Sampling and
Consideration of
Variability (Temporal and
Spatial) For Monitoring of
Recreational Waters
&EPA
U.S. Environmental Protection Agency
Office of Water
EPA-823-R-10-005
December 2010
-------
This page is intentionally blank.
-------
U.S. Environmental Protection Agency—Office of Water
Contents
Executive Summary 1
CHAPTER 1 Introduction and Background 5
1.1. Purpose 5
1.2. EPA Monitoring Research for New or Revised Recreational Water Quality Criteria 5
1.3. Summary of Previous EPA Recommended Water Quality Criteria 5
1.3.1. Previous EPA Recreational Ambient Water Quality Criteria 6
1.3.2. Current EPA Recreational AWQC 6
1.3.3. EPA National BEACH Guidance and Required Performance Criteria for Grants 6
1.4. Organization of this Report 7
CHAPTER 2 Findings on Indicator Density and Temporal Variability 9
2.1. Variability with Time Scales Less than 1 Hour 10
2.1.1. Coastal Sites 10
2.1.2. River and Stream Sites 10
2.1.3. Summary 11
2.2. Diurnal Variations 11
2.2.1. Coastal Sites 12
2.2.2. River and Stream Sites 13
2.2.3. Summary 13
2.3. Variations Related to Tidal Processes 13
2.4. Variability Attributable to Rainfall and Runoff (Event-Scale Variability) 15
2.4.1. River and Stream Sites 15
2.4.2. Coastal Sites 16
2.4.3. Summary 17
2.5. Monthly and Seasonal Variability 18
2.5.1. River and Stream Sites 18
2.5.2. Coastal Sites 20
2.5.3. Summary 20
2.6. Predictive Power of Prior Day's Indicator Density 21
CHAPTER 3 Findings on Indicator Density and Spatial Variability 23
3.1. Indicator Density Variation with Water Depth at the Point of Sample Collection 25
3.1.1. Coastal Sites 25
3.1.2. River and Stream Sites 28
3.1.3. Summary 28
3.2. Indicator Density Variation with Depth Below Surface of Sample Collection 30
3.2.1. Inland Sites 30
3.2.2. Coastal Sites 31
3.2.3. Summary 31
3.3. Along-Shore Spatial Variability 31
3.4. Longitudinal Spatial Variation, Streams and Lakes 36
3.5. Cross-Sectional Variations in Indicator Densities in Streams 38
3.6. Beach Features Promoting Non-Homogeneity in Indicator Density 39
3.7. Inferences from Single and Composite Samples 40
CHAPTER 4 Development of Monitoring Approach 43
4.1. Factors to Consider for Monitoring Plan Development 43
4.2. Site Characterization 43
4.2.1. Sanitary Surveys 43
December 2010
-------
U.S. Environmental Protection Agency—Office of Water
4.2.2. Pilot Studies 44
4.3. Monitoring Approaches and Statistics for Assessing Water Quality 46
4.3.1. Statistical Considerations: Variability, Confidence Estimates, and Sample Numbers 47
4.3.2. Statistical Considerations for Spatial Sampling 48
4.3.3. Statistical Considerations for Temporal Sampling 49
4.4. Summary of Monitoring Approaches and Considerations 50
4.4.1. Review of Indicator Density Variability 50
4.4.2. Where to Sample 51
4.4.3. When to Sample 52
4.4.4. How to Sample 53
4.4.5. How Often to Sample 53
CHAPTER 5 References 55
Tables
Table 1. Summary statistics for distribution of E. coll density over 24 hours for three streams 11
Table 2. Fraction of exceedances missed for different sampling schemes 21
Figures
Figure 1. Distribution of indicator sources in a coastal setting near a point source 23
Figure 2. Transport processes in the surf zone (plan view) 24
Figure 3. Indicator density variation with water depth at sample collection, 9:00 a.m. samples,
EMPACT Study (E. coli for West Beach and Belle Isle and Enterococcus for other
beaches) 27
Figure 4. Indicator density variation with water depth at sample collection, 2:00 p.m. samples,
EMPACT Study (E. coli for West Beach and Belle Isle and Enterococcus for other
beaches) 27
Figure 5. Dispersion, advection, and removal in the surf zone 32
Figure 6. Variation in fecal coliform density downstream of a sewage outfall 37
Figure 7. E. coli density along cross-stream transects of the Charles River 39
Figure 8. Illustration of beach features promoting non-uniform indicator density in parts of a beach 40
Figure 9. Illustration of uneven fecal pollution loading and potential sample locations 45
Figure 10. Illustration of beach features interfering with mixing 45
Figure 11. Confidence intervals tighten with increasing number of samples 47
December 2010
-------
U.S. Environmental Protection Agency—Office of Water
Abbreviations and Acronyms
AFRI
ANN
ANOVA
AWQC
BEACH Act
CAFO
CPU
CL
CPSP
CSO
CV
CWA
DGD
DPLSR
EMC
EMPACT
EPA
EU
FIB
GI
HCGI
LOAEL
MF
MPN
MS4
MSE
MTF
NEEAR
NOAEL
NPDES
NFS
NRC
OLSR
ORP
PC
PCR
POTW
acute febrile respiratory illness
artificial neural network (model)
analysis of variance
ambient water quality criteria
Beaches Environmental Assessment and Coastal Health Act
concentrated animal feeding operation
colony forming unit
confidence limit
Critical Path Science Plan
combined sewer overflow
coeffi ci ent of vari ati on
Clean Water Act
discrete growth distribution
dynamic partial least-squares regression (model)
event mean concentration
Environmental Monitoring for Public Access and Community Tracking
U.S. Environmental Protection Agency
European Union
fecal indicator bacteria
gastrointestinal
highly credible gastrointestinal illness
lowest observed adverse effect level
membrane filtration
most probable number
municipal separate storm sewer system
mean sum of errors
multiple tube fermentation
National Epidemiological and Environmental Assessment of Recreational
no-observed-adverse-effect level
National Pollutant Discharge Elimination System
nonpoint source (pollution)
National Research Council
ordinary least squares regression (model)
oxidation reduction potential
prospective cohort
polymerase chain reaction
publicly owned (sewage/wastewater) treatment works
December 2010
-------
U.S. Environmental Protection Agency—Office of Water
qPCR
QPCRCE
RMSE
ROC
SSM
SSO
TMA
TMDL
U.K.
WHO
WQS
WWTP
quantitative polymerase chain reaction
qPCR cell equivalents
root mean square error
receiver operating characteristic (analysis)
single sample maximum
sanitary sewer overflow
transcription-mediated amplification
total maximum daily load
United Kingdom
World Health Organization (United Nations)
water quality standard[s]
wastewater treatment plant
December 2010
IV
-------
U.S. Environmental Protection Agency—Office of Water
Executive Summary
This report reviews the literature on temporal and spatial variability of fecal indicator organism
density at recreational sites and the implications of variability for the design of sampling plans.
For all sites, the greatest temporal variability in indicator densities is from rain events. For
coastal water quality sampling locations, the greatest spatial variabilities are those related to
sample depth and site features, such as the alignment of fecal pollution sources with a beach. For
inland recreational sites, along-stream variability is most important. For coastal sites, pilot
monitoring and sanitary surveys are useful tools for collecting site information. These should be
performed before development of monitoring plans and sampling microbial water quality in the
morning, at waist depth, and at multiple locations selected according to the site characteristics.
For inland sites, sample locations can be selected on the basis of known or suspected locations of
fecal pollution sources and the locations where recreational activity is likely.
Methodology
A literature review was performed to identify and compile the information used to develop this
report. The review included specific searches for information on physical and biological
processes at temporal and spatial scales relevant to indicator organism variability for coastal and
inland waters. On the basis of the results of the review, the report summarizes key findings to
help in the design of appropriate recreational water quality sampling schemes that are protective
of human health.
This report emphasizes research and findings primarily from studies using culture-based
methods. Non-culture-based methods (e.g., quantitative polymerase chain reaction) are
mentioned and discussed where information is available. Such information, however, is not well
described in the literature. Accordingly, this report acknowledges the expected future importance
and relevance of non-culture-based methods for developing and implementing EPA's new or
revised recreational water quality criteria. In addition, the attributes of current fecal indicators
and available enumeration methods, along with their inherent uncertainties, are not discussed in
this report, despite their importance in interpreting monitoring results.
Summary of Key Findings
The literature review revealed that several factors influence temporal and spatial variability of
fecal indicators in recreational waters, although with different degrees of importance. The
ranking of those factors is illustrated in Exhibit 1. Discrete events (e.g., precipitation events or
combined sewer overflow [CSO] discharges) have by far the greatest impact on temporal
variability, while sample depth and along-stream sampling have the greatest impact on spatial
variability for coastal and inland sites, respectively. Most important, specific knowledge of a
recreational site is crucial, and appropriate site investigation is paramount to achieving an
accurate and comprehensive understanding of the factors influencing fecal pollution and
associated risks to human health at that site.
December 2010
-------
U.S. Environmental Protection Agency—Office of Water
Exhibit 1. Ranking of factors influencing variability of fecal
indicators in natural systems
TEMPORAL VARIABILITY
SPATIAL VARIABILITY
COASTAL SITES
Temporal variability
This report points to the global
importance of climatic features
(e.g., temperature, storm events,
day/night duration, tide intensity)
on indicator variability along with
the indirect consequences on
loading through increased
recreational activities and
associated risks in warmer
seasons and locales. The
importance of human-made
events (e.g., treated wastewater
effluent discharges) is also
highlighted.
Spatial variability: coastal sites
The sample depth, related to the
swimmer's distance from the
shoreline (e.g., ankle- and waist-
depth) exhibits higher spatial
variability than along-shore
variations or variations with depth
at which a sample is drawn. Site features that either promote or prevent mixing can have a strong
influence on the distribution of indicators along a coastline. The impact of site features highlights
the importance of sanitary surveys in developing monitoring schemes.
Spatial variability: inland sites
Variability along (longitudinally) streams and estuaries is generally greater than that associated
with vertical depth of sampling from the water surface or cross-stream variability. As with
coastal sites, this finding emphasizes the importance of identifying fecal pollution sources
through a sanitary survey before developing water quality monitoring plans.
Statistical assessment of water quality
Along-shore variation of indicator density at coastal sites appears best characterized by a
lognormal distribution. When interpreting the results of multiple samples taken at a site, the
geometric mean of indicator densities is considered the best metric for characterizing water
quality, because the geometric has been demonstrated to correlate with incidence of illness in
epidemiological studies conducted at coastal sites. In general, for large sites requiring multiple
samples to characterize water quality, discrete sampling at multiple points is suggested; although,
using composite sampling could provide a valuable tradeoff between cost and effort and
precision for assessing fecal indicator densities.
Note: Temporal variability at a short time scale is ranked lowest, except for
samples obtained at ankle depth and shallower.
December 2010
-------
U.S. Environmental Protection Agency—Office of Water
Monitoring Considerations
All the above factors influencing the variability of fecal indicator densities need to be taken into
account when designing a monitoring scheme for a specific recreational site. On the basis of the
factors illustrated in Exhibit 1 and specific features at a site, the following approach can be used
to help design a monitoring plan (Exhibit 2). Pilot monitoring studies and sanitary surveys are
the best tools available for collecting data required to develop effective site-specific monitoring
plans.
Exhibit 2. Monitoring considerations for recreational waters
WHERE
Area allowing best and
most efficient
characterization:
>^ Link to fecal pollution
>^ No native sources
>^ Small variability
Coastal
>^ Knee-deep or greater
>^ Knowledge of
hydrodynamics
Inland
>^ Knowledge of stream
>^ Top 6 inches of water
column
WHEN
Morning samples yield
conservative results
relating water quality to
human health effects when
using culture methods,
whereas the use of qPCR
methods yields results that
are relatively stable
throughout the day.
Sample collection
frequency could be
related to site charac-
terization, site usage, or
practical constraints.
HOW
Multiple approaches for
choice of location and
number of samples,
based on site specific
constraints and
historical data:
>^ Power-curve approach
> sampling based on
site-specific variance
>^ Limited sampling
> based on constraints
v' Composite sampling
December 2010
-------
U.S. Environmental Protection Agency—Office of Water
This page is intentionally blank.
December 2010
-------
U.S. Environmental Protection Agency—Office of Water
CHAPTER 1 Introduction and Background
1.1. PURPOSE
The purpose of this document is to meet one of the elements (Project P-12) in the U.S.
Environmental Protection Agency's (EPA'sj Critical Path Science Plan for Development of New
or Revised Recreational Water Quality Criteria (CPSP) (USEPA 2007a).1 The intent of Project
P-12 is to provide detailed reference information so that EPA can "design and evaluate a
monitoring approach that will characterize the quality of beach waters that takes into account the
spatial and temporal variability associated with water sampling." After publication of its new or
revised recreation water quality criteria, EPA expects to use information from this report and
other materials to develop implementation recommendations.
1.2. EPA MONITORING RESEARCH FOR NEW OR REVISED RECREATIONAL
WATER QUALITY CRITERIA
In 2002 EPA published Environmental Monitoring for Public Access and Community Tracking
(EMPACT) Beaches Project: Time-Relevant Beach and Recreational Water Quality Monitoring
and Reporting (USEPA 2002a) and in 2005 the EMPACT Beaches Project: Results from a Study
on Microbiological Monitoring in Recreational Waters (USEPA 2005). Both of those projects
were part of the EMPACT Program. Given its obvious relevance to this report, data from the
EMPACT Program is discussed and analyzed in Chapter 4 of this report.
EPA has also been conducting the National Epidemiological and Environmental Assessment of
Recreational (NEEAR) Water Study, which is a series of prospective cohort (PC)
epidemiological studies beginning in 2002 at several Great Lakes (freshwater) recreational
beaches and continuing at marine beaches. The purpose of the NEEAR epidemiology studies is
to determine the association of swimming illness with fecal indicator levels in recreational
waters.
1.3. SUMMARY OF PREVIOUS EPA RECOMMENDED WATER QUALITY
CRITERIA
A brief review of the microbiological guidelines and standards/criteria for recreational waters
and their context for development and implementation, as addressed by the EPA, is presented
below.
1 Report is at http://www.epa. gov/waterscience/criteria/recreation/plan/index.html.
2 Further information about the NEEAR Water Study is at http://www.epa.gov/nheerl/neear/index.html.
December 2010
-------
U.S. Environmental Protection Agency—Office of Water
1.3.1. PREVIOUS EPA RECREATIONAL AMBIENT WATER QUALITY CRITERIA
The ambient water quality criteria (AWQC) for the United States that were proposed in 1968 and
recommended again in 1976 were established on the basis of the epidemiological studies
conducted during the late 1940s and early 1950s by the U.S. Public Health Service (Stevenson
1953). Those criteria for recreational waters were, "As determined by multiple-tube fermentation
or membrane filter procedures and based on a minimum of not less than five samples taken over
not more than a 30-day period, the fecal coliform content of primary contact recreation waters
shall not exceed a log mean of 200/100 millilters [mL], nor shall more than 10 percent of total
samples during any 30-day period exceed 400/100 mL."
1.3.2. CURRENT EPA RECREATIONAL AWQC
EPA's current water quality criteria for recreational exposure to surface waters (USEPA 1986)
are based on the observed occurrence of gastrointestinal (GI) illness associated with swimming
in fresh (USEPA 1984) or marine (USEPA 1983) recreational waters as determined through
several PC epidemiology studies conducted in the 1970s and early 1980s.
For marine recreational waters, based on a statistically significant number of samples (generally
not less than five samples equally spaced over a 30-day period), a steady state (i.e., dry weather
conditions) geometric mean indicator density of 35 CPU (colony forming units)/!00 mL of
enterococci was recommended; for fresh recreational waters, a steady state geometric mean
indicator density of 33 CFU/100 mL for enterococci or 126 CFU/mL for Escherichia coll was
recommended. In addition, no single sample should exceed a one-sided confidence limit (CL)
value calculated for each indicator according to four different levels of beach usage (i.e.,
established single sample maximums [SSMs]). In this regard, the 1986 bacteria criteria
recommended different SSMs depending on beach usage levels. The levels correspond to the
following four SSMs: designated bathing beach for the 75 percent (most protective) CL,
moderate use for bathing for the 82% CL, light use for bathing for the 90 percent CL, and
infrequent use for bathing for the 95 percent CL. Thus, where a given recreational area has a
greater potential for more people to be exposed, a higher degree of protectiveness (i.e., a lower
SSM) was recommended.
Those recommended criteria are in effect and required for use at coastal and Great Lakes waters
designated for swimming or similar water contact activities, except where the state or territory
has in place EPA-approved criteria that are as protective of human health as EPA's 1986
recommendations (USEPA 2004). EPA also published a fact sheet (USEPA 2006a) that
addresses questions regarding the appropriate risk level (or levels) a state may choose when
adopting into the state's water quality standards (WQS) bacteria criteria to protect its coastal
recreation waters. Another fact sheet (USEPA 2006b) addresses the appropriate use of the SSM
values component of EPA's 1986 bacteria criteria in coastal recreation waters.
1.3.3. EPA NATIONAL BEACH GUIDANCE AND REQUIRED PERFORMANCE
CRITERIA FOR GRANTS
EPA's National Beach Guidance and Required Performance Criteria for Grants (USEPA
2002b) provides performance criteria for monitoring and assessment of coastal recreation waters
December 2010
-------
U.S. Environmental Protection Agency—Office of Water
adjacent to beaches, and for prompt public notification of any exceedance or likelihood of
exceedance of applicable WQS for pathogens and pathogen indicators for coastal recreation
waters. It also outlines the eligibility requirements for monitoring and notification program
implementation grants under CWA section 406(b).
That beach guidance document provides EPA's current requirements and recommendations for
monitoring beach waters. Chapter 3 of that guidance establishes procedures for states to evaluate
and rank their beaches according to risk or usage (or both) and establish a priority tiering system.
Chapter 4 of that document requires that states develop a Tiered Monitoring Plan, consistent with
the priority ranking of their beaches. Requirements and recommendations are included for a
variety of monitoring circumstances and other monitoring/assessment issues. For each of the
tiers, it offers recommendations such as when to conduct basic sampling; when to conduct
additional sampling; where to collect samples; what depth to collect samples, and such. More
detailed monitoring considerations are discussed in Appendix H of the document. Chapter 5 of
that document sets forth the public notification requirements and recommendations for a tiered
notification system.
1.4. ORGANIZATION OF THIS REPORT
To prepare this report, a detailed literature search and retrieval was conducted. Chapter 2
provides findings from the literature on temporal variability of indicator density for all relevant
time scales. Chapter 3 provides findings from the literature on spatial variability of indicator
density at all relevant length scales and directions. Chapter 4 draws and builds on Chapters 2 and
3 to describe when, where, and how monitoring could be conducted such that it is consistent with
and accounts for the spatial and temporal variability inherent in fecal indicator organism
densities in natural systems. Last, on the basis of findings from the literature and analyses,
Chapter 4 also lays out factors to consider in determining where to sample, when to sample, and
how to sample for recreational microbial water quality purposes.
It is important to note that culture-derived quantification methods (e.g., membrane filtration,
Enterolert® and Colilert®) are the only EPA-approved methods for regulatory monitoring of
fecal indicators. Therefore, the majority of the phenomena described in this report relate
indicator variability (temporal and spatial) for indicator densities enumerated via culture
methods. It is not suggested that variability will be the same when different methods are used,
only that the body of literature available for assessing variability for culture-independent
methods is relatively small. Particularly in the case of the quantitative polymerase chain reaction
assay (qPCR), the variability of the indicator signal in both space and time can differ from that of
the culture signal. The persistence of genetic material differs from that of live, viable cells; the
uncertainty of molecular methods could be significantly different from that of culture methods.
However, in the past decade, culture-independent enumeration methods (e.g., qPCR) have grown
widely in use and sophistication and are likely to become standardized as a regulatory
monitoring tool, mainly thanks to their rapidity and ability to enumerate non-culturable
organisms. Thus, relevant information related to such methods is cited where appropriate in this
report. As discussed in Section 4.8, work is under way to assess the inherent variability of the
methods for use as a monitoring tool for recreational waters.
December 2010
-------
U.S. Environmental Protection Agency—Office of Water
This page is intentionally blank.
December 2010
-------
U.S. Environmental Protection Agency—Office of Water
CHAPTER 2 Findings on Indicator Density and
Temporal Variability
This chapter discusses the temporal variability of indicator organism density. The phenomena
described in this chapter and Chapter 3 form the basis for considerations and suggestions in
Chapter 4 about where, when, and how recreational water quality sampling should be conducted.
More specifically, this section presents findings from a literature review of published studies on
the temporal variability of indicator density for relevant time scales, sorted by waterbody type
(coastal versus inland rivers and streams).
Temporal variability in indicator density—at time scales ranging from minutes to months—has
been observed in time series analyses of indicator density. Variations with time scales on the
order of minutes are important because such considerations influence the number of samples
needed to accurately characterize microbial water quality and the confidence with which to
ascribe results of sampling events. Variations with times scales on the order of tens of minutes
are important because they have the same time scale as that of typical recreational use episodes.
Variations with time scales on the order of a day are important because their knowledge allows
comparison between samples taken at different times of the day or between samples taken on
successive days.
The tradeoff between sampling cost and effort and protection of public health is illustrated by
Fleisher (1985, 1990). In those two studies, reanalysis of total coliform data collected over a 3-
year period shows that variability in indicator density resulted in potential mis-classification of
water quality for 33 percent, 64 percent, and 71 percent of sampling dates for the first, second,
and third years of the study, respectively. Reanalysis entailed classifying sample results as above
or below the criterion on the basis of their 95 percent confidence interval, rather than a simple
arithmetic mean or geometric mean of samples taken over a given period (for more discussion of
arithmetic mean versus geometric mean, see Section 4.1.1). The authors also found that
contradictory water quality determinations could often be made on the basis of morning and
afternoon sample results. Method uncertainty and temporal variability both contributed to the
overall uncertainty in water quality. The observations led Fleisher (1990) to recommend replicate
samples be drawn at bathing sites and that replicate laboratory analyses be performed on sample
splits.
The use of lognormal distribution for describing the distribution of indicator densities at a site or
between sites is described in greater detail in sections that follow, and thus is only briefly
discussed below. In general, time series non-log-transformed indicator data are characterized by
long tails at high indicator organism densities. The long tails result from the very high indicator
densities associated with rain events and the frequency with which such events occur. Thus, the
temporal distribution of indicators at a coastal site is often assumed well characterized by a
lognormal distribution. For example, results from enumeration of enterococci in 11,000 bathing
water samples collected from marine sites in the U.K. were fit with a lognormal distribution with
a mean of 0.9337 (i.e., a geometric mean of 9 enterococci/100 mL), standard deviation of 0.8103
and a 95th percentile value of 2.267 (i.e., 185 enterococci/100 mL) (Kay et al. 2004). Kim and
Grant (2004) also found that a lognormal distribution provided a good fit to a relatively large
December 2010
-------
U.S. Environmental Protection Agency—Office of Water
(n = 860) data set of enterococci observations (goodness of fit was assessed via a Kolmogorov-
Smirnov test for normality); however, the distribution mean and standard deviation were not
reported.
2.1. VARIABILITY WITH TIME SCALES LESS THAN 1 HOUR
2.1.1. COASTAL SITES
At two marine beaches, Boehm (2007) noted very high variability in enterococci density at time
scales less than 1 hour. The high short-time-scale temporal variability was determined to not be
the result of method uncertainty (the Enterolert® most probable number (MPN) method was used
for bacteria enumeration in that study) and was not random (white noise). Rather, enterococci
time series were found to be fractal, with variability in densities related to physical and
biological processes occurring at the sample locations. For samples drawn at 10-minute intervals,
average variability (as change in concentration between consecutive samples) was 60 percent and
as high as 700 percent. To achieve a coefficient of variation of 50 percent around the one-hour
mean, the number of samples at the four sampling points (on two beaches) evaluated in the study
was estimated to be 6, 5, 4, and 4, respectively. To achieve a 20 percent coefficient of variation,
the number of samples was estimated to be 39, 31, 25, and 25 for the four sampling locations.
For samples taken at 1-minute intervals at a single sample location (samples taken at ankle depth
on incoming waves and analyzed via Enterolert , with results reported as MPN/100 mL), Boehm
(2007) again observed high variability, with an average enterococci density change between
consecutive samples of 34 MPN/100 mL/minute and a maximum change of 140 MPN/100
mL/minute. To achieve coefficients of variation of 20 percent and 50 percent relative to the 10-
minute mean enterococci density, 10 and 2 samples would have to be drawn, respectively.
In an earlier study of short time-scale temporal variation in indicator density, Boehm et al.
(2002) noted high variability between samples taken at 10-minute intervals. Samples in that
study were collected at ankle depth for incoming waves. Many observances of samples
significantly below WQS followed by samples significantly exceeding the same WQS were
reported. Transport of pulses of enterococci via rip currents (time scale on the order of hours)
was inferred from observation of elevated enterococci densities at five locations along the beach.
The authors estimated that, for the water quality monitored during the studies, 70 percent of
single sample exceedances (104 MPN/100 mL) have durations of less than 1 hour, and 40
percent have durations of less than 10 minutes.
2.1.2. RIVER AND STREAM SITES
Variability over short time scales has been observed at inland streams and at coastal sites. Meays
et al. (2006) studied E. coli variability in three streams—two in areas dominated by agricultural
and forested land use and one downstream of an area of recreational use. The mean, minimum,
maximum and standard deviation of E. coli density (data not log-transformed) for samples drawn
at 15-minute intervals for 24-hour monitoring are presented in Table 1. Both between-sample
and longer time scale variabilities were observed in E. coli densities. A period of elevated E. coli
December 2010 10
-------
U.S. Environmental Protection Agency—Office of Water
density was observed during the afternoon hours at the site with the highest mean E. coli density,
which was attributed to a rainfall event that occurred on the morning of the study.
Table 1.
Summary statistics for distribution of E. coli density over 24 hours for three streams
Creek
Duteau Creek (primarily agricultural and
forest land use)
Deer Creek (primarily agricultural and forest
land use)
BX Creek (downstream from a recreation
area)
Mean
4
19
156
Minimum
0
6
22
Maximum
13
79
696
Standard
deviation
2.3
11.8
181.4
Source: Meays et al. 2006
Because variability differs between streams and arises from a complex set of factors, the authors
recommend that an understanding of the sources for a site (e.g., by executing a sanitary survey)
be developed before designing and implementing monitoring programs.
2.1.3. SUMMARY
Significant short time-scale variability has been observed at shallow (ankle-depth) coastal sites
(Boehm 2007). Extreme variability in indicator density is generally limited to shallow sites and
is likely related to mobilization of sediment-associated indicator bacteria by wave action. Short-
term variability is less pronounced at locations with greater water depth. Two strategies for
overcoming short time-scale variability when assessing bacteriological water quality are to select
sample sites with less variability (e.g., sites at greater water depth) or to use composite samples if
sampling at locations with high variability cannot be avoided or is required.
Short-term variability (time scales of less than 1 hour) has also been observed in streams. Event-
scale and diurnal variability are generally greater than short-term variability in streams; although,
sudden loading can result in rapid changes in stream indicator density. Because short time-scale
variability in streams is less significant than other variabilities, short time-scale fluctuations are
not a significant factor in developing sampling plans for stream sites.
2.2. DIURNAL VARIATIONS
Several studies have identified diurnal variation in indicator density in marine and freshwater
coastal environments, streams, and non-flowing inland waters (e.g., Brenniman et al. 1981; Boehm
et al. 2002; Whitman et al. 2004a; Whitman and Nevers 2004b; Noble et al. 2005; Liu et al. 2006;
Meays et al. 2006; Rosenfeld et al. 2006; Traister and Anisfeld 2006; He et al. 2007). All other
factors being equal, when measured by culture methods, fecal indicator bacteria demonstrate a
predictable pattern of highest density in the morning, decreasing density during the day (often by
several orders of magnitude), reaching the lowest density in the mid-afternoon, and followed by a
sharp rebound of density in the late evening. The decrease of indicator bacteria during daylight
hours results from inactivation of organisms by incident solar radiation (Sinton et al. 2002) and
December 2010
11
-------
U.S. Environmental Protection Agency—Office of Water
possibly from increased removal of organisms via predation (Menon et al. 2003; Boehm et al.
2005a). The rapid rebound of indicator density during evening hours remains incompletely
understood (Boehm 2007). Although the likely cause of rapid rebound is resuscitation of viable but
non-culturable cells, it is possible that other processes such as replenishment of viable indicators
from other sources (sediments, influent waters) also play a role.
In contrast to indicator diurnal variation by culture methods, indicator density diurnal variation
for qPCR methods is lower, with relatively stable indicator density reported for samples taken
throughout the day (e.g., as observed at Great Lakes beaches by Wade et al. 2006). This is
apparently due to the different persistence and sensitivity to light molecular material versus
viable culture cells. Those differences result in differences in diurnal variation in indicator
densities when measured by the two techniques. In studies of light and dark marine water
mesocosms, Walters et al. (2009) found that decay rates of naked genetic material were the same
in both types of mesocosms, whereas inactivation of culturable cells was faster in light
mesocosms than dark ones. Further, the persistence of naked genetic material was significantly
longer than that of intact viable cells in marine water and in sewage. The findings suggest that
variability of indicators as measured by qPCR is likely different from that of indicators as
measured by culture-based methods. That difference is expected to be pronounced for diurnal
variability of indicator densities in streams, inland lakes, and coastal sites.
2.2.1. COASTAL SITES
In a comparison of fecal coliform, total coliform, and enterococci survival in marine
environments and mesocosms, Boehm et al. (2002) found that mesocosm indicator organism
densities declined when mesocosms were exposed to natural sunlight, but they did not rebound
during evening and nighttime hours. In contrast, bacteria populations in the surf zone exposed to
the same solar radiation rebounded rapidly, reaching morning density levels by approximately
8:00 in the evening. Reasons for differences between mesocosm and in situ populations include
rapid replenishment of bacteria from sediments or other sources, or growth outpacing
inactivation/removal in situ during periods of low solar radiation intensity. In general, because of
the predictable variation in microbiological water quality during the course of a day, morning
water quality assessments are good predictors of afternoon water quality determinations. For
example, in a study of marine beaches, Corbett et al. (1993) found that a strong correlation
existed between passing water quality determination (in this case, geometric mean fecal coliform
count less than 300/100 mL) in a morning test and sub sequent pass in an afternoon test, while
there was a 50 percent chance of water quality failing the afternoon test on days when the
morning test resulted in a failure.
However, in support for EPA epidemiological studies conducted at inland (Great Lakes) bathing
beaches, Haugland et al. (2005) and Wade et al. (2008) observed that, in contrast to culture-
based method results, qPCR counts of enterococci in Great Lakes waters were relatively constant
during the day, which is consistent with the explanation provided at the end of Section 1.3.
December 2010 12
-------
U.S. Environmental Protection Agency—Office of Water
2.2.2. RIVER AND STREAM SITES
Traister and Anisfeld (2006) observed diurnal variation of E. coli in five temperate streams
except on days in which loads of E. coli from rainfall/runoff masked the die-off of bacteria in the
afternoon. The daily fluctuations in E. coli density on streams were found to be more pronounced
on slower-flowing, less shaded stream reaches than on smaller, more shaded ones. Meays et al.
(2006) observed that stream indicator density response to rainfall was much greater than diurnal
variability due to UV radiation or temperature effects and die-off
A potential anthropogenic cause for diurnal indicator density fluctuations is the variable loading
of surface waters of raw (untreated) and treated sewage. Bordalo (2003) observed that fecal
coliform density in raw sewage discharged to a river 3.3 kilometers (km) upstream of its mouth
1 9
exhibited high temporal variability, reaching a peak concentration around 10 CFU/100 mL
around 9:00 a.m., a second, less distinguishable peak around 108 CFU/100 mL around 8:00 p.m.,
and a low value of less than 10 CFU/100 mL at 10:00 p.m. Indicator density and loadings for
treated sewage are also expected to vary with time of day, although not as radically as for raw
sewage.
2.2.3. SUMMARY
Regardless of the cause of diurnal fluctuations in indicator density as measured using culture-
based methods, the universal observance of the fluctuations dictates that sampling should be
conducted at the same time each day if water quality is to be compared between days and that
sampling in the morning provides the most conservative measure of the health risk posed by
recreational water. An additional benefit of morning sampling is delivery and analysis of the
samples at laboratories early in the day. That allows the availability of results of 24-hour tests
before the beginning of recreational activities on the following day for culture methods and can
expedite reporting of results from qPCR methods.
Sampling strategies that account for diurnal variations in indicator density include the following
(Whitman and Nevers 2004b):
• Collecting samples at a standard time of day at which maximum exposure is anticipated.
• Using early morning samples for developing conservative estimates of water quality.
• Using adaptive sampling (collecting supplemental samples on the basis of the results of
earlier sampling events).
2.3. VARIATIONS RELATED TO TIDAL PROCESSES
Tides influence indicator organism density via dilution (during flooding tides); through drainage
of indicator organisms from sands, sediments, and coastal wetlands (during ebb tides) by
establishing a connection between the tidal waters and nearhore surface waters; and through tidal
currents (Boehm and Weisberg 2005b). The extent to which tides influence indicator density
depends on the size of the tide because dilution is directly related to the tide height and because
the distribution of indicator organisms in nearshore sediments and waters varies spatially. To
determine which elements of the tidal cycle (spring versus neap and ebb versus flowing) have the
greatest influence on indicator (enterococci) density at marine beaches in Southern California,
Boehm and Weisberg performed statistical analyses of a large database of indicator density and
December 2010 13
-------
U.S. Environmental Protection Agency—Office of Water
tide conditions. On the basis of observation of signals in indicator density associated with tidal
phenomena and on an N-factor analysis of variance (ANOVA), the authors concluded that spring
tides and the spring-ebb tide cycle were associated with rises in indicator density at the majority
of beaches studied, regardless of the proximity to known point sources of fecal pollution. Those
results indicate that the presence of indicator organisms at coastal sites during spring tides and
the spring-ebb tide cycle may not have a direct relationship to sources of fecal pollution. Rather,
they may be related to other sources or reservoirs of indicators, including birds, and organisms
stored or growing in sediments, wrack, and water within the beach aquifer.
In a study of another Southern California beach, Boehm at el. (2003) used the increased
incidence of indicator bacteria in the water column during ebb tide to deduce that shore—rather
than offshore or intermittent—sources of indicator bacteria were the likely cause of frequent
exceedances of WQS at that beach. The finding is consistent with and explained by subsequent
research (Santoro and Boehm 2007; Yamahara et al. 2007), in which enterococci densities in
sediments decreased significantly when tides submerged the sediments, presumably mobilizing
loosely bound bacteria from sediments and introducing them to the water column. Rough
estimates of the number of enterococci mobilized from sediments during a rising tide were very
close to estimates of the increase in number of enterococci in the water column during the same
period.
In a study of the same shoreline, other researchers (Rosenfeld et al. 2006) confirmed the
association of higher indicator densities with spring tides. The trend was observed before and
after disinfection was initiated at a wastewater treatment plant (WWTP) discharging to a deep-
water outfall in the study area. The lack of change in indicator relationship with tides after
implementation of disinfection suggests that interaction of tidal processes with the outfall plume
is not responsible for indicator loads along the section of beach studied. The association of
elevated indicator density with spring tides was also observed at Hong Kong beaches (Cheung et
al. 1991). Contrary to other findings, indicator densities at Hong Kong beaches tended to be low
during ebb tides. The observed fecal indicator density trends were attributed to the transport of
fecal pollution to the beaches from sources outside the beaches.
A less direct, though still important, influence of tides on indicator density was shown by Boehm
et al. (2004) in a study of the covariation between sea surface temperature and total coliform
density along a 23-km stretch of Southern California coast. Water temperature was found to have
a fortnightly variation, potentially resulting in upwelling and subsequent transport of offshore
pollutant plumes toward shore. Because the source and transport mechanisms are complex, the
authors could not conclusively verify their importance and recommended further investigation.
In summary, low tides are associated in most cases with higher indicator organism densities at
coastal sites. This association is a result of mobilization of indicators from sediments as tide
waters recede. In a minority of circumstances, such as when rising tides cause waters with high
indicator density to become hydrologically connected to coastal waters, high tides can be
associated with high indicator densities. In general, tidal variability is minor compared with
diurnal variability and rainfall event-related variability. Approaches for accounting for tidal
variation of indicator density in developing sampling schemes include (1) sampling without
regard to tidal cycles, or (2) sampling at low tide or the portion of the tidal cycle during which
indicator density is highest (all other factors being equal).
December 2010 14
-------
U.S. Environmental Protection Agency—Office of Water
2.4. VARIABILITY ATTRIBUTABLE TO RAINFALL AND RUNOFF
(EVENT-SCALE VARIABILITY)
Rainfall and subsequent runoff can increase indicator density through loading (e.g., wash-off of
indicators with surface flow, washout of indicators from beach sands or river bank sediments,
initiation of combined sewer overflow [CSO] or sanitary sewer overflow [SSO] events), or can
decrease it by dilution (Gentry et al. 2006; Koirala et al. 2008; Vidon et al. 2008). The complex
relationship between hydrology and indicator density results in frequent poor correlation
between hydrologic variables (e.g., stream flow and precipitation) and indicator organism density
but better correlation between hydrologic variables and indicator load (Gentry et al. 2006; Vidon
et al. 2008). As noted by Petersen et al. (2005), "bacterial pollution is characterized in terms of
concentrations, but concentration data may be misleading if not related to the flows from each
source as loads are additive, while concentrations are not."
2.4.1. RIVER AND STREAM SITES
Traister and Anisfeld (2006) noted that stream E. coli density varied greatly between storms and
was not simply related to precipitation depth. They also reported that change in E. coli density
can be related to land use, with more urbanized stream reaches showing a smaller response
(change in density) for a given storm than less urbanized reaches.
Astrom et al. (2009) developed a predictive model for indicator and pathogen density for a large
river receiving indicator and pathogen loads from WWTP effluent and CSO and SSO discharges.
Triangular distributions were assumed for the density of indicators (E. coli, spores ofClostridia
spp. [potential pathogenic organisms], and somatic coliphages) and of pathogens (norovirus,
Giardia, Cryptosporidium) in raw sewage and dilution of microorganism loads by runoff were
assumed lognormally distributed. A Monte Carlo simulation of water quality in the receiving
water indicated the importance of single emergency events (SSO events) occurring in dry or wet
weather. The model tended to underpredict median indicator and pathogen densities but
overpredict the upper 95 percent confidence level for densities.
The response of stream indicator density (the pollutograph) to rainfall events varies significantly
from storm to storm (Dorner et al. 2007) and within storms (Baxter-Potter and Gilliland 1988;
Jamieson et al. 2005). Although correlated with stream flow, indicator density varies with stream
flow in a complex manner. For example, intensive monitoring of fecal coliform density during a
single storm demonstrated consistently higher density of the indicator for a given stream
discharge during the rising limb of the hydrograph than the falling limb (Baxter-Potter and
Gilliland 1988; Olyphant and Whitman 2004). During the early portion of storms, wash-off of
indicators into streams is high, whereas loads are lower later in storms because surface sources of
microorganisms are depleted (Traister and Anisfeld 2006; Dorner et al. 2007). In studies of
indicator density changes in streams during storms, Jamieson et al. (2005), Edwards et al. (1997),
and Haack et al. (2003) also observed higher indicator density associated with the rising limb of
the hydrograph. Jamieson et al. (2005) speculated that indicator densities are higher during the
rising limb because there is a greater availability of particle-associated bacteria to be
resuspended; during the falling limb, most of the bacteria available for resuspension have been
depleted. The importance of resuspension of sediment indicators was also noted by Edwards et
al. (1997) and McDonald et al. (1982). During controlled releases of water from reservoirs
December 2010 15
-------
U.S. Environmental Protection Agency—Office of Water
during dry periods, pollutographs of fecal coliforms and total coliforms similar to those
associated with storms are observed (i.e., high densities during the rising limb and lower
densities during recession) (McDonald et al. 1982). That observation emphasizes the importance
of resuspension in the mass balance of indicator organisms in streams.
Rainfall influences on indicator densities in both streams and coastal sites near the mouths of
streams have been observed in relatively undeveloped watersheds and in those dominated by
stormwater or publicly owned treatment works (POTW) discharges. In an agriculture- and
woodland-dominated watershed in Jersey, U.K., indicator density (total coliforms, E. coli, and
streptococci [enterococci]) was strongly influenced by rain events at coastal and inland sites,
with enterococci density increasing by more than three orders of magnitude at the outlet of the
stream after one storm (Wyer et al. 1995a). Wyer and colleagues concluded that indicator
organism loading from captive birds (swans and ducks) played an important role in elevation of
indicator densities during storm events in that watershed. That conclusion was based on a
sanitary survey and comparison of indicator densities at key locations in the catchment.
Interestingly, in that study, a significant reduction in indicator density was observed downstream
of the bird sources; the decrease was attributed to sedimentation and is further evidence of the
complex interactions between precipitation, loading, and geography that give rise to temporal
changes in indicator organism density.
The importance of individual source contributions in determining the indicator density can vary
with rainfall. For example, using combined water quality data and microbial source tracking
(MST) data, Shehane et al. (2005) observed that a coastal stream was more affected by animal
sources during a period of drought and more affected by human sources during periods of normal
precipitation. In that same study, it was shown that a composite index based on measurements of
multiple indicator organisms was a better indicator of water quality and correlated better with
rainfall than any individual indicator organism; that finding is consistent with the observation
that multiple sources influence the water quality and that their relative importance changes
temporally.
Thresholds at which rainfall and runoff produce large changes in fecal indicator density differ
between rivers and for a given river according to the conditions antecedent to the rainfall. For an
estuary along the North Carolina coast, it was determined that indicator (fecal coliform and
Enterococcus) density was significantly different after storms with net precipitation greater than
or equal to 2.5 centimeters (cm) when rainfall was less than 2.5 cm and that rainfall amounts
above 3.81 cm were associated with indicator densities above an action level (Coulliette and
Noble 2008). The difference was observed at stations relatively near the coast (within 250 meters
[m]) and for stations further offshore.
2.4.2. COASTAL SITES
The effect of the duration of a rainfall event on fecal indicator bacteria on a coastal site is
variable. For a coastal beach in harbors receiving stormwater runoff in urbanized areas, rainfall
in the prior 24 hours accounted for 5 to 10 times more variability in a regression model than
rainfall in other periods (prior 48 hours, 22 hours, and so on) (Hose et al. 2005). Chigbu et al.
(2005) observed that, in an estuary on the Gulf of Mexico, the time required for fecal coliform
density in the estuary to fall to a geometric mean of 14 fecal coliforms MPN per 100 mL ranged
December 2010 16
-------
U.S. Environmental Protection Agency—Office of Water
from 0.3 to 12.9 days. Haramoto et al. (2006) found thatE. coli levels in marine coastal sites fell
to pre-storm levels within a few days of the rain event in Tokyo Bay.
The influence of rainfall events on beach water quality along the California coast was observed
to be much higher near storm drains, particularly those in urbanized areas, and to persist for more
than 36 hours after a large storm (Noble et al. 2003a). After a large (spatially) storm of total
precipitation between 2.7 and 7.8 cm, 87 percent of beaches in close proximity to urban runoff
outlets failed to meet WQS (10,000 MPN or CPU per 100 mL total coliforms, 400 MPN or CPU
per 100 mL fecal coliforms, or 104 MPN or CPU per 100 mL enterococci) on the basis of single
samples, with enterococci standard exceeded in 100 percent of samples exceeding either of the
other two standards. The extent of shoreline exceeding criteria following the storm was 10 times
greater than for dry weather. Among samples whose indicator density exceeded criteria, the
indicator density was generally far in excess of the standard. In contrast, exceedances during dry
weather tend to be only slightly above criteria. This study indicates both the importance of
rainfall events on coastal sites and the persistence of effects of rainfall on water quality for a
significant period following the end of rain. Put differently, dilution cannot be assumed to
completely mitigate rainfall effects at coastal sites or to ensure rapid return of indicator density
to pre-storm levels.
The lag between a rainfall event and a subsequent change in indicator density at a coastal site can
vary significantly with the orientation of the site to stormwater outfalls, river mouths, or other
point or contained sources of indicator organisms. Haack et al. (2003) noted that on the Grand
Traverse Bay, Lake Michigan, a 48-72 hour lag existed between rainfall and elevated E. coli
density at southern shoreline beaches, but no such lag was observed for western and eastern
shoreline beaches.
Rainfall and runoff suspend indicator organism loads from sands and sediments on beaches and
release them from external sources such as storm drains and stream discharges. Whitman et al.
(2006) observed E. coli response to a rainfall event for hydrologically connected sand, pore
water, and lake water. E. coli density in all three media increased in the early stage of the rainfall
event, and sand E. coli density fell sharply and faster than density in the other two media after
the rainfall event. That observation indicates the potential for high loading of indicator organisms
originating from beach sands or stream sediments early in storms, and lower loadings after
sediments and sands are depleted, late in rain events. The observation is consistent with the
findings of Yamahara et al. (2007), who observed mobilization of enterococci during a rising tide
or because of wave action, followed by reduced loading as sediment indicator bacteria were
depleted.
2.4.3. SUMMARY
Event scale variability causes the greatest variability (including both temporal and spatial
variabilities) in indicator density for coastal and inland waters. During events, indicator densities
at all types of sites can undergo orders-of-magnitude changes, and events account for a large
fraction of indicator organism loadings to drinking water source waters,3 inland lakes and
3 Although the main intent of this section is describing variability in recreational waters, several studies on drinking
water reservoir loading are cited and described because they provide data that informs the understanding of inland
lake loading and indicator variability.
December 2010 17
-------
U.S. Environmental Protection Agency—Office of Water
reservoirs, and coastal sites. For inland sites, indicator densities correlate poorly with rainfall
amounts and stream gage due to dependence of indicator response (pollutograph) on factors such
as antecedent rainfall (which relates to soil capacity to retain stormwater and the number of
indicator organisms available for runoff into receiving waters) and the input of indicator bacteria
from sources such as CSO discharges. In general, indicator density peaks during the rising limb
of the storm hydrograph when loading to the stream is high and streams are turbulent, promoting
resuspension of sediment-associated indicators. The lag period between the beginning of rainfall
events and sharp rises in indicator density varies among sites, with small, flashy streams
exhibiting shorter lag periods and coastal sites exhibiting longer lag periods. Generally, indicator
densities decline faster than the hydrograph because of depletion of indicators from land surfaces
and other reservoirs as they are washed out. The time required for the indicator density in a
stream or lake to recede to pre-storm levels is highly variable among drainages and even for a
given drainage. Similar trends have been observed for coastal sites: indicator densities rise
quickly during storms because of loading from stormwater runoff, nearshore sands, and
increased wave action and mobilization of indicators from sediments. Presumably, dilution
would cause event-scale variability to be less at coastal sites than on streams, though poor
mixing in the vicinity of stream mouths and stormwater outfalls appears to contribute to extreme
event-driven changes in indicators.
2.5. MONTHLY AND SEASONAL VARIABILITY
2.5.1. RIVER AND STREAM SITES
On an inland lake near the Texas-Oklahoma border and with relatively low rainfall during
summer months, E. coli density was variable, but generally lower during summer months than
winter months (An et al. 2002). At the Texas-Oklahoma site, low summer month densities likely
are due to low loading as a result of lower rainfall (those months tended to be drier than other
seasons) and higher die-off and removal via predation with increasing water temperature.
Observations made in the study can differ from those of lakes in other regions of the United
States with different seasonal rain patterns. Note that low loading of lakes during summer is not
inconsistent with typically high indicator densities in streams during summer months—although
indicator density can be high, stream discharge is often low during summer months. Monthly
mean water temperature was not reported in that study, precluding the comparison of rainfall and
temperature effects. In contrast, in a small stream without point sources of fecal pollution in an
area with more even yearly distribution of rainfall (northern Indiana), E. coli density was
generally higher during summer months than winter months. At that site, the peakE. coli
occurrence (based on weekly sampling) was during warmer months (in the late summer)
(Byappanahalli et al. 2003).
The combined effects of indicator organism loading (including from nonpoint sources where
growth can occur along with sedimentation/resuspension) and dilution determine the indicator
density at a station and time (Gentry et al. 2006; Vidon et al. 2008). Thus, Vidon and colleagues
observed that in two agriculture-dominated watersheds, E. coli density (number of bacteria per
volume of stream water) did not change significantly with season, although E. coli loading
(number of bacteria per time) was higher during winter months than summer months.
Obiri-Danso and Jones (1999) observed relatively steady levels of fecal streptococci in two
December 2010 18
-------
U.S. Environmental Protection Agency—Office of Water
highly polluted streams during a 12-month period, despite wide differences in loading during the
period. Die-off rates for E. coli and fecal streptococci (similar to enterococci) are similar and
dependent on the same factors. Higher loading during winter months could indicate more direct
connection between E. coli sources (e.g., failed septic systems) and receiving waters or could be
the result of improved survival of E. coli at lower temperature. Koirala et al. (2008) observed a
seasonal trend in total coliform density in a stream in a mixed-use watershed in Tennessee.
Interpretation of their results warrants caution in the context of this report, because many non-
fecal sources of total coliforms exist. However, because that study was one of few identified in
which longer-term indicator trends were described for inland streams, the findings and
implications of the study are presented here. Monthly geometric mean total coliforms were
highest during summer months (periods of high temperature [possible regrowth] and low flow
[low dilution]). On the basis of time series analysis, Koirala and colleagues also noted total
coliform density exhibited long-term persistence (period from 4 weeks to 1 year), perhaps related
to stocks of total coliforms in stream sediments or stream bank soils. Seasonal trends in E. coli
for multiple stations were observed in a mixed-use watershed (Traister and Anisfeld 2006), with
apparent increases in E. coli during summer months for samples taken under baseflow
conditions.
Likewise, Tiefenthaler et al. (2009) observed higher enterococci and E. coli densities during
baseflow in unaffected streams in Southern California during summer months and attributed the
trend to summer conditions promoting growth or regrowth in streams, to increased loads from
sources such as wildlife and birds, or to reduced streamflow (lower dilution) during summer
months. Reischer et al. (2008) observed seasonally highE1. coli density during summer and early
fall on an Alpine spring-fed stream, with the highest loadings to the stream coming from
summertime rain events. In that study, seasonality in the detection of ruminant-specific BacR
marker was also observed and attributed to seasonal variation in the discharge of springs to the
stream (dilution). Edwards et al. (1997) observed high fecal coliform and fecal streptococcus
densities during summertime on two streams whose watersheds were primarily pasture lands and
deciduous forest; although, the authors noted that periodic observations of indicator organisms in
the fall and spring were at the same level as those observed in the summer. Those high spring
and fall observations might have been associated with wet weather, indicating that rainfall and
runoff play a more significant role in variability of indicator density than season. As with the
total coliform trends observed by Koirala et al. (2008) above, the findings of Edwards and
colleagues should be interpreted with the understanding that many potential non-fecal sources of
fecal coliforms and fecal streptococci exist.
Seasonal variations tend to be more pronounced for smaller streams and headwaters than near the
mouth of streams or for large streams (Shanks et al. 2006). A tropical stream in Hawaii exhibited
higher fecal indicator densities roughly in December to March than in the rest of the year (Roll
and Fujioka 1997). In their MST study of a catchment with agricultural and POTW impacts,
Shanks et al. (2006) observed different seasonal variations in indicator density and source-
specific indicators on different parts of the drainage. The differences could be, in part, attributed
to source and to rainfall/runoff Indicators are loaded sporadically from agriculture, with loading
occurring during rainfall and dependent on die-off of indicators in land-applied waste between
rain events. POTW loading is relatively steady (independent of rain events) and expected to
exhibit seasonal variations that differ from agricultural loadings.
December 2010 19
-------
U.S. Environmental Protection Agency—Office of Water
2.5.2. COASTAL SITES
In an estuary in North Carolina, both fecal coliforms and enterococci densities (determined via
Enterolert® and Colilert® with E. coli assumed to comprise the majority of fecal coliforms) were
generally highest during summer months and lowest during winter months (Coulliette and Noble
2008). Trowbridge and Jones (2009) also observed higher fecal coliform densities in an estuary
during summer months. Higher summer indicator organism densities in estuaries are likely
caused by the same factors promoting higher summer indicator densities in streams: lower flow
rates [lower dilution] during relatively dry summer months; and higher loading (possible growth
in sediments and higher loadings from agricultural and wildlife sources) during higher-
temperature summer months. Sayler et al. (1975) observed lower summertime indicator densities
and maximum densities in December on the Chesapeake Bay, with the exception of a sampling
location at the bay mouth where high densities were observed in the summertime.
On the basis of a study of E. coli occurrence in upland soils and stream headwaters, downstream
waters, beach soils and sediments and coastal waters, Whitman et al. (2006) demonstrate that
soils upland from beaches can serve as steady nonpoint sources of fecal indicator bacteria that
persist throughout the year. Like E. coli and other indicators growing in beach sands, the
occurrence on a beach of those indicators from nonpoint watershed sources does not necessarily
coincide with fecal pollution events, whose loading and seasonality can be significantly different
from those of the non-enteric, environmental population.
In a one-year study of Southern California beach sites (Turbow et al. 2003), seasonal variation in
enterococci densities differed from that typically observed in temperate climate streams; higher
indicator densities were observed during late winter and early spring at the California coastal
sites, whereas higher densities in temperate streams and estuaries were reported for summer
months when temperature is high and rainfall low. Interestingly, Tiefenthaler et al. (2009) report
higher enterococci counts in reference (natural) streams in Southern California during summer
months. That observation is at odds with the observed seasonal trend at coastal sites and points to
the importance of anthropogenic indicator bacteria sources and complex dynamics at the coastal
sites. Pednekar et al. (2005) were able to attribute 69 percent of variation in total coliform
density at a Southern California bay to rainfall, indicating that stormwater runoff is the most
significant source of indicator bacteria in urbanized areas of Southern California.
2.5.3. SUMMARY
Most U.S. inland streams experience higher indicator densities during the summer than the
winter. That phenomenon arises from generally lower precipitation and runoff during summer
months combined with greater loading from sources such as wildlife and domestic animals
(particularly those with seasonal access to streams) and bacteria growing in nearshore soils or
sediments. In locales with tropical climates such as Hawaii, Puerto Rico, south Florida and
others, differences in seasonal precipitation trends and other climatic factors can give rise to peak
indicator density in a season other than summer. For sites where the recreational use season
spans only summer months, variation in indicator density with season does not influence design
of monitoring programs. Similarly, seasonal and monthly variability of fecal indicators at coastal
sites is difficult to assess and tends to be linked to the wide range of climates existing along the
U.S. shoreline and its indirect consequences on indicator density (e.g., loading patterns that vary
December 2010 20
-------
U.S. Environmental Protection Agency—Office of Water
with season). At both inland and coastal settings, site type, seasonal, and monthly variability of
fecal indicator organisms is of lesser significance than event-scale variability.
2.6. PREDICTIVE POWER OF PRIOR DAY'S INDICATOR DENSITY
Leecaster and Weisberg (2001) analyzed a large data set of total coliform and fecal coliform data
from samples collected at Southern California beaches in an attempt to associate sample
collection frequency with misidentification of indicator density in exceedance of standards. No
consideration was made of the lag time between sampling and completion of analysis. The
number of missed exceedances for four sampling schemes is presented in Table 2. An
explanation for the poor performance of the schemes considered is the frequency of exceedances
of single-day duration; approximately 70 percent of exceedances lasted only one day. The
exceedances were characterized by water quality only slightly exceeding standards. Given the
variabilities and uncertainties associated with sample collection and analysis, there is a high
probability for misclassification of water quality for samples whose indicator level is near the
standard.
In a study of the impact of deep-water outfalls on marine beach water quality, Armstrong et al.
(1996) recognized that loading of fecal pollution at monitored beaches was episodic, despite a
relatively constant flux of indicator organisms in the presumptive sources (outfalls) of beach
indicators. In that same study, the predictive power of rainfall on the day of sampling and
indicator density from samples drawn two days before sampling was found to be greater than
that of rainfall alone or visual indicators of pollution alone. The improvement was not quantified,
and the authors noted factors that could confound the improvement in fit of general linear models
using sampling day rainfall and two days' prior indicator density as covariates.
Olyphant and Whitman (2004) performed a regression analysis to determine the relationship
between E. coli density on a given sample day and E. coli density on the prior day at the same
time for samples taken at a Great Lakes beach. The resulting correlation coefficient was not
statistically different from zero, indicating virtually no correlation inE. coli density for
successive days. Correlation was, however, observed between E. coli density in samples taken at
different times on the same day. On the basis of those result, the authors note the need for a
warning system that operates semi-continuously.
Table 2.
Fraction of exceedances missed for different sampling schemes
Sampling scheme
5 days per week (weekdays only)
3 times per week
Once per week
Once per month
% Missed exceedances
20%
45%
75%
95%
Source: Adapted from Leecaster and Weisberg 2001
December 2010
21
-------
U.S. Environmental Protection Agency—Office of Water
This page is intentionally blank.
December 2010 22
-------
U.S. Environmental Protection Agency—Office of Water
CHAPTER 3 Findings on Indicator Density and
Spatial Variability
This chapter discusses variations in indicator organism density and uncertainty attributable to
spatial factors. The phenomena described in this chapter support the suggestions made in Chapter
4 regarding where and how recreational water quality sampling should be conducted. More
specifically, this chapter presents findings from a literature review of published studies on spatial
variability of indicator density at all relevant length scales and directions. Spatial variability
within a site relates to the alignment of sources within the site (Figure 1), advection, and the
distribution of mixing on the site.
As described below, transport processes in coastal settings are complex and highly variable. The
most important transport processes are shown in Figure 2, a schematic illustrating water transport
at a coastal site.. Those processes include along-shore flow (littoral drift), turbulent dispersion,
offshore transport in jet plumes, and rip tides. Turbulent dispersion, rip tides, and along-shore
flow all disperse indicators, although at different length scales and with different mechanisms.
As described in Section 4.1.1, rip tides might play an important role in the dispersion and
transport of fecal pollution plumes. Rip tides might remove indicators from the surf zone, then
redeposit them at a location further up or down the coast from the location where they were
extracted, resulting in irregular, patchy indicator distribution along a beach. Tidal flows (not
shown in Figure 2) also play significant roles in the determination of indicator density and
distribution along a beach. Tidal influences and conditions promoting high and low indicator
densities are described in Section 2.3.
Pointsource
Diffuse source
Figure 1. Distribution of indicator sources in a coastal setting near a point source.
December 2010
23
-------
U.S. Environmental Protection Agency—Office of Water
Turbulent
dispersion!
-------
U.S. Environmental Protection Agency—Office of Water
Various statistical distributions have been proposed for describing spatial variability of indicators
in the along-shore direction at coastal sites, the most important of which are the Poisson
distribution, the negative binomial distribution, and the lognormal distribution. The Poisson
distribution describes well-mixed (homogeneous) sites, whereas the negative binomial
distribution describes sites with a high degree of heterogeneity. The lognormal distribution has
been suggested as adequately describing the along-shore variability for coastal sites (USEPA
2005). The lognormal distribution n is the most familiar of the distributions for the regulated
community and is relatively easy to use in common spreadsheet programs.
3.1. INDICATOR DENSITY VARIATION WITH WATER DEPTH AT THE POINT OF
SAMPLE COLLECTION
The most significant factors contributing to indicator density differences along transects
extending perpendicular to the shore (i.e., the sampling zone) at a coastal site or perpendicular to
the streamlines of inland flowing waters are the following:
• Proximity to sources (especially sediments).
• Settling.
• Mixing.
• Dilution.
Outside the mixing zones of point sources, the processes governing the distribution of indicators
along a transect are (a) generation of indicators from sediments, (b) dispersion of indicators to
lower-density waters, (c) settling of indicators, and (d) dilution. Resuspension of fecal indicators
from sediments into the water column occurs in relatively shallow water where mixing (turbulent
kinetic energy) is vigorous and dilution is relatively low.
The importance of the various factors listed above can vary depending on the analytical method
used for enumerating the indicator. For example, the proximity to source can have different level
of influence depending on whether one uses a qPCR or a culture technique; genetic material and
culturable cells from the same source can have different persistences.
3.1.1. COASTAL SITES
Factors that have been used for selecting water depth at sampling locations for coastal sites include
the depth at which adults are most likely to ingest water, depth at which children are most likely to
ingest water, and the increasing likelihood that sediments will be disturbed and influence sample
indicator density for samples taken at shallow locations (Kleinheinz et al. 2006).
Four studies of variation in indicator density with water depth at Great Lakes beaches produced a
general indication that indicator density decreases with increasing distance from shore, although
results of the studies are somewhat contradictory. In a study of two Lake Erie beaches,
Brenniman et al. (1981) did not identify a significant relationship between water depth at the
sample collection location and the indicator density. One of the beaches monitored in that study
was approximately 1,600 m west of a large wastewater plant discharge and 300 m from the
nearest stormwater discharge. The second beach was believed to have no fecal pollution sources
in the vicinity of the beach. At both beaches, three samples were collected along a transect
December 2010 25
-------
U.S. Environmental Protection Agency—Office of Water
extending from the center life guard station. All samples were collected 10 cm below the water
surface and at knee depth, chest depth, and at the furthest location (from the shore) at which
swimming was allowed. Samples were also collected at chest depth at the western and eastern
extents of the beaches. Samples were collected three times per day (9:00 a.m., 12:00 p.m., and
3:00 p.m.) on both weekend days for three consecutive weekends (total number of samples taken
at each location, n= 18).
Samples were analyzed for total coliforms, fecal coliforms, E. coli, fecal streptococci, P.
aeruginosa, and total staphylococci. The authors compared mean concentrations at sample sites
via ANOVA and determined that there was no significant difference in indicator densities related
to sample location. They speculated that the mixing in the two beaches was thorough and
cautioned against assuming homogeneous indicator density in the bathing area for beaches where
dispersion of pollution might be poor. Haack et al. (2003) also found no variation in indicator
density with water depth at point of sample collection for beaches on the Grand Traverse Bay of
Lake Michigan. In that study, samples collected at ankle depth and knee depth were compared
and no significant difference related to sample depth was noted (based on Mest) for either E. coli
or enterococci. As in the study by Brenniman et al. (1981), beaches had relatively low average
enterococci densities. In addition, Haack and colleagues noted that the beaches studied were
primarily coarse sand, a medium believed to harbor relatively low numbers of indicator bacteria.
Contrary to the lack of association between sample collection water depth and indicator density
observed by Brenniman et al. (1981), Whitman et al. (2004b) observed consistent dependence of
E. coli density with depth of water at sample locations for a Chicago Lake Michigan beach. The
difference in indicator density at different water depths was observed for samples taken at
different times of day and for samples taken under different weather conditions (sunny versus
non-sunny). Using hourly E. coli density data, Whitman and colleagues estimated the first-order
E. coli decay constant to be k = 0.468 hr"1 for samples drawn at a water depth of 45 cm and k =
0.418 hr"1 for samples drawn at a water depth of 90 cm. Note that those are net decay rates and
include effects such as sedimentation and predation in addition to inactivation (or conversion to
viable but nonculturable state). Like other researchers, Whitman et al. (2004b) observed rapid
rebound of E. coli densities at night. The authors state that the two most plausible explanations
for the rebound are replenishment from sources such as beach sands and resuscitation of
nonculturable cells. On the basis of results of in situ microcosm experiments, replenishment is
regarded as the more likely cause of nighttime rebound.
Similar to the work of Whitman et al. (2004b), EPA's BMP ACT study (USEPA 2005) found depth
of water at the sample location (referred to as a zone in the original study) was the single most
important influence on indicator density. Other factors explored in that study were horizontal
(shore-wise) location, depth at which the sample was drawn, and time of day. Beaches in that study
included a Lake Michigan freshwater beach, one beach on a slow-flowing portion of the Detroit
River, two marine beaches, and one estuarine beach. Samples were taken at 15 cm (ankle depth),
50 cm (knee depth), and 150 cm (chest depth). For freshwater sites (West Beach and Belle Isle) the
indicator was E. coli enumerated via the modified mTEC agar membrane filter method. For the
other beaches (marine and estuarine) the indicator was Enterococciis, analyzed by the mEI agar
membrane filter method (Method 1600). Geometric mean indicator densities at ankle, knee, and
chest depth for samples taken at the five beaches in the morning are shown in Figure 3, and
geometric means for samples taken at 2:00 p.m. are shown in Figure 4. Although all beaches
exhibited an association of indicator density with zone, the impact was more pronounced at
December 2010 26
-------
U.S. Environmental Protection Agency—Office of Water
o
o
E?
o
o> K -
Ankle depth
Knee depth
Chest depth
West Beach
Source: USEPA 2005
Belle Isle
Wollaston
Imperial Beach Miami Beach Park
Figure 3. Indicator density variation with water depth at sample collection, 9:00 a.m. samples,
EMPACT Study (E. col/for West Beach and Belle Isle and Enterococcus for other beaches)
o
o
en
to
-
(D
o
Ankle depth
Knee depth
Chest depth
West Beach
Source: USEPA 2005
Belle Isle
Wollaston
Imperial Beach Miami Beach Park
Figure 4. Indicator density variation with water depth at sample collection, 2:00 p.m. samples,
EMPACT Study (E. co//for West Beach and Belle Isle and Enterococcus for other beaches)
December 2010
27
-------
U.S. Environmental Protection Agency—Office of Water
beaches with higher indicator concentrations. Clear trends toward lower indicator density with
increasing depth at which sample is drawn are seen for both morning and afternoon samples.
In a reanalysis of the EMPACT Beaches study data using the random forest means of decisional
analysis, Parkhurst et al. (2005) confirmed the importance of water depth at point of sampling
and noted that the three predictors strongly related to indicator density were water depth at
sampling point, day of the week, and density 24 hours earlier. That finding is important because
it applies to the five very different beaches monitored and analyzed in the EMPACT study and
because the random forest method is believed to be an effective method for distinguishing
between explanatory variables whose effects could be nonlinear and correlated.
Kleinheinz et al. (2006) used a sampling grid similar to that employed by Brenniman et al.
(1981) in a study of five Lake Michigan beaches and five Lake Superior beaches. Samples were
taken at depths of 30 cm, 60 cm and 120 cm along a transect from the center of the beaches and
at 30 cm and 60 cm on transects at the beach edges. Samples were collected three times a week
at Lake Michigan beaches and twice a week at Lake Superior beaches. Samples were collected at
the same time of day (unspecified) at sample locations at a depth of 15 cm-30 cm below the
water surface. ANOVA was used to determine whether mean E. coli concentrations at different
depths were significantly different. Significant variation in mean E. coli density with sample
location depth was observed for 60 percent of Lake Michigan beaches and for the Lake Michigan
data pooled by sample depth. For Lake Superior beaches, only 20 percent of beaches exhibited
significant differences in indicator density with sample location depth, although pooled Lake
Superior data did show a significant difference in indicator density related to sample location
depth. Differences between Lake Superior and Lake Michigan beaches were attributed to the
relatively low density of E. coli at Lake Superior beaches.
The trend toward exponential reduction in indicator bacteria with depth of sample location was
also observed for a marine bay in Southern California (Boehm et al. 2003) and for a Lake
Michigan beach in Chicago (Whitman and Nevers 2004b). In the marine beach study, the density
of three indicator bacteria (total coliforms, E. coli and enterococci) were roughly one order of
magnitude less at locations in waist-deep water than at locations with ankle-deep water. In the
Lake Michigan study, shallower stations had consistently higher indicator counts; those higher
counts were believed to be a result of the release of indicator bacteria (E. coli) from nearshore
beach sands.
3.1.2. RIVER AND STREAM SITES
The literature review identified no studies providing detailed information on the dependence of
indicator density on depth of water at the sample collection site.
3.1.3. SUMMARY
The studies discussed indicate that, for coastal sites, there is a general trend toward decreasing
indicator density with water column depth at the sample location. That finding indicates the
importance of consistency in water column depth at sample location. Whitman and Nevers
(2004b) suggest that sampling at a shallow depth (e.g., 45 cm) results in water quality estimates
that are protective of human health, including the health of children who tend to swim at
December 2010 28
-------
U.S. Environmental Protection Agency—Office of Water
shallower depths than adults. Sampling at shallow depths, as suggested by Whitman and Nevers
(2004b) could result in an overly conservative estimate of water quality; because nearshore sands
are an important source of bacterial indicators that might not be directly related to fecal
pollution, sampling at shallow depths has the potential to overstate risk relative to measured
indicator density. On the other hand, as noted by Heaney et al. (2009), an association between
contact with sands and GI illness was observed at multiple beaches (freshwater and marine).
That finding indicates that indigenous indicator bacteria, perhaps in sands, soils, or sediments,
might be associated with nonpoint fecal pollution sources and associated pathogenic organisms.
In that and other considerations, the distinction between pathogens (whose dose is related to
health effects) and indicators (whose presence is an indication of fecal pollution but is not
associated with a dose) should be used in relating indicator density to health effects.
The relationship between indicator density and human health effects is complex, given
• That indicators are related to the presence of fecal pollution and do not cause illness,
per se.
• The non-static nature of recreational activities.
• Differences in ingestion rates among individuals and between age groups.
• Variability in indicator density, particularly variations in indicator density with depth at
which samples are collected.
It is hypothesized that (1) the best indications of water quality are those shown to correlate best
with human health outcomes in epidemiological studies and (2) that sampling/monitoring
locations should be chosen on the basis of correlation between water quality at those locations
and observed human health outcomes from epidemiological studies when this information is
available. Among epidemiology studies reporting depth at which indicator densities were
measured, associations between human health outcomes and water quality were observed when
samples were taken at both knee and waist depth. On the basis of indicator bacteria (total
staphylococci, fecal coliforms, and fecal streptococci [enterococci]) in samples collected at a
water depth of 50 cm, Seyfried et al. (1985) found obvious trends and strong correlations
between total staphylococci and fecal coliform densities and the adjusted odds of illness. There
was also an apparent trend toward increased odds of illness with increasing fecal streptococcus
density, though the correlation of illness and fecal streptococcus density was not as strong as that
for the other indicators. Wade et al. (2006, 2008) determined that the incidence of GI illness
correlated with for the geometric mean of samples collected at knee and waist depth. The
relationship between indicator density at shin and waist depth was observed for samples
analyzed via membrane filtration and via qPCR, with stronger relationships observed for the
qPCR data.
Observed relationships between indicator density at knee to waist depth and human health
effects, lower short-term variability (temporal) in indicator density at greater water depths, and
the importance of consistent sampling at a single water depth, suggest that sampling in waist-
deep water might be a practical approach that balances the need for a practical sampling location
in terms of ability to collect a sample with sampling at a depth where water quality appears to
relate to human health. That option might not be available for all settings because of surf and
other local factors. Water quality (as the geometric mean of knee and waist depth samples) was
strongly associated with odds of GI illness in children (Wade et al. 2006), indicating that
although children tend to spend more time in waters shallower than waist depth, indicator
December 2010 29
-------
U.S. Environmental Protection Agency—Office of Water
densities based on samples collected deeper than waters where children concentrate their time
are still predictive of health effects for children. Those considerations notwithstanding, note that
the prevalent practice at California Pacific Ocean beaches is to sample at ankle depth because of
the practical limitations of sampling in deeper water (Weisberg, Steven, Southern California
Coastal Water Research Project. 2010. Personal communication), which could result in overly
conservative estimates of fecal indicators.
A more recent study notes the importance of exposure to beach sands in the odds of illness for
children and adults (Heaney et al. 2009). The beach sands study implies that indicators from
sands (runoff) and possibly sediments could be associated with nonpoint fecal pollution sources.
This finding does not appear to affect the selection of in-water sampling location.
3.2. INDICATOR DENSITY VARIATION WITH DEPTH BELOW SURFACE OF
SAMPLE COLLECTION
Fewer studies describing the variation in indicator density with depth in the water column were
identified in the literature search other than those documenting indicator density variation with
water depth at the sample collection location. In this context, depth of sample collection denotes
the vertical distance into the water column from the water surface at which a sample is collected,
independently from the distance to the shoreline, while water depth at sample collection location
is directly characterized by the distance from the shore where a sample is collected, and the
swimmer's location (e.g., ankle or waist depth).
3.2.1. INLAND SITES
Indicator (E. coif) density at inland lakes was, in general, found to increase with depth at which
the sample was drawn (Canale et al. 1991; An et al. 2002; Brookes et al. 2005), although in a
study of fecal streptococci in inland lakes in Scotland, higher indicator counts were observed at
30-cm depth than at 100 cm (PHLS Water Surveillance Group 1995). In the case of the higher
indicator density near the water surface, additional information about the sample location and its
proximity to a lake inlet should be considered. Potential reasons for higher concentrations near
lake bottoms are lower temperature (and die-off), association of bacteria with particles, high
indicator density in the sediments relative to that in the water column, and density/stratification
effects. In a study of an inland lake (An et al. 2002), E. coli density was found to be generally
higher in the water column within one foot (ft) of the lake bottom than at one ft below the water
surface, although significant instances were observed in which the concentration near the water
surface was significantly higher than that near the lake bottom. Two studies (Canale et al. 1991;
Brookes et al. 2005) of inland lakes indicate the relationship between difference in temperature
between inflowing waters and lake/reservoir waters and increased indicator density at greater
depth. In both of those studies, indicator-laden influent waters had lower temperature than the
average lake and reservoir temperature. Under those conditions, plumes of influent water sank,
resulting in higher indicator density with increasing depth and higher indicator density near the
mouth of the influent river. In another study of an inland lake (Dan and Stone 1991), fecal
coliform and enterococci densities were consistently higher near the lake bottom than in surface
waters at stations along the length of the lake. The researchers attributed the higher indicator
density at greater depths to sedimentation. In a mixed-used reservoir in California, enterococci,
December 2010 30
-------
U.S. Environmental Protection Agency—Office of Water
fecal coliform, and E. coli densities were generally higher near the lake bottom than for surface
samples for all sites sampled in an intensive monitoring effort (Davis et al. 2005). That trend
held for both shallow and deep sites and during periods of stratification. In cooler months during
which there was no stratification, enterococci density did not vary with water depth.
3.2.2. COASTAL SITES
In contrast to the variation of indicator densities with depth in inland lakes and reservoirs,
Boehm et al. (2003) found that indicator density (total coliform, E. coli, and enterococci) had
little relation to the depth at which a sample was drawn in a marine coastal setting. In that study,
indicator densities corresponding to surface water and one-m depth were compared for offshore
sites and indicator densities at ankle depth and waist depth were compared for shoreline samples.
In a study of a marine beach with particularly low enterococci density, Le Fevre and Lewis
(2003) observed a statistically significant difference in enterococci density between samples
taken at a depth of 10 cm from the surface and 10 cm from the bottom in the surf zone at a water
depth of 1.0 m-1.5 m but did not observe a difference related to depth at which the sample was
taken for offshore sites. The authors surmised the difference between offshore and surf zone
sights was related to resuspension of indicators from sediments.
In EPA's BMP ACT (USEPA 2005) studies of variability at freshwater and marine beaches,
Wymer found that geometric means (over the duration of the study and over all transects for each
type of depth sample) of indicator densities at different depths were not significantly different.
However, they observed that 44 percent of the time, exceedance of the standard was observed at
one depth, it was not observed at the other, despite the number of exceedances for each of the
depths being the same.
3.2.3. SUMMARY
The variation in indicator density with depth of sample collection appears to be much smaller
than other spatial variabilities, such as variability with distance from shore (depth at location of
sample collection) or along-shore or stream-wise variability (described below). For inland lakes,
there is a general trend toward higher indicator density at the bottom of the water column,
probably because of increased bacterial mortality at the water surface and persistence of
indicator bacteria in lake sediments. In inland lakes, the configuration of influent streams and the
difference between influent water temperature and ambient lake temperature might influence the
distribution of indicators in the water column. At coastal sites, differences in indicator density
associated with depth in the water column could be assumed minor. Given those findings and the
greater likelihood of recreational activity and incidental water ingestion occurring near the water
surface, samples taken a short depth below the water surface appear to offer an adequate measure
of water quality and human exposure.
3.3. ALONG-SHORE SPATIAL VARIABILITY
Tracer experiments have demonstrated the importance of littoral drift, dispersion, mixing via rip
currents, and tidal processes in the transport of chemicals and microorganisms at coastal sites
(e.g., Clarke et al. 2007) and have shown that such processes exhibit high spatial and temporal
December 2010 31
-------
U.S. Environmental Protection Agency—Office of Water
variability. The flow features are illustrated in Figure 5. In their tracer study, Clarke and
colleagues injected dye into the discharge of streams and storm drains into the ocean and
observed the passage of the tracer at stations 25, 50, and 100 m from the dye injection point.
Tracer concentrations at the station 100 m from the dye injection point indicate bulk advective
flow and dispersion of the plume. On four different days, the rates of bulk flow and dispersion
were very different; the minimum and maximum times for the plume to pass the 100-m station
were less than 10 minutes and more than 30 minutes, respectively. Rip currents were found to
have a significant role in tracer transport along the coast. Rip currents were observed visually as
transport of the dye offshore and resulted in rapid decrease in dye concentration between
adjacent stations. Along-shore dispersion was estimated on the basis of a dye transport model
and found to be as many as four orders of magnitude higher when rip currents were observed
than when they were not.
Exchange with
water outside the
surf zone
-qTN\hdy
Figure 5. Dispersion, advection, and removal in the surf zone
Dye dispersion experiments conducted in coastal waters of Southern California beaches (Clarke
et al. 2007) indicated that tracer transport occurred via alongshore advection accompanied by
dispersion and other mixing phenomena related to rip currents. When present, rip currents were
observed to withdraw tracer from the nearshore region and reintroduced the dye to the nearshore
region at a down-current location.
December 2010
32
-------
U.S. Environmental Protection Agency—Office of Water
In their study of two Lake Erie beaches, Brenniman et al. (1981) did not find significant
differences in indicator concentration between transects at the center and edges of the either of
the beaches. In that study, one of the beaches was not near any known point fecal pollution
sources, while the other beach was in the vicinity of a large WWTP outfall and several
stormwater outfalls. Homogeneity of indicators in the waters of the beach near point pollution
sources is somewhat surprising and indicates that both alignment of fecal pollutions sources and
mixing warrant consideration when assessing whether indicators will be homogeneously
distributed in the waters. In a similar study of 10 Lake Michigan and Lake Superior beaches,
Kleinheinz et al. (2006) also found no significant variation in indicator density between
horizontal (along-shore) samples. Point and nonpoint fecal pollution sources for the 10 beaches
were not described.
In a study of three Lake Erie beaches, Bertke (2007) observed relatively low spatial variability in
samples taken at the center and limits of the beach for two beaches. At the third beach, the range
of indicator concentrations for most sampling events was 0.5 to 1.0 logs (data presented
graphically). The beach with high apparent variability was not well described in the report. The
fecal pollution sources believed to be in the vicinity of the beach were stormwater runoff and
wild birds. The author did not state whether indicator densities were consistently higher at a
location on the third beach. A sanitary survey would help in interpretation of indicator densities
at the beach with high variability.
When a point source of fecal pollution is near a bathing beach, gradients in fecal indicator
bacteria might be observed. For example, in a study of the impact of stormwater on beach water
quality (Ahn et al. 2005), significant spatial variations in total coliform, fecal coliform, and
enterococci densities were observed around the outlet of a large river into marine waters.
Indicator density was observed to be dependent on distance from the stormwater discharge, wave
direction, and transport and dilution of the fecal pollution plume by rip currents. The influence of
the stormwater discharge was observed to extend less than 5 km on each side of the river mouth
(criteria for making this determination were not described). A similar general trend in spatial
variability in indicator density related to a point source (storm drain) was observed for a marine
coastal beach (Boehm 2007). Despite the general tendency toward higher enterococci density
nearer the storm drain, Boehm noted high temporal variability in indicator density resulted in
periodic observance of lower indicator density at the location nearer the storm drain. Elmanama
et al. (2006) found consistently higher fecal indicator (fecal coliform and fecal streptococci)
densities in the vicinity of outfalls for untreated sewage than at other locations on a marine
beach.
Boehm (2003) and Boehm et al. (2005a) developed a simple model for approximating the
variation in indicator density for a beach near a point source of fecal indicators. Under the
assumptions of well-mixed water in the surf zone (along a transect), steady and known
inactivation rate, grazing rate and net along-shore transport velocity, the pollutant density, N
(organisms per water volume), varies with distance along the beach from the point source, y,
according to the relation
\n(N/N0 ) = -— [Equation 1 ]
December 2010 33
-------
U.S. Environmental Protection Agency—Office of Water
where No is indicator density at the point source discharge and leff is the effective length scale
for microbial fate and transport. As seen in equation 1, leff is the distance over which the
indicator density is reduced to e~l (37 percent) of its value at the source. Processes considered in
that study determining the reduction of indicator organisms from a point source (and determining
the effective length scale) are dilution (via transfer with waters outside the surf zone due to rip
tides), die-off, and grazing. Along-shore dispersion was not explicitly included in the model;
rather, along-shore transport was assumed dominated by littoral drift, whose transport velocity
was calculated via the Longuet-Higgens equation. Including dispersion explicitly in the
formulation would produce faster decay of the indicator density with distance from the point
source. Boehm et al. (2003, 2005a) estimated the effective indicator decay length scale by linear
regression of total coliform (Boehm 2003) and enterococci observations (Boehm et al. 2005a) to
equation 1. The effective length scale was highly variable, ranging and dependent on prevailing
wave direction and differing between beaches studied. Total coliform effective length scales near
drains were determined to be largely in the range of 3,000-4,500 m, although at least 6 percent of
estimated leff were > 10,000 m. For the second drain studied, effective length scale was much
less (typically > 2,000 m). The authors questioned the validity of the model for the second drain
because of the presence of a jetty in the study area. For Enterococcus studies, length scale for
one site during upcast flow was quite low (< 10 m). For other sites, effective length was much
different during up- and downcast periods. Under all conditions, the majority of estimates for
effective length were in the range 1,000 m < leff< 5,000 m, with some estimates as high as
13,000m.
The highest spatial variability along a coast corresponds to the lowest effective length scale. An
estimate of the variation in concentrations that can be expected at a beach similar to those studied
by Boehm (2003) and Boehm et al. (2005a) could be developed using equation 1 and for several
estimates of the effective length scale. Referring to the edge of a beach nearest a point source as
y\, the beach edge furthest from the point source as_y2, and the length (span) of the beach as
Ibeach, equation 1 leads to the following estimate of the change in indicator density along the
beach:
\n(N2 /N1) = - y*~yi = - — [Equation 2]
-------
U.S. Environmental Protection Agency—Office of Water
Spatially resolved total coliform densities along a Southern California beach were observed to
decay rapidly with distance along the beach from a point source (Chen et al. 1991). At ebb tide,
total coliform density was observed to decay from a value of > 900 MPN/100 mL at the
discharge of a duck pond to the surf zone to < 10 MPN/100 mL at a location approximately
750 m from the discharge of the duck pond. The short distance over which the fecal pollution
influenced the beach was attributed to the beginning of flood tide and dilution of high indicator
density waters.
Spatial variation in indicator density somewhat contradictory to that of equation 1 was observed
by Rosenfeld et al. (2006). In a study of fecal indicator organism variability in the vicinity of the
Santa Ana River (the presumptive point source of fecal pollution) and an offshore WWTP
outfall, Rosenfeld and colleagues found that peak indicator densities (fecal coliform, total
coliform, and enterococci) were frequently observed in a band approximately 900 m-3,700 m
north of the river and outfall. In addition, bands of elevated enterococci were observed almost
simultaneously from 2,700 m south of the river discharge into the ocean to 4,600 m north of the
river discharge. Turbow et al. (2003) also observed bands of high enterococci density north of
the Santa Ana river mouth but in their one-year study found the station with the highest mean
enterococci density was at the river mouth.
Along-shore variability in water column fecal indicators can arise from non-uniform loading of
indicators from nonpoint sources. Bonilla et al. (2007) conducted high-resolution observations of
enterococci in dry and wet sands at a marine beach to quantify the variability in indicator density
and to determine the spreading rate of enterococci from a single source such as a bird pellet.
Extreme variability in enterococci density (as organisms per 100 grams [g] of sand) was
observed over a 2-m distance. At the most extreme, densities of ND to 17,672/100 g were
observed within a 2-m distance. Those findings indicate the potential for differential loading of
beaches from nonpoint sources; although, in the same study, the variability in observed indicator
densities in the water column was far below that observed for the sands. For example, if birds or
wildlife favor a site on a beach, enterococci density at that site could be high and could
contribute to high loadings of the water column. Non-uniform loading from diverse surface water
sources of fecal pollution was observed in a study of the Rhode River sub-estuary of the
Chesapeake Bay (Carney et al. 1975). In that study, peak indicator (fecal coliform) density
occurred in different months for stations at different locations in the estuary. Stations near
streams emanating from watersheds with high agricultural use tended to have higher indicator
density during summer months, whereas the indicator density for sites near the middle of the
estuary and at the mouth had the highest observed density in springtime months.
E. coli density was observed to vary between sites on an inland lake (reservoir) where there is
significant recreational activity (including boating) (An et al. 2002). In that study, potential
causes for spatial variation in indicator density were resuspension in regions of heavy power boat
use, direct loading of fecal indicators during recreational activities, and preference of waterfowl
for certain regions in the reservoir. Other studies documenting spatial variability of indicators in
inland lakes noted a trend toward high indicator density in the vicinity of influent indicator-laden
streams and decreasing indicator density with distance from the stream (e.g., Dan and Stone
1991; Brookes et al. 2005; Davis et al. 2005). As high-indicator density waters enter lakes and
reservoirs, dilution, predation, inactivation, and sedimentation all play roles in the reduction of
indicator populations.
December 2010 35
-------
U.S. Environmental Protection Agency—Office of Water
Spatial variations in indicator density could be related to bather loading and activities such as
resuspension of indicators in sediments or liberation of indicators from sands. Experiments
conducted with volunteers indicated that mean number of enterococci shed by bathers during a
15-minute swimming episode was 5.5xl05 CPU (an estimate comparable to those made in earlier
studies) and that between 1.8 and 16 percent of the enterococci were from sands that adhered to
bathers during recreation (Elmir et al. 2007). Depending on the turbulent dispersion in swimming
areas, the density of enterococci in the vicinity of individual swimmers or groups of swimmers
might be much higher than the background density of indicators in a swimming area.
3.4. LONGITUDINAL SPATIAL VARIATION, STREAMS AND LAKES
Mixing and loading are two primary causes for longitudinal variations in microbial water quality
along streams. Variability along a stream related to loading could arise from point sources (e.g.,
Fernandez-Molina et al. 2004) or nonpoint sources (Baxter-Potter and Gilliland 1988). In their
review paper, Baxter-Potter and Gilliland (1988) note that factors such as livestock management,
manure handling, land use, and antecedent soil conditions are important in determining the
spatial and temporal distribution of indicators in agricultural watersheds. A general trend of
relatively low indicator density in headwaters and increasing indicator density with river mile
has been reported (e.g., Haack et al. 2003; Petersen et al. 2005, 2006; Shanks et al. 2006),
although land use in the headwaters can influence indicator density significantly. In an extensive
study of temperate streams in Oregon (Shanks et al. 2006), longitudinal changes (increases) in E.
coli density could be related directly to land use (concentrated animal feed operations,
residential) and point sources. Point and nonpoint fecal pollution inputs were related to
longitudinal variations in fecal coliform density on the Rio Grande River (Ryu et al. 2005), with
increases in indicator density related to runoff from dairy farms and urban areas and decreases in
indicator density associated with river reaches without an apparent indicator source.
Indicator density in the water column and sediments can vary differently along a river reach. For
example, in a 17-month study (sampling frequency not provided) Goyal et al. (1977) observed
the fecal coliform densities shown in Figure 6. Water samples were drawn at the surface and
sediment samples were collected using an Eckman dredge. Along the river reach studied, water
column fecal coliform density fell sharply between the first and second stations, possibly to a
value relatively near the background concentration (upstream of the sewage outfall). In contrast,
the sediment concentration decayed logarithmically (note that Figure 6 is a semi-log plot) with
distance from the sewage outfall and was consistently much higher than the density in the water
column.
Association of indicators with particles and settling and mixing processes might also give rise to
variation in indicator density along a river. For example, indicator (E. coli, enterococci, somatic
coliphage, Clostridium perfringens) densities and pathogen (Cryptosporidium and Giardid)
densities were measured along transects at four locations in a reservoir, including the dam
headwall (Brookes et al. 2005). Non-uniformities in indicator density observed in the reservoir
were related to the process of mixing of inflowing river water with reservoir water and settling of
particle-associated organisms near the dam headwall, a location of relatively low flow. During
the study, river temperature water was lower than that of the receiving reservoir, resulting in
stratification of influent river water. During storms, river water indicator density was elevated,
resulting in higher microorganism density deeper in the reservoir.
December 2010 36
-------
U.S. Environmental Protection Agency—Office of Water
CD
¥ -I
0)
o 8 j
S t-1
t/i
O
o
"re
o
01
0
CO
¥ -I
0
CN
¥ -J
0
o
O Sediments
v Water Column
O
O
O
v
0 50 100 150
Distance from sewage outfall (m)
Source: Goyal etal. 1977
Figure 6. Variation in fecal coliform density downstream of a sewage outfall.
Along a small stream (greatest depth < 20 cm and greatest width < 3.5 m) with no known point
fecal pollution inputs, E. coli was observed to be significantly higher at downstream sites than at
upstream ones (Byappanahalli et al. 2003). The lowest observed E. coli densities were in a marsh
and its receiving waters. Two potential causes of lower marsh concentrations are low velocities
promoting settling of particle associated bacteria and longer detention times and increased solar
radiation dose. In contrast, Roll and Fujioka (1997) observed higher indicator organism (fecal
coliforms, total coliforms, and enterococci) densities at upstream locations on a tropical
(Hawaiian) stream. The trend was attributed to the storage and growth of indicator organisms in
stream sediments and bank soils and was noted by the authors as a reason to choose Clostridium
perfringem as an alternate indicator for tropical waters.
Land use in the vicinity of first-order, tidally influenced streams in South Carolina was a strong
determinant of the spatial distribution in the streams (DiDonato et al. 2009); for second and third
order streams the relationship was weaker or nonexistent. An explanation for the greater
sensitivity of first-order streams to indicator loads is the relatively low capacity for dilution in
comparison with that in higher-order streams. The wide variation in observed indicator densities
among streams in the same watershed was also noted by Dorner et al. (2007), who ascribed the
differences to different sources and loadings associated with sub water sheds. The reach with the
highest observed E. coli densities was dominated by agricultural (livestock) land use; the second-
highest densities were observed on an urban reach with abundant ducks and geese.
December 2010
37
-------
U.S. Environmental Protection Agency—Office of Water
The types of fecal indicators can exhibit different spatial variations along streams. For example,
in tidally influenced streams (Mill et al. 2006), enterococci density was much higher in upstream
reaches of the creek than at the mouth of the creek, whereas E. coll density was more uniform.
Those trends were observed during both flood and ebb tides.
Indicator density spatial variability in estuaries often relates to the configuration of streams
discharging to an estuary or bay. Chigbu et al. (2005) noted a general trend toward lower
indicator density with distance away from the mouths of rivers discharging to an estuary in the
Gulf of Mexico. A similar trend was observed for Newport Bay in Southern California (Pednekar
et al. 2005); indicator density was highest in streams feeding the bay, lower at the mouth of the
streams, and lowest at the bay location farthest from the mouths of streams. Sayler et al. (1975)
observed the highest fecal indicator bacteria (total coliforms, fecal coliforms, and fecal
streptococci) at the discharge of the Susquehanna River to the Chesapeake Bay and indicator
density generally decreasing in the direction of the mouth of the bay. For the Sydney Harbor,
Australia, non-metric multidimensional scaling showed that indicator counts were generally
lower in the vicinity of the harbor mouth than at points further upstream (Hose et al. 2005).
Like estuaries, distribution of water quality in inland lakes is generally determined by the
configuration of inlets and outlets and the distribution of nonpoint sources of fecal indicator
bacteria. A study of five inland lakes in Scotland (PHLS Water Surveillance Group 1995)
showed that indicator bacteria (E. coli and fecal streptococci) counts were consistently higher
near inlets for all the lakes with clearly identifiable inlets. However, the spatial variability in
indicator density in lakes is less than the temporal variability associated with rainfall events.
3.5. CROSS-SECTIONAL VARIATIONS IN INDICATOR DENSITIES IN STREAMS
In a study of canal waters in coastal Texas (Goyal et al. 1977), the distribution of fecal coliforms
at the surface across a stream was found to be uniform, but the sediment fecal coliform densities
in the middle of the cross section were significantly higher than that in sediments near the banks.
Masopust (2005) performed a highly resolved survey of E. coli in the Charles River after two
storms. Transects along the river and across the river were sampled at 25-m spacing between
sample locations. Two of the cross-stream transects (D and E) were immediately downstream of
CSOs and a third transect (C) was more than 300 m downstream of the nearest CSO. Plots
showing the variation in E. coli density at the three transects for storm 1 (S1) and storm 2 (S2)
are shown in Figure 7. The two storms had very different characteristics: one storm followed an
extended rainy period and was lower intensity; the second storm followed a drier period and was
higher intensity. During and immediately after storm 1, the E. coli density was essentially
uniform across the river at all three cross-river transects. During and after storm 2, the E. coli
density was uniform across the stream for the transect significantly downstream of all CSOs, but
it was non-uniform for the transects near CSOs. That finding indicates that loading and mixing
can result in highly non-uniform distributions of indicators across rivers and that such non-
uniformity can persist some distance downstream from the source.
December 2010 38
-------
U.S. Environmental Protection Agency—Office of Water
700
600
^
E
o 500 -
400
£ 300
c
o
O
=§ 200
6
ui
100-
0
Boston
shore
SL/Transect C
Si/Transect D
Si/Transect E
S2/Transect C
S2/Transect D
S2/Transect E
0
100
20C
Kid
300 400 500
Distance from Boston shore (meters)
Source: Masopust 2005
Figure 7. E. co//density along cross-stream transects of the Charles River.
700
3.6. BEACH FEATURES PROMOTING NON-HOMOGENEITY IN INDICATOR
DENSITY
Features of beaches promoting non-uniform distribution of indicator bacteria are breakwaters or
groins, exposure to sunlight, or type of use (Bertke 2007) and degree of shelter (Yamahara et al.
2007). For example, Bordalo (2003) reports not only significant differences in bacterial water
quality, but in temperature and salinity for two beaches separated by a 250-m jetty. A schematic
drawing showing the beach and relevant features is presented in Figure 8. Observed trends at
both beaches (response to rainfall events, diurnal variation in indicator density, variations with
tidal cycle) were similar, but one beach had consistently higher indicator density. The beach with
the consistently higher density was confined on both sides by jetties, whereas the other beach
was described as more open to the ocean. Higher densities in the confined waters can be
explained by reduced dilution arising from the inhibition of mixing by the jetties.
On a Lake Michigan beach, breakwaters are also believed to influence mixing, retaining
indicators (and other pollution) originating from terrestrial sources (beach sands, runoff) and
carried in along-shore currents at Chicago beaches (Whitman and Nevers 2008). Further, among
the 23 Lake Michigan beaches studied by the researchers, E. coli densities exhibited similar time
variation at all beaches but three during a 5-year study; it was surmised that the physical features
of the three beaches, particularly the presence of breakwaters, were the cause of the different
temporal fluctuations observed at those beaches. Among the 23 beaches studied, an additional
physical feature of the beach influencing indicator density and temporal fluctuations was the
beach location relative to the mouth of the Chicago River—beaches south of the river mouth had
consistently higher E. coli densities than those north of the mouth.
December 2010
39
-------
U.S. Environmental Protection Agency—Office of Water
The mobilization of indicator
bacteria from sands and sediments
is related to waves, which, in turn,
are related to the beach physical
configuration. Yamahara et al.
(2007) used an TV-way ANOVA to
determine which factors influenced
presence/absence and density of
enterococci and E. coli in beach
sands at multiple beaches along the
California coast. Among other
factors, presence and density were
most influenced by wave action and
presence of a source. Sheltered
beaches (low wave action) with an
indicator organism source had the
highest sand enterococci densities
among beaches studied.
Again, note that the relationships
observations discussed above are
based almost entirely on the
analysis of cultural data. The extent
to which those observations can be
extrapolated to qPCR observations
has not, for the most part, been
discussed in the literature.
71
100m
Open beach p0rto
Sheltered beach
Oungo beach
PastorM beach
Primary
indicator
source
Source: Bordalo2003
Figure 8. Illustration of beach features promoting
non-uniform indicator density in parts of a beach.
3.7. INFERENCES FROM SINGLE AND COMPOSITE SAMPLES
The use of composite samples offers the potential of improved characterization of beach water
quality via use of multiple sample locations without the additional analysis costs. Concerns over
the use of composite samples generally relate to the following:
• Sampling errors (because the portions of a composite sample amount to samples of
smaller volume than typical individual samples with higher sampling error).
• Use of an arithmetic mean (the density of indicators in composite samples is the
arithmetic mean of the indicator densities of the samples composited) as a water quality
measure rather than the geometric mean.
Several studies evaluated the potential benefits of composite sampling over use of the geometric
mean of a small number of samples for assessing water quality.
In a study of three Lake Erie beaches, Bertke (2007) determined that (1) no significant difference
existed in E. coli concentration between the average of multiple point samples and the E. coli
density in a single composite sample, (2) the water quality assessment (whether the beach should
be closed) was similar when composite and multiple point sampling was employed, and (3) use
of composite samples is considerably less costly than use of multiple samples, regardless of the
December 2010
40
-------
U.S. Environmental Protection Agency—Office of Water
level of sampling. Samples in the study were drawn at a water depth of approximately one m and
were collected approximately 30 cm below the water surface. Two of the beaches investigated in
the study had relatively low spatial variation in indicator density across the beach on given
sample days; one of the beaches, however, had a 0.5- to 1-log variation in indicator density for
samples taken at multiple locations along the beach. Fecal pollution at the beach with high
spatial variability had unknown origin, but it was thought to include stormwater runoff and birds.
The observation that the arithmetic mean of multiple samples was not significantly different
from the density in the composite sample indicates that sampling variability was not an
impediment to using composite samples for these beaches. The other two observations (similar
water quality assessment for composite samples and multiple samples and lower cost of
composite sampling) suggest that composite sampling may provide a favorable policy option
balancing precise determination of water quality via geometric mean and sample analysis costs.
A more recent study of Lake Michigan beaches (Reicherts and Emerson 2009) had similar
findings to those of Bertke (2007). In their study, sampling events were conducted weekly and
composed of collecting three samples at different locations on a beach. Composite samples were
made by sampling equal volumes from the three samples collected at each sampling event. The
arithmetic average of E. coli density from the three samples used to create the composite sample
was not significantly different from that of the composite sample, indicating that sampling error
was not significant. In a retrospective analysis of water quality data from multiple Lake
Michigan beaches, Reicherts and Emerson (2009) found 26 occurrences (2 percent of sample
days) on which the geometric mean exceeded 300 CFU/100 mL (the standard used in the study).
The composite samples (arithmetic mean) also exceeded 300 CF/100 mL for all 26 samples. The
arithmetic mean exceeded the standard for an additional six sampling events (0.5 percent) on
which the geometric mean was below the standard. The authors suggest this tradeoff (0.5 percent
of events incorrectly classified versus a reduction in analysis costs by two-thirds) is favorable,
particularly because using composite samples (arithmetic means) produces a more conservative
estimate of water quality than using geometric means (the more appropriate measure of water
quality for beaches with lognormal spatial indicator distribution).
The EMPACT report (USEPA 2005) provide a mathematical analysis of differences in
composite samples and geometric means of multiple individual samples. First, the authors note
that, if the spatial distribution of indicators at a site is known to be lognormal, the geometric
mean estimates the median of the distribution while the mean is estimated by the arithmetic
mean. These two parameters of the log normal distribution are related by
Median = Mean x 10~L15I/where Fis the variance of the logic densities. That relationship could
be used to estimate the geometric mean (median) using the arithmetic mean (mean) from a
composite sample. Use of the relation requires a priori knowledge of the variance from sufficient
historical results of individual samples. The EMPACT report further suggests that the variance
be based on observations from at least 50 samples. In a subsequent analysis of the influence of
spatial variability on the use of composite sampling, Wymer (Wymer 2007) noted that the
number of composite samples required to achieve equal precision to a given number of
individual samples is a function of sampling variance (variance of logio[indicator density] per
100 mL). Note that the number of composite samples need not be constrained to the number of
individual samples, as was done in the studies by Bertke (2007) and Reicherts and Emerson
(2009). At beaches with low sampling variance (e.g., 0.1) the number of composite samples to
provide equivalent precision to three individually analyzed samples is five. In such a case,
December 2010 41
-------
U.S. Environmental Protection Agency—Office of Water
analysis costs are reduced by two-thirds. When the sampling variance is 0.3, compositing
17 samples yields an estimate with precision equivalent to that of 9 individually analyzed
samples. Although few beaches collect 9 samples for analysis, in this example, the cost of
analyzing a single composite of 17 samples is one-ninth the cost of analyzing 9 individual
samples.
In summary, composite sampling appears to offer a suitable tradeoff between analysis resource
constraints and accurate estimation of beach water quality. The results of composite samples can
be compared against samples as a conservative estimator of water quality, or they can be
converted to an equivalent geometric mean under the assumption of lognormal distribution of
indicators across the beach with known variance. In any event, locations for sampling, whether
via individual samples or composite samples, should be based on site-specific data obtained in a
sanitary survey or other means and on historic indicator data obtained from pilot studies with
high-density sampling.
December 2010 42
-------
U.S. Environmental Protection Agency—Office of Water
CHAPTER 4 Development of Monitoring
Approach
This chapter builds on the findings reported in Chapters 2 and 3 describing when, where, and
how monitoring could be conducted such that it is consistent with and accounts for the spatial
and temporal variability inherent in fecal indicator organism densities in recreational waters.
4.1. FACTORS TO CONSIDER FOR MONITORING PLAN DEVELOPMENT
Monitoring programs to evaluate health risk and demonstrate that water quality is appropriate for
the site should be designed with due consideration of the following factors:
• The variability in water quality at the site.
• The degree to which that variability is known.
• Other practical concerns such as optimizing the public health benefit of limited resources,
access to sampling points, and ability to deliver samples to laboratories within acceptable
holding times.
4.2. SITE CHARACTERIZATION
Regardless of the site type, a monitoring approach should be developed on the basis of site-
specific data for indicator variability at the site; likely fecal pollution sources; and, if possible,
the correlation of indicator density with other measurable features of the site, such as rainfall or
wind speed. Such data could be generated by using a sanitary survey, which would also be useful
in advance of a pilot monitoring study, or studies conducted for developing predictive models.
This section presents several approaches to monitoring. In all cases, the variability of indicator
density at a specific site and the number of samples taken on a sample event determine how
results of the sampling event are compared to WQSs. The variability in indicator density at a site
can be evaluated via use of sanitary surveys, developing predictive models, and pilot monitoring
programs.
4.2.1. SANITARY SURVEYS
Sanitary surveys entail identifying fecal pollution sources with the potential for affecting a site
and features of the site that can be used in risk management (USEPA 2002b). The results of the
survey can be used in developing sampling plans, in risk management, or as inputs or
information for use in predictive models (USEPA 2002b, 2008). Results from sanitary surveys
inform where to sample and can provide an indication of the minimum number of samples
required to characterize water quality on a beach.
When developing sampling plans, it is useful to identify the following elements of sanitary
surveys:
• Fecal pollution sources with the potential to load beaches unevenly.
December 2010 43
-------
U.S. Environmental Protection Agency—Office of Water
• Features hampering mixing (along the beach).
• Temporal variations in fecal pollution sources (e.g., seasonal or weekly variations in
bather density).
Suspected uneven fecal pollution loading indicates the desirability of sampling on multiple
transects along the beach, with one sample location chosen as the location either nearest the fecal
pollution source or at the location expected to receive the greatest fecal pollution loading. Other
sampling locations could be chosen either because they are at the location on the beach farthest
from the fecal pollution source (for small beaches) or at a locations typically sampled, such as
the beach center or a transect with the highest incidence of swimming. A simplified illustration
of uneven loading and associated sampling locations is presented in Figure 9. For this beach, the
location of a fecal pollution source near the beach suggests that samples should be taken at the
northernmost edge of the swimming area and at another location, such as a high swimmer
density area, or the southernmost edge of the swimming area (shown flags).
When sanitary surveys identify features that hamper mixing at a beach (e.g., jetties,
breakwaters), samples should be taken in all the distinct regions of the beach. Information from a
sanitary survey alone might not be sufficient to assess whether beach features hinder mixing
significantly. Beach managers might need to supplement the survey with a modeling study,
tracer experiments, or other activities to adequately characterize transport at a site. Figure 10
illustrates such a scenario. At that site, beaches are separated by breakwaters along the discharge
of a channel connecting Muskegon Lake with Lake Michigan. Long-shore currents and uneven
distribution of the channel effluent to the beaches, north and south of the channel, can cause
significant differences in indicator density on opposite sides of the breakwater. In such a case,
the potential for significant differences on opposite sides of the breakwater is obvious. It might
be identified in the course of a sanitary survey, perhaps by analyzing satellite images. In other
cases, the need to sample different sections of a beach might not be obvious.
Sanitary surveys also canm produce information about the preferred timing and frequency of
sampling. In a sanitary survey developed for the Great Lakes (USEPA 2008), surveyors were
asked to use historic observations to produce a qualitative assessment of the correlation between
environmental conditions and bacteria levels. Surveyors also characterized variations in
populations of swimmers, shorebirds, wildlife, and domestic animals during the bathing season.
Among those, the key data differ between sites but are generally the rainfall and swimmer
density and, for coastal sites, the wave height, wind direction, and information regarding
prevailing currents.
4.2.2. PILOT STUDIES
Historic data alone might not suffice to adequately characterize the temporal or spatial variations
of indicator density at the scales needed to develop a comprehensive monitoring plan. Therefore,
pilot monitoring studies should be considered to develop additional quantitative data to help
decide how often to sample and how many locations should be sampled per sampling event. Pilot
study designers should evaluate the heterogeneity of indicator density along a given beach.
December 2010 44
-------
U.S. Environmental Protection Agency—Office of Water
Fecal
pollution
source
Sample
locations
Figure 9.
Illustration of uneven fecal pollution loading
and potential sample locations.
Figure 10.
Illustration of beach features interfering with
mixing.
Prior research indicates that variation of indicator density with water depth at sample location
and variation in indicator density with the depth (below the water surface) at which samples are
drawn are relatively consistent among sites. Given that optimal sample collection water depth
appears to be between shin and waist depth and optimal sample collection depth appears to be
10 cm-30 cm below the water surface, most pilot studies might not require designs with
sampling at multiple water depths and sample collection depths. Exceptions include sites for
which sanitary surveys indicate spatial variability differs from typical sites or sites where
sampling might need to occur at a depth other than knee depth. For example, if sampler safety
dictates sampling at ankle depth, the temporal variability of indicator density in ankle-depth
samples should be characterized as a component of monitoring plan development.
Ideally, a pilot study would involve a relatively high density of samples taken per sample day
and would be conducted for a sufficient number of days that the conditions spanning those
expected during a recreational season would be encountered. To determine whether indicators
are homogeneously distributed at a site, one should assess the indicator density distribution
among samples taken at different along-shore positions. A Fisher's dispersion test can be used to
determine whether the samples indicate dispersion (heterogeneity). It is possible that the
inference drawn from the Fisher's dispersion test differs from one sampling event to the next.
The conservative assumption is that the indicators are not homogeneously distributed on a beach
and that, except for relatively small swimming areas, a single sample does not characterize water
quality adequately.
A pilot study can provide some of the most important information for developing operational
monitoring plans for beaches that are monitored on daily—the along-shore spatial variability of
indicators. A typical assumption is that indicators are lognormally distributed in the along-shore
December 2010
45
-------
U.S. Environmental Protection Agency—Office of Water
direction. Thus, the standard deviation for the spatial distribution of the log-transformed
indicator density could be used either to estimate the number of samples that should be collected
to achieve an estimate of the indicator density at the beach with a given precision or to estimate
the precision with which a set of one or more observations predicts the mean indicator density.
The along-shore variability of indicator density is likely to differ among sample days during a pilot
monitoring study. Several options could be used for selecting a representative along-shore variance
for use in developing sampling and analysis plans. The best estimate of variance to use is,
arguably, the variance observed when water quality for the sampling event was marginal (with
respect to a single-sample criterion). In most cases, it can be assumed that variability increased
with increasing indicator density. When water quality is poor (indicator density approaches or
exceeds a single-sample criterion) the choice of representative variance is irrelevant; regardless of
the choice of variance, the mean density of the samples indicates exceedance of the criterion.
Likewise, at low indicator density, unless a site is subject to extremely high spatial variability, the
inference from a set of samples will be that the water quality meets a criterion. Thus, the variance
at marginal water quality (e.g., 90 percent of the criterion) appears to be the critical variance for
selecting number of samples or for making inferences on the basis of a single sample.
Alternatives to using the variance at marginal quality as the characteristic variance include the
following:
• Use of an average variance based on all samples taken during the pilot study.
• Assumption that the along-shore variances for all sample days are characterized by an F-
distribution.
• Use of a choice of the characteristic variance based on the daily variance estimates and
using a confidence interval.
The latter method for estimating characteristic variance assumes the variances are for samples
for a normal (or lognormal) distribution. For along-shore variability, that is not strictly true, since
the along-shore indicator density is lognormally distributed and the temporal distribution of
indicators (distribution of the daily geometric mean) is also lognormally distributed.
Pilot studies can provide an indication of temporal variability at a site and spatial variability,
although limitations on study duration can make the historical record of indicator densities at a
site a better data set for establishing temporal variability. If other data (rainfall depths, wind
directions) are collected with indicator density data, pilot studies can be used to develop
relationships such as rainfall rating curves that associate indicator density with rainfall depth or
other event conditions. Again, because pilot studies could be limited in duration, historic
indicator records might be better suited than pilot study data for developing such curves,
assuming that environmental data are available for their corresponding indicator densities.
4.3. MONITORING APPROACHES AND STATISTICS FOR ASSESSING WATER
QUALITY
This section discusses methodologies and statistical approaches for assessing water quality.
Monitoring plans will include specification of the spatial and temporal sampling strategies. Both
December 2010 46
-------
U.S. Environmental Protection Agency—Office of Water
the spatial and temporal strategies will likely differ among the types of sites. The monitoring
approaches presented below were developed on the basis of the following considerations:
• Sites with generally good water quality should not require as much monitoring as sites
with relatively poor water quality.
• The use of confidence intervals calculated on the basis of site-specific variance in
indicator density (rather than mean or median values for a set of samples) is suggested.
• Statistical approaches used for developing monitoring plans and evaluating criteria
should be relatively simple to use.
• The approaches should be consistent with the approach that underlies the new criteria.
The first two of the above considerations suggest that monitoring schemes should be site-specific
and based on results of a pilot monitoring study or the historic record of water quality at a site.
The third and fourth considerations take into account the broad range of statistical expertise
among those charged with developing and evaluating water quality monitoring plans.
o
o
Z)
LL
O
0
T3
O
O
O
O
O
UJ
o
o -
o
CO
o
CD
O
OJ
4.3.1. STATISTICAL CONSIDERATIONS: VARIABILITY, CONFIDENCE ESTIMATES,
AND SAMPLE NUMBERS
As a preface to the approaches for developing monitoring plans, this section provides an example
from EPA's EMPACT study to illustrate the relationship between variability, confidence in
estimates of indicator density, and number of samples. Assuming indicator density data at West
Beach, one of the Great Lakes beaches
studied in the EMPACT study
(USEPA 2005), are lognormally
distributed, a single sample may
produce an estimate of indicator
density that is significantly different
from the true mean density (which is
presumably the best measure of
exposure). The uncertainty in the true
population mean decreases as the
number of samples used to estimate
water quality is increased. That trend
is illustrated in Figure 11. In Figure
11, if a set of n samples were taken at
a site with lognormally distributed
Enterococcus density and a log-mean
density of 19 enterococci/100 mL
(shown as a dotted line), the geometric
mean of the n samples would lie
within the confidence intervals shown
95 percent of the time. It is important
to note that the confidence intervals
are calculated using a ^-distribution
^
L
\
L
L
\
\
I
l '
i
A L
Criterion
Sample Event Mean
-A- Confidence interval
i 1 1 1 1 1
^ 1 A A A A A
10
Number of samples
because the number of samples used to
characterize the mean is relatively low
Figure 11.
Confidence intervals tighten with increasing
number of samples.
December 2010
47
-------
U.S. Environmental Protection Agency—Office of Water
(usually less than 30). Figure 11 shows that to state with 97.5 percent confidence that the
observed sample mean is below the criterion (35 enterococci per 100 mL; shown as a dashed
line), it would be necessary to take four samples.
Figure 11 shows that when microbial water quality is good, a relatively small number of samples
is required to conclude with a given level of confidence that the mean indicator density is below
a specified criterion. In this illustration, the criterion is associated with a confidence level. This
confidence level accounts for variability inherent at a site and might be selected on the basis of a
tradeoff between beach closures and management of human health risk.
In actual practice, the number of samples is likely to vary according to beach characteristics and
the sources of pollution at a beach. The analysis of one of the few sets of results from EPA
epidemiology studies conducted to assess the relationship between GI illness and qPCR
monitoring results has shown that six samples collected at a single sample time are usually
sufficient to maintain an adequate correlation with GI illness (Wade et al. 2006).
4.3.2. STATISTICAL CONSIDERATIONS FOR SPATIAL SAMPLING.
Building on that background information, several approaches can be used to choose the number
of samples taken per sampling event. Those approaches include the following:
• Selecting the number of samples on the basis of a power curve (considering the
difference from the criterion that must be detected and the acceptable type I and type II
errors).
• Acquiring a composite sample composed of a sufficient number of subsamples to provide
a precision approaching some specified value (e.g., an equivalent to a specified number
of individual samples).
• Selecting a small number of samples (as few as one) on the basis of economic or other
constraints.
The use of a power curve for establishing the number of samples required along a beach per
sampling event is described in detail by Wymer (USEPA 2005; Wymer, 2007). Briefly, low
(typically 0.05) and high (typically 0.95) acceptable risks are assigned to illness rates above and
below the rate selected as acceptable for primary contact recreation. Those low and high
indicator illness rates can be converted to indicator densities via health effects relations
developed in epidemiological studies, and the resulting indicator densities constitute a tolerance
interval or a detectable difference that must be possible with the number of samples selected in a
monitoring plan. Assuming the along-shore (between transect) spatial variance, V, at the site has
been established through a pilot monitoring study or via adopting a reference value for sites with
similar features and fecal pollution sources, the number of samples, «, required to limit type I
errors to a and to be consistent with a tolerance interval of L is given by (Devore 1991):
[Equation 3]
December 2010 48
-------
U.S. Environmental Protection Agency—Office of Water
where zoc is the upper a* percentile of the standard normal distribution. As described above, the
spatial variance can be estimated from data from pilot monitoring and can be chosen on the basis
of some average variance observed among sampling events or on the basis of a value observed
when water quality was marginal.
When inferences are based on a single sample, the tolerance interval cannot be specified but
rather will be determined on the basis of type I errors. If estimates of the spatial along-shore
variance are based on a relatively small number of samples, the upper confidence level estimate
for the indicator density based on a single sample in which the density was found to be x
organisms per 100 mL is given by
[Equation 4]
where X is the log-mean indicator density for the samples and, eris an estimate standard deviation
of the log-transformed densities. Because the number of samples is not chosen consistent with a
tolerance interval, the single sample results in a higher incidence of false positives (indicator
density assessed as above the standard when it is not) than if more samples were used.
As noted in the discussion of composite samples presented in Section 4.3, the precision of a
composite sample can be raised by increasing the number of sub-samples composing the
composite sample. So an alternative approach to estimating the sample size via equation 3 is to
use a composite sample with a sufficient number of sub-samples to give the same precision as
the number of individual samples as calculated in equation 3.
4.3.3. STATISTICAL CONSIDERATIONS FOR TEMPORAL SAMPLING.
As noted above, sites are assessed differently and have different temporal sampling
requirements. Sampling schemes could be regular, random, or event driven/adaptive. For sites
with many swimmers and high economic consequences associated with closing beaches, the
objective of sampling is to assess the water quality on the day the sample is drawn. That is
particularly relevant for samples analyzed via rapid methods; the value in the rapid method lies
in the ability to use information from the sample to inform swimmers of potential water quality
problems before (on the day) they swim.
Two temporal sampling approaches appear appropriate for sites with designated beaches:
sampling at a regular interval (preferably day) and event-driven sampling. Here, event-driven
sampling refers to sampling conducted when a prior indicator measurement, environmental
condition at the site, or an anomalous event such as a sewage spill indicate a high potential for
exceedance of a target indicator level. Studies that have lead to developing predictive models
(e.g., Kinzelman et al. 2004; Nevers and Whitman 2005), which demonstrated that variables such
as rainfall, wind direction, wind speed and wave height can account for a significant portion of
the variability in observed indicator densities. In event-driven sampling, the occurrence of
conditions associated with increased indicator density can be used to trigger sampling. Once
samples indicate an exceedance of the target indicator level, the site should be considered out of
compliance until a subsequent sample indicates water quality within the target level and event
conditions are no longer present. Selecting the level of an environmental variable at which
sampling should be triggered can be done on the basis of a pilot sampling study or historical data
December 2010 49
-------
U.S. Environmental Protection Agency—Office of Water
and through use of constructs such as a rainfall rating curve. In a rainfall rating curve, historical
indicator densities measured at a site are plotted against rainfall amounts. Ideally, such a curve
will demonstrate a clear relationship between indicator density and rainfall in the vicinity of the
indicator density of concern.
4.4. SUMMARY OF MONITORING APPROACHES AND CONSIDERATIONS
In summary, except at beaches with historical data on both temporal and spatial indicator
variability, development of site-specific sampling plans should be considered as part of an
approach to site characterization using a sanitary survey and a pilot monitoring study as first
steps. Those studies will identify how fecal pollution sources are aligned with respect to the
beach and whether any beach features divide the site into hydrologically distinct zones that
should be sampled separately. After variability is assessed, one can use several options for spatial
and temporal sampling to establish whether water quality meets a specified target.
On the basis of findings from the literature and analyses, summary information is provided
below for where to sample, when to sample, and how to sample.
4.4.1. REVIEW OF INDICATOR DENSITY VARIABILITY
Indicator densities exhibit high variability at multiple time and length scales. While conditions
can cause variability to differ between sites, the relative magnitudes of temporal variations in
indicator density for both coastal and inland sites at different time scales are the following:
MONTH/
SEASON
Event variability refers to the change in indicator density associated with rain events. Among the
documented temporal variabilities, event scale variability is, by far, the greatest. Indicator
organism density can change multiple logs during a single precipitation event. Indicator loading
also varies significantly during rain events, although not necessarily following the same temporal
pattern as indicator density variations. Because event variability is so great, beach managers
might use alternatives to indicator counts for assessing water quality after and during events. For
example, historic rainfall and indicator data might be used to estimate a rainfall depth threshold
at which beaches are likely to be out of compliance.
Although site variations can alter the relative dependence of indicator density on sample
location, the general dependence of indicator density variability with location for coastal sites is
expected to be the following:
December 2010
50
-------
U.S. Environmental Protection Agency—Office of Water
SITE m ALONG
FEATURESJlT SHORE
DEPTH
BELOW
SURFACE
For inland water sites, the variation is expected to be the following:
ALONG
STREAM
DEPTH
BELOW
SURFACE
The physical alignment of both coastal and inland sites with features that promote or inhibit
mixing and point sources of fecal indicator organisms plays a significant role in the distribution
of indicators at those sites. These features can be identified during a sanitary survey of
recreational sites. Site features that appear important to identify during sanitary surveys include
the presence of the following:
• Jetties, dams, or other features that influence mixing at a site or promote retention of
indicators at a site.
• Point sources, particularly POTW discharges, in the vicinity of the site.
• Nonpoint sources, particularly livestock operations and areas where wild birds and
animals congregate, near sites.
4.4.2. WHERE TO SAMPLE
Sampling locations should be selected on the basis of the ability of a small number of samples to
adequately describe water quality at the site. Site-specific sampling plans should take into
account historic water quality and variability and the presence of physical features (point
sources, bird nesting areas, structures that influence mixing, and such) known to affect
distribution of indicators. Sanitary surveys offer a vehicle for identifying important physical
features and an alternative to intensive sampling as a means for assessing spatial variability. In
general, samples should be drawn from where water quality can be best characterized. Those are
locations where indicator organisms are most likely to be associated with a fecal pollution source
(e.g., away from areas where resuspended or indigenous organisms might be suspended) and at
locations where variability is not excessive. Such an approach will optimize monitoring such that
samples properly reflect the water quality of the site and are related to health effects data.
Where to Sample: Coastal Sites
Water column depth zone. The water column depth zone can be considered an area parallel to the
shore where one collects a sample. Taking a sample in the zone where the water depth is
approximately knee deep or greater appears to offer some advantages. Indicator density tends to
December 2010
51
-------
U.S. Environmental Protection Agency—Office of Water
vary less in deeper waters than in shallower zones. In addition, correlations between indicator
organism density measured at that depth tend to have better correlations with GI illness incidence
rates. Indicator density in shallower water is higher than that in deeper areas because of the
resuspension of indicator organisms growing or sheltered in sediments. Resuspended indicators
might not be indicative of fresh fecal pollution and, therefore, samples with a high number of
resuspended organisms might not provide a good means to assess water quality. In some cases,
considerations other than the locations for optimal sampling can play a role in selecting the
appropriate sample depth zone. In California marine waters, for example, samples are taken at
ankle depth in part to protect the safety of the sampler from the threat posed by incoming waves.
Depth of sample collection below surface. Collecting one's sample near the water surface offers
some advantages. The depth for the collection device (i.e., distance below the water surface)
appears to be less critical than the depth zone (e.g. knee depth) where sampling is conducted.
Some studies have demonstrated higher indicator density near bottom sediments than in
overlying waters and their findings support sampling in the top 15 cm (~ 6 inches) of the water
column. Additional positive features of sampling near the water surface include ease of sample
collection and avoidance of water in the vicinity of sediments where resuspension of indicator
bacteria is possible.
Sample locations in the along-shore direction should be chosen on the basis of knowledge of the
mixing characteristic of the beach and location of sources of fecal contamination as determined
by sanitary surveys. When there are beach features influencing hydrodynamics (mixing), regions
of the beach with different hydraulics cannot be expected to have similar indicator densities and,
hence, should be sampled separately.
Where to Sample: Inland Sites
For streams, except in the vicinity of point sources, indicator density is expected to vary in the
downstream direction (because of indicator inactivation in the water column or resuspension
from sediments) and to be higher near the sediments than at the water surface. Sampling of
streams within the top 15 cm of water allows characterization of water quality at the most likely
site of human exposure and away from resuspended indicator organisms that might not be
indicative of recent fecal pollution events. In the downstream direction, monitoring locations can
be selected on the basis of knowledge of the location of point and nonpoint sources of pollution.
4.4.3. WHEN TO SAMPLE
Collection of samples in the morning appears to offer the best balance between practicality and
generation of data that protects human health. If culture methods are used for enumerating
indicator bacteria, morning samples could generate results that would allow posting of health
advisories the next day or two. If qPCR or other rapid methods are used, the faster evaluation of
samples might allow same day notification. However, practical limitations (such as sample
transport and other factors) could delay such notifications.
Diurnal variation in indicator density is observed in both inland and coastal waters, with the
variation in indicator density possibly higher for coastal sites where there is less shading and
greater water surface are than for inland sites.
December 2010 52
-------
U.S. Environmental Protection Agency—Office of Water
In general, when culture methods are used for indicator bacteria enumeration, high indicator
densities are observed at least until 8:00 a.m. and perhaps later for some sites. Depending on the
insolation on a given day and at a site, the lowest indicator density typically occurs between 2:00
and 3:00 p.m., according to culture results. Such a diurnal trend might not apply to indicator
density measurements made using qPCR or might apply to a different extent.
4.4.4. How TO SAMPLE
Sampling should target areas of beaches in closer proximity to fecal pollution sources and
portions of the beach with significantly different mixing should be sampled separately. From a
monitoring perspective, the most important information derived from pilot monitoring studies is
along-shore spatial variance characteristic for a site. The along-shore variance can be estimated
on the basis of an aggregate measure of daily along-shore variances observed during the pilot
study or on the basis of a characteristic variance such as the variance observed when water
quality is marginal (approaching a level of concern).
Operational beach monitoring could be configured on the basis of (1) a power curve approach
(acceptance sampling), with the number of samples selected according to site-specific variance
estimate and a tolerance interval selected using a range of tolerable risks and an epidemiological
relationship between indicator density and human health effects, or (2) a composite sampling
strategy in which the number of samples composited is chosen to provide a precision
approaching that associated with the number of samples estimated using the power curve
approach. In each case, inferences about water quality are based on comparison of some
confidence interval around sampling for a sampling event estimate of variance with a target
value. Confidence intervals about sample means are based on site-specific variance estimates
derived in site characterization.
With regard to the actual number of samples required to adequately characterize water quality for
public health protection at beaches, the number is likely to vary according to beach
characteristics and the sources of pollution at a beach.
4.4.5. How OFTEN TO SAMPLE
Although literature on this topic based on qPCR data is limited to EPA studies and those of a few
other researchers, the analysis of the results from EPA's epidemiology studies conducted to
assess the relationship between GI illness and qPCR monitoring results has shown that six
samples collected at a single sample time are usually sufficient to maintain an adequate
correlation with GI illness. Further research based on the statistical analyses of results from
qPCR water quality determinations will be required to further elucidate these relationships.
The available basis for research on all aspects of the representativeness of qPCR sampling results
will be augmented by the implementation and widespread use of qPCR for monitoring at
recreational beaches. Additional focused statistical analysis of results from qPCR water quality
determinations, including qPCR monitoring data from current epidemiology and other studies
being and recently conducted by EPA and other agencies, would further inform the temporal,
spatial, and statistical basis for sampling requirements for the protection of public health at
beaches.
December 2010 53
-------
U.S. Environmental Protection Agency—Office of Water
This page is intentionally blank.
December 2010 54
-------
U.S. Environmental Protection Agency—Office of Water
CHAPTERS References
Ahn, J.H., S.B. Grant, C.Q. Surbeck, P.M. DiGiacomo, N.P. Nezlin, and S. Jiang. 2005. Coastal
water quality impact of stormwater runoff from an urban watershed in Southern
California. Environmental Science and Technology 39(16):5940-5953.
An, Y. J., D.H. Kampbell, and G.P. Breidenbach. 2002. Escherichia coli and total coliforms in
water and sediments at lake marinas. Environmental Pollution 120(3):771-778.
Armstrong, I, S. Higham, G. Hudson, and T. Colley. 1996. The Beachwatch pollution
monitoring programme: Changing priorities to recognize changed circumstances. Marine
Pollution Bulletin 33(7-12):249-259.
Astrom, J., TJ. Pettersson, T.A. Stenstrom, and O. Bergstedt. 2009. Variability analysis of
pathogen and indicator loads from urban sewer systems along a river. Water Science and
Technology 59(2):203-212.
Baxter-Potter, W.R., and M.W. Gilliland. 1988. Bacterial pollution run-off from agricultural
lands. Journal of Environmental Quality 17(l):27-34.
Bertke, E.E. 2007. Composite analysis for Escherichia coli at coastal beaches. Journal of Great
Lakes Research 33(2):335-341.
Boehm, A.B. 2003. Model of microbial transport and inactivation in the surf zone and
application to field measurements of total coliform in northern Orange County,
California. Environmental Science and Technology 37(24):5511-5517.
Boehm, A.B. 2007. Enterococci concentrations in diverse coastal environments exhibit extreme
variability. Environmental Science and Technology 41(24):8227-8232.
Boehm, A.B., S.B. Grant, J.H. Kim, S.L. Mowbray, C.D. McGee, C.D. Clark, D.M. Foley, and
D.E. Wellman. 2002. Decadal and shorter period variability of surf zone water quality at
Huntington Beach, California. Environmental Science and Technology 36(18):3885-
3892.
Boehm, A.B., J.A. Fuhrman, R.D. Mrse, and S.B. Grant. 2003. Tiered approach for identification
of a human fecal pollution source at a recreational beach: Case study at Avalon Bay,
Catalina Island, California. Environmental Science and Technology 37(4):673-680.
Boehm, A.B., D.B. Lluch-Cota, K.A. Davis, C.D. Winant, and S.G. Monismith. 2004.
Covariation of coastal water temperature and microbial pollution at interannual to tidal
periods. Geophysical Research Letters 31(6).
Boehm, A.B., D.P. Keymer, and G.G. Shellenbarger. 2005a. An analytical model of enterococci
inactivation, grazing, and transport in the surf zone of a marine beach. Water Research
39:3565-3578.
December 2010 55
-------
U.S. Environmental Protection Agency—Office of Water
Boehm, A.B., and S.B. Weisberg. 2005b. Tidal forcing of enterococci at marine recreational
beaches at fortnightly and semidiurnal frequencies. Environmental Science and
Technology 39(15):5575-5583.
Bonilla, T.D., K. Nowosielski, M. Cuvelier, A. Hartz, M. Green, N. Esiobu, D.S. McCorquodale,
J.M. Fleisher, and A. Rogerson. 2007. Prevalence and distribution of fecal indicator
organisms in South Florida beach sand and preliminary assessment of health effects
associated with beach sand exposure. Marine Pollution Bulletin 54(9): 1472-1482.
Bordalo, A. A. 2003. Microbiological water quality in urban coastal beaches: The influence of
water dynamics and optimization of the sampling strategy. Water Research 37(13):3233-
3341.
Brenniman, G.R., S.H. Rosenberg, andR.L. Northrop. 1981. Microbial sampling variables and
recreational water quality standards. American Journal of Public Health 71(3):283-289.
Brookes, J.D., M.R. Hipsey, M.D. Burch, R.H. Regel, L.G. Linden, C.M. Ferguson, and J.P.
Antenucci. 2005. Relative value of surrogate indicators for detecting pathogens in lakes
and reservoirs. Environmental Science and Technology 39(22):8614-8621.
Byappanahalli, M.N., M. Fowler, D. Shively, and R.L. Whitman. 2003. Ubiquity and persistence
of Escherichia coll in a midwestern coastal stream. Applied and Environmental
Microbiology 69(8):4549-4555.
Canale, R.P., M.T. Auer, E.M. Owens, T.M. Heidtke, and S.W. Effler. 1991. Modeling fecal
coliform bacteria - II. Model development and application. Water Research 27(4):703-714.
Carney, J.F., C.E. Carty, and R.R. Colwell. 1975. Seasonal occurrence and distribution of
microbial indicators and pathogens in the Rhode River of Chesapeake Bay. Applied and
Environmental Microbiology 30(5):771-780.
Chen, C.W., L.E. Gomez, C.L. Chen, and D.B. Jacobsen. 1991. Investigation of beach
contamination using tracer. Journal of Environmental Engineering 117(1): 101-115.
Cheung, W.H.S., K.C.K. Chang, and R.P.S. Hung. 1991. Variations in microbial indicator
densities in beach waters and health-related assessment of bathing water quality.
Epidemiology and Infection 106(2): 3 29-3 44.
Chigbu, P., S. Gordon, and T.R. Strange. 2005. Fecal coliform disappearance rates in a north-
central Gulf of Mexico estuary. Estuarine, Coastal and Shelf Science 65(1-2):309-318.
Clarke, L.B., D. Ackerman, and J. Largier. 2007. Dye dispersion in the surf zone: Measurements
and simple models. Continental Shelf Research 27:650-669.
Corbett, S.J., G.L. Rubin, G.K. Curry, and D.G. Kleinbaum. 1993. The health effects of
swimming at Sydney beaches. The Sydney Beach Users Study Advisory Group.
American Journal of Public Health 83(12): 1701-1706.
Coulliette, A.D., and R.T. Noble. 2008. Impacts of rainfall on the water quality of the Newport
River Estuary (Eastern North Carolina, USA). Journal of Water and Health 6(4):473-482.
December 2010 56
-------
U.S. Environmental Protection Agency—Office of Water
Dan, T.B.-B., and L. Stone. 1991. The distribution of fecal pollution indicator bacteria in Lake
Kinneret. Water Research 25(3):263-270.
Davis, K.A., M.A. Anderson, and M.V. Yates. 2005. Distribution of indicator bacteria in Canyon
Lake, California. Water Research 39:1277-1288.
Devore, J.L. 1991. Probability and Statistics for Engineering and the Sciences. Duxbury Press,
Belmont, CA.
DiDonato, G.T., J.R. Stewart, D.M. Sanger, BJ. Robinson, B.C. Thompson, A.F. Holland, and
R.F. Van Dolah. 2009. Effects of changing land use on the microbial water quality of
tidal creeks. Marine Pollution Bulletin 58(1):97-106 E-published ahead of print.
Dorner, S.M., W.B. Anderson, T. Gaulin, H.L. Candon, R.M. Slawson, P. Payment, and P.M.
Huck. 2007. Pathogen and indicator variability in a heavily impacted watershed. Journal
of Water and Health 5(2):241-257.
Edwards, D.R., M.S. Coyne, T.C. Daniel, P.P. Vendrell, J.F. Murdoch, and P.A.J. Moore. 1997.
Indicator bacteria concentrations of two northwest Arkansas streams in relation to flow and
season. Transactions of the American Society of Agricultural Engineers 40(1): 103-109.
Elmanama, A.A., S. Afifi, and S. Bahr. 2006. Seasonal and spatial variation in the monitoring
parameters of Gaza Beach during 2002-2003. Environmental Research 101(l):25-33.
Elmir, S.M., M.E. Wright, A. Abdelzaher, H.M. Solo-Gabriele, L.E. Fleming, G. Miller, M.
Rybolowik, M.T. Peter Shih, S.P. Pillai, J.A. Cooper, andE.A. Quaye. 2007. Quantitative
evaluation of bacteria released by bathers in a marine water. Water Research 41(1):3-10.
Fernandez-Molina, M.C., A. Alvarez, and M. Espigares. 2004. Presence of hepatitis A virus in
water and its relations with indicators of fecal contamination. Water, Air, and Soil
Pollution 159(1): 197-208.
Fleisher, J.M. 1985. Implications of coliform variability in the assessment of the sanitary quality
of recreational waters. Journal of Hygiene 94(2): 193-200.
Fleisher, J.M. 1990. The effects of measurement error on previously reported mathematical
relationships between indicator organism density and swimming-associated illness: A
quantitative estimate of the resulting bias. International Journal of Epidemiology
19(4): 1100-1106.
Gentry, R.W., J.F. McCarthy, A.C. Layton, L.D. McKay, D.E. Williams, S.R. Koirala, and G.S.
Sayler. 2006. Escherichia coll loading at or near base flow in a mixed-use watershed.
Journal of Environmental Quality 3 5 (6): 2244-2249.
Goyal, S.M., C.P. Gerba, and J.L. Melnick. 1977. Occurrence and distribution bacterial
indicators and pathogens in canal communities along the Texas coast. Applied and
Environmental Microbiology 34(2):139-149.
Grant, S.B., J.H. Kim, B.H. Jones, T.M. Jenkins, J. Wasyl, and C. Cudaback. 2005. Surf-zone
entrainment, along-shore transport, and human health implications of pollution from tidal
outlets. Journal of Geophysical Research 110:20.
December 2010 57
-------
U.S. Environmental Protection Agency—Office of Water
Haack, S.K., L.R. Fogarty, and C.C. Wright. 2003. Escherichia coli and Enterococci at beaches
in the Grand Traverse Bay, Lake Michigan: Sources, characteristics, and environmental
pathways. Environmental Science and Technology 37(15):3275-3283.
Haramoto, E., H. Katayama, K. Oguma, Y. Koibuchi, H. Furumai, and S. Ohgaki. 2006. Effects
of rainfall on the occurrence of human adenoviruses, total coliforms, and Escherichia coli
in seawater. Water Science and Technology 54(3):225-230.
Haugland, R.A., S.C. Siefring, LJ. Wymer, K.P. Brenner, and A.P. Dufour. 2005. Comparison
of Enterococcus measurements in freshwater at two recreational beaches by quantitative
polymerase chain reaction and membrane filter culture analysis. Water Research
39(4):559-568.
He, L.-M., J. Lu, and W. Shi. 2007. Variability of fecal indicator bacteria in flowing and ponded
waters in Southern California: Implications for bacterial TMDL development and
implementation. Water Research 41(14):3132-3140.
Heaney, C.D., E. Sams, S. Wing, S. Marshall, K. Brenner, A.P. Dufour, and T.J. Wade. 2009.
Contact with beach sand among beachgoers and risk of illness. American Journal of
Epidemiology 170(2): 164-172.
Hose, G.C., G. Gordon, F.E. McCullough, N. Pulver, and B.R. Murray. 2005. Spatial and rainfall
related patterns of bacterial contamination in Sydney Harbour Estuary. Journal of Water
and Health 3(4):349-358.
Jamieson, R.C., D.M. Joy, H. Lee, R. Kostaschuk, and RJ. Gordon. 2005. Resuspension of
sediment-associated Escherichia coli in a natural stream. Journal of Environmental
Quality 34:581-589.
Jin, G., A.J. Englande, and A. Liu. 2003. A preliminary study on coastal water quality
monitoring and modeling. Journal of Environmental Science and Health, Part A
Toxic/Hazardous Substances and Environmental Engineering 38(3):493-509.
Kay, D., J. Bartram, A. Priiss, N. Ashbolt, M.D. Wyer, J.M. Fleisher, L. Fewtrell, A. Rogers, and
G. Rees. 2004. Derivation of numerical values for the World Health Organization
guidelines for recreational waters. Water Research 38(5): 1296-1304.
Kim, J.H., and S.B. Grant. 2004. Public mis-notification of coastal water quality: A probabilistic
evaluation of posting errors atHuntington Beach, California. Environmental Science and
Technology 38(9):2497-2504.
Kinzelman, J., S.L. McLellan, A.D. Daniels, S. Cashin, A. Singh, S. Gradus, and R. Bagley.
2004. Non-point source pollution: Determination of replication versus persistence of
Escherichia coli in surface water and sediments and correlation of levels to readily
measurable environmental parameters. Journal of Water and Health 2(2): 103-114.
Kleinheinz, G.T., C.M. McDermott, M.C. Leewis, andE. Englebert. 2006. Influence of sampling
depth on Escherichia coli concentrations in beach monitoring. Water Research
40(20):3831-3837.
December 2010 58
-------
U.S. Environmental Protection Agency—Office of Water
Koirala, S.R., R.W. Gentry, E. Perfect, J.S. Schwartz, and G.S. Sayler. 2008. Temporal variation
and persistence of bacteria in streams. Journal of Environmental Quality 37(4): 1559-1566.
Le Fevre, N.M., and G.D. Lewis. 2003. The role of resuspension in enterococci distribution in
water at an urban beach. Water Science and Technology 47(3):205-210.
Leecaster, M.K., and S.B. Weisberg. 2001. Effect of sampling frequency on shoreline
microbiology assessments. Marine Pollution Bulletin 42(11): 1150-1154.
Liu, L., M.S. Phanikumar, S.L. Molloy, R.L. Whitman, D.A. Shively, M.B. Nevers, DJ.
Schwab, and J.B. Rose. 2006. Modeling the transport and inactivation of E. coli and
enterococci in the near-shore region of Lake Michigan. Environmental Science and
Technology 40(16):5022-5028.
Masopust, P. 2005. High-Resolution spatial and Temporal Variability and Patterns of
Escherichia coli in the Charles River. Master of Science thesis. Northeastern University,
Civil and Environmental Engineering. Boston, MA.
McDonald, A., D. Kay, and T. Jenkins. 1982. Generation of fecal and total coliform surges by
stream flow manipulation in the absence of normal hydrometeorological stimuli. Applied
and Environmental Microbiology 44(2): 292-3 00.
Meays, C.L., K. Broersma, R. Nordin, A. Mazumder, and M. Samadpour. 2006. Diurnal
variability in concentrations and sources of Escherichia coli in three streams. Canadian
Journal of Microbiology 52(11): 1130-1135.
Menon, P., G. Billen, and P. Servair. 2003. Mortality rates of autochthonous and fecal bacteria in
natural aquatic systems. Water Research 37:4151-4158.
Mill, A., T. Schlacher, and M. Katouli. 2006. Tidal and longitudinal variation of fecal indicator
bacteria in an estuarine creek in south-east Queensland, Australia. Marine Pollution
Bulletin 52:881-891.
Nevers, M.B., and R.L. Whitman. 2005. Nowcast modeling of Escherichia coli concentrations at
multiple urban beaches of southern Lake Michigan. Water Research 39(20):5250-5260.
Noble, R.T., S.B. Weisberg, M.K. Leecaster, C.D. McGee, J.H. Dorsey, P. Vainik, and V.
Orozco-Borbon. 2003. Storm effects on regional beach water quality along the Southern
California shoreline. Journal of Water and Health 1(1):23-31.
Noble, M.A., J.P. Xu, L.K. Rosenfeld, C.D. McGee, and G.L. Robertson. 2005. Temporal and
spatial patterns for surf zone bacteria before and after disinfection of the Orange County
sanitation district effluent. Oceans 1-3:382-387.
Obiri-Danso, K., and K. Jones. 1999. Distribution and seasonality of microbial indicators and
thermophilic campylobacters in two freshwater bathing sites on the River Lune in
northwest England. Journal of Applied Microbiology 87(6):822-832.
Olyphant, G.A., and R.L. Whitman. 2004. Elements of a predictive model for determining beach
closures on a real time basis: The case of 63rd Street Beach Chicago. Environmental
Monitoring and Assessment 98(1-3): 175-190.
December 2010 59
-------
U.S. Environmental Protection Agency—Office of Water
Parkhurst, D.F., K.P. Brenner, A.P. Dufour, and LJ. Wymer. 2005. Indicator bacteria at five
swimming beaches—Analysis using random forests. Water Research 39(7): 1354-1360.
Pednekar, A.M., S.B. Grant, Y. Jeong, Y. Poon, and C. Oancea. 2005. Influence of climate
change, tidal mixing, and watershed urbanization on historical water quality in Newport
Bay, a saltwater wetland and tidal embayment in Southern California. Environmental
Science and Technology 39(23):9071-9082.
Petersen, T.M., H.S. Rifai, M.P. Suarez, and A.R. Stein. 2005. Bacteria loads from point and
nonpoint sources in an urban watershed. Journal of Environmental Engineering
131(10):1414-1425.
Petersen, T.M., M.P. Suarez, H.S. Rifai, P. Jensen, Y.C. Su, and R. Stein. 2006. Status and trends
of fecal indicator bacteria in two urban watersheds. Water Environment Research
78(12):2340-2355.
PHLS Water Surveillance Group. 1995. Preliminary study of microbiological parameters in eight
inland recreational waters. Letters in Applied Microbiology 21(4):267-271.
Reischer, G.H., J.M. Haider, R. Sommer, H. Stadler, K.M. Keiblinger, R. Hornek, W. Zerobin,
R.L. Mach, and A.H. Farnleitner. 2008. Quantitative microbial faecal source tracking
with sampling guided by hydrological catchment dynamics. Environmental Microbiology
10(10):2598-2608.
Roll, B.M., and R.S. Fujioka. 1997. Sources of faecal indicator bacteria in a brackish, tropical
stream and their impact on recreational water quality. Water Science and Technology
35(11-12):179-186.
Rosenfeld, L.K., C.D. McGee, G.L. Robertson, M.A. Noble, and B.H. Jones. 2006. Temporal
and spatial variability of fecal indicator bacteria in the surf zone off Huntington Beach,
CA. Marine Environmental Research 61(5):471-93.
Ryu, H., A. Alum, M. Alvarez, J. Mendoza, and M. Abbaszadegan. 2005. An assessment of
water quality and microbial risk in Rio Grande Basin in the United States-Mexican
border region. Journal of Water and Health 3(2):209-218.
Santoro, A.E., and A.B. Boehm. 2007. Frequent occurrence of the human-specific Bacteroides
fecal marker at an open coast marine beach: Relationship to waves, tides and traditional
indicators. Environmental Microbiology 9(8):2038-2049.
Sayler, G.S., J.D. Nelson Jr., A. Justice, and R.R. Colwell. 1975. Distribution and significance of
fecal indicator organisms in the Upper Chesapeake Bay. Applied and Environmental
Microbiology 30(4):625-638.
Seyfried, P.L., R.S. Tobin, N.E. Brown, and P.F. Ness. 1985. A prospective study of swimming-
related illness. II. Morbidity and the microbiological quality of water. American Journal
of Public Health 75(9): 1071-1075.
Shanks, O.C., C. Nietch, M. Simonich, M. Younger, D. Reynolds, and K.G. Field. 2006. Basin-
wide analysis of the dynamics of fecal contamination and fecal source identification in
Tillamook Bay, Oregon. Applied and Environmental Microbiology 72(8):5537-5546.
December 2010 60
-------
U.S. Environmental Protection Agency—Office of Water
Shehane, S.D., VJ. Harwood, I.E. Whitlock, and J.B. Rose. 2005. The influence of rainfall on
the incidence of microbial faecal indicators and the dominant sources of faecal pollution
in a Florida river. Journal of Applied Microbiology 98(5): 1127-1136.
Sinton, L.W., C.H. Hall, P.A. Lynch, and RJ. Davies-Colley. 2002. Sunlight inactivation of
fecal indicator bacteria and bacteriophages from waste stabilization pond effluent in fresh
and saline waters. Applied and Environmental Microbiology 68(3):1122-1131.
Stevenson, A.H. 1953. Studies of bathing water quality and health. American Journal of Public
Health 45:529-539.
Tiefenthaler, L.L., E.D. Stein, and G.S. Lyon. 2009. Fecal indicator bacteria (FIB) levels during
dry weather from Southern California reference streams. Environmental Monitoring and
Assessment 155(l-4):477-492.
Traister, E., and S.C. Anisfeld. 2006. Variability of indicator bacteria at different time scales in the
Upper Hoosic River watershed. Environmental Science and Technology 40(16):4990-4995.
Trowbridge, P.R., and S.H. Jones. 2009. Detecting water quality patterns in New Hampshire's
estuaries using National Coastal Assessment probability-based survey data.
Environmental Monitoring and Assessment 150(1-4):129-142.
Turbow, D.J., N.D. Osgood, and S.C. Jiang. 2003. Evaluation of recreational health risk in
coastal waters based on Enterococcus densities and bathing patterns. Environmental
Health Perspectives 111(4):598-603.
USEPA (U.S. Environmental Protection Agency). 1983. Health Effects Criteria for Marine
Recreational Waters. EPA-600/1-80-031 U.S. Environmental Protection Agency,
Cincinnati, OH.
USEPA (U.S. Environmental Protection Agency). 1984. Health Effects Criteria for Fresh
Recreational Waters. EPA-600-1-84-004 U.S. Environmental Protection Agency, Office
of Research and Development, Cincinnati, OH.
USEPA (U.S. Environmental Protection Agency). 1986. Ambient Water Quality for Bacteria.
EPA440/5-84-002 U.S. Environmental Protection Agency, Washington, DC.
USEPA (U.S. Environmental Protection Agency). 2002a. Environmental Monitoring for Public
Access and Community Tracking (EMPACT) Beaches Project: Time-Relevant Beach and
Recreational Water Quality Monitoring and Reporting. EPA/625/R-02/017 U.S.
Environmental Protection Agency, Washington, DC.
USEPA (U.S. Environmental Protection Agency). 2002b. National Beach Guidance and
Required Performance Criteria for Grants. EPA-823-B-02-004 U.S. Environmental
Protection Agency, Washington DC.
USEPA (U.S. Environmental Protection Agency). 2004. Water Quality Standards for Coastal
and Great Lakes Recreation Water. Federal Register 69(220): 67217-67243. U.S.
Environmental Protection Agency.
December 2010 61
-------
U.S. Environmental Protection Agency—Office of Water
USEPA (U.S. Environmental Protection Agency). 2005. EMPACTBeaches Project: Results
from a Study on Microbiological Monitoring in Recreational Waters. EPA 600/R-04/023
U.S. Environmental Protection Agency, Office of Research and Development, National
Exposure Research laboratory, Cincinnati, OH.
USEPA (U.S. Environmental Protection Agency). 2006a. Water Quality Standards For Coastal
Recreation Waters: Considerations for States as They Select Appropriate Risk Levels.
EPA-823-F-06-012 U.S. Environmental Protection Agency, Washington, DC.
USEPA (U.S. Environmental Protection Agency). 2006b. Water Quality Standards for Coastal
Recreation Waters: Acceptable Risk Levels in Great Lakes Waters. EPA-823-F-06-013
U.S. Environmental Protection Agency, Washington, DC.
USEPA (U.S. Environmental Protection Agency). 2007a. Critical Path Science Plan for
Development of New of Revised Recreational Water Quality Criteria. EPA 823-R-08-002
U.S. Environmental Protection Agency, Washington, DC.
USEPA (U.S. Environmental Protection Agency). 2008. Great Lakes Beach Sanitary Survey
User Manual. EPA-823-B-06-001 U.S. Environmental Protection Agency, Office of
Water, Washington, DC.
Vidon, P., L.P. Tedesco, J. Wilson, M.A. Campbell, L.R. Casey, and M. Gray. 2008. Direct and
indirect hydrological controls on E. coli concentration and loading in midwestern
streams. Journal of Environmental Quality 37(5): 1761-1768.
Wade, T. J., R.L. Calderon, E. Sams, M. Beach, K.P. Brenner, A.H. Williams, and A.P. Dufour.
2006. Rapidily measured indicators of recreational water quality are predictive of
swimming-associated gadtrointestinal illness. Environmental Health Perspectives
114(l):24-28.
Wade, T.J., R.L. Calderon, K.P. Brenner, E. Sams, M. Beach, R.A. Haugland, L. Wymer, and
A.P. Dufour. 2008. High sensitivity of children to swimming-associated gastrointestinal
illness. Epidemiology 19(3):375-383.
Walters, S.P., K.M. Yamahara, and A.B. Boehm. 2009. Persistence of nucleic acid markers of
health-relevant organisms in seawater microcosms: Implications for their use in assessing
risk in recreational waters. Water Research 43(19):4929-4939.
Whitman, R.L., M.B. Nevers, G.C. Korinek, and M.N. Byappanahalli. 2004a. Solar and temporal
effects on Escherichia coli concentration at a Lake Michigan swimming beach. Applied
and Environmental Microbiology 70(7): 4276-428 5.
Whitman, R.L., and M.B. Nevers. 2004b. Escherichia coli sampling reliability at a frequently
closed Chicago Beach: Monitoring and management implications. Environmental Science
and Technology 38(16):4241-6.
Whitman, R.L., M.B. Nevers, and M.N. Byappanahalli. 2006. Examination of the watershed-
wide distribution of Escherichia coli along southern Lake Michigan: An integrated
approach. Applied and Environmental Microbiology 72(11):7301-7310.
December 2010 62
-------
U.S. Environmental Protection Agency—Office of Water
Whitman, R.L., and M.B. Nevers. 2008. Summer E. coli patterns and responses along 23
Chicago beaches. Environmental Science and Technology 42(24):9217-9224.
Wyer, M.D., D. Kay, G.F. Jackson, H.M. Dawson, J. Yeo, and L. Tanguy. 1995a. Indicator
organism sources and coastal water quality: A catchment study on the island of Jersey.
Journal of Applied Bateriology 78(3):290-296.
Wymer, L. 2007. The lognormal distribution and use of the geometric mean and arithmetic mean
in recreational water quality measurement In Statistical Framework for Recreational
Quality Criteria andModeling, pp. 91-112. John Wiley and Sons, West Sussex, U.K.
Yamahara, K.M., A.C. Layton, A.E. Santoro, and A.B. Boehm. 2007. Beach sands along the
California coast are diffuse sources of fecal bacteria to coastal waters. Environmental
Science and Technology 41:4515-4521.
December 2010 63
------- |