Abstract

Taiwan is surrounded by oceans, and therefore numerous pleasure beaches attract millions of tourists annually to participate in recreational swimming activities. However, impaired water quality because of fecal pollution poses a potential threat to the tourists' health. This study probabilistically characterized the health risks associated with recreational swimming engendered by waterborne enterococci at 13 Taiwanese beaches by using quantitative microbial risk assessment. First, data on enterococci concentrations at coastal beaches monitored by the Taiwan Environmental Protection Administration were reproduced using nonparametric Monte Carlo simulation (MCS). The ingestion volumes of recreational swimming based on uniform and gamma distributions were subsequently determined using MCS. Finally, after the distribution combination of the two parameters, the beta-Poisson dose–response function was employed to quantitatively estimate health risks to recreational swimmers. Moreover, various levels of risk to recreational swimmers were classified and spatially mapped to explore feasible recreational and environmental management strategies at the beaches. The study results revealed that although the health risks associated with recreational swimming did not exceed an acceptable benchmark of 0.019 illnesses daily at all beaches, they approached to this benchmark at certain beaches. Beaches with relatively high risks are located in Northwestern Taiwan owing to the current movements.

INTRODUCTION

Swimming is the most prevalent water recreational activity and involves whole-body contact with water (World Health Organization (WHO) 2003). However, natural water typically includes large amounts of pathogenic bacteria, which may be harmful to swimmers if they accidentally swallow the water. Waterborne bacteria mainly originate from human and animal feces. Untreated household sewage and agricultural wastewater with feces are discharged into rivers or oceans, constituting a fecal–oral transmission route for beach swimmers (Gerba et al. 1996; Haas et al. 2000). Typically, enterococci are an adequate microbial indicator associated with fecal contamination in marine waters and can be adopted to evaluate a swimmer's health risks via the aforementioned route (WHO 2003; Health Canada 2012; US Environmental Protection Agency (EPA) 2012). The US EPA (1986) suggested that the reference value of enterococci is the geometric mean concentration of 35 most probable number (MPN)/100 mL for recreational marine water. Moreover, the US EPA (1986); WHO (2003), and Health Canada (2012) have recommended 19 gastrointestinal (GI) illnesses per 1,000 swimmers daily as an acceptable risk benchmark.

Taiwan is located in the Western Pacific and is an island; therefore, it has many pleasure beaches. Millions of tourists visit the beaches annually and participate in swimming and surfing. In Taiwan, the reference threshold of enterococci is 50 MPN/100 mL for recreational marine water (Taiwan Environmental Protection Administration (EPA) 2016). However, the Taiwan EPA has monitored the water quality of coastal beaches over the long term and indicated that enterococci concentrations frequently exceed the aforementioned threshold at many beaches. Accordingly, assessing health risks to recreational swimmers at Taiwanese beaches is critical for preventing public health hazards.

Quantitative microbial risk assessment (QMRA), first proposed by Haas (1983), is a common tool for gauging the health risks to water users (Haas et al. 2014; US EPA 2014; WHO 2016). The exposure parameters in QMRA typically include pathogenic microorganism concentrations and water ingestion volumes. Monte Carlo simulation (MCS) is adopted to characterize the uncertainty or variability of the exposure parameters in QMRA. On the basis of epidemiological surveys, an exponential, beta-Poisson, or empirical model is used to quantitatively determine a dose–response relationship between bacteria and GI illnesses. Previous QMRA studies have analyzed health risks associated with recreational swimming activities in freshwater (e.g., Gerba et al. 1996; Haas et al. 2000; Steyn et al. 2004; Till et al. 2008; Soller et al. 2010; Schets et al. 2011; Chigor et al. 2014; Sunger & Haas 2015) and in marine water (e.g., Craig et al. 2003; Ashbolt et al. 2010; Schoen & Ashbolt 2010; Schets et al. 2011; Tseng & Jiang 2012; Dickinson et al. 2013; Ming et al. 2014). In addition, a few studies have investigated the volume of water consumed by coastal swimmers compared with pool or freshwater swimmers. For example, Ashbolt et al. (1997) presented observational data of 20–50 mL per coastal swimming event that were subsequently used for QMRA (Craig et al. 2003). Schets et al. (2011) also evaluated water ingestion volumes during coastal swimming activities according to exposure cases self-reported in questionnaires. Their study result indicated that the average water volumes swallowed per swimming event are 31 mL for children and 27 mL and 18 mL for adult men and women, respectively, and the swallowed water volumes follow a gamma distribution. To model the variability of pathogenic bacteria exposure, this study applied the data on water ingestion volumes of coastal swimming activities.

The present study applied QMRA to probabilistically characterize the health risks associated with recreational swimming that are engendered by waterborne enterococci at 13 Taiwanese beaches. First, enterococci concentrations at coastal beaches, as surveyed by the Taiwan EPA, were collected and reproduced using nonparametric MCS. The water ingestion volumes of recreational swimming were subsequently considered, as reported by Ashbolt et al. (1997) and Schets et al. (2011), and determined using MCS based on uniform and gamma distributions, respectively. Finally, after the distribution combination of the aforementioned parameters, the beta-Poisson dose–response function was employed to quantitatively determine health risks to recreational swimmers. Moreover, different levels of risks to recreational swimmers were classified and spatially mapped to explore feasible recreational and environmental management strategies at Taiwanese beaches.

MATERIALS AND METHODS

Study area

Taiwan is enclosed by the East China Sea to the north, the Bashi Channel to the south, the Taiwan Strait to the west, and the Pacific Ocean to the east (Figure 1). Taiwan has an area of approximately 36,000 km2 and mainly consists of Taiwan and Penghu Islands. Because of the surrounding oceans, Taiwan has abundant recreational resources of coastal beaches. Moreover, Taiwan has approximately 13 famous beach sites: Xinjinshan (XJS), Qianshuiwan (QSW), Qiding (QD), Tongxiao (TX), Mashagou (MSG), Xiziwan (XZW), Qijin (QJ), Kending (KD), Shanyuan (SY), Jiqi (JQ), Waiao (WA), Fulong (FL), and Guanyinting (GYT). According to statistical data published by the Taiwan Tourism Bureau, several beaches attract more than one million tourists annually.

Figure 1

Map of the Taiwan and beach sites.

Figure 1

Map of the Taiwan and beach sites.

Data collection for enterococci concentrations at coastal beaches

Since 2001, the Taiwan EPA has monitored the water quality, including enterococci concentrations, at 13 Taiwanese beaches during the summer (June–August). In general, three samples were taken from the left, central, and right sides of each beach. Because the WHO (2003) suggested that a 5-year period dataset can be adopted to assess microbial water quality, this study used the enterococci data between 2011 and 2015 reported by the Taiwan EPA (2016). Table 1 lists the statistics for the enterococci data recorded during this period at 13 beaches. From each beach, 30–69 samples were collected. The geometric mean and geometric standard deviation of the enterococci concentrations and the ratio exceeding 50 MPN/100 mL were the highest at the TX beach. The maximum of the enterococci concentrations was found at the GYT beach. Figure 2 shows the histograms of the observed enterococci concentrations at all beaches. This study applied Oracle Crystal Ball Release 11 to identify the distributions of the enterococci data by using the Kolmogorov–Smirnov (K–S) and chi-squared tests, indicating that the enterococci distributions at 11 beaches did not meet any specified standard distribution types. Accordingly, a nonparametric method was appropriate for modeling the enterococci data.

Table 1

Statistics for enterococci concentrations measured by the Taiwan EPA (2016) between 2011 and 2015

Beaches Data number Geometric meana (MPN/100 mL) Geometric standard deviationa (MPN/100 mL) Maximumb (MPN/100 mL) Ratio (%) of >50 MPN/100 mL 
Xinjinshan (XJS) 69 2.69 5.02 310 8.7 
Qianshuiwan (QSW) 30 7.75 7.10 340 16.7 
Qiding (QD) 51 6.25 6.48 150 15.7 
Tongxiao (TX) 69 11.57 8.30 1,300 24.6 
Mashagou (MZG) 69 4.67 6.59 190 13.0 
Xiziwan (XZW) 57 4.73 6.91 200 19.3 
Qijin (QJ) 48 3.04 4.23 86 2.1 
Kending (KD) 66 6.32 4.86 170 7.6 
Shanyuan (SY) 66 4.82 7.08 440 15.2 
Jiqi (JQ) 54 5.51 4.82 63 5.6 
Waiao (WA) 66 2.56 3.80 52 1.5 
Fulong (FL) 66 4.21 6.66 330 16.7 
Guanyinting (GYT) 69 3.32 7.97 2,900 13.0 
Beaches Data number Geometric meana (MPN/100 mL) Geometric standard deviationa (MPN/100 mL) Maximumb (MPN/100 mL) Ratio (%) of >50 MPN/100 mL 
Xinjinshan (XJS) 69 2.69 5.02 310 8.7 
Qianshuiwan (QSW) 30 7.75 7.10 340 16.7 
Qiding (QD) 51 6.25 6.48 150 15.7 
Tongxiao (TX) 69 11.57 8.30 1,300 24.6 
Mashagou (MZG) 69 4.67 6.59 190 13.0 
Xiziwan (XZW) 57 4.73 6.91 200 19.3 
Qijin (QJ) 48 3.04 4.23 86 2.1 
Kending (KD) 66 6.32 4.86 170 7.6 
Shanyuan (SY) 66 4.82 7.08 440 15.2 
Jiqi (JQ) 54 5.51 4.82 63 5.6 
Waiao (WA) 66 2.56 3.80 52 1.5 
Fulong (FL) 66 4.21 6.66 330 16.7 
Guanyinting (GYT) 69 3.32 7.97 2,900 13.0 

aLess than a detection limit for enterococci (<10 MPN/100 mL) is regarded as 1 MPN/100 mL for the calculation of the geometric mean and geometric standard deviation.

bMinimum is not showed at the beaches because less than the detection limit.

Figure 2

Histograms of enterococci concentrations monitored by the Taiwan EPA (2016).

Figure 2

Histograms of enterococci concentrations monitored by the Taiwan EPA (2016).

Quantitative microbial risk assessment

Haas et al. (2014) and the WHO (2016) have documented the QMRA framework to explore the health risks associated with recreational water. Typically, the QMRA involves hazard identification, exposure assessment, dose–response assessment, and risk characterization (US EPA 2001; 2014). Because of the discharge of untreated human sewage and agricultural wastewater, coastal marine water in Taiwan occasionally contains substantial amounts of enterococci, as reported by the Taiwan EPA (2016). The risk of exposure to enterococci during recreational swimming results from accidental ingestion via the fecal–oral transmission route (WHO 2003); thus, recreational swimmers at coastal beaches may suffer from GI illnesses. The beta-Poisson dose–response model applied in this study is defined as follows (Haas et al. 2014; US EPA 2014; Haas 2015):  
formula
(1)
 
formula
(2)
where P is the probability of disease infection (illnesses/user/day), dEnt is the exposure dose of enterococci (MPN), N50 is the median infective dose (MPN), α is the slope parameter for the beta-Poisson model (dimensionless), CEnt is the enterococci concentration (MPN/100 mL), and WIV is the water ingestion volume of recreational swimming (mL/exposure). Table 2 presents a summary of the estimation methods, values or distributions, and data sources of the parameters used in this study. In addition, this study considered a level of 19 GI illnesses per 1,000 swimmers daily as an acceptable risk benchmark set by US EPA (1986). This study also assumed one swimming exposure event daily for each swimmer.
Table 2

Estimation methods and data sources of parameters used in QMRA

Parameters Estimation methods Values or distributions Data sources 
α Point 0.16 Sunger & Haas (2015)  
N50 (MPN) Point 59,938 Sunger & Haas (2015)  
CEnt (MPN/100 mL) Probability distribution (MCS) Nonparametric distribution Taiwan EPA (2016)  
WIV (mL/exposure) Probability distribution (MCS) U(20, 50)a
Ga(0.45,60)b for adult men
Ga(0.51,35)b for women
Ga(0.58,55)b for children 
Ashbolt et al. (1997) and Schets et al. (2011)  
Parameters Estimation methods Values or distributions Data sources 
α Point 0.16 Sunger & Haas (2015)  
N50 (MPN) Point 59,938 Sunger & Haas (2015)  
CEnt (MPN/100 mL) Probability distribution (MCS) Nonparametric distribution Taiwan EPA (2016)  
WIV (mL/exposure) Probability distribution (MCS) U(20, 50)a
Ga(0.45,60)b for adult men
Ga(0.51,35)b for women
Ga(0.58,55)b for children 
Ashbolt et al. (1997) and Schets et al. (2011)  

aU(LB,UB) represents the uniform distribution with a lower bound of LB (mL/exposure) and an upper bound of UB (mL/exposure).

bGa(r,λ) represents the gamma distribution with a shape parameter of r and a scale parameter of λ.

Monte Carlo simulation

Because of limited observational data, MCS is typically employed to model the uncertainty or variability of parameters in QMRA (Haas et al. 2014; US EPA 2014; WHO 2016). Generally, the K–S or chi-squared test is first adopted to examine the distribution of the observational data. Some statistical characteristics are subsequently determined according to the examined distribution. Finally, on the basis of the statistical characteristics and distributions of the observational data, multiple simulated values are randomly yielded to propagate the uncertainty or variability of the parameters in QMRA.

The water ingestion volume for coastal swimmers was a uniform distribution with a lower bound of 20 mL/exposure and an upper bound of 50 mL/exposure (i.e., U(20,50)) according to Ashbolt et al. (1997) and was the gamma distribution with a shape parameter of 0.45 and a scale parameter of 60 for adult men (i.e., Ga(0.45,60)), a shape parameter of 0.51 and a scale parameter of 35 for adult women (i.e., Ga(0.51,35)), and a shape parameter of 0.58 and a scale parameter of 55 for children (i.e., Ga(0.58,55)) according to Schets et al. (2011). This study applied MCS to generate 5,000 data for WIV on the basis of U(20,50), Ga(0.45,60), Ga(0.51,35), and Ga(0.58,55). However, most dEnt data exhibited nonparametric distributions. This study adopted nine percentiles (10th, 20th, 30th, 40th, 50th, 60th, 70th, 80th, and 90th percentiles) of observational data to establish the cumulative distribution function (CDF) at each beach (Figure 3). Goovaerts (1997) recommended that the CDF is extrapolated toward nil at the lower tail (y < 0.1) by using a negatively skewed power function, that it is linearly established within two bounding thresholds (xk, xk−1) and estimated probabilities (yk, yk−1) for 0.1≦y≦0.9, and that it is extrapolated toward an infinite upper bound at the upper tail (y > 0.9) by using a hyperbolic function. The CDF functions are expressed as follows (Goovaerts 1997):  
formula
(3)
 
formula
(4)
 
formula
(5)
where wp is 2.5 and wh is 1.5 (Goovaerts 1997). This study followed the aforementioned functions for determining CDF and used MCS to obtain 5,000 simulated data (i.e., y data) according to a uniform distribution between 0 and 1 (i.e., U(0,1)). The enterococci concentrations at each beach (x value) were determined using Equations (3)–(5) according to the established CDF functions and the y data. Moreover, this study used Oracle Crystal Ball Release 11 to execute MCS.
Figure 3

Interpolation and extrapolation of CDF for nonparametric MCS. The nine percentile values (x1, x2, .., x9) were statistically determined according to observed data. The CDF values (y data) were randomly generated using MCS based on U(0,1). The corresponding x data are calculated according to the randomized y data and the power, linear, and hyperbolic CDF models.

Figure 3

Interpolation and extrapolation of CDF for nonparametric MCS. The nine percentile values (x1, x2, .., x9) were statistically determined according to observed data. The CDF values (y data) were randomly generated using MCS based on U(0,1). The corresponding x data are calculated according to the randomized y data and the power, linear, and hyperbolic CDF models.

RESULTS

Distribution estimation of enterococci concentrations and water ingestion volume by using MCS

Because most enterococci concentrations at coastal beaches were nonparametrically distributed, we first determined the 10th, 20th, 30th, 40th, 50th, 60th, 70th, 80th, and 90th percentiles of the observed enterococci concentrations at each beach to examine the distribution. Table 3 lists the enterococci concentrations at the nine percentiles for different beach sites; the enterococci concentrations at the nine percentiles were x1, x2, …, and x9, as shown in Figure 3. Five thousand simulated data were obtained using MCS on the basis of U(0,1) (y data in Figure 3). The corresponding 5,000 simulated enterococci concentrations (x data in Figure 3) were calculated according to the 5,000 y data and established CDF functions (Equations (3)–(5)). Figure 4 shows the histograms of simulated enterococci concentrations at 13 beach sites. Moreover, 5,000 simulated water ingestion volume data of recreational swimming were individually acquired using MCS on the basis of U(20,50), Ga(0.45,60), Ga(0.51,35), and Ga(0.58,55) (Figure 5).

Table 3

Measured enterococci concentrations (MPN/100 mL) at nine percentiles used for nonparametric MCS

Beach sites Percentiles
 
10th 20th 30th 40th 50th 60th 70th 80th 90th 
XJS 0a 10 24.2 
QSW 10 10 23.3 41 108.5 
QD 10 20 20 42 87 
TX 10 10 20 42 68.4 188 
MSG 10 10 30.4 69.6 
XZW 10 20 39 89.6 
QJ 10 10 20 
KD 10 10 10 20 20 36 
SY 10 20 31 79 
JQ 10 10 20 20 38 
WA 10 10 20 
FL 20 31 69 
GYT 20 62.2 
Beach sites Percentiles
 
10th 20th 30th 40th 50th 60th 70th 80th 90th 
XJS 0a 10 24.2 
QSW 10 10 23.3 41 108.5 
QD 10 20 20 42 87 
TX 10 10 20 42 68.4 188 
MSG 10 10 30.4 69.6 
XZW 10 20 39 89.6 
QJ 10 10 20 
KD 10 10 10 20 20 36 
SY 10 20 31 79 
JQ 10 10 20 20 38 
WA 10 10 20 
FL 20 31 69 
GYT 20 62.2 

aLess than a detection limit for enterococci (<10 MPN/100 mL) is regarded as zero.

Figure 4

Histograms of enterococci concentrations reproduced by nonparametric MCS.

Figure 4

Histograms of enterococci concentrations reproduced by nonparametric MCS.

Figure 5

Water ingestion volumes estimated using MCS according to (a) Ashbolt et al. (1997), and Schets et al. (2011) for (b) adult men, (c) adult women, and (d) children.

Figure 5

Water ingestion volumes estimated using MCS according to (a) Ashbolt et al. (1997), and Schets et al. (2011) for (b) adult men, (c) adult women, and (d) children.

Health risks associated with recreational swimming by using QMRA

This study probabilistically determined the distributions of enterococci concentrations and water ingestion volumes used in QMRA because the parameters were relatively uncertain and highly variable. Moreover, the combination of the two parameters was addressed before risk assessment by using the beta-Poisson dose–response function. This study followed the procedure for parameter combination and uncertainty treatment recommended by Goovaerts et al. (2001) and Jang et al. (2006). The joint distributions of CEnt×WIV were obtained for each beach to propagate the uncertainty or variability of QMRA and were subsequently adopted to determine health risks by using the aforementioned function.

Figures 6 and 7 show the box-and-whisker plot for estimating health risks at different beach sites according to the WIV data obtained from Ashbolt et al. (1997) and Schets et al. (2011), respectively. The US EPA (2001, 2014) and WHO (2016) have recommended that the 95th percentile of risk distributions can be used to quantify health risks to human health. Consequently, this study determined the 95th percentile of the risk distributions and considered the risk value at the 95th percentile as a representative risk level at each beach site. Moreover, this study classified the health risks into three categories: high risk (>0.019), medium risk (0.008–0.019), and low risk (<0.008).

Figure 6

Box-and-whisker plot for estimating health risks associated with recreational swimming at beaches according to WIV data obtained from Ashbolt et al. (1997).

Figure 6

Box-and-whisker plot for estimating health risks associated with recreational swimming at beaches according to WIV data obtained from Ashbolt et al. (1997).

Figure 7

Box-and-whisker plot for estimating health risks associated with recreational swimming at beaches according to WIV data obtained from Schets et al. (2011) for (a) adult men, (b) adult women, and (c) children.

Figure 7

Box-and-whisker plot for estimating health risks associated with recreational swimming at beaches according to WIV data obtained from Schets et al. (2011) for (a) adult men, (b) adult women, and (c) children.

For the water ingestion volume reported by Ashbolt et al. (1997), the risks ranged from 0.0020 to 0.0181 at the 13 beaches. No beaches were in the high risk category. TX, QSW, QD, and XZW beaches were in the medium risk category. The remaining beaches were in the low risk category. Meanwhile, for the water ingestion volume reported by Schets et al. (2011), the risks for adult men ranged from 0.0020 to 0.0156 at the 13 beaches. TX, QSW, and QD beaches were in the medium risk category. The risks for adult women ranged from 0.0013 to 0.0099 at the 13 beaches. TX beach was in the medium risk category. The risks for children ranged from 0.0023 to 0.0176 at the 13 beaches. TX, QSW, XZW, and QD beaches were in the medium risk category. The risk categories of the beaches were spatially mapped, as illustrated in Figure 8. Most beaches in the medium risk categories were located in Northwestern Taiwan.

Figure 8

Spatial map for health risks associated with recreational swimming at the 95th percentile at beaches according to WIV data obtained from (a) Ashbolt et al. (1997) and Schets et al. (2011) for (b) adult men, (c) adult women, and (d) children.

Figure 8

Spatial map for health risks associated with recreational swimming at the 95th percentile at beaches according to WIV data obtained from (a) Ashbolt et al. (1997) and Schets et al. (2011) for (b) adult men, (c) adult women, and (d) children.

DISCUSSION

Because most enterococci data at the beaches were nonparametrically distributed, this study used nonparametric MCS to propagate their uncertainty and variability. Comparing the results in Figures 2 and 4 revealed that the simulated enterococci concentrations were similar to those observed at the 13 beach sites, indicating that nonparametric MCS is reliable for characterizing the uncertainty and variability of enterococci concentrations in QMRA.

Because gamma distributions were more positively skewed than uniform distributions (Figure 5), the risks estimated at the 50th and 75th percentiles using water ingestion volume data obtained from Ashbolt et al. (1997) were higher than those from Schets et al. (2011). Moreover, at the 95th percentile, the risks estimated using water ingestion volume data obtained from Ashbolt et al. (1997) were considerably close to those from Schets et al. (2011) for adult men and children.

Although the health risks of recreational swimming at all beaches were below an acceptable benchmark of 0.019 illnesses daily, they approached this benchmark at certain beaches, particularly at TX beach. To avoid public health hazards, suitable environmental and recreational management strategies of beaches should be proposed in Taiwan. Heavy rainfall runoff and sewer overflow transport a substantial amount of urban and agricultural contaminants to coastal oceans; therefore, the coastal oceans were reported to have a significantly high level of pathogenic microorganism load during a precipitation event (Hsu & Huang 2008; Tseng & Jiang 2012). Dorsey (2010) recommended that low-flow diversions and wetland establishment are the effective control strategies to reduce the amount of runoff and sewer overflow with pathogenic bacteria from reaching beaches. Furthermore, according to the online public database on sanitary sewer construction through the Taiwan Construction and Planning Agency, the complete ratio of sanitary sewer systems was 51.48% in Taiwan in 2015. The incomplete sanitary sewer systems in Taiwan were the major cause of serious fecal contamination in waterborne environments because of the discharge of untreated human sewage and agricultural wastewater. To enhance recreational and environmental quality, the sanitary sewer systems in Taiwan should be completed as rapidly as possible. In addition, because heavy rainfall frequently results in increased fecal pollution in marine water, to conserve public health, swimming activities should be temporarily prohibited at the beaches with medium risks after heavy rainfall.

Generally, the estimated risks were greater at the western coastal beaches than at the eastern ones because household, livestock, industry, and river pollution is more severe in Western Taiwan than in Eastern Taiwan (Fan 2001). Moreover, most beaches in the medium risk category were located in Northwestern Taiwan. The nearshore current is the principal force dispersing coastal pollutants (Fan 2001) and upper-ocean currents around Taiwan primarily move northward toward the East China Sea owing to the Kuroshio Current (Liang et al. 2003), resulting in the northward movement of currents carrying coastal fecal contaminants and significant increases of the enterococci density at the northwestern coastal beaches.

This study characterized an inferred health risk of recreational swimming at Taiwanese beaches using QMRA. However, real disease surveys for beach swimmers were lacking in Taiwan. Thus, to compare estimated health risks with real disease occurrence, infectious cases for swimmers self-reported in questionnaires should be recorded at Taiwanese beaches in future studies.

CONCLUSION

This study focused on estimating health risks of recreational swimming at 13 Taiwanese beaches by using QMRA. The QMRA framework can provide an excellent insight into the public health issue of recreational swimming at beaches because of contaminated marine environments. Because most enterococci concentrations are nonparametrically distributed, this study proposes nonparametric MCS to propagate the parameter uncertainty and variability. Moreover, the probabilistic risk findings at beaches can serve as a guideline for government administrators to establish safe recreational water management strategies in Taiwan. The study results reveal that although the health risks associated with recreational swimming did not exceed an acceptable benchmark of 0.019 illnesses daily at all beaches, they approached to this benchmark at certain beaches, particularly at TX beach. The health risks to recreational swimmers are typically greater at the western Taiwanese beaches than at the eastern ones. Furthermore, beach sites with medium risks are mainly located in Northwestern Taiwan because of current movements. Accordingly, to avoid public health hazards, suitable environmental and recreational management strategies should be implemented at medium risk beaches.

ACKNOWLEDGEMENTS

The authors would like to thank the Taiwan Environmental Protection Administration for generously providing the data on enterococci concentrations at 13 Taiwanese beaches and the Taiwan Ministry of Science and Technology for financially supporting this research under Contract No. MOST 105-2410-H-424-015.

REFERENCES

REFERENCES
Ashbolt
,
N. J.
,
Reidy
,
C.
&
Haas
,
C. N.
1997
Microbial health risk at Sydney's coastal bathing beaches
. In:
Proceedings of the 17th Australian Water and Wastewater Association Meeting
,
vol. 2
.
Australian Water and Wastewater Association
,
Melbourne
, pp.
104
111
.
Ashbolt
,
N. J.
,
Schoen
,
M. E.
,
Soller
,
J. A.
&
Roser
,
D. J.
2010
Predicting pathogen risks to aid beach management: the real value of quantitative microbial risk assessment (QMRA)
.
Water Research
44
(
16
),
4692
4703
.
Craig
,
D. L.
,
Fallowfield
,
H. J.
&
Cromar
,
N. J.
2003
Effectiveness of guideline faecal indicator organism values in estimation of exposure risk at recreational coastal sites
.
Water Science and Technology
47
(
3
),
191
198
.
Dickinson
,
G.
,
Lim
,
K. Y.
&
Jiang
,
S. C.
2013
Quantitative microbial risk assessment of pathogenic vibrios in marine recreational waters of Southern California
.
Applied and Environmental Microbiology
79
(
1
),
294
302
.
Fan
,
K. L.
2001
Some coastal environmental problems in Taiwan
.
Acta Oceanographica Taiwanica
39
,
1
10
.
Gerba
,
C. P.
,
Rose
,
J. B.
,
Haas
,
C. N.
&
Crabtree
,
K. D.
1996
Waterborne rotavirus: a risk assessment
.
Water Research
30
(
12
),
2929
2940
.
Goovaerts
,
P.
1997
Geostatistics for Natural Resources Evaluation
.
Oxford University Press
,
New York
, pp.
259
368
.
Goovaerts
,
P.
,
Semrau
,
J.
&
Lontoh
,
S.
2001
Monte Carlo analysis of uncertainty attached to microbial pollutant degradation rates
.
Environmental Science & Technology
35
,
3924
3930
.
Haas
,
C. N.
2015
Microbial dose response modeling: past, present, and future
.
Environmental Science & Technology
49
(
3
),
1245
1259
.
Haas
,
C. N.
,
Thayyar-Madabusi
,
A.
,
Rose
,
J. B.
&
Gerba
,
C. P.
2000
Development of a dose-response relationship for Escherichia coli O157:H7
.
International Journal of Food Microbiology
1748
,
153
159
.
Haas
,
C. N.
,
Rose
,
J. B.
&
Gerba
,
C. P.
2014
Quantitative Microbial Risk Assessment
,
2nd edn
.
Wiley
,
New York
, pp.
267
321
.
Health Canada
2012
Guidelines for Canadian Recreational Water Quality
.
3rd edn
.
Water, Air and Climate Change Bureau, Healthy Environments and Consumer Safety Branch, Health Canada
,
Ottawa
,
Ontario
, pp.
26
.
Hsu
,
B. M.
&
Huang
,
Y. L.
2008
Intensive water quality monitoring in a Taiwan bathing beach
.
Environmental Monitoring and Assessment
144
(
1
),
463
468
.
Liang
,
W. D.
,
Tang
,
T. Y.
,
Yang
,
Y. J.
,
Ko
,
M. T.
&
Chuang
,
W. S.
2003
Upper-ocean currents around Taiwan
.
Deep Sea Research Part II: Topical Studies in Oceanography
50
(
6–7
),
1085
1105
.
Ming
,
H. X.
,
Zhu
,
L.
,
Feng
,
J. F.
,
Yang
,
G.
&
Fan
,
J. F.
2014
Risk assessment of rotavirus infection in surface seawater from Bohai Bay, China
.
Human and Ecological Risk Assessment
20
,
929
940
.
Schets
,
F. M.
,
Schijven
,
J. F.
&
de Roda Husman
,
A. M.
2011
Exposure assessment for swimmers in bathing waters and swimming pools
.
Water Research
45
,
2392
2400
.
Schoen
,
M. E.
&
Ashbolt
,
N. J.
2010
Assessing pathogen risk to swimmers at non-sewage impacted recreational beaches
.
Environmental Science & Technology
44
(
7
),
2286
2291
.
Soller
,
J. A.
,
Schoen
,
M. E.
,
Bartrand
,
T.
,
Ravenscroft
,
J. E.
&
Ashbolt
,
N. J.
2010
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
.
Steyn
,
M.
,
Jagals
,
P.
&
Genthe
,
B.
2004
Assessment of microbial infection risks posed by ingestion of water during domestic water use and full-contact recreation in a mid-southern African region
.
Water Science and Technology
50
(
1
),
301
308
.
Taiwan Environmental Protection Administration (EPA)
2016
Environmental Water Quality Information
.
Environmental Protection Administration, Executive Yuan
,
Taiwan
. .
Till
,
D.
,
McBride
,
G. .
,
Ball
,
A.
,
Taylor
,
K.
&
Pyle
,
E.
2008
Large-scale freshwater microbiological study: rationale, results and risks
.
Journal of Water and Health
6
(
4
),
443
460
.
US Environmental Protection Agency (EPA)
1986
Ambient Water Quality Criteria for Bacteria – 1986
.
EPA 440-5-84-002
.
US Environmental Protection Agency
,
Washington, DC
, pp.
15
.
US Environmental Protection Agency (EPA)
2001
Risk Assessment Guidance for Superfund (RAGS) Volume III – Part A: Process for Conducting Probabilistic Risk Assessment. Office of Emergency and Remedial Response
.
US Environmental Protection Agency
,
Washington, DC
, pp.
3-1
3-27
.
US Environmental Protection Agency (EPA)
2012
Recreational Water Quality Criteria. Office of Water
.
United States Environmental Protection Agency
,
Washington, DC
,
(EPA-820-F-12-058)
14
pp.
US Environmental Protection Agency (EPA)
2014
Microbiological risk assessment (MRA) tools, methods, and approaches for water media
.
Office of Science and Technology Office of Water US Environmental Protection Agency
,
Washington, DC
, (
EPA-820-R-14-009
) pp.
90
94
and pp. 104–110
.
World Health Organization (WHO)
2003
Guidelines for Safe Recreational Water Environments
.
Vol. 1
.
Coastal and Fresh Waters
.
World Health Organization
,
Geneva
,
Switzerland
, pp.
82
87
.
World Health Organization (WHO)
2016
Quantitative Microbial Risk Assessment: Application for Water Safety Management
.
World Health Organization
,
Geneva
,
Switzerland
, pp.
171
179
.