Search

Pond irrigation water comprises a major pathway of pathogenic bacteria to fresh produce. Current regulatory methods have been shown to be ineffective in assessing this risk when variability of bacterial concentrations is large. This paper proposes using mechanistic modeling of bacterial transport as a way to identify improved strategies for mitigating this risk pathway. If the mechanistic model is successfully tested against observed data, global sensitivity analysis (GSA) can identify important mechanisms to inform alternative, preventive bacterial control practices. Model development favored parsimony and prediction of peak bacterial concentration events. Data from two highly variable surface water irrigation ponds showed that the model performance was similar or superior to that of existing pathogen transport models, with a Nash-Sutcliffe efficiency of 0.48 and 0.18 for the two ponds. GSA quantified bacterial sourcing and hydrology as the most important processes driving pond bacterial contamination events. Model analysis has two main implications for improved regulatory methods: that peak concentration events are associated with runoff-producing rainfall events and that intercepting bacterial runoff transport may be the best option to prevent bacterial contamination of surface water irrigation ponds and thus fresh produce. This research suggests the need for temporal management strategies. IMPORTANCE Preventive management of agricultural waters requires understanding of the drivers of bacterial contamination events. We propose mechanistic modeling as a way forward to understand and predict such events and have developed and tested a parsimonious model for rain-driven surface runoff contributing to generic Escherichia coli contamination of irrigation ponds in Central Florida. While the model was able to predict the timing of peak events reasonably well, the highly variable magnitude of the peaks was less well predicted. This indicates the need to collect more data on the fecal contamination inputs of these ponds and the use of mechanistic modeling and global sensitivity analysis to identify the most important data needs. [ABSTRACT FROM AUTHOR]

The U.S. Food and Drug Administration (FDA) has defined standards for the microbial quality of agricultural surface water used for irrigation. According to the FDA produce safety rule (PSR), a microbial water quality profile requires analysis of a minimum of 20 samples for Escherichia coli over 2 to 4 years. The geometric mean (GM) level of E. coli should not exceed 126 CFU/100 mL, and the statistical threshold value (STV) should not exceed 410 CFU/100 mL. The water quality profile should be updated by analysis of a minimum of five samples per year. We used an extensive set of data on levels of E. coli and other fecal indicator organisms, the presence or absence of Salmonella, and physicochemical parameters in six agricultural irrigation ponds in West Central Florida to evaluate the empirical and theoretical basis of this PSR. We found highly variable log-transformed E. coli levels, with standard deviations exceeding those assumed in the PSR by up to threefold. Lognormal distributions provided an acceptable fit to the data in most cases but may underestimate extreme levels. Replacing censored data with the detection limit of the microbial tests underestimated the true variability, leading to biased estimates of GM and STV. Maximum likelihood estimation using truncated lognormal distributions is recommended. Twenty samples are not sufficient to characterize the bacteriological quality of irrigation ponds, and a rolling data set of five samples per year used to update GM and STV values results in highly uncertain results and delays in detecting a shift in water quality. In these ponds, E. coli was an adequate predictor of the presence of Salmonella in 150-mL samples, and turbidity was a second significant variable. The variability in levels of E. coli in agricultural water was higher than that anticipated when the PSR was finalized, and more detailed information based on mechanistic modeling is necessary to develop targeted risk management strategies. [ABSTRACT FROM AUTHOR]