FEASIBILITY
ANALYSIS
OF
MERCURY
MONITORING
TECHNIQUES
12
May
2004
Draft
Report
Infrastructure,
buildings,
environment,

communications
U.
S.
Environmental
Protection
Agency
Clean
Air
Markets
Division
FEASIBILITY
ANALYSIS
OF
MERCURY
MONITORING
TECHNIQUES
Draft
Report
Brent
Hall
Project
Manager
Wojciech
Jozewicz
Project
Advisor
Laura
Nessley
Quality
Assurance/
Quality
Control
Specialist
Prepared
for:

U.
S.
Environmental
Protection
Agency
Clean
Air
Markets
Division
Dr.
Ruben
D.
Deza
Prepared
by:

ARCADIS
G&
M,
Inc.

4915
Prospectus
Drive
Suite
F
Durham
North
Carolina
27713
Tel
919
544
4535
Fax
919
544
5690
Our
Ref.:

RN096402.0001.00002
Date:

12
May
2004
i
Table
of
Contents
CONTENTS
List
of
Acronyms
iii
1.
Introduction
1
1.1
Scope
1
1.2
Regulatory
Framework
2
1.2.1
Precision
Criterion
4
1.2.2
Reliability
Criterion
4
1.2.3
Accessibility
and
Timeliness
Criteria
5
1.2.4
Summary
of
Subparts
F
and
G
6
1.3
Methods
8
1.3.1
Ontario
Hydro
Method
8
1.3.2
CEMS
9
1.3.3
Sorbent
Trap
Method
10
1.3.4
Fuel
Sampling
11
2.
Systems
Comparison
14
2.1
Precision
14
2.1.1
Variance
(
F
Test)
14
2.1.2
Correlation
Analysis
16
2.1.3
Bias
(
t­
test)
17
2.2
Reliability
19
2.3
Accessibility
and
Timeliness
21
2.3.1
Ontario
Hydro
Method
21
2.3.2
CEMS
21
2.3.3
Sorbent
Trap
Method
22
2.3.4
Fuel
Sampling
23
2.4
Missing
Data
Substitution
25
ii
Table
of
Contents
2.5
Qualitative
Comparison
of
CEMS
and
Sorbent
Traps
26
3.
Summary
28
3.1
Precision
28
3.2
Reliability
28
3.3
Accessibility
and
timeliness
29
iii
Table
of
Contents
List
of
Acronyms
ng/
L
Nanograms
Per
Liter
AAS
Atomic
Absorption
Spectroscopy
AMS
Alternative
Monitoring
System
ASTM
American
Society
for
Testing
and
Materials
CEMS
Continuous
Emission
Monitoring
System
CFR
Code
of
Federal
Register
CMS
Continuous
Monitoring
System
CO2
Carbon
Dioxide
CRM
Certified
Reference
Material
CSA
Clear
Skies
Act
CVAAS
Cold
Vapor
Atomic
Absorption
Spectroscopy
CVAFS
Cold
Vapor
Atomic
Fluorescence
Spectroscopy
DQI
Daily
Quality
Indicator
EPA
U.
S.
Environmental
Protection
Agency
H2O2
Hydrogen
Peroxide
Hg
Mercury
ICR
Information
Collection
Request
ISO
International
Organization
for
Standardization
KCl
Potassium
Chloride
KMnO3
Potassium
Permanganate
LOI
Loss
on
Ignition
iv
Table
of
Contents
ME
Measurement
Error
NIST
National
Institute
of
Standards
&
Technology
NOX
Nitrogen
Oxide
OH
Ontario
Hydro
PS
Performance
Standard
QA
Quality
Assurance
QC
Quality
Control
RA
Relative
Accuracy
RATA
Relative
Accuracy
Test
Audit
RDF
Refuse
Derived
Fuel
RPD
Relative
Percent
Difference
RSD
Relative
Standard
Deviation
SO2
Sulfur
Dioxide
U.
S.
United
States
USGS
United
States
Geological
Survey
1
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
1.
Introduction
1.1
Scope
This
report
evaluates
and
compares
the
various
mercury
(
Hg)
emission
monitoring
methods
in
light
of
potential
qualitative
performance
in
a
Hg
emissions
quantification
program
such
as
a
market­
based
cap­
and­
trade
program.
Such
an
evaluation
must
be
based
on
certain,
pre­
defined
criteria
or
attributes.
In
this
report,
the
methods
are
evaluated
on
the
basis
of
the
attributes
defined
in
40
CFR
75,
which
include
precision,
reliability,
accessibility,
and
timeliness.
The
40
CFR
75
pertains
to
the
pollutants
sulfur
dioxide
(
SO2)
and
nitrogen
oxide
(
NOX)
(
as
well
as
carbon
dioxide
[
CO2],
opacity
and
volumetric
flow)
but
it
does
not
address
mercury.

According
to
40
CFR
75.40,
the
owner
of
an
affected
unit
is
required
to
install
a
continuous
emissions
monitoring
systems
(
CEMS),
but
may
apply
to
the
Environmental
Protection
Agency
(
EPA)
administrator
for
approval
of
an
alternative
monitoring
system
(
AMS)
to
determine
average
hourly
emission
data
for
SO2
and
NOX
and/
or
volumetric
flow
by
demonstrating
that
the
AMS
has
the
same
or
better
precision,
reliability,
accessibility,
and
timeliness
as
that
provided
by
the
CEMS.
Subsequently,
AMSs
are
compared
to
the
CEMS
method,
where
the
CEMS
can
be
seen
as
a
"
base"
system.
AMSs
may
also
be
compared
to
a
parallel
operating
reference
method.
Thus,
40
CFR
75.40
is
written
for
comparison
of
one
AMS
against
an
existing
CEMS
(
or
one
reference
method).
Some
Hg
CEMS
are
commercially
available
and
in
refinement
stage,
while
others
are
in
their
infancy
stage;
however,
all
are
currently
referenced
with
American
Society
for
Testing
and
Materials
(
ASTM)
D6784­
02,
Standard
Test
Method
for
Elemental,
Oxidized,
Particle­
Bound
and
Total
Mercury
in
Flue
Gas
Generated
from
Coal­
Fired
Stationary
Sources,
or
the
Ontario
Hydro
(
OH)
Method.
Accordingly,
for
mercury,
no
"
base"
system
has
been
established
at
this
time.

In
this
report,
not
two,
but
four
mercury
"
monitoring
methods"
are
evaluated.
These
include:
CEMS,
sorbent
trap
sampling,
fuel
sampling,
and
the
mercury
reference
method,
Ontario
Hydro.
These
are
introduced
in
Section
1.3.
2
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
It
should
be
pointed
out
that
fuel
sampling
is
not
a
monitoring
method,
nor
is
it
a
continuous
method,
but
it
is
of
interest
as
a
sampling
method.
1
Sorbent
trap
sampling
could
be
considered
as
a
semi­
continuous
method,
because
it
is
unable
to
provide
continuous
independent
samples.
Instead,
it
provides
a
continuous
composite
sample
that
is
analyzed
by
an
off­
site
laboratory.

1.2
Regulatory
Framework
On
December
15,
2003,
EPA
proposed
new
regulations
on
mercury
emissions
for
Coal­
Fired
Electric
Utility
Steam
Generating
Units.
Under
the
proposed
regulation
§
63.10007,
affected
units
must
monitor
Hg
emissions
with
a
Continuous
Monitoring
System
(
CMS)
comprised
of
CEMS
or
Method
324
(
Sorbent
Trap
Sampling).
The
proposed
regulation
does
not
appear
to
specifically
exclude
use
of
alternate
CMS
and
provides
an
avenue
for
substitution
of
an
alternate
CMS
for
the
listed
CMSs.

In
the
proposed
Clear
Skies
Act
of
2003
(
CSA),
CEMS
were
proposed
as
a
way
to
quantify
Hg
emissions
for
a
market
driven
cap
and
trade
system.
Under
the
CSA,
the
administrator
would
specify
requirements
for
CEMS.
Additionally,
any
alternative
monitoring
system
could
be
used
as
long
as
it
provides
comparable
precision,
reliability,
accessibility,
and
timeliness
as
a
CEMS.
These
performance
criteria
are
the
same
indicators
used
in
evaluating
alternative
monitoring
systems
under
40
CFR
75.

The
regulations
established
under
40
CFR
75
include
general
requirements
for
the
installation,
certification,
operation,
and
maintenance
of
continuous
emission
or
opacity
monitoring
systems
and
specific
requirements
for
the
monitoring
of
SO2
emissions,
volumetric
flow,
NOX
emissions,
opacity,
CO2
emissions
and
SO2
emissions
removal.
Specifications
for
the
installation
and
performance
of
CEMS,
certification
tests
and
procedures,
and
quality
assurance
tests
and
procedures
are
included
in
appendices
A
and
B
to
40
CFR
75.

1
It
may
be
possible
that
in
the
future,
continuous
analytical
technologies
become
available
that
can
analyze
mercury
in
solid
samples.
3
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
Proposed
amendments
to
40
CFR
Part
75
set
forth
the
specific
monitoring
and
reporting
requirements
for
Hg
mass
emissions
and
include
the
additional
provisions
necessary
for
a
cap
and
trade
program.
Part
75
is
used
in
both
the
Acid
Rain
and
the
NOx
Budget
Trading
programs,
and
most
sources
affected
by
this
rulemaking
are
already
meeting
the
requirements
of
Part
75
for
one
or
both
of
those
programs.

In
order
to
ensure
program
integrity,
EPA
proposes
to
require
states
to
include
year
round
Part
75
monitoring
and
reporting
for
Hg
for
all
sources.
Proposed
deadlines
for
monitor
certification
and
other
details
are
specified
in
the
model
trading
rule.
EPA
believes
that
emissions
will
then
be
accurately
and
consistently
monitored
and
reported
from
unit
to
unit
and
from
State
to
State.

Part
75
also
specifies
reporting
requirements.
As
is
currently
required
for
sources
subject
to
both
the
Acid
Rain
program
and
the
NOx
Budget
Trading
program,
EPA
proposes
to
require
year
round
reporting
of
emissions
and
monitoring
data
from
each
unit
at
each
affected
facility.
Similarly,
this
data
would
be
provided
to
EPA
on
a
quarterly
basis
in
a
format
specified
by
the
Agency
and
submitted
to
EPA
electronically
using
EPA
provided
software.
EPA
has
found
this
centralized
reporting
requirement
necessary
to
ensure
consistent
review,
checking,
and
posting
of
the
emissions
and
monitoring
data
at
all
affected
sources,
which
contributes
to
the
integrity
and
efficiency
of
the
trading
program.

Criteria
for
alternative
monitoring
systems
and
provisions
to
account
for
missing
data
from
certified
CEMS
or
approved
alternative
monitoring
systems
are
found
in
40
CFR
75.41­
44.
In
these
sections,
the
alternative
monitoring
system
is
compared
to
a
"
base"
CEMS,
both
installed
and
operating
in
parallel
at
the
same
site.
The
base
system
is
defined
as
a
parallel
operating,
fully
certified
CEMS
or
reference
method.
In
order
to
apply
the
"
alternative"
system,
it
must
be
demonstrated
that
the
AMS
has
the
same
or
better:
precision,
reliability,
accessibility,
and
timeliness
as
that
provided
by
the
original
"
base"
CEMS.

Of
particular
importance
for
the
accessibility
and
timeliness
criteria
are
Subpart
F,
record
keeping
requirements
and
Subpart
G,
reporting
requirements,
which
are
briefly
discussed
in
the
ensuing
section.
However,
it
should
be
kept
in
mind
that
40
CFR
75
was
published
in
November
1990.
Since
then,
much
has
changed
in
regard
to
record
keeping
and
reporting
techniques
as
a
result
of
advances
in
data
acquisition,
handling,
and
storage.
New
techniques
have
become
available
that,
in
addition
to
continuous
monitoring,
include
real
time
data
processing
and
near­
real
time
output
in
tabular
as
well
as
graphic
format.
Also
the
capacity
for
electronic
data
storage
and
handling
has
become
greatly
enhanced.
For
these
reasons,
it
is
4
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
possible
that
certain
sections
of
Subparts
F
and
G
are
not
pertinent
to
the
existing
situation.

Precision,
reliability,
accessibility
and
timeliness
are
discussed
below.

1.2.1
Precision
Criterion
In
40
CFR
75.41(
a),
precision
is
determined
by
performance
of
an
F
test,
a
correlation
analysis,
and
a
t­
test
to
address
bias.
The
F
test
and
correlation
analysis
must
be
conducted
according
to
procedures
spelled
out
in
40
CFR
75.41.
The
procedure
for
the
t­
test
is
included
in
40
CFR
75,
Appendix
A,
Specifications
and
Test
Procedures.
The
CFR
precision
indicator
appears
to
include
both
precision
and
bias,
and
probably
would
now
be
called
"
accuracy"
using
EPA's
more
current
terminology
for
data
quality
indicators.

1.2.2
Reliability
Criterion
Reliability
should
be
demonstrated
by
showing
that
the
AMS
is
capable
of
providing
valid
1­
hour
averages
for
95.0
percent
or
more
of
the
unit
operating
hours
over
a
one­
year
period
and
that
the
system
meets
the
applicable
requirements
of
40
CFR
75,
Appendix
B,
Quality
Assurance
(
QA)
and
Quality
Control
(
QC)
procedures.

40
CFR
75,
Appendix
B
includes
the
following
main
topics:

1.
General
QA
and
QC
program
1.1
Requirements
for
all
systems
(
preventive
maintenance
procedures,
record
keeping,
maintenance
records)

1.2
Specific
requirements
for
CEMS
(
calibration,
relative
accuracy
test
audit
[
RATA]
procedures)

2.
Frequency
of
testing
2.1
Daily
Assessments
(
calibration
error
test,
daily
flow
interference
check,
data
validation,
QA,
recording)

2.2
Quarterly
Assessments
(
linearity
check,
leak
check,
data
validation,
flow­
to­
load
ratio
or
gross
heat
rate
evaluation)
5
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
2.3
Semiannual
and
Annual
Assessments
(
RATA,
data
validation,
bias
adjustment
factor)

2.4
Recertification
2.5
Other
Audits.

The
reliability
criterion
can
be
seen
as
having
two
components.
The
first
component
requires
that
the
continuous
monitor
produces
certain
data,
95
percent
of
the
time.
In
other
words,
that
the
monitor
is
working
reliably.
The
second
component
pertains
to
the
data
quality
and
format.
The
data
must
be
valid
1­
hour
averages.
The
term
"
Valid"
is
not
further
defined
in
the
40
CFR
75,
and
must
be
defined
first.
It
is
recommended
to
use
EPA's
definition
of
"
valid"
which
can
be
found
in
EPA's
Introduction
to
Data
Quality
Indicator's
(
DQIs).
2
1.2.3
Accessibility
and
Timeliness
Criteria
Accessibility
and
timeliness
are
closely
related
and
are
discussed
together.

The
accessibility
attribute
under
40
CFR
75
is
described
as
follows:
"
To
demonstrate
accessibility
equal
to
or
better
than
the
CEMS,
the
owner
or
operator
shall
provide
reports
and
on­
site
records
of
emissions
data
to
demonstrate
that
the
AMS
provides
data
meeting
the
requirements
of
Subparts
F
and
G."

From
this
definition
it
is
clear
that
it
pertains
to
data
output,
and
not
to
physical
accessibility
of
the
monitor,
which
is
not
further
discussed.
Accessibility
can
thus
be
interpreted
as
the
capability
of
the
monitoring
system
to
provide
both
personnel
and
other
authorized
parties
unimpeded
access
to
data
records
that
adhere
to
predefined
criteria,
while
fulfilling
the
specific
requirements
under
Subparts
F
and
G.

The
timeliness
attribute
under
40
CFR
75
is
described
as
follows:
"
To
demonstrate
timeliness
equal
to
or
better
than
the
CEMS,
it
must
be
demonstrated
that
the
AMS
can:

2
Introduction
to
Data
Quality
Indicators
Course.
EPA's
Quality
website.

(
http://
www.
epa.
gov/
QUALITY/
trcourse.
html#
intro_
dqi)
6
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report

Meet
the
requirements
of
subparts
F
and
G

Provide
a
continuous
quality
assured,
permanent
record
of
certified
emissions
data
on
an
hourly
basis

Issue
a
record
of
data
for
the
previous
day
within
24
hours"

Per
definition,
timeliness
pertains
to
data
only,
and
it
can,
thus
be
seen
as
a
subattribute
to
data
accessibility.
Timeliness
can
be
interpreted
as
the
capability
of
the
monitor
to
provide
data
on
or
before
a
specified
time,
while
fulfilling
the
requirements
under
Subparts
F
and
G.

1.2.4
Summary
of
Subparts
F
and
G
As
mentioned
before,
Subpart
F
(
CFR
75)
of
40
CFR
75
includes
record
keeping
requirements,
while
Subpart
G
(
CFR
75)
includes
reporting
requirements.

Subpart
F
has
three
sections
that
are
potentially
pertinent
to
the
accessibility
and
timeliness
criteria:

75.53
Monitoring
plan
75.57
General
record
keeping
provisions
75.59
Certification,
quality
assurance
and
quality
control
record
keeping
provisions
Section
75.53
mandates
that
the
monitoring
plan
should
contain
a
description
of
the
monitor
site
location
for
each
monitoring
component
in
the
continuous
emission
(
or
opacity)
monitoring
systems
(
c)(
3)
and
(
c)(
4),
as
well
as
a
table
that
describes
the
identifiable
components.
Also,
a
data
acquisition
and
handling
system
table
is
required
that
includes
identification
and
description
of
all
major
components
of
the
automated
data
acquisition
and
handling
system.
(
c)(
5).
This
includes
a
copy
of
the
test
results
verifying
accuracy
of
the
automated
data
acquisition
and
handling
system
(
once
such
results
are
available).
Also
required
is
a
schematic
stack
diagram
for
the
units
monitored
by
a
continuous
emissions
(
or
opacity)
monitoring
system.

Section
75.57
provides
separate
record
keeping
provisions
for
SO2,
NOX,
and
CO2,
which
are
fairly
similar
in
content.
The
provisions
for
SO2
are
summarized
below,
as
they
can
possibly
be
used
as
a
future
basis
for
provisions
for
Hg,
because
they
may
be
pertinent
to
accessibility
and
timeliness.
SO2
emission
record
provisions,
(
c)(
1),
(
c)(
2)
and
(
c)(
4),
state
that
the
owner
or
operator
shall
record
for
each
hour
7
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
for
each
affected
unit
during
unit
operation,
as
measured
and
reported
from
each
certified
monitor:


Hourly
average
SO2
concentration
and
mass
emission
rate

Hourly
average
SO2
concentration
and
mass
emission
rate
adjusted
for
bias
if
this
adjustment
is
required

Percent
SO2
monitor
data
availability

Method
of
determination
for
hourly
average
SO2
concentration

Hourly
average
volumetric
flow
rate

Hourly
average
volumetric
flow
rate
adjusted
for
bias
if
this
adjustment
is
required

Percent
volumetric
flow
monitor
data
availability,
method
of
determination
for
hourly
average
flow
rate
In
addition,
under
Certification,
quality
assurance
and
quality
control
record
keeping
provisions
(
Section
75.59),
numerous
QA/
QC
parameters
need
to
be
recorded
for
the
CMS
that
include
calibration
and
RATA
testing.

Subpart
G
addresses
submission
of
records,
confidentiality,
notifications,
monitoring
plan
submittals,
certification
and
recertification
submittals,
quarterly
reports
and
petitions.
It
is
not
likely
that
the
requirements
under
Subpart
G
will
have
an
effect
on
the
evaluation
of
the
different
CMSs
and
Subpart
G
is
not
discussed
in
further
detail.
8
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
1.3
Methods
1.3.1
Ontario
Hydro
Method
The
ASTM
recently
issued
the
OH
method
under
designation
D6784­
02
approved
April
10,
2002.
The
OH
method
is
an
impinger
train
gas
sampling
method
that
differentiates
mercury
into
particle­
bound,
oxidized,
and
elemental
fractions.
The
OH
method
was
widely
used
as
the
basis
for
sampling
plans
under
the
recent
mercury
Information
Collection
Request
(
ICR)
phase
III.

The
OH
method
is
applicable
to
the
determination
of
elemental,
oxidized,
particlebound
and
total
mercury
emissions
from
coal­
fired
stationary
sources.
Resulting
emissions
are
expressed
in
concentration
terms,
µ
g/
Nm3.
Conversion
of
OH
results
into
an
annual
mass
emission
will
require
application
of
additional
calculations.
Though
the
method
explicitly
claims
applicability
to
coal­
fired
stationary
sources,
the
method
does
not
appear
to
exclude
the
use
of
other
fuels
such
as
coke,
orimulsion,
oil,
tires,
refuse
derived
fuel
(
RDF),
or
biomass.
Use
of
alternate
fuels
may
produce
ash
that
may
further
confound
speciation
bias
in
the
OH
method
but
is
not
expected
to
effect
total
mercury.

In
the
OH
method,
sample
is
extracted
isokinetically
from
the
flue
gas
through
a
probe
and
filter
assembly.
Particulate
matter
is
removed
from
the
sample
with
a
filter.
The
sample
is
then
passed
through
a
series
of
seven
impingers
contained
in
an
ice
bath
to
sequentially
remove
Hg.
The
first
three
impingers
remove
oxidized
Hg
from
the
sample
with
a
potassium
chloride
(
KCl)
solution.
Elemental
Hg
is
then
removed
in
the
fourth
impinger
containing
an
acidic
hydrogen
peroxide
(
H2O2)
solution,
followed
by
three
additional
impingers
containing
acidic
potassium
permanganate
(
KMnO3)
solution.
The
remaining
gas
is
then
dried
in
an
impinger
containing
a
desiccant,
and
the
Hg
content
in
the
dry
flow
is
determined.
Particlebound
Hg
is
recovered
from
the
"
front
half"
of
the
sampling
train,
the
probe
and
filter
assembly,
and
analyzed
by
cold
vapor
atomic
absorption
spectroscopy
(
CVAAS)
or
cold
vapor
atomic
fluorescence
spectroscopy
(
CVAFS).
Oxidized
Hg
recovered
from
the
KCl
impingers
and
connectors
is
measured
by
CVAAS
or
CVAFS.
Elemental
Hg
is
determined
by
adding
the
amount
of
Hg
recovered
in
the
impinger
containing
acidic
H2O2
and
its
connector
and
the
impinger
containing
KMnO3
and
their
connectors.
9
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
There
are
no
known
biases
to
the
OH
method
for
total
Hg,
however
several
conditions
can
affect
the
speciation
of
the
sample.
3
As
stated
in
the
method,
the
relative
standard
deviations
for
gas
phase
elemental
and
oxidized
Hg
were
less
than
11
percent
for
concentrations
greater
than
3
µ
g/
m3
and
less
than
34
percent
for
concentrations
less
than
3
µ
g/
m3.4
A
number
of
revisions
to
the
OH
method
have
been
made
since
the
Method
301
evaluation
cited
in
the
OH
method.
The
precision
of
the
OH
method
is
likely
to
have
improved
since
the
initial
Method
301
evaluation.

1.3.2
CEMS
Hg
CEMS
are
undergoing
continuing
development,
with
current
technology
being
tested
under
various
programs.
Few,
if
any,
commercial
applications
at
coal­
fired
power
plants
currently
exist
in
the
United
States,
but
the
technology
is
developing
rapidly.
Many
CEMS
vendors
have
participated
in
various
field
tests,
including
some
conducted
by
EPA.
Most
manufacturers
are
in
the
process
of
making
significant
improvements
or
refinements
to
their
monitors
as
time
proceeds.

CEMS
can
be
classified
according
to
their
method
of
converting
oxidized
Hg
to
elemental
Hg
as
dry
or
wet.
Wet
systems
convert
Hg
in
impinger
solutions,
while
dry
systems
thermally
or
catalytically
convert
the
Hg
species.
Most
CEMS
can
provide
speciated
Hg
emissions
values,
but
the
development
trend
in
industry
is
toward
total
Hg
measurement.
Based
on
operational
experience
and
industry
knowledge,
the
dry
systems
appear
to
offer
more
operational
availability
and
require
less
maintenance,
compared
to
wet
systems.
These
CEMS
provide
quick
real­
time
Hg
concentrations
and
many
have
inherent
auto
zeroing
and
system
checks
built
into
them.
Based
on
operational
experience,
the
dry
systems
are
easier
to
operate
overall
because
of
not
having
any
chemical
interactions.

As
previously
mentioned,
EPA
has
proposed
regulations
for
Hg
emissions
from
coal­
fired
utilities
and
has
also
established
a
draft
performance
standard
(
PS)
12A,

3
D6784­
02
Standard
Test
Method
for
Elemental,
Oxidized,
Particle­
Bound,
and
Total
Mercury
in
Flue
Gas
Generated
from
Coal­
Fired
Stationary
Sources
(
Ontario
Hydro
Method),
ASTM
International,
West
Conshohoken
Pennsylvania,
June
1,
2002.

4
ibid.
10
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
against
which
Hg
CEMS
can
be
evaluated.
The
current
proposed
regulation
for
Hg
requires
monitoring
of
total
Hg
only.

Evaluation
of
accuracy
and
precision
of
Hg
CEMS
is
confounded
by
the
rapid
and
continuing
development
of
Hg
CEMSs
by
vendors
and
the
lack
of
published
information
on
CEMS
performance.
Evaluation
of
the
performance
of
CEMS
will
improve
when
suitable
reference
materials
or
procedures
become
available.
A
National
Institute
of
Standards
&
Technology
(
NIST)
approved
elemental
mercury
gas
standard
is
still
under
development.
EPA
has
developed
PS
12A,
which
defines
relative
accuracy
requirements
for
Hg
CEMS
measurement
along
with
certain
other
QA
requirements.

1.3.3
Sorbent
Trap
Method
The
sorbent
trap
sampling
method
for
total
Hg
is
in
the
laboratory/
pilot
phase
and
is
being
tested
on
a
limited
scale.
No
references
were
found
that
support
that
the
method
has
been
applied
commercially.

The
sorbent
trap
technique
uses
3/
8
to
½
­
inch
diameter
cylindrical
tubes
containing
activated
carbon
impregnated
with
potassium
iodide
to
adsorb
mercury
from
flue
gas.
The
sampling
system
consists
of
a
sample
probe,
Teflon
®
tubing,
a
glass
wool
plug
to
collect
particulate
matter,
a
two­
part
sorbent
trap,
a
sample
pump,
and
a
gas
meter,
mounted
in
a
console.
The
front
section
of
the
sorbent
trap
is
used
to
chemically
adsorb
mercury.
The
back
section,
separated
from
the
front
by
glass
wool,
collects
any
mercury
that
breaks
through
the
front
section.
The
flue
gas
is
extracted
through
the
carbon
sorbent­
packed
trap
where
the
Hg
is
adsorbed.
The
sample
pump
mounted
in
the
console,
along
with
a
flow
meter
and
temperature
gauges,
creates
the
vacuum
on
the
sorbent
trap.
The
flow
meter
can
be
adjusted
to
achieve
any
desired
flow
rate
through
the
sampling
system
according
to
the
time
period
of
sampling
and
the
flow
rate
of
the
stack
gas.
Typically,
the
sorbent
trap
and
the
probe
that
secures
it
are
the
only
things
extending
into
the
flue
gas
path.
Once
the
sample
has
passed
the
sorbent
trap
and
probe,
it
travels
through
a
heated
Teflon
line
to
prevent
condensation
of
any
Hg
that
may
adhere
to
the
walls.
Then
the
sample
travels
to
the
sample
pump
where
the
moisture
is
removed
before
passing
through
the
dry
gas
meter.

Once
the
sorbent
trap
method
has
reached
commercial
status,
it
should
be
applicable
to
capturing
total
mercury
with
close
to
100
percent
efficiency
from
coalfired
stationary
sources.
The
method
takes
a
continuous,
cumulative
sample
over
a
given
time
period
that
can
range
from
one
hour
to
one
week
and
possibly
up
to
a
11
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
month.
After
this
time
period
the
sorbent
trap
is
exchanged
for
a
fresh
one,
and
the
used
sorbent
trap
is
sent
to
an
off­
site
laboratory
to
be
analyzed
according
to
standard
methods.
Currently,
there
is
one
known
laboratory
that
analyzes
these
sorbent
traps.
The
quantity
of
mercury
is
measured
using
a
method
with
a
detection
limit
of
0.5
nanograms
per
liter
(
ng/
L)
or
5
ppt
in
the
leachate.
The
known
analysis
laboratory
estimates
that
this
detection
limit
is
equivalent
to
about
1
ng/
m3
for
a
one­
week
sampling
period.
In
addition,
the
laboratory
maintains
that
they
can
easily
match
the
OH
Method
detection
limits
and
performance
criteria
for
total
mercury.

1.3.4
Fuel
Sampling
Fuel
sampling
(
coal
sampling
and
analysis)
as
an
emissions
measurement
technique
implicitly
utilizes
an
emission
factor
approach;
mercury
emissions
are
some
fraction
of
all
the
mercury
inputs.
Mercury
in
the
fuel
is
generally
considered
the
sole
source
of
mercury
emissions
even
though
other
mercury
inputs
could
easily
be
incorporated
into
this
approach
if
deemed
necessary.
Mercury
content
in
the
fuel
and
the
amount
of
fuel
burned
are
used
to
determine
the
mercury
inputs
to
a
unit.
In
the
extreme
or
maximum
emissions
case,
mercury
emissions
from
a
unit
could
be
estimated
as
the
total
measured
mercury
inputs,
discounting
mercury
retained
in
residuals
such
as
fly
ash
or
gypsum.

An
alternate
approach
is
to
apply
an
emission
factor,
which
can
either
be
based
on
unit­
specific
emission
measurements
or
on
average
emission
measurements
for
a
defined
class
of
technology
employed.
Measuring
mercury
content
in
the
fuel
and
fuel
inputs
to
units
appears
to
be
an
integral
part
of
the
proposed
mercury
control
legislation.
Mercury
inputs
provide
the
baseline
which
emissions
are
compared
to
in
control­
based
legislation.
Similarly,
the
proposed
Clear
Skies
Act
of
2003
reverts
to
a
control­
based
mercury
emissions
requirement
based
on
fuel
mercury
content,
should
the
administrator
not
establish
the
emissions
trading
system.
5
Fuel
sampling
will
determine
the
total
amount
of
mercury
in
the
fuel.
Mercury
in
the
fuel
can
be
expressed
on
a
mass
basis,
either
wet
or
dry,
or
can
be
normalized
to
the
heating
value
of
the
fuel
and
be
expressed
on
an
energy
basis.
When
the
fuel
is
burned,
the
mercury
in
the
fuel
will
be
converted
to
elemental,
oxidized,
and
particle­
bound
mercury.
Though
the
fuel
sampling
is
most
directly
related
to
total
5
"
Clear
Skies
Act
of
2003"
accessed
at
http://
www.
epa.
gov/
air/
clearskies/
Air_
005.
pdf
12
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
mercury
emissions,
it
may
be
possible
to
estimate
the
partitioning
of
the
mercury
based
on
other
fuel
components
and
fuel
firing
conditions.

Accuracy
and
precision
for
mercury
fuel
sampling
appear
to
be
dependent
upon
both
the
analytical
method
and
the
specific
matrix
being
analyzed.
The
standard
test
method
for
total
mercury
in
coal
(
ASTM
D­
3684­
01)
is
reported
to
have
a
bias
with
respect
to
three
reference
materials
analyzed.
Bias
was
significant
at
a
95
percent
confidence
level
for
all
three
reference
materials
but
appears
to
be
dependent
on
the
specific
fuel.
The
mean
Hg
content
from
ICR
phase
II
for
bituminous
coals
at
electric
utilities
was
0.11
ppm6.
This
was
quite
comparable
to
the
Certified
Reference
Material
(
CRM)
used
to
evaluate
bias
in
the
ASTM
method.
Repeatability
of
the
method
is
reported
as
0.036
ppm,
while
reproducibility
of
the
method
is
reported
as
0.054
ppm.
7
In
practice,
fuel
sampling
can
either
be
conventional
or
on­
line
(
automatic).
The
ASTM
and
the
International
Organization
for
Standardization
(
ISO)
set
standard
practices
for
the
sampling,
sample
preparation,
and
analysis
of
coal
samples.

Conventional
analysis
is
essentially
a
series
of
batch
operations
that
entail
the
incremental
removal
of
a
specified
series
of
relatively
small
quantities
of
coal
from
the
process
stream.
Sample
preparation
covers
a
set
of
conditioning
procedures
such
as
drying,
mixing,
crushing,
reducing
and
dividing,
so
the
series
of
samples
is
transformed
into
a
test
sample
that
is
appropriate
for
subsequent
analytical
procedures.
Mercury
is
typically
determined
with
the
flameless
atomic
absorption
spectroscopy
(
AAS)
method,
where
the
powdered
coal
is
first
digested
with
oxidizing
acids.
Mercury
in
the
sample
solution
is
reduced
to
its
elemental
state
with
stannous
chloride
and
then
aerated
from
solution
onto
a
silver
screen
placed
in
the
vapor
train.
The
silver
screen
is
subsequently
heated
and
the
mercury
vapor
is
carried
by
an
airstream
to
the
absorption
cell.
(
United
States
Geological
Survey
6
National
Risk
Management
Research
Laboratory,
"
Control
of
Mercury
Emissions
from
Coal­
Fired
Electric
Utility
Boilers:
Interim
Report
Including
Errata
Dated
3­
21­
02,"
EPA­
600/
R­
01­
109,
April
2002.

7
D3684­
01
Standard
Test
Method
for
Total
Mercury
in
coal
by
the
Oxygen
Bomb
Combustion/
Atomic
Absorption
Method,
ASTM
International,
West
Conshohoken
Pennsylvania
13
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
[
USGS]
Circular
735:
Methods8).
AAS
is
usually
conducted
in
specialized
laboratories,
implying
that
the
samples
will
have
to
be
shipped
and
that
results
are
available
some
time
after
sampling
and
shipment
(
assume
one
week).

On­
line
analysis
is
a
continuous
series
of
rapid
operations,
where
the
product
is
conditioned
for
analysis
and
subsequently
analyzed.
During
analysis,
the
coal
is
subjected
to
certain
physical
processes
that
interact
with
the
sample
and
correlate
the
analyte
quantity
to
a
database
of
values
that
were
previously
obtained
from
coals
that
have
been
analyzed
by
conventional
analysis.

At
this
point,
on­
line
analysis
technology
for
mercury
from
solid
samples
is
not
yet
commercially
available.
Thus,
mercury
sampling
by
means
of
fuel
analysis
constitutes
conventional
analysis.

8
http://
pubs.
usgs.
gov/
circ/
c735/
method.
htm
14
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
2.
Systems
Comparison
2.1
Precision
As
per
40
CFR
75,
the
precision
criteria
to
be
used
to
evaluate
performance
of
an
Hg
CEMS
compared
to
a
reference
method
include
variance,
correlation,
and
bias.
These
precision
criteria
are
related
to
performance
of
a
reference
method
as
opposed
to
some
absolute
value.
The
criteria
are
used
because
precision
obtained
from
an
alternative
monitoring
system
is
often
required
to
be
as
good
or
better
than
the
standard
monitoring
system;
such
is
the
case
in
40
CFR
75.41.

In
part
75,
precision
between
the
methods
is
compared
using
three
statistical
tests:
an
F
test,
a
correlation
test,
and
a
t­
test.
The
F
test
is
a
comparison
of
the
variance
of
emissions
determined
with
the
standard
monitoring
system
and
the
variance
of
the
alternative
monitoring
system.
The
correlation
test
provides
a
comparison
of
the
responses
of
standard
and
alternate
monitoring
systems
for
linearity.
The
t­
test
compares
the
average
response
of
the
standard
and
alternate
monitoring
system
to
ensure
they
are
not
significantly
different.
These
three
statistical
tests,
as
they
would
be
supplied
to
each
method,
are
discussed
in
detail
below.

2.1.1
Variance
(
F
Test)

Variance
is
determined
with
the
F
test.
The
F
test
ensures
the
variability
in
emissions
encountered
with
an
alternate
monitoring
system
is
not
significantly
larger
than
the
variability
encountered
with
the
standard
monitoring
system.
This
F
test
includes
the
variability
of
the
source
and
the
variability
of
the
monitoring
system.
Since
the
variability
associated
with
the
source
is
common
to
both
systems,
the
F
test
compares
the
variability
of
the
alternative
and
standard
monitoring
system
without
addressing
the
variability
associated
with
the
source.
Alternatively,
the
variance
associated
with
the
method
can
be
determined
through
co­
located
and
coincident
sampling.

The
reference
method
for
mercury
used
in
the
recently
proposed
PS
12A
is
Ontario­
Hydro
(
mercury)
or
Method
29
(
mercury
and
other
metals).
PS
12A
requires
duplicate
sampling
for
mercury
and
sets
a
limit
on
the
agreement
of
each
test.
The
duplicate
analysis
can
provide
an
estimate
of
method
variance
for
the
reference
method.
PS
12A
requires
the
Relative
Percent
Difference
(
RPD)
of
the
duplicate
analysis
to
be
less
than
or
equal
to
10
percent
and
Relative
Standard
Deviation
[
RSD]
less
than
14
percent
for
mean
mercury
concentrations
above
1
µ
g/
m3.
The
Method
301
evaluation
published
with
the
Ontario­
Hydro
method,
which
utilized
15
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
quad­
trains,
resulted
in
total
vapor
phase
mercury
RSD
of
8.8
percent
for
baseline
conditions
and
lower
RSD
for
spiked
conditions.
A
number
of
improvements
to
the
Ontario­
Hydro
method
have
been
incorporated
since
the
Method
301
evaluation.
As
a
result,
the
precision
of
the
Ontario­
Hydro
method
should
be
improved
relative
to
the
301
evaluation.

The
variability
of
mercury
CEMS
is
to
be
governed
primarily
by
the
ability
of
the
CEMS
to
pass
a
RATA
as
described
in
PS
12A.
The
relevant
requirement
is
that
the
Relative
Accuracy
(
RA)
be
no
greater
than
20
percent.
The
RA
is
defined
as
the
sum
of
the
absolute
value
of
the
difference
in
the
two
methods
plus
the
confidence
coefficient,
a
function
of
the
standard
deviation
of
the
differences
in
the
two
methods,
divided
by
the
mean
of
the
reference
method,
or
the
applicable
emission
limit.
Under
Part
75,
the
bias
could
be
removed
by
application
of
a
bias
adjustment
factor.
When
the
differences
in
the
CEMS
and
the
standard
method
(
e.
g.,
Ontario­
Hydro)
are
not
significant,
the
differences
are
a
result
of
the
random
variability
of
the
CEMS
and
the
random
variability
of
the
standard
method.
Recently
published
RATA
results
of
EPA
sponsored
CEMS
research
indicated
initial
RAs
of
4­
15
percent
for
CEMS
evaluated
and
final
RAs
between
14
and
41
percent.
9
Insufficient
detail
was
published
to
determine
if
significant
bias
was
present
from
the
CEMS.
Further
CEMS
evaluation
may
be
available
in
the
near
future
with
EPA
reports
on
recent
CEMS
evaluations.

The
quality
assurance
required
of
sorbent
traps,
including
the
variability
of
the
analysis,
appears
to
be
governed
in
recently
proposed
mercury
regulations
by
site
specific
quality
assurance
programs
yet
to
be
developed.
Sorbent
trap
results
for
short
duration
tests,
coincident
with
a
reference
method,
may
be
evaluated
using
the
RATA
approach
similar
to
CEMS.
Two
such
RATA
style
tests
were
reported
from
recent
EPA
sponsored
tests
resulting
in
RAs
of
9
and
12
percent10.
Insufficient
detail
was
published
to
determine
if
significant
bias
was
present
from
the
sorbent
traps.

9
Midwest
Research
Institute,
"
Hg
CEMS
Field
Observations:
May­
July
2003."
in
Proceedings
of
International
Conference
on
Air
Quality
IV,"
EERC/
EPRI/
US
EPA/
US
DOE,
Arlington
VA,
September
22­
24,
2003.

10
ibid.
16
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
The
variability
of
fuel
sampling
for
determining
emissions
incorporates
several
sources
of
error
that
are
not
typically
evaluated
simultaneously.
The
emissions
estimate
resulting
from
fuel
sampling
depends
on
representative
fuel
sampling,
sample
analysis,
and
process
operations.
In
comparing
the
fuel
sampling
emissions
to
a
reference
method,
such
as
Ontario­
Hydro,
a
constant
emission
factor
would
typically
be
applied;
variability
caused
by
process
operations
would
be
manifested
in
the
reference
method
but
not
in
the
fuel
sampling
emissions
estimate.
Variation
due
to
fuel
sampling
and
analysis
will
both
be
included
in
a
fuel
sampling
emissions
estimate.
Several
methods
could
potentially
be
used
for
analyzing
mercury
in
the
fuel
and
there
is
no
clear
choice
as
to
which
is
the
best.
Variability
results
of
ASTM
methods
resulting
from
recent
inter­
laboratory
studies
are
incorporated
in
methods
D3684,
D6414,
and
D6722.
For
the
three
methods,
variability
of
coal
sample
analysis
at
a
single
laboratory
using
the
same
lab
equipment
(
repeatability
conditions)
results
in
RSDs
of
6­
19,
5­
11,
and
2­
17
percent,
respectively.
Repeatability
of
ASTM
methods
for
mercury
is
typically
represented
as
having
a
fixed
component
and
a
proportional
component.
Available
data
sets
suggest
there
may
be
distinctions
between
the
performance
of
low
rank
and
high
rank
coals.
Variability
resulting
from
sampling
itself
and
from
coincident
heating
value
analysis
should
result
in
increased
variability
over
mercury
analysis
alone
in
estimating
mercury
emissions.

2.1.2
Correlation
Analysis
The
correlation
analysis
in
part
75.41
is
performed
to
ensure
the
response
of
the
AMS
is
linearly
related
to
the
response
of
a
CEMS
at
each
operating
level.
The
criterion
in
part
75.41
is
for
the
paired
observation
to
be
correlated
with
a
Correlation
Coefficient
of
0.8
or
better.
In
a
general
context,
the
correlation
analysis
demonstrates
that
the
relative
bias
does
not
change
with
the
measured
value.
The
linearity
of
the
reference
method
(
such
as
Ontario­
Hydro)
is
inferred
by
the
lack
of
bias
associated
with
the
method.
CEMS
are
not
currently
required
to
demonstrate
that
their
output
is
linearly
related
to
reference
method
results.
However,
PS
12A
provides
for
determining
the
linearity
of
the
CEMS
by
requiring
measurement
error
(
ME)
of
no
greater
than
5
percent
of
span
at
zero,
mid­
level,
and
high
level
gas
standards.
Various
measures
could
be
employed
as
part
of
a
quality
assurance
program
to
ensure
the
analytical
linearity
of
sorbent
trap
analysis
and/
or
fuel
analysis.
Analytical
methods
involving
spiked
samples
can
demonstrate
linearity
or
the
presence
of
matrix
effects.
17
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
For
sorbent
traps,
the
appropriate
loading
of
sorbent
can
be
demonstrated
by
the
clean
recovery
of
the
back­
half
of
the
sorbent
column.
Ultimately,
sorbent
traps
(
like
CEMS
or
fuel
sampling)
could
be
challenged
with
a
correlation
test.

Fuel
sampling
is
likely
to
fail
in
many
instances
because
mercury
control
in
the
process
equipment
is
not
likely
to
be
linearly
related
to
input
in
all
situations.
Ultimately,
fuel
sampling
(
like
CEMS
or
sorbent
traps)
could
be
challenged
with
a
correlation
test.

2.1.3
Bias
(
t­
test)

The
determination
of
a
statistically
significant
difference
in
two
averages
(
bias)
is
performed
with
a
t­
test.
In
this
case,
t­
test
can
be
used
to
compare
the
average
mercury
emissions
from
a
standard
and
a
proposed
method.
The
t­
test
is
therefore
an
evaluation
of
the
relative
accuracy
of
the
two
emission
measurement
methods.
If
the
methods
are
considered
unbiased,
the
t­
test
becomes
an
evaluation
of
accuracy.
In
Part
75.41,
the
t­
test
compares
a
CEMS
and
an
AMS
over
a
long
operating
period.
In
Part
75,
Appendix
A,
a
similar
evaluation
is
made
in
applying
a
Bias
Adjustment
Factor
in
the
evaluation
of
RATA
tests.
The
RATA
tests
are
typically
paired
observations
whereas
the
CEMS/
AMS
comparison
may
not
be
completely
paired
or
precisely
concomitant.
The
evaluation
depends
largely
on
the
variability
of
the
test
method
with
respect
to
the
standard
method
and
the
number
of
data
points
in
the
average.

The
ultimate
accuracy
of
the
CEMS
emissions
estimate
is
judged
against
a
reference
method.
In
recently
proposed
PS
12A
the
reference
method
for
mercury
is
specified
as
either
Ontario­
Hydro
or
Method
29.
The
PS
12A
regulates
total
vapor
phase
mercury.
Neither
Ontario­
Hydro
nor
Method
29
indicates
a
bias
in
total
mercury,
though
Ontario­
Hydro
indicates
biases
may
exist
in
the
speciation
of
total
mercury.
For
the
purposes
of
total
vapor
phase
emission
estimates,
adsorption
or
desorption
from
collected
particulate
and/
or
from
the
front­
half
equipment
remains
the
only
relevant
bias.
Such
bias
is
minimized
by
sampling
after
the
last
control
device,
resulting
in
nearly
particulate
free
sample,
and
in
controlling
probe
temperatures
using
method
procedures.

The
bias
of
a
mercury
CEMS
system
can
be
evaluated
over
a
short
term
through
the
use
of
a
RATA
as
prescribed
in
the
proposed
PS
12A.
The
RATA
ensures
the
CEMS
is
installed
to
provide
representative
results.
The
amount
of
the
acceptable
bias
per
PS
12A
will
vary
depending
on
the
variability
of
the
test
methods
and
the
number
of
tests
performed.
Recently
published
RATA
results
of
EPA
sponsored
18
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
CEMS
research
indicated
initial
RAs
of
4­
15
percent
for
CEMS
evaluated
and
final
RAs
between
14
and
41
percent.
11
There
does
not
appear
to
be
provisions
for
correcting
statistically
significant
bias
in
the
proposed
regulation
but
this
could
be
incorporated.
With
the
development
of
standard
gases
for
elemental
and
oxidized
mercury,
continued
operation
of
the
Hg
CEMS
could
be
validated
daily
in
the
same
manner
as
Part
75
CEMS.

The
bias
of
sorbent
traps
will
be
ascertained
by
quality
assurance
plans
to
be
developed
for
each
source.
In
principal,
such
a
plan
could
ensure
the
representativeness
of
the
sample
through
the
use
of
RATA
tests;
this
procedure
may
require
modification
to
accommodate
differences
in
sample
times.
Recent
sorbent
trap
RATAs
resulted
in
RAs
of
9
and
12
percent
respectively.
12
A
quality
assurance
plan
would
likely
address
determination
of
the
quantity
of
mercury
captured
as
well
as
quantity
of
sample
volume.
Both
procedures
appear
to
have
well­
established
precedent.

The
bias
of
fuel
sampling
for
mercury
emissions
estimates
is
necessarily
confounded
with
a
determination
of
an
emission
factor
through
the
use
of
some
reference
method.
The
results
from
fuel
sampling
must
be
multiplied
by
an
emission
factor
to
adjust
for
inherent
mercury
control
resulting
from
process
equipment.
The
ASTM
methods
indicate
a
fuel
specific
bias
is
associated
with
each
analytical
technique:
D3684,
D6414,
and
D6722.
The
bias
on
certified
reference
material
ranged
from
­
24
percent
to
9
percent,
­
12
percent
to
insignificant,
and
­
7
percent
to
insignificant
for
the
three
methods,
respectively.
One
unbiased
method
to
correct
these
analyses,
failing
the
certification
of
an
unbiased
fuel
mercury
analytical
technique,
is
through
the
use
of
the
emission
factor
for
each
fuel
used.
In
this
case,
the
performance
of
the
process
equipment
is
then
confounded
with
the
unknown
analytical
bias.
Furthermore,
fuel
sampling
would
only
measure
inputs
on
a
semi­
continuous
basis
as
opposed
to
emissions.
Changes
in
process
equipment
performance
which
alter
mercury
emissions
could
not
be
recorded.
Process
equipment
performance
may
change
as
a
result
of
various
factors
including
changes
resulting
from
fuel
source,
load
factor,
Loss
on
Ignition
(
LOI),
control
equipment
operation,
and
temporal
factors.
Therefore,
fuel
sampling
cannot
currently
be
used
to
estimate
maximum
possible
emissions
with
certainty.
Development
of
both
an
11
ibid.

12
ibid.
19
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
unbiased
fuel
mercury
analysis
technique
and
an
accurate
predictive
model
of
process
performance
relative
to
mercury
emissions
are
necessary
to
provide
an
independent
validation
of
fuel
sampling­
based
emissions.

2.2
Reliability
Under
40
CFR
75,
reliability
must
be
demonstrated
by
showing
that
the
monitoring
system
is
capable
of
providing
valid
1­
hour
averages
for
95.0
percent
or
more
of
the
unit
operating
hours
over
a
1­
year
period
and
that
the
system
meets
the
applicable
requirements
of
40
CFR
75,
Appendix
B,
QA
and
QC
procedures.
The
reliability
criterion
can
be
seen
as
having
two
components,
which
are
discussed
below.

The
first
component
requires
that
the
continuous
monitor
produces
verifiable
data
95
percent
of
the
time.
In
other
words,
data
that
can
be
used
to
prove
the
monitor
is
working
reliably.
This
aspect
of
reliability
may
be
comparable
to
ruggedness.

The
second
component
pertains
to
the
data
quality
and
format.
The
data
must
be
valid
1­
hour
averages.
"
Validity"
is
not
further
defined
in
the
40
CFR
75,
and
can
be
defined
within
the
context
of
EPA's
DQIs13,
which
include
precision,
bias,
representativeness,
comparability,
completeness,
and
sensitivity.
Validity
as
used
within
the
reliability
context
under
40
CFR
75
is
probably
comparable
to
representativeness.
Representativeness
is
the
measure
of
the
degree
to
which
data
suitably
represent
a
characteristic
of
a
population,
parameter
variations
at
a
sampling
point,
a
process
condition,
or
an
environmental
condition.
Representativeness
DQIs
are
qualitative
and
quantitative
statements
regarding
the
degree
to
which
data
reflect
the
true
characteristics
of
a
well
defined
population.
A
number
of
DQIs
relate
to
representativeness
 
in
fact;
precision,
bias,
and
completeness
all
play
integral
roles.
In
this
way,
representativeness
is
an
over
arching
indicator.

CEMS
are
rapidly
developing
and
some
commercially
available
CEMS
are
capable
of
satisfying
the
first
component
of
reliability
criterion,
i.
e.,
operating
and
producing
data
for
95
percent
of
the
time.
Sorbent
traps
are
in
the
early
pilot
phase
and
are
not
yet
proven
of
being
capable
of
providing
data
95
percent
of
the
time.
Although,
the
data
on
sorbent
traps
(
consistency
throughout
the
report)
that
is
13
Introduction
to
Data
Quality
Indicators
Course.
EPA's
Quality
website.

(
http://
www.
epa.
gov/
QUALITY/
trcourse.
html#
intro_
dqi)
20
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
available
does
suggest
that
they
may
be
able
to
meet
this
criterion.
This
component
of
the
reliability
criterion
is
not
applicable
to
the
OH
method,
because
it
is
an
incidental
short­
term
verification
method
and
does
not
operate
year­
round.
(
This
is
one
reason
the
OH
method
is
used
as
the
reference
method.)
Fuel
sampling
is
a
year
round
method
that
is
able
to
produce
data
95
percent
of
the
time
on
a
yearly
basis,
due
to
its
simplicity.
It
can
be
described
as
rugged,
but
as
mentioned
before,
fuel
sampling
is
a
discontinuous
method,
so
a
comparison
may
be
skewed.

Newly
drafted
quality
reliability
requirements
for
mercury
CEMS
need
to
take
into
account
the
specific
requirements
for
routine
maintenance.
For
example,
wet­
based
CEMS
can
produce
waste
streams
of
several
liters
per
day
that
require
handling,
or
the
instruments
may
require
routine
biannual
maintenance.
Most
dry
CEMS
manufacturers
require
minimal
maintenance,
such
as
changing
the
catalyst
or
the
filter
at
varying
intervals.
These
change
outs
depend
on
the
flue
gas
being
sampled
and
the
conditions
of
the
gas
stream.
The
majority
of
dry­
based
CEMS
have
minimal
or
no
calibration
requirements.

The
second
component
of
the
reliability
criterion,
data
must
be
valid
1­
hour
averages,
can
definitely
be
satisfied
by
CEMS.
Most
CEMS
produce
data
every
second,
which
can
then
be
manipulated
by
the
included
software
to
generate
the
required
valid
one­
hour
average.
The
OH
method
can
provide
valid
1­
hour
averages,
but
it
is
only
for
two
or
three
hours
at
a
time.
In
addition,
this
method
is
generally
used
for
verification
of
the
other
AMS
data.
Sorbent
traps
sampling
can
produce
1­
hour
samples
only
in
theory;
when
the
sorbent
traps
are
exchanged
every
hour
on
an
around­
the­
clock
basis.
Fuel
sampling
can
also
in
theory
produce
1­
hour
samples.
However,
these
must
still
be
translated
into
verifiable
mercury
flue
gas
emissions.

Data
collected
by
CEMS
and
the
OH
method
are
believed
to
be
most
representative
of
the
actual
dataset.
It
is
expected
that
the
same
statement
can
be
made
for
sorbent
traps,
once
this
technology
matures,
however,
sorbent
traps
can
only
produce
averaged
data
over
a
certain
time­
frame
and
no
"
snap­
shot"
data.
Fuel
sampling
data
are
expected
to
be
less
representative
of
the
true
dataset,
because
the
fuel
sampling
data
need
to
be
extrapolated
to
represent
the
actual
mercury
concentrations
with
the
help
of
emission
factors.
At
this
time,
reliable
mercury
emission
factors
for
various
coal
burning
and
pollutant
control
scenarios
still
must
be
developed.
21
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
2.3
Accessibility
and
Timeliness
2.3.1
Ontario
Hydro
Method
The
Ontario
Hydro
method
is
a
method
designated
by
ASTM
D6784­
02.
The
method
involves
isokinetic
extraction
of
a
sample
from
the
flue
gas
through
a
probe
and
filter
assembly.
Particulate
matter
is
removed
from
the
sample
with
a
filter.
The
sample
is
then
passed
through
a
series
of
seven
impingers
each
filled
with
specific
solutions.
The
different
mercury
species
are
recovered,
combined
and
usually
analyzed
by
CVAAS
or
CVAFS.
Both
CVAAS
and
CVAFS
are
sophisticated
methods
that
are
typically
conducted
by
specialized
laboratories
off­
site,
requiring
the
shipment
of
samples.

The
OH
method
was
not
developed
for
permanent
operation
but
is
used
as
a
reference
method
to
verify
the
performance
of
other
methods,
such
as
CEMS.
Because
of
the
need
to
manually
collect
and
handle
individual
samples,
which
are
analyzed
off­
site,
the
OH
method
is
not
continuous.
The
method
states
operation
for
at
least
two
hours,
but
not
more
then
three
hours.
In
regard
to
the
accessibility
attribute
under
40
CFR
75,
the
OH
method
can
meet
the
requirements
as
they
pertain
to
submittal
of
a
monitoring
plan
and
provision
of
test
results.
However,
the
requirements
regarding
timeliness
can
not
easily
be
met,
for
practical
reasons.
In
order
to
meet
the
timeliness
requirements,
on­
going
sampling
would
have
to
occur
on
an
hourly
basis,
while
the
samples
would
have
to
be
analyzed
and
reported
within
24
hours.
Implementation
of
such
a
scenario
does
not
appear
to
be
likely.

2.3.2
CEMS
Mercury
CEMS
are
undergoing
continuing
development,
with
current
technology
being
tested
under
various
programs.
Few,
if
any,
commercial
applications
at
coalfired
power
plants
currently
exist
in
the
United
States,
but
the
technology
is
developing
rapidly.
CEMS
provide
quick
real­
time
mercury
concentrations
and
many
have
inherent
auto
zeroing
and
system
checks
built
into
them.

Because
CEMS
are
truly
continuous,
they
meet
the
requirements
for
continuous
emission
monitors
under
40
CFR
75
and
can
serve
as
the
"
base"
system
discussed
in
Section
1.1.

In
regard
to
accessibility,
CEMS
and
their
ancillary
data
processing
units
are
capable
of
providing
full
and
unimpeded
access
to
data
records.
Furthermore,
the
22
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
requirements
under
40
CFR(
f)
can
be
met,
because
this
text
was
specifically
written
for
continuous
monitors.
Specifically,
all
of
the
following
criteria
can
be
met.


Hourly
average
mercury
concentration
and
mass
emission
rate

Hourly
average
mercury
concentration
and
mass
emission
rate
adjusted
for
bias
if
this
adjustment
is
required

Percent
mercury
monitor
data
availability

Method
of
determination
for
hourly
average
mercury
concentration

Hourly
average
volumetric
flow
rate

Hourly
average
volumetric
flow
rate
adjusted
for
bias
if
this
adjustment
is
required

Percent
volumetric
flow
monitor
data
availability,
method
of
determination
for
hourly
average
flow
rate
Mercury
CEMS
technology
is
developing
rapidly.
Improvements
focus
on,
as
previously
stated,
reduction
of
contaminants
and
on
improved
ruggedness
and
reliability
of
the
instruments,
as
well
as
reduced
maintenance.
When
the
technology
has
become
fully
mature,
CEMS
are
expected
to
be
able
to
meet
the
specific
requirements
for
timeliness
in
40
CFR
75.

Thus,
in
the
near
future,
CEMS
are
expected
to
be
able
to
provide
a
continuous
quality
assured,
permanent
record
of
certified
emissions
data
on
an
hourly
basis
and
issue
a
record
of
data
for
the
previous
day
within
24
hours.
Currently,
for
most
CEMS,
data
output
is
immediate
and
the
accessibility
and
timeliness
criterion
are
satisfied.
Modern
data
processing
technology
is
expected
to
allow
for
manipulation
of
the
output
in
most
desired
formats
in
textual,
tabular
or
graphic
form.

2.3.3
Sorbent
Trap
Method
The
sorbent
trap
method
is
in
the
laboratory/
pilot
phase
and
is
being
tested
on
a
limited
scale.
The
sorbent
trap
technique
uses
cylindrical
tubes
up
to
½
­
inch
diameter
containing
activated
carbon
impregnated
with
potassium
iodide
to
adsorb
mercury
from
flue
gas.
Once
the
sorbent
trap
method
has
reached
commercial
status,
it
should
be
applicable
to
capturing
total
mercury
with
close
to
100
percent
23
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
efficiency
from
coal­
fired
stationary
sources.
The
method
takes
a
continuous,
cumulative
sample
over
a
given
time
period
that
can
range
from
one
hour
to
one
week
and
possibly
up
to
a
month.
After
this
time
period
the
sorbent
trap
is
exchanged
for
a
fresh
one,
and
the
used
sorbent
trap
is
sent
to
an
off­
site
laboratory
to
be
analyzed
according
to
standard
methods.
Currently,
only
one
laboratory
in
the
nation
is
commercially
conducting
sorbent
trap
analysis
for
Hg
determination
for
coal­
fired
flue
gas
streams.

The
sorbent
trap
method
is
not
a
monitoring
method
but
a
method
that
takes
physical
samples
over
a
certain
time
period,
which
have
to
be
analyzed
off­
site.
Although
the
sorbent
trap
sampling
method
is
described
as
continuous,
it
only
provides
one
cumulative
sample
over
the
sampling
period.
If
this
sampling
period
is
set
for
one
hour,
this
method
could,
in
theory,
provide
hourly
average
samples
as
required
under
40
CFR
75(
f).
At
this
point,
sampling
sorbent
trap
removal
and
placement
is
a
manual
task,
as
is
preparation
for
shipment.
The
hourly
frequency
of
sampling
may
not
be
practical
for
coal­
fired
power
plants.

In
addition,
the
samples
need
to
be
shipped
off­
site
for
analysis.
The
timeliness
requirements
under
40
CFR
75
include
that
a
record
of
data
from
the
previous
day
must
be
available
within
24
hours.
It
is
unlikely
that
these
conditions
can
be
met
under
the
current
arrangement
while
an
off­
site
laboratory
has
to
analyze
the
sorbent
traps.

For
the
sorbent
trap
method
to
become
an
acceptable
alternate
monitoring
system
under
the
40
CFR
75
accessibility
and
timeliness
criteria,
the
sorbent
trap
replacement
would
have
to
occur
hourly
and
would
have
to
be
automated,
with
analysis
occurring
on­
site
with
data
reporting
within
24
hours.

2.3.4
Fuel
Sampling
Fuel
sampling
will
determine
the
total
amount
of
mercury
in
the
coal.
When
the
coal
is
burned,
the
mercury
in
the
coal
will
be
converted
to
elemental,
oxidized,
and
particle­
bound
mercury.
Though
the
fuel
sampling
is
most
directly
related
to
total
mercury
emissions,
it
may
be
possible
to
estimate
the
partitioning
of
the
mercury
based
on
other
fuel
components
and
fuel
firing
conditions.
Emission
factors
could
be
developed
which
would
be
based
on
unit
specific
emission
measurements
or
on
average
emission
measurements
for
a
defined
class
of
technology
employed.

In
general
and
in
practice,
fuel
sampling
can
either
be
conventional
or
on­
line
(
automatic)
for
certain
chemicals
or
constituents.
On­
line
analysis
is
a
continuous
24
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
series
of
rapid
operations,
where
the
product
is
conditioned
for
analysis
and
subsequently
analyzed.
At
this
point,
on­
line
analysis
technology
for
mercury
from
solid
samples
is
not
yet
commercially
available,
thus,
mercury
sampling
by
means
of
fuel
analysis
constitutes
conventional
analysis.

Conventional
analysis
is
essentially
a
series
of
batch
operations
that
entail
the
incremental
removal
of
a
specified
series
of
relatively
small
quantities
of
coal
from
the
process
stream.
Sample
preparation
covers
a
set
of
conditioning
procedures
such
as
drying,
mixing,
crushing,
reducing
and
dividing,
so
the
series
of
samples
is
transformed
into
a
test
sample
that
is
appropriate
for
subsequent
analytical
procedures.
Mercury
is
typically
determined
with
the
flameless
atomic
absorption
spectroscopy
the
method
where
the
powdered
coal
is
first
digested
with
oxidizing
acids.
AAS
is
usually
conducted
in
specialized
laboratories,
implying
that
the
samples
will
have
to
be
shipped
and
that
results
are
available
some
time
after
sampling
and
shipment.

As
with
sorbent
traps,
fuel
sampling
is
not
a
monitoring
method,
it
is
a
sampling
method.
Because,
the
40
CFR
75(
f),
requires
a
"
monitoring
plan"
the
accessibility
criterion
can
not
be
met
with
the
current
format
of
the
text.
The
requirements
for
the
monitoring
plan
have
been
summarized
in
Section
1.2.5
and
include
descriptions
of
the
monitor
site
location,
monitoring
components,
data
acquisition
and
handling
system.

In
addition,
the
fuel
sampling
method
does
not
sample
the
flue
gas
and
hence,
deliver
an
emission
rate;
it
samples
the
coal
before
it
is
combusted.
As
stated
before,
the
mercury
content
of
the
coal
samples
can
either
be
used
to
quantify
the
upper
boundary
of
mercury
emissions,
or
it
can
be
combined
with
an
emission
factor
to
establish
an
emissions
estimate.
In
either
case,
the
samples
are
taken
before
the
flue
gas
is
produced
and
need
to
be
correlated
to
the
mercury
quantities
in
the
actual
flue
gas.
In
addition
the
samples
will
have
to
be
sent
off­
site
for
analysis,
unless
the
plant
is
furnished
with
an
AAS
laboratory.

The
40
CFR
75(
f)
timeliness
criterion
requires
that
hourly
averages
for
the
mercury
in
the
flue
gas
are
recorded,
which
are
available
within
24
hours,
while
adhering
to
specified
QA
requirements.
Under
the
current
fuel
sampling
procedures
and
analysis,
this
criterion
can
not
be
met.

For
the
fuel
sampling
method
to
become
acceptable
under
the
40
CFR
75
accessibility
and
timeliness
criteria,
it
would
have
to
evolve
into
a
true
monitoring
method,
with
automated,
on­
line
sampling
and
analysis.
In
addition,
the
correlation
25
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
of
the
mercury­
in­
coal
samples
with
the
actual
emissions
will
have
to
be
established
through
the
development
of
verifiable
emission
factors.
This
could
prove
complicated
at
best
because
of
the
variability
of
pollution
control
technology
performance
and
the
combinations
of
control
technology
at
various
coal­
fired
plants.

2.4
Missing
Data
Substitution
Missing
data
substitution
criteria
procedures
are
included
in
Subpart
D
to
40
CFR
75
and
are
summarized
below.
The
procedures
are
summarized
and
the
implications
for
the
different
monitoring
methods
under
review
are
briefly
discussed.

When
the
unit
has
a
certified
redundant
backup
CEMS,
a
like­
kind
replacement
non­
redundant
backup
analyzer,
or
a
backup
reference
method
monitoring
system,
data
generated
by
such
a
system
can
be
used
(
par.
75.30).
When
prior
quality
assured
data
exist,
these
can
be
used
as
substitutes
in
certain
cases
which
are
identified
in
par.
75.31.
When
no
prior
quality
assured
data
exist,
the
maximum
potential
pollutant
concentration
must
be
used.
This
parameter
is
described
in
Appendix
A,
2.1.1.1.
to
40
CFR
75,
and
it
is
based
on
fuel
analysis.

An
important
part
of
the
missing
data
substitution
procedure
is
the
calculation
of
percent
monitor
data
availability
(
par.
75.32).
This
parameter
is
used
to
determine
what
method
should
be
used
for
missing
data
substitution,
together
with
the
duration
of
the
CEMS
outage.
(
See
Table
1
in
par.
75.33
for
SO2,
which
was
used
as
a
surrogate
for
mercury).
The
higher
the
availability
and
the
shorter
the
outage
for
the
CEMS,
the
data
substitution
procedure
becomes
more
lenient
and
simpler.
For
example,
when
data
availability
is
high
(>
95
percent)
and
the
duration
of
the
outing
is
short
(<
24
hours),
it
suffices
to
take
the
average
of
the
average
of
the
hour
before
and
the
average
of
the
hour
after
the
outage.
On
the
other
hand,
when
data
availability
drops
below
80
percent,
the
maximum
potential
pollutant
concentration
must
be
used.

Subpart
D
of
40
CFR
75
was
clearly
drafted
with
CEMS
as
the
standard
method.
At
this
point,
only
the
Ontario
Hydro
method
can
provide
hourly
certified
data
and
this
method
could
be
used
as
a
certified
backup
analyzer
for
short
periods
of
time.
As
indicated,
fuel
analysis
is
incorporated
into
the
method
to
calculate
maximum
potential
pollutant
concentrations.
It
appears
that
this
would
rule
out
the
use
of
fuel
analysis
as
primary
data
collection
method.
Because
the
missing
data
procedure
26
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
uses
average
hourly
data,
sorbent
trap
data
can
potentially
be
used
as
backup
data,
provided
that
the
method
becomes
certified
and
is
able
to
generate
hourly
data.

2.5
Qualitative
Comparison
of
CEMS
and
Sorbent
Traps
Given
the
variability
in
the
type,
operation
and
fuel
mix
of
sources
in
the
proposed
Hg
cap
and
trade
program,
EPA
believes
that
emissions
must
be
monitored
continuously
in
order
to
ensure
the
precision,
reliability,
accuracy,
and
timeliness
of
emissions
data
that
support
the
cap
and
trade
program.
This
implies
that
CEMS
and
sorbent
trap
monitoring
systems
are
the
most
appropriate
systems.
EPA
is
considering
two
alternative
approaches
for
the
continuous
monitoring
of
Hg
emissions.
Alternative
1
represents
EPA's
traditional
approach
to
implementing
an
emissions
trading
program,
and
Alternative
2
is
a
variation
of
Alternative
1.

In
Alternative
1,
most
sources
would
be
required
to
use
continuous
emission
monitoring
systems
(
CEMS),
while
low­
emitting
sources
with
Hg
emissions
at
or
below
a
specified
threshold
value
would
be
allowed
to
use
sorbent
trap
monitoring
systems.
As
of
March
2004,
the
proposed
threshold
value
is
9
lbs
(
144
ounces)
of
mercury
emissions
per
year,
although
EPA
is
accepting
comment
on
three
additional
alternative
thresholds
of
29,
46
and
76
lbs/
yr.

In
Alternative
2,
all
sources
would
be
allowed
to
use
either
CEMS
or
sorbent
trap
monitoring
systems.
But
those
sources
choosing
the
sorbent
trap
monitoring
system,
whose
Hg
emissions
are
above
the
specified
emission
threshold
would
be
required
to
implement
quarterly
relative
accuracy
audits
(
RAAs)
so
that
the
qualityassurance
(
QA)
procedures
are
comparable
to
those
for
a
CEMS.
However,
the
exact
QA
procedures
are
still
being
be
defined.

The
use
of
sorbent
trap
monitoring
systems
as
an
alternative
to
CEMS
for
monitoring
Hg
emissions
has
been
the
subject
of
recent
field
studies,
and
has
been
proposed
by
EPA
as
an
alternative
to
CEMS
for
determining
compliance
with
the
Hg
MACT
standard
which
is
the
subject
of
EPA's
December
15,
2003,
proposed
rulemaking.
Since
these
studies
have
indicated
that
sorbent
trap
monitoring
systems
are
capable
of
providing
accurate
measurements
of
Hg
concentration,
EPA
is
also
proposing
to
allow
the
use
of
sorbent
trap
monitoring
systems
for
the
Hg
trading
program
within
limitations.
These
limitations
are
based
on
EPA's
concerns
that
sorbent
trap
monitoring
systems
cannot
be
subjected
to
the
same
level
of
quality
assurance
and
quality
control
as
traditional
CEMS
for
long­
term
use
in
monitoring
Hg
emissions.
Insufficient
quality
assurance
of
the
emissions
data
inspires
a
lack
of
confidence
in
the
commodity
being
traded,
i.
e.
the
emission
allowances.
27
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
Furthermore,
this
would
be
inconsistent
with
the
monitoring
requirements
of
EPA's
two
existing
successful
emissions
cap
and
trade
programs,
the
Acid
Rain
program
and
the
NOx
Budget
Trading
program.

Congress
required
quality
assurance
for
CEMS
or
alternative
monitoring
systems
used
in
the
Acid
Rain
program.
EPA
requires
the
CEMS
to
undergo
rigorous
initial
certification
testing
and
periodic
quality
assurance
testing,
including
daily
calibrations,
quarterly
linearity
checks,
and
annual
or
semi­
annual
RATAs.
A
standard
set
of
data­
validation
rules
and
substitute
data
procedures
apply
to
all
of
the
CEMS.
These
stringent
requirements
provide
an
accurate
hourly
accounting
of
the
mass
emissions
from
each
affected
source,
and
provide
prompt
feedback
if
the
monitoring
system
is
operating
incorrectly.
This
ensures
a
level
playing
field
among
the
regulated
sources
with
accurate
accounting
for
every
ton
of
emissions,
which
inspires
confidence
in
the
trading
of
allowances.

Although
EPA
is
proposing
initial
certification
testing
and
periodic
quality
assurance
testing
for
a
sorbent
trap
monitoring
system,
these
proposed
requirements
do
not
provide
the
same
frequency
of
quality
assurance
testing
as
for
a
CEMS.
Daily
calibrations
and
quarterly
linearity
checks,
or
their
equivalents,
cannot
be
performed
on
a
sorbent
trap
monitoring
system.
EPA
solicits
comment
as
to
whether
other
daily,
monthly,
or
quarterly
quality
assurance
test
approaches
have
been
developed
which
parallel
those
required
for
CEMS.
To
address
this
difference
in
quality
assurance
testing
between
the
sorbent
trap
monitoring
system
and
the
CEMS,
EPA
is
proposing
additional
quarterly
RAAs
for
those
sources
above
the
specified
emissions
threshold
who
would
elect
to
use
sorbent
trap
monitoring
systems
under
the
second
alternative
approach.

This
requirement
for
more
frequent
quality
assurance
testing
reflects
the
difference
in
the
purposes
of
monitoring
for
an
emissions
trading
program
compared
to
monitoring
for
an
emissions
limitation
program
such
as
a
MACT
standard.
Monitoring
for
the
trading
program
requires
frequent,
periodic
testing
to
ensure
continued
accuracy
of
the
measurement
method,
to
accurately
account
for
all
emissions
since
each
unit
of
emissions
is
tied
to
an
allowance
which
is
tradeable
at
any
time
throughout
the
year.
It
is
important
for
source
owners
to
know
how
much
"
money
is
in
the
bank"
at
any
given
time.
This
necessity
was
recognized
by
Congress
when
it
required
the
use
of
CEMS
in
the
Acid
Rain
Program,
which
serves
as
the
model
for
both
the
NOx
Budget
Trading
program
and
the
proposed
Hg
trading
program.
28
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
3.
Summary
This
report
evaluates
and
compares
the
various
Hg
emission
monitoring
methods
in
light
of
potential
qualitative
performance
in
a
Hg
emissions
quantification
program
such
as
a
market­
based
cap­
and­
trade
program.
The
monitoring
methods
are
evaluated
on
the
basis
of
the
attributes
defined
in
40
CFR
75,
which
include
precision,
reliability,
accessibility,
and
timeliness.
The
40
CFR
75
pertains
to
the
pollutants
SO2
and
NOX
(
as
well
as
CO2,
opacity
and
volumetric
flow)
but
it
does
not
address
mercury.
However,
this
report
uses
these
attributes
and
applies
them
to
mercury
for
consistency.

3.1
Precision
In
40
CFR
75.41(
a),
precision
is
expressed
as
including
an
F
test,
a
correlation
analysis,
and
a
t­
test
to
address
bias.
The
F
test
is
a
comparison
of
the
variance
of
emissions
determined
with
the
standard
monitoring
system
and
the
variance
of
the
alternative
monitoring
system.
The
correlation
test
provides
a
comparison
of
the
responses
of
standard
and
alternate
monitoring
systems
for
linearity.
The
t­
test
compares
the
average
response
of
the
standard
and
alternate
monitoring
system
to
ensure
they
are
not
significantly
different.

Fuel
sampling
is
likely
to
fail
in
many
instances
because
mercury
control
in
the
process
equipment
is
not
likely
to
be
linearly
related
to
input
in
all
situations.
Ultimately,
fuel
sampling
(
like
CEMS
or
sorbent
traps)
could
be
challenged
with
a
correlation
test.
There
does
not
appear
to
be
provisions
for
correcting
statistically
significant
bias
in
the
proposed
regulation
but
this
could
be
incorporated.
With
the
development
of
standard
gases
for
elemental
and
oxidized
mercury,
continued
operation
of
the
Hg
CEMS
could
be
validated
daily
in
the
same
manner
as
Part
75
CEMS.
Development
of
both
an
unbiased
fuel
mercury
analysis
technique
and
an
accurate
predictive
model
of
process
performance
relative
to
mercury
emissions
are
necessary
to
provide
an
independent
validation
of
fuel
sampling­
based
emissions.

3.2
Reliability
Reliability
should
be
demonstrated
by
showing
that
the
monitoring
method
is
capable
of
providing
valid
1­
hour
averages
for
95.0
percent
or
more
of
the
unit
operating
hours
over
a
one­
year
period
and
that
the
system
meets
the
applicable
QAQC
requirements
of
40
CFR
75,
Appendix
B.
29
Feasibility
Analysis
of
Mercury
Monitoring
Techniques
Draft
Report
CEMS
are
rapidly
developing
and
some
commercially
available
CEMS
are
capable
of
operating
and
producing
data
for
95
percent
of
the
time.
Sorbent
traps
are
in
the
early
pilot
phase
and
are
not
yet
proven
capable
of
providing
data
95
percent
of
the
time,
although,
the
data
on
sorbent
traps
that
is
available
does
suggest
that
they
may
be
able
to
meet
this
criterion.
The
OH
method
is
an
incidental
short­
term
verification
method
and
does
not
operate
year­
round.

The
second
component
of
the
reliability
criterion,
data
must
be
valid
1­
hour
averages,
can
definitely
be
satisfied
by
CEMS.
Most
CEMS
produce
data
every
second,
which
can
then
be
manipulated
by
the
included
software
to
generate
the
required
valid
one­
hour
average.
The
OH
method
can
provide
valid
1­
hour
averages,
but
it
is
only
for
two
or
three
hours
at
a
time.
Sorbent
traps
sampling
can
produce
1­
hour
samples
only
in
theory;
when
the
sorbent
traps
are
exchanged
every
hour
on
an
around­
the­
clock
basis.
Fuel
sampling
can
also
in
theory
produce
one­
hour
samples.
However,
these
must
still
be
translated
into
verifiable
mercury
flue
gas
emissions.

3.3
Accessibility
and
timeliness
Accessibility
can
be
interpreted
as
the
capability
of
the
monitoring
system
to
provide
unimpeded
access
to
data
records.
Similarly,
timeliness
requires
that
a
continuous
quality
assured,
permanent
record
of
certified
emissions
data
on
an
hourly
basis
is
provided
within
24
hours.

Currently,
for
most
CEMS,
data
output
is
immediate
and
the
accessibility
and
timeliness
criteria
can
be
satisfied.
Modern
data
processing
technology
is
expected
to
allow
for
manipulation
of
the
output
in
most
desired
formats
in
textual,
tabular
or
graphic
form.

At
this
point,
sampling
sorbent
trap
removal
and
placement
is
a
manual
task,
as
is
preparation
for
shipment
and
the
hourly
frequency
of
sampling
may
not
be
practical
for
coal­
fired
power
plants.
The
timeliness
criterion
includes
that
a
record
of
data
from
the
previous
day
must
be
available
within
24
hours.
It
is
unlikely
that
these
conditions
can
be
met
under
the
current
arrangement
while
an
off­
site
laboratory
has
to
analyze
the
sorbent
traps.
Under
the
current
fuel
sampling
and
analysis
procedures,
the
timeliness
criterion
cannot
likely
be
met,
because
hourly
average
data
can
not
likely
be
available
within
24
hours.
