Risk Assessment Report
for the Sterigenics Facility in Willowbrook, Illinois
EPA's Office of Air Quality Planning and Standards
Office of Air and Radiation
August 2019
-------
Table of Contents
Executive Summary 3
1 Introduction 6
2 Methods 7
2.1 Emissions and source data 8
2.2 Dispersion modeling for inhalation exposure assessment 8
2.3 Estimating chronic human inhalation exposure 15
2.4 Acute risk screening and refined assessments 16
2.5 Dose-response assessment 17
2.5.1 Sources of chronic dose-response information 17
2.5.2 Sources of acute dose-response information 20
2.6 Risk characterization 22
3 Risk results for the Sterigenics facility in Willowbrook, IL 23
3.1 Risk assessment results for baseline emissions 24
3.2 Risk assessment results for the illustrative future scenario 25
4 General discussion of uncertainties in the risk assessment 28
4.1 Emissions inventory uncertainties 28
4.2 Exposure modeling uncertainties 28
4.3 Uncertainties in the dose-response relationships 30
5 References 38
Index of Tables and Figures
Table 2.2 - 1. AERMOD version 18081 Model Options for Risk Assessment Modeling 9
Table 2.2 - 2. Monitor Receptors 11
Figure 2.2-1. Census Block and Monitor Location Receptors 12
Figure 2.2 - 2. Gridded Residential and Commercial/Industrial Receptors 13
Figure 2.2 - 3. Polar Receptors 14
Table 3.1 - 1. Inhalation Risks for the Sterigenics Willowbrook, Illinois Facility - Baseline
Emissions 24
Figure 3.1 - 1. Modeled Lifetime Cancer Risks for Sterigenics in Willowbrook, IL 26
Figure 3.1 - 2. Modeled Non-Residential Cancer Risks for Sterigenics in Willowbrook, IL.. 27
Appendices
Appendix 1 Development of Ethylene Oxide Emissions Rates Used for Risk Assessment
Appendix 2 Technical Support Document for HEM-3 Modeling
Appendix 3 Meteorological Data for HEM-3 Modeling
Appendix 4 U.S. EPA Risk Assessment for Sterigenics-Willowbrook (Slides from May 29,
2019, Public Meeting)
1
-------
Index of Acronyms
AERMOD
American Meteorological Society/EPA Regulatory Model
AEGL
Acute exposure guideline level
ASTDR
US Agency for Toxic Substances and Disease Registry
CalEPA
California Environmental Agency
ERPG
Emergency Response Planning Guideline
HAP
Hazardous Air Pollutant(s)
HEM
Human Exposure Model
HI
Hazard index
HQ
Hazard quotient
IRIS
Integrated Risk Information System
MACT
Maximum Achievable Control Technology
MIR
Maximum Individual Risk
MOA
Mode of action
NAC
National Advisory Committee
NAAQS
National Ambient Air Quality Standards
NATA
National Air Toxics Assessment
NEI
National Emissions Inventory
PB-HAP
Persistent and Bioaccumulative - HAP
PAH
Polycyclic aromatic hydrocarbon
POM
Polycyclic organic matter
REL
Reference exposure level
RfC
Reference concentration
RfD
Reference dose
RTR
Risk and Technology Review
TOSHI
Target-organ-specific hazard index
LIRE
Unit risk estimate
2
-------
Executive Summary
This document describes the risk assessment that the U.S. Environmental Protection Agency
(EPA) conducted to assess the human health risks posed by emissions of the hazardous air
pollutant (HAP) ethylene oxide (EtO) from the Sterigenics facility in Willowbrook, IL. The
facility is a commercial sterilizer subject to the National Emission Standards for Hazardous
Air Pollutants (NESHAP) for Ethylene Oxide Commercial Sterilization and Fumigation
Operations under 40 CFR part 63, subpart O. Facilities in the commercial EtO sterilization
source category, including the Sterigenics facility in Willowbrook, are engaged in commercial
sterilization or fumigation using EtO as a sterilant for heat- and moisture-sensitive products
and as a fumigant to control microorganisms or insects. Generally, EtO is used to sterilize or
fumigate medical equipment (e.g., syringes and surgical gloves), spices, pharmaceuticals, and
cosmetics. Emission points included in the assessment are those where EtO can be released
during the sterilization cycle, including: sterilization chamber vent(s); sterilization chamber
vacuum pump drain; chamber exhaust vent(s) (i.e., the "backvent"); aeration room vent(s);
and fugitives.
The EPA conducts risk assessments for regulatory and non-regulatory purposes. The risk
assessment described in this document is not part of a regulatory activity, however, the
approaches the EPA used in this assessment are similar to those used in the regulatory
residual risk and technology review (RTR) program. Typically, the risk assessments we
perform are conducted under Section 112 of the Clean Air Act (CAA), which establishes a
two-stage regulatory process for addressing emissions of HAP from stationary sources. In the
first stage, the EPA must promulgate technology-based NESHAP for categories of sources.
For NESHAP that require maximum achievable control technology (MACT) standards, the
EPA is required to complete a second stage of the regulatory process eight years after
adopting the MACT standards, which is known as the residual risk review. In this second
stage, the EPA is required to assess the health and environmental risks that remain after
implementation of the technology-based standards. The EPA must also review each of the
technology-based standards at least every eight years and revise them, as necessary, taking
into account developments in practices, processes and control technologies. For efficiency,
the Agency includes the analyses for both reviews in the same regulatory package and calls
these rulemakings Risk and Technology Reviews (RTRs). The EPA completed the RTR
review for the commercial EtO sterilization NESHAP in 2006.
This risk assessment examined two scenarios: (1) a baseline scenario reflecting operations of
the facility prior to a February 2019 Seal Order issued by the State of Illinois (facility
emissions under this scenario are approximately 4,000 pounds per year); and (2) an
illustrative future scenario in which all emission points are routed to a control device and are
released to the atmosphere from a single 26.5 m (87 ft) stack (facility emissions under this
scenario are 26 pounds per year). EtO was the only pollutant included in this risk assessment.
We only assessed human health risks from EtO inhalation exposures. EtO is not a persistent
and bioaccumulative HAP (PB-HAP), therefore a multipathway risk assessment is not
warranted. The EPA evaluates 8 HAP for adverse environmental effects. These
3
-------
"environmental HAP" were selected by the EPA based on their persistence and
bioaccumulation potential, magnitude of emissions, and relative environmental toxicity.
Because EtO is not an environmental HAP, an environmental risk screening assessment is not
warranted.
Several key points about this risk assessment are worth noting. The assessment:
• Assumes people are exposed to ethylene oxide 24 hours a day, 365 days a year for 70
years to represent lifetime exposures (non-residential exposure1 durations are lower).
• Estimates the risk of getting cancer that is in addition to people's overall risk of
getting cancer for other reasons.
• Focuses only on the risk from ethylene oxide emissions from the Sterigenics facility (it
does not address comprehensive risk from all pollutants and all air pollution sources).
• Projects risk going forward. It does not estimate past risk.
• Provides general estimates of risk to populations. It cannot predict any one person's
risk of developing cancer.
• Is more likely to over-estimate risk than underestimate risk due to what we call
"health-protective" assumptions
The table below summarizes the results of the baseline risk assessment for the facility. The
results of the chronic (long-term, i.e., 70-year lifetime) inhalation cancer risk assessment
indicate that the maximum lifetime (residential) individual cancer risk is 1,000-in-l million.2
The total estimated cancer incidence3 from this facility is 0.3 excess cancer cases per year, or
one excess case in every three years. Approximately 7.7 million people live within 50
kilometers of this facility and 60 people are estimated to have cancer risks equal to 1,000-in-l
million from EtO emitted from this facility. The estimated non-residential maximum cancer
risk is also 1,000-in-l million. It is a coincidence that the risk results for non-residents and
residents were the same: the analyses for these two populations were based on different
modeled ambient concentrations and different exposure assumptions (see Section 2.3 for
details). Population risks are not estimated for the non-resident scenario because there are no
data available to estimate with specificity where people would be, or for how long, or how
many people there would be at specific locations.
1 Non-residential locations are where people could spend a significant amount of time, but less than a lifetime
(for example, an offsite worker).
2 Risk results are typically presented by the EPA using one significant figure in light of the uncertainties inherent
in these analyses - see, for example, Section 4 of this document.
3 In this risk assessment context, estimated cancer incidence is the predicted (based on modeling) number of
excess cancer cases per year due to emissions of ethylene oxide from Sterigenics. It is not a count of actual
cancer cases, which might be provided in other types of studies.
4
-------
Risk Summary for the Sterigenics Facility in Willowbrook, Illinois
"Baseline" Scenario Reflecting Emissions Prior to Seal Order
Inhalation
Cancer Risk
Population Cancer Risk
Max Chronic
Individual
Noncancer Risk
Max Acute
Noncancer
Risk
Maximum
Individual Risk
(in 1 million)
Cancer
Incidence
(cases per Year)
Cancer
Incidence
(Years for 1 case)
= 1000 in 1
million
>100 in 1
million
> 1 in 1
million
Hazard Index
(TOSHI)
Hazard
Quotient (HQ)
Residential
1,000
0.3
3
60
11,500
6,500,000
0.01
0.02
Non-
Residential
1,000
NAa
NAa
NAa
NAa
NAa
0.01
0.02
a NA = not applicable. Population risks were not estimated for non-residents because there are no data available to estimate with specificity where people would
be, or for how long, or how many people there would be at specific locations.
The EPA also examined noncancer risk as part of the assessment, finding the residential
maximum chronic noncancer hazard index (neurological) for the facility is 0.01. Of the
approximately 7.7 million people living within 50 kilometers of the facility, no one is exposed
to noncancer hazard index levels above 1. The non-residential maximum chronic noncancer
hazard index for the facility is 0.01. The low hazard index estimates indicate that we do not
expect any chronic noncancer effects to occur.
Regarding acute (short-term) noncancer health risks posed by baseline emissions, the highest
screening acute hazard quotient is estimated to be 0.02 using the AEGL-24 value for EtO. This
hazard quotient is based on a 1-hour exposure anywhere off facility property, so there is no
distinction made between resident and non-resident. The low hazard quotient estimates
indicate that we do not expect any acute noncancer effects to occur.
The table below summarizes the results of the risk assessment for the illustrative future
scenario. The maximum lifetime (residential) individual cancer risk is 1 -in-1 million, which
occurs at a single residential grid receptor. All cancer risks at census blocks are less than
1-in-l million. The total estimated cancer incidence is 0.002 excess cancer cases per year, or
one excess case in every 700 years within the entire modeling domain. Over 70 years, the
estimated number of cancer cases is less than 1 (0.1). Approximately 70,000 people are
estimated to have cancer risks between 0.1- and 1-in-l million, so the remaining 7.6 million
people within the modeling domain have estimated cancer risk less than 0.1-in-l million. The
maximum chronic noncancer hazard index is 6E-6 (neurological). For non-residential
exposures, the maximum cancer risk is 0.08-in-l million, and the maximum chronic
noncancer hazard index is 9E-7 (neurological). The highest screening acute HQ was 4E-6
(based on the 1-hr AEGL-2 value for EtO). These estimates indicate low cancer risk and we
do not expect any chronic or acute noncancer effects to occur.
4 Acute exposure guideline levels (AEGLs) describe the human health effects from once-in-a-lifetime, or rare,
exposure to airborne chemicals. The AEGL-2 is the airborne concentration (expressed as ppm or mg/m3) of a
substance above which it is predicted that the general population, including susceptible individuals, could
experience irreversible or other serious, long-lasting adverse health effects or an impaired ability to escape.
5
-------
Risk Summary for the Sterigenics Facility in Willowbrook, Illinois
Illustrative Future Scenario
Inhalation
Cancer Risk
Population Cancer Risk
Max Chronic
Individual
Noncancer Risk
Max Acute
Noncancer
Risk
Maximum
Individual Risk
(in 1 million)
Cancer
Incidence
(cases per Year)
Cancer
Incidence
(Years for 1 case)
= 1000 in 1
million
>100 in 1
million
> 1 in 1
million
Hazard Index
(TOSHI)
Hazard
Quotient (HQ)
Residential
la
0.002
700
0
0
0
6E-6
4E-6
Non-
Residential
0.08
NAb
NAb
NAb
NAb
NAb
9E-7
4E-6
a The maximum risk of 1 -in-1 million occurs at a single residential receptor. All cancer risk estimates at census blocks are less than 1 -in-1 million, so the
population estimated to be greater than or equal to 1 -in-1 million is zero.
b NA = not applicable. Population risks were not estimated for non-residents because there are no data available to estimate with specificity where people would
be, or for how long, or how many people there would be at specific locations.
This document summarizes the methods and results of the risk assessment for this facility.
Section 1 provides an introduction to the risk assessment, including key questions to be
addressed. Methods described in Section 2 include those used by the EPA to develop refined
estimates of chronic inhalation exposures and human health risks for cancer and noncancer
endpoints, as well as those used to screen for acute health risks. The risk assessment results
are presented in Section 3. Section 4 contains a discussion of the uncertainties of the risk
assessment, including uncertainties in the exposure assessment and in the dose-response
values. The appendices to this risk report contain detailed descriptions of the methods used to
develop emissions estimates, process meteorological data, and conduct dispersion modeling.
1 Introduction
The EPA conducts risk assessments for regulatory and non-regulatory purposes. The risk
assessment described in this document is non-regulatory, however the approaches the EPA
used in this assessment are similar to those used in the regulatory residual risk and technology
review (RTR) program. More information on the RTR program, source categories included in
the program, the EPA's statutory authorities, and our risk-related framework for decision
making can be found on the RTR website at https://www3.epa.gov/ttn/atw/rrisk/rtrpg.html.
The EPA conducted this risk assessment for EtO emissions from the Sterigenics facility in
Willowbrook, Illinois to answer several questions:
• What is the estimated maximum cancer risk in the area of highest concentration where
people live?
• What is the estimated maximum cancer risk in the area of highest concentration where
people work (offsite - not at the facility)?
• How many people have different levels of risk in the neighboring communities?
• What is the estimate of possible cancer cases per year?
The assessment is not designed to predict any individual's risk. Also, it cannot look
retrospectively at potential risk experienced in the past, e.g., from the time the facility opened
until today. It is designed to assess risks from EtO emissions from this specific facility, not
6
-------
all risks from EtO exposure that an individual may face. Additional limitations or
uncertainties are described in Section 4.
The remaining sections of the document contain the methods we used to conduct the risk
assessment (Section 2), the results of the risk assessment (Section 3), and a description of
associated uncertainties (Section 4). More detailed information about some of the inputs can
be found in the appendices.
2 Methods
A risk assessment consists of four steps: 1) hazard identification, 2) dose-response
assessment, 3) exposure assessment, and 4) risk characterization. The first step, hazard
identification, determines whether the pollutants of concern can be linked to the health effects
in question (cancer and/or noncancer). In our regulatory program, Section 112 of the CAA
identifies the HAP to be considered in the risk assessment for a source category. For this
facility-specific risk assessment, we are assessing the HAP EtO in the hazard identification
step. The second step is the dose-response assessment, which quantifies the relationship
between the dose of a pollutant and the resultant health effects. Dose-response assessments
are performed by the EPA through the Integrated Risk Information System (IRIS) process as
well as by other agencies, such as the Agency for Toxic Substances and Disease Registry
(ATSDR). See Section 2.5 of this document for more information on dose-response
assessments. The third and fourth steps, the exposure assessment and the risk characterization,
respectively, are specific to the facility and are described throughout this report. The exposure
assessment includes characterization of HAP emissions, environmental fate and transport, and
population exposure for the inhalation pathway. The fourth and final step, risk
characterization, integrates all the information from the previous steps and describes the
outcome of the assessment. This four-step approach to risk assessment was endorsed by the
National Academy of Sciences in its publication "Science and Judgment in Risk Assessment"
(NAS, 1994) and subsequently was adopted in the EPA's "Residual Risk Report to Congress"
(USEPA, 1999).
The EPA conducts risk assessments that provide estimates of the maximum individual risk
(MIR) posed by the HAP emissions from each source, the target-organ-specific hazard index
(TOSHI) for chronic exposures to HAP with potential to cause chronic (or long-term)
noncancer health effects, and the hazard quotient (HQ) for acute exposures to HAP with the
potential to cause acute (or short-term) noncancer health effects. The MIR is defined as the
cancer risk associated with a lifetime of exposure at the highest concentration of HAP where
people are likely to live. The HQ is the ratio of the potential exposure to the HAP to the level
at or below which no adverse effects are expected; the TOSHI is the sum of chronic HQs for
HAP that affect the same target organ or organ system. The risk assessment also provides
estimates of the distribution of cancer risks within the exposed residential populations as well
as cancer incidence. The following sections describe how we estimate HAP emissions and
conduct steps three and four of the risk assessment. The methods used to assess risks are
consistent with those peer-reviewed by a panel of the EPA's Science Advisory Board (SAB)
in 2009 (USEPA, 2009a) and described in their peer review report issued in 2010 (USEPA
2010). In 2017, we submitted updated methodologies to SAB for review. The updated
methodologies are described in, "Screening Methodologies to Support Risk and Technology
7
-------
Reviews (RTR): A Case Study Analysis" (USEPA, 2017a). The SAB's findings for this
review, "Review of EPA's draft technical report entitled Screening Methodologies to Support
Risk and Technology Reviews (RTR): A Case Study Analysis" (USEPA, 2018a) were
submitted to the EPA in September 2018.
2.1 Emissions and source data
The Sterigenics Willowbrook facility consists of two buildings separated by approximately
100 meters (m). To develop baseline emissions estimates and other source data for the facility,
we used information provided by Sterigenics regarding their operations and estimated
emissions rates and operational parameters from both the controlled and uncontrolled sources.
We used this information and derived site-specific emission factors from previous stack testing
results for the "controlled" sources and estimated site-specific emission factors for the
uncontrolled or "fugitive" emissions. Emissions factors are representative values that attempt
to relate the quantity of a pollutant released to the atmosphere with an activity associated with
the release of that pollutant and are generally assumed to be representative of long-term
averages. Using dispersion modeling, the EPA evaluated the accuracy of these factors and
made the necessary adjustments to these factors to better match the observed ambient
measurement values at the monitoring sites near the facility with the modeled value. The total
EtO baseline emissions from the facility are approximately 2 tons per year and come from the
two buildings. Each building has the following sources: sterilizer vacuum pump, aeration
room, sterilizer back vent, and fugitives. Details on the development of the source data,
emissions, and associated uncertainties for the baseline emissions data for this facility can be
found in Appendix 1 {Development of Ethylene Oxide Emissions Rates Usedfor Risk
Assessment). We also assessed an illustrative future scenario, where we assumed that all
emissions come from one building, and that all remaining emissions come from one stack. We
assumed that all fugitives are captured and routed to a control device. Future case emissions
are estimated at 26 lbs/yr.5
2.2 Dispersion modeling for inhalation exposure assessment
For risk analyses, we estimate both long- and short-term inhalation exposure concentrations
and associated health risks from each facility of interest. To do this, we use the Human
Exposure Model 3 (HEM-3), which includes the American Meteorological Society/EPA
Regulatory Model (AERMOD) for dispersion modeling. HEM-3 performs three main
operations: atmospheric dispersion modeling, estimation of individual human exposures and
health risks, and estimation of population risks. The approach used in applying this modeling
system for the assessment of Sterigenics is outlined below and is similar to the approach used
for assessments conducted under the RTR program. Details on the use of HEM-3 for RTR
5 This scenario was developed considering information available to EPA in April/May 2019, such as a draft
permit application for another commercial sterilizer in Illinois and conversations with the state and the company
on a possible control scenario. Subsequently, a draft permit for Sterigenics was issued by Illinois EPA on July
15, 2019, based on a permit application submitted by the company on June 24, 2019. The draft permit (and
associated permit application) reflect similar, albeit not identical, emissions and operating parameters. For
example, allowable emissions in the draft permit, while lower than our estimated baseline emissions, are
somewhat higher than our illustrative future emissions. As a result, calculated risks (for these higher future
emissions) would be greater than those modeled in this assessment but are still in the range of 1 - to 10-in-l
million.
8
-------
assessments are provided in Appendix 2 to this document (Technical Support Document for
HEM-3 Modeling). This section focuses on the dispersion modeling component.
The dispersion model in HEM-3, AERMOD version 18081, is a state-of-the-science Gaussian
plume dispersion model that is preferred by the EPA for modeling point, area, and volume
sources of continuous air emissions from facility applications (USEPA, 2017b). Further
details on AERMOD can be found in the AERMOD User's Guide (USEPA, 2018b) and the
AERMOD Implementation Guide (USEPA, 2018c). The model is used to estimate annual (or
multi-year) average ambient concentrations through the simulation of hour-by-hour dispersion
from the emission sources into the surrounding atmosphere. Unless data are available on the
hours of operation for a source category, default hourly emission rates used for this simulation
are generated by evenly dividing the total annual emission rate from the inventory into the
8,760 hours of the year.
The first step in the application of HEM-3 is to predict ambient concentrations at locations of
interest. The AERMOD model options used for this assessment are summarized in
Table 2.2-1 and are discussed further below.
Table 2.2 -1. AERMOD version 18081 Model Options for Risk Assessment Modeling
Modeling Option
Selected Parameter for chronic exposure
Type of calculations
Hourly ambient concentration
Source types
Point
Receptor orientation
Polar (13 rings and 16 radials)
Discrete (census block centroids, monitor locations, and
additional gridded receptors)
Terrain characterization
Actual from USGS 1/3-arc-second DEM data
Building downwash
Included
Plume deposition/depletion
Not included
Urban source option
Urban (population = 50,000)
Meteorology
5-year representative data from nearby sites (Argonne
National Lab and Midway Airport) for years 2014-2018
In HEM-3, meteorological data are ordinarily selected from a list of more than 800 National
Weather Service (NWS) surface observation stations across the continental United States,
Alaska, Hawaii, and Puerto Rico, and HEM-3 defaults to the station closest to each modeled
facility. We use data from other stations in special circumstances if we have reason to believe
that other data are more representative for certain facilities. The NWS station closest to the
Sterigenics facility is Chicago Midway International Airport (approximately 16 km east).
While Midway can be considered adequately representative of the facility in the absence of
other data, given the proximity of Argonne National Laboratory to the facility (7 km
southwest), the EPA concluded that meteorological data collected at Argonne would be more
representative of conditions at the facility than data from Midway. The Argonne
9
-------
meteorological tower had measurements of wind, temperature, and turbulence (standard
deviation of wind direction) at 10 m and 60 m vertical levels, making a more robust dataset
over standard airport observations which have one level of data without turbulence
measurements. Missing data for some parameters in the Argonne data were supplemented
with data from Midway. Upper air data were obtained from the nearest NWS site with such
data available, which is Davenport Municipal Airport in Davenport, Iowa. We processed 5
years of data for the years 2014 through 2018 (the most recent five full years available) using
the AERMET meteorological data preprocessor. In 2016, the Agency released to the public on
the EPA's Support Center for Regulatory Atmospheric Modeling (SCRAM) website both
AERMET and AERMOD (version 18081). Appendix 3 to this document (Meteorological
Data for HEM-3 Modeling) provides detailed information on the sources of meteorological
data, why we selected the data we used, and how we processed those data for use in
AERMOD.
The HEM-3 model estimates ambient concentrations at the geographic centroids of populated
census blocks (using the 2010 Census) and at a set of "polar" receptors, which are the
intersection points of a set of concentric rings and outward radials that are centered on the
facility. Census blocks are the finest resolution data available in the Census, and each block
contains approximately 50 people or about 20 households based on national averages. The 50
km (radius) modeling domain centered on the Sterigenics facility is more densely populated
than the national average, with the average block in the modeling domain containing about 70
people. We calculate long-term exposure and risk at the census blocks, and we also model
short-term concentrations at the blocks. The population data for the census blocks are used to
calculate cancer incidence and population risks. The polar receptors are used to estimate long-
and short-term exposures at locations that may be closer to the facility than the census blocks
(for example, to represent a residence that is closer). The polar receptors are also used to
interpolate values for census blocks far from the facility because by default HEM-3 only
explicitly models (in AERMOD) block locations within 3 km of the facility. For this
assessment, we used polar receptors based on the HEM-3 default of 13 concentric rings and
16 radials (one every 22.5 degrees), but HEM-3 does allow the user to change the number of
rings and radials. In addition to the census blocks and polar receptors, we also included a set
of nested grid receptors, which were spaced 50 m apart within a 1 km square centered on the
facility and spaced 100 m apart within a 2 km square centered on the facility. Using these
dense grid receptors near the facility allowed for the estimation of exposures at potential non-
residential locations where people could spend a significant amount of time, but less than a
lifetime (for example, an offsite worker). Finally, we included as receptors the locations of
ambient monitors that collected air samples from mid November 2018 to the end of
March 2019. The coordinates of the monitors are given in Table 2.2-2, along with the distance
and direction from the facility.
10
-------
Table 2.2 - 2. Monitor Receptors
Monitor
Longitude
Latitude
Distance and Direction from
Facility
EPA warehouse
-87.938738
41.747442
100 mSE
Gower Elementary School
-87.956186
41.748843
1.2 km W
Gower Middle School
-87.933929
41.743473
700 m SE
Hinsdale South High
School
-87.948504
41.753694
900 mNW
Village Hall
-87.941100
41.748598
100 mNW
Water tower
-87.939173
41.755373
800 mN
West neighborhood
-87.945561
41.748773
400 mW
Willow pond park
-87.939850
41.763988
1.7 km N
Figure 2.2-1 shows the populated census blocks near the facility, along with the boundaries of
those blocks. The monitor locations are also given in this figure. Figure 2.2-2 shows the
nested grid of receptors, distinguished by whether they fall in residential areas or non-
residential (commercial/industrial) areas. Figure 2.2-3 shows the first five rings of the polar
receptors, with the first ring set by default to include all emission points at the facility.
HEM-3 accounts for the effects of multiple facilities when estimating concentration impacts
at each block centroid. We typically combine the impacts of all facilities within the same
source category and assess chronic exposure and risk for all census blocks with at least one
resident (i.e., locations where people may reasonably be assumed to reside rather than
receptor points at the fenceline of a facility). For this assessment, we considered only the
Sterigenics facility. We calculate long-term ambient concentrations as the annual (or multi-
year) average of all estimated short-term (one-hour) concentrations at each receptor. We do
not consider possible future residential use of currently uninhabited areas, but this would not
impact this assessment because the areas around the facility are already fully developed.
We determine census block elevations for HEM-3 nationally from the US Geological Survey
1/3 Arc Second National Elevation Dataset, which has a spatial resolution of about 10 meters.
We also used these elevation data to estimate elevations of the nested grid receptors. Each
polar receptor is assigned the highest elevation of any census block in its neighborhood (all
blocks closer to that polar receptor than any other polar receptor). If an elevation is not
provided for an emission source, HEM-3 uses the average elevation of all polar receptors on
the innermost polar ring. However, we used the National Elevation Dataset to estimate source
elevations. There is very little elevation variance near the facility, with differences less than
five meters within several hundred meters of the facility.
We ran AERMOD in urban mode (versus rural mode), which accounts for the dispersive
nature of the "convective-like" boundary layer that forms during nighttime conditions due to
the urban heat island effect. We concluded the urban mode is most appropriate for modeling
the Sterigenics facility. The facility is located within the Chicago-Joliet-Naperville urbanized
area, and although Willowbrook is considered suburban and the land use around the facility is
mostly low to middle density developed areas, we considered the potential for urban heat
island influences across the full modeling domain which includes the nearby large urban area
11
-------
*Willowb]ro&M
Figure 2.2 -1. Census Block and Monitor Location Receptors
Sterigenics
Receptors
A Monitors
Populated Census Blocks
Census Block Boundaries
0 100 200 300 400 500 Meters!
12
-------
HollS
• Elm«fct
Hinsdale South
-M-ligh'School
Pliddle
sr
tx •
Figure 2.2 - 2. Gridded Residential and Commercial/Industrial Receptors
¦¦ Sterigenics
Receptors
o Commercial/Industrial
o Residential
100 200 300 400 500 Meters
13
-------
HollS
Roger Rd
CB^SSalmSouth \
Pill^'Schooi
Midv^y-D-
I, iGower^Middle
^UpflgoT
co 80th St
8 2nd-St-
Figure 2.2 - 3. Polar Receptors
¦ Sterigenics
Receptors
o Polar Receptors
0 100 200 300 400 500 Meters
14
-------
of Chicago. Also, most of the areas around the facility have a population density that exceeds
the 750 people per square kilometer criteria recommended in the AERMOD Implementation
Guide for inclusion as urban. The magnitude of the urban effect in AERMOD is based on an
empirical relationship between urban/rural temperature differences and population, and
AERMOD requires a population value when in urban mode. Because using the population of
the entire metropolitan area (about 9.5 million people) could overstate the urban heat island
effect, and to be health protective, we used the minimum population allowed in HEM-3,
which is 50,000 people.
To assess the potential impacts from short-term exposures, we estimated worst-case one-hour
concentrations at the census block centroids and at points closer to the facility (using either
the polar receptors or the grid receptors) where people may be present for short periods. Note
that this differs from the estimation of ambient concentrations for evaluating long-term
exposures, which we perform only for populated census blocks and residential grid and polar
receptors. Because short-term emission rates are needed to screen for the potential hazard
from acute exposures, but the emissions data typically contain only annual emission totals, for
RTR assessments we generally use the assumption that the maximum one-hour emission rate
from each source is ten times the average annual hourly emission rate for that source.
Sterilization operations are batch in nature in that individual chambers are charged with EtO,
then vented to a control device after sufficient time to sterilize products in the chamber. This
batch nature likely leads to some variability in emissions, although with multiple chambers
operating simultaneously and at different stages of the sterilization process, we would not
expect as much variability as for a truly batch operation. Emissions from aeration room vents
and fugitive emissions would not be as variable as those from the chamber. Given these
process characteristics, and without process-specific data on hourly emissions variations, we
conclude that the short-term emissions factor of ten should be sufficient to estimate hourly
emissions. Further discussion of the acute risk assessment can be found in Section 2.4.
2.3 Estimating chronic human inhalation exposure
We considered two chronic human inhalation exposure scenarios: residential and non-
residential. For the residential scenario, we use the estimated 5-year average ambient air
concentration at each census block centroid as a surrogate for the lifetime inhalation exposure
concentration of all the people who reside in the census block. We also use the grid and polar
receptors for lifetime inhalation exposure concentration if they fall in residential areas. The
residential exposure scenario does not consider either the short-term or long-term behavior
(mobility) of the exposed populations and its potential influence on their exposure. For
example, we do not reduce exposure durations to reflect that people leave their home census
blocks to go to work or school in other blocks. We do not consider that indoor concentrations
(of pollutants emitted from outdoor sources) may be higher or lower than outdoor ambient
concentrations. However, for gaseous pollutants like EtO, we have no reason to conclude
there would be significant differences between indoor and outdoor concentrations caused by
outdoor sources.
We do not address long-term migration or population growth or decrease over the 70-year
exposure period. Instead, we assume that each person's predicted exposure is constant over
the course of their lifetime, which is assumed to be 70 years. The assumption of not
15
-------
considering short- or long-term population mobility does not bias the estimate of the
theoretical MIR (assumes a person stays in one location for 70 years) nor does it affect the
estimate of cancer incidence since the total population number remains the same. It does,
however, affect the shape of the distribution of individual risks across the affected population,
shifting it toward higher estimated individual risks at the upper end and reducing the number
of people estimated to be at lower risks, thereby increasing the estimated number of people at
higher risk levels.
For the non-residential scenario, we consider all receptors, and we apply an exposure factor to
the estimated 5-year average ambient air concentrations to reflect less than lifetime exposure.
This scenario is based on an offsite worker as described by ATSDR, which assumes an
8.5-hour workday, 250 days a year, for 25 years (ATSDR, 2016). We use an exposure factor
that is slightly different from that used by ATSDR in that the 25-year working time is
compared to the EPA's 70-year lifetime assumption rather than ATSDR's 78-year lifetime,
resulting in an exposure factor of 0.087. Workers at the Sterigenics facility would be covered
under the Occupational Safety and Health Administration (OSHA) EtO standard (29 CFR
1910.1047).
2.4 Acute risk screening and refined assessments
In establishing a scientifically defensible approach for the assessment of potential health risks
due to acute exposures to HAP, we follow a similar approach to that for chronic health risk
assessments under the residual risk program, in that we begin with a screening assessment and
then, if appropriate, perform a refined assessment.
The approach for the acute health risk screening assessment is designed to eliminate from
further consideration those facilities for which we have confidence that no acute adverse
health effects of concern will occur. For this screening assessment, we use available data and
conservative assumptions for emission rates, meteorology, and exposure location that, in
combination, approximate a worst-case exposure.
The following are the steps we take and assumptions we make in the acute screening
assessment:
• When available, we use peak 1-hour emission data obtained from data collection
efforts or estimated based on the operating characteristics and engineering judgement
of facility emission sources; otherwise, we use a default emission adjustment factor of
10.
• We assume that the peak emissions occur at all emission points at the same time.
• For facilities with multiple emission points, 1-hour concentrations at each receptor are
assumed to be the sum of the maximum concentrations due to each emission point,
regardless of whether those maximum concentrations occurred during the same hour.
• Worst-case meteorology (from five years of local meteorology) is assumed to occur at
the same time the peak emission rates occur. The recommended EPA local-scale
dispersion model, AERMOD, is used for simulating atmospheric dispersion.
• A person is assumed to be located downwind at the point of maximum modeled
impact during this same worst-case 1-hour period.
16
-------
As a result of this screening assessment, the maximum pollutant concentration is compared to
multiple acute dose-response values for the HAP being assessed to determine whether a
possible acute health risk might exist. The acute dose-response values are described in
section 2.5.2 of this report.
A facility will either be found to pose no potential acute health risks (i.e., it will "screen out")
or will need to undergo a more refined assessment. When we identify levels of a HAP that
exceed its acute health benchmarks, we perform a more refined assessment, if possible.
Situations in which we have used engineering judgement to estimate emissions, a refinement
may be to obtain facility-specific data on HAP emissions. Other refinements may include the
temporal pattern of emissions (number of working hours, batch vs continuous operation), the
location of emission points, the boundaries of the facility, and/or the local meteorology. In
some cases, all of these site-specific data are used to refine the assessment; in others, lesser
amounts of site-specific data may be used to determine that acute exposures are not a concern,
and significant additional data collection is not necessary. For the Sterigenics facility,
modeled concentrations of EtO are well below the available acute health benchmarks, so we
did not perform any refinement of the acute assessment.
2.5 Dose-response assessment
2.5.1 Sources of chronic dose-response information
Dose-response assessments (carcinogenic and non-carcinogenic) for chronic exposure (either
by inhalation or ingestion) for the HAP reported in the emissions inventory for this source
category are based on the EPA Office of Air Quality Planning and Standards' (OAQPS)
existing recommendations for HAP (USEPA, 2018d). This information has been obtained
from various sources and prioritized according to (1) conceptual consistency with EPA risk
assessment guidelines and (2) level of peer review received. The prioritization process was
aimed at incorporating into our assessments the best available science with respect to dose-
response information. The recommendations are based on the following sources, in order of
priority:
1) U.S. Environmental Protection Agency (EPA). The EPA has developed dose-
response assessments for chronic exposure for many HAP. These assessments
typically provide a qualitative statement regarding the strength of scientific data and
specify a reference concentration (RfC, for inhalation) or reference dose (RfD, for
ingestion) to protect against effects other than cancer and/or a unit risk estimate (URE,
for inhalation) or slope factor (SF, for ingestion) to estimate the probability of
developing cancer. The RfC is defined as an "estimate (with uncertainty spanning
perhaps an order of magnitude) of a continuous inhalation exposure to the human
population (including sensitive subgroups) that is likely to be without an appreciable
risk of deleterious effects during a lifetime." The RfD is "an estimate (with
uncertainty spanning perhaps an order of magnitude) of a daily oral exposure to the
human population (including sensitive subgroups) that is likely to be without an
appreciable risk of deleterious effects during a lifetime." The URE is defined as "the
17
-------
upper-bound excess cancer risk6 estimated to result from continuous lifetime exposure
to an agent at a concentration of 1 |ig/m3 in air." The SF is "an upper bound,
approximating a 95 percent confidence limit, on the increased cancer risk from a
lifetime exposure to an agent. This estimate, [is] usually expressed in units of
proportion (of a population) affected per mg/kg-day.
The EPA disseminates dose-response assessment information in several forms, based
on the level of review. The Integrated Risk Information System (IRIS) is an EPA
database that contains scientific health assessment information, including dose-
response information. All IRIS assessments since 1996 have also undergone
independent external peer review. The current IRIS process includes review by EPA
scientists, interagency reviewers from other federal agencies, and the public, as well as
peer review by independent scientists external to the EPA. New IRIS values are
developed and old IRIS values are updated as new health effects data become
available. Refer to the IRIS Agenda for detailed information on status and scheduling
of current individual IRIS assessments and updates. The EPA's science policy
approach, under the current carcinogen guidelines, is to use linear low-dose
extrapolation as a default option for carcinogens for which the mode of action (MOA)
has not been identified. We expect future EPA dose-response assessments to identify
nonlinear MO As where appropriate, and we will use those analyses (once they are
peer reviewed) in our risk assessments. At this time, however, there are no available
carcinogen dose-response assessments for inhalation exposure that are based on a
nonlinear MOA.
2) U.S. Agency for Toxic Substances and Disease Registry (ATSDR). ATSDR, which
is part of the US Department of Health and Human Services, develops and publishes
Minimal Risk Levels (MRLs) for inhalation and oral exposure to many toxic
substances. As stated on the ATSDR web site: "Following discussions with scientists
within the Department of Health and Human Services (HHS) and the EPA, ATSDR
chose to adopt a practice similar to that of the EPA's Reference Dose (RfD) and
Reference Concentration (RfC) for deriving substance specific health guidance levels
for non-neoplastic endpoints." The MRL is defined as "an estimate of daily human
exposure to a substance that is likely to be without an appreciable risk of adverse
effects (other than cancer) over a specified duration of exposure." ATSDR describes
MRLs as substance-specific estimates to be used by health assessors to select
environmental contaminants for further evaluation.
3) California Environmental Protection Agency (CalEPA). The CalEPA Office of
Environmental Health Hazard Assessment has developed dose-response assessments
for many substances, based both on carcinogenicity and health effects other than
cancer. The process for developing these assessments is similar to that used by the
EPA to develop IRIS values and incorporates significant external scientific peer
review. As stated in the CalEPA Technical Support Document for developing their
6 Upper-bound lifetime cancer risk is a likely upper limit to the true probability that a person will contract cancer
over a 70-year lifetime due to a given hazard (such as exposure to a toxic chemical). This risk can be measured
or estimated in numerical terms (for example, one chance in a hundred).
18
-------
chronic assessments (CalEPA, 2008), the guidelines for developing chronic inhalation
exposure levels incorporate many recommendations of the U.S. EPA (USEPA, 1994)
and NAS (NAS, 1994). The noncancer information includes available inhalation
health risk guidance values expressed as chronic inhalation reference exposure levels
(RELs). CalEPA defines the REL as "the concentration level at or below which no
health effects are anticipated in the general human population." CalEPA's quantitative
dose-response information on carcinogenicity by inhalation exposure (CalEPA, 2009)
is expressed in terms of the URE, defined similarly to the EPA's URE. The EPA may
also look to other state dose-response assessments as appropriate.
For certain HAP, to address data gaps, increase accuracy, and avoid underestimating risk, we
make additional changes to some of the chronic inhalation exposure values to take into
account their mutagenic mode of action. For carcinogenic chemicals acting via a mutagenic
mode of action (i.e., chemicals that cause cancer by damaging genes), we estimate risks to
reflect the increased carcinogenicity of such chemicals during childhood. This approach is
explained in detail in the Supplemental Guidance for Assessing Susceptibility from Early-Life
Exposure to Carcinogens (USEPA. 2005a). Where available data do not support a chemical-
specific evaluation of differences between adults and children, the Supplemental Guidance
recommends using the following default adjustment factors for early-life exposures: increase
the carcinogenic potency by 10-fold for children up to 2 years old and by 3-fold for children 2
to 15 years old. These adjustments have the aggregate effects of increasing by about 60
percent the estimated risk (a 1.6-fold increase) for a lifetime of constant inhalation exposure.
The EPA uses these default adjustments only for carcinogens known to be mutagenic for
which data to evaluate adult and juvenile differences in toxicity are not available.
In December 2016, the EPA finalized its Evaluation of the Inhalation Carcinogenicity of
Ethylene Oxide (USEPA, 2016) in IRIS, which addresses the potential carcinogenicity from
long-term inhalation exposure to EtO. The EPA characterizes EtO as "carcinogenic to
humans" by the inhalation route of exposure based on the total weight of evidence, in
accordance with the EPA's 2005 Guidelines for Carcinogen Risk Assessment (Cancer
Guidelines) (U.S. EPA, 2005b). The lines of evidence supporting this characterization
include: (1) strong, but less than conclusive on its own, epidemiological evidence of
lymphohematopoietic cancers and breast cancer in EtO-exposed workers, (2) extensive
evidence of carcinogenicity in laboratory animals, including lymphohematopoietic cancers in
rats and mice and mammary carcinomas in mice following inhalation exposure, (3) clear
evidence that EtO is genotoxic and sufficient weight of evidence to support a mutagenic mode
of action for EtO carcinogenicity, and (4) strong evidence that the key precursor events are
anticipated to occur in humans and progress to tumors, including evidence of chromosome
damage in humans exposed to EtO. Overall, confidence in the hazard characterization of EtO
as "carcinogenic to humans" is high.
In this risk assessment, to estimate lifetime cancer risk from residential exposures we used the
IRIS full lifetime cancer unit risk estimate for EtO of 0.005 per (J,g/m3, which includes age-
dependent adjustment factors to account for early-life susceptibility. For non-residential
exposures, we used the IRIS unit risk estimate (0.003 per (j,g/m3) without age-dependent
adjustment factors because those are not relevant for an adult offsite worker. For noncancer
19
-------
effects, EtO has not been assessed under the IRIS program, nor does ATSDR have a chronic
MRL for EtO. Therefore, in this assessment we used the CalEPA chronic REL for EtO, which
is 0.03 mg/m3. In recent and forthcoming rulemakings, the EPA seeks public comment on the
use of certain hazard identification and dose-response information for key source categories.
2.5.2 Sources of acute dose-response information
Hazard identification and dose-response assessment information for acute inhalation exposure
assessments is based on the existing recommendations of OAQPS for HAP (USEPA, 2018e).
When the benchmarks are available, the results from acute screening assessments are
compared to both "no effects" reference levels for the general public, such as the California
Reference Exposure Levels (RELs), and to emergency response levels, such as Acute
Exposure Guideline Levels (AEGLs) and Emergency Response Planning Guidelines
(ERPGs), with the recognition that the ultimate interpretation of any potential risks associated
with an estimated exceedance of a particular reference level depends on the definition of that
level and any limitations expressed therein. Comparisons among different available inhalation
health effect reference values (both acute and chronic) for selected HAP can be found in an
EPA document of graphical arrays (USEPA, 2009b).
California Acute Reference Exposure Levels (RELs). CalEPA has developed acute dose-
response reference values for many substances, expressing the results as acute inhalation
RELs. The acute REL is defined by CalEPA (CalEPA, 2016) as "the concentration level at or
below which no adverse health effects are anticipated for a specified exposure duration. RELs
are based on the most sensitive, relevant, adverse health effect reported in the medical and
toxicological literature. RELs are designed to protect the most sensitive individuals in the
population by the inclusion of margins of safety. Since margins of safety are incorporated to
address data gaps and uncertainties, exceeding the REL does not automatically indicate an
adverse health impact." Acute RELs are developed for 1-hour (and 8-hour) exposures. The
values incorporate uncertainty factors similar to those used in deriving the EPA's inhalation
RfCs for chronic exposures.
Acute Exposure Guideline Levels (AEGLs). AEGLs are developed by the National
Advisory Committee on Acute Exposure Guideline Levels (NAC/AEGL) for Hazardous
Substances and then reviewed and published by the National Research Council. As described
in the Committee's Standing Operating Procedures (NAS, 2001), AEGLs "represent threshold
exposure limits for the general public and are applicable to emergency exposures ranging
from 10 min to 8 h " Their intended application is "for conducting risk assessments to aid in
the development of emergency preparedness and prevention plans, as well as real time
emergency response actions, for accidental chemical releases at fixed facilities and from
transport carriers." The document states that "the primary purpose of the AEGL program and
the NAC/AEGL Committee is to develop guideline levels for once-in-a-lifetime, short-term
exposures to airborne concentrations of acutely toxic, high-priority chemicals." In detailing
the intended application of AEGL values, the document states, "It is anticipated that the
AEGL values will be used for regulatory and nonregulatory purposes by U.S. Federal and
State agencies, and possibly the international community in conjunction with chemical
emergency response, planning, and prevention programs. More specifically, the AEGL values
will be used for conducting various risk assessments to aid in the development of emergency
20
-------
preparedness and prevention plans, as well as real-time emergency response actions, for
accidental chemical releases at fixed facilities and from transport carriers."
The NAC/AEGL defines AEGL-1 and AEGL-2 as:
"AEGL-1 is the airborne concentration (expressed as ppm or mg/m3) of a substance above
which it is predicted that the general population, including susceptible individuals, could
experience notable discomfort, irritation, or certain asymptomatic nonsensory effects.
However, the effects are not disabling and are transient and reversible upon cessation of
exposure."
"AEGL-2 is the airborne concentration (expressed as ppm or mg/m3) of a substance above
which it is predicted that the general population, including susceptible individuals, could
experience irreversible or other serious, long-lasting adverse health effects or an impaired
ability to escape."
"Airborne concentrations above AEGL-1 represent exposure levels that can produce mild
and progressively increasing but transient and nondisabling odor, taste, and sensory
irritation or certain asymptomatic, nonsensory effects. With increasing airborne
concentrations above each AEGL, there is a progressive increase in the likelihood of
occurrence and the severity of effects described for each corresponding AEGL. Although
the AEGL values represent threshold levels for the general public, including susceptible
subpopulations, such as infants, children, the elderly, persons with asthma, and those with
other illnesses, it is recognized that individuals, subject to unique or idiosyncratic
responses, could experience the effects described at concentrations below the
corresponding AEGL."
Emergency Response Planning Guidelines (ERPGs). The American Industrial Hygiene
Association (AIHA) has developed ERPGs for acute exposures at three different levels of
severity. These guidelines represent concentrations for exposure of the general population
(but not particularly sensitive persons) for up to 1 hour associated with effects expected to be
mild or transient (ERPG-1), irreversible or serious (ERPG-2), and potentially life-threatening
(ERPG-3).
ERPG values are described in their supporting documentation as follows: "ERPGs are air
concentration guidelines for single exposures to agents and are intended for use as tools to
assess the adequacy of accident prevention and emergency response plans, including
transportation emergency planning, community emergency response plans, and incident
prevention and mitigation."
ERPG-1 and ERPG-2 values are defined by AIHA's Standard Operating Procedures (AIHA,
2018) as follows:
"ERPG-1 is the maximum airborne concentration below which nearly all individuals
could be exposed for up to 1 hour without experiencing more than mild, transient health
effects or without perceiving a clearly defined objectionable odor."
21
-------
"ERPG-2 is the maximum airborne concentration below which nearly all individuals
could be exposed for up to 1 hour without experiencing or developing irreversible or other
serious adverse health effects or symptoms that could impair an individual's ability to take
protective action."
There is no California acute REL available for EtO, nor is there an AEGL-1 or ERPG-1 for
EtO. Values for AEGL-1 were not derived because concentrations causing mild sensory
irritation are above the AEGL-2 values and would not serve as a warning of potential
exposure (NAS, 2010). In this risk assessment, we used the 1-hour AEGL-2 value of
81 mg/m3.
2.6 Risk characterization
The final product of the risk assessment is the risk characterization, in which the information
from the previous steps is integrated and an overall conclusion about risk is synthesized that is
complete, informative, and useful for decision makers. In general, the nature of this risk
characterization depends on the information available, the application of the risk information
and the resources available. In all cases, major issues associated with determining the nature
and extent of the risk are identified and discussed. Further, it is the EPA's policy that a risk
characterization be prepared in a manner that is clear, transparent, reasonable, and consistent
with other risk characterizations of similar scope prepared across programs in the Agency.
These principles of transparency and consistency have been reinforced by the Agency's Risk
Characterization Handbook (USEPA, 2000a), in the Agency's information quality guidelines
(USEPA, 2002a), and in the Office of Management and Budget (OMB) Memorandum on
Updated Principles for Risk Analysis (OMB, 2007), and they are incorporated in this
assessment.
Estimates of health risk are presented in the context of uncertainties and limitations in the data
and methodology. We have attempted to reduce both uncertainty and bias to the greatest
degree possible in this assessment. We provide summaries of risk metrics (including
maximum individual cancer risks and noncancer hazards, as well as cancer incidence
estimates) along with a discussion of the major uncertainties associated with their derivation.
For each carcinogenic HAP included in an assessment for which a potency estimate is
available, individual and population cancer risks are calculated by multiplying the
corresponding lifetime average exposure estimate by the appropriate URE. This calculated
cancer risk is defined as the upper-bound probability of developing cancer over a 70-year
period (i.e., the assumed human lifespan) at that exposure. Because UREs for most HAP are
upper-bound estimates, actual risks at a given exposure level may be lower than predicted.
Increased cancer incidence for the entire population within the area of analysis is estimated by
multiplying the estimated lifetime cancer risk for each census block by the number of people
residing in that block, then summing the results for the entire modeled domain. This lifetime
population incidence estimate is divided by 70 years to obtain an estimate of the number of
cancer cases per year. We did not estimate cancer incidence for the non-residential scenario
22
-------
because we do not have data on where or how many people would be at specific locations, nor
how long they would be there. Also, calculating incidence in such cases could double count
cases because the same people likely live in a nearby census block for which we are
calculating incidence under the residential scenario.
Unlike linear dose-response assessments for cancer, noncancer health hazards generally are
not expressed as a probability of an adverse occurrence. Instead, the estimated human health
risk for noncancer effects is expressed by comparing an exposure to a reference level as a
ratio. The hazard quotient (HQ) is the estimated exposure divided by a reference level (e.g.,
the RfC). For a given HAP, exposures at or below the reference level (HQ < 1) are not likely
to cause adverse health effects. As exposures increase above the reference level (HQs
increasingly greater than 1), the potential for adverse effects increases. For exposures
predicted to be above the RfC, the risk characterization includes the degree of confidence
ascribed to the RfC values for the compound(s) of concern (i.e., high, medium, or low
confidence) and discusses the impact of this on possible health interpretations.
The risk characterization for chronic effects other than cancer is developed using the HQ for
inhalation, calculated for each HAP at each census block centroid. As discussed above, RfCs
incorporate generally conservative uncertainty factors in the face of uncertain extrapolations,
such that an HQ greater than 1 does not necessarily suggest the onset of adverse effects. The
target-organ-specific hazard index (TOSHI) is the sum of hazard quotients for substances that
affect the same target organ or organ system and approximates the aggregate effect on a
specific target organ (e.g., the lungs). The HQ and TOSHI cannot be translated to a
probability that adverse effects will occur, and it is unlikely to be proportional to adverse
health effect outcomes in a population.
Screening for potentially significant acute inhalation exposures also follows the HQ approach.
We divide the maximum estimated acute exposure by each available acute dose-response
value to develop an array of HQs. In general, when none of these HQs is greater than one,
there is no potential for acute risk. When one or more HQ is above 1, we evaluate additional
information (e.g., proximity of the facility to potential exposure locations) to determine
whether there is a potential for significant acute risks.
3 Risk results for the Sterigenics facility in Willowbrook, IL
This section presents the results of the risk assessment for the Sterigenics facility in
Willowbrook, Illinois based on the modeling methods described in the previous sections. All
baseline risk results were developed using the best estimates of actual EtO emissions before
the Seal Order issued in February 2019 by the state of Illinois. The basic chronic inhalation
risk estimates presented here are the maximum individual lifetime cancer risk, the maximum
chronic hazard index, and the cancer incidence. We also present results from our acute
inhalation screening assessment in the form of maximum hazard quotients. This section also
presents the risk results for the illustrative future scenario.
23
-------
3.1 Risk assessment results for baseline emissions
Table 3.1-1 summarizes the chronic and acute inhalation risk results for this facility based on
baseline emissions. The results of the chronic inhalation cancer risk assessment indicate that
the maximum lifetime (residential) individual cancer risk posed by the facility is 1,000-in-l
million. The total estimated cancer incidence is 0.3 excess cancer cases per year, or one
excess case in every 3 years within the entire modeling domain. Over 70 years, the estimated
number of cancer cases is approximately 20. Estimated maximum lifetime individual cancer
risks of 100-in-l million extend out to about 2 km (1.4 mi) from the facility, cancer risks of
50-in-l million extend out to about 4 km (2.7 mi) from the facility, cancer risks of 10-in-l
million extend out to about 9 km (6 mi) from the facility, and cancer risks of 1 -in-1 million
extend out to about 40 km (25 mi) from the facility. Approximately 60 people are estimated to
have cancer risks equal to 1,000-in-l million, 11,500 people are estimated to have cancer risks
greater than or equal to 100-in-l million, 230,000 people are estimated to have cancer risks
greater than or equal to 10-in-l million, and 6.5 million people are estimated to have cancer
risks greater than or equal to 1 —in-1 million.
The maximum cancer risk from non-residential exposures is also 1,000-in-l million, but it is
only coincidence that this estimate matches the lifetime residential risk estimate. The
residential and non-residential risk estimates are based on different exposure concentrations
and different cancer unit risk estimates. Estimated maximum non-residential cancer risks of
100-in-l million extend out to about 400 m (400 yds) from the facility, cancer risks of 50-in-l
million extend out to about 600 m (700 yds) from the facility, cancer risks of 10-in-l million
extend out to about 2 km (1 mi) from the facility, and cancer risks of 1-in-l million extend out
to about 7 km (5 mi) from the facility.
Table 3.1-1. Inhalation Risks for the Sterigenics Willowbrook, Illinois Facility -
Baseline Emissions
Result
Residential
Non-Residential
Cancer Risks
Maximum Individual Lifetime Cancer Risk (in 1
1,000
1,000
million)
Chronic Noncancer Risks
Maximum Neurological Hazard Index
0.01
0.01
Acute Noncancer Screening Results
Maximum Acute Hazard Quotient
0.02
0.02
Population Exposure
Number of People Living Within 50 km of Facility
7,700,000
n/a
Number of People Exposed to Cancer Risk:
Greater than or equal to 1,000-in-l million
60
n/a
Greater than or equal to 100-in-l million
11,500
n/a
Greater than or equal to 1-in-l million
6,500,000
n/a
Estimated Cancer Incidence
0.3
n/a
(excess cancer cases per year)
Estimated number of years for 1 cancer case
3
n/a
Estimated number of cancer cases over 70 years
20
n/a
24
-------
The maximum chronic noncancer hazard index is 0.01 (neurological) for both residential and
non-residential exposures, and no one is exposed to TOSHI levels above 1. Worst-case acute
HQs were calculated and are as shown in Table 3.1-1. The highest screening acute HQ was
0.02 (based on the 1-hr AEGL-2 value for EtO). Acute exposures are estimated at all
receptors (residential and non-residential) assuming someone could be at the receptor location
for an hour, so no distinction is made between residential and non-residential acute exposures.
Since the screening HQ was not greater than 1, further refinement of the estimate was not
warranted.
Figure 3.1-1 shows the estimated lifetime cancer risk contours near the facility. The figure also
shows the commercial/industrial (non-residential) areas adjacent to the facility. The risk
contours are not applicable in the non-residential areas because lifetime exposures are relevant
only for residential locations. Figure 3.1-2 shows the estimated cancer risk contours for the
non-residential scenario. These estimates are based on an offsite worker who is exposed
8.5 hours per day, 250 days per year, for 25 years. Similar maps were presented at a public
meeting in Willowbrook, Illinois on May 29, 2019, and are provided in Appendix 4. The risk
contours in the maps in Appendix 4 are slightly different than those in Figures 3.1-1 and 3.1-2
because they do not reflect limiting the displayed values to one significant digit. For example,
the risk contour in Figure3.1-1 for the 100- to 200-in-l million range displays data from 95-to
249-in-l million, whereas the corresponding risk contour in the Appendix 4 map displays data
strictly between 100- and 200-in-l million.
3.2 Risk assessment results for the illustrative future scenario
In addition to assessing the baseline scenario, we also assessed an illustrative future scenario,
where all emission sources at the facility are routed to a control device, and the post-control
emissions (26 lb/yr) are released from a single 26.5 m (87 ft) stack. The maximum lifetime
(residential) individual cancer risk under this scenario is 1-in-l million, which occurs at a
single residential grid receptor. All cancer risks at census blocks are less than 1-in-l million.
The total estimated cancer incidence is 0.002 excess cancer cases per year, or one excess
case in every 700 years within the entire modeling domain. Over 70 years, the estimated
number of cancer cases is less than 1 (0.1). Approximately 70,000 people are estimated to
have cancer risks between 0.1- and 1-in-l million, so the remaining 7.6 million people
within the modeling domain have estimated cancer risk less than 0.1-in-l million. The
maximum chronic noncancer hazard index is 6E-6 (neurological). For non-residential
exposures, the maximum cancer risk is 0.08-in-l million, and the maximum chronic
noncancer hazard index is 9E-7 (neurological). The highest screening acute HQ was 4E-6
(based on the 1-hr AEGL-2 value for EtO).
As discussed in Section 2.1, the emissions and release parameters modeled for the future
scenario are similar but not identical to those data in the actual permit application for the
Willowbrook facility. The emissions in the permit application are approximately three times
higher than the emissions modeled for this assessment, so the calculated risks for these higher
future emissions would be greater than those modeled in this assessment but are still in the
range of 1- to-10-in-l million.
25
-------
59th-St
ChestnutLnr
Harvester-^
[H7g£Sch£^
W/illowbrookj
Figure 3.1 -1. Modeled Lifetime Cancer Risks for Sterigenics, Willowbrook, IL
Lifetime
Ethylene Oxide
Cancer Risk
(in 1 million)
500 - 1000
Sterigenics
Facility
l\\1 Non-Residential Areas
0 200 400 600 800 1,000 Meter:
26
-------
Figure 3.1 - 2. Modeled Non-Residential Cancer Risks for Sterigenics, Willowbrook, IL
Executive Dr
Midway-Dn
tWi llowproAkl
Non-Residential
Ethylene Oxide
Cancer Risk
(in 1 million)
H 100 - 200
200 - 500
H 500 - 1000
I—| Sterigenics
Facility
50 100 150 200 250 Meters
27
-------
4 General discussion of uncertainties in the risk assessment
The uncertainties in virtually all risk assessments can be divided into three areas: 1)
uncertainties in the emission data sets, 2) exposure modeling uncertainties, and 3)
uncertainties in the dose-response relationships. Uncertainties in the emission estimates and in
the air quality models lead to uncertainty in air concentrations. Uncertainty in exposure
modeling can arise due to uncertain activity patterns, the locations of individuals within a
census block, and the microenvironmental concentrations as reflected in the exposure model.
Finally, uncertainty in the shape of the relationship between exposure and effects, the URE
and the RfC, also contributes to uncertainties in the risk assessment. These three areas of
uncertainty are discussed below.
4.1 Emissions inventory uncertainties
Appendix 1 of this document describes how we developed EtO emission estimates for the
Sterigenics facility, starting with information provided to us by Sterigenics regarding their
operations and estimated emissions rates and operational parameters for both the controlled
and uncontrolled sources. We took this information and derived site-specific emission factors
from previous stack testing results for the "controlled" sources and estimated site-specific
emission factors for the uncontrolled or "fugitive" emissions. Emission factors are calculated
values that relate the quantity of a pollutant released to the atmosphere with an activity
associated with the release of that pollutant and are generally assumed to be representative of
long-term averages. Using dispersion modeling, we evaluated the accuracy of these site-
specific emission factors and made adjustments to these factors so that the modeled results
would better correspond with the ambient air values measured at the monitoring sites near the
facility. Since the estimated emissions are representative of long-term averages, they do not
reflect short-term fluctuations during the course of a year or variations from year to year.
For the acute effects screening assessment, in the absence of available specific estimates or
measurements we use estimates of peak hourly emission rates. These estimates typically are
calculated by first estimating the average annual hourly emissions rates by evenly dividing the
total annual emission rate into the 8,760 hours of the year. An emission adjustment factor that
is intended to account for emission fluctuations during normal facility operations is then
applied to these average annual hourly emission rates. The adjustment factor can be based on
actual fluctuations seen in the available emission data or on engineering judgment; in the
absence of such information a default factor is applied, as was done for this assessment.
4.2 Exposure modeling uncertainties
We did not include the effects of human mobility on exposures in the assessment.
Specifically, short-term mobility and long-term mobility between census blocks in the
modeling domain were not considered. (Short-term mobility is movement from one micro-
environment to another over the course of hours or days. Long-term mobility is movement
from one residence to another over the course of a lifetime.) The approach of not considering
short or long-term population mobility does not bias the estimate of the theoretical MIR (by
definition), nor does it affect the estimate of cancer incidence because the total population
number remains the same. It does, however, affect the shape of the distribution of individual
risks across the affected population, shifting it toward higher estimated individual risks at the
28
-------
upper end and reducing the number of people estimated to be at lower risks, thereby
increasing the estimated number of people at specific high-risk levels (e.g., l-in-10 thousand
or 1-in-l million).
We also do not factor in the possibility of a source closure occurring during the 70-year
chronic exposure period, leading to a potential upward bias in both the MIR and population
risk estimates. Nor do we factor in the possibility of population growth during the 70-year
chronic exposure period, which could lead to a potential downward bias in both the MIR and
population risk estimates. Finally, we do not factor in time an individual spends indoors. The
exposure estimates used in these analyses assume chronic exposures to ambient (outdoor)
levels of pollutants. Because people spend most of their time indoors, actual exposures may
not be as high, depending on the characteristics of the pollutants modeled. For many HAP,
indoor levels are roughly equivalent to ambient levels, but for very reactive pollutants or
larger particles, indoor levels are typically lower. This factor has the potential to result in an
overestimate of 25 to 30 percent of exposures (USEPA, 2001).
We estimated the chronic exposures at the centroid of each populated census block as
surrogates for the exposure concentrations for all people living in that block. Using the census
block centroid to predict chronic exposures tends to over-predict exposures for people in the
census block who live farther from the facility and under-predict exposures for people in the
census block who live closer to the facility. Thus, using the census block centroid to predict
chronic exposures may lead to a potential understatement or overstatement of the true
maximum impact, but is an unbiased estimate of average risk and incidence. We reduce this
uncertainty by analyzing large census blocks near facilities using aerial imagery and adjusting
the location of the block centroid to better represent the population in the block, as well as
adding additional receptor locations where the block population is not well represented by a
single location. In this assessment, we used many additional receptors which cover the areas
near the facility, so we likely have not missed the location of maximum exposure.
The assessment evaluates the cancer inhalation risks associated with pollutant exposures over
a 70-year period, which is the assumed lifetime of an individual. In reality, both the length of
time that modeled emission sources at facilities actually operate (i.e., more or less than 70
years) and the domestic growth or decline of the modeled industry (i.e., the increase or
decrease in the number or size of domestic facilities) will influence the future risks posed by a
given source or source category. Depending on the characteristics of the industry, these
factors will, in most cases, result in an overestimate both in individual risk levels and in the
total estimated number of cancer cases. However, in the unlikely scenario where a facility
maintains, or even increases, its emissions levels over a period of more than 70 years,
residents live beyond 70 years at the same location, and the residents spend more of their days
at that location, then the cancer inhalation risks could potentially be underestimated.
However, annual cancer incidence estimates from exposures to emissions from these sources
would not be affected by the length of time an emissions source operates.
For the acute screening assessment, the results are intentionally biased high, and thus health-
protective, by assuming the co-occurrence of independent factors, such as hourly emission
rates, meteorology and human activity patterns. Furthermore, in cases where multiple acute
29
-------
dose-response values for a pollutant are considered scientifically acceptable, we choose the
most conservative of these dose-response values, erring on the side of overestimating
potential health risks from acute exposures. In cases where these results indicate the potential
for exceeding acute HQs, we refine our assessment by developing a better understanding of
the geography of the facility relative to potential exposure locations.
Appendix 3 of this document includes the analyses performed to support the use of
meteorological data from the Argonne National Laboratory, but there are always uncertainties
regarding the spatial and temporal representativeness of any meteorological data. Section
8.4.1 of The Guideline on Air Quality Models states that the meteorological data should be
adequately representative of the modeling domain, including proximity of the meteorological
station to the source, terrain complexity, exposure of the meteorological tower, and period of
time the data were collected relative to the modeled period. While there can be uncertainties
in the meteorological data for the modeling domain, such as potential wind direction changes
across the domain or surface characteristics of the source versus the meteorological site, these
uncertainties are mitigated by the choice of adequately representative meteorological data for
the model domain. For example, there will always be variations in winds across a domain
especially on an hourly basis, but for the long term the meteorological data selected for this
assessment are adequately representative of the model domain.
4.3 Uncertainties in the dose-response relationships
In the sections that follow, separate discussions are provided on uncertainty associated with
cancer potency factors and for noncancer reference values. Cancer potency values are derived
for chronic (lifetime) exposures. Noncancer dose-response values are generally derived for
chronic exposures (up to a lifetime) but may also be derived for acute (less than 24 hours),
short-term (from 24 hours up to 30 days), and subchronic (30 days up to 10 percent of
lifetime) exposure durations, all of which are derived based on an assumption of continuous
exposure throughout the duration specified. For the purposes of assessing all potential health
risks associated with the emissions included in an assessment, we rely on both chronic (cancer
and noncancer) and acute (noncancer) dose-response values, which are described in more
detail below.
Cancer assessment
The discussion of dose-response uncertainties in the estimation of cancer risk below focuses
on the uncertainties associated with the specific approach currently used by the EPA to
develop cancer potency factors. In general, these same uncertainties attend the development
of cancer potency factors by CalEPA, the source of peer-reviewed cancer potency factors
used where EPA-developed values are not yet available. According to the EPA's Cancer
Guidelines, "The primary goal of EPA actions is protection of human health; accordingly, as
an Agency policy, risk assessment procedures, including default options that are used in the
absence of scientific data to the contrary, should be health protective." The approach adopted
in this document is consistent with this approach as described in the Cancer Guidelines.
For cancer endpoints the EPA usually derives an oral slope factor for ingestion and a unit risk
value for inhalation exposures. These values allow estimation of a lifetime probability of
developing cancer given long-term exposures to the pollutant. Depending on the pollutant
30
-------
being evaluated, the EPA relies on both animal bioassay and epidemiological studies to
characterize cancer risk. As a science policy approach, consistent with the Cancer Guidelines,
the EPA uses animal cancer bioassays as indicators of potential human health risk when other
human cancer risk data are unavailable.
Extrapolation of study data to estimate potential risks to human populations is based upon the
EPA's assessment of the scientific database for a pollutant using EPA guidance documents
and other peer-reviewed methodologies. The EPA Cancer Guidelines describe the Agency's
recommendations for methodologies for cancer risk assessment. The EPA believes that cancer
risk estimates developed following the procedures described in the Cancer Guidelines and
outlined below generally provide an upper bound estimate of risk. That is, the EPA's upper
bound estimates represent a plausible upper limit to the true value of a quantity (although this
is usually not a true statistical confidence limit). In some circumstances, the true risk could be
as low as zero; however, in other circumstances the risk could also be greater.7 When
developing an upper bound estimate of risk and to provide risk values that do not
underestimate risk, the EPA generally relies on conservative default approaches.8 The EPA
also uses the upper bound (rather than lower bound or central tendency) estimates in its
assessments, although it is noted that this approach can have limitations for some uses (e.g.
priority setting, expected benefits analysis).
Such health risk assessments have associated uncertainties, some which may be considered
quantitatively, and others which generally are expressed qualitatively. Uncertainties may vary
substantially among cancer risk assessments associated with exposures to different pollutants,
since the assessments employ different databases with different strengths and limitations and
the procedures employed may differ in how well they represent actual biological processes for
the assessed substance. Some of the major sources of uncertainty and variability in deriving
cancer risk values are described more fully below.
(1) The qualitative similarities or differences between tumor responses observed in
experimental animal bioassays and those which would occur in humans are a source of
uncertainty in cancer risk assessments. In general, the EPA does not assume that tumor sites
observed in an experimental animal bioassay are necessarily predictive of the sites at which
7 The exception to this is the URE for benzene, which is considered to cover a range of values, each end of
which is considered to be equally plausible, and which is based on maximum likelihood estimates.
8 According to the NRC report Science and Judgment in Risk Assessment (NRC, 1994) "[Default] options are
generic approaches, based on general scientific knowledge and policy judgment, that are applied to various
elements of the risk-assessment process when the correct scientific model is unknown or uncertain." The 1983
NRC report Risk Assessment in the Federal Government: Managing the Process defined default option as "the
option chosen on the basis of risk assessment policy that appears to be the best choice in the absence of data to
the contrary" (NRC, 1983a, p. 63). Therefore, default options are not rules that bind the Agency; rather, the
Agency may depart from them in evaluating the risks posed by a specific substance when it believes this to be
appropriate. In keeping with EPA's goal of protecting public health and the environment, default assumptions
are used to ensure that risk to chemicals is not underestimated (although defaults are not intended to overtly
overestimate risk). See EPA 2004 An Examination of EPA Risk Assessment Principles and Practices.
EPA/100/B-04/001.
31
-------
tumors would occur in humans.9 However, unless scientific support is available to show
otherwise, the EPA assumes that tumors in animals are relevant in humans, regardless of
target organ concordance. For a specific pollutant, qualitative differences in species responses
can lead to either under-estimation or over-estimation of human cancer risks.
(2) Uncertainties regarding the most appropriate dose metric for an assessment can also lead
to differences in risk predictions. For example, the measure of dose is commonly expressed in
units of mg/kg/d ingested or the inhaled concentration of the pollutant. However, data may
support development of a pharmacokinetic model for the absorption, distribution, metabolism
and excretion of an agent, which may result in improved dose metrics (e.g., average blood
concentration of the pollutant or the quantity of agent metabolized in the body). Quantitative
uncertainties result when the appropriate choice of a dose metric is uncertain or when dose
metric estimates are themselves uncertain (e.g., as can occur when alternative
pharmacokinetic models are available for a compound). Uncertainty in dose estimates may
lead to either over or underestimation of risk.
(3) For the quantitative extrapolation of cancer risk estimates from experimental animals to
humans, the EPA uses scaling methodologies (relating expected response to differences in
physical size of the species), which introduce another source of uncertainty. These
methodologies are based on both biological data on differences in rates of process according
to species size and empirical comparisons of toxicity between experimental animals and
humans. For a particular pollutant, the quantitative difference in cancer potency between
experimental animals and humans may be either greater than or less than that estimated by
baseline scientific scaling predictions due to uncertainties associated with limitations in the
test data and the correctness of scaled estimates.
(4) EPA cancer risk estimates, whether based on epidemiological or experimental animal data,
are generally developed using a benchmark dose (BMD) analysis to estimate a dose at which
there is a specified excess risk of cancer, which is used as the point of departure (or POD) for
the remainder of the calculation. Statistical uncertainty in developing a POD using a
benchmark dose (BMD) approach is generally addressed though use of the 95 percent lower
confidence limit on the dose at which the specified excess risk occurs (the BMDL),
decreasing the likelihood of understating risk. The EPA has generally utilized the multistage
model for estimation of the BMDL using cancer bioassay data (see further discussion below).
(5) Extrapolation from high to low doses is an important source of uncertainty in cancer risk
assessment. The EPA uses different approaches to low dose risk assessment (i.e., developing
estimates of risk for exposures to environmental doses of an agent from observations in
experimental or epidemiological studies at higher dose) depending on the available data and
understanding of a pollutant's mode of action (i.e., the manner in which a pollutant causes
cancer). The EPA's Cancer Guidelines express a preference for the use of reliable, compound-
specific, biologically-based risk models when feasible; however, such models are rarely
available. The mode of action for a pollutant (i.e., the manner in which a pollutant causes
9 Per the EPA Cancer Guidelines: "The default option is that positive effects in animal cancer studies indicate
that the agent under study can have carcinogenic potential in humans." and "Target organ concordance is not a
prerequisite for evaluating the implications of animal study results for humans."
32
-------
cancer) is a key consideration in determining how risks should be estimated for low-dose
exposure. A reference value is calculated when the available mode of action data show the
response to be nonlinear (e.g., as in a threshold response). A linear low-dose (straight line
from POD) approach is used when available mode of action data support a linear (e.g.,
nonthreshold) response or as the most common default approach when a compound's mode of
action is unknown. Linear extrapolation can be supported by both pollutant-specific data and
broader scientific considerations. For example, the EPA's Cancer Guidelines generally
consider a linear dose-response to be appropriate for pollutants that interact with DNA and
induce mutations. Pollutants whose effects are additive to background biological processes in
cancer development can also be predicted to have low-dose linear responses, although the
slope of this relationship may not be the same as the slope estimated by the straight line
approach.
The EPA most frequently utilizes a linear low-dose extrapolation approach as a baseline
science-policy choice (a "default") when available data do not allow a compound-specific
determination. This approach is designed to not underestimate risk in the face of uncertainty
and variability. The EPA believes that linear dose-response models, when appropriately
applied as part of the EPA's cancer risk assessment process, provide an upper bound estimate
of risk and generally provide a health protective approach. Note that another source of
uncertainty is the characterization of low-dose nonlinear, non-threshold relationships. The
National Academy of Sciences (NAS, 1994) has encouraged the exploration of sigmoidal type
functions (e.g., log-probit models) in representing dose-response relationships due to the
variability in response within human populations. Another National Research Council report
(NRC, 2006) suggests that models based on distributions of individual thresholds are likely to
lead to sigmoidal-shaped dose-response functions for a population. This report notes sources
of variability in the human population: "One might expect these individual tolerances to vary
extensively in humans depending on genetics, coincident exposures, nutritional status, and
various other susceptibility factors..." Thus, if a distribution of thresholds approach is
considered for a carcinogen risk assessment, application would depend on ability of modeling
to reflect the degree of variability in response in human populations (as opposed to responses
in bioassays with genetically more uniform rodents). Note also that low dose linearity in risk
can arise for reasons separate from population variability: due to the nature of a mode of
action and additivity of a chemical's effect on top of background chemical exposures and
biological processes.
As noted above, the EPA's current approach to cancer risk assessment typically utilizes a
straight line approach from the BMDL. This is equivalent to using an upper confidence limit
on the slope of the straight line extrapolation. The impact of the choice of the BMDL on
bottom line risk estimates can be quantified by comparing risk estimates using the BMDL
value to central estimate BMD values, although these differences are generally not a large
contributor to uncertainty in risk assessment (Subramaniam et. al., 2006). It is important to
note that earlier EPA assessments, including the majority of those for which risk values exist
today, were generally developed using the multistage model to extrapolate down to
environmental dose levels and did not involve the use of a POD. Subramaniam et. al. (2006)
also provide comparisons indicating that slopes based on straight line extrapolation from a
33
-------
POD do not show large differences from those based on the upper confidence limit of the
multistage model.
(6) Cancer risk estimates do not generally make specific adjustments to reflect the variability
in response within the human population — resulting in another source of uncertainty in
assessments. In the diverse human population, some individuals are likely to be more
sensitive to the action of a carcinogen than the typical individual, although compound-specific
data to evaluate this variability are generally not available. There may also be important life
stage differences in the quantitative potency of carcinogens and, with the exception of the
recommendations in the EPA's Supplemental Guidance for carcinogens with a mutagenic
mode of action, risk assessments do not generally quantitatively address life stage differences.
However, one approach used commonly in EPA assessments that may help address variability
in response is to extrapolate human response from results observed in the most sensitive
species and sex tested, resulting typically in the highest URE which can be supported by
reliable data, thus supporting estimates that are designed not to underestimate risk in the face
of uncertainty and variability.
Chronic noncancer assessment
Chronic noncancer reference values represent chronic exposure levels that are intended to be
health-protective. That is, the EPA and other organizations, such as the ATSDR, which
develop noncancer dose-response values use an approach that is intended not to underestimate
risk in the face of uncertainty and variability. When there are gaps in the available
information, uncertainty factors (UFs) are applied to derive reference values that are intended
to be protective against appreciable risk of deleterious effects. Uncertainty factors are
commonly default values10 (e.g., factors of 10 or 3) used in the absence of compound-specific
data; where data are available, uncertainty factors may also be developed using compound-
specific information. When data are limited, more assumptions are needed and more default
factors are used. Thus, there may be a greater tendency to overestimate risk—in the sense that
further study might support development of reference values that are higher (i.e., less potent)
because fewer default assumptions are needed. However, for some pollutants it is possible
that risks may be underestimated.
For noncancer endpoints related to chronic exposures, the EPA derives a reference dose (RfD)
for exposures via ingestion, and a reference concentration (RfC) for inhalation exposures. As
stated in the IRIS Glossary, these values provide an estimate (with uncertainty spanning
10 According to the NRC report Science and Judgment in Risk Assessment (NRC, 1994) "[Default] options are
generic approaches, based on general scientific knowledge and policy judgment, that are applied to various
elements of the risk-assessment process when the correct scientific model is unknown or uncertain." The 1983
NRC report Risk Assessment in the Federal Government: Managing the Process defined default option as "the
option chosen on the basis of risk assessment policy that appears to be the best choice in the absence of data to
the contrary" (NRC, 1983a, p. 63). Therefore, default options are not rules that bind the Agency; rather, the
Agency may depart from them in evaluating the risks posed by a specific substance when it believes this to be
appropriate. In keeping with the EPA's goal of protecting public health and the environment, default
assumptions are used to ensure that risk to chemicals is not underestimated (although defaults are not intended to
overtly overestimate risk). See EPA 2004 An examination of EPA Risk Assessment Principles and Practices,
EPA/100/B-04/001.
34
-------
perhaps an order of magnitude) of daily oral exposure (RfD) or of a continuous inhalation
exposure (RfC) to the human population (including sensitive subgroups) that is likely to be
without an appreciable risk of deleterious effects during a lifetime. To derive values that are
intended to be "without appreciable risk," the EPA's methodology relies upon an uncertainty
factor (UF) approach (USEPA, 1994) which includes consideration of both uncertainty and
variability.
The EPA begins by evaluating all of the available peer-reviewed literature to determine
noncancer endpoints of concern, evaluating the quality, strengths and limitations of the
available studies. The EPA typically chooses the relevant endpoint that occurs at the lowest
dose, often using statistical modeling of the available data, and then determines the
appropriate POD for derivation of the reference value. A POD is determined by (in order of
preference): (1) a statistical estimation using the BMD approach; (2) use of the dose or
concentration at which the toxic response was not significantly elevated (no observed adverse
effect level— NOAEL); or (3) use of the lowest observed adverse effect level (LOAEL).
A series of downward adjustments using default UFs is then applied to the POD to estimate
the reference value (USEPA, 2002b). While collectively termed "UFs", these factors account
for a number of different quantitative considerations when utilizing observed animal (usually
rodent) or human toxicity data in a risk assessment. The UFs are intended to account for: (1)
variation in susceptibility among the members of the human population (i.e., inter-individual
variability); (2) uncertainty in extrapolating from experimental animal data to humans (i.e.,
interspecies differences); (3) uncertainty in extrapolating from data obtained in a study with
less-than-lifetime exposure (i.e., extrapolating from subchronic to chronic exposure);
(4) uncertainty in extrapolating from a LOAEL in the absence of a NOAEL; and
(5) uncertainty when the database is incomplete or there are problems with applicability of
available studies. When scientifically sound, peer-reviewed assessment-specific data are not
available, default adjustment values are selected for the individual UFs. For each type of
uncertainty (when relevant to the assessment), the EPA typically applies an UF value of 10 or
3 with the cumulative UF value leading to a downward adjustment of 10-3000 fold from the
selected POD. An UF of 3 is used when the data do not support the use of a 10-fold factor. If
an extrapolation step or adjustment is not relevant to an assessment (e.g., if applying human
toxicity data and an interspecies extrapolation is not required) the associated UF is not used.
The major adjustment steps are described more fully below.
1) Heterogeneity among humans is a key source of variability as well as uncertainty.
Uncertainty related to human variation is considered in extrapolating doses from a subset or
smaller-sized population, often of one sex or of a narrow range of life stages (typical of
occupational epidemiologic studies), to a larger, more diverse population. In the absence of
pollutant-specific data on human variation, a 10-fold UF is used to account for uncertainty
associated with human variation. Human variation may be larger or smaller; however, data to
examine the potential magnitude of human variability are often unavailable. In some
situations, a smaller UF of 3 may be applied to reflect a known lack of significant variability
among humans.
35
-------
2) Extrapolation from results of studies in experimental animals to humans is a
necessary step for the majority of chemical risk assessments. When interpreting animal data,
the concentration at the POD (e.g. NOAEL, BMDL) in an animal model (e.g. rodents) is
extrapolated to estimate the human response. While there is long-standing scientific support
for the use of animal studies as indicators of potential toxicity to humans, there are
uncertainties in such extrapolations. In the absence of data to the contrary, the typical
approach is to use the most relevant endpoint from the most sensitive species and the most
sensitive sex in assessing risks to the average human. Typically, compound specific data to
evaluate relative sensitivity in humans versus rodents are lacking, thus leading to uncertainty
in this extrapolation. Size-related differences (allometric relationships) indicate that typically
humans are more sensitive than rodents when compared on a mg/kg/day basis. The default
choice of 10 for the interspecies UF is consistent with these differences. For a specific
chemical, differences in species responses may be greater or less than this value.
Pharmacokinetic models are useful to examine species differences in pharmacokinetic
processing and associated uncertainties; however, such dosimetric adjustments are not always
possible. Information may not be available to quantitatively assess toxicokinetic or
toxicodynamic differences between animals and humans, and in many cases a 10-fold UF
(with separate factors of 3 for toxicokinetic and toxicodynamic components) is used to
account for expected species differences and associated uncertainty in extrapolating from
laboratory animals to humans in the derivation of a reference value. If information on one or
the other of these components is available and accounted for in the cross-species
extrapolation, a UF of 3 may be used for the remaining component.
3) In the case of reference values for chronic exposures where only data from shorter
durations are available (e.g., 90-day subchronic studies in rodents) or when such data are
judged more appropriate for development of an RfC, an additional UF of 3 or 10-fold is
typically applied unless the available scientific information supports use of a different value.
4) Toxicity data are typically limited as to the dose or exposure levels that have been
tested in individual studies; in an animal study, for example, treatment groups may differ in
exposure by up to an order of magnitude. The preferred approach to arrive at a POD is to use
BMD analysis; however, this approach requires adequate quantitative results for a meaningful
analysis, which is not always possible. Use of a NOAEL is the next preferred approach after
BMD analysis in determining a POD for deriving a health effect reference value. However,
many studies lack a dose or exposure level at which an adverse effect is not observed (i.e., a
NOAEL is not identified). When using data limited to a LOAEL, a UF of 10 or 3-fold is often
applied.
5) The database UF is intended to account for the potential for deriving an
underprotective RfD/RfC due to a data gap preventing complete characterization of the
chemical's toxicity. In the absence of studies for a known or suspected endpoint of concern, a
UF of 10 or 3-fold is typically applied.
36
-------
Acute noncancer assessment
Many of the UFs used to account for variability and uncertainty in the development of acute
reference values are quite similar to those developed for chronic durations. For acute
reference values, though, individual UF values may be less than 10. UFs are applied based on
chemical- or health effect-specific information or based on the purpose of the reference value.
The UFs applied in acute reference value derivation include: 1) heterogeneity among
humans; 2) uncertainty in extrapolating from animals to humans; 3) uncertainty in LOAEL to
NOAEL adjustments; and 4) uncertainty in accounting for an incomplete database on toxic
effects of potential concern. Additional adjustments are often applied to account for
uncertainty in extrapolation from observations at one exposure duration (e.g., 4 hours) to
arrive at a POD for derivation of an acute reference value at another exposure duration (e.g., 1
hour).
Not all acute dose-response values are developed for the same purpose and care must be taken
when interpreting the results of an acute assessment of human health effects relative to the
reference value or values being exceeded. Where relevant to the estimated exposures, the lack
of dose-response values at different levels of severity should be factored into the risk
characterization as potential uncertainties.
37
-------
5 References
American Industrial Hygiene Association. 2018. 2018 Emergency Response Planning
Guidelines/Workplace Environmental Exposure Levels Handbook, https://www.aiha.org/get-
involved/AIHAGuidelineFoundation/EmergencvResponsePlanningGuidelines/Documents/20
18%20ERPG-WEEL%20Handbook%20(ERPGYpdf
Agency for Toxic Substances and Disease Registry. 2016. Exposure Dose Guidance for
Determining Life Expectancy and Exposure Factor.
California Environmental Protection Agency. 2008. Technical Support Document
For the Derivation of Noncancer Reference Exposure Levels.
https://oehha.ca.gov/media/downloads/crnr/appendixdlfinal.pdf
California Environmental Protection Agency. 2009. Technical Support Document for Cancer
Potency Factors 2009. https://oehha.ca.gov/air/crnr/technical-support-document-cancer-
potencv-factors-2009.
California Environmental Protection Agency. 2016. OEHHA Acute, 8-hour and Chronic
Reference Exposure Level (REL) Summary, https://oehha.ca. gov/air/ general -info/oehha-
acute-8-hour-and-chronic-reference-exposure-level-rel-summary.
National Academy of Sciences. 1994. National Research Council. Science and Judgement in
Risk Assessment, https://www.nap.edu/catalog/2125/science-and-iudgment-in-risk-
assessment.
National Academy of Sciences. 2001. National Research Council. Standing Operating
Procedures for Developing Acute Exposure Guideline Levels for Hazardous Chemicals.
https://www.nap.edu/read/10122/chapter/1
National Academy of Sciences. 2010. National Research Council. Acute Exposure Guideline
Levels for Selected Airborne Chemicals, Volume 9.
https://www.nap.edu/catalog/12978/acute-exposure-guideline-levels-for-selected-airborne-
chemical s-volume-9.
Office of Management and Budget. 2007. Memorandum for the Heads of Executive
Departments and Agencies - Updated Principles for Risk Analysis (September 19, 2007),
from Susan E. Dudley, Administrator, Office of Information and Regulatory Affairs, Office of
Management and Budget; and Sharon L. Hays, Associate Director and Deputy Director for
Science, Office of Science and Technology Policy.
https://georgewbush-whitehouse.archives.gov/omb/memoranda/fv2007/m07-24.pdf
R.P. Subramaniam, P. White and V.J. Cogliano. 2006. Comparison of cancer slope factors
using different statistical approaches, Risk Anal. Vol 26, p. 825-830.
38
-------
U.S. Environmental Protection Agency. 1994. Methods for Derivation of Inhalation
Reference Concentrations and Application of Inhalation Dosimetry. EPA/600/8-90/066F.
https://www.epa.gov/risk/methods-derivation-inhalation-reference-concentrations-and-
application-inhalation-dosimetry
U.S. Environmental Protection Agency. 1999. Residual Risk Report to Congress. 453R-99-
001. https://www3.epa.gov/airtoxics/rrisk/risk rep.pdf
U.S. Environmental Protection Agency. 2000a. Risk Characterization Handbook. EPA 100-B-
00-002. https://www.epa.gov/sites/production/files/2015-
10/documents/osp risk characterization handbook 2000.pdf
U.S. Environmental Protection Agency. 2001. National-Scale Air Toxics Assessment for
1996. EPA 453/R-01-003. January 2001. page 85.
U.S. Environmental Protection Agency. 2002a. EPA's Guidelines for Ensuring and
Maximizing the Quality, Objectivity, Utility, and Integrity of Information Disseminated by
the Environmental Protection Agency. EPA/260R-02-008.
https://www.epa.gov/qualitv/guidelines-ensuring-and-maximizing-qualitv-obiectivitv-utilitv-
and-integritv-information
U.S. Environmental Protection Agency. 2002b. A Review of the Reference Dose and
Reference Concentration Processes, https://www.epa.gov/osa/review-reference-dose-and-
reference-concentration-processes
U.S. Environmental Protection Agency. 2005a. Supplemental guidance for assessing early-life
exposure to carcinogens. EPA/630/R-03003F.
https://www3.epa.gov/ttn/atw/childrens supplement final.pdf
U.S. Environmental Protection Agency. 2005b. Guidelines for Carcinogen Risk Assessment.
EPA/630/P-03/001F. https://www.epa.gov/risk/guidelines-carcinogen-risk-assessment.
U.S. Environmental Protection Agency. 2009a. Risk and Technology Review (RTR) Risk
Assessment Methodologies: For Review by the EPA's Science Advisory Board with Case
Studies - MACT I Petroleum Refining Sources and Portland Cement Manufacturing. EPA-
452/R-09-006. https://cfpub.epa.gov/si/si public record report.cfm?dirEntryID=238928
U.S. Environmental Protection Agency. 2009b. Graphical Arrays of Chemical-Specific Health
Effect Reference Values for Inhalation Exposures [Final Report], EPA/600/R-09/061, 2009.
http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid=211003
U.S. Environmental Protection Agency. 2010. Science Advisory Board. Review of EPA's
draft entitled, "Risk and Technology Review (RTR) Risk Assessment Methodologies: For
Review by the EPA's Science Advisory Board with Case Studies - MACT I Petroleum
Refining Sources and Portland Cement Manufacturing."
39
-------
https://vosemite.epa.gOv/sab/sabproduct.nsf/0/b031ddf79cffded38525734fD0649caflQpenDoc
ument&TableRow=2.3#2
U.S. Environmental Protection Agency. 2016. Evaluation of the Inhalation Carcinogenicity of
Ethylene Oxide (Final Report).
https://cfpub.epa.gov/ncea/iris drafts/recordisplav.cfm?deid=329730.
U.S. Environmental Protection Agency. 2017a. Screening Methodologies to Support Risk and
Technology Reviews (RTR): A Case Study Analysis.
https://www3.epa.gov/ttn/atw/rrisk/ScreeningMethodologiestoSupportRTRs ACaseStudvAna
lvsis.pdf.
U.S. Environmental Protection Agency. 2017b. Revisions to the Guideline on Air
Quality Models: Enhancements to the AERMOD Dispersion Modeling System and
Incorporation of Approaches to Address Ozone and Fine Particulate Matter. 40 CFR Part 51.
https://www3.epa. gov/ttn/ scram/guidance/ guide/appw 17.pdf
U.S. Environmental Protection Agency. 2018a. Science Advisory Board. Review of EPA's
draft technical report entitled "Screening Methodologies to Support Risk and Technology
Reviews (RTR): A Case Study Analysis."
https://vosemite.epa.gov/sab/sabproduct.nsf/LookupWebReportsLastMonthBOARD/7A84AA
DF3F2FE04A85258307005F7D70/$File/EPA-SAB-18-004+.pdf
U.S. Environmental Protection Agency. 2018b. User's Guide for the AMS/EPA Regulatory
Model (AERMOD). EPA-454/B-18-001.
https://www3.epa.gov/ttn/scram/models/aermod/aermod userguide.pdf
U.S. Environmental Protection Agency. 2018c. AERMOD Implementation Guide. EPA-
454/B-18-003.
https://www3.epa.gov/ttn/scram/models/aermod/aermod implementation guide.pdf
U.S. Environmental Protection Agency. 2018d. Office of Air Quality Planning and Standards.
Prioritized Chronic Dose-Response Values (6/18/2018). https://www.epa.gov/fera/dose-
response-assessment-assessing-health-risks-associated-exposure-hazardous-air-pollutants
U.S. Environmental Protection Agency. 2018e. Office of Air Quality Planning and Standards.
Acute Dose-Response Values for Screening Risk Assessments (6/18/2018).
https://www.epa.gov/fera/dose-response-assessment-assessing-health-risks-associated-
exposure-hazardous-air-pollutants
40
-------
Appendix 1 to the Risk Assessment Report for the Sterigenics Facility in Willowbrook,
Illinois:
Development of Ethylene Oxide Emissions Rates Used for Risk Assessment
Introduction
We (the EPA) developed ethylene oxide (EtO) emission estimates for the Sterigenics facility in
Willowbrook, Illinois (Willowbrook 1 and Willowbrook 2 buildings), starting with information
provided to us by Sterigenics regarding their operations, estimated emissions rates, and
operational parameters for both the controlled and uncontrolled sources. We took this
information and derived site-specific emission factors from previous stack testing results for the
"controlled" sources, and estimated site-specific emission factors for the uncontrolled or
"fugitive" emissions. Emission factors are calculated values that relate the quantity of a
pollutant released to the atmosphere with an activity associated with the release of that pollutant
and are generally assumed to be representative of long-term averages. Using dispersion
modeling, we evaluated the accuracy of these site-specific emission factors and made
adjustments to the factors so that the modeled results would better correspond with the ambient
air concentrations measured at the monitoring sites near the facility. Tables 1 and 2 give the site-
specific emission factors for each emission point type used for the risk assessment.
Table 1. Willowbrook 1 and Willowbrook 2 site-specific emission factors used for the risk assessment
Facility
Sterilizer vacuum vent
(lbs EtO emitted/ton used)
Aeration room and backvent
(lbs EtO emitted/ton used)
Fugitives
(lbs EtO emitted/ton used)
Willowbrook 1
0.9
0.5
12.0
Willowbrook 2
9.4
0.5
13.0
The EPA used the site-specific emission factors and annual EtO usage rates for each building to
determine the EtO emission rate for each emission point. An emission rate is the mass of a
pollutant emitted over a period of time. The emission rate for each emission point was calculated
as:
£r = EF * UD *K
Where:
Er = Emission Rate (Ib/hr) EF = Emission Factor (lbs EtO emitted/ton used)
Ud = 2017 Facility Usage12 (ton/year) K = 0.000114, conversion from lbs/year to Ibs/hr
The emission rates for all sources at Willowbrook 1 and Willowbrook 2 were combined to yield
the emissions estimates in Table 2.
Table 2. Willowbrook 1 and Willowbrook 2 emission estimates used for the risk assessment
Emission Rate (Ibs/hr)
Willowbrook 1
0.28
Willowbrook 2
0.19
Methodology
The emission factors in Table 1 were developed in part based upon ambient sampling that was
performed by the EPA in Willowbrook, Illinois, from November 13, 2018 to March 31, 2019.
11 Combined output for all fugitive emission sources.
12 2017 usage rates Willowbrook 1 (142 tons), Willowbrook (70 tons).
1-1
-------
Sampling was conducted at eight total locations, two of which are very near the facility
(Willowbrook Village Hall and EPA warehouse), and six additional sampling locations in the
surrounding community. For the purposes of this analysis, only the sample data for Willowbrook
Village Hall and the EPA warehouse were used, and only for the dates on which the facility was
actively processing EtO.13 The EtO samples were collected and analyzed according to EPA
Compendium Method TO-15, Determination of Volatile Organic Compounds (VOCs) in Air
Collected in Specially Prepared Canisters and Analyzed by Gas Chromatography/Mass
Spectrometry (GC/MS),14 and the Quality Assurance Project Plan (QAPP) for the Field
Sampling Plan for Ambient Air Ethylene Oxide Monitoring Near Sterigenics Facility,
Willowbrook, IL, dated November 17, 2018.15 The ambient air samples were collected on a 1-in-
3 day schedule16 throughout the program with the exception of periods in which sampling was
collected off-schedule to accommodate holidays or when weather was not conducive to
sampling.
Sterigenics provided information to the EPA regarding the locations of expected EtO emissions
points for both controlled and fugitive emissions, as well as emission factors for these sources.
This information included the exact location, release height above ground, exit velocity,
temperature, and other parameters needed for dispersion modeling. In addition to this
information, the company also provided daily EtO usage rates17 for each building for the entire
sampling period, which were used to determine the daily emission rates for the individual
emission points.
Air dispersion modeling of the emission points18 was conducted using the latest version of the
American Meteorological Society/EPA Regulatory Model (AERMOD) atmospheric dispersion
model (version 18081). Meteorological data used for the dispersion modeling came from a
temporary weather station located on the roof of the EPA warehouse building. Where
meteorological data were not available from this location due to data availability or quality
concerns, alternate data were acquired from Midway Airport, located approximately 16 km east
of the facility. For each day in which samples were collected, modeling runs were performed
using the established modeling parameters (all emission locations), the meteorological data for
that day, and calculated daily emission rates (all emission locations combined) to determine the
projected impact (i.e., concentrations) of EtO in the areas surrounding the facility. The modeling
does not consider any background concentrations of EtO that may be present in the ambient air;
it only takes into account EtO emissions from emission points at the facility. To compare the
measured ambient values against the modeled values, the EPA corrected the modeling results to
include background concentrations19 of EtO by adding the corresponding background
concentration observed at the upwind location for each sampling day. Upwind locations were
13 November 13, 2018 - February 11, 2019.
14 USEPA. 1999. "Air Method, Toxic Organics-15 (TO-15): Compendium of Methods for the Determination of
Toxic. Organic Compounds in Ambient Air, Second Edition: Determination of Volatile Organic Compounds
(VOCs) in Air Collected in Specially-Prepared Canisters and Analyzed by Gas Chromatography/Mass Spectrometry
(GC/MS)." EPA 625/R-96/010b. https://www.epa.gov/homeland-securitv-research/epa-air-method-toxic-organics-
15-15 -determination-volatile-organic.
15 https://www.epa.gov/sites/production/files/2018-ll/documents/aapp eto willowbrook vl.4 final signed.pdf.
16 See addendum for sampling days and the sample results for all locations (Table A-l).
17 See addendum for EtO usage for Willowbrook 1 and Willowbrook 2 (Table A-2).
18 See addendum for emission point details (Table A-3).
19 See addendum for daily background EtO levels (Table A-4).
1-2
-------
identified based on daily meteorology to determine which residential sampling location was not
affected by emissions from the facility.
We made a number of assumptions regarding the other sources of EtO emissions in the area of
the facility and the emissions from and modeling parameters for the Sterigenics fugitive emission
points that could not be verified from previous testing. We evaluated all known sources of EtO in
the area and did not identify any significant sources. To confirm this assumption, we used a
diagnostic mapping tool called a polarPlot20 that shows EtO concentrations by wind speed and
direction and allows us to identify any potential sources of EtO. This tool identified no sources of
EtO other than Sterigenics. Additionally, while there are no test data to verify the exact location
of the fugitive sources at the company and their associated modeling parameters, the information
provided by the company seemed appropriate based on our understanding of the processes at the
facility.
Emission Factor Development and Evaluation
The development of the site-specific emission factors was predicated on the ability to achieve
agreement between the modeled values with the observed values from the ambient sampling. To
do this, we used an iterative process to evaluate different emission factors and modeling
parameters to predict emissions versus the observed ambient values within the accuracy of the
model (factor of +/- 2). This was done by determining the impact at the location of the ambient
monitoring sites using modeling of each emission point (controlled and fugitive) at the facility.
As a starting point, we performed a sensitivity analysis for each of the site-specific emission
factors provided by Sterigenics against a "strawman" scenario representing a decrease in the
control efficiency of those controlled sources and an increase in fugitives for a number of
ambient sampling days.21 We took the site-specific emission factors combined with the
corresponding daily usage rate data for each building to determine the daily EtO emission rate
for each emission point. The emission rates for each sampling day were calculated in the same
manner as for the risk assessment, but the daily usage rate was used to determine an emission
rate specific to the sampling day. Table 3 gives the emission factors used for the sensitivity
analysis.
Table 3. Site Specific Emission Factors Used for Sensitivity Analysis
Whole site emission
Sterilizer vacuum
Aeration room and
Fugitives
Building
factor (lbs/ton)
vent (lbs/ton)
backvent (lbs/ton)
(lbs/ton)
Sterigenics Emission Factor
Willowbrook 1
1.4
0.01
0.4
1.0
Willowbrook 2
2.5
1.1
0.4
1.0
Strawman
Willowbrook 1
5.9
1.9
1.0
3.0
Willowbrook 2
5.9
1.9
1.0
3.0
Table 4 gives the average model-to-monitor comparison for the sensitivity analysis. The results
of this analysis indicated that the results of the modeling using the emission factors used for both
the Sterigenics and the EPA Strawman were significantly underpredicting the observed values.
20 See addendum of polarPlot maps (Figure A-l).
21 December 6, 13, and 26, 2018; and January 17.
1-3
-------
Table 4. Model to Monitor Comparison for the Sensitivity Analysis
Location
Observed
(Hg/m )
Sterigenics emission
factors (ng/m )
Strawman emission factor
(Hg/m )
Willowbrook Village Hall
4.69
0.13
0.61
EPA Warehouse
8.41
0.49
2.23
Based on these results, we chose to modify the emission factors in Table 3 for the controlled
emissions from the EPA strawman to be in-line with manufacturer guarantees for similar
pollution control equipment installed at the facility. We also reviewed the modeling parameters
and compared them against previous test data at the facility as well as other test data from similar
sources. This review yielded some seasonal corrections to the modeling parameters to better
reflect the likely exit temperatures of the exhaust points during the winter months. With the
controlled emission factors set, we incrementally increased the emission factors for the fugitive
sources until the objectives were met for the comparison of the modeled results to the observed
values. During this period, we were in contact with the company regarding the modifications
being made to the facility air handling system and how these changes would affect the fugitive
sources. We made revisions to the modeling parameters as new information was received, and
these revisions were used for all modeling going forward. Figure 1 gives the ambient monitoring
results (observed) plotted against the values developed from the dispersion modeling (modeled)
based on the final emission factors and modeling parameters, for all monitor locations. This plot
compares the monitored to the modeled results in a manner consistent with past evaluations of
AERMOD22 by comparing the monitored and modeled results unpaired in time and space, called
a Q-Q plot. The monitored and modeled concentration distributions are both sorted and plotted
against each other based on rank, so the highest monitored concentration is compared against the
highest modeled concentration, regardless of the location and time of occurrence.
Figure 1. Modeled value vs. observed value comparison (11/19/2018 - 02/11/2019)
Scenario 2: 11/19/18-02/11/19
All monitors and r>o outliers
We did a model-to-monitor comparison using a statistic called the Robust Highest Concentration
(RHC) and fractional bias. This comparison focuses on the higher concentrations in the
distribution. The RHC coupled with fractional bias is the preferred methodology in the EPA's
22 USEPA. 2003. "AERMOD: Latest Features and Evaluation Results." EPA-454/R-03-003.
https://www3.epa.gov/scram001/7thconf/aermod/aermod mep.pdf.
1-4
-------
Protocol for Determining the Best Performing Model.23 Normally, the protocol evaluates 1-hour,
3-hour, and 24-hour average concentrations. Since the ambient monitoring data for Sterigenics
are only 24-hour averages, we focused only on 24-hour averages. The RHC is calculated at each
monitoring location for observed concentrations and modeled concentrations.
The RHC is calculated as:
3 N
RHC = X(JV) + [X - X(JV)] x In
Where X(N) is the Nth highest concentration, and X is the average of N-l values where N is
typically set to 26 values for most model evaluations. However, given the small sample size at
each monitor, we started with N=11 and evaluated results up to N=20 (the fewest number of
observations across the monitors). As stated above, the RHC is calculated at each monitor for
observed concentrations and modeled concentrations. Next a fractional bias is calculated using
the maximum observed RHC and maximum modeled RHC as:
FB = 2
OB - PR
YOB + PR\
Where FB is the fractional bias, OB is the maximum observed RHC, and PR is the maximum
modeled RHC. A positive (negative) fractional bias indicates model underprediction
(overprediction). Fractional biases within ± 0.67 are not considered statistically different. Also,
note that the two RHC values in the fractional bias may not be from the same monitor location.
This is done to assess the model's ability to assess concentrations for regulatory purposes, that is,
how well the model predicts maximum concentrations regardless of the spatial location. Table 5
gives the fractional biases and monitors used for the calculations for a range of values of N using
the meteorology at the EPA warehouse and the estimated emissions factors.
Table 5. Fractional Bias Estimates Using All Monitors
N
Observed
RHC
Modeled
RHC
Fractional
Bias
Observed monitor
location
Modeled monitor
location
11
20.8
8.0
0.89
EPA Warehouse
EPA Warehouse
12
19.8
7.5
0.90
EPA Warehouse
EPA Warehouse
13
19.0
7.3
0.9
EPA Warehouse
EPA Warehouse
14
17.9
7.0
0.9
EPA Warehouse
EPA Warehouse
15
16.9
6.8
0.8
EPA Warehouse
EPA Warehouse
16
16.7
6.7
0.9
EPA Warehouse
EPA Warehouse
17
16.1
7.0
0.8
EPA Warehouse
EPA Warehouse
18
16.2
6.9
0.8
EPA Warehouse
EPA Warehouse
19
14.4
6.5
0.8
EPA Warehouse
EPA Warehouse
20
13.7
6.3
0.7
EPA Warehouse
EPA Warehouse
We also generated a Q-Q plot of the concentrations at only the Willowbrook Village Hall and the
EPA warehouse, shown in Figure 2. The plot indicates good agreement on the low end of the
concentration distribution, and underprediction at the middle to high end of the concentration
23 USEPA. 1992. Protocol for Determining the Best Performing Model. EPA-454/R-92-025.
1-5
-------
distribution, but within a factor of 2, which is acceptable performance. At the highest end of the
distribution, the model is just slightly underpredicting compared to the observed maximum.
Figure 2. Q-Q plot
Scenario 2 (Warehouse and ¥llag« Hall: no ouSiars)
Scenario 2; 11/19/18-02/11/19
Warehouse and Village Hall: no outliers
In addition to the RHC analysis and Q-Q plots, we also did a direct comparison of the modeled
values against the observed values at Willowbrook Village Hall and the EPA warehouse. For this
analysis, all data points were included in the comparison unless a sample was invalided, elevated
background concentrations were observed, or when a result was considered an outlier. A total of
47 data points was used for this analysis, 26 from sampling events at the Willowbrook Village
Hall monitoring location and 21 from the EPA warehouse monitoring location. The modeled
value agreed (within a factor of 2) with the observed value for approximately 65 percent of the
sampling events, with the model overpredicting 15 percent and underpredicting 20 percent of the
time. A comparison of the means of the modeled versus the observed or monitored results, the
observed mean was within the accuracy of the model, although the model appears to
underpredict. The mean observed value is heavily influenced by the elevated values observed
after January 12, 2019, following a maintenance event at Willowbrook 1. Tables 6 and 7 present
the results of the model-to-monitor comparison for the entire sampling period and for the period
prior to the maintenance event at Willowbrook 1, respectively.
Table 6. Model-to-monitor comparison 11/19/2019 - 02/11/2019
Location
Mean Observed Value
(Hg/m )
Mean Modeled Value
(Hg/m )
Willowbrook Village Hall
2.83
1.53
EPA Warehouse
3.14
2.02
24 Corrected for background.
1-6
-------
Table 7. Model-to-monitor comparison 11/19/2019 - 01/09/2019
Location
Mean Observed Value
(Hg/m )
Mean Modeled Value
(Hg/m )
Willowbrook Village Hall
2.85
2.05
EPA Warehouse
2.31
2.69
The model-to-monitor comparison showed reasonable results when comparing mean results at
the monitor location, but the model had difficulty predicting the elevated results at these
locations on a few of the days when samples were collected. Disparities in the modeled versus
the observed results can be attributed to the model's sensitivity to errors in the meteorology or to
the other activities at the facility or happening in the surrounding area that could affect plume
magnitude or dispersion. This could explain the closer relationship observed at the EPA
Warehouse sampling location which was near the temporary weather station located on the EPA
Warehouse building.
Conclusions
The site-specific estimated emission factors from which the emission rates were derived and
modeling parameters developed for the risk assessment appear to adequately predict the expected
concentrations surrounding the facility and, while these factors appear to underpredict the
emissions from the facility, the results are well within the acceptable performance of the model.
The results of this analysis provide an estimation of the emission of the EtO emissions for the
purposes of the risk assessment. These results only provide emission estimates for the period in
time when ambient samples were collected and analyzed. A more refined assessment of these
emissions was problematic due to the limited number of monitoring locations near the facility
and the relatively small sample size. While additional measurements were collected from the
residential areas, these were not used for this analysis due to the significant proportion of EtO
concentrations present in the ambient air not attributed to the company.
The tools used to perform this analysis were adequate due to the magnitude of the emissions
from the facility. Any changes made to the facility or similar facilities which would result in a
significant decrease in EtO emissions would result in a need to revise the way emissions are
characterized. Any future assessment should incorporate direct measurement of all emission
points at the facility during all aspects of operation to more effectively determine emission
factors. As these sources become better controlled (e.g., improved capture and control of
fugitives), emission characterization using ambient measurements will become more difficult
because the contribution from the facility would be less distinguishable from levels found in the
ambient air.
25 Corrected for background.
1-7
-------
Addendum to Appendix 1
Table A-l. Ambient monitoring results ((.ig/nr" for Willowbrook village hall and EPA warehouse
locations
Sample Start
Date
Willowbrook
village hall
EPA
warehouse
Sample Start
Date
Willowbrook
village hall
EPA
warehouse
11/13/2018
Invalid
2.37
1/27/2019
19.3
1.11
11/16/2018
0.824
1.81
2/1/2019
0.954
0.133
11/19/2018
6.11
6.62
2/2/2019
0.383
0.228
11/23/2018
0.284
0.180
2/5/2019
17.3
26.4
11/25/2018
4.10
Invalid
2/8/2019
0.725
5.04
11/28/2018
1.83
0.248
2/11/2019
3.98
ND
12/1/2018
1.68
0.456
2/14/2019
0.178
0.745
12/6/2018
5.39
11.7
2/19/2019
0.239
0.150
12/7/2018
0.737
2.26
2/20/2019
0.260
0.159
12/10/2018
0.300
0.269
2/21/2019
0.144
ND
12/13/2018
2.04
0.436
2/22/2019
0.123
0.121
12/16/2018
0.871
2.11
2/23/2019
0.128
0.132
12/19/2018
0.521
0.345
2/26/2019
0.166
0.119
12/22/2018
0.981
3.09
3/1/2019
ND
0.103
12/26/2018
10.8
Invalid
3/4/2019
0.161
ND
12/28/2018
0.672
1.42
3/7/2019
0.099
0.096
1/2/2019
0.251
0.237
3/10/2019
Invalid
0.075
1/3/2019
0.372
ND
3/13/2019
0.204
0.122
1/6/2019
7.59
ND
3/16/2019
0.461
0.171
1/9/2019
3.81
Invalid
3/19/2019
0.136
0.056
1/12/2019
1.57
ND
3/22/2019
0.060
0.117
1/15/2019
0.672
14.2
3/25/2019
0.078
0.134
1/17/2019
0.517
13.1
3/28/2019
0.114
0.181
1/22/2019
1.51
4.10
3/31/2019
0.057
ND
1/24/2019
0.262
0.280
-
-
-
1-8
-------
Table A-2. Daily ethylene oxide usage rates (lbs) fed to the sterilization chamber
Date
Willowbrook 1
Willowbrook 2
Date
Willowbrook 1
Willowbrook 2
11/13/2018
755 (820)
482 (477)
12/30/2018
853
0
11/14/2018
753
495
12/31/2018
510
0
11/15/2018
794
258
1/1/2019
622
0
11/16/2018
864 (935)
611 (385)
1/2/2019
598 (491)
0(0)
11/17/2018
877
489
1/3/2019
732 (718)
0(0)
11/18/2018
938
465
1/4/2019
795
151
11/19/2018
880 (981)
517 (529)
1/5/2019
703.3
420
11/20/2018
1057
413
1/6/2019
110 (517)
279 (487)
11/21/2018
946
694
1/7/2019
0.3
485
11/22/2018
808
339
1/8/2019
0
274
11/23/2018
827(1036)
690 (593)
1/9/2019
0
338
11/24/2018
844
538
1/10/2019
0
242
11/25/2018
665 (729)
131 (487)
1/11/2019
613.9
485
11/26/2018
844
0
1/12/2019
940 (895)
315 (468)
11/27/2018
789
0
1/13/2019
693.7
489
11/28/2018
851 (864)
0(0)
1/14/2019
911.4
333
11/29/2018
902
0
1/15/2019
764 (805)
318 (336)
11/30/2018
943
0
1/16/2019
950.7
58
12/1/2018
793 (908)
11(11)
1/17/2019
813 (760)
344 (128)
12/2/2018
837
515
1/18/2019
857.7
420
12/3/2018
975
341
1/19/2019
800.2
343
12/4/2018
1035
390
1/20/2019
803.6
484
12/5/2018
972
445
1/21/2019
1068.2
317
12/6/2018
1054(1105)
347 (317)
1/22/2019
787 (1003)
298 (417)
12/7/2018
697 (839)
262 (480)
1/23/2019
862.1
373
12/8/2018
948
447
1/24/2019
653 (859)
340 (426)
12/9/2018
1020
415
1/25/2019
960.9
396
12/10/2018
852 (892)
412 (494)
1/26/2019
759.7
444
12/11/2018
843
414
1/27/2019
888 (875)
286 (313)
12/12/2018
797
416
1/28/2019
916.1
313
12/13/2018
1064 (852)
476 (441)
1/29/2019
866.4
358
12/14/2018
671
59
1/30/2019
607.1
289
12/15/2018
574
0
1/31/2019
928.1
357
12/16/2018
626 (786)
293 (222)
2/1/2019
892
345
12/17/2018
964
470
2/2/2019
829
340
12/18/2018
669
384
2/3/2019
821.5
188
12/19/2018
826 (988)
402 (312)
2/4/2019
795.1
282
12/20/2018
878
351
2/5/2019
773
344
12/21/2018
784
342
2/6/2019
974.6
131
12/22/2018
685 (953)
0(283)
2/7/2019
790.4
312
12/23/2018
797.2
0
2/8/2019
847
470
12/24/2018
736
350
2/9/2019
929.6
352
12/25/2018
893
399
2/10/2019
657.3
553
12/26/2018
631 (796)
471 (471)
2/11/2019
814
260
12/27/2018
784
360
2/12/2019
69.5
302
12/28/2018
593 (684)
295 (293)
2/13/2019
818.7
442
12/29/2018
671
228
2/14/2019
852.8
408
1-9
-------
Note: BOLD values are days in which ambient sampling was taken. Additionally, the values in (parenthesis) for
sample dates from 11/13/2018 - 1/27/2019 are the estimated mass of ethylene oxide sent to the pollution controls.
Table A-3. Willowbrook 1 and Willowbrook 2 emission points and locations
EtO
Source
Easting
Northing
Emissions
Building
ID
Source Description
(x)"-
(Y)"
(Yes/No)
Emission Type
Controlled emissions from the chamber vent
WB1
STK1
Deoxx
421892.07
4622242.11
Yes
WB1
Controlled emissions from the aeration rooms
STK2
AAT Scrubber
421897.15
4622252.27
Yes
and backvent
WB1
1EF11
1-EF-ll Work Aisle
421896.70
4622230.30
Yes
EtO fugitive emission point
l-EF-15 Process Storage/East
Former fugitive emission point, exhaust fan has
WB1
1EF15
Aeration
421911.94
4622211.67
No
been turned off effective January 2019 (assumed)
WB1
1EF3
l-EF-3 Shipping
421835.32
4622206.80
Yes
EtO fugitive emission point
WB1
l-EF-4 Process
Yes
EtO fugitive emission point
1EF4
Storage/Central Aeration
421868.72
4622224.47
WB1
1EF10
l-EF-10 Maintenance Aisle
421897.74
4622213.58
No
Former fugitive emission point
WB1
l-EF-9 Work Aisle/Boiler
Yes
EtO fugitive emission point
1EF9
Room
421888.14
4622229.62
WB1
Former fugitive emission point, exhaust fan has
1EF13
l-EF-13 Chamber A or 9
421904.23
4622241.98
No
been turned off
WB1
l-EF-20 Chamber B Cubical
Former fugitive emission point, exhaust fan has
1EF20
Exhaust
421922.88
4622241.05
No
been turned off
WB1
l-EF-21 Aat Scrubber Room
No emission expected
1EF21
Exhaust
421925.04
4622249.06
No
WB1
1EF8
l-EF-8 Pump Aisle
421879.63
4622243.03
No
No emission expected
WB1
l-EF-12 Chamber A Gassing
No
Former fugitive emission point, exhaust fan has
1EF12
Room
421908.04
4622241.75
been turned off
WB1
1EF16
l-EF-16 Chamber A Cubicle
421913.64
4622241.08
No
No emission expected
WB1
l-EF-19 Chamber E Cubical
No
No emission expected
1EF19
Exhaust
421921.00
4622223.31
WB1
l-EF-18 Chamber C Cubical
No
No emission expected
1EF18
Exhaust
421916.72
4622238.97
Yes
Controlled emissions from chamber vent,
WB2
A
AAT Scrubber
421701.70
4622357.89
aeration room, and backvents
No
Former EtO emission point, routed to AAT
WB2
B
3 Chamber Backvent
421708.37
4622378.69
scrubber July 2018
No
Former EtO emission point, routed to AAT
WB2
C
1 Chamber Backvent
421709.16
4622354.88
scrubber July 2018
WB2
P
Chamber Room Exhaust Fan
421736.89
4622335.04
Yes
EtO fugitive emission point
WB2
Q
Work Aisle Exhaust Fan
421736.30
4622328.70
Yes
EtO fugitive emission point
Former fugitive emission point, exhaust fan has
WB2
T2
North Wall Vent West
421713.72
4622390.70
No
been turned off effective January 2019 (assumed)
No
Former fugitive emission point, exhaust fan has
WB2
T3
North Wall Vent East
421742.29
4622390.70
been turned off effective January 2019 (assumed)
26 Coordinates reflect UTM NAD83, Zone
27 Coordinates reflect UTM NAD83, Zone
16
16
1-10
-------
Table A-4. Daily background ethylene oxide levels
Modeled
Corrected
Background
Background Location
Background value
background value
Date
(Hg/m )
(Hg/m )
(Hg/m )
11/19/2018
0.164
Gower ES
0.016
0.148
11/23/2018
0.197
Gower MS
0.007
0.190
11/25/2018
0.345
Willow Pond Park
0.046
0.299
11/28/2018
0.656
Gower MS
0.064
0.592
12/1/2018
0.211
Willow Pond Park
0.013
0.198
12/6/2018
0.082
Willow Pond Park
0.022
0.060
12/7/2018
0.164
Gower ES
0.030
0.134
12/10/2018
0.138
Gower ES
0.017
0.121
12/13/2018
0.211
Water Tower
0.060
0.151
12/16/2018
0.732
Gower ES
0.011
0.721
12/19/2018
0.360
Gower MS
0.028
0.332
12/22/2018
0.360
Gower ES
0.027
0.333
12/26/2018
0.082
Gower MS
0.084
-0.002
12/28/2018
0.133
Gower ES
0.010
0.123
1/2/2019
0.210
Gower ES
0.004
0.206
1/3/2019
0.082
West Neighborhood
0.040
0.042
1/6/2019
0.082
Willow Pond Park
0.006
0.076
1/9/2019
0.295
Hinsdale South High School
0.027
0.268
1/12/2019
0.082
Gower MS
0.007
0.075
1/15/2019
0.082
Gower ES
0.008
0.074
1/17/2019
0.144
Willow Pond Park
0.008
0.136
1/22/2019
0.349
Hinsdale South High School
0.059
0.290
1/24/2019
0.095
Gower ES
0.005
0.090
1/27/2019
0.155
Gower MS
0.045
0.110
2/1/2019
0.101
Gower MS
0.039
0.062
2/2/2019
0.371
Gower MS
0.016
0.355
2/5/2019
0.174
Willow Pond Park
0.006
0.168
2/8/2019
0.202
Gower ES
0.010
0.192
2/11/2019
0.089
Willow Pond Park
0.001
0.088
1-11
-------
Figure A-l. EtO Concentration Plots for the Willowbrook Village Hall and EPA Warehouse Monitors
Willow Brook EtO Monitoring Network
75th St
IHJJ
Wmuvrvt
to
i
o
15th PI 2 &
< c
I
%
\
Wmt EiO CorMntnvMiw Uigin '>from 11/13 OKI
U
O
Executive Dr
1
U ..!»
01
a.6
//^
£
* ->°
AKF ~
Mwn EiO ConcttftriUion* mu m from 11 «13-C3.'31
79th St
ov
6^
th St
«
*
mi
«0
**
I
1-12
-------
Appendix 2 to the Risk Assessment Report
for the Sterigenics Facility in Willowbrook, Illinois:
Technical Support Document for HEM-AERMOD Modeling
2-1
-------
Modeling for the Residual Risk and Technology Review
Using the Human Exposure Model 3 - AERMOD Version
Updated 4/24/2019
Technical Support Document
Prepared for:
U.S. Environmental Protection Agency
Office of Air and Radiation
Office of Air Quality Planning and Standards
Health and Environmental Impacts Division
Air Toxics Assessment Group
Research Triangle Park, NC 27711
Prepared by:
EC/R Incorporated
(A wholly owned subsidiary of SC&A, Inc.)
501 Eastowne Drive, Suite 250
Chapel Hill, North Carolina 27514
2-2
-------
Disclaimer
The research described in this document has been funded by the United States
Environmental Protection Agency contracts 68-D-01-076, EP-D-06-119, and
currently EP-W-12-011 to EC/R Incorporated. Although it has been subject to the
Agency's review, it does not necessarily reflect the views of the Agency, and no
official endorsement should be inferred.
2-3
-------
1. Introduction
Contents
6
2. Overview of the HEM-3 - AERMOD System 7
2.1 Preparation of Dispersion Modeling Inputs 8
2.1.1 Compiling Emission Source Data 8
2.1.2 Defining the Modeling Domain and Receptors 9
Treatment of Nearby Census Blocks and Screening for Overlapping Blocks 10
Polar receptor network 10
Elevations and hill heights for model receptors 11
2.1.3 Selection of Meteorological Data 11
2.2 Running of AERMOD 11
2.2.1 AERMOD Dispersion Options Used by HEM-3 11
2.2.2 Use of Dilution Factors 13
2.3 Postprocessing of AERMOD Results in HEM-3 13
2.3.1 Calculation of Impacts at Modeled Receptors 14
2.3.2 Interpolation of Impacts at Outer Census Blocks 15
2.3.3 Calculation of Population Exposures and Incidence 16
2.3.4 Model Outputs 17
2.4 Data Libraries Used in HEM-3 18
2.4.1 Chemical Health Effects Information 18
2.4.2 Census Block Locations and Elevation Data 19
2.4.3 Meteorological Data 20
2.4.4 Gaseous Deposition Parameters 20
3. Modeling for the Residual Risk Technology Review 21
3.1 Emission Source Inputs 21
3.2 Pollutant Cross-Referencing 22
3.3 Meteorological Data 22
3.4 Model Options Selected 25
3.4.1 Urban or Rural Dispersion Characteristics 25
2-4
-------
3.4.2 Deposition and Plume Depletion 25
3.4.3 Cutoff Distance for Modeling of Individual Blocks 26
3.4.4 Facility Boundary Assumptions 26
3.5 Modeling of Multiple Facilities 28
4. Quality Assurance 29
4.1 Engineering Review 29
4.2 Geographic Pre-Modeling Checks 29
4.3 Geographic Post-Modeling Checks 30
5. Uncertainties 32
6. References 38
Figures
Figure 3-1. HEM-3 Meteorological Stations 24
Tables
Table 2-1. Parameters Used to Delineate the Modeling Domain in HEM-3 9
Table 3-1. HEM-3 Domain and Set-Up Options As Used in the Residual Risk and Technology
Review Assessments 27
Table 5-1. Summary of General Uncertainties Associated with Risk and Technology Review
Risk Assessments 34
2-5
-------
1. Introduction
This document describes the general modeling approach used to estimate the risks to
human populations in support of the Residual Risk and Technology Review (RTR) currently
being carried out by the U.S. Environmental Protection Agency (EPA). It is important to note
that risk characterizations of individual source categories under the RTR program may not
follow every item/approach noted in this document. The reader is referred to the main body of
the risk assessment document for more details on source category specific approaches that may
have been included in the analysis.
The model used in these risk assessments is the Human Exposure Model, Version 3
(HEM-3). HEM-3 incorporates AERMOD, a state of the science air dispersion model developed
under the direction of the American Meteorological Society / Environmental Protection Agency
Regulatory Model Improvement Committee (AERMIC).
Section 2 of this report provides an overview of the HEM-3-AERMOD system; and
Section 3 describes inputs and choices made in implementing the model for the RTR program.
Quality assurance efforts undertaken in the modeling effort are discussed in Section 4, and
uncertainties associated with the modeling effort are discussed in Section 5.
2-6
-------
2. Overview of the HEM-3 - AERMOD System
HEM-3 performs three main operations: dispersion modeling, estimation of population
exposure, and estimation of human health risks. The state-of-the-art American Meteorological
Society (AMS) / EPA Regulatory Model (AERMOD)1'2 is used for dispersion modeling.
AERMOD can handle a wide range of different source types which may be associated with an
industrial source complex, including stack (point) sources, area and polygon sources, volume
sources, line and buoyant line sources.
To prepare dispersion modeling inputs and carry out risk calculations, HEM-3 draws
primarily on three data libraries, which are provided with the model. The first is a library of
meteorological data for over 800 stations, which are used for dispersion calculations. A second
library of Census block ("centroid") internal point locations and populations provides the basis
of human exposure calculations. The Census library also includes the elevations of every Census
block, which are used in the dispersion calculations for the RTR assessments. A third library of
pollutant unit risk estimates and reference concentrations is used to calculate population risks.
These unit risk estimates (URE) and reference concentrations (RfCs) are based on the latest dose
response values recommended by EPA for hazardous air pollutants (HAPs) and other toxic air
pollutants. A fourth data library, which contains deposition parameters for gaseous pollutants, is
also provided with HEM-3 but used only when the user opts to compute gaseous deposition with
or without plume depletion. (Note: Deposition has not been computed for the RTR assessments
to date).
HEM-3 has been implemented in two versions: a single facility version ("Single HEM-
S''), and a multiple facility version ("Multi HEM-3"). Multi HEM-3 is used in the RTR risk
assessment modeling. Both versions operate under the same general principles. In essence, Multi
HEM-3 provides a platform for running the single facility version multiple times. In both
versions, source location and emissions data are input through a set of Excel™ spreadsheets. The
main difference is in the user interface for other model inputs. Single HEM-3 includes a
graphical user interface (GUI) for the selection of various dispersion modeling options. In Multi
HEM-3, a control file replaces many of these GUI inputs.
The model estimates cancer risks and non-cancer adverse health effects due to inhalation
exposure at Census block internal point locations (or "centroids"), at concentric rings
surrounding the facility center, and at other receptor locations that can be specified by the user.
Cancer risks are computed using EPA's recommended unit risk estimates for HAPs and other
toxic air pollutants. The resulting estimates reflect the excess cancer risk for an individual
breathing the ambient air at a given receptor site 24-hours per day over a 70-year lifetime. The
model estimates the numbers of people exposed to various cancer risk levels. In addition, HEM-3
estimates the total incremental cancer risks for people living within different distances of the
modeled emission sources.
Potential non-cancer health effects due to chronic exposures are quantified using hazard
quotients and hazard indices for various target organs. The "hazard quotient" (HQ) for a given
chemical and receptor site is the ratio of the ambient concentration of the chemical to the
reference concentration. The "hazard index" (HI) for a given organ is the sum of hazard
quotients for substances that affect that organ. HEM-3 computes target-organ-specific hazard
2-7
-------
indices (TOSHIs) for HAPs and other toxic air pollutants, and estimates the numbers of people
exposed to different hazard index levels. In addition, short term ("acute") concentrations are
computed for all pollutants, and concentrations are compared with various threshold levels for
acute health effects.
The following sections outline the methodologies used in the HEM-3-AERMOD system.
Section 2.1 describes the preparation of dispersion modeling inputs, Section 2.2 describes the
running of AERMOD, Section 2.3 describes calculations performed by HEM-3 to calculate risks
and exposures, and Section 2.4 details the sources and methods used to produce HEM-3 's data
libraries. The HEM-3 User's Manuals - for Single HEM-3 and Multi HEM-3 - provide
additional details on the input data and algorithms used in the model.3 Specific model options
used in the RTR assessments are discussed in Chapter 3.
2.1 Preparation of Dispersion Modeling Inputs
HEM-3 compiles data that will be needed for dispersion modeling, and prepares an input
file suitable for running AERMOD. The dispersion modeling inputs can be divided into three
main components: emission source data, information on the modeling domain and receptors for
which impacts will be computed, and meteorological data.
2.1.1 Compiling Emission Source Data
A series of Excel™ spreadsheet files are used to specify the emissions and configuration
of the facility to be modeled. At a minimum, two files are needed: a HAP emissions file, and an
emissions location file. The HAP emissions file includes an emission source identification code
for each emission source at the facility, the names of pollutants emitted by each source, and the
emission rate for each pollutant. In addition, if the model run is to incorporate deposition or
plume depletion, the HAP emissions file must also specify the percentage of each pollutant that
is in the form of particulate matter. The balance is assumed to be in gaseous/vapor form.
The emissions location file includes the coordinates of each source, as well as
information on the configuration and other characteristics of the source. HEM-3 can analyze
point sources, area and polygon sources, volume sources, and line and buoyant line sources -
configurations that are described in AERMOD's documentation.1'2 For stack (point) sources,
such as a vertical non-capped, capped or horizontal stacks the emissions location file must
provide the stack height, stack diameter, exit velocity, and emission release temperature. The file
must also provide dimensions for each area, polygon, volume or line source, as well as the height
of the source above the ground. For area sources, the angle of rotation from north can also be
specified. The user can also provide the terrain elevation at the base of each source. (The
controlling hill height is also used in AERMOD's flow calculations. Calculation of the
controlling hill height by HEM-3 is discussed in Section 2.4.2.) If the terrain elevations are not
provided by the user, HEM-3 will calculate elevations and controlling hill heights based on
elevations and hill heights provided by the Census database for the Census blocks nearest to the
facility.
If buoyant line source types are to be considered, particularly when computing building
downwash effects, then HEM-3 requires an additional input file to specify the source type's
2-8
-------
parameters. For buoyant line sources, the average buoyancy parameter, the average building
dimensions (i.e., average building length, height, and width), the average line source width, and
the average separation distance between buoyant lines are required inputs for an associated
buoyant line parameters input file.
If particulate deposition and plume depletion are to be considered, then HEM-3 requires
an input file to specify the particle size distribution. This input file must include the average
particle diameter, the mass fraction percentage, and the average particle density for each size
range emitted. Another optional file can be used to specify building dimensions if building wake
effects are to be modeled.
2.1.2 Defining the Modeling Domain and Receptors
HEM-3 defines a modeling domain for each facility that is analyzed based on parameters
specified by the model user or calculated by the model. These parameters are summarized in
Table 2-1. The modeling domain is circular, and is centered on the facility, with a radius
specified by the user. For the RTR analysis, the radius of the modeling domain is 50 kilometers
(km). HEM-3 identifies all of the Census block locations in the modeling domain from its
Census database, and divides the blocks into two groups based on their distance from the facility.
For the inner group of Census blocks (closest to the facility), each block location is modeled as a
separate receptor in AERMOD. The cutoff distance for modeling individual Census blocks is
generally set to 3,000 m (3 km) for the RTR assessments, although it can be set differently by the
model user. The model user can also provide an Excel™ spreadsheet specifying additional
locations to be included as model receptors in AERMOD. These additional discrete "user
receptors" may include facility boundary locations, monitoring sites, individual residences,
schools, or other locations of interest.
Table 2-1. Parameters Used to Delineate the
Modeling Domain in HEM-3
Parameter
Tvpical
\ al ne
Modeling domain size - maximum radial distance to
be modeled from facility center
50 km
Cutoff distance for modeling of individual blocks3
3,000 m
Overlap distance - where receptors are considered
on facility property3
30 m
Polar receptor network:
Distance to the innermost ringb
>100 m
Number of concentric rings
13
Number of radial directions
16
a Measured from each stack at the facility, and from the edges of each area or
volume source.
b Generally model-calculated to encompass all emission sources but not less
than 100 meters from the facility center.
For Census blocks in the outer group, beyond this modeling cutoff distance, emission impacts are
interpolated based on modeling results for a polar receptor network. The user also specifies an
2-9
-------
"overlap" distance, within which Census block coordinates will be considered to be on facility
property. The following paragraphs provide more details on the treatment of blocks near the
facility, on the polar receptor network, and on the determination of receptor elevations and
controlling hill heights to be used in AERMOD.
Treatment of Nearby Census Blocks and Screening for Overlapping Blocks
Census block locations near the facility are modeled as separate receptors within
AERMOD. The cutoff distance for modeling of individual Census blocks may be chosen by the
user, but is typically 3,000 meters for the RTR assessments. This distance is not measured from
the center of the facility, but is the minimum distance from any source at the facility. Therefore,
any Census block location that is within the cutoff distance from any emission source is treated
as a discrete AERMOD receptor.
HEM-3 checks Census blocks that are very close to the facility in order to assess whether
they overlap any point, area, volume, line or buoyant line emission sources. In addition, the user
can specify an overlap distance, within which receptors will be considered to be on facility
property. The default value for the overlap distance is 30 meters, or approximately equal to the
width of a narrow buffer and a roadway. HEM-3 tests each nearby receptor to determine whether
it is within this distance from any stack or from the perimeter of any area, volume, line or
buoyant line source. If a receptor falls within this distance, HEM-3 will not calculate risks based
on the location of that receptor, but will instead assume that the risks associated with the receptor
are the same as the highest predicted value for any receptor that is not overlapping. The location
for calculating the default impact may be either another Census block, one of the polar grid
receptors, or one of the additional discrete user-specified receptor locations. [Note: An exception
to this occurs when modeling polygon sources. Unlike other sources, when modeling polygons,
the overlap function is disabled. This allows the impacts for a census tract modeled as a polygon
source (e.g. mobile source emissions modeled uniformly across a census tract) to be calculated
within the census tract being modeled.]
Polar receptor network
The polar receptor network used in HEM-3 serves three functions. First, it is used to
estimate default impacts if one or more Census locations are inside the overlap cutoff distance
used to represent the facility boundary. Second, it is used to evaluate potential acute effects that
may occur due to short-term exposures in locations outside the facility boundary. Third, the polar
receptor network is used to interpolate long-term and short-term impacts at Census block
locations that are outside the cutoff distance for modeling of individual blocks.
Generally, the model calculates the inner radius (or first ring distance) for the polar
receptor network to be just outside the emission source locations, but not less than 100 meters
from the facility center. However, the user can override the default distance calculated by the
model to fit the size and shape of the facility properties to be modeled. Likewise, the model will
also use default values for the number of concentric rings to be analyzed (13 rings by default),
and the number of radial directions (16 radials by default), although these default values can also
be changed by the user to meet the needs of a specific modeling study. The inner radius of the
2-10
-------
polar network should be the minimum distance from the facility center that is generally outside
of facility property. (For complex facility shapes, it is sometimes useful to specify an inner ring
that encroaches on facility property in some directions.) HEM-3 will distribute the radial
directions evenly around the facility. For the concentric rings, the model will generate a
logarithmic progression of distances starting at the inner ring radium and ending at the outer
radium of the modeling domain.
Elevations and hill heights for model receptors
HEM-3 includes terrain elevations by default for the RTR assessments, but the user can
choose to exclude terrain effects when running AERMOD. If the default terrain option is used,
HEM-3 obtains elevations and controlling hill heights for Census block receptors from its
internal Census location library. Section 2.4.2 describes the derivation of these elevations and
hill heights.
Elevations and controlling hill heights for the polar grid receptors are also estimated
based on values from the Census library. HEM-3 divides the modeling domain into sectors based
on the polar receptor network, with each Census block assigned to the sector corresponding to
the closest polar grid receptor. Each polar grid receptor is then assigned an elevation based on
the highest elevation for any Census block in its sector. The controlling hill height is also set to
the maximum hill height within the sector. If a sector does not contain any blocks, the model
defaults to the elevation and controlling hill height of the nearest block outside the sector.
2.1.3 Selection of Meteorological Data
In addition to source and receptor information, AERMOD requires surface and upper air
meteorological observations in a prescribed format. The model user can select a meteorological
station from the HEM-3 meteorological data library, or add new files to the library if site-specific
data are available. If the user does not specify a meteorological station, HEM-3 will select the
closest station to the center of the modeling domain, as is generally done for the RTR
assessments.
2.2 Running of AERMOD
Based on the user input data and other data described in the previous section, HEM-3
produces an input file suitable for AERMOD. HEM-3 then runs AERMOD as a compiled
executable program. No changes have been made from the version of AERMOD released to the
public by EPA. The following sections give additional information on how AERMOD is used
within HEM-3.
2.2.1 AERMOD Dispersion Options Used by HEM-3
AERMOD provides a wide array of options for controlling dispersion modeling
calculations. In general, HEM-3 uses the regulatory default options when running AERMOD.1
These options include the following:
2-11
-------
• Use stack-tip downwash (except for Schulman-Scire downwash);
• Use buoyancy-induced dispersion (except for Schulman-Scire downwash);
• Do not use gradual plume rise (except for building downwash);
• Use the "calms processing" routines;
• Use upper-bound concentration estimates for sources influenced by building downwash
from super-squat buildings;
• Use default wind profile exponents;
• Use low wind speed threshold;
• Use default vertical potential temperature gradients;
• Use of missing-data processing routines; and
• Consider terrain effects.
The following additional AERMOD options are available to the HEM-3 user:
• Calculation of wet and dry deposition rates for gaseous and particulate pollutants;
• Consideration of plume depletion (due to deposition) when calculating air concentrations;
• Consideration of building wake effects;
• Calculation of short term (acute) impacts;
• Use of the FASTALL option, which conserves model runtime by simplifying the
AERMOD algorithms used to represent meander of the pollutant plume; and
• Use of the buoyant line plume option.
As noted in Section 2.1, the calculation of deposition or depletion and the consideration of
building wake effects require additional user inputs.
The user can opt to analyze short term impacts on a number of different time scales (i.e.,
1 hour, 2 hours, 3 hours, 4 hours, 6 hours, 8 hours, 12 hours, or 24 hours) however only one short
term time scale can be selected per run. If the user chooses to analyze short term (acute) impacts,
a multiplier must be specified to reflect the ratio between the maximum short term emission rate
and the long term average emission rate. If available, acute multipliers specific to source
classification codes (SCCs) are used in RTR assessments. If SCC-specific acute multipliers are
not available, the default multiplier for short term emissions is a factor of 10. This means that in
the default case the maximum short term emission rate is assumed to be 10 times the long term
average emission rate. The multiplier can be set to one (1) if emissions from the facility are
known to be constant. For RTR assessments, acute impacts are generally included in the
modeling and the default multiplier of 10 is used, unless more source-specific information is
available upon which to base the acute factor for the source category being modeled.
2-12
-------
2.2.2 Use of Dilution Factors
To save computer run time when analyzing the impacts of multiple pollutants, HEM-3
does not model each pollutant separately. Instead, AERMOD is used to compute a series of
dilution factors, specific to each emission source and receptor. The dilution factor for a particular
emission source and receptor is defined as the predicted ambient impact from the given source
and at the given receptor, divided by the emission rate from the given source.
If the user chooses not to analyze deposition (with or without plume depletion), the
dilution factor does not vary from pollutant to pollutant. If deposition and/or depletion is chosen
as a model option, separate dilution coefficients must be computed for each gaseous pollutant. In
addition, separate dilution factors must be computed for different components of particulate
matter if the components do not have the same particle size distribution. In the current version of
HEM-3, this can be done by creating a separate emission record for each pollutant emitted by
from each source. (Common location data and source configurations can be used for different
pollutant records representing the same emission source.)
2.3 Postprocessing of AERMOD Results in HEM-3
HEM-3 estimates total excess cancer risks and potential chronic non-cancer health effects
for all Census block locations in the modeling domain, all user-defined receptors, and all points
in the polar receptor network. Potential chronic non-cancer health effects are expressed in terms
of TOSHI. Based on the results for Census blocks and other receptors, HEM-3 estimates the
maximum individual risk (MIR) and maximum TOSHI for populated receptors, and determines
the locations of these maximum impacts. The model also determines the concentrations of
different pollutants at the site(s) of maximum risk and maximum TOSHI, and the contributions
of different emission sources to these locations of maximum impact. It should be noted that the
locations of maximum impact may differ for the maximum individual cancer risk and for the
hazard indices of different target organs.
For acute impacts, HEM-3 calculates the 99th percentile maximum short term
concentrations for all pollutants emitted by the facility. These short term concentrations are
compared with various threshold levels for acute health effects (e.g., the California EPA
reference exposure level [REL] for no adverse effects).
At the option of the model user, HEM-3 will also compute the long term and short term
predicted ambient concentrations of all pollutants emitted by the facility at all of the receptors in
the modeling domain. In addition, pollutant contributions from each emission source at the
facility are computed under this option. In RTR assessments, this option is standard and
concentrations are computed for all receptors.
Section 2.3.1 describes methods used to calculate cancer risks and hazard indices for
receptors that are explicitly modeled using AERMOD. Section 2.3.2 describes the interpolation
approach used to estimate cancer risks and hazard indices at Census blocks that are not explicitly
modeled.
2-13
-------
2.3.1 Calculation of Impacts at Modeled Receptors
As noted in Section 2.2.2, HEM-3 does not model each pollutant separately unless
deposition or depletion is being analyzed. Instead, AERMOD is used to compute a series of
dilution factors, specific to each emission source and receptor. The following algorithms are used
to compute cancer risks and TOSHI for chronic non-cancer health effects.
For cancer risk:
For TOSHI:
where:
CRt =
I» =
CRi,j =
DF„ =
CF =
2> =
Ei,k =
UREk =
TOSHIx =
TOSHIij =
RfCk =
CRt Xij CRij
CR, , = DFij x CF x £k [E,.k x UREk]
TOSHIx = £ij TOSHIy
TOSHIy = DFij x CF x £k [Ei.k / RfCk]
total cancer risk at a given receptor (probability for one person)
the sum over all sources i and pollutant types j (particulate or gas)
cancer risk at the given receptor for source i and pollutant type j
dilution factor [(|ig/m3) / (g/sec)] at the given receptor for source i and
pollutant type j
conversion factor, 0.02877 [(g/sec) / (ton/year)]
sum over all pollutants k within pollutant type j (particulate or gas)
emissions of pollutant k from source i and in pollutant type j
cancer unit risk factor for pollutant k
total target-organ-specific hazard index at a given receptor
target-organ-specific hazard index at the given receptor for source i and
pollutant type j
non-cancer health effect reference concentration for pollutant k
The above equations are equivalent to the following simpler equations:
CRt = £i,k ACi,k x UREk
TOSHIx = £i,k AC, k/RCk
where:
ACi,k = ambient concentration (|ig/m3) for pollutant k at the given receptor. This is the same
as [E;,k x DFi j x CF]
However, use of these simpler equations would require modeling all pollutants individually in
AERMOD, and performing separate risk calculations for each pollutant.
2-14
-------
If the cancer unit risk estimate is not available for a given chemical, then that chemical is
not included in the calculation of cancer risk. Likewise, if the non-cancer reference concentration
is not available for a given chemical, that chemical is not included in the calculation of hazard
indices. Note also that separate reference concentrations are used for acute and chronic hazard
indices.
HEM-3 computes short term concentrations and records the highest short term
concentration for each pollutant. In addition, the user can opt to compute and record the short
term and long concentrations at each receptor. Concentrations are computed as follows.
Long term concentrations:
ACT)k = ZiACg£
ACi,k = Ei,kxDFij x CF
Short term concentrations:
ACx = £iACi,k
ACi,k = Ei,kxDFij x CF x M
where:
ACtj< = total estimated ambient concentration for pollutant k at a given receptor
£i = the sum over all sources i (|ig/m3)
ACi,k= estimated ambient concentration of pollutant k at the given receptor as a result
of emissions from source i (|ig/m3)
M = ratio between the estimated maximum short term emission rate and the long
term average emission rate (dimensionless)
2.3.2 Interpolation of Impacts at Outer Census Blocks
For Census blocks outside of the cutoff distance for individual block modeling, HEM-3
estimates cancer risks and hazard indices by interpolation from the polar receptor network.
HEM-3 estimates impacts at the polar grid receptors using AERMOD modeling results and the
algorithms described in Section 2.3.1. If terrain elevation is part of the modeling, then an
elevation is estimated for each polar receptor. HEM-3 estimates elevations and controlling hill
heights for the polar grid receptors based on values from the census library. HEM-3 divides the
modeling domain into sectors based on the polar grid receptor network, with each census block
assigned to the sector corresponding to the closest polar grid receptor.
HEM-3 then assigns each polar grid receptor an elevation based on the highest elevation
for any census block in its sector. The controlling hill height is also set to the maximum hill
height within the sector. If a sector does not contain any blocks, the model defaults to the
elevation and controlling hill height of the nearest block outside the sector.
HEM-3 interpolates the impacts at each outer Census block from the four nearest polar
grid receptors. The interpolation is linear in the angular direction, and logarithmic in the radial
direction, as summarized in the following equations:
2-15
-------
Ia,r — lAl,r + (lA2,r — lAl,r) x (a — Al) / (A2 — Al)
lAi,r= exp{(ln (Iai,ri) + [(In (Iai,r2) - In (Iai,ri)] x [(In r) - ln(Rl)] / [ln(R2) - ln(Rl)]}
lA2,r= exp{(ln (I \2.ri ) + [(In (L\2.r2) - In (Ia2,ri)] x [(In r) - ln(Rl)] / [ln(R2) - ln(Rl)]}
where:
la,r= the impact (cancer risk, hazard index, or concentration) at an angle, a, from
north, and radius, r, from the center of the modeling domain
a = the angle of the target receptor, from north
r = the radius of the target receptor, from the center of the modeling domain
Al = the angle of the polar network receptors immediately counterclockwise from
the target receptor
A2 = the angle of the polar network receptors immediately clockwise from the target
receptor
R1 = the radius of the polar network receptors immediately inside the target receptor
R2 = the radius of the polar network receptors immediately outside the target
receptor
2.3.3 Calculation of Population Exposures and Incidence
Using the predicted impacts for Census blocks, HEM-3 estimates the numbers of people
exposed to various cancer risk levels and TOSHI levels. This is done by adding up the
populations for receptors that have predicted cancer risks or TOSHI above the given threshold.
The model also estimates the annual excess cancer risk (incidence) for the entire
modeling region. The following equation is used:
TCR = Im [CRm x Pm ] / LT
where:
TCR = the estimated annual cancer incidence (excess cancers/year) to the population
living within the modeling domain
Im = the sum over all Census blocks m within distance the modeling domain
CRm = the total lifetime cancer risk (from all modeled pollutants and emission sources)
at Census block m
Pm — the population at Census block m
LT = the average lifetime used to develop the cancer unit risk factor, 70 years
HEM-3 also estimates the contributions of different chemicals and emission sources to total
annual cancer incidence for the overall modeling domain using the following equations:
TCRij = [(Ik Ei,k x UREk) x DFy,m x CF / LT]
TCRi,k = TCRij x E,.k x UREk / (Ik E,.k x UREk)
where:
TCRij = the estimated total annual cancer incidence (cancers/year) to the population in the
modeling domain due to emissions from pollutant type j (1 = particulate, 2 = gas)
and emission source i
2-16
-------
Xm = the sum over all Census blocks m within distance the modeling domain
Xk = the sum over all pollutant k, within pollutant type j
Ei.k = emissions of pollutant k from source i (tons/year)
UREk = unit risk factor for pollutant k
DFi^m = dilution factor at receptor m, for emissions of pollutant type j (which includes
pollutant k), from source i
CF = conversion factor, 0.02877 [(g/sec) / (tons/year)]
TCRi.k = the estimated annual cancer incidence (cancers/year) of the population in the
modeling domain due to emissions of pollutant k (in pollutant type j) from
emission source i
2.3.4 Model Outputs
The following is a summary of the outputs produced by HEM-3. These are written to a
collection of files in Excel™ and dBase™ format (dbf).
• Long term impacts at populated locations
o maximum long term ambient concentration for each chemical
o maximum lifetime individual cancer risk (MIR)
o maximum TOSHI for the following health effects
- respiratory system effects
- liver effects
- neurological system effects
- developmental effects
- reproductive system effects
- kidney effects
- ocular system effects
- endocrine system effects
- hematological system effects
- immunological system effects
- skeletal system effects
- spleen effects
- thyroid effects
- whole body effects
o locations of the maximum cancer risk and maximum TOSHIs
o Census block identification codes for the maximum concentration, maximum cancer
risk and maximum TOSHIs, and number of people in the Census block
o contributions of different chemicals and emission sources to the maximum risk and
TOSHI
• Acute impacts
o 99th percentile maximum short term ambient concentration for each chemical
2-17
-------
o threshold levels for acute health effects of each chemical (compared with the 99th
percentile maximum short term concentrations)
o locations of the 99th percentile maximum impacts for different chemicals (often
polar receptors)
o Census block identification codes at the locations of 99th percentile maximum
concentration, and number of people in the block
o contribution of each emission source at the facility to the 99th percentile
maximum short term concentration of each chemical
• Outputs for all receptors
o maximum individual cancer risk and TOSHI (all target organs) for each Census
block and each user-specified discrete receptor (monitoring sites, etc.)
o maximum individual cancer risk and TOSHI (all target organs) for each polar grid
receptor
o estimated deposition flux (optional)
o predicted ambient concentration resulting from each emission source at each
Census block and polar grid receptor (optional)
• Population exposures and total cancer risk, or incidence
o estimated numbers of people exposed to different levels of lifetime individual
cancer risk (1 in a million, 1 in 100,000, etc.)
o estimated numbers of people exposed to different levels of TOSHI (1, 2, 10, etc.)
o total cancer risk, or incidence, in estimated cancer deaths per year, over the entire
modeling domain, and for each pollutant and source combination
2.4 Data Libraries Used in HEM-3
2.4.1 Chemical Health Effects Information
HEM-3 includes a library of available health effects data for HAPs. For each pollutant,
the library includes the following parameters, where available:
2-18
-------
• unit risk estimate (URE) for cancer;
• reference concentration (RfC) for chronic non-cancer health effects;
• reference benchmark concentrations for acute health effects; and
• target organs affected by the chemical for chronic non-cancer health effects.
Unit risk estimates and reference concentrations included in the HEM-3 chemical library have
been taken from EPA's database of recommended dose-response factors for HAPs, which is
updated periodically, consistent with continued research on these parameters.4 The URE
represents the upper-bound excess lifetime cancer risk estimated to result from continuous
exposure to an agent (HAP) at a concentration of 1 microgram per cubic meter ([j,g/m3) in air
(e.g., if the URE = 1.5 x 10"6 per (j,g/m3, then 1.5 excess tumors are expected to develop per 1
million people if all 1 million people were exposed daily for a lifetime to 1 microgram of the
chemical in 1 cubic meter of air). UREs are considered plausible upper limits to the true value;
the true risk is likely to be less but could be greater.5
The RfC is a concentration estimate of a continuous inhalation exposure to the human
population that is likely to be without an appreciable "risk" of deleterious non-cancer health
effects during a lifetime (including to sensitive subgroups such as children, asthmatics and the
elderly). No adverse effects are expected as a result of exposure if the ratio of the potential
exposure concentration to the RfC, defined as the hazard quotient (HQ), is less than one. Note
that the uncertainty of the RfC estimates can span an order of magnitude.5 Target organs are
those organs (e.g., kidney) or organ systems (e.g., respiratory) which may be impacted with
chronic non-cancer health effects by exposure to the chemical in question. The hazard index (HI)
is the sum of HQs for substances that affect the same target organ or organ system, also known
as the target organ specific hazard index (TOSHI).
The reference benchmark concentration for acute health effects, similar to the chronic
RfC, is the concentration below which no adverse health effects are anticipated when an
individual is exposed to the benchmark concentration for 1 hour (or 8 hours, depending on the
specific acute benchmark used and the formulation of that benchmark). A more in-depth
discussion of the development and use of these parameters for estimating cancer risk and non-
cancer hazard may be found in the EPA's Air Toxics Risk Assessment Library.6
The model user can add pollutants and associated health effects to HEM-3 's chemical
health effects (dose-response and target organ endpoints) library, as needed.
2.4.2 Census Block Locations and Elevation Data
The HEM-3 Census library includes Census block identification codes, locations,
populations, elevations, and controlling hill heights for all of the over 6 million Census blocks
identified in the 2010 Census and the over 5 million Census blocks identified in the 2000
Census. The model user may choose to use either Census database according to their modeling
needs. The location coordinates reflect the internal "centroid" of the block, which is a point
2-19
-------
selected by the Census to be roughly in the center of the block. For complex shapes, the internal
point may not be in the geographic center of the block. Locations and population data for Census
blocks in the 50 states and Puerto Rico were extracted from the LandView® database For the
2000 Census7 and from the U.S. Census Bureau website for the 2010 Census.8 Locations and
populations for blocks in the Virgin Islands were obtained from the U.S. Census Bureau website.
U.S. Geological Survey data was used to estimate the elevation of each census block in
the continental U.S. and Hawaii. The data used for the 2000 Census elevations have a resolution
of 3 arc-seconds, or about 90 meters.9 The data used for the 2010 Census elevations have a
resolution of 1/3 of an arc second, or about 10 meters.10 Using analysis tools (ArcGIS® 9.1
software application for the 2000 Census, and ArcGIS® 10 for the 2010 Census), elevation was
estimated for each census block in Alaska and the U.S. Virgin Islands. The point locations of the
census blocks in Alaska and the U.S. Virgin Islands were overlaid with a raster layer of North
American Digital Elevation Model (DEM) elevations (in meters).9 An elevation value was
assigned to each census block point based on the closest point in the ArcGIS elevation raster file.
An algorithm used in AERMAP, the AERMOD terrain processor, is used to determine
controlling hill heights.11'12 These values are used for flow calculations within AERMOD.
To save run time and resources, the HEM-3 census block elevation database is substituted for the
DEM data generally used in AERMAP. As noted above, the census block elevations were
originally derived from the DEM database. To determine the controlling hill height for each
census block, a cone is projected away from the block centroid location, representing a 10%
elevation grade. The controlling hill height is selected based on the highest elevation above that
10% grade (in accordance with the AERMAP methodology). The distance cutoff for this
calculation is 100 km. (This corresponds to an elevation difference at a 10% grade of 10,000 m,
which considerably exceeds the maximum elevation difference in North America.)
2.4.3 Meteorological Data
HEM-3 includes an extensive library of meteorological data to support the AERMOD
dispersion model. Currently over 800 meteorological stations have been preprocessed for
AERMOD as part of the RTR effort. Section 3.3 includes a depiction of these meteorological
stations and Appendix 3 discusses the preparation of meteorological data for the RTR in more
detail.
2.4.4 Gaseous Deposition Parameters
HEM-3 provides options to compute the deposition of air pollutants, and to take into
account the impacts of plume depletion due to deposition of gaseous and particulate pollutants. If
the deposition and depletion option is selected by the model user for gaseous pollutants, a
number of pollutant properties are required by AERMOD. (These include the diffusivity of the
pollutant in air, the diffusivity of the pollutant in water, the Henry's Law constant, and a
parameter reflecting the cuticular resistance to uptake of the pollutant by leaves rci).13 HEM-3
includes a library of these parameters for approximately 130 gaseous HAPs. This library is based
on a compendium of gaseous deposition parameters developed by Argonne National
Laboratories.14 The HEM-3 user can edit these values, if appropriate, including adding additional
2-20
-------
pollutant values available in the literature or calculated based on recommended methodology, as
discussed in the Single HEM-3 User's Guide.3 It should be noted, however, that the deposition
and depletion option of HEM-3 and AERMOD have not been used to date for the RTR
assessments.
3. Modeling for the Residual Risk Technology Review
This section discusses the general approach used to implement the HEM-3 AERMOD
system for the RTR modeling analyses. Separate reports have been prepared for each of the
emission source categories analyzed to date. These reports provide information on the emissions
inputs and results for specific emission categories.
3.1 Emission Source Inputs
HEM-3 and AERMOD require detailed data on emissions from each emission source
included in the modeling analysis. These data include, for example:
• pollutants emitted;
• emission rate for each pollutant;
• emission source coordinates;
• stack height (or emission height for fugitive and other area sources);
• stack diameter (or configuration of fugitive and other area sources);
• emission velocity; and
• emission temperature.
Emissions data for the RTR assessments are compiled from a variety of data sources
(e.g., the National Emissions Inventory (NEI)15, information collection requests). Each source
category evaluated under the RTR program utilizes the best available data. These data include
HAP emission rates, emission source coordinates, stack heights, stack diameters, flow rates, exit
temperatures, and other emission parameters depending on the emission source types modeled.
EPA performs an engineering review of the NEI data. In cases where new or better data are
known to exist for a particular source category, that information is integrated into the data used
in modeling that category. For each source category, the emissions are summarized in the source
category specific report. Detailed computer files containing all emission and release
characteristics are available in the docket prepared for the specific RTR source category under
proposed or final rulemaking.
As noted in the previous section, industrial emission sources can be characterized in
AERMOD as point (vertical, capped and horizontal), area, polygon, volume, line or buoyant line
2-21
-------
sources. Fugitive emissions are generally characterized as low point sources with minimal exit
velocities. For some categories, additional information is available on the configuration of
fugitive emission sources. This information is incorporated into the emissions database as part of
the engineering review. For example, fugitive emission sources are characterized as area or
volume sources when sufficient configuration information is available.
3.2 Pollutant Cross-Referencing
Because the NEI is developed from a number of different data sources, a single chemical
may be listed in the inventory under different names (i.e., a "common name" and one or more
structure-based names). In addition, pollutant groupings such as polycyclic organic matter
(POM), can be listed in the NEI under the names of individual member compounds, and under
different synonyms (e.g. polynuclear aromatic hydrocarbons). HEM-3 requires an exact match
with the chemical name in order to link emissions to the appropriate dose-response factors. The
model will not process any pollutant that is not specifically listed in the chemical library.
Therefore, all of the HAP names used in the NEI are linked to the appropriate chemical names in
the HEM-3 reference file.
Pollutant-specific dose response values are used in the HEM-3 modeling whenever
available, including when modeling POM pollutants and metal compounds. Pollutant groupings,
such as POM groupings, are used for POMs without a chemical-specific unit URE's. These
POMs are assigned a URE associated with various POM compounds having similar
characteristics. The "Technical Support Document - EPA's 2011 National-scale Air Toxics
Assessment" 2015 document16 provides more details regarding POM modeling, including (p.
121):
[S]ome emissions of POM were reported in [the] NEI as "7-PAH" or "16-PAH,"
representing subsets of certain POM, or simply as "total PAH" or "polycyclic organic
matter." In other cases, individual POM compounds are reported for which no
quantitative cancer dose-response value has been published in the sources used for
NATA. As a result, simplifying assumptions that characterize emissions reported as POM
are applied so that cancer risk can be quantitatively evaluated for these chemicals without
substantially under- or overestimating risk (which can occur if all reported emissions of
POM are assigned the same URE). To accomplish this, POM emissions as reported in
NEI were grouped into categories. EPA assigns dose-response values based on the known
or estimated toxicity for POM within each group and on information for the POM
speciation of emission sources, such as wood fires and industrial processes involving
combustion.
Toxicity values used for metal compounds are also discussed in EPA's 2011 National Air Toxics
Assessment Technical Support Document, including the treatment of chromium (VI)
compounds, lead and nickel compounds.16
3.3 Meteorological Data
Nationwide meteorological data files are accessed by HEM-3 and used for the RTR
modeling. The current HEM-3 AERMOD Meteorological Library includes over 800 nationwide
2-22
-------
locations, depicted in Figure 3-1. This library contains surface and upper air 2016 meteorological
data from National Weather Service (NWS) observation stations, which span the entire U.S. as
well as Puerto Rico and the U.S. Virgin Islands. AERMOD requires surface and upper air
meteorological data that meet specific format requirements.17'18 Appendix 3 discusses the
preprocessing performed on the meteorological data used by AERMOD and includes a detailed
listing of the 824 meteorological surface and upper air station pairs, including coordinates,
ground elevation and anemometer height for each station.
2-23
-------
Figure 3-1.
PR and VI
2-24
-------
3.4 Model Options Selected
HEM-3 presents a number of options for characterizing the modeling domain and data
sources. As many sources are generally modeled in RTR assessments, established defaults and
common practices are relied on to make these choices. The choices available to a HEM-3 user
and the selections that are made in most RTR assessments are presented in Table 3-1. Some of
the key selections are discussed in more detail in the paragraphs below.
It should be noted that although routine emissions are not expected to vary significantly
with time, nonroutine (upset) emissions can be significant relative to routine emissions. Upset
emissions occur during periods of startup, shutdown, and malfunction. Upset emissions are not
likely for equipment or storage tanks, but do result from malfunctioning control devices and
leaks in cooling tower heat exchangers. There is some limited data on upset emissions
available,19 but no facility-specific analyses of these data were performed to characterize short-
term emissions from these emission sources, and upset emissions are generally not modeled for
the RTR risk assessments.
3.4.1 Urban or Rural Dispersion Characteristics
Current RTR source category assessments which use the 2010 Census are based on either
urban or rural dispersion characteristics, depending on the land characteristics surrounding each
modeled facility. The EPA provides guidance on whether to select urban or rural dispersion
coefficients in its Guideline on Air Quality Models.20 In general, the urban option is used if (1)
the land use is classified as urban for more than 50% of the land within a 3-kilometer radius of
the emission source, or (2) the population density within a 3-kilometer radius is greater than 750
people per square kilometer. Of these two criteria, the land use criterion is more definitive.
Using the 2010 Census, the HEM-3 model determines, by default, whether to use rural or
urban dispersion characteristics. HEM-3 will find the nearest census block to the facility center
and determine whether that census block is in an urban area, as designated by the 2010 Census.21
The population of the designated urban area will be used to specify the population input for
AERMOD's urban mode. (Alternatively, a user may select the rural or urban option to override
determination by the model. If a user selects an urban dispersion environment, then the user must
provide the urban population as well.)
For the 2008 and prior screening-level RTR assessments of 51 source categories, the rural
option was chosen to be most conservative (i.e., more likely to overestimate risk results). The
rural option is also chosen by default by the HEM-3 model whenever the 2000 Census is selected
by the user.
3.4.2 Deposition and Plume Depletion
The RTR modeling analysis to date has not taken into account the depletion of pollutant
concentrations in the plume due to wet or dry deposition, although HEM-3 can model deposition
with or without depletion using AERMOD. In addition, reactivity and decay have not been
considered. It is possible that this approach may overestimate air concentrations and therefore
2-25
-------
risk. However, one of the main metrics used by EPA in the residual risk program is the risk to
the individual most exposed (the maximum individual risk, or MIR). Because the maximum risk
usually occurs at a receptor very close to the emission source, it is unlikely to be influenced by
altered plume dispersion characteristics of this type. For more refined, multipathway
assessments, EPA may consider deposition and depletion.
3.4.3 Cutoff Distance for Modeling of Individual Blocks
The cutoff distance for modeling individual Census blocks is initially set to 3 km by
default. This distance generally ensures that the maximum individual cancer risk and the
maximum TOSHI are modeled explicitly and not interpolated. Following a modeling run, the
results for each facility are checked to determine whether the maximum impacts are located
inside the modeling cutoff distance. If the maximum impacts are outside the cutoff distance, and
if any of the impacts are significant, then HEM-3 is rerun for the facility with a cutoff distance
greater than 3 km. In general, this is done if the cancer risk exceeds 1 in 1 million or any TOSHI
exceeds one. However, the risks for such facilities are generally very low, since the maximum
impacts are in most cases only interpolated when the nearest Census block is more than 3 km
from the facility (i.e., in sparsely populated areas).
3.4.4 Facility Boundary Assumptions
The main input mechanisms for incorporating facility boundary information in HEM-3
are the overlap distance, the distance to the innermost polar receptor ring, and user-specified
receptor locations. The NEI does not provide information on facility boundaries. However,
satellite/aerial images are used to locate residential populations that are closer to a facility than
the Census block centroid. User-specified receptor locations are used in such assessments to
avoid underestimating risk. Conservative default assumptions are used for the overlap distance
and the innermost polar receptor ring. However, these are adjusted for some categories where
facility sites are known to be large. In addition, satellite imagery is used to check the facility
boundary assumptions for facilities with large projected impacts. These checks are discussed
further in the section on Quality Assurance (Section 4).
2-26
-------
Table 3-1. HEM-3 Domain and Set-Up Options As Used in the Residual Risk and
Technology Review Assessments
Option
Selection
Dispersion model
AERMOD
Census database: 2010 or 2000
2010, unless
retrospective analysis
Type of analysis: chronic, acute, or both
Both
Averaging time for short term impacts
1-hour
Multiplier for short term emissions
Source type-specific
factors are used if
available; a factor of
10 used otherwise
Dispersion characteristics: urban or rural, as determined by model,
based on closest 2010 Census block to each facility (when using
2010 Census). Rural by default, when using the 2000 Census.
Urban or Rural based
on facility location;
Include terrain impacts
Yes
Include building wake effects
No
Calculate deposition (wet, dry, or both) & include impacts of plume
depletion
Nod
User-specified receptor locations (for residential population
locations, facility boundary sites, or other sites of interest)
Yes, for some
facilities
Modeling domain size - maximum distance to be modeled
50 km
Cutoff distance for modeling of individual blocks
3 kma
Overlap distance where receptors are considered to be on facility
property - measured from each source measured from each source
30 mb
Polar receptor network specifications:
Distance from the facility center to the innermost ring
> 100 m°
Number of rings
13
Number of directions
16
Meteorology data
Closest site
a The individual block modeling cutoff is increased for categories and for some facilities to ensure that the
maximum individual risk values are not interpolated.
b The overlap distance is adjusted for some facilities to avoid modeling locations that are on facility property
(see section 4.2).
0 HEM-3 sets the innermost ring distance to be just outside the emission sources but not < 100 m.
d RTR assessments typically do not calculate deposition and/or depletion, although the option to use AERMOD
to model deposition with or without depletion is available in HEM-3.
2-27
-------
3.5 Modeling of Multiple Facilities
HEM-3 models one facility at a time. However, clusters of nearby facilities may impact
the same people, resulting in higher risk to those people. To account for this situation, risks are
summed at each Census block for all facilities affecting the Census block.
As described earlier (Section 2.3.4), HEM-3 produces detailed output tables containing
the risk and population for every Census block in the modeling domain. These detailed tables are
combined for all facilities in a source category and the risk for each Census block is summed,
using the RTR Summary Program add-on module to the Multi HEM-3 model, as described in the
Multi HEM-3 User's Guide.3 Thus, the effect of multiple facilities in the same source category
on the same receptor are estimated. The resulting "combined facility" or "cluster-effect" census
block risks are used to calculate population exposure to different cancer risk levels, non-cancer
hazard indices, and source category incidence.
2-28
-------
4. Quality Assurance
The National Emissions Inventory (NEI) is subject to an extensive program of quality
assurance (QA) and quality control (QC). The QA/QC program for the point source component
of the NEI is documented in a separate report, available from the NEI website.22 This section
describes QA activities carried out under the RTR modeling analysis.
4.1 Engineering Review
In addition to the standardized QA steps taken for the entire NEI, EPA performs an
engineering review of NEI data for the emission source categories included in the RTR analysis.
This engineering review includes two main components. The first component addresses the list
of facilities included in each source category. EPA engineers review independent sources of
information to identify all sources in the category that are included in the NEI. In addition, EPA
reviews the list of sources represented as part of each category in the NEI to ensure that the
facilities actually manufacture products characteristic of the source category.
The second component of the engineering review focuses on the appropriateness of
facility emissions. EPA reviews the list of HAPs reportedly emitted by each facility to ensure
that the pollutants are appropriate to the source category. In addition, EPA engineers review the
magnitude of those HAP emissions. In cases where new or better data are known to exist for a
particular source category, that information is integrated into the data used in the HEM-
3/AERMOD modeling for that category. In these cases, the source category specific documents
provide additional details on the emissions inputs used.
4.2 Geographic Pre-Modeling Checks
The NEI QA process includes some basic checks on location data for point sources. The
coordinates for each source are checked to ensure that they are in the county that has been
specified for the source. If this is not the case, or if no geographic coordinates are available for
the emission source, then the coordinates are set to a default location based on the nature of the
emission source category.22 In addition, coordinates for all emission sources at a given facility
are checked to ensure that they are within 3 km of one another. These QA checks happen prior to
HEM-3 modeling and the results of such checks are reflected in the HEM-3 input files.
Another pre-modeling geographic QA check regards the location of the census block
receptors. As noted above, to estimate ambient concentrations for evaluating long-term
exposures, the HEM-3 model uses the census block centroids as dispersion model receptors. The
census block centroids are often good surrogates for where people live within a census block. A
census block generally encompasses about 40 people or 10-15 households. However, in cases
where a block centroid is located on industrial facility property, or where a census block is large
and the centroid less likely to be representative of the block's residential locations, the block
centroid may not be an appropriate surrogate.
2-29
-------
Census block centroids that are on facility property can sometimes be identified by their
proximity to emission sources. In cases where a census block centroid is within 300 meters of
any emission source, aerial images of the facility are reviewed to determine whether the block
centroid is likely located on facility property. The selection of the 300-meter distance reflects a
compromise between too few and too many blocks identified as being potentially on facility
property. Distances smaller than 300 meters would identify only block centroids very near the
emission sources and could exclude some block centroids that are still within facility boundaries,
particularly for large facilities. Distances significantly larger than 300 meters would identify
many block centroids that are outside facility boundaries, particularly for small facilities. Block
centroids confirmed to be located on facility property are moved to a location that best represents
the residential locations in the block.
In addition, census block centroids for blocks with large areas may not be representative
of residential locations. Risk estimates based on such centroids can be understated if there are
residences nearer to a facility than the centroid, and overstated if the residences are farther from
the facility than the centroid. To avoid understating the maximum individual risk associated with
a facility, block centroids are relocated in some cases, or additional user-specified receptors are
added to a block. Aerial images of all large census blocks within one kilometer of any emission
source are examined. Experience from previous risks characterizations show that in most cases
the MIR is generally located within 1 km of the facility boundary. If the block centroid does not
represent the residential locations, it is relocated in the HEM-3 input files to better represent
them. If residential locations cannot be represented by a single receptor (that is, the residences
are spread out over the block), additional user-specified receptors are included in the HEM-3
input files to represent residences nearer to the facility than the centroid.
4.3 Geographic Post-Modeling Checks
As part of the RTR modeling analysis, additional geographical QA checks are made for
some facilities, after initial HEM-3 modeling results are reviewed. Facilities subjected to these
additional checks include:
• cases where the initial estimates of maximum risks are particularly high
o maximum individual cancer risk of over 1 in 10,000
o any maximum TO SHI above 10
• cases where no Census blocks are identified by the model within 3 km of the facility
HEM-3 produces a detailed Google Earth™ map of the modeled point, area, polygon,
volume, line and buoyant line emission sources and surrounding receptors (including Census
block centroids, polar receptors and user-specified receptors) overlaying Google Earth™'s
satellite imagery. This map allows a QA check of the specific source locations, as well as an
approximate check of the facility boundaries. The emission source coordinates are reviewed for
each of these facilities and compared with the address reported for the facility. If the address and
the coordinates represent the same location, then the coordinates are taken to be correct. For
2-30
-------
more recent modeling of source categories, the emission coordinates initially modeled by HEM-
3 tend to be correct, as they undergo pre-modeling scrutiny and QA checks (as discussed in
Section 4.2).
More rarely, the modeled emission coordinates will be determined post initial modeling
not to be located on facility property. If the facility and emission coordinate locations are
different, then the satellite imagery for the address and the coordinate location are reviewed to
determine whether either photograph includes an industrial facility. If emission source
coordinates are found to be incorrect, HEM-3 is rerun using corrected coordinates. These
changes are described in the source category documents.
For the high-risk facilities, the coordinates used to represent the most impacted Census
blocks are also reviewed. This review draws on detailed Census block boundary maps and
satellite imagery. Large industrial facilities will frequently occupy one or more entire Census
blocks. However, these blocks may also include one or more residences on the periphery of the
industrial land. Generally, the centroid coordinates listed for a Census block are near the center
of the block. In these cases of mixed industrial and residential blocks, the coordinates may be on
facility property.
In general, block coordinates are considered to be on facility property if they are located
between the different emission source locations listed for the facility. In these situations, HEM-3
is rerun with an expanded overlap distance, in order to exclude the Census block coordinates that
appear to be located on facility property. The distance to the innermost polar receptor ring is also
adjusted to ensure that this ring is not on facility property, but as close to the apparent facility
boundaries as possible.
2-31
-------
5. Uncertainties
The RTR risk assessments using HEM-3 and AERMOD are subject to a number of
uncertainties. For instance, model verification studies for AERMOD show predicted maximum
annual concentrations ranging from 0.3 to 1.6 times measured values, with an average of 0.9.
Predicted maximum short term (1 to 24 hours) concentrations were 0.25 to 2.5 times measured
values, with an average of one.23
In addition, a number of simplifying assumptions are made in these modeling analyses.
First, the coordinates reported by the U.S. Census Bureau for Census block internal points
("centroids") have been used as a surrogate for long-term population exposures. Locations of
actual residences have not been modeled. In addition, the current version of HEM-3 does not
take into account the movement of people from one Census block to another during the course of
their lives, or commuting patterns during a given day. Nor does the model take into account the
attenuation of pollutant from outside emission sources in indoor air. Ideally, risks to individuals
would be modeled as they move through their communities and undertake different activities.
However, such modeling is time-and resource-intensive and can only capture a portion of the
uncertainty associated with the full range of human activities. In general, it is expected that long-
term exposures will be overstated for high-end estimates (as most individuals will not spend all
their time at their highly affected residences), but may understate the total population exposed
(as some individuals living outside the modeled area may regularly commute into the area for
work or school).
When considering long-term or lifetime exposures, it should be noted that relatively few
people in the United States reside in one place for their entire lives. For the purposes of this
assessment, cancer risk estimates are based on a lifetime exposure at the Census-identified place
of residence. While it is impossible to know how this assumption affects the risk experiences by
a particular individual (as people can move into higher- or lower-risk areas), it is likely that this
assumption will overstate the exposure to those most exposed (i.e., people already living in high
exposure areas are unlikely to move to yet higher exposure areas). However, this assumption will
also tend to underestimate the total number of people exposed and population risk (i.e.,
incidence) because population levels are generally increasing.
In the current analyses, only direct inhalation is modeled. Other pathways such as the
deposition of pollutants to drinking water, and to bioaccumulation of deposited pollutants in the
food supply may be a significant source of exposure for persistent and bioaccumulative
hazardous air pollutants (PB HAP). Screening level evaluations of the potential human health
risks associated with emissions of PB HAP from the modeled facilities are used to determine if
additional analyses are needed, but these analyses are outside the scope of this document.
Because the HEM-3 AERMOD analyses are restricted to the inhalation pathway and depleting
the plume would not be a conservative approach to modeling air concentrations, the impacts of
plume depletion due to deposition are not taken into account. Thus, inhalation impacts may be
overestimated for some pollutants, but exposures through other pathways would be
underestimated.
2-32
-------
A number of other simplifications are made in the dispersion modeling analyses, as noted
in Table 3-1. For instance, building wake effects are not considered. In addition, meteorological
observations are based on the closest station in the HEM-3 meteorological library (see Figure
3-1). Alternative meteorological stations may be more appropriate for some facilities. Ideally,
facility-specific meteorological observations would be used. A single year of meteorological data
(2016) is currently used for AERMOD's dispersion modeling. (The 2008 and prior screening-
level RTR assessments of 51 source categories used meteorological data based on the year
1991.) When considering off-site meteorological data most site specific dispersion modeling
efforts will employ up to five years of data to capture variability in weather patterns from year to
year. However, because of the large number of facilities in the analyses and the extent of the
dispersion modeling analysis (national scale), it is not practical to model five years of data. Other
national studies such as NATA also consider only a single year of meteorological data. A
sensitivity analyses performed by the NATA assessment found that variability attributable to the
selection of the meteorology location/time (both temporal and spatial) resulted in a 17-84%
variation in predicted concentrations at a given station.24
Finally, risk and exposure factors are also subject to uncertainty. Not all individuals
experience the same degree of exposure or internal dose of a given pollutant due to individual-
specific parameters such as weight, age, and gender. While the health benchmarks used in the
analyses crudely account for sensitive populations, a prototypical human (e.g., body weight,
ventilation rate) is used to define the benchmark. Because of the variability of these parameters
in the population, this factor will result in a degree of uncertainty in the resulting risk estimate.
Table 5-1 summarizes the general sources of uncertainty for the RTR modeling analyses.
The table also gives a qualitative indication of the potential direction of bias on risk estimates.
The sources of uncertainty in Table 5-1 are divided into four categories, based on the major
components of the analyses:
• emissions inventory;
• fate and transport modeling;
• exposure assessment; and
• toxicity assessment.
It must also be noted that individual source categories may be subject to additional uncertainties.
These are discussed in separate reports which are prepared for each emission source category
included in the RTR assessments.
2-33
-------
Table 5-1. Summary of General Uncertainties Associated with Risk and Technology Review Risk Assessments
Paramclcr
Assumption
I nccrlainl\/Variabilis Discussion
Polcniial Direction of liias on
Risk I'lslimalcs
Emissions Inventory
Individual HAP emissions
rates and facility
characteristics (stack
parameters, property
boundaries)
Emissions and facility characteristics
from the NEI provide an accurate
characterization of actual source
emissions.
Our current emissions inventory is based on source
category specific ICR and/or the latest NEI, our
internal review, and public comments received. The
degree to which the data in our inventory represents
actual emissions is likely to vary across sources. For
the 2008 screening level assessments, nearly half of
the sources in a given source category submitted a
review of their emissions and facility characteristics
data. Some detailed data, such as property boundary
information is not available for most facilities. This
is an important consideration in determining acute
impacts.
Unbiased overall, magnitude
variable
Multiplier for short-term
emission rates
Generally, maximum short term
emission rates are estimated by
applying a simple multiplier (a factor
of 10) to average annual emissions.
The ratio between short-term and long-term average
emission rates may vary among the different
emission sources at a facility. In addition, the use of a
simple multiplier means that impacts of maximum
short term emissions are modeled with the 99th
percentile meteorological conditions and assuming
these conditions for population exposure.
Potential overestimate due to
the fact that worst-case
emissions are assumed to
occasionally coincide with 99th
percentile worst-case
meteorology.
Overestimate due to lack of
actual information on short-
term emission rates.
Fate and Transport Modeling
Atmospheric dispersion
model choice
AERMOD is EPA's recommended
dispersion model for assessing
pollutant concentrations from
industrial facilities
Field testing of dispersion models, including
AERMOD, have shown results to generally be within
a factor of 2 of measured concentrations.
Unbiased overall
Building downwash
Not included in assessments
Use of this algorithm in AERMOD could improve
the dispersion calculations at individual facilities.
However, data are not readily available to utilize this
option.
Potential underestimate of
maximum risks near facility.
No effect on risks further out.
Plume deposition and
depletion
Not included in assessments
Ignoring these impacts for pollutants that deposit
minimally, and whose risks derive predominantly
from inhalation, should have minimal effect on risk
estimates.
Unbiased or minimal
overestimate.
2-34
-------
Table 5-1. Summary of General Uncertainties Associated with Risk and Technology Review Risk Assessments
(continued)
Piinimcler
Assumption
1 nccrl;iinh/Y;iri;il>ilil\ Discussion
Polcnlhil Direction ol° liiiis on
Risk I'lsliniiiles
Meteorology
One year of meteorological data from
the nearest weather station (selected
from 824 nationwide) is representative
of long-term weather conditions at the
facility.
The use of one year of data rather than the five or
more adds uncertainty based on whether that year is
representative of each location's climatology. Use of
weather station data rather than on-site data can add
to uncertainty. Additionally, the use of default
surface parameters in the generation of the
meteorological datasets imparts uncertainty to the
results from any individual facility.
Minimal underestimate or
overestimate.
Reactivity
Not included in the assessments
Chemical reactions and transformations of individual
HAP into other compounds due to solar radiation and
reactions with other chemicals happens in the
atmosphere. However, in general, the HAP in this
assessment do not react quickly enough for these
transformations to be important near the sources,
where the highest individual risks are estimated.
Further, most of the HAP do not react quickly
enough for these transformations to be important to
risk estimates in the entire modeled domain (i.e.,
within 50 km of the source).
No impact on maximum risk
estimates. Minimal impact on
population risks and incidence.
Maximum modeling
distance
50 kilometers from center of facility
This distance is considered to be the maximum
downwind distance for a Gaussian plume model such
as AERMOD. This is because, in general, winds
cannot be considered to follow straight line
trajectories beyond this distance.
No effect on maximum
individual risks. Minimal
underestimation of incidence.
Exposure Assessment
Locations and short-term
movements of individuals
Ambient concentration at centroid of
each off-site census block is equal to
the exposure concentration for all
people living in that census block.
Effect of human activity patterns on
exposures is not included in the
assessment.
People live at different areas within block that may
have higher or lower exposures than at the centroid.
Individuals also move from outdoors to indoors and
from home to school/work to recreation, etc., and this
can affect their total exposure from these sources.
Unbiased across population for
most pollutants and individuals,
likely overestimate for most
exposed and underestimate for
least exposed persons.
2-35
-------
Table 5-1. Summary of General Uncertainties Associated with Risk and Technology Review Risk Assessments
(continued)
Piinimclcr
Assumption
1 nccrl;iinh/Y;iri;il>ilil\ Discussion
Polcnlhil Direction ol° liiiis on
Risk I'lsliniiiles
Long-term movements of
individuals
MIR individual is exposed
continuously to the highest exposure
concentration for a 70-year lifetime.
Maximum individual risk (MIR) is defined in this
way to be a maximum theoretical risk at a point
where a person can actually reside.
Unbiased for most individuals,
likely overestimate for the
actual individual most exposed
and likely underestimate for the
least exposed. Incidence
remains unbiased unless
population around facilities
increases or decreases over 70
years.
Toxicity Assessment
Reference concentrations
(RfC)
Consistent with EPA guidance, RfCs
are developed including uncertainty
factors to be protective of sensitive
subpopulations. Additionally, RfCs
are developed based on the level
producing an effect in the most
sensitive target organ or system.
While other organ systems may be impacted at
concentrations above the RfC, these are not included
in the calculation of target organ-specific hazard
indices.
In general, EPA derives RfCs
using procedures whose goal is
to avoid underestimating risks
in light of uncertainty and
variability. The greater the
uncertainties, the greater the
potential for overestimating
risks.
Unit Risk Estimate
(URE)
Use of unit risk estimates developed
from dose-response models such as
linear low-dose extrapolation.
Uncertainty in extrapolating the impacts from short-
duration, high-dose animal or work-related exposures
to longer duration, lower-dose environmental
impacts.
Overestimate of risks for
nonlinear carcinogens and for
linear carcinogens with sparse
health effects data. In general,
EPA derives URE values using
procedures aimed at
overestimating risks in light of
uncertainty and variability.
Toxicity of mixtures
Cancer risks and non-cancer hazard
quotients were calculated for each
HAP individually and then summed
into a total risk or hazard index
(assumption of additivity).
Concurrent exposures to multiple chemicals may
result in either increased or decreased toxicity due to
chemical interactions but the data needed to quantify
these effects are generally not available.
Unbiased overall. Some
mixtures may have
underestimated risks, some
overestimated, and some
correctly estimated.
2-36
-------
Table 5-1. Summary of General Uncertainties Associated with Risk and Technology Review Risk Assessments
(continued)
Piimmolcr
Assumption
1 nci'r(;iinl\/Y;iri;il)ili(\ Discussion
Polcnlhil Direction ol° liiiis on
Risk r.sliiiiiik's
Surrogate dose-response
values for HAPs without
values
In the case of groups of HAPs such as
glycol ethers, the most conservative
dose-response value of the chemical
group was used as a surrogate for
missing dose-response values in the
group. For others, such as unspeciated
metals, we have applied speciation
profiles appropriate to the source
category to develop a composite dose-
response value for the group.
For HAP which are not in a group and
for which no URE's or RfC's are
available from credible sources, no
assessment of risk is made.
Rather than neglecting the assessment of risks from
some HAPs lacking dose response values,
conservative assumptions allow the examination of
whether these HAPs may pose an unacceptable risk
and require further examination, or whether the
conservative level examination with surrogates
screens out the HAPs from further assessment.
Overestimate where most
conservative values used.
Unbiased where category-
specific profiles applied.
There is the potential to
underestimate risks for
pollutants which are not
included in the assessment.
2-37
-------
6. References
1. EPA. 2017. AERMOD: Model Formulation and Evaluation. EPA-454/R-17-001, U.S.
Environmental Protection Agency, Research Triangle Park, NC. December 2016.
https://www3.epa.gov/ttn/scram/models/aermod/aermod mfed.pdf (Website last accessed
July 2017.)
2. EPA. 2016. User's Guide for the AMS/EPA Regulatory Model (AERMOD). EPA-454/B-
16-011, U.S. Environmental Protection Agency, Research Triangle Park, NC.
https://www3.epa.gov/ttn/scram/models/aermod/aermod userguide.pdf (Website last
accessed July 2017.)
3. EPA. 2016. The Human Exposure Model (HEM): Single-facility HEM-3 User's Guide;
and Multi-facility HEM-3 andRTR Summary Programs User's Guide. Prepared for the
U.S. Environmental Protection Agency, Research Triangle Park, NC, by EC/R Inc.,
Chapel Hill, NC. June 2016. https://www.epa.gov/fera/human-exposure-model-hem-3-
users-guides
4. EPA. 2016. Dose-Response Assessment for Assessing Health Risks Associated With
Exposure to Hazardous Air Pollutants. U.S. Environmental Protection Agency, Research
Triangle Park, NC. http://www.epa.gov/fera/dose-response-assessment-assessing-health-
risks-associated-exposure-hazardous-air-pollutants (Website last accessed September
2016.)
5. EPA. 2015. Glossary of Key Terms. Technology Transfer Network Air Toxics, 2011
National-Scale Air Toxics Assessment. U.S. Environmental Protection Agency.
http://www.epa.gov/national-air-toxics-assessment/nata-glossary-terms (Website last
accessed September 2016)
6. EPA. 2015. Air Toxics Risk Assessment Reference Library, EPA-453-K-04-001 A, U.S.
Environmental Protection Agency, Research Triangle Park, NC. October 2015.
http://www.epa.gov/fera/risk-assessment-and-modeling-air-toxics-risk-assessment-
reference-library (Website last accessed September 2016)
7. Census. 2000. LandView® 5 on DVD. U.S. Census Bureau, Washington, DC.
8. Census. 2010. Census Summary File 1 - United States:
http://www2.census.gov/census 2010/04-Summary File 1/ prepared by the U.S. Census
Bureau, Washington, DC, 2011. See also Technical Documentation for the 2010 Census
Summary File 1. (Website last accessed September 2016)
9. USGS. 2000. US GeoData Digital Elevation Models - Fact Sheet 040-00 (April 2000).
2-38
-------
U.S. Department of the Interior - U.S. Geological Survey, Washington, DC.
http://data.geocomm.com/sdts/fs0400Q.pdf (Website last accessed September 2016)
10. USGS. 2015. USGS Seamless Data Warehouse. U.S. Department of the Interior - U.S.
Geological Survey, Washington, DC. http://nationalmap.gov/viewer.html (Website last
accessed September 2016)
11. EPA. 2004. User's Guide for the AERMOD Terrain Preprocessor (AERMAP). EPA-
454/B-03-003, U.S. Environmental Protection Agency, Research Triangle Park, NC.
October 2004. http://www.epa.gov/scram001/dispersion related.htm#aermap (Website
last accessed September 2016)
12. EPA. 2011. Addendum to the User's Guide for the AERMOD Terrain Preprocessor
(AERMAP) (EPA-454/B-03-003, October 2004), U.S. Environmental Protection Agency,
Research Triangle Park, NC. March 2011.
http://www.epa.gov/scram001/dispersion related.htm#aermap (Website last accessed
September 2016)
13. EPA. 2015. Addendum to the User's Guide for the AMS/EPA Regulatory Model -
AERMOD (EPA-454/B-03-001, September 2004) (Section 2.2 Deposition Algorithm
Inputs and Options). U.S. Environmental Protection Agency, Research Triangle Park,
NC. June 2015. http://www.epa.gov/scram001/dispersion prefrec.htm#aermod. See also
the "AERMOD Deposition Algorithms - Science Document (Revised Draft)" at
http://www.epa.gov/ttn/scram/7thconf/aermod/aer scid.pdf (Websites last accessed
September 2016)
14. Wesely, M.L., P.V. Doskey, and J.D. Shannon. 2002. Deposition Parameterizationsfor
the Industrial Source Complex (ISC3) Model. ANL/ER/TR-01/003, Argonne National
Laboratory, Argonne, Illinois 60439. June 2002. See "AERMOD Deposition
Parameterizations Document" pdf link under "Model Supporting Documents" section of
TTN's Preferred/Recommended Models webpage at
http://www.epa.gov/scram001/dispersion prefrec.htm#aermod (Website last accessed
September 2016)
15. EPA. 2016. 2014 National Emissions Inventory (NEI) Documentation. U.S.
Environmental Protection Agency, Research Triangle Park, NC.
https://www.epa.gov/air-emissions-inventories/2014-national-emissions-inventory-nei-
documentation (Website last accessed October 2016)
16. EPA. 2015. Technical Support Document - EPA's 2011 National-scale Air Toxics
Assessment - 2011 NATA TSD. Office of Air Quality, Planning, and Standards,
Research Triangle Park, NC, December 2015, pp 20, 121. https://www.epa.gov/national-
2-39
-------
air-toxics-assessment/2011 -nata-technical-support-document (Website last accessed
September 2016)
17. EPA. 2004. User's Guide for the AERMOD Meteorological Preprocessor (AERMET),
EPA-454/B-03-002, U.S. Environmental Protection Agency, Research Triangle Park,
NC. http://www.epa.gov/scramOO 1 /7thconf/aermod/aermetugb.pdf (Website last accessed
September 2016)
18. EPA. 2015. Addendum to the User's Guide for the AERMOD Meteorological
Preprocessor (AERMET) (EPA-454/B-03-002, November 2004), U.S. Environmental
Protection Agency, Research Triangle Park, NC. June 2015.
http://www.epa.gov/ttn/scram/metobsdata procaccprogs.htm#aermet (Website last
accessed September 2016)
19. Texas Commission on Environmental Quality. 2016. Reports on Air Emission Events.
http://www.tceq.texas.gov/field/eventreporting (Website last accessed October 2016)
20. EPA. 2005. Revision to the Guideline on Air Quality Models: Adoption of a Preferred
General Purpose (Flat and Complex Terrain) Dispersion Model and Other Revisions.
Appendix W of 40 CFR Part 51. November 2005.
http://www.epa.gov/ttn/scram/guidance/guide/appw 05.pdf (Website last accessed
September 2016)
21. Federal Register. 2012. Qualifying Urban Areas for the 2010 Census. FR 77:59 (27
March 2012). p. 18652. http://www.gpo.gov/fdsvs/pkg/FR-2012-03-27/pdf/2012-69Q3.pdf
(Website last accessed September 2016)
22. EPA. 2006. NEI Quality Assurance and Data Augmentation for Point Sources. U.S.
Environmental Protection Agency, Emission Inventory Group, RTP, NC. February 2006.
https://www.epa.gov/sites/production/files/2Q15-
1 l/documents/nei2002 qa augmentation 0206.pdf. See also "Quality Assurance details"
for the 2011 NEI at https://www.epa.gov/air-emissions-inventories/2011-national-
emissions-inventory-nei-documentation. (Websites last accessed September 2016)
23. EPA. 2003. AERMOD: Latest Features and Evaluation Results. EPA-454/R-03-003,
U.S. Environmental Protection Agency, Research Triangle Park, NC. June 2003.
https://www3.epa.gov/ttn/scram/7thconf/aermod/aermod mep.pdf. (Website last
accessed September 2016)
24. EPA, 2002, National-Scale Air Toxics Assessment for 1996; EPA-453/R-01-003; EPA
454/B-03-003, U.S. Environmental Protection Agency, Research Triangle Park, NC.
https://archive.epa.gov/airtoxics/nata/web/html/index-2.html (Website last accessed
September 2016)
2-40
-------
Appendix 3 to the Risk Assessment Report for the Sterigenics Facility in Willowbrook,
Illinois:
Meteorological Data for HEM-3 Modeling
3.1 Introduction
As part of the risk assessment for Sterigenics, 2014-2018 meteorological data from Argonne
National Laboratory were processed in AERMET for subsequent input to AERMOD (USEPA,
2018a). Argonne is approximately 7 km southwest of the Sterigenics facility (Figure 1). The
closest National Weather Service (NWS) station, Midway airport, is approximately 16 km east of
Sterigenics. While Midway can be considered adequately representative of the Sterigenics
facility in the absence of other data, given the proximity of Argonne to the facility, the EPA
concluded that meteorological data collected at Argonne would be more representative of
conditions at Sterigenics than data from Midway. The Argonne meteorological tower also had
measurements of wind, temperature, and turbulence (standard deviation of wind direction, oe) at
10 m and 60 m vertical levels, making a more robust dataset over standard airport observations
which only have one level of data without turbulence measurements. Sections 3.4 and 3.5
describe the methodology and results to support the EPA's decision to use Argonne data for the
risk assessment.
3.2 Meteorological data processing
Meteorological data for Argonne are available for download at
http://www.atmos.anl.gov/ANLMET/. Both hourly averaged data and data in 15-minute
intervals are available for download. For the purposes of the risk assessment, the hourly
averaged data were used. The following variables from Argonne were input to AERMET
(USEPA, 2018b):
• Solar insolation
• Surface pressure
• 10 m wind speed
• 10 m wind direction
• 10 m temperature
• 10 m standard deviation of wind direction (oe)
• 60 m wind speed
• 60 m wind direction
• 60 m temperature
• 60 m standard deviation of wind direction (oe)
The wind speed threshold used in AERMET to define valid wind speeds was set to 0.1 m/s. In
accordance with the EPA's Guideline on Air Quality Modeling (USEPA, 2017), since the
Argonne data included turbulence data (oe), the adjustment to the surface friction velocity
(adjusted u* option) was not utilized.
3-1
-------
Figure 1. Locations of Argonne National Laboratory tower and Midway Airport relative
to Sterigenics.
Sterfgenics
Argonne
Sour?*: Ss
-------
AERSURFACE is limited to the 1992 NLCD. While the 2019 version is draft, it can be used for
regulatory purposes if run with the default 1 km radius for surface roughness estimates, use of
landcover, impervious surface data, and tree canopy data for the selected NLCD year, and in
consultation with the appropriate reviewing authority (U.S EPA, 2019). For this risk assessment,
2011 data were used. Year-specific monthly surface characteristics were calculated for 2014-
2018 because there are two inputs to AERSURFACE that can vary by year: 1) moisture
conditions for the year (average, wet, or dry year based on precipitation), and 2) the presence of
continuous snow cover during the winter. The assumptions of moisture conditions and winter
conditions were assumed to be the same for both Argonne and Midway. These assumptions
were based on climatological data for Midway for 1989-2018. The assignments for wet, dry, and
average rainfall are based on guidance in the AERSURFACE user's guide (USEPA, 2019).
Because the lookup tables used by AERSURFACE are based on seasons, when calculating
monthly surface characteristics, each month must be assigned to a season. Table 1 lists the
seasonal assignments by month for each modeled year as well as the moisture conditions for
each year.
Table 1. Seasonal assignments by month and year for AERSURFACE processing.
Year
Season
2014 (wet)
2015 (wet)
2016 (average)
2017 (average)
2018 (average)
Winter (no
November,
November,
November,
November,
November,
snow)
December,
December,
January,
January,
December,
March
March
February,
February,
January,
March
March
March
Winter
January,
January,
December
December
February
(continuous
February
February
snow)
Spring
April, May
April, May
April, May
April, May
April, May
Summer
June, July,
June, July,
June, July,
June, July,
June, July,
August
August
August
August
August
Autumn
September,
September,
September,
September,
September,
October
October
October
October
October
Surface roughness was calculated for four sectors for Argonne (Figure 2) and three sectors for
Midway (Figure 3). AERSURFACE also allows for different treatment of surface roughness for
a sector depending on whether the land use around the site in that sector is more like an airport or
non-airport. This choice is used when a sector contains impervious surfaces such as buildings,
roads, runways, parking lots, etc. If a sector contains mostly flat impervious surfaces such as
roads or parking lots, the sector can be treated as an airport even if the site is not an airport. If
the sector contains mostly buildings, then it can be treated as non-airport even if the site is an
airport but the sector contains the terminal buildings, for example. All sectors at Argonne were
treated as non-airport sectors. Sector 1 at Midway was treated as an airport sector while the
other two sectors were treated as non-airport. Sector 1 is treated as an airport sector because
most of the land use in that sector is a developed category with large flat developed spaces such
as runways. The other two sectors are treated as non-airport because they are developed spaces
3-3
-------
Figure 2. Argonne surface roughness sectors.
Kilo meters]
gfapftS OiES'Ajftajjl
iwCcmmu'ity1
3-4
-------
Figure 3. Midway surface roughness sectors.
I Kilo meter;
that are not flat spaces and composed of developed structures such as buildings. See the
AERSURFACE guide (USEPA, 2019) for more details on sector treatment.
3.4 Meteorological comparisons for the ethylene oxide sampling period
To determine the representativeness of Argonne for Sterigenics, wind and temperature data from
Argonne, Midway, and the meteorological instrument at the EPA warehouse near Sterigenics
were compared for the ambient air sampling period of November 13, 2018 through March 31,
2019. Figure 4 shows the location of the EPA warehouse meteorological i nstrument relative to
the two Sterigenics buildings, Willowbrook 1 (WB1) and Willowbrook 2 (WB2). The EPA
instrument is located approximately 150 m southwest of WB1 and approximately 300 m from
WB2. The height of the EPA instrument is 8.5 m above ground and is indicated by the green
triangle in Figure 4. The EPA instrument collected temperature, wind, oe, relative humidity,
pressure, and precipitation measurements. The EPA data were processed in AERMET with the
inputs listed above except for precipitation, which is only needed for AERMOD simulations
involving deposition calculations. The draft 2019 AERSURFACE was run for all three sites for
January through March 2019 assuming average moisture conditions, continuous snow for
January, and no continuous snow for February and March. For 2018, all three sites used the
3-5
-------
moisture conditions and seasonal-month assignments outlined in Table 1 for November and
December. AERSURFACE was run for four surface roughness sectors (all non-airport) (Figure
5) for the EPA site. Midway was used as the representative NWS site with surface characteristics
as described in the previous section with 5.7 percent of the hours in the data period subsituted
with Midway data. As with Argonne, since the EPA warehouse site collected turbulence data,
the surface friction velocity adjustment was not performed. AERMET was also run for the
sampling period for Midway only to assess how well the representative NWS site performed.
Since Midway did not collect turbulence data, the surface friction velocity adjustment was
included in the AERMET processing.
Wind roses for the monitoring period are shown for all three locations in Figure 6. The roses
indicate that the overall flow pattern among the three sites is similar. Flowever, the EPA site
tends to have stronger signals of southerly and northerly flows compared to the other two sites.
The differences in flow patterns could be due to building effects near the EPA instrument while
the other two sites are in open locations and would represent the more general flow for the area.
Figure 4. Location of EPA meteorological instruments relative to the Sterigenics buildings.
MetersI
Esri. DigitalGlot
SgTappingA^
3-6
-------
Figure 5. EPA surface roughness sectors.
Kilometer:
3-7
-------
Figure 6. Argonne, EPA, and Midway wind roses for November 13, 2018 - March 31, 2019.
-22
17-21
~ 1T-21
Argonne
EPA
VWD SPEED
(Knots)
I 1 17-21
¦ 11.17
Midway
128%
Analyses of wind speeds, directions, and temperatures were conducted among the three sites.
Winds and temperatures at the 10 m level for Argonne were compared to the 8.5 m level winds
and temperatures for the EPA site, and to the 10 m winds and 2 m temperature for Midway, on
an hourly basis. Table 2 lists the minimum, mean, median, and maximum wind speed
differences among the three sites. Table 3 lists the minimum, mean, median, and maximum
wind direction differences among the three sites28. There were 2,920 hours where all three sites
had wind data out of a possible 3,300 hours (the EPA instruments started at 13:00 LST on
November 13, 2018). The results in Table 2 indicate that Argonne tended to have higher wind
speeds than the EPA site. In fact, of the 2,920 hours, there were 2,639 hours where Argonne was
higher than the EPA site. Conversely, Argonne tended to have lower wind speeds than Midway
(2,537 hours) as did the EPA site when compared to Midway (2,853 hours). When looking at the
number of hours where the sites' wind speeds were within 1 m/s of each other, there were 1,515
28 The maximum difference between two directions is 180°. For example, the difference between a 10° direction
and 350° direction is 20. after accounting for the 360° crossover on the compass0, not 340° based on a straight
arithmetic difference between 350° and 10°.
3-8
-------
hours where Argonne and the EPA site were within ± 1 m/s, 1,388 hours where Argonne and
Midway were within ± 1 m/s, and 409 hours where the EPA site and Midway within ± 1 m/s.
Table 2. Hourly wind speed differences among Argonne, EPA site, and Midway.
Difference
Minimum (m/s)
Mean (m/s)
Median (m/s)
Maximum (m/s)
Argonne - EPA
-8.30
1.07
1.00
5.20
Argonne - Midway
-5.34
-1.08
-1.02
3.00
EPA - Midway
-7.38
-2.16
-2.08
8.63
The wind direction differences in Table 3 indicate the wind direction tended to vary within 20°
among the three sites, with only a few hours where the winds were in almost opposite directions.
There were 1,322 hours where Argonne and the EPA site wind directions were within 10°, 1,573
hours where Argonne and Midway directions were within 10°, and 1,268 hours where the EPA
site and Midway directions were within 10°. The number of hours where winds were in almost
opposite directions (> 170°) were few. There were only three hours where Argonne and the EPA
site direction differences exceeded 170°, one hour where Argonne and Midway direction
differences exceeded 170°, and 11 hours where the EPA site and Midway direction differences
exceeded 170°.
Table 3. Hourly wind direction differences among Argonne, EPA, and Midway.
Difference
Minimum (°)
Mean (°)
Median (°)
Maximum (°)
Argonne - EPA
0
13
11
178
Argonne - Midway
0
16
9
173
EPA - Midway
0
17
12
180
Table 4 lists the minimum, mean, and maximum hourly temperatures for each site for each
month of the sampling period. These statistics were calculated for each site independently of the
other two. The results in Table 4 indicate that, on average, the temperatures among the three
sites are similar.
3-9
-------
Table 4. Monthly minimum, mean, and maximum temperatures for Argonne, EPA site,
and Midway.
Temperature
(°C)
Site
November
December
January
February
March
Minimum
Argonne
-8.40
-10.20
-31.0
-17.6
-19.9
EPA
-7.80
-9.90
-30.2
-17.7
-19.5
Midway
-10.76
-11.26
-32.26
-18.66
-21.96
Mean
Argonne
-0.72
0.51
-6.12
-3.30
1.37
EPA
-1.20
0.60
-5.63
-2.76
1.66
Midway
-2.72
-1.92
-8.19
-5.48
-1.01
Maximum
Argonne
9.70
11.50
12.20
10.30
16.90
EPA
7.90
11.60
12.0
10.60
17.90
Midway
7.16
9.24
9.74
7.64
15.24
Table 5 lists the minimum, mean, median, and maximum hourly temperature differences among
the three sites. There were 3,135 hours where all three sites had temperature data.
Table 5. Hourly temperature differences among Argonne, EPA site, and Midway.
Difference
Minimum (°C)
Mean (°C)
Median (°C)
Maximum (°C)
Argonne - EPA
-4.50
-0.35
-0.3
4.2
Argonne - Midway
-0.74
2.20
2.16
7.96
EPA - Midway
-1.94
2.57
2.46
8.26
While the minimum and maximum hourly differences were greater than 1° for Argonne and the
EPA site, the mean and median differences indicated little difference between the two sites. In
fact, for the 3,135 hours of temperature data, 2,803 hours had temperature differences within ±
1°C between Argonne and the EPA site. There were larger differences between Midway and the
other two sites, with only 111 hours of temperature differences within ± 1°C between Midway
and Argonne, and 34 hours of temperature differences within ± 1°C between Midway and the
EPA site. These comparisons indicate that the Argonne data seem to better represent the
Willowbrook area, supporting the use of the Argonne meteorological data for the risk
assessment.
3.5 AERMOD simulations
To further evaluate the representativeness of Argonne, the EPA site, and Midway, AERMOD
simulations using day-specific ethylene oxide usage were conducted for 28 of the sampling days.
AERMOD performance for the 28 sampling days at the monitors using Argonne, EPA site, and
Midway meteorological data was evaluated using methodology from the EPA Protocol for
Determining the Best Performing Model (USEPA, 1992) for regulatory application, which
focuses on the higher concentrations in the concentration distribution. Normally, the protocol
evaluates 1-hour, 3-hour, and 24-hour average concentrations. Since the monitor data for
3-10
-------
Sterigenics are only 24-hour averages, the EPA focused only on 24-hour averages. The protocol
uses a statistic call Robust Highest Concentration (RHC) and fractional bias for evaluation of
model performance. The RHC is calculated at each monitor location for observed concentrations
and modeled concentrations. The RHC is calculated as:
3 N -
RHC = X(JV) + [X - X(JV)] x In
where X(N) is the Nth highest concentration, X is the average of N-l values, and N is typically
set to 26 values for most model evaluations. However, given the small sample size at each
monitor, we started with N=5 to determine performance for the higher concentrations and
evaluated results up to N=18 (the fewest number of observations across the monitors) to
determine performance across the entire concentration distribution. As stated above, the RHC is
calculated at each monitor for observed concentrations and modeled concentrations. Next, a
fractional bias is calculated using the maximum observed RHC and maximum modeled
(predicted) RHC as:
FB = 2
OB - PR
YOB + PRi
where FB is the fractional bias, OB is the maximum observed RHC, and PR is the maximum
modeled RHC. A positive fractional bias indicates model underprediction, and a negative
fractional bias indicates model overprediction. Fractional biases within ± 0.67 are not considered
statistically different. Also, note that the two RHC values in the fractional bias may not be from
the same monitor location. This is done to assess the model's ability to assess concentrations for
regulatory purposes, that is, how well the model predicts maximum concentrations regardless of
the spatial location. Table 6 lists the fractional biases for three values of N for Argonne, the EPA
site and Midway. For all three sample sizes of N, the EPA site performed best, while Argonne
outperformed Midway, which supports the use of the Argonne meteorological data for the risk
assessment.
Table 6. Fractional biases for N= 5,10, and 18 for Argonne, Midway, and the EPA site.
N
Argonne fractional
bias
Midway fractional bias
EPA fractional bias
5
1.05
1.29
0.98
10
1.05
1.23
0.98
18
0.85
1.10
0.84
3.6 2014-2018 Argonne vs. Midway meteorological data comparisons
Comparisons of winds and temperatures between Argonne and Midway were made for the full
period of 2014-2018, with an additional emphasis on the November-March period over all five
years, to ensure that the November 2018-March 2019 period was not an outlier relative to other
years. Figures 7 and 8 show the wind roses for Argonne and Midway, respectively, for the entire
2014-2018 period. Figures 9 and 10 show the 2014-2018 wind roses for November-March only,
3-11
-------
to coincide with the sampling period from November 2018-March 2019. For the entire 5-year
period, while there are some differences, the wind roses are similar in the overall pattern of
winds. Both stations exhibit a strong northeasterly wind component and south to west
Figure 7. Argonne 2014-2018 wind rose.
NORTH
12%
WEST
WIND SPEED
(Knots)
SOUTH
Calms: 0.00%
3-12
-------
Figure 8. Midway 2014-2018 wind rose.
NORTH
12%
WEST;
WIND SPEED
(Knots)
SOUTH
Calms: 0.66%
3-13
-------
Figure 9. Argonne November-March 2014-2018 wind rose.
NORTH'
12%
9.6%
4.8%
WEST
7.2%
: EAST;
WIND SPEED
(Knots)
SOUTH
>= 22
17-21
11 -17
7 -11
4-7
1 -4
Calms: 0.00%
3-14
-------
Figure 10. Midway November-March 2014-2018 wind rose.
12%
'O
EAST
WIND SPEED
(Knots)
Calms: 0.16%
3-15
-------
component (Figures 7 and 8). For November-March periods over the five years, both stations
exhibit the same general pattern, with Midway having a higher frequency of mid-range wind
speeds (11-17 knots) than Argonne.
Hourly wind difference analyses were conducted between Argonne and Midway for 2014-2018.
Table 7 gives the hourly wind speed differences for the entire 5-year period, as well as the
November-March period. The distribution of differences for both the entire period and the
November to March period were comparable to the distributions in Table 2. Of the 39,043 hours
of winds where both sites had data for the full period, 12,937 hours had a wind speed difference
within ± 1 m/s. For the November-March months, there were 16,850 hours where both sites had
data and 6,417 hours had a wind speed difference within ± 1 m/s. Table 8 lists the wind
direction differences between Argonne and Midway, and the distributions of differences in Table
8 compared well with the Table 3 differences. For the wind direction differences, there were
19,144 hours where the wind direction difference was less than 10° for the full 5-year period and
9,566 hours for the November-March period with wind direction differences less than 10°.
Table 7. Hourly wind speed differences between Argonne and Midway for 2014-2018.
Difference
Minimum (m/s)
Mean (m/s)
Median (m/s)
Maximum (m/s)
Argonne - Midway
(full period)
-9.05
-1.50
-1.39
4.56
Argonne - Midway
(November-March)
-9.05
-1.39
-1.25
3.2
Table 8. Hourly wind direction differences between Argonne and Midway for 2014-2018.
Difference
Minimum (°)
Mean (°)
Median (°)
Maximum (°)
Argonne - Midway
(full period)
0
17
10
180
Argonne - Midway
(November-March)
0
13
9
179
Table 9 lists the 5-year average minimum, mean, and maximum temperatures by month for
Argonne and Midway. As with the November 2018-March 2019 period, the temperatures are
similar across all months between the two stations. Also, the statistics for November-March do
not indicate that the November 2018-March 2019 differences (Table 4) were unusual when
compared to the 5-year averages.
3-16
-------
Table 9. 5-year average monthly minimum, mean, and maximum temperatures (°C) for
Argonne and Midway.
Month
Argonne
Midway
T min
—1
O)
<
era
T max
T min
—1
O)
<
era
T max
January
-23.30
-4.96
10.22
-21.18
-3.74
11.12
February
-18.46
-3.28
14.24
-16.70
-2.17
14.72
March
-12.58
2.75
20.58
-10.90
3.58
21.18
April
-3.60
9.23
26.30
-2.32
9.82
26.92
May
3.70
16.34
31.20
4.60
17.13
32.46
June
12.58
21.85
31.88
11.3
22.47
34.26
July
12.74
22.63
32.10
14.64
24.19
33.72
August
12.38
22.35
31.46
14.12
24.00
33.86
September
7.18
19.6
32.40
8.58
21.00
33.70
October
0.14
12.48
27.26
1.42
13.56
27.96
November
-9.66
4.29
18.24
-8.02
5.48
18.62
December
-16.06
-0.57
13.30
-14.2
0.60
14.44
Table 10 lists the hourly temperature difference statistics between Argonne and Midway. There
were 42,291 hours where both sites had data for the entire period and 18,037 hours where both
sites had data for the months of November-March. Argonne seems to have slightly cooler
temperatures than Midway, possibly due to Midway being in a more urban environment than
Argonne. The November-March statistics do vary from the November 2018-March 2019 results
in Table 5, especially for the minimum and maximum temperature differences. This would not
be unexpected when looking at an individual period (November 2018-March 2019) compared to
a longer-term period of 5 years for the same months, but overall the differences for the 5-year
period are comparable to the differences for November 2018-March 2019.
Table 10. Hourly temperature differences between Argonne and Midway for 2014-2018.
Difference
Minimum (°C)
Mean (°C)
Median (°C)
Maximum (°C)
Argonne - Midway
(full period)
-6.44
-1.14
-1.24
14.86
Argonne - Midway
(November-March)
-4.84
-1.10
-1.14
11.16
Based on the analyses in this section, there is nothing to indicate that Argonne would not be
representative of Sterigenics for the 2014-2018 period and the analysis of Section 3.5 using
November 2018-March 2019 would be valid for the entire period of 2014-2018.
The meteorological analyses presented here indicate that both Midway and Argonne can be
considered representative of Sterigenics. A statistical analysis of AERMOD output using
3-17
-------
methodology from the EPA's protocol for determining the best performing model shows that
Argonne meteorological data outperformed Midway data. These analyses support the conclusion
that while both Midway and Argonne are adequately representative meteorological sites for the
risk assessment, Argonne would be the most representative of the two sites, given proximity to
Sterigenics, available data, and how those data influence model output.
3.7 References
USEPA. 1992. Protocol for Determining the Best Performing Model, EPA-454/R-92-025. U.S.
Environmental Protection Agency, Research Triangle Park, NC.
USEPA. 2013. AERSURFACE User's Guide. U.S. Environmental Protection Agency. EPA
454/B-08-001. Revised January 16, 2013.
USEPA. 2015. AERMINUTE User's Guide. U.S. Environmental Protection Agency. EPA
454/B-15-006.
USEPA. 2017. Revisions to the Guideline on Air Quality Models: Enhancements to the
AERMOD Dispersion Modeling System and Incorporation of Approaches to Address Ozone and
Fine Particulate Matter. 40 CFR Part 51.
https://www3.epa.eov/ttn/scram/eiiidance/guide/appw 17.pdf
USEPA. 2018a. User's Guide for the AMS/EPA Regulatory Model - AERMOD. U.S.
Environmental Protection Agency. 454/B-18-001.
USEPA. 2018b. User's Guide for the AERMOD Meteorological Processor (AERMET). U.S.
Environmental Protection Agency. EPA-454/B-18-002.
USEPA. 2019. User's Guide for Draft AERSURFACE Tool (Version 19039 DRFT). U.S.
Environmental Protection Agency. EPA 454/B-19-001.
3-18
-------
Appendix 4 to the Risk Assessment Report
for the Sterigenics Facility in Willowbrook, Illinois:
U.S. EPA Risk Assessment for Sterigenics-Willowbrook (Slides from May 29, 2019,
Public Meeting)
-------
U.S. EPA Risk
Assessment for
Sterigenics-
Willowbrook
Kelly Rimer
Leader, Air Toxics Assessment Group
United States Environmental Protection Agency
-------
What we'll cover
~ Key Terms
~ EPA's Sterigenics Willowbrook Risk Assessment
~ What the Assessment Examined
~ Areas the Assessment Covered
~ Limitations and Uncertainties
~ Review of Results
-------
Two Key Terms
~ Air toxics are pollutants that are known or suspected to
cause cancer or other serious health effects
~ Also known as "hazardous air pollutants"
~ Ethylene oxide is an air toxic
~ Cancer risk refers to the chance that breathing in an air
toxic will cause people to develop cancer
~ Separate from the risk of developing cancer from other causes
~ EPA describes that chance as a number in 1 million people
~ For example, 1 in 1 million means that 1 person in 1 million
people could develop cancer from breathing air toxics
-------
Areas the risk assessment covered
~ This risk assessment estimates the risks for
several communities including:
~ Willowbrook
~ Burr Ridge
~ Hinsdale
~ Darien
~ Indian Head Park
~ Western Springs
-------
We evaluated two scenarios
1. Potential risks from the Sterigenics-Willowbrook
facility that exist after the emission controls that
were installed in July 2018
~ Called the "Pre-Seal Order"
2. Potential risks assuming that the emissions from
the facility is more highly controlled
~ Called the "Illustrative Future Case"
-------
Assumptions in the scenarios
~ For both scenarios the assessment estimates:
~ Risk in areas where people live
~ Risk in areas where people work close to the facility (but not
at the facility)
~ For areas where people live, we assume continuous 24/7
exposure for 70 years
~ For areas where people work close to the facility, we
assume people are exposed 8.5 hours a day, 5 days a week,
50 weeks a year for 25 years
-------
Limitations and Uncertainties
This risk assessment:
~ Focuses on risks from the Sterigenics facility only
~ Does not assess comprehensive risk from all air pollution sources
~ Provides general estimates of a population's risk of getting
cancer due to EtO emissions from the Sterigenics-Willowbrook
plant
~ Cannot be used predict an individual's chance of getting cancer
~ Is more likely to over-estimate risk than underestimate risk due
to what we call 'health-protective assumptions'
-------
Estimated Residential Lifetime Cancer Risk from ethylene oxide
emissions from Sterigenics Willowbrook
Pre-Seal Order Conditions
Darien
Western
Springs
Hinsdale
Indian
Head Park
Willowbrook
Lifetime
Ethylene Oxide
Cancer Risk
(in 1 million)
H 100 - 200
200 - 500
500 - 1000
Non-Residential
Areas
Sterigenics
Illustrative Future Case
12 3 Kilometers
_J l l l I l l l I
Western
Springs
Hinsdale
Indian
Head Park
Willowbrook
Burr Ridge
Darien
3 Kilometers
J l l l I l l l I
Based on operations before seal order (Reflects emissions reductions
from controls installed in Summer 2018)
Based on the facility being more highly controlled. Estimated risks
would be below 100 in 1 million - and potentially as low as 1 in 1
million.
-------
Willowbrook
Zooming in:
Estimated Residential Lifetime Cancer Risk
Lifetime
Ethylene Oxide
Cancer Risk
(in 1 million)
M 100 - 200
200 - 500
500 - 1000
/.. Non-Residential
X*' Areas
¦ Sterigenics
0 200 400 600 800 1,000 Meters
1 I I I I I
Pre-Seal Order Conditions
Based on operations before
seal order (Reflects emissions
reductions from controls
installed in Summer 2018)
9
-------
Estimated Residential Lifetime Cancer Risk from ethylene oxide
emissions from Sterigenics Willowbrook
Willowbrook
Western
Springs
Hinsdale
Lifetime
Ethylene Oxide
Cancer Risk
(in 1 million)
M 100 - 200
200 - 500
500 - 1000
Non-Residential
Areas
Sterigenics
Indian
Head Park
Darien
0 12 3 Kilometers
1 I i l I i I I I l l l I
Pre-Seal Order Conditions
Based on operations before
seal order (Reflects emissions
reductions from controls
installed in Summer 2018)
10
-------
Zooming in:
Estimated Residential Lifetime Cancer Risk
Steriqenics
Willowbrook
Burr Ridge
llustrative Future Case
Based on the facility being
more highly controlled.
Estimated risks would be
below 100 in 1 million - and
potentially as low as 1 in 1
million
0 200 400 600 800 1,000 Meters
1 I I I I I
11
-------
Executive
Willowbrook
Occupational
Ethylene Oxide
Cancer Risk
(in 1 million)
H 100 - 200
200 - 500
500 - 1000
Sterigenics
Estimated Occupational Lifetime Ethylene Oxide Cancer Risk
from Sterigenics Willowbrook
Pre-Seal Order Conditions Illustrative Future Case
Executive Or
Executive Dr
0 50 100 150 200 250 Meters
1 I I I I I
250 Meters
_l
Based on operations before seal order (Reflects emissions reductions Based on the facility being more highly controlled. Estimated risks would
from controls installed in Summer 2018) be below 100 in 1 million - and potentially as low as 1 in 1 million.
Willowbrook
Midway Dr
-------
Estimated Occupational Lifetime Ethylene Oxide Cancer Risk
from Sterigenics Willowbrook
Pre-Seal Order Conditions
Based on operations before
seal order (Reflects emissions
reductions from controls
installed in Summer 2018)
Occupational
Ethylene Oxide
Cancer Risk
(in 1 million)
100 - 200
200 - 500
500 - 1000
C.y.-p-fcuM
Executive ur
Sterigenics
Wi owbrook
250 Meters
-------
Estimated Occupational Lifetime Ethylene Oxide Cancer Risk
from Sterigenics Willowbrook
Sterigenics
ExecutiV!
Executive Dr
Midway
Willowbrook
%
m
if
ft,.
. V
.$•
-------
Thank You
15
------- |