Modeling Fecal Indicator Bacteria Concentrations in Natural Surface Waters: A Review

Microbiological quality of waters must be assessed to ensure that no health risk due to pathogenic microorganisms is associated with its use. As it is impossible to measure the abundance of all possible pathogens, it is general practice to quantify the abundance only of one or a few fecal indicator bacteria (FIB), organisms which are selected to be indicative of fecal pollution and can therefore serve as general indicators of microbiological water quality. Yet, even by focusing on this limited number of indicator organisms, it is still unfeasible to experimentally monitor their levels at the high spatiotemporal resolution often needed in real applications. Therefore, direct FIB measurements are increasingly combined with the use of models. The aim of this review is to present and evaluate the wide variety of models used so far in the scientific literature to simulate and predict FIB concentrations in natural surface waters. First, the distinction is made between regression-based and mechanistic models. While the first are particularly useful in operational contexts and indeed can produce reliable short-term predictions, they do not allow an in-depth understanding of the processes. Because the sources and processes are not modeled explicitly, they cannot be used to test the effect of changes in these internal and external forcings, e. g., to evaluate the impact of different management options. These questions can only be addressed by the use of mechanistic models, which are often based on rather complex computational methods. These models explicitly consider the effect on FIB concentrations due to horizontal transport, external sources, decay, and/or sediment-related processes. It is not possible to make statements about best practices, given the broad range of study domains and questions. Instead, we attempted to compile the modeling approaches published so far in a comprehensive and transparent way, hoping that the resulting overview would help to better understand current models and more efficiently set up future ones.