Historical Analysis of Recreational Beach Enterococci Levels; Using Logistic Regression as an Advisory Tool

Enterococci levels are measured to assess water safety in recreational beaches through a state surveillance program. This surveillance informs the public of beach safety, yet the sampling methodology is limited to only making an advisory posting one sample at a time. This methodology poses a challenge for managers such as: 24 hour advisory waiting period, untested days and extreme variability of enterococci levels in the environment. Therefore, there is a need to integrate adaptive management methodologies that can assist managers to proactively assess beach water safety. This study explored the utility of a historical analysis and logistic regression modeling as a method and as an advisory tool. The analysis utilized 10 years of enterococci surveillance data (7,422 samples) from 15 sub-tropical beaches in Miami-Dade County, Florida. It was determined that Miami beaches have historical low enterococci exceedance counts (3% of total data), that there are some beaches that are more propense to higher exceedance counts than others and that the wet season overall did not readily appear to affect exceedances counts. The logistic regression model utilized an exceedance/ non-exceedance dichotomy and spatial, temporal and annual variables. The model indicated that the overall range of probability of having an exceedance for the sampled beaches under each variable was less than 10%. The ability to use this model and get probability results showed that logistic regression is an accurate statistical tool that provides the historical probabilities of an exceedance on a beach and can complement a random sampling methodology. Furthermore it’s a simple and inexpensive methodology that provides the ability to categorize and recognize patterns estimating the surveillance-managed sample sites probabilities that provides foresight as to where to focus resources in order to reduce risk and facilitating beach management. Through the use of a historical analysis and a logistic regression model, it is possible to address dynamic recreational beach environments with a large-scale view and in a historically comprehensive manner, instead of only making management choices sample by sample.