FIFRA
Scientific
Advisory
Panel
Briefing
Document
for
a
Consultation
on:

Monitoring
Strategies
for
Pesticides
in
Surface
Derived
Drinking
Water
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
2
Table
of
Contents
Introduction
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4
Background
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5
Occurrence
of
Pesticides
in
Surface
Water
and
Relevance
to
Survey
Design
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5
Use
of
Drinking
Water
(
Surface)
Monitoring
Data
in
Risk
Assessment
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
6
Use
of
Drinking
Water
(
Surface)
Monitoring
Data
in
Risk
Management
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
7
Previous
or
On­
Going
Surface
Water
Monitoring
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
8
Design
Elements
of
Submitted
Monitoring
Studies
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
9
Submitted
Proposal
for
a
National
Surface
Water
Drinking
Water
Survey
for
Pesticides.
.
.
.
.
.
10
Critical
Design
Elements
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
11
Target
Population
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
11
Pesticide
Selection
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
11
Risk
Based
Selection
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
12
Pesticide
Use
and
Crop
Based
Selection
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
12
Pesticide
Usage
Based
Analyte
Selection
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
13
Environmental
Fate
and
Transport
Properties
Based
Analyte
Selection
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
13
Regulatory
Based
Analyte
Selection
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
13
Analytical
Methodology
Based
Selection
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
14
Combination
of
Several
Criteria
in
Analyte
Selection
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
14
Design
Framework
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
15
Domains
of
Interest
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
15
Pesticide
Use
Domains
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
15
Regional
Domains
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
16
Crop
Use
Area
Domains
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
16
Vulnerability
Domains
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
16
Stratification
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
16
Vulnerability
Strata
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
17
Regional
Strata
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
18
Site
Selection
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
18
ACPA
Site
Selection
Strategy
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
18
ARP­
Acetochlor
Site
Selection
Strategy
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
19
Watershed­
based
Site
Selection
Strategy
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
19
Site
Selection
Tools
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
19
Number
of
Sites
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
22
Number
of
Samples
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
24
Sample
Timing
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
26
Number
of
Samples
v.
Number
of
Sites
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
27
Raw
and
Finished
Water
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
27
Finished
Water
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
28
Raw
Water
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
28
Raw
and
Finished
Water
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
28
Duration
of
the
Study
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
29
One
Year
Study
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
30
One
Year
Study
With
Limited
Monitoring
Beyond
One
Year
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
30
Multiple
Years
of
Monitoring
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
30
Ancillary
Data
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
31
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
3
EFED's
Proposed
Monitoring
Design
Framework
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
31
Assumptions,
Constraints,
and
Biases
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
31
Design
Structure
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
32
Design
Efficiency
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
33
Sampling
Frequency
and
Timing
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
33
Study
Duration
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
33
Chemical
Selection
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
33
Raw
and
Finished
Water
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
34
Ancillary
Data
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
34
Census
for
Major
Facilities
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
34
List
of
Tables
Table
1.
Examples
of
surface
water
monitoring
studies
used
in
drinking
water
assessments.
.
.
.
.
.
.
.
.
.
.
.
9
List
of
Figures
Figure
1.
Sample
QA
map
for
verifying
CWS
location.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
20
Figure
2.
Upstream
and
surrounding
cataloging
units
for
delineated
basin.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
21
Figure
3.
Delineated
basin
boundary
within
an
individual
cataloging
unit.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
22
Figure
4.
Yearly
concentration
profiles
for
atrazine
and
bromoxynil
simulated
by
PRZM­
EXAMS.
.
.
.
24
Figure
5.
Variations
due
to
28­
day
sampling.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
25
Figure
6.
The
range
of
yearly
averages
possible
with
a
7­
day
sampling
interval
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
25
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
4
Introduction
The
Food
Quality
and
Protection
Act
(
FQPA)
of
1996
mandates
that
the
Office
of
Pesticide
Programs
(
OPP)
complete
aggregate
and
cumulative
human
health
risk
assessments
for
pesticides
with
food
tolerances.
Aggregate
risk
assessments
consider
all
routes
of
exposure,
including
drinking
water.
Cumulative
risk
assessments
consider
risks
for
multiple
pesticides
with
common
modes
of
action.
The
focus
for
this
Scientific
Advisory
Panel
(
SAP)
is
to
discuss
issues
associated
with
designing
a
survey
to
collect
pesticide
monitoring
data
for
drinking
water
derived
from
surface
water
for
use
in
chronic
aggregate
and
cumulative
human
health
risk
assessments,
and
risk
management
decisions.
Assessments
of
acute
exposure
will
not
be
considered
at
this
time.

The
Agency
uses
a
tiered
approach
for
assessing
pesticide
concentrations
in
drinking
water.
This
approach
includes
using
screening
models,
pesticide
occurrence
data
in
ambient
waters
[
e.
g.,
National
Water
Quality
Assessment
(
NAWQA)
Program],
and
when
available,
pesticide
occurrence
data
in
drinking
water.
Developing
a
systematic
methodology
for
monitoring
pesticide
concentrations
at
drinking
water
intakes
on
surface
water
systems
in
pesticide
use
areas
is
an
important
and
integral
part
of
this
approach,
especially
in
the
short
term.
For
many
reasons,
ranging
from
cost
considerations
to
practical
limitations,
OPP
believes
over
the
long
term
we
must
develop
and
validate
models
which
can
accurately
predict
pesticide
concentrations
at
drinking
water
intakes
on
surface
water
systems
in
pesticide
use
areas.
In
March
2000,
the
Environmental
Fate
and
Effects
Division
(
EFED)
and
the
United
States
Geological
Survey
(
USGS)
brought
two
regression
based
models
to
the
SAP.
The
March
SAP
was
very
favorable
in
its
review
and
comments
and
strongly
encouraged
EFED
to
continue
down
this
path.
Also,
in
August
1998,
the
Agency
presented
to
the
SAP,
physically
based
basin
scale
models
that
could
be
used
to
estimate
pesticide
concentrations
in
drinking
water.
Work
is
continuing
in
EFED
and
the
USGS
to
develop
these
models,
but
surface
water
monitoring
data
is
needed
to
develop,
test
and
effectively
use
them.
Thus,
it
is
EFED's
view
that
given
the
high
cost
of
collecting
drinking
water
monitoring
data,
surface
water
monitoring
data
collected
in
the
near
term
should
both
support
the
longer
term
goal
of
model
development
and
should
provide
sound
and
meaningful
data
which
can
be
used
today
in
aggregate
human
health
risk
assessments.

Approximately
two
years
ago,
OPP
initiated
an
effort
to
work
with
the
Office
of
Water,
other
federal
Agencies,
and
ACPA
to
develop
a
plan
to
collect
data
on
drinking
water
derived
from
surface
water
sources
nationally
for
use
in
human
dietary
pesticide
risk
assessments.
Last
year,
ACPA
presented
OPP
with
a
proposal
for
a
one­
year
survey
for
pesticide
occurrence
in
drinking
water
derived
from
surface
water
resources
addressing
acute
and
chronic
endpoints
for
an
approximate
cost
of
$
7
­
$
10
million.
Although
we
were
in
agreement
with
elements
of
the
chronic
design,
OPP
had
concerns
about
a
number
of
the
critical
elements
of
the
ACPA
proposal.
OPP
is
seeking
to
obtain
input
from
SAP
members
on
how
to
best
approach
the
design
of
a
multi­
pesticide
surface­
water
based
drinking
water
monitoring
study
whose
goal
is
to
define
for
each
targeted
pesticide,
the
distribution
of
annual
average
concentrations
in
surface
water
which
people
drink
across
the
pesticide's
use
areas
for
an
approximate
cost
of
$
7
­
$
10
million.
OPP
would
use
the
distribution
of
the
annual
average
concentrations
for
each
targeted
pesticide
in
the
aggregate
human
health
risk
assessment.
Further,
other
information
collected
during
the
survey
(
e.
g.
information
on
pesticide
use
and
watershed
characteristics
around
the
sampled
systems)
would
be
used
by
OPP
to
help
target
risk
mitigation
actions
to
those
areas
at
highest
risk.

For
this
SAP
consultation,
EFED
is
presenting
six
critical
design
elements,
including
the
elements
proposed
by
ACPA,
for
discussion.
In
addition,
EFED
is
offering
its
current
thoughts
on
what
it
views
as
a
preferred
approach
for
elements
of
this
survey.
Critical
design
issues
are:

1.
How
to
select
pesticides?
2.
How
to
approach
community
water
system
(
CWS)
site
selection?
1Larson
S.
J.,
P.
D.
Capel,
and
M.
S.
Majewski.
(
1997?)
Pesticides
in
Surface
Waters:
Distribution
Trends
and
Governing
Factors.
Ann
Arbor
Press,
Inc.
Chelsea
Michigan.
373
pp.

Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
5
a.
How
to
define
CWS
vulnerability
(
both
urban
and
agricultural)?
b.
Should
the
design
be
stratified
based
on
CWS
vulnerability?
3.
What
to
sample
 
raw
and/
or
finished
drinking
water?
4.
How
frequently
should
samples
be
taken
at
each
system?
Given
cost
constraints,
how
to
select
between
the
number
of
sites
and
the
number
of
samples?
What
role
should
seasonality
of
pesticide
use
play
in
determining
when
samples
are
collected?
5.
For
how
many
years
should
sampling
continue?
6.
What
type
of
ancillary
data
should
be
collected
on
watershed
characteristics
at
the
time
of
sampling?

These
six
design
issues
ultimately
influence
the
information
available
to
answer
risk
management
questions.
OPP
recognizes
there
is
a
great
deal
of
interplay
between
science,
policy,
and
economics
in
the
design
of
such
a
survey.
Decisions
on
key
survey
design
issues
define
the
nature
of
the
monitoring
program.
These
issues
are
influenced
by
the
needs
of
risk
managers,
which
include
answers
to
the
following
questions:

1.
Is
there
a
significant
number
of
people
and/
or
sub­
populations
(
e.
g.,
the
people
served
by
a
set
of
surface
water­
based
CWS's)
who
will
be
exposed
to
annual
average
concentrations
of
a
pesticide
in
drinking
water
at
levels
which
would
cause
a
significant
chronic
risk?
2.
Can
we
identify
and
mitigate
areas
of
high
risk
in
a
scientifically
defensible
manner.

Background
To
characterize
the
occurrence
of
and
exposure
to
pesticides
in
drinking
water,
the
Agency
uses
all
available
reliable
data;
initially
screening
individual
pesticides
and
gradually
refining
occurrence
and
exposure
assessments.
Refinements
use
more
sophisticated
models
and
incorporate
available
monitoring
data.
When
the
Agency
has
reason
to
believe
significant
exposure
could
occur,
but
uncertainties
exist
in
the
occurrence
estimates,
pesticide­
specific
monitoring
has
been
required
to
better
characterize
pesticide
occurrence.
This
section
provides
background
on
occurrence,
exposure,
and
risk
management
issues
specific
to
drinking
water
assessment
as
well
as
a
description
of
several
surface
water
monitoring
programs
which
have
been
conducted
or
proposed
to
provide
surface
water
monitoring
data
to
the
Agency.

Occurrence
of
Pesticides
in
Surface
Water
and
Relevance
to
Survey
Design
Pesticides
in
surface
water
tend
to
occur
in
pulses
that
can
last
from
days
to
weeks
to
months,
depending
on
the
type
of
water
body
and
the
pesticide1.
The
occurrence
of
most
pesticides
in
surface
water
from
agricultural
uses
is
seasonal,
and
is
related
to
the
timing
of
pesticide
applications,
rainfall
or
irrigation,
and
the
size
of
the
watershed.
The
predominant
source
of
agricultural
pesticide
residues
in
surface
water
is
related
spatially
to
where
the
pesticide
is
used1.
Thus,
the
occurrence
of
a
pesticide
in
drinking
water
at
a
CWS
above
a
level
of
concern
is
not
a
random
event
but
is
heavily
influenced
by
whether
that
pesticide
is
used
upstream
of
a
CWS
intake.
Seasonal
occurrence
patterns
of
urban­
use
pesticides
are
less
distinct
in
part
because
application
occurs
over
longer
periods
of
time
than
for
agricultural
use
pesticides1.

Since
the
water
a
person
drinks
typically
comes
from
one
or
two
sources,
a
monitoring
survey
must
be
capable
of
characterizing
the
pesticide
concentrations
to
which
an
individual
is
exposed
at
local
or
substantial.
It
also
needs
to
account
for
the
fact
that
if
a
pesticide
occurs,
it
will
probably
be
there
for
several
days
in
a
row.
In
order
for
the
Agency
to
sufficiently
characterize
pesticide
occurrence
in
drinking
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
6
water
derived
from
surface
water,
monitoring
data
must
be
able
to
estimate
the
following
parameters
for
sampled
CWS's
throughout
the
pesticide
use
area:
how
frequently
a
pesticide
occurs
at
an
individual
CWS;
if
it
is
likely
to
recur
at
the
CWS
at
the
same
time
each
year;
how
long
it
persists
at
the
CWS
when
it
occurs,
and
the
magnitude
of
the
pesticide
concentrations
that
occur.
To
fully
assess
the
impact
of
a
pesticide
on
drinking­
water
quality
we
also
need
to
characterize:
1)
how
the
parameters
described
above
vary
from
one
CWS
sampled
to
another
across
the
pesticide
use
area;
2)
uncertainty
in
extrapolating
from
one
sampled
CWS
to
other
CWS's
in
the
pesticide
use
area;
3)
and
the
error
in
extrapolating
from
the
time
period
sampled
to
other
times
(
for
example,
one
year
to
another).

Most
importantly,
any
survey
design
must
well
distinguish
situations
where
a
pesticide
may
be
of
concern
from
those
where
it
is
not.
This
is
critical
to
reasoned
risk
management
decisions.
A
good
risk
manager
seeks
always
to
reduce
risk
in
the
most
cost­
effective
manner,
by
strategically
tailoring
use
restrictions
or
limitations.
The
Agency
must
have
sufficient
confidence
in
the
monitoring
results
that,
given
no
detections
for
a
specific
pesticide,
we
can
reasonably
conclude
that
there
will
be
at
most
negligible
exposure,
rather
than
the
absence
of
detections
were
the
consequence
of
a
design
flaw.

Use
of
Drinking
Water
(
Surface)
Monitoring
Data
in
Risk
Assessment
Estimating
exposure
to
a
pesticide
in
drinking
water
entails
estimating
both
the
occurrence
of
a
pesticide
as
well
as
estimating
the
consumption
of
water
by
the
relevant
population.
The
Agency
needs
to
be
able
to
characterize
pesticide
exposure
(
occurrence
and
consumption)
in
drinking
water
at
an
appropriate
regional
scale
for
sensitive
sub­
populations
and
for
the
overall
population
in
that
region.
EFED
expects
the
results
of
this
monitoring
to
be
used
by
the
Health
Effects
Division
(
HED)
in
combination
with
drinking
water
consumption
estimates
to
develop
either
a
point
estimate
of
chronic
human
drinking
water
exposure
or
a
distribution
of
chronic
human
drinking
water
exposure
across
the
use
area
of
the
pesticide.
Either
of
these
would
be
combined
with
estimates
of
chronic
exposure
to
the
pesticide
in
food
and/
or
chronic
exposure
to
the
pesticide
from
use
in
or
around
the
home.

The
monitoring
should
provide
distributions
of
concentrations
at
individual
sampled
CWS's.
From
those
data,
one
can
also
calculate
an
annual
average
concentration
for
each
CWS
and
develop
a
distribution
of
annual
average
concentrations
across
sites.
The
full
distribution,
the
distribution
of
the
site
means,
or
alternatively
a
selected
point
estimate
(
for
example,
the
95th
percentile
site
in
the
distribution
of
annual
means
for
CWS's
in
a
region)
can
be
used
to
estimate
pesticide
occurrence
at
CWS's
within,
for
example,
the
pesticide's
use
area.

HED
is
currently
developing
tools
to
estimate
exposure
through
food
consumption
based
on
the
principle
that
exposure
occurs
to
an
individual
over
time.
Accounting
for
the
timing
of
pesticide
occurrence
in
drinking
water
is
important
to
consider
in
exposure
assessment
because
higher
concentrations
may
occur
simultaneously
with
increased
residential
exposure
as
all
these
routes
of
exposure
are
associated
with
the
application
of
the
pesticide.
The
survey
design
needs
to
provide
occurrence
data
at
the
appropriate
level
of
spatial
and
temporal
resolution
so
that
risks
from
water
could
be
appropriately
added
to
other
sources
of
dietary
risks
in
a
probabilistic
manner
(
if
desired)
as
well
as
to
other
sources
of
risk
in
an
aggregate
human
health
risk
assessment.

Currently
EFED
provides
HED
with
model­
based
annual
mean
concentrations
which
are
equaled
or
exceeded
every
10
yrs
at
a
site.
The
site
used
in
the
model
scenario
is
considered
to
be
an
upper
90th
percentile
use
site;
that
is,
exposure
at
the
use
site
is
expected
to
be
greater
than
exposure
at
90%
of
sites
used
to
grow
a
specific
crop.
There
are
no
confidence
bounds
on
the
predicted
concentrations.
Model
inputs
are
generally
chosen
to
be
representative
of
mean
values
for
sites
being
modeled.
Chemical
specific
inputs
(
e.
g.
aerobic
soil
half­
life)
are
generally
chosen
to
represent
90th
percentile
values.
The
cumulative
result
of
model
site
selection
and
chemical
inputs
is
intended
to
predict
concentrations
at
the
higher
range
of
the
distribution
concentrations
likely
to
occur
in
surface
water.
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
7
The
standard
for
pesticide
concentration
data
quality
has
four
elements:
(
1)
the
summary
statistic
used
to
describe
the
exposure,
(
2)
the
fraction
of
the
population
protected,
(
3)
the
frequency
at
which
an
adverse
concentration
will
occur,
and
(
4)
the
confidence
in
the
accuracy
of
the
assessment.

1.
For
chronic
human
health
risk,
OPP
is
currently
using
an
annual
mean
as
the
statistic
to
represent
exposure
in
deterministic
risk
assessments.
The
mean
over
time
at
a
site
is
better
approximated
using
time­
weighting
for
the
estimate
rather
than
a
straight
arithmetic
mean.
This
is
generally
necessary
because
samples
are
not
usually
collected
at
equal
intervals
throughout
the
year,
but
concentrated
when
they
are
most
likely
to
occur.
Time­
weighting
accounts
for
the
bias
generated
by
these
sampling
methods.

2.
HED
uses
the
average
dietary
consumption
(
foods
and
water)
of
10,000
individuals
over
a
3­
day
period
along
with
pesticide
occurrence
on
individual
food
items
to
estimate
chronic
exposure.
Food
residues
are
typically
estimated
using
either
monitoring
data
or
residue
field
trials.
The
exposure
through
all
dietary
sources
(
all
foods
and
water)
will
be
summed
to
estimate
aggregate
dietary
exposure.
The
dietary
exposure
is
then
compared
to
a
toxic
reference
point
(
e.
g.
%
RfD)
and
the
Agencies
determination
wether
the
risk
is
acceptable
or
if
mitigation
is
needed.

To
provide
HED
with
data
on
pesticide
exposure
through
drinking
water,
EFED
is
proposing
to
identify
95th
percentile
CWS's
(
the
facilities
at
which
the
annual
average
concentration
is
at
the
95th
percentile),
rather
than
the
90th
percentile
site
used
in
modeling.
Because
of
the
uneven
distribution
of
population
among
water
supplies,
and
because
small
water
supplies
tend
to
be
more
vulnerable
to
pesticide
contamination,
this
will
protect
a
substantially
greater
portion
of
the
population.
Population
weighting
can
be
used
to
convert
the
design
from
a
CWS
basis
to
a
population
basis.

3.
Since
surface
water
concentrations
vary
over
time
it
is
necessary
to
identify
a
specific
frequency
at
which
a
level
will
occur
as
a
standard
to
characterize
exposure
at
a
given
site.
The
standard
for
return
frequency,
or
the
frequency
at
which
a
level
will
occur,
for
modeled
estimates
of
pesticide
concentrations
in
surface
water
is
once
in
ten
years.
In
chronic
dietary
exposure
assessments,
average
annual
food
residues
may
change
due
to
use/
usage
changes
resulting
from
climate,
pest
pressure,
or
market
forces.
Currently
HED
uses
food
residues
averaged
over
five
years.
In
the
near
future,
HED
intends
to
identify
durations
of
exposure
(
six
months
to
one
year
for
chronic
exposure),
exceeding
a
toxic
endpoint,
that
occur
over
a
five
year
period.
Food
residue
monitoring
data
collect
within
five
years
is
used
in
exposure
assessments.

4.
There
is
no
current
Agency
standard
for
data
confidence.
However,
a
standard
is
needed
to
design
a
statistically­
based
survey.
EFED
is
proposing
to
use
a
95%
confidence
value
as
a
standard
for
the
study.

Use
of
Drinking
Water
(
Surface)
Monitoring
Data
in
Risk
Management
EFED
expects
the
results
of
this
monitoring
survey
to
be
used
by
risk
managers
to
target
drinking
water
related
risk
mitigation
actions
in
those
areas
posing
unacceptably
high
risk.
The
local
nature
of
drinking
water
exposure
allows
for
geographically
targeted
risk
management
options.
It
is
important
to
understand,
once
geographical
areas
have
been
identified,
the
relationship
between
the
timing
of
pesticide
application
and
occurrences
of
high
concentrations.
More
information
regarding
the
cause
of
the
high
concentrations
will
permit
risk
managers
to
impose
the
most
tailored
mitigation
measures
with
the
least
impact
to
growers.
Without
this
information,
mitigation
measures
will
be
less
discriminating.
For
example,
if
monitoring
data
shows
that
high
concentrations
in
drinking
water
for
a
pesticide
are
associated
with
2
We
are
defining
validation
in
this
case
as
"
A
comparison
of
measured
values
with
values
calculated
by
some
model
in
order
to
determine
the
level
agreement
and
circumstances
or
causes
where
there
is
substantial
lack
of
agreement.

Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
8
applications
during
the
non­
growing
season
in
northern
California
(
a
dormant
application),
risk
mitigation
could
be
focused
on
this
specific
practice
in
this
location.
Without
these
data,
the
use
of
the
pesticide
may
be
canceled
on
certain
crops,
not
allowed
in
the
western
part
of
the
US,
or
the
dormant
application
of
the
pesticide
may
not
be
allowed
nationwide.

Previous
or
On­
Going
Surface
Water
Monitoring
For
previous
studies,
the
Agency
has
looked
at
eight
data
objectives:
four
assessment
and
four
developmental.
The
assessment
objectives
are:

1.
Drinking
water
exposure
assessment
 
estimate
the
frequency,
magnitude
and
duration
of
exposure
in
drinking
water
across
the
national
population
and
identifiable
sub­
populations.
This
objective
is
meant
to
support
the
human
health
risk
assessments.

2.
Aquatic
exposure
assessment
 
estimate
the
frequency,
magnitude
and
duration
of
exposure
to
nontarget
life
in
aquatic
ecosystems.
This
objective
supports
the
ecological
risk
assessments.

3.
Water
quality
assessment
 
determine
the
magnitude
and
frequency
of
the
occurrence
of
a
pesticide
in
water
resources.
This
objective
supports
the
Agencies
pollution
prevention
goals
and
is
not
keyed
to
any
particular
identifiable
risk.

4.
Fate
and
transport
assessment
 
determine
the
relative
importance
of
the
different
routes
and
mechanisms
of
transport
to
and
through
a
watershed
or
aquifer.
This
objective
provides
understanding
of
the
pesticide's
behavior
in
the
environment
and
aids
in
making
informed
and
appropriate
risk
management
decisions
for
the
pesticide.

The
development
objectives
are:
1.
The
development
of
regression
based,
or
empirical,
models
for
describing
the
occurrence
of
pesticides
in
water
resources.
A
sound
tool
of
this
nature
would
aid
in
differentiating
pesticides
requiring
further
refinement
of
drinking
water
concentrations
from
those
that
do
not.
The
Agency
is
currently
working
with
the
USGS
to
develop
models
of
this
nature.

2.
The
evaluation
and
validation2
of
physically­
based
models.
The
Agency
may
also
use
physically
based
basin
scale
models
for
risk
assessment.
These
models
have
different
capabilities
from
empirical
models
and
are
particularly
useful
for
evaluating
potential
risk
mitigation
options.
Physical
models
must
be
validated
against
measured
field
data
in
order
to
verify
that
the
model
does
in
fact
reflect
reality.
Successful
validation
exercises
should
build
the
confidence
necessary
to
use
these
tools
in
a
regulatory
setting.

3.
Development
and
testing
of
a
vulnerability
assessment
tool
or
index
for
water
resources.
The
ability
to
determine
relative
likelihood
that
a
site
will
have
frequent
and/
or
high
magnitude
occurrence
of
a
pesticide
is
a
critical
need
for
the
OPP.
A
site
vulnerability
index
would
be
valuable
for
many
purposes
including,
designing
future
monitoring
studies,
putting
data
from
older
studies
into
context,
identifying
vulnerable
sub­
populations
and
using
data
from
studies
of
that
were
not
done
at
drinking
water
sites
and
using
it
to
assess
drinking
water
exposure.
This
subject
is
discussed
in
substantially
more
detail
in
the
site
selection
process
in
the
body
of
this
document.
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
9
4.
Development
of
improved
and
more
powerful
statistical
methods
for
the
interpretation
of
monitoring
studies.
An
improved
understanding
of
the
structure
and
relationships
that
underlay
pesticide
concentration
data
will
allow
the
development
of
improved
tools
for
analyzing
monitoring
studies.
These
improved
tools
will
allow
regulatory
decisions
to
made
with
increased
confidence,
or
with
less
underlying
data
in
the
future.

OPP
has
in
the
past
required
surface
water
monitoring
programs
for
specific
pesticides
as
a
condition
of
registration
or
re­
registration.
In
addition,
the
Agency
requires
compliance
monitoring
for
16
currently
registered
pesticides
under
the
Safe
Drinking
Water
Act
(
SDWA)
at
community
water
systems
(
CWS).
A
key
source
of
surface
water
monitoring
data
is
provided
to
OPP
by
USGS
National
Water
Quality
Assessment
Program
(
NAWQA).
Overall,
there
has
been
limited
drinking
water
monitoring
of
surface
water,
and
the
scope
and
nature
of
all
monitoring
program
designs
has
been
variable.

Design
Elements
of
Submitted
Monitoring
Studies
Table
1
provides
a
summary
of
several
surface
water
monitoring
programs.
A
review
of
survey
design
elements
reveals
that
there
is
a
great
deal
of
variety
in
approaches
to
survey
design.
The
variety
in
design
frameworks
is
influenced
by
the
purpose,
and
the
nature
of
the
pesticides
in
the
study.
For
example,
the
NAWQA
study
is
designed
to
characterize
water
quality
across
the
nation.
Pesticides
considered
in
the
NAWQA
survey
were
selected
to
represent
those
with
greater
use.
Only
ambient
water
from
rivers
and
streams
was
sampled.
In
contrast,
the
purpose
of
the
Unregulated
Contaminant
Monitoring
Regulation
(
UCMR)
is
to
provide
data
to
further
refine
drinking
water
regulations
as
required
under
the
SDWA.
Pesticides
in
the
UCMR
appear
on
the
Unregulated
Contaminant
List.
Finished
drinking
water
is
only
sampled
in
the
UCMR.
Other
monitoring
studies
have
been
designed
to
focus
on
apparently
highly
vulnerable
sites.
These
studies
are
generally
conducted
for
a
single
pesticide
and
can
be
used
for
screening
purposes
to
place
other
monitoring
data
into
context.
None
of
the
study
designs
provide
a
complete
design
framework
for
a
national­
scale
drinking
water
monitoring
program.

Table
1.
Examples
of
surface
water
monitoring
studies
used
in
drinking
water
assessments.

Study
Identification
Crops
Type
of
Water
Study
Duration
Sample
Location
Watershed
Vulnerability
Considerations
Number
of
Sites
Sample
Frequency
Acetochlor
Registration
Partnership
corn
Finished
with
raw
samples
on
sites
with
activated
carbon
7
years
CWS
Corn
acreage
in
the
watershed
surface
water
body
source
175
14
samples
per
year
with
biweekly
sampling
during
use
period
Atrazine
corn
Finished
6
years
CWS
None
5044
Quarterly
Samples
Bromoxynil
corn,
small
grains,
cotton
Finished
Raw
1
year
CWS
Watershed
size
Pesticide
Use
Atrazine
Detections
16
Pre­
and
Post
treatment,
two
intervals
during
the
growing
season
and
monthly
interval
thereafter
Table
1.
Examples
of
surface
water
monitoring
studies
used
in
drinking
water
assessments.

Study
Identification
Crops
Type
of
Water
Study
Duration
Sample
Location
Watershed
Vulnerability
Considerations
Number
of
Sites
Sample
Frequency
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
10
Fipronil
corn
Finished
Raw
1
year
CWS
Watershed
Six
Pesticide
Use
Atrazine
Detections
13
Pre­
and
Post
treatment,
two
intervals
during
the
growing
season
and
monthly
interval
thereafter
Carbaryl
corn
&
non­
ag
uses
Finished
Raw
1
year
CWS
Watershed
Size
Pesticide
Use
Atrazine
Detections
20
Pre­
and
Post
treatment,
two
intervals
during
the
growing
season
and
monthly
interval
thereafter
NAWQA
Agricultu
ral
land,
Forrest,
&
urban
areas
Ambient
1991­
Ambient
Targeted
Land
uses.
-
60
intensive
sites
Variable­
depends
on
site
designation
Pilot
Reservoir
Monitoring
Cotton
Corn
Peaches
Non­
Ag
Finished
Raw
1
year
CWS
Pesticide
Use
Reservoir
Size
Reservoir
Location
12
reservoirs
Variable
Unregulated
Contaminant
Monitoring
Regulation
None
Finished
1
year
every
3
years
CWS
None
-
6400
Quarterly
Samples
Submitted
Proposal
for
a
National
Surface
Water
Drinking
Water
Survey
for
Pesticides.

The
American
Crop
Protection
Association
(
ACPA)
has
proposed
a
national­
scale
design
framework
for
assessing
daily
peak
and
annual
mean
pesticide
concentrations
in
surface
water
used
for
drinking
water.
Based
on
EFED's
review
of
the
ACPA
submission
the
following
are
the
essential
features
of
the
chronic
survey
component
of
the
monitoring
strategy
proposed:

°
ACPA's
site
selection
process
proposes
a
national
drinking
water
survey
based
upon
five
geographically
based
analysis
domains:
four
food
regions
and
the
coterminous
United
States.
An
additional
five
domains
will
represent
"
vulnerable"
CWS's
within
these
geographic
areas.
°
The
counties
in
the
contiguous
48
States
comprising
the
top
quartile
of
pesticide
use
intensity
for
any
one
of
25
target
pesticides
are
identified
as
a
"
high
pesticide
use
intensity
area".
When
a
CWS
with
a
"
small
or
medium"
sized
watershed
occurs
in
the
"
high
pesticide
use
intensity
area"
it
is
defined
as
a
"
vulnerable
CWS".
To
identify
CWS
that
are
vulnerable
to
chronic
exposure
ACPA
will
select
"
CWS
using
reservoirs
and
having
a
higher
use
of
products
with
chronic
exposure
issues".
°
Monitor
finished
drinking
water
for
one
year,
with
triggers
for
extension
if
"
regional
pesticide
use
is
too
low
to
meet
FQPA
requirements".
°
Use
an
optimized
sampling
strategy
based
on
stratified,
random
sampling.
For
chronic
exposure
parameters,
water
samples
will
be
collected
systematically
throughout
the
year
at
the
required
number
of
CWS
in
each
domain.
Strata
will
"
be
based
on
factors
thought
to
be
related
to
overall
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
11
CWS
vulnerability:
reservoirs
vs
flowing
water,
pesticide
use
intensity,
population
served,
geographic
diversity,
etc."
°
Collect
samples
from
each
CWS
an
average
of
10
times
in
a
year
depending
on
the
type
of
water
body.
Sampling
frequency
will
be
adjusted
based
on
the
hydrologic
nature
of
the
water
source,
with
less
frequent
sampling
from
reservoirs
and
lakes.
Possibly
composite
samples
seasonally
to
reduce
to
4
"
seasonal"
samples
per
site.
°
Analyze
each
sample
for
all
118
pesticides
on
the
USGS
multi­
analyte
schedules
°
Develop
a
distribution
of
seasonal
and
annualized
mean
concentrations
at
each
CWS
for
all
118
pesticides,
including
non­
targeted
compounds
"
with
a
maximum
value
to
exceed
the
95th
percentile
of
CWS­
years".
°
Use
seasonal
and
annual
average
concentrations
to
represent
regional
and
national
average
concentrations
for
"
vulnerable"
CWS
and
for
all
CWS.

OPP
believes
the
ACPA
design
framework
for
assessing
annual
mean
concentrations
could
be
an
acceptable
survey
design
with
some
modifications,
including
the
site
selection
process,
sampling
frequency,
vulnerability
classification
system,
sampling
raw
and
finished
water,
selection
of
priority
group
of
pesticides,
and
study
duration.

Critical
Design
Elements
Target
Population
OPP
is
defining
the
target
population
for
the
survey
as
CWS's
within
the
use
areas
of
the
selected
pesticides.
The
population
served
by
each
CWS
will
be
known
and
can
be
used
to
convert
the
data
from
a
CWS
basis
to
a
population
basis.
By
defining
the
target
population
in
this
manner,
it
is
assumed
runoff
is
the
predominant
route
through
which
pesticides
contaminate
drinking
water.
Contamination
resulting
from
routes
other
than
runoff
is
expected
to
result
in
lower
concentrations
that
are
unlikely
to
exceed
levels
of
concern.

Ultimately
the
Agency
is
interested
in
the
human
population
consuming
surface­
derived
drinking
water
and
link
these
exposures
to
dietary
assessments.
Dietary
(
food­
based)
exposure
assessments
assume
the
food
supply
of
the
U.
S.
is
evenly
distributed.
Drinking
water,
unlike
food,
is
supplied
from
local
sources
 
it
is
not
nationally
distributed.
Surface
source
drinking
water
is
generally
processed
at
local
public
water
facilities.
Thus,
pesticide
exposure
needs
to
be
evaluated
on
sub­
populations
of
humans
in
the
U.
S.
deriving
drinking
water
originating
in
a
watershed
where
a
pesticide
is
used
(
i.
e.
a
CWS).
Because
each
CWS
serves
multiple
households,
targeted
monitoring
of
public
water
facilities
is
expected
to
be
the
most
efficient
approach
in
a
monitoring
study.
Monitoring
of
in­
house
tap
water
would
provide
the
most
refined
exposure
concentrations
of
pesticides
in
drinking
water.
However,
a
study
of
tap
water
does
not
allow
the
Agency
to
meet
its
other
regulatory
objectives
(
e.
g.
model
validation
or
information
for
risk
management
decisions).
Further,
tap
water
monitoring
may
be
cost
prohibitive
given
the
variability
in
distribution
systems
and
networks.

ACPA
defines
the
target
population
of
there
study
as
CWS's
obtaining
some
or
all
of
there
source
water
from
surface
water
sources.

Pesticide
Selection
Many
strategies
can
be
employed
to
select
pesticides
as
analytes
in
a
national
survey
of
pesticides
in
drinking
water
and
each
strategy
has
advantages
and
disadvantages.
It
is
critical
to
be
aware
that
there
is
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
12
an
intimate
link
between
site
selection,
chemical
selection,
sampling
frequency,
and
the
cost
constraints
of
the
survey.
Although
one
can
make
an
independent
decision
about
which
chemicals
are
most
important
to
target
and
analyze
in
such
a
survey,
the
framework
of
the
design
and
characteristics
of
pesticide
use
in
the
proximity
of
a
water
treatment
facility
during
the
survey
will
largely
determine
how
well
survey
data
represent
the
water
quality
impacts
of
each
individual
pesticide.
It
is
reasonable
to
assume
that
national
survey
results
will
best
represent
the
impact
on
surface­
water
quality
from
pesticides
that
are
widely
used
throughout
the
country,
rather
than
from
pesticides
used
on
speciality
crops
or
pesticides
with
major
nonagricultural
uses,
whose
actual
use
areas
are
poorly
defined.
The
methodology
used
to
select
target
pesticides
as
analytes
must
be
taken
into
account
in
determining
the
process
to
select
appropriate
monitoring
locations.

The
analysis
of
water
samples
for
pesticide
residues
is
the
major
factor
in
the
cost
of
a
monitoring
survey.
When
costs
are
constrained,
as
they
are
in
this
survey,
one
way
to
maximize
the
information
obtained
and
minimize
analytical
costs
is
to
use
multi­
analyte
methods,
which
provide
concentrations
for
a
large
number
of
pesticides
in
one
analysis.
A
downside
of
multi­
analyte
methods
is
that
the
data
obtained
may
not
be
equally
meaningful
for
all
analytes.
For
example,
if
a
pesticide
is
not
used
near
any
sampling
locations,
a
multi­
analyte
method
will
still
provide
data.
However,
one
cannot
conclude
that
these
survey
results
for
that
pesticide
represent
the
impact
on
water
quality
in
areas
where
that
particular
pesticide
is
actually
used.

Possible
chemical
selection
strategies
based
on
risk,
use
site
(
crops),
usage,
environmental
fate
properties,
anticipated
regulatory
decisions,
analytical
methodologies,
or
a
combination
of
these
criteria
are
described
below.
In
the
end
risk
managers
play
a
major
role
in
selecting
chemicals
for
this
survey.

Risk
Based
Selection
Pesticides
with
relatively
higher
risks
can
be
targeted
as
analytes
based
on
toxicity
to
humans
and
occurrence
in
drinking
water.
Toxicity
and
risk
determinations
can
be
based
on
studies
reviewed
by
the
Agency,
and
are
typically
summarized
in
Re­
registration
Eligibility
Decision
(
RED)
documents
and
in
emergency
exemptions
(
Section
18)
and
Section
3
science
chapters.

The
advantage
of
selecting
pesticides
based
upon
risk
is
that
one
can
ensure
that
pesticides
with
potentially
high
risks
(
and
which
need
refined
estimates)
will
be
included
in
the
national
drinking
water
survey
and
that
data
will
be
available
to
the
Agency
for
use
in
risk
assessment
and
risk
management.
One
possible
drawback
of
using
this
factor
alone
in
selecting
target
analytes
is
that
new
data
can
change
the
Agency's
conclusions
about
toxicity,
or
provide
information
about
new
areas
of
risk
(
e.
g.
endocrine
disruption
effects)
about
which
the
Agency
is
unaware
at
the
time
the
survey
is
designed.

Pesticide
Use
and
Crop
Based
Selection
Pesticides
can
be
targeted
as
analytes
for
the
survey
based
upon
the
crops
that
they
are
used
on.
For
example,
the
herbicides,
insecticides,
fungicides,
and
fumigants
used
on
cotton,
corn
and
soybeans,
and
apples
could
be
selected
a
s
survey
analytes.
Good
data
on
land
use
and
pesticide
use
will
be
needed
for
this
approach,
as
crop
use
areas
will
determine
the
geographic
scope
of
the
monitoring.
An
advantage
of
this
approach
is
that,
like
the
Pesticide
Data
Program
(
PDP)
which
supplies
pesticide
residue
data
on
a
crop
specific
basis
for
dietary
assessments,
these
data
clearly
identify
impacts
of
a
specific
pesticide
use
on
drinking
water
quality.
Also,
the
"
national"
scope
of
the
program
would
be
geographically
defined
as
the
pesticide
use
areas
of
these
crops.
Additionally,
risk
managers
will
have
high
quality
comparative
data
about
the
relative
impacts
of
alternative
pesticides
for
a
certain
use
on
drinking
water
quality.
One
disadvantage
of
this
approach
is
that,
like
PDP
data,
targeted
data
will
not
be
gathered
for
all
uses
of
a
pesticide,
and
additional
monitoring
may
be
needed
on
an
on­
going
basis
focusing
on
other
pesticide
uses.
3Goss,
D.
W.
1992.
Screening
Procedure
for
Soils
and
Pesticides
for
Potential
Water
Quality
Impacts.
Weed
Technol.
6(
3):
701­
708.

4Hornsby,
A.
G.
1992.
Site­
Specific
Pesticide
Recommendations:
The
Final
Step
in
Environmental
Impact
Prevention.
Weed
Technol.
6(
3);
736­
742.

Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
13
Pesticide
Usage
Based
Analyte
Selection
Pesticides
can
be
targeted
as
analytes
for
the
survey
based
on
their
overall
national
usage,
high
application
rates,
or
on
use
intensity.
EPA's
National
Survey
of
Pesticides
in
Drinking
Water
Wells
selected
analytes
based
in
part
on
this
criterion
(
at
least
one
million
pounds
or
more
of
use
nationally
in
a
specific
year).
An
advantage
of
this
approach,
since
pesticide
usage
or
use
intensity
is
strongly
correlated
with
occurrence
in
surface
water.
Also,
model
predictions
are
directly
correlated
with
pesticide
application
rate.
Thus,
there
would
be
a
greater
likelihood
that
analytes
identified
based
on
these
criteria
would
be
found
in
such
a
survey.
A
disadvantage
is
that
some
compounds
would
not
be
included
if
they
are
not
heavily
used,
yet
these
compounds
could
have
an
impact
on
water
quality
in
local
areas.

Environmental
Fate
and
Transport
Properties
Based
Analyte
Selection
Environmental
fate
and
transport
properties
of
a
pesticide
determine
in
part
if
there
is
the
potential
for
that
pesticide
to
be
transported
via
surface
runoff
to
surface
water
which
may
be
a
source
of
drinking
water.
Pesticide
environmental
fate
properties
affecting
runoff
potential
include:
water
solubility,
acid­
base
dissociation
constants,
ionic
properties,
soil/
sediment
sorption
coefficients,
and
environmental
stability
or
persistence.
Some
examples
of
screening
procedures
to
determine
if
a
pesticide
is
not
likely
to
occur
in
runoff
are
the
methods
of
Goss3,
Hornsby4
and
the
FIRST
and
PRZM
models.
In
general,
highly
water
soluble
pesticides
tend
to
stay
in
the
aqueous
phase,
and
can
have
a
greater
potential
for
runoff.
Ionic
chemicals
and
pesticides
that
exhibit
acid­
base
characteristics
may
be
more
prone
to
offsite
transport
in
the
dissolve
phase,
which
is
typical
when
the
solution
pH
enhances
the
formation
of
anionic
species.
For
example,
DCPA
dimethyl
ester
is
immobile;
however,
when
hydrolyzed
to
mono­
and
di­
acid
species
it
is
quite
mobile
and
easily
transported
in
runoff.
Pesticides
with
lower
sorption
coefficients
are
generally
more
prone
to
runoff
in
the
dissolved
phase
than
those
with
higher
sorption
values.
Another
factor
affecting
runoff
potential
is
persistence,
which
can
be
predicted
from
aerobic
and
aquatic
metabolism
half­
lives.
Pesticides
which
are
more
persistent
are
generally
more
available
to
be
transported
in
runoff.

One
important
advantage
of
using
this
criterion
is
that
chemicals
that
are
persistent
and
mobile
and
therefore
have
a
greater
likelihood
of
being
detected
in
surface
water
will
be
selected,
if
they
are
not
removed
or
transformed
during
water
treatment
processes.
Another
advantage
is
that
if
one
can
identify
a
sub­
group
of
pesticides
which
have
common
or
comparable
fate
and
transport
properties
and
select
one
compound
to
represent
each
sub­
group,
survey
results
could
be
applied
to
pesticides
which
were
not
analytes
but
which
are
members
of
that
sub­
group.
A
disadvantage
of
this
approach
is
that
there
may
not
be
adequate
fate
data
available
to
draw
conclusions
about
all
pesticides,
degradation
products,
or
transformation
products.

Regulatory
Based
Analyte
Selection
The
Office
of
Pesticide
Programs
is
operating
under
congressionally
mandated
deadlines
to
reassess
tolerances
for
all
pesticides.
The
Food
Quality
Protection
Act
(
FQPA)
sets
a
timetable
for
tolerance
reassessment.
Based
on
this
criterion,
one
could
select
the
next
group
of
pesticides
for
which
monitoring
data
is
needed
for
tolerance
reassessment
decision
making.
Another
option
is
to
use
the
pesticides
on
the
Office
of
Water's
Unregulated
Contaminant
Candidate
list
(
UCCL)
(
established
by
the
SDWA)
as
analytes
in
the
survey.
The
Agency
will
consider
developing
drinking
water
standards
(
maximum
contaminant
levels)
for
these
pesticides
based
in
part
on
monitoring
data.
An
advantage
of
using
this
selection
strategy
is
that
when
regulatory
decisions
are
made
the
best
scientific
data
will
be
available.
A
disadvantage
is
that
risk
management
decisions
could
be
made
in
the
near­
term
that
change
target
pesticide
uses,
use
areas
and
5Larson,
S.
J.,
R.
J.
Gilliom,
and
P.
D.
Capel.
1999.
Pesticides
in
Streams
of
the
United
states
 
Initial
Results
from
the
National
Water­
Quality
Assessment
Program.
U.
S.
Geological
Survey.
Water­
Resources
Investigations
Report
98­
4222.
Sacramento,
California.

6American
Crop
Protection
Association
(
ACPA).
1999.
A
National
Survey
of
Pesticides
in
Drinking
Water
Obtained
from
Surface
Sources.
Paper
submitted
to
OPP
EFED.

Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
14
application
rates
that
would
make
the
survey
results
inapplicable.

Analytical
Methodology
Based
Selection
Using
this
criterion,
pesticides
and
degradation
products
with
validated
analytical
methodologies
of
acceptable
precision
and
accuracy
would
be
included
analytes.
The
USGS
NAWQA
program
has
established
multi­
analyte
analytical
methods
with
very
low
limits
of
detection
for
a
large
number
of
pesticides
(
118)
5;
however
some
of
these
methods
have
low
analytical
recoveries.
The
major
advantage
of
this
selection
strategy
is
that
several
analytical
methods
already
exist,
and
if
these
methods
are
determined
to
be
acceptable,
analytical
method
development
will
not
affect
the
time
required
to
implement
the
survey.
A
disadvantage
is
that
certain
pesticides,
pesticide
degradates,
or
pesticide
transformation
products
which
are
toxic
may
not
be
included
because
analytical
methods
are
not
available.
Also,
multi­
analytes
methods
may
not
provide
data
of
equivalent
quality
for
all
compounds
or
if
compounds
are
not
used,
because
the
method
is
not
optimized
for
all
compounds.
Detections
of
analytes
for
which
methods
are
not
optimized
may
in
some
instances
be
of
some
use
in
defining
lower
bounds
of
exposure
in
FQPA
drinking
water
assessments
on
a
case
by
case
basis.

The
ACPA
proposal
for
a
national
survey
adopted
this
approach
for
analyte
selection.
Analytes
were
identified
for
which
there
is
a
validated
NAWQA
analytical
method
(
identified
as
NAWQA
schedules:
2001,
2002,
or
2060).
About
118
analytes
would
be
analyzed
using
this
approach.
Data
would
only
be
reported
for
those
compounds
whose
registrants
agree
to
participate
in
the
Survey6.

Combination
of
Several
Criteria
in
Analyte
Selection
A
combination
of
the
criteria
listed
above
can
be
used
to
rank
potential
analytes
from
which
a
target
list
can
be
selected.
This
combination
would
take
into
account
risk
or
toxicity
concerns,
use
intensity
and
rate,
surface
water
runoff
potential,
and
the
availability
of
acceptable
analytical
methodologies.
For
each
pesticide
one
would
consider
and
weight
these
factors
separately
(
e.
g,
large,
medium,
and
small
use).
The
cumulative
or
total
score
is
used
in
the
relative
ranking
and
final
selection
of
the
target
analytes.

Using
this
weighting,
ranking,
and
selection
method
has
the
advantage
that
a
number
of
different
key
factors
can
be
considered
simultaneously
in
selection
of
analytes.
For
example,
including
the
criteria
described
in
the
example
above,
selected
analytes
would
be
toxic
to
humans,
be
more
likely
to
occur
in
surface
waters,
and
have
available
validated
analytical
methods
of
acceptable
sensitivities.
If
different
criteria
are
included,
analytes
would
be
selected
that
reflect
those
criteria.
For
example
if
analytical
methodology,
use
site,
and
environmental
fate
and
transport
criteria
are
chosen,
selected
analytes
would
have
existing
analytical
methods,
be
used
on
specified
crops,
and
be
more
likely
to
occur
in
surface
waters.
One
possible
disadvantage
of
using
multiple
factors
in
analyte
selection
is
the
relatively
longer
time
needed
to
collect
all
the
required
information
before
the
ranking
and
final
selection
of
the
target
analytes
can
be
completed.
Other
disadvantages
are
described
above
under
individual
factors.

Design
Framework
The
sampling
strategy
framework
is
the
foundation
of
a
monitoring
program.
There
are
two
sampling
approaches,
probabilistic
and
non­
probabilistic.
Probabilistic
methods
use
random
sampling,
but
may
divide
the
sample
space
into
predefined
subsets.
Non­
probabilistic
methods
such
as
haphazard
sampling
and
judgement
sampling
do
not
sample
randomly
and
thus
require
a
homogeneous
population
over
time
and
7(
Gilbert,
1987)

Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
15
space
for
an
unbiased
estimate.

Probabilistic
sampling
strategies
include
simple
random
sampling,
stratified
random
sampling,
twostage
sampling,
cluster
sampling,
systematic
sampling,
and
double
sampling7.
Each
of
these
sampling
strategies
are
useful
to
optimize
the
design
efficiency
for
addressing
spatial
and
temporal
distributions
of
contaminants.
The
simplest
probabilistic
sampling
strategy
is
random
sampling.
This
sampling
approach
is
useful
when
there
is
no
prior
knowledge
of
the
population
of
interest.
The
stratified
random
sampling
strategy
is
useful
when
the
population
of
interest
can
be
divided
into
homogeneous
subgroups.
The
stratified
random
sampling
approach
should
allow
for
equivalent
or
more
precise
estimation
of
the
mean
than
the
simple
random
sampling
strategy.
Cluster
sampling
strategies
are
useful
when
the
units
of
the
population
cannot
be
sampled
individually,
such
as
a
school
of
fish.
Systematic
sampling
strategies
are
useful
for
sampling
a
dispersed
population
of
interest
using
systematic
or
random
sampling
along
a
predefined
grid
pattern.
Systematic
sampling
can
lead
biased
estimates
of
the
mean
because
of
unknown
patterns
in
the
population
of
interest.
Also,
the
distinct
patterns
or
non­
randomness
in
the
population
of
interest
may
prevent
an
accurate
estimate
of
the
sampling
error.

Non
 
probabilistic
sampling
methods
such
as
haphazard
sampling
are
not
appropriate
for
a
national
drinking
water
survey
because
of
the
non­
random
nature
of
pesticide
use.
Judgement
sampling
is
a
more
appropriate
approach
because
sampling
can
be
focused
within
pesticide
use
areas.
The
heterogeneous
highly
structured
nature
of
judgement
sampling
is
expected
to
bias
estimates
of
the
mean
pesticide
concentrations.
This
type
of
sampling
can
be
associated
with
current
targeted
surface
monitoring
studies
for
assessing
pesticide
concentrations
at
vulnerable
CWS's
within
pesticide
use
areas.

ACPA
has
proposed
a
stratified
random
design
based
on
five
domains.

Domains
of
Interest
The
study
is
designed
around
domains,
which
are
distinct
portions
of
the
population
for
which
data
collected
will
meet
minimal
standards
of
accuracy.
In
this
survey,
the
domains
are
collections
of
CWS's
grouped
according
to
common
factors
such
as
pesticide
usage,
crop
areas,
regions,
and
watershed
vulnerability.
For
example,
to
draw
conclusions
about
CWS's
in
a
pesticide
use
area,
CWS's
would
be
the
target
population
and
the
pesticide
use
area
would
be
the
logical
domain.
Domains
need
to
be
constructed
in
a
manner
allowing
for
calculation
of
a
statistical
inference.
For
instance,
the
minimum
desired
statistical
inference
may
be
the
mean
annual
concentration
at
the
95th
percentile
site
with
95%
confidence
with
a
two
year
return
frequency.
Domains
should
be
chosen
to
reflect
the
target
population.
Domains
may
overlap
(
e.
g.
pesticide
use
area
for
multiple
compounds)
or
CWS's
may
be
assigned
to
separate
domains
(
e.
g.
national
food
region
domains).

Pesticide
Use
Domains
Pesticide
use
domains
are
CWS's
with
target
pesticide
use
in
their
watersheds.
The
advantage
of
using
pesticide
use
as
a
domain
is
that
survey
results
will
provide
information
about
drinking
water
quality
at
CWS's
in
watersheds
where
the
pesticide
is
used.
Additionally,
the
use
of
pesticide
use
as
a
domain
should
make
the
design
more
efficient
because
CWS's
with
multiple
target
pesticides
will
be
included
in
several
pesticide
domains.
Importantly,
data
will
be
collected
for
assessing
pesticide
co­
occurrence
for
cumulative
exposure
assessments.
A
disadvantage
is
that
site
selection
is
somewhat
more
difficult
because
the
same
CWS
may
fall
into
multiple
domains.
ACPA's
proposal
dealt
with
the
issue
of
overlapping
pesticide
use
areas
by
making
one
joint
domain
consisting
of
multiple
pesticides
defined
by
"
higher
use
intensity."
A
drawback
to
this
approach
is
that
domains
are
not
optimized
for
any
pesticide
in
8
(
Gilbert,
1987)

Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
16
particular.
This
means
(
1)
the
data
for
each
specific
pesticide
will
be
of
poorer
quality,
and
(
2)
it
becomes
more
difficult
to
make
risk
management
decisions
for
individual
pesticides.
Regional
Domains
Regional
domains
are
spatially
chosen
to
provide
a
regional
estimate
of
pesticide
concentrations
in
drinking
water.
Regional
domains
may
be
based
on
geopolitical
boundaries,
physiographic
regions
(
e.
g.,
coastal
plain
region,
etc.),
hydrologic
boundaries
(
e.
g.
USGS
Hydrologic
Unit
Code),
or
a
grid
system.
Regional
domains
do
not
overlap
(
i.
e.
individual
CWS's
fall
in
unique
domains),
which
simplifies
site
selection.
Monitoring
in
regional
domains
should
capture
a
broader
array
of
environmental
conditions
because
of
the
geographic
dispersion
of
sampling
sites.
This
type
of
information
is
useful
in
model
development,
identification
of
regional
mitigation
factors,
and
regional
assessment
of
pesticide
exposure
for
sub­
populations
of
people.
A
disadvantage
of
using
large
regional
domains
(
Western
US
v.
North
Eastern
US)
is
that
there
will
likely
be
a
substantial
number
of
sites
selected
in
which
there
is
little
or
no
target
pesticide
use.
This
makes
risk
management
decisions
abut
individual
pesticides
more
difficult.

ACPA
has
proposed
five
regional
domains
(
one
national
and
4
USDA
food
regions).
They
have
also
recommended
using
"
vulnerable"
strata
as
a
sub­
domain
in
each
region.

Crop
Use
Area
Domains
Crop
use
area
domain
can
defined
as
a
specific
area
where
a
particular
crop
is
grown
(
e.
g.,
cotton,
corn,
etc.).
An
advantage
of
using
crop
use
area
domains
is
the
ability
to
capture
a
broad
array
of
compounds
used
on
common
crops.
Overlap
is
likely
for
CWS's
in
domains
of
crops
used
in
common
crop
rotations
(
e.
g.,
corn
and
soybean),
and
this
should
increase
the
efficiency
of
the
design.
A
disadvantage
is
major
pesticides
are
used
on
multiple
crops.

Vulnerability
Domains
Vulnerability
domains
can
be
defined
as
the
set
of
CWS's
likely
to
have
high
concentrations
of
a
pesticide
or
a
set
of
pesticides
in
water
(
a
fuller
discussion
of
vulnerability
is
provided
below).
An
advantage
of
this
approach
is
that
risk
managers
will
have
available
better
data
for
these
more
vulnerable
CWS's.
If
vulnerability
is
used
as
the
stratification
criteria,
the
most
vulnerable
strata
could
be
used
as
a
sub­
domain.
A
drawback
to
this
approach
is
that
it
can
substantially
increase
the
number
of
samples
needed
in
the
survey.

Stratification
Stratification
is
the
process
of
separating
homogeneous
groupings
within
a
domain
for
more
efficient
characterization
of
the
distribution8.
Stratification
enables
one
to
reduce
the
number
of
samples
needed
or
to
increase
the
level
of
confidence
using
the
same
number
of
samples.
Because
we
do
not
have
good
information
about
concentration
distributions,
the
degree
of
improved
confidence
by
stratification
is
uncertain.
Vulnerability
may
be
described
using
watershed
characteristics,
hydrologic
characteristics
of
the
surface
water
body,
and
proximity
of
pesticide
application
area
to
water
source
supply.
Geographic
criteria
can
be
used
as
regional
strata.

Vulnerability
Strata
For
the
purpose
of
this
drinking
water
survey
design,
it
would
be
advantageous
to
have
a
simple
system
for
defining
vulnerability
categories
for
comparing
the
relative
risks
of
surface
water
contamination
between
watersheds
for
each
chemical.
A
vulnerable
CWS
is
one
likely
to
have
frequent
pesticide
detections
and/
or
high
levels
of
pesticides
in
raw
water.
Vulnerability
should
be
related
to
pesticide
usage
and
watershed
characteristics.
Using
a
vulnerability
index
offers
several
potential
advantages
in
designing
a
drinking
water
monitoring
study.
A
vulnerability
index
would
allow
a
more
focused
design
aimed
at
9A
Framework
for
Estimating
Pesticide
Concentrations
in
Drinking
Water
for
Aggregate
Exposure
Assessments.
International
Life
Sciences
Institute.
Washington,
D.
C.,
1999.

Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
17
watersheds
expected
to
produce
greater
than
average
pesticide
concentrations
in
surface
water.
It
might
also
be
used
to
identify
specific
types
of
water
bodies
associated
with
CWS
intakes
expected
to
have
higher
pesticide
concentrations.
By
developing
a
stratified
sampling
strategy
based
on
vulnerability,
the
study
design
can
have
more
statistical
power
to
address
its
goals.
In
a
study
attempting
to
characterize
upper
percentile
occurrence,
sampling
would
be
focused
on
those
watersheds
and
water
bodies
with
higher
vulnerability.

The
International
Life
Sciences
Institute
(
ILSI),
recommends
that
the
vulnerability
of
the
CWS's
be
used
as
strata
in
survey
design
for
assessing
pesticides
in
drinking
water9.
The
ILSI
guidance
for
drinking
water
survey
design
suggests:
"
The
sampling
population
should
be
stratified
by
classifying
the
use
regions
and
assessment
areas
into
higher­
and
lower­
risk
isopleth's
and
ranking
individual
supplies
by
vulnerability"
and
further
suggests
using
GIS
as
a
tool
for
quantifying
vulnerability.
Specific
characteristics
associated
with
increased
risk
were
identified
as
"
intensity
of
pesticide
use,
climatic
factors
(
e.
g.
rainfall
history),
contributing­
area
characteristics
(
e.
g.
geology,
slope),
source
of
supply,
type
of
water
delivery
function,
and
proximity
to
the
application
area."

The
type
and
size
of
the
water
body
from
which
drinking
water
is
derived
can
have
an
effect
on
the
magnitude
and
duration
of
pesticide
loadings.
Persistent
chemicals
may
accumulate
in
water
bodies
with
long
hydrologic
residence
times,
such
as
reservoirs
and
lakes,
while
water
bodies
with
shorter
residence
times
are
less
likely
to
accumulate
pesticides.
The
size
of
the
water
body
is
also
expected
to
affect
vulnerability.
For
instance,
small
rivers,
reservoirs,
and
lakes
are
likely
to
have
higher
pesticide
concentrations
than
large
ones
because
pesticide
loads
into
larger
water
bodies
would
be
diluted.
In
order
to
design
a
study
for
characterizing
pesticide
exposures
through
surface
derived
drinking
water
it
is
necessary
to
monitor
a
range
of
water
body
types.
Sampling
of
drinking
water
intakes
associated
with
different
types
of
water
bodies
(
e.
g.
rivers
and
reservoirs)
in
watersheds
with
a
range
of
predicted
vulnerabilities
should
be
included
in
the
study
design.

It
is
likely
the
effectiveness
of
certain
criteria
in
predicting
vulnerability
will
vary
from
chemical
to
chemical.
For
a
pesticide,
historic
levels
of
runoff
may
be
a
good
predictor
of
vulnerability,
while
for
another
pesticide,
usage
in
the
watershed
may
be
the
most
effective
indicator.
Since
it
is
not
known
which
factors
are
likely
to
be
the
best
indicators
of
vulnerability,
it
may
be
advantageous
to
use
more
than
one
vulnerability
indicator.
Stratifying
using
two
independent
potential
indicators
of
vulnerability
would
double
the
chances
of
producing
results
consistent
with
the
stratification
and
provide
twice
as
much
information
on
means
to
mitigate
surface
water
contamination.
The
simplest
way
to
use
multiple
criteria
of
vulnerability
would
be
to
use
two
vulnerability
criteria,
each
with
two
categories.
For
instance,
the
two
criteria
might
be
historic
runoff
levels
and
pesticide
usage
within
the
CWS's
watershed.
Each
of
these
two
criteria
may
have
two
categories:
high
and
low.
This
would
create
a
two
dimensional
matrix
with
four
categories
of
risk:
lowlow
low­
high,
high­
low,
and
high­
high.
This
type
of
vulnerability
index
does
not
attempt
quantify
the
magnitude
of
difference
between
different
categories
but
could
be
used
to
stratify
CWS's
allowing
for
higher
risk
CWS's
to
be
over­
sampled.

ACPA
has
proposed
to
stratify
by
vulnerability,
dividing
CWS's
selected
for
monitoring
in
the
chronic
survey
into
high
and
low
vulnerability
groups,
and
selecting
"
CWS's
using
reservoirs
and
having
a
higher
use
of
products
with
chronic
exposure
issues".
ACPA
will
identify
a
"
higher
pesticide
use
intensity
region"
based
on
the
total
use
intensity
of
target
pesticides.

Regional
Strata
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
18
Regional
strata
can
be
developed
to
group
CWS's
with
similar
geographical
properties,
such
as
climatic
conditions,
soil
types,
land
management
practices,
etc.
This
stratification
system
assumes
CWS's
within
a
defined
region
will
be
more
similar
than
CWS's
in
different
regions.
An
advantage
is
the
relative
ease
of
stratification.
A
disadvantage
is
that
CWS's
may
be
in
different
portions
of
the
distribution
even
if
they
are
geographically
close.

A
number
of
approaches
to
site
selection
have
been
used
in
past
monitoring
programs.
An
overview
for
some
of
these
approaches
is
presented.

Site
Selection
ACPA
Site
Selection
Strategy
The
site
selection
process
proposed
by
ACPA
for
chronic
exposure
is
based
upon
five
geographically
based
analysis
domains
 
the
conterminous
United
States
and
the
4
food
regions
defined
by
USDA.
Monitoring
sites
are
CWS's
which
derive
their
water
from
surface
water.
Sites
located
within
each
domain
are
classified
as
vulnerable
or
non­
vulnerable.
ACPA
uses
a
three
step
process
to
identify
and
select
the
sites:

1.
a.
Identify
the
counties
in
the
lower
48
states
which
represent
the
top
quartile
of
use
intensity
(
based
on
USGS
estimates)
for
each
of
25
target
pesticides.
This
area
is
referred
to
as
the
"
higher
pesticide
use
intensity
area".
Note
that
if
as
few
as
one
of
the
25
pesticides
is
used
in
a
county,
that
county
is
included
in
the
"
higher
pesticide
use
intensity
area",
even
if
no
other
pesticide
is
used
b.
Identify
all
CWS's
which
obtain
some
or
all
of
their
source
water
from
surface
water.
Classify
each
CWS
based
on
the
size
of
its
watershed
(
small/
medium/
large).
2.
a.
Identify
all
vulnerable
CWS's
that
occur
in
medium
and
small
watersheds
in
the
"
higher
pesticide
use
intensity
area".
To
identify
CWS's
that
are
vulnerable
to
chronic
exposure
ACPA
will
select
"
CWS's
using
reservoirs
and
having
a
higher
use
of
products
with
chronic
exposure
issues".
This
would
exclude
CWS's
using
major
continental
rivers
or
Great
Lakes,
which
would
be
left
in
the
"
non­
vulnerable"
category."
b.
Identify
all
CWS's
that
are
not
vulnerable
to
"
chronic
exposure."

(
It
is
not
clear
how
product
use
factors
into
identifying
CWS's
that
are
vulnerable
to
"
chronic
exposure".
Sites
initially
meeting
the
criteria
for
vulnerable
and
non­
vulnerable
would
be
verified
via
phone
contact
to
assess
the
accuracy
of
the
intake
surface
water
source,
pesticide
use
data
and
other
ancillary
information
such
as
population
served.)

3.
In
each
of
4
food
regions,
identify
59
vulnerable
and
24
non­
vulnerable
CWS's.
(
It
is
not
clear
how
CWS's
will
be
selected
from
the
list
of
candidates)

This
results
in
a
total
of
236
vulnerable
and
96
non­
vulnerable
surface
water
based
CWS's
being
selected
for
monitoring
nationally,
or
a
total
of
332
out
of
the
10,728
CWS's
where
monitoring
is
occurring.
For
urban
sites
ACPA
proposes
to
work
with
the
Agency
to
obtain
non­
crop
pesticide
use
data
for
compounds
of
interest
and
include
any
counties
in
the
higher
pesticide
use
intensity
area
that
would
reflect
such
use.

ARP­
Acetochlor
Site
Selection
Strategy
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
19
The
Acetochlor
Registration
Partnership
(
ARP)
is
conducting
an
ongoing
surface
water
monitoring
study
to
support
the
registration
of
acetochlor.
In
addition
to
acetochlor,
the
ARP
has
analyzed
for
atrazine,
and
alachlor.
This
study
includes
175
sites
in
12
states.
Sites
were
selected
from
among
surface
water
source
supplies
in
these
states
that
did
not
blend
with
ground
water
and
that
had
more
than
5%
corn
acreage
in
the
watershed
above
the
drinking
water
facility.
The
original
intent
was
to
select
25
sites
in
each
of
seven
states,
Iowa,
Illinois,
Indiana,
Kansas,
Nebraska,
Minnesota,
and
Wisconsin,
representing
80%
of
the
anticipated
acetochlor
use
area,
but
it
was
determined
that
several
states
such
as
Minnesota
and
Wisconsin
did
not
have
25
facilities
that
met
the
criteria.
The
number
of
sites
in
other
states
were
expanded,
and
two
other
corn
belt
states,
Ohio,
and
Missouri
were
added.
In
addition
three
east
coast
states
Delaware,
Maryland,
and
Pennsylvania
were
added
in
deference
to
the
needs
of
the
state
agencies
rather
than
to
meet
the
needs
of
OPP
for
risk
assessment.

The
sites
were
selected
using
a
stratified,
rather
than
a
statistical
random
approach.
Sites
were
placed
in
one
of
five
bins.
Two
bins
were
based
on
hydrology
of
the
source
water,
a
Great
Lakes
bin,
and
a
Continental
Rivers
bin.
The
remaining
sites
were
placed
in
bins
based
on
corn
acreage
in
the
watershed,
5
to
10%,
10
to
20%,
and
greater
than
20%.
More
sites
were
selected
from
the
bins
with
higher
corn
intensity.

Watershed­
based
Site
Selection
Strategy
The
Agency
is
currently
developing
a
watershed­
based
approach
to
site
selection.
This
approach
has
some
similarities
to
the
ARP
design
but
supports
a
statistical
selection
rather
than
one
based
on
judgement.
ILSI,
1999
also
recommends
using
pesticide
use
area
as
a
domain.
To
facilitate
the
site
selection
process
the
Agency
believes
geographic
information
systems
(
GIS)
should
be
used
to
evaluate
the
potential
vulnerability
of
a
CWS
to
pesticides
in
surface
water.

Site
Selection
Tools
The
process
of
identifying
vulnerable
sites
is
dependent
upon
various
data
sources
describing
the
location
of
the
CWS's,
the
hydrography
of
the
watershed
in
which
they
are
located,
land
use
within
the
watershed,
pesticide
use
within
the
watershed,
and
other
ancillary
data
(
e.
g.
soil
runoff
potential
or
annual
precipitation).

The
first
step
in
creating
a
site
selection
tool
is
identification
of
the
CWS
intake
locations.
According
to
EPA's
Office
of
Water
there
were
approximately
170,000
public
water
systems
(
PWS)
in
the
United
States
as
of
1999.
A
public
water
system:
"
provides
water
for
human
consumption
through
pipes
or
other
constructed
conveyances
to
at
least
15
service
connections
or
serves
at
least
25
people
for
at
least
60
days
a
year."
EPA
has
defined
three
subcategories
of
PWS's:

°
Community
Water
System
(
CWS):
A
PWS
that
supplies
water
to
the
same
population
year
around
°
Non­
Transient
Non­
Community
Water
System
(
NTNCWS):
A
PWS
that
regularly
supplies
water
to
at
least
25
of
the
same
people
at
least
six
months
per
year,
but
not
year
around.
(
e.
g.
schools,
factories,
office
buildings
or
hospitals)

°
Transient
Non­
Community
Water
Systems
(
TNCWS):
A
PWS
that
provides
water
in
a
place
such
as
a
gas
station
or
campground
where
people
do
not
remain
for
long
periods
of
time.

Currently
OPP
and
USGS
have
a
quality
assured
(
QA)
database
containing
the
location
of
intakes
for
CWS's
drawing
from
surface
water
serving
a
population
of
10,000
or
greater.
This
database
was
constructed
in
a
joint
effort
with
EPA's
Office
of
Water
and
USGS.
This
data
set
represents
approximately
20
percent
of
all
surface
water
CWS's
and
167
million
people.
The
USGS
National
Water
Quality
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
20
Figure
1.
Sample
QA
map
for
verifying
CWS
location.
Assessment
(
NAWQA)
program
expects
to
complete
quality
assurance
of
a
database
containing
location
information
for
all
surface
water
intakes
for
CWS's
serving
25
people
or
more
by
the
end
of
summer
2000.
Initially
EPA
will
use
the
data
set
of
CWS's
serving
10,000
and
greater
to
delineate
watershed
boundaries
and
then
augment
this
data
set
with
the
CWS's
serving
25­
10,000
once
it
is
available.
The
Office
of
Water
is
currently
updating
the
Safe
Drinking
Water
Information
System
(
SDWIS)
database
which
includes
locations
for
all
PWS's.
However,
until
this
updated
database
undergoes
complete
QA/
QC
its
utility
may
be
limited.
Therefore,
NTNCWS's
and
TNCWS's
are
not
included
in
the
site
selection
tool
under
development.

The
next
step
in
the
process
is
to
associate
the
CWS
intake
location
with
a
specific
river
reach
or
water
body
and
then
delineate
a
watershed
boundary.
EPA
is
using
the
National
Hydrography
Data
set
(
NHD)
for
this
purpose.
The
NHD
combines
USGS
digital
line
graph
hydrography
data
with
the
EPA
river
reach
file
version
3
(
RF3)
creating
an
enhanced
data
set
which
describes
surface
water
features
such
as
lakes,
streams,
rivers,
springs
and
wells
at
1:
100,000
scale.
Currently
the
NHD
is
available
for
most
of
the
conterminous
United
States.
It
is
expected
missing
regions
will
be
completed
soon.

Preliminary
analyses
indicate
that
a
manual
interactive
step
is
essential
to
determine
that
the
proper
reach
is
selected.
In
many
cases,
the
nearest
reach
is
not
the
actual
reach
where
the
intake
is
located.
This
is
important
because
identifying
the
wrong
reach
can
have
an
impact
on
the
delineated
watershed.
To
verify
the
intake
is
assigned
to
the
correct
reach
USGS
Digital
Raster
Graphs
(
DRG)
or
Digital
Ortho
Quads
1:
24k
or
similar
product
are
needed
to
verify
site
location
(
e.
g.,
use
location
name
and
information
from
DRG
to
verify
site
is
located
correctly).
As
a
quality
assurance
step
a
standardized
map
of
site
location
with
DRG
streams
can
be
plotted
and
sent
to
the
CWS
for
verification
(
Figure
1).

Once
the
CWS
has
been
assigned
to
the
proper
reach
a
watershed
boundary
is
delineated
within
a
cataloging
unit
(
CU).
The
CU's
of
the
NHD
are
based
on
8­
digit
hydrologic
unit
code
(
HUC).
Initial
work
has
shown
the
CU­
to­
CU
connectivity
in
NHD
is
poor.
Therefore
our
method
will
delineate
a
watershed
10National
Land
Cover
Data
read
me
file
for
Washington
State
version
06­
01­
99.

Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
21
Figure
2.
Upstream
and
surrounding
cataloging
units
for
delineated
basin.
within
a
CU
and
superimpose
boundaries
for
upstream
CU's
during
a
subsequent
step.
This
processes
is
illustrated
in
Figures
2
and
3.

Once
a
watershed
boundary
has
been
defined
for
a
specific
CWS
various
data
sets
can
be
intersected
with
it
and
watershed
statistics
calculated.
The
first
data
layer
to
be
added
will
describe
land
use
and
land
cover
(
LU/
LC).
There
are
two
data
sets
the
agency
proposes
using
for
this
purpose:
the
National
Land
Cover
Data
(
NLCD);
and
the
Geographic
Information
Retrieval
and
Analysis
System
(
GIRAS)
LU/
LC.

The
NLCD
is
based
on
30­
meter
Landsat
thematic
mapper
data
collected
in
the
early
90'
s.
As
with
all
LU/
LC
data
there
are
certain
limitations.
For
example
the
NLCD
data
for
Washington
state
contains
the
caveat
"
Crop
types
in
the
irrigated
areas
were
difficult
to
reliably
distinguish;
row
crops
are
likely
to
be
under
represented
where
no
field
observations
or
other
ancillary
information
was
incorporated10."
Further
the
NLCD
is
not
complete
at
this
time.
Fifteen
states
remain
to
be
completed
including
Texas
and
the
Dakotas.
Since
the
NLCD
is
incomplete
a
separate
data
source
is
needed
to
fill
in
the
blanks.

Where
as
the
NLCD
is
based
upon
a
single
data
source,
GIRAS
is
a
compilation
of
aerial
photographs
and
previous
LU/
LC
maps
with
some
of
the
data
dating
to
the
early
70'
s.
Also,
the
spatial
resolution
is
200­
400­
meters
as
opposed
to
30­
meters
for
NLCD.
Although
GIRAS
is
composed
of
older
data
it
does
provide
a
complete
coverage
for
the
contiguous
United
States.
For
the
GIS
tool
being
created
GIRAS
will
serve
two
purposes:
Fill
in
the
gaps
of
the
NLCD;
and
compared
with
the
NLCD
to
determine
the
variation
in
LU/
LC
statistics
(
e.
g.
percent
agriculture
in
a
watershed).

Once
the
LU/
LC
data
is
intersected
with
the
watershed
boundary
the
amount
of
agricultural
or
urban
land
within
the
watershed
can
be
calculated.
Then
the
amount
of
agricultural
land
within
a
basin
can
have
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
22
Figure
3.
Delineated
basin
boundary
within
an
individual
cataloging
unit.
pesticide
use
data
apportioned
to
it
and
an
estimate
of
pounds
active
ingredient
applied
in
the
watershed
calculated.
Since
there
are
no
known
sources
of
pesticide
use
data
in
urban
or
nonagricultural
areas
urban
sites
can
be
selected
based
upon
the
amount
of
urban
land
in
the
CWS
watershed.
The
greater
amount
of
urban
land
use
in
the
watershed
the
more
likely
a
CWS
would
be
selected
for
sampling.

The
pesticide
use
data
is
the
same
used
for
the
USGS
National
Water
Quality
Assessment
Pesticide
National
Synthesis
Project
annual
use
maps.
This
data
set
was
chosen
because
it
is
accessible,
non­
proprietary,
and
is
from
data
corresponding
temporally
with
the
NLCD.
The
disadvantage
is
that
the
data
represents
an
average
use
of
a
particular
pesticide
for
the
entire
state
and
does
not
necessarily
represent
local
practices.
Also,
Census
of
Agriculture
data
for
county
level
crop
acreage
is
sometimes
censored
due
to
non­
disclosure
rules.
Therefore,
the
utility
of
basing
a
site
selection
strategy
on
use
alone
is
of
questionable
value.
This
data
allows
for
determining
potential
areas
where
the
pesticide
of
concern
is
used
and
may
allow
for
categorizing
use
areas
into
a
high,
low
and
no
use
category.
Further,
the
smaller
the
watershed
delineated,
the
less
likely
use
data
and
to
a
certain
extent
LU/
LC
data
will
be
accurate.
This
illustrates
the
need
to
follow
up
the
selection
process
with
telephone
interviews
or
reconnaissance
trips
to
the
potential
monitoring
sites.

Number
of
Sites
The
number
of
CWS's
sampled
needs
to
be
balanced
against
the
number
of
samples
collected
from
each
CWS
in
order
to
reduce
the
total
uncertainty
in
the
study.
Uncertainty
in
the
results
from
pesticide
sampling
can
come
from
four
main
areas:
year
to
year
variations,
variations
due
to
sampling
location,
variations
due
to
the
method
in
which
the
mean
annual
concentration
is
measured,
and
uncertainty
due
to
chemical
analysis.
In
a
one
year
study
we
have
no
control
over
year
to
year
variations
and
the
uncertainty
due
to
chemical
analysis
is
fixed
by
the
analysis
method,
so
the
most
important
compromise
is
between
the
number
of
sites
measured
and
the
number
of
samples
taken
from
each
site
for
calculating
the
annual
mean
concentration.
11Guttman,
I,
S.
S.
Wilks,
and
J.
S.
Hunter.
1982.
Introductory
Engineering
Statistics,
3rd
edition.
John
Wiley,
New
York.

Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
23
The
best
compromise
depends
on
the
goals
of
the
study,
more
locations
will
batter
describe
the
spatial
variation
in
pesticide
concentrations
in
the
use
areas,
while
more
samples
per
location
will
be
able
to
provide
more
accurate
data
on
the
temporal
variability
of
pesticide
concentrations.
The
pesticides
of
interest
and
the
fate
properties
of
those
pesticides
are
thus
required
in
order
to
properly
determine
how
to
distribute
samples.

The
number
of
sampling
locations
required
to
determine
a
given
percentile
of
a
distribution
with
a
given
level
of
confidence
can
be
determined
by
using
order
statistics
with
minimal
assumptions
about
the
distribution.
Knowledge
of
a
percentile
alone
may
not
be
sufficient
to
determine
if
a
chemical
poses
no
risk
to
a
definable
sub­
population.
Distributions
of
pesticides
may
be
highly
skewed,
meaning
that
the
water
systems
above
the
known
percentile
could
have
pesticide
levels
significantly
higher
than
any
sampled
sites.

Calculation
of
the
sample
size
required
to
estimate
a
given
percentile
with
a
given
confidence
is
outlined
as
follows.
If
we
sample
n
sites
and
measure
a
quantity
x
in
each
of
them,
we
can
write
the
probability
that
the
largest
measurement,
x(
n),
is
less
than
some
quantity
u,
as
the
probability
that
all
measurements
xi
are
less
than
u
P(
x(
n)
<
u)
=
P(
x1
<
u,
x2
<
u,
...,
xn
<
u)

Assuming
that
all
the
xi
are
independent,
we
can
multiply
the
individual
probabilities
together
P(
x(
n)
<
u)
=
P1(
x1
<
u)
P2(
x2
<
u)
...
Pn(
xn
<
u)

and
assuming
the
cumulative
distribution
functions
(
cdf's)
are
identical,
P(
x<
u)
=
F(
u),
this
can
be
shortened
to
P(
x(
n)
<
u)
=
Fn(
u)
=
G(
u)

Where
F
is
the
cdf
for
the
xi
and
G
is
the
cdf
for
x(
n).
We
can
then
differentiate
this
to
get
a
probability
density
function
(
pdf)
for
the
largest
element,

g(
u)
=
nFn­
1(
u)
f(
u).

The
assumption
of
independent
identically
distributed
random
variables,
is
critical
because
without
it
we
can't
make
the
step
of
multiplying
the
individual
cdf's
together
and
the
function
above
does
not
describe
the
pdf
of
the
largest
element
of
our
sample.
The
pdf
above
can
then
be
used
to
show11
that
P[
F(
x(
n))
>
$
]
=
1
­
$
n
which
says
that
at
least
a
fraction
$
of
the
population
has
values
of
x
less
than
x(
n)
with
probability
1
­
$
n.
If
we
let
(
=
P[
F(
x(
n))
>
$
]
then
1
­
$
n
=
(
,
where
$
is
the
quantile,
(
the
confidence
limit,
and
n
is
the
number
of
samples
taken.
Solving
for
n
gives
n
=
ln(
1­
(
)
/
ln
$
.

For
example,
to
ensure
95%
confidence
that
the
largest
sample
is
at
or
above
the
95th
percentile
of
the
distribution
would
require
n
=
ln(
1­
0.95)/
ln(
0.95)
=
58.4
samples
and
we
would
round
up
to59
to
be
sure
we
were
within
the
set
bounds.
Similarly,
90%
confidence
in
the
99.9th
percentile
would
require
2302
sampling
sites.
This
example
is
the
same
as
what
ACPA
has
proposed.
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
24
Atrazine
Day
of
Year
0
100
200
300
Concentration
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
Bromoxynil
Day
of
Year
0
100
200
300
Concentration
0.00
0.01
0.02
0.03
0.04
0.05
0.06
Figure
4.
Yearly
concentration
profiles
for
atrazine
and
bromoxynil
simulated
by
PRZM­
EXAMS.
Number
of
Samples
Choosing
the
number
of
samples
required
to
define
the
average
value
at
a
site
is
not
as
straightforward.
Samples
taken
from
a
water
body
are
clearly
not
independent,
identically
distributed
random
variables,
as
required
by
the
approach
outlined
above.
Instead,
the
concentrations
follow
a
pattern
over
the
year,
and
how
the
sampling
times
coincide
with
this
pattern
can
have
an
important
effect
on
the
average
value
of
the
samples.
To
calculate
an
annual
mean
value
with
perfect
accuracy
would
require
sampling
at
twice
the
rate
of
the
fastest
fluctuation,
but
would
never
fit
within
budget
constraints.

Because
we
can't
sample
enough
to
capture
every
nuance
of
the
signal,
we
would
like
to
know
how
much
error
is
associated
with
sampling
at
longer
time
intervals.
Without
knowing
the
shape
of
the
yearly
concentration
curve
it
is
impossible
to
quantitatively
determine
the
error
associated
with
less
frequent
sampling
or
to
define
an
ideal
schedule
of
less
frequent
samples.
Instead
to
illustrate
the
possible
errors
we
look
at
some
examples.

The
shape
of
a
yearly
concentration
curve
(
or
chemograph)
is
dependent
on
many
variables
including
the
fate
characteristics
of
the
chemical,
the
use
of
the
chemicals,
the
weather
in
the
area
and
the
hydrologic
characteristics
of
the
watershed.
Of
these,
the
weather
and
hydrologic
properties
cannot
be
defined
well
enough
to
aid
in
determining
the
deeded
sampling
frequency.
However,
the
fate
characteristics
of
the
chemicals
to
be
sampled
can
be
used
with
modeling,
to
help
determine
the
number
of
samples
needed
to
define
a
yearly
average
concentration.
Models
would
be
run
(
as
in
Figure
4)
to
get
an
estimate
of
the
shape
of
the
chemograph,
and
this
curve
would
then
be
used
to
determine
the
number
of
samples
needed.

Measuring
an
average
yearly
concentration
requires
measuring
the
area
under
the
concentration
curve
for
the
entire
year.
Persistent
compounds,
in
which
concentrations
vary
slowly
throughout
the
year
will
thus
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
25
Atrazine
Year
0
10
20
30
Fractional
Deviation
from
True
Mean
­
1.0
­
0.5
0.0
0.5
1.0
Bromoxynil
Year
0
10
20
30
Fractional
Deviation
from
True
Mean
­
1.0
­
0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Figure
5.
Variations
due
to
28­
day
sampling.
exhibit
smaller
errors
with
long
sampling
intervals
than
shortlived
compounds
because
a
simple
interpolation
between
sampling
points
will
more
closely
approximate
the
true
concentration
profile
over
the
year.
For
example,
Figure
4
shows
PRZM/
EXAMS
simulations
of
two
pesticides
under
identical
environmental
and
cropping
conditions
for
36
years.
(
The
PRZM/
EXAMS
model
was
used
because
we
did
not
have
access
to
daily
field
data
for
two
such
pesticides
under
similar
conditions.)
Atrazine,
the
more
persistent
of
the
two,
is
present
(
at
least
at
low
concentrations)
throughout
the
entire
year
but
bromoxynil
degrades
rapidly
and
disappears
approximately
one
month
after
application.
Even
with
no
simulation
of
sampling,
it
is
clear
that
monthly
sampling
is
inadequate
to
characterize
the
mean
concentration
of
bromoxynil.
With
a
peak
lasting
only
a
month,
it
would
be
quite
possible
to
miss
the
presence
of
bromoxynil
entirely.

Figures
5
and
6
show
the
variability
associated
with
sampling
the
model
output
shown
in
Figure
4
at
7
and
28
day
intervals.
These
figures
show
the
difference
between
the
sampling
at
a
7
or
28­
day
interval
and
sampling
daily,
and
are
reported
as
a
fraction
of
the
annual
mean.
The
simulated
sampling
was
carried
out
for
each
possible
starting
date,
i.
e.
for
7­
day
sampling,
sampling
simulations
were
run
with
the
first
samples
taken
on
the
1st,
2nd,
3rd,
4th,
5th,
6th
and
7th
day
of
the
year
and
repeated
every
7
days.
Naturally
for
the
28­
day
interval
(
Figure
6)
there
were
28
sampling
simulations.
ACPA
has
proposed
collecting
an
average
of
10
samples
or
four
seasonally
composited
samples
at
each
CWS.
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
26
Atrazine
Year
0
10
20
30
Fractional
Deviation
from
True
Mean
­
1.0
­
0.5
0.0
0.5
1.0
Bromoxynil
Year
0
10
20
30
Fractional
Deviation
from
True
Mean
­
1.0
­
0.5
0.0
0.5
1.0
Figure
6.
The
range
of
yearly
averages
possible
with
a
7­
day
sampling
interval
Sample
Timing
Modeling,
as
described
above,
is
one
way
to
estimate
the
sampling
frequency
required
for
an
acceptable
measurement
of
yearly
average
concentration.
Different
sampling
intervals
for
atrazine
and
bromoxynil
show
much
higher
variability
with
less
frequent
sampling
and
higher
variability
with
the
more
quickly
degrading
bromoxynil.
Sampling
atrazine
at
a
28­
day
interval
gives
measurements
between
+
20%
to
!
10%
of
the
true
mean
(
in
this
case
the
"
true
mean"
is
the
mean
calculated
from
daily
model
output
values).
Sampling
bromoxynil
at
the
same
interval
can
yield
results
as
much
as
2
times
the
true
mean
or
as
little
as
25%
of
the
true
mean.
The
deviations
from
the
true
modeled
mean,
as
shown
in
Figures
5
and
6,
should
be
less
than
the
inaccuracy
in
the
analytical
method
in
order
to
keep
overall
measurement
errors
to
a
minimum.

Disadvantages
to
using
this
approach
are
that
concentrations
in
nature
can
fluctuate
more
rapidly
than
model
predictions.
More
rapid
fluctuations
in
concentration
require
more
frequent
sampling
and
thus
this
approach
give
an
underestimate
of
the
error
associated
with
each
sampling
interval.

To
reduce
the
number
of
samples,
sampling
could
be
concentrated
around
the
time
of
application,
when
chemograph
fluctuations
are
most
pronounced.
This
method
should
yield
results
similar
to
sampling
all
year
at
the
higher
rate,
because
fluctuations
in
concentration
are
smaller
and
less
rapid
outside
of
the
time
of
application.
For
example,
if
half
of
all
samples
were
taken
between
days
100
and
200
(
inspection
of
Figure
4
shows
this
is
the
period
of
greatest
change)
and
the
remaining
half
take
at
longer
intervals
during
the
rest
of
the
year.
The
period
of
more
frequent
sampling
would
depend
on
local
use
patterns
for
each
pesticide
in
the
survey
and
should
vary
both
between
pesticides
and
between
geographical
regions.
The
ARP
study
collected
14
samples
per
year
with
biweekly
samples
during
the
spring
and
summer.
If
the
concentrations
at
any
particular
site
remained
elevated
past
August,
the
ARP
continued
biweekly
sampling
until
levels
declined
to
background
levels.
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
27
This
approach
assumes
there
are
no
rapid
fluctuations
in
concentration
during
the
periods
of
infrequent
sampling,
an
assumption
that
holds
up
well
with
modeled
output
but
may
be
less
valid
with
real
data.
If
there
are
important
concentration
spikes
during
the
infrequent
sampling
period,
these
may
show
up
in
some
of
the
CWS
chemograph's
and
can
be
used
(
after
completion
of
sampling)
to
assess
the
assumption
of
little
fluctuation
during
that
period.

Another
way
to
get
around
the
problems
associated
with
too
few
samples
would
be
to
use
composite
samples.
If
samples
were
taken
in
small
amounts
at
rapid
intervals
and
mixed
together,
then
the
result
would
be
the
mean
value
of
all
the
small
samples
over
the
time
of
sampling.
The
problem
with
composite
samples
is
that
it
is
more
difficult
to
keep
samples
fresh.
In
a
30­
day
composite,
by
definition,
the
first
sample
is
30
days
older
than
the
last,
and
if
the
chemical
of
interest
changes
in
those
times
the
final
result
will
not
be
an
accurate
mean
value
of
concentration
over
the
30
day
time.

ACPA
has
proposed
to
sample
each
CWS
systematically
through
out
the
year
with
"
less
frequent
sampling
of
water
from
reservoir's
and
lakes."

Number
of
Samples
v.
Number
of
Sites
The
tradeoff
between
how
many
sites
to
sample
and
how
many
samples
taken
from
each
site
is
difficult
to
quantify
because
of
the
unknown
nature
of
the
shape
of
the
chemograph
at
each
site.
We
can
use
statistical
methods
to
help
define
the
number
of
samples
needed
to
achieve
a
defined
confidence
in
a
given
percentile,
but
that
number
is
valid
only
when
taking
a
simple
measurement
with
no
error.
Any
practical
sampling
method
will
introduce
errors
and
sampling
at
infrequent
intervals
will
introduce
large
errors.
These
errors
must
be
balanced
against
the
desired
accuracy
when
making
the
tradeoff
between
sampling
many
sites
and
sampling
more
often.

Because
sampling
must
be
carried
out
frequently
for
an
accurate
measurement
of
the
annual
average
concentration,
it
will
also
yield
a
reasonable
estimate
of
the
peak
concentration.
For
a
limited
study,
as
proposed
here,
the
sampling
is
limited
to
long­
lived
compounds
in
which
the
peaks
decay
slowly
and
thus
can
by
found
by
infrequent
sampling.
If
peak
decay
can
be
shown
to
follow
a
recognizable
decay
curve,
it
may
also
be
possible
to
extrapolate
a
peak
value
from
that
curve.

Raw
and
Finished
Water
Water
is
typically
treated
before
distribution
to
consumers.
Most
treatment
plants
use
conventional
treatment
such
as
coagulation/
flocculation,
sedimentation,
filtration,
and
disinfection
with
a
chlorine
compound
or
ozone.
Some
plants
also
use
advanced
filtration
processes
(
e.
g.
activated
carbon
and
reverse
osmosis).
Some
water
treatment
facilities
located
in
agricultural
areas
adjust
treatment
during
times
of
pesticide
use
to
reduce
concentrations
and
subsequent
exposure,
at
added
cost.
Because
some
pesticides
are
not
stable
in
the
presence
of
strong
oxidizing
agents
such
as
chlorine,
or
at
elevated
pH,
pesticides
concentrations
may
be
reduced
or
removed
through
treatment.
In
addition
to
degradation,
during
chemical
12Adams,
C.
D.
and
S.
J.
Randtke.
1992.
Removal
of
Atrazine
from
Drinking
Water
by
Ozonation.
Jour.
Amer.
Water
Works
Assoc.
84(
9):
91­
102;

13Magara,
Y.,
T.
Aizawa,
N.
Matumoto,
and
F.
Souna.
1994.
Degradation
of
Pesticides
by
Chlorination
during
Water
Purification.
Wat.
Sci.
Tech.
30(
7):
119­
128;

14Acero,
J.
L.,
K.
Stemmler,
and
U.
V.
Gunten.
2000.
Degradation
Kinetics
of
Atrazine
and
Its
Degradation
Products
with
Ozone
and
OH
Radicals:
A
predictive
Tool
for
Drinking
Water
Treatment.
Environ.
Sci.
Technol.
34(
4):
591­
597.

15A
Framework
for
Estimating
Pesticide
Concentrations
in
Drinking
Water
for
Aggregate
Exposure
Assessments.
International
Life
Sciences
Institute.
Washington,
D.
C.,
1999.

Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
28
disinfection
with
chlorine
and
ozone
certain
pesticides
are
transformed
to
other
compounds
which
may
have
toxicological
significance.
12,13,14
Ideally,
tap
water
should
be
sampled
and
analyzed
for
levels
of
pesticide
residues
in
drinking
water.
However,
tap
water
represents
a
composite
of
source
waters
from
supply
intakes
subjected
to
varying
treatment
effects,
distribution
allocation,
and
mixing
in
distribution
systems15.
The
sampling
options
for
a
national
survey
design
are
described
for:
finished
water
only;
raw
water
only;
and
a
combination
of
raw
and
finished
water.

Finished
Water
Finished
water
samples
are
collected
at
the
outflow
from
a
water
treatment
plant.
Sampling
finished
water
more
closely
represents
the
water
people
drink,
so
survey
results
can
be
used
directly
to
assess
human
exposure.
In
addition,
the
magnitude
of
residues
(
parent
and
transformation
product)
measured
represent
the
combined
effects
of
treatment.

Sampling
only
finished
water
means
no
information
will
be
collected
on
the
concentration
of
pesticides
entering
water
treatment
facilities,
so
the
Agency
cannot
assess
the
burden
placed
upon
CWS's
to
remove
pesticides.
Also,
pesticide
concentrations
in
finished
water
cannot
be
linked
to
watershed
characteristics
or
other
factors
that
could
be
used
by
risk
mangers
to
mitigate
exposure.
Since
treatment
is
variable
between
water
treatment
plants
and
within
the
same
plant
over
time,
an
additional
source
of
variability
is
introduced.
To
assess
CWS
vulnerability
during
the
site
selection
process
one
would
have
to
know
the
treatment
processes
used
at
the
CWS
and
the
effects
of
treatment
on
all
pesticides.
If
this
is
not
done,
a
system
with
advanced
treatment
methodologies
which
remove
pesticides
could
erroneously
be
characterized
as
"
vulnerable."
Further,
finished
drinking
water
data
is
not
adequate
for
model
development
or
validation,
because
simulation
and
regression
models
simulate
natural
hydrologic
processes
and
agronomic
practices,
not
treatment
effects.
Finally,
data
cannot
be
used
to
evaluate
the
effectiveness
of
treatment
for
reducing
pesticide
concentrations
without
collection
and
analysis
of
temporally
paired
raw
and
finished
water
samples.
ACPA
has
proposed
sampling
only
finished
water
in
their
survey.

Raw
Water
Raw
water
is
sampled
at
a
CWS
intake
prior
to
treatment.
Since
raw
water
does
not
reflect
the
effects
of
treatment,
data
interpretation
(
including
identification
of
causative
factors)
is
simplified.
Based
on
raw
water
analyses
and
ancillary
data
collected
in
the
watershed,
pesticide
concentrations
can
be
correlated
to
hydrologic
factors
and
sources
of
contamination
when
identifying
possible
measures
to
mitigate
high
concentrations.
Analyses
of
raw
water
samples
will
enable
predictive
computer
models
to
be
developed
and
validated.
In
addition,
since
raw
water
samples
represent
the
impact
that
pesticides
have
on
the
aquatic
environment
as
a
result
of
their
application,
this
occurrence
data
is
directly
relevant
for
ecological
assessment.

Few
people
drink
water
directly
from
a
surface
water
body.
Pesticide
concentrations
in
raw
water
can
serve
as
a
conservative
estimate
of
pesticide
residues
in
drinking
water
or,
if
the
effectiveness
of
treatment
16Larson
S.
J.,
P.
D.
Capel,
and
M.
S.
Majewski.
(
1997?)
Pesticides
in
Surface
Waters:
Distribution
Trends
and
Governing
Factors.
Ann
Arbor
Press,
Inc.
Chelsea
Michigan.
373
pp.

17Carsel,
R.
F.,
J.
C.
Imhoff,
P.
R.
Hummel,
J.
M.
Cheplick,
and
A.
S.
Donigian,
Jr.
PRZM­
3,
Model
for
Predicting
Pesticide
and
Nitrogen
Fate
in
the
Crop
Root
and
Unsaturated
Soil
Zones:
Users
Manual
for
Release
3.0.
National
Exposure
Research
Laboratory.
Office
of
Research
and
Development.
US
EPA
30605­
2720.

Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
29
technologies
at
removal
of
a
specific
pesticide
can
be
reliably
predicted,
they
can
be
used
to
estimate
finished
drinking
water
concentrations.
Also,
data
on
transformation
products
cannot
be
obtained
if
solely
raw
water
samples
are
collected.
Finally,
data
cannot
be
used
to
evaluate
the
effectiveness
of
treatment
at
reducing
pesticide
concentrations
from
raw
to
finished
water
without
collection
and
analysis
of
temporally
paired
raw
and
finished
water
samples.

Raw
and
Finished
Water
Collecting
raw
and
finished
water
samples
could
significantly
affect
the
overall
cost
of
the
survey.
Therefore,
a
number
of
alternatives
were
considered:
fully
paired
samples
(
one
raw
and
one
finished);
fully
paired
samples
(
one
raw
and
one
finished)
with
reactive
analysis
(
the
finished
sample
is
analyzed
only
if
residues
are
detected
in
the
raw
sample);
and
paired
samples
for
the
samples
collected
during
the
pesticide
use
season,
with
reactive
analysis.

There
are
many
advantages
in
sampling
both
raw
and
finished
water,
as
described
above.
Finished
water
analyses
provide
a
better
representation
of
the
water
people
drink,
including
the
opportunity
to
characterize
transformation
product
concentrations,
if
desired.
Raw
water
analyses
provide
a
measure
of
pesticide
loading
to
a
water
treatment
facility,
results
can
be
linked
to
possible
mitigation
options,
and
data
can
be
used
to
develop
and
validate
predictive
models.
Raw
water
data
can
be
used
directly
for
ecological
risk
assessment.
Importantly,
data
for
paired
samples
can
be
used
to
evaluate
the
combined
effects
of
treatment
and
processing
on
pesticide
concentrations.

Although
this
option
resolves
many
of
the
disadvantages
identified
above
in
sampling
just
raw
or
finished
water,
a
disadvantage
is
increased
cost
of
analysis.
If
fully
paired
samples
are
collected
and
analyzed
at
each
sampling
interval,
the
analytical
costs
would
essentially
double.
Reactive
sampling
would
reduce
added
costs
of
paired
samples.
Further,
it
is
difficult
to
collect
temporally
paired
raw
and
finished
water
samples.

Duration
of
the
Study
The
mean
annual
concentration
for
a
given
pesticide
at
particular
CWS
within
the
pesticide's
use
area
varies
from
year
to
year.
Factors
affecting
variability
at
a
particular
CWS
may
be
related
to
climactic
conditions
(
e.
g.
timing
and
intensity
of
rain
events),
long­
term
usage
trends
(
e.
g.
increasing
market
share
of
the
pesticide
or
changing
cropping
patterns),
and/
or
short­
term
trends
(
e.
g.
a
change
in
pest
pressure).

Rainfall
can
vary
greatly
from
year
to
year.
In
years
with
low
rainfall,
pesticide
runoff
is
lower.
Seasonal
variation
in
rainfall
can
also
greatly
affect
pesticide
runoff.
Rain
events
occurring
shortly
after
pesticide
applications
produce
higher
runoff
concentrations.
This
is
confirmed
by
monitoring
and
modeling
results
for
many
pesticides.
In
monitoring
studies
atrazine,
alachlor,
cyanazine,
and
metolachlor
concentrations
are
strongly
correlated
with
application
timing
and
intensity
and
year­
to­
year
variability
in
peak
measured
concentration
is
large16.
In
PRZM17
modeling
used
to
estimate
runoff
from
a
variety
of
soil
types
under
different
management
practices,
pesticide
runoff
loads
are
strongly
correlated
to
simulated
rainfall
intensity
and
timing.

Pesticide
use
varies
greatly
from
year
to
year.
Climactic
conditions
and
other
factors
influence
pest
intensity
and
thus
pesticide
use.
Additionally
label
restrictions
may
limit
the
availability
of
a
given
chemical
18Aspelin,
A.
L.
and
A.
H.
Grube.
(
1999)
Pesticide
Industry
Sales
and
Usage
1996
and
1997
Market
Estimates.
Biological
and
Economic
Analysis
Division,
Office
of
Pesticide
Programs,
US
EPA
20460.
Available
at
http://
www.
epa.
gov/
oppbead1/
pestsales/
97pestsales/
97pestsales.
pdf
19US
Department
of
Agriculture.
(
1991)
National
Boll
Weevil
Cooperative
Control
Program:
Final
Environmental
Impact
Statement
­
1991
volume
1.
Animal
and
Plant
Health
Inspection
Service.
Hyattsville
Maryland,
20782.

Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
30
in
certain
regions
from
year
to
year
which
may
in
turn
lead
to
higher
use
of
other
chemicals.
Alachlor
and
metam
sodium
exemplify
long
term
trends
of
changing
pesticide
usage
on
a
national
scale.
In
1995,
approximately
19­
24
million
pounds
of
alachlor
and
49­
54
million
pounds
of
metam
sodium
were
used
in
agricultural
crop
production18.
In
1983,
55­
60
million
pounds
of
alachlor
and
5­
8
million
pounds
of
metam
sodium
were
used.
On
a
regional
scale,
the
changing
pest
pressures
can
have
an
even
greater
effect
on
short
term
trends
on
pesticide
usage
and
application
frequency.
After
completion
of
the
USDA
Boll
Weevil
Eradication
Program
(
BWEP)
insecticide
applications
to
cotton
in
the
Southeast
decreased
88%
19.
Growers
prior
to
the
BWEP
made
10
to
12
applications
per
year.
After
the
BWEP
growers
averaged
two
applications
per
year.

One
Year
Study
A
one
year
study
will
meet
the
resource
and
cost
constraints
and
while
the
data
collected
from
one
year
of
sampling
will
be
useful,
it
will
also
have
limitations.
A
one
year
study
is
unable
to
capture
year
to
year
variability,
and
a
study
conducted
over
this
period
may
not
be
representative
of
typical
conditions
and
is
unlikely
to
reflect
upper
percentile
values.
Because
pesticide
concentrations
in
surface
water
are
highly
variable,
results
of
any
study
conducted
over
a
single
year
must
be
viewed
within
the
time
frame
of
the
study.
The
magnitude
of
pesticide
concentrations
in
drinking
water
sources
must
be
considered
in
relation
to
usage
and
to
weather
during
that
year.
Any
statistical
analysis
(
e.
g.
percentile
or
tolerance
bound)
of
a
1­
year
study
would
be
limited
to
the
year
in
which
the
data
was
collected.
For
instance,
the
95th
percentile
value
would
represent
the
95th
percentile
concentration
during
that
year.

One
Year
Study
With
Limited
Monitoring
Beyond
One
Year
A
one
year
study
with
continued
monitoring
of
a
fraction
of
the
original
sites
in
subsequent
years
will
conserve
resources
and
focus
monitoring
on
sites
previously
found
to
have
high
pesticide
concentrations.
However,
the
full
monitoring
program
would
last
only
one
year,
so
year
to
year
variability
across
sites
would
not
be
measured.
In
addition,
limited
CWS
monitoring
and
the
method
of
site
selection
after
the
first
year,
would
make
it
difficult
to
draw
any
statistical
inference
regarding
the
results.

A
survey
design
using
one
year
of
full
monitoring
and
supplemental
monitoring
for
subsequent
years
would
be
more
costly
than
a
one
year
study
and
the
results
from
subsequent
years
would
not
improve
the
statistical
power
of
the
study.
However,
data
collected
from
subsequent
years
could
be
used
to
estimate
year
to
year
variation
in
the
annual
mean
concentration
at
certain
sites.
With
the
proper
ancillary
data
at
multiple
year
monitoring
sites,
it
may
be
possible
to
correlate
annual
mean
concentrations
with
particular
factors.
ACPA
has
proposed
sampling
for
one
year
with
unspecified
triggers
for
extended
monitoring
at
some
sites.

Multiple
Years
of
Monitoring
A
multiple
year
monitoring
study
will
better
account
for
climatic
variability,
temporal
variability,
variation
in
pesticide
usage
and
other
factors.
ILSI
recommends
a
multi­
year
study
for
adequately
estimating
pesticide
concentrations
in
surface
water.
The
major
limitation
in
existing
monitoring
data
is
the
absence
of
systematic,
long­
term
studies
that
collect
frequent
samples
over
time
to
identify
long­
term
trends
of
concentrations
in
surface
waters:
"
From
the
studies
reviewed,
little
can
be
concluded
about
the
long­
term
trends
in
pesticide
occurrence
in
surface
waters
of
the
United
States
for
several
reasons.
The
major
reason
20SPARROW
SAP,
March
2000
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
31
is
the
general
lack
of
consistent
long­
term
studies
in
which
the
same
sites
are
sampled
over
a
number
of
years.....
The
normal
seasonal
variations
in
concentration,
combined
with
year­
to­
year
variations
caused
by
differences
in
weather
and
agricultural
practices,
make
comparison
of
recent
and
past
concentration
data
tenuous."
16
A
survey
design
conducted
over
three
or
more
years
will
be
more
expensive
than
the
options
mentioned
above,
however,
the
data
could
be
more
easily
used
to
statistically
analyze
year
to
year
variability
from
site
to
site
and
improve
upon
many
existing
studies.

Ancillary
Data
Ancillary
data
are
needed
for
watershed
characterization,
classification
of
watershed
vulnerability,
population
weighting,
assessment
of
water
treatment
processes,
and
development
of
watershed
scale
modeling
approaches
for
source
water
protection
and
dietary
exposure.
Additionally,
ancillary
data
are
needed
to
confirm
or
"
ground­
truth"
the
data
used
in
the
watershed
characterization
tool.

Watershed
characterization
requires
data
on
pesticide
use
including
application
rate,
timing
,
total
pesticide
applied,
types
of
crops,
types
of
soils,
agronomic
practices,
hydrologic
characterization
(
flow
rate),
precipitation;
physical
watershed
properties
including
watershed
size,
topography,
and
proximity
of
the
pesticide
application
area
to
the
CWS.
Several
of
these
properties
including
pesticide
use
intensity
and
areal
extent
of
B
soils
have
been
identified
by
ILSI9
and
USGS20
as
critical
factors
in
predicting
pesticide
run­
off.

Ancillary
data
for
water
treatment
processes
are
needed
to
assess
the
impact
of
water
treatment
on
pesticide
removal
and
transformation.
Additionally,
information
on
the
population
served
is
required
for
population
weighting.
Each
CWS
in
the
survey
requires
identification
of
the
source
water,
geographic
coordinates
of
the
CWS,
type
and
timing
water
treatment
processes,
source
water
quality
information,
flow
rate
through
the
CWS,
and
population
served.
To
date
ACPA
has
not
committed
to
collect
ancillary
data
for
their
study.

EFED's
Proposed
Monitoring
Design
Framework
EFED
is
assessing
design
criteria
for
conducting
a
statistically­
based
survey
to
assess
annual
average
pesticide
concentrations
in
surface
source
drinking
water
for
FQPA
exposure
assessments.
This
section
presents
the
general
thinking
of
EFED
regarding
the
design
criteria
discussed
above.

Assumptions,
Constraints,
and
Biases
It
is
assumed
that
pesticide
exposure
through
drinking
water
occurs
at
CWS's
with
pesticide
application
in
their
upstream
watersheds.
This
assumption
is
based
on
the
premise
that
pesticide
transport
into
surface
waters
is
controlled
through
runoff
and
short­
range
spray
drift.
Such
an
approach
does
not
consider
pesticide
deposition
from
long­
range
atmospheric
transport
including
fog
banks,
wind
erosion
of
sediment,
or
rainfall.
Further,
it
is
assumed:

°
$
10
million
is
available
to
support
a
one
year
monitoring
study.

°
The
minimum
analytical
cost
is
$
500.00
per
sample
for
a
single
analyte
or
a
single
group
of
analytes
(
e.
g.,
USGS
analytical
schedule).
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
32
°
Based
on
a
$
10
million
budget
and
a
$
500
sample
analysis
cost,
the
survey
will
be
able
to
process
no
more
than
20,000
samples.

°
There
are
a
limited
number
of
facilities
capable
of
handling
20,000
samples
per
year.

°
Urban
land
use/
land
cover
density
in
the
basin
can
serve
as
a
surrogate
for
non­
agricultural
pesticide
use.

°
If
there
is
a
substantial
lack
of
cooperation
from
CWS's,
survey
design
requirements
will
not
be
met.

°
There
is
error
in
county­
level
pesticide
use,
and
further
error
may
be
introduced
when
interpolating
any
county­
level
data
to
new
geographic
boundaries
(
i.
e.
watershed
boundaries).

The
survey
needs
to
address
both
the
spatial
and
temporal
variability
of
pesticide
concentrations
in
drinking
water.
Climatic
conditions
(
magnitude
and
timing
of
precipitation),
surface
water
body
type,
watershed
size,
and
pesticide
use
area,
intensity
and
environmental
fate
properties,
are
factors
controlling
pesticide
concentrations
in
surface
water.
The
survey
needs
to
consider
these
variables
in
assessing
the
vulnerability
of
each
CWS
to
pesticide
contamination.
Within
financial
constraints,
the
design
framework
must
be
capable
of
assessing
sub­
populations
of
CWS's
serving
small,
identifiable
human
populations.
Effects
of
water
treatment
on
pesticide
removal
and
transformation
need
to
be
evaluated
in
order
to
assess
pesticide
levels
in
finished
drinking
water.
Also,
pesticide
use
and
watershed
characteristics
must
be
verified
to
facilitate
data
interpretation.
The
survey
results
should
provide
data
for
model
development
and
evaluation,
and
for
identification
of
mitigation
strategies.
The
recommended
data
quality
standard
for
the
survey
is
a
distribution
of
annual
mean
values
across
CWS's
so
that
an
upper
95%
confidence
bound
can
be
placed
on
the
upper
95th
percentile
of
the
distribution
for
the
year
of
the
study.
The
prediction
accuracy
was
selected
to
guard
against
the
cumulative
error
of
assessing
distributions
for
multiple
pesticides
(
ILSI,
1999).
A
total
of
59
sites
per
domain
are
required
to
achieve
this
accuracy,
but
this
number
does
not
guard
against
the
cumulative
error
of
assessing
multiple
compounds.
When
assessing
multiple
compounds,
the
number
of
sites
in
any
domain
could
increase
depending
on
the
designation
of
an
"
acceptable"
minimum
prediction
accuracy
for
describing
a
population
or
sub­
population
in
the
design
framework.

Design
Structure
OPP
believes
a
random
stratified
design
is
an
efficient
sampling
strategy
for
a
national
scale
drinking
water
survey
for
assessing
annual
mean
pesticide
concentrations.
Chemicals
selected
in
the
survey
should
be
selected
according
to
their
relative
risk,
extent
of
use,
and
environmental
fate
properties.
Compounds
with
high
risk
for
chronic
non­
cancer
or
cancer
effects,
high
extent
of
use,
and
which
exhibit
high
mobility
and
persistence
in
terrestrial
and
aquatic
environments
will
be
considered
as
high
priority
pesticides.
The
list
of
priority
pesticides
will
define
the
target
population
for
the
survey.
The
target
population
is
the
CWS's
within
each
of
the
pesticide
use
areas.
This
assumes
that
pesticide
occurrence
in
drinking
water
is
dependent
on
runoff.

The
use
area
of
each
selected
pesticide
will
be
survey
domains.
CWS's
should
be
selected
randomly
from
within
each
pesticide
use
domain
using
watershed
vulnerability
strata.
A
vulnerability
classification
system
needs
to
be
developed
to
establish
the
stratification
system.
The
addition
of
regional
and
vulnerability
sub­
domains
may
be
considered
in
optimizing
the
study
design
framework
.
EFED
proposes
using
a
GIS
site
selection
tool
to
evaluate
which
domains
and
strata
are
used.
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
33
Balancing
the
number
of
sites
and
samples
per
site
requires
consideration
of
the
number
of
domains
in
the
study,
the
minimum
prediction
accuracy
required
in
the
survey,
development
of
an
acceptable
sampling
frequency
for
measuring
annual
mean
pesticide
concentrations,
and
the
water
sampling
strategy
for
raw
and
finished
drinking
water.
The
maximum
number
of
domains,
in
light
of
assumptions,
uncertainties,
and
limitations
in
the
study
design,
is
approximately
16.
This
implies
that16
pesticides
may
be
evaluated
in
the
survey.
The
number
of
selected
pesticides
considered
in
the
survey
is
expected
to
decrease
with
the
addition
of
sub­
domains
in
the
survey,
increased
sampling
frequency,
and
the
addition
of
paired
raw
and
finished
water
samples.
However,
it
is
possible
to
expand
the
number
of
selected
pesticides
in
the
survey
through
relaxation
of
the
minimum
prediction
accuracy
and
accounting
for
common
CWS's
within
the
pesticide
use
areas.

Design
Efficiency
Accounting
for
overlap
of
pesticide
use
in
the
survey
could
increase
the
design
efficiency
through
inclusion
of
CWS's
with
multiple
selected
pesticide
use
in
the
watershed.
However,
EFED
is
concerned
if
CWS
selection
is
solely
based
on
location
in
multiple
pesticide
use
domains
the
CWS's
will
not
be
selected
randomly.
The
selection
of
pesticides
with
similar
national
use
areas
is
expected
to
increase
the
likelihood
of
selecting
CWS's
with
multiple
pesticides
used
in
the
their
watersheds.
The
areal
extent
of
pesticide
use
is
one
factor
considered
in
the
recommended
chemical
selection
process.

Sampling
Frequency
and
Timing
The
sampling
frequency
at
each
CWS
is
an
important
factor
in
measuring
the
annual
mean
pesticide
concentration
in
drinking
water.
OPP
is
recommending
that
14
samples/
year
with
focused
sampling
across
the
pesticide
use
period
is
the
minimum
sampling
frequency
for
assessing
the
annual
mean
pesticide
concentration
in
drinking
water.
This
sampling
frequency
was
selected
as
a
minimum
design
element
because
empirical
evaluation
of
monitoring
data
indicate
14
samples/
year
appears
to
be
reliable
for
assessing
annual
means
for
corn
herbicides
in
static
and
flowing
surface
water
bodies.
However,
a
universal
sampling
frequency
in
a
multi­
pesticide
design
may
not
be
appropriate
because
the
sampling
frequency
should
be
designed
to
capture
concentration
peaks
for
a
reliable
measure
of
annual
mean
pesticide
concentrations.
OPP's
assessment
suggests
sampling
frequency
appears
to
be
linked
to
environmental
persistence;
less
persistent
pesticides
will
require
a
higher
sampling
frequency.
Other
factors
expected
to
impact
sampling
frequency
are
the
flow
rate
of
the
surface
water
body
and
proximity
of
the
pesticide
application
sites
to
the
location
of
the
water
intake.
Flowing
surface
water
bodies
(
such
as
rivers
and
streams)
are
expected
to
have
higher
variability
in
pesticide
concentrations
and
thus
may
need
higher
sampling
frequencies.

Study
Duration
The
recommendation
of
a
one
year
study
is
based
on
financial
constraints,
rather
than
on
scientific
merit.
While
a
one
year
study
can
provide
useful
information,
it
is
not
capable
of
providing
information
on
the
year
to
year
variability.
The
study
could
be
lengthened
by
reducing
the
number
of
pesticides.
However,
this
approach
was
not
chosen
because
it
would
limit
the
number
of
selected
pesticides
in
the
survey
design.

Based
only
on
the
scientific
merits,
the
Agency
recommends
that
monitoring
studies
be
conducted
for
multiple
years.
A
multiple
year
study
will
allow
the
assessment
of
climactic
variability
and
variations
in
pesticide
management
practices
on
the
annual
mean
concentrations.

Chemical
Selection
Drinking
Water
Survey
SAP
Document
File
Name:
SAP
DWS
Consultation
Document­
Final.
wpd
34
OPP
recommends
a
combination
approach
that
considers
the
pesticide
use
pattern,
environmental
fate
properties
of
the
pesticide,
risk,
and
availability
of
analytical
method.
We
believe
this
type
of
approach
will
identify
the
pesticides
that
can
be
appropriately
assessed
with
a
national
survey
and
provide
the
Agency
with
the
necessary
data
for
risk
assessment
and
management
decisions.

Raw
and
Finished
Water
In
light
of
funding
constraints,
OPP
believes
a
minimum
sampling
approach
is
to
collect
paired
raw
and
finished
water
samples
with
analysis
of
all
raw
water
samples
and
fewer
finished
water
samples.
Reactive
analysis
of
finished
water
samples
is
recommended
based
on
the
detection
of
target
pesticide
concentrations
in
raw
source
water.
The
feasibility
of
this
approach
is
dependent
on
the
stability
of
each
pesticide
in
finished
water.

A
preliminary
literature
review
suggests
that
most
pesticides
are
unlikely
to
be
removed
but
may
be
transformed
through
conventional
water
treatment
processes.
However,
granular
activated
carbon
has
been
shown
to
remove
most
pesticides.
Also,
water
softening,
pH­
adjustment
for
corrosion
control,
disinfection
with
chlorine
and
ozonation
have
been
shown
to
remove
or
transform
some
pesticides.
The
ability
to
assess
the
impact
of
water
treatment
processes
in
the
survey
design
is
another
important
variable.

Ancillary
Data
Ancillary
data
are
required
for
watershed
characterization
are
pesticide
use
information
including
the
total
amount
and
application
rate,
types
of
crops,
type
of
soils,
agronomic
practices,
hydrologic
characteristics,
precipitation,
and
physical
watershed
characteristics.
These
data
are
needed
to
assess
mitigation
options
for
source
water
protection.
Information
on
the
water
treatment
processes
including
type
of
water
treatment
processes,
source
water
quality
information,
and
flow
rate
through
the
CWS's
are
needed
to
evaluate
the
impact
of
water
treatment
on
pesticide
removal
and
transformation.
The
population
served
by
each
CWS
in
the
survey
is
also
needed
to
allow
for
population
weighting
pesticide
concentrations
for
the
dietary
exposure
assessment.

Census
for
Major
Facilities
OPP
is
considering
as
part
of
a
design
framework
that
CWS's
serving
large
populations
of
people
(
e.
g.,
New
York,
Chicago,
and
Los
Angeles)
be
automatically
included
in
the
design.
