Modeling system for predicting enterococci levels at Holly Beach

Abstract

This paper presents a new modeling system for nowcasting and forecasting enterococci levels in coastal recreation waters at any time during the day. The modeling system consists of (1) an artificial neural network (ANN) model for predicting the enterococci level at sunrise time, (2) a clear-sky solar radiation and turbidity correction to the ANN model, (3) remote sensing algorithms for turbidity, and (4) nowcasting/forecasting data. The first three components are also unique features of the new modeling system. While the component (1) is useful to beach monitoring programs requiring enterococci levels in early morning, the component (2) in combination with the component (1) makes it possible to predict the bacterial level in beach waters at any time during the day if the data from the components (3) and (4) are available. Therefore, predictions from the component (2) are of primary interest to beachgoers. The modeling system was developed using three years of swimming season data and validated using additional four years of independent data. Testing results showed that (1) the sunrise-time model correctly reproduced 82.63% of the advisories issued in seven years with a false positive rate of 2.65% and a false negative rate of 14.72%, and (2) the new modeling system was capable of predicting the temporal variability in enterococci levels in beach waters, ranging from hourly changes to daily cycles. The results demonstrate the efficacy of the new modeling system in predicting enterococci levels in coastal beach waters. Applications of the modeling system will improve the management of recreational beaches and protection of public health.