Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Microbial Risk Analysis xxx (xxxx) xxxx

Contents lists available at ScienceDirect

Microbial Risk Analysis


journal homepage: www.elsevier.com/locate/mran

Refined ambient water quality thresholds for human-associated fecal


indicator HF183 for recreational waters with and without co-occurring gull
fecal contamination
A.B. Boehma, J.A. Sollerb,

a
Department of Civil & Environmental Engineering, Stanford University, Stanford, CA, United States 94305-4020
b
Soller Environmental, Berkeley, California, United States

ARTICLE INFO ABSTRACT

Keywords: A number of studies have used quantitative microbial risk assessment (QMRA) to derive risk-based water quality
Microbial risk assessment thresholds (RBTs) for a diverse set of novel water quality indicators in recreational waters. Over time, the QMRA
Recreational water approach has been refined with respect to model inputs and hazard characterization. Recent work considered
HF183 differential decay of pathogens and indicators, and mixtures of contamination of diverse ages. In the present
Alternative indicator
study, we present a refined ambient water quality RBT for the human-associated marker HF183. The new es­
timate updates previous work as it specifically considers contamination aging through the inclusion of tem­
perature-specific organism decay, and the presence of mixtures of human sewage contamination of diverse ages.
Based on these analyses we derived a RBT of 525 HF183 copies/100 mL as representative of conditions con­
sistent with those described in the 2012 Recreational Water Quality Criteria (32 illnesses /1000). In recreational
waters, microbial contamination due to gulls is also common. To account for this, we consider the case where
human contamination from sewage co-occurs with contamination from gull feces. The resultant proposed RBTs
for HF183 range from 1 to 525 copies/100 ml and are a function of the amount of gull fecal contamination that is
present in the water. The proposed RBTs can be considered for use in the evaluation of recreational water
quality.

1. Introduction health from recreational exposures as compared to fecal contamination


from other sources, such as gulls, whereas FIB sources can be diverse
The protection of public health is the cornerstone of recreational and widespread (Soller et al., 2010a; Soller et al., 2010b; Korajkic et al.,
water management worldwide (WHO, 2003; USEPA, 2012a). Histori­ 2018). It is also accepted that there can be site specific conditions
cally, epidemiological studies have been used to establish water quality where alternative indicators may yield better estimates of adverse
criteria values or recommendations to ensure public health protection health outcomes, depending on the sources of contamination to a par­
(Pruss, 1998; Zmirou et al., 2003; Wade et al., 2003). Most recreational ticular waterbody (USEPA, 2012b; USEPA, 2014a; USEPA, 2014b;
water epidemiological studies and subsequent water quality criteria or USEPA, 2015).
regulations for recreational waters have relied on fecal indicator bac­ To address the disparity between potential FIB contributions from
teria (FIB) such as fecal coliform or enterococci as indices of water wide-ranging sources and the most important contributors of risk to
quality, since those FIB have generally demonstrated good correspon­ human health, scientists have advanced the use of alternative in­
dence with adverse health effects, and are inexpensive and convenient dicators, such as molecular markers that are specific to the source of
to enumerate. Those water quality criteria and related regulations are contamination (Boehm et al., 2013; Shanks and Korajkic, 2020; Bae and
considered to be protective of public health under a wide variety of Wuertz, 2012; Staley et al., 2012; Symonds et al., 2016). These mole­
conditions (USEPA, 2012a; USEPA, 1986). However, the scientific un­ cular markers were originally developed for microbial source tracking
derstanding of risk to human health from recreational water exposure purposes, and have gained broad acceptance among water quality sci­
has evolved significantly over the last decade – it is now generally ac­ entists and water quality managers. Success with molecular markers
cepted that human fecal contamination poses the greatest risk to human within the context of microbial source tracking has led to interest in


Corresponding author.
E-mail address: jsoller@sollerenvironmental.com (J.A. Soller).

https://doi.org/10.1016/j.mran.2020.100139
Received 24 April 2020; Received in revised form 31 August 2020; Accepted 5 September 2020
2352-3522/ © 2020 Elsevier B.V. All rights reserved.

Please cite this article as: A.B. Boehm and J.A. Soller, Microbial Risk Analysis, https://doi.org/10.1016/j.mran.2020.100139
A.B. Boehm and J.A. Soller Microbial Risk Analysis xxx (xxxx) xxxx

their wider use as potential water quality management tools. Their use inclusion of temperature-specific organism decay, and considering a
could allow remediation efforts to focus on sources of fecal con­ scenario where human contamination from sewage co-occurs with
tamination that are the greatest source of risk to human health. The contamination from gull feces. These new scenarios were considered in
Recreational Water Quality Criteria in the United States (USEPA, the present study to address a growing desire from the regulatory
2012a), are now based on specific levels of human health protection (32 community for guidance on the use of alternative fecal indicators for
or 36 illnesses per 1000) rather than specific levels of FIB, as was the monitoring bathing water quality for the protection of swimmer health.
case previously (USEPA, 1986). Therefore, there is now the possibility
for a site-specific, cost-effective, and health protective approach to
2. Methods
water quality management (Given et al., 2006) that focuses on specific
sources of contamination of critical interest rather than on generic FIB
Exposure to untreated, raw sewage of known age: HF183. A
(Environmental Incentives and EcoNorthwest 2017). In addition to the
static QMRA was used to estimate gastrointestinal illness risk from
potential use of a viral indicator (USEPA, 2015), molecular markers are
swimming in surface waters with varying concentrations of HF183 from
another promising tool that can be used to help facilitate and imple­
untreated sewage of different ages using R (Version 1.1.463). The in­
ment this new approach.
fluence of immunity and secondary transmission was not considered in
Boehm et al. (2015) developed quantitative estimates of the risk
the models (Soller and Eisenberg, 2008). In the QMRA, HF183 serves as
associated with specific levels of molecular markers in recreational
an index for the amount of sewage present in surface water. The QMRA
waters. They used quantitative microbial risk assessment (QMRA) to
considers the cumulative risk from exposure to reference pathogens
simulate the risk of gastrointestinal (GI) illness associated with re­
adenovirus (not considered by Boehm et al. (2015)), norovirus, Giardia,
creating in waters containing different concentrations of human-asso­
Cryptosporidium, E. coli O157:H7, Campylobacter, and Salmonella as re­
ciated fecal markers (HF183 and HumM2) from raw sewage. HF183
commended by USEPA and used extensively in bathing water QMRAs
was of particular interest in that study and is currently considered to be
(Soller et al., 2010a; Soller et al., 2017a; Soller et al., 2015d;
the state-of-the-art human molecular marker because it has been eval­
Schoen and Ashbolt, 2010).
uated in a number of studies and been shown to be highly specific to
The volume fraction of raw sewage in surface water (Vs) serves as an
human feces with little to no cross reactivity with other host feces
input to the model and varies between 10−8 and 1 in increments of 0.5
(Boehm et al., 2013; Layton et al., 2013; Shanks et al., 2009;
log10 units. The age of the contamination (the time it has spent in
Shanks et al., 2010; Griffith et al., 2003). Further justification for the
surface water after being released from an untreated, raw sewage
use of HF183 in this regard is provided by a recent review article which
source) is τ. τ serves as a model input and varies between 0 (unaged)
summarized the sensitivity and specificity metrics for the HF183 assay -
and 15 days (d) in 0.5 d increments. A total of 527 Vs-τ combinations
a collective sensitivity of 83.1% calculated from 1242 human waste
(31 distinct ages and 17 distinct Vs) were modeled. The increment size
samples and specificity of 94.6% based on a collection of 2966 non­
for Vs and τ were chosen to be fine enough to elucidate trends in the
human fecal samples was reported (Ahmed et al., 2016). Additionally, a
results without requiring excessive computational time; choosing dif­
national study from the US reported 100% sensitivity of HF183 qPCR
ferent increment sizes did not affect the results (data not shown). The
DNA target determined from 54 sewage samples collected from 39
choice of 15 days for the largest age was arbitrary and chosen to be
different states (Shanks et al., 2010). Albeit rare, the HF183 DNA target
large enough to illustrate the effect of aging.
has been reported in a small proportion of chicken, dog, gull, and deer
After Vs and τ are specified as model inputs, the concentration of ith
reference fecal samples almost always at concentrations considerably
reference pathogen in surface waters (Ci_surface) is modeled as follows:
lower than levels observed in human waste (Shanks and Korajkic, 2020;
Nguyen et al., 2018; Ahmed et al., 2012). Ci _ surface = Vs Ci _ sewage e ki
(1)
Since the Boehm et al. (2015) study, a number of other studies have
used the QMRA approach to identify concentrations for human-asso­ where Ci_sewage and ki are, respectively, the ith reference pathogen
ciated genetic markers that correspond to defined public health risk concentration in sewage, and ki is the ith reference pathogen's first order
benchmarks for sewage-impacted recreational water (Zhang et al., decay rate constant. In Eq. (1), Ci_sewage are described by distributions
2019; Boehm et al., 2018; Ahmed et al., 2018). These risk-based (Table 1) that were obtained from the literature and have been used in
threshold estimates (RBT) are valuable research tools to compare the previous QMRA studies (Brown et al., 2017b; Soller et al., 2017b;
utility of various indicators to each other and across different experi­ Boehm, 2019).
mental scenarios based on a set of defined model assumptions. A recent Two systematic reviews and meta-analysis of decay rate constants in
review review indicates that a diverse set of novel water quality in­ surface waters indicated that decay rate constants of allochthonous
dicators have been used for this purpose under ever broadening en­ microorganisms in surface waters are a function of water temperature
vironmental conditions (Zhang et al., 2019). Recent work has con­ (Boehm et al., 2018; Boehm et al., 2019). Those meta-analyses in­
sidered differential decay of pathogens and indicators and mixtures of dicated little evidence that salinity or nutrient content of the water
contamination of diverse ages. For example, Boehm et al. (2018) ex­ significantly affected decay rate constants, so those factors are not
tended their earlier work to consider aging of contamination; considered herein. Here we considered ki values at 20°C. This tem­
Brown et al. (2017a) developed RBTs for gull-associated fecal marker perature represents an annual average temperature of waters off San
concentrations for the case that gull feces were not aged in the en­ Diego, CA based on data obtained from Scripps pier (data not shown).
vironment; Brown et al. (2017b) considered mixtures of un-aged human An annual average temperature was chosen for this work to identify
and gull-associated contamination and marker levels; and Crank et al. risk-based thresholds that would be appropriate any time of year.
(2019) developed RBTs with exposure to the crAssphage human marker Previous work compiled raw surface water decay rate constants for
and pepper mild mottle virus from fresh, unaged sewage. Wu et al. HF183 and the reference pathogens used in this study for surface waters
(2020) considered the risk associated with exposure to feces with a and noted the experimental conditions, including temperatures, at
range of ages. which the rate constants were derived (Boehm et al., 2018;
In the present study, we present refined ambient water quality RBTs Boehm et al., 2019). Those systematic reviews only included results
for human-associated HF183 and the gull and pigeon-associated marker from experiments conducted with raw water and excluded experiments
Catellicoccus marimammalium (Sinigalliano et al., 2013; Lee et al., that were conducted with sterilized water or artificial waters; therefore,
2013). The new estimates update previous work by adding adenovirus the decay rate constants represent estimates of decay in real environ­
to the previous reference pathogens for human-associated fecal con­ mental waters. We derived distributions for the rate constants needed in
tamination, specifically considering contamination aging through the Eq. (1) (ki) for temperatures of 20 °C using the methods outlined in

2
A.B. Boehm and J.A. Soller Microbial Risk Analysis xxx (xxxx) xxxx

probability of infection, Pill|inf is probability of becoming ill after infection. Note that units of pathogens are per liter and for indicators is per ml to reflect the units used in the literature for these parameters. CFU is colony
forming unit, MPN is most probable number, copy refers to gene copy number, IU is infectious unit, and PFU is plaque forming unit. 1F1 is the hypergeometric function. When specified, Pill|inf are represented by a range of
parameters, as indicated, drawn from a uniform distribution. Pill|inf for Campylobacter is dose-dependent with r = 2.44 × 108 and n = 3.63 × 10−9. References (Refs) for Pinf and Pill|inf are provided in the last column.
Untreated sewage concentrations for reference enteric pathogens and indicators, and dose-response relations, and Pill|inf for reference enteric pathogens. Unit (refs) is the concentration in sewage, µ is the dose, Pinf is
Table 2
Normal distributions of log10-transformed k for the organisms and indicators

Teunis et al., 2008; Ludwig et al., 2002; Werber et al., 2008;


used in the QMRA. Mean and standard deviation (SD) are provided. Units of k
before they were log10-transformed are d− 1. CAT k is described by a uniform
distribution bounded by 0.11 and 3.5 per day, as described in the text.

Haas et al., 1999; Fazil, 1996; Teunis et al., 1999


Organism/ mean SD

Rose and Gerba, 1991; Eisenberg et al., 1996


Indicator log10-k log10-k

Salmonella spp. −0.13 0.058


Campylobacter 0.45 0.16
E. coli O157:H7 −0.36 0.044
Cryptosporidium −1.01 0.16
Giardia −1.59 0.21
Bielaszewska et al., 1997

norovirus −1.01 0.14


adenovirus −0.34 0.15
Teunis et al., 2005

Teunis et al., 2008


Teunis et al., 2016
HF183 0.063 0.34
enterococci 0.068 0.49
US EPA 2006

Boehm (2019). Briefly, a statistical model relating ki to temperature was


Refs

used to generate predictions for ki at the chosen temperature. The


prediction includes a mean and variance describing a predicted dis­
Pill|inf (distribution)

0.17–0.4 (uniform)

tribution. ki were subsequently, randomly drawn from their respective


0.2–0.6 (uniform)

0.2–0.7 (uniform)
0.3–0.8 (uniform)
0.3–0.7(uniform)

distributions (Table 2). Note that these ki values are estimates for the
0.5 (uniform)
1-(1+nµ)−r

pathogens as enumerated using culture-based methods (Boehm, 2019).


In deriving Eq. (1), it is assumed that the advection and dispersion of
indicator and reference pathogens are identical, and any non-con­
NA
NA

servative behavior of water quality targets is adequately captured by


1- 1F1 (0.04, 0.04+ 0.055, -µ)

first order kinetics.


1–1F1(0.024,0.024+0.011,-µ)

1- 1F1 (5.11, 5.11+ 2.8, -µ)

It is assumed that the volume (V) of water ingested by a swimmer


per swimming event follows the log10-normal distribution with a mean
−0.3126

1-(1+ µ/48.8)−0.248

1 - exp(−0.0199 µ)

of 1.20 and standard deviation of 0.68; units of V are ml (DeFlorio-


1 - exp(−0.09 µ)
1-(1+ µ/2884)

Barker et al., 2018). The dose of pathogen i, µi, is given by Ci_surfaceV.


The dose was used as input to the reference pathogen dose-response
functions (Table 1) to determine the probability of infection (Pinf_i). The
The two values separated by a comma are the mean and standard deviation of a log10-normal distribution.
The two values separated by a comma are the minimum and maximum of the log10-uniform distribution.
Pinf

probability of illness (Pill_i) was calculated by multiplying the prob­


NA
NA

ability of infection by the probability of illness given infection Pill|inf_i


(Table 1). Pill|inf_i was randomly drawn from a uniform distribution for
each model iteration except for the case of Campylobacter which used a
dose-dependent formula (Table 1). The cumulative risk of illness from
oocysts/L (Jian et al., 2015; Crockett, 2007; Harwood et al., 2005;

Lower range is not detected and −1 is used as a lower bound. NA means not applicable.
IU/L Soller et al., 2017b; Hurst et al., 1988; Hewitt et al., 2011
CFU/L (Koivunen et al., 2003; Lemarchand and Lebaron, 2003)

exposure to all reference pathogens (Pill) is given by

Pill = 1 (1 Pilli)
(2)
cysts/L (Harwood et al., 2005; Kitajima et al., 2014)

It was assumed that infection and illness for each pathogen is in­
References for sewage concentration range are provided adjacent to the unit.

dependent. 50,000 iterations were obtained for each Vs-τ combination


(Staley et al., 2012; Sinigalliano et al., 2013). The median, interquartile
range, and 10th and 90th percentiles of Pill for each Vs -τ combination
Nasser, 2016; Schoen et al., 2017)
CFU/L (García-Aljaro et al., 2004)

were calculated from the respective 50,000 iterations. Pill was com­
copies/ml (Shanks et al., 2010)

pared to the value 32/1000 which is equal to one of the key risk
CFU/ml (Soller et al., 2010a)
MPN/L (Stampi et al., 1993)

copy/L (Eftim et al., 2017)

thresholds published by USEPA for bathing waters (USEPA, 2012). For


each τ, the Vs for which the median simulated risk was 32/1000 was
determined – this value of Vs is hereafter referred to as “risk-based
threshold Vs”. Note that Vs is independent of the indicator used to
measure fecal contamination.
Unit (refs)

The concentration of HF183 in the surface water when the volume


fraction of sewage is equal to risk based threshold Vs and aged a time τ
was determined as follows:
[5.212, 0.566]^
#

[1.75,3.84] #
[−0.52, 3.7]
[−1,3.3] #,*

(3)
CHF183 = Vs CHF183_ sewage e k HF183
#
[2.9,4.6] #

[4.0,1.1] ^
[0.51,4.2]
Ci_sewage

[2.8, 5]#
#

where CHF183_sewage and kHF183 are, respectively, the concentration of


[0.5,5]

HF183 in sewage and the decay rate constant of HF183 in surface


waters, and Vs is the risk based threshold Vs. In Eq. (3), CHF183_sewage is
Organism/Target

described by a distribution (Table 1), and kHF183 is also described by a


E. coli O157:H7

Cryptosporidium
Salmonella spp.
Campylobacter

distribution obtained using the same methods as described for ki


enterococci
adenovirus
norovirus

(Table 2). To find CHF183, values were drawn from the distributions
Giardia
Table 1

HF183

50,000 times and the median CHF183 as well as other percentiles de­
#

scribing the resultant distribution were recorded. The median of this


^

3
A.B. Boehm and J.A. Soller Microbial Risk Analysis xxx (xxxx) xxxx

distribution is referred to as the “risk-based threshold for HF183”. Table 3


Exposure to untreated sewage of unknown age: HF183. The age Concentrations of Salmonella, Campylobacter, and CAT in gull feces.
of sewage in surface water is usually unknown and furthermore, surface Organism/indicator C_i_feces units ref
water contamination may represent a mixture of contamination of di­
verse ages. We repeated the QMRA for a scenario where the indicator
#
Salmonella [2.3,9] CFU/g (Lévesque et al., 2000)
Campylobacter [3.3,6]# CFU/g (Lévesque et al., 2000)
HF183 is measured in surface waters, and the contamination age is
CAT [8.26,8.73]* copies/g (Brown et al., 2017a)
unknown. We used the same methods as described by Boehm et al.
(2018; 2019). The concentration of HF183 as measured in surface water #
The two values separated by a comma are the minimum and maximum of
(CHF183) was specified as an input to the model and varied between 100 the log10-uniform distribution.
and 105 copies/100 ml in increments of 0.2 log10 units. The con­ ⁎
The two values separated by a comma are the shape and scale of a Weibell
centrations of pathogens in surface waters (Ci_surface) was modeled as distribution where the concentrations were log10-transformed prior to being fit
follows: to the distribution. CAT was measured using the LeeSeaGull assay (Lee et al.,
2013).
CHF183 Ci_ sewage k
Ci _ surface = e
CHF183_ sewage (4) possible amount of feces. The age of the contamination τ serves as a
model input and varies between 0 (unaged) and 7 d in 0.5 d increments.
where Δk = kHF183 - ki, and Ci_sewage and CHF183_sewage are, respectively, A total of 315 Mb -τ combinations (15 distinct ages and 21 distinct Mb)
the ith reference pathogen and HF183 concentrations in sewage, and ki were modeled.
and kHF183 are their first order decay rate constants. These variables After Mb and τ are specified as model inputs, the concentration of ith
were all drawn from their respective distributions (Tables 1 and 2). τ reference pathogen in surface waters (Ci_surface) is modeled as follows:
was drawn from a uniform distribution ranging from 0 to a maximum
(6)
Ci _ surface = Mb Ci _ feces e ki
realistic value (τmax) given the specified Cmeas:
th
where Ci_feces and ki are, respectively, the i reference pathogen con­
max = 1/m [log10Cmeas b] (5)
centration in gull feces, and ki is the ith pathogen's first order decay rate
where b is the log10-transformed median concentration of HF183 in raw constant, both described by distributions (Tables 2 and 3). The dose of
sewage (CHF183), and m is -kmed*log10e where kmed is the median HF183 pathogen i (µi) was subsequently determined by the following equation:
decay rate constant (kHF183) in surface waters (Table 4). Further con­ µi = Ci _ surface Vpi (7)
ceptual discussion of τmax can be found in Boehm (2019). Ci_surface was
used to determine dose, the dose was input into the dose-response where V was drawn from the same distribution described in the pre­
equations, and then Pill was determined, as described in the previous vious sections, and pi represents the fraction of that pathogen from gull
section. feces that is human infectious. pi can vary between 0.01 and 0.4 and is
Model output, including calculation of the risk-based threshold, was drawn from a uniform distribution; as described in our previous work
handled as described in the previous section. Note that this model re­ (Brown et al., 2017a). The dose was subsequently used in the dose-
quires Cmeas to be input rather than Vs since τmax is dependent speci­ response equations and a cumulative illness risk (Pill) was determined.
fically on Cmeas (Eq. (5)). Model output, including calculation of the risk-based threshold for
Exposure to untreated sewage of unknown age: ENT. We re­ Mb, was handled as described in the previous sections.
peated the analysis described in the previous section, but instead of The concentration of CAT in the surface water (CCAT) when the mass
using HF183 as an index for the amount of sewage present in the sur­ concentration of feces is equal to the risk-based threshold Mb and aged a
face water, we used culturable enterococci (ENT). The modeling time τ was determined as follows:
method is the same except CENT was used in place of CHF183, kENT was
CCAT = Mb CCAT_ feces e k CAT
(8)
used in place of kHF183, and CENT_sewage was used in place of
CHF183_sewage in Eq. (4); and the m and b in Eq. (5) are those for ENT where CCAT_feces and kCAT are respectively, the concentration of CAT in
(Table 4). CENT_sewage and kENT are described by distributions (Tables 1 gull feces and the decay rate constant of CAT in surface waters, and Mb
and 2). kENT was extracted from the literature review completed by is the risk-based threshold Mb. CCAT_feces and kCAT are described by
Brooks and Field (2016). We downloaded their supplementary data of distributions (Tables 2 and 3). There are limited data on kCAT in the
compiled kENT. After removing suspected duplicates, and converting literature. Brown and Boehm (2015) report kCAT for seawater under
reported T90 and T99 values to k values, we log10-transformed the kENT dark, and intensely sunlit conditions. For the present study, we deemed
and took their mean and standard deviation to define a log-normal the kCAT reported by those authors for sunlit conditions would be un­
distribution for modeling (Table 2). We were unable to account for the realistic for field conditions – they would need to be corrected for flu­
effect of temperature using their data as the appropriate meta-data were ence and screening of photons by the water column to be applicable to a
not provided . field setting (Nelson et al., 2018). A well-mixed water column of a
Exposure to gull feces of known age: CAT. A molecular marker natural water is likey to have low incident UVA and UVB, the most
for a bacterium Catellicoccus marimammalium (CAT) (Sinigalliano et al., germicidal wavelengths of sunlight (Nelson et al., 2018), and thus may
2013; Lee et al., 2013; Lu et al., 2008) has proven to be a sensitive and be better approximated as dark rather than intensely sunlit. Therefore,
specific marker of bird feces (gull and pigeon) in surface waters we opted to use kCAT they report for dark conditions. We used the lower
(Boehm et al., 2013). We investigated the risk-based threshold for CAT estimate and upper estimate for dark kCAT as bounds on a uniform
in recreational waters assuming they are contaminated with gull feces distribution for kCAT for the QMRA model, 0.11 and 3.5 per day, re­
of a known age. This method mirrors that described above for in­ spectively.
vestigation exposure to raw sewage using HF183 as an indicator, except To find CCAT, values were drawn from the distributions 50,000 times
reference pathogens are Salmonella and Campylobacter (Soller et al., and the median CCAT as well as other distribution percentiles were re­
2010a; Brown et al., 2017a; Schoen and Ashbolt, 2010). This approach corded. The median of this distribution is referred to as the “risk-based
mirrors that used by Brown et al. (2017a) with slight modifications. threshold for CAT”.
The mass concentration of bird feces in surface water (Mb) serves as
an input to the model and varies between 10−8 and 100 g/100 ml in 2.1. Exposure to gull feces unknown age: CAT
increments of 0.5 log10 units. The upper limit of 100 g/ 100 ml implies
the water is ~ 100% gull feces by mass and represents a maximum We repeated the QMRA for a scenario where the CAT is measured in

4
A.B. Boehm and J.A. Soller Microbial Risk Analysis xxx (xxxx) xxxx

surface waters, and the gull fecal contamination age is unknown. The Table 5
methods are similar as described above for unknown age exposure The risk-based threshold for HF183 given CCAT. The values in all rows come
scenarios. The concentration of CAT as measured in surface water was from the line of constant risk in Fig. 6 of 32/1000. The value for the first row
specified as an input to the model and varied between 100 and 108 (when CCAT = 0) is from the simulation for HF183 when raw sewage con­
tributes exclusively to risk and the age is unknown (Fig. 2).
copies/100 ml in increments of 0.5 log10 units. The concentrations of
reference pathogens in surface waters (Ci_surface) was modeled as fol­ CCAT [copies / 100 ml] Risk-based threshold for HF183 [copies/100 ml]
lows:
0 525
CCAT Ci_ feces k 1 370
Ci _ surface = e
CCAT _ feces (9) 10 275
100 175
where Δk = kCAT - ki, and Ci_feces and CCAT_feces are, respectively, the ith 300 120
reference pathogen and CAT concentrations in feces, and ki and kCAT 1000 70
3000 30
are their first order decay rate constants. These variables were all drawn
10,000 7
from their respective distributions (Tables 2 and 3). τ was drawn from a 22,500 1
uniform distribution ranging from 0 to a maximum realistic value (τmax)
given the specified Cmeas as described in Eq. (5). Here b is the log10-
transformed median concentration of CAT in feces (Cfeces) multiplied by different CCAT and CHF183) was extracted and used to create a contour
a maximum realistic mass concentration of feces in water Mbmax, and m plot illustrating median simulated risk as a function CCAT and CHF183.
is -kmed*log10e where kmed is the median CAT decay rate constant (kCAT) The data were interpolated to identify the risk contour representing 32/
in surface waters (Table 4). We assumed Mbmax was 100 g/ 100 ml. 1000 to elucidate the risk-based threshold for HF183 as a function of
Ci_surface was used to determine dose, accounting for pi (as in Eq. (7)), CAT concentration.
the dose was input into the dose-response equations, and then Pill was
determined. 3. Results
Model output, including calculation of the risk-based threshold, was
handled as described in the previous sections. Note that this model Exposure to untreated, raw sewage of known and unknown
requires Cmeas to be input rather than Mb since τmax is dependent spe­ age. The risk-based volume fraction of sewage (Vs) increases with the
cifically on Cmeas (Eq. (5)). age of contamination (Fig. 1) because pathogen concentrations de­
crease as the sewage ages. In contrast, the risk-based threshold for
2.2. Exposure to a mixture of gull feces and sewage of unknown ages: HF183 declines with the age of contamination, as reported previously
CATand HF183 (Boehm et al., 2018). The decline occurs because the decay rate con­
stant of norovirus, on average, is smaller than that of HF183 and even
This QMRA model estimates risk from swimming in surface waters with the addition of adenovirus in these simulations, norovirus con­
containing a mixture of gull feces and sewage, measured using CAT and tinued to dominate risk (Figure S1). When sewage is fresh (unaged,
HF183, respectively, where the ages of gull and raw sewage con­ τ = 0), the risk-based threshold is approximately 9000 copies/100 ml.
tamination are both unknown and independent. The concentrations of After approximately 9 days, the risk-based threshold declines to ~ 1
CAT and HF183 measured in surface waters are used as inputs to the copy/100 ml, which is at a possible, but very sensitive, limit of quan­
model. CCAT is varied from 100 to 108 copies/100 ml, and CHF183 is tification for the assay. This implies that after about sewage has aged
varied from 100 to 105 copies/100 ml, both in 0.5 log10 units resulting approximately 9 days, the risk-based threshold for HF183 is lower that
in 187 different simulations that combined the different CAT and what can be readily measured in surface waters. As indicated above,
HF183 concentrations. As described above, each simulation included norovirus is the pathogen that contributes the most to risk at all ages, as
50,000 Monte Carlo iterations. reported in previous studies (Boehm et al., 2015; Boehm et al., 2018).
The concentrations of pathogens in surface waters from sewage When the age of contamination is unknown per the scenario de­
Ci_surface_sewage are computed using Eq. (4) where τ is drawn from a scribed in the methods, the risk-based threshold for HF183 is 525 co­
uniform distribution between 0 and τmax_HF183. The concentration of pies/100 ml (Fig. 2). If the risk-based threshold is defined as where the
pathogens in surface waters from gull feces Ci_surface_feces is calculated 55th or 45th percentile of risk is equal to 32/1000, then the threshold
using Eq. (9) where τ is drawn from a uniform distribution between 0 would be 240 or 1000 copies/100 ml, respectively. These alternative
and τmax_CAT. τmax_HF183 and τmax_CAT are computed using Eq. (5) using percentiles are chosen arbitrarily; any percentile could be chosen by a
the indicator specified m and b (Table 5) described in the previous policy maker depending on their needs. This scenario assumes that the
sections for the non-mixture exposure scenario for HF183 and CAT, age of contamination could be any number between 0 and τmax where
respectively. The doses of Cryptosporidium, Giardia, E. coli O157:H7, τmax is the oldest age contamination could be in surface waters. As for
norovirus and adenovirus are computed as µi= Ci_surface_sewageV as there the known age scenario, exposure to norovirus accounts for the ma­
is no contribution from gull feces. The doses of Salmonella and Cam­ jority of the risk (Boehm et al., 2018).
pylobacter are calculated as µi= (Ci_surface_sewage + Ci_surface_fecespi) V as We repeated the QMRA to calculate risk of illness from exposure to
there are contributions from both pathogen sources. The doses are used ENT from raw sewage of unknown age to evaluate the approach and
to calculate cumulative risk (Pill) as described above. assumptions used for the HF183 analysis. The risk-based threshold for
The median Pillof each of the simulations (each representing a ENT is calculated to be 47 CFU/100 ml (Fig. 3). This number is close to
the USEPA surface water criterion for ENT (at the same level of health
Table 4 protection) of 32 CFU/100 ml (USEPA, 2012a).
m and b for calculating tmax via Eq. (5). These values presume that the units
ofτmax are days, and the unit of CHF183, CENT, and CCAT in Eq. (5) are per 3.1. Exposure to gull feces known and unknown age: CAT
100 ml.
Indicator b m The risk-based mass concentration of gull feces (Mb) increases with
the age of contamination (Fig. 4) because associated pathogen con­
HF183 7.2 −0.50
centrations decrease as the fecal contamination ages. The increase is
ENT 5.9 −0.51
CAT 10.35 −0.78 biphasic because initially, exposure to Campylobacter contributes to the
total risk (~60% of the risk, averaged all concentrations for t = 0 days),

5
A.B. Boehm and J.A. Soller Microbial Risk Analysis xxx (xxxx) xxxx

Fig. 1. Panel A. The risk-based threshold for Vs (volume fraction of sewage) as a function of the age of sewage contamination. Panel B. The risk-based threshold of
HF183 as a function of age of sewage contamination. The 45th and 55th percentiles are shown as well as the median to give a sense of variability in the model output.
The horizontal dotted line shows a sensitive lower limit of quantification of the HF183 assay for comparison purposes.

Fig. 2. The median, 45th, and 55th percentiles of calculated risk for a measured
concentration of HF183 in surface waters when the age of contamination is Fig. 3. The median, 45th, and 55th percentiles of calculated risk for a measured
unknown, and thus assumed to be any possible age in the Monte Carlo simu­ concentration of ENT in surface waters when the age of contamination is un­
lations. The median risk in the figure is the median of 50,000 Monte Carlo known, and thus assumed to be any possible age in the Monte Carlo simula­
simulations for each CHF183. Reference line corresponds to 32/1000 illness tions. The median risk in the figure is the median of 50,000 Monte Carlo si­
level. mulations for each CENT. Reference line corresponds to 32/1000 illness level.

described in the methods (Fig. 5). Based on this analysis, the risk based
but after ~2.5 days, due to its relatively high decay rate constant threshold for CAT is 200,000 copies/100 ml. If the risk-based threshold
(Table 2, median 2.8 day−1), Campylobacter contributes little to the risk were redefined to be the CAT concentration at which the 55th or 45th
(average 25% or less of the total risk, averaged across concentrations), percentile of risk were equal to 32/1000, then the threshold would be
and most of the risk is then attributable to Salmonella. 100,000 or 400,000 copies/100 ml. Since the CAT RBT for gull fecal
The risk-based threshold for CAT is relatively constant, compared to contamination of known age is a relative weak function of age (Fig. 4),
that of HF183, with the age of contamination (Fig. 4). There is an initial the CAT RBT for the unknown age case is not very sensitive to the
increase in the RBT and then a decrease after 2.5 days. The trend can be functional relationship between τmax and CCAT. Changing Mbmax from 1
explained by differences in the relative decay of the pathogens and to 10 to 100 g/ 100 ml does not substantially change the CAT RBT for
CAT. The initial increase occurs because the pathogen contributing unknown age (results not shown).
most of the risk, Campylobacter, has a higher decay rate constant, on
average, than CAT (medians of 2.8 per day for Campylobacter compared 3.2. Exposure to a mixture of gull feces and sewage of unknown ages:
to 1.8 day−1 for CAT). After about 2.5 days, when Salmonella overtakes CATand HF183
Campylobacter as contributing most of the risk, the RBT begins to de­
crease because the decay rate constant for Salmonella (median of 0.74 We considered a hazard scenario where surface water contains both
day−1) is smaller, on average than the decay rate constant for CAT. gull feces and raw sewage, detected using the CAT marker and HF183,
When gull fecal contaminations is fresh (unaged, τ = 0), the risk-based respectively. The age of contamination is unknown in each case. The
threshold is approximately 45,000 copies/100 ml. After 7 days, the risk- resultant median calculated risk given different co-occurring con­
based threshold declines to approximately 8500 copies/100 ml. The centrations of CAT and HF183 is shown in the form of a contour plot
geometric mean RBT for CAT over all ages considered in the simulation (Fig. 6). The constant risk line of 32/1000 is shown in black (it is la­
is 94,000 copies/100 ml. beled with the log10-transform 32/1000). This line shows different risk-
We conducted a QMRA assuming that the age was unknown as based thresholds for HF183 given CCAT. A summary of the values

6
A.B. Boehm and J.A. Soller Microbial Risk Analysis xxx (xxxx) xxxx

Fig. 4. Panel A. The risk-based threshold for Mb as a function of the age of fecal contamination. Panel B. The risk-based threshold of CAT as a function of age of
contamination. The 45th and 55th percentiles are shown as well as the median – which is the risk-based threshold for CAT.

Fig. 5. The median, 45th, and 55th percentiles of calculated risk for a measured Fig. 6. Median calculated risk as a function of CHF183 and CCAT. The lines re­
concentration of CAT in surface waters when the age of contamination is un­ present lines of constant median risk and are labeled with the log10-transform of
known, and thus assumed to be any possible age in the Monte Carlo simula­ the risk value. The line of constant risk = 32/1000 is shown in black and is
tions. The median risk in the figure is the median of 50,000 Monte Carlo si­ labeled with the log10(32/1000)= −1.49485. The symbols represent the va­
mulations for each CCAT. lues used to create Table 5. Similar contours for the 45th and 55th percentiles of
risk can be found in the supporting material (Figures S3 and S4).
corresponding to a risk of 32/1000 is provided in Table 5. As CCAT
increases from 0 to 10,000 copies/100 ml, the risk-based threshold for present in surface waters, and the proportion of each component of the
HF183 decreases from 525 to 7 copies/100 ml. In other words, as the mixture and its age is known, it is possible to estimate the RBT for
amount of gull contamination (of unknown age) increases in the water, HF183. However, it is rarely the case that anything specific is known
the amount of raw sewage (of unknown age) that can be present de­ about the age of contamination, and the proportion of each aged
creases in order to ensure the median calculated risk does not exceed component that constitute the mixture. Given these uncertainties, we
32/1000. considered the case where the age of contamination is unknown, and
bound by 0 and a maximum possible age (τmax). The maximum possible
age is a function of the concentration of HF183; it is the age for which
4. Discussion
that HF183 concentration would exist in the water if the water were
100% sewage decaying with a rate constant equivalent to the median k
The risk-based threshold (RBT) for a water quality indicator is de­
for HF183. In this scenario, it is assumed the sewage contamination is
fined as the concentration of the indicator for which the median si­
present in equal proportions of each possible age. Under this scenario,
mulated illness risk is 32/1000. The RBT of HF183 is a strong function
which we term hereafter “the unknown age” scenario, we derived a RBT
of the age of contamination, as described previously (Boehm et al.,
of 525 copies/100 ml for HF183. This represents an update to the RBT
2018). This is due to the fact that norovirus, the pathogen to which
reported by Boehm et al. (2018) of 4100 copies/ 100 ml for a similar
exposure drives the risk estimates, has a smaller decay rate constant on
hazard scenario. The present RBT derivation includes refined tem­
average, than HF183 (Boehm et al., 2018). Sewage contamination in a
perature-adjusted decay rate constants, and additional risk from ex­
water body is rarely of one single known age, unless there was a spe­
posure to adenovirus, updated pathogen concentrations in sewage for
cific, short duration release of sewage into an otherwise clean water­
Cryptosporidium, and, in our opinion, a more defensible formulation for
body. If there is a mixture of sewage contamination of diverse ages

7
A.B. Boehm and J.A. Soller Microbial Risk Analysis xxx (xxxx) xxxx

the maximum age of contamination. The difference between the ap­ bird fecal marker (hereafter, CAT). We used risks from gull con­
proach used herein for estimating the maximum age of contamination tamination as our basis since gulls are commonly present near coastal
and that used by Boehm et al. (2018) is discussed in detail elsewhere waters and pathogen data in gull feces were previously reported
(Boehm, 2019). (Soller et al., 2010a). We found that the CAT RBT varied with fecal age
The RBT for HF183 for the unknown scenario is sensitive to the in a bimodal fashion; initially increasing with age and then decreasing
relationship between τmax and CHF183. This sensitivity was described by owing to different pathogen exposures dominating the risk as the
Boehm (2019) in her analysis of RBTs for coliphage. The RBT for HF183 contamination ages (first Campylobacter and then Salmonella). Overall,
decreases when τmax is larger for a specific CHF183. For the present however, the RBT is a much weaker function of contamination age than
analysis, τmax represents the age for which that HF183 concentration HF183 in sewage contaminated waters because of the more similar
would exist in the water if the water were 100% sewage decaying with decay rate constants of the indicator CAT and the pathogens Salmonella
a rate constant equivalent to the median for HF183. Given that we and Campylobacter compared to the decay rates of HF183 and nor­
understand that recreational water would never be 100% raw sewage, ovirus. As with the human sewage hazard, it is often unknown how
and a lower percent of raw sewage at time 0 would decrease τmax for a aged bird fecal contamination is in a surface water, or if the con­
given CHF183 by reducing “b” in Eq. (5), this formulation for τmax pro­ tamination is composed of a diverse mixture of ages. We therefore
vides a margin of safety around the estimate of the risk-based threshold. considered the unknown age scenario for gull fecal contamination and
An additional margin of safety allowed for in our analysis is our use of found that CAT RBT to be 200,000 copies/100 ml.
the non-aggregated hypergeometric norovirus dose-response curve We then considered a hazard of a mixture of gull feces and human
published by Teunis et al. (2008) which predicts a larger probability of sewage in surface waters, considering the unknown age scenario for
infection given low doses of norovirus than the fractional beta-poisson both. The results of that analysis indicate that RBT for HF183 declines
model described by Messner et al. (2014) by additional margin of as more CAT is present in the water body. This is because as CAT in­
safety, we mean that we provide an estimate for the RBT that assumes a creases in the water, exposure to bird feces adds risk meaning that less
relatively high level of risk from exposure to norovirus. Using the human sewage can be in the water if the allowable risk is capped at 32/
fractional beta-poisson model to describe the norovirus dose-response 1000 illnesses. For example, if CAT is present at 1000 copies/100 ml,
provides a lower risk estimate for exposure to norovirus, and adeno­ and the manager wishes to ensure that the total risk from exposure to
virus contributes the most to the cumulative risk (data not shown). gull feces and human sewage is less than 32/1000, then HF183 RBT
Using a dose-response curve that weights the contributions of both the would be 70 copies/100 ml.
norovirus dose response curves gives a similar estimate of risk as using All the hazards we considered in this work included contributions of
the non-aggregated hypergeometric norovirus dose-response untreated sewage. Another potential source of pathogens to the en­
(Boehm, 2019). vironmental is treated wastewater effluent. While treated effluent cer­
To evaluate the credibility of the approach for calculating the tainly contains far fewer pathogens than untreated sewage, it may
HF183 RBT for the unknown age scenario, we repeated the analysis contain viruses or other hardy pathogens that persist through the
using culturable ENT instead of HF183. We found that the RBT for ENT treatment train (Brown et al., 2017b), and it also can contain HF183
for the unknown age scenario was 47 CFU/100 ml, a value close to, but (~104 copies/ml) (Brown et al., 2017b). We previously conducted a
higher than, the USEPA ambient water quality criterion of 30 CFU/ QMRA to calculate the RBT for HF183 when the source of contamina­
100 ml, derived via a completely independent method. The fact that the tion is unaged, treated effluent; we found that the RBT was approxi­
unknown age scenario-derived RBT for ENT and the national ambient mately 10 times higher than it was if the source of contamination is
water quality criteria are the same order of magnitude, and quite close unaged, untreated sewage (Brown et al., 2017b). We thus expect that
together, adds credence to this numerical simulation-based approach. RBT established for untreated sewage to be protective for exposure to
Depending on the concentration of HF183, the maximum age of treated effluent.
contamination can sometimes be quite large (for example 10 days when This work, along with prior publications evaluating the potential for
HF183 is 160 copies/100 ml). We considered the case where the molecular marker-based RBTs (Boehm et al., 2015; Boehm et al., 2018;
maximum age of contamination was capped at 3 days, and reran a Brown et al., 2017a; Brown et al., 2017b; Crank et al., 2019), indicates
scenario similar to the unknown age scenario (hereafter, “capped age” that molecular markers have potential to be valuable tools in recrea­
scenario) (Figure S2). This duration may represent, for example, when tional water management and potentially for water quality regulations.
there is a local source of contamination (and no non-local source that These molecular markers exhibit desirable characteristics which could
would introduce aged contamination) in a surface water that has a allow water quality managers to efficiently identify salient sources of
maximum residence time of 3 d The capped age scenario resulted in a contamination in a waterbody and focus their remediation efforts on
HF183 RBT of 2400 copies /100 ml. However, it is unclear the extent to contamination sources that can yield the greatest impact on improved
which this or similar scenarios can be applied, as they require that no human health risk – the basis for current recreational water quality
non-local source has the potential to introduce aged contamination into criteria, recommendations, and/or regulations worldwide. This overall
the waterbody. approach – focusing on risk-relevant contamination is consistent with
Many coastal and inland waterbodies have bird fecal contamination recent revisions to water quality recommendations in the USA (USEPA,
present. For example, along the California coast, several recent studies 2012b). In fact, to facilitate this overall approach, the USEPA has
have measured bird-specific fecal source tracking markers at California published guidelines describing the development of site-specific alter­
beaches. For example, Russell et al. (2013) measured the LeeSeaGull native recreational water quality criteria which include alternative in­
bird fecal marker (CAT) at Cowell Beach in Santa Cruz and measured dicators and methods (USEPA, 2014a; USEPA, 2014b). This work fun­
between 102 and 105 copies/100 ml in the seawater. Ervin et al. (2014) damentally fits within the overarching framework described there and
measured a bird fecal marker (Gull2) at a beach in Santa Barbara and provides water quality managers with an additional tool that could be
found ~103 copies/100 ml. Steele et al. (2018) report the LeeSeaGull useful in specific circumstances. We acknowledge that making HF183
marker (CAT) in stormwater in San Diego between below their detec­ measurements requires qualified laboratory staff and specialized
tion limit and 103 copies/100 ml. equipment. While some agencies that conduct water quality measure­
We were interested in determining how much untreated human ments may have reliable access to the appropriate staff and equipment,
sewage could be present in a waterbody if that waterbody also con­ others may not.
tained bird feces while maintaining a calculated risk less than 32/1000. We understand that this field is quickly evolving and that refine­
We first updated previous work by Brown et al. (2017a; 2017b) to ments to this work will undoubtedly be published in both near and
consider the effect of bird feces aging on the RBT for the LeeSeaGull longer terms. This raises an important and philosophically interesting

8
A.B. Boehm and J.A. Soller Microbial Risk Analysis xxx (xxxx) xxxx

question about when a new approach or indicator should be adopted Markers in Recreational Water Contaminated with Fresh Untreated and Secondary
and/or how much information is required. Although there is no uni­ Treated Sewage. Environ Int 117, 243–249. https://doi.org/10.1016/j.envint.2018.
05.012.
versal answer to this question, we believe that, in this case, the science Bae, S., Wuertz, S., 2012. Survival of Host-Associated Bacteroidales Cells and Their
has sufficiently advanced and the potential benefits of focusing on risk- Relationship with Enterococcus Spp., Campylobacter Jejuni, Salmonella Enterica
Serovar Typhimurium, and Adenovirus in Freshwater Microcosms as Measured by
relevant contamination outweigh the drawbacks associated with un­ Propidium Monoazide-Quantitative PCR. Appl Environ Microbiol 78 (4), 922–932.
certainties and future iterative refinements, and thus encourage the https://doi.org/10.1128/AEM.05157-11.
consideration of the proposed RBTs for use in the evaluation of re­ Bielaszewska, M., Janda, J., Bláhová, K., Minaríková, H., Jíková, E., Karmali, M.A.,
Laubová, J., Sikulová, J., Preston, M.A., Khakhria, R., Karch, H., Klazarová, H., Nyc,
creational water quality for those agencies capable of conducting mo­ O., 1997. Human Escherichia Coli O157:H7 Infection Associated with the
lecular measurements. Future efforts might also consider monitoring for Consumption of Unpasteurized Goat’s Milk. Epidemiol. Infect. 119 (3), 299–305.
Boehm, A.B., Van De Werfhorst, L.C., Griffith, J.F., Holden, P.A., Jay, J.A., Shanks, O.C.,
viral pathogens which appear to be the main drivers of risk in recrea­
Wang, D., Weisberg, S.B., 2013. Performance of Forty-Three Microbial Source
tional waters. Tracking Methods: a Twenty-Seven Lab Evaluation Study. Water Res. 47 (18),
Sensitivity analyses of the models described in this paper have lar­ 6812–6828.
Boehm, A.B., Soller, J.A., Shanks, O.C., 2015. Human-Associated Fecal QPCR
gely been completed previously (Boehm et al., 2015; Boehm et al., Measurements and Simulated Risk of Gastrointestinal Illness in Recreational Waters
2018; Brown et al., 2017a; Brown et al., 2017b) and the sensitivity of Contaminated with Raw Sewage. Environ Sci Technol Lett 2, 270–275. https://doi.
the HF183 RBT to certain assumptions and parameter choices have org/10.1021/acs.estlett.5b00219.
Boehm, A.B., Graham, K.E., Jennings, W.C., 2018. Can We Swim yet? Systematic Review,
already been discussed herein. An important parameter that strongly Meta-Analysis, and Risk Assessment of Aging Sewage in Surface Waters. Environ. Sci.
affects the HF183 RBT is the decay rate constant for norovirus. Since Technol. 52 (17), 9634–9645.
Boehm, A.B., 2019. Risk-Based Water Quality Thresholds for Coliphages in Surface
norovirus is difficult to culture using a cell line, estimates for this Waters: effect of Temperature and Contamination Aging. Environ. Sci.: Processes
parameter come from studies that quantify the virus using RT-QPCR or Impacts 21 (12), 2031–2041. https://doi.org/10.1039/C9EM00376B.
by using a viral surrogate like murine norovirus (Boehm et al., 2019). Boehm, A.B., Silverman, A.I., Schriewer, A., Goodwin, K., 2019. Systematic Review and
Meta-Analysis of Decay Rates of Waterborne Mammalian Viruses and Coliphages in
Future refinement of our understanding of norovirus persistence in Surface Waters. Water Res., 114898. https://doi.org/10.1016/j.watres.2019.114898.
environmental waters can improve the QMRA described herein. The Brooks, L.E., Field, K.G., 2016. Bayesian Meta-Analysis to Synthesize Decay Rate Constant
decay rate constants used in the QMRA models were derived from Estimates for Common Fecal Indicator Bacteria. Water Res. 104, 262–271. https://
doi.org/10.1016/j.watres.2016.08.005.
published systematic reviews of decay rate constants of pathogens in Brown, K.I., Graham, K.E., Boehm, A.B., 2017a. Risk-Based Threshold of Gull-Associated
raw waters and this represent the best available estimates to date; Fecal Marker Concentrations for Recreational Water. Environ. Sci. Technol. Lett. 4
(2), 44–48. https://doi.org/10.1021/acs.estlett.6b00473.
however, further refinements on those estimates are encouraged. Ad­ Brown, K.I., Graham, K.E., Soller, J.A., Boehm, A.B., 2017b. Estimating the Probability of
ditionally, there are few data on CAT decay in water and additional Illness Due to Swimming in Recreational Water with a Mixture of Human- and Gull-
data are needed. It is also important to note that the HF183 RBT is for Associated Microbial Source Tracking Markers. Environ. Sci.: Processes Impacts 19
(12), 1528–1541. https://doi.org/10.1039/C7EM00316A.
HF183 measured in the manner used to obtain the HF183 distribution Brown, K.I., Boehm, A.B., 2015. Comparative Decay of Catellicoccus Marimmalium and
in raw sewage (Shanks et al., 2010). HF183 measured using another Enterococci in Beach Sand and Seawater. Water Res. 83, 377–384. https://doi.org/
10.1016/j.watres.2015.06.055.
method (including both the concentration and quantification steps)
Crank, K., Petersen, S., Bibby, K., 2019. Quantitative Microbial Risk Assessment of
would need to be shown to be equivalent to the method of Shanks et al. Swimming in Sewage Impacted Waters Using CrAssphage and Pepper Mild Mottle
(2010) for the HF183 RBTs reported herein to be applicable. Virus in a Customizable Model. Environ. Sci. Technol. Lett. https://doi.org/10.1021/
acs.estlett.9b00468.
Crockett, C.S., 2007. The Role of Wastewater Treatment in Protecting Water Supplies
Acknowledgements Against Emerging Pathogens. Water Environment Research 79 (3), 221–232. https://
doi.org/10.2175/106143006X111952.
DeFlorio-Barker, S., Arnold, B.F., Sams, E.A., Dufour, A.P., Colford Jr., J.M., Weisberg,
This work was funded via an MOU between the City and County of S.B., Schiff, K.C., Wade, T.J., 2018. Child Environmental Exposures to Water and
San Diego, and County of Orange. Funding was provided through the Sand at the Beach: findings from Studies of over 68,000 Subjects at 12 Beaches. J.
Exposure Science And Environmental Epidemiology 28, 93–100.
City of San Diego's contract with Tetra Tech, Inc. We thank Ruth Kolb, Eisenberg, J.N.S., Seto, E.Y.W., Olivieri, A.W., Spear, R.C., 1996. Quantifying Water
Victoria Kalkirtz, and Drew Kleis (City of San Diego), Todd Snyder and Pathogen Risk in an Epidemiological Framework. Risk Anal 16 (4), 549–563.
Jo Ann Weber (County of San Diego), and Grant Sharp and Cindy Rivers Eftim, S.E., Hong, T., Soller, J., Boehm, A., Warren, I., Ichida, A., Nappier, S.P., 2017.
Occurrence of Norovirus in Raw Sewage – A Systematic Literature Review and Meta-
(County of Orange) for their support of this work. Helen Yu from the Analysis. Water Res. 111 (Supplement C), 366–374. https://doi.org/10.1016/j.
San Diego Water Quality Control Board provided technical direction watres.2017.01.017.
and oversight of this work. Clint Boschen (TetraTech, Inc.) and Ken Environmental Incentives and EcoNorthwest. Cost-Benefit Analysis: san Diego Region
Bacteria Total Maximum Daily Loads; San Diego Regional Water Quality Control
Schiff (Southern California Coastal Water Research Project) were on the Board, 2017.
technical team and provided review and thoughtful feedback Ervin, J.S., Van De Werfhorst, L.C., Murray, J.L.S., Holden, P.A., 2014. Microbial Source
Tracking in a Coastal California Watershed Reveals Canines as Controllable Sources
throughout the course of this work. Clint Boschen was the Project of Fecal Contamination. Environ. Sci. Technol. 48 (16), 9043–9052. https://doi.org/
Manager for the TetraTech contract and served as liaison with the 10.1021/es502173s.
funders. Fazil, A.M., 1996. A Quantitative Risk Assessment Model For Salmonella. Drexel
University, Philadelphia, PA.
The authors declare no conflict of interest in the conduct of this García-Aljaro, C., Bonjoch, X., Blanch, A.R., 2004. Combined Use of an Immunomagnetic
work. Separation Method and Immunoblotting for the Enumeration and Isolation of
Escherichia Coli O157 in Wastewaters. J. Appl. Microbiol. 98 (3), 589–597. https://
doi.org/10.1111/j.1365-2672.2004.02497.x.
Supplementary materials Given, S., Pendleton, L.H., Boehm, A.B., 2006. Regional Public Health Cost Estimates of
Contaminated Coastal Waters: a Case Study of Gastroenteritis at Southern California
Beaches. Environ. Sci. Technol. 40, 4851–4858.
Supplementary material associated with this article can be found, in Griffith, J.F., Weisberg, S.B., McGee, C.D., 2003. Evaluation of Microbial Source Tracking
the online version, at doi:10.1016/j.mran.2020.100139. Methods Using Mixed Fecal Sources in Aqueous Test Samples. J Water Health 1,
141–151.
Haas, C.N., Rose, J.B., Gerba, C.P., 1999. Quantitative Microbial Risk Assessment; Lett.
References Appl. Microbiol. J.W. Wiley, Inc.
Harwood, V.J., Levine, A.D., Scott, T.M., Chivukula, V., Lukasik, J., Farrah, S.R., Rose,
J.B., 2005. Validity of the Indicator Organism Paradigm for Pathogen Reduction in
Ahmed, W., Hughes, B., Harwood, J.V., 2016. Current Status of Marker Genes of
Reclaimed Water and Public Health Protection. Appl. Environ. Microbiol. 71 (6),
Bacteroides and Related Taxa for Identifying Sewage Pollution in Environmental
3163–3170. https://doi.org/10.1128/aem.71.6.3163-3170.2005.
Waters. Water (Basel) 8 (6). https://doi.org/10.3390/w8060231.
Hewitt, J., Leonard, M., Greening, G.E., Lewis, G.D., 2011. Influence of Wastewater
Ahmed, W., Masters, N., Toze, S., 2012. Consistency in the Host Specificity and Host
Treatment Process and the Population Size on Human Virus Profiles in Wastewater.
Sensitivity of the Bacteroides HF183 Marker for Sewage Pollution Tracking. Lett.
Water Res. 45 (18), 6267–6276. https://doi.org/10.1016/j.watres.2011.09.029.
Appl. Microbiol. 55 (4), 283–289. https://doi.org/10.1111/j.1472-765X.2012.
Hurst, C.J., McClellan, K.A., Benton, W.H., 1988. Comparison of Cytopathogenicity,
03291.x.
Immunofluorescence and In Situ DNA Hybridization as Methods for the Detection of
Ahmed, W., Hamilton, K.A., Lobos, A., Hughes, B., Staley, C., Sadowsky, M.J., Harwood,
Adenoviruses. Water Res. 22 (12), 1547–1552. https://doi.org/10.1016/0043-
V.J., 2018. Quantitative Microbial Risk Assessment of Microbial Source Tracking

9
A.B. Boehm and J.A. Soller Microbial Risk Analysis xxx (xxxx) xxxx

1354(88)90167-4. Soller, J.A., Schoen, M.E., Bartrand, T., Ravenscroft, J.E., Ashbolt, N.J., 2010a. Estimated
Jian, Yang, Schneider Orren, D., Jjemba Patrick, K., Lechevallier Mark, W., 2015. Human Health Risks from Exposure to Recreational Waters Impacted by Human and
Microbial Risk Modeling for Main Breaks. J Am Water Works Assoc 107 (2), Non-Human Sources of Faecal Contamination. Water Res. 44 (16), 4674–4691.
E97–E108. https://doi.org/10.5942/jawwa.2015.107.0010. Soller, J.A., Bartrand, T., Ashbolt, N.J., Ravenscroft, J., Wade, T.J., 2010b. Estimating the
Kitajima, M., Haramoto, E., Iker, B.C., Gerba, C.P., 2014. Occurrence of Cryptosporidium, Primary Aetiologic Agents in Recreational Freshwaters Impacted by Human Sources
Giardia, and Cyclospora in Influent and Effluent Water at Wastewater Treatment of Faecal Contamination. Water Res. 44 (16), 4736–4747.
Plants in Arizona. Sci. Total Environment 484, 129–136. https://doi.org/10.1016/j. Soller, J.A., Eisenberg, J.N.S., 2008. An Evaluation of Parsimony for Microbial Risk
scitotenv.2014.03.036. Assessment Models. Environmetrics 19 (1), 61–78.
Koivunen, J., Siitonen, A., Heinonen-Tanski, H., 2003. Elimination of Enteric Bacteria in Soller, J.A., Schoen, M., Steele, J.A., Griffith, J.F., Schiff, K.C., 2017a. Incidence of
Biological–Chemical Wastewater Treatment and Tertiary Filtration Units. Water Res. Gastrointestinal Illness Following Wet Weather Recreational Exposures: harmoniza­
37 (3), 690–698. https://doi.org/10.1016/S0043-1354(02)00305-6. tion of Quantitative Microbial Risk Assessment with an Epidemiologic Investigation
Korajkic, A., McMinn, B.R., Harwood, V.J., 2018. Relationships between Microbial of Surfers. Water Res. 121, 280–289. https://doi.org/10.1016/j.watres.2017.05.017.
Indicators and Pathogens in Recreational Water Settings. Int J Environ Res Public Soller, J.A., Eftim, S., Wade, T.J., Ichida, A.M., Clancy, J.L., Johnson, T.B., Schwab, K.,
Health 15 (12). https://doi.org/10.3390/ijerph15122842. Ramirez-Toro, G., Nappier, S., Ravenscroft, J.E., 2015. Use of Quantitative Microbial
Layton, B., Cao, Y., Ebentier, D., Hanley, K., Van De Werfhorst, L., Wang, D., Madi, T., Risk Assessment to Improve Interpretation of a Recreational Water Epidemiological
Whitman, R., Byappanahalli, M., Ballesté, E., Meijer, W., Schriewer, A., Wuertz, S., Study. Microbial Risk Analysis. https://doi.org/10.1016/j.mran.2015.04.001.
Converse, R., Noble, R., Srinivasan, S., Rose, J., Lee, C.S., Lee, J., Shields, J., Stewart, Soller, J.A., Eftim, S.E., Warren, I., Nappier, S.P., 2017b. Evaluation of Microbiological
J., Reischer, G., Farnleitner, A., Sinigalliano, C., Rodrigues, R., Lozach, S., Risks Associated with Direct Potable Reuse. Microbial Risk Analysis 5, 3–14. https://
Gourmelon, M., Peed, L., Shanks, O., Jay, J., Holden, T., Boehm, A., Griffith, J., 2013. doi.org/10.1016/j.mran.2016.08.003.
Performance of Human Fecal Anaerobe-Associated PCR-Based Assays in a Multi- Staley, C., Gordon, K.V., Schoen, M.E., Harwood, V.J., 2012. Performance of Two
Laboratory Method Evaluation Study. Water Res. 47 (18), 6897–6908. Quantitative PCR Methods for Microbial Source Tracking of Human Sewage and
Lee, C., Marion, J.W., Lee, J., 2013. Development and Application of a Quantitative PCR Implications for Microbial Risk Assessment in Recreational Waters. Appl Environ
Assay Targeting Catellicoccus Marimammalium for Assessing Gull-Specific Fecal Microbiol 78 (20), 7317–7326. https://doi.org/10.1128/AEM.01430-12.
Contamination at Lake Erie Beaches. Sci. Total Environment 454–455, 1–8. Stampi, S., Varoli, O., Zanetti, F., De Luca, G., 1993. Arcobacter Cryaerophilus and
Lemarchand, K., Lebaron, P., 2003. Occurrence of Salmonella Spp. and Cryptosporidium Thermophilic Campylobacters in a Sewage Treatment Plant in Italy: two Secondary
Spp. in a French Coastal Watershed: relationship with Fecal Indicators. FEMS Treatments Compared. Epidemiol. Infect. 110 (3), 633–639. https://doi.org/10.
Microbiol. Letts. 218 (1), 203–209. 1017/S0950268800051050.
Lévesque, B., Brousseau, P., Bernier, F., Dewailly, É., Joly, J., 2000. Study of the Bacterial Steele, J.A., Blackwood, A.D., Griffith, J.F., Noble, R.T., Schiff, K.C., 2018. Quantification
Content of Ring-Billed Gull Droppings in Relation to Recreational Water Quality. of Pathogens and Markers of Fecal Contamination during Storm Events along Popular
Water Res. 34 (4), 1089–1096. https://doi.org/10.1016/S0043-1354(99)00266-3. Surfing Beaches in San Diego, California. Water Res. 136, 137–149. https://doi.org/
Lu, J., Santo Domingo, J., Lamendella, R., Edge, T., Hill, S., 2008. Phylogenetic Diversity 10.1016/j.watres.2018.01.056.
and Molecular Detection of Bacteria in Gull Feces. Appl. Environ. Microbiol. 74 (13), Symonds, E.M., Sinigalliano, C., Gidley, M., Ahmed, W., McQuaig-Ulrich, S.M., Breitbart,
3969–3976. M., 2016. Faecal Pollution along the Southeastern Coast of Florida and Insight into
Ludwig, K., Sarkim, V., Bitzan, M., Karmali, M.A., Bobrowski, C., Ruder, H., Laufs, R., the Use of Pepper Mild Mottle Virus as an Indicator. J. Appl. Microbiol. 121 (5),
Sobottka, I., Petric, M., Karch, H., Müller-Wiefel, D.E., 2002. Shiga Toxin-Producing 1469–1481. https://doi.org/10.1111/jam.13252.
Escherichia Coli Infection and Antibodies against Stx2 and Stx1 in Household Teunis, P.F., Moe, C.L., Liu, P., S, E.M., Lindesmith, L., Baric, R.S., Le Pendu, J., Calderon,
Contacts of Children with Enteropathic Hemolytic-Uremic Syndrome. J. Clin. R.L., 2008a. Norwalk Virus: how Infectious Is It? J. Med. Virol. 80 (8), 1468–1476.
Microbiol. 40 (5), 1773–1782. https://doi.org/10.1128/JCM.40.5.1773-1782.2002. Teunis, P., Nagelkerke, N.J., Haas, C.N., 1999. Dose-Response Models for Infectious
Messner, M.J., Berger, P., Nappier, S.P., 2014. Fractional Poisson–a Simple Dose- Gastroenteritis. Risk Analysis 19 (6), 1251–1260.
Response Model for Human Norovirus. Risk Analysis 34 (10), 1820–1829. Teunis, P., Van den Brandhof, W., Nauta, M., Wagenaar, J., Van den Kerkhof, H., Van Pelt,
Nasser, A.M., 2016. Removal of <em>Cryptosporidium</Em>by Wastewater W., 2005. A Reconsideration of the Campylobacter Dose-Response Relation.
Treatment Processes: a Review. J Water Health 14 (1), 1. https://doi.org/10.2166/ Epidemiol. Infect. 133 (4), 583–592.
wh.2015.131. Teunis, P., Ogden, I.D., Strachan, N.J.C., 2008b. Hierarchical Dose Response of E. Coli
Nelson, K.L., Boehm, A.B., Davies-Colley, R.J., Dodd, M.C., Kohn, T., Linden, K.G., Liu, Y., O157:H7 from Human Outbreaks Incorporating Heterogeneity in Exposure.
Maraccini, P.A., McNeill, K., Mitch, W.A., Nguyen, T.H., Parker, K.M., Rodriguez, Epidemiol. Infect. 136 (6), 761–770. https://doi.org/10.1017/S0950268807008771.
R.A., Sassoubre, L.M., Wigginton, K.R., Zepp, R.G., 2018. Sunlight-Mediated Teunis, P., Schijven, J., Rutjes, S., 2016. A Generalized Dose-Response Relationship for
Inactivation of Health-Relevant Microorganisms in Water: a Review of Mechanisms Adenovirus Infection and Illness by Exposure Pathway. Epidemiol. Infect. 144 (16),
and Modeling Approaches. Environ. Sci. Processes Impacts 20, 1089–1122 In revi­ 3461–3473. https://doi.org/10.1017/S0950268816001862.
sion. USEPA. Ambient Water Quality Criteria for Bacteria; Office of Water, Series Ed.; EPA440/5-
Nguyen, K.H., Senay, C., Young, S., Nayak, B., Lobos, A., Conrad, J., Harwood, V.J., 2018. 84-002; Washington, DC, 1986.
Determination of Wild Animal Sources of Fecal Indicator Bacteria by Microbial USEPA. National Primary Drinking Water Regulations: long Term 2 Enhanced Surface Water
Source Tracking (MST) Influences Regulatory Decisions. Water Res. 144, 424–434. Treatment Rule (LT2ESWTR); Final Rule. 40CFR Parts 9, 141 and 142, volume 71,
https://doi.org/10.1016/j.watres.2018.07.034. Number 654; 2006.
Pruss, A., 1998. Review of Epidemiological Studies on Health Effects from Exposure to USEPA. Recreational Water Quality Criteria; Office of Water 820-F-12-058; 2012.
Recreational Water. Int. J. Epidemiol. 27 (1), 1–9. USEPA. Recreational Water Quality Criteria. https://www.epa.gov/wqc/2012-recreational-
Rose, J.B., Gerba, C.P., 1991. Use of Risk Assessment for Development of Microbial water-quality-criteria, 2012a.
Standards. Water Science and Technology 24, 29–34. USEPA. Framework for Human Health Risk Assessment to Inform Decision Making; 601-D12-
Russell, T.L., Sassoubre, L.M., Wang, D., Masuda, S., Chen, H., Soetjipto, C., Hassaballah, 001; 2012b.
A., Boehm, A.B., 2013. A Coupled Modeling and Molecular Biology Approach to USEPA. Site-Specific Alternative Recreational Criteria Technical Support Materials For
Microbial Source Tracking at Cowell Beach, Santa Cruz, CA, USA. Environ. Sci. Alternative Indicators and Methods; EPA-820-R-14-011; 2014a.
Technol. 47, 10231–10239. USEPA. Overview of Technical Support Materials: a Guide to the Site-Specific Alternative
Schoen, M.E., Ashbolt, N.J., 2010. Assessing Pathogen Risk to Swimmers at Non-Sewage Recreational Criteria TSM Document; 820-R-14–010; Office of Water: Washington DC,
Impacted Recreational Beaches. Environ. Sci. Technol. 44, 2286–2291. 2014b.
Schoen, M.E., Ashbolt, N.J., Jahne, M.A., Garland, J., 2017. Risk-Based Enteric Pathogen USEPA. Review of Coliphages as Possible Indicators of Fecal Contamination for Ambient Water
Reduction Targets for Non-Potable and Direct Potable Use of Roof Runoff, Quality; 820-R-15–098; USEPA: Washington DC, 2015.
Stormwater, and Greywater. Microbial Risk Analysis 5, 32–43. https://doi.org/10. Wade, T.J., Pai, N., Eisenberg, J., Colford, J.M., 2003. Do USEPA Water Quality
1016/j.mran.2017.01.002. Guidelines for Recreational Waters Prevent Gastrointestinal Illness? A Systematic
Shanks, O.C., Kelty, C.A., Sivaganesan, M., Varma, M., Haugland, R.A., 2009. Review and Meta-Analysis. Environ. Heal. Perspec. 111 (8), 1102–1109.
Quantitative PCR for Genetic Markers of Human Fecal Pollution. Appl. Environ. Werber, D., Mason, B.W., Evans, M.R., Salmon, R.L., 2008. Preventing Household
Microbiol. 75 (17), 5507–5513. Transmission of Shiga Toxin–Producing Escherichia Coli O157 Infection: promptly
Shanks, O.C., White, K., Kelty, C.A., Sivaganesan, M., Blannon, J., Meckes, M., Varma, M., Separating Siblings Might Be the Key. Clinical Infectious Diseases 46 (8), 1189–1196.
Haugland, R.A., 2010. Performance of PCR-Based Assays Targeting Bacteroidales https://doi.org/10.1086/587670.
Genetic Markers of Human Fecal Pollution in Sewage and Fecal Samples. Environ. WHO, 2003. Guidelines For Safe Recreational Water Environments. Volume 1 Coastal and
Sci. Technol. 44 (16), 6281–6288. Fresh Waters. World Health Organization, Geneva.
Shanks, O.C., Korajkic, A., 2020. Chapter 6 - Microbial Source Tracking: characterization Wu, B., Wang, C., Zhang, C., Sadowsky, M.J., Dzakpasu, M., Wang, X.C., 2020. Source-
of Human Fecal Pollution in Environmental Waters with HF183 Quantitative Real- Associated Gastroenteritis Risk from Swimming Exposure to Aging Fecal Pathogens.
Time PCR. In: Budowle, B., Schutzer, S., Morse, S. (Eds.), Microbial Forensics, 3rd Environ. Sci. Technol. 54 (2), 921–929. https://doi.org/10.1021/acs.est.9b01188.
Edition. Academic Press, pp. 71–87. https://doi.org/10.1016/B978-0-12-815379-6. Zhang, Q., Gallard, J., Wu, B., Harwood, V.J., Sadowsky, M.J., Hamilton, K.A., Ahmed,
00006-4. W., 2019. Synergy between Quantitative Microbial Source Tracking (QMST) and
Sinigalliano, C., Ervin, J., Van De Werfhorst, L., Lee, J., Goodwin, K., Wanless, D., Quantitative Microbial Risk Assessment (QMRA): a Review and Prospectus. Environ
Bartkowiak, J., Wuertz, S., Schriewer, A., Holden, T., Jay, J., Boehm, A.B., Layton, B., Int 130, 104703. https://doi.org/10.1016/j.envint.2019.03.051.
Raith, M., Meijer, W., Sadowsky, M., Gourmelon, M., Noble, R., Griffith, J., Zmirou, D., Pena, L., Ledrans, M., Letertre, A., 2003. Risks Associated with the
Byappanahalli, M., Wang, D., SantoDomingo, J., 2013. A Multi-Laboratory Microbiological Quality of Bodies of Fresh and Marine Water Used for Recreational
Comparative Assessment on the Performance of Seabird-Host-Specific Microbial Purposes: summary Estimates Based on Published Epidemiological Studies. Arch
Source Tracking PCR Assays Targeting Catellicoccus Marimammalium. Water Res. 47 Environ Health 58 (11), 703–711.
(18), 6897–6908.

10

You might also like