MC, with members representing
Baltimore City, Baltimore County, Anne Arundel County, and the
State, managed- and oversaw the IEMP and set its overall policy
direction. The TAG, composed of technical managers from the City
of Baltimore, the two counties, the State, as well as
representatives from the Maryland Regional Planning Council and
the academic community, advised the MC on the technical and
scientific aspects of the project, oversaw and commented on all
EPA and consultant work, and recommended and suggested issues to
study. EPA provided administrative, technical, and analytical
support. In the second phase of the project, a Risk Assessment
Review Panel, consisting of scientists from Johns Hopkins
University, advised the committees on scientific and technical
questions related to public health. Also in Phase II, special
workgroups with members from both the TAG and representatives
from industry, public interest groups, government, and academia
were organized around each priority issue.
1-2
-------
FIGURE 1-1
BALTIMORE IEMP STUDY AREA
PRINCE
GEORGE'S
CO.
1-3
-------
4. THE RESULTS QF THE PHASE I PRIORITY-SETTING PROCESS
The major task in Phase I was to identify environmental
issues of concern in the study area and to set priorities among
them for further study and development of control strategies in
Phase II. The Baltimore IEMP set priorities on the basis of
available information, supplemented by data from a brief ambient
monitoring effort conducted by EPA.1
Five environmental topics were chosen for further
examination in Phase II of the Baltimore IEMP:
1) Ambient air toxics. The goal was to investigate and
demonstrate alternative approaches for reducing risks
from ambient air toxics in urban areas like Baltimore
at the least cost to society.
2) Multi-media metals. The goal was to develop cost-
effective techniques for lead paint removal and dust
abatement in Baltimore homes.
3) Underground storage tanks. The goal was to develop a
strategy for identifying those ground-water resources
at greatest potential risk if underground tanks leak.
4) Indoor air pollution. The goal was to develop the
information necessary to support (1) discussion of
possible programs to reduce exposures to indoor air
pollution and (2) the expansion of local government
capability to respond to inquiries concerning indoor
air pollution.
5) Baltimore Harbor. The goals were to define current and
future uses of the Harbor's waters and identify
actions, additional research, and institutional
arrangements necessary to help environmental decision
makers improve water quality and habitat in the Harbor
to achieve the desired uses.
5. RELATIONSHIP TO THE EPA AND AMA AIR TOXICS STRATEGIES
While this report summarizes primarily the results of the
Baltimore IEMP air toxics study, it is important to note the ways
in which the Baltimore IEMP—especially the Baltimore air toxics
'please see Chapter IV of Baltimore integrated Environmental
Management Prolect: Phase I Report (hereafter referred to as the
Phase I report) for a detailed account of the priority-setting
process in the first phase of the IEMP.
1-4
-------
study—has complemented the air toxics strategies recently
developed by EPA and the Maryland Air Management Administration
(AMA).
EPA's national air toxics strategy was first announced in
June 1985 and was developed considering both routine and
accidental releases.2 For routine releases of hazardous air
pollutants (the focus of the IEMP, as well as the AMA's program),
EPA's strategy was envisioned as a three-pronged effort
consisting of:
(1) An enhanced and refocused air toxics control program
for problems of national scope, building on the
existing National Emissions Standards for Hazardous Air
Pollutants (NESHAPS) program
(2) A new federal program to build state capability to deal
with air toxics within their boundaries
(3) An expanded effort to devise strategies to reduce risks
from multi-media/multi-source pollutants in specific
localities where problems may exist
The IEMP in its design and implementation can have its largest
impact on the third component of the strategy. However, the
extensive data collection activities that accompany an IEMP can
also provide state and local officials with the tools they need
to start building their own programs.
The air toxics program developed by Maryland's Air
Management Administration (AMA) consists of three activities:
(1) continued enforcement of federal programs, such as NESHAPS
regulations, (2) recently adopted state regulations for
industrial point sources, and (3) an urban air toxics initiative
to address risks from exposures to complex chemical mixtures
(sometimes referred to as "urban soup") in metropolitan areas
like Baltimore.3 This program is very consistent with the EPA
air toxics strategy. The Baltimore IEMP air toxics study has its
greatest influence on the AMA urban air toxics initiative. It
complements the State's activities in this area by providing
insights on the relative risks for different types of sources and
pollutants, as well as the relative cost-effectiveness of
alternative control strategies for reducing these risks.
2For further information, see: U.S. EPA, "A Strategy to
Reduce Risks to Public Health from Air Toxics," June 1985.
3The recently adopted industrial point source regulations
can be found in: Code of Maryland Regulations 26. 11. 15, Toxic
Air Pollutants.
1-5
-------
6. OVERVIEW OF THE REPORT
This report is organized into six chapters including this
introduction. These chapters are:
• Chapter lit Ob1actives and Design of the Baltimore
Ambient Air Toxics Study
• Chapter III; Summary of the Principles of Risk
Assessment and Risk Management as Used in the Ambient Air
Toxics Study
• Chapter IV: Selecting Pollutants. Evaluating Toxicitv.
and Estimating Exposure
• Chapter V; Risk Assessment Screening Results
• Chapter VI; Risk Management Strategies for Reducing
Risks; A Demonstration
In addition, this report contains an Executive Summary that
summarizes the objectives and findings of the Baltimore IEMF
Ambient Air Toxics Study. An Epilogue follows the Executive
Summary that presents -the Ambient Air Toxics Workgroup's insights
on the EPA/State relationship/ the overall IBMP process/ and
technical issues. Appendixes are included to provide more detail
on some of the supporting data. Additional background
information is referenced throughout the report and is available
from EPA's RID.
1-6
-------
II. OBJECTIVES AMD DESIGN OF THE BALTIMORE AMBIENT AIR TOXICS
STUDY
This chapter presents an overview of the Baltimore air
toxics study, focusing on the study objectives, the study design,
and how the results of this study should be viewed.
1. STUDY OBJECTIVES
The Baltimore IEMP air toxics study was designed to test a
new approach for using risk assessment to set control and
research priorities among sources and pollutants contributing to
the complex mixture of ambient air toxics (often referred to as
"urban soup") in the Baltimore area.
The Ambient Air Toxics Workgroup identified four major
project objectives of the Baltimore air toxics study:
• To compile the best possible air toxics emissions
inventory given available resources
• To estimate annual average ambient air concentrations of
selected air toxics resulting from emissions from both
point (industrial) and area (e.g., cars and dry cleaners)
sources; these compounds included volatile organic
compounds (VOCs), semi-volatile products of incomplete
combustion (polycyclic organic matter), and metals
• To estimate increased risks to human health posed by
these compounds and by individual sources and source
categories
• To demonstrate alternative approaches for establishing
priorities by developing and analyzing control strategies
to reduce the adverse health and environmental effects
from emissions of selected air toxics
These objectives would allow us to answer the following
analytical questions:
• Can we develop a screening methodology to identify and
characterize health risks posed by selected air toxics
from exposures occurring in the "urban soup?"
II-l
-------
What is the relative contribution of point sources versus
area sources to urban air toxics in the Baltimore area?
What is the relative importance of specific point or area
sources to urban air toxics risk?
What is the relative importance of specific air
contaminants, by source, to urban air toxics risk?
What are the relative costs and effectiveness of
alternative control options to reduce human health risk
from air toxics in the Baltimore area?
Will the control of carcinogenic emissions lead to
reduced .concern for noncarcinogenic health effects from
air toxics?
Will the control of air toxic emissions also result in
lower emissions of criteria air pollutants, in particular
precursors to ozone, such as volatile organic compounds
(VOCs) and particulates?
What are the health and welfare benefits that accrue from
reduced emissions of VOCs and particulates which result
from controlling air toxics emissions at specific
sources?
2. STUDY DESIGN
The Baltimore air toxics study consisted of three major
activities: (1) an initial pollutant selection designed to focus
the air toxics study on the most important pollutants; (2) a
screening investigation of the risks associated with exposure to
the selected air toxics, and (3) a demonstration of alternative
analytical approaches for identifying strategies to reduce the
risks (i.e., cost-effectiveness and benefit-cost analysis). In
this section, we describe briefly the general steps that we
performed as part of each of these activities. The pollutant
selection process and results are discussed in more detail in
Chapter IV. The results of the exposure and risk
characterizations are detailed in Chapters IV and V,
respectively. Finally, Chapter VI summarizes the approach and
results of our demonstration project to explore alternative risk
management strategies.
a. Pollutant Selection
Because of time and resource constraints, the first task in
the air toxics study was to identify the most important
pollutants that could be examined in detail. Two sets of
compounds were identified for the purpose of air dispersion
II-2
-------
modelling and for ambient air monitoring. These lists were
compiled using data we collected on emissions, exposures,
potencies, the potential for control, and criteria developed by
the Workgroup.
b. Risk Characterization of the Selected Pollutants
After completing the pollutant selection, we focused our
efforts on collecting the additional data that would allow us to
perform a quantitative risk assessment. As discussed in Chapter
III, there are four steps in risk assessment: hazard
identification, dose-response evaluation, human exposure
evaluation, and risk characterization. The Baltimore IEMP did
not perform the first two steps. We relied on EPA determinations
regarding hazard identification and dose-response for our
studies. To quantify human exposure, information was needed on
ambient air levels, duration of exposure, and the number of
exposed individuals.
We generated data on ambient levels using results from
predictive (i.e., fate and transport) modelling and ambient air
monitoring. To use dispersion models, one must have information
on pollutant emissions by source and meteorological conditions.
We generated the emissions data by compiling an emissions
inventory of point and area sources using available information
from the AMA and engineering estimates prepared by our technical
contractor. These data were quality controlled and entered into
our data management system, PIPQUIC.
The monitoring data used in this study were collected by
(1) EPA's Office of Research and Development as part of a special
monitoring study called TEAM (Total Exposure Assessment
Methodology)1 that they were conducting in the Baltimore area,
(2) EPA's Regulatory Integration Division during Phase I of the
Baltimore IEMP, and (3) AMA as part of their routine and special
project toxics monitoring.
To calculate exposure to the estimated ambient air
concentrations, we relied on standard EPA exposure assumptions,
which are highly conservative in that they assume continuous
exposure to predicted concentrations for a lifetime of 70 years.
Chapter III and Appendix A provide more information on the EPA
exposure assumptions. Information on the population exposed by
'The focus of the TEAM study is to develop methods for
quantifying exposures to pollutants present in the immediate
vicinity of an individual as he or she goes about his or her
daily activities. Readers interested in this study should
contact U.S. EPA, Office of Research and Development,
Environmental Monitoring Systems Laboratory, for more
information.
II-3
-------
geographic area was obtained from the 1980 U.S. Census at the
block grid/enumeration district level.
The risk characterization combined all of the information
generated as part of the above steps. For carcinogenic effects/
we estimated (1) maximum increased lifetime individual cancer
risk/ (2) average increased lifetime individual cancer risk/ and
(3) annual excess cancer incidence.
For noncancer effects/ we calculated the ratio of the
estimated ambient air concentration to the no-effect threshold
for individual pollutants. An exceedance of the threshold
indicates that pollutant-specific exposures require additional
investigation. Because individuals in urban environments are
exposed to complex chemical mixtures, we also explored whether
exposure to multiple pollutants with the same noncancer effects
required additional study using a "Hazard Index." The hazard
index is calculated simply by summing the ratios of estimated
ambient air concentration to the no-effect threshold for all
pollutants with the same noncancer effect. If the index exceeds
one, there is a need to further examine these exposure levels.
c. Risk Management: A Demonstration
Because of the considerable uncertainties underlying our
risk estimates, as well as the limited resources available for
developing appropriate control options and costs, the Workgroup
decided that the risk management phase of this study should be
conducted as a demonstration. To this end/ we analyzed different
approaches for identifying strategies to reduce the risks
examined.
For the sources selected as the focus of our demonstration,
we identified feasible control options and estimated their
associated costs, efficiencies, and benefits by employing
engineering estimates and EPA technical documents developed to
support various regulatory activities. We then packaged these
options into strategies for consideration by local risk managers
using two approaches: cost-effectiveness and benefit-cost
analysis.
3. HOW THE RESULTS OF THE AIR TOXICS STUDY SHOULD BE VIEWED
The Phase II study of air toxics in the Baltimore area was
intended as a demonstration of an analytical approach for ranking
strategies to reduce the increased risk from toxics in the
II-4
-------
ambient air.2 We made no determinations regarding the absolute
magnitude of risk nor the need for regulatory action to reduce
emissions of air toxics, in general. The risk estimates are
intended only for ranking risks and establishing pollutant and
source category priorities for more detailed evaluation. Because
of conservative assumptions (that is, protective of public
health) regarding the potency of compounds studied and the length
of exposure (i.e., an assumed continuous lifetime (70 year)
exposure), it is highly unlikely that the true risks would be as
high as the estimates, and they could be significantly lower.
We identified control options only for the purpose of
demonstrating a risk management approach. Rough ranges of
expected costs associated with controls and the possible
reductions in emissions that they realize were used in the
demonstration. Economic and engineering considerations may
preclude their use for the specific sources identified.
Finally, we emphasize that this study was based on rough
emissions data developed at the start of the project. Budget and
time constraints prevented us from updating the risk estimates
when new or better emissions data became available. Thus, the
results summarized in this report may not reflect current
conditions and are best used to indicate where additional
analysis is needed to verify our estimated exposures and risks.
2This study estimated the increase in cancer and noncancer
risk resulting from exposure to ambient (i.e., outdoor)
concentrations of air toxics. It did not consider exposures
resulting from (1) indoor air, (2) the workplace, and (3) other
pathways (e.g., ingestion).
II-5
-------
III. SUMMARY OF THE PRINCIPLES OF RISK ASSESSMENT AND RISK
MANAGEMENT AS USED IN THE AMBIENT AIR TOXICS STUDY
The Baltimore IBMP air toxics study was a demonstration of
techniques for making decisions on policies that address
environmental and human health risks. These techniques fall into
the two categories of environmental decision-making: risk
assessment and risk management. Below, we provide a general
overview of each of these concepts. We then describe the key
tasks associated with risk assessment and risk management as
applied to the Baltimore IEMF air toxics study.
1. RISK ASSESSMENT AND RISK MANAGEMENT
As recently defined by the National Academy of Sciences
(NAS), risk assessment is the scientific activity of evaluating
the toxic properties of a chemical and the conditions of human
exposure to it both to ascertain the likelihood that exposed
humans will be adversely affected, and to characterize the nature
of the effects they may experience.1 A risk assessment can be
either quantitative, emphasizing a reliance on numerical results,
or qualitative. For cancer, the Baltimore IEMP uses quantitative
risk assessment to establish priorities among pollutants and
sources. For noncancer health effects, we can only identify
whether exposures—either modelled or measured—exceeding an
estimated threshold are generally more appropriate subjects for
further investigation because of data limitations.
It is important to emphasize that our risk estimates were
intended to help decision makers understand the relative
importance of different problems and set priorities among them.
They serve as aids in developing public policies for controlling
environmental risks. The risk assessments do not examine disease
incidence in the local population and then attempt to link
incidence with environmental exposures. This was not an
epidemiological study; the risk assessments were not intended to
and do not answer questions such as what caused a statistically
higher rate of cancer in one neighborhood or part of the
community. Nor do they make a definitive statement concerning
*Risk Assessment in the Federal Government; Managing the
Process. (Washington, D.C.: National Academy Press, 1983).
Additional information in this section is drawn from Principles
of Risk Assessment; A Nontechnical Review. (U.S. EPA, 1985).
III-l
-------
the absolute risk posed by a particular substance, pollutant, or
exposure pathway. Rather the data provide a rough, but
conservative indication of potential human health risks. The
actual risk is not likely to be much higher than the estimated
risk, but it could be considerably lower. The reader is referred
to Appendix A for a more detailed discussion of the issues
surrounding EPA's method of quantifying cancer risks.
In contrast to risk assessment, the NAS describes risk
management as a process of evaluating (1) whether an assessed
risk is sufficiently high to present a public health concern and
(2) the appropriate response to the potential chronic health
hazard. The risk management process entails consideration of
political, social, economic, and engineering information during
the evaluation process. It also involves value judgments,
especially when viewing the acceptability of risk and the
reasonableness of the costs of control.
The sections that follow describe in more detail the steps
involved in completing a quantitative risk ^assessment for the
selected air toxics studied in Baltimore. This chapter concludes
with a discussion of the demonstration project that was conducted
to assist in the risk management phase of the air toxics study.
2. STEPS IN RISK ASSESSMENT
Risk assessment can be divided into four major steps:
• Hazard identification
• Dose-response evaluation
• Human exposure evaluation
• Risk characterization
Each of these steps is discussed below in the context of the
Baltimore air toxics study.
a. Hazard Identification
Hazard identification involves reviewing relevant human
epidemiological studies, animal experiments, and toxicological
tests bearing on whether or not an agent can cause an increase in
the incidence of a health condition (e.g., cancer, birth
defects). It is important to note that in the Baltimore IEMP, we
relied solely on the hazard determinations already made by EPA.
In deciding the potential hazard of a compound, EPA has
evaluated the available studies using sound biological and
statistical considerations and procedures. In addition, it has
III-2
-------
characterized the nature and strength of the evidence of
causation, which is an important component of the hazard
identification process. • For compounds associated with noncancer
effects, EPA has included uncertainty factors into the
calculation of an acceptable exposure level or threshold. These
uncertainty factors account for differences in animal and human
sensitivity, differences in human sensitivities, and differences
in experimental versus actual (or expected) exposure levels. For
agents identified as potential human carcinogens, EPA has
developed a weight-of-evidence stratification scheme that
reflects a synthesis of the conclusions drawn from the different
pieces of information on the carcinogenicity of the agent. There
are five levels of the EPA stratification scheme:
• Group A: Human Carcinogen
• Group Bt Probable Human Carcinogen
• Group C: Possible Carcinogen
• Group D: Not Classified
• Group E: No Evidence of Carcinogenicity for Humans
Appendix A discusses these groups in more detail.
Agents that are judged to be in the EPA weight-of-evidence
stratification groups A and B are regarded as suitable for
quantitative risk assessments. Agents judged to be in group C
are generally regarded as suitable, but judgments in this regard
are made on a case-by-case basis. For the agents assigned to
groups A, B, and, to a lesser extent, C, a quantitative
dose-response relationship has been developed.
b. Dose-Response Evaluation
A dose-response assessment is the process of characterizing
the relationship between the dose of an agent administered or
received and the extent of toxic injury or disease. The
information used in this type of assessment is derived from
animal studies or, less frequently, from studies in exposed human
populations. Thus, in most cases, the dose-response evaluation
involves extrapolation between species and exposure levels (i.e.,
from high to low doses). In completing our risk assessment for
selected air toxics in the Baltimore area, we relied primarily on
the "dose-response relationships" or potency estimates already
developed by EPA for three effect categories: cancer, cancer
caused by exposures to complex mixtures, and noncancer. Each of
these is described below.
III-3
-------
i. Cancer
EPA'3 Carcinogen Assessment Group (CAG) relies heavily on
data from animal tests in determining the toxicity (especially
carcinogenicity) of chemicals and in estimating their potency in
producing adverse health effects.2 The reasons are as follows:
• Human testing is not possible for ethical reasons.
• Use of animals as proxies for humans is consistent with
current scientific knowledge of the mechanism of
carcinogenicity in humans, although certain species of
animals may not be suitable models for the action of
certain chemicals. Known human carcinogens, with the
single possible exception of arsenic, have been found to
be carcinogenic in some animal system. Although the
reverse is not necessarily true, it does indicate the
potential for carcinogenicity in a mammalian species.
Consequently, "in the absence of adequate data on humans,
it is reasonable, for practical purposes, to regard
chemicals for which there is sufficient evidence of
carcinogenicity in animals as if they presented a
carcinogenic risk to humans."3
• Epidemiological studies of the chronic health effects
(such as cancer) of long-term environmental exposures to
chemicals of concern are problematic; they require
extensive high-quality data on historical exposures and
causes of disease and death and can generally only be
applied to situations where exposures have occurred for
twenty or more years.
• A high-quality negative epidemiological study, while
useful, cannot prove the absence of an association
between chemical exposure and human cancer. For the
population and levels of exposure studied, it can
determine an upper-limit or range for the estimates of
risk and the statistical likelihood of the study to
detect an effect.
• Understanding the mechanism of action of a chemical in
producing an adverse health effect through studies on
animals is important in determining its toxicity and
2For more information on the methodology used by CAG in its
does-response evaluation, please see: "Guidelines for
Carcinogenic Risk Assessment, as published in the Federal
Register. September 24, 1986.
3 Preamble, International Agency for Research on Cancer
(IARC) monograph.
III-4
-------
potency at environmental exposures, as opposed to
occupational exposures for which most epidemiological
data exist.
The use of animal data in assessing the carcinogenicity of
chemicals and developing potency estimates is not without
problems. There are major uncertainties in extrapolating from
animals to humans. There are important species differences in
uptake, metabolism, and organ distribution of carcinogens, as
well as species and strain differences in target-site
susceptibility. And human populations are variable with respect
to genetic constitution, diet, occupational and home environment,
activity patterns, and other cultural factors.
Another major area of uncertainty involves estimating the
potency of chemicals at the low environmental exposures found in
IBMP studies from the high doses or exposure levels observed in
occupational or experimental settings—a process called "low-dose
extrapolation." For practical reasons, responses at low doses
cannot be measured directly either by animal experiments or by
epidemiological studies. Low-dose extrapolation must, therefore,
be based on current understanding of the mechanisms of
carcinogenesis.
In estimating the potency of compounds at low environmental
exposure levels, CAG uses mathematical models that are designed
to produce conservative estimates. The potency estimates made on
the basis of these models (where actual exposures are consistent
with exposure assumptions that underlie the model) should be
regarded as representing the plausible upper-bound for the risks;
the true value of the risk is unknown and may be as low as zero.
They are not best estimates of cancer potency.4 We emphasize
that use of the upper-bound potency estimates does not
necessarily give a realistic prediction of actual risks to
humans. In addition, because we present risk estimates as sums
of the separate risks from multiple compounds, the uncertainty
regarding the final estimate is commensurately increased.
The potency estimates are generally expressed as "unit
cancer risk factors." The unit cancer risk factor combines
estimates of cancer potency with EPA standard exposure
assumptions. The unit cancer risk factor represents the upper-
bound estimate of the increase in cancer risk associated with
continuous lifetime (i.e., 70 years) exposure to 1 ug/m3 of a
specific compound.
* Alternative procedures for cancer risk assessment exist
that account for conflicting cancer risk estimates. However, the
suitability of the procedure depends on the uses to which the
risk estimates will be used.
III-5
-------
ii. Cancer Caused by Exposures to Complex Mixtures
The complex mixtures of chemicals produced by the incomplete
combustion of fossil fuels (e.g., automobile exhaust, coke oven
emissions, smoke from woodburning stoves) have long been known to
contain compounds, such as polycyclic organic matter (POM) which
includes benzo(a)pyrene, a human carcinogen. However, these
mixtures present special problems in estimating carcinogenic
potency because combustion of fuels can produce thousands of
organic gaseous and particulate pollutants, as well as inorganic
materials, such as metals. Also, the relative composition of
these pollutants within these mixtures, and hence potency of the
overall mixture, varies with fuel source. Testing and developing
potency estimates for every constituent of the products of
incomplete combustion from a source represents a formidable task.
Previous studies9 used one constituent of a carcinogenic
class of compounds within the complex mixture, such as
benzo(a)pyrene (BaP), as a surrogate for all carcinogenic species
and applied its potency estimate to the class, such as POMs. An
alternative approach assumed that the carcinogenic potency of all
combustion products was the same, regardless of the source.6
Neither of these approaches has been found to be consistent with
current scientific understanding of the relationship of
mutagenicity and human carcinogenicity or with the relative
composition of complex mixtures.
In the Baltimore ambient air toxics study, we used the
comparative potency method. Scientists estimate the risk of a
suspect carcinogen (e.g., diesel emissions), for which there are
no epidemiological cancer data, by comparing the potency of the
agent in mutagenicity and carcinogenicity bioassays to those of
known human carcinogens. The known human carcinogens are coke
oven, roofing tar, and cigarette smoke tar emissions. The
following equation is used in the example for diesel emissions:
Estimated Hunan Risk* • Human Risk, x B
Where:
d = diesel
c = coke ovens
B = the relative bioassay potency of the bioassay potency
(diesel) to the bioassay potency (coke ovens)
3 One example of such a study is U.S. EPA, Santa Clara
Vallev Integrated Environmental Management Protect; Revised Stage
One Report. Office of Policy, Planning, and Evaluation, May 30,
1986.
6 U.S. EPA, Office of Policy Analysis, The Air Toxics
Problem in the United States: an Analysis of Cancer Risks for
Selected Air Toxics. U.S. EPA, May 1985.
III-6
-------
The relative bioassay potency is obtained by calculating the
ratio of the observed mutagenic activity in tests of diesel
emissions to the observed mutagenic activity in tests of coke
oven emissions (the slopes of the dose response functions from
the same in vitro or in vivo bioassays).
iii. Noncancer Health Effects
The noncancer health effects we examined included systemic
effects, such as liver and kidney toxicity, fetal developmental
(teratological), neurobehavioral, and blood effects. Only
effects associated with long-term exposures were considered.
They cover a very broad range of severities with regard to human
health—from life-threatening to only mildly debilitating.
Agents that give rise to these effects are often referred to by
EPA as "systemic toxicants" because they affect the functioning
of the various organs. Based on scientists' understanding of
physiological mechanisms, these systemic toxicants are assumed to
have identifiable exposure thresholds (both for the individual
and the population) below which the adverse effects are not
observable.
EPA uses animal data to calculate these thresholds.
Threshold values, expressed in terms of mg/kg/day are called
"reference doses" (RfOs). As discussed above, the estimates
include uncertainty factors to account for differences in animal
and human sensitivity, differences in human sensitivities, and
differences in experimental versus actual (or expected) exposure
levels.
EPA has established a method for calculating RfDs for
exposures that occur only through ingestion—oral RfDs. Thus,
there are no Agency-approved inhalation RfDs for target
pollutants in this air toxics study. We have calculated
inhalation thresholds using either Agency-approved oral RfDs, if
they existed for our target pollutants, or threshold values
estimated by RID toxicologists using the same method employed by
EPA in developing oral RfDs. The procedures we followed have
considerable precedence at the Agency.
Because the effects for which the thresholds were developed
cover a wide gamut of impacts on human health, exceedance of a
threshold generally does not imply a life-threatening situation.
Furthermore, it is assumed that exposures occur over a lifetime
for most effects (excluding fetal developmental) that could be
caused by environmental exposures. Our exposure assessment
represents only a snapshot in time and not the kind of analysis
necessary for an epidemiological study. The exceedance of a
threshold for an individual pollutant indicates the need to
further investigate this compound.
III-7
-------
In Baltimore, as in any urban environment, environmental
exposures generally involve numerous compounds. To account for
exposures to complex chemical mixtures, one can also sum the
threshold ratios for pollutants that contribute to the same
noncancer health effect. By summing these ratios, one creates a
so-called "Hazard Index," which is discussed in more detail later
in this chapter. If the estimated Hazard Index exceeds "1," it
identifies exposures that need further exploration.
We refer the reader to Appendix A for a more comprehensive
discussion of the analytical approach for developing thresholds.
c. Human Exposure Evaluation
This step in the risk assessment process involves measuring
or estimating the intensity, frequency, and duration of human
exposures to a compound in the environment. It can also require
determining the size of the exposed population. There are two
major ways the Baltimore IEMP estimated exposures to chemical
compounds in the urban air: (1) dispersion modelling of ambient
concentrations, and (2) monitoring of ambient concentrations.
These approaches are traditionally used by air quality programs
and have been adopted by the IEMP.
The IEMP approach requires knowing the concentrations of a
chemical in the ambient air at a residential location. The
information on concentrations at receptor sites was coupled with
assumptions about how ambient concentrations relate to actual
human exposures, or dose, to arrive at an estimate of exposure
for an individual or population. We followed EPA standard
assumptions, called exposure constants, that account for the
amount of air or water a typical person takes in each day, and
the weight of an average person. Specifically, EPA assumes that
an average person weighs 70 kilograms, breathes 20 cubic meters
of air and drinks two liters of water each day.
We estimated ambient air exposures at fixed points
throughout the study area using dispersion (i.e., predictive
transport and fate) modelling and monitoring at fixed outdoor
sites. Ambient monitoring data often provide the most direct
measure of environmental conditions. However, there are two
important limitations to their use:
• Monitoring data cannot generally be used to predict
maximum levels as data are often taken from a few
specific points and must be extrapolated over the area of
concern. In contrast, modelling can take into account
geographic variability and can pinpoint the sites of
7 Versar, Inc., A Screening Methodology for Air Quality
Analysis. U.S. EPA, April 1985 draft.
III-8
-------
maximum ambient concentrations. Modelling can also be
used to evaluate source contribution.
• Monitoring data most often cannot link observed ambient
concentrations to sources, which is essential for making
risk management decisions. Phase II of the Baltimore
IEMP focused on gathering the data needed to identify
sources.
Consequently, for the purpose of estimating risks at the highest
exposure levels, we relied primarily on dispersion modelling. To
estimate risks at average exposure levels, we used both modelled
and monitoring data. We used monitoring data to evaluate how
well our models predict actual concentrations. It is difficult,
to use modelling alone to predict total ambient exposures.
Because background levels of pollutants are rarely known, they
cannot be incorporated into the dispersion model when it is run
to estimate ambient air concentrations. The result is a range of
likely concentrations, bordered by our modelled and monitored
values.
d. Risk Characterization
Risk characterization is the process of estimating the
incidence of a health effect under various exposure scenarios.
Generally, it involves the integration of the data and analysis
stemming from the three steps of risk assessment described above.
The type of risk characterization that we performed for
carcinogens versus systemic toxicants was somewhat different, as
detailed below.
i. Cancer
Cancer risk is defined as either 1) the increased
probability that an individual continuously exposed for a
lifetime to one or more chemicals will contract cancer, referred
to in this report as either the maximum or average increased
lifetime individual cancer risk, or 2} where a number of people
are likely to be exposed, the excess number of cancer cases per
year in the population exposed to one or more chemicals, referred
to as the annual excess cancer incidence. Where simultaneous
exposures to multiple compounds is assumed, these risk estimates
for individual compounds are then summed to produce a cumulative
estimate of risk. The methods used to calculate each risk
measure are discussed separately.
Maximum Increased Lifetime Individual Cancer Risk. We
define the maximum increased lifetime individual cancer risk as
the increased probability that an individual exposed to the
greatest amount of one or more chemicals will contract cancer
over a lifetime (70 years). Generally, this calculation is
computed by multiplying the maximum ambient concentration at a
III-9
-------
discrete receptor location (generated through our air dispersion
modelling exercise) by the unit cancer risk factor. In the
Baltimore IEMP, a discrete receptor location was defined as a
residential area. Thus, high concentrations in areas not defined
as residential by U.S. Geological Survey maps or by visual siting
were not considered possible maximum individual risk sites.
increased KMIM™ Unit Cancer
Lifetime Individual - Concentration^ x Risk Factor!
Cancer Risk
Where :
i = Target pollutant i
k = Receptor location k
We then summed the maximum increased lifetime individual cancer
risks for each target pollutant to estimate the total maximum
increased lifetime individual cancer risk from all study
pollutants .
Average Increased Lifetime Individual Cancer Risk. We
define the average increased lifetime individual cancer risk as
the increased probability that an individual exposed to the
average amount of a chemical will contract cancer over a lifetime
(i.e., 70 years). We calculated the average increased lifetime
individual cancer risk for each pollutant by multiplying the
average ambient pollutant concentration in a defined area by the
exposure constant and the appropriate unit cancer risk factor.
These values were summed to arrive at the aggregate average
increased lifetime individual cancer risk posed by all target
pollutants .
The estimates of average ambient air concentration were
based on both measured and modelled data. The results of the
dispersion modelling were used to estimate average increased
lifetime individual cancer risks for the study area, as a whole.
The results of the available ambient air monitoring were used to
estimate average increased lifetime individual cancer risks at
each of the fixed monitoring site locations, as well as for the
study area.
Annual Excess Cancer Incidence. We define annual excess
cancer incidence as the excess number of cancer cases per year in
a defined population resulting from specified environmental
exposures. Annual cancer incidence is calculated by dividing the
8The unit cancer risk factor combines estimates of cancer
potency with EPA standard exposure assumptions. The unit cancer
risk factor represents the upper-bound estimate of the increase
in cancer risk associated with continuous lifetime (i.e., 70
years) exposure to 1 ug/m3 of a specific compound.
111-10
-------
lifetime excess cancer incidence by 70, the assumed human
lifetime. The lifetime excess cancer incidence is calculated by
multiplying the average increased individual cancer risk for each
pollutant, as discussed above, by the population and summing
across all pollutants. These calculations are represented by the
following equations:
Annual Excess - V^ Lifetime Excess ' 70 years
Cancer Incidence Z-* Cancer Incidence!
1
-or-
Annual Excess • V^ V^ Average Increased Estimated • 70
Cancer L^ Z-* Individual Cancer * Popula-
Incldence 1 j
Where t
i = Target pollutant i
j = Either grid cell j (dispersion modelling) or the study
area (available monitoring data)
The annual excess cancer incidence was calculated for the
entire study area and the grid cell of highest predicted
incidence. Dispersion modelling was used to generate area-wide
annual excess cancer incidence estimates, as well as to identify
the grid cell of maximum annual excess cancer incidence. As
discussed in more detail in Chapter IV, the dispersion model was
programmed to generate estimates of ambient concentration for a
series of small boxes or grid cells that are overlayed on the
study area.
As a policy check on the incidence results generated using
the dispersion modelling results, we developed rough estimates of
annual excess cancer incidence for the study area using the
available area-wide monitoring data. The incidence calculations
were performed for each pollutant by multiplying the measured
concentration (averaged across all monitoring locations) by the
appropriate unit cancer risk factor (which combines the cancer
potency scores and the exposure constants) and an estimate of the
total exposed population in the study area. Although there are
numerous limitations with this approach, the results can provide
the decision maker with additional insight on the range of
potential human health risks posed by selected air toxics using
both monitoring and modelled data.
ii. Noncancer
There are two risk measures for noncancer health effects.
For single pollutants, the measure of noncancer health effects is
the ratio of the estimated pollutant ambient air concentration —
based on either modelling or monitoring — to its no-effect
threshold. Exposures that are less than the no-effect threshold
Ill-ll
-------
are not likely to be associated with noncancer health effects and
are, therefore, less likely to be of regulatory concern.
Conversely, as the frequency of exposures exceeding the threshold
increases and as the size of the excess increases, the
probability also increases that adverse effects may be observed
in a human population. Such exceedances indicate the need to
examine the exposures more carefully. Because a single substance
may have many different thresholds for the different health
effects associated with it, we examine each health effect for
each substance separately. We can also explore the need to
further investigate exposures to multiple compounds that have the
same noncancer effect, as discussed below.
Because most environmental exposures (especially in urban
areas) involve concurrent or sequential exposures to a complex
mixture of compounds that may induce similar or dissimilar
effects over time, we can calculate a "hazard index" to
assess the potential hazard from cumulative exposures to the
target compounds. The hazard index is used to indicate when
simultaneous exposures to multiple pollutants with the same
systemic effects could require additional investigation, although
exposures to each individual pollutant did not exceed the no-
effect threshold. The index is based on the assumption of dose
additivity and is defined by the following equation:
HI - Ei/AIi! + B2/ALj + ...+
Where :
HI = Hazard index for a particular category of health
effect
Et = Ambient concentration of pollutant i
ALj. = Threshold value for pollutant i for a particular
health effect category
After first calculating the ratio of the ambient concentration
for the pollutant to the threshold for the health effect
(Et/ALt), we summed these ratios across all pollutants with the
same effect.
In this study, the hazard index is intended to be a
numerical indication of whether exposures to complex mixtures of
pollutants in the environment deserve additional study. As the
index approaches 1, concern for the potential hazard of the
chemical mixture increases. If the index exceeds 1, the concern
is the same as if a no-effect threshold were exceeded by the same
amount by an individual pollutant, thus indicating the need to
investigate further the observed exposure levels.
9 U.S. EPA, Environmental Criteria and Assessment Office,
Guidelines for the Health Risk Assessment of Chemical Mixtures.
Final Report, September 1985, pp. 12-14.
111-12
-------
3. INTERPRETING RISK ASSESSMENT RESULTS IN THE BALTIMORE AIR
TOXICS STUDY
We have already highlighted some of the uses and limitations
of risk assessment. We restate and elaborate these points below
to ensure that the reader properly interprets the estimates of
risk presented in Chapters V and VI of this report:
• The estimates of maximum and average increased lifetime
individual cancer risk and annual excess cancer incidence
from exposure to toxics should not be interpreted as
precise or absolute estimates of future health effects.
The simplifying assumptions and uncertainties in both the
toxicology and exposure components are simply too great
to justify a high level of confidence in the precision of
the results. They are approximations of the potential
upper-bound risks to human health and are designed to
permit relative comparison across different sources,
pollutants, and exposure pathways.
• The potency and threshold estimates used in this study
are consistently conservative (i.e., upper-bound) in the
direction of overestimating risk for the particular
pollutants and exposure pathways we have assessed. There
is the possibility that the cancer risk from any or all
of the pollutants could be zero.
• We may understate increased risks from ambient air toxics
exposures to the extent that we have not considered all
pollutants, sources, and exposure scenarios (such as
short-term exposures to very high levels). Current
technology, for example, does not allow scientists to
identify all compounds in air or water. This is because,
first of all, monitoring techniques are selective in the
kinds of compounds they can effectively and reliably
capture. Second, laboratory analysis to determine the
identity and quantity of monitored samples must be
targeted at specific chemical or types of chemicals of
concern. Modelling, on the other hand, is limited by the
quality of the data on emissions of pollutants. And
third, even if one were able to identify all compounds to
which one is being exposed through a particular medium,
one would have only a rough notion of the toxicity of
only a very small fraction of the total number of
compounds.
l°This study estimated only the increase in cancer and
noncancer risk resulting from exposure to ambient (i.e., outdoor)
concentrations of air toxics. It did not consider exposures
resulting from (1) indoor air, (2) the work place, and (3) other
pathways (e.g., ingestion).
111-13
-------
4. RISK MANAGEMENT
A major focus of the Phase II IEMP work was in demonstrating
an approach to identify the best (namely, the most cost-
effective) strategy for managing the estimated risks. The
National Academy of Sciences defines risk management as the
process of evaluating whether an assessed risk is sufficiently
high to present a public health concern and the appropriate
response to the potential human health hazard. Risk management
considers not only the level of risk posed by a particular
pollutant or source of pollution but also the feasibility and
cost of control, public preferences, and institutional
capabilities. Setting priorities for research into the risk
posed by pollutants is also a key aspect of risk management. Key
activities are investigating control alternatives, estimating
their efficiency in controlling pollution, developing estimates
of their cost, and then assessing their cost-effectiveness.
In the Baltimore air toxics study, the criteria of
cost-effectiveness (i.e., the cost of control per unit of human
health risk reduced) and benefit-cost (i.e., the cost of control
per dollar of benefits accruing to society) were used to evaluate
alternative control strategies. Benefits estimates were included
in this analysis to capture additional positive outcomes from
controlling air toxics (other than risk reduction) that may
offset the costs of control. For example, controlling benzene
emissions from cars can also lead to the reduction of volatile
organic compounds (VOCs) that cause smog. These analyses can
consider risks that result from exposure to primary as well as
secondary (intermedia) releases. However, the Baltimore air
toxics study considered only primary ambient air releases.
We developed rough information on the costs and efficiencies
for a set of potentially feasible control options. We then
ranked these control options by their cost-effectiveness, as well
as their benefits. This information was used in a mathematical
model, referred to as a mixed integer program, to obtain the
lowest cost combination of control options to achieve a given
level of benefits or risk reduction. The control options and the
computer model are discussed in more detail in Chapter VI and
Appendix C.
The estimates of the costs of control are, like the risk
estimates, highly uncertain. They are best used in developing a
relative ranking of control strategies, not in generating
definitive estimates of the resources necessary to solve the
selected environmental problems.
111-14
-------
IV. SELECTING POLLUTANTS. EVALUATING TOXICITY. AND ESTIMATING
EXPOSURE
In this chapter, we describe the activities undertaken to
select the study pollutants and characterize the risks and
exposures associated with these pollutants. First, we discuss
the approach we used to identify the study pollutants, including
a description of the data gathering activities that supported the
pollutant screening. Second, we present what is known about the
toxicity of each compound evaluated. This discussion combines
the risk assessment steps of hazard identification and dose-
response evaluation. It also discusses the weight-of-evidence
issue, which is important to consider when viewing the potential
toxicity of each pollutant examined. We then summarize how we
assessed the extent to which Baltimore residents could be exposed
to the target pollutants in the ambient air. The results of the
risk assessment for the study pollutants are presented in
Chapter V.
1. POLLUTANT SELECTION METHODOLOGY AND RESULTS
We selected two sets of compounds to analyze: one for air
dispersion modelling and the other for ambient air monitoring.
The air dispersion modelling was performed as part of the
Baltimore IBMP ambient air toxics study. Air monitoring was
conducted by EPA's Office of Research and Development (ORD) as
part of its Baltimore TEAM study.1 Although the Baltimore IEMP
was not part of the Baltimore TEAM study, we coordinated with ORO
to take advantage of the opportunity to obtain additional site-
specific air monitoring data. In addition to the TEAM monitoring
data, we gathered all readily available monitoring data for the
Baltimore area. These data were drawn from sampling programs
conducted by the AMA over the past three years and the IEMP
during 1983/1984.
'The TEAM study is a study designed by EPA's Office of
Research and Development to assess the significance of indoor
versus outdoor exposures to selected chemicals. Readers
interested in this study should contact U.S. EPA, Office of
Research and Development, Environmental Monitoring Systems
Laboratory for more information.
iv-1
-------
Different factors influenced our choice of study compounds
for dispersion modelling and ambient air monitoring. For the air
dispersion modelling/ we were constrained by the set of chemicals
for which we had estimates of emissions from sources. Even for
these chemicals, we could not always model a compound with any
reasonable degree of accuracy (e.g., a compound, such as
formaldehyde, may be expected to be produced in significant
amounts from other compounds through atmospheric processes). For
the ambient air monitoring conducted by ORD, we were limited by
sampling technology and resources. Furthermore, we were
constrained to select compounds for monitoring which could be
expected to be found in both the outdoor and indoor environments
to complement the Baltimore TEAM study mentioned earlier.
In what follows, we describe the criteria used to select
compounds for air dispersion modelling, the procedure followed to
choose compounds included in the TEAM ambient air monitoring
program, and the final list of pollutants analyzed in the
Baltimore air toxics study.
a. Pollutant Selection for Modelling
We used the following three criteria in selecting compounds
for modelling:
1. Likelihood of exposure or presence in the ambient air.
To make this determination, we developed a rough
emissions inventory of air toxics from both point and
area sources. We also used available monitoring data.
The following sources were used:
i. 1983 Survey for the Maryland Toxic Substance
Registry fTSRl. This survey (referred to in this
document as the 1983 TSR survey) compiled
information on toxic substances used, produced or
handled by industry. The State's 1985 TSR survey
had not yet been completed when the pollutant
selection was completed.
ii. 1983 IEMP monitoring data. As discussed in the
Phase I report, IEMP sampled for ten VOCs for a
period of three months starting in November of
1983.
The rough inventory—which was entered into a computer
database that could be managed through the IEMP data
management system (PIPQUIC)—includes process and boiler
emissions from approximately 250 point sources, as well
as emissions from area sources (e.g., dry cleaners and
degreasers) and nontraditional sources (e.g., wastewater
IV-2
-------
treatment plants).2 The inventory was a compilation of
data generated by the AMA and EPA. More detail is
provided on the emissions inventory later in this chapter
under Section 3 entitled "Human Exposure Evaluation."
Existence of GAG unit cancer risk factors or threshold
values for noncancer effects.3The Agency has approved
numerous RfDs for the oral route of exposure, but none
for inhalation. Toxicologists from EPA's Regulatory
Integration Division (RID) and EPA's Air Office have
adjusted these oral RfDs, where possible, for the
inhalation route of exposure. We acknowledge that these
are not official EPA values and are useful only for
identifying sources and pollutants that warrant further
investigation. The unit cancer risk factors are
developed by EPA's Carcinogen Assessment Group.
Sources t
i. EPA's Cancer Assessment Group fCAGU CAG unit
cancer risk factors represent the Agency's
consensus on estimates for carcinogens and are
commonly used by all EPA's program offices.4
ii. EPA's Office of Air Quality Programs and Standards
(OAQPS1. OAQPS is charged with developing
regulations mandated by the Federal Clean Air Act
for controlling air toxics.
iii. Regulatory Integration Division toxicolooists.
RID toxicologists and health scientists estimated
thresholds for noncancer health effects based on
reviews of toxicological literature.
2An overview of the PIPQUIC system is contained in: PIPQUIC
User's Guide. American Management Systems, Inc. for U.S. EPA,
Office of Policy Analysis, Regulatory Integration Division,
November 1985.
3The unit cancer risk factor combines estimates of cancer
potency with EPA standard exposure assumptions. The unit cancer
risk factor represents the upper-bound estimate of the increase
in cancer risk associated with continuous lifetime (i.e., 70
years) exposure to 1 ug/m3 of a specific compound.
*The unit cancer risk factors can be obtained directly from
CAG or through ORD's Integrated Risk Information System (IRIS).
Information on IRIS may be obtained by contacting EPA's Office of
Health and Environmental Assessment.
IV-3
-------
3. Feaaibilitv of control. Pollutants with limited
potential for reducing exposure to the chemical through
engineering measures or changes in production processes
were excluded. For example, some toxics are emitted from
natural sources. Control of such emissions cannot be
achieved readily through an air toxics regulatory
program.
Sources:
i. Maryland's Air Management Administration engineers
ii. EPA contractors to the Baltimore IEMP air toxics
study
Source: EPA's Office of Research and Development.
To account for known differences in the quality of the data,
we used professional judgment in applying the above criteria. In
general, any individual compound met the first two criteria if 1)
over a metric ton per year was being emitted into the air and it
was carcinogenic or it had an RfD that could conceivably be
exceeded by environmental levels, or 2) it was emitted in
substantial quantities from a single source (over a metric ton
per year) and was toxic or carcinogenic.
In addition to individual compounds, we also chose to
examine emissions of the larger molecular weight, semi-volatile
class of compounds called polycyclic organic matter (POM) that is
produced by incomplete combustion of organic matter from a
variety of combustion processes. A recent nationwide study of
air toxics conducted by EPA had shown that the products of
incomplete combustion could be significant contributors to
airborne risk (U.S. EPA, Office of Policy Analysis, The Air
Toxics Problem in the United Statest An Analysis of Cancer Risks
for Selected Air Toxics, May 1985). Unfortunately, the unit
cancer risk factor that we have for POM from different categories
of sources are generally not of the same quality as for many of
the volatile organic compounds included in the study, as detailed
in Section 2 of this chapter. [The exception is the unit cancer
risk factors for POM from coke oven emissions which is based on
human epidemiological data.] Nevertheless, we decided that it
was preferable to provide rough estimates of potential risk from
particulate organics than to ignore this important source of risk
altogether.3
3Joellen Lewtas, "Combustion Emissions: Characterization
and Comparison of Their Mutagenic and Carcinogenic Activity," in
Hans Stich, ed., Carcinogens and Mutacrens in the Environment,
Volume V (CRC Press, Inc., Baton Rouge, 1985), pp. 59-72. Also,
Roy Albert, J. Lewtas, S. Nesnow, T. Thorslund, and E. Anderson,
IV-4
-------
The compounds that were selected for modelling are shown
in Table IV-1, ranked by rough estimates of total annual
emissions in the study area. Since the emissions data were drawn
from different EPA and State databases generated for different
purposes, they vary greatly in quality. The toxicities of the
compounds are discussed under Section 2 of this chapter.
b. Pollutant Selection for TEAM Monitoring
The monitoring data used in the Phase II Baltimore air
toxics study came from several sources; however, the only
monitoring that took place while the air toxics study was being
conducted came from ORD as part of their Baltimore TEAM study.
We coordinated with the ORD monitoring effort because we saw the
potential to use the TEAM monitoring data in our dispersion model
performance evaluation.6 To ensure success of this project, we
worked with ORD to develop a list of pollutants for sampling that
included many of the pollutants modelled in the IEMP ambient air
toxics study.
Building on previous indoor air toxics work, the Workgroup
and ORD identified the pollutants for monitoring starting with
four lists of compounds. The first had been compiled by Dr.
Joyce McCann at the University of California's Lawrence Berkeley
Laboratory. Under a contract with the U.S. Department of
Energy, Dr. McCann and her coworkers had conducted a
comprehensive review of data on indoor air pollutants. The list
she provided us contained those chemicals that were suspected to
be present in the indoor environment and were known to be toxic
at environmental levels, but whose concentrations had not been
adequately determined for the purpose of estimating risks to the
general population. The second was a list of compounds that the
Baltimore IEMP Indoor Air Workgroup had identified as candidates
for inclusion in the TEAM study. This list in turn was developed
on the basis of knowledge of chemicals that are used in common
consumer products and that are toxic to some degree. The third
list included a set of priority compounds developed by the Air
Toxics Workgroup using professional judgment. Finally, the
fourth was a list of compounds that TEAM had studied in previous
"Comparative Potency Method for Cancer Risk Assessment:
Application to Diesel Particulate Emissions," Risk Analysis. Vol.
3, No. 2, 1983.
6The model performance evaluation is discussed later in this
chapter (see Section 3(a)(iv)).
7Personal communication between Andrew Manale, EPA and Dr.
Joyce McCann, Lawrence Berkeley Laboratory and the University of
California, Berkeley, March 1986.
IV-5
-------
TABLE IV-1
CHEMICALS SELECTED FOR AIR DISPERSION MODELLING
IN THE BALIMORE IEMP AIR TOXICS STUDY
(Data Drawn from the 1983 Survey for the Maryland Toxic
Substances Registry and 1983 National Emissions Data System)
POLLUTANT
TOTAL
EMISSIONS
(kkg/yr)1
SINGLE SOURCE
EMISSIONS
(kkg/yr)
USE BY MAJOR
POINT SOURCES
(kkg/yr)
Oraanic oases
Toluene
Benzene
Tetrachloroethylene
Xylene
Methyl chloroform
Methylene chloride
Trichloroethylene
Formaldehyde
Glycol ethers
Ethyl benzene
Ethylene oxide
Chloroform
Phenol
Carbon tetrachloride
Ethylene dichloride
Ethylene dibromide
Oraanic Particulates
POM2
Benzo(a)pyrene
Metals
Chromium
Cadmium
Arsenic
5923.00
3038.00
2341.00
1762.00
1696.00
1097.00
769.00
709.00
638.00
232.00
45.00
30.00
19.00
3.80
3.35
1.71
N.A.
N.A.
6105.00
1.60
1.00
2056.44
2585.52
N.A.
83.40
16.00
77.36
197.70
82.60
15.64
4.46
66.16
51.80
0.01
6.40
N.A.
N.A.
N.A.
N.A.
2.30
1.40
0.40
22618.49
68839.70
974.33
20265.98
1817.85
938.68
457.23
1188.43
N.A.
N.A.
2530.63
15.88
897.22
1652.92
26.62
N.A.
N.A.
N.A.
46108.76
2430.84
83.92
Note: The data in this chart are of varying quality.
N.A. = not available
1 Metric tons/year
2 POM = polycyclic organic matter. POM includes carbon
containing molecules having more than one ring structure.
Polyaromatic compounds, such as benzo(a)pyrene, are included in
this class of compounds.
IV-6
-------
studies. Budget considerations caused EPA to restrict the number
of target compounds for TEAM monitoring to 27. These compounds
are listed in Table IV-2.
2. HAZARD IDENTIFICATION/DOSE-RESPONSE EVALUATION
We present below the hazard assessments for the substances
included in the Baltimore ambient air toxics study. Table IV-3
lists all compounds included in the study, as well as the effects
(i.e., cancer versus noncancer) that could result from exposure
to these substances.8 As discussed earlier in this report, we
relied primarily on hazard determinations and dose-response
evaluations already made by EPA. The one exception is for
polycyclic organic matter (POM), which results from the
incomplete combustion of organic material. This is because POM
represents a large class of compounds and because the
constituents of POM and their portion of the total composition of
emissions vary with source category. EPA recently identified
products of incomplete combustion as posing potentially
significant risks from exposure.9
a. Cancer Effects
i. Weight of Evidence
Before we present the dose-response information for the
carcinogenic compounds examined in this study, it is important to
emphasize that not all carcinogens are equal in terms of potency
or weight of evidence. As discussed in Chapter III, EPA has
developed a stratification scheme that classifies compounds into
one of five groups (i.e., A, B, C, D, E) based on weight of
evidence. In general, compounds that are judged as either Group
A or B carcinogens are regarded as suitable for quantitative risk
assessment. Group C carcinogens are also suitable for
quantitative risk assessment, but a decision to include these
pollutants must be made on a case-by-case basis. Because the
weight-of-evidence determination is critical information for the
decision maker, we have included this designation on all tables
summarizing risk results by pollutant. It was impossible to
assign an overall weight-of-evidence score in instances where we
8Several compounds included in the TEAM sampling were
excluded from the Baltimore ambient air toxics study because of
limited health effects information. These compounds are
limonene, decane, dodecane, nonane, octane, and undecane.
9For more information, see U.S. EPA, Office of Policy
Analysis, The Air Toxics Problem in the United States: An
Analysis of Cancer Risks for Selected Pollutants, May 1985. This
study is also referred to as the "Six Month Study."
IV-7
-------
TABLE IV-2
CHEMICALS SELECTED FOR ORD'S TEAM AMBIENT AIR MONITORING
AS PART OF THE BALTIMORE IEMP AIR TOXICS STUDY
STRAIGHT CHAIN
VOLATILE ALIPHATIC
ORGANICS METALS HYDROCARBONS
Benzene1 Arsenic1 Decane3
Carbon Cadmium1 Dodecane
tetrachloride2 Chromium1 Nonane3
Chlorobenzene3 Lead3 Octane3
Chloroform1 Undecane3
Dichlorobenzene3
Ethyl benzene1
Ethylene dibromide
(EDB)1
Formaldehyde1
Limonene3
Methyl chloroform
(TCA)1
Methylene chloride
Styrene3
Tetrachloroethane
Tetrachloroethylene1
(Perchloroethylene)
Trichloroethylene
(TCE)1
Vinyl chloride3
Vinylidene chloride3
Xylene1
1 The ambient concentrations of these compounds were modelled
as well as monitored.
2 Although we had originally chosen carbon tetrachloride for
both air dispersion modelling and ambient air monitoring
based on 1983 emissions data/ more recent data from 1985
indicated that the compound was little used by industry in
the Baltimore area. Consequently, we dropped it from the
modelling study but included it in the monitoring effort.
3 This compound was chosen primarily because of its interest
to the TEAM monitoring study. Modelling was not possible
due to lack of data on either emissions from point or area
sources or the likelihood that it is produced through
atmospheric transformations of other air contaminants.
IV-8
-------
1-1
VO
TABLE IV-3
SUMMARY OF POLLUTANTS SELECTED FOR MODELLING AND MONITORING
IN THE BALTIMORE AMBIENT AIR TOXICS STUDY
AVAILABLE
POLLUTANT MODELLED (1)
ORGANIC GASES:
Benzene
Carbon tetrachloride
Chlorobenzene
Chloroform
Dichlorobenzene
Ethyl benzene
Ethylene dibromide
Ethylene dichloride
Formaldehyde
Glycol ethers
Methyl chloroform
Methylene chloride
Methyl isobutyl ketone
Perch loroethylene
Phenol
Propylene dichloride
Styrene
Toluene
Trichloroethylene
Vinyl chloride
Vinyl idene chloride
Xylene
ORGANIC P ARTICULATES:
Benzo(a)pyrene
POM
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
TEAM
MONITORING (I
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
MONITORING
) DATA (1)
X
X
X
X
X
X
X
X
X
X
X
X
X
X
CANCER
EFFECT
X
X
X
X
X
X
X
X
X
X
X
X
X
NONCANCER
EFFECT
X
X
X
X
X
X
X
X
X
' X
X
X
X
X
X
X
X
X
X
X
(continued)
-------
TABLE IV-3
SUMMARY OF POLLUTANTS SELECTED FOR MODELLING AND MONITORING
IN THE BALTIMORE AMBIENT AIR TOXICS STUDY
AVAILABLE
TEAM MONITORING CANCER NONCANCER
POLLUTANT MODELLED (1) MONITORING (1) DATA (1) EFFECT EFFECT
METALS:
Arsenic XX XX
Cadmium XX XX
Chromium XX XX
Lead XX X
(1) Please see Tables IV-4 and IV-5 for more information on the health effects associated
with these compounds. Several compounds in the TEAM monitoring were not included in
the Baltimore Ambient Air Toxics Study because of limited health effects data.
These compounds include: tetrachloroethane, limonene, decane, dodecane, nonane,
octane, and undecane.
-------
summed risks across pollutants and/or sources. However, the
uncertainty surrounding these summed risk estimates is
proportionately increased.
ii. Dose-Response Evaluation
Table IV-4 presents data on the carcinogenicity of the
target compounds, including POM. The data are grouped according
to whether or not the target compounds were modelled, monitored,
or both and by general chemical class, i.e., volatile organic
(organic gas) or related compound, metal, or POM. Two types of
information are presented! 1) a qualitative evaluation of the
strength of evidence that a substance is carcinogenic and 2) a
quantitative estimate of the unit cancer risk developed by EPA's
Carcinogen Assessment Group (GAG), except where noted. We must
again remind the reader that POM, unlike the other compounds,
represents a large group of chemicals that differs in composition
according to source category.
With the exception of coke oven emissions, the hazard
assessment for POM was conducted by the Genetic Bioassay Branch,
Genetic Toxicology Division, Health Effects Research Laboratory
within EPA's Office of Research and Development. A comparative
potency approach was used in developing the unit cancer risk
estimates. The approach involves comparing the tumorigenic and
mutagenic potencies of particulates from different source
categories with combustion and pyrolysis products (coke ovens,
roofing tar, and cigarette smoke) known to cause lung cancer in
humans.
b. Noncancer Effects
i. Weight of Evidence
Similar to the situation for pollutants identified as
carcinogens, there can be considerable uncertainty surrounding
whether an agent poses noncancer effects. As a result, EPA
includes uncertainty factors into the calculation of acceptable
exposure levels or thresholds for noncarcinogens based on a
synthesis of all available information. These uncertainty
factors account for differences in animal and human sensitivity,
differences in human sensitivities, and differences in
10Joellen Lewtas, "Combustion Emissions! Characterization
and Comparison of Their Mutagenic and Carcinogenic Activity," in
Hans Stich, ed.. Carcinogens and Mutagena in the Environment,
Volume V (CRC Press, Inc., Baton Rouge, 1985), pp. 59-72. Also,
Roy Albert, J. Lewtas, S. Nesnow, T. Thorslund, and E. Anderson,
"Comparative Potency Method for Cancer Risk Assessment:
Application to Diesel Particulate Emissions," Risk Analysis. Vol.
3, no. 2, 1983.
IV-11
-------
TABLE IV-4
UNIT CANCER RISK FACTORS AND THE WEIGHT OF EVIDENCE
FOR CARCINOGENICITY FOR THE BALTIMORE IEMP TARGET COMPOUNDS
Target Compounds
Unit Cancer
Risk Factor1
Weight of Evidence
Classification2
Organic gases
Benzene
Benzo ( a ) pyrene
Carbon tetrachloride
Chloroform
Chlorobenzene
Dichlorobenzene
Ethyl benzene
Ethylene dibromide
Ethylene dichloride
Ethylene oxide
Formaldehyde
Glycol ethers
Methyl chloroform
Methylene chloride
Perchloroethylene
Phenol
Propylene dichloride
Styrene
Tetrachloroethylene
Toluene
Trichloroethylene
Vinyl chloride
Vinyl idene chloride
Xylene
Metals
Arsenic
Cadmium
Chromium VI
Lead
8.0 x 10'6
3.3 x 10°
3.7 x 10-6
2.3 x 10'5
N/A See Note 3
N/A See Note 4
N/A See Note 3
2.2 x 10'*
2.6 x 10'3
1.0 x 10'4
1.3 x 10'3
N/A See Note 6
N/A See Note 7
4.1 x 10'6
5.8 x 10'3
N/A See Note 3
1.8 x 10'3
N/A See Note 9
4.8 x 10'J
N/A See Note 3
1.3 x 10'6
2.6 x lO'6
5.0 x 10'3
N/A See Note 3
4.3 x 10°
1.8 x 10°
1.2 x 10'2
N/A See Note 9
A
B2
B2
B2
B2
B2
Bl
Bl (see note 7)
B2
C
C
B2
B2
A
C
A
Bl
A
(continued)
IV-12
-------
TABLE IV-4 (continued)
UNIT CANCER RISK FACTORS AND THE WEIGHT OF EVIDENCE
FOR CARCINOGENICITY FOR THE BALTIMORE IEMP TARGET COMPOUNDS
Target Compounds
Unit
Cancer
Risk Factor1
Weight
of Evidence
Classification2
Emission source
Automobiles:
Gasoline-catalyst
Gasoline-non-catalyst
Diesel
Trucks
Heavy duty diesel
Light duty gasoline
(non-catalyst—van)
Residential heating
Oil
Coal
Wood
Industrial & Utility
Power Plants
Oil
Coal
Municipal Waste
Incineration
Coke ovens
See Note 11
5.1 x 1(
1.6 x 10
3.0 x 1(T5
5
'5
See Note 12
2.0 x 10'
3.0 x 10
-*
See Note 12
9.0 x 10-6
1.0 x 10'5
1.0 x 10'5
See Note 12
See Note 12
8.0 x 10-B
See Note 13
6.5 x 10-s
N.A.
N.A.
N.A.
N.A.
3.0 x 10'7
8.0 x 10'8
N.A.
Notes:
N/A - Not applicable; compound is not currently considered to be a
human carcinogen through inhalation. See pollutant-specific
notes for more detail.
N.A. = Not available
(continued)
IV-13
-------
TABLE IV-4 (continued)
NOTES
The unit cancer risk factors are developed by EPA's GAG,
unless otherwise indicated. The unit cancer risk factor
combines estimates of cancer potency with EPA standard
exposure assumptions. The unit cancer risk factor
represents the upper-bound estimate of the increase in
cancer risk associated with continuous lifetime (i.e.,
70 years) exposure to 1 ug/m3 of a specific compound.
EPA classification: A • carcinogenic in humans,
Bl = probably carcinogenic in humans (limited human data),
B2 = probably carcinogenic in humans (insufficient human
evidence but sufficient animal evidence), C = possibly
carcinogenic in humans.
At this time, EPA's Cancer Assessment Group (CAG) has not
found this compound to be carcinogenic from inhalation
exposures and, therefore, not suitable for quantitative risk
assessment.
Within this group of compounds, p-dichlorobenzene is a
CERCLA (Superfund) hazardous substance that has been
identified as a potential carcinogen by CAG. The weight of
evidence for p-dichlorobenzene is B2/C.
The unit cancer risk factor and evaluation of weight of
evidence are from EPA's Office of Toxic Substances.
We examined methyl and butyl cellosolve as the
representative glycol ethers in this study. Insufficient
data exist for an adequate evaluation of the potential
carcinogenicity of these compounds. Therefore, they are not
analyzed as carcinogens in this report.
The carcinogenicity of methyl chloroform is under review by
EPA's CAG. Until a final determination is made by CAG, we
cannot include this compound in our risk assessment.
Styrene is one of the CERCLA (Superfund) hazardous
substances that has been identified as a potential
carcinogen by CAG.
EPA's Office of Toxic Substances is "moderately concerned11
about lead and is evaluating the potential carcinogenicity
of lead. Memo from Jim Cogliano, CAG to Jack Kooyoomjian,
Emergency Response Division, Office of Emergency and
Remedial Response, USEPA, "Potential carcinogens for
designation or updating," June 2, 1987.
IV-14
-------
TABLE IV-4 (continued)
NOTES
10 The kinds and amounts of compounds within this chemical
class vary depending on the combustion source.
Consequently, the unit cancer risk factor varies with source
type. The unit risk values are to be used with estimates of
total particle exposures (i.e., the combination of POM and
elemental carbon).
11 These numbers were generated by the comparative potency
method for estimating unit cancer risk. Generally, there is
more uncertainty surrounding these numbers than for CAG unit
cancer risk factors. (Source: Joelien Lewtas, "Combustion
Emissions: Characterization and Comparison of Their
Mutagenic and Carcinogenic Activity," in Hans Stich, ed.,
Carcinogens and Mutaaens in the Environment. Volume V (CRC
Press, Inc., Baton Rouge, 1985), pp. 59-72. Also, Roy
Albert, J. Lewtas, S. Nesnow, T. Thorslund, and E. Anderson,
"Comparative Potency Method for Cancer Risk Assessment:
Application to Diesel Particulate Emissions, "Risk Analysis.
Vol. 3, No. 2, 1983.
12 There is greater uncertainty surrounding these comparative
potency estimates than for other compounds because of
limited tumorigenicity data.
13 The potency estimate for POM from coke oven emissions is
based on CAG's potency estimate which is derived from human
epidemiological data. Because the CAG unit cancer risk
factor is only for POM, the value was adjusted to reflect
total particle emissions (i.e., POM and elemental carbon).
IV-15
-------
experimental versus actual exposure levels. EPA also includes an
additional modifying factor—ranging from greater than zero to
ten—to reflect qualitative professional judgments regarding
scientific uncertainties not covered by uncertainty factors, such
as the completeness of the overall database and the number of
animals in the study.11 Thus, the reader should keep in mind
that the quality of evidence underlying the threshold values used
in this report vary widely by pollutant.
ii. Dose-Response Evaluation
To assess the noncancer health risks for the selected
pollutants, we compared measured and modelled concentrations with
estimated human threshold values for noncancer health effects.
The threshold values were derived from EPA-developed reference
doses (RfD) or estimates from EPA's Regulatory Integration
Division (RID) and Office of Air Quality Planning and Standards
(OAQPS) toxicologists. In general, the RfD is an estimate (with
the uncertainty spanning potentially an order of magnitude or
greater) of a continuous lifetime human exposure (considering
sensitive subpopulations) that is not likely to present an
appreciable risk of adverse health effects. Doses below the RfD
are not likely to be associated with any health risks. However,
as the frequency and size of exposures exceeding the RfD
increase, the probability that adverse health effects may be
observed in the exposed population increases, thus indicating a
need for further investigation. The RID threshold values were
developed using the same procedures used in deriving RfDs.
However, these threshold values have not undergone EPA review.
In converting to dose from concentration in the ambient air,
standard EPA exposure assumptions were used, which assume that
the typical person weighs 70 kg and breathes 20 cubic meters of
air each day for 70 years.
The threshold values are expressed mostly in terms of six
broad health effects categories! renal (kidney), hepatic
(liver), reproductive (male and female), fetal developmental
(teratological), neurobehavioral, and blood effects. Additional
broad health effects categories—mutagenicity, nonspecific
cellular effects, gastrointestinal, respiratory and
cardiovascular—were examined but were not included in the final
analysis because lower thresholds for the target pollutants were
available for the other health endpoints. Few of the target
pollutants are known or have been shown to produce these other
effects at environmental levels. If a compound was shown to be
generally toxic, but was not necessarily linked with the
"U.S. EPA, Office of Solid Waste Workshop on Risk and
Decision Making, "Principles of Risk Assessments A Nontechnical
Review," p. IV-3.
IV-16
-------
malfunctioning of a specific organ, we grouped its threshold
value under the rubric of "non-specific cellular" effects.
Existence of a threshold does not imply a dose-response
relationship nor is a threshold for a health effect for one
pollutant necessarily comparable in severity to that of another
pollutant. Thus, for example, the health effect for pollutant A
which falls within the category of liver effects may be life-
threatening. On the other hand, the health effect for pollutant
B, which does or does not fall within the same effects category,
may not be.
We present possible noncancer health effects, estimated
thresholds, and sources used in deriving the thresholds in
Table IV-5. The compounds are grouped according to their
chemical structures into four subgroups since some are potential
metabolic precursors for others. The reader should keep in mind
that the quality of evidence underlying these threshold values
varies widely.
3. HUMAN EXPOSURE EVALUATION
To perform an exposure evaluation, it is necessary to have
information on ambient concentrations and the number of
individuals exposed to these levels. We estimated ambient air
levels of the target pollutants using air dispersion modelling
and ambient air monitoring. Ambient monitoring that took place
during the study was conducted by EPA's TEAM with the assistance
of EPA's Region III Central Regional Laboratory. In addition, we
gathered existing monitoring data for the Baltimore area to use
in our exposure determinations.
a. Air Dispersion Modelling
Air dispersion modelling requires (1) emissions estimates
for the selected pollutants by source category and (2) computer
models to estimate ambient concentrations resulting from source
emissions. This section describes the preliminary emissions
inventory that we compiled as part of this study and the steps
taken to refine our emission estimates for key sources. It also
describes the dispersion models used in this analysis and the
results of our model performance evaluation. The model
performance evaluation was designed to ascertain our model's
precision in predicting actual ambient levels by comparing the
modelled results with measured values.
i. Emissions Assessment
The rough preliminary emissions inventory compiled for the
risk assessment phase of the air toxics study—which was updated
from the one used in the pollutant selection process—includes
IV-17
-------
TABLE IV-5
NONCANCER HEALTH EFFECTS AND THRESHOLD VALUES
FOR TARGET COMPOUNDS IN THE BALTIMORE IEMP AIR ANALYSIS
Substance
Health Effect Threshold Value1
Category (ug/nr) Source1
Modelled compounds
Organic gasesi
Benzene and derivatives
Benzene
Ethyl benzene
Phenol
Toluene
Xylene
blood
fetal/dvlopnntl
liver
kidney
liver
kidney
liver
kidney
reproductive
blood
neurobehavioral
respiratory
fetal/dvlpmntl
liver
neurobehavioral
respiratory
cardiovascular
blood
kidney
reproductive
fetal/dvlpmntl
2.451 Snyder et al, 1980
41.3 1986 RID analysis
350.0 IRIS RfD*
350.0 IRIS RfD*
40.0 IRIS RfD*
40.0 IRIS RfD*
1050.0 IRIS RfD*
1050.0 IRIS RfD*
500.0 Matsumoto et al., 1971
1050.0 CITT, 1980
580.0 Hanninen et al, 1976;
Seppalainen et al.,
1978
1010.0 Von Dettingen et al.,
1942; Bruckner &
Peterson, 1981
476.0 Hudak & Ungvary, 1978
215.0 Bowers et al, 1982
215.0 EPA, 1984; Savolainen
et al, 1979
215.0 Hipolito, 1980
215.0 Hipolito, 1980
215.0 Hipolito, 1980;
Browning, 1965
215.0 EPA, 1984
52.8 Ungvary et al., 1980
52.8 Ungvary et al., 1980
IV-18
-------
TABLE IV- 5 (continued)
NONCANCER HEALTH EFFECTS AND THRESHOLD VALUES
FOR TARGET COMPOUNDS IN THE BALTIMORE I BMP AIR ANALYSIS
Substance
Health Effect
Category
Threshold value1
(ug/ms) Source
Halogenated Aliphatics
Carbon
tetrachloride
Chloroform
Ethylene
dichloride
Ethylene
dibromide
Methyl
chloroform
Methylene
chloride
liver
neurobehavioral
kidney
reproductive
fetal/dvlpmntl
fetal/dvlpmntl
liver
neurobehavioral
kidney
reproductive
liver
neurobehavioral
kidney
gastrointestinal
reproductive
(male)
reproductive
(female)
liver
kidney
nonspecific
cellular
liver
fetal/dvlpmntl
kidney
neurobehavioral
Perchloroethylene liver
kidney
fetal/dvlpmntl
2.45 IRIS RfD; EPA, 1980 &
1984; Smyth,1935 & 36
2.45 Holler, 1973
108.0 EPA, 1980
430,0 Adams et al., 1952
24.2 Schwetz et al., 1974
2.4 Murray et al., 1979
Schwetz et al., 1974
35.0 IRIS RfD4
11.7 EPA 1985, Challen et
al., 1858
22.5 Heywood et al., 1979
2.4 Schwetz et al., 1974
26.0 Kozik, 1957
26.0 Kozik, 1957
26.0 Heppel et al., 1946
Hoffman et al., 1971
26.0 EPA, I982a
1.75 Hurtt & Zenick, 1985
NTP, 1982
11.9 Short et al., 1977
12.8 NTP, 1982
12.8 NTP, 1982
315.0 IRIS RfD*
210.0
210.0
699.0
8600.0
69.9
69.9
909.0
IRIS RfD*
Schwetz et al, 1975
NTP 1985 draft
EPA, 1985a
IRIS RfD*
IRIS RfD*
Nelson et al.,
1979
IV-19
-------
TABLE IV-5 (continued)
NONCANCER HEALTH.EFFECTS AND THRESHOLD VALUES
FOR TARGET COMPOUNDS IN THE BALTIMORE IEMP AIR ANALYSIS
Substance
Health Effect
Category
Threshold Value1
(ug/m3)
Source
Propylene
dichloride
liver
kidney
Trichloroethylene kidney
(TCE) liver
neurobehaviora1
Other Qrqanicflt
Benzo(a)pyrene
308.0 Basu, et al. 1984
NTP 1983
308.0 NTP 1983
3770.0 Tucker, 1982
25.9 EPA, 1984
25.9 Grandjean et al, 1955
Ethylene oxide3
Glycol ethers6
Formaldehyde7
Metale;
Arsenic
Cadmium
Chromium VI
fetal/dvlpmntl 93.3
reproductive
blood, kidney
liver
nonspec. cell. 12.37
nonspecific 13.3
cellular effects
fetal/dvlpmntl 21.0
liver 0.4
respiratory 2.0
kidney 0.24
reproductive 119.0
fetal/dvlpmntl 17.0
liver 17.0
reproductive 17.0
nonspecific cell. 17.5
Hanley et a1.,1984
OTS, 1986
Heywood & Sortel,'79
Schroeder &
Kitchener, 1971
Friberg, 1950
Comm. Eur. Com.,1978
Kjellstrom et al,'77
Scharpf et al., 1972
Gale, 1978
NIOSH, 1975
Gale, 1978
IRIS RfD4
IV-20
-------
TABLE IV-5 (continued)
NONCANCER HEALTH EFFECTS AND THRESHOLD VALUES
FOR TARGET COMPOUNDS IN THE BALTIMORE IEMP AIR ANALYSIS
Substance
Health Effect
Category
Threshold Value
(ug/m3)
Source
Compounds onlv monitored
Organic aaaeai
Benzene derivatives
Chlorobenzene8 nonspec. cell. 104.9 Inne et al, 1969
Dichlorobenzene liver
(para & ortho blood
forms) kidney
neurobehavioral
Styrene
cardiovascular
respiratory
liver
blood
Halogenated aliphatics
Vinylidene liver
Chloride
Vinyl chloride liver
Metalat
Lead
nonspecific
cellular
315.0 NTP, 1982
315.0 Varashavskaya,1967
315.0 NTP, 1982
315.0 Hollingsworth et
al. 1956
315.0 NTP, 1982
315.0 NTP, 1982
699.3 IRIS RfD*
699.3 IRIS RfD4
31.5 IRIS RfD*
4.5 EPA Office of
Drinking Water
1.5 Health Effects
Assessment Document
prepared by the
Environmental Criteria
and Assessment Office,
USEPA, Cincinnati, Ohio
(May 1986)
IV-21
-------
TABLE IV-5 (continued)
NOTES
1 Quality of the data for the thresholds varies greatly. Those
derived from RfDs are generally more reliable. Noncancer health
risks were assessed in two ways: (1) for individual pollutants/ the
modelled and measured ambient air concentrations were compared with
the estimated threshold values, and (2) for mixtures of pollutants
with the same systemic effect, a hazard index was created (see
Chapter III). If the index exceeds the value "1", this indicates
the need to further investigate these exposure levels. The hazard
index is created by summing the ratio of the predicted or measured
ambient air concentration to the threshold value for all pollutants
associated with the same health effect category.
2 All individual studies referenced can be obtained by contacting the
Regulatory Integration Division, U.S. EPA, Washington, D.C.
3 This threshold value is being reviewed by EPA's Environmental
Criteria and Assessment Office.
4 IRIS = Integrated Risk Information System. IRIS is an EPA database
that maintains the most recent information of RfDs and CA6 values
for use in risk assessment/risk management. Information on IRIS
can be obtained by contacting EPA's Office of Health and
Environmental Analysis, Office of Research and Development.
3 The Risk Analysis Branch of EPA's Office of Toxic Substances has
identified tentative threshold values for ethylene glycol
monomethyl and monoethyl ether, not for the monobutyl ether form
which is the glycol ether of primary concern in this study on the
basis of industrial use of glycol ethers in the Baltimore study
area. Toxicological data on ethylene glycol butyl ether suggest
that its toxicity is similar to that of ethylene glycol monomethyl
ether, producing primarily blood, liver, and kidney effects. The
threshold values given are for ethylene glycol monomethyl ether and
should serve, for the purpose of screening noncancer risks, as
conservative estimates of threshold values for the glycol ether
group.
6 RID scientists estimated the threshold for health effects based
upon the results of an evaluation of toxicological data on health
risks that was conducted by the Risk Analysis Branch of EPA's
Office of Toxic Substances (Assessment of Health risks to Garment
Workers and Certain Home Residents from Exposure to Formaldehyde,
December 1986). The report has yet to be formally peer or
administratively reviewed.
7 The threshold value should be-considered tentative because it has
been subject to very limited EPA review.
IV-2 2
-------
manufacturing process and boiler emissions from approximately 250
point sources. It also includes estimates of emissions from area
sources and nontraditional sources (such as cooling towers and
sewage treatment plants). The inventory covered more than 200
pollutants. Although this inventory was relatively
comprehensive/ it is important to note that the quality of the
emissions estimates varies considerably by source.
It should also be noted that the Workgroup made a conscious
decision in 1986 to stop making revisions to the emissions
database because of time and resource constraints. New and more
refined emissions information for some of the major sources
included in this study became available after 1986, but were not
included in this report.
(1) Point Sources
We estimated emissions from point (industrial) sources
using data from the 1985 TSR survey.12 This survey generated
data on hazardous materials that were being used, produced or
handled at selected industrial facilities. Based on this
information, AHA estimated plant-wide daily emissions using
registration files on plant operations, EPA emission factors for
certain kinds of manufacturing facilities, and engineering
judgment. The daily emission rates were then converted to annual
rates using plant operating schedules.
There were three major limitations in using the Maryland
1985 TSR survey to estimate emissions. First, the survey did not
consider all sources and pollutants. Second, it was not possible
to differentiate between fugitive emissions and stack emissions
at specific point sources. Third, the survey did not cover
emissions from fuel combustion (i.e., metals, aldehydes, and
polycyclic organic matter—POMs). Finally, only total plant-wide
use information was available to estimate emissions (i.e.,
emissions were- not broken down within a plant by process or
stack). For large facilities with emissions coming from many
locations at the plant, our inability to identify the location of
these emissions could limit the accuracy of results from the
dispersion model.
We supplemented the point source emissions inventory with
emissions estimates for source categories that were not
originally included, such as boilers. For large industrial
facilities, we assigned emissions to processes or areas within
the facility where the emissions were likely to occur. We
12After pollutant selection, source emissions were
reestimated using the more recent TSR survey from the fall of
1985. This more recent survey was not yet available when
emissions estimates were generated for the pollutant selection.
IV-23
-------
assumed that all emissions were released from stacks in our
modelling.
We calculated volatilization of organics during wastewater
treatment at the sewage treatment plants using data from
influent, effluent/ and sludge monitoring. We used mass balances
and estimates of destruction rates during treatment to estimate
emissions of the volatiles. We also calculated emissions
resulting from sewage sludge incineration on the basis of our
estimates on the type of incinerator, air pollution controls,
sludge composition and throughput, and literature estimates of
destruction and removal efficiencies.13
To complement the existing process data we developed an
algorithm and a set of emission factors to estimate boiler
emissions for the point sources in the toxics inventory.14 These
unique emission factors were based on fuel type, boiler type and
size, and assumed in-place controls. We then multiplied the
factors by fuel consumption reported in the AMA registration
files to estimate metals, formaldehyde, and*1 total suspended
particulate (TSP) emissions from point source boilers. We
included TSPs so that we could later estimate risk from exposure
to POMs, as described below.
We estimated emissions of POMs using a technique developed
in conjunction with EPA's Office of Air Quality Programs and
Standards (OAQPS). We calculated POM emissions, defined as the
benzene-extractable portion of combustion particulate matter, by
first estimating TSP emissions from point and area combustion
sources. EPA's Genetic Bioassay Laboratory provided estimates of
the percentage of extractable organics by source category. The
resulting POM emission rates were later modelled and multiplied
by source-specific potency scores also provided by the Genetic
Bioassay Laboratory. For point sources, emissions of POMs were
estimated for industrial coal and oil combustion, coke ovens, and
municipal waste incinerators.
Because hexavalent chromium is highly potent, we made an
attempt to estimate hexavalent chromium, as opposed to total
chromium, emissions as accurately as the data would allow. This
The emission calculations are documented in: U.S. EPA,
Office of Policy Analysis, Regulatory Integration Division,
Baltimore Phase II Air Toxics Emissions Assessment, Versar Inc.
for EPA, EPA Contract No. 68-01-7053, November 6, 1987.
"ibid.
IV-24
-------
determination was made using best engineering judgment and EPA
emission factors.15
(2) Area Sources
We designated all sources not included in the category of
major point sources as area sources. Examples of area sources
that were included in the analysis are: motor vehicles; gasoline
service stations; solvent usage by commercial and small
industrial facilities (dry cleaning establishments; businesses
that do degreasing, surface coating, printing and publishing; and
manufacturers and users of rubber and plastics); industrial,
commercial, institutional, and residential heating (including
wood stoves and fireplaces); cooling towers; the drinking water
distribution system from which trihalomethanes volatilize; waste
oil combustion; agricultural burning, and minor point sources not
modelled individually. We calculated emissions from these
sources, except the minor point sources, using an assortment of
emission factors and algorithms developed by RID and EPA's Office
of Air and Radiation.
Emissions from cars and trucks were estimated using a
computer model called MOBIL3 that calculates total hydrocarbon
emissions on the basis of vehicle miles traveled in rural and
urban sections of the study area. These estimates of total
hydrocarbon emissions were subsequently disaggregated into
estimates for specific pollutant emissions on the basis of
emission factors developed by OAQPS and published in the
scientific literature.
Area sources of POMs included gasoline- and diesel-powered
road vehicles; small industrial boilers not included in the point
source inventory; residential, institutional, and commercial
heating; and wood stoves and fireplaces. We used techniques
similar to those described earlier for point sources to estimate
POM emissions as source category-specific percentages of TSP
emissions.17 All of the emissions data were entered into the
PIPQUIC database.
l5The emission calculations are documented in: U.S. EPA,
Office of Policy Analysis, Regulatory Integration Division,
Baltimore Phase II Air Toxics Emissions Assessment. Versar Inc.
for EPA, EPA Contract No. 68-01-7053, November 6, 1987.
"ibid.
"The emission calculations are documented in: U.S. EPA,
Office of Policy Analysis, Regulatory Integration Division,
Baltimore Phase II Air Toxics Emissions Assessment. Versar Inc.
for EPA, EPA Contract No. 68-01-7053, November 6, 1987.
IV-25
-------
Area sources are not uniformly distributed within a
geographic area, Thus, we distributed the area source emissions,
except road vehicles, on the basis of USGS land use data. The
entire study area was subdivided into quadrangles of irregular
shape and size composing the area source grid shown in
Figure IV-1. Land use within each quadrangle was then broken
down into various use categories, including residential,
commercial and services, industrial, and transportation
categories. Emissions associated with different area source
activities were then apportioned into the grids depending on use.
For example, dry-cleaning emissions were distributed
proportionately to commercial and service land use categories.
For motor vehicles, emissions were distributed over the area
source grid cells on the basis of the computer model MOBIL3,
previously described.
(3) Quality Control
Quality assurance/quality control was an integral part of
our data gathering efforts. The number of emission estimates and
factors, algorithms, combined with multiple applications and
manipulations of data necessitated that we continuously review
the data and algorithms to assure that all data entered into
PIPQUIC were as accurate as possible and that PIPQUIC tools and
algorithms were operating as designed. Thus, the massive size
and interrelational nature of the PIPQUIC database meant that
quality control became a routine part of all data gathering and
analysis.
Some of the specific measures taken were the following:
• EPA's technical contractors reviewed AMA's plant-wide
point source emissions estimates, which were based on the
1985 TSR survey, to identify any apparent errors.
• Technical staff at EPA's RID reviewed all emission
factors and algorithms developed by the technical
contractor for area sources, nontraditional sources, and
point source boilers, to assure that values were correct
and applied properly. Many of these factors underwent
further review by other EPA Offices, such as the Office
of Air Quality Planning and Standards.
• As soon as data were entered into PIPQUIC, EPA's
technical contractors checked printouts again to assure
that all algorithms and factors were correct.
• EPA's technical contractors periodically generated new
printouts to compare with previous printouts to guarantee
that any internal modifications to PIPQUIC tools did not
incorrectly change PIPQUIC output values.
IV-26
-------
FIGURE IV-1
Area Source Grid System for Baltimore Study
4401
331 331 341 344 391 3» 311 3M 371 371 3*1 3M 391
Note:
County Line
Area Source Grid
UTM - Universal Transverse Mercator
IV-2 7
-------
ii. Emissions Characterization for Kev Sources
Because of the large number of point sources that make up
the emissions inventory, we ranked facilities to identify which
sources to emphasize in our analyses. We estimated the portion
of the plant-wide emissions for these facilities to be assigned
to a particular process or stack. This step was necessary to
improve the accuracy of the modelling because (as mentioned in
the previous section) the rough emissions inventory provided only
a single figure for the emissions of a pollutant from a plant.
We ranked facilities using three approaches: (1) total
emissions of study pollutants, (2) emissions of study pollutants
weighted by unit cancer risk factors, and (3) emissions of study
pollutants weighted by both exposed population and the unit
cancer risk factor. In the first ranking based on emissions, one
point source accounts for nearly 50 percent of the total point
source emissions of study pollutants in the study area; the top
10 facilities contribute 80 percent of the point source
emissions.
While the product of emissions and unit cancer risk factors
does not reflect actual risk, it does consider two important
factors that contribute to human health risk: the volume of
emissions and the toxicity of the pollutant(s). The same point
source that accounted for most of the emissions also dominated
the second ranking in terms of toxicity-weighted emissions.
Several smaller sources of emissions ranked high in the second
approach because of the estimated high toxicity-weighted
emissions of hexavalent chromium.
Multiplying the risk-weighted value by the population in the
vicinity of each facility gave an indication of which sources are
likely to be exposing the largest populations to the greatest
volumes of the most potent carcinogens. The relative rankings of
point sources by this measure differed very little from the
rankings based on toxicity-weighted emissions. This indicates
that variation in potencies of the pollutant is more significant
than any variation in population among receptor grid cells.
On the basis of these three rankings, we selected 15
facilities as major point sources.
For complex integrated major point sources, we broke down
emissions by process, plant area, or stack. We distinguished
between process and boiler emissions and assigned representative
or unique stack parameters for each release, basing our selection
on each facility's largest VOC or TSP stack as listed in the
AMA's registration files. We visited some major point sources to
update and verify emissions, gather modelling input information,
such as stack parameters and building dimensions inside and
IV-28
-------
adjacent to the facility, and gather data on feasible control
options (see Chapter VI).
We classified approximately 40 additional facilities, for
modelling purposes, as "intermediate point sources." For these
plants, emissions from boilers and process emissions were
modelled separately. We chose a single process stack and one
boiler stack based on registry data to be representative of
process and boiler emissions, respectively. We made no attempt
to further break down emissions for these operations. As a
general rule, we chose stacks emitting the largest amount of VOC
as being the representative process stack, while the boiler stack
emitting the greatest quantity of particulates was selected as
the representative boiler stack (since the pollutant of concern
would be POM).
Because of the small volume of emissions from the remaining
200 very small point sources, we modelled these "minor point
sources" as area sources, with area source release specifications
(height, temperature, and exit velocity). Each of these minor
point sources is still listed in PIPQUIC as a point source.
The decision rules used in our modelling of large,
intermediate, and area sources may have resulted in
underestimates of maximum increased lifetime individual cancer
risk and the need to further investigate noncancer effects. By
assuming that emissions from all large and intermediate sources
were released from the stack with the greatest VOC or particulate
emissions, we often modelled the tallest stack. Emissions from a
taller stack will disperse better, thus resulting in lower
maximum concentrations. Similarly, modelling all small volume
emitters as area sources—although practical—eliminated our
ability to consider peak concentrations.
iii. Dispersion Modelling
(1) Choosing the Model
In choosing an air guality model for evaluating urban
exposures in Baltimore, we selected a model that is (1) cost-
effective in estimating annual exposures, and (2) capable of
addressing urban area sources as well as point sources. While a
number of readily available models fulfill these desired
characteristics, we chose two off-the-shelf EPA Gaussian
dispersion models: the Climatological Dispersion Model (COM) and
the Industrial Source Complex Long-Term (ISCLT) model.18
18U.S. EPA, Guideline on Air Quality Models {Revised!. U.S.
Environmental Protection Agency, Office of Planning and
Standards, Research Triangle Park, N.C., EPA-450/2-78-027R, 1986.
IV-29
-------
The CDM is an EPA preferred model that is appropriate for
estimating the annual average concentrations for both point and
area sources located in flat, urban regions.19 We used CDM to
estimate the larger/ urban area source impact. The integration
technique used in the CDH model for area sources theoretically
produces more accurate results than are obtained through the more
simplified treatments found in most other models.
The ISCLT model is also an EPA recommended model for
evaluating point source releases (including small industrial area
sources) located in regions of flat or gently rolling terrain.20
ISCLT is generally the model of choice when evaluating major
industrial sources, especially facilities for which detailed data
are available on release specifications such as building downwash
and vent releases. We chose this model to estimate the annual
concentrations in the neighborhoods surrounding each major
facility.
(2) Running the Model
We modelled 55 release points at 52 major and intermediate-
size facilities (one of the large facilities had three release
points modelled). The remaining minor facilities were grouped
into 23 area sources and modelled. We used detailed release
specifications (when available) to assess the microscale
concentrations (that is, concentrations at discrete receptor
sites). With the exception of a few large facilities, most
facilities were modelled using one representative stack.
We used hourly meteorological data that had been collected
at a downtown monitoring site during Phase I of the Baltimore
IEMP study in 1983/1984. These data consist of wind direction,
wind speed, and horizontal and vertical turbulence information.
We derived the temperature and stability data by using the
National Weather Service stations at Baltimore-Washington
International (BWI) Airport, and obtained mixing height data from
Washington (Dulles) Airport. When data were missing from the
downtown station, we used surface data from BWI Airport. We then
reformatted the hourly meteorological data into day/night joint
frequency distributions (frequency of wind directions, wind
speeds, and stabilities), as required by both our air dispersion
"irwin, J.S., T. Chico, and J. Catalano, CDM 2.0
Climatoloqical Dispersion Model—User's Guide. U.S. Environmental
Protection Agency.
ZOU.S. EPA, Industrial Source Complex (ISC^ Dispersion Model
User's Guide. Second Edition, Volume 1. U.S. Environmental
Protection Agency, Research Triangle Park, N.C., Publication No.
EPA-450/4-86-005a, 1986.
IV-30
-------
models, Industrial Source Complex Long-term (ISCLT) Model and
Climatological Dispersion Model (COM).
iv. Model Performance Evaluation
We conducted a limited evaluation of the performance of the
dispersion model using monitoring data that had been collected
for three months at ten stations in 1983-84 during Phase I of the
Baltimore IEMP.21 The evaluation allowed us to assess the
adequacy of the modelling conducted for the exposure assessment
for the Baltimore IEMP project.
We ran COM (for urban area sources) and ISCLT (for point
sources) to estimate the pollutant concentrations at locations
corresponding to the monitoring sites and monitoring periods. We
then compared the modelled and measured concentrations.
Table IV-6 displays the modelled and measured ambient air
concentrations for the ten compounds and ten monitoring sites.
For all pollutants considered, except trichloroethylene, the
models consistently underpredicted ambient air concentrations.
We statistically compared measured and modelled data for
bias and correlation. Good correlation indicates that our
modelling reliably predicts ambient levels. Bias tells us the
extent to which the model under- or overpredicts actual levels.
Table IV-7 displays the bias and correlation coefficients
for the Baltimore IEMP study (with and without the contribution
of point source emissions) and the Philadelphia IEMP study. When
comparing model performance in the Baltimore and Philadelphia
IEMP studies, the bias for the Baltimore model performance is
substantially greater for most compounds and the correlation
substantially lower. The higher bias could be attributed to the
following factors: (1) a potential underestimation of the
Baltimore point or area source emissions, (2) failure to account
for background concentrations imported from outside the study
area or by chemical transformations, or (3) that the models are
systematically underpredicting ambient levels due to the
treatment of atmospheric and meteorological conditions.
"The monitoring data collected in 1987 as part of the TEAM
study, previously described, were not available in time for this
evaluation.
2ZA more detailed model performance check would also include
partitioning of the results as a function of meteorological
conditions, and performing a follow-up verification of point
source emissions data. However, we did not conduct a detailed
performance check because the 1983-84 data are outdated.
IV-31
-------
M
Ul
Isj
Table IV-6
Comparison of Measured vs. Predicted Concentration* for Baltimore Monitoring Sites
During the Period 11/20/83 - 2/16/84 for Model Hun llrbun Mode I
Compound
Measured Cone. (mg/m3)
Benzene
Cartxm Telrachtoiide
ChkMDfbim
Eltiyl Beniene
Xytone
Toluene
1.2CMct)kxoelhane
1 .2 OfcNw opropane
Tilchloroelhylene
Peichtoroelhylene
Predicted Cone. (mo,/m3)
Benzene
Carbon TeUachtorUe
Chloroform
Ethyl Benzene
Xytene
Toluene
1.2DJcWoroelhane
1.2Olchloropropene
Trkhloraethylene
Peicrttoroeinytene
Ratio ol Pred./Meaa.
Benzene
Caibon Tetrechlortde
Chloroform
Eltiyl Benzene
Xytone
Toluene
l.20fchtoroelhane
1 .2 OlcMofopr opane
Trfchtoroethylene
Perchloroetliytene
t
1200
1 10
210
930
3040
1640
020
030
090
4 BO
1
421
000
005
062
283
479
001
000
144
434
1
035
000
002
007
009
029
005
000
160
090
2
950
090
040
640
21 10
900
020
020
050
540
2
223
000
007
080
205
312
001
000
160
578
2
023
000
017
012
010
035
006
000
320
107
3
1020
060
020
800
2080
990
030
020
050
700
3
261
000
004
049
181
320
001
000
140
3 10
3
026
000
018
006
009
032
002
000
279
044
4
1290
130
070
740
2010
1600
050
070
390
600
4
397
000
003
038
214
404
001
000
092
244
4
031
000
004
005
Oil
025
001
000
024
041
5
1260
090
too
470
1370
930
040
030
1 40
930
5
928
000
003
036
1 98
267
001
000
104
240
5
074
000
003
008
014
029
002
000
074
026
6
550
090
320
290
570
750
260
010
040
1 SO
6
992
000
002
025
1 45
206
001
000
064
163
e
1 80
000
001
009
025
027
000
000
161
109
7
1210
120
260
620
1720
7.60
070
200
1 10
320
7
930
002
006
038
917
738
001
000
076
186
7
077
001
002
006
053
097
001
000
069
058
a
1060
1 40
470
570
1640
480
020
040
100
390
8
669
000
003
041
303
438
001
000
090
225
8
063
000
001
00?
018
091
004
000
090
058
9
780
070
060
520
1360
550
100
040
020
290
9
660
000
002
023
185
399
001
000
061
ISO
9
085
000
003
004
014
072
001
000
306
052
10
1030
060
1 10
630
1920
740
020
020
030
240
10
S96
000
002
020
1 19
268
000
000
056
1 35
10
058
000
002
003
006
036
002
000
185
056
Note: BalUmoto monitoring data used In tins analysis wen collected m 1983/84
Sites
1 Gulltord
2NEPO
3SWPD
4 HoteWrd
9 Uumtaft
6 Chesapeake Terrace*
7 Sun A Chesapeake
6 Fl Henry
9 Coasl Guutti
10 Rivet a Beach
-------
Table IV- 7
Comparison of Measured vs. Modelled Data
for the Baltimore and Philadelphia IEMP Studies [1]
Bias [2]
Baltimore IEMP Baltimore IEMP
Compound
Benzene
Carbon Tetrachlonde
Chloroform
Ethyl Benzene
Xylene
Toluene
1,2 Dichloroethane
1,2 Oichloropropane
Trichloroethylene
Perchloroethylene
Philadelphia IEMP
Paint and Area
Compound
Benzene
Carbon Tetrachlonde
Chloroform
Ethyl Benzene
Xylene
Toluene
1,2 Dichloroethane
1,2 Dichloropropane
Trichloroethylene
Perchloroethylene
m mm wuiv
9.74
0.96
1.63
5.82
16.93
7.89
0.62
0.48
0.04
2.03
r wi ••% «ip»« ^^ • •• ••
4.27
0.96
1.62
5.80
15.07
5.51
0.62
0.48
0.03
1.97
3.70
2.00
2.70
6.00
16.60
8.60
0.20
0.40
1.30
1.70
Correlation
Baltimore IEMP Baltimore IEMP
Philadelphia IEMP
Paint and Area
Mrvai wniv
0.36
0.00
-0.27
0.56
0.49
0.19
-0.45
0.00
0.01
0.39
-0.20
0.36
-0.04
0.55
0.08
0.08
-0.45
0.00
0.01
0.39
0.62
0.28
•0.49
0.38
0.51
0.53
0.92
0.99
0.02
0.80
[1] For the compoeae period 11/83 • 3/84
[2] Average measured concentration - Average modelled concentrauon
Note: Baton
tormg data ueed in ttiia analysis were collected in 1983/84.
IV-3 3
-------
Of the ten compounds considered, we found that ethyl
benzene, perchloroethylene, xylene, and benzene had the best
correlation in the model runs. Of these four, ethyl benzene
correlated highest in the Baltimore area with a value of 0.56.
The remaining six compounds, chloroform, carbon tetrachloride,
toluene, 1,2-dichloroethane, 1,2-dichloropropane, and
trichloroethylene (TCE) had negligible lower correlations.
This analysis indicates that our air dispersion modelling
underestimates actual ambient concentrations of the selected
pollutants, with the exception of TCE. Predicted ambient air
concentrations of benzene and benzene-ring compounds, like
xylene, toluene, and ethyl benzene, are likely to be
underestimated by at least a factor of two. TCE may have been
overestimated by a factor of two. The values for the other
compounds are at best seen as lower estimates of ambient
concentrations. Consequently, our exposure estimates based on
modelling are understated; however, our estimates of risk using
these modelled values are still more likely to overstate than
understate true risk.
b. Ambient Air Monitoring
EPA's Office of Research and Development—in coordination
with the AHA and the Baltimore County Department of Health—
conducted monitoring of the ambient air at two fixed sites
(Dundalk and Parkville) in the Baltimore study area as part of
its TEAM study (see Figure IV-2). The sampling focused on 27
compounds and their isomeric forms, of which 18 were VOCs
(volatile organic compounds, i.e., organic gases), 5 were
straight-chain aliphatic hydrocarbons, and 4 were metals. The
modelling analysis addressed 13 of these 27 pollutants.
In addition, we compiled readily available monitoring data
on organics and metals for the Baltimore area. These data came
from various AMA studies and the 1983/1984 IEMP monitoring study.
There is considerable uncertainty surrounding these data and not
all of the results have undergone extensive peer review.
Nonetheless, we felt that it was important to include this
information to demonstrate a different approach for estimating
exposure.
i. TEAM VOC Sampling
The sampling of all VOCs and straight-chain aliphatic
hydrocarbons occurred concurrently at the two fixed sites over a
six-week period between March 23 and May 1, 1987. Table IV-8
presents a summary of the maximum and average measured pollutant
concentrations of the target VOCs and hydrocarbons included in
the Baltimore ambient air toxics study.
17-34
-------
FIGURE IV-2
Baltimore Integrated Environmental Management Project
Map of Ambient Air Monitoring in 1987
t = Monitoring Site
Parkvilte Fixed
Monitoring Site
Dundalk Fixed
Monitoring Site
IV-3 5
-------
There are several limitations to the data that should be
noted in properly interpreting the monitoring results. In
Table IV-8, we reported the level of confidence in each pollutant
exposure level. In particular, a'confidence level of 3 means
that the exposure estimate is probably invalid. A low confidence
level was assigned in situations where:
(1) The quality control techniques could not confirm a
pollutant's presence (vinyl chloride)
(2) There were analytical problems in quantifying ambient
concentrations of methylene chloride at Parkville, and
vinylidene chloride at both Dundalk and Parlcville, as
well as specific isomers of a compound (e.g., xylene)
(3) Pollutant emissions from sources near the monitoring
site were suspected of interfering with the ambient
sampling (dichlorobenzene)
(4) Measured concentrations may have been overstated
because of coelution with another substance (styrene
with xylene)
The concentrations of pollutants at the Dundalk site,
downwind of major industrial sources, were generally lower than
anticipated. Vinylidene chloride was the clear exception;
however, there are no known sources of vinylidene chloride in the
area and, most importantly, the measured concentrations are
highly uncertain. Additional exploration is needed to verify
whether vinylidene chloride are, in fact, present in the ambient
air.
The dichlorobenzenes were also unexpectedly detected in the
ambient air. The known sources of emissions of the
dichlorobenzenes in Baltimore are only consumer products.
Statistical analysis of the relationship of their concentrations
with wind direction suggested that at least some portion of the
emissions may have come from vents from the building upon which
the monitoring station was located. We were not able to identify
products used in the building which contained the compounds.
The Parkville monitoring results for pollutants that have
been sampled in the past were generally at anticipated levels,
based on the results of previous lEMPs. The surprises were for
methylene chloride, vinylidene chloride, and the
dichlorobenzenes. However, the measured levels of methylene
chloride and vinylidene chloride are suspect because of the
analytical problems in quantifying their concentrations.
Methylene chloride, like the dichlorobenzenes, is heavily used in
consumer products.
IV-36
-------
Table IV-8
BALTIMORE IEMP AIR TOXICS STUDY
AVERAGE MEASURED AMBIENT CONCENTRATIONS .
AT THE TWO FIXED MONITORING SITES: TARGET VOLATILE ORGANICS1
(ug/m3)
Dundalk
Compound
Average ,
Concentration
Maximum
Measured Value
Confidence
in
Exposure.
Estimate
Benzene
Carbon tetrachloride
Chlorobenzene
Chloroform .
m-Dichlorobenzene
o-Dichlorobenzene
p-Dichlorobenzene
Ethylenendibromide {EDB)
Formaldehyde .
Methylene Chloride5
Percnloroethylene
Styrene
Trichloroethane (TCA)
Trichloroethylene (TCE)
Vinyl chlorioV ,
Vinylidene chloride7
m-Xylene
3S:
3.8
21.9
75.0
22!
s:
.0
Parkville
Compound
Average ,
Concentration
Confidence
in
Maximum Exposure.
Measured Value Estimate
Benzene
Carbon
Chlorol
Chloro:
m-Dich
o-Dich
p-DichJ
tetrachloride
Benzene
:orm
.orobenzene
.orobenzene
[orobenzene
Ethyl benzene
Ethylene dibromide (EDB)
Formaldehyde ,
Methylene Chloride
Percnlorpethylene
Styrene
Trichloroethane (TCA)
Trichloroethylene (TCE)
Vinyl chloride8^ , ;
Vinylidene chloride
m-Xylener
5.2
1.1
1.5
N.D.
7.2
4.8
5.7
5.9
0.1
25610
ii:l
5.9
0.01
N.D.
68.4
15.4
19
16
48
39
27
55
J
27
109
0
ili
.2
.0
.6
. 3
. 8
.3
'.7
:jj
!4
:8
2
1
2
3
3
3
1
3
2
1
1
N.D.: Not detected
(continued)
IV-3 7
-------
Table IV-8 (continued)
NOTES
Monitoring occurred between March 26, 1987 and May 1, 1987 at
Parkville Middle School in the Parkyille section.of Baltimqrr
County and Merritt Point Mental Health Facility in the Dundalk
section of Baltimore County. The air monitoring was conducted as
part of the 1987 Baltimore Total Exposure Assessment Methodology
fTEAM) Study.
Monitoring samples for which no concentration of the target L
compounds was detected were given the value zero. On the other
Sana, samples for which trace amounts of the target compounds were
etected but the concentrations were below quantification were
assigned half the value for the level of detection for that sample
and target compound.
There are three levels of confidence in exposure estimates. Level
1 means that the exposure estimate is probably good. Level 2 means
that the exposure estimate may be good. Finally, level 3 means
that the exposure estimate is probably not good. The evaluation is
based on EPA audit reports and statistical analysis by RID staff.
It was not possible to quantify the concentration of m-
dichlorobenzene at Dundalk.
There is some evidence that methylene chloride was being emitted
from sources within or on top of the building where the monitoring
station was located. Consequently, the measured values may not be
representative of concentrations to which people may be exposed in
the ambient, as opposed to indoor, environment.
Quality control procedures could not confirm the presence of vinyl
chloride in monitoring samples.
There were analytical problems in quantifying ambient air
at both Parkville ar__
are highly uncertain and
niBiw wwtw anaj.ytiuaj. UCUUJ.«HIIS J.IL uueijii.ii,
concentrations of vinylidene chloride at both Parkville and
Dundalk. Thus, these measured values arc
deserve further investigation.
8 The laboratory was only able to quantify the concentration of the
meta isomer of xylene.
9 The laboratory analytical level of detection (0.37 micrograms per
cubic meter) was not adequate to measure chloroform in tne ambient
air of Parkville.
10 The value given for styrene is probably high because of coelution
with the xylene isomer o-xylene.
IV-38
-------
Because the 1983 IEHP monitoring effort (discussed in the
Phase I report) did occur under similar meteorological
conditions, we compared the averages for compounds present in the
two sets of samples. The results shown in Table IV-9 are
generally similar, differing by less than a factor of two. The
earlier data appear to be the higher of the two. The exceptions
are chloroform/ trichloroethylene, and, to a lesser extent,
perchloroethylene.
ii. TEAM Metala Monitoring
EPA's ORD collected 26 24-hour samples of both total
suspended particulates (TSP) and small respirable particulates of
10 microns and smaller (PM10) over a six week period from March
26, 1987 to May 1, 1987 using high volume evacuated canisters.
The sampling occurred concurrently with the VOC monitoring at the
two fixed monitoring sites in Parkville and Dundalk. The samples
were analyzed for only four metals: arsenic, cadmium, chromium,
and lead.
Table IV-10 summarizes the monitoring results for the
metals. The concentrations represent simple arithmetic averages.
The highest ambient air concentrations are reported for lead.
For all the metals, the maximum value for any sample was
generally twice the average measured value. The averages were
roughly the same for each site.
The limitations of the results are again emphasized. At
both monitoring sites, cadmium and chromium were detected in only
a small number of the samples. Furthermore, the analysis could
not discern between trivalent and hexavalent chromium.
Consequently, the values are for total chromium. At Dundalk,
arsenic was not found in any of the samples. The reported
concentration for arsenic at this site represents half of the
analytical detection limit.
ORD's treatment of samples with no detectable levels of
metals differed from their treatment of samples with no
detectable levels of VOCs. There is no standard method for
calculating average levels of pollutants where pollutants are
present at levels below the levels of detection. EPA felt that
if a particular VOC was regularly not detected, there was a
reasonable likelihood that it was indeed not present in the
ambient air or was present in negligible quantities. On the
other hand, it was known from data on sources and emissions that
all of the metals were likely to be present in the ambient air on
the sampling days, but the sensitivity of the analytical
procedures was simply not adequate to detect and quantify them.
ORO also collected samples for metals over a much smaller period
of time than for VOCs (roughly 40 days versus about 14 days for
metals) and hence the confidence in the estimated ambient levels
were likely to be lower. Finally, the targeted metals are
IV-3 9
-------
Table IV-9
COMPARISON OF 1987 AND 1983/84 MONITORING RESULTS1
(ug/m3)
1983-1984 1987
Compound Monitoring2 Monitoring
Benzene 10.4 4.6
Carbon tetrachloride 1.0 0.7
Chloroform 1.7 O.I3
Ethyl benzene 6.2 3.7
Xylene 17.8 10.54
Trichloroethylene 1.0 0.2
Perchloroethylene 4.6 1.4
Values represent averages for values from all monitoring sites.
Although more compounds were monitored, this table presents
only those compounds that overlapped in the two monitoring .-.
programs.
2 Source: Versar, Inc., Draft Data Report for Monitoring and
Analytical Activities to Determine Ambient Air Concentrations
of Selected Toxic Pollutants in Baltimore. Maryland. April 18.
1984.
3 Because chloroform was detected at only one site, only this
average value for this one site is given.
* Value is for ra-xylene only.
IV-40
-------
Table IV-10
AVERAGE MEASURED AMBIENT CONCENTRATIONS: METALS AT FIXED
MONITORING SITES
.3,1
(nanograms per cubic meter of air—ng/m )
Avg. Across
Compound1
Arsenic
maximum
Cadmium6
maximum
Chromium7
maximum
Lead
maximum
Dundalk
TSP
1.83
-
1.6
7.6
3.1
5.8
43.0
88.0
PM10
1.83
-
1.1
3.6
2.0
4.2
36.7
73.0
Parkville Sites
TSP
1.8*
4.1
0.9
3.2
1.7
3.6
37.0
87.0
PM10
1.93
3.6
0.9
3.0
2.0
—
33.3
57.0
TSP PM10
1.8 1.8
1.2 1.0
2.4 2.0
40.0 35.0
1 One nanogram is equivalent to one one-thousandth of a micro-
gram, which, in turn, is equal to one one-millionth of a gram.
2 The value for one-half the detection limit was given for those
samples for which the target compound was not detected. The
figures in the table are simple arithmetic averages. Eighteen
samples were collected in Dundalk and seventeen in Parkville.
3 in the samples collected in Dundalk no arsenic was detected
within the laboratory's analytical detection limit. Conse-
quently, the value for one half the detection limit for the
samples was used in estimating exposure levels.
* Arsenic was detected in only one TSP sample in Parkville.
3 Arsenic was detected in only one PM10 sample in Parkville.
6 Cadmium was detected in eight TSP and five PM10 samples in
Dundalk and two TSP and 2 PM10 samples in Parkville.
7 Chromium was detected in eight TSP and two PM10 samples in
Dundalk and in only two TSP samples in Parkville.
IV-41
-------
generally much more potent carcinogens than the VOCs such that
even relatively low levels can pose significant risks. Thus,
assigning the value of half the detection limits to samples with
no detected metal would likely provide more reasonable estimates
of exposure. Clearly, if actual ambient concentrations are even
less than half the detection limit, risks may be overstated.
iii. Available Area-Wide Sampling Data
To supplement the TEAM monitoring data, we compiled readily
available air toxics monitoring data for several sites throughout
the Baltimore metropolitan area. This summary consists of
monitoring data generated by several sampling programs conducted
by the AMA and the 1983/1984 IEMP VOC monitoring. Table IV-11
presents the lowest and highest measured pollutant value at a
given monitoring site and the average measured pollutant
concentrations across all monitoring locations (the supporting
data can be found in Appendix B). We emphasize that these data
are highly uncertain; however, we feel that it is important to
include them in this analysis to demonstrate the importance of
evaluating area-wide exposures for screening purposes using both
modelled and measured data.
c. Exposed Population
We subdivided the study area into 5 km squares (grid cells)
composing what we call the standard arid and centered a refined
grid consisting of 2.5 km grid cells over the most densely
populated portion of the study area. Figures IV-3 and
IV-4 show the siting of the standard and refined grids,
respectively. The total area covered by the standard grid is
1600 km (40 km by 40 km) with 64 nodes in all and a total area of
400 km (20 km by 20 km) covered by the refined grid, again with
64 nodes. The refined grid is offset from the standard grid by
1.25 km because we placed the refined grid over the most densely
populated portion of the study area.
We used the 1980 Census Bureau Block Grid/Enumeration .
District data to determine the population within the squares of
each grid. The total population in the study area is about 1.6
million. We used these population estimates to estimate the
annual excess cancer incidence, above background, that would
occur as a consequence of exposure to the modelled concentrations
within each square. These population estimates were also used to
determine the number of people at risk of noncancer effects as a
result of average pollutant levels within the squares.
Each grid cell is identified by the Universal Transverse
Mercator (UTM) coordinates for its southwest node. We henceforth
refer to a specific cell or location within the study area by its
UTM coordinates.
IV-4 2
-------
TABLE IV-11
AVAILABLE AREA-HIDE SUMMARY MONITORING DATA FOR BALITMORE (1)
M
POLLUTANT
LOWEST AND HIGHEST
AVGERAGE MEASURED
CONCENTRATION
(ug/m3)
AVERAGE MEASURED
CONCENTRATION ACROSS
ALL MONITORING SITES
(ug/m3)
NUMBER OF
MONITORING SITES
Arsenic
Benzene
Benzo ( a ) pyrene
Cadmium
Carbon tetrachloride
Chloroform
Chromium
Ethylene dichloride
Propylene dichloride
Ethyl benzene
Lead
Methylene chloride
Methyl isobutyl ketone
Perchloroethylene
Toluene
Methyl chloroform
T r i ch 1 or oet hy 1 ene
VinyL chloride
Xylene
<0.01
3.90
<0.01
<0.01
0.20
0.10
-------
FIGURE IV-3
Baltimore Study Area
Map of Boundaries of the Modelling Domain
for the 5 km (Standard) Grid System
IV-4 4
-------
FIGURE IV-4
Baltimore Study Area
Map of Boundaries of the Modelling Domain
for the 2.5 km (Refined) Grid System
IV-4 5
-------
We also identified 118 locations, called discrete receptors.
for use in estimating the highest increased lifetime individual
risks. The 118 discrete receptor locations were chosen because
of their location relative to ten major emission sources. They
represent discrete geographic locations or points. The locations
for these discrete receptors are also identified by their UTM
coordinates. We show the location of these discrete receptors in
Figure IV-5.
IV-4 6
-------
FIGURE IV-5
Baltimore Study Area
Map of Boundaries of the Modeling Domain
Leaend:
+ Location of discrete modelling receptor.
iv-4 7
-------
V. RISK ASSESSMENT SCREENING RESULTS
This chapter presents the results of the risk assessment
screening analyses used to address the key questions raised in
Chapter II:
• Can we develop a screening methodology to identify and
characterize health risks posed by selected air toxics
from exposures occurring in the "urban soup"? l|2
• What is the relative importance of point sources versus
area sources to urban air toxics in the Baltimore area?
• What is the relative contribution of specific point and
area source categories to air toxics risk?
In the Baltimore air toxics study, a relative ranking of
pollutants and sources was - developed using two approaches.
First, at the recommendation of the Johns Hopkins Risk Assessment
Review Panel (RARP), a ranking of pollutants and sources by
cancer potency-weighted ambient concentrations was developed.
This approach is advantageous because it (1) avoids relying on
highly uncertain exposure assumptions used by EPA in its risk
assessments (e.g./ continuous exposures for 70 years), and (2)
does not incorporate population weighing factors. The second
approach consisted of conducting a quantitative risk assessment
for the pollutants and sources examined. This latter analysis
considered both cancer and noncancer effects.
We emphasize that the risk estimates presented in this
chapter are highly uncertain and were calculated only for
screening purposes. For example, the preliminary emissions
estimates used in the modelling effort are based on limited
*This study estimated the increase in cancer and noncancer
risks resulting from exposure to ambient (i.e., outdoor)
concentrations of air toxics. It did not consider exposures
resulting from (1) indoor air, (2) the workplace, and (3) other
pathways (e.g., ingestion).
2As discussed in Chapter II, Summary of the Principles of
Risk Assessment and Management, we define "urban soup" as the
combination of toxic emissions from both point and area sources
to form elevated concentrations of pollutants in urban areas.
v-l
-------
information that was available in 1985. Since this time,
emission estimates at some sources have changed significantly
(e.g., Point Source A), while important new sources have been
introduced. Because of these changes, the numbers and the
conclusions presented below may no longer be valid. Also,
because of the generally conservative bias in the underlying data
used in the risk assessment, it is highly unlikely that the true
risks would be as high as the risk estimates, and they could be
considerably lower.
This chapter is organized into four sections. The first
describes the methodology and results for the RARP cancer
potency-weighted ranking scheme. The second summarizes the
results of the quantitative risk assessment for cancer and
noncancer effects. The third section provides perspective on the
risk assessment screening results by comparing (1) the Baltimore
risk screening estimates with the results of other IBMP and
urban-scale studies of air toxics, (2) the approach used in the
Baltimore study to estimate risks from the products of incomplete
combustion with other approaches used by EPA, and (3) the
Baltimore risk screening results for drinking water from Phase I
with the results of this more recent study of air toxics. The
final section of this chapter summarizes the major findings from
the risk assessment screening analyses.
1. RANKING OF TARGET COMPOUNDS AND SOURCES BY CANCER POTENCY-
WEIGHTED AMBIENT CONCENTRATIONS
As stated in Chapter III, quantitative risk assessment
incorporates numerous assumptions that are not readily
verifiable. Some of the more important ones are (1) the
percentage of time that people spend at given locations in the
study area, (2) the amount of air inhaled, (3) body-weight, and
(4) lifetime. Where these assumptions are inappropriate, the
cancer risk estimates and the conclusions drawn from these
estimates may be misleading.
in this section, at the suggestion of the Johns Hopkins
RARP, we describe how we ranked pollutants and sources by a
method that is simpler than the IEMP risk assessment approach
described in Chapter III. This alternative method avoids the
reliance on assumptions that relate to population exposure by
leaving out the population weighting factor from the quantitative
risk equation. We calculated the average area-wide concentration
of each target pollutant using the 5 km modelling grid system and
then multiplied this value by the pollutant's corresponding unit
cancer risk factor. The product is a cancer potency-weighted
average ambient concentration for each target pollutant in the
study area. The cancer potency-weighted average concentration is
simply a measure of each pollutant's potential human health risk
in the Baltimore area. Pollutants were compared and ranked by
V-2
-------
normalizing against the cancer potency-weighted concentration for
one of the more common pollutants in the ambient air.
We chose the potency-weighted average concentration of
chloroform as the base value and then divided the values for the
other pollutants by this number. The result, as shown in
Table V-l, is a ranking of pollutants in descending order of
potency-weighted concentration ratios. POM ranked highest with a
score approximately four times higher than the next highest
pollutant, chromium, and roughly twice as high as the summed
scores for all other pollutants. Carbon tetrachloride produced
the lowest score.
To rank sources, we first summed the potency-weighted
average concentrations resulting from each source or source
category and then normalized against the value for gasoline
marketing. Again, the potency-weighted concentration is simply a
measure of the potential risk to human health posed by a source
or source category in the Baltimore area. We included only those
sources that ranked high on the basis of quantitative risk
assessment (see Table v-8) to allow comparison of the two
approaches. We present the results of this''-ranking in Table V-2.
Point Source A ranks highest with a score approximately seven
times higher than the next highest source, road vehicles, and
roughly twice as high as the summed scores for all other sources.
Gasoline marketing produced the lowest score.
2. QUANTITATIVE RISK ASSESSMENT SCREENING RESULTS
In the Baltimore ambient air toxics study, risks were
quantified considering two health endpointst cancer and
noncancer effects. As discussed in Chapter IV, the calculations
for estimating the risks associated with ambient air exposures
used data on (1) ambient air concentrations based on either
dispersion modelling or monitoring, (2) population data (1980
U.S. Census), and (3) EPA unit cancer risk factors (see
Table IV-2) and threshold values for noncancer effects
(Table IV-3).3 We summarize the results of the risk screen for
cancer and noncancer effects below. We also list the limitations
that the reader must consider to interpret and use the results
presented in this chapter.
3The unit cancer risk factor combines estimates of cancer
potency with EPA standard exposure assumptions. The unit cancer
risk factor represents the upper-bound estimates of the increase
in cancer risk associated with continuous lifetime (i.e., 70
years) exposure to 1 ug/m3 of a specific compound.
v-3
-------
TABLE V-l
RELATIVE RANKING OF TARGET COMPOUNDS
BY CANCER POTENCY-WEIGHTED
AMBIENT CONCENTRATIONS
POLLUTANT
POM
CHROMIUM- 6
BENZENE
FORMALDEHYDE
METHYLENE
CHLORIDE
ARSENIC
CADMIUM
PERCHLOROETHYLENE
TRICHLOROETHYLENE
CHLOROFORM
ETHYLENE DICHLORIDE
ETHYLENE DIBROMIDE
ETHYLENE OXIDE
CARBON
TETRACHLORIDE
UNIT
CANCER RISK
FACTOR-
6.50E-05
1.20E-02
8.00E-06
1.30E-05
4.10E-06
4.30E-03
1.80E-Q3
4.80E-07
1.30E-06
2.30E-05
2.60E-05
2.20E-04
l.OOE-04
3.70E-06
AVERAGE
CONCENTRATION
-------
TABLE V-2
RELATIVE RANKING OF SOURCES BY CONTRIBUTION
TO CANCER POTENCY-WEIGHTED
AMBIENT CONCENTRATIONS
SOURCE
UNIT
CANCER
RISK
POLLUTANT FACTOR
AVERAGE
CONCEN-
TRATION
(ug/m3) (1)
POTENCY-
WEIGHTED
CONC.
SUM OF RATIO OF
POTENCY- SOURCES
WEIGHTED (gas
CONC. mrktng-1)
Point sources:
Point Source A POM 6.50E-05
Chrom.-6 1.20E-02
Benzene 8.00E-06
Point Source B Chrom.-6 1.20E-02
Point Source C Chrom.-6 1.20E-02
•Point Source D Chrom.-6 1.20E-02
0.86406 0.0000562 0.0000905 489
0.00245 0.0000294
0.60625 0.0000049
0.00027 0.0000033 0.0000033 18
0.00026 0.0000031 0.0000031 17
0.00011 0.0000013 0.0000013 7
POM 0.00005
Fnnldhyde 1.30E-05
Benzene 8.00E-06
Area sources:
Road vehicles
Solvent usage
Heating
Gas marketing Benzene 8.00E-06
Met. CL
PERC
TCE
POM
Arsenic
Cadmium
4.10E-06
4.80E-07
1.30E-06
0.00001
4.30E-03
1.80E-03
0.11494 0.0000057
0.27422 0.0000036
0.41250 0.0000033
1.34609 0.0000055
0.68156 0.0000003
0.68000 0.0000009
0.45625 0.0000046
0.00088 0.0000038
0.00100 0.0000018
0.0000126
0.0000067
0.0000101
68
36
55
0.02313 0.0000002 0.0000002
(1) Modelled concentration.
V-5
-------
a. Cancer Risks
We used different measures of cancer risk to characterize
urban soup in the Baltimore area. For area-wide risks, the most
important risk measures were: average increased lifetime
individual cancer risk and annual excess cancer incidence. For
so-called "hotspot" areas, the two risk measures of interest
were: the highest increased lifetime individual cancer risk at
three non-contiguous locations, and annual excess cancer
incidence in the grid cell of highest predicted cancer
incidence.*'5
i. Area-Wide Risks
(1) Average Increased Lifetime Individual Cancer
Risk
Results Based on Dispersion Modelling. We estimated the
average increased lifetime individual cancer risk posed by the
pollutants examined in the study area by (1) multiplying the
estimated average ambient air concentration by pollutant for each
5 km grid cell by the appropriate unit cancer risk factor, (2)
calculating an arithmetic average across all grid cells
evaluated, and (3) summing across all pollutants examined. The
resulting value indicated an average increased lifetime
individual cancer risk for the study area of approximately
1.5 x 10"*. The study area is shown with its UTM (XY)
coordinates in Figure V-l.
In general, we found that the average increased lifetime
individual cancer risk was fairly uniform across all grid cells.
The one exception was in the heavily industrialized southeast
section of the study area. These findings are graphically
presented in Figure v-2.
Figure V-2 also shows the relative contribution of point and
area sources to the estimated average increased lifetime
individual cancer risk by grid cell. As shown, point source
emissions are the major contributor to the average increased
lifetime individual cancer risk in the study area. Area source
emissions pose more or less uniform average increased lifetime
individual cancer risks in the area of downtown Baltimore; lower
average increased lifetime individual cancer risks were estimated
at the fringes of the study area. The higher average increased
*See Chapter II, Summary of the Principles of Risk
Assessment and Risk Management, for detail on the different risk
measures and their estimation.
5The term "hotspot" refers to specific areas in which people
are exposed to higher than average concentrations of the target
Dollutants.
pollutants
v-6
-------
FIGURE V-1
MODELLING DOMAIN
BY UTM COORDINATES
U
T
M
BALTIMORE COUNTY
4356
4346
PARKMUE
OUMMLK
BALTIMORE
CITY
ANNE ARUNDEL > .
COUNTY
UTM X
SPARROWS PONT
V-7
-------
FIGURE V-2
Baltimore IEMP Air Toxics
Point & Area Source Contribution to Average
Lifetime Individual Cancer Risk by Grid Cell
Average Lifetime Individual Cancer Risk
Point Source Contribution to Average Lifetime
Individual Cancer Risk
Area Source Contribution to Increased
Individual Cancer Risk
0.000
UTM COORDINATES
.Vote: The nsk scale for the area source contribution
is significant? larger than that used for the average
lifetime individual cancer risk and point source contribution
graphs.
V-8
-------
lifetime individual cancer risks in the heavily industrialized
southeast section of the study area (see the right-hand corner of
the second graph in Figure V-2) are attributable primarily to
point source emissions. Point sources also contribute more to
the lower average increased lifetime individual cancer risks
estimated in the northwest section of the study area, but to a
lesser extent that in the southeast section.
Results Based on Monitoring Data. Similar to the risk
analysis based on the modelled data, we used available monitoring
data from the past four years for the Baltimore area to
characterize area-wide risks. These data came from sampling
programs conducted by the AMA in 1985 through 1987 and from the
IEMF air toxics monitoring performed in 1983/1984. We also
estimate the average increased lifetime individual cancer risks
at two specific locations within the study area based on the
monitoring data available at two sites used in the TEAM study—
Dundalk and Parkvilie.
We estimated the average increased lifetime individual
cancer risk for the Baltimore study area by (1) calculating the
average concentration by pollutant across all of the AHA and IEMP
monitoring sites (see Appendix B for more detail), (2)
multiplying by the appropriate unit cancer risk factor, and (3)
summing across pollutants. The same steps were also taken to
estimate the average increased lifetime individual cancer risk at
the Dundalk .and Parkville sites. In determining the average
increased lifetime individual cancer risk, we assumed that all of
the detected chromium was the hexavalent or highly carcinogenic
form. In actuality, this is unlikely to be the case. Thus, the
average increased lifetime individual cancer risk from chromium
exposure must be seen as worse-case.
Table V-3 summarizes the total and pollutant-specific
average increased lifetime individual cancer risks using the
available monitoring data for the Baltimore area. All pollutants
individually pose increased lifetime individual cancer risks
equal to or greater than 1 x 10"£ (i.e., greater than one chance
in one million). Combined, the total average increased lifetime
individual cancer risk is approximately 5.0 x 10** area-wide.
Three pollutants account for most (approximately 87 percent) of
the estimated average increased lifetime individual cancer risk.
Perchloroethylene accounts for almost half (48 percent) of the
total, followed by hexavalent chromium (24 percent) and benzene
(15 percent).
Table V-4 summarizes the total and pollutant-specific
average increased lifetime individual cancer risks using the TEAM
monitoring data gathered at the Dundalk and Parkville sites.
The reader should note that the average increased lifetime
individual cancer risk is reported individually for vinylidene
chloride, but is not included in the summed risks because of the
V-9
-------
TABLE V-3
AVERAGE INCREASED LIFETIME INDIVIDUAL CANCER RISK
USING AVAILABLE MONITORING DATA (1/2)
RESULTS INTENDED FOR POLICY DEVELOPMENT ONLY
AVERAGE INCREASED
POLLUTANT LIFETIME INDIVIDUAL
[Weight of Evidence] CANCER RISK
Arsenic [Bl]
Benzene [A]
Benzo(a)pyrene [B2]
Cadmium [Bl]
Carbon tetrachloride [B2]
Chloroform [B2]
Chromium VI [A]
Ethylene dichloride [B2]
Propylene dichloride [C]
Ethyl benzene [N/A]
Lead [N/A]
Methylene chloride [B2]
Methyl isobutyl ketone [N/A]
Perchloroethylene [B2]
Toluene [N/A]
Methyl chloroform [N/A]
Trichloroethylene [B2]
Vinyl chloride [A]
Xylene [N/A]
7.0E-06
7.5E-05
2.5E-06
2.2E-06
3.4E-06
3.2E-05
1.2E-04
1.6E-05
8.6E-06
N/A
N/A
9.8E-06
N/A
2.4E-04
N/A
N/A
1.1E-06
N/A
N/A
TOTAL 5.2E-04
N/A: Not Applicable
(1) Source of monitoring data: See Appendix B.
(2) This study uses conservative estimates of increased cancer
risk from ambient (i.e., outdoor) exposure to establish
priorities among pollutants and sources. The risk estimates
are calculated using modelled or monitored concentrations
and EPA unit cancer risk factors. There is considerable
uncertainty in the estimated concentrations, which could
either overstate or understate the true concentrations (see
Chapter IV). Unit cancer risk factors combine CAG potency
estimates with EPA exposure assumptions. The CAG potency
estimates provide a plausible upper limit to the cancer risk
of a compound (see Appendix A); however, the true value of
the risk is unknown and may be as low as zero. The exposure
assumptions are extremely conservative in that they assume
continuous exposure to outdoor air for 70 years. Because of
the generally conservative bias in the information, it is
highly unlikely that the true risks would be as high as the
estimates and they could be considerably lower.
V-10
-------
TABLE V-4
BALTIMORE IEMP AIR TOXICS STUDY .
PHASE II RESULTS INTENDED FOR POLICY DEVELOPMENT1
AVERAGE INCREASED LIFETIME INDIVIDUAL CANCER RISKS
FROM EXPOSURE TO TARGET COMPOUNDS,
AT THE TWO FIXED MONITORING SITES2
Average"
Target Compound
Increased Lifetime CAG
Individual. Classificat
Cancer Risk
Level'
I Confidence
ion in Exposure
Estimate
Volatile OrganicBi
Benzene , .,
Carbon tetrachloride
Chlorobenzene
Chloroform 6
mDichlorobenzene
oDichlorobenzene
pDichlorobenzene
Ethylene dibromide
Formaldehyde
Methyl chloroform
Methylene Chloride
Percnloroethylene
Styrene ,
Tetrachloroethane
Trichloroethylene „
Vinylidene chloride
Metalst
3.1 x IQ'l
7.4 x 10'7
2.3~x 10'*
N.Q.
4.4~x 10^
2.2 x 10'3
1.7 x 10
2.9 x 10
-5
-1
-7
N.D.
5.2 x 10 ,
1.7 x 10°
A
B2
B2
IZ/2
B2
B2
C
B2
B2
C
Cumulative Lifetime Individual Cancer Risk*10
1.5 x 10'
Par]
Average
•kville
1
1
2
1
1
3
Arsenic!
Cadmium ,
Chromium- VI
7.7 x 10'*
2.0 x 10'*
2.4 x ID'5
A
Bl
A
2
Target Compound
Increased Lifetime CAG
Individual classification
Cancer Risk Level
Confidence
in Exposure
Estimate
Volatile organicai
Benzene
Carbon tetrachloride
Chlorobenzene
Chloroform
mDichlorobenzene
oDichlorobenzene
§Dichlorobenzene
thylene dibromide
Formaldehyde
Methyl chloroform ,,
Methylene chloride"
Percnloroethylene
Styrene .,
Tetrachloroethane
Trichloroethylene rt
Vinylidene chloride
Metalsi
.2 x
.1 x
N.D.
2.2 x
2.1 x
1.0
1.0
N.D.
1.3 x 1
3.4 x I
-5
-6
-S
-5
-3
-6
.
'!
'3
t;
B2
B2
Bl/2
B2
B2
C
B2
B2
C
2
1
1
1
3
1
Arsenic!
Cadmium ,
Chromium-vi
8.2
1.6
2.4
: x 10'*
x 10't
x 10
A
Bl
A
3
2
2
Cumulative Lifetime Individual Cancer Risk
114
1.2 x 10'
V-ll
-------
1
TABLE V-4 (CONTINUED)
NOTES
N.Q. = Not quantified
N.D. = Not detected
This study uses conservative estimates of increased cancer risk from
ambient (i.e., outdoor) exposure to establish priorities among
pollutants and sources. The risk estimates are calculated using
modelled or monitored concentrations and EPA unit cancer risk
factors. There is considerable uncertainty in the estimated
concentrations, which could either overstate or understate the true
concentrations (see Chapter IV). Unit cancer risk factors combine
CAG potency estimates with EPA exposure assumptions. The CAG
potency estimates provide a plausible upper limit to the cancer risl
of a compound (see Appendix A); however, the true value of the risk
is unknown and may be as low as zero. The exposure assumptions are
extremely conservative in that they assume continuous exposure to
outdoor air for 70 years. Because of the generally conservative
bias in the information, it is highly unlikely that the true risks
would be as high as the estimates, and they could be considerably
lower.
2 The two fixed ambient monitoring stations were located at the
Parkville Middle School in the Parkville section of Baltimore County
and the Merritt Point Mental Health Facility in the Dundalk section
of Baltimore County respectively.
3 Average monitored values are from the sampling period March 26 to
June 1, 1987.
* Unit risk (potency) estimates are current as of May 1987.
5 There are three levels of confidence in exposure estimates. Level 1
means that the exposure estimate is probably good. Level 2 means
that the exposure estimate may be good. Finally, level 3 means that
the exposure estimate is probably no good. The evaluations are
based on quality control audits.
6 m-dichlorobenzene levels could not be quantified at Dundalk.
7 NO tetrachloroethane was detected at a detection limit of 0.37
micrograms per cubic meter.
8 Average concentrations for the PM10 samples (i.e., particle sizes
equal to or smaller than 10 microns) were used in estimating risks.
9 Average chromium concentrations for the PM10 samples were used in
estimating exposure levels. We assumed that all of the chromium
that was collected was chromium-VI. The risk from exposure to
chromium is therefore overestimated for the purpose of this
screening exercise since the actual percentage of chromium-VI in the
sample is likely to be considerably lower.
(continued)
V-12
-------
TABLE V-4 (CONTINUED)
NOTES
10 The cumulative lifetime individual cancer risk estimate does not
include vinylidene chloride because of low confidence in the
exposure estimate.
11 For the monitoring samples in Parkville, no chloroform was detected
at the analytical level of detection of the laboratory/ which was
0.3 micrograms per cubic meter.
12 The monitoring results for methylene chloride in the ambient air of
Parkville are highly questionable due to analytical problems in
quantifying the amounts in the monitoring samples. Consequently,
for the purposes of comparison, the risk result is given, but it is
not included in the cumulative risk estimate.
13 No tetrachloroethane was detected at a detection limit of 0.37
micrograms per cubic meter.
The cumulative lifetime individual cancer risk estimate does not
include the estimate for methylene chloride and vinyl chloride
because of low confidence in the exposure estimates.
Vinylidene chloride was detected in the sampling, but there were
analytical problems in quantifying ambient concentrations. Because
vinylidene chloride could be an important compound, the individual
risk estimate for vinylidene chloride is reported in the risk
screening results; however, it is not included in the hazard index.
Additional analysis is needed to confirm the presence of vinylidene
chloride in the ambient air.
14
V-13
-------
extreme uncertainty surrounding its actual measured value.6 The
average increased lifetime individual cancer risk summed across
all pollutants at the Dundalk site is 1.5 x 10~* or two chances
in ten thousand and 1.2 x 10"* or roughly one chance in ten
thousand at Parkville. At the Dundalk site, ethylene dibromide
and benzene exposures account for approximately half of the
predicted average increased lifetime individual cancer risk. An
additional 42 percent of the risk is accounted for by hexavalent
chromium (16 percent), formaldehyde (15 percent), and methylene
chloride (11 percent). At Parkville, roughly 55 percent of the
estimated average increased lifetime individual cancer risk is
attributable to benzene (35 percent) and hexavalent chromium (20
percent); ethylene dibromide and formaldehyde each contribute an
additional 18 percent.
For the pollutants common to the TEAM sampling and the
available area-wide monitoring, there is close agreement in the
estimated risks for most compounds. With the exception of
perchloroethylene, which was found at a lower ambient level in
the TEAK study, the average increased lifetime individual cancer
risks by pollutant at the Dundalk and Parkville sites fall within
the range of the estimated average increased lifetime individual
cancer risks using the lowest and highest average measured values
from the available Baltimore area monitoring data (see
Table V-5).
We cannot easily compare the estimated ambient levels and,
hence, the average increased lifetime individual cancer risks for
our target pollutants from the monitoring and modelling studies.
The monitoring efforts were short-term (for example, the TEAM
sampling covered roughly six weeks during the transitional period
when winter ends and spring begins), whereas the modelling
estimates are annual averages that take into account the
different meteorological and industry operating conditions that
occur throughout the year. Nevertheless, the average increased
lifetime individual cancer risks from just the compounds that are
common to both the modelling and the monitoring studies are
close. From modelling, we would expect the average increased
lifetime individual cancer risks to be upwards of 1 x 10" . The
average increased lifetime individual cancer risks based on
monitoring data for this joint set of pollutants within the study
area are approximately 5 x 10~*. The average increased lifetime
6As discussed in Chapter IV, there were analytical problems
in quantifying the concentrations of vinylidene chloride. Based
on the advice of EPA's Office of Research and Development, we
show the estimated risk for vinylidene chloride in Table V-4 to
indicate its potential hazard as part of the risk screen;
however, we have not included this in the summed risks. ORD has
recently completed additional monitoring in Baltimore that may
help to confirm the ambient air concentrations of vinylidene
chloride. The results will be available in early 1989.
V-14
-------
TABLE V-5
AVERAGE INCREASED LIFETIME INDIVIDUAL CANCER RISKS
BASED ON AVAILABLE MONITORING DATA:
SITES WITH THE HIGHEST AND LOWEST MEASURED VALUES (1), (2)
RESULTS INTENDED FOR POLICY DEVELOPMENT ONLY
POLLUTANT
[Weight of Evidence]
Arsenic [Bl]
Benzene [A]
Benzo(a)pyrene [B2]
Cadmium [Bl]
Carbon tetrachloride [B2]
Chloroform [B2]
Chromium VI [A]
Ethylene dichloride [B2]
Propylene dichloride [CJ
Ethyl benzene [N/A]
Lead [N/A]
Methylene chloride [B2]
Methyl isobutyl. ketone [N
Perchloroethylene [B2]
Toluene [N/A]
Methyl chloroform [N/A]
Trichloroethylene [B2]
Vinyl chloride [A]
Xylene [N/A]
AVERAGE LIFETIME INDIVIDUAL
CANCER RISK
LOWEST MEASURED HIGHEST MEASURED
VALUE VALUE
1,
7,
6.9E-06
3.1E-05
6.6E-09
.8E-06
.4E-07
2.3E-06
2.4E-05
5.2E-06
1.8E-06
N/A
N/A
2.9E-06
N/A
3.5E-05
N/A
N/A
1.3E-08
N/A
N/A
7.3E-06
1,
3
1,
5,
1,
•OE-04
2E-06
8E-06
2E-06
1E-04
2.2E-04
6.8E-05
3.6E-05
N/A
N/A
1.7E-05
N/A
5.4E-04
N/A
N/A
5.1E-06
N/A
N/A
N/A: Not Applicable
(1) Source of monitoring data: See Appendix B.
(2) This study uses conservative estimates of increased cancer
risk from ambient (i.e., outdoor) exposure to establish
priorities among pollutants and sources. The risk estimates
are calculated using modelled or monitored concentrations
and EPA unit cancer risk factors. There is considerable
uncertainty in the estimated concentrations, which could
either overstate or understate the true concentrations (see
Chapter IV). Unit cancer risk factors combine CAG potency
estimates with EPA exposure assumptions. The CAG potency
estimates provide a plausible upper limit to the cancer risk
of a compound (see Appendix A); however, the true value of the
risk is unknown and may be as low as zero. The exposure
assumptions are extremely conservative in that they assume
continuous exposure to outdoor air for 70 years. Because of
the generally conservative bias in the information, it is
highly unlikely that the true risks would be as high as the
estimates, and they could be considerably lower.
v-15
-------
individual cancer risks for Dundalk and Parkville are 1.3 x 10"*
and 1.0 x 10"*, respectively.
(2) Annual Excess Cancer Incidence
Results Based on Dispersion Modelling. We determined the
estimated annual excess cancer incidence by pollutant for the
study area using the average increased lifetime individual cancer
risk estimates for each 5 km grid cell/ multiplying this value by
the number of people estimated to live in the grid cell, summing
across all cells, and then dividing this number by 70 years (an
assumed lifetime). We then summed across pollutants to estimate
area-wide annual excess cancer incidence.
The results of each step in the calculation of annual excess
cancer incidence are illustrated in Figure V-3. Each block in
the graph at the bottom corresponds to the number of people who
live in the area represented by the grid cell. As shown, the
greatest number of people live in downtown Baltimore. The blocks
in the middle graph show the average increased lifetime
individual cancer risk for each grid cell. Finally, in the top
graph, we show the estimated annual excess cancer incidence due
to the combination of the average increased lifetime individual
cancer risk and population living in the area represented by the
grid cell. The highest blocks correspond to the downtown area of
Baltimore.
Although the average increased lifetime individual cancer
risk was highest in the southeast section of the study area, the
highest predicted annual excess cancer incidence was for the
downtown area because of the high concentration of people living
in this area. Clearly, the differences in population among cells
have the greatest impact on our estimated annual excess cancer
incidence for a given pollutant. Blocks representing the
relative importance of the estimated annual excess cancer
incidence are superimposed upon a map of the study area in
Figure V-4. The weighing of our average increased lifetime
individual cancer risk estimates by population allows us to
identify the particular sources and pollutants that contribute
more to exposures where most individuals currently reside.
Table V-6 details the estimated annual excess cancer
incidence associated with ambient air exposures to each of the
pollutants (from all sources) included in the study. As shown,
the modelling analysis indicates that exposure to the selected
sources and pollutants analyzed in Baltimore will result in a
conservative estimate of approximately four excess cancer cases
each year. This estimate is highly uncertain because of the
limitations in the emissions calculations and the dispersion
modelling, as discussed in Chapter IV.
V-16
-------
FIGURE V-3
Baltimore IEMP Air Toxics
Annual Excess Cancer Incidence, Average Lifetime
Cancer Risk and Exposed Population by Grid Cell
Annual Excess Cancer Incidence
Average Lifetime Individual Cancer Risk
Exposed Population
V-17
-------
Annual Excess
Cancer Incidence
0.44
FIGURE V-4
Annual Excess Cancer Incidence by
Grid Cell for the Study Area
-------
TABLE V-6
ESTIMATED ANNUAL EXCESS CANCER INCIDENCE FOR SELECTED
POLLUTANTS MODELLED IN THE BALTIMORE IEMP AIR TOXICS STUDY
(TOTAL STUDY AREA, 5 KM GRID SYSTEM)1
RESULTS INTENDED FOR POLICY DEVELOPMENT ONLY
ANNUAL EXCESS
CANCER INCIDENCE PERCENTAGE
POLLUTANT (Weight of Evidence) OP TOTAL
Polycyclic organic
matter (POM)r 2.05 (N.A.) 57.4
Chromium (VI) 0.53 (A) 14.8
Methylene chloride 0.20 (62) 5.6
Benzene 0.20 (A) 5.6
Arsenic 0.20 (A) 5.6
Formaldehyde 0.19 (Bl) 5.3
Cadmium 0.09 (Bl) 2.5
Perchloroethylene (PCE) 0.04 (B2) 1.1
Trichloroethylene (TCE) 0.03 (B2) 0.8
Chloroform 0.02 (B2) 0.6
Ethylene dichloride (EDC) 0.01 (B2) 0.3
Ethylene dibromide (EDB) <0.01 (B2) 0.1
Benzo(a)pyrene (B(a)P) <0.01 (B2) <0.1
Carbon tetrachloride <0.01 (B2) <0.1
TOTAL1 3.56 100.0
N.A. = Not Available~~~
1 This study uses conservative estimates of increased cancer
risk from ambient (i.e., outdoor) exposure to establish
priorities among pollutants and sources. The risk estimates
are calculated using modelled or monitored concentrations and
EPA unit cancer risk factors. There is considerable
uncertainty in the estimated concentrations, which could
either overstate or understate the true concentrations (see
Chapter IV). Unit cancer risk factors combine CAG potency
estimates with EPA exposure assumptions. The CAG potency
estimates provide a plausible upper limit to the cancer risk
of a compound (see Appendix A); however, the true value of the
risk is unknown and may be as low as zero. The exposure
assumptions are extremely conservative in that they assume
continuous exposure to outdoor air for 70 years. Because of
the generally conservative bias in the information, it is
highly unlikely that the true risks would be as high as the
estimates, and they could be considerably lower.
2 The POM risk estimates are, in part, based on source-specific
unit cancer potency factors that have not undergone extensive
peer review. Thus, these numbers are subject to change.
3 Totals may not sum because of rounding.
V-19
-------
Almost all of the estimated annual excess cancer incidence
(approximately 94 percent) can be attributed to six pollutants:
POMs, chromium (VI), methylene chloride, benzene, arsenic, and
formaldehyde. Most of the estimated annual excess cancer
incidence (approximately 57 percent) results solely from POM
exposures. However, the POM estimates are based on unit cancer
risk factors that have not been extensively reviewed and are thus
more uncertain. An additional 15 percent of the estimated annual
excess cancer incidence can be attributed to hexavalent chromium.
Because there was no practical way for us to measure the actual
levels of hexavalent chromium in the ambient air, we could not
ascertain the accuracy of our exposure estimates (which assumed
that all chromium concentrations were hexavalent, its most potent
form) and hence the estimates of risk from this chemical.
Methylene chloride, benzene, arsenic, and formaldehyde each
account for between 5 and 6 percent of the total.
Table V-7 expands on Table V-6 by showing the portions of
the estimated annual excess cancer incidence attributable to
point and area sources. In the Baltimore study area, point and
area sources account for fairly similar percentages of the
estimated annual excess cancer incidence (45 percent and 55
percent, respectively). Figure V-5 shows this graphically.
The roughly even contribution of point and area sources to
the estimated annual excess cancer incidence does not hold for
individual pollutants. This is easily seen when the two key
pollutants in the analysis of cancer incidence, POM and
hexavalent chromium, are examined more closely. Whereas both
point and area sources account for the estimated annual excess
cancer incidence associated with POM (approximately 44 percent
versus 56 percent, respectively), point source emissions account
for almost all (99.6 percent) of the estimated annual excess
cancer incidence associated with hexavalent chromium.
Table V-8 adds an important dimension to the analysis by
identifying the specific source categories that account for the
estimated annual excess cancer incidence in the Baltimore area.
As shown, approximately 81 percent of the estimated annual excess
cancer incidence can be attributed to three sources: road
vehicles, Point Source A, and heating. These three sources also
account for all of the estimated annual excess cancer incidence
associated with POM, benzene, formaldehyde, ethylene dibromide,
and ethylene dichloride. Approximately 89 percent of the
estimated animal excess cancer incidence resulting from exposure
to POM ambient air concentrations comes from road vehicles and
Point Source A. The majority of the estimated annual excess
cancer incidence associated with hexavalent chromium
(approximately 87 percent) can be attributed to Point Sources A,
B, and C.
V-20
-------
TABLE V-7
NI
ESTIMATED ANNUAL EXCESS CANCER INCIDENCE FOR SELECTED
POLLUTANTS MODELLED IN THE BALTJMORE, IEMP AIR TOXICS STUDY BY SOURCE (1)
(TOTAL STUDY AREA. 5KM GRID SYSTEM)
RESULTS INTENDED FOR POLICY DEVELOPMEHT ONLY
POLLUTANT
[Weight of Evidence]
A.J
POLYC ORG MATTER (POM)(2) [N
CHROMIUM HEXAVALENT [A]
METHYLENE CHLORIDE [B2]
BENZENE [A)
ARSENIC [A]
FORMALDEHYDE [Bl]
CADMIUM [Bl)
PERCHLOROETHYLENE (PCE)
TRICHLOROETHYLENE (TCE)
CHLOROFORM [B2]
ETHYLENE DICHLORIDE (EDC) [B2]
ETHYLENE BROMIDE (EDB) [B2]
CARBON TETRACHLORIDE [B2]
ETHYLENE OXIDE [Bl]
TOTAL (3)
1C]
[B2]
ANNUAL
EXCESS
INCIDENCE:
POINT
SOURCES
0.90
0.53
0.01
0.09
O.OS
<0.01
0.02
<0.01
<0.01
<0.01
<0.01
<0.01
PERCENTAGE
OF TOTAL:
POINT SOURCES
25.2
14.8
0.2
2.5
1.5
0.1
0.6
<0.1
0.1
0.1
ANNUAL
EXCESS
INCIDENCE:
AREA
SOURCES
1.15
<0.01
0.20
0.11
0.15
0.18
0.07
0.04
0.03
0.02
<0.01
<0.01
PERCENTAGE
OF TOTAL:
AREA SOURCES
32.4
0.1
5.5
3.2
4.1
5.2
1.9
1.0
0.9
0.6
0.1
0.1
1.60
44.99
1.95
54.93
ANNUAL
EXCESS
CANCER
INCIDENCE
2.05
0.53
0.20
0.20
0.20
0.19
0.09
0.04
0.03
0.02
<0.01
<0.01
<0.01
<0.01
3.56
N.A. = Not Available
(1) This study uses conservative estimates of increased cancer risk from ambient (i.e., outdoor)
exposure to establish priorities among pollutants and sources. The risk estimates are calculated
using modelled or monitored concentrations and EPA unit cancer risk factors. There is
considerable uncertainty in the estimated concentrations, which could either overstate or
understate the true concentrations (see Chapter IV). Unit cancer risk factors combine CAG
potency estimates with EPA exposure assumptions. The CAG potency estimates provide a plausible
upper limit to the cancer risk of a compound (see Appendix A); however, the true value of the
risk is unknown and may be as low as zero. The exposure assumptions are extremely conservative
in that they assume continuous exposure to outdoor air for 70 years. Because of the generally
conservative bias in the information, it is highly unlikely that the true risks would be as high
as the estimates, and they could be considerably lower.
(2) The POM risk estimates are based on source specific unit risk factors that have not undergone
extensive peer review. Thus, these numbers are subject to change.
(3) The numbers shown may not sum to the totals because of rounding.
-------
FIGURE V-5
Baltimore IEMP Air Toxics
Point & Area Source Contribution to Annual
Excess Cancer Incidence by Grid Call
Annual Excess Cancer Incidence
Point Source Contribution to Annual
Excess Cancer Incidence
Area Source Contribution to Annual
Excess Cancer Incidence
0.26
0.23
0.20
0.17
0.14
0.11
0.09
0.06
0.03
UTM COORDINATES
V-22
-------
f
N)
OJ
POLLUTANT
(Weight of Evidence]
TABLE V-B
BALTIMORE 1EMP AIR TOXICS STUDY
ESTIMATED ANNUAL EXCESS CANCER INCIDENCE
BASED ON MODELLING BY SOURCE AND POLLUTANT (1)
(TOTAL STUDY AREA. 5 KM GRID SYSTEM)
RESULTS INTENDED FOR POLICY DEVELOPMENT ONLY
ROAD POINT SOLVENT POINT POINT POINT ALL
VEHICLES SOURCE A HEATING USAGE SOURCE B SOURCE C SOURCE D OTHER
TOTAL ANNUAL
EXCESS CANCER
INCIDENCE
ARSENIC [A]
BENZENE [A]
CADMIUM (Bl]
CARBON TETRACHLORIDE [B2]
CHLOROFORM [B2]
CHROMIUM HEXAVALENT [A]
ETHYLENE OXIDE [Bl]
FORMALDEHYDE [Bl]
METHYLENE CHLORIDE [B2]
POLYC ORG MATTER (POM)[N.A.] (2)
ETHYLENE DIBROMIDE
-------
In reviewing these results, it should be noted that there is
considerable uncertainty underlying the modelled estimates of
ambient concentration. The model performance evaluation (see
Chapter IV) suggests that we have underpredicted actual ambient
levels for some pollutants. There is also information, according
to the Maryland AHA, showing that modelled ambient air
concentrations at Point Source A are generally much greater than
those that have been found at the facility using a detailed
monitoring network. Because Point Source A contributes so
heavily to the estimated human health risks, any revision to its
estimated emissions will have a significant impact on our risk
results. Finally, our emissions estimates were based on the
rough information available at the start of this study.
Completion of a more detailed analysis considering the changes
over recent years could lead to different findings.
Results Based on Monitoring. Using available monitoring
data for the Baltimore area, we estimated the annual excess
cancer incidence by simply multiplying the area-wide average
increased lifetime individual cancer risks by the total exposed
population used in the modelling analysis, approximately 1.6
million. Table V-9 presents the results of these calculations.
As shown/ we predict a larger annual excess cancer incidence—
roughly 12 annual cases—using monitoring versus modelled data.
For comparison, Table V-9 shows the modelled annual excess cancer
incidence for those pollutants common to the monitoring and
modelling studies. As shown, the difference in the predicted
annual excess cancer incidence is explained by significantly
higher incidence estimates based on the monitoring data for
perchloroethylene, chromium VI, and benzene.
There is considerable uncertainty surrounding the incidence
calculations based on monitoring data. Although these monitoring
data represent the best readily available information for the
Baltimore area (other than the TEAM results), the results used in
this analysis are drawn, by pollutant, from different studies,
and not all results have undergone extensive peer review. Also,
exposures were estimated by assuming that all individuals are
exposed to the same ambient pollutant levels represented by the
average of the measured concentrations across the monitoring
sites. Despite these limitations, the annual excess cancer
incidence estimates based on the available monitoring data
suggest that some exposures could be dramatically higher than
those characterized by dispersion modelling. Accordingly, the
actual ambient air concentrations and sources of
perchloroethylene, chromium VI, and benzene deserve further
attention.
ii. "Hotspot" Risks
A hotspot is defined as an area in which people are exposed
to higher than average concentrations. Given available data and
resources, dispersion modelling is the only practical way to
V-24
-------
TABLE V-9
AREA-WIDE ANNUAL EXCESS CANCER INCIDENCE
USING AVAILABLE MONITORING DATA (1),(2)
RESULTS INTENDED FOR POLICY DEVELOPMENT ONLY
POLLUTANT
[Weight of Evidence]
ANNUAL EXCESS
CANCER INCIDENCE
ANNUAL EXCESS
CANCER INCIDENCE
(MODELLED FROM
TABLE V-6)
Arsenic [Bl]
Benzene [A]
Benzo(a)pyrene [B2]
Cadmium [Bl]
Carbon tetrachloride [B2]
Chloroform [B2]
Chromium VI [A]
Ethylene dichloride [B2]
Propylene dichloride [C]
Ethyl benzene [N/A]
Lead [N/A]
Methylene chloride [B2]
Methyl isobutyl ketone [N/A]
Perchloroethylene [ B2 ]
Toluene [N/A]
Methyl chloroform [N/A]
Trichloroethylene [B2]
Vinyl chloride [A]
Xylene [N/A]
0.16
1.72
0.06
0.05
0.08
0.73
2.74
0.37
0.20
N/A
N/A
0.22
N/A
5.44
N/A
N/A
0.03
N/A
N/A
0.20
0.20
<0.01
0.09
<0.01
0.02
0.53
0.01
N.A.
N/A
N/A
0.20
N/A
0.04
N/A
N/A
0.03
N.A.
N/A
TOTAL
11.79
1.32
N/A: Not applicable; not a proven human carcinogen
N.A.: Not available
(1) Source of monitoring data: See Appendix B.
(2) This study uses conservative estimates of increased cancer risk
from ambient (i.e., outdoor) exposure to establish priorities among
pollutants and sources. The risk estimates are calculated using
modelled or monitored concentrations and EPA unit cancer risk
factors. There is considerable uncertainty in the estimated
concentrations, which could either overstate or understate the true
concentrations (see Chapter IV). Unit cancer risk factors combine
CAG potency estimates with EPA exposure assumptions. The CAG
estimates provide a plausible upper limit to the cancer risk of a
compound (see Appendix A); however, the true value of the risk is
(see Appendix A); however, the true value of the risk is unknown
and may be as low as zero. The exposure assumption are extremely
conservative in that they assume continuous exposure to outdoor
air for 70 years. Because of the generally conservative bias
in the information, it is highly unlikely that the true risks
would be as high as the estimates, and they could be
considerably lower.
V-25
-------
pinpoint those areas with above-average exposure levels. We
relied on the results of our dispersion modelling to locate (1)
discrete receptor locations with the highest estimated increased
lifetime individual cancer risk, and (2) the grid cell (using the
2.5 km refined grid system) of highest predicted annual excess
cancer incidence.
(1) The Highest Estimated Increased Lifetime
Individual Cancer Risks at Three Non-
Contiguous Sites
As Figure V-2 suggests, emissions from point sources can
lead to relatively high localized average increased lifetime
individual cancer risks. We examined these average increased
lifetime individual cancer risks in greater detail by focusing
our attention on the three non-contiguous hotspot locations
(separated by more than 2 kilometers) with the highest increased
lifetime individual cancer risks. Figure V-6 shows the location
of the three hotspots. The purpose of choosing non-contiguous
sites was to analyze hotspots resulting from the emissions from
different combinations of sources. These hotspot locations were
modelled as discrete receptor locations rather than as a grid
cell.
Table V-10 shows the increased lifetime individual cancer
risks by pollutant at the three hotspots. Their exact location
is identified by UTM coordinates. There is approximately a
factor of two difference between the highest and the lowest of
the three individual cancer risks (1.3 x 10° versus 5.4 x 10' ).
The hotspot with the highest increased lifetime individual cancer
risk is approximately a factor of nine higher than the estimated
average area-wide increased lifetime individual cancer risk
(1.5 x 10"*); the hotspot with the lowest increased lifetime
individual cancer risk is approximately four times higher than
the area-wide average.
At the hotspot exhibiting the highest increased lifetime
individual cancer risk/ POM and hexavalent chromium account for
roughly 96 percent of the increased lifetime individual cancer
risk. In addition, exposure to benzene in the ambient air
results in an increased lifetime individual cancer risk of
2.2 x 10'3. These pollutants are also the key compounds of
concern at the hotspot with the second highest increased lifetime
individual cancer risk. At the third hotspot site, POM and
hexavalent chromium continue to account for most (approximately
79 percent) of the estimated increased lifetime individual cancer
risks. However, arsenic (6.1 x 10'5) and cadmium (2.6 x 10' )
also contribute significantly to these increased lifetime
individual cancer risks.
Table V-ll summarizes the increased lifetime individual
cancer risks by pollutant and individual source for each of the
three hotspot locations in order of highest to lowest risk. As
V-26
-------
FIGURE V-6
LOCATION OF HOTSPOTS
EVALUATED IN THE AIR TOXICS STUDY
BALTIMORE COUNTY
\
BALTIMORE
\ CITY
/ANNE ARUNDEL
COUNTY/^
HOTSPOT 3
UTM: 4350.90 364.55
HOTSPOT 1
UTM: 4343.00 374.12
HOTSPOT 2
UTM: 4339.85 374.88
V-27
-------
TABLE V-10
POLLUTANT CONTRIBUTIONS BASED ON MODELLING TO
MAXIMUM INCREASED LIFETIME INDIVIDUAL CANCER RISK
AT THREE HOTSPOT LOCATIONS (1)
RISK RESULTS INTENDED FOR POLICY DEVELOPMENT ONLY
MAXIMUM LIFETIME
POLLUTANT INDIVIDUAL CANCER
[Weight of Evidence]
Hotspot 1: 4343.00 374.12
CHROMIUM [A]
POLYCLYCLIC ORGANIC MATTER (POM) [N.A.] (3)
BENZENE [A]
ARSENIC JA]
FORMALDEHYDE fBll
METHYLENE CHLORIDE [B2]
CADMIUM [Bl]
TRICHLOROETHYLENE [B2]
PERCHLOROETHYLENE [C]
CARBON TETRACHLORIDE [82)
ETHYLENE DICHLORIDE [B2]
ETHYLENE OXIDE [B1J
ETHYLENE DIBROMIDE [B2]
CHLOROFORM [B2]
TOTAL (4)
Hotspot 2: 4339.85 374.88
POLYCLYCLIC ORGANIC MATTER
-------
TABLE V-10 (CONTINUED)
NOTES
1 Only the sites with the highest increased lifetime individual
cancer risk that were separated by at least two kilometers were
chosen. The purpose was to examine the sites of maximum
lifetime individual cancer risk affected by the emissions from
different combinations of sources.
2 This study uses conservative estimates of increased cancer risk
from ambient (i.e., outdoor) exposure to establish priorities
among pollutants and sources. The risk estimates are
calculated using modelled or monitored concentrations and EPA
unit cancer risk factors. There is considerable uncertainty in
the estimated concentrations, which could either overstate or
understate the true concentrations (see Chapter IV). Unit
cancer risk factors combine CAG potency estimates with EPA
exposure assumptions. The CAG potency estimates provide a
plausible upper limit to the cancer risk of a compound (see
Appendix A); however, the true value of the risk is unknown and
may be as low as zero. The exposure assumptions are extremely
conservative in that they assume continuous exposure to outdoor
air for 70 years. Because of the generally conservative bias
in the information, it is highly unlikely that the true risks
would be as high as the estimates, and they could be
considerably lower.
3 The POM risk estimates are based on source-specific unit risk
factors of which only POM from coke oven emissions has
undergone extensive peer review. Thus, these numbers are
subject to change. There is no overall weight-of-evidence
determination for POM.
* Totals may not sum because of rounding.
V-29
-------
TABLE V-ll
TOTAL SOURCE AND POLLUTANT CONTRIBUTION
BASED ON MODELLING AT EACH HOTSPOT LOCATION (1,2)
(HOTSPOT LOCATION 1: 4343.00 374.12)
(URBAN SCALE, 2.5 KM GRID SYSTEM)
RESULTS INTENDED FOR POLICY DEVELOPMENT ONLY
POLLUTANT
ROAD
[Weight of Evidence] VEHICLES
ARSENIC (A]
BENZENE [A] 3
CADMIUM [Bl] <1
CARBON TETRACHLORIDE [B2]
CHLOROFORM [ B2 ]
CHROMIUM HEXAVALENT [A]
ETHYLENE OXIDE [B1J
FORMALDEHYDE [ Bl ] 3
METHYLENE CHLORIDE [B2]
POLY ORG MATTER (POM) [NA](3) 2
ETHYLENE DI BROMIDE (EDB)[B2] <1
ETHYLENE DICHLORIDE (EDC)fB2]
-------
TABLE V-ll (CONTINUED)
TOTAL SOURCE AND POLLUTANT CONTRIBUTION
BASED ON MODELLING AT EACH HOTSPOT LOCATION(1,2)
(HOTSPOT LOCATION 2: 4339.85 374.88)
(URBAN SCALE. 2.5 KM GRID SYSTEM)
RESULTS INTENDED FOR POLICY DEVELOPMENT ONLY
POLLUTANT
ROAD POINT SOLVENT POINT POINT
[Weight of Evidence] VEHICLES SOURCE A HEATING USAGE SOURCE B SOURCE C
ARSENIC [A]
BENZENE [A] 2
CADMIUM [Bl] <1
< CARBON TETRACHLORIDE [B2]
' CHLOROFORM [ B2 ]
£ CHROMIUM HEXAVALENT [A]
ETHYLENE OXIDE (Bl]
FORMALDEHYDE [ Bl ] 3
METHYLENE CHLORIDE [B2]
POLY ORG MATTER (POM) [NA](3) 2
ETHYLENE DIBROMIDE (EDB)[B2] <1
ETHYLENE DICHLORIDE (EDC)[B2]<1
TRICHLOROETHYLENE [B2]
PERCHLOROETHYLENE 1 C 1
<1. OOE-06 2. OOE-06 <1. OOE-06 <1. OOE-06
.OOE-06 3.00E-05
.OOE-06 <1. OOE-06 1. OOE-06 <1. OOE-06
8.20E-05 3. OOE-06 4. OOE-06
.OOE-06 1. OOE-06
4. OOE-06
. 70E-05 6.00E-04 2. OOE-06
.OOE-06
.OOE-06
<1. OOE-06 <1. OOE-06
<1. OOE-06
TOTAL
INCREASED
LIFETIME
INDIVIDUAL
POINT ALL CANCER
SOURCE D OTHER
2.50E-06
2. OOE-06
1.20E-06
4. OOE-09
3.90E-07
2. OOE-06 2. OOE-06
6.20E-OB
6. OOE-07
4. OOE-07
<1. OOE-06
9.60E-08
1.20E-07
7.60E-07
6.90E-07
RISK
4.50E-06
3.40E-OS
2.20E-06
4. OOE-09
3.90E-07
9.30E-05
6.20E-08
4.60E-06
4.40E-06
6.29E-04
9.60E-08
1.20E-07
7.60E-07
6.90E-07
TOTAL INCREASED LIFETIME
INDIVIDUAL CANCER RISK (4)
3.20E-05 7.12E-04 6.OOE-06 4.OOE-06 3.OOE-06 4.OOE-06 2.OOE-06 1.08E-05 7.74E-04
(continued)
-------
to
TABLE V-ll (CONTINUED)
TOTAL SOURCE AND POLLUTANT CONTRIBUTION
BASED ON MODELLING AT EACH HOTSPOT LOCATION (1,2)
(HOTSPOT LOCATION 3: 4350.90 364.55)
(URBAN SCALE, 2.5 KM GRID SYSTEM)
POLLUTANT
[Weight of Evidence]
ROAD POINT
VEHICLES SOURCE A
HEATING
ARSENIC (A]
BENZENE [A] 4.OOE-06
CADMIUM [Bl] <1.00E-06
CARBON TETRACHLORIDE [B2]
CHLOROFORM [B2]
CHROMIUM HEXAVALENT [A]
ETHYLENE OXIDE [Bl]
FORMALDEHYDE [Bl] 4 . DOE-06
METHYLENE CHLORIDE [B2]
POLYC ORG MATTER (POM)[NA](3) 3.70E-05
ETHYLENE DIBROMIDE (EDB)[B2] <1.00E-06
ETHYLENE DICHLORIDE (EDC)[B2] <1.00E-06
TRICHLOROETHYLENE [B2]
PERCHLOROETHYLENE [C]
i.OOE-06
9.OOE-06
.OOE-06
.OOE-06
.OOE-06
2.OOE-06
5.00E-05 4.OOE-06
DEVELOPMENT ONLY
SOLVENT POINT POINT
USAGE SOURCE B SOURCE C
5.40E-05 <1.00E-06
2.20E-05
3.01E-04 l.OOE-05
6.00E-06
l.OOE-06
l.OOE-06
POINT ALL
SOURCE 0 OTHER
3.00E-06
L.60E-06
2.00E-06
l.OOE-08
6.90E-07
3.00E-06 <1.00E-06
9.00E-08
9.00E-07
1.50E-06
4. OOE-06
1.40E-07
1.80E-07
2.00E-07
3.00E-07
TOTAL
INCREASED
LIFETIME
INDIVIDUAL
CANCER
RISK
6.10E-OS
9.60E-06
2.60E-05
l.OOE-08
6.90E-07
3.23E-04
9.00E-08
6.90E-06
7.SOE-06
9.50E-05
1.40E-07
1.80E-07
1.20E-06
1.30E-06
TOTAL INCREASED LIFETIME
INDIVIDUAL CANCER RISK (4)
NA = Not Available
4.50E-05 6.30E-05 1.20E-OS 8.OOE-06 3.77E-04 l.OOE-05 3.OOE-06 1.46E-05 S.33E-04
(1) This study uses conservative estimates of increased cancer risk from ambient (i.e., outdoor) exposure to
establish priorities among pollutants and sources. The risk estimates are calculated using modelled or
monitored concentrations and EPA unit cancer risk factors. There is considerable uncertainty in the estimated
concentrations, which could either overstate or understate the true concentrations (see Chapter IV). Unit
cancer risk factors combine CAG potency estimates with EPA exposure assumptions. The CAG potency estimates
provide a plausible upper limit to the cancer risk of a compound (see Appendix A); however, the true value of
the risk is unknown and may be as low as zero. The exposure assumptions are extremely conservative in that
they assume conintuous exposure to outdoor air for 70 years. Because of the generally conservative bias in the
information, it is highly unlikely that the true risks would be as high as the estimates, and they could be
considerably lower.
(<2) Risk numbers are in scientific notation. Thus 1E-06 is the same as 1 x 10 to the sixth power or one in a million.
(3) The POM risk estimates are based on source specific unit risk factors that have not undergone
extensive peer review. Thus, these numbers are subject to change.
(4) The numbers shown may not sum to the totals because of rounding
-------
shown, the lifetime individual cancer risks at each of these
sites are dominated by emissions from Point Source A. At the
first two receptor locations, Point Source A accounts for
approximately 91 percent of the increased lifetime individual
cancer risks. At the third hotspot location, Point Source A is
still a major player but its contribution to the increased
lifetime individual cancer risk is lower, approximately 69
percent. Point Sources A, B, and C together account for roughly
83 percent of the increased lifetime individual cancer risks at
the third site. Road vehicles and heating account for an
additional 11 percent. The pollutants of concern at each of the
hotspot locations are also of important contributors to the
estimated annual excess cancer incidence.
(2) Estimated Annual Excess Cancer Incidence in
the Grid Cell of Highest Predicted Annual
Excess Cancer Incidence
In Table V-12, we provide a detailed examination of the 2.5
km grid with the highest estimated area-wide annual excess cancer
incidence. For this analysis, we used the more accurate
estimates associated with the 2.5 kilometer (refined) grid system
(see Figure IV-4).
The grid with the highest estimated annual excess cancer
incidence can be found in the center of the study area. The
average increased lifetime individual cancer risk at this
location is 3 x 10"*, and the predicted annual excess cancer
incidence is approximately 0.15. The grid, which covers less
than 1 percent of the study area, accounts for about 4 percent of
the estimated area-wide annual excess cancer incidence. Almost
99 percent of the annual excess cancer cases are attributable to
seven compounds: hexavalent chromium, POMs, arsenic, cadmium,
benzene, methylene chloride, and formaldehyde. Hexavalent
chromium and POMs are the primary contributors to the annual
excess cancer incidence in the grid cell of highest incidence, as
well as the estimated annual excess cancer incidence for the
study area. However, hexavalent chromium, and not POM, explains
most of the estimated annual excess cancer incidence in the grid
cell of highest incidence.
Table V-13 arrays the estimated annual excess cancer
incidence in the grid cell of highest incidence by facility and
pollutant. Almost 91 percent is attributable to four sources:
Point Source A, Point Source B, road vehicles, and heating.
Point Source B accounts for almost half of the estimated annual
excess cancer incidence at the location of maximum incidence.
Approximately 85 percent of the estimated annual excess cancer
incidence associated with hexavalent chromium is also
attributable to Point Source B. Almost all of the estimated
annual excess cancer incidence associated with POM can be
explained by environmental releases from Point Source A, road
vehicles, and heating.
V-33
-------
TABLE V-12
BALTIMORE IEMP AIR TOXICS STUDY
ESTIMATED ANNUAL EXCESS CANCER INCIDENCE
BASED ON MODELLING BY POLLUTANT IN THE GRID CELL OF
HIGHEST PREDICTED ANNUAL EXCESS CANCER INCIDENCE (1)
RESULTS FOR POLICY DEVELOPMENT ONLY
POLLUTANT
[Weight of Evidence]
Chromium (VI) [AJ
Polycyclic organic
matter (POM) [N.A.] (2)
Arsenic [A]
Cadmium [Bl]
Benzene [A]
Methylene chloride [B2]
Formaldehyde [Bl]
Perchloroethyiene (Perc) [C]
Trichloroethylene (TCE) [B2]
Chloroform [B2]
Ethylene dichloride (EDC) [B2]
Ethylene dibromide (EDB) [B2]
Ethylene oxide [Bl]
Carbon tetrachloride [B2]
ANNUAL EXCESS
CANCER
INCIDENCE
0.10
0.04
0.01
Oft 1
. Ul
<0.01
<0.01
<0.01
<0.01
<0.01
f\ f\ 1
<0 . 01
<0.01
<0.01
<0.01
<0.01
PERCENTAGE
OF TOTAL
47.2
29.5
9.6
a i
*t « x
3n
. u
2.7
2.6
0.5
0.4
01
• i
0.1
<0.1
TOTAL (3)
0.15 100.0
N.A. = Not available.
This study uses conservative estimates of increased cancer risk
from ambient (i.e., outdoor) exposure to establish priorities
among pollutants aAd sources. The risk estimates are calculated
using modelled or monitored concentrations and EPA unit cancer
risk factors. There is considerable uncertainty in the estimated
concentrations, which could either overstate or understate the
t?Se concentrations (see Chapter IV). Unit Cancer risk factors
combine CAG potency estimates with EPA exposure assumptions. The
CAG potency estimates provide a plausible upper limit to the
cancer risfc of a compound (see Appendix A); however, the true risk
is unknown and may be as low as zero. The exposure.assumptions
are eSremely conservative in that they assume continuous exposure
II outdoor air for 70 years.. Because °f the generally conservative
bias in the information, it is highly unlikely that the true risks
would be as high as the estimates and they could be considerably
lower.
i7\ The POM risk estimates are based, in part, on source-specific
( } unit risk factors that have not undergone extensive peer review.
Thus, these numbers are subject to change.
(3) Totals may not sum because of rounding.
V-34
-------
OJ
Ul
TABLE V-13
BALTIMORE IEMP AIR TOXICS STUDY
ESTIMATED MAXIMUM ANNUAL EXCESS CANCER INCIDENCE
BY POLLUTANT AND SOURCE IN THE GRID CELL OF HIGHEST
ESTIMATED ANNUAL EXCESS CANCER INCIDENCE (1)
(URBAN SCALE, 2.5 KM GRID SYSTEM)
RESULTS INTENDED FOR POLICY DEVELOPMENT ONLY
POLLUTANT
[Weight of Evidence]
POINT POINT ROAD
SOURCE B SOURCE A VEHICLES
HEATING
SOLVENT POINT POINT
USAGE SOURCE C SOURCE D
ARSENIC [A]
BENZENE [A]
CADMIUM [Bl]
CARBON TETRACHLORIDE [B2]
CHLOROFORM [B2]
CHROMIUM HEXAVALENT [A]
ETHYLENE OXIDE [Bl]
FORMALDEHYDE [Bl]
METHYLENE CHLORIDE [62]
POLYC ORG MATTER (POM)[N.A.]
ETHYLENE DIBROMIDE (EDB)[B2]
ETHYLENE DICHLORIDE (EDC)[B2]
TRICHLOROETHYLENE [B2]
PERCHLOROETHYLENE [C]
TOTAL ANNUAL EXCESS
CANCER INCIDENCE (3).(4)
PERCENTAGE OF TOTAL (3).(4)
0.06
(2)
0.07
47.6Z
<0.01
0.02
0.02
13.8Z
<0.01
0.02
<0.01
<0.01
0.02
13.52
<0.01
<0.01
<0.01
0.00
O.OZ
TOTAL
ANNUAL EXCESS
ALL CANCER
OTHER INCIDENCE
0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
0.00
o.oz
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01 0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
0.01
<0.01
<0.01
<0.01
<0.01
0.07
<0.01
<0.01
<0.01
0.04
<0.01
<0.01
<0.01
<0.01
0.00
O.OZ
0.00
O.OZ
0.04
26.71
0.15
100.OZ
N.A. = Not Available
(1) This study uses conservative estimates of increased cancer risk from ambient (i.e.. outdoor) exposure to
establish priorities among pollutants and sources. The risk estimates are calculated using modelled or
monitored concentrations and EPA unit cancer risk factors. There is considerable uncertainty in the estimated
concentrations, which could either overstate or understate the true concentrations (see Chapter IV). Unit
cancer risk factors combine CAG potency estimates with EPA exposure assumptions. The CAG potency estimates
provide a plausible upper limit to the cancer risk of a compound (see Appendix A); however, the true value of
the risk is unknown and may be as low as zero. The exposure assumptions are extremely conservative in that
they assume continuous exposure to outdoor air for 70 years. Because of the generally conservative bias in the
information, it is highly unlikely that the true risks would be as high as the estimates, and they could be
considerably lower.
(continued)
-------
TABLE V-13 (CONTINUED)
(2) The POM risk estimates are based, in part, on source-specific unit risk factors that have
not undergone extensive peer review. Thus, these numbers are subject to change.
(3) The POM risk estimates are based on source specific unit risk factors that have not undergone
extensive peer review. Thus, these number are subject to change.
(A) The numbers shown may not sum to the totals because of rounding.
-------
b. Noncancer Risks
i. Results Based on Dispersion Modelling
The analysis of noncancer risks considered two exposure
scenarios: (1) modelled concentrations that were generated for a
subset of the study area using the refined (i.e., 2.5 km) grid
system, and (2) modelled concentrations resulting from an in-
depth analysis of ambient air concentrations at 118 discrete
locations near roughly 40 sources.7 The analysis of noncancer
risks focused on 18 pollutants, of which 12 are also
carcinogenic.
As discussed earlier in Chapter II, we analyzed noncancer
effects in two ways. First, we divided the predicted ambient air
concentration of each pollutant evaluated in the modelling
exercise by the no-effect threshold(s) for those noncancer
effects (e.g., liver toxicity, kidney toxicity, reproductive,
neurological, fetal or blood) relevant to the pollutant, If the
resulting ratio exceeded one, we identified these pollutant-
specific exposures as requiring additional investigation.
Second, we developed a "hazard index" that .summed individual
pollutant ratios by effect category. The latter analysis was
aimed at examining the impact of exposure to complex chemical
mixtures in the ambient air. If the index exceeded one, we
identified these exposures as deserving further analysis.
(1) Pollutant Specific
In both the refined grid and discrete receptor analysis, we
found locations at which the estimated benzene and xylene ambient
air concentrations exceed several no-effect thresholds. The
refined (2.5 km) grid analysis indicates that there is a need to
further examine benzene exposures in the southeast quadrant of
the study area (see Table V-14) where benzene ambient concentra-
tions exceed the no-effect threshold for blood effects. The
results from the discrete receptor analysis support these
findings and show additional sites at which benzene exposures
should also be studied in greater detail. The discrete receptor
analysis also identifies a section of the study area in which
xylene exposures may need to be more carefully examined because
they exceed the threshold levels for the following noncancer
7See Chapter IV for an overview of the three levels at which
modelling and population exposures were considered: 5 km grid
cell, 2.5 km grid cell, and discrete receptor locations. Figures
IV-2 through IV-4 illustrate these grid locations graphically.
8For the purpose of this analysis, we assumed that a ratio
greater than or equal to 0.95 was equivalent to 1.0.
'ibid.
V-37
-------
TABLE V-14
RECEPTOR LOCATIONS WARRANTING FURTHER INVESTIGATION
FOR NONCANCER EFFECTS: POLLUTANT-SPECIFIC (1)
RESULTS INTENDED FOR POLICY DEVELOPMENT ONLY
RECEPTOR
LOCATION
REFINED GRID
CONCENTRATION
TO THRESHOLD
RATIO (2)
POLLUTANT
OP CONCERN
NONCANCER
EFFECT (3)
4342.25
4342.25
4342.25
4342.75
4342.75
4342.75
364.75
367.25
369.75
364.75
367.25
369.75
1.1
1.6
5.0
1.3
1.1
1.4
BENZENE
BENZENE
BENZENE
BENZENE
BENZENE
BENZENE
BLOOD
BLOOD
BLOOD
BLOOD
BLOOD
BLOOD
DISCRETE SITE
4339.85
4340.22
4340.65
4340.85
4341.15
4341.41
4342.00
4342.52
4343.00
4343.15
4350.92
4350.92
4350.92
4350.92
4350.92
4350.92
374.86
374.95
375.10
375.31
375.30
375.30
374.25
374.00
374.12
373.90
370.55
370.55
370.55
370.55
370.55
370.55
1.7
1.6
1.3
1.2
1.2
1.1
1.3
1.2
1.1
1.1
1.5
1.5
6.2
1.5
6.2
1.5
BENZENE
BENZENE
BENZENE
BENZENE
BENZENE
BENZENE
BENZENE
BENZENE
BENZENE
BENZENE
XYLENE
XYLENE
XYLENE
XYLENE
XYLENE
XYLENE
BLOOD
BLOOD
BLOOD
BLOOD
BLOOD
BLOOD
BLOOD
BLOOD
BLOOD
BLOOD
LIVER
KIDNEY
REPRODUCTIVE
NEUROLOGICAL
FETAL/DEVELOPMENT
BLOOD
(1) The noncancer effects analysis did not consider personal exposures
(e.g., indoor and work).
(2) The threshold values underlying these ratio calculations are based
on data of uneven quality. Furthermore, ezceedance of a threshold
value of greater than one does not necessarily indicate severity
of effect; it simply indicates the need to further explore these
exposure levels. For the purpose of this study, it was assumed
that a ratio greater than or equal to 0.95 was equivalent to "1"
through rounding.
(3) The blood effect thresholds for benzene and xylene have not yet
undergone peer review, thus the results are subject to change.
V-38
-------
effects: liver, kidney, reproductive, neurological, fetal, and
blood.
We explored source contribution for the refined grid cell
and the discrete site with the highest concentration-to-threshold
ratios. For the refined grid cell, benzene emissions from Point
Source A account for approximately 94 percent of the ambient
concentrations that appear to pose a concern for blood effects.
This finding is subject to the limitation stated above regarding
the uncertainty (potentially overstated) of the modelled results
for Point Source A. At the discrete receptor site with the
highest concentration-to-threshold ratios, almost all of the
ambient air concentrations of xylene can be attributed to Point
Source E.
(2) Complex Pollutant Mixtures
Using the hazard index we identified additional discrete
receptor locations where exposures to multiple pollutants may
deserve a closer examination because of concern for blood
effects. As shown in Table V-15, there sites are primarily in
the southeast quadrant of the study area, Similar to the
pollutant specific results. The compound principal concern is
benzene.
ii. Results Based on Monitoring Data
We investigated the increased concern for noncancer effects
looking at (1) average area-wide exposures based on the available
monitoring data for the Baltimore area and (2) exposures in the
vicinity of individual monitoring sites from the TEAM study
(Dundalk and Parkville) and the available area-wide Baltimore
monitoring data. We explored whether the measured ambient air
concentration of each pollutant examined exceeded any relevant
no-effect thresholds. We also calculated a hazard index for each
noncancer effect. We should note again that the existence of a
threshold value does not suggest anything about the severity of
effect. Also, the data underlying the threshold values are of
varying quality.
(1) Area-Wide
Using the average area-wide ambient pollutant levels
suggested by the available Baltimore monitoring data (see
Table IV-11), it appears that benzene exposures warrant further
investigation for blood effects in the Baltimore study area.
Because of the limitations in using data from a limited number of
fixed monitoring sites to typify an entire region, further
sampling and analysis are needed to validate these findings. The
hazard index calculation did not indicate the need to further
study additional pollutant exposures.
V-39
-------
TABLE V-15
RECEPTOR LOCATIONS WARRANTING FURTHER INVESTIGATION
FOR NONCAHCER EFFECTS: COMPLEX POLLUTANT MIXTURES (1)
RESULTS INTENDED FOR POLICY DEVELOPMENT ONLY
DISCRETE
RECEPTOR
LOCATION
4342.93
4343.20
4343.45
4343.58
4343.70
4346.27
4346.68
362.98
362.95
362.78
362.55
362.28
367.88
367.77
HAZARD
INDEX (2}
1.0
1.0
0.96
1.0
0.95
PRIMARY POLLUTANT
OF CONCERN
(RATIO VALUE)
BENZENE
BENZENE
BENZENE
BENZENE
BENZENE
BENZENE
BENZENE
(0.92)
(0.92)
(0.94)
(0.93)
(0.88)
(0.93)
(0.87)
NONCANCER
EFFECT
BLOOD
BLOOD
BLOOD
BLOOD
BLOOD
BLOOD
BLOOD
(1) The noncancer effects analysis did not consider personal exposures
(e.g., indoor and work).
(2} The threshold values underlying these hazard indexes are based
on data of uneven quality. Furthermore, exceedance of a threshold
value or a hazard index of greater than one does not necessarily
indicate severity of effect; it simply indicates the need to further
investigate these exposure levels. For the purpose of this study.
it was assumed that a ratio greater than or equal to 0.95 was
equivalent to "1" through rounding. The hazard index represents the
sum of all pollutant-specific ratios with the same systemic effect
at a particular location.
(3) The blood effect threshold for benzene has not yet undergone peer
review, thus the results are subject to change.
V-40
-------
(2) Individual Monitoring Sites
TEAM Data (Dundalk and Parkville). Table V-16 presents the
concentration-to-threshold ratios calculated by pollutant and
noncancer effect, as well as the hazard indexes. These
calculations were made using the average measured pollutant
concentrations from the two fixed TEAM monitoring sites. Note
that we did not include the concentration-to-threshold ratios for
methylene chloride and vinylidene chloride in calculating the
hazard indexes because of our concern regarding the validity of
the monitoring data (see Chapter IV).
Our analysis of individual pollutants indicates that benzene
exposures warrant further investigation at the Dundalk and
Parkville sites; the measured benzene concentrations in these
areas exceed the threshold for blood effects. In addition, if
verified, the vinylidene chloride ambient levels at each site and
methylene chloride at Parkville may require additional analysis
because of an increased concern for liver effects. The hazard
index calculations did not identify additional pollutants
warranting further study.
Available Monitoring Data for the Baltimore Area. Based on
the available monitoring data for the Baltimore area (see
Appendix B for a summary of the average measured concentrations
by monitoring site), benzene exposures—even at the monitoring
site with lowest average measured level—need to be examined more
closely. This finding is supported by the TEAM data and the
results from the dispersion modelling.
c. Limitations
The risk calculations presented in this section build
directly on the estimates of exposure (either measured or
modelled) and dose-response. The underlying uncertainties
associated with the exposure and dose-response calculations are
briefly discussed below.
i. Exposure Limitations Baaed on Dispersion Modelling
Although time constraints have not permitted a thorough
evaluation of model performance using the 1987 TEAM ambient data,
a similar exercise has been completed using the ambient air data
gathered during 1983 to 1984 in Baltimore. The results of this
effort suggest that the model may be systematically underpredict-
ing ambient air risks for some pollutants. However, facility-
specific monitoring performed by the AMA suggests that ambient
levels associated with Point Source A are most likely overstated.
There are four major factors introducing uncertainty into
our estimates of exposure:
V-41
-------
TABLE V-16
BALTIMORE IEHP AIR TOXICS STUDY
NONCANCER RATIO CALCULATIONS FROM EXPOSURE TO TARGET COMPOUNDS
AT THE TWO FIXED TEAM MONITORING SITES (1), (2)
Dundalk
POLLUTANT (3)
REPRO- NEURO-
LIVER KIDNEY DUCTIVE BEHAVIORAL
FETAL BLOOD (4)
Arsenic
Benzene
Cadmium
Carbon tetrachloride
Chlorobenzene
Chloroform
Chromium-VI (5)
mDichlorobenzene
oDichlorobenzene
pDichlorobenzene
Ethyl benzene
Ethylene dibronu.de
Formaldehyde (6)
Lead
Methyl chloroform
Methylene chloride
Perchloroe thy lane
Styrene
Trichloroethylene
Vinylidene chloride (7)
m-Xylene
0.003
0.061
0.000
0.002
0.000
N.Q.
0.009
0.018
0.004
0.017
0.024
0.005
0.162
0.009
0.006
0.015
1.061
0.026
0.005
0.001
0.004
N.Q.
0.009
0.018
0.004
0.017
0.024
0.005
0.049
0.009
0.000
0.026
0.000
0.000
0.042
0.000
0.123
0.024
0,005
0.105
0.061
0.009
N.Q.
0.009
0.018
0.024
0.005
0.004
0.015
0.026
0.093
0.000
0.006
0.042
0.000
0.009
0.018
0.024
0.005
0.162
0.001
0.105
0.000
1.572
0.000
N.Q.
0.138
0.024
0.005
0.006
0.026
HAZARD INDEXES (8)
0.3
0.2
0.3
0.2
0.4
1.8
POLLUTANT (3)
Parkville
REPRO- NEURO-
LIVER KIDNEY DUCTIVE BEHAVIORAL
FETAL BLOOD £4)
Arsenic
Benzene
Cadmium
Carbon tetrachloride
Chlorobenzene
Chloroform
Chromium-VI (5)
mDichlorobenzene
oDichlorobenzene
pDichlorobenzene
Ethyl benzene
Ethylene dibromide
Formaldehyde (6)
Lead
Methyl chloroform
Methylene chloride
Perchloroethylene
Styrene
Trichloroethylene
Vinylidene chloride (7)
m-Xylene
HAZARD INDEXES (9)
0.002
0.408
0.014
N.D.
0.000
0.023
0.015
0.018
0.017
0.006
0.022
0.019
1.219
0.031
0.017
0.000
2.171
0.071
0.6
0.004
0.009
Q.014
N.D.
0.023
0.015
0.018
0.017
0.006
0.022
0.019
0.366
0.031
0.000
0.071
0.2
0.000
0.002
0.014
N.D.
0.000
0.043
0.022
0.019
0.291
0.4
0.408
0.014
N.D.
0.023
0.015
0.018
0.022
0.019
0.000
0.071
0.6
0.127
0.000
0.041
0.014
N.D.
0.000
0.022
0.019
1.219
0.002
0.291
0.5
0.000
2.135
0.014
0.000
0.023
0.015
0.018
0.130
0.022
0.019
0.017
0.071
2.5
N.Q. - Not Quantified in sampling
N.D. - Not detected in sampling
(continued)
V-42
-------
TABLE V-16 (CONTINUED)
NOTES
1. The two fixed ambient monitoring stations were located at
the Parkville Middle School in the Parkville section of
Baltimore County and the Merritt Point Mental Health
Facility at the Dundalk section of Baltimore County.
2. The noncancer ratios are defined as (1) the ratio of the
measured ambient concentration of the pollutant and the
threshold for a specific noncancer health effect category,
and (2) a "hazard index" which is calculated by summing the
pollutant-specific ratios for each noncancer health effect.
3. Risk assessments for metals used the average concentrations
in the PM10 samples.
4. For the purpose of this screening analysis, the health
category of blood effects includes non-specific cellular
effects.
5. We assumed that all of the chromium that was collected was
chromium VI. The risk from exposure to chromium is
therefore overestimated for the purpose of this screening
exercise since the actual percentage of chromium VI in the
sample is likely to be considerably lower.
6. The noncancer health effects of formaldehyde were included
in this category even though the known effects are of a less
specific nature.
7. Vinylidene chloride was detected in the sampling, but there
were analytical problems in quantifying ambient
concentrations. Because vinylidene chloride could be an
important compound, the ratio of the ambient concentrations
to the threshold value is given for the purposes of
comparison; however, it is not included in the hazard index.
Additional analysis is needed to confirm the presence of
vinylidene chloride in the ambient air.
8. The monitoring results for vinylidene chloride at Dundalk
are highly questionable due to analytical problems in
quantifying ambient concentrations. Consequently, it is not
included in the cumulative hazard index. Additional
analysis is needed to confirm the presence of vinylidene
chloride in the ambient air.
9. The monitoring results for methylene chloride and vinylidene
chloride in the ambient air of Parkville are highly
questionable due to analytical problems in quantifying the
amounts in the monitoring samples. Consequently, these
compounds are not included in the cumulative hazard index.
V-43
-------
• First, exposure modelling can be limited by the quality
and coverage of the emissions database. Although the
air toxics inventory that we compiled for this study
represents a significant data collection effort, it is
impossible to account for all sources. For example,
the inventory does not address fugitive emissions which
can be a major contributor to ambient levels. Based on
the model performance evaluation, our modelling
analysis significantly underestimated exposures to all
pollutants, except for trichloroethylene (a more
thorough discussion on this topic can be found in
Chapter IV).
• Second, modelling can be hampered by the lack of
adequate detail on releases and meteorological
conditions at a facility. Simplifying assumptions
concerning release parameters and meteorological
conditions could result in either underestimates or
overestimates (as in the case of Point Source A) of
exposure.
• Third, dispersion modelling cannot adequately address
the products of atmospheric chemical transformation,
which may be the driving force behind the problem of
"urban soup." As scientific knowledge expands/
dispersion modelling may be able to accommodate trans-
formation products more easily. For the moment,
however, existing models are best suited for quantify-
ing ambient levels of relatively stable compounds.
• Fourth, as we begin to add the pollutant-specific
risks, the uncertainties in the risk calculations are
compounded.
ii. Exposure Limitations Based on Monitoring
Many of the limitations of the TEAM monitoring program and
the available monitoring data for the Baltimore area have already
been discussed, but a few of them are worth repeating for the
reader:
• The quality control analysis in the TEAM study
indicated that actual concentrations may be overstated
by 30 percent for the samples collected at Parkville.
However, it does not appear that this overestimate
changes our findings on individual risks and noncancer
effects dramatically: average increased lifetime
individual cancer risks at Parkville could fall from
3.5 x 10~3 to 2.5 x 10"3, and the hazard indexes for
liver and blood (the two noncancer effect categories of
concern) will still remain above one.
v-44
-------
• The available monitoring data for the Baltimore area
are highly uncertain. These data stem from different
monitoring programs conducted at different times.
Also, not all results have undergone extensive peer
review.
• Unlike dispersion modelling, ambient data alone cannot
be used to identify sources and their relative
contributions.
• Additional ambient air monitoring should be performed
to obtain a better baseline estimate of methylene
chloride ambient levels.
• Vinylidene chloride, a possible human carcinogen
(Class C), could be an important factor in the risk
calculations and hence its measured concentrations in
the ambient air need to be confirmed. As a result,
additional analysis is needed to confirm its presence
in the ambient air.
• As we begin to add the pollutant-specific risks, the
uncertainties in the risk calculations are compounded.
iii. Dose-Response Limitations
Chapter. Ill and Appendix A describe in detail the principals
and limitations of quantitative risk assessment. The major
caveats relating to the unit cancer risk factors and no-effect
thresholds for noncancer effects are briefly restated below:
(1) Cancer
• As discussed throughout the report, EPA relies on
models that yield a plausible, upper-bound estimate of
potency rather than a "best guess" estimate. Thus, the
true risks are unlikely to be as high as our risk
estimates suggest, and could be significantly lower.
• The POM unit cancer risk factors have not undergone
peer review and are thus subject to change.
(2) Noncancer
• The no-effect threshold for blood effects (the
threshold most often exceeded) has not undergone peer
review and is, therefore, subject to change.
• Numerous uncertainty factors have been incorporated
into the setting of a no-effect threshold. Thus, the
quality of evidence underlying the threshold values
used in this report vary widely by pollutant.
V-45
-------
• Exceedance of a threshold does not indicate severity of
an effect; it merely indicates the need to further
examine an observed exposure level.
3. RISK RESULT COMPARISONS
To provide perspective on the modelled risk estimates for
the Baltimore study area, we compared our findings on modelled
risk with the results from other lEMPs and studies of urban
areas. Because we used a new approach to estimate the lifetime
individual cancer risks from exposure to the semi-volatile
fraction of the products of incomplete combustion (POMs), we also
compared our POM risk results with the estimated increased
lifetime individual cancer risks from the products of incomplete
combustion summarized in a 1985 EPA study. ° Finally, we
compared modelled risks attributable to air toxics in the
Baltimore area with those associated with drinking water. Each
of these comparisons is detailed below.
a. Comparison with IBMP and Other Urban-Scale Analyses
Table V-17 shows a comparison of the average increased
lifetime individual cancer risks and Table V-18 the maximum
increased lifetime individual cancer risks in Baltimore with the
results of seven other urban air toxic studies. The air toxics
studies developed estimates of average and maximum increased
lifetime individual cancer risks for individual toxic pollutants.
These studies are briefly described below.
The EPA "Six Month Study." The first attempt by EPA, using
a variety of techniques based largely on existing data and on
extrapolations from existing data, to estimate the scope and
character of the air toxics problem nationally. It comprises a
series of studies, including the following: "35 County Study,"
which produced estimates of air toxics risk in 35 U.S. counties;
the "Ambient Air Quality Study," which used available ambient air
data to estimate cancer risk; and the "NESHAPs" study, which
analyzed air toxics risk based on available national emissions
data.
IOU.S. EPA, The Air Toxics Problem in the United States; An
Analysis of Cancer Risks for Selected Pollutants, June 1985 (the
so-called "Six Month Study)."
V-46
-------
TABLE V-17
A COMPARISON OF PREDICTED AVERAGE LIFETIME INDIVIDUAL CANCER RISK
IN BALTIMORE WITH OTHER STUDIES
RESULTS INTENDED FOR POLICY DEVELOPMENT ONLY1-2
POLLUTANTS
[WEIGHT OP
EVIDENCE 1
Arsenic [A]
Benzene [A]
Benzo-a-
pyrene [B2]
Cadmium [Bl]
Carbon Tet. [B2]
Chloroform [B2J
Chromium VI [A]
1.2Dichloro-
ethane [B2]
Ethylene
Dibromide [B2]
Ethylene
Oxide [Bl]
Forma Idehyde(Bl)
Methylene
Chloride [B2]
Perchloro-
ethylene (C)
POMs [N.A.]
Trichloro-
NESHAPS
STUDY
1.4E-06
9.8E-06
4.9E-06
4.2E-06
<7.00E-07
7.7E-06
1.3E-05
8.4-06
1.5E-05
<7.00E-07
<7.00E-07
2.8E-06
35
COUNTY
STUDY
1.4E-05
2.7E-OS
1.4E-06
1.4E-06
2.8E-07
2.1E-07
2.0E-05
2.8E-06
1.4E-06
1.5E-05
9.8E-06
1.1E-05
AMBIENT
AIR
QUALITY
STUDY
1.8E-05
7.6E-OS
1.4E-06
4.2E-06
2.6E-OS
3.2E-OS
7.4E-05
5.8E-05
7.7E-06
7.7E-06
SOUTH
COAST
l.OE-09
3.9E-04
6.7E-06
l.OE-08
7.1E-04
5.0E-08
2.4E-OS
3.0E-06
SANTA
CLARA
IEMP
1.5E-05
2.0E-05
4.0E-06
l.OE-OS
6.0E-08
2.0E-05
2.0E-07
2.0E-06
6.0E-07
2.0E-06
l.OE-07
KANAWHA
IEMP
1.3E-05
2.9E-05
4.3E-06
2.0E-06
1.5E-05
1.4E-04
5.9E-04
2.6E-05
1.5E-06
1.8E-06
PHILA BALT
IEMP STUDY
8.6E-06
1.9E-OS 8.7E-06
9.9E-09
3.8E-06
1.5E-06 3.7E-09
4.6E-06 9.0E-07
2.2E-05
2.6E-06 2.1E-07
1.7E-07
8.4E-08
8.1E-06
8.7E-06
1.7E-06 1.7E-0
8.9E-05
1.3E-06 1.5E-06
ethylene [B2]
(continued)
-------
TABLE V-17 (CONTD.)
NOTES
N.A. - Not available.
1 This study uses conservative estimates of increased cancer risk from ambient (i.e., outdoor) exposure to
establish priorities among pollutants and sources. The risk estimates are calculated using modelled
concentrations and EPA unit cancer risk factors. There is considerable uncertainty in the estimated
concentrations, vhich could either overstate or understate the true concentrations (see Chapter IV). Unit
cancer risk factors combine CAG potency estimates with EPA exposure assumptions. The CAG patency estimate
provide a plausible upper limit to the cancer risk of a ocmpound (see Appendix A); however, the true value of
the risk is unknown and may be ae low as zero. The exposure assumptions are extremely conservative in that
they assume continuous exposure to outdoor air for 70 years. Because of the generally conservative bias in the
information, it is highly unlikely that the true risks would be as high as the estimates, and they could be
considerably lower.
* The unit cancer risk factors used to estimate risk vary among the studies. The reader is directed to each
report for details. Differences in modelling assumptions, which vary across studies, can slightly affect
predictions of risk.
i
•I*
CO
-------
TABLE V-1B
A COMPARISON OF PREDICTED MAXIMUM LIFETIME INDIVIDUAL CANCER RISK
IN BALTIMORE WITH OTHER STUDIES
RESULTS INTENDED FOR POLICY DEVELOPMENT ONLY1'2
POLLUTANTS
[WEIGHT OF
EVIDENCE 1
Arsenic [A]
Benzene [A]
Benzo-a-
pyrene [B2]
Cadmium (Bl]
Carbon Tet. [B2]
Chloroform [B2]
Chromium VI [A]
*. 1,2 Dichloro-
«• ethane [B2]
Ethylene
Dibromide [B2J
Ethylene
Oxide [Bl]
Formaldehyde [Bl]
Methylene
Chloride [B2]
Perchloro-
ethylene [C]
POMs [N.A.]
35
NESHAPS COUNTY
STUDY STUDY
6.5E-03
8.0E-03
7.5E-04
5.8E-04 6.0E-06
3.0E-03
1.6E-01 3.6E-03
2.9E-04
1.6E-04
6.8E-03
6.1E-04
l.OE-05
4.6E-04 l.OE-05
AMBIENT
AIR SANTA
QUALITY CLARA KANAWHA
STUDY IEMP IEMP
4.0E-03
1.5E-04 2.0E-04 9.0E-09
2.5E-05
1.5E-03
1.5E-04 4.0E-04
7.7E-05 2.0E-06 2.3E-03
1.4E-03
6.0E-06
2.0E-04 4.6E-03
4.9E-05
<1.0E-05 5.0E-04
1.9C-05 l.OE-05
BALT
STUDY
6.1E-05
3.4E-OS
3.2E-08
2.6E-06
4.8E-08
4.5E-06
7.5E-04
2.3E-07
1.8E-07
6.8E-07
8.7E-06
4.2E-05
3.8E-06
6.3E-04
Trichloro-
ethylene [B2]
l.OE-04
2.1E-05
2.6E-05
2.0E-05
6.0E-06
1.8E-06
(continued)
-------
TABLE V-18 (CONTD.)
NOTES
N.A. - Not available.
1 This study uses conservative estimates of increased cancer risk from ambient (i.e., outdoor) exposure to
establish priorities among pollutants and sources. The risk estimates are calculated using modelled
concentrations and EPA unit cancer risk factors. There is considerable uncertainty in the estimated
concentrations, which could either overstate or understate the true concentrations (see Chapter IV). Unit
cancer risk factors combine CAG potency estimates with EPA exposure assumptions. The CAG potency estimate
provide a plausible upper limit to the cancer risk of a ocmpound (see Appendix A); however, the true value of
the risk is unknown and may be as low as zero. The exposure assumptions are extremely conservative in that
they assume continuous exposure to outdoor air for 70 years. Because of the generally conservative bias in the
information, it is highly unlikely that the true risks would be as high as the estimates, and they could be
considerably lower.
2 The unit cancer risk factors used to estimate risk vary among the studies. The reader is directed to each
report for details. Differences in modelling assumptions, which vary across studies, can slightly affect
predictions of risk.
m
o
-------
Philadelphia Study flEMPl. Like the Baltimore IEMP project,
the study used dispersion modelling and ambient air and drinking
water monitoring to examine the extent of toxic pollutant
releases in the Philadelphia metropolitan area.
Kanawha Vallev Study HEMP1. A Joint EPA-IEMP/West Virginia
study of chronic exposures to toxic air pollution and other
toxics-related issues in this heavily-industrialized area.'2
Santa Clara Study flEMPK A multimedia toxics study
performed by EPA-IBMP in conjunction with the California Bay Area
Air Quality Management District.13
South Coast Study. A study of the Los Angeles basin
performed by the California South Coast Air Quality Management
District. It involved both ambient air monitoring and dispersion
modelling.14
To compare the pollutant-specific lifetime individual cancer
risks from these preceding studies with the Baltimore IEMP, we
developed estimates of maximum and average increased lifetime
individual cancer risks by pollutant. We identified the average
increased lifetime individual cancer risk by multiplying
population-weighted average exposure data for each pollutant by
its CAG unit cancer risk factors. We determined the maximum
increased lifetime individual cancer risk by identifying the
location with the highest concentration of the particular
pollutant and multiplying this number by its unit cancer risk
factor. This is different from the hotspot exposure analysis
discussed earlier in this chapter, which considered only the
highest lifetime individual cancer risks summed across all
pollutants.
"Source: U.S. EPA, Office of Policy Analysis, Final Report
of the Philadelphia Integrated Environmental Management Project,
December 1986.
"Source: U.S. EPA, Kanawha Vallev. West Virginia Toxics
Screening Study Report. Environmental Services Division (U.S. EPA
Region 3) and the Regulatory Integration Division, Philadelphia,
Pennsylvania, 1987.
"Source: U.S. EPA, Office of Policy Analysis, Santa Clara
Vallev Integrated Environmental Management Protect; Revised
Stage One Report. May 30, 1986.
"Shikiya, D.L. Chung, E. Nelson, R. Rapport, The Magnitude
of Ambient Air Toxics Impacts from Existing Sources in the South
Coast Air Basin: 1987 Air Quality Management Plan Working Paper.
Revision #31, South Coast Air Quality Management District, El
Monte/ California; 1987.
V-51
-------
For those compounds for which comparison data were
available, average increased lifetime individual cancer risks
were generally in the same range as those predicted in the other
urban studies. Estimates of maximum increased lifetime
individual cancer risks for individual pollutants, however,
tended to be generally somewhat lower.
Caution must be employed when comparing estimates of maximum
and average increased lifetime individual cancer risk across the
various studies. Differences in modelling assumptions regarding
release specifications and receptor placement can significantly
affect predictions of lifetime individual cancer risk.
b. Comparison of Alternative Approaches to Estimating
Risks from the Products of Incomplete Combustion
In the EPA "Six Month Study," ambient concentrations of PICa
were calculated using the ambient levels of benzo(a)pyrene
(B(a)P) as surrogates for the levels of the semi-volatile or
polycyclic organic (POM) component of the products of incomplete
combustion (PIC). The study quantified these risks using a unit
cancer risk factor of 0.42 (ug/m3)'1 derived from a review of
existing epidemiological data. Although there were numerous
limitations with this approach, which are discussed in the "Six
Month Study," it was the only available tool for exploring PIC
risks at the time the study was completed. In Table V-19 we
present the estimated annual excess cancer incidence for the
study area and the estimated increased lifetime individual cancer
risks at each of the three hotspot locations using the two
approaches.
The B(a)P surrogate approach results in over three times
greater estimates of annual excess cancer incidence than the POM
approach used in this study. On the other hand, the surrogate
approach gives much lower estimates of estimated increased
lifetime individual cancer risks for the three hotspot sites.
The reason for this seeming contradiction is that the B(a)P
surrogate approach uses only one unit cancer risk factor for all
emissions of particulates that result from incomplete combustion
of organic matter. In contrast, the unit risk factors we used
for the semi-volatile fraction of the product of incomplete
combustion in the Baltimore IEMP are specific to types of sources
of combustion products. As shown in Table IV-5, EPA's ORD has
found that the relative carcinogenic potency of POM from
different source types can differ by over five orders of
magnitude (x 100,000). Thus, the risk estimates depend heavily
on the type of combustion sources located in the study area.
Additional analytical work and peer review of both approaches are
required before the results of one approach over the other can be
supported.
V-52
-------
TABLE V-19
COMPARISON OF THE RESULTS OF ALTERNATIVE APPROACHES
TO ESTIMATING THE RISKS FROM THE SEMI-VOLATILE
COMPONENT OF THE PRODUCTS OF INCOMPLETE COMBUSTION1
RESULTS INTENDED FOR POLICY DEVELOPMENT ONLY
POM B(a)P SURROGATE
APPROACH APPROACH
Annual Excess Cancer
Incidence 2.05 6.8
Estimated Lifetime
Individual Cancer
Risk at Hotspot 3.7 x 10'4 7.9 x 10°
Site 1
Estimated Lifetime
Individual Cancer
Risk at Hotspot 6.3 x 10"4 1.4 x 10"
Site 2
Estimated Lifetime
Individual Cancer
Risk at Hotspot 6.3 x 10"4 1.3 x 10
Site 3
1 This study uses conservative estimates of increased cancer
risk from ambient (i.e., outdoor) exposure to establish
priorities among pollutants and sources. The risk estimates
are calculated using modelled or monitored concentrations and
EPA unit cancer risk factors. There is considerable
uncertainty in the estimated concentrations, which could
either overstate or understate the true concentrations (see
Chapter IV). Unit cancer risk factors combine CAG potency
estimates with EPA exposure assumptions. The CAG potency
estimates provide a plausible upper limit to the cancer risk
of a compound (see Appendix A); however, the true value of the
risk is unknown and may be as low as zero. The exposure
assumptions are extremely conservative in that they assume
continuous exposure to outdoor air for 70 years. Because of
the generally conservative bias in the information, it is
highly unlikely that the true risks would be as high as the
estimates/ and they could be considerably lower.
V-53
-------
c. Comparison of Risks in Air and Drinking Water
In Phase I of the Baltimore IBMP, the potential cancer risks
from ingestion of drinking water were examined. In particular,
the analysis looked at trihalomethanes (THMs). The Phase I risk
assessment was performed using average ambient data for finished
drinking water at the Ashburton and Montebello drinking water
treatment plants. Only exposures from ingestion of drinking
water were modelled; there was no consideration of alternative
exposure pathways, such as inhalation of volatilized compounds
from showering.
The Phase I results suggested that the cancer risks from
chloroform, a THM, are roughly comparable to those from air
toxics» the annual excess cancer incidence was estimated at
approximately three cases; average increased lifetime individual
cancer risks were estimated at approximately 1 x 10" . The Phase
II air toxics results indicated that the estimated annual excess
cancer incidence is somewhat higher, but the average increased
lifetime individual cancer risks are generally in the same order
of magnitude.
V-54
-------
VI. RISK MANAGEMENT STRATEGIES FOR REDUCING RISKS!
A DEMONSTRATION
This chapter demonstrates how two analytical techniques—
cost-effectiveness and benefit-cost analysis—can be used to
identify control priorities in the risk management phase of the
IBMP. In conducting these analyses, we identified (1) control
options, (2) rough estimates of control costs, (3) the
effectiveness of the control options in reducing risks to human
health, and (4) the dollar benefits afforded by each control
option in terms of reduced mortality, morbidity, and adverse
welfare effects (e.g., materials damage, visibility, and
agricultural effects).
In the cost-effectiveness analysis, we ranked control
options and identified the set of control options that would
reduce risk to a specified level at the least cost. In the
benefit-cost analysis, we identified the control strategy—that
is, the set of control options—that would lead to the greatest
monetary benefit to society at a given level of cost. We
examined these two approaches because the risk manager may decide
to implement a specified control strategy on the basis of various
factors, including baseline risks, cost-effectiveness of the
remedial program, and the economic benefits to society of the
control plan.
The results of these two studies can be used to shed light
on policy questions relating to risk management that were raised
in Chapter II of this document, namely:
What are the relative coats and effectiveness of alternative
control options and control option strategies to reduce
human health risk from air toxics in the Baltimore area?
Will the control of carcinogenic emissions lead to reduced
concern for noncarcinogenic health effects from air toxics?
Will the control of air toxic emissions also result in lower
emissions of criteria air pollutants, in particular ozone
precursors, such as volatile organic compounds (VOCs), and
particulatea?
VI-l
-------
What are the health and welfare benefits that will accrue
from reducing emissions of VOCa and participates as part of
controls to lower air toxics emissions at specific sources?
We emphasize that the results presented in this chapter are
highly uncertain and are included primarily to illustrate
different approaches for establishing risk management priorities.
This chapter is divided into four sections. The first
describes the process by which sources in the Baltimore area were
selected for inclusion in our risk management demonstration
project. The second section describes the control options
identified for the screened sources based on a preliminary
engineering analysis. It also presents rough estimates of
control costs and control effectiveness from the standpoint of
reducing human health risk and increasing dollar benefits to
society. The third and fourth sections describe the methodology
and results for the cost-effectiveness and benefit-cost analyses,
respectively.
1. SELECTION OF SOURCES
The first step in conducting either a cost-effectiveness or
benefit-cost analysis is to select the sources for which control
options will be evaluated. An analysis of feasible control
options can be very time consuming/ even at a preliminary level.
Because of the large number of sources and pollutants in the
study area, we focused the analysis on the sources and pollutants
that were likely to contribute most to cancer risk.
The Ambient Air Toxics Workgroup established criteria to
identify the point and area sources that were included in the
analysis. A point or area source needed to meet only one of the
screening criteria to be included in the analysis:
(1) Contributes at least 1 percent (through rounding) of
the estimated excess cancer incidence in the study area
(2) Projected to pose an average increased lifetime
individual cancer risk of 5 x 10 or greater (through
rounding) for the target pollutants considered
(3) Contributes at least 1 percent (through rounding) to
the incidence at the refined grid cell of highest
incidence.
The value for the increased lifetime individual cancer risk
criterion was chosen to ensure—given the uncertainty of the risk
estimates—that point sources with emissions of carcinogens that
result in risks in excess of 1 x 10"5 (a common de minimis risk
VI-2
-------
level used by various states and EPA programs) would be included
in the analysis. We also decided to use the 1 percent criterion
to focus the control analysis on those sources that contribute
most to cancer risk. Although we did not explicitly consider
noncancer effects in selecting sources, we did evaluate the
impact of controls to reduce cancer risks on exposures that lead
to noncancer effects.
We did not arbitrarily choose these criteria. Our review of
the data in Chapter V suggested that these criteria represented a
clear break in the importance of sources and pollutants in terms
of overall cancer risk from ambient air toxics.
we show a comparison of the selected sources and pollutants
with regard to their estimated annual excess cancer incidence in
Table VI-1. The area source categories discussed in Chapter V
are broken out in greater detail here for the purpose of
identifying control options. Each of these subcategories also
had to meet at least one of the selection criteria to be
included. The category of road vehicles comprises four different
types of road vehicles, Solvent usage is represented by the
major subcategories of solvent use: degreasing, dry cleaning,
and the catchall category for methylene chloride use,
miscellaneous industrial methylene chloride use. Heating is
characterized by residential woodstoves, commercial oil.
combustion, residential oil combustion, and residential coal
combustion. We also chose to include gasoline marketing (that
is, gasoline stations, truck stops, and other commercial
operations involving the sale of gasoline). This was done
because the AMA had already estimated control costs and VOC
removal efficiencies.
The sources chosen on the basis of maximum increased
lifetime individual cancer risk are listed in Table VI-2. Four
of the six point sources, were also selected because they
contribute approximately 1 percent or more to the total estimated
excess cancer incidence.
The final list of 15 sources and their pollutants for which
controls were identified is presented in Table vi-3. Constraints
on time and resources prevented us from analyzing control options
for heavy and light duty gasoline road vehicles, Point Source E,
and Point Source F. However, because this was a demonstration
project, it was unnecessary to include all sources.
Although the control-options analysis focused on the removal
of toxics, we estimated co-control of other study pollutants as
well. For example, we assumed that controls to reduce benzene
emissions at Point Source A, would also reduce ambient air
releases of toluene and xylene. [Only a detailed engineering
examination of the facility could ensure that the control options
for one air toxic would indeed control other discharges at the
VI-3
-------
Table VI-1
BALTIMORE IBMP AIR TOXICS STUDY
ESTIMATES OF EXCESS ANNUAL CANCER INCIDENCE ,
FOR SOURCES SELECTED FOR ANALYSIS OF CONTROL OPTIONS1
Sources
POINT SOURCES
Point Source A
POM2--coke ovens
Chromium- 6
Benzene
Arsenic
Cadmium
Point Source B
Chromiura-6
Arsenic
Cadmium
Point Source C
Chromium- 6
Arsenic
Point Source D
Chromium- 6
AREA SOURCES
Road vehicles:
Light duty gasoline
POM
Formaldehyde
Benzene
EDB
EDC
Cadmium
Heavy duty gasoline
POM
Formaldehyde
Benzene
EDB
EDC
Current
Annual
Emissions
t kkor/vr 1
863 63
0.20
0.10
0.10
0.28
0.07
0.19
TOTAL3:
vehicles
34.63
!3I:?8
3:18
0.08
vehicles
10!60
3?:SS
1.32
Light duty diesel vehicles
POM— light duty diesel 24.44
Formaldehyde 32.50
Control
Estimated Annual
Excess Cancer
Incidence
rweiaht of Evidence 1
1.18
0.90
8:i§
8.00
.00
N.A.]
'A] J
AJ
'AJ
•Bl]
0.21
0.17
8:8?
A]
A]
si]
0.12
0.11
0.01
A]
A]
0.05
0.05 [A]
1.56
0.72
0.52
Q»o|
0.10
o!oo
N.A.]
Bl] J
A!
B21
B2
Bl
0.35
8:33
0.01
0.00
0.00
[N.A.I
Bl]
Al
B2]
B2
0.12
8:4!
N.A.]
Bl]
Percentage
of Total
Cases
32.59
24.82
4.98
2.77
0.01
0.01
5.77
4.58
0.83
0.36
3.44
3.16
0.28
1.30
1.30
43.10
19.85
14.47
2-S2
0.06
0.05
9.67
l-Q
0.09
o!o2
3.39
3.04
0.35
(continued)
VI-4
-------
Table VI-1 (continued)
BALTIMORE IEMP AIR TOXICS STUDY
ESTIMATES OF EXCESS ANNUAL CANCER INCIDENCE
FOR SOURCES SELECTED FOR ANALYSIS OF CONTROL OPTIONS
Annual
Emissions
Sources t kka/vr >
Heavy duty diesel vehicles
POM— heavy duty diesel 59.31
Forma Idehyde 9.60
Solvent usage
" Methylene chloride 172.00
Percnloroethylene 315.00
Trichloroethylene 660.00
-Drycleaning
Perchloroethylene 1 , 825 . 00
-Miscellaneous industrial
Methylene chloride use 753.00
Heating
-Woodstoves
POM 112.00
-Commercial oil combustion
POM— distilled oil 12.60
POM-com. residual oil 14.00
Forma Idehyde 0.60
Chromium- o 0 . 0 0 1
Cadmium 0.51
Arsenic 0.34
-Residential oil combustion
POM— distilled oil 83.60
Formaldehyde 7.80
Chromium-5 0.003
Cadmium 0.33
Arsenic 0.05
-Residential coal
POM 133.30
Cadmium 0.01
Arsenic 0.24
-Gasoline marketing
Benzene 25.20
EDC 2.10
EDB 0.21
Estimated Annual
Excess Cancer
Incidence
fWeioht of Evidence!
0.02
0.01
0.00
Ul
0.01
0.03
O.*03
[N.A. ]
1B1]
ci]
B2]
[C]
0.12
0.12 [B2]
0!03 [N.A.]
J:JJ
oioo
o'.os
8. 10
.07
0.00
ul
0.09
0.05
0.00
0.04
8:81
8:88
N.A. ]
N.A.]
sa]
1],
[N.A.]
|j j
[N.A.]
I"1
I®}
;B2J
Percentage
of Total
Cases
0.42
0.32
0.10
1.76
0.75
3.27
3.27
0.91
0.91
CL36
0.01
0.01
0.83
1.42
2.92
2.06
0.10
0.03
2.40
1.38
0.01
1.01
8:8
8:81
TOTAL RISK SUBJECT TO
CONTROL ANALYSIS:
3.3 annual cases
1.76 48.57%
91.7% of total risk
VI-5
-------
TABLE VI-1 (CONTD.)
NOTES
N.A. = Not Available
1 This study uses conservative estimates of increased cancer
risk from ambient (i.e., outdoor) exposure to establish
priorities among pollutants and sources. The risk estimates
are calculated using modelled or monitored concentrations and
EPA unit cancer risk factors. There is considerable
uncertainty in the estimated concentrations, which could
either overstate or understate the true concentrations (see
Chapter IV). Unit cancer risk factors combine GAG potency
estimates with EPA exposure assumptions. The CAG potency
estimates provide a plausible upper limit to the cancer risk
of a compound (see Appendix A); however/ the true value of the
risk is unknown and may be as low as zero. The exposure
assumptions are extremely conservative in that they assume
continuous exposure to outdoor air for 70 years. Because of
the generally conservative bias in the information, it is
highly unlikely that the true risks would be as high as the
estimates, and they could be considerably lower.
2 POM - Polycyclic organic matter produced through incomplete
combustion of fossil fuels. With the exception of coke oven
emissions, POM unit risk factors used in the assessment of
risks have not undergone extensive EPA review and thus are
subject to change.
3 The numbers shown may not sum to the totals because of
rounding.
VI-6
-------
TABLE VI-2
SOURCES POSING GREATER THAN 5 X 10~6
INCREASED LIFETIME INDIVIDUAL CANCER RISK1
Risk Estimates Are for Policy Development Only
INCREASED LIFETIME INDIVIDUAL
SOURCE CANCER RISK
Point Source A 1.1 x 10~3
Point Source B 3.8 x 10"J
Point Source C 9.9 x 10~^
Road vehicles 7.1 x 10"^
Point Source E 3.8 x 10";?
Heating 2.4 x 10"|
Solvent usage 1.4 x 10~£
Point Source D 1.2 x 10~-|
Point Source F 6.3 x 10~6
This study uses conservative estimates of increased cancer
risk from ambient (i.e., outdoor) exposure to establish
priorities among pollutants and sources. The risk estimates
are calculated using modelled or monitored concentrations and
EPA unit cancer risk factors. There is considerable
uncertainty in the estimated concentrations, which could
either overstate or understate the true concentrations (see
Chapter IV). Unit cancer risk factors combine CAG potency
estimates with EPA exposure assumptions. The CAG potency
estimates provide a plausible upper limit to the cancer risk
of a compound (see Appendix A); however, the true value of the
risk is unknown and may be as low as zero. The exposure
assumptions are extremely conservative in that they assume
continuous exposure to outdoor air for 70 years. Because of
the generally conservative bias in the information, it is
highly unlikely that the true risks would be as high as the
estimates, and they could be considerably lower.
VI-7
-------
TABLE VI-3 SOURCES & POLLUTANTS INCLUDED IN
THE CONTROL ANALYSIS
Source
Point Sources
Point Source A
Point Source B
Point Source C
Point Source 0
Area Sources
Degreasing (solvent usage)
Dry Cleaning (solvent usage)
Other (misc.) Solvent Usage
Light Duty Diesel Road Vehicles
Heavy Duty Diesel Road Vehicles
Commercial Oil Heating
Residential Oil Heating
Residential Coal Heating
Residential Wood Stoves
Gasoline Marketing*
Pollutants
poiycyctic organic matter, hexavalent chromium,
benzene, arsenic, cadmium, toluene, xylerte
hexavalent chromium, nickel, cadmium, arsenic
hexavalent chromium, nickel, arsenic
hexavalent chromium
tnchloroethylene. methylene chloride.
perchloroethylene
perchloroethylene
methylene chloride
polycyclic organic matter, cadmium, formaldehyde
polycyclic organic matter, cadmium, formaldehyde
hexavalent chromium, nickel, formaldehyde
toluene, benzene, polycychc organic matter
hexavalent chromium, nickel, formaldehyde
toluene, benzene, poiycyctic organic matter
arsenic, cadmium, hexavalent chromium, nickel.
formaldehyde, toluene, benzene, polycydie organic
matter
polycyclic organic matter, arsenic, cadmium,
hexavalent chromium, formaldehyde,
benzene, toluene, nickel, phenol
benzene, EDC, EDa xylene. toluene, ethyl benzene
Gasoline Marketing was included in the analysis despite the fact that the source
contributed less than 1% of the modeled incidence in the study ares.
VI-8
-------
same level of effectiveness.] In addition, we generated data on
the co-removal of VOCs and TSFs which were used in the benefit-
cost analysis, as described in Section 2 below.
2. ESTIMATING CONTROL OPTIONS COSTS. EFFECTIVENESS. AND
BENEFITS
a. Feasible Control Options and their Cost-Effectiveness
We analyzed one or more potentially feasible air pollution
controls for each of the key point and area sources using
engineering estimates. We then approximated capital and annual
operating costs and the effectiveness in pollutant removal for
each control. For the cost-effectiveness analysis, the pollutant
removal efficiencies were specific to individual toxic
constituents to permit calculation of the reductions in risk
after control. For the benefit-cost analysis, the removal
efficiencies were identified for the pollutant categories of VOCs
and particulates to permit calculation of benefits.
Actual costs and removal efficiencies for a control at a
particular site may differ from these engineering estimates.
Thus, a detailed site-specific analysis is required to confirm
whether an identified opti- is feasible. We do not claim that
all chosen control options 3 practical in consideration of the
cost of reducing emissions, -it only that each option is probably
technically feasible.
We based the pollution sntrol analysis on literature
review, engineering judgmer: discussions with AHA staff, and
telephone conversations wit:, pollution control equipment vendors.
Alternative control technologies were grouped into one or more
appropriate control "options." In some cases, a control option
is a single removal technology, while in other cases an option
represents a series of controls on a single emission point or
separate controls on multiple release points within a given
facility. Controls for point sources were identified by first
consulting with AMA staff responsible for monitoring permit
compliance of the respective facilities. Each State inspector
familiar with a facility was asked to identify control options
that could be applied to reduce toxic emissions.
Tables VI-4 and VI-5 summarize the options for point and
area sources respectively. Costs and effectiveness of controls
in reducing emissions were calculated on the basis of the
literature and estimates provided by vendors. We computed
capital and annual (operating and maintenance) costs in May 1987
dollars. For Point Source A, we grouped controls by the major
location at the facility where the emissions occur. We present a
more detailed description of baseline emissions, the
effectiveness of each control option in reducing emissions from
VI-9
-------
TABLE VI-4 DESCRIPTION of CONTROL OPTIONS for POINT SOURCES
Source
Point Source A
Plant Area
A&B Plants
Benzene/Lit ol
Plant
Storage Tanks
Mills
Release Pt.
Tar decanter, tar intercepting
sump, flushing liquor
circulation tank
Tar storage tanks, tar
dewatering tanks
Excess ammonia liquor
storage tanks
Light oil condensor
Light oil sump
Benzene condensor
Light oil benzene mixture
storage tanks
Benzene storage tanks
Chrome plating line
Option/Alternative
Current Control
Alternative 1
Current Control
Alternative 2
Alternative 3
Current Control
Alternative 4
Alternative 5
Current Control
Alternative 6
Current Control
Alternative 7
Current Control
Alternative 8
Current Control
Alternative 9
Alternative 10
Current Control
Alternative 1 1
Alternative 12
Current control
Alternative 13
Option Description
Uncontrolled
Coke oven gas blanketing system
Uncontrolled
Wash Oil Scrubber
Coke oven gas blanketing system
Uncontrolled
Wash Oil Scrubber
Coke oven gas blanketing system
Uncontrolled
Coke oven gas blanketing system
Uncontrolled
Cover
Uncontrolled
Coke oven gas blanketing system
Uncontrolled
Wash Oil Scrubber
Coke oven gas blanketing system
Uncontrolled
Wash Oil Scrubber
Coke oven gas blanketing system
Packed bed wet scrubber
Packed bed wet scrubber
Efficiency
97
90
97.9
90
97.7
97.9
98.2
97.9
89.9
98.2
90
97.9
50
I
»-•
O
-------
TABLE VI-4 DESCRIPTION of CONTROL OPTIONS lor POINT SOURCES (continued)
Source
Point Source A
[continued)
Plant Area
Coke Ovens
Release Pt.
11 and 12 Batteries
A Battery
1
Option/Alternative
Current Control
Option 1
Option 2
Current Control
Option Description
Meets proposed NESHAP limitations for
coke oven doors, i.e.. 10% leaking
doors, lor 75% control. Charges are
currently every 32 seconds lor 97%
control. Four percent of lids leak.
Construct two new batteries to replace
the existing batteries. The new
batteries wouild be designed to meet
the propuosed NESHAP.
Modify charging to 16 second intervals.
lor 98% control. Unit leaking lids to 3%
ol topside ports through maintenance
and inspection
Meets proposed NESHAP
Efficiency
1.3
13
—
M
I
-------
TABLE VI-4 DESCRIPTION of CONTROL OPTIONS for POINT SOURCES (continued)
Source
Point Source B
Point Source C
Point Source D
Plant Area
Release Pt.
Electric arc furnace (E AF)
Argon-oxygen decarbunzation
vessel (ADD)
Electric arc furnace (EAF)
Argon-oxygen decarbunzation
vessel (AOO)
EAF and AOO
Process Line
Option/Alternative
Current Control
Option 1
Current Control
Option 2
Current Control
Option 1
Current Control
Option 2
Option 3
Current Control
Option 1
Option Description
Direct evacuation control, segmented
canopy hood
Scavenger ducts, cross draft partitions,
closed roof
Close Filling Hoods
Segmented canopy hood, scavenger
ducts, cross draft partitions, and
closed roof
Segmented canopy hood, closed roof
Direct evacuation control, scavenger
ducts, and cross draft partitions
Close Fitting Hoods
Segmented canopy hood, scavenger
ducts, and cross draft partitions
Vinyl strip doors to block wind
circulation
Horizontal packed bed wet scrubber,
venturi scrubber
Second packed bed wet scrubber
Efficiency
99%
99%
99%
99%
50%
40%
I
!-•
N>
* Control alternatives for the A&B Plant, Benzene/Lrtol Plant, and Storage Tanks were grouped into a series of Options as shown in Table VI-6
The selection of control options and the estimation ol capital and annual costs and removal efficiencies are projections and estimations based
on discussions with AMA staff, literature review and engineering judgement. Actual costs as well as feasibility and effectiveness could differ
considerably from our estimates.
-------
TABLE VI-S DESCRIPTION of CONTROL OPTIONS lor AREA SOURCES
Area Source
Current
Control
Option 1
Option 2
Option 3
DRY CLEANING
None
Quickly identify and repair leaks
Regenerative filter wastes not to
exceed 25kg of solvent per 100 kg
of wet waste material
Distillation wastes not to exceed
60 kg of solvent per 100 kg of wet
waste materials
Better housekeeping
Overall control efficiency = 20%
Quickly identify and repair leaks
Regenerative filler wastes not to
exceed 25kg of solvent per 100 kg
of wet waste material
Distillation wastes not lo exceed
60 kg of solvent per 100 kg of wet
waste materials
Belter housekeeping
Carbon adsorption systems for all
commercial dry cleaners
Overall control efficiency = 58%
I
t-«
Ul
Quickly identify and repair leaks
Regenerative filter wastes not lo
exceed 25kg of solvent per 100 kg
of wet waste material
Distillation wastes not to exceed
60 kg of solvent per 10O kg of wet
waste materials
Belter housekeeping
Carbon adsorption systems tor all
coin-op and commercial dry cleaners
Overall control efficiency = 65%
DECREASING
Nona
Cold Cleaners
Cover during idle time (control
efficiency of 90%)
Dram racks with 30 second dram
and control efficiecy of 50%
during operation of degreaser
Open Top Vapor Degieasers
Cover during ide time (control
efficiency of 90%)
Increase freeboard ratio lo 0 75
dunng operation (control
efficiency of 27%)
Cold Cleaners
Cover dunng idle time (control
efficiency of 90%)
Dram racks with 30 second drain
and control efficiecy of 50%
ddurmg operation of degreaser
Open Top Vapor Degreasers
Cover during ide time (control
efficiency of 90%)
Use refrigerated freeboard device
(control efficiency of 60% for
vaporization losses and 29% of
earn/out losses)
Cold Cleaners
Cover during idle lime (control
efficiency of 90%)
Drain racks with 30 second dram
and control efficiecy of 50%
ddurmg operation of degreaser
Open Top Vapor Degreasers
Cover during ide time (control
efficiency of 90%)
Use carbon adsorption unit
(control efficiency of 70% for
vaporization losses and 3O% of
carryoul losses)
-------
TABLE VI-S DESCRIPTION of CONTROL OPTIONS for AREA SOURCES (continued)
Area Source
OTHER INDUSTRIAL
GAS MARKETING
ROAD VEHICLES
Light Duty Gas
Road Vehicles
Light Duty Diesel
Road Vehicles
Heavy Duty Diesel
Road Vehicles
Current
Control
None
Stage 1
Annual Inspection
None
None
Option 1
Enlarged condensation zone
Waste recovery facility
Manual enclosure
Overall control efficiency 21%
Stage II (vapor balance system at pump
Overall control efficiency 70 5%
Cease Annual Inspection
Catalytic Convenor
Catalytic Convenor
Option 2
Enlarged condensation zone
Waste recovery facility
Automatic enclosure
Overall control efficiency 32 8%
)
Option 3
Enlarged condensation zone
Waste recovery facility
Refrigerated condenser
Overall control efficiency 36 6%
M
I
-------
TABLE VI-S DESCRIPTION of CONTROL OPTIONS for AREA SOURCES (continued)
Area Source
HEATING
Residential
Wood Stoves
Commercial Oil
Heating
Residential Oil
Heating
Residential Coal
Heating
Current
Control
None
None
None
None*
Option 1
Stop using wood stoves Instead
use mam fossil fuel heating system
(gas.coal or oil)
Replace oil-lired furnaces with
natural gas systems
Replace oil-firedfurnaces with
natural gas sytems
Replace coal-fired units with
natural gas systems
Option 2
Replace existing burner with dual
burner and bum natural gas
Option 3
U»
^ • The burning of coal is illegal, but does occur
The selection of control options and the estimation oi capital and annual costs and removal efficiencies are projections and estimations based
on discussions with AMA staff, literature review and engineering judgement Actual costs as well as feasibility and effectiveness could differ
considerably from our estimates
-------
the baseline, and capital and annual costs for each option for
each of the four point sources in Appendix C.1
We calculated the costs and removal efficiencies of controls
for area sources for the total study area, rather than the
individual facility. Because of the large number of small
facilities making up each area source category, we assumed that
the frequency of facilities within a source category per unit
population (e.g., dry cleaners) was the same for Baltimore as for
the nation as a whole. We then identified existing and potential
air pollution controls and estimated removal efficiencies and
associated capital and annual costs on the basis of the
following: EPA publications, such publications as background
documents for proposed and promulgated standards and control
technology assessment documents, and national data on the
distributions of sources of different sizes. A summary of the
estimated costs and emission reduction associated with the
selected control options for each area source can be found in
Appendix C.
The control option costs for both point and area sources do
not take into account the age or usefulness of current equipment
that must be replaced under the option. For example, a
forty-year old coke oven at Point Source A that has already
outlived its useful life would/ under Control Option #1, be
replaced by a new oven that meets current federal emission
standards. To the extent that Point Source A would make the
investment in new equipment regardless of the control option or
the control only fractionally reduces the useful life of existing
equipment, we overestimate the actual costs. Furthermore, we do
not take into account the resale or income from replaced
equipment or scrap.
Similarly, both commercial facilities and private residences
would, under the respective control options, be expected to
switch from coal and oil combustion to natural gas for heating
purposes. Yet a certain, probably high, percentage of both
commercial facilities and private residences—given the age of
the property in the study area—have old equipment that must be
replaced regardless of the needs of air quality. The costs to
these residences of purchasing natural gas-burning equipment
rather than oil or coal is substantially less than the costs we
estimate.
Thus, our estimates of costs, particularly for the more
expensive options, may be greatly exaggerated. We could not
*For more information, see: Versar Inc., Baltimore Phase II
Air Toxics Control Options Analysis, prepared for U.S. EPA,
Office of Policy Analysis, EPA Contract No. 68-01-7053, November
6, 1987.
VI-16
-------
generate more realistic estimates without a much more thorough
analysis. Yet these cost estimates should suffice for a general
comparison of options and strategies for control and the setting
of priorities in this demonstration project. They are not
adequate for an accurate and realistic determination of the
financial burden of specific controls.
Table VI-6 presents a summary of the total annualized costs,
the annual reduction in the estimated annual excess cancer
incidence for each control option, the annualized cost per case
reduced, and the percent reduction in cases expected.
Annualization is the process of spreading the present value of an
expenditure over time considering an appropriate discount rate.
In this analysis, the capital costs of the control equipment were
annualized over the estimated life of the equipment using a 10
percent discount rate. Total annualized costs were estimated by
adding the annualized capital costs to the annual operating
costs.
The greatest annual reduction in excess cancer incidence
comes with control option 3 on Point Source B. The lowest
annualized cost per cancer case reduced—a cost savings—is
roughly $200,000 for control option 2 for reclaiming
trichloroethylene (TCE) from degreasing, which would lead to an
estimated reduction in the estimated annual excess cancer
incidence of 0.014 (an annualized cost per case of approximately
$13 million). At the opposite end of the spectrum, controls on
the use of oil for residential heating would cost about $150
million annually and would lead to only about 0.045 fewer cases
per year—an annualized cost per case of $3.3 billion. What
should be clear from Table VI-6 is that no one control option
alone is likely to suffice in reducing total cancer incidence,
should the risk manager deem current predictions of excess cancer
incidence to be unnecessarily high.
Table VI-6 is an extremely important table for the risk
manager. By showing the annual cost per case of each option
evaluated, the risk manager can assess the cost-effectiveness of
implementing controls at individual facilities, as well as at a
combination of facilities. By ordering the cost-effectiveness
ratios for each option from highest to lowest, the risk manager
can develop a strategy for achieving increasing levels of risk
reduction at least cost. In a more complicated analysis that,
for example, also considered intermedia transfers resulting from
pollution control, it would not be as easy to identify the
optimal control strategy. Thus, computer modelling is often
utilized to assist with such analyses. In this demonstration
project, the latter approach was utilized and is discussed in
Section 3 below.
VI-17
-------
SOURCE
HD DIESEL VEHICLES
LO DIESEL VEHICLES
POINT SOURCE A-COKE OVENS
POINT SOURCE A
POINT SOURCE A-BENZENE
POINT SOURCE B
POINT SOURCE C
POINT SOURCE D
GASOLINE MARKETING
OEGREASING-PERCHLOROETHYLENE
DEGREAS1NG-TRICHLORQETHYLEME
OEGREASING-HETHYL. CHLORIDE
ORYCLEANING
MISC. IND.-METHYL. CHLORIDE
HOOOSTOVES
COMMERCIAL OIL COMBUSTION
RESIDENTIAL COAL COMBUSTION
RESIDENTIAL OIL COMBUSITON
CONTROL
OPTION
1
1
1
2
1
1
2
3
4
5
1
2
3
1
2
3
4
1
1
I
2
3
1
2
3
1
1
2
3
1
2
3
1
1
2
1
1
CAPITAL
COST
$5.941.000
S956.320
1238,000.000
$433.000
{460.000
S770.000
WO. 000
$200.000
1150,000
$210,000
$70,000
1240,000
$310.000
$120,000
$160.000
$280,000
$5,000
$10,000
$8,300,000
$60.000
$140.000
$390.000
$120.000
$330,000
$900.000
$6,000
$0
$2,750.000
$4.470,000
$390.000
$830,000.
$1,210,000
$0
$163.110.000
$20.040,000
$18.100.000
$699,270,000
TABLE VI - 6
CONTROL OPTIONS & ANNUALIZED COSTS (1.2)
.CAPITAL ANNUAL ANNUALIZED ANN. CASE ANNUAL COST \ REDCTN
LIFE OPERATING COST COST REDUCTN PER CASE IN CASES
10
10
25
25
15
15
15
15
15
15
10
10
10
10
10
10
10
15
10
15
15
15
15
15
15
15
10
10
15
15
15
35
35
35
35
$0
$0
$243,000
$243.000
$200.000
$158.000
$64.000
$4,000
$54,000
$60.000
$20.000
$60.000
$80.000
$30.000
$40.000
$70,000
$200
$60,000
$970,000
($50.000)
($80.000)
$70,000
($130.000)
($220.000)
($220.000)
($50.000)
$110,000
$170.000
$110,000
($120.000)
($130,000)
($130,000)
$220,000
$8,920.000
$8,920,000
$1.300.000
$77.700.000
$966,600
$155,593
$26,423,000
$290.630
$259.800
$259.255
$117.904
$30.290
$73.721
$87,609
$31.392
$99,059
$130,450
$49.529
$66,039
$115,569
$1.014
$61,015
$2,320,0*00
($42.112)
($61.594)
$121.274
($114,223)
($176.614)
($101,674)
($49.211)
$110.000
$277.549
$837,449
($68,726)
($20.877)
$29,082
$220,000
$25.830,000
$10,998,000
$3,180.000
$150.000,000
0.004
0.021
0.080
0.007
0.090
0.073
0.059
0.036
0.009
0.023
0.040
0.154
0.194
0.001
0.004
0.005
0.003
0.019
0.006
0.001
0.002
0.003
0.007
0.014
0.016
0.010
0.006
0.017
0.020
0.025
0.039
0.044
0.033
0.104
0.104
0.096
0.045
$226.370,023
$7,251,007
$332,066,427
$42,207,676
$2.886.667
$3,572.411
$2.001,572
$839.058
$8.230,415
$3,832,894
$792.042
$642,704
$671.434.
$69,340.600
$15,409,100
$23.654.474
$362.143
$3.251,553
$375.057.737
($31.697,204)
($27.567.647)
$47.267.149
($15.677.667)
($12.745.340)
($6,354.625)
($5.065.838)
$18.246.445
$15.924,943
$42,789,365
($2,764.839)
($535,308)
$665,275
$6.579.228
$248.878.183
$105,968,341
$33.095.450
$3,322.784.810
0.1181
0.5929
2.2016
0.1905
2.4901
2.0079
1.6298
0.9988
0.2478
0.6324
1.0966
4.2644
5.3755
0.0198
0.1186
0.1352
0.0775
0.5217
0.1711
0.0368
0.0618
0.0710
0.2016
0.3834
' 0.4427
0.2688
0.1668
0.4822
0.5415
0.6877
1.0791
1.2095
0.9252
2.6715
2.8715
2.6585
1.2490
1 The selection of control options and the estimation of capital and annual costs and removal efficiencies
are projections and estimations based on discussions with AHA staff, literature review and engineering
judgement. Actual costs as well as feasibility and effectiveness could differ considerably from
our estimates.
2 This study uses conservative estimates of Increased cancer risk from ambient (I.e.. outdoor) exposure to establish
priorities among pollutants and sources. The risk estimates are calculated using modelled or monitored
concentrations and EPA unit cancer risk factors. There is considerable uncertainty in the estimated
concentrations, which could either overstate or understate the true concentrations (see Chapter IV). Unit cancer
risk factors combine CAG potency estimates with EPA exposure assumptions. The CAG potency estimates provide
a plausible upper limit to the cancer risk of a compound (see Appendix A); however, the true value of the risk
is unknown and may be as low as zero. The exposure assumptions are extremely conservative in that they assume
continuous exposure to outdoor air for 70 years. Because of the generally conservative bias In the information.
it is highly unlikely that the true risks would be as high as the estimates, and they could be considerably lower.
VI-18
-------
The results in Table VI-6 are presented only as part of our
demonstration project. There is considerable uncertainty
surrounding our estimates of emissions, risks, and cost. A more
detailed analysis would be necessary to validate these results.
b. Estimating the Benefits Associated with the Control
Options
An important economic benefit that can result from control
of toxic air pollutants is the concomitant reduction in emissions
of pollutants that impair visibility and cause smog, including
the adverse respiratory health effects associated with smog.
These pollutants—photochemically reactive compounds, such as
VOCs in general, that contribute to ozone formation; total
suspended particulate (TSP) matter often emitted by combustion
sources; oxides of sulfur and nitrogen; and carbon monoxide (also
called "criteria pollutants"2)—are frequently controlled by the
same control technologies designed to reduce health risks
attributed to toxic air pollutants. Information on the changes
in health and environmental effects resulting from reductions in
emissions and exposure levels to criteria air pollutants can
change the relative effectiveness and outcomes of control
strategies directed at toxic air pollutants.
To conduct a benefit-cost analysis, we developed benefits
estimates for each control option selected to reduce air toxics
emissions. The benefits estimates required three pieces of
information: (1) the effectiveness of each control option in
reducing emissions of particulates and VOCs, (2) the benefits
categories related to particulates and VOCs that are of interest,
and (3) an estimate of the economic value for each ton of
particulate and VOC reduced by benefit category. We used the
data presented in Appendix C on the effectiveness of control
options for air toxics emissions in reducing emissions of
particulates and VOCs. We based our analysis of economic
benefits on the quantified health and welfare effects listed in
Table VI-7. Appendix D provides more detail on the benefits
calculations and the sources of the benefits numbers by category.
The procedure we followed to estimate benefits required several
simplifying assumptions:
• For the control options analyzed, a reduction of
1 percent in total particulate emissions translates
into a 1 percent reduction in ambient concentrations of
particulate matter.
2 They are called criteria pollutants because they serve to
define ambient air quality under the Federal Clean Air Act.
VI-19
-------
TABLE VI-7
Health and Welfare Effects Quantified
Pollutant
Total suspended
particulate matter
Effects
Mortality
- Carcinogenic
- Noncarcinogenic
(Adverse Respiratory
Effects)
Morbidity
- Acute Effects
(Work Loss Days,
Restricted Activity
Days, Medical Costs)
Welfare
- Material Effects
- Visibility Effects
Volatile organic compounds
Mortality
Carcinogenic
Morbidity
- Acute Effects
(Work Loss Days,
Restricted Activity
Days, Medical Costs,
Eye Irritation, Coughing,
Chest Discomfort, Shortness
of Breath, Headaches)
- Chronic Effects
(Respiratory Disease)
Welfare
- Agricultural Effects
- Material Effects
- Ornamental Plant Effects
VI-20
-------
• For the control options analyzed, a reduction of 1
percent in VOC emissions results in a 0.6 percent
reduction in ambient air concentrations of ozone.
• Reductions in the ambient air concentrations of
particulate matter and ozone are evenly distributed
over the Baltimore study area.
We estimated the expected economic benefits that would
result from reducing one ton of emissions for both particulate
matter and volatile organic compounds using the results of
readily available statistical analyses. These statistical
studies provided a quantitative estimate of the relationship
between changes in health and welfare effects and predicted
changes in the levels of ambient concentrations of particulates
and ozone.
To arrive at an estimate of the economic benefits of each
control option, we multiplied the total reductions in emissions
of particulates and VOCs that each control option would provide
by our estimates of economic benefit per ton. The results of
this exercise are summarized in Table VI-8. The expected
economic benefits from reducing emissions of particulate matter
are about $14,000 per ton reduced. The expected benefits of
reducing ozone levels resulting from controls on the emissions of
VOCs are $345 per ton reduced. Both of these estimates of
benefits fall within the ranges of the estimated economic
benefits of reducing TSP and ozone concentrations performed for
other cities under conditions similar to those modelled in
Baltimore.
3. ANALYSIS OF THE COST-EFFECTIVENESS OF CONTROL STRATEGIES
In this section, we present the results of our analysis of
cost-effective-strategies for reducing risks from ambient air
toxics and the results. Figure VI-1 illustrates the key pieces
of information required to conduct a cost-effectiveness analysis.
The information necessary for the activities represented in the
boxes in the left-hand side of the diagram was generated in the
risk assessment phase of the study. The activities represented
on the right-hand side were conducted once key sources and
pollutants posing risk had been identified, as described above.
We used a computer (mixed integer programming) model to select
those options that would provide a given percent reduction in
risk at the lowest cost for the sources and pollutants examined.
a. Cost-Effectiveness Analysis Overview
For the purposes of our study, we defined cost-effectiveness
as the cost of control per unit of health risk reduced. We
defined effectiveness as the percent reduction in either the
VI-21
-------
TABLE VI-B
CONTROL OPTIONS, ANNUALIZEO BENEFITS. AND COSTS FOR TSP AND VOC CONTROLS U.2)
ANNUAL ANNUAL
VOC TSP ANNUAL
CONTROL REDUCED REDUCED REDUCED
SOURCE Tin OPTION EMISSIONS EMISSIONS CAHCERS
(KKC/TR) (KKC/YR)
M
1
to
to
HEAVY-DUTY DIESEL VEHICLES
LIGHT-DUTY DIESEL VEHICLES
POINT SOURCE A - COKE OVENS
POINT SOURCE A - METALS
POINT SOURCE A - BENZENE
POINT SOURCE B - METALS
POINT SOURCE C - METALS
POINT SOURCE C
GASOLINE MARKETING
DECREASINC-PERCBLOROETHYLENE
DECREAS1NG-TRICHLOROETHYLENE
DEGREASING-METHYLEHE. CHLORIDE
DRYCLEAN1NC-PERCBLOROBTHYLENE
MISC. INDUSTRIAL-METHYL. CHLORIDE
WOODSTOVES-RESIDENTIAL HEATING
COMMERCIAL OIL COMBUSTION
RESIDENTIAL COAL COMBUSTION
RESIDENTIAL OIL COMBUSTION
1119
309
0
0
0
648
522
299
81
223
0
0
0
0
0
0
0
0
2955
80
135
155
155
290
335
62
368
1061
1188
158
248
278
93.6
-3.5
-5.3
146.12
0.4
20
12.5
77
7
0
0
0
0
0
0
39
34
72
a
7
15
17
0.94
0
0
0
0
0
0
0
0
0
0
0
0
0
0
148
71.09
71.09
534.94
213.9
0.004
0.021
0.080
0.007
0.090
0.073
0.059
0.036
0.009
0.023
0.040
0.154
0.194
0.001
0.004
0.005
0.003
0.019
0.006
O.OOI
0.002
0.003
0.007
0.014
0.016
0.010
0.006
0.017
0.020
0.025
0.039
0.044
0.033
0.104
0.104
0.096
0.045
ANNUAL
CANCER CONTROL
BENEFITS
(1987 $)
$8,540
$42,857
$159,143
$13,771
$180,000
$145,143
$117,811
$72,200
$17,914
$45,714
$79,269
$308,257
$388,571
$1,429
$8,571
$9.771
$5,600
$37,714
$12.371
$2,657
$4,469
$5.131
$14,571
$27.714
$32,000
$19,429
$12,057
$34,857
$39,143
$49,714
$78.000
$87.429
$66,877
$207,571
$207,571
$192,171
$90,286
VALUE
PER REDUCED
CAHCERS
TOTAL
ANNUAL TSP
BENEFITS
(1987 $)
$274,000
$171,250
$1.054,900
$95,900
$0
$0
$0
$0
$0
$0
$534,300
$465,800
$986,400
$109,600
$95,900
$205,500
$232.900
$12.878
$0
$0
$0
$0
$0
$0
$0
$0
$0
$0
$0
$0
$0
$0
$2.027,600
$973,933
$973,933
$7.328,678
$2,930,430
TSP
BENEFIT
PER TON
TOTAL
ANNUAL VOC
BENEFITS
(1987 $>
$386,055
$106,605
$0
$0
$0
$223,560
$180,090
$103,155
$27,945
$76,935
$0
$0
$0
$0
$0
$0
$0
SO
$1,019,475
$27,600
$46,575
$53,475
$53,475
$100,050
$115,575
$21.390
$126,960
$366,045
$409,860
$54,510
$85,560
$95,910
$12,292
($1,898)
($1,898)
$50,411
$138
VOC
BENEFIT
PER TON
TOTAL
ANNUAL
BENEFITS
(1987 $)
$668,595
$320,712
$1,214,043
$109.671
$180,000
$368.703
$297,901
$175,355
$45,859
$122,649
$613,569
$774,057
$1,374.971
$111.029
$104,471
$215,271
$238,500
$50,592
$1,031,846
$30,257
$51,044
$58,606
$68,046
$127,764
$147,575
$40,819
$139,017
$400,902
$449,003
$104,224
$163,560
$183,339
$2,126.769
$1,179,607
$1,179,607
$7,571,261
$3,020,834
TOTAL
ANNUAL
CONTROL
COSTS
(1987 $)
$966,600
$155.593
$26,423,000
$290,630
$259,800
$259,255
$117,904
$30,290
$73,721
$87,609
$31,392
$99,059
$130,450
$49,529
$66.039
$115,569
$1,014
$61,315
$2.320,000
($42,112)
($61,594)
$121.274
($114,223)
($176.614)
($101,674)
($49,211)
$110.000
$277,549
$837,449
($68,726)
($20,877)
$29;082
$220,000
$25,830.000
$10,998,000
$3,180,000
$150,000,000
TOTAL
ANNUAL
NET BENEFITS
(1987 $)
($298,003)
$165.119
($25.208.957)
($180,959)
($79.800)
$109.448
$179,997
$145,065
($27,862)
$35,040
$582,177
$674.998
$1,244,521
$61 , 500
$38,432
$99.702
$237,486
($10,723)
($1,288,154)
$72,369
$112.638
($62,668)
$182,269
$304,378
$249.249
$90,030
$29,017
$123.351
($388,446)
$172,950
$184,437
$134,257
$1,906,769
($24,650,193)
($9,818,393)
$4.391,261
($146.979.146)
$2,000,000 $13700
$345
1. The selection of control options and the estimation of capital and annual coses and removal efficiencies are projections
and estimations based on discussions with AHA staff, literature review and engineering Judgment., Actual costs as well as
-------
TABLE Vl-B (continued)
2. This study usea conservative estimates of increased cancer risk fron ambient (i.e., outdoor) exposure to establish
priorities among poilutanta and sources. The risk estimates are calculated using nodelled or monitored concentrations
and EPA unit cancer risk factors. There is considerable uncertainty in the estimated concent radons, which could either
overstate or understate the true concentrations, (see Chapter IV). Unit cancer risk factors combine CAG potency estimates
with EPA exposure assumptions. The CAC potency estimates provide a plausible upper limit to the cancer risk of a compound
(see Appendix A); however, the true value of the risk is unknown and may be as low aa gero. The exposure assumptions are
extremely conservative ta chat they assume continuous exposure to outdoor air for 10 years. Because of the generally
conservative bias in the Information, it is highly unlikely that the true risks would be as high as the estimates, and
they could be considerably lower.
<
M
(O
-------
Figure VI -1
Baltimore IEMP Air Toxics Study
Analysis Of The Cost-Effectiveness
Of Control Options 1
Information Flow
Sources
(1987 control levels)
• Point • Area
Pollutant Release
Feasible
Control Option
Exposure Pathway
Dispersion Modeling
Control Costs
Human Health Effects
(Baseline Cancer Risks)
• annual excess cancer
• maximum lifetime
individual cancer nsk
Computer
Model
Control Effectiveness
(Percent removal
efficiency)
Cost-Effective Control Strategies
•Cost of the reduction in annual excess
cancer incidence
•Cost of the reduction in maximum
lifetime individual cancer risk
Cost-effecfivness is defined as the cost of control per unit of health nsk reduced.
VI-24
-------
annual excess cancer incidence or the maximum increased lifetime
individual cancer risk. It is important to note that this
analysis did not attempt to reassess the cost-effectiveness of
controls currently in use; we accepted these as part of the
baseline.
A cost-effectiveness analysis of control options can provide
information useful to local risk managers in setting control
priorities. In particular this analysis took an approach that
allows the risk manager to determine (1) which individual control
options are most cost-effective, and (2) the cheapest set (or
combination) of control options from for all sources included in
the analysis that will achieve any predetermined percentage risk
reduction level. We used a computerized optimization technique,
known as a mixed integer program/ to identify the least-cost
control strategy to meet several specified risk reduction levels,
We chose the different percent reductions in risk to human
health to identify a series of control strategies for different
levels of risk reduction. As we requested progressively higher
levels of risk reduction, the computer model reconfigured the set
of control options (above current controls) to determine the
optimum strategy to provide the desired level of risk reduction
at the least cost. We employed this strategy because we did not
know the risk manager's preferences ahead of time. Comparing
costs to the percent reduction in cancer risk (above current
controls) aids the risk manager in deciding how much risk
reduction to target.
The results of the cost-effectiveness analysis are provided
below for two measures of human health risk: annual excess
cancer incidence (both area-wide and for the grid cell of maximum
annual excess cancer incidence), and maximum increased lifetime
individual cancer risk. In some situations, the risk manager may
be faced with a need to implement two different control
strategies to reduce both measures of risk. In our discussion
that follows/ we explore whether this situation occurs in the
Baltimore area. We also assess qualitatively how the different
control strategies will affect noncancer effects. This section
concludes by summarizing the major limitations of the analysis.
b. Coat-Effectiveness Analysis Results
i. Results; Annual Excess Cancer Incidence
(1) Area-Wide Control Options
Individual Options. Table VI-6 lists the cost-effectiveness
(i.e./ the annual cost per cancer case avoided) for all control
options evaluated in this analysis. As shown/ the options range
from a net annual saving of about $32 million per case avoided
from perchloroethylene degreasing (control option 1) to a maximum
vi-25
-------
of over $3 billion from residential oil combustion (control
option 1). By inspecting the cost-effectiveness of each option,
in conjunction with a policy determination of risk reduction
goals, the decision maker can explore where additional control
may be warranted. For example, of the individual options to
reduce at least annual excess cancer incidence over ten years,
control option 2 at Point Source B ranks best at annualized cost
of approximately $643,000. It is also interesting to note the
numerous control options that could result in a savings while
reducing cancer risk.
Control Strategies. Using our automated model, we were able
to ranking and compile options that would achieve various
predetermined risk reduction levels. Table VI-9 shows the least-
cost control strategies for reducing different levels of the
estimated upper-bound annual excess cancer incidence posed by the
target pollutants. This table also displays the total cost,
average cost per case reduced, and incremental cost per case
reduced for each control strategy. As can be seen from this
table as well as from the graph in Figure VI-2, the total
annualized cost of implementing the control strategies associated
with each of the points A through K ranges from a low of negative
$200,000 (that is, a cost savings) to a high of almost $211
million for the controls considered.
We estimate that we can achieve a modest reduction in the
estimated annual excess cancer incidence (by roughly 7 percent)
and save money by implementing solvent recovery programs for area
sources emitting perchloroethylene, TCE, and methylene chloride
(again at cost savings), and controls at Point Sources B and C at
modest cost. The cost savings for solvent usage are the result
of controls that lead to the recovery of solvent and reduced
solvent consumption.
Roughly one excess cancer case every two years (see point 0)
could be avoided at a cost of $1.4 million per year (that is,
about $2.8 million per case avoided). This is within the range
of values often given by economists for the value of a
statistical life saved ($400,000 to $7,000,000).3
The cost per excess cancer case avoided for a given set of
control strategies increases with increasing requirements on risk
reduction. The costs begin to rise rapidly with the reduction of
0.51 cancer cases per year at point D (a 14 percent reduction
from baseline risk) because the incremental (or marginal) cost
for each additional case reduced begins to rise sharply.
Controls are required on all of the point sources and most of the
area sources. The jump in cost is explained by the selection of
3 U.S. EPA, "Regulatory Impact Guidelines," December 1983,
EPA 230-01-84003.
VI-26
-------
Table Vl-9
BALTIMORE IEMP AIR TOXICS STUDY
SCHEDULE OF CONTROL STRATEGIES FOR REDUCING CANCER INCIDENCE
PHASE II RESULTS INTENDED FOR POLICY DEVELOPMENT1-2
Point
on Cases
Figure VI-2 Reduced
Graph per Year
A 0.25
B 0.27
C 0.34
D 0.51
Percent Avg. Cost per Incremental
Reduction in Total Case Reduced Cost per
Cancer Aromatized from Current Incremental Pollution Controls Implemented
Incidence from Cost Control Case Reduced
Current Control (tlOOO/vearl (11000/case) (SlOQO/case) Source Control Potion
7 -{195 -J780 $39,000 Point Source A-benzene controls
Point Source 6
Perc degreasing
TCE degreasing
Methylene chloride degreasing
Misc. industrial
methylene chloride use
7.5 J73 $268 {13,400 Point Source A-benzene controls
Point Source B
Point Source C
Point Source 0
Perc degreasing
TCE degreasing
Methylene chlonde-degreasing
Misc. industrial
methylene chloride use
9.5 $183.4 J534.9 $1,572 Point Source A-chrone controls
Point Source A-benzene controls
Point Source B
Point Source C
Perc degreasing
TCE degreasing
Methylene degreasing
Misc. industrial
methylene chloride use
14 Jl. 416.2 JZ.800.4 $7,573 Light duty diesel vehicles
Point Source A-coke ovens
controls
Point Source A-chrome controls
Point Source A-benzene controls
Point Source B
Point Source C
Point Source D
Perc degreasing
TCE degreasing
Methylene chloride degreasing
Dryc lean ing
Misc. industrial
methylene chloride use
Woodstoves
3
3
2
2
1
1
2
3
4
1
2
2
1
1
1
3
3
4
2
2
1
1
L
2
1
1
3
I
1
2
2
1
2
2
L
VI-27
-------
Table Vl-9 (continued)
BALTIMORE IEMP AIR TOXICS STUDY
PHASE II RESULTS INTENDED FOR POLICY DEVELOPMENT
SCHEDULE OF CONTROL STRATEGIES FOR REDUCING CANCER INCIDENCE1'2
Point
on
Figure VI-2
Graph
Cases
Reduced
per Year
0.69
Percent
Reduction in
Cancer
Incidence from
Current Control
19
Total
Annualized
Cost
fllQOO/vear)
$7.730
Aug. Cost per Incremental
Case Reduced Cost per
from Current Incremental Pollution Controls Implemented
Control Case Reduced
(SlOOO/case) (tlOOO/case) Source
Control Option
$11.273
$39.463
Light duty diesel vehicles 1
Point Source A-coke oven controls 2
Point Source A-chrome controls 1
Point Source A-benzene controls 1
Point Source B 3
Point Source C <
Point Source 0 1
Gasoline marketing 1
Perc degrees ing 2
TCE degreasing 2
Methylene chloride degreasing 1
Dryc lean ing 3
Misc. industrial 2
methylene chloride use
woodstoves 1
Residential coal 1
0.72
20
$15.592
$21.570
$211.344
Light duty diesel vehicles 1
Point Source A-coke oven controls 2
Point Source A-chrome controls
Point Source A-benzene controls
Point Source B
Point Source C
Point Source D
Perc degreasing
TCE degreasing
Methylene chloride degreasing
Dryc lean ing
Misc. industrial
methylene chloride use
Woodstoves
Commercial oil controls
Residential coal
0.76
21.0
$45,084
$59.433
$826.106 Heavy duty diesel vehicles
Light duty diesel vehicles
1
PoTnt Source A-coke oven controls 1
Point Source A-chrome controls 1
Point Source A-benzene controls 1
Point Source B 3
Point Source C 2
Point Source 0 1
Gasoline marketing 1
Perc degreasing 2
TCE degreasing 3
Methylene chloride degreasing 1
Drycleaning 2
Misc. industrial 2
methylene chloride use
Woodstoves 1
Conmercial oil controls 2
Residential coal 1
VI-28
-------
Table VI-9 (continued)
BALTIMORE IEMP AIR TOXICS STUDY
PHASE II RESULTS INTENDED FOR POLICY DEVELOPMENT
SCHEDULE OF CONTROL STRATEGIES FOR REDUCING CANCER INCIDENCE2
Percent Avg. Cost per Incremental
Point Reduction in Total Case Reduced Cost per
on Cases Cancer Annualized from Current Incremental Pollution Controls Implemented
Figure VI-2 Reduced Incidence from Cost Control Case Reduced
Graph per Year Current Control (SlOOO/vear) (SlOOO/case) (SlOOO/case) SourceControl Option
H 0.80 22.0 199.976 $130.000 $1.479.569 Heavy duty diesel vehicles L
Light duty diesel vehicles 1
Point Source A-coke oven controls 1
Point Source A-chrome controls 1
Point Source A-benzene controls 1
Point Source B 3
Point Source C 3
Point Source 0 I
Gasoline marketing 1
Perc degreasing 3
TCE decreasing 3
Methylene chloride degreasing 1
Dryclean ing 3
Misc. industrial 2
methylene chloride use
Woodstoves 1
Conrercial oil controls 2
Residential coal 1
I 0.83 23.0 S191.800 1230.000 $2.572.101 Light duty diesel vehicles 1
Point Source A-coke oven controls 1
Point Source A-chrome controls 1
Point Source A-benzene controls 1
Point Source B 3
Point Source C 2
Point Source 0 1
Perc degreasing 2
TCE degreasing 3
Methylene chloride degreasing 1
Drycleaning 2
Misc. industrial 2
methylene chloride use
Woodstoves 1
Comnercial oil controls 2
Residential coal
Residential oil
J 0.84 23.3 $195,876 $230,000 $407,600 Heavy duty diesel vehicles
Light duty diesel vehicles
Point Source A-coke oven controls
Point Source A-chrome controls
Point Source A-benzene controls
Point Source B
Point Source C
Point Source 0
Gasoline marketing
Perc degreasing 3
TCE degreasing 3
Methylene chloride degreasing 1
Orycleaning 3
Misc. industrial 2
methylene chloride use
Woodstoves 1
Comnercial oil controls 2
Residential coal 1
Residential oil 1
VI-29
-------
Table VI-9 (continued)
BALTIMORE IEHP AIR TOXICS STUDY
PHASE II RESULTS INTENDED FOR POLICY DEVELOPMENT
SCHEDULE OF CONTROL STRATEGIES FOR REDUCING CANCER INCIDENCE1'2
Point
on
Figure VI-2
Graph
Cases
Reduced
per Year
0.85
Percent
Reduction in
Cancer
Incidence from
Current Control
Total
Annualized
Cost
(SlOOO/vear)
Avg. Cost per Incremental
Case Reduced Cost per
from Current Incremental Pollution Controls Implemented
Control Case Reduced
HlOOO/case) HlOOO/case) Source
Control Option
23.5 5210,916
1250.000 (1.744.186 Heavy duty diesel vehicles 1
Light duty diesel vehicles 1
Point Source A-coke oven controls 1
Point Source A-chrome controls 1
Point Source A-benzene controls 1
Point Source B 3
Point Source C 3
Point Source D 1
Gasoline marketing 1
Perc degreasing 3
TCE degreasing 3
Methylene chloride degreasing 1
Drycleaning 3
Misc. industrial 3
methylene chloride use
Woodstoves 1
Comnercial oil controls 1
Residential coal 1
Residential oil 1
This study uses conservative estimates of increased cancer risk from ambient (i.e.. outdoor) exposure to establish
priorities among pollutants and sources. The risk estimates are calculated using modelled or monitored concentrations
and EPA unit cancer risk factors. There is considerable uncertainty in the estimated concentrations, which could either
overstate or understate the true concentrations (see Chapter IV). Unit cancer risk factors combine CAG potency
estimates with EPA exposure assumptions. The CAG potency estimates provide a plausible upper limit to the cancer risk
of a compound (see Appendix A); however, the true value of the risk is unknown and may be as low as zero. The exposure
assumptions are extremely conservative in that they assume continuous exposure to outdoor air for 70 years. Because of
the generally conservative bias in the information, it is highly unlikely that the true risks would be as high as the
estimates, and they could be considerably lower.
Cost estimates were based on engineering estimates and best professional judgment and may significantly over- or
uhderstimate real costs.
VI-30
-------
FIGURE VI-2
TOTAL ANNUALIZED COST VERSUS REDUCTION IN
ANNUAL EXCESS CANCER INCIDENCE
i_
-------
area source controls, such as the control options for woodstoves
and dry cleaners, and controls at Point Source D that are costly
given the amount of risk reduction that they afford.
Overall, we can identify control strategies that can reduce
only about 24 percent of the estimated cancer incidence in the
Baltimore area. Implementation of the most expensive strategy
identified would reduce the estimated annual excess cancer
incidence from approximately 3.6 to only about 2.8.*
The data suggest that any cost-effective control strategy
would need to include controls on area-wide uses of solvents and
relatively less expensive controls on emissions from Point
Sources A, B, and C. To achieve any modest reductions in risk
(14 to 19 percent reduction from current baseline levels posed by
the study pollutants), we would also need to impose controls on
most area sources, such as light-duty trucks, dry-cleaning
establishments, and the use of woodstoves for home heating.
Significant reductions in incidence (greater than 20 percent)
would require relatively expensive controls on all the point
sources included in the analysis, as well as controls on area
sources.
The results in Table VI-9 are presented only as part of our
demonstration project. There is considerable uncertainty
surrounding our estimates of emissions, risks/ and cost. A more
detailed analysis would be necessary to validate these results.
(2) Control Strategies for the Grid Cell of
Maximum Annual Excess Cancer Incidence
We also explored the impact that each of the strategies
identified in Table VI-9 would have on the grid cell of maximum
incidence—the location of the highest population-weighted
average individual risk. As discussed in the previous chapter,
the estimated annual excess cancer incidence in this section of
the study area is approximately 0.15, with three sources
accounting for most of the quantified incidence (approximately 91
percent): Point Source B, Point Source A, and road vehicles.
Metals emissions at Point Source B account for 58 percent of the
total; POM releases from Point Source B and road vehicles
accounts for the remainder.
At all points, the combination of control options would
reduce the annual excess cancer incidence from 0.15 to 0.06 (a
percentage reduction of approximately 43 percent) by controlling
emissions of metals, primarily hexavalent chromium, from Point
'Transportation control measures were not included in the
analysis. Transportation control measures seek to reduce
emissions by reducing single-passenger automotive use.
Vl-32
-------
Source B (Option 3). Although the more costly control strategies
also include controls for Point Source A and light and heavy duty
diesel vehicles (a subset of road vehicles), these controls would
not result in significant reductions in POM ambient air releases.
The results for the grid cell of highest predicted annual
excess cancer incidence are presented only as part of our
demonstration project. There is considerable uncertainty
surrounding our estimates of emissions, risks, and cost. A more
detailed analysis would be necessary to validate these results.
(3) Noncarcinogenic Pollutants
Although we did not explicitly evaluate control options to
reduce the emissions of noncarcinogenic pollutants, we did
explore the extent to which our control strategies to reduce
annual excess cancer incidence also lower the concern for
noncancer effects. In the previous chapter, we indicated a need
to further investigate ambient air concentrations at two
locations because of a concern for noncancer effects. At one
location, identified using the refined 2.5 km grid system (UTM
coordinates 4342.25 369.75), benzene exposures exceeded the no-
effect threshold for blood effects, thus indicated a need to
further explore these exposure levels. Point Source A accounted
for 94 percent of the benzene ambient air concentrations at this
location. All control strategies selected at points D and higher
on Table VI-9 include Option 1 for benzene at Point Source A;
Option 1 has a benzene removal efficiency of approximately 97.7
percent. If Option 1 were implemented at Point Source A, the
ratio of benzene ambient concentrations to the no-effect
threshold for blood effects would drop from approximately 5.1 to
0.4, thus indicating that there is less need to investigate these
exposures.
If the control strategies associated with points A through C
on Table VI-9 were implemented, the need to further examine
benzene exposures would still exist. Although some benzene
controls at Point Source A are part of the control strategies at
points A through C/ the recommended control strategies include
either Option 3, which has a removal efficiency for benzene of
only 50.5 percent, or Option 2, which has a benzene removal
efficiency of approximately 79 percent. Option 3 would result in
a decrease in the ratio of benzene ambient air concentrations to
the no-effect threshold for blood effects from approximately 5.1
to 2.7. A ratio value of one or more indicates a need to further
examine these exposures. Similarly, Option 2 would result in a
ratio above 1.0.
At the second highest location of maximum increased lifetime
individual cancer risk, the pollutant of concern was xylene from
Point Source E. None of the control strategies considered Point
Source E. Thus the results of our analysis cannot be used to
VI-33
-------
identify strategies for reducing xylene exposures at this
location.
These results are presented only as part of our
demonstration project. There is considerable uncertainty
surrounding our estimates of emissions, risks, and cost. A more
detailed analysis would be necessary to validate these results.
ii. Results; Maximum Increased Lifetime Individual
Cancer Risk
For the control options analysis, we limited our evaluation
of cost-effective control options to the site within the
Baltimore study area with the maximum increased lifetime cancer
risk. As discussed in the previous chapter, the maximum
increased lifetime individual cancer risk posed by the study
pollutants at this location is approximately 1.3 x 10" . Similar
to other risk findings, Point Source A accounts for approximately
91 percent of the estimated maximum increased lifetime individual
risk at this site.
Table VI-10 arrays the least-cost control strategies for
reducing the estimated increased lifetime individual cancer risk.
This table also displays the total annualized cost and
incremental cost (that is, the fraction of the total cost
expended to move from one risk reduction scenario to the next)
for each control strategy. Figure Vl-3 graphically portrays the
total cost of achieving specified levels of individual risk
reduction.
Unlike the situation for reducing annual excess cancer
incidence posed by the study pollutants, few strategies exist for
reducing the maximum increased lifetime individual cancer risks
at this maximum increased lifetime individual cancer risk site.
Point Source A dominates because (1) it is the largest
contributor to the maximum increased lifetime individual cancer
risk and (2) the potential control costs at this source would be
significant. [Note, however, that the costs of control are
likely to be overestimated to the extent that construction of new
equipment has a utility to this producer separate from its value
in reducing pollutant emissions, we were not able to factor in
the savings to Point Source A resulting from the greater
production efficiency afforded by the control options.]
Implementation of control strategy A, shown as point A on
Figure VI-3, would result in a very minor (3 percent) reduction
in risk at possible cost-savings. Control strategy A combines
relatively low-cost options at Point Source A and Point Source C,
and cost-saving options that require improved solvent recovery by
industrial and commercial solvent users. Control strategy B
would provide a significant reduction in risk (an additional 45
percent) by requiring large outlays for construction of new plant
VI-34
-------
Table VI-10
BALTIMORE IEHP AIR TOXICS STUDY
SCHEDULE OF CONTROL STRATEGIES FOR REDUCING LIFETIME INDIVIDUAL CANCER RISK
AT SITE OF MAXIMUM INDIVIDUAL CANCER RISK1
PHASE II RESULTS INTENDED FOR POLICY DEVELOPMENT
Point
on
Figure VI -3
Graph
A
B
Percent
Reduction in Total
Expected Cancer Risk Annual Ized
Lifeline from Cost
Cancer Risk Current Control ($1000/year)
1.3 x 10'3 3 -$85
7.4 x ID'4 45 $26,000
Incremental Pollution Controls Implemented
Cost
($1000/yr) Source Control
Potion
-$85 Point Source A-chrome controls 1
Point Source C 4
Perc degreasing 2
TCE degreasing 2
Methylene chloride degreasing 1
Misc. industrial 1
methylene chloride use
$26,085 Point Source A-coke oven controls 1
Point Source A-chrome controls 1
Perc degreasing 2
TCE degreasing 2
Methylene chloride degreasing 1
Misc. industrial 1
methylene chloride use
7.2 x 10
-4
46.9
$42,000
7.1 x 10
,-4
47.1
$211,000
$16,000 Light duty diesel vehicles 1
Point Source A-coke oven controls 1
Point Source A-chrome controls 1
Point Source A-benzene controls 1
Point Source B 3
Point Source C 3
Perc degreasing 3
TCE degreasing 3
Methylene chloride degreasing 1
Dry cleaning 2
Misc. industrial 3
methylene chloride use
Woodstoves 1
Comnercial oil heating 2
Residential coal heating 1
$169,000 Heavy duty diesel vehicles 1
Light duty diesel vehicles 1
Point Source A-coke oven controls 1
Point Source A-chrome controls 1
Point Source A-benzene controls 1
Point Source B 3
Point Source C 3
Point Source 0 1
Gasoline marketing 1
Perc degreasing 3
TCE degreasing 3
Methylene chloride degreasing 1
Drycleaning 3
Misc. industrial 3
methylene chloride use
Woodstoves 1
Conrercial oil controls 1
Residential coal 1
Residential oil 1
(continued)
VI-35
-------
Table VI-10 (continued)
NOTES
This study uses conservative estimates of increased cancer risk from ambient (i.e.. outdoor) exposure to establish
priorities among pollutants and sources. The risk estimates are calculated using modelled or monitored concentrations
and EPA unit cancer risk factors. There is considerable uncertainty in the estimated concentrations, which could
either overstate or understate the true concentrations (see Chapter IV). Unit cancer risk factors combine CAG potency
estimates with EPA exposure assumptions. The CAG potency estimates provide a plausible upper limit to the cancer risk
of a compound (see Appendix A); however, the true value of the risk is unknown and may be as low as zero. The exposure
assumptions are extremely conservative in that they assume continuous exposure to outdoor air for 70 years. Because or
the generally conservative bias in the information, it is highly unlikely that the true risks would be as high as the
estimates, and they could be considerably lower.
Cost estimates were based on engineering estimates and best professional judgment and may significantly over - or
underestimate real costs.
VI-36
-------
FIGURE VI-3
TOTAL ANNUALIZED COST VERSUS REDUCTION IN
MAXIMUM LIFETIME INDIVIDUAL CANCER RISK
ZED COST
DOLLARS)
D h-
3?
ZfA
-------
equipment. We can achieve only minor additional reductions in
risk and large increases in cost through imposition of strategies
C and D. The resulting risk, about 7 x 10' , is still
significantly above the area-wide individual risk of about 1 to
2 x 10'*.
Regarding noncancer effects, the control strategies for
reducing the maximum increased lifetime individual cancer risk
that include high removal of benzene releases from Point Source A
(e.g., Option 1) will succeed in lowering the benzene ambient
concentrations to levels that require less investigation. Again,
the strategies considered in this analysis will not reduce xylene
ambient air concentrations at its maximum point (identified by
the discrete point analysis) because the control options analysis
did not consider the main contributor of xylene releases at this
location, Point Source E.
These results are presented only as part of our
demonstration project. There is considerable uncertainty
surrounding our estimates of emissions, risks, and cost. A more
detailed analysis would be necessary to validate these results.
iii. Limitations
We emphasize that the cost-effectiveness analysis was
designed primarily to demonstrate a methodology for evaluating
alternative pollution control strategies. Although the results
of the analysis can provide the risk manager with an indication
of possible remedial programs to pursue, additional analysis
needs to be performed before any strategy can be implemented.
Specifically, some of the major limitations of the analysis
include:
• The cost-effectiveness analysis could not consider all
sources and pollutants because of gaps in data.
• The analysis is based on preliminary estimates of
emissions, exposures, and risks by source. As we
discussed in Chapters IV and V, we believe that we have
overstated the exposures associated with Point Source A
. because of limitations in our ability to adequately
model this site.
• The analysis did not consider all possible control
options. A detailed engineering analysis is needed to
determine if a technology is, in fact, feasible.
• We overestimate the cost of certain control options to
both industry and area sources to the extent that we do
not take into account the age or usefulness of equip-
ment that is replaced under the proposed option or the
income from the sale of replaced equipment or scrap.
vi-38
-------
The cost estimates were rough approximations based on
limited data and were developed to compare relative.
and not absolute, costs. Additional work is needed in
this area.
4. MAXIMIZING THE BENEFITS TO SOCIETY IN CONTROLLING AIR TOXICS
Significant economic benefits can result from control of
toxic air pollutants by co-controlling emissions of VOCs, which
impair visibility and cause smog, and particulates, which cause
crop and materials damage. Consideration of these additional
effects could greatly alter how the decision maker views the
relative effectiveness and environmental impact of alternative
control strategies directed at toxic air pollutants. In this
section, we present the results of our preliminary evaluation of
least-cost control strategies for reducing air toxics emissions
that will also achieve the highest level of economic benefits.
a. Results of the Analysis of Strategies to Maximize
Benefits
In identifying control strategies, we chose the combinations
of control options that provided the most economic benefits at
arbitrarily chosen maximum costs to the regulated sources. We
summarize these results in Table VI-11.
As shown, we can achieve $3.6 million in economic benefits
at zero net cost. At point A in Table VI-11, costs of controls
for all point sources, dry cleaners, residential heating
(woodstoves and bans on the use of coal for residential heating),
and trucks are offset by cost savings resulting from the control
and recovery of solvents. The greatest total benefit for the
levels of costs that we have chosen occurs at point J. However,
the total annualized control costs for the selected control
strategy exceed total benefits. Thus, the net benefits are
negative. The point that maximizes the net benefits occurs at
point D. At an annualized cost to the affected sources of $5
million, we achieve estimated net benefits of $8.5 million and
about a 16 percent reduction in annual excess cancer incidence.
b. Comparison of Benefit and Cost-Effectiveness Analyses
We compared control strategies (the combination of control
options) to achieve a requisite reduction of cancer risk against
control strategies that maximized economic benefits subject to
cost constraints. The cost-effectiveness analysis generated a
set of control strategies that provides selected reductions of
cancer risk—measured as either annual excess cancer incidence or
maximum increased lifetime individual cancer risk—at the lowest
annualized implementation cost to sources. For the benefit-cost
VI-39
-------
TABLE VI-11
CONTROL STRATEGIES
BASED ON BENEFITS ANALYSIS (1)
Percent
Reduction in Total Total
Cases Cancer Annualued Annuahzed Net Pollution Control Implemented
Reduced Incidence from Cost Benefits Benefits
Point per Year Current Cntrl (Smil./yr) (Jmil./yr) (Jmil./yr) SourceOption
—X JT2 STTO JO JTB J3.6 Point Source A-coke oven cntr
chrome cntrIs
benzene cntrIs 3
Point Source B 1
Point Source C 4
Point Source 0 1
Gasoline marketing
Solvent usage-perc 2
TCE 2
methyl, chl. 1
Dry Cleaners
Misc. methylene chloride use 2
Woodstoves 1
Heating-comnercial oil
residential coal
residential oil
Heavy duty trucks
Light duty trucks
B 0.51 14.00% $1.0 IS.4 $4.4 Point Source A-coke oven cntr
chrome cntrIs 1
benzene cntrIs 2
Point Source B 3
Point Source C 4
Chemical Plant 1
Gasoline marketing
Solvent usage-perc 2
TCE 2
methylene chlor 1
Dry Cleaners 2
Misc. methylene chloride use 3
Woodstoves 1
Heating-connercial oil
residential coal
residential oil
Heavy duty trucks
Light duty trucks 1
0.54 14.90% $2.6 $6.3 $3.7 Point Source A-coke oven cntr 2
chrome cntrIs 1
benzene cntrIs 1
Point Source B 3
Point Source C 4
Point Source 0 1
Gasoline marketing
Solvent usage-perc 2
TCE 3
methylene chlor 1
Dry Cleaners 2
Misc. methylene chloride use 3
Hoodstoves I
Heating-connercial oil
residential coal
residential oil
Heavy duty trucks 1
Light duty trucks 1
VI-40
-------
TABLE VI-11
CONTROL STRATEGIES
BASED ON BENEFITS ANALYSIS (1)
Percent
Reduction in Total Total
Cases Cancer Annual)zed Annual*zed Net Pollution Control Implemented
Reduced Incidence from Cost Benefits Benefits
Point per Year Current Cntrl (Jmll./yr) ($mil./yr) (Jrail./yr) SourceOption
—B Q757 rSTTH JO JITS SB.5 Point Source A-coke oven cntr
chrome cntrIs 1
benzene cntrIs 2
Point Source B 3
Point Source C 4
Point Source 0
Gasoline marketing
Solvent usage-perc 2
TCE 2
methylene en lor 1
Dry Cleaners 2
Nisc. methylene chloride use 1
Hoodstoves 1
Heating-coimercial oil
residential coal 1
residential oil
Heavy duty trucks 1
Light duty trucks 1
0.59 16.32* $5.4 S13.8 S3.4 Point Source A-coke oven cntr 2
chrome cntrIs 1
benzene cntrIs 1
Point Source B 3
Point Source C 4
Point Source D
Gasoline marketing
Solvent usage-perc 2
TCE 2
methylene chlor 1
Dry Cleaners 2
Misc. methylene chloride use 1
Hoodstoves 1
Heating-conmercial oil
residential coal 1
residential oil
Heavy duty trucks 1
Light duty trucks 1
0.6 16.69% $7.5 S14.7 S7.2 Point Source A-coke oven cntr
chrome cntrIs 1
benzene cntrIs 1
Point Source B 3
Point Source C 4
Point Source 0
Gasoline marketing 1
Solvent usage-perc 2
TCE 2
methylene chlor 1
Dry Cleaners 2
Misc. methylene chloride use 2
Woodstoves 1
Heatlng-cornnercial oil
residential coal 1
residential oil
Heavy duty trucks 1
Light duty trucks 1
VI-41
-------
TABLE VI-11
CONTROL STRATEGIES
BASED ON BENEFITS ANALYSIS (1)
Percent
Reduction in Total Total
Cases Cancer Annual!zed Annual!zed Net Pollution Control Implemented
Reduced Incidence from Cost Benefits Benefits
Point per Year Current Cntrl (Smil./yr) ($mil./yr) ($m11./yr) SourceOption
—5 ITW 17^61 5577 STO 56.2 Point Source A-coke oven cntr T
chrome cntrIs 1
benzene cntr Is 1
Point Source B 3
Point Source C 4
Point Source D 1
Gasoline marketing 1
Solvent usage-perc 3
TCE 3
methylene chlor 1
Dry Cleaners 3
Hisc. methylene chloride use 3
Hoodstoves 1
Heating-commercial oil
residential coal 1
residential oil
Heavy duty trucks 1
Light duty trucks 1
H 0.64 17.66% $8.7 $14.9 $6.2 Point Source A-coke oven cntr 2
chrome cntrIs 1
benzene cntrIs 1
Point Source B 3
Point Source C 4
Point Source 0 1
Gasoline marketing 1
Solvent usage-perc 3
TCE 3
methylene chlor 1
Dry Cleaners 3
Misc. methylene chloride use 3
Hoodstoves 1
Heating-commercial oil
residential coal 1
residential oil
Heavy duty trucks 1
Light duty trucks 1
0.71 19.60% $19.5 $14.0 ($5.5)Po1nt Source A-coke oven cntr 2
chrome cntrIs 1
benzene cntrIs 1
Point Source B 3
Point Source C 4
Point Source 0 1
Gasoline marketing 1
Solvent usage-perc 3
TCE 3
methylene chlor 1
Dry Cleaners 3
Misc. methylene chloride use 3
Hoodstoves 1
Heating-commercial oil 2
residential coal 1
residential oil
Heavy duty trucks 1
Light duty trucks 1
VI-42
-------
TABLE VI-11
CONTROL STRATEGIES
BASED ON BENEFITS ANALYSIS (1)
Percent
Reduction in Total- Total
Cases Cancer Annualized Annualized Net Pollution Control Implemented
Reduced Incidence from Cost Benefits Benefits
Point per Year Current Cntrl (Smil./yr) ($mil./yr) (Jmil./yr) SourceOption
—] OI ZTTSftJ3TB JT7T2 (J28.4)Point Source A-coke oven cntr T
chrome cntrIs 1
benzene cntrIs 1
Point Source B 3
Point Source C 4
Point Source 0 1
Gasoline marketing 1
Solvent usage-perc 3
TCE 3
methylene chlor 1
Dry Cleaners 3
Misc. methylene chloride use 3
Uoodstoves 1
Heatlng-cormercial oil 2
residential coal 1
residential oil
Heavy duty trucks 1
Light duty trucks 1
This study uses conservative estimates of increased cancer risk from ambient (i.e.. outdoor)
exposure to establish priorities among pollutants and sources. The risk estimates are calculated
using modelled or monitored concentrations and EPA unit cancer risk factors. There is
considerable uncertainty in the estimated concentrations, which could either overstate or
understate the true risk (see Chapter IV). Unit cancer risk factors combine CAG potency
estimates with EPA exposure assumptions. The CAG potency estimates provide a plausible upper
limit to the cancer risk of a compound (see Appendix A); however, the true value of the risk is
unknown and may be as low as zero. The exposure assumptions are extremely conservative in that
they assume continuous exposure to outdoor air for 70 years. Because of the generally
conservative bias in the information, it is highly unlikely that the true risks would be as high
as the estimates, and they could be considerably lower.
VI-43
-------
analysis, we sought control strategies that provided optimal
monetary benefits (including both health, cancer and non-cancer,
and welfare benefits) subject to an arbitrarily chosen set of
maximum expenditures that could be required of the regulated
sources.
We compared the two approaches by asking whether optimal
strategies for obtaining equivalent reductions in annual excess
cancer incidence under the two different approaches result in
different combinations of pollution control options. In other
words, does the omission of noncancer health and welfare benefits
from the cost-effectiveness scenarios result in strategies that
obtain the desired reduction in cancer risks, but do not provide
the greatest net economic benefits? Or, does the least cost
(i.e., the optimal) set of control options to achieve a desired
level of excess cancer risk result in less net economic benefit
to residents of the Baltimore area than an alternative set that
imposes greater cost on the regulated sources?
In Figure VI-4, we display the total costs of controlling
air toxics associated with a reduction in risk under each
approach. As shown, the control costs to reduce cancer risk rise
gradually up to about the 20 percent reduction level. Above this
point, the costs of reducing additional cancer risk rise sharply.
The selection of control options using either the benefit-cost
method or cost-effectiveness method appear to be fairly
consistent when looking at the relative positions of the point on
the figure.
Figure VI-5 displays the total benefits for each risk
reduction strategy under the benefit-cost and cost-effectiveness
approaches. The points in the lower left quadrant indicate that
for low levels of risk reduction, there are tradeoffs between
maximizing benefits and minimizing cancer risks. Some controls
that provide greater total benefits do a better job of reducing
noncancer risks than other options providing a greater reduction
in the estimated excess cancer incidence. For cancer incidence
reduction rates of 14 to 20 percent, the selection of control
options is, again, fairly consistent across the two methods.
The final figure, Figure VI-6, combines the information from
the total cost and total benefit figures in order to discern
which control options provide the greatest net benefits. The
rise in net benefits is relatively small up to a 16 percent
reduction in the estimated annual excess cancer incidence. After
this point, net benefits begin to fall, reaching zero around the
19 percent risk reduction level. The selection of options
reducing annual excess cancer incidence beyond this point are
costly, and the benefits from reducing annual excess cancer
incidence and the risk of noncancer effects fail to keep pace
with the cost increases.
VI-44
-------
FIGURE VI-4
SUMMARY OF COSTS OF CONTROLLING AIR TOXICS IN BALTIMORE
COMPARISON OF BENEFIT METHOD AND COST-EFFECTIVE SCENARIOS
1
•£•
01
-------
FIGURE VI-5
SUMMARY OF BENEFITS OF CONTROLLING AIR TOXICS IN BALTIMORE
COMPARISON OF BENEFIT METHOD AND COST-EFFECTIVE SCENARIOS
0\
i^^
CO
5
3
0
Q
1
IL
O
CO
z
o
3
I
4i -
20 -
19 -
18 -
17 -
16 -
15 -
14 -
13 -
12 -
11 -
10 -
9 -
w
8 -
7 -
6 -
5 -
4 -
3 -
2 -
1 -
+f
+
+ o
o +
o + a +
a
at-
a
+ + +
IIIIIMIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIMIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIlllllllll
2 4 6 8 10 12 14 16 18 20 22 24
PERCENT REDUCTION IN ANNUAL EXCESS CANCER INCIDENCE
a BENEFIT METHOD
CE METHOD
-------
FIGURE VI-6
SUMMARY OF NET ECONOMIC BENEFITS OF CONTROLLING AIR TOXICS IN BALTIMORE
COMPARISON OF BENEFITS METHOD AND COST-EFFECTIVE SCENARIOS
-------
The peak in Figure VI-6 is the point where the additional
benefits from control exceed the incremental costs by the
greatest amount. Additional reductions in risk cost more than
the cancer and noncancer benefits afforded by their
implementation. An examination of the net benefits column in
Table VI-8 shows which options can provide benefits that exceed
their implementation costs. Presentation of information in this
format allows decision-makers to consider the additional benefits
and costs of adopting each control option, and can help in the
efficient selection of controls where the costs borne by the
affected facilities may face constraints, or where a particular
level of risk reduction is a goal for the community. Even in
those instances where controls need to be undertaken to meet
required reductions in risk—and the costs of these controls
exceed the benefits they provide—the use of benefit information
can lead to more efficient uses of society's resources than
decisions made in the absence of this information. These
economic frameworks not only provide an opportunity to address
the multiple benefits of controlling toxic emissions, but also a
framework to compare the benefits from environmental controls
with other risk reduction actions that may be available.
VI-48
-------
APPENDIX A
GENERAL METHODOLOGY
FOR AN
INTEGRATED ENVIRONMENTAL MANAGEMENT PROJECT
-------
APPENDIX A. GENERAL METHODOLOGY
This Appendix provides a general overview of the analytical
steps which EPA employs in Phase I of an Integrated Environmental
Management Project (IEMP) in the assessment of risks. It also
defines the concepts of risk assessment and risk management and
describes the analytical methods used for risk assessment. By
and large, the Baltimore IEMP followed these steps, with some
variations.
APPLICATION OF METHODOLOGY
This section presents a comprehensive description of EPA's
general methodology for conducting an IEMP. When we actually
apply this approach, inevitable practical limitations (such as
time and resource constraints, or limits on the state of know-
•ledge) and the characteristics of the particular site force us
to tailor our efforts to our conditions. As a result, we do not
necessarily apply and develop the full framework and its analyti-
cal tools in any one IEMP. Moreover, we also believe that we
can achieve the most progress if we are flexible in applying the
approach and if we adapt to new circumstances as information
becomes available during the course of the analysis. Therefore,
the framework should not be considered to be a rigid blueprint;
.its specific application can vary from one study to another,
although the general approach will remain unchanged.
There are significant limitations and uncertainties associ-
ated with the lEMP's methodology, which warrant consideration be-
fore examining the actual procedures. First, the risk estimates
are based primarily on existing knowledge about pollutant potency,
releases, and ambient conditions; these data,- however, vary wide-
ly in quality and are almost always incomplete. Second, the ex-
posure estimates incorporate a series of simplifying assumptions;
although these assumptions are necessary, they they remain open
to question and may be controversial. Third, the potency esti-
mates are necessarily based on current knowledge of the toxicolo-
gical effects of various substances. Considerable controversy
exists about the degree of hazard posed by different pollutants,
and about whether some are hazardous at all. Finally, resource
and time constraints and the breadth of our focus prevent us from
analyzing individual issues in as much depth as might be possible.
We have attempted to strike a balance between the desire for
exhaustive and definitive analysis and the need for results at a
reasonable cost.
As noted at the beginning of this Appendix, the range of
potential environmental issues at any site is so large that it
A-l
-------
would be impossible, given limited resources, to conduct the in-
depth analysis that would be required to study all of the more
complex problems. Therefore, to manage an integrated study
effectively it is necessary to focus on a small set of study
topics. Although the screening process is designated as a single
step in Phase I, screening actually takes place continually
throughout the project as new information becomes available.
In its simplest form an IEMP consists of seven steps orga-
nized into two phases. An outline of the steps is provided below
and is followed by a more detailed discussion.
Phase I
1. Establishing arrangements between institutions cooper-
ating on the project
2. Defining the scope of the project:
--Setting geographic boundaries
—Making an initial selection of pollutants and issues
for study
—Establishing risk assessment approaches, e.g., selec-
ting the health effects of concern
3. Collecting information on sources, pollutants, and
exposure pathways for entry into a computerized data-
base
4. Performing a screening analysis on the initial selec-
tion of pollutants and sources to determine which
of those should receive further attention in Phase II.
The screen involves two complementary approaches:
—Evaluating risks to determine which pollutants,
sources, and exposure pathways are most significant
—Qualitatively assessing analytical feasibility; rele-
vance to EPA, state, and local program objectives;
and potential for effective response
Phase II
5. Gathering additional data to confirm and refine the
risk assessments performed in Phase I, and to adjust
priorities accordingly
6. Analyzing and evaluating the cost-effectiveness of
alternative control options
7. Developing conclusions
A-2
-------
allows like comparisons rather than having to confront the dilem-
ma of comparing different subjects for which there is no obvious
common denominator. This aspect of scoping as it applies to the
Baltimore IEMP is discussed in more detail in Chapter IV. [Some
issues can be placed into more than one reference category? toxic
contamination of fish, for example, can not only be an ecological
problem, but at the same time a public health threat.]
After issues have been grouped in reference categories,
project participants may decide to consider only one, or perhaps
just a subset, of these groupings of issues. Such was the case in
both Santa Clara and Philadelphia. At the outset of these pro-
ject participants decided to study only issues relating to human
health risks.
Developing a Database
The next step is to design a database that is appropriate to
the scope and objectives of the study. One approach is to gather
and inspect all readily available information on the issues
selected, i.e., sources, environmental releases, and exposed
populations for each medium. State, county, and city agencies,
permit writers, EPA, and local industrial facilities are the pri-
mary sources of these data, which should be collected and evalu-
ated before new data is generated.
After the available data have been reviewed, engineering
estimates, in many cases, can be used to fill the gaps and pro-
vide the needed information to compare issues. In almost all
cases, engineering estimates are needed to calculate intermedia
transfers. (Intermedia transfers are the relocation of pollu-
tants from one medium to another. For example, the incineration
of waste reduces — but does not eliminate. — the volume of
solid material requiring landfilling. However, the process may
also produce air-borne pollutants.) For certain common sources,
such as dry cleaners, degreasers, mobile sources (air), and non-
point run-off (water), EPA's Regulatory Integration Division in
the Office of Policy Analysis has developed algorithms to esti-
mate pollutant loadings to all relevant media that are easily
adapted to different geographic areas. In some situations, some
new monitoring may be initiated at this stage if it is deemed
necessary for setting priorities.
The Screening Process
The screening process selects and sets priorities for the
issues to be studied in Phase II. The process is emphatically
not one that can be performed in a mechanical way, but instead
relies on continuous evaluation of data, especially when compar-
ing potential risks.
A-4
-------
The screening process involves two complementary steps.
The first step quantifies risks to human health or the environ-
ment to identify the sources, pollutants, and exposure pathways
of greatest concern. This "risk screen" represents a preliminary
attempt at quantitative assessment, using available data, and
therefore may not generate results sufficiently complete to use
in setting priorities.
For that reason, we also use a second step which focuses on
issues that are not readily quantified. This approach qualita-
tively assesses an issue's analytical feasibility; its relevance
to EPA, state, and local program objectives; and its potential
for effective response (i.e., controllability).
Step 1; Risk Screen. The risk screen is a preliminary
quantitative assessment of the risks to human health or the envi-
ronment, performed to identify the sources, pollutants, and
exposure pathways of greatest concern. This assessment can be
made in a variety of ways, depending on the topic of interest.
For example, we can examine either the variety of sources that
contribute to pollution in one medium, or we can address the
sources of pollution in several media, each of which contributes
to the same health effect.
It may take several cuts at the entire set of issues within
a reference category to develop the best common denominator. Once
this common denominator is developed, a subset of issues will
become particularly significant. For example, if the study par-
ticipants determine that carcinogenic effects should be the
"common currency" of evaluation, those issues or conditions that
might be expected to cause greater levels of cancer will become
more important than those that might cause other types of health
consequences.
3ecause the focus of the IEMP analysis is for policy pur-
poses and due to inevitable resource constraints, the IEMP pro-
ject manager must reserve the bulk of the study resources * for
the later steps. For that reason, the risk estimates generated
for the screen are not precise estimates but rather very rough
approximations that are just suitable for setting priorities.
In fact, the information available to drive the risk screen is
not likely to be adequate even to identify the relative ranking
of issues with much certainty; instead, it serves only to group
the issues into three broad categories: (1) those that are likely
to pose high risk, (2) those with the potential for high risk
but little substantive evidence, and (3) those that are likely
to pose relatively low risk. Discretion must be used in deter-
mining when enough data have been gathered to support a success-
ful risk screen.
A-5
-------
Step 2; Qualitative Evaluation. After the issues identi-
fied in Phase I as potential candidates for Phase II analyses are
separated into the three groups based on quantitative measures of
risk, the issues classified in the first two groups (high risks)
are further evaluated using nonrisk or secondary criteria. These
secondary criteria, applied to the first two groups of high risk
issues, provide broader perspectives for setting priorities than
those based solely on risk assessment criteria. As more informa-
tion becomes available, further reclassification of the issues
may be warranted. The secondary criteria are discussed below.
Like the rest of the priority-setting process, they should not
be regarded as being inflexible. Other groups conducting inte-
grated studies may want to modify or add to these criteria. For
example, the Baltimore IEMP added an explicit criterion of avoid-
ing duplication of existing analyses of control programs.
Analytical Feasibility
The primary elements determining analytical feasibility are
the amount of supporting data available, the availability of
analytical methods, and the level of effort required to generate
new data. These are basically program management considerations
in that they indicate how much effort would be needed for basic
data gathering before developing management alternatives (e.g.,
for control). There is an important tradeoff implicit in the
criterion of analytical feasibility: the greatest payoff in
terms of identifying and controlling toxics may be in studying
issues that have not been studied in the past or have been avoid-
ed because of the complexity or magnitude of the issue. On the
other hand, within the time and budget available for the project,
it may not be feasible to characterize these issues adequately,
and it may be necessary to curtail the analysis short of develop-
ing complete risk management strategies.
Relevance to EPA, State, and Local Program Objectives
The elements of this criterion involve the significance of
the issue for national EPA programs and state and local interest
in the issue. One of our objectives is to use the IEMP projects
to indicate where shifts in EPA priorities and methods are appro-
priate, so the relationship of the issue to national ^PA programs
is an important factor. State and local interest affects the
extent of support that local and state participants will give to
the detailed study effort, and also strongly affects the feasi-
bility of implementing control strategies, when warranted.
Potential for Effective Response
The level of existing control provides a rough measure of
the likelihood that additional controls are cost-effective, since
cost-effectiveness generally declines with increasing control.
A-6
-------
The feasibility of additional controls is also important; if no
technological, institutional, and political means are available
to reduce risk, the issue is probably not a useful candidate for
study.
Risk Management Challenges in Interpreting Results and
Applying the Process.
In the process of establishing priorities, study partici-
pants are likely to encounter several basic problems involving
the interpretation and application of environmental risk informa-
tion. Some specific issues are discussed below.
Maximum Exposed Individual Risks Versus Total Population
Risks
The management or project advisory group selecting study
topics will have to confront one of the classic trade offs in
risk management--that of maximum exposed individual versus cumu-
lative population risks. Inevitably, some of the potential
study topics will involve situations where a relatively small
number of people experience risks higher than those faced by the
aggregate population; these topics do not necessarily coincide
with those where there are widespread risks. Decisions may have
to be made whether to spend study resources to help a few people
by a substantial margin or to spend them to help many people by a
modest margin. This issue is one of the most difficult issues in
setting priorities.
Comparing Across Effects
Another difficult issue involves the comparison of risks
across different health effects (e.g., kidney damage versus can-
cer). Typically, this is done on the basis of informed judgment.
RID is developing an approach to setting priorities that attempts
to account for the severity of different effects. There have
been several other attempts to scale different health effects
(for instance, EPA's Office of Solid Waste has developed such a
scale), but this issue may be most effectively resolved by con-
sidering local concern for the severity of different effects.
Comparing Human Health Effects to Environmental Effects
In cases where the scope of the study, as in the Baltimore
IEMP, includes human health as well as ecological issues, diffi-
cult value judgments must be made. These judgments are not
generally made within the same policy decision analysis because
the quantitative means of equating the two very different public
policy objectives are lacking. It is clearly important to protect
both public health and our ecosystems, but the value placed on
A-7
-------
human health is not the same as that placed on pur other environ-
mental goals. Decisions concerning the setting of priorities
between these two types of issues (because the quantitative
measures used to assess adverse impacts for each are not scienti-
fically or analytically comparable) involve the difficult and
necessarily subjective balancing of both public policy objec-
tives.
Phase I Product
This Report contains information about the risks we quanti-
fied, identifies priority issues based on our risk assessments,
and describes the methods we used to assess exposure and risks.
Furthermore, it describes the process by which these products
were made possible. In this regard, we elaborate on the institu-
tional arrangements necessary for the making of key policy deci-
sions and management judgements and the manner in which these
decisions were made.
Phase II; Risk Management Control Options
After Xfcfcselecting Phase I issues warranting further analy-
sis, we begin Phase II. Phase II consists of three major tasks:
additional data gathering to improve our exposure and risk
assessments, a scientific review of our risk assessments, and
analysis of control options.
Additional Data Gathering
By definition, in Phase I we identify the study topics for
further analysis using existing and special engineering esti-
mates. While this is reasonable for setting priorities, the
quality of the data becomes even more important at the later
stages of the study. Depending on the specific issue, in Phase
II we generally engage in further data gathering in order to:
(1) confirm whether an issue warrants further investment of time
and resources, (2) ensure that data used in later steps are as
accurate as possible, and (3) revise Phase I priorities where
appropriate.
In some cases, the objective of additional data collection
may be to improve our understanding of the significance of parti-
cular pollutants and sources. This situation may only call for
limited monitoring.
In other cases, which are really at the core of the Phase II
activities, the objective will be to analyze the costs of poten-
tial control alternatives, where effectiveness is measured by the
reduction in health risk — either Most Exposed Individual (MEl)
(defined on p. A. 15) risk or total population incidence. In these
A-8
-------
situations, we will need to ensure that our estimates of pollu-
tant loadings and ambient concentrations, for example, are rea-
sonable and representative of average annual conditions. These
efforts may include consulting local environmental agencies and
plant managers to confirm underlying data on ambient releases,
as well as initiating an extensive monitoring program.
As additional information is gathered in Phase II, it is
helpful to periodically re-evaluate the Phase II issues in terms
of potential risk. This can often be done by conducting a sensi-
tivity analysis using the revised data. New information may in-
dicate that the risks are not likely to be as great as once
thought, or they may be greater than originally anticipated.
Sensitivity analyses of new information provide a way to further
focus the use of project resources during Phase II.
Developing Pollution Control Options
The next step in Phase II is to assess alternative control
strategies for the subset of environmental issues identified for
this worK. A very significant objective is to provide analysis
of how to reduce risks to health (total population as well as to
the maximum exposed individual) or the environment (the measure
of effectiveness) at the minimum cost. This occurs through the
lEMP's cost-effectiveness analysis, which presents the tradeoffs
of costs and risk reductions that decision makers usually take
into consideration when formulating regulatory strategies, The
analysis considers risks that result from exposure to primary as
well as secondary (intermedia) releases. It is our intention
that community decision-makers use it to shape their general
strategies for providing additional environmental protection in
the area.
There are three ma^or components to an analysis of control
options: quantitative measures of ambient concentrations at the
point of exposure; estimates of exposure and risk; and the costs
and efficiencies of feasible control options. Data collected in
Phase I and supplemented in Phase II provide the necessary infor-
mation on pollutant loadings estimates. We generally employ EPA
fate and transport models to develop ambient concentrations at
the point of exposure. Exposure and risk calculations are then
made using standard EPA assumptions, which we discuss later
in this chapter.
We identify feasible control options and estimate their
associated costs and efficiencies by employing engineering
assessments and EPA technical documents developed to support
various regulatory activities.
The application of this process will inevitably vary from
one project to the next. In Philadelphia, we implemented the
A-9
-------
basics of this methodology. However, because Philadelphia was
our first project, we did not have experience to guide us in
assembling the steps in an efficient manner. Largely because of
lack of sophistication with the methodology, we chose not to
examine other issues which would involve ecological and non-
cancer health effects.
In the Baltimore IEMP, we have expanded the scope of the
project to include ecological impacts and have decided to examine
non-cancer health effects in Phase II. We will again be treading
new ground in developing analytic methodologies to examine eco-
logical impacts and the risk of non-cancer effects.
RISK ASSESSMENT AND RISK MANAGEMENT
The two key organizing concepts of an integrated environmen-
tal management project are risk assessment and risk management.
Risk assessment is the central task of Phase I; risk management
is the central task of Phase II. Conceptually, risk assessment
should be done independently of risk management.1 In IEMP
studies, we separate assessment from management as much as
possible; however, this is not always possible so that some of
the activities in Phase I and Phase II overlap.
We emphasize that our discussion of the IEMP methodology is
presented at a conceptual level; our actual application of the
methodology may vary from one geographic project to the next.
Furthermore, we may use different methods in response to the
needs of a particular IEMP. For example, project participants
in the Baltimore IEMP had identified issues pertaining to local
ecology and groundwater resources, in addition to health issues,
for study in Phase I. To assess the significance of these issues
and to decide which of them warranted further examination in
Phase II, we needed different analytical tools from those we used
in setting priorities among human health issues in Philadelphia
and the Santa Clara Valley. (Note that Philadelphia and Santa
Clara Valley are areas where we conducted earlier geographic
studies.) We describe these other (nonhealth related) methods in
Chapters VI (Analysis of sources with Potential Adverse Impacts on
Ground-water) and VII (Analysis of Ecological Impact) and discuss
only our methods for analyzing risks to human health in this
Appendix.
Phase I Risk Assessment and Priority-Setting
There are always more environmental issues to study than
resources with which to address them in a manner that will lead
to their successful resolution. A key task of Phase I of an IEMP
is, therefore, this: to choose which issues demand the immediate
A-10
-------
attention of environmental officials—and the expenditure of
their time and resources—and which can be addressed at a later
date. Such comparing of issues and setting of priorities re-
quires that all issues be reduced to a common element. (Note
that an integrated approach to all issues means we try to examine
the same issue across media. An IEMP is not a comprehensive look
at all the environmental issues of the study area.) Risk, either
to human health or the environment (or both), is the common ele-
ment we use for comparing environmental issues in geographic
studies. Risk can be categorized in several different ways. As
appropriate, we can examine risk associated with exposures from
a specific pollutant source. We can examine risks associated
with exposure to pollutants, irrespective of source. We can
look at risk associated with exposures in a single environmental
medium, such as air, water, or food. We can examine risks asso-
ciated with exposure pathways. In the absence of other impor-
tant considerations, the more significant the risk, the more
likely the issue will rank high in priority for examination in
Phase II, the risk management phase. Important considerations
that may alter the priority ranking include, for example,
resource limitations or the fact that a particular issue is
already being handled by the Federal, state of local government.
The risks we examine are associated with exposure to toxic
chemicals in the environment and represent estimates of the poten-
tial for these exposures to cause adverse effects. Chemicals can
adversely affect human health, and thus pose a risk, when people
are exposed to toxic levels in the air, water, food, or contami-
nated soil. Chemical exposures can adversely affect the environ-
ment by disrupting the ecology of surface waters or the dynamics
of biological communities. Furthermore, toxic chemicals can so
contaminate a natural resource that future generations can, for
all practical purposes, no longer use or enjoy it or must pay
greatly to do so.
In order to examine potential impacts to human health from
environmental exposures in a particular geographical location,
we perform a risk assessment. Based on the National Research
Council's recommendations,1 an IEMP risk assessment contains the
following elements:
o Hazard Identification
Does the agent cause the adverse effect?
For the chemicals of interest, RID toxicologists and con-
tractors prepare individual Profile reports that summarize
the information on animal and human health effects. Each
report examines both oral and inhalation exposure data for
the following ten health effect categories: cancer, liver,
kidney, reproductive, fetal developmental, neurobehavioral,
mutagenicity, respiratory, cardiovascular, and other. A
A-ll
-------
weight-of-evidence evaluation is provided for each health
effect category. For cancer, we use the EPA weight-of-
evidence evaluations that are prepared by the Agency's
Cancer Assessment Group (CAG) according to the 1986 cancer
risk assessment guidelines.3 EPA does not have weight-of-
evidence evaluation schemes for the other health effect
categories, so we use ranking schemes developed specifically
for our work. Our ranking schemes are currently being
reviewed by scientists within and outside EPA.
o Dose-Response Assessment
(What is the relationship between dose and incidence of
effect in humans?)
This information is also provided in the chemical profile
reports. With the exception of mutagenicity, a separate
evaluation is provided for each of the health effect cate-
gories. For cancer, we use the dose-response evaluations,
or "unit risk factors" that are prepared, by EPA's Cancer
Assessment Group (CAG). For the remaining non-cancer
health effect categories, EPA does not have a satisfac-
tory method for quantifying dose-response relationships.
Therefore, as an alternative we use the EPA Reference
Doses (RfDs) as benchmarks to estimate exposure levels
that are "safe" and for which we assume the incidence of
effects is insignificant.*
o Exposure Assessment
(What exposuresare currently experienced or anticipated
under different conditions?)
The IEMP studies include an assessment of environmental
exposure levels for selected chemicals in the media of
interest, usually air and water. Environmental exposures
are estimated primarily by modelling. As feasible, field
monitoring is conducted and the collected data are used
to validate the modelling. In addition estimates of
exposed populations are made by examining census data.
0 Risk Characterization
(What is the estimated incidence of the adverse effect in
a given population?)
For the environmental media of interest, we use our expo-
sure assessment to estimate both the concentrations of
*In so doing, IEMP follows EPA practice. However, other
authorities may use other methods. For example, for
airborne contaminants, Threshold Limit Values (TLVs)
may be more suitable under certain circumstances.
A-12
-------
specific chemicals and the number of people exposed at
those concentrations. We then link this information to
our chemical-specific dose-response assessments in order
to make rough estimates of the incidence of disease that
may occur above the baseline rate in the study area.
Note that because of limitations in our ability to
quantify dose-response relationships for each of the
eight health effect categories, we must analyze cancer
differently from the non-cancer health effects. For can-
cer, we treat the CAG unit risk factor as the slope of
the dose-response curve; therefore, we can use this value
to estimate a crude incidence. For the non-cancer health
effect categories, we do not have estimates of the slopes
of the dose-response curves. Instead, we have the RfDs,
which we use as estimates of exposure levels that are
"safe", in terms of non-cancer health risks. So for non-
cancer effects, we do not calculate the incidence of
effects. Rather, we calculate the number of people ex-
posed to levels above the RfDs in order to estimate how
many may be at some increased risk.
The risk characterization also includes a discussion of
the uncertainties in the hazard identification, dose-
response assessment, and exposure assessment. For the
hazard identification, this discussion focuses on the
weight-of-evidence evaluations. For the dose-response and
exposure assessments, the discussion focuses on uncertain-
ties in the databases and models used.
During Phase I of our first geographic project in Philadelphia
and then again in Baltimore, we also relied on priority-setting
methods other than quantitative risk assessment; e.g., the use of
expert judgement to establish relative ranking of issues.
In order to perform an IEMP risk assessment, as outlined
above, we rely heavily on approved EPA quantitative methods, CAG
un-.t risk values, RfDs and other appropriate standards and crite-
ria. In addition, we often use other approaches and criteria
to refine our priority ranking of issues. These criteria include
analytical feasibility, relevance to EPA, state, and local pro-
gram objectives, and the potential for effective response. These
criteria (which incorporate risk management perspectives) are
discussed in Chapter VIII of this report.
Bear in mind that the objectives of the IEMP in the use of
risk assessment procedures are to allow for relative comparisons
of potential problems and to provide a general sense of the re-
lative significance of an issue, rather than to make definitive
A-13
-------
statements about the absolute risks posed by a particular sub-
stance, source, or pathway. It must be emphasized that we conduct
risk assessments; we do not conduct epidemiological studies.
Thus, we do not, and cannot examine the disease rates in the
local population in order to correlate health effects with current
or past environmental exposures. Moreover, the exposure assess-
ments conducted as part of an IEMP are not designed to be used in
nor are they appropriate for an epidemiological study.
Although different in their objectives, design and interpre-
tation, risk assessments and epidemiological studies are actually
complementary. Risk assessment can help to identify populations
and geographic areas that appear to be at risk and therefore
might be appropriate subjects for an in-depth epidemiological
study. Epidemiological studies increase the scientific under-
standing of the relationship between exposure and health effects,
thereby strengthening the basis of risk assessment. Epidemiolo-
gical studies may also, in some cases, be useful for confirming
specific hypotheses suggested by risk assessment.
Other qualifications regarding the use of quantitative and
other health risk assessment methods are discussed later in this
appendix.
Phase II Risk Management
In Phase II, we will focus the resources of the project on
considerations of how best to manage risk in selected areas.
Risk management is the process of evaluating and selecting
approaches to reduce the risks identified through risk assess-
ment. Risk management considers not only the level of risk
posed by a particular pollutant or source of pollution but also
factors such as the feasibility and cost of control, public pre-
ferences, and institutional capabilities. Setting priorities
for research or government action is also a key aspect of risk
management. The objectives of this Phase are to investigate
control alternatives, estimate their efficiency in controlling
pollution, develop estimates of their cost, and then assess their
cost-effectiveness; in the Baltimore IEMP it also includes devel-
opment of analytical tools (i.e., a model for helping local govern-
ments establish priorities for addressing potential threats to
ground-water resources from underground storage tanks) as- well
as a research plan (the Harbor blueprint). Where necessary, we
supplement the analysis by developing detailed exposure estimates
and other specific data required to refine the quantification of
risks assessed in Phase I.
The criterion of cost-effectiveness (i.e., cost per change in
some unit of effectiveness) is generally expressed as the amount
of human health risk reduced by alternative controls in relation
A-14
-------
co che cost of implementing those controls. However, in some
cases, such comparisons cannot be made and other economic crite-
ria or definitions of cost-effectiveness must be used (e.g., the
costs of preventing damage to a resource).
Even where mechanisms are already in place for controlling
chemical exposures, an analysis of the cost-effectiveness of
these control strategies can be useful to local communities. An
analysis of the cost-effectiveness of control options may lead
to the identification of control strategies that attain the same
level of control for less cost.
GENERATION OF RISK ESTIMATES
The following overview of methods for quantifying health
risks theoretically applies to both cancer and non-cancer health
effects. In practice, appropriate quantitative methods have
only been developed for estimating carcinogenic risks. There-
fore, unless noted otherwise, the following discussion deals
only with estimation of carcinogenic risk to an exposed popula-
tion.
In IEMP studies, we calculate risk using three measures of
exposure: risk to the most exposed individual (MEI), risk to
the average exposed individual (AEI), and the excess aggregate
population incidence. We define risk to the MEI as the increased
probability that an individual chronically exposed to the highest
concentration of one or more chemicals will have exposure-related
cancer during the course of his or her lifetime. For the MEI,
the exposure is calculated as either the highest modelled concen-
tration or the average concentration obtained for the monitoring
site with the highest daily or annual average monitored value for
one or more chemicals. We define risk to the AEI as the increas-
ed probability that an individual exposed to the area-wide aver-
age concentration of one or more chemicals will have exposure-
related cancer during the course of his or her lifetime. Aggre-
gate population risk is the estimate of the increased incidence
of cancer, above the background rate, in an exposed population.
We use the standard EPA assumption that exposure is for 70 years.
Thus, elsewhere when we refer to "lifetime risk" or "70-year life-
time", this means 70 years of exposure. A more detailed discussion
of our assumptions and approach for calculating these risks is
provided later in this Appendix.
For a given population, the IEMP risk assessment involves
linking estimates of chemical potency, derived from animal and
human health effects data, with estimates or measurements of
area-wide contamination or exposure levels. A relatively high
degree of uncertainty unavoidably underlies this approach. In
A-15
-------
most cases, the risk estimates cannot easily be verified or dis-
proven by observation. The approach is useful, however, because
it is straightforward and allows application of the most current
scientific information to the analysis of adverse health effects
from current and projected exposures.
The two key elements in estimating health risk are: 1) the
calculation of the potency of a chemical, and 2) the determina-
tion of the level of exposure to that chemical.
1. Estimating Potency
A. Carcinogens
Potency is estimated from an analysis of relevant animal and
human (i.e., occupational, epidemiological) data concerning the
toxicity of a chemical. At EPA, data are obtained from the open
literature and from proprietary studies submitted to che Agency.
Chemical potency is determined by: 1) evaluating qualitatively
the preponderance (or weight) of evidence that a chemical causes
the adverse health effect in question; and, if so, (2) estimating
quantitatively the relationship between the dose (that is, the
•amount of chemical an animal or person is to exposed over a
'period of time) and the incidence of the effect within an exposed
•population of a given size. The latter "dose-response relation-
•ship" describes the potency of the chemical. That is, the lower
•the threshold and the steeper the slope of the dose-response
curve, the more potent the chemical. For assessing carcinogenic
risks, potency estimates can be used to relate the levels of ex-
posure to the probability that individual(s) will have cancer.
The EPA Cancer Assessment Group (CAG) calculates chemical-
specific cancer potency values called "unit risk" estimates. We
use these unit risk estimates in our IEMP studies and many Pro-
gram Offices throughout EPA also use these values for a variety
of risk assessment and risk management activities.
o Qualitative evaluation
The qualitative, or weight-of-evidence evaluation involves
reviewing the scientific literature, both human and animal,
to establish the sufficiency of evidence for a chemical's
carcinogenic potential in humans. This evaluation of the
weight-of-evidence for a causal association between the
chemical exposure and cancer is an important part of risk
assessment. It is necessary because the extent and quality
of chemical-specific databases rarely allow for an unambi-
guous determination of carcinogenic potential.
A-16
-------
Two evaluation schemes are available for assessing cancer
weight-of-evidence. One was developed by the International
Agency for Research on Cancer (IARC) and the other was developed
by EPA. The EPA scheme is a modification of the IARC system.
All chemicals evaluated for carcinogenicity by EPA's CAG are now
classified according to the EPA scheme.
Briefly, the IARC scheme has the following three categories:2
Group I; Known Human Carcinogen. This category is used
only when there is sufficient evidence from epidemiological
studies to support a causal association between the exposure and
cancer.
Group II; Probable Human Carcinogen. This category
includes exposures for which, at one extreme, the evidence of
human carcinogenicity is almost "sufficient" and at the other
extreme, it is "inadequate". To reflect this range, the category
is divided into Group IIA and Group I IB to indicate higher and
lower degrees of evidence, respectively.
Group III; This category includes exposures that can not be
classified as to their carcinogenicity in humans.
The EPA scheme has the following five categories; 3
Group A; Human Carcinogen. This category is used only when
there is sufficient evidence from epidemiological studies to
support a causal association between exposure to the agent and
cancer in humans.
Group B; Probable Human Carcinogen. This category includes
agents for which the evidence based on epidemiological studies
is "limited" and agents for which the evidence based on animal
studies is "sufficient". To reflect this range, the category is
divided into Group Bl and Group B2, indicating higher and lower
degrees of evidence, respectively. Group Bl has agents with
"limited" evidence from epidemiological studies. Group B2 has
agents with "sufficient" evidence from animal studies and "inade-
quate" evidence or "no data" from epidemiological studies.
Group C; Possible Human Carcinogen. This category is used
for agents with "limited" evidence of carcinogenicity in animals
and an absence of human data.
Group D; Not Classifiable as to Human Carcinogenicity. This
category is used for agents with "inadequate" human and animal
evidence of carcinogenicity or for which no data are available.
A-17
-------
Group E: Evidence of Non-Carcinoqenicity for Humans.
This category is used for agents that show no evidence of
carcinoqenicity in at least two adequate animal tests in different
soecies or in both adequate eoidemiological and animal studies.
This designation is based on the available evidence ind should
not be interpreted as a definitive conclusion that the agent will
••not be a carcinogen under any circumstances.
-Pr3sentlv, there is no similar scheme Cor assessing the weight-
'of-evidance for agents causing non-cancer health effects.3
0 Quantitative evaluation
In general, the EPA Carcinogen Assessment Group (CAG)
estimates the cancer potency of agents that are classified
Group B and higher. Potency is expressed as the "unit
risk". This unit risk estimate for an air or water oollu-
tant is defined as "the incremental Vifetime cancer risk
occurring in a hypothetical population in which all indi-
viduals are exposed continouslv from birth throughout
their lifetimes to a concentration of 1 uq/m3 of the agent
in the air they breathe, or to 1 uq/L in the water they
drink. -This calculation is done to estimate in quantita-
tive terms the impact of the agent as a carcinogen. Unit
risk estimates are used for two purposes: (1) to compare
the carcinogenic potency of several agents with each other,
and (2) to give a crude indication of the population risk
that might be associated with air or water exposure to
these agents, if the actual exposures are known."4
0 Key simplifying assumptions of the quantitative evaluation
In most cases, CAG uses a so-called linearized multistage
model to calculate chemical-specific cancer potency or
unit risk estimates. A mathematical model is needed in
order to calculate the unit risk factor because animal and
human carcinogenicity data are always associated with exoo-
sure levels that are much higher than those normally found
in environmental settinqs. Therefore, in order to esti-
mate potential cancer risk for the low exposure levels of
concern, it is necessary to extrapolate from data associ-
ated with high exposures. This is done through the use
of aporopriate models.
In explaining their choice of model, CAG notes the follow-
ing important considerations:
"The unit risk estimate represents an extrapolation
below the doss range of experimental data. There is cur-
rently no solid scientific basis for any mathematical
A-18
-------
extrapolation model that relates exposure to cancer risk
at the extremely low concentrations,... that must be dealt
with in evaluating environmental hazards. For practical
reasons the correspondingly low levels of risk cannot be
measured directly either by animal experiments or by epi-
demiological studies. Low-dose extrapolation must, there-
fore, be based on current understanding of the mechanisms
of carcinogenesis. At the present time the dominant view
of the carcinogenic process involves the concept that most
cancer-causing agents also cause irreversible damage to
DNA. This position is based in part on the fact that a
very large proportion of agents that cause cancer are
also mutagenic. There is reason to expect that the quan-
tal response that is characteristic of mutagenesis is
associated with a linear (at low doses) nonthreshold
dose-response relationship. Indeed, there is substantial
evidence from mutagenicity studies with both ionizing
radiation and a wide variety of chemicals that this type
of dose-response model is the appropriate one to use. This
is particularly true at the lower end of the dose-response
curve; at high doses, there can be an upward curvature,
probably reflecting the effects of multistage processes
on the mutagenic response. The low-dose linear nonthres-
hold dose-response relationship is also consistent with
the relatively few epidemiologic studies of cancer respon-
ses to specific agents that contain enough information to
make the evaluation possible (e.g., radiation-induced leu-
kemia, breast and thyroid cancer, skin cancer induced by
arsenic in drinking water, liver cancer induced by afla-
toxins in the diet). Some supporting evidence also exists
from animal experiments (e.g., the initiation stage of the
two-stage carcinogenesis model in rat liver and mouse skin).
Because its scientific basis, although limited, is the best
of any of the current mathematical models, the nonthreshold model,
which is linear at low doses, has been adopted as the primary
basis for risk extrapolation to low levels of the dose-response
relationship. The risk estimates made with such a model should
be regarded as conservative, representing a plausible upper limit
for the risk; i.e., the true risk is not likely to be higher than
the estimate, but it could be, and probably is, lower.
For several reasons, the risk estimate based on animal bioas-
says is only an approximate indication of the absolute risk in
populations exposed to known carcinogen concentrations. First,
there are important species differences in uptake, metabolism,
and organ distribution of carcinogens, as well as species diffe-
rences in target site susceptibility, immunological response,
hormone function, dietary factors, and disease. Second, the
concept of equivalent doses for humans compared to animals on a
A-19
-------
mg/surface area basis is virtually without experimental verifica-
tion as regards carcinogenic response. Finally, human popula-
tions are variable with respect to genetic constitution and
diet, living environment, activity patterns, and other cultural
factors.
The risk estimate can give a rough indication of the rela-
tive potency of a given agent compared with other carcinogens.
Such estimates are, of course, more reliable when the comparisons
are based on studies in which the test species, strain, sex,
and routes of exposure are similar.
The mathematical formulation chosen to describe the linear
(at low dose) nonthreshold dose-response relationship is the
linearized multistage model. This model employs enough arbitrary
constants to be able to fit almost any monotonically increasing
dose-response data, and it incorporates a procedure for estima-
ting the largest possible linear slope (in the 95% confidence
limit sense) at low extrapolated doses that is consistent with
the data at all dose levels of the experiment." 4
The EPA Carcinogen Risk Assessment Guidelines indicate the
following concerning the choice of extrapolation model:
"The Agency will review each assessment as to the evidence
on carcinogenesis mechanisms and other biological or statistical
extrapolation model. ...A rationale will be included to justify
the use of the chosen model. In the absence of adequate informa-
tion to the contrary, the linearized multistage procedure leads
to a plausible upper limit to the risk that is consistent with
some proposed mechanisms of carcinogenesis. Such an estimate,
however, does not necessarily give a realistic prediction of the
risk. The true value of the risk is unknown, and may be as low
as zero."3
B. Mutagens
In the IEMP studies, we restrict our evaluation of mutagen-
icity to a qualitative assessment. We simply indicate if the
chemical is considered to be a mutagen or if there is insuffi-
cient dat'a to make such a determination. Most of the experimen-
tal dates addressing muta'genicity involve in vitro testing of
mammalian cell cultures and non-mammalian organisms such as
bacteria. At present, we do not have methodologiess for using
these kinds of data to calculate dose-response curves amd quanti-
fy risk to humans. EPA assumes that a dose-response relation-
ship for mutagenicity, like carcinogenicity, would have no thres-
hold; therefore, any exposure Level would be associated with
some degree of risk. Evidence of mutagenicity often, but not
always, suggests carcinogenic potential.
A-20
-------
C. Non-Carcinogens
Chemicals that give rise to toxic endpoints other than cancer
and gene mutations are often referred to by EPA as "systemic
toxicants" because they affect the functioning of various organ
system, such as liver, kidney, cardiovascular, and respiratory
systems. Generally, based on our understanding of physiological
mechanisms, systemic toxicants are treated as if there is an
identifiable exposure threshold (both for the individual and for
the population) below which adverse effects are not observable.
From the viewpoint of risk assessment and risk management, this
characteristic distinguishes systemic toxicity from carcinogeni-
city and mutagenicity, since the latter two are usually treated
as non-threshold processes.
Because of this concept of identifiable thresholds, systemic
effects traditionally have been evaluated by EPA through the cal-
culation of "safe" exposure levels. This is unlike EPA's ap-
proach for evaluating carcinogens, for which all exposures are
assumed to involve some measure of risk. For many years, the
concepts of "acceptable daily intake" (ADI) and "margin of safe-
ty" have been at the heart of EPA's approach to risk assessment
for non-cancer health effects. Although there are limits to
some of these approaches, EPA is often called upon to apply these
concepts when making and explaining decisions concerning the
significance to human health of certain chemical exposures in the
environment. Thus, the threshold concept for non-cancer effects
is extremely important in the regulatory and risk management con-
text.
More recently, the Agency has tried to come to grips with
some of the technical and philosophical issues inherent in defin-
ing a "safe" exposure level for systemic toxicants. As an out-
growth of this effort, the concept of the "reference dose" (RfD)
has been recommended to replace that of the ADI. The RfD is
calculated in the same manner as an ADI; however, it is viewed
as a benchmark level without the value-laden connotations of
absolute safety or acceptability versus absolute unacceptability.
Thus, exposures that are less than the RfD are not likely to be
associated with non-cancer health effects and are therefore less
likely to be of regulatory concern. Conversely, as the frequency
of exposures exceeding the RfD increases and as the size of the
excess increases, the probability also increases that adverse
effects may be observed in a human population. Nonetheless, a
clear conclusion cannot be categorically drawn that all doses
below the RfD are "acceptable" and that all doses in excess of
the RfD are "unacceptable."^
To date, the Agency's calculation and use of ADIs and RfDs
has been applied only to oral exposure routes, such as drinking
A-21
-------
water and food. EPA has an established methodology for calculat-
ing oral RfDs and there are values for approximately 400 chemi-
cals. In contrast, the Agency does not have a methodology for
calculating similar benchmark levels for inhalation exposures.
EPA has not yet established a procedure for estimating such
benchmark values for chronic inhalation exposures; a methodology
is under development by a special committee in FY'87 and the
values will be called inhalation RfDs. Thus, with the exception
of the six or seven heavily studied criteria air pollutants
(ozone, carbon monoxide, sulfur oxides, etc.) there are no bench-
mark values for chronic inhalation exposures that are appropriate
for use in risk assessment. Because the Agency's RfDs are calcu-
lated for lifetime oral exposures only, and IEMP studies examine
both oral and inhalation exposure routes, it has been necessary
for us to estimate IEMP benchmark levels for inhalation exposures.
IEMP studies rely on EPA oral RfDs and on other "threshold"
values computed by our toxicologists and consultants. Our "thres-
holds" for various non-cancer effects have been calculated employ-
ing the same methodology used to estimate oral RfDs. Highlights
of the RfD methodology are as follows:
RfDs are calculated by collecting the available
animal and human data, noting the various dose
levels (in milligrams per Kg body weight per day)
at which different non-cancer health effects are
seen, and identifying the highest NOEL (No Observed
Effect Level). The highest NOEL represents the
highest dose at which biologically or statistically
significant effects were not seen. The scientists
also try to identify the LOEL (Lowest Observed
Effect Level), which is the lowest dose at which a
biologically or statistically significant non
cancer effect was seen.
Many different NOELs and LOELs can be identified
for each chemical, depending on the doses
selected, the spacing between the doses and the
health effects examined and reported by the
researchers. Toxicological research is very
expensive and each experiment cannot be exhaustive
in terms of the number of doses tested and the
number of effects studied. Similarily,
epidemiological research is expensive and subject
to many problems with both measurement of exposures
and identification of exposure-related effects.
This problem introduces a degree of uncertainty in
the identification of the "true" NOEL and LOEL of a
chemical and this uncertainty is carried through
the calculation of the RfD.
A-22
-------
The RfD is calculated by dividing the highest
reliable NOEL (or lowest reliable LOEL, if
appropriate) by uncertainty factors. The selection
of appropriate uncertainty factors is based,
primarily, on the nature of the study from which
the NOEL, or LOEL, was derived. Each factor has a
value from 1-10 and represents an extra degree of
uncertainty to account for: 1) extrapolation from
the average human to sensitive members of the human
population, 2) extrapolation from the average
animal to the average human, 3) extrapolation from
subchronic exposures to chronic exposures and 4)
extrapolation from the LOEL to the NOEL.5
By definition, the RfD is based on the critical non-cancer
health effect. This is the effect first seen as exposures in-
crease above zero. Since IEMP studies often examine exposures
that are above the RfD for a particular chemical, it would be
valuable to know what kind of effect(s) might be seen as expo-
sures get higher and higher above the RfD. This type of infor-
mation is not obtainable from the single RfD that is calculated
for a chemical. To help solve this problem, we have taken the
following approach in our IEMP studies:
1) non-cancer health effects are divided into the six
broad health effect categories of liver, kidney, reproduc-
tive, neurobehavioral, fetal developmental, and other. The
category "other" includes a variety of effects such as cardi
ovascular, respiratory, and gastrointestinal.
2) for each chemical of interest, examine both the oral
and inhalation data for humans and animals.
3) using the RfD methodology, and route-specific toxicity
data, we calculate a separate inhalation and oral "threshold
for each of the six health effect categories. Obviously, the
more toxic a chemical, and the more it has been studied, the
more health effect categories for which we can esti-
mate "thresholds".
2. ) Assessing the Level of Exposure
The second key element (see p. A-16) in estimating health risk
is determining the level of exposure to that chemical. Most of
our exposure assessments attempt to estimate ambient concentra-
tions of substances in the air and drinking water. We then make
certain assumptions about how ambient concentrations relate to
actual human exposure or dose level. For example, we take no
account of population mobility. These assumptions — the exposure
constants regarding how much air a person breathes or water he
or she drinks--are described below.
A-23
"
-------
We estimate ambient concentrations in two ways: direct moni-
toring or simulation modelling. There are a number of advantages
to the modelling of ambient concentrations over direct monitoring
alone:
o Modelling may provide the only way to estimate ambient
concentrations under alternative exposure scenarios.
o Modelling can take into account the geographic variabi-
lity of a large area. Because monitoring data are often
from a few specific points, they best serve as refe-
rence points for evaluating the performance of air dis-
persion models.
o Modelling is often less costly than extensive monitoring.
o Modelling enables predictions of exposures in any loca-
tion (and, in particular, the location of the most ex-
posed individual), whereas monitoring can only provide
an indication of the exposures in the vicinity of the
sampling sites.
o Modelling links concentration estimates, and hence expo-
sures, to sources. Such source information is important
as a risk management tool in that it allows us to esti-
mate the impact of various pollution control options.
o For some pollutants, there are no accepted methods for
monitoring them in ambient air.
On the other hand, constructing a model of pollutant releases
and resultant ambient concentrations involves making assumptions
about the important processes between pollutant source and human
receptor. Building such a model thus requires an understanding
of those processes, which is not necessary if one can simply
monitor ambient concentrations directly.
When carrying out an IEMP, we first employ existing data-
bases which have been reviewed for currency and quality. We
conduct monitoring as an adjunct as resources allow and use
monitoring data to verify ambient concentrations estimated from
our models. Under certain circumstances, such as situations
where we do not know sources and emissions, monitoring data are
crucial for estimating exposures. Furthermore, an advantage of
reliable, long-term monitoring data is that they provide a more
direct and often simpler means of estimating ambient conditions.
Even with direct monitoring, however, interpretation of results
can be difficult if data are limited or if lab or sample contami-
nation is suspected.
A-24
-------
The key steps in modelling exposures are:
(1) estimating the number of sources of emissions;
(2) gathering data on emissions (in g/sec.) of target
compounds from selected sources;
(3) gathering release specifications for chemical emis-
sions, such as building dimensions, stack height,
location of release points;
(4) characterizing the processes and pathways by which a
pollutant is transported in air, soil, and water,
including the speed of transport, the extent of dilu-
tion or dispersion, and any chemical transformation
the pollutant might undergo (such as degradation to
a nontoxic, or possibly even more toxic form).
(5) locating receptor sites; and
(6) assigning populations to receptor sites.
We try to estimate resulting ambient concentrations at
various distances from the source. Estimating long-term average
concentration levels, which are of concern for evaluating chronic
health impacts, is simpler than short-term modeling, which must
take greater account of variations in meteorological conditions.
RISK CHARACTERIZATION
As noted on pages A-12-13, risk characterization is the part
of risk assessment that brings together the exposure assessment
and the dose-response information in order to estimate the inci-
dence of an adverse effect in a given population. As further
noted, because of limitations in our ability to quantify dose-
response relationships for each of our health effect categories,
we must analyze cancer differently from the non-cancer health
effects. For cancer, we treat the CAG unit risk factor as the
slope of the dose-response curve; therefore, we can use this
value with the results of our exposure assessment to estimate a
crude incidence. For the non-cancer health effect categories,
we do not have estimates of the slopes of the dose-response
curves. Instead, we have RfDs and our own "thresholds" for each
of the effect categories. Therefore, for non-cancer, instead of
calculating the incidence of effects, we calculate the number of
people exposed to levels of chemicals that are above the RfDs and
other "thresholds" in order to estimate how many may be at some
increased health risk. For each chemical of interest, we examine
each of the six non-cancer health effect categories separately.
A-25
-------
If exposures to a substance are below the threshold for a
particular non-cancer health effect category, then we expect
little increased risk of effect from that substance. If expo-
sures exceed the estimated threshold for a particular health
effect category, we then indicate the possibility of increased
risk of that effect and try to estimate the size of the popula-
tion exposed to such concentrations. Exposures exceeding an
estimated threshold are generally more appropriate subjects for
• further investigation.
Note that the use of GAG potency estimates to evaluate can-
cer risk and the use of RfDs to identify exposures of potential
concern for non-cancer health effects are relatively straightfor-
ward practices and have been used throughout EPA for many years.
Risk Characterization as Applied to an IEMP
In IEM studies, we calculate cancer risk using three measures
of exposures: risk to the most exposed individual (MEI), risk
to the average exposed individual (AEI), and the excess aggregate
population incidence. For all calculations, we use the following
standard assumptions: 1) the average lifetime is 70 years, 2)
the average adult breathes 20 cubic meters of air per day, and
3) the average adult drinks 2 liters of water per day.
We express individual risk as either MEI risk or AEI risk.
We define risk to the MEI as the increased probability that an
.individual exposed to the highest concentration of one or more
chemicals will have exposure-related cancer during the course of
his or her lifetime. For the MEI, the exposure is calculated as
either the highest modelled concentration or the average concen-
tration obtained for the monitoring site with the highest daily
or annual average monitored value for one or more chemicals. We
define risk to the AEI as the increased probability that an indi-
vidual exposed to the areawide average concentration of one or
more chemicals will contract cancer during the course of his or
her lifetime. Aggregate population risk is the estimate of the
increased incidence of cancer, above the background rate, in an
exposed population.
To estimate the risk to the most exposed individual, we
•typically need to know how far an individual lives from the maxi-
mum pollutant concentration near a source. To estimate average
individual risk, we estimate an average pollutant concentration
to determine exposure. To estimate population risk, we must iden-
tify the number of people exposed to a given pollutant concentra-
tion. In our exposure estimates, we are assuming that people are
exposed to outdoor ambient air concentrations 24 hours a day for
a lifetime. In reality, people spend most of their time indoors,
either at home or at work. While our assumption overstates
A-26
-------
actual exposure to outdoor air, the bias introduced by this pro-
cedure may not be too great, since recent EPA (TEAM) studies
show that many outdoor air contaminants are also found indoors
at equal or greater concentrations. This represents current
standard practice at EPA.
Under these assumptions and exposure scenarios, lifetime
cancer risk to the exposed individual is simply the product of
exposure and potency:
R E x P (1)
individual risk exposure potency factor
As discussed above, exposure is the product of the ambient
concentration of the pollutant in the medium of concern (air or
drinking water) and exposure constants (i.e., the standard assump-
tions of body weight, breathing rate and water consumption):
E • Y x Z (2)
exposure ambient concentration exposure constants
in medium of concern
and
R = Y x Z x P (3)
individual ambient exposure potency
risk cone, in medium constants factor
of concern
SAMPLE CALCULATION
The following is a simple example to illustrate how we would
calculate the lifetime risk to the MEI, as well as the lifetime
risk to a population, associated with an ambient air exposure of
ten micrograms per cubic meter (ug/m3) of benzene. We have
divided the geographic area of concern into sections which we
will call "grids."
In our example, we have four grid sections. For each grid, we
have the following average ambient air concentrations and popula-
tion data:
Annual Average
Grid *
1
2
3
4
Benzene
Concentration
(ug/m3 )
10
8
8
5
Exposed
Population
20,000
30,000
25,000
25,000
A-27
-------
The MEI risk of cancer for an individual exposed to 10 ug/m3
of benzene in the air (the highest benzene exposure) is calculat-
ed by using equation (3). The ambient concentration (Y) is 10
ug/rH, which is equivalent to 0.01 milligrams per cubic -neter or
0.01 mg/m3. The potency factor (P) for inhaled benzene, devel-
oped by EPA's Carcinogenic Assessment Group (GAG), is 0.029
(mg/kg/day)-1. This value means that an individual inhaling one
milligram (mg) of benzene per kilogram body weight per day for a
lifetime has an estimated increased probability of developing
cancer of about three in one hundred (upper-bound estimate).
Using equation (3) and the maximum ambient concentration
value of 0.01 mg/m3, we calculate the risk to the MEI as follows:
R = Y x z x P (4)
MEI Avg. annual exposure potency
Risk ambient cone. constants factor
where, the exposure constants (Z) are adult body weight (70 kg.)
and adult breathing rate (20 m3/day)
therefore,
R = 0.01 mg/m3 x (20 n3/day x 1/70 kg) x 0.029 (mg/kg/day)-1
= 8.3 x 10-5
The lifetime upper-bound estimate of risk to the MEI (R) in this
example is 8.3 "x -10-5 or roughly eight chances in 100,000 of
developing cancer over a lifetime given constant exposure, to 0.01
mg/m3 per day.
The risk to the average exposed individual (AEI) is calcu-
lated the same way, using an average exposure determined from
either monitoring data or air dispersion models.
The final step in our example is to estimate the increased
incidence of cancer in the total population in our four grid
geographic area. To do this, we multiply the risk to the AEI in
each grid by the number of people in that grid. We then sum
these estimates of incidence across all grids to calculate the
lifetime aggregate incidence of cancer for our four grids.
Generally, EPA presents incidence as the expected number of
excess cancer cases per year. Dividing the upper-bound lifetime
estimate by 70, we arrive at the upper-bound estimate of the
annual average number of excess cancer cases in the population of
concern.
We estimate individual and population risks from ingestion
of drinking water in exactly the same way we estimate the risks
from inhalation. The only difference is that the potency factors
A-2 8
-------
have different values and we use the water consumption exposure
constant of 2 Liters/ day. Because the potency of a chemical may
differ for different routes of exposure, we must use the potency
factor that is appropriate for ingestion. This potency factor is
provided to us by the Carcinogen Assessment Group. Concentrations
of selected chemicals in the drinking water are measured or esti-
mated and then used with the oral potency factors in equation
(3) to estimate risk. We calculate both maximum individual risk
and aggregate excess incidence in a population in the same way
indicated in our four grid example.
Risk Characterization as Applied to the Baltimore IEMP
The Baltimore Phase I activities emphasized cancer as the
primary health effect of concern but some attention was given to
non-cancer effects of chemicals that were addressed as carcino-
gens. In Phase II, we will more broadly consider noncarcinogenic
effects to the extent that available data will allow. We empha-
sized cancer in Phase I because the public has expressed frequent
concern about the possible link between exposure to environmental
pollution and the incidence of cancer Also, the quantitative risk
assessment methodology for cancer has more scientific credibility
within EPA than do similar methods for assessing non-cancer
effects.
INTERPRETING RISK ASSESSMENT RESULTS
The estimates of individual health risk and aggregate inc-
idence from exposure to toxics should not be interpreted as pre-
cise or absolute estimates of future health effects. The simpli-
fying assumptions and uncertainties in both the toxicology and
exposure components are simply too great to justify a high level
of confidence in the precision of the results.
The potency and the threshold estimates used in this study
are consistently conservative in the direction of overestimating
risk; they may overestimate the likely effects of chemical expo-
sure but are unlikely to underestimate them. Such an estimate,
however, does not necessarily give a realistic prediction of the
risk. The true value of the risk is unknown, and may be as low
as zero." This leads to assessments that have biases for health
protection when the results are used. By contrast, our exposure
estimates are not as clearly conservative; some assumptions are
conservative while others are our best guess of an actual value.
Overall, we have tried to be somewhat conservative in our expo-
sure assessments, as is appropriate in a priority-setting exer-
cise. It is important to read the detailed chapters carefully to
understand what confidence to place in a particular risk estimate.
A-29
-------
On the other hand, we may understate risks to the extent
that we do not estimate exposure and risks from all sources and
pollutants that may have toxic effects. We have tried to identi-
fy the sources and pollutants of greatest concern and address
those that data and methods allow us to cover. We acknowledge
that we have not addressed all sources and pollutants.
Because of the uncertainties involved, the results should
not be interpreted too literally. For example, one should not
conclude that a source projected to cause three cases of cancer
per year is clearly worse than a source projected to cause two
cases; given the overall uncertainty of the analysis, these two
results are indistinguishable. On the other hand, for example,
it is reasonable to conclude that a source projected to result in
one case per year represents less risk to the population than a
source projected to result in ten cases per year. Despite the
uncertainties, our risk estimates are useful for roughly assess-
ing the potential magnitude of the overall risks from particular
pollutants, sources, and pathways; comparing issues with one
another; and setting priorities among environmental issues and
concerns.
A-30
-------
REFERENCES FOR APPENDIX A
1. National Research Council (1983). Risk Assessment in the
Federal government: Managing the Process. Committee on the
Institutional Means for Assessment of Risks to Public Health,
Commission on Life Sciences, national Academy of Sciences.
Rational Academy Press, Washington, D.C.
2. International Agency for Research on Cancer (1982). IARC
MONOGRAPHS on the EVALUATION OF THE CARCINOGENIC RISK OF CHEMI-
CALS TO HUMANS. World Health Organization, Volumes 1 to 29,
Supplement 4, pp. 13, 14.
3. U.S. Environmental Protection Agency (1986). Guidelines for
Carcingenic Risk Assessment. Federa'l Register, Vol. 51, Mo. 135,
33993 - 34003.
4. U.S. Environmental Protection Agency (1985). Health Assess-
ment Document for 1,2-Dichloroethane (Ethylene Dichloride).
Office of Health and Environmental Assessment; Washington, D.C.
pp. 9-235 to 9-241.
5. U.S. Environmental Protection Agency (1986). Reference Doses
(RfD): Description and Use in Health Risk Assessments. Draft
prepared by the RfD Workgroup for the Risk Assessment Forum,
May 1986.
A-31
-------
APPENDIX B
SUMMARY OF AMBIENT AIR MONITORING
PROVIDED BY THE MARYLAND AMA
-------
MONITORING SITES
BALTIMORE AND ANNE ARUNDEL COUNTIES
KEY SITE NAME
CG Coast Guard
0 Dundalk
P Parkville
20 Glen Burnie
23- Riviera Beach
28 Essex
32 Dundalk
33 Chesapeake Terrace Elementary
-------
FIG. 1 - MONITOR SITES - BALTIMORE AND ANNE ARUNDEL COUNTIES
-------
MONITORING SITES
BALTIMORE CITY
KEY SITE NAME
BM Baltimore Museum of Industry
BP Baltimore Polytechnic Institute
LF Lacy's Foundry
P7 Pier 7
SG Spring Gardens
35 Fire Dept. Headquarters
36 Old Town Fire Station
37 Allegany Pepsi
38 N. E. Police Station
40 S. E. Police Station
41 S. W. Police Station
42 Sun Sreet
46 Holabird Elementry School
47 Canton Pier II
48 Guilford
50 Fort McHenry
54 1-95 & Moravia Road
-------
FIG. 2 MONITOR SITES - BALTIMORE CITY
'HERRING .GaraenvTiie :j ,
«-•> I
to»J. Park,.0!ll, 'il^'^LS
n I ' ^—S '^~^T««,« v. t
^^^^^2^M^"
Fairficlcl
^ 'V-Curlis Qny
— i • , %i p
-------
DOCUMENTATION OF DATA (Cont'd)
Fire Dept. - Headquarters
Old Town Fire Station
Allegany Pepsi
N. E. Police Station
S. £. Police Station
S. W. Police Station
Sun Street
Holabird Elementary
Canton Pier II
Guilford
Fort McHenry
1-95 and Moravia Road
Maryland Air Management, 1986 Maryland Air
Quality Data Report, 1986, p 21.
Maryland Air Management, 1986 Summary of
Baltimore, Maryland Vinyl Chloride Air Sampling,
September - October 1986, p. 3.
Maryland Air Management, 1986 Maryland Air
Quality Data Report, 1986, p. 21.
EPA, 1987, Baltimore Integrated Environmental
Management Project, Phase I Report, Table V-3
Maryland Air Management, 1986 Maryland Air
Quality Data Report, 1986, p. 22
Maryland Air Management, 1986 Maryland Air
Quality Data Report, 1986, p. 22
EPA, 1987 Baltimore Integrated Environmental
Management Project, Phase I Report, Table V-3
Maryland Air Management, 1986 Summary of
Baltimore, Maryland Vinyl Chloride In Air
Sampling, September - October 1986, p. 3.
Maryland Air Management, 1986 Summary of
Baltimore, Maryland Hydrocarbon and Chlorinated
Hydrocarbon Air Sampling February - September,
1986, Table 2
EPA, 1987 Baltimore Integrated Environmental
Management Project, Phase I Report, Table V-3
EPA, 1987 Baltimore Integrate Environemtnal
Management Project, Phase I Report, Table V-3
Maryland Air Management, 1985, Maryland Air
Quality Data Report, 1985, p. 32
EPA, 1987 Baltimore Integrated Environmental
Management Project, Phase I Report, Table V-3
Maryland Air Management, 1986 Maryland Air
Quality Data Report, 1986, p. 21
EPA, 1987 Baltimore Integrated Environmental
Management Project, Phase I Report, Table V-3
Maryland Air Management, 1986 Maryland Air
Quality Data report, 1986 p. 21
-------
DOCDMENTION OF DATA
Baltimore Museum of Industry
Baltimore Polytechnic Institute
Coast Guard
Dundalk
Lacy's Foundry
Parkville
Pier - 7
Spring Gardens
Glen Burnie
Riviera Beach
Essex
Chesapeake Terrace Elementary
Maryland Air Management, 1987 Determination of
Ambient Background Concentrations of Total
Suspended Paniculate and Chromium in the
Vicinity of the Allied Chemical Plant, Table 2.
Conney, Maryland Air Management 1986 sorbent
tube sampling at Polytechnic Institute for acetic
acid and other volatile organics, Table 1
EPA, 1987, Baltimore Integrated Environmental
Management Project, Phase I Report, Table V-3.
EPA, 1987 Baltimore Integrated Environmental
Management Project, Phase I Report, Table V-3.
EPA, 1988 Baltimore Integrated Environmental
Management Project. Phase n Draft Final Report,
Ambient Air Toxics, Table IV-9.
Maryland Air Management, 1987 Determination of
Ambient Background Concentrations of Total
Suspended Particulate and Chromium in the
Vicinity of the Allied Chemical Plant, Table 2.
EPA 1988 Baltimore Integrated Environmental
Management Project, Phase II Draft Final Report,
Ambient Air Toxics, Table IV-9
Maryland Air Management, 1987 Determination of
Ambient Background Concentrations of Total
Suspended Particlate and Chromium in the
Vicinity of the Allied Chemical Plant, Table 2.
O'Melia, 1987 Sampling Results from BGdcE Spring
Gardens facility.
Maryland Air Managment, 1986, Maryland Air
Quality Data Report, 1986, p 20.
EPA, 1987 Baltimore Integrated Environmental
Management Project, Phase I Report, Table V-3
Maryland Air Management, 1986 Maryland Air
Quality Data Report, 1986, p 22.
EPA, 1987 Baltimore Integrated Environmental
Management Project, Phase I Report, Table V-3
-------
AVERAGE MBASUKKU AMUIENT CONCENTRATIONS: VOLATILE OKGANICS (ug/m3)
CHEMICAL BN[ BP CGD LF£ P7SG2023283233353637
ARSENIC .0017 .0016 .0016
BENZENE 7.8 3.9 5.2 10.3 12.6 5.5
WiNZO-A-PYRENE .002 .92 .92 .95
CADMIUM .001 .001 .001
CARUONTETRACIILORIDE .7 .2 1.1 .6 .9 .9
CHLOROFORM .6 .1 ND I.I |. 3.2
CHROMIUM .002 .018 .004 .007 .007 .007
1,2 IJICI1LOUOETIIANE 1.0 .2 .4 2.6
1,2 DICIILOROPROPANE .4 .2 .3 .1
ETHYL BENZENE 5.2 6.3 4.7 2.9
LEAD .058 .060 .084 .074
METHYLI-NE CHLORIDE 4.l->
MI-THYL ISOUUTYL KETONE
PEHCIILOROETIIYLENE 2.9 .6 2.1 2.4 9.3 1.5
TOLUENE 5.4 5.5 7.4 9.3 7.5
1,1,1 TRICIILOROETIIANE 1.5 5.9
TRICIILOROETHYLENE .2 .4 .01 .3 1.4 .4
VINYL CHLORIDE ND NL>
XYLENES 8.35 13.6 5.5 15.4 19.2 13.7 5.7
-------
AVERAGE MEASURED AMBIENT CONCENTRATIONS: VOLATILE OKGANICS (ug/m3)
CHEMICAL
ARSENIC
DENZENE
HENZO-A-PYRENE
CADMIUM
CAUDONTETRACHLORIDE
CIII.OROFROM
CHROMIUM
1.2 DICULOUETIIANE
1,2 DICHLOROPROHANE
ETHYL BENZENE
I,I-AO
METIIYI.ENE CHLORIDE
METHYL ISO8UTYI, KETONE
PERCIILOROETIIYLENE
TUI.UKNE
1,1,1 TKICIILORETUANE
TIUCtll.OUOl-THYLENIi
VINYL CHLORIDE
XYLENES
38 40
9.5
.96
.001
.9
.4
.009 .018
.2
.2
6.4
.111 .143
5.4
9.0
.5
21.1
41
10.2
.6
.2
.011
.3
.2
8.0
.057
7.0
9.9
.5
20.8
42 46
12.1 12.9
1.2 1.3
2.6 .7
.7 .5
2.0 .7
6.2 7.4
.7
.3
3.2 6.0
7.6 16
.6
l.l 3.9
ND
17.2 20.1
NO. of
47 48 SO 54 Sites
3
12.0 10.6 12
5
.002 5
I.I 1.4 12
2.1 4.7 12
.016 10
.2 .2 10
.3 .4 10
9.3 5.7 10
.079 .059 .081 10
2
1
4.8 3.9 12
16.4 4.8 1 1
3
.9 1.0 12
3
30.4 16.4 13
Total
.0049
112.6
3.752
.006
10.9
16.7
.099
6.3
4.8
62. t
.806
4.8
.3
49.1
98.8
8
10.61
-
207.45
AVG. COMENTS
.00163
9.38
.75
.0012
.91
1.39 ND: NO DETECTABLE LEVEL
.01
.63
.48
6.2
.081
2.4
.3
4.1
8.98
2.7
.88
ND: NO DETECTABLE LEVEL
15.90
-------
APPENDIX C
SUMMARY OF CONTROL OPTIONS COST AND
REMOVAL EFFICIENCIES
-------
TABLE C-1 DESCRIPTION Of CONTROL OPTIONS for POINT SOURCES
Source
Point Source A
Plant Area
A&B Plants
Benzene/Litol
Plani
Storage Tanks
Mills
Release Pt.
Tar decanter, tar intercepting
sump, flushing liquor
circulation tank
Tar storage tanks, tar
dewatering tanks
Excess ammonia liquor
storage tanks
Light oil condenser
Light oil sump
Benzene condensor
Light oil benzene mixture
storage tanks
Benzene storage tanks
Chrome plating line
Option/Alternative
Current Control
Alternative 1
Current Control
Alternative 2
Alternative 3
Cui rent Control
Alternative 4
Alternative 5
Onrrnni Conjrol
.-:6
Current Control
Alternative 7
Current Control
Alternative 8
Current Control
Alternative 9
Alternative 10
Current Control
Alternative 11
Alternative 12
Current control
Alternative 13
Option Description
Uncontrolled
Coke oven gas blanketing system
Uncontrolled
Wash Oil Scrubber
Coke oven gas blanketing system
Uncontrolled
Wash Oil Scrubber
Coke oven gas blanketing system
Uncontrolled
Coke oven gas blanketing system
Uncontrolled
Cover
Uncontrolled
Coke oven gas blanketing system
Uncontrolled
Wash Oil Scrubber
Coke oven gas blanketing system
Uncontrolled
Wash Oil Scrubber
Coke oven gas blanketing system
Packed bed wet scrubber
Packed bed wet scrubber
Efficiency
97
90
97.9
90
97.7
97.9
98.2
97.9
89.9
98.2
90
97.9
50
-------
TABLE C-1 DESCRIPTION Of CONTROL OPTIONS for POINT SOURCES (continued)
Source
Point Source A
(continued)
Plant Area
Coke Ovens
Release Pt.
11 and 12 Batteries
A Battery
Option/Alternative
Current Control
Option 1
Option 2
Current Control
Option Description
Meets proposed NESHAP limitations for
coke oven doors, I.e., 10% leaking
doors, for 75% control. Charges are
currently every 32 seconds for 97%
control. Four percent of lids leak.
Construct two new batteries to replace
the existing batteries. The new
batteries wouild be designed to meet
the propoosed NESHAP.
Modify charging to 16 second intervals.
for 98% control. Limit leaking lids to 3%
of topside ports through maintenance
and inspection
Meets proposed NESHAP
Efficiency
"
1.3
13
—
-------
TABLE C-1 DESCRIPTION of CONTROL OPTIONS for POINT SOURCES (continued)
Source
Point Source B
Point Source C
Point Source D
Plant Area
Release Pt.
Electric arc furnace (EAF)
Argon-oxygen decarburization
vessel (AOD)
Electric arc furnace (EAF)
Argon-oxygen decarburization
vessel (AOD)
EAF and AOD
Process Line
Option/Alternative
Current Control
Option 1
Current Control
Option 2
Current Control
Option 1
Current Control
Optbn 2
Option 3
Current Control
Option 1
Option Description
Direct evacuation control, segmented
canopy hood
Scavenger ducts, cross draft partitions,
closed roof
Close Fitting Hoods
Segmented canopy hood, scavenger
ducts, cross draft partitions, and
closed roof
Segmented canopy hood, closed roof
Direct evacuation control, scavenger
ducts, and cross draft partitions
Close Fitting Hoods
Segmented canopy hood, scavenger
ducts, and cross draft partitions
Vinyl strip doors to block wind
circulation
Horizontal packed bed wet scrubber,
venturi scrubber
Second packed bed wet scrubber
Efficiency
99%
99%
99%
99%
50%
40%
* Control alternatives for the A&B Plant. Benzene/Litol Plant, and Storage Tanks were grouped into a series of Options as shown in Table vi-6
The selection of control options and the estimation of capital and annual costs and removal efficiencies are projections and estimations based
on discussions with AM A staff, literature review and engineering judgement. Actual costs as well as feasibility and effectiveness could differ
considerably from our estimates.
-------
TABLE C-2 DESCRIPTION of CONTROL OPTIONS for AREA SOURCES
Area Source
Current
Control
Option 1
Option 2
Option 3
DRY CLEANING
None
Quickly identify and repair leaks.
Degenerative filter wastes not to
exceed 25kg of solvent per 100 kg
of wet waste material.
Distillation wastes not to exceed
60 kg of solvent per 100 kg of wet
waste materials.
Better housekeeping
Overall control efficiency - 20%
Quickly identify and repair leaks
Regenerative filter wastes not to
exceed 25kg of solvent per 100 kg
of wet waste material
Distillation wastes not to exceed
60 kg of solvent per 100 kg of wet
waste materials
Better housekeeping
Carbon adsorption systems for all
commercial dry cleaners
Overall control efficiency - 58%
Quickly identify and repair leaks
Regenerative filter wastes not to
exceed 25kg of solvent per 100 kg
of wet waste material
Distillation wastes not to exceed
60 kg of solvent per 100 kg of wet
waste materials
Better housekeeping
Carbon adsorption systems for all
coin-op and commercial dry cleaners
Overall control efficiency - 65%
DECREASING
None
Cold Cleaners
Cover during idle time (control
efficiency of 90%)
Drain racks with 30 second drain
and control efficiecy of 50%
during operation of degreaser
Open Top Vapor Degroasers
Cover during ida time (control
efficiency of 90%)
Increase freeboard ratio to 0.75
during operation (control
efficiency of 27%)
Cold Cleaners
Cover during idle time (control
efficiency of 90%)
Drain racks with 30 second drain
and control efficiecy of 50%
ddurmg operation of degreaser
Open Top Vapor Degreasers
Cover during ide time (control
efficiency of 90%)
Use refrigerated freeboard device
(control efficiency of 60% for
vaporization losses and 29% of
carryout losses)
Cold Cleaners
Cover during idle time (control
efficiency of 90%)
Dram racks with 30 second drain
and control efficiecy of 50%
dduring operation of degreaser
Open Top Vapor Degreasers
Cover during ide time (control
efficiency of 90%)
Use carbon adsorption unit
(control efficiency of 70% for
vaporization losses and 30% of
carryout losses)
-------
TABLE C-2 DESCRIPTION of CONTROL OPTIONS for AREA SOURCES (continued)
Area Source
Current
Control
Option 1
Option 2
Option 3
OTHER INDUSTRIAL
None
Enlarged condensation zone
Waste recovery facility
Manual enclosure
Overall control efficiency: 21%
Enlarged condensation zone
Waste recovery facility
Automatic enclosure
Overall control efficiency: 32.8%
Enlarged condensation zone
Waste recovery facility
Refrigerated condenser
Overall control efficiency: 36.6%
GAS MARKETING
Stage I
Stage II (vapor balance system at pumps)
Overall control efficiency: 70.5%
ROAD VEHICLES
Light Duty Diesel
Road Vehicles
Heavy Duty Diesel
Road Vehicles
None
None
Catalytic Converter
Catalytic Converter
-------
TABLE C-2 DESCRIPTION of CONTROL OPTIONS for AREA SOURCES (continued)
Area Source
HEATING
Residential
Wood Stoves
Commercial Oil
Heating
Residential Oil
Heating
Residential Coal
Heating
Current
Control
None
None
None
None*
Option 1
Stop using wood stoves. Instead
use main fossil fuel heating system
(gas,coal or oil)
Replace oil-fired furnaces with
natural gas systems
Replace oil-f iredfurnaces with
natural gas sytems
Replace coal-fired units with
natural gas systems
Option 2
Replace existing burner with dual
burner and burn natural gas
Option 3
* The burning of coal is illegal, but does occur
The selection of control options and the estimation of capital and annual costs and removal efficiencies are projections and estimations based
on discussions with AMA staff, literature review and engineering judgement. Actual costs as well as feasibility and effectiveness could differ
considerably from our estimates.
-------
TABLE C-3
SUMMARY OF CONTROL OPTIONS FOR POINT SOURCE A
Benzene Sources
Capital Costs ($)
Annual Costs ($/yr)
TSP Emissions (kkg/yr)
VOC Emissions (kkg/yr)
Benzene Emissions (kkg/yr)
Toluene Emissions (kkg/yr)
Xylene Emissions (kkg/yr)
Percent Benzene Emitted
Percent Toluene Emitted
Percent Xylene Emitted
Percent VOC Emitted
Chrome Plating (Mills)
Current Control Option 1 Option 2 Option 3 Option 4 Option 5
664
565
2.0
0.9
100
100
100
100
Current Control Option 1
770.000
158,000
16
13
0.05
0.02
2.3
2.5
2.2
2.3
410.000
64.000
142
117
0.4
0.6
20.7
20.0
66.7
21.4
200.000
4.000
365
291
1.0
0.7
51.5
50.0
77.8
55.0
150.000
54.000
583
497
1.8
0.3
88.0
90.0
33.3
87.8
210.000
60.000
441
391
1.5
0.8
69.2
75.0
88.9
66.4
Capital Costs ($)
Annual Costs ($/yr)
TSP Emissions (kkg/yr)
VOC Emissions (kkg/yr)
Chromium-6 Emissions (kkg/yr)
Percent Emitted
Coke Ovens
Capital Costs ($)
Annual Costs ($/yr)
TSP Emissions (kkg/yr)
VOC Emissions (kkg/yr)
POMCO Emissions (kkg/yr)
Percent Emitted
—
—
—
—
1
100
Current Control
—
—
864
—
864
100
460.000
200.000
—
—
0.5
50
Option 1
238,000,000
243,000
787
—
787
91
Option 2
433,000
243.000
857
—
857
99
The selection of control options and the estimation of capital and annual costs and removal efficiencies are projections and
estimations based on discussions with MA A staff, literature reviews and engineering judgement. Actual costs as well as
feasibility and effectiveness could differ considerably from our estimates.
-------
TABLE C-4
DESCRIPTION Of POINT SOURCE A BENZENE CONTROL OPTIONS
Benzene Control
Control Options Alternatives
(see Table VI-4)
Current Control
#1 1,3,5,6,7,8,10,12
#2 1,6,8
#3 6,8
#4 4.7,9.12
#5 1
The selection of control options and the estimation of capital and annual costs and
removal efficiencies are projections and estimations based on discussions with AMA
staff, literature review and engineering judgement. Actual costs as well as feasibility
and effectiveness could differ considerably from our estimates.
-------
TABLE C-5 SUMMARY of CONTROL OPTIONS for POINT SOURCE B
Current Control Option 1 Option 2 Options
(Alternative 1) (Alternative 2) (Alternatives 1&2)
Capital Costs ($)
Annual Costs ($/yr)
TSP Emissions (kkg/yr)
VOC Emissions (kkg/yr)
Chromium-6 Emissions (kkg/yr)
Nickel Emissions (kkg/yr)
Cadmium Emissions (kkg/yr)
Arsenic Emissions (kkg/yr)
Percent TSP'emittad
Percent Chromium emitted
Percent Nickel emitted
Percent Cadmium emitted
Percent Arsenic emitted
193E-01
2.25E+00
100E-01
100E-01
70.000
20.000
191E-01
1 49E+00
109E-02
1.09E-02
51
98
66
11
11
240.000
60,000
1.35E-02
931E-01
1 OOE-01
1 OOE-01
57
7
4
100
100
310.000
80.000
7
1 17E-02
1.78E-01
109E-02
109E-02
9
6
B
11
11
The selection of control options and the estimation of capital and annual costs and removal efficiencies are projections and
estimations based on discussions with AMA staff, literature review and engineering judgement Actual costs as well as
feasibility and effectiveness could differ considerably from our estimates.
TABLE C-6 SUMMARY of CONTROL OPTIONS for POINT SOURCE C
Capital Costs ($)
Annual Costs ($/yr)
TSP Emissions (kkg/yr)
VOC Emissions (kkg/yr)
Chromium-6 Emissions (kkg/yr)
Nickel Emissions (kkg/yr)
Arsenic Emissions (kkg/yr)
Percent TSP emitted
Percent Chromium-6 emitted
Percent Nickel emitted
Percent Arsenic emitted
Current Control
247
2.82E-01
382E+00
6.50E-02
Option 1
(Alternative 1)
120.000
30.000
239
282E-01
379E+00
608E-02
97
100
99
94
Option 2
(Alternative 2)
160.000
40,000
240
2.68E-01
364E+00
6 50E-02
97
95
96
100
Options
(Alternatives 1&2)
280.000
70,000
232
266E-01
3.61E+00
608E-02
94
95
94
94
Opflon4
(Alternative 3)
5.000
200
230
2.73E-01
3.69E+00
5 98E-02
93
97
97
92
The selection of control options and the estimation of capital and annual costs and removal efficiencies are projections and
estimations based on discussions with AMA staff, literature review and engineenng ludgement. Actual costs as well as
feasibility and effectiveness could differ considerably from our estimates.
-------
TABLE C-7 SUMMARY Of CONTROL OPTIONS for POINT SOURCE D
CURRENT CONTROL OPTION 1
CAPITAL COST ($) — —
ANNUAL COST ($/YR) — 220,000
CHROMIUM-6TOAIR(kko/yr) 0.19 0.11
TSPTOAIR(kkg/yr) 2.36 1.42
PERCENT EMITTED 100 60
7779 selection of control options and the estimation of capital and annual costs and
removal efficiencies are projections and estimations based on discussions with AMA
staff, literature review and engineering judgement. Actual costs as well as
feasibility and effectiveness could differ considerably from our estimates.
-------
TABLE C-8 SUMMARY Of DECREASING CONTROL OPTIONS
PERCHLOROETHYLENE DECREASING
CURRENT OPTION 1 OPTION 2 OPTIONS
CONTROL
CAPITAL COSTS ($)
ANNUAL COSTS ($/yr)
PERCENT EMITTED
PERCLOROETHYLENE TO AIR (kkg/yr)
VOC TO AIR (kkg/yr)
TSPTOAIR(kkg/yr)
100
315
315
60,000
(50,000)
75
235
235
140,000
(80,000)
57
180
180
390,000
70,000
51
160
160
TRICHLOROETHYLENE DECREASING
CAPITAL COSTS ($)
ANNUAL COSTS ($/yr)
PERCENT EMITTED
TRICHLOROETHYLENE TO AIR (kkg/yr)
VOCTOAIR(kkg/yr)
TSP TO AIR (kkg/yr)
CURRENT OPTION 1 OPTION 2 OPTIONS
CONTROL
— 120,000 330.000 900,000
— (130,000) (220,000) (220,000)
100 77 56 49
660 505 370 325
660 505 370 325
METHYLENE CHLORIDE DECREASING
CURRENT OPTION 1
CONTROL
CAPITAL COSTS ($)
ANNUAL COSTS ($/yr)
PERCENT EMITTED
METHYLENE CHLORIDE TO AIR (kkg/yr)
VOC TO AIR (kkg/yr)
TSP TO AIR (kkg/yr)
100
172
0
6,000
(50,000)
64
110
0
The selection of control options and the estimation of capital and annual costs and removal
efficiencies are projections and estimations based on discussions with AMA staff, literature
review and engineering judgement. Actual costs as well as feasibility and effectiveness
could differ considerably from our estimates.
-------
TABLE C-9 SUMMARY of DRY CLEANING CONTROL OPTIONS
PERCHLOROETHYLENE DRY CLEANERS
CURRENT OPTION 1 OPTION 2 OPTION 3
CONTROL
CAPITAL COSTS ($) — 0 2.750.000 4.470.000
ANNUAL COSTS ($/yr) — 110.000 (170,000) 110,000
PERCENT EMITTED 100 80 42 35
PERCLOROETHYLENETOAIR(kkg/yr) 1,825 1,460 765 640
VOCTOAIR(kkg/yr) 1,825 1,460 765 640
TSPTOAIR(kkgtyr) — — — —
MISCELLANEOUS INDUSTRIAL METHYLENE CHLORIDE USAGE
CURRENT OPTION 1 OPTION2 OPTIONS
CONTROL
CAPITAL COSTS ($) — 390,000 830,000 121,000
ANNUAL COSTS ($/yr) — (120,000) (130,000) (130,000)
PERCENT EMITTED 100 79 67 63
METHYLENE CHLORIDE TO AIR (kkg/yr) 753 595 505 475
VOCTOAIR(kkg/yr) 00 00
TSPTOAIR(Kkgtyr) — — — —
The selection of control options and the estimation of capital and annual costs and removal
efficiencies are projections and estimations based on discussions with AMA staff, literature
review and engineering judgement. Actual costs as well as feasibility and effectiveness
could differ considerably from our estimates.
-------
TABLE C-10 SUMMARY Of COMMERCIAL OIL CONTROL OPTIONS
CURRENT CONTROL OPTION 1
CAPITAL COST ($) —
ANNUAL COST ($/YR) —
ARSENIC TO AIR (kkg/yr) 3.37E-01
CADMIUM TO AIR (kko/yr) 5.12E-01
CHROMIUM-TTO AIR (kkg/yr) 2.04E-01
CHROMIUM-6 TO AIR (kkg/yr) 7.45E-04
NICKEL TO AIR (kkg/yr) 2.24E+00
FORMALDEHYDE TO AIR (kkg/yr) 5.97E-01
TOLUENE TO AIR {kkg/yr)
BENZENE TO AIR (kkg/yr)
POMCDO TO AIR (kkg/yr) 1.26E+01
POMCRO TO AIR (kkg/yr) 1.40E+01
TSPTOAIR • 8.06E+01
VOC TO AIR (kkg/yr) 1.13E+01
PERCENT EMITTED* 100
PERCENT FORMALDEHYDE EMITTED 100
$163,110,000
$8,920,000
1.96E+00
9.79E-01
4.88E-01
O.OOE+00
O.OOE+00
9.51 E+00
1.68E+01
0
33
OPTION 2
$20,040,000
$8,920,000
1.96E+00
9.79E-01
4.88E-01
O.OOE+00
O.OOE+00
9.51 E+00
1.68E+01
0
33
* arsenic, cadmium, chromium, chromium, chromium-6, nickel, toluene, benzene, TSP,
VOC. POMCDO, POMCRO
77?e selection of control options and the estimation of capital and annual costs and removal
efficiencies are projections and estimations based on discussions with AMA staff, literature
review and engineering judgement. Actual costs as well as feasibility and effectiveness
could differ considerably from our estimates.
-------
TABLE C-11 SUMMARY of RESIDENTIAL COAL
and RESIDENTIAL OIL CONTROL OPTIONS
Residential Coal Controls
CURRENT CONTROL
OPTION 1
CAPITAL COST ($) - $18.100,000
ANNUAL COST ($/YR) - $1.300.000
ARSENIC TO AIR (kkg/yr) 2.38E-01 0 OOE+00
CADMIUM TO AIR (kkg/yr) 7.75E-03 O.OOE+00
CHROMIUM-TTOAIR(kKg/yr) 6.03E-02 O.OOE+00
CHROMIUMS TO AIR (Kkg/yr) 448E-05 0 OOE+00
NICKEL TO AIR (kkg^r) 6 03E-02 0 OOE+00
FORMALDEHYDE TO AIR (kkg/yr) 387E-02 2.65E+00
TOLUENE TO AIR (kkg/yr) 0 OOE+00 1 33E+00
BENZENE TO AIR (kkg/yr) O.OOE+00 661E-01
POMRAN TO AIR (kkg/yr) 998E+01 0 OOE+00
POMRBT TO AIR (kkg/yr) 345E+01 O.OOE+00
TSP TO AIR (kkg/yr) 5 36E+02 1.06E+00
VOC TO AIR (kkg/yr) 148E+02 1.88E+00
3 100 0
PERCENT FORMALDEHYDE EMITTED 100 680
PERCENT BENZENE & TOLUENE EMITTED 0 100
Residential Oil Controls
CURRENTCONTROL OPTION 1
CAPITAL COST ($) — $699.270.000
ANNUAL COST ($/YR) - $77.700.000
ARSENIC TO AIR (kkg/yr) 4 51E-02 0 OOE+00
CADMIUM TO AIR (kkg/yr) 329E-01 0 OOE+00
CHROMIUM-T TO AIR (kkg/yr) 329E-02 0 OOE+00
CHROMIUMS TO AIR (kkg/yr) 2.82E-03 0 OOE+00
NICKEL TO AIR (kkg/yr) 3.09E+00 0 OOE+00
FORMALDEHYDE TO AIR (kkg/yr) 7.80E+00 103E+02
TOLUENE TO AIR (kkg/yr) O.OOE+00 5.13E+01
BENZENE TO AIR (kkg/yr) 0 OOE+00 2S6E+01
POMRDOTOAIR(kkg/yr> 836E*01 0 OOE+00
POMRRO TO AIR (kkg/yr) 3.73E-04 0 OOE+00
TSP TO AIR (kkg/yr) 2 55E+02 411E+01
VOC TO AIR (kkg/yr) 7.30E+01 7 26E+01
PERCENT EMITTED* 100 0
PERCENT FORMALDEHYDE EMITTED 100 1300
PERCENT BENZENE & TOLUENE EMITTED 0 100
•except benzene, toluene, and formaldehyde
The selection al control options and the estimation of capital and annual costs and
removal efficiencies an protections ana" estimations based on discussions with AMA
staff, literature review and engineering ludgement. Actual costs as well as
feasibility and effectiveness could differ considerably from our estimates.
-------
TABLE C-12 SUMMARY of RESIDENTIAL WOOD STOVE CONTROLS
CAPITAL COST ($)
ANNUAL COST ($/YR)
ARSENIC TO AIR (kkg/yr)
CADMIUM TO AIR (kkg/yr)
CHROMIUM-T TO AIR (kkg/yr)
CHROMIUM-6 TO AIR (kkg/yr)
FORMALDEHYDE TO AIR (kkg/yr)
BENZENE TO AIR (kkg/yr)
TOLUENE TO AIR (kkg/yr)
NICKEL TO AIR (kkg/yr)
PHENOL TO AIR (kkg/yr)
POMRBT TO AIR (kkg/yr)
POMRAN TO AIR (kkg/yr)
POMRDO TO AIR (kkg/yr)
POMRRO TO AIR (kkg/yr)
POMRWS TO AIR (kkg/yr)
VOC TO AIR (kkg/yr)
TSP TO AIR (kkg/yr)
CURRENT CONTROL
9.00E-04
2.50E-04
7.00E-03
NA
1.50E+00
1 .OOE-02
112
93.9
149
OPTION 1
220,000
3.00E-04
3.50E-04
1 .OOE-04
3.00E-06
1.40E-01
6.50E-02
3.00E-02
8.00E-03
1.00E+00
4.00E-02
1.00E-01
9.00E-02
5.50E-04
O.OOE+00
0.25
1
The selection of control options and the estimation of capital and annual costs and
removal efficiencies are projections and estimations based on discussions with AM A
staff, literature review and engineering judgement. Actual costs as well as
feasibility and effectiveness could differ considerably from our estimates.
-------
TABLE C-13 SUMMARY of ROAD VEHICLE CONTROL OPTIONS
Light Duty Diesel
CURRENT CONTROL OPTION 1
CAPITAL COST ($) _ 956.320
ANNUAL COST ($/YR) — _
CADMIUM TO AIR (Kkg/yr) 6 OOE-09 6.00E-09
FORMALDEHYDE TO AIR (kkg/yr) 32.5 49
POMLDD (kkg/yr) 24.5 2
TSP TO AIR (Kkg/yr) 122.5 110
VOC TO AIR (Kkg/yr) 364 55
PERCENT EMITTTED (cadmium)) 100 100
PERCENT EMITTTED (POMLDD. TSP) 100 90
PERCENT EMITTTED (formaldehyde. VOC) 100 15
Heavy Duty Dlmal
CURRENTCONTROL OPTION 1
CAPITAL COST ($) _ 5,941.000
ANNUAL COST (S/YR) — —
CADMIUM TO AIR (kkg/yr) 6.00E-09 6.00E-09
FORMALDEHYDE TO AIR (kkg/yr) 10 2
POMHDD (kkg/yr) 59 S3
TSP TO AIR (kkg/yr) 197 177
VOC TO AIR (Kkg/yr) 1316 197
PERCENT EMITTTED (cadrmum)) 100 100
PERCENT EMITTTED (POMLDD. TSP) 100 90
PERCENT EMITTTED (formaldehyde. VOC) 100 15
77»e selection of control options and the estimation of capital and annual costs and
removal efficiencies are projections and estimations based on discussions with AMA
staff, literature review and engineering lodgement Actual costs as well as feasibility
and effectiveness could differ considerably from our estimates.
-------
TABLE C-14 SUMMARY Of GASOLINE MARKETING CONTROL OPTIONS
CAPITAL COSTS ($)
ANNUAL COSTS ($/yr)
PERCENT EMITTED
BENZENE TO AIR (kkgfyr)
ETHYLENE CHLORIDE (EDC) (kkg/yr)
ETHYLENE DIBROMIDE (EDB) (kkg/yr)
XYLENE (kkg/yr)
TOLUENE (kkg/yr)
ETHYL BENZENE (kkg/yr)
VOC TO AIR (kkg/yr)
TSPTOAIR (kkg/yr)
CURRENT
CONTROLS
100
25.1
2.1
0.2
8.4
46.2
0.4
4,185
OPTION 1
8,300,000
970.000
29
7
0.6
0.06
2.4
13.4
0.1
1,230
The selection of control options and the estimation of capital and annual costs and removal
efficiencies are projections and estimations based on discussions with AMA staff, literature
review and engineering judgement. Actual costs as well as feasibility and effectiveness could
differ considerably from our estimates.
-------
TABLE C-15 GLOSSARY OF TERMS
POMRDO
POMRRO
POMCDO
POMCRO
POMRAN
POMRBT
Polycyclic organic matter from residential distillate oil
Polycyclic organic matter from residential residual oil
Polycyclic organic matter from commercial distillate oil
Polycyclic organic matter from commercial residual oil
Polycyclic organic matter from residential anthracite coal
Polycyclic organic matter from residential bituminous coal
-------
APPENDIX D
BACKGROUND DOCUMENTATION FOR BENEFITS ESTIMATES
-------
APPENDIX D
BACKGROUND DOCUMENTATION FOR BENEFITS ESTIMATES
This appendix describes the sources of available benefits
numbers for TSP and VOC and how these benefits numbers were used
in the Baltimore IBMP. As discussed in Chapter VI, it is
desirable to know the economic value of an outcome to put into
perspective the benefit and costs of changing pollution levels
when devising pollution control programs. In many situations,
considering the benefits, and not just cost-effectiveness, could
results in different the risk management decisions.
We begin by describing the general techniques available for
quantifying benefits (i.e., valuation using direct and indirect
economic techniques). We then present the sources and benefits
results for the TSF and VOC effects considered in the Baltimore
IBMP. We conclude with a discussion of how the benefits
estimates were used to calculate the overall benefits afforded by
individual control options evaluated in the Baltimore IEMP.
GENERAL VALUATION TECHNIQUES
Hedonic Property Method
One approach frequently used to value environmental
conditions that are related to the price of goods is the hedonic
property method. Hedonic property studies assume that property
values reflect the economic value of environmental quality
characteristics in the vicinity of the property relative to other
areas. Given that pollutant levels vary within most urban areas,
and property owner preferences are affected by pollutants, the
market prices for properties will capture these qualities. Using
data collected over a range of property prices, pollution
conditions and other attributes related to property values, and
conducting the appropriate statistical tests, can provide
estimates of the impacts of pollution on market prices and
individual preferences for different pollution conditions. The
pollution price will reflect any health and welfare risks
perceived by the individual, but there is little empirical
evidence to suggest how the total price should be attributed to
specific outcomes. Therefore, this technique is not well-suited
for use in instances where control programs are evaluated on the
basis of their ability to reduce specific risks.
D-l
-------
Hedonic Wage Method
Hedonic wage studies are based upon the assumption that
wage rates vary across occupations for a variety of reasons, one
of which is the risk of injury and death inherent in an
occupational environment. Each occupation can be characterized
as having a probability of death or lesser injury associated
with the nature of work. The risk of injury is presumed to be
known by individuals and employers across different occupations.
Given this information, individuals attempt to select jobs where
the marginal adjustment in the wage rate is equated with the
incremental level of risk they expect to face when choosing an
occupation. Combining the marginal change in wage accepted with
the marginal change in risk of injury yields an implicit value
for the acceptance of various risks. By observing the variations
in wages across different occupational risks, a functional
relationship can be estimated that describes how much wages must
change to accept incremental changes in risk.
Travel-Cost Method
A third market-oriented method of valuing changes in
pollution condition is the travel-cost method. This method
assumes individuals produce activities that require inputs, for
example, travelling to a beach to swim in order to enjoy some
level of satisfaction. The individual, in selecting which beach
to visit, considers the time and resources necessary to get to
each beach, plus the environmental and social amenities provided
at each beach. If there are sufficient numbers of observations
on beach visits and variations in environmental amenities at the
beaches, then it is possible to infer the value of the
environmental amenities through the participation rates and costs
(e.g., travel and time costs). Because most of the expenses are
incurred in travelling to these sites, the method is referred to
as the travel-cost method.
Resource Cost Method
The health and welfare outcomes of pollution exposures are
often observed in market transactions for goods other than homes
and salaries directly affected by the pollutant, or goods used to
mitigate the risks posed by pollutants. After calculating a
dose-response function to measure the change in outcomes (e.g.,
health effects, agricultural effects) for a given change in
exposure to pollution, outcomes are valued using either the price
individuals must pay to recover from the damages once they have
occurred, or preferably, the price they are willing to pay to
protect against suffering the damaging outcomes of pollution.
This approach can be used in those instances where the outcomes
are observed, but recognition of the contribution of pollution to
the outcome need not be well understood by the individual. The
price of preventing or rectifying the effects are then used as
D-2
-------
implicit values for benefits of reducing pollutants. The
relevant proportion pollutants contribute to the estimated
outcomes can be used to estimate the economic benefits of
changing the contribution of pollution to these outcomes. For
human health outcomes, the technique of assigning a value to
human health outcomes based on the cost of medical care resources
is referred to as the cost-of-illness approach.
The use of cost-of-illness and other resource cost methods
may fail to capture the total extent of damages attributable to
pollution. Concentration-response functions do not fully account
for defensive behavioral responses that may occur in polluted
environments. Estimating concentration-response functions using
health statistics would not capture the actions of individuals
taken to avoid these damages. Therefore, the valuation of
pollutant outcomes should look beyond the direct effects and
include the costs of additional actions taken to reduce pollution
risks. In addition, there is a failure to account for the pain
and suffering caused by pollutants that are not captured in the
costs of providing medical care. The magnitude of these other
economic values will be a function of the ability of individuals
to recognize the relationship between pollution and effects, and
their ability to take personal actions to prevent the damages
from occurring. Therefore, rather than rely solely on the
opportunity cost of medical resources, the individual's
preferences should also be considered when valuing the benefits
of controlling pollutants.
Contingent Valuation Method
The potential inability of the labor, housing, and other
markets to reflect preferences for risk reductions and the values
of reducing risks have led to other alternative mechanisms. The
contingent valuation methods uses a survey format to gain
information from individuals on their preferences for the
perceived benefits from reductions in pollution levels.
Individuals respond to questions that pose the existence of a
market where they can exchange goods for dollars, even though
such a market may not exist in reality. If the questions are
posed in a manner so that there are no particular gains to be had
on the part of the individual from revealing other than their
true preferences, then this method can provide values that are
both meaningful and useful for evaluating environmental benefits.
values obtained using this method are capable of including the
value of reduced pain and suffering, and the value of maintaining
activities in accordance with their preferences and not those
associated with mitigating behavior.
D-3
-------
PARTICULATE MATTER BENEFITS ESTIMATES
Health Effects
Mortality Effects
Mazumdar (1982) investigated the mortality rates in the
greater London area from the period 1958-1972 and tested whether
the daily variations in these rates were correlated with
variations in British Smoke (BS) and S02 concentrations Schwartz
(1986) reanalyzed the data and established more robust
statistical relationships between daily mortality figures and BS
levels less than 200 ug/m3. After controlling for the effects of
temperature and humidity, the relationship reads:
change in daily mortality rates in London
= 0.138 x (change in daily BS concentrations).
To convert BS measures into total suspended particulate
concentrations, several studies have been conducted with BS and
TSP measures taken in London during the same period when
mortality rates were analyzed. The relationship between BS and
TSP in London is that a 1 ug/m3 change in BS is likely to result
in a 1.41 ug/m3 change in TSP, for BS levels less than 200 ug/m3.
Using this relationship to convert BS to TSP, the mortality
equation now is:
change in daily mortality rates in London
= (0.138)/(1.41) x (change in daily TSP concentrations).
Given the London population exposed to TSP concentrations
was about 8 million persons, the change in daily mortality rates
per 100,000 persons exposed to TSP would be 1.22 x 10~3 mortality
cases per day per 100,000 people exposed. On an annual basis:
change in annual mortality cases per 100,000 persons
= 0.446 x (change in annual average TSP concentration).
For example, if annual average TSP concentrations fall from
75 to 70 ug/m' in a city with population of 1 million persons,
the estimated annual mortality rate will fall by 22.3 cases per
year. It is likely that many of these estimated cases will be
reduced for persons aged 65 and over, or those suffering from
chronic illnesses that put them at greater risk.
Mortality Values
Most of the studies that deal with valuing changes in
mortality rates have not dealt with the value of preventing the
D-4
-------
death of a particular individual. Instead, most studies examine
the value of a relatively small reduction in the probability of a
death occurring, and then extrapolate this to value the total
number of cases reduced. Most studies have looked at wage
premiums associated with different Job-related risks, the theory
being that higher risk occupations will be rewarded with greater
wages to induce individuals to participate in riskier occupa-
tions. Summaries of the empirical analyses performed to date
(ERC: 1983, 1986) suggest that individuals are willing to accept
an additional risk of 1 x 10~4 on an annual basis for $100 to
$700 per year. Extrapolating this to one case yields a
willingness to accept this risk for $1 to $7 million per case.
The general rule is to use $2 million per case, or $2 per 1 x
10~*> reduction in premature mortality risk.
Acute Morbidity Effects
Ostro (1983,1987) used data collected in the Health
Interview Survey on restricted activity days (RADs) and
information on fine particulate matter (FP) to develop
concentration-response relationships between RADs and FP
concentrations. The form of the estimated relationship was as
follows:
change in RAD per 2-week period
= 0.0048 x (baseline RADs for 2-weeks)(change in 2-week FP).
To convert the relationship to one based on annual average
number of RADs and changes in FP,
change in RADs per person per year
• (0.0048) x (baseline RADs per year)(change in annual FP)
For example, if the baseline annual average number of RADs
is 20 days per year, and the change in FP is from 40 ug/mj to 35
ug/nr, then:
change in RADs per person per year
= (0.0048)(20)(5) =0.48 fewer RADs per person exposed.
Given that TSP is a more standard measure of particulate
matter, it may be desirable to express the above relationship in
terms of TSP rather than FP. The relationship between TSP and FP
varies depending on the emission sources of TSP. For example,
should the relationship be 1 ug/m3 TSP equals approximately 0.25
ug/m^ FP, and TSP concentrations drop from 90 ug/mj to 75 ug/m-3,
then:
D-5
-------
change in RADs per person per year
= (0.0048)(20)(15)(0.25) =0.26 fewer RADs per person.
Samet (1981) examined how changes in the number of emergency
room visits varied with changes in ambient concentrations of TSP,
S02/ and N02 during early spring and fall months of 1974-1979.
The estimated relationship between daily emergency room visits in
Steubenville, Ohio (exposed population group) and TSP is:
change in visits to emergency rooms per person per year
= fO.Oin (365)(change in annual average TSP levels)
(31000 persons in Steubenville),
= (1.29 x 10~4) (change in annual average TSP).
For example, if TSP levels fell from 90 ug/m3 to 75 ug/m3 in
a city having a population of 1 million persons, the expected
change in the number of emergency room visits would be:
change in number of emergency room visits per year
= (1.29 x 10-4)(15)(1 x 106),
= 1940 fewer emergency room visits per year.
Chronic Morbidity Effects
Ferris (1978) examined how the number of chronic
respiratory disease cases in Berlin, New Hampshire varied in
accordance with variations in ambient TSP concentrations. Ferris
did not report concentration-response functions, but rather
reported on age and sex-standardized rates of prevalence for
chronic non-specific respiratory diseases for the two years
examined (1961 and 1967). The ratios suggest that TSP reductions
from 180 ug/m3 to 130 ug/m3 result in significant changes in the
number of chronic respiratory disease incidents, but changes
below 130 ug/m3 did not result in measurable changes in
respiratory disease. For those instances where Ferris found a
relationship between TSP and chronic respiratory disease cases,
the equations are as follow:
change in number of chronic respiratory disease
incidents per person per year
= (b)(change in TSP)(average annual rate of chronic
respiratory disease incidents per
person),
D-6
-------
where b = 0.0027 for males and b = 0.0019 for females. For
example, if TSP levels fall from 145 to 130 ug/m^, average annual
incident rates for males are 0.247, and rates for females are
0.281, then:
change in number of incidents for males per year
= (0.0027)(15)(0.247) =0.01 fewer incidents per year,
and
change in number of incidents for females per year
= (0.0019)(15)(0.281) = 0.008 fewer incidents per year.
Morbidity Values
Most dollar estimates for the value of preventing morbidity
effects related to PM have been based on a cost-of-illness
approach. The cost of providing care to correct for the health
effect and the opportunity cost of time and productivity lost
during the illness are used to estimate the value of reducing
health effects attributable to PM. Economic theory suggests that
these estimates are underestimates of the willingness to pay on
the part of exposed individuals to avoid incurring the estimated
health effects. Cost-of-illness studies fail to capture the
discomfort and inconvenience imposed upon individuals. Efforts
to estimate willingness to pay using surveys has been successful
for some morbidity effects, primarily those related to ozone.
Some of the effects attributed to PM may be of a similar nature
to those caused by ozone, however, most RADs attributable to PM
are considered to be of a more severe nature than RADs caused by
ozone.
Dollar values for chronic and acute health effects that lead
to RADs are estimated using wage information for working persons
and medical expenditures. If a person suffers a work loss day,
the economic value of the day lost is estimated to be equivalent
to the productive value of the individual. Economic theory
suggests that the productive value is equal to the individual's
wages. Therefore, if an individual is removed from the work
force due to illness caused by PM exposure, the benefits of
reducing the health damages should be at least equal to the value
in foregone productivity as measured by wages. As stated above,
this figure does not include the inconveniences imposed on the
individual and other persons at work or in the home, so it is
likely to underestimate the true value of reducing health
effects. For the U.S., the average wage rate is $10/hour, so a
lost work day would be valued at $80/day for full-time persons.
Persons working part time, persons working in services that do
not receive hourly wages (self-employed sectors of the economy,
homemakers) and retired persons are also assumed to have their
D-7
-------
more severe RADs valued using average wage rates. Convention
dictates that these RADs be valued at the same rate. For those
RADs of a less severe nature that allow the individual to
continue working/ but presumably at some diminished level or in a
state of mild discomfort, the value is often assumed to be some
fraction of the average daily wage. Most studies use one-half to
one-quarter the daily wage rate, although these values are
arbitrarily chosen.
An additional cost included in the value of RADs is the
direct medical expenditures necessary to combat the health
effects of PM. Based upon national health statistics for medical
expenditures and estimated number of RAD incidents, the average
cost per incident is around $2 for acute effects. The average
medical cost incurred attributable to chronic health effects of
PM are $10 per RAD incident. Severe acute and chronic attacks
will result in significantly higher medical expenditures than
those presented here, but these figures are intended to represent
the average expense aggregated over a considerable number of
cases. In many instances, less severe attacks will not require
that individuals spend money for medical services. In those
cases where more severe episodes occur, it Is likely that the
individual will have to take advantage of acute care provided by
emergency rooms.
Dollar value estimates for emergency room visits use the
same theory as RADs, but supplement the daily wage estimate with
the average cost of providing emergency room care to the
individual. Average medical expenditures for an emergency room
visit amount to approximately $180 per out-patient visit. This
value includes costs covered through medical health insurance.
Assuming that an emergency room visit results in a lost work day
or its equivalent for non-full time employees, then the total
value for an emergency room visit would be $260 ($180 + $80). As
with other less severe health effects, this dollar estimate does
not include the values attributed to reducing potential pain and
anxiety associated with the estimated health effects. Therefore,
this estimate is expected to be an underestimate of individual's
willingness to pay to avoid the outcome. Few studies have
examined the magnitude of the underestimation (ERC, 1986), but
for those conducted to date, the willingness-to-pay estimate
tends to be 2-3 times larger than the cost-of-illness value.
Welfare Effects
Materials Effects and Values
Estimates of materials damages due to TSP are developed from
using aggregate consumer purchase data on household goods and
changes in these purchases as a function of TSP concentrations.
The relationship is directly between TSP concentrations and
expenditures, so there is no intermediate relationship between
D-8
-------
TSF and household material damages. The data was collected on 24
Standard Metropolitan Statistical Areas (SMS As) and expenditures
on a set of commodities. Demand equations are estimated for
seven different commodity groups: food, shelter, household
operations, home furnishings, clothing, transportation, and
personal care. Air quality data concentrations and expenditures
data was analyzed using conditions present during 1972 and 1973.
A system of demand equations are estimated, each having pollutant
variables that are introduced to explain variations in purchases
for goods on the basis of different pollution conditions in the
24 SMS As.
The TSP variable was significant (at the 95% level) in the
demand equations for home repair activities, and laundry and
cleaning operations. The most robust measure of TSP was based on
the second-highest monitored 24-hour concentration of TSP for all
monitors in the SMSA. Through a somewhat complicated series of
estimation procedures, the welfare estimates for changes in TSP
are estimated using compensated demand functions and expenditure
functions. Therefore, it is difficult to express the benefit
estimate procedure as a simple formula.
A similar procedure was followed to estimate the impacts of
particulate matter on the manufacturing industry. Examination of
cost functions and TSP concentrations showed that TSP was a
significant variable in estimating cost variations across
manufacturers of fabricated metals and machinery. As with the
household sector model developed by Mathtech, there is no simple
equation that can be used to estimate effects. However, attempts
to quantify benefits on the basis of TSP emissions have shown
that the benefits to household and manufacturing sectors of
reducing TSP emissions in urban areas may range between $0 - $50
per ton, with an average benefit of $10 per ton. The range is a
function of the proximity and density of household and industrial
operations to TSP emission sources.
Visibility Effects
Trijonis (1982) developed an equation that described the
relationship between visibility as measured by visual range, and
the presence of TSP, NOX, sulfates (804), and relative humidity
(RH). He estimated the following relationship for California:
Visual range in miles
24.3
_
(.12 + .04(S04/(1-(RH/100)) + .03(NOX/(1-(RH/100))
+ .003((TSP-S04-NOX)/(1-(RH/100))),
where RH = relative humidity. As this relationship relies on
knowing several different pollution variables, it is somewhat
cumbersome. However, to illustrate an example, if we assume that
D-9
-------
at the nearest ambient monitoring station to the individual's
place of residence. The statistical analysis was conducted using
the number of respiratory restricted activity days (RRADs)
experienced over the two-week period as the dependent variable,
and the pollution and socioeconomic variables as the independent
variables. The ozone variable was positively correlated with the
number of RRADs and the coefficient on the ozone variable was
significant. Two different functional forms were examined that
described the variability of RRADs and ozone. The first
regression was:
number of RRADs per person per 2-week period
= exp(aZ + b(ozone)),
where Z = other independent variables, and a and b are estimated
parameters. When using this functional form to estimate a change
in RRADs for a change in ozone, one must know the baseline
number of RRADs,
change in number of RRADs per person per 2-week period
- (average number of RRADs){b(change in ozone)).
The second functional form used the square root of the
observed ozone concentration as an independent variable, so that
the regression equation read:
number of RRADs per person per 2-week period
= exp(aZ + b(ozone)-5),
and
change in number of RRADs per person per 2-week period
= (average number of RRADsU.5HbWchange in ozone} .
(beginning ozone level)'3
The coefficients for the above equations were b = 6.883 for the
linear equation, and b = 4.926 for the square root function. The
average number of RRADs = .162 days per person during a two-week
period, and E(ozone) = .042 ppm in the study. Using these as the
baseline conditions, the calculated change in RRADS for a 10%
reduction in the two-week arithmetic average of the daily high
hourly ozone levels would be:
D-ll
-------
linear model: change in expected number of RRADs
per person per 2-week period
= (.162)(6.883){.042 - (.9)(.042)>
= .005 fewer cases per person during a two-week period.
and
quadratic model: change in expected number of RRADs
per person per 2-week period
= f .162W.5W4.926U0.42 - f.9W0.421V
(.042)-3
= .008 fewer cases per person during a two-week period.
These health effects were estimated for adult populations,
including both asthmatic and non-asthmatic individuals. However,
a different techniques is used to estimate ozone effects for
asthmatic individuals.
Holguin (1984) developed a prospective epidemiological study
to determine the impacts of ozone on asthmatics. Information on
pollutant levels as measured at monitors in proximity to their
place of residence were used to estimate exposure levels.
Holguin used information on ozone, N02, pollen, and weather data
to supplement socioeconomic characteristics that would explain
variations in asthma symptoms across his sample. Seventy-five
percent of his sample were individuals between the ages of 10 and
19 years and all individuals were non-smokers. The ozone
exposure variable related to the presence or absence of an
asthma attack over a 12-hour period was the maximum 12-hour
period exposure concentration. Holguin used a weighted logistic
model to determine the change in probability of suffering asthma
attack given a change in the maximum exposure concentration for
ozone. The estimated equation is:
probability of experiencing an asthma attack per asthmatic
per 12-hour period
1
(l-i- exp(az + b(ozone))
where a = vector of coefficients for other independent variables,
z = other independent variables, b = estimated coefficient for
ozone (6.20 in Hoguin's sample), and ozone = ozone concentration
(12-hour period 1-hour maximum concentration in ppm). For
changes in ozone, the change in expected asthma attacks would be:
D-12
-------
change in probability of having an asthma attack during a
per asthmatic per 12-hour period
= (l - pi fexofbfOzone^-Ozonein - 11
(!+((!- p)/p)exp(b(0zone2 -
where p - baseline attack rate per 12-hour period, Ozone^ =
initial 1-hour maximum ozone concentration, and Ozone2 = final 1-
hour maximum ozone concentration. If the asthmatic population
constitutes 8 percent of the population exposed, the baseline
attack rate is 0.10, and ozone the maximum ozone concentration
falls from 0.13 ppm to 0.12 ppm, then the number of attacks
reduced per 12 -hour period would be:
change in number of asthma attacks per asthmatic per
12 -hour period
= ri - 0.10 Wexpf 6.20U0.12 - 0.1311 - llf population HO. 081
(1 + ((1 - 0. 10)/0.10)exp(6. 20(0. 12-0. 13)) )
= (- 4.576 x 10~4) (population exposed).
If the same 0.01 ppm reduction in the 12-hour period maximum
1-hour ozone concentration occurs over the entire ozone season
then the change in number of asthma attacks per ozone season is :
= (- 4.576 x 10~4) (exposed population) x
(number of 12-hour periods in the ozone season) .
McDonnell (1983) used six groups of exercising male
volunteers in a clinical human exposure study to test whether
exposures to different ozone concentrations over short (2.5
hours) time periods led to changes in pulmonary function,
ventilatory function and other health symptoms. Using a sigmoid
function, regression equations were used to estimate to what
extent different ozone levels could explain the variations in
pulmonary, ventilatory and symptomatic effects across individuals
in the sample. The tests showed small differences from the
baseline ozone levels and levels around 0.12-0.18 ppm. Larger
differences were observed at the upper levels tested (0.24-0.40
ppm).
The functional form of the regression leads to the following
relationship for coughing symptoms is:
number of coughing incidents per 3-hour period per
exercising adult non-smoking male
1 (non-smoking male population) x
(1 + exp(-a-b(ozone))(% exercising at moderate level),
D-13
-------
where a and b are estimated coefficients (a = - 1.527 and b =
12.78) and ozone = 3-hour period average ozone exposure level.
To determine the change in the number of coughing incidents for
exercising non-smoking males when ozone exposure concentrations
are reduced, use the following:
change in number of number of coughing incidents per 3-hour
period per exercising adult non-smoking male
= (I - pUexpfbfOzoneg-Ozonein - 11 x
(!+((!- p)/p)exp(b(0zone2 -Ozonei))
(percent time spent exercising heavily) x
(adult male non-smoking population),
with p = initial probability of coughing at initial ozone level,
Ozonei = initial ozone concentration and Ozone2 • new
concentration.
If adult males spend approximately 1.5 percent of their
waking hours (15 hours per day) exercising heavily, and ozone
levels decrease from 0.10 to 0.09 ppm over the 3-hour period they
are exercising (one-fifth of all waking hours) then:
change in number of coughing incidents per 3-hour period per
exercising non-smoking adult male
= fl - Q.438Wexofl2.78fO.09-0.10n - 1UO.015/51 x
(1 + ((1 - 0.438)/0.438)exp(12.78(0.09 - 0.10))
(adult male non-smoking population)
= ( - 9.50 x 10~5)(adult male non-smoking population).
McDonnell also determined how periods of shortness in breath
changed with ozone exposures. He used the same functional
relationship as that for coughing and estimated new coefficients
(a = -0.182 and b = 8.106). Therefore, the change in number of
shortness of breath incidents, under the same assumed changes in
ozone as above would lead to the following:
change in shortness of breath incidents per 3-hour period
per exercising non-smoking adult male
= il - Q.652Hexpf8.106(0.09-0.10n - 1HO.015/51 x
(1 + ((1 - 0.652)/0.652)exp(8.106(0.09 - 0.10))
(adult male non-smoking population).
= - 5.45 x 10"5 (adult male non-smoking population).
D-14
-------
Hammer (1974) conducted a prospective epidemiological study
of 112 nurses in the Los Angeles area. The nurses were asked to
keep a daily diary of their activity patterns and medical
history. The nurses were asked to report on the presence/absence
and severity of respiratory disease symptoms. He supplemented
their diaries with pollution data for oxidants, CO, NC>2/ and
weather data.
Hasselblad and Svendsgaard (1975) reanalyzed the Hammer data
and estimated a relationship between eye irritation and daily
maximum one-hour concentration of ozone levels. The relationship
is:
probability of experiencing eye irritation per day per
person
= (0.0407) + _ fl - 0.04071 _ .
(1 + exp(4.96 - 9.07(ozone))
The change in the probability of experiencing eye irritation for
a change in ozone levels would be:
change in probability of having an eye irritation case per
day per person
(I - 0.0407U1 - pi fexof 9. 07 f Ozones-Ozone! 11 - 11 .
(!+((!- p)/(p - 0.0407) )exp(9.07(0zone2 -
If the initial daily maximum ozone level is 0.13 ppm, and the new
daily maximum is 0.12 ppm, then the change in probability is:
change in probability of having an eye irritation case per
day per person
= (I - 0.0407U1 - O.Q621Uexpf9. 07/0. 12-0.1311 - 11
(1 +(( 1-0. 0621 )/( 0.0621-0.0407) )exp(9. 07(0.12-0.13)))
= - 1.90 x 10~3.
To determine the change in number of cases, the change in
probability would be multiplied by the number of persons exposed
to the assumed ozone concentrations .
Likewise, Hammer estimated the relationship between ozone
levels and changes in chest discomfort. Using the same
functional form, Hasselblad and Svensgaard estimated the
following relationship:
D-15
-------
probability of suffering chest discomfort per day per person
= (0.017) + ___ (1 - 0.0171 .
(1 + exp(3.53 - 0.23(initial ozone level)))
Therefore, the change in the probability of suffering chest
discomfort, given the above scenario (the reduction of 0.13 to
0.12 ppm in daily maximum ozone concentration) yields the
following:
change in probability of suffering chest discomfort per day
per person
= II - 0.017W1 - Plfexpf0.23(0zone2-0zonein - 11
(!+((!- p)/(P - 0.017))exp(0.23(0zone2 -Ozone!)))
= fl - Q.017W1 - Q.045BUexpf0.23(0.12 - 0.1311 - 11 ,
(1 + ((1 - 0.0458)/(0.0458 - 0.017))exp(0.23(0.12 - 0.13)))
= - 6.33 x 10~5.
Lastly, Hammer with Hasselblad and Svendsgaard estimated the
relationship between headaches and ozone exposures. Using the
same functional relationship,
probability of experiencing a headache per day per person
= (0.0976) + fl - 0.09761
(1 + exp(4.88 - 4.7(initial ozone))).
If we assume a change in daily maximum ozone concentrations from
0.13 to 0.12 ppm, then the change in probability of having a
headache will be:
change in probability of a headache per day per person
= fl - 0.0976H1 - O.lllfexpf4.7f0.12 - 0.1311 - 11— ,
(1 + ((1 - 0.11)7(0.11 - 0.0976))exp(4.7(0.12 - 0.13)))
- - 5.24 x 10~4.
Schwartz et al. (1987) also re-examined data collected in
the prospective study by Hammer (1974) to test for the effects of
ozone on young female adults. Using a logistic functional form
for the concentration-response function, Schwartz estimated the
following:
D-16
-------
Probability of incurring a cough per day per person
= 1 + dv,
( 1 + exp(az + b(Ozone))
where z = other variables, a = estimated coefficients for z
variables, b = coefficient for ozone, Ozone = maximum 1-hour
observation, v = lagged error variables, and d = coefficients for
lagged error variables. If the baseline probability of incurring
a cough is 0.095 and ozone levels decrease from 0.13 to 0.12 ppm,
given b = 0.61, the change in coughing would be the following:
change in probability of coughing per person per day
= 11 - O.Q95UexpfO.61(0.13-0.12n - 11 ,
(1 + ((1 - 0.095)/0.095)exp(0.61(0.13 - 0.12))
= - 5.26 x 10~4.
Schwartz also examined the effects of ozone on eye
discomfort. Using the same functional form, with b = 2.02 and
the probability of suffering eye irritation = 0.063, a change in
maximum ozone concentrations from 0.13 to 0.12 would yield the
following:
change in probability of suffering from eye irritation
per person per day
= (i - 0.063WexDf2.02fO. 13-0.121 - 11 ,
(1 + ((1 - 0.063)/(0.063)exp(2.02(0.13 - 0.12))
= - 1.20 x lO'3.
This estimate is smaller than the estimated change in probability
given by Hasselblad and Svensgaard in their analysis of the same
data set. The difference is due in part to Schwartz's attempts
to control for autocorrelation in the data given the time-series
aspects of the data. Additional analysis of the effects of ozone
on chest discomfort and headaches found that ozone was not a
significant explanatory variable in these regression equations.
Chronic Health Effects
The relationship between exposure to ozone concentrations
and chronic respiratory disease was established for humans on the
basis of animal studies conducted by Fujinaka(1985) and
extrapolated to the estimate likely impacts on humans. The
concentration-response function developed by McKee (1987) using
the animal study data is:
D-17
-------
Proportion of individuals contracting chronic respiratory
disease per year
= (exposed population)(l/.64){(7/12)(O.OSppm ozone) +
(5/12)(annual average daily ozone concentration)},
where the annual average daily ozone level is based on
measurements taken between 10:00 a.m. and 7:00 p.m. For example,
if one million persons are exposed to ozone concentrations that
fall from 0.07 ppm to 0.06 ppm, and of the one million persons
exposed one in 160 is at risk of contracting chronic respiratory
disease, then the estimated number of fewer chronic respiratory
cases would be:
change in number of chronic respiratory disease cases per
year
= (1 x 106 persons)(l/160)(l/.64){(5/12)(0.06) -
(5/12)(0.07)>,
= - 40.7, or 40.7 fewer chronic cases per year.
Morbidity Values
For the non-asthmatic population, the health effects
literature suggested that individuals would suffer from coughing,
shortness of breath, eye irritations, and headaches when exposed
to larger ozone concentrations. In order to attribute values to
these effects, a series of contingent valuation studies were
conducted to ascertain how much individuals would be willing to
pay to avoid suffering from each of the above symptoms. The
average values for each of the willingness-to-pay estimates have
been summarized by Krupnick (1986) and are listed below for each
symptom:
Symptom Willingness to Pay Value
Per Incident
cough $4.00
shortness of breath 8.00
chest discomfort 6.00
eye irritation 5.00
headache 5.00
For those individuals expected to suffer from a respiratory
restricted activity day (RRAD), the unit value was estimated to
be $21.00 per day. Since the symptoms given above can
D-18
-------
individually, or collectively, contribute to a respiratory
reduced activity day, it is necessary to develop some protocol to
avoid double-counting of health effects and economic benefits.
Although there is no "correct" way to do this, it is suggested
that the number of RRADs and symptom days be calculated
individually using the relevant exposure groups and concentration
response functions. The sum of the symptom effects excluding eye
irritation effects can be compared with the estimated number of
RRADs. It expected that individuals experiencing a RRAD will
have one or more of the above symptoms. Therefore, it is not
clear what relationship may exist between the total number of
cases estimated using the concentration-response relationships
for the symptomatic effects and the number of RRADs. Therefore,
it is suggested that the values for the two estimates be
individually calculated and the arithmetic mean of the two be
used to calculate the benefits of changing ozone concentrations
for all symptoms other than eye irritation, which should be
calculated separately.
Most of the health effect studies have focused on male adult
populations. It is likely that some of these same health effects
will occur for females and children when exposed to ozone,
although not necessarily to the same extent. Several
epidemiological studies have failed to show that ozone
concentrations are statistically correlated with bed rest days or
lost school days experienced by children due to respiratory
illness (Portney, 1984). The analyst should use some discretion
in determining which populations should be fitted to each
concentration-response function.
The average value of reducing asthma incidents is estimated
to be $25.00 per incident. This value is developed using
contingent valuation techniques that capture both the cost of
mitigating the effects plus the pain and suffering avoided for a
moderate asthma attack. This value should be used for both adult
and children asthmatic populations.
The average value of reducing chronic disease associated
with ozone exposures is based on cost-of-illness estimates for
expenditures undertaken for medical expenses and the value of
lost work days. Data on national expenditures for chronic
respiratory diseases suggest that the average annual cost-of-
illness for individuals suffering from chronic respiratory
diseases is $740 per year.
Since the cost-of-illness estimate fails to include the
value of avoiding the pain and suffering that individuals would
be willing to pay to avoid having chronic respiratory diseases,
Rowe (1985) attempted to estimate the ratio of cost-of-illness
values to willingness-to-pay values for respiratory diseases.
Their study suggested that willingness-to-pay values are
approximately.two times the values estimated using cost-of-
D-19
-------
illness measures for asthmatics. Therefore, doubling the chronic
respiratory disease value to $1480 per year may yield a more
accurate estimate of the value of reducing chronic diseases from
ozone.
Since the individuals suffer from the effects of ozone
induced chronic diseases not only for the year in which the
disease's effects first appear, but also in subsequent years, the
costs should be estimated over the entire time period that the
health effects occur. Therefore, a present value estimate of the
cost-of-illness should be used. When estimating present values,
it is appropriate to use a discount value on future expenditures
that reflects the opportunity cost of resources used to combat
the disease. The current convention suggests that rates between
five and ten percent are appropriate, depending on the
opportunity cost of capital used in the supply of medical
services.
Welfare Effects
Agricultural Effects
Heck (1982,1983) studied the impacts of ozone on yields for
several major types of crops using ambient and experimental
concentrations of ozone. Dose-response functions for each crop
were estimated using several different functional forms. The
Weibull function was used to combine test results conducted at
different test sites for the same species of plant in order to
arrive at a common proportional yield response to ozone effects.
The functional form is:
Yield = (Maximum yield)(exp(-(ozone concentration/b)c)),
where b = estimated ozone concentration when yield is
(0.37)(maximum yield) and c = estimated dimensionless shape
parameter for the function. Using this form, the percentage
change in yield for a given change in ozone relative to the
background ozone level (average measured at 0.025 ppm) would be:
percent reduction in yield relative to 0.025 ppm per year
= 1 - .
The individual or common response ozone concentration
parameters estimated by Heck are:
D-20
-------
Crop b (ppmV c
Soybeans 0.150 1.28
Corn (Coker 16) 0.221 4.46
Corn (others) 0.158 3.53
Wheat 0.174 2.90
Peanut 0.111 2.11
Cotton 0.197 1.12
Kidney Bean 0.287 1.77
Lettuce 0.098 1.22
Turnip 0.093 2.75
Spinach 0.135 2.08
As an example, if the average ozone concentrations during
the growing season falls from 0.040 ppm to 0.030 ppm, then for
soybeans crops, the estimated percent change in yield rates would
be:
percent reduction in soybean yield per year
= 1 - {exp((0.040/0.ISO)1-28 - (0.025/0.ISO)1-28)}
= 1 - {1.087}
= -0.087, or an increase in yield of 8.7 percent relative
to that harvested at 0.040 ppm ozone.
Agricultural Values
Kopp (1983) took these dose-response functions and, with a
behavioral model of farming practices, determined how aggregate
crop yields and revenues would change under different ambient
ozone concentration scenarios. In his original analysis, Kopp
assumed that agricultural markets and prices were established in
a competitive market structure. Subsequent reviews suggested
that government price support structures may lead to inefficient
and distorted agricultural markets. When this alternative view
of agricultural markets is considered, then the expected changes
in behavior and agricultural revenues from changes in ozone
concentrations will be different from the more simplistic model
of the agricultural sector.
The model provides estimated changes in revenues by
agricultural producing area and crop type for different ozone
concentration scenarios. In order to estimate welfare changes,
one must have some knowledge of the agricultural market
structure.
D-21
-------
Forest Effects
Tree growth rates have been shown to be sensitive to ambient
ozone concentrations, with reduced growth rates observed at
higher ozone concentrations. Several studies have been conducted
on controlled exposures to young seedlings/ comparisons of growth
of young seedlings in uncontrolled environments over different
time periods and ozone concentrations/ and radial growth patterns
for sensitive and insensitive species.
There is a limited amount of data available to draw upon and
construct dose-response functions for different tree species
exposed to ozone concentrations. The controlled exposure studies
are considered to have the most robust dose-response
relationships/ although the exposure patterns varied between 6
and 12 hours per day, so it is difficult to easily draw a
relationship between percent changes in growth patterns and daily
exposure levels.
Wilhour (1986) summarized the empirical results developed
using experimental studies exposing seedlings to controlled ozone
concentrations. He concluded that the average change in percent
growth was 1.8 percent for softwood trees per 0.01 ppm change in
ambient ozone concentration level. For hardwood species, the
change in growth rate was 1.9 percent per .01 ppm change in
ozone.
Forest Values
The changes in growth patterns can be transformed into
changes in optimal harvest rates, given there exists a particular
size of tree that is desirable to harvest when maximizing
revenues. Producers may delay harvests to allow for the tree to
mature to the size it would reach if no ozone effects were
present. Producers may also find that they cannot wait for the
tree to meet full maturity and therefore harvest a smaller tree.
In either event the effects of ozone can result in a reduction in
social welfare over time. As with agricultural crops, the
existence of market distortions in the lumber business, should
they exist, would need to be reflected in the calculation of
welfare changes attributable to changes in ozone concentrations.
Materials Effects and Values
Ozone has been shown to damage products made with certain
elastomers. Tires exposed to ozone concentrations observed today
in urban areas can lead to cracking and a reduction in
flexibility that leads to premature changing or retreading of
tires, or attempts on the part of tire manufacturers to introduce
anti-oxidant materials into tires. The McCarthy analysis
calculated the required levels of anti-oxidants that would need
to be introduced into rubber tires for alternative annual average
D-22
-------
ozone concentrations. McCarthy considered tire-use life, oxidant
use per tire, and annual tire production based on national
statistics. In addition, several studies have shown that larger
ozone concentrations may influence the expected life of dyes,
fabrics, and other materials. Collectively, these damages are
reflected in the concentration-response function reported below:
change in annual per capita materials damages due to ozone
= $ 28.8 x (change in annual average ozone concentration (ppm)).
BENEFITS OF REDUCING TSP AND VOC EMISSIONS IN BALTIMORE
Introduction
Based on the methodologies used to estimate risks from TSP
and ozone, and the subsequent values of reducing those risks, we
estimated the potential benefits of reducing TSP and ozone
concentrations in the Baltimore area. Because the cost analysis
was prepared in a manner that allowed comparisons of costs of
pollution reduction on a per ton basis, it was determined that
the best way to incorporate economic benefit information into the
analysis would also require that benefits be calculated on a per
ton basis.
However, most of the equations used to calculate benefits of
TSP and ozone reduction required that benefits first be
calculated using changes in ambient concentrations. Therefore,
information on current ambient concentrations and EPA standards
were used to formulate the necessary changes in ambient
concentrations. After calculating the necessary percentage
changes in ambient concentrations that must occur to meet the TSP
and ozone standards, baseline information on particulate matter
and volatile organic compound (VOC) emissions from sources in the
Baltimore area were used to calculate the necessary changes in
emissions to meet the desired change in ambient concentrations.
Combining the dollar benefits estimated with changes in ambient
concentrations with the emission reductions necessary to achieve
the change in ambient concentration, provided a dollar benefit
per ton estimate for both TSP and ozone. The techniques use to
calculate changes in ambient concentration and emissions are
described below in greater detail.
Benefits from Changes in Total Suspended Particulate Matter
Methodology
To calculate dollar benefits per ton for TSP reduction, the
analysis made several assumptions about emissions and ambient
concentrations in the Baltimore area. The first assumption was
D-23
-------
that a given percentage reduction in ambient concentration could
be achieved through an equal percentage reduction in emissions.
For example, if it was necessary to reduce average annual ambient
concentrations by !%/ then a reduction in annual emissions of 1%
would achieve this objective. Information on the total number of
particulate emissions in the Baltimore area were used to
calculate the baseline emissions, and a 1% change in the baseline
was used to calculate the number of tons of particulate matter
that would have to be reduced to meet that objective. In this
case, we used an annual emissions estimate of 17,789 metric tons
per year (mtpy) provided by Versar. Therefore, a 1% reduction in
emissions would require a reduction of 178 mtpy.
Using the 1% reduction in ambient concentrations with the
health and welfare effects equations and valuation methods
outlined earlier, we then calculated the total expected benefits
that would result from a 1% reduction in TSP concentrations. The
total benefits of a 1% reduction in TSP in Baltimore was
estimated to be $2.5 million per year. Dividing this dollar
benefit by the number of tone necessary to achieve these benefits
(178 mtpy) gave us an estimated dollar benefit per ton of $13,800
per mtpy. This benefit estimate was then applied to any
projected reduction in particulate matter emissions discussed in
the emission control scenarios.
The benefit per ton value does not include the benefits from
reduced carcinogenic risks. In order to calculate the additional
benefits of reducing carcinogenic risks, the expected reduction
in cancer risks, measured as fewer expected cases per year
(cumulative cancer risk), was multiplied by $2 million per cancer
case avoided. As discussed in the valuation section, this value
is estimated using willingness-to-pay estimates of small changes
in risks, which are then extrapolated to cases (i.e., $2 for 1 in
1 million risk leads to $2 million for 1 in 1 risk). As these
cancer risks are statistical in nature (i.e., we have not
identified the particular individuals expected to have cancer),
this method is suitable to use. If we can identify the
individual(s) who will contract cancer, then this method is
inappropriate.
Limitations
Because of the relative simplicity of the method used to
ascribe benefits to particulate matter emission reductions, we
should note that there are several issues to consider when using
this approach. Some of the benefits equations used to draw
connections between TSP and effects were estimated under
conditions atypical to those experienced in Baltimore. Several
of the relationships have been drawn during high TSP
concentration episodes. There remains some uncertainty as to
whether there are threshold concentrations, below which health
and welfare effects are reduced. We used threshold information
D-24
-------
where pertinent, given Baltimore's ambient conditions, but some
uncertainty remains.
We also did not directly link the location of emission
reductions to exposed populations, except for the carcinogenic
benefit estimates. Most of the benefit estimates are based upon
aggregate (area-wide) average changes in emissions and ambient
concentrations. Most of the equations were not developed using
detailed emission source information. Therefore, drawing
relationships between average changes in ambient concentrations,
and site-specific changes in emissions may serve to bias the
benefit estimates. We should note, that the benefits could be
either understated or overstated for any of the measured effects
(e.g., mortality, morbidity). The location of the sources
relative to populations, while taking account of dispersion
characteristics and behavioral patterns, will provide some
indication of the direction of the error.
Benefits from Changes in Ozone
Methodology
The method used to draw the connection between changes in
ozone concentrations and economic benefits is somewhat comparable
to that used for TSP, although there are several important
differences. The benefit equations required changes in ambient
ozone concentrations, but the control strategies reported changes
in VOC emissions. In order to reconcile these different units,
we were forced to make several assumptions.
Previous work prepared by EPA had drawn relationships
between changes in VOC emissions and changes in ozone
concentrations. Work prepared by Ostro and McGartland (1985)
estimated that a 1% reduction in VOC emissions in an urban area
would, on average, reduce peak ambient ozone concentrations in
the urban area by 0.6%. Therefore, in order to obtain a 1%
reduction in peak ozone readings, VOC emissions would have to be
reduced by about 1.7%. Using Versar's estimate on the current
amount of VOC emissions in the area (94,017 mtpy), a 1% reduction
in peak ozone readings requires a 1570 mtpy reduction in VOC
emissions. Using the benefits equations outlined earlier in the
study, we then calculated the total benefits resulting from a 1%
reduction in peak ozone concentrations. We estimated that the
total benefits of a 1% reduction in ozone concentrations would be
$540,000. Dividing this benefit estimate into the fewer tons of
VOCs emitted necessary to achieve this benefit, resulted in an
estimated dollar benefit per ton of $345 per mtpy. This benefit
estimate was then applied to VOC emission reductions analyzed in
the control scenarios.
The carcinogenic risks of VOC emissions were not included in
this benefit per ton estimate. The procedure used to calculate
D-25
-------
cancer reduction benefits used with TSP emissions was duplicated
for VOC emissions. A value of $2 million per cancer case avoided
was used in conjunction with the non-carcinogenic benefits
accounted for in the $345 per mtpy value.
Limitations
The approach used to draw connections between VOC emissions
and ozone concentrations suffer from the same problems
experienced for particulate matter emissions and ambient TSP
concentrations. Some of the health effects studies were
conducted in settings that are not normally present in the
Baltimore area. The existence of threshold values was considered
in the analysis, but the specific conditions of Baltimore may
result in an overestimate or underestimate of the benefits of
reducing ozone concentrations. The relationship between VOC
emissions and ozone concentrations is dependent upon many factors
which were not controlled for to capture Baltimore's specific
conditions. The relationship between VOC emissions and ozone
concentrations depends upon many variables, but the relationship
used in the analysis was felt to be representative of average
conditions found in urban settings.
The contribution of the specific VOC emission sources
examined in the analysis to ozone concentrations and risks may be
more direct than that attributed to the TSP emissions and risks.
Since the VOC emissions are likely to affect a broader area given
the atmospheric characteristics of ozone formation, the
relationship between VOC emissions and ozone health and welfare
effects is likely to be stronger than that assumed to exist
between particulate matter emissions and TSP health and welfare
effects.
Conclusions
To calculate the economic benefits of reducing particulate
matter and VOC emissions, we made use of many established
methodologies that attempt to draw linkages between exposures to
pollutants and effects. After establishing the extent of the
effect, we then use economic theory and techniques to calculate
the economic value of preventing these effects from occurring.
This approach allows us to estimate the economic benefits of
reducing pollution emissions, which can then be compared against
the economic costs of installing and operating equipment, or
making other changes in production processes.
We have estimated that each ton of TSP that is prevented
from entering Baltimore's air results in an average savings of
$13,700. Similarly, each ton of VOC that is not emitted into the
air results in an average savings of $345. Both of these
estimates do not include the value of reducing carcinogenic risks
D-26
-------
identified with these emissions. A method is available to add
these benefits to the non-carcinogenic benefits in order to
calculate the total benefits of reducing pollution emissions from
each of the sources identified in the air toxics study.
To conclude, we illustrate an example of using this
information to calculate the economic benefits of reducing
emissions from one of the sources identified in the study.
Woodstoves used for residential heating emit particulate and
VOCs, for which some of these components contribute cancer risks.
The control option proposed would be to discontinue using
woodstoves for heat, and switch to other conventional forms of
heat (e.g., gas, electric).
By eliminating residential woodstove emissions, annual TSP
emissions would be reduced by 93.6 mtpy and VOC emissions would
be reduced by 148 mtpy. The cancer risks associated with these
emissions is calculated to contribute 0.033 cancer cases per
year. Using the TSP and VOC benefit per mtpy values ($13,700 and
$345, respectively), the annual non-carcinogenic benefits of
reducing TSP emissions are $2,127,769, and the annual non-
carcinogenic benefits of reducing VOC emissions are $32,292.
Using $2,000,000 per reduced annual cancer case, the cancer
benefits are valued at $66,877. The total benefits are
$2,126,769. This can be compared against the annual costs of
limiting the use of woodstoves, which is estimated to be $220,000
per year. Therefore, the net benefits (benefit minus costs) are
$1,906,769 per year.
To compare this control option with other possible control
options, it is desirable to calculate the benefit-cost ratio
(benefits divided by cost). Comparing ratios across control
options allows one to examine which additional expenditure
(control cost) provides the greatest additional benefit (benefit
estimate). If there is no limit on the amount that can be spent
to control emissions, it is economically efficient to first adopt
those control options that provide the greatest incremental
benefits (i.e., have the highest benefit-cost ratio). The
benefit-cost ratio for the control of residential woodstoves does
not have the largest benefit-cost ratio, but its ratio of 9.9 is
one of the higher ones among the control options examined in the
study.
D-27
-------
REFERENCES
Chestnut, L.G., R.D. Rowe, and J. Murdoch. 1987. "Review of
Establishing and Valuing the Effects of Improved Visibility
in Eastern United States." Energy and Resource Consultants,
Inc. prepared for the Office of Policy Analysis, U.S.
Environmental Protection Agency. Contract No. 68-01-7033,
October.
Ferris, B.6. Jr. 1978. "Health Effects of Exposure to Low
Levels of Regulated Air Pollutants: A Critical Review,"
Journal of the Air Pollution Control Association. 28 (May):
482-97.
Fujinaka, L.E., D.M. Hyde, C.G. Plopper, W.S. Tyler, D.L.
Dungworth, L.O. Lollini. 1985. "Respiratory Bronchitis
Following Long-Term Ozone Exposure in Bonnet Monkeys: A
Morphometric Study," Experimental Luna Research. 8:167-190.
Hammer, D. I., V. Hasselblad, B. Portnoy, and P.F. Wehrle. 1974.
"Los Angeles Student Nurse Study: Daily Symptom Reporting
and Photochemical Oxidants," Archives of Environmental
Health. 28: 255-260.
Hasselblad, V., and D. Svendsgaard. 1975. "Reanalysis of the
Los Angeles Student Nurse Study," staff report, Health
Effects Research Laboratory, U.S. Environmental Protection
Agency, Research Triangle Park, N.C.
Heck, W.W., O.C. Taylor, R. Adams, G. Bingham, J. Miller, E.
Preston, and L. Weinstein. 1982. "Assessment of Crop Loss
for Ozone," Journal of the Air Pollution Control
Association. 32 (4): 353-61.
Heck, W.W., R. Adams, W.W. Cure, A.S. Heagle, H.E. Heggestad,
R.J. Kohut, L.W. Kress, J.O. Rawlings, O.C. Taylor. 1983. "A
Reassessment of Crop Loss from Ozone," Environmental Science
and Technology. 17 (12): 572-81.
Holguin, A.H., P. fluffier, C. Constant, T. Stock, D. Kotchmar, B.
Hsi, D. Jenkins, B. Gehan, L. Noel, and M. Mei. 1984. "The
Effects of Ozone on Asthmatics in the Houston Area," Air
Pollution control Association Transactions on Ozone/Oxidants
Standards. Houston, Texas, November, 1984: 262-280.
Kopp, R.J., W.J. Vaughan, and M. Hazilla. 1983. Agricultural
Sector Benefits Analysis of Ozone: Methods Evaluation and
Demonstration. Prepared by Resources for the Future for
Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency, Research Triangle Park,
N.C. September.
D-28
-------
REFERENCES (CONTD.)
Krupnick, A.J., J.R. Kurland, and T. Narel. 1986. A Preliminary
Analvaia for the Control of Photochemical Oxidants.
Prepared by Resources for the Future for Office of Air
Quality Planning and Standards, U.S. Environmental
Protection Agency, Research Triangle Park, N.C., Contract
No. 68-02-3844.
Mazumdar, S., H. Schimmel. and I.T.T. Higgins. 1982.
"Relationship of Daily Mortality to Air Pollution: An
Analysis of 14 London Winters, 1958/59 - 1971/72." Archives
of Environmental Health 37 (July/August): 213-20.
McDonnell, W.F., D.H. Horstman, M.J. Hazucha, E. Seal, E.D. Haak,
S. Salaam, and D.E. House. 1983. "Pulmonary Effects of
Ozone Exposure During Exercise: Dose-Response
Characteristics," Journal of Applied Physiology. 54 (5)i
1345-1352.
McCarthy, E.F., et al. 1983. Damage Coat Model for Pollution
Effects on Material. U.S. Environmental Sciences Research
Laboratory, Research Triangle Park, N.C.
McKee, D. and J. Graham. 1987. "Estimated Dose-Response Function
for Chronic Respiratory Disease Associated with Ozone
Exposures," Staff Paper, Office of Air Quality Planning and
Standards, U.S. Environmental Protection Agency, Research
Triangle Park, N.C.
Ostro, B.D. 1983. "Urban Air Pollution and Morbidity: A
Retrospective Approach," Journal of Environmental Economics
and Management. 10 (December). 371-82.
Ostro, B.D. 1987. "Air Pollution and Morbidity Revisited: A
Specification Test'" Journal of Environmental Economics and
Management. 13 (December).
Portney, P.R. and J. Mullahy. 1986. "Urban Air Quality and Acute
Respiratory Illness," Journal of Urban Economics. 20
(July): 21-38.
Rowe, R.D. and L.G. Chestnut. 1985. Qxidants and Asthmatics in
Los Anaeles: A Benefits Analysis. Prepared by Energy and
Resource Consultants, Inc. for Office of Policy Analysis,
U.S. Environmental Protection Agency, EPA-230-07-85-010.
Wa s hington, D.C.
D-29
-------
REFERENCES (CONTD.)
Samet, J.M., Y. Bishop, F.E. Speizer, J.D. Spongier, and E.G.
Ferris. 1981. "The Relationship Between Air Pollution and
Emergency Room visits in an Industrial Community," Journal
of the Air Pollution Control Association. 31 (March): 236-
40.
Schwartz, J. and A. Marcus. 1986. "Statistical Reanalysis of
Data Relating Mortality to Air Pollution During London
Winters 1958-72." Working Paper, U.S. Environmental
Protection Agency, Washington, D.C., October.
Trijonis, J. 1982. "Visibility In California," Journal of the
Air Pollution Control Association. 32 (February): 165-69.
Violette, D.M., and L.G. Chestnut. 1983. Valuing Reductions in
Risks; A Review of the Empirical Estimates. EPA-230-05-83-
003. Prepared by Energy and Resource Consultants to Office
of Policy Analysis, U.S. Environmental Protection Agency,
Washington, D.C., June.
Violette, D.M., L.G. Chestnut, and A. Fisher. 1986. "Valuing
Risks to Human Health," Toxics Law Reporter. September.
Wilhour, R. 1986. "Benefit to Forests of Reducing Ambient Ozone
Concentrations," Memorandum to Office of Air Quality
Planning and Standards and Office of Policy Analysis, U.S.
Environmental Protection Agency, Washington D.C. September
22, 1986.
D-30
------- |