Walt
Stevenson/
RTP/
USEPA/
US
04/
14/
2006
10:
40
AM
To
edmond_
toy@
omb.
eop.
gov
cc
RobertJ
Wayland/
RTP/
USEPA/
US@
EPA,
Brian
Shrager/
RTP/
USEPA/
US@
EPA
Subject
UCL
99.7
and
99
/
LMWC
5­
ye
review
Edmond
Attached
below
is
the
revised
text
we
have
used
to
hopefully
better
explain
the
UCL
issue.

take
care
Walt
H.
What
emission
variability
factor
is
appropriate?

Although
most
commenters
and
EPA
used
similar
statistical
methodology,
differences
were
identified
in
assumptions
used
to
develop
emission
variability
factors
(
in
particular,
the
upper
confidence
limit
(
UCL)).
The
UCL
addresses
how
often,
on
average,
one
would
estimate
that
an
emissions
level
may
exceed
a
certain
value
(
the
standard).

For
analysis
of
CEMS
data,
such
as
sulfur
dioxide
or
nitrogen
oxides,
where
365
tests
(
24­
hr
CEMS
average)
are
conducted
per
year,
EPA
and
commenters
agreed
the
emission
limit
should
be
set
at
a
level
that
would
be
expected
to
be
exceeded
only
once
per
year
at
a
well­
operated
MWC
plant.

Once
per
year
translates
into
a
99.7
percent
UCL
(
364/
365
=
0.997
or
99.7
percent).
A
number
of
commenters
suggested
the
use
of
a
99.7
percent
UCL
for
development
of
limits
using
both
CEMS
data
(
sulfur
dioxide
and
nitrogen
oxide)

and
stack
test
data
(
cadmium,
lead,
mercury,
particulate
matter,
dioxin,
and
hydrogen
chloride).
This
is
one
area
where
EPA
disagrees
with
these
commenters.
EPA
concludes
a
different
assumption
is
appropriate
for
stack
test
compliance
procedures.
For
compliance
determination
by
stack
test,
EPA
used
a
different
and
lower
UCL.
One
must
remember
when
using
a
stack
test
for
determining
compliance
(
cadmium,
lead,
mercury,
particulate
matter,
dioxin,
and
hydrogen
chloride),
only
one
stack
test
is
conducted
per
year,
in
contrast
to
the
continuous
measurements
provided
by
CEMS.
Over
the
30­
year
life
of
a
facility,
only
30
stack
tests
would
have
occurred.
This
needs
to
be
reflected
in
the
emission
limits.
The
actual
stringency
of
an
emissions
limit
is
affected
by
how
often
one
tests.
In
fact,
using
the
same
data
distribution,
same
emission
limit,
and
same
UCL,
the
stringency
of
standard
is
affected
by
adjusting
just
the
testing
schedule.
For
example,
assuming
a
99.7
UCL
,
if
one
tests
every
day
using
a
CEMS
over
the
30
year
life
of
an
MWC
as
compared
to
testing
only
once
per
year
over
the
same
30
year
life
of
the
MWC,
one
will
find
the
probability
of
exceeding
the
standard
noticably
higher
with
the
CEMS.
In
this
example,
one
would
in
fact
find
the
probability
of
exceeding
the
standard
about
ten
times
as
high
if
CEMS
are
used.
((
1­(.
997^
30x365))/(
1­
(.
997^
30))=
11.6
)
In
consideration
of
this,
EPA
reduced
the
UCL
for
stack
test
analysis
to
99
percent.
Even
lower
UCL
could
be
considered,
however,
EPA
reviewed
actual
year
2000
to
2005
data
to
judge
the
reasonableness
of
the
estimates
and
conclude
99
percent
UCL
presented
reasonable
estimates
for
stack
test
pollutants.
