
1
III.
HOW
TO
IDENTIFY
SOURCES
"
SUBJECT
TO
BART"

Once
you
have
compiled
your
list
of
BART­
eligible
sources,

you
need
to
determine
whether
(
1)
to
make
BART
determinations
for
all
of
them
or
(
2)
to
consider
exempting
some
of
them
from
BART
because
they
may
not
cause
or
contribute
to
any
visibility
impairment
in
a
Class
I
area.
If
you
decide
to
make
BART
determinations
for
all
the
BART­
eligible
sources
on
your
list
(
i.
e.,
that
all
of
them
should
be
"
subject
to
BART"),
you
should
make
those
determinations
by
applying
the
five
statutory
factors
discussed
in
Section
IV
below.

On
the
other
hand,
you
may
want
to
conduct
further
analysis
to
determine
whether
and
the
degree
to
which
a
particular
BARTeligible
source
or
group
of
sources
contributes
to
visibility
impairment
in
nearby
Class
I
areas.
If
your
analysis,
or
analysis
submitted
by
the
source,
shows
that
an
individual
source
or
group
of
sources
(
or
certain
pollutants
from
those
sources)

cannot
reasonably
be
anticipated
to
cause
or
contribute
to
any
visibility
impairment
in
a
Class
I
area,
then
you
do
not
need
to
make
BART
determinations
for
that
source
or
group
of
sources
(
or
for
certain
pollutants
from
those
sources).
In
such
a
case,
the
source
is
not
"
subject
to
BART"
and
you
do
not
need
to
apply
the
five
statutory
factors
to
make
a
BART
determination.

This
section
briefly
discusses
the
factors
you
should
consider
in
deciding
whether
to
make
BART
determinations
for
all
2
your
BART­
eligible
sources.
It
then
discusses
several
approaches
that
you
can
use
to
exempt
sources
from
BART
even
though
they
are
BART­
eligible.

A.
What
Factors
Should
I
Consider
in
Deciding
Whether
All
BARTEligible
Sources
Should
be
Subject
to
BART?

States
have
considerable
discretion
in
making
BART
determinations
and
in
determining
which
sources
should
be
subject
to
BART.
In
exercising
this
discretion,
you
may
want
to
consider
all
your
state's
air
quality
planning
needs,
including
the
need
to
meet
the
national
ambient
air
quality
standards
(
NAAQS)
for
ozone
and
PM2.5.
The
pollutants
that
are
primarily
responsible
for
visibility
impairment
(
SO2,
NOX,
and
PM)
are
also
the
primary
contributors
to
PM2.5
nonattainment.
In
addition,
NOX
is
an
ozone
precursor
that,
in
most
areas,
needs
to
be
controlled
during
the
ozone
season
to
reduce
the
formation
of
ground­
level
ozone.
You
may
not
want
to
invest
significant
resources
in
conducting
visibility
analysis
for
individual
BART­
eligible
sources
if
your
state
is
likely
to
control
SO2,
NOX,
or
PM
from
those
same
sources
in
order
to
meet
its
attainment
needs.

On
the
other
hand,
if
there
are
not
PM2.5
or
ozone
nonattainment
areas
in
your
state
or
in
nearby
states,
then
you
may
want
to
consider
various
approaches
that
would
allow
you
to
determine
that
your
BART­
eligible
sources
cannot
reasonably
be
anticipated
to
contribute
to
visibility
impairment
in
any
Class
I
3
area.
In
particular,
if
most
of
your
sources
are
relatively
small
NOX­
only
sources
that
are
located
relatively
far
away
from
Class
I
areas,
you
should
consider
the
approaches
discussed
below
that
would
allow
you
to
find
that
individual
sources
or
groups
of
sources
are
not
subject
to
BART.

B.
What
Steps
Do
I
Follow
to
Determine
Whether
A
Source
or
Group
of
Sources
Cause
or
Contribute
to
Visibility
Impairment
for
Purposes
of
BART?

1.
How
Do
I
Establish
a
Threshold?

One
of
the
first
steps
in
determining
whether
sources
cause
or
contribute
to
visibility
impairment
for
purposes
of
BART
is
to
establish
a
threshold
(
measured
in
deciviews)
against
which
to
measure
the
visibility
impact
of
one
or
more
sources.
A
single
source
that
is
responsible
for
a
1.0
deciview
change
or
more
should
be
considered
to
"
cause"
visibility
impairment;
a
source
that
causes
less
than
a
1.0
deciview
change
can
still
contribute
to
visibility
impairment
and
thus
be
subject
to
BART.

Because
of
varying
circumstances
affecting
different
Class
I
areas,
the
appropriate
threshold
for
determining
whether
a
source
"
contributes
to
any
visibility
impairment"
for
the
purposes
of
BART
may
reasonably
differ
across
States.
As
a
general
matter,

any
threshold
that
you
use
should
not
be
higher
than
0.5
deciviews,
and
could
be
as
low
as
0.1
deciviews
or
less.

In
setting
a
threshold
for
"
contribution,"
you
should
1
We
expect
that
regional
planning
organizations
will
have
modeling
information
that
identifies
sources
affecting
visibility
in
individual
class
I
areas,
and
the
magnitude
of
their
impacts.

4
consider
the
number
of
emissions
sources
affecting
the
Class
I
areas
at
issue
and
the
magnitude
of
the
individual
sources'

impacts.
1
In
general,
a
larger
number
of
sources
causing
impacts
in
a
class
I
area
would
warrant
a
lower
contribution
threshold.

Depending
on
the
circumstances
and
number
of
sources
affecting
a
Class
I
area
and
their
modeled
impacts,
you
should
consider
a
threshold
captures
the
BART­
eligible
sources
responsible
for
the
vast
majority
of
the
total
visibility
impacts
from
such
sources,

while
still
excluding
other
sources
with
relatively
small
impacts.

2.
What
Pollutants
Do
I
Need
to
Consider?

You
must
look
at
SO2,
NOx,
and
PM10
and
PM2.5
emissions
in
determining
whether
sources
cause
or
contribute
to
visibility
impairment.
Consistent
with
the
approach
for
identifying
your
BART­
eligible
sources,
you
do
not
need
to
consider
less
than
de
minimis
emissions
of
these
pollutants
from
a
source.

As
explained
in
Section
II,
you
must
use
your
best
judgement
to
determine
whether
VOC
or
ammonia
emissions
are
likely
to
have
an
impact
on
visibility
in
an
area.
In
addition,
you
may
use
PM10
as
an
indicator
for
particulate
matter
in
determining
whether
a
source
is
BART­
eligible.
In
determining
whether
a
5
source
contributes
to
visibility
impairment,
however,
you
should
distinguish
between
the
fine
and
coarse
particle
components
of
direct
particulate
emissions.
Although
both
fine
and
coarse
particulate
matter
contribute
to
visibility
impairment,
the
longrange
transport
of
fine
particles
is
of
particular
concern
in
the
formation
of
regional
haze.
Air
quality
modeling
results
used
in
the
BART
determination
will
provide
a
more
accurate
prediction
of
a
source's
impact
on
visibility
if
the
inputs
into
the
model
account
for
the
relative
particle
size
of
any
directly
emitted
particulate
matter
(
i.
e.
PM10
vs.
PM2.5).

3.
What
Kind
of
Modeling
Should
I
Conduct
to
Determine
Which
Sources
and
Pollutants
Need
Not
Be
Subject
to
BART?

This
section
presents
several
options
for
determining
that
certain
sources
need
not
be
subject
to
BART.
These
options
rely
on
different
modeling
and/
or
emissions
analysis
approaches.
They
are
provided
for
your
guidance.
You
may
also
use
other
reasonable
approaches
for
analyzing
the
visibility
impacts
of
an
individual
source
or
group
of
sources.

Option
1:
Individual
Source
Attribution
Approach
(
Dispersion
Modeling)

You
can
use
dispersion
modeling
to
determine
that
an
individual
source
cannot
reasonably
be
anticipated
to
cause
or
contribute
to
visibility
impairment
in
a
Class
I
area
and
thus
is
not
subject
to
BART.
You
can
conduct
this
modeling
yourself
or
2
In
determining
whether
sources
of
particulate
matter
are
subject
to
BART,
you
should
try
to
estimate
the
PM2.5
fraction
of
direct
particulate
emissions
as
correctly
as
possible.
This
is
because,
as
discussed
in
section
II.
A.
3.
above,
long­
range
transport
of
fine
particles
is
of
particular
concern
in
the
formation
of
regional
haze,
and
because
we
believe
that
air
quality
modeling
results
will
be
more
meaningful
if
your
inputs
account
for
the
relative
particle
size
of
directly
emitted
particulate
matter
(
e.
g.
PM10
vs.
PM2.5).

3
The
model
code
and
its
documentation
are
available
at
no
cost
for
download
from
http://
www.
epa.
gov/
scram001/
tt22.
htm#
calpuff.

6
allow
the
submission
of
modeling
from
an
outside
expert.
Under
this
option,
you
can
analyze
an
individual
source's
impact
on
visibility
as
a
result
of
its
emissions
of
SO2,
NOx
and
direct
particulate
matter
emissions
(
PM10
and
PM2.5)
2
Dispersion
modeling
cannot
currently
be
used
to
estimate
the
predicted
impacts
on
visibility
from
an
individual
source's
emissions
of
VOC
or
ammonia.
You
may
use
a
more
qualitative
assessment
to
determine
on
a
case­
by­
case
basis
which
sources
of
VOC
or
ammonia
emissions
may
be
likely
to
impair
visibility
and
should
therefore
be
subject
to
BART
review,
as
explained
in
section
II.
A.
3.
above.

You
can
use
CALPUFF3,
or
another
EPA
approved
model,
to
predict
the
visibility
impacts
from
a
single
source
at
a
Class
I
area.
CALPUFF
is
the
best
regulatory
modeling
application
currently
available
for
predicting
a
single
source's
contribution
to
visibility
impairment
and
is
currently
the
only
EPA­
approved
model
for
use
in
estimating
single
source
pollutant
4
The
Guideline
on
Air
Quality
Models
addresses
the
regulatory
application
of
air
quality
models
for
assessing
criteria
pollutants
under
the
CAA,
and
describes
further
the
procedures
for
using
the
CALPUFF
model,
as
well
as
for
obtaining
approval
for
the
use
of
other,
nonguideline
models.

7
concentrations
resulting
from
the
long
range
transport
of
primary
pollutants.
4
It
can
also
be
used
for
some
other
purposes,
such
as
the
visibility
assessments
addressed
in
today's
rule,
to
account
for
the
chemical
transformation
of
SO2
and
NOx.

There
are
several
steps
for
making
an
individual
source
attribution
using
a
dispersion
model:

a.
Submit
a
modeling
protocol
to
EPA.

If
you
are
doing
the
modeling,
you
should
submit
a
modeling
protocol
to
EPA
in
advance
so
that
any
issues
can
be
discussed
before
you
run
the
models.
Similarly,
if
you
are
allowing
the
source
owner
or
operator
to
do
the
modeling,
it
should
prepare
and
submit
a
modeling
protocol
in
advance
to
assure
that
relevant
technical
issues
are
addressed
up
front.

Some
critical
items
to
include
in
the
protocol
are
the
meteorological
and
terrain
data
that
will
be
used,
as
well
as
the
source­
specific
information
(
stack
height,
temperature,
exit
velocity,
elevation,
and
emission
rates
of
applicable
pollutants),
and
receptor
data
from
appropriate
Class
I
areas.

We
recommend
following
EPA's
Interagency
Workgroup
on
Air
Quality
Modeling
(
IWAQM)
Phase
2
Summary
Report
and
Recommendations
for
5
Interagency
Workgroup
on
Air
Quality
Modelig
(
IWAQM)
Phase
2
Summary
Report
and
Recommendations
for
Modeling
Long
Range
Transport
Impacts,
U.
S.
Environmental
Protection
Agency,
EPA­
454/
R­
98­
019,
December
1998.

8
Modeling
Long
Range
Transport
Impacts5
for
parameter
settings
and
meteorological
data
inputs.
You
may
use
other
settings
from
those
in
IWAQM,
but
you
should
identify
these
settings
and
explain
your
selection
of
these
settings.

One
important
element
of
the
protocol
is
in
establishing
the
receptors
that
will
be
used
in
the
model.
The
receptors
that
you
use
should
be
located
in
the
nearest
Class
I
area
with
sufficient
density
to
identify
the
maximum
impact
of
the
source.
(
In
general,
one­
kilometer
spacing
should
be
used.)
For
other
Class
I
areas
in
relatively
close
proximity
to
a
BART­
eligible
source,

you
may
model
a
few
strategic
receptors
to
determine
whether
impacts
at
those
areas
may
be
greater
than
at
the
nearest
Class
I
area.
For
example,
you
might
chose
to
locate
receptors
at
these
areas
at
the
closest
point
to
the
source,
at
the
highest
and
lowest
elevation
in
the
Class
I
area,
at
the
IMPROVE
monitor,
and
at
the
approximate
expected
plume
release
height.
If
the
highest
modeled
impacts
are
observed
at
the
nearest
Class
I
area,
you
may
choose
not
to
analyze
the
other
Class
I
areas
any
further,
as
additional
analyses
might
be
unwarranted.
In
any
event,
a
full
field
of
receptors
should
be
modeled
for
the
Class
I
area
with
the
highest
impact.
9
You
should
bear
in
mind
that
some
receptors
within
the
relevant
Class
I
area
may
be
less
than
50
km
from
the
source
while
other
receptors
within
that
same
Class
I
area
may
be
greater
than
50
km
from
the
same
source.
As
indicated
by
the
Guideline
on
Air
Quality
Models,
this
situation
may
call
for
the
use
of
two
different
modeling
approaches
for
the
same
Class
I
area
and
source,
depending
upon
the
State's
chosen
method
for
modeling
sources
less
than
50
km.
In
situations
where
you
are
assessing
visibility
impacts
for
source­
receptor
distances
less
than
50
km,
you
should
use
expert
modeling
judgment
in
determining
visibility
impacts,
giving
consideration
to
both
CALPUFF
and
other
EPA­
approved
methods.

b.
Run
the
model
in
accordance
with
the
accepted
protocol
and
compare
the
predicted
visibility
impacts
with
your
threshold
for
"
contribution."

You
should
calculate
daily
visibility
values
for
each
receptor
as
the
change
in
deciviews
compared
against
natural
visibility
conditions.
You
can
use
EPA's
"
Guidance
for
Estimating
Natural
Visibility
Conditions
Under
the
Regional
Haze
Rule,"
EPA­
454/
B­
03­
005
(
September
2003)
in
making
this
calculation.
To
determine
whether
a
source
may
reasonably
be
anticipated
to
cause
or
contribute
to
visibility
impairment
at
Class
I
area,
you
then
compare
the
impacts
predicted
by
the
model
against
the
threshold
that
you
have
selected.
10
In
making
this
comparison,
you
should
note
that
the
modeling
approach
we
recommend
includes
several
features
that
likely
understate
the
potential
visibility
impacts
of
the
source
being
modeled.
The
emissions
estimates
used
in
the
models
are
intended
to
reflect
steady­
state
operating
conditions
during
periods
of
high
capacity
utilization.
We
do
not
generally
recommend
that
emissions
reflecting
periods
of
start­
up,
shutdown,
and
malfunction
be
used,
as
such
emission
rates
could
produce
higher
than
normal
impacts
than
would
be
typical
of
most
facilities.
In
addition,
the
monthly
average
relative
humidity
is
used,
rather
than
the
daily
average
humidity
 
an
approach
that
effectively
lowers
the
peak
values
in
daily
model
averages.

For
these
reasons,
if
you
use
the
modeling
approach
we
recommend,
you
should
compare
your
"
contribution"
threshold
against
either
the
maximum
daily
value
or
the
99th
percentile
of
values.
Comparing
values
to
the
99th
percentile
of
values
effectively
excludes
approximately
4
days
per
year,
or
18­
20
days
in
any
5­
year
period,
with
the
worst
visibility.
Such
an
approach
affords
added
protection
against
the
possibility
that
poor
visibility
might
be
caused
by
unusual
meteorology
and
thus,

in
our
view,
represents
a
reasonable
lower­
bound
for
comparison.

If
the
maximum
daily
visibility
value
or
the
99th
percentile
value
from
your
modeling
is
less
than
your
threshold,
then
you
may
conclude
that
the
source
does
not
contribute
to
visibility
11
impairment
and
is
not
subject
to
BART.

You
may
wish
to
use
another
reasonable
approach,
such
as
use
of
the
maximum
24­
hr
potential
emission
rate
of
the
source
and
more
realistic
assumptions
regarding
meteorology
(
e.
g.
daily
relative
humidity
values
rather
than
a
30
day
average),
comparing
results
to
some
other
value
than
the
maximum
24­
hour
visibility
value
or
99th
percentile.
If
so,
you
must
consult
with
EPA,

providing
specific
facts
supporting
any
alternative
approach.

Option
2:
Use
of
Model
Plants
to
Exempt
Individual
Sources
with
Common
Characteristics.

Under
this
option,
analysis
of
model
plants
could
be
used
to
exempt
certain
BART­
eligible
sources
that
share
specific
characteristics.
It
may
be
most
useful
to
use
this
type
of
analysis
to
identify
the
types
of
small
sources
that
do
not
cause
or
contribute
to
visibility
impairment
for
purposes
of
BART,
and
thus
should
not
be
subject
to
a
BART
review.
Due
to
the
unique
characteristics
of
different
Class
I
areas,
you
should
use
special
care
to
ensure
that
the
criteria
you
develop
are
appropriate
for
all
applicable
cases.

In
carrying
out
this
approach,
you
could
use
modeling
analyses
of
representative
plants
to
sort
specific
sources
with
important
common
characteristics.
Based
on
these
analyses,
you
may
find
that
certain
types
of
sources
are
reasonably
anticipated
to
cause
or
contribute
to
visibility
impairment.
You
6
[
insert
reference
to
Mark's
docket
memo]

12
could
then
choose
to
categorically
require
those
types
of
sources
to
undergo
a
BART
determination.
Conversely,
you
may
find
based
on
representative
plant
analyses
that
certain
types
of
sources
are
clearly
not
reasonably
anticipated
to
cause
or
contribute
to
visibility
impairment.
To
do
this,
you
should
conduct
your
own
modeling
to
establish
emission
levels
and
distances
from
Class
I
areas
on
which
you
can
rely
to
exempt
sources
with
those
characteristics.
For
example,
based
on
your
modeling
you
might
choose
to
exempt
all
NOx­
only
sources
that
emit
less
than
a
certain
amount
per
year
(
such
as
500
tons),
and
which
are
located
a
certain
distance
from
a
Class
I
area
(
such
as
100
kilometers).

You
could
then
choose
to
categorically
exempt
such
sources
from
BART.
In
designing
your
analyses,
you
should
ensure
that
all
sources
of
that
type
would
not
contribute
any
more
to
visibility
impairment
than
the
modeled
representative
facility.
In
doing
such,
you
should
also
consider
the
number
of
sources
with
these
characteristics
contributing
to
visibility
impairment
in
your
State.

Our
analyses
of
visibility
impacts
from
model
plants
provide
a
useful
example
of
the
type
of
analyses
that
might
be
used
to
exempt
categories
of
sources
from
BART.
6
In
our
analysis,
we
developed
model
plants
(
EGUs
and
non­
EGUs),
with
representative
plume
and
stack
characteristics,
for
use
in
considering
the
13
visibility
impact
from
emission
sources
of
different
sizes
and
compositions
at
distances
of
50,
100
and
200
kilometers
from
two
hypothetical
Class
I
areas
(
one
in
the
East
and
one
in
the
West).

Since
the
plume
and
stack
characteristics
of
these
model
plants
were
developed
considering
the
broad
range
of
sources
within
the
EGU
and
non­
EGU
categories,
they
do
not
necessarily
represent
any
specific
plant
in
the
United
States.
However,
the
results
of
these
analyses
may
be
instructive
in
the
development
of
an
exemption
process
for
any
Class
I
area.

In
our
analyses,
for
example,
a
larger
EGU
in
the
eastern
domain
(
see
Run
#
9
in
the
example
runs
where
the
SO2
emission
rate
is
30,000
TPY
and
NOx
is
10,000
TPY)
would
almost
certainly
contribute
to
visibility
impairment
and
should
be
subject
to
BART.
On
the
other
hand,
a
representative
EGU
in
the
east
or
west
with
much
lower
SO2
emission
rates
(
well
controlled
for
SO2
for
example)
may
suggest
that
similar
EGUs
could
be
exempt
from
BART.
You
should
note
that
the
model
plants
in
our
analyses
that
represented
industrial
boilers
or
other
BART­
eligible
source
categories
with
similar
emission
profiles
to
EGUs,
in
some
instances,
have
comparable
or
higher
visibility
impacts
than
the
EGUs
(
especially
for
closer
distances
to
the
Class
I
area
­
see
Runs
#
12
and
#
13).
This
may
be
due
to
relatively
lower
stack
heights
and
flow
rates.

In
a
manner
similar
to
these
analysis
and
using
information
14
from
model
plant
runs
that
take
into
account
local,
regional,
and
other
relevant
factors
(
such
as
meteorology,
sulfur
dioxide,

nitrogen
dioxide,
and
ammonia),
you
could
compare
your
sources
to
the
model
plants,
make
appropriate
adjustments
to
account
for
small
differences
in
emissions
and
stack
parameters,
and
construct
an
attribution
process
to
apply
to
your
BART
eligible
sources.
Any
State
wishing
to
pursue
this
approach
should
discuss
it,
along
with
a
proposed
protocol
for
modeling,
with
your
appropriate
Regional
Office
before
undertaking
the
exercise.

In
preparing
our
hypothetical
examples,
we
have
made
a
number
of
assumptions
and
exercised
certain
modeling
choices;

some
of
these
have
a
tendency
to
lend
conservatism
to
the
results,
overstating
the
likely
impacts,
while
others
may
understate
the
modeling
results.
On
balance,
when
all
of
these
factors
are
considered,
we
believe
that
our
examples
reflect
realistic
treatments
of
the
situations
being
modeled.

Option
3:
Cumulative
Modeling
to
Show
that
No
sources
in
a
State
are
subject
to
BART
You
may
also
submit
to
EPA
a
demonstration,
based
on
an
analysis
of
overall
visibility
impacts,
that
the
sum
of
all
emissions
from
BART­
eligible
sources
in
your
State
do
not
cause
or
contribute
to
any
visibility
impairment
in
a
Class
I
area
and
thus
no
source
should
be
subject
to
BART.
You
may
do
this
on
a
pollutant
by
pollutant
basis
or
for
all
visibility­
impairing
15
pollutants
to
determine
if
emissions
from
these
sources
contribute
to
visibility
impairment.

For
example,
emissions
of
SO2
from
your
BART­
eligible
sources
may
clearly
cause
or
contribute
to
visibility
impairment,

while
direct
emissions
of
PM2.5
from
these
sources
may
not
contribute
to
impairment.
If
you
can
make
such
a
demonstration,

then
you
may
reasonably
conclude
that
none
of
your
BART­
eligible
sources
are
subject
to
BART
for
a
particular
pollutant
or
pollutants.
As
noted
above,
your
demonstration
should
take
into
account
the
interactions
among
pollutants
and
their
resulting
impacts
on
visibility
before
making
any
pollutant­
specific
determinations.

Analyses
may
be
conducted
using
several
alternative
modeling
approaches.
First,
you
may
use
the
CALPUFF
or
another
EPAapproved
model
as
described
in
Option
1
to
evaluate
the
impacts
of
individual
sources
on
downwind
Class
I
areas,
aggregating
those
impacts
to
determine
the
collective
contribution
of
all
BART­
eligible
sources
to
visibility
impairment.
You
may
also
use
a
photochemical
grid
model.
As
a
general
matter,
the
larger
the
number
of
sources
being
modeled,
the
more
appropriate
it
may
be
to
use
a
photochemical
grid
model.
However,
because
such
models
are
significantly
less
sensitive
than
dispersion
models
to
the
contributions
of
one
or
a
few
sources,
as
well
as
to
the
interactions
among
sources
that
are
widely
distributed
16
geographically,
if
you
wish
to
use
a
grid
model,
you
should
consult
with
the
appropriate
EPA
Regional
Office
to
develop
an
appropriate
modeling
protocol.
