Date of Award
Director of Thesis
Dr. Edsel Peña
Dr. John Grego
This project addresses the need for predictive modeling tools to forecast expected concentrations of fecal bacteria in recreational waters in the Charleston, SC area. Data was provided by Charleston Waterkeeper, a water quality monitoring organization that has been measuring Enterococcus faecalis concentrations at 15 recreational sites since 2013. The data contain a non-negligible number of censored and missing observations, so three distinct imputation methods were developed and compared in terms of their effect on final predictive model characteristics. The best performing method relied on drawing samples from a truncated normal distribution to replace censored values, and using a partial model built from all non-missing observations to impute missing values. Finally, a predictive model of Enterococcus in terms of precipitation in the past 72 hours, tidal stage, and sample site was developed. Results from this project may be used for forecasting Enterococcus concentrations in practice, as well as for informing the imputation phases of future studies.
Allen, Carter Alexander, "A Comparison of Imputation Algorithms for Modeling Water Quality" (2017). Senior Theses. 128.