https://orcid.org/0000-0002-1814-7487
Figures
Abstract
At many public beaches, routine monitoring of beach water quality using fecal indicator bacteria is conducted to evaluate the risk of recreational water illness. Results from water sample analysis can take over 24-hr, which may no longer accurately reflect current water quality conditions. This study aimed to assess which combination of environmental factors best predicts fecal contamination (E. coli) levels at two of the most popular beaches on Lake Winnipeg, Manitoba (Gimli and Grand Beach), by linking water quality data and publicly available environmental data from 2007 to 2021. We developed separate mixed effects models for each beach for two outcomes, linear (continuous log-transformed E. coli concentration) and categorical (200 CFU/100 ml threshold), to explore differences in the predictors of E. coli concentrations and exceedances of the provincial health risk threshold, respectively. We used a Directed Acyclic Graph to choose which predictor variables to include in the models. For both beaches, we identified clustering of the E. coli outcomes by year, suggesting year-specific variation. We also determined that extreme weather days, with higher levels of rainfall in the preceding 48-hr, previous day average air temperature, and previous day E. coli concentration could result in a higher probability of E. coli threshold exceedances or higher concentrations in the water bodies. In Grand Beach, we identified that days with lower average UV levels in the previous 24-hr and antecedent dry days could result in a higher probability of E. coli threshold exceedances or higher concentrations. The findings can inform possible trends in other freshwater settings and be used to help develop real-time recreational water quality predictive models to allow more accurate beach management decisions and warrant enhancement of beach monitoring programs for extreme weather events as part of the climate change preparedness efforts.
Citation: Desta BN, Sanchez J, Heasley C, Young I, Tustin J (2024) Environmental factors associated with Escherichia coli concentration at freshwater beaches on Lake Winnipeg, Manitoba, Canada. PLOS Water 3(4): e0000143. https://doi.org/10.1371/journal.pwat.0000143
Editor: Ricardo Santos, Universidade Lisboa, Instituto superior Técnico, PORTUGAL
Received: May 19, 2023; Accepted: February 26, 2024; Published: April 4, 2024
Copyright: © 2024 Desta et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The environmental data presented in this study are openly available online from the Environment and Climate Change Canada historical data repository (https://climate.weather.gc.ca/climate_data/daily_data_e.html?hlyRange=2006-09-22%7C2022-05-10&dlyRange=2006-09-19%7C2022-05-10&mlyRange=2007-05-01%7C2007-11-01&StationID=45007&Prov=MB&urlExtension=_e.html&searchType=stnName&optLimit=yearRange&StartYear=1840&EndYear=2022&selRowPerPage=25&Line=3&searchMethod=contains&Month=5&Day=13&txtStationName=gimli&timeframe=2&Year=2022) (accessed on 5 May 2022), the Government of Canada’s Water Office (https://wateroffice.ec.gc.ca/report/historical_e.html?stn=05SA003) (accessed on 5 May 2022), the department of Oceans and Fisheries buoy (https://www.meds-sdmm.dfo-mpo.gc.ca/isdm-gdsi/waves-vagues/data-donnees/index-eng.asp) (accessed on 25 April 2022), and the NASA database of solar and meteorological parameters (https://power.larc.nasa.gov/data-access-viewer/) (accessed on 25 April 2022). Manitoba E. coli data can be requested from the Environment and Climate (Government of Manitoba; Phone number: 1-800-282-8069 (ext. 0002) and Email: waterquality@gov.mb.ca). The study’s analysis SAS code and R script files are available at https://github.com/bndesta/ManitobaRecWater.
Funding: This research was funded by the Public Health Agency of Canada, grant number 2021- HQ-000017 (funder’s website: https://www.canada.ca/en/public-health.html, recipients: IY, JT). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Beachgoers who come into contact with water contaminated with fecal pollution are at increased risk for recreational water illnesses [1]. At some public beaches, beach advisories and closures are issued following routine monitoring, i.e., where fecal indicator bacteria (FIB) are monitored and observed concentrations indicate a threat to public health [2–4]. Regular water quality monitoring involves measuring the FIB concentration, using threshold levels, and comparing results against threshold levels as specified in the respective guidelines to determine follow-up actions to protect beachgoer health [4, 5]. A review study indicated that increased concentration of the FIB (E. coli) counts in recreational freshwater bodies is associated with increased risk of recreational water illnesses such as acute gastrointestinal illnesses [6].
Many countries, including Canada, the European Union Member States, and the United States, use guidelines (e.g., the European Bathing Water Directive and the US Recreational Water Quality Guideline) to monitor recreational water quality [4, 7–9]. Escherichia coli is the main FIB used for fresh recreational waters in those countries [4, 7–9]. According to the Canadian recreational water quality guideline, geometric mean concentrations >200 CFU/100 ml of E. coli (calculated based on samples collected at one time at multiple sites) are considered a high-risk threshold for acute gastrointestinal illness [10]. However, a new version of the guideline is under development (2022), establishing a new beach action value of 235 CFU/100 ml for any sample [9]. Based on this routine monitoring approach, the beach advisories rely on the previous day’s testing results, as the culture-based sampling and laboratory procedures take 1–2 days to process [3, 11]. As the water quality could change within hours, these beach advisories may no longer accurately reflect current water quality conditions [5]. Multiple pollution sources and various environmental factors play a role in the fecal contamination of recreational waters [12]. Numerous studies have found that fecal contamination of fresh lake water is associated with recent rainfall [13–15]. The growth and survival of FIB in freshwater can also be determined by air and water temperature [14, 16]. The FIB concentration in fresh lake water also tends to decrease with increased ultraviolet (UV) solar radiation [14]. As climate change is expected to influence all of these parameters, these trends may impact the risk of exposure to recreational water illness in the future [5]. However, the relative importance of these factors may differ regionally and by beach. Identifying such environmental predictors of fresh lake water quality would complement water quality monitoring activities to address the gap in untimely beach advisories of E. coli concentration. In Manitoba, Williamson et al. (2004) conducted a study at Lake Winnipeg Beaches and found a significant correlation between E. coli densities at Gimli and West Grand Beaches with daily water level changes; however, in recent years, limited studies have been conducted on beaches to evaluate environmental predictors of E. coli concentration [3, 17]. This study aimed to determine the association between environmental factors and E. coli concentration at two popular beaches on Lake Winnipeg, Manitoba, by linking water quality data and publicly available environmental data. The primary purpose of this analysis was to assess which combination of environmental factors best predicts fecal contamination (E. coli) levels at these two beaches, with results used to inform future predictive modelling, climate change preparedness, and water quality monitoring in the province.
Materials and methods
Study area
We examined water quality at two public beaches on Lake Winnipeg: Grand Beach and Gimli Beach. The beaches were chosen based on their popularity among beachgoers in the province. Located in Grand Beach Provincial Park, Grand Beach comprises an East and West beach located 80 km northeast of Winnipeg; it is Manitoba’s most popular beach. While the beach only makes up part of the provincial park, thousands of people visit the park on the most active days in the summer. Gimli Beach, located on the west side of Lake Winnipeg in the town of Gimli, is another popular beach destination. Gimli also has a commercial fishing industry and a marina near the beach. Located on opposite sides of the Lake, these beaches allow for an exploration of water quality across the popular sites of Lake Winnipeg (Fig 1).
The data on the map was plotted on ArcGIS using a base map shapefile provided by World Topo by ESRI [18].
Water sample collection
Water quality sampling is operated by the Department of Environment and Climate of the Manitoba government, which monitored around 60 beaches in Manitoba and provided the data used in this study. The beaches monitored are chosen based on recreational intensity and historical bacterial data. Water quality data were collected at both beaches from 2003 to 2021. Sampling typically begins in late May and continues until late August or early September. As of 2015, both beaches are monitored at least weekly or more often if re-sampling is required; from 2003 through 2014, there was more frequent (i.e., daily) sampling to gather continuous data. Five replicate samples are collected at equidistant locations across the length of the beach, with the geometric mean of the five samples reported. Grand Beach is divided into West and East Beach; both beaches have five replicate sample locations. Water is sampled at approximately thigh-deep or 1-metre-deep, based on a standardized procedure from the Manitoba government, and then shipped to the laboratory (ALS Laboratories in Winnipeg) in a cooler with ice [19]. Beaches are re-sampled immediately when the lower tail of a 60% confidence interval around the geometric mean of all the E. coli replicate samples exceeds 200 CFU/100 mL. Before 2008, fecal coliforms were tested in addition to E. coli, using the same criteria. Since 2008, swimming advisory signs have been posted at Lake Winnipeg beaches if the original samples meet the abovementioned requirements for re-sampling.
Environmental data
Starting in 2005, the province collected air and water temperatures at the time of water sample collection. Beginning in 2010, wind direction, wind speed, wave conditions, cloud conditions, presence of an algal bloom, bird count, and other qualitative observations were collected. The environmental variables collected with the E. coli measurement contained large proportions of missing values for most study years, including environmental variables (temperature and wind measurements) and observation field notes (wave condition, number of shorebirds and people) (See S1 Table). Thus, we used publicly available data, made available to the open public for research and other related purposes, from other sources for this study. We collected daily air temperature and precipitation data from the nearest Environment and Climate Change Canada weather station to both beaches, located in Gimli, from 2006 to 2021 [20]. We also collected data on the water level in Lake Winnipeg from a water level station at Victoria Beach, north of Grand Beach, operated by the Government of Canada’s Water Office since 2006 [21]. We collected wave height and wind speed from a Department of Oceans and Fisheries buoy in the middle of the south basin (i.e., not near the sampling locations), with data going back to 2012 [22]. We obtained UV index data from a National Aeronautics and Space Administration (NASA, of the United States) database of solar and meteorological parameters, with data starting in 2006 [23].
Statistical analysis
We used SAS 9.4 (SAS Institute, Cary, NC) and R version 4.2.2 to analyze data. In our exploratory analysis, we checked for normality and linearity of all variables and natural log-transformed the E. coli geometric means to reduce its skewness. We excluded missing data for specific variables from the analysis of that variable. Our research spanned from 2007 to 2021 to include the most complete data on environmental factors. We merged the water quality data collected separately for the east and west Grand beaches since most of the environmental predictors (in this analysis) are less likely to vary between the two beaches. We found no statistically significant difference in E. coli geometric mean concentration between the two sides (assessed via a t-test, equal variance), where a slight fluctuation between the two sides was detected in most of the years by the visual comparison of the annual mean of E. coli geometric mean concentration across the years (see S1 Fig). As we expected E. coli concentration to cluster at the year level, we chose to fit a mixed effects model to address the multi-level data structure. We developed separate mixed effects models for each beach for two outcomes, linear (continuous natural log-transformed E. coli concentration) and categorical (200 CFU/100 ml threshold), to explore differences in the predictors of E. coli concentrations and exceeding the provincial recreational water quality objective value, respectively.
The environmental factors considered for analysis included same-day values of mean air temperature, total precipitation, speed of the maximum gust of wind, water level above a standard (mean sea level), average UV index, and number of days since the last rainfall (antecedent dry days). We developed and used a Directed Acyclic Graph (DAG; S2 Fig) with path analysis (which is used to determine the effects of a set of predictors on an outcome variable through multiple pathways [24]) to choose which of these variables to include in the models [25]. We excluded water level from the final models, following the path analysis modelling outcome (S2 and S3 Tables), which indicated the mediator effect (needs no controlling [26]) of water level in the relation of rainfall and air temperature with E. coli concentration (see S3 and S4 Figs). Additionally, the speed of the maximum gust of wind variable was ultimately excluded from the final models due to its small number of observations (i.e., many days had missing values; See S1 Table). We also assessed values from the previous day to evaluate the temporality of mean air temperature, geometric mean E. coli concentration, and average UV index. A sum of values for the last two days from the day of water sample collection was included for rainfall (i.e., 48-hr cumulative rainfall in the prior two days). As a sensitivity analysis to assess the usefulness of the variable excluded from our final models due to limited number of observations, we added two variables (24-hr speed of max wind gust and mean wave height) and ran the models. We conducted correlation analysis (see S4 Table) and produced scatterplots (with a regression line; see S5–S14 Figs) of the predictor variables included in the final models against the log E. coli geometric mean values. The assumption of independence was tested via multicollinearity diagnostics, where the variance inflation factor (VIF) corresponding to each variable included in the models should be below two (conventionally, a small value showing little evidence for multicollinearity) [27]. A two-way interaction effect of selected variables (temperature and rainfall; antecedent dry days and average UV index) was tested on the E. coli outcomes. We developed intercept-only models (containing year as the random effect portion) without including fixed effects to examine the presence of a within-group dependence of observations. We assessed the intraclass correlation coefficient (ICC) values and the respective model chi-square test to check the appropriateness of the mixed effects approach for this analysis [28, 29]. The final models were chosen to fit the data for the two responses, linear and categorical thresholds, containing all variables (Table 1) regardless of their statistical significance, based on the DAG (S2 Fig) [25]. In the logistic mixed regression model (categorical response), the average marginal effects of the predicted probability of exceeding the 200 CFU/100 ml E. coli recreational water quality objective by different levels (minimum, median, 95th and 99th percentiles) of significant predictors were plotted. These plots allow us to predict the marginal effect of predictor variables on the exceedance predicted probability while considering all other predictor variables in the final models.
Results
Descriptive data
A total of 1625 calculated E. coli geometric mean values between 2007 and 2021 across both of the study beaches were included. The E. coli concentration varied substantially across the years (Fig 2). Gimli Beach had a higher concentration in all the study years than Grand Beach except in 2019, where the highest overall mean annual geometric E. coli concentration was observed (Fig 2).
Gimli Beach had a higher overall percentage of days per season that exceeded the 200 CFU/mL health risk threshold compared to Grand Beach (Fig 3). The exceedances of thresholds varied across the years at both beaches (Fig 3).
Mixed effects models
The models only containing year as a random effect of the continuous measure of E. coli concentration for Gimli and Grand Beach showed ICCs of 0.098 (p-value: 0.011) and 0.133 (p-value: 0.008), respectively. These ICC values, along with Fig 1, suggested considerable clustering of observations at the year level and indicated the appropriateness of the multi-level modelling approach. Similarly, including year only in the model as a random effect of the categorical outcome (200 CFU/100 ml threshold) presented ICCs of 0.227 (p-value: 0.044) and 0.116 (p-value: 0.043) in Gimli and Grand Beach, respectively, also supported the multi-level modelling approach.
Gimli Beach
In the Gimli Beach models (linear and categorical), the fixed portion showed that 48-hr cumulative rainfall, 24-hr mean air temperature, and 24-hr log geometric mean of E. coli were positively associated with E. coli in both models (Table 2). All of the interaction tests were insignificant.
Fig 4 showed that the predicted marginal effect (after taking into consideration the impact of other variables in the final logistic mixed effect Gimli model) at extreme values (95th and 99th percentile; 22.8 and 45.0 mm, respectively) of 48-hr cumulative rainfall results in a higher probability of exceeding the 200 CFU/100 E. coli threshold compared to the minimum or median values (0.0 and 0.5 mm, respectively). Similarly, Fig 5 presented the predicted probability at extreme values (95th and 99th percentile; 23.1 and 25.2°C, respectively) of 24-hr mean air temperature results in a higher probability of exceeding the 200 CFU/100 E. coli concentration threshold than at the minimum or median values (0.0 and 16.7°C, respectively), while holding other predictors constant. Higher values (95th and 99th percentile; 658.5 and 2392.3 CFU/100ml, respectively) of the 24-hr geometric mean of E. coli corresponds to a predicted higher probability of exceeding the 200 CFU/100 E. coli threshold compared to the minimum or median values (1.0 and 35.2 CFU/100ml, respectively), while controlling for other predictors in the model (Fig 6).
(Here, the four lines correspond to the different values of 48-hr rainfall, and the respective colour-shaded area indicates the prediction intervals for each value).
(Here, the four lines correspond to the different 24-hr mean air temperature values, and the respective colour-shaded area indicates the prediction intervals for each value).
(Here, the four lines correspond to the different values of the 24-hr log geometric mean of E. coli, and the respective colour-shaded area indicates the prediction intervals for each value).
Grand Beach
The fixed portion of the Grand Beach models presented that the 24-hr air average temperature and 24-hr log geometric mean of E. coli were positively associated with E. coli in both the linear and categorical models. In contrast, 24-hr average UV index was negatively associated with both outcomes (Table 3). An increased 48-hr cumulative rainfall was positively associated with only E. coli concentrations as a linear response, while an increased number of antecedent dry days since the last rain was negatively associated with E. coli concentrations (Table 3). All of the interaction tests were insignificant.
Fig 7 presented that the predicted marginal effect at extreme values (95th and 99th percentile; 23.1 and 25.0°C, respectively) of 24-hr mean air temperature results in a higher probability of exceeding the 200 CFU/100 E. coli threshold compared to the minimum or median values (0.0 and 17.1°C, respectively), after taking into considerations of the effect of other variables in the final logistic mixed effect Grand model. Likewise, higher values (95th and 99th percentile; 247.2 and 897.8 CFU/100 ml, respectively) of the 24-hr geometric mean of E. coli corresponds to a predicted higher probability of exceeding the 200 CFU/100 E. coli threshold compared to the minimum or median values (0.0 and 16.4 CFU/100 ml, respectively), while controlling for other predictors in the model (Fig 8). Fig 9 illustrates that the predicted probability at extreme values (95th and 99th percentile; 2.0 and 2.2 UV index, respectively) of 24-hr mean UV index results in a lower probability of exceeding the 200 CFU/100 E. coli concentration threshold than at the minimum or median values (0.0 and 1.4 UV index, respectively), while holding other predictors constant.
(Here, the four lines correspond to the different 24-hr mean air temperature values, and the respective colour-shaded area indicates the prediction intervals for each value).
(Here, the four lines correspond to the different values of the 24-hr log geometric mean of E. coli, and the respective colour-shaded area indicates the prediction intervals for each value).
(Here, the four lines correspond to the different values of the 24-hr mean UV index, and the respective colour-shaded area indicates the prediction intervals for each value).
The sensitivity analysis that included variables (24-hr speed of max wind gust and mean wave height) changed the relationship and effect shown in the final models (see S5 and S6 Tables). Still, the results were less credible, given that the sensitivity analysis models used only one-third of the observations (Gimli Beach Models = 139 and Grand Beach Models = 216) for the final models.
Discussion
This study aimed to determine the association between environmental factors and E. coli concentrations (and exceeding the provincial recreational water quality thresholds) at two beaches on Lake Winnipeg, Manitoba. Higher exceedances (of the 200 CFU/ml geometric mean threshold) of E. coli concentration were seen in recent years at both beaches. Still, they were notably more elevated at Gimli than Grand Beach. Isolated incidents such as sewage spills [30, 31] would provoke increased emergency sampling and result in a greater number of exceedance results, although the events were controlled. The initial models, which included year only as a random effect, identified clustering of the E. coli outcomes at the year level. This clustering was also noted in the average annual geometric mean values at each beach and signifies the importance of yearly fluctuations in E. coli levels and the influence of contamination sources.
Increased rainfall in the preceding 48hr was positively associated with E. coli outcomes at both beaches, except for the categorical outcome at Grand Beach. In previous studies, increased E. coli concentration in recreational water has been associated with preceding rainfall, which could be linked to its impact on increasing surface water runoff, urban stormwater discharge, and water level changes [14, 32]. Rainfall-induced runoff from streambed sediments could cause contamination of beach water due to the runoff generation and sediment mobilization that occurs after rainfall initiation [33]. Regardless of the precipitation, E. coli concentration in stormwater runoff was reported to exceed the recreational water quality standards in a US study, indicating the runoff-related fecal contamination of beach water [34].
Consistent with a similar study in Ontario, Canada [14], increased previous day average air temperature was positively associated with E. coli concentration at both beaches and for both linear and categorical outcomes. Warmer air temperature (a) could facilitate bacterial survival, growth, and reproduction, including E. coli [35], and (b) could also result in more beach visitors, stirring up the sediment or contributing E. coli to the beach water [36]. Higher temperature with extended sunlight facilitates vegetable growth, thereby increasing the blockage and bactericidal effect from UV exposure and increasing the survival rate of microbes [37]. Many people visit beaches during hot days for relief, thereby allowing direct fecal deposition from humans at the beach, creating local resuspension of the microbes there, and facilitating the transportation of the microbes from the streambed to the beach water [36].
In Grand Beach only, average UV levels in the previous 24-hr was negatively associated with E. coli when tested for both the linear and categorical outcome. Other studies reported a similar negative association [13, 38], which aligns with what has been known of UV radiation’s bactericidal and cell growth inhibitory effects [39, 40]. In addition, the survival of E. coli bacteria depends on the dosage of UV radiation [41, 42], where the E. coli concentration in water was reported to be lower during day time (midday, with maximum daily UV light level) than at night time. This suggests that there might be value in sampling water quality at a few different times in a day when bathers could be present to evaluate daily fluctuations in outcomes, instead of just once per day [38].
An increased number of antecedent days without rainfall was negatively associated with E. coli concentration at Grand Beach, which could be due to the decrease in survival of these indicator bacteria the longer they stayed in the environment before being washed (surface runoff) into the recreational water bodies [43]. Other studies have reported a positive or no association of antecedent dry days with E. coli levels in freshwater [43–45], indicating its effect on water quality may vary. Predictive model development for beach water quality monitoring should consider including a variable on antecedent dry days to evaluate its impact, especially when no data are available to measure nearby surface water (e.g., river, stream) runoff directly.
Consistent with a prior study in Ontario [14], previous day E. coli concentration was positively associated with E. coli concentration and threshold exceedances at both beaches. This could indicate the persistence of contamination levels in recreational freshwater, given that other environmental factors could still play a role in impacting the water quality. Like other government agencies that monitor recreational water quality across Canada, the Manitoba Environment and Climate Department (Government of Manitoba) uses previous-day geometric mean E. coli to inform current decision-making on beach posting. Thus, developing predictive models containing previous day E. coli concentration with other daily environmental data would enable real-time choices to be made by the responsible authorities of current recreational water quality status. Additionally, recent advances in molecular and rapid testing for FIB are now recommended in the proposed new Health Canada guidelines [9]. Manitoba and other authorities could also consider adopting those methods for faster turnaround.
Moreover, the marginal effects of the predicted probability plots in this study indicated that all the previously-discussed environmental predictors have varying impacts at different levels on exceedances of the 200 CFU/100 E. coli threshold in the water. The extreme values (rainfall in the preceding 48-hr, the previous day average air temperature, and the previous day E. coli concentration) increase the probability of exceeding the E. coli threshold. In contrast, the minimum average UV levels in the last 24-hr result in higher exceedances. Overall, these results suggest that extreme values of these environmental predictors are essential and could significantly impact water quality and recreation water illness risks. Additionally, as severe weather becomes more likely with climate change, we might need more enhanced monitoring of such events as part of a beach monitoring program in the summer as a climate change preparedness measure.
The present study has some limitations. Potential predictor variables, obtained and linked from multiple sources, were excluded from the final model mainly due to data incompleteness. The two variables (i.e., direction and speed of maximum gust of wind) obtained from Environment and Climate Change Canada were not measured for several days (range of missed days: 10–54) across each study year and were not reported at all in one of the study years. Moreover, the variables obtained from Fisheries and Oceans Canada, including wave height, wind speed and direction, gust speed, and water surface temperatures, were only available for 9/15 study years. Manitoba Environment, Climate and Parks collected several environmental variables, including wind speed and direction, air and water temperature, wave size, presence and count of shorebirds, and the number of people on the beach; however, there were significant discrepancies and irregularities on the measurements, with very low (or no) numbers of observations and measurements in almost all of the study years. A more consistent collection of such variables is recommended in future years to support historical and trend analyses and future predictive modelling. Other environmental factors known to influence FIB levels, such as turbidity, shore wave height, and waterfowl presence [14], were not available but should be considered for future data collection as these factors could act as predictors of water quality. Furthermore, we used proximity and data availability when selecting the weather and buoy stations. However, data from these stations might not represent the actual beach situation due to possible local variations in weather patterns. Notwithstanding these limitations, we were able to utilize the available and usable information in this study. Also, we employed multiple analysis techniques, including path analysis, multilevel modelling, and marginal effects plots, to ensure the credibility of the findings.
In conclusion, the identified environmental factors associated with recreational water quality at two popular beaches in Manitoba could help to inform beach management decisions to safeguard the health of beachgoers. We examined the predictors of threshold exceedance and the predictors showing a linear association with E. coli concentration on the two beaches. We presented that extreme weather days, with higher levels of rainfall in the preceding 48-hr, previous day average air temperature and previous day E. coli concentration could result in a higher probability of E. coli threshold exceedances or higher E. coli concentration in the water bodies. Also, in Grand Beach, we identified that days with lower average UV levels in the previous 24-hr and antecedent dry days could result in a higher probability of E. coli threshold exceedances or higher E. coli concentration. These findings warrant enhancing beach monitoring programs for extreme weather events as part of climate change preparedness efforts. In both beaches, E. coli concentration clustered at the year level, suggesting year-specific variation; beach management activities and future studies could benefit from a holistic approach that explores and considers year-specific phenomena linked with beach water quality. The findings of this study could inform possible trends in other freshwater settings and be used to help develop real-time recreational water quality predictive models to allow more accurate beach management decisions.
Supporting information
S1 Table. Data completeness by variable corresponding to the geometric mean of E. coli (n = 1625) of Manitoba Beaches, Canada, 2007–2021.
https://doi.org/10.1371/journal.pwat.0000143.s001
(DOCX)
S2 Table. Model fit of Manitoba beaches path models, 2007–2021.
https://doi.org/10.1371/journal.pwat.0000143.s002
(DOCX)
S3 Table. Estimates of mediation of Manitoba beaches path models, 2007–2021.
https://doi.org/10.1371/journal.pwat.0000143.s003
(DOCX)
S4 Table. Estimates of correlation of each predictor variable with log E. coli included in the final models of the two Manitoba beach models, 2007–2021.
https://doi.org/10.1371/journal.pwat.0000143.s004
(DOCX)
S5 Table. Linear and logistic mixed-effects models with environmental factors (including variables regardless of completeness) as the fixed effects, E. coli concentration (linear and categorical) as the outcome, and year as a random effect variable at Gimli Beach, Manitoba, 2007–2021.
https://doi.org/10.1371/journal.pwat.0000143.s005
(DOCX)
S6 Table. Linear and logistic mixed-effects models with environmental factors (including variables regardless of completeness) as the fixed effects, E. coli concentration (linear and categorical) as the outcome, and year as a random effect variable at Grand Beach, Manitoba, 2007–2021.
https://doi.org/10.1371/journal.pwat.0000143.s006
(DOCX)
S1 Fig. Mean annual geometric mean at Manitoba beaches, 2007–2021.
https://doi.org/10.1371/journal.pwat.0000143.s007
(TIF)
S2 Fig. Directed Acyclic Graph (DAG) of the relationship between variables affecting E. coli.
https://doi.org/10.1371/journal.pwat.0000143.s008
(TIF)
S3 Fig. Path diagram of Gimli beach, 2007–2021. Grey-highlighted estimates indicate a non-significant path at 0.05.
https://doi.org/10.1371/journal.pwat.0000143.s009
(TIF)
S4 Fig. Path diagram of Grand beach, 2007–2021. Grey-highlighted estimates indicate a non-significant path at 0.05.
https://doi.org/10.1371/journal.pwat.0000143.s010
(TIF)
S5 Fig. Scatter plot (with regression line) of 48-hr rainfall with log E. coli at Gimli Beach, 2007–2021.
https://doi.org/10.1371/journal.pwat.0000143.s011
(TIF)
S6 Fig. Scatter plot (with regression line) of 24-hr mean air temperature with log E. coli at Gimli Beach, 2007–2021.
https://doi.org/10.1371/journal.pwat.0000143.s012
(TIF)
S7 Fig. Scatter plot (with regression line) of 24 h log geometric mean of E. coli with log E. coli at Gimli Beach, 2007–2021.
https://doi.org/10.1371/journal.pwat.0000143.s013
(TIF)
S8 Fig. Scatter plot (with regression line) of 24 h mean UV with log E. coli at Gimli Beach, 2007–2021.
https://doi.org/10.1371/journal.pwat.0000143.s014
(TIF)
S9 Fig. Scatter plot (with regression line) of days since rain with log E. coli at Gimli Beach, 2007–2021.
https://doi.org/10.1371/journal.pwat.0000143.s015
(TIF)
S10 Fig. Scatter plot (with regression line) of 48-hr rainfall with log E. coli at Grand Beach, 2007–2021.
https://doi.org/10.1371/journal.pwat.0000143.s016
(TIF)
S11 Fig. Scatter plot (with regression line) of 24-hr mean air temperature with log E. coli at Grand Beach, 2007–2021.
https://doi.org/10.1371/journal.pwat.0000143.s017
(TIF)
S12 Fig. Scatter plot (with regression line) of 24 h log geometric mean of E. coli with log E. coli at Grand Beach, 2007–2021.
https://doi.org/10.1371/journal.pwat.0000143.s018
(TIF)
S13 Fig. Scatter plot (with regression line) of 24 h mean UV with log E. coli at Grand Beach, 2007–2021.
https://doi.org/10.1371/journal.pwat.0000143.s019
(TIF)
S14 Fig. Scatter plot (with regression line) of days since rain with log E. coli at Grand Beach, 2007–2021.
https://doi.org/10.1371/journal.pwat.0000143.s020
(TIF)
Acknowledgments
We thank Dylan Lyng, Kelly-Anne Richmond and Andrew Burton for providing the water quality parameter data collected by the Department of Environment and Climate in the Manitoba government. We also would like to specifically thank Dylan Lyng for reviewing and editing the manuscript.
References
- 1. Fewtrell L, Kay D. Recreational Water and Infection: A Review of Recent Findings. Curr Environ Heal Reports. 2015;2: 85. pmid:25821715
- 2. Korajkic A, McMinn BR, Harwood VJ. Relationships between Microbial Indicators and Pathogens in Recreational Water Settings. Int J Environ Res Public Health. 2018;15. pmid:30551597
- 3. Heasley C, Sanchez J, Young I, Tustin J. Beach water monitoring practices and challenges in Ontario Public Health units. Environ Heal Rev. 2022;65: 17–24.
- 4. Tiwari A, Oliver DM, Bivins A, Sherchan SP, Pitkänen T. Bathing Water Quality Monitoring Practices in Europe and the United States. Int J Environ Res Public Heal 2021, Vol 18, Page 5513. 2021;18: 5513. pmid:34063910
- 5. Young I, Sanchez JJ, Tustin J. Recreational water illness in Canada: a changing risk landscape in the context of climate change. Can J Public Heal 2022. 2022; 1–4. pmid:36112242
- 6. Prüss A. Review of epidemiological studies on health effects from exposure to recreational water. Int J Epidemiol. 1998;27: 1–9. pmid:9563686
- 7. United States Environmental Protection Agency. Recreational Water Quality Criteria and Methods. US EPA. 2013. Available: https://www.epa.gov/wqc/recreational-water-quality-criteria-and-methods
- 8. European Commission. Directive 2006/7/EC of the European Parliament and of the Council of 15 February 2006 concerning the management of bathing water quality and repealing Directive 76/160/EEC. Off J Eur Union. 2006;L 064: 37–51. Available: https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=celex%3A32006L0007
- 9. Health Canada. Guidelines for Canadian Recreational Water Quality: Indicators of Fecal Contamination. 2021.
- 10. Health Canada. Guidelines for Canadian recreational water quality–third edition. 2012. Available: https://www.canada.ca/en/health-canada/services/publications/healthy-living/guidelines-canadian-recreational-water-quality-third-edition.html
- 11. Heasley C, Sanchez JJ, Tustin J, Young I. Systematic review of predictive models of microbial water quality at freshwater recreational beaches. PLoS One. 2021;16. pmid:34437625
- 12. Turgeon P. Monitoring Recreational Waters: How to Integrate Environmental Determinants. J Environ Prot (Irvine, Calif). 2012;3: 798–808.
- 13. Sanchez JJ, Young I, Heasley C, Kelly J, Habjan A, Waterhouse R, et al. Environmental factors associated with freshwater recreational water quality in Niagara Region, Ontario, Canada: A path analysis. Epidemiol Infect. 2021;149: e217.
- 14. Sanchez J, Tustin J, Heasley C, Patel M, Kelly J, Habjan A, et al. Region-specific associations between environmental factors and Escherichia coli in freshwater beaches in Toronto and Niagara region, Canada. Int J Environ Res Public Health. 2021;18. pmid:34886567
- 15. Kleinheinz GT, McDermott CM, Hughes S, Brown A. Effects of Rainfall on E. coli Concentrations at Door County, Wisconsin Beaches. Int J Microbiol. 2009;2009. pmid:20182543
- 16. Francy DS, Stelzer EA, Duris JW, Brady AMG, Harrison JH, Johnson HE, et al. Predictive Models for Escherichia coli Concentrations at Inland Lake Beaches and Relationship of Model Variables to Pathogen Detection. Appl Environ Microbiol. 2013;79: 1676. pmid:23291550
- 17. Williamson BD, Ralley W, Bourne A, Armstrong N, Fortin R. Principal factor affecting Escherichia Coli at Lake Winnipeg Beaches, Manitoba, Canada—Interim Report. Manitoba Water Steward. 2004 [cited 18 Aug 2022]. Available: http://www.togz.ca/E coli report lk wpg_beach_report_interim-040129.pdf
- 18. ESRI World Topo. QuickMapServices: discover, share and use geospatial data services. [cited 4 Sep 2023]. Available: https://qms.nextgis.com/geoservices/519/
- 19. Manitoba Water Stewardship Water Science and Management Branch. Manitoba Water Quality Standards, Objectives, and Guidelines. Manitoba Water Steward Rep 2011–01. 2011. Available: https://www.gov.mb.ca/waterstewardship/water_quality/quality/pdf/mb_water_quality_standard_final.pdf
- 20. Environment and Climate Change Canada. Daily Data Report for May 2022—Climate. [cited 9 May 2023]. Available: https://climate.weather.gc.ca/climate_data/daily_data_e.html?hlyRange=2006-09-22%7C2022-05-10&dlyRange=2006-09-19%7C2022-05-10&mlyRange=2007-05-01%7C2007-11-01&StationID=45007&Prov=MB&urlExtension=_e.html&searchType=stnName&optLimit=yearRange&StartYear=18
- 21. Government of Canada’s Water Office. Daily Water Level Graph for LAKE WINNIPEG AT VICTORIA BEACH (05SA003) [MB]—Water Level and Flow. [cited 9 May 2023]. Available: https://wateroffice.ec.gc.ca/report/historical_e.html?stn=05SA003
- 22. Oceans and Fisheries Canada. Wave data available on-line. [cited 9 May 2023]. Available: https://www.meds-sdmm.dfo-mpo.gc.ca/isdm-gdsi/waves-vagues/data-donnees/index-eng.asp
- 23. National Aeronautics and Space Administration of the United States. POWER | Data Access Viewer—Prediction Of Worldwide Energy Resource. [cited 9 May 2023]. Available: https://power.larc.nasa.gov/data-access-viewer/
- 24. Streiner DL. Finding Our Way: An Introduction to Path Analysis. 2005;50: 115–122. pmid:15807228
- 25. Digitale JC, Martin JN, Glymour MM. Tutorial on directed acyclic graphs. J Clin Epidemiol. 2022;142: 264–267. pmid:34371103
- 26. Zhang Z. Distinguishing between mediators and confounders is important for the causal inference in observational studies. AME Med J. 2019;4: 35–35.
- 27. Thompson CG, Kim RS, Aloe AM, Becker BJ. Extracting the Variance In flation Factor and Other Multicollinearity Diagnostics from Typical Regression Results. Basic Appl Soc Psych. 2017;39: 81–90.
- 28. Ene M, Leighton EA, Blue GL, Bell BA. Multilevel Models for Categorical Data Using SAS ® PROC GLIMMIX: The Basics. SAS Glob Forum. 2015; 1–12. Available: https://www.mendeley.com/catalogue/986a804e-f72b-3665-9ad5-56fd26f41b6e/?utm_source=desktop&utm_medium=1.19.8&utm_campaign=open_catalog&userDocumentId=%7B377f2b5e-dad4-47c9-b2a9-9ef8441a4057%7D
- 29. Bell BA, Ene M, Smiley W, Schoeneberger JA. A Multilevel Model Primer Using SAS PROC MIXED. SAS Glob Forum. 2013; 1–19.
- 30. Stapleton J. Warnings lifted for Gimli swim areas affected by raw sewage leak. In: CBC News [Internet]. 2018 [cited 26 Oct 2022] p. July 28. Available: https://www.cbc.ca/news/canada/manitoba/gimli-ecoli-water-warning-lifted-1.4765832
- 31. Media Bulletin—Manitoba. Province Advises of Sewage Spill at Spruce Sands Beach. In: News Releases, Province of Manitoba [Internet]. 2022 [cited 26 Oct 2022] p. July 20. Available: https://news.gov.mb.ca/news/?archive=&item=55581
- 32. Hong Y, Soulignac F, Roguet A, Li C, Lemaire BJ, Martins RS, et al. Impact of Escherichia coli from stormwater drainage on recreational water quality: an integrated monitoring and modelling of urban catchment, pipes and lake. Environ Sci Pollut Res. 2021;28: 2245–2259. pmid:32876821
- 33. Frey SK, Gottschall N, Wilkes G, Grégoire DS, Topp E, Pintar KDM, et al. Rainfall-Induced Runoff from Exposed Streambed Sediments: An Important Source of Water Pollution. J Environ Qual. 2015;44: 236–247. pmid:25602339
- 34. Parker JK, McIntyre D, Noble RT. Characterizing fecal contamination in stormwater runoff in coastal North Carolina, USA. Water Res. 2010;44: 4186–4194. pmid:20617564
- 35. Jang J, Hur HG, Sadowsky MJ, Byappanahalli MN, Yan T, Ishii S. Environmental Escherichia coli: ecology and public health implications-a review. J Appl Microbiol. 2017;123: 570–581. pmid:28383815
- 36. Weiskerger CJ, Brandão J, Ahmed W, Aslan A, Avolio L, Badgley BD, et al. Impacts of a changing earth on microbial dynamics and human health risks in the continuum between beach water and sand. Water Res. 2019;162: 456–470. pmid:31301475
- 37. Tiwari A, Kauppinen A, Pitkänen T. Decay of Enterococcus faecalis, vibrio cholerae and MS2 coliphage in a laboratory mesocosm under brackish beach conditions. Front Public Heal. 2019;7: 480435.
- 38. Whitman RL, Nevers MB, Korinek GC, Byappanahalli MN. Solar and temporal effects on Escherichia coli concentration at a Lake Michigan swimming beach. Appl Environ Microbiol. 2004;70: 4276–4285. pmid:15240311
- 39. Mori M, Hamamoto A, Nakano M, Akutagawa M, Takahashi A, Ikehara T, et al. Effects of ultraviolet LED on bacteria. IFMBE Proc. 2007;14: 1327–1330.
- 40. Park SK, Jo DM, Kang MG, Khan F, Hong SD, Kim CY, et al. Bactericidal effect of ultraviolet C light-emitting diodes: Optimization of efficacy toward foodborne pathogens in water. J Photochem Photobiol B Biol. 2021;222: 112277. pmid:34364078
- 41. Vermeulen N, Keeler WJ, Nandakumar K, Leung KT. The bactericidal effect of ultraviolet and visible light on Escherichia coli. Biotechnol Bioeng. 2008;99: 550–556. pmid:17680675
- 42. Varna K, Meghan J. The Effects of Ultraviolet Light on Escherichia coli. Emerg Investig. 2015.
- 43. Paule-Mercado MA, Ventura JS, Memon SA, Jahng D, Kang JH, Lee CH. Monitoring and predicting the fecal indicator bacteria concentrations from agricultural, mixed land use and urban stormwater runoff. Sci Total Environ. 2016;550: 1171–1181. pmid:26895037
- 44. Gonzalez RA, Conn KE, Crosswell JR, Noble RT. Application of empirical predictive modeling using conventional and alternative fecal indicator bacteria in eastern North Carolina waters. Water Res. 2012;46: 5871–5882. pmid:22981488
- 45. Gikas GD, Tsihrintzis VA. Effect of first-flush device, roofing material, and antecedent dry days on water quality of harvested rainwater. Environ Sci Pollut Res. 2017;24: 21997–22006. pmid:28785942