Abstract
Aquatic recreation is a popular pastime, but recreators often incur the risk of acquiring illnesses through exposure to waterborne pathogens. Several standards advise stakeholders on a waterbody's recreational suitability using indicator bacteria concentrations. However, quantifying indicator organisms often takes several hours or days, resulting in a lack of real-time information. Microbe concentrations can also fluctuate dramatically over short timeframes, making it difficult to systematically assess and communicate risk using discrete samples. This study investigated turbidity and streamflow as real-time predictors of Escherichia coli (E. coli) in two major Iowa waterways (the Des Moines and Raccoon Rivers). A large historical dataset (n > 4000 at both sites) was leveraged to construct logistic models that predicted the probability of E. coli exceeding two recreational water quality standards (235 and 2880 organisms/100 mL). High-resolution turbidity measurements were collected to implement these models and comprehensively estimate long-term exceedance levels of the standards. Models were created at each site that used turbidity as the sole predictor-streamflow proved unnecessary. While these models were viable, their performance was relatively weak (pseudo R(2) between 0.19 and 0.38, sensitivities between 6.8 and 68%), suggesting that turbidity and streamflow alone are likely insufficient for producing real-time predictions of risk at accuracy levels acceptable to stakeholders. Still, there may be utility in deploying similar models to describe a waterway's propensity to exceed E. coli thresholds over the course of several months or years and to prioritize improvement efforts.