PREPARED BY: Environmental Health

DATE: 2026-02-05

BACKGROUND/PURPOSE

The Windsor-Essex County Health Unit (WECHU) conducts beach water sampling in accordance with the Recreational Water Protocol, 2019 and following the Operational Approaches for Recreational Water Guideline, 2018. Seasonal beach monitoring includes surveillance of water and weather conditions, and potential pollutants, as well as weekly water sampling to test for Escherichia coli (E. coli), bacteria that can cause severe stomach cramps, diarrhea, fever, and vomiting. To determine the water quality, the WECHU staff collect water samples weekly at eight local beaches.

As a follow up to the February 6^th, 2025, Board of Health Meeting report, the WECHU implemented a predictive model to enhance the 2025 routine beach monitoring and notification program. This enhancement helps to address limitations related to the delay in receiving water quality results and the day-to-day water quality changes. The predictive model provides community access to daily estimations of what the water quality results may be. This supplementation to weekly beach monitoring supported residents on making an informed decision on whether to use the participating public beaches.

DISCUSSION

The predictive modeling program was launched in collaboration with local municipalities that agreed to participate. Five local public beaches  participated in year one—Sandpoint Beach, Lakeshore Lakeview Park West, Point Pelee North West, Colchester, and Holiday Beach (Sandpoint beach was unable to participate since the City of Windsor closed the beach due to safety concerns). The WECHU formalized these partnerships through Memorandum of Understandings and provided training and required equipment (turbidity monitors and thermometers) to ensure consistent data collection. On weekdays (excluding holidays) staff from participating beaches obtained and submitted daily turbidity and water temperature readings through an online survey. These readings were inputted into the model, which generated a risk level (low, moderate, or high). The predicted results were published on the WECHU’s website and was updated on the beach hotline. Regardless of the model’s predictive results, if the beach was closed based on the weekly routine sampling, the predictive model’s output was marked as “Not Applicable” for that period, and the beach would remain closed.

The WECHU conducted a comprehensive evaluation of the predictive modeling program to assess its accuracy, effectiveness, and impact on public health decision-making through stakeholder and beach user surveys, website analytics, and statistical validation of the model. 

The stakeholders survey achieved a 75% response rate, and the respondents expressed satisfaction with the training, instructional materials, and resources provided by the WECHU, noting that the turbidity monitors were user-friendly and the online submission process was efficient. The respondents indicated the model was valuable for delivering timely, same-day water quality information and strengthening public health communication. However, some operational challenges were identified by the respondents, including difficultly in collecting water samples due to beach conditions, limited staffing capacity, and occasional missed submissions. 

The Beach User Evaluation Survey, conducted at two beaches with 18 participants, confirmed that local beaches are valued as affordable, family‑friendly recreational spaces. The awareness of the WECHU’s beach water monitoring program was mixed, and none of the respondents were familiar with the predictive modeling component. Among those aware of routine beach water testing, all reported checking results online or through the hotline before visiting the beach and adjusted their plans accordingly. Website analytics confirmed strong community interest in the program. During the 2025 season, the Beach Water Testing webpage received over 32,000 visits and more than 313,000 interactions. Residents actively engaged with interactive maps, weekly test results, predictive forecasts, and educational content. This level of engagement highlights the program’s reach and its role in empowering residents to make informed recreational water use decisions.

The validity of the predictive model in forecasting daily E. coli risk levels was assessed by comparing the estimated E. coli concentrations and predicted risk categories (high, moderate, low) with actual water sample results (beach closed, open with warning, or open with no warning). A total of 55 assessments were conducted. The comparison showed strong overall alignment, with 93% of the predicted values demonstrating moderate or optimal validity. Optimal validity was observed in 38% of assessments, where the predicted risk matched the actual beach status. Another 55% demonstrated moderate validity, reflecting slight over or underestimation of results. Only 7% of assessments showed low validity, where predicted and actual conditions did not align.

Overall, the predictive modeling program has proven to be a valuable enhancement to our beach monitoring program, offering timely and reliable estimates of E. coli levels that support informed public decision-making. The next steps will focus on strengthening stakeholder engagement, enhancing public communication, and refining the predictive model. Key actions include expanding participation to all routinely monitored public beaches by re‑engaging current partners and conducting targeted outreach to partners not yet involved. A comprehensive communication strategy that uses the website, beach hotline, social media, and media partnerships to promote the predictive modeling program will be developed. At the same time, ongoing model refinement will aim to improve accuracy by exploring enhancements to existing predictors and potential new variables. Together, these efforts will support broader adoption and utilization of the predictive modeling program in our region.