Certain strains of Escherichia coli have been reported to bloom in the environment, resulting in high concentrations in waters in the absence of apparent fecal input or human pathogens, and in turn, undermining its reliability as an indicator of recent fecal contamination. Given the capacity of environmental strains of E. coli to replicate in the environment, the objective of this research work was to investigate whether any of the routinely measured parameters could predict the onset of an E. coli bloom in drinking water reservoirs. Information from historical catchment, weather, and water quality data were extracted for a number of Australian reservoirs that experienced E. coli blooms. Data were preprocessed and analyzed with time series analysis, linear and nonlinear regression, and self-organizing maps. Findings suggest that warm water, dry catchments, algal blooms, and nutrient availability were important factors in increasing the propensity for a bloom. Nutrient availability can be affected by many extrinsic factors that are often not well characterized, such as bushfires and back burning, decomposition of aquatic species, and dust storms. Based on data analysis outputs, a data-driven Bayesian Network model was developed, which, considering the paucity of data for some key input parameters, should only be used to trigger more intensive monitoring programs whenever the predicted risk of a bloom exceeds predetermined key thresholds. Such new data could be fed into the model to continuously improve its accuracy, and to eventually predict and proactively manage future blooms.