1
Fax­
On­
Demand
Fax
Number:
(
202)
401­
0527
Item:
6044
ESTIMATING
THE
DRINKING
WATER
COMPONENT
OF
A
DIETARY
EXPOSURE
ASSESSMENT
OFFICE
OF
PESTICIDE
PROGRAMS
ENVIRONMENTAL
PROTECTION
AGENCY
(
November
2,
1999)

I.
EXECUTIVE
SUMMARY
The
Office
of
Pesticide
Programs
(
OPP),
EPA,
is
issuing
a
revised
science
policy
paper
to
describe
changes
in
its
approach
to
estimate
pesticide
concentrations
in
drinking
water
as
part
of
its
assessment
of
dietary
exposures
to
pesticides.
This
document
was
developed
from
the
Agency's
science
policy
paper
entitled
Science
Policy
5:
Estimating
the
Drinking
Water
Component
of
a
Dietary
Exposure
Assessment
(
12/
22/
98
Draft),
that
was
released
for
public
comment
on
January
4,
1999
(
64
FR
162).
The
Agency
received
comments
from
various
organizations.
Each
of
the
commentors
offered
recommendations
for
improving
the
science
policy.
All
comments
were
extensively
evaluated
and
considered
by
the
Agency.
This
revised
version
embodies
many
recommendations
of
the
commentors,
as
well
as
recommendations
from
a
May
1999
FIFRA
Scientific
Advisory
Panel
which
evaluated
the
proposed
approach
for
incorporating
a
"
crop
area
adjustment
factor"
along
with
a
drinking
water
reservoir
scenario
in
the
Agency's
surface
water
screening
models.
The
public
comments,
as
well
as
a
detailed
summary
of
the
Agency's
response
to
the
comments
are
also
available
in
the
docket
for
this
notice.

For
some
time
the
Agency
has
been
using
screening
models
to
estimate
pesticide
concentrations
in
ground
water
and
surface
water
to
identify
those
food­
use
pesticides
that
are
not
expected
to
contribute
enough
exposure
via
drinking
water
to
result
in
unacceptable
levels
of
aggregate
risk.
The
Agency
uses
monitoring
data,
where
available
and
reliable,
to
refine
its
assessments
in
those
cases
where
the
use
of
the
screening
models
does
not
result
in
"
clearing"
the
pesticide
(
i.
e.,
indicates
a
low
risk)
from
a
drinking
water
perspective.
This
paper's
description
of
the
models
and
approaches
EPA
generally
intends
to
follow
is
not
meant
to
restrict
interested
parties
from
commenting
on
the
appropriateness
of
these
models
and
approaches,
either
generally
or
in
regard
to
a
specific
application,
or
from
proposing
new
or
different
models
or
approaches.

In
response
to
public
comments,
OPP
made
a
number
of
significant
changes
to
its
drinking
water
assessment
approaches,
primarily
to
refine
existing
screening
methods
for
identifying
pesticides
which
may
be
present
in
drinking
water
at
levels
of
concern.
These
refinements
will
2
enable
OPP
to
more
accurately
determine
whether
a
pesticide
has
the
potential
to
result
in
significant
risks
to
the
public
and
sensitive
populations
such
as
infants
and
children.
Specifically,
in
1999,
OPP
will
change
its
screening
level
drinking
water
assessment
by
replacing
the
"
farm
field
pond"
scenario
in
its
surface
water
screening
models
with
a
"
drinking
water
reservoir"
scenario
and
will
begin
incorporating
into
the
model
a
factor
to
account
for
the
area
surrounding
the
reservoir
that
is
cropped.
To
start,
percent
cropped
area
factors
will
be
used
for
corn,
soybeans,
cotton,
and
wheat.
Additional
factors
for
other
major
crops
will
be
added
in
late
1999
and
early
2000.
These
changes,
which
better
represent
actual
drinking
water
conditions,
will
improve
the
accuracy
of
EPA's
initial
screening
assessment.
The
Agency
is
also
evaluating
several
watershed­
scale
surface
water
models
for
use
in
future
drinking
water
assessments.

EPA
will
continue
to
use
SCI­
GROW
(
Screening
Concentration
In
GROund
Water)
as
an
initial
screening
model
for
ground
water
sources
of
drinking
water.
An
evaluation
of
models
and
procedures
for
a
second­
tier
assessment
of
pesticide
exposure
in
ground
water
is
beginning.
In
the
meantime,
the
Agency
will
rely
on
ground
water
monitoring
studies
to
estimate
concentrations
in
ground
water
for
those
pesticides
which
do
not
pass
through
the
SCI­
GROW
screen.

The
Agency
believes
its
risk
assessments
would
be
strengthened
by
additional
monitoring
data
and
is
working
on
a
number
of
levels
to
fill
in
the
gaps
in
monitoring
data
and
acquire
more
high
quality
data
on
pesticide
concentrations
in
drinking
water
sources.
Efforts
include
requesting
monitoring
and
runoff
studies
on
individual
pesticides,
working
with
the
U.
S.
Geological
Survey
(
USGS)
to
obtain
more
regional­
and
national­
scale
monitoring
data
on
multiple
pesticides,
and
exploring
design
considerations
for
a
national
survey
of
pesticides
in
drinking
water
with
various
government
agencies
and
industry
groups
and
associations.

As
a
result
of
the
comments,
OPP
has
identified
two
issues
regarding
drinking
water
that
will
be
addressed
in
separate
science
policy
papers
within
the
next
six
months.
In
an
effort
to
better
estimate
pesticide
concentrations
in
tap
water,
the
Agency
will
issue
a
paper
on
the
effectiveness
of
water
treatment
in
reducing
pesticide
levels
in
drinking
water
and
an
approach
for
addressing
treatment
issues
in
the
assessment
process.
EPA
also
plans
to
issue
for
public
comment
a
paper
on
approaches
for
utilizing
available
data
and
models
to
develop
quantitative
estimates
of
pesticide
concentrations
in
drinking
water
and
estimates
of
people
exposed
for
pesticides
which
pose
a
particularly
high
potential
for
contaminating
drinking
water.

II.
BACKGROUND
A.
Why
is
EPA
concerned
about
including
exposure
to
pesticides
in
drinking
water
in
its
decisions
about
acceptable
levels
of
pesticides
on
food?

With
the
passage
of
the
Food
Quality
Protection
Act
(
FQPA)
in
August
1996,
Congress
directed
EPA
to
consider
"
all
anticipated
dietary
exposures
and
all
other
exposures
for
which
there
is
reliable
information"
in
determining
whether
pesticide
residues
in
food
are
safe.
EPA
considers
drinking
water
to
be
an
anticipated
dietary
exposure
route
for
certain
pesticides.
Because
a
number
of
pesticides
have
been
found
in
ground
water
and
surface
water
throughout
the
United
States,
EPA
will
continue
to
address
pesticide
exposure
through
drinking
water
in
the
EPA
uses
a
"
reference
dose"
concept
to
represent
the
sum
of
exposures
from
dietary
and
1
non­
dietary
sources
that,
together,
do
not
exceed
a
maximum
safe
daily
intake.
Each
source
of
pesticide
exposure
(
food,
residential
exposure,
and
drinking
water
for
each
pesticide
use)
creates
a
risk
that
consumes
part
of
the
reference
dose.
The
reference
dose
for
a
pesticide
may
allow
for
a
number
of
crop­
specific
uses
as
long
as
the
aggregate
exposure
and
risk
from
all
of
those
uses
do
not
exceed
the
maximum
safe
daily
intake.
Reference
doses
have
been
established
for
short­
term
exposure
(
days
to
weeks),
as
well
as
for
lifetime
exposure.

3
aggregate
exposure
assessment
process.
The
picture
emerging
from
available
federal,
state
and
local
water
monitoring
efforts
is
complex.
Typically,
a
mix
of
pesticides
is
detected
in
water
at
low
levels
with
seasonal
pulses
of
higher
concentrations.
Of
the
major
sources
of
monitoring
data
that
OPP
routinely
uses
 
the
United
States
Geological
Survey's
(
USGS)
National
Water
Quality
Assessment
Program
(
NAWQA),
Toxic
Substances
Hydrology
Program
(
TSHP),
and
National
Stream
Quality
Accounting
Network
(
NASQAN);
and
the
EPA's
National
Pesticide
survey
 
a
majority
of
the
streams
(
up
to
95
percent)
and
half
of
the
wells
near
agricultural
and
urban
areas
contain
detectable
levels
of
at
least
one,
and
often
two
or
more,
pesticides.
Most
ground
water
aquifers
and
half
of
the
streams
investigated
by
these
programs
are
direct
sources
of
drinking
water.

Prior
to
FQPA,
OPP's
strategy
for
managing
pesticides
which
had
the
potential
to
contaminate
water
was
to
emphasize
prevention
 
requiring
mitigation
measures
such
as
geographic
restrictions
on
pesticide
use
and
"
buffer
zones"
near
water
bodies
where
pesticide
use
is
prohibited.
Since
FQPA,
OPP
has
routinely
assessed
exposure
to
pesticides
in
drinking
water
as
a
part
of
its
dietary
risk
assessments
process.

B.
What
has
been
EPA's
general
process
since
the
passage
of
the
FQPA
for
including
drinking
water
exposure
in
its
decisions
about
acceptable
levels
of
pesticides
on
food?

August
1996­
November
1997
While
it
developed
a
more
science­
based
policy
for
estimating
drinking
water
exposure
and
for
interpreting
available
monitoring
data
in
the
initial
months
after
the
enactment
of
FQPA,
OPP
used
an
interim
approach
which
assumed
that
up
to
10%
of
what
it
considered
acceptable
exposure
to
a
pesticide
could
occur
via
the
drinking
water
route
(
PR
Notice
97­
1).
Therefore,
OPP
reserved
10%
of
the
"
reference
dose
"
for
drinking
water
related
risks
and
allowed
food
1
residues
and
other
routes
of
exposure
to
consume
no
more
than
90%
of
the
reference
dose.
This
10%
value
for
drinking
water
was
a
default
assumption
that
OPP
knew
was
likely
to
overestimate
actual
exposure
in
many
cases,
while
potentially
underestimating
actual
exposures
in
some
others.

Overview
of
EPA's
Approach
Since
November
1997
In
November
of
1997,
OPP
ceased
using
the
10%
default
assumption
and
formally
adopted
the
following
interim
process
for
addressing
drinking
water
exposures
for
all
pesticides
with
outdoor
uses.
"
High­
end"
refers
to
a
combination
of
events
and
conditions
such
that,
taken
together,
2
produces
conceivable
risk
to
greater
than
90
percent
of
the
population
subject
to
the
risk
assessment,
but
less
than
the
maximally
exposed
population.
"
High­
end"
is
defined
in
"
Guidelines
for
Exposure
Assessment,"
FRN
Vol.
57,
No.
104,
Fri,
May
29,
1992,
and
"
Guidance
on
Risk
Characterization
for
Risk
Managers
and
Risk
Assessors,"
F.
Henry
Habicht
II,
Deputy
Administrator,
U.
S.
EPA,
to
Assistant
and
Regional
Administrators,
U.
S.
EPA,
Feb.
26,
1992.

The
Drinking
Water
Level
of
Comparison
(
DWLOC)
is
the
theoretical
concentration
of
a
3
chemical
in
drinking
water
that
would
be
acceptable
as
an
upper
limit
in
light
of
total
aggregate
exposure
to
that
chemical
from
food,
water,
and
non­
occupational
(
residential)
sources.
It
is
based
on
the
difference
between
the
maximum
daily
intake
(
the
reference
dose)
and
the
sum
of
the
exposure
from
food
and
residential
sources.
OPP
originally
used
the
term
"
Drinking
Water
Level
of
Concern,"
but
felt
that
this
term
conveyed
more
regulatory
significance
than
is
intended.
The
DWLOC
is
not
a
regulatory
standard
for
drinking
water,
but
is
the
theoretical
upper
limit
of
"
acceptable"
exposure
after
considering
food
and
residential
exposures
as
sources.

To
assure
the
protection
of
infants/
children
and
adults,
and
to
assure
consistency
across
the
4
Agency,
OPP
has
adopted
the
same
assumptions
(
bodyweights,
daily
intakes
and
the
percentiles
of
consumption
for
children
and
adults)
that
are
used
by
EPA's
Office
of
Drinking
Water
in
setting
national
drinking
water
standards.
Source:
"
Exposure
Factors
Handbook,
Vol.
1,
General
Factors,"
August,
1997,
Office
of
Research
and
Development,
EPA/
600/
P­
95/
002Fa.

4
1.
OPP
scientists
review
all
available
laboratory
and
field
data
submitted
by
the
registrant
to
determine
whether
a
particular
pesticide
will
easily
move
to
ground
water
or
surface
water,
will
degrade
quickly
or
persist,
and
will
form
toxic
breakdown
products
as
it
degrades.

2.
OPP
uses
pesticide­
specific
data
from
these
studies
in
mathematical
screening
models
to
estimate
pesticide
concentrations
in
water
in
pesticide
use
areas.
Peer
reviews
of
these
models
(
section
II.
C.)
generally
support
OPP's
view
that
the
estimates
coming
out
of
these
models
are
high­
end
estimates
of
potential
pesticide
concentrations
in
drinking
2
water
derived
from
the
upper
regions
of
major
watersheds.

3.
OPP
compares
the
screening
estimates
to
human
health­
based
"
drinking
water
levels
of
comparison"
(
DWLOC)
,
which
are
derived
after
first
considering
all
food­
related
and
3
residential
exposures
for
which
EPA
has
reliable
information.
Specifically,
OPP
compares
the
estimated
potential
daily
intake
of
pesticide
residues
by
a
10­
kg
child
(
age
12
months)
consuming
1
liter
of
water
per
day
(
approximates
90th
percentile)
and
by
a
70­
kg
adult
consuming
2
liters
of
water
(
approximates
80th
percentile)
per
day
to
the
daily
intake
that
would
be
permitted
by
the
DWLOC.
This
comparison
determines
whether
OPP
clears
4
the
pesticide
from
a
drinking
water
perspective
or
attempts
to
refine
its
estimates
of
pesticide
concentrations
in
drinking
water
to
reflect
more
representative
and
realistic
conditions.
In
some
cases,
the
DWLOC
may
be
very
low,
not
because
the
pesticide
is
particularly
toxic,
but
because
contributions
from
food­
related
uses
and
other
pathways
of
exposure
are
so
great
that
very
little
of
the
reference
dose
or
target
exposure
is
available
to
allow
for
any
exposure
via
drinking
water.
Alternatively,
some
pesticides
(
particularly
5
newer
pesticides)
may
have
a
very
high
DWLOC
solely
because
they
have
very
few
food
uses
or
other
uses
which
result
in
exposure,
leaving
the
majority
of
the
reference
dose
available
for
exposures
through
drinking
water.
If
additional
uses
are
added
for
a
pesticide,
the
DWLOC
will
decrease
in
relation
to
the
exposure
added
from
the
new
uses.

4.
In
those
instances
where
the
model
estimates
suggest
a
potential
for
concern
(
i.
e.,
estimated
exposure
exceeds
the
DWLOC),
additional
steps
taken
by
the
Agency
are
determined
on
a
case­
by­
case
basis,
depending
on
how
much
monitoring
data
are
available
and
the
extent
of
available
information
on
use
and
management
practices,
which
are
pesticide­
specific.
These
additional
steps
focus
on
gathering
more
information
to
reduce
the
uncertainty
in
the
drinking
water
estimates,
analyzing
and
evaluating
existing
monitoring
data,
or
requesting
additional
monitoring
data
that
can
be
related
to
drinking
water
sources.

5.
If
monitoring
data
are
not
available
or
are
not
sufficient
for
purposes
of
refining
the
screening
level
estimates,
OPP
will
first
attempt
to
reduce
the
uncertainty
in
screening
model
estimates
by
requesting
additional
pesticide
fate
data
to
fill
in
any
gaps
in
the
model
inputs.
The
Agency
also
evaluates
pesticide
usage
data
in
relation
to
potential
drinking
water
sources
to
determine
the
potential
for
a
pesticide
to
reach
drinking
water.
Such
refinements
would
be
used
to
determine
whether
ground
water
and/
or
surface
water
monitoring
is
needed.
Generally,
OPP
does
not
base
significant
risk
management
action
(
e.
g.,
revocation
or
denial
of
a
tolerance)
solely
on
screening
model
estimates.
However,
OPP
uses
screening
models
to
judge
the
effectiveness
of
management
options
that
may
be
taken
to
reduce
potential
exposures
to
drinking
water.

6.
If
sufficient
and
reliable
monitoring
data
are
available,
OPP
scientists
analyze
the
data
and
consult
with
risk
managers
as
to
how
the
data
fit
specific
risk
endpoints
being
addressed
in
the
human
health
risk
assessment.
The
Agency
evaluates
the
monitoring
data
in
relation
to
accompanying
information
on
spatial
and
temporal
distribution
of
the
sample
points,
the
water
body
(
or
bodies)
represented
by
the
sampling,
the
characteristics
of
the
site
surrounding
the
water
body,
and
the
weather/
environmental
conditions
represented
by
the
study.
EPA
also
attempts
to
determine
whether
the
sampled
water
bodies
represent
real
or
likely
drinking
water
sources
and
whether
the
data
represent
potentially
vulnerable
sites.
If
the
evaluation
indicates
that
the
data
represent
drinking
water
sources,
then
such
data
may
be
used
quantitatively
in
aggregate
exposure
assessments.
Appropriate
shortterm
(
for
acute
effects)
and/
or
longer­
term
average
(
for
chronic
effects
or
cancer)
drinking
water
concentrations
are
selected.
In
keeping
with
the
Agency
definition
of
"
high­
end"
exposure
estimates
(
refer
to
footnote
2),
the
Agency
selects
a
"
high­
end"
value,
but
not
necessarily
the
highest
monitoring
data
value,
for
use
in
its
risk
assessments.
The
values
from
monitoring
data
used
in
the
human
health
risk
assessment
are
usually
less
than
the
model
estimates
but,
in
a
few
cases,
may
be
greater
than
that
predicted
by
OPP's
screening
models.

7.
Estimates
of
pesticide
concentrations
in
drinking
water,
derived
from
monitoring
data,
are
combined
with
estimates
of
water
consumption
to
estimate
human
exposure
via
drinking
For
a
more
detailed
description
of
these
screening
models
and
their
use
in
the
drinking
water
5
assessments,
see
the
SAP
documents
(
1997
and
1998)
listed
in
the
bibliography.

6
water.
This
estimate
of
exposure
is
then
added
to
estimates
of
food
and
residential
exposure
to
complete
the
aggregate
exposure
assessment.

8.
If
sufficient
to
do
so,
the
monitoring
data
may
be
used
to
produce
a
regionally­
based
picture
of
the
distribution
of
measurements.
However,
this
is
rarely
possible
due
to
the
variability
and
uncertainty
associated
with
existing
monitoring
data
and
the
lack
of
an
extensive
monitoring
data
base
for
most
pesticides.

C.
EPA's
Use
of
Screening
Models
to
Estimate
Pesticide
Concentrations
in
Drinking
Water
1.
Surface
Water
Screening
Models
OPP
uses
two
mathematical
screening
models
to
assess
whether
pesticides
are
likely
or
unlikely
to
occur
at
significant
levels
in
drinking
water
derived
from
surface
water
.
The
model
5
GENEEC
(
GENeric
Estimated
Environmental
Concentrations)
provides
an
initial
screening
level
assessment
of
pesticide
concentrations
in
surface
water
while
the
linked
Pesticide
Root
Zone
Model
(
PRZM)
and
EXposure
Analysis
Model
System
(
EXAMS)
models
provide
a
more
refined
screen.
GENEEC
and
PRZM/
EXAMS,
initially
used
by
OPP
for
ecological
risk
assessments,
are
the
only
mechanistic
models
available
to
OPP
for
rapidly
and
cost­
effectively
producing
"
highend
estimates
of
pesticide
levels
in
surface
water.

GENEEC
uses
readily­
available
pesticide
properties
to
estimate
peak
and
time­
averaged
pesticide
concentrations
in
a
"
field
pond,"
20
million
liters
(
5.3
million
gallons)
in
capacity,
located
at
the
edge
of
a
10­
hectare
(
approximately
25
acres)
treated
cotton
field.
The
GENEEC
model
assumes
that
no
buffer
exists
between
the
pond
and
the
treated
field,
that
runoff
exactly
equals
water
losses
due
to
evaporation,
and
that
the
pesticide
is
uniformly
mixed
throughout
the
water
body.
The
model
is
likely
to
overestimate
pesticide
concentrations
in
drinking
water
because
it
is
designed
to
represent
a
water
body
in
the
upper
reaches
of
the
watershed
while
drinking
water
intakes
are
typically
found
lower
in
the
watershed
and
receive
drainage
from
a
greater
land
area.
Pesticide
concentrations
in
water
bodies
in
the
upper
reaches
of
agricultural
watersheds
are
generally
greater
because
there
may
be
less
dilution
from
non­
agricultural
runoff.
For
this
reason,
screening
model
estimates
have
generally
been
considered
to
be
"
bounding"
estimates
for
drinking
water
for
the
purpose
of
comparison
to
a
DWLOC.
GENEEC
simulates
a
single
pesticide
application
or
series
of
applications
to
bare
soil
followed
by
a
single
rainfall
event
two
days
after
the
final
application.
Depending
on
the
propensity
of
the
pesticide
to
move
into
water
or
stay
with
the
soil,
this
storm
will
wash
up
to
10%
of
the
pesticide
remaining
in
the
top
inch
of
soil
at
the
time
of
the
storm
into
the
pond.

If
the
surface
water
estimates
using
GENEEC
do
not
exceed
the
DWLOC,
then
OPP
concludes
that
the
pesticide
is
not
expected
to
pose
an
unacceptable
risk
from
combined
food
and
7
drinking
water
exposure
as
a
consequence
of
runoff
into
surface
water
and
no
further
evaluation
of
surface
water
exposure
is
necessary.
If
the
GENEEC
results
indicate
a
potential
concern,
then
the
coupled
PRZM
and
EXAMS
models
are
used
to
refine
the
estimates
of
potential
pesticide
levels
in
surface
water
by
including
more
pesticide­
specific
properties,
simulating
multiple
years
to
reflect
climatic
variations,
and
modeling
on
a
crop­
specific
basis.
In
comparison
to
GENEEC,
PRZM/
EXAMS
includes
more
site­
specific
information
in
the
scenario
details
regarding
application
method
and
temporal
distribution
with
weather,
and
better
accommodates
chemicalspecific
parameters.
However,
it
still
uses
the
same
20­
million
liter
pond,
which
represents
a
water
body
in
an
upland
area
from
which
few
people
would
derive
their
drinking
water.
Thus,
having
a
body
of
water
which
is
more
reflective
of
drinking
water
sources
is
an
important
revision
to
EPA's
drinking
water
exposure
assessment
(
see
Section
III.
A.
1).

2.
Ground
Water
Screening
Model
OPP
developed
SCI­
GROW
(
Screening
Concentration
In
GROund
Water)
as
an
initial
screening
model
to
estimate
pesticide
concentrations
in
ground
water
under
reasonable,
vulnerable
conditions.
SCI­
GROW
was
developed
by
comparing
selected
pesticide
properties
to
pesticide
concentrations
measured
in
ten
prospective
ground­
water
monitoring
studies
conducted
by
pesticide
registrants.
The
studies
were
conducted
by
applying
the
pesticide
at
maximum
allowed
rates
and
frequency
to
hydrogeologically­
vulnerable
sites
(
i.
e.,
shallow
aquifers,
sandy,
permeable
soils,
and
substantial
rainfall
and/
or
irrigation
to
maximize
leaching).
The
highest
three
consecutive
monthly
data
points
from
a
selected
well
in
each
study
were
averaged
to
represent
90­
day
peak
pesticide
concentrations.
A
predictive
equation,
adjusted
for
the
application
rate,
was
developed
by
comparing
the
90­
day
peak
ground­
water
concentrations
to
a
pesticide
leaching
potential
index
that
is
based
on
its
persistence
in
soil
(
half­
life)
and
affinity
to
adsorb
to
soil
(
soilwater
partitioning
coefficient).

Using
data
on
pesticide
persistence
(
in
particular,
soil
metabolism
half­
life
values)
and
soil
adsorption,
and
the
application
rate,
SCI­
GROW
estimates
the
concentration
of
a
pesticide
in
shallow
ground
water
(
average
depth
15
feet)
beneath
sandy,
highly
permeable
soils.
If
the
ground
water
estimates
using
SCI­
GROW
do
not
exceed
the
DWLOC,
then
OPP
concludes
that
the
pesticide
is
not
expected
to
pose
an
unacceptable
risk
as
a
consequence
of
leaching
into
ground
water
and
no
further
evaluation
of
ground
water
exposure
is
necessary.
If
the
SCIGROW
results
indicate
a
potential
concern,
OPP
reviews
available
monitoring
data
for
exposure
refinement
and,
if
necessary,
requests
additional
information,
usually
in
the
form
of
prospective
ground
water
monitoring
studies.

D.
EPA's
Approach
to
Evaluating
and
Incorporating
Drinking
Water
Monitoring
Data
into
Human
Health
Risk
Assessments
If
the
estimates
of
pesticide
concentrations
in
drinking
water
from
modeled
surface­
water
sources
(
using
GENEEC
or
PRZM/
EXAMS)
or
ground­
water
sources
(
using
SCI­
GROW)
do
not
exceed
the
DWLOC,
then
OPP
concludes
that
the
pesticide
is
not
expected
to
pose
an
unacceptable
risk
via
exposure
to
drinking
water
and
no
further
evaluation
is
necessary.
However,
if
any
of
the
model
estimates
exceed
the
DWLOC,
OPP
gathers
additional
data
in
order
8
to
refine
model
estimates
as
well
as
available
water
monitoring
data
and
uses
these
data
to
characterize
the
anticipated
human
exposure
to
the
pesticide
via
drinking
water.
By
the
time
a
pesticide
reaches
this
stage
of
review,
OPP
believes
that
the
pesticide
has
some
potential
to
reach
surface
water
and/
or
ground
water
and
that
it
has
some
potential
to
be
present
at
levels
of
concern
to
human
health.

Typical
sources
of
monitoring
data
include
USGS's
NAWQA,
NASQAN,
and
Toxic
Substances
Hydrology
programs
(
USGS,
1998),
EPA
Office
of
Water's
STORET
database
(
US
EPA
OW,
1998),
OPP's
Pesticides
in
Ground
Water
Data
Base
(
US
EPA
OPP,
1992),
and
the
National
Pesticide
Survey
(
US
EPA,
1990).
OPP
may
also
seek
additional
water
monitoring
data
from
open
literature,
state
agencies,
or
other
monitoring
studies
such
as
the
Lake
Erie
Basin
data
collected
at
Heidelburg
(
OH)
College,
the
Acetochlor
Registration
Partnership
(
ARP)
study,
and
Novartis
Crop
Protection's
Atrazine
Volunteer
Monitoring
program.

The
availability
of
adequate
temporal
and
spatial
monitoring
data
can
reduce
uncertainty
associated
with
models,
and
can
provide
a
more
accurate
estimate
of
the
distribution
of
drinking
water
concentrations
in
areas
of
use.
In
a
few
cases,
EPA
will
have
considerable
water
monitoring
data
available
for
a
particular
pesticide,
including
registrant­
sponsored
monitoring
studies
and
monitoring
data
from
state,
local
and
federal
programs.
Nevertheless,
even
when
such
data
are
available,
they
may
have
been
collected
in
a
manner
that
limits
their
usefulness
for
estimating
the
distribution
of
drinking
water
concentrations
in
areas
of
use.
Therefore,
EPA
must
exercise
considerable
judgement
concerning
the
best
use
and
interpretation
of
these
data,
and
how
to
interpret
exposures
and
risk
estimates
calculated
from
them.
This
is
particularly
true
when
trying
to
characterize
exposures
from
a
region
where
there
may
be
more
than
one
source
of
water
monitoring
data.

In
evaluating,
characterizing
and
interpreting
water
monitoring
data,
OPP
scientists
attempt
to
collect
as
much
information
as
is
readily
available
on
the
design
of
the
studies.
That
is,
OPP
scientists
try
to
determine
how
the
samples
were
collected
and
analyzed,
why
they
were
collected,
and
where
and
when
they
were
collected.
To
complete
the
FQPA
assessment,
OPP
scientists
review
the
reliability/
validity
of
the
monitoring
data
and
present
the
range
of
values
reported,
the
highest
values
reported,
various
return
frequencies
(
e.
g.,
1
in
10
year
concentration),
and
the
mean
and
median
values.
If
OPP
has
adequate
data
to
produce
a
regional
"
picture"
of
the
distribution
of
reported
values,
this
is
completed
as
well.

Because
of
the
level
of
variability
and
uncertainty
associated
with
existing
monitoring
data,
the
selection
of
a
value
or
values
to
be
incorporated
into
the
human
health
risk
assessment
can
be
challenging.
Sometimes
valid
reported
values
vary
from
one
region
to
another
by
several
orders
of
magnitude.
Without
having
specific
information
on
the
history
of
the
use
of
the
pesticide
in
the
sampled
area,
it
is
very
difficult
to
fully
understand
the
reasons
for
these
differences.
Interpreting
the
results
of
studies
which
include
a
large
number
of
samples
with
no
residues
(
i.
e.,
"
non­
detects")
poses
additional
difficulties.
Non­
detects
can
indicate
that
the
pesticide
of
concern
is
not
reaching
the
drinking
water
source.
However,
non­
detects
can
also
result
when
the
samples
are
taken
from
areas
where
the
pesticide
is
not
applied
or
at
times
when
the
pesticide
is
not
being
used.
Limitations
with
analytical
methods
may
also
result
in
non­
detects
9
(
i.
e.,
the
pesticide
may
be
present
in
the
water
at
concentrations
that
are
less
than
the
limits
of
detection
for
the
analytical
method).
For
these
reasons,
the
Agency
must
consider
such
information
in
interpreting
non­
detects
in
monitoring
data
sets.
EPA
often
lacks
data
to
verify
that
reported
"
non­
detects"
were
in
actual
areas
of
use
and,
thus,
has
difficulty
concluding
that
the
pesticide,
when
used,
is
not
in
fact
reaching
water
frequently
enough
to
be
of
concern.
Further,
EPA
is
not
always
able
to
discern
whether
samples
were
taken
from
potential
drinking
water
sources
or
waters
that
would
be
representative
of
such
drinking
water
sources.

Despite
the
challenge
of
analyzing
and
interpreting
these
data,
OPP
will
choose
a
value
or
values
from
valid
monitoring
data,
when
reliable
data
are
available,
to
make
decisions
in
the
human
health
risk
assessment.
Values
have
been
chosen
from
valid
monitoring
data
even
when
the
data
were
limited
in
time
or
location.
As
OPP
has
gained
experience
in
reviewing
and
incorporating
monitoring
data
into
its
risk
assessments,
it
has
generally
chosen
"
reasonable,
highend
monitoring
values.
That
is,
OPP
has
selected
a
value
on
the
"
high
end"
(
as
defined
in
footnote
2)
of
the
range
rather
than
the
highest
measured
value.
This
"
high
end"
estimate
is
characterized
in
terms
of
its
representativeness
to
drinking
water
sources
and
the
degree
of
uncertainty
present
in
the
estimate.
Such
characterizations
are
used
to
determine
the
reliability
of
the
estimate.
While
the
ultimate
goal
of
OPP
is
to
estimate
pesticide
concentrations
in
tap
water,
such
information
(
either
in
the
form
of
monitoring
data
or
of
the
effect
of
various
water
treatment
processes
on
pesticide
concentrations)
is
rarely
available.

E.
Workshops
and
Peer
Reviews
of
Screening
Models
OPP
has
sought
and
obtained
external
scientific
review
of
its
interim
approach
and
of
the
models
it
uses
to
complete
screening
level
assessments
from
both
the
FIFRA
Scientific
Advisory
Panel
(
SAP)
and
expert
panels
convened
by
the
International
Life
Science
Institute
(
ILSI).
Most
of
the
external
review
to
date
has
focused
on
evaluating
the
tools
and
methods
used
as
initial
screens
to
estimate
pesticide
concentrations
in
drinking
water.

1.
International
Life
Science
Institute
(
ILSI)
Risk
Science
Institute
OPP
is
working
through
ILSI
to
review
its
current
model
screening
approach
and
to
recommend
improvements
which
could
be
implemented
in
the
short
term
to
improve
the
accuracy
of
its
estimates.
This
cooperative
effort
is
also
evaluating
how
to
refine
screening
level
model
estimates
and
how
to
use
and
interpret
monitoring
data.
ILSI
is
an
independent,
nonprofit
foundation
established
to
advance
the
understanding
of
scientific
issues
related
to
nutrition,
food
safety,
toxicology,
and
the
environment.
Through
its
Risk
Science
Institute,
ILSI
brings
together
experts
from
academia,
industry,
government,
and
public
interest
groups
to
address
cutting­
edge
scientific
issues.
These
expert
groups
meet
in
sessions
open
to
the
public
and
prepare
reports
for
the
Agency
which
are
also
distributed
to
the
public.
In
October
1997,
ILSI
convened
a
working
group
of
scientists
with
expertise
in
the
fate,
transport
and
occurrence
of
pesticides
in
surface
water
and
ground
water
to
evaluate
OPP's
tools
and
methods
for
estimating
potential
concentrations
of
pesticides
in
drinking
water.

The
ILSI
working
group
concluded
that
(
ILSI,
1998):
10
°
Screening
tools
are
needed
to
quickly
identify
pesticides
and
pesticide
uses
that
are
unlikely
to
contaminate
drinking
water
AND
that,
in
general,
the
screening
models
being
used
by
OPP
are
of
the
appropriate
type
and
level
of
detail
to
rapidly
identify
pesticides
that
are
unlikely
to
occur
in
drinking
waters
above
a
level
of
concern;

°
Preliminary
evaluations
indicate
that
these
models
are
reliable
for
screening
purposes.
However,
documentation
and
testing
of
the
screening
models
against
field
observations
is
not
yet
sufficient
to
conclude
that
they
will
reliably
meet
this
objective.

°
The
screening
models
should
be
improved
so
that
non­
problem
pesticides
(
from
a
drinking
water
perspective)
can
be
more
accurately
identified
in
the
initial
screen.

The
ILSI
working
group
provided
recommendations
on
the
types
of
information
on
drinking
water
needed
to
complete
aggregate
exposure
assessments
in
its
April
2,
1998,
report,
Assessment
of
Methods
to
Estimate
Pesticide
Concentrations
in
Drinking
Water
Sources
(
ILSI,
1998).
The
ILSI
report
advised
that
work
toward
developing
probability
distributions
(
as
frequency
of
exceedance)
for
peak
and
long
term
average
drinking
water
concentrations
within
a
pesticide's
use
region(
s)
is
needed.
Ideally,
the
estimates
of
peak
and
long
term
average
concentrations
should
be
derived
from
full,
temporal
distributions
in
actual
drinking
water.
These
are
the
kinds
of
concentration
data
which
are
needed
for
inclusion
with
the
more
refined,
probabilistic
exposure
assessments
for
residues
on
food
performed
using
Monte
Carlo
analysis
methods.

2.
FIFRA
SAP
Review
In
December
1997,
OPP
presented
its
interim
methods
for
estimating
exposure
to
pesticide
residues
in
drinking
water
to
the
Federal
Insecticide,
Fungicide,
and
Rodenticide
Act
(
FIFRA)
Scientific
Advisory
Panel
(
SAP).
After
commending
OPP's
work,
the
SAP
encouraged
OPP
to
develop
a
longer
term
plan
for
improving
tools
and
methods
to
produce
more
refined
and
accurate
estimates
of
drinking
water
concentrations.
In
response
to
specific
questions
from
OPP,
the
SAP
provided
the
following
important
comments:

°
Many
Panel
members
agreed
that
SCI­
GROW
generates
appropriately
conservative
estimates
of
pesticide
concentrations
in
drinking
water
from
ground
water
sources
for
use
in
an
initial
screen.
Most
members
believed
the
estimates
need
to
be
further
tested
and
verified
against
monitoring
data.

°
Nearly
all
Panel
members
agreed
that
estimates
produced
by
GENEEC
are
most
likely
overly
conservative
and
that
some
adjustments
should
be
made
to
account
for
the
percent
cropped
area
around
a
water
body
and
the
percent
of
that
crop
treated
with
the
pesticide.

°
Most
Panel
members
considered
PRZM/
EXAMS
a
reasonable
second
tier
modeling
approach
for
refining
estimates
generated
using
GENEEC
because
of
its
use
of
more
specific
crop,
weather,
and
site
geophysical
data
and
more
extensive
use
of
pesticide
fate
and
transport
data.
However,
as
with
GENEEC,
many
Panel
members
recommended
11
incorporating
an
adjustment
factor
into
the
model
for
the
percentage
of
cropped
area
within
the
reservoir
drainage
area.
Additionally,
the
Panel
was
unanimous
in
recommending
a
rigorous
effort
to
validate
PRZM/
EXAMS
by
(
1)
comparing
model
results
with
data
from
monitoring
studies
to
determine
the
limitations
and
(
2)
performing
a
systematic
sensitivity
analysis
of
the
model
input
parameters.

°
The
Panel
stated
that
OPP
needs
to
develop
databases
and
methods
that
effectively
use
monitoring
both
in
assessments
and
model
validation.
OPP
should
(
1)
invest
time
and
resources
in
the
development
of
geographic
information
system
(
GIS)
tools
related
to
soil
type,
crop
coverage
and
water
monitoring
sampling
points;
(
2)
describe
and
document
all
variables
in
its
models
and
methods,
and
better
articulate
the
relative
impact
of
these
variables
on
its
drinking
water
assessment;
and
(
3)
compare
predictions
from
its
screening
models
with
monitoring
data
to
better
understand
how
these
relate.

In
July
1998,
OPP
presented
to
the
SAP
its
proposed
methods
for
basin­
scale
estimation
of
pesticide
concentrations
in
flowing
water
and
a
proposal
for
using
a
reservoirs
scenario
in
screening
level
assessments.
The
SAP
commended
OPP
for
developing
a
more
realistic
reservoir
modeling
scenario
as
an
improvement
over
the
"
pond"
scenario.
Additionally,
the
SAP
reacted
favorably
to
the
Agency's
proposed
strategy
to
develop
a
higher
tier
"
watershed­
scale"
model
for
use
in
FQPA
drinking
water
assessments.
In
response
to
specific
questions
from
OPP,
the
SAP
provided
the
following
comments
relative
to
the
Index
Drinking
Water
Reservoir
and
the
Basinscale
Model
Evaluation:

°
The
Panel
overall
agreed
with
the
Agency
on
moving
to
a
more
"
realistic"
approach
to
estimating
pesticide
concentrations
for
use
in
drinking
water
assessments
using
the
Index
Drinking
Water
Reservoir
(
IDWR)
scenario
and
agreed
that
the
approach
to
the
selection
of
the
IDWR
is
reasonable.
Several
Panel
members
noted
that
location
of
treated
fields
in
a
watershed
and
soil/
crop
management
factors
are
integral
to
potential
reservoir
contamination
and
should
be
considered
in
evaluating
the
IDWR
approach.

°
The
Panel
agreed
that
the
Agency
should
move
forward
in
utilizing
the
IDWR
and
encouraged
the
Agency
to
seek
further
scientific
peer
review
as
additional
refinements
are
made.
The
Panel
agreed
that
the
proposed
IDWR
scenario
is
conservative
even
though
the
conditions
of
nearly
30
percent
of
the
nation's
reservoirs
are
likely
to
be
more
conducive
to
pesticide
runoff
than
conditions
around
the
chosen
IDWR.
This,
they
concluded,
may
be
a
result
of
poor
reservoir
design,
thus,
these
systems
may
require
special
protection.

°
Most
Panel
members
concluded
that
the
Agency's
approach
to
evaluating
basin­
scale
models
was
sound
and
the
Panel
listed
five
basic
properties
that
should
be
considered
in
the
model
evaluation.
The
Panel
viewed
these
models
as
a
cost­
effective
means
of
providing
information
on
pesticide
concentrations
in
surface
water
given
the
complexities
in
monitoring
water
quality
for
a
broad
range
of
pesticides.

°
Panel
members
were
of
the
opinion
that
a
single
basin­
scale
model
may
not
be
sufficient
to
12
answer
all
the
Agency's
needs.
Each
model
with
its
inherent
strengths
and
weaknesses
may
be
applicable
to
a
subset
of
scenarios
and,
thus,
the
Agency
may
need
several
basinscale
models
in
its
modeling
suite
for
upper
tier
assessments.
Additionally,
the
Panel
agreed
that
the
two
monitoring
data
sets
on
which
model
performance
will
be
evaluated
are
adequate
for
the
initial
evaluation.
However,
because
a
complete
evaluation
cannot
be
accomplished
on
two
data
sets,
the
Panel
encouraged
the
Agency
to
continue
to
develop
monitoring
data
to
further
evaluate
basin­
scale
models.

In
May
1999,
OPP
presented
to
the
SAP
a
proposed
approach
for
incorporating
both
a
"
crop
area
adjustment
factor"
and
a
drinking
water
reservoir
scenario
in
the
PRZM
and
EXAMS
modeling.
The
Panel
agreed
that
application
of
a
"
Percent
Cropped
Area"
(
PCA)
factor
"
worked
reasonably
well
with
major
crops
in
the
Midwest
and
can
be
comfortably
applied
under
those
conditions."
The
SAP
recommended
that
more
consideration
be
given
to
low­
cost,
targeted
monitoring,
especially
in
the
case
of
minor­
use
crops
where
modeling
efforts
tend
to
be
imprecise.
The
Panel
identified
several
limitations
to
the
PCA
approach:

°
The
PCA
adjustment
may
underpredict
chemical
losses
as
some
croplands
contribute
disproportionately
to
runoff
(
e.
g.,
in
a
watershed
with
both
row
crops
and
forest,
the
cropland
would
contribute
more
significantly
to
runoff
and
watershed
discharge
and,
in
such
instances,
the
model
estimates
may
not
be
conservative
enough).

°
The
PCAs
were
derived
from
hydrologic
units
which
average
over
1,000
square
miles
in
size
and
county­
based
crop
acreages,
while
most
drinking
water
supplies
are
fed
by
smaller
watersheds.
The
discrepancies
between
calculated
PCA
values
and
the
PCA
of
the
actual
watershed
surrounding
a
reservoir
are
likely
to
be
more
of
a
problem
with
minor
crops
than
with
major
crops.

°
The
scale
of
the
watershed
does
not
allow
certain
factors
such
as
landscape
position,
soil
type,
or
slope
to
be
taken
into
account.
This
may
be
important
with
minor
crops
that
are
not
grown
on
the
typical
soil
type
modeled
in
the
watershed.

°
In
the
proposed
approach,
a
single
maximum
PCA
will
be
applied
universally
for
a
given
pesticide
across
all
regions
and
climatic
zones
while
the
model
will
use
region­
specific
soil
and
climatic
data.
For
consistency,
the
Panel
recommended
either
choosing
a
single
"
worst
case"
scenario
for
a
pesticide­
crop
combination
by
the
county
with
the
maximum
PCA
for
a
given
crop
or
doing
simulations
with
climatic
and
region­
specific
soil
and
PCA
data.

Most
Panel
members
agreed
that
the
Agency
should
consider
percent
crop
treated
in
future
model
refinements.
The
SAP
noted
that
relatively
high
uncertainties
may
be
encountered
for
chemicals
which
are
applied
to
less
than
10
%
of
the
cropped
area.
One
Panel
member
disagreed
with
the
recommendation,
commenting
that
the
use
of
percent
crop
treated
data
moves
beyond
the
original
intent
of
the
screening
approach.

The
May,
1999,
SAP
encouraged
the
Agency
to
continue
to
evaluate
watershed­
based
13
models,
as
suggested
by
the
July,
1998,
SAP.
The
Panel
recommended
that
the
Agency
consider
watershed­
based
models
in
a
GIS
environment.
While
watershed­
based
regression
models
"
may
be
appropriate
and
should
be
considered,"
the
Panel
cautioned
that
extrapolation
of
such
model
estimates
to
regions,
times,
and
site
conditions
beyond
the
range
under
which
the
model
was
developed
may
provide
inaccurate
estimates.

III.
POLICY
CHANGES
TO
BE
IMPLEMENTED
The
long­
term
goal
of
OPP
is
to
move
toward
the
use
of
probabilistic
drinking
water
exposure
assessments
for
regulatory
decisions
under
FQPA.
That
is,
OPP
wants
to
produce
information
on
the
number
of
people
likely
to
be
exposed
to
different
levels
of
pesticide
residues
in
drinking
water
and
use
this,
along
with
information
on
the
distribution
of
consumption
values
(
i.
e.,
the
number
of
people
who
drink
different
amounts
of
water
each
day),
to
generate
a
probabilistic
drinking
water
exposure
assessment.
OPP
also
wants
to
develop
watershed­
scale
models
for
use
in
refined
estimates
of
pesticide
concentrations
in
drinking
water
when
a
screening
level
model
estimate
indicates
a
significant
risk
may
exist.
However,
the
Agency
realizes
how
difficult
it
will
be
to
accomplish
both
of
these
goals.
Much
remains
to
be
done
to
develop
adequate
and
reliable
methods
and
the
data
necessary
to
use
these
approaches.
In
the
meantime,
OPP
is
considering
refinements
in
its
existing
screening
models
and
in
its
use
of
monitoring
data
for
estimating
concentrations
of
pesticides
in
drinking
water.

Mathematical
models
allow
OPP
to
rapidly
screen
pesticides
to
determine
whether
the
Agency
can
confidently
conclude
that
residues
are
unlikely
to
occur
in
drinking
water
at
levels
that
will
result
in
exceedances
of
the
DWLOC
(
when
combined
with
food
and
residential
exposure)
or
whether
the
Agency
needs
more
information
to
complete
an
assessment.
The
screening
model
estimates
need
to
be
protective
to
minimize
the
potential
for
"
passing"
a
pesticide
which
may
pose
a
concern
while
not
being
overly
protective
such
that
those
pesticides
which
will
truly
pose
no
concern
would
fail
the
screen.

For
drinking
water
assessments
involving
surface
water,
EPA
will
replace
its
current
field
pond
scenario
used
in
screening
assessments
with
an
"
index"
reservoir
based
on
an
actual
drinking
water
reservoir
(
Section
III.
A.
1).
To
more
realistically
reflect
watershed­
scale
use,
the
model
will
also
be
adjusted
for
the
percentage
of
the
watershed
feeding
the
reservoir
that
is
actually
in
agricultural
production
(
Section
III.
A.
2).
In
the
longer
term,
EPA
will
move
to
a
watershed­
scale
model
which
more
accurately
captures
basin­
area
processes
and
would
be
more
appropriate
for
drinking
water
assessments
(
Section
III.
A.
3).

One
challenge
facing
the
Agency
is
gathering
enough
reliable
monitoring
data
to
evaluate
model
estimates.
EPA
will
continue
to
seek
existing
and
new
monitoring
data
to
use
in
evaluating
and
strengthening
its
models.
Results
of
a
preliminary
evaluation
of
PRZM/
EXAMS
model
estimates
against
monitoring
data
presented
to
the
FIFRA
Scientific
Advisory
Panel
(
SAP)
in
May
Proposed
Methods
for
Determining
Watershed­
derived
Percent
Crop
Areas
and
Considerations
6
for
Applying
Crop
Area
Adjustments
to
Surface
Water
Screening
Models;
electronic
copy
available
from
the
EPA
home
page
under
the
Office
of
Pesticide
Programs
(
OPP)
at
http://
www.
epa.
gov/
pesticides/
SAP/
1999/
may/
pca_
sap.
pdf
.

14
1999
indicate
that
the
surface­
water
screening
models
may
not
be
consistent
in
overestimating
6
pesticide
concentrations.
Comparisons
made
with
limited
monitoring
data
from
the
Midwestern
U.
S.
that
represent
atrazine,
simazine,
and
metolachlor
concentrations
in
drinking
water
and
on
chlorpyrifos
and
simazine
concentrations
in
rivers
and
streams
in
the
San
Joaquin
River
Valley
of
California
showed
that
the
highest
values
observed
in
monitoring
fell
below
the
crop­
area
adjusted
estimated
"
high­
end"
(
i.
e.,
1­
in­
10
year)
peak
(
single
day)
concentrations.
However,
timeweighted
annual
average
concentrations
of
atrazine
in
surface
water
sources
of
drinking
water
in
one
monitoring
study
exceeded
the
average
annual
model
estimates,
adjusted
for
the
cropped
area,
in
at
least
one
year
for
five
of
37
surface
water
sources.
The
modeled
peak
concentration
of
diazinon
was
equivalent
(
plus
or
minus
a
factor
of
3)
to
peak
monitoring
results
in
the
San
Joaquin
Valley,
and
only
an
order
of
magnitude
greater
than
the
lowest
monitoring
concentrations
reported.
A
lack
of
extensive
monitoring
on
diazinon
in
other
use
areas
makes
it
difficult
to
determine
whether
the
San
Joaquin
Valley
data
represent
high­
end
or
typical
concentrations.
A
preliminary
survey
of
eleven
recent
drinking
water
exposure
assessments
found
that
six
screening
model
estimates
resulted
in
predictions
of
pesticide
concentrations
of
more
than
an
order
of
magnitude
greater
than
available
monitoring
data
while
five
assessments
resulted
in
model
predictions
that
were
similar
to
available
monitoring
data.
These
monitoring
data
represented
either
actual
drinking
water
sources
or
water
bodies
that
could
be
used
as
drinking
water
sources.
The
Agency
is
concerned
that
the
differences
between
model
estimates
and
monitoring
data
are
uneven
and
not
readily
predictable.

The
Agency
plans
to
assess
the
capability
of
PRZM
and
EXAMS
to
provide
high­
end
exposure
estimates
through
(
1)
a
sensitivity
analysis
of
the
models
to
determine
what
input
factors
most
influence
the
model
results
and
(
2)
a
more
thorough
comparison
of
modeling
and
monitoring
data
to
identify
specific
chemical,
site,
or
use
characteristics
that
could
lead
to
inconsistencies
in
the
modeling
results.
Results
of
the
evaluation
could
determine
whether,
for
certain
pesticides
or
uses,
revisions
to
the
models
are
needed
or
whether
another
form
of
screening
is
necessary.

Reliable
and
representative
data
on
measured
pesticide
residues
in
drinking
water
are
a
valuable
assessment
tool
when
available.
However,
because
pesticide
concentrations
vary
greatly
in
location
(
some
drinking
water
sources
are
more
vulnerable
than
others)
and
time
(
both
seasonally
and
year­
to­
year),
most
existing
monitoring
data
provide
little
more
than
a
piece
of
a
complex
puzzle.
OPP
will
continue
to
evaluate
ways
to
better
use
existing
monitoring
data
and
seek
options
for
obtaining
additional
monitoring
data
for
pesticides
that
will
allow
for
improved
assessments
of
pesticide
concentrations
in
drinking
water.

A.
Refinement
of
Screening
Models
for
Use
in
Estimating
Pesticide
Concentrations
in
Drinking
Water
OPP
plans
to
continue
using
mathematical
screening
models
as
a
part
of
its
tiered
In
public
comments
at
the
July
1998
SAP,
one
person
expressed
concern
that
Shipman
City
7
Lake
was
impacted
by
a
point
source
(
a
pesticide
loading
facility),
making
it
unsuitable
for
use
as
an
index
drinking
water
reservoir.
In
follow­
up
investigations,
OPP
determined
that
the
loading
facility
was
shut
down
and
had
not
been
in
operation
during
the
period
that
monitoring
was
conducted.
Despite
that
potential
concern,
the
SAP
concluded
that
OPP
could
continue
to
develop
the
index
drinking
water
reservoir
using
Shipman
City
Lake.

15
approach
to
assessing
the
potential
exposure
to
pesticides
in
drinking
water
in
order
to
effectively
focus
resources
on
the
potential
problem
chemicals.
Further,
modeling
and
other
forms
of
extrapolation
of
data
are
the
only
assessment
tools
currently
available
to
estimate
potential
concentrations
of
new
pesticides.
By
October
31,
1999,
EPA
will
make
the
following
modifications
to
its
approach
in
order
to
provide
a
more
effective
screen
that
identifies
those
pesticides
for
which
a
potential
risk
may
exist.

1.
The
Use
of
An
Index
Drinking
Water
Reservoir
in
Surface
Water
Modeling
Scenarios
In
July,
1998,
OPP
presented
to
the
FIFRA
SAP
a
proposed
"
index"
reservoir
scenario
to
replace
the
"
field
pond"
scenario
currently
used
in
its
screening
models
(
GENEEC
and
PRZM/
EXAMS)
to
estimate
pesticide
concentrations
in
drinking
water
derived
from
surface
water.
OPP
initially
proposed
to
replace
the
existing
farm
pond
scenario
in
drinking
water
screening
assessments
with
Shipman
City
(
IL)
Lake
because
it
was
representative
of
a
number
of
reservoirs
in
the
central
Midwest
that
are
known
to
be
vulnerable
to
pesticide
contamination.
These
reservoirs
tend
to
be
small
and
shallow
with
small
watersheds,
and
frequently
have
Safe
Drinking
Water
Act
compliance
problems
with
atrazine,
a
herbicide
widely
used
on
corn
grown
in
these
watersheds.
Shipman
City
Lake
is
13
acres
(
5.3
ha)
in
area,
averages
9
feet
(
2.7
m)
in
depth,
and
has
a
watershed
area
of
427
acres
(
178
ha)
and
a
normal
capacity
of
144,320
m
.
3
In
July
1998,
the
FIFRA
SAP
called
the
approach
to
selecting
index
drinking
water
reservoirs
reasonable,
but
also
recommended
additional
scientific
review
and
refinements
.
OPP
7
compiled
and
screened
a
list
of
82
candidate
reservoirs
of
varying
sizes
on
the
basis
of
the
percentage
of
the
watershed
that
is
cropped
(
in
this
case,
in
corn),
the
ratio
of
drainage
area
to
normal
reservoir
capacity,
and
the
availability
of
monitoring
data
on
corn
herbicides.
The
initial
list
was
trimmed
to
four
reservoirs
and
the
monitoring
data
and
physical
characteristics
of
these
reservoirs
were
compared.
After
this
evaluation,
the
Agency
determined
that
the
Shipman
City
Lake
was
still
appropriate
for
use
as
an
index
drinking
water
reservoir.
The
index
drinking
water
reservoir
characteristics
will
be
incorporated
into
the
PRZM
and
EXAMS
models
and
will
be
implemented
in
conjunction
with
percent
cropped
area
adjustment
(
see
Section
III.
A.
2
for
discussion
of
the
Percent
Cropped
Area
and
for
timing
of
implementation).

Estimates
of
pesticide
concentrations
in
drinking
water
based
on
a
Midwestern
index
drinking
water
reservoir
may
not
be
representative
of
residue
levels
in
drinking
water
sources
in
other
parts
of
the
country.
The
modeling
scenarios
currently
account
for
region­
specific
rainfall,
soil,
and
hydrologic/
runoff
factors.
The
incorporation
of
an
index
drinking
water
reservoir
is
the
latest
step
in
an
interim
process
that
will
eventually
include
basin­
scale
modeling.
The
Agency
In
the
12/
22/
98
draft
science
policy
document,
the
crop
adjustment
factor
was
referred
to
as
the
8
"
crop
area
factor"
or
"
CAF."
OPP
has
changed
this
term
to
"
percent
cropped
area"
or
"
PCA"
to
be
more
in
line
with
the
terminology
used
for
"
percent
crop
treated."

The
major­
use
crops
and
corresponding
percent
crop
area
adjustments
(
based
on
8­
unit
HUC
9
watersheds)
presented
at
the
May
1999
SAP
are:
corn
(
0.46),
soybeans
(
0.41),
wheat
(
0.56),
and
cotton
16
recognizes
the
need
to
develop
scenarios
for
regional
reservoirs
for
advanced
tiers
of
modeling
as
well
as
for
basin­
scale
modeling.
However,
this
step
is
hampered
by
the
lack
of
monitoring
data
outside
of
the
Midwest
that
is
of
sufficient
quality
and
extent
to
develop
scenarios
for
additional
reservoirs.
As
these
data
become
available,
EPA
will
develop
and
utilize
regional
reservoir
scenarios
in
addition
to
the
current
index
drinking
water
reservoir
scenario.

2.
Accounting
for
the
Percentage
of
Area
Cropped
in
the
Index
Drinking
Water
Reservoir
Models
OPP
has
developed
the
necessary
data
bases
and
Geographical
Information
System
(
GIS)
tools
to
enable
it
to
consider
the
percentage
of
the
area
around
the
index
drinking
water
reservoir
that
is
cropped
(
i.
e.,
the
"
Percent
Cropped
Area"
or
"
PCA"
)
and,
thus,
potentially
treated
with
a
8
pesticide
when
it
uses
its
model
to
predict
pesticide
levels
in
a
drinking
water
reservoir.
OPP
presented
its
plan
for
implementing
the
percent
cropped
area
(
PCA)
as
a
refinement
to
the
FQPA
drinking
water
assessment
process
to
the
FIFRA
Scientific
Advisory
Panel
(
SAP)
in
May
1999.
The
SAP
agreed
with
the
concept
of
the
PCA
as
an
"
appropriate
and
reasonable"
adjustment
for
major­
use
crops
while
still
providing
a
protective
(
i.
e.,
"
highly
vulnerable")
assessment.
It
observed
that
the
PCA
"
provides
a
technically
defensible
approach
to
reduce
estimates
of
acute
and
chronic
pesticide
exposures
to
levels
similar
to
those
found
in
monitoring
data."
However,
the
SAP
also
identified
several
limitations
to
the
approach,
which
have
been
outlined
in
Section
II.
E.
2.
[
SAP
Report
No.
99­
03C,
May
27,
1999;
available
via
the
public
docket].

Using
1992
Agricultural
Census
data,
OPP
ranked
counties
by
PCA
(
since
the
data
are
reported
on
a
county
basis).
For
each
crop,
OPP
used
GIS
tools
to
select
the
small
watershed
(
the
8­
digit
Hydrologic
Unit
Code
was
the
basis
for
evaluation)
which
has
the
highest
PCA.
PCAs
were
derived
on
a
watershed
basis
in
response
to
recommendations
from
the
December
1997
SAP.
The
May
1999
SAP
expressed
concern
that
the
Agency
would
be
unable
to
validate
PCAs
for
minor
crops
and
recommended
that
EPA
consider
requesting
low­
cost,
targeted
monitoring
data
to
evaluate
pesticide
contamination
from
use
on
minor
crops
[
SAP
Report
No.
99­
03C,
May
27,
1999].
The
May
1999
SAP
also
recommended
that,
for
multiple
crop
use,
the
Agency
could
derive
PCAs
based
on
the
maximum
combined
acreage
of
crops
in
a
watershed.
If
pesticide
application
rate
and
timing
vary
from
crop
to
crop,
an
aggregate
pesticide
concentration
estimate
could
be
made
by
separately
simulating
each
crop
in
the
watershed
and
then
summing
the
individual
model
estimates.
EPA
plans
to
incorporate
the
SAP
recommendations
when
it
implements
the
PCA.

OPP
will
implement
the
"
index"
reservoir
and
percent
crop
area
factors
for
the
major­
use
crops
presented
to
the
May
1999
SAP
in
its
Tier
2
(
PRZM/
EXAMS)
surface
water
screening
9
(
0.20).

17
models
by
October
31,
1999.
Once
the
Tier
2
model
is
in
place,
OPP
will
develop
a
Tier
1
index
drinking
water
reservoir
model,
similar
to
that
of
GENEEC.
Based
on
recommendations
from
the
July
1998
SAP,
PCAs
will
not
be
used
with
the
Tier
1
model.
The
method
for
deriving
watershed­
based
percent
cropped
area
(
PCA)
correction
factors
will
be
converted
into
guidance
for
developing
PCAs
for
major
crops
and
cropping
combinations
through
early
2000.

3.
The
Use
of
Watershed­
scale
Models
OPP
completed
and
presented
to
the
FIFRA
SAP
in
July
1998
its
preliminary
evaluation
of
seven
watershed­
scale
surface
water
models.
Further
efforts
to
evaluate
these
models
against
actual
monitoring
data
are
ongoing.
This
model
evaluation
effort
is
expected
to
provide
an
understanding
of
the
relative
accuracy
of
each
of
these
models.
OPP
expects
that
one
or
more
of
these
watershed­
scale
models
will
ultimately
be
used
to
produce
more
refined
estimates
of
pesticide
concentrations
in
drinking
water.
EPA
is
aware
of
the
difficulties
in
developing
and
evaluating
a
watershed­
scale
model
and
is
investing
considerable
effort
in
this
area
in
FY2000.

4.
Ground
Water
Screening
Model
Approach
OPP
will
continue
to
use
SCI­
GROW
as
an
initial
screening
tool
to
determine
the
potential
of
a
pesticide
to
contaminate
ground
water
sources
of
drinking
water
at
concentrations
high
enough
to
indicate
a
potential
for
risk.
On
the
basis
of
recommendations
of
the
FIFRA
SAP
in
December
1997
and
the
experience
of
OPP
in
using
SCI­
GROW
as
an
initial
screen
for
drinking
water
assessments,
OPP
plans
to
systematically
evaluate
SCI­
GROW
against
additional
ground
water
monitoring
data.
Included
in
the
evaluation
will
be
an
assessment
of
the
potential
limitations
in
the
predictive
capacity
of
SCI­
GROW.
For
instance,
do
certain
classes
of
chemicals
or
certain
environmental
fate
parameters
exist
for
which
SCI­
GROW
is
not
well
suited?
Depending
on
the
outcome
of
the
assessment,
some
changes
in
OPP's
approach
to
the
initial
screening
tier
for
ground
water
may
occur.

OPP
also
plans
to
evaluate
models
and
develop
a
procedure
for
a
second
tier
assessment
of
pesticides
in
ground
water.
The
Agency
has
evaluated
approximately
fifty
candidate
models
and
has
selected
six
models
for
detailed
evaluation.
OPP
plans
to
use
data
from
existing
prospective
ground
water
monitoring
studies
to
evaluate
the
ability
of
the
models
to
predict
pesticide
concentrations
in
ground
water.
To
date,
OPP
has
completed
a
preliminary
evaluation
with
one
data
set.
A
similar
evaluation
with
data
sets
from
at
least
two
other
pesticides
representing
other
crops,
pesticide
groups,
use
patterns
and
areas
of
the
country
is
pending.
As
these
evaluations
are
completed,
the
Agency
intends
to
solicit
external
peer
review
and
comment.

B.
Use
of
Monitoring
Data
in
Estimating
Pesticide
Concentrations
in
Drinking
Water
The
Agency
believes
its
risk
assessments
would
be
strengthened
by
additional
monitoring
data
and
is
working
on
a
number
of
levels
to
fill
in
the
gaps
in
monitoring
data
and
acquire
more
18
high
quality
data
on
pesticide
concentrations
in
drinking
water
sources.
At
pesticide­
specific
levels,
the
Agency
is
requesting
registrant­
sponsored
monitoring
and
runoff
studies
when
screening
models
indicate
a
potential
for
concern.
For
multiple
pesticides
on
the
regional
and
national
scales,
the
Agency
is
working
with
the
U.
S.
Geological
Survey
(
USGS)
on
a
pilot
reservoir
monitoring
study
that
will
partially
address
missing
data
on
pesticide
concentrations
in
drinking
water
reservoirs.
EPA
is
also
exploring
design
considerations
for
a
national
survey
of
pesticides
in
drinking
water
with
various
government
agencies
and
industry
groups.

Currently,
standardized
guidance
on
assessing
water
monitoring
data
does
not
exist;
the
criteria
for
such
assessments
will
depend
on
whether
the
data
will
be
used
for
model
validation
or
to
make
decisions
on
an
individual
pesticide.
OPP
does
include
valid
monitoring
data
in
its
risk
assessments.
OPP
does
not
always
have
information
to
determine
whether
the
available
monitoring
data
are
representative
of
particularly
vulnerable
drinking
water
sources.
The
factors
the
Agency
takes
into
account
in
evaluating
the
usefulness
of
the
monitoring
data
in
a
risk
assessment
include
distribution
across
the
cropped
region
and
pesticide
use
area,
design
and
purpose
of
the
study,
vulnerability
of
the
sites,
representativeness
of
actual
drinking
water
sources,
monitoring
of
both
source
and
treated
water,
sampling
frequency
sufficient
to
capture
occurrence
over
time,
analytical
detection
levels
adequate
to
support
aggregate
analysis,
and
inclusion
of
important
metabolites
and
degradates.
A
complete
characterization
of
watershed,
cropping
patterns,
pesticide
application,
water
treatment,
and
water
quality
assist
the
Agency
in
interpreting
monitoring
results.

Some
issues
the
Agency
is
attempting
to
address,
based
on
its
experience
in
evaluating
existing
monitoring
data
sets,
include:

°
Reliance
on
limited
monitoring
data
(
e.
g.,
data
that
do
not
necessarily
cover
the
range
of
use
areas
or
span
a
sufficient
time
to
capture
seasonal
and
multi­
year
variations
in
pesticide
concentrations)
may
lead
to
a
decision
that
a
pesticide
does
not
pose
a
risk
via
the
drinking
water
route
under
certain
conditions
when
in
fact
it
does
under
other
existing
conditions.
Existing
monitoring
data
may
suggest
that,
on
a
national
basis,
the
pesticide
in
question
is
not
occurring
in
drinking
water
at
a
frequency
of
concern.
However,
in
certain
vulnerable
areas,
the
pesticide
may
be
found
in
concentrations
high
enough
to
be
of
toxicological
concern.

°
A
monitoring
data
set
may
include
non­
detects,
particularly
in
a
national
monitoring
program.
Non­
detects
may
result
when
the
pesticide
occurs
in
concentrations
that
are
below
the
limit
of
detection
for
the
analytical
method
or
when
the
pesticide
is
not
present
at
all
in
the
water
sample.
The
absence
of
the
pesticide
in
water
may
indicate
that
the
pesticide
is
not
likely
to
occur
in
drinking
water
sources;
it
may
also
result
when
samples
are
taken
in
areas
where
the
pesticide
is
not
used
or
during
times
of
the
year
when
the
pesticide
is
not
used.
Information
needed
to
evaluate
the
significance
of
non­
detects
is
rarely
included
in
the
data
set.

°
The
frequency
of
sample
collection
in
monitoring
studies
is
rarely
adequate
to
capture
peak
pesticide
concentrations
or
to
estimate
a
reasonable
maximum
exposure.
19
°
Concentrations
of
pesticide
transformation
products
which
are
also
of
toxicological
concern
are
rarely
included
in
monitoring
studies.

°
Monitoring
data
based
on
untreated
water
samples
do
not
account
for
removal
or
dilution
of
pesticides
or,
in
some
cases,
the
formation
of
more
toxic
compounds,
that
may
occur
in
water
treatment.
However,
because
of
the
variability
in
treatment
processes
(
which
may
include
no
treatment
in
the
case
of
private
wells),
data
gathered
from
treated
samples
may
not
be
representative
of
minimal,
typical
or
high­
end
treatment
processes
(
see
Section
III.
D).

C
OPP
intends
to
pursue
data
on
distributions
of
pesticide
concentrations
in
drinking
water
for
use
in
aggregate
and
cumulative
exposure
and
risk
assessments
from
statisticallydesigned
surveys
that
reflect:
pesticide
usage
on
a
compound­
specific
basis,
size
of
community
water
systems,
water
source,
treatment
(
where
warranted),
system
vulnerability
to
pesticide
contamination,
and
temporal
variability.

Developing
criteria
for
evaluating
monitoring
data
will
not
only
aid
in
the
evaluation
of
current
data,
but
will
help
guide
the
design
of
future
monitoring
studies.
As
noted
earlier,
standardized
guidance
on
assessing
water
monitoring
data
does
not
exist
and
will
vary
depending
on
how
the
data
will
be
used.
The
Agency
will
be
looking
to
various
sources
for
guidance
in
assessing
the
usability
of
monitoring
data
in
pesticide
exposure
assessments.

GIS
tools,
coupled
with
more
detailed
site,
meteorological,
and
use
characterizations,
will
also
assist
in
characterizing
and
evaluating
new
and
existing
monitoring
data.
It
will
help
the
Agency
assess
potentially
exposed
populations
and
identify
gaps
in
existing
data
in
order
to
better
target
additional
monitoring.
The
Agency
continues
to
seek
and
develop
such
tools
to
improve
its
assessment
of
pesticide
exposure
from
drinking
water
sources.
At
the
same
time,
the
Agency
believes
that
more
monitoring
data,
and
more
ancillary
information
(
weather,
site
and
usage
characteristics),
will
be
needed
to
take
full
advantage
of
the
GIS
capabilities
at
hand.

C.
Drinking
Water
Vulnerability
Assessments
The
Agency
will
continue
to
seek
and
evaluate
tools
that
would
aid
in
assessing
the
vulnerability
of
water
resources.
Such
tools
would
be
useful
not
only
in
identifying
areas
of
potential
concern,
but
also
in
evaluating
monitoring
data
and
modeling
estimates
and
in
developing
site
selection
criteria
for
surveys.
Whether
the
drinking
water
assessment
is
conducted
on
a
regional
or
national
scale,
the
ultimate
goal
of
such
an
assessment
is
to
identify
where
the
risk
occurs
since
drinking
water
exposure
is
localized
in
nature.
Vulnerability
can
be
defined
as
the
tendency
or
likelihood
for
contaminants
to
reach
a
surface­
or
ground­
water
system
after
introduction
at
some
location
within
the
watershed
(
National
Research
Council,
1993).
Vulnerability
depends
on
a
combination
of
factors
relating
to
pesticide
usage,
site/
environmental
factors,
crop
and
pest
management,
and
weather
patterns.
As
noted
in
the
National
Research
Council
report,
defining
what
constitutes
"
most
vulnerable"
is
a
challenge.
Conceptual
models
of
vulnerability
exist,
but
differ
in
what
is
considered
vulnerable
and
what
factors
are
included.
20
Some
factors
affecting
vulnerability
of
drinking
water
sources
may
differ
for
surface
water
and
ground
water.
While
a
portion
of
the
vulnerability
assessment
would
be
attributable
to
intrinsic
site
factors
that
can
be
mapped,
other
portions,
such
as
weather
patterns
and
management
practices,
are
more
dynamic
and
would
require
a
different
approach.

D.
Accounting
for
Drinking
Water
Treatment
Effects
The
ultimate
goal
of
the
Agency
is
to
conduct
risk
assessments
for
drinking
water
based
on
exposure
to
the
consumer
at
the
tap.
However,
differences
of
opinion
exist
as
to
how
the
Agency
should
address
drinking
water
treatment
effects
in
its
drinking
water
assessment.
The
type
and
degree
of
drinking
water
treatment
varies
among
community
water
systems.
A
significant
portion
of
the
nation's
population
consumes
drinking
water
that
is
not
treated
with
technology
that
is
likely
to
reduce
pesticide
concentrations
(
i.
e.,
private
drinking
water
sources,
community
drinking
water
from
ground
water,
or
small
surface
water
community
systems).
Private
wells
are
likely
to
receive
little
or
no
treatment.
The
smaller
community
water
systems
will,
at
a
minimum,
add
a
disinfectant
and
may
adjust
the
pH
of
the
water,
which
may
affect
some
classes
of
pesticides
but
not
others.
In
some
cases,
the
disinfectant
treatment
may
result
in
transformation
to
a
toxic
degradate.
Representatives
of
community
water
system
operators
have
urged
the
Agency
to
focus
its
decision­
making
criteria
on
drinking
water
at
the
intake
(
source
water)
and
consider
the
impact
of
pesticides
on
water
entering
drinking
water
supplies
and
private
wells.

The
Agency
is
in
the
process
of
gathering
information
on
the
extent
of
drinking
water
treatments
in
use
and
the
effectiveness
of
these
treatments
on
reducing
the
level
of
pesticides
in
water.
Consideration
of
water
treatment
effects
on
pesticides
in
water
must
take
into
account
not
only
effectiveness
in
pesticide
removal,
but
also
the
secondary
formation
of
any
transformation
products
of
toxicological
concern
as
a
result
of
the
treatment
process.
The
area
and
population
served
by
the
particular
treatments,
as
well
as
temporal
variations
in
drinking
water
treatment
effectiveness,
must
also
be
considered.
By
the
end
of
1999,
the
Agency
plans
to
issue
a
paper
addressing
how
it
will
incorporate
drinking
water
treatment
effects
in
its
drinking
water
exposure
assessment.
The
public
will
be
invited
to
comment
on
this
science
policy
document.

E.
Using
Model
Estimates
and
Monitoring
Data
in
Quantitative
Assessment
of
Uses
of
Concern
for
Drinking
Water
In
the
Agency's
experience,
many
pesticides
pass
the
PRZM/
EXAMS
screen
and
no
additional
assessment
is
needed.
In
those
instances
where
the
model
estimates
suggest
a
potential
for
concern
(
i.
e.,
the
estimated
pesticide
concentration
in
water
exceeds
the
drinking
water
level
of
comparison),
additional
steps
taken
by
the
Agency
are
determined
on
a
case­
by­
case
basis,
depending
on
how
much
monitoring
data
are
available
and
the
extent
of
available
information
on
use
and
management
practices,
which
are
pesticide­
specific.
These
additional
steps
focus
on
gathering
more
information
to
reduce
the
uncertainty
in
the
drinking
water
estimates
or
requesting
additional
monitoring
data
that
can
be
related
to
drinking
water
sources.
EPA
plans
to
issue
a
paper,
"
Quantitative
Assessment
of
Uses
of
Concern
for
Drinking
Water,"
which
will
propose
using
available
data
and
models
to
develop
quantitative
estimates
of
pesticide
concentrations
in
21
drinking
water
for
pesticides
which
EPA
is
particularly
concerned
and
to
estimate
the
potential
size
of
the
population
exposed
to
these
levels.
This
paper,
which
is
expected
to
be
released
for
public
comment
in
the
Spring
of
2000,
will
describe
how
the
Agency
proposes
to
use
these
estimates
in
certain
cases
in
quantitative
aggregate
human
health
risk
assessments.
22
IV.
BIBLIOGRAPHY
ILSI.
1998.
Assessment
of
Methods
to
Estimate
Pesticide
Concentrations
in
Drinking
Water
Sources.
Report
of
a
workgroup
on
Aggregate
Exposure
Assessment,
International
Life
Science
Institute,
Washington,
DC.

National
Research
Council.
1993.
Ground
water
vulnerability
assessment
:
predicting
relative
contamination
potential
under
uncertainty.
Committee
on
Techniques
for
Assessing
Ground
Water
Vulnerability,
Water
Science
and
Technology
Board,
Commission
on
Geosciences,
Environment,
and
Resources.

US
EPA.
1990.
National
Pesticide
Survey.
Summary
of
Results
of
WPA's
National
Summary
of
Pesticides
in
Drinking
Water
Wells.
Office
of
Water
and
Office
of
Pesticides
and
Toxic
Substances,
US
EPA,
Washington,
D.
C.

US
EPA
OPP.
1992.
Pesticides
in
Ground
Water
Database
 
A
Compilation
of
Monitoring
Studies:
1971­
1991
National
Summary.
EPA
734­
12­
92­
001.
US
EPA
Office
of
Pesticide
Programs,
Washington,
DC.

US.
EPA
FIFRA
Scientific
Advisory
Panel.
1997.
FIFRA
SAP
Report
on
A
Set
of
Scientific
Issues
Being
Considered
by
the
Agency
in
Connection
with
Estimating
Drinking
Water
Exposure
as
a
Component
of
the
Dietary
Risk
Assessment.
Electronic
copy
available
from
the
EPA
OPP
home
page
at
http://
www.
epa.
gov/
pesticides/
SAP/
1997/
december/
finaldec.
pdf
.

U.
S.
EPA
FIFRA
Scientific
Advisory
Panel.
1998.
FIFRA
SAP
Report
on
Proposed
Methods
for
Basin­
Scale
Estimation
of
Pesticide
Concentrations
in
Flowing
Water
and
Reservoirs
for
Tolerance
Reassessment.
Electronic
copy
available
from
the
EPA
OPP
home
page
at
http://
www.
epa.
gov/
pesticides/
SAP/
1998/
index.
htm#
october
.

U.
S.
EPA
Office
of
Pesticide
Programs.
1997.
Pesticide
Regulation
(
PR)
Notice
97­
1:
Agency
Actions
Under
the
Requirements
of
the
Food
Quality
Protection
Act.
Electronic
copy
available
from
the
EPA
OPP
home
page
at
http://
www.
epa.
gov/
opppmsd1/
PR_
Notices/
.

U.
S.
EPA
Office
of
Pesticide
Programs.
1999.
Proposed
Methods
for
Determining
Watershedderived
Percent
Crop
Areas
and
Considerations
for
Applying
Crop
Area
Adjustments
to
Surface
Water
Screening
Models.
Electronic
copy
available
from
the
EPA
home
page
under
the
Office
of
Pesticide
Programs
(
OPP)
at
http://
www.
epa.
gov/
pesticides/
SAP/
1999/
may/
pca_
sap.
pdf
.

US
EPA
Office
of
Water.
1998.
U.
S.
EPA
Office
of
Water
Storage
and
Retrieval
Database
(
STORET).
<
http://
www.
epa.
gov/
owowwtr1/
STORET/
overview.
html>

USGS.
1998.
The
National
Water
Quality
Assessment
(
NAWQA)
Program,
National
Stream
Quality
Accounting
Network
(
NASQAN),
and
Toxic
Substances
Hydrology
Program
can
be
reached
through
the
U.
S.
Geological
Survey's
main
web
page
at
<
http://
usgs.
gov/>.
23
THE
AGENCY'S
RESPONSES
TO
PUBLIC
COMMENTS
ON
THE
DRAFT
FQPA
SCIENCE
POLICY
DOCUMENT:

"
Estimating
the
Drinking
Water
Component
of
a
Dietary
Exposure
Assessment"
(
Announced
January
4,
1999;
64
FR
162;
FRL­
6054­
8.)
(
October
19,
1999)

The
Agency
reviewed
all
comments
pertaining
to
this
draft
science
policy
document
that
were
submitted
specifically
under
this
docket
(
OPP­
00577)
or
in
relation
to
the
Tolerance
Reassessment
Advisory
Committee
(
TRAC).
At
the
end
of
the
document
is
a
listing
of
the
names
and
affiliations
of
the
individuals
submitting
comments.
In
revising
the
document,
the
Agency
extensively
reviewed
and
considered
all
comments.
The
comments
addressed
a
broad
range
of
issues
and,
in
many
instances,
provided
no
general
consensus.
These
differences
in
opinion
highlight
the
difficulties
the
Agency
faces
in
improving
its
existing
science­
based
policy
for
estimating
pesticide
concentrations
in
drinking
water.
In
addition,
the
Agency
has
incorporated
comments
from
a
May
1999
FIFRA
Scientific
Advisory
Panel
that
evaluated
the
proposed
approach
for
incorporating
a
crop
area
adjustment
factor
along
with
a
drinking
water
reservoir
scenario
in
its
surface
water
screening
models.
The
Agency
grouped
all
comments
according
to
the
nature
of
the
comment
and
the
issue
or
section
of
the
document
which
they
addressed.
For
the
substantive
comments
that
follow,
contrasting
opinions
are
presented,
along
with
EPA's
response.
The
numbers
used
in
the
summary
below
correspond
to
specific
commentors
(
listed
at
the
end
of
this
document).

A.
Drinking
Water
As
a
Source
of
Dietary
Exposure
in
Aggregate
Exposure
Assessments
Differences
of
opinion
exist
as
to
whether
drinking
water
should
even
be
considered
in
aggregate
exposure
assessments.
One
commentor
[
3]
argued
that
drinking
water
should
be
considered
as
"
other"
nonoccupational
exposure
rather
than
as
a
"
dietary"
source
of
exposure
and
thus
should
only
be
included
in
aggregate
exposure
calculations
when
"
reliable
information"
exists.
Another
commentor
[
1]
agreed
with
the
Agency
position
that
drinking
water
is
a
dietary
exposure
source.

Agency
Response:
Ample
data
show
that
many
pesticides
can
move
from
the
site
of
application,
by
leaching,
spray
drift,
and
runoff,
into
surface
water
bodies
or
ground
water
that
are
used
as
sources
of
drinking
water.
In
the
case
of
some
pesticides
and
some
locations,
a
high
potential
exists
for
pesticide
contamination
of
drinking
water
at
levels
of
significance
for
public
health.
Therefore,
OPP
believes
it
is
prudent
to
consider
routinely
the
contribution
to
overall
risk
made
by
ingestion
of
drinking
water.
As
for
the
issue
of
whether
OPP
should
consider
drinking
water
exposure
as
a
dietary
source
of
exposure,
water
is
an
important
component
of
the
human
diet
and
as
such
should
be
considered
for
purposes
of
FQPA
assessments
as
a
dietary
source
of
exposure.

B.
Screening
Approach
For
Drinking
Water
Exposure
Assessments
and
the
Conservative
Nature
of
the
Model
Estimates
"
High­
end"
is
defined
in
"
Guidelines
for
Exposure
Assessment,"
FR
Vol.
57,
No.
104,
May
29,
10
1992,
and
"
Guidance
on
Risk
Characterization
for
Risk
Managers
and
Risk
Assessors."
F.
Henry
Habicht
II,
Deputy
Administrator,
U.
S.
EPA,
to
Assistant
and
Regional
Administrators,
U.
S.
EPA,
Feb.
26,
1992.

24
1.
The
Role
of
Screening
Models
Philosophical
differences
were
evident
in
the
comments
on
the
role
and
nature
of
screening
models
in
drinking
water
exposure
assessments.
Some
commentors
[
3,
7]
felt
that
advanced
model
tiers
should
shift
to
more
typical
use
scenarios
that
provide
central
tendency
estimates
(
i.
e.,
mean,
mode
or
median
values
across
all
use
sites,
not
just
high­
end
exposure
sites).
However,
other
commentors
[
1,
6]
felt
that
the
Agency
should
focus
on
developing
models
that
provide
a
"
credible
worst
case
scenario."
One
commentor
[
1]
observed
that
the
worst­
case
scenario
is
necessary
because
of
uncertainties
in
pesticide
runoff
and
leaching,
and
the
absence
of
state
or
local
management
programs
for
mitigation.
In
noting
that
more
realistic
estimates
do
not
necessarily
equate
to
central
tendency
estimates,
the
commentor
recommended
that
decisions
based
on
screening
models
err
toward
obtaining
more
and
better
quality
data
on
which
to
base
a
decision,
and
that
models
result
in
"
realistic
upper
bound
estimates"
for
protective
decisionmaking
This
commentor
felt
that
conservatism
(
i.
e.,
estimates
of
high­
end
exposure)
in
modeling
is
needed
because
it
is
unlikely
that
adequate
monitoring
data
will
be
available
to
support
registration
or
reregistration
decisions
in
the
near
term.

Agency
Response:
The
Agency
uses
a
tiered
approach
to
conducting
drinking
water
assessments
because
this
is
a
cost­
effective
way
to
factor
drinking
water
exposure
into
risk
assessments
under
FQPA.
The
initial
screening
tiers
of
the
drinking
water
assessment
process
are
purposefully
designed
to
provide
OPP
with
a
high
degree
of
confidence
that
pesticides
which
are
cleared
will
not
in
fact
pose
a
drinking
water
concern.
However,
an
initial
screen
that
is
too
"
fine"
results
in
wasted
Agency
review
resources.
Accordingly,
OPP's
goal
is
to
minimize
the
potential
for
"
passing"
a
pesticide
which
may
pose
a
concern
while
not
being
overly
cautious
such
that
those
pesticides
which
will
truly
pose
no
concern
fail
the
screen.
In
order
to
achieve
this
goal,
OPP
considers
pesticide
concentrations
from
a
high­
end
exposure
scenario
rather
than
a
centraltendency
scenario
which,
by
nature,
approximates
a
middle
level
of
exposure.
As
defined
in
the
draft
science
policy
document,
"
high­
end"
refers
to
a
combination
of
events
and
conditions
such
that,
taken
together,
produces
conceivable
risk
greater
than
that
estimated
to
be
experienced
by
90
percent
of
the
population,
but
less
than
the
maximally
exposed
risk
.
The
Agency's
approach
10
to
refining
estimates
of
pesticide
concentrations
in
drinking
water
in
higher
tiers
is
to
reduce
the
uncertainties
in
the
estimate,
which
often
requires
obtaining
more
and
better
quality
data
with
which
to
make
a
decision.

2.
The
Conservatism
of
the
Screening
Models
Commentors
disagreed
on
the
conservative
nature
of
surface
water
screening
models.
Some
[
3,
7]
felt
that
the
existing
screening
models
generally
overestimated
pesticide
concentrations,
sometimes
by
several
orders
of
magnitude.
They
recommended
that
the
models
be
calibrated
with
"
representative
monitoring
data"
to
ensure
that
the
degree
of
conservatism
is
"
understood,
recognized,
and
reasonable."
However,
other
commentors
[
1,
6]
questioned
the
Proposed
Methods
for
Determining
Watershed­
derived
Percent
Crop
Areas
and
Considerations
11
for
Applying
Crop
Area
Adjustments
to
Surface
Water
Screening
Models;
electronic
copy
available
from
the
EPA
home
page
under
the
Office
of
Pesticide
Programs
(
OPP)
at
http://
www.
epa.
gov/
pesticides/
SAP/
1999/
may/
pca_
sap.
pdf
.

Monitoring
data
for
the
midwestern
U.
S.
included
the
37
drinking
water
sources
for
which
12
untreated
water
was
sampled
in
the
Acetochlor
Registration
Partnership
(
ARP)
(
see
the
EPA
OPP
home
page
at
http://
www.
epa.
gov/
oppefed1/
aceto/
index.
htm
for
more
information)
and
76
midwestern
reservoirs
sampled
by
USGS
(
Scribner
et
al.
1996.
Concentrations
of
selected
herbicides,
herbicide
metabolites,
and
nutrients
in
outflow
from
selected
midwestern
reservoirs,
April
1992
through
September
1993.
U.
S.
Geol.
Surv.
Open­
File
Report
96­
393.
Prepared
as
part
of
the
Toxic
Substances
Hydrology
Program.
Laurence,
KS.).
Monitoring
data
for
the
San
Joaquin
valley
came
from
USGS's
NAWQA
program,
from
14
sites
sampled
in
1993
and
1994.

25
assumed
conservatism
of
the
existing
models
without
more
data
to
support
the
assessment.
One
commentor
[
1]
noted
that
pesticide
detections
at
some
water
treatment
plants,
resulting
in
exceedances
of
MCLs
and
installation
of
additional
treatment
measures,
suggest
otherwise.

Agency
Response:
As
is
explained
earlier
is
the
response
to
Question
B1,
OPP's
goal
is
to
minimize
the
potential
for
"
passing"
a
pesticide
which
may
pose
a
concern
while
not
being
overly
cautious
such
that
those
pesticides
which
will
truly
pose
no
concern
fail
the
screen.
In
order
to
achieve
this
goal,
OPP
considers
pesticide
concentrations
from
a
high­
end
exposure
scenario
rather
than
a
central­
tendency
scenario
which,
by
nature,
approximates
a
middle
level
of
exposure.
Although
OPP
believes
that
its
screening
level
approach
achieves
this
goal,
recent
work
to
compare
model
estimates
to
monitoring
data
is
informative
as
to
whether
model
estimates
grossly
overestimate
pesticide
concentrations
in
surface
water.
The
Agency
completed
a
preliminary
evaluation
of
PRZM/
EXAMS
model
estimates
against
monitoring
data
for
the
index
reservoir
/
percent
crop
area
modeling
revisions
presented
to
the
FIFRA
Scientific
Advisory
Panel
(
SAP)
in
May
1999
.
Initial
comparisons
were
made
with
limited
monitoring
information
on
atrazine,
11
simazine,
and
metolachlor
in
the
Midwestern
U.
S.
and
on
chlorpyrifos,
simazine,
and
diazinon
in
the
San
Joaquin
River
Valley
of
California.
For
simazine
and
metolachlor
in
the
midwest,
the
12
highest
values
observed
in
monitoring
fell
below
the
crop­
area
adjusted
estimated
"
high­
end"
(
i.
e.,
1­
in­
10
year)
one­
day
and
average
annual
concentrations.
For
atrazine,
the
crop­
area
adjustment
still
resulted
in
estimated
1­
in­
10­
year
peak
concentrations
greater
than
most
monitoring
peaks.
A
number
of
time­
weighted
annual
average
concentrations
from
the
ARP
monitoring
sources
(
5
of
the
37
reservoirs)
exceeded
the
crop­
adjusted
average
annual
estimates
in
at
least
one
year
of
the
study.
In
the
San
Joaquin
Valley,
the
crop­
adjusted
model
estimates
of
1­
in­
10
year
peak
concentrations
for
chlorpyrifos
and
simazine
were
greater
than
the
highest
observed
monitoring
concentrations
by
a
factor
or
two
or
more.
In
the
case
of
diazinon,
however,
the
estimated
peak
concentrations
were
in
the
same
range
as
monitoring
results,
and
lower
than
four
monitoring
values.
The
confidence
in
the
comparison
between
peak
monitoring
and
modeling
values
is
low
because
the
low
sampling
frequencies
in
the
monitoring
studies
were
not
adequate
to
capture
peak
concentrations.

In
the
May
1999
SAP
presentation,
OPP
presented
a
preliminary
survey
of
eleven
recent
26
drinking
water
exposure
assessments.
In
six
of
those
assessments,
screening
model
estimates
with
PRZM
and
EXAMS,
using
the
farm
pond
scenario
and
assuming
100%
of
the
drainage
area
is
in
crop
and
is
treated
with
the
pesticide
at
the
same
time,
resulted
in
predictions
of
pesticide
concentrations
of
more
than
an
order
of
magnitude
greater
than
available
monitoring
data.
However,
in
five
assessments,
the
model
predictions
were
similar
to
available
monitoring
data
(
assuming
that
the
monitoring
data
captured
the
true
peak
concentration).
The
Agency
has
some
concern
that
the
differences
between
model
estimates
and
monitoring
data
are
uneven
and
not
readily
predictable.

The
SAP
[
SAP
Report
No.
99­
03C,
May
27,
1999]
noted
that
model
inconsistencies
could
come
from
many
sources,
including
inaccurate
process
representation
in
the
model
itself
(
preferential
flow,
ground
water
discharge,
etc.),
quality
of
input
data
(
soil,
climate,
chemical
use,
etc.),
quality
of
monitoring
data
(
sampling
frequency,
duration
of
study,
etc.),
and
differences
in
the
size
of
the
modeled
watershed
versus
the
size
of
the
watersheds
in
which
the
data
was
collected.
They
concurred
with
the
steps
identified
by
the
Agency
to
assess
the
capability
of
the
model
to
provide
high­
end
exposure
estimates,
including
a
sensitivity
analysis
of
the
models
to
determine
what
input
factors
most
influence
the
model
results
and
a
more
thorough
comparison
of
modeling
and
monitoring
data
to
identify
specific
chemical
or
scenario
characteristics
that
could
lead
to
inconsistencies
in
the
modeling
results.
Results
of
the
evaluation
could
determine
whether,
for
certain
pesticides
or
uses,
corrections
to
the
models
are
needed
or
whether
another
form
of
screening
is
necessary.
OPP
is
in
the
process
of
completing
a
sensitivity
analysis
and
these
results
should
be
available
within
the
next
6
months.

3.
Steps
Beyond
the
Screening
Tiers
Two
commentors
[
3,
5]
expressed
concern
that
the
draft
science
policy
document
lacked
a
discussion
of
refinements
beyond
the
initial
screening
tiers
or
a
"
clear
definition
of
procedures
for
making
science­
based
risk
decisions
when
reliable
information
is
not
available
or
is
insufficient
for
purposes
of
refining
the
screening
level
estimates
of
pesticide
exposure
through
drinking
water."

Agency
Response:
The
initial
science
policy
document
spelled
out
the
general
screening
approach
and
identified
areas
where
more
information
or
tools
are
needed.
In
the
Agency's
experience,
many
pesticides
pass
the
tier
2
screen
and
no
additional
assessment
is
needed.
In
those
instances
where
the
model
estimates
suggest
a
potential
for
concern
(
i.
e.,
the
estimated
pesticide
concentration
in
water
exceeds
the
drinking
water
level
of
comparison),
additional
steps
taken
by
the
Agency
are
determined
on
a
case­
by­
case
basis,
depending
on
how
much
monitoring
data
are
available
and
the
extent
of
available
information
on
use
and
management
practices,
which
are
pesticide­
specific.
These
additional
steps
focus
on
gathering
more
information
to
reduce
the
uncertainty
in
the
drinking
water
estimates
or
requesting
additional
monitoring
data
that
can
be
related
to
drinking
water
sources.

EPA
plans
to
issue
a
paper,
"
Quantitative
Assessment
of
Uses
of
Concern
for
Drinking
Water,"
which
will
propose
using
available
data
and
available
models
in
certain
circumstances
to
develop
quantitative
estimates
of
pesticide
concentrations
in
drinking
water
and
to
estimate
the
potential
size
of
the
population
exposed
to
these
levels.
OPP
believes
that
it
needs
to
be
able
to
27
develop
such
estimates
in
those
cases
where
all
available
data
and
information
present
a
compelling
case
that
a
pesticide
has
a
very
high
probability
of
resulting
in
significant
human
health
risks
through
the
drinking
water
pathway.
This
paper,
which
is
expected
to
be
released
for
comment
in
early
2000,
will
describe
how
the
Agency
proposes
to
use
these
estimates
in
certain
cases
in
quantitative
aggregate
human
health
risk
assessments.

4.
The
Suitability
of
Screening
Models
for
Drinking
Water
Exposure
Assessments
Opinions
regarding
the
suitability
of
screening
models
for
drinking
water
assessments
varied.
One
commentor
[
3]
noted
that
screening
models
are
reliable
for
showing
a
lack
of
concern
about
drinking
water
exposure,
but
cannot
show
that
a
certain
level
of
exposure
is
likely.
Another
[
6]
thought
that
the
FQPA
provision
of
"
reasonable
certainty
of
no
harm"
should
put
the
burden
on
the
registrant
to
show
a
reasonable
certainty
of
no
harm
rather
than
on
the
Agency
to
show
a
reasonable
certainty
of
harm.
This
commentor
proposed
that
the
Agency
first
look
at
existing
monitoring
data;
if
any
value
exceeds
the
drinking
water
level
of
comparison
(
DWLOC),
then
EPA
cannot
conclude
a
reasonable
certainty
of
no
harm.
If
no
value
exceeds
the
DWLOC,
then
EPA
should
move
to
screening
models.

Agency
Response:
Under
FIFRA
and
FFDCA,
the
ultimate
burden
of
proving
safety
of
a
given
pesticide
rests
on
the
proponent
of
a
FIFRA
registration
(
or
reregistration)
of
the
pesticide
or
the
proponent
of
establishing
or
maintaining
a
FFDCA
tolerance
or
exemption.
The
draft
science
policy
document
explains
the
approach
that
EPA
will
use
for
factoring
drinking
water
into
tolerance
decisions
under
FQPA.
The
policy
document
is
not
intended
in
any
way
to
address
or
alter
the
burden
of
proof
as
established
by
existing
statutes
and
court
decisions.
As
for
looking
at
monitoring
data
as
the
first
screen
and
then
proceeding
to
models,
the
Agency
relies
on
models
as
the
first
screen
because
monitoring
data
are
not
always
available
and
the
data
that
are
available
require
considerable
scientific
judgment
and
time
to
evaluate
their
suitability.
Issues
related
to
the
evaluation
and
interpretation
of
monitoring
data
are
elaborated
upon
in
a
later
section
of
this
comment/
response
paper.
The
screening
model
approach
allows
for
an
effective
use
of
the
Agency's
limited
resources.

C.
Scale
of
the
Drinking
Water
Assessment
Commentors
[
3,
6,
7]
provided
different
reasons
for
why
a
complete
national
assessment
of
the
drinking
water
contribution
to
aggregate
exposure
from
a
pesticide
was
not
always
necessary.
One
[
3]
stated
that
aggregate
exposure
assessments
should
be
done
on
a
regional,
not
national,
scale.
A
second
[
7]
recommended
that
EPA
begin
on
the
local
scale,
with
high
vulnerability/
high
pesticide
use
areas
representing
the
"
worst­
case."
If
the
local
scenario
indicates
potential
concerns
from
pesticide
exposure
in
drinking
water,
EPA
would
proceed
to
a
regional
scale
incorporating
more
watersheds.
Regional
or
national
scales
would
be
more
appropriate
for
probabilistic
drinking
water
assessments.
A
third
commentor
[
6]
stated
that
if
EPA
anticipates,
or
if
existing
data
show,
some
portion
of
the
population
will
consume
pesticides
in
drinking
water
at
a
level
greater
than
the
DWLOC,
then
the
Agency
cannot
conclude
that
a
reasonable
certainty
of
no
harm
exists
and
a
broader,
national
assessment
will
not
change
that
conclusion.
This
commentor
recommended
that
EPA
focus
its
assessment
not
to
protect
"
most"
people
or
an
Ground
Water
Vulnerability
Assessment
:
Predicting
Relative
Contamination
Potential
under
13
Uncertainty
/
Committee
on
Techniques
for
Assessing
Ground
Water
Vulnerability,
Water
Science
and
Technology
Board,
Commission
on
Geosciences,
Environment,
and
Resources,
National
Research
Council,
1993.

28
"
average"
child,
but
to
protect
those
who
are
most
likely
to
be
exposed
to
the
highest
levels
of
pesticides
in
tap
water.

Agency
Response:
The
Agency's
goal
in
conducting
a
drinking
water
assessment
is
to
identify
where
risk
from
the
drinking
water
pathway
will
occur
and
the
magnitude
of
the
risk
(
including
the
number
of
people
exposed
above
levels
of
concern
from
a
human
health
perspective)
from
this
pathway
of
exposure.
If
the
severity
and
magnitude
of
exposure
is
significant
enough
from
this
pathway
(
that
is,
exposure
is
expected
to
significantly
impact
a
significant
subpopulation),
then
this
exposure
would
be
factored
into
the
aggregate
exposure
assessment
for
tolerance
setting
purposes.
Since
drinking
water
exposure
is
localized
in
nature,
determining
the
severity
and
magnitude
of
risk
from
drinking
water
exposure
requires
evaluation
at
the
watershed
scale
and
then
estimation
of
the
number
of
watersheds
and
people
exposed
at
larger
scales
(
e.
g.,
regional
and
national).
One
approach
is
to
evaluate
the
potential
exposure
at
vulnerable
locations
and
use
the
results
of
this
evaluation
to
extrapolate
to
other
locations.

Vulnerability
can
be
defined
as
the
tendency
or
likelihood
for
contaminants
to
reach
a
surface­
or
ground­
water
system
after
introduction
at
some
location
within
the
watershed
.
13
Vulnerability
depends
on
a
combination
of
factors
relating
to
pesticide
usage,
site/
environmental
factors,
crop
and
pest
management,
and
weather
patterns.
As
noted
in
the
National
Research
Council
report,
defining
what
constitutes
"
most
vulnerable"
is
a
challenge.
Conceptual
models
of
vulnerability
exist,
but
differ
in
what
is
considered
vulnerable
and
what
factors
are
included.
Some
factors
affecting
vulnerability
of
drinking
water
sources
may
differ
for
surface
water
and
ground
water.
While
a
portion
of
the
vulnerability
assessment
would
be
attributable
to
intrinsic
site
factors
that
can
be
mapped,
other
portions,
such
as
weather
patterns
and
management
practices,
are
more
dynamic
and
would
require
a
different
approach.
The
Agency
will
continue
to
seek
and
evaluate
tools
that
would
aid
it
in
assessing
the
vulnerability
of
water
sources.
Such
tools
would
be
useful
not
only
in
identifying
areas
of
potential
concern,
but
also
in
evaluating
monitoring
data
and
modeling
estimates.

D.
Sources
of
Pesticide
Contamination
Not
Considered
One
commentor
[
6]
observed
that
the
assessment
approach
did
not
account
for
additional
exposure
sources
from
non­
agricultural
uses
of
pesticides,
such
as
pesticide
use
in
urban
and
suburban
areas,
which
have
been
found
to
be
significant
sources
of
contamination
in
the
U.
S.
Geological
Survey
(
USGS)
monitoring
studies,
or
exposure
from
mixing/
loading
or
storage
areas.

Agency
Response:
The
Agency
is
aware
that
non­
agricultural
uses
of
pesticides
can
contribute
to
pesticide
concentrations
in
water
but,
for
the
most
part,
lacks
the
tools
to
estimate
the
extent
of
contributions
for
many
non­
agricultural
uses,
especially
home
uses.
To
the
extent
Proposed
Methods
for
Basin­
Scale
Estimation
of
Pesticide
Concentrations
in
Flowing
Water
14
and
Reservoirs
for
Tolerance
Reassessment;
electronic
copy
available
from
the
EPA
OPP
home
page
at
http://
www.
epa.
gov/
pesticides/
SAP/
1998/
index.
htm#
october
.

29
possible,
EPA
uses
available
monitoring
data
to
evaluate
the
impact
of
non­
agricultural
pesticide
use
on
water
quality.
However,
where
monitoring
data
are
absent
or
of
limited
scope,
the
Agency
can
provide
only
a
qualitative
assessment
of
the
impact
of
non­
agricultural
use
of
pesticides
on
drinking
water.
In
these
instances,
the
Agency
will
often
require
additional
data,
including
monitoring
studies,
in
order
to
make
an
assessment
of
the
non­
agricultural
contributions
of
pesticide
contamination
to
drinking
water
sources.
The
Agency
also
note
this
type
of
limitation
in
its
drinking
water
exposure
assessments.

E.
Implementation
of
the
Index
Reservoir
(
IR)
in
Surface
Water
Screening
Models
1.
Replacing
the
Farm
Pond
With
an
Index
Reservoir
Most
of
the
commentors
[
1,
2,
3,
5,
6,
7]
expressed
concern
with
the
Agency's
plan
to
replace
the
current
farm
pond
with
an
index
reservoir
in
its
tier
2
surface
water
screening
model.
However,
their
reasons
for
concern
differed.
Two
commentors
[
1,
6]
were
concerned
that
the
index
reservoir
would
not
be
representative
of
highly
vulnerable
reservoirs
while
other
commentors
[
2,
3,
5,
7]
were
concerned
that
the
index
reservoir
would
be
biased
toward
sites
which
have
unreasonably
high
vulnerability.
Two
[
3,
7]
disagreed
with
the
Agency's
assessment
that
the
index
reservoir
would
provide
more
realistic
estimates
of
pesticide
concentrations
in
surface
water
for
the
following
reasons:

°
an
increased
ratio
of
treated
area
to
water
volume
will
increase
estimated
concentrations;
°
no
clear
relationship
exists
between
drainage
area
to
reservoir
normal
capacity
(
DA/
NC)
and
concentration;
°
the
assumption
that
the
treated
fields
are
adjacent
to
the
reservoir
does
not
account
for
spatial
distribution,
buffering
effects,
or
agronomic/
management
practices;
°
drift
is
not
likely
to
contribute
significantly
to
pesticide
levels
in
larger
bodies
of
water
and
should
be
eliminated
from
the
model.

Agency
Response:
The
reservoir
parameters
chosen
were
similar
to
a
number
of
small
reservoirs
located
in
intensively
cropped
areas
in
the
Midwestern
United
States.
The
Agency
evaluated
actual
monitoring
data
to
select
a
site
that
represented
a
"
high­
end"
site
in
terms
of
concentration
of
pesticides
in
the
reservoir.
In
so
doing,
the
Agency
believes
that
it
has
selected
a
site
that
is
representative
of
vulnerable
reservoirs
but
is
not
unreasonably
vulnerable.

The
ratio
of
watershed
area
to
volume
used
for
the
index
reservoir
is
based
on
parameters
from
an
actual
reservoir.
Since
the
Agency
plans
to
use
the
index
reservoir
in
combination
with
the
percent
cropped
area,
the
ratio
of
cropped
watershed
area
to
volume
will
decrease
for
virtually
all
uses.
Indeed,
model
scenario
comparisons
presented
in
the
May
1999
SAP
report
and
the
July
1998
SAP
report
show
this
to
be
the
case.
14
More
information
on
the
ARP
study
is
available
from
the
EPA
OPP
home
page
at
15
http://
www.
epa.
gov/
oppefed1/
aceto/
index.
htm
.

30
Two
commentors
referred
to
data
presented
for
13
selected
reservoirs
in
the
July
1998
SAP
report
to
make
their
case
that
no
relationship
exists
between
DA/
NC
and
pesticide
concentration.
The
Agency's
assessment
of
the
DA/
NC
­
concentration
relationship
was
based
on
an
evaluation
of
all
the
reservoirs
included
in
the
Acetochlor
Registration
Partnership
(
ARP)
study,
which
is
monitoring
175
surface
water
sources
of
drinking
water
from
the
corn­
growing
region
of
the
midwest
east
to
Pennsylvania
and
Delaware
.
When
all
reservoirs
are
considered,
15
higher
pesticide
concentrations
in
water
tended
to
be
associated
with
the
larger
DA/
NCs.
The
13
reservoirs
selected
for
further
consideration
were
those
considered
to
be
vulnerable
based
on
monitoring
data
and
do
not
represent
the
full
range
in
pesticide
concentrations
or
DA/
NCs.

The
current
Tier
2
surface
water
model
is
not
capable
of
accounting
for
the
spatial
distribution
of
cropping
patterns,
pesticide
use,
or
location
of
water
bodies
within
a
watershed.
The
Agency
anticipates
that
a
watershed­
scale
model,
when
it
becomes
available,
would
be
able
to
account
for
spatial
distribution
of
crops
in
the
drainage
basin.
Buffers
and
management
practices
are
not
always
specified
on
the
label
nor
consistently
implemented
in
the
field,
making
it
difficult
for
the
Agency
to
include
them
in
the
modeling
scenario.
If
specific
buffers
or
management
practices
are
included
in
the
pesticide
label,
these
additional
factors
are
addressed
by
EPA
in
its
characterization
of
the
risk
posed
by
the
pesticide
or
in
further
model
refinements,
if
warranted.

The
spray
drift
contribution
to
a
larger
water
body
is
influenced
by
factors
such
as
size
and
geometry
of
the
water
body,
location
of
the
treated
field
in
relation
to
the
receiving
reservoir
and
streams
which
flow
into
the
reservoir,
and
will
be
reevaluated.
While
the
contribution
by
pesticide
drift
to
drinking
water
residues
is
likely
to
be
lower
for
a
reservoir
than
that
for
the
farm
pond,
the
Agency
does
not
believe
it
should
be
ignored
altogether.
The
Agency
plans
to
use
the
Spray
Drift
Task
Force
data
on
spray
drift
to
determine
an
appropriate
estimate
of
drift
loading
to
the
index
reservoir.
EPA
also
plans
to
evaluate
drift
loading
in
the
watershed­
scale
model
evaluations.
This
work
is
currently
in
progress.

2.
Appropriateness
of
Using
Small
Reservoirs
One
commentor
[
3]
questioned
the
appropriateness
of
small
reservoirs,
arguing
that
they
are
generally
more
prone
to
local
factors,
especially
cropping
patterns
and
pesticides
used,
that
"
have
little
or
no
relevance
to
national
exposures,
even
for
the
99
percentile."
However,
two
th
commentors
[
1,
6]
were
concerned
that
the
index
reservoir,
coupled
with
cropping
adjustments,
may
not
be
"
truly
representative
of
high
risk
scenarios."
None
of
the
commentors
provided
concrete
evidence
to
support
their
concerns.

Agency
Response:
The
Agency
disagrees
that
an
assessment
based
on
a
small
reservoir
which
may
be
prone
to
local
factors
has
"
little
or
no
relevance"
to
the
estimation
of
drinking
water
exposure
for
purposes
of
aggregate
risk
assessment
under
FQPA.
The
intent
of
the
screening
model
is
to
determine
whether
estimated
high­
end
pesticide
concentrations
in
drinking
31
water
could
exceed
a
DWLOC.
If,
as
available
monitoring
data
seem
to
indicate,
the
estimates
based
on
EPA's
index
reservoir
represent
a
high­
exposure
scenario,
then
the
Agency
can
conclude
with
confidence
that
the
pesticide
does
not
pose
a
concern
nationally
if
the
screening
model
estimates
are
below
the
DWLOC.
If,
however,
the
estimates
are
similar
to
or
greater
than
the
DWLOC,
then
the
model
estimates
indicate
a
potential
concern.
Whether
a
significant
subpopulation
would
be
exposed
to
concentrations
of
concern
would
depend
on
how
frequently
those
conditions
would
be
expected
to
occur
nationally.
Arguably,
some
number
of
identifiable
communities
of
25
people
or
more
could
be
viewed
as
a
significant
subpopulation
from
a
drinking
water
perspective.

Available
monitoring
data
seem
to
indicate
that
the
index
reservoir
does
represent
a
high
exposure
scenario.
The
index
reservoir
was
selected
from
the
175
surface
water
sources
of
drinking
water
included
in
the
ARP
study.
The
study
is
biased
toward
vulnerable
watersheds
in
intensive
corn­
growing
regions,
although
it
also
includes
larger
bodies
of
water
(
including
the
Great
Lakes),
flowing
water,
and
reservoirs
in
less
intensively
cropped
areas.
Thus,
Shipman
City
Lake,
the
selected
index
reservoir,
is
ranked
8
of
the
175
surface
water
sources
of
drinking
water
th
(
95
percentile)
in
terms
of
concentration
of
atrazine,
from
a
group
of
reservoirs
located
in
an
th
intensive
pesticide
use
region
of
the
country.

3.
Representativeness
of
the
Midwestern
Index
Reservoir
Several
commentors
[
2,
3,
7]
questioned
whether
a
midwestern
index
reservoir,
which
applies
primarily
to
corn,
would
be
representative
outside
the
Midwestern
U.
S.
Two
commentors
[
3,
7]
recommended
a
series
of
"
representative"
reservoirs
for
each
geographic
region
and/
or
crop.
They
asserted
that
advanced
(
tier
2)
screening
models
should
incorporate
region­
specific
parameters
(
crop
area
factors,
rainfall,
watershed
size,
soil
and
watershed
variations,
etc)
to
provide
"
more
realistic
concentrations"
than
Tier
1
screening
models.

Agency
Response:
The
Agency
agrees
that
estimates
of
pesticides
in
drinking
water
based
on
the
index
reservoir
may
not
be
as
accurate
for
other
parts
of
the
country
as
they
are
for
Midwest
locations.
However,
the
modeling
scenarios
currently
do
account
for
region­
specific
rainfall,
soil,
and
hydrologic/
runoff
factors.
The
incorporation
of
an
index
reservoir
is
the
latest
step
in
an
ongoing
effort
to
improve
EPA's
drinking
water
risk
assessments
that
will
eventually
include
basin­
scale
modeling.
The
Agency
recognizes
the
need
to
develop
regional
reservoirs
for
advanced
tiers
of
modeling
as
well
as
for
basin­
scale
modeling.
However,
this
step
is
hampered
by
the
lack
of
monitoring
data
outside
of
the
Midwestern
U.
S.
that
are
of
sufficient
quality
and
extent
to
develop
scenarios
for
additional
reservoirs.
As
these
data
become
available,
EPA
will
develop
regional
reservoir
scenarios
to
go
with
the
current
index
reservoir.

The
Agency
is
using
what
it
believes
to
be
a
vulnerable
reservoir
scenario
in
a
screening
model
to
determine
whether
a
potential
concern
exists.
If
the
model,
when
used
in
other
regions,
shows
pesticide
concentrations
in
water
at
or
above
levels
of
concern,
then
region­
specific
factors
will
be
considered
in
further
refinements.
The
refinements
may
include
targeted
monitoring
in
areas
where
the
pesticide
is
used.
In
its
comments
to
the
Agency's
May
1999
report,
the
SAP
recommended
that
"
more
consideration
should
be
given
to
low­
cost,
targeted
monitoring,
32
especially
in
the
case
of
minor­
use
crops
where
modeling
efforts
tend
to
be
imprecise"
[
SAP
Report
No.
99­
03C,
May
27,
1999].

4.
Peer
Review
and
Calibration
of
the
Index
Reservoir
Model
Two
commentors
[
3,
7]
opposed
adopting
the
index
reservoir
until
the
percent
crop
area
(
PCA)
is
available
and
the
model
is
"
properly
calibrated
with
reliable
and
representative
monitoring
data."
They
urged
peer
review
of
the
model,
with
further
presentation
to
the
SAP
and
stakeholder
groups
before
adoption.
Indeed,
several
commentors
[
1,
2,
3,
6,
7]
urged
the
Agency
to
compare
model
predictions
to
available
monitoring
data
for
calibration.
Some
commentors
[
2,
3,
7]
felt
that
model
calibration/
validation
was
essential
before
the
index
reservoir
is
used
for
regulatory
purposes.
They
also
recommended
that
the
model
predictions
be
calibrated
against
alternate
models
such
as
the
regression­
based
Surface
Water
Mobility
Index
(
SWMI)
developed
by
Novartis
Crop
Protection,
Inc.,
and
the
American
Crop
Protection
Association
(
ACPA)
[
3,
7].
Other
commentors
[
1,
6]
felt
an
evaluation
of
model
predictions
against
monitoring
data
is
essential
to
determine
whether
the
assumptions
will
include
the
most
highly
exposed
populations
and
whether
the
index
reservoir
"
is
truly
representative
of
high
risk
scenarios."

Agency
Response:
The
Agency
presented
methods
to
derive
watershed­
based
cropping
area
adjustments
to
the
SAP
in
May
1999
and
plans
to
implement
the
index
reservoir
in
conjunction
with
a
crop
area
adjustment.
The
SAP
generally
supported
the
Agency's
tiered
approach
for
drinking
water
assessments,
including
the
use
of
a
percentage
crop
area
adjustment
coupled
with
the
index
reservoir.
The
Agency
agrees
that
peer
review
and
extensive
model
evaluation
are
needed
and
continues
with
efforts
to
do
so.
The
SAP
presentations
referred
to
in
this
comment/
response
document
and
in
the
drinking
water
science
policy
paper
are
part
of
this
continuing
process.
In
the
interim,
EPA
is
tasked
with
completing
drinking
water
assessments
as
part
of
FQPA
and
is
doing
them
with
what
it
believes
to
be
the
best
tools
available.
Comparisons
of
model
estimates
with
monitoring
data
were
done
in
evaluating
the
index
reservoir
and
percent
cropped
area
adjustments
[
July
1998
and
May
1999
SAP
reports]
and
will
continue
as
additional
monitoring
data
are
obtained.
The
Agency
feels
that
the
index
reservoir
and
percent
cropped
area
modifications
are
an
improvement
over
existing
modeling
approaches
and,
for
that
reason,
is
moving
forward
with
those
changes.

The
Agency
disagrees
with
the
recommendation
to
calibrate
one
model
against
another.
In
particular,
OPP
has
concluded
that
SWMI
is
not
considered
a
valid
model
for
concerns
that
are
addressed
in
the
later
section
on
Watershed­
or
Basin­
Scale
Modeling
(
Section
G).

F.
Accounting
For
the
Percent
Cropped
Area
in
Surface
Water
Screening
Models
1.
Use
of
a
Percent
Cropped
Area
(
Cropped
Area
Factor)

Comments
were
divided
on
the
use
of
a
correction
factor
to
account
for
the
area
of
the
watershed
that
is
planted
to
a
specific
crop
or
crops
of
concern
[
referred
to
as
the
Crop
Area
Factor
(
CAF)
in
the
initial
science
policy
document,
the
term
has
been
changed
to
Percent
Cropped
Area
(
PCA)
to
be
more
in
line
with
the
Percent
Crop
Treated
terminology
already
in
use
33
in
the
Agency].
Two
commentors
[
3,
7]
agreed
with
the
approach
while
two
others
[
1,
6]
questioned
whether
the
PCA
approach
would
represent
a
"
highly
vulnerable"
system.
One
commentor
[
1]
warned
against
selecting
a
factor
that
represented
the
"
average"
crop
mixture
in
a
watershed,
noting
that
drinking
water
treatment
is
"
challenged"
where
cropping
is
intense,
best
management
practices
are
inadequate
or
not
applied,
precipitation
events
carry
recently­
applied
pesticides
to
streams,
and
multiple
pesticide
applications
are
made.
Another
commentor
[
6]
encouraged
the
Agency
to
conduct
a
"
nationwide
review
of
data"
to
determine
whether
PCAadjusted
model
estimates
capture
high­
exposure
scenarios
before
applying
the
factor
to
screening
model
results.

Agency
Response:
Since
submitting
its
initial
science
policy
document
for
public
comment,
the
Agency
presented
its
plan
for
implementing
the
percent
cropped
area
(
PCA)
as
a
refinement
to
the
FQPA
drinking
water
assessment
process
to
the
FIFRA
Scientific
Advisory
Panel
(
SAP)
in
May
1999.
The
SAP
agreed
with
the
concept
of
the
PCA
as
an
"
appropriate
and
reasonable"
adjustment
for
major­
use
crops
while
still
providing
an
effective
initial
screen
(
i.
e.,
one
that
provides
EPA
with
a
high
degree
of
confidence
that
"
cleared"
pesticides
will
in
fact
not
pose
a
significant
risk
of
exposure
through
the
drinking
water
route).
The
Panel
observed
that
the
PCA
"
provides
a
technically
defensible
approach
to
reduce
estimates
of
acute
and
chronic
pesticide
exposures
to
levels
similar
to
those
found
in
monitoring
data."
However,
the
SAP
also
identified
several
limitations
to
the
approach
and,
particularly
for
assessments
involving
minor­
use
crops,
recommended
that
the
Agency
consider
requesting
"
low­
cost,
targeted
monitoring"
and
field
experiments
using
rainfall
simulators
and
other
techniques
to
evaluate
pesticide
loss
under
extreme
rainfall
conditions.
Such
information
would
improve
the
Agency's
ability
to
predict
pesticide
concentrations
in
water
and
increase
the
amount
of
data
available
to
evaluate
models
[
SAP
Report
No.
99­
03C,
May
27,
1999].

2.
Percent
Cropped
Treated
Two
commentors
[
3,
7]
recommended
that
the
Agency
go
beyond
the
Percent
Cropped
Area
(
PCA)
approach
to
use
"
product­
specific
area
factors"
(
e.
g.,
percent
crop
treated)
when
market
share
information
is
available.

Agency
Response:
The
use
of
percent
crop
treated
is
limited
by
the
availability
of
data
at
a
sufficiently
detailed
scale
in
order
to
determine
pesticide
use
distribution
within
a
watershed.
Currently,
only
two
states
(
California
and
New
York)
collect
pesticide
usage
data
at
such
a
detailed
scale.
The
Agency
is
concerned
that
incorporation
of
percent
crop
treated
would
compromise
the
protective
nature
of
the
screening
models.
A
national
average
percent
crop
treated
would
not
be
appropriate,
since
it
does
not
reflect
the
variation
in
percentage
of
crop
treated
across
the
country
(
i.
e.,
while
in
some
watersheds,
the
percentage
of
crop
treated
may
be
less
than
the
national
average,
in
other
watersheds,
it
may
be
much
greater).
Such
an
adjustment
would
also
need
to
account
for
temporal
changes
in
pesticide
usage
resulting
from
changes
in
pest
pressures,
management
practices,
and
alternate
treatments.
As
model
improvements
are
made
and
as
more
and
better
quality
monitoring
and
pesticide
usage
data
become
available
nationwide,
the
Agency
may
consider
such
adjustments
in
the
future
for
advanced
tier
refinements
in
a
basin­
scale
model.
34
The
majority
of
the
May
1999
SAP
agreed
that
the
Agency
should
consider
percent
crop
treated
in
future
model
refinements.
The
Panel
advised
OPP
to
analyze
the
existing
New
York
and
California
data
to
determine
the
extent
of
discrepancy
between
"
percent
crop
area"
and
"
percent
crop
treated"
and
then
"
make
an
educated
decision
on
how
to
handle
this
issue."
The
SAP
noted
that
relatively
high
uncertainties
may
be
encountered
for
chemicals
which
are
applied
to
less
than
10
%
of
the
cropped
area.
One
Panel
member
disagreed
with
the
recommendation,
commenting
that
the
use
of
percent
crop
treated
data
moves
beyond
the
original
intent
of
a
screening
approach.
Use
of
this
data
would
increase
the
site­
specificity
of
the
technique
and
may
not
appropriately
represent
a
worst­
case
scenario
for
screening
purposes.
In
its
presentation
to
the
May
SAP,
the
Agency
illustrated
the
effect
of
the
percent
crop
treated
adjustment
for
diazinon
in
the
San
Joaquin
Valley.
This
adjustment
reduced
the
screening
model
estimates
to
a
value
that
was
an
order
of
magnitude
lower
than
the
peak
concentrations
detected
in
the
NAQWA
monitoring
data.

3.
Technical
Comments
on
PCA
Implementation
Additional
comments
related
to
the
specifics
of
deriving
and
implementing
the
PCA
factor
included
using
county­
based
rather
than
watershed­
derived
PCAs
[
3],
determining
PCAs
for
minor­
use
crops
[
7],
and
accounting
for
more
than
one
crop
use
in
a
watershed
[
3,
7].
One
commentor
[
3]
encouraged
the
Agency
to
use
county­
based
PCAs
because
they
were
readily
available.
Another
commentor
[
7]
noted
the
challenges
in
deriving
and
validating
PCAs
for
minor
crops
and
recommended
that
the
Agency
either
(
1)
validate
PCAs
with
major
crops
and
extrapolate
to
minor
crops,
or
(
2)
use
a
default
PCA
reflective
of
minor
crop
status.
Recommendations
for
accounting
for
pesticide
use
on
more
than
one
crop
in
a
watershed
focused
on
weighting
for
crop
area
and
accounting
for
temporal
differences
in
application
[
3,
7].

Agency
Response:
In
December
1997,
the
SAP
recommended
that,
since
the
Agency
is
assessing
a
watershed
process,
it
develop
PCAs
on
a
physiographically­
based
unit
(
i.
e.,
watershed
boundary)
rather
than
on
a
political
unit
(
i.
e.,
county
boundary).
This
recommendation
was
supported
by
the
latest
SAP
in
May
1999.
The
May
1999
SAP
also
expressed
concern
that
the
Agency
would
be
unable
to
validate
PCAs
for
minor
crops
because
of
a
lack
of
monitoring
data
for
evaluation
and
recommended
that
EPA
consider
requesting
low­
cost,
targeted
monitoring
data
to
evaluate
pesticide
contamination
from
use
on
minor
crops
[
SAP
Report
No.
99­
03C,
May
27,
1999].

For
multiple
crop
use,
the
May
1999
SAP
recommended
that
the
Agency
could
derive
PCAs
based
on
the
maximum
combined
acreage
of
crops
in
a
watershed.
If
pesticide
application
rate
and
timing
vary
from
crop
to
crop,
an
aggregate
pesticide
concentration
estimate
could
be
made
by
separately
simulating
each
crop
in
the
watershed
and
then
summing
the
individual
model
estimates.
EPA
plans
to
incorporate
the
SAP
recommendations
when
it
implements
the
PCA.

G.
Watershed­
or
Basin­
Scale
Models
In
general,
the
commentors
[
1,
3,
7]
supported
the
Agency's
plan
to
develop
watershedscale
models,
but
were
not
aware
of
any
validated
watershed­
scale
models
currently
in
existence.
FIFRA
Scientific
Advisory
Panel
report
on
A
Set
of
Scientific
Issues
Being
Considered
by
the
16
Agency
in
Connection
with
Estimating
Drinking
Water
Exposure
as
a
Component
of
the
Dietary
Risk
Assessment;
electronic
copy
available
from
the
EPA
OPP
home
page
at
http://
www.
epa.
gov/
pesticides/
SAP/
1997/
december/
finaldec.
pdf
.

A
May
7,
1999,
memo
(
Review
Comments
on
"
A
Simple
Regression
Model
for
Predicting
Surface
17
Water
Concentrations
Resulting
from
Agricultural
Field
Runoff
and
Erosion")
from
Parker,
Hetrick,
and
the
Water
Quality
Tech
Team,
through
Joe
Merenda,
EFED
Director,
to
Keefer,
Gilding,
and
the
ACPA
Drinking
Water
Exposure
Workgroup
provided
detailed
comments
on
the
SWMI
model.

35
These
commentors
noted
the
complexity
of
factors
involved
in
watershed­
scale
modeling
as
well
as
the
difficulty
in
developing
such
a
model.
In
addition
to
the
models
the
Agency
discussed
in
the
July
1998
SAP
report,
two
commentors
[
3,
7]
recommended
using
a
regression­
based
model
such
as
the
Surface
Water
Mobility
Index
(
SWMI)
developed
by
Novartis
Crop
Protection,
Inc.,
a
pesticide
registrant,
and
the
American
Crop
Protection
Association
(
ACPA),
the
pesticide
industry
trade
association,
because
"
it
is
based
on
real
monitoring
data
that
reflect
all
relevant
landscape
factors."

Agency
Response:
The
Agency
generally
agrees
with
these
comments;
OPP
wants
to
develop
watershed­
scale
models
for
use
in
refined
estimates
of
pesticide
concentrations
in
drinking
water
when
a
screening
level
model
estimate
indicates
a
potential
risk
may
exist.
As
noted
earlier,
the
Agency
intends
to
shift
its
resources
toward
developing
watershed­
or
basinscale
modeling
and
away
from
further
refinements
in
its
screening
approach/
method.
A
validated,
mechanistic
basin­
scale
model
will
address
many
of
the
concerns
that
arise
from
applying
fieldscale
models,
such
as
PRZM,
to
basin­
scale
assessments.
EPA
is
aware
of
the
difficulties
in
developing
and
evaluating
such
a
model,
and
expects
that
a
usable
basin­
scale
model
is
several
years
away.
In
commenting
on
the
Agency's
May
1999
presentation
on
its
proposed
method
for
incorporating
the
percent
cropped
area
into
surface
water
models,
the
SAP
expressed
the
opinion
that
PRZM
is
not
the
model
to
use
if
the
Agency
is
seeking
to
develop
a
representative
watershed
or
basin
configuration.
However,
the
SAP
also
noted
that
addressing
the
limitations
in
the
current
modeling
process
will
require
"
considerable
investment
of
resources."

The
SAP
encouraged
the
Agency
to
continue
evaluating
other
models
as
suggested
in
the
July
1998
SAP
report,
especially
watershed­
based
models
in
a
GIS
environment.
The
Panel
noted
that
watershed­
based
regression
models
could
be
developed
with
increasing
availability
of
highquality
data,
but
that
extrapolation
to
areas,
times,
and
conditions
beyond
the
range
of
available
data
may
provide
inaccurate
estimates
[
SAP
Report
No.
99­
03C,
May
27,
1999].
Similar
concerns
about
the
transportability
of
the
regression
models
beyond
the
range
of
the
data
set
used
for
development
were
expressed
in
the
December
1997
SAP
response
.
16
The
Agency
previously
reviewed
and
commented
on
the
SWMI
model
developed
by
Novartis
and
ACPA
.
While
the
data
set
used
for
model
development
(
Lake
Erie
Basin
data
17
collected
by
Baker
and
others
at
Heidelburg,
OH,
College)
is
one
of
the
more
complete
pesticide
data
sets
for
the
Midwest,
the
Agency
is
concerned
about
the
ability
to
use
this
model
on
a
national
scale
for
a
number
of
reasons.
The
model
is
a
regression
based
on
data
collected
from
a
36
limited
geographic
region
(
six
watersheds
in
Ohio).
Pesticide
concentrations
reported
in
monitoring
studies
from
other
geographic
areas
are
frequently
an
order
of
magnitude
or
more
higher
than
the
concentrations
measured
in
the
Lake
Erie
basin.
Thus,
the
Lake
Erie
basin
data
may
not
necessarily
reflect
the
more
vulnerable
watersheds
of
the
region.
This
becomes
a
major
concern
since
a
regression­
based
model,
unlike
a
mechanistic
model,
is
limited
to
the
geographical
and
temporal
boundaries
of
the
data
set.

H.
Ground
Water
Screening
Models
1.
SCI­
GROW
Two
commentors
[
3,
7]
expressed
an
opinion
that
SCI­
GROW
is
an
appropriate
tool
for
screening
in
"
highly
vulnerable"
shallow
aquifers,
but
recommended
further
"
verification"
with
existing
monitoring
data
to
establish
the
range
in
pesticide
properties
over
which
the
model
would
be
reliable.

Agency
Response:
The
December
1997
SAP
recommended
that
OPP
fully
document
and
publish
SCI­
GROW,
characterize
uncertainties
in
the
estimates,
and
determine
the
likelihood
of
false
negatives
in
a
SCI­
GROW
screen.
The
Agency
plans
to
complete
documentation
of
the
SCI­
GROW
model
by
the
end
of
1999.
Once
completed,
the
documentation
will
be
published
as
recommended
by
the
SAP.
Further
steps
involve
evaluating
the
model
against
other
monitoring
data
to
improve
and
expand
the
current
range
in
pesticide
properties
over
which
the
model
can
provide
predictions
and
to
subject
the
model
to
an
internal
and
external
peer
review
and
evaluation
process.

2.
Tier
2
Ground
Water
Model
Two
commentors
[
3,
7]
urged
the
Agency
to
develop
a
higher
tier
model
capable
of
predicting
"
more
realistic
pesticide
concentrations
in
drinking
wells."
Another
commentor
[
1]
recommended
incorporating
decision
criteria
into
the
"
more
complex"
screening
models
to
distinguish
four
management
options:

­
register
and
provide
a
tolerance
­
conditionally
register
and
provide
a
tolerance
­
deny
registration
pending
additional
information
­
deny
registration
Agency
Response:
The
Agency
plans
to
develop
a
tier
2
ground
water
model
after
it
completes
work
on
SCI­
GROW.
Although
no
tier
2
model
currently
exists
for
ground
water,
the
Agency
requests
additional
information,
usually
in
the
form
of
prospective
ground
water
monitoring
studies,
when
SCI­
GROW
results
in
estimated
concentrations
of
pesticides
in
ground
water
that
exceed
the
DWLOC.

The
initial
science
policy
document
was
intended
to
provide
a
framework
to
describe
how
science
will
be
used
in
doing
drinking
water
assessments.
It
was
not
intended
to
address
37
management
options,
such
as
the
registration
triggers
suggested
by
one
commentor.
Such
decisions
take
into
account
a
broader
range
of
factors,
of
which
the
scientific
data
are
one
part.

I.
Probabilistic
Drinking
Water
Assessments
1.
Development
of
Probabilistic
Models
One
commentor
[
3]
noted
that
because
drinking
water
varies
locally,
a
probabilistic
approach
to
drinking
water
exposure
assessment
must
incorporate
a
geospatial
distribution
based
on
monitoring
and/
or
modeling.
Much
remains
to
be
done
to
develop
"
adequate
and
reliable
probabilistic
methods"
and
the
necessary
data
for
these
methods.
Moreover,
the
commentor
encouraged
OPP
to
work
with
other
offices
in
the
Agency
(
Office
of
Research
and
Development
and
Office
of
Water)
as
well
as
other
government
agencies
and
stakeholder
groups
to
develop
the
necessary
tools
and
data.
OPP
must
first
determine
that
it
has
"
reliable
information
on
the
distribution
of
exposure
to
a
pesticide
via
drinking
water
on
a
national
or
regional
basis"
[
3].

Agency
Response:
While
the
long­
term
goal
of
the
Agency
is
to
use
probabilistic
drinking
water
exposure
assessments
for
tolerance
assessments
under
FQPA,
the
tools
and
data
for
performing
probabilistic
assessments
are
not
yet
available.
Since
pesticide
concentrations
vary
both
in
time
and
in
location,
a
probabilistic
approach
for
drinking
water
exposure
assessments
will
be
more
valuable
if
it
can
incorporate
spatial
and
temporal
distributions.
However,
at
this
point
the
Agency
has
only
limited
information
on
the
spatial
and
temporal
distributions
in
pesticide
concentrations
and
would
have
to
make
numerous
non­
verifiable
simplifying
assumptions
for
distributions
that
are
not
available.

2.
Potential
Conflict
of
Interest
In
the
draft
document,
the
Agency
noted
that
it
was
working
cooperatively
with
the
International
Life
Science
Institute
(
ILSI)
to
advance
the
development
of
probabilistic
drinking
water
assessments.
One
commentor
[
6]
raised
a
concern
about
a
potential
conflict
of
interest
with
ILSI,
which
"
has
an
economic
interest
in
pesticide
regulations,
and
is
governed
by
the
very
industry
EPA
is
charged
with
regulating.
EPA's
reliance
on
ILSI
for
advice
here
violates
the
Federal
Advisory
Committee
Act
and
the
Agency's
own
conflict
of
interest
requirements."

Agency
Response:
ILSI
is
an
independent,
nonprofit
foundation
established
to
advance
the
understanding
of
scientific
issues
related
to
nutrition,
food
safety,
toxicology,
and
the
environment.
Through
its
Risk
Science
Institute,
ILSI
brings
together
experts
from
academia,
industry,
government,
and
public
interest
groups
to
address
cutting­
edge
scientific
issues.
These
expert
groups
meet
in
sessions
open
to
the
public
and
prepare
reports
for
the
Agency
which
are
also
distributed
to
the
public.
EPA's
cooperative,
open
and
public
relationship
with
ILSI
neither
presents
a
conflict
of
interest
nor
a
violation
of
the
Federal
Advisory
Committee
Act.

J.
Interpreting
and
Using
Monitoring
Data
in
Drinking
Water
Assessments
1.
Adequacy
of
Existing
Monitoring
Data
38
Differences
of
opinion
exist
on
the
quality
and
quantity
of
existing
monitoring
data
available
to
the
Agency
and
how
these
data
should
be
incorporated
into
drinking
water
exposure
assessments.
While
one
commentor
[
3]
believed
"
the
science
and
technology,
including
monitoring
databases,
exist
to
develop
and
support
realistic
risk
assessment
tiers
for
drinking
water,
beyond
the
preliminary
screening
tiers"
and
pushed
for
use
of
monitoring
over
modeling,
another
[
1]
said
that
the
Agency
would
need
to
rely
on
modeling
because
monitoring
data
adequate
to
support
registration
decisions
is
unlikely
to
be
available
in
the
near
term.
Some
commentors
[
3,
7]
felt
that
the
Agency
does
not
use
monitoring
data
as
much
as
it
could
while
another
[
6]
recommended
that
EPA
abandon
the
approach
of
relying
on
data
that
may
not
be
representative
of
especially
vulnerable
water
sources.

Agency
Response:
As
noted
in
the
draft
science
policy
document,
the
Agency
gathers
available
monitoring
data
for
use
in
the
drinking
water
exposure
assessment
for
a
pesticide
if
the
screening
model
estimates
are
close
to
or
exceed
the
drinking
water
level
of
comparison
(
DWLOC).
The
science
policy
document
listed
typical
sources
of
monitoring
data
used
by
the
Agency.
Commentors
[
3,
7]
mentioned
additional
sources,
such
as
Lake
Erie
Basin
data
collected
at
Heidelburg
(
OH)
College,
the
ARP
study,
and
Novartis'
Atrazine
Volunteer
Monitoring
program.
The
Agency
is
aware
of
and
uses
these
data
sources,
as
well
as
others,
in
its
assessments.
To
help
interpret
the
monitoring
data,
the
Agency
gathers
as
much
information
as
it
can
on
where
and
when
the
samples
were
collected,
the
circumstances
surrounding
the
collection,
how
the
samples
were
collected
and
analyzed,
sample
locations
in
relation
to
pesticide
usage,
timing
of
samples
in
relation
to
time
of
pesticide
application
in
the
sample
area,
nature
and
size
of
the
water
body,
and
size
and
characteristics
of
the
area
draining
into
the
water
body.

The
quantity
and
quality
of
data
varies
from
pesticide
to
pesticide.
Even
the
data
sources
mentioned
above
are
limited
in
geographic
extent
(
primarily
in
the
Midwestern
U.
S.)
and
in
time.
Therefore,
the
Agency's
decision
to
use
monitoring
data
involves
consideration
of
many
factors
and
each
judgment
is
case­
by­
case.
In
some
instances,
the
monitoring
data
does
not
provide
helpful
or
adequate
information
to
characterize
reasonable
high­
end
exposures.

2.
Evaluating
Monitoring
Data
Several
commentors
[
3,
7]
urged
the
Agency
to
develop
"
clear,
formal
guidance"
on
how
monitoring
data
is
evaluated
for
usefulness
and
to
use
"
all
monitoring
data
available
and
meeting
predefined
quality
criteria"
for
drinking
water
exposure
assessments.
Although
none
of
the
commentors
provided
specific
criteria
for
evaluating
monitoring
data,
several
[
1,
3,
6,
7]
identified
factors
that
need
to
be
considered,
including
distribution
across
the
cropped
region
and
pesticide
use
area,
design
and
purpose
of
the
study,
vulnerability
of
the
sites,
representativeness
of
actual
drinking
water
sources,
monitoring
of
both
source
and
treated
water,
sampling
frequency
sufficient
to
capture
occurrence
over
time,
analytical
detection
levels
adequate
to
support
aggregate
analysis,
inclusion
of
important
metabolites
and
degradates,
and
complete
characterization
of
watershed,
cropping
patterns,
pesticide
application,
water
treatment,
and
water
quality.
One
commentor
[
3]
referred
to
an
upcoming
ILSI
report
on
the
amount
of
sampling
needed
to
address
local,
regional,
and
national
assessments.
39
Agency
Response:
The
Agency
agrees
that
developing
criteria
for
evaluating
monitoring
data
will
not
only
aid
in
the
evaluation
of
current
data,
but
will
help
guide
the
design
of
future
monitoring
studies.
Currently,
standardized
OPP
guidance
on
assessing
water
monitoring
data
do
not
exist;
the
criteria
for
such
assessments
will
depend
on
whether
the
data
will
be
used
for
model
validation
or
estimating
drinking
water
exposures
for
a
single
pesticide.
The
Agency
will
be
looking
to
the
ILSI
report
and
to
other
sources
for
guidance
in
assessing
the
usability
of
monitoring
data
in
pesticide
exposure
assessments.
OPP
plans
to
develop
interim
guidance
during
FY2000.

3.
Use
of
Geographical
Information
Systems
(
GIS)

One
commentor
[
3]
urged
the
Agency
to
take
advantage
of
Geographic
Information
System
(
GIS)
tools
in
evaluating
monitoring
data.

Agency
Response:
The
Agency
believes
that
GIS
tools,
coupled
with
more
detailed
site,
environment,
and
use
characterizations,
will
assist
in
characterizing
and
evaluating
existing
monitoring
data.
GIS
tools
will
also
help
the
Agency
assess
potentially
exposed
populations
and
identify
gaps
in
existing
data
in
order
to
better
target
additional
monitoring.
The
Agency
continues
to
seek
and
develop
such
tools
to
improve
its
assessment
of
pesticide
exposure
from
drinking
water
sources.
At
the
same
time,
the
Agency
believes
that
more
monitoring
data,
and
more
ancillary
information
(
site
and
usage
characteristics),
will
be
needed
to
take
full
advantage
of
the
GIS
capabilities
at
hand.

4.
Additional
Monitoring
Data
Needs
Commentors
[
1,
3,
6,
7]
noted
that
additional
monitoring
is
needed
for
pesticides
on
a
number
of
scales.
While
one
commentor
[
7]
felt
the
gaps
should
be
filled
with
"
government
sponsored
monitoring
programs"
targeted
at
drinking
water
sources,
another
[
1]
noted
that
conditional
registrations
with
monitoring
provisions
will
be
necessary
to
obtain
the
data.
A
third
commentor
[
3]
urged
the
Agency
to
grant
"
sufficient
time"
to
allow
for
collection
of
new
monitoring
data
"
of
sufficient
quality
to
refine
screening
level
estimates."

Agency
Response:
The
Agency
is
working
on
a
number
of
levels
to
fill
in
the
gaps
in
monitoring
data
and
acquire
more
high
quality
data
on
pesticide
concentrations
in
drinking
water
sources.
At
pesticide­
specific
levels,
the
Agency
is
requesting
registrant­
sponsored
monitoring
and
runoff
studies
when
screening
models
indicate
a
potential
for
concern.
On
regional
and
national
scales
involving
multiple
pesticides,
the
Agency
is
working
with
the
U.
S.
Geological
Survey
(
USGS)
on
a
pilot
reservoir
monitoring
study
that
will
fill
in
missing
data
on
pesticide
concentrations
in
drinking
water
reservoirs.
EPA
is
also
discussing
design
considerations
for
a
national
pesticide
in
drinking
water
survey
with
various
government
agencies
and
industry
groups.
Such
efforts
cannot
rely
solely
on
"
government
sponsored
monitoring
programs."

5.
Treatment
of
High­
End
Monitoring
Data
In
addition
to
differences
in
what
is
considered
"
adequate
and
reliable"
monitoring
data
for
40
use
in
drinking
water
assessments,
commentors
disagreed
on
how
the
monitoring
data
should
be
used.
One
commentor
[
6]
stated
that
EPA
should
not
"
arbitrarily"
discard
the
highest
reported
values
from
monitoring,
but
use
them
in
the
assessment
if
they
represent
"
real
world"
monitoring
data
for
drinking
water.
However,
another
commentor
[
7]
recommended
using
high­
end
monitoring
data
points
only
as
"
upper­
bound"
estimates
for
screening
purposes,
but
not
quantitatively
in
aggregate
exposure/
risk
assessments.

Agency
Response:
There
is
no
simple
rule
regarding
whether
using
or
excluding
high­
end
monitoring
data
values
is
appropriate
for
risk
assessments.
The
Agency
evaluates
the
monitoring
data
in
relation
to
accompanying
information
on
spatial
and
temporal
distribution
of
the
sample
points,
the
water
body
(
or
bodies)
represented
by
the
sampling,
the
characteristics
of
the
site
surrounding
the
water
body,
and
the
weather/
environmental
conditions
represented
by
the
study.
EPA
also
attempts
to
determine
whether
the
sampled
water
bodies
represent
real
or
likely
drinking
water
sources
and
whether
the
data
represent
potentially
vulnerable
sites
(
as
noted
earlier,
the
concept
of
vulnerability
is
not
a
simple,
easy­
to­
define
concept).
If
the
evaluation
indicates
that
the
data
do
represent
drinking
water
sources,
then
such
data
may
be
used
quantitatively
in
aggregate
exposure
assessments.

6.
Non­
Detects
Several
recommendations
were
made
regarding
interpreting
non­
detections
in
monitoring
data
sets
[
3,
6,
7].
One
commentor
[
6]
notes
that
non­
detects
in
monitoring
data
can
be
misleading
and
should
be
considered
only
in
light
of
the
concerns
listed
in
the
science
policy
document.
Another
commentor
[
7]
recommends
that
the
Agency
confirm
and
"
validate"
nondetects
in
the
same
manner
as
detects.
If
non­
detects
are
from
areas
where
the
product
is
not
used,
the
commentor
recommends
reporting
the
value
as
0;
if
the
non­
detects
are
in
areas
of
use,
report
the
value
as
as
one­
half
of
the
limit
of
detection
(
LOD)
or
quantification
(
LOQ).
This
commentor
encouraged
the
development
of
statistical
"
imputation
methods"
to
define
the
shape
of
data
distributions
below
LOD/
LOQ.

Agency
Response:
Non­
detects
can
indicate
that
the
pesticide
of
concern
is
not
reaching
the
drinking
water
source.
However,
non­
detects
can
also
result
when
the
pesticide
is
analyzed
in
areas
where
or
times
when
the
pesticide
is
not
being
used.
Limitations
with
analytical
methods
may
also
result
in
non­
detects.
All
of
these
factors
will
influence
how
the
Agency
interprets
nondetects
in
monitoring
data
sets.
This
information,
particularly
on
pesticide
usage
in
relation
to
the
monitoring
sites,
is
not
always
available
for
the
pesticide
monitoring
data
provided
to
the
Agency,
thus
making
it
difficult
to
interpret
the
meaning
of
some
reported
non­
detects.
The
Agency
does
not
believe
that
it
is
appropriate
at
this
time,
given
the
current
state
of
monitoring
data
and
ancillary
information,
to
establish
a
policy
such
as
that
which
has
been
applied
to
non­
detects
on
food
(
i.
e.,
to
assume
that
a
pesticide
is
present
at
½
the
LOD
whenever
a
non­
detect
is
reported).

K.
Considering
Water
Treatment
Effects
in
Drinking
Water
Assessments
Two
commentors
[
3,
7]
stated
that
the
Agency
should
always
consider
treated,
or
41
"
finished,"
water
in
its
drinking
water
assessments,
if
such
data
are
available,
since
this
represents
the
actual
exposure
to
the
population.
Both
encouraged
EPA
to
develop
default
pesticide
reduction
factors
by
evaluating
effects
of
drinking
water
treatment
on
pesticide
concentrations
and
effectiveness
variability
among
plants.
Another
commentor
[
6]
cited
statistics
showing
that
a
"
large
number"
of
people
consume
drinking
water
that
is
not
treated
(
17%
of
the
nation
uses
private
drinking
water
sources
which
are
not
usually
treated;
roughly
half
of
drinking
water
comes
from
ground
water,
which
is
generally
not
treated
with
technology
that
is
likely
to
reduce
pesticide
concentrations;
"
large
numbers"
of
drinking
water
facilities
do
not
use
modern
technology
which
would
"
substantially"
reduce
pesticide
concentrations)
to
support
the
position
that
EPA
should
not
focus
on
the
effectiveness
of
drinking
water
treatment
in
pesticide
exposure
assessments.

A
commentor
representing
a
number
of
community
water
system
operators
[
1]
recommended
that
an
appropriate
default
treatment
should
be
"
the
minimum
treatment
allowed
by
federal
regulations
governing
public
water
supplies
and
private
wells."
For
ground
water
supplies,
this
would
be
equivalent
to
no
treatment;
for
community
surface
water
systems,
this
would
be
disinfection
only
(
e.
g.,
chlorine)
for
some
systems,
or
conventional
treatment
(
rapid
mix­
coagulation­
flocculation­
sedimentation­
filtration­
disinfection)
which
covers
the
majority
of
surface
water
systems
in
the
U.
S.
[
1]
However,
this
commentor
urged
the
Agency
to
focus
its
decision­
making
criteria
on
drinking
water
at
the
intake
(
source
water)
and
consider
the
impact
of
pesticides
on
water
entering
drinking
water
supplies
and
private
wells.

Agency
Response:
The
Agency
is
in
the
process
of
gathering
information
on
the
extent
of
drinking
water
treatments
in
use
and
the
effectiveness
of
these
treatments
on
reducing
the
level
of
pesticides
in
water.
Consideration
of
water
treatment
effects
on
pesticides
in
water
must
take
into
account
not
only
effectiveness
in
pesticide
removal,
but
also
the
secondary
formation
of
any
transformation
products
of
toxicological
concern
as
a
result
of
the
treatment
process.
The
area
and
population
served
by
the
particular
treatments,
as
well
as
temporal
variations
in
drinking
water
treatment
effectiveness,
must
also
be
considered.
The
Agency
plans
to
issue
a
paper
addressing
drinking
water
treatment
effects
in
its
drinking
water
exposure
assessment
in
early
2000.

L.
Use
of
the
Drinking
Water
Level
of
Comparison
(
DWLOC)
in
Aggregate
Exposure
Assessments
1.
Consistency
in
DWLOC
and
MCL
Approaches
Two
commentors
[
1,
6]
felt
that
OPP's
DWLOC
was
not
consistent
with
the
Office
of
Water's
Maximum
Contaminant
Level
(
MCL)
and
the
Safe
Drinking
Water
Act
(
SDWA)
standard­
setting
process.
They
urged
OPP
to
establish
enforceable
standards
that
support
source
water
protection
and
drinking
water
program
compliance
with
pesticide
standards.
One
commentor
[
8],
noting
that
the
DWLOC
procedure
uses
the
Reference
dose
(
RfD)
as
the
maximum
daily
intake,
requested
that
the
document
clarify
that
this
is
a
conservative
approach
since
the
true
safe
level
is
some
value
above
the
RfD.
42
Agency
Response:
The
MCL
established
by
the
Office
of
Water
is
a
regulatory
standard
based
on
the
RfD
that
also
takes
into
consideration
such
factors
as
what
is
technologically
achievable
in
terms
of
contaminant
removal
and
the
costs
associated
with
such
removal.
The
DWLOC
does
not
factor
in
such
economic/
technology­
based
considerations.
Moreover,
in
practice
the
DWLOC
is
a
"
tool"
that
OPP
uses
in
its
human
health
aggregate
risk
assessment
process
to
identify
pesticides
which
require
further
evaluation
and
more
sophisticated
assessments.
The
DWLOC
is
an
expression
of
the
allowable
level
of
a
pesticide
in
drinking
water
which
when
combined
with
food
exposures
will
not
result
in
exposure
above
an
acceptable
level
of
risk.
OPP
strongly
supports
the
need
to
prevent
drinking
water
contamination
and
has
available
to
it
federal
authority
to
control
pesticide
use
and
application
methods
to
mitigate
contamination
of
water.

2.
Current
Approach
For
Determining
the
DWLOC
In
the
draft
document,
the
Agency
asked
for
comment
on
its
current
approach
of
calculating
DWLOCs
for
drinking
water
exposure
after
contributions
from
food
and
residential
sources
have
been
considered.
Responses
to
this
question
were
divided.
Two
commentors
[
6,
7]
called
the
approach
reasonable,
although
one
[
7]
said
the
process
is
only
as
valid
as
the
dietary
and
nondietary
assessments.
If
the
assessments
"
grossly
overestimate
risk,"
then
the
DWLOC
is
not
really
a
level
of
concern.
One
commentor
[
1]
saw
"
no
clear
logic
to
making
calculation
of
exposure
through
drinking
water
contingent
on
the
contributions
from
food
and
residential
exposure"
and
recommended
that
all
routes
of
exposure
should
be
considered
equally.
Another
commentor
[
3]
stated
that
total
aggregate
exposure
should
not
depend
on
the
order
in
which
the
potential
routes
of
exposure
are
considered.
This
commentor
expressed
concern
that
current
risk
evaluation
procedures
for
the
three
routes
of
exposure
are
"
at
vastly
different
levels
of
development"
and
encouraged
the
Agency
to
develop
more
sophisticated
risk
refinement
methods
for
all
important
routes
of
exposure.

Agency
Response:
The
DWLOC
was
originally
designed
as
a
tool
to
be
used
along
with
screening
level
estimates
of
pesticide
concentrations
in
drinking
water
to
determine
the
need
for
more
sophisticated
analyses
of
aggregate
exposure.
In
many
cases
the
comparison
of
the
DWLOC
to
the
screening
level
drinking
water
estimate
has
allowed
EPA
to
cost­
effectively
clear
the
pesticide
from
the
perspective
of
total
aggregate
risk.
Commentors
raise
a
very
legitimate
issue,
however,
with
regard
to
the
approach
that
the
Agency
takes
to
mitigation
or
additional
monitoring
requirements
in
those
cases
where
the
DWLOC
is
met
or
exceeded
and
further
refinements
in
the
drinking
water
estimate
of
exposure
still
do
not
allow
the
pesticide
to
be
cleared
from
an
aggregate
risk
perspective.
It
would
seem
that
a
regulatory
approach
which
would
have
as
its
objective
the
identification
of
the
most
cost­
effective
method
for
reducing
aggregate
risk
(
taking
into
consideration
all
pertinent
regulatory
authorities
 
SDWA
as
well
as
FIFRA
and
FFDCA
and
perhaps
even
relevant
State
and
local
regulatory
authorities)
would
be
the
preferable
approach.
43
List
of
Public
Commentors
and
Affiliations
I.
D.
#
Name
and
Organization
1.
John
H.
Sullivan,
Deputy
Executive
Director,
American
Water
Works
Association.
Denver,
Colorado.

2.
Arthur
L.
Craigmill,
Extension
Toxicity
Specialist,
Extension
Toxicity
Network
(
EXTOXNET),
University
of
California.
Davis,
California.

3.
FQPA
Implementation
Working
Group
(
IWG).
IWG
is
a
non­
governmental
coalition
of
farm,
food,
pest
management,
and
manufacturing
organizations.

4.
Sam
Moore,
President,
Kentucky
Farm
Bureau
Federation.
Louisville,
Kentucky.

5.
Jack
Laurie,
President,
Michigan
Farm
Bureau.
Lansing,
Michigan.

6.
Erik
D.
Olson,
Senior
Attorney,
Natural
Resources
Defense
Council
(
NRDC).
Washington,
D.
C.

7.
Dave
Whitacre,
Vice
President,
Development,
Novartis
Crop
Protection,
Inc.
Greensboro,
North
Carolina.

8.
Elaine
Francis,
Director,
Pesticides,
Toxics
and
Multimedia
Staff,
Office
of
Science
Policy,
Office
of
Research
and
Development,
U.
S.
Environmental
Protection
Agency.
Washington,
D.
C.
44
Sources
Cited
in
the
Agency
Responses
Habicht,
F.
H.
1992.
Guidance
on
Risk
Characterization
for
Risk
Managers
and
Risk
Assessors.
Memo
from
F.
Henry
Habicht
II,
Deputy
Administrator,
U.
S.
EPA,
to
Assistant
and
Regional
Administrators,
U.
S.
EPA,
Feb.
26,
1992.

Hackett,
A.
G.
1996,
1997,
1998.
Surface
Drinking
Water
Monitoring
Program
for
Acetochlor
and
Other
Corn
Herbicides.
Acetochlor
Registration
Partnership.
St.
Louis,
MO.
See
the
EPA
OPP
home
page
at
http://
www.
epa.
gov/
oppefed1/
aceto/
index.
htm
for
more
information.

National
Research
Council.
1993.
Ground
Water
Vulnerability
Assessment
:
Predicting
Relative
Contamination
Potential
under
Uncertainty
/
Committee
on
Techniques
for
Assessing
Ground
Water
Vulnerability,
Water
Science
and
Technology
Board,
Commission
on
Geosciences,
Environment,
and
Resources,
National
Research
Council,
1993.

Parker,
R.,
J.
Hetrick,
et
al.
1999.
Review
Comments
on
"
A
Simple
Regression
Model
for
Predicting
Surface
Water
Concentrations
Resulting
from
Agricultural
Field
Runoff
and
Erosion".
A
May
7,
1999,
memo
from
Parker,
Hetrick,
and
the
Water
Quality
Tech
Team,
through
Joe
Merenda,
EFED
Director,
to
Keefer,
Gilding,
and
the
ACPA
Drinking
Water
Exposure
Workgroup.

Scribner,
E.
A.,
D.
A.
Goolsby,
E.
M.
Thurman,
M.
T.
Meyer,
and
W.
A.
Battaglin.
1996.
Concentrations
of
Selected
Herbicides,
Herbicide
Metabolites,
and
Nutrients
in
Outflow
from
Selected
Midwestern
Reservoirs,
April
1992
Through
September
1993.
U.
S.
Geol.
Surv.
Open­
File
Report
96­
393.
Prepared
as
part
of
the
Toxic
Substances
Hydrology
Program.
Laurence,
KS.

U.
S.
EPA.
1992.
Guidelines
for
Exposure
Assessment.
FR
Vol.
57,
No.
104,
Fri,
May
29,
1992.

US.
EPA
FIFRA
Scientific
Advisory
Panel.
1997.
FIFRA
SAP
Report
on
A
Set
of
Scientific
Issues
Being
Considered
by
the
Agency
in
Connection
with
Estimating
Drinking
Water
Exposure
as
a
Component
of
the
Dietary
Risk
Assessment.
Electronic
copy
available
from
the
EPA
OPP
home
page
at
http://
www.
epa.
gov/
pesticides/
SAP/
1997/
december/
finaldec.
pdf
.

U.
S.
EPA
Office
of
Pesticide
Programs.
1998.
Proposed
Methods
for
Basin­
Scale
Estimation
of
Pesticide
Concentrations
in
Flowing
Water
and
Reservoirs
for
Tolerance
Reassessment.
Electronic
copy
available
from
the
EPA
OPP
home
page
at
http://
www.
epa.
gov/
pesticides/
SAP/
1998/
index.
htm#
october
.

U.
S.
EPA
Office
of
Pesticide
Programs.
1999.
Proposed
Methods
for
Determining
Watershedderived
Percent
Crop
Areas
and
Considerations
for
Applying
Crop
Area
Adjustments
to
Surface
Water
Screening
Models.
Electronic
copy
available
from
the
EPA
home
page
under
the
Office
of
Pesticide
Programs
(
OPP)
at
http://
www.
epa.
gov/
pesticides/
SAP/
1999/
may/
pca_
sap.
pdf
.
