SAP
Minutes
No.
2004­
06
A
Set
of
Scientific
Issues
Being
Considered
by
the
Environmental
Protection
Agency
Regarding:

FUMIGANT
BYSTANDER
EXPOSURE
MODEL
REVIEW:
PROBABILISTIC
EXPOSURE
AND
RISK
MODEL
FOR
FUMIGANTS
(
PERFUM)
USING
IODOMETHANE
AS
A
CASE
STUDY
AUGUST
24
and
25,
2004
FIFRA
Scientific
Advisory
Panel
Meeting,
held
at
the
Holiday
Inn
­
National
Airport,
Arlington,
Virginia
2
of
43
NOTICE
These
meeting
minutes
have
been
written
as
part
of
the
activities
of
the
Federal
Insecticide,
Fungicide,
and
Rodenticide
Act
(
FIFRA),
Scientific
Advisory
Panel
(
SAP).
The
meeting
minutes
represent
the
views
and
recommendations
of
the
FIFRA
SAP,
not
the
United
States
Environmental
Protection
Agency
(
Agency).
The
content
of
the
meeting
minutes
does
not
represent
information
approved
or
disseminated
by
the
Agency.
The
meeting
minutes
have
not
been
reviewed
for
approval
by
the
Agency
and,
hence,
the
contents
of
these
meeting
minutes
do
not
necessarily
represent
the
views
and
policies
of
the
Agency,
nor
of
other
agencies
in
the
Executive
Branch
of
the
Federal
government.
Nor
does
mention
of
trade
names
or
commercial
products
constitute
a
recommendation
for
use.

The
FIFRA
SAP
is
a
Federal
advisory
committee
operating
in
accordance
with
the
Federal
Advisory
Committee
Act
and
established
under
the
provisions
of
FIFRA
as
amended
by
the
Food
Quality
Protection
Act
(
FQPA)
of
1996.
The
FIFRA
SAP
provides
advice,
information,
and
recommendations
to
the
Agency
Administrator
on
pesticides
and
pesticiderelated
issues
regarding
the
impact
of
regulatory
actions
on
health
and
the
environment.
The
Panel
serves
as
the
primary
scientific
peer
review
mechanism
of
the
EPA,
Office
of
Pesticide
Programs
(
OPP),
and
is
structured
to
provide
balanced
expert
assessment
of
pesticide
and
pesticide­
related
matters
facing
the
Agency.
Food
Quality
Protection
Act
Science
Review
Board
members
serve
the
FIFRA
SAP
on
an
ad
hoc
basis
to
assist
in
reviews
conducted
by
the
FIFRA
SAP.
Further
information
about
FIFRA
SAP
reports
and
activities
can
be
obtained
from
its
website
at
http://
www.
epa.
gov/
scipoly/
sap/
or
the
OPP
Docket
at
(
703)
305­
5805.
Interested
persons
are
invited
to
contact
Myrta
R.
Christian,
SAP
Designated
Federal
Official,
via
e­
mail
at
christian.
myrta@.
epa.
gov.

In
preparing
the
meeting
minutes,
the
Panel
carefully
considered
all
information
provided
and
presented
by
the
Agency
presenters,
as
well
as
information
presented
by
public
commenters.
This
document
addresses
the
information
provided
and
presented
by
the
Agency
within
the
structure
of
the
charge.
3
of
43
CONTENTS
PARTICIPANTS
............................................................................................................
5
INTRODUCTION..........................................................................................................
7
PUBLIC
COMMENTERS..............................................................................................
8
SUMMARY
OF
PANEL
DISCUSSION
AND
RECOMMENDATIONS
.......................
9
PANEL
DELIBERATIONS
AND
RESPONSE
TO
CHARGE
.....................................
12
REFERENCES.............................................................................................................
39
4
of
43
SAP
Minutes
No.
2004­
06
A
Set
of
Scientific
Issues
Being
Considered
by
the
Environmental
Protection
Agency
Regarding:

FUMIGANT
BYSTANDER
EXPOSURE
MODEL
REVIEW:
PROBABILISTIC
EXPOSURE
AND
RISK
MODEL
FOR
FUMIGANTS
(
PERFUM)
USING
IODOMETHANE
AS
A
CASE
STUDY
AUGUST
24
and
25,
2004
FIFRA
Scientific
Advisory
Panel
Meeting,
held
at
the
Holiday
Inn
­
National
Airport,
Arlington,
Virginia
Myrta
R.
Christian,
M.
S.
Stephen
M.
Roberts,
Ph.
D.
Designated
Federal
Official
FIFRA
SAP,
Session
Chair
FIFRA
Scientific
Advisory
Panel
FIFRA
Scientific
Advisory
Panel
Date:
November
9,
2004
Date:
November
9,
2004
5
of
43
Federal
Insecticide,
Fungicide,
and
Rodenticide
Act
Scientific
Advisory
Panel
Meeting
August
24
and
25,
2004
FUMIGANT
BYSTANDER
EXPOSURE
MODEL
REVIEW:
PROBABILISTIC
EXPOSURE
AND
RISK
MODEL
FOR
FUMIGANTS
(
PERFUM)
USING
IODOMETHANE
AS
A
CASE
STUDY
PARTICIPANTS
FIFRA
SAP
Chair
Stephen
M.
Roberts,
Ph.
D.,
Professor
&
Program
Director,
University
of
Florida,
Center
for
Environmental
&
Human
Toxicology,
Gainesville,
FL
Designated
Federal
Official
Myrta
R.
Christian,
M.
S.,
FIFRA
Scientific
Advisory
Panel
Staff,
Office
of
Science
Coordination
and
Policy,
EPA
FIFRA
Scientific
Advisory
Panel
Members
Steven
G.
Heeringa,
Ph.
D.,
Research
Scientist
&
Director
for
Statistical
Design,
University
of
Michigan,
Institute
for
Social
Research,
Ann
Arbor,
MI
Kenneth
M.
Portier,
Ph.
D.,
Associate
Professor,
Statistics,
Institute
of
Food
and
Agricultural
Sciences,
University
of
Florida,
Gainesville,
FL
FQPA
Science
Review
Board
Members
Daniel
C.
Baker,
Ph.
D.,
Senior
Consultant,
Environmental
Computing
Shell
Global
Solutions
US,
Houston,
TX
Paul
W.
Bartlett,
M.
A.,
Research
Associate,
Center
for
the
Biology
of
Natural
Systems
Queens
College,
City
University
of
New
York,
N.
Y.

Adel
F.
Hanna,
Ph.
D.,
Research
Professor,
Carolina
Environmental
Program,
University
of
North
Carolina
at
Chapel
Hill,
Chapel
Hill,
NC
Michael
S.
Majewski,
Ph.
D.,
Research
Chemist,
US
Geological
Survey,
Sacramento,
CA
David
R.
Maxwell,
M.
P.
A.,
M.
B.
A.,
C.
C.
M.,
Environmental
Specialist/
Air
Quality
Monitoring
Specialist,
National
Parks
Service,
Denver,
CO
Li­
Tse
Ou,
Ph.
D.,
Scientist,
Soil
&
Water
Science
Department,
University
of
Florida,
Gainesville,
FL
6
of
43
James
N.
Seiber,
Ph.
D.,
Director,
U.
S.
Department
of
Agriculture
/
ARS/,
Western
Regional
Research
Center,
Albany,
CA
Frederick
M.
Shokes,
Ph.
D.,
Center
Director
and
Professor,
Tidewater
Agricultural
Research
and
Extension
Center,
Virginia
Tech
University,
Suffolk,
VA
Mitchell
J.
Small,
Ph.
D.,
H.
John
Heinz
III
Professor
of
Environmental
Engineering
Carnegie
Mellon
University,
Pittsburgh,
PA
Thomas
O.
Spicer,
III,
Ph.
D.,
Professor
and
Head,
Department
of
Chemical
Engineering,
University
of
Arkansas,
Fayetteville,
AR
Dong
Wang,
Ph.
D.,
Associate
Professor,
Department
of
Soil,
Water
&
Climate,
University
of
Minnesota,
St.
Paul,
MN
Eric
D.
Winegar,
Ph.
D.,
QEP,
Applied
Measurement
Science,
Fair
Oaks,
CA
Scott
R.
Yates,
Ph.
D.,
Interim
Research
Leader,
U.
S.
Department
of
Agriculture/
ARS/
GEBJr,
Salinity
Lab.,
Soil
Physics
&
Pesticides
Research
Unit,
Riverside,
CA
7
of
43
INTRODUCTION
On
August
24­
25,
2004,
August
26­
27,
2004,
and
September
9­
10,
2004,
the
Federal
Insecticide,
Fungicide,
and
Rodenticide
Act
(
FIFRA),
Scientific
Advisory
Panel
(
SAP)
held
three
separate
meetings
to
consider
and
review
three
fumigant
bystander
exposure
models.
These
meeting
minutes
focus
on
the
FIFRA
SAP
meeting
held
August
24­
25,
2004
to
review
the
Probabilistic
Exposure
and
Risk
model
for
FUMigants
(
PERFUM)
using
iodomethane
as
a
case
study.
The
FIFRA
SAP
also
met
on
August
26­
27,
2004
to
review
the
Fumigant
Exposure
Modeling
System
(
FEMS)
using
metam
sodium
as
a
case
study
and
on
September
9­
10,
2004
to
review
the
SOil
Fumigant
Exposure
Assessment
System
(
SOFEA
©
)
using
telone
as
a
case
study.
Minutes
from
each
of
these
FIFRA
SAP
meetings
are
available
from
the
FIFRA
SAP
website
at
http://
www.
epa.
gov/
scipoly/
sap/
or
the
OPP
Docket
at
(
703)
305­
5805.

Advance
notice
of
the
August
24­
25,
2004
meeting
was
published
in
the
Federal
Register
on
July
23,
2004.
The
review
was
conducted
in
an
open
Panel
meeting
held
in
Arlington,
Virginia
and
was
chaired
by
Stephen
M.
Roberts,
Ph.
D.
Mrs.
Myrta
R.
Christian
served
as
the
Designated
Federal
Official.

EPA's
Office
of
Pesticide
Programs
is
engaged
in
pesticide
tolerance
reassessment
activities
as
mandated
by
the
Food
Quality
Protection
Act
(
1996).
As
part
of
that
process,
the
Agency
is
currently
involved
in
the
development
of
a
comparative
risk
assessment
for
six
soil
fumigant
pesticides
that
include
chloropicrin,
dazomet,
iodomethane,
methyl
bromide,
metamsodium
and
telone.
Each
of
these
chemicals
has
a
degree
of
volatility
associated
with
it
which
is
a
key
characteristic
needed
to
achieve
a
satisfactory
measure
of
efficacy.
This
volatility,
however,
can
contribute
to
human
exposures
because
these
chemicals
can
travel
to
non­
target
receptors,
such
as
nearby
human
populations.
Commonly
referred
to
as
bystander
exposure,
it
is
considered
by
the
Agency
to
be
the
primary
pathway
through
which
human
exposure
to
fumigants
may
occur.

In
order
to
address
bystander
exposures,
the
Agency
developed
a
method
based
on
a
deterministic
use
of
the
Office
of
Air
model
entitled
Industrial
Source
Complex
Short
Term
Model
(
ISCST3)
that
is
routinely
used
for
regulatory
decisions.
ISCST3
is
publicly
available
from
the
following
Agency
website
(
http://
www.
epa.
gov/
scram001/
tt22.
htm#
isc.
In
this
approach,
the
Agency
uses
chemical­
specific
measures
of
volatility
to
quantify
field
emission
rates
for
modeling
purposes.
Additionally,
the
Agency
uses
standardized
meteorological
conditions
which
represent
a
stable
atmosphere
and
unidirectional
wind
patterns
that
provide
conservative
estimates
of
exposure.

Stakeholders
expressed
concern
that
the
conditions
represented
by
the
current
approach
provide
results
that
are
not
sufficiently
refined
for
regulatory
actions
such
as
risk
mitigation.
In
response,
the
Arvesta
Corporation,
the
registrant
for
iodomethane,
has
submitted
a
model
entitled
Probabilistic
Exposure
and
Risk
model
for
FUMigants
(
PERFUM)
for
consideration
as
a
possible
refinement
to
the
Agency's
approach.
The
Agency
believes
that
this
model
also
may
have
the
potential
to
be
used
generically
to
calculate
risks
for
the
six
soil
fumigants
being
evaluated
in
the
current
risk
assessment.
The
key
differences
between
PERFUM
and
the
current
8
of
43
Agency
approach
are
that
it
incorporates
ranges
of
both
field
emission
rates
and
5
years
of
meteorological
data
from
stations
in
areas
where
iodomethane
is
used.

The
purpose
of
this
meeting
of
the
FIFRA
Scientific
Advisory
Panel
(
SAP)
was
to
evaluate
the
approaches
contained
in
PERFUM
for
integrating
actual
meteorological
data
into
ISCST3
analyses.
Additionally,
the
Agency
was
seeking
a
specific
evaluation
of
the
methods
used
pertaining
to
field
emission
rates,
statistical
approaches
for
data
analysis,
receptor
locations,
and
defining
the
exposed
populations.
Finally,
the
Agency
was
seeking
a
determination
as
to
the
scientific
validity
of
the
overall
approach
included
in
PERFUM.
The
agenda
for
this
SAP
meeting
involved
an
introductory
overview
of
the
current
risk
assessment
approach
by
the
EPA
provided
by
Mr.
Jeffrey
Dawson
(
Health
Effects
Division,
Office
of
Pesticide
Programs).
On
behalf
of
the
Arvesta
Corporation,
a
detailed
presentation
of
the
PERFUM
model
was
given
by
Dr.
Richard
Reiss
of
Sciences
International,
Inc.,
located
in
Alexandria,
Virginia.
Dr.
Terri
Barry
and
Dr.
Randy
Segawa
from
California
Department
of
Pesticide
Regulation
also
participated
with
EPA
in
this
SAP
meeting.
Mr.
James
J.
Jones
(
Director,
Office
of
Pesticide
Programs)
and
Ms.
Margaret
Stasikowski
(
Director,
Health
Effects
Division,
Office
of
Pesticide
Programs)
offered
opening
remarks
at
the
meeting.

In
preparing
these
meeting
minutes,
the
Panel
carefully
considered
all
information
provided
and
presented
by
the
Agency
presenters,
as
well
as
information
presented
by
public
commenters.
This
document
addresses
the
information
provided
and
presented
at
the
meeting,
especially
the
response
to
the
charge
by
the
Agency.

PUBLIC
COMMENTERS
Oral
statements
were
presented
as
follows:

Ms.
Shelley
Davis,
on
behalf
of
the
California
Rural
Legal
Assistance
Foundation
and
the
Farmworker
Justice
Fund
James
Platt,
Ph.
D.
on
behalf
of
the
Arvesta
Corporation
9
of
43
SUMMARY
OF
PANEL
DISCUSSION
AND
RECOMMENDATIONS
1.
The
Panel
commended
the
Agency
for
convening
this
scientific
review
of
the
PERFUM
bystander
exposure
assessment
model.
The
Agency
is
addressing
concerns
voiced
by
stakeholders
that
the
conditions
represented
in
the
current
EPA
approach
for
exposure
assessment
modeling
provide
results
that
are
not
sufficiently
refined
to
make
fair
and
accurate
regulatory
actions
such
as
risk
mitigation.
By
its
design,
PERFUM
addresses
a
number
of
these
concerns.
PERFUM
uses
standard
EPA­
approved
regulatory
models
in
a
manner
that
can:
utilize
historical
meteorological
datasets
to
more
fully
characterize
potential
downwind
concentrations,
determine
the
upper
percentile
of
the
distribution
of
buffer­
zone
distances,
provide
a
distribution
of
margins
of
exposure
(
MOEs)
for
a
user­
supplied
buffer
zone,
and
incorporate
emission
profiles
from
a
variety
of
sources
(
i.
e.,
constant,
variable
strength,
point,
or
area
sources).

2.
The
description
of
the
model
components
was
considered
to
be
scientifically
sound
and
a
reasonably
skilled
user
could
replicate
the
calculations.
The
algorithms
in
the
code
appear
to
perform
the
function
intended
in
the
documentation.
However,
the
discussion
on
how
margins
of
exposure
and
exceedance
probabilities
are
computed
for
buffer
zone
lengths
should
be
presented
more
clearly,
additional
documentation
of
the
field
trials
is
needed,
and
the
direct
flux
measurements
and
data
reduction
techniques
should
be
better
documented.

3.
PERFUM
estimates
period
flux
rates
using
an
indirect
(
back
calculation)
method.
However,
large
variations
in
period
flux
rates
have
been
observed
from
different
approaches
to
calculate
flux,
even
when
the
same
experimental
data
(
i.
e.,
air
concentration,
wind
speed,
temperature)
is
used
for
each
approach.
The
Panel
thought
that
direct
methods,
such
as
the
aerodynamic
gradient
methods,
need
to
be
investigated
to
determine
if
they
are
more
accurate.
The
Panel
also
thought
that
additional
concentration
data
in
the
vertical
dimension
are
needed
(
along
with
the
1.5
meter
data)
to
improve
the
estimate
of
the
emission
flux
obtained
from
the
indirect
method.
The
availability
of
vertical
concentration
profiles
may
be
one
other
way
to
improve
the
calibration
of
the
model.

4.
The
Panel
thought
that
representing
emission
flux
as
a
probability
distribution
to
address
uncertainty
and
real
variation
is
a
worthy
objective.
There
should
be
a
distinction
between
uncertainty
arising
from
measurements,
uncertainty
from
the
limited
sampling,
uncertainty
from
measuring
emission
flux
from
back
calculations
by
the
dispersion
model,
and
uncertainty
from
the
actual
variability
of
emissions.

5.
PERFUM
accounts
for
variability
in
the
period
flux
measurement
by
calculating
the
standard
error
between
the
measured
and
model
concentrations.
This
gives
a
measurement
of
the
error/
variability
for
only
that
field
study.
However,
total
emissions
tend
to
be
more
accurately
measured
compared
to
short­
interval
emissions
and
variability
within
a
study
tends
to
be
smaller
than
the
variation
between
studies.
This
could
make
the
implementation
of
PERFUM
problematic.
10
of
43
6.
It
is
unlikely
that
generalizing
the
use
of
PERFUM
to
a
regional
or
state
scale
will
be
appropriate
if
the
emission
data
were
collected
at
a
single
(
or
relatively
few)
location(
s).
To
fully
capture
the
variability/
uncertainty
in
emission
and
concentrations,
an
extensive
data
set
is
needed
to
account
for
soil
and
environmental
conditions,
differences
in
meteorology,
and
seasonal
effects.

7.
The
Panel
members
expressed
concern
with
the
manner
in
which
PERFUM
selects
the
uncertain
emission
rates
in
sequential
time
periods.
Since
many
of
the
causes
of
variability
are
likely
to
persist
over
time,
sampling
the
uncertain
emission
rates
in
sequential
time
periods
independently
of
the
previous
time
periods
will
create
unrealistic
patters
in
the
daily
emission
rate,
and
may
also
underestimate
the
frequency
of
high­
end
emissions
and
associated
high­
end
buffer
zone
lengths.
A
method
is
needed
to
incorporate
temporal
persistence
into
the
estimate
of
flux
rate.

8.
Several
Panel
members
suggested
that
soils­
based
mechanistic
models
should
be
investigated
and,
potentially,
incorporated
into
PERFUM.
Models
of
this
type
can
integrate
soil
moisture,
temperature,
organic
matter,
and
other
soils
factors
into
the
prediction
of
fumigant
fate,
soil
transport
and
volatilization.
A
stand­
alone
model
could
be
used
to
obtain
period­
averaged
emission
rates
for
use
with
PERFUM.
The
flexibility
to
incorporate
period­
flux
measurements
from
other
sources
is
desirable.

9.
The
PERFUM
modeling
system
uses
a
traditional
square
grid
instead
of
radial
grid.
A
more
adaptive
grid
approach
that
allows
coarse
grid
information
to
be
used
to
suggest
additional
grid
points
to
improve
buffer
zone
estimates
should
be
examined.
A
nested
grid
might
significantly
improve
the
computational
efficiencies,
while
maintaining
equivalent
accuracy.
The
Panel
supported
any
effort
that
would
increase
computational
efficiency
and
recommended
consideration
of
irregularly
shaped
fields.

10.
The
primary
source
of
meteorological
data
should
be
in
the
following
order:
(
1)
National
Weather
Service
(
NWS);
(
2)
ASOS
(
FAA)
data
(
ASOS
came
on
line
in
early
1990,
a
very
rich
data
source
that
covers
the
country);
(
3)
State
climatology
or
agricultural
weather
stations
(
e.
g.,
CIMIS,
FAWN,
SCAN,
ECONet,
etc.);
and
(
4)
meteorological
data
selected
from
regulatory
agency
sites
or
industrial
sites
if
the
data
are
quality
assured.
The
NWS
has
cloud
cover
data
as
one
of
its
measured
parameters,
and
cloud
cover
data
are
required
by
the
Turner
method
for
calculating
atmospheric
stability.
Observational
networks
from
Florida
and
California
do
not
include
this
parameter.

11.
Filling
in
missing
data
is
tricky,
especially
for
cloud
cover,
which
is
one
of
the
parameters
used
by
the
ISCST3
model
to
designate
a
stability
index.
The
EPA
recommendations
are
to
use
data
within
a
few
hours
of
the
missing
period.
While
this
might
be
a
reasonable
approach
for
parameters
such
as
temperature,
a
highly
variable
and
uncertain
parameter
such
as
cloud
cover
will
be
much
less
accurate
when
using
the
interpolation
approach.

12.
The
report
describes
various
sources
of
uncertainty
in
the
study
and
provides
an
initial
assessment
of
the
sensitivity
of
model
predictions
to
selected
alternative
model
assumptions
and
parameter
values.
The
factors
discussed
include:
the
flux
emission
estimate/
profile,
11
of
43
meteorological
data
and
sources
of
datasets,
anemometer
height,
calms
processing,
nonmeteorological
environmental
factors,
"
Gaussian
formulated"
model­
model
comparisons,
indoor
exposure,
human
activity
patterns,
coincidental
temporally/
spatially
close
multi­
field
applications,
seasonality,
and
horizontal
placement
of
monitors
during
field
studies
for
flux
estimation.
The
uncertainty
associated
with
these
factors
should
be
considered
in
an
integrated
manner.

13.
The
Panel
thought
that
many
of
the
limitations
in
the
ISCST3
model
(
the
core
of
PERFUM)
would
be
alleviated
when
the
Agency
adopts
AERMOD.
The
Panel
thought
approving
the
AERMOD
model
should
be
a
high
priority
of
the
Agency.
12
of
43
PANEL
DELIBERATIONS
AND
RESPONSE
TO
CHARGE
The
specific
issues
addressed
by
the
Panel
are
keyed
to
the
Agency's
background
documents,
references,
and
the
Agency's
charge
questions.

Charge
Critical
Element
1:
Documentation
Question
1:
The
background
information
presented
to
the
SAP
Panel
by
the
PERFUM
developers
provides
both
user
guidance
and
a
technical
overview
of
the
system.

a)
Please
comment
on
the
detail
and
clarity
of
this
document.
Are
the
descriptions
of
the
specific
model
components
scientifically
sound?

Response
The
report
presented
to
the
FIFRA
SAP
appears
to
be
preliminary
because
there
is
an
indication
that
additional
field­
testing
was
conducted,
but
has
yet
to
be
included
in
the
report.
Dr.
Richard
Reiss
discussed
the
added
field
test
information
as
part
of
his
presentation.
Also,
Dr.
Reiss
discussed
other
minor
modifications
made
to
the
model
in
the
time
between
preparation
of
the
report
and
the
meeting
such
as
a
modification
of
the
method
for
calculating
random
deviations
in
the
flux.
In
his
presentation
to
the
Panel,
Dr.
Reiss
discussed
details
concerning
how
flux
data
were
obtained
from
the
field
tests
on
iodomethane.
Dr.
Reiss
indicated
that
the
documentation
would
be
updated
to
reflect
the
added
content
included
in
his
presentation,
including
background
information
on
the
physical
and
chemical
properties
of
iodomethane.
The
report
and
presentation
were
well
received
by
the
Panel.
Some
Panel
members
voiced
concern
as
to
whether
supporting
documentation
would
be
published
in
a
peer­
reviewed
journal
article
and
Dr.
Reiss
indicated
that
there
were
plans
to
do
this.

Other
details
that
should
be
addressed
with
regard
to
documentation
include:

 
Because
PERFUM
is
built
around
ISCST3,
it
was
suggested
that
a
flow
diagram
be
included
in
the
documentation
that
outlines
the
simulation
process
in
PERFUM,
including
details
of
ISCST3.

 
It
was
observed
that
parts
of
the
PERFUM
system
have
been
extensively
validated,
and
the
report
would
benefit
from
referencing
this
peer­
reviewed
work.

 
The
discussion
on
how
margins
of
exposure
are
computed
and
how
exceedance
probabilities
are
computed
for
buffer
zone
lengths
should
be
presented
more
clearly.

 
Additional
documentation
of
the
field
trials
is
needed,
including
information
about
soil
characteristics
(
i.
e.,
soil
type,
organic
matter
content,
soil
moisture,
etc.),
meteorological
data,
and
field
preparation
methods.
13
of
43
 
ISCST3,
as
used
in
PERFUM
was
not
shown
in
the
documentation
to
include
processes
of
atmospheric
degradation,
deposition,
or
interaction
with
atmospheric
moisture,
(
e.
g.,
fog
and
rain).
The
effects
of
complex
terrain
were
not
included
in
the
documentation,
but
should
be
since
hilly,
uneven
terrain
often
requires
fumigant
treatment.
Even
though
photodegradation
would
not
be
expected
to
affect
the
buffer
zone
distances
for
iodomethane
because
this
process
is
relatively
slow
in
comparison
to
the
travel
times
to
buffer
zone
distances,
an
example
of
this
would
be
useful.
Examples
of
ISCST3'
s
capability
of
including
the
effects
of
deposition
and
interaction
with
atmospheric
moisture
(
e.
g.,
fog)
would
also
be
useful.

 
The
direct
flux
measurements
and
data
reduction
techniques
should
be
documented.

 
Example
simulations
using
PERFUM
to
address
irregularly
shaped
fields
would
be
useful.

 
Discussion
regarding
terrain
and
obstacle
effects
should
be
included.

 
The
discussion
of
how
PERFUM
deals
with
calm
wind
conditions
in
simulations
(
predictive
mode)
as
well
as
flux
determination
from
field
tests
should
be
more
extensive.

The
description
of
the
model
components
was
considered
to
be
scientifically
sound
and
a
reasonably
skilled
user
could
replicate
the
calculations.
The
algorithms
in
the
code
appear
to
perform
the
function
intended
in
the
documentation
with
the
exceptions
of
the
suggested
report
revisions
discussed
above.

b)
Do
the
algorithms
in
the
annotated
code
perform
the
functions
as
defined
in
this
document?
Please
discuss
any
difficulties
encountered
with
respect
to
loading
the
software
and
evaluating
the
system,
including
the
presented
case
study.

Response
Comments
varied
widely
depending
on
a
Panel
member's
level
of
expertise
with
computers
and
programming,
and
the
amount
of
time
they
invested
with
the
program
running
the
example
cases
provided
before
the
meeting.
Each
of
the
following
was
reported
by
at
least
one
Panel
member:
not
being
able
to
access
the
files
because
of
file
attribute
problems
(
files
copied
from
a
CD
are
set
as
"
read­
only"
by
default);
getting
an
error
message
after
following
the
instructions
in
the
README
file
included
on
the
CD;
having
sorted
out
the
problems
to
execute
PERFUM
but
could
not
execute
PERFUM_
MOE
in
the
time
available;
and
being
able
to
execute
PERFUM
and
PEFUM_
MOE.
Some
problems
seemed
to
be
caused
by
file
name
discrepancies.
Other
comments
with
regard
to
the
software
include:

 
There
were
some
minor
FORTRAN
issues
that
may
limit
portability
between
compilers.
A
few
lines
of
program
code
used
tabs
to
align
with
column
7,
suggesting
a
limitation
with
previous
FORTRAN
standards.
This
may
lead
to
substandard
performance
on
some
compilers.
MAX
and
MOD
functions
are
presently
coded
for
use
with
the
Lahey
14
of
43
compiler,
which
may
be
incompatible
with
other
compilers.
It
was
noted
that
implementation
of
a
GUI
interface
may
be
more
easily
accomplished
with
other
compilers
such
as
the
Digital
Visual
FORTRAN
compiler.

 
PERFUM
is
presently
executed
in
a
(
DOS)
command
window.
Some
Panel
members
thought
that
using
this
interface
may
hinder
inexperienced
users
from
successfully
utilizing
PERFUM,
and
that
a
windows­
based
GUI
would
help
future
users.

 
The
format
of
the
PERFUM
input
file
was
generally
found
to
be
quite
usable,
particularly
because
each
line
had
a
single
input
parameter,
and
documentation
about
the
input
value
could
be
included
on
the
same
line
after
its
value.
Some
Panel
members
thought
experienced
users
may
prefer
the
present
interface
to
a
GUI
interface.

 
Some
Panel
members
suggested
that
PERFUM
should
be
modified
to
allow
for
different
simulation
durations.

 
Some
Panel
members
recommended
that
model
output
be
tailored
to
a
standard
graphical
output
package.
Dr.
Reiss
responded
that
such
a
project
might
be
considered
in
the
future,
but
that
the
market
may
not
support
an
extensive
effort.

Critical
Element
2:
System
Design/
Inputs
Question
2:
In
Section
2.3:
Development
of
the
PERFUM
Modeling
System
of
the
background
document,
a
series
of
detailed
individual
processes
and
components
included
in
PERFUM
are
presented.
The
key
processes
include
(
1)
incorporation
of
ISCST3
into
PERFUM,
(
2)
probabilistic
treatment
of
flux
rates;
and
(
3)
development
of
a
receptor
grid.

a)
Please
comment
on
these
proposed
processes,
the
nature
of
the
components
included
in
PERFUM
and
the
data
needed
to
generate
an
analysis
using
PERFUM.

Response
(
1)
Incorporation
of
ISCST3
into
PERFUM
ISCST3
is
a
dispersion
model
required
by
the
USEPA
for
use
in
applications
intended
for
air
quality
regulation.
ISCST3
is
run
within
PERFUM
as
a
callable
subroutine.
There
are
additional
program
manipulations
to
extract
necessary
information
that
ISCST3
would
otherwise
not
provide
as
output.
The
Panel
agreed
that
this
is
very
desirable
since
additional,
valuable
information
is
obtained
compared
to
using
the
ISCST3
output
alone.
The
approach
is
appropriately
described
in
the
documentation
and
increases
the
overall
usefulness
for
estimating
buffer
zone
distances
around
fields
fumigated
with
iodomethane.
The
approach
also
yields
reduced
output
file
sizes,
reduced
run
times,
allows
for
24­
hour
average
concentration
estimates
for
time
periods
other
than
midnight
to
midnight,
and
provides
hourly
flux
values
as
a
probabilistic
variable.
15
of
43
Although
ISCST3
was
developed
for
use
with
emissions
from
industrial­
source
complexes,
some
justification
should
be
provided
showing
that
it
is
appropriate
to
use
ISCST3
to
predict
the
movement
of
agricultural
fumigants
down­
wind
from
treated
fields.

Data
required
by
PERFUM
include
a
source
flux
term
along
with
various
meteorological
data.
The
source
flux
term
was
obtained
through
calibration
using
measured
data
from
six
field
studies
(
five
conducted
in
California
and
one
in
Florida).
An
Indirect
Flux
Method
was
used
to
estimate
the
source­
flux
term.
For
each
period,
this
involves
selecting
a
default
source­
flux
term,
running
ISCST3,
plotting
the
measured
vs.
modeled
air
concentrations
at
all
the
sample
locations,
and
fitting
a
regression
line
to
the
plotted
values.
The
slope
of
the
regression
line
is
the
best
estimate
of
the
emission
rate.
In
one
experimental
study,
the
Indirect
Flux
Method
was
compared
to
a
flux
estimate
obtained
using
the
Aerodynamic
Gradient
Method
(
a
direct
flux
method),
with
good
results.
For
the
experiments
considered,
the
first
24­
h
period
after
application
was
found
to
be
the
most
important
in
terms
of
the
highest
downwind
air
concentrations,
and
the
model
output
focuses
on
determining
the
buffer
zone
based
on
this
time
period.

Although
the
actual
field
emission
flux
depends
on
a
variety
of
factors
such
as
the
application
method,
injection
depth,
tarp
type
and
thickness,
soil
properties,
and
the
physical/
chemical
properties
of
iodomethane,
the
model
obtains
flux
values
independently
of
these
important
variables.
Many
soil,
environmental,
and
application
factors
have
a
significant
effect
on
volatilization
rates.
The
Panel
had
concerns
that
results
obtained
using
data
from
only
a
few
similar
studies
may
not
be
appropriately
applied
in
other
geographic
areas,
at
other
times,
or
for
other
fumigant
application
methods.
It
may
be
appropriate
to
estimate
buffer
zones
for
the
particular
application
method,
location,
and
time
for
which
data
are
available.
The
Panel
recommended
more
field
tests
to
increase
confidence
in
the
model
output.

If
iodomethane
is
approved
for
use,
it
is
possible
that
concurrent
applications
in
high­
use
areas
may
increase
the
background
concentrations
and
may
affect
the
buffer
zone
estimate.
The
Panel
agreed
that,
initially,
the
background
levels
would
be
low
compared
to
the
source
term
and
may
not
affect
the
buffer
zone
boundary.
As
the
fumigant
use
increases,
however,
concurrent
applications
might
occur
and
the
model
should
be
configured
to
allow
multiple
source
inputs.

The
ISCST3
model
is
always
run
in
the
rural
mode,
even
for
applications
of
iodomethane
occurring
near
homes,
urban
areas,
or
in
areas
with
trees
or
other
topography/
obstructions
that
might
cause
deviations
from
rural
conditions.
For
situations
that
produce
increased
turbulence,
running
the
model
in
the
rural
mode
gives
longer
buffer
zones
that
might
be
needlessly
conservative.

The
buffer
zone
estimates
were
calculated
using
hourly
meteorological
data
from
three
data
sets
containing
five
years
of
records
 
NWS
(
nationwide),
CIMIS
(
California),
and
ASOS
(
FAA)
or
FAWN
(
Florida).
These
networks
provide
usable
data,
but
only
the
NWS
data
are
taken
at
the
heights
used
by
standard
methods
for
estimating
atmospheric
stability.
The
Panel
considers
the
NWS
network
to
be
the
most
complete
and
have
the
best
data
quality
control;
therefore,
these
data
are
the
preferred
source
of
meteorological
data.
16
of
43
Overall,
the
modeling
results
using
the
5­
year
meteorological
data
from
these
stations
were
relatively
similar.
The
Panel
recommended
that
the
meteorological
data
from
the
station(
s)
near
to
the
actual
application
area
should
be
used
in
the
buffer
zone
estimations
even
though
there
were
no
statistically
significant
differences
found
between
the
coastal
and
inland
stations
or
stations
in
agricultural
regions
compared
to
more
urban
stations.
Also,
since
wind
can
vary
dramatically
in
more
complex
situations,
e.
g.,
narrow
valleys,
it
may
be
appropriate
to
install
weather
stations
in
these
areas
to
obtain
data
needed
to
improve
the
estimation
of
buffer
zones.

(
2)
Probabilistic
Treatment
of
Flux
Rates
PERFUM
estimates
period
flux
using
an
indirect
(
back
calculation)
method.
Direct
methods,
such
as
the
Aerodynamic
Gradient
Methods
need
to
be
investigated
to
determine
if
this
approach
is
accurate,
and
more
comparisons
should
be
provided
in
the
PERFUM
documentation.
The
uncertainty
in
flux
estimates
can
be
obtained
for
both
approaches,
but
only
accounts
for
experimental
and
model
errors.
Uncertainty
due
to
differences
in
soil
types,
environmental
factors,
and
regional
and
temporal
translocation
are
not
included.

There
is
a
need
for
a
clear
discussion
on
the
estimation
of
standard
error
of
the
predicted
flux
rate
(
slide
22
error
and
relationship
to
slide
43{
SAP
Presentation:
PERFUM
Probabilistic
Exposure
and
Risk
Model
for
FUMigants
by
Dr.
Richard
Reiss
in
EPA,
OPP
Docket,
Document
ID
#
OPP­
0240­
0015}).
A
Panel
Member
questioned
whether
this
standard
error
includes
projection
bounds
(
for
future
values)
rather
than
confidence
bounds
(
expected
value
of
estimate).

Agricultural
fields
tend
to
be
large
in
area
and
are
treated
with
pesticides
one
row
or
segment
at
a
time.
At
the
beginning
of
the
application
process,
the
segment
first
treated
begins
to
emit
while
other
segments,
not
yet
treated,
do
not.
Thus
during
application,
the
field
is
not
a
uniform
emitting
source.
There
is
a
need
for
further
discussion
about
how
this
might
affect
or
skew
the
flux
term
calculation
during
the
application
period.
An
example
is
needed
showing
how
PERFUM
responds
(
i.
e.,
how
sensitive
it
is)
to
a
temporally
and
spatially
varying
area
source.

The
Panel
thought
that
representing
emission
flux
as
a
probability
distribution
to
address
uncertainty
and
real
variation
is
a
worthy
objective.
The
characterization,
"
measurement
uncertainty,"
can
be
confusing
and
should
be
clarified.
There
should
be
a
distinction
between
uncertainty
arising
from
measurements,
uncertainty
from
the
limited
sampling,
uncertainty
from
measuring
emission
flux
from
back
calculations
by
the
dispersion
model,
and
uncertainty
from
the
actual
variability
of
emissions.
There
should
also
be
a
discussion
of
the
physical­
chemical
processes
of
emission
that
the
probability
distribution
represents.

(
3)
Development
of
a
Receptor
Grid
The
PERFUM
modeling
system
uses
a
traditional
square
grid
instead
of
a
radial
grid.
Although
this
is
easier
to
generate
using
GIS,
there
is
no
reason
why
an
irregular
grid
could
not
be
used
around
irregularly
shaped
fields.
The
Panel
was
assured
that
PERFUM
could
handle
irregular
shaped
source
fields;
an
example
should
be
included
in
the
documentation.
The
obvious
benefit
of
the
grid
structure
as
presented
is
that
when
the
wind
direction
is
into
a
corner,
17
of
43
the
traditional
radial
grid
has
fewer
points
available
for
interpolation.
The
resulting
receptor
points
are
concentrated
at
each
corner
and
the
side­
by­
side
distance
increases
for
each
receptor
ring
away
from
the
field.
This
doesn't
happen
with
the
receptor
points
associated
with
the
sides
of
the
field.
This
results
in
much
of
the
computation
occurring
in
close
proximity
to
the
field,
yet
the
threshold
boundaries
may
be
further
out.
A
more
adaptive
grid
approach
that
allows
course
grid
information
to
be
used
to
suggest
additional
grid
points
to
improve
boundary
estimates
should
be
examined.
A
nested
grid
might
significantly
improve
the
computational
efficiencies,
while
maintaining
equivalent
accuracy.

b)
Are
there
any
other
potential
critical
sources
of
data
or
methodologies
that
should
be
considered?

Response
Models
that
correlate
flux
with
vapor
pressure
and
surface
type,
(
e.
g.,
Woodrow
et
al.,
2001)
as
well
as
mechanistic
models,
(
e.
g.,
Jury
et
al.,
1983;
Baker
et
al.,
1996;
Wang
et
al.,
1997;
Yates
et
al.,
2002),
should
to
be
considered
and
discussed
in
terms
of
their
applicability
to
PERFUM.
There
are
a
wide
variety
of
environmental
factors
that
affect
emission
rates.
Several
Panel
members
suggested
that
soils­
based
mechanistic
models
(
such
as
CHAIN_
2D)
should
be
investigated
and,
potentially,
incorporated
into
PERFUM.
This
type
of
model
can
integrate
the
soil
moisture,
temperature,
organic
matter,
and
other
soils
factors
into
the
computation
of
fumigant
fate,
soil
transport
and
volatilization.
Although
the
authors
state
that
a
soils­
based
emission
model
could
not
be
directly
incorporated
into
the
PERFUM
system,
a
stand­
alone
model
could
be
used
to
obtain
period­
averaged
emission
rates.
The
flexibility
to
incorporate
period­
flux
measurements
from
other
sources
would
be
desirable.

The
Panel
suggested
that
meteorological
data
produced
by
meteorological
model
simulations
may
be
considered
for
use
in
PERFUM.
Such
data
sources
will
help
in
filling
data
gaps
for
areas
where
credible
meteorological
measurements
or
observations
are
not
available.
An
example
of
such
models
are
the
National
Center
for
Atmospheric
Research
(
NCAR),
Mesoscale
Model
version
5
(
MM5),
and
the
Colorado
State
Regional
Atmospheric
Modeling
System
(
RAMS).
Compared
to
the
integrity
of
meteorological
data
from
other
models,
such
as
the
wind
field
models,
MM5
and
RAMS
have
broader
interpretability.

The
meteorological
preprocessor
for
handling
FAWN
and
CIMIS
station
data
to
develop
ISCST3­
ready
input
and
the
emissions
preprocessor
should
be
made
available.

Question
3:
The
determination
of
appropriate
flux/
emission
rates
is
critical
to
the
proper
use
of
the
PERFUM
model
as
these
values
define
the
source
of
fumigants
in
the
air
that
can
lead
to
exposures.
Upon
its
review
of
how
flux
rates
can
be
calculated,
the
Agency
has
identified
a
number
of
questions
it
would
like
the
Panel
to
consider.
In
PERFUM,
flux
rates
were
treated
as
a
probabilistic
variable
with
an
uncertainty
developed
from
the
statistical
bounds
of
the
flux
calculation.
For
each
measurement
period
a
standard
error
is
generated
that
reflects
the
measurement
uncertainty
of
the
flux
rate.
PERFUM
then
perturbs
the
concentration
estimates
within
each
period
by
the
standard
error
using
Monte
Carlo
methods
to
simulate
the
uncertainty
in
the
flux
estimates.
18
of
43
a)
What,
if
any,
refinements
are
needed
for
this
process
including
the
manner
in
which
flux
values
were
calculated
for
each
monitoring
period
to
generate
the
standard
error
estimates?

Response
The
calculation
of
the
fumigant
emission
rate
using
the
"
indirect
method"
(
i.
e.,
where
observed
downwind
ambient
concentrations
following
an
application
are
used
with
ISCST3
to
back­
calculate
the
aerial
emission
rate)
is
appropriate
for
estimating
emissions
at
a
given
time,
for
a
given
application,
and
at
the
given
site.
Furthermore,
the
method
presented
in
the
PERFUM
documentation
for
fitting
the
emissions
rate
and
characterizing
its
uncertainty
is
statistically
sound
and
reasonable
as
a
first
approximation.

Uncertainty
in
emission
rates
is
generated
in
the
model
by
randomly
sampling
the
estimated
slope
(
and
inferred
value
of
E)
based
on
a
t­
distribution
(
now
a
normal
distribution)
with
a
standard
deviation
defined
by
the
standard
error
of
the
slope
estimate.
While
appropriate
in
concept,
the
method
is
susceptible
to
error
for
a
number
of
reasons.
Furthermore,
it
cannot
capture
the
predominant
uncertainty
that
is
present
when
it
is
applied
in
an
extrapolation
mode,
for
new
conditions
or
at
different
sites.

A
second
concern
expressed
by
Panel
members
is
the
way
that
the
method
is
implemented.
This
involves
the
random
selection
of
uncertain
emission
rates
in
sequential
time
periods.
Since
the
causes
of
variability
are
likely
to
persist
over
time,
sampling
the
uncertain
emission
rates
in
sequential
time
periods
independently
from
each
other
will
likely
underestimate
the
uncertainty
variance
of
the
daily
emission
rate,
and
also
underestimate
the
frequency
of
highend
emissions
and
associated
high­
end
buffer
zone
lengths.
A
method
is
needed
to
incorporate
the
temporal
persistence
of
the
uncertainty
in
the
flux
estimate.

Some
Panel
members
thought
that
improvements
could
be
achieved
by
utilizing
more
advanced
methods
such
as
mechanistic
approaches
that
incorporate
physical/
mass­
balance
constraints
and
consider
the
effects
of
chemical
properties
(
vapor
pressure,
solubility
and
soil
adsorption
coefficient),
soil
properties
(
porosity,
organic
matter
fraction,
and
tortuosity),
application
method,
and
atmospheric
conditions
(
wind
speed,
atmospheric
stability,
temperature,
and
pressure).
Such
a
procedure
could
be
used
with
confidence
at
the
tested
site
for
the
varying
(
e.
g.,
5­
year)
meteorological
conditions,
and
would
provide
a
better
basis
for
prediction
at
other
times
and
at
other
locations
with
different
soil,
environmental,
and
topographic
properties.

b)
How
appropriate
is
it
to
use
a
flux/
emission
factor
from
a
single
monitoring
study
(
or
small
number
of
studies)
and
apply
it
to
different
situations
such
as
for
the
same
crop
in
a
different
region
of
the
country?

Response
It
is
unlikely
that
generalizing
the
use
of
PERFUM
to
regional
or
state
scale
will
be
appropriate
if
the
emission
data
were
collected
at
a
single
(
or
relatively
few)
location(
s).
To
19
of
43
fully
capture
the
variability/
uncertainty
in
the
measurements,
model,
soil
and
environmental
conditions,
differences
in
meteorology,
and
seasonal
effects
will
require
the
availability
of
an
extensive
data
set.
This
is
not
a
failing
of
the
model,
rather,
the
data
supporting
the
model.
As
is
true
with
all
models,
the
quality
of
the
output
information
is
directly
related
to
the
quality
of
the
input
information.

The
developers
and
users
of
PERFUM
need
to
develop
a
more
robust
data
set
of
emissions
of
iodomethane
under
different
field/
environmental
conditions
(
e.
g.,
soil
type,
temperature,
atmospheric
and
soil
moisture,
application
type,
etc.).
Additional
information
will
increase
reliability
of
the
method.

c)
Please
comment
on
PERFUM's
capability
to
adequately
consider
multiple,
linked
application
events
as
well
as
single
source
scenarios.
Does
PERFUM
appropriately
address
situations
where
data
are
missing?

Response
A
simplified
example
of
the
effect
of
multiple
fields
was
presented
to
the
Panel.
It
is
not
clear
from
the
presentation
if
more
complex
field
patterns
could
be
evaluated
(
e.
g.,
arbitrary
arrangement,
varying
field
sizes,
varying
application
rates,
arbitrary
orientation
of
field
boundaries
with
respect
to
North
and
South).
In
reality,
multiple
sources
or
fumigated
fields
tend
to
cluster
around
the
receptor
areas
and
be
of
varying
size
and
source
strength
(
i.
e.,
flux
rates
may
differ).
The
developers
of
PERFUM
should
elaborate
on
the
capability
of
simulating
multiple
linked
sources,
especially
by
considering
several
combinations
of
spatial
distribution
of
the
fumigated
fields
versus
the
receptor
points.

PERFUM
authors
should
be
encouraged
to
develop
generalized
routines
applicable
to
emissions
from
multiple
fields.
ISCST3
and
CALPUFF
have
previously
been
shown
to
be
adaptable
to
this
situation
(
See
P.
S.
Honaganihalli
and
J.
N.
Seiber.
2000).
PERFUM
should
be
applicable
as
well,
and
thus
should
be
adapted
to
this
fairly
realistic
scenario
of
emissions
contributions
from
more
than
one
field.

d)
In
the
back­
calculation
approach
used
for
estimating
emission
rates,
the
regression
of
measured
versus
modeled
values
can
be
forced
through
the
origin
or
not.
Which
approach
does
the
Panel
prefer
and
what
are
the
implications
of
each
approach?

Response
One
of
the
issues
raised
by
the
Panel
involves
the
way
that
the
intercept
is
treated
in
the
analysis
of
the
flux
using
the
indirect
method.
In
the
regression
model,
the
intercept
has
a
physical
meaning
 
it
corresponds
to
a
"
background"
concentration
at
the
receptor
(
i.
e.,
if
emissions
are
0).
A
key
issue
is
the
assessment
of
whether
there
is
a
reason
to
believe
that
a
background
concentration
could
or
should
be
present
at
the
site,
so
that
a
non­
zero
intercept
does
(
or
does
not)
make
sense.
This
issue
has
not
been
addressed
in
an
adequate
manner
in
the
PERFUM
documentation.
20
of
43
Modeled
concentrations
that
are
zero
at
locations
where
measured
concentrations
are
non­
zero
give
rise
to
the
non­
zero
intercept.
When
the
background
concentration
is
known
to
be
zero,
such
a
situation
indicates
that
the
real
plume
extends
outside
the
predicted
plume
behavior.
This
behavior
is
likely
caused
by
inadequately
predicting
the
plume
direction
or
the
dispersion
coefficients.

 
Since
atmospheric
stability
is
used
to
predict
dispersion
coefficients,
and
atmospheric
stability
is
expressed
as
a
category
number,
the
predicted
dispersion
coefficients
may
be
in
error
because
true
dispersion
coefficients
behave
in
more
of
a
continuous
fashion.

 
Furthermore,
atmospheric
stability
may
be
improperly
inferred
from
measurements
especially
around
sunset
or
sunrise
when
stability
is
rapidly
changing.
Periods
around
sunset
are
of
most
concern
because
the
transition
to
stable
conditions
will
cause
nearsurface
concentrations
to
be
largest.

 
Regardless
of
the
cause
of
the
discrepancy,
there
is
a
flux
of
material
in
the
real
plume
unaccounted
for
by
the
modeled
plume
even
if
the
modeled
concentrations
are
fit
to
the
measured
concentrations
within
the
modeled
plume
boundary.
The
flux
estimates
so
determined
will
underestimate
the
actual
values.
Put
another
way,
having
a
non­
zero
intercept
indicates
that
the
flux
will
be
underestimated
because
there
is
a
flux
of
material
in
the
real
plume
that
cannot
be
captured
in
the
modeled
plume.

 
In
the
iodomethane
case
study,
sensors
were
located
fairly
close
to
the
source
at
a
uniform
height.
At
distances
close
to
the
source,
it
is
important
to
check
the
vertical
concentration
distribution
to
ensure
that
larger
concentrations
are
not
present
at
lower
elevations.

A
second
issue
addressed
as
part
of
the
model
development
is
whether
the
data
should
be
log­
transformed
prior
to
the
regression.
Log­
transformation
could
be
motivated
by
the
positively­
skewed,
non­
negative
nature
of
measured
air
concentrations,
which
tend
to
be
lognormally
distributed.
One
Panel
member
stated
that
the
log­
transform/
linear
regression
approach
is
not
recommended
since
it
violates
the
linear,
mass­
balance
assumptions.

PERFUM
accounts
for
variability
in
the
period
flux
measurement
by
calculating
the
standard
error
between
the
measured
and
model
concentrations.
This
gives
a
measurement
of
the
error/
variability
for
only
that
field
study.
There
are
a
number
of
considerations
to
this
approach:

 
Total
emissions
tend
to
be
more
accurately
measured
compared
to
short­
interval
emissions.
The
range
of
total
emissions
from
several
methyl
bromide
experiments
is
high
with
a
mean
total
emission
of
about
50%
and
a
standard
deviation
of
approximately
35%
(
see
Slide
1).
The
bounding
lognormal
distributions
shown
in
this
slide
demonstrate
the
variability
among
these
studies.
To
capture
other
sources
of
variability
(
e.
g.,
variations
in
local
conditions
and
seasons),
it
might
be
better
to
sample
from
a
distribution
similar
to
the
one
shown
in
Slide
1.
A
beta
distribution
offers
a
potentially
useful
parametric
form
for
the
distribution
of
fraction
emitted,
since
it
is
bounded
by
zero
and
one.
21
of
43
 
The
error
bars
on
the
results
from
two
experiments
shown
in
Slide
1
suggest
that
intrastudy
variability
would
be
smaller
than
the
variation
between
studies.

 
Slide
2
shows
the
period
emission
(
2
or
4
hours)
for
8
days
after
application
of
methyl
bromide.
Three
methods
were
used
to
obtain
emission
values:
aerodynamic,
theoretical
profile
shape,
and
the
instantaneous
horizontal
flux
methods.
Slide
2
shows
that
there
is
large
variation
in
period
emissions
between
these
approaches
to
calculate
emission
rates,
even
though
the
same
experimental
data
(
i.
e.,
air
concentration,
wind
speed,
temperature)
are
used
by
each
approach.
Also,
note
that
the
three
methods
give
nearly
equivalent
cumulative
emissions.

 
The
use
of
flux
chambers
also
shows
large
differences
in
the
period
emission
values
compared
to
the
other
flux
estimation
approaches
(
Slide
3).
These
data
suggest
a
need
to
include
regional
variability
into
a
PERFUM
assessment.

 
For
flat­
fume
applications,
the
plastic
film
can
be
a
dominant
factor
controlling
emissions
and
the
permeability
depends
on
fumigant
chemical
and
ambient
temperature.
Slide
4
shows
that
high­
density
polyethylene
(
HDPE)
is
more
permeable
to
iodomethane
than
methyl
bromide.
This
suggests
that,
everything
else
held
constant,
emissions
of
iodomethane
through
HDPE
will
be
greater
than
methyl
bromide.

 
Slides
5­
7
show
that
a
numerical
model
may
provide
a
good
description
of
the
volatilization
process
when
atmospheric
data
are
available.
The
numerical
model
describes
water,
heat
and
chemical
transport
following
the
equations
shown
in
Slide
5.
Slide
6
shows
two
volatilization
boundary
conditions
that
were
used.
The
mass
transfer
boundary
condition
used
in
CHAIN­
2D
is
coupled
to
processes
occurring
in
the
atmosphere
and
provides
an
overly
simplified
flux
history
(
see
Slide
3,
solid
lines).
Even
adding
temperature
dependence
to
these
boundary
conditions
is
insufficient
in
characterizing
the
variation
in
the
flux
rate
(
Slide
3).
Utilizing
a
mass
transfer
boundary
condition
(
SOLUTE.
EXE)
that
is
directly
coupled
to
atmospheric
processes
(
Baker
et
al.,
1996),
(
i.
e.,
wind
and
stability),
provides
a
much
more
accurate
flux
history
(
see
Slide
7).
Also,
the
variation
in
the
predicted
flux
more
closely
matches
the
variability
in
the
measured
values.
A
model
like
this
offers
the
potential
to
determine
the
period
flux
values
at
a
new
location
(
or
fumigation
methodology,
etc.)
easily
and
less
expensively.
This
might
be
one
way
to
obtain
input
data
and
avoid
expensive,
time­
consuming,
and
possibly
cost­
prohibitive
studies.
22
of
43
Slide
1
Slide
2
Slide
3
Slide
4
Slide
5
Slide
6
23
of
43
Slide
7
Question
4:
The
integration
of
actual
time­
base
meteorological
data
into
ISCST3
is
one
of
the
key
components
that
separate
the
PERFUM
methodology
from
that
being
employed
by
the
Agency
in
its
current
assessment.
There
are
several
potential
sources
of
these
data
including
the
National
Weather
Service,
Federal
Aviation
Administration,
California
Irrigation
Management
Information
System
(
CIMIS),
and
the
Florida
Automated
Weather
Network
(
FAWN).
The
Agency
is
also
aware
that
there
are
several
approaches
that
can
be
used
to
process
meteorological
data,
and
acknowledges
that
PERFUM
used
PCRAMMET,
which
is
a
standard
Agency
tool
for
this
purpose,
as
well
as
other
techniques
in
some
cases
(
e.
g.,
for
the
FAWN
&
CIMIS
data).
Various
datasets
from
both
California
and
Florida
were
used
as
the
basis
for
the
PERFUM
case
study.

a)
Please
comment
on
the
methods
used
to
select
monitoring
station
locations.

Response
Terrain
effects
(
topography)
need
to
be
addressed
when
selecting
station
locations.

The
Panel
recommended
a
review
of
state
climatological
agricultural
weather
stations
{
see
examples
in
response
4
(
c)}
for
potential
sources
of
rural
meteorological
data.
In
the
current
study
CIMIS
data
from
California
may
meet
these
criteria,
but
FAWN
data
in
Florida
are
not
very
reliable.

b)
What
criteria
should
be
used
to
identify
meteorological
regions
for
analysis
and
how
should
specific
monitoring
data
be
selected
from
within
each
region?

Response
24
of
43
When
selecting
monitoring
stations
we
should
consider
inland
as
well
as
coastal
regions.
We
should
also
consider
areas
with
complex
terrain.
Those
locations
usually
have
different
meteorological
conditions.
For
example,
the
diurnal
wind
cycle
in
coastal
and
near
water
areas
exhibits
an
opposite
wind
direction
during
day
and
night
associated
with
the
land
and
sea
breeze
phenomena
triggered
by
the
gradient
in
temperature
between
land
and
sea.

c)
What
criteria
should
be
used
to
identify
meteorological
regions
for
analysis?

Response
Precipitation
(
linked
to
cloud
cover).

Precipitation
not
only
contributes
to
the
removal
of
pollutants,
but
also
to
moisturizing
the
soil,
which
might
be
an
issue
in
post
precipitation
application.

Temperature
field.

The
flux
rate
of
emissions
is
affected
by
ambient
and
land
(
skin)
temperature.
Temperature
profiles
at
several
regions
in
the
country
show
a
temperature
inversion
during
the
early
morning
hours.
These
inversion
events
are
associated
with
very
stable
atmospheric
conditions
and
the
occurrence
of
early
morning
fog.
Relative
humidity
is
typically
very
high
during
these
morning
hours
until
the
lifting
of
the
inversion
by
the
daylight
heating.

Locality
(
terrain,
physical
characteristics
of
the
land
surface,
etc.).

Terrain
and
topography
may
affect
the
dispersion
of
the
plume,
both
horizontal
and
vertical.
The
current
structure
of
the
ISCST3
model
does
not
handle
terrain
very
well.
The
AERMOD
approach
seems
to
be
more
adequate.
It
is
the
understanding
of
the
Panel
that
the
Agency
is
moving
toward
the
adoption
of
AERMOD
as
an
approved
regulatory
model.
This
could
then
be
integrated
into
PERFUM.

The
primary
source
of
meteorological
data
should
be
in
the
following
order:
(
1)
National
Weather
Service;
(
2)
ASOS
(
FAA)
data
(
ASOS
came
on
line
in
early
1990,
a
very
rich
data
source
that
covers
the
country);
(
3)
State
climatology
or
agricultural
weather
stations
(
e.
g.,
CIMIS,
FAWN,
SCAN,
ECONet,
etc.);
(
4)
meteorological
data
selected
from
regulatory
agency
sites
or
industrial
sites
if
the
data
are
quality
assured.

The
closest
quality­
assured
data
source
should
be
used
for
model
input.
The
data
source
should
be
well
documented.

d)
Please
comment
on
the
manner
that
data
from
the
selected
various
stations
were
processed.

Response
25
of
43
Processing
the
data
followed
the
normal
procedure.
Missing
data
however
is
the
tricky
part,
especially
for
cloud
cover,
which
is
one
of
the
parameters
used
by
the
ISCST3
model
to
designate
a
stability
index
(
letter).
When
filling
gaps
of
missing
data,
the
EPA
recommendations
are
to
use
data
within
a
few
hours
of
the
missing
period.
This
might
be
a
reasonable
approach
for
parameters
such
as
temperature.
However
a
highly
variable
and
uncertain
parameter
such
as
cloud
cover
will
be
much
less
accurate
when
using
the
interpolation
approach.

As
was
reported,
the
NWS
data
seem
to
have
the
highest
accuracy,
with
quality
control
and
quality
assurance
procedures
being
implemented.
Data
from
the
three
other
data
sources
seem
to
be
less
accurate;
especially
the
Florida
database.
The
lack
of
quality
control
in
the
Florida
system
could
increase
uncertainty
in
the
system.
Also
while
the
NWS
has
cloud
cover
data
as
one
of
its
measured
parameters,
observational
networks
from
Florida
and
California
do
not
include
this
parameter.
The
Turner
method
for
calculating
stability
requires
information
on
the
cloud
cover.
An
alternative
method
was
used
to
calculate
the
stability
parameter.
That
diminishes
consistency
when
comparing
the
results.

The
Panel
suggested
the
use
of
other
meteorological
data
sources
for
appropriate
regions
in
the
country.
Many
state
climate
offices
have
good
collections
of
weather
data.
For
example,
the
State
Climate
Office
of
North
Carolina
maintains
the
CRONOS/
ECONet
database.
The
ECONet
network
is
a
part
of
the
CRONOS
Database.
Data
from
216
stations
are
archived
in
this
database
for
retrieval.
The
primary
focus
is
on
North
Carolina
stations,
but
data
are
also
archived
from
stations
in
portions
of
South
Carolina,
Georgia,
Virginia,
and
Tennessee.
The
FAA,
NWS,
or
other
government
agencies
maintain
most
stations.
A
total
of
24
stations
(
ECONet)
are
maintained
directly
by
the
State
Climate
Office
of
NC.

Soil
Climate
Analysis
Network
(
SCAN)
stations
focus
on
the
agricultural
areas
of
the
United
States.
Maintained
by
the
National
Resources
Conservation
Service,
the
SCAN
sites
are
used
to
monitor
drought
development,
for
soil
classification
and
moisture
assessment,
for
input
into
global
circulation
models,
and
for
various
water
table
assessments
that
are
important
to
local
crops,
woodlands,
and
wetlands.
There
are
6
SCAN
stations
in
the
NC
CRONOS
Database.
Portable
weather
stations
can
be
deployed
at
the
SCAN
site
or
at
specific
locations
within
the
SCAN
network
for
a
particular
case
study.

Other
sources
of
data
include
the
AmeriFlux
network
(
http://
public.
ornl.
gov/
ameriflux)
established
in
1996
to
provide
continuous
observations
of
surface
and
weather
parameters,
and
the
Peanut/
Cotton
Infonet
in
southeastern
Virginia.

The
Panel
also
suggested
the
use
of
meteorological
simulation
modeling
as
another
source
of
meteorological
information.
Models
such
as
the
Regional
Atmospheric
Modeling
System
(
RAMS)
of
Colorado
State
University
and
the
NCAR
Mesoscale
Model
Version
5
(
MM5)
have
been
run
on
a
fine
resolution
(
4km
or
less)
over
the
continental
US
for
a
period
of
a
year
or
more.
Information
based
on
those
models
can
be
used
to
supplement
data
when
measurements
are
not
available.
Data
from
these
models
could
be
used
for
scanning,
screening
and
comparisons
among
areas
to
determine
where
different
site
data
might
be
needed
for
PERFUM.
26
of
43
Having
mentioned
various
meteorological
data
sources,
the
main
point
is
to
identify
the
biases
and
errors
in
each
dataset
in
order
to
truly
understand
its
limitations.

The
Panel
noted
that
different
types
of
data
are
needed
if
PERFUM
is
used
in
developing
national
and/
or
regional
buffer
zone
strategies
compared
to
the
use
of
PERFUM
in
helping
with
decision­
making
at
the
local
level;
for
example
in
decisions
regarding
permitting
of
a
proposed
iodomethane
application
in
the
vicinity
of
subdivisions,
schools,
etc.
If
PERFUM
is
used
in
the
latter
situation,
site­
specific
meteorological
data
should
be
collected
that
are
as
close
as
possible
to
the
proposed
application
site
and
taken
near
the
time
of
application.
Meteorological
data
collected
over
a
few
months,
or
even
weeks
or
days
that
are
proximate
to
the
proposed
application
might
be
more
relevant
in
this
situation
than
5­
year
historical
data.

e)
Data
quality
and
uncertainty
associated
with
these
data
vary
with
the
source.
Does
the
Panel
agree
with
the
approaches
used
to
characterize
these
factors?

Response
Generally,
the
National
Weather
Service
data
are
preferred
since
cloud
cover
is
identified
and
thus
Turner
Stability
Class
can
be
determined.
In
addition,
the
data
are
collected
at
10
meters
above
ground
level.

f)
Anemometer
sampling
height
has
been
identified
as
a
concern
by
the
Agency
in
preparation
for
this
meeting.
What
are
the
potential
impacts
of
using
data
collected
with
different
anemometer
heights
in
an
analysis
of
this
nature?

Response
Some
of
the
observational
data
used
in
PERFUM
are
at
10
meters
heights;
the
NWS
measurements
are
typically
between
6
and
10
meters.
Wind
data
from
California
are
at
2
meters.
What
are
the
differences?
Are
they
important?
The
surface
layer
is
defined
as
the
lowest
10%
of
the
boundary
layer
height,
typically
the
lowest
100
meters
above
the
surface.
In
general
there
are
differences
between
the
wind
speeds
and
directions
measured
at
different
heights.
The
magnitudes
of
those
differences
depend
on
the
stability
conditions
of
the
atmosphere.
For
example,
in
the
daytime
over
land
during
clear
weather
conditions,
vertical
wind
profiles
will
have
less
speed
and
wind
direction
changes
than
during
stable
night
time
conditions.
Plume
centerlines
would
vary
with
data
collected
at
2
meters
as
opposed
to
10
meters.
Winds
are
more
variable
at
lower
heights
above
ground
level,
which
impacts
stability
class
determination
in
the
ISCST3
model.
Surface
roughness
factors
(
roughness
length)
must
be
considered
for
data
collected
at
2
meters
elevation.
The
roughness
length
varies
widely
depending
on
the
physical
characteristics
of
the
surface.
The
friction
velocity
sets
the
level
of
the
velocity
fluctuations
in
the
surface
layer.
In
describing
the
relationship
between
the
2
meters
and
10
meters
wind
measurements,
a
logarithmic
wind
profile
might
be
considered
with
the
following
cautionary
remark.

At
2
meters
the
air
could
be
more
stable
with
lower
wind
speeds
leading
to
underestimation
if
the
10
meters
data
would
be
used.
In
some
areas
the
differences
between
2
27
of
43
and
10
meters
could
be
quite
large.
There
is
a
great
source
of
uncertainty
when
adjusting
from
2
to
10
meters
using
the
power
formula
for
sites
where
both
heights
are
measured.
Note
that
the
ISCST3
model
does
not
make
adjustments
below
10
meters.
Uncertainty
in
the
wind
speed
measurement
below
10
meters
is
greater
than
measurements
taken
at
10
meters.
Wind
direction
is
more
variable
at
lower
heights,
increasing
uncertainty,
another
reason
to
try
to
use
the
NWS
data
when
possible.

In
summary,
there
are
differences
between
2
meters
and
10
meters
wind
speed
and
wind
direction
but
it
might
not
be
easily
addressed.
It
will
be
better
to
standardize
on
one
height
(
in
the
case
of
ISCST3).
Or,
even
better,
to
have
multiple
height
measurements,
at
the
same
time
and
location,
to
more
accurately
account
for
turbulence
(
using
AERMOD).

g)
Does
PERFUM
treat
stability
class
inputs
appropriately?

Response
Apparently
PERFUM
treats
stability
class
inputs
appropriately.
There
is
not
much
difference
between
the
major
ways
to
estimate
stability
class
via
the
Turner,
Sigma
Theta,
and
Delta
Temperature/
Solar
Radiation
methods.
However,
stability
class
in
ISCST3
is
fixed
during
the
one­
hour
simulation
and
cannot
be
altered.

h)
Does
PERFUM
appropriately
calculate
boundary
air
concentration
estimates
by
concurrently
using
upper­
bound
meteorological
and
emission/
flux
inputs?

Response
The
PERFUM
model
apparently
incorporates
"
worst­
case"
emission
flux
with
five
years
of
meteorological
data
where
poor
dispersion
hours
could
coincide
with
"
worst­
case"
emission
flux
hours.

The
horizontal
dispersion,
Sigma­
Y,
and
vertical
dispersion
coefficient,
Sigma­
Z,
need
to
be
examined
probabilistically.
Uncertainty
related
to
horizontal
and
vertical
dispersion
may
be
included
in
model
formulation
by
randomly
selecting
a
multiplier
to
Sigma­
Y
and
Sigma­
Z
in
the
ISCST3
model
code.
The
multiplier
can
be
introduced
on
an
hourly
or
daily
basis.
This
multiplier
is
based
on
the
use
of
a
cumulative
distribution
function
(
CDF)
to
represent
the
uncertainty
(
based
on
the
difference
between
measurements
at
field
experiments
and
the
model
calculated
Sigmas).

There
is
a
need
to
keep
variability
and
uncertainty
separate
in
the
runs
of
PERFUM.
The
5
years
of
climate
data
are
used
to
examine
the
effect
of
variability
on
the
results.
The
changes
to
the
flux
rate
(
question
3)
address
uncertainty.
Typically
in
a
Monte
Carlo
analysis,
one
set
of
flux
rates
would
be
run
for
the
full
5
years
and
the
results
stored,
then
a
new
set
of
flux
rates
for
each
period
generated
and
the
5
years
re­
run.
The
runs
associated
with
uncertainty
are
there
to
add
confidence
intervals
to
the
probability
statements.
The
current
PERFUM
runs
confound
the
uncertainty
and
variability,
something
that
will
have
to
be
decoupled
when
the
model
is
used
to
develop
management
tables.
There
are
other
"
parameters"
that
have
associated
uncertainty
that
28
of
43
would
have
to
be
looked
at
in
the
2D
Monte
Carlo
format
as
well.
There
are
a
number
of
agencies
that
tried
to
incorporate
variability
and
uncertainty
into
the
ISCST3
model.

Question
5:
The
Agency
model,
ISCST3
is
the
basis
for
the
PERFUM
approach.
This
model
has
been
peer
reviewed
and
is
commonly
used
for
regulatory
purposes
by
the
Agency.
PERFUM
also
uses
other
Agency
systems
such
as
PCRAMMET.

a)
Please
recommend
any
parameters
that
should
be
altered
to
optimize
the
manner
that
they
are
used
in
PERFUM.

Response
The
Panel
did
not
have
any
specific
recommendations
with
respect
to
PCRAMMET.
The
Panel
recommended
that
meteorological
pre­
processors
developed
within
the
PERFUM
project
for
other
meteorological
datasets
besides
NWS
stations
(
ASOS,
CIMIS,
FAWN,
etc.)
should
be
made
available.

It
was
noted
that
precipitation
during
the
summer
in
Florida
could
wash
the
material
out
of
the
atmosphere.
Inclusion
of
precipitation
would
require
inclusion
of
a
separate
database,
which
is
not
very
complete.
Special
consideration
of
precipitation
may
not
be
warranted
given
that
fumigant
application
is
not
planned/
recommended
if
precipitation
is
forecast.
This
clarification
led
to
a
discussion
on
sorting,
and
the
idea
of
eliminating
days
for
which
precipitation
occurs
was
discussed.
Again,
the
database
is
limited
and
the
added
value
of
this
sorting
was
not
obvious.
Decreases
in
barometric
pressure
(
potentially
enhancing
the
fumigant
flux)
were
also
discussed,
but
its
potential
significance
was
debatable.

Winter
versus
summer
differences,
for
example,
could
be
explored
with
real
data
from
the
same
sites.
Related
Panelist
experience
doing
this
at
the
same
site
showed
very
different
conditions
in
soil
climate.
The
use
of
flux
rates
from
one
season
for
the
whole
year
introduces
uncertainty.

b)
Does
the
Panel
agree
with
the
manner
in
which
the
receptor
grid
was
developed,
and
if
not,
please
provide
suggestions
for
improving
this
approach?

Response
The
Panel
agreed
with
the
development
of
the
receptor
grid
for
the
example
of
the
square
field,
but
also
supported
any
effort
that
would
increase
computational
efficiency.
In
addition,
the
Panel
recommended
consideration
of
rectangular
fields
and
other
irregularly
shaped
fields
 
both
from
a
gridding
consideration
and
for
better
representation
of
these
field
geometries.

c)
ISCST3,
as
integrated
into
PERFUM,
was
run
assuming
rural,
flat
terrain
which
would
be
typical
of
treated
farm
fields
but
might
not
be
typical
of
surrounding
residential
areas.
Does
the
Panel
concur
with
this
approach?
What
are
the
implications
of
such
an
approach?
What
improvements
can
be
made
to
this
approach?
29
of
43
Response
The
Panel
agreed
that
within
the
ISCST3
framework,
the
selection
of
rural
mode
would
yield
the
highest
predicted
downwind
concentrations
for
this
source
type.
However,

 
Outside
the
ISCST3
model
framework,
it
was
noted
that
the
fumigant
air
plume
from
this
type
of
ground
level
emission
source,
in
the
event
that
it
encounters
a
dense
obstacle
field
(
subdivision
housing
was
specifically
mentioned),
may
have
reduced
horizontal
flow
and
reduced
dispersion
within
the
obstacle
field
even
though
turbulence
is
enhanced
above
the
obstacle
field.

 
Wind
breaking
through
the
use
of
a
row
of
trees
or
high
bushes
was
discussed,
particularly
for
eastern
locations.
This
would
increase
the
atmospheric
turbulence
and
reduce
concentrations
beyond
the
break
in
wind.
Dense
obstacle
field
effects,
if
they
were
realized
due
to
the
windbreak,
would
be
contained
within
an
area
under
owner
control
to
exclude
exposure.

 
A
somewhat
similar
comment
was
made
regarding
atmospheric
turbulence
introduced
due
to
tall
crops
like
corn
in
neighboring
fields.

d)
ISCST3,
as
integrated
into
PERFUM,
was
run
in
regulatory
mode
which
includes
the
use
of
the
"
calms"
processing
routine.
Does
the
Panel
concur
with
this
approach?
If
not,
please
suggest
a
suitable
alternative.

Response
The
Panel
supported
the
use
of
calms
processing
as
being
consistent
with
traditional
regulatory
practice.
The
calm
periods
that
have
been
captured
in
field
studies
used
in
the
calibration
of
ISCST3,
as
well
as
calm
periods
captured
in
the
fumigant
field
studies,
if
any,
should
be
largely
self­
correcting
for
the
limitations
of
the
Gaussian
formulation
that
is
inherent
in
ISCST3.
The
Panel
supported
consideration
of
additional
field
studies,
which
should
further
improve
this
situation.
However,
not
all
Panel
members
were
confident
in
the
degree
of
selfcorrection
Some
Panel
members
expressed
concern
for
calm
periods
not
being
fully
captured
in
the
calibration
of
ISCST3,
thus
an
under­
prediction
of
the
buffer
zone
was
implied.
Also,
some
Panel
members
expressed
concern
for
how
calm
periods
affect
the
concentration
data
collected
during
the
fumigant
field
studies
and
the
implication
this
has
for
emission
flux
estimates.
Issues
on
the
back­
calculation
method
of
the
fumigant
flux,
when
calm
periods
were
encountered
in
the
field
studies,
somewhat
overlap
with
similar
issues
on
the
intercept
(
and
the
cluster
of
data
near
the
intercept)
during
regression
analysis.

Critical
Element
3:
Results
Question
6:
Soil
fumigants
can
be
used
in
different
regions
of
the
country
under
different
conditions
and
they
can
be
applied
with
a
variety
of
equipment.
30
of
43
a)
Please
comment
on
whether
the
methodologies
in
PERFUM
can
be
applied
generically
in
order
to
assess
a
wide
variety
of
fumigant
uses.

Response
PERFUM
and
its
associated
components
(
back­
calculation
of
flux,
ISCST3
model,
and
MOE
calculation)
appeared
to
the
Panel
to
be
generally
applicable,
and
could
be
applied
(
with
minimal
adaptation/
modification)
to
most
growing
regions
in
the
U.
S.

PERFUM
could
probably
be
applied
generically
to
evaluate
other
fumigants
in
other
regions
but
in
its
present
configuration
is
probably
best
for
highly
volatile
fumigants
with
high
initial
emission
from
the
soil.
To
use
the
model
in
other
areas
it
certainly
will
be
essential
to
use
regional
or
local
weather
data
as
close
to
the
area
of
concern
as
possible.

Calibration
and/
or
validation
studies
will
need
to
be
done
to
fit
regions
and
sites,
in
other
words,
applicability
will
need
to
be
demonstrated
when
moving
from
one
cropping
system
to
another,
or
from
one
growing
region
to
another.
Such
studies
should
be
designed
to
address
variables
that
might
influence
flux
rates,
downwind
residue
fate,
and
unusual
atmospheric
conditions.
Making
the
States
aware
of
the
need
for
certain
types
of
data
will
likely
bring
about
improvement
in
some
of
the
data
development.
Data
are
frequently
available
today
through
local
weather
networks,
e.
g.,
Peanut/
Cotton
Infonet
in
southeastern
Virginia.

Since
PERFUM
was
developed
based
on
the
methodologies
of
the
California
DPR
model
for
methyl
bromide,
PERFUM
should
accurately
predict
methyl
bromide
exposure.
But
the
three
liquid
fumigants
1,3­
D,
chloropicrin,
and
MITC
have
much
lower
vapor
pressure
than
methyl
bromide
and
the
case
study
chemical,
iodomethane.
Therefore,
validation
should
be
conducted
for
these
liquid
fumigants.

b)
What
considerations
with
regard
to
data
needs
and
model
inputs
should
be
considered
for
such
an
effort?

Response
A
statement
was
made
in
the
PERFUM
report
("
A
Probabilistic
Exposure
and
Risk
Model
for
Fumigant
Bystander
Exposures
Using
Iodomethane
as
a
Case
Study")
that
the
flux
rate
is
likely
dependent
on
factors
such
as
soil
type,
soil
temperature,
organic
matter
content,
etc.
Yet
it
is
noted
that
effects
of
each
of
these
factors
have
not
been
quantified
for
fumigants
at
field
scale
and
that
it
would
be
difficult
to
do
so.
A
robust
model
should
include
factors
or
variables
that
are
known
to
be
important
in
the
fate
and
transport
of
fumigants.
For
example,
the
following
are
considered
to
be
important:

 
Temperature
(
Air
and
Soil).
Air
temperature
can
affect
atmospheric
stability
terms,
and
(
along
with
atmospheric
moisture
 
humidity
or
fog)
such
things
as
sampling
efficiency
of
iodomethane
through
charcoal
adsorbent.
Soil
temperature
can
influence
flux
rates
and
soil
degradation
rates,
including
microbial
degradation.
Microbial
degradation
needs
31
of
43
to
be
considered
for
iodomethane,
particularly
in
fields
that
are
treated
more
than
once
(
microbial
adaptation
and
enhanced
degradation).

 
Water
evaporation
rate.
In
future
studies,
consideration
should
be
made
of
compiling
and
potentially
correlating
water
evaporation
rate
(
pan
evaporation)
data.
There
may
be
a
correlation
between
flux
of
iodomethane,
other
fumigants,
and
water
flux.

 
Atmospheric
Moisture.
Rain
and
fog
should
be
taken
into
account
in
terms
of
their
influence
on
PERFUM
modeling.
Rain
can
reduce
flux
if
rainwater
accumulates
on
the
surface
of
a
tarp
or
at
the
soil
surface
by
plugging
the
soil
pores
and
reducing
vapor
phase
diffusion.
Later,
rainwater
can
increase
flux
as
moisture
competes
for
adsorptive
sites
on
or
in
soil,
and
by
providing
a
mass
transport
mechanism
to
the
surface
under
evaporative
conditions.
Rain
can
also
wash
out
downwind
residues
before
they
reach
receptors.
The
effect
of
fog
on
fumigant
airborne
residues,
if
any,
will
need
to
be
determined.

 
Obstructions
to
airflow.
Trees
planted
as
windbreaks,
as
ornamentals
or
in
forested
wooded
areas
can
affect
surface
roughness
and
could
potentially
subtract
residues
from
the
air
by
absorptive
deposition.
Downwind
crop
cover,
corn
for
example,
might
similarly
affect
conditions
around
a
field
subject
to
fumigation,
or
downwind
residue
content.

 
Application
variables.
Type
of
irrigation
(
sprinkler,
drip,
and
flood),
use
or
not
of
tarps,
types
of
tarps,
and
depth
and
methods
of
injection,
might
differ
between
crops
and
growing
regions.

 
Efficacy
considerations
and
using
the
model
in
a
predictive
mode.
It
seems
likely
that
if
a
model
incorporates
soil
parameters
as
well
as
characteristics
of
the
fumigant,
then
the
proper
conditions
for
safest
and
most
effective
application
might
be
ascertained.

 
Airshed
considerations.
Ambient
residue
levels
may
be
of
increasing
concern
in
airsheds
where
iodomethane
might
be
used
frequently.
PERFUM
could
potentially
be
adapted
to
help
address
this
by
extending
its
use
to
longer
distances
from
the
treated
field,
and
to
situations
where
more
than
one
field
is
applied
simultaneously
or
in
close
proximity.

 
Other
considerations.
The
range
of
flux
values
might
be
different
in
differing
soil
types
and
with
differing
application
methods
used
in
various
growing
regions.
Timing
of
applications
could
differ.
Topography
will
differ,
affecting
the
use
of
the
ISCST3
dispersion
model.

It
is
noted
that
the
model
adequately
considers
the
atmospheric
stability
and
computes
the
buffer
zones.
The
fact
that
it
does
not
consider
soil
type,
soil
moisture,
organic
matter,
etc.,
may
be
of
little
consequence
with
other
fumigants
as
volatile
as
iodomethane.
However,
these
factors
could
be
significant
with
other
fumigants,
especially
in
the
initial
fumigation
process.
These
parameters
might
affect
the
efflux
of
fumigant
gases
from
the
soil.
Without
some
way
of
accounting
for
these
'
uncertain'
factors,
errors
might
occur
in
calculating
the
buffer
zones.
These
32
of
43
factors
should
be
investigated
and
incorporated,
if
possible,
into
the
model
to
improve
its
usefulness
for
other
fumigants
in
other
environments.

There
are
significant
uncertainties
associated
with
PERFUM
because
soil
factors
are
not
taken
into
account
in
the
model.
Some
of
these
factors
could
affect
the
rate
of
flux
of
some
fumigants.
Another
factor
is
windrows,
tree
barriers,
etc.,
a
common
occurrence
in
many
areas
in
the
southeastern
U.
S.
A
Panel
member
mentioned
that
such
barriers
would
increase
the
turbulence
and
could
affect
the
size
of
buffer
zones.
Rain
is
also
a
factor
that
could
mitigate
downwind
atmospheric
concentrations.

A
Panel
member
mentioned
that
a
model
is
being
developed
that
will
address
some
of
these
effects.
It
seems
likely
that
if
a
model
incorporates
soil
parameters
as
well
as
characteristics
of
the
fumigant,
then
the
proper
conditions
for
the
safest
and
most
effective
application
might
be
ascertained.
This
could
work
positively
toward
reducing
the
dose,
increasing
the
efficacy,
and
decreasing
the
atmospheric
concentration
at
any
given
point
in
time.
This
could
work
favorably
toward
decreasing
calculated
buffer
zones.
It
seems
that
the
use
of
large
buffer
zones
would
increase
the
safety
of
bystander
exposure
but
does
suggest
some
problems
with
marking
and
enforcing
those
zones.
Such
zones
might
preclude
many
areas
from
agricultural
use
because
of
the
proximity
of
houses,
etc.
While
having
buffer
zones
to
ensure
safety
of
people
certainly
seems
wise,
it
appears
that
anything
that
can
be
done
to
minimize
such
zones
would
be
practical.

Question
7:

a)
Please
comment
on
whether
PERFUM
adequately
identifies
and
quantifies
airborne
concentrations
of
soil
fumigants
that
have
migrated
from
treated
fields
to
sensitive
receptors.

Response
The
ISCST3
model
has
been
tested/
validated
and
is
therefore
considered
to
be
as
accurate
as
other
EPA­
approved
models
subject
to
the
limitations
of
its
design.
The
application
of
the
ISCST3
model
to
compute
ambient
concentrations
is
appropriate
for
regulatory
purposes.
However,
as
noted
in
the
Panel's
response
to
Question
3,
the
uncertainty
in
emission
rates
for
new
sites
and
conditions
is
likely
underestimated
by
the
procedures
used
for
characterizing
the
uncertainty
in
fluxes
for
a
test
site
and
data
set.

The
question
of
the
accuracy
of
this
model
is
dependent
on
the
dispersion
code
that
is
incorporated
into
it.
PERFUM
has
been
described
as
having
the
source
code
for
ISCST3
incorporated
into
its
computational
routines,
making
it
equivalent
to
ISCST3.
ISCST3
has
been
validated
by
a
number
of
field
studies,
and
it
has
been
used
to
accurately
predict
airborne
concentrations
in
a
variety
of
physical
situations.
Therefore,
to
the
extent
that
ISCST3
is
accurate
for
a
given
dispersion
scenario,
PERFUM
follows
suit.

Since
air
samplers
were
installed
at
distances
that
were
much
smaller
than
typical
buffer
zones,
new
more­
rigorous
studies
are
needed
that
test
the
accuracy
of
the
buffer
zone
calculation
33
of
43
using
samplers
installed
at
500
to
1500
m.
This
would
allow
evaluation
of
the
model
reliability
at
the
lower
concentrations.

Little
information
is
available
to
determine
the
accuracy
of
the
flux
and
down­
wind
concentration
estimates.
Nothing
in
the
report
addresses
this
directly.
It
may
be
possible
to
test
PERFUM's
ability
to
quantify
airborne
concentrations
at
sensitive
receptors
by
using
the
emission
rate
data
from
one
experiment
as
an
input
parameter
to
determine
the
appropriate
buffer
zones
using
meteorological
data
appropriate
for
another
experiment.
Once
the
analysis
is
completed,
the
"
predicted"
buffer
zone
could
be
compared
to
the
measured
air
concentration
data
that
were
not
used
in
the
buffer
zone
calculation,
to
determine
the
accuracy
of
PERFUM.
This
kind
of
cross­
site
validation
is
needed
and
does
not
require
any
additional
data.

Flux
studies
using
direct
flux
methods
could
be
performed
to
better
parameterize
the
PERFUM
model
and
to
determine
the
variability
in
emission
rate
as
a
function
of
soil
type,
region,
and
timing.
This
information
would
also
help
in
determining
if
the
indirect
method
is
accurately
estimating
flux
rates.

Although
the
Panel
thought
that
direct
flux
measurements
would
probably
be
more
accurate,
the
use
of
the
indirect
flux
method
assists
in
calibrating
the
modeling
system
to
a
particular
application
site.
This
would
improve
the
accuracy
for
a
particular
field
study.
The
appropriateness
of
model
input
parameters
to
risk
assessment
was
addressed
in
Dr.
Reiss'
presentations
on
August
24,
2004
and
also
in
some
of
the
other
Panel
members'
responses
to
questions.

Several
studies
have
compared
ISCST3­
based
model
predictions
with
measured
downwind
air
concentrations
of
fumigants.
These
include
studies
for
methyl
bromide,
telone,
metam
sodium,
and
possibly
others.
The
PERFUM
developers
should
add
these
comparative
data,
or
at
least
reference
it,
in
the
documentation.
Some
of
the
studies
can
be
found
in
Symposium
Proceedings
by
J.
N.
Seiber,
J.
A.
Knuteson,
J.
E.
Woodrow,
N.
L.
Wolfe,
M.
V.
Yates
and
S.
R.
Yates,
1996.

The
indirect
flux
method
bypasses
the
use
of
more
sophisticated,
but
data­
intensive,
physical
models
that
potentially
would
be
able
to
predict
emissions
from
fundamental
soil
properties
such
as
soil
type,
moisture
content,
sorption
rate,
etc.
The
use
of
the
indirect
flux
method
allows
the
ISCST3
model
to
be
tied
to
a
physical
situation
(
one
field
location
and
time)
without
having
to
resort
to
additional
data
collection
and
a
different
model
to
predict
emissions.
Although
the
use
of
soil­
based
models
may
prove
difficult,
users
of
PERFUM
face
equal
challenges
in
obtaining
high­
quality
emissions
data
that
represents
conditions
over
a
large
region
and
throughout
the
year.
It
may
be
useful
to
have
the
ability
to
use
more
detailed
models
that
handle
soil,
temperature,
and
the
soil­
atmosphere
interface.
Such
a
model
can
provide
probabilistic
analysis
for
risk
assessment,
and
may
prove
more
reliable
predicting
flux
at
new
locations
or
for
new
conditions.

b)
The
Agency
is
particularly
concerned
about
air
concentrations
in
the
upper
ends
of
the
distribution.
Are
these
results
presented
in
a
clear
and
concise
manner
that
34
of
43
would
allow
for
appropriate
characterization
of
exposures
that
could
occur
at
such
levels?

Response
PERFUM
and
the
incorporated
methodology
assume
that
the
ISCST3
model
can
accurately
predict
atmospheric
concentrations
at
receptor
points
with
accuracy,
including
rare
events.
Once
a
buffer
zone
has
been
determined
in
a
study
area,
a
post­
analysis
of
the
PERFUM
model
would
be
useful.
Without
such
information
it
is
difficult
to
determine
how
well
PERFUM
quantifies
the
upper
ends
of
the
distribution.

As
with
the
first
question,
the
accuracy
of
any
output
of
the
model
is
dependent
on
the
accuracy
of
the
internal
computational
engine,
ISCST3.
Therefore,
based
on
the
above
factors,
both
low
and
high­
end
concentrations
should
be
as
accurate
as
the
limitations
of
ISCST3.

Another
concern
that
must
still
be
addressed
in
order
to
estimate
the
high­
end
probability
distribution
of
ambient
concentrations
and
exposures
is
the
treatment
of
calm­
wind
periods
in
the
simulation
procedure.
Since
the
model
is
intended
for
risk
management
for
a
procedure
that
could
be
repeated
often
at
different
sites,
reducing
the
probability
of
a
serious
exposure
at
any
one
of
many
sites,
e.
g.,
in
a
given
year,
will
require
a
much
lower
probability
of
occurrence
at
each
individual
site.
In
particular,
to
keep
the
probability
of
one
or
more
serious
exposures
at
N
sites
in
a
given
year
below
PN,
then
(
assuming
independence)
the
probability
of
a
serious
exposure
at
each
site
must
be
kept
below
P1,
where
1­
PN
=
(
1­
P1)
N,
so
that
P1
=
1
­
(
1­
PN)
1/
N.
For
example,
if
PN
=
0.05
and
N
=
100,
then
P1
=
0.000513,
that
is,
you
must
be
99.95%
confident
of
no
serious
exposure
at
each
of
the
100
sites.
As
such,
very
high­
end
(
e.
g.,
99%,
99.5%,
or
higher)
protection
should
be
sought
for
individual
applications.

Some
Panel
members
were
concerned
that
the
field
studies
sometimes
seemed
to
show
some
experimental
weaknesses.
Some
of
the
studies
lost
the
first
24
hours
of
data,
the
most
important
time
period.
The
studies
seemed
to
have
only
one
set
of
downwind
samples,
which
may
not
allow
a
complete
characterization
of
the
downwind
environment.
Other
Panel
members
have
mentioned
the
possible
need
for
a
vertical
placement
of
samplers
to
test
that
component.

c)
The
PERFUM
model
calculates
the
concentration
distributions
both
in
all
directions
and
for
only
the
maximum
concentration
direction.
Can
the
Panel
comment
on
how
accurately
the
model
approximates
both
of
these
distributions?

Response
Some
Panel
members
suggested
that
allowing
only
the
maximum
concentration
distribution
introduces
unnecessary
conservatism
into
the
model
output.
Therefore,
it
is
recommended
that
the
entire
distribution
of
wind
directions
be
used
in
order
to
maintain
an
appropriate
physical
representation
of
the
emission
distribution
and
exposure
at
the
receptors.
However,
it
may
be
useful
for
PERFUM
to
allow
the
option
of
an
analysis
based
on
the
maximum
concentration
direction
as
a
tool
for
examining
potential
impacts.
35
of
43
In
addition,
it
is
recommended
that
the
output
of
PERFUM
be
modified
so
that
GIS
or
graphics
programs
can
access
the
output
for
visualization.
Better
decisions
regarding
the
buffer
zones
can
be
made
through
examining
the
physical
impact
of
buffer
zones
to
actual
field
configurations.

The
use
of
portable
air
quality
samplers
using
charcoal
filters
was
performed
in
all
directions
within
a
range
of
30
feet
to
140
feet
from
the
application
area.
However,
there
are
significant
uncertainties
in
predicting
buffer
zones
and
concentrations
at
those
locations.

An
additional
concern
regarding
the
accuracy
is
the
use
of
meteorological
data
sets
that
are
not
associated
with
the
flux
at
a
particular
location.
It
is
recommended
that
additional
work
be
performed
to
provide
assurance
that
the
appropriate
wind
data
are
used
in
the
eventual
model.

Question
8:
A
sensitivity/
uncertainty
analysis
has
been
conducted
and
is
described
in
the
PERFUM
background
document.

a)
What
types,
if
any,
of
additional
contribution/
sensitivity
analyses
are
recommended
by
the
Panel
to
be
the
most
useful
in
making
scientifically
sound,
regulatory
decisions?

Response
The
report
includes
a
reasonable
discussion
of
the
various
sources
of
uncertainty
in
the
study
and
provides
an
initial
assessment
of
the
sensitivity
of
model
predictions
to
selected
alternative
model
assumptions
and
parameter
values.
Several
factors
were
considered
on
an
individual
basis,
but
not
on
a
comprehensive
basis;
however,
this
seems
appropriate
for
the
current
stage
of
model
development.
The
factors
discussed
include:
the
flux
emission
estimate/
profile,
meteorological
data
and
sources
of
datasets,
anemometer
height,
calms
processing,
non­
meteorological
environmental
factors,
(
Gaussian
formulated)
model­
model
comparisons,
indoor
exposure,
human
activity
patterns,
coincidental
temporally/
spatially
close
multi­
field
applications,
seasonality,
and
horizontal
placement
of
monitors
during
field
studies
for
flux
estimation.

The
Panel
recommended
separating
out
emissions
related
issues
from
meteorological/
air
dispersion
model
issues;
and
then,
to
the
extent
possible,
separating
out
uncertainty
and
variability.
The
discussion
of
emission
rate
uncertainty
needs
more
clarification
in
the
report.

Comparing
vertical
concentration
profiles
may
be
one
other
way
to
improve
the
calibration
of
the
model.
Currently,
only
data
in
the
horizontal
direction
are
being
collected
and
used.

b)
What
should
be
routinely
reported
as
part
of
a
PERFUM
assessment
with
respect
to
inputs
and
outputs?

Response
36
of
43
One
of
the
variations
considered
around
the
flux
estimation
includes
the
use
of
a
higher
coefficient
of
variation
for
emission
rates
(
47%)
and
the
use
of
higher
base­
case
(
mean)
emission
rate,
corresponding
to
the
75th
percentile
of
the
estimated
value.
As
noted
in
the
Panel's
response
to
Question
3,
the
use
of
the
47%
CV,
while
more
likely
representative
of
site­
to­
site
variation
than
the
smaller
CV
determined
by
fitting
a
single
flux
study
at
a
single
site,
the
assumption
that
uncertainties
in
sequential
simulation
periods
are
random
and
independent
could
underestimate
the
variation
associated
with
persistent
meteorological
conditions
(
that
affect
the
emission
rate)
and
systematic
error
in
the
model.

More
mechanistic
investigation
of
the
emission
flux
(
soil
temperature,
tarp
parameters,
etc.)
could
guide
additional
data
collection,
analysis,
and
reporting.
The
state­
of­
development
of
mechanistic
soil
fumigant
flux
modeling
is
such
that
the
need
for
field
studies
will
not
be
eliminated;
however,
it
may
be
mutually
beneficial
to
explore
what
can
be
done
in
a
complimentary
fashion.

In
the
final
presentation
of
the
PERFUM
model
it
would
be
important
to
clearly
articulate
whether
an
input
is
a
"
random
input"
or
an
"
uncertain
input"
(
i.
e.,
flux
rate).

c)
Are
there
certain
tables
and
graphs
that
should
be
reported?

Response
No
analysis
was
presented
to
address
the
implications
of
upset
or
unusual
conditions
in
fumigant
handling
and
application
that
could
lead
to
especially
high
exposures.

The
tables
and
graphs
in
the
report
should
be
provided
for
most
of
the
user's
needs.

d)
What
types
of
further
evaluation
steps
does
the
Panel
recommend
for
PERFUM?

Response
A
simple
relationship
is
anticipated
between
the
ground
level
emission
source
and
the
1.5
m
collection
cartridges
during
the
field
studies.
Nevertheless,
the
absence
of
any
concentration
data
collected
in
the
vertical
dimension
precludes
challenging
this
assumption.
Additionally,
concentration
data
in
the
vertical
dimension
can
be
used
along
with
the
1.5
m
concentration
data
to
improve
the
emission
flux
estimation
obtained
from
back­
calculation
­­
similar
to
the
additional
placement
of
monitors
in
the
horizontal
dimension
being
planned
for
a
future
field
study.

The
discussion
of
the
calibration
of
ISCST3
to
the
robust
highest
concentration
(
measured)
using
meteorological
datasets
that
include
calm
conditions,
and
have
been
processed
for
calm
conditions,
was
not
adequate
to
convince
all
Panelists
that
calm
periods
have
been
accounted
for.

A
meteorological
sensitivity
run
was
briefly
suggested
for
crops
with
specific
windows
of
application
time
in
specific
regions
(
i.
e.,
January
and
February
applications
for
strawberries
in
37
of
43
Florida).
The
monthly
sensitivity
runs
captured
a
component
of
this,
though
the
region
specific
feature
may
not
be
captured.

Concluding
Discussions
of
the
Panel
A
number
of
issues
not
directly
related
to
the
Agency
questions
were
introduced
and
discussed
at
the
end
of
the
meeting.

The
PERFUM
technical
documentation
should
include
descriptions
of
the
canister
sampling
methods.
Whether
the
charcoal
canisters
were
evacuated
was
not
clear
to
some
of
the
Panel
members.
A
query
of
the
developers
indicated
that
indeed
they
were
attached
to
a
vacuum
pump.
This
led
to
a
discussion
of
potential
alternative
sampling
devices.
One
such
methodology,
Suma
Canisters
(
Compendium
Method
TO­
15,
EPA/
625/
R­
96/
010b
(
http://
www.
epa.
gov/
ttn/
amtic/
airtox.
html),
involving
the
use
of
evacuated
spheres
as
quick
air
samplers,
was
described
to
the
Panel
and
the
effectiveness
of
these
devices
briefly
discussed.

There
were
several
references
to
charcoal
canisters
used
in
the
field
studies.
The
word
"
canister"
implies
some
other
technology
that
is
commonly
used
in
air
sampling
as
described
above,
so
a
mis­
application
of
this
nomenclature
is
misleading
and
should
be
clarified.

It
appears
that
the
field
studies
made
use
of
charcoal
sorbent
tubes.
These
tubes
consist
of
a
small
glass
tube
filled
with
a
specially
prepared
charcoal
sorbent
material.
A
calibrated
pump
pulls
air
through
the
sorbent
tube
at
an
appropriate
rate,
after
which
the
adsorbed
material
is
chemically
desorbed
in
the
laboratory
and
analyzed,
typically
by
gas
chromatography.
This
technique
is
very
commonly
used,
but
may
suffer
from
relatively
high
detection
limits,
particularly
for
short
sampling
times,
as
the
sensitivity
of
the
method
is
dependent
on
the
amount
of
air
that
is
pulled
through
the
sorbent
tube.
The
implication
for
the
flux
studies
is
that
the
use
of
this
approach
may
limit
the
duration
of
the
sampling
periods
to
times
longer
than
might
be
advantageous
for
a
more
detailed
examination
of
the
flux
behavior.
For
example,
hourly
samples
that
correlate
with
the
hourly
meteorological
data
collection
and
the
hourly
output
of
ISCST3
might
be
useful
to
more
fully
characterize
the
first
important
24
hours
post­
application.

An
alternative
sampling
method
based
on
evacuated
canisters,
called
Summa
canisters
as
referenced
above,
can
provide
from
very
short
sampling
periods
(
1­
2
minutes)
to
24
hours
or
more.
This
method
may
prove
useful
for
future
flux
validation
studies.

There
was
discussion
of
whether
and
how
field­
sampling
layouts
(
sampling
design)
might
be
modified
to
improve
the
results
from
the
back
calculation
of
flux.
It
was
suggested
that
different
layouts
could
be
examined
using
expectations
from
the
fitted
model.
Clearly,
if
there
were
cheaper
measurement
instruments,
even
if
they
were
less
precise,
their
use
in
conjunction
with
the
expensive
instruments
would
greatly
increase
the
information
available
for
flux
estimation.
The
cheap
and
expensive
could
be
used
together
at
a
few
locations
and
the
relationship
between
the
two
used
to
calibrate
the
cheap
method
values
to
the
expensive
method
values.
There
is
a
need
to
consider
standardization
of
the
sampling
design
if
the
methodology
is
to
be
established
as
useful
for
other
regions
of
the
country.
38
of
43
The
use
of
alternative
sampling
designs,
such
as
the
single
point
design,
or
designs
that
include
samples
taken
in
different
vertical
positions
(
gradient
methods)
was
discussed.
Some
of
the
Panel
thought
there
was
a
need
for
alternative
designs,
and
that
the
back
calculation
method
applied
to
single
height
data
was
not
adequate
for
the
purposes
to
which
the
results
of
the
PERFUM
model
are
to
be
used.
Gradient
methods
have
some
appeal
but
there
are
significant
field
constraints
that
limit
their
use,
often
resulting
in
too
few
measurement
locations.
None
of
the
measurement
methods
would
be
useful
for
the
very
low
concentrations
that
might
be
observed
at,
e.
g.,
1
km.

One
Panel
member
justified
the
need
for
vertical
data
based
on
recent
experiments
that
generated
more
data
than
was
available
for
this
study.
The
reason
for
observing
zero
concentrations
in
locations
where
the
model
says
there
should
be
concentration
can
partially
be
explained
with
these
vertical
data.
Typically,
zero
concentrations
are
the
result
of
input
factor
levels,
for
example
stability,
affecting
the
dispersion
parameters
of
the
model.
By
using
the
predicted
dispersion
coefficients,
even
slight
changes
in
elevation
are
shown
to
be
able
to
make
big
differences.
In
reality,
dispersion
coefficients
are
not
step
functions
but
are
continuous
and
it
is
not
a
trivial
task
to
determine
how
to
establish
sufficient
vertical
samples
to
get
useful
data
for
estimation.

The
Panel
discussed
the
use
of
log
transformations
in
the
back
calculation
of
flux
and
its
relationship
to
the
mass
balance
assumptions
made
in
the
model.
The
decision
to
use
a
regression
line
that
is
forced
through
the
origin
(
the
ratio
model)
versus
an
intercept
linear
model
cannot
be
justified
simply
on
statistical
terms,
especially
in
light
of
the
relatively
small
sample
sizes
typically
seen.
It
would
be
best
if
there
were
a
physical
interpretation
for
the
intercept
term
(
for
example,
nonlinearity
of
relationship
between
observed
and
expected
concentrations
close
to
zero).
If
no
physical
reason
can
be
found,
it
makes
sense
to
force
the
intercept
to
zero.
It
is
possible
that
some
of
the
studies
that
have
been
done
on
other
fumigants
could
shed
light
on
this
issue.

There
was
discussion
about
the
practical
limitations
of
laying
cover
tarps
and
the
impact
of
non­
uniform
tarp
cover
on
directional
changes
in
concentrations.
Imperfections
may
allow
emission
concentrations
in
situations
where
the
model
says
the
concentrations
should
be
zero.
Some
of
this
is
lost
when
the
data
are
averaged
over
an
hour
(
or
longer)
to
allow
use
in
the
ISCST3
model.

Finally,
there
was
some
discussion
that
the
research
on
methyl
bromide
had
indicated
that
degradation
of
the
chemical
is
a
simple
process.
The
PERFUM
model
seems
to
assume
that
this
will
be
the
same
for
other
chemicals
as
illustrated
by
the
case
study
using
iodomethane.
The
Panel
wondered
if
this
assumption
is
warranted
and
whether
there
are
data
to
support
this
assumption.
39
of
43
REFERENCES
Baker,
J.
M.,
W.
C.
Koskinen
and
R.
H.
Dowdy.
1996.
Volatilization
of
EPTC:
simulation
and
measurement.
J.
Environ.
Qual.
25:
169­
177.

Compendium
of
Methods
for
the
Determination
of
Toxic
Organic
Compounds
in
Ambient
Air:
Method
TO­
15,
Determination
of
Volatile
Organic
Compounds
(
VOCs)
in
Air
Collected
in
Specially­
Prepared
Canisters
and
Analyzed
by
Gas
Chromatography/
Mass
Spectrometry
(
GC/
MS).
Second
Edition.
U.
S.
Environmental
Protection
Agency,
Office
of
Research
and
Development,
OH.
EPA/
625/
R­
96/
010b.
January
1999.
(
http://
www.
epa.
gov/
ttn/
amtic/
airtox.
html)

Honaganahalli,
P.
S.
and
J.
N.
Seiber.
2000.
Measured
and
predicted
air
shed
concentrations
of
methyl
bromide
in
an
agricultural
valley
and
applications
to
exposure
assessment.
Atmos.
Environ.
34:
3511­
3523.

Jury,
W.
A.,
W.
F.
Spencer
and
W.
J.
Farmer.
1983.
Behavior
assessment
model
for
trace
organics
in
soil:
1.
Model
description.
J.
Environ.
Qual.
12:
558­
564.

Seiber,
J.
N.,
J.
A.
Knuteson,
J.
E.
Woodrow,
N.
L.
Wolfe,
M.
V.
Yates
and
S.
R.
Yates
(
Eds.).
1996.
Fumigants:
Environmental
fate,
exposures,
and
analysis.
ACS
Symposium
Series
652,
American
Chemical
Society,
Washington,
D.
C.
pp135­
153.

Wang,
D.,
S.
R.
Yates
and
J.
Gan.
1997.
Temperature
effect
of
fate
and
transport
of
methyl
bromide
in
soil
fumigation.
J.
Environ.
Qual.
26:
1072­
1079.

Woodrow,
J.
E.,
J.
N.
Seiber
and
C.
Dary.
2001.
Predicting
pesticide
emissions
and
downwind
concentrations
using
correlations
with
estimated
vapor
pressures.
J.
Agric.
Food
Chem.
49:
3841­
3846.

Yates,
S.
R.,
D.
Wang,
S.
K.
Papiernik
and
J.
Gan.
2002.
Predicting
pesticide
volatilization
from
soils,
Environmetrics.
13:
569
 
578.
40
of
43
Appendix
1.

Information
on
the
indirect
flux
estimation
method
provided
by
Dr.
Mitchell
J.
Small
following
the
meeting.

The
indirect
method
for
estimating
emissions
from
ambient
concentrations
is
based
on
the
relationship:

T
E
b
C
×
+
=
1
(
A.
1)

where
C
is
the
ambient
concentration,
E
is
the
emission
rate,
b1
is
a
background
concentration
that
would
be
present
even
if
there
were
no
emission
sources
at
the
site
(
the
background
concentration
could
be
present
as
result
of
emissions
from
other
local
or
regional
sources),
and
T
is
a
linear
source­
receptor
transfer
coefficient
between
the
emission
source
and
the
receptor.
This
equation
assumes
a
linear
fate­
and­
transport
process
(
as
is
assumed
for
most
applied
atmospheric
dispersion
models)
and
is
applicable
either
at
a
particular
time
or
over
some
averaging
period.
The
parameters
of
the
equation
relating
C
to
T
(
the
intercept
b1
and
the
slope
E)
are
estimated
in
the
PERFUM
methodology
by
regressing
observed
concentrations
(
Ymeas)
at
many
locations
following
a
fumigant
application
vs.
the
concentrations
(
XISC)
predicted
by
the
ISCST3
model
for
a
unit
or
nominal
emission
(
Fluxnominal):

ISC
meas
X
m
b
Y
×
+
=
(
A.
2)

The
intercept
obtained
with
this
method,
b,
is
equivalent
to
b1
in
the
equation
above,
while
the
slope,
m,
allows
E
to
be
calculated
as
follows:

al
no
est
Flux
m
Flux
E
min
×
=
=
(
A.
3)

The
indirect
approach
for
emissions
estimation
is,
in
general,
appropriately
formulated
and
implemented
in
the
PERFUM
model.

One
of
the
issues
raised
as
part
of
the
third
question
to
the
Panel
involves
the
way
that
the
intercept,
b,
in
equation
A.
2
is
treated
in
the
analysis.
As
noted
above,
the
intercept
in
equation
A.
1
and
A.
2
has
a
physical
meaning
 
it
corresponds
to
the
"
background"
concentration
that
would
be
present
at
the
receptor,
for
the
time
period
of
interest,
had
no
emission
been
present,
i.
e.,
E
=
0.
Since
ideally,
when
a
fumigant
is
introduced
at
a
site,
it
should
be
the
only
significant
source
of
the
chemical,
there
is
a
preference
expressed
that
the
intercept,
b,
be
equal
to
zero.
In
previous
and
the
PERFUM­
proposed
approaches,
when
the
estimated
slope
is
not
significantly
different
from
zero,
this
is
considered
an
"
acceptable"
result
and
the
estimated
slope
m
that
is
used
to
estimate
the
emissions,
can
be
used
directly.
However,
when
b
is
significantly
different
from
zero,
the
data
are
processed
further,
by
matching
the
rank
order
of
the
model
predictions
with
that
of
the
observed
data
and/
or
the
regression
is
repeated
with
the
intercept
forced
to
be
zero,
in
order
to
provide
a
"
more
appropriate"
estimate
of
the
slope
and
associated
emission
flux
rate.
41
of
43
It
should
be
noted
that
the
opposite
inference
and
estimation
procedure
could
be
(
and
often
is)
used
by
statisticians
and
scientists,
given
the
result
that
an
estimated
slope
in
a
linear
regression
is,
or
is
not,
significantly
different
from
zero.
In
this
alternative
approach,
the
result
that
an
estimated
intercept
is
not
significantly
different
from
zero
implies
that,
since
there
is
not
sufficient
confidence
that
the
intercept
is
different
from
zero,
the
intercept
term
should
be
deleted
from
the
model
and
the
equation
re­
estimated
with
the
intercept
set
equal
to
zero.
In
contrast,
when
an
intercept
is
found
to
be
significantly
different
from
zero,
this
indicates
that
the
non­
zero
intercept
term
is
"
real",
and
should
be
left
in
the
model.
A
key
issue
in
these
alternative
interpretations
and
approaches
is,
the
assessment
of
whether
there
is
physical
reason
to
believe
that
a
background
concentration
could
or
should
be
present
at
the
site,
so
that
a
non­
zero
intercept
does
(
or
does
not)
make
physical
sense.
Has
the
fumigant
been
applied
previously
at
the
site
or
at
nearby
sites
(
this
would
especially
be
the
case
if
a
sequence
of
test
applications
were
conducted
at
a
site
for
the
purpose
of
model
calibration)?
Has
the
fumigant
been
used
for
a
long­
enough
time
so
that
a
regional
(
or
perhaps
even
a
global)
background
is
present
in
the
environment?
Furthermore,
depending
on
how
very
low,
below­
detection­
limit
concentration
measurements
are
determined
and
reported
(
e.
g.,
at
the
detection
limit,
at
one­
half
the
detection
limit,
or
at
zero),
an
apparent
background
concentration
could
be
inferred
as
a
result
of
the
way
the
data
are
treated.
1
These
issues
have
not
been
addressed
in
an
adequate
manner
in
the
PERFUM
documentation,
and
should
provide
the
motivation
for
seeking
either
to
"
leave
the
intercept
term
in
the
model,"
in
the
case
where
there
is
a
physical
basis
for
a
non­
zero
background
concentration
at
the
site,
or
preferring
to
"
remove
the
intercept
term
from
the
model,"
in
the
case
where
physical
arguments
and
evidence
(
e.
g.,
monitoring
upwind
of
the
site)
suggest
that
the
background
term
is
indeed
zero.

A
second
issue
addressed
as
part
of
the
model
development
is
whether
the
data
should
be
log­
transformed
prior
to
the
regression.
This
approach
could
be
motivated
by
the
positivelyskewed
non­
negative
nature
of
measured
air
pollution
concentrations,
which
tend
to
be
lognormally
distributed.
The
log­
transform/
linear
regression
approach
is
not
recommended,
since
it
violates
the
linear,
mass­
balance
assumptions
inherent
to
equation
A.
1.
However,
it
may
be
the
case
that
the
errors
in
the
regression
are
non­
normal,
perhaps
lognormal,
especially
when
these
errors
are
large
and/
or
for
very
low
observed
and
predicted
concentrations,
since
negative
concentrations
could
be
inferred
for
these
cases
when
a
normal
error
structure
is
assumed.
The
assumption
of
a
non­
normal
error
structure
for
predicted
concentrations
would
require
the
regression
model
to
be
estimated
using
advanced
numerical
methods,
rather
than
with
standard
linear
regression
procedures
available
in
common
spreadsheet
and
statistical
packages.
Since,
as
noted
below,
the
major
source
of
error
and
variability
associated
with
the
use
of
the
PERFUM
model
does
not
involve
its
use
to
estimate
emissions
at
a
given
site
and
time
where
ambient
concentrations
have
been
measured,
but
rather
its
use
to
predict
emissions
under
different
conditions
and
for
different
sites,
the
extra
effort
needed
to
pin
down
emissions
and
their
uncertainty
for
a
particular
set
of
experiments
places
the
emphasis
in
the
wrong
place.
As
such,
unless
there
is
strong
evidence
of
large,
highly­
skewed
regression
errors,
the
use
of
these
advanced
methods,
in
conjunction
with
a
non­
normal
error
assumption
for
predicted
concentrations
in
the
indirect
method
for
emissions
estimation
is
not
recommended.

1
The
PERFUM
model
developers
indicate
that
the
LOD
data
are
assumed
equal
to
zero
in
the
regression
method,
so
this
could
lead
to
a
minor
underestimation
of
the
intercept.
42
of
43
Uncertainty
in
emission
rates
is
generated
in
the
model
by
randomly
sampling
the
estimated
slope
(
and
inferred
value
of
E)
based
on
a
t­
distribution
(
now
a
normal
distribution)
with
a
standard
deviation
defined
by
the
standard
error
of
the
slope
estimate.
While
appropriate
in
concept,
the
method
is
susceptible
to
error
for
a
number
of
reasons.
Furthermore,
it
cannot
capture
the
predominant
uncertainty
that
is
present
when
it
is
applied
in
an
extrapolation
mode,
for
new
conditions
or
at
different
sites.
First,
in
the
case
where
the
error
in
observed
and
predicted
concentrations
is
non­
normal
and
a
small
sample
of
predicted­
observed
concentration
pairs
are
used
to
fit
the
model,
the
error
structure
of
the
estimated
slope
and
inferred
flux
rate
will
likewise
be
non­
normal
(
nor
like
a
t­
distribution,
which
is
also
symmetric).
Again,
consideration
of
this
possibility
would
require
the
use
of
advanced
statistical
procedures
that,
given
the
overall
uncertainty
in
the
models,
observations,
and
approach,
do
not
seem
warranted.
A
second
concern
with
the
way
that
the
method
is
implemented
involves
the
random
selection
of
uncertain
emission
rates
in
sequential
time
periods,
which
in
the
applications
described
to
the
Panel
involves
time
steps
of
a
few
hours
up
to
12
hours,
with
typically
~
4
h
time
periods
used
for
a
1­
day
simulation.
Since
the
variable
and
uncertain
factors
that
could
make
the
flux
rate
higher
(
or
lower)
for
a
given
field
and
time
period
than
that
estimated
from
the
nominal
slope
 
e.
g.,
high
(
or
low)
winds
and/
or
temperature,
site­
specific
soil
or
application
conditions,
or
a
systematic
error
in
the
overall
estimation
procedure
 
are
likely
to
persist
over
time
for
that
field
(
or,
beyond
that,
be
applicable
to
all
of
the
time
periods
for
that
field),
sampling
the
uncertain
emission
rates
in
sequential
time
periods
as
independent
from
each
other
will
underestimate
the
uncertainty
variance
of
the
daily
emission
rate,
and
therefore
also
underestimate
the
frequency
of
high
end
emissions
and
associated
high­
end
buffer
zone
lengths.
Indeed,
this
may
be
one
of
the
reasons
that
the
uncertainties
in
the
model
estimates
for
these
outputs
appear
to
be
small,
in
this
case,
for
an
inappropriate
reason.
A
method
is
needed
to
incorporate
the
temporal
persistence,
or
whole­
period
nature,
of
the
uncertainty
in
the
flux
estimate.
2
Alternatively,
if
shorter
averaging
times
are
used
to
compute
exposure
concentrations
for
the
buffer
zone
calculation
(
i.
e.,
concentration
averaging
times
that
are
of
the
same
duration
as
the
period
for
which
the
uncertainties
in
emission
rates
are
sampled
within
the
model),
then
this
would
obviate
the
need
to
consider
serial
persistence
in
emission
rate
uncertainty.

The
current
model
basis
predicted
emission
rates
at
a
site
on
the
profile
of
measured
emissions,
indexed
by
time
of
day.
While
the
time
of
day
does
provide
a
first
surrogate
for
atmospheric
processes
that
affect
emission
rates
 
typically
stability
and
wind
speed
conditions
 
it
is
only
a
partial
surrogate.
Direct
consideration
of
these
conditions
would
provide
a
more
robust
approach.
Furthermore,
it
creates
confusion
when
the
application
in
the
test
case
occurs
at
different
times
of
the
day,
since
the
emission
rate
is
pinned
only
to
that
time
of
day,
and
not
to
the
amount
of
time
that
has
elapsed
since
the
application
was
made.
A
"
first­
morning"
emission
will
be
very
different
if
it
occurs
immediately
following
an
early
morning
application,
as
opposed
to
the
case
where
it
occurs
on
the
day
after
an
early
afternoon
application,
since
in
the
latter
case
a
significant
portion
of
the
mass
will
have
already
been
lost
by
the
time
this
"
first
morning"
occurs.
As
such,
some
procedure
for
considering
the
amount
of
time
that
has
elapsed
(
and
the
amount
of
mass
that
has
been
lost)
since
the
application
is
needed.
This
again
highlights
the
advantage
of
a
more­
mechanistic,
mass­
balance
based
approach
for
predicting
emissions.

2
If
this
is
done,
care
should
be
taken
to
ensure
that
the
high­
end
emission
rates
determined
in
this
way
do
not
yield
total
emissions
that
exceed
100%
of
the
applied
fumigant.
This
concern
would
be
addressed
with
the
use
of
a
model
that
ensures
that
mass
balance
is
maintained,
as
discussed
later.
43
of
43
A
mechanistic
model
for
emissions
for
a
field
would
consider
the
mass
of
fumigant
applied
to
the
field,
the
mass
balance
of
the
fumigant
on
or
in
the
soil,
and
that
which
is
lost
to
the
atmosphere.
It
might
also
consider
an
explicit
equation
relating
the
emission
rate
to
the
soil
properties,
chemical
properties
(
if
multiple
chemicals
are
considered),
and
the
meteorological
conditions,
especially
wind
speed
and
possibly
stability,
pressure,
and
temperature.
The
effects
of
these
meteorological
factors
on
the
emission
rate
are
now
implicitly
included
by
consideration
of
the
time­
of­
day
in
the
flux
rate
used
in
the
model,
but
a
direct
functional
fit
of
the
estimated
emission
rate
to
the
wind
speed
(
and
possibly
the
other
meteorological
variables)
present
at
the
time
of
the
experiment,
would
allow
a
more
robust
emission
relationship
to
be
determined.
The
wind
speed
and
stability
conditions
are
already
considered
in
computing
the
atmospheric
transport
of
the
fumigant
once
it
is
emitted.
The
effects
of
these
factors
on
the
estimated
flux
rate
should
also
be
explored.
The
resulting
emissions
function
could
be
applied
to
other
sites
and
time
periods
with
different
meteorological
conditions
with
more
confidence
than
the
current
method
that
is
based
on
the
time­
of­
day
alone.

While
it
is
clear
that
a
mechanistic
model
that
incorporates
mass
balance
is
preferred
and
should
be
explored
for
eventual
Agency
use,
this
type
of
model
may
not
be
available
to
meet
the
Agency's
near­
term
needs
for
risk
management.
In
the
interim,
an
empirical
approach
that
considers
the
constraints
of
mass
balance
and
the
reality
of
site­
to­
site
variability
could
be
explored
for
capturing
the
key
uncertainties
 
those
that
occur
under
new
(
un­
monitored)
atmospheric
and
site
conditions.
A
possible
approach
is
to
focus
on
the
fraction
of
the
applied
mass
that
is
emitted,
either
over
the
entire
duration
of
an
experiment,
or
during
the
first
day.
Since
this
fraction
is
constrained
to
be
between
zero
and
one,
a
beta
distribution
could
be
used
to
characterize
and
simulate
the
variability
and
uncertainty
of
this
emission
fraction.
Once
the
total
mass
emitted
has
been
simulated
(
calculated
as
the
product
of
the
applied
mass
and
the
fraction
emitted),
a
temporal
distribution
of
the
emission
rate
over
the
day
that
is
consistent
with
this
total
mass
emission
could
be
generated.
Simulation
should
consider
overall
exponential
decay
in
the
average
emission
rate
with
time
(
consistent
with
a
first­
order
loss
process),
the
time
of
day
(
as
a
surrogate),
or
atmospheric
conditions
in
the
PERFUM­
accessed
meteorological
file
(
preferably).
Random
errors
of
the
type
that
have
been
observed
in
(
direct
or
indirect)
flux
studies
from
the
literature
should
be
added
to
the
emission
rate
for
each
of
the
time
periods
(
but
still
be
constrained
so
that
the
sum
yields
the
desired
mass
fraction).
When
the
model
is
applied
at
new
sites
or
for
use
in
determining
generally­
applicable
rules
(
e.
g.,
minimal
buffering
zone
lengths
designed
to
be
protective
of
all
sites),
the
beta
distribution
for
the
fraction
of
the
applied
mass
that
is
emitted
should
have
a
variance
that
reflects
the
(
large)
variation
observed
in
emitted
mass
fractions
estimated
in
previous
studies
at
different
sites
and
under
different
weather
conditions.
For
use
at
a
given
site,
previous
studies
at
that
site
(
i.
e.,
direct
or
indirect
flux
estimation
studies)
could
be
used
to
determine
the
(
smaller)
variation
that
is
appropriate
for
use
at
that
site.
