Rainfall and Turbidity as Simple Alternative Predictors of Beach Water Quality in Chicago

Monitoring beach water quality is an essential component in protecting recreator health and preventing gastrointestinal illness at beaches. Historically, laboratory methods used for monitoring bacteria required at least 18-24 hours before the public could be notified of water quality conditions. This gap period of 18-24 hours led to inaccuracies of current water quality conditions at beaches, as these methods assessed conditions of the day prior. These methods are still used by monitoring programs across the United States and, until recently, Chicago. Starting in 2015, the Chicago Park District switched to new DNA-based methods that produce reportable results within approximately four hours. Despite this, DNA-testing is prohibitively expensive and a gap period of four hours still remains between sample collection and notification. This study of Chicago beach water quality was conducted using turbidity and rainfall as simple alternatives for predicting beach action value exceedance. This study used 4,624 sampling days the from 2015-2018 Chicago beach seasons to develop beach-specific decision trees using turbidity and precipitation nodes to produce near real time probabilities of beach action value exceedances. The goal of this study was to develop decision trees that predict beach status more accurately than using an always-open approach where 90% of sampling days are correctly non-exceeding. The final low-turbidity decision tree design has approximately half the error rate (5.04% and 9.31%) compared to the always-open approach (9.68% and 14.72%) at predicting no exceedances during dry and wet conditions, respectfully. During high-turbidity conditions, a conservative tree for predicting an exceedance (≥33% accuracy) application could be applied to 7 out of 19 beaches during wet conditions only.