United States Office of Water EPA 815-R-99-003
Environmental Protection 4607 August 1999
Agency
vvEPA National Representative
Sample of Small Public
Water Systems:
Statistical Design and
State Plans for the
Unregulated Contaminant
Monitoring Regulation
Printed on Recycled Paper
-------
-------
UCMR Statistical Design and State Plans August 1999
[This page left intentionally blank]
-------
UCMR Statistical Design and State Plans
August 1999
Foreword
Under §1445(a)(2)(A) of the Safe Drinking Water Act (SDWA), as amended in 1996, the
Environmental Protection Agency (EPA) is to promulgate regulations for an unregulated
contaminant monitoring program by August 1999. The existing unregulated contaminant monitoring
program has been performed according to the program described in CFR141.40. The 1996 SDWA
Amendments direct a substantially revised Unregulated Contaminant Monitoring Regulation
(UCMR). The revised UCMR has a new list of contaminants and changes the number of public
water systems (PWSs) that must conduct monitoring and the frequency and schedule for monitoring.
This document provides technical background information on the statistical process used to select
the nationally representative sample of small public water systems (that is, those serving 10,000 or
fewer people) that will be included in the revised UCMR Program. The data collected through this
program will be used to support the development of the Contaminant Candidate List (CCL), to
support the Administrator's determination of whether to regulate a contaminant, and to develop
regulations. The revised monitoring program is one of the cornerstones of the sound science
approach to future drinking water regulation that is an aim of the 1996 SDWA Amendments.
-------
UCMR Statistical Design and State Plans August 1999
[This page left intentionally blank]
11
-------
UCMR Statistical Design and State Plans August 1999
Disclaimers
This guidance document is designed to implement national policy concerning this UCMR Program.
The document does not, however, substitute for the SDWA or EPA's regulations nor is this
document itself a regulation. Thus, it cannot impose legally-binding requirements on EPA, States,
or the regulated communiry, and may not apply to a particular situation based upon the
circumstances. EPA and State decision makers retain the discretion to adopt approaches on a case-
by-case basis that differs from this guidance where appropriate. EPA may change this guidance in
the future.
Mention of trade names or commercial products does not constitute endorsement or recommendation
for use.
111
-------
. UCMR Statistical Design and State Plans . August 1999
[This page left intentionally blank]
IV
-------
•*•• •,•..(,.; '•"'•».!,. ,- ; '!
UCMR Statistical Design and State Plans " August 1999
Acknowledgments
This document was prepared in support of the Unregulated Contaminant Monitoring Regulation
(UCMR) for EPA's Office of Ground Water and Drinking Water. Charles Job served as EPA's team
leader for development of the UCMR with James Taft as Targeting and Analysis Branch Chief.
Rachel Sakata served as Work Assignment Manager. The UCMR Work Group provided technical
•guidance throughout, in particular, Andrew Schulman, Christopher Frebis, and Dennis Helsel (of the
U.S. Geological Survey), provided scientific and editorial guidance. External expert reviewers and
miny stakeholders provided valuable advice to improve the UCMR and this document. The major
contributions ofKim Clemente, Jonathan Koplos, Maureen Devitt, and Piyali Talukdar are gratefully
acknowledged. George Hallberg served as Cadmus' Project Manager.
v
-------
UCMR Statistical Design and State Plans August 1999
[This page left intentionally blank]
VI
-------
UCMR Statistical Design and State Plans ' August1999
Table of Contents
Section Page
Foreword \
' j
Disclaimers iii
.. ' i
Acknowledgments v
1. Introduction 1
1.1 Background 2
2. Selecting the Statistical Population for Systems Serving 10,000 or Fewer People 5
2.1 Determining the Population 5
2.2 Stratifying the Population 5
2.3 Tribal Water Systems as an Individual Stratum 7
2.4 Consistency of State Plans 7
3. Selecting the Representative Sample for Systems Serving 10,000 or Fewer People .... 8
3.1 Determining the Size of the National Sample 8
3.2 Selecting a Representative Sample of Systems in Individual States 11
4. Selecting Systems for Initial Plan List and the Replacement List in Each State 12
j
5. Selecting Systems for the State Plan 12
6. Index System Monitoring 13
i
7. Assessment Monitoring 14
8. Screening Surveys 17
9. Pre-Screen Testing 25
10. References 26
Vll
-------
UCMR Statistical Design and State Plans August 1999
[This page left intentionally blank]
vm
-------
UCMR Statistical Design and State Plans
August 1999
List of Tables and Figures
Title
Page
Systems Serving 10,000 or Fewer People 6
Number
Table 1.
Table 2. Population-Weighted National Representative Sample
Distributed by System Size Category and Water Source Type 11
Table 3. Distribution of Index Systems in the Representative Sample 14
Table 4. Allocation of CWSs Conducting Assessment Monitoring with
the Associated Margins of Error for an Estimated Exposure of 1% 16
.IK " i i .
Table 5. Comparison of Confidence Intervals at the 99 Percent Confidence
Level for CWSs 18
Table 6. Allocation of Systems for Screening Surveys by Size Category
with the Associated Confidence Levels and Margins of Error 19
,,'.!', if
Table 7. Distribution of Small Systems Required to Conduct Assessment
Monitoring and Screening Survey in Each State/Tribe/Territory 23
Figure 1. Time-line for National Representative Sample Selection,
Assessment Monitoring, Screening Surveys, Pre-Screen Testing
and other Related UCMR Activities 3
Figure 2. Number and Probability of Small Systems Chosen for Assessment
Monitoring and Screening Surveys for the UCMR Years 22
Appendices
Appendix A. Statistical Formulas A-l
Appendix B. Acronyms B-l
Appendix C. Definitions C-l
IX
-------
UCMR Statistical Design and State Plans August 1999
[This page left intentionally blank]
x
-------
UCMR Statistical Design and State Plans August 1999
1. Introduction
Under Section 1445(a)(2) of SDWA, as amended in 1996, the Environmental Protection Agency
(EPA) is required to establish criteria for a revised monitoring program for unregulated contaminants
and, by August ll, 1999, to publish a list of contaminants to be monitored. To conform to the 1996
Amendments, EPA has proposed substantial revisions in its proposed Unregulated Contaminant
Monitoring Regulation (UCMR), described in the UCMR Preamble and Rule (64 FR ). The
purpose of the UCMR Program is to collect unregulated contaminant occurrence data to help
determine which of those contaminants EPA should regulate based on the concentrations of the
contaminants) in public water systems and their potential adverse health effects through
contaminant exposure.
This proposed rule will take the place of the regulations currently in 40 CFR §§141.35,141.40, and
142.15(c)(3) and modify § 142.16. The revisions cover the following: (1) the frequency and schedule
for monitoring based on public water system (PWS) size, water source, and likelihood of finding the
contaminants; (2) a new shorter list of contaminants to be monitored, (3) procedures for selecting
and monitoring a national representative sample of PWSs serving 10,000 or fewer people, and (4)
procedures for placing the monitoring data in the National Drinking Water Contaminant Occurrence
Database (NCOD), as required under Section 1445. The data generated by this rule will be used to
support the development of the Contaminant Candidate List (CCL), the Administrator's
determinatioijpfwhetherornotto regulate a contaminant, and to develop drinking water regulations.
The proposed revised UCMR Program is a cornerstone of the sound science approach to future
drinking water regulation, which is one of the aims of the SDWA Amendments.
„ , i'i,,' •' ' • • " ' „ ' "• I
The purpose of this document is to describe the general statistical design, rationale, and specific
rnethods used to select the representative sample of small systems that are required to undertake
UCMR monitoring. Portions of this document also describe how this process relates to the UCMR
monitoring plans of individual States. Under the UCMR, the listed unregulated contaminants will
be monitored between 2001-2005. All large PWSs (systems serving more than 10,000 people) are
required to monitor for these unregulated contaminants. Section 1445(a)(2) of SDWA mandates that
only a representative sample of small PWSs (systems serving 10,000 or fewer people) may be
required to monitor under the UCMR. The representative sample must be of adequate size and
quality to obtain the necessary and valid contaminant occurrence information upon which to base
regulatory determinations while minimizing burden to the water system.
The objective of the statistical approach for the UCMR is to estimate contaminant exposure and
occurrence in a nationally representative sample of small systems which will enable extrapolations
of exposure and occurrence nationwide. For contaminant exposure assessments (that is, the fraction
of population that is exposed to a contaminant), the representative sample design is primarily
population-weighted. However, information on contaminant occurrence is also necessary. The
context of occurrence (for example, the size of a water system or its water source) is a factor when
evaluating potential future regulatory implementation. Therefore, the representative sampling design
incorporates a stratified sampling approach and allocates some samples among strata to enable
evaluations of occurrence relative to system size (based on population served), water source type
(surface water or groundwater) and, to some degree, geographic distribution. Although this
statistical design is not strictly optimal for either exposure or occurrence, the design still meets the
data quality objective criteria for exposure estimates (99% confidence level with ±1% error, at 1%
exposure) while providing important occurrence estimates for categories of small systems.
-------
UCMR Statistical Design and State Plans , August 1999
1.1 Background
There are three major components of the proposed UCMR Program: Assessment Monitoring,
Screening Surveys, and Pre-Screen Testing. Assessment Monitoring will be conducted in the first
three years (2001 -2003) of the five-year UCMR cycle (2001 to 2005), and will be conducted for the
contaminants with currently available analytical methods. The Assessment Monitoring contaminants
are listed in § 141.40(a)(3) Table 1, UCMR (1999) List 1. Assessment Monitoring will be required
for all large systems and a representative sample of small systems. A statistically representative,
stratified random sample will be used to select 800 small systems to undertake Assessment
Monitoring (See Section 7). Analytical monitoring results from Assessment Monitoring will be used
for evaluations of contaminant exposure and occurrence.
Screening Surveys will be conducted by a statistically selected subset of the large and small systems
conducting Assessment Monitoring. The Screening Surveys (see Section 8) will 'be conducted
between2001 and2003 for contaminants listed in § 141.40(a)(3) Table 1, UCMR (1999) List 2. The
analytical methods for these contaminants currently are being refined. Analytical monitoring results
from Screening Surveys will be used for evaluations of contaminant occurrence in water systems.
Screening Survey results will help guide decisions of whether or not to move List 2 contaminants
to List 1 for Assessment Monitoring.
Pre-Screen Testing will be conducted by up to 200 small and large systems that are determined to
be most vulnerable to occurrence of the contaminants listed in § 141.40(a)(3) Table 1, UCMR (1999)
List 3. The Pre-Screen Testing (see Section 9) will be conducted in 2003 or 2004. The analytical
methods for the Pre-Screen Testing contaminants currently are in the early stages of development
Pre-Screen Testing will be conducted to determine whether the analytical methods in early
development can provide adequate results hi conditions under which the contaminants are most
likely to occur.
Figure 1 illustrates the time-line for the national representative sample selection, Assessment
Monitoring, Screening Surveys, Pre-Screen Testing and other related UCMR activities.
Under SDWA and the UCMR, States can develop State Monitoring Plans (State Plans) for small
system monitoring. Discussions with States and other stakeholders indicate the need to select a
representative sample of systems across all States to ensure both confidence in the UCMR results
and a comprehensive spatial distribution. To ensure that the sample is representative of the nation
and to reduce the burden on the States, EPA will statistically select a nationally representative
sample of systems serving 10,000 or fewer people for the UCMR. To further reduce the burden on
States, EPA will directly implement the UCMR.
The States can participate hi the program through the State Plan as established by Memorandums
of Agreement (MOAs) with EPA. The State Plans will be established by States that enter into
MO As. Through the MO As, States take a partnership role hi the development of the State Plan and
implementation of the UCMR. All steps involved with sample selection described throughout this
document assume that a State has entered into an MO A with the appropriate EPA Regional Office.
As described later in this document, a list of the statistically-selected systems will first be provided
by EPA to the States. The list will be comprised of an "initial list" and "replacement list" of systems.
These lists will be provided to the States for then- review and inclusion hi their State Plans. States
can either (1) respond by accepting the primary list as their representative plans, or (2) propose an
alternative plan by selecting other system(s) from the replacement list, in cases where EPA's initial
-------
U£MR Statistical Design and State Plans
August 1999
Figure 1. Time-line for National Representative Sample Selection, Assessment
Monitoring, Screening Surveys, Pre-Screen Testing and other Related UCMR
Activities.
1999
2000
2001
2002
2003
2004
2005
UCMR (1999): IMPLEMENTATION ACTIVITIES
UCMR Issued:
Guidance
Available
State MOAs
and State
Plans
Developed:
Inform PWSs
Representative
Sample selected
by EPA
National
Contaminant
Occurrence
Database
Operational
EPA Contract
Laboratories
Operational
Assessment Monitoring:
List 1 (1999) Contaminants
All 2,774 Large and 800 Small PWSs
Index System Monitoring
"Up to 30 Small PWSs (2001-2005)"
Screening Surveys:
List 2 (1999) Contaminants
2 Groups of 300 Large and Small PWSs
Screening Survey
Group 1
(2001 or 2002)
Screening Survey
Group 2
(2003 or 2004)
Pre-Screening Testing:
States Specify
Vulnerable Systems
Pre-Screen Testing:
List 3 (1999)
Contaminants
Up to 200 Large
and Small PWSs
(2003 or 2004)
Analyze Results
'and Data Quality"
Next ^ Next UCMR
Candidate List Issued
Contaminant
List Issued
-------
UCMR Statistical Design and State Plans August 1999
plan identifies system(s) that no longer exist, because of merger or closure, or have switched to
purchased water.
One-third of the systems in each State Plan will be sampled each year over the three-year
Assessment Monitoring period. EPA will pay for the testing costs for the Assessment
Monitoring of small systems (and will also pay for the costs of the Screening Survey and Pre-
Screen Testing programs) as long as the systems are part of the State Plans. EPA may also
modify the size of the representative sample to reflect available funds.
Additionally, EPA will select up to 30 small PWSs as "Index" systems and will monitor these
systems each year during the five year UCMR Listing Cycle. Index systems are a subset of the
national representative sample and will be selected randomly from the sample. Index system
monitoring will provide added quality control and more detailed information on temporal
variations in contaminant occurrence, and on environmental and operating conditions of small
systems so that future regulations can better reflect small system characteristics and conditions.
Further information about the UCMR Program can be found in the UCMR Preamble of the final
Rule (64 FR ), and supporting guidance and technical documents. These documents are
available from the EPA Water Docket, (202) 260-3027, Docket Number W-98-02. General
information can also be obtained from the EPA Safe Drinking Water Hotline, (800) 426-4791, or
through the EPA Office of Ground Water and Drinking Water Internet Home page at
www.epa.gov/ogwdw.
-------
•' ;ff"
UCMR Statistical Design and State Plans August 1999
2. Selecting the Statistical Population for Systems Serving 10,000 or Fewer People
III,I, , ,! Jll' i ! I
2.1 Determining the Population
• • . i • "•••' - • •' ... i
The total population of small PWSs is comprised of community water systems (CWSs), non-
transient non-community water systems (NTNCWSs) and transient non-community water systems
(TNCWSs). Two categories of PWSs are being excluded from the population for selecting the
sample. PWSs that purchase their entire water supply from another PWS are generally exempt from
the regulation, since monitoring at these systems could result hi double counting of systems using
the same source.1 Additionally, TNCWSs will be excluded from the UCMR, since projecting
contaminant exposure from monitoring results is difficult and inconclusive due to the transient
nature of the population that use these sources of drinking water.
EPA estimates that there are approximately 65,636 non-purchased CWSs and NTNCWSs, based on
the 1997 Safe Drinking Water Information Systems' (SDWIS) PWS inventory.2 Table 1 illustrates
the national number of non-purchased CWSs and NTNCWSs serving 10,000 or fewer people in each
size category (serving 25 to 500,501 to 3,300 and 3,301 to 10,000 people) by the source water type
(ground or surface water), from the SDWIS inventory as of January 1997. These systems provide
an example of the statistical population from which the national representative sample will be drawn.
The actual sample used for UCMR sampling will be derived from the most current SDWIS inventory
available at the time of sample selection, so the actual number of systems in Table 1 may change
slightly.
2.2 Stratifying the Population
In developing the representative sample, EPA must consider factors such as (1) geographic location,
(2) population served, and (3) water source. The UCMR accomplishes this at various levels by
stratifying the sample by categories of population served, allocating samples proportionately to each
State by system size, and then by water source (surface water and ground-water supplied systems).
NTNCWSs are selected as a separate category since these systems may be a significant source of
water consumed by residents of a community.
Sources of water may not be evenly distributed across any given State. Cities transfer water across
watershed boundaries, or move water from one State to another. To account for the proportion of
the population served by a specific type of water source (surface or ground water), EPA proposes
to define "geographic location" in the representative sample as the location of the water source and
stratify the sample further by source of water supply (ground and surface water). For example, if
10 percent of the population in a State obtains their water from surface water supplied PWSs that
1 Note that purchased water systems may be required to monitor UCMR (1999) List 2 and UCMR (1999)
List 3 microbiological contaminants where the system is considered to be the distribution line with the
maximum residence time.
2 For the purposes of estimates in this report, the SDWIS January 1997 inventory of PWSs is used. EPA
will re-compute the sample size at the time of actual statistical sample selection using the most current
inventory.
-------
oe
CO
•s
a,
e
o
I
s.
es
H
00
O
II
c S
12
**^
O
I
I
en
CO
O
"Sb
s
a,
o
i
Nationa
V
CO
o
^
1
o
CO
IB
3 ti
§'
1!
I
6C
1
CO
I
CO
1-H
cu
os
CO
CO
O
CO
O
U
oe
oo
Os
»— i
C4
*\
-
«\
\o
oo
CO
CO
*v
CN
ON
CN
O
•*
o
VO
CO
vo^
•^
o
o
in
rs
CS
S
CO
t~~
«s
ON
ON
ON
O
CM
ON
CN
03
ON
0)
O
O
CO
co"
CN
ON
ON
CO
00
cs
oo
O
CO
oo
NO
CN
ON
vo
ON
NO
1 0
-H O
00
VJ
XO
09
00
oo
®«
ON
09
i—I
ss
oo
fO
ON
CO
ON
a\
CO
oo
NO
co
t>-
CN
oo^
O
CN
vo
o
CN
CO
O^
CN"
ON
co^
to"
o
o
OO
NO
oo
l-H
NO
oT
VJ
T-l
eC
in
Vi
v
vo
^
VO
fO
¥-1
eT
VO
oo
oo
oo
e— g
llf
£•» fe*
*- C I-
"GO &0 ^
o-g c
ill
S?J
•Ceu.2
.§11
lit
11!
g s S
f Tn O
2.2-S
111
o.S <""O
!§a3
g ra Q> TQ
11II
llil
fill
.3
ll'll
r
— ^ 2 = QJ
•£ 2 S m
l|s*
S H
-S S
Sil
to 3_
' "3—
-------
LfCMR Statistical Design and State Plans " August 1999
serve less than 500 individuals, then 10 percent of the sample systems in that State should come from
the PWSs in this size and source category. The distribution of systems across the State, therefore,
is accommodated by the population-weighted statistical sample selection.
Therefore, the sample is stratified by system type (C WSs and NTNC WSs) and by source water type
(ground water and surface water) within each small system size category (categories 1 through 3)
in each State.
ii
23 Tribal Water Systems as an Individual Stratum
„ ,, :|
Small PWSs that are located on Tribal lands in each of the ten EPA Regions will be grouped into
a single category for the representative sample; this Tribal category is equivalent to a State for the
statistical selection process. Tribal systems will have the same probability of being selected as other
water systems in the stratified random selection process that weighs systems by water source and
size class by population served. Using this discrete stratum ensures that the Tribal systems are
selected as part of the national representative sample. The systems selected will comprise the "State
Plan" for Tribal water systems.
2.4 Consistency of State Plans
: , ... ,; . 1
... ( .'; ',: ' • "' . 1
EPA will select the representative sample from the population of CWSs and NTNCWSs nationally,
then the sample will be allocated on a State-by-State basis, weighted approximately for the
proportion of population served by each service size category and water source type. Based on a
stratified random selection process applied to CWSs and NTNCWSs, the sample size is weighted
by population served (to enable exposure assessments from Assessment Monitoring results) and
water source type (to enable comparisons between surface or ground water) while allocated
proportionately amongst States (to ensure geographic coverage) within service size category
(categories 1 through 3). EPA will select two to three times as many CWSs and NTNCWSs as
required for a national representative sample. These systems will appear on an "initial plan" list and
a "replacement" list. The initial plan list of systems will identify those systems selected for each
State.
States can include the EPA-selected systems on the initial plan list hi their State Plan. If, however,
the State Review determines that a system on the initial plan list has closed or merged, the system
can be removed from the State Plan List. To remove a system from the State Plan List and replace
it with another, the State must notify EPA of the reasons for removal. Valid reasons for removal
include system closure, system merger, or a determination that a system operates exclusively with
purchased water.3 To identify a replacement system for the system removed, States will select the
first water system (from the appropriate category) from the existing replacement list. (See Section
4 for a more detailed discussion of initial plan and replacement list selection procedures.)
When the list of systems is finalized, States will inform the EPA Regional Office of the States'
choice of plans (including the details of any modified plans). As needed, the EPA Regional Office
will work with the State to develop an acceptable modified plan. The State Plan will include a
Purchased water systems may be required to monitor for UCMR microbiological contaminants that occur
primarily in distribution lines.
-------
UCMR Statistical Design and State Plans August 1999
process for the State to inform the systems of their selection as part of the representative sample and
the systems' responsibilities related to their participation in the UCMR. This approach ensures a
nationally consistent system selection process (whether a State accepts the initial plan list or
generates a modified plan list of systems to sample) and enables acceptable plan development with
minimal State burden.
If the EPA Regional Office does not receive the notice of a State Plan within 60 days, EPA will
assume that systems on the initial plan represent the State Plan. The plan will also specify the timing
of the monitoring. The State may adjust this time in the Plan to coordinate with compliance
monitoring, if appropriate. States may also designate the most vulnerable period for sampling of
systems in their jurisdiction. States will also be asked to nominate systems vulnerable to UCMR
(1999) List 3 contaminants for Pre-Screen Testing, though this may be done in a separate request.
3. Selecting the Representative Sample for Systems Serving 10,000 or Fewer People
3.1 Determining the Size of the National Sample
The general population of small systems from which the representative sample will be drawn is
found hi Table 1. The data from the representative sample must be of high quality to estimate
national occurrence and exposure, and to be dependable hi guiding development of possible
regulatory alternatives. In the UCMR, national occurrence is defined as the fraction of systems
where a contaminant occurs, and exposure is defined as the fraction of the population exposed to a
contaminant. Hence, the first step hi designing the sample is to define the acceptable or allowable
measurement error by setting precision and confidence levels. Two facets of the allowable
measurement error are described below.
The first type of potential error describes the precision or tolerance. Precision reflects the
understanding that the sample value may not precisely reflect the value for the population as a whole.
EPA will use a one percent allowable error level, i.e., the (estimated) sample value will fall within
one percent (above or below) the true value for the population as a whole. This allowable error is
also known as the confidence interval.
The second type of error may be expressed using the confidence level. The confidence level
describes the probability that the sample results will be within the range set by the precision figure.
For example, EPA will use a confidence level of 99 percent, which suggests that 99 times out of 100,
the (estimated) sample results will fall within one percent of the true population.
EPA has specified these stringent statistical parameters to ensure high quality data and dependable
monitoring results. In general, many similar random surveys with continuous variables use a lower
level of confidence (95 percent) and/or a larger allowable error (plus or minus 5 percent). However,
use of a larger possible error is unacceptable for this program. Examination and analysis of current
occurrence data shows that many contaminants that are currently regulated, or being considered for
regulation, occur in one percent or less of systems on a national basis. However, for many
contaminants, a one percent occurrence nationally reflects a substantially larger occurrence
regionally. Even a small percentage of systems with detections of a contaminant can translate into
exposure of a significant population. By accepting a greater margin of error, and the resultant
smaller sample size, such small national occurrence might be missed entirely.
-------
UCMR Statistical Design and State Plans August 1999
EPA must also have the capability to evaluate contaminant occurrence hi relation to source waters
and different size categories of systems. Many statutes and current regulations are implemented
differently for systems of different size, or for different source water categories. Combining the
representative (small system) sample with the results from all large systems provides increased
power hi the total sample, but EPA must also be able to evaluate occurrence, and possible regulatory
options, related to the small systems themselves. SDWA and many current rules focus on burden
reduction for small systems when feasible.
There are also other uncertainties and sources of variance in such a sample program. For example,
all contaminants have censored distributions (i.e., "less than the detection level" analytical results)
and there are a myriad of factors that affect variability and vulnerability of ground water systems.
It is not clear how normal sampling theory accommodates some of these sources of variation.
Hence, the high confidence level, low allowable error, and larger sample size should help to ensure
adequate data to meet the objectives of the UCMR Program.
To, define the sample size required to meet the values for allowable error and confidence level, the
occurrence and variability within the population must also be accounted for. The variance is, of
course, unknown before the survey is conducted, but it can be estimated within various bounds. In
the following discussion, p will represent the true proportion of systems with contaminants, and 1 -
p is the true proportion of systems without any contaminants.
• mi ,N :, M'i j R, ii ' ' i1 ih - ' ' :i
Given a population with the true proportion of systems with contaminants assumed asp, the sample
size (ri) required to produce the specified confidence level and allowable error (d) estimate of the
population mean is,4
K_z2*p(l-p) (1).
d2
Here z is the critical value at 99 percent confidence, taken from the table of the normal distribution.
The underlying assumptions of the approximation used to derive equation (1) are: (1) that we take
a simple random sample from the population of systems; (2) that the sample is large enough for a
normal approximation to hold; and (3) that in each system we can determine with certainty whether
or not a contaminant is present. These assumptions are only approximately correct, so the sample
size calculation is also only approximately correct. The more complicated stratified sampling plan,
as opposed to the simple random sampling plan assumed here, is taken into account when confidence
intervals are calculated in Section 7.
The largest value of the product p(l -p) is 0.25 occurring when/> = 0.5. In other words, without
precise data on each system, the most conservative assumption least likely to underestimate the
potential error is to use an estimate of 50 percent variability (p = 0.5). This results in the largest
value for » (with other factors held constant). However, analysis of regulated and prior unregulated
4 The normal approximation to the binomial distribution was used. For rare events (e.g., one percent
occurrence) the Poisson distribution may provide a better approximation. However, as sample size
increases the differences diminish. The sample size estimated for the UCMR is great enough that there is
no substantive difference hi the estimated «.
-------
UCMR Statistical Design and State Plans _ August 1999
contaminant monitoring occurrence results indicate that only a very small proportion of systems on
a national level are likely to exhibit any contaminant occurrence. Hence an estimate of 1%
occurrence, or p = 0.01, was selected. This assumption is believed to be appropriate and accurate
while balancing cost and logistical considerations with sample size.
Given the standard normal distribution, the critical value for z at the 99 percent confidence level is
2.58. The one percent margin of error (<£=0.01) suggests:
sample fraction - .01 < population fraction < sample fraction + .01
The necessary sample size is then calculated as:
(2)-
(0.01)2
Thus, for the population of 65,636 small systems, a representative sample size of approximately 659
systems will provide a confidence level of 99 percent with an allowable margin of error of ± 1
percent. However, to provide for a broad geographic coverage, a portion of the sample needs to be
distributed among all States. EPA will specify the stratified sample process so as to ensure a
minimum of two representative systems allocated to each State. It is important that all States, and
at least several systems in each State, contribute to UCMR occurrence results. These results will be
proportionate to the population served by PWSs within each system size category and by water
source type (surface or ground water). Contaminant occurrence results will be proportionate to
population served to enable EPA decision making based on exposure assessments.
Given the potential small individual State sample size, no statistically valid conclusions may be
drawn at the State level. However, EPA still considers it important that all States are represented
and have the opportunity to participate in the UCMR State Plan. Some contaminants, such as some
pesticides, may only be used intensively in specific regions of the country. It is possible that with
the relatively small number of systems in the representative sample, monitoring may miss
contaminants with such targeted regional use patterns. However, including systems in every state
in proportion to the population served should ensure that contaminants with regional use patterns,
to the extent that they potentially contaminate water supplies, are proportionately represented by the
national sampling design. These factors were also considered in selecting the number of systems
for the national representative sample.
EPA proposes to use a sample size of 72 1 CWSs and 79 NTNCWSs for a total of 800 small systems
for the representative sample. This sample size was selected for various statistical and budgetary
considerations. A sample size of 721 CWSs is more than the minimum 659 CWSs needed to ensure
a 99 percent confidence level, (see Equation 2 above) and allows sampling of at least two CWSs in
each State. The number of NTNCWSs selected for inclusion in the sample reflects the proportion
of the NTNCWSs, and the proportion of the population served by NTNCWSs.
Table 2 illustrates the composition of the nationally representative sample of 800 systems by system
size and type (CWSs vs. NTNCWSs and ground water vs. surface water). In order to gather more
information about the systems in the very small category of CWSs (Category 1) and improve the
statistical power of results hi this category, the number of very small ground water systems was
increased by 6, and the number of very small surface water systems was increased by 56, compared
to an allocation that would be strictly proportional to the population served by these systems. Due
10
-------
UCMR Statistical Design and State Plans
August 1999
to the small sample size of the NTNCWSs hi the ground water and surface water categories within
each size category (Categories 1 through 3), statistical conclusions about NTNCWSs must be
analyzed with caution. Conclusions about NTNCWSs cannot be based on source water type since
the margin of error would be too great.
Table 2. Population-Weighted National Representative Sample Distributed by
System Size Category and Water Source Type
Size Category
(by population
served)
Category 1
Category 2
Category 3
500 and
Under
501 to
3,000
3,001 to
10,000
Total
Number of
CWSs
Ground
Water
67
186
189
442
Surface
Water
64
74
141
279
Number of
NTNCWSs
Ground
Water
33
36
4
73
Surface
Water
2
3
1
6
Subtotal of AH
Systems by Water
Source Type
Ground
Water
100
222
193
515
Surface
Water
66
77
142
285
Total
166
299
335
800
CWS * Community Water Systems, NTNCWS = Non-Transient Non-Community Water Systems
Note: The population-weighted distribution of samples indicated above is based on population and water system information from
the 1997 SDwIS database inventory. The table above provides an illustration of the anticipated approximate distribution of water
systems selected for the national sample as based on population served by system size category, water source type, and water system
type. When the actual representative sample selection process is conducted, the most current population data will be used. The
actual, final distribution of the number of systems by category will change based on the updated population inventory, but the total
sample number of 800 systems will remain unchanged.
3.2 Selecting a Representative Sample of Systems in Individual States
Given the UCMR Program's requisite 99 percent confidence level (± 1 percent margin of error) and
the other considerations discussed, the total number of systems hi the representative sample (800)
is then allocated approximately in proportion to the population served, by source water type and
equally across States, Territories, and Tribes, by three system size categories (Table 2).
Approximately 62 extra systems were added to Category 1 to increase statistical power, therefore
the sample is not strictly allocated by the proportion of the population served. In each State, the total
number of systems is allocated in proportion to the population served in the State. For example,
presume hi State Y, 156,718 individuals are served by ground-water supplied CWSs in category 1.
The number of individuals served nationally in this size category is estimated as 4,636,537 (Table
1). The number of systems hi this category in the national sample is 66. Therefore, in category 1,
the number of ground-water CWSs required in the representative sample, x, in State Y is calculated
as:
11
-------
- UCMR Statistical Design and State Plans August 1999
xa 156'718 *66=2.23^2 (3).
4,636,537
The number of systems in each State will be rounded so that the total number of systems in the
survey equals 800, and so that no State is represented by less than 2 systems. The procedure is
repeated for all system size and source water type categories to estimate the total number of systems
hi each State.
4. Selecting Systems for Initial Plan List and the Replacement List in Each State
EPA will choose the PWSs for the national sample through a stratified random selection process.
For each system size category, stratified by system type (CWSs and NTNCWSs) and source water
supply (surface or ground water), EPA will: (1) list and systematically assign a number to each of
the systems within these groups for each State/Tribe; and (2) generate a random sequence of uniform
integers between one and the total number of systems hi that class, Using a uniform random number
generator.
The initial plan list includes the actual number of samples initially allocated to a State through the
statistical random selection process. The replacement list is a list of systems comprised of additional
randomly-selected systems that may be used to replace systems on tile initial plan list if necessary.
Replacement of a system on the initial plan list may occur only when the initial plan list system has
closed, merged, or is identified as purchasing all of its water supply. Note that in the case of
microbiological sampling, purchased water systems may remain on the initial plan list. EPA will
develop guidance on the use of purchased water systems in conjunction with guidance on monitoring
for UCMR (1999) List 2 and UCMR (1999) List 3 contaminants.
To illustrate system selection, we can continue with the example from Equation 3. State Y has 494
ground-water supplied small CWSs in Category 1. The 494 systems are listed hi order by PWSID
and are numbered from 1 to 494. According to the proportion of population in State Y served by this
system-size category and type, two systems will be selected for UCMR monitoring from this group
(Equation 3). EPA will randomly select two to three times this required number of systems to
provide a replacement list for each size category for each State. The random number generator
provides a list of six numbers: 26, 294, 48, 137, 303, and 2 (with the numbers corresponding to
PWSID numbers). Hence, systems number 26 and 294 (i.e., the first two system numbers randomly
selected that fit the required size and type categories) will be the two systems selected for State Y's
national representative sample initial plan list. The four additional systems .randomly selected (i.e.,
systems numbered 48,137,303, and 2) will be placed on State Y's replacement list.
5. Selecting Systems for the State Plan
Each State, tribe and territory will have 60 days to review the initial plan list. The State/Tribe will
either: (1) accept the selections as its State Plan and notify the Regional Administrator of its
acceptance along with its procedure to inform the selected systems of their responsibilities for
monitoring; or (2) propose deletions from the initial plan list and select alternates from the
replacement list as its State Plan, including the reasons for the changes, informing the Regional
12
-------
UCMR Statistical Design and State Plans August 1999
Administrator of the proposed changes; or (3) take no action within 60 days, allowing the Regional
Administrator to specify the portion of the representative sample applicable to the State as its State
Plan. In the second case, the Regional Administrator will have 60 days to work with the State to
develop a suitable plan, if problems are encountered. (For further information on specific guidance
on developing State Plans, please refer to EPA's Unregulated Contaminant Monitoring Regulation:
Implementation Guidance for States.*)
Any system(s) removed from the initial plan list must be replaced by the system(s) selected next in
sequence on the replacement list, as they are ordered by random selection. Continuing with our
example in State Y, systems 26 and 294 were randomly selected as State Y's two representative
systems (again, according to the required size and water type categories) to be included on the State's
initial plan list. However, State Y verifies that system 26 was taken out of service within the
previous year. The State then chooses the first system on the replacement list, in this example,
system number 48, to replace system number 26. The modified State Plan list is then comprised of
systems 48 and 294. If system 294 was also not suitable, system 137 (next hi the replacement
sequence) would become its replacement; if 137 was not suitable, then system 303 would become
the replacement, and so forth.
In summary, each State/Tribe reviews its portion of the national representative sample to determine
that the systems selected have the appropriate operational status. The State/Tribe submits its
representative sample listing to the EPA Regional Office, with all changes from the initial list
marked and the reasons for any changes noted. This list is included in the State Plan, and becomes
a portion of the national representative sample.
States/Tribes may also sample additional systems. However, any additional sites sampled by States
would not be combined with those of the representative sample for the purpose of computing
national estimates of contamination. EPA cannot pay for the testing of these additional systems.
These additional systems, though providing useful information, will bias the national set of systems
if included with those selected using the stated national criteria. However, if the States provide the
results of such monitoring, EPA will receive the data through SDWIS for input to the National
Drinking Water Contaminant Occurrence Database (NCOD).
6. Index System Monitoring
EPA will identify up to 30 systems from the representative sample to be "Index" systems. The data
collected from the Index systems will be used partly for added quality control and to better
characterize monitoring results and operating characteristics of small systems. These systems will
be monitored every year for five years (one complete UCMR cycle). This will provide some detailed
information regarding temporal variations during the course of the UCMR monitoring, as well as
possible effects related to operational changes. EPA will pay for this monitoring, including
provisions for sampling equipment, labor for sample collection, shipment of samples, testing and
analysis. Additional water quality and operational data from these systems will also be collected at
the same time, with minimal burden to the systems. The Index systems will be selected hi
proportion to the population served in each size category and water source type. Among each system
size category and source water type, systems will be chosen randomly from the national
representative sample used for Assessment Monitoring. Table 3 illustrates the number of systems
cbxjsen in each size category as Index systems from the representative sample.
13
-------
UCMR Statistical Design and State Plans
August 1999
Table 3. Distribution of Index Systems in the Representative Sample
Size Category
Ground Water
500 and Under
501 to 3,000
3,001 to 10,000
Number of
Non-Index Systems
96
213
185
Number of
Index Systems
4
9
8
Surface Water
500 and Under
501 to 3,000
3,001 to 10,000
Number of Systems
in the Representative
Sample
66
74
136
770
0
3
6
30
Note: The distribution of samples indicated above is based on the 1997 SDWIS database inventory.
The table above provides an illustration of the anticipated approximate distribution of sample systems
based on population served by system size category. When the actual representative sample selection
process is conducted, the most current population data will be used. The actual, final distribution of
the number of systems by population size category will change based on the updated population
inventory, but the total sample numbers by system type (non-index vs index), as illustrated above,
will remain unchanged.
7. Assessment Monitoring
The UCMR requires all 2,774 large and 800 small CWSs and NTNCWSs (the national
representative sample) to monitor for the contaminants for which established analytical methods are
available. This first stage of the UCMR is called Assessment Monitoring and includes twelve
organic chemicals (listed in§141.40(a)(3) Table 1, UCMR (1999) List 1) for monitoring. EPAplans
to have one-third of the representative sample (267 systems) monitor in each of the three Assessment
Monitoring years (2001 to 2003). This distribution of sampling effort is designed to facilitate
laboratory scheduling and other logistical considerations. The small systems will be delegated to
a sampling year by random selection with a 33 percent probability that each system will be selected
in each of the three years. As stated earlier, Assessment Monitoring is being conducted to assess
exposure to contaminants, as well as contaminant occurrence. An exposure assessment estimates
the fraction of the population that is exposed to a given contaminant, while contaminant occurrence
estimates the fraction of systems in which a contaminant occurs.
After the sampling year is selected, each system will be assigned specific weeks during the year
when theur samples will be collected, with four sampling tunes per year for surface water systems
and two per year for ground water systems. One sampling period must be during the most
vulnerable period, as designated in the regulation. The specification of year-month-week will not
14
-------
. I;
UCMR Statistical Design and State Plans _ August 1999
only facilitate scheduling of laboratory resources, but will ensure that sampling covers vulnerable
periods and all seasons to assess some aspects of temporal occurrence patterns. To provide States
with flexibility in determining vulnerable periods, EPA is allowing the vulnerable period samples
to be taken within 2 weeks (plus or minus 2 weeks) of the scheduled sampling date. EPA is also
allowing the second ground water sample to be taken within 5 to 7 months of the initial vulnerable
period sample. While Index systems sample during all five years of the UCMR cycle, each Index
System will also be assigned an "official" sample year. Only the data from the official sample year
will be used in the national summary of results from the Assessment Monitoring, for consistency
with the sample design.
The UCMR cannot specify any particular year for Assessment Monitoring for the large PWSs, but
does specify that they must conduct their monitoring within the first three years (200 1 -2003) of the
UCMR cycle. EPA expects that large system UCMR monitoring for unregulated contaminants
coincide, whenever possible, with required monitoring for regulated contaminants. Since monitoring
schedules for regulated chemicals depend on system size and detection history, compliance
schedules vary significantly. EPA recognizes that although it will be desirable to collect UCMR
samples concurrently with compliance samples for regulated chemical contaminants, sometimes it
may be difficult to coordinate the two sampling events. Large systems are required to bear the costs
of sampling, testing and reporting the results, and coincident monitoring will help reduce the burden.
' '. • ' • • t ,;' ' i
Since all the systems in the representative sample (i.e., 800) will be required to conduct Assessment
Monitoring, the allocation of systems follows the example given hi Equation 3. More generally, the
number of systems conducting Assessment Monitoring (A ni) in category n in State/Tribe i, is
calculated as:
(4).
NP
n
Wjiere Pni is the population served by system category n in State/Tribe i, and NPn is the total national
population served by systems in this category, 7n is the number of systems allocated to category n.
Table 4 illustrates the total number of CWSs conducting Assessment Monitoring, the sample number
(n) distribution according to system size, and the associated margins of error at 99 percent and 95
percent confidence levels. Note that Table 4 does not include data for NTNCWSs. Given the small
number of samples allocated to the NTNCWSs (79), statistical inferences specific to this system-type
category can riot be made under the necessary confidence levels and margins of error. Also, there
is a potential overlap between the populations served by CWSs and NTNCWSs. Therefore, given
the possibility of potential double-counting of populations served and the related complexities of
appropriately estimating population exposure to identified contaminants, the monitoring results for
the NTNCWSs will not be combined with the CWSs' for national exposure estimates. However,
the monitoring results for the NTNCWSs will, nonetheless, provide important qualitative
contaminant occurrence information for populations served by those systems and in preliminary
comparisons to CWSs' monitoring results.
" ..... '! < " , ' . , 1 ' ".
The information in Table 4 indicates that results from Assessment Monitoring of representative
sample systems can be extrapolated for national occurrence and exposure estimates within
reasonable confidence levels (99 or 95 percent) and margins of error (±1 to ±2 percent) for: (1) all
small CWSs, and (2) all ground water- and surface water-supplied systems. However, within each
15
-------
UCMR Statistical Design and State Plans
August 1999
size category, the sample size can be as low as 64 systems, raising the margin of error of estimation
slightly (to ±3 percent) for some subcategories of the sample. As noted elsewhere, results can not
be used to estimate occurrence at the State level.
The error ranges in Table 4 are computed by the statistical formulas in Appendix A, using the normal
approximation shown in Equation (1). This approximation is valid when n and/? are large enough.
For example, Casella and Berger (1990) recommend n*pz5, while Parzen (1960) recommends
n*pzlQ. In Table 4, where n = 721 and/? = 0.01, n*p = 7.21. By this measure, the normal
approximation may not be valid. However there is a clear problem with the normal approximation
in Table 4: some of the error bounds are so wide that they include negative occurrence fractions
within the margin of error. For example, among very small surface water systems, when the
observed fraction of systems with a contaminant is 1 percent, a 95 percent confidence interval for
the true fraction is 0.01 ± 0.02, or [-0.01, 0.03]. This interval allows the possibility of a negative
fraction of occurrence, which cannot logically occur. The interval may be truncated to [0,0.03],
but the need to truncate suggests that the normal approximation is not the most appropriate method
to calculate the actual confidence level and margin of error.
Table 4. Allocation of C WSs Conducting Assessment Monitoring with
the Associated Margins of Error for an Estimated Exposure of 1%
Size Category
500 and Under
501 to 3,300
3,301 to 10,000
Total
Ground Water-
Supplied Systems
n1
67
186
189
442
99% 2
±3
±2
±2
±1
95% 2
±2
±1
±1
±1
Surface Water-
Supplied Systems
ri1
64
74
141
279
99% 2
,±3
±3
±2
±2
95% 2
±2
±2
±2
±1
Total
n1
131
260
330
721
99% 2
±3
±2
±1
±1
95% 2
±2
±1
±1
±1
Values in the columns with the heading of "n" indicate the number of CWSs allocated to a specific system size category.
2 These column t
columns are the...
calculations in the
The distribution of samples across the three size categories in Table 4 is approximately proportional to population served by each
of the respective community water system size categories. An extra 62 systems were allocated to the smallest system size
category to increase the statistical power of the sample, by decreasing the margin of error. The population figures used to
determine this distribution are based on data from the 1997 SDWIS database inventory. The sample distribution, confidence
levels, and margins of error included in this table are intended only to illustrate the UCMR Program design details and, therefore,
are not final program parameters. When the actual representative sample selection process is conducted, the most current
population data will be used and the related, final program sample distribution, confidence levels, and margins of error will be
calculated. Note that the total sample size indicated above will remain unchanged in the final program design.
See Appendix A for details on how to calculate the normal and Wilson score confidence intervals.
The normal-based confidence interval is only one of several possible confidence intervals for an
estimated proportion. Newcombe (1998) compares seven such intervals, including two varieties of
16
-------
UCMR Statistical Design and State Plans
August 1999
the normal interval. Of these, the Wilson score interval without continuity correction (Wilson, 1927)
has good statistical properties (e.g., the stated confidence level is approximately correct for a wide
range of w and/?), is simple to compute, and unlike the normal interval, always gives confidence
limits between 0 and 1. Given an estimated occurrence fraction/? from a sample of size n, the
Wilson score interval forp is computed as:
2np + z2 ± z-y/z2 + 4np(l - p)
2(n+z2)
(5).
where z is the critical value of the normal distribution at the desired confidence level. Returning to
the example of very small surface water systems, for an estimated occurrence fraction of j:?=0.01
from a sample of size «=64, a 95 percent Wilson score interval for the true proportion is:
2(64)(0.01)+ (1.96)2 ± 1.96V(1.96)2
2(64+(1.96)2)
4(64)(0.01)(0.99)
= [0.0013,0.0742]
(6).
So with 95 percent confidence, the true proportion lies somewhere between 0.1 percent and 7.4
percent By comparison, the normal interval for this example was -1 percent to +3 percent.
Although the Wilson interval in this example is wider than the normal interval, it is more believable
in part because it does not include negative occurrence values.
Table 5 compares the Wilson and normal confidence intervals, still assuming an estimated
occurrence fraction ofp=0.01 and using the sample sizes given in Table 4. A simple interpretation
of these intervals is that the normal interval equals/?, the estimated fraction, plus or minus some
amount, while the Wilson interval equals p times or divided by some amount.
In summary, the normal-based error ranges in Table 4 are useful as a rough guide to the expected
precision of an estimated occurrence fraction. Moreover the normal approximation yields the simple
formula in Equation (1) for estimating the sample size needed to achieve a given precision with
given confidence. However when computing confidence intervals for the estimated proportion, the
Wilson score interval is preferred, both because of its good statistical properties and because it
avoids the possibility of including negative occurrence values.
8. Screening Surveys
The second monitoring component of the UCMR is called the Screening Survey. A second group
of contaminants will be monitored in the Screening Surveys. This group of contaminants includes
thirteen organic contaminants, one microbiological contaminant, and one radiological contaminant
ready for wide-scale, general use. When further developed and ready for use, these methods will
need to be used initially under close quality-controlled conditions. Therefore, a designated group
17
-------
UCMR Statistical Design and State Plans
August 1999
of EPA-certified laboratories will be used for screening survey analytical methods to ensure data
quality of the new methods. With these concerns, the Screening Surveys will be conducted through
a small subset of systems monitoring under the UCMR to provide a national screening for the
UCMR (1999) List 2 contaminants and to better establish the analytical methods used.
Table 5. Comparison of Confidence Intervals at the 99 Percent Confidence Level for CWSs
Size Category
500 and under
501 to 3,300
3^01 to 10,000
Total
Wilson Score Confidence Intervals
Ground Water-
Supplied Systems
[0.001, 0.108]
[0.002, 0.052]
[0.002, 0.052]
[0.003, 0.031]
Surface Water-
Supplied Systems
[0.001, 0.108]
[0.001, 0.100]
[0.002, 0.063]
[0.002, 0.046]
All
[0.001, 0.091]
[0.002, 0.042]
[0.003, 0.037]
[0.004, 0.026]
Size Category
500 and under
501 to 3,300
3301 to 10,000
Total
Normal Confidence Intervals
Ground Water-
Supplied Systems
[-0.021, 0.041]
[-0.009, 0.029]
[-0.008, 0.028]
[-0.002, 0.022]
Surface Water-
Supplied Systems
[-0.021, 0.041]
[-0.019, 0.039]
[-0.010, 0.030]
[-0.006, 0.026]
All
[-0.018, 0.038]
[-0.006, 0.026]
[-0.003, 0.023]
[0.000, 0.020]
Each matrix above represents the confidence intervals using both the Wilson Score method and the normal distribution
method. The confidence level in each matrix (i.e., for ground water-supplied systems serving 500 and under) is
specified as 99 percent.
The Screening Surveys are being conducted to assess contaminant occurrence in PWSs, and not to
determine exposure assessment by population (as is the purpose of Assessment Monitoring). EPA
estimates that there will be two different groups of systems involved in the Screening Surveys. Each
group will be comprised of 300 large and small CWSs andNTNCWSs. The first group will conduct
the Screening Survey in the year 2001 or 2002, the second group will conduct the Screening Survey
in 2003. The two groups will monitor for different contaminants based on the estimated timing of
when methods will be available. Sampling schedules have been established, in part, to enable
Screening Survey water sample collection coincident with the Assessment Monitoring sample
collection whenever possible to minimize labor burden for sample collection.
18
-------
UCMR Statistic?! Design and State Plans
August 1999
EPA is examining general thresholds to evaluate Screening Survey results, relative to the margin of
error in the sample. For example, if a contaminant occurs over a certain threshold (i.e., in a
percentage of systems/population served), the contaminant may then be placed on the Assessment
Monitoring list and monitored in the next round of the UCMR by all large systems and a
representative sample of small systems. If the contaminant occurrence is below this threshold, it is
possible that no further testing will be required. Factors such as health effects levels will also need
to be considered; hence, thresholds may vary by contaminant.
Both large and small systems will be included in the Screening Surveys. Systems will be selected
from all the size and water source categories. However, selection will not be proportionately
weighted by the population served, or by the proportion of system size category. If the sample is
weighted by population served, a disproportionate number of large systems would be included in the
Screening Surveys. However, if the sample is weighted by the number of systems in each size
category, a disproportionate number of small systems would be represented. Therefore, each size
category is given equal importance with 60 systems selected from each size category, with the
selected systems distributed proportionately between surface water and ground water systems. This
results in 180 small systems and 120 large systems hi each of the Screening Surveys (i.e., a total of
360 small systems and 240 large systems hi the two Screening Surveys). To make national
population estimates, or system estimates, the resultant data will need to be weighted in relation to
these sample distributions.
1 , ' ,,,' i ; , i .' ••.,•,•„• ", • • , i
Table 6 illustrates the allocation of systems hi each size category in each group for each Screening
Survey and the associated margin of errors of estimation at the 99 and 95 percent confidence levels
to evaluate the measurement precision for the sample of 300 systems. Even though there are a total
of 600 systems involved, there will be, as noted, two Screening Surveys performed, by two mutually
exclusive groups of systems, analyzing water samples for two different sets of contaminants.
Table 6. Allocation of Systems for Screening Surveys by Size Category
with the Associated Confidence Levels and Margins of Error
Size Category
500 and Under
501 to 3,300
3,301 to 10,000
Subtotal
Small Systems
10,001 to 50,000
50,001 and over
Subtotal
Large Systems
Ground Water-
Supplied Systems
n1
54
44
35
133
51
20
71
99% 2
±3 ,
±4
±4
±2.2
±4
±6
±3.0
95% 2
±3
±3
±3
±1.7
±3
±4
±2.3
Surface Water-
Supplied Systems
n1
6
16
25
47
9
40
49
99% 2
±10
±6
±5
±3.7
±9
±4
±3.7
95% 2
±8
±5
±4
±2.8
±7
±3
±2.8
Total
n1
60
60
60
180
60
60
120
99% 2
±3
±3
±3
±1.9
±3
±3
±2.3
95% 2
±2.5
±2.5
±2.5
±1.5
±2.5
±2.5
±1.8
19
J ..
-------
UCMR Statistical Design and State Plans
August 1999
Size Category
Total
Ground Water-
Supplied Systems
n'
204
99% 2
±1.8
95% 2
±1.4
Surface Water-
Supplied Systems
n1
96
99% 2
±2.6
95% 2
±2.0
Total
n1
300
99% 2
±1.5
95% 2
±1.1
Values in the columns with the heading of "n" indicate the number of CWSs allocated to a specific system size
category.
These column headings indicate the confidence level used for evaluation. The values preceded by "±" listed in these
columns are the margins of error associated with the designated confidence level (either 99 percent or 95 percent).
Error calculations in the table are based on an estimated occurrence fraction of p = 0.01.
Monitoring results for Screening Surveys must be analyzed and interpreted with careful consideration to appropriate
population-weighting. Since the number of systems within each size category in this sample is equal (60 systems), the
monitoring results must be weighted by population within each service size category before interpretation.
The distribution of samples across all categories in Tabje 6 is proportional to the respective population served. The
population and water system information used to determine this distribution are based on data from the 1997 SDWIS
database inventory. The sample distribution, confidence levels, and margins of error included in this table are intended
only to illustrate the UCMR Program design details and, therefore, are not final program parameters. When the actual
representative sample selection process is conducted, the most current population data will be used and the related, final
program sample distribution, confidence levels, and margins of error will be calculated. Note that the total sample sizes
indicated above will remain unchanged in the final program design.
Results from the Screening Surveys are likely only suitable for aggregate national estimates given
the 99 percent confidence level and ±1.5 percent margin of error. Only aggregated national
estimates are appropriate because the error margin may be too large in small subcategories (e.g.,
surface or ground water systems in a given size category) to be conclusive, particularly in cases
where no detections occur. For example, a contaminant most likely to occur in small surface water
systems could have a zero occurrence in the sample, but this contaminant might actually occur in
up to 3.5 percent of these surface water systems nationally. Note also that since the total number of
systems allocated to each size category is equal (60 systems per category), the monitoring results
will have to be weighted by the proportion of the population within each service size category.
Monitoring results will have to be carefully analyzed to correctly assess the possible implications
of such results.
To implement the Screening Surveys, EPA will select 180 small PWSs from the set of 267 systems
(i.e., one-third of the 800 systems in the national representative sample), scheduled to conduct
Assessment Monitoring either in year 2001 or 2002 (for the first group) and again in year 2003 (for
the second group). The probability of a system being selected for Assessment Monitoring (A) in any
given year n, is 267/800, or Pn(A)= 33 percent. Given that a system is first selected for Assessment
Monitoring (A) in any given year n, the probability of that system also being selected for Screening
Survey (S), is:
(7).
20
-------
UCMR Statistical Design and State Plans
August 1999
Overall, there is a 22.5 percent probability that a system will be selected for both the Screening
Surveys and Assessment Monitoring in the same year (67 percent chance of being selected for
Screening Surveys multiplied by a 33 percent chance of being selected for Assessment Monitoring).
However, if the first Screening Survey is conducted in the year 2002, the systems selected to conduct
Assessment Monitoring in the year 2001 have no chance of being selected for a Screening Survey.
Overall, there is a 45 percent chance for a small system to be selected for both Assessment
Monitoring and a Screening Survey simultaneously. Therefore, the probability of a system being
selected only for Assessment Monitoring is estimated as 55 percent. Figure 2 depicts the number
of systems and the probability of a system being chosen for Assessment Monitoring and a Screening
Survey.
•i
Similarly, for the large CWSs and NTNCWSs, the probability of a system being required to
participate in a Screening Survey (5) is:
large
, 240 ,
^2774"
(8).
Therefore, there is approximately a 9 percent probability that a large system will be chosen for a
Screening Survey.
Again, based on the proportion of population served by small CWSs and NTNCWSs in each State,
the number of systems selected for the two groups of Screening Surveys (Sa) in each State/Tribe n,
is calculated as:
Jni
•-NP.
(9).
where Pni is the population served by small systems in State/Tribe n in category i, and NP, is the total
national population served in system category i, and Zf is the total number of systems required to
conduct the survey in that category i.
Table 7 illustrates the allocation of systems conducting Assessment Monitoring and Screening
Surveys in each State/Tribe based on the population served by the systems. The allocation of PWSs
in table 7 is based on the January 1997 inventory estimates from the SDWIS database. The actual
allocation of samples within States, Tribes, and Territories may change when the represented State
sample is selected from current inventory numbers.
21
-------
UCMR Statistical Design and State Plans
August 1999
Figure 2. Number and Probability of Small Systems Chosen for Assessment Monitoring and
Screening Surveys for the UCMR Years
800
Representative
Systems
O00%)
UCMR
Monitoring
Year
1
t> 2001 *•
»• 2002 ^
*2003 »
Assessment
Monitoring
in any
given year
1
266 systems
(Ajyooi suu system)
Assessment Monitoring
Screening Survey
1
Assessment
Monitoring
Only
266 systems
systems)
267 systems
(33% of 800 systems)
180 systems
(67%of267or22.5%of
800 systems)
87 systems
(33%of267orll%of
" 800 systems)
267 systems
(33% of 800 systems)
180 systems
*• (67%of267or22.5%of
800 systems)
87 systems
(33%of267or
> ll%of
800 systems)
* Overall Probability (over three years)
22
-------
UCMR Statistical Design and State Plans
August 1999
Table 7. Distribution of Small Systems Required to Conduct Assessment Monitoring and
Screening Survey in Each State/Tribe/Territory
State/Tribes/
Territories
Tribes4
U.S. Territory5
Alaska
Alabama
Arkansas
Arizona
California
Colorado
Connecticut
Washington DC6
Delaware
Florida
Georgia
Hawaii
Iowa
Idaho
Illinois
Indiana
Kansas
Kentucky
Louisiana
Massachusetts
Maryland
Maine
Michigan
Minnesota
Missouri
Mississippi
Montana
Population Served by
Small Systems
(10,000 or less people)1
(Pn)
406,922
664,719
273,699
1,437,985
1,067,162
706,010
2,994,866
632,197
425,457
0
139,300
2,086,859
1,277,566
210,684
1,089,434
449,557
2,463,266
1,370,421
820,989
1,193,681
1,655,137
831,358
510,175
337,773
1,842,695
983,999
1,405,886
1,741,105
354,745
Number of Small Systems
Conducting Assessment
Monitoring,2
(AJ
6
10
4
21
15
10
43
9
6
0
2
30
18
3
16
6
35
20
12
17
24
12
7
5
26
14
20
25
5
Number of Small
Systems Conducting
Screening Surveys,3
-------
UCMR Statistical Design and State Plans
August 1999
State/Tribes/
Territories
North Carolina
North Dakota
Nebraska
New Hampshire
New Jersey
New Mexico
Nevada
New York
Ohio
Oklahoma
Oregon
Pennsylvania
Rhode Island '
South Carolina
South Dakota
Tennessee
Texas
Utah
Virginia
Vermont
Washington
Wisconsin
West Virginia
Wyoming
Total
Population Served by
Small Systems
(10,000 or less people)1
(Pn)
1,688,147
234,434
468,742
361,738
1,118,886
473,735
261,755
2,377,365
1,967,750
1,195,980
653,761
2,452,599
114,592
841,608
314,992
1,235,268
5,362,371 .
515,146
1,131,704
281,185
1,242,213
1,265,579
781,716
194,170
55,909,083
Number of Small Systems
Conducting Assessment
Monitoring,2
(AJ
24
3
7
5
16
7
4
34
28
17
9
35
2
12
5
18
77
7
16
4
18
18
11
3
800
Number of Small
Systems Conducting
Screening Surveys,3
(Sn)
11
2
3
2
7
3
2
15
13
8
4
16
1
5
2
8
35
3
7
2
8
8
' 5
1
360
Ine distribution of samples above is based on the population and water system information in the 1997 SDWIS
database inventory, pie sample distributions and populations included in this table are intended only to illustrate the
UCMR Program design details and, therefore, are not final program parameters. When the actual representative
sample selection process is conducted, the most current population and water system data will be used, and the related,
final program sample distribution will be calculated. Note that the total numbers of systems conducting sampling
indicated above will remain unchanged in the final program design.
This column represents the total number of small systems allocated in individual State/Tribe from the national
representative sample of 800 systems.
There are 360 small systems shown for Screening Surveys; 180 for each of the two Screening Survey groups. Note
that each Screening Survey Group of 120 large systems will also be required to monitor. Therefore, there is a total
of 300 small and large systems (a total of 600 Screening Survey systems) in each Survey.
24
-------
UCMR Statistical Design and State Plans
August 1999
* The number of Tribal water systems includes Tribal systems in each of the 10 EPA Regions. Tribal systems were
aggregated as a State to ensure that Tribal systems were represented in the national representative sample of small
systems in the UCMR.
s U.S. Territories include American Samoa, Guam, Northern Mariana Islands, Puerto Rico, and the Virgin Islands.
Territories were aggregated as a State to ensure that Territories were represented in the national representative sample
of small systems in the UCMR.
6 The Washington DC water supply is provided exclusively by large PWSs.
9. Pre-Sereen Testing
The third monitoring component of the UCMR is Pre-Screen Testing. EPA established this third
tier of the UCMR monitoring with its stakeholders for contaminants of concern for which analytical
methods are in the early stages of development and/or whose methods are currently too expensive
for wide-scale monitoring. Pre-Screen Testing may also address contaminants that have recently
emerged or been identified as a concern, such as through the Governors' petition process. The
purpose of this monitoring component will be to determine whether the methods in early
development will provide adequate analytical results in conditions under which the contaminants
are most likely to occur. Currently, there are seven microbiological contaminants and one
radiological contaminant listed in §141.40(a)(3) Table 1, UCMR (1999) List 3 that are candidates
for Pre-Screen Testing in 2004.
,•' ' • ' ' ' ' 1 ,
EPA will ask each State who has entered into a MOA with the appropriate EPA Region to identify
a list of between 5 and 25 systems that might be most vulnerable to the UCMR (1999) List 3 Pre-
Screen Testing contaminants. EPA will identify the number of systems selected to monitor hi each
State based on the population served by CWSs and NTNCWSs in the State. From this list, EPA will
randomly select up to 200 large and small CWSs and NTNCWSs nationally for Pre-Screen Testing
of the specified contaminants. The systems selected for Pre-Screen Testing will be added to the
State Plans. The Pre-Screen Testing will use analytical results from a small sample to evaluate and
improve methods, and to conduct an initial assessment of occurrence. Given the small number of
Pre-Screen Testing systems, the monitoring results cannot be used to estimate national occurrence
of UCMR (1999) List 3 contaminants hi a statistically rigorous manner.
25
-------
UCMR Statistical Design and State Plans August 1999
10. References
Casella, G. and Berger, R. (1990), Statistical Inference, Pacific Grove, Ca, Wadsworth.
Cochran, W. G. (1977), Sampling Techniques (3rd ed.), New York, Wiley.
Newcombe, R.G. (1998), "Two-sided confidence intervals for the single proportion:
comparison of seven methods," Statistics in Medicine, 17: 857-872.
Parzen, E. (1960), Modern Probability Theory and Its Applications, New York: Wiley.
Wilson, E.B. (1927), "Probable inference, the law of succession, and statistical
inference," Journal of the American Statistical Association, 22: 209-212.
26
-------
UCMR Statistical Design and State Plans
August 1999
Appendix A
Statistical Formulas
This appendix contains statistical formulas for the estimation of probabilities of exposure and
occurrence, and for means, variances, and confidence intervals for the estimates.
ii 'i , • 1
In the following definitions, let h stand for any one of H strata being considered. Typically a stratum
will be a combination of system size (Category 1 , 2, or 3) and source water type (surface or ground
water), so there are H=6 strata. We could also consider just the 3 size strata or 2 source type strata.
> " '(I"'" : " ' ' ' • • I •- i, ,',-"" '. 1 ' '
We consider a single contaminant of interest where:
. ., i
Mhi — # of people served by the z-th system in h
Mh = mean # of people served by each system in h
Nh = # of systems in h
nh — # of systems sampled from h (assume these are system numbers Z=!,...,HA)
yhi = 1 if the contaminant occurs in the z-th system in h; 0 otherwise.
pe_h = probability of exposure for a person served by a system in h
pe = mean probability of exposure for a person of any stratum
= exposure weight of h
Now make the simplifying assumption that all systems in a stratum serve the same number of
people, that is, Mhi = Mh for all z". There are three reasons for this assumption. First, it simplifies
the estimation and variance formulas below. Second, estimates of the Mhi are not readily available,
while estimates of the Mh are easily obtained from Table 1. Third, the error due to replacing the Mhi
„ , „ „ „, l,
by Mh is likely to be small, since the greatest differences in population per system occur between
strata rather than within strata.
Under the above assumption, the expression farpeh simplifies to:
Pc.k -
(A-l).
A-l
-------
UCMR Statistical Design and State Plans August 1999
Now assuming simple random sampling within each stratum, sample estimates ofpe_h andpe are:
By Theorem 2.1 of Cochran (1977), peh is unbiased forj?^, and so£cis unbiased for^e. By
Theorem 2.2 of Cochran (1977):
nh
where:
and therefore:
A normal-based 100(l-or)% confidence interval for^>e is then pe ± z^V(J>e~) , where z is the
1 - a/2 quantile of the standard normal distribution, and V(^c) is an estimate of Var(/?e)
obtained by substituting peh foTpeh.
The Wilson score interval without continuity correction (Newcombe, 1998; Wilson, 1927) can be
adapted to stratified sampling, under the assumption that Pe h- Pe for all h. In this case
equations (A-4) and (A-5) give:
) = vaPe (i - p. ) (A-6).
where:
-)
A)
A-2
-------
UCMR Statistical Design and State Plans
August 1999
Then by repeating the derivation of the Wilson interval, one finds the same limits as in Equation (5),
but with n replaced everywhere by K"1.
liiiiiiii • in • n 'i Mii i »' . ii
The above equations and derivations for exposure are nearly identical for occurrence. The
probability of occurrence, Po , is defined as the probability that a contaminant occurs within any
(randomly selected) system. The probability of occurrence within a given stratum h is:
Po,h
__l_v"*
'•* ~ NL ^'=1
Vht
(A-8).
(the same aspeh under the assumption that Mhi = Mh for all i, above), and:
ZH
h
(A-9).
where:
(A-10).
The remainder of the derivation of means, variances, and confidence intervals is exactly as
above, with the subscript e replaced everywhere by o.
A-3
•<*
il1,,, "
-------
UCMR Statistical Design and State Plans August 1999
[This page left intentionally blank]
A-4
-------
1:
U£MR Statistical Design and State Plans
August 1999
2,4-DNT
2,6-DNT
4,4'-DDE
Alachlor ESA
AOAC
APHA
ASDWA
ASTM
BGM
CAS
CASRN
CCL
CCR
CERCLA
CFR
CPU
CFU/mL
CWS
DCPA
DCPA mono- and
di-acid degradates
DDT
DNA
EDL
EPA
EPTC
EPTDS
ESA
FACA
FIE
Appendix B
Acronym List
" , ' ' . f \ ••
- 2,4-dinitrotoluene
- 2,6-dinitrotoluene
- 4,4'-dichloro dichlorophenyl ethylene, a degradation product of DDT
' , „ ', ' '.',', h|r ,' , Hi [ ' ,
- alachlor ethanesulfonic acid, a degradation product of alachlor
- Association of Official Analytical Chemists
- American Public Health Association
- Association of State Drinking Water Administrators
- American Society for Testing and Materials
' ' ' ' ' ' ' '••' •• •• 1 •
- Buffalo Green Monkey cells, a specific cell line used to grow viruses
- Chemical Abstract Service
- Chemical Abstract Service Registry Number
- Contaminant Candidate List
- Consumer Confidence Reports
- Comprehensive Environmental Response, Compensation & Liability Act
- Code of Federal Regulations
- colony forming unit
- colony forming units per milliliter
- community water system
- dimethyl tetrachloroterephthalate, chemical name of the herbicide
dacthal
- degradation products of DCPA
- dichloro dichlorophenyl ethylene, a degradation product of DDT
- dichloro diphenyl trichloroethane, a general insecticide
- deoxyribonucleic acid
- estimated detection limit
- Environmental Protection Agency
- s-ethyl-dipropylthiocarbamate, an herbicide
- Entry Point to the Distribution System
- ethanesulfonic acid, a degradation product of alachlor
- Federal Advisory Committee Act
- full-time equivalent
GC
GLI method
GW
GUDI
HPLC
- gas chromatography, a laboratory method
- Great Lakes Instruments method
- ground water
- ground water under the direct influence (of surface water)
- high performance liquid chromatography, a laboratory method
B-l
-------
UCMR Statistical Design and State Plans
August 1999
ICR
IRFA
IMS
IRIS
IS
LLE
MAC
MOA
MCL
MDL
MRL
MS
MS
MSD
MTBE
NAWQA
NCOD
NDWAC
NERL
NFS
NTIS
NTNCWS
NTTAA
OGWDW
OMB
PAH
PB
PBMS
pCi/L
PCR
2>°Pb
210p0
PWS
PWSF
QA
QC
RDX
RFA
RPD
RSD
- Information Collection Request / Rule
- initial regulatory flexibility analysis
- immunomagnetic separation
- Integrated Risk Information System
- internal standard
- liquid/liquid extraction, a laboratory method
- Mycobacterium aviumcomplex
- Memorandum of Agreement
- maximum contaminant level
- method detection limit
- minimum reporting level
- mass spectrometry, a laboratory method
- sample matrix spike
- sample matrix spike duplicate
- memyl-tertiary-butyl-ether, a gasoline additive
- National Water Quality Assessment Program
- National Drinking Water Contaminant Occurrence Database
- National Drinking Water Advisory Council
- National Environmental Research Laboratory
- National Pesticide Survey
- National Technical Information Service
- non-transient non-community water system
- National Technology Transfer and Advancement Act
- Office of Ground Water and Drinking Water
- Office of Management and Budget
- Poly-aromatic hydrocarbon
- particle beam
- Performance-Based Measurement System
- picocuries per liter
- polymerase chain reaction
- Lead-210 (also Pb-210), a lead isotope and radionuclide; part of the
uranium decay series
- Polonium-210 (also Po-210), a polonium isotope and radionuclide; part
of the uranium decay series
- Public Water System
- Public Water System Facility
- quality assurance
- quality control
- royal demolition explosive, hexahydro-l,3,5-trimtro-l,3,5-triazine
- Regulatory Flexibility Act
- relative percent difference
- relative standard deviation
B-2
-------
UCMR Statistical Design and State Plans
August 1999
SBREFA
SD
SDWA
SDWIS
SDWIS FED
SM
SMF
SOC
SPE
SRF
STORET
SW
1BD
•mews
UCMR
UCM
UMRA
USEPA
UV
v6c
• Small Business Regulatory Enforcement Fairness Act
• standard deviation
• Safe Drinking Water Act
• Safe Drinking Water Information System
• the Federal Safe Drinking Water Information System
• Standard Methods
• Standard Compliance Monitoring Framework
• synthetic organic compound
- solid phase extraction, a laboratory method
• State Revolving Fund
• Storage and Retrieval System
- surface water
• to be determined
• transient non-community water system
• Unregulated Contaminant Monitoring Regulation/Rule
• Unregulated Contaminant Monitoring
• Unfunded Mandates Reform Act of 1995
• United States Environmental Protection Agency
• ultraviolet
• volatile organic compound
• micrograms per liter
B-3
, HSi' f , [ill
i.-i,'!. Ri,:" l:i
:"'"li!,!i i
-------
UCMR Statistical Design and State Plans August 1999
[This page left intentionally blank]
B-4
-------
UCMR Statistical Design and State Plans August 1999
Appendix C
"' 111 , |, ' | ' 'I'll ,,'" 'j,l l| :'!
Definitions
All monitored systems means all community water systems serving more than 10,000 people, and
the national representative sample of community and non-transient non-community water systems
serving 10,000 or fewer people that are selected to be part of a State Plan for the UCMR.
:, : ' ' ": " '. " : , :.". ' : 1
Assessment Monitoring means sampling, testing, and reporting of listed contaminants that have
available analytical methods and for which preliminary data indicate their possible occurrence hi
drinking water. All monitored systems must conduct Assessment Monitoring. Assessment
Monitoring will be conducted for the UCMR (1999) List 1 contaminants.
Intfex Systems means a limited number of small CWSs andNTNCWSs, randomly selected from the
systems inState Plans, that must monitor for UCMR contaminants and also additionally must report
information on system operating conditions (such as water source, pumping rates, and environmental
setting). These systems must monitor and report quarterly each year of the 5-year UCMR cycle with
EPA paying for all reasonable monitoring costs. This more detailed and regular monitoring of
contaminants and operating conditions will provide important information with which EPA can
more fully evaluate conditions under which systems operate and will enable comparisons between
system operations of similar size and characteristics.
"i:;,,i ::: • ••• • ;_ • •• • • • I ,
Listed contaminant means a contaminant identified as an analyte in Table 1, 141.40(a)(3) of the
Unregulated Contaminant Monitoring Regulation (UCMR). To distinguish the current 1999 UCMR
listed contaminants from potential future UCMR listed contaminants, all references to UCMR
contaminant lists will identify the appropriate year hi parenthesis immediately following the
acronym UCMR and before the referenced list. For example, the contaminants included hi the
UCMR (1999) List include the component lists identified as UCMR (1999) List 1, UCMR (1999)
List 2 and UCMR (1999) List 3 contaminants.
i1, • • Tli • u "!> •• ' !' I
!"" . .. ; , v »i? • ! :; '. " '-' ., v-:". ... •!, '; :. ' •..
Listing cycle means the 5-year period for which each revised UCMR list is effective and during
whjch no more than 30 unregulated contaminants from the list may be required to be monitored.
EPA is mandated to develop and promulgate a new UCMR List every 5 years.
I ; ••' ! It' ' ' ' i'l • f '' . ' " ' • • " ''•','',' ,;!> ' "'i,- . ' . k. " ' '••- : " ' '
Monitoring means (as distinct from Assessment Monitoring), all aspects of determining the quality
of drinking water relative to the listed contaminants. These aspects include drinking water sampling
and testing, and the reviewing, reporting, and submission to EPA of analytical results.
Most vulnerable systems (ox Systems most vulnerable) means a subset of 5 to not more than 25
systems of all monitored systems in a State that are determined by that State in consultation with the
EPA Regional Office to be most likely to have the listed contaminants occur hi their drinking waters,
considering the characteristics of the listed contaminants, precipitation, system operation, and
environmental conditions (soils, geology and land use).
Pre-Screen Testing means samplhig, testing, and reporting of the listed contaminants that may have
newly emerged as drinking water concerns and, in most cases, for which methods are hi an early
stage of development. Pre-Screen Testing must be conducted by a limited number of systems (up
to 200). The Pre-Screen Testing systems will be selected through the use of a random number
generatorj and from a list comprised of the States' nominations of up to 25 of the most vulnerable
:r. • , . ' • • • , ., C-l
-------
UCMR Statistical Design and State Plans August 1999
systems per State. Pre-Screen Testing will be performed to determine whether a listed contaminant
occurs in sufficient frequency in the most vulnerable systems or sampling locations to warrant its
being included in future Assessment Monitoring or Screening Surveys. Pre-Screen Testing will be
conducted for the UCMR (1999) List 3 contaminants.
Random Sampling is a statistical sampling method by which each member of the population has an
equal probability (an equal random chance) of being selected as part of a sample (the sample being
a small subset of the population which represents the population as a whole).
Representative Sample means a subset of community and non-transient non-community water
systems serving 10,000 or fewer people which EPA selects using a random number generator to
obtain public water system identification numbers to place them on the first representative sample
list. The selection is weighted by population served within a State, water source and then by -size
categories of 10,000 to 3,301 people, 3,300 to 501 people, and 500 or fewer people; a State may
substitute systems from a replacement list of such systems derived through the same method for
systems in the first list because a system on the first list is closed, merged or purchases water from
another system.
Sampling means the act of collecting water from the appropriate location in a public water system
(from the applicable point from an intake or well to the end of a distribution line, or in some limited
cases, a residential tap) following proper methods for the particular contaminant or group of
contaminants.
Sampling Point means a unique location where UCMR samples are to be collected.
Screening Survey means sampling, testing, and reporting of the listed contaminants for which
analytical methods are recently developed and have uncertain potential for occurrence in drinking
water by a subset of approximately 300 systems from all monitored systems selected through use
of a random number generator for public water system identification numbers. These systems must
conduct the Screening Survey for the listed contaminants after public notice and comment to
determine whether a listed contaminant occurs at a sufficient frequency and concentration (or
density) to warrant being included in future Assessment Monitoring. Two Screening Surveys will
be conducted for the UCMR (1999) List 2 contaminants.
State means, for the purposes of this section, each of the fifty States, the District of Columbia, U.S.
Territories, and Tribal lands. For the national representative sample, Guam, the Commonwealth of
Puerto Rico, the Northern Mariana Islands, the Virgin Islands, American Samoa, and the Trust
Territories of the Pacific Islands are treated as a State. Any Indian Tribe which has status as a State
under Section 1451 of the Safe Drinking Water Act for this program will be considered as a State.
State Monitoring Plan (or State Plan) means a State's portion of the national representative sample
of CWSs and NTNCWSs serving 10,000 or fewer people which must monitor for unregulated
contaminants. A State Plan may be developed by a State's acceptance of EPA's representative
sample for that State, or by a State's selection of systems from a replacement list for systems
specified in the first list that are closed, merged or purchase water from another system. A State Plan
also includes a process by which the State will inform each public water system of its selection for
the plan and of its responsibilities to monitor. A State Plan will also include the systems required
to conduct Pre-Screen Testing, selected from the State's designation of vulnerable systems.
Stratified Random Sampling is a procedure to draw a random sample from a population that has been
divided into subpopulations or strata, with each stratum comprised of a population subset sharing
C-2
-------
'I !"••"
UCMR Statistical Design and State Plans
August 1999
common characteristics. Random samples are selected from each stratum proportional to that
stratum's proportion of the entire population. The aggregate random sample (compiled from all the
strata samples) provides a random sample of the entire population that reflects the proportional
distribution of characteristics of the population. In the context of the UCMR, the population of
public water systems was stratified by size category (based on population served by the water
system) and by water source type supplying the water system (ground water or surface water). This
stratification was done to ensure that systems randomly selected as nationally representative sample
systems would proportionally reflect the actual number of size and water type categories nationally.
Testingmeans, for the purposes of tiie UCMR and distinct from Pre-Screen Testing, the submission
reporting of the sample's analytical results for evaluation. Testing is a subset of activities defined
as monitoring.
. „ i1 j mil • ,' < "in ii
Unregulated contaminants means chemical, microbiological, radiological and other substances that
occur in drinking water or sources of drinking water that are not currently regulated under the federal
drinking water program. EPA,has not issued standards for these substances hi drinking water (i.e.,
maximum contaminant levels or treatment technology requirements). EPA is required by Congress
to establish a program to monitor for selected unregulated contaminants hi public water systems to
determine whether they should be considered for future regulation to protect public health. The
selected contaminants are listed hi 141.40(a)(3), Table 1, the UCM List.
Vulnerable time (or vulnerable period) means the tune (or, in some cases, the 3-month quarter) of
the year determined as the most likely to have the listed group of contaminants present at their
highest concentrations or densities in drinking water. The vulnerable determination, in the case of
the UCMR, is made by the EPA or by the State (under arrangement with the EPA) for a system,
subset of systems, or all systems hi a State. The vulnerable determination is based on characteristics
of the contaminants, precipitation, system operations, and environmental conditions such as soil
types, geology, and land use. This determination does not indicate or imply that the listed
contaminants will be identified in the drinking water with certainty, but only that sampling
conducted during the vulnerable period presumably has the highest likelihood of identifying those
contaminants in higher concentrations relative to other sampling times of the year, if and when the
contaminants occur.
C-3
------- |