£% PQA United States
Environmental Protection
hI Agency
Office of Water
EPA 822-R-24-003
April 2024
Technical Support Materials: Developing
Alternative Recreational Criteria for
Waters Contaminated by Predominantly
Non-Human Fecal Sources
-------
Technical Support Materials: Developing Alternative
Recreational Criteria for Waters Contaminated by
Predominantly Non-Human Fecal Sources
U.S. Environmental Protection Agency
Office of Water (4304T)
Office of Science and Technology
Health and Ecological Criteria Division
Washington, DC 20460
April 2024
-------
Disclaimer
While this document cites statutes and regulations that contain legally binding requirements, it does
not itself impose legally binding requirements on the U.S. Environmental Protection Agency, states,
Tribes, other regulatory authorities, or the regulated community. The EPA, state, Tribal, and other
decision-makers retain the discretion to adopt approaches on a case-by-case basis that differ from
those discussed in this document as appropriate and consistent with statutory and regulatory
requirements. This document does not confer legal rights or impose legal obligations upon any
member of the public. This document does not constitute a regulation, nor does it change or substitute
for any Clean Water Act (CWA) provision or the EPA regulations. The EPA can update this document as
new information becomes available. The EPA and its employees do not endorse any products, services,
or enterprises. Mention of trade names or commercial products in this document does not constitute
an endorsement or recommendation for use.
Acknowledgements
This document was prepared by the Health and Ecological Criteria Division, Office of Science and
Technology, Office of Water (OW) of the EPA. The OW lead for the Technical Support Materials (TSM)
is John Ravenscroft. OW staff who provided valuable contributions to the development of the TSM
include Shamima Akhter, PhD; Shari Barash; Betsy Behl (retired); Tracy Bone; Elizabeth Doyle (retired);
Susan Euling, PhD; Samantha Fontenelle; Denise Hawkins (retired); Sharon Nappier, PhD; Jamie Strong,
PhD; John Wathen, and Lars Wilcut. The agency gratefully acknowledges the valuable contributions of
EPA staff from the Regional Offices, including Diane Evans; Terry Fleming (retired); and Ann Lavaty.
Contractor support from ICF International EPA Contracts EP-C-11-005, EP-C-16-011, and
68HERC19D0003 included: Audrey Ichida, PhD; Jeffrey Soller; Alexandria Boehm, PhD; Sorina Eftim,
PhD; Kaedra Jones; Arun Varghese; Lucas Rocha Melogno, PhD; and Laura Tuhela-Reuning, PhD. The
document, including the approach and methods, presented in this document were externally peer-
reviewed by Eastern Research Group Inc. under EPA contract number EP-C-13-009 in April 2015. Peer
reviewers included: Patrick W.M.H. Smeets, PhD (Senior Scientific Researcher, Microbial Water Quality
and Health, KWR, Watercycle Research Institute; The Netherlands); Jay M. Fleischer, PhD (Associate
Professor of Public Health, College of Osteopathic Medicine, NOVA Southeastern University); and Kelly
A. Reynolds, MSPH, PhD (Associate Professor [Tenured], Community, Environment and Policy, Mel and
Enid Zuckerman College of Public Health, The University of Arizona).
-------
Table of Contents
Disclaimer ii
Acknowledgements ii
Acronyms ix
1.0 Introduction 1
1.1 Purpose and Scope 3
1.2 Approaches for Evaluating Relative Risks for Nonhuman Fecal Sources 5
1.3 Application of QMRA to Develop Alternative Criteria to Address Nonhuman Fecal Sources... 10
1.3.1 Summary of QMRA Literature Review 13
1.3.2 Overview of the TSM Framework 15
2.0 Scope and Problem Formulation: Characterizing Human Health Risks from Nonhuman
Sources of Fecal Contamination in Recreational Waters Using QMRA 18
2.1 Rationale for use of Stressors and Surrogates 21
2.1.1 Reference and Index Pathogens 23
2.1.2 Surrogates: Fecal Indicator Bacteria 49
2.2 Problem Formulation 52
2.2.1 Conceptual Model 52
2.2.2 Analysis Plan 57
3.0 Site-Specific Alternative Criteria Development for Waters Predominantly Impacted by
Nonhuman Fecal Contamination 83
3.1 Step 1: Sanitary Survey: Identify the Contributing Fecal Sources 84
3.2 Step 2: Source Confirmation: Conduct a Water Quality Study 86
3.3 Step 3: HHRA—Forward and Relative QMRA modeling 93
3.3.1 Forward QMRA Approach 1: Using Data Collected from Water Quality Monitoring 96
3.3.2 Forward QMRA Approach 2: Using a Combination of Collected and Literature-based
Data 99
3.3.3 Interpreting Results 101
3.4 Step 4: Derive Site-Specific Alternative Criteria 105
3.4.1 Reverse QMRA and Calculating the GM 108
3.4.2 Derive STV and BAV 112
4.0 Effects Characterization 114
4.1 Methods to Evaluate Health-Protective Values for Different Susceptible Subpopulations.... 114
4.1.1 Considering and Identifying Susceptible Subpopulations 114
4.1.2 Accounting for Differential Susceptibility 115
4.2 Sensitivity Analyses 119
4.2.1 Compare Gull Analysis from Soller et al. (2014) to this TSM 119
4.2.2 Compare Children's Exposure to General Population 122
4.2.3 Confirmation of Viral Etiology for Human Fecal Sources 123
4.3 Strengths and Uncertainties of Approach 126
iii
-------
4.3.1 QMRA 127
4.4 Future Directions/Research Needs 130
5.0 References 132
Appendix A: Literature Search Strategies A-l
Appendix B: QMRA Sanitary Survey Form B-l
Appendix C: Example Sampling and Analysis Plan C-l
Appendix D: Water Quality Standards Submission Checklist D-l
Appendix E: Example Code for Step 3: Approach 1 E-l
Appendix F: Example Code for Step 3: Approach 2 F-l
Appendix G: Reverse QMRA Gull Case Study G-l
Appendix H: Python Code for Appendix G H-l
Appendix I: Example Code for STV and BAV Calculation 1-1
Appendix J: Computer Code for Reverse QMRA J-l
iv
-------
Tables
Table 1-1. Example classification matrix for fecal contamination of recreational water
environments* (adapted from NHMRC [2008] and WHO [2021]) 6
Table 1-2. Selected QMRA examples from the published literature 8
Table 2-1. Estimated annual illnesses in the United States from known pathogens.a,b 24
Table 2-2. Density of reference pathogens in human and selected animal fecal waste (adapted
from Soller et al. [2010b, 2018]) 29
Table 2-3. Literature-reported prevalence (%) of reference pathogens in cattle, chicken, gulls, and
pigs (adapted from Soller et al. [2010b]) 31
Table 2-4. Fraction of human infectious strains of reference pathogens in cattle, chickens, gulls,
and pigs (adapted from Soller et al. [2010b]) 32
Table 2-5. Literature-reported dose-response relationships for reference pathogens 43
Table 2-6. Literature-reported values for the proportion of infections resulting in illness from
exposure to reference pathogens 47
Table 2-7. Literature-reported density of FIB in human and selected animal fecal waste 50
Table 2-8. Comparison of estimated or measured ingestion volumes per recreational event for
Dufour et al. (2017) and DeFlorio-Barker et al. (2017) 60
Table 2-9. Approaches 1 and 2: definitions and sources of variables used in equations 73
Table 2-10. Reverse QMRA: definitions and sources for variables in equations 81
Table 3-1. Comparing forward QMRA and water quality monitoring results to the EPA's 2012
RWQC 101
Table 3-2. Predicted GM enterococci densities (CFU per 100 mL) in mixed sources of fecal
contamination that correspond to 36 NGI per 1,000 recreators. Values presented are
for illustrative purposes and are not meant to be used as alternative criteria 108
Table 3-3. Enterococci values corresponding to the target illness rate for selected nonhuman and
human fecal mixtures 112
Table 4-1. Parameters that differ between Soller et al. (2014) and the gull case study in this TSM 120
Table 4-2. Predicted median enterococci densities that correspond to illness levels of 36 NGI per
1,000 recreators (RBT) for waters impacted by mixed sea gull and human fecal
contamination 121
Table 4-3. Enterococci RBTs (CFU/100 mL) at different fractions of human fecal contamination
(%) 122
Table 4-4. Parameter comparison between Soller et al. (2010a) and the 2022 Etiology Analysis 124
Table 4-5. Illness rate per 1,000 swimmers for the POTW effluent-based approach 125
Table A-l. 2021 literature search for selected parameters A-3
Table C-l. Sampling locations C-4
v
-------
Table C-2. Field team members C-5
Table C-3. Analytes and methods C-8
Table E-l. Key for annotated code E-l
Table E-2. Step 3, Approach 1 forward QMRA: definitions and sources for variables in equations E-l
Table F-l. Key for annotated code F-l
Table F-2. Step 3, Approach 2 forward QMRA: definitions and sources for variables in equations F-l
Table G-l. Dose-response functions and parameters (from TSM Table 2-5, Section 2.1.1.3.2) G-3
Table G-2. Probability of illness functions and parameters (from TSM Table 2-6, Section 2.1.1.3.3).... G-4
Table G-3. Pathogen and enterococci levels in fecal source (from TSM Table 2-2 in Section 2.1.1.1
and Table 2-7 in Section 2.1.2.1) G-4
Table G-4. Prevalence and infectious fraction of reference pathogens in gull contamination (from
TSM Tables 2-3 and 2-4 in Section 2.1.1.1) G-5
Table G-5. RBTs (enterococci CFU/100 mL) for waters impacted by human and gull contamination
at 36 NGI per 1,000 recreators G-ll
Table G-6. RBTs (enterococci CFU/100 mL) for waters impacted by human and nonpathogenic
contamination at 36 NGI per 1,000 recreators G-12
Table G-7. STVs (enterococci CFU/100 mL) for waters impacted by human and gull
contamination G-12
Table G-8. STVs (enterococci CFU/100 mL) for waters impacted by human and nonpathogenic
contamination G-12
Table G-9. BAVs (enterococci CFU/100 mL) for waters impacted by human and gull
contamination G-13
Table G-10. BAVs (enterococci CFU/100 mL) for waters impacted by human and nonpathogenic
contamination G-13
Table H-l. Constants for simulated human mixtures for other target illness rates and FIB H-l
Table H-2. Key for annotated code H-l
Table H-3. Appendix G Gull Case Study: definitions and sources for variables in equations H-2
Table 1-1. Key for annotated code 1-1
Table J-l. Key for annotated code J-l
Table J-2. Soller et al. (2014) reverse QMRA: definitions and sources for variables in equations J-2
vi
-------
Figures
Figure 1-1. Flow diagram for considering QMRA in developing site-specific alternative criteria 16
Figure 2-1. Framework for HHRA to inform decision-making (U.S. EPA, 2014c) 19
Figure 2-2. Factors affecting the viability of pathogens and indicators from deposition to surface
waters (adapted from Rosen [2000]) 33
Figure 2-3. Routes of avian fecal contamination at a recreational freshwater beach 34
Figure 2-4. Epidemiologic triangle (adapted from CDC [2011]) 35
Figure 2-5. Conceptual model of exposure pathways to enteric pathogens associated with fecal
contamination in surface waters while recreating 53
Figure 2-6. Overview diagram of the process discussed in this TSM for developing alternative
criteria using QMRA for waters affected by predominantly nonhuman fecal sources 62
Figure 2-7. Conceptual diagram of the overall QMRA analytical approach described in this TSM 64
Figure 2-8. Analysis plan for a forward QMRA 68
Figure 2-9. Analytical approach for "relative" QMRA anchored at a specific FIB level 75
Figure 2-10. Analytical approach for reverse QMRA 76
Figure 3-1. Information collected in Step 1 84
Figure 3-2. Water quality study 87
Figure 3-3. Step 3: HHRA using two QMRA approaches 94
Figure 3-4. An example of an integrated, multimedia modeling framework linking the problem
definition, data access, retrieval, and processing, and IEM (including the health models
used in QMRA) to quantify risk at receptor locations (adapted from Whelan et al.
[2014a]) 95
Figure 3-5. Comparison of illness risks from exposure to undiluted runoff from land-applied
manures from three different species (from Soller et al. [2015]) 99
Figure 3-6. Box and whisker plot displaying the relative QMRA probability of illness across fecal
contamination from different species (from Soller et al. [2015]). The line in the middle
of each box is the median. The upper and lower edges of each box are the 75th and
25th percentiles. The bars (whiskers) extend to the 90th and 10th percentiles. The
diamonds are at the 95th and 5th percentiles. The dotted line represents the EPA's
recommended target illness rate of 36 NGI per 1,000 recreators 104
Figure 3-7. Calculate the GM associated with the health-based goal 106
Figure 3-8. Predicted GM enterococci densities corresponding to 36 NGI per 1,000 recreators for
waters impacted by mixed sources of fecal contamination for three animal species
(CFU per 100 mL) (adapted from Soller et al. [2014]). The output from the reverse
QMRA is a FIB density that can be used as a GM corresponding to the target illness
rate and is called a RBT 107
vii
-------
Figure 3-9. Reverse QMRA results for the gull case study. Each data point on the graph represents
the enterococci GM associated with the 36 NGI per 1,000 recreator target illness rate
for varying proportions of gull and human fecal mixtures Ill
Figure 4-1. Enterococci RBTs for recreational water contaminated by gulls and humans, resulting
in 36 NGI per 1,000 recreators. Dotted horizontal line is 35 enterococci CFU/100 mL
(the RWQC magnitude) 121
Figure 4-2. Enterococci criteria levels for recreational water contaminated by gull and human
fecal mixture, resulting in 36 NGI per 1,000 recreators. Dotted horizontal line is
35 enterococci CFU/100 mL 123
Figure C-l. Flow chart of laboratories C-7
Figure C.A-1. Sampling setup at the lagoon site (there will be two identical setups, because two
filters will be used) C-16
Figure C.A-2. Sampling setup at the beach, POTW, and package plant sites (there will be two
identical setups, because two filters will be used) C-20
Figure C.E-1. UF backflush setup C-30
Figure G-l. Base analysis results—enterococci RBT for waters impacted by human and gull fecal
mixtures G-9
Figure G-2. Enterococci RBTs corresponding to 36 NGI per 1,000 recreators for the general
population using estimates of incidental ingestion from Dufour et al. (2017) and
DeFlorio-Barker et al. (2017) G-10
Figure G-3. Enterococci RBTs corresponding to 36 NGI per 1,000 recreators for the general
population and children ages 6 to 10 years G-ll
Text Boxes
Text Box 1-1. RWQC 2012 Target Illness Rates 2
Text Box 1-2. Options for Developing Water Quality Criteria 4
Text Box 2-1. Reference and Index Pathogens 25
Text Box 3-1. Decision Point Step 1 84
Text Box 3-2. Decision Point Step 2 86
Text Box 3-3. Reference Pathogens 88
Text Box 3-4. Lognormal Distribution and Geometric Mean 91
Text Box 3-5. Decision Points Step 3 93
Text Box 3-6. Decision Points Step 4 105
Text Box G-l. Risk-Based Threshold G-l
viii
-------
Acronyms
AGI acute gastrointestinal illness
AIDS acquired immune deficiency
syndrome
AWQC ambient water quality criteria
BAV beach action value
BMP best management practices
°C degrees Celsius
CAFO concentrated animal feeding
operation
CAWS Chicago Area Waterway System
CDC Centers for Disease Control and
Prevention
CFR Code of Federal Regulations
CFU colony-forming units
CI confidence interval
CM centimeters
CSO combined sewer overflow
CWA Clean Water Act
DNA deoxyribonucleic acid
DQO data quality objectives
E. coli Escherichia coli
EJ environmental justice
FFU focus-forming units
FIB fecal indicator bacteria
g gram
GIS geographic information system
GM geometric mean
H high
HCGI highly credible gastrointestinal
illness
HHRA human health risk assessment
HUS hemolytic urea syndrome
ID infectious dose ("identification"
in Appendix C)
IEM integrated environmental
modeling
ILSI International Life Sciences
Institute
IQ intelligence quotient
kDA kilodalton
L liter (or "low" in Table 2-4 or
"litter" in Table 2-7)
Ln
LOD
log
logSD
LOQ
LT2ESWTR
M
MAC
mL
MPN
MRA
MS
MST
NA
ND
NEEAR
NGI
NIST
NORS
NR
OW
PCR
PFU
pH
POTW
QMRA
qPCR
RBT
RNA
RO
RP
RWQC
SAP
natural logarithm
limit of detection
logarithm
log standard deviation
limit of quantification
Long Term 2 Enhanced Surface
Water Treatment Rule
medium
microbial water quality
assessment category
milliliters
most probable number
microbial risk assessment
Microsoft
microbial source tracking
not applicable
not detected
National Epidemiological and
Environmental Assessment of
Recreational Water
NEEAR gastrointestinal illness
National Institutes of Standards
and Technology
National Outbreak Reporting
System
none reported
Office of Water
polymerase chain reaction
plaque-forming units
potential of hydrogen
publicly owned treatment works
quantitative microbial risk
assessment
quantitative polymerase chain
reaction
risk-based threshold
ribonucleic acid
reverse osmosis
recursive partitioning
recreational water quality criteria
sampling and analysis plan(s)
ix
-------
SCCWRP
Southern California Coastal
TSM
technical support material(s)
Water Research Project
UF
ultrafilter
SD
standard deviation
UV
ultraviolet
SIC
sanitary inspection category
VBNC
viable, but nonculturable
SOP
standard operating procedure
WHO
World Health Organization
SRM
standard control material
WQC
water quality criteria
STEC
Shiga toxin-producing E. coli
WQS
water quality standard(s)
STV
statistical threshold value
WWTP
wastewater treatment plant
TMDL
total maximum daily load
X
-------
1.0 Introduction
In 2007, 43 national and international experts from academia, numerous states, public-interest groups,
the U.S. Environmental Protection Agency, and other federal agencies participated in the Experts
Scientific Workshop on Critical Research and Science Needs for the Development of New or Revised
Recreational Water Quality Criteria ("Expert Workshop") to identify research needs in the area of
recreational water quality research and implementation issues. A key recommendation from the
Expert Workshop was to develop more flexible recreational water criteria because "one size fits all"
criteria would be inadequate for both public health protection and Clean Water Act (CWA) compliance
activities (U.S. EPA, 2007a,b). There is recognition that a single indicator measured to gauge water
quality may derive from many sources (i.e., humans, animals [agricultural and wildlife], and
environmental) and the health risks to humans vary depending on which of these sources is
contributing to the indicator being measured and the pathogens that can cause illness. Public health
warning systems based on the single indicator approach do not provide an equal level of health risk
protection in all recreational waters. Notification and remediation activities, such as beach postings,
listings, and total maximum daily load (TMDL) determinations, are not necessarily focused on the
recreational waters representing the greatest potential health risk. These activities occur in
recreational waters where indicator levels exceeding standards can result from "natural" or
environmental sources and processes (U.S. EPA, 2007b).
The experts identified critical path research for supporting the development of new criteria to include
provisions that account for potential differences in human health risks associated with human versus
nonhuman sources of fecal contamination. As noted in the Experts Workshop Report (U.S. EPA, 2007b):
It is widely believed that human feces pose a larger health risk than animal feces to
swimmers and other primary contact recreational water users. This belief derives from
the basic concept that virtually all enteric pathogens of humans are infectious to other
humans, while relatively few of the enteric pathogens of animals are infectious to
humans...The bottom line is that there are few data to demonstrate whether animal
feces pose a lower, greater, or equivalent heath risk to swimmers than human feces. If
there is a difference, it would be important to know the magnitude of the difference, in
order for EPA to make appropriate criteria recommendations. The only way to get a
better sense of the health risk to swimmers posed by animal feces is to conduct
targeted studies.
To help address the identified data gap, the experts recommended enhancing epidemiological study
designs, using quantitative microbial risk assessment (QMRA) to help characterize potential human
health impacts, evaluating loading, fate and transport of pathogens, and developing quantitative
sanitary investigations. Experts also ranked agricultural animals (e.g., cattle, poultry, and sheep) and
aquatic birds as the highest priority sources to begin assessing the relative risk differential (U.S. EPA,
2007b).
In response to the Expert Workshop report recommendations and the Consent Decree and Settlement
Agreement of 2008 (U.S. EPA, 2008a),1 the EPA developed a QMRA-based framework for application to
1 The EPA agreed to "conduct QMRA (based on measurement of pathogenic organisms and indicators) to estimate illness at
a freshwater beach impacted by agricultural animal sources of fecal contamination." Case 2;06-cv-04843-PSG-JTL Document
159-3 Files 08/08/2008 Page 3 of 15.
1
-------
recreational waters and analyzed the potential human health risks from various sources of fecal
contamination impacting surface waters based on published data for zoonotic pathogens associated
with waterborne disease (U.S. EPA, 2007c, 2010a). The QMRA results from this and other follow-on
analyses conducted by the EPA and analyses conducted by others demonstrated the potential for a
lower mean probability of gastrointestinal illness, by approximately two to four orders of magnitude,
for selected animal sources of fecal contamination; also, the physical processes associated with fate,
transport, and environmental attenuation of pathogens (see Section 2.1.1.2) can influence the burden
of illness among the exposed population (Soller et al., 2010b; U.S. EPA, 2010a; WERF, 2011; McBride et
al., 2013; Soller et al., 2015). Therefore, the source, magnitude, and mechanism of the fecal loading to
a surface water influences the pathogen profile present and the potential for illness in recreators. A
fecal "source" could be human (point or nonpoint inputs) or be a specific animal (e.g., pigs) or group of
animals (e.g., wildlife). The magnitude of fecal loading refers to the amount of mass or volume of feces
entering a waterbody. The mechanism of fecal loading includes consideration of how the feces reaches
a waterbody and the receptor (e.g., dry weather point source discharge, wet weather mobilized
manure, direct fecal deposition).
The EPA's 2012 Recreational Water Quality Criteria (RWQC; Section 6.2.2) includes a discussion of
deriving site-specific alternative criteria using QMRA (U.S. EPA, 2012). Part of Section 6.2.2 of the
RWQC document highlights QMRA as a tool that could be useful for developing alternative RWQC for
sites impacted by nonhuman sources (Boehm, 2009; Dorevitch et al., 2010; Soller et al., 2010a,b; U.S.
EPA, 2007a,b).
Site-specific alternative water quality criteria (WQC) could be helpful in cases when a waterbody has
characteristics that differ from those of waterbodies that the EPA studied when developing the 2012
RWQC. For example, the epidemiological data that informed the development of the RWQC were
conducted at beaches predominantly affected by human fecal contamination. This document provides
information on how to use a risk-based approach to develop WQC that are equally health-protective as
the EPA's 2012 RWQC for waterbodies predominantly affected by nonhuman fecal loading. The
alternative WQC developed may be numerically higher or lower than the 2012 RWQC-recommended
water quality values and would be scientifically defensible and protective of the recreational
designated use. The purpose of alternative WQCs is to reflect local conditions, site-specific
characteristics, and local risk management factors that differ from the basis underlying the 2012
RWQC.
In the 2012 RWQC, the EPA recommended using criteria values associated with specific target illness
rates. This approach was taken because the EPA had translated the historically accepted target illness
rate from the previous definition of highly credible gastrointestinal illness (HCGI) to the newer
definition of illness used in the EPA's National Epidemiological and Environmental Assessment of
Recreational Water (NEEAR) study, called NEEAR gastrointestinal illness (NGI). The EPA's 2012 RWQC
geometric mean (GM) values are associated with a defined NGI rate (Text Box 1). In deriving the 2012
RWQC, the EPA relied on epidemiological data at beaches where
human fecal sources, such as poorly treated sewage and
secondary-treated and disinfected wastewater effluent, were
identified as the predominant fecal source affecting water
quality (Cabelli, 1983; Dufour, 1984; U.S. EPA, 1986, 2012; Wade
et al., 2003, 2008). Research indicates that human enteric
viruses (e.g., adenovirus, enterovirus, and norovirus) are the
Text Box 1-1. RWQC 2012 Target
Illness Rates
The target illness rate associated
with the recommended national
2012 RWQC is 32 or 36 NGI per
1,000 recreators.
2
-------
primary disease-causing agent of concern in predominately human-impacted waters (Cabelli et al.,
1982; Soller et al., 2010a; Arnold et al., 2013; Colford et al., 2012; Dorevitch et al., 2012).
However, animal feces can contain zoonotic bacterial and protozoan pathogens and waters
predominantly affected by animal fecal loading can present less potential for human health risks
compared to waters affected by human feces (WHO, 2003; Roser et al., 2006; Schoen and Ashbolt,
2010; Soller et al., 2010b). Differing characteristics of the three main pathogen classes (i.e., bacteria,
protozoa, and viruses), such as occurrence, prevalence, and infectivity, influence the potential risk of
gastrointestinal illness to those exposed to waters containing differing pathogen profiles.
Contamination patterns and mass loadings of fecal material to surface waters can differ for human
versus nonhuman fecal sources (e.g., continuous point source discharges, event-driven nonpoint
source inputs).
Delineating the potential human health risks that nonhuman sources of fecal contamination pose
relative to those risks when the fecal source is predominantly human has been a major area of
uncertainty for risk managers and regulators. Therefore, for waterbodies that are not predominantly
affected by human fecal sources, the QMRA approach described in this Technical Support Materials
(TSM) document may be used to estimate health risks for fecal loading scenarios, including for those
scenarios where epidemiological approaches have not been successful or have not been studied.
QMRA provides a scientifically defensible approach to characterize potential risks for different fecal
loading scenarios. QMRA is a formal process for quantifying or estimating the risk of illness due to
exposures to microbes. Through the process of using a microbial risk assessment (MRA) framework to
inform human health protection, diverse information pertaining to a specific pathogen or set of
pathogens is compiled in a logical, transparent, and scientifically defensible manner. Decision-makers
can use the results of QMRA to decide how to protect public health from fecal contamination,
including obtaining answers to their specific questions or examining specific exposure scenarios
(Petterson and Ashbolt, 2016). This document describes how to use QMRA to adjust the national
RWQC based on the sources of fecal indicator bacteria (FIB) and pathogens affecting a waterbody.
1.1 Purpose and Scope
This TSM supports the implementation of the 2012 RWQC using QMRA to derive site-specific
alternative WQC2 for ambient recreational waters in which the predominant contamination is from
nonhuman fecal sources (Text Box 1-2). Specifically, this TSM describes a peer-reviewed QMRA-based
process for developing alternative WQC that modify the EPA's 2012 RWQC recommended criteria for
2 The term "water quality criteria" is used in two sections of the CWA, Section 304(a)(1) and Section 303(c)(2). The term has
a different program impact in each section. In Section 304, the term represents a scientific assessment of ecological and
human health effects that the EPA recommends to states and authorized Tribes for establishing WQS that ultimately
provide a basis for controlling discharges or releases of pollutants. Ambient water quality criteria (AWQC) associated with
specific stream uses, when adopted as state or Tribal WQS under Section 303, define the maximum levels of a pollutant
necessary to protect designated uses in ambient waters. The water quality criteria adopted in the state or Tribal WQS could
have the same numerical limits as the criteria developed under Section 304. However, in many situations, states and
authorized Tribes may want to adjust water quality criteria developed under Section 304 to reflect local environmental
conditions and human exposure patterns before incorporation into WQS. When adopting their water quality criteria, states
and authorized Tribes have four options: (1) adopt the EPA's 304(a) recommendations; (2) adopt 304(a) criteria modified to
reflect site-specific conditions; (3) develop criteria based on other scientifically defensible methods; or (4) establish
narrative criteria where numeric criteria cannot be determined.
3
-------
predominantly nonhuman fecal source contributions. The
QMRA framework is based on the current state of the
science, the EPA's FIB recommendations for enterococci
or Escherichia coli (E. coli), and associated enumeration
methods (EPA Methods 1600,1603, or equivalent). In
Sections 2, 3 and 4 of this document, additional
information about the exposure scenario, descriptions of
the reference pathogens, QMRA parameters, such as
dose-response functions, information on risk
characterization, and examples of the types of useful
supporting evidence that can be included is discussed.
The intended audience of this document is risk assessors
and those who prepare or review water quality standard
(WQS) packages. This document describes the technical
steps and explains how risk assessors can perform the
analyses needed to develop alternative WQC, including
compiling supporting documentation, such as water
quality data, and identifying technical and policy decision
points for consideration within the TSM approach. This
document describes the scientific basis and process for
developing the alternative WQC, which can be considered
scientifically defensible and protective of the primary
contact recreational designated use at the same level of health protection as the 2012 RWQC
recommendations. This TSM also allows for straightforward iteration as states revisit and update WQS
based on changing needs, scientific advancement, and potential changes in fecal sources contributing
to a waterbody.
This TSM document helps users address three main questions:
• Are the predominant fecal contributions to the site from nonhuman sources? See Section 3,
Steps 1 and 2.
• What is the potential risk from recreational exposure to feces-associated waterborne
pathogens at the site? See Section 3, Step 3.
• How do I calculate alternative criteria to ensure that the proposed site-specific criteria are
protective of the primary contact recreational designated use? See Section 3, Step 4.
The EPA recommends first reading the 2012 RWQC (U.S. EPA, 2012), which describes the scientific
bases for the national recommended water quality values and the associated target illness rates. For
developing site-specific alternative criteria for other indicators and enumeration methods, please see
Site-Specific Alternative Recreational Criteria Technical Support Materials for Alternative Indicators and
Methods (U.S. EPA, 2014a).
Text Box 1-2. Options for Developing
Water Quality Criteria
Under EPA's implementing regulations for
§303 of the CWA, states must adopt WQC
that contain sufficient parameters or
constituents to protect the designated
use of a waterbody and are based on
sound scientific rationale. States may
establish WQC for waterbodies or a
portion of a waterbody and, therefore,
could establish WQC for a specific site. A
"site" may be a beach, a waterbody, a
particular watershed, or a collection of
waterbodies exhibiting similar water
quality or fecal loading characteristics.
Alternative criteria could characterize a
potentially high-risk scenario and be
applied more broadly. The applicability of
alternative criteria is a science policy
decision. The generalizability of the
results beyond the study site might
require additional information.
4
-------
Using the approach described in this document helps to develop a transparent, clear, consistent, and
reasonable3 QMRA for nonhuman fecal sources of contamination. Deriving QMRA-based WQC might also
consider policy, regulatory, and other local considerations; these evaluations should be documented in
addition to the risk assessment. Transparent documentation of the results in the supporting material for
a QMRA-based submission of a new or revised WQS will facilitate the EPA's evaluation.
1.2 Approaches for Evaluating Relative Risks for Nonhuman Fecal Sources
Similar to the EPA's recommended RWQC, recreational water recommendations and guidelines issued
by international entities typically rely on epidemiological data collected in waters affected by human
fecal contamination (EU, 2006; NHMRC, 2008; U.S. EPA, 2012; Health Canada, 2012; WHO, 2021).
There are limited examples of addressing potential risk from nonhuman fecal sources in these
recommendations and guidelines, which include indirectly addressing these sources based on a
recommended classification and assessment approach (World Health Organization [WHO] and the
National Health and Medical Research Council) and directly addressing these sources in guideline
development (New Zealand).
The WHO discusses the importance of identifying fecal sources to inform beach classification as part of
the recommended sanitary inspection survey (WHO, 2016; WHO, 2021). The inspection survey
emphasizes human fecal sources and considers animal sources as generally less important. Australian
guidelines also consider human fecal inputs the most important factor in determining the sanitary
inspection category (SIC) (NHMRC, 2008). However, nonhuman fecal sources, such as avian fecal
inputs, can be associated with elevated FIB levels in a waterbody. Other animal fecal sources, such as
those associated with animal husbandry, can contribute zoonotic pathogens to surface waters (WHO,
2021). Both WHO and Australian guidelines discuss the importance of documenting major animal fecal
sources as part of the sanitary inspection because zoonotic pathogens potentially present in animal
fecal contamination can cause an increased probability of illness in recreators.
The Australian and the WHO guidance documents classify recreational waters using a matrix combining
the SIC (i.e., a waterbody's susceptibility to fecal influence) with the microbial water quality
assessment category (MAC) (i.e., the 95th percentiles of intestinal enterococci per 100 milliliters [mL])
(Table 1-1). The MAC cut points are defined by the level of excess Gl illness risk among bathers
exposed to fecally-contaminated recreational water during epidemiological studies (e.g., category "A"
corresponds to an expected Gl Illness risk <1%) (NHMRC, 2008). The matrix emphasizes human fecal
sources relative to nonhuman sources by evaluating a waterbody's susceptibility to human fecal
contamination. When a waterbody classified as having a low or very low susceptibility to (human) fecal
influence demonstrates elevated enterococci densities (e.g., > 201/100 mL), nonsewage sources of
fecal indicators are implied and should be verified by a sanitary survey (NHMRC, 2008). The
epidemiological data underlying the MAC are typically collected under fair weather conditions. On the
other hand, exceptional circumstances, such as sewer breaks, extreme flooding, and human or
zoonotic disease outbreaks, can affect waters in any SIC or MAC. Water quality managers should be
aware of potential exceptional circumstances that could affect their waterbodies and what actions to
take in response (WHO, 2021).
3 Transparent, clear, consistent, and reasonable are the underlying principles for good risk characterization. The elements of
a sound risk characterization (e.g., key findings, policy choices, uncertainty, and variability) describe, straightforwardly, the
critical points that make it valuable in decision-making (U.S. EPA, 2000a).
5
-------
Table 1-1. Example classification matrix for fecal contamination of recreational water environments*
(adapted from NHMRC [2008] and WHO [2021]).
Sanitary
inspection
category
(Susceptibility to
fecal influence)
Microbial water quality assessment category
(95th percentiles of intestinal enterococci/100 mL)
A
B
C
D
Exceptional
<40
41-200
201-500
>500
circumstancesG
Very low
Very good
Very good
Follow upb
Follow upb
Low
Very good
Good
Follow upb
Follow upb
Moderate
Good3
Good
Poor
Poor
ACTION
High
Good3
Fair3
Poor
Very poor
Very high
Follow up3
Fair3
Poor
Very poor
Exceptional
circumstancesG
ACTION
Notes:
* In certain circumstances, there may be a risk of transmission of pathogens associated with more severe health effects through
recreational water use. The human health risk depends greatly on specific (often local) circumstances. Public health authorities should
be engaged in identifying and interpreting such conditions.
a. Indicates possible discontinuous/sporadic contamination (often driven by results such as rainfall). This is most commonly associated
with the presence of sewage-contaminated stormwater. These results should be investigated further, and initial follow-up should
include verification of the SIC and ensuring that samples recorded include "event" periods. Confirm analytical results and review
possible analytical errors.
b. Implies nonsewage sources of fecal indicators (e.g., livestock), which need to be verified.
c. Exceptional circumstances are known periods of higher risk, such as during an outbreak involving a human or other pathogen that may
be waterborne (e.g., avian botulism—where outbreaks of avian botulism occur, swimming or other aquatic recreational activities
should not be permitted), or the rupture of a sewer in a recreational water catchment area, etc. Under such circumstances, the
classification matrix may not fairly represent risk/safety. Goals for actions associated with these circumstances can prevent exposure
and/or remediate the hazard. WHO (2021) discusses approaches for identifying and communicating a predicted or detected hazardous
condition and implementing an incident response plan among the relevant agencies involved in order to address a deterioration in
water quality.
Risks from nonhuman fecal sources can be influenced, however, by the magnitude of fecal loading.
New Zealand's Microbiological Water Quality Guidelines for Marine and Freshwater Recreational Areas
(New Zealand, 2003) document also discusses the use of the SIC and MAC to assess the suitability of a
site for recreation but adapted with local health and water quality data. Guidelines for marine waters
were informed by epidemiological investigations at seven different types of beaches: two control
(minimal fecal impact), two rural (animal waste inputs), and three human-affected (inputs from
oxidation ponds) (McBride et al., 1998). The study reported an association between illness and culture-
enumerated enterococci, but no statistical difference in illness risk based on the source of fecal
material (i.e., human or animal) was identified. Illness risks were significantly lower at the control
beaches compared to the other beaches (McBride et al., 1998). The freshwater guidelines were
developed based on the results of a water quality study, including monitoring data for 10 fecal
indicator organisms and pathogens, and the QMRA (McBride et al., 2002; Till et al., 2008). Twenty-five
sites, representing fecal inputs from birds, dairy farming, humans, and sheep, as well as
undeveloped/forested areas, across New Zealand were routinely monitored for 15 months. Results of
6
-------
the risk assessment demonstrated the importance of Campylobacter4 and human adenoviruses as
causes of waterborne illness in recreational waters. The authors estimated 4% of all notified cases of
campylobacteriosis in New Zealand could be attributed to recreational exposures (McBride et al.,
2002). The potential for increased Campylobacter infection was used to derive the guideline E. coli
value for freshwaters (New Zealand, 2003).
Numerous published QMRA examples have been conducted to address gaps in the available data and
to aid in decision-making associated with water quality in the peer-reviewed scientific literature.
Selected QMRAs from the literature addressing risks from nonhuman fecal sources, deriving risk-based
threshold (RBT) values for an indicator or source marker, or using QMRA outputs to inform risk
management practices are included in Table 1-2. These examples highlight the utility of QMRA to add
additional insight for risk management in waters not represented by traditional epidemiological-based
approaches or where it may be impractical to conduct epidemiological studies. Further, combining the
use of epidemiological and QMRA approaches can improve the interpretation of empirical health data
and provide additional context for risk management decisions (Till and McBride, 2004; Soller et al.,
2016) because recreational water epidemiological studies do not characterize the pathogens present in
feces-contaminated waters that could lead to illness and, in some cases, epidemiological approaches
fail to establish predictive relationships between the FIB and reported illnesses.
In addition to the summarized literature studies included here (Table 1-2), several other EPA resources
provide information about MRA, the development of the QMRA framework, and assessing risk from
nonhuman fecal sources in surface waters, including:
1. Review of zoonotic pathogens in ambient waters and an overview of the potential risks of
animal fecal contamination (U.S. EPA, 2009b).
2. Summary of studies reporting the prevalence of Cryptosporidium, Giardia, Campylobacter,
Salmonella, and E. coli 0157:1-17 in cattle, swine, poultry, other domestic animals, wildlife, and
environmental samples (U.S. EPA, 2010a).
3. QMRA to estimate illness at freshwater beaches impacted by agricultural animal sources of
fecal contamination (U.S. EPA, 2010a).
4. Summary of the farm factors that influence the presence of zoonotic pathogens in livestock
(U.S. EPA, 2010a: Annex 3).
5. QMRA model to evaluate the relative impacts to human health risks from animal-impacted
recreational waters (U.S. EPA, 2010a: Annex 2).
6. Microbiological Risk Assessment Tools, Methods, and Approaches for Water Media (U.S. EPA,
2014b) and an interagency Microbial Risk Assessment Guideline Pathogenic Microorganisms
with Focus on Food and Water (U.S. EPA and USDA, 2012).
4 Both the 1998 and 2002 study reports included sites highly affected by animal fecal loading due to direct deposition and
runoff containing animal wastes in the watersheds examined. For example, about 11 million cattle and up to
50 million sheep were resident in New Zealand compared to approximately 3.5 million people (McBride et al., 1998). Sheep
are known reservoirs of Campylobacter (Yang et al., 2014).
7
-------
Table 1-2. Selected QMRA examples from the published literature.
Study
Approach
Pathogens included
Indicators included
Relevance
Boehm et
al. (2015)
Derived risk-based values for
human fecal source markers,
HF183 and HumM2.
Salmonella spp.;
Campylobacter; E. coli
0157:1-17; Cryptosporidium;
Giardia; norovirus
HF183, HumM2
Recreational exposures modeled. Target illness
rate = 3% NGI. Risk-based approach informs interpretation
of data for human fecal source markers, HF183 and
HumM2, from water affected by raw sewage.
Epidemiological data for the indicators are not available.
Boehm et
al. (2018)
Derived a risk-based value for
HF183.
Salmonella spp.;
Campylobacter; E. coli
0157:1-17; Cryptosporidium;
Giardia; norovirus
HF183
Compiled decay rate constants for pathogens and indicator
into QMRA. Presented the effect of aging sewage
contamination on the RBT for HF183.
Brown et al.
(2017a,b)
Derived a RBT for the gull fecal
marker, Catellicoccus (CAT).
Salmonella and
Campylobacter
CAT
Recreational exposure modeled. Target illness
rate = 3% NGI. Addresses nonhuman fecal source, gulls,
which can contribute substantially to FIB levels, but not
necessarily risk.
Goh et al.
(2021)
Derived risk-based values for
human fecal source markers in
tropical watershed.
Adenovirus, norovirus
Enterococci, human
polyomavi ruses,
Bacteroides
thetaiotaomicron,
Methanobrevibacter
smithii
Recreational exposure modeled. Validated specificity and
sensitivity of human markers in tropical watersheds.
Identified threshold levels of the markers corresponding to
3.6% NGI. Included decay and dilution in modeling.
McBride et
al. (2013)
Characterized potential illness
risks from exposure to
stormwater discharges from
human and nonhuman fecal
sources.
Salmonella spp.;
Campylobacter,
Cryptosporidium; Giardia;
norovirus, rotavirus,
enterovirus, adenovirus
Not reported
Reported risk profiles for children's ingestion exposure.
Reported risks associated with short-term, event-based
scenarios. Human-associated viruses dominated the
predicted risks in discharges. Reference pathogens
associated with nonhuman sources had lower risk profiles.
Roser et al.
(2006)
QMRA results were used to
inform and adapt risk
management approaches to
reduce bather exposure to fecal
pathogens.
Cryptospori di um,
Salmonella, Campylobacter
E. coli, enterococci, fecal
coliforms, Clostridium
perfringens, sulfate-
reducing Clostridia.
During dry weather, wild birds were predominant fecal
sources. Results demonstrated that human sources
became notable after > 10 mm rainfall in 24 hours. Results
indicated that recreational activities would need to be
controlled after larger runoff events due to human
contamination.
Schoen and
Ashbolt
(2010)
Characterized potential illness
risk from sewage and gull fecal
sources at a recreational beach.
Campylobacter jejuni,
Salmonella enterica,
norovirus, Giardia
intestinalis, and
Cryptosporidium spp.
Enterococci
Risks from gull feces much lower than sewage at the same
level of indicator. Campylobacter dominated the risk from
gulls while norovirus dominated the risk from sewage.
8
-------
Study
Approach
Pathogens included
Indicators included
Relevance
Soller et al.
(2010b)
Characterized potential risks
from direct deposition of gull,
chicken, pig or cattle fecal
contamination compared to
human fecal sources in a
recreational water.
Norovirus, Cryptosporidium
spp., Giardia lamblia,
Campylobacter jejuni,
Salmonella enterica, E. coli
0157:H7
E. coli, enterococci
Gastrointestinal illness risks from fresh cattle feces not
significantly different from human fecal loading. Fresh gull,
chicken, or pig feces were associated with substantially
lower risks.
Soller et al.
(2014)
Evaluated the influence of
multiple sources of enterococci
in recreational water bodies on
potential human health risk.
Compared human and animal
sources, human and
nonpathogenic sources, and
animal and nonpathogenic
sources of enterococci.
Norovirus, Cryptosporidium
spp., Giardia lamblia,
Campylobacter jejuni, E.
coli0157:H7, Salmonella
enterica
Enterococci
Risks vary with the proportion of enterococci derived from
mixtures of human, nonhuman and nonfecal sources.
Study reports corresponding enterococci densities for
these source combinations equivalent to the target illness
rate of the RWQC. Risks are influenced by the potency of
the source (e.g., human) rather than the source
contributing the most indicator. For mixtures with low
human contributions, corresponding enterococci densities
could be substantially greater than the RWQC.
Soller et al.
(2015)
Characterized potential risks
from exposure to freshwater
affected by rainfall-induced
runoff containing agricultural
animal fecal material.
Cryptosporidium spp.,
Giardia lamblia,
Campylobacter jejuni, E.
coli 0157:H7, Salmonella
enterica
E. coli, enterococci
Risks from animal fecal sources at least an order of
magnitude lower than the target illness rate in the RWQC.
Equivalently protective enterococci densities in waters
receiving these animal fecal sources could be significantly
higher.
Till et al.
(2008)
Evaluated microbial indicators
and pathogens at 25 freshwater
recreational and water supply
sited throughout New Zealand.
Campylobacter,
Cryptosporidium, Giardia,
human enteroviruses,
human adenoviruses,
Salmonella
Clostridium perfringens,
E. coli, F-RNA
bacteriophage, somatic
coliphage
Campylobacter and human adenoviruses were most likely
to cause waterborne illness in recreational freshwaters.
Calculated a threshold value for E. coli as an indicator of
increased campylobacteriosis. Results were used to inform
national water quality guidelines in New Zealand.
U.S. EPA
(2010a)
Estimated illness in freshwater
affected by selected agricultural
animal fecal sources.
Salmonella spp.;
Campylobacter; E. coli
0157:H7; Cryptosporidium;
Giardia; norovirus
E. coli, enterococci
Recreational exposure modeled. Target illness
rates = 0.8% HCGI. Modeled direct deposition and runoff
from land-applied manures. Median risks from animal fecal
sources are generally lower than human fecal sources at
the same level of FIB. Median risks from direct deposition
of cattle manure equal to RWQC illness rate.
9
-------
1.3 Application of QMRA to Develop Alternative Criteria to Address Nonhuman
Fecal Sources
Recreational exposure to waters containing fecal contamination can be a public health concern due to
the potential for adverse health effects, including gastrointestinal illness, respiratory illness, and other
health endpoints (NRC, 2004; Craun et al., 2005; Wade et al., 2006, 2008, 2010; Fleming et al., 2006,
2008; Parkhurst et al., 2007). Microbial hazards in recreational water contaminated by feces can
include pathogenic bacteria, viruses, and parasitic protozoa of human and/or animal origin. Risks to
swimmers can differ depending on the source of the fecal loading (i.e., human or animal) because
(1) the pathogens in animal manure differ in type, occurrence, and abundance from those in human
sewage (WHO, 2004) and (2) the pathways by which human-infectious pathogens of animal origin
(zoonoses) reach swimmers can differ from the routes of human enteric pathogens (e.g., intermittent
rainfall with mobilization and overland transport as compared to wastewater treatment plant (WWTP)
effluent with the relatively constant flow) (U.S. EPA, 2010a).
The potential human health impacts from recreational exposure to waters contaminated by feces have
been largely informed by the available health surveys and epidemiological studies since the 1940s and
1950s (Dufour and Schaub, 2007). A common approach in a number of these studies has been to assess
whether there is an association between water quality (as measured by culture-enumerated FIB used
as a metric of the magnitude of fecal contamination) and reported human health effects. For example,
health studies conducted in the United States in the early 1950s provided evidence of a detectable
health effect when total coliform densities were observed above approximately 2,300 colony-forming
units (CFU) per 100 mL (Stevenson, 1953). In the late 1970s and early 1980s, the EPA conducted
epidemiological studies at freshwater and marine beaches primarily impacted by human fecal
contamination that demonstrated a positive statistical association between reported health effects
over a range of water quality measured by culturable E. coli and enterococci (Cabelli, 1983; Dufour,
1984). More recently, the EPA completed the NEEAR study at freshwater and marine beaches primarily
impacted by secondary treated and disinfected WWTP effluent (Wade et al., 2006, 2008, 2010; U.S.
EPA, 2010c) and examined the associations between rapid, gene-based enumeration methods and
incidence of gastrointestinal illness. The sites characterized in these studies were chosen because
human fecal sources contain human pathogens that are infective to other humans; thus, an observable
effect was expected. Additionally, wastewater treatment more effectively reduces culturable FIB
compared to viral and protozoan pathogens and recreational waters receiving this type of human fecal
source can have elevated illness at lower levels of FIB (Wade et al., 2006, 2008, 2010, 2022; U.S. EPA,
2007a; Petrinca et al., 2009; Arnold et al., 2016; Rames et al., 2016; Sidhu et al., 2018; Teixeira et al.,
2020). Surface waters affected by these human fecal sources, therefore, represent a scenario of one
type of public health concern (i.e., human exposure to human-excreted pathogens) and a potentially
higher probability of gastrointestinal illness per level of culturable FIB compared to other types of fecal
sources.
Due to the differences in pathogen profiles between human and non-human sources and the pathways
those sources can take to reach surfaces waters, it can be technically and logistically difficult to
conduct epidemiology studies in waters predominantly affected by animal sources of fecal
contamination because, but not limited to, nonpoint fecal sources can be associated with wet weather,
different fate and transport behaviors between FIB and pathogens associated with nonpoint fecal
sources, and beaches without human fecal sources may not visited by high numbers of bathers.
10
-------
Epidemiological data collected from predominantly nonhuman or nonpoint source-impacted waters
have not consistently demonstrated a statistically significant association between water quality, as
measured by culturable FIB, and human illness (Calderon et al., 1991; Priiss, 1998; Haile et al., 1999;
Wade et al., 2003; Colford et al., 2007, 2012; Dufour et al., 2012). One possible reason is that
differences in the source, magnitude, and mechanism of loading of a nonhuman fecal source to a
waterbody in the different studies influence the fate and transport of pathogens in the environment
and their effect on human illness. For example, waters with high animal waste runoff and direct
deposition might have a similar potential to cause human health effects as human fecal-impacted
waters (McBride et al., 1998; Soller et al., 2010b). Thus, the lack of a predictive epidemiology-based
association between FIB and adverse human health effects does not mean no potential for human
health effects exists. Indeed, some epidemiological evidence shows that swimmers report an increased
rate of illness compared to nonswimmers in nonpoint source or nonhuman fecal-impacted waters,
although the reported illnesses were not statistically associated with the measures of water quality at
the beaches studied (Calderon et al., 1991; Colford et al., 2007, 2012; Fleisher et al., 2010; Sinigalliano
et al., 2010; Dufour et al., 2012; Arnold et al., 2013). As indicated by this uncertainty, epidemiology-
based approaches have mainly been uninformative regarding the extent of health effects in
recreational waters receiving nonhuman fecal inputs in the context of the current regulatory
framework.
Recreational water epidemiological studies remain costly and can lack the statistical power to assess
the potential health risks from recreational exposure to waters affected by nonhuman fecal sources.
Epidemiological approaches traditionally require a large number of participants and a detectable
difference in the rate of swimmer-associated illness compared to nonswimmers to have enough
statistical power to assess associations. Also, if the actual risk of illness due to swimming is low, the
epidemiological analysis might not establish a statistical association between illness and FIB.5 Ethical
constraints are also associated with exposing recreators to waters containing known levels of
pathogens. Given the reported risk differential between human and some nonhuman fecal sources, the
highest-level exposures would likely have to occur on study days with elevated pathogen densities to
be detected via current epidemiological study designs. Elevated culturable FIB can still be predictive of
fecal contamination or potential human health risks in nonhuman contamination scenarios, but the
relationship between illness and indicator can be different, which affects the level of indicator
corresponding to the level of public health concern.
MRA is an alternative approach that uses modeling approaches to characterize potential human health
risks from recreational exposure to water containing fecal contamination. QMRA is a formal
investigative process, analogous to the chemical risk assessment process that the EPA and others use
to characterize the nature and magnitude of potential health risks to humans from contaminants that
can be present in the environment and to inform decision-making. QMRA is a framework and approach
that brings the available information together with mathematical models to address the spread of
microbial agents through environmental exposures and to characterize the nature of the adverse
outcomes (CAMRA, 2021).6 As the field of MRA developed, some complexities associated with
modeling infectious diseases that are unique to pathogens became clear. Thus, some features of
5 FIB provide an estimation of the amount of feces and, indirectly, the presence and quantity of fecal pathogens in the
water (NRC, 2004).
6 An online resource for QMRA includes a QMRAwiki developed by CAMRA at Michigan State University:
http://qmrawiki.org (last accessed June 23, 2023).
11
-------
QMRA necessitate the use of techniques and data in ways that differ from the assessment of chemicals
and other risks and include:
• Variation in the ability of individual organisms in a population of pathogens to initiate infection.
• Wide variation in the susceptibility (immunity) of human and animal hosts to infection.
• Wide variation in expression of disease symptoms.
• Risks of secondary (person-to-person) transmission of pathogens.
• Growth of pathogens in vivo and, for a subset of pathogens, growth in the environment.
• High variability (spatial and temporal) in the occurrence of pathogens in the environment.
• Difficulty in recovery and enumeration of pathogens.
QMRA can be used to address specific questions that risk managers might have, such as examining
specific exposure scenarios that might be of concern. QMRA can be used for a variety of reasons,
including to:
• Assess the potential for human risk associated with exposure to human or zoonotic pathogens
or both (Soller et al., 2010a,b, 2016, 2017; Schoen et al., 2011; Vergara et al., 2016; Ahmed et
al., 2018).
• Evaluate the potential efficacy of mitigation approaches, such as watershed protection
measures (Soller et al., 2006).
• Evaluate specific treatment processes to reduce, remove, or inactivate various pathogens
(Soller et al., 2003; Schoen and Garland, 2017; Schoen et al., 2017; Nappier et al., 2018).
• Predict the consequences of various management options for reducing risk (Roser et al., 2006;
Ashbolt et al., 2010; Viau et al., 2011; Brown et al., 2017a, Soller et al., 2018).
• Identify and prioritize research needs (Ashbolt et al., 2010; WERF, 2011).
• Assist in the interpretation of epidemiological investigations (Soller et al., 2016, 2017).
• Derive WQC linked to a specified level of protection and reflective of specific exposure
scenarios and sources of fecal contamination (Roser et al., 2006; Boehm et al., 2009; Viau et al.,
2011; Boehm etal., 2015).
Researchers have demonstrated QMRA to be a scientifically valid and effective tool for evaluating and
managing the public health impacts of nonhuman fecal sources in recreational waters and for
examining the relative pathogenicity of organisms found in animal fecal material (McBride et al., 2002,
2013; Till and McBride, 2004; Roser et al., 2006; Schoen and Ashbolt, 2010; Soller et al., 2010b, 2014;
U.S. EPA, 2010a; WERF, 2011). Findings from QMRA relative risk evaluations between waters affected
by human and nonhuman fecal contamination have been used to refine the understanding of the lack
of health correlations in epidemiological studies in waters predominately impacted by nonhuman fecal
contamination. Further benefits of QMRA include the lack of requirement for human subjects and the
ability to address risk management for sensitive subpopulations, such as the potential swimming-
12
-------
associated risk differential between children and adults. Questions related to these areas can be
important to risk managers.
In this TSM, QMRA is the general technical approach used to estimate the risk of illness due to
recreational exposures to surface waters containing enteric pathogens due to fecal contamination.
Additionally, the associated framework discussed can be used as a tool to develop alternative criteria
when the sources of FIB are predominantly nonhuman or nonfecal. This framework can be adapted to
account for specific microbes, dosage levels, and illnesses of concern, among other parameters. When
using this TSM, it is important to completely document collected data, follow accepted practices, and
rely on scientifically defensible approaches to be transparent in decision-making.
1.3.1 Summary of QMRA Literature Review
Since the 1990s, the EPA has been collecting and reviewing literature related to QMRA. This effort has
included reviewing published information on exposure assessment, human health effects assessment
(including dose-response modeling), and risk characterization for waterborne pathogens. The agency
also has conducted literature searches to identify studies describing QMRAs or commonly used QMRA
techniques relevant to characterizing risks from recreational water exposure and studies describing
novel or cutting-edge QMRA-related techniques (Appendix A; U.S. EPA, 2010a). The textbook by Haas
et al. (1999, 2014) also provides an extensive overview of the field of QMRA. The EPA has synthesized
the information available to evaluate potential risks from animal-derived pathogens (U.S. EPA, 2010a).
The EPA examined the most relevant and highest quality QMRA studies in detail to evaluate the
following objectives:
• Develop a list of QMRA studies from which to draw elements of future study designs.
• Identify the pathogens addressed in QMRAs and the potential reasons the study authors
selected those pathogens.
• Identify how variability has been addressed in QMRA studies, particularly source variability and
consumption/ingestion variability.
• Assess the tendency for risk analysts to include secondary transmission in the estimate of
overall risk.
• Compare the practices used by different QMRA researchers and practitioners, particularly
sensitivity analyses and risk characterization.
Studies relevant to exposure pathways other than water were not included. For example, high-quality
studies in the literature on food were not evaluated in detail, such as studies primarily concerned with
post-slaughter processes or the preparation of food products. For example, studies of Listeria
monocytogenes growth in delicatessen meat storage are not directly relevant to recreational
waterborne exposure and were excluded. However, several studies providing novel techniques for
incorporating the growth of Listeria monocytogenes into exposure assessment are included. Drinking
water studies focused on the treatment process in determining risk were excluded as not relevant;
however, several studies that assessed the role of source water quality in finished drinking water risk
were included.
13
-------
Based on the studies reviewed, some general observations about the use of QMRA were noted. First,
the utility of QMRA has been demonstrated in a wide variety of scenarios, including ambient waters
receiving human and nonhuman fecal contamination and the nonfecal contribution of FIB (Till and
McBride, 2008; Schoen and Ashbolt, 2010; Soller et al., 2010b, 2015; de Man et al., 2014). QMRA has
been used to characterize risk in scenarios where risk estimation by other techniques, such as
epidemiological methods, has been difficult to conduct, not yet conducted or would be cost-prohibitive
to conduct. Studies by McBride et al. (2004, 2013), Roser et al. (2006), Schoen and Ashbolt (2010),
Soller et al. (2010a,b, 2014, 2015), Sunger et al. (2019), Till and McBride (2008), Timm et al. (2016),
Vegara et al. (2006), Viau et al. (2011b), and Wong et al. (2009) are examples of QMRA used to address
risk management questions and support decisions related to understanding important aspects of
waterborne illness. QMRAs typically can evaluate variability when assessing potential health hazards.
For example, many of the QMRAs reviewed for this report accounted for the variability in pathogen or
indicator density by treating them as stochastic variables.
Several observations can be drawn from the review and comparison of the identified QMRA studies.
First, the identified studies focused on a small subset of the pathogens potentially important in
waterborne exposure during recreation. The two pathogens analyzed most frequently, rotavirus and
Cryptosporidium, may be important contributors to the risk of gastrointestinal illness, primarily due to
their low infectious dose (ID)io (or another measure of low-dose infection), frequent occurrence in
sewage, and, particularly for Cryptosporidium, relatively high persistence in environmental matrices.
One potential reason for frequent selection of rotavirus and Cryptosporidium is that the available peer-
reviewed dose-response models are based on oral ingestion. Numerous studies (Petterson and
Ashbolt, 2001; Ottoson and Stenstrom, 2003; Soller et al., 2003, 2006; Eisenberg et al., 2004, 2008;
Hamilton et al., 2006; Bastos et al., 2008; Katukiza et al., 2014; Ahmed et al., 2020) used rotavirus as a
surrogate for enteric viruses. When considering the potential illness risk posed by enteric viruses to the
recreating general population, this approach can be considered health-protective, given that rotavirus
is routinely found in sewage and has a high probability of a single organism initiating infection given
known dose-response relationships. The norovirus dose-response model has also been used more
frequently in recent years (Soller et al., 2010a,b, 2014; Eregno et al., 2016; Murphy et al., 2016;
Chaudhry et al., 2017; Ahmed et al., 2018; Simhon et al., 2020; Jahne et al., 2023) and is notable
because it is based on doses in units of genome copies (polymerase chain reaction [PCR] based units)
(Teunis et al., 2008a, 2020). Soller et al. (2010b) evaluated the Teunis et al. (2008a) norovirus dose
response to understand the potential etiologic agents causing the reported illness in epidemiological
studies conducted on the Great Lakes; they found it represented the empirical results well. Although
the focus of this TSM is for recreational waters that are predominantly affected by nonhuman fecal
sources, some loading of human fecal contamination can occur and be accounted for in the QMRA
framework by including human-related reference pathogens, such as enteric viruses. However, as
discussed in Section 3, the potential risk of gastrointestinal illness can be quickly driven by the human
source(s) as the proportion of human inputs increases (e.g., greater than approximately 30% of fecal
loading attributable to humans). Waterbodies receiving treated and disinfected effluent would likely
not qualify for alternative criteria using the process described in this TSM due to the particular risk
profile associated with this source (Soller et al., 2010b, 2014; Schoen et al., 2011).
The review of the scientific literature indicates that the exponential and approximate beta-Poisson
models are two commonly used dose-response models. The exact beta-Poisson relationship (i.e., the
hypergeometric function [Haas et al., 1999]) was seldom used because it is complex and
computationally difficult, nor were empirical models popular in food dose-response studies (e.g., as
14
-------
described in Buchanan et al. [2000] and Moon et al. [2004]). A disadvantage of the exponential and
approximate beta-Poisson models is that they often do not account for variability and uncertainty in
dose-response model parameters. There is a need for dose-response models corresponding to
different exposure routes (e.g., ingestion, inhalation) because some waterborne pathogens (e.g.,
adenovirus) can initiate infection via multiple routes. Variability in dose-response model parameters or
the response of the exposed population is rarely considered or addressed. A likely cause for the latter
is that dose-response model studies do not consistently provide confidence intervals (CI) for model
parameters and seldom present quantitative information on the shape of the distribution for
parameter estimates.
The review also indicates pathogen density can be relatively straightforward to model. Modeling
variability in pathogen density appears to be limited by the scarcity of available data. Among the
studies reviewed for this report, many studies employed point estimates for pathogen density. The two
most common methods for accounting for pathogen variability among the studies are (1) the use of
empirical distributions for pathogen density based on relatively short time series and (2) the
assumption of lognormal distribution of pathogen densities. Among studies using distributions to
describe pathogen variability, normal, triangular, lognormal, negative binomial, uniform, and Poisson
distributions were the most common.
Another finding from the review of the scientific literature is that secondary transmission and
(temporary) immunity are often not included in risk estimation. Several studies (Eisenberg et al., 1996,
1998; Soller et al., 1999, 2003, 2006; Koopman et al., 2002) have demonstrated that consideration of
secondary transmission and immunity can, in some cases, significantly influence overall risk associated
with exposure to pathogens. The scope of this TSM focuses on primary contact recreation; it does not
include secondary transmission in the health modeling because the potential risks are characterized,
and the guideline values for recreational waters are recommended for recreators engaging in primary
contact recreation. Risk managers are tasked with determining whether a waterbody is safe for
recreation.
1.3.2 Overview of the TSM Framework
This TSM includes a multi-step decision process (see Figure 1-1) that can be used to determine whether
the predominant source(s) of fecal contamination affecting a waterbody are nonhuman and, if so, how
to apply QMRA in developing site-specific alternative criteria for recreational waters. The steps in this
TSM also include risk management and policy decisions to help plan and refine water quality studies to
support decision-making.
The process starts with acquiring information on the sources and loadings of fecal contamination to a
waterbody of interest to determine whether it is predominantly affected by nonhuman fecal sources
or if the human contribution is sufficient for risks to be comparable to the human contamination
scenario characterized in epidemiological studies informing the EPA's 2012 RWQC. Four major steps
comprise the overall process. At each step, evaluating the information collected and analyzed will help
address specific decision points for the risk manager to consider when proceeding to the next step.
15
-------
k A
Identify fecal sources
• How: Sanitary Survey (see Appendix B).
• Decision Point: Is there clear evidence that nonhuman fecal sources
impact and possibly dominate the waterbody?
A
• Collect water quality information
• How: Water Quality Study (see examples in Appendix C).
• Decision Point: Are nonhuman fecal contamination sources
confirmed to be affecting the waterbody?
• Evaluate human health risks
• How: "forward" QMRA.
• Decision Points:
• How do the water quality and illness estimates compare to the
water quality and illness targets of the applicable WQS?
• Does this information support deriving alternative water quality
criteria for the waterbody?
A
• Derive alternative water quality criteria
• How: Calculate GM, statistical threshold value (STV) and beach
action value (BAV) corresponding to health-based goal using
"reverse" QMRA.
• Decision Points:
• Are the decisions and risk assessment assumptions transparently
documented?
• Determine the magnitude, duration, and frequency elements to be
included in the criteria.
• Are the proposed criteria scientifically defensible and protective of
the designated use? >
Figure 1-1. Flow diagram for considering QMRA in developing site-specific alternative criteria.
16
-------
Step 1 consists of identifying the fecal sources in the watershed that can affect the waterbody to
determine if QMRA is appropriate. Typically, this is accomplished by using a standardized sanitary
survey to determine the sources of fecal contamination. If the sanitary survey findings support the
original premise that human sewage does not dominate the fecal contamination entering the
waterbody, proceed to the next step.
Step 2 involves collecting water quality data on the waterbody, including information about fecal
indicators, pathogens, and fecal source identification and tracking. This information will help
characterize the extent of the fecal loading in the waterbody. Steps 1 and 2 together comprise the
"Sanitary Characterization," which can provide substantive lines of evidence to support moving
forward with conducting a QMRA and potentially developing site-specific alternative criteria. If the
sanitary characterization reveals that the waterbody's susceptibility to human contamination is low,
such as the absence of significant human fecal loadings, then a decision to proceed to the next step is
supported. Lines of evidence supporting a categorical determination of "low" susceptibility to human
contamination are also provided in Step 2.
Step 3 consists of evaluating human health risks from recreational exposure to the waterbody. These
risks are characterized using a forward QMRA process and the water quality information collected from
the watershed. A forward QMRA uses pathogen occurrence as an input and calculates the probability
of illness. The results of the QMRA analysis in this step will quantify the potential human health risk
from recreational exposure in the waterbody. Conducting a relative QMRA analysis may be useful for
interpreting the results of the forward QMRA and should include the same target illness rate
associated with the applicable WQS or the EPA's RWQC.
Step 4 describes how to derive alternative WQC and calculate adjusted GM, statistical threshold value
(STV) and BAV values for culture-enumerated enterococci or E. coli corresponding to the target illness
rate.
The EPA also provides a WQS submission checklist to help the user assemble the information that the
EPA will use to evaluate a WQS submission (Appendix D).
17
-------
2.0 Scope and Problem Formulation: Characterizing Human Health Risks
from Nonhuman Sources of Fecal Contamination in Recreational Waters
Using QMRA
While the generalized risk assessment conceptual framework used in chemical risk assessment can be
applied to MRA, differences between chemical contamination and microbial contamination require
specific approaches to assessing risk from human infection and resulting illness following exposure to
pathogens. To address this need, the EPA's Office of Water developed a conceptual framework in
conjunction with the International Life Sciences Institute (ILSI) to assess the risks of human infection
associated with pathogenic microorganisms (ILSI, 2000). The purpose of the ILSI framework was to
describe a generic approach to identifying scientific information that should be considered when
quantitatively assessing human health risks associated with exposure to infectious agents in water
(ILSI, 2000). The framework supports the conduct of MRA for various types of microorganisms under
various exposure scenarios. The U.S. Department of Agriculture's Food Safety and Inspection Service
and the EPA developed the Microbial Risk Assessment Guideline: Pathogenic Microorganisms with
Focus on Food and Water ("MRA Guideline") as a resource for developing MRAs (U.S. EPA and USDA,
2012). The MRA Guideline document discusses a flexible framework including general fundamental risk
assessment principles that specifically address microbial risks. In 2014, the EPA published Microbial
Risk Assessment Tools, Methods, and Approaches for Water Media ("MRA tools") to facilitate the
conduct of MRAs for water-related media (U.S. EPA, 2014b, Ichida et al., 2016). The MRA tools
document describes a flexible human health risk assessment (HHRA) framework that can be used to
inform public health decisions related to microbial hazards in various water-based media. Users can
adapt, modify, or tailor the elements within the framework to address specific risk management
questions (U.S. EPA, 2014b). These framework documents support risk assessors in developing
consistent, high-quality evaluations of microbial risks and informing decisions focusing on the risks
posed by infectious microorganisms. The approaches described in these documents help risk assessors
characterize exposure sources, causative agents (including naturally occurring pathogens and
pathogens introduced via contamination), contributing immunity factors, associated symptoms, acute
disease and chronic sequelae endpoints, and potential control points to reduce risk. The EPA's HHRA
framework, Framework for Human Health Risk Assessment to Inform Decision Making ("HHRA
framework") (U.S. EPA, 2014c), was developed to support risk-based decision-making to maximize the
utility of risk assessment. The HHRA framework discusses the importance of the design of risk
assessments in the decision-making process and integrates planning, scoping, and problem formulation
with traditional risk assessment approaches to support a "fit for purpose" assessment. The approach
discussed in this TSM incorporates aspects of the MRA and HHRA frameworks (Figure 2-1). The QMRA
analyses described herein are designed to quantitatively assess risk from enteric pathogens associated
with fecal contamination in surface waters. The MRA steps are integrated into a decision-based
process for risk managers to utilize when addressing questions about the risks posed by nonhuman
fecal contamination of surface waters.
18
-------
Public,
Stakeholder
and
Community
Involvement
Planning and Scoping
and Problem Formulation
Conceptual
Analysis
Model
Plan
Risk Assessment
Exposure
Assessment
Effects Assessment
Hazard Identification
Dose Response
Risk Characterization
ai
in
O
Q.
w
3
CL
Informing Decisions
Figure 2-1. Framework for HHRAto inform decision-making (U.S. EPA, 2014c).
The planning, scoping, and problem formulation stage is used to identify: (1) the purpose of the risk
assessment, (2) the critical issues to be addressed, and (3) how the results might be used to protect
public health. As part of the initial planning process, key stakeholders interested in recreational water
quality and safeguarding the public from pathogen pollutants in waters where people recreate can be
identified and invited to participate. Stakeholders may include public health officials, the business
community, environmental groups, and landowners. Stakeholder input can help define the scope of
the effort and ensure that various science, policy, and management perspectives are incorporated into
steps of the alternative criteria evaluation process. Throughout the process discussed in this TSM,
periodic stakeholder engagement can be used to define the "fit for purpose"—or suitability and
usefulness of the activities—for informing risk management decisions. Stakeholder input can also help
inform the various decision points identified in this TSM, including when developing and conducting
both the sanitary survey and the water quality study supporting the conduct of a QMRA.
Problem formulation is a process for generating and evaluating preliminary hypotheses about why
human health effects have occurred or might occur as the result of exposure to sources of human and
nonhuman fecal contamination. The scoping and problem formulation steps include the development
of the conceptual model and analysis plan and are part of the framework described in this document
(Section 2.2). The framework presented in this document also incorporates risk assessment purposes,
including to:
1. Establish and quantify the potential human health risk from recreational exposure to waters
receiving predominantly nonhuman fecal contamination,
2. Compare the potential risk differential to exposure to human fecal contamination; and,
19
-------
3. Translate the potential risk differential into adjusted values for culturable E. coli or enterococci
that are protective to the same level as EPA's 2012 RWQC recommendations.
Users of this document may include additional or updated elements, such as alternative health
endpoints, alternative or additional pathogens, or new modeling information. All parameter and model
choices discussed in this document are documented in a transparent fashion. The QMRA sanitary
survey form (see Appendix B) or portable device application aids users in delineating fecal sources. An
example Sampling and Analysis Plan (SAP) describing the process for generating water quality data to
inform the QMRA is included (see Appendix C).
During the risk assessment process, information about the exposure to and health effects of fecal-
associated pathogens is compiled and summarized. The quantitative and qualitative evaluation of
these data, expert opinion, and other information are integrated into the risk characterization,
including the associated assumptions and uncertainties. Using QMRA to describe human health risks
and inform alternative WQC is a data-driven exercise—from conducting a sanitary survey and
characterizing water quality to estimating potential risks and translating risks into alternative indicator
values. The QMRA framework allows consideration of specific risk management questions involving
sensitive subpopulations, such as the potential swimming-associated risk differential between differing
age groups, such as younger children, older children, and adults. Importantly, decisions made during
the process discussed in thisTSM should be supported by data that characterize the potential sources
of fecal-associated pathogens in a waterbody that have the potential to adversely affect human health.
Documented data analysis and interpretation support the scientific defensibility and effectiveness of
alternative WQC. Iterative engagement with stakeholders to discuss the QMRA parameter selections
and results is recommended.
The QMRA framework described in this TSM incorporates the state of the science for QMRA and
aspects of previous analyses conducted by the EPA and others (Ashbolt et al., 2010; Schoen and
Ashbolt, 2010; Soller et al., 2010a, b, 2014, 2015, 2016; U.S. EPA, 2010a; WERF 2011; McBride et al.,
2013; Sunger et al., 2019). Consistent with previously published QMRAs, the QMRA scenario discussed
in this TSM considers recreational events as independent events and are treated probabilistically. The
independence of recreational events means that the probability of an adverse health effect on one day
is not influenced by recreational events that occurred on prior days. To evaluate recreational events
probabilistically, statistical distributions, rather than point estimates, of the QMRA input parameters
were used. The probabilistic approach produces different model results each time the model is run due
to the effects of chance. Each time the QMRA model is run, a random value within each input
parameter distribution is used; this process runs for thousands of iterations to obtain the result, which
is also a distribution.
Conducting a QMRA in recreational water requires four data inputs:
1. Volume of water ingested during recreational activities,
2. Density of reference pathogens in the ingested water,
3. A quantitative relationship between the number of pathogens ingested (dose) and the
probability of infection (dose-response relationship), and
4. The conditional probability of illness given infection (probability of symptomatic response).
20
-------
These parameters and how they are used in QMRA to estimate the probability of illness in a waterbody
using locally collected data are described in this document. Outputs from the data analysis and
quantitative health models can support understanding the risk of gastrointestinal illness in the
waterbody studied (Section 3.3) and provide the basis for adjusting the EPA's 2012 RWQC
recommendations due to the nature of the source(s) of fecal contamination (Section 3.4). Additional
information on susceptible subpopulations and sensitivity analyses can be found in Section 4. The
discussion in Sections 3 and 4 can help stakeholders integrate and interpret the QMRA-based outputs,
characterize uncertainty, and support decision-making.
2.1 Rationale for use of Stressors and Surrogates
Human and animal fecal waste can contain a wide array of pathogenic microbes from three main
pathogen categories: viruses, bacteria, and protozoa. The level of each pathogen type can vary in the
host and in the receiving waters due to dilution, decay, and other factors (Teixeira et al., 2020). Some
of these pathogens are host-specific (e.g., human enteric viruses), while other pathogens can infect
and cause illness in both animals and humans (i.e., zoonotic pathogens) (U.S. EPA, 2009b). Fecal
contamination of surface waters can result in the transmission of these pathogens, causing waterborne
disease (Savichtcheva and Okabe, 2006). Sources of fecal contamination can include human and
nonhuman contributions and be categorized as "point source" (e.g., raw sewage, treated effluent from
WWTPs, etc., characteristically discharged from discrete and/or permitted sources) or "nonpoint
source" (e.g., runoff from agriculture, wildlife, and urban land usages) inputs that can adversely affect
water quality. This latter category may enter waterbodies as overland flow, via drainage ditches and
creeks, or through less-regulated stormwater conveyances. Animal fecal inputs to surface waters can
range from wildlife and vermin sources to animal husbandry operations (e.g., concentrated animal
feeding operations [CAFOs]) and agronomic applications of manures.
Plowright et al. (2017) describe "zoonotic spillover" as the transmission of a pathogen from animal to
human. Spillover transmission depends on multiple dynamic processes and host interactions that can
result in zoonotic infection and illness in humans. The probability of the spillover occurring is
influenced by three functional phases: pathogen pressure, pathogen exposure, and human host
characteristics (Plowright et al., 2017). First, the presence of pathogens in feces depends on their
occurrence in infected individuals in a population and the rate of shedding. Once shed by the animal,
pathogens can be disseminated in the environment and can be attenuated by numerous
environmental factors. Some excreted pathogens further develop in water as part of their natural
lifecycle or find a new host (U.S. EPA, 2009b). Second, human behavior can influence pathogen
exposure, including the route and dose of exposure. For the third phase, Plowright et al. (2017)
describe the human host's genetic, physiological, and immunological attributes that, given a sufficient
dose, affect the probability and severity of infection and resulting illness. Other studies suggest the
potential for attenuation of the infectivity, virulence, and disease severity to humans from animal-
derived ("passaged") zoonotic pathogens (U.S. EPA, 2009b).
The diversity of pathogens that can be present, typically at low densities in environmental waters, and
the challenges associated with frequent, widespread monitoring for those pathogens has resulted in
the use of FIB as surrogates for fecal contamination, such as culture-enumerated E. coli and
enterococci, to routinely evaluate water quality (Schwab, 2007; Korajkic et al., 2018). Pathogen
occurrence in surface waters can be variable and their enumeration methods can be complex, labor-
intensive, technically challenging, and in some cases not feasible (Savichtcheva and Okabe, 2006;
21
-------
Korajkic et al., 2018). Not detecting a particular pathogen does not indicate the absence of other
pathogens (Ahmed et al., 2018). Because pathogens in the surface water matrix can be associated with
high temporal and spatial variability, further complicated by enumeration method sensitivity and
variability, the lack of detection may not indicate the absence of the microorganism, just that the level
of the target could be below the limit of detection (LOD)/limit of quantification (LOQ). Frequent
nondetects or results below the LOQ can be challenging to interpret quantitatively.
In lieu of monitoring for all pathogens that may be present, FIB, such as E. coli and enterococci, are
routinely enumerated with simple, straightforward culture-based laboratory methodologies. Although
FIB can have environmental, nonfecal sources, elevated quantifications above background, especially
during baseflow conditions, can be considered indicative of fecal contamination and, indirectly, of the
potential presence of pathogens. Traditional culturable FIB are not well correlated with the presence
or absence of specific pathogens in environmental waters; however, published epidemiological
examples provide statistically significant associations between increasing FIB levels and increases in
illness, mainly in waters receiving human fecal contamination (Cabelli, 1983; Dufour, 1984; Kay et al.,
1994; McBride et al., 1998). In waters affected by animal fecal sources, the epidemiological evidence is
inconsistent, but participants in those studies reported increased illness given recreational exposures
(Calderon et al., 1991; McBride et al., 1998; Sinigalliano et al., 2010). Calderon et al. (1991) attributed
the reported illnesses to swimmer-to-swimmer transmission due to illness being associated with high
numbers of swimmers per day and elevated levels of staphylococci. McBride et al. (1998) observed no
differences in reported HCGI at marine beaches receiving animal waste inputs or human effluents from
oxidation ponds. Sinigalliano et al. (2010) found an increase in self-reported gastrointestinal,
respiratory, and skin illnesses among bathers compared to nonbathers in nonpoint source subtropical
recreational marine waters.
The EPA published a review of potential human illnesses caused by recreational exposure to water
contaminated by animal feces (U.S. EPA, 2009b). The available recreational water monitoring data,
public health reporting, epidemiological studies, illness outbreak reports, and dose-response studies
suggest that a subset of pathogens is representative of the majority of hazards due to recreational
exposure to human and animal feces-impacted recreational water (Scallan et al., 2011a,b). The review
focused on warm-blooded animals (mammals and birds) whose fecal material is detected by currently
recommended FIB. Seventy pathogens from warm-blooded animals were evaluated for the following
attributes:
1. The pathogen spends part of its lifecycle within one or more warm-blooded animal species.
2. Within the pathogen's lifecycle, it is probable or conceivable that some lifestage will enter water.
3. Transmission of the pathogen from the animal source to human is through a water-related route.
4. The pathogen causes infection or illness in humans.
Twenty pathogens have been shown to possess all four of these attributes. Based on their relevance in
the United States, six of these 20 waterborne, zoonotic pathogens from warm-blooded animals were
selected for further evaluation. Five of the six were linked to illness outbreaks in ambient, untreated
recreational water, and the sixth was included based on outbreak reports in drinking water (U.S. EPA,
2009b). The six key waterborne zoonotic pathogens identified in the 2009 review included pathogenic
E. coli, Campylobacter, Salmonella, Leptospira, Cryptosporidium, and Giardia.
22
-------
Pathogens not covered in this TSM are nonzoonotic pathogens, pathogens generally found in the
environment, such as free-living protozoa, and pathogens with cold-blooded hosts, such as snails or
copepods (U.S. EPA, 2009b). Additionally, some common zoonotic pathogens were excluded because
they do not have a well-documented waterborne transmission but may be transmitted by other
routes, such as food or soil.
Reviews of data on waterborne transmission of zoonotic pathogens have identified pathogens of
primary concern based on their occurrence in water, abundance in animal feces, and persistence and
ability to multiply in the environment (Rosen, 2000; Bicudo and Goyal, 2003; Goss and Richards, 2008;
U.S. EPA, 2009a,b, 2013a,b). Based on these criteria, the protozoa Cryptosporidium and Giardia, the
bacterial pathogens Salmonella, Campylobacter, and pathogenic E. coli (represented by E. coli
0157:1-17) are the primary pathogens of concern in livestock waste (Rosen, 2000; Bicudo and Goyal,
2003; Goss and Richards, 2008). Recreational water illnesses and their etiological agents of secondary
concern include yersiniosis (caused by Yersinia enterocolitica), cercarial dermatitis (caused by
schistosomes), shigellosis (caused by Shigella spp.), and leptospirosis (caused by Leptospira spp.)
(Rosen, 2000; Levett, 2001; Bicudo and Goyal, 2003; Moyer and Degnan, 2006). There is little evidence
of transmission of fecal-associated viruses of animal origin to humans, but some interspecies
transmission of rotaviruses has been demonstrated experimentally (Rosen, 2000). Most viruses are
host-specific, but a few are more general and are capable of infecting one or more species of host.
Human and animal fecal wastes contain a variety of different kinds of viruses, and some animal viruses
(e.g., hepatitis E) have the potential to infect humans (U.S. EPA, 2009b). However, there are
considerable uncertainties about the waterborne transmission of zoonotic viruses to humans.
Outbreaks of waterborne gastrointestinal illness are commonly attributed to human enteric viruses,
such as norovirus (Hlavsa et al., 2015).
QMRA models estimate potential risk from exposure to specific pathogens. As part of the QMRA-based
analyses discussed in this TSM, the EPA chose a short list of "reference" pathogens based on the key
waterborne zoonotic pathogens found in feces and common human enteric viral pathogens. The
specific pathogens, the justification for selecting the pathogens, and pathogen-specific parameters
used in QMRA are presented along with information on the FIB, E. coli and enterococci, found in
nonhuman fecal sources.
2.1.1 Reference and Index Pathogens
The U.S. Centers for Disease Control and Prevention (CDC) estimates that known pathogens account for
an estimated 33 million illnesses each year in the United States, including 4.2 million due to bacteria,
2 million due to parasites, and 27 million due to viruses (Table 2-1) (Scallan et al., 2011a,b). Of those
illnesses, 9.3 million were thought to be foodborne, leaving 24 million illnesses, some of which were due
to waterborne exposures, including recreational water contact. Of the 24 million nonfoodborne illnesses
per year, 21.5 million illnesses are associated with a viral etiology, with norovirus estimated to account
for approximately 15.3 million of the nonfoodborne illnesses. There is an overlap between the few
pathogens attributed to most nonfoodborne illnesses in the United States and the list of key feces-
associated waterborne zoonotic pathogens the EPA identified (U.S. EPA, 2009b). Research has
demonstrated that bacteria (Campylobacter spp., E. coli 0157:H7, and Salmonella enterica) and protozoa
(Cryptosporidium and Giardia spp.) can be zoonotic pathogens (Till et al., 2008; Ahmed et al., 2009; U.S.
EPA, 2009b; Gabriel et al., 2010; McBride et al., 2013; Schmidt et al., 2013; Hlavsa et al., 2015; Goh et al.,
2019; Korajkic et al., 2019). These five zoonotic pathogens, which appear on both the CDC's and
23
-------
Table 2-1. Estimated annual illnesses in the United States from known pathogens.a,b
Pathogen
class
Pathogen
Total
estimated
annual cases
Percent
foodborne
illness
Number of
foodborne
illnesses
Number of
nonfoodborne
illnesses0
Bacterial
Bacillus cereus
63,411
99.98
63,400
11
Botulism, foodborne
55
100
55
0
Brucella spp.
1,679
50
839
840
Campylobacter spp.
1,058,387
80
845,024
213,363
Clostridium perfringens
966,120
99.98
965,958
162
E. coli 0157 :H7
93,094
68
63,153
29,941
E. coli, non-0157 STEC
138,063
82
112,752
25,311
E. coli, enterotoxigenic
17,897
99.98
17,894
3
E. coli, other diarrheagenic
39,739
30
11,982
27,757
Listeria monocytogenes
1,607
99
1,591
16
Salmonella enterica serotype Typhi
1,897
96
1,821
76
Salmonella, nontyphoidal
1,095,079
94
1,027,561
67,518
Shigella spp.
421,048
31
131,254
289,794
Staphylococcus aureus foodborne
241,188
100
241,188
0
Streptococcus, foodborne
11,219
99.98
11,217
2
Vibrio cholerae, toxigenic
84
100
84
0
V. vulnificus
203
47
96
107
Vibrio, other
30,727
57
17,564
13,163
Yersinia enterocolitica
108,490
90
97,656
10,834
Subtotal
4,289,987
3,611,089
678,898
Parasitic
Cryptosporidium spp.
678,828
8
57,616
621,212
Cyclospora cayetanensis
11,522
99
11,407
115
Giardia intestinalis
1,121,864
7
76,840
1,045,024
Toxoplasma gondii
173,415
50
86,686
86,729
Trichinella spiralis
156
100
156
0
Subtotal
1,985,485
232,705
1,753080
Viral
Norovirus
20,796,079
26
5,461,731
15,334,348
Rotavirus
3,089,868
< 1
15,433
3,074,435
Astrovirus
3,089,868
< 1
15,433
3,074,435
Hepatitis A
21,041
7
1,566
19,475
Subtotal
26,996,856
5,494,163
21,502,693
Total
33,272,328
9,337,957
23,934,493
Notes:
a. Source: Scallan et al. (2011a,b)
b. Shaded pathogens represent the standard list of reference pathogens the EPA discusses in this TSM.
c. These can include recreational waterborne illnesses.
24
-------
the EPA's lists, plus the human enteric viruses norovirus and rotavirus, are highlighted in gray in
Table 2-1. To this list, the EPA added human adenovirus, a nonenveloped virus resistant to common
wastewater treatment (because it has been identified as an etiologic agent in untreated recreational
water outbreaks), documented to occur in sewage year-round and measured in waters affected by
human fecal contamination (Lee et al., 2014; Hlavsa et al., 2015; Rames et al., 2016; Goh et al., 2019;
Graciaa et al., 2018). Together, these eight pathogens can be considered reference pathogens for
evaluating waterborne illnesses in QMRA. The use of reference pathogens to represent the infectivity
and the likely environmental fate and transport of each microbial group is a widely accepted practice in
the field of QMRA (WHO, 2004; Roser et al., 2007; Soller et al., 2010b; Haas et al., 1999, 2014).
The QMRA reference pathogens included in the QMRA modeling in this TSM (Text Box 2-1) were
selected for recreational waters because they:
1. Represent the three classes of pathogens of
concern (bacteria, viruses, and protozoa),
2. Are found in a wide variety of sources,
3. Are transmitted via water exposure,
4. Are known to cause gastrointestinal illness,
5. Have associated peer-reviewed dose-response
relationships in the literature, and
6. Account for more than 97% of nonfoodborne
illnesses in the United States (Mead et al., 1999;
Scallan et al., 2011a,b).
The reference pathogens include the following:
• Bacteria: Campylobacter spp., Salmonella enterica,
and E. coli 0157:H7 (as an index for the broader
group of pathogenic E. coli and other gram-negative, facultative anaerobic, nonspore-forming
bacteria that can cause diarrheal diseases such as Shigella spp.).
• Protozoa: Cryptosporidium and Giardia spp.
• Viruses: Norovirus (as an index for the broader group of enteric viruses7), rotavirus, adenovirus
(as an index for viruses that are resistant to wastewater disinfection and have low attenuation
rates in the environment).
Text Box 2-1. Reference and Index
Pathogens
Reference and index pathogens represent
the infectivity and the fate and transport
characteristics of the pathogens most
likely to occur in waterbodies affected by
warm-blooded animal (including human)
fecal sources. The reference pathogens in
this list are associated with a significant
proportion of waterborne gastrointestinal
illnesses. The index pathogens listed
represent a broader group of pathogens
that, as a group, are important. This set of
reference and index pathogens is useful
because it covers the major waterborne
pathogens associated with nonfoodborne
illness occurrence in the United States.
7 In this TSM, norovirus is considered an index pathogen because published dose-response functions have been shown to
effectively model gastrointestinal illnesses reported in epidemiological studies (Soller et al., 2010a). Because norovirus
occurrence in the population can vary over the year (with peaks during winter in the northern hemisphere) and the
epidemiological data were collected during the summer, norovirus was unlikely to be the primary viral etiologic agent.
Other enteric viruses, such as adenoviruses and enteroviruses, causing similar symptoms and resulting in quick-onset,
usually self-resolving acute gastrointestinal illness (AGI) were likely present because the beaches characterized in the
epidemiological studies were affected by nearby discharges of secondary treated and disinfected effluent and the
symptomology of illness was consistent with a viral etiology. See Wade et al. (2006, 2008, 2010, 2022) for a discussion of
the health effects reported during EPA's NEEAR epidemiological studies.
25
-------
Of the eight reference pathogens described above, the CDC estimated the number of annual illnesses
for seven of these pathogens (gray rows in Table 2-1). Adenovirus, norovirus, and E. coli 0157:H7 are
also considered index pathogens because they represent a broader class of pathogens that share
similar infectivity or environmental fate and transport characteristics, can be waterborne, and have
been included in published QMRAs (WHO, 2004; Pond, 2005; Roser et al., 2007; Soller et al., 2010a,b;
Ferguson et al., 2008; Schoen et al., 2011; Hlavsa et al., 2015; Graciaa et al., 2018). For the purposes of
this TSM, the index pathogens are considered a subset of the reference pathogens. Hereafter, these
eight pathogens are referred to collectively as reference pathogens.
Adenoviruses are nonenveloped, double-stranded deoxyribonucleic acid (DNA) viruses that are very
stable after chemical or physical agent exposure or adverse potential of hydrogen (pH) conditions, are
stable and persistent with a wide distribution in a range of water matrices, found consistently in
sewage, have high resistance to ultraviolet disinfection, detected in chlorinated wastewater, and occur
in higher abundance compared to other enteric viruses (Pond, 2005; Rames et al., 2016). Adenovirus is
not a reportable disease agent in the United States, so the health burden from adenovirus exposures in
recreational waters is likely underestimated but is known as one of the leading causes of recreationally
associated waterborne disease worldwide (Sinclair et al., 2009). Adenovirus is the most prevalent type
of human enteric viruses in water and has been identified as the etiologic agent responsible for some
outbreaks in untreated recreational water (Pond, 2005; Jiang, 2006; Hlavsa et al., 2015; Graciaa et al.,
2018). There are available culture and gene-based assays to enumerate adenoviruses, which are easier
to detect than ribonucleic acid (RNA) viruses (Rames et al., 2016; Hess et al., 2021). Because
adenoviruses can replicate and achieve a high viral load in the gut before excretion in feces,
contaminated water has been a documented source of exposure for adenoviruses, either through
ingestion, inhalation, or direct contact with the eyes (Singh-Naz and Rodriguez, 1996; Glass, 2001; Kelly
and Birch, 2004; U.S. EPA, 2009b). Environmental monitoring studies have documented the occurrence
of adenovirus in surface waters (Choi and Jiang, 2005; U.S. EPA, 2009b; Marion et al., 2014). Marion et
al. (2014) reported significant positive associations between adenovirus exposure and diarrhea and
gastrointestinal illness in recreators.
Norovirus was selected as a reference pathogen because it is a leading cause of recreational
waterborne illness including outbreaks of gastrointestinal illness (Scallan et al., 2011a,b; Matthews et
al., 2012; Gibson, 2014, Eftim et al., 2017; Graciaa et al., 2018). Norovirus is also considered an index
pathogen because it can represent the broader class of enteric viruses present in human fecal
contamination (Soller et al., 2010a; McBride et al., 2013). In Soller et al. (2010a), the Teunis et al.
(2008a) norovirus dose response effectively indexed the vast majority of illnesses reported in the EPA's
NEEAR water study.
E. coli 0157:H7 is an index for pathogenic E. coli because it is representative of other pathogenic E. coli
and other gram-negative, facultative anaerobic, nonspore-forming bacteria (e.g., Shigella spp.); is
highly infectious; has peer-reviewed dose-response relationships; and can exhibit more serious health
outcomes, such as hemolytic urea syndrome (HUS). The EPA selected E. coli 0157:H7 to ensure that
these types of outcomes would be included, even though the reported number of cases per year is
relatively low.
While users of this TSM are not limited to these reference pathogens, they were selected for
recreational waters based on a thorough review of the scientific literature and their use in other QMRA
and water quality studies. Users of this TSM should provide supporting documentation if other
26
-------
reference pathogens are included in the QMRA or if the standard reference pathogens are not
included. If the collected source tracking information demonstrates little to no human inputs to a
waterbody, a decision not to include the viral reference pathogens (adenovirus, norovirus, rotavirus)
could be made because they are associated with human fecal sources. If human sources are present
but are not dominating the fecal loading, monitoring for one or more of these viruses could be used as
additional information on the extent of human fecal loading to the water being studied. For example,
adenoviruses are known to persist in environmental waters, are consistently found in wastewater
throughout the year and can be detected with both culture- and gene-based methods (Rames et al.,
2016). In mixed-source scenarios with low human inputs, recent advances made in methods for viral
enumeration could be considered (Haramoto et al., 2018). Given this information, the selection of the
reference pathogens should be based on the sources of contamination.
This section discusses the QMRA parameters related to the reference pathogens, including:
1. Occurrence of reference pathogens in feces and water.
a. Abundance of pathogens in fecal material for humans, cattle, gulls, pigs, and sheep
(Table 2-2).
b. Prevalence of pathogens in animal feces (proportion of animals that are shedding)
(Table 2-3).
c. Environmental fate and transport of pathogens.
2. Infectivity and illness associated with reference pathogens.
a. Human infection potential of pathogens in animal feces (Table 2-4).
b. Dose-response relationships and parameters for dose-response models for each
reference pathogen (Table 2-5).
c. Probability of illness given infection for each reference pathogen (Table 2-6).
The QMRA framework presented in this TSM follows the infectious disease process of exposure,
infection, and illness chronologically. For example, the first step is exposure to a pathogen through
some behavioral activity. Once exposure occurs, the pathogen interacts with the host, which can lead
to infection. However, not all exposures result in infection. Illness may or may not occur in an infected
host. The human immune system can mediate the extent and severity of the exposure, infection, and
illness process.
The process of environmental fate and transport of pathogens from excretion to entering surface
waters is also presented to consider to what extent fate and transport parameters are included within
the QMRA. A health-protective and parsimonious evaluation of illness risk can be associated with a
direct deposition scenario where exposure occurs to fresh animal feces contaminating a recreational
waterbody (Soller et al., 2010b). However, including some fate and transport information, if known,
may provide information about fecal loading dynamics to surface waters, hazardous conditions, and
evaluation of risk (Soller et al., 2015; Wu et al., 2020).
27
-------
If cattle are identified as a contributing source of fecal loading to a waterbody, added considerations
for planning and scoping are warranted. Soller et al. (2010b) conducted a QMRA evaluating the relative
risks of gastrointestinal illness between human and animal fecal sources. They found that the potential
risk from recreational exposure to direct deposition of fresh cattle feces in an untreated ambient
recreational water body can be equivalent to the risk from treated human point source discharges at
the same level of culturable enterococci. The fecal pathogen profiles differ between cattle and human
fecal sources. Conversely, when using culturable E. coli to estimate the potential risk of illness, direct
deposition of cattle feces was less risky than treated human discharges but similar to the risk posed by
raw sewage at the same level of culturable E. coli. The difference in risk estimates can be affected by
the relative levels of pathogens reported in fecal sources compared to levels of the FIB (see Table 2-7).
Cattle excrete, on average, much more E. coli compared to enterococci. Furthermore, when
considering certain fate and transport characteristics, such as land deposition and event-driven runoff
on the potential risks from pastured cattle, the probability of gastrointestinal illness from recreational
exposures to cattle feces can be reduced compared to the direct deposition scenario (WERF, 2011;
Soller et al., 2015). Other QMRA-based analyses have reported the potential to overestimate health
risks when microbial decay characteristics are not considered in the modeling (Wu et al., 2020). When
best management practices (BMPs) prevent direct contact of cattle with water and cattle feces are well
managed to prevent mass loading to water, the risk of gastrointestinal illness from cattle feces can be
reduced (U.S. EPA, 2010c; Soller et al., 2015; Wu et al., 2020).
2.1.1.1 Prevalence and Abundance of Pathogens in Animal Feces
The observed ranges of pathogens in the feces of different animals and humans (Table 2-2) have been
reported in the scientific literature (Schoen and Ashbolt, 2010; Soller et al., 2010b, 2018; U.S. EPA,
2010c; Boehm and Soller, 2020). For this TSM, the criteria used to select pathogen prevalence and
abundance data and numerical ranges from the literature for use in QMRA were:
1. Preferred data were from large-scale studies of long duration.
2. Conducted in the United States.
3. Based on individual (not composite) samples.
The values in Table 2-2 can be used to conduct QMRA as outlined in Section 3, Step 3. The ranges in
Table 2-2 can be included in the QMRA modeling as a distribution between the high and low values,
with each value equally likely (i.e., uniformly distributed). Table 2-2 includes human (raw and treated
sewage), and animal (cattle, chickens, gulls, and pigs) sources. E. coli 0157:1-17 has been found in cattle,
human, and pig feces but has not been reported in chicken or gull feces. Campylobacter and
Salmonella were found in human, pig, chicken, and gull feces. Cryptosporidium and Giardia were found
in human and pig feces but not reported in chicken or gull feces. Norovirus was found only in human
feces. The six pathogens, E. coli 0157:1-17, Campylobacter, Salmonella, Cryptosporidium, Giardia, and
norovirus, were all detected in raw sewage (Stampi et al., 1993; Wallis et al., 1996; Grant et al., 1996;
Jimenez-cisneros et al., 2001; Lemarchand and Lebaron, 2003; Metcalf and Eddy, 2003; Rose et al.,
28
-------
Table 2-2. Density of reference pathogens in human and selected animal fecal waste (adapted from Soller et al. [2010b, 2018]).
Fecal source
Density3 of:
E.coli 0157:H7 Campylobacter spp. Salmonella spp. Cryptosporidium spp. Giardia lamblia
Norovirus
mirr
max"
min
max
min
max
min
max
min
max
Raw sewage
Ranged
NDe
3.3
2.95
4.6
0.48
7.38
-0.52
4.38
0.51
4.95
Log mean = 4.7
(Log StDev = 1.5)
Attenuation range
(logio)f
2.2 to 3.0
NA1
Human Secondary treated
and chlorinated
effluent
Ranged
ND
ND
ND
-1.0
1.5
-1.0
2.1
Cattle
Ranged
Basis'
Typek
3.1g
8.4
1.2
7.3
3.0"
5.8
2.3
3.9
4.9
W
C
W
D
W
c
w
c
w
D
Animal Chicken
Ranged
Basis'
Typek
Manure""
NR
2.8
6.5
-1.0
4.5
ND
NR
NA
W
D
F
D
D1
F
Gulls
Ranged
NR"
3.3
6.0
2.3
9.0
NR
NR
NA
PigsP
Ranged
Typek
ND
7.0
2.0
5.7
4.9
1.7g
3.6
6.8°
NA
D
D
D
Notes:
a. Values represent exponent in the power of 10 exponential (e.g., n = lOn)
b. Denotes minimum observed value
c. Denotes maximum observed value
d. For raw sewage and secondary chlorinated effluent, units of minimum and maximum observations are CFU per liter (L) or oocysts per L or cysts per L; for livestock wastes, units
are CFU per gram (g) or oocysts per g or cysts per g; norovirus is in units of genomes per L.
e. Not detected (ND).
f. Removal range (rather than range of density). Attenuation in treatment (in loglO units) is assumed to be uniformly distributed.
g. GM (minimum observed density not reported).
h. Low end of range of values "typically measured in cattle manure." Actual minimum not presented.
i. Not applicable (NA), not usually considered to be present in this source.
j. Basis refers to weight basis for manure. D denotes dry weight and W denotes wet weight,
k. Sample type is either composite (C) or direct (D).
I. Samples were taken at random from the top of the litter pile. Because the droppings were fresh, they presumably came from a single bird,
m. Chicken manure type is litter (L) or fresh (F).
n. None reported (NR); no data were found in the literature to quantify densities in this source,
o. Estimated from data presented graphically.
p. All pig fecal abundances reported are for solid, fresh fecal samples (not slurries or treated manure).
29
-------
2004; Garcia-Aljaro et al., 2004, 2005; Haramoto et al., 2006; Robertson et al., 2006; Katayama et al.,
2008). In effluent from secondary treatment and chlorination, E. coli 0157:H7, Campylobacter, and
Salmonella, were not detected (ND) (Stampi et al., 1993; Lemarchand and Lebaron, 2003; Rose et al.,
2004; Garcia-Aljaro et al., 2004; Haramoto et al., 2006; Whiley et al., 2013). Campylobacter and
Salmonella were the only pathogens included for gulls in the QMRA model (Levesque et al., 2000;
Fogarty et al., 2003; Haack et al., 2003; Schoen and Ashbolt, 2010). For pigs, E. coli 0157:H7,
Campylobacter, Salmonella, Cryptosporidium, and Giardia were included in the QMRA model (Weijtens
et al., 1999; Cornick and Helgerson, 2004; Hutchison et al., 2004; Maddox-Hyttel et al., 2006; Peu et al.,
2006). For chickens, Campylobacter and Salmonella were the only pathogens included in the QMRA
model (Kraft et al., 1969; Terzich et al., 2000; Cox et al., 2002; Hutchison et al., 2004; Doane et al.,
2007; Brooks et al., 2009). For cattle, fecal densities of E. coli 0157:H7, Campylobacter, Salmonella,
Cryptosporidium, and Giardia were found in the literature (Hutchison et al., 2004; Atwill et al., 2006;
Moriarty et al., 2008).
Note that the units for the different enumeration methods differ. For example, culture-based methods
are usually in CFU or most probable number (MPN), whereas molecular methods can be in genome
copies, cell equivalents, or PCR units. Transparent risk assessment documentation includes a discussion
of the units for the various parameters, specifically where occurrence data do not have the same units
as the dose-response relationship. In some cases, a conversion factor for units might be necessary
(McBride et al., 2013).
The proportion of animals shedding pathogens, also known as prevalence, can differ among animals,
with cattle producing a high proportion off. coli 0157:H7 and Cryptosporidium loading, poultry and
dairy cattle contributing significantly to Campylobacter loading, and swine and poultry contributing to
loadings of Salmonella. Manure handling methods vary widely among U.S. farms, with some directly
depositing manure on pastures as solids or slurries (Bicudo and Goyal, 2003; U.S. EPA, 2013a). Manures
might also be stored for variable amounts of time before application, which can differentially affect the
attenuation of pathogens and the viability of bacterial indicators (Sobsey et al., 2006; Soller, 2015). The
exposure modeling includes pathogen shedding (Table 2-3). The prevalence columns in Table 2-3
represent the percentage of animals shedding within a herd or flock at any given time. The average of
the minimum and maximum values was calculated as a point estimate and included in the dose
estimation calculations. For example, for Campylobacter in pigs, 72% was used (0.72 probability). For
gulls, of the zoonotic pathogens, only Campylobacter and Salmonella were reported (Levesque et al.,
2000). Campylobacter and Salmonella in gulls had a lower bound based on levels detected by
quantitative polymerase chain reaction (qPCR) (Lu et al., 2011; Rodriguez et al., 2012) and an upper
bound of 100% based on composite samples (Levesque et al., 2000). Of the five zoonotic pathogens
listed, all but Campylobacter had less than 50% prevalence in pigs (Heitman et al., 2002; Cornick and
Helgerson, 2004; Dorner et al., 2004; Hutchison et al., 2004; Xiao et al., 2006; Dorr et al., 2009). In
chickens, Campylobacter and Salmonella had higher than 50% prevalence (Ley et al., 1988; Chapman et
al., 1997; Byrd, 1998; Martin et al., 1998; Cox et al., 2002; El-Shibiny et al., 2005; Doane et al., 2007).
30
-------
Table 2-3. Literature-reported prevalence (%) of reference pathogens in cattle, chicken, gulls, and
pigs (adapted from Soller et al. [2010b]).
Source
E. coli 0157:H7
Campylobacter spp.
Salmonella enterica
Cryptosporidium spp.
Giardia lamblia
mina
maxb
min
max
min
max
min
max
min
max
Cattle
9.7
28
5
38
5
18
0.6
23
0.2
37
Chickens
0
0
57
69
0
95
6C
27°
NRd
Gulls6
NR
54
100
75
100
NR
NR
Pigs
0.1
12
46
98
7.9
15
0
45
3.3
18
Notes:
a. Denotes minimum observed value.
b. Denotes maximum observed value.
c. The Cryptosporidium strains in chickens are not infectious to humans (Ley et al., 1988). In Tables 2-2 and 2-4 (below), human-infectious
Cryptosporidium was considered undetected in chickens.
d. NR, no data were found in the literature to quantify densities in this source.
e. For gulls, fecal prevalence and abundance data were based on observations from composite samples. All samples yielded
Campylobacters and salmonellae, so an upper bound of 100% prevalence was used. No samples yielded £ coli 0157:1-17,
Cryptosporidium, or Giardia.
The relative fraction of human-infectious strains of each reference pathogen in the nonhuman sources
can be variable; therefore, assigning an infectious fraction value for any particular source has an
associated uncertainty. Insufficient data were available to assign consistent quantitative values for this
parameter. Thus, a qualitative assessment of this parameter was used where categorical values of low
(L), medium (M), or high (H) were assigned to each pathogen for each nonhuman source (Table 2-4).
The qualitative potential for human infection was based on the known proportion of human-infectious
species/strains/serotypes/isolates in animal feces. For human sources, the relative fraction of human-
infectious strains is assumed to be 1.0 because the indicator and pathogen data are from sewage and
not individual fecal samples and, therefore, already account for the fraction of humans who shed at
any given time. For cattle, chickens, gulls, and pigs, point estimates were used to characterize the
fractions of human-infectious strains based on the midpoint of the ranges of 0%-33% for low, 33%-
66% for medium, and 67%-100% for high. Therefore, low is 16.5%, medium is 50%, and high is 83.3%.
In gull feces, Campylobacter and Salmonella were considered to have a high fraction of human
infectious strains (Levesque et al., 2000). In pig feces, Cryptosporidium was considered to have a low
fraction of human infectious strains, Salmonella a medium fraction, and E. coli 0157:1-17,
Campylobacter, and Giardia a high fraction (Heitman et al., 2002; Cornick and Helgerson, 2004; Dorner
et al., 2004; Hutchison et al., 2004; Xiao et al., 2006; Dorr et al., 2009). In chicken feces, Campylobacter
and Salmonella were considered to have a medium fraction of human infectious strains (Chapman et
al., 1997; Byrd, 1998; Ley et al., 1988; Martin et al., 1998; Cox et al., 2002; El-Shibiny et al., 2005;
Doane et al., 2007). In cattle feces, E. coli 0157:1-17, Campylobacter, Cryptosporidium, and Giardia were
considered to have a high fraction of human infectious strains and Salmonella was considered to have
a medium fraction of human infectious strains (Wesley et al., 2000; Fayer et al., 2000; Wade et al.,
2000; Hoar et al., 2001; Sturdee et al., 2003; Hutchison et al., 2004; Fossler et al., 2005; Atwill et al.,
2006; Berry et al., 2007).
31
-------
Table 2-4. Fraction of human infectious strains of reference pathogens in cattle, chickens, gulls, and
pigs (adapted from Soller et al. [2010b]).
Source
E. coli
0157:H7
Campylobacter
spp.
Salmonella
enterica
Cryptosporidium
spp.
Giardia lamblia
Cattle
H
H
M
H
H
Chickens
NRa
M
M
NR
NR
Gulls
NR
L
L
NR
NR
Pigs
H
H
M
L
H
Notes: Potential for human infection (H = high, M = medium, L = low) was based on the prevalence of known human-infectious
species/strains/serotypes/ isolates in animal feces,
a. NR; no data were found in the literature to quantify densities in this source.
2.1.1.2 Environmental Fate and Transport
Upon excretion, enteric pathogens are exposed to very different conditions on land or in water
compared to the gastrointestinal tracts of animals or humans in which they previously resided.
Numerous physicochemical (abiotic) and biotic factors can affect the viability and infectivity of these
pathogens, and thus the magnitude of potential health risks, as they move from the point of excretion
to the point of exposure (Bradford et al., 2013; Ahmed et al., 2018; Korajkic et al., 2018). Pathogen
growth, persistence, and attenuation vary between pathogens, between environmental media, and
with other conditions such as moisture content (or water activity), sunlight, temperature, pH,
humidity, salinity, predation, and organic matter (Gale, 2005; Englehardt et al., 2009; U.S. EPA, 2010c:
Annex 3; Ahmed et al., 2018). Zoonotic pathogens may survive in the environment for days to months,
depending on the extant conditions, but survival rates for the specific pathogens of concern may not
be well-defined for all possible conditions (Roberts et al., 2016). Uncertainty in the estimates of
pathogen loads can be large, particularly for pathogens like Cryptosporidium spp. that can be highly
infectious and for pathogens with frequent nondetection. These variability and uncertainty issues
around model parameters have been characterized in previous QMRAs (Signor and Ashbolt, 2006;
Petterson et al., 2007, 2009).
Atwill et al. (2002) describe the steps in the pathway followed by enteric microorganisms from fecal
material to a specified downslope location (in this case, a receiving stream) as follows:
1. Rainfall of sufficient intensity erodes the top layer of the fecal material, releasing pathogenic
organisms onto the wetted soil surface;
2. Rainfall intensity reaches infiltration capacity, and pathogenic organisms are carried downslope
via sheet flow, preferential rill flow, or exfiltration in variable source areas;
3. Pathogens are transported downslope or infiltrate into the subsurface; and
4. Vegetative buffer strips intercept flows laden with pathogenic organisms and enhance
infiltration.
Ferguson et al. (2003) divided processes governing the relationship between watershed pathogens and
densities in surface waters into those most important in organism inactivation (water/osmotic
potential, temperature, sunlight, pH, and inorganic and organic nutrients) and those most important in
transport (adsorption/desorption effects, hydrological movement, and mechanical or biological
32
-------
movement). Figure 2-2 illustrates the various factors affecting animal waste as it moves from the point
of excretion to receiving waters via multiple pathways. The transport of pathogens and fecal indicator
organisms in the waste is influenced both by the pathways for feces entering waterbodies and by
additional factors that impact survival and attenuation of the microbes; each process and factor have
associated variability.
Avian fecal contamination can also reach recreational waters by multiple routes (Figure 2-3), including:
• Direct deposition as feces into the water column.
• Direct deposition via mechanical transfer (e.g., carried to receiving water on the legs of birds
wading in sewage or sewage-impacted water) into the water column.
• Resuspension of fecal-associated pathogens from sediment.
• Precipitation-induced runoff of pathogens (either deposited or progeny of deposited
organisms) from soil, vegetation, or impervious areas near the recreation area.
• Advection of bird-origin feces from stocks (e.g., in wetlands hydraulically connected to
recreation site waters during high tides or flooding, as noted by He et al., 2007).
Figure 2-2. Factors affecting the viability of pathogens and indicators from deposition to surface
Barnyard runoff
/'/ / ' /'
Storage
system
leakage
Wate
waters (adapted from Rosen [2000]).
33
-------
Direct-deposition:
* In fecal material
~ Mechanical
Advection from
offsite
Resuspension ( <*'
from sediments " ^ i*"
Figure 2-3. Routes of avian fecal contamination at a recreational freshwater beach.
Microbial runoff from soil is affected by numerous factors. Rather than characterize each factor
separately, the EPA conducted field experiments for a defined typical rain event (Appendix E in U.S.
EPA, 2010c). Plots receiving cattle, chicken, and swine manure applications at typical agronomic rates
were subjected to an intense (< 100-year return period) rain event and the plot-scale runoff was
monitored for reference pathogens and FIB. The calculated mobilization fractions for the zoonotic
reference pathogen and FIB were subsequently incorporated into a QMRA, and risks from recreational
exposure to the undiluted runoff were characterized (Soller et al., 2015).
Selection of pathogen enumeration methods can also affect fate and transport and potentially
influence risk characterization (Rogers et al., 2011; Eichmiller et al., 2014). The use of culture-
dependent or culture-independent (e.g., gene-based) enumeration methods can provide different
information on the environmental decay of a pathogen and may or may not provide information
characterizing the infectivity of the target. For example, the ultraviolet (UV) light component of
sunlight typically increases decay rates, but the magnitude of this effect is influenced by other
environmental conditions and the methodology used to measure the target (Korajkic et al., 2018). UV
exposure can attenuate viable cells, which affects culture-based enumerations more readily than the
corresponding nucleic acids. In low-nutrient, high-stress environments, many pathogens can enter a
"viable, but non-culturable" (VBNC) state (Roberts et al., 2016). VBNC pathogens that are not
detectable using culture-based methods may be detected by a gene-based enumeration method, such
as qPCR.
The EPA used a health-protective approach in its QMRA framework that minimizes the variability
associated with these fate and transport processes. In the exposure scenario to estimate human health
effects, the EPA assumes direct deposition of feces to a waterbody or indirect deposition characterized
as exposure to zoonotic pathogens in runoff because the levels of pathogens in these scenarios are the
highest and would result in health-protective estimates. Although these assumptions might
overestimate the potential for illness associated with inputs from a given source of fecal
34
-------
contamination, the estimates generated with this approach minimize the uncertainties accompanying
the multiple microbial fate and transport and loading parameters. Sunger et al. (2019) estimated illness
risk to swimmers exposed to treated effluent-affected waters and incorporated the literature-reported
microbial decay rates in freshwaters. Additional fate parameters can be included in the QMRA model
discussed in this TSM, but the EPA's approach is more straightforward to conduct and allows for
consistent application of the framework across many sites with varying characteristics. Illness
probability estimates using this QMRA framework represent an upper bound on the potential for
human health effects from recreational exposure to a nonhuman source of fecal contamination.
2.1.1.3 Infectivity and Illness
This subsection presents summary information on the reference pathogens' ability to cause infection
given exposure and the probability of illness given infection. Recreators' exposure to fecal
contamination in ambient waters can result in illness and, therefore, be a public health concern.
Infectivity refers to the ability of a pathogen to cause or transmit an infection (U.S. EPA and USDA,
2012). The probability of illness in an exposed individual results from an interaction between the
presence of the pathogen in the environment, the virulence and pathogenicity of the pathogen, and
the immune status of the host. Figure 2-4 displays the epidemiological triangle, which is a model
conceptualizing the agent-environment-host interaction as a means to consider parameters that
influence public health outcomes (CDC, 2011; U.S. EPA, 2014c). Illness occurs when a pathogen capable
of causing disease infects a host that is vulnerable to infection (Gulis and Fujino, 2015).
Figure 2-4. Epidemiologic triangle (adapted from CDC [2011]).
35
-------
Recreational epidemiological studies (Cabelli, 1983; Dufour, 1984; Kay et al., 1994; Priiss, 1998; Wade
et al., 2003, 2006, 2008, 2010; Fewtrell and Kay, 2015) have evaluated health endpoints among
recreators with self-reported symptoms following exposure to waters affected by fecal contamination
and include:
1. "Gastrointestinal illness," typically defined by any of the following within 10-12 days after
swimming: (a) diarrhea (three or more loose stools in a 24-hour period), (b) vomiting, (c)
nausea and stomachache, or (d) nausea or stomachache and impact on daily activity;
2. "Upper respiratory illness," which was defined as any two of the following: sore throat, cough,
runny nose, cold, or fever;
3. "Rash," which was defined as a rash or itchy skin;
4. "Eye ailments," which were defined as either an eye infection or a watery eye;
5. "Earache," which was defined as ear pain, ear infection, or runny ears; or
6. "Infected cut," which was defined as a cut or wound that became infected.
At a marine subtropical beach with nonpoint source inputs, Sinigalliano et al. (2010) reported an
increase in gastrointestinal, respiratory, and skin illnesses among bathers versus nonbathers. Colford et
al. (2007) characterized health outcomes among recreators at Mission Bay, California, a nonpoint
source-affected beach. They found an increase in diarrhea and rashes among those having any water
contact and increased reports of cramps and eye ailments in those who swallowed water. Calderon et
al. (1991) studied a freshwater recreation site with nonpoint source animal fecal inputs and no known
human fecal inputs; they reported an increase in gastrointestinal illness among bathers, although
increases in illness were associated with increasing bather density. Recreational water epidemiological
studies generally do not directly characterize the etiologic agents of the reported illness in the waters
studied. Some of the health endpoints listed above can be caused by one or more pathogens, and, in
some cases, multiple endpoints can be attributed to a single pathogen.
Published QMRA studies characterizing different fecal contamination sources and dynamics have
demonstrated that the nature (i.e., type of source and how the source affects surface waters) and
magnitude of the fecal source(s) affecting a waterbody directly influence the probability of illness in
recreators exposed to the contamination (Table 1-2). The etiologic agents of concern differ for human
and nonhuman sources of contamination (Schoen and Ashbolt, 2010; Soller et al., 2010a,b; McBride et
al., 2013; Sunger et al., 2019). Using QMRA, the EPA evaluated direct and indirect deposition from
human and animal fecal contamination and mixtures of human and animal fecal contamination sources
for the probability of illness, the potential etiologic drivers of illness, and estimating equivalently
protective FIB levels (compared to the EPA's 2012 RWQC) for waters predominantly affected by
nonhuman fecal sources (U.S. EPA, 2010a; Soller et al., 2014, 2015). The case study examples described
in this document illustrate that the potential risk associated with a specific level of FIB from nonhuman
fecal contamination can be significantly lower compared to human sources (Soller et al., 2010b, 2015).
The risk differential is influenced by the difference in pathogen profiles present in different sources,
the loading of that source to a waterbody and the relative level of culturable FIB among fecal sources.
Other QMRA examples from the literature demonstrate the source with the highest pathogenicity, not
necessarily the source(s) of greatest fecal loading, influences the risk of adverse health effects (Schoen
36
-------
et al., 2011; Soller et al., 2014). These examples help to illustrate that considering the risk from
nonhuman fecal loading can allow for the development of alternative WQC.
The following subsections describe the reported health outcomes, dose-response relationships, and
morbidity reported in the literature for each reference pathogen included in the QMRA framework
discussed in this TSM document.
2.1.1.3.1 Hazard Identification for Reference Pathogens
The eight reference pathogens discussed in this document are agents that are known to cause
waterborne illness and account for more than 97% of nonfood borne illness in the United States (Mead
et al., 1999; U.S. EPA, 2009b, 2010a; Scallan et al., 2011a,b). The CDC reported 140 recreational water-
associated outbreaks caused by pathogen, toxin, or chemical exposure from 2000 to 2014 (Graciaa et
al., 2018). Enteric pathogens accounted for 80 documented outbreaks of waterborne disease. Although
waterborne disease outbreak surveillance information represents an underreporting of the actual
potential health burden, the reference pathogens discussed in this document were reported to
account for approximately 53% of all recreational water-associated outbreaks from 2000 to 2014,
where an etiologic agent was determined (Graciaa et al., 2018).8 Salmonella and rotavirus were not
identified as a known etiologic agent in these outbreaks.
Vanden Esschert et al. (2020) report on illness outbreaks associated with untreated recreational water
in California, Maine, and Minnesota during 2018 and 2019. These outbreaks were voluntary entries in
the CDC's National Outbreak Reporting System (NORS). Summary information provided on 31 states
for the 2009-2019 period included 88 outbreaks with confirmed etiologies: norovirus (19 [22%]
outbreaks; 1,858 cases); Shiga toxin-producing E. coli (STEC) (19 [22%] outbreaks; 240 cases);
Cryptosporidium (17 [19%] outbreaks; 237 cases); and Shigella (14 [16%] outbreaks; 713 cases)
(Vanden Esschert et al., 2020). Among the reference pathogens discussed in this TSM, E. coil 0157:1-17
is the most commonly identified STEC in the United States. (Minnesota Department of Health, 2022).
In this section, information about infectivity and adverse health outcomes reported in the literature for
each of the reference pathogens is summarized. Discussion of how and where the pathogen(s) initiates
infection is helpful for understanding the context of how different exposures lead to infections. The
mechanisms of infection are not covered in detail in this document because the multiple reference
pathogens cover the three major groups of pathogens (i.e., viruses, bacteria, and protozoa) and their
mechanisms differ between and among the groups. For all reference pathogens discussed in this
document, gastrointestinal illness is a common acute adverse health outcome. The following
summaries are not exhaustive but are provided to establish each pathogen's relevance to recreational
water illness.
Adenovirus. Adenovirus infections have been identified as a causal agent of lower and upper respiratory
disease, acute gastrointestinal illness (AGI), acute conjunctivitis, cystitis, pharyngoconjunctival fever, and
diseases of the central nervous system and urinary tract (Enriquez and Thurston-Enriquez, 2006;
Teixeiria et al., 2020). Adenoviral lower respiratory tract infections are infrequent, sporadic, and most
commonly associated with adenovirus types 3, 5, and 7 (Murtagh et al., 1993; Mandell, 2000).
Epidemic keratoconjunctivitis is associated with adenovirus serotypes 8, 19, and 37. Acute respiratory
8 Graciaa et al. (2018) report outbreaks for Campylobacter, E. coli, Shigella, Cryptosporidium, Giardia, adenovirus, and
norovirus. In this TSM, E. coli 0157:H7 is an index pathogen for the broader group of toxin-producing E. coli and other gram-
negative, facultative anaerobic, nonspore-forming bacteria capable of causing waterborne disease (e.g., Shigella sp.).
37
-------
disease is most often associated with adenovirus types 4 and 7. Enteric adenoviruses 40 and 41 cause
gastroenteritis, usually in children (Wilhelmi et al., 2003). Many adenovirus serotypes can multiply in the
small intestine, but only types 40 and 41 have been strongly associated with gastroenteritis (Grimwood
et al., 1995). Adenoviruses have a low infective dose with approximately 5 viral particles needed to cause
disease in a susceptible individual (Teixeiria et al., 2020).
Children, infants, older adults, the immunocompromised, and other sensitive populations can be
especially vulnerable to developing infections (Jiang, 2006). Infection is most often spread via direct
contact with an infected individual, the fecal-oral route, aerosol inhalation, or recreational water
activities (which can include both oral and inhalation exposure). Adenovirus exposure has been
significantly associated with gastrointestinal illness in recreational waters (Marion et al., 2014).
Although not thought to be the most common route of transmission, two drinking water outbreaks
have been reported in Europe in which enteric adenoviruses were suspected as the cause of the
reported acute gastroenteritis (Mena and Gerba, 2009). Adenoviruses in drinking water also might
contribute to viral infections of unknown etiology (Ko et al., 2003).
Neurological sequelae from Reye's syndrome following infections by adenovirus have been reported
(Daugherty and Heubi, 1985). For example, studies of Reye's syndrome survivors have shown sequelae
ranging from severe psychomotor retardation to mild specific perceptual and language impairments,
intelligence quotient (IQ) deficits, and chronic behavioral deficits that lasted from six to 18 months
after infection (Davidson et al., 1978; Brunner et al., 1979; Shaywitz et al., 1982).
Norovirus. Noroviruses are highly contagious and transmitted primarily through the fecal-oral route-
most commonly by ingesting fecally contaminated food or water or through person-to-person spread
(Schwab and Hurst, 2006; CDC, 2023). Three of the six norovirus genotypes are infective to humans
and are referred to as genogroups I, II, and IV. Genogroup II noroviruses are thought to be the
predominant cause of infections (Teixeira et al., 2020). In temperate latitudes, norovirus is associated
with "winter vomiting disease" (Hutson et al., 2004). Environmental and fomite9 contamination can
also be an important source of transmission and infection. Matthews et al. (2012) conducted a meta-
analysis of outbreaks (n = 902) and found that the primary attack rate for food (50%) and water (38%)
was higher than the attack rate for other exposures (e.g., person-to-person [27%], fomites [26%]).
Waterborne outbreaks of norovirus disease in community settings are often caused by sewage
contamination of wells and recreational water (CDC, 2023) and are widely considered to be the
causative agents in many viral waterborne outbreaks of unknown etiology (Maunula et al., 2005; U.S.
EPA, 2006c). Although pre-symptomatic viral shedding can occur, shedding usually begins with the
onset of symptoms and can continue for up to two weeks after recovery from illness (CDC, 2023). The
incubation period for norovirus-associated AGI in humans is usually between 24 hours and 48 hours
(median in outbreaks: 33-36 hours), but cases can occur within as little as 12 hours of exposure.
Infection usually presents as acute-onset vomiting, watery nonbloody diarrhea with abdominal cramps,
and nausea. A low-grade fever also can occur, although dehydration is the most common complication,
especially among young children and older adults. Symptoms can usually last 2-4 days (Teixeira et al.,
2020). Adverse health outcomes are higher among young children, older adults, and
immunocompromised persons (Hutson et al., 2004). A human volunteer feeding study was conducted
to determine the infectivity of norovirus in healthy adults to predict the likelihood of enteric infection
following ingestion exposure of norovirus in water (Teunis et al., 2008a). Although asymptomatic
9 A fomite is an inanimate object or substance that can transmit infectious organisms from one individual to another.
38
-------
infection can occur in as many as 30% of infections, its role in calicivirus/norovirus transmission
remains poorly understood.
Rotavirus. Rotavirus infection is the leading cause of acute severe gastroenteritis in young children and
infants in the United States and worldwide (Abbaszadegan, 2006). Symptoms of rotavirus infection
include diarrhea, fever, and vomiting, leading to dehydration—especially in young children and infants.
Rotaviruses are excreted in very large quantities in the feces of infected persons with diarrhea
(1011 per mL), so they are present in relatively high densities in wastewater and ambient water (Ansari
et al., 1991). Rotavirus is highly contagious (Gerba et al., 1996). Infection is most often spread via the
fecal-oral route, contact with feces or virus-contaminated objects, direct contact with an infected
individual, or contaminated food or water. In 2008, the CDC reported that each year, rotavirus
infections accounted for an estimated 55,000 to 70,000 hospitalizations in the United States (CDC,
2008). The CDC estimates that introducing a rotavirus vaccine in 2006 has prevented an estimated
40,000 to 50,000 hospitalizations annually among U.S. infants and young children (CDC, 2018). Since
the introduction of vaccines against rotavirus in 2006, the number of cases in the United States has
declined. Although not common, drinking-water-related rotavirus outbreaks have occurred in the
United States (e.g., Hopkins et al., 1984, 1985). Rotaviruses have been detected in treated drinking
water (Abbaszadegan et al., 1999) and nondisinfected well water (Borchardt et al., 2003). Rotaviruses
in drinking water also can contribute to waterborne outbreaks of unknown etiology and unknown
levels of endemic disease.
Campylobacter jejuni. Campylobacter sp., most notably C. jejuni, is a major zoonotic pathogen. The
primary adverse health endpoint associated with C. jejuni infection is AGI, characterized by diarrhea,
fever, and abdominal cramps (Butzler, 2004). The incubation period is typically two to five days but can
extend up to 10 days. In about 50% of patients, diarrhea is preceded by a febrile period associated with
malaise, myalgia, and abdominal pain; fresh blood can also appear in the stool by the third day. Most
C. jejuni infections are sporadic in nature and are often associated with outbreaks where the source of
pathogen is not determined (Hanninen et al., 2003). In drinking water outbreaks attributed to C. jejuni
(e.g., Jones and Roworth, 1996; Holme, 2003), the drinking water source is usually shown to be fecally-
contaminated by runoff of surface water after rain or by leakage of a sewage pipe close to the drinking
water pipeline.
McBride et al. (2002) found that 60% of all samples collected from 25 freshwater recreational water
sites in New Zealand over 15 months contained at least one species of Campylobacter. This finding led
to an inference that 4% of campylobacteriosis cases in New Zealand were due to water contact
recreation (McBride et al., 2002).
C. jejuni is increasingly recognized as a risk factor for Guillain-Barre syndrome, a common cause of
neuromuscular dysfunction (Bunning et al., 1997) that occurs in one in 1,000 infections (Butzler, 2004).
Guillain-Barre syndrome is the most serious complication of Campylobacter infection (Kaldor and
Speed, 1984; Nachamkin et al., 2000; Yuki, 2001; Ang et al., 2002; Gilbert et al., 2004). Studies by
McCarthy and Giesecke (2001) have shown that the risk of developing Guillain-Barre syndrome during
the two months following a symptomatic episode of C. jejuni infection is approximately 100 times
higher than the risk in the general population.
Reactive arthritis is a rare complication of infection (FDA, 2012). Arthritis typically occurs within two
weeks after the onset of gastrointestinal illness, but onset time can range from four to 35 days
(Anonymous, 1998). The condition is characterized by infection at a distant site, whereby joint
39
-------
inflammation occurs without typical evidence of sepsis at the affected joint(s) (Shirtliff and Mader,
2002).
Reiter's syndrome is an aseptic arthritis (Kim et al., 2007) that is triggered by an infectious agent
located outside the joint. It also has a strong association with human leukocyte antigen-B27. Pope et al.
(2007) reviewed the epidemiological literature on Campylobacter-associated reactive arthritis and
found that follow-up for long-term sequelae related to Campylobacter and Reiter's syndrome was
largely unknown.
Immunoproliferative small intestinal disease is an infectious pathogen-associated occurrence of human
lymphomas (Al-Saleem and Al-Mondhiry, 2005), and molecular and immunohistochemical studies have
demonstrated an association with C. jejuni (Lecuit et al., 2004). This disease involves the proximal small
intestine, resulting in malabsorption, diarrhea, and abdominal pain, as well as weight loss, intestinal
obstructions, and abdominal masses (Gilinsky et al., 1987; Rambaud et al., 1990; Fine and Stone, 1999;
Al-Saleem and Al-Mondhiry, 2005). Symptoms are often chronic, and patients can experience mild
symptoms for five to 10 years before developing higher-grade illness or lymphoplasmacytic and
immunoblastic lymphomas (Al-Saleem and Al-Mondhiry, 2005).
E. coli Q157:H7. E. coli 0157:1-17 has been found worldwide in feces-contaminated surface water,
groundwater, nonchlorinated or inadequately treated drinking water and swimming pools, soil and
sediment, and a wide variety of foods (Muniesa et al., 2006). It has been documented to cause both
endemic illness and outbreaks in humans through fecal-oral transmission, consumption of feces-
contaminated food and water, and contact with infected persons and animals (Rangel et al., 2005). The
clinical symptoms of infection vary from nonbloody diarrhea to bloody diarrhea, referred to as
hemorrhagic colitis. Hemorrhagic colitis is a serious, life-threatening condition and can lead to
hemolytic uremic syndrome, which results in renal damage and possibly death (Tarr, 1995).
Enterohemorrhagic E. coli infections can lead to long-term or permanent kidney damage and renal
disease. Persons who develop chronic kidney failure could require lifelong dialysis or a kidney
transplant. Garg et al. (2003) conducted a comprehensive review and meta-analysis of the current
literature along with expert consultations to quantify the long-term renal prognosis of patients with
diarrhea-associated hemolytic uremic syndrome. A higher severity of acute illness was strongly
associated with a worse long-term prognosis. Studies with a higher proportion of patients with central
nervous system symptoms (coma, seizures, or stroke), had a higher proportion of patients who died or
developed permanent end-stage renal disease at follow-up. Death or end-stage renal disease occurs in
about 12% of patients with diarrhea-associated hemolytic uremic syndrome, and 25% of survivors
demonstrate long-term renal sequelae.
Pathogenic E. coli are also implicated in rheumatoid diseases (Locht and Krogfelt, 2002; Schiellerup et
al., 2008), including reactive arthritis and Reiter's syndrome with similar pathologies as described
above. Pathogenic E. coli are phylogenetically related to Shigella, and some strains can produce shiga
toxin (M0lbak and Scheutz, 2004)
Salmonella enterica. Most Salmonella enterica infections can be traced to contaminated food products,
although water-related outbreaks (including drinking water outbreaks), have been reported (Schuster
et al., 2005; CDC, 2006; Covert and Meckes, 2006; Craun et al., 2006). The incubation period ranges
from 18-48 hours after ingestion, and illness is usually characterized by acute, self-limiting AGI
(although some infections can be severe), fever, and septicemia, lasting 2-5 five days (Covert, 1999).
40
-------
A broad range of chronic sequelae has been reported as consequences of Salmonella infection, such as
reactive arthritis, Reiter's syndrome, rheumatoid syndromes, pancreatitis, osteomyelitis, myocarditis,
colitis, cholecystitis, and meningitis (Thomson et al., 1995; Dworkin et al., 2001; Motarjemi, 2002).
Cryptosporidium sp. Cryptosporidium is a zoonotic waterborne pathogen of global public health
importance. The CDC estimates that ~700,000 cases of Cryptosporidium infections a year occur in the
United States (Scallan et al., 2011a). Worldwide, the estimated number of annual cases of
cryptosporidiosis exceeds several million, and infection has been documented in almost 100 countries
and on every continent except Antarctica (Casemore et al., 1997; Fayer et al., 1997; Caccio, 2005).
Cryptosporidium is a well-known cause of opportunistic infections among acquired immune deficiency
syndrome (AIDS) patients (Roy et al., 2004) and a common cause of outbreaks of gastrointestinal illness
(CDC, 2008). Surveillance data indicate that infections are more common among immunosuppressed
individuals and that 90% of cases are not involved in outbreaks (Dietz et al., 2000). Valid species of
Cryptosporidium have been described in more than 155 mammalian species, more than 30 avian
species, 57 reptilian species, one species offish, and two amphibian species (O'Donoghue, 1995; Fayer,
2004).
Most clinical cases of cryptosporidiosis involve infection by C. parvum or C. hominis. The most common
feature of cryptosporidiosis is profuse, watery diarrhea. Other clinical signs of infection include
dehydration, fever, anorexia, weight loss, weakness, and progressive loss of condition (O'Donoghue,
1995). Recovery is usually spontaneous within one to two weeks of infection. Developmental stages of
the parasite are often seen within the small intestine and occasionally elsewhere (stomach, colon,
liver, lungs). In general, the development of cryptosporidiosis depends on the host's species, age, and
immune status (Fayer et al., 1997). Younger persons with less developed or compromised immune
systems are often more susceptible to severe infection than healthy adults (O'Donoghue, 1995).
Several human volunteer feeding studies have been conducted to determine the infectivity of C.
parvum and C. hominis in healthy adults to predict the likelihood of enteric infection following
exposure to contaminated drinking water (DuPont et al., 1995; Chappell et al., 1996; Okhuysen et al.,
1999, 2002). A summary of the dose-response data for all six tested isolates has been published
(U.S. EPA, 2005). The EPA has also conducted additional dose-response modeling on Cryptosporidium
(Messner and Berger, 2016).
Giardia sp. Giardia intestinalis (also known as G. lamblia and G. duodenalis) is a zoonotic waterborne
pathogen of global public health concern. It is the most common intestinal parasite identified by public
health laboratories in the United States (Rose et al., 1991; Mead et al., 1999). CDC estimates that
approximately 1.2 million to 2 million illnesses occur annually in the United States due to Giardia
(Mead et al., 1999; Scallan et al., 2011a). Of the U.S. general population, 30% has seropositivity for
Giardia (indicates current or past infection), and 7% of small children are asymptomatically infected
with Giardia (Frost and Craun, 1998). More than 100 waterborne giardiasis outbreaks have been
reported worldwide from the beginning of the previous century to 2004 (Plutzer et al., 2010). Giardia
also infects various domestic and wild mammals (e.g., cats, dogs, cattle, deer, beavers) (Thompson,
2000; Ballweber et al., 2010).
41
-------
High-risk groups for giardiasis include infants and young children, travelers to developing countries, the
immunocompromised, and persons who consume untreated water from lakes, streams, and shallow
wells (U.S. EPA 1998a; CDC, 2008). A wide spectrum of symptoms is associated with giardiasis, from
asymptomatic infection and acute self-limiting AGI to persistent chronic diarrhea, which might fail to
respond to treatment. Asymptomatic infection is very common (50%-75% of infected persons are
symptomatic) (Mintz et al., 1993; U.S. EPA 1998a). Symptoms of giardiasis include diarrhea, abdominal
cramps, bloating, weight loss, and malabsorption (Rodriguez-Hernandez et al., 1996; Hellard et al.,
2000; Thompson, 2000). Case reports also indicate that giardiasis might be associated with developing
reactive arthritis (Tupchong et al., 1999). Giardia infection is frequently self-limited, but
immunocompromised persons can have more serious and prolonged infection (Benenson, 1995).
Hospitalizations and deaths due to giardiasis are relatively rare; CDC estimates that giardiasis causes
approximately 10-30 deaths and 3,500-5,000 hospitalizations annually in the United States (Mead et
al., 1999; Scallan et al., 2011a).
In some patients, symptoms last for only three or four days; in others, the symptoms last for months.
Usually, patients commonly resolve their infections spontaneously, with acute disease lasting from one
to four weeks (Smith and Wolfe, 1980). In some patients, however, the acute stage can persist for
months (Wolfe, 1990). The period of communicability lasts for the entire duration of infection, but the
shedding of cysts can be intermittent (Benenson, 1995).
Clinical data suggest that Giardia cysts are highly infective for humans (Rendtorff, 1954a,b; Rose et al.,
1991; Teunis etal., 1996).
2.1.1.3.2 Zoonotic Infectivity Potential
Upon exposure to a pathogen, there is some probability of an infection being initiated. Infection is
influenced by the pathogenicity or virulence of the pathogen and the susceptibility of the host.
Mathematically, the probability of infection is described by a "dose response" function. Dose-response
relationships for the reference pathogens presented in Table 2-5 were identified from the peer-
reviewed literature and are for an infection endpoint. Definitions of infection most often used in dose-
response models were seroconversion and shedding of pathogens in feces. The definition of reference
pathogen illness also varied but was usually related to the incidence of diarrhea or vomiting. Narrative
descriptions of the justification for the selected values follow.
Adenovirus. Teunis et al. (2016) provided a two-level dose-response model describing the distributions
of infectivity and pathogenicity in various challenge studies of adenovirus, incorporating differences in
inoculation route as shifts in average infectivity and pathogenicity. The combined challenge studies
included a total of 91 subjects challenged in 23 dose groups. The infection dose-response relationship
is an exact beta-Poisson (aka hypergeometric) relationship, and the conditional probability of illness
given infection is dose-dependent. The hierarchical model was implemented in a Bayesian framework,
the result of which was used to make predictions for the infectivity of adenovirus specific to any of the
four studied inoculation methods (ingestion, inhalation, intranasal, and intraocular). The mean
infection dose-response parameter estimates for oral exposures are a = 5.11, and 3 = 2.8.
42
-------
Table 2-5. Literature-reported dose-response relationships for reference pathogens.
Reference pathogen
Distributional forma
Parameter of
distribution13
Parameter values
Reference
Rotavirus
Beta-Poisson
a
P
0.26
0.42
Regli et al., 1991
Norovirus
Hypergeometric0
a
P
0.393
0.767
Teunis et al., 2020
W
mean = -0.608
variance = 1.79
covariance = -1.03
z
mean = 0.194
variance = 2.54
covariance = -1.03
Adenovirus
Hypergeometricd
a
P
5.11
2.8
Teunis et al., 2016
Cryptosporidium spp.
Exponential
r
0.09
U.S. EPA, 2006a
Giardia lamblia
Exponential
r
0.0199
Rose and Gerba,
1991
Campylobacter jejuni
Hypergeometric0
a
P
0.44
0.51
Teunis et al., 2018
W
mean = -0.177
variance = 1.303
covariance = -0.041
z
mean = 0.054
variance = 1.070
covariance = -0.041
Campylobacter jejuni
Beta-Poisson
a
0.145
Medema et al.,
P
7.59
1996
E. coli 0157:H7
Beta-Poisson
a
P
0.248
48.8
Teunis et al., 2008b
Salmonella enterica
Beta-Poisson
a
0.3126
Haas et al., 1999;
P
2884
Fazil, 1996
Notes:
a. Endpoints are infection. Dose-response model equations:
Exponential (Rose et al., 1991): Pinfect(dose\ r) = 1 — e~r dose
+ ~p~)
Hypergeometric (Teunis et al., 2008a): Pinfect(dose\ a,/3) = 1— 1F1(a, a +/?, — dose)
The 1F1 function represents the confluent hypergeometric function that does not have a close form but is approximated by the beta-
Poisson distribution when a « P and P » 1.
b. The parameters listed in this column correspond to the dose-response model (distributional form) listed for the pathogen.
c. Beta-Poisson model with the confluent hypergeometric function can use point estimates (a and P) or distributional estimates of a and
P derived from the intermediate parameters w and z. The w and z distributions are defined by their mean, variance, and covariance.
d. Teunis et al. (2016) only reported point estimates for a and P for adenovirus.
43
-------
Norovirus. Teunis et al. (2020) combined data from volunteer challenge studies and outbreaks in a
multilevel dose-response framework that built on earlier Bayesian modeling of norovirus dose
response (Teunis et al., 2008a). Infectivity and pathogenicity were analyzed by secretor status as a host
factor (Se+ or Se-) and genogroup as a pathogen factor (Genogroup I or Genogroup II). Nonsecretors
(Se-) are resistant to several norovirus genogroups (Nordgren and Svensson, 2019). The values
selected in Table 2-5 correspond to the Se+ and Genogroup I modeling parameters provided by Teunis
et al. (2020). The EPA selected this scenario because, in Se+ subjects, Genogroup I viruses appear
slightly more infectious than Genogroup II viruses (Teunis et al., 2020). The Teunis et al. (2020) meta-
analysis confirms the high infectivity of norovirus, with an estimated mean infection risk of 0.28 when
exposed to one genome copy (qPCR unit) of Genogroup I norovirus, and 0.076 for one genome copy of
Genogroup II norovirus, both in Se+ subjects.
Rotavirus. The rotavirus dose-response model is commonly used in risk assessment and is well
accepted. The model was developed using data from human feeding studies (Ward et al., 1986a,b).
Volunteers in the study were adult males 18-45 years old. Overall, the ratio of ill-to-infected
individuals was 0.67, and the progression of infection to illness did not appear to be dose-dependent.
Rose and Gerba (1991) used these data and reported parameter values of a = 0.232 and 3 = 0.247.
Haas et al. (1993) reported a beta-Poisson dose-response relationship for rotavirus infectivity with
parameter values of a = 0.2531 and 3 = 0.4265. Regli et al. (1991) published the widely accepted dose-
response parameter values for rotavirus of a = 0.26 and 3 = 0.42 (Table 2-5).
An issue unresolved based on the available studies is that the viral units used in the feeding studies
were reported as focus-forming units (FFU) rather than individual viral particles (WERF, 2011). FFU
assays determine virus titer by fluorescent staining of the viral antigen expressed in cells during
infection (Keiser et al., 2021). The most commonly used assumption—that plaque-forming units (PFUs)
of rotavirus are equivalent to the FFUs from the feeding study—could overestimate the risk associated
with rotavirus (WERF, 2011) because the FFUs reported in the feeding study might have contained
more than one PFU per FFU. PFUs can arise from more than one virus particle.
Campylobacter jejuni. Teunis et al. (2018) provided a multilevel model utilizing data from a set of
different volunteer challenge studies and outbreak data on C. jejuni. Challenge studies have not
included low doses because the primary objective of the studies was to achieve high illness rates to
assess interventions. Outbreak data are from situations where contaminated raw milk was consumed.
Neither situation is optimally reflective of the types of low-dose water matrix exposures that would be
relevant for recreational water exposures. Teunis indicates that it appears that in the C. jejuni
challenge studies, the doses required for infection and the doses required for causing symptoms
(illness) have different orders of magnitude, so it might be concluded that a high dose is needed to
cause illness. The dose-response parameters in Table 2-5 were used in conjunction with a conditional
probability of illness, as discussed below.
E. coli Q157:H7. The E. coli 0157:H7 dose-response model was derived using data from eight outbreaks
(Teunis et al., 2008b) and from an assumption that doses ingested in each of those outbreaks were
Poisson-gamma distributed. The exposure model was refined by adjusting the gamma distribution
parameter for exposure to reflect the dispersion associated with each outbreak. An exploration of
various models led Teunis et al. (2008b) to select a beta-Poisson dose-response model (infection
endpoint). Teunis developed and made available 10,000 pairs of dose-response parameters. The point
44
-------
estimate values are reported in the paper as best estimates; however, using the 10,000 a and 3 pairs is
recommended whenever possible.
Teunis et al. (2004) also published a dose-response model for E. coli 0157:H7 for a single outbreak that
occurred in Japan. A beta-Poisson model was also used in that study, and dose-response parameter
values were developed separately for children and adults. However, a single outbreak provides limited
information because the dose range is usually small, not several orders of magnitude as in
experimental studies.
The recommended dose-response model in this TSM from Teunis et al. (2008b) was selected because it
is considered more comprehensive as it covers a broader range of exposure conditions. The outbreaks
included in the model encompass various transmission routes, including hand-to-mouth
contamination, various foods, and water. Estimated exposure in these incidents covered a range of
doses, from a few bacteria to approximately 10,000, and the attack rates ranged from 0.5%-80%.
Salmonella enterica. Haas et al. (1999) estimated Salmonella infectivity using a beta-Poisson model and
pooled data from human feeding studies conducted by McCullough and Eiselle (1951a,b,c). The studies
investigated the infectivity of five Salmonella species (S. newport; S. derby; S. bareilly; S. anatum strains
I, II, and II; and S. maleagridis strains I, II and III). Haas et al. (1999) determined that the beta-Poisson
model best fit the pooled study data when four outliers (of 50 data points) were removed; one outlier
did not appear to represent monotonic behavior, and the remaining three data points lacked
consistency with adjacent doses in the same experiment. The pooled model was not statistically
different from models for the individual studies and had a = 0.3126 and 3 = 2,885 (Fazil, 1996). Haas et
al. (1999) validated the model using data on a waterborne salmonellosis outbreak in California from
Boring et al. (1971) and Riverside County Health Department et al. (1971).
Other researchers also have reported dose-response relationships for Salmonella, based on the same
feeding studies (Coleman and Marks, 2000; Coleman et al., 2004), using very different fundamental
assumptions. That work highlights the potential importance of strain variability, as the ID50S appear to
vary by several orders of magnitude. Bollaerts et al. (2009) report a generalized linear mixed model for
the dose-response relationship based on a series of foodborne outbreaks. These outbreaks
demonstrate that the vehicle of transmission can be important (i.e., different types of food) and that a
higher level of infectivity at substantially lower doses than those reported by Haas et al. (1999) is
possible. Although a substantial amount of uncertainty is associated with the potential importance of
strain variability in the dose-response relationship reported by Haas et al. (1999), it is widely used in
QMRA. Strain variability is recommended in this TSM because it is based on human feeding study data
and covers a wide range of potentially important environmental strains of Salmonella.
Cryptosporidium sp. The dose-response model for Cryptosporidium is based on the analysis conducted
forthe LongTerm 2 Enhanced Surface Water Treatment Rule (LT2ESWTR) (U.S. EPA, 2005). The
LT2ESWTR analysis includes feeding study data from six isolates of Cryptosporidium. Those isolates
included Iowa (DuPont et al., 1995; Okhuysen et al., 1999), TAMU (Okhuysen et al., 1999), UCP
(Okhuysen et al., 1999), Moredun (Okhuysen et al., 2002), TU502 (Chappell et al., 2006), and 16W. In
the experimental dose-response studies, human response to different isolates of Cryptosporidium
parvum varied widely (Okhuysen et al., 1999, 2002; Messner et al., 2001). With analyses based on
Messner et al. (2001), the LT2ESWTR Cryptosporidium dose-response model was developed using
Bayesian analyses of individual and combined data sets for different isolates and outbreak data. The
LT2ESWTR dose-response model is exponential with model parameter r = 0.09. Uncertainty within the
45
-------
dose-response model is evaluated by allowing the model parameter to vary uniformly across the range
of 0.04 to 0.16, consistent with the range reported in the LT2ESWTR (U.S. EPA, 2005). Several
researchers previously reported dose-response relationships for Cryptosporidium (Haas et al., 1996;
Okhuysen et al., 1999; Teunis et al., 2002a,b; Englehardt and Swartout, 2004, 2006; WHO, 2009). The
EPA's dose-response model for Cryptosporidium is used for this TSM because it accounts for all strains
of Cryptosporidium reported in the literature and is the most comprehensive analysis reported. More
recently, Messner and Berger et al. (2016) reported on the application of a fractional Poisson dose-
response model that functions as an upper bound to the exponential model and assumes that all
oocysts are capable of initiating infection (an assumption that may not describe risks from zoonotic
sources of Cryptosporidium).
Giardia lamblia. The Giardia lamblia dose-response model used in this TSM has been used in other
published QMRA (Regli et al., 1991; Rose et al., 1991; Rose and Gerba, 1991; Teunis et al., 1996; Haas
et al., 1999). In this dose-response model, Giardia infectivity is estimated using data from previously
conducted human feeding studies conducted by Rendtorff (1954a,b) and Rendtorff and Holt (1954a,b).
In the studies, volunteers were fed doses of one to 106 Giardia cysts, and infection was defined as cyst
excretion in the feces. Rose and Gerba (1991) estimated infectivity using an exponential dose-response
model, assuming a Poisson distribution of Giardia, and calculated r = 0.0199 (Table 2-5).
2.1.1.3.3 Morbidity
Morbidity is a term used to describe the proportion of infections that progress to a symptomatic
response (illness) (U.S. EPA and USDA, 2012). For each of the reference pathogens, morbidity is
expressed as a range to the extent that supporting data are available. Table 2-6 summarizes the
parameter values used to characterize the proportions of infections that result in illness. The EPA
presents point estimates and ranges for some of the reference pathogens to define a distribution for
input into the risk calculations. Narrative descriptions of the justification for the selected values follow.
Adenovirus. Teunis et al. (2016) provide point estimates to be used in the dose-dependent infection-
to-illness rate for adenovirus. The values provided in Table 2-6 are integral to the dose-response model
chosen in Table 2-5. Teunis et al. (2016) provided the n and r point estimates, so Bayesian modeling is
not required for running the model.
Norovirus. Teunis et al. (2020) highlighted previously reported results that the conditional probability
of illness in an infected individual may also depend on the dose to which that the individual was
exposed (Teunis et al., 1999). Thus, at high doses, the risk that an infection is symptomatic could be
higher than at low doses (Teunis et al., 2008a). Teunis et al. (2020) provide a concise description of the
posterior dose-response parameter distributions comprising a vector of mean values and a covariance
matrix defining a bivariate normal distribution for the illness dose-response parameters.
Rotavirus. The CDC (2012) reports that infection with rotavirus is nearly universal, with almost all
children infected by five years of age. Recovery from a first rotavirus infection usually does not lead to
permanent immunity. After a single natural infection, 38% of children are protected against any
subsequent rotavirus infection, 77% are protected against rotavirus diarrhea, and 87% are protected
against severe diarrhea. Reinfection can occur at any age. Subsequent infections confer progressively
greater protection and are often less severe than the first. Recurrent rotavirus infections affect people
of all ages. Recurrent infections are usually asymptomatic or result in mild diarrhea that can be
46
-------
Table 2-6. Literature-reported values for the proportion of infections resulting in illness from
exposure to reference pathogens.
Reference pathogen
Parameter point estimate
Parameter description
Reference
Rotavirus
0.4a
Midpoint of the range [0.1 to 0.7]
Haffejee et al., 1995
Ward et al., 1986a,b
Norovirus
n = 0.801
r = 3.19
Point estimate values for dose-
dependent relationship15
Teunis et al., 2020
w = (mean = 1.74,
variance = 5.55,
covariance = -0.708)
Intermediate parameters to derive a
distribution of n and r
z = (mean = 1.82,
variance = 4.64,
covariance = -0.708)
Adenovirus
H = 6.53
r = 0.41
Point estimate values for dose-
dependent relationship15
Teunis et al., 2016
Cryptosporidium spp.
0.5a
Midpoint of the range [0.2 to, 0.7]
U.S. EPA, 2006a
Giardia lamblia
0.45a
Midpoint of the range [0.2 to 0.7]
Eisenberg et al., 1996
Campylobacter jejuni
H =0.88
r = 0.06
Point estimate values for dose-
dependent relationship15
Teunis et al., 2018
w = (mean = -2.744,
variance = 1.337,
covariance = 0.010)
Intermediate parameters to derive a
distribution of n and r
z = (mean = -4.89 x 10"3,
variance = 0.993
covariance = 0.010)
E. coli 0157:H7
0.4a
Midpoint of the range [0.2 to 0.6]
Bielaszewska et al., 1997
Ludwig et al., 2002
Werber et al., 2008
Salmonella enterica
0.285a
Midpoint of the range [0.17 to 0.4]
Teunis et al., 1999
Notes:
a. Parameter values are probabilities, where 1 = 100%, so 0.6 means 60% of infections result in illness. For pathogens where a point
estimate is given, the EPA assumed that illness is independent of dose.
b. Illness is assumed to be dependent on dose; r| and r are point estimates used to model the dose-dependent illness.
preceded or accompanied by vomiting and low-grade fever. Based on these data, approximately 62%
are susceptible to subsequent rotavirus infection, and the overall probability of symptomatic response
to a nonneonatal rotavirus infection is no greater than approximately 23% (CDC, 2012).
In the feeding study reported by Ward et al. (1986a,b) on which the dose-response relationship is
based, 17 of 30 (57%) infected subjects became ill with doses equivalent to that required for infection.
Haffajee et al. (1995) noted that diarrhea has been reported to occur in ll%-70% of adults infected
with rotavirus. Based on these reported data, for the purposes of this TSM, the range of infections
leading to illness is estimated to be 0.1-0.7.
47
-------
Campylobacter jejuni. While infection induced by a pathogen may be asymptomatic, illness symptoms
can be considered conditional on infection; in the absence of infection, illness is assumed not to have
occurred. For Campylobacter jejuni, Teunis et al. (2005, 2018) contend that there appears to be
evidence indicating a dose-dependent illness response among infected subjects. Teunis et al. (2018)
calculated the (unconditional) probability of illness as the product of the probability of infection and
the conditional probability of illness given infection. The paper uses empirical data from previous
studies to derive a mathematical relationship that describes the conditional probability of illness given
infection as a function of ingested dose.
E. coli Q157:H7. The progression from infection to symptomatic illness for E. coli 0157:H7 is assumed
to be in the range of 0.2-0.6 based on outbreak data (Bielaszewska et al., 1997), the percentage of
symptomatic and asymptomatic individuals who were household contacts of hemolytic uremic
syndrome patients (Werber et al., 2008), and the occurrence of anti-Stx2 IgG (Ludwig et al., 2002). This
range is consistent with the proportion of illnesses reported in an analysis of an E. coli 0157:H7
outbreak (Teunis et al., 2004).
Salmonella enterica. The progression from infection to symptomatic illness for Salmonella enterica is
assumed to be in the range of 0.17-0.4. Teunis et al. (1999) report the response of human volunteers
from a human feeding study using Salmonella enterica serovar Meleagridis (McCullough and Eisele,
1951a). The Salmonella isolate was produced from market samples of high moisture, spray-dried,
whole egg powder, and administered in a glass of eggnog to healthy male prisoners (4 doses tested).
Three strains were originally tested, and Teunis et al. (1996) reported the results for strain II, the most
virulent. Fecal samples were taken daily for culturing to determine infections, and illness was defined
as having diarrhea. At the two lowest doses, 1.58 x 105 organisms and 1.5 x 106 organisms ingested,
one of six and five of six individuals were infected, respectively, but none became ill. At the third dose,
7.675 x 106 organisms ingested, all participants (six) were infected, and one became ill (17% of those
infected). At the highest dose, 1.0 x 107 organisms ingested, five of six individuals were infected, and
two became ill (40% of those infected).
Cryptosporidium sp. The progression from infection to symptomatic illness for Cryptosporidium is
based on research to develop the LT2ESWTR (U.S. EPA, 2005, 2006a). In that analysis, the available
literature and identified studies were analyzed. DuPont et al. (1995) found that 39% of those infected
had clinical cryptosporidiosis. Haas et al. (1996) provided information based on the same data also
suggesting a morbidity rate of 39% but computed 95% confidence limits of 19% and 62%. More
recently, a study found that after repeated exposure to C. parvum (IOWA strain), the morbidity rate
was the same as for the initial exposure in reinfected subjects (Okhuysen et al., 1998). Okhuysen et al.
(1998) also found that 58% of their subjects who received Cryptosporidium doses developed diarrhea,
an underestimate of morbidity because symptoms other than diarrhea contribute to the morbidity
rate. Based on these data, the progression from infection to symptomatic illness for Cryptosporidium is
assumed to be in the range of 0.2-0.7.
Giardia lamblia. Giardia infection is often asymptomatic, with asymptomatic cases representing as
much as 50%-75% of infected persons (Mintz et al., 1993). In a study at the Swiss Tropical Institute,
27% of 158 patients who had Giardia cysts in their feces exhibited symptoms (Degremont et al., 1981).
Based on these data, the progression from infection to symptomatic illness for Giardia is assumed to
be in the range of 0.2-0.7. This range is consistent with that reported by Eisenberg et al. (1996), who
48
-------
summarized the proportion of infections by Giardia lamblia that result in illness as reported in
Flanagan (1992), Birkhead and Vogt (1989), and Dupont and Sullivan (1986).
2.1.2 Surrogates: Fecal Indicator Bacteria
For over a century, culture-enumerated FIB have facilitated important gains in public health protection
against pathogens in drinking and recreational waters that are associated with disease. Due to the
potential diversity of pathogens associated with fecal contamination and their variable occurrence,
often at low densities, in environmental waters, routine pathogen monitoring can be costly, technically
challenging, and not feasible in some circumstances (Korajkic et al., 2018). Instead, public health
agencies monitor levels of surrogates—the FIB—in waters to measure the extent of fecal
contamination. Elevated levels of FIB have a low correlation with the presence (or absence) of specific
pathogens (Savichtcheva and Okabe, 2006; Wu et al., 2011). However, FIB are highly abundant in the
intestinal tract and feces of warm-blooded animals and humans. FIB co-occur with enteric pathogens in
the feces of infected animals and humans, so their use as metrics of fecal contamination in ambient
waters can indirectly protect the public from exposure to the waterborne pathogens associated with
fecal contamination. Their use as indicators of disinfection efficacy during wastewater treatment has
supported the reduction of bacterial pathogens, which can cause diseases like cholera and typhoid
fever, in effluent discharges. Additionally, epidemiological studies have established a statistically
significant association between increasing FIB abundance and gastrointestinal illness in some types of
waters (Priiss, 1998; Wade et al., 2003; Zmirou et al., 2003; Ahmed et al., 2018).
Currently, culture-enumerated E. coli and enterococci are commonly recommended by public health
agencies worldwide as the primary indicators of fecal contamination. Previous fecal indicator groups,
such as total coliforms and fecal coliforms, have been used and are still being used in some areas. The
EPA's 2012 RWQC recommended water quality values for culture-enumerated E. coli and enterococci,
associated with two specific target illness rates, to protect primary contact recreation in fresh and
marine waters (U.S. EPA, 2012).
The underlying epidemiological data supporting the EPA's RWQC were collected in waters receiving
human fecal contamination, such as poorly or partially treated sewage and treated effluent discharges
(Cabelli, 1983; Dufour, 1984; Wade et al., 2006, 2008, 2010). Elevated FIB levels in ambient waters
lacking direct point source impacts have not demonstrated a significant potential for adverse human
health effects linked to the abundance of FIB (Calderon et al., 1991; Colford et al., 2007, 2012). Eel
grass beds, drying kelp (wrack), green algae (Cladophora), other aquatic vegetation, biofilms, sands,
and sediments are examples of potential nonfecal sources of FIB (U.S. EPA, 2010b; Stewart et al., 2008;
Fujioka and Byappanahalli, 2003; Yan et al., 2011; Byappanahalli et al., 2012). Nonfecal sources of FIB
can potentially contribute substantial FIB loads to a waterbody (Byappanahalli et al., 2006, 2012; Ksoll
et al., 2007; Litton et al., 2010; Skinner et al., 2010). Data demonstrate that these bacterial sources are
not usually sources of fecal pathogens (Byappanahalli et al., 2006, 2012).
Although recreational water epidemiological studies conducted in waters where the predominant
impacts are from nonhuman or nonpoint sources have not provided clear, consistent associations
between FIB and adverse human health outcomes, a growing body of QMRA-based studies suggests
animal fecal sources can cause fewer diseases and pose less risk than human fecal contamination
(Schoen and Ashbolt, 2010; Soller et al., 2010b, 2015; WERF, 2011; McBride et al., 2013). Human
health risks from recreational exposures can be quite low when point sources, such as secondary
treated and disinfected WWTP effluent, do not impact a waterbody (Soller et al., 2016). Pathogens of
49
-------
public health concern from different fecal sources also can differ relative to the level of FIB measured
in environmental waters (U.S. EPA, 2009a,b, 2010b; Soller et al., 2015). The following sections provide
information on the abundance of E. coli and enterococci in animal feces and discuss the fate and
transport characteristics of these FIB in environmental waters.
2.1.2.1 Abundance of E. coli and Enterococci in Animal Feces
The FIB, enterococci and E. coli, are abundant in the feces of warm-blooded animals, although the
reported range for each FIB varies among different fecal sources (Table 2-7). Generally, the range of
FIB is from 105 to 1010 per gram (g) of feces in humans, pigs, chickens, and gulls. When considered
alongside the range of reference pathogens in the context of the health modeling, the level of FIB at a
specific target illness level can be different for different fecal sources. Similar to the pathogen ranges,
the ranges shown in Table 2-7 can be included in the QMRA modeling as a distribution between the
high and low values, with each value being equally likely (i.e., uniformly distributed).
Table 2-7. Literature-reported density of FIB in human and selected animal fecal waste.
Density3 of:
Fecal source
Enterococci
E. coli
minb
maxc
min
max
Raw sewage
Ranged
5.8
8.0
6.7
8.0
Human
Secondary treated and
chlorinated effluent
Ranged
0.5
2.7
0.5
4.0
Ranged
2.0
5.1
5.0
6.7
Cattle
Basis6
D
W
Typef
C
c
Ranged
5.0
7.0
5. is
10.9
Animal
Chickens
Basis6
W
w
Typef
c
c
Manureh
L
L
Gulls
Ranged
4.114
9.447
3.934
10.301
Pigs'
Ranged
5.3
7.2
6.1
7.3
Typef
D
C
Notes:
a. Represent power of 10 exponential value (e.g., n = 10").
b. Denotes minimum observed value.
c. Denotes maximum observed value.
d For raw sewage and secondary chlorinated effluent, units of minimum and maximum observations are CFU per L; for livestock wastes,
units are CFU per g.
e. Basis refers to the weight basis for manure. D denotes dry weight, and W denotes wet weight.
f. Sample type is either composite (C) or direct (D).
g. Reported as mean per state. Range presented is the range from the state with lowest mean density to the state with highest mean
density.
h. Chicken manure type is litter (L) or fresh (F).
i. All pig fecal abundances reported are for solid, fresh fecal samples (not slurries or treated manure).
50
-------
It is important to note that there are potential nonfecal sources of FIB, including autochthonous E. coli
and enterococci found in sand, soil and sediments, aquatic and terrestrial vegetation, and ambient
waters (Byappanahalli et al., 2006, 2012; Ferguson and Signoretto, 2011). Multiple studies have
demonstrated that FIB are widely distributed in nature, including in habitats with little or no input of
human or animal fecal contamination (Byappanahalli et al., 2012). Research has also demonstrated
these FIB can persist for extended periods in sands and other reservoirs (Byappanahalli et al., 2006).
Under certain circumstances, such as wet weather, the mobilization of FIB in nonhuman fecal sources
and environmental (nonfecal) FIB can occur. Nonfecal indicator loading can be a significant proportion
of the indicator enumerated. Enumeration methods routinely used to measure the FIB do not
distinguish between FIB from fecal and nonfecal sources or human and nonhuman fecal sources.
The values shown in Table 2-7 for raw sewage, secondary treated and chlorinated effluent, cattle,
chickens, and pigs are from Soller et al. (2010b). The values for gulls are from Goodwin et al. (2017).
2.1.2.2 Environmental Fate and Transport
Upon entering ambient waters, multiple environmental abiotic and biotic factors can affect the fate
and transport characteristics of FIB (Korajkic et al., 2018). The primary stressors affecting the
persistence of FIB in ambient waters include sunlight, rainfall, temperature, salinity, starvation
(available nutrient levels), predation and competition, and disinfection (if discharged via a point
source) (U.S. EPA, 2010b; Byappanahalli et al., 2012; Korajkic et al., 2018). These environmental
parameters may have different and potentially significant effects on the presence and persistence of
FIB depending upon the climate (tropical, subtropical, or temperate) (U.S. EPA, 2010b). FIB can
partition within the water column and to sediment. Once in the sediment, the FIB can be resuspended
and move further "downstream." Beach sand is recognized as a potential reservoir of FIB and
pathogens (Whitman et al., 2014). The source of the FIB can also affect the ability of FIB to persist in
the environment, with increased persistence noted in FIB that originates from ruminants (Korajkic et
al., 2018).
Sunlight, particularly the UV component, can increase the decay rates of FIB. FIB not associated with
particles, such as organic matter, decay faster than FIB associated with particles (Gutierrez-Cacciabue
et al., 2016; Boehm et al., 2019). Gutierrez-Cacciabue et al. (2016) also noted that the measurable DNA
of FIB persisted longer than the culturable cells. Culturable FIB are known to enter a VBNC state in
ambient environments, allowing increased persistence and potential regrowth when conducive
conditions allow (Ferguson and Signoretto, 2011; Li et al., 2014).
Eregno et al. (2018) characterized the decay rates of FIB in seawater samples from different depths.
Decay rates for FIB were low at low water temperatures (e.g., 4 degrees Celsius [°C]). At 20 °C, decay
rates for FIB were higher than adenovirus of coliphages. The authors state that accounting for different
decay rates at different temperatures, water qualities, and meteorological conditions in hydrodynamic
modeling can improve input parameters for QMRA (Eregno et al., 2018).
The EPA conducted a literature review to identify studies that characterized various fate and transport
factors affecting the persistence of FIB in ambient waters (U.S. EPA, 2010b). The review did identify
multiple studies that provide more discussion on these factors.
51
-------
2.2 Problem Formulation
This section describes and discusses the major factors considered for evaluating risk from exposure to
recreational waters contaminated by predominantly nonhuman fecal sources. Typically, the problem
formulation stage is considered the analytical phase of a risk assessment in which "the purpose for the
assessment is articulated, the problem is defined, and a plan for analyzing and characterizing risk is
determined" (U.S. EPA 1998b, 2014d). In this TSM, the factors considered (e.g., target illness level,
exposed population) and the process used for evaluating human health risks from pathogen exposure
in recreational waters are consistent with the EPA's 2012 RWQC. The EPA has included the selection of
risk assessment parameters (e.g., reference pathogens, exposure profiles) in the health modeling
presented here to facilitate a consistent approach for evaluating risk from nonhuman fecal sources.
Other risk management and technical considerations may be included if they are clearly identified and
transparently described in alternative WQC developed using this guidance.
2.2.1 Conceptual Model
A conceptual model was developed to provide useful information that characterizes and
communicates the potential health risks related to exposure to fecal contamination in recreational
waters. The model depicts the sources of the pathogens in these waters, the recreational routes of
exposure for potentially sensitive biological receptors of concern, and the potential adverse health
effect endpoints (e.g., gastrointestinal illness, respiratory illness) (Figure 2-5). Note that the conceptual
model is the same for different sources of fecal contamination. Recreational waters can receive fecal
loading from human and nonhuman sources.
2.2.1.1 Description of Conceptual Model (Figure)
The conceptual model is intended to explore potential relationships between exposure to a
contaminant or stressor and the adverse health endpoints important for management goals. In the
case of this TSM, the recommended 2012 RWQC can be adjusted to account for waters affected by
predominantly nonhuman fecal sources. Because of the linkage to the RWQC, aspects of the
conceptual model and the analysis discussed below are constrained by the decisions and assumptions
underlying the RWQC, such as:
1. Primary contact recreational scenario: The RWQC was designed to protect the primary contact
recreation designated use. Incidental ingestion while recreating is considered the predominant
exposure pathway for primary contact recreation.
2. General Population: The values recommended in the RWQC are informed by epidemiological
data collected from the general population (i.e., study participants of all ages combined). The
QMRA modeling discussed in Sections 2 and 3 includes exposure data for the general
population. Section 4 presents a sensitivity analysis comparing exposure for the general
population compared to children.
3. Gastrointestinal illness: The target illness rates discussed in the RWQC are based on the
gastrointestinal illness endpoint.
52
-------
STRESSORS
Pathogens associated
with feces of humans
and other warm-
blooded animals
SOURCES
EXPOSURE
ROUTES
Point source
discharges of fecal
contamination
T
Nonpoint and diffuse
sources of fecal
contamination
Direct deposition of
fecal contamination
Oral
E
Dermal
Incidental
ingestion while
recreating
Dermal
contact while
recreating
Incidental
inhalation while
recreating
RECEPTORS
General
population
(adults and
children)
_L
ENDPOINTS
Gastrointestinal
illness (e.g., NGI,
HCGI)
Respiratory
Effects
Eye
Ailments
Earache
Rashes;
Infected cut
Legend
Dark Gray: Primary contact
pathway and quantitative data
available to evaluate via QMRA
Light Gray: Pathways where
quantitative and qualitative data are
available, but were not used to
evaluate primary contact recreation
White: Data are insufficient
or incomplete for evaluating
primary contact rec.
Figure 2-5. Conceptual model of exposure pathways to enteric pathogens associated with fecal
contamination in surface waters while recreating.
53
-------
4. Boxes associated with the different components of the conceptual model shown in Figure 2-5
are shaded to signify the primary scenario discussed in this TSM to adjust the RWQC for
predominantly nonhuman fecal sources. Dark gray boxes denote the primary contact pathway
and include quantitative parameter data presented below to evaluate via QMRA. Lighter gray
boxes denote the inhalation exposure pathway where quantitative and qualitative data are
available but were not used to evaluate primary contact recreation. Inhalation exposure is less
than incidental ingestion (Backer et al., 2010; Butler et al., 2012). White denotes the dermal
exposure pathway and includes items for which data are insufficient or incomplete for
evaluating primary contact recreation.
2.2.1.2 Conceptual Model to Define Scope of Assessment
2.2.1.2.1 Factors Included in the Conceptual Model
Stressors
The stressors are pathogens associated with the feces of humans and animals that enter surface water
from various contamination pathways. Feces-associated, also referred to as enteric, pathogens of
public health concern belong to three major groups: viruses, bacteria, and protozoa. Eight reference
pathogens, including representatives from each pathogen group accounting for more than 97% of
nonfoodborne illnesses in the United States, are considered as part of the QMRA framework in this
TSM. Section 2.1.1 describes the reference pathogens and why each was selected.
Sources
Sources of fecal contamination can be characterized as point and nonpoint sources that contaminate
surface waters via direct and/or indirect pathways. Point sources of feces include WWTPs that treat
human sewage and subsequently discharge treated effluent directly to a river, lake, or coastal
waterbody. Large farms that raise livestock, such as cows, pigs, and chickens, known as CAFOs,10 also
can be considered point sources of fecal contamination (NOAA, 2022). Untreated animal waste from
CAFOs can enter nearby waterbodies as raw sewage. Nonpoint sources of feces include more diffuse
sources of fecal contamination, such as septage from septic systems, stormwater runoff from urban,
suburban, or agricultural areas, inputs from wildlife, including avian, and leaky infrastructure.
Generally, nonpoint fecal sources can be associated with wet weather mobilization leading to overland
runoff or subsurface flow of fecal pathogens to a waterbody. Some sources, such as shorebirds and
livestock with access to surface waters, can contribute feces via direct deposition. The potential
loading of fecal material from direct deposition from livestock operations can sometimes have a
relevant public health concern similar to the risks posed by human contamination (Soller et al., 2010b).
Routes of Exposure
Exposure to enteric pathogens in recreational water can occur via oral exposure (i.e., incidental
ingestion while recreating); dermal exposure (i.e., contact of exposed parts of the body with water
during recreational activities such as swimming, wading, or water skiing); and inhalation exposure to
aerosols (i.e., while recreating). Epidemiological data used to support the 2012 RWQC found that
(1) swimmers reported more illness than nonswimmers; (2) there were increased adjusted odds ratios
10 The EPA defines CAFO as any animal feeding operation with more than 1,000 animal units confined on-site for more than
45 days during the year. Any animal-feeding operation discharging manure or wastewater into a waterbody is considered a
CAFO, regardless of size (https://www.epa.gov/npdes/animal-feeding-operations-afos).
54
-------
with increasing levels of body exposure to water at the beach for recreators reporting body and head
immersion, a surrogate for incidental ingestion; and (3) swimmers who swallowed water reported the
highest rate of illness incidence. For consistency with the RWQC, incidental ingestion while recreating
is considered the primary route of exposure in this TSM. Incidental ingestion parameter values are
discussed below for use in QMRA.
Receptors
Humans who recreate in a water body where fecal contamination is present could be exposed to
enteric pathogens through ingestion, dermal contact, and inhalation of aerosols while recreating on or
in contaminated surface waters. The EPA relied on recreational water epidemiological studies that
characterized health effects for the general population to inform the 2012 RWQC. To be consistent
with the RWQC, this TSM includes QMRA parameters for the general population of recreators.
Endpoints
Published epidemiological and QMRA results and outbreak reports provide information about
associations between exposure to recreational waters containing fecal contamination and illness in
recreators. Common illness endpoints and their definitions include (Cabelli, 1983; Dufour, 1984; Kay et
al., 1994; Priiss, 1998; Wade et al., 2003, 2006, 2008, 2010; U.S. EPA, 2012; Fewtrell and Kay, 2015):
1. Gastrointestinal:11 any of the following within 10-12 days after swimming: (a) diarrhea (three
or more loose stools in a 24-hour period), (b) vomiting, (c) nausea and stomachache, or (d)
nausea or stomachache and impact on daily activity;
2. Respiratory: any two of the following: sore throat, cough, runny nose, cold, or fever;
3. Rashes: a rash or itchy skin;
4. Eye ailments: either an eye infection or a watery eye;
5. Earaches: ear pain, ear infection, or runny ears; and
6. Infected cut: a cut or wound that became infected.
Gastrointestinal illness is a commonly observed endpoint in health studies conducted at beaches
affected by fecal contamination. In epidemiology studies, significant increases for other illness
endpoints among swimmers are less commonly reported. The epidemiological data used to inform the
2012 RWQC identified gastrointestinal illness as the most sensitive endpoint among swimmers
participating in the studies. The fecal-oral route in the transmission of waterborne disease is an
important factor associated with exposure to recreational waters affected by fecal contamination (de
Graaf et al., 2017). Pathogens associated with feces can initiate colonization and infection in the
gastrointestinal tract of a new host (Clements et al., 2012; Kolling et al., 2012; Stecher, 2015). For
consistency with the RWQC, this TSM focuses on gastrointestinal illness as the primary adverse health
outcome among recreators.
11 Case definitions of gastrointestinal illness differ and have changed over time. Two common definitions in use include
HCGI and NGI. Fever is included as a required symptom for HCGI, but not for NGI.
55
-------
2.2.1.2.2 Factors not included
Recreating children can be at increased risk from exposure to fecal contamination because they spend
more time in contact with the water compared to adults, they incidentally ingest more water than
adults while recreating, and they can be more susceptible to infection and illness due to their
immature immune systems (Dufour er al., 2017; Wade et al., 2022). To be consistent with the 2012
RWQC recommendations tied to illness rates reported in the general population, which includes
children, this TSM describes the general population of receptors within the QMRA framework
presented. However, the EPA evaluates and discusses differences between lifestages in Section 4.1.2.1
of the Effects Characterization. Data for incidental ingestion by children are included in the sensitivity
analysis (Section 4.2.2).
2.2.1.2.2.1 Rationale for Not Considering Other Endpoints
Although the RWQC and this TSM focus on gastrointestinal illness as the primary health endpoint
related to incidental ingestion exposure to fecal contamination in recreational waters, other endpoints
may result after exposure and have been reported in some health studies. For example, in an
epidemiological study conducted by Sinigalliano et al. (2010) at a nonpoint source affected beach, an
increase in gastrointestinal, respiratory, and skin illnesses was reported by recreators, and a significant
association between exposure and skin illnesses and culturable enterococci was observed. Some
pathogens can initiate infection at more than one site in the body depending on exposure,
susceptibility, and the type of pathogen. Human adenovirus infection is known to cause respiratory
and eye infections in addition to gastrointestinal illness (Radke and Cook, 2018). Ear infection can occur
in the outer ear, otitis externa, or the inner ear (due to a wider sinus or upper respiratory infection).
These other illnesses generally occur at a lower rate than gastrointestinal illness (Fleisher et al., 1998;
McBride et al., 1998; Haile et al., 1999; Wade et al., 2008). For example, Wade et al. (2008) reported a
mean overall gastrointestinal illness incidence of 7.3%, upper respiratory infection incidence of 5.7%,
rash incidence of 2.7%, and eye irritations and infections of 2.9%. Kay et al. (1994) and Fleisher et al.
(1998) reported 14.8% gastrointestinal illness in swimmers and 9.7% in nonswimmers, 4.7% incidence
of respiratory infection in swimmers and 3% in nonswimmers, and 4.2% incidence of ear ailments in
swimmers and 4.8% in nonswimmers. Additionally, significant associations between these other
endpoints and FIB are less commonly reported in epidemiological studies.
While users of this TSM can consider other health endpoints in a QMRA, it is recommended that the
gastrointestinal endpoint be considered to facilitate consistent comparisons with the recommended
RWQC.
2.2.1.2.2.2 Other potential exposure routes
Exposure to enteric pathogens in contaminated recreational waters can occur via inhalation and
dermal contact. Inhalation can occur from exposures to wave-associated aerosols, personal watercraft,
and boat spray. Respiratory illness has been reported in epidemiology studies conducted at beaches
affected by human fecal contamination, although at lower rates than gastrointestinal illness (Kay et al.,
1994; Wade et al., 2008; Sinigalliano et al., 2010). Dermal exposure can occur through recreational
water contact. Although one of the skin's main functions is to provide a physical barrier to external
hazards, recreators with breaks in the skin can be at risk of infections at those sites (Oliver, 2005). A
caveat to this point is that many bacteria can be considered opportunistic pathogens, and wound
infection is possible from microorganisms that are not typically considered pathogenic. Additionally,
56
-------
some naturally occurring pathogens in aquatic environments, such as Vibrio vulnificus, are not
associated with fecal contamination. Vibrio wound infections can be serious (CDC, 2022).
Except for the published adenovirus dose-response function, which is characterized for respiratory
infections, the other dose responses included in this TSM are specific for gastrointestinal infections. For
QMRA discussed in this TSM, the dose response for adenovirus is assumed to be the same for both
respiratory and gastrointestinal infections.
2.2.2 Analysis Plan
The QMRA method for the human health risk evaluation discussed in this TSM is a static, stochastic
Monte Carlo simulation-based approach selected because it is relatively simple, straightforward, and
parsimonious compared to dynamic modeling (Soller et al., 2008). In this type of analysis, individual
recreational events are considered independent and treated probabilistically. Statistical distributions
rather than point estimates are used to characterize input parameters. In contrast to a deterministic
(point estimate-based) model, a stochastic QMRA model considers the variability associated with the
input parameters and yields a distribution of outcomes rather than a single-value outcome. The Monte
Carlo-based approach involves running the model multiple times; each run is called an iteration, using
randomly selected values within the range of each parameter distribution included in the model (U.S.
EPA, 1994, 2007d). A stochastic model will produce different results each time it is run because of
statistical chance. Over the thousands of iterations conducted, a distribution of probable outcomes is
generated. For example, one output of the QMRA framework in this TSM is that an estimate of the
probability of illness is generated that is associated with recreational exposure to the reference
pathogens. When the QMRA methodology is applied to derive the FIB level corresponding to a single
target illness rate, it represents a single iteration of the model in which all inputs must be specified as
static point estimates; the output represents a single FIB level corresponding to the target illness rate.
The recommended QMRA approaches in this TSM are consistent with the EPA's Microbiological Risk
Assessment Tools, Methods, and Approaches for Water Media (U.S. EPA, 2014b).
The EPA conducted extensive literature searches between 2007 to 2021 (Appendix A). The EPA
reviewed the published literature and compiled information on the reference pathogens, exposure
parameters, and health effects associated with primary contact in feces-contaminated recreational
waters (U.S. EPA 2009a,b, 2010a,b, 2013a). The parameter distributions and values specified in this
TSM are based on data from scientific literature. Parameters that are provided as standard values are
justified below.
2.2.2.1 Measures of Effect
This TSM is focused on gastrointestinal illness as the measure of effect because it is the most common
endpoint reported from recreational water epidemiological studies. The EPA's 2012 RWQC (U.S. EPA,
2012) provides the health effects basis for developing criteria expressed as culture-enumerated E. coli
and enterococci. The recreational water epidemiological studies informing the development of those
criteria were conducted in waters affected by human fecal contamination. Results demonstrated a
predictive positive association between gastrointestinal illness and culturable E. coli and enterococci
(Cabelli, 1983; Dufour, 1984). In these studies, gastrointestinal illness is expressed as NGI, which is
defined as: "any of the following [within ten to 12 days after swimming]: (a) diarrhea (three or more
loose stools in a 24-hour period), (b) vomiting, (c) nausea and stomachache, or (d) nausea or
stomachache and impact on daily activity" (Wade et al., 2006). To facilitate comparison with the
57
-------
RWQC, it is recommended that one of the two recommended target NGI illness rates (i.e., 32 or 36 NGI
per 1,000 recreators) be used in the QMRA framework discussed in this TSM.
2.2.2.2 Measures of Exposure
This TSM is focused on incidental ingestion because it is the primary source of exposure during primary
contact recreational activities. In HHRA, exposure is defined as human contact with a biological,
physical, or chemical agent—usually through ingestion, inhalation, or dermal contact. Risk assessment
can be performed for specific target populations or for an individual target organism, such as a human
with a defined exposure pattern. Exposure assessment involves determining or estimating
(qualitatively or quantitatively) the magnitude, frequency, duration, and route(s) of exposure (U.S. EPA,
2011b). A primary purpose of exposure estimation is to support dose estimation (U.S. EPA, 1998b).
Dose is the amount of a pathogen that enters or interacts with a host (ILSI, 2000). For nearly all QMRA
contexts, dose refers to potential dose (i.e., the number of pathogens ingested in a specified period)
because the actual number of pathogens that an individual is exposed to is usually unknown. The dose
is typically calculated as a function of the level of pathogens enumerated from a specific volume of the
exposure medium (e.g., the most probable number per liter of water), and the volume of that medium
that is ingested or inhaled. Ingestion of contaminated water is the main exposure pathway to enteric
pathogens, although inhalation and dermal contact can also be important in certain circumstances
(Hauchman, 2008). Empirical health data collected from recreational water epidemiological studies
report gastrointestinal illness symptoms more frequently than respiratory or dermal illness symptoms
(Wadeetal., 2008, 2010).
Because this TSM serves to support the EPA's 2012 RWQC, its scope is purposefully focused on
recreational exposures in ambient waters. The EPA selected incidental ingestion during primary
contact activities (such as swimming) for input to the QMRA framework because (1) data suggest that
incidental ingestion can be considered the highest potential exposure pathway for enteric pathogens
while recreating, (2) the health relationship underpinning the RWQC is based on this scenario, and
(3) the approach is consistent with the derivation of RWQC (U.S. EPA, 2019a). Dorevitch et al. (2011)
studied the volume of water ingested during a range of recreational activities in the Chicago Area
Waterway System (CAWS) and at a public outdoor swimming pool. Study participants took part in one
of the following activities on the CAWS: canoeing, fishing, kayaking, motor boating, or rowing. In the
swimming pool, participants took part in canoeing, fishing, kayaking, swimming, or wading/splashing.
The study results indicate that the odds of ingesting a teaspoon or more of water are significantly
higher among swimmers than among those who just immersed their head in a swimming pool or those
who participated in the other, more limited contact activities on surface waters. Therefore, the EPA
determined that using a swimmer scenario for exposure as the basis for the criteria is protective of
these other aquatic activities.
In this TSM, oral exposure is defined by incidental ingestion by the general population participating in
primary contact recreation where immersion and incidental ingestion are likely. In QMRA, incidental
ingestion is an event-based parameter (i.e., the volume of water ingested per recreational event).
Multiple recreational events over time are considered independent exposures. This exposure
parameter approach differs from a chemical risk assessment, which is based on the ingestion rate in
units of volume per day. For pathogen exposures, the exposure input value determines the dose of
pathogens ingested, which is then combined with the pathogen dose-response parameters to calculate
a probability of infection per event for each reference pathogen.
58
-------
The EPA identified and reviewed seven studies found in the scientific literature, reporting data on
incidental ingestion by recreators (U.S EPA, 2019a). Of the seven studies identified, two studies, Dufour
et al. (2017) and DeFlorio-Barker et al. (2017), reported quantitative data that may be used as input
parameters for QMRA. See U.S. EPA (2019a) for a more complete discussion of the strengths and
limitations of the seven studies identified in the literature.
Dufour et al. (2017) used cyanuric acid as an indicator of the amount of water ingested while
swimming in an outdoor pool. Cyanuric acid acts to stabilize hypochlorite levels in pool water exposed
to sunlight. Ingested cyanuric acid is not metabolized and excreted in urine. Pool water samples were
collected before the start of swimming activities, and urine from 549 participants, aged six to 81 years
and approximately evenly divided by gender and lifestage, was collected for 24 hours after the
swimming event ended; pool water and urine samples were analyzed for cyanuric acid. Study results
reported that males ingested more than females and younger children ingested more than older
children or adults. For the general population, the average amount of water ingested was 32 mL
(range: 0 mL-280 mL) (Dufour et al., 2017).
While Dufour et al. (2017) report quantitative data for water ingested while recreating, estimates were
gathered from participants recreating in a swimming pool. There is uncertainty associated with how
recreating in a swimming pool might compare with recreational behavior in ambient waters. Separate
ingestion estimates were not provided for participants with varying degrees of body exposure (e.g.,
waist-deep, submerged head). The distribution of ingestion reported by Dufour et al. (2017) did not
differ significantly from the distribution reported previously in a pilot-scale study by Dufour et al.
(2006).
DeFlorio-Barker et al. (2017) combined ingestion data from Dufour et al. (2017) and participant-
estimated time spent in the water data from 12 cohorts of previously conducted epidemiological
studies at fresh and marine beaches to provide an estimate of the volume of water ingested per
swimming event. The NEEAR Study and Southern California Coastal Water Research Project (SCCWRP)
epidemiological studies included 68,685 recreators at four freshwater and eight marine beaches. The
study participants estimated how much time they spent in the water. DeFlorio-Barker et al. (2017)
combined the volume of ingestion from Dufour et al. (2017) and the self-reported time spent in the
water by the epidemiological study participants to simulate an estimated distribution of the volume of
water ingested per event. The DeFlorio-Barker study reported similar trends to Dufour and other
ingestion studies, including that males ingested more than females and younger children ingested
more than older age groups. The similar trends (e.g., gender and age-related incidental ingestion)
reported by both studies may indicate that recreational behavior between swimming pools and
ambient waters may not be dramatically different. While the larger dataset provides more participants
by gender and age, the underlying ingestion data comes from the swimming pool study conducted by
Dufour et al. (2017). The estimated distribution assumes a linear relationship between the volume of
water ingested and the total time spent in the water (DeFlorio-Barker et al., 2017). However, it is
possible that some individuals may have more exposure in less time if the behavior includes head
immersion activities compared to other individuals who wade for a longer time in knee-depth or waist-
deep water. The epidemiological studies included self-reported data of swimming exposures and time
spent in the water that may be subject to recall bias (DeFlorio-Barker et al., 2017).
59
-------
A comparison of the mean and upper 90th percentile ingestion estimates for both Dufour et al. (2017)
and DeFlorio-Barker et al. (2017) is presented in Table 2-8. For this TSM, the ingestion distribution for
the general population is used as the input for exposure in QMRA because the 2012 RWQC focused on
the general population. The EPA chose the general population data from Dufour et al. (2017) for the
base analysis because it represents empirically measured ingestion data, whereas the ingestion results
from DeFlorio-Barker et al. (2017) are modeled from estimated time spent in the water for NEEAR
study participants combined with the Dufour ingestion data. See Section 4.1.2.1 in the Effects
Characterization for discussion on ingestion rates by lifestage.
2.2.2.3 Expression of Criteria (GM, STV and BAV)
The EPA's 2012 RWQC recommendations consist of a magnitude, duration, and frequency of exposure.
The magnitude of exposure is the numeric expression of the maximum amount of the contaminant that
may be present in a water body that supports the designated use. Duration is the period over which the
magnitude is calculated. The frequency of excursion describes the number of times the contaminant may
be present above the magnitude over the specified period (duration). A criterion is derived such that the
combination of magnitude, duration, and frequency protects the designated use (e.g., primary contact
recreation). Alternative criteria should include the same basic elements as the national RWQC.
In the 2012 RWQC, the EPA recommends expressing the criteria magnitude as a GM12 value
corresponding to the 50th percentile of the water quality distribution and an STV corresponding to the
90th percentile of the same water quality distribution. For duration and frequency, the EPA
recommends that the waterbody GM not be greater than the selected GM magnitude in any 30-day
interval. There should not be greater than a 10% excursion frequency of the selected STV magnitude in
the same 30-day interval. Calculating the STV requires having the logarithm (log) standard deviation
(logSD)13 of the data for the indicator in the waterbody.
Table 2-8. Comparison of estimated or measured ingestion volumes per recreational event for
Dufour et al. (2017) and DeFlorio-Barker et al. (2017).
Study
Age group
Number of
participants
Median (mL)
Mean (mL)
Upper 90th
percentile (mL)
Dufour et al.
(2017)
All ages
549
19
32
(In mean -3.98;
In SD 1.43)a
77
DeFlorio-Barker
et al. (2017)
All ages
68,685b
16
44
105
Notes: In = natural logarithm
a. For the forward QMRA, the distribution for ingestion is incorporated into the calculations (Section 2.2.2.4.3). For the reverse QMRA,
the point estimate is used (Section 2.2.2.4.5).
b. This value represents the number of participants in epidemiological studies conducted across 12 beach locations which did not
measure or estimate ingestion for each participant. DeFlorio-Barker modeled the distribution of water ingested by the 549 participants
from the Dufour et al. (2017) study for the participants in the epidemiological studies.
12 The GM can be defined as the average of the log values converted back to base 10. For microbial monitoring data,
transform each value to the base 10 log, find the arithmetic mean of the logio values, and convert the result back to
base 10.
13 The logSD is the SD of the base 10 logs of the data. The antilog of the logSD would be the geometric SD.
60
-------
2.2.2.4 Approach for adjusting the EPA's 2012 RWQC recommended FIB values for nonhuman
sources.
As shown in Figure 1-1, four main steps are involved in the approach for adjusting the EPA's
recommended 2012 RWQC FIB values to account for predominantly nonhuman fecal sources affecting
a waterbody (see Section 3). At each step, there are risk management decision points to address,
including whether to proceed with the TSM or adopt the nationally recommended criteria. Figure 2-6
displays a flow diagram of the overall process discussed in this TSM, including the risk management
questions, major decision points, data collection activities, and QMRA analyses.
Step 1 (Section 3.1) is to conduct a sanitary survey to gain an understanding of the sources of fecal
contamination in the watershed.
Step 2 (Section 3.2) is to conduct a water quality study to better understand both the initial
assumptions and the results of the sanitary survey. Collectively, Steps 1 and 2 comprise a sanitary
characterization. The sanitary characterization is critical for identifying and describing the fecal sources
affecting the waterbody and for deriving site-specific criteria and will inform multiple science and
policy decisions that are made to identify the predominant fecal source, conduct the QMRA, interpret
the results, and derive alternative WQC.
Importantly, Steps 1 and 2 will also provide the information needed to decide if the human proportion
of the fecal contamination in the watershed is small enough to warrant continuing to evaluate site-
specific criteria for the watershed. For example, based on the data collected in Steps 1 and 2, it might
be decided not to proceed with this process at this location and find another more representative site
or instead adopt the nationally recommended values. It might be prudent to consider whether other
factors, such as public perception or the ability to ensure protection for downstream waters, would
influence the desirability or acceptability of site-specific criteria.
Step 3 (Section 3.3) is to conduct an HHRA based on the information/data collated and collected to
determine human health risks due to the fecal sources identified in Steps 1 and 2. This step describes
two QMRA-based approaches for using literature-derived and site-collected data to calculate pathogen
and FIB densities in water (Sections 3.3.1 and 3.3.2). In all cases, the QMRA documentation should be
clear and transparent. Importantly, at the end of this step, a determination is made of whether
alternative criteria (including a fecal source and water quality evaluation and the target illness rate)
can be developed or whether the recreational use for the study location is protected by the nationally
recommended water quality and associated illness rates.
Step 4 (Section 3.4) is to calculate a GM, STV and BAV reflecting the predominantly nonhuman fecal
source loading and corresponding to the chosen target illness rate. Once the sanitary characterization
(Steps 1 and 2 above) and the HHRA (Step 3 above) are completed, the QMRA output can be used to
determine what level of indicator (i.e., criteria magnitude) is protective of the primary contact
recreational use in the waterbody.
61
-------
Non-human Fecal Sources TSM
(RWQC Section 6.2.2)
sthepredon nantfeca
ontan riat on source non-
human?
Inform at on that suggests
that sources are
predcm nantVnon-human
Section 3.1 Step 1: Sanitary
Suiw:
=3
Understand ng Sources
a
Impacting 'A'atershed
9
as
Sect on3.2 StepZ'Water
?
Qua tv Study: *eca nd cator
•?.
and pathogen mon tor ng,
St
source trc ck ng
a
f Sect on3 ? Step
J Forward QMRfi
! Hunanheathr sts
* era uat on
u dethe i 'aterqua
and nes est nares
compareta the2?12
RWQC c TWOS'3
Sect on3.- Step". S te-
Spec ^cS'ternat veCr tet a
der vat on. Revei seCMRA
and GM, S~». and BAV
ca'cu a on correspond ng
to target inessrate
Legend
Boxes-activities
Diamonds - questions and decisions
Ctrcies-Criteria
Rounded boxes-information
Figure 2-6. Overview diagram of the process discussed in this TSM for developing alternative criteria
using QMRAfor waters affected by predominantly nonhuman fecal sources.
62
-------
The scientific literature describes two general types of QMRA modeling to evaluate potential human
health risks that differ in the use of the input parameters. Deterministic modeling uses point estimates
for each parameter. Stochastic modeling incorporates the distribution of values associated with each
parameter. This TSM includes both types of QMRA modeling to address the risk management
questions. For estimating the potential for gastrointestinal illness from primary contact recreational
exposure, the suggested QMRA method for the human health risk evaluation is a stochastic Monte
Carlo simulation-based model because it provides additional information on the potential ranges of the
modeled outcomes. In this type of analysis, recreational events are treated probabilistically, and
statistical distributions are used to characterize input parameters rather than point estimates. In
contrast to a deterministic (point estimate-based) model, a stochastic QMRA model can include many
thousands of iterative runs to yield a distribution of outcomes rather than a single-valued outcome. A
deterministic QMRA approach is used in Step 4 to estimate the density of FIB corresponding to the
target illness rate chosen.
Figure 2-7 displays the overall analytical plan for the QMRA conducted as part of this TSM. Pathogen
densities in recreational waters can be considered for two main scenarios: (1) direct deposition and
(2) indirect deposition. Direct deposition represents fresh feces from the fecal source affecting the
waterbody with no environmental or treatment attenuation of pathogens considered before inputting
the pathogen data into the health model. Indirect deposition represents a scenario where fecal sources
may be land applied, as in the case of agronomic applications of manures, or from land-deposited fecal
pats. The indirect deposition scenario includes consideration of physical and environmental fate and
transport processes (e.g., runoff, sunlight, fecal age) that can result in microbial attenuation. The
indirect scenario might be considered if pathogens are measured in land-applied feces and an estimate
of fecal loading to a waterbody is evaluated. Measuring pathogens in feces can be helpful if levels in
water are low and variable in occurrence (e.g., nonpoint sources dependent on wet weather-
associated influx). See Soller et al. (2015) for an example of an indirect deposition scenario providing
input into a QMRA. As discussed above in Sections 2.1.1.2 and 2.1.2.2, multiple potential fate and
transport factors could be considered; generating data for all of them to provide input into this
approach can be involved and resource-intensive. As discussed in Soller et al. (2015) the fate and
transport construct can be simplified to one or a couple of important factors, such as stormwater
runoff, coupled with health-protective assumptions, to provide informative illness estimates.
Whichever deposition scenario is chosen, the user is encouraged to transparently document choices
and provide substantiating data.
In this TSM, the health model can be run using three different approaches to address the different
questions being addressed at each step. Discussion of the three main QMRA approaches can be found
in Sections 2.2.2.4.3, 2.2.2.4.4, and 2.2.2.4.5. For example, if the starting point is the pathogen
occurrence in recreational waters and the desired output is the total probability of illness, refer to
Section 2.2.2.4.3 for the correct QMRA approach.
2.2.2.4.1 Summary of the EPA's Illustrative Scenario
The EPA has customized the QMRA framework in this TSM to specifically evaluate a recreational
exposure scenario. The parameters needed for conducting QMRA, including exposure scenarios, are
presented in this document to help simplify the overall approach and encourage consistency in the
process and evaluation at the state level of potential alternative WQC reflecting nonhuman fecal
source inputs. Therefore, some of the QMRA parameters are based on standard assumptions and
63
-------
Pathogen densities in
recreational waters*
Indirect
Fate and Transport
e.g., Land application,
runoff*
Sediment and
Resuspension
Partition/die-off*
Direct
Deposition
Volume of water ingested*
Pathogen dose-response
functions and model
parameters*
Dose of pathogen ingested
Probability
of infection
Probability of infection
leading to illness*
Indicator densities in
recreational waters
(site-specific data)*
* indicates inputs;
boxes within the darker
gray boxes are
computations;
arrows denote the
conceptual flow of the
QMRA components
Figure 2-7. Conceptual diagram of the overall QMRA analytical approach described in this TSM.
others on site-specific information and data from Step 2 (Section 3.2). The EPA developed this
approach and subjected its various components to independent scientific peer review to bolster the
risk estimates' scientific defensibility and to encourage states to consider these tools within their
respective WQS programs.
The EPA developed a specific exposure scenario (the "primary contact" scenario) for QMRAs
addressing human exposures to recreational waters. The EPA purposefully constructed this exposure
scenario to simplify the health modeling. For evaluating alternative criteria for waters contaminated by
nonhuman fecal sources, the primary contact exposure scenario is as follows:
• The waterbody is predominantly impacted by a nonhuman source of contamination.
• The nonhuman source directly contributes14 FIB or pathogens, or both, into the waterbody of
interest.
14 This is a health-protective assumption. Although the watershed may not have direct deposition, this assumption can still
be chosen.
64
-------
• FIB and reference pathogens occur in the fecal material of the source or in the nonfecal source
at levels consistent with those reported in the peer-reviewed literature.
• Primary contact recreation occurs near and within a short period of time to the contamination
(i.e., no die-off of either FIB or reference pathogens occurs).
• Primary contact recreation results in ingestion of water consistent with studies reported in the
peer-reviewed literature (Dufour et al., 2017).
The EPA has previously published information to justify using the primary contact exposure scenario
and this general approach (U.S. EPA, 2010a; Schoen and Ashbolt, 2010; Soller et al., 2010b, 2015). This
exposure scenario is considered health-protective because it uses a hypothetical situation involving the
direct deposition of fresh fecal material from a source to a waterbody in close proximity to recreators.
Using these scenario assumptions, QMRA results indicate the mean probability of gastrointestinal
illness associated with recreational exposure to waters impacted by fresh gull, chicken, or pig feces is
substantially lower than in waters impacted by human sources (Soller et al., 2010b). The potential for
gastrointestinal illness associated with recreational exposure to waters receiving direct deposition of
fresh cattle feces is not substantially different from waters impacted by human sources. In waters
affected by human fecal sources, risks were primarily associated with human enteric viruses, while in
waters receiving bovine inputs, the pathogen profile associated with risk included a combination of
zoonotic bacteria and protozoan pathogens (U.S EPA, 2010a; Soller et al., 2010a). This exposure
scenario is also health-protective because pathogens are not attenuated via fate and transport or
mobilization processes, nor does loading from additional sources occur that could increase indicator
densities but without a commensurate increase in pathogens.
Refinement of the scenario to include land application of fecal material with mobilization of FIB and
pathogens via undiluted runoff (introduction of fate and transport parameters), referred to as the
"indirect deposition" scenario in Figure 2-7 above, can result in a reduction of estimates of potential of
illness for all nonhuman sources, including cattle sources (U.S. EPA 2010a; Soller et al., 2015). These
analyses demonstrate that the nature and magnitude of fecal contamination help to define the
potential for human health impacts. Given the variability and uncertainty associated with the input
parameters, the EPA decided to use this relatively straightforward and specific exposure scenario in its
approach and highlight where additional fate and transport information could be included. This
parsimonious approach can be extended to include more complicated or sophisticated fate and
transport modeling, if desired, for specific waterbody conditions.
2.2.2.4.2 Sanitary Characterization
A central input into the QMRA-based framework discussed in this TSM is the need to identify and
characterize the source(s) of fecal contamination affecting the waterbodies of interest. Different fecal
sources (e.g., human versus nonhuman) can contribute different enteric pathogen profiles to fecal
contamination, which can result in different levels of potential risks for recreators exposed to waters
affected by different sources of contamination. The risk differential can manifest as different FIB values
associated with the two target illness rates discussed in the RWQC. Nonfecal sources of E. coli and
enterococci, such as sands and soils, algal mats and plant surfaces, and biofilms, also exist. Although an
increase in illness has been demonstrated to be associated with sand exposure (Heaney et al., 2009),
FIB enumerated from the beach sand have not been demonstrated to be predictive of the potential for
human health effects (Halliday and Gast, 2011; Viau et al., 2011; Shibata and Solo-Gabriele, 2012).
65
-------
Characterizing fecal source inputs will help to delineate the potential risks from nonhuman fecal
sources relative to our understanding of the risks associated with exposure to human fecal
contamination.
A key decision point in the overall framework discussed in this TSM is evaluating multiple lines of
evidence to demonstrate that predominantly nonhuman sources of FIB influence the waterbody. Due
to the importance of fecal source identification in determining fecal loading to a waterbody, the EPA
has developed a "sanitary characterization" approach consisting of a two-step process for collating
existing information on the potential fecal sources impacting the selected waterbody, and then
collecting water quality data to substantiate the existence of nonhuman fecal sources and the absence
of substantial human fecal inputs, such as wastewater effluent. Collecting data under different
conditions (e.g., dry and wet weather periods) to understand contamination dynamics may be needed
to fully characterize potential conditions and risks. The epidemiological data used to inform the 2012
RWQC recommendations were conducted in dry weather conditions. Wet weather scenarios can result
in different indicator and health relationships (Arnold et al., 2017; Soller et al., 2017). It is critical to this
QMRA framework that due diligence is used to identify and characterize the fecal source(s) affecting
the waterbodies used to support adjusting the 2012 RWQC to account for nonhuman fecal sources.
Consider all evidence the sanitary survey and the water quality study provide. If a waterbody is in a
more urbanized watershed, more evidence to demonstrate nonhuman fecal sources predominate
might be needed because such waterbodies can be more susceptible to human fecal contamination
and other anthropogenic inputs. Identifying and delineating human fecal contamination from nonfecal
anthropogenic and nonanthropogenic inputs (e.g., FIB associated with algal mats and biofilm
contributions of FIB) in urbanized watersheds can be challenging (Byappanahalli et al., 2003; Skinner et
al., 2010; Piggot et al., 2012). Alternatively, less evidence might be needed for a remote location
lacking significant surrounding urbanization, point sources, or both, to demonstrate adequately that
human fecal contamination is low. For remote and pristine waterbodies, the initial sanitary survey
might be sufficient evidence to demonstrate that nonhuman fecal sources are predominant. Decision-
makers should also consider the "representativeness" of the waterbody studied if there is interest in
generalizing these findings to other waterbodies with nonhuman fecal inputs.
2.2.2.4.2.1 Sanitary Survey
The first step in the sanitary characterization is a sanitary survey. A sanitary survey is a method of
investigating the sources of fecal contamination to a waterbody (U.S. EPA, 2022). Although the
literature includes many examples of sanitary surveys, including approaches used for drinking water
source waters, shellfish, and watershed protection, the EPA has developed a sanitary survey form
specific for use in the QMRA framework discussed in this TSM (see Appendix B for a QMRA sanitary
survey form). The form found in Appendix B was based on the routine and annual beach sanitary
surveys the EPA published previously. Additionally, the EPA has made available sanitary survey tools
for fresh and marine recreational waters, including an application for Apple and Android operating
systems (U.S. EPA, 2022). The information identified and collated as a part of this survey activity will
help decide whether to proceed with the next step in the sanitary characterization or halt the process.
This decision can involve both a science and a policy component, which can be transparently discussed
as part of the documentation.
66
-------
2.2.2.4.2.2 Water Quality Study
The second step of the sanitary characterization is to collect water quality information from the
selected waterbody and watershed to substantiate the sanitary survey conclusions. The types of
information included in a water quality study (see Section 3, Step 2) and the level of evidence needed
could depend on local conditions. A robust understanding of the occurrence and prevalence of FIB and
the types and loading of pathogens associated with the fecal sources in the watershed should be
generated in the water quality study. Monitoring approaches should reflect conditions that result in
fecal contamination. For example, if animal fecal contamination could occur from runoff of manures
applied to fields, then monitoring following wet weather events is justified. Pre-existing monitoring
data can be considered in this context as well.
Appendix C presents an example of a SAP for conducting a water quality study. This example includes a
range of information that can be collected, and not all aspects need to be presented in the water
quality study. A water quality study should be representative of the meteorological and hydrological
conditions in the selected watershed. The two main purposes of this additional information are to
develop scientifically defensible evidence that predominantly nonhuman fecal sources impact the
waterbody and to supply the information necessary to estimate potential human health effects in the
QMRA.
As part of the water quality study, it is recommended that microbial source identification and tracking
(MST) techniques be incorporated to help confirm that the fecal sources that were identified in the
sanitary survey are affecting the waterbody. The application of MST can help identify sources that the
survey may have missed. Use of the microbial genetic source marker methods can also be helpful in
understanding the dynamics of fecal contamination (i.e., sources affecting a waterbody under baseflow
or wet weather conditions) and BMPs aimed at reducing contamination (Bushon et al., 2017; Li et al.,
2019, 2021; Nevers et al., 2022; Shrestha et al., 2020). Characterized nonhuman microbial genetic
source markers include bovine/ruminant (Shanks et al., 2008; Mieszkin et al., 2010; Raith et al., 2013;
Xue and Feng, 2019), canine (Ervin et al., 2014; Green et al., 2014), avian/gulls/waterfowl (Green et al.,
2012; McMinn et al., 2019), and swine (Mieszkin et al., 2009; Fan et al., 2017). Forty-one source
markers have been evaluated for selectivity and specificity in a multi-laboratory trial (Boehm et al.,
2013). The literature also includes several examples of the derivation of RBTs of selected MST markers
(Boehm et al., 2015, 2018; Brown et al., 2017a; Ahmed et al., 2018; Schoen et al., 2020; Goh et al.,
2021). Multiple human fecal source markers are available that can help confirm the presence of human
fecal sources, such as sewage and effluent, and help evaluate if human fecal contamination
predominates in the waterbody (Cao et al., 2018; Stachler et al., 2018; Schoen et al., 2020). The EPA
has published draft Methods 1696 and 1697 for the characterization of human fecal contamination in
water by HF183 and HumM2, respectively, via qPCR (U.S. EPA, 2019b,c). Recently, the EPA partnered
with the National Institutes of Standards and Technology (NIST) to develop a high-quality standard
control material (NIST SRM 2917) for use in MST qPCR methods (Willis et al., 2022). Choosing a well-
characterized source marker can provide robust information on fecal sources in a water quality study
and clear substantiation for decision points discussed in this TSM.
67
-------
2.2.2.4.3 "Forward" QMRA modeling
A forward QMRA is an application of the health model using pathogen occurrence data as an input to
calculate an estimate of the probability of illness. Forward QMRA includes the water quality
information about the fecal source(s) and any monitoring data collected in the sanitary
characterization. The resulting output of the analysis will provide a quantitative estimate of human
health risks from recreational exposure to the waterbody of interest and can be used to compare with
a target illness rate. The suggested QMRA method for the human health risk evaluation is a stochastic
Monte Carlo simulation-based model. In this type of analysis, recreational events are treated
probabilistically, and statistical distributions are used to characterize input parameters rather than
point estimates. Figure 2-8 displays the overall process for a forward QMRA.
Indirect
Pathogen densities in
sources or recreational
waters*
Fate and Transport
e.g., Land application,
runoff*
Sediment and
Resuspension
Partition/die-off*
Direct
Deposition
Volume of water ingested*
Pathogen dose-response
functions and model
parameters*
Probability of infection
leading to illness*
* indicates inputs;
boxes within the darker
gray boxes are
computations;
arrows denote the
conceptual flow of the
QMRA components
Dose of pathogen ingested
Probability
of infection
Illnesses attributable to
individual pathogens
Total Gl illness rate per
1000 recreation events
Figure 2-8. Analysis plan for a forward QMRA.
68
-------
Forward QMRAs conducted in Step 3 (Section 3.3) bring together exposure and human health data in a
logical way to estimate a predicted level of illness. As Figure 2-8 shows, reference pathogen densities
are inputs to the human health risk evaluation. This TSM discusses two approaches to estimate the
reference pathogen densities in Section 3.3.
• Approach 1: Estimate the waterbody pathogen density for each reference pathogen based on
the reference pathogen data in the waterbody gathered during Step 2.
• Approach 2: Estimate the waterbody pathogen density for each reference pathogen from the
FIB density in the waterbody, FIB density in the source, and reference pathogen density in the
source following the methods of Schoen and Ashbolt (2010) and Soller et al. (2010b).
Both Approaches 1 and 2 provide valuable information for understanding the potential human health
risks associated with recreational activities in the waterbody. Users of this TSM will need to do both
approaches because they provide complementary perspectives. For example, the utility of Approach 1
is limited in cases where the enumeration of low levels of pathogens in ambient waters is limited by
assay limits of detection (and quantification). A nondetection or nonquantified value in monitoring can
mean that the pathogen is present at levels lower than the LOD or quantification of the enumeration
method, not that the pathogen is absent. See Appendix C for a discussion of a sampling and analysis
approach used at Boqueron Beach utilizing both quantitative enumeration methods and PCR-based
presence-absence tests for specific pathogens. PCR-based methods can be more sensitive than culture-
based methods, particularly at low levels of the target organism. However, they also may detect
pathogens that are noninfectious or nonviable. Because pathogen occurrence in recreational waters
varies temporally and spatially, monitoring results might not necessarily be representative of pathogen
levels in the waterbody. Therefore, any subsequent risk analysis relying solely on these results (from
Approach 1) might be misleading. In contrast, the direct deposition of fecal material (as in Approach 2),
which does not account for pathogen or FIB decay, is expected to provide QMRA results at the higher
end of the potential risk estimate. Taken together, QMRA results based on data from both
Approaches 1 and 2 provide valuable insight for understanding the potential human health risks in the
watershed.
Forward QMRA Approach 1: Use of data collected from water-quality monitoring
Approach 1 estimates the probability of gastrointestinal illness from primary contact recreational
activities based on reference pathogen data sampled and measured in the waterbody during Step 2.
Calculating the probability of illness by the method of forward QMRA involves computing the dose of
each reference pathogen ingested during a recreational event, the probability of illnesses from each
reference pathogen, and the combined total probability of illness from all reference pathogens using
Equations 1A-1D.
[irv=V X C™ [Eq. 1A]
where:
Hrp is the dose ingested of reference pathogen rp (number of pathogens).
V is the volume of water ingested during the primary recreation activity (L; baseline
assumption = 0.019 L).
69
-------
C™p is the measured density of reference pathogen rp in the waterbody (from site-specific data
in units of pathogens/L).
C/ = fa-r (M [Eq. IB]
where:
is the probability of infection from reference pathogen rp (unitless).
(Mrp) is the mathematical dose-response function for reference pathogen rp (presented in
Table 2-5) which estimates the probability of infection at a given dose (unitless).
Pm = Pamnf X PSr [Eq-1C]
where:
Pfrf is the probability of illness from reference pathogen rp (unitless).
Pm\inf's proportion of individuals infected with reference pathogen rp who experience
illness (unitless; presented in Table 2-6).
Pill = 1 - rirp(l - Pm ) [Eq-1D]
where:
Pm is the probability of illness from primary contact recreational activities in the waterbody
accounting for the effects of all reference pathogens (unitless; assumes the independence of
each reference pathogen).
Implementing equations 1A-1D using a single measurement of pathogen density will result in a point
estimate of the probability of illness. Obtaining a statistical distribution of the probability of illness
requires multiple measurements of pathogen density in the waterbody. These measurements can be
used to conduct a Monte Carlo simulation, which produces a statistical distribution of the probability
of illness. One method of using the pathogen density measurements in a Monte Carlo simulation is to
randomly draw a single value of pathogen density from the range of measured values and estimate the
probability of illness using this value and Equations 1A-1D above. Repeating this process thousands of
times while replacing the drawn measurement value during each repetition generates a distribution of
the probability of illness. An alternative method of using the measured pathogen densities in a Monte
Carlo simulation is to first fit a statistical distribution to the measured density values using standard
statistical techniques, randomly draw thousands of times from the fitted distribution of densities and
process the randomly drawn density value each time through Equations 1A-1D. This procedure also
will result in a statistical distribution for the probability of illness. In addition to randomly drawing
pathogen densities, the Monte Carlo simulation can also randomly draw from statistical distributions
for the volume of water ingested and the various dose-response parameters.
This approach is still valid if reported pathogen densities are below the LOQ—or LOD—or if the dataset
contains many nondetects. It is not unexpected for reported reference pathogen densities to be
consistently below detectable limits, partly due to the variability in prevalence in the hosts and
occurrence in environmental matrices. However, there are approaches that can be used to address
nondetects. One approach is to assume that the average organism density is at the respective
70
-------
detection limit of the enumeration method being used (Soller et al., 2016). Because the actual
pathogen density is not greater than the detection limit in this case, this approach can provide a
reasonable upper-bound estimate of pathogen density. Other approaches for addressing nondetects
and left-censored data can be found in the literature (Helsel, 2012; Schmidt et al., 2013; Canales et al.,
2018).
Appendix E contains annotated computer code in R that can be used to implement the suggested
stochastic QMRA approach.15
Forward QMRA Approach 2: Using a combination of collected and literature-based data
In Approach 2, the waterbody reference pathogen densities are estimated based on the measured
FIB16 density in the waterbody, FIB density in the source, and reference pathogen density in the source
following the methods of Schoen and Ashbolt (2010) and Soller et al. (2010b). This process involves
sequentially estimating:
1. The density of each reference pathogen in the waterbody deriving from each source.
2. The ingested dose of each reference pathogen attributable to each source.
3. The probability of illness from each reference pathogen attributable to each source.
4. The combined probability of illness counting the contribution of all reference pathogens for
each source.
5. The total probability of illness combining the contribution of all pathogens from all sources in
the waterbody.
Equations 2A-2F summarize these steps.
RS
Crp = Pfib x Cfib x [Eq. 2A]
kfib
where:
Cyp is the estimated density of the reference pathogen (number of pathogens or genomes/L) in
the waterbody deriving from Source S.
PpIBis the proportion of FIB in the waterbody attributable to Source S (unitless; from the site-
specific information).
CFIB is the waterbody density of FIB measured using a culture-based method (CFU/L) (from the
site-specific data). Note to users: For example, enterococci measurements in water are
frequently reported in units of CFU/100 mL. However, Equation 2A requires that the
measurements be converted into units of CFU/L.
RpIB is the density of FIB using a culture-based method (CFU/L or CFU/g) in Source S (Table 2-7
or the source).
15 R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna,
Austria. URL http://www.R-proiect.org/.
16 In the examples in this TSM, the EPA used enterococci. E. coli can also be used.
71
-------
RrP is the density of the reference pathogen (number of pathogens or genomes/L or number of
pathogens or genomes/g) in Source S (Table 2-2 or the source).
[l£p = X p^p X Ijip XV [Eq. 2B]
where:
is the estimated dose of human infectious reference pathogens (number of pathogens or
genomes) attributable to the source.
PrP is the fraction of human infectious pathogenic strains from Source S (unitless; presented in
Table 2-4).
IrP is the fraction prevalence of infection in the Source S (unitless; presented in Table 2-3).
V is the volume of water ingested during the recreational event (L; baseline
assumption = 0.019 L).
P%S = fa-r (P.rp) [Eq- 2C]
where:
P[£f is the probability of infection from reference pathogen rp originating from source S
(unitless).
fd-r (lxrp) 's the mathematical dose-response function for reference pathogen rp (presented in
Table 2-5), which estimates the probability of infection at a given dose (unitless).
PS* = Pmtnf X C/ [Eq- 2D]
where:
Pj^'S is the probability of illness from a specific reference pathogen rp originating from Source S
(unitless).
Pm\inf's the proportion of individuals infected with reference pathogen rp who experience
illness (unitless; presented in Table 2-6).
Pill = 1 - rirp(l - PmS) [Eq- 2E]
where:
Pfu is the estimated probability of illness from a specific source of contamination S (unitless;
assumes the independence of each reference pathogen in Source S).
Pia = 1 - ns(l - P?u) [Eq. 2F]
where:
Pui is the estimated total probability of illness from all sources of contamination and all
pathogens in the waterbody (unitless; assumes the independence of each contamination source).
72
-------
As discussed in Approach 1, creating a distribution of the estimated probability of illness is possible
using Monte Carlo simulation. Creating a distribution requires random drawing with replacement
multiple times from the FIB distribution and, potentially, random drawing from the statistical
distributions of the volume of water ingested and the dose-response parameters.
Table 2-9 summarizes the definitions of the variables in the above equations, provides the notation for
each variable, and indicates where the data for the variable can be located or indicates that the data
are from the water quality study. Appendix F contains annotated computer code in R that is publicly
available and can be used to implement Approach 2.
Table 2-9. Approaches 1 and 2: definitions and sources of variables used in equations.
Variable
Definition
Data source
fM
L>rp
Measured density of reference pathogen in
waterbody (pathogens per L)
Site-specific reference pathogen data
cs
°rp
Estimated density of reference pathogen in
waterbody derived from Source S (pathogens per L)
Calculated based on Equation 2A
Cfib
Waterbody density of FIB using a culture method
(CFU per L)
Site-specific FIB (enterococci or E. coli) data (GM
calculated for the waterbody in Step 2)
D S
nFIB
Density of FIB using a culture method (CFU/L or
CFU/g) in the source
Table 2-7 data are from the peer-reviewed scientific
literature. For other fecal sources, directly gather
enterococci density in the sources at the site or find
published information.
D S
n.rp
Density of each reference pathogen from Source S
(number of pathogens or genomes/L or number of
pathogens or genomes/g)
Table 2-2 data are from the peer-reviewed scientific
literature. For other fecal sources, directly gather
pathogen densities in the sources at the site or find
published information.
[Xrp
Estimated dose of reference pathogen derived from
Source S (number of pathogens ingested)
Calculated based on Equation 1A or 2B
Pfib
Proportion of FIB deriving from source S (unitless)
Site-specific information
Prp
Fraction of human infectious pathogenic strains
(unitless)
Table 2-4
Irp
Prevalence of infection in the nonhuman source
(unitless)
Table 2-3
V
Volume of water ingested (L)
Described in the text (point estimate of 0.019 L per
person per primary contact recreation event)
prp
inf
Probability of infection from reference pathogen rp
(unitless)
Calculated based on Equation IB
prp,S
inf
Probability of infection from reference pathogen rp
originating from Source S (unitless)
Calculated based on Equation 2C
prp
ill
Probability of illness from reference pathogen rp
(unitless)
Calculated based on Equation 1C
prp,S
ill
Probability of illness from a specific reference
pathogen rp from Source S (unitless)
Calculated based on Equation 2D
pS
"ill
Probability of illness from Source S from all
pathogens (unitless)
Calculated based on Equation 2E
Pill
Probability of illness from all sources (unitless)
Calculated based on Equation ID or2F
73
-------
2.2.2.4.4 "Relative" QMRA modeling
A relative QMRA complements a forward QMRA by normalizing risks from specific fecal sources to a
specified FIB density (Soller et al., 2015). A relative QMRA is an application of the health model in a
scenario where both the monitored FIB and estimated risk from the forward QMRA are elevated above
the recommended water quality value and target illness rate, respectively. If the sanitary
characterization indicates that nonhuman sources predominate, then a relative QMRA can be used to
understand what the estimated target illness rate would be in the waterbody when it is meeting the
EPA's recommended RWQC FIB levels. The level and types of pathogens determine the risk due to fecal
contamination, but WQS are based on the level of FIB; therefore, to compare the water at the study
site to the EPA's recommended water quality values and target illness rate, there is a need to
normalize the risk estimates to the EPA's recommended level of FIB. The analytical approach displayed
in Figure 2-9 shows how a QMRA can be designed to answer the question: What would the predicted
illness level be if the waterbody water quality was at 35 enterococci CFU per 100 mL? Conducting a
relative QMRA can assist users in interpreting the forward QMRA results and provide additional
substantiation for alternative WQC developed in Step 4 (Section 3.4).
Comparing Figure 2-9 to the forward QMRA displayed in Figure 2-8, the input data differ from those
used in the forward QMRA. For the relative QMRA, a specific level of FIB is assumed in the
waterbody—in this case, the current RWQC levels for enterococci or E. coli. These FIB levels are used in
conjunction with the FIB and reference pathogen levels in the land-applied material, the fraction of
human-infectious pathogenic strains in each fecal source of interest, the prevalence of infection in the
nonhuman source (proportion of animals shedding the pathogen), the proportions of FIB and reference
pathogens that mobilize during a rain event, and the volume of water ingested (Schoen and Ashbolt,
2010; Soller et al., 2010b). The relative QMRA model output is the probability of infection and illness
associated with exposure to water during recreation for each source of interest referenced to the
chosen level of FIB. Once the data are normalized to the EPA's recommended water quality value, two
outcomes are possible. If the mean of the predicted illness level is below the target illness rate, then
continue to Step 4. If the mean predicted illness level is above the target illness rate, then the
nationally recommended criteria apply and do not proceed further with this TSM at this study location.
See further discussion on interpreting results of the forward QMRA for cases where predicted illness
levels are higher than the illness level benchmark in Section 3.3.3.
2.2.2.4.5 "Reverse" QMRA modeling
Commonly, QMRAs are conducted using microbial density as an input, and the output is the estimated
risk level as was done for Approaches 1 and 2 in Step 3. In contrast, a "reverse" QMRA is when the
input is the risk level, and the corresponding microbial density is calculated (Figure 2-10). Reverse
QMRA is used in Step 4 (Section 3.4) to solve for FIB density rather than risk level. Conducting a reverse
QMRA in Step 4 results in the calculation of the alternative FIB GM value associated with the
predominant source of nonhuman fecal contamination affecting the waterbody. The same QMRA
parameters are used as described above in Step 3: (1) the volume of water ingested during swimming;
(2) the density of pathogens in the specific sources; (3) the mathematical dose-response relationships
and the parameter values for each reference pathogen; and (4) the conditional probability of illness
given infection for different pathogens. In contrast to the stochastic approach used in the forward
QMRA, the reverse QMRA uses a deterministic approach because a stochastic reverse QMRA can be
highly computationally intensive, and the output desired is a single value (i.e., a single FIB GM) instead
74
-------
++ Fate and Transport modeling is optional
Boxes outside of the darker gray box are model inputs and boxes
within the darker gray box are computations.
Arrows denote the conceptual flow of the relative QMRA model
Figure 2-9. Analytical approach for "relative" QMRA anchored at a specific FIB level.
75
-------
Pathogen and
FIB density in
human and
non-human
sources
Fate and
Transport+
Proportion of
FIB in
waterbody
deriving from
human and
non-human
sources
Dose-Response
functions and
parameters for
each reference
pathogen
Morbidity for
each reference
pathogen
Revise estimate
for RBTFIB
density in
waterbody
Provide initial estimate for the FIB density in
waterbody corresponding to target
probability of illness (plausible level for the
model to start iterative solving)
Compute density of each reference pathogen
in the waterbody deriving from the human
and non-human sources, respectively
+
Compute dose of each reference pathogen
received during a recreational event from
the human and non-human sources,
respectively
+
Compute probability of illness from exposure
to each reference pathogen deriving from
the human and non-human sources,
respectively
Compute probability of illness accounting for
all reference pathogens deriving from the
human and non-human sources, respectively
Compute total probability of illness
accounting human and non-human sources
Output final estimate for RBT FIB density in
waterbody corresponding to target
probability of illness
Volume of
water ingested
during a
recreational
event
Prevalence of
infection in
the non-
human
sources
Human
infectious
potential from
non-human
sources
++ Fate and
Transport modeling
is optional
Boxes outside of the
darker gray boxare
model inputsand
boxes within the
darker gray boxare
computations.
Arrows denote the
conceptualflow of
the relative Q.MRA
model
"Yes" indicates
model convergence.
"No" indicates
solving continues for
another iteration.
Figure 2-10. Analytical approach for reverse QMRA.
76
-------
of a range. These QMRA calculations can estimate the culturable enterococci or E. coli level for various
mixtures of human and nonhuman contamination. A case study example is provided in Step 4 to
illustrate the process.
The EPA has conducted and documented QMRAs for waters affected by human, cattle, pig, chicken,
gull, or nonpathogenic fecal sources (Schoen and Ashbolt, 2010; Soller et al., 2010a,b, 2014, 2015;
Schoen et al., 2011). An important result noted in the QMRAs evaluating mixed fecal sources was that
the risks are determined predominantly by the proportion of the contamination source with the
greatest ability to cause human infection, not necessarily the source that contributes the greatest
number of FIB (Schoen et al., 2011; Soller et al., 2014). When human sources are a component of the
mixed fecal loading, the viral pathogens in human waste pose the highest risk of gastrointestinal illness
to recreators. Soller et al. (2014) demonstrated that the estimated risks of gastrointestinal illness from
recreational exposures to fecal mixtures were up to 50% lower for mixtures with an approximately 30%
human component compared to fecal loadings of 100% human as measured by enterococci. If there
was gull, pig, or chicken fecal contamination and a low human contribution (i.e., less than the
approximate 30% contribution), reverse QMRA-predicted enterococci GMs were substantially higher
than the 2012 RWQC water quality recommendations but based on the same level of health
protection.
For reverse QMRA, estimating the FIB levels for recent human and animal contamination mixtures that
correspond to a specific target illness rate requires conducting a set of analyses, which essentially
perform the inverse of the Step 3 QMRA analyses described above for a specific level of predicted
illness. The steps below (which correspond to the example Python code Appendix G presents) are the
same as those used in Soller et al. (2010b, 2015).
This QMRA initially uses a preliminary value for the FIB density (enterococci in this example) in the
waterbody to compute the density of each reference pathogen attributable to each source as
described in Equations 3A and 3B. Derive this FIB density via numerical iteration as described below.
d NHS
CrT = P$iHbS X CFIB x [Eq. 3A]
kfib
where:
C^s is the estimated density of the reference pathogen (number of pathogens or genomes/L)
in the waterbody deriving from the nonhuman source NHS.
Pf/b5's the proportion of FIB in the waterbody that is attributable to the nonhuman source NHS
(unitless; from site-specific information).
CFIB is an initial estimate for FIB density (CFU/L) in the waterbody that is assumed to
correspond to the required target probability of illness (this estimate is iteratively updated).
RypSis the density of the reference pathogen (number of pathogens or genomes/L or number
of pathogens or genomes/g) in the nonhuman source NHS (Table 2-5 or the source at the site).
RfibS's ^e density of FIB (CFU/L or CFU/g) in the nonhuman source NHS (Table 2-7 or the
source at the site).
77
-------
Crp — Pfib x CfjB X -Ths~ [Eq- 3B]
kfib
where:
Cf?p is the estimated density of the reference pathogen (number of pathogens or genomes/L) in
the waterbody deriving from the human source HS.
PpfBis the proportion of FIB in the waterbody that is attributable to the human source HS,
which is equivalent to (1 - p^jg5) defined above (unitless).
CFIB is an initial estimate for FIB density (CFU/L) in the waterbody that is assumed to
correspond to the required target probability of illness (this estimate is iteratively updated).
Rrp is the density of the reference pathogen (number of pathogens or genomes/L or number of
pathogens or genomes/g) in the human source HS.
RpfB is the density of FIB (CFU/L or CFU/g) in the human source HS.
If the human source is a mixture of raw and secondary treated and disinfected WWTP effluent,
compute the ratio of pathogens to FIB in the human source in Equation 3B above as follows (also see
Soller et al., 2010b):
r, HS nraw „ WWTP
= 0.9889^ + 0.0111-j^wfp [Eq. 3C]
kF1B kF1B kF1B
where:
Rpfs 's the density of FIB (CFU/g) in raw human contamination (Table 2-7).
RrpW is the density of the reference pathogen (number of pathogens or genomes/g) in raw
human contamination (Table 2-5).
Rf^TP is the density of FIB (CFU/L) in the WWTP-treated human contamination (Table 2-7).
R^vtp is the density of the reference pathogen (number of pathogens or genomes/L) in
WWTP-treated human contamination (Table 2-5).
Note that the constants in Equation 3C are derived by anchoring a probability of 36 NGI per
1,000 recreators17 to an enterococci level of 35 CFU per 100 mL for an ingestion volume of 0.019 L (i.e.,
anchoring to the EPA's recommended target illness rate and corresponding water quality value). If the
anchor levels are adjusted either for the illness rate or the culture-enumerated FIB, the constants in
Equation 3C may also need to be adjusted.
Then, compute the dose of each reference pathogen attributable to each source per Equations 3D and
3E.
= V X CrNpHS X p?pHS X I™5 [Eq. 3D]
17 The recommended health goal associated with the national 2012 RWQC is 32 or 36 NGI per 1,000 primary contact
recreation events. These two values are not significantly different.
78
-------
= VxCPpS [Eq. 3E]
where:
fJ-rpS is the estimated dose of human infectious reference pathogens (number of pathogens or
genomes) attributable to the nonhuman source.
V is the volume of water ingested during the recreational event (0.019 L; Table 2-8).
PrpSis the fraction of human infectious pathogenic strains from the nonhuman source NHS
(unitless; presented in Table 2-4).
I™s is the fraction prevalence of infection in the nonhuman source NHS (unitless; presented in
Table 2-3).
jj.rp is the estimated dose of human infectious reference pathogens (number of pathogens or
genomes) attributable to the human source.
Compute the probability of infection from exposure to each reference pathogen derived from each
source per Equations 3F and 3G.
= ti-r (HrpS) [Eq- 3p]
PSr = fa-r «p ) [Eq- 3G]
where:
PinfNHS's t'ne probability of infection from a specific reference pathogen rp originating from the
nonhuman source NHS (unitless).
PfnfHS is the probability of infection from a specific reference pathogen rp originating from the
human source HS (unitless).
f^r is the mathematical dose-response function for reference pathogen rp (presented in
Table 2-5) that estimates the probability of infection at a given dose (unitless).
Compute the probability of illness from exposure to each reference pathogen derived from each
source per Equations 3H and 31.
nrp,NHS rp .. nrp,NHS ri- n
PiU = Pill\inf x pinf [Eq-3H]
nrp,HS rp w nrp,HS ri-
Pill ~ Pill\inf X ^inf [Ecl- 3|]
where:
Pfif'NHS is the probability of illness from a specific reference pathogen rp originating from the
nonhuman source NHS (unitless).
P[if'HS is the probability of illness from a specific reference pathogen rp originating from the
human source HS (unitless).
79
-------
Piii\inf's t'ne ProP0|"tion of individuals infected with reference pathogen rp who experience
illness (unitless; presented in Table 2-6).
Compute the probability of illness from exposure to each source, accounting for the effect of all
reference pathogens within that source, per Equations 3J and 3K.
PlNuHS = l-I\rPa-ParNHS) [Eq. 3J]
= 1 " IWl " PafHS) [Eq. 3K]
where:
is the estimated probability of illness from the nonhuman source of contamination NHS
(unitless; assumes the independence of each reference pathogen).
Piifis the estimated probability of illness from the human source of contamination HS (unitless;
assumes the independence of each reference pathogen).
Compute the total probability of illness accounting for exposure to both the human and nonhuman
sources using Equation 3L.
Pa, = 1 - (1 - P$r) X (1 - PUf) [Eq. 3L]
where:
Piu is the estimated total probability of illness accounting for exposure to both the nonhuman
and human sources of contamination in the waterbody (unitless; assumes the independence of
risk from human and nonhuman sources).
If the computed value of the total probability of illness in Equation 3L differs from the target
probability of illness, the initial estimate for the FIB density in the waterbody must be iteratively
adjusted until the computed value agrees with the target value of probability of illness. Usually, that
can be readily accomplished using analytical software packages. The FIB density level in the final
iteration of this procedure is the waterbody FIB density level that corresponds to the target probability
of illness for the specified contamination mixture. Table 2-10 summarizes the variables in the above
text and includes where the information for the variables can be found in the TSM. In this TSM
example, enterococci was used as the FIB.
The values generated from modeling would apply if the sources affecting the waterbody were well
understood and characterized. In addition, the land use patterns in the watershed also would need to
be consistent with the conditions under which the water quality study was conducted (Step 2). Note
that it is unlikely that any waterbody with a WWTP outfall would qualify as having a predominantly
nonhuman source. Waterbodies potentially affected by sanitary infrastructure failures need to be
evaluated carefully in this process by understanding under what conditions exfiltration from
infrastructure, septage in poorly draining soils, and other potential human inputs may occur (e.g.,
rainfall-associated).
80
-------
Table 2-10. Reverse QMRA: definitions and sources for variables in equations.
In-text variable
Definition
Data sources
/-¦NHS
L.rp
Estimated density of reference pathogen in
waterbody derived from the nonhuman source
NHS (pathogens per L)
Calculated based on equation 3A
pHS
L,rp
Estimated density of reference pathogen in
waterbody derived from the nonhuman source HS
(pathogens per L)
Calculated based on equation 3B
Cfib
Waterbody density of FIB based on a culture
method (CFU/L)
Initial estimate that is iteratively updated
-NHS
Pfib
Proportion of FIB deriving from the nonhuman
source NHS (unitless)
Site-specific information
Vfib
Proportion of FIB deriving from the human source
HS (unitless)
Calculated based on (1 - p?,1s)
d NHS
nFIB
Density of FIB (CFU/L or CFU/g) in the nonhuman
source NHS
Table 2-7 data are from the peer-reviewed
scientific literature. For other fecal sources,
directly gather FIB densities in sources at the site
or find published information.
dHS
kFIB
Density of FIB (CFU/L or CFU/g) in the human
source HS
Calculated as ratio to pathogen density
d NHS
l\rp
Density of each reference pathogen in the
nonhuman source NHS (number of pathogens or
genomes/g or number of pathogens or
genomes/L)
Table 2-2 data are from the peer-reviewed
scientific literature. For other fecal sources,
directly gather reference pathogen density in
sources at the site or find published information.
D HS
i\ rp
Density of each reference pathogen in the human
source HS (number of pathogens or genomes/g or
number of pathogens or genomes/L)
Calculated as ratio to FIB density
nraw
kFIB
Density of FIB (CFU/g) in raw human
contamination
Table 2-7 data are from the peer-reviewed
scientific literature. For other fecal sources,
directly gather FIB density in sources at the site or
find published information.
uWWTP
kFIB
Density of FIB (CFU/L) in WWTP-treated human
contamination
Table 2-7 data are from the peer-reviewed
scientific literature. For other fecal sources,
directly gather FIB density in sources at the site or
find published information.
nraw
[\rp
Density of each reference pathogen in raw human
contamination (number of pathogens or
genomes/g)
Table 2-7 data are from the peer-reviewed
scientific literature. For other fecal sources,
directly gather reference pathogen density in
sources at the site or find published information.
DrP
WWTP
Density of reference pathogens in WWTP-treated
human contamination (number of pathogens or
genomes/L)
Table 2-7 data are from the peer-reviewed
scientific literature. For other fecal sources,
directly gather reference pathogen density in
sources at the site or find published information.
..NHS
H-rp
Estimated dose of reference pathogen derived
from nonhuman source NHS (number of
pathogens)
Calculated based on equation 3D
V-rp
Estimated dose of reference pathogen derived
from human source HS (number of pathogens)
Calculated based on equation 3E
81
-------
In-text variable
Definition
Data sources
-NHS
rrp
Fraction of human infectious pathogenic strains of
the reference pathogen in the nonhuman source
(unitless)
Table 2-4
i NHS
'rp
Prevalence of infection in the nonhuman source
for the reference pathogen (unitless)
Table 2-3
V
Volume of water ingested (L)
Described in the text (point estimate of 0.019 L
per person per primary contact recreation event)
prp.NHS
inf
Probability of infection from a specific reference
pathogen rp deriving from the nonhuman source
NHS (unitless)
Calculated based on equation 3F
prp.HS
inf
Probability of infection from a specific reference
pathogen rp deriving from the human source HS
(unitless)
Calculated based on equation 3G
prp.NHS
Ull
Probability of illness from a specific reference
pathogen rp deriving from the nonhuman source
NHS (unitless)
Calculated based on equation 3H
prp.HS
Ull
Probability of illness from a specific reference
pathogen rp deriving from the human source HS
(unitless)
Calculated based on equation 31
n NHS
"ill
Probability of illness from all pathogens in the
nonhuman source NHS (unitless)
Calculated based on equation 3J
pHS
rill
Probability of illness from all pathogens in the
human source HS (unitless)
Calculated based on equation 3K
Pill
Probability of illness from the human and
nonhuman sources (unitless)
Calculated based on equation 3L
82
-------
3.0 Site-Specific Alternative Criteria Development for Waters
Predominantly Impacted by Nonhuman Fecal Contamination
This section discusses each step in developing alternative criteria by adjusting the EPA's
recommendations for culturable E. coli and enterococci by accounting for the potential lower human
health risk associated with exposure to primarily nonhuman fecal sources. The information is
presented with a level of detail sufficient for a technical audience (i.e., those conducting evaluations to
develop the criteria).
To initiate the data collection and analysis steps discussed in this section, users of this document will
have completed the planning and scoping process described in Section 2 and have identified one or more
potential study locations where evidence suggests nonhuman fecal sources are likely to predominate.
The process described in this section incorporates the microbial and exposure parameters information
presented in Section 2 and the user-supplied information into the QMRA-based framework pictured in
Figure 2-6 (see Section 2.2.2.4). Steps 1-4 described in this section correspond to the four boxes in
Figure 2-6. In each section, a more detailed figure for each box corresponding to Figure 2-6 is presented,
decision points are identified, and an illustrative example of the process for each step is provided.
Documenting the details in a transparent and clear way in supporting documentation will facilitate public
comment by stakeholders and the EPA's evaluation of WQS submissions.
Feces-contaminated recreational waters can contain pathogens with the potential to cause human
illness. Because different sources of fecal material can be associated with different types and levels of
pathogens, the nature and loading of fecal source(s) are important variables to characterize. Further,
characterizing whether the water body is predominantly contaminated with nonhuman fecal sources
will be important to establish to support adjusting the 2012 RWQC. To begin, information on fecal
sources and the hydrodynamics of fecal loading to the waterbody or waterbodies being characterized
is first collected via a process termed "sanitary characterization" (Section 1.3.2). The sanitary
characterization includes two steps: (1) a sanitary survey (Step 1) and (2) a water quality study (Step 2).
If the predominance of nonhuman fecal inputs is established, then the next step (Step 3) is to combine
the information collected in Steps 1 and 2 with the information on pathogenicity and exposure,
discussed in Section 2, by conducting a QMRA. The QMRA estimates the potential human health risks
from primary contact recreation in the waterbody with predominantly nonhuman fecal inputs.
Step 4 describes how to derive alternative WQC using the information and results from Steps 1 through
3. The alternative criteria adjust the magnitude of the EPA's national recommendations to reflect the
predominance of nonhuman fecal inputs. Any alternative WQC developed using this TSM will include a
GM, STV, and a BAV consistent with the EPA's national recommendations.
Users of this TSM should understand the representativeness of the location chosen for conducting the
sanitary characterization and the potential for applying the results of this TSM process beyond the
location studied. Consider the variable fecal source loading dynamics, such as differences between
baseflow and wet weather conditions or between tidal states. Site selection (including the
consideration of unique watershed characteristics) for the sanitary characterization steps (including
the data generated and, by extension, the QMRA results) can influence the decision on the applicability
of the results beyond the study location. It is recommended that a sanitary characterization be
conducted at multiple sites where the nonhuman fecal source is thought to predominate to document
83
-------
the variability of conditions, fecal loading and resulting risk estimates. A multi-site characterization
could provide additional substantiation to decisions on the broader applicability of results.
3.1 Step 1: Sanitary Survey: Identify the Contributing Fecal Sources
In Step 1, information about the fecal sources affecting the
waterbody is collected through a sanitary survey and other
available sources of information to help understand the
potential sources affecting the waterbody more clearly (Text
Box 3-1). The EPA developed a new QMRA sanitary survey
form to guide data collection and documentation for this
step. (Refer to Appendix B, QMRA Sanitary Survey Form.)
The QMRA sanitary survey form can be augmented with
other useful documentation (Figure 3-1). For example, the
management history of the watershed might provide
information on historical water quality and identif
previous source control efforts. The history of
BMPs and other mitigation efforts intended to
attenuate fecal sources may also be useful
information to document as part of the sanitary
characterization. Impervious surfaces and
constructed drainage features or other built
structures are also useful information that may
impact fecal sources.
The sanitary survey form in Appendix B includes
three main sections:
1. A summary form at the top of the sanitary
survey form to integrate information
collected about potential fecal sources
identified in the watershed that can affect
the particular waterbody.
2. A watershed information form (Section 1)
for documenting details about the
watershed and existing water quality
monitoring location(s).
3. A series of worksheets that are used to
document details about potential sources i
fecal contamination in a watershed or at a recreational site.
a. Worksheet A: Description of land use, maps, and physical conditions.
b. Worksheet B: Weather conditions, water quality sampling and modeling.
c. Worksheet C: Human fecal sources.
Text Box 3-1. Decision Point Step 1
Based on the available information, a
conclusion is made that the
predominant sources of fecal
contamination impacting the waterbody
are nonhuman. Include nonfecal
source(s) of FIB if significant.
Substantiate this conclusion using a
sanitary survey and other lines of
documented evidence.
Step 1: Sanitary Survey:
Understanding Sources Impacting
Watershed
• Conduct a QMRA sanitary survey
Compile information on:
• Existingand historical management
practices
• Past monitoring data
c
• Existing information on fecal source
o
identification and tracking
"ro
N
'
• Survey of wildlife, agricultural, and
CD
•4—>
domesticanimals
U
ro
• Future planning and zoning
ro
_c
U
>.
information (if available)
• Maps
fC
-1—>
• Hydrological data
'c
ro
• Meteorological data
LT)
Step 2: WaterQualityStudy: fecal
indicator and pathogen monitoring,
source tracking
1
Figure 3-1. Information collected in Step 1.
(This figure shows details of Step 1 in Figure 2-6.)
84
-------
d. Worksheet D: Stormwater, urban runoff, drains, streams, wetlands.
e. Worksheet E: Agricultural animals and wildlife.
An explanation for the data requested in specific sections is provided following the worksheets.
Any existing fecal source identification information can be helpful in identifying sources of FIB and
pathogens (U.S. EPA, 2011a). Summarize the fecal source information, including the basis of the
information, such as peer-reviewed publications, watershed reports prepared by local or state
agencies, discharge permits, or waste load allocation information. The uncertainties, variability, and
assumptions associated with the available fecal source identification studies need to be summarized
clearly, concisely, and transparently. The relevance of the existing studies should also be described
(e.g., how recent are the studies, are existing conditions similar to the conditions extant at the time of
the study, did the existing study characterize the hazardous condition[s]) and focus on points that
apply to the waterbody and are relevant with respect to microbial WQC. Existing data for microbial
source markers can be useful for understanding the nature of fecal inputs in the watershed and can be
included. Please note for the purposes of this TSM, source marker data do not necessarily reflect the
potential risk from that source. Examples of RBTs for microbial source markers are available in the
scientific literature (Brown et al., 2017a; Boehm et al., 2018; Wu et al., 2020; Boehm and Soller, 2020).
The sanitary survey, coupled with the other information discussed in Figure 3-1, can be useful for
establishing evidence of source(s) that can be confirmed in Step 2.
Document the occurrence and prevalence of wildlife (e.g., warm-blooded animals) that could be a
source of FIB, pathogens, or both. For example, waterfowl can contribute substantially to fecal
indicator loads (Hansen et al., 2011; Sinigalliano et al., 2013). Exposure to gull fecal contamination has
been shown to pose significantly lower human health risks than exposure to human fecal sources at
equivalent culturable FIB levels because the pathogen profile present in gull feces differs from the
pathogen profile in human feces (Schoen and Ashbolt, 2010; Soller et al., 2010b). The differences in
pathogen profiles relative to the loading of culturable FIB from each source result in a risk differential
relevant to risk management decisions and public health protection. A wildlife survey can document
the types of wildlife that are present, seasonal fluctuations, and other discernible wildlife patterns.
Sanitary surveys also include documentation of the occurrence of agricultural animals, such as livestock
and poultry operations, recreational animals, such as horses, and the presence of domestic animals,
such as dogs allowed at the beach. Published QMRA evaluating recreational exposure to the direct
deposition of cattle feces into surface water demonstrates a similar magnitude of gastrointestinal
illness risks as human fecal sources (Soller et al., 2010b); however, the risk from the indirect
deposition, or rainfall-induced runoff, of cattle manure can be substantially lower due to differences in
attenuation and mobilization of pathogens and FIB in the applied manures prior to rainfall events (U.S.
EPA, 2010a; Soller et al., 2015). Effective manure management practices can encourage pathogen
attenuation, reduce offsite transport of animal feces in runoff, and decrease fecal contamination of
surface waters, so it may be important to record such practices if direct deposition might not represent
the situation at the study location (ASCE and U.S. EPA, 2000; CWP, 2007).
Geographic information system (GIS) mapping can also provide important insights about a watershed.
A map can be included in the initial sanitary survey identifying any WWTP outfalls or other features
that could contribute (e.g., septic systems or pit toilets) or attenuate (BMPs) FIB and pathogens on the
map. Note that a WWTP in the watershed is considered a significant human source even if treatment is
85
-------
effectively reducing FIB. A properly designed and operated WWTP reduces culturable FIB more
effectively than pathogenic viruses and protozoa; therefore, waters affected by effluent can pose
elevated risks to human health that are not predicted by culturable FIB (Wade et al., 2006, 2008, 2010,
2022).
Any information on currents and tidal effects can help characterize how the FIB and pathogens move
throughout the watershed. In some cases, hydrological data might be available in map format. This
information can also be important for designing the sampling strategy for the Step 2 water quality
study.
If the sanitary survey indicates that the sources appear to be predominantly nonhuman, and other
factors such as public perception and available resources indicate that alternative criteria would be
desirable and that downstream waters would still be protected, then proceed with Step 2, the water
quality study. The information from Step 1 might also indicate the potential benefit of management
actions (e.g., initiating manure management practices, installing fencing in the watershed to prevent
livestock from having direct access to surface water, implementing sanitation practices at beaches to
deter waste scavenging by birds and other animals) that could positively affect water quality. If this is
the case, ideally, the management actions could be implemented before Step 2 is initiated. The EPA's
Handbook for Developing Watershed Plans to Restore and Protect Our Waters walks users through the
basic watershed planning and implementation process (U.S. EPA, 2008b, 2013c).
Alternatively, if the results of the sanitary survey indicate a higher contribution of human fecal
contamination than initially thought, then a decision not to proceed with developing alternative
criteria could be made. The presence of point sources, such as WWTP effluent discharges or combined
sewer overflow (CSO) outfalls, can contribute significant loading of human enteric pathogens to
surface waters. Higher densities of older septic systems in areas with poorly draining soils can also
result in significant contributions of human fecal contamination (Peed et al., 2011). Published QMRA
results have suggested human contributions of approximately 20%-30% as a potential threshold for
determining if a water is predominantly affected by nonhuman fecal sources (Schoen et al., 2011;
Soller et al., 2014). As the proportion of human fecal loadings increases, a clear understanding of how
and when these inputs occur is needed so that the potential higher-risk scenarios these sources can
cause are identified. Information on source loading dynamics can help inform the monitoring approach
and analytical targets included in the water quality study discussed in Step 2.
3.2 Step 2: Source Confirmation: Conduct a Water Quality Study
In Step 2, a water quality study is conducted to characterize
fecal indicators, source markers, and pathogens (Text Box
3-2). The water quality study can (1) be used to verify the
results of the sanitary survey and (2) provide data for input
into QMRA in Step 3. The reference pathogens chosen for
monitoring during the water quality study can be based on
the fecal sources identified in Step 1.
Ideally, the information collected during the water quality
study supports the results of the sanitary survey—that the
predominant fecal loading to the waterbody is nonhuman.
The scope of the water quality study can include FIB and
Text Box 3-2. Decision Point Step 2
The collected information should
characterize the source(s) of fecal
contamination in the watershed. A
conclusion is made that the
predominant sources of fecal
contamination affecting the waterbody
are nonhuman. Substantiate this
conclusion using the data collected via
monitoring and modeling as part of the
water quality study.
86
-------
pathogen monitoring in the waterbody or
waterbodies of interest, FIB and pathogen
monitoring in the feces of the principal sources of
concern, fate and transport modeling, and fecal
source identification and tracking information,
such as microbial source markers (Figure 3-2).
The water quality study can also help
investigators understand the magnitude of fecal
loading and under what conditions the fecal
loading occurs. The water quality study approach
and design consider the fecal loading dynamics,
and the monitoring program reflects the
hazardous condition when the fecal loading is
thought to occur. For example, if the fecal loading
from a source is associated with wet weather,
then the monitoring program would reflect wet
weather events. For event-driven fecal loadings,
monitoring the base flow conditions is not
necessarily informative nor provides robust data
for QMRA. Users are encouraged to consider
monitoring at locations with the potential for
fecal loading (e.g., if a nonhuman source is
thought to enter surface waters via runoff, then
sampling close to that location when runoff is
occurring can be helpful to understand the
potential fecal loading and contamination dynamics). Monitoring at locations proximal to the fecal
loading can aid in enumerating pathogens that may occur in the source.
To begin the water quality study process, users of this document develop a SAP, which describes the
processes for characterizing water quality and fecal sources and analyzing the data collected, including
quality assurance plans. The SAP should be designed to capture the hazardous conditions where the
identified fecal sources in Step 1 are thought to contribute to the loading of fecal material to surface
waters. If fate and transport parameters, such as accounting for mobilization or environmental decay,
or integrated environmental modeling (IEM) are anticipated, the SAP discusses the data being collected
and how it is used in a model. Appendix C includes an example SAP used to characterize levels of FIB,
source markers, and pathogens in a study the EPA conducted to support a QMRA at a beach in
Boqueron, Puerto Rico. Having a detailed SAP and having subject matter experts review it can help
ensure the scientific defensibility of the effort. Quality assurance procedures for data collection are
also an important part of the SAP. The SAP is expected to be representative of the meteorological and
hydrological conditions in the watershed. The EPA has published guidance on systematic planning
using the data quality objectives (DQO) process (U.S. EPA, 2006b). The DQO process can be helpful
during planning when environmental data are used to estimate contamination levels and inform
decision-making. Refer to Chapter 7 of U.S. EPA (2006b), "Step 7: Develop the Plan for Obtaining Data,"
to inform the development of a SAP for the water quality study.
Fecal source identification information is important evidence for demonstrating that predominant
impacts on the site are nonhuman. Document the specificity and selectivity of source markers used for
Step 1: Sanitary Survey:
UnderstandingSources Impacting Watershed
Step 2: Water Quality Study: fecal indicator
c
o
and pathogen monitoring, source tracking
05
M
• Create a SAP for conducting the WQ
°i_
CD
study
u
03
• Select reference pathogens based on
ro
_C
source(s)
U
>
• Determine the FIB densities in the
ro
+->
waterbody and presumed fecal source(s)
'c
03
• Determine reference pathogen densities
I/)
in the waterbody and presumed fecal
source(s)
• Other studies that provide information
on sources
Predominantfecal
contamination source?
Figure 3-2. Water quality study.
(This figure shows details of Step 2 in Figure 2-6.)
87
-------
fecal identification. It is recommended to evaluate fecal samples or discharges from the presumed
source(s) to test that the source identification markers chosen can reliably detect that source. For
example, if birds are suspected of being an important source, test the marker on fecal samples from
the birds thought to be a contributing source. This is an active area of research, and examples of
relevant studies include Boehm et al. (2013), Shanks et al. (2010, 2016), Sinigalliano et al. (2013), Unno
et al. (2018), and Holcomb and Stewart (2020). The fecal source identification characterization can
indicate whether the sources are predominantly nonhuman and, in turn, inform the selection of
reference pathogens for that waterbody.
When the information collected in Step 1 indicates that humans are not a contributing fecal source,
the EPA recommends using human MST marker(s) to confirm that significant loading of human fecal
material is not occurring. A pattern of positive results in the water quality study using a human marker
when the sanitary survey indicated little to no human inputs suggests that potential human source(s)
contributing to the waterbody may have been missed; thus, it might be necessary to revisit the sanitary
survey step or reevaluate the study site as a location to collect data supporting the development of
alternative criteria. If source markers indicate human inputs (qualitatively or quantitatively), it is
important to have a better understanding of the potential magnitude and contamination pattern of the
human input. The presence of human markers below the LOQ of the detection methods is best
evaluated in context to the overall fecal loading to a waterbody. Detection of microbial source markers
does not necessarily indicate elevated human health risks, but marker detection can, for example,
inform on the monitoring approach in the water quality study and choice of reference pathogens to
include in the QMRA.
Select reference pathogens that reflect the enteric
pathogens potentially present in the identified fecal sources.
The EPA is providing information on the standard QMRA
reference pathogens adenovirus, Campylobacter spp.,
Cryptosporidium spp., E. coli 0157:H7, Giardia spp.,
norovirus, rotavirus, and Salmonella enterica (Text Box 3-3).
For example, if gulls are the source of the fecal
contamination, the pathogens of concern are Salmonella
enterica and Campylobacter spp. For pigs, the pathogens of
concern are E. coli 0157:H7, Campylobacter spp.,
Salmonella enterica, Cryptosporidium spp., and Giardia spp.
However, users of this TSM are not limited to these
reference pathogens. The EPA selected these reference
pathogens to represent waterborne illness in recreational
waters for several reasons: (1) they represent the three
classes of pathogens of concern (viruses, protozoa, and
bacteria); (2) they are found in a wide variety of sources;
(3) they have associated peer-reviewed dose-response
relationships in the literature; and, (4) they account for
more than 97% of nonfoodborne illness in the United States
(Mead et al., 1999; Scallan et al., 2011a,b). Thus, these pathogens can be considered responsible for
the largest proportion of nonfoodborne illness as reference pathogens for waterborne illnesses. The
reference pathogens norovirus, adenovirus, and E. coli 0157:1-17 are also considered index pathogens
because they represent a broader class of pathogens that share similar infectivity or environmental
Text Box 3-3. Reference Pathogens
Reference pathogens represent the
pathogens that are most likely to occur
in waterbodies with warm-blooded
animal (including human) fecal sources.
Campylobacter spp., Cryptosporidium
spp., E. coli 0157:H7, Giardia spp., and
Salmonella enterica represent
pathogens that are potentially zoonotic
and associated with human and
nonhuman fecal contamination (U.S.
EPA, 2009a). Adenovirus, norovirus,
and rotavirus are associated with
human fecal material and are not
zoonotic. This set of reference
pathogens is useful because it covers
the major waterborne pathogens
associated with non-foodborne illness
occurrence in the United States.
88
-------
fate and transport characteristics (WHO, 2004, 2005; Soller et al., 2010a; Ferguson et al., 2008). Using
an index pathogen to represent a broader group of pathogens is a widely accepted practice in the field
of QMRA (Roser et al., 2007; Soller et al., 2010b; Schoen et al., 2011).18 Research has demonstrated
that the bacteria (Campylobacter spp., E. coli 0157:H7, and Salmonella enterica) and protozoa
(Cryptosporidium and Giardia spp.) on the list can be zoonotic pathogens, that is, pathogens
transmitted from animals to humans. For more discussion of these pathogens, see Section 2.1.1.
In cases where the fecal source information collected demonstrates little to no human fecal inputs, a
decision not to include the virus reference pathogens (specifically adenovirus, norovirus, rotavirus)
could be made because these are exclusively associated with human sources. However, in this case,
limited monitoring for the reference viruses could be conducted to evaluate the assumption that there
are little to no human fecal sources. It is recommended that human MST markers, such as HF 183 (U.S.
EPA, 2019b), be used for this evaluation because MST markers tend to occur in higher abundance and
with more consistency compared to enteric viruses. When monitoring for both enteric viruses and
human markers, human markers may be detected and quantified, but viral pathogen monitoring may
have many nondetects due to the abundance difference and because the pathogens might not be
circulating in the population contributing to the source (e.g., seasonal occurrence of norovirus in
sewage). For suspected human source contributions, investigators might want to consider increasing
sample volumes in combination with ultrafiltration (Bofill-Mas and Rusinol, 2020; Korajkic et al., 2022).
Designing a monitoring approach to reflect when a suspected human source is reaching a surface
water can help characterize the magnitude of human fecal loading; see Zimmer-Faust et al. (2020) for
an example of a high-resolution sampling approach. Collecting samples at the point of discharge or
proximal to where the suspected source reaches surface water can aid in the detection of pathogens.
The other five reference pathogens (i.e., bacteria and protozoa) can be of zoonotic origin (U.S. EPA,
2009b). Using this information, select the reference pathogens based on the fecal sources identified
and/or suspected.
The water quality measurements for FIB, MST markers, and pathogens are considered because these
data are needed to characterize current conditions at the study site(s). Describe which FIB, MST, and
pathogen data are being collected in the SAP. The environmental monitoring data collected from the
study site(s) will be used in the next step (Step 3). In waters contaminated by feces, pathogens can be
present in much lower densities than FIB and MST markers. Pathogen occurrence in feces can be
transitory and affected by population prevalence and lifestage, which is partly why surrogates of fecal
contamination like FIB are relied on to routinely serve as metrics for water quality determinations
rather than pathogens (Text Box 3-3). Microbial enumeration methods tend not to be efficient when
low levels of the target are present and can have poor recovery. Characterize and describe the
recovery efficiency enumeration methods as part of the SAP development. Discuss enumeration
method LOD and LOQ in the SAP.
Process monitoring data for FIB and pathogens by handling nondetects and samples below the LOQ in
a defensible and transparent manner. LOD is the level below which the method does not detect the
presence of microbes. A "nondetect" result does not necessarily mean that the actual sample value is
zero but rather that the microbial target is present below the LOD of the assay used (Silvestri et al.,
18 Definitions are:
Reference pathogen: A pathogen responsible for a significant proportion of waterborne gastrointestinal illness.
Index pathogen: A pathogen that does not necessarily cause a large proportion of illnesses but represents a broader group
of pathogens that as a group are important.
89
-------
2017). The LOQ is the level at which the method can enumerate the microbe. For culture-based
methods, the LOQ is usually reported as a countable range, such as 20 CFU-200 CFU for membrane
filter plates (Silvestri et al., 2017). Below the LOQ, the method may detect but cannot accurately
enumerate the organisms. Above the upper level of the LOQ, the method cannot enumerate accurately
due to factors like overgrowth or colony growth interference (Be'er et al., 2009; ASTM, 2020).
Problems with upper LOQ can be attenuated by diluting the sample, so only the lower LOQ is
considered in the discussion that follows. For assays that enumerate microbes using culture methods,
the LOD can be the LOQ. Molecular methods, such as qPCR, can have an LOD that is below the LOQ.
Therefore, discussing both the LOD and LOQ of qPCR assays is essential. Understanding these
thresholds is important for preparing datasets for analysis.
Some water quality monitoring results may be below the LOQ or below the LOD. Depending on
whether enumeration of FIB or pathogens is used, results that are less than the LOD or LOQ will be
treated differently. FIB are often present at densities at or above commonly used assay LODs (which in
most cases is also the LOQ). For this reason, many samples will have an actual measured density of FIB.
Also, higher variability is expected below the lower end of the LOQ due to various biological, method
and matrix factors (Silvestri et al., 2017). Occasionally, a sample might be below the LOD. The scientific
literature contains precedent for replacing results below the LOD with the LOD, or V* the LOD, although
more complex replacement methods exist. Replacing nondetect values for FIB with the LOD or V* LOD
can be used, depending on the goal of subsequent statistical analysis. Provided the substitution does
not drastically change the data distribution including central tendency measures or the standard
deviation (SD), substitution will likely not significantly affect the analyses. For FIB, the nondetects are
most likely in the lower tail of the water quality distribution. Provided the replaced LOD values
represent a small fraction of the total data observations (< 15%) (U.S. EPA, 2000b), as is likely for FIB
data used in this context, replacement is acceptable. See Site-Specific Alternative Recreational Criteria
Technical Support Materials for Alternative Indicators and Methods for additional discussion of
handling nondetects for FIB (U.S. EPA, 2014a).
The ability to detect and quantify pathogens in surface waters is influenced by their occurrence and
prevalence in the population that is contributing to the fecal loading to a waterbody. If there are few
infected individuals excreting the pathogen, the population of infected excreters is small, or amount of
feces excreted per infected animal is low, then the overall pathogen loading to a waterbody can be
low. Pathogen occurrence may be affected by lifestage (e.g., super-shedding of Cryptosporidium and
E. coli 0157:1-17 by cattle) or season (Lai et al., 2012; Thomson et al., 2019; Antaki-Zukoski et al., 2021).
Because of the complex environmental/host/pathogen interactions, pathogens might be present in
ambient waters at densities near or below the assay LOD. For example, it is common in environmental
datasets for pathogens to be "ND" in more than 50% of samples assayed. In such cases, replacing
values below the LOD with the LOD or V* LOD (or LOQ) is not advised because it would drastically
change the data distribution. The EPA addressed this issue when estimating densities of
Cryptosporidium in U.S. drinking water sources (U.S. EPA, 2005). The goal was to obtain the distribution
of Cryptosporidium densities in water using densities measured across the country. Many of the
measurements, however, were below the LOD. The EPA used a Monte Carlo simulation to estimate
densities given the volumes of samples the participants assayed, the participating lab's recovery, and
the count of Cryptosporidium oocysts obtained for the various samples (many were zero) (U.S. EPA,
2005). This technique resulted in a data distribution despite the numerous measurements that were
below the LOD. Alternatively, if the fecal sources affecting the waterbody were shown to be
predominantly nonhuman and the human fecal inputs do not exist or are negligible, nondetects can be
90
-------
replaced with zero for the human viral pathogens. If results are above the LOD but below the LOQ,
replacing those results with the LOQ is likely acceptable. With molecular analytical methods in
particular (for both pathogens and MST markers), a transparent description and documentation of the
LOD and LOQ thresholds is critical for demonstrating that data were prepared for analysis in a
scientifically defensible manner. For some pathogen enumeration methods with high LOQs, generating
quantitative environmental data may be challenging without applying analysis methodologies to
support data interpretation. This example illustrates the importance of the sanitary characterization
for this approach. Refer to Helsel (2012) for additional approaches for including nondetects in
environmental datasets.
Step 3 involves fitting the microbial data to a statistical distribution for input into the QMRA. A
common distribution for this purpose is a lognormal distribution (Text Box 3-4), which can be fully
characterized in terms of an estimated GM and the logSD.19 For example, the EPA recommended water
quality values are expressed as a GM of a lognormal distribution. The GM and logSD of the distribution
can be readily computed from the data collected in Step 2. A spreadsheet such as one in Microsoft
(MS) Excel20 is an effective and straightforward tool for this purpose. For example, using MS Excel,
compute the baselO log of each enterococci observation (the syntax in MS Excel is
=loglO(observation)). Then calculate the average (=average(loglO(observations))), and the SD of those
log values or the logSD (=stdev(loglO(observations))). The GM of the distribution is obtained by
computing the antilog of the computed average of the log values, and the geometric SD of the
distribution is obtained by computing the antilog of the computed logSD.21
Text Box 3-4. Lognormal Distribution and Geometric Mean
Water quality at a beach tends to be lognormally distributed because water quality data typically span many
orders of magnitude, and the distributions are right-skewed since they are bound by zero on the left
(Bartram and Rees, 2000; Kay et al., 2004; Wyer et al., 1999). General environmental sampling data tend to
follow lognormal distributions, a phenomenon discussed by Esmen and Hammad (1977). The data sets from
some beaches may not fit a lognormal distribution. However, based on data from the NEEAR beaches and
other studies from the literature, when enough data are collected and the samples below the detection limit
are treated in a statistically rigorous manner, the lognormal distribution tends to be a good fit for water
quality monitoring. For example, WHO based their recreational water Guidelines (WHO, 2003) on a
lognormally distributed dataset from the EU from over 11,000 bathing locations and over
121,000 enterococci observations (Kay, et al., 2004).
The use of the GM for RWQC is discussed in detail by Wymer and Wade (2007). Numerous researchers have
used the GM to discuss the central tendency of water quality at beaches (for examples, see Boehm et al.,
2009 and Whitman et al., 2004). EPA's 2012 RWQC recommends GM water quality values because the GM is
significantly associated with gastrointestinal illness in epidemiological studies. Note that this health
relationship is not tied to individual sample results.
19 The logSD is the SD of the baselO log of the data. The antilog of the logSD would be the geometric SD.
20 Microsoft. Microsoft Excel. Redmond, Washington: Microsoft, 2010. Computer Software.
21 This information will be used in Section 3, Step 3 (below) to calculate the density of FIB in the waterbody.
91
-------
The data and information from Steps 1 and 2 together constitute a sanitary characterization and
provide evidence about the sources of FIB and pathogens in the watershed sufficient to decide that the
predominant source affecting the waterbody is nonhuman. Waters with a greater susceptibility to
human fecal contamination would need a greater weight of evidence to demonstrate that nonhuman
sources are predominant. Depending on the level of evidence needed, approaches for demonstrating
source loading can range from qualitative to quantitative. Apportioning sources quantitatively solely
based on source marker data is not a straightforward process, although research in this area is
ongoing. However, multiple approaches exist, including source marker, FIB, watershed, and
meteorological information providing evidence of fecal contamination dynamics. Investigators have
used source marker data combined with GIS land use and E. coli measurements to characterize fecal
contamination trends on a watershed scale (Li et al., 2019). Human, canine, and avian source marker
data were evaluated in combination with FIB and precipitation information to identify sources
associated with an exceedance of the BAV (Shrestha et al., 2020). In another study, paired
measurements of viral and bacterial fecal indicators and human, ruminant, canine, and avian source
markers were assessed from six Great Lakes locations over a recreation season to estimate fecal score
ratios (Li et al., 2021). Approaches for estimating fecal mass loading from a watershed to a waterbody
and the process for developing TMDL apportionment can also be applied to inform this decision (U.S.
EPA, 2001).22
The EPA's researchers have employed recursive partitioning (RP) to evaluate MST marker and FIB data
(Bradshaw et al., 2016; McKee et al., 2020, 2021) using water quality data that are log-transformed
prior to linear regression. In Bradshaw et al. (2016), RP was used to predict pathogen presence based
on an analysis of independent variables. RP analysis assigned a score to the identified independent
variables to define the level of association. Inconsistent relationships between individual pathogens,
FIB and source markers were found. However, in a mixed land use system, a combination of FIB, MST,
and other water quality measurements, such as pH and water temperature, were helpful in assessing
microbial water quality (Bradshaw et al., 2016). RP can also be used to determine which marker, and at
what level, best predicts the presence of pathogens or density of FIB. For example, McKee et al. (2020)
found the dog source marker (DogBact) had the largest R2 value and was the best predictor of elevated
E. coli levels above the BAV. McKee et al. (2021) evaluated the spatial and temporal patterns of fecal
contamination in Congaree National Park, SC, which is affected by feral swine populations. The swine
marker was the most frequently detected of the various source markers included. Swine marker levels
above 43 gene copies per mL were associated with E. coli levels above the BAV (McKee et al., 2021).
The EPA is not suggesting a specific maximum allowable quantitative level of human fecal
contamination for developing alternative values for culturable E. coli or enterococci because there is a
wide range of variability of potential risk among human fecal sources (e.g., secondary treated and
disinfected effluent, raw or poorly treated sewage, septage), so the level of human fecal inputs that
would qualify as human predominated is dependent on the nature and magnitude of the human
inputs. However, the level of human fecal contamination is generally addressed in this TSM by
evaluating the susceptibility of the waterbody to human inputs. As discussed in Steps 3 and 4,
evaluation of human and nonhuman fecal mixtures and recreational exposures demonstrated a lower
risk of gastrointestinal illness when human fecal contributions were less than approximately 30% of the
mixture (Soller et al., 2014). For the purposes of this document, susceptibility to human fecal
contamination is considered "low" when the human contribution is below approximately one-third of
22 For the EPA's guidance on TMDLs, see https://www.epa.gov/tmdl (last accessed May 25, 2023).
92
-------
the total fecal loading to a waterbody. As the proportion of human contribution increases, human
enteric viruses begin to dominate the overall estimated risk (Schoen et al., 2011; Soller et al., 2014).
Two additional categories of human fecal susceptibility, "medium" (34%-67% human contribution) and
"high" (> 67% human contribution), have results similar to the 2012 RWQC recommendations (see
examples in Figure 3-8 and Table 3-2) (U.S. EPA, 2010a; Soller et al., 2014). The EPA recommends that
only waters falling into the low-susceptibility category be considered for alternative WQC. The
uncertainty associated with source apportionment and risk estimates will be affected by the quantity
and quality of the data collected. The range of variability associated with the data collected and the
implications of uncertainty for the decision-making process should be evaluated and understood. As
the human fecal contribution increases, the calculated indicator value will approach that of the 2012
RWQC, which is based on data collected in waters predominantly affected by human fecal sources.
Temporal variability of source contributions can be considered as part of source determination. Two
examples of how the sources of fecal contamination might vary over a year include manure slurry
applications during winter or sewer overflows during seasonal wet weather. The EPA and others have
demonstrated that nonhuman fecal sources can pose less potential for human health effects at the
same level of fecal indicator (Schoen and Ashbolt, 2010; Soller et al., 2010b; U.S. EPA, 2010a; WERF,
2011). Consider the information from Steps 1 and 2, along with stakeholders' feedback, when deciding
whether to proceed with the alternative criteria development process.
3.3 Step 3: HHRA—Forward and Relative QMRA modeling.
Step 3 uses the information collected in Steps 1 and 2 to
estimate the potential human health risks associated with
recreational exposure to the waterbody. In this step, the risk
assessment process is considered a "forward" QMRA
because pathogen occurrence is an input for calculating a
health risk estimate as the output. This step describes two
forward QMRAs (Figure 3-3). The outputs from both
approaches are then compared to the target illness rate
chosen and the EPA's recommended water quality values to
inform a decision on proceeding with developing alternative
criteria values (Text Box 3-5).
Conduct the Step 3 human health risk evaluation using the
QMRA model parameters summarized in Section 2.1 and the
data collected in Section 3.2 (Step 2) in two different but
complementary approaches to estimate the waterbody pathogen density:
1. Estimate the waterbody pathogen density for each reference pathogen based on the reference
pathogen data gathered in Step 2.
2. Estimate the waterbody pathogen density for each reference pathogen from the FIB density in
the waterbody, FIB density in the source, and reference pathogen density in the source using
the methods of Schoen and Ashbolt (2010) and Soller et al. (2010b).
As mentioned in Step 2 (above), pathogen levels in receiving waters can vary over time and space, and
the recovery efficiency of pathogen enumeration methods can be poor, particularly when levels of
pathogens are low in ambient waters. Therefore, estimate the densities of reference pathogens based
Text Box 3-5. Decision Points Step 3
• What target illness rate and water
quality values are chosen?
• How do the QMRA estimates of
illness compare to the target
illness rate?
• How does the water quality at the
study location compare to the
recommended water quality
values?
• Do the QMRA estimates support
proceeding with developing
adjusted water quality criteria?
93
-------
on both data collected and the range that
the literature reports in fecal material
(Section 2.1.1.1). Both approaches are
reasonable ways to address the
evaluation of potential human health
risks. Literature values can be used to
provide an upper-bound estimate for the
potential illness. In both cases, the
pathogen densities and the estimates of
the volume of water ingested during
recreation can be gathered from a dose-
response relationship from the peer-
reviewed literature and the conditional
probability of illness given infection from
the literature to estimate illness levels.
Results from these two approaches might
indicate a range of potential human
health risks, or they could yield
comparable results (e.g., pathogen data
collected in the waterbody or the
identified fecal source(s) might be in the
same range as reported in the peer-
reviewed literature). More detail on each
approach can be found in Sections 3.3.1
(Approach 1) and 3.3.2 (Approach 2).
Guidance for interpreting the results of
both approaches can be found in Section
3.3.3.
Step 3: Forward QMRA
Do Beth Approach 1 and2
• FIS and reference pathogen dataco llected at the site
and insc . i :¦= "•: "! Step2);
•In tsosqyantf icatiorifocthepathcgen methods;
¦ :e::' raliability and uncertaintf;
• Add t c r = : r.
• S t e-spear* factors ar.d datarct included a nc! the
just.f.cai.on for not including those factors;e.g. temporal
shifts in loading!.
Approach 1:
Approach 2:
input WQ data for the
input pathof ens and FIB in
feces and FIB in waterbodf
waterbocjy (see Appendix
(seeAppendixFfor
Efor eiampie)
eiampie)
Character zetheF Bd str butcnfrcm
data co ected n5tep2
\
\
Z7
X
dothe*i,aterqua*ty'"^»^
and "tieasest'nates
conparetc the
benchmarks'
Figure 3-3. Step 3: HHRA using two QMRA approaches.
(This figure shows details of Step 3 in Figure 2-6.)
As discussed in Section 2.2.2.4.1, the EPA developed an illustrative exposure scenario for use in the
QMRA analyses recommended in this TSM. The illustrative scenario assumes the direct deposition of
feces to a waterbody in close proximity to recreation. Users of this TSM may be interested in
incorporating fate and transport terms to describe the behavior of the pathogens, FIB, and microbial
source markers from the source to the study location. Refinement of the illustrative scenario with
additional fate and transport information, such as including indirect fecal deposition pathways, is
possible, but be aware that specific fate and transport characteristics in the watershed studied may
limit the scope of the applicability of the QMRA output. Using fate and transport modeling or
mobilization data to estimate pathogen densities in the studied waterbody can introduce additional
complexity and a wider range of assumptions compared to the direct deposition scenario described by
the EPA. A more generic indirect fecal deposition scenario was developed by the EPA in which animal
manures were land applied at agronomic rates and mobilized by rainfall (U.S. EPA, 2010a; Soller et al.,
2015). In this scenario, mobilized fecal pathogens were characterized at the edge of the area where
manures were applied. Pathogen densities in the runoff were used to estimate risk. This indirect
deposition scenario incorporates a common mechanism—wet weather-induced mobilization of
manures—and could be utilized if, for example, grazing animals are an identified fecal source in the
watershed studied and do not have direct access to the waterbody studied or its tributaries. The EPA
94
-------
decided to use a simplified exposure scenario (i.e., direct deposition) in its approach, given the
variability and uncertainty associated with the input parameters.23
The illustrative exposure scenario may not be representative for the user's study location, and there
may be interest in incorporating fate and transport because primary contact recreation occurs
downstream a significant distance from the predominant contamination source. To consider the
various decision points, the EPA has developed an IEM framework (Figure 3-4) that includes fate and
transport considerations and the health models used in QMRA to link the risk estimation between
source and receptor locations in a watershed (Whelan et al., 2014a,b; U.S. EPA, 2017a,b,c,d,e,f,g,h,i).
The IEM approach is more data-intensive than the approach described in this TSM, but both can
provide quantitative risk estimates that can inform decisions on proceeding to Step 4.
Problem Definition
Pathogens
Identification
Etiologies
Properties
Scenarios
Baseline
Alternatives
ZE
Data Access, Retrieval, and Processing
Field and
Laboratory
Monitoring
Sampling
Measurements
Sorption studies
I n acti vati on an d d i e-off rates
Mortality kinetics
Impacts: Sunlight, Temp
Data Access,
Retrieval, and
Processing
Met data
Soils &Topography
Land use/cover
Watershed/Stream Delineations
Policy-related Issues
Risk target(s)
IE
QMRA Investigations
Site characteristics and pathogens
Model scale and resolution
Risks by varying hydrologicconditions
Source apportionment
Risk by pathogen, fecal source, water type
Sensitivity/Uncertainty
Integrated Modeling Framework
_A_
Source-term
Loadings
Fate & Transport
Watershed
Hydrology
Water body
Network
Receiving
Waters
Health
Impacts
Dose-
Response
Relationship
Epidemiology
Studies
Iterate on sources, watersheds, water bodies and receptor locations
Risk
Quantification
Figure 3-4. An example of an integrated, multimedia modeling framework linking the problem
definition, data access, retrieval, and processing, and IEM (including the health models used in
QMRA) to quantify risk at receptor locations (adapted from Whelan et al. [2014a]).
23 Variability is differences attributable to true heterogeneity or diversity in a population or exposure parameter. Variability
implies real differences among members of that population. For example, different individuals have different intakes and
susceptibility. Intraindividual variability refers to differences over time for a given individual. Interindividual variability
refers to differences across members of a population at a given time. Variability in MRA cannot be reduced, but it can be
more precisely characterized. Uncertainty occurs because of lack of knowledge regarding the true value of a quantity, such
as a specific characteristic (e.g., mean, variance) of a distribution for variability, or regarding the appropriate and adequate
inference options to use to structure a model or scenario. The EPA also refers to these as model uncertainty and scenario
uncertainty. Lack of knowledge uncertainty can be reduced by obtaining more information through research and data
collection, such as through research on mechanisms, larger sample sizes, or more representative samples (U.S. EPA, 2007).
95
-------
Rather than recommending one specific method for modeling this type of exposure scenario, the EPA
suggests using health-protective assumptions to reduce complexity. If fate and transport parameters
are included, it is important to thoroughly characterize and carefully document the inclusion of fate
and transport information at all steps (FIB and pathogens on a site-specific basis). Also, when including
specific fate and transport information, it would be important to consider the applicability of the
resulting illness estimates more broadly if the information is used to develop alternative criteria
discussed in Section 3.4 (Step 4).
3.3.1 Forward QMRA Approach 1: Using Data Collected from Water Quality Monitoring
In Approach 1, the reference pathogen data collected in Step 2 water quality study is used to estimate
the waterbody pathogen density, which is then used in a forward QMRA to estimate human health risk
from recreational exposure. A QMRA example (U.S. EPA 2010a; Soller et al., 2015) is described to
illustrate the process associated with Approach 1. Although the analysis presented in the example was
not conducted for a specific waterbody, it does follow the steps in the forward QMRA Approach 1 (see
Section 2.2.2.4.3). The following data are needed for the forward QMRA process in Approach 1:
1. Reference pathogen densities: these data were collected as part of the water quality study in
Step 2.
2. Volume of water ingested: data for incidental ingestion of water while recreating is provided in
Section 2.2.2.2: Measures of Exposure.
3. Reference pathogen dose-response parameters: this information can be found in Section
2.1.1.3.2: Zoonotic Infectivity Potential.
4. Probability of illness given infection for each reference pathogen: this information can be found
in Section 2.1.1.3.3: Morbidity.
A forward QMRA consists of three steps: (1) a calculation of a dose for each reference pathogen; (2) a
calculation of the probability of infection for each reference pathogen; and (3) a calculation of the
probability of illness given infection for each reference pathogen. The probability of illness for each
reference pathogen is then summed to calculate a total estimated risk of illness from recreational
exposure.
U.S. EPA (2010a) and Soller et al. (2015), hereafter referred to as "Example 1," discuss a forward QMRA
conducted to estimate risk from recreational exposure to a hypothetical waterbody receiving rainfall-
induced runoff containing cattle, pig or chicken fecal material that was land applied at recommended
agronomic rates. In the scenario characterized in Example 1, FIB and reference pathogens were
mobilized by the rainfall and moved offsite to an adjacent waterbody where primary contact
recreation exposure occurs via incidental ingestion of undiluted runoff. This scenario was intentionally
constructed to provide health-protective estimates of risk (U.S. EPA, 2010a). The reference pathogens
included in the analysis were Cryptosporidium spp., Giardia lamblia, Campylobacter jejuni, Salmonella
enterica, and E. coli 0157:1-17. Human enteric virus reference pathogens, such as norovirus or
adenovirus, were not included because only nonhuman fecal sources were being characterized.
In Example 1, note that the actual parameter values used were current at the time of publication, and
there are differences in some of the values discussed in Soller et al. (2015) compared to the values
presented in this TSM (Section 2.1). For some parameters, newer information was available in the
96
-------
scientific literature and has been included and discussed in this TSM. It is recommended that the user
of this TSM be aware of how these values are incorporated into a forward QMRA (Figure 2-8) and be
familiar with the equations in Section 2.2.2.4.3. Over time, new information for these parameters may
be published in the peer-reviewed literature, and users could evaluate the quality of that information
and include it in the QMRA analysis when this occurs as needed. It is worth noting that the newer
published information included in this TSM did not result in substantial changes in the risk output or
the characterization of a QMRA analysis for recreational exposure to gull feces (see Section 4.2 for a
comparison).
In both Example 1 and the forward QMRA R code supplied in Appendix E, equations for each step of a
forward QMRA are combined so that reference pathogen densities, either user-supplied from data
collected in Step 2 or estimated using pathogen mobilization rates as in the example, are combined
with the literature-based parameter values to provide a total probability of illness as an output. The R
code uses a stochastic approach (some inputs are distributions), but users could also use a
deterministic approach with point estimates as inputs using the equations in Section 2.2.2.4.3. For
transparency, the three steps are described in more detail below, and the example is referenced to
illustrate the process.
Calculation of dose of pathogen ingested. The dose of each reference pathogen will be influenced by
the density of the pathogen in the water. Water quality data collected in Step 2 are used to determine
the summary statistics (i.e., mean density, median density, minimum and maximum density, and the
SD) for each reference pathogen (Text Box 3-3). In equation 1A (Section 2.2.2.4.3) and in the forward
QMRA R code provided in Appendix E, the pathogen density is multiplied by the volume of water
incidentally ingested (Section 2.2.2.2) to estimate a dose. Equation 1A shows both pathogen density
and ingested volume as a point estimate rather than a distribution; however, whether a point estimate
or a distribution is used, the basic relationship the equation describes is the same. The forward QMRA
R code (Appendix E) combines user-supplied pathogen data with the exposure distribution parameters
for the volume of water ingested to calculate a dose. To use this R code, the user supplies (1) the log
mean and logSD of a distribution of incidental ingestion volume and (2) a spreadsheet containing
monitoring data for pathogen density collected in Step 2. Note the R code requires that the pathogen
data be in units per L, rather than units per 100 mL. The R code implements a Monte Carlo approach,
which runs 1,000 iterations of the calculations. In each iteration, random values for each of the model
parameters are drawn and used in the mathematical formulas.
In the example, direct pathogen measurement in surface water receiving runoff containing animal
manures was not conducted. Manures can have low and variable levels of pathogens (Soller et al.,
2015). The EPA conducted field experiments to empirically determine mobilization rates of reference
pathogens and FIB from land-applied manures. The mobilization rates were combined with literature-
based values for the proportion of animals shedding each pathogen and the density of the reference
pathogens in cattle, pig and chicken manures to provide an estimate of pathogen densities in water
(Soller et al., 2015). The approach used in the example included the following assumptions: intense
rain-inducing runoff, no attenuation of the pathogen between runoff and entry into surface water, and
ingestion of undiluted runoff (Soller et al., 2015). Also, in the example, the volume of water ingested
was modeled as a log-normal distribution with a log mean and logSD of 2.92 mL and 1.43 mL,
respectively (equivalent to a GM of 18.5 mL per recreational event; Dufour et al., 2006). Although the
approach described for characterizing pathogen levels is more complex than what the TSM discusses in
Step 2, the additional complexity helped address the variable nature of pathogen occurrence in animal
97
-------
fecal materials and resulted in scientifically defensible estimates of pathogen densities in runoff from
the experimental plots. The pathogen density estimates were then used to calculate a pathogen dose
given incidental ingestion, as discussed above.
Calculation of probability of infection from an exposure event. The dose of each reference pathogen is
then used to estimate the probability of infection from each reference pathogen. Equation IB (in
Section 2.2.2.4.3) shows that the probability of infection is a function of the dose-response
relationship. The dose-response information for each reference pathogen can be found in Table 2-5 in
Section 2.1.1.3.2. The forward QMRA R code provided in Appendix E includes the dose-response
models (e.g., Beta Poisson, Exponential, and Hypergeometric) and the pathogen-specific parameters
that are used in the model (e.g., alpha, beta, rho, eta are unitless parameters). The Approach 1 R code
includes dose-response information for Salmonella, Cryptosporidium, Giardia, Campylobacter,
adenovirus, and norovirus.
In the forward QMRA Example 1, the dose for each reference pathogen calculated above was
combined with the dose-response function to calculate the probability of infection for each reference
pathogen. It was assumed that infection from each pathogen is independent of the others (Soller et al.,
2015).
Calculation of probability of illness given infection and total gastrointestinal illness rate. The probability
of infection from each reference pathogen is multiplied by a pathogen-specific morbidity ratio to
produce a probability of illness (U.S. EPA, 2010a). The morbidity factor is the percentage of infections
that lead to illness for each reference pathogen. Equation 1C (in Section 2.2.2.4.3) shows how the
probability of infection multiplied by the proportion of infected individuals that become ill results in
the probability of illness. Table 2-6 in Section 2.1.1.3.3 includes literature-reported values for the
proportion of infections resulting in illness for each reference pathogen. Appendix E's forward QMRA R
code includes point estimates for the probability of illness, given infection for each reference
pathogen. The last step in the R code sums the risks from all the reference pathogens and calculates
the mean and median illnesses per 1,000 events, which is the output of the forward QMRA model. This
summing step is shown in Equation ID in Section 2.2.2.4.3. The median illness rate is used to compare
to the EPA's recommendations in Section 3.3.
In the example, the total probability of illness was calculated for three separate hypothetical
recreational exposure scenarios for waters containing cattle, pig, or chicken fecal material. The total
risk is displayed as a total gastrointestinal illness rate, and the GM culturable enterococci density is
shown in Figure 3-5 (Soller et al., 2015). The predicted median risk of illness from a recreational
exposure to the undiluted runoff from each of the land-applied fecal materials was: (1) 46 illnesses per
1,000 recreation events with an associated GM 14,000 enterococci CFU per 100 mL from runoff
containing cattle manure; (2) 15 illnesses per 1,000 recreation events with an associated GM of
36,000 enterococci CFU per 100 mL from runoff containing pig feces; and (3) 18 illnesses per
1,000 recreation events with an associated GM of 1,800 enterococci CFU per 100 mL from runoff
containing chicken feces (Soller et al., 2015). In Figure 3-5, the line in the middle of the box is the
median. The upper and lower edges of the box are the 75th percentile and the 25th percentile,
respectively. The bars (whiskers) extend to the 90th and 10th percentiles. The diamonds are at the
95th and 5th percentiles. The dotted line represents the EPA's recommended target illness rate of
36 NGI per 1,000 recreators.
98
-------
average
14,000 CFU/lOOmL
average
36,000 CFU/lOOmL
average
1,800 CFU/lOOmL
10°
Figure 3-5. Comparison of illness risks from exposure to undiluted runoff from land-applied
manures from three different species (from Soller et al. [2015]).
The results from this example forward QMRA show that cattle manure is associated with the highest
risk and pig slurry with the lowest risk among the three nonhuman fecal sources within this exposure
scenario. In all three scenarios, enterococci densities were substantially elevated compared to the
2012 RWQC recommendations; however, for the pig and chicken scenarios, risks were well below the
target illness rate associated with the 2012 RWQC. Although risks in the cattle scenario were higher
than the EPA's recommended target illness rate, the risk occurred at a much higher enterococci GM.
The interpretation of the results of the forward QMRA is an important aspect of thisTSM. The results
from the example are also discussed below in Section 3.3.3 to illustrate a decision point on proceeding
to Step 4.
3.3.2 Forward QMRA Approach 2: Using a Combination of Collected and Literature-based
Data
Approach 2, like Approach 1, is a forward QMRA; the difference between Approach 1 and Approach 2
is how the pathogen densities are derived for input into the QMRA and how the pathogen densities are
adjusted for abundance in the source and human infectivity. In Approach 2, pathogen density for each
reference pathogen is estimated by adjusting literature-based data on pathogen density in the source
by the proportion of measured FIB density (from Step 2) and literature-based data on FIB density in the
source (See Section 2.2.2.4.3 Eq. 2A). This calculation provides an upper bound for the level of
pathogens in the water because pathogen attenuation via environmental fate and transport is not
directly considered. The estimated pathogen densities, like Approach 1, are then input into a forward
QMRA to estimate human health risk from recreational exposure.
99
-------
The prevalence and abundance of reference pathogens in selected animal fecal waste can be found in
Table 2-2 (Section 2.1.1.1). Densities of FIB in selected animal fecal waste can be found in Table 2-7
(Section 2.1.2.1). Data for other animal source(s) can be found by searching the scientific literature
using database searches, such as Google Scholar, PubMed, and other literature databases, including
subscription-based services. To account for a mixed fecal source scenario, Schoen et al. (2011) describe
an approach to adjust the proportion of FIB in a waterbody that is attributable to a specific source. This
term (pf/B) is included in Equation 2A. The sanitary survey supports the estimation of each source
contribution, thus an estimate of FIB from each source can be developed. For example, if the sanitary
survey and/or fecal source identification from the water quality study indicates that less than 10% of
the fecal loading is from human sources and the rest (90%) is likely from avian and predominantly
seagulls, then the calculation described below is performed once for the human source
(proportion = 0.1) and once for the gull source (proportion = 0.9). The proportions of all the sources
equal 1 (100% of FIB accounted for). If the sanitary survey indicates that there are multiple sources of
FIB, those sources can be grouped together if the source with the highest potential levels of each
reference pathogen is used as the representative group for that source. For example, if avian species
are known to be the major source of FIB and there is a mixture of wild birds and chickens in the
watershed, users can combine the avian sources together from a loading perspective and choose the
reference pathogen and FIB data from the subsource with the highest density. Selecting the highest
density is consistent with a health-protective approach for the Approach 2 analysis. If a single source is
determined to be a predominant loading of nonhuman fecal waste, then this term can be considered
equal to "1.0."
Please note that the FIB data from Step 2 could be collected close to where the source is thought to
contaminate surface water and under representative conditions when the source is thought to
mobilize to surface water. If multiple sources are identified, having data collected proximal to the
source loadings and data collected from a study site downstream can support the estimation of source
contribution.
Once the density of each reference pathogen is determined, the user can proceed with the calculations
for dose, probability of infection, and probability of illness. The output of Equation 2A is used in
Equation 2B to calculate the dose for each reference pathogen. The calculation of dose in Approach 2
is slightly different than in Approach 1. In Approach 1, there is multiplication of pathogen density by
ingested volume of water. In Approach 2, Equation 2B accounts for the density of pathogens by
including the fraction of pathogens that are infectious to humans (Section 2.1.1.1, Table 2-4) and the
prevalence of animals that harbor the pathogen (percentage infected) (Section 2.1.1.1, Table 2-3). The
volume of water incidentally ingested will be the same as used for Approach 1. The subsequent steps
of dose response and infections that lead to illness are the same in Approaches 1 and 2. Users can refer
to the description of Approach 1 above for these steps of the forward QMRA.
The example forward QMRA R code in Appendix F performs forward QMRA for a hypothetical
waterbody that contains a 100% gull fecal source. The parameters described above (from Tables 2-2, 2-
3, 2-4, and 2-7) for gulls are used in the R code in Appendix F. The dose-response relationships
(Table 2-5) and probability of infection leading to illness (Table 2-6) for Campylobacter and Salmonella
are included in the R code in Appendix F. Users supply a spreadsheet of input data that contain
enterococci densities in CFUs per liter (L). The estimated mean and median NGI per 1,000 recreators
are outputs.
100
-------
3.3.3 Interpreting Results
Sanitary survey information and water quality data were used to conduct a QMRA estimating risk at
the study location receiving fecal contamination from one or more sources. The output from the
forward QMRA analyses can be used to address the following question: "What is the potential human
health risk from recreational exposure to water at the study location contaminated by the fecal
source(s) in the watershed?" The transparent comparison of the QMRA results to current water quality
recommendations and target illness rates or the current WQS constitutes an important decision point.
Understanding how the results compare to the EPA recommendations will inform the decision about
whether to move forward with deriving alternative criteria.
In this section, the results of the QMRAs from Approaches 1 and 2 and from the water quality
distribution will be compared to the EPA's 2012 recommendations for water quality and target illness
rate. Table 3-1 outlines the different steps that can be taken based on the results of the comparison.
Note that the case where the water quality level at the study location is below the recommended
water quality is included in Table 3-1 for completeness. Although any location could be evaluated for
alternative criteria, the EPA expects that most sites having water quality levels below current
recommendations would not be evaluated for alternative criteria; instead, the EPA-recommended
criteria or the current applicable WQS would apply. Waters affected by secondary-treated and
disinfected wastewater effluent are a notable exception to this point. WWTP effluent-affected waters
represent a condition where water quality can meet the recommended criteria; however, elevated
illness can exist in downstream waters due to the differential treatment efficacy between culturable
FIB and pathogens such as enteric viruses and protozoa. This combination of outcomes is beyond the
scope of this TSM because this document focuses on adjusting the 2012 RWQC water quality values for
less risky nonhuman fecal sources. However, the same tools described in this TSM (e.g., sanitary
characterization and QMRA) could be used to further characterize risk in high-risk scenarios but are not
the focus here. At this point in the TSM process, it is assumed that waters affected by substantial
inputs of human fecal contamination, such as treated human wastewater effluent, are not being
considered for alternative culturable FIB-based criteria based on predominantly nonhuman fecal
sources.
Table 3-1. Comparing forward QMRA and water quality monitoring results to the EPA's 2012 RWQC.
Relationship of water quality and
target illness recommendations
Monitoring results are:
< Recommended GM
> Recommended GM
QMRA results < target illness rate
National RWQC or applicable WQS may
apply, or alternative criteria could be
derived if waterbody is predominantly
affected by nonhuman fecal sources
(Go to Step 4).
Go to Step 4.
QMRA results > target illness rate
Beyond the scope of this TSM.
National RWQC or applicable WQS may
apply. Relative QMRA if nonhuman
fecal sources predominate.
101
-------
Together, the results from Approaches 1 and 2 above can result in a range of estimated risks (i.e., the
range between the median estimates) for a given nonhuman fecal source. The range results from the
assumptions, including pathogen loading and recreational exposure, applied in each approach. For
example, pathogen densities in water can be characterized using monitoring data from a waterbody
some distance downstream of a source or from water upstream near where the source is thought to
enter the water. Although the EPA recommends that monitoring data reflect the hazardous condition
that results in the source entering the water, pathogen loadings can be associated with variability, and
monitoring results below the LOQ of the enumeration method can occur. Approach 2 does not account
for pathogen fate and transport and can be considered a "direct deposition" scenario. It is expected
that Approach 2 results could be at the higher end of potential risk from the nonhuman fecal source.
Taken together, QMRA results based on data from both Approaches 1 and 2 provide valuable insight
for understanding the potential human health risks in the watershed.
There are three key outcome categories to consider for a range of risk estimates resulting from
Approaches 1 and 2 compared to the EPA's recommended target illness range: (1) the range is wholly
below the EPA's recommended target illness level; (2) the range straddles the target illness level, or
(3) the range is wholly above the target illness level. When the range is below, users can proceed to
Step 4. If the comparison of results indicates "Go to Step 4" (Table 3-1), and the other lines of evidence
from the sanitary characterization support predominantly nonhuman fecal sources, proceeding to
Step 4 to calculate a GM value that corresponds to the target illness rate can be substantiated.
As shown in Figure 3-5, the hypothetical waterbodies with pig and chicken fecal sources have
enterococci levels (36,000 enterococci CFU per 100 mL and 1,800 enterococci CFU per 100 mL for pigs
and chickens, respectively) greater than the RWQC recommended GM of 35 enterococci CFU per
100 mL. Looking at Table 3-1, they are in the column where monitoring results are "> Recommended
GM." Next, the predicted illness rates are 15 illnesses and 18 illnesses per 1,000 recreation events for
pig- and chicken-impacted waterbodies, respectively. These predicted illness rates are below the
RWQC associated target illness rate of 36 NGI per 1,000 recreators (row in Table 3-1: "QMRA
results < target illness rate"). Considering both pieces of information—waterbody FIB GM and
predicted illness level— and referring to Table 3-1, these two nonhuman source scenarios fall in the
upper right quadrant of the table. For these two scenarios, the user could proceed to Step 4 of this
TSM to derive criteria.
When the range straddles the recommended target illness level, users can apply their best professional
judgment to determine whether to go directly to Step 4 or decide to perform a relative QMRA. For
example, if the recommended target illness rate is relatively close to the upper end of the estimated
range, then the decision to move on to Step 4 is based on expert judgment. Approach 2 contains
health-protective assumptions that may overestimate actual risk at a specific location. If the target
illness rate is closer to the bottom end of the estimate risk range, users can decide to conduct the
relative QMRA before going on to Step 4 to provide substantiation that the waterbody would be
protective at the EPA's recommended level of water quality.
102
-------
If a comparison of results demonstrates both water quality and estimated illness are above the
recommended GM and the associated target illness rate, then there are two choices:
1. Apply the EPA's RWQC recommendation or the applicable WQS.
2. If the evidence collected indicates that nonhuman fecal sources predominate, then a relative
QMRA could be conducted to understand what the estimated target illness rate would be if the
waterbody met the EPA's recommended criteria value.
As shown in Figure 3-5, the hypothetical waterbody with cattle as a fecal source has enterococci levels
greater than 14,000 CFU per 100 mL compared to the RWQC recommended GM 35 enterococci CFU
per 100 mL. Looking at Table 3-1, this waterbody falls in the column where monitoring results are
"> Recommended GM." The predicted illness rate for the cattle-impacted simulated waterbody is
46 NGI per 1,000 recreators, above the RWQC target illness rate of 36 NGI per 1,000 recreators. This
corresponds to Table 3-1 row: "QMRA results > target illness rate." Taking both these pieces of
information—waterbody FIB GM and predicted illness level—and referring to Table 3-1, this example
cattle-impacted waterbody falls in the lower right quadrant of the table. In this situation, the TSM user
can apply the EPA's recommended RWQC, or a relative QMRA can be conducted to determine if the
waterbody met the EPA's target illness rate at the RWQC recommended GM (in this case,
35 enterococci CFU per 100 mL).
A relative QMRA complements a forward QMRA by estimating risks from specific fecal sources at a
specified FIB density (Soller et al., 2015). The specified FIB density would be one of the EPA-
recommended values, so this analysis estimates the potential risk for the nonhuman source relative to
the EPA's RWQC. The relative QMRA process is diagrammed in Section 2.2.2.4.4. It addresses the
following question: "What would the predicted illness level be if the waterbody water quality was at
the EPA's recommended criteria value (e.g., 35 enterococci CFU per 100 mL)?" For example, if
monitored water quality results from the study location is a GM of 1,000 enterococci CFU per 100 mL
and the QMRA-estimated median illness level is 40 NGI per 1,000 recreators, and assuming the choice
of target illness rate was 36 NGI per 1,000 recreators, the relative QMRA step would predict the illness
level for the waterbody studied if the GM were 35 enterococci CFU per 100 mL.24 This comparison can
demonstrate whether the waterbody is protected at the EPA-recommended target illness rate. If the
relative QMRA step demonstrates that the predicted illness rate is below the EPA's recommended
target illness rate, users can decide to proceed to the next step. Note that the results for the forward
QMRA where both water quality and estimated illness are above the EPA recommendations may
indicate that the study site characteristics and/or conditions could be substantially affecting water
quality and risk estimates, which also may limit the generalization of derived alternative criteria to
other locations where nonhuman sources predominate. In the example discussed above (Figure 3-5),
pathogen mobilization was induced via simulated rainfall at a rate of 6.75 centimeters (cm)/hour
(Soller et al., 2015). These conditions may not be reflective of other locations.
The example discussed above in Section 3.3.1 (Soller et al., 2015) includes a relative QMRA step for
cattle, pig, and chicken fecal sources, and it is useful for illustrating relative risk from nonhuman
sources compared to the EPA's RWQC recommendations. In this example, the probability of illness for
24 The recommended health goal associated with the national 2012 RWQC is 32 or 36 NGI per 1,000 primary contact
recreation events. Enterococci at 35 CFU per 100 mL is associated with 36 NGI per 1,000 primary contact recreation events.
Recreators are those who recreate on or in the water.
103
-------
each source is expressed in terms of the FIB (£. coli or enterococci) from each source to the target
illness rate associated with the EPA's RWQC (Figure 3-6). Note that in the forward QMRA, each source
was associated with FIB levels well above the RWQC recommendations, although the risk of illness was
lower for pigs and chickens. In Figure 3-6, the risk of illness from each source is compared as if the
water were meeting the RWQC. Although QMRA typically focuses on risk from exposure to a pathogen,
<~n
IS)
CD
_C
o
>
-M
_Q
03
_Q
O
C?
ST
E-8 cattle enterococci
pig enterococci
^ chicken enterococci
cattle E. coli
@ pig E. coli
[£E) chicken E. coli
oV
fS>
>
^*
Figure 3-6. Box and whisker plot displaying the relative QMRA probability of illness across fecal
contamination from different species (from Soller et al. [2015]). The line in the middle of each box is
the median. The upper and lower edges of each box are the 75th and 25th percentiles. The bars
(whiskers) extend to the 90th and 10th percentiles. The diamonds are at the 95th and
5th percentiles. The dotted line represents the EPA's recommended target illness rate of 36 NGI per
1,000 recreators.
104
-------
expressing the risk in terms of FIB levels is helpful for comparing to the EPA's recommendations and/or
existing state WQS. For a waterbody affected by cattle fecal material, the predicted illness rate is
30 times lower (~1.1 NGI per 1,000 recreators) compared to the EPA target illness rate. For a pig-
impacted waterbody, the predicted illness rate is 65 times lower (~0.5 NGI per 1,000 recreators)
compared to the EPA target illness rate. For a chicken-impacted waterbody, the predicted illness rate is
25 times lower (~1.3 NGI per 1,000 recreators) compared to the EPA target illness rate (Soller et al.,
2015).
Given the results of the relative QMRA, namely that the predicted level of illness for the hypothetical
cattle-impacted waterbody is 1.1 NGI per 1,000 recreators (if the enterococci levels were at 35 CFU per
100 mL), this example could proceed to Step 4 of the TSM. It meets the condition that the predicted
level of illness is below 36 NGI per 1,000 recreators. If the predicted level of illness had been above
36 NGI per 1,000 recreators, then users are advised to apply the nationally recommended RWQC. In
this case, users may want to consider characterizing another nonhuman fecal source-affected
waterbody for the development of alternative WQC.
3.4 Step 4: Derive Site-Specific Alternative Criteria
Step 4 describes the derivation of alternative WQC (i.e., GM, STV, and
BAV) that adjust the EPA's 2012 RWQC for waters where nonhuman
fecal sources predominate. Users need to choose one of the two
illness rates recommended by the EPA to support the derivation (Text
Box 3-6). Examples presented in Step 4 incorporate the 36 NGI per
1,000 recreators target illness rate because this benchmark facilitates
comparisons to many existing state WQS. The methods and
approaches discussed in Step 4 (see Figure 3-7) would apply equally
well if the 32 NGI per 1,000 recreator rate was selected.
The target illness rate chosen is a policy decision that should be clearly described in the documentation
for the WQS submission. When evaluating WQC submissions, the EPA may return submissions to
request additional information or to disapprove submissions that do not document sufficient detail (40
Code of Federal Regulations [CFR] 131.5 and 131.6) or scientific rationale.
Most commonly, QMRAs are conducted using microbial density as an input, and the output is the
estimated risk level, as was discussed for Approaches 1 and 2 in Step 3. In contrast to a forward QMRA,
a "reverse" QMRA considers associations in the opposite direction by starting with a target illness rate
and then calculating a microbial density. The same QMRA parameters and values are used as described
above in Step 3: (1) the volume of water ingested during swimming; (2) the density of pathogens in the
specific sources; (3) the mathematical dose-response relationships and the parameter values for each
reference pathogen; and (4) the conditional probability of illness given infection for different
pathogens. Reverse QMRA is discussed in more detail in Section 2.2.2.4.5.
The EPA and others have conducted and documented QMRAs for scenarios consisting of waters
affected by human, cattle, pig, chicken, gull, and nonpathogenic fecal sources, and other nonhuman
and mixed fecal discharges to a watershed (U.S. EPA, 2010a; Schoen and Ashbolt, 2010; Soller et al.,
2010a,b, 2014, 2015, 2016; Schoen et al., 2011; McBride et al., 2013). Some QMRA analyses have
considered single nonhuman sources to estimate relative risks compared to human sources (U.S. EPA,
Text Box 3-6. Decision
Points Step 4
Choose one of the EPA's
target illness levels or a lower
level. The magnitude,
duration, and frequency are
scientifically defensible and
protective of the use.
105
-------
2010a; Soller et al., 2010). Other QMRAs have
included mixed fecal contribution scenarios
consisting of human and nonhuman fecal sources
(Schoen and Ashbolt, 2010; Schoen et al., 2011;
Soller et al., 2014). Mixed source contributions in
a watershed may be the more commonly
encountered scenario. An important result noted
in the QMRAs evaluating mixed fecal sources was
that the risks are determined predominantly by
the proportion of the contamination source with
the greatest ability to cause human infection, not
necessarily the source that contributes the
greatest number of FIB (Schoen et al., 2011; Soller
et al., 2014).
When human sources are a component of the
mixed fecal loading, the viral pathogens in human
waste can pose the highest potential of
gastrointestinal illness to recreators. For example,
Soller et al. (2014) simulated the influence of
multiple sources of culturable enterococci in a
recreational water on the illness potential those
sources present. Combinations of human and nonhuman fecal sources were considered and varied
from zero to 100% of the mixture. Results demonstrated that the estimated risks of gastrointestinal
illness from recreational exposures to fecal mixtures were up to 50% lower for mixtures with an
approximately 30% human component compared to fecal loadings of 100% human as measured by
enterococci (Figure 3-8). In this example, gull, pig, or chicken sources with a low human contribution
(i.e., less than the approximate 30% contribution) resulted in numerically higher, but equivalently
protective, enterococci levels at the 36 NGI per 1,000 recreators target illness rate. Nonpathogenic
sources include autochthonous enterococci and E. coli (Byappanahalli et al., 2006).
Please note that the results presented in Figure 3-8 represent reverse QMRA results based on the
information available when the analysis was conducted. The information included parameter values,
some of which have been updated as newer science has become available (e.g., dose responses for
norovirus and other enteric pathogens and incidental ingestion data). Figure 3-8 is presented as a
demonstration based on data from three nonhuman sources of higher FIB values possible at the same
target illness rate as the EPA recommends. Note the trend of decreasing enterococci values as the
human contribution increases relative to the nonhuman contribution. Results would support a science
policy decision of approximately 30% human contribution (i.e., FIB values at this percentage are
approximately a half log higher than at 100% human) as an inflection point above which risks from
human sources predominate the risk profile.
Step 4: Derive Site-Specific Alternative Criteria,
Conduct Reverse QMRA
(see Appendix H for example QMRA code)
Calculate GM
Derive STV and BAV
(Section 3.4.1)
(Section 3.4.2)
\
/
Figure 3-7. Calculate the GM associated with
the health-based goal.
106
-------
400
350
300
E 250
O
o
D
u_
U
GQ
C£
u
u
O
200
150
53 100
50
i
/
O Pig
~ Chicken
i
A Gull
o
X Non-pathogenic
RWQC 35 enterocci CFU/100 mL
I
s
*
~
~
~
!
i
IS
SI
1£
3
si
1£
10 20 30 40 50 60 70
Fraction of Human Source Contribution (%)
80 90
100
Figure 3-8. Predicted GM enterococci densities corresponding to 36 NGI per 1,000 recreators for
waters impacted by mixed sources of fecal contamination for three animal species (CFU per 100 mL)
(adapted from Soller et al. [2014]). The output from the reverse QMRA is a FIB density that can be
used as a GM corresponding to the target illness rate and is called a RBT.
Table 3-2 presents selected results shown in Figure 3-8 to help support the interpretation of the effect of
mixed fecal sources (e.g., nonhuman and human fecal contributions) affecting a waterbody. In this
example, the susceptibility of a waterbody to human fecal contamination is "binned" into low, medium,
and high susceptibility, representing human contributions of 33% or less, 34%-67%, and 68% or greater,
respectively. Due to the potential variability associated with estimating levels of FIB based on monitoring,
the "medium" and "high" bins are considered to be similar to the human-dominated scenario
underpinning the 2012 RWQC recommended criteria. Waterbodies where human contributions
represent less than a third of the total fecal loading fall into the "low" human fecal susceptibility bin.
When states decide to adopt new or revised WQC into their WQS, the criteria must be scientifically
defensible and protective of the designated uses of the waterbodies. The EPA's regulation 40 CFR
§131.11(b)(1) provides that "In establishing criteria, states should (1) Establish numerical values based
on (i) 304(a) Guidance; or (ii) 304(a) Guidance modified to reflect site-specific conditions; or (iii) Other
scientifically defensible methods."
Include the same basic elements as the national RWQC—magnitude, duration, and frequency—for the
alternative criteria developed using this TSM. Magnitude is the numeric expression of the maximum
amount of the pollutant that may be present in a waterbody that supports the designated use.
Duration is the period of time over which the magnitude is calculated. The frequency of excursion
describes the maximum number of times the pollutant may be present above the magnitude over the
specified time period (duration). A criterion is set in a WQS such that the combination of magnitude,
duration, and frequency protects the designated use (such as primary contact recreation).
107
-------
Table 3-2. Predicted GM enterococci densities (CFU per 100 mL) in mixed sources of fecal
contamination that correspond to 36 NGI per 1,000 recreators. Values presented are for
illustrative purposes and are not meant to be used as alternative criteria.
Nonhuman source of
fecal contamination
Susceptibility of waterbody to human fecal contamination (percent human contribution)
Low
Medium High
< 10%
<25%
<33%
34%-67% > 67%
Nonhuman source #1°
95
83
76
Nationally recommended RWQC apply
Nonpathogenic source^
350
140
106
Notes:
a. Nonhuman fecal source examples with relative risk evaluations reported in the peer-reviewed scientific literature include cow, chicken
and pig manures, bird/gull feces, and mixed human and nonhuman fecal contributions. Contributions can be "fresh" and directly
deposited in a water body, deposited on land due to livestock grazing, or may be aged manures applied at agronomic rates to fields.
b. Nonpathogenic sources include autochthonous or other sources of enterococci and E. coli not directly associated with the loading of
feces to a waterbody.
The magnitude is expressed in terms of a GM and a STV. The EPA's 2012 RWQC recommends
expressing the criteria magnitude as a GM value corresponding to the 50th percentile of the water
quality distribution and an STV corresponding to the 90th percentile of the same water quality
distribution. In the RWQC, two sets of GM and STV are recommended, each associated with a specific
target illness level. The GM calculated in Step 4 is related to one or the other target illness rate in the
2012 RWQC. For duration and frequency, the EPA recommends that the waterbody GM not be greater
than the selected GM magnitude in any 30-day interval. The state would need to define the duration in
their proposed WQS. There should not be greater than a 10% excursion frequency of the selected STV
magnitude in the same 30-day interval. The EPA does not specify sampling frequency or the number of
samples required. States can determine the sampling frequency and the number of samples that
provide the degree of statistical confidence they want. Some resources for evaluating sample sizes
include Smeets et al. (2010) and Helsel (2012).
Although states have flexibility in selecting the construct of their WQS, the submissions will need to
clearly and transparently document that the standards are scientifically defensible and protective of
the designated use (see Appendix D for the WQS checklist). Supporting documentation for any
deviations from the EPA's recommendations is recommended. WQS submissions adopting the EPA's
recommended construct will be easier to evaluate relative to the 2012 RWQC recommendations.
3.4.1 Reverse QMRA and Calculating the GM
As discussed in Section 2.2.2.4.5, a reverse QMRA approach is used to estimate the FIB levels for
waterbodies predominated by nonhuman fecal sources, including human and animal fecal
contamination mixtures, at a specified target illness rate. A reverse QMRA approach can be used to
estimate the level of a pathogen that results in an estimated or observed level of illness (e.g., what
level of norovirus is needed to cause the level of gastrointestinal illness reported in a waterborne
outbreak). In Soller et al. (2010b), densities of each reference pathogen were computed corresponding
to the illness reported on each day in the NEEAR epidemiological study, resulting in a pathogen profile
for the fecal loading at the study locations. The reverse QMRA approach discussed in this TSM is
computationally more involved because the relative level of FIB from each fecal source is considered in
the derivation of the FIB GM anchored to the target illness level selected. An example is provided to
108
-------
aid the user in conducting the reverse QMRA analysis for this step. At the end of this step, a GM for
enterococci or E. coli that corresponds to either 32 or 36 NGI per 1,000 recreators will be derived.
3.4.1.1 Reverse QMRA modeling description and assumptions
This TSM provides a case study that simulates a hypothetical scenario of a waterbody affected by a
mixture of gull and human fecal sources in different proportions (Appendix G). In this scenario, a
reverse QMRA analysis is conducted to derive an alternative FIB GM and is presented as an illustrative
example of the process described in this TSM. The gull and human fecal inputs are varied in
proportions between 10% and 100% human. The reverse QMRA case study is summarized in this
section, with full documentation described in Appendix G and the associated reverse QMRA Python
code provided in Appendix H.
The reverse QMRA uses the following inputs to derive a GM for enterococci anchored to a
recommended target illness rate for each of the fecal source mixtures:
• Target illness rate: A target illness rate of 36 NGI per 1,000 recreators was recommended in the
2012 RWQC (U.S. EPA, 2012).
• Ingestion volume: A median general population ingestion volume of 19 mL per recreational
event (Dufour et al., 2017).
• Human fecal contamination: Norovirus was selected as the primary index pathogen. Norovirus
and enterococci levels in human sewage are literature-based (Table 2-2).
• Gull fecal contamination: Campylobacter jejuni and Salmonella were selected as reference
pathogens. Pathogen and enterococci levels in gull feces are literature-based (Table 2-2).
• Dose-response: The infectivity of the three reference pathogens was characterized by
mathematical dose-response relationships (Table 2-5). The probability of infection leading to
illness was characterized as a constant for Salmonella and using a hazard model function for
Campylobacter and norovirus (Table 2-6).
• Fecal Source Apportionment: The percentage of enterococci contributed by human fecal
sources was modeled separately at 10%, 20%, 25%, 30%, 33%, 50%, 60%, 67%, 75%, 90%, and
100% to reflect a range of possibilities providing context for the results.
• Simulated human mixture: To anchor the reverse QMRA to the 2012 RWQC a simulated human
source was created that provides model results of 35 enterococci CFU per 100 mL at the illness
level of 36 NGI per 1,000 recreators. The simulated human mixture is comprised of secondary
treated effluent with some raw sewage. In this mixture, the effluent contributes pathogens
(i.e., risk of gastrointestinal illness), and the raw sewage primarily contributes the FIB (See
Schoen et al., 2011 and Soller et al., 2014 for more details).
Generally, a reverse QMRA begins with a target illness level and ends with the quantity of stressor
associated with the input target illness level. In this reverse QMRA, the pathogen (stressor) density is
linked to an enterococci level, so the output is a GM density of enterococci that is associated with the
input target illness rate. Equations 3A through 3L found in Section 2.2.2.4.5 are used to combine the
information in the bullets above. Conceptually the reverse QMRA is shown in Figure 2-10. An initial
109
-------
enterococci density estimate is provided to start the iterative process of calculating the FIB level
associated with the target illness level, which is shown as the loop between the diamond and the
beginning of the model in Figure 2-10. The density of each pathogen in the human or nonhuman
source is used to compute a dose. The dose-response functions are used to compute the probability of
infection, given the dose and a recreational exposure, which is then used to compute an illness rate
based on the probability of infection leading to illnesses. The computed probability of illness is then
compared to the target illness rate. If they do not match, then the process is repeated in an iterative
fashion until the computed illness rate matches the input target illness rate. Once the computed illness
rate matches the target illness rate, the corresponding enterococci density is considered the RBT for
enterococci CFU per 100 mL.
The reverse QMRA Python code in Appendix H is annotated so users can follow the purpose of each
line of code.25 Briefly, the Python code begins with defining the target illness rate and the ingestion
volume point estimate. The input values of the density of the enterococci for each source are then
assigned to variables in the code (as shown in Appendix G: Table G-3). Input values for the prevalence
of each pathogen in each source are assigned in the next section of the code, followed by the fraction
of infectious pathogens in each source. The structure of the reverse QMRA requires iterative solving,
which is where an initial value of FIB density corresponding to the target risk is estimated and used by
an iterative formula that specifies the mathematical relationship between FIB density and risk
consistent with the input assumptions and generates a sequence of improving approximate solutions
for the FIB RBT. In the case study example, an enterococci level of 50 CFU per mL (500 CFU per L in the
code) was provided as an initial estimate of the enterococci density corresponding to the target risk
rate for each source. This value is refined by computer code through numerical iteration to derive the
specific enterococci density for each source and at each simulated human source contribution level,
resulting in 11 fecal mixture proportions (i.e., model convergence). The code includes an internal
forward QMRA to confirm that the model has reached convergence for each scenario.
The output of the reverse QMRA Python code (Appendix H) is a table that contains enterococci
densities (CFU per 100 mL) for the 11 apportionment scenarios for each of the four sources included in
the code (gulls, swine, chicken, and nonpathogenic). The reverse QMRA code example implements
TSM Equations 3A through 3L found in Section 2.2.2.4.5.
Figure 3-9 displays a graphic representation of the results from the reverse QMRA presented in
Appendix G. Each point represents the enterococci GM corresponding to the 36 NGI per
1,000 recreator target illness rate for the various proportional gull and human fecal mixtures. The
dotted line on the graph represents 35 enterococci CFU per 100 mL. At 100% human contribution, the
model output is anchored at 35 enterococci CFU per 100 mL and 36 NGI per 1,000 recreators. As the
percentage of human sources decreases and the percentage of gull fecal sources correspondingly
increases, the level of enterococci that corresponds to the target illness rate increases. The
hypothetical waterbody with 10% human fecal source and 90% gull source results in 350 enterococci
CFU per 100 mL at the recommended target illness rate.
25 The python code in Appendix H also includes the parameter information for swine and chicken fecal sources (results not
shown).
110
-------
c
O Enterococci RBT for General
I • , ¦ lis' n >
;corr
:ion
o
o
3
u_ -
U
II™
DQ -
OCT ^
'u
u
O 1
u J
O
j™.
cu ^
E 3
-O-
Fraction of Human Source Contribution
-JO
Figure 3-9. Reverse QMRA results for the gull case study. Each data point on the graph represents the
enterococci GM associated with the 36 NGI per 1,000 recreator target illness rate for varying
proportions of gull and human fecal mixtures.
Table 3-3 presents the threshold enterococci level corresponding to the target risk level (36 NGI per
1,000 recreators) for two different source mixtures (human-gull mixture and human-nonpathogenic
mixture) considered in the low susceptibility of human influence (see Table 3-2 for more detail). The
gull row shows the numerical values for three of the points in Figure 3-9 above. The nonpathogenic
source was not plotted in Figure 3-9 because the points overlap the gull results. The levels of human
contribution in Table 3-3 were selected because these levels are conceptually straightforward to
visualize and estimate based on the qualitative and quantitative information gathered as part of the
Sanitary Characterization (Step 1, Section 3.1, and Step 2, Section 3.2). Additionally, a one-third
proportion of human fecal loading results in an approximate half log increase in the GM enterococci
value, while a 10% human loading corresponds to a one log increase in the GM enterococci value.
The values presented in Table 3-3 represent the output from the reverse QMRA code using the data
inputs identified above providing a GM threshold for the hypothetical waterbody associated with one
of the EPA's recommended target illness rates. In this example, if the sanitary characterization
identified that the waterbody was receiving fecal loading that was predominantly gull, with a quarter
of the loading from humans, an enterococci GM value of 140 enterococci CFU per 100 mL would
correspond to the chosen target illness rate. The enterococci values in Table 3-3 illustrate the relative
risk difference between gull and human fecal inputs but should not be interpreted as criteria
recommendations for a specific nonhuman source. These values are provided as examples only.
Ill
-------
Table 3-3. Enterococci values corresponding to the target illness rate for selected nonhuman and
human fecal mixtures.
Human source contribution3
Nonhuman source
< 10%
<25%
<33%
Gull
349
140
106
Nonpathogenic
350
140
106
Notes:
a. The three percentages of human source contribution are associated with a low susceptibility for human fecal influence. Output from
Case Study Example in Appendix G, enterococci RBT (CFU/100 mL) corresponding to the target illness rate.
To complete criteria derivation, the STV, duration, and frequency components are also needed.
Calculation of the STV and BAV corresponding to the GM is discussed in Section 3.4.2 below. In this
example, the STV and BAV values corresponding to an enterococci GM of 140 CFU per 100 mL are 513
(or could be rounded to 510) and 277 (or could be rounded to 280) enterococci CFU per 100 mL,
respectively (Appendix G).
3.4.2 Derive STV and BAV
STV. Calculating the STV requires having the logSD26 of the data for the indicator in the waterbody. The
EPA based the 2012 RWQC STV values on the logSDs of FIB they observed during the EPA
epidemiological studies. The STV represents the 90th percentile of the water quality distribution
centered on the recommended GM values, so the frequency of samples above the STV can be up to
10%, and the waterbody being measured is still considered in attainment. At the NEEAR study sites, the
pooled estimate for the logSD of culturable enterococci was 0.44 CFU per 100 mL (U.S. EPA, 2012).27
The EPA used the following equation to calculate the STV at the 90th percentile for the corresponding
GM value for enterococci or E. coli in the RWQC:
• Assuming a logSD28 of 0.44, Equation 4A gives the 90th percentile of the distribution of a
waterbody with a GM:
STV = antilogio [logio(GM) + Z90x 0.44] [Eq. 4A]
Where:
STV is the statistical threshold value.
Z90 = 1.28 is the 90th percentile of the standard normal distribution (unitless).
GM is the geometric mean of the waterbody in units CFU per 100 mL.
0.44 (units enterococci CFU per 100 mL) is the logSD of the data.
26 The logSD is the SD of the base 10 log of the data. The antilog of the logSD would be the geometric SD.
27 The EPA's 2012 RWQC page 40.
28 The EPA based the SDs on water quality data from the EPA epidemiological studies. The logSD for the 1986 freshwater E.
coli studies was 0.40 CFU per 100 mL; in the NEEAR studies, using the marine and freshwater beaches combined (not
including Surfside or Boqueron), the logSD for culturable enterococci was 0.44 CFU per 100 mL.
112
-------
Different waterbodies can have different levels of water quality variability. Use either the logSD that
the EPA used in the RWQC or calculate the logSD for the FIB data collected in Step 2. The logSD can be
calculated using various software packages, including MS Excel. The first step is to log-transform the
data collected. In Excel, for example, use the "STDEV" or "STDEV.S" functions to calculate the SD of the
log-transformed values. This value can replace the 0.44 value in the above equation.
BAV. The EPA provided recommended BAVs, which correspond to the 75th percentile of the water
quality distribution. The BAV is calculated using the same SD used to calculate the STV. BAVs are not
used for determining use attainment. The BAV provides a decision point for beach managers based on
a single monitoring sample result for use in beach notification programs. Recreational exposure risks
are not linked to, nor predicted by, any single monitoring sample results, so the BAV represents a
precautionary tool on any given day when the GM is unknown. Because the BAV represents the 75th
percentile, water quality can be expected to exceed this value 25% of the time and the waterbody
being monitored would still be attaining the use.
• Assuming a logSD of 0.44, Equation 4B gives the 75th percentile of the distribution of a
waterbody with a GM:
BAV = antilogio [logio(GM) + Z75X 0.44] [Eq. 4B]
Where:
STV is the statistical threshold value.
Z75 = 0.675 is the 75th percentile of the standard normal distribution (unitless).
GM is the geometric mean of the waterbody in units CFU per 100 mL.
0.44 (units enterococci CFU per 100 mL) is the logSD of the data.
113
-------
4.0 Effects Characterization
4.1 Methods to Evaluate Health-Protective Values for Different Susceptible
Subpopulations
Susceptible subpopulations can be characterized by intrinsic traits (e.g., age, gender, genetic traits,
immune status and other biological factors) and acquired traits (e.g., exposure, behaviors), which have
the potential to influence both the exposure profile and the health outcomes in these subgroups (U.S.
EPA, 2000c, 2014b). The 2012 RWQC recommendations are based on studies reporting AGI from event-
based recreational exposure to ambient waters receiving human fecal contamination (U.S. EPA, 2012).
Reported recreational exposures are characterized by either interviewing people who participate in
recreational activities (Schets et al., 2011; Deflorio-Barker et al., 2017; Wade et al., 2006, 2008, 2010,
2022) or by estimating incidental ingestion in controlled settings (Dorevitch et al., 2011; Dufour et al.,
2017). Additionally, because exposure can be influenced by behaviors, some studies have
characterized time spent for various recreational activities (Deflorio-Barker et al., 2017; Ferguson et al.,
2019, 2021). Adverse health outcomes following exposure to feces-contaminated recreational water
are characterized in three main types of health studies: epidemiological, QMRA, and outbreak
compilations. The etiologic agent(s) of disease are not often identified in epidemiological studies and
outbreak reports. Typically, the risk from fecal contamination, and hence the pathogens that may be
present in feces, is expressed as statistical association with a level of a nonpathogenic surrogate, such
as FIB. The statistical association between illness and level of surrogate characterized by an
epidemiological study is dependent on the pathogen profile present, which is dependent on the source
of feces, the mass loading of feces and the dynamics of contamination (see Section 2.1).
4.1.1 Considering and Identifying Susceptible Subpopulations
Considering subpopulations is important because there are groups that are typically considered more
susceptible to infection and illness, including more severe illness outcomes, compared to healthy
adults: pregnant women; neonates and children; people over 65 years old; individuals residing in
nursing homes or related care facilities; and cancer, organ transplant, and AIDS patients (Haas et al.,
2014). Subgroup differentiation is not necessary unless there is evidence for relevant differences
between the subgroups (U.S. EPA, 2014b). QMRA documentation presents a scientific rationale for
dividing subgroups as well as data that directly pertain to that subgroup or could be adjusted to
address that subgroup. When considering risks from pathogen exposures, factors that influence the
identification of a "susceptible subpopulation" include the exposure profile, susceptibility to infection,
susceptibility to illness given infection, and severity of illness. Exposure is influenced by factors such as
biology and behavior. Susceptibility to infection and illness can be affected by an individual's immune
and gastrointestinal status or by a genetic host factor such as a pathogen receptor, where individuals
without the receptor are not susceptible to infection and individuals with the receptor are susceptible
to infection upon exposure to a specific pathogen (Nordgren and Svensson, 2019). Infection, illness,
and illness severity are also influenced by the virulence, pathogenicity, and host specificity of the
pathogen (U.S. EPA, 2014b).
People with compromised immune systems and/or concurrent health conditions are considered a high-
risk subgroup because they can be more susceptible to infection and subsequent illness and can
experience more severe illness symptoms and health outcomes, including death (U.S. EPA, 2014b). Risk
communication for this subgroup regarding recreational exposure to ambient waters focuses on
114
-------
preventing exposure because it is not possible to achieve a "safe" exposure to ambient water that may
contain feces-associated and/or naturally occurring pathogens.
Age groups are considered as subpopulations in the context of acute risk of illness from exposure to
fecal-associated pathogens because health study results published in the scientific literature
demonstrate a distinct difference in exposure, illness burden and potential severity of health outcomes
among broad age groupings (Arnold et al., 2016; Dufour et al., 2017; Deflorio-Barker et al., 2017;
Verhougstraete et al., 2020; Wade et al., 2008, 2022). Risk differences can be influenced by one or
more of the following: (1) children's immunological, digestive, and other bodily systems that are still
developing; (2) children's greater exposure because they ingest more water and breathe more air in
proportion to their body weight than adults; and (3) children's behavior, such as increased time spent
in water and more vigorous activity, that might result in increased exposure in comparison to adults.
Published epidemiological and outbreak information has demonstrated that children have a higher risk
of illness compared to adults when exposed to human fecal contamination in recreational waters
(Arnold et al., 2016; Mosnier et al., 2018; Schets et al., 2018; Sips et al., 2020; Wade et al., 2008, 2020).
Statistically significant increased water ingestion among children 6-10 years old was documented by
Dufour et al. (2017). Modeling the ingestion data together with the time spent recreating
demonstrated that children less than 10 years old exhibit a higher-exposure profile (Deflorio-Barker et
al., 2017). The exposure data support the health study results; together, they demonstrate that
children are more susceptible to gastrointestinal illness in feces-contaminated ambient recreational
waters. Due to the nature of the data available, the age groupings used to characterize risk from fecal
contamination are less granular than the lifestage breakouts typically considered in chemical risk
assessments. In contrast to chemical contaminants in water, the adverse health effects associated with
human exposure to waterborne pathogens have been best documented for event-related (i.e., short-
term, single exposure) rather than chronic exposure over extended periods of time (U.S. EPA, 2014b).
4.1.2 Accounting for Differential Susceptibility
The development of the 2012 RWQC was supported by data collected during epidemiological studies
that included adults and children. In the older epidemiological studies, participants under 19 years old
comprised approximately 45% of swimmers (Dufour, 1984). In the NEEAR study, children aged 10 years
and younger comprised approximately 17%-30% of swimmers at the beach locations studied (U.S. EPA,
2012). According to the U.S. Census data for 2009, children younger than 10 years old make up
approximately 14% of the U.S. population (Census, 2010).
The 2012 RWQC recommendations are based on health data collected for the general population (i.e.,
all study participants). The exposure parameters described in Section 2 of this TSM are based on data
from the general population, and the pathogen dose-response data was largely based on challenge
studies with healthy adults. However, data documenting child-specific parameters are available.
Dufour et al. (2017) and Deflorio-Barker et al. (2017) report exposure data, including incidental
ingestion volumes for younger children, older children, and adults. Arnold et al. (2016) and Wade et al.
(2008, 2010, 2022) report epidemiological-based health relationships for children under 10 years old.
Ferguson et al. (2019) conducted an observational study focused on children ages 1 to 6 years old and
their behavior patterns at recreational marine beaches in Florida and Texas. Ferguson et al. (2021)
used a virtual timing device to quantify real-time, sequential micro-activity pattern data collected from
videos of 120 children at four different beaches.
115
-------
4.1.2.1 Exposure: Ingestion Rate Differences by Lifestage
The EPA's 2019 Recommended Human Health Recreational Ambient Water Quality Criteria (AWQC) or
Swimming Advisories for Microcystins and Cylindrospermopsin (Cyanotoxin AWQC; U.S. EPA, 2019a)
reviewed seven studies to evaluate recreation-associated incidental ingestion (Dufour et al., 2006,
2017; Schijven and de Roda Husman, 2006; Dorevitch et al., 2011; Schets et al., 2011; Suppes et al.,
2014, DeFlorio-Barker et al., 2017). The Cyanotoxin AWQC includes graphical visualizations of ingestion
volumes for different age groups, information on the length of time individuals spend in the water
(duration of exposure), and a calculation of the incidental ingestion rate, which shows that children
ages 6 to 10 years old have an incidental ingestion rate higher than the other age groups characterized.
The results for children from the studies are briefly summarized below.
Dufour et al. (2017) used excreted cyanuric acid to measure incidental ingestion (mL/minute) in ages 6
and above and reported higher ingestion volumes among children 6 to 10 years old compared to
adults. Suppes et al. (2014) used a similar measurement method as Dufour et al. (2006, 2017), with the
addition of videos to estimate time in water, to evaluate the rate of water ingested by 16 children ages
5 to 17 years. They found that children, on average, ingested pool water at a higher rate than adult
participants.
DeFlorio-Barker et al. (2017) combined the incidental ingestion volumes from Dufour et al. (2017) and
the self-reported time spent in the water for 12 cohorts of participants of epidemiological studies to
calculate the volume of water ingested per event. The results of this study corroborate other studies
that demonstrate that, on average, children have higher incidental ingestion than adults when
recreating.
Schets et al. (2011) reported incidental ingestion volumes and durations of recreational events for
children under 15 years old. However, the study did not further divide this cohort into younger
children and older children. The incidental ingestion data for children under 15 years old represent
parental estimates of volumes of freshwater incidentally ingested by their children. The exposure
durations were also parental estimates. The study reported higher incidental ingestion volumes, on
average, for children compared to adults (Schets et al., 2011).
Although these studies used different methodologies and have limitations regarding reporting
information for different age group categories, their results show a similar pattern for children
incidentally ingesting water at a higher rate than adults (U.S. EPA, 2019a).
4.1.2.2 Infection and Illness
Reported health study data show that children (e.g., < 10 years old) experience more illness compared
to adolescents and adults, and associations between water quality and health effects are stronger
compared to adults (Arnold et al., 2016; Verhougstraete et al., 2020; Wade et al., 2022). Arnold et al.
(2016) conducted a retrospective pooled analysis of 13 prospective cohorts at marine and freshwater
beaches in the United States and found that gastroenteritis risk and associated health burden was
higher in young children. Multiple beaches evaluated were affected by human fecal sources, but a few
had little to no identified human fecal inputs (e.g., Malibu, Mission Bay, Surfside). The authors focused
on incident diarrhea, defined as three or more loose or watery stools in 24 hours following exposure,
as the primary adverse health outcome. Gastrointestinal illness, defined consistently with previous
studies, was evaluated as a secondary health outcome. Increases in incident diarrhea were observed as
the level of exposure (body immersion, head immersion, swallowed water) increased compared to
116
-------
nonswimmers with a greater effect observed for children than adults. Children under 4 years old and
5-10 years old had the most water exposure, exhibited stronger associations between levels of water
quality and illness, and accounted for the largest attributable illness burden.
Verhougstraete et al. (2020) developed adjusted risk difference models (excess gastrointestinal illness
per swimming event) for children (< 10 years old) and nonchildren (> 10 years old) and compared the
results of these models to those used to develop WHO's guidelines on recreational water risks. The
authors used epidemiological data published in Lamparelli et al. (2015) and additional enterococci
water quality data from routine monitoring at five sewage-impacted urban beaches in Brazil, polluted
by stormwater runoff, partially treated sewage, and uncontrolled wastewater discharges. The risk
model included a concentration-response function that is site-specific and based on enterococci
(measured using culture methods) from the 1999 epidemiological study (Lamparelli et al., 2015). The
risk models indicated that children less than 10 years old can have twice the risk of gastrointestinal
illness than recreators greater than 10 years old. Elevated enterococci levels resulted in an excess of
96 NGI cases per 1,000 swimming children. In addition, the enterococci levels were higher in Brazil
than in the primary United Kingdom study (Kay et al., 1994) that informed WHO guidelines for
recreational water.
Wade et al. (2022) reported that children who participated in the NEEAR studies were at higher risk of
illness associated with exposure to fecal contamination than adults, as measured by Enterococcus
qPCR. Wade et al. (2022) analyzed epidemiological data from 13 beaches to compare illness risks and
water quality, as measured with both enterococci culture and qPCR, for different age groups, including
exposure levels, beach sites, and health endpoints. Children 12 and under comprised approximately
25% of the study population (total enrollment = 83,452 participants) across the various site categories.
Seventy percent of children had at least some contact with water, and 27% stayed in the water for
60 minutes or more. Gastrointestinal Illness was the most sensitive health endpoint, and the strongest
associations were observed with Enterococcus measured by qPCR at sites with human fecal
contamination. Under several exposure scenarios, odds ratios for NGI among swimmers were higher
among children compared to adolescents and adults. Respiratory symptoms were also associated with
Enterococcus spp. exposures among young children at sites affected by human fecal sources, although
small sample sizes resulted in imprecise estimates for these associations. The highest odds ratios were
among young children at the core NEEAR sites (OR = 2.32, 95% CI = 1.33-4.06). For severe
gastrointestinal illness at the core NEEAR sites, statistically significant associations were observed
among children under 6 years old who stayed in the water 60 minutes or longer with Enterococcus
qPCR CE (8.13, 95% CI = 1.92-36.2) but were not significantly associated with Enterococcus CFU (Wade
et al., 2022).
4.1.2.3 Availability of data on other susceptible subpopulations
A discussion of susceptible subpopulations related to pathogen exposures can be found in the EPA's
Microbial Risk Assessment (MRA) Tools, Methods, and Approaches for Water Media (U.S. EPA, 2014b).
Subgroupings can be delineated by age, immune status, genetic background, pregnancy, nutritional
status, and social, cultural, or age-related behavior traits or geographic location (e.g., environmental
justice [EJ] communities). The young and older adults generally have less resistance to infections.
Children, especially malnourished children, may be more likely to exhibit severe effects of AGI after
exposure to some pathogens (e.g., pathogenic E. coli, some enteric viruses). However, some pathogens
(e.g., Hepatitis A, poliovirus) may cause less clinical illness in children than in adults (Gerba et al.,
117
-------
1996). Age can also contribute to different exposure patterns due to behavior. For example, as
discussed above, children have higher levels of incidental ingestion of water during swimming than
adults (Dufour et al., 2017; DeFlorio-Barker et al., 2017).
Populations considered immunocompromised or immunosuppressed due to recent or concurrent
illness or medical treatment may be defined as subpopulations that risk assessment could address
(Effler et al., 2001). For clarity, definitions of subpopulations included in a risk assessment would
include the criteria used to classify individuals as immunocompromised and could be limited to specific
identifiable types of immune defects. For example, extreme physical or emotional stress can lower
immune competency (Schneiderman et al., 2005; Dhabhar, 2014). The host gastrointestinal
environment can vary in ways that affect pathogens, and innate immunity also plays a role in infection
dynamics (Yoo et al., 2020). Malnourished individuals tend to have weaker immune defenses than
well-nourished individuals (Schaible et al., 2007). Persons with concurrent illness or undergoing
medical treatments may have increased susceptibility (Morris et al., 1997; Dropulic and Lederman,
2016). Genetic background can also affect immune status but may play a larger role in the mechanism
of infection and disease progress (U.S. EPA, 2014b).
Previous exposure can confer limited, short-term, or longer-term protective immunity for some
pathogens (Frost et al., 2005). The converse of this may also be true; when individuals or populations
that have not previously been exposed to particular pathogens, infection and illness rates can be
higher than would otherwise be anticipated. "Traveler's diarrhea" is a well-known observed
phenomenon that exemplifies this type of situation. Pregnancy may cause women to be more
susceptible to a pathogen. For example, Hepatitis E, which causes a self-limiting disease in most
infected persons, can cause up to 20% mortality in women in the third trimester of pregnancy (Jameel,
1999). Pregnancy can also affect behaviors, such as increased water consumption.
Social and cultural behavioral traits primarily affect exposure patterns. For example, a relatively small
proportion of the population is responsible for consuming the majority of raw and partially cooked
shellfish (FDA, 2005). Cultural traditions such as subsistence fishing can increase exposure to water.
Location combined with cultural traditions, such as Asian and Pacific Islander (e.g., Hawaiians)
engagement with the ocean, can influence exposure levels. Tribal exposures may not fall into primary
or secondary contact categories, as lifestyle exposures go beyond both of those categories. Surfers also
have lifestyles that have higher levels of exposure to recreational waters than is normally considered
(Arnold et al., 2017).
EJ is the fair treatment and meaningful involvement of all people, regardless of race, color, national
origin, or income, with respect to the development, implementation, and enforcement of environmental
laws, regulations, and policies.29 Executive Order 12898, Federal Actions to Address Environmental
Justice in Minority Populations and Low-Income Populations, directs federal agencies to:
• Identify and address the disproportionately high and adverse human health or environmental
effects of their actions on minority and low-income populations to the greatest extent
practicable and permitted by law.
• Develop a strategy for implementing EJ.
29 https://www.epa.gov/environmentaliustice
118
-------
• Promote nondiscrimination in federal programs that affect human health and the environment
and provide minority and low-income communities access to public information and public
participation.
EJ communities may have higher proportions of risk factors due to limited access to medical care and
affordable and nutritious food.
4.2 Sensitivity Analyses
A sensitivity analysis is the practice of changing a parameter value in a risk assessment to determine
how big of an effect that parameter has on the results and to better understand the results of the base
analysis. Three sensitivity analyses were performed to evaluate how choosing different parameters or
parameters for different populations affected the QMRA results. The questions these analyses
addressed were:
1. Considering the nonhuman fecal source, seagulls, how does more recently available
information in the scientific literature affect previously published QMRA health modeling
results (Soller et al., 2014)? The analysis in Soller et al. (2014) is compared to an updated QMRA
analysis incorporating revised parameter choices based on available scientific literature and
presented in thisTSM (Section 4.2.1).
2. What effect does the choice of ingestion volume parameter have on QMRA results characterizing
a mixture of human and nonhuman fecal contamination when the ingestion for the general
population is compared to the ingestion volume for children 6 to 10 years old (Section 4.2.2)?
3. Does the inclusion of more recently available dose-response information for norovirus (Teunis
et al., 2020) affect the viral etiology conclusions from Soller et al. (2010a) (Section 4.2.3)?
4.2.1 Compare Gull Analysis from Soller et al. (2014) to this TSM
Soller et al. (2014) reported the predicted risk for a mixture of human and nonhuman (gulls, pigs, or
chickens) fecal contamination in varying proportions from 0%-100% human and predicted enterococci
densities that correspond to the 36 NGI per 1,000 recreators target illness rate in the 2012 RWQC
(these densities are referred to hereafter as RBTs). A recent literature search of relevant QMRA
parameters identified multiple potential updates that were included and discussed in this TSM (Section
2.1). As part of that update, the EPA updated the analysis presented in Soller et al. (2014) to
understand the effect of these changes on the QMRA modeling output. This sensitivity analysis
compares the RBT enterococci values reported for gulls in Soller et al. (2014) to the results from the
Gull Case Study presented in this TSM (Appendix G).
Table 4-1 outlines the parameters that differ between the Soller et al. (2014) analysis and the revised Gull
Case Study. The analysis used a reverse QMRA approach following the process presented in Section
2.2.2.4.5. While the Soller et al. (2014) modeling was originally conducted in the Mathcad computational
environment, the EPA previously converted the Mathcad code to Python to make an open-source code
option available. The converted Python code corresponding to the Soller et al. (2014) analysis is in
Appendix J. The Gull Case Study in this TSM was conducted using Python code, which is available in
Appendix H. While Appendix G comprehensively documents the Gull Case Study, the principal differences
between the Soller et al. (2014) and the Gull Case Study may be broadly summarized as:
119
-------
Table 4-1. Parameters that differ between Soller et al. (2014) and the gull case study in this TSM.
Parameter
Soller et al. (2014)
This TSM (gull case study; Appendix G)
Ingestion volume
19 mL/event
Table 2-8
Density of norovirus in raw
sewage (10" gene copies/L)
Uniform distribution with
Min 3.0
Max 6.0
Log normal distribution
(Table 2-2)
Density of enterococci in gull
feces (10n CFU/L)
Min = 6
Max = 8
Table 2-7
Density off. coli in gull feces
(10n CFU/L)
Min = 5
Max = 9
Table 2-7
Dose-response norovirus
a = 0.04
Table 2-5
P = 0.055
Stochastic z and w intermediate parameters
used to estimate distributions of a and p
Dose-response Campylobacter
a = 0.024
Table 2-5
P = 0.011
Stochastic z and w intermediate parameters
used to estimate distributions of a and p
Infection to illness norovirus
Point estimate 0.6 (60%)
Table 2-6
Stochastic z and w intermediate parameters
used to estimate distribution of r and n
Infection to illness
Campylobacter
r = 2.44 x 10s
nu = 3.63 x 10"9
Table 2-6
Stochastic z and w intermediate parameters
used to estimate distribution of r and n
• Updates to the density of pathogen and indicator levels found in human and gull
contamination, which were implemented by straightforward parameter updates to the 2014
analysis code.
• Updates to the dose-response functions used to describe the relationship between pathogen
ingestion and risk of illness, which required more structural changes to the 2014 analysis code.
Specifically, the new dose-response functions required the intermediate generation of
stochastic estimates (statistical distributions) of the parameters of the dose-response functions,
which were then used to estimate a distribution of risk at the modeled dose. The median of the
risk distribution was thereafter used as a point estimate of the risk corresponding to the
modeled dose.
• Recalculation of the relative contribution of raw human contamination and human effluent to
the human source mix consistent with the new pathogen and indicator densities.
120
-------
The results of the sensitivity analysis are presented in Table 4-2 and Figure 4-1. Note that the results
are so similar that the icons mostly overlap on Figure 4-1. Given these results, it can be concluded that
the updated parameters, including the updated norovirus dose-response function, did not have a
meaningful impact on the results.
Table 4-2. Predicted median enterococci densities that correspond to illness levels of 36 NGI per
1,000 recreators (RBT) for waters impacted by mixed sea gull and human fecal
contamination.
Nonhuman source
Percent human contribution
10%
20%
30%
50%
100%
Gull (Soller et al., 2014)
339
174
116
70
35
Gull (this TSM gull case study)
349
175
117
70
35
E
o
o
z>
U_
U
CO
cc
o
o
O
o
O
a>
4->
£
400
350
300
250
200
_ 150
100
50
0
~ Enterococci RBT from Soller et al. (2014)
O Enterococci RBT from TSM Gull Case Study
Reference criteria
~
~
n
¦-a
20 40 60 80
Fraction of Human Source Contribution (%)
100
Figure 4-1. Enterococci RBTs for recreational water contaminated by gulls and humans, resulting in
36 NGI per 1,000 recreators. Dotted horizontal line is 35 enterococci CFU/100 mL (the RWQC
magnitude).
121
-------
4.2.2 Compare Children's Exposure to General Population
The EPA conducted a sensitivity analysis to compare the results of the reverse QMRA approach
presented in this TSM depending on whether the incidental ingestion parameter was for the general
population or children 6 to 10 years old (see Section 2.2.2.4.4 for the methodological description and
Appendix G for the base analysis). The median incidental ingestion parameter for the general
population is 19 mL per recreational event (base analysis) and 25 mL per recreational event for
children aged 6 to 10 (sensitivity analysis). The analysis considered different fractions of contamination
from human sources (10%, 25%, and 33%) in relation to the contribution of seagull fecal
contamination.
This sensitivity analysis demonstrates that the ingestion parameter noticeably influences the calculated
enterococci RBT. For each proportion of human fecal contamination in the mixture, the calculated
enterococci levels corresponding to 36 NGI per 1,000 recreators were consistently lower for children
than the general population. As the percentage of human contribution increases, the enterococci RBT
decreases. Including the increased incidental ingestion for children 6 to 10 years old also corresponded
to a lower RBT for enterococci, but the child-based RBT is still greater than the recommended value in
the 2012 RWQC. In the scenario where gull contamination is mixed with human sources, the
enterococci criteria levels would be 24% lower than the levels for the general population, regardless of
the level of human contamination. This corresponds to 25 CFU lower at the 33% human contamination
level and 84 CFU lower at the 10% human contamination level (Table 4-3). Results from a
nonpathogenic source of enterococci are shown in Appendix G and are similar enough to the gull
results that the icons would overlap on Figure 4-2.
Figure 4-2 shows the derived enterococci criteria levels on the Y-axis in CFU per 100 mL and the
percentage of human fecal contribution on the X-axis. The figure shows six different contamination
scenarios: 10%, 20%, 25%, 30%, 33%, and 50% contribution of human fecal contamination. The
remaining nonhuman contribution of enterococci is from seagulls. This sensitivity analysis shows that
hypothetical enterococci criteria levels for this illustrative scenario, derived through the approach
described in this TSM, would be lower for children (squares) compared to the general population
(circles).
Table 4-3. Enterococci RBTs (CFU/100 mL) at different fractions of human fecal contamination (%).
Nonhuman source of fecal
contamination & ingestion
volume
Susceptibility of waterbody to human fecal contamination
(percent human contribution)
Low Medium High
< 10% < 25% < 33% 34-67% > 67%
Gu s (19 mL median ingested „„„ „ „„ „„„
, , ¦ , 349 140 106
volume, gen. population)
Nationally recommended RWQC apply
Gulls (25 mL median ingested „„„
& 265 106 81
volume, children)
122
-------
400
O Enterococci RBT for General Population (gulls' model)
350 O
~ Enterococci RBT for Children (gulls' model)
2012 RWQC Recommendation
300
I ~
g 250
rH
D
U 200
fe O
0£
u 150 _
S ~ o
o m O
-------
both infection (a, J3) and illness (r, r|). Specifically, the median values for infection from norovirus
genogroup I (a, J3) for secreter-positive individuals and the median values for illness (r, r|) based on the
outbreak-based observations were used (Tables 2-5 and 2-6, Section 2.1.1).
The EPA also included updated parameters for incidental ingestion and dose-response relationships for
the reference pathogens adenovirus and Campylobacter. For adenovirus, the median values for infection
(a, |3) and illness (r, r|) were based on the oral route of ingestion (Teunis et al., 2016; also evaluated three
other routes of exposure). For Campylobacter, the median values for infection (a, (3) and illness (r, r|)
were based on the values predicted by Teunis et al. (2018) based on a compilation of challenge study
data (noting that challenge studies and outbreaks resulted in similar infection dose-response
relationships; in outbreaks, the doses required to cause illness were lower than in challenge studies).
The original Mathcad code (not shown) used for Soller et al. (2010a) was modified as summarized in
Table 4-4. In these sensitivity analyses, the EPA revisited the Soller et al. (2010a) analysis to evaluate a
set of parameter changes on the model outputs to test that the main conclusions in Soller et al.
(2010a) were still relevant. Specifically, the EPA evaluated changes to the ingestion, norovirus dose
response, adenovirus dose response, and Campylobacter dose-response parameter selections
(Table 4-4; parameters that were not changed are not shown in Table 4-4.)
The 2010 analyses were conducted using two different approaches to estimate pathogen densities
consistent with reported illness rates. The first, a "health-based approach," assumed that swimming-
associated gastrointestinal illnesses occurred in the same proportion as projected illnesses from known
pathogen occurrence in the United States. The second, a publicly owned treatment works (POTW)
"POTW effluent-based approach," assumed pathogens occurred in the recreational waters in the same
proportion as in disinfected secondary effluent. In these sensitivity analyses, the same approach was
maintained as was used in 2010 to facilitate comparisons and identify the extent to which parameter
selection impacts the overall findings reported in 2010.
Table 4-4. Parameter comparison between Soller et al. (2010a) and the 2022 Etiology Analysis.
Parameters
Parameter detail
2010 (Soller etal., 2010a)
2022 Etiology Analysis
Ingestion
Mean general population
33 mL (Mean)
44 mL
Dose-response
Norovirus hypergeometric
a = 0.04
P = 0.055
a = 0.393
P = 0.767
Adenovirus exponential
0.4172
a = 5.11
P = 2.8
Campylobacter hypergeometric
0.024
0.011
a = 0.44
P = 0.51
Probability of illness
Norovirus
0.6 (60%)
r = 3.19
given infection
H =0.801
Adenovirus
0.5 (50%)
r= 0.41
H = 6.53
Campylobacter
0.028 (2.8%)
r= 0.06
n =0.88
124
-------
Using the reported health association and the daily measured water quality from the freshwater
NEEAR study sites resulted in a mean illness rate of 30.6 illnesses per 1,000 recreators (Soller et al.,
2010b). For the health-based approach, the results (in terms of illness rate per 1,000 swimmers) for
these sensitivity analyses were the same as reported in 2010 because the ratio of illnesses was a fixed
as part of the methodology. The results of these sensitivity analyses for the "POTW effluent-based"
approach are shown in Table 4-5. The results from the 2010 analysis are presented first for comparison
(Base Analysis). Changing the ingestion from 33 mL to 44 mL did not impact the results, as shown in the
Sensitivity Analysis 1 column of Table 4-5. Updating the norovirus dose response based on the results
reported by Teunis et al. (2020) had a minor influence on the quantitative results, where the fraction of
illnesses attributed to norovirus in the analysis decreased from 29.7 to 29.3 NGI per 1,000 recreators
(Sensitivity Analysis 2); in contrast, the fraction of illnesses attributed to other pathogens changed less
than 0.1 NGI per 1,000 recreators. Updating all the parameters to the most recent values (Teunis et al.,
2016, 2018, 2020, and updated ingestion values) yielded results very similar to the Base Analysis
(Sensitivity Analysis 3).
Taken together, these sensitivity analyses support the original findings from the 2010 analyses that
human enteric viruses, as indexed by the norovirus dose response, were likely responsible for the vast
majority of the reported swimming-associated NGI at the NEEAR study sites, and that norovirus can be
considered as an index pathogen when characterizing potential risks from exposure to human fecal
sources.
Table 4-5. Illness rate per 1,000 swimmers for the POTW effluent-based approach.
Base analysis
Sensitivity analysis 1
Sensitivity analysis 2
Sensitivity analysis 3
Pathogen
2010 ingestion and
dose response
2022 ingestion and
2010 dose response
2022 ingestion and
2022 norovirus dose
response
All updated
parameter values*
Rotavirus
0.3
0.3
0.4
0.4
Norovirus
29.7
29.7
29.3
29.8
Adenovirus
0.3
0.3
0.4
0.00015
Cryptosporidium spp.
0.2
0.2
0.3
0.34
Giardia lamblia
0.01
0.01
0.02
0.02
Campylobacter jejuni
0.1
0.1
0.2
0.003
E. coli 0157:H7
0.001
0.001
0.001
0.001
Salmonella enterica
0.0003
0.0003
0.0004
0.0004
Note: * Includes updated dose response and probability of illness given infection for norovirus, adenovirus, and Campylobacter (as shown
in Table 4-4).
125
-------
4.3 Strengths and Uncertainties of Approach
The strengths and uncertainties of this TSM approach are largely based on how the user implements
the approach. Throughout the TSM, users are provided with information to help them conduct each
step with transparency and flexibility to improve parameters with location-specific data. When high-
quality location-specific data are included, the approach provides the best fit for public protection at
the location(s) characterized.
Site selection is an important aspect of the approach and can result in increased strength or,
alternatively, greater uncertainty. Whether site selection is a strength or a limitation depends on how
representative the selected sites are of the locations where potential alternative criteria developed
using the process described in this TSM will apply. The representativeness of the sites can affect the
applicability of the RBTs developed and decisions made using the TSM approach.
Because pathogen densities in fecally contaminated ambient waters can be highly variable and the
associated monitoring results can include many nondetects, the representativeness of the monitoring
approach included in the SAP is also important to consider. Users should have a plan to address
nondetects in their analysis. A robust distribution of pathogen data above the detection limit will result
in better-quality risk estimates in the QMRA. It is important to reiterate that the monitoring approach
used to enumerate the reference pathogens reflects the expected hazardous condition, i.e., when
pathogen loading to the waterbody being studied is expected to occur. If the sanitary survey
information collected in Step 1 identifies that nonhuman fecal source loading to surface waters is
expected following rain events, then the monitoring approach should reflect that pattern. If fecal
loading is event-driven, monitoring base flow conditions may not reflect the hazardous condition.
A transparent documentation of representativeness allows for a better assessment of the
uncertainties. For example, if the user is characterizing beach sites affected by gulls, and the overall
goal is to apply alternative criteria developed using this TSM broadly to gull-affected beaches in a state,
the user would need to identify specific site characteristics and water quality data collection that
support representativeness.
The approach presented in this TSM is designed to be a methodical stepwise process to inform
transparent decision-making. As shown in Sections 1 and 2, the approach includes a sanitary
characterization (sanitary survey and water quality study), evaluation of human health risks (forward
QMRA), and derivation of alternative RBTs for culturable enterococci or E. coli (reverse QMRA). The
approach in this TSM is to provide transparent documentation of each of step when deriving
alternative WQC.
A strength of the sanitary characterization approach in this TSM is that it includes two types of
information: a sanitary survey and a water quality study that are specific to the location being studied.
Together, these sources of information can be used to identify fecal sources, confirm or dispel
assumptions about fecal loading and contamination dynamics, and delineate the representativeness of
the data collected. Identification of the fecal sources is an important outcome of the sanitary
characterization. The better the characterization of fecal sources, the better the confidence in the
outcome of the TSM approach.
There are inherent limitations of water quality methods used to enumerate microorganisms and with
monitoring approaches that do not characterize matrix variability; these limitations apply broadly to
microbial water quality assessment, and a specific discussion on these topics is beyond the scope of
126
-------
thisTSM. However, a discussion of the specific limitations of the water quality methods that are
chosen for the water quality study is helpful for assessing the quality of the data. For example, the
characterization of limits of detection, target specificity, number of replicates, and sample size are
important for transparent interpretation and communication of water quality data.
4.3.1 QMRA
QMRA is a popular approach to aid in understanding the risks associated with exposure to recreational
waters (see Table 1-2 in Section 1.2 for summaries of QMRA studies). QMRA has been used in several
studies to examine the effects of recreational exposures to both human and animal fecal
contamination on human health (Ashbolt et al., 2010; Schoen and Ashbolt, 2010; Soller et al., 2010a,b;
U.S. EPA, 2010a; Schoen et al., 2011). It has also been highlighted as a tool that could be useful for
developing alternative RWQC for sites impacted by nonhuman sources (U.S. EPA, 2007a,b; Boehm,
2009; Dorevitch et al., 2010; Soller et al., 2010a,b).
The QMRAs presented in this TSM have some features that are related to choices that define the scope
of the scenarios that are covered by the QMRA. Scope-related features include:
• Gastrointestinal illness as the health outcome of primary concern: Based on epidemiological
investigations, skin infection and disease, conjunctiva infection and disease, and ear infections
and disease have not been typically shown to be associated with culturable FIB (Priiss, 1998;
Wade et al., 2003; Zmirou et al., 2003). Although an association with culturable FIB has, in some
cases, been shown with respiratory infection and illness (Fleisher et al., 1996) or skin illness
(Sinigalliano et al., 2010), associations with gastrointestinal illness have been documented more
frequently; therefore, gastrointestinal illness rates predicted by the QMRA or established in
epidemiological studies are assumed to be protective for respiratory illness.
• Selection of reference pathogens: The reference pathogens that were selected all have illness
endpoints that include gastrointestinal illness. Additional or other reference pathogens could
be chosen to cover other illness endpoints.
• Incidental ingestion during recreational activities is the exposure route of interest: Other
routes of exposure, such as inhalation and dermal contact, have not been shown to
substantially add to the risk associated with ingestion.
• The approach does not include secondary transmission in the health modeling: Because
criteria for recreational waters are recommended for recreators engaging in primary contact
recreation, the health modeling does not include secondary transmission. In this context, illness
is considered the result of event-based exposure to recreational waters containing fecal
contamination. Secondary transmission is not considered within the scope of determining a
health burden or target level of protection of criteria recommendations. In the QMRA
framework discussed in this TSM, the scope of the scenario does not include person-to-person
illnesses, nor does it explicitly address infectious disease transmission attributes (e.g., person-
to-person transmission and immunity). Although person-to-person transmission and immunity
can influence risk in unintuitive ways (Riley et al., 2003; Eisenberg et al., 2004, 2008; Soller et
al., 2006; Soller, 2009), including these parameters does not substantially affect the estimated
risks for the recreational exposure scenarios considered in this TSM (Soller and Eisenberg,
2008).
127
-------
Assumptions underlying the QMRA include the following:
• Loss of pathogen virulence due to passage through nonhuman hosts or exposure to a
nonenteric environment can be characterized as "high," "medium," and "low" in the QMRA
based on the relative occurrence of species that infect humans and strains and serotypes
present in typical livestock wastes.
• Human dose-response models adequately predict infection or illness risks for reference
pathogens, regardless of the source (though variability in host-pathogen system response can
be included in dose-response modeling).
4.3.1.1 Strengths
Applying QMRA to evaluate recreational waters has developed over the last several decades and has
established a strong scientific foundation in the peer-reviewed literature. The general QMRA approach
is well-accepted and grounded in peer-reviewed approaches and practices that are thoroughly well-
documented. Numerous QMRA-based studies have been peer reviewed and published in scientific
journals, and a standard of practice is now well established.
The QMRA methodology discussed in this TSM is sufficiently flexible for evaluating a range of source
inputs and exposure scenarios. Published QMRA examples in the literature have characterized human
health risks from exposure to many media types, including drinking water, recreational water,
biosolids, and potable and nonpotable recycled water. Data are also available to address recreational
exposure differences among lifestages.
The QMRA process described in this TSM is transparent in the assumptions made and input parameters
chosen. QMRA can also be used to (1) evaluate sensitivity and uncertainty analyses, which can help
test parameter assumptions and choices, and (2) identify potential data gaps, which can inform
research planning. The Problem Formulation in this TSM (Section 2.2) describes the parameter choices
and the rationale for those choices. Further, as new data and information become available, they can
be straightforwardly incorporated into subsequent QMRA studies and the existing QMRAs can be
efficiently revisited and revised if needed. Before new published parameter information is
incorporated, it is recommended that users of this TSM evaluate the need to modify the existing
parameters and the effect of the new information, such as with a sensitivity analysis. As with the
discussion of incorporating newer dose-response information for norovirus (Section 4.2.1), the added
complexity with the newer dose response did not substantially alter the results, so users can evaluate
the need and potential tradeoffs when substantiating parameter selection in the risk assessment.
QMRA can directly address the etiological cause of illness, which is only indirectly addressed in
epidemiological studies. In this TSM, a list of reference pathogens is recommended to estimate risk
because these pathogens account for more than 97% of nonfood borne illnesses in the United States
(Mead et al., 1999; Scallan et al., 2011a,b). In contrast, recreational epidemiological studies most often
characterize water quality in terms of a nonspecific surrogate for fecal contamination, such as
culturable fecal indicator organisms. Additional reference pathogens could be added if desired.
Finally, an important strength of QMRA is the ability to evaluate situations that are difficult or
impossible to evaluate otherwise. As discussed in Section 1, a strength of QMRA is its ability to add
additional insight for risk management in waters not represented by traditional epidemiological-based
128
-------
approaches or in waters where it may be impractical to conduct epidemiological studies (i.e., too few
swimmers). In addition, in some cases, epidemiological approaches fail to establish predictive
relationships between the FIB and reported illnesses. A strength of QMRA is that it does not have these
potential limitations of epidemiological studies and can be cheaper to conduct than epidemiological
studies.
4.3.1.2 Uncertainties/limitations (data availability)
The QMRA approaches presented in this TSM have associated limitations and uncertainties. These
limitations and uncertainties fall into two main categories: (1) methodological and (2) data limitations.
In some cases, the EPA has attempted to address these limitations in the ways discussed below.
Methodological:
The QMRA framework discussed in this TSM includes eight reference pathogens that are surrogates for
the three major groups of pathogens and account for more than 97% of nonfood borne illnesses in the
United States (Mead et al., 1999; Scallan et al., 2011a,b). The TSM approach is flexible and can include
additional pathogens in the framework. A limitation of the reference pathogen approach is that it does
not include all pathogens. Although many pathogens are known, not all are known, and new ones can
emerge over time. It is also the nature of a QMRA to generate a predicted level of illness rather than
describe a reported case symptomology as in the case of an epidemiological study. The EPA has
attempted to address this limitation by characterizing, or "anchoring," the performance of the health
models using the assumptions and information listed above against the empirical outcomes reported in
recreational water epidemiological studies. For example, in Soller et al. (2010b), incubation time and
illness duration in swimmers and nonswimmers in the NEEAR epidemiological study were used to
evaluate the plausibility of the QMRA results. The time-to-onset of illness data for nonswimmers and
the predicted pathogens for the epidemiology studies were used to estimate the distribution of the
time-to-onset of gastrointestinal illness in swimmers. That distribution was then compared to the time-
to-onset of gastrointestinal illness in swimmers reported during the water epidemiology studies to
evaluate the feasibility of the predictions about the pathogens present during the water epidemiology
studies described in the previous sections. The results of the estimated time-to-onset of illness in
swimmers were very similar to the observed time-to-onset of illness in swimmers who participated in
the NEEAR study.
In other examples, QMRA was conducted in concert with epidemiological studies to provide additional
information and interpretation of the results (Soller et al., 2016; Soller et al., 2017). Soller et al. (2016)
describe a QMRA conducted using pathogen data collected during the EPA's epidemiological study in
Boqueron, Puerto Rico. The QMRA calculated a mean swimming-associated illness level of
approximately 2 NGI per 1,000 recreators, a rate of illness below the statistical power the
epidemiological study was designed to detect (i.e., 17 NGI per 1,000 recreators). Soller et al. (2017)
incorporated site-specific pathogen monitoring data of stormwater collected during an epidemiological
study characterizing risks from wet weather impacts in an urbanized watershed. Results were
consistent with enteric viruses, such as norovirus, as an important cause of gastrointestinal illness
among recreators. Coupling QMRA with an epidemiological study at a single site provides additional
utility for understanding illness and informing decisions.
129
-------
Data Limitations:
Data availability and data representativeness are important concerns for selecting parameter values. A
wide range of data quality is associated with potential datasets that could be used in QMRA studies.
• Variations Among Pathogen Strains: Accounting for strain heterogeneity in pathogens is
challenging because dose-response data are not available for all strains, and not all pathogen
strains have been identified or tested. The simplest means to predict response for exposure to
a variety of strains is to ignore inter-strain variations by using a dose-response model based on
pooled data, using a dose-response model based on the most virulent strain among the strains
considered, or selecting a dose-response model for a "representative strain." In some cases,
models of pooled data might not exhibit goodness of fit (Coleman and Marks, 1998). A more
systematic technique for addressing strain-to-strain variation is described by Soller et al. (2007).
In that study and drawing from the previous work of Coleman and Marks (1998, 2000),
Gompertz-log dose-response models (alternatively called the Weibull dose-response model)
were fitted to data for all strains of Salmonella for which data were available.
• Accounting for Differential Susceptibility in Dose Response: Intrinsic and extrinsic factors
modify the effect of a specific exposure on the risk or severity of outcomes in an individual
population (Balbus et al., 2000). Intrinsic factors include age, gender, prior disease, immune
status, pregnancy, and diabetes, and extrinsic factors include residence, income, co-exposures,
access to health care, and behaviors (Lanciers et al., 1999; Currie et al., 2000; Parkin et al.,
2003; Makri et al., 2004; Parkin, 2004; WHO, 2005; Dietert et al., 2010). Dose-response is
related to the intrinsic factors that affect how likely a person is to become infected or ill after
exposure to a pathogen.
• Fate and Transport of Microbes in Environmental Waters: Hydrodynamic and transport models
have been used to characterize changes in pathogen densities for various scenarios, including
watershed scale (a review of watershed transport models is found in Coffey et al. [2007]);
coastal waters (e.g., by Liu et al., 2006); for mixing in a river reach downstream of WWTP
discharge (Soller et al., 2003); CSO-impacted river and bay (King County Department of Natural
Resources, 1999); in an impaired waterbody (Soller et al., 2006); and for fate and transport of
manure-based pathogens (Martinez et al., 2014).
4.4 Future Directions/Research Needs
The scope of thisTSM is based on the scope of the 2012 RWQC. This discussion is limited to potential
future information relevant to the current scope of the TSM. Therefore, directions that would expand
the scope of the TSM are not included here.
Additional information that could be informative at the local scale includes:
• Fate and transport of pathogens and indicators: improved understanding of specific scenarios
to provide information for IEM.
• Quantitative source characterization: improved methods for identifying and quantifying fecal
sources, including more data collected at locations of interest.
130
-------
• Pathogen profiles: more characterization of pathogen profiles from fecal sources across
geographic regions.
• Exposure scenarios: expand understanding of different exposure scenarios.
131
-------
5.0 References
Abbaszadegan, M., P. Stewart, and M. LeChevallier. 1999. A strategy for detection of viruses in
groundwater by PCR. Applied and Environmental Microbiology 65(2):444-449.
Abbaszadegan, M. 2006. Rotaviruses. Chapter 46 in Waterborne Pathogens: Manual of Water Supply
Practices. AWWA Manual M48. 2nd ed. 295-298 pp. American Water Works Association,
Denver, CO.
Ahmed, W., A. Goonetilleke, D. Powell, K. Chauhan, and T. Gardner. 2009. Comparison of molecular
markers to detect fresh sewage in environmental waters. Water Research 43(19):4908-4917.
Ahmed, W., K.A. Hamilton, A. Lobos, B. Hughes, C. Staley, M.J. Sadowsky, and V.J. Harwood. 2018.
Quantitative microbial risk assessment of microbial source tracking markers in recreational
water contaminated with fresh untreated and secondary treated sewage. Environment
International 117:243-249.
Ahmed, J., L.P. Wong, Y.P. Chua, N. Channa, R.B. Mahar, A. Yasmin, J.A. VanDerslice, and J.V. Garn.
2020. Quantitative Microbial Risk Assessment of Drinking Water Quality to Predict the Risk of
Waterborne Diseases in Primary-School Children. International Journal of Environmental
Research and Public Health 17(8).
Al-Saleem, T., and H. Al-Mondhir. 2005. Immunoproliferative small intestinal disease (IPSID): A model
for mature B-cell neoplasms. Blood 105(6):2274-2280.
Ang, C.W., Noordzij, P.G., de Klerk, M.A., Endtz, H.P., van Doom, P.A., Laman, J.D. 2002. Ganglioside
mimicry of Campylobacter jejuni lipopolysaccharides determines antiganglioside specificity in
rabbits. Infection and Immunity 70(9):5081-5085.
Anonymous. 1998. Case records of the Massachusetts General Hospital. Weekly clinicopathological
exercises. Case 19-1998. A 70-year-old man with diarrhea, polyarthritis, and a history of Reiter's
syndrome. New England Journal of Medicine 338(25):1830-1836.
Ansari, S.A., V.S. Springthorpe, and S.A. Sattar. 1991. Survival and vehicular spread of human
rotaviruses: Possible relation to seasonality of outbreaks. Reviews of Infectious Disease
13(3):448-461.
Antaki-Zukoski, E.M., X. Li, B. Hoar, J.M. Adaska, B.A. Byrne, and E.R. Atwill. 2021. Understanding the
transmission dynamics of Escherichia coli 0157:H7 super-shedding infections in feedlot cattle.
PeerJ. Computer Science 9:el2524.
Arnold, B.F., K.C. Schiff, J.F. Griffith, J.S. Gruber, V. Yau, C.C. Wright, T.J. Wade, S. Burns, J.M. Hayes, C.
McGee, M. Gold, Y. Cao, S.B. Weisberg, and J.M. Colford, Jr. 2013. Swimmer illness associated
with marine water exposure and water quality indicators: Impact of widely used assumptions.
Epidemiology 24(6):845-853.
Arnold, B.F., T.J. Wade, J. Benjamin-Chung, K.C. Schiff, J.F. Griffith, A.P. Dufour, S.B. Weisberg, and J.M.
Colford, Jr. 2016. Acute Gastroenteritis and Recreational Water: Highest Burden Among Young
US Children. American Journal of Public Health 106(9):1690-1697.
132
-------
Arnold, B.F., K.C. Schiff, A. Ercumen, J. Benjamin-Chung, J.A. Steele, J.F. Griffith, S.J. Steinberg, P. Smith,
C.D. McGee, R. Wilson, C. Nelsen, S.B. Weisberg, and J.M. Colford, Jr. 2017. Acute Illness Among
Surfers After Exposure to Seawater in Dry- and Wet-Weather Conditions. American Journal of
Epidemiology 186(7):866-875.
ASCE (American Society of Civil Engineers) and U.S. EPA (U.S. Environmental Protection Agency). 2000.
Determining Urban Stormwater Best Management Practice (BMP) Removal Efficiencies.
American Society of Civil Engineers and U.S. Environmental Protection Agency, Washington, DC.
Ashbolt, N.J., M.E. Schoen, J.A. Soller, and D.J. Roser. 2010. Predicting pathogen risks to aid beach
management: The real value of quantitative microbial risk assessment (QMRA). Water Research
44(16) :4692-4703.
ASTM. 2020. Standard Practice for Determining Microbial Colony Counts from Waters Analyzed by
Plating Methods [D5465-16(2020)]. ASTM International, Washington, DC.
Atwill, E.R., L. Hou, B.M. Karle, T. Harter, K.W. Tate, and R.A. Dahlgren. 2002. Transport of
Cryptosporidium parvum oocysts through vegetated buffer strips and estimated filtration
efficiency. Applied and Environmental Microbiology 68(ll):5517-5527.
Atwill, E.R., M.D. Pereira, L.H. Alonso, C. Elmi, W.B. Epperson, R. Smith, W. Riggs, L.V. Carpenter, D.A.
Dargatz, and B. Hoar. 2006. Environmental load of Cryptosporidium parvum oocysts from cattle
manure in feedlots from the central and western United States. Journal of Environmental
Quality 35(l):200-206.
Backer, L.C., S.V. McNeel, T. Barber, B. Kirkpatrick, C. Williams, M. Irvin, Y. Zhou, T.B. Johnson, K.
Nierenberg, M. Aubel, R. LePrell, A. Chapman, A. Foss, S. Corum, V.R. Hill, S.M. Kieszak, and Y.S.
Cheng. 2010. Recreational exposure to microcystins during algal blooms in two California lakes.
Toxicon 55(5):909-921.
Balbus, J., R. Parkin, and M. Embrey. 2000. Susceptibility in microbial risk assessment: Definitions and
research needs. Environmental Health Perspectives 108(9):901-905.
Ballweber, L.R., L. Xiao, D.D. Bowman, G. Kahn, and V.A. Cama. 2010. Giardiasis in dogs and cats:
Update on epidemiology and public health significance. Trends in Parasitology 26(4):180-189.
Bartram, J., and G. Rees. 2000. Monitoring bathing waters: A practical guide to the design and
implementation of assessments and monitoring programmes. London, UK, and New York, NY: E
& FN Spon for WHO, U.S. EPA, and Commission of the European Communities.
Bastos, R.K., P.D. Bevilacqua, C.A. Silva, and C.V. Silva. 2008. Wastewater irrigation of salad crops:
Further evidence for the evaluation of the WHO guidelines. Water Science and Technology
57(8):1213-1219.
Be'er, A., H.P. Zhang, E.L. Florin, S.M. Payne, E. Ben-Jacob, and H.L. Swinney. 2009. Deadly competition
between sibling bacterial colonies. Proceedings of the National Academy of Sciences of the
United States of America 106(2) :428-433.
Benenson, A.S., ed. 1995. Giardiasis. Chapter in Control of Communicable Diseases Manual. 16th ed.
American Public Health Association, Washington, DC.
133
-------
Berry, E.D., B.L. Woodbury, J.A. Nienaber, R.A. Eigenberg, J.A. Thurston, and J.E. Wells. 2007. Incidence
and persistence of zoonotic bacterial and protozoan pathogens in a beef cattle feedlot runoff
control vegetative treatment system. Journal of Environmental Quality 36(6):1873-1882.
Bicudo, J.R., and S.M. Goyal. 2003. Pathogens and manure management systems: A review.
Environmental Technology 24(1):115-130.
Bielaszewska, M., J. Janda, K. Blahova, H. Minarfkova, E. Jfkova, M.A. Karmali, J. Laubova, J. Sikulova,
M.A. Preston, R. Khakhria, H. Karch, H. Klazarova, and 0. Nyc. 1997. Human Escherichia coli
0157:H7 infection associated with the consumption of unpasteurized goat's milk. Epidemiology
and Infection 119(3):299-305.
Birkhead, G., and R.L. Vogt. 1989. Epidemiologic surveillance for endemic Giardia lamblia infection in
Vermont. The roles of waterborne and person-to-person transmission. American Journal of
Epidemiology 129(4) :762-768.
Boehm, A.B., Ashbolt, N.J., Colford, J.M., Jr., Dunbar, L.E., Fleming, L.E., Gold, M.A., Hansel, J.A., Hunter,
P.R., Ichida, A.M., McGee, C.D., Soller, J.A., Weisberg, S.B. 2009. A sea change ahead for
recreational water quality criteria. Journal of Water and Health, 7(l):9-20.
Boehm, A.B., L.C. Van De Werfhorst, J.F. Griffith, P.A. Holden, J.A. Jay, O.C. Shanks, D. Wang, and S.B.
Weisberg. 2013. Performance of forty-one microbial source tracking methods: A twenty-seven
lab evaluation study. Water Research 47(18):6812-6828.
Boehm, A.B., J.A. Soller, and O.C. Shanks. 2015. Human-associated fecal quantitative polymerase chain
reaction measurements and simulated risk of gastrointestinal illness in recreational waters
contaminated with raw sewage. Environmental Science & Technology Letters 2(10):270-275.
Boehm, A.B., K.E. Graham, and W.C. Jennings. 2018. Can we swim yet? Systematic review, meta-
analysis, and risk assessment of aging sewage in surface waters. Environmental Science &
Technology 52(17):9634-9645.
Boehm, A.B., A.I. Silverman, A. Schriewer, and K. Goodwin. 2019. Systematic review and meta-analysis
of decay rates of waterborne mammalian viruses and coliphages in surface waters. Water
Research 164:114898.
Boehm, A.B., and J.A. Soller. 2020. Refined ambient water quality thresholds for human-associated
fecal indicator HF183 for recreational waters with and without co-occurring gull fecal
contamination. Microbial Risk Analysis 16:100139.
Bofill-Mas, S., and M. Rusinol. 2020. Recent trends on methods for the concentration of viruses from
water samples. Current Opinion in Environmental Science & Health 16:7-13.
Bollaerts, K.E., W. Messens, L. Delhalle, M. Aerts, Y. Van der Stede, J. Dewulf, S. Quoilin, D. Maes, K.
Mintiens, and K. Grijspeerdt. 2009. Development of a quantitative microbial risk assessment for
human salmonellosis through household consumption of fresh minced pork meat in Belgium.
Risk Analysis 29(6):820-840.
134
-------
Borchardt, M.A., P.D. Bertz, S.K. Spencer, and D.A. Battigelli. 2003. Incidence of enteric viruses in
groundwater from household wells in Wisconsin. Applied and Environmental Microbiology
69(2):1172-1180.
Boring, J.R., 3rd, W.T. Martin, and L.M. Elliott. 1971. Isolation of Salmonella typhimurium from
municipal water, Riverside, California, 1965. American Journal of Epidemiology 93(l):49-54.
Bradford, S.A., V.L. Morales, W. Zhang, R.W. Harvey, A.I. Packman, A. Mohanram, and C. Welty. 2013.
Transport and fate of microbial pathogens in agricultural settings. Critical Reviews in
Environmental Science and Technology 43(8):775-893.
Bradshaw, J.K., B.J. Snyder, A. Oladeinde, D. Spidle, M.E. Berrang, R.J. Meinersmann, B. Oakley, R.C.
Sidle, K. Sullivan, and M. Molina. 2016. Characterizing relationships among fecal indicator
bacteria, microbial source tracking markers, and associated waterborne pathogen occurrence in
stream water and sediments in a mixed land use watershed. Water Research 101:498-509.
Brooks, J.P., A. Adeli, J.J. Read, and M.R. McLaughlin. 2009. Rainfall simulation in greenhouse
microcosms to assess bacterial-associated runoff from land-applied poultry litter. Journal of
Environmental Quality 38(l):218-229.
Brown, K.I., K.E. Graham, and A.B. Boehm. 2017a. Risk-based threshold of gull-associated fecal marker
concentrations for recreational water. Environmental Science & Technology Letters 4(2):44-48.
Brown, K.I., K.E. Graham, J.A. Soller, and A.B. Boehm. 2017b. Estimating the probability of illness due to
swimming in recreational water with a mixture of human- and gull-associated microbial source
tracking markers. Environmental Science. Process & Impacts 19(12):1528-1541.
Brunner, R.L., D.J. O'Grady, J.C. Partin, J.S. Partin, and W.K. Schubert. 1979. Neuropsychologic
consequences of Reye syndrome. Journal of Pediatrics 95(5 Pt 1):706-711.
Buchanan, R.L., J.L. Smith, and W. Long. 2000. Microbial risk assessment: Dose-response relations and
risk characterization. International Journal of Food Microbiology 58(3):159-172.
Bunning, V.K., J.A. Lindsay, and D.L. Archer. 1997. Chronic health effects of microbial foodborne
disease. World Health Statistics Quarterly 50(l-2):51-56.
Bushon, R.N., A.M. Brady, E.D. Christensen, and E.A. Stelzer. 2017. Multi-Year Microbial Source
Tracking Study Characterizing Fecal Contamination in an Urban Watershed. Water Environment
Research 89(2):127-143.
Butler, N., J. Carlisle, and R. Linville. 2012. Toxicological Summary and Suggested Action Levels to
Reduce Potential Adverse Health Effects of Six Cyanotoxins. California Environmental Protection
Agency, Office of Environmental Health Hazard Assessment, Sacramento, CA.
Butzler, J.P. 2004. Campylobacter, from obscurity to celebrity. Clinical Microbiology and Infection
10(10):868-876.
Byappanahalli, M.N., D.A. Shively, M.B. Nevers, M.J. Sadowsky, and R.L. Whitman. 2003. Growth and
survival of Escherichia coli and enterococci populations in the macro-alga Cladophora
(Chlorophyta). FEMS Microbiology Ecology 46(2):203-211.
135
-------
Byappanahalli, M.N., R.L. Whitman, D.A. Shively, W.T. Ting, C.C. Tseng, and M.B. Nevers. 2006.
Seasonal persistence and population characteristics of Escherichia coli and enterococci in deep
backshore sand of two freshwater beaches. Journal of Water and Health 4(3):313-320.
Byappanahalli, M.N., M.B. Nevers, A. Korajkic, Z.R. Staley, and V.J. Harwood. 2012. Enterococci in the
environment. Microbiology and Molecular Biology Reviews 76(4):685-706.
Byrd, J.A., D.E. Corrier, J.R. Deloach, D.J. Nisbet, and L.H. Stanker. 1998. Horizontal transmission of
Salmonella typhimurium in broiler chicks. Journal of Applied Poultry Research 7(l):75-80.
Cabelli, V.J. 1983. Health Effects Criteria for Marine Recreational Waters. EPA 600/1-80-031. U.S.
Environmental Protection Agency, Office of Research and Development, Health Effects
Research Laboratory, Research Triangle Park, NC.
Cabelli, V.J., A.P. Dufour, L.J. McCabe, and M.A. Levin. 1982. Swimming-associated gastroenteritis and
water quality. American Journal of Epidemiology 115(4):606-616.
Caccio, S.M. 2005. Molecular epidemiology of human cryptosporidiosis. Parassitologia 47(2):185-192.
Calderon, R.L., E.W. Mood, and A.P. Dufour. 1991. Health effects of swimmers and nonpoint sources of
contaminated water. International Journal of Environmental Health Research 1(1):21-31.
CAMRA (Center for Advancing Microbial Risk Assessment). 2021. QMRA Wiki. Center for Advancing
Microbial Risk Assessment at Michigan State University. Accessed June 23, 2023.
http://qmrawiki.org.
Canales, R.A., A.M. Wilson, J.I. Pearce-Walker, M.P. Verhougstraete, and K.A. Reynolds. 2018. Methods
for handling left-censored data in quantitative microbial risk assessment. Applied and
Environmental Microbiology 84(20) :e01203-18.
Cao, Y., M. Sivaganesan, C.A. Kelty, D. Wang, A.B. Boehm, J.F. Griffith, S.B. Weisberg, and O.C. Shanks.
2018. A human fecal contamination score for ranking recreational sites using the
HF183/BacR287 quantitative real-time PCR method. Water Research 128:148-156.
Casemore, D.P., Wright, S.E., Coop, R.L. 1997. Cryptosporidiosis—Human and animal epidemiology.
Chapter 3 in Cryptosporidium and Cryptosporidiosis. R. Fayer, ed. 65-92 pp. CRC Press, New
York, NY.
CDC (Centers for Disease Control and Prevention). 2006. Surveillance for waterborne disease and
outbreaks associated with recreational water— United States, 2003-2004. Morbidity and
Mortality Weekly Report, Surveillance Summaries 55(SS12):l-24.
CDC (Centers for Disease Control and Prevention). 2008. Surveillance for waterborne-disease and
outbreaks associated with recreational water use and other aquatic facilities—United States,
2005-2006. Morbidity and Mortality Weekly Report 57:1-72.
CDC (Centers for Disease Control and Prevention). 2011. Genomics in Public Health Preparedness:
Chance Favors the Prepared Mind, https://blogs.cdc.gov/genomics/2011/09/22/genomics-in-
public-health-preparedness-chance-favors-the-prepa red-mind/.
136
-------
CDC (Centers for Disease Control and Prevention). 2012. Epidemiology and Prevention of Vaccine-
Preventable Diseases. The Pink Book: Course Textbook. 12th ed. U.S. Department of Health and
Human Services, Centers for Disease Control and Prevention, Atlanta, GA.
CDC (Centers for Disease Control and Prevention). 2018. Rotavirus in the U.S. Accessed June 23, 2023.
https://www.cdc.gov/rotavirus/surveillance.html.
CDC (Centers for Disease Control and Prevention). 2022. Vibrio vulnificus & Wounds. Accessed June 23,
2023. https://www.cdc.gov/vibrio/wounds.html.
CDC (Centers for Disease Control and Prevention). 2023. Norovirus. Accessed July 13, 2023.
https://www.cdc.gov/norovirus/index.html.
Census. 2010. Table 1. Annual Estimates of the Resident Population by Sex and Five-Year Age Groups
for the United States: April 1, 2000 to July 1, 2009 (NC-EST2009-01). Release date: June 2010.
U.S. Census Bureau, Population Division.
Chapman, P.A., C.A. Siddons, A.T. Gerdan Malo, and M.A. Harkin. 1997. A 1-year study of Escherichia
coli 0157 in cattle, sheep, pigs and poultry. Epidemiology and Infection 119(2):245-250.
Chappell, C.L., P.C. Okhuysen, C.R. Sterling, and H.L. DuPont. 1996. Cryptosporidium parvum: Intensity
of infection and oocyst excretion patterns in healthy volunteers. The Journal of Infectious
Diseases 173(l):232-236.
Chappell, C.L., P.C. Okhuysen, R. Langer-Curry, G. Widmer, D.E. Akiyoshi, S. Tanriverdi, and S. Tzipori.
2006. Cryptosporidium hominis: Experimental challenge of healthy adults. The American Journal
of Tropical Medicine and Hygiene 75(5):851-857.
Chaudhry, R.M., K.A. Hamilton, C.N. Haas, and K.L. Nelson. 2017. Drivers of microbial risk for direct
potable reuse and de facto reuse treatment schemes: The impacts of source water quality and
blending. International Journal of Environmental Research and Public Health 14(6):635.
Choi, S., and S.C. Jiang. 2005. Real-time PCR quantification of human adenoviruses in urban rivers
indicates genome prevalence but low infectivity. Applied and Environmental Microbiology
71(11) :7426-7433.
Clements, A., J.C. Young, N. Constantinou, and G. Frankel. 2012. Infection strategies of enteric
pathogenic Escherichia coli. Gut Microbes 3(2):71-87.
Coffey, R., E. Cummins, M. Cormican, V.O. Flaherty, and S. Kelly. 2007. Microbial exposure assessment
of waterborne pathogens. Human and Ecological Risk Assessment 13(6):1313-1351.
Coleman, M., and H. Marks. 1998. Topics in dose-response modeling. Journal of Food Protection
61(11) :1550-1559.
Coleman, M.E., and H.M. Marks. 2000. Mechanistic modeling of salmonellosis. Quantitative
Microbiology 2(3):227-247.
Coleman, M.E., H.M. Marks, N.J. Golden, and H.K. Latimer. 2004. Discerning strain effects in microbial
dose-response data. Journal of Toxicology and Environmental Health Part A 67(8-10):667-685.
137
-------
Colford, J.M., Jr., T.J. Wade, K.C. Schiff, C.C. Wright, J.F. Griffith, S.K. Sandhu, S. Burns, M. Sobsey, G.
Lovelace, and S.B. Weisberg. 2007. Water quality indicators and the risk of illness at beaches
with nonpoint sources of fecal contamination. Epidemiology 18(l):27-35.
Colford, J.M., Jr., K.C. Schiff, J.F. Griffith, V. Yau, B.F. Arnold, C.C. Wright, J.S. Gruber, T.J. Wade, S.
Burns, J. Hayes, C. McGee, M. Gold, Y. Cao, R.T. Noble, R. Haugland, and S.B. Weisberg. 2012.
Using rapid indicators for Enterococcus to assess the risk of illness after exposure to urban
runoff contaminated marine water. Water Research 46(7):2176-2186.
Cornick, N.A., and A.F. Helgerson. 2004. Transmission and infectious dose of Escherichia coli 0157:H7
in swine. Applied and Environmental Microbiology 70(9):5331-5335.
Covert, T.C. 1999. Salmonella. Chapter in Waterborne Pathogens: Manual of Water Supply Practices.
AWWA Manual M48. 1st ed. American Water Works Association, Denver, CO.
Covert, T.C., and M.C. Meckes. 2006. Salmonella. Chapter 17 in Waterborne Pathogens: Manual of
Water Supply Practices. AWWA Manual M48. 2nd ed. 135-139 pp. American Water Works
Association, Denver, CO.
Cox, N.A., N.J. Stern, M.T. Musgrove, J.S. Bailey, S.E. Craven, P.F. Cray, R.J. Buhr, and K.L. Hiett. 2002.
Prevalence and level of Campylobacter in commercial broiler breeders (parents) and broilers.
Journal of Applied Poultry Research 11(2):187-190.
Craun, G.F., R.L. Calderon, and M.F. Craun. 2005. Outbreaks associated with recreational water in the
United States. International Journal of Environmental Health Research 15(4):243-262.
Craun, M.F., G.F. Craun, R.L. Calderon, and M.J. Beach. 2006. Waterborne outbreaks reported in the
United States. Journal of Water and Health 4(Suppl 2):19-30.
Currie, B.J., D.A. Fisher, D.M. Howard, J.N. Burrow, D. Lo, S. Selva-Nayagam, N.M. Anstey, S.E. Huffam,
P.L. Snelling, P.J. Marks, D.P. Stephens, G.D. Lum, S.P. Jacups, and V.L. Krause. 2000. Endemic
melioidosis in tropical northern Australia: A 10-year prospective study and review of the
literature. Clinical Infectious Diseases 31(4):981-986.
CWP (Center for Watershed Protection). 2007. National Pollutant Removal Performance Database
(Version 3). Center for Watershed Protection.
Daugherty, C.C., and J.E. Heubi. 1985. Reye's syndrome associated with adenovirus infections.
American Journal of Diseases of Children 139(11):1076.
Davidson, P.W., R.H. Willoughby, L.A. O'Tuama, C.N. Swisher, and D. Benjamins. 1978. Neurological and
intellectual sequelae of Reye's syndrome. American Journal of Mental Deficiency 82(6) :535-
541.
de Graaf, M., R. Beck, S.M. Caccio, B. Duim, P. Fraaij, F.S. Le Guyader, M. Lecuit, J. Le Pendu, E. de Wit,
and C. Schultsz. 2017. Sustained fecal-oral human-to-human transmission following a zoonotic
event. Current Opinion in Virology 22:1-6.
138
-------
de Man, H., H.H. van den Berg, E.J. Leenen, J.F. Schijven, F.M. Schets, J.C. van der Vliet, F. van Knapen,
and A.M. de Roda Husman. 2014. Quantitative assessment of infection risk from exposure to
waterborne pathogens in urban floodwater. Water Research 48:90-99.
DeFlorio-Barker, S., B.F. Arnold, E.A. Sams, A.P. Dufour, J.M. Colford, Jr., S.B. Weisberg, K.C. Schiff, and
T.J. Wade. 2017. Child environmental exposures to water and sand at the beach: Findings from
studies of over 68,000 subjects at 12 beaches. Journal of Exposure Science & Environmental
Epidemiology 00:1-8.
Degremont, A., D. Sturchler, E. Wolfensberger, and B. Osterwalder. 1981. [Clinical and therapeutic
study of a group of 217 patients with intestinal giardiasis and amebiasis]. Schweizerische
Medizinische Wochenschrift 111(52):2039-2046. [In German.]
Dhabhar, F.S. 2014. Effects of stress on immune function: The good, the bad, and the beautiful.
Immunologic Research 58(2-3):193-210.
Dietert, R.R., J.C. DeWitt, D.R. Germolec, and J.T. Zelikoff. 2010. Breaking patterns of environmentally
influenced disease for health risk reduction: Immune perspectives. Environmental Health
Perspectives 118(8): 1091-1099.
Dietz, V., D. Vugia, R. Nelson, J. Wicklund, J. Nadle, K.G. McCombs, and S. Reddy. 2000. Active,
multisite, laboratory-based surveillance for Cryptosporidium parvum. The American Journal of
Tropical Medicine and Hygiene 62(3):368-372.
Doane, C.A., P. Pangloli, H.A. Richards, J.R. Mount, D.A. Golden, and F.A. Draughon. 2007. Occurrence
of Escherichia coli 0157:H7 in diverse farm environments. Journal of Food Protection 70(1):6-
10.
Dorevitch, S., N.J. Ashbolt, C.M. Ferguson, R. Fujioka, C.D. McGee, J.A. Soller, and R.L. Whitman. 2010.
Meeting report: Knowledge and gaps in developing microbial criteria for inland recreational
waters. Environmental Health Perspectives 118(6):871-876.
Dorevitch, S., S. Panthi, Y. Huang, H. Li, A.M. Michalek, P. Pratap, M. Wroblewski, L. Liu, P.A. Scheff, and
A. Li. 2011. Water ingestion during water recreation. Water Research 45(5):2020-2028.
Dorevitch, S., M.S. Dworkin, S.A. Deflorio, W.M. Janda, J. Wuellner, and R.C. Hershow. 2012. Enteric
pathogens in stool samples of Chicago-area water recreators with new-onset gastrointestinal
symptoms. Water Research 46(16):4961-4972.
Dorner, S.M., P.M. Huck, and R.M. Slawson. 2004. Estimating potential environmental loadings of
Cryptosporidium spp. and Campylobacter spp. from livestock in the Grand River Watershed,
Ontario, Canada. Environmental Science & Technology 38(12):3370-3380.
Dorr, P.M., D.A. Tadesse, B.M. Zewde, P. Fry, S. Thakur, and W.A. Gebreyes. 2009. Longitudinal study of
Salmonella dispersion and the role of environmental contamination in commercial swine
production systems. Applied and Environmental Microbiology 75(6):1478-1486.
Dropulic, L.K., and H.M. Lederman. 2016. Overview of Infections in the Immunocompromised Host. In
Diagnostic Microbiology of the Immunocompromised Host. R.T. Hayden, D.M. Wolk, K.C. Carroll,
Y.-W. Tang, eds. 1-50 pp. Wiley-Blackwell.
139
-------
Dufour, A.P. 1984. Health effects criteria for fresh recreational waters. Research Triangle Park, NC: U.S.
Environmental Protection Agency. EPA 600/1-84-004.
Dufour, A.P., 0. Evans, T.D. Behymer, and R. Cantu. 2006. Water ingestion during swimming activities
in a pool: A pilot study. Journal of Water and Health 4(4):425-430.
Dufour, A., and S. Schaub. 2007. The evolution of water quality criteria in the United States, 1922-
2003. Chapter 1 in Statistical Framework for Recreational Water Quality Criteria and
Monitoring. L.J. Wymer, ed. 1-11 pp. John Wiley and Sons, Ltd., Chichester, UK.
Dufour, A., T.J. Wade, and D. Kay. 2012. Epidemiological Studies on Swimmer Health Effects Associated
with Potential Exposure to Zoonotic Pathogens in Bathing Beach Water—A Review. Chapter 11
in Animal Waste, Water Quality and Human Health. A. Dufour, J. Bartram, R. Bos, V. Gannon,
eds. 415-428 pp. IWA Publishing for the World Health Organization, London, UK.
Dufour, A.P., T.D. Behymer, R. Cantu, M. Magnuson, and L.J. Wymer. 2017. Ingestion of swimming pool
water by recreational swimmers. Journal of Water and Health 15(3):429-437.
Dupont, H.L., and P.S. Sullivan. 1986. Giardiasis: The clinical spectrum, diagnosis and therapy. Pediatric
Infectious Disease 5(1 Suppl):S131-138.
DuPont, H.L., C.L. Chappell, C.R. Sterling, P.C. Okhuysen, J.B. Rose, and W. Jakubowski. 1995. The
infectivity of Cryptosporidium parvum in healthy volunteers. New England Journal of Medicine
332(13):855-859.
Dworkin, M.S., P.C. Shoemaker, M.J. Goldoft, and J.M. Kobayashi. 2001. Reactive arthritis and Reiter's
syndrome following an outbreak of gastroenteritis caused by Salmonella enteritidis. Clinical
Infectious Diseases 33(7):1010-1014.
Effler, P., M.C. Leong, A. Kimura, M. Nakata, R. Burr, E. Cremer, and L. Slutsker. 2001. Sporadic
Campylobacter jejuni infections in Hawaii: Associations with prior antibiotic use and
commercially prepared chicken. The Journal of Infectious Diseases 183(7):1152-1155.
Eftim, S.E., T. Hong, J. Soller, A. Boehm, I. Warren, A. Ichida, and S.P. Nappier. 2017. Occurrence of
norovirus in raw sewage—A systematic literature review and meta-analysis. Water Research
111:366-374.
Eichmiller, J.J., A.J. Borchert, M.J. Sadowsky, and R.E. Hicks. 2014. Decay of genetic markers for fecal
bacterial indicators and pathogens in sand from Lake Superior. Water Research 59:99-111.
Eisenberg, J.N., E.Y. Seto, A.W. Olivieri, and R.C. Spear. 1996. Quantifying water pathogen risk in an
epidemiological framework. Risk Analysis 16(4):549-563.
Eisenberg, J.N., E.Y. Seto, J.M. Colford, Jr., A. Olivieri, and R.C. Spear. 1998. An analysis of the
Milwaukee cryptosporidiosis outbreak based on a dynamic model of the infection process.
Epidemiology 9(3):255-263.
Eisenberg, J.N., J.A. Soller, J. Scott, D.M. Eisenberg, J.M. Colford, Jr. 2004. A dynamic model to assess
microbial health risks associated with beneficial uses of biosolids. Risk Analysis 24(l):221-236.
140
-------
Eisenberg, J.N., K. Moore, J.A. Soller, D. Eisenberg, and J.M. Colford, Jr. 2008. Microbial risk assessment
framework for exposure to amended sludge projects. Environmental Health Perspectives
116(6):727-733.
El-Shibiny, A., P.L. Connerton, and I.F. Connerton, I.F. 2005. Enumeration and diversity of
Campylobacters and bacteriophages isolated during the rearing cycles of free-range and organic
chickens. Applied and Environmental Microbiology 71(3):1259-1266.
Englehardt, J.D., and J. Swartout. 2004. Predictive population dose-response assessment for
Cryptosporidium parvum: Infection endpoint. Journal of Toxicology and Environmental Health
Part A 67(8-10) :651-666.
Englehardt, J.D., and J. Swartout. 2006. Predictive Bayesian microbial dose-response assessment based
on suggested self-organization in primary illness response: Cryptosporidium parvum. Risk
Analysis 26(2):543-554.
Englehardt, J., J. Swartout, and C. Loewenstine. 2009. A new theoretical discrete growth distribution
with verification for microbial counts in water. Risk Analysis 29(6):841-856.
Enriquez, C., and J. Thurston-Enriquez. 2006. Adenoviruses. Chapter 38 in Waterborne Pathogens:
Manual of Water Supply Practices. AWWA Manual M48. 2nd ed. 253-257 pp. American Water
Works Association, Denver, CO.
Eregno, F.E., I. Tryland, T. Tjomsland, M. Myrmel, L. Robertson, and A. Heistad. 2016. Quantitative
microbial risk assessment combined with hydrodynamic modelling to estimate the public health
risk associated with bathing after rainfall events. The Science of the Total Environment 548-
549:270-279.
Eregno, F.E., I. Tryland, M. Myrmel, A. Wennberg, A. Oliinyk, M. Khatri, and A. Heistad. 2018. Decay
rate of virus and faecal indicator bacteria (FIB) in seawater and the concentration of FIBs in
different wastewater systems. Microbial Risk Analysis 8:14-21.
Ervin, J.S., L.C. Van De Werfhorst, J.L. Murray, and P.A. Holden. 2014. Microbial source tracking in a
coastal California watershed reveals canines as controllable sources of fecal contamination.
Environmental Science & Technology 48(16):9043-9052.
Esmen, N.A., and Y.Y. Hammad. 1977. Log-normality of environmental sampling data. Journal of
Environmental Science and Health. Part A: Environmental Science and Engineering 12(l-2):29-
41.
EU (European Union). 2006. Directive 2006/7/EC of the European Parliament and of the Council of 15
February 2006 Concerning the Management of Bathing Water Quality and Repealing Directive
76/160/EEC. European Union.
Fan, L., J. Shuai, R. Zeng, H. Mo, S. Wang, X. Zhang, and Y. He. 2017. Validation and application of
quantitative PCR assays using host-specific Bacteroidales genetic markers for swine fecal
pollution tracking. Environmental Pollution 231(Pt 2):1569-1577.
Fayer, R., C.A. Speer, and J.P. Dubey. 1997. The General Biology of Cryptosporidium. Chapter In
Cryptosporidium and Cryptosporidiosis. R. Fayer, ed. CRC Press, New York, NY.
141
-------
Fayer, R., J.M. Trout, T.K. Graczyk, and E.J. Lewis. 2000. Prevalence of Cryptosporidium, Giardia and
Eimeria infections in post-weaned and adult cattle on three Maryland farms. Veterinary
Parasitology 93(2):103-112.
Fayer, R. 2004. Cryptosporidium: A water-borne zoonotic parasite. Veterinary Parasitology 126(1-
2):37-56.
Fazil, A.M. 1996. A Quantitative Risk Assessment Model for Salmonella. M.S. thesis. Drexel University,
Philadelphia, PA.
FDA (U.S. Food and Drug Administration). 2005. Quantitative Risk Assessment on the Public Health
Impact of Pathogenic Vibrio Parahaemolyticus in Raw Oysters. U.S. Department of Health and
Human Services, Food and Drug Administration, Center for Food Safety and Applied Nutrition,
College Park, MD.
FDA (U.S. Food and Drug Administration). 2012. Bad Bug Book: Handbook of Foodborne Pathogenic
Microorganisms and Natural Toxins. Accessed June 23, 2023.
https://www.fda.gov/food/foodborne-pathogens/bad-bug-book-second-edition.
Ferguson, C., A.M.D. de Roda Husman, N. Altavilla, D. Deere, and N. Ashbolt. 2003. Fate and transport
of surface water pathogens in watersheds. Critical Reviews in Environmental Science and
Technology 33(3):299-361.
Ferguson, C.M., K. Charles, and D.A. Deere. 2008. Quantification of microbial sources in drinking-water
catchments. Critical Reviews in Environmental Science and Technology 39(l):l-40.
Ferguson, D., and C. Signoretto. 2011. Environmental persistence and naturalization of fecal indicator
organisms. Chapter 17 in Microbial Source Tracking: Methods, Applications, and Case Studies. C.
Hagedorn, A.R. Blanch, V.J. Harwood, eds. 379-398 pp. Springer, New York, NY.
Ferguson, A., C. Del Donno, E. Obeng-Gyasi, K. Mena, T. Kaur Altomare, R. Guerrero, M. Gidley, L.
Montas, and H.M. Solo-Gabriele. 2019. Children exposure-related behavior patterns and risk
perception associated with recreational beach use. International Journal of Environmental
Research and Public Health 16(15):2783.
Ferguson, A., A. Dwivedi, F. Adelabu, E. Ehindero, M. Lamssali, E. Obeng-Gyasi, K. Mena, and H. Solo-
Gabriele. 2021. Quantified activity patterns for young children in beach environments relevant
for exposure to contaminants. International Journal of Environmental Research and Public
Health 18(6):3274.
Fewtrell, L., and D. Kay. 2015. Recreational water and infection: A review of recent findings. Current
Environmental Health Reports 2(l):85-94.
Fine, K.D., and M.J. Stone. 1999. Alpha-heavy chain disease, Mediterranean lymphoma, and
immunoproliferative small intestinal disease: A review of clinicopathological features,
pathogenesis, and differential diagnosis. American Journal of Gastroenterology 94(5):1139-
1152.
Flanagan, P.A. 1992. Giardia diagnosis, clinical course and epidemiology. A review. Epidemiology and
Infection 109(l):l-22.
142
-------
Fleisher, J.M., D. Kay, R.L. Salmon, F. Jones, M.D. Wyer, and A.F. Godfree. 1996. Marine waters
contaminated with domestic sewage: Nonenteric illnesses associated with bather exposure in
the United Kingdom. American Journal of Public Health 86(9):1228-1234.
Fleisher, J.M., D. Kay, M.D. Wyer, and A.F. Godfree. 1998. Estimates of the severity of illnesses
associated with bathing in marine recreational waters contaminated with domestic sewage.
International Journal of Epidemiology 27(4):722-726.
Fleisher, J.M., L.E. Fleming, H.M. Solo-Gabriele, J.K. Kish, C.D. Sinigalliano, L. Piano, S.M. Elmir, J.D.
Wang, K. Withum, T. Shibata, M.L. Gidley, A. Abdelzaher, G. He, C. Ortega, X. Zhu, M. Wright, J.
Hollenbeck, and L.C. Backer. 2010. The BEACHES Study: Health effects and exposures from non-
point source microbial contaminants in subtropical recreational marine waters. International
Journal of Epidemiology 39(5):1291-1298.
Fleming, L.E., H. Solo Gabriele, J.M. Fleisher, K. Goodwin, L. Backer, S. Elmir, and J. Wang. 2006. The
Pilot Epidemiologic Assessment of Microbial Indicators for Monitoring Recreational Water
Quality in Marine Sub/Tropical Environments (FL DOH External Funding Proposal). University of
Miami, NOVA Southeastern University, NOAA Southeast Miami Lab, Miami Dade County Public
Health Department, and Centers for Disease Control and Prevention (CDC).
Fleming, L.E., H. Solo Gabriele, J.M. Fleisher, S. Elmir, C. Sinigalliano, L. Piano, and J. Wang. 2008. Final
Report: The Pilot Epidemiologic Assessment of Microbial Indicators for Monitoring Recreational
Water Quality in Marine Sub/Tropical Environments. NSF NIEHS Oceans and Human Health
Center, University of Miami, Miami, FL.
Fogarty, L.R., S.K. Haack, M.J. Wolcott, and R.L. Whitman. 2003. Abundance and characteristics of the
recreational water quality indicator bacteria Escherichia coli and enterococci in gull faeces.
Journal of Applied Microbiology 94(5):865-878.
Fossler, C.P., S.J. Wells, J.B. Kaneene, P.L. Ruegg, L.D. Warnick, L.E. Eberly, S.M. Godden, L.W. Halbert,
A.M. Campbell, C.A. Bolin, and A.M. Zwald. 2005. Cattle and environmental sample-level factors
associated with the presence of Salmonella in a multi-state study of conventional and organic
dairy farms. Preventive Veterinary Medicine 67(l):39-53.
Frost, F., and G. Craun. 1998. The Importance of Acquired Immunity in the Epidemiology of
Cryptosporidiosis and Giardiasis. OECD Workshop Molecular Methods for Safe Drinking Water,
Interlaken, Switzerland. EPA and Organisation for Economic Co-operation and Development
(OECD).
Frost, F.J., M. Roberts, T.R. Kunde, G. Craun, K. Tollestrup, L. Harter, and T. Muller. 2005. How clean
must our drinking water be: The importance of protective immunity. The Journal of Infectious
Diseases 191(5):809-814.
Fujioka, R.S., and M.N. Byappanahalli, eds. 2003. Proceedings and Report: Tropical Water Quality
Indicator Workshop. University of Hawaii at Manoa, Water Resources Research Center,
Honolulu, HI.
143
-------
Gabriel, E., D.J. Wilson, A.J. Leatherbarrow, J. Cheesbrough, S. Gee, E. Bolton, A. Fox, P. Fearnhead, C.A.
Hart, and P.J. Diggle. 2010. Spatio-temporal epidemiology of Campylobacter jejuni enteritis, in
an area of Northwest England, 2000-2002. Epidemiology and Infection 138(10):1384-1390.
Gale, P. 2005. Land application of treated sewage sludge: Quantifying pathogen risks from
consumption of crops. Journal of Applied Microbiology 98(2):380-396.
Garcia-Aljaro, C., M. Muniesa, J. Jofre, and A.R. Blanch. 2004. Prevalence of the stx2 gene in coliform
populations from aquatic environments. Applied and Environmental Microbiology 70(6):3535-
3540.
Garcia-Aljaro, C., X. Bonjoch, and A.R. Blanch. 2005. Combined use of an immunomagnetic separation
method and immunoblotting for the enumeration and isolation of Escherichia coli 0157 in
wastewaters. Journal of Applied Microbiology 98(3):589-597.
Garg, A.X., R.S. Suri, N. Barrowman, F. Rehman, D. Matsell, M.P. Rosas-Arellano, M. Salvadori, R.B.
Haynes, and W.F. Clark. 2003. Long-term renal prognosis of diarrhea-associated hemolytic
uremic syndrome: A systematic review, meta-analysis, and meta-regression. The Journal of the
American Medical Association 290(10):1360-1370.
Gerba, C.P., J.B. Rose, C.N. Haas, and K.D. Crabtree. 1996. Waterborne rotavirus: A risk assessment.
Water Research 30(12):2929-2940.
Gibson, K.E. 2014. Viral pathogens in water: Occurrence, public health impact, and available control
strategies. Current Opinion in Virology 4:50-57.
Gilbert, M., P.C. Godschalk, M.F. Karwaski, C.W. Ang, A. van Belkum, J. Li, W.W. Wakarchuk, and H.P.
Endtz. 2004. Evidence for acquisition of the lipooligosaccharide biosynthesis locus in
Campylobacter jejuni GB11, a strain isolated from a patient with Guillain-Barre syndrome, by
horizontal exchange. Infection and Immunity 72(2):1162-1165.
Gilinsky, N.H., B.H. Novis, J.P. Wright, D.M. Dent, H. King, and I.N. Marks. 1987. Immunoproliferative
small-intestinal disease: Clinical features and outcome in 30 cases. Medicine (Baltimore)
66(6):438-446.
Glass, R.I., J. Bresee, B. Jiang, J. Gentsch, T. Ando, R. Fankhauser, J. Noel, U. Parashar, B. Rosen, and S.S.
Monroe. 2001. Gastroenteritis viruses: An overview. Novartis Foundation Symposium 238:5-19;
discussion 19-25.
Goh, S.G., N. Saeidi, X. Gu, G.G.R. Vergara, L. Liang, H. Fang, M. Kitajima, A. Kushmaro, and K.Y. Gin.
2019. Occurrence of microbial indicators, pathogenic bacteria and viruses in tropical surface
waters subject to contrasting land use. Water Research 150:200-215.
Goh, S.G., L. Liang, and K.Y.H. Gin. 2021. Assessment of human health risks in tropical environmental
waters with microbial source tracking markers. Water Research 207:117748.
Goodwin, K.D., A. Schriewer, A. Jirik, K. Curtis, and A. Crumpacker. 2017. Consideration of natural
sources in a bacteria TMDL-lines of evidence, including beach microbial source tracking.
Environmental Science & Technology 51(14):7775-7784.
144
-------
Goss, M., and C. Richards. 2008. Development of a risk-based index for source water protection
planning, which supports the reduction of pathogens from agricultural activity entering water
resources. Journal of Environmental Management 87(4):623-632.
Graciaa, D.S., J.R. Cope, V.A. Roberts, B.L. Cikesh, A.M. Kahler, M. Vigar, E.D. Hilborn, E.J. Wade, L.C.
Backer, S.P. Montgomery, W.E. Secor, V.R. Hill, M.J. Beach, K.E. Fullerton, J.S. Yoder, and M.C.
Hlavsa. 2018. Outbreaks associated with untreated recreational water—United States, 2000-
2014. Morbidity and Mortality Weekly Report 67(25):701-706.
Grant, S.B., C.P. Pendroy, C.L. Mayer, J.K. Bellin, and C.J. Palmer. 1996. Prevalence of
enterohemorrhagic Escherichia coli in raw and treated municipal sewage. Applied and
Environmental Microbiology 62(9):3466-3469.
Green, H.C., L.K. Dick, B. Gilpin, M. Samadpour, and K.G. Field. 2012. Genetic markers for rapid PCR-
based identification of gull, Canada goose, duck, and chicken fecal contamination in water.
Applied and Environmental Microbiology 78(2):503-510.
Green, H.C., K.M. White, C.A. Kelty, and O.C. Shanks. 2014. Development of rapid canine fecal source
identification PCR-based assays. Environmental Science & Technology 48(19):11453-11461.
Grimwood, K., R. Carzino, G.L. Barnes, and R.F. Bishop. 1995. Patients with enteric adenovirus
gastroenteritis admitted to an Australian pediatric teaching hospital from 1981 to 1992. Journal
of Clinical Microbiology 33(1):131-136.
Gulis, G., and Y. Fujino. 2015. Epidemiology, population health, and health impact assessment. Journal
of Epidemiology 25(3):179-180.
Gutierrez-Cacciabue, D., A.G. Cid, and V.B. Rajal. 2016. How long can culturable bacteria and total DNA
persist in environmental waters? The role of sunlight and solid particles. The Science of the
Total Environment 539:494-502.
Haack, S.K., L.R. Fogarty, and C. Wright. 2003. Escherichia coli and enterococci at beaches in the Grand
Traverse Bay, Lake Michigan: Sources, characteristics, and environmental pathways.
Environmental Science & Technology 37(15):3275-3282.
Haas, C.N., J.B. Rose, C. Gerba, and S. Regli. 1993. Risk assessment of virus in drinking water. Risk
Analysis 13(5):545-552.
Haas, C.N., C.S. Crockett, J.B. Rose, C.P. Gerba, and A.M. Fazil. 1996. Assessing the risk posed by oocysts
in drinking water. Journal of the American Water Works Association 88(9):131-136.
Haas, C.N., J.B. Rose, and C.P. Gerba. 1999. Quantitative Microbial Risk Assessment. J. W. Wiley, Inc.,
New York, NY.
Haas, C.N., J.B. Rose, and C.P. Gerba. 2014. Quantitative Microbial Risk Assessment. 2nd ed. J. W.
Wiley, Inc, Hoboken, NJ.
Haffejee, I.E. 1995. The epidemiology of rotavirus infections: A global perspective. Journal of Pediatric
Gastroenterology and Nutrition 20(3):275-286.
145
-------
Haile, R.W., J.S. Witte, M. Gold, R. Cressey, C. McGee, R.C. Millikan, A. Glasser, N. Harawa, C. Ervin, P.
Harmon, J. Harper, J. Dermand, J. Alamillo, K. Barrett, M. Nides, and G. Wang. 1999. The health
effects of swimming in ocean water contaminated by storm drain runoff. Epidemiology
10(4):355-363.
Halliday, E., and R.J. Gast. 2011. Bacteria in beach sands: An emerging challenge in protecting coastal
water quality and bather health. Environmental Science & Technology 45(2):370-379.
Hamilton, A.J., F. Stagnitti, R. Premier, A.M. Boland, and G. Hale. 2006. Quantitative microbial risk
assessment models for consumption of raw vegetables irrigated with reclaimed water. Applied
and Environmental Microbiology 72(5):3284-3290.
Hanninen, M.L., H. Haajanen, T. Pummi, K. Wermundsen, M.L. Katila, H. Sarkkinen, I. Miettinen, and H.
Rautelin. 2003. Detection and typing of Campylobacter jejuni and Campylobacter coli and
analysis of indicator organisms in three waterborne outbreaks in Finland. Applied and
Environmental Microbiology 69(3):1391-1396.
Hansen, D.L., S. Ishii, M.J. Sadowsky, and R.E. Hicks. 2011. Waterfowl abundance does not predict the
dominant avian source of beach Escherichia coli. Journal of Environmental Quality 40(6):1924-
1931.
Haramoto, E., H. Katayama, K. Oguma, H. Yamashita, A. Tajima, H. Nakajima, and S. Ohgaki. 2006.
Seasonal profiles of human noroviruses and indicator bacteria in a wastewater treatment plant
in Tokyo, Japan. Water Science and Technology 54(ll-12):301-308.
Haramoto, E., M. Kitajima, A. Hata, J.R. Torrey, Y. Masago, D. Sano, and H. Katayama. 2018. A review on
recent progress in the detection methods and prevalence of human enteric viruses in water.
Water Research 135:168-186.
Hauchman, F. 2008. Assessing Exposure to Waterborne Pathogens. Air & Waste Management
Association, Pittsburgh, PA.
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.
Health Canada. 2012. Guidelines for Canadian Recreational Water Quality. 3rd ed. Catalogue No H129-
15/2012E. Health Canada, Water, Air and Climate Change Bureau, Healthy Environments and
Consumer Safety Branch, Ottawa, Ontario.
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.
Heitman, T.L., L.M. Frederick, J.R. Viste, N.J. Guselle, U.M. Morgan, R.C. Thompson, and M.E. Olson.
2002. Prevalence of Giardia and Cryptosporidium and characterization of Cryptosporidium spp.
isolated from wildlife, human, and agricultural sources in the North Saskatchewan River Basin in
Alberta, Canada. Canadian Journal of Microbiology 48(6):530-541.
146
-------
Hellard, M.E., M.I. Sinclair, G.G. Hogg, and C.K. Fairley. 2000. Prevalence of enteric pathogens among
community based asymptomatic individuals. Journal of Gastroenterology and Hepatology
15(3):290-293.
Helsel, D.R. 2012. Statistics for Censored Environmental Data Using Minitab and R. 2nd ed. John Wiley
& Sons, Inc., Hoboken, NJ.
Hess, S., R. Niessner, and M. Seidel. 2021. Quantitative detection of human adenovirus from river
water by monolithic adsorption filtration and quantitative PCR. Journal of Virological Methods
292:114128.
Hlavsa, M.C., V.A. Roberts, A.M. Kahler, E.D. Hilborn, T.R. Mecher, M.J. Beach, T.J. Wade, and J.S.
Yoder. 2015. Outbreaks of Illness Associated with Recreational Water—United States, 2011-
2012. Morbidity and Mortality Weekly Report 64(24):668-672.
Hoar, B.R., E.R. Atwill, C. Elmi, and T.B. Farver. 2001. An examination of risk factors associated with
beef cattle shedding pathogens of potential zoonotic concern. Epidemiology and Infection
127(1):147-155.
Holcomb, D.A., and J.R. Stewart. 2020. Microbial Indicators of Fecal Pollution: Recent Progress and
Challenges in Assessing Water Quality. Current Environmental Health Reports 7(3):311-324.
Holme, R. 2003. Drinking water contamination in Walkerton, Ontario: Positive resolutions from a tragic
event. Water Science and Technology 47(3):l-6.
Hopkins, R.S., G.B. Gaspard, F.P. Williams, Jr., R.J. Karlin, G. Cukor, and N.R. Blacklow. 1984. A
community waterborne gastroenteritis outbreak: Evidence for rotavirus as the agent. American
Journal of Public Health 74(3):263-265.
Hopkins, R.S., P. Shillam, B. Gaspard, L. Eisnach, and R.J. Karlin. 1985. Waterborne disease in Colorado:
three years' surveillance and 18 outbreaks. American Journal of Public Health 75(3):254-257.
Hutchison, M.L., L.D. Walters, S.M. Avery, B.A. Synge, and A. Moore. 2004. Levels of zoonotic agents in
British livestock manures. Letters in Applied Microbiology 39(2):207-214.
Hutson, A.M., R.L. Atmar, and M.K. Estes. 2004. Norovirus disease: Changing epidemiology and host
susceptibility factors. Trends in Microbiology 12(6):279-287.
Ichida, A., S. Schaub, J. Soller, S. Nappier, and J. Ravenscroft. 2016. Microbial risk assessment tools,
methods, and approaches for water media. Microbial Risk Analysis 1:12-12.
ILSI (International Life Sciences Institute). 2000. Revised Framework for Microbial Risk Assessment.
International Life Sciences Institute, Washington, DC.
Jahne, M.A., M.E. Schoen, A. Kaufmann, B.M. Pecson, A. Olivieri, S. Sharvelle, A. Anderson, N.J. Ashbolt,
and J.L. Garland. 2023. Enteric pathogen reduction targets for onsite non-potable water
systems: A critical evaluation. Water Research 233:119742.
Jameel, S. 1999. Molecular biology and pathogenesis of hepatitis E virus. Expert Reviews in Molecular
Medicine 1999:1-16.
147
-------
Jiang, S.C. 2006. Human adenoviruses in water: Occurrence and health implications: a critical review.
Environmental Science & Technology 40(23):7132-7140.
Jimenez-Cisneros, B.E., C. Maya-Rendon, and G. Salgado-Velazquez. 2001. The elimination of helminth
ova, faecal coliforms, Salmonella and protozoan cysts by various physicochemical processes in
wastewater and sludge. Water Science and Technology 43(12): 179-182.
Jones, I.G., and M. Roworth. 1996. An outbreak of Escherichia coli 0157 and campylobacteriosis
associated with contamination of a drinking water supply. Public Health 110(5):277-282.
Kaldor, J., and B.R. Speed. 1984. Guillain-Barre syndrome and Campylobacter jejuni: A serological
study. British Medical Journal (Clinical Research Ed.) 288(6434):1867-1870.
Katayama, H., E. Haramoto, K. Oguma, H. Yamashita, A. Tajima, H. Nakajima, and S. Ohgaki. 2008. One-
year monthly quantitative survey of noroviruses, enteroviruses, and adenoviruses in
wastewater collected from six plants in Japan. Water Research 42(6-7):1441-1448.
Katukiza, A.Y., M. Ronteltap, P. van der Steen, J.W. Foppen, and P.N. Lens. 2014. Quantification of
microbial risks to human health caused by waterborne viruses and bacteria in an urban slum.
Journal of Applied Microbiology 116(2):447-463.
Kay, D., J.M. Fleisher, R.L. Salmon, F. Jones, M.D. Wyer, A.F. Godfree, Z. Zelenauch-Jacquotte, and R.
Shore. 1994. Predicting likelihood of gastroenteritis from sea bathing: Results from randomised
exposure. Lancet 344(8927):905-909.
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.
Keiser, P.T., M. Anantpadma, H. Staples, R. Carrion, and R.A. Davey. 2021. Automation of infectious
focus assay for determination of filovirus titers and direct comparison to plaque and TCID(50)
assays. Microorganisms 9(1): 156.
Kelly, H., and C. Birch. 2004. The causes and diagnosis of influenza-like illness. Australian Family
Physician 33(5):305-309.
Kim, S.K., J.Y. An, M.S. Park, and B.J. Kim. 2007. A case report of Reiter's syndrome with progressive
myelopathy. Journal of Clinical Neurology 3(4):215-218.
King County Department of Natural Resources. 1999. King County Combined Sewer Overflow Water
Quality Assessment for the Duwamish River and Elliott Bay. Volume 1: Overview and
Interpretation. King County Department of Natural Resources, Wastewater Treatment Division
& Water and Land Resources Division, Seattle, WA.
Ko, G., T.L. Cromeans, and M.D. Sobsey. 2003. Detection of infectious adenovirus in cell culture by
mRNA reverse transcription-PCR. Applied and Environmental Microbiology 69(12):7377-7384.
Kolling, G., M. Wu, and R.L. Guerrant. 2012. Enteric pathogens through life stages. Frontiers in Cellular
Infection Microbiology 2:114.
148
-------
Koopman, J.S., S.E. Chick, C.P. Simon, C.S. Riolo, and G. Jacquez. 2002. Stochastic effects on endemic
infection levels of disseminating versus local contacts. Mathematical Biosciences 180:49-71.
Korajkic, A., B.R. McMinn, and V.J. Harwood. 2018. Relationships between microbial indicators and
pathogens in recreational water settings. International Journal of Environmental Research and
Public Health 15(12):2842.
Korajkic, A., B.R. McMinn, N.J. Ashbolt, M. Sivaganesan, V.J. Harwood, and O.C. Shanks. 2019. Extended
persistence of general and cattle-associated fecal indicators in marine and freshwater
environment. The Science of the Total Environment 650(Pt 1):1292-1302.
Korajkic, A., J. Kelleher, O.C. Shanks, M.P. Herrmann, and B.R. McMinn. 2022. Effectiveness of two
wastewater disinfection strategies for the removal of fecal indicator bacteria, bacteriophage,
and enteric viral pathogens concentrated using dead-end hollow fiber ultrafiltration (D-HFUF).
The Science of the Total Environment 831:154861.
Kraft, D.J., C. Olechowski-Gerhardt, J. Berkowitz, M.S. Finstein. 1969. Salmonella in wastes produced at
commercial poultry farms. Applied Microbiology 18(5):703-707.
Ksoll, W.B., S. Ishii, M.J. Sadowsky, and R.E. Hicks. 2007. Presence and sources of fecal coliform bacteria
in epilithic periphyton communities of Lake Superior. Applied and Environmental Microbiology
73(12) :3771-3778.
Lai, A., S. Hales, N. French, and M.G. Baker. 2012. Seasonality in human zoonotic enteric diseases: A
systematic review. PLoS One 7(4):e31883.
Lamparelli, C.C., K. Pogreba-Brown, M. Verhougstraete, M.I. Sato, A. de Castro Bruni, T.J. Wade, and
J.N. Eisenberg. 2015. Are fecal indicator bacteria appropriate measures of recreational water
risks in the tropics: A cohort study of beach goers in Brazil? Water Research 87:59-68.
Lanciers, S., B. Despinasse, D.I. Mehta, and U. Blecker. 1999. Increased susceptibility to Helicobacter
pylori infection in pregnancy. Infectious Diseases in Obstetrics and Gynecology 7(4):195-198.
Lecuit, M., E. Abachin, A. Martin, C. Poyart, P. Pochart, F. Suarez, D. Bengoufa, J. Feuillard, A. Lavergne,
J.I. Gordon, P. Berche, L. Guillevin, and O. Lortholary. 2004. Immunoproliferative small intestinal
disease associated with Campylobacter jejuni. New England Journal of Medicine 350(3):239-
248.
Lee, C.S., C. Lee, J. Marion, Q. Wang, L. Saif, and J. Lee. 2014. Occurrence of human enteric viruses at
freshwater beaches during swimming season and its link to water inflow. The Science of the
Total Environment 472:757-766.
Lemarchand, K., and P. Lebaron. 2003. Occurrence of Salmonella spp and Cryptosporidium spp in a
French coastal watershed: Relationship with fecal indicators. FEMS Microbiology Letters
218(l):203-209.
Levesque, B., P. Brousseau, F. Bernier, E. Dewailly, and J. Joly. 2000. Study of the bacterial content of
ring-billed gull droppings in relation to recreational water quality. Water Research 34(4):1089-
1096.
149
-------
Levett, P.N. 2001. Leptospirosis. Clinical Microbiology Reviews 14(2):296-326.
Ley, D.H., M.G. Levy, L. Hunter, W. Corbett, and H.J. Barnes. 1988. Cryptosporidia-positive rates of
avian necropsy accessions determined by examination of auramine O-stained fecal smears.
Avian Diseases 32(1):108-113.
Li, L., N. Mendis, H. Trigui, J.D. Oliver, and S.P. Faucher. 2014. The importance of the viable but non-
culturable state in human bacterial pathogens. Frontiers in Microbiology 5:258.
Li, X., M. Sivaganesan, C.A. Kelty, A. Zimmer-Faust, P. Clinton, J.R. Reichman, Y. Johnson, W. Matthews,
S. Bailey, and O.C. Shanks. 2019. Large-scale implementation of standardized quantitative real-
time PCR fecal source identification procedures in the Tillamook Bay Watershed. PLoS One
14(6):e0216827.
Li, X., C.A. Kelty, M. Sivaganesan, and O.C. Shanks. 2021. Variable fecal source prioritization in
recreational waters routinely monitored with viral and bacterial general indicators. Water
Research 192:116845.
Litton, R.M., J.H. Ahn, B. Sercu, P.A. Holden, D.L. Sedlak, and S.B. Grant. 2010. Evaluation of chemical,
molecular, and traditional markers of fecal contamination in an effluent dominated urban
stream. Environmental Science & Technology 44(19):7369-7375.
Liu, L., M.S. Phanikumar, S.L. Molloy, R.L. Whitman, D.A. Shively, M.B. Nevers, D.J. 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 & Technology 40(16):5022-5028.
Locht, H., and K.A. Krogfelt. 2002. Comparison of rheumatological and gastrointestinal symptoms after
infection with Campylobacter jejuni/coli and enterotoxigenic Escherichia coli. Annals of the
Rheumatic Diseases 61(5):448-452.
Lu, J., H. Ryu, J.W. Santo Domingo, J.F. Griffith, and N. Ashbolt. 2011. Molecular detection of
Campylobacter spp. in California gull (Larus californicus) excreta. Applied and Environmental
Microbiology 77(14):5034-5039.
Ludwig, K., V. Sarkim, M. Bitzan, M.A. Karmali, C. Bobrowski, H. Ruder, R. Laufs, I. Sobottka, M. Petric,
H. Karch, and D.E. Muller-Wiefel. 2002. Shiga toxin-producing Escherichia coli infection and
antibodies against Stx2 and Stxl in household contacts of children with enteropathic hemolytic-
uremic syndrome. Journal of Clinical Microbiology 40(5):1773-1782.
Maddox-Hyttel, C., Langkjaer, R.B., Enemark, H.L., Vigre, H. 2006. Cryptosporidium and Giardia in
different age groups of Danish cattle and pigs—occurrence and management associated risk
factors. Veterinary Parasitology 141(l-2):48-59.
Makri, A., R. Modarres, and R. Parkin. 2004. Cryptosporidiosis susceptibility and risk: A case study. Risk
Analysis 24(l):209-220.
Mandell, G.L., J.E. Bennett, R. Donlin, and R.G. Douglas, eds. 2000. Mandell, Douglas, and Bennett's
Principles and Practice of Infectious Diseases. Churchill Livingstone, Philadelphia, PA.
150
-------
Marion, J.W., C. Lee, C.S. Lee, Q. Wang, S. Lemeshow, T.J. Buckley, L.J. Saif, and J. Lee. 2014. Integrating
bacterial and viral water quality assessment to predict swimming-associated illness at a
freshwater beach: A cohort study. PLoS One 9(ll):ell2029.
Martin, S.A., M.A. McCann, and W.D. Waltman. 1998. Microbiological survey of Georgia poultry litter.
Journal of Applied Poultry Research 7(l):90-98.
Martinez, G., Y.A. Pachepsky, G. Whelan, A.M. Yakirevich, A. Guber, and T.J. Gish. 2014. Rainfall-
induced fecal indicator organisms transport from manured fields: Model sensitivity analysis.
Environment International 63:121-129.
Matthews, J.E., B.W. Dickey, R.D. Miller, J.R. Felzer, B.P. Dawson, A.S. Lee, J.J. Rocks, J. Kiel, J.S. Montes,
C.L. Moe, J.N. Eisenberg, and J.S. Leon. 2012. The epidemiology of published norovirus
outbreaks: A review of risk factors associated with attack rate and genogroup. Epidemiology
and Infection 140(7):1161-1172.
Maunula, L., I.T. Miettinen, and C.H. von Bonsdorff. 2005. Norovirus outbreaks from drinking water.
Emerging Infectious Diseases 11(11) :1716-1721.
McBride, G.B., Salmond, C.E., Bandaranayake, D.R., Turner, S.J., Lewis, G.D., Till, D.G. 1998. Health
effects of marine bathing in New Zealand. International Journal of Environmental Health
Research 8(3): 173-189.
McBride, G., D. Till, T. Ryan, A. Ball, G. Lewis, S. Palmer, and P. Weinstein. 2002. Freshwater
Microbiology Research Programme Report: Pathogen Occurrence and Human Health Risk
Assessment Analysis. Ministry for the Environment, Wellington, NZ.
McBride, G.B. 2004. Quantitative microbial risk assessment issues. Chapter 29 in Waterborne
Zoonoses: Identification, Causes, and Control. J.A. Cortruvo, A.P. Dufour, G. Rees, J. Bartram, R.
Carr, D.O. Cliver, G.F. Craun, R. Fayer, V.P.J. Gannon, eds. 460-470 pp. IWA Publishing for the
World Health Organization, London, UK:.
McBride, G.B., R. Stott, W. Miller, D. Bambic, and S. Wuertz. 2013. Discharge-based QMRA for
estimation of public health risks from exposure to stormwater-borne pathogens in recreational
waters in the United States. Water Research 47(14):5282-5297.
McCarthy, N., and J. Giesecke. 2001. Incidence of Guillain-Barre syndrome following infection with
Campylobacter jejuni. American Journal of Epidemiology 153(6):610-614.
McCullough, N.B., and C.W. Eisele. 1951a. Experimental human salmonellosis. I. Pathogenicity of
strains of Salmonella meleagridis and Salmonella anatum obtained from spray-dried whole egg.
The Journal of Infectious Diseases 88(3):278-289.
McCullough, N.B., and C.W. Eisele. 1951b. Experimental human salmonellosis. II. Immunity studies
following experimental illness with Salmonella meleagridis and Salmonella anatum. Journal of
Immunology 66(5):595-608.
McCullough, N.B., and C.W. Eisele. 1951c. Experimental human salmonellosis. III. Pathogenicity of
strains of Salmonella newport, Salmonella derby, and Salmonella bareilly obtained from spray-
dried whole egg. The Journal of Infectious Diseases 89(3):209-213.
151
-------
McKee, B.A., M. Molina, M. Cyterski, and A. Couch. 2020. Microbial source tracking (MST) in
Chattahoochee River National Recreation Area: Seasonal and precipitation trends in MST
marker concentrations, and associations with E. coli levels, pathogenic marker presence, and
land use. Water Research 171:115435.
McKee, A.M., P.M. Bradley, D. Shelley, S. McCarthy, and M. Molina. 2021. Feral swine as sources of
fecal contamination in recreational waters. Scientific Reports 11(1):4212.
McMinn, B.R., S. Klemm, A. Korajkic, K.M. Wyatt, M.P. Herrmann, R.A. Haugland, J. Lu, E.N. Villegas,
and C. Frye. 2019. A constructed wetland for treatment of an impacted waterway and the
influence of native waterfowl on its perceived effectiveness. Ecological Engineering 128:48-56.
Mead, P.S., L. Slutsker, V. Dietz, L.F. McCaig, J.S. Bresee, C. Shapiro, P.M. Griffin, and R.V. Tauxe. 1999.
Food-related illness and death in the United States. Emerging Infectious Diseases 5(5):607-625.
Medema, G.J., P.F. Teunis, A.H. Havelaar, and C.N. Haas. 1996. Assessment of the dose-response
relationship of Campylobacter jejuni. International Journal of Food Microbiology 30(1-2):101-
111.
Mena, K.D., and C.P. Gerba. 2009. Waterborne adenovirus. Reviews of Environmental Contamination
and Toxicology 198:133-167.
Messner, M.J., C.L. Chappell, and P.C. Okhuysen. 2001. Risk assessment for Cryptosporidium: A
hierarchical Bayesian analysis of human dose response data. Water Research 35(16):3934-
3940.
Messner, M.J., and P. Berger. 2016. Cryptosporidium Infection Risk: Results of New Dose-Response
Modeling. Risk Analysis 36(10):1969-1982.
Metcalf & Eddy, Inc. 2003. Wastewater Engineering: Treatment and Reuse. 4th edition. Tchobanoglous,
G., F.L. Burton, and H.D. Stensel, eds. McGraw-Hill, Boston, MA.
Mieszkin, S., J.P. Furet, G. Corthier, and M. Gourmelon. 2009. Estimation of pig fecal contamination in a
river catchment by real-time PCR using two pig-specific Bacteroidales 16S rRNA genetic
markers. Applied and Environmental Microbiology 75(10):3045-3054.
Mieszkin, S., J.F. Yala, R. Joubrel, and M. Gourmelon. 2010. Phylogenetic analysis of Bacteroidales 16S
rRNA gene sequences from human and animal effluents and assessment of ruminant faecal
pollution by real-time PCR. Journal of Applied Microbiology 108(3):974-984.
Minnesota Department of Health. 2022. Escherichia coli (E. coli). Accessed March 9, 2022.
https://www.health.state.mn.us/diseases/ecoli/index.html.
Mintz, E.D., M. Hudson-Wragg, P. Mshar, M.L. Cartter, and J.L. Hadler. 1993. Food borne giardiasis in a
corporate office setting. The Journal of Infectious Diseases 167(l):250-253.
M0lbak, K., and F. Scheutz. 2004. Verocytotoxin-producing Escherichia coli and Other Diarrhoeagenic E.
coli. Chapter 13 in Waterborne Zoonoses: Identification, Causes, and Control. J.A. Cortruvo, A.P.
Dufour, G. Rees, J. Bartram, R. Carr, D.O. Cliver, G.F. Craun, R. Fayer, and V.P.J. Gannon, eds.
213-227 pp. IWA Publishing for the World Health Organization, London, UK.
152
-------
Moon, H., J.J. Chen, D.W. Gaylor, and R.L. Kodell. 2004. A comparison of microbial dose-response
models fitted to human data. Regulatory Toxicology and Pharmacology 40(2):177-184.
Moriarty, E.M., L.W. Sinton, M.L. Mackenzie, N. Karki, and D.R. Wood. 2008. A survey of enteric
bacteria and protozoans in fresh bovine faeces on New Zealand dairy farms. Journal of Applied
Microbiology 105(6) :2015-2025.
Morris, J.G., Jr., and M. Potter. 1997. Emergence of new pathogens as a function of changes in host
susceptibility. Emerging Infectious Diseases 3(4):435-441.
Mosnier, E., N. Martin, R. Razakandrainibe, F. Dalle, G. Roux, A. Buteux, L. Favennec, P. Brousse, B.
Guarmit, D. Blanchet, L. Epelboin, C. Girouin, E. Martin, F. Djossou, M. Nacher, and M. Demar.
2018. Cryptosporidiosis outbreak in immunocompetent children from a remote area of French
Guiana. The American Journal of Tropical Medicine and Hygiene 98(6):1727-1732.
Motarjemi, Y. 2002. Chronic sequelae of foodborne infections. In Food borne Pathogens. C.D. Blackburn
P.J. and McClure, eds. 501-513 pp. CRC Press, Boca Raton, FL.
Moyer, N., and A. Degnan. 2006. Shigella. Chapter 19 in Waterborne Pathogens: Manual of Water
Supply Practices. AWWA Manual M48. 2nd ed. 145-148 pp. American Waterworks
Association, Denver, CO.
Muniesa, M., J. Jofre, C. Garcfa-Aljaro, and A.R. Blanch. 2006. Occurrence of Escherichia coli 0157:H7
and other enterohemorrhagic Escherichia coli in the environment. Environmental Science &
Technology 40(23):7141-7149.
Murphy, H.M., M.K. Thomas, P.J. Schmidt, D.T. Medeiros, F.S. Mc, and K.D. Pintar. 2016. Estimating the
burden of acute gastrointestinal illness due to Giardia, Cryptosporidium, Campylobacter, E. coli
0157 and norovirus associated with private wells and small water systems in Canada.
Epidemiology and Infection 144(7):1355-1370.
Murtagh, P., C. Cerqueiro, A. Halac, M. Avila, and A. Kajon. 1993. Adenovirus type 7h respiratory
infections: A report of 29 cases of acute lower respiratory disease. Acta Paediatrica 82(6-
7):557-561.
Nachamkin, I., B.M. Alios, and T.W. Ho. 2000. Campylobacter jejuni infection and the association with
Guillain-Barre' syndrome. In Campylobacter 2000. 2nd ed. I. Nachamkin and M.J. Blaser, eds.
155-178 pp. American Society for Microbiology (ASM) Press, Washington, DC.
Nappier, S.P., J.A. Soller, and S.E. Eftim. 2018. Potable water reuse: What are the microbiological risks?
Current Environmental Health Reports 5(2):283-292.
Nevers, M.B., P.M. Buszka, M.N. Byappanahalli, T. Cole, S.R. Corsi, P.R. Jackson, J.L. Kinzelman, C.H.
Nakatsu, and M.S. Phanikumar. 2022. Microbial source tracking and evaluation of best
management practices for restoring degraded beaches of Lake Michigan. Journal of Great Lakes
Research 48(2):441-454.
153
-------
NHMRC (National Health and Medical Research Council). 2008. Guidelines for Managing Risks in
Recreational Water. National Health and Medical Research Council, Australian Government,
Canberra, ACT. Accessed June 23, 2023.
https://www.nhmrc.gov.au/sites/default/files/images/guidelines-for-managing-risks-in-
recreational-water.pdf.
NOAA (National Oceanic and Atmospheric Administration). 2022. Point Source Pollution Tutorial.
Accessed October 17, 2022.
https://oceanservice.noaa.gov/education/tutorial pollution/03pointsource.html.
Nordgren, J., and L. Svensson. 2019. Genetic susceptibility to human norovirus infection: An update.
Viruses 11(3):226.
NRC (National Research Council). 2004. Indicators for Waterborne Pathogens. National Academy Press,
National Research Council, Washington, DC.
New Zealand. 2003. Microbiological Water Quality Guidelines for Marine and Freshwater Recreational
Areas. New Zealand Ministry for the Environment, Wellington, NZ.
O'Donoghue, P.J. 1995. Cryptosporidium and cryptosporidiosis in man and animals. International
Journal for Parasitology 25(2):139-195.
Okhuysen, P.C., C.L. Chappell, C.R. Sterling, W. Jakubowski, and H.L. DuPont. 1998. Susceptibility and
serologic response of healthy adults to reinfection with Cryptosporidium parvum. Infection and
Immunity 66(2):441-443.
Okhuysen, P.C., C.L. Chappell, J.H. Crabb, C.R. Sterling, and H.L. DuPont. 1999. Virulence of three
distinct Cryptosporidium parvum isolates for healthy adults. The Journal of Infectious Diseases
180(4) :1275-1281.
Okhuysen, P.C., S.M. Rich, C.L. Chappell, K.A. Grimes, G. Widmer, X. Feng, and S. Tzipori. 2002.
Infectivity of a Cryptosporidium parvum isolate of cervine origin for healthy adults and
interferon-gamma knockout mice. The Journal of Infectious Diseases 185(9):1320-1325.
Oliver, J.D. 2005. Wound infections caused by Vibrio vulnificus and other marine bacteria. Epidemiology
and Infection 133(3):383-391.
Ottoson, J., and T.A. Stenstrom. 2003. Faecal contamination of greywater and associated microbial
risks. Water Research 37(3):645-655.
Parkhurst, D.F., G.F. Craun, and J.A. Soller. 2007. Conceptual bases for relating illness risk to indicator
concentrations. Chapter 3 in Statistical Framework for Recreational Water Quality Criteria and
Monitoring. L.J. Wymer, ed. 19-43 pp. John Wiley and Sons, Ltd., Chichester, UK.
Parkin, R.T., J.A. Soller, and A.W. Olivieri. 2003. Incorporating susceptible subpopulations in microbial
risk assessment: Pediatric exposures to enteroviruses in river water. Journal of Exposure
Analysis and Environmental Epidemiology 13(2):161-168.
Parkin, R.T. 2004. Addressing susceptibility in microbial pathogen risk assessment. Human and
Ecological Risk Assessment 10(1):135-149.
154
-------
Peed, L.A., C.T. Nietch, C.A. Kelty, M. Meckes, T. Mooney, M. Sivaganesan, and O.C. Shanks. 2011.
Combining land use information and small stream sampling with PCR-based methods for better
characterization of diffuse sources of human fecal pollution. Environmental Science &
Technology 45(13):5652-5659.
Petrinca, A.R., D. Donia, A. Pierangeli, R. Gabrieli, A.M. Degener, E. Bonanni, L. Diaco, G. Cecchini, P.
Anastasi, and M. Divizia. 2009. Presence and environmental circulation of enteric viruses in
three different wastewater treatment plants. Journal of Applied Microbiology 106(5):1608-
1617.
Petterson, S.R., and N.J. Ashbolt. 2001. Viral risks associated with wastewater reuse: Modeling virus
persistence on wastewater irrigated salad crops. Water Science and Technology 43(12):23-26.
Petterson, S.R., R.S. Signor, and N.J. Ashbolt. 2007. Incorporating method recovery uncertainties in
stochastic estimates of raw water protozoan concentrations for QMRA. Journal of Water and
Health 5(Suppl l):51-65.
Petterson, S.R., N. Dumoutier, J.F. Loret, and N.J. Ashbolt. 2009. Quantitative Bayesian predictions of
source water concentration for QMRA from presence/absence data for E. coli 0157:H7. Water
Science and Technology 59(ll):2245-2252.
Petterson, S.R., and N.J. Ashbolt. 2016. QMRA and water safety management: Review of application in
drinking water systems. Journal of Water and Health 14(4):571-589.
Peu, P., H. Brugere, A.M. Pourcher, M. Kerouredan, J.J. Godon, J.P. Delgenes, and P. Dabert. 2006.
Dynamics of a pig slurry microbial community during anaerobic storage and management.
Applied and Environmental Microbiology 72(5):3578-3585.
Piggot, A.M., J.S. Klaus, S. Johnson, M.C. Phillips, and H.M. Solo-Gabriele. 2012. Relationship between
enterococcal levels and sediment biofilms at recreational beaches in South Florida. Applied and
Environmental Microbiology 78(17):5973-5982.
Plowright, R.K., C.R. Parrish, H. McCallum, P.J. Hudson, A.I. Ko, A.L. Graham, and J.O. Lloyd-Smith. 2017.
Pathways to zoonotic spillover. Nature Reviews. Microbiology 15(8):502-510.
Plutzer, J., J. Ongerth, and P. Karanis. 2010. Giardia taxonomy, phylogeny and epidemiology: Facts and
open questions. International Journal of Hygiene and Environmental Health 213(5):321-333.
Pond, K. 2005. Water Recreation and Disease. Plausibility of Associated Infections: Acute Effects,
Sequelae and Mortality. IWA Publishing for the World Health Organization, London, UK.
Pope, J.E., A. Krizova, and A.X. Garg, H. Thiessen-Philbrook, and J.M. Ouimet. 2007. Campylobacter
reactive arthritis: A systematic review. Seminars in Arthritis and Rheumatism 37(l):48-55.
Priiss, A. 1998. Review of epidemiological studies on health effects from exposure to recreational
water. International Journal of Epidemiology 27(l):l-9.
Radke, J.R., and J.L. Cook. 2018. Human adenovirus infections: Update and consideration of
mechanisms of viral persistence. Current Opinion in Infectious Disease 31(3):251-256.
155
-------
Raith, M.R., C.A. Kelty, J.F. Griffith, A. Schriewer, S. Wuertz, S. Mieszkin, M. Gourmelon, G.H. Reischer,
A.H. Farnleitner, J.S. Ervin, P.A. Holden, D.L. Ebentier, J.A. Jay, D. Wang, A.B. Boehm, T.G. Aw,
J.B. Rose, E. Balleste, W.G. Meijer, M. Sivaganesan, and O.C. Shanks. 2013. Comparison of PCR
and quantitative real-time PCR methods for the characterization of ruminant and cattle fecal
pollution sources. Water Research 47(18):6921-6928.
Rambaud, J.C., M. Halphen, A. Galian, and A. Tsapis. 1990. Immunoproliferative small intestinal disease
(IPSID): Relationships with alpha-chain disease and "Mediterranean" lymphomas. Springer
Seminars in Immunopathology 12(2-3):239-250.
Rames, E., A. Roiko, H. Stratton, and J. Macdonald. 2016. Technical aspects of using human adenovirus
as a viral water quality indicator. Water Research 96:308-326.
Rangel, J.M., P.H. Sparling, C. Crowe, P.M. Griffin, and D.L. Swerdlow. 2005. Epidemiology of
Escherichia coli 0157:1-17 outbreaks, United States, 1982-2002. Emerging Infectious Diseases
ll(4):603-609.
Regli, S., J.B. Rose, C.N. Haas, and C.P. Gerba. 1991. Modeling the risk from Giardia and viruses in
drinking-water. Journal of the American Water Works Association 83(ll):76-84.
Rendtorff, R.C. 1954a. The experimental transmission of human intestinal protozoan parasites. I.
Endamoeba coli cysts given in capsules. American Journal of Hygiene 59(2):196-208.
Rendtorff, R.C. 1954b. The experimental transmission of human intestinal protozoan parasites. II.
Giardia lamblia cysts given in capsules. American Journal of Hygiene 59(2):209-220.
Rendtorff, R.C., and C.J. Holt. 1954a. The experimental transmission of human intestinal protozoan
parasites. III. Attempts to transmit Endamoeba coli and Giardia lamblia cysts by flies. American
Journal of Hygiene 60(3):320-326.
Rendtorff, R.C., and C.J. Holt. 1954b. The experimental transmission of human intestinal protozoan
parasites. IV. Attempts to transmit Endamoeba coli and Giardia lamblia cysts by water.
American Journal of Hygiene 60(3):327-338.
Riley, S., C. Fraser, C.A. Donnelly, A.C. Ghani, L.J. Abu-Raddad, A.J. Hedley, G.M. Leung, L.M. Ho, T.H.
Lam, T.Q. Thach, P. Chau, K.P. Chan, S.V. Lo, P.Y. Leung, T. Tsang, W. Ho, K.H. Lee, E.M. Lau,
N.M. Ferguson, and R.M. Anderson. 2003. Transmission dynamics of the etiological agent of
SARS in Hong Kong: Impact of public health interventions. Science 300(5627):1961-1966.
Riverside County Health Department, California State Department, Centers for Disease Control and
Prevention, and National Center for Urban and Industrial Health. 1971. A waterborne epidemic
of salmonellosis in Riverside, California, 1965. Epidemiologic aspects. A collaborative report.
American Journal of Epidemiology 93(l):33-48.
Roberts, B.N., R.H. Bailey, M.R. McLaughlin, and J.P. Brooks. 2016. Decay rates of zoonotic pathogens
and viral surrogates in soils amended with biosolids and manures and comparison of qPCR and
culture derived rates. The Science of the Total Environment 573:671-679.
Robertson, L.J., L. Hermansen, and B.K. Gjerde. 2006. Occurrence of Cryptosporidium oocysts and
Giardia cysts in sewage in Norway. Applied and Environmental Microbiology 72(8):5297-5303.
156
-------
Rodriguez-Hernandez, J., A. Canut-Blasco, and A.M. Martin-Sanchez. 1996. Seasonal prevalences of
Cryptosporidium and Giardia infections in children attending day care centres in Salamanca
(Spain) studied for a period of 15 months. European Journal of Epidemiology 12(3):291-295.
Rodriguez, F., J. Moreno, R. Ortega, C. Mathieu, A. Garcia, F. Cerda-Leal, and D. Gonzalez-Acuna. 2012.
Evidence for Kelp Gulls (Larus dominicanus) and Franklin's Gulls (Leucophaeus pipixcan) as
carriers of Salmonella by real-time polymerase chain reaction. Journal of Wildlife Diseases
48(4):1105-1108.
Rogers, S.W., M. Donnelly, L. Peed, C.A. Kelty, S. Mondal, Z. Zhong, and O.C. Shanks. 2011. Decay of
bacterial pathogens, fecal indicators, and real-time quantitative PCR genetic markers in
manure-amended soils. Applied and Environmental Microbiology 77(14):4839-4848.
Rose, J.B., and C.P. Gerba. 1991. Use of risk assessment for development of microbial standards. Water
Science and Technology 24(2):29-34.
Rose, J.B., C.N. Haas, and S. Regli. 1991. Risk assessment and control of waterborne giardiasis.
American Journal of Public Health 81(6):709-713.
Rose, J.B., S.R. Farrah, V.J. Harwood, A.D. Levine, J. Lukasik, P. Menendez, and T.M. Scott. 2004.
Reduction of Pathogens, Indicator Bacteria, and Alternative Indicators by Wastewater
Treatment and Reclamation Processes: Water for Reuse Final Report 2004. 00-PUM-2T. IWA
Publishing and Water Environment Research Foundation (WERF), London, UK.
Rosen, B.H. 2000. Waterborne Pathogens in Agricultural Watersheds. U.S. Dept. of Agriculture, Natural
Resources Conservation Service, Watershed Science Institute, Ithaca, NY.
Roser, D.J., C.M. Davies, N.J. Ashbolt, and P. Morison. 2006. Microbial exposure assessment of an
urban recreational lake: A case study of the application of new risk-based guidelines. Water
Science and Technology 54(3):245-252.
Roser, D., N. Ashbolt, C. Davies, W. Glamore, K. Hawker, and B. Miller. 2007. Application of TMDL and
risk assessment principles for pathogen management at an urban recreational lake. In
Watershed Management to Meet Water Quality Standards and TMDLS, 4th Conference
Proceedings, 10-14 March 2007, San Antonio, Texas, 420-426 pp. The American Society of
Agricultural and Biological Engineers (ASABE).
Roy, S.L., S.M. DeLong, S.A. Stenzel, B. Shiferaw, J.M. Roberts, A. Khalakdina, R. Marcus, S.D. Segler,
D.D. Shah, S. Thomas, D.J. Vugia, S.M. Zansky, V. Dietz, M.J. Beach, and Emerging Infections
Program FoodNet Working Group. 2004. Risk factors for sporadic cryptosporidiosis among
immunocompetent persons in the United States from 1999 to 2001. Journal of Clinical
Microbiology 42(7):2944-2951.
Savichtcheva, O., and S. Okabe. 2006. Alternative indicators of fecal pollution: Relations with
pathogens and conventional indicators, current methodologies for direct pathogen monitoring
and future application perspectives. Water Research 40(13):2463-2476.
Scallan, E., R.M. Hoekstra, F.J. Angulo, R.V. Tauxe, M.A. Widdowson, S.L. Roy, J.L. Jones, and P.M.
Griffin. 2011a. Foodborne illness acquired in the United States—major pathogens. Emerging
Infectious Diseases 17(1):7-15.
157
-------
Scallan, E., P.M. Griffin, F.J. Angulo, R.V. Tauxe, and R.M. Hoekstra. 2011b. Foodborne illness acquired
in the United States—unspecified agents. Emerging Infectious Diseases 17(l):16-22.
Schaible, U.E., and S.H. Kaufmann. 2007. Malnutrition and infection: Complex mechanisms and global
impacts. PLoS Medicine 4(5):ell5.
Schets, F.M., J.F. Schijven, and A.M. de Roda Husman. 2011. Exposure assessment for swimmers in
bathing waters and swimming pools. Water Research 45(7):2392-2400.
Schets, F. M., H.H.J.L. van den Berg, H. Vennema, M.T.M. Pelgrim, C. Colle, S.A. Rutjes, and W.J. Lodder.
2018. Norovirus Outbreak Associated with Swimming in a Recreational Lake Not Influenced by
External Human Fecal Sources in The Netherlands, August 2012. International Journal of
Environmental Research and Public Health 15(11):2550.
https://doi.org/10.3390/iierphl511255Q.
Schiellerup, P., K.A. Krogfelt, and H. Locht. 2008. A comparison of self-reported joint symptoms
following infection with different enteric pathogens: Effect of HLA-B27. The Journal of
Rheumatology 35(3):480-487.
Schijven, J., and A.M. de Roda Husman. 2006. A survey of diving behaviour and accidental water
ingestion among Dutch occupational and sport divers to assess the risk of infection with
waterborne pathogenic microorganisms. Environmental Health Perspectives 114(5):712-717.
Schmidt, P.J., K.D. Pintar, A.M. Fazil, C.A. Flemming, M. Lanthier, N. Laprade, M.D. Sunohara, A.
Simhon, J.L. Thomas, E. Topp, G. Wilkes, and D.R. Lapen. 2013. Using Campylobacter spp. and
Escherichia coli data and Bayesian microbial risk assessment to examine public health risks in
agricultural watersheds under tile drainage management. Water Research 47(10):3255-3272.
Schneiderman, N., G. Ironson, and S.D. Siegel. 2005. Stress and health: Psychological, behavioral, and
biological determinants. Annual Review of Clinical Psychology 1:607-628.
Schoen, M.E., and N.J. Ashbolt. 2010. Assessing pathogen risk to swimmers at non-sewage impacted
recreational beaches. Environmental Science & Technology 44(7):2286-2291.
Schoen, M.E., J.A. Soller, and N.J. Ashbolt. 2011. Evaluating the importance of faecal sources in human-
impacted waters. Water Research 45(8):2670-2680.
Schoen, M.E., and J. Garland,. 2017. Review of pathogen treatment reductions for onsite non-potable
reuse of alternative source waters. Microbial Risk Analysis 5:25-31.
Schoen, M.E., N.J. Ashbolt, M.A. Jahne, and J. Garland. 2017. Risk-based enteric pathogen reduction
targets for non-potable and direct potable use of roof runoff, stormwater, and greywater.
Microbial Risk Analysis 5:32-43.
Schoen, M.E., A.B. Boehm, J. Soller, and O.C. Shanks. 2020. Contamination scenario matters when
using viral and bacterial human-associated genetic markers as indicators of a health risk in
untreated sewage-impacted recreational waters. Environmental Science & Technology
54(20) :13101-13109.
158
-------
Schuster, C.J., A.G. Ellis, W.J. Robertson, D.F. Charron, J.J. Aramini, B.J. Marshall, and D.T. Medeiros.
2005. Infectious disease outbreaks related to drinking water in Canada, 1974-2001. Canadian
Journal of Public Health 96(4):254-258.
Schwab, K.J., and C. Hurst. 2006. Human caliciviruses. Chapter in Waterborne Pathogens: Manual of
Water Supply Practices. AWWA Manual M48. 2nd ed. 281-286 pp. American Water Works
Association, Denver, CO.
Schwab, K.J. 2007. Are existing bacterial indicators adequate for determining recreational water illness
in waters impacted by nonpoint pollution? Epidemiology 18(l):21-22.
Shanks, O.C., E. Atikovic, A.D. Blackwood, J. Lu, R.T. Noble, J.S. Domingo, S. Seifring, M. Sivaganesan,
and R.A. Haugland. 2008. Quantitative PCR for detection and enumeration of genetic markers
of bovine fecal pollution. Applied and Environmental Microbiology 74(3):745-752.
Shanks, O.C., K. White, C.A. Kelty, M. Sivaganesan, J. Blannon, M. Meckes, M. Varma, and R.A.
Haugland. 2010. Performance of PCR-based assays targeting Bacteroidales genetic markers of
human fecal pollution in sewage and fecal samples. Environmental Science & Technology
44(16) :6281-6288.
Shanks, O.C., C.A. Kelty, R. Oshiro, R.A. Haugland, T. Madi, L. Brooks, K.G. Field, and M. Sivaganesan.
2016. Data Acceptance Criteria for Standardized Human-Associated Fecal Source Identification
Quantitative Real-Time PCR Methods. Applied and Environmental Microbiology 82(9):2773-
2782.
Shaywitz, S.E., P.M. Cohen, D.J. Cohen, E. Mikkelson, G. Morowitz, and B.A. Shaywitz. 1982. Long-term
consequences of Reye syndrome: A sibling-matched, controlled study of neurologic, cognitive,
academic, and psychiatric function. Journal of Pediatrics 100(l):41-46.
Shibata, T., and H.M. Solo-Gabriele. 2012. Quantitative microbial risk assessment of human illness from
exposure to marine beach sand. Environmental Science & Technology 46(5):2799-2805.
Shirtliff, M.E., and J.T. Mader. 2002. Acute septic arthritis. Clinical Microbiology Reviews 15(4):527-
544.
Shrestha, A., C.A. Kelty, M. Sivaganesan, O.C. Shanks, and S. Dorevitch. 2020. Fecal pollution source
characterization at non-point source impacted beaches under dry and wet weather conditions.
Water Research 182:116014.
Sidhu, J.P.S., K. Sena, L. Hodgers, A. Palmer, and S. Toze. 2018. Comparative enteric viruses and
coliphage removal during wastewater treatment processes in a sub-tropical environment. The
Science of the Total Environment 616-617:669-677.
Signor, R.S., and N.J. Ashbolt. 2006. Pathogen monitoring offers questionable protection against
drinking-water risks: A QMRA (quantitative microbial risk analysis) approach to assess
management strategies. Water Science and Technology 54(3):261-268.
159
-------
Silvestri, E.E., C. Yund, S. Taft, C.Y. Bowling, D. Chappie, K. Garrahan, E. Brady-Roberts, H. Stone, and
T.L. NicholsL. 2017. Considerations for estimating microbial environmental data concentrations
collected from a field setting. Journal of Exposure Science & Environmental Epidemiology
27(2):141—151.
Simhon, A., V. Pileggi, C.A. Flemming, G. Lai, and M. Manoharan. 2020. Norovirus risk at a golf course
irrigated with reclaimed water: Should QMRA doses be adjusted for infectiousness? Water
Research 183:116121.
Sinclair, R.G., E.L. Jones, and C.P. Gerba. 2009. Viruses in recreational water-borne disease outbreaks: A
review. Journal of Applied Microbiology 107(6):1769-1780.
Singh-Naz, N., and W. Rodriguez. 1996. Adenoviral infections in children. Advances in Pediatric
Infectious Diseases 11:365-388.
Sinigalliano, C.D., J.M. Fleisher, M.L. Gidley, H.M. Solo-Gabriele, T. Shibata, L.R. Piano, S.M. Elmir, D.
Wanless, J. Bartkowiak, R. Boiteau, K. Withum, A.M. Abdelzaher, G. He, C. Ortega, X. Zhu, M.E.
Wright, J. Kish, J. Hollenbeck, T. Scott, L.C. Backer, and L.E. Fleming. 2010. Traditional and
molecular analyses for fecal indicator bacteria in non-point source subtropical recreational
marine waters. Water Research 44(13):3763-3772.
Sinigalliano, C.D., J.S. Ervin, L.C. Van De Werfhorst, B.D. Badgley, E. Balleste, J. Bartkowiak, A.B. Boehm,
M. Byappanahalli, K.D. Goodwin, M. Gourmelon, J. Griffith, P.A. Holden, J. Jay, B. Layton, C. Lee,
J. Lee, W.G. Meijer, R. Noble, M. Raith, H. Ryu, M.J. Sadowsky, A. Schriewer, D. Wang, D.
Wanless, R. Whitman, S. Wuertz, and J.W. Santo Domingo. 2013. Multi-laboratory evaluations
of the performance of Catellicoccus marimammalium PCR assays developed to target gull fecal
sources. Water Research 47(18):6883-6896.
Sips, G.J., M.J.G. Dirven, J.T. Donkervoort, F.M. van Kolfschoten, C.M.E. Schapendonk, M.V.T. Phan, A.
Bloem, A.F. van Leeuwen, M.E. Trompenaars, M.P.G. Koopmans, A.A. van der Eijk, M. de Graaf,
and E.B. Fanoy. 2020. Norovirus outbreak in a natural playground: A One Health approach.
Zoonoses and Public Health 67(4):453-459.
Skinner, J.F., J. Kappeler, and J. Guzman. 2010. Regrowth of enterococci & fecal coliform in biofilm.
Stormwater Solutions July/August:28-34.
Smeets, P.W., L.C. Rietveld, J.C. van Dijk, and G.J. Medema. 2010. Practical applications of quantitative
microbial risk assessment (QMRA) for water safety plans. Water Science and Technology
61(6):1561-1568.
Smith, J.W., and M.S. Wolfe. 1980. Giardiasis. Annual Review of Medicine 31:373-383.
Sobsey, M.D., L.A. Khatib, V.R. Hill, E. Alocilja, and S. Pillai. 2006. Pathogens in animal wastes and the
impacts of waste management practices on their survival, transport and fate. Chapter in Animal
Agriculture and the Environment: National Center for Manure and Animal Waste Management
White Papers. J.M. Rice, D.F. Caldwell, F.J. Humenik, eds. ASABE, St. Joseph, Ml.
Soller, J.A., J.N. Eisenberg, and A.W. Olivieri. 1999. Evaluation of Pathogen Risk Assessment Framework.
Case Study: Human Infection Through Drinking Water Exposure to Human Infectious Rotavirus.
Eisenberg, Olivieri and Associates, Inc., Oakland, CA.
160
-------
Soller, J.A., A.W. Olivieri, J. Crook, R.C. Cooper, G. Tchobanoglous, R.T. Parkin, R.C. Spear, and J.N.
Eisenberg. 2003. Risk-based approach to evaluate the public health benefit of additional
wastewater treatment. Environmental Science & Technology 37(9):1882-1891.
Soller, J.A., J.N. Eisenberg, J.F. DeGeorge, R.C. Cooper, G. Tchobanoglous, and A.W. Olivieri. 2006. A
public health evaluation of recreational water impairment. Journal of Water and Health 4(1):1—
19.
Soller, J.A., E.Y. Seto, and A.W. Olivieri. 2007. Application of Microbial Risk Assessment Techniques to
Estimate Risk Due to Exposure to Reclaimed Waters. WateReuse Foundation, Final Project
Report.
Soller, J.A., and J.N.S. Eisenberg. 2008. An evaluation of parsimony for microbial risk assessment
models. Environmetrics 19(l):61-78.
Soller, J.A. 2009. The potential implications of person-to-person transmission of viral infection for US
EPA's Groundwater Rule. Journal of Water and Health 7(2):208-223.
Soller, J.A., T. Bartrand, N.J. Ashbolt, J. Ravenscroft, and T.J. Wade. 2010a. Estimating the primary
etiologic agents in recreational freshwaters impacted by human sources of faecal
contamination. Water Research 44(16):4736-4747.
Soller, J.A., Schoen, M.E., Bartrand, T., Ravenscroft, J.E., Ashbolt, N.J. 2010b. Estimated human health
risks from exposure to recreational waters impacted by human and non-human sources of
faecal contamination. Water Research 44(16):4674-4691.
Soller, J.A., M.E. Schoen, A. Varghese, A.M. Ichida, A.B. Boehm, S. Eftim, N.J. Ashbolt, and J.E.
Ravenscroft. 2014. Human health risk implications of multiple sources of faecal indicator
bacteria in a recreational waterbody. Water Research 66:254-264.
Soller, J., T. Bartrand, J. Ravenscroft, M. Molina, G. Whelan, M. Schoen, and N. Ashbolt. 2015.
Estimated human health risks from recreational exposures to stormwater runoff containing
animal faecal material. Environmental Modelling & Software 72:21-32.
Soller, J.A., S. Eftim, T.J. Wade, A.M. Ichida, J.L. Clancy, T.B. Johnson, K. Schwab, G. Ramirez-Toro, S.
Nappier, and J.E. Ravenscroft. 2016. Use of quantitative microbial risk assessment to improve
interpretation of a recreational water epidemiological study. Microbial Risk Analysis 1:2-11.
Soller, J.A., M. Schoen, J.A. Steele, J.F. Griffith, and K.C. Schiff. 2017. Incidence of gastrointestinal illness
following wet weather recreational exposures: Harmonization of quantitative microbial risk
assessment with an epidemiologic investigation of surfers. Water Research 121:280-289.
Soller, J.A., A.M. Parker, and A. Salveson. 2018. Public health implications of short duration, off-
specification conditions at potable reuse water treatment facilities. Environmental Science &
Technology Letters 5(ll):675-680.
Stachler, E., B. Akyon, N.A. de Carvalho, C. Ference, and K. Bibby. 2018. Correlation of crAssphage qPCR
Markers with Culturable and Molecular Indicators of Human Fecal Pollution in an Impacted
Urban Watershed. Environmental Science & Technology 52(13) :7505-7512.
161
-------
Stampi, S., 0. Varoli, F. Zanetti, and G. De Luca. 1993. Arcobacter cryaerophilus and thermophilic
Campylobacters in a sewage treatment plant in Italy: Two secondary treatments compared.
Epidemiology and Infection 110(3):633-639.
Stecher, B. 2015. The Roles of Inflammation, Nutrient Availability and the Commensal Microbiota in
Enteric Pathogen Infection. Microbiology Spectrum 3(3).
Stevenson, A.H. 1953. Studies of bathing water quality and health. American Journal of Public Health
and the Nation's Health 43(5 Pt l):529-538.
Stewart, J.R., R.J. Gast, R.S. Fujioka, H.M. Solo-Gabriele, J.S. Meschke, L.A. Amaral-Zettler, E. del
Castillo, M.F. Polz, T.K. Collier, M.S. Strom, C.D. Sinigalliano, P.D. Moeller, and A.F. HollandF.
2008. The coastal environment and human health: microbial indicators, pathogens, sentinels
and reservoirs. Environmental Health 7(Suppl 2):S3.
Sturdee, A.P., A.T. Bodley-Tickell, A. Archer, and R.M. Chalmers. 2003. Long-term study of
Cryptosporidium prevalence on a lowland farm in the United Kingdom. Veterinary Parasitology
116(2):97—113.
Sunger, N., K.A. Hamilton, P.M. Morgan, and C.N. Haas. 2019. Comparison of pathogen-derived 'total
risk' with indicator-based correlations for recreational (swimming) exposure. Environmental
Science and Pollution Research International 26(30):30614-30624.
Suppes, L.M., L. Abrell, A.P. Dufour, and K.A. Reynolds. 2014. Assessment of swimmer behaviors on
pool water ingestion. Journal of Water and Health 12(2):269-279.
Tarr, P.I. 1995. Escherichia coli 0157:H7: Clinical, diagnostic, and epidemiological aspects of human
infection. Clinical Infectious Diseases,20(l):l-8; quiz 9-10.
Teixeira, P., D. Salvador, J. Brandao, W. Ahmed, M.J. Sadowsky, and E. Valerio. 2020. Environmental
and adaptive changes necessitate a paradigm shift for indicators of fecal contamination.
Microbiology Spectrum 8(2).
Terzich, M., M.J. Pope, T.E. Cherry, and J. Hollinger. 2000. Survey of pathogens in poultry litter in the
United States. Journal of Applied Poultry Research 9(3):287-291.
Teunis, P.F.M., O.G. van der Heijden, J.W.B. van der Giessen, and A.H. Havelaar. 1996. The Dose-
Response Relation in Human Volunteers for Gastro-lntestinal Pathogens. Rijksinstituut voor
Volksgezondheid en Milieu (RIVM), Bilthoven, The Netherlands.
Teunis, P.F., N.J. Nagelkerke, and C.N. Haas. 1999. Dose response models for infectious gastroenteritis.
Risk Analysis 19(6): 1251-1260.
Teunis, P.F., C.L. Chappell, and P.C. Okhuysen. 2002a. Cryptosporidium dose response studies: variation
between isolates. Risk Analysis 22(1):175-183.
Teunis, P.F., C.L. Chappell, and P.C. Okhuysen. 2002b. Cryptosporidium dose-response studies: variation
between hosts. Risk Analysis 22(3):475-485.
Teunis, P., K. Takumi, and K. Shinagawa. 2004. Dose response for infection by Escherichia coli 0157:H7
from outbreak data. Risk Analysis 24(2):401-407.
162
-------
Teunis, P., W. Van den Brandhof, M. Nauta, J. Wagenaar, H. Van den Kerkhof, and W. Van Pelt. 2005. A
reconsideration of the Campylobacter dose-response relation. Epidemiology and Infection
133(4):583-592.
Teunis, P.F., C.L. Moe, P. Liu, S.E. Miller, L. Lindesmith, R.S. Baric, J. Le Pendu, and R.L. Calderon. 2008a.
Norwalk virus: How infectious is it? Journal of Medical Virology 80(8):1468-1476.
Teunis, P.F., I.D. Ogden, and N.J. StrachanJ. 2008b. Hierarchical dose response of E. coli 0157:H7 from
human outbreaks incorporating heterogeneity in exposure. Epidemiology and Infection
136(6):761-770.
Teunis, P., J. Schijven, and S. Rutjes. 2016. A generalized dose-response relationship for adenovirus
infection and illness by exposure pathway. Epidemiology and Infection 144(16):3461-3473.
Teunis, P.F.M., A. Bonacic Marinovic, D.R. Tribble, C.K. Porter, and A. Swart. 2018. Acute illness from
Campylobacter jejuni may require high doses while infection occurs at low doses. Epidemics
24:1-20.
Teunis, P.F.M., F.S. Le Guyader, P. Liu, J. Ollivier, and C.L. Moe. 2020. Noroviruses are highly infectious
but there is strong variation in host susceptibility and virus pathogenicity. Epidemics 32:100401.
Thompson, R.C. 2000. Giardiasis as a re-emerging infectious disease and its zoonotic potential.
International Journal for Parasitology 30(12-13):1259-1267.
Thomson, G.T., D.A. DeRubeis, M.A. Hodge, C. Rajanayagam, and R.D. Inman. 1995. Post -Salmonella
reactive arthritis: Late clinical sequelae in a point source cohort. American Journal of Medicine
98(1):13—21.
Thomson, S., E.A. Innes, N.N. Jonsson, and F. Katzer. 2019. Shedding of Cryptosporidium in calves and
dams: Evidence of re-infection and shedding of different gp60 subtypes. Parasitology
146(11) :1404-1413.
Till, D.G., and G.B. McBride. 2004. Potential public health risk of Campylobacter and other zoonotic
waterborne infections in New Zealand. Chapter 12 in Waterbome Zoonoses: Identification,
Causes, and Control. J.A. Cortruvo, A.P. Dufour, G. Rees, J. Bartram, R. Carr, D.O. Cliver, G.F.
Craun, R. Fayer, and V.P.J. Gannon, eds. 191-208 pp. IWA Publishing for the World Health
Organization, London, UK.
Till, D., G. McBride, A. Ball, K. Taylor, and E. Pyle. 2008. Large-scale freshwater microbiological study:
Rationale, results and risks. Journal of Water and Health 6(4):443-460.
Timm, C., S. Luther, L. Jurzik, I.A. Hamza, and T. Kistemann. 2016. Applying QMRA and DALY to assess
health risks from river bathing. International Journal of Hygiene and Environmental Health
219(7 Pt B):681-692.
Tupchong, M., A. Simor, and C. Dewar. 1999. Beaver fever—a rare cause of reactive arthritis. The
Journal of Rheumatology 26(12):2701-2702.
163
-------
U.S. EPA (U.S. Environmental Protection Agency). 1986. Ambient Water Quality Criteria for Bacteria -
1986. EPA/440/5-84/002. U.S. Environmental Protection Agency, Office of Water Regulations
and Standards, Criteria Standards Division, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 1994. Use of Monte Carlo Simulation in Risk
Assessments. Accessed October 28, 2022.
https://www.epa.gov/risk/use-monte-carlo-simulation-risk-assessments.
U.S. EPA (U.S. Environmental Protection Agency). 1998a. Giardia: Human Health Criteria Document.
EPA-823-R-002. U.S. Environmental Protection Agency, Office of Water, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 1998b. Guidelines for Ecological Risk Assessment.
Federal Register, May 14,1998, 63(93): 26846-26924. EPA/630/R-95/002F.
U.S. EPA (U.S. Environmental Protection Agency). 2000a. Risk Characterization Handbook. EPA/100/B-
00/002. U.S. Environmental Protection Agency, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2000b. Guidance for Data Quality Assessment:
Practical Methods for Data Analysis. EPA/600/R-96/084. U.S. Environmental Protection Agency,
Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2000c. Report to Congress EPA Studies on Sensitive
Subpopulations and Drinking Water Contaminants. EPA 815-R-00-015. U.S. Environmental
Protection Agency, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2001. Protocol for Developing Pathogen TMDLs. 1st
ed. EPA-841-R-00-002. U.S. Environmental Protection Agency, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2005. Occurrence and Exposure Assessment for the
Final Long Term 2 Enhanced Surface Water Treatment Rule. EPA-815-R-06-002. U.S.
Environmental Protection Agency, Office of Water, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2006a. National Primary Drinking water regulations:
Long Term 2 Enhanced Surface Water Treatment Rule (LT2ESWTR); Final Rule (40 CFR Parts 9,
141 and 142, Volume 71, Number 654). Federal Register, March 6, 2006, 71(3):654-786.
U.S. EPA (U.S. Environmental Protection Agency). 2006b. Guidance on Systematic Planning Using the
Data Quality Objectives Process: EPA QA/G-4. EPA/240/B-06/001. U.S. Environmental
Protection Agency, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2006c. National Primary Drinking Water Regulations:
Ground Water Rule; Final Rule (40 CFR Parts 9, 141, and 142). Federal Register, November 8,
2006, 71(216):65574-65660.
U.S. EPA (U.S. Environmental Protection Agency). 2007a. Report of the Experts Scientific Workshop on
Critical Research Needs for the Development of New or Revised Recreational Water Quality
Criteria: Draft Executive Summary by Workgroup Chairs. EPA-823-R-07-007. U.S. Environmental
Protection Agency, Office of Water, Office of Research and Development, Washington, DC.
164
-------
U.S. EPA (U.S. Environmental Protection Agency). 2007b. Report of the Experts Scientific Workshop on
Critical Research Needs for the Development of New or Revised Recreational Water Quality
Criteria. EPA-823-R-07-006. U.S. Environmental Protection Agency, Office of Water, Office of
Research and Development, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2007c. Critical Path Science Plan for the Development
of New or Revised Recreational Water Quality Criteria. EPA-823-R-08-002. U.S. Environmental
Protection Agency, Office of Water, Office of Research and Development, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2007d. Thesaurus of Terms Used in Microbiological
Risk Assessment. U.S. Environmental Protection Agency, Office of Water, Office of Science and
Technology, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2008a. Consent Decree and Settlement Agreement of
2008. Case 2;06-cv-04843-PSG-JTL document 159-3 files 08/08/2008. U.S. Environmental
Protection Agency, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2008b. Handbook for Developing Watershed Plans to
Restore and Protect our Waters. EPA-841-B-08-002U.S. Environmental Protection Agency,
Office of Water, Nonpoint Source Control Branch, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2009a. Review of Published Studies to Characterize
Relative Risks from Different Sources of Fecal Contamination in Recreational Water. EPA-822-R-
09-001. U.S. Environmental Protection Agency, Office of Water, Health and Ecological Criteria
Division, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2009b. Review of Zoonotic Pathogens in Ambient
Waters. EPA 822-R-09-002. U.S. Environmental Protection Agency, Office of Water, Health and
Ecological Criteria Division, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2010a. Quantitative Microbial Risk Assessment to
Estimate Illness in Freshwater Impacted by Agricultural Animal Sources of Fecal Contamination.
EPA-822-R-10-005. U.S. Environmental Protection Agency, Office of Water, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2010b. Assessment of the Extra-Enteric Behavior of
Fecal Indicator Organisms in Ambient Waters. U.S. Environmental Protection Agency, Office of
Water, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2010c. Report on 2009 National Epidemiologic and
Environmental Assessment of Recreational Water Epidemiology Studies (NEEAR 2010—Surfside
& Boqueron). EPA/600/R-10/168. U.S. Environmental Protection Agency, Office of Research and
Development, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2011a. Using Microbial Source Tracking to Support
TMDL development and Implementation. U.S. Environmental Protection Agency, Region 10
Watersheds Unit, Seattle, WA.
U.S. EPA (U.S. Environmental Protection Agency). 2011b. Exposure Factors Handbook: 2011 Edition.
EPA/600/R-090/052F. U.S. Environmental Protection Agency, Office of Research and
Development, National Center for Environmental Assessment, Washington, DC.
165
-------
U.S. EPA (U.S. Environmental Protection Agency). 2012. Recreational Water Quality Criteria. 820-F-12-
058. U.S. Environmental Protection Agency, Office of Water, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2013a. Literature Review of Contaminants in
Livestock and Poultry Manure and Implications for Water Quality. EPA-820-R-13-002. U.S.
Environmental Protection Agency, Office of Water, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2013b. Marine Beach Sanitary Survey User Manual.
EPA-820-B-13-001. U.S. Environmental Protection Agency, Office of Water, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2013c. A Quick Guide to Developing Watershed Plans
to Restore and Protect our Waters. EPA-841-R-13-003. U.S. Environmental Protection Agency,
Office of Wetlands, Oceans, and Watersheds, Nonpoint Source Control Branch (4503T),
Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2014a. Site-Specific Alternative Recreational Criteria
Technical Support Materials for Alternative Indicators and Methods. EPA-820-R-14-011. U.S.
Environmental Protection Agency, Office of Water, Office of Science and Technology, Health
and Ecological Criteria Division, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2014b. Microbiological Risk Assessment (MRA) Tools,
Methods, and Approaches for Water Media. EPA-820-R-14-009. U.S. Environmental Protection
Agency, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2014c. Framework for Human Health Risk
Assessment to Inform Decision Making. Final. EPA/100/R-14/001. U.S. Environmental
Protection, Risk Assessment Forum, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2017a. Quantitative Microbial Risk Assessment
Tutorial—SDMProjectBuilder: Import Local Data Files to Identify and Modify Contamination
Sources and Input Parameters. EPA/600/B-15/316. U.S. Environmental Protection Agency,
Office of Research and Development, National Exposure Research Laboratory, Ecosystems
Research Division, Athens, GA.
U.S. EPA (U.S. Environmental Protection Agency). 2017b. Microbial Source Module (MSM):
Documenting the Science and Software for Discovery, Evaluation, and Integration. EPA/600/B-
15/315. U.S. Environmental Protection Agency, Office of Research and Development, National
Exposure Research Laboratory, Ecosystems Research Division, Athens, GA.
U.S. EPA (U.S. Environmental Protection Agency). 2017c. Quantitative Microbial Risk Assessment
Tutorial: Using NLDAS and NCDC Meteorological Data. EPA/600/B-15/299. U.S. Environmental
Protection Agency, Office of Research and Development, National Exposure Research
Laboratory, Athens, GA.
U.S. EPA (U.S. Environmental Protection Agency). 2017d. Navigate the SDMPB and identify an 8-digit
HUC of interest- Updated 2017. EPA/600/B-15/273.U.S. Environmental Protection Agency,
Office of Research and Development, National Exposure Research Laboratory, Athens, GA.
166
-------
U.S. EPA (U.S. Environmental Protection Agency). 2017e. Quantitative Microbial Risk Assessment
Tutorial: Land-Applied Microbial Loadings Within a 12-Digit HUC. EPA/600/B-15/298. U.S.
Environmental Protection Agency, Office of Research and Development, National Exposure
Research Laboratory, Athens, GA.
U.S. EPA (U.S. Environmental Protection Agency). 2017f. Quantitative Microbial Risk Assessment
Tutorial: Pour Point Analysis of Land-Applied Microbial Loadings and Comparison of Simulated
and Gaging Station Results. EPA/600/B-15/290. U.S. Environmental Protection Agency, Office of
Research and Development, National Exposure Research Laboratory, Athens, GA.
U.S. EPA (U.S. Environmental Protection Agency). 2017g. Quantitative Microbial Risk Assessment
Tutorial: Point Source and Land-Applied Microbial Loadings Within a 12-Digit HUC (In Review).
U.S. Environmental Protection Agency, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2017h. Quantitative Microbial Risk Assessment
Tutorial: Publishing a Microbial Density Time Series as a Txt File. EPA/600/B-15/274. U.S.
Environmental Protection Agency, Office of Research and Development, National Exposure
Research Laboratory, Athens, GA.
U.S. EPA (U.S. Environmental Protection Agency). 2017i. Quantitative Microbial Risk Assessment
Tutorial: Installation of Software for Watershed Modeling in Support ofQMRA. EPA/600/B-
15/276. U.S. Environmental Protection Agency, Office of Research and Development, National
Exposure Research Laboratory, Athens, GA.
U.S. EPA (U.S. Environmental Protection Agency). 2019a. Recommended Human Health Recreational
Ambient Water Quality Criteria or Swimming Advisories for Microcystins and
Cylindrospermopsin. EPA-822-R-19-001. U.S. Environmental Protection Agency, Washington,
DC.
U.S. EPA (U.S. Environmental Protection Agency). 2019b. Method 1696: Characterization of Human
Fecal Pollution in Water by HF183/BacR287 TaqMan® Quantitative Polymerase Chain Reaction
(qPCR) Assay. EPA-821-R-19-002. U.S. Environmental Protection Agency, Office of Water,
Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2019c. Method 1697: Characterization of Human
Fecal Pollution in Water by HumM2 TaqMan ® Quantitative Polymerase Chain Reaction (qPCR)
Assay. EPA-821-R-19-003. U.S. Environmental Protection Agency, Office of Water, Washington,
DC.
U.S. EPA (U.S. Environmental Protection Agency). 2022. Sanitary Surveys for Recreational Waters.
Accessed October 17, 2022.
https://www.epa.gov/beach-tech/sanitarv-surveys-recreational-waters.
U.S. EPA (U.S. Environmental Protection Agency) and USDA (U.S. Department of Agriculture). 2012.
Microbial Risk Assessment Guideline: Pathogenic Microorganisms With Focus on Food and
Water. EPA/100/J-12/001; USDA/FSIS/2012-001.
167
-------
Unno, T., C. Staley, C.M. Brown, D. Han, M.J. Sadowsky, and H.G. Hur. 2018. Fecal pollution: New
trends and challenges in microbial source tracking using next-generation sequencing.
Environmental Microbiology 20(9):3132-3140.
Vanden Esschert, K.L., M.C. Mattioli, E.D. Hilborn, V.A. Roberts, A.T. Yu, K. Lamba, G. Arzaga, M. Zahn,
Z. Marsh, S.M. Combes, E.S. Smith, T.J. Robinson, S.R. Gretsch, J.P. Laco, M.E. Wikswo, A.D.
Miller, D.M. Tack, T.J. Wade, and M.C. Hlavsa. 2020. Outbreaks associated with untreated
recreational water—California, Maine, and Minnesota, 2018-2019. Morbidity and Mortality
Weekly Report 69(25):781-783.
Vergara, G., J.B. Rose, and K.Y.H. Gin. 2016. Risk assessment of noroviruses and human adenoviruses in
recreational surface waters. Water Research 103:276-282.
Verhougstraete, M.P., K. Pogreba-Brown, K.A. Reynolds, C.C. Lamparelli, M.I. Zanoli Sato, T.J. Wade,
and J.N.S. Eisenberg. 2020. A critical analysis of recreational water guidelines developed from
temperate climate data and applied to the tropics. Water Research 170:115294.
Viau, E.J., D. Lee, and A.B. Boehm. 2011. Swimmer risk of gastrointestinal illness from exposure to
tropical coastal waters impacted by terrestrial dry-weather runoff. Environmental Science &
Technology 45(17):7158-7165.
Wade, S.E., H.O. Mohammed, and S.L. Schaaf. 2000. Prevalence of Giardia sp. Cryptosporidium parvum
and Cryptosporidium andersoni (syn. C. muris) [correction of Cryptosporidium parvum and
Cryptosporidium muris (C. andersoni)] in 109 dairy herds in five counties of southeastern New
York. Veterinary Parasitology 93(1): 1-11.
Wade, T.J., N. Pai, J.N. Eisenberg, and J.M. Colford, Jr. 2003. Do U.S. Environmental Protection Agency
water quality guidelines for recreational waters prevent gastrointestinal illness? A systematic
review and meta-analysis. Environmental Health Perspectives 111(8):1102-1109.
Wade, T.J., R.L. Calderon, E. Sams, M. Beach, K.P. Brenner, A.H. Williams, and A.P. Dufour. 2006.
Rapidly measured indicators of recreational water quality are predictive of swimming-
associated gastrointestinal illness. Environmental Health Perspectives 114(l):24-28.
Wade, T.J., R.L. Calderon, K.P. Brenner, E. Sams, M. Beach, R. Haugland, L. Wymer, and A.P. Dufour.
2008. High sensitivity of children to swimming-associated gastrointestinal illness: Results using
a rapid assay of recreational water quality. Epidemiology 19(3):375-383.
Wade, T.J., E. Sams, K.P. Brenner, R. Haugland, E. Chern, M. Beach, L. Wymer, C.C. Rankin, D. Love, Q.
Li, R. Noble, and A.P. Dufour. 2010. Rapidly measured indicators of recreational water quality
and swimming-associated illness at marine beaches: A prospective cohort study. Environmental
Health, 9:66.
Wade, T.J., B.F. Arnold, K. Schiff, J.M. Colford, Jr., S.B. Weisberg, J.F. Griffith, and A.P. Dufour. 2022.
Health risks to children from exposure to fecally-contaminated recreational water. PLoS One
17(4):e0266749.
168
-------
Wallis, P.M., S.L. Erlandsen, J.L. Isaac-Renton, M.E. Olson, W.J. Robertson, and H. van Keulen. 1996.
Prevalence of Giardia cysts and Cryptosporidium oocysts and characterization of Giardia spp.
isolated from drinking water in Canada. Applied and Environmental Microbiology 62(8):2789-
2797.
Ward, R.L., D.I. Bernstein, E.C. Young, J.R. Sherwood, D.R. Knowlton, and G.M. Schiff. 1986. Human
rotavirus studies in volunteers: Determination of infectious dose and serological response to
infection. The Journal of Infectious Diseases 154(5):871-880.
Ward, R.L., D.R. Knowlton, and P.E. Winston. 1986. Mechanism of inactivation of enteric viruses in
fresh water. Applied and Environmental Microbiology 52(3):450-459.
Weijtens, M.J., R.D. Reinders, H.A. Urlings, and J. Van der Plas. 1999. Campylobacter infections in
fattening pigs; excretion pattern and genetic diversity. Journal of Applied Microbiology
86(1):63—70.
Werber, D., B.W. Mason, M.R. Evans, and R.L. Salmon. 2008. Preventing household transmission of
Shiga toxin-producing Escherichia coli 0157 infection: Promptly separating siblings might be the
key. Clinical Infectious Diseases 46(8):1189-1196.
WERF (Water Environment Research Foundation). 2011. Quantification of Pathogens and Sources of
Microbial Indicators for QMRA In Recreational Waters: Pathogens and Human Health Final
Report. WERF Report PATH2R08. IWA Publishing and Water Environment Research Foundation
(WERF), London, England.
Wesley, I.V., S.J. Wells, K.M. Harmon, A. Green, L. Schroeder-Tucker, M. Glover, and I. Siddique 2000.
Fecal shedding of Campylobacter and Arcobacter spp. in dairy cattle. Applied and Environmental
Microbiology 66(5): 1994-2000.
Whelan, G., K. Kim, M.A. Pelton, J.A. Soller, K.J. Castleton, M. Molina, Y. Pachepsky, and R. Zepp. 2014a.
An integrated environmental modeling framework for performing Quantitative Microbial Risk
Assessments. Environmental Modelling & Software 55:77-91.
Whelan, G., K. Kim, M.A. Pelton, K.J. Castleton, G.F. Laniak, K. Wolfe, R. Parmar, J. Babendreier, and M.
Galvin. 2014b. Design of a component-based integrated environmental modeling framework.
Environmental Modelling & Software 55:1-24.
Whiley, H., B. van den Akker, S. Giglio, and R. Bentham. 2013. The role of environmental reservoirs in
human campylobacteriosis. International Journal of Environmental Research and Public Health
10(11) :5886-5907.
Whitman, R.L., M.B. Nevers, G.C. Korinek, and M.N. Byappanahalli. 2004. Solar and temporal effects on
Escherichia coli concentration at a Lake Michigan swimming beach. Applied and Environmental
Microbiology 70(7):4276-4285.
Whitman, R., V.J. Harwood, T.A. Edge, M. Nevers, M. Byappanahalli, K. Vijayavel, J. Brandao, M.J.
Sadowsky, E.W. Aim, A. Crowe, D. Ferguson, Z. Ge, E. Halliday, J. Kinzelman, G. Kleinheinz, K.
Przybyla-Kelly, C. Staley, Z. Staley, and H.M. Solo-Gabriele. 2014. Microbes in beach sands:
Integrating environment, ecology and public health. Reviews in Environmental Science and
Bio/Technology 13(3):329-368.
169
-------
WHO (World Health Organization). 2003. Guidelines for Safe Recreational Water Environments. Volume
1, Coastal and Fresh Waters. World Health Organization, Geneva, Switzerland.
WHO (World Health Organization). 2004. Waterbome Zoonoses: Identification, Causes and Control.
World Health Organization, Geneva, Switzerland.
WHO (World Health Organization). 2009. Risk Assessment of Cryptosporidium in Drinking-Water.
WHO/HSE/WSH/09.04.World Health Organization, Geneva, Switzerland.
WHO (World Health Organization). 2016. Quantitative Microbial Risk Assessment: Application for
Water Safety Management. World Health Organization, Geneva, Switzerland.
WHO (World Health Organization). 2021. Guidelines on Recreational Water Quality. Volume 1: Coastal
and Fresh Waters. License: CC BY-NC-SA 3.0 IGO. World Health Organization, Geneva,
Switzerland.
Wilhelmi, I., E. Roman, and A. Sanchez-Fauquier. 2003. Viruses causing gastroenteritis. Clinical
Microbiology and Infection 9(4):247-262.
Willis, J.R., M. Sivaganesan, R.A. Haugland, J. Kralj, S. Servetas, M.E. Hunter, S.A. Jackson, and O.C.
Shanks. 2022. Performance of NIST SRM® 2917 with 13 recreational water quality monitoring
qPCR assays. Water Research 212:118114.
Wolfe, M.S. 1990. Clinical symptoms and diagnosis by traditional methods. Chapter in Giardiasis. E.A.
Meyer, ed. 175-186 pp. Elsevier Science Publishers, Amsterdam, The Netherlands.
Wong, M., L. Kumar, T.M. Jenkins, I. Xagoraraki, M.S. Phanikumar, and J.B. Rose. 2009. Evaluation of
public health risks at recreational beaches in Lake Michigan via detection of enteric viruses and
a human-specific bacteriological marker. Water Research 43(4):1137-1149.
Wu, J., S.C. Long, D. Das, and S.M. Dorner. 2011. Are microbial indicators and pathogens correlated? A
statistical analysis of 40 years of research. Journal of Water and Health 9(2):265-278.
Wu, B., C. Wang, C. Zhang, M.J. Sadowsky, M. Dzakpasu, and X.C. Wang. 2020. Source-associated
gastroenteritis risk from swimming exposure to aging fecal pathogens. Environmental Science &
Technology 54(2):921-929.
Wyer, M.D., D. Kay, J.M. Fleisher, R.L. Salmon, F. Jones, A.F. Godfree, G. Jackson, and A. Rogers. 1999.
An experimental health-related classification for marine waters. Water Research 33(3):715-
722.
Wymer, L.J., and T.J. Wade. 2007. The lognormal distribution and use of the geometric mean and the
arithmetic mean in recreational water quality measurement. Chapter 6 in Statistical Framework
for Recreational Water Quality Criteria and Monitoring. L.J. Wymer, ed. 91-111 pp. John Wiley
& Sons, Chichester, UK.
Wymer, L.J., T.J. Wade, and A.P. Dufour. 2013. Equivalency of risk for a modified health endpoint: A
case from recreational water epidemiology studies. BMC Public Health 13:459.
170
-------
Xiao, L., J.E. Moore, U. Ukoh, W. Gatei, C.J. Lowery, T.M. Murphy, J.S. Dooley, B.C. Millar, P.J. Rooney,
and J.R. Rao. 2006. Prevalence and identity of Cryptosporidium spp. in pig slurry. Applied and
Environmental Microbiology 72(6):4461-4463.
Xue, J., and Y. Feng. 2019. Comparison of microbial source tracking efficacy for detection of cattle fecal
contamination by quantitative PCR. The Science of the Total Environment 686:1104-1112.
Yan, T., D.K. Goto, and F. Feng. 2011. Concentration Dynamics of Fecal Indicators in Hawaii's Coastal
and Inland Sand, Soil, and Water During Rainfall Events. PATH6R09. Water Environment
Research Foundation Alexandria, VA, and IWA Publishing.
Yang, R., C. Jacobson, G. Gardner, I. Carmichael, A.J. Campbell, and U. Ryan. 2014. Longitudinal
prevalence, faecal shedding and molecular characterisation of Campylobacter spp. and
Salmonella enterica in sheep. Veterinary Journal 202(2):250-254.
Yoo, J.Y., M. Groer, S.V.O. Dutra, A. Sarkar, and D.I. McSkimming. 2020. Gut microbiota and immune
system interactions. Microorganisms 8(10).
Yuki, N. 2001. Infectious origins of, and molecular mimicry in, Guillain-Barre and Fisher syndromes. The
Lancet. Infectious Diseases l(l):29-37.
Zimmer-Faust, A.G., J.A. Steele, J.F. Griffith, and K. Schiff. 2020. The challenges of microbial source
tracking at urban beaches for Quantitative Microbial Risk Assessment (QMRA). Marine Pollution
Bulletin 160:111546.
Zmirou, D., L. Pena, M. Ledrans, and A. Letertre. 2003. Risks associated with the microbiological quality
of bodies of fresh and marine water used for recreational purposes: Summary estimates based
on published epidemiological studies. Archives of Environmental Health 58(11):703-711.
171
-------
Appendix A: Literature Search Strategies
This appendix provides information on how literature was found for this TSM. The EPA developed a set
of literature search strategies based on the Critical Path Science Plan that were previously published.30
The peer-reviewed scientific literature was searched for recreational exposure information and
relevant data for reference pathogens and fecal indicators. The resulting literature from the following
published literature search strategies were examined for relevance for this TSM. These literature
search strategies can be found in:
• Report on the 2nd Five-Year Review of EPA's Recreational Water Quality Criteria (U.S. EPA,
2023).
• 2017 Five-Year Review of the 2012 Recreational Water Quality Criteria (U.S. EPA, 2018).
(Includes a targeted literature search for relevant QMRAs.)
• Appendix A of Review of Published Studies to Characterize Relative Risks from Different Sources
of Fecal Contamination in Recreational Water (U.S. EPA, 2009a).
• Appendix B of Review of Zoonotic Pathogens in Ambient Waters (U.S. EPA, 2009b).
• Literature Search Strategy in State-of-the-Science Review of Quantitative Microbial Risk
Assessment: Estimating Risk of Illness in Recreational Waters which is Annex 1 of Quantitative
Microbial Risk Assessment to Estimate Illness in Freshwater Impacted by Agricultural Animal
Sources of Fecal Contamination (U.S. EPA, 2010a).
• Appendix B of Assessment of the Extra-Enteric Behavior of Fecal Indicator Organisms in
Ambient Waters (U.S. EPA, 2010b).
Additionally, citation mining (reverse searches) was conducted for the following publications:
• Microbiological Risk Assessment (MRA) Tools, Methods, and Approaches for Water Media (U.S.
EPA, 2014b).
• Recreational Water Quality Criteria (U.S. EPA, 2012).
• Recommended Human Health Recreational Ambient Water Quality Criteria or Swimming
Advisories for Microcystins and Cylindrospermopsin (U.S. EPA, 2019).
• Guidelines on Recreational Water Quality. Volume 1: Coastal and Fresh Waters (WHO, 2021).
• Soller, J.A., M.E. Schoen, T. Bartrand, J.E. Ravenscroft, and N.J. Ashbolt. 2010b. Estimated
human health risks from exposure to recreational waters impacted by human and nonhuman
sources of faecal contamination. Water Research 44(16):4674-4691. (Includes primary citations
for parameters related to pathogens and FIB in feces and dose-response models.)
30 https://www.epa.gov/wqc/research-supporting-development-2012-recreational-water-qualitv-criteria
A-l
-------
Subject matter experts, authors, and reviewers listed in the acknowledgements page suggested
additional references to include.
Targeted literature searches were conducted from 2011 to 2021 during the development of the TSM to
supplement the literature searches listed above. The most recent literature search to update
parameters is shown in Table A-l. Of the 122 title/abstracts returned by the literature search, full text
was reviewed for 37, and 12 of those had new information on parameters. Some of the 12 had
duplicate information, so only adenovirus and norovirus had new information that was incorporated
into the latest iteration of the modeling.
A targeted literature search was conducted for gull feces parameters. The pathogen prevalence,
density of pathogens, and fraction that are human infectious, were the parameters of interest. The
following searches were conducted:
• Google Scholar on 5/2/2022 for terms "enterococci and gull and feces and culture" with
restrictions of 2011-2022.
• Google Scholar on 5/4/2022 for terms "gull and feces and Campylobacter" with restriction to
2011-2022.
The searches resulted in 1,800 titles, sorted by relevance. The top 30 titles were screened from each
search. Six articles were reviewed in full text. The cited articles from two reviews, Minias (2020) and
Smith et al. (2020), were used to identify additional articles that contained prevalence and density data
for E. coli, Campylobacter, and Salmonella in gull feces. In addition, 39 articles were reviewed that cited
the most relevant EPA article, Lu et al. (2011).
References
Lu, J., H. Ryu, J.W. Santo Domingo, J.F. Griffith, and N. Ashbolt. 2011. Molecular detection of
Campylobacter spp. in California gull (Larus californicus) excreta. Applied and Environmental
Microbiology 77(14):5034-5039.
Minias, P. 2020. Contrasting patterns of Campylobacter and Salmonella distribution in wild birds: a
comparative analysis. Journal of Avian Biology 51.
Smith, O.M., W.E. Snyder, and J.P. Owen. 2020. Are we overestimating risk of enteric pathogen
spillover from wild birds to humans? Biological Reviews 95(3):652-679.
Soller, J.A., Schoen, M.E., Bartrand, T., Ravenscroft, J.E., Ashbolt, N.J. 2010b. Estimated human health
risks from exposure to recreational waters impacted by human and non-human sources of
faecal contamination. Water Research 44(16):4674-4691.
U.S. EPA (U.S. Environmental Protection Agency). 2009a. Review of Published Studies to Characterize
Relative Risks from Different Sources of Fecal Contamination in Recreational Water. EPA-822-R-
09-001. U.S. Environmental Protection Agency, Office of Water, Health and Ecological Criteria
Division, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2009b. Review of Zoonotic Pathogens in Ambient
Waters. EPA 822-R-09-002. U.S. Environmental Protection Agency, Office of Water, Health and
Ecological Criteria Division, Washington, DC.
A-2
-------
Table A-l. 2021 literature search for selected parameters.
Volume of water ingested during a primary contact recreation event
Number of records
Search date
PubMed31 search string
identified
("recreational water ingestion" OR "incidental
ingestion of water" OR "oral exposure to water"
12/3/2021
OR "oral exposure recreational activities") AND
("volume of water") AND ("2017/01/01"[Date -
Publication] : "2022/04/13"[Date - Publication])
6
Dose-response relationships and parameters for dose-response models for each
reference pathogen
Number of records
Pathogen
Search date
PubMed search string
identified
("rotavirus") AND ("dose response" OR "dose-
Rotavirus
12/3/2021
response") AND ("water") AND
("2017/01/01"[Date - Publication] :
"2022/04/13"[Date - Publication])
3
("norovirus") AND ("dose response" OR "dose-
Norovirus
12/3/2021
response") AND ("water") AND
("2017/01/01"[Date - Publication] :
"2022/04/13"[Date - Publication])
16
("adenovirus") AND ("dose response" OR "dose-
Adenovirus
12/3/2021
response") AND ("water") AND
("2017/01/01"[Date - Publication] :
"2022/04/13"[Date - Publication])
6
("Cryptosporidium") AND ("dose response" OR
Cryptosporidium
12/3/2021
"dose-response") AND ("water") AND
("2016/01/01"[Date - Publication] :
"2022/04/13"[Date - Publication])
9
("giardia" OR "giardia lamblia") AND ("dose
Giardia lamblia
12/3/2021
response" OR "dose-response") AND ("water")
AND ("2017/01/01"[Date- Publication] :
"2022/04/13"[Date - Publication])
6
("Campylobacter") AND ("dose response" OR
Campylobacter
12/3/2021
"dose-response") AND ("water") AND
("2017/01/01"[Date - Publication] :
"2022/04/13"[Date - Publication])
6
("E. coli 0157:H7" OR "Escherichia coli 0157:H7")
E. coli 0157:H7
12/3/2021
AND ("dose response" OR "dose-response") AND
("water") AND ("2017/01/01"[Date - Publication] :
"2022/04/13"[Date - Publication])
2
("salmonella enterica") AND ("dose response" OR
Salmonella enterica
12/3/2021
"dose-response") AND ("water") AND
("2017/01/01"[Date - Publication] :
"2022/04/13"[Date - Publication])
31 https://pubmed.ncbi.nlm.nih.gov
A-3
-------
Probability of illness given infection for each reference pathogen
Pathogen
Search date
PubMed search string
Number of records
identified
Rotavirus
12/5/2021
("rotavirus") AND ("probability of infection" OR
"probability of illness" OR "morbidity") AND
("water") AND ("2017/01/01"[Date - Publication] :
"2022/04/13"[Date - Publication])
22
Norovirus
12/5/2021
("norovirus") AND ("probability of infection" OR
"probability of illness" OR "morbidity") AND
("water") AND ("2017/01/01"[Date - Publication] :
"2022/04/13"[Date - Publication])
16
Adenovirus
12/5/2021
("adenovirus") AND ("probability of infection" OR
"probability of illness" OR "morbidity") AND
("water") AND ("2017/01/01"[Date - Publication] :
"2022/04/13"[Date - Publication])
7
Cryptosporidium
12/5/2021
("Cryptosporidium") AND ("probability of infection"
OR "probability of illness" OR "morbidity") AND
("water") AND ("2017/01/01"[Date - Publication] :
"2022/04/13"[Date - Publication])
27
Giardia lamblia
12/5/2021
("giardia" OR "giardia lamblia") AND ("probability
of infection" OR "probability of illness" OR
"morbidity") AND ("water") AND
("2017/01/01"[Date - Publication] :
"2022/04/13"[Date - Publication])
21
Campylobacter
12/5/2021
("Campylobacter") AND ("probability of infection"
OR "probability of illness" OR "morbidity") AND
("water") AND ("2017/01/01"[Date - Publication] :
"2022/04/13"[Date - Publication])
15
E. coli 0157:H7
12/5/2021
("E. coli 0157:H7" OR "Escherichia coli 0157:H7")
AND ("probability of infection" OR "probability of
illness" OR "morbidity") AND ("water") AND
("2017/01/01"[Date - Publication] :
"2022/04/13"[Date - Publication])
3
Salmonella enterica
12/5/2021
("salmonella enterica") AND ("probability of
infection" OR "probability of illness" OR
"morbidity") AND ("water") AND
("2017/01/01"[Date - Publication] :
"2022/04/13"[Date - Publication])
14
A-4
-------
U.S. EPA (U.S. Environmental Protection Agency). 2010a. Quantitative Microbial Risk Assessment to
Estimate Illness in Freshwater Impacted by Agricultural Animal Sources of Fecal Contamination.
EPA-822-R-10-005. U.S. Environmental Protection Agency, Office of Water, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2010b. Assessment of the Extra-Enteric Behavior of
Fecal Indicator Organisms in Ambient Waters. U.S. Environmental Protection Agency, Office of
Water, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2012. Recreational Water Quality Criteria. 820-F-12-
058. U.S. Environmental Protection Agency, Office of Water, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2014c. Microbiological Risk Assessment (MRA) Tools,
Methods, and Approaches for Water Media. EPA-820-R-14-009. U.S. Environmental Protection
Agency, Washington, DC.
U.S. EPA (U.S. Environmental Protection Agency). 2018. 2017 Five-Year Review of the 2012 Recreational
Water Quality Criteria. EPA-823-R-18-001. U.S. Environmental Protection Agency, Washington,
DC.
U.S. EPA (U.S. Environmental Protection Agency). 2019a. Recommended Human Health Recreational
Ambient Water Quality Criteria or Swimming Advisories for Microcystins and
Cylindrospermopsin. EPA-822-R-19-001. U.S. Environmental Protection Agency, Washington,
DC.
U.S. EPA (U.S. Environmental Protection Agency). 2023. Report on the 2nd Five-Year Review of EPA's
Recreational Water Quality Criteria. EPA-822-R-23-003. U.S. Environmental Protection Agency,
Washington, DC.
WHO (World Health Organization). 2021. Guidelines on Recreational Water Quality. Volume 1: Coastal
and Fresh Waters. License: CC BY-NC-SA 3.0 IGO. World Health Organization, Geneva,
Switzerland.
A-5
-------
Appendix B: QMRA Sanitary Survey Form
The purpose of this Pollutant Source Characterization form is to provide a convenient and consistent
mechanism to collect information about microbiological contamination from fecal sources for
(1) short-term QMRA or epidemiological applications (e.g., site characterization) or (2) longer-term
applications (e.g., AWQC site classification or site-specific AWQC).
The form is comprised of an overview form that is to be used to summarize high-level information
about potential pollutant sources, followed by a watershed information form and a series of
worksheets that are used to summarize details related to specific potential sources of fecal
contamination in a watershed or at a recreational site. Justification for the data requested in specific
sections is provided following the worksheets at the end of the document.
Logistically, it is expected that the watershed information form and the series of worksheets will be
completed first. Then, that information will be used to provide the high-level summary information on
the overview form.
The EPA has published a sanitary survey app, which can be accessed at: https://www.epa.gov/beach-
tech/sanitary-surveys-recreational-waters.
B-l
-------
SUMMARY OF POTENTIAL POLLUTANT SOURCES
Overall Evaluation of Pollutant Sources in Waterbody (narrative describing predominant source of contamination and
justification)
Class and type of source
Supporting
worksheet
cross
reference
Distance
from site of
recreational
waterbody
(feet)
Travel time to
recreational
waterbody
(base / peak
flow)
Relative
contribution of
contamination
(H, M, L, NA)
(base 1 peak
flow)
Summary justification
(Use numbers to cross
reference with description
in overall evaluation above)
HUMAN SOURCES
Wastewater discharges
A, B, C
Type of treatment,
disinfection, etc.
Sewage overflows
and combined sewer
overflows
A, B, C
Frequency, magnitude,
duration
Septic systems and
Unsewered areas
A, B, C
Number, proximity
Subsurface sewage
disposal
A, B, C
Proximity, volume
Marinas, harbors,
mooring boats
A, B, C
Proximity, number, temporal
trends
Bathhouses
A, B, C
Number, proximity, potential
for leakage
Designated swimming
areas, beaches
A, B, C
Number of recreators,
temporal trends
B-2
-------
Class and type of source
Supporting
worksheet
cross
reference
Distance
from site of
recreational
waterbody
(feet)
Travel time to
recreational
waterbody
(base / peak
flow)
Relative
contribution of
contamination
(H, M, L, NA)
(base / peak
flow)
Summary justification
(Use numbers to cross
reference with description
in overall evaluation above)
NONHUMAN INDIRECT
SOURCES
Stormwater outfalls
A, B, D
Number, volume, temporal
trends
Urban runoff, industrial
waste
A, B, D
Number, volume, temporal
trends
Other drains and pipes
A, B, D
Number, volume, temporal
trends
Wetland drainage
A, B, D
Stream volume, temporal
trends
Groundwater seepage
A, B, D
Volume, relative contribution
Streams or natural
outfalls
A, B, D
Volume, relative contribution
ANIMAL SOURCES
CAFOs or AFOs
A, B, E
Number of animals, which
animals
Other agriculture runoff
A, B, E
Number, which animals
Wildlife
A, B, E
Number, which animals
Domestic animals
A, B, E
Number, which animals
Birds
A, B, E
Number, temporal patterns
OTHER (specify)
B-3
-------
1.A. WATERSHED INFORMATION
Name of Watershed or Beach:
Date(s) of Survey:
Watershed ID:
Number of Routine Surveys Used:
T own/City/County/State:
Name(s) of Surveyor(s):
Sampling Station(s)/ID:
Surveyor Affiliation:
STORET Organizational ID:
1.B INFORMATION ON WATER QUALITY SAMPLING LOCATION
Description of Sample Points
Sample point name/ID
Location
Description
Sample frequency
Time of day of
sample collection
Description of hydrometric network (note that this is a network of monitoring stations that collect data such as rainfall and stream
flow)
Comments/Observations:
B-4
-------
WORKSHEET A
DESCRIPTION OF LAND USE, MAPS, AND PHYSICAL CONDITIONS
A.l CURREINI
T LAND USE IN WATERSHED OR BEACH
Type
Residential
Industrial
Commercial
Agricultural
Other (specify):
Percentage
Overall Development Describe
% undeveloped
% developed
Waterbody Uses: ~ Boating ~ Fishing ~ Surfing ~ Windsurfing ~ Diving ~ Swimming
Are maps of the watershed attached? ~ yes ~ no
~ Other (Specify)
List maps and their sources:
Sample Points
~ yes
~ no
(explain)
Hydrometric Network
~ yes
~ no
(explain)
Pollutant Sources
~ yes
~ no
(explain)
Boat Traffic
~ yes
~ no
(explain)
Marinas
~ yes
~ no
(explain)
Boat Dockage
~ yes
~ no
(explain)
Fishing
~ yes
~ no
(explain)
Bathing / Swimming
~ yes
~ no
(explain)
Bounding Structures
~ yes
~ no
(specify: do they impact circulation?):
Residential Areas
~ yes
~ no
(explain)
WWTP Outfalls
~ yes
~ no
(explain)
Stormwater Outfalls
~ yes
~ no
(explain)
Other Drains
~ yes
~ no
(explain)
Sanitary Facilities
~ yes
~ no
(explain)
Septic Systems or
Unsewered Areas
~ yes
~ no
(explain)
CAFOs or AFOs
~ yes
~ no
(explain)
Wetlands
~ yes
~ no
(explain)
Other
~ yes
~ no
(explain)
Image Number
Date/Time
File Name
Description of Photograph
(Include Pictures of High Watermark Locations and Corresponding
Fixed Objects)
B-5
-------
A.2. PHYSICAL CONDITIONS OF WATERSHED OR BEACH
Recreational waterbody length or dimensions (indicate Z1, Z2, and Z3 on a map)
Length (m):
Width average
(m):
Depth min, average,
max (m):
Width Z1 (m):
Width Z2 (m):
Width Z3 (m):
Local water level variation:
feet
inches
Hydrographic influences:
Characterize any longshore or nearshore currents and their potential effects based on bacteria sampling results
Approximate beach slope at swim area %
(If appropriate):
Summary description of recreational waterbody
Other Comments/Observations:
Outfalls
Number
Description or Comment
Streams or natural outfalls
Storm drains
WWTPs
Other (specify):
Other (specify):
Beach Materials/Sediments:
~ Sandy ~ Mucky Q Rocky ~ Other:
Is sand or sediment known to contribute to bacterial concentrations? (explain)
Habitat around beach: (Provide distance from recreational waterbody)
Dunes ~ Wetlands ~ River/Stream Forest ~ Park ~ Protected Habitat or Reserve
~ Other:
B-6
-------
WORKSHEET B
WEATHER CONDITIONS, WATER QUALITY SAMPLING, AND MODELING
B.1. WEATHER
Examine the weather data collected over the prior recreational water season(s) along with bacteria sampling results.
Do the fecal indicator concentrations at this watershed or beach area appear to correlate with any of the following?
Rainfall
~ yes
~ no
(explain)
Air Temperature
~ yes
~ no
(explain)
Water Temperature
~ yes
~ no
(explain)
Cloud Cover
~ yes
~ no
(explain)
Wind Speed
~ yes
~ no
(explain)
Wind Direction
~ yes
~ no
(explain)
Other Weather
~ yes
~ no
(explain)
Have any statistical analyses been done to calculate the degree of correlation? ~ yes ~ no
Describe any analyses done, and any trends or correlations found (add lines if needed to describe in detail):
Rainfall total for the beach season (in):
Average rainfall for all beach seasons (in):
Number of significant rain events:
What constitutes "significant?" (e.g., 1 inch or more rain)
Additional Comments/Observations:
B.2. WATER QUALITY SAMPLING
Name of laboratory:
Distance to laboratory:
miles
Is there a sampling and analysis plan? ~ yes ~ no
Is it adequate? ~ yes ~ no (explain)
Are the sampling staff properly trained on sampling techniques, equipment maintenance, and calibration ~ yes ~ no
procedures?
Microbiological Sampling
Do you test for Escherichia co//? ~ yes ~ no Analytical Method Used:
Do you test for Enterococcus? dyes ~ no Analytical Method Used:
Do you test for fecal coliform? Dyes ~ no Analytical Method Used:
List any additional indicators or pathogens tested and associated analytical methods:
How do this past season's bacteria results compare to that of previous years? (explain if possible)
Do the indicator or pathogen results correlate to other parameters, such as water quality, weather, flow, bather load, algae, or
wildlife?
(If yes, explain) ~ yes ~ no
Describe in detail analyses that were performed on historical data or that typically are performed on data that are collected.
B-7
-------
Were there any unusual results, such as extremely high or low values detected, or unusual trends?
If yes, explain what was found and any potential causes:
~ yes ~ no
Are water quality annual trend data attached? ~ yes ~ no
Comments/Observations:
B.3. ADVISORIES/CLOSINGS
List any advisories and closings that occurred, whether bacteria levels were high, and any possible reasons for advisory or closing
or high bacteria level, such as stormwater runoff, sewage spill, or wild
ife on the beach.
Advisory or closing
(specify one)
Start and end dates
Length of
advisory or
closing (days)
Did bacteria
concentrations
exceed GM or
SSM criteria?
Reason for advisory or closing or
possible contributing factors
Total number of closings issued: Total number of days under an advisory:
Total number of advisories issued: Total number of days beach was closed:
Comments/Observations:
B.4. MODELING
Are models being used to predict bacteria concentrations or the exceedance of bacteria standards? ~ yes ~ no
If yes, list types of models being used and a brief description of the models:
Comments/Observations:
B-8
-------
WORKSHEET C
Human Sources of Pollution
C.1. WASTEWATER DISCHARGE
Do the presence of wastewater discharges contribute to the fecal indicator concentrations at this watershed or beach area?
~ yes ~ no
If yes, please answer the following:
Facility name:
Does the facility discharge directly to the recreational water? ~ yes ~ no
Proximity to recreational water: (approximate distance in feet) ~ < 500 ~ 500-2,500 ~ > 2,500
Discharge rate:
Approximate travel time to waterbody:
Is there a history of unintended discharge events? ~ yes ~ no
If yes, explain:
Describe treatment and disinfection employed at the treatment facility (Include temporal variability in treatment efficacy, if
appropriate)
Do wet weather events affect the wastewater discharge? (e.g., additional flows, observed impacts to the recreational water)
If yes, describe:
C.2. SEWAGE OVERFLOWS
Does all of the watershed have municipal sewer? ~ yes ~ no
Are there sanitary sewer overflow pipes within the watershed?
~ yes ~ no ~ unknown
If yes, describe location relative to recreational water:
Do the presence of sewage overflows contribute to the fecal indicator concentrations at this watershed or beach area?
~ yes ~ no
If yes, please answer the following:
Frequency of overflows:
Magnitude of overflows: (e.g., overflow volume)
Duration of overflows:
Are there specific conditions under which sewage overflows increase?: (e.g., rainfall events)
B-9
-------
C.3. COMBINED SEWER OVERFLOWS
Are there combined sewer overflow pipes and/or drains within the recreational watershed?
~ yes ~ no ~ unknown
Do the presence of combined sewer overflows (CSOs) contribute to the fecal indicator concentrations at this watershed or beach
area?
~ yes ~ no
If yes, please answer the following:
Frequency of CSOs:
Magnitude of CSOs: (e.g., overflow volume)
Duration of CSOs:
Are there specific conditions under which CSOs increase?: (e.g., rainfall events)
Under conditions when CSOs occur, describe treatment and/or disinfection that occurs, if any:
C.4. SEPTIC SYSTEM AND UNSEWERED AREAS
Do the presence of septic systems or unsewered areas contribute to the fecal indicator concentrations at this watershed or beach
area?
Septic systems: ~ yes ~ no Unsewered areas: ~ yes ~ no
If yes, please answer the following:
Proximity of septic system or unsewered area to recreational water: (approximate distance in feet)
Estimated volume of septic system or unsewered area:
Estimated area of septic system or unsewered area:
C.5. SUBSURFACE SEWAGE DISPOSAL
Is there subsurface sewage disposal at this watershed or beach area? ~ yes ~ no
If yes, please answer the following:
Proximity of subsurface sewage disposal to recreational water: (approximate distance in feet)
Estimated volume of subsurface sewage disposal:
C.6. MARINAS AND MOORING BOATS
Is there a marina or other boat moorings at this watershed or beach area?
Boats moored within watershed: ~ contain bilge tanks
Number of moorings:
Approximate number of boats present: (daily average)
Proximity of marina or moorings to the recreational water: (approximate distance in feet)
Does the presence of boats vary throughout the year (explain)?: (e.g., are there seasonal trends?)
~ yes ~ no
~ do not contain bilge tanks
B-10
-------
C.7 DESCRIPTION OF BATHHOUSES OR SANITARY FACILITIES
Bathhouses:
"otal number of bathhouses/toilet facilities? Relevant to watershed or beach:
Number or
ID
Location description
Condition
(good, fair, poor)
Distance from beach (feet)
Frequency of cleaning
(daily, weekly, monthly)
Describe further. Include number of toilets, s
lowers, sinks, etc., and whether these facilities are adequate to support beach use.
C.8 BATHER LOAD (# OF BEACH USERS)
Is the number of bathers or recreators monitored? ~ yes ~ no
If yes, describe how beachgoer numbers are calculated (i.e., turnstile, counting at noon, photographs):
Beach Use
Beachgoer category
Number of People Per
Day Using the Beach
Peak use for
the season
(daily use)
Seasonal
average
(daily use)
Holiday
average
(daily use)
Weekend
average
(daily use)
Weekday
average
(daily use)
Off-season
average, if
applicable
(daily use)
Total beach users
People in the water
People on beach but
not swimming
People boating
People fishing
People surfing
People windsurfing
People diving
Other (specify):
Frequency of
measurements
(e.g., biweekly, weekly,
monthly)
Examine bather density data along with microbial sampling results for the past beach season(s). Look at each sampling point.
Does bather density appear to correlate with bacteria concentrations at any of these sampling points? Has a statistical analysis
been done? Describe:
Comments/Observations:
B-ll
-------
WORKSHEET D
STORMWATER, URBAN RUNOFF, DRAINS, STREAMS, WETLANDS
D.1. STORMWATER
Are there stormwater overflow pipes and drains within the watershed?
~ yes Q no ~ unknown
Do the presence of stormwater runoff contribute to the fecal indicator concentrations at this watershed or beach area?
~ yes ~ no
If yes, please answer the following:
Type of impact: ~ direct discharge ~ indirect nonpoint source ~ both direct and nonpoint
sources
Proximity of discharge to recreational water: (approximate distance in feet)
Frequency of discharge:
Magnitude of discharge: (e.g., overflow volume)
Duration of discharge:
Is there a known or suspected predominant source of fecal contamination in the stormwater?
~ yes ~ no ~ unknown
If yes, explain:
D.2. URBAN RUNOFF (NONSTORMWATER)
Does nonstormwater, urban runoff contribute to the fecal indicator concentrations at this watershed or beach area?
~ yes ~ no
If yes, please answer the following:
Are there any best management practices in place to control sources of urban runoff? ~ yes ~ no
Proximity of discharge to recreational water: (approximate distance in feet)
Frequency of discharge:
Magnitude of discharge: (e.g., overflow volume)
Duration of discharge:
Is there a known or suspected predominant source of fecal contamination in the urban runoff?
~ yes ~ no ~ unknown
If yes, explain:
D.3. OTHER DRAINS AND PIPES
Are there other drains and pipes at this watershed or beach area?
~ yes ~ no
If yes, please answer the following:
Identify the type of drain or pipe:
Proximity of discharge to recreational water: (approximate distance in feet)
Frequency of discharge:
Magnitude of discharge: (e.g., overflow volume)
Duration of discharge:
Is there a known or suspected predominant source of fecal contamination in the drains or pipes?
~ yes ~ no ~ unknown
If yes, explain:
B-12
-------
D.4. WETLAND DRAINAGE
Is there a wetland drainage at this watershed or beach area?
~ yes ~ no
If yes, please answer the following:
Proximity of wetland drainage to the recreational water: (approximate distance in feet)
Frequency of drainage:
Magnitude of drainage: (i.e., drainage volume)
Duration of drainage:
Describe the conditions during which the wetland drains: (e.g., rainfall events)
D.5. GROUNDWATER SEEPAGE
Is there evidence of groundwater seepage at this watershed or beach area?
~ yes ~ no
If yes, please answer the following:
Proximity of groundwater seepage to recreational water: (approximate distance in feet)
Frequency of groundwater seepage:
Magnitude of groundwater seepage: (i.e., seepage volume)
Duration of groundwater seepage:
Are there any conditions under which groundwater seepage increases?: (e.g., rainfall events)
D.6. STREAMS OR NATURAL OUTFALLS
Are there streams or natural outfalls contributing to the fecal indicator concentrations at this watershed or beach area?
Streams: ~ yes ~ no Natural outfalls: ~ yes ~ no
Number of streams and natural outfalls:
Proximity of discharges to recreational water: (approximate distance in feet)
Frequency of discharges:
Magnitude of discharges: (i.e., stream or outfall volume)
Duration of discharges:
Are there any conditions under which stream or natural outfall discharge increases?: (e.g., rainfall events)
B-13
-------
WORKSHEET E
AGRICULTURAL ANIMALS AND WILDLIFE
E.1. CAFOS/AFOS
Do the presence of animal feeding operations contribute to the fecal indicator concentrations at this watershed or beach area?
~ yes ~ no
If yes, please answer the following:
Type of impact: ~ direct discharge ~ indirect nonpoint source ~ both direct and nonpoint sources
Are there any best management practices in place to control sources of runoff? ~ yes ~ no
Proximity of discharge to recreational water: (approximate distance in feet)
Frequency of discharge:
Does the frequency of discharge vary temporally? (e.g., monthly, by season)
Magnitude of discharge: (e.g., overflow volume)
Duration of discharge:
Are there any conditions during which CAFO/AFO drainage increases? (e.g., rainfall events)
What animals are present in the CAFO/AFO?
E.2. OTHER AGRICULTURAL RUNOFF
Are there other sources of agricultural runoff that contribute to the fecal indicator concentrations at this watershed or beach area?
~ yes ~ no
If yes, please answer the following:
Are there any best management practices in place to control sources of agricultural runoff? ~ yes ~ no
Proximity of runoff to recreational water: (approximate distance in feet)
Frequency of runoff:
Does the frequency of runoff vary temporally? (e.g., by season)
Magnitude of runoff: (e.g., overflow volume)
Duration of runoff:
Are there any conditions during which other sources of agricultural runoff increase? (e.g., rainfall events)
What animals contribute to the agricultural runoff?
B-14
-------
E.3. PRESENCE OF ANIMALS AND WILDLIFE
Presence of Wildlife and Domestic Animals
Type
Degree of
presence
(low, mod,
high)
Does the presence appear
to correlate with bacteria
results? (yes, no, don't
know)
Describe further (include whether fecal droppings are seen and
are a problem)
Geese
Gulls
Dogs
Horses:
Other (specify):
Other (specify):
E.4. DESCRIPTION OF OTHER FACILITIES
List facilities in the watershed or beach area that indicate presence homeless encampments or animals, such as dog parks.
Facility name/type
Location
Condition
(good, fair, or poor)
Distance from
beach (feet)
How might this facility
contribute to water quality
problems?
Comments/Observations:
B-15
-------
DATA NEEDS JUSTIFICATION
Summary of Potential Pollutant Sources:
The information that is presented on the summary sheet provides an overview of the predominant
fecal pollution at a particular site, that is, whether it is from human sources or not. This information
will be useful in the short term for QMRA purposes, as it will determine the specific pathogens of
interest. This information is also important for site-specific alternative criteria because it is used to
determine which pathogens the site-specific QMRA will focus on. More specifically, the distance from
and travel time to the recreational waterbody will be useful for determining the relative impact of a
specific pollutant source on a specific site. Again, the sources of fecal contamination will determine the
pathogens of interest, and that information will directly feed into the exposure assessment component
of a QMRA.
Section 1: Basic Information
It is assumed that the information in Sections 1A and IB will be needed for record keeping purposes
for any of the short- or longer-term potential uses of the survey. None of this information will be used
for specific components of a QMRA analysis, but as indicated, would likely be needed to keep track of
the areas to which the information collected later in the form refers.
Information related to sampling location: In the same way that the information in Section 1A is
needed, it is also assumed that the information will be needed for record keeping purposes for any of
the short- or longer-term potential uses of the survey. None of this information would be used for
specific components of a QMRA analysis, but as indicated, would likely be needed to keep track of the
areas to which the information collected later in the form refers.
Worksheet A
Information related to land use: These characteristics will be important for determining whether the
predominant fecal pollution at a particular site is from human sources or not. If the land uses are
residential, industrial, or commercial, it could be inferred that there is the potential for human
contamination, whereas if the land use is all (or nearly all) agricultural, it may be that the primary
source is nonhuman. This information would be useful in the short term for QMRA as it will determine
the specific pathogens of interest. This information would also be useful in the longer term as part of
the AWQC site classification (that is, which AWQC site type[s] is [are] most appropriate for a particular
location—human impacted or impacted by other types of sources). Another potential long-term use is
to develop site-specific criteria, in which case this information would be used to determine the
pathogens on which the site-specific QMRA would be focused.
Information related to maps and photographs: These characteristics will be important for determining
whether the predominant fecal pollution at a particular site is from human sources or not. If any of the
anthropogenic-based map categories are present, it could lead to a presumption that there is the
potential for human contamination. Similar to land use information, this information would be useful
in QMRA for determining which pathogens would be of most interest (if human sources are not
present, then viruses are likely less critical). The same type of logic would be used for longer-term
purposes (site classification or development of site-specific criteria).
Information related to physical conditions: The information in this section is important for describing
the beach area, but does not provide information directly related to QMRA. However, longshore or
B-16
-------
nearshore currents could be considered as events that could potentially be managed if specific tides
and/or currents correlated with increased exposures to indicators and pathogens. Again, this
information would be useful in the exposure assessment component of QMRA.
Worksheet B
Information related to weather conditions: The information in this worksheet could be used in
predictive models and/or by beach managers to identify conditions which are known to be of higher
risk and to manage recreational areas (post, close, etc.), provided that this type of approach can be
approved by the EPA. In these cases, the exposure assessment would result in increased levels of
indicator (and presumably pathogen) exposures as compared to nominal conditions.
Information related to water quality sampling: These characteristics will be important for longer term
uses, such as determining whether, or demonstrating that, site-specific criteria may be needed. For
example, if there are no human sources, but there are bird sources, and the indicator levels are such
that the new/revised AWQC appear to be not applicable for a specific site, these data would be needed
to make that determination. Information related to unusual results would be used to demonstrate that
specific hazardous events impact water quality.
Information related to advisories and/or closings The information in this section is important for
describing the beach area, but do not provide information directly related to QMRA.
Information related to modeling: The information in this section is important for describing the beach
area, but is not necessarily needed to provide information directly related to QMRA.
Worksheet C
Information collected on this worksheet will be used to identify human sources of fecal contamination
that could impact a watershed or beach area. As indicated above, this information will be useful in the
short term for QMRA as it will be used to determine the specific pathogens of interest and in the
longer term as part of the AWQC site classification (that is, which AWQC site type[s] is [are] most
appropriate for a particular location—human impacted or impacted by other types of sources).
Another potential long-term use is to develop site-specific criteria, in which case this information
would be used to determine the pathogens on which the site-specific QMRA would be focused.
Specifically, these data would inform the exposure component of a QMRA, whether the purposes of
that QMRA are short or long-term.
Information related to wastewater discharge: The information in this section is used to identify the
potential water quality impacts of wastewater treatment facilities discharging to the waterbody of
interest. In addition to being used to identify the presence of such a discharge, data are requested to
characterize the proximity, discharge rate, travel time, and the frequency of any unintended discharge
events. The type of treatment and disinfection employed are also important, as they will determine the
relative level of indicators and pathogens that potentially could be present.
Information related to sewage overflows and CSOs: The information in this section is used to identify
the potential water quality impacts of sewage overflows and CSOs. In addition to being used to locate
the presence of such types of events, data are requested to characterize the proximity, discharge rate,
travel time, and frequency of such events. The presence of these types of events would provide strong
evidence of human-based fecal contamination.
B-17
-------
Information related to unsewered areas and subsurface sewage disposal: This information will be
important for determining whether the predominant fecal pollution at a particular site is from human
sources or not. If unsewered areas or subsurface sewage disposal are close to a particular watershed, it
may be difficult to consider the primary fecal pollution source as nonhuman.
Information related to marinas and sanitary facilities: These characteristics will be important for
determining whether the predominant fecal pollution at a particular site is from human sources or not.
If marinas or sanitary facilities are close to a particular beach, it may be difficult to consider the primary
fecal pollution source as nonhuman.
Information related to bather load: These characteristics will be important for determining whether
the predominant fecal pollution at a particular site is from human sources or not. This information
could be used in QMRA to determine which pathogens are of highest concern in a particular
waterbody. If human sources are considered important, the exposure assessment of QMRA will be
used to account for human enteric viruses in addition to the pathogens that would be of interest from
animal sources. Similarly, these data would be useful for longer term needs such as indicating the
potential presence of human contamination in a waterbody (AWQC site classification). From a policy
perspective, it is possible that QMRA could be used to identify a threshold bather density, such that
any waterbody that exceeds some number (threshold) of recreators per volume of water would need
to be classified as a human-impacted waterbody, even if all "upstream" sources are nonhuman.
Worksheet D
Information collected on this worksheet will be used to identify fecal contamination from stormwater,
urban runoff, drains, streams, and wetlands that could impact a watershed or beach area. This
information will be useful in the short term for QMRA as it will determine the specific pathogens of
interest and in the longer term as part of the AWQC site classification (that is, which AWQC site type[s]
is [are] most appropriate for a particular location—human impacted or impacted by other types of
sources). Another potential long-term use is to develop site-specific criteria, in which case this
information would be used to determine the pathogens on which the site-specific QMRA would be
focused.
Information related to stormwater. urban runoff, and drains: The data collected in these sections will
be used to document the impacts of stormwater, urban runoff, and drains on the waterbody of
interest. Unlike data in the previous worksheet, these sources may be predominantly due to human or
nonhuman sources. In each case, the proximity, frequency, magnitude, and duration will help to
quantify the impacts of the sources. The question related to the known or suspected predominant
source will be critical for use in QMRA related activities.
Information related to wetlands, groundwater seepage, and natural outfalls: The data collected in
these sections will be used to document the impacts of wetlands, groundwater seepage, and natural
outfalls on the waterbody of interest. Unlike data in the previous section, these sources may give a
strong indication that contamination could be predominantly due to nonhuman sources. In each case,
the proximity, frequency, magnitude, and duration will help to quantify the impacts of the sources. Like
the previous section, the question related to the known or suspected predominant source will be
critical for use in QMRA related activities.
B-18
-------
Worksheet E
The information collected in this worksheet will be important for determining whether the
predominant fecal pollution at a particular site is from animal sources and documenting that impact. If
the animal sources are present and human sources are not, it could be inferred that the potential for
human contamination is low. This information will be useful for the exposure assessment component
of QMRA because the presence or absence of particular fecal sources will help to determine which
pathogens will be of highest concern. This information will also be useful for longer-term uses (that is
which AWQC site type[s] is [are] most appropriate for a particular location—human impacted or
impacted by other types of sources). Another potential long-term use is to develop site-specific
criteria, in which case this information would be used to determine the pathogens on which the site-
specific QMRA would be focused.
Information related to presence of CAFOs and AFOs and other agricultural runoff: The information in
these sections document the presence of agricultural animals in the watershed of interest. Data are
also requested to document the proximity, frequency, magnitude, duration, and temporal nature of
the runoff. Data indicating the predominant animals are also requested.
Information related to presence of wildlife and domesticated animals: Similar to the data from the
previous sections in this worksheet, data requested in this section will be used to document the
potential for animals and wildlife to impact the waterbody of interest. If data are available to link these
sources with water quality impacts, those results should be highlighted and stressed. Due to limited
health related data in the scientific literature on this topic, the extent to which these data can be useful
for QMRA in a quantitative sense is not currently known. However, a lack of human-related inputs and
documented impacts due to wildlife, would be valuable information for site characterization.
Information related to other facilities: The information in this section highlights potential nonhuman
fecal contamination.
B-19
-------
Appendix C: Example Sampling and Analysis Plan
Introduction
A sampling and analysis plan (SAP) describes the process for sampling and how the data will be
analyzed, including quality assurance plans. A detailed SAP that is reviewed by others can help ensure
that resources are allocated effectively. Data collection quality assurance procedures are also an
important part of the SAP. The SAP also needs to include the indicator method that is associated with
the water quality standard (WQS) being modified for site-specific alternative criteria. The SAP also
needs to be representative of the meteorological and hydrological conditions in the watershed.
This appendix includes an example SAP for a water quality study that was conducted as part of a
sanitary characterization. Initial information from the watershed suggested specific predominant
sources of fecal contamination in the waterbody of concern. Furthermore, a sanitary survey was
conducted to understand the potential sources of contamination in the waterbody and confirmed the
initial information about the likely sources of contamination. Based on the results, water quality and
fecal source studies were planned, developed, and carried out. The SAP describes the studies that were
conducted to evaluate FIB and pathogenic microorganisms in the waterbody and the presumed
source(s) of contamination.
The SAP in this appendix demonstrates that there is considerable flexibility in the experimental design
for the water quality (fecal indicator, pathogen monitoring, source tracking) study included with the
sanitary characterization step. In the SAP presented, the emphasis was on characterizing the presence
and levels of human pathogens present at the beach site and in waters that were thought to have the
potential to contribute fecal contamination to the beach site. This example SAP did include source
tracking because the initial information suggested human contamination was likely. The example SAP
provides an overview that can be taken into consideration during the iterative development of a SAP
for water quality studies that are conducted as part of a sanitary characterization.
C-l
-------
Boqueron SAP
Purpose of the Data Collection
The purpose of the EPA's Collection of Data for Calibrating the EPA's Quantitative Microbial Risk
Assessment Methodology and Anchoring Results to an Epidemiology Study is to provide pathogen and
indicator surveillance data in conjunction with an epidemiological study occurring during the summer
of 2009 at Boqueron Beach, Puerto Rico. Potential effects on the recreational beach include a
wastewater treatment plant (WWTP) outfall, two packaging plant outfalls, and other unknown fecal
pollution sources.
Water samples from the beach and major fecal pollution source locations will be collected on
Saturdays and Sundays for nine weeks beginning June 13, 2009. Water samples will be analyzed for a
suite of pathogens and indicator organisms. The microbial water quality data will be used to anchor a
quantitative microbial risk assessment (QMRA) model; therefore, concentrations of indicators and
pathogens will be quantified when possible.
The information from this data collection effort will not be used for the development of a recreational
water quality criteria (RWQC) value or for any other regulatory or rulemaking purposes. QMRA will be
used to support the implementation of new criteria. Some of the methods used for pathogen analysis
are research or experimental but have been determined to be the most viable in order to obtain the
results.
Sampling Schedule and Locations
Water samples will be collected from each sampling site once on Saturday and once on Sunday, every
weekend for nine consecutive weekends beginning June 13. The weekend of June 6 served as a training
weekend for the field and laboratory staff, during which a dry run of an entire 2-day sampling event
was conducted by all sampling personnel under the guidance of the Laboratory A Field Sampling
Manager (for more detail on the training, see Section III).
A reconnaissance mission of the area was conducted during the week of May 25, 2009, to determine
the greatest potential sources of pathogens to Boqueron Beach, after which exact sampling points
were confirmed. Fecal contamination sources to be sampled include (1) the effluent of a publicly
owned treatment work (POTW) (Prassa WWTP) before discharging to the outfall piping, (2) the effluent
of a small package WWTP (Package Plant #2) before discharging to the outfall piping, and (3) the
mouth of a lagoon into which multiple fecal contamination sources likely flow.
In addition, one combined sample will be collected from Boqueron Beach. This sample will be a
combination of three sampling locations corresponding to the waist-depth sample points at the three
transects being sampled in the EPA epidemiological study. Three field team members will each carry a
carboy, wading into the water, following in the footsteps of field team members of the EPA's
epidemiological study being conducted at the same time. Once the field team members have collected
their samples, they will walk approximately 10 feet south in the water before walking straight back to
the sand (out of the water). The EPA determined this was the least disruptive way to collect samples
and minimize disruption to the sediment. In addition, collecting the samples immediately after the
epidemiological study field team (approximately 8:00 a.m.) will provide the maximum time possible
C-2
-------
between the morning and noon samples collected by the epidemiological study field team, minimizing
disruption to their noon sample. Details about the sampling are included in Annex A.
Table C-l includes descriptions of the sample locations for the study. Global positioning system (GPS)
coordinates for each sampling site will be used during the first field event to verify that samples are
collected from the correct locations. If, for some reason, the source sampling locations cannot be
accessed during future sampling events (e.g., because of weather or a docked vessel), samples will be
collected from a location as close as possible to the designated locations, and the alternate location
will be noted on the Field Form.
Training Field Team Members
The weekend of June 6 served as a training weekend for the field and laboratory staff, during which a
dry run of an entire 2-day sampling event was conducted by all sampling personnel under the guidance
of the Laboratory A Field Sampling Manager. Both teams collected samples following the sampling
protocol (provided in Annex A) on Saturday and Sunday at all four sampling sites. Samples were
collected, placed on ice in coolers, and brought back to Laboratory B (grab samples) and Laboratory C
(ultrafilters [UFs]). At Laboratory B, samplers were trained on the filtration method for the qPCR
samples for E. coli, Enterococcus, and Bacteroidales (collected as a subsample from the grab sample
bottles). For the filtration procedure, see Annex B. Samplers then prepared the UFs for further shipping
(Section 5.E). At Laboratory C, samplers placed the UFs in the refrigerator until Monday, when they
packaged the filters for shipping. Samplers also prepared and cleaned all the sampling equipment and
supplies at Laboratory C.
For the rest of the sampling events, there will be two sampling teams, and each team will sample two
sites per day for the duration of the project. One team will sample the Beach site and the POTW. The
other team will sample the Lagoon site and the Package Plant. However, all sampling team members
have been trained to sample each location and can switch teams if needed. Table C-2 provides the
names and affiliations of the sampling team members. The names in bold are the team leaders.
Additional members might be added in the future. If this occurs, they will be trained appropriately by
one of the field team leaders.
C-3
-------
Table C-l. Sampling locations.
Sample site
number (from
north to south)
Sample site
name
Sample location
description
Access
Comments
1
POTW-
Discharge point
Drive to the POTW,
Samples will be collected at the discharge monitoring point, pumped directly from the
Prassa
of the POTW
access the site on
plant's discharge chamber. The POTW effluent will first be pumped into large tanks (with
WWTP
foot
new clean garbage bags as liners), sodium thiosulfate will be added to neutralize any
chlorine present, and samples will be pumped from the tank through the filter apparatus.
Grab samples will be collected as the tanks are filled, into sterile bottles containing
sufficient sodium thiosulfate to neutralize the chlorine. Sample collection time will be
following collection of the beach sample.
2
Boqueron
The same three
Drive to the beach,
Samplers will arrive at the beach before 8:00 a.m. The sample from each transect will be
Beach
transects as used
park in the public
collected immediately after the epidemiological study water samplers collect their waist-
for the EPA
lot, access the beach
depth sample and exit the water at each transect. Three sampling team members will
epidemiological
to collect the sample
simultaneously wade into the water to the waist-deep sampling point of the transect.
study (waist-
on foot, wading
Each team member will submerge and fill a 20-liter (L) sterile carboy, aiming for a
depth point)
from the beach
sampling depth of about 1 foot below the water surface, taking care to minimize
disruption to the sediments. The samplers will then exit the transect by walking to their
right (south) at least 10 feet (parallel to the beach) before turning back toward shore and
walking out of the water. The carboys might be hauled away from the water with a small
wagon. As the samples are collected in carboys from each transect, they will be
transported out of the water and placed into a vehicle and hauled off the beach to a
staging area for compositing and filtering. The staging area is along the road/parking lot
near the mangroves between the beach site and the lagoon site. Discharge water from
the filters will be discretely poured back onto the sand near the mangroves. No grab
samples will be collected from the beach location.
3 Package Discharge point Drive to the package Samples will be collected at the discharge monitoring point, pumped directly from the
Plant #2 of the package plant, access on foot plant's discharge chamber. The package plant effluent will first be pumped into large tanks
plant (with new clean garbage bags as liners), sodium thiosulfate will be added to neutralize any
chlorine present, and samples will be pumped from the tank through the filter apparatus.
Grab samples will be collected as the tanks are filled, into sterile bottles containing
sufficient sodium thiosulfate to neutralize the chlorine. The time of sample collection will be
either before or after the lagoon sample, depending on the time of low tide.
4 Mangrove At the mouth of Boat (limited shore At low tide, samples will be collected at the mouth of the lagoon, at the midpoint
Lagoon the lagoon access because of between the two channel markers. Samples will be pumped or grabbed (depending on
mangroves) the sample type) directly from the water and into the sample bottles and filter apparatus
on the boat.
C-4
-------
Table C-2. Field team members.
Name
Affiliation
Jane Doe
List affiliation
John Doe
List affiliation
Team member #1
List affiliation
Team member #2
List affiliation
Team member #3
List affiliation
Team member #4
List affiliation
Team member #5
List affiliation
Team member #6
List affiliation
Access/Health and Safety
Water samples from the WWTP will be collected from a trough at the plant. Field team members will
wear appropriate clothing and protection devices if required by the plant (such as closed or steel-toed
shoes and hard hats); however, plant operators have not yet requested this. Shore access to the lagoon
sampling location is limited by the mangrove forest, and samples will be collected by boat. Boats will
be operated by experienced drivers, and personal flotation devices will be available to all members of
the sampling team.
Any field team member who wades out into the water to collect the beach sample will know how to
swim and will not go out if conditions are dangerous (e.g., bad weather, rough water).
Sample Collection Procedures
A. Field Forms, Notebooks, and Photograph Log
The field teams will complete a Field Form at each sampling site during each sampling event (Annex C).
Original copies of the forms will be saved and submitted to the EPA at the end of the project. Copies
will be sent to the EPA periodically during the project when requested by the WAM. The forms will be
completed in their entirety. If anything does not apply, "NA" will be noted on the form.
The samplers will use a field notebook to note any other observations made during a sampling event or
any issues that occurred in addition to what was noted on the Field Forms.
Samplers will take at least one set of photographs at each sample point, to document the sample
location and sample setup. If needed, samplers will take photographs of any unusual occurrences or
issues that might affect sampling results. Samplers will keep a record of photographs in the field
notebook, noting the time, date, location, and description of each image taken. Any photographs taken
and copies of the photograph log will be submitted to the EPA after the last sampling event.
C-5
-------
B. Sample Collection and Handling
1. Grab Samples
For those indicators with short hold times, grab samples will be collected in sterile 1-L Nalgene
containers and split for multiple analyses at Laboratory B. The indicator analyses to be performed at
Laboratory B include enterococci and E. coli by culture methods and filtering for future qPCR analyses
(Enterococcus spp., E. coli, and Bacteroidales). Grab samples will be collected at the POTW, Package
Plant, and Lagoon sites. Grab samples will not be collected at the Beach site.
For grab samples that will be processed in Laboratory B: 3 1-L bottles will be used for sample
collection. Grab samples will be chilled in a cooler immediately after sampling and will be delivered to
and processed at Laboratory B within a 6-hour hold time. Processing will be complete within an 8-hour
hold time (i.e., petri dishes in incubators and membrane filters for qPCR in extraction tubes in a
-20 degrees Celsius [°C] freezer). For each of the qPCR analyses (Enterococcus spp., E. coli, and
Bacteroidales), the analysts will filter 100 mL through each of three 47-mm diameter with 0.4 |am pore
size filters.
2. Ultrafilter Samples
The use of UFs (30,000 kilodaltons [kDa]) operated in the dead end mode was selected for use in this
study because the goal of the study is to analyze pathogens and indicators from four sample sites of
different matrices. This method has been used successfully by scientists from Laboratory A in previous
studies where the focus is multiple pathogen concentration. This is a data collection effort and UF was
determined to be the best way to concentrate samples for pathogen recovery and detection.
UF samples will be collected at all four sites (POTW, Package Plant, Lagoon, and Beach), with two filters
used at each site for each set of samples. At each sample location, up to 200 L of water will be
concentrated for pathogens and indicators using a dead-end ultrafiltration procedure similar to that
developed by the Center for Disease Control and Prevention32 and commonly used for outbreak
investigations. Water samples will be concentrated on two Asahi REXEED-25S or 25SX dialyzers (also
referred to as UFs) using two centrifugal pumps, pushing no more than 100 L water through each filter.
For water samples that are collected at the Lagoon site by boat, the intake tubing will be tied to
weighted line off the side of the boat and positioned about 1 foot below the water surface. Pumps will
be run on 12 V batteries at the Beach and Lagoon sites and on electricity at the POTW and Package
Plant sites, and there will be two dedicated pumps for each sampling location. These filters can be
operated at 2 L-3 L/min, and at 2 L/min, the target volume can be sampled within 1 hour. However,
filters can clog, and flow rates might diminish. Pumping will be terminated if the filters clog and the
flow rate becomes less than 0.5 L/min or if significant backpressure builds up in the line. Only two
filters will be used per sampling location for each sampling event, regardless of what volume of water
is passed through the filter. The volumes filtered through each filter will be recorded by the sampling
crew. Pumps will be disinfected after each use (chlorination followed by dechlorination).
After filtering, the UFs will be capped to seal the filters with the sample water inside the cartridge. The
two UFs from one sampling site will be bagged together. The UFs will be placed in a cooler and then
transferred to the Laboratory B refrigerator until they are shipped on ice Monday afternoon for priority
32 Hill, V.R., A.L. Polaczyk, D. Hahn, J. Narayanan, T.L. Cromeans, J.M. Roberts, and J.E. Amburgey. 2005. Development of a
rapid method for simultaneous recovery of diverse microbes in drinking water by ultrafiltration with sodium polyphosphate
and surfactants. Applied and Environmental Microbiology 71:6878-6884.
C-6
-------
overnight (Tuesday, no longer than 72 hours) delivery to Laboratory A. Filters must not freeze during
shipment. Laboratory A will take the temperature of each filter immediately after receiving the
shipment and will record this on the Chain of Custody Form (Annex D). Filters which have not
maintained the proper temperature will not be analyzed and will be documented. The filters will be
backflushed upon arrival at Laboratory A on Tuesday and subsamples of the UF concentrates will be
shipped priority overnight on Tuesday (to arrive Wednesday morning) to Laboratory D for culturable
virus analyses. Similarly, samples for norovirus analysis will be shipped to Laboratory E and frozen
there and analyzed as a set when all samples have been received. Bacterial analysis at Laboratory A will
begin Tuesday afternoon as soon as UF concentrates are ready.
Figure C-l shows a flow chart that explains how the samples will arrive at each lab and what assays
each lab will perform.
C. Sample Processing and Analysis
Water samples will be processed and analyzed for the list of pathogens and indicators included in Table
C-3. Table C-3 also includes the laboratories that will be performing each analysis. Indicator methods
will be the same as those being performed for the epidemiological study being conducted concurrently
at Boqueron Beach and other EPA beach studies, where possible. Pathogen methods might not be
standard methods—research methods or modified standard methods may be used to ensure the
detection of pathogens and adapt to the sampling conditions.
Grab Samples
Laboratory B
(E coli and
Enteroccocus
culture)
Laboratory F
(.Enteroccocus and
Bacteroidales qPCR)
Laboratory C
(filter, frozen, and
ship in a batch)
EPA ORD
(E coli qPCR)
UF Samples
Laboratory A
(Backflush UF, analyze samples for Cryptosporidium,
Giardia, infectious Cryptosporidium, toxigenic
E coli, non-0157 shiga-toxin strains,
Campylobacter, Salmonella, and C. perfringens;
ship remaining eluate to other labs
Laboratory D
(Enterovirus and
Adenovirus culture)
Laboratory E
(Norovirus,
Adenovirus, and
Enterovirus qPCR)
Figure C-l. Flow chart of laboratories.
C-7
-------
Table C-3. Analytes and methods.
Microbe
Analysis method
Grab or UF
concentrate or other
Equivalent original
sample volume
analyzed
Laboratory
performing
analysis
Pathogens
Norovirus
qPCR
UF concentrate
Up to 20 L
Laboratory E
Adenovirus
Cell Culture Assay for Adenovirus
with A549 Cell Line
UF concentrate
Up to 20 L
Laboratory D
Adenovirus
qPCR
Supernatant supplied
from Laboratory D
To be determined
Laboratory E
Enterovirus
Enterovirus Culture Method Most-
Probable Number Method
UF concentrate
Up to 20 L
Laboratory D
Enterovirus
CIPC-Qpcr
Supernatant supplied
from Laboratory D
To be determined
Laboratory E
Cryptospori dium/
Giardia
EPA Method 1623
UF concentrate
Up to 10 L
Laboratory A
Infectious
Cryptospori di um
Infectious Cryptosporidium Oocyst
by cell culture IF antibody detection
(Texas A&M Method)
UF concentrate
Up to 10 L
Laboratory A
Toxigenic E. coli
0157:H7
Culture Method (IMS, CTSMAC, latex
agglutination confirmation)
UF concentrate
1 L-10 L
Laboratory A
Toxigenic E. coli
0157:H7
qPCR of UF Concentrate
UF concentrate
To be determined
Laboratory A
Non-0157 shiga-
toxin strains
Culture Method (broth enrichment,
plating on Rainbow agar and
CTSMAC agar)
UF concentrate
1 L-10 L
Laboratory A
Non-0157 shiga-
toxin strains
qPCR of UF Concentrate
UF concentrate
To be determined
Laboratory A
Campylobacter
Modification of ISO 17995 (2005),
adapted to MPN test, use latex
agglutination chips to confirm
UF concentrate
1 L-10 L
Laboratory A
Salmonella spp.
Modification of ISO 6340 (1995),
adapted to MPN test, use
enterotubes to confirm
UF concentrate
1 L-10 L
Laboratory A
Indicators
Clostridium
perfringens
MF on MCP agar (modified ICR
method)
UF concentrate
Up to 1 L
Laboratory A
Enterococcus spp.
and Bacteroidales
EPA's duplex TaqMan qPCR method
for Enterococcus spp. and
Bacteroides
Grabs filtered using
0.4 um polycarbonate
filters in Laboratory B
Up to 100 mL in
duplicate
Laboratory F
Enterococci
EPA Method 1600
Grab
Up to 500 mL (in
duplicate or triplicate)
Laboratory B
Male-specific F+
Coliphage
EPA Method 1602
UF concentrate
Up to 1 L
Laboratory A
E. coli
SM9221F (tube fermentation)
Grab
Up to 500 mL (in
duplicate or triplicate)
Laboratory B
E. coli
EPA qPCR (23S)
Grabs filtered using
0.4 um polycarbonate
filters in Laboratory B
Up to 100 mL in
duplicate
EPAORD
C-8
-------
1. Grab Samples
Three 1-L bottles of water will be collected as a grab sample at the Lagoon, POTW, and Package Plant
sites. No grab samples are collected at the Beach site. These samples will be used for the following:
• Duplicate or triplicate 100 milliliter (mL) samples and dilutions will be analyzed locally at
Laboratory B for E. coli using SM9221F and for Enterococcus spp. using EPA Method 1600
(membrane filtration and mEI agar).
• Triplicate 100 mL volumes will be filtered for E. coli, Bacteroidales and Enterococcus spp. qPCR
following the EPA protocols. One filter will be used per analyte, following the protocol used in
the EPA's epidemiological studies. Membrane filters will be frozen for the duration of the
project. Soon after the last sampling event has occurred, the frozen filters will be shipped on
dry ice and sent for deoxyribonucleic acid (DNA) extraction and qPCR at the appropriate
subcontracted laboratory (Bacteroidales and Enterococcus spp. assays will be performed by
Laboratory F,33 and E. coli assays will be performed by the EPA ORD). The EPA standard
operating procedure (SOP) for filtering water samples for qPCR and a quick reference guide
(Annex B) has been provided to the Laboratory B staff for this project.
2. UF Sample Processing and Analysis
• UFs will be received at Laboratory A each Tuesday (shipped from Laboratory B each Monday).
Upon receipt, the filters will be backflushed with 500 mL of backflush solution as described in
Annex E. The resulting UF concentrates from the two filters collected at each site will be
combined for a total volume of approximately 1 L per sample. The final concentrate volumes
will be recorded on bench sheets. The concentrates will then be immediately split for the
various analyses listed in Table C-3.
o For some analyses, UF concentrates will require secondary concentration using
centrifugation and analysis of the resuspended centrifugation pellet, or PEG precipitation
or filtration through Centricon units. (This depends on the sample volume collected and if
salt water or backflush solution needs to be rinsed away before analysis.)
• At Laboratory A, UF concentrate will be analyzed for C. perfringens using membrane filtration
and mCP agar (hold times should have minimal effect on spores).
• At Laboratory A, a UF concentrate volume equivalent to 1 L-10 L original sample volume will be
analyzed for each bacterial pathogen.
o Salmonella spp. by enrichment broth culture (most probable number [MPN]) of UF
concentrate (modification of ISO 6340 [1995]).
o Campylobacter by enrichment broth culture (MPN) of UF concentrate (modification of
ISO 17995 (2005).
o E. coli 0157 and toxigenic non-0157 E. coli by enrichment broth culture (MPN) of UF
concentrate.
33 Any filters which have not maintained the proper temperature will not be analyzed and will be documented.
C-9
-------
• At Laboratory A, a UF concentrate volume equivalent to 1 L-10 L original sample volume will be
analyzed for male-specific coliphage by modification EPA Method 1602 (Double Agar Layer).
• At Laboratory A, a UF concentrate volume equivalent to 20 L original sample volume will be
centrifuged, and the resuspended centrifugation pellet will be split equally and analyzed for
Cryptosporidium and Giardia by EPA Method 1623 (December 2005 version) and for infectious
Cryptosporidium.
• A UF concentrate volume equivalent to up to 40 L original sample volume will be shipped
priority overnight on Tuesday on ice to Laboratory D (to arrive Wednesday morning).
Laboratory D will perform cell culture/MPN analyses for enteroviruses and adenoviruses. In
addition, portions of the supernatant from these analyses will be sent to Laboratory E for PCR
analyses. The shipment of the samples for Laboratory E will depend on the availability of their
lab staff, but will likely be in a batch after the last sampling event.
Any remaining UF concentrate will be frozen and stored at Laboratory A, for potential repeat or
alternate analyses during the course of study or for filtering for future qPCR analyses.
Recovery of Bacteria, Protozoa, and Viruses Using Dead-end Ultrafiltration
Dead-end Ultrafiltration was chosen for this project as the filters are easy to use, can accommodate
large sample volumes (up to 100 L), allow capture of virus particles (molecular weight
cutoff = 30,000 d), and are easily backflushed for recovery of target organisms. To assess recovery of
target organisms of interest, the following study will be conducted at Laboratory A. A total of
eight samples will be analyzed to develop performance data. The study design will be similar to that
used in method performance assessments in Method 1622/23 (Annex F). The goal of this study is to
assess the ability of the method to recover target organisms in reagent water and matrix samples.
Initial Precision and Recovery Experiments: Four filters will be spiked with 100 L reagent water seeded
with Salmonella (~104/L) Giardia and Cryptosporidium (~10/L), and MS234 (~104/L). These organisms
represent the target groups—bacteria, protozoan parasites, and viruses. The seed organisms will be
spiked into 50 L of water and the samples filtered as described in the SAP for sample collection from a
sanitized container. The filters will be spiked in the first half of sample collection, and followed by
rinsing in the remaining 50 L reagent water. Samples will be backflushed and processed as described in
the SAP for each method: Giardia and Cryptosporidium using Method 1623; MS2 using the double agar
overlay; and Salmonella by two methods—the MPN as described in the SAP and by direct plating in XLD
and SS agars. Samples for each target organism will be analyzed in replicate as total sample volume
allows. The mean and standard deviation (SD) for each organism will be calculated.
Matrix Spike Samples: To examine the effect of the beach water matrix, Laboratory A will receive
three 90-L samples of beach water from Boqueron Beach on the last two sampling rounds. The 90-L
seawater samples will be seeded as described above with Salmonella, Giardia, Cryptosporidium, and
MS2 and processed in the same manner. Recoveries and SDs will be calculated for the beach water
matrix.
34 MS2 is a virus that infects E. coli, a bacteriophage. It is a male-specific, also known as F+, bacteriophage. It is often used
for laboratory seeding and recovery experiments and used as a positive control for F+ bacteriophage assays.
C-10
-------
Method Blank: A method blank using 100 L reagent water will be sampled and processed as described
to assess the presence/absence of any of the analytes in the sampling and analysis process.
D. Shipping
All samples shipped from Puerto Rico will be shipped with priority overnight delivery on the Monday
after sample collection on a Saturday and Sunday, to arrive at the labs on Tuesday (no longer than
72 hours) morning. Samplers will compare the Chain-of-Custody form with the samples in the cooler to
make sure all samples are present. Samplers will ensure compliance with U.S. Department of
Transportation and the International Air Transport Association regulations regarding the transfer of
hazardous substances and environmental samples. Samples being shipped from Laboratory B to
Laboratory A will be shipped in standard coolers, double lined with trash bags. UFs will be placed in
resealable bags inside the inner liner. Frozen ice packs will be placed in resealable bags between the
inner and outer cooler liners, with sufficient ice packs used to keep samples properly chilled but not
frozen. Both bags will be tied in a knot or with zip ties. A copy of the Chain of Custody form will be
placed inside a resealable bag and taped to the inside of the cooler lid. Coolers will be taped shut and
dropped off at the shipping center.
Samples that are shipped to Laboratory F and the EPA ORD (filters for qPCR analyses) will be shipped in
coolers using dry ice. After filtering, each filter will be placed in an extraction tube, and all the tubes
will be placed in a Styrofoam tube box in the freezer (-20 °C35) until ready for shipping (after the last
sampling event in August). When ready to ship, the Styrofoam tube box will be placed inside a cooler.
Up to (but not exceeding) 5 lbs of dry ice will be placed inside the cooler, surrounding the tube box. A
Chain of Custody form will be placed inside a resealable bag inside the cooler. The cooler will be sealed
and shipped with the proper labels on the outside of the box, indicating that the shipment contains dry
ice.
Samplers will contact the labs that are receiving shipment to let them know what is being shipped and
to give them a tracking number for each package.
Any field team member who wades out into the water to collect the beach sample will know how to
swim and will not go out if conditions are dangerous (e.g., bad weather, rough water).
Annex A: Sampling Protocol Grab Sampling and Ultrafiltration
The sampling protocol listed below is to be used for collecting samples at Boqueron Beach, Puerto
Rico, during the summer of 2009. There are four sample locations, and sample procedures vary slightly
between each sample site. This protocol assumes that two sampling teams will each collect samples at
two sites (four sites total) for two days in a row, every Saturday and Sunday for nine weeks. The teams
will collect grab samples at each location except the beach and two UFs at each location. One team will
collect samples at the Beach and the POTW sites, and the other team will collect samples at the Lagoon
and the Package Plant sites. However, all team members will be trained to sample each location.
35 The EPA qPCR SOP says filters can be stored at -20 °C or -70 °C. Laboratory B and Laboratory C can only store at -20 °C.
C-ll
-------
A.l Equipment
Table C.A-1 is a list of recommended equipment that each sampling team should have with it for
sampling in Boqueron. Before leaving for the sampling sites, the teams should review the list to make
sure they have all items needed for sampling. This list might be adjusted slightly after the first sampling
event occurs.
A.2. Sampling Setup
Packing of field supplies and equipment should be performed Friday and Saturday evenings using the
equipment and supply list. As part of the sampling preparation, bleach and sodium thiosulfate
solutions will need to be prepared for cleaning the equipment after use and for dechlorinating the
water before sampling at the POTW and Package Plant.
Cleaning Equipment
Before sampling begins each weekend, each team needs to prepare a carboy of 1% bleach36 (100 mL
bleach + 9.9 L distilled water) (based on recommendations in the EPA's ICR Microbial Laboratory
Manual (EPA 1996)37 and a carboy of 0.05% sodium thiosulfate (0.5 g sodium thiosulfate in 10 L
distilled water). The bleach may be used for up to a month as long as residuals remain high (chlorine
level will be measured periodically using a chlorine test kit). These items will be used for cleaning the
equipment after each sampling event (see the Post Sampling Procedures at the end of this appendix).
The sodium thiosulfate carboy should be prepared fresh each weekend.
Dechlorination of water at the POTW and Package Pant
Eight small bottles of 38 g sodium thiosulfate should be prepared, mixed with a small amount of
distilled water (just enough to dissolve) each week for dechlorination at the wastewater sites (POTW
and Package Plant). Two bottles will be used at each of the two sites on each weekend day. This is
likely to be more than what is needed to dechlorinate the wastewater. However, this ensures that
enough sodium thiosulfate will be added in the event that chlorine levels are high (chlorine will be
measured before and after adding the sodium thiosulfate).
A.3 Sampling Procedures
Sampling procedures vary slightly between each sample site. For the Lagoon site, samples will be
collected directly from the lagoon water. For the beach site, carboy samples will be collected from
three transects. These samples will be composited into two large sterile containers, where the water
will be continually recirculated using sterile, dedicated bilge pumps. Sterile, dedicated centrifugal
pumps will be used to pull water from this container and through the UFs. For the POTW and Package
Plant sites, water will also be collected into large containers and continually recirculated using sterile,
dedicated bilge pumps. This water will be dechlorinated using a sodium thiosulfate solution. After
dechlorination, the sanitized, dedicated centrifugal pumps will be used to pull water from the
container through the UFs. Grab samples will be collected during initial filling of the tank into sterile
bottles containing sodium thiosulfate.
3610% bleach is not needed for this monitoring program
37 U.S. EPA (U.S. Environmental Protection Agency). 1996. ICR Microbial Laboratory Manual. EPA/600/R-95/178.
Government Printing Office, Washington, DC.
C-12
-------
Table C.A-1. Equipment list for field teams.
Site 2 (Beach) Quantity ? Site 3 (Package Plant #2) Quantity
Filtering
5' Intake Tubing
2
12V Pump
2
12V Battery
2
Filter Kit
1
Filter
2
5' Discharge Tubing
2
Graduated Discharge Tanks
2
Cleaning
Bleach Carboy
1
Na Thiosulfate Carboy
1
Incidentals
Tubing Cutters
1
Labeling Tape
1 roll
Duct Tape
1 roll
Zip Ties
1 bag
Sharpies/Pens
2
Gloves
1 box
Sanitizer
1
Trash Bags
2
Sampling
20L Carboys
3
Sample Tanks
2
Sterile Tank Liners
2
Bilge Pump
2
12V Battery
1
1-L Bottles
3
Sampling
High Flow Pump
Extension Cord
Sample Tanks
Clean Tank Bags
Bilge Pump
Chlorine Kit
38g Na Thiosulfate Bottle
1-L Bottles
Filtering
5' Intake Tubing
Electric Pump
Filter Kit
Filter
5' Discharge Tubing
Graduated Discharge Tanks
Incidentals
Tubing Cutters
Labeling Tape
Duct Tape
Zip Ties
Sharpies/Pens
Gloves
Sanitizer
Trash Bags
1 roll
1 roll
1 box
1
Site 1 (PRASA POTW)
Quantity ?
Site 4 (Lagoon)
Quantity ?
Sampling
High Flow Pump
1
Extension Cord
1
Sample Tanks
2
Clean Tank Bags
2
Bilge Pump
2
Chlorine Kit
1
38g Na Thiosulfate Bottle
2
1 -L Bottles
3
Filtering
5' Intake Tubing
2
Electric Pump
2
Filter Kit
1
Filter
2
5' Discharge Tubing
2
Graduated Discharge Tanks
2
High Flow Pump Fittings
2
Sampling
1-L Bottles | 3 ~f
No additional equipment needed. Lagoon
water is pumped directly through the filter.
Filtering
Intake Tubing w/Dumb Bell
1
12V Pump
2
12V Battery
2
Filter Kit
1
Filter
2
3' Discharge Tubing
2
Graduated 5-gallon Buckets
4
Incidentals
Tubing Cutters
1
Labeling Tape
1 roll
Duct Tape
1 roll
Zip Ties
1 bag
Sharpies/Pens
2
Gloves
1 box
Sanitizer
1
Trash Bags
2
Teflon Tape
1 roll
Incidentals
Tubing Cutters
1
Labeling Tape
1 roll
Duct Tape
1 roll
Zip Ties
1 bag
Sharpies/Pens
2
Gloves
1 box
Sanitizer
1
Trash Bags
2
C-13
-------
Sampling Procedure for the Lagoon Site
1. Arrive at the site by boat, using GPS coordinates if possible to make sure you are at the correct
point. After first sampling, simply use channel markers to verify correct sampling location.
2. Fill out the Field Form.
3. Using label tape, label each filter with the Sample ID*. Leave space on the label for volume and
times.
Sample ID:
Filter Start Time:
Example of a field label for filters
*The Sample ID incorporates the sample date and site; therefore, these are not listed
separately on the sample labels.
4. Label the 1 L bottle for grab samples with Sample ID. Leave space on the label for times.
Sample ID:
Example of a field label for grab samples
5. Pour the excess water out of the two UFs (through the end port and side port), if using a 25S
(this is not necessary for the 25SX). Attach the molded DIN adapter and molded port cap as
shown in Figure C.A-1.
6. Assemble two filtration systems as shown in Figure C.A-1, except do not connect the UF. Use
new tubing from the box and cut to length with knife (discharge tubing may be reused).
7. Turn on the pumps and flush sample water through the sampling apparatus for 30 seconds and
to prime the pumps.
8. Turn the valve to keep prime, turn off the pumps and reassemble the filtration system with the
UFs as shown in Figure C.A-1. Remember to use hose clamps where directed on the diagram.
9. Turn on the pumps and begin pumping the sample water through the filtration system. Note
the start filtration time on the Field Form.
10. During filtration, collect grab samples and place in cooler with ice. Record the sample collection
time on the bottle label and the Field Form.
C-14
-------
11. During filtration, direct discharge hoses to graduated 5-gallon buckets. Each time the bucket
fills to the 10-L mark, switch out with a new bucket. Dump first bucket over the opposite side of
the boat as the intake line. Keep track of how many 10-L volumes have been dumped. This
procedure will determine the volume collected through the filters.
12. Stop filtration when one of the following has happened:
a. 100 L has been filtered.
b. The pressure increases and tubing begins to bulge.
c. The flow rate has slowed to 0.5 L/min.
13. Record the end filtration time Field Form.
14. On the Field Form record the total volume filtered through each filter and write the volume on
the filter label.
15. Replace the original caps on the UF side port and replace other end port with another molded
cap.
16. Place both UFs in one 2-gallon resealable bag.
17. Place sample bottles and filters in a cooler with ice.
18. Return to the local lab for processing and shipping, or continue to the next sampling site.
C-15
-------
Flow
~
Sampling Setup—1
(for lagoon site, samples collected from a boat)
Pump
Figure C.A-1. Sampling setup at the lagoon site (there will be two identical setups, because two filters will be used).
C-16
-------
Sampling Procedure for the POTW and Package Plant Sites
1. Arrive at the site and go to the effluent discharge sampling point.
2. Set up 30-gallon trashcans and insert clean garbage bag liners and sterile, dedicated bilge
pumps.
3. Fill out the Field Form.
4. Test the effluent using the chlorine test kit. Note the chlorine level on Field Form.
5. Add the pre-measured sodium thiosulfate solution to the tanks.
6. Set up the pump according to Figure C.A-2.
7. Pump water into the tank until about 100-L level, making sure to continuously recirculate the
water using the sterile, dedicated bilge pumps.
8. Collect grab samples as tanks fill and place in cooler on ice. Record the sample collection time
on the bottle label and the Field Form.
9. Once the tank is full, turn off the pump.
10. Recheck the chlorine level and ensure it is zero.
11. Set up the sampling apparatus according to Figure C.A-2.
12. Using label tape, label each filter with the Sample ID. Leave space on the label for volume and
times.
13. Label the 1-L bottles for grab samples with the Sample ID. Leave space on the label for times.
14. Pour the excess water out of two UFs (through the end port and side port), if using a 25S (this is
not necessary for the 25SX). Attach the molded DIN adapter and molded port cap as shown in
Figure C.A-2.
15. Assemble two filtration systems as shown in Figure C.A-2, except do not connect the UF.
16. Use new tubing from the box and cut to length with a knife (discharge tubing may be reused).
17. Turn on the pumps and flush sample water through the sampling apparatus for 30 seconds and
to prime the pumps.
18. Turn the valves to hold the prime and turn off the pumps and reassemble the filtration system
with the UFs as shown in Figure C.A-2. Remember to use hose clamps where directed on the
diagram.
19. Begin pumping the sample water through the filtration system. Note the start filtration time on
the Field Form.
20. Stop filtration when one of the following has happened:
a. 100 liters have been filtered.
C-17
-------
b. The pressure increases and tubing begins to bulge.
c. The flow rate has slowed to 0.5 L/min.
21. Record the end filtration time on the Field Form.
22. Record on the Field Form the total volume filtered through each filter and also write the
volume on the filter label.
23. Replace the original caps on the UF side port and replace other end port with another molded
cap.
24. Place both UFs in one 2-gallon resealable bag.
25. Place sample bottles and filters in a cooler with ice.
26. Dump or pump the discharge water back into the plant's discharge stream.
27. Return to the local lab for processing and shipping, or continue to the next sampling site.
Sampling Procedure for the Beach Site
1. Arrive at the beach at 7:30 a.m. to begin sampling at approximately 8:00 a.m. Drive your vehicle
with sampling gear as close as possible to the transects, while taking care not to disturb the
beach itself, lifeguards, or beachgoers.
2. In the back of the truck, prepare two 30-gallon tanks with sterile tank liner, sterile, dedicated
bilge bumps, and clean intake tubing.
3. Begin filling out the Field Form.
4. Sampling begins immediately after the epidemiology study samplers finish sampling the first
transect and exit the water to move down the beach.
5. All three sampling team members (each with one 20 L carboy) wade into the water at the first
transect until approximately waist-deep (1 meter), lining up with the lifeguard tower right rail
and palm tree.
6. Lower each carboy into the water approximately one foot below the water surface. When
carboys are full, walk to the right (south) parallel to the shoreline approximately 10 feet and
walk out of the water onto the beach. Using the wagon, transport the carboys to the truck.
Pour 30 L into each of the two tanks.
7. Walk to the second transect and repeat Steps 4-6, piggy-backing the epidemiology study
samplers as discretely as possible.
8. Walk to the third transect and repeat Steps 4-6, piggy-backing the epidemiology study
samplers as discretely as possible.
C-18
-------
9. Check the set-up of the two 30-gallon trash cans with sterile liners in the back of the vehicle.
Pour approximately half of each carboy into the first lined trash can, and half into the second.
Tie the top of the liners with zip ties. Load all sampling gear into the vehicle. Drive the vehicle
off the beach, down the road, and pull off to the side near the nearby pathway to the mangrove
lagoon for sample filtering.
10. Using label tape, label each filter with a Sample ID. Leave space on the label for volume and
times.
11. Pour the excess water out of two UFs (through the end port and side port), if using a 25S (this is
not necessary for the 25SX). Attach the molded DIN adapter and molded port cap.
12. Assemble two filtration systems as shown in Figure C.A-2, except without the UF. Use new
tubing from the box and cut to length with a knife. Discharge tubing may be reused.
13. Turn on the pumps and flush sample water through the sampling apparatus for 30 seconds
minute and to prime the pumps.
14. Turn the valve to closed to hold the prime and turn off the pumps. Reassemble the filtration
system with the UFs as shown in Figure C.A-2. Remember to use hose clamps where directed on
the diagram.
15. Begin pumping the sample water through the filtration system. Note the start filtration time on
the Field Form.
16. Stop filtration when 90 L (the entire volume in the tank) has been filtered.
17. Record the end filtration time on the Field Form.
18. Record on the Field Form the total volume filtered through each filter and also write the
volume on the filter label.
19. Replace the original caps on the UF side port and replace other end port with another molded
cap.
20. Place both UFs in one 2-gallon resealable bag.
21. Place filters in a cooler with ice.
22. Dump the discharge water discretely onto the sand (not in the grassy area), which will flow
through the mangroves and into the lagoon.
23. Return to local lab for processing and shipping, or continue to the next sampling site.
C-19
-------
Sampling Setup—2
(for beach, WTP, and package plant sites)
Flow
Centrifugal
Bilge pump (for
reciruculating water)
Figure C.A-2. Sampling setup at the beach, POTW, and package plant sites (there will be two identical setups, because two filters will be
used).
C-20
-------
Post-Sampling Procedures
This procedure applies to samples collected from all four sites.
1. Take grab samples to Laboratory B for processing each sampling day.
2. Transfer UFs from coolers to dedicated refrigerator at Laboratory C each sampling day.
3. At Laboratory C, hose down inside and outside of trash cans and set out to dry at the end of each
sampling day.
4. At Laboratory C, clean up chlorine kits each sampling day.
5. Dedicated Centrifugal Pumps—pumps are dedicated for one field site and are to be cleaned in the
field after each sampling.
a. Keep the tubing from sampling attached to pump for this procedure.
b. Place inlet and outlet tubing in the bleach water carboy; recirculate bleach solution for
5 minutes using the pump.
c. Drain water from the pump and lines back into the carboy.
d. Place inlet and outlet tubing in the sodium thiosulfate carboy; recirculate sodium thiosulfate
solution for 5 minutes using the pump.
e. Drain water from the pump and lines onto the ground.
f. Discard tubing.
g. Place the clean pumps in a new garbage bag.
h. Tie up the bag and label with Site #, date and Clean.
6. Carboys (there are three)—for beach sampling; clean back at Laboratory B at end of each sampling
day.
a. Hose down outside of the carboys.
b. Triple rinse inside with mild soap, rinse with tap water until all suds are gone.
c. Pour 100 mL bleach into each carboy, fill with tap water to very top, let soak at least
30 minutes or overnight.
d. Dump the bleach water down the drain, pouring water over carboy lids to disinfect those as
well.
e. Triple rinse with sodium thiosulfate solution from pump-cleaning carboys (250 mL of stock
two percent sodium thiosulfate solution + 9,750 mL reverse osmosis [RO] water), rinse the
carboy lids as well.
f. Make sure bleach odor is completely gone from the carboys and lids.
C-21
-------
g. Triple rinse each carboy with RO water.
h. Replace lids and set carboys aside for next beach sampling, label carboys as Clean and with
the date.
7. Dedicated Bilge Pumps—pumps are dedicated for one field site and are to be cleaned back at
Laboratory B at the end of each sampling day.
a. Place pumps and most of wire in a 5-gallon bucket. Tape the end of the wire to the side of the
bucket so that battery connections do not corrode.
b. Pour in 1% bleach water to cover pumps and wires. Make sure water goes down inside pumps.
c. Soak for five minutes in bleach water.
d. Remove the pumps and shake out all excess bleach water.
e. Transfer the pumps and wires similarly into another 5-gallon bucket of sodium thiosulfate
solution from the pump-cleaning carboys (250 mL of stock 2% sodium thiosulfate
solution + 9,750 mL RO water).
f. Soak for five minutes in the sodium thiosulfate solution. Make sure the solution goes down
inside pumps.
g. Remove the pumps and shake out excess solution.
h. Pour RO water over the pumps for final rinse.
i. Place the clean pumps in resealable bags and label as Clean and with Site # and Date.
8. Other important tasks
a. Sundays
1. Check the charge on the batteries and begin recharging, if needed.
2. Take inventory of supplies and e-mail Trisha Johnson ordering requests.
b. Monday (or throughout week)
1. Prepare eight sodium thiosulfate bottles for the next weekend sampling (38 g in enough RO
water to dissolve, made in autoclaved plastic bottles).
c. Friday
1. Check chlorine level in the bleach carboy using the chlorine test kit.
2. Prepare the fresh sodium thiosulfate carboy.
3. Pack for weekend sampling.
C-22
-------
Annex B: Quick Reference—Filtration of QPCR Samples at Laboratory B for Boqueron
Study Summer 2009
Materials:
1. Dedicated workstation for water filtrations.
2. Water filtration apparatus, including disposable filter funnels and bases for 47-mm filter;
vacuum manifold and trap with appropriate tubing connected to vacuum source (line, aspirator
or electric pump).
3. Autoclaved extraction tubes with glass beads and tube rack.
4. Polycarbonate filters, 47-mm, 0.4-|am pore size.
5. PCR-grade water and autoclaved38 squeeze bottles.
6. Pipettor and sterile disposable pipettes.
7. Two membrane filter forceps.
8. Ethanol, 95%.
9. Flame source.
10. Permanent marking pen for labeling tubes.
11. Bench sheets/lab notebook and black pen.
12. Styrofoam tube box for storage of extraction tubes in a freezer.
13. Freezer at <-20 °C.39
Procedures:
1. Remove the funnel, place a polycarbonate filter on the filter base and replace the funnel.
2. Shake the water sample bottle vigorously to suspend the bacteria uniformly. For lagoon
samples, filter 100 mL. For wastewater samples, filter at least 50 mL and up to 100 mL if
possible. Filter each sample in triplicate (3 filters).
3. Thoroughly rinse the sides of the funnel twice with PCR-grade water using an autoclaved
squeeze bottle.
4. Turn off the vacuum and remove the funnel from the filter base.
5. Using sterile forceps, fold the filter on the filter base into a cylinder with the sample side facing
inward and insert the filter into an extraction tube with glass beads. (Note: Handle the filter
38 Standard times for autoclaving will be used, as written in EPA's methods.
39 The EPA qPCR SOP says filters can be stored at -20 °C or -70 °C. Laboratory B and Laboratory C can only store at -20 °C.
C-23
-------
with the forceps at the edges; do not touch the portion of the filter exposed to the water
sample. Try not to allow the filter to contact the outside edges of the extraction tube.)
6. Cap the extraction tube tightly and label it.
7. Replace used filter funnels with new sterile funnels. Repeat Steps 1-6 until all water samples
have been processed.
8. For filter blanks, filter 50 mL of PCR-grade water using one new sterile filter funnel per filter
blank.
9. Samples must be filtered within six hours and placed in the -20 °C freezer within eight hours of
sample collection time.
10. Perform one filter blank for every six sample filters processed.
Label each tube with the Sample ID (e.g., 7JUN09-1A), and label field blanks as date-FB (e.g., 7JUN09-
FB); if you do more than one filter blank in a day, label as date-FBI, -FB2, etc.
C-24
-------
Annex C: Field Form
Field Form for EPA QMRA Study—Boqueron Summer 2009
Date: Members of Field Crew:
Weather Conditions:
Sampling Information:
Asahi Kasei
Rexeed
Dialyzer Type
(circle one)
Sampling
Site1
Filter Field ID2
Start
Filtration
Time
End
Filtration
Time
Volume
Filtered
CD
Total Chlorine
Residual before
Dechlorinaton
(mg/L)
Total Chlorine
Residual after
Dechlorination
(mg/L)3
Grab
Sample
Collection
Time
Carboy Sample
Collection Time at each
Beach Transect (T3,
T2, Tl)
25 S or 25 SX
25S or 25SX
Comments/Observations:
25S or 25SX
25 S or 25 SX
Comments/Observations:
'Sites: 1 (Prassa WWTP) 2 (Boqueron Beach) 3 (Package Plant #2) 4 (Lagoon)
2For Filter ID, use Date (e.g. 06 JUN09), Hyphen (-), Site # (1-4). and letter "A" for Saturday event or letter "B" for Sunday event.
3Each tank of 100 L wastewater effluent is deehlorinated with 38 g sodium thiosulfate.
C-25
-------
Annex D: Sample Chain of Custody Form
Clancy Environmental Consultants, Inc.
CHAIN OF CUSTODY RECORD
PROJECT:
EPA QMRA Study—Boqueron Summer 2009
Destination
Facility (shipped from):
CECIA-IAUPR
CEC
Courier:
20 Mapleville Depot
Airbill#
St. Albans, VT 05478
Sample
Collection
Point ID
Sample ID
(Filter Field ID)
Sample Date
Sample
Collection
Start Time
Sample
Collection
End Time
Volume
Filtered
(L)
Sample Type
Temperature (°C)
of Filter upon
Receipt in Lab
(for CEC use)
LAB ID
(for CEC use)
Remarks
1
1
2
2
3
3
4
4
1
1
2
2
3
3
4
4
Dispatched by: (Signature)
Date
Time
Received by: (Signature)
Date
Time
C-26
-------
Annex E: Instructions for Preparing the Backflush Solution
The solution will be used for backflushing UFs collected for the Puerto Rico Project. Sixteen 500-mL
bottles of solution need to be prepared every Monday following the steps below. The backflush solution
consists of 0.5% Tween 80, 0.01% sodium polyphosphate, and 0.001% antifoam Y-30 emulsion, filter
sterilized.
1. Prepare a Stock Solution weekly by adding the following ingredients to 440 mL of distilled
water:
a. 0.8 g sodium polyphosphate
b. 0.08 mL (80 |aL) antifoam Y-30 emulsion
c. 40 mL Tween 80
Place the stock solution on a stir plate with a stir bar. The solution will take approximately
25 minutes to completely dissolve.
2. Add 30 mL stock solution to 470 mL distilled water in the funnel of a 0.2 |am disposable
sterilizing filter funnel. Use the pipette to thoroughly mix.
3. Attach the funnel to vacuum and filter sterilize.
4. Repeat Steps 1-3 another 15 times for a total of sixteen 500 mL bottles of Backflush Solution,
using a fresh sterilizing filter funnel each time.
5. Label each bottle with Backflush Solution, Date, and the preparer's initials.
6. Refrigerate the bottles.
Additional Prep
(Items needed by Tuesday morning of each week)
1. Wash and autoclave eight squat-form 500 mL plastic graduated cylinders.
2. Wash and autoclave eight 1 L glass or plastic bottles.
C-27
-------
Instructions for Ultrafilter Backflush
*Sixteen UFs collected for the Tetra Tech Puerto Rico Project will arrive at Laboratory A every Tuesday
morning. Sixteen 500 mL bottles of backflush solution will be prepared before the arrival of the filters.
UFs will be backflushed according to the procedure described below.
1. UFs from the Beach and Lagoon sites will be backflushed at the bench; UFs from the POTW and
Package Plant sites will be backflushed in the Biological Safety Cabinet (BSC).
2. Wear gloves.
3. Waste generated from the Beach/Lagoon UF processing can go in the regular trash. Waste
generated from the POTW/Package Plant UF processing must go in a biohazard bag to be
autoclaved and then discarded.
4. Lay down an underpad and set up a backflush station with a peristaltic pump and a ringstand as
shown in Figure C.E-1.
5. Label a sterile 1 L glass or plastic bottle with the Lab ID, Date, and Field ID. Leave a space to
record volume.
6. Remove two bottles of backflush solution from fridge and take to bench/BSC.
7. Remove a bag of two UFs (having the same Field ID and Lab ID) from the cooler/fridge and take
to the bench/BSC.
8. Record the Field ID and Lab ID on the benchsheet.
9. Remove one filter from the bag and open both side ports to dump excess water trapped in the
filter housing into a waste beaker. Replace the side port caps.
10. Place the UF in the ringstand, remove the bottom port cap (blue end of UF), and position the UF
over the 1 L sterile glass bottle as shown in Figure C.E-1.
11. Remove the bottom side port cap from the UF.
12. Cut fresh L/S 36 tubing to the needed length (using a scalpel or tubing cutter) and attach to side
port of UF as shown in Figure C.E-1.
13. Be careful to keep the end of the tubing that will go into backflush solution clean; try not to lay
it down on the work surface.
Materials:
Underpads
Gloves
Ringstand and clamp
Peristaltic pump
L/S 36 tubing
16 bottles sterile Backflush Solution
8 sterile 500 mL grad cylinders
8 sterile 1 L glass or plastic bottles
Benchsheet
Black marker and black pen
Scalpel or tubing cutter
Labeling tape
Trash can
Large biohazard bag
C-28
-------
14. Load the tubing into peristaltic pump head.
15. Place the other end of tubing into a fresh bottle of backflush solution.
16. Check the direction of the pump flow and turn the pump on at a rate of 650 mL/min to push the
backflush solution into the UF. Water should begin to drain from the UF into the glass bottle.
17. On the benchsheet, note the time backflushing begins.
18. Tilt the bottle of the backflush solution as it empties to drain as much solution from the bottle
as possible.
19. When the flow from the UF slows from a steady stream to a drip, immediately turn off the
pump.
20. Reverse the pump flow direction and turn on the pump for 10 seconds. This will release the
pressure that has built up in the UF.
21. Remove the UF from the setup, cap all ports, and discard the UF.
22. Using the same 1 L glass bottle and the same tubing, repeat Steps 9 through 21 for the other UF
in the bag (which has the same Sample ID).
23. Using the sterile graduated cylinder, measure the total volume of UF concentrate combined
from the two filters. (The easiest way to do this is to measure 500 mL of the UF concentrate,
pour it into one of the empty backflush solution bottles, measure another 500 mL and pour it
into the other empty backflush solution bottle, and measure the last small volume in the grad
cylinder. Then pour the entire amount back into the 1 L bottle).
24. Record the total volume on the 1-L bottle label and on the benchsheet.
25. Place the UF concentrate in the refrigerator and continue with backflushing the remaining UFs.
26. Each pair of UFs will require a new grad cylinder, 1-L glass bottle, two bottles of backflush
solution, and new tubing.
27. Wipe down the work area with bleach water and dispose of the underpad. Turn on the
ultraviolet (UV) light in the BSC for 30 minutes.
C-29
-------
Labeled 1-L sterile bottle
L/S 36 tubing
Figure C.E-1. UF backflush setup.
C-30
-------
Annex F: Laboratory Standard Operating Procedures and Protocols
Laboratory procedures specific to this study: microbiological measurement protocols and procedures
not included herein (reference methods published under copyright) will be maintained by the
respective laboratories and made available for review on request.
• Quality Assurance/Quality Control Guidance for Laboratories Performing PCR Analyses on
Environmental Samples
Pathogens:
• Norovirus qPCR
o Dr. Schwab's Norovirus PCR Procedure and Bench Sheet
• Adenovirus Culture Method
o Laboratory D's Virus Cell Culture Manual
o Laboratory D's Cell Culture and Assay for Adenovirus with A549 Cell Line
o Laboratory D's Quantitation of Total Culturable Virus Data Sheets
• Adenovirus qPCR Method
o Dr. Schwab's Adenovirus PCR Procedure and Bench Sheet
• Enterovirus Culture Method Most-Probable Number Method (Supernatant Used for Enterovirus
PCR Method
o USEPA Manual of Methods for Virology. Chapter 14: Concentration and Processing of
Waterborne Viruses by Positive Charge 1MDS Cartridge Filters and Organic Flocculation
o USEPA Manual of Methods for Virology. Chapter 15: Total Culturable Virus Quantal
Assay
o Laboratory D's Quantitation of Total Culturable Virus Data Sheets
• Enterovirus Using RT-PCR Method for Enterovirus PCR
o Dr. Schwab's Enterovirus PCR Procedure and Bench Sheet
• Cryptosporidium and Giardia using EPA Method 1623 for Cryptosporidium and Giardia in Water
by Filtration/IMS/FA
o EPA Method 1623: Cryptosporidium and Giardia in Water by Filtration/IMS/FA
o EPA Method 1622/23 Benchsheet and Text
o EPA Method 1623 Hemacytometer Data Sheet for Cryptosporidium or Giardia and Text
o EPA Method 1622/1623 Spiking Suspension Enumeration Form: Cryptosporidium or
Giardia and Text
C-31
-------
o EPA's Method 1622/1623 Cryptosporidium Report Form and Text
o EPA's Method 1622/1623 Giardia Report Form and Text
• Infectious Cryptosporidium Culture/Research Method
o EPA's ICR Microbial Laboratory Manual, Section VII
o AwwaRF 3021 Detection of Infectious Cryptosporidium in Filtered Drinking Water SOP
(Giovanni, 2006)
o EPA's ICR Microbial Laboratory Manual, Appendix VII-4, Cryptosporidium Report Form
• Toxigenic E. coli 0157:H7 and Non-0157 Shiga-Toxin Strains Culture/Research Method
o Laboratory A's Toxigenic E. coli 0157:H7 and Non-0157 Shiga-Toxin Protocol and
Benchsheet
• Campylobacter Culture/Research Method
o Laboratory A's Campylobacter Protocol and Benchsheet
• Salmonella spp. Culture/Research Method and qPCR Research Method
o Laboratory A's Salmonella Culture Protocol and Benchsheet
• Toxigenic E. coli 0157:H7 and Non-0157 Shiga-Toxin Strains and Salmonella spp. qPCR Research
Methods
o E. coli 0157:H7 and Non-0157 Shiga-Toxin Strains and Salmonella spp. qPCR Protocols
and Benchsheets (including sample processing protocol, DNA extraction, and PCR run
protocols
¦ Laboratory A's Preliminary Information on Quantification of Bacterial Pathogens
Using qPCR
¦ Laboratory A's qPCR Centrifugation Log
¦ Laboratory A's qPCR Filtering Log
¦ EZ1® DNA Investigator Handbook (QIAGEN, 2009)
¦ Rotor-Gene® Q—Pure Detection (QIAGEN, 2009)
Indicators:
• Clostridium perfringens
o EPA's ICR Microbial Laboratory Manual, Section XI
o Laboratory A's SOP for Detecting Clostridium perfringens in Water and Wastewater
Using the Membrane Filter Method
o Laboratory A's Clostridium Laboratory Benchsheet
C-32
-------
Male-specific (F+) Coliphage
o EPA Method 1602: Male-specific (F+) and Somatic Coliphage in Water by Single Agar
Layer Procedure
o Laboratory A's Standard Operating Procedure for Production and Enumeration of Male
Specific Coliphages
o Laboratory A's Standard Operating Procedure for Male Specific (F+) Coliphages in
Water—Single Agar Layer
o Laboratory A's Initial Precision and Recovery Trials for Male-Specific Coliphage Using
U.S. EPA Method 1602
o Laboratory A's Method 1602—Male Specific Coliphage in Water
E. coli SM9211F
o Laboratory B's E. coli SM9211F Protocol
Enterococcus EPA Method 1600 (Membrane-Enterococcus lndoxyl-|3-D-Glucoside Agar
Membrane Filter Method for Detection of Enterococci)
o Method 1600: Enterococci in Water by Membrane Filtration Using Membrane-
Enterococcus lndoxyl-|3-D-Glucoside Agar
o Laboratory B's Method 1600 Protocol
Enterococcus & Bacteroides EPA Duplex qPCR Research Method
o EPA's Enterococcus spp. and Bacteroidales Method described in Laboratory E's Rapid,
PCR Based Method for Measuring Total Enterococcus spp. and Bacteriodales in Water
Samples
o Laboratory F's Example Setup Page
o Laboratory F's Example Run Report
C-33
-------
Appendix D: Water Quality Standards Submission Checklist
This appendix provides a checklist of additional information that is useful to include in the WQS
submission that includes alternative WQC. Following this list will help the EPA determine if the WQS
submission is based on sound science and protective of the designated use, as required in 40 CFR §
131.6 and 40 CFR § 131.11. Users provide substantiation for the identified decision points at each step.
This list complements the information in the EPA's Water Quality Standards Handbook chapter on
"Procedures for Review and Revision of Water Quality Standards," which describes the process for
submitting and the EPA review of WQS.40 A reevaluation of the sanitary survey once every three years
(e.g., at the time of a state or Tribal triennial review) should be sufficient to confirm that the sources
do not substantially differ from those characterized in Steps 1 and 2.
Information Provided bytheTSM User
Step 1: Understanding Sources Impacting the Watershed
Decision Point: Is there clear evidence that nonhuman fecal sources impact and possibly
dominate the waterbody?
QMRA sanitary survey results41
Existing and historical management practices
Past monitoring data
Existing information on fecal source identification and tracking
Survey of wildlife, agricultural, and domestic animals
Future planning and zoning information (if available)
Maps
Hydrological data
Meteorological data
Step 2: Conduct Water Quality Study
Decision Point: Are nonhuman fecal contamination sources confirmed to be the predominant
fecal source affecting the waterbody?
Sampling and analysis plan for conducting the water quality study
Select reference pathogens based on source
FIB densities (geometric mean [GM] of enterococci in marine water and enterococci or E. coli
in freshwater) in the waterbody and presumed fecal source
Reference pathogen densities in the waterbody and presumed fecal source
Fecal source identification
Other studies that provide information on sources
Step 3: Human Health Risk Evaluation
Decision Points:
Identify water quality value and associated target illness rate to adjust for nonhuman
fecal sources.
40https://www. epa.gov/wqs-tech/water-qualitv-standards-handbook
41A Sanitary Survey form developed for recreational water locations for which QMRAs will be conducted can be found in
TSM Appendix B. The EPA has also provided a mobile application for conducting beach sanitary surveys:
https://www.epa.gOv/beach-tech/sanitarv-survevs-recreational-waters#epa
D-l
-------
How do the water quality and illness estimates compare to the water quality and illness
targets of the applicable water quality standard?
Does this information support deriving alternative water quality criteria for the
waterbody?
FIB and reference pathogen data collected at the site and in sources that impact the site (from
Step 2)
Limits of quantification for the pathogen methods
Sources of variability and uncertainty
Additional site-specific assumptions
Site-specific factors and data not included and the justification for not including those factors
(e.g., temporal shifts in loading)
Step 4: Calculate alternative WQC GM that corresponds with the selected benchmark illness level
Decision Points:
Are the decisions and risk assessment assumptions transparently documented?
Determine the magnitude, duration, and frequency elements to be included in the
criteria.
Are the proposed criteria scientifically defensible and protective of the designated use?
Transparent calculation of the GM that corresponds with the target illness rate.
Transparent documentation and justification of any parameters used in the reverse QMRA by
either pointing to the TSM or, if using values not discussed in the TSM, pointing to information
supporting those parameters.
Calculate the statistical threshold value (STV) and beach action value (BAV) corresponding to
the GM and include the duration component of the alternative criteria and transparent
documentation of calculations.
D-2
-------
Appendix E: Example Code for Step 3: Approach 1
This appendix provides an annotated code example for Step 3, Approach 1 forward QMRA. The code
estimates the mean and median illness levels for ambient recreational waters for which measured
pathogen densities are available. The code may be adapted for other pathogens by defining and
appropriately including parameters. It is recommended that this code be utilized by experienced
programmers after a careful reading of its syntax and structure with reference to the appropriate
equations described in the TSM. This is an example and should not be considered as a tool. A key for
the annotated code is presented in Table E-l and the definitions and sources for variables for the
Step 3, Approach 1 forward QMRA is presented in Table E-2.
Table E-l. Key for annotated code.
KEY
BOLD^Actual R code
(R available at http://www.r-proiect.org/)
PLAIN—Comments for guidance
Table E-2. Step 3, Approach 1 forward QMRA: definitions and sources for variables in equations.
In-text
variable
Code Variable
X represents the pathogen of interest
Definition
fM
L.rp
testdata$_X
Note: _X refers to different pathogen names
Measured density of reference pathogen in waterbody
(pathogens/L).
V
rand_vol
Volume of water ingested (L)
Mrp
Estimated dose of reference pathogen ingested,
computed directly as product of
CrP and V (number of pathogens)
prP
ill
piN_x
Probability of illness from a specific reference
pathogen (unitless)
Pill
Tpill
Estimated probability of illness (unitless)
frp
lil—r
pill_bpa, pill_exp, pill_hyp, pill_hyp_cj
Dose-response functions
n>
Pill\inf
drp_X$c_pill
Note: _X refers to different pathogen names
Probability of illness following infection (unitless)
#!/usr/bin/Rscript
#############################################################################################
### Example code for QMRA Approach 1 based on measured pathogen densities
### Example relies on an R dataframe named "testdata" which contains data on simultaneously ###### measured
pathogen densities in separate columns in units of pathogens/L
### The "testdata" dataset is not provided. The dataset name and column names will differ in the user's dataset.
#############################################################################################
rm(list=ls()); # Removes all previous objects from the R environment
set.seed(l); ## Ensures numerical replicability of simulation
E-l
-------
##############################################################################
### The code below checks whether or not required libraries "hypergeo" have been installed previously.
### If not, the code downloads and installs the required packages first.
##############################################################################
list.of.packages <- c("hypergeo","MASS")
new.packages <- list.of.packages[!(list.of.packages %in% installed.packages()[,"Package"])]
if(length(new.packages)) install.packages!new.packages)
require!" hypergeo");
require("MASS");
#######################################
### Dose Response Functions
#######################################
###################################################################
### The following lines define generic dose-response functions
### which may be applied to one or more pathogens when supplied
### with the appropriate parameters. These functions implement
### the formulas provided in Table 2-5.
###################################################################
###########################################
### Beta Poisson Dose Response Function
###########################################
pill_bpa <- function(a, b, c_pill, d){
return((l-(l+d/b)A-a)*c_pill)
};
##########################################
### Exponential Dose Response Function
##########################################
pill_exp <- function(r, c_pill, d){
return((l-exp(-r*d))*c_pill)
};
##############################################
### Confluent Hypergeometric Expression
### of Beta Poisson Dose Response Function
##############################################
pill_hyp <- function(alpha, beta, c_pill, d){
return((l-genhypergeo(alpha, alpha+beta, -d))*c_pill)
};
##############################################################
### Confluent Hypergeometric Expression
### of Beta Poisson Dose Response Function
### Implements Teunis 2018 / 2020 Bayesian parameter method
##############################################################
### Function sample w_z jointly samples w and z (transformed parameters) from posterior distribution expressed as ###
bivariate normal
sample_w_z<-function(means, vcov,n){
return(mvrnorm(n, mu = means, Sigma = vcov)) ### inputs are vector of means and the variance covariance matrix
}
## Function hyplF_t20 computes probability of infection based on w, z parameter distribution using the confluent
## hypergeometric function
hyplFl_t20<-function(inf_wz_dist,d){
w<-inf_wz_dist[,l]; z<-inf_wz_dist[,2] ### Inputs are joint w,z distribution and dose
u<-exp(w)/(l+exp(w)); v<-exp(z)
alpha<-v*u; beta<-v*(l-u)
pill<-c()
for (i in l:n_samp){
E-2
-------
p<-l-genhypergeo(alpha[i],alpha[i]+beta[i],-d)
p<-pmin(l,p)
p<-pmax(0,p)
pill<-c(pill,p)
}#end of for
return(pill)}
### Function cpill_t20 computes conditional probability of illness based on w, z parameter distribution
cpill_t20<-function(ill_wz_dist,d){
w<-ill_wz_dist[,l]; z<-ill_wz_dist[,2] ### Inputs are joint w,z distribution and dose
u<-exp(w)/(l+exp(w)); v<-exp(z)
r<-v*u; eta<-v*(l-u)
cpill<-c()
for (i in l:n_samp){
p<-l-(l+d/eta[i])**-r[i]
p<-pmin(l,p)
p<-pmax(0,p)
cpill<-c(cpill,p)
}# end of for
return(cpill)}
### Function pill_hyp_t20 computes median probability of illness combining probability of infection and condititonal
#### probability of illness given infection
pill_hyp_t20 <- functionfin f_wz_d i st, i 11 _wz_d i st, dose){
p_inf<-hyplFl_t20(inf_wz_dist,dose) ### Inputs are joint w,z distribution for both infection and illness, and dose
c_pill<-cpill_t20(ill_wz_dist,dose)
p_ill<-c_pill*p_inf
return(median(p_ill))
};
##### Dose Response Parameters for Pathogens of Interest
###############################################################
#########################################################################
### The following lines define and save the dose response parameters
### for each of the pathogens occurring at the site in "data frames".
### These values are drawn from Table 2-5.
#########################################################################
####################################################################
### Salmonella Dose Response Parameters (for Beta Poisson DRF)
### c_pill is actually a uniformly distributed range of values
### (0.17 to 0.4). The mean value has been used.
####################################################################
drp_sal <- data.frame(a=0.3126, b=2884, c_pill=0.285);
#################################################################
### Adenovirus Dose Response Parameters (for Confluent Hypergeometric DRF
### with dose dependent conditional probability of illness)
#################################################################
drp_ade<- data.frame(alpha=5.11, beta=2.8, r=0.41, eta=6.53);
######################################################################
### Cryptosporidium Dose Response Parameters (for Exponential DRF)
### c_pill in the line below is actually a uniformly distributed
### range of values (.3 to .7). The mean value has been used.
######################################################################
drp_cryp <- data.frame(r=0.09, c_pill=0.5);
#####################################################################
### Giardia Dose Response Parameters (for Exponential DRF)
### c_pill in the line below is actually a uniformly distributed
E-3
-------
### range of values (.2 to .7). The mean value has been used.
#####################################################################
drp_gia <- data.frame(r=0.0199, c_pill=0.45);
#################################################################
### Norovirus Dose Response Parameters
### (Confluent Hypergeometric Expression of Beta Poisson DRF)
### using the Teunis dose dependent dose response function
### Both sets of parameters are specified as distributions per Teunis 2020)
#################################################################
drp_nor<- list(inf_wz_means=c(-0.608, 0.194),inf_wz_vcov=matrix(c(1.79, -1.03, -1.03, 2.54),2),ill_wz_means=c(1.74,
1.82),ill_wz_vcov=matrix(c(5.55, -0.708, -0.708, 4.64),2));
#################################################################
### Campylobacter jejuni Dose Response Parameters
### (Confluent Hypergeometric Expression of Beta Poisson DRF
### using the Teunis dose dependent dose response function
### Both sets of parameters are specified as distributions per Teunis 2018)
#################################################################
drp_camp <- list(inf_wz_means=c(-0.177,0.054),inf_wz_vcov=matrix(c(l.303,-0.041,-0.041,1.070),2),ill_wz_means=c(-
2.744,-4.89E-3),ill_wz_vcov=matrix(c(1.337, 0.010, 0.010, 0.993),2));
#######################################################################################
### The following line specifies the number of simulations that are to be performed
#######################################################################################
numsim <- 1000; ### number of simulated swim events
n_samp<-1000; ### number of samples from Bayesian posterior parameter distribution for Teunis 2018 and 2020 DRFs
#################################################################
##################################
### Defining matrix dimensions
##################################
tpill <- matrix(0, numsim); # tpill is the matrix that will contain multiple estimates of the estimated total probability of
illness from a single swim event
############################################################
### Sample from normalized parameter posterior distributions
### for DRFs based on Teunis 2018 and 2020
###(Generate w,z distributions for infection and illness)
############################################################
nor_inf_wz_dist<-sample_w_z(drp_nor$inf_wz_means, drp_nor$inf_wz_vcov,n_samp)
nor_ill_wz_dist<-sample_w_z(drp_nor$ill_wz_means, drp_nor$ill_wz_vcov,n_samp)
camp_inf_wz_dist<-sample_w_z(drp_camp$inf_wz_means, drp_camp$inf_wz_vcov,n_samp)
camp_ill_wz_dist<-sample_w_z(drp_camp$ill_wz_means, drp_camp$ill_wz_vcov,n_samp)
##################################################
### Beginning the swim event simulation loop
##################################################
for (i in l:numsim){
#### Randomly select pathogen densities and volume ingested for each swim event####
rand_row <- floor(runif(l, 1, nrow(testdata)+l)); # Randomly picks a row of data from the data file
# (this is known as bootstrap sampling with replacement)
rand_vol <- rlnorm(l, -3.98,1.43); # Randomly picks an ingested volume from a lognormal distribution
# with characteristics of In mean=-3.98 and In sd=1.43. (Dufour, 2017)
# Units adjusted to output volume in L. These parameter values are from Table 2-8.
#############################################################################
### The following lines compute the individual probability of illness
### from each pathogen in the data during the simulated single swim event
#############################################################################
### Estimate Salmonella Risk (Equation IB and 1C)
pill_sal <- pill_bpa(drp_sal$a, drp_sal$b, drp_sal$c_pill, testdata$sal_c[rand_row]*rand_vol);
E-4
-------
### Estimate Adenovirus Risk (Equation IB)
pill_ade <- pill_hyp(drp_ade$alpha, drp_ade$beta,drp_ade$r, drp_ade$eta,
testdata$aden_c[rand_row]*rand_vol);
### Estimate Cryptosporidium Risk (Equation IB and 1C)
pill_cryp <- pill_exp(drp_cryp$r, drp_cryp$c_pill, testdata$crypl623[rand_row]*rand_vol);
### Estimate Giardia Risk (Equation IB)
pill_gia <- pill_exp(drp_gia$r, drp_gia$c_pill, testdata$giarl623[rand_row]*rand_vol);
### Estimate Norovirus Risk (Equation IB)
pill_nor <- pill_hyp_t20(nor_inf_wz_dist,nor_ill_wz_dist, testdata$tcv[rand_row]*rand_vol);
### Estimate Campylobacter Risk (Equation IB)
pill_camp <- pill_hyp_t20(camp_inf_wz_dist,camp_ill_wz_dist, testdata$camp_c[rand_row]*rand_vol);
### The following line computes illnesses per 1000 events from all pathogens in the data during the simulated
single swim event (Equation ID)
tpill[i] <- 1000*(l-(l-pill_sal)*(l-pill_ade)*(l-pill_cryp)*(l-pill_gia)*(l-pill_nor)*(l-pill_camp)); #
Estimate and Save Total Risk from all pathogens
}# End the simulation loop
### Output summary statistics relating to the total probability of illness from swimming ###
print(paste("The estimated mean illnesses per 1000 events using Approach 1 is", round(mean(tpill, na.rm=T), digits=2)));
### Output summary statistics relating to the total probability of illness from swimming ###
print(paste("The estimated median illnesses per 1000 events using Approach 1 is", round(median(tpill, na.rm=T),
digits=2)));
### Output summary statistics relating to the total probability of illness from swimming ###
printf'Summary statistics of the estimated illnesses per 1000 events using Approach 1 are:")
summary(tpill, na.rm=T)
df<-data.frame(unclass(summary(tpill, na.rm=T)), check.names = FALSE, stringsAsFactors = FALSE)
write.csv(df,"Yourpath/YourFilename.csv", row.names = FALSE)
E-5
-------
Appendix F: Example Code for Step 3: Approach 2
This appendix provides an annotated code example for Step 3, Approach 2 forward QMRA. The code
estimates the mean and median illness levels for ambient recreational waters that are contaminated
by seagull sources. The code may be adapted for other animal sources by defining and appropriately
including parameters for those sources. It is recommended that this code by utilized by experienced
programmers after a careful reading of its syntax and structure with reference to the appropriate
equations described in the TSM. This is an example and should not be considered a tool. In this
example code all the references to FIB refer to enterococci, however, it can be adapted to other FIB,
such as E. coli. A key for the annotated code is presented in Table F-l and the definitions and sources
for variables for the Step 3, Approach 2 forward QMRA is presented in Table F-2.
Table F-l. Key for annotated code.
KEY
BOLD^Actual R code
(R available at http://www.r-proiect.org/)
PLAIN—Comments for guidance
Table F-2. Step 3, Approach 2 forward QMRA: definitions and sources for variables in equations.
In-text
variable
Code variable
X represents the pathogen of interest
Definition
Cs
°rp
Crp_X
Note: _X refers to different pathogen names
Estimated density of reference pathogen in the
waterbody derived from the Source S (pathogens/L)
Cfib
testdata$ent
Measured waterbody density of enterococci using a
culture method (CFU/L)
D S
kfib
Rfib
Density of enterococci (CFU/L or CFU/g) in the source S
D S
n.rp
Rrp_X
Note: _X refers to different pathogen names
Density of reference pathogens (number of pathogens
or genomes/L or number of pathogens or genomes/g)
in the Source S
[Xrp
dose_x
Note: _X refers to different pathogen names
Estimated dose ingested of reference pathogen derived
from the Source S (number of pathogens)
Prp
lnf_x$prp
Note: _X refers to different pathogen names
Fraction of human infectious pathogenic strains in the
Source S (unitless)
Irp
lnf_x$irp
Note: _X refers to different pathogen names
Prevalence of infection in the nonhuman Source S
(unitless)
V
rand_vol
Volume of water ingested (L)
pS
rill
tpill
Estimated probability of illness from a specific source of
contamination (unitless)
pRP
rill
drp_X$c_pill
Note: _X refers to different pathogen names
Probability of illness from a specific reference pathogen
(unitless)
F-l
-------
#!/usr/bin/Rscript
#############################################################################################
### Example code for QMRA approach based on measured FIB density assuming seagulls as
### the source of fecal contamination. Example relies on an R dataframe named "testdata" which
### contains data on measured FIB density in units of CFU/L The "testdata" dataset is not provided. The dataset name and
column names will differ in the user's dataset.
#############################################################################################
rm(list=ls()); # Removes all previous objects from the R environment
set.seed(l); ## Ensures numerical replicability of simulation
##############################################################################
### The code below checks whether or not required libraries have been installed previously.
### If not, the code downloads and installs the required packages first.
##############################################################################
list.of.packages <- c("hypergeo","MASS")
new.packages <- list.of.packages[!(list.of.packages %in% installed.packages()[,"Package"])]
if(length(new.packages)) install.packages!new.packages)
require("hypergeo");
require("MASS");
#######################################
### Dose Response Functions
#######################################
###################################################################
### The following lines define generic dose-response functions
### which may be applied to one or more pathogens when supplied
### with the appropriate parameters. These functions implement
### the formulas provided in Table 2-5.
###################################################################
###########################################
### Beta Poisson Dose Response Function
###########################################
pill_bpa <- function(a, b, c_pill, d){
return((l-(l+d/b)A-a)*c_pill)
};
##############################################################
### Confluent Hypergeometric Expression
### of Beta Poisson Dose Response Function
### Implements Teunis 2018 / 2020 Bayesian parameter method
##############################################################
### Function sample w_z jointly samples w and z (transformed parameters) from posterior distribution expressed as ###
bivariate normal
sample_w_z<-function(means, vcov,n){
return(mvrnorm(n, mu = means, Sigma = vcov)) ### inputs are vector of means and the variance covariance matrix
## Function hyplF_t20 computes probability of infection based on w, z parameter distribution using the confluent
## hypergeometric function
hyplFl_t20<-function(inf_wz_dist,d){
w<-inf_wz_dist[,l]; z<-inf_wz_dist[,2] ### Inputs are joint w,z distribution and dose
u<-exp(w)/(l+exp(w)); v<-exp(z)
alpha<-v*u; beta<-v*(l-u)
F-2
-------
pill<-c()
for (i in l:n_samp){
p<-l-genhypergeo(alpha[i],alpha[i]+beta[i],-d)
p<-pmin(l,p)
p<-pmax(0,p)
pill<-c(pill,p)
} # end of for
retu rn(pill)}
### Function cpill_t20 computes conditional probability of illness based on w, z parameter distribution
cpill_t20<-function(ill_wz_dist,d){
w<-ill_wz_dist[,l]; z<-ill_wz_dist[,2] ### Inputs are joint w,z distribution and dose
u<-exp(w)/(l+exp(w)); v<-exp(z)
r<-v*u; eta<-v*(l-u)
cpill<-c()
for (i in l:n_samp){
p<-l-(l+d/eta[i])**-r[i]
p<-pmin(l,p)
p<-pmax(0,p)
cpill<-c(cpill,p)
}# end of for
return(cpill)}
### Function pill_hyp_t20 computes median probability of illness combining probability of infection and conditional
probability of illness given infection
pill_hyp_t20 <- f u net io n (i nf_wz_dist,iIl_wz_dist, dose){
p_inf<-hyplFl_t20(inf_wz_dist,dose) ### Inputs are joint w,z distribution for both infection and illness, and dose
c_pill<-cpill_t20(ill_wz_dist,dose)
p_ill<-c_pill*p_inf
return(median(p_ill))
};
###############################################################
##### Dose Response Parameters for Pathogens of Interest
###############################################################
###############################################################################
### The following lines define and save the dose response parameters
### for each of the pathogens occurring in the source (seagulls)
### in "data frames". These values are drawn from Table 2-5.
###############################################################################
####################################################################
### Salmonella Dose Response Parameters (for Beta Poisson DRF)
### c_pill is actually a uniformly distributed range of values
### (0.17 to 0.4). The mean value has been used.
####################################################################
drp_sal <- data.frame(a=0.3126, b=2884, c_pill=.285);
#################################################################
### Campylobacter jejuni Dose Response Parameters
### (Confluent Hypergeometric Expression of Beta Poisson DRF
### using the Teunis dose dependent dose response function)
F-3
-------
#################################################################
drp_camp <- list(inf_wz_means=c(-0.177,0.054),inf_wz_vcov=matrix(c(l.303,-0.041,-0.041,1.070),2),ill_wz_means=c(-
2.744,-4.89E-3),ill_wz_vcov=matrix(c(1.337, 0.010, 0.010, 0.993),2));
#################################################################################
### The following lines define and save the infectivity parameters for
### each of the pathogens occurring in the source (seagulls) in "data frames"
### The parameters irp and prp are from Table 2-4 and Table 2-3 respectively.
#################################################################################
inf_sal <- data.frame(prp=(75+100)/2/100, irp=0.165); # Defining prp (fraction of human infectious pathogenic strains from
###source) and irp (prevalence of infection in source) for Salmonella
inf_cam <- data.frame(prp=(54+100)/2/100, irp=0.165); # Defining prp and irp for Campylobacter
#######################################################################################
### The following line specifies the number of simulations that are to be performed
#######################################################################################
numsim <-1000 # Number of simulations
n_samp<-1000; ### number of samples from Bayesian posterior parameter distribution for Teunis 2018 and 2018 DRFs
### Defining matrix dimensions ###
tpill <- matrix(0, numsim) # tpill is the matrix that will contain multiple estimates of the estimated total probability of
### illness from a single swim event
############################################################
### Sample from normalized parameter posterior distributions
### for DRFs based on Teunis 2018 and 2020
###(Generate w,z distributions for infection and illness)
############################################################
camp_inf_wz_dist<-sample_w_z(drp_camp$inf_wz_means, drp_camp$inf_wz_vcov,n_samp)
camp_ill_wz_dist<-sample_w_z(drp_camp$ill_wz_means, drp_camp$ill_wz_vcov,n_samp)
##################################################
#### Beginning the swim event simulation loop
##################################################
for (i in l:numsim){
### The following lines randomize (for each swim event) the FIB density, Volume ingested, FIB density in source,
# and Pathogen density in source ###
rand_row <- floor(runif(l, 1, nrow(testdata)+l)); # Randomly selects a row of data from the data file (this
#is known as bootstrap sampling with replacement)
rand_vol <- rlnorm(l, -3.98,1.43); # Randomly picks an ingested volume from a lognormal distribution
# with characteristics of In mean=-3.98 and In sd=1.43. (Dufour,
#2017)
# Units adjusted to output volume in L. These parameter values are from Table 2-8.
### The pathogen and FIB density parameters used in this block are from Table 2-2 and 2-7 ###
Rfib <- 10Arunif(l, 4.114, 9.447); # Randomly simulates the density of fecal indicator enterococci in source
# in CFU/g. If you are using E. coli, change the values of this parameter.
Rrp_sal <- 10Arunif(l, 2.3, 9.0); # Randomly simulates the density of Salmonella in source
#organisms/g
Rrp_cam <- 10Arunif(l, 3.3, 6.0); # Randomly simulates the density of Campylobacter in source
#organisms/g
F-4
-------
### The following lines estimate the dose ingested during the swim event ###
Crp_sal <- (testdata$ent[rand_row]/Rfib)*Rrp_sal; # EQUATION 2A: Calculates the density of
### Salmonella in water in pathogens/L
dose_sal <- Crp_sal*inf_sal$prp*inf_sal$irp*rand_vol; # EQUATION 2B: Dose calculation for Salmonella.
Crp_cam <- (testdata$ent[rand_row]/Rfib)*Rrp_cam; # EQUATION 2A: Calculates the
### density of Campylobacteria in water in pathogens/L
dose_cam <- Crp_cam*inf_cam$prp*inf_cam$irp*rand_vol; # EQUATION 2B: Dose calculation for
### Campylobacteria
### The following lines compute the individual probability of illness from each pathogen in the source during the
### simulated single swim event
pill_sal <- pill_bpa(drp_sal$a, drp_sal$b, drp_sal$c_pill, dose_sal);
# EQUATION 2C: Estimate Salmonella Risk
pill_cam <- pill_hyp_t20(camp_inf_wz_dist,camp_ill_wz_dist, dose_cam);
# EQUATION 2C: Estimates Campylobacter Risk
### The following line computes illnesses per 1000 events from all pathogens found in the source during the
### simulated single swim event
tpill[i] <- 1000*(l-(l-pill_sal)*(l-pill_cam)) # EQUATION 2D: Estimate and Save Total Risk from all
### pathogens
}# End the simulation loop
print(paste("The estimated mean illnesses per 1000 events using Approach 2 is", round(mean(tpill, na.rm=T),
digits=2)));# Output summary statistics relating to the total probability of illness from swimming
print(paste("The estimated median illnesses per 1000 events using Approach 2 is", round(median(tpill, na.rm=T),
digits=2)));# Output summary statistics relating to the total probability of illness from swimming
### Output summary statistics relating to the total probability of illness from swimming ###
print("Summary statistics of the estimated illnesses per 1000 events using Approach 1 are:")
summary(tpill, na.rm=T)
## Output summary stats in a file
df<-data.frame(unclass(summary(tpill, na.rm=T)), check.names = FALSE, stringsAsFactors = FALSE)
write.csv(df,"YourFilePath/YourFileName.csv", row.names = FALSE)
#####################################################################################################
##########################################################################
### ADDITIONAL NOTE: If the summary outputs an error message, it is most likely because the simulated dose has
exceeded the range of R's confluent hypergeometric function. ###
### Rerunning the code usually corrects this problem. The reason this occurs is attributable to the high-end values for
pathogen densities in the source.
### The greater the numbers of simulations that are run, the more likely the extreme case will be simulated
############################################################################################
F-5
-------
Appendix G: Reverse QMRA Gull Case Study
Introduction
This appendix describes a reverse QMRA case study that implements the reverse QMRA approach
discussed in Section 2.2.2.4.5 of this TSM and corresponds to Step 4 in the process to estimate FIB
levels for waterbodies predominated by nonhuman fecal sources (Section 3.4). The EPA constructed
this illustrative example to assist users with application of this TSM. The scenarios characterized in this
case study are hypothetical waterbodies predominantly affected by sea gull feces or by a nonfecal
source of indicator (i.e., no enteric pathogen contribution). Soller et al. (2010b) presented a QMRA-
based analysis characterizing the influence of multiple sources of enterococci (human, animal and
nonfecal) on potential human health risk. Soller et al. (2014) evaluated mixtures of nonhuman and
human fecal sources and reported corresponding enterococci densities equivalent to the EPA's
recommended target illness rate. These studies concluded that RBTs for waters with predominantly
nonhuman sources are likely to be higher than for waters impacted by predominantly human sources.
Soller et al. (2010b) concluded that risks associated with exposure to recreational waters impacted by
fresh gull, chicken, or pig feces can be substantially lower than waters impacted by human sources,
when the water quality is anchored at 35 CFU per 100 mL enterococci or 126 CFU per 100 mL E. coli.
Soller et al. (2014) concluded the predicted culturable enterococci densities that correspond to the RBT
are substantially greater than the 2012 RWQC value for
mixtures of human/nonpathogenic, human/gull,
human/pig, and human/chicken inputs provided the human
contribution was relatively low (Text Box G-l). Likewise,
recreational waters containing a large fraction of culturable
enterococci from nonhuman contamination sources may
have reduced gastrointestinal illness risks to bathers
compared to waters containing the same level of
enterococci from human fecal sources.
Recent literature searches (Appendix A) have provided updated values for some QMRA parameters
compared to those reported in Soller et al. (2010b, 2014). As part of this case study, the EPA conducted
a sensitivity analysis comparing the new (base) analysis and Soller et al. (2014). See TSM Section 4.2.1
for a discussion of the sensitivity analysis.
By Step 4 (Section 3.4) of this TSM, users will have collected information to understand and confirm
the fecal sources affecting the waterbody being characterized and have conducted a "forward" QMRA
to understand the potential human health risk posed by the source loading to that waterbody. If the
waterbody is shown to pose less risk compared to the EPA's target illness rates, then a "reverse"
QMRA can be conducted to calculate a FIB density corresponding to a specific target illness rate. This
appendix provides an example of how to adjust the 2012 RWQC recommendations to reflect risks from
non-human sources, including using a "reverse" QMRA to calculate the GM. This case study example
demonstrates an approach to document the reverse QMRA portion of this TSM. Examples of
calculations for the STV and BAV are also provided. Users may find this case study example helpful for
collating and documenting the type of information needed in preparing and submitting a WQS package
including adjusted water quality criteria values to the EPA.
Text Box G-l. Risk-Based Threshold
The output from the reverse QMRA is a
FIB density that can be used as a
geometric mean (GM) corresponding to
the target illness rate and is called a Risk-
Based Threshold (RBT).
G-l
-------
Methods
Each of the parameter values and model components used in this example reverse QMRA are indicated
within tables in this section. Point estimates rather than stochastic distributions were used to describe
all model parameters to ensure straightforward analytical computations in the reverse QMRA
framework and for consistency with the form of the model output because the target illness rate was
represented as a point estimate rather than a risk distribution. Information on model inputs is provided
below.
Target Illness Rate
The case study example includes a target illness rate of 36 NGI per 1,000 recreators, which represents
the upper illness rate included in the EPA's 2012 RWQC recommendations (U.S. EPA, 2012).
Ingestion Volume
Four different ingestion volume scenarios were modeled to capture variability across age groups and
uncertainty in ingestion estimates. The ingestion parameters evaluated include the following point
estimates:
• For the general population:
o A median ingestion volume of 19 mL per swim event (Dufour et al., 2017).
o A median ingestion volume of 16 mL per swim event (DeFlorio-Barker et al., 2017).
• For the 6-10 years old age group:
o A median ingestion volume of 25 mL per swim event (Dufour et al., 2017).
o A median ingestion volume of 40 mL per swim event (DeFlorio-Barker et al., 2017).
The general population (Dufour et al., 2017) provides the base analysis parameter for ingestion (19 mL
per event), and the other three estimates are used in the sensitivity analyses. Section 2.2.2.2 describes
these studies and the justification for the base analysis. Henceforth Dufour et al. (2017) is referred to
as "Dufour" and DeFlorio-Barker et al. (2017) is referred to as "DFB."
The Python code (Appendix H) allows for one ingestion volume point estimate, so the base analysis and
the three sensitivity analyses were run as separate iterations of the model.
Reference Pathogens in Human and Gull Contamination
The following reference pathogens were modeled for each source of contamination:
1. Human contamination: norovirus; as discussed in Section 2.1.1, norovirus was reported to
account for over 70% of nonfood borne illness with a viral etiology and approximately 63% of
all nonfoodborne illness reported in the United States per year (Scallan et al., 2011a,b).
Additionally, the norovirus dose-response relationship serves as an index for NGI caused by
enteric viruses (Soller et al., 2010a).
2. Gull contamination: Campylobacter jejuni and Salmonella; as shown in Table 2-2 (in Section
2.1.1.1), these are the two pathogens that are most likely to occur in gull feces.
G-2
-------
Dose-response Functions and Parameters
Table G-l summarizes the dose-response functions and parameters used to estimate the probability of
infection for each reference pathogen. The Beta Poisson model requires two parameters to define the
probability of infection. The Beta Poisson model is typically approximated by: probability of infection
equals 1 - (1 + dose/3)"01. Both a and 3 are point estimates that provide the information that is unique
for each pathogen. The Beta Poisson model with the confluent hypergeometric function can also use
point estimates a and 3, however in this case study a distribution was used for both of these
parameters, called w and z.42 The distributions are defined by the mean, variance, and covariance of
the distribution as shown in Table G-l.
Table G-l. Dose-response functions and parameters (from TSM Table 2-5, Section 2.1.1.3.2).
Reference pathogen
Dose-response function
Parameter
Value/distribution
Norovirus
Beta Poisson model with the confluent
hypergeometric function
mean = -0.608
variance = 1.79
covariance = -1.03
mean = 0.194
variance = 2.54
covariance = -1.03
Campylobacter jejuni
Beta Poisson model with the confluent
hypergeometric function
mean = -0.177
variance = 1.303
covariance = -0.041
mean = 0.054
variance = 1.070
covariance = -0.041
Salmonella
Beta Poisson model approximation
0.3126
P
2884
Note: * The w,z bivariate normal distributions can also be used to derive distributions for the a and (3 dose response parameters per the
methods in Teunis et al. (2018) and Teunis et al. (2020).
Table G-2 summarizes the functions and parameters used to estimate the probability of illness given
infection for each reference pathogen. The probability of illness given infection for Salmonella was
modeled as a point estimate percentage. For norovirus and Campylobacter, the probability of illness
given infection was based on the approach presented in Teunis et al. (2018) and Teunis et al. (2020),
where this function is dependent on the dose and is called the hazard model function. The hazard
model function can use either two point estimates or two distributions. The point estimates r and n
can alternatively be represented by distributions w and z, which are incorporated into the model using
the mean, variance, and covariance.43
42 The choice of the symbols a, p, w, and z is consistent with the symbols used by Teunis et al. (2020) for these parameters.
43 The choice of the symbols r, x\, w, and z is consistent with the symbols used by Teunis et al. (2020) for these parameters.
G-3
-------
Table G-2. Probability of illness functions and parameters (from TSM Table 2-6, Section 2.1.1.3.3).
Reference pathogen Probability function Parameter Value/distribution
Norovirus Hazard model function w* mean = 1.74
variance = 5.55
covariance = -0.708
z* mean = 1.82
variance = 4.64
covariance = -0.708
mean = -2.744
variance = 1.337
covariance = 0.010
mean = -4.89 x 10"3
variance = 0.993
covariance = 0.010
Salmonella Constant c_pill (0.17 + 0.40)/2.0
midpoint of the
range [0.17 to 0.4]
Note: * The w,z bivariate normal distributions can be used to derive distributions for the r and r| hazard model parameters per the
methods in Teunis et al. (2018) and Teunis et al. (2020).
Pathogen and FIB Levels in Contamination
Table G-3 shows the pathogen and enterococci levels in each type of fecal source. Each value is a point
estimate that represents the midpoint of a reported range or a central tendency value for the
parameter. The information from TSM Tables 2-2 and 2-7 were entered into the python code
(Appendix H) as equations as shown in Table G-3. The value that the equation represents is also shown
to illustrate the differences between sources in a transparent fashion. The level of norovirus in effluent
was computed by dividing the level in human raw sewage by the average attenuation factor achieved
in treatment plants.
Table G-3. Pathogen and enterococci levels in fecal source (from TSM Table 2-2 in Section 2.1.1.1 and
Table 2-7 in Section 2.1.2.1).
Organism in fecal source (units)
Density in source
Density in source as shown in the python code
Enterococci in human raw (CFU/g)
7,943,282
10A((5.8+8.0)/2.0)
Enterococci in human effluent (CFU/L)
40
10A((0.5+2.7)/2.0)
Enterococci in gulls (CFU/g)
6,032,537
10A((0.5+2.7)/2.0)
Norovirus in human raw (genomes/g)
50,119
10A(4.7)
Norovirus in human effluent (genomes/L)
122
10A(4.7)/(10A((2.23+3)/2))
Salmonella in gulls (organisms/g)
446,684
10A((2.3+9.0)/2.0)
Campylobacter in gulls (organisms/g)
44,668
10A((3.3+6.0)/2.0)
Notes: The values in this table represent point estimates for the midpoint of the indicator and pathogen distributions. The value shown in
the middle column is represented by an equation in the Appendix H python code (far right column). The equation in the python code
uses the ranges shown in Table 2-2 and 2-7 (expressed as logio exponents) divided by 2 to arrive at the midpoint of the range, except
for norovirus in human raw, which is a central tendency point estimate.
Campylobacter jejuni Hazard model function
G-4
-------
Simulated Human Source Mix
The simulated human fecal source is a mixture of two sources: raw human and treated human effluent.
The QMRA is anchored with two inputs from the EPA's 2012 RWQC, the target illness rate and the
criteria magnitude (GM) for the FIB, in this case enterococci. The EPA's target illness rate of 36 NGI per
1,000 recreators is ultimately based on the EPA's epidemiological studies conducted in the 1970s
(Dufour, 1984).44 In order to have the level of enterococci at 35 CFU per 100 mL (the criteria
magnitude) and the illness rate at 36 NGI per 1,000 recreators the human fecal source needs to have a
blend of enterococci and pathogen levels that allow for this relationship. Neither human raw feces nor
human effluent alone provides the correct ratio of pathogens to enterococci to result in the illness
level of 36 NGI per 1,000 recreators at 35 enterococci CFU per 100 mL. Therefore a "simulated"
mixture of human raw and effluent was computed to result in the correct anchoring values. The
percentage of human source enterococci contributed by raw feces versus effluent was determined to
be 1.11% and 98.89% respectively. This ratio of human source inputs allows 35 enterococci CFU per
100 mL to result in an illness level of 36 NGI per 1,000 recreators, thus anchoring the modeling to the
epidemiological relationship. This simulated mixture was referred to as "idealized" by Schoen et al.
(2011) and Soller et al. (2014), who first published this anchoring approach.
Pathogen Prevalence and Infectious Fraction Parameters
Table G-4 presents the prevalence and infectious fraction of the reference pathogens found in gull
contamination. The minimum and maximum prevalence values from TSM Table 2-3 are converted into
a point estimate that is the midpoint of the range. The percentages and the form of the values that are
included in the python code (Appendix H) are both shown in Table G-4 for ease of understanding and
transparency.
In human contamination, the prevalence and infectious fraction of norovirus were assumed to be
1 (100%). In other words, the prevalence of norovirus in the human source is 100% and the fraction of
norovirus in the human source that is infectious is 100%.
Table G-4. Prevalence and infectious fraction of reference pathogens in gull contamination (from
TSM Tables 2-3 and 2-4 in Section 2.1.1.1).
Reference pathogen
Prevalence
percentage
Prevalence as shown
in python code*
Infectious
percentage
Infectious fraction as
shown in python
code**
Campylobacter jejuni
11%
(0.54+l)/2
16.7%
0.167
Salmonella
87.5%
(0.75+l)/2
16.7%
0.167
Notes:
* The prevalence values in this table represent point estimates for the midpoint of the minimum and maximum pathogen prevalence
estimates.
** The fraction of human infectious strains represents a qualitative estimate of "low" that was converted to 0.167 as the midpoint of 0
and 0.33 per Soller et al. (2010b).
44 As described in the 2012 RWQC, EPA translated the level of HCGI that corresponded with 35 enterococci CFU per 100 mL
in Dufour (1984) to the level of NGI from the EPA NEEAR studies that corresponded with 35 enterococci CFU per 100 mL
(Wymer et al., 2013).
G-5
-------
Proportion of FIB from Human Fecal Sources
The percentage of enterococci contributed by human fecal sources was calculated at multiple levels to
reflect a range of scenarios, including 10%, 20%, 25%, 30%, 33%, 50%, 60%, 67%, 75%, 90%, and 100%.
The Python Code in Appendix H includes these fecal source apportionment scenarios in a single run.
Initial FIB Level Estimates
Due to the iterative nature of the reverse QMRA process as applied in Step 4 of this TSM, an initial
value for the FIB level was needed to initiate the modeling. An FIB level of 50 CFU/100 mL (in the code
500 CFU/L) was provided as an initial estimate of the threshold FIB level corresponding to the target
risk level for each mixture type. This value was refined by computer code through numerical iteration
to derive the specific threshold FIB level applicable for each mixture type and at each simulated human
source contribution level as described in the analytical approach.
Analytical Approach
This section presents the analytical approach for:
• The reverse QMRA approach used to derive the adjusted FIB GM.
• The statistical approach used to derive the STV and BAV criteria based on the QMRA results.
The analytical approach references the TSM to which users may refer for detailed explanation of the
equations and terminology.
Reverse QMRA Approach
A preliminary estimate for the enterococci RBT for the different example scenarios was applied to
compute the density of each reference pathogen attributable to each source as described in TSM
Equations 3A and 3B. This initial RBT estimate was later refined by numerical iteration during the
reverse QMRA as described below. For definition of parameters see Section 2.2.2.4.5.
p NHS
CrT = P$iHbS X CFIB x [Eq. 3A]
kfib
rhs
Crp = Pfib x Cpib X —ws~ [Eq - 3B]
kfib
As the human source was a mixture of raw and secondary treated and disinfected WWTP effluent, the
ratio of pathogens to FIB in the human source in TSM Equation 3B was computed as follows:
d HS nraw p WWTP
= 0.9889^ + 0.0111-j^wfp [Eq. 3C]
kF1B kF1B kfib
Note that the constants in TSM Equation 3C were derived by anchoring a probability of illness of 36 per
1,00045 to an enterococci level of 35 CFU per 100 mL for an ingestion volume of 0.019 L (i.e., anchoring
to the EPA's recommended target illness rate and corresponding water quality value).
45 The recommended health goal associated with the national 2012 RWQC is 32 or 36 NGI illnesses per 1,000 primary
contact recreation events. These two values are not significantly different.
G-6
-------
The dose of each reference pathogen attributable to each source was computed perTSM Equations 3D
and 3E.
ll^s = V X CrNpHS X p^s X l\lHS [Eq. 3D]
= V XC%? [Eq. 3E]
The probability of infection from exposure to each reference pathogen derived from each source was
computed perTSM Equations 3F and 3G.
p%nhs = [Eq-3F]
Pin?5 = fir G4£) [Eq- 3G]
The probability of illness from exposure to each reference pathogen derived from each source was
computed perTSM Equations 3H and 31.
j-.rp.NHS rp w j-.rp.NHS ri- .,
Pill = Pill\inf X Pinf [Eq-3H]
nrp,HS rp w nrp,HS ri-
Pill ~ Pill\inf X ^nf [Ecl- 3|]
The probability of illness from exposure to each source, accounting for the effect of all reference
pathogens within that source was computed perTSM Equations 3J and 3K.
^5 = l-n,P(l-P,rH5) [Eq. 3J]
p$is = i-UrP(.i-pS'HS) [Eq-3K]
The total probability of illness accounting for exposure to both the human and nonhuman sources was
computed using TSM Equation 3L.
Pm = 1 - (1 - Pltr) X (1 - Pith [Eq. 3L]
As the computed value of the total probability of illness in Equation 3L differs from the target
probability of illness, the initial estimate for the RBT in the waterbody was iteratively adjusted by the
computer program until the computed value agreed with the target value of probability of illness. The
FIB RBT in the final iteration of this procedure is the waterbody FIB density level that corresponds to
the target probability of illness for the specified contamination mixture.
The approach for this analysis has been also implemented in computer code, which is included in
Appendix H (reverse QMRA example in Python code). This code example may be modified for specific
applications and may be further adapted to work in batch mode to run sensitivity analyses. The Python
code was run four separate times to account for each of the ingestion scenarios. The fecal source
apportionment scenarios are built into the code, so are run as a batch each time the code is run. The
user defines a location for an output file, which presents the results in a table with 5 rows (a header
and each of the four sources modeled in the code) and 12 columns (the row labels, and
11 apportionment scenarios).
G-7
-------
Derivation of STVand BAV
The example STVs and BAVs were derived based on the magnitude of the site-specific criteria obtained
from the reverse QMRA approach. The statistical formulas for deriving the STV and BAV are provided in
TSM Section 3.4.2. In this case study the standard deviation (SD) utilized by the EPA in the 2012 RWQC
was applied (0.44 logSD for enterococci).46
The approach for this analysis has been also implemented in computer code, which is included in
Appendix I (STV and BAV calculation example in R code). This code example may be modified for
specific applications and may be further adapted to work in batch mode to run sensitivity analyses.
Results
RBT Threshold—GM
Figure G-l shows the results of the base analysis, which includes the median ingestion volume of 19 mL
for varying gull/human fecal mixtures. The nonpathogenic/human mixture results are not shown in
Figures G-l, G-2, and G-3 because they are so similar to the gull/human results that the icons would
overlap.
Each of the points on Figure G-l represent an enterococci density consistent with a 36 NGI per
1,000 recreators target illness rate. The dotted horizontal line represents the EPA's recommended
enterococci value of 35 CFU enterococci per 100 mL. The point at 100% human source is, by definition,
anchored at 35 CFU enterococci per 100 mL because the epidemiological-based health relationship was
characterized at beaches affected by human fecal sources (Cabelli, 1983; Dufour, 1984). At 33% human
source (and 66% gull source) there is about a half log difference between the RBT and the RWQC. At
10% human source (and 90% gull source) there is a log difference between the RBT and the RWQC.
46 As noted in Section 3.4.2, if estimates of the SD of the indicator levels for the evaluated waterbody are not available, the
SDs utilized by EPA in the 2012 RWQC can be applied (0.44 logSD for enterococci or 0.40 logSD for E. coli).
G-8
-------
500
450
400
3 350
E
o
o 300
& 250
S 200
"u
8 150
o
S_
(D
£ 100
LU
50
0
O Enterococci RBT for General
Population (gulls' model -
Dufour)
o
2012 RWQC
; Recommendation
o
o
o
o
O
O
CL 0
10 20 30 40 50 60 70 80 90 100
Fraction of Human Source Contribution (%)
Figure G-l. Base analysis results—enterococci RBT for waters impacted by human and gull fecal
mixtures.
Figure G-2 compares the RBTs using the general population ingestion volumes from Dufour and DFB.
The lower ingestion volume from the DFB study (16 mL/event) versus Dufour (19 mL/event) results in
higher RBTs than the base analysis. The incidental ingestion volume is the primary exposure pathway
considered within the scope of the QMRA.
G-9
-------
500
450
400
3 350
E
o
o 300
& 250
S 200
"u
8 150
o
S_
(D
£ 100
LU
50
O Enterococci
Population i
RBT for General
[gulls' model -
~
Dufour)
o
~ Enterococci
Population i
RBT for General
[gulls' model - DFB)
~
° ~
~ o
~ o
o
~
° £
* 8
0
fcL £
10 20 30 40 50 60 70 80 90 100
Fraction of Human Source Contribution
Figure G-2. Enterococci RBTs corresponding to 36 NGI per 1,000 recreators for the general population
using estimates of incidental ingestion from Dufour et al. (2017) and DeFlorio-Barker et al. (2017).
Figure G-3 compares the RBTs for the general population and the 6 to 10 years old age group using the
median ingestion volumes from Dufour. The higher ingestion volumes for children ages 6 to
10 years old (25 mL/event) results in RBTs that are lower than the base analysis. It is worth noting that
at 100% human contribution, the enterococci RBT for children 6-10 years old is lower than the RWQC
recommendation because the RWQC is benchmarked to the general population health relationship.
G-10
-------
400
350
300
o250
o
¦ 200
U
CO
qc 150
<5
o
o
glOO
(D
4-»
w 50
O Enterococci RBT for General
Population (gulls' model)
~ Enterococci RBT for Children
(gulls' model)
2012 RWQC Recommendation
u
~
O
~
~ o
o
o
D ~
c
o
o ~
¦
— —
_p_ _ .
o
—~—
10
20
30 40 50 60 70 80
Fraction of Human Source Contribution (%)
90
100
Figure G-3. Enterococci RBTs corresponding to 36 NGI per 1,000 recreators for the general population
and children ages 6 to 10 years.
Table G-5 presents the enterococci RBTs for gull/human fecal source mixtures at different fractions of
human fecal contributions. All four ingestion scenarios are shown (base analysis and three sensitivity
analyses). The model was anchored at 35 CFU enterococci per 100 mL with the general population
Dufour ingestion scenario (base analysis). Similarly, Table G-6 presents the enterococci RBT for
nonpathogenic/human source mixtures. The RBTs for gull/human and nonpathogenic/human mixtures
are the same except for two values at the 10% level where the CFU differ by 1.
Table G-5. RBTs (enterococci CFU/100 mL) for waters impacted by human and gull contamination at
36 NGI per 1,000 recreators.
Ingestion scenario
Percentage of human source FIB contribution
< 10%
<25%
<33%
General population (Dufour)
349
140
106
General population (DFB)
415
166
126
6 to 10 (Dufour)
265
106
81
6 to 10 (DFB)
166
66
50
G-ll
-------
Table G-6. RBTs (enterococci CFU/100 mL) for waters impacted by human and nonpathogenic
contamination at 36 NGI per 1,000 recreators.
Ingestion scenario
Percentage of human source FIB contribution
< 10%
<25%
<33%
General population (Dufour)
350
140
106
General population (DFB)
415
166
126
6 to 10 (Dufour)
266
106
81
6 to 10 (DFB)
166
66
50
STV and BA V Calculations
The values in Table G-5 were used to calculate the corresponding STVs (Table G-7) and BAVs (Table G-
9) for gull/human sources. The values in Table G-6 were used to calculate the corresponding STVs
(Table G-8) and BAVs (Table G-10) for nonpathogenic/human sources. Example R code for calculating
the STV and BAV is available in TSM Appendix I.
Table G-7. STVs (enterococci CFU/100 mL) for waters impacted by human and gull contamination.
Ingestion scenario
Percentage of human source FIB contribution
< 10%
<25%
<33%
General population (Dufour)
1279
512
388
General population (DFB)
1519
609
461
6 to 10 (Dufour)
972
389
295
6 to 10 (DFB)
608
243
184
Table G-8. STVs (enterococci CFU/100 mL) for waters impacted by human and nonpathogenic
contamination.
Ingestion scenario
Percentage of human source FIB contribution
< 10%
<25%
<33%
General population (Dufour)
1282
513
388
General population (DFB)
1522
609
461
6 to 10 (Dufour)
974
390
295
6 to 10 (DFB)
609
244
184
G-12
-------
Table G-9. BAVs (enterococci CFU/100 mL) for waters impacted by human and gull contamination.
Ingestion scenario
Percentage of human source FIB contribution
< 10%
<25%
<33%
General population (Dufour)
691
277
210
General population (DFB)
821
329
249
6 to 10 (Dufour)
526
211
160
6 to 10 (DFB) 328 132 100
Table G-10. BAVs (enterococci CFU/100 mL) for waters impacted by human and nonpathogenic
contamination.
Ingestion scenario
Percentage of human source FIB contribution
< 10%
<25%
<33%
General population (Dufour)
693
277
210
General population (DFB)
823
329
249
6 to 10 (Dufour)
527
211
160
6 to 10 (DFB) 329 132 100
Discussion
The enterococci RBTs derived above reflect scenarios where gull feces (or nonpathogenic FIB sources)
are the predominant source of enterococci in the waterbody. Varying levels of human to gull source
contributions were modeled to provide context. The enterococci RBTs provide a GM threshold for the
hypothetical site that is consistent with the EPA's recommended target illness rate. For example, if the
waterbody was determined to have predominantly fecal loading from gull feces with an approximate
25% human fecal contribution, the reverse QMRA output GM value of 140 enterococci CFU/100 mL
(Table G-5) could be chosen as the adjusted water quality criteria value. Table G-5 only shows human
fecal sources below 33% because waters with low human inputs are candidates for alternative RBTs.
Figure G-l shows the enterococci density corresponding to the target illness rate for the base analysis
for 11 different fecal apportionment scenarios. The model anchors the 100% human source at
35 enterococci CFU per 100 mL. The fecal apportionment scenarios with less human fecal source and
more gull fecal source have higher corresponding enterococci densities corresponding to the target
illness rate. For example, at < 10% human fecal sources the corresponding enterococci density RBT is
349 enterococci CFU per 100 mL (Table G-5).
Figure G-2 shows the base scenario compared to the ingestion volume for the general population
calculated in DFB. The lower ingestion rate from DFB results in higher enterococci densities
corresponding to the target illness rate for each of the fecal apportionment scenarios. The EPA chose
the general population data from Dufour for the base analysis because it represents empirically
measured ingestion data, whereas the ingestion results from DFB are modeled from estimated time
spent in the water for NEEAR study participants combined with the Dufour ingestion data.
Figure G-3 shows the base scenario compared to the ingestion volume for children ages 6 to 10 years
old from Dufour. The higher ingestion rate for children 6 to 10 years old results in lower enterococci
G-13
-------
densities corresponding to the target illness rate for each of the fecal apportionment scenarios. At
100% human source the base scenario is anchored at 35 enterococci CFU per 100 mL and for children 6
to 10 years old the 100% human source is below the anchor line (Figure G-3).
For the base scenario a BAV value of 277 and an STV value of 513 enterococci CFU/100 mL could be
chosen consistent with the GM (Tables G-9 and G-7). The EPA expects users to apply the frequency
consistent with the unadjusted WQS and determine the duration as part of any alternative WQS
submitted to the EPA.
The results of this case study confirm earlier analyses, including Soller et al. (2014) and Soller et al.
(2010b), that found sea gulls represent a very low risk fecal source when they are the predominant
source of contamination. The case study illustrates how GM RBTs, STVs, and BAVs, may be derived, and
the sensitivity of the QMRA output to alternative assumptions for mixtures with degrees of nonhuman
contamination. In each case, the RBTs are considerably higher than if only human source
contamination were present. The almost identical results for human and nonpathogenic mixtures,
compared to human and sea gull contamination mixtures, is further evidence that sea gull
contamination contributes a very low level of risk.
References
Cabelli, V.J. 1983. Health Effects Criteria for Marine Recreational Waters. EPA 600/1-80-031. U.S.
Environmental Protection Agency, Office of Research and Development, Health Effects
Research Laboratory, Research Triangle Park, NC.
Dufour, A.P. 1984. Health effects criteria for fresh recreational waters. Research Triangle Park, NC: U.S.
Environmental Protection Agency. EPA 600/1-84-004.
Dufour, A.P., T.D. Behymer, R. Cantu, M. Magnuson, and L.J. Wymer. 2017. Ingestion of swimming pool
water by recreational swimmers. Journal of Water and Health 15(3):429-437.
DeFlorio-Barker, S., B.F. Arnold, E.A. Sams, A.P. Dufour, J.M. Colford, Jr., S.B. Weisberg, K.C. Schiff, and
T.J. Wade. 2017. Child environmental exposures to water and sand at the beach: Findings from
studies of over 68,000 subjects at 12 beaches. Journal of Exposure Science & Environmental
Epidemiology 00:1-8.
Scallan, E., R.M. Hoekstra, F.J. Angulo, R.V. Tauxe, M.A. Widdowson, S.L. Roy, J.L. Jones, and P.M.
Griffin. 2011a. Foodborne illness acquired in the United States—major pathogens. Emerging
Infectious Diseases 17(1):7-15.
Scallan, E., P.M. Griffin, F.J. Angulo, R.V. Tauxe, and R.M. Hoekstra. 2011b. Foodborne illness acquired
in the United States—unspecified agents. Emerging Infectious Diseases 17(l):16-22.
Soller, J.A., T. Bartrand, N.J. Ashbolt, J. Ravenscroft, and T.J. Wade. 2010a. Estimating the primary
etiologic agents in recreational freshwaters impacted by human sources of faecal
contamination. Water Research, 44(16):4736-4747.
Soller, J.A., Schoen, M.E., Bartrand, T., Ravenscroft, J.E., Ashbolt, N.J. 2010b. Estimated human health
risks from exposure to recreational waters impacted by human and non-human sources of
faecal contamination. Water Research, 44(16):4674-4691.
G-14
-------
Soller, J.A., M.E. Schoen, A. Varghese, A.M. Ichida, A.B. Boehm, S. Eftim, N.J. Ashbolt, and J.E.
Ravenscroft. 2014. Human health risk implications of multiple sources of faecal indicator
bacteria in a recreational waterbody. Water Research 66:254-264.
Teunis, P.F.M., A. Bonacic Marinovic, D.R. Tribble, C.K. Porter, and A. Swart. 2018. Acute illness from
Campylobacter jejuni may require high doses while infection occurs at low doses. Epidemics
24:1-20.
Teunis, P.F.M., F.S. Le Guyader, P. Liu, J. Ollivier, and C.L. Moe. 2020. Noroviruses are highly infectious
but there is strong variation in host susceptibility and virus pathogenicity. Epidemics 32:100401.
U.S. EPA (U.S. Environmental Protection Agency). 2012. Recreational Water Quality Criteria. 820-F-12-
058. U.S. Environmental Protection Agency, Office of Water, Washington, DC.
Wymer, L.J., T.J. Wade, and A.P. Dufour. 2013. Equivalency of risk for a modified health endpoint: A
case from recreational water epidemiology studies. BMC Public Health 13:459.
G-15
-------
Appendix H: Python Code for Appendix G
This appendix provides the annotated code for the Appendix G Gull Case Study. The code estimates the
FIB (in this case enterococci) risk-based thresholds (RBTs) corresponding to a target illness rate for the
gull/human fecal mixtures and nonpathogenic/human fecal mixtures. Although the Appendix G Gull
Case Study does not include swine or chicken sources, the code also includes swine (pig)/human and
chicken/human fecal mixtures. It is recommended that this code be utilized by experienced
programmers after a careful reading of its syntax and structure with reference to the appropriate
equations described in the TSM. This code should not be considered a tool. In this example code all the
references to FIB refer to enterococci, however, the code can be adapted for other FIB such as E. coli.
Note that the code outputs RBTs in units of CFU/100 mL. If the simulated human fecal source needs to
be adjusted because a different target illness rate or FIB is used, this code could be adapted. Example
constants for the simulated "idealized" human mixture of effluent to raw are in Table H-l. A key for the
annotated code is presented in Table H-2 and the definitions and sources for variables for the
Appendix G Gull Case study is presented in Table H-3.
Table H-l. Constants for simulated human mixtures for other target illness rates and FIB.
2012 RWQC target illness
rate
FIB density
ratio_hum_raw
rat io_h u m_wwt p
36 NG 1/1,000 recreators
35 CFU/100 mLa
enterococci (culture-based)
0.9889
0.0111
32 NG 1/1,000 recreators
30 CFU/100 mLb
enterococci (culture-based)
0.9882
0.0118
36 NG 1/1,000 recreators
126 CFU/100 mLc
E. coli (culture-based)
0.9869
0.0131
32 NG 1/1,000 recreators
100 CFU/100 mLc
E. coli (culture-based)
0.9845
0.0155
Notes:
a. This is the mixture shown in the code below and corresponds to the scenario described in Appendix G.
b. In this example code, 100% human source is benchmarked to 36 NGI per 1,000 recreators and 35 CFU enterococci per 100 mL. If
32 NGI per 1,000 recreators is chosen as the target illness rate, the associated enterococci is 30 CFU per 100 mL. The two values for
ratio_hum_raw and ratio_hum_wwtp would need to be replaced by the values shown in the table.
c. If E. coli is used, the two values for ratio_hum_raw and ratio_hum_wwtp would need to be replaced by the values shown in the table
corresponding to the target illness rate chosen. In addition, all the FIB densities for each source would also need to be changed to
correspond with £ coli densities.
Table H-2. Key for annotated code.
Key
BOLD^Actual code for python
(software available at http://www.python.org/getit/)
PLAIN—Comments for guidance
H-l
-------
Table H-3. Appendix G Gull Case Study: definitions and sources for variables in equations.
In-text variable
Code variable
X represents the pathogen of interest
Definition
rNHS rHS
LIfp f
No explicit variable defined.
Computation embedded within the dose
function for program efficiency.
Estimated waterbody density of reference pathogen
(number of pathogens or genomes/L) in nonhuman and
human sources.
~NHS
Vfib
pcfu
Proportion of FIB density attributable to the nonhuman
source NHS (unitless; evaluated here as a vector
assuming multiple values)
Cfib
fib_dens
Initial estimate for FIB RBT (CFU/L) corresponding to
the target probability of illness for each proportion.
d NHS
l\rp
xx_dens['PathogenName']
Density of reference pathogens (number of pathogens
or genomes/L or number of pathogens or genomes/g)
in the nonhuman source NHS
n NHS
kFIB
xx_dens['ent']
Density of FIB (CFU/L or CFU/g) in the nonhuman
source NHS
d raw
r\rp
hum_raw_dens['nor']
Density of each reference pathogen (number of
pathogens or genomes/g) in raw human contamination;
nWWTP
t\rp
hum_wwtp_dens['nor']
Density of each reference pathogen (number of
pathogens or genomes/L) in WWTP-treated human
contamination;
nraw
kFIB
hum_raw_dens['ent']
Density of FIB (CFU/g) in raw human contamination
nWWTP
kFIB
hum_wwtp_dens['ent']
Density of FIB (CFU/L) in WWTP-treated human
contamination
..NHS i/HS
l*rp > r i'j)
dose['fib_dens'/PathogenName']
Estimated dose ingested of reference pathogen from
nonhuman and human sources respectively (number of
pathogens). Common implementation in dose function.
V
voljng
Volume of water ingested (L)
~NHS
Hrp
sw_frinf['PathogenName']
Fraction of human infectious pathogenic strains
(unitless)
i NHS
'rp
sw_prev['PathogenName']
Prevalence of infection in the nonhuman source
(unitless)
rp
Pill\inf
drp_X['c_pill']
Proportion of infections resulting in illness (unitless)
frp
J d-r
drf_bp,drf_exp,drf_hyp. Dose response
parameters for these functions are
found in drp_X.
Dose-response functions for various pathogens
(unitless)
prp.NHS prp,HS
ill ' ill
pill_X. (No explicit superscripting was
required for NHS and HS because the
pathogens found in the two sources did
not overlap.)
Probability of illness from a specific reference pathogen
rp originating from the nonhuman source NHS and
human source HS, respectively (unitless)
nNHS nHS
"ill > "ill
Directly implemented in tpill function
below.
Estimated probability of illness from the nonhuman
source NHS and human source HS, respectively
(unitless)
Pill
tpill
Estimated total probability of illness accounting for
exposure to both the nonhuman and human sources of
contamination (unitless)
H-2
-------
#!/usr/local/bin/python
#########################################################################################
### Example code for implementing a "reverse" QMRA. Determines FIB density
### corresponding to specified level of risk for contamination mixtures deriving
### from human and alternative animal (swine, chicken, gull, non-pathogenic) sources.
### Uses the Teunis 2020 and 2018 Bayesian dose response function for norovirus and camp, jejuni
#########################################################################################
from numpy import * ### Imports Python NumPy module
from scipy.optimize import fsolve ### Imports necessary sub-module from Python SciPy module
from scipy.special import hyplfl ### Imports necessary sub-module from Python SciPy module
random.seed(l) ### Ensures numerical replicability
print ("REVERSE QMRA EXAMPLE FOR ANIMAL AND HUMAN MIXTURES USING THE BAYESIAN TEUNIS DRFs FOR NORO
and G ") ## Prints output message
print("")
print( "FINDS FIB DENSITY CORRESPONDING TO SPECIFIED LEVEL OF RISK") ## Prints output message
print("")
risk=0.036 ### Target illness rate from the 2012 RWQC, specified as a probability.
### In this case, the chosen level of risk corresponds to 36 illnesses per 1000 swim events (unitless).
vol_ing=0.019 ### Volume of water ingested during the recreational event, in L. General population median value from
Dufour et al. 2017.
### The following lines create arrays of FIB and pathogen densities found in swine, chicken and gull contamination
### based on the midpoint of the range of values in TSM Section 2 Tables 2-2 and 2-7.
### Array of FIB (cfu/g) and pathogen (organisms/g) densities found in swine contamination
sw_dens={,ent':10**((5.3+7.2)/2.0),,cj,:10**((2.0+5.7)/2.0),,g,:10**((0+6.8)/2.0),,s,:10**((2.8+4.9)/2.0),,cr,:10**((1.7+3.6)
/2.0),'c57':10**((0.0+7.0)/2.0)}
### Array of FIB (cfu/g) and pathogen densities (organisms/g) found in chicken contamination
ch_dens={'ent':10**((5.0+7.0)/2.0),'cj':10**((2.8+6.5)/2.0),'g':0,'s':10**((-1.0+4.5)/2.0),'cr':0,'c57':0}
### Array of FIB (cfu/g) and pathogen (organisms/g) densities found in gull contamination
gu_dens={'ent':10**((4.114+9.447)/2.0),'cj':10**((3.3+6.0)/2.0),'g':0,'s':10**((2.3+9.0)/2.0),'cr':0,'c57':0}
### Array of FIB (cfu/L) and pathogen (organisms/L) densities found in non-pathogenic contamination.
### Artificial density value used for enterococci density to avoid 0/0 error when calculating path:fib ratio.
np_dens={'ent':0.1,'cj':0,'g':0,'s':0,'cr':0,'c57':0}
### Nested dictionary of animal source pathogen and FIB densities.
dens_source={'swine':sw_dens,'chicken':ch_dens,'gulls': gu_dens,'non-path':np_dens}
### The following lines create arrays of pathogen prevalence rates for swine, chicken and gull contamination
### based on the midpoint of the range of values in TSM Section 2 Table 2-3.
### Array of pathogen prevalence rates for swine contamination (unitless)
sw_prev={'cj':(0.46+0.98)/2,'g':(0.033+0.18)/2.0,'s':(0.079+0.15)/2.0,'cr':(0+0.45)/2.0,'c57':(0.001+0.12)/2.0}
### Array of pathogen prevalence rates for chicken contamination (unitless)
ch_prev={'cj':(0.57+0.69)/2,'g':0,'s':(0.0+0.95)/2.0,'cr':(0.06+0.27)/2.0,'c57':0}
### Array of pathogen prevalence rates for gull contamination (unitless)
gu_prev={'cj':(54+100)/2/100,'g':0,'s':(75+100)/2/100,'cr':0,'c57':0}
### Array of pathogen prevalence rates for non-pathogenic contamination (unitless)
np_prev={'cj':0.0,'g':0,'s':0.0,'cr':0,'c57':0}
### Nested dictionary of pathogen prevalence rates.
prev_source={'swine':sw_prev,'chicken':ch_prev,'gulls':gu_prev,'non-path':np_prev}
H-3
-------
### The following lines create arrays of pathogen infectious fraction estimates for swine, chicken and
### gull contamination based on TSM Section 2 Table 2-4.
### Array of pathogen infectious fraction estimates for swine contamination (unitless).
sw_frinf={,cj':0.833,,g,:0.833,,s,:0.5,,cr,:0.167,,c57,:0.833}
### Array of pathogen infectious fraction estimates for chicken contamination (unitless).
ch_frinf={,cj':0.5,,g,:0,,s,:0.5,,cr,:0.167,,c57,:0}
### Array of pathogen infectious fraction estimates for gull contamination (unitless).
gu_frinf={,cj,:0.167,,g':0,,s,:0.167,,cr,:0,,c57,:0}
### Array of pathogen infectious fraction estimates for non-pathogenic contamination (unitless)
np_frinf={,cj,:0.0,,g,:0,,s,:0.0,'cr,:0,,c57,:0}
### Nested dictionary of infectious fraction rates.
frinf_source={,swine,:sw_frinf,,chicken,:ch_frinf,'gulls,:gu_frinf,,non-path,:np_frinf}
### The following line creates an array of FIB and pathogen densities found in raw human contamination
### based on the midpoint of the range of values in TSM Section 2 Tables 2-2 and 2-7.
### Array of FIB (cfu/L) and pathogen (organisms/L) densities found in raw human contamination
hum_raw_dens={'ent,:10**((5.8+8.0)/2.0),,nor,:10**(4.7)}
### Norovirus attenuation level achieved in WWTP plants based on the midpoint of the range of values in Soller, et al.
(2010b).
wwtp_atten=10**((2.23+3)/2)
### The following line creates an array of FIB and pathogen densities found in WWTP-treated human contamination based
### on the midpoint of the range of values in TSM Section 2 Table 2-3 and the attenuation value specified in the line above.
### Array of FIB (cfu/L) and pathogen (organisms/L) densities found in WWTP-treated human contamination
hum_wwtp_dens={,ent,:10**((0.5+2.7)/2.0),'nor,:hum_raw_dens[,nor,]/wwtp_atten}
### Ratio of pathogen to FIB levels in raw human contamination
ratio_hum_raw=hum_raw_dens[,nor,]/hum_raw_dens[,ent']
### Ratio of pathogen to FIB levels in WWTP-treated human contamination
ratio_hum_wwtp=hum_wwtp_dens[,nor,]/hum_wwtp_dens[,ent']
### Ratio of pathogen to enterococci level in mixture of raw and WWTPtreated human contamination.
### The fractions of raw and WWTP
### contamination in the mixture were determined such that an ingestion volume of 0.019 L results in an illness rate
### of 36/1000 and an enterococci level of 35 cfu/lOOmL.
### This implementation is based on Equation 3C and represents the simulated human source
ratio_dens_hmix=0.9889*ratio_hum_raw+0.0111*ratio_hum_wwtp
### Array specifies alternative levels of non-human source contribution of total FIB level in surface water. For
### instance, 0 corresponds to 100% human contribution; 0.1 corresponds to 10% non-human source and 90% human
### contribution, etc.
pcfu=array([0,0.1,0.25,0.33,0.4,0.5,0.67,0.7,0.75,0.8,0.9])
### Array contains an initial estimate of FIB levels in CFU/L corresponding to specified risk for the different
### percentages of the mixture defined in the line above.
fib_dens=array([500,500,500,500,500,500,500,500,500,500,500])
### The function "dose" defined below computes the dose of the specified pathogen ingested during a swim event by
### estimating the % of FIB from the source in which that pathogen is found, multiplying it by the ratio of pathogens
### per FIB in that source, and then multiplying that by the volume of water ingested.
### This implementation is based on Equations 3A, 3B, 3D and 3E.
def dose(fib_dens,PathogenName,AnimalSource):
if PathogenName=='nor':
result = (l-pcfu)*fib_dens*ratio_dens_hmix*vol_ing
else:
result =
pcfu*fib_dens*dens_source[AnimalSource][PathogenName]/dens_source[AnimalSource]['ent,]*prev_source[AnimalSou
rce][PathogenName]*frinf_source[AnimalSource][PathogenName]*vol_ing
return result
H-4
-------
### Beta Poisson Dose Response Function based on TSM Section 2 Table 2-5.
def drf_bp(a,b,c_pill,d):
return (l-(l+d/b)**-a)*c_pill
### Exponential Dose Response Function based on TSM Section 2 Table 2-5.
def drf_exp (r,c_pill,d):
return(l-exp(-r*d))*c_pill
#####################################################################################################
## The functions below implement the Teunis 2018 and Teunis 2020
### dose response functions with parameters jointly estimated by the Bayesian posterior approach
#####################################################################################################
#
n_samp=10000 # number of samples to draw from posterior distribution
### This function jointly samples w and z (transformed parameters) from posterior distribution expressed as bivariate
normal
### inputs are vector of means and the variance covariance matrix
def sample_w_z(means, vcov,n):
return(random.multivariate_normal(means,vcov,n).T)
### This function computes probability of infection based on w, z parameter distribution using the confluent
hypergeometric function
### Inputs are joint w,z distribution and dose
def pinf_t20(inf_dist,d):
w=inf_dist[0]; z=inf_dist[l]
u=exp(w)/(l+exp(w)); v=exp(z)
alpha=v*u; beta=v*(l-u)
p=l-hyplfl(alpha,(alpha+beta),-d)
p=clip(p, 0,1)
return(p)
### This function computes conditional probability of illness based on w, z parameter distribution
### Inputs are joint w,z distribution and dose
def cpill_t20(ill_dist,d):
w=ill_dist[0]; z=ill_dist[l]
u=exp(w)/(l+exp(w)); v=exp(z)
r=v*u; eta=v*(l-u)
p=l-(l+d/eta)**-r
p=clip(p, 0,1)
return(p)
### This function computes the median probability of illness combining probability of infection and conditional probability
of illness given infection
### Inputs are joint w,z distribution for both infection and illness, and dose
def drf_hyp_t20(inf_dist,ill_dist,dose):
pill_vector=[]
for d in dose:
pinf=pinf_t20(inf_dist,d)
cpill=cpill_t20(ill_dist,d)
pill=pinf*cpill
pill_vector.append(median(pill))
return(array(pill_vector))
H-5
-------
####################################################################################################
### Dose Response Parameters for salmonella based on Table 2-5
drp_s={,a,:0.3126,,b':2884,,c_pill,:(0.17+0.40)/2.0}
### Dose Response Parameters for giardia based on
drp_g={,r,:0.0199,'c_pill,:(0.2+0.7)/2.0}
### Dose Response Parameters for Cryptosporidium based on Table 2-5
drp.cr^'r'iO.OS/c.pill'iO.S}
### Dose Response Parameters for e-coli 0157:1-17 based on Table 2-5
drp_c57={,a,:0.248,,b':48.8,,c_pill,:(0.2+0.6)/2.0}
### Dose Response Parameters for norovirus based on Table 2-5
drp_nor={'inf_means':[-0.608, 0.194],'inf_vcov':[[1.79, -1.03], [-1.03, 2.54]],,ill_means': [1.74,1.82],'ill_vcov,:[[5.55, -
0.708], [-0.708, 4.64]]}
### Dose Response Parameters for Campylobacter jejuni based on Table 2-5
drp_cj={'inf_means':[-0.177,0.054],'inf_vcov':[[ 1.303,-0.041], [-0.041,1.070]],'ilLmeans': [-2.744,-4.89E-
3],'ill_vcov,:[[1.337, 0.010], [0.010, 0.993]]}
############################################################
### Sample from normalized parameter posterior distributions
### for DRFs based on Teunis 2018 and 2020
###(Generate w,z distributions for infection and illness)
############################################################
nor_inf_dist=sample_w_z(drp_nor['inf_means'], drp_nor['inf_vcov'],n_samp)
nor_ill_dist=sample_w_z(drp_nor[,ill_means,], drp_nor[,ill_vcov'],n_samp)
cj_inf_dist=sample_w_z(drp_cj['inf_means'], drp_cj['inf_vcov'],n_samp)
cj_ill_dist=sample_w_z(drp_cj[,ill_means'], drp_cj['ill_vcov,],n_samp)
### The following functions are implementations of the generic representation in Equation 3F, 3G, 3H, and 31
### The function pill_cj computes the probability of illness from exposure to Campylobacter jejuni using the dose
### response function, dose response parameters and dose for that pathogen,
def pill_cj(fib_dens,animalsource):
return drf_hyp_t20(cj_inf_dist,cj_ill_dist,dose(fib_dens,'cj,,animalsource))
### The function pill_s computes the probability of illness from exposure to salmonella using the dose response function,
### dose response parameters and dose for that pathogen,
def pill_s(fib_dens,animalsource):
return drf_bp(drp_s[,a,],drp_s[,b,],drp_s[,c_piN,],dose(fib_dens,,s,,animalsource))
### The function pill_g computes the probability of illness from exposure to giardia using the dose response function,
### dose response parameters and dose for that pathogen,
def pill_g (fib_dens,animalsource):
return drf_exp(drp_g[,r,],drp_g[,c_piM,],dose(fib_dens,,g,,animalsource))
### The function pill_cr computes the probability of illness from exposure to Cryptosporidium using the dose response
### function, dose response parameters and dose for that pathogen,
def pill_cr (fib_dens,animalsource):
return drf_exp(drp_cr[,r,],drp_cr[,c_piM,],dose(fib_dens,,cr,,animalsource))
### The function pill_c57 computes the probability of illness from exposure to e-coli 0157:1-17 using the dose response
### function, dose response parameters and dose for that pathogen,
def pill_c57(fib_dens,animalsource):
return drf_bp(drp_c57[,a,],drp_c57[,b,],drp_c57[,c_piN,],dose(fib_dens,,c57,,animalsource))
H-6
-------
### The function pill_nor computes the probability of illness from exposure to norovirus using the dose response
### function, dose response parameters and dose for that pathogen,
def pill_nor(fib_dens,animalsource):
return drf_hyp_t20(nor_inf_dist,nor_ill_dist,dose(fib_dens,,nor,,animalsource))
### The function tpill computes the probability of illness from exposure to all the pathogens listed above.
### This implementation is based on equations 3J,3K, and 3L
def tpill(fib_dens,animalsource):
return(l-(l-pill_cj(fib_dens,animalsource))*(l-pill_s(fib_dens,animalsource))*(l-pill_g(fib_dens,animalsource))*(l-
pill_cr(fib_dens,animalsource))*(l-pill_c57(fib_dens,animalsource))*(l-pill_nor(fib_dens,animalsource)))
### The function deltarisk estimates the difference between the target risk level and the current value computed
### using an initial estimate for fib_dens
def deltarisk (fib_dens,animalsource):
return (risk-tpill(fib_dens,animalsource))
### The following lines define the different types of animal source available in this code.
animalsourcel=(,swine')
animalsource2=('chicken')
animalsource3=(,gulls')
animalsource4=('non-path')
### The following lines cause the initial estimate for fib_dens to be adjusted until deltarisk equals zero.
### In other words, it finds the FIB density levels that correspond to the target risk level for different
### mixtures of swine and human contamination.
fib_criticall=fsolve(deltarisk, fib_dens, args=animalsourcel)
fib_critical2=fsolve(deltarisk, 0.5*fib_dens, args=animalsource2) # Note: a factor to initial estimate of fib_dens to ensure
#convergence
fib_critical3=fsolve(deltarisk, fib_dens, args=animalsource3)
fib_critical4=fsolve(deltarisk, fib_dens, args=animalsource4)
fib_critical4[fib_critical4 < 0] = nan # replace negative values in 100% non pathogenic contamination with NA
print("")
print("THE FIB RBTs in CFU/lOOmL CORRESPONDING TO THE TARGET RISK AT DIFFERENT LEVELS OF SWINE
CONTRIBUTION TO THE ANIMAL-HUMAN CONTAMINATION MIXTURE ARE: ") # Prints this message
print("")
print( pcfu, (fib_criticall/10).round(0))
print!"")
print! "THE FIB RBTs in CFU/lOOmL CORRESPONDING TO THE TARGET RISK AT DIFFERENT LEVELS OF CHICKEN
CONTRIBUTION TO THE ANIMAL-HUMAN CONTAMINATION MIXTURE ARE: ") # Prints this message
print!"")
print! pcfu, (fib_criticall/10).round(0))
print!"")
print! "THE FIB RBTs in CFU/lOOmL CORRESPONDING TO THE TARGET RISK AT DIFFERENT LEVELS OF GULL
CONTRIBUTION TO THE ANIMAL-HUMAN CONTAMINATION MIXTURE ARE: ") # Prints this message
print!"")
print! pcfu, (fib_criticall/10).round(0))
print!"")
print! "THE FIB RBTs in CFU/lOOmL CORRESPONDING TO THE TARGET RISK AT DIFFERENT LEVELS OF NON-PATHOGENIC
CONTRIBUTION TO THE ANIMAL-HUMAN CONTAMINATION MIXTURE ARE: ") # Prints this message
print!"")
print! pcfu, (fib_criticall/10).round(0))
### Postprocessing results. Generates output results as Excel file after appropriate transformations, transposing, and ###
### conversion to CFU/lOOmL to be consistent with TSM tables
import pandas as pd
H-7
-------
hc=['10%', '20%', '25%', '30%','33%','50%', '60%','67%', '75%', '90%','100%'] # he is an array with percent human
### contribution to fib but in reverse order compared to the nonhuman contribution array in the code above
## The lines below creates a dataframe using the arrays but reverses the order to correspond to he, rounds them, and
### divides by 10 to generate output in CFU/lOOmL
df=pd.DataFrame({'% Human FIB Contribution':hc,'Human-Swine Mixture':(fib_criticall[::-l]/10).round(0),'Human-
Chicken Mixture':(fib_critical2[::-l]/10).round(0),'Human-Gull Mixture':(fib_critical3[::-l]/10).round(0),'Human-Non-
pathogentic Mixture':(fib_critical4[::-l]/10).round(0)})
df=df.transpose()
df.to_excel(r'C:\YourOutputFolder\RBT_Simulation_Output.xlsx',index=True,header=None)
##########QA TO ENSURE CONVERGENCE
######## IF THESE VALUES DO NOT MATCH THE TARGET RISK, ADJUST THE INITIAL ESTIMATE FOR FIB_DENS UNTIL YOU
### SEE CONVERGENCE.
print("Convergence check foranimalsourcel, tpill(fib_criticall,animalsourcel))
print("Convergence check foranimalsource2, tpill(fib_critical2,animalsource2))
print("Convergence check for", animalsource3, tpill(fib_critical3,animalsource3))
print("Convergence check for", animalsource4, tpill(fib_critical4,animalsource4))
H-8
-------
Appendix I: Example Code for STV and BAV Calculation
This appendix provides an annotated code example for the computation of statistical threshold values
(STVs) and beach action values (BAVs). This code is an example and should not be considered a tool.
The computation of STVs and BAVs requires as input the magnitude of the indicator criterion in units of
CFU/100 mL (the geometric mean value of the criterion) and the site-specific standard deviation (SD) of
the logio transformed indicator values in units of logio CFU/100 mL. If the SD of the logio transformed
indicator values for the waterbody used in the water quality study is not available, users may use the
values from the EPA's 2012 RWQC: 0.44 logio CFU/100 mL for enterococci and 0.40 logio CFU/100 mL
for E. coli. This example uses enterococci as the indicator with a logio SD of 0.44 logio CFU/100 mL. A
key for the annotated code is presented in Table 1-1.
Table 1-1. Key for annotated code.
KEY
BOLD^Actual R code
(R available at http://www.r-proiect.org/)
PLAIN—Comments for guidance
#!/usr/bin/Rscript
####################################################################################
#
### Example code for STV (Statistical Threshold Value) and BAV (Beach Action Value) Calculation
### Includes EPA values for loglO standard deviation of enterococci and E. coli (within annotations)
### Include examples for single use. May be modified by analysts to process a larger batch of inputs.
### This script assumes the input geometric means and loglO standard deviation of the indicator are
specified in units ofCFU/lOOmL
####################################################################################
#
#### Functions to compute STV and BAV
z90<-qnorm(0.90) # the z score corresponding to the 90th percentile of a normal distribution
z75<-qnorm(0.75) # the z score corresponding to the 75th percentile of a normal distribution
compute_stv<-function(GM,loglOsd){ # function to compute the STV
stv_val<- 10A(logl0(GM)+logl0sd*z90)
st v_va I <-ro u n d (st v_va 1,0)
return(stv_val)
}
compute_bav<-function(GM,loglOsd){ # function to compute the BAV
bav_val<-10A(logl0(GM)+logl0sd*z75)
bav_val<-round(bav_val,0)
return(bav_val)
}
1-1
-------
#### EXAMPLES OF USE FOR A SINGLE GM
GM<-35 # user must input here the geometric mean of the indicator in units of CFU/lOOmL
Iogl0sd_indicator<-0.44 ## user must input here the loglO standard deviation (standard deviation of
the loglO transformed indicator data) of the indicator in units of loglO CFU/lOOmL or use the values
from EPA RWQC 2012 of 0.44 loglO CFU/lOOmL for enterococci or 0.4 loglO CFU/lOOmL if using E. coli
as indicator
stv<-compute_stv(GM,loglOsd_indicator) # call function to compute stv
bav<-compute_bav(GM,loglOsd_indicator) # call function to compute bav
print (c("The calculated STV value is round(stv,0)," CFU/100mL")) ## Print the calculated STV
print (c("The calculated BAV value is round(bav,0)," CFU/100mL")) ## Print the calculated BAV
1-2
-------
Appendix J: Computer Code for Reverse QMRA
This appendix provides an annotated code example for replicating Soller et al. (2014) reverse QMRA.
This code was used to compute the values shown in TSM Section 4.2.1. The code estimates the risk-
based threshold for the enterococci level corresponding to a target illness rate for the following fecal
mixtures: swine (pig)/human; chicken/human; gull/human; and nonpathogenic/human. This code can
be utilized by experienced programmers after a careful reading of its syntax and structure with
reference to the appropriate equations described in Soller et al. (2014) and parameters described in
Soller et al. (2010b) as cited in Soller et al. (2014). This code should not be considered a tool. Note that
the code outputs RBTs in units of CFU/100 mL. A key for the annotated code is presented in Table J-l
and the definitions and sources for variables for the Soller et al. (2014) reverse QMRA is presented in
Table J-2.
Citations:
Soller, J.A., Schoen, M.E., Bartrand, T., Ravenscroft, J.E., Ashbolt, N.J. 2010b. Estimated human health
risks from exposure to recreational waters impacted by human and non-human sources of
faecal contamination. Water Research, 44(16):4674-4691.
Soller, J.A., M.E. Schoen, A. Varghese, A.M. Ichida, A.B. Boehm, S. Eftim, N.J. Ashbolt, and J.E.
Ravenscroft. 2014. Human health risk implications of multiple sources of faecal indicator
bacteria in a recreational waterbody. Water Research 66:254-264.
Teunis, P., W. Van den Brandhof, M. Nauta, J. Wagenaar, H. Van den Kerkhof, and W. Van Pelt. 2005. A
reconsideration of the Campylobacter dose-response relation. Epidemiology and Infection
133(4):583-592.
Teunis, P., W. Van den Brandhof, M. Nauta, J. Wagenaar, H. Van den Kerkhof, and W. Van Pelt. 2005. A
reconsideration of the Campylobacter dose-response relation. Epidemiology and Infection
Table J-l. Key for annotated code.
KEY
BOLD^Actual code for Python
(software available at http://www.python.org/getit/)
PLAIN—Comments for guidance
J-l
-------
Table J-2. Soller et al. (2014) reverse QMRA: definitions and sources for variables in equations.
In-text variable
Code variable
X represents the pathogen of interest
Definition
rNHS rHS
No explicit variable defined.
Estimated waterbody density of reference pathogen
Computation embedded within the dose
(number of pathogens or genomes/L) in nonhuman and
function for program efficiency.
human sources.
-NHS
Pfib
pcfu
Proportion of FIB density attributable to the nonhuman
source NHS (unitless; evaluated here as a vector
assuming multiple values)
Cfib
fib_dens
Initial estimate of critical FIB density (CFU/L)
corresponding to the target probability of illness for
each proportion.
d NHS
t\rp
xx_dens ['PathogenName']
Density of reference pathogens (number of pathogens
or genomes/L or number of pathogens or genomes/g)
in the nonhuman source NHS
p NHS
kFIB
xx_dens['ent']
Density of FIB (CFU/L or CFU/g) in the nonhuman
source NHS
praw
t\rp
hum_raw_dens['nor']
Density of each reference pathogen (number of
pathogens or genomes/g) in raw human contamination;
dWWTP
l\rp
hum_wwtp_dens['nor']
Density of each reference pathogen (number of
pathogens or genomes/L) in WWTP-treated human
contamination;
praw
kFIB
hum_raw_dens['ent']
Density of FIB (CFU/g) in raw human contamination
nWWTP
kFIB
hum_wwtp_dens['ent']
Density of FIB (CFU/L) in WWTP-treated human
contamination
..NHS ..HS
r-rp ' r-rp
dos e['fib_dens',' PathogenName']
Estimated dose ingested of reference pathogen from
nonhuman and human sources respectively (number of
pathogens). Common implementation in dose function.
V
voljng
Volume of water ingested (L)
~NHS
Hrp
s w_f r i r\f['Pathogen Name']
Fraction of human infectious pathogenic strains
(unitless)
i NHS
'rp
sw_pr e\/['Pathogen Name']
Prevalence of infection in the nonhuman source
(unitless)
rp
Pill\inf
drp_X['c_pill']
Proportion of infections resulting in illness (unitless)
frp
drf_bp,drf_exp,drf_hyp. Dose-response
Dose-response functions for various pathogens
parameters for these functions are
found in drp_X.
(unitless)
prp.NHS prp,HS
ill ' ill
pill_X. (No explicit superscripting was
Probability of illness from a specific reference pathogen
required for NHS and HS because the
rp originating from the nonhuman source NHS and
pathogens found in the two sources did
human source HS, respectively (unitless)
not overlap.)
pNHS pHS
"ill > "ill
Directly implemented in tpill function
Estimated probability of illness from the nonhuman
below.
source NHS and human source HS, respectively
(unitless)
J-2
-------
In-text variable
Code variable
X represents the pathogen of interest
Definition
Pill
tpill
Estimated total probability of illness accounting for
exposure to both the nonhuman and human sources of
contamination (unitless)
# coding: utf-8
mill
Created on Thu Apr 28 09:41:12 2022
MINI
#!/usr/local/bin/python
#########################################################################################
### This is code replicating Soller et al. (2014) "reverse" QMRA. Determines FIB density
### corresponding to specified level of risk for contamination mixtures deriving
### from human and alternative animal (swine, chicken, gull, non-pathogenic) sources.
#########################################################################################
from numpy import * ### Imports Python NumPy module
from scipy.optimize import fsolve ### Imports necessary sub-module from Python SciPy module
from scipy.special import hyplfl ### Imports necessary sub-module from Python SciPy module
print ("REVERSE QMRA EXAMPLE FOR ANIMAL AND HUMAN MIXTURES") ## Prints output message
print("")
print( "FINDS FIB DENSITY CORRESPONDING TO SPECIFIED LEVEL OF RISK") ## Prints output message
print("")
risk=0.036 ### Target illness rate from the 2012 RWQC, specified as a probability.
### In this case, the chosen level of risk corresponds to 36 illnesses per 1000 swim events (unitless).
vol_ing=0.019 ### Volume of water ingested during the recreational event, in L General population median value from
###### Dufour et al. 2017.
### The following lines create arrays of FIB and pathogen densities found in swine, chicken and gull contamination
### based on the midpoint of the range of values in Soller et al. (2010b) as cited in Soller et al. (2014)
### Array of FIB (cfu/g) and pathogen (organisms/g) densities found in swine contamination
sw_dens={,ent':10**((5.3+7.2)/2.0),,cj,:10**((2.0+5.7)/2.0),,g,:10**((0+6.8)/2.0),,s,:10**((2.8+4.9)/2.0),,cr,:10**((1.7+3.6)
/2.0),'c57':10**((0.0+7.0)/2.0)}
### Array of FIB (cfu/g) and pathogen densities (organisms/g) found in chicken contamination
ch_dens={'ent':10**((5.0+7.0)/2.0),'cj':10**((2.8+6.5)/2.0),'g':0,'s':10**((-1.0+4.5)/2.0),'cr':0,'c57':0}
### Array of FIB (cfu/g) and pathogen (organisms/g) densities found in gull contamination
gu_dens={'ent':10**((6.0+8.0)/2.0),'cj':10**((3.3+6.0)/2.0),'g':0,'s':10**((2.3+9.0)/2.0),'cr':0,'c57':0}
### Array of FIB (cfu/L) and pathogen (organisms/L) densities found in non-pathogenic contamination.
### Artificial density value used for enterococci density to avoid 0/0 error when calculating path::fib ratio.
np_dens={'ent':0.1,'cj':0,'g':0,'s':0,'cr':0,'c57':0}
### Nested dictionary of animal source pathogen and FIB densities.
dens_source={'swine,:sw_dens,,chicken':ch_dens,'gulls': gu_dens,'non-path':np_dens}
### The following lines create arrays of pathogen prevalence rates for swine, chicken and gull contamination
### based on the midpoint of the range of values in Soller et al. (2010b) as cited in Soller et al. (2014)
### Array of pathogen prevalence rates for swine contamination (unitless)
sw_prev={'cj':(0.46+0.98)/2,'g':(0.033+0.18)/2.0,'s':(0.079+0.15)/2.0,'cr':(0+0.45)/2.0,'c57':(0.001+0.12)/2.0}
### Array of pathogen prevalence rates for chicken contamination (unitless)
ch_prev={'cj':(0.57+0.69)/2,'g':0,'s':(0.0+0.95)/2.0,'cr':(0.06+0.27)/2.0,'c57':0}
J-3
-------
### Array of pathogen prevalence rates for gull contamination (unitless)
gu_prev={,cj':1.0,,g,:0,,s,:1.0,,cr,:0,,c57,:0}
### Array of pathogen prevalence rates for non-pathogenic contamination (unitless)
np_prev={,cj,:0.0,,g,:0,'s,:0.0,,cr,:0,,c57,:0}
### Nested dictionary of pathogen prevalence rates.
prev_source={,swine,:sw_prev,,chicken,:ch_prev,,gulls':gu_prev,,non-path,:np_prev}
### The following lines create arrays of pathogen infectious fraction estimates for swine, chicken and
### gull contamination based on Soller et al. (2010b) as cited in Soller et al. (2014)
### Array of pathogen infectious fraction estimates for swine contamination (unitless).
sw_frinf={,cj':0.833,,g,:0.833,,s,:0.5,,cr,:0.167,,c57,:0.833}
### Array of pathogen infectious fraction estimates for chicken contamination (unitless).
ch_frinf={,cj,:0.5,,g,:0,,s,:0.5,,cr,:0.167/c57,:0}
### Array of pathogen infectious fraction estimates for gull contamination (unitless).
gu_frinf={,cj,:0.167,,g,:0,,s,:0.167/cr,:0/c57,:0}
### Array of pathogen infectious fraction estimates for non-pathogenic contamination (unitless)
np_frinf={,cj':0.0,,g,:0,,s,:0.0,,cr,:0,,c57,:0}
### Nested dictionary of infectious fraction rates.
frinf_source={,swine,:sw_frinf,,chicken,:ch_frinf,'gulls,:gu_frinf,,non-path,:np_frinf}
### The following line creates an array of FIB and pathogen densities found in raw human contamination
### based on the midpoint of the range of values in Soller et al. (2010b) as cited in Soller et al. (2014)
### Array of FIB (cfu/L) and pathogen (organisms/L) densities found in raw human contamination
hu m_raw_dens={'ent,:10* *((5.8+8.0)/2.0),'nor':10**((3.0+6.0)/2.0)}
### Norovirus attenuation level achieved in WWTP plants based on the midpoint of the range of values in Soller et al.
(2010b) as cited in Soller et al. (2014)
wwtp_atten=10**((2.23+3)/2)
### The following line creates an array of FIB and pathogen densities found in WWTP-treated human contamination based
### on the midpoint of the range of values in Soller et al. (2010b) as cited in Soller et al. (2014) and the attenuation value
specified in the line above.
### Array of FIB (cfu/L) and pathogen (organisms/L) densities found in WWTP-treated human contamination
hum_wwtp_dens={'ent,:10**((0.5+2.7)/2.0),,nor,:hum_raw_dens[,nor,]/wwtp_atten}
### Ratio of pathogen to FIB levels in raw human contamination
ratio_hum_raw=hum_raw_dens[,nor,]/hum_raw_dens[,entl]
### Ratio of pathogen to FIB levels in WWTP-treated human contamination
ratio_hum_wwtp=hum_wwtp_dens[,nor,]/hum_wwtp_dens[,ent']
### Ratio of pathogen to FIB levels in mixture of raw and WWTPtreated human contamination. The fractions of raw and
WWTP
### contamination in the mixture were determined such that an ingestion volume of 0.019 L results in an illness rate
### of 36/1000 and an FIB level of 35 cfu/lOOmL.
ratio_dens_hmix=0.9901*ratio_hum_raw+0.0099*ratio_hum_wwtp
### Array specifies alternative levels of non-human source contribution of total FIB level in surface water. For
### instance, 0 corresponds to 100% human contribution; 0.1 corresponds to 10% non-human source and 90% human
### contribution, etc.
pcfu=array([0,0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9])
### Array contains an initial estimate of FIB levels in CFU/L corresponding to specified risk for the different
### percentages of the mixture defined in the line above.
fib_dens=array([100,100,100,100,100,100,100,100,100,100])
J-4
-------
### The function "dose" defined below computes the dose of the specified pathogen ingested during a swim event by
### estimating the % of FIB from the source in which that pathogen is found, multiplying it by the ratio of pathogens
### per FIB in that source, and then multiplying that by the volume of water ingested.
### This implementation is based on Soller et al. (2014) Equation 1.
def dose(fib_dens,PathogenName,AnimalSource):
if PathogenName=='nor':
result = (l-pcfu)*fib_dens*ratio_dens_hmix*vol_ing
else:
result =
pcfu*fib_dens*dens_source[AnimalSource][PathogenName]/dens_source[AnimalSource]['ent,]*prev_source[AnimalSou
rce][PathogenName]*frinf_source[AnimalSource][PathogenName]*vol_ing
return result
### Beta Poisson Dose Response Function based on Table 1 in Soller et al. (2014)
def drf_bp(a,b,c_pill,d):
return (l-(l+d/b)**-a)*c_pill
### Exponential Dose Response Function based on Table 1 in Soller et al. (2014)
def drf_exp (r,c_pill,d):
return(l-exp(-r*d))*c_pill
### 1F1 Confluent Hypergeometric Dose Response Function based on Table 1 in Soller et al. (2014)
def drf_hyp(alpha,beta,c_pill,d):
return(l-hyplfl(alpha,(alpha+beta),-d))*c_pill
### Dose Response Parameters for Campylobacter jejuni based on Table 1 in Soller et al. (2014) and Teunis dose dependent
### conditional probability of illness
drp_cj={'alpha,:0.024,,beta,:0.011,,nu,:3.63*10**-9,,r,:2.44*10**8}
### The following function computes the conditional probability of illness given infection for cj based on the
### Teunis dose dependent relationship
def c_pill_cj_dd(fib_dens,PathogenName,AnimalSource):
return (l-(l+drp_cj[,nu,]*dose(fib_dens,PathogenName,AnimalSource))**-drp_cj[,r'])
### Dose Response Parameters for salmonella based on Table 1 in Soller et al. (2014)
drp_s={,a,:0.3126/b,:2884/c_piN,:(0.17+0.40)/2.0}
### Dose Response Parameters for giardia based on Table 1 in Soller et al. (2014)
drp_g={,r,:0.0199/c_piN':(0.2+0.7)/2.0}
### Dose Response Parameters for Cryptosporidium based on Table 1 in Soller et al. (2014)
drp_cr={,r,:0.09, ,c_pill,:0.5}
### Dose Response Parameters for e-coli 0157:H7 based on Table 1 in Soller et al. (2014)
drp_c57={'a,:0.248,,b,:48.8,,c_piN,:(0.2+0.6)/2.0}
### Dose Response Parameters for norovirus based on Table 1 in Soller et al. (2014)
drp_nor={,alpha,:0.04/beta,:0.055,,c_piM,:0.6}
### The following functions are based on Soller et al. (2010b) Table 3 and Teunis et al. (2005)
### The function pill_cj computes the probability of illness from exposure to Campylobacter jejuni using the dose
### response function, dose response parameters and dose for that pathogen,
def pill_cj(fib_dens,animalsource):
return drf_hyp(drp_cj[,alpha,],drp_cj[,beta,],c_pill_cj_dd(fib_dens,,cj,,animalsource),dose(fib_dens,,cj,,animalsource))
### The function pill_s computes the probability of illness from exposure to salmonella using the dose response function,
### dose response parameters and dose for that pathogen,
def pill_s(fib_dens,animalsource):
return drf_bp(drp_s[,a,],drp_s[,b,],drp_s[,c_piN,],dose(fib_dens,,s,,animalsource))
### The function pill_g computes the probability of illness from exposure to giardia using the dose response function,
J-5
-------
### dose response parameters and dose for that pathogen,
def pill_g (fib_dens,animalsource):
return drf_exp(drp_g[,r,],drp_g[,c_piM,],dose(fib_dens,,g,,animalsource))
### The function pill_cr computes the probability of illness from exposure to Cryptosporidium using the dose response
### function, dose response parameters and dose for that pathogen,
def pill_cr (fib_dens,animalsource):
return drf_exp(drp_cr[,r,],drp_cr[,c_piM,],dose(fib_dens/cr,,animalsource))
### The function pill_c57 computes the probability of illness from exposure to e-coli 0157:H7 using the dose response
### function, dose response parameters and dose for that pathogen,
def pill_c57(fib_dens,animalsource):
return drf_bp(drp_c57[,a,],drp_c57[,b,],drp_c57[,c_piN,],dose(fib_dens,,c57,,animalsource))
### The function pill_nor computes the probability of illness from exposure to norovirus using the dose response
### function, dose response parameters and dose for that pathogen,
def pill_nor(fib_dens,animalsource):
return drf_hyp(drp_nor[,alpha,],drp_nor[,beta,],drp_nor[,c_piM,],dose(fib_dens/nor,,animalsource))
### The function tpill computes the probability of illness from exposure to all the pathogens listed above.
### This implementation is based on Soller et al. (2014) Section 2.1.4.
def tpill(fib_dens,animalsource):
return(l-(l-pill_cj(fib_dens,animalsource))*(l-pill_s(fib_dens,animalsource))*(l-pill_g(fib_dens,animalsource))*(l-
pill_cr(fib_dens,animalsource))*(l-pill_c57(fib_dens,animalsource))*(l-pill_nor(fib_dens,animalsource)))
### The function deltarisk estimates the difference between the target risk level and the current value computed
### using an initial estimate for fib_dens
def deltarisk (fib_dens,animalsource):
return (risk-tpill(fib_dens,animalsource))
### The following lines define the different types of animal source available in this code.
animalsourcel=(,swine')
animalsource2=('chicken')
animalsource3=(,gulls')
animalsource4=('non-path')
### The following lines cause the initial estimate for fib_dens to be adjusted until deltarisk equals zero.
### In other words, it finds the FIB density levels that correspond to the target risk level for different
### mixtures of swine and human contamination.
fib_criticall=fsolve(deltarisk, fib_dens, args=animalsourcel)
fib_critical2=fsolve(deltarisk, fib_dens, args=animalsource2)
fib_critical3=fsolve(deltarisk, fib_dens, args=animalsource3)
fib_critical4=fsolve(deltarisk, fib_dens, args=animalsource4)
fib_critical4[fib_critical4 < 0] = nan # replace negative values in 100% non pathogenic contamination with NA
print("")
print("THE FIB RBTs in CFU/lOOmL CORRESPONDING TO THE TARGET RISK AT DIFFERENT LEVELS OF SWINE
CONTRIBUTION TO THE ANIMAL-HUMAN CONTAMINATION MIXTURE ARE: ") # Prints this message
print("")
print( pcfu, (fib_criticall/10).round(0))
print!"")
print! "THE FIB RBTs in CFU/lOOmL CORRESPONDING TO THE TARGET RISK AT DIFFERENT LEVELS OF CHICKEN
CONTRIBUTION TO THE ANIMAL-HUMAN CONTAMINATION MIXTURE ARE: ") # Prints this message
print!"")
print! pcfu, (fib_criticall/10).round(0))
print!"")
J-6
-------
print( "THE FIB RBTs in CFU/lOOmL CORRESPONDING TO THE TARGET RISK AT DIFFERENT LEVELS OF GULL
CONTRIBUTION TO THE ANIMAL-HUMAN CONTAMINATION MIXTURE ARE: ") # Prints this message
print("")
print( pcfu, (fib_criticall/10).round(0))
print!"")
print! "THE FIB RBTs in CFU/lOOmL CORRESPONDING TO THE TARGET RISK AT DIFFERENT LEVELS OF NON-PATHOGENIC
CONTRIBUTION TO THE ANIMAL-HUMAN CONTAMINATION MIXTURE ARE: ") # Prints this message
print!"")
print! pcfu, (fib_criticall/10).round(0))
### Postprocessing results. Generates output results as Excel file after appropriate transformations, transposing, and
###conversion to CFU/lOOmL to be consistent with TSM tables
import pandas as pd
hc=['10%', '20%', '30%','40%','50%', '60%','70%', '80%', '90%','100%'] # he is an array with percent human contribution to
fib but in reverse order compared to the nonhuman contribution array in the code above
### The lines below creates a dataframe using the arrays but reverses the order to correspond to he, rounds them, and
###divides by 10 to generate output in CFU/lOOmL
df=pd.DataFrame({'% Human FIB Contribution':hc,'Human-Swine Mixture':(fib_criticall[::-l]/10).round(0),'Human-
Chicken Mixture':(fib_critical2[::-l]/10).round(0),'Human-Gull Mixture':(fib_critical3[::-l]/10).round(0),'Human-Non-
pathogentic Mixture':(fib_critical4[::-l]/10).round(0)})
df=df.transpose()
df.to_excel(r'C:\YourOutputFolder\RBT_Simulation_Output_Soller2014_Method.xlsx',index=True,header=None)
######### QATO ENSURE CONVERGENCE
######### IF THESE VALUES DO NOT MATCH THE TARGET RISK, ADJUST THE INITIAL ESTIMATE FOR FIB_DENS UNTIL YOU
######### SEE CONVERGENCE.
print("Convergence check foranimalsourcel, tpill(fib_criticall,animalsourcel))
print("Convergence check foranimalsource2, tpill(fib_critical2,animalsource2))
print("Convergence check foranimalsource3, tpill(fib_critical3,animalsource3))
print("Convergence check foranimalsource4, tpill(fib_critical4,animalsonurce4))
J-7
------- |