c/EPA
             United States
             Environmental Protection
             Agency
            Region 5
            Air and Radiation Division
            230 South Dearborn Street
            Chicago, Illinois 60604
January 1989
Estimation and
Evaluation of Cancer
Risks Attributed to Air
Pollution in Southeast
Chicago
   DRAFT
                                    905R89103

-------
                          REQUEST FOR PUBLIC COMMENTS


The United States Environmental  Protection Agency is soliciting public comments
on this draft report.  Comments submitted by March 31.1989, will  be considered
in preparing the final report.  Comments should be submitted to:

                         John Summerhays (bAR-26)
                         U.S. Environmental  Protection Agency
                         Air and Radiation Division
                         230 South Dearborn  Street
                         Chicago, Illinois  60604

This report makes reference to various supporting documents, particularly two
reports documenting the emissions inventory  used in this risk assessment.
These reports may be obtained by writing Mr. Summerhays at the above address or
calling him at (312) 886-6067.

-------
         Estimation and Evaluation of



         Cancer Risks Attributable to



      Air Pollution in Southeast Chicago
                    DRAFT
               John Surnmerhays



          Air and Radiation Division



United States Environmental Protection  Ayency



                   Region V



              Chicago , 111 inois
                January  1989

-------
                                      111

                             Acknowledgements


A study of this magnitude is not completed by a single individual.  This
report reflects knowledge possessed by numerous people with expertise on
various source types and pollutants.  Both in the technical development of
emissions and risk estimates and in the  documentation of this study, the
assistance and advice from many people made this study far better than would
otherwise have been possible.

Particularly noteworthy are the contributions by Tom Lahre, of the
Noncriteria Pollutant Programs Branch of the Office of Air Quality Planning
and Standards (OAQPS).  Through Tom's arrangements, the Noncriteria
Pollutant Programs Branch provided contractual  assistance for the dispersion
and risk analyses' in this study, without which this study would not have
been possible.  Tom also provided valuable information, feedback, and
comments in both the emissions estimation and risk analysis phases of
this study.

The Illinois Environmental Protection Agency and the Indiana Department of
Environmental Management made important  contributions to this study.  These
agencies sent out questionnaires to industrial  facilities and supplied key
information used in the study.

The author wishes to acknowledge important assistance from other employees of
Region V working on this study.  Dr. Harriet Croke compiled emissions estimates
for many industrial facilities and managed a contract to develop emissions
estimates for waste handling.  Also assisting in developing emissions estimates
were Barry Bolka and March Klevs.  Special appreciation is also extended to
Carole Bell and Melody Noel  who typed this report.

Several other individuals made significant contributions.  Dr. Milton Clark,
of Region V's Office of Health and Environmental Assessment, provided useful
advice and comments on the report.  Fred Hauchman, of the Pollutant Assessment
Branch of OAQPS, served an important role as a central  source of information
on unit risk factors.  Dr. Ila Cote, also of the Pollutant Assessment Branch,
provided significant comments and feedback on health impact assessment.  Loren
Hall, of the Design and Development Branch of the Office of Toxic Substances,
provided useful information and constructive comments in both the emissions
estimation and risk analysis phases of the study.  Jacob Wind, of American
Management Systems, provided contractual  assistance in loading and refining
PIPQUIC, a data handling system for urban risk assessments.  Chuck Vaught, of
Midwest Research Institute, provided contractual assistance in assessing
emissions from waste handling facilities.  Valuable review and comments were
provided by Penny Carey (Office of Mobile Sources), Cheryl Siegel-Scott (Office
of Toxic Substances).  Finally, a lengthy list of other individuals contri-
buted other information on emissions from particular source types or on other
aspects of the study.

-------
                                        IV


                              TABLE    OF   CONTENTS
Sect_i_qn                                               Pjjge


Tables                                                  v


Figures                                                vi


Summary                                               vii


Introduction                                            1


Study Design                                            3


Emissions Estimation                                    7
Estimation of Concentrations by                        14
 Atmospheric Dispersion Modeling
Comparison of Model iny and Monitoring                  16
 Concentration Estimates
Evaluation of Cancer Risk Factors                      25


Incidence and Risk Estimates                           30


Conclusions                                            45


References                                             50

-------
                                         V

                                       TABLES


Number                                                                      Paje

 la.  Emissions in Source Area by Source Category and Pollutant               12
 Ib.  Other Substances in Study                                               13

  2.  Monitoring Studies Conducted in Southeast Chicago                       20

  3.  Comparison of Modeled-Versus Monitored-based Concentration
      Estimates for Organic Toxicants                                        21

  4.  Comparison of Modeled-Versus Monitored-based Concentration
      Estimates for PCBs                                                     23

  5.  Comparison of Modeled-Versus Monitored-based Concentration
      Estimates for Particulate Toxicants                                    24

  6.  Carcinogenicity of Inventoried Pollutants                               28

  7.  Contributions to Area Cancer Cases by Source Type and
      Pollutant Across the Study Area                                        33

  8.  Estimated Contributions to Lifetime Cancer Risk at the Grid
      with the Highest Estimated Number of Cancer Cases                      43

-------
                                          VI

                                       FIGURES

 Number                                                                     Page


 A. Contribution to Estimated Annual Cancer Cases by Source Type             ix

 la. Southeast Chicago Study Area - Source Area                               4
 Ib. Southeast Chicago Study Area - Receptor Area                             5

 2. Map of Estimated Coke Oven Pollutant Concentrations                      17

 3. Map of Concentrations of Polycyclic Organic Matter                       18

 4. Contributions to Estimated Annual  Cancer Cases by Source Type            32

 5. Relative Distribution of Estimated Lifetime Cancer Cases                 34

 6. Breakdown by Source Category of Contributions to Estimated
     Cases                                                                   35

 7. Contributions to Estimated Cases from Consumer-
     oriented Sources                                                        37

 8. Contributions to Estimated Annual  Cancer Cases
     by Pollutant                                                            39

 9. Map of Estimated Lifetime Cancer Risks from Air Pollutants
     in Southeast Chicago                                                    40

10. Estimated Lifetime Cancer Risks from Air Pollutants                      41

11. Contributions to Estimated Risk at the Peak Incidence Location           44

-------
                                    SUMMARY

Increasing concern has developed that air pollution may cause significant
cancer risks in urban areas due to the combined effects of multiple sources and
multiple pollutants.  Given the density of exposed populations in urban  areas,
the possibility of high risks would further suggest that the number of incidences
of resulting cancer cases may also be relatively high.  The Southeast  Chicago
area has both a substantial concentration of industrial  and non-industrial
emission sources and a relatively high population density exposed to these
emissions.  This study was undertaken to evaluate the extent to which  this
exposure to ambient (outdoor) air contaminants may be a public health  problem
and to provide an informed basis for determining what emission reductions,  if
any, might be warranted to reduce the exposure.

The study sought to use as broad a base of information as possible in
evaluating air pollution-related cancer risks in the Southeast Chicago area.
The study considered every air toxicant for which the United States Environ-
mental  Protection Agency (USEPA) can estimate a quantitative relationship
between the exposure to the air toxicant and the resulting increase in the
probability of contracting cancer.  All source types for which emissions of the
identified pollutants could be quantitatively estimated were included.  Estimates
were made of emissions in a relatively broad area, so that impacts both  from
nearby sources and from more distant sources could be included.

The National Academy of Sciences has defined risk assessment as a process having
four steps: hazard identification, exposure assessment, assessment of  dose-
response relationships, and risk characterization.  The hazard identified  for
assessment in this study is cancer due to ambient air contamination.  The
exposure assessment principally involves estimation of ambient atmospheric
concentrations, which, for most pollutants, were estimated by first deriving an
inventory of emissions, and then estimating atmospheric dispersion of  these
emissions.  The assessment of dose-response relationships involves derivation
of a unit risk factor, which expresses the probability or risk of contracting
cancer that is associated with exposure to a unit concentration of air pollution.

Finally, risk characterization involves deriving various measures of risk.   The
simplest measure of risk is individual risk, representing the risk attributable
to air contaminants at a specific geographic location.  An alternative measure
of risk is the number of cancer cases attributable to air contaminants estimated
to occur among the population in the study area.  In addition to estimating
these general  measures of cancer risk, this study also investigated the  origins
of these risks and incidences, i.e., which source types and which pollutants
are the most significant probable causes of these individual  and area-wide
risks estimated to result from air pollution in the Southeast Chicago  area.

It must be noted that the risk estimates presented in this report should be
regarded as only rough approximations of total cancer cases and individual
1ifetime risks, and are best used in a relative sense.  Estimates for  indivi-
dual pollutants are highly uncertain and should be used with particular  caution.

-------
                                      VI 11

This study found atmospheric emissions of 30 pollutants in the study area which
USEPA considers to be carcinogens.  Some  of these pollutants have been shown
to be carcinogenic based on human exposure data, and others have been impli-
cated by animal studies.

The cumulative total  number of cancer cases that this study estimated to be
attributable to air pollution is about 85 cases over 70 years or about 1 per
year.  The area for which exposure was assessed has a population of about
393,000 residents.  Therefore, the average risk across the area due to air
pollution as estimated by this study is approximately 2.2x10'^, or about 2
chances in 10,000.  It should be noted that, as a national average across
the United States, the chance of contracting cancer over a lifetime from a
number of factors (including both voluntary and involuntary exposures) which
are not fully understood, is about one chance in three.  One in seven people
die from cancer.  "

Several  types of sources appear to contribute significantly to the cancer cases
estimated to result from air pollution in Southeast Chicago.  Figure A is a
pie chart of the contributions of various source types to cancer cases in the
area.  The most significant source type is steel mills, particularly the coke
ovens found at steel  mills.  Steel mills  appear to contribute almost 34% of
the total  estimated cancer incidence.  Emissions from other industrial facili-
ties, primarily chrome platers, are estimated to cause approximately 16% of
the incidence.  Consumer-oriented area sources (e.g., home heating and gaso-
line marketing) contribute about 14%, and roadway vehicles are also estimated
to cause about 14% of the total cancer cases.  Furthermore, the background
pollutant impacts from formaldehyde and carbon tetrachloride contribute almost
the entire remaining  22%.  Together, these source types account for about
99.8% of the estimated air pollution-related cancer risks in the area.

This study also provides useful information on what source categories in the
area make only minor  contributions to the total  estimated cancer risks.  In
terms of estimated contributions to overall  area cancer incidence, wastewater
treatment plants contribute 0.1% of the total , and facilities for the handling
and disposal  of hazardous and non-hazardous waste (including landfills, two
hazardous waste incinerators, and liquid  waste storage tanks) also contribute
0.1% of the total.  Thus, these facilities are clearly estimated to cause much
less risk in the Southeast Chicago area than the more dominant source types
discussed previously.

It is useful  to apportion the estimated total  number of cancer cases according
to the weight of evidence that the pollutants are carcinogenic.  According to
USEPA's review of the weight of evidence  of carcinogenicity, the 30 pollutants
for which risks were  estimated in this study include 6 "known human carcinogens",
22 "probable human carcinogens, and 2 "possible human carcinogens".  Of the
estimated 85 cancer cases per 70 years, almost 53% are attributable to pollu-
tants that USEPA labels "known human carcinogens," about 47% are attributable
to "probable human carcinogens," and about  0.02% are attributable to "possible
human carcinogens."

This study also estimated lifetime individual  risks in an array of locations.
A peak lifetime risk  of about 5xlO~3 (or  about 5 chances in 1,000) is estimated

-------
      0
      Q.
      0)
      O

      3
      O
(0
      0)
      CD
      (ft
      (Q
     O
      o
      c
      CO
     O

     "5

      c
>•   I
§   1
11
O   LU
<   2
2   (0
z   c
o   o
CO   3
CO

-------
in the study area.  However, available Census Bureau information does not
indicate any residents in this area.  The square kilometer with the highest
estimated number of cancer cases has an estimated lifetime risk of about 1x10"^
(1 in 1,000).  In general, risks are greatest in the northeast part of the
area and are relatively lower in the southern and western part of the area.
The average lifetime risk across the area is about 2.2xlO~^ (about 2 in  10,000).

Consideration of the results of this study should include consideration  of
various uncertainties inherent in the study.  The estimation of emissions
generally relies on extrapolation of studies of emission sources elsewhere to
the sources in the Southeast Chicago area.  In addition to uncertainties in
quantitative emissions estimates, there is also qualitative uncertainty  since
we may not be aware of some sources and source types for some pollutants.
Atmospheric dispersion modeling also introduces uncertainty in the estimation
of ambient (outdoor) concentrations.  Finally, there are significant uncer-
tainties in the unit risk factors used in this study, due to the necessity for
various extrapolations from the exposure conditions in the studies deriving
the risk factors to the exposure conditions in the Southeast Chicago area.

It is difficult to judge whether the risks in this study are more likely to be
underestimated or overestimated.  Comparison of monitoring data to the modeling
data used in this study suggests that most pollutants are reasonably well
addressed, but some pollutants appear underestimated.  Thus, this comparison
suggests that actual  risks may in fact be higher than indicated in this  study.
Conversely, the conservatism underlying the unit risk factors used in this
study implies that actual  risks may be lower.  Both types of uncertainty appear
to be relatively modest for some pollutants and relatively major for other
pollutants.  Thus, the risk estimates derived in this study may either overstate
or understate actual  risks.

This study did not evaluate routes of exposure to environmental  contaminants
other than ambient air pollution.  While most if not all  the water consumed in
the area is from Lake Michigan, and not groundwater, drinking water is another
potential  source of risk.  Other environmental  exposures include indoor  air
pollution (including radon gas), fish consumption and dermal  exposure.
Further, there may be other potential carcinogens or source categories which
have not yet been identified.

This study identifies various aspects of air toxics exposure in Southeast
Chicago that warrant further study.  Several  such investigations are currently
underway.

At the same time, the study suggests that options for reducing risks due to
air pollution in Southeast Chicago should be investigated.  This study
identifies the source categories which contribute most to risk in the area and,
therefore, most warrant control.  The States and USEPA are working toward
regulating several  of the important source types that this study indicates
are significant.  It is hoped that this study will  form a basis for further
discussions concerning the reduction of cancer risks potentially attributable
to air toxic emissions in the Southeast Chicago area.

-------
Increasing national  attention has focused on the health risks from "toxic"
(non-criteria) air pollutants that arise in urban areas where a concentrated
level of industrial  activity coexists with high population density.  Within
Region V, an area that combines concentrated industrial activity with high  popu-
lation density is Southeast Chicago.  In particular, Southeast Chicago and  the
surrounding area is one of the nation's foremost locations for integrated steel
production and a wide range of other manufacturing activity.   This area also
has one of the nation's five facilities permitted for polychlorinated biphenyls
(PCB) incineration and has a variety of other facilities for  treating, storing
and disposing of hazardous waste.  Therefore, Region V of the United  States
Environmental  Protection Agency (USEPA), with assistance from the Illinois
Environmental  Protection Agency (IEPA)  and the Indiana Department of  Environ-
mental Management (IDEM), has completed an extensive study of air toxicants  in
the Southeast Chicago area.

The goal  of this study has been to obtain a broad understanding of the risks of
cancer that may be attributable to inhalation of ambient air  pollutants found
in the Southeast Chicago area.  The National  Academy of Sciences defines four
steps of risk assessments: hazard identification, exposure assessment, evaluation
of dose-response relationships for the  pollutants in the study, and estimation
and characterization of risk.  Hazard identification involves identifying an
exposure scenario, in this case inhalation of air contaminants, which may be
causing adverse health effects.  Exposure assessment involves evaluating
the ambient concentrations of the pollutants to which the public is exposed.
The principal  method for assessing exposure in this study is  to estimate emis-
sions and then estimate atmospheric dispersion of these emissions.  The evalua-
tion of dose-response relationships in  this study involves the estimation of
cancer risk factors, representing the cancer risk estimated to result from
breathing a unit concentration (e.g., one millionth of a gram per cubic meter
of air).  Finally, estimation and characterization of risk involves compiling
and analyzing all this information in a way that provides useful statements
about risk.

A more direct means  of considering the  impact of environmental  contaminants  on
cancer rates is to conduct an epidemiol ogical  evaluation of cancer statistics.
Unfortunately, due to the difficulties  of distinguishing environmental  factors
from other factors,  such studies are often inconclusive.  Further, such studies
generally do not even attempt to consider the separate influences of  the various
sources of the various environmental  contaminants.  The study described in this
report thus has different purposes from the purposes of epidemiological  studies.
Epidemiol ogical  studies, if conclusive, can provide a better  evaluation of the
correlation between  air pollution and cancer statistics.  However, this study
provides a more detailed data base on the potential  relative  significance of
different source types and different pollutants.  Further, due to the long
periods of exposure  that are considered to be involved in cancer induction,
current cancer statistics probably reflect exposures over the last several
decades.   In contrast, this study addresses cancer risks that USEPA methods
of risk assessment would associate with current air pollutant concentrations.
(This study may be considered to estimate future risks if air pollutant
concentrations were  to remain constant  at current levels for  the next several
decades.)   Furthermore, given the mobility of population in the United  States,
cancer statistics reflect exposure in multiple areas where members of the
studied population have lived.  In contrast, this study focuses specifically

-------
on estimated impacts of exposure to pollutant concentrations in  the  Southeast
Chicago area.  Thus, this study more serves the purpose of evaluating  which
source types and which pollutants are best addressed  in order to reduce  the
future cancer risks that current risk assessment methods suggest may result
from air pollution in the Southeast Chicago area.

This study may be considered in the context of national  concern  about  urban  air
toxics issues.  A USE PA report entitled The Air Toxics Problem in the  United
States: An Analysis of Cancer Risks for Selected Pollutants (dated May 1985)
estimates that as many as 1800 to 2400 cancer cases per year may be  attributed
nationally to air pollution (not including indoor radon).   This  report further
finds that while individual  industrial  operations may lead to high localized
risks, a much greater share of the cumulative risk from air toxicants  comes
from activities that are more population-oriented, such as driving motor
vehicles and heating (with fireplaces and wood stoves).  In fact, limited
monitoring data in some large cities indicates that risks  even in residential
and commercial  areas approach the risks found near the highest risk  industrial
facilities.  Further, various studies suggest that cancer  risks  from air
pollution throughout urban areas are commonly in the  range of lxlO~3 (i.e.,
1 case per thousand people exposed for a lifetime) to lxlO~4 (1  case in  10,000).
These risks arise from the multiple sources of emissions of multiple pollutants
that exist in all  urban areas.  Since 61% of the United States population  lives
in urbanized areas, and the exposure to high urban toxics  risks  extends
throughout these urban areas, this urban air toxics exposure appears to  contri-
bute the major share of the cases of cancer attributable to air  pollution.   The
purpose of the Southeast Chicago study, then, given the general  national picture
of urban air toxics risks, is to define, in more detail , the relative  contri-
butions of various source types to that risk in this  geographic  area.

Conducting a study like this requires substantial  computerized data  handling.
Data handling for developing emissions estimates required  specifically developed
computer programs.  Dispersion modeling, risk estimation,  and cancer incidence
estimation relied heavily on a data handling system known  as PIPQUIC (Program
Integration Project Queries Using Interactive Commands).  PIPQUIC also provided
many of the figures shown later in this report.

This report includes eight sections.  This introduction has focused  on the
context in which this study was conducted.  The next  section describes
several of the general  features of the design of this study.  The third
section summarizes the procedures and results of the  emissions inventory phase.
The fourth section describes the exposure assessment, particularly describing
the atmospheric dispersion modeling used as the principal  method for esti-
mating pollutant concentrations, and also providing a sampling of the  concen-
tration outputs of this study.  The fifth section compares the modeled concen-
tration estimates against concentration estimates based on monitoring.  The
sixth section describes the dose-response relationships (i.e., the health
impacts associated with given concentrations)  used to estimate risks.  The
seventh section then presents results of the risk estimations, discussing
the estimated magnitude of the cancer risk attributable to air pollutants,
the relative contributions of different source types  and pollutants, and the
spatial distribution of the risks over the studied receptor area.  The
final  section summarizes the conclusions of this study.

-------
Study Design

The first step in this study was to plan a study design.  A key decision here
was whether to develop a screening study covering multiple pollutants and
multiple source types using only readily available information or whether to
develop a more focused inventory investigating only a few pollutants and source
types.  This study was designed for screening purposes, to provide an overview
of excess cancer risks that may be attributable to ambient air pollution in  the
area.

This study has been designed to be comprehensive in several respects.  F:irst,
it has attempted to include all source types that emit any of the substances
being studied.  Second, although the focus of this study is on exposure in a
moderately sized area (approximately 65 square miles), a much broader area was
inventoried to i-nclude all  sources with potentially significant impacts in the
selected receptor area.  Third, this study attempted to address a comprehensive
list of potential carcinogens.

With respect to source types, this study included all source types for which
air toxics emissions could  be estimated.  A special  aspect of this study was
the inclusion of the volatilization from wastewater treatment plants, emissions
from hazardous waste treatment, storage, and disposal facilities (TSDF's), and
emissions from landfills for municipal waste.  Emissions from these source
categories are difficult to estimate and are not included in traditional  air
pollutant emissions inventories.  However, they were included in this study
due to national and local  interest in their relative contribution to risk.
Also included were source  types which have more traditionally been inventoried,
such as industrial  facilities, population-oriented sources (e.g., dry cleaning)
and highway vehicles.  Although a greater ability has been developed to estimate
emissions from these types  of sources, the derivation of emissions factors for
the substances inventoried  in this study nevertheless required substantial
literature research and then development of factors suitable for use in this
kind of inventory.   This study did not involve direct emissions measurements;
instead, emissions estimates reflected production rates of sources in the area
(e.g. tons of steel produced) in conjunction with results from various studies
of the relationship between production and emissions (e.g., pounds of emissions
per ton of steel  produced).

With respect to spatial coverage, Figure la is a map showing the broad "source
area" included in the inventory, and Figure Ib is a map showing the smaller
target "receptor area" for  the exposure analysis.  The focus of this study
is on air pollutant concentrations in the receptor area and on the cancer
impacts that exposure to these air pollutants in this area may cause.  However,
it is clear that the air quality in this area is affected by emissions that
can be transported in from  a much broader area.  Consequently, emissions were
inventoried for a much broader area.

For purposes of this study, the "Southeast Chicago" receptor area was defined
as an area that is approximately a 13 kilometer (8 mile) square, having a
total area of 169 square kilometers (65 square miles).  This area covers much
of the southeast corner of  the City of Chicago plus portions of adjoining
suburbs, ranging specifically from 87th Street to Sibley Boulevard and from
Western Avenue to the Indiana State line.  This area has a population of about
393,000.

-------
*    KISJ

-------
3NI1 31V1S SIONmi VNVICNI

-------
By comparison, the inventoried source area covers a 46 kilometer (about 29 mile)
square area, with a total area of 2136 square kilometers (about 817 square
miles).  Since the prevailing winds in the area are from the southwest quadrant,
the source area is skewed toward the south and west of the receptor area.  The
specific boundaries of the source area are, in terms of UTM (Universal  Trans-
verse Mercator) coordinates, from 4584 to 4630 kilometers northing and from 420
to 466 kilometers easting in zone 16.  This source area extends 30 kilometers
south and west and 16 kilometers north and east of the center of the receptor
area.  Thus, the emissions study area includes roughly a third of the City of
Chicago, most of the city's southern and southwestern suburbs, and a portion
of Northwest Indiana.  This source area has a population of about 2,361,000.
The inventory further includes a few additional  point sources outside of this
source area which were judged to be potentially significant sources.

With respect to pollutants, this study included all potential  carcinogens for
which a quantitative relationship between air concentration and risk has been
estimated.  During the initial  design of the study, unit risk factors had been
estimated for 47 of the 51 substances on the targeted pollutant list.  However,
further review led to the conclusion that for many of these 47 substances, the
evidence of carcinogenicity is too weak or the cancer risk factor estimates are
too unreliable to use in this study.  This further review concluded that 32
substances had reasonable evidence of being carcinogenic and risks could rea-
sonably be quantified.  Thus, the study list of 51 substances includes 15 sub-
stances which may or may not be carcinogenic, but could not be quantitatively
analyzed, and 4 substances that were included only on the basis of potential
noncarcinogenic impacts.  (As will be discussed below, all  but 2 of the 32
quantifiably carcinogenic pollutants were found to have atmospheric emissions
in the studied source area.)

Analysis of systemic, noncarcinogenic health effects was considered beyond the
scope of this study.  First, Agency-reviewed dose-response data for systemic
effects due to inhalation of air contaminants were not available at the inception
of this study.  Second, analysis of systemic health effects generally requires
consideration of concentration thresholds below which no adverse health effects
are observed.  Therefore, it is necessary to conduct a substantially different
and more complicated exposure assessment to evaluate the extent and frequency
with which the threshold may be exceeded.  Thus, this study focused on cancer
effects of the 32 pollutants with agency-reviewed risk factors.

As indicated above, this study primarily used emissions estimates in conjunction
with atmospheric dispersion modeling rather than using monitoring data to esti-
mate ambient concentrations of the pollutants being studied.  Both methods
have advantages and disadvantages as approaches for estimating ambient concen-
trations.  The advantages of modeling include the ability to address concentra-
tions across an entire geographic area, to address long term average concentra-
tions, and to estimate concentrations belo  the concentration levels that avail-
able monitoring methods can detect.  The c ..'responding disadvantages of monitoring
data are that resource constraints generally limit the collectable data to une
or a few locations and for relatively short time periods.  Additionally, moni-
toring methods are not available for some pollutants, and for other pollutants,
monitoring cannot detect some of the concentrations of interest.  A further

-------
                                       7

advantage of the emissions estimation/dispersion modeling  approach  is that  it
readily identifies the separate contributions of sources and source categories
to any given concentration, which monitoring data alone cannot do.   For  these
reasons, the emissions estimation/dispersion modeling  approach was  judged a
better means of evaluating concentrations throughout the study area and  judged
to be a more informative approach, particularly in describing  relative contri-
butions of different source types.  On the other hand, monitoring data have the
advantage that for the time and location being monitored,  and  if concentrations
are detectable, the uncertainties are generally less than  the  uncertainties
inherent in emissions inventorying and dispersion model ing .   For this reason,
monitoring data can be used to obtain a "reality check", to  suggest at least
for the locations and pollutants successfully measured whether or not the
modeled concentrations are approximately correct.

A further advantage of monitoring is the ability to assess concentrations (at
least if concentrations are above detection limits) of atmospheric  contamination
which is not the direct result of current emissions.  Conversely, a disadvantage
of the emissions estimation/dispersion modeling approach is  that this approach
is unable to consider such "background impacts".  For  most pollutants in this
study, "background concentrations" may be presumed to  be overwhelmed by  urban
area emissions, and such background concentrations may reasonably be ignored.
However, two pollutants in this study are presumed to  have origins  other than
current emissions: formaldehyde and carbon tetrachloride.  Although current
emissions of these pollutants contribute to ambient concentrations, most of the
ambient concentrations are attributable to other origins.  Much of  the formal-
dehyde concentration is presumed to be attributable to atmospheric  photochemical
reaction of other organics.  Since carbon tetrachloride remains unreacted in the
atmosphere for a very long time, current concentrations are  largely the  result
of an accumulation of historic emissions over wide geographic  areas.  Thus,
monitoring data were used in this study to indicate the concentrations of these
two pollutants from origins not addressed by the emissions estimation/dispersion
modeling approach.  The term "background pollutants" is used in this report to
identify these origins of risk.

Emi s s ion Est1 mat i on

The emissions  inventory is described in separate reports.  A detailed description
of the inventory is given in a July 1987 report entitled "Air  Toxics Emissions
Inventory for  the Southeast Chicago Area", authored by John  Summerhays and
Harriet Croke.  This report documents emissions estimates  for  a wide range  of
source types,  including source types that are traditionally  inventoried  in  air
pollution studies as well as some source types that are not  traditionally
inventoried such as volatilization from wastewater at  sewage treatment plants.
An addendum to this report (dated January 1989) updates this report by describ-
ing limited revisions to the previously described inventory  and by  describing
procedures and results of estimating air emissions from the  treatment, storage,
and disposal  of hazardous waste, and from landfills storing  municipal  waste.
Further details on the estimation of air emissions from the  handling of  hazar-
dous and nonhazardous waste are provided in two reports by the Midwest Research
Institute:  "Estimation of Hazardous Air Emissions in  Southeast Chicago  Contri-
buted by TSDF's", covering air emissions from the treatment, storage, and
disposal  of hazardous waste, and "Estimation of Hazardous  Air  Emissions  From

-------
Sanitary Landfills", covering air emissions from landfills for ordinary muni-
cipal  solid waste.  The reader interested in more details of the procedures,
data sources, and emissions estimates should consult these separate reports.
The discussion that follows will  present only an overview of the development
and results of the emissions inventory.

This study involved no direct measurement of emissions.  Instead, emissions
estimates in this study were generally based on local  activity levels (e.g.,
point by point steel  production or local  traffic levels)  in  conjunction with
the results of measurement studies elsewhere establishing the relationship
between activity levels and emissions (e.g., emissions per ton of steel  produced
or per mile driven).  This approach is used partly because emissions measurements
even just for the 88 industrial  facilities in this study  would be prohibitively
expensive, and partly because limited emissions measurements do not necessarily
provide representative long-term  data on emissions.

The sources considered in this study include industrial sources, consumer-
oriented sources (e.g. dry cleaning and gasoline marketing), roadway vehicles,
facilities for handling hazardous and municipal waste, and wastewater treatment
plants.  From another perspective, many of the industrial  sources as well  as
the waste handling facilities and the wastewater treatment plants are at clearly
identified locations, and are labeled "point sources," whereas other industrial
activities, as well as all of the consumer-oriented sources  and roadway vehicles,
are more broadly distributed, and are labeled "area sources."  The distinction
between point and area sources leads to the use of different methods for esti-
mating emissions.

For industrial point sources, three emission estimation methods were used.  The
first method may be labeled the questionnaire method.   Questionnaires were sent
to 29 companies considered candidates for being significant  sources of air
toxics emissions.  These questionnaires requested the  annual  emissions for each
pollutant in this study, as well  as stack data necessary  for dispersion modeling.
These questionnaires were sent by the Illinois Environmental  Protection Agency
and the Indiana Department of Environmental  Management.  Region V then reviewed
these company responses to assure that complete and reasonable emissions esti-
mates would be used for these facilities.  The second  method may be labeled
the species fraction method.  This method, used for 59 other identified facili-
ties, begins with estimates of emissions of total  organic emissions and total
suspended particulate emissions,  estimates which are based on the best available
information on plant operating rates and estimated emissions per unit operation.
This method then calls for multiplying these emissions totals times species
fractions, expressed as the ratios of the particular species emissions versus
the total  emissions, thereby estimating species emissions.  For example,
particulate emissions from blast  furnaces (e.g., Standard Classification
Code 3-03-008-25) were estimated  to be 0.013% arsenic, and so a blast furnace
casthouse that emitted 20 tons per year of particulate matter would be estimated
to anit 0.0026 tons per year of arsenic.  The third method may be labeled the
emission factor approach.  This method uses a direct emission factor, expressing
the quantity of a particular species emitted per unit  activity level  (e.g. per
1000 gallons of paint solids).  The emission factor is multiplied times the
actual  level  of activity to estimate total emissions.   This  method was only
used for one type of source (coke by-product recovery  plants), since for all
other point source types the direct emission factors were either not available
or the source types were not found in the Southeast Chicago  area.

-------
For area-type sources, both the species fraction method and the emission  factor
method were used.  As an example of the species fraction method, roadway  vehicles
were inventoried by multiplyiny total  emissions of oryanics times measured or
derived species fractions.  As an example of the emission factor method,  wood
combustion emissions were estimated by multiplyiny estimates of wood quantities
burned in fireplaces and wood stoves times an emission factor of the quantity
of the pollutant, polycyclic organic matter, per pound of wood burned.   The
companion emissions inventory reports  provide more details of the methods used
for each cateyory in this study, as well  as a discussion of the advantages and
disadvantages of the two methods.

A further issue to be addressed in inventorying area and mobile sources is the
spatial distribution of these emissions.   The impacts of given quantities of
emissions at any particular location are  a function of how distant and  how
frequently upwind the emission sources are from the impact location.  By
definition, area sources are collections  of sources too numerous and too  dis-
persed to identify the location of each source.  The solution to this problem
used in this study was to distribute emissions according to the distribution of
"surrogate parameters" such as population, housing, or manufacturing employment.
For example, it would not have been feasible to identify locations of the
estimated 2650 buildings with air conditioner cooling towers, not to mention
identifying the approximately 15% of those towers which use chromium as a
corrosion inhibitor.  Instead, these emissions were distributed in accordance
with the known distribution of nonmanufacturing , nonretail  employment.   Simi-
larly, roadway vehicle emissions on freeways and other roadways were distributed
according to traffic estimates for freeway and other roadway travel .

In addition to inventorying the above, which are relatively traditional  air
pollution source types, this study also included several  source types that
have not traditionally been included in air pollution inventories.  One such
source category is hazardous waste treatment, storage, and disposal  facilities
(TSDFs).  The Southeast Chicago study area includes a total of 43 facilities
regulated under the Resource Conservation and Recovery Act to handle hazardous
waste.  Included among these facilities is one of the nation's five  incinerators
of polychlorinated biphenyls (PCBs), a second incinerator handling non-PCB
hazardous waste, a hazardous waste landfill, several  facilities storing waste
in storage tanks, and a majority of facilities loading wastes into drums  or
trucks .

Estimating emissions for TSDFs required several  steps.  The first step  was
identifying facilities.  The second step  was obtaining data on the quantity of
each type of waste handled by each facility.  The third step was reviewing
studies of the composition of various  waste streams to estimate the  quantity of
individual  pollutants in the waste streams at each facility.  Finally,  emissions
estimation models were used, relying on the derived estimates of waste  quantities
and often relying on assumptions about operating procedures to estimate emissions
of each pollutant at each facility.  Most of these emissions estimates  were
derived by Midwest Research Institute  under contract to USEPA Region V, with
Region V deriving a few additional emission estimates.

A second type of facility which has not traditionally been included  in  air
pollution studies, but was included in this study, is municipal  waste landfills.

-------
                                      10

Biodegradation in landfills generates methane, and this methane can carry trace
amounts of contaminants contained in household and industrial  solid waste into
the atmosphere.

The first step in estimating these emissions was to review available data on
the contaminant concentrations found in gases emanating from landfills.   The
second step was to estimate landfill gas generation rates based on the estimated
volumes of landfill gases for each landfill  in the study area.   The third step
multiplied the results of the first two steps to estimate the  emissions  of each
species of concern from each landfill.  These estimates were again developed by
Midwest Research Institute under contract to USEPA Region V.

A third source type not traditionally included in air pollution studies  but
included in this study was wastewater treatment.  The focus in  this study was
on two wastewater treatment plants handling  the largest volumes of industrial
wastewater in the source area, i.e. the Calumet and the West-Southwest treat-
ment plants.  For each of these plants, the  Metropolitan Sanitary District of
Greater Chicago made measurements of the volatile organic concentrations in the
wastewater entering and exiting each of these facilities for seven consecutive
days.  Daily quantities of volatile organics were computed by  multiplying the
wastewater concentrations of each compound of interest times the respective
day's volume of wastewater, after which the  seven days' quantities were  averaged.
The next step of the analysis was to address the fate of these contaminants.
Possible fates for contamination in the influent wastewater include volatili-
zation to the atmosphere, biodegradation in  the treatment plant, sludge, and
treated wastewater leaving the plant.  Contaminants in the wastewater leaving
the treatment plant, where significant, were addressed by subtracting outgoing
contaminant quantities from incoming contaminant quantities.  Partitioning to
sludge was in all cases insignificant.  Nevertheless, volatilization from
sludge is included, insofar as sludge contamination was inventoried as if the
contaminants remained in the wastewater available to volatilize.  Most wastewater
contamination either volatilizes or biodegrades.  Based on studies measuring
volatilization and biodegradation for nonpolar organic solvents (the most
significant contaminants considered here) at other wastewater  treatment  facili-
ties, it was assumed that volatilization accounts for 40% of incoming contami-
nation (minus any adjustment for contamination in outgoing wastewater) and
biodegradation accounts for the remaining 60%.

This study also addressed several  other source categories which may be relatively
unimportant with respect to the "traditional" (criteria) air pollutants  but
which have the potential  to be significant with respect to toxic air pollutant
emissions.  While these categories generally emit relatively small  quantities of
the traditional  pollutants, the materials being emitted appear  to be highly
toxic.  Examples of such source categories included in this study are chrome
el ectropl aters (emitting chromium), wood combustion in fireplaces and wood
stoves (emitting polycyclic organic matter,  a component of "wood smoke", as a
product of incomplete combustion), and hospitals (emitting ethylene oxide
used in some sterilizing operations).

It should be noted that all  emissions estimates were, in general , compiled for
a 1985 base year.  A minor deviation from use of 1985 data is  the deletion of
sources which are known to have permanently  shut down since that time.  In

-------
                                      11

addition, the estimates compiled in this study are for typical  actual  emissions.
No attempt was made to evaluate emissions for the scenario in which all  plants
emit maximum allowable amounts, because this scenario is unlikely to persist
continuously over a 70 year lifetime.

An important influence on emissions from many source categories is the existence
of emission controls.  This study sought to develop emissions estimates
appropriate to 1985 levels of emission control .   A special  effort was  made
to assure that steel mill emissions estimates reflect the current status of
controls.  For other point sources, it is less clear whether emission  controls
adopted according to various regulations are, in fact, represented in  the emis-
sion estimates used in this study, though again, the goal  was to use emission
estimates that correspond to 1985 levels of control.  For roadway vehicles,
the emission estimates reflected elaborate, computer-assisted evaluation of
what portion of the vehicle fleet had what degree of emission control  as of the
1985 inventory date.  In particular, the MOBILE  3 emission factor model  was used
in conjunction with some updates for the consideration of evaporative  emissions.
It is noted that more recent information suggests that evaporative emissions may
be much higher due to "running losses."  For other types of sources, for the  few
source categories where emissions controls are in place, this study attempted
to use emissions estimates that reflect these controls.

One special element of the emissions inventory development was  the use of data
on facility emissions that Section 313 of the Superfund Amendments and
Reauthorization Act requires companies to submit.  In particular, companies are
required under this Section to develop and report emissions estimates  for
numerous pollutants including most of the pollutants in this study.  These
data were compared with the emissions estimates  that were independently  derived
in this study.  Unfortunately, these reports do  not address area sources.
Nevertheless, these data were used for additional  refinement of the industrial
source component of the Southeast Chicago area inventory.

Table la summarizes the emissions of known or suspected carcinogens found in
this study.  In the study area, 30 pollutants were found which  USEPA considers
carcinogenic.  This table distinguishes emissions from steel mills, other
industrial  operations, consumer-oriented sources, roadway vehicles, hazardous
waste treatment storage, and disposal  facilities (also including municipal
waste landfills), and wastewater treatment plants.  This table  shows that 30 of
the 32 known or suspected carcinogens were found to be emitted  in the  Southeast
Chicago study area.  The significance of the emissions shown here is best
interpreted in terms of risk assessment results, so this topic  will  be discussed
in the section discussing risk estimates.

As shown in Table Ib, this study found no emissions of allyl chloride  or radio-
nuclides.  This reflects the fact that either this study found  no methods for
quantifying emissions of these pollutants, or no sources were identified in this
area.  The emissions inventory phase of this study also attempted to include 19
substances without unit risk factors;   as described in the inventory reports,
13 of these 19 substances had quantifiable emissions in the study area.

A variety of uncertainties apply to the emissions inventory used in this study.
Emissions measurements were not conducted in the Southeast Chicago area, and so
it was necessary to apply emission facturs (i.e., emissions per unit operation)

-------
                                            12
    Table la.  Emissions in Source Area by  Source  Category  and  Pollutant (in metric tons/year)
Ac rylamide



Acrylonitrile



Arsenic



Asbestos



Benzene



Beryl 1ium



Butadiene



Cadmium



Carbon Tet.



Chi oroform



Chromium**



Coke Oven  Em.



Dioxin



Epichlorohydrin



Eth . Dibromide



Eth. Dichloride



Eth. Oxide



Formaldehyde



Gas. Vapors



Hex-chl-benz.



Methyl  Chi .



Methylene  Chi .



Perchloroeth.



PCB's
Steel
Mills


3.9

3044.2

.2
4.3


.07
388.0




14.6






Other
Industrial Consumer
Sources Sources
.02
1.0
1 .2
.02
55.2 37.1
.0008
.5
.2
.0003
.0003 31.1
2.5 .5

.0002
.09
54.6
61.5 11.2
12.6 110.0
216.2 4737.2
.07
.3 10.9
287.3 1084.0
383.7 802.0
.0002
Mobile Waste
Sources Facilities

.002

.04
812.8 12.0

73.1 .2
.02
2.7
.2


.0000007
.00002
.2

353.5 .04
14376.0
.5
.0003
61.9
.7
.001
Sewage
Treatment
Plants Total
.02
1.0
5.1
.06
.7 3962.0
.0008
74.0
4.6
2.7
.7 32.0
3.2
388.0
.0002
.09
.
.7 55.5
72.7
491.7
19329.2
1.3 1.8
.07 11.3
8.6 1441.7
6.0 1192.3
.001

-------
                                             13
Compound*

POM

Prop. Oxide

Styrene

Trichloroeth.

Vinyl Chi .

Vinylidene  Chi
                Table la.  (Continued)

                             Other
                  Steel       Industrial  Consumer
                  Mills      Sources     Sources
                                .02

                                .9

                              11.5

                             374.7

                               2.3

                                .4
16.9
Mobile
Sources

 8.0
                                                                          Sewage
                                                                 Waste     Treatment
icil ities


1.5
27.8
4.0
.8
Plants


2.4
1.9

.01
Total
24.9
.9
15.4
404.4
6.3
1.2
*Abbreviations :
       Carbon Tet .
       Eth.
       Gas.
       Hex-chl-benz.
                     - Carbon tetrachl oride
                     - Ethyl ene
                     - Gasol ine
                     - Hexachl orobenzene
           Chi .   - Chloride
           PCB's  - Polychlorinated biphenyls
           POM    - Polycyclic oryanic. matter
           Prop.  - Propylene
**Estimates are for hexavalent (+6) form of chromium.
                Table Ib. Other Substances in  Study
Substances without Unit Risk Factors
found in Southeast Chicago Area

     Acetone
     Diethanolamine
     Dioctylphthalate
     Ethyl Acrylate
     Ethylene
     Melamine
     Mercury
     Nickel
     Nitrobenzene
     Pentachlorophenol
     Titanium Dioxide
     Toluene
     Xylene
                                          Substances without  Unit
                                          Risk Factors  not  found

                                          Dimethylnitrosami ne
                                          Isopropylidene  Diphenol
                                          Methylene Dianiline
                                          Nitrosomorpholine
                                          Propylene Dichloride
                                          Terepththalic Acid
                           Substances with Unit
                           Risk Factors not found

                              Allyl Chloride
                              Radionuclides

-------
                                       14

measured elsewhere in estimating emissions in the Southeast Chicago area.   This
extrapolation from sources elsewhere to sources in the Southeast  Chicago area
is probably the greatest cause of uncertainty in the emissions inventory.   On
the one hand, this extrapolation is probably fairly good for some source types,
especially for area and mobile sources.  For example, roadway vehicles  in  South-
east Chicago are probably similar to the roadway vehicles in other places  in the
United States.  On the other hand, for other source types, source-to-source
differences in the raw materials used and differences in source operations may
yield significant differences in emissions, not just in  the quantity of emissions,
but even in whether particular substances are emitted at all.  A  second major
uncertainty is that some sources of some pollutants may  be missing in this
inventory either due to lack of awareness of the source  or source type  or  due
to unavailability of information with which to quantify  emissions.  This is
likely to be a particular problem for relatively unknown pollutants and for
pollutants that are difficult to measure.

Lesser uncertainties exist in various aspects of the emissions estimation
process.  Data used to estimate emissions in this study  include source  operating
rates, emission factors for particulate and organic emissions, data on  composition
of these emissions, data on the extent of emissions producing activities (e.g.,
pounds of wood combusted), and data used for area sources to spatially  distribute
these emissions.  For each of these types of data, the best reasonably  available
data were used, but even the best reasonably available data have  uncertainties
in their measurement and in their adequacy in representing emissions in the
Southeast Chicago area.

Estimation of Concentrations by Atmospheric Dispersion Modeling

The principal method used in this study to estimate concentrations is to model
the atmospheric dispersion of the emissions estimates described in the  previous
section.  Atmospheric dispersion is a function of several  factors.  From the
standpoint of selecting atmospheric dispersion models, two important factors
are the averaging times of the concentrations and the nature of the emissions
sources.  With respect to averaging times, some dispersion models are designed
to estimate short term average (e.g., 1 hour average) concentrations, and  other
models are designed to estimate long term average (e.g., annual average) concen-
trations.  The health effect being addressed in this study, cancer, is  most
appropriately addressed by evaluating lifetime cumulative doses (Cf. the "USEPA
Guidelines for Carcinogen Risk Assessment", 51FR33992).   Therefore, dispersion
models for estimating long term average concentrations were selected.   With
respect to the emissions sources, some dispersion models are designed to address
point sources (i.e., stacks or other similarly localized emission points), and
other dispersion models are designed to address area sources.  This study
includes both types of sources.  Therefore this study used one model  for point
sources and a second model  for area sources.

The models used in this study were the Industrial  Source Complex  Long Term
model (ISCLT) for point sources and version 2 of the Cl imatol ogical  Dispersion
Model (CDM-2) for area sources.  The two models reflect  the obvious differences
in initial  dispersion (e.g., the broad dispersion of area source  emissions even
at the moment of emission).  However, the degree of atmospheric dispersion
assumed in the application of these two models was the same.  One parameter  in
both models is the choice of dispersion coefficients. Separate sets of disper-
sion coefficients are available for urban versus rural areas to represent  the

-------
                                       15

degree of atmospheric mixing under various meteorological  conditions.  In this
study, for both models, Briggs' urban dispersion coefficients were used.  A
second parameter in both models is the meteorological  data used.  As a simpli-
fication in estimating long-term average concentrations, both models in this
study use stability array (STAR) data showing the joint frequency distribution
of winds in each of six classes of wind speed and six  classes of atmospheric
stability for each of 16 wind directions.  Both models estimate concentrations
for each wind speed/stability/wind direction category.  These models then
estimate an annual  average concentration by averaging  the category-specific
concentrations, weighted according to the frequency of each meteorological
category.  For both models, the meteorological  frequency distribution was based
on 1973 to 1977 data collected at Midway Airport, representing the nearest, most
recent, and most representative complete data set available.  Further, both
models assumed relatively flat terrain.  Finally, it should be noted that both
of these models are state-of-the-art models which are  routinely used for regu-
latory applications where estimates of atmospheric transport and dispersion are
necessary.  In fact, both of these models are reference models noted in USEPA's
Guideline on Air Quality Models (Revised), July 1986,  (EPA-450/2-78-027R).
Although this guideline does not address the pollutants in this study, the
study uses the models recommended in the guideline for the general  type of
modeling being conducted here.

The discussion of emissions estimation has noted that  point sources in this
study include steel mills, most other industrial  sources,  waste handling
facilities, and wastewater treatment plants.  That discussion also noted that
area sources include a few industrial source types (chrome platers, degreasing,
and barge loading), consumer-oriented sources,  and roadway vehicles.  This same
distinction applies to selection of a dispersion model  for addressing each source
type.  An important exception is that a selected set of steel  mill  operations
were simulated with a small  but finite initial  dispersion, reflecting the
modest area from which these emissions arise.  These emissions were simulated
using the area source algorithm of ISCLT.  For  example, a  typical  coke oven was
simulated by distributing emissions into three  neighboring 40 foot squares.
This approach was intended to simulate more realistically  the dispersion of
these emissions, and was used for coke ovens and for roof  monitors at steel-
making furnaces.  A second exception is chrome  platers.  In Illinois, it appeared
that a sufficient listing of electroplaters was available  to treat these emis-
sions as point sources, assigning the area's emissions to  the identified plater
locations.  This treatment has the advantage of providing  more realistic treat-
ment of the dispersion characteristics of these sources.  Note that in Indiana,
where no listing of sources was available, this source category was both inven-
toried and modeled  as area sources.  A third exception is  municipal  waste
landfills, which were simulated as area sources using  CDM-2 using  landfill-
specific dimensions.

An unavoidable element of uncertainty is introduced in estimating  atmospheric
dispersion.  In general, the data and equations used to estimate atmospheric
dispersion are an approximation of real  atmospheric phenomena.  Specifically,
in Southeast Chicago, the proximity of Lake Michigan may cause alterations in
the frequency of some wind directions and wind  speeds  and  may also affect the
extent of dispersion in this area as compared to the meteorology at Midway
Airport.  Generally, atmospheric dispersion models are considered  accurate

-------
                                       16

within a factor of two.  Although actual  uncertainties for annual  average
concentration estimates are difficult to  quantify, this generalization does
give a sense of the uncertainties in the  modeling element of this  study.

A sample of the concentrations estimated  in this study is shown  in  Figure 2.
This figure shows a map of coke oven pollutant concentrations.   This  map
highlights the grid system used in estimating  concentrations.   The  area was
divided into 1 kilometer squares, and concentrations were estimated at the
center of each square.  The geographic coordinate system used  in this study was
the Universal  Transverse Mercator (UTM) system.  In UTM coordinates,  the  square
with the highest coke oven pollutant concentrations extends from 4614.5 kilo-
meters to 4615.5 kilometers north and from 452.5 kilometers to  453.5  kilometers
east in zone 16.  In Chicago streets, this square extends roughly  from 117th
Street to 112th Street and from almost a  kilometer west of Torrence Avenue to  a
little east of Torrence Avenue.  The concentration estimate used for  this grid
square was estimated at 4615 kilometers north/453 kilometers east,  which  is
near 114th Street and Torrence Avenue.

Although the receptor resolution (i.e., the estimation of concentrations  at
1 kilometer intervals) is adequate for the purposes of this screening study, it
must be understood that a finer receptor  resolution (i.e., estimation of  concen-
trations at more closely spaced intervals) would be expected to  yield a higher
peak concentration.  This is because estimation of concentrations  at  more
locations can be expected to identify some locations with somewhat  higher
concentrations.  That is, the actual peak concentration for coke oven pollutants
is probably somewhat higher than the 6.1  ug/m^ shown in figure  2.   However, the
design and scope of this study was not to obtain a precise peak  concentration
estimate but rather to address area-wide  impacts from multiple  pollutants and
multiple sources.

The estimate of area-wide exposure to specific pollutants would  also  be more
precise if a finer receptor resolution were used.  However, concentrations
generally do not change dramatically more than a few kilometers  from  a given
source, so the use of a finer receptor network would not be expected  to alter
the area-wide exposure estimates significantly.

Figure 3 shows a map of concentrations of polycyclic organic matter.   (This map
and Figure 2 were both produced by PIPQUIC.)   This figure shows  concentrations
generally increasing toward the center of Chicago, reflecting  the  increase in
population density and, therefore, density of  sources of polycyclic organic
matter (particularly mobile sources and homes  being heated) as  one  approaches
the center of Chicago.

Similar concentration estimates were made for  the other pollutants  in this study,
However, the most meaningful  way of addressing multiple pollutants  is to  use
the common denominator of risk.  This discussion will  be included  later in this
report.

Comparison of Modeling and Monitoring Concentration Estimates

This study uses monitoring data in two ways.   The first use is  to  compare with
dispersion model  estimates, to provide an indication of the reliability of the
model  estimates.  The second use, applicable  to formaldehyde and carbon tetra-
chloride, is for quantifying concentrations of "background pollutants" which are

-------
                                                                         17
T3
 >
                                                                                      -C
                                                                                      4-J
                                                                                        CO
                                                                                                                           >>
                                                                                                                           O)
                                                                                                                          _Q

                                                                                                                          OO
                                                                                                                                              V
 I/)
 c
 O
 C
 
O

 O)
-ii
 O
-o
 O)
   OO
 i
•r-    CD
+->    3
 1/5
LU    C
 O

 Q.
S_
3
o>
                       en
                                                                                    °°r_
                                           un        ^t        n        
-------
                                                             18
                              X
                              13
-M
  r^
  co
S-
o>

-------
                                       19

not the direct result of current emissions.  Various monitoring programs  have
been conducted in the Southeast Chicago area to measure concentrations of
pollutants in this study.  Table 2 summarizes the studies from which data were
available.  This table shows the organization conducting the  monitoring,  the
location(s) of the monitoring site(s), the monitoring method, the  sampling
period, the number of samples, the sampling duration (frequency and  averaging
time) and the pollutants monitored.

Table 3 presents a comparison of modeled versus monitored concentration estimates
for the organic substances for which monitoring data are available.   For  each
comparison, the monitoring data represent the average over the full  time  period
for which reliable data are available.  The modeling data in  effect  are 5 year
averages (since the underlying meteorological  data are 5 year averages and  the
underlying emissions data are intended to be similarly long-term averages).
The modeling results are also specifically interpolated to the location of  the
monitor from the concentrations estimated at the nearest modeling  grid points.
Although in a few cases such interpolated results may differ  significantly  from
the results that would be obtained by direct modeling for concentrations  at the
monitor location, particularly near major sources where spatial  gradients may
be high, in most cases these differences should be small.

The best comparison on Table 3 is for benzene.  For this pollutant,  the monitored
values are within a factor of two to three higher than the modeled concentrations.
Given the relative sparsity of monitoring data (in no study were more than  about
30 days sampled), the uncertainty of the monitoring methods at concentrations
close to the detection limit (generally not more than around  three times  the
detection limit), and the uncertainties in the emissions inventory and modeling
analysis, these results should be considered quite comparable.  Note that although
the modeled estimates could be adjusted to include the benzene component  of
coke oven emissions, this would only be a few percent increase.  Less encouraging
are the comparisons for toluene and xylene, where monitored values are between
one and two orders of magnitude greater than modeled estimates.  The same may
be true of chloroform, whereas the comparison for perchloroethylene  and
trichloroethylene appear to be as close as the comparison for benzene. However,
the concentrations that the Illinois Institute of Technology  (IIT) and the
Hazardous Waste Research Information Center (HWRIC) identify  for perchloroethylene,
trichl oroethyl ene, chloroform, and carbon tetrachl oride are below  the monitoring
detection limits that Radian identifies for these compounds,  so these comparisons
may not be reliable.

It has been noted previously that a substantial  portion of formaldehyde and
carbon tetrachloride concentrations may be attributed to origins other than
current emissions.  In this study, the emissions estimation/dispersion modeling
approach is considered the best means of addressing the impacts of current
emissions.  For formaldehyde and carbon tetrachloride, monitoring  data provide
the best indication of the sum of direct impacts from current emissions plus
indirect impacts from other causes.  Thus, in this study, background concen-
trations for these two pollutants were evaluated by determining a  total con-
centration from available monitoring data and then subtracting the concentration
attributable to current emissions.  These background concentrations  were  assumed
to be uniform throughout the Southeast Chicago area.  Total concentrations  of
these two pollutants at each of the receptor locations were then derived  by

-------An error occurred while trying to OCR this image.

-------
21
•o
cu
'aj
•a
o
*gf.
"O CU
TO !-
c as o
o •*-*
4-> C
TO ^
C
O
c
o 1 cu
3 5
C/l O
TO * S
-° S£l
' ?'-o
cu i— cu
Of O
4-> 4-> - — - 4~>
•r- C OO 'I-
C TO E C
o o -^ o
s •-- ai s:
X 3
3 h- C
s_ o
Ol -l~ f i"O
> C C OJ
1 TO 0 "-
-a en •!- ai
— O TO O
01 s_ * 2:
-Q S_ 4J 1 —
S 4— Ol i — i "O
o cu
<4- l/l CZ t_
O Ol O O
•MO +->
C TO -i-
.2 -i n §
W1 -M 
• 1 I 1 • 1 • 1 O O
^_ o o • •
•»! * *
* * * * -k
>* 00 ^J~ ro ioco^i-oo
ro • • • CT> i vo • «C>
• V V V CM — 1 V V OO •
00 r- ( =d" CM
CU
^2
1 —
C
LO CTt p<~
, — i — CM 1 1 LO
r~- 0*1 i i i i • i i CM c
• • 1 1 1 1 O , — CU
CD
0)
TO
, — oo i i i i a~i cri to
i i i i • i i • £
«3" CM CM 1 I CTi TO
Ol
o
a.
c
'-O  oo '^i i o «OICM o
• • CM 1 ' r— O ^- 4J
,— O O CD • -r-
LOOO^HCO i — COO <£> 4->
r^- CM CTt i^^ i "-.o C71 r — i co c
. « • . 1 ... i . '•*- O
^± CM O CM "^f CTl CM r--~ O 'r—
1/1 o
•r- 4.)
+-> -a
cu
•0 T3
oo *3- r-~ cu
CM I 1 i — 1 1 LO ^- 1 , — •D-'-
• 1 ICDl 1 "i — ICM C<4-
OO • O CD CM TO •!-
"O d
c "O
i — i i CTi i i oo ^ ^n cu o>
• t 1 • 1 1 • • 1 OO T3 r-
U"> , 	 <*O O 1 « 4.)
r— ^- S
TO O
U Ol Ol i/l Ol
C C C E J3
CU CU O) >•>
'>,'>, r- >, D CU
£ .C O) JZ 4-> i- •!-
4->4-> "O 'C_5OO OM—
CU CU g >» -t-J TO -i—
ooCjc oioi^a c
S-t-OCUOl 1 — C O>O)
cu o o "4- ~o c o. oi  i/i T-
c i — i — o , — oi^c — 01 oi'-o
OJ-C-C'-TO. — O. O >) C -C
NOUOE^^-Q-^OI h- V
C S_ -1- r— S_ JZ — !_+->^-
OJCU'--COJ-)3'OOI>> •)«•)«
±aa_i — ou-uji — osx *

-------
                                       22

adding the uniform concentration representing background impacts  plus  the vari-
able concentrations representing direct emissions impacts.

As seen in Tables 2 and 3, formaldehyde was monitored at one location  in  the
area.  Data are available for September 1987 to March 1988.   While these  are
the best data available, it must be noted that the absence  of data from the
summer, when photochemical  formation of formaldehyde is greatest, indicates that
available data probably understate the annual average formaldehyde concentration.
In any case, the average of available data is a concentration of  2.93  ug/m^.
At the monitor location, the impact of direct emissions is  estimated to be 0.27
ug/m^.  Therefore, the formaldehyde concentration attributed to  photochemical
formation is the difference of 2.66 ug/m^.

Tables 2 and 3 show that carbon tetrachloride was monitored  at three locations
in the Southeast Chicago area.  However, Table 2 also shows  that  two of the
three studies (by I IT and HWRIC) include only a small  number of  samples,  and
Table 3 shows that the third study (by Illinois EPA/Radian)  did  not report any
detectable concentrations.  Atmospheric accumulation of carbon tetrachloride
over prior decades may be presumed fairly uniformly distributed  in the global
atmosphere, and so a more reliable indicator of the atmospheric  accumulation  of
carbon tetrachloride is from more thorough studies elsewhere.  In areas of the
United States that may be presumed not to have significant  sources of  carbon
tetrachloride , available monitoring data suggest concentrations generally
between 0.6 and 0.8 ug/m^.  An average value of 0.76 ug/m^  is used as  the average
value in the Southeast Chicago area.  Most of this concentration  is assumed
uniform throughout the area;  only the minor portion of the  concentration
attributable to current emissions is treated as varying from location  to  location

Table 4 compares PCS concentrations monitored by IEPA with modeled concen-
trations.  Possible explanations for this relatively poor comparison include
missing emission sources, uncertainties in the monitoring method, a short and
therefore possibly unrepresentative monitoring period, and  the long atmospheric
residence of PCBs.

Table 5 compares particulate matter monitoring data with modeled  concentrations.
For arsenic, cadmium, and chromium, the two sets of concentrations are quite
similar, indeed well  within the uncertainty ranges for the monitoring  and
modeling data. (Note that for chromium, both the monitoring  and modeling  data
show total  chromium concentrations.)  The other pollutant shown in Table 5,
benzo( a) pyrene, again seems to show a close comparison between monitored  and
modeled concentrations.  This comparison is complicated by the fact that  the
monitoring method measures specifically benzo(a)pyrene, a compound which  in
the inventory is included in the class of compounds labeled  polycyclic organic
matter (POM) as well  as in coke oven emissions.  The designated modeled value
in Table 5 was estimated as a somewhat arbitrary 1% of the combined mass  of POM
plus coke oven pollutants.  Given the uncertainty in this comparison,  no  firm
conclusions can be drawn from the similarity of these monitored  and modeled
data .

These comparisons of modeled versus monitored concentration  estimates  appear  to
support two generalizations: (1) for most pollutants, the modeled and  monitored
concentration estimates agree reasonably well, and (2) the differences between

-------
                                       23
          Table 4. Comparison of Modeled- Versus Monitored-based



                        Concentration Estimates for PC6s



                         (all concentrations in ug/m^)
Bright School





Washington School





Grissom School
.OU19






.0003






.0005
Modeled






.000004






.000001






.000004

-------An error occurred while trying to OCR this image.

-------
                                       25

modeled and monitored concentrations, whether these differences are large or
small , in essentially all cases show higher monitored concentrations than
modeled concentrations.  The first generalization suggests that for most
pollutants, this study provides a reasonable assessment of the concentrations
of these pollutants.  The second generalization suggests that even these
reasonably assessed pollutant concentrations are slightly underestimated, and
concentrations for a few other pollutants may be substantially underestimating
actual ambient concentrations.  This in turn suggests the possibility that the
emissions inventory of this study underestimates emissions, perhaps by under-
estimating emissions at identified facilities and perhaps by failing to identify
some  sources of emissions.  Supporting this hypothesis, some of the pollutants
which appear most underestimated by modeling (particularly PCBs and chloroform)
are also among the more difficult pollutants for which to estimate emissions.

Evaluation of Cancer Risk Factors

Once concentration estimates have been made for the identified pollutants, it
is then necessary to estimate the relationship between concentration and the
increased probability or risk of contracting cancer that exposure to each pollu-
tant may cause.  This relationship is commonly expressed in terms of a unit
factor, representing the risk estimated to result from exposure to a unit con-
centration of a pollutant.  For example, if a pollutant has a unit risk factor
of lxlO'4 per ug/m^, then lifetime exposure to 1 ug/rn^ (1 millionth of a gram
of the pollutant per cubic meter of air) would be estimated to increase the
probability of contracting cancer by 1x10"^ or 1 chance in 10,000.  The pro-
bability or risk of contracting cancer is generally treated as linear within
the range of actual  exposure conditions, so that in the example above, exposure
to a concentration of 3 ug/m^ would be estimated to increase cancer risks 1:0
3xlO-4 or 3 chances in 10,000.

There is a lack of data where large numbers of people are exposed to typical
environmental  concentrations, where the concentrations and the resulting number
of cancer cases are well  defined for several  subpopulations, and where confounding
influences from other causes of cancer can be clearly factored out.  Therefore,
a variety of methods, scientific judgements and assumptions are used to assess
the relationship between exposure to a pollutant and the resulting risk of con-
tracting cancer.

For some pollutants, sufficient data do exist for specifiable human exposure
circumstances to estimate the exposure levels and to evaluate the cancer risks
that apparently result.  The interpretation of these statistical  data is gene-
rally designed to derive a maximum likelihood estimate of the unit risk factor
(i.e., deriving a unit risk factor which the data suggest will  have the greatest
likelihood of accurately representing the ratio between exposure and cancer
risk for the conditions of the study).  In general , the exposures that can be
studied are higher than typical  ambient concentrations, and so extrapolation of
the exposure-cancer risk relationship must be performed.  This extrapolation of
the dose-response relationship down to lower exposure levels uses conservative
methods, so as to decrease the likelihood of underestimating risks.

For a majority of pollutants, however, no human exposure situation can be
sufficiently characterized to support the derivation of a unit risk factor.
The only data for deriving unit risk factors for these pollutants, then, will

-------
                                       26

generally be from studies involving animals.  These studies provide statistical
data which by various interpretations can yield alternative unit risk factor
estimates.  The usual interpretation method is to select a 95% upper confidence
level value.  This signifies that the selected unit risk factor is the value
which has a 95% likelihood of not understating the true risk factor indicated
by the data.  It should be noted that this discussion refers only to the con-
servatism inherent in the statistical interpretation of cancer data, which  is
not the only element of conservatism in  the unit risk factor.  As with the
maximum likelihood estimate, a downward  extrapolation from studied exposures  to
ambient exposures is necessary, and this extrapolation is done in a way that
adds conservatism.  (For animal studies, practical  considerations generally
require studied exposures to be higher than ambient exposures.  For example,  a
study involving 100 animals cannot provide a meaningful  result if the risk  is
1 in 1,000,000.)  The extrapolation of the unit risk factor applicable to
typical  ambient concentrations involves  best scientific judgement of a plausible
yet conservative extrapolation.  With animal  studies, an additional  adjustment
is made from animal  carcinogenicity to human carcinogenicity based on differences
in body weight and breathing rate, again involving best scientific judgement  of
a plausible yet conservative extrapolation.  Thus, the methods of extrapolating
unit risk factors add some conservatism  to the conservatism inherent in the use
of a 95% upper confidence limit.

The relationship between pollutant concentration and cancer risk is a function
of both the quantity of pollutant inhaled and the body's reaction to the inhaled
quantity.  Unit risk factors are designed to estimate the cancer risk resulting
from inhaling a unit concentration for 24 hours a day for a 70 year lifetime.
Similarly, cancer risks in this study are estimated by assuming that Southeast
Chicago area residents are exposed to the estimated concentrations for 24
hours per day for a 70 year lifetime.  Clearly, these residents spend some  time
outside the study area and spend some time indoors, but the absence of knowledge
of pollutant concentrations in these other environments makes it impossible to
make upward or downward adjustments according to these other exposures.

In addition to variability in carcinogenic strength, there is also variability
in how much evidence exists to indicate  more fundamentally whether individual
pollutants are in fact carcinogenic.  Therefore, USEPA has established a
classification system describing the weight of experimental  evidence that a.
pollutant is carcinogenic.  The classifications used by the U.S. EPA are:
A - human carcinogen; B - probable human carcinogen; C - possible human car-
cinogen; D - not classifiable as to human carcinogenicity; and E - evidence of
noncarcinogenicity in humans.  These ratings reflect the following types of
evidence:  A - "sufficient" human data show carcinogenicity; B - is subdivided
into Bl and B2, in which either "limited" human data or "sufficient" animal
data show carcinogenicity; C - human data are inadequate or nonexistent but
limited animal  data show carcinogenicity; D - data to assess carcinogenicity
are inadequate or nonexistent; and E - well  designed studies suggest that the
pollutant is noncarcinogenic.  More detailed definitions of these classifications
can be found in USEPA's Risk Assessment  Guidelines of 1986.  For clarity,
references to group A pollutants in this report will  use the term "known human
carcinogen."

The classifications in the weight of evidence approach are intended  to indicate
the strength of the evidence of carcinogenicity independently of any evaluation
of carcinogenic strength.  As yet, no equivalent system has been developed  to

-------
                                       27

address the accuracy of the unit risk factors.  For some pollutants,  a greater
weight of evidence of carcinogenicity also signifies a better data base from
which to estimate unit risk factors, but this is not the case for  all  pollu-
tants.

This study found and quantified emissions for 30 presumed carcinogens.  USEPA's
evaluation of the weight of evidence is that these 30 pollutants include 6 known
human carcinogens, 22 probable human carcinogens, and 2 possible human carcino-
gens.  Table 6 provides the names of these pollutants, the weight  of  evidence
classification, the unit risk factor used in this study, and whether  this risk
factor is calculated as a 95% upper confidence level (UCL), a maximum  likeli-
hood estimate value (MLE), or a best estimate (BE).  This table also  shows
which USEPA office developed the unit risk factor.  In this table, IRIS (Inte-
grated Risk Information System) signifies risk factors that have received
agency-wide review.  Other values have not received agency-wide review but have
been developed by the Office of Health and Environmental  Assessment in the
Office of Research and Development (designated OHEA), by the Office of Air
Quality Planning and Standards (designated OAQPS), or by the Office of Toxic
Substances (designated OTS).

Several  of the pollutants in Table 6 represent mixtures of compounds.   One
such mixture is designated in Table 6 as "Benzo(a)pyrene (POM)."  Benzo(a)
pyrene is the most studied member of the class of compounds known  as  polycyclic
organic matter (POM).  This study inventoried emissions and estimated  concen-
trations of the full  class of POM compounds, and then estimated risk by multi-
plying the POM concentrations times the benzo(a)pyrene unit risk factor.  While
some POM compounds are probably more carcinogenic and other POM compounds are
less carcinogenic, this approach in effect assumes that the average cancer
potency of the full  range of POM compounds equals the cancer potency of
benzo(a)pyrene.

Another mixture shown in Table 6 is coke oven emissions.  For this mixture, a
unit risk factor for the full mixture has been developed (based on epidemio-
logical  analysis of occupational  exposure data).  This mixture includes sub-
stantial quantities of other pollutants in this study, including polycyclic
organic matter and benzene.  However, no effort was made to assess emissions  or
risk from these coke oven gas constituents individually.  Instead, the emissions
estimates, the unit risk factor, and the risk estimates for coke oven  emissions
are designed to address the emissions, toxicity, and risk of the full  mixture
emitted from coke batteries.

A third mixture shown in Table 6 is dioxin.  In this study "dioxin" represents
a class of 75 chlorinated dibenzo-dioxins and 135 chlorinated dibenzo-furans.
The unit risk factor shown in Table 6 is for 2,3,7,8 - tetrachloro-dibenzo-
dioxin (2,3,7,8 - TCDD), the best studied dioxin.  Other dioxins were  inventoried
on the basis of toxic equivalents, i.e., what mass of 2,3,7,8 - TCDD would have
equivalent toxicity to the given mass of identified dioxin.  For example, 10
grams of 2,3,7,8 - tetrachl oro-dibenzo-furan, having an estimated  toxicity
equivalence factor of 0.1, would be inventoried as if it were 1 gram of
2,3,7,8 - TCDD.  Quantitative details are given in the inventoried documentation.

Two other mixtures shown in Table 6 are gasoline vapors and polychl orinated
biphenyls (PCBs).  The unit risk factor for gasoline vapors was derived from  a

-------
                      28
Table 6.  Carcinogenicity of Inventoried Pollutants
Pol lutant
Acryl amide
Acrylonitrile
Arsenic
Asbestos
Benzene
Benzo(a)pyrene (POM)
Beryl 1 i urn
Butadiene
Cadmi urn
Carbon Tetrachl oride
Chi oroform
Chromium
Coke Oven Emissions
Di oxin
Epichl orohydrin
Ethylene Dibromide
Ethylene Dichl oride
Ethylene Oxide
Formaldehyde
Gasol ine Vapors
Hexachl orobenzene
Methyl Chloride
Weight of
Evidence
Rating*
B2
Bl
A
A
A
B2
B2
Bl
Bl
B2
B2
A
A
B2
B2
B2
B2
B1-B2
Bl
B2
B2
C
Unit Risk
Factor
(in (ug/m3)-!)
1.2 x 10-3
6.8 x 10-5
4.3 x lO-3
8.1 x ID'3
8.3 x ID'6
1.7 x ID'3
2.4 x ID'3
1.1 x ID'4
1.8 x ID'3
1.5 x ID'5
2.3 x ID'5
1.2 x ID'2
6.2 x ID'4
3.3 x 10+1
1.2 x ID'6
2.2 x lO'4
2.6 x ICr5
1.0 x ID'4
1.3 x 1CT5
6.6 x ID'7
4.9 x ID'4
3.6 x 10-6
Type of
Risk
Factor**
UCL
UCL
MLE
BE
MLE
UCL
UCL
UCL
MLE
UCL
UCL
MLE
UCL
UCL
UCL
UCL
UCL
UCL
UCL
UCL
UCL
UCL
Source
of
Data***
IRIS
IRIS
IRIS
IRIS
IRIS
OAQPS
IR]S
IRIS
IRIS
IRIS
IRIS
IRIS
OHEA
OHEA
IRIS
IRIS
IRIS
OHEA
OTS
OAQPS
OHEA
OHEA

-------
                                           29
                                  Table 6. (Continued)
 Pollutant

 Methylene Chloride

 Perchlorethylene

 PCB's

 Propylene Oxide

 Styrene

 Trichloroethylene

 Vinyl Chloride

 Vinylidene Chloride
Weight of
Evidence
Rating*
B2
B2
B2
B2
B2
B2
A
C
Unit Risk
Factor
(in (ug/m3)~l)
4.7 x 10'7
5.8 x 10~7
2.2 x 10-3
3.8 x ID'6
5.7 x ID'7
1.7 x lO"6
4.1 x 10~6
5.0 x lO'5
Type of
Risk
Factor**
UCL
UCL
UCL
UCL
UCL
UCL
UCL
UCL
Source
of
Data***
OHEA
OHEA
OHEA
OHEA
OHEA
OHEA
OAQPS
IRIS
   - As discussed in text, these ratings signify:
     A - Known human carcinogen
     B - Probable human carcinogen
     Bl - Based on "limited"  human data
     B2 - Based on "sufficient" animal  studies
     C - Possible human carcinogen
 ** _
The three types of risk factors used in this study are;
UCL - 95% upper confidence limit
MLE - maximum likelihood estimate
 BE - best estimate
*** IRIS - Integrated Risk Information System
    OAQPS - Office of Air Quality Planning and Standards
    OHEA - Office of Health and Environmental  Assessment
    OTS - Office of Toxic Substances
    Note: As described in the text, each unit risk factor estimates risk from
          lifetime exposure to a unit pollutant concentration.

-------
                                       30

study of the full mixture, though it does not include the impact of gasoline's
benzene component.  The unit risk factor for PCBs  was derived  for a representative
compound of this set of compounds.

Chromium and ethylene oxide also warrant special  comment.  For chromium,  both
the emissions estimates and the unit risk factor  are only for  the hexavalent
(+6) form of chromium.  For ethyl ene oxide, the classification B1-B2 refers to
the fact there there is both "limited" human evidence and "sufficient"  animal
evidence of the carcinogenicity of this compound.

The above discussion addresses the calculation of  risks  from individual  pollu-
tants.  This study also seeks to estimate the combined impact  of all  the
pollutants included in this study.  The methodology recommended in  the  "Chemical
Mixtures Risk Assessment Guidelines" (part of USEPA's Risk Assessment Guidelines
of 1986) is to estimate total risks as a linear sum of the individual pollutant
risks, in the absence of information suggesting otherwise.  It is possible that
exposure to some combinations of pollutants may cause a  greater risk (synergism)
or a lesser risk (antagonism) than the sum of the  risks  resulting from  exposure
to the substances individually.  However, there are no clear means  of quantify-
ing any synergistic or antagonistic effects from  exposure to the complex  and
variable mixtures in the Southeast Chicago area atmosphere, if in fact  such
effects are occurring.  Therefore, the method for  combining risks used  in this
study was to sum the risks estimated for individual  pollutants.

The unit risk factors used in this study reflect  the best judgements of USEPA
scientists in evaluating available evidence both  as to the interpretation of
specific studies and as to the procedures that most reliably extrapolate  unit
risk factors from these studies.  Nevertheless, the uncertainties in the  unit
risk factors are probably the greatest uncertainties in  this study.  These
uncertainties arise from the significant extrapolations  such as from high
concentrations to lower concentrations and from rats or  mice to humans  that
are necessary to estimate the risk factors.

The Risk Assessment Guidelines of 1986 discuss the significant assumptions and
therefore the significant uncertainties that are  necessary in  developing  unit
risk factors.  In summary, these assumptions and  uncertainties are  as follows:
(1) Exposure to any amount of the substance, no matter how small , is assumed to
represent an increased probability of cancer.  There is  uncertainty that  cancer
impacts may occur only above some pollutant-specific threshold concentration;
(2) For risk factors based on animal studies, the  development  of cancer in
humans is analogous to the development of cancer  in the  animals.  There is
uncertainty that the biological  process of cancer  formation is the  same process
in humans as in animals.  For this and other reasons, there is also uncertainty
in the quantitative extrapolation of the relationship between  cancer risks and
exposure for humans from the relationship for animals;  (3) Information  on the
carcinogenicity of substances at "high" concentrations can be  used  to predict
the effects at "low" concentrations; and (4) The  increased probability  of cancer
incidence is proportional  to the concentration of  the substance at  low
concentrations.

Incidence and Risk Estimates

As indicated previously, risks at a given location were  estimated by multi-
plying, for each pollutant, the modeled concentration estimate (in  ug/rn-^) times
the risk per ug/m^ of that pollutant, and then summing for all  pollutants.
These risks are commonly expressed in exponential  form,  where, for  example,

-------
                                       31

2xlO~3 equals 2 chances in 1,000.  Thus, a person residing for a lifetime at
such a location will  have 2 chances in 1000 of contracting cancer from this
exposure.

Incidence is a more population-oriented measure of pollutant impacts.  By multi-
plying the risk in a given grid times the number of people in that grid, one
can estimate a probable number of cancer cases contracted  as a result of the
exposure.  For example, if a grid square with an estimated lifetime risk of
2xlO~3 has a population of 2000, one would estimate that a lifetime of exposure
would lead to 4 cancer cases.  This figure is sometimes translated to an annual
probability: a probability estimate of 4 cases divided by a 70 year lifetime
suggests a probability estimate of 4/70 or 0.057 cases per year or one case per
17.5 years.  This calculation is done for each grid square; the total  across
all grids is then the estimated number of cancer cases in  the entire study area
attributable to air pollution.

It should be noted that the risk estimates presented in this report should be
regarded as only rough approximations of total  cancer cases and individual
lifetime risks, and are best used in a relative sense.  Estimates for indivi-
dual pollutants are highly uncertain and should be used with particular caution.

The total  cancer incidence estimated in this study is approximately 85 cases
over 70 years, or about 1 case per year.  Figure 4 is a pie chart illustrat-
ing the contributions of the various source types to this  estimated incidence.
(This figure is titled as contributions to annual  cancer cases, but percentage
contributions to the lifetime (70 year) number of cases are the same.)  Table 7
presents a more detailed table breaking down the contributions by pollutant
and by source type.  Figure 5 shows the spatial  distribution of cancer incidence
estimates.  (Areas with zero incidence estimates, such as  in the Lake Calumet
area, represent areas with no residents.)

Figure 4 and Table 7 show that the greatest contribution to cancer incidence  in
the Southeast Chicago area appears to result from emissions at steelmaking
facilities.  In total, the various integrated steel mills  in the southern Lake
Michigan area were estimated to cause about 29 cancer cases over 70 years,
representing almost 34% of the total.  Coke ovens in particular appear to
contribute more than any other individual operation to air pollution-related
cancer incidence.  Specifically, the emissions from coke ovens, including the
emissions from charging coal  into the ovens and from leaks around the oven
doors, lids, and offtakes were estimated to contribute 24  cases over 70 years,
or about 85% of the steel  mill  contribution to area incidence.  Coke oven
by-product recovery plants, which refine the gases baked out of coal  by the
coke ovens, are estimated to contribute an additional  2 cases over 70 years.
Since this operation may be considered an adjunct to coke  production, the
total  risk estimated for coke production is 26 cases over  70 years, or about
92% of the steel mill  contribution.  The remaining 8% of the steel  mill  con-
tribution to incidence arises principally from arsenic, cadmium, and chromium
emissions from basic oxygen furnaces, electric arc furnaces, blast furnaces,
and sintering operations.

Figure 4 and Table 7 indicate other significant contributors to estimated
air pollution related cancer cases in the Southeast Chicago area.  Also use-
ful here is Figure 6, showing a more detailed breakdown than Figure 4 of
specific source types to area cancer case estimates.  After steel  production,

-------
                                          32
      0

      Q.
 O)

 S-
<
LU
cc
o
      0
      O


      o
     CO
 CO
 0
 CO
 CO
o

 0
 o
 c
 CO
o

"5

 c
 c
 0
+*
 CO
3   2
2   (0
z   c
o   o

I-   '§
CO   3

CO

-------
                                           33
                       Table 7.  Contributions to Estimated Area Cancer Cases by
                                 Source Type and Pollutant Across the Study Area
                                       (in cases per 70 years)
  Compound*

Acryl amide
Acrylonitrile
Arsenic
Asbestos
Benzene

Beryl 1ium
Butadiene
Cadmium
Carbon Tet.
Chioroform

Chromi urn
Coke Oven Em.
Dioxin
Epichlorohydrin
Eth. Di bromide

Eth. Dichloride
Eth. Oxide
Formaldehyde
Gas. Vapors
Hex-chl-benz.

Methyl  Chi.
Methylene Chi.
Perchloroeth.
PCB's
POM

Prop.  Oxide
Styrene
Trichloroeth.
Vinyl  Chi.
Vinylidene Chi.

Steel
Mills


1.4

2.2

m
.8
'

.06
24.2





m












Other
Industrial
Sources
m**
m
.1

.06
m
m
.01
m
m
13.0

.2
m

.05
.2
m
m
m
m
.04
.04
m
m
m
m
.2
m
m

Consumer Mobile
Sources Sources



.04 .1
.1 1.9

2.2
.01

.2
1.7



.05

.3
.4 1.3
.9 2.6

.01
.1
.1

8 3.7





Sewage
Waste Treatment Background
Handling Plants Pollutants

m


m m

m

.01 4.5
m m


.01
m

m m

m 13.6

.06 .1
m
m m
m m
m


m m
.01 m
m
m m


Total
m
m
1.5
.1
4
m
2
.8
5
.2
14
24
.2
m
.05
.05
.5
15
3
.2
.01
.2
.2
m
12
m
m
.2
m
m
   TOTAL***           29

   *Abbreviations:
          Carbon Tet.
          Eth.
          Gas.
          Hex-chl-benz.
        14
12
12
.1
- Carbon tetrachloride
- Ethylene
- Gasoline
- Hexachlorobenzene
   **,
     'm - minor' (<.005 cases per 70 years)
               Chi.
               PCB's
               POM
               Prop.
.1
18
85
            - Chloride
            - Polychlorinated biphenyls
            - Polycyclic organic matter
            - Propylene
   *** Most figures have been rounded to nearest whole number.
      BECAUSE OF UNCERTAINTY IN PROCEDURES, METHODS, ASSUMPTIONS AND DATA, THESE RISK  NUMBERS
      SHOULD BE REGARDED AS ONLY ROUGH APPROXIMATIONS AND ARE BEST USED  IN A  RELATIVE  SENSE.
      ESTIMATES FOR INDIVIDUAL POLLUTANTS ARE HIGHLY UNCERTAIN AND SHOULD BE  USED WITH
      PARTICULAR CAUTION.

-------
                             34
 en
   0)
   CO
   CO
   O

   <5
   o
   c
   /? 3)
   VJ DC
   0) 2

   .1 *
   tj> g

      o
CD 2

!!
"£ X

o  ™ S
S  *
5  5
S  ®
O  >
I-  '+3
CO  (0
<  "5
UJ
o
CO

-------
                                              35
 o
 U)
 a)
"o
u

 0)
 u

 D
 O
 o
TJ
^
 O
 0)
 i_
CD


ID

 0)
 u
 D
 D)
•—
Lu
 V)
 0)
 W
 O
CJ

-o
 0)
•*—
 D

 E
      C
      O
_Q
 C
 O
u

-------
                                       36

the most significant contribution to risk is labeled "background  pollutants,"
which contribute 18 cases over 70 years or about 21% of the total .   This
category includes air pollution which is not caused by current emissions,
but rather represent "background concentrations" from other causes.   Specifi-
cally, this category includes formaldehyde, which results from atmospheric
photochemical  reactions of other currently emitted organics, and  carbon
tetrachloride, which results largely from atmospheric accumlation of previous
carbon tetrachloride emissions.  As seen in Figure 6, photochemically formed
formaldehyde appears to be one of the most significant air pollution-related
causes of area cancer cases.  Atmospheric accumulation of carbon  tetrachloride
also appears significant, although less significant than other contributions
in Figure 6.

A third significant contribution to risk as shown in Figure 4 is  industrial
operations other than steelmaking, which cause about 14 cases over  70 years
or about 16% of the total .  As seen in Figure 6, this risk is predominantly
due to chrome plating operations.  Degreasing and miscellaneous other manu-
facturing operations add a fairly modest contribution to estimated  cancer  cases
relative to other causes of air pollution-related risks.  A fourth  significant
contributor to risk is identified on Figure 4 as consumer sources,  which cause
about 12 cases over 70 years, or about 14% of the total.  This category  includes
several  activities engaged in by the general  public which result  in  emissions
of presumed carcinogens.  A fifth significant contributor to risk is from  road-
way vehicles such as cars and trucks traveling on streets and highways,  also
causing about  12 cases over 70 years, or about 14% of the total.

Figure 7 provides more detailed information on the contribution of  consumer-
oriented sources of air emissions to the estimated number of area cancer cases
across the study area.  Home heating appears to be the most significant  such
source type.  Although some of this estimated risk is from combustion by-products
(polycyclic organic matter and formaldehyde)  from gas furnaces, the  bulk of
this contribution to risk is from wood combustion in fireplaces and  wood stoves.
Although the actual  quantity of wood burned in the study's source area is
small, a significant fraction of this wood is transformed during  combustion
into polycyclic organic matter.  This pollutant is sufficiently toxic, so  that
wood combustion emerges as a relatively significant source category.

Figure 7 also  shows contributions from other types of consumer-oriented  source
categories.  The activities aggregated as "miscellaneous activities" each  are
estimated to contribute less than 0.3 cases over 70 years (about  4  cases in
1,000 years),  and include, in order of estimated significance: sterilizing
operations at  hospitals using ethylene oxide, chloroform from chlorinated
drinking water, formaldehyde in miscellaneous consumer products,  dry cleaning,
methylene chloride in aerosol  cans and paint stripping, various minor constituents
of house paint, and asbestos from demolition and renovation of asbestos-containing
structures.  The total  contribution of these "miscellaneous activities"  is esti-
mated to be about 1 case over 70 years.

Collectively,  the impacts of steel  mills, other industrial  operations, back-
ground pollutants, consumer sources, and roadway vehicles contribute all but
0.2% of the total  estimated number of cancer  cases attributable to  air pollution.
Using more narrowly drawn source categories,  the collective impact  of steel mills,

-------
                                                   37
                                                                                   *
                                                                                   o
                                                                                   o
                                                                                   X
Q)
S

c c
0 0
00
.
. £
I"- 0
k.
03 **-
l_
AM
3 w
ZL. 0)
.P W

U. D
O


x™ *
X
IT)
DO
x^
n
D
0
ID
[^
^
"»
u
M
2
                     O)
                     c
                     o
                    2

                     to
                     0
                    o
O)
                                             o
                                             o
                                            o

-------
                                       38

chrome plating, background formaldehyde, background carbon tetrachl oride, home
heating, and roadway vehicles contribute almost 95% of the total  estimated air
pollution related cancer cases. (Other industrial  sources and other consumer
sources contribute the remainder of the 99.8% referenced above ("all  but 0.2%").)

Figures 4 and 6 and Table 7 also suggest what source categories make relatively
minor contributions to estimated risks.  In particular, both wastewater treat-
ment plants and waste handling facilities (including both hazardous waste and
municipal solid waste) each contribute only about  0.1% of the total  air
pollution-related number of cancer cases in the area.

Figures 4 and 6 and Table 7 suggest a variety of additional  statements on the
contribution of various air pollution source types to cancer incidence in the
Southeast Chicago area.  One proposition supported by these  data  is that
while steelmakiny i.s the most significant contributor to estimated lifetime
risk, several  other specific source categories also make significant contributions,
These source categories include chrome platers, background pollutants, roadway
vehicles, and home heating, each estimated  to contribute between  10 to 20% of
the total incidence.  The remaining 5% of risks are divided  among a wide range
of additional  source types.

Another finding is that while the most significant pollutant is coke oven
emissions, several other pollutants also make significant contributions to air
pollution-related cancer risks.  This point is illustrated in Figure 8, showing
a pie chart of the contributions of various pollutants to total estimated number
of cases across the study area.  (As with Figure 4, the percentage contributions
to the lifetime (70 year) number of cases are the  same as the contributions to
annual  cases.)  This figure illustrates the  fact that the combined contribution
of the five most significant pollutants yields only 83% of the total  estimated
number of cancer cases.  The contributions  from the 10 most  significant pollu-
tants must be included to explain 98% of the cases.

A second means of examining cancer impacts  of air  pollution  in Southeast Chicago
is to evaluate individual risks.  Figure 9  presents a map of the  individual
risks estimated in the Southeast Chicago area.  This same information is
presented in a different format in Figure 10.  These figures include background
pollutants, which are assumed to be uniform throughout the area,  representing
a risk of 5x10"^ in each grid square.

The highest estimated lifetime risk in the  study area is about 5xlO~3 (5 chances
in 1,000), at the square centered near 114th Street and Torrence  Avenue.
However, this grid is indicated by Census Bureau data to have no  residents.
(If this grid in fact has any residents, these residents would be included in
exposure estimates for a neighboring grid.   Note that this study  does not address
exposure in nonresidential  locations such as workplaces in this or other grids.)
The grid with the greatest  estimated number of cancer cases  attributed
to air pollution is a grid  where individual  risks  are somewhat lower but where
substantial  population lives.  Specifically, the grid square with the highest
estimated number of cancer cases is centered on the UTM coordinates at 4616
N/455 E (ranging from about 107th Street to 112th  Street and from about Burley
Avenue to Avenue J).  Risk in this grid square is  estimated  to be slighly less
than lxlO~3 (i  in 1,000).  The incidence at this site is estimated to be almost
4 cases (about  3.8) over 70 years.  The estimated  average risk across the entire
study area is about 2.2xlO"4 (about 2 chances in 10,000).

-------
                                               39
      *-
 CO


 O)
 cr>
      (0
      0)
      t/>
      CD
     O
 O
 c
 CD
O

"5

 c
LU

CO   LLJ

o    o
CD   *-
<    W
o    c
<    i;
LU
                                 o
                                 o.
(A
!•
0
Q.
£
0
mc
"o
w
(0
CD
0)
•o
>•
0)
•o
(0
E
£


-------
                                                                    40
                                                                                             00
                                                                                            co
                          J	I
c
03
 O
Q-
 O
 o:
 
-------
                                             41
     CO

     c
     (0
     o
     0.
     <

     o
 •
     s
     E
z   *<
I    <°
o    E
H   ^
(/)    (0
<   UJ
UJ
X

D
O
CO

-------
                                       42

Table 8 summarizes the various contributions to risk at the grid with  the
highest incidence location, showing the components of the risks  in terms of
source type and pollutant.  Clearly steel sources, particularly  coke ovens, are
the dominant source of estimated risk at this location.  Figure  11 shows
contribution to risk at the peak incidence locations in another  format.

Figures 9 and 10 also suggest findings about the spatial  distribution  of risk
in the Southeast Chicago area.  Although the spatial  resolution  in this  study
is insufficient for a detailed examination of spatial  variations, these  figures
do suggest that the highest risks are generally in the predominant downwind
direction (northeast) of the steel  mills near Lake Calumet, and  that risks  in
the southern and western parts of the area are more uniform and  relatively  lower.

Figures 9 and 10 may also be compared with Figure 5.  Figure 5 shows estimated
cancer cases, which reflect population exposed as well  as individual  risks.
Figure 5 shows the most cases in the northern and northeastern parts of  the
study area.

The risk estimates presented in this report should be regarded as only rough
approximations, and should be considered in the context of the substantial
uncertainties that are inherent in  state-of-the-art of risk assessment techniques.
The discussions of emissions estimates, dispersion analysis, and estimation of
unit risk factors have each identified various uncertainties.  A useful  illus-
tration of these unavoidable uncertainties is the improvements that have become
available even during the last six  months.  In estimating emissions, recent
information suggested that chrome plating facilities emit 25%  more than  pre-
viously thought, and information was recently discovered  that  permitted  the
estimation of asbestos emissions from two types of sources.  In  unit risk  factors,
recent revisions indicated a roughly 2-fold increase in the PCB  unit risk
factor, a 9-fold reduction in the methylene chloride unit risk factor, and
deletion of melamine from consideration as a carcinogen.   It is  reasonable  to
presume that even the best estimates that can be developed today are prone  to
have errors of these general  magnitudes.

A further indication of the degree  of uncertainty in this study  can be obtained
by reviewing the comparison of modeled versus monitored concentrations.  This
comparison suggests that some pollutants (particularly the metals)  seem  to  be
reasonably accurately characterized, some pollutants are  suggested to  be
underestimated two- to four-fold, and a few pollutants appear  to be understated
by as much as two or three orders of magnitude.

It should be noted that an assessment of the actual  health effects attributable
to air pollution in the study area  could only be answered by an  epidemiological
study.  Unfortunately, epidemiol oyical studies often produce inconclusive
results, due to the difficulties of obtaining the necessary detailed cancer
statistics, of distinguishing effects of exposure to outdoor air versus  the
effects of other exposures (e.g., indoor air, occupational  exposure, and air
inhaled outside the study receptor  area), and of distinguishing  the effects of
exposure to air pollutants from potentially much larger effects  such as  cigarette
smoking.  In any case, an epidemiol ogical analysis was outside the scope of this
study.  Thus, it is only possible to make qualitative  statements about the
uncertainties inherent in the risk  factors.  In particular, risk factors based
on human data will  generally have less uncertainty than risk factors based  on
animal data.

-------
                                                   43

                       Table 8. Estimated Contributions to Lifetime Cancer Risk at the
                                Grid with the Highest Estimated Number of Cancer Cases
Compound*
Acryl amide
Aery! onitril e
Arsenic
Asbestos
Benzene
Beryl 1 i urn
Butadiene
Cadmium
Carbon Tet .
Chi oroform
Chromium
Coke Oven Em.
Dioxin
Epichl orohydrin
Eth. Dibromide
Eth . Dichloride
Eth. Oxide
Formal dehyde
Gas. Vapors
Hex-chl -benz.
Methyl Chi .
Methyl ene Chi .
Perch! oroeth.
PCB's
POM
Prop. Oxide
Styrene
Trichl oroeth .
Vinyl Chi .
Vinyl idene Chi .
Other
Steel Industrial
Mills Sources
m**
m
2E-5 6E-7

6E-5 3E-7
m
m m
1E-5 4E-8
m
m
4E-7 4E-5
7E-4
5E-7
m
1E-7
5E-7
m m
m
m
m
2E-7
2E-7
m
m
m
m
8E-7
m
Consumer
Sources

9E-8
2E-7



4E-7
3E-6




2E-7
8E-7
2E-6
3E-8
3E-7
2E-7

2E-5


Mobile
Sources

2E-7
4E-6
4E-6
2E-8





9E-8


2E-6
5E-6


7E-6


Waste
Handling
m

m
m

2E-7
m


3E-8
m
1E-8

9E-7
3E-8
m
m

m
1E-7
6E-8
6E-8
Sewaye
Treatment Background
Plants Pollutants Total
m
m
2E-5
3E-7
m 6E-5
m
4E-6
1E-5
1E-5 1E-5
m 4E-7
5E-5
7E-4
6E-7
m
9E-8
m 2E-7
7E-7
3E-5 4E-5
7E-6
2E-7 1E-6
m 3E-8
m 5E-7
m 4E-7
m
3E-b
m
m m
m 9E-7
6E-8
m 6E-8
   TOTAL

*Abbreviations:
      Carbon Tet.
      Eth.
      Gas.
      Hex-chl-benz.
8E-4
5E-5
3E-5
  Carbon tetrachloride
  Ethylene
  Gasoline
  Hexachlorobenzene
2E-5
                      Chi .
                      PCB's
                      POM
                      Prop.
IE-6
2E-7
5E-5
                   - Chioride
                   - Polychlorinated  biphenyls
                   - Polycyclic  organic  matter
                   - Propylene
9E-4
                                                                  or  9x10
                                                                         -4
**To emphasize the higher contributions to risk, three formats are used:
  m - minor - designates risks below IxlO"8 (0.00000001)
      exponential  format used  for risks above lxlO~^:  for example, 6E-5  = 6xlO~^ = .00006

BECAUSE OF UNCERTAINTY IN PROCEDURES, METHODS, ASSUMPTIONS AND DATA,  THESE RISK NUMBERS SHOULD
BE REGARDED AS ONLY ROUGH APPROXIMATIONS AND ARE BEST USED IN A RELATIVE SENSE.  ESTIMATES FOR
INDIVIDUAL POLLUTANTS ARE HIGHLY UNCERTAIN AND SHOULD BE USED WITH PARTICULAR CAUTION.

-------
                                                      44
 0)  C

 O  °

 ED
^  «j
 w  o
UJ -I

 o  <»
-I-  U
     S
     0)
 (1)
 V.
 L.

-------
                                       45

A related issue is whether this study is likely to understate or overstate
actual risks.  The comparison of modeling and monitoring  data suggests that
this  study may understate exposure for many pollutants.   On the other  hand,  the
risk  factors are designed to be more likely too high than too low,  particularly
those risk factors designed as 95% upper confidence limits.  Indeed, for some
pollutants the risk may even be zero, since some of the  30 pollutants  in this
study may not actually be carcinogenic to humans at ambient air concentrations.
However, the net effect of these causes of understatement and overstatement  of
risk  is not cl ear .

An additional  perspective on uncertainty is to review the relative  significance
of those groups of pollutants which are more or less uncertain.  A  measure of
uncertainty of the health data, at least a measure of the uncertainty  that
specific pollutants are indeed carcinogenic, are the classifications of weight
of evidence of carcinogenicity.  Thus, one means of addressing  the  uncertainty
with  respect to health data is to sum the estimated number of cancer cases
estimated for each group of pollutants (i.e. known human  carcinogens,  probable
human carcinogens, and possible human carcinogens).  A means of addressing the
uncertainty in exposure estimates can be derived from the comparison of modeling
versus monitoring concentration estimates.  Using the results in Table 7, the 6
known human carcinogens contribute almost 53% of the estimated  cases,  the 22
probable human carcinogens contribute 47% of the estimated cases, and  the 2
possible human carcinogens contribute less than 0.02% of  the estimated cases.
Thus, the pollutants with the least uncertainty (at least with  respect to the
fact of being carcinogenic) are the most significant, and the pollutants with
the most uncertainty (at least with respect to being carcinogenic)  appear to
be relatively insignificant pollutants.

The modeling-monitoring comparison suggests that the pollutants which  appear
to be most underestimated also appear relatively insignificant.  For example,
even if concentrations of PCBs and chloroform were arbitrarily  assumed to be
100 times higher, PCBs would still only contribute 0.04%  of total air  toxics-
related cancer cases and chloroform would be increased to about 20% of these
cases, producing only a modest increase in the total  estimate of air toxics
related cases per year.  Combining this review with the  review  of carcinogeni-
city, it appears that while some pollutants and some source types may  be signi-
ficantly mischaracterized, the most significant pollutants and  source  types  in
this study also appear to be have relatively more reliable risk estimates than
the less significant pollutants and source types in this  study.

Conclusions

The risk estimates presented in this report jhould be regarded  as only rough
approximations of total cancer cases and individual lifetime risks, and are
best used in a relative sense.  Estimates for individual  pollutants are highly
uncertain and should be used with particular caution.   More detailed  discus-
sions of the uncertainties are included in the respective individual sections
on Emission  Estimation, Estimation of Concentrations by  Atmospheric Dispersion
Modeling, Evaluation of Cancer Risk Factors and Incidence and Risk  Estimates.

This project found atmospheric emissions of 30 pollutants in the study area
which USEPA considers carcinogenic.  Some of these pollutants have  been shown
to be carcinogenic based on human exposure data, and others have been  impli-
cated by animal  studies.

-------
                                       46

This study suggests that about 85 cases of cancer over 70 years or  about  1
cancer case per year in this study area may be attributable to air  pollution.
Further, a peak lifetime risk of about 5xlO~3 (or about 5 chances  in  1,000)  is
estimated in the study area.  However, Census Bureau information does not
indicate any residents in this area.  The square kilometer with the highest
estimated number of cancer cases has an estimated lifetime risk of  slightly
less than lxlO~3 (1 in 1,000).  There is some geographic variability  in  the
risks across the study area.  In general , risks are greatest in the northeast
part of the area and are relatively lower in the southern and western part of
the area.  An average lifetime risk across the area is about 2.2x10"^ (about
2 in 10,000).

In evaluating the sources of airborne risk in this area, steel  mills  contribute
to about 29 cancer cases over a 70 year lifetime (almost 34% of the total).
Emissions from other industrial  facilities, primarily chrome platers, are
estimated to cause approximately 14 cancer cases over a 70 year lifetime
(approximately 16% of the total), and consumer-oriented area sources  (e.g.,
home heating and gasoline marketing) contribute approximately 12 cancer cases
over 70 years (about 14% of the total).  Roadway vehicles are also  estimated to
cause about 12 cases over 70 years (about  14% of the total).  Furthermore, the
background pollutant impacts from formaldehyde and carbon tetrachl oride,  which
contribute an estimated 18 cases of cancer over 70 years (almost the  entire
remaining 22% of the total  incidence) may also be attributed principally  to
industrial facilities, consumer-oriented sources, and roadway vehicle emissions.
In total, these source categories contribute about 99.8% of the risk.

Correspondingly, there are some source categories which appear to contribute
relatively little to airborne risk in this area.  This study suggests that the
sum of air toxic based risks attributable to the handling of both hazardous
and municipal  wastes equals about 0.1% of the total  air pollution related
cancer risk in the area, or about 0.07 cases over 70 years (1 case  in 1,000
years).  Thus, these sources are estimated to pose considerably less  air
toxic risks than other source types evaluated in this study.

Another relatively minor source type is air emissions from wastewater treatment
plants, which were estimated to lead to about 0.14 cancer cases over  70 years
(about 2 cases in 1,000 years), or about 0.1% of the total  area's air
pollution-related incidence.  These risks are somewhat greater than those
from handling hazardous and municipal  wastes, but are still  much smaller  than
several other source types evaluated in this study.

It is useful  to apportion the estimated total  number of cancer cases  according
to the weight of evidence that the pollutants are carcinogenic. According to
USEPA's review of the weight of evidence of carcinogenicity, the 30 pollutants
for which risks were estimated in this study include 6 "known human carcinogens"
22 "probable human carcinogens, and 2 "possible human carcinogens".  Of the
estimated 85 cancer cases per 70 years, almost 53% are attributable to "known
human carcinogens," about 47% are attributable to "probable human carcinogens,"
and about 0.02% are attributable to "possible human carcinogens."

To put the air toxics risk in perspective, it would be desirable to discuss
cancer risks due to other forms of environmental  pollution.  However, this

-------
                                       47

study focused on air pollution risks and did not evaluate risks from other
forms of exposure to environmental  contamination.  Other exposure routes
include exposure through drinking water, skin contact, eating fish or swimming
in lakes (e.g., Wolf Lake) which may contain contaminants, and exposure to
indoor air contaminants including radon.  Also complicating any review of the
relative significance of air pollution is the potential  for other air pollutants
which cannot currently be quantitatively evaluated but nevertheless cause signi-
ficant risks.  Air pollution appears to be an important  cause of environmental
pollution-related cancer cases in this area, but a comparison of airborne risks
to risks from other environmental exposures is outside the scope of this study.

Although specific estimates of individual risks and area-wide cancer cases have
been given in this report, the uncertainties underlying  these estimates dictate
that these estimates be used cautiously.  The specific types of uncertainty
inherent in these estimates have been described in various sections of the
report, and include various uncertainties in estimating  emissions, uncertainty
in quantifying atmospheric dispersion, and various uncertainties in developing
unit risk factors from available human or animal  data.  Also, evidence in this
study suggests that concentrations may generally be understated, whereas unit
risk factors are designed to be more likely to overstate than to understate
risks.  Thus, this study may either overstate or understate risks, and in either
case may provide estimates which differ substantially from true risks.
This study was designed as a screening study of a broad  range of source types
and air pollutants, rather than as a more intensive study of any single source
type or pollutant.  As such, more reliable results could be obtained by further
investigation of several  elements of the study.  Given the evolving nature of
knowledge for the pollutants in this study, a new review of the literature
would likely suggest several modifications in the emissions estimates used in
the study.

The evolutionary nature of these types of study is illustrated by the numerous
improvements that became available during the course of  this study.  In parti-
cular, the emissions inventory update documents several  source categories for
which improvements became available between mid-1987 and the end of 1988.
Similar improvements were developed during the same period for several  unit
risk factors.  It is likely that similar improvements for various source cate-
gories and pollutants will be discovered in the future.

Several  additional  studies are underway which should also help improve the
reliability of the study.  Two studies are underway to  evaluate the signifi-
cance of home wood combustion in the Southeast Chicago area.  One study is
analyzing atmospheric monitoring data for characteristic pollutants emitted
by wood combustion to discern the relative contribution  of home wood combustion.
A second study is a telephone survey polling Southeast Chicago area residents
on their actual wood usage.  Another study is underway to evaluate emissions
from abandoned hazardous  waste sites.  A fourth study, being conducted in
Cincinnati, Ohio, is evaluating the extent to which volatilization of contami-
nants in wastewater occurs in sewers, i.e., prior to arrival  at wastewater
treatment plants.

Another relevant study is being performed by the Illinois Cancer Council  as
mandated by the Illinois  General  Assembly.  This study is an epidemiol ogical
investigation of leukemias and lymphomas.

-------
                                       48

Other studies could also be conducted to improve the reliability of the risk
estimates in this study.  Further studies of unit risk factors, while expensive,
could substantially improve risk estimates.  Forthcoming summertime monitoring
data for formaldehyde could be incorporated.  Monitoring focusing on carbon
tetrachloride could be conducted.  Further investigation to resolve discrepancies
between monitored and modeled concentrations could be conducted.

In addressing the risks identified in this study, the USEPA is subject  to some
important limitations in the legal authority provided in the Clean Air  Act and
other legislation for air toxics regulations.  The development of regulations
under these statutes requires intensive investigations.  Also, USEPA's  policy
is to address source categories which are significant from a national  perspec-
tive, not to address individual  sources which are significant in a given local
area.  Thus, State^and local air pollution agencies in the area have an impor-
tant role in identifying and adopting regulations to address the risks  estimated
in this study.  At the same time, the State and local air pollution agencies
are also subject to limitations in statutory authority for addressing  these
issues.

Despite its limitations on authority, USEPA is developing regulations at a
national  level  for numerous categories.  For coke ovens, USEPA has proposed
regulations to require more control  of leaks from doors, lids, offtakes, and
from charging.  For coke oven by-product recovery plants, USEPA has proposed
regulations to require numerous measures to reduce benzene emissions.   For
chrome platers, the Agency is nearing completion of a background information
document which is necessary to provide technical  support for regulating these
sources.   For gasoline vapor emissions from automobile refueling, the  Agency
has proposed two alternatives (controlling these vapors either with equipment
built into automobiles or with equipment installed at gasoline stations) and
is working to resolve technical  concerns about these alternatives.  For facili-
ties that treat, store, and dispose of hazardous waste, an assortment  of
regulations are being developed for proposal.  For chromium emissions  from
comfort cooling towers, the Agency has proposed banning the use of chromium in
these units and, thereby, eliminating these emissions.

An important relevant type of State activity is the adoption of regulations
designed  to reduce emissions of volatile organic pollutants and particulate
matter and thus reduce air toxics emissions.  One example of such a regulation
is the regulation for the coke by-product recovery plants adopted by Indiana,
which was designed to reduce volatile organic compound emissions for ozone
control, but which will  significantly reduce benzene emissions.  A second
example is motor vehicle inspection and maintenance programs adopted by both
Illinois and Indiana, which again was designed to reduce volatile organic
compound  emissions in general, but which will also simultaneously reduce emis-
sions of several  individual  organic species.  Reductions in organic emissions
also have the effect of reducing the risks from formaldehyde which is  photo-
chemical ly created in the atmosphere.  Note that for both of these examples,
the required emission reductions have not been reflected in the emissions and
risks estimated in this study because these programs had not been effectively
started up in the year selected for this study.  A third example is enforcement
of existing regulations controlling volatile organic compound and particulate
matter emissions.  Illustrative of this activity are enforcement negotiations

-------
                                       49

which are leading to improved control  of one of the coke batteries in  the
area.  Further reductions in emissions of volatile organic  compounds  and
particulate matter are mandated both for Southeast Chicago  and for Northwest
Indiana, which can be expected to further reduce emissions  and risks  from the
pollutants in this study.

Other State programs more directly address air toxics  issues.   Illinois currently
considers air toxics in reviewing companies'  air pollution  permit  applications,
and both Illinois and Indiana are in the process of developing more formalized
air toxics programs.  USEPA believes this report documents  sufficient  risk to
warrant investigation of possible means for its reduction.   It is  hoped that
this study will  lead to informed discussions on how to design  USEPA1 s  and the
States'  control  programs to achieve effective reductions to cancer risks  in the
Southeast Chicago area.

-------
                                       50

                                Refernces
     Many references were used in developing emissions estimates.  The
     inventory is described in the following two reports, which provide
     detailed data on references:

     J. Sutmerhays , H, Croke, "Air Toxics Emission Inventory for the Southeast
     Chicago Area," Region V, U.S. Environmental Protection Agency, July 1987.

     J. Summerhays , "Update to an Air Toxics Emission Inventory for the
     Southeast Chicago Area," Region V, U.S. Environmental  Protection Agency,
     January 1989.

Dispersion Modeling

     Guideline on Air Quality Models (Revised), EPA-450/2-78-027R, Office of
     Air Quality Planning and Standards, U.S. Environmental Protection Agency,
     1986.

     Industrial Source Complex (ISC) Dispersion Model User's Guide - Second
     Edition (Revised), EPA-4 50/4-88-002 , Office of Air Quality Planning
     and Standards, U.S. Environmental  Protection Agency, December 1987.

     J. Irwin, T. Chi co, J. Catal ano , COM 2.0 - C1 imatological  Dispersion
     Model User's Guide, EPA-60U/8-85-029, Atmospheric Sciences Research
     Laboratory, U.S.  Environmental  Protection Agency, 1985.

Monitoring

     Urban Air Toxics Program (UATP), First Quarterly Report, Fourth Quarter
     1987, Radian Corp., April 1988.

     Urban Air Toxics Program (UATP) Second Quarterly Report, First Quarter
     1988, Radian Corp., July 1988.

     Toxic Air Monitoring System Status Report, Office of Air Quality Planning
     and Standards, U.S. Environmental  Protection Agency, February 1988.

     P. Aronian, P. Scheff, R. Madden,  "Winter Time Source-Reconciliation
     of Ambient Organics,"  Illinois Institute of Technology/University
     of Illinois at Chicago,  Paper presented at Annual Meeting  of Air
     Pollution Control  Association, June 1988.

     C. Sweet, D. Gatz, Atmosphere Research and Monitoring  Study of Hazardous
     Substances:  Third Annual Report,  Hazardous Waste Research and Information
     Center, Illinois  Department of Energy and Natural Resources, November 1987,

-------
                                       51
An Interim Rep or t__o i_n_t _he_ jtesuj_t_s_ jp t_
                                  '
                                                         -e
                      __ __ __     ___   _   _               _
     Area of Southeast Chicago, Divisio'n of Air Pollution Control, IllTnofs"
     Environmental  Protection Agency, May 1987.

Unit Risk Factor Data

     Each pollutant in this study has a body of literature that was considered
     in the development of the unit risk factor.  However, Region V, in con-
     ducting this study, did not itself develop any unit risk  factor and
     did not conduct any associated literature review.   Readers interested
     in the literature relevant to unit risk factors for particular
     pollutants are advised to consult the office identified  in Table 6
     as the source  of the factor.

General References

     E. Haemisegger, A. Jones, B. Steigerwald, V. Thomson, The Ai r Toxi cs
     Problem in the United States: An Analysis of Cancer Risks for Selected
     Pollutants, Office of Air and Radiation/ Off ice of  Policy, Planning and
     Evaluation, U.S. Environmental  Protection Agency,  May 1985.

     The Risk Assessments Guidelines of 1986,  U.S.  Environmental  Protection
     Agency, originally published in 51 Federal Register 33992-34052,
September 24,
August 1987.
                   1986, and reprinted as EPA Report #600/8-87-045 in

-------