Meta-Analysis Addressing the Implications of Model Uncertainty in Understanding the Persistence of Indicators and Pathogens in Natural Surface Waters.

Abstract

This study evaluates the impact persistence model selection has on the prediction of persistence values of interest and the identification of influential water quality and environmental factors for microorganisms in natural surface waters. Five persistence models representing first-order decay and nonlinear decay profiles were fit to a comprehensive database of 629 data sets for fecal indicator bacteria (FIB), bacteriophages, bacteria, viruses, and protozoa mined from the literature. Initial periods of minimal decay and decay rates tapering off over time were often observed, and a two-parameter model, based on the logistic probability distribution, provided the best fit to the data most frequently. First-order decay kinetics provided the best fit to less than 20% of the analyzed data. Using the best fitting models in this analysis, T90 and T99 metrics were calculated for each data set and used as the dependent variable in a variety of exploratory factor analyses. Random forest methods identified temperature and predation as some of the most important water quality factors influencing persistence, and the protozoa target type differed the most from FIB. This analysis further confirmed the interactions between temperature and predation and suggests that pH and turbidity be more frequently documented in persistence studies to further elucidate their impact on target persistence. The findings from this analysis and the calculated persistence metrics can be used to better inform quantitative microbial risk assessments and may lead to improved predictions of human health risks and water management decisions.