A
PROPOSED
DATA
COLLECTION
EFFORT
TO
ESTABLISH
DEFAULT
FUEL
CONTAMINANT
INPUT
VALUES
FOR
CERTAIN
SOLID
FUEL
CLASSES
July
8,
2003
Introduction
The
forest
products
industry
is
interested
in
establishing
default
fuel
contaminant
input
values
for
one
or
more
classes
of
solid
fuels.
As
described
in
the
white
paper
previously
provided
to
EPA
(
Establishing
Default
Fuel
Contaminant
Input
Values
for
Fuel
Classes
and
Performing
Compliance
Demonstrations
for
Boiler
MACT),
facilities
would
have
the
option
of
using
this
default
value
in
lieu
of
a
site­
specific
value.

At
the
June
5th
meeting
with
forest
products
industry
representatives,
EPA
technical
staff
conceptually
supported
such
default
values.
However,
they
were
unable
to
provide
definitive
guidance
concerning
the
number
of
samples
that
will
be
needed
to
establish
a
default
value
for
a
given
fuel
class
and
how
conservative
the
default
value
should
be.
They
did
indicate
an
average
would
probably
not
be
acceptable
as
a
national
default
value
to
be
used
in
compliance
demonstrations.
They
were
inclined
towards
inherently
conservative
default
values
to
ensure
facilities
opting
to
use
these
values
would
be
have
long­
term
average
emission
rates
less
than
the
boiler
MACT
emission
limits.
EPA
noted
the
procedure
for
establishing
default
values
would
have
to
be
applicable
to
any
class
of
solid
fuel,
not
just
classes
having
very
low
contaminant
values.

With
these
considerations
in
mind,
the
forest
products
industry
is
proposing
a
statistical
procedure
and
associated
sampling
program
to
establish
default
fuel
contaminant
input
values
for
mercury
and
chloride
for
one
class
of
solid
fuels,
hog
fuel.
This
paper
outlines
the
proposed
program.
It
is
being
provided
to
EPA
technical
staff
for
review
and
comment.
The
data
collection
effort
will
commence
if
EPA
finds
it
to
be
an
acceptable
approach
for
developing
the
default
values.

Proposed
Program
Solid
non­
fossil
fuels
of
interest
to
the
forest
products
industry
include
bark,
stemwood
(
wood
fines,
planer
shavings,
sawdust),
hog
fuel
(
typically
a
mix
of
bark
and
stemwood),
panel
trim,
sanderdust,
creosote
treated
wood,
urban
waste
wood,
agricultural
debris
(
orchard
trimmings,
fruit
pits,
peanut
shells,
cottonseeds,
rice
hulls),
wastewater
treatment
plant
residuals,
rejects
from
processing
of
old
corrugated
containers
(
OCC
rejects),
tire
derived
fuel,
and
paper
pellets.
Solid
fossil
fuels
of
interest
are
bituminous
and
subbituminous
coal
and
petroleum
coke.
Although
default
values
would
be
desirable
for
all
of
these
fuel
classes,
hog
fuel
has
been
chosen
to
demonstrate
the
proposed
procedures.
Hog
fuel
is
the
most
widely
used
non­
fossil
fuel
in
the
forest
products
industry,
with
annual
consumption
of
at
least
60
million
tons.

Prior
to
sample
collection,
a
decision
must
be
reached
on
the
statistical
procedure
that
will
be
used
for
establishing
a
default
value.
This
will
govern
the
number
of
facilities
where
samples
must
be
collected
and
the
number
of
samples
to
be
collected
at
each.
Collecting
and
analyzing
samples
from
all
facilities
burning
hog
fuel
would
defeat
the
purpose
of
establishing
a
default
value.
Thus,
a
subset
of
the
facilities
must
be
selected
so
that
appropriate
statistical
analyses
can
be
performed
to
obtain
a
reasonable
default
value.

The
industry
believes
a
default
value
can
be
established
from
a
distribution
of
mean
values
for
fuel
samples
obtained
at
a
reasonable
number
of
facilities
that
burn
hog
fuel.
In
2000,
there
were
at
least
300
boilers
at
forest
products
manufacturing
facilities
burning
hog
fuel.
Detailed
NCASI
survey
information
showed
126
pulp
and
paper
mill
boilers
burned
wood
fuels
at
96
mills,
with
total
fuel
consumption
on
the
order
of
42
million
wet
tons.
Less
comprehensive
information
for
panel
plants
and
sawmills
indicated
there
were
at
least
170
boilers
burning
wood
fuels
at
120
different
locations.
NCASI
does
not
have
a
complete
data
base
for
sawmill
boilers,
although
the
quantity
of
hog
fuel
burned
in
them
should
be
small
compared
to
quantities
burned
at
pulp
and
paper
mills
and
wood
panel
plants.
Obtaining
fuel
samples
from
25
to
30
locations,
selected
to
represent
the
major
geographic
areas
with
forest
products
manufacturing
facilities,
should
provide
a
reasonable
sample
of
the
national
population
of
hog
fuel
being
burned.
Based
on
the
distribution
of
mean
values
from
the
25
to
30
locations,
an
appropriate
statistic,
e.
g.
the
90th
percentile
of
the
mean
values,
can
be
selected
for
the
default
value.

Because
of
the
heterogeneous
nature
of
hog
fuel,
several
samples
from
each
location
will
be
obtained
to
compute
the
mean
fuel
contaminant
input
value
for
the
location.
Data
on
the
variability
of
mercury
and
chloride
in
hog
fuel
is
lacking,
so
it
is
difficult
to
ascertain
how
many
samples
are
needed
to
get
a
good
estimate
of
the
mean
value
(
in
lb/
106
Btu)
for
a
given
location.
While
three
samples
may
be
sufficient
if
variability
is
small,
more
are
desirable
when
variability
is
high.
As
a
starting
point,
getting
seven
samples
per
location
seems
reasonable.
This
will
result
the
analysis
of
approximately
200
fuel
samples.

Each
facility
will
obtain
fuel
samples
following
the
suggested
sampling
protocol
in
Appendix
A.
Sample
collection
will
be
done
at
appropriate
intervals
over
the
course
of
a
one
month
period.
All
samples
will
be
shipped
to
the
NCASI
Southern
Regional
Center,
where
subsamples
will
be
prepared
and
ground.
The
processed
samples
will
be
shipped
to
a
single
laboratory
for
analysis
of
mercury,
chloride
and
higher
heating
value.
NCASI
will
retain
a
duplicate
of
each
sample
for
reanalysis,
should
problems
necessitate
a
reanalysis.
The
laboratory
selected
for
the
analysis
must
have
the
capability
to
obtain
mercury
quantitation
levels
on
the
order
of
1
to
2
ppb
(
dry
basis)
and
about
10
ppm
for
chloride.
Preliminary
estimates
from
three
contract
laboratories
suggest
it
will
cost
roughly
$
150
to
analyze
each
fuel
sample
for
mercury,
chloride
and
higher
heating
value.

Data
Analysis
and
Selection
of
a
Default
Value
The
mercury
content
will
be
calculated
in
lb/
1012Btu
and
the
chloride
content
in
lb/
106
Btu
for
each
sample.
One­
half
the
method
detection
limit
will
be
used
for
the
mercury
or
chloride
concentration
if
the
measured
value
is
below
the
method
detection
limit.
For
each
location,
the
mercury
and
chloride
data
sets
will
be
scanned
for
outliers
(
both
high
and
low)
using
the
Extreme
Value
Test
(
Dixon's
test)
recommended
by
EPA
(
Guidance
for
Data
Quality
Assessment
­
Practical
Methods
for
Data
Analysis,
2000)
for
data
sets
containing
less
than
25
values.
This
test
can
be
used
if
the
data
distribution
(
without
the
suspected
outlier)
is
found
to
be
normal
using
the
Studentized
Range
Test
(
EPA,
2000).
If
this
test
shows
the
data
distribution
is
not
normal,
then
the
suspected
outlier
will
not
be
rejected.
If
it
is
normal,
the
suspect
outlier
will
be
rejected.

The
mean
values
from
each
location
will
be
arrayed
from
lowest
to
highest.
This
distribution
of
mean
values
will
be
used
to
select
a
default
value.
Available
choices
that
might
be
acceptable
to
EPA
include
a
percentile
value
from
the
distribution
of
the
means,
an
upper
confidence
limit
for
the
mean
for
all
facilities
sampled,
and
the
overall
mean
plus
one
or
more
standard
deviations.
Without
a
priori
knowledge
of
this
distribution,
it
is
difficult
to
decide
on
the
most
appropriate
statistic
for
the
default
value.
As
noted
earlier,
EPA
technical
staff
believe
the
default
must
be
higher
than
the
mean
or
median
value
of
this
distribution,
but
less
than
the
highest
value
in
the
distribution.
It
is
understood
EPA
technical
staff
supports
a
choice
that
would
ensure
the
majority
of
facilities
using
the
default
value
will
have
actual
annual
average
fuel
contaminant
input
values
less
than
the
default
value.
Industry
representatives
have
discussed
the
concept
of
default
values
with
OECA
personnel,
but
not
specific
procedures
and
statistics.
Further
discussions
with
EPA
staff
are
needed
to
get
a
better
sense
of
what
statistical
approach
and
default
parameter
would
be
acceptable
to
the
Agency,
as
these
choices
may
influence
the
scope
of
the
sample
collection
effort.
Appendix
A
Draft
Procedure
for
Bark,
Hog
Fuel,
and
Other
Fuel
Sampling
at
Forest
Products
Industry
Sources
Sample
Collection
1.
Sampling
From
a
Belt
(
or
Screw)
Feeding
a
Boiler
or
Silo
a.
Stop
the
belt
and
withdraw
a
6­
inch
wide
(
or
wider)
sample
from
the
full
cross­
section
of
the
stopped
belt
to
obtain
approximately
2
lbs
of
sample.
Transfer
the
sample
to
a
clean
plastic
bag.

b.
Collect
a
minimum
of
three
samples
at
approximately
equal
intervals
during
the
testing
period.

c.
Transfer
all
the
samples
to
a
clean
plastic
bag
to
obtain
a
single
combined
sample
for
further
processing.

2.
Sampling
from
a
Pile
If
a
moving
belt
is
not
accessible,
samples
may
be
obtained
from
a
pile.
For
sampling
from
a
pile,
the
following
procedure
may
be
used:

a.
Select
a
minimum
of
5
sampling
locations
uniformly
spaced
over
the
surface
of
the
pile.

b.
At
each
sampling
site,
dig
into
the
pile
to
a
depth
of
18
inches.
Then,
insert
a
clean
flat,
square
shovel
into
the
hole
and
withdraw
a
sample,
making
sure
that
large
pieces
do
not
fall
off
during
sampling.

c.
Transfer
all
the
samples
to
a
clean
plastic
bag
for
further
processing.

Sample
Subdivision
and
Grinding
1.
Thoroughly
mix
and
pour
the
entire
sample
over
a
clean
plastic
sheet.
2.
If
possible,
break
sample
pieces
larger
than
3
inches
into
smaller
sizes.
3.
Make
a
pie
shape
with
the
entire
sample
and
subdivide
it
into
four
equal
parts.
4.
Separate
one
of
the
quarter
samples
as
the
first
subset.
5.
If
this
subset
is
too
large
for
grinding,
repeat
procedure
3
with
the
quarter
sample
and
obtain
a
one­
quarter
subset
from
this
sample.
6.
Grind
this
sample
in
a
Wiley
mill.
7.
Use
the
procedure
in
3
above
to
obtain
a
one­
quarter
sub­
sample
for
analysis.
If
the
quarter
sample
is
too
large,
subdivide
it
further
using
the
same
procedure.
Appendix
B
Potential/
Proposed
Analytical
Methods
Mercury
There
are
currently
only
two
commercially
available
methods
that
can
provide
reliable
quantification
of
mercury
in
solids
down
to
nominally
1
ppb.

1.
EPA
has
published
(
EPA­
821­
R­
01­
013)
an
appendix
to
Method
1631
giving
procedures
for
the
digestion
of
solids.
Analysis
of
the
resulting
digestates
using
EPA
Method
1631
results
in
sub­
ppb
Method
Detection
Limits
(
MDLs)
and
Minimum
Levels
on
the
order
of
1
ppb
(
ng/
g
in
solids).
Although
the
laboratory
selection
process
is
not
yet
completed,
NCASI
has
obtained
bids
from
two
laboratories
that
can
perform
this
analysis.

2.
There
are
a
number
of
commercially
available
instruments
implementing
a
pyrolysis
sample
preparation
step
for
determining
mercury
in
solids.
However,
NCASI
has
only
been
able
to
find
one
laboratory
that
offers
this
analysis
on
a
commercial
basis.
This
specific
laboratory
uses
the
Milestone
Direct
Mercury
Analyzer
(
DMA­
80),
which
was
the
instrument
used
to
develop
Draft
EPA
Method
7473.
Although
NCASI
has
not
completed
it's
assessment
of
this
methodology
(
i.
e.,
the
laboratory
has
not
yet
provided
"
qualification"
data),
this
approach
should
be
capable
of
delivering
sub­
ppb
MDLs
and
MLs
equivalent
to
Method
1631
(
in
solids).

Depending
on
the
outcome
of
ongoing
discussions
with
all
three
of
the
laboratories
noted
above,
one
of
these
two
methods
(
1631
or
7473)
will
be
used
to
determine
mercury
in
the
hog
fuel
samples
at
a
ML
of
nominally
1
ppb.

Chlorine
Total
chlorine
will
be
determined
using
EPA
Method
9076
(
pyrolysis
followed
by
titration),
or
by
analysis
of
the
HHV
rinsate
using,
e.
g.,
EPA
Method
300
(
ion
chromatography).
Note
that
the
analysis
of
the
HHV
rinsate
is
functionally
equivalent
to
a
determination
by
EPA
Method
5050.

Higher
Heating
Value
(
HHV)/
Gross
Calorific
Value
ASTM
D2015
will
be
used
to
determine
HHV.
