Minimum treatment requirements are set in response to established or anticipated levels of enteric pathogens in the source water of drinking water treatment plants (DWTPs). For surface water, contamination can be determined directly by monitoring reference pathogens or indirectly by measuring fecal indicators such as Escherichia coli (E. coli). In the latter case, a quantitative interpretation of E. coli for estimating reference pathogen concentrations could be used to define treatment requirements. This study presents the statistical analysis of paired E. coli and reference protozoa (Cryptosporidium, Giardia) data collected monthly for two years in source water from 27 DWTPs supplied by rivers in Canada. E. coli/Cryptosporidium and E. coli/Giardia ratios in source water were modeled as the ratio of two correlated lognormal variables. To evaluate the potential of E. coli for defining protozoa treatment requirements, risk-based critical mean protozoa concentrations in source water were determined with a reverse quantitative microbial risk assessment (QMRA) model. Model assumptions were selected to be consistent with the World Health Organization (WHO) Guidelines for drinking-water quality. The sensitivity of mean E. coli concentration trigger levels to identify these critical concentrations in source water was then evaluated. Results showed no proportionalities between the log of mean E. coli concentrations and the log of mean protozoa concentrations. E. coli/protozoa ratios at DWTPs supplied by small rivers in agricultural and forested areas were typically 1.0 to 2.0-log lower than at DWTPs supplied by large rivers in urban areas. The seasonal variations analysis revealed that these differences were related to low mean E. coli concentrations during winter in small rivers. To achieve the WHO target of 10⁻⁶ disability-adjusted life year (DALY) per person per year, a minimum reduction of 4.0-log of Cryptosporidium would be required for 20 DWTPs, and a minimum reduction of 4.0-log of Giardia would be needed for all DWTPs. A mean E. coli trigger level of 50 CFU 100 mL⁻¹ would be a sensitive threshold to identify critical mean concentrations for Cryptosporidium but not for Giardia. Treatment requirements higher than 3.0-log would be needed at DWTPs with mean E. coli concentrations as low as 30 CFU 100 mL⁻¹ for Cryptosporidium and 3 CFU 100 mL⁻¹ for Giardia. Therefore, an E. coli trigger level would have limited value for defining health-based treatment requirements for protozoa at DWTPs supplied by small rivers in rural areas.

Temporal variations in concentrations of pathogenic microorganisms in surface waters are well known to be influenced by hydrometeorological events. Reasonable methods for accounting for microbial peaks in the quantification of drinking water treatment requirements need to be addressed. Here, we applied a novel method for data collection and model validation to explicitly account for weather events (rainfall, snowmelt) when concentrations of pathogens are estimated in source water. Online in situ β-d-glucuronidase activity measurements were used to trigger sequential grab sampling of source water to quantify Cryptosporidium and Giardia concentrations during rainfall and snowmelt events at an urban and an agricultural drinking water treatment plant in Quebec, Canada. We then evaluate if mixed Poisson distributions fitted to monthly sampling data ((Formula presented.) = 30 samples) could accurately predict daily mean concentrations during these events. We found that using the gamma distribution underestimated high Cryptosporidium and Giardia concentrations measured with routine or event-based monitoring. However, the log-normal distribution accurately predicted these high concentrations. The selection of a log-normal distribution in preference to a gamma distribution increased the annual mean concentration by less than 0.1-log but increased the upper bound of the 95% credibility interval on the annual mean by about 0.5-log. Therefore, considering parametric uncertainty in an exposure assessment is essential to account for microbial peaks in risk assessment.

The identification of appropriately conservative statistical distributions is needed to predict microbial peak events in drinking water sources explicitly. In this study, Poisson and mixed Poisson distributions with different upper tail behaviors were used for modeling source water Cryptosporidium and Giardia data from 30 drinking water treatment plants. Small differences (<0.5-log) were found between the “best” estimates of the mean Cryptosporidium and Giardia concentrations with the Poisson–gamma and Poisson–log-normal models. However, the upper bound of the 95% credibility interval on the mean Cryptosporidium concentrations of the Poisson–log-normal model was considerably higher (>0.5-log) than that of the Poisson–gamma model at four sites. The improper choice of a model may, therefore, mislead the assessment of treatment requirements and health risks associated with the water supply. Discrimination between models using the marginal deviance information criterion (mDIC) was unachievable because differences in upper tail behaviors were not well characterized with available data sets ((Formula presented.)). Therefore, the gamma and the log-normal distributions fit the data equally well but may predict different risk estimates when they are used as an input distribution in an exposure assessment. The collection of event-based monitoring data and the modeling of larger routine monitoring data sets are recommended to identify appropriately conservative distributions to predict microbial peak events.