SAP
Minutes
No.
2004­
01
A
Set
of
Scientific
Issues
Being
Considered
by
the
Environmental
Protection
Agency
Regarding:

REFINED
(
LEVEL
II)
TERRESTRIAL
AND
AQUATIC
MODELS
PROBABILISTIC
ECOLOGICAL
ASSESSMENTS
FOR
PESTICIDES:
Level
II
Aquatic
Model
Session
APRIL
1
and
2,
2004
FIFRA
Scientific
Advisory
Panel
Meeting,
held
at
the
Crowne
Plaza
Washington­
National
Airport
Hotel,
Arlington,
Virginia
2
of
44
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
pesticide­
related
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
44
CONTENTS
PARTICIPANTS...........................................................................................................
5
INTRODUCTION.........................................................................................................
7
CHARGE.......................................................................................................................
7
SUMMARY
OF
PANEL
DISCUSSION
AND
RECOMMENDATIONS.................
11
PANEL
DELIBERATIONS
AND
RESPONSE
TO
CHARGE.................................
12
REFERENCES............................................................................................................
41
4
of
44
SAP
Minutes
No.
2004­
01
A
Set
of
Scientific
Issues
Being
Considered
by
the
Environmental
Protection
Agency
Regarding:

REFINED
(
LEVEL
II)
TERRESTRIAL
AND
AQUATIC
MODELS
PROBABILISTIC
ECOLOGICAL
ASSESSMENTS
FOR
PESTICIDES:
Level
II
Aquatic
Model
Session
APRIL
1
and
2,
2004
FIFRA
Scientific
Advisory
Panel
Meeting,
held
at
the
Crowne
Plaza
Washington­
National
Airport
Hotel,
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:
June
16,
2004
Date:
June
16,
2004
5
of
44
Federal
Insecticide,
Fungicide,
and
Rodenticide
Act
Scientific
Advisory
Panel
Meeting
April
1
and
2,
2004
REFINED
(
LEVEL
II)
TERRESTRIAL
AND
AQUATIC
MODELS
PROBABILISTIC
ECOLOGICAL
ASSESSMENTS
FOR
PESTICIDES
­
Level
II
Aquatic
Model
Session
PARTICIPANTS
FIFRA
SAP,
Session
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
Stuart
Handwerger,
M.
D.,
Director,
Division
of
Endocrinology,
Cincinnati
Children's
Hospital
Medical
Center,
University
of
Cincinnati,
Cincinnati,
OH
Steven
G.
Heeringa,
Ph.
D.,
Research
Scientist
&
Director
for
Statistical
Design,
University
of
Michigan,
Institute
for
Social
Research,
Ann
Arbor,
MI
Gary
E.
Isom,
Ph.
D.,
Professor
of
Toxicology,
School
of
Pharmacy
and
Pharmacal
Sciences,
Purdue
University,
West
Lafayette,
IN
FQPA
Science
Review
Board
Members
Xuefeng
Chu,
Ph.
D.,
Assistant
Professor,
Annis
Water
Resources
Institute,
Grand
Valley
State
University,
Muskegon,
MI
Peter
Delorme,
Ph.
D.,
Senior
Evaluator,
Environmental
Assessment
Division,
PMRA,
Health
Canada,
Ottawa,
ON,
Canada
Philip
Dixon,
Ph.
D.,
Professor,
Statistics,
Iowa
State
University,
Ames,
IA
Paul
W.
Eslinger,
Ph.
D.,
Staff
Scientist,
Pacific
Northwest
National
Laboratory,
Richland,
WA
Christian
Grue,
Ph.
D.,
Associate
Professor
&
Leader,
Washington
Cooperative
Fish
and
Wildlife
Research
Unit,
University
of
Washington,
Seattle,
WA
6
of
44
Chad
Jafvert,
Ph.
D.,
Professor,
School
of
Civil
Engineering,
Purdue
University,
West
Lafayette,
IN
Stephen
J.
Klaine,
Ph.
D.,
Dept.
of
Biological
Sciences,
Graduate
Program
in
Environmental
Toxicology,
Clemson
University,
Pendleton,
SC
Thomas
La
Point,
Ph.
D.,
Professor
and
Director,
Biological
Sciences
and
Institute
of
Applied
Sciences,
University
of
North
Texas,
Denton,
TX
Peter
D.
M.
Macdonald,
D.
Phil.,
Professor
of
Mathematics
and
Statistics,
McMaster
University,
Hamilton,
Ontario,
Canada
Dwayne
Moore,
Ph.
D.,
Cantox
Environmental,
Inc.,
Ottawa,
Ontario,
Canada
Michael
C.
Newman,
Ph.
D.,
Professor
of
Marine
Science,
School
of
Marine
Science,
Virginia
Institute
of
Marine
Science,
College
of
William
&
Mary,
Gloucester
Point,
VA
Gary
M.
Rand,
Ph.
D.,
Associate
Professor,
Department
of
Environmental
Studies
and
Southeast
Environmental
Research
Center,
Ecotoxicological
Laboratory,
Florida
International
University,
Biscayne
Bay
Campus,
Miami,
FL
Geoffrey
Scott,
Ph.
D.,
Acting
Director,
U.
S.
Department
of
Commerce,
NOAA,
National
Ocean
Service,
Center
for
Coastal
Environmental
Health
&
Biomolecular
Research,
Charleston,
SC
Paul
K.
Sibley,
Ph.
D.,
Assistant
Professor,
Department
of
Environmental
Biology,
University
of
Guelph,
Guelph,
Ontario,
Canada
Tammo
S.
Steenhuis,
Ph.
D.,
Professor
of
Watershed
Management,
Department
of
Biological
and
Environmental
Engineering,
Cornell
University,
Ithaca,
NY
7
of
44
INTRODUCTION
The
Federal
Insecticide,
Fungicide,
and
Rodenticide
Act
(
FIFRA),
Scientific
Advisory
Panel
(
SAP)
has
completed
its
review
of
the
set
of
scientific
issues
being
considered
by
the
Agency
pertaining
to
refined
(
Level
II)
aquatic
model
probabilistic
ecological
assessment
for
pesticides.
Advance
notice
of
the
meeting
was
published
in
the
Federal
Register
on
February
20,
2004.
The
review
was
conducted
in
an
open
Panel
meeting
held
in
Arlington,
Virginia,
on
April
1
and
2,
2004.
Dr.
Stephen
M.
Roberts
chaired
the
meeting.
Mrs.
Myrta
R.
Christian
served
as
the
Designated
Federal
Official.

The
FIFRA
SAP
met
to
review
the
Agency's
Level
II
Terrestrial
and
Aquatic
Models
(
Version
2.0).
The
previous
version
of
these
models
was
reviewed
by
the
SAP
during
a
session
held
March
13
­
16,
2001.
The
terrestrial
and
aquatic
models
are
a
key
component
of
the
Agency's
initiative
to
revise
the
ecological
risk
assessment
process,
focusing
on
the
development
of
tools
and
methodologies
to
conduct
probabilistic
ecological
risk
assessments
for
pesticides.

Some
modifications
to
the
models
were
in
response
to
the
2001
SAP
comments
and
recommendations.
Other
modifications
were
based
on
the
suggestions
made
by
the
Ecological
Committee
on
FIFRA
Risk
Assessment
Methods,
a
stakeholder
workgroup
which
provided
recommendations
to
the
Agency
when
this
initiative
first
began.
These
suggestions,
which
were
evaluated
in
the
context
of
the
2001
SAP
review,
were
discussed
within
the
Agency
and
in
national
and
international
scientific
professional
meetings.

The
Agency
was
interested
in
any
general
comments
and
recommendations
from
the
SAP
regarding
the
modifications
to
the
models.
In
addition,
the
Agency
requested
that
the
SAP
respond
to
specific
questions
regarding
the
Terrestrial
and
Aquatic
Level
II
Models.

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.

CHARGE
1.
Varying
Volume
Water
Model
(
VVWM).
For
aquatic
risk
assessments,
OPP
currently
uses
a
water
body
fate
model
that
has
a
fixed
volume
and
does
not
consider
hydrologic
inputs
and
outputs.
The
SAP
2001
suggested
that
adding
volume
variations
and
overflow
to
the
Level
II
fate
model
would
improve
the
characterization
of
the
water
body
and
improve
estimates
of
aquatic
pesticide
concentrations.

In
response,
a
new
model
has
been
developed
that
allows
volume
variations
and
overflow
8
of
44
in
the
water
body.
The
new
model
also
allows
for
meteorologically­
dependent
parameters,
such
as
temperature
and
wind
speed,
to
vary
on
a
daily
basis,
rather
than
a
monthly
basis,
to
better
capture
temporal
variability.
In
addition,
the
model
was
constructed
to
improve
runtime
because
of
the
potential
use
in
Monte
Carlo
simulations.

a.)
Please
discuss
the
new
model's
capability
to
capture
the
most
salient
processes
influencing
the
variations
in
water
body
volume,
and
also
discuss
the
modification
allowing
daily
variations
in
meteorologically­
dependent
variables.
b.)
Inputs
of
mass
on
a
given
day
are
assumed
to
occur
instantaneously.
Please
discuss
the
advantages
and
disadvantages
of
this
assumption
with
specific
consideration
for
the
trade­
off
between
runtime,
accuracy
and
the
consideration
that
input
data
are
given
as
daily
values.
What,
if
any,
additional
approaches
regarding
modeling
input
mass
would
the
SAP
recommend?
Please
provide
a
discussion
of
the
pros
and
cons
as
compared
to
the
current
method.
c.)
What
additional
model
characterization
or
documentation
is
required
to
ensure
clarity
and
transparency?

2.
Exposure
Model
Testing.
The
QA/
QC
testing
of
the
aquatic
Level
II
Version
2.0
exposure
model
demonstrated
that
the
refined
risk
assessment
shell
is
consistent
with
the
Level
II
Version
1.0
shell
(
PE4)
for
launching
PRZM
and
is
compatible
with
all
crop
scenarios
and
meteorological
files.
The
testing
also
showed
that
the
dissipation
algorithms
in
the
VVWM
are
consistent
with
EXAMS
and
that
the
volume
and
overflow
algorithms
are
correct.
Evaluation
of
the
VVWM
showed
the
potential
effect
that
a
varying
volume
water
body,
using
current
standard
field
size
and
water
body
volume
and
surface
area,
can
have
on
estimated
environmental
concentrations
due
to
dilution,
evaporation,
and
overflow.

a.)
What
additional
testing,
evaluation
and/
or
sensitivity
analysis
can
the
SAP
recommend
to
ensure
that
the
aquatic
Level
II
exposure
model
meets
the
Agency
objectives
of
transparent
processes,
and
clear,
consistent
and
reasonable
products
suitable
for
risk
characterization?
b.)
Based
on
the
evaluation
performed
using
the
VVWM
under
standard
field
(
10
ha)
and
standard
surface
water
scenario
conditions
(
1
ha
surface
area,
20,000
m3
volume),
please
discuss
the
advantages
or
disadvantages
to
characterizing
risk
by
replacing
a
single
standard
with
multiple,
crop
scenario­
specific
standards
at
Level
II.

3.
Field
Drainage
Area
and
Water
Body
Size
Selection.
At
Level
II,
the
risk
assessment
approach
is
aimed
at
addressing
the
risk
to
aquatic
species
in
high
exposure,
edge­
of­
field
situations.
The
surrogate
surface
water
used
for
Level
II
consists
of
a
small,
perennial
surface
water
body
at
the
edge
of
an
agricultural
field.
This
water
body
is
capable
of
being
supported
by
agricultural
field
runoff
alone,
and
of
supporting
an
aquatic
community.
Crop
scenario­
specific
input
values
for
field
size,
surface
water
volume,
surface
area,
and
depth
were
developed
and
systematically
explored
using
three
methods.
9
of
44
The
methods
used
readily
available
drainage
area
to
volume
capacity
(
DA/
VC)
ratios
and
associated
water
depth
guidance
for
construction
of
small
permanent
surface
waters
of
the
continental
U.
S.

a.)
The
U.
S.
Department
of
Agriculture's
(
1997)
DA/
VC
ratios
and
depth
guidelines
for
construction
of
small
permanent
water
supplies
(
e.
g.,
irrigation,
livestock,
fish
and
wildlife)
were
used
as
the
source
of
national
and
regional
DA/
VC
ratios
and
associated
water
depths.
What
additional
existing
sources
of
national
or
regional
DA/
VC
ratios
for
small,
permanent
surface
waters
(
e.
g.,
wetlands,
pools,
ponds)
should
be
considered?
b.)
Please
describe
the
merits
or
limitations
to
the
approaches
and
assumptions
evaluated
for
using
the
U.
S.
Department
of
Agriculture's
(
1997)
guidelines
to
derive
field
size,
surface
water
volume,
and
surface
area
input
values
for
specific
crop
scenarios?
What,
if
any,
additional
approaches
and
assumptions
should
be
considered?
c.)
A
default
minimum
depth
was
set
as
0.01
m.
What
minimum
depth
would
the
SAP
recommend
as
a
criterion
to
evaluate
the
biological
relevancy
of
the
scenario?
d.)
Simulations
with
the
PRZM/
VVWM
were
performed
using
both
the
crop­
specific
surface
water
area
and
volume
and
the
historic
standard
values
(
DA/
VC
=
1.5
acres/
acre­
ft)
to
characterize
effect
on
exposure
outputs
for
a
relatively
arid
growing
region
(
DA/
VC
=
50
acres/
acre­
ft)
and
a
wetter
climate
(
DA/
VC
=
1
acre/
acre­
ft)
for
both
a
short­
lived
and
a
long­
lived
pesticide.
In
addition,
the
effect
on
volume
in
the
surface
water
body
was
characterized
for
all
crop­
specific
scenarios.
Please
discuss
what,
if
any,
additional
crop
scenario/
pesticide
evaluations
should
be
performed
to
further
characterize
the
impact
to
exposure
outputs,
and/
or
to
volume.
e.)
What
are
the
advantages
or
disadvantages
to
characterizing
exposure
for
small,
perennial
surface
waters
at
the
edge
of
treated
fields
using
the
method
selected
for
setting
crop
scenario­
specific
DA/
VC
ratio,
depth,
surface
area
and
volume
input
values?
What
adjustments
or
changes
to
the
method
does
the
SAP
recommend,
and
what
are
their
advantages
and
disadvantages?
f.)
Please
describe
the
weaknesses
and
strengths
of
using
simulated
exposure
concentrations
from
these
crop
scenario­
specific
water
bodies
as
a
surrogate
for
a
low­
order
stream
at
the
edge
of
a
field,
for
a
temporary
pool
or
pond,
and
for
a
small
tidal
creek
or
estuary.
g.)
Simulations
with
PRZM/
EXAMS,
a
fixed
volume
surface
water
model,
will
be
performed
using
both
the
crop­
specific
DA/
VC
approach
and
the
historic
standard
values
to
characterize
effect
on
exposure
outputs
for
relatively
arid
growing
regions
(
DA/
VC
=
50
and
80)
and
a
wetter
climate
(
DA/
VC
=
1)
for
both
a
shortlived
and
a
long­
lived
pesticide.
Please
discuss
what,
if
any,
additional
crop
scenario/
pesticide
evaluations
should
be
performed
to
further
characterize
the
impact
to
exposure
outputs
in
a
fixed
volume
situation.
h.)
Please
discuss
sources
or
approaches
for
national
or
regional
DA/
VC
ratios
and
associated
water
depth
and
size
information
for
temporary
pool
and
pond
aquatic­
10
of
44
life
resources.

4.
Curve
Number.
The
SAP
2001
recommended
that
additional
characterizations
of
variability
should
be
given
to
those
parameters
in
the
exposure
model
that
have
a
major
impact
on
exposure
concentrations.
The
curve
number
is
perhaps
the
most
influential
parameter
in
PRZM,
and
it
has
been
interpreted
in
recent
literature
as
a
random
variable.
PRZM
currently
treats
the
curve
number
as
a
function
of
soil
moisture,
although
recent
literature
suggests
that
the
curve
number
may
more
appropriately
be
interpreted
as
a
random
variable.

a.)
Please
discuss
the
pros
and
cons
of
assuming
strict
dependence
of
curve
number
on
calculated
soil
moisture
versus
treatment
as
a
random
variable
unrelated
to
soil
moisture
as
a
means
of
characterizing
runoff
variability.
Please
identify
and
discuss
alternative
methods.
b.)
Since
the
curve
number
was
not
designed
for
use
in
continuous
modeling,
what
problems
may
arise
when
the
curve
number
is
used
in
this
manner?
Could
a
probabilistic
interpretation
address
some
of
these
issues?
If
so,
how?
c.)
What
is
the
impact
on
interpretation
of
probabilistic­
simulated
exposure
values
when
the
curve
number
is
used
as
a
random
variable
and
autocorrelation
of
temporally­
varying
physical
properties
that
may
impact
runoff
is
ignored?
d.)
A
lognormal
distribution
is
being
investigated
to
characterize
variability
in
certain
curve
number
parameters.
Is
it
reasonable
to
assume
such
a
distribution
has
stationary
properties
(
constant
mean
and
variance)
for
all
rain
events
(
e.
g.,
large
and
small)?
Please
provide
rationale.
e.)
Monte
Carlo
modeling
is
being
investigated
as
a
method
of
integrating
the
potential
variability
of
curve
numbers
into
exposure
modeling.
Can
the
SAP
recommend
other
methods
available
to
incorporate
variable
and
uncertain
curve
numbers
into
a
continuous
runoff
model?
Please
discuss
the
pros
and
cons
of
these
methods
versus
Monte
Carlo.
11
of
44
SUMMARY
OF
PANEL
DISCUSSION
AND
RECOMMENDATIONS
The
FIFRA
SAP
reviewed
the
Agency
Document
and
made
suggestions
for
the
Aquatic
(
Level
II)
model.
Additional
related
issues
are
also
noted.
Below
is
a
summary
of
the
Panel's
findings
and
recommendations.

1.
The
Panel
commended
the
Agency
for
the
initiative
of
developing
a
methodology
and
tool
for
refining
the
Level
II
aquatic
probabilistic
ecological
assessments
for
pesticides.
The
Varying
Volume
Water
Model
(
VVWM)
with
daily
varying
input
parameters
is
an
important
advance
over
the
constant
volume
EXAMS
model
using
monthly
averaged
data.
The
degree
of
complexity
included
in
the
conceptualization
of
the
field­
pond
system
is
appropriate
for
Level
II
analysis.
The
temporal
variations
in
water
level
and
volume
and
the
resultant
pesticide
concentrations
appear
to
be
reasonably
predicted.
The
model
captures
the
important
hydrological
and
pesticide
fate
processes
and
appears
to
give
reasonable
and
realistic
predictions
and
refined
estimates
of
exposure.
However,
implementation
of
the
hydrology
and
fate
processes
varies
regionally.

2.
Verification
procedures
to
ensure
that
the
VVWM
code
correctly
replicates
the
corresponding
code
in
EXAMS
have
been
well
thought
out.
The
Panel
noted,
however,
that
it
would
be
useful
to
know
what
QA/
QC
procedures
were
previously
undertaken
to
ensure
that
the
EXAMS
code
is
correct.

3.
Additional
sensitivity
analyses
are
needed.
Given
the
large
number
of
input
parameters
for
the
VVWM
model,
formal
sensitivity
analyses
to
identify
those
variables
that
most
influence
the
results
are
needed.
Such
analyses
should
consider
the
relative
influence
of
the
standardized
inputs.

4.
There
was
general
agreement
that
a
regional
approach
was
needed
for
defining
DA/
VC
and
pond
depths.
The
watershed
size
chosen
by
EPA
of
0.1
to
1.0
km2
falls
into
the
most
vulnerable
watersheds
for
pesticide
contamination.
Sources
for
regionally
characterizing
types
of
ponds
and
their
locations
in
the
landscape
were
identified.

5.
The
merits
of
including
subsurface
flow
for
Level
II
risk
assessment
models
were
discussed.
The
most
immediate
effect
of
the
presence
of
groundwater
is
dampening
the
fluctuations
in
pond
water
elevations.
Neglecting
ground
water
pesticide
inputs
was
appropriately
conservative
for
Level
II
analysis,
with
the
possible
exception
of
cases
where
tile
drainage
lines
play
a
significant
role
in
watershed
transport.

6.
Additional
modeling
scenarios
for
running
the
Level
II
risk
assessment
should
be
considered.
For
a
comprehensive
evaluation
to
determine
which
of
the
many
12
of
44
variables
are
important,
the
use
of
appropriate
experimental
designs
(
e.
g.
Kleijnen
2004)
is
recommended.

7.
The
Agency
is
commended
for
attempting
an
innovative
solution
to
the
difficult
problem
of
simulating
runoff.
It
is
possible
to
spend
considerable
time
and
energy
on
a
detailed
infiltration
model
based
on
physical
principles.
While
such
a
model
might
be
appropriate
for
Level
III
or
IV
assessments,
it
is
not
clear
that
such
a
model
is
needed
at
Level
II.

8.
Panel
members
felt
that
EPA
should
consider
modifying
the
code
to
include
physical
relationships
(
CN
linkage
to
important
physical
parameter(
s))
and
probabilistic
aspects.
The
Panel
proposed
an
alternate
probabilistic
approach
that
will
aid
in
fusing
physical
and
probabilistic
issues.

9.
Independent
of
which
approach
is
used,
the
final
model
should
be
tested
under
a
wide
range
of
conditions
(
different
catchments
sizes,
size
and
intensity
of
rain
events,
etc.)
in
order
to
adequately
account
(
if
possible)
for
the
unexplained
sources
of
variability.
The
final
code
itself
should
be
well
documented
and
published.

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
1.
Varying
Volume
Water
Model
(
VVWM).
For
aquatic
risk
assessments,
OPP
currently
uses
a
water
body
fate
model
that
has
a
fixed
volume
and
does
not
consider
hydrologic
inputs
and
outputs.
The
SAP
2001
suggested
that
adding
volume
variations
and
overflow
to
the
Level
II
fate
model
would
improve
the
characterization
of
the
water
body
and
improve
estimates
of
aquatic
pesticide
concentrations.

In
response,
a
new
model
has
been
developed
that
allows
volume
variations
and
overflow
in
the
water
body.
The
new
model
also
allows
for
meteorologically
dependent
parameters,
such
as
temperature
and
wind
speed,
to
vary
on
a
daily
basis,
rather
than
a
monthly
basis,
to
better
capture
temporal
variability.
In
addition,
the
model
was
constructed
to
improve
runtime
because
of
the
potential
use
in
Monte
Carlo
simulations.

a.)
Please
discuss
the
new
model's
capability
to
capture
the
most
salient
processes
influencing
the
variations
in
water
body
volume,
and
also
discuss
the
modification
allowing
daily
variations
in
meteorological
dependent
variables.
13
of
44
Response
The
Panel
agreed
that
the
VVWM
with
daily
varying
input
parameters
is
an
important
advance
over
the
constant
volume
EXAMS
model
using
monthly
averaged
data.
The
degree
of
complexity
included
in
the
conceptualization
of
the
field­
pond
system
is
appropriate
for
Level
II
analysis.
The
temporal
variations
in
water
level
and
volume
and
the
resultant
pesticide
concentrations
appear
to
be
reasonably
predicted.
The
model
captures
the
important
hydrological
and
pesticide
fate
processes
and
appears
to
give
reasonable
and
realistic
predictions
and
refined
estimates
of
exposure.
However,
implementation
of
the
hydrology
and
fate
processes
varies
regionally.
Whether
the
model
captures
the
most
salient
processes
depends
on
the
similarity
between
the
conceptual
model
and
the
real
system.
Special
attention
should
be
paid
to
the
simplifying
assumptions
when
the
model
is
used
for
any
real
world
applications.

The
Panel
had
several
comments
on
the
model
structure
and
suggested
possible
refinements.
These
are
presented
for
the
following
areas:

Pond
Inflow
In
the
VVWM,
surface
runoff
is
considered
as
the
primary
water
and
pesticide
source
for
the
surface
water
body
adjacent
to
the
field.
The
Panel
is
supportive
of
the
daily
time
step
calculations
of
surface
runoff
for
the
Level
II
model.
Currently,
the
Agency
uses
the
SCS
curve
number
method
(
in
PRZM)
for
runoff
estimation.
The
curve
number
method
is
essentially
an
empirical,
event­
based
approach.
Thus,
caution
should
be
used
when
the
curve
number
method
is
used
for
runoff
simulation.
Detailed
comments
are
given
in
the
response
to
question
4.

In
arid
regions,
irrigation
is
the
primary
water
source
for
growing
crops.
Incorporating
irrigation
is
critical
to
pesticide
fate,
transport
and
risk
assessment.
Moreover,
runoff
of
irrigation
water
provides
inflow
to
the
pond
and
maintains
the
water
level
whereas
under
rainfall­
only
conditions
in
arid
regions
the
pond
might
evaporate
completely.
PRZM
is
capable
of
simulating
irrigation.
The
Panel
recommends
that
EPA
include
water
addition
by
irrigation
in
the
Level
II
model
pesticide
risk
assessment.

In­
Pond
Processes
One
of
the
most
significant
modifications
from
EXAMS
to
VVWM
is
the
way
that
pesticide
movement
across
the
sediment­
water
interface
is
computed.
The
resulting
ingenious
analytical
solution
is
ideally
suited
for
running
Monte
Carlo
simulations.
However,
the
explanation
given
(
in
the
March
4,
2004
document)
as
a
mass
transfer
process
between
the
littoral
and
benthic
zones
can
be
more
realistically
explained
as
a
mixing
process.
Note
that
the
term
describing
the
rate
of
pesticide
transfer
between
the
two
zones,
A
×
D/ 
x
(
pages
36­
37
in
chapter
IV
of
the
March
4,
2004
Document),
has
14
of
44
units
of
volume
per
time.
This
is
exactly
like
a
mixing
term
where,
on
a
daily
basis,
a
fraction
of
the
volume
of
one
compartment
is
added
to
the
other
and
vice
versa.
For
two
water
compartments,
obviously,
this
is
a
turnover
rate,
where
the
same
volume
is
transferred
between
the
two
compartments
at
each
time
interval,
and
this
exchanged
volume
is
then
completely
mixed
into
the
new
compartment.
This
would
be
similar
for
the
mixing
of
sediments.
The
mixing
of
each
phase
(
water
into
water,
sediment
into
sediment)
is
likely
driven
by
biological
activity.
Rather
than
view
this
as
a
diffusional
flux
requiring
a
characteristic
length
(
as
is
true
for
the
case
of
large,
compartmentalized
lakes),
this
should
be
viewed
as
a
turnover
rate.
The
question
is,
how
frequently
does
the
water
in
the
first
5
cm
of
sediment
`
turn
over'
to
the
littoral
zone
due
to
biological
activity?
Any
water
pumped
out
of
the
sediment
(
which
includes
excretion
from
worms)
must
be
replaced
with
water
from
the
littoral
zone.
The
selection
of
5
cm
as
the
depth
of
the
benthic
layer
is
consistent
with
the
biologically­
active
layer
in
sediments.

Pond
Overflow
Overflow
from
the
pond
is
an
important
addition
to
the
new
model
and
results
in
predictions
of
persistent
pesticide
concentrations
in
the
pond
water
that
are
more
realistic
than
those
of
the
EXAMS
model.
It
is
also
an
important
improvement
for
better
prediction
of
downstream
(
spatial)
effects.
One
concern
is
that
the
simulated
exposure
concentrations
(
especially
for
persistent
pesticides)
overly
depend
on
the
selection
of
the
shape
and
volume
of
the
surface
water
body.
The
Panel
recommends
that
the
Agency
consider
realistic
regional
variable
pond
depths
and
adjust
the
shape
to
more
appropriate
(
simple)
geometric
forms.
More
details
are
given
in
the
response
to
question
3.

Ground
Water
and
Surface
Water
Interaction
The
Panel
unanimously
agreed
that
the
groundwater
and
surface
water
interaction
is
an
important
issue.
In
some
regions,
subsurface
flow
and
its
impact
on
pesticide
exposure
levels
in
the
surface
water
body
can
be
significant.
Subsurface
flow,
especially
from
tile
drainage
lines,
may
contain
high
levels
of
pesticides
and
may
result
in
inputs
over
longer
periods
of
time.
To
enhance
the
capability
of
handling
a
wide
range
of
real­
world
problems
across
the
United
States,
the
groundwater­
surface
water
(
GW­
SW)
interactions
should
be
taken
into
account,
even
if
only
at
a
qualitative
level.
Indeed,
considering
the
complexity
of
this
issue
(
tremendous
modeling
efforts
are
required
to
characterize
the
dynamic
interaction
between
surface
and
subsurface
water
systems),
most
of
the
Panel
members
recommended
that
this
issue
be
addressed
in
the
models
at
Levels
III
and
IV.

The
Panel
suggested
that,
for
conditions
where
groundwater
determines
the
water
table
height,
rather
than
assume
that
all
excess
volume
above
the
design
maximum
goes
to
outflow
at
the
end
of
each
day,
it
would
be
preferable
to
assume
that
this
excess
volume
decays
to
the
design
maximum
through
a
simple
first­
order
process
with
a
half­
life
on
the
order
of
3
days
to
1
week.
15
of
44
b.)
Inputs
of
mass
on
a
given
day
are
assumed
to
occur
instantaneously.
Please
discuss
the
advantages
and
disadvantages
of
this
assumption
with
specific
consideration
for
the
trade
off
between
runtime,
accuracy
and
the
consideration
that
input
data
are
given
as
daily
values.
What,
if
any,
additional
approaches
regarding
modeling
input
mass
would
the
SAP
recommend?
Please
provide
a
discussion
of
the
pros
and
cons
as
compared
to
the
current
method.

Response
The
model
is
computationally
efficient
because
analytical
solutions
were
found
for
the
differential
equations
describing
the
processes
in
the
pond.
To
obtain
these
analytical
solutions,
individual
pond
processes
are
assumed
to
be
at
steady
state
and
mass
inputs
to
the
pond
are
instantaneous.
Instead
of
instantaneous
inputs,
continuous
daily
varying
steady
state
application
could
be
used,
but
no
significant
differences
are
expected.
The
Panel
endorsed
the
current
Level
II
modeling
approach
involving
daily
instantaneous
inputs,
especially
because
instantaneous
mass
loadings
likely
are
to
be
more
conservative
in
the
risk
assessment.

In
the
VVWM,
additions
of
the
soil
itself
 
originating
from
erosion
from
the
field
to
the
pond
 
are
neglected.
Pesticide
masses
adsorbed
to
the
soil
are
distributed
evenly
and
instantaneously
to
the
littoral
and
benthic
zone
(
50%
each).
The
Panel
would
like
to
see
additional
justification
for
this
assumption.
Some
Panel
members
suggested
the
need
for
sediment
transport
modeling
in
the
water
column,
and
interaction
between
the
water
column
and
the
benthic
layer
by
quantitatively
simulating
processes
such
as
settling,
resuspension,
and
sedimentation.
Most
Panel
members
did
not
think
this
is
necessary
for
a
Level
II
risk
assessment
model
and
that
physically­
based
modeling
of
sediment
transport
should
be
considered
only
in
Level
III
or
IV
risk
assessment
models.

c.)
What
additional
model
characterization
or
documentation
is
required
to
ensure
clarity
and
transparency?

Response
While
the
Panel
is
impressed
with
the
significant
progress
in
modeling
Level
II
exposure
risk
assessment,
at
the
same
time
it
acknowledges
that
further
extensions
can
be
made
to
the
current
VVWM.
These
are
presented
in
the
following
areas:

Probabilistic
distribution
of
input
parameters
The
VVWM
can
easily
incorporate
probabilistic
variables,
as
is
the
case
with
other
modules
of
the
Level
II
aquatic
model.
EPA
has
indicated
that
probability
distributions
are
being
developed
for
key
input
parameters
for
future
iterations
of
the
VVWM.
These
distribution
inputs
would
be
a
good
improvement.
It
is
important
that
this
effort
not
be
restricted
to
the
curve
number
only;
other
input
variables
are
equally
important.
To
16
of
44
improve
the
computational
efficiency
in
the
Monte
Carlo
simulations,
consideration
should
be
given
to
moving
PRZM
to
a
faster
platform.

Refinement
of
modeling
processes
The
modeling
system
involves
simulations
of
both
water
flow
and
pesticide
fate
and
transport
in
the
crop
field
and
the
adjacent
surface
water
body
(
pond).
The
final
exposure
levels
of
pesticides
are
affected
by
a
number
of
physical
and
biochemical
processes
that
may
vary
both
spatially
and
temporally.
For
some
processes
such
as
photolysis
and
metabolism,
separation
into
daytime
and
nighttime
segments
(
i.
e.,
12­
hour
periods)
could
be
important.
Seasonal
variations
in
solar
intensity
could
be
important
too.

Other
refinements
mentioned
were:

 
pH­
dependent
hydrolysis
for
some
chemicals.

 
Pesticide
mass­
balance
for
the
benthic
zone.
No
toxicity
is
currently
expressed
in
this
layer;
however,
without
reporting
concentrations
for
this
layer,
the
analysis
is
incomplete.

 
Freezing
effects.
On
the
small
permanent
surface
water
body,
ice
formation
is
not
simulated
in
VVWM.
In
PRZM,
snow
melt
is
included.

 
Impervious
field
conditions.
Impervious
plastic
(
mulch)
culture
affects
the
rate
of
runoff,
the
amount
of
sediment
scour,
and
the
frequency
with
which
runoff
events
occur,
particularly
for
lower
intensity
rainfall
events.

Documentation
Several
Panel
members
expressed
the
need
for
documentation
of
model
testing
with
fieldobserved
data
and
documentation
of
sensitivity
analyses.
Details
are
given
in
the
response
to
question
2
and
in
"
Additional
General
Comments
from
the
SAP."

Major
assumptions
should
be
listed
in
table
form
for
clarity.
Model
application
conditions
and
limitations
should
be
described.
For
example,
if
the
water
depth
is
large,
the
assumption
of
complete
mixing
may
not
be
applicable.

Future
Improvements
For
Levels
III
and
IV,
it
would
be
useful
to
have
the
capability
of
modeling
different
types
of
water
bodies
(
e.
g.,
small,
low­
order
perennial
streams
receiving
multiple
inputs
from
adjacent
fields),
as
is
often
the
case
in
the
real
world.
It
would
also
be
useful
to
incorporate
the
influence
of
geometry
and
water
quantity/
quality
interactions
between
groundwater
and
surface
water.
17
of
44
Without
simulating
settling,
sediments
eroded
from
the
crop
field
might
be
accumulated
in
the
water
column
under
certain
conditions.
Due
to
settling,
contaminated
sediments
can
be
buried
and
are
thus
no
longer
available
for
the
mass
exchange
between
the
two
zones.
Thus,
settling
and
sedimentation
can
be
very
important
for
a
standing
surface
water
body.
It
is
recommended
that
these
processes
be
addressed
in
higher
level
models.

Summary
In
summary,
there
are
a
number
of
processes
and
factors
that
should
be
considered
or
clarified.
Considering
data
availability,
however,
the
Panel
agreed
that
some
processes,
such
as
GW­
SW
interactions,
need
to
be
considered
quantitatively
only
at
a
higher
level
(
Level
III
and
IV),
although
these
processes
should
be
considered
qualitatively
when
defining
certain
variables
in
the
Level
II
model,
such
as
minimum
volume.
The
decision
of
which
processes
to
consider
further
at
Level
II
should
include
consideration
of
whether
each
makes
the
model
more
or
less
conservative.
If
ignoring
a
process
will
lead
to
underestimation
of
exposure,
then
the
process
should
be
considered
at
Level
II.

2.
Exposure
Model
Testing.
The
QA/
QC
testing
of
the
aquatic
Level
II
Version
2.0
exposure
model
demonstrated
that
the
refined
risk
assessment
shell
is
consistent
with
the
Level
II
Version
1.0
shell
(
PE4)
for
launching
PRZM
and
is
compatible
with
all
crop
scenarios
and
meteorological
files.
The
testing
also
showed
that
the
dissipation
algorithms
in
the
VVWM
are
consistent
with
EXAMS
and
that
the
volume
and
overflow
algorithms
are
correct.
Evaluation
of
the
VVWM
showed
the
potential
effect
that
a
varying
volume
water
body,
using
current
standard
field
size
and
water
body
volume
and
surface
area,
can
have
on
estimated
environmental
concentrations
due
to
dilution,
evaporation,
and
overflow.

a.)
What
additional
testing,
evaluation
and/
or
sensitivity
analysis
can
the
SAP
recommend
to
ensure
that
the
aquatic
Level
II
exposure
model
meets
the
Agency
objectives
of
transparent
processes,
and
clear,
consistent
and
reasonable
products
suitable
for
risk
characterization?

Response
Verification
The
Panel
agreed
with
the
procedures
used
to
ensure
that
the
VVWM
code
correctly
replicates
the
corresponding
code
in
EXAMS.
It
appears
to
have
been
well
thought
out
and
indicates
that
the
VVWM
is
operating
correctly,
as
evidenced
by
the
lack
of
differences
in
the
output
from
EXAMS
and
VVWM
for
water
concentrations.
Whereas
the
side­
by­
side
testing
showed
that
water
concentrations
were
consistent
between
EXAMS
and
the
VVWM,
no
data
were
shown
for
sediment
or
pore­
water
concentrations
predicted
by
EXAMS.
The
Panel
received
verbal
confirmation
from
Agency
staff
that
18
of
44
sediment
and
pore­
water
concentrations
were
replicated
by
the
VVWM.
The
PE4
or
RRA
shell
launched
PRZM
successfully.
The
few
minor
problems
with
various
input
parameters
from
the
standard
scenarios
were
identified
and
corrected,
or
necessary
code
modifications
were
made.
The
Panel
noted,
however,
that
while
everything
appears
to
function
correctly,
the
procedures
assume
that
the
original
code
in
EXAMS
is
correct.
It
would
be
useful
to
know
what
QA/
QC
procedures
were
previously
undertaken
to
ensure
that
the
EXAMS
code
is
correct.
In
addition,
while
recognizing
that
the
model
is
still
in
the
development
and
testing
phase,
the
Panel
recommended
that
at
some
point
the
code
should
be
disclosed
and
a
code
audit
undertaken
to
ensure
its
integrity.

In
response
to
a
request
from
EPA
staff
for
references
on
the
statistical
design
of
a
sensitivity
study,
the
following
references
are
suggested
as
a
starting
point:
Kleijnen,
J.
P.
C.
(
1997
and
2004);
Sacks,
J.,
W.
J.
Welch,
T.
J.
Mitchell,
and
H.
P.
Wynn
(
1989).

Sensitivity
Analysis
Panel
members
agreed
that
additional
sensitivity
analyses
are
needed.
Although
a
limited
number
of
sensitivity
analyses
were
presented,
they
were
all
"
inward­
looking"
with
respect
to
the
variable
volume
water
model.
In
this
respect,
the
Agency
was
cautioned
that
the
sensitivity
analysis
conducted
to
determine
the
influence
of
maximum
pond
depth
(
Dmax)
on
concentration
indicated
that
variation
in
Dmax
had
a
relatively
minor
effect
on
average
daily
concentration,
particularly
for
high
precipitation
areas.
However,
the
analyses
only
varied
Dmax
slightly
(
2.44
to
3.05
m
in
California,
1.83
to
2.13
m
in
Florida).
Thus,
pesticide
concentrations
could
vary
quite
substantially
between
systems
with
widely
varying
maximum
depths.

Given
the
large
number
of
input
parameters
for
the
VVWM
model,
formal
sensitivity
analyses
are
needed
to
identify
those
variables
that
most
influence
the
results.
Such
analyses
should
consider
the
relative
influence
of
the
inputs
that
are
standardized
(
e.
g.
fraction
of
organic
carbon,
light
attenuation
factor,
benthic
dispersion
coefficient,
boundary
layer
thickness,
O2
exchange
coefficient,
etc.).
Many
of
these
standardized
input
values
are
based
on
empirical
data,
but
the
variability
associated
with
them
is
not
discussed
and
their
effects
on
model
outcomes
appear
to
be
unknown.
The
need
for
further
examination
of
"
standard"
input
values
was
referred
to
in
the
document
as
an
area
needing
review
but
was
not
identified
in
the
ongoing/
future
activities.

Specific
suggestions
were
made
regarding
sensitivity
analyses
for
several
parameters
in
the
model.
These
parameters
included
pH,
biomass
and
total
suspended
solids
(
TSS).
For
example
it
was
suggested
that
TSS
could
range
from
10
to
20,000
mg/
L.
This
is
important
for
highly
erodible
soils
such
as
Loess
soils
(
which
are
found
in
western
Tennessee
and
elsewhere)
and
would
affect
strongly­
sorbed
(
high
Koc)
chemicals.
At
high
TSS,
the
consequences
of
mass
transport
could
be
significant
even
for
pesticides
with
low
Koc
values.
Ultimately,
the
suspended
solids
load
could
be
generated
from
the
mass
of
soil
identified
in
PRZM
output
(
maybe
at
Level
III?).
The
pH
was
identified
in
the
19
of
44
document
as
a
candidate
parameter
for
further
examination
of
its
influence
on
model
results.
pH
was
identified
by
the
SAP
in
2001
as
being
a
parameter
that
should
be
varied
as
it
can
have
a
direct
influence
on
hydrolysis
rates
and
on
dissociation
of
certain
compounds.
While
there
may
be
merit
in
varying
this
on
a
scenario
basis,
there
may
also
be
merit
in
considering
using
a
distribution
or
using
a
minimum/
maximum/
mean
approach
in
those
instances
where
hydrolysis
or
dissociation
is
affected
by
pH
across
the
normal
range
found
in
aquatic
systems.
One
possible
approach
would
be
to
include
this
at
Level
II,
but
hold
constant
in
Level
I.
This
type
of
approach
could
be
used
for
several
other
influential
parameters.

In
addition,
sensitivity
analyses
for
compound­
specific
parameters,
i.
e.,
physical
and
chemical
properties
and
degradation
rates,
were
suggested.
This
should
include
determining
the
influence
of
varying
Kd/
Koc
on
the
output.
Both
the
short­
lived
and
persistent
chemicals
that
were
used
for
some
of
the
QA/
QC
work
had
similar
Koc
values
(
487
vs.
422).
Because
this
parameter
can
have
a
major
effect
on
eventual
partitioning
and
fate,
particularly
in
the
variable
volume
scenarios,
there
needs
to
be
assurance
that
the
partitioning
is
consistent
between
the
two
models.

Sensitivity
analysis
was
also
suggested
to
examine
the
influence
of
assumptions
made
with
regard
to
dynamic
processes
in
the
model.
It
was
noted
that
particulate
surface­
or
humic
acid­
catalyzed
hydrolysis
was
assumed
not
to
occur.
A
simple
analysis
toggling
this
factor
on
or
off
could
provide
insight
into
its
importance.

Additional
evaluation
of
the
approach
for
determining
the
appropriate
DA/
VC
should
be
considered.
According
to
the
USDA
(
USDA
1997),
the
values
for
DA/
VC
can
vary
greatly
in
a
local
area
when
drainage
areas
have
unique
characteristics.
USDA
recommends
reducing
DA/
VC
values
by
as
much
as
25
percent
for
drainage
areas
having
extreme
runoff­
producing
characteristics
and
increasing
them
by
50
percent
or
more
for
low
runoff­
producing
characteristics.

Risk
Assessment
The
Panel
acknowledges
that
VVWM
intended
for
use
in
a
Level
II
assessment
is
still
relatively
conservative.
It
is
suggested
that
the
Agency
consider
conducting
case
studies
to
look
at
the
effect
of
using
the
VVWM
and
using
a
regional
scenario­
based
approach
on
risk
assessment
conclusions.

The
receiving
water
body
scenario
has
been
derived
from
information
on
pond
construction
from
USDA.
However,
an
analysis
to
relate
receiving
water
scenarios
to
vulnerable
aquatic
ecosystems
is
needed.
This
can
be
done
by
comparing
hydrologic
parameters
(
i.
e.,
area,
depth,
and
volume)
of
real
ecosystems
with
the
basic
parameters
of
the
pond
characterizations
derived
from
the
USDA
Handbook.
A
more
difficult
task
will
be
to
characterize
the
drainage
area.
There
seems
to
be
a
paucity
of
data
for
these
types
of
receiving
environments.
This
could
help
to
move
away
from
the
misconception
that
a
20
of
44
farm
pond
is
being
simulated
rather
than
a
more
relevant
natural
ecosystem
such
as
a
wetland.

From
the
perspective
of
the
overall
risk
assessment
process,
the
Agency
may
want
to
compare
results
from
the
VVWM
regional
scenarios
with
GENEEC
for
a
range
of
chemicals
to
determine
if
there
are
some
which
result
in
higher
concentrations
from
the
VVWM.
If
there
are,
several
options
might
be
available,
including
dropping
the
use
of
GENEEC
in
Level
I,
Tier
I,
and
replacing
this
model
with
the
peak
values
from
PRZM/
VVWM
for
the
scenario
which
consistently
results
in
the
highest
runoff
concentrations.

Comparisons
with
Observed
Data
Given
the
flexibility
that
VVWM
now
has,
it
would
seem
relatively
easy
to
obtain
existing
pesticide
monitoring
data
sets
from
edge­
of­
field
ponds
and
compare
these
monitoring
data
results
to
VVWM
model
predictions
derived
from
comparable
site­
specific
scenarios.
With
this
information,
model
performance
could
be
determined.
If
possible,
model
performance
should
be
estimated
for
a
range
of
regions
and
crop
scenarios
to
determine
how
well
the
model
performs
under
different
scenarios.
Alternatively,
evaluation
of
model
performance
could
be
conducted
under
controlled
situations
or
in
selected
watersheds.

One
Panel
member
noted
that
EPA
already
has
data
submitted
by
registrants
which
could
be
useful
for
both
characterizing
the
receiving
water
body
morphometry
and
for
examining
model
performance.
The
Agency
may
want
to
evaluate
data
generated
from
farm
pond
studies
and
from
constructed
mesocosm
and
microcosm
studies
conducted
by
registrants
for
the
EPA
between
the
late
1970s
and
the
mid­
1990s.
Pesticide
registrants
conducted
these
studies
to
satisfy
data
requirements
for
ecological
effects
rather
than
exposure.
However,
they
may
contain
relevant
information/
data
that
would
be
useful
to
benchmark
(
ground
truth)
the
models
discussed.
Farm
pond
studies
should
be
available
from
the
late
1970s
to
the
late
1980s
and
mesocosm/
microcosm
studies
from
the
1980s
through
the
1990s.
Farm
pond
studies
were
conducted
with
ponds
(
1
to
3
acres
in
surface
area)
with
watersheds
approximately
10
or
more
times
larger.
One
Panel
member
was
aware
of
farm
pond
studies
submitted
for
a
range
of
compounds,
including
synthetic
pyrethrins,
endosulfan
and
organophosphates.
Parameters
measured
included
(
but
were
not
limited
to):
meteorological
conditions,
pond
morphometry
(
for
farm
ponds,
depth,
area)
and
chemical
concentrations.
Pesticide
application
in
these
types
of
studies
was
done
according
to
typical
label
application
practices
for
the
crop
and
chemical
combination.
Edge­
of
field
surface
water
runoff
(%)
values
were
also
generated
in
these
pond
studies.
Mesocosm
and
microcosm
studies
were
conducted
with
a
host
of
pyrethrins,
organophosphates
and
herbicides.
Constructed
mesocosms
were
up
to
0.25
acres
in
surface
area
and
six
feet
deep,
with
similar
parameter
measurements
as
farm
ponds.
Meteorological
conditions
are
likely
available
for
some
of
the
mesocosm
studies
conducted
in
the
midwest,
southwest
and
southeast
parts
of
the
U.
S.
21
of
44
Although
the
model
construct
may
not
represent
the
reality
of
hydrologic
and
other
physical/
mechanistic
processes,
it
is
a
tool
to
estimate
environmental
concentrations.
It
is
important
that
the
model
and
the
associated
scenarios
reasonably
represent
the
concentrations
likely
to
occur
in
the
environment
and
meet
the
needs
of
the
conceptual
model
for
a
Level
II
assessment.
Additional
refinement
of
the
model
and
scenarios
will
be
necessary
for
Level
III
assessments.

b.)
Based
on
the
evaluation
performed
using
the
VVWM
under
standard
field
(
10
ha)
and
standard
surface
water
scenario
conditions
(
1
ha
surface
area,
20000
m3
volume),
please
discuss
the
advantages
or
disadvantages
to
characterizing
risk
by
replacing
a
single
standard
with
multiple,
crop
scenario­
specific
standards
at
Level
II.

Response
Advantages
of
Multiple
Crop­
Scenario
Approach
 
This
approach
can
be
used
to
better
characterize
risk
on
a
regional
basis,
thereby
allowing
the
Agency
to
focus
its
assessment
efforts
on
the
crops
and/
or
areas
where
potential
problems
are
identified.

 
This
approach
recognizes
that
there
are
regional
differences
in
water
bodies
and
rainfall
patterns
and
gives
the
Agency
the
ability
to
account
for
regional
differences.

 
If
peak
values
from
PRZM/
VVWM
are
incorporated
into
Level
I,
then
this
information
could
identify
which
regions
and/
or
crops
might
need
additional
refinement
in
a
Level
II
assessment.

 
The
approach
for
Level
II
should
be
to
have
regional
representations
of
ponds
and
surface
water
scenarios.
Otherwise,
the
present
approach
is
good.
There
are
advantages
in
having
a
probabilistic
statement
that
can
be
made
over
a
wider
aspect
of
sizes
(
and
maybe
shapes)
of
ponds.
Shallow
systems
can
also
be
biologically
important
for
macroinvertebrates,
amphibians,
etc.

 
There
is
potential
for
improved
ability
to
evaluate
model
performance,
because
the
VVWM
model
can
be
tailored
to
match
scenarios
for
which
monitoring
data
have
been
collected.

 
The
Agency
should
examine
the
effect
on
sediment/
pore
water
concentrations
for
the
different
crop­
scenarios
using
the
VVWM.

Disadvantages
of
Multiple
Crop­
Scenario
Approach
22
of
44
 
The
results
of
risk
assessments
may
be
more
challenging
to
communicate
to
risk
managers.

 
A
possible
disadvantage
is
the
potential
for
increased
resource
requirements.
However,
given
the
cost
of
the
resources
to
be
protected,
it
may
be
worth
the
time
and
effort.

3.
Field
Drainage
Area
and
Water
Body
Size
Selection.
At
Level
II,
the
risk
assessment
approach
is
aimed
at
addressing
the
risk
to
aquatic
species
in
high
exposure,
edge­
offield
situations.
The
surrogate
surface
water
used
for
Level
II
consists
of
a
small,
perennial
surface
water
body
at
the
edge
of
an
agricultural
field.
This
water
body
is
capable
of
being
supported
by
agricultural
field
runoff
alone,
and
of
supporting
an
aquatic
community.
Crop
scenario­
specific
input
values
for
field
size,
surface
water
volume,
surface
area,
and
depth
were
developed
and
systematically
explored
using
three
methods.
The
methods
used
readily
available
drainage
area
to
volume
capacity
(
DA/
VC)
ratios
and
associated
water
depth
guidance
for
construction
of
small
permanent
surface
waters
of
the
continental
U.
S.

a.)
The
U.
S.
Department
of
Agriculture's
(
1997)
DA/
VC
ratios
and
depth
guidelines
for
construction
of
small
permanent
water
supplies
(
e.
g.,
irrigation,
livestock,
fish
and
wildlife)
were
used
as
the
source
of
national
and
regional
DA/
VC
ratios
and
associated
water
depths.
What
additional
existing
sources
of
national
or
regional
DA/
VC
ratios
for
small,
permanent
surface
waters
(
e.
g.,
wetlands,
pools,
ponds)
should
be
considered?

Response
There
was
general
agreement
that
a
regional
approach
was
needed
for
defining
DA/
VC
and
pond
depths.
The
watershed
size
chosen
by
EPA
of
0.1
to
1.0
km2
falls
into
the
most
vulnerable
watersheds
for
pesticide
contamination
(
Schulz,
2004).
This
is
especially
important
for
coastal
wetlands
where
tidal
range,
width
or
expanse
of
wetlands
in
terms
of
buffer
size,
and
groundwater
considerations
must
be
taken
into
account
on
a
regional
basis.
Other
important
regional
differences
that
need
to
be
taken
into
account
are
the
depths
to
groundwater.

The
following
sources
can
be
used
to
obtain
regional
information
for
types
of
ponds
and
their
location
in
the
landscape:

°
NOAA's
Oil
Spill
Environmental
Sensitivity
Index
(
ESI)
mapping
which
identifies
the
most
vulnerable
habitats
based
upon
exposure
duration
(
e.
g.
depositional
sheltered
habitats
are
most
vulnerable
and
sensitive
areas).
The
ESI
maps
also
denote
locations
of
bird
rookeries,
marine
mammal
haulouts,
and
sea
turtle
nesting
areas.
These
maps
are
prepared
regionally
for
coastal
areas.
23
of
44
°
Canada
is
in
the
process
of
developing
a
suitable
receiving
water
scenario
for
pesticide
ecological
assessments
using
PRZM/
EXAMS.
However,
not
having
a
similar
guidance
as
the
USDA
document,
they
took
a
somewhat
different
approach.
They
first
identified
the
type
of
ecosystem
of
concern
and
then
proceeded
to
characterize
the
surface
area
(
SA),
depth,
volume
and
drainage
area
by
examining
available
data.
While
not
finalized,
the
results
for
SA,
volume,
depth,
and
DA/
VC
appear
to
fall
within
the
ranges
in
the
proposed
scenario­
specific
parameters.
The
data
to
do
this
type
of
evaluation
are
limited.
EPA
has
on­
going
activities,
including
a
GIS­
based
approach
that
was
suggested
by
several
Panel
members
to
determine
the
relevance
of
regional
scenarios.

°
Empirical
information
is
available
in
several
recent
publications.
Some
of
the
information
is
from
certain
areas
in
Canada,
but
would
be
relevant
for
northern
states
(
e.
g.,
Hayashi
&
Vander
Kamp,
2000;
Price,
1993;
Brooks
&
Hayashi,
2002;
Wiens
2002).

°
Information
can
be
obtained
by
contacting
other
government
agencies
(
USGS)
or
groups
such
as
Ducks
Unlimited
that
may
have
pertinent
unpublished
information.

°
Data
are
available
from
Agricultural
Experiment
Stations,
such
as
the
standard
"
farm
pond"
construction
guides
and
recommendations.
Local
GIS
surveys
or
data
from
actual
use
patterns
are
still
desirable
and
should
supersede
the
instructions
in
the
manual.

b.)
Please
describe
the
merits
or
limitations
to
the
approaches
and
assumptions
evaluated
for
using
the
U.
S.
Department
of
Agriculture's
(
1997)
guidelines
to
derive
field
size,
surface
water
volume,
and
surface
area
input
values
for
specific
crop
scenarios?
What,
if
any,
additional
approaches
and
assumptions
should
be
considered?

Response
The
Panel
discussed
whether
to
include
subsurface
flow
at
Level
II.
The
most
immediate
effect
of
the
presence
of
groundwater
is
dampening
the
fluctuations
in
pond
water
elevations.
In
general,
it
was
thought
that
neglecting
ground
water
pesticide
inputs
was
appropriately
conservative
for
the
Level
II
analysis,
with
the
possible
exception
of
cases
where
tile
drainage
lines
play
a
significant
role
in
watershed
transport.
The
presence
of
groundwater
can
be
obtained
from
regional
maps.
Alternatively,
as
suggested
by
one
Panel
member,
high
percolation
rates
predicted
by
PRZM
output
can
possibly
be
used
as
a
"
red
flag"
for
the
presence
of
groundwater.

An
alternative
approach
to
developing
scenarios
is
to
use
PRZM
runoff
volumes
for
a
set
drainage
area
 
balanced
by
evaporation
 
to
derive
the
sustainable
volumes
for
permanent
ponds
in
the
various
regions.
There
are
a
number
of
methods
that
might
then
be
used
to
24
of
44
determine
pond
morphometry
parameters
such
as
surface
area
and
depth.
This
approach
assumes
that
the
runoff
estimates
generated
by
PRZM
are
reasonable.

Additionally,
the
USDA
guidance
for
pond
construction
indicates
that,
if
known,
runoff
volumes
for
an
area
should
be
used
rather
than
the
generic
values
given.
If
available,
EPA
might
want
to
consider
taking
advantage
of
these
data.
Limitations
arise
in
arid
areas
where
ponds
either
do
not
naturally
exist
or
only
periodically
contain
water
and
thus
have
no
permanent
aquatic
resident
species.
In
those
cases,
a
more
realistic
scenario
might
be
to
model
a
small
stream
near
the
sprayed
field.

Finally,
the
guidelines
were
developed
for
water
supply
purposes
(
construction
of
small
permanent
water
supplies,
such
as
irrigation)
and
not
for
"
natural
ponds".
The
depth
was
determined
according
to
the
handbook
by
the
expected
rate
of
infiltration.
Because
percolation
from
ponds
is
not
included
in
the
VVWM
model,
the
pond
depth
can
be
smaller
than
what
is
proposed
in
the
handbook.

c.)
A
default
minimum
depth
was
set
as
0.01
m.
What
minimum
depth
would
the
SAP
recommend
as
a
criterion
to
evaluate
the
biological
relevancy
of
the
scenario?

Response
In
general,
the
Panel
was
of
the
opinion
that
a
minimum
depth
of
1
cm
was
too
small,
with
the
exception
for
coastal
wetlands,
where
the
shallow
water
depths
proposed
are
extremely
important
for
providing
accurate
exposure
scenarios
for
wetland
and
estuarine
habitats.
In
particular,
the
shallow
coastal
waters
contain
the
most
sensitive
life­
history
stages
of
fish
and
shellfish,
and
are
the
point
at
which
pesticides
enter
tidal
creeks
and
bind
to
organically
rich
sediments.

In
other
cases,
a
depth
of
15
to
30
cm
seemed
to
be
more
appropriate
for
the
Level
II
model.
If
the
exposure
simulation
indicates
that
reduction
to
very
low
depths
(
e.
g.,
5
to
15
cm)
is
expected,
it
would
be
appropriate
to
redo
the
effects
SSD
such
that
species
requiring
larger
water
volumes
(
e.
g.,
largemouth
bass,
pike)
are
removed
from
the
analysis.
In
all
cases,
it
is
important
to
have
a
better
identification
of
assessment
endpoints
for
the
Level
II
assessment.

As
noted
previously,
in
the
current
document
several
references
are
made
to
the
littoral
zone.
In
the
background
documentation,
references
to
the
"
littoral
zone"
are
in
fact
a
reference
to
the
water
in
the
pond
scenario,
which
is
a
misleading
use
of
the
term.
From
a
limnological
perspective
the
littoral
zone
or
region
is
described
as
the
interface
area
between
the
land
of
the
drainage
basin
and
open
water
of
lakes.
This
zone
includes
both
the
water,
sediments
and
associated
biota
in
near
shore
areas
and
is
important
from
an
ecological
perspective
as
an
area
with
generally
high
biological
activity
when
compared
with
pelagic
(
open
water)
and
profundal
(
deeper
bottom
areas)
regions.
Currently,
the
assumption
for
Level
I
and
II
assessments
is
that
pesticides
entering
water
are
25
of
44
instantaneously
diluted
into
the
entire
water
body.
While
this
is
likely
suitable
for
a
Level
II
assessment,
it
does
not
account
for
the
mixing
which
would
initially
occur
in
the
littoral
zone
of
any
receiving
water.
The
consideration
of
such
a
mixing
zone
might
be
suitable
for
a
Level
III
assessment.

d.)
Simulations
with
the
PRZM/
VVWM
were
performed
using
both
the
crop­
specific
surface
water
area
and
volume
and
the
historic
standard
values
(
DA/
VC
=
1.5
acres/
acre­
ft)
to
characterize
effect
on
exposure
outputs
for
a
relatively
arid
growing
region
(
DA/
VC
=
50
acres/
acre­
ft)
and
a
wetter
climate
(
DA/
VC
=
1
acre/
acre­
ft)
for
both
a
short­
lived
and
long­
lived
pesticide.
In
addition,
the
effect
on
volume
in
the
surface
water
body
was
characterized
for
all
crop­
specific
scenarios.
Please
discuss
what,
if
any,
additional
crop
scenario/
pesticide
evaluations
should
be
performed
to
further
characterize
the
impact
to
exposure
outputs,
and/
or
to
volume.

Response
The
Panel
suggested
many
additional
scenarios
that
should
be
considered
and
are
listed
below.
It
was
clear
from
the
discussions
that
some
of
these
scenarios
would
be
more
appropriate
for
a
Level
III
assessment.
In
all
cases,
regional
differences
were
considered
important.
In
addition,
the
Panel
felt
that
checking
the
model
results
with
available
pond
pesticide
data
would
be
a
good
idea.
In
this
regard,
data
collected
on
pesticide
concentrations
in
ponds
by
Dr.
Cobb
in
Texas
might
be
useful
for
checking
the
models
(
see
report:
Refined
(
Level
II)
Terrestrial
and
Aquatic
Models
Probabilistic
Ecological
Assessments
for
Pesticides:
Level
II
Terrestrial
Model
Session).
The
Panel
had
the
following
suggestions:

°
Pesticides
that
include
chemicals
with
a
range
of
physical
and/
or
chemical
parameters
from
different
major
chemical
classes
should
be
further
evaluated.
In
addition,
subsequent
evaluations
might
include
new
scenario­
based
parameters
(
such
as
timing
of
application)
and
include
characterization
of
other
application
methods,
particularly
in­
furrow
or
sub­
surface
applications.

°
Eventually,
the
goal
is
to
move
from
running
specific
scenarios
to
a
more
comprehensive
evaluation
in
order
to
determine
which
variables
are
important.
It
is
easy
to
run
a
scenario,
but
it
is
harder
to
pick
those
that
are
the
most
important
for
an
assessment.
In
a
situation
with
a
need
to
determine
which
factor(
s)
among
many
factors
actually
drive
the
model,
it
is
appropriate
to
use
fractional
factorial
designs
(
Taguchi,
1986).
Hicks
and
Turner
(
1999),
Kleijnen
(
1997,
2004),
and
Schulz
(
2004)
have
developed
experimental
designs
specifically
for
sensitivity
analysis
of
simulation
models
like
these.
This
is
a
much
more
appropriate
approach
in
contrast
to
simply
testing
scenario
after
scenario
as
they
come
to
mind.
One
of
these
designs
could
be
applied
to
the
model
globally,
or
within
a
regionalized
model.
26
of
44
°
The
regional
approaches
for
predicting
effects
result
in
more
realistic
models
that
take
into
account
differences
in
regional
soil
types,
meteorological
conditions
and
farming
practices.
EPA
should
consider
regional
factors
such
as
plasti­
culture
and
other
farming
and/
or
meteorological
factors,
which
will
result
in
larger
volumes
of
runoff
and
more
frequent
connectivity
between
agricultural
fields
and
surface
waters.

°
Type
of
irrigation
(
e.
g.
drip
versus
surface­
applied)
and
time
of
application
may
be
critical
to
environmental
exposure
outputs.

e.)
What
are
the
advantages
or
disadvantages
to
characterizing
exposure
for
small,
perennial
surface
waters
at
the
edge
of
treated
fields
using
the
method
selected
for
setting
crop
scenario­
specific
DA/
VC
ratio,
depth,
surface
area
and
volume
input
values?
What
adjustments
or
changes
to
the
method
does
the
SAP
recommend,
and
what
are
their
advantages
and
disadvantages?

Response
The
proposed
approach
represents
a
good
first
step
and
is
appropriate
for
Level
II
risk
assessments.
The
method
seems
reasonable.
Advantages
are
that
the
method
of
calculating
pesticide
concentrations
is
transparent,
easy
to
calculate,
keeps
field
size
in
a
reasonable
range,
and
can
take
regional
differences
into
account.

The
Panel
was
impressed
with
the
ability
of
the
model
to
estimate
overflow.
This
parameter
can
be
used
as
a
surrogate
for
estimating
edge­
of­
field
(
within
the
pond)
and
downstream
(
overflow)
effects.
This
allows
EPA
to
focus
on
spatial
effects
and
determining
where
the
majority
of
risk
will
be.
The
model
output
clearly
shows
that
for
short­
lived
chemicals,
there
is
little
in
the
way
of
downstream
effects,
whereas
for
longlived
chemicals
there
could
be
a
clear
downstream
effect.
The
disadvantage
is
that
the
edge­
of­
field
contaminant
losses
are
lumped
and
the
distributed
response
is
not
known.
However,
distributed
modeling
is
clearly
a
Level
III
or
IV
assessment
process.

EPA
is
cautioned
that
in
the
USDA
Handbook
590
guidance
on
constructing
ponds,
specific
advice
is
offered
on
reducing
DA/
VC
for
high
runoff
soils:
"
To
apply
the
information
given
in
Figure
10
in
USDA
(
1997)
some
adjustments
may
be
necessary
to
meet
local
conditions.
Modify
the
values
in
the
figure
for
drainage
areas
other
than
normal.
Reduce
the
values
by
as
much
25%
for
drainage
areas
having
extreme
runoff
producing
characteristics.
Increase
them
by
as
much
by
50%
or
more
for
low
runoff
producing
characteristics."
With
this
in
mind
 
and
remembering
that
crop
scenarios
are
currently
chosen
to
represent
a
high
runoff
soil
and
high
rainfall
 
some
consideration
needs
to
be
given
to
the
combination
of
DA/
VC
and
the
scenario.
The
model
should
be
fine­
tuned
based
on
the
results
of
any
sensitivity
analyses.

f.)
Please
describe
the
weaknesses
and
strengths
of
using
simulated
exposure
concentrations
from
these
crop
scenario­
specific
water
bodies
as
a
surrogate
for
27
of
44
a
low­
order
stream
at
the
edge
of
a
field,
for
a
temporary
pool
or
pond,
and
for
a
small
tidal
creek
or
estuary.

Response
The
strength
of
the
model
is
that
it
runs
quickly
and
uses
the
available
data
to
its
best
advantage.
However,
EPA
is
advised
to
more
clearly
define
the
endpoints
of
the
assessment
and
their
relation
to
the
types
of
habitat.
This
in
turn
would
help
in
determining
the
nature
of
the
temporary
habitat
(
e.
g.,
amphibians,
waterfowl,
etc.).

Weaknesses
listed
by
individual
Panel
members
were:

°
The
fate
parameters
in
estuarine
areas
are
poorly
characterized
because
all
fate
processes
are
determined
in
freshwater.
Some
information
for
older
chemicals
in
salt
water
might
exist
in
the
open
literature.

°
Daily
time
steps
are
not
appropriate
for
tidal
creeks.

°
To
estimate
exposure
concentrations
in
low­
order
streams,
it
would
be
useful
to
consider
linking
the
VVWM
to
models
that
predict
mixing
zones
and
downstream
concentrations
in
streams
following
inputs
from
outfalls
or
other
point
sources
(
e.
g.,
CORMIX,
7Q10).

°
The
modeled
ponds
have
no
littoral
edge
to
them.

°
The
effect
of
surface
area
on
evaporation
rates
and
photolysis
should
be
considered
in
more
depth.

°
Simulated
exposure
concentration
with
VVWM
for
a
stagnant
water
body
might
not
be
representative
for
streams
where
advection
is
important.

g.)
Simulations
with
PRZM/
EXAMS,
a
fixed
volume
surface
water
model,
will
be
performed
using
both
the
crop­
specific
DA/
VC
approach
and
the
historic
standard
values
to
characterize
effect
on
exposure
outputs
for
relatively
arid
growing
regions
(
DA/
VC
=
50
and
80)
and
a
wetter
climate
(
DA/
VC
=
1)
for
both
a
short­
lived
and
long­
lived
pesticide.
Please
discuss
what,
if
any,
additional
crop
scenario/
pesticide
evaluations
should
be
performed
to
further
characterize
the
impact
to
exposure
outputs
in
a
fixed
volume
situation.

Response
Clarification
was
sought
from
EPA
staff
on
the
interpretation
of
the
question.
EPA
clarified
the
statement
by
indicating
that
they
are
currently
considering
a
Level
I,
Tier
II
assessment
that
would
utilize
the
fixed
volume
receiving
water
body
in
PRZM/
EXAMS
28
of
44
using
two
DA/
VC
ratios
(
1
&
50),
and
then
use
a
regional
approach
with
PRZM/
VVWM
in
Level
II
assessments.
Because
PRZM/
EXAMS
is
already
well
characterized
with
respect
to
crop
scenario
inputs
and
pesticides,
no
further
characterization
is
required.
However,
EPA
may
want
to
evaluate
how
this
approach
will
impact
the
overall
risk
assessment
process
and,
in
particular,
the
movement
from
Level
I
to
Level
II.

In
general
the
approach
was
considered
good.
In
some
cases,
groundwater,
plasti­
culture
and
wet
vs.
dry
weather
scenarios
(
El
Ni

o
vs.
La
Ni

a
scenarios)
should
be
taken
into
account.
EPA
should
also
compare
pesticides
with
differing
Koc
levels
to
test
how
sediment
partitioning
is
affected
by
Koc.

h.)
Please
discuss
sources
or
approaches
for
national
or
regional
DA/
VC
ratios
and
associated
water
depth
and
size
information
for
temporary
pool
and
pond
aquatic­
life
resources.

Response
In
addition
to
government
sources
for
data
on
the
hydrometric
dimensions
for
receiving
waters
(
surface
area,
volume,
depth,
drainage
area)
other
organizations
may
also
have
collected
relevant
information.
For
example,
in
Canada,
Ducks
Unlimited
has
compiled
a
database
from
Landsat
imagery
of
3,061,000
wetlands
covering
a
total
area
of
almost
11
million
acres
(
4.4
million
hectares).
This
covers
about
90%
of
the
prairie
ecozone.
This
database
was
obtained
from
the
Ducks
Unlimited
National
Headquarters
in
Oak
Hammond
Marsh
Conservation
Centre,
Stonewall,
Manitoba
(
Wetland
Habitat
Inventory
for
Prairie
Pothole
Region
of
Canada).
Summary
data
provided
by
Ducks
Unlimited
staff
included
distribution
of
surface
area
size
for
wetlands
in
the
prairie
region
of
Canada.
Similar
data
may
be
available
from
this
or
other
groups
in
the
U.
S.

In
Canada
data
from
Ducks
Unlimited
were
used
to
develop
one
estimate
of
a
typical
drainage
area
for
a
1­
ha
water
body.
While
still
under
development,
and
not
yet
available
for
release,
the
approach
used
can
be
outlined.
The
database
includes
the
area
of
water
in
each
quarter
section
(
160
acres)
in
the
prairie
ecozone
derived
from
Landsat
imagery.
The
database
represented
a
compilation
of
Landsat
scenes
from
36
varied
dates
ranging
from
1984
to
1995
and
during
the
months
April,
May
and
June.
Data
were
screened
to
eliminate
those
areas
which
might
skew
results
(
e.
g.
areas
with
incomplete
data
or
which
included
major
rivers).
These
data
were
analyzed
by
calculating
the
overall
ratio
between
the
area
of
the
wetland
and
the
total
area
(
water
and
land).
This
ratio
indicates
the
overall
fraction
of
the
surface
area
of
the
landscape
that
is
occupied
by
wetlands.
The
inverse
of
this
fraction
is
the
mean
catchment
area,
where
catchment
area
is
defined
as
the
total
watershed
area.
The
mean
drainage
area
is
calculated
by
subtracting
the
mean
wetland
area
from
the
mean
catchment
area.

The
main
advantage
of
this
analysis
is
the
very
large
size
of
the
database
that
covers
most
of
the
Canadian
prairie
ecozone.
The
main
disadvantage
is
the
unknown
degree
of
29
of
44
uncertainty
introduced
by
the
climatic
and
temporal
variations
in
the
data.

Texas
A&
M
and
the
University
of
North
Texas
have
high
resolution
(
3
m)
GIS
maps
of
marshes
along
eastern
Texas.
We
are
sure
that
other
state
agencies
are
increasingly
developing
these.

An
older,
yet
very
influential,
text
by
Hutchinson
(
1957)
on
Limnology
contains
descriptions
on
types
of
ponds
and
other
water
bodies,
the
importance
of
morphometry
in
determining
biological
activity,
and
changes
in
shape
and
its
influence
on
biological
dynamics.

Manuscripts
by
Schulz
(
2004)
and
Pennington
et
al.
(
2001)
have
information
on
the
relationship
between
drainage
area
and
pond
size.

Other
Information:
Satellite
Data:
LIDAR
Data
and
specifically
the
Beaufort
County
Special
Area
Management
Plan
in
South
Carolina
is
a
good
example.

4.
Curve
Number.
The
SAP
2001
recommended
that
additional
characterizations
of
variability
should
be
given
to
those
parameters
in
the
exposure
model
that
have
a
major
impact
on
exposure
concentrations.
The
curve
number
is
perhaps
the
most
influential
parameter
in
PRZM,
and
it
has
been
interpreted
in
recent
literature
as
a
random
variable.
PRZM
currently
treats
the
curve
number
as
a
function
of
soil
moisture,
although
recent
literature
suggests
that
the
curve
number
may
more
appropriately
be
interpreted
as
a
random
variable.

General
Comments
The
Agency
is
commended
for
attempting
an
innovative
solution
to
a
difficult
problem.
It
is
possible
to
spend
considerable
time
and
energy
on
a
detailed
infiltration
model
based
on
physical
principles.
While
such
a
model
might
be
appropriate
for
a
Level
III
or
IV
assessment,
it
is
not
clear
that
such
a
model
is
needed
at
Level
II.
The
Panel
generally
supports
the
proposed
curve
number
approach
with
the
qualifications
detailed
in
the
response
to
the
questions.

In
order
to
better
understand
the
limitation
of
the
curve
number
approach,
a
short
overview
on
its
development
is
given
here.
The
original
form
of
the
SCS
equation
as
proposed
by
Mockus
was
otherwise
0
,
if
)
(
2
=
>
+
 
 
=
Q
I
P
S
I
P
I
P
Q
a
a
a
where
S
is
the
watershed
storage
and
Ia
is
the
initial
abstraction
or
the
amount
of
rainfall
before
runoff
occurs
(
see
also
Eq
4.61
in
the
March
4,
2004
Document).
In
the
SCS
30
of
44
Handbook
4
a
relationship
was
proposed
between
initial
abstraction
and
S
(
Figure
1).
Based
on
this
data
it
was
proposed
that
Ia=
0.2S.
As
can
be
seen
from
Figure
1
this
relationship
between
S
and
Ia
is
tentative
at
best.
This
should
not
be
a
surprise
since
S
in
the
original
theory
of
Mockus
is
a
function
of
the
amount
of
water
that
can
be
stored
in
the
watershed
after
runoff
has
started.
This
S
is
a
function
of
the
depth
of
the
soils
in
the
watershed
and
should
be
only
minimally
dependent
on
the
rainfall
history.
The
initial
abstraction,
Ia,
defined
previously
as
the
amount
of
rainfall
before
runoff
starts,
is
thus
mainly
a
function
of
the
moisture
content
in
the
watershed
and
consequently
the
rainfall
history.
Thus
S
and
Ia
are
fundamentally
different
parameters,
each
with
their
own
distributions,
and
should
not
be
related
to
each
other.
This
is
important
when
considering
probabilistic
approaches
using
the
curve
number
equation
to
predict
runoff.

Ia
vs.
S
0.01
0.1
1
10
0.1
1
10
100
S
Ia
Figure
1:
Plot
of
initial
abstraction,
Ia,
and
watershed
storage
S
(
both
in
inches).
Redrawn
from
SCS
Handbook
4.

a.)
Please
discuss
the
pros
and
cons
of
assuming
strict
dependence
of
curve
number
on
calculated
soil
moisture
versus
treatment
as
a
random
variable
unrelated
to
soil
moisture
as
a
means
of
characterizing
runoff
variability.
Please
identify
and
discuss
alternative
methods.

Response
Two
methods
are
proposed
for
calculation
of
the
runoff
with
the
curve
number
(
CN):
Method
1,
Physically­
based
deterministic
CN
method;
and
Method
2,
Soil
moisture
independent
probabilistic
CN
method.
31
of
44
Deterministic
Approach
For
the
physically­
based
runoff
model
(
Method
1)
there
are
two
prevailing
theories
for
the
runoff
generation
hydrologic
process.
The
dependence
of
CN
on
moisture
content
depends
on
which
runoff
mechanism
dominates.
The
first
theory
assumes
that
runoff
begins
when
rainfall
rates
exceed
infiltration
capacities
of
soil.
This
theory
is
referred
to
as
infiltration­
excess,
also
called
Hortonian
flow,
in
honor
of
the
related
research
by
Robert
Horton
(
1933,
1940).
According
to
Hortonian
flow
theory,
runoff
amounts
are
directly
controlled
by
characteristics
that
influence
soil
infiltration,
such
as
land
use
and
soil
type.
For
watersheds
where
runoff
is
generated
by
infiltration
excess,
the
dependence
on
moisture
content
is
tentative.

Generally,
high
water
content
in
the
topsoil
may
indicate
a
higher
runoff
potential
while
dry
soil
tends
to
absorb
more
infiltrating
water.
However,
no
direct
and
definite
relationship
between
water
content
and
runoff
exists.
The
second
prevailing
overland
flow
theory
assumes
most
runoff
is
generated
via
direct
precipitation
onto
or
exfiltration
from
saturated
areas
in
the
landscape
through
a
process
termed
saturation
excess
overland
flow
(
e.
g.,
Dunne
and
Black,
1970).
The
extent
of
these
saturated
areas
can
vary
with
season
and
depends
on
the
moisture
content,
and
can
be
predicted
with
the
CN
method
(
Steenhuis
et
al.,
1995).

For
watersheds
where
saturation
excess
overland
flow
dominates,
a
relationship
between
runoff
amount
and
moisture
content
is
expected.
In
watersheds
where
saturation
excess
overland
flow
is
the
main
mechanism
of
producing
runoff,
the
original
SCS
equation
can
well
simulate
the
runoff
pattern
with
a
constant
S
value
and
with
Ia
calculated
with
a
water
balance
(
Thornthwaite
Mather,
1955)
for
the
shallowest
soil
in
the
watershed,
which
will
produce
the
first
runoff
(
Steenhuis
et
al.,
1995;
Lyon
and
Steenhuis,
2004).
Figure
2
is
an
example
from
the
Town
Brook
Watershed
in
the
Catskills
region
of
New
York
32
of
44
Figure
2:
Prediction
of
Runoff
for
two
watersheds
in
the
Catkills
Region
of
New
York.

For
the
traditional
curve
number
equation
(
i.
e.,
with
Ia=
0.2S;
See
also
Eq.
4­
62
in
the
March
4,
2004
Document),
it
is
assumed
that
CNI,
CNII,
CNIII
represent
the
10th,
50th,
and
90th
percentile
curve
numbers,
respectively.
Statistically,
this
method
is
sound.

Physically,
however,
failure
to
use
the
available
simulated
soil
moisture
may
miss
some
valuable
information
as
indicated
above.
Note
that
the
dependence
of
CN
on
moisture
content
occurs
because
it
assumed
that
Ia
=
0.2
S.
If
we
decouple
the
initial
abstraction
from
S,
then
the
CN
should
be
independent
of
initial
moisture
content.
Based
on
the
above,
further
work
is
needed
to
evaluate
the
applicability
of
the
presented
log
normal
distribution
with
mean
=
­
1.609
and
standard
deviation
=
0.67
(
note
that
the
mean
value
is
derived
from
Ia/
S
=
0.2).

Probabilistic
Approach
Although
incorporating
variability
by
generating
a
random
CN
is
a
positive
step,
abandoning
linkage
among
precipitation,
infiltration,
and
runoff
may
cause
other
issues
to
become
problematic.

There
are
several
ways
that
these
concerns
could
be
addressed.
A
linear
regression
model
is
one
straightforward
alternative
that
could
include
both
soil
moisture
dependence
and
variability.
Analysis
of
the
available
data
would
quickly
indicate
the
relative
importance
of
the
two.
The
physically­
based
components
could
be
incorporated
into
the
probabilistic
code.
One
of
the
ways
that
EPA
could
consider
changing
the
probabilistic
approach
is
as
follows:
33
of
44
CNI
and
CNIII
were
early
attempts
to
introduce
variability
into
a
deterministic
model.
The
correlation
between
CN
and
antecedent
rainfall
is
too
weak
to
be
useful.
The
Agency
wants
a
model
for
runoff
that
is
simple
and
interpretable,
but
still
good
enough
for
Level
II
risk
assessments.
The
presented
attempts
to
describe
the
unpredictability
in
runoff
in
terms
of
a
random
CN
are
unnecessarily
convoluted
and
consequently
confusing.
We
suggest
that
EPA
apply
the
following
approach
instead
of
the
current
one.

In
this
approach,
CN
is
considered
to
be
a
global
property
of
a
given
terrain.
Temporal
and
spatial
variations
in
the
soil
are
modeled
by
random
variables.
Pick
CN
(
CNII)
according
to
the
terrain
from
the
standard
table
or,
even
better,
derive
S
directly
from
watershed
outflow
data
if
available
as
this
avoids
having
to
assume
anything
about
the
relationship
between
CN
and
S.
The
model
can
then
generate
log­
normal
random
variables
X1
and
X2
to
give
runoff
Q
by
the
following
formulas:

S
=
1000
CN
 
10
I
a
=
X
1
S
or
I
a
=
X
1
 
Q
=
X
2
(
P
 
I
a
)
2
P
 
I
a
+
S
P
>
I
a
0
P
 
I
a





where
 
is
a
variable
that
indicates
the
moisture
status
in
the
watershed,
calculated
with
a
simple
water
balance
procedure
such
as
introduced
by
Thornthwaite
&
Mather
(
1955,
1957)
or
calculated
directly
with
PRZM.
If
we
assume
that
I
a
=
0.2S
and
that
the
value
of
Q
is
centered
on
the
deterministic
formula,
we
could
take
X
1
~
lnorm(
log(.
2),
 
1)
X
2
~
lnorm(
log(
1),
 
2)

This
leaves
two
unknown
parameters,
the
lognormal
standard
deviations,
which
can
be
set
to
reasonable
values,
perhaps
by
matching
quantiles.
We
already
have
an
accepted
value
 
1
=
0.67
.
Figure
3
presents
some
simulation
results
showing
runoff
for
given
rainfall,
assuming
 
1
=
1,
 
2
=
0.5,
and
the
results
appear
to
be
quite
realistic.

Final
Comment
It
is
important
to
realize
that
the
approach
in
Chapter
IV
(
March
4,
2004
Document)
was
applied/
tested
in
a
very
limited
manner
(
e.
g.,
the
example
used
was
based
on
a
single
dataset
from
a
very
small
catchment).
The
final
code
should
be
tested
under
a
much
wider
range
of
conditions
(
different
catchment
sizes,
duration
and
intensity
of
rain
events,
etc.)
in
order
to
adequately
account
(
if
possible)
for
the
unexplained
sources
of
variability.
34
of
44
Panel
members
concluded
that
EPA
should
consider
modifying
the
code
to
include
physical
relationships
(
CN
linkage
to
important
physical
parameter(
s))
and
probabilistic
aspects.
The
Panel
proposed
an
alternate
probabilistic
approach
that
will
aid
in
fusing
physical
and
probabilistic
issues.

b.)
Since
the
curve
number
was
not
designed
for
use
in
continuous
modeling,
what
problems
may
arise
when
the
curve
number
is
used
in
this
manner?
Could
a
probabilistic
interpretation
address
some
of
these
issues?
If
so,
how?

Response
The
Panel
expressed
several
concerns
regarding
this
issue.
The
concern
was
high
because,
as
stated
in
the
EPA
document,
the
CN
is
so
important
in
predictions.
The
manner
in
which
the
program
resets
conditions
at
the
beginning
of
each
day
was
a
major
concern.
The
scale
within
which
the
CN
approach
was
being
used
was
seen
as
an
issue
to
be
addressed
in
the
near
future
by
EPA.
The
original
CN
application
was
for
watersheds
and
annual
extreme
rain
events;
however,
application
here
is
for
shorter
temporal
and
smaller
spatial
scales.

Spatial
and
Temporal
Scales
Spatial
and
temporal
scales
are
important
in
addressing
this
issue.
The
choice
of
scale
is
important
here
because
the
different
applications
of
the
method
differ
markedly
in
their
spatial
and
temporal
scales.
The
original
development
of
CNs
in
hydrological
engineering
was
for
a
much
larger
spatial
and
temporal
scale
than
the
proposed
farm
field/
daily
application.
Data
collected
on
gauged
watersheds
are
also
at
a
large
scale.
Experimental
runoff
data
(
e.
g.,
Wauchope
et
al.,
1999)
are
collected
on
smaller
spatial
and
temporal
scales
than
the
proposed
application.
Figure
3:
Probabilistic
simulation
of
runoff
35
of
44
Spatial
scale
Temporal
scale
Classic
curve
number
Watershed
Annual
extreme
rainfall
event
Watershed
data
Watershed
Rainfall
event
EPA
application
Farm
field
(
10
ha)
Daily
Experimental
data
Small
plot
2
hours?

The
Agency
should
carefully
consider
scale
when
it
examines
CN
data.
Changing
scales
usually
changes
variability,
but
the
direction
of
the
change
is
not
obvious.
Changing
from
an
event
time
scale
to
a
daily
time
scale
adds
a
concern
about
autocorrelation
discussed
in
the
next
question.
A
probabilistic
interpretation
does
not
address
this
concern.
If
anything,
it
hides
the
issue
when
estimates
of
variability
from
one
scale
are
applied
without
change
to
other
spatial
or
temporal
scales.
The
Agency
should
evaluate
whether
the
proposed
standard
deviation
from
very
large­
scale
phenomena
is
appropriate
for
daily
field­
scale
data.

When
the
Agency
examines
data,
they
should
also
consider
the
components
of
variability
that
might
be
present.
Do
the
data
represent
variability
among
events
on
the
same
field?
Variability
among
years
on
a
single
field?
Variability
among
fields?
All
combinations
are
available
and
will
not
have
the
same
population
standard
deviation.

Rainfall
events
of
several
days
A
very
important
issue
is
how
one
defines
a
rainfall
event.
Rainfall
events
extending
through
two
or
more
days
would
be
treated
in
the
code
as
separate
(
parameter
values
being
independent).
Especially
in
regions
where
saturation
excess
runoff
is
dominant,
this
leads
to
gross
underprediction
of
runoff
events.
In
the
Northeastern
US,
for
example,
any
rainfall
event
producing
more
than
15
cm
of
rainfall
over
several
days
will
cause
significant
flooding.
Currently,
PRZM
resets
the
CN
at
the
beginning
of
each
day.
Making
the
CN
probabilistic
in
such
a
case
would
ignore
useful
information
from
previous
days.
Because
the
CN
is
so
important,
it
would
seem
that
the
consequences
of
this
shortcoming
should
be
explored
more
thoroughly.
Would
it
be
profitable
to
explore
the
addition
of
a
subroutine
that
partially
addresses
this
issue
for
such
events?
Relative
to
multiday
events,
on
each
rainfall
event,
the
simulation
could
determine
the
number
of
days
it
will
last
and
then
modify
the
simulation
for
those
days.
The
CN
model
was
derived
from
event­
scale
information,
yielding
total
runoff
for
the
event.
The
proposed
use
is
to
apply
the
same
model
to
daily
precipitation
to
give
daily
runoff.
An
event­
total
runoff
could
be
computed
from
the
sum
of
the
daily
runoff
events.
At
a
minimum,
the
properties
of
the
total
daily
runoff
should
be
similar
to
the
event
runoff.
The
shape
of
the
runoff
distribution
depends
on
the
relationship
between
P
and
Ia.
It
is
left
skewed
at
low
precipitation
and
right
skewed
at
high
precipitation
(
Figure
4).

The
sum
of
the
daily
runoff
will
have
a
different
distribution
than
the
event­
scale
runoff.
The
annual
total
runoff
is
the
sum
of
event
runoffs
or
daily
runoffs.
The
distribution
of
the
annual
total
will
also
be
different.
36
of
44
The
Ia
is
commonly
approximated
as
0.2
S
for
event­
scale
data.
Applying
this
value
to
daily
data
will
underestimate
the
total
runoff.
When
the
model
is
applied
to
daily
precipitation
data,
Ia
values
are
subtracted
from
each
day's
precipitation
in
a
multi­
day
rainfall
event
and
will
underestimate
the
runoff
unless
the
same
adjustment
to
Ia
and
S
is
made
on
subsequent
days.
Choosing
and
justifying
such
an
adjustment
will
be
difficult
because
of
the
lack
of
available
data.
One
approach
to
evaluate
the
magnitude
of
the
issue
would
use
daily
precipitation
sequences.
For
each
sequence,
classify
days
into
rainfall
events.
Some
events
may
be
single
day;
others
may
be
multi­
day.
Compare
event­
level
predictions
to
the
sum
of
daily
predictions.
This
approach
could
also
be
used
to
evaluate
different
methods
to
deal
with
autocorrelation.

0.2
2
5
10
0
2
4
6
8
10
runoff
0.2
2
5
10
0.0
0.2
0.4
0.6
0.8
1.0
runoff/
precip
Figure
4:
The
shape
of
the
runoff
distribution
depends
on
the
relationship
between
P
and
Ia.
It
is
left
skewed
at
low
precipitation
and
right
skewed
at
high
precipitation
c.)
What
is
the
impact
on
interpretation
of
probabilistic­
simulated
exposure
values
when
the
curve
number
is
used
as
a
random
variable
and
autocorrelation
of
temporally
varying
physical
properties
that
may
impact
run
off
is
ignored?

Response
Autocorrelation
between
random
variables
influences
the
variance
but
not
the
mean.
Regardless,
the
presence
of
autocorrelations
could
insert
error
in
predictions
from
the
probabilistic
model.
Several
methods
were
identified
by
Panel
members
for
coping
with
potential
autocorrelation.
They
include
permutation
tests,
including
the
autocorrelations
in
future
versions
of
the
probabilistic
models
(
i.
e.,
using
correlation
coefficients
in
Monte
Carlo
simulations),
and
running
a
crude
sensitivity
case
study.
These
methods
are
discussed
below.
37
of
44
A
data
set
provided
to
the
Panel
indicated
no
apparent
dependence
of
CN
on
rainfall
in
the
previous
5
days,
thus
ignoring
autocorrelation
may
have
no
significant
impact
most
of
the
time.
Potential
impacts
are
difficult
to
determine
heuristically.
The
effect
of
ignoring
autocorrelation
(
interpreted
as
a
subsequent
event
having
higher
runoff
than
would
be
calculated
under
the
random
model)
would
lead
one
to
conclude
that
the
pesticide
concentration
in
the
pond
would
be
slightly
overstated.

If
no
dependencies
exist
between
CN
and
temporally­
varying
physical
properties,
the
autocorrelation
of
the
latter
properties
is
of
no
concern.
However,
if
such
dependencies
exist,
they
should
be
incorporated
in
future
versions
of
the
probabilistic
models
(
e.
g.,
use
of
correlation
coefficients
in
Monte
Carlo
analysis).
In
the
latter
situation,
CN
would
need
to
be
reselected
for
each
time
step
of
the
analysis.

As
with
any
data
that
are
autocorrelated,
ignoring
this
relationship
can
reduce
the
confidence
in
statistical
analyses.
Because
this
particular
situation
is
not
one
of
statistical
significance
testing
per
se,
autocorrelation
may
be
less
of
a
problem.
There
are
a
number
of
statistical
approaches
that
could
be
used
to
handle
spatially­
or
temporally­
autocorrelated
data
including
permutation
tests.
The
latter
can
be
used
in
situations
where
data
are
independent
(
default)
or
known
to
be
autocorrelated.
Permutation
methods
may
also
be
appropriate
because
this
form
of
testing
is
already
incorporated
as
part
of
the
refined
Level
II
RA
approach
(
e.
g.,
Figure
4.1)
and
is
being
considered
(
in
the
form
of
Monte
Carlo
testing)
as
a
method
of
integrating
the
potential
variability
of
CN
into
exposure
modeling
(
e.
g.,
Question
4e).

d.)
A
lognormal
distribution
is
being
investigated
to
characterize
variability
in
certain
curve
number
parameters.
Is
it
reasonable
to
assume
such
a
distribution
has
stationary
properties
(
constant
mean
and
variance)
for
all
rain
events
(
e.
g.,
large
and
small)?
Please
provide
rationale.

Response
The
Panel
understands
the
context
(
Level
II)
within
which
the
lognormal
distribution
was
being
proposed.
Regardless,
several
group
members
recommended
that
more
work
is
needed
because
there
is
little
evidence
suggesting
that
the
distribution
will
remain
stable.
It
seemed
unlikely
that
one
lognormal
distribution
would
provide
adequate
predictions
for
all
relevant
scenarios.
The
distribution
can
change
even
within
a
rain
event.
Expansion
in
the
near
future
to
include
these
issues
is
recommended.

e.)
Monte
Carlo
modeling
is
being
investigated
as
a
method
of
integrating
the
potential
variability
of
curve
numbers
into
exposure
modeling.
Can
the
SAP
recommend
other
methods
available
to
incorporate
variable
and
uncertain
curve
numbers
into
a
continuous
runoff
model.
Please
discuss
the
pros
and
cons
of
these
methods
versus
Monte
Carlo.
38
of
44
Response
The
general
response
is
that
Monte
Carlo
is
the
method
of
choice,
but
as
detailed
in
specific
Panel
member
comments,
there
are
some
computational
issues
that
should
be
considered.
This
is
in
addition
to
what
is
presented
above.
Below,
an
important
issue
relative
to
how
one
does
the
Monte
Carlo
simulations
is
highlighted,
providing
a
specific
alternate
approach.

Monte
Carlo
Simulations
There
may
be
a
more
appropriate
Monte
Carlo
method.
The
current
method
has
three
important
characteristics:
1)
A
lognormal
distribution
of
Ia/
S,
based
on
Hawkins
et
al.
(
1985);
(
2)
It
can
be
adapted
to
specific
regions
and
crops
by
setting
the
median
CN
according
to
standard
tables;
and
(
3)
The
output
is
a
CN
to
feed
to
PRZM.
The
approach
proposed
by
EPA
is
to:

Calculate
the
value
of
S
for
the
tabular
value
of
CN
Generate
values
of
Ia/
S
from
a
log
normal
distribution
and
multiply
by
S
to
get
Ia
Calculate
Q
from
P,
S,
and
Ia,
Recompute
the
Se
that
corresponds
to
that
Q
and
P
Convert
Se
back
to
a
curve
number
This
approach
puts
all
the
variability
into
Ia.
The
alternative
is
to
put
the
variability
into
S,
by
treating
Se
as
a
lognormal
random
variable,
e.
g.
log
Se~
N(
log
S,
s.
d.)
where
s.
d.
is
the
same
s.
d.
used
to
simulate
Ia/
S,
e.
g.
the
Hawkins
value,
0.67.
When
the
precipitation
is
high,
this
approach
generates
a
similar
CN
distribution
as
the
EPA
approach.
It
does
not
do
so
for
low
precipitation,
because
of
the
truncation
when
P
<
Ia.
The
practical
effect
of
the
difference
in
distributions
may
be
small.
The
largest
differences
between
distributions
occur
for
low
precipitation
and
moderate
CN.
These
are
conditions
with
relatively
small
amounts
of
precipitation.
The
CN
distributions
are
similar
for
conditions
expected
to
produce
large
amounts
of
runoff:
high
precipitation
and
large
CN.

An
important
advantage
of
this
alternate
method
is
easy
explanation.
The
current
Agency
method
is
indirect
and
hard
to
follow.
The
alternative
method
has
the
advantage
of
clarity
and
transparency.
In
the
alternative
method,
a
tabular
value
is
used
to
set
the
median
of
the
distribution.
The
remaining
attributes
of
the
distribution
(
standard
deviation
and
distributional
form)
come
from
Hawkins'
(
1985)
model.

This
alternative
method
is
not
the
only
possible
one.
There
are
two
random
quantities,
Ia
and
S.
If
data
were
available,
distributions
could
be
constructed
for
both
quantities.

The
distribution
of
CNs
should
be
compared
to
field
data
sets,
even
though
there
are
few
appropriate
data
sets.
The
upper
portion
of
the
distribution
(
e.
g.
CN
>
90)
should
be
given
the
most
attention.
Are
the
quantiles
of
the
probabilistic
distribution
similar
to
the
empirical
quantiles
in
the
field
data?
This
comparison
should
be
restricted
to
comparable
39
of
44
events.
Because
the
field
data
come
from
rainfall
simulator
experiments,
the
field
data
do
not
include
an
event
unless
there
was
runoff.
Hence,
the
appropriate
comparison
is
to
the
probabilistic
distribution,
truncated
to
omit
the
zero
values.
The
comparison
of
Figure
4.31
(
in
chapter
IV
of
the
March
4,
2004
Document)
suggests
they
are
similar,
but
it
is
hard
to
compare
the
distributions
in
a
plot
like
figure
4.31.
A
quantile­
quantile
plot
provides
an
easier
way
to
interpret
comparison
of
distributions.

The
Panel
discussions
included
other
approaches.
Second­
order
Monte
Carlo
techniques
are
often
used
to
separate
uncertainty
due
to
variability
and
uncertainty
due
to
lack
of
knowledge.
Probability
bounds
analysis
can
also
be
used
to
separate
and
estimate
the
relative
importance
these
two
sources
of
uncertainty.
If
assuming
constant
mean
and
variance
is
inappropriate
for
either
theoretical
reasons
or
because
the
sample
size
is
low,
then
second­
order
Monte
Carlo
analysis
or
probability
bounds
analysis
would
be
useful
techniques
to
deal
with
this
situation.

5.
Additional
General
Comments
from
the
SAP
At
the
conclusion
of
the
Panel
discussions
on
the
questions,
Panel
members
were
given
the
opportunity
to
make
additional
comments.
Comments
were
on
subjects
previously
discussed
and
also
on
subjects
not
specifically
included
in
the
Panel
discussions
of
questions.

Several
Panel
members
have
been
involved
with
this
process
from
its
early
days
and
recognized
that
a
significant
amount
of
work
has
already
been
done
and
that
EPA
has
moved
forward
in
its
proposed
risk
assessment
methods.
The
currently
proposed
approach
is
reasonable
given
the
Panel's
understanding
of
the
Agency's
goals
for
a
Level
II
assessment.
Given
that
deterministic
assessments
are
currently
used
for
decision
making,
some
Panel
members
lauded
EPA's
desire
to
start
implementation
within
the
next
8
­
12
months.

While
this
is
a
good
start,
additional
work
will
be
necessary.
EPA
must
move
beyond
the
current
conceptual
model,
which
is
suitable
for
agricultural
uses
of
pesticides,
and
identify
which
other
conceptual
models
should
be
developed
for
other
use
patterns.
During
the
course
of
discussions,
several
use
patterns/
scenarios
were
identified,
e.
g.,
mosquito
control,
forestry,
urban
uses
or
receiving
water
scenarios
other
than
a
permanent
water
body
such
as
temporary
pools.
A
number
of
Panel
members
identified
the
need
for
a
conceptual
model
that
includes
tile
drainage
as
an
input
source
to
receiving
waters.

Several
Panel
members
noted
that
the
topic
for
this
SAP
focused
exclusively
on
estimation
of
exposure
concentrations.
Panel
members
had
the
following
comments
with
respect
to
the
characterization
of
toxicity
and
the
use
of
toxicity
data
in
risk
assessment.

The
species
sensitivity
distribution
(
SSD)
approach
currently
being
proposed
in
the
Level
40
of
44
II
aquatic
model
relies
on
LC/
EC50s
derived
by
probit
analysis.
Probit
analysis
is
appropriate
for
quantal
endpoints,
e.
g.,
mortality,
but
for
other
types
of
endpoints,
e.
g.,
count
or
continuous
variables,
other
types
of
models
must
be
used.
The
Generalized
Linear
Model
(
GLiM)
framework
described
by
Kerr
and
Meador
(
1996)
and
Bailer
and
Oris
(
1997)
is
a
useful
framework
for
deriving
concentration­
response
relationships
for
a
variety
of
toxicity
test
endpoints,
e.
g.,
quantal,
count
and
continuous
endpoints.
The
framework
involves
using
link
functions
to
transform
effects
metrics,
e.
g.,
probit
or
logit
link
functions
for
quantal
responses,
log
transformation
for
count
and
continuous
variables,
and
assigning
appropriate
error
distributions,
e.
g.,
binomial
distribution
for
quantal
responses,
Poisson
distribution
for
count
variables,
normal
distribution
for
continuous
variables.
Linear
regression
can
then
be
conducted
on
the
transformed
data
to
derive
the
concentration­
response
relationship.
Thus,
the
framework
can
be
used
for
all
available
types
of
response
variables.
By
adding
a
quadratic
term
to
the
linear
model,
the
framework
can
be
adapted
to
incorporate
simulations
at
low
concentrations.

The
use
of
lower
percentiles,
e.
g.,
LC/
EC10s,
should
be
considered
in
deriving
SSDs,
given
that
the
goal
of
Level
II
assessments
is
to
err
on
the
side
of
conservatism.

Newman
et
al.,
2001,
showed
that
formal
hypothesis
testing
results
in
rejection
of
the
lognormal
distribution
in
more
than
half
of
50
species
sensitivity
data
sets
assessed.
Therefore,
using
a
lognormal
model
for
all
SSD
analyses
may
not
be
optimal,
given
that
the
lower
part
of
the
curve
that
is
used
for
doing
predictions
will
show
the
most
difference
among
the
models.

Other
distributions
can
and
are
used;
see
other
chapters
in
the
book
in
which
Newman
et
al.
(
2001)
were
published.
Also
there
are
bootstrap
methods
that
can
be
used
that
avoid
assumptions
of
any
particular
distribution
such
as
presented
in
Grist
et
al.
(
2002).

Where
peaks
in
ponds
are
short­
lived,
it
may
be
inappropriate
to
rely
on
96
h
toxicity
test
endpoints.
Shorter
duration
toxicity
test
endpoints
should
be
used
in
these
situations.
Effects
at
different
exposure
durations
can
be
matched
to
exposure
peak
durations
through
use
of
time­
to­
effect
modeling.
This
assumes
that
toxicity
results
are
available
for
multiple
times
during
the
test.

With
respect
to
the
topic
of
exposure
duration,
one
Panel
member
provided
the
following
publications
and
reports
which
provide
detailed
explanations
of
pulsed
dose
responses:
Clark
et
al.
(
1986,
1987),
Scott
et
al.
(
1989,
1992),
and
Moore
et
al.
(
1989).

For
assessment
of
chronic
effects,
it
was
suggested
that
endpoints
from
chronic
studies
need
to
move
from
NOAELs
(
hypothesis
testing
study
design)
to
regression­
based
endpoints.
This
is
currently
being
done
with
endpoints
for
acute
effects,
for
pesticides
with
sufficient
data,
(
e.
g.,
re­
registration
pesticides
that
are
fairly
persistent
and
have
large
aquatic
toxicity
data
sets).
Some
noted
that
this
would
require
a
change
to
the
existing
protocols
for
these
types
of
studies.
41
of
44
It
may
be
appropriate
to
compare
the
model
soil
loss
predictions
with
agricultural
runoff
(
soil
loss
and
runoff)
and
golf
course
runoff
(
reduced
soil
loss).
Both
have
ponds
and
should
provide
interesting
comparisons.
Golf
courses
with
ponds
and
a
vegetative
cover
throughout.
may
experience
a
greatly
reduced
loss
of
sediment
because
of
this
vegetative
cover.

REFERENCES
Bailer,
A.
J.
and
J.
T.
Oris.
1997.
Estimating
inhibition
concentrations
for
different
response
scales
using
generalized
linear
models.
Environ.
Toxicol.
Chem.
16:
1554­
1559.

Brooks,
R.
T.
and
M.
Hayashi.
2002.
Depth­
area­
volume
and
hydroperiod
relationships
of
ephemeral
(
vernal)
forest
pools
in
southern
New
England.
Wetlands
22(
2):
247­
255.

Clark,
J.
R.,
P.
W.
Borthwick,
L.
R.
Goodman,
J.
M.
Patrick,
E.
M.
Lores,
and
J.
C.
Moore.
1987
Comparison
of
Laboratory
Toxicity
Test
Results
with
Responses
of
Estuarine
Animals
Exposed
to
Fenthion
in
the
Field.
Environ.
Toxicol.
Chem.
6(
2):
151­
160.

Clark,
J.
R.,
L.
R.
Goodman,
P.
W.
Borthwick,
J.
M.
Patrick,
and
J.
C.
Moore.
1986.
Field
and
Laboratory
Toxicity
Tests
with
Shrimp,
Mysids,
and
Sheepshead
Minnows
Exposed
to
Fenthion.
Environmental
Research
Lab.,
Gulf
Breeze,
FL.
Report
Number:
EPA/
600/
D­
86/
036.
In:
Aquatic
Toxicology
and
Environmental
Fate:
Ninth
Volume,
ASTM
STP
921.
EPA/
600/
D­
86/
036.
T.
M.
Poston
and
R.
Purdy,
Editors.
American
Society
for
Testing
Materials,
Philadelphia,
PA.
Pp.
161­
176.
(
ERL,
GB539).
(
Avail.
from
NTIS,
Springfield,
VA:
PB86­
158649).

Dunne,
T.
and
R.
D.
Black.
1970.
Partial
area
contributions
to
storm
runoff
in
a
small
New
England
watershed.
Water
Resour.
Res.
6(
5):
1296­
1308.

Grist,
E.
P.
M.,
K.
M.
Y.
Leung,
J.
R.
Wheeler,
and
M.
Crane.
2002.
Better
bootstrap
estimation
of
hazardous
concentration
thresholds
for
aquatic
assemblages.
Environ.
Toxicol.
Chem.
24:
1515­
1524.

Hawkins,
R.
H.,
A.
T.
Hjemfelt
and
A.
W.
Zevenberger.
1985.
Runoff
probability,
storm
depth
and
curve
number.
J.
Irrigation
Drainage
Engineering
111(
4):
330­
340.

Hayashi
M.,
G.
van
der
Kamp.
2000.
Simple
equations
to
represent
the
volume­
areadepth
relations
of
shallow
wetlands
in
small
topographic
depressions.
J.
Hydrol.
237(
1­
2):
74­
85.

Hicks,
C.
R.
and
K.
V.
Turner,
Jr.
1999.
The
Taguchi
approach
to
the
design
of
experiments.
Chapter
14,
in
Fundamental
Concepts
in
The
Design
of
Experiments.
Oxford:
Oxford
University
Press.
42
of
44
Horton,
R.
E.
1933.
The
role
of
infiltration
in
the
hydrologic
cycle.
Trans.
AGU
14:
446
460.

Horton,
R.
E.
1940.
An
approach
toward
a
physical
interpretation
of
infiltration
capacity.
Soil
Sci.
Soc.
Amer.
Proceedings
4:
399­
417.

Hutchinson,
G.
E.
1957.
A
treatise
on
limnology,
[
by]
G.
Evelyn
Hutchinson.
Published:
New
York,
Wiley.

Kerr,
D.
R.
and
J.
P.
Meador.
1996.
Modeling
dose
response
using
generalized
linear
models.
Environmental
Toxicology
and
Chemistry
15:
395­
401.

Kleijnen,
J.
P.
C.
1997.
Sensitivity
analysis
and
related
analyses:
A
review
of
some
statistical
techniques.
Journal
of
Statistical
Computation
and
Simulation
57:
111­
142.

Kleijnen,
J.
P.
C.
2004.
An
overview
of
the
design
and
analysis
of
simulation
experiments
for
sensitivity
analysis.
European
Journal
of
Operational
Research,
in
press.
Available
online
at
http://
center.
uvt.
nl/
staff/
kleijnen/
ejor_
review_
proof.
pdf
Lyon,
S.
W.,
M.
T.
Walter,
P.
Gérard­
Marchant,
and
T.
S.
Steenhuis.
2004.
Using
a
topographic
index
to
distribute
variable
source
area
runoff
predicted
with
the
SCS­
curve
number
equation.
Hydrol.
Proc.
[
In
Press].

Moore,
D.
W.,
M.
D.
Schluchter,
and
G.
I.
Scott.
1990.
Use
of
Hazard
Models
in
Evaluating
the
Effect
of
Exposure
Duration
on
the
Acute
Toxicity
of
Three
Pesticides.
In:
Aquatic
Toxicology
and
Risk
Assessment:
Thirteenth
Volume,
ASTM
STP
1096.
EPA/
600/
A­
95/
010.
W.
G.
Landis
and
W.
H.
van
der
Schalie,
Editors.
American
Society
for
Testing
and
Materials,
Philadelphia,
PA.
pp.
247­
263.

Newman,
M.
C.,
D.
R.
Ownby,
L.
C.
A.
Mézin,
D.
C.
Powell,
T.
R.
L.
Christensen,
S.
B.
Lerberg,
B.
A.
Anderson,
and
T.
V.
Padma.
2001.
Species
sensitivity
distributions
in
ecological
risk
assessment:
Analysis
of
distributional
assumptions,
alternate
bootstrap
techniques,
and
estimation
of
adequate
number
of
species.
In:
The
Use
of
Species
Sensitivity
Distributions
in
Ecotoxicology.
CRC
Press
LLC,
Boca
Raton,
FL.
pp.
119­
132.

Pennington,
P.
L.,
J.
W.
Daugomah,
A.
C.
Colbert,
M.
H.
Fulton,
P.
B.
Key,
B.
C.
Thompson,
E.
D.
Strozier,
and
G.
I.
Scott.
2001.
Analysis
of
pesticide
runoff
from
mid­
Texas
estuaries
and
risk
assessment
implications
for
marine
phytoplankton.
Journal
of
Environmental
Science
and
Health
B
36(
1):
1­
14.

Price,
J.
S.
1993.
Water
level
regimes
in
prairie
sloughs.
Canadian
Water
Resources
Journal
18(
2):
95­
105.
43
of
44
Sacks,
J.,
W.
J.
Welch,
T.
J.
Mitchell,
and
H.
P.
Wynn.
1989.
Design
and
analysis
of
computer
experiments
(
with
discussion).
Statistical
Science
4:
409­
435.

Schulz,
R.
2004.
Field
studies
on
exposure,
effects,
and
risk
mitigation
of
aquatic
nonpoint­
source
insecticide
pollution:
a
review.
Journal
of
Env.
Quality
33:
419­
448.

Scott,
G.
I.,
D.
W.
Moore,
M.
H.
Fulton,
T.
W.
Hampton,
J.
M.
Marcus,
G.
T.
Chandler,
K.
L.
Jackson,
D.
S.
Baughman,
A.
H.
Trim,
C.
J.
Louden,
and
E.
R.
Patterson.
1989.
Agricultural
Insecticide
Runoff
Effects
on
Estuarine
Organisms:
Correlating
Laboratory
and
Field
Toxicity
Testing
With
Ecotoxicological
Biomonitoring.
Volumes
I
and
II.
2nd
Annual
Report.
US
Environmental
Protection
Agency,
Gulf
Breeze,
FL.
688pp.

Scott,
G.
I.,
M.
H.
Fulton,
M.
C.
Crosby,
P.
B.
Key,
J.
W.
Daugomah,
J.
T.
Walden,
E.
D.
Strozier,
C.
J.
Louden,
G.
T.
Chandler,
T.
F.
Bidleman,
K.
L.
Jackson,
T.
W.
Hampton,
T.
Hoffman,
A.
Shultz
and
M.
Bradford.
1992.
Agricultural
nonpoint
runoff
effects
on
estuarine
organisms:
correlating
laboratory
and
field
bioassays
and
ecotoxicological
biomonitoring.
Final
Report.
U.
S.
Environmental
Protection
Agency,
Gulf
Breeze,
FL.
281pp.

Steenhuis,
T.
S.,
M.
Winchell,
J.
Rossing,
J.
A.
Zollweg,
and
M.
F.
Walter.
1995.
SCS
runoff
equation
revisited
for
variable­
source
runoff
areas.
Journal
of
Irrigation
and
Drainage
Engineering­
ASCE
121(
3):
234­
238.

Su
M.,
W.
J.
Stolte
and
G.
van
der
Kamp.
2000.
Modelling
Canadian
prairie
wetland
hydrology
using
a
semi­
distributed
streamflow
model.
Hydrol.
Process
14(
14):
2405­
2422.

Taguchi,
G.
1986.
Introduction
to
Quality
Engineering.
Tokyo:
Asian
Productivity
Organization.

Thornthwaite,
C.
W.
and
J.
R.
Mather.
1955.
The
water
balance.
Publ.
Climatol.
8(
1).

Thornthwaite,
C.
W.
and
J.
R.
Mather.
1957.
Instructions
and
tables
for
computing
potential
evapotransporation
and
the
water
balance.
Publ.
Climatol.
10(
3).

USDA.
1997.
SCS
Agricultural
Handbook
number
590.
Ponds­
Planning,
Design
Construction
(
revised).

Wauchope,
R.
D.,
H.
R.
Summer,
C.
C.
Truman,
A.
W.
Johnson,
C.
C.
Dowler,
J.
E.
Hook,
G.
J.
Gascho,
J.
G
Davis,
and
L.
D.
Chandler.
1999.
Runoff
from
a
cornfield
as
affected
by
tillage
and
corn
canopy:
A
large­
scale
simulated­
rainfall
hydrologic
data
set
for
model
testing.
Water
Resources
Research
35(
9):
2881­
2885.

Wiens,
L.
H.
2001.
A
Surface
Area­
Volume
Relationship
for
Prairie
Wetlands
in
the
44
of
44
Upper
Assiniboine
River
Basin,
Saskatchewan.
Canadian
Water
Resources
Journal
26(
4):
503­
513.
