RESPONSES
TO
PEER
REVIEW
COMMENTS
ON
THE
DRAFT
RISK
ASSESSMENT
DOCUMENT
FOR
COKE
OVEN
BATTERIES
Risk
and
Exposure
Assessment
Group
Emissions
Standards
Division
Office
of
Air
Quality
Planning
and
Standards
U.
S.
Environmental
Protection
Agency
Research
Triangle
Park,
North
Carolina
May
20,
2004
1
I.
INTRODUCTION
The
U.
S.
Environmental
Protection
Agency
(
EPA)
submitted
the
April
2003
draft
report,
Risk
Assessment
Document
for
Coke
Oven
MACT
Residual
Risk,
to
a
panel
of
three
independent
experts
for
peer
review.
The
peer
reviewer
comments
responded
to
six
charge
questions
we
provided.
Their
comments
are
detailed
in
the
document,
Peer
Review
Report
on
the
Draft
Risk
Assessment
for
Coke
Oven
MACT
Residual
Risk
Assessment,
dated
June
13,
2003.
Our
responses
to
the
peer
review
comments
are
incorporated
into
the
risk
assessment
document
for
the
proposed
amendments,
which
can
be
found
in
the
rulemaking
docket
(
Docket
Number
OAR­
2003­
0051).
This
report
documents
our
responses
to
each
of
the
major
(
i.
e.,
comments
that
addressed
technical
rather
that
editorial
issues)
comments,
including
those
additional
comments
made
in
response
to
questions
that
arose
during
their
review.

II.
SUMMARY
OF
EPA
RESPONSES
TO
PEER
REVIEW
COMMENTS
Charge
Question
1:
Given
the
specific
emission
characteristics
of
this
source
and
the
tools
available
for
characterizing
these
emissions,
is
the
methodology
we
applied,
(
i.
e.,
using
the
Buoyant
Line
Plume
[
BLP]
model
to
estimate
plume
height
and
the
Industrial
Source
Complex
[
ISCST3]
model
to
disperse
the
emissions),
a
reasonable
approach
for
modeling
the
atmospheric
transport
of
the
affected
emissions?
Additionally,
is
our
use
of
virtual
stacks
a
reasonable
approach
for
handling
the
emissions
coming
from
this
type
of
source?

Peer
Review
Comments:
The
reviewers
were
generally
supportive
of
the
use
of
the
modeling
approach
we
used,
(
i.
e.,
the
joining
of
the
BLP
and
ISC
models
and
the
use
of
14
"
virtual
stacks")
to
disperse
the
emissions
across
these
batteries.
They
requested
clarification
on
these
and
a
few
other
topics.
Their
major
comments
focused
on
the
modeling
assumptions
or
options
chosen.
Specifically,
their
comments
were
that
more
documentation
and
justification
for
the
modeling
assumptions
were
needed,
especially
in
the
area
of
final
plume
rise,
buoyancy
induced
dispersion,
and
building
downwash.
One
reviewer
also
pointed
out
a
possible
error
in
the
buoyancy
flux
term,
which
could
underestimate
the
maximum
ground
level
concentration.

Response:
To
address
their
comments
requesting
additional
clarification,
we
revised
the
risk
assessment
document
to
add
more
information
to
the
text
and
to
the
tables.
In
regards
to
the
modeling
assumptions,
we
added
more
text
to
Appendix
D
as
well
as
Table
D­
3
to
identify
the
modeling
assumptions
and
support
how
they
were
used.
We
also
added
some
additional
text
to
clarify
the
bias
table
(
Table
2­
10).
Table
D­
3
provides
a
description
of
possible
modeling
options
and
how
each
was
handled
in
setting
up
the
models.
For
the
parameters
upon
which
the
reviewers
focused,
we
used
the
building
downwash
algorithm
contained
within
the
BLP
model
but
did
not
include
the
option
for
plume
depletion,
buoyancy
induced
dispersion,
or
stack­
tip
downwash.
We
performed
a
sensitivity
evaluation
of
these
features
and
concluded
that
their
inclusion
as
options
in
the
modeling
would
have
minimal
effect
on
the
predicted
ambient
concentration
and
deposition
at
critical
offsite
locations.
More
specifically,
these
evaluations
2
show:

(
1)
When
gradual
plume
rise
not
considered,
(
i.
e.,
the
final
plume
height
is
assumed
to
be
reached
directly
above
the
battery),
the
effect
is
minimal
because
under
most
meteorological
conditions,
final
plume
rise
is
reached
before
the
plume
reaches
the
facility
fenceline.

(
2)
Under
typical
conditions,
buoyancy
induced
dispersion
will
result
in
an
average
increase
of
plume
size
by
about
5%.

(
3)
The
difference
in
ambient
concentrations
estimates
resulting
from
different
plume
downwash
calculations
(
i.
e.,
between
the
BLP
algorithm
and
the
newer
ISC
algorithms)
is
minimal
at
offsite
locations.

(
4)
Increasing
the
number
of
virtual
or
representative
stacks
from
14
to
36
results
in
ambient
concentration
predictions
that
vary
by
less
than
10%.

We
checked
the
buoyancy
flux
term
in
the
draft
risk
assessment
document
and
found
that
the
term
F
should
be
F'
(
F
prime),
but
the
calculations
were
correct.
The
text
has
been
corrected
and
our
procedures
match
those
of
Sciences
International,
who
developed
the
protocol
.

Charge
Question
2:
A
multi­
pathway
screening
analysis
was
done
in
order
to
determine
if
pathways
other
than
inhalation
were
a
source
of
significant
(
by
comparison
with
inhalation)
risk
to
the
surrounding
population.
This
screening
analysis
was
conducted
in
two
steps.
First,
hazardous
air
pollutants
(
HAP)
identified
as
being
constituents
of
coke
oven
emissions
were
crossed
with
lists
of
chemicals
identified
as
being
a
PBT
(
persistent,
bioaccumulative,
toxic)
concern
to
EPA.
Second,
the
resulting
list
of
constituent
HAP
was
modeled
using
EPA's
Indirect
Exposure
Model
(
IEM).
Is
this
two
step
approach
an
appropriate
method
for
screening
HAPs
for
multi­
pathway?
Is
the
level
of
the
analysis
sufficient
for
making
the
determination
that
ingestion
risk
is
not
enough
of
a
contributor
to
population
risk
to
warrant
more
refined
or
extensive
analysis?

Peer
Review
Comments:
There
was
concern
that
some
of
the
precipitation
data
on
SAMSON
CD
may
not
have
been
complete.
One
reviewer
questioned
the
need
for
a
two­
step
selection
process
for
determining
which
HAP
would
be
included
in
a
multipathway
screening
assessment,
but
agreed
that
what
we
did
was
done
correctly
and
was
comfortable
with
the
level
of
the
assessment.
However,
the
document
should
include
explanation
for
7
(
out
of
a
possible
maximum
ranking
of
9)
as
the
cut­
off
criterion
for
PBT
chemicals.
This
reviewer
also
suggested
it
would
have
been
just
as
easy
to
run
all
HAP
rather
than
use
a
process
of
selection
to
narrow
the
focus.
There
was
also
a
comment
that
while
the
IEM
model
provided
an
appropriate
basis
for
the
final
decision
that
inhalation
was
the
dominant
pathway,
there
was
a
need
for
more
documentation
of
the
IEM
model
inputs
and
options.
1
Air
Toxics
Risk
Assessment
Reference
Library:
Technical
Resources
Manual,
Vol.
I.,
EPA:
OAQPS,
EPA­
453­
K­
04­
001A
and
Facility­
Specific
Assessment,
Vol.
II,
EPA­
453­
K­
04­
001B,
April
2004.
These
documents
are
available
at
http:
www.
epa.
gov/
ttn/
fera/
risk_
atoxic.
html.

3
Response:
We
checked
the
SAMSON
data
and
found
that
the
precipitation
data
for
the
meteorological
sites
we
used
in
this
assessment
were
complete.
In
response
to
the
need
for
more
clarification,
we
added
specific
text
to
the
document
to
better
explain
the
selection
process
and
identified
the
references
which
contain
the
"
options
and
switches"
needed
to
run
the
IEM.
Had
the
multipathway
analysis
progressed
beyond
a
screening­
level
assessment
and
had
the
noninhalation
cancer
risk
and
hazard
added
more
to
the
total
risk
and
hazard,
we
would
have
added
more
detailed
multipathway
modeling
information.

One
reviewer
did
not
think
we
needed
to
use
a
two­
step
selection
process
for
determining
which
HAP
to
include
in
the
screening
multipathway
assessment.
The
approach
we
used
was
to
select
HAP
from
the
universe
of
HAP
known
to
be
emitted
from
coke
oven
facilities
that
were
found
on
EPA
listings
of
PBT
HAP.
This
method
is
consistent
with
our
current
approach
suggested
in
our
Technical
Resource
Manual.
1
That
is,
we
focused
on
the
subset
of
HAP
that
persist
and
may
bioaccumulate,
then
remove
them
from
further
analysis
if
they
are
shown
to
contribute
a
small
fraction
to
the
total
risk
estimate.
In
that
document,
we
suggest
a
list
of
HAP
that
should
be
assessed
via
a
multipathway
assessment.
Once
the
HAP
are
selected,
the
screening­
level
analyses
may
assume
relatively
high
exposure
factors
to
determine
whether
risk
associated
with
a
specific
pathway
appears
to
be
significant
enough
to
warrant
more
robust
analysis.
Subsequent
tiers
of
analysis,
using
more
realistic
exposure
factors
and
perhaps
sampling
and
analysis,
are
generally
undertaken
only
if
lower­
tier
analyses
continue
to
indicate
the
potential
for
risk.
The
analysis
we
conducted
for
coke
ovens
assumed
central
tendency
as
well
high­
end
exposure
factors.
When
the
risks
resulting
from
the
use
of
high­
end
factors
were
estimated,
we
determined
that
the
non­
inhalation
risk
was
less
by
an
order
of
magnitude
than
risks
due
to
inhalation
exposures.
Therefore,
we
did
not
perform
any
additional
tiers
or
refinements
of
the
assessments.
A
detailed
discussion
of
this
process
is
given
in
the
risk
assessment
document.

As
requested,
we
enhanced
the
text
with
more
explanation
of
the
choice
of
7
as
the
cut­
off
criterion.
In
our
methodology,
we
compared
the
coke
oven
emissions
constituent
HAP
list
with
the
EPA
list
(
see
the
risk
report
for
more
complete
information)
that
represents
the
relative
ranking
of
chemicals
based
on
their
persistence,
bioaccumulation
factor,
and
toxicity
(
PBT).
This
PBT
list
ranks
each
chemical
for
persistence,
bioaccumulation
and
toxicity
using
scores
ranging
from
1
to
3
for
each
characteristic;
consequently,
the
maximum
ranking
for
any
chemical
for
PBT
is
9.
Although
persistence,
bioaccumulation,
and
toxicity
are
predictors
of
potential
chronic
risk,
these
chemical
properties
and
their
scores
do
not
indicate
absolute
risk,
but
are
a
starting
point
in
assessing
their
risk.
To
screen
coke
oven
constituents
for
their
PBT
concern,
we
selected
a
cut­
off
score
of
7.
This
approach
ensured
that
HAP
of
interest
represented
at
least
a
moderately
high
concern
for
PBT.

Charge
Question
3:
This
analysis
models
coke
oven
emissions
to
generate
air
4
concentrations
and
deposition
rates
within
a
50
kilometer
(
km)
radius
around
each
facility.
Risk
distributions
are
calculated
using
the
2000
census
data
and
variations
in
exposure
frequency.
Overall
analysis
uncertainty
is
qualitatively
described
in
the
bias
table
in
the
report,
but
more
quantitative
uncertainty
was
estimated
only
for
the
emissions
from
the
sources
subject
to
the
1993
NESHAP,
which
are
considered
the
critical
element
in
this
assessment.
Only
variability
around
exposure
duration
is
presented.
Does
the
evaluation
and
discussion
of
uncertainty
adequately
characterize
the
uncertainty
and
variability
inherent
in
this
assessment
at
a
level
that
is
appropriate
for
the
level
of
the
decision
being
made,
(
i.
e.,
to
determine
the
need
for
and
level
of
a
risk­
based
rule,
recognizing
that
the
second
decision
involves
cost
and
feasibility
as
well
as
a
risk
component)?

Peer
Review
Comments:
One
reviewer
suggested
that
ISCST3
may
underestimate
the
long­
term
average
for
elevated
sources
in
flat­
terrain
(
based
on
recent
model
comparisons)
rather
than
overestimate
under
the
conditions
we
stated.
He
also
wanted
to
see
the
actual
values
(
rather
than
just
our
reference
to
the
GIRAS
data)
used
to
represent
roughness
in
the
model.
The
other
reviewers
commented
on
the
limited
scope
of
our
uncertainty
analysis
(
e.
g.,
focusing
on
the
emissions
from
the
MACT
1
sources
only).
One
reviewer
pointed
to
the
uncertainty
analysis
performed
for
the
utility
risk
assessment
as
a
possible
approach.
They
also
commented
that
the
current
analysis
does
not
provide
sufficient
basis
for
the
decision,
but
qualified
that
statement
(
in
their
response
to
Charge
Question
4)
by
saying
they
didn't
think
a
more
rigorous
uncertainty
analysis
would
have
affected
the
decision
made.

Response:
In
our
response
to
the
modeling
question,
we
did
not
make
any
adjustments
to
the
document.
The
ISC
is
EPA's
current
regulatory
model,
and
as
a
general
rule,
will
overestimate
risks
rather
then
underestimate
risk,
particularly
when
the
inputs
and
model
options
are
chosen
to
be
more
health
protective
as
they
were
in
the
coke
oven
risk
assessment.

In
response
to
the
comment
on
the
GIRAS
data,
we
added
extensive
footnotes
to
Table
D­
5
which
describe
the
input
values
and
assumptions
for
roughness
used
in
the
ISCST3
modeling.
The
input
value
we
used
wasn't
as
high
as
the
reviewer
thought
it
was,
but,
in
addition,
we
also
determined
that
the
impact
of
roughness
on
the
assessment
would
not
be
significant.

In
response
to
the
comment
on
uncertainty,
we
added
additional
text
to
the
document,
but
did
not
consider
re­
doing
the
analysis
based
on
the
reviewer's
suggestion.
In
the
utility
risk
assessment
which
the
reviewer
mentioned,
dispersion
and
exposure
parameters
were
evaluated
as
a
ratio,
(
i.
e.,
the
deterministic
value
used
was
divided
by
various
values
from
each
parameter
distribution).
When
this
was
done
for
the
previous
risk
assessment,
the
conclusion
was
that
the
baseline
deterministic
risk
assessment
results
were
conservative
estimates
of
risk
which
are
more
likely
to
overestimate
than
underestimate
risk.
The
deterministic
estimates
of
risk
ranged
from
the
71st
percentile
to
the
98th
percentile
when
compared
with
the
probabilistic
results.
The
95th
percentile
of
the
overall
distribution
was
found
to
be
roughly
twice
the
original
deterministic
risk
estimate
suggesting
that
the
deterministic
estimate
fell
within
the
range
of
a
typical
high­
end
5
estimate.
The
uncertainty
analysis
we
did
for
the
coke
oven
risk
assessment
was
not
as
encompassing
as
what
was
done
in
the
previous
risk
assessment,
(
i.
e.,
we
only
addressed
uncertainty
around
the
emissions
estimates),
but
the
results
of
this
analysis
suggested
that
our
estimates
of
emissions
may
be
as
low
as
2
times
or
as
high
as
3
times
the
emission
estimates
we
used
in
the
coke
oven
risk
assessment.
These
values
suggest
our
uncertainty
around
these
estimates
is
small.
When
considering
the
results
of
the
analysis
we
did
for
the
coke
oven
risk
assessment
and
the
results
of
previous
analyses
done
on
exposure
parameters,
our
conclusion
was
that
while
a
more
detailed
uncertainty
analysis
may
be
more
desirable,
it
is
not
expected
to
have
an
impact
on
the
risk
levels
such
that
a
change
in
any
regulatory
decision
would
be
anticipated.
The
reviewer
who
suggested
the
more
rigorous
analysis
also
came
up
with
the
same
conclusion.

Charge
Question
4:
Is
the
scope
of
this
assessment
appropriate
and
does
it
provide
sufficient
information
to
support
the
regulatory
decision?

Peer
Review
Comments:
The
reviewers
generally
felt
the
scope
of
the
assessment
was
adequate
but
that
some
clarifications
in
the
text
and
tables
would
help.
One
reviewer
suggested
strengthening
our
case
for
not
including
mercury
in
our
multipathway
screening
assessment.

Response:
We
added
more
text
and
information
to
various
tables
to
clarify
the
assessment
(
e.
g.,
footnotes
to
Appendix
I
tables
describing
the
data
used
for
calculating
cumulative
probabilities).
In
the
revised
risk
assessment
document,
we
conclude
that
mercury
should
not
be
assessed
in
the
multipathway
assessment
because
the
test
data
for
similar
coke
facilities
showed
mercury
emissions
to
be
below
detectable
levels
(
see
data
in
Appendix
C
).
This
result
was
supported
by
the
European
study
cited
in
the
risk
assessment
document
on
metals
emitted
from
coke
ovens
which
concluded
that
mercury
in
the
coal
was
transferred
to
the
byproduct
plant
and
removed
with
the
recovered
by­
products.
Based
on
these
data,
we
decided
not
to
assess
mercury
as
a
constituent
of
coke
oven
emissions.
In
responding
to
the
request
for
more
IEM
model
information,
we
would
have
expanded
the
document
with
additional
multipathway
modeling
information,
had
the
multipathway
analysis
progressed
beyond
a
screening­
level
assessment
and
had
the
non­
inhalation
cancer
risk
and
hazard
added
more
to
the
total
risk
and
hazard
.

Charge
Question
5:
Does
the
report
present
results
and
highlight
the
major
assumptions
and
uncertainties
in
a
clear,
concise,
and
transparent
manner?

Peer
Review
Comments:
The
reviewers
had
a
mixed
response
to
this
question.
One
said
the
report
"
presents
results
and
highlights
the
major
assumptions
and
uncertainties
in
a
clear,
concise,
and
transparent
manner",
but
suggested
improvement
in
documentation.
However,
another
reviewer
said
he
didn't
think
the
assumptions
were
well
described.
This
latter
comment
was
directed
more
towards
the
reports
organization
and
documentation
rather
than
the
completeness
of
the
information.

Response:
We
revised
the
document
to
respond
to
these
comments.
Some
2
Mercury
Study
Report
to
Congress.
EPA:
OAQPS.
EPA­
452/
R­
97­
004,
December
1997.
This
document
is
available
at
http://
www.
epa.
gov/
oar/
mercury.
html.

6
organizational
changes
were
made
to
give
the
reader
a
better
sense
of
the
sequence
of
assessment
activities,
(
e.
g.,
a
flowchart
was
added).
We
also
added
more
descriptive
text
and
footnotes
to
text
and
tables
to
help
explain
what
was
done
and
what
information
sources
we
used
(
e.
g.,
the
explanation
of
the
BLP/
ISC
modeling
was
enhanced
in
Appendices
D
and
E
with
more
text
and
tables
of
assumptions
and
modeling
options).
We
did
not
significantly
expand
the
discussion
of
the
multipathway
assessment
in
the
document.
The
IEM
model
is
very
data
intensive
and
although
we
have
included
many
tables
of
input
parameters,
we
didn't
add
more
details
on
the
IEM
model
"
options."
We
rely
on
the
many
references
to
other
analyses
which
used
the
IEM
model
as
the
source
of
these
data.
For
example,
in
the
initial
screen
we
used
a
default
value
for
waterbody
size,
flow
rate,
watershed
size,
and
other
parameters
based
on
the
health
protective
scenario
analyzed
in
the
Mercury
Report
to
Congress.
2
Had
the
multipathway
analysis
progressed
beyond
a
screening­
level
assessment
and
had
the
non­
inhalation
cancer
risk
and
hazard
added
more
to
the
total
risk
and
hazard,
we
would
have
expanded
the
current
document
with
additional
specific
multipathway
modeling
information.

Charge
Question
6:
Has
any
information
been
neglected
which
would
likely
impact
or
alter
the
regulatory
decision?

Peer
Reviewer
Comments:
The
reviewers
did
not
identify
any
information
that
was
neglected
and
which
would
have
a
likely
impact
on
the
regulatory
decision.
However,
they
reemphasized
earlier
comments
on
modeling
assumptions
and
the
need
for
better
documentation.

Response:
Responses
to
their
concerns
about
documentation
and
clarity
were
addressed
under
other
charge
questions.
Several
revisions
were
made
to
improve
clarity,
explain
modeling
assumptions,
and
to
provide
better
documentation.

Additional
Comments
and
Responses:
Two
of
the
reviewers
made
additional
comments
that
were
not
identified
with
a
specific
charge
question.
Many
of
these
were
editorial
in
nature
or
requested
additional
clarification.
These
were
addressed
directly
in
the
risk
assessment
document.
For
example,
a
flow
chart
was
included
describing
the
sequence
of
activities
undertaken
in
this
assessment.
Also,
we
added
an
outline
at
the
beginning
of
section
2
and
rewrote
the
text
to
match
that
outline.
We
also
made
revisions
to
the
risk
assessment
document
to
respond
to
their
suggestions
for
clarifying
the
relationship
between
the
BSO
components
and
the
overall
risk
coefficient
for
inhalation,
how
the
noncancer
hazard
was
determined,
how
risks
were
aggregated,
the
criterion
of
an
HI
value
of
0.2
for
screening,
how
risks
were
generated,
exposure
duration
assumptions,
apportioned
emissions
contributions
for
MACT
I
and
II
sources,
exclusion
of
populations
in
Canada,
and
why
50
kilometers
is
used
as
the
modeling
distance.
