The value of stochastic modelling for the characterisation of maturation pond performance and health risk assessment of effluent reuse options

Treated wastewater from Waste Stabilisation Ponds (WSPs) is a potential resource, especially in regional and remote areas facing water shortages. As the ultimate goal of reusing treated wastewater is ensuring that it is fit-for-purpose, the comprehensive assessment, management and communication of potential health risks are a necessity. Although not mandated by policy in Australia, Quantitative Microbial Risk Assessment (QMRA) is recommended in the Australian Guidelines for Water Recycling (AGWR) as a methodology for estimating the potential levels of health risk associated with exposures to microbial pathogens. With the rapid uptake of QMRA modelling in the water sector in Australia, it is necessary to explore the sensitivity of QMRA models to underlying assumptions. This stochastic modelling study investigated the impacts of uncertainty in key input parameters underpinning QMRA and evaluated the sensitivity of a QMRA model to a range of underlying assumptions. This was conducted using the @Risk software program within the Palisade Decision Suite and operational monitoring data from a pond-based, Waste Water Treatment Plant (WWTP) in regional New South Wales, Australia. This study was conducted in two phases. The first addressed two research questions regarding: 1: The impact of assumed sampling regimes (weekly, fortnightly or monthly), and 2: The impact of seasonal variability in characterising pond performance as measured by Log Reduction Values (LRVs). The fortnightly and monthly datasets were compiled from the weekly monitoring data collected from both the inlet and outlet of the 3 maturation-pond system within the overall treatment system. The weekly data were also stratified by season for investigation of potential seasonal patterns. The distributions of these LRVs were generated through Monte Carlo simulations and impacts of temporal variation investigated statistically. The second phase focused on estimating potential health risks associated with the hypothetical scenario of irrigating lettuce with pond effluent using QMRA. The results from the first phase were used as input data in the QMRA stochastic modelling. The QMRA incorporated the four steps: hazard identification, dose-response, exposure assessment and risk characterisation. Monte Carlo simulation was used to generate a probabilistic distribution of health risk estimates, and the built-in sensitivity analysis functions in @Risk were used to rank the input parameters by their effect on the estimated risk levels. The results from the first phase revealed no significant difference between how weekly, fortnightly or monthly datasets characterised the microbial water quality from the maturation pond system in terms of LRVs. This suggests that the frequency of monitoring at the WWTP could be reduced without compromising the information value of the dataset. This would, in turn, reduce expenses. Seasonality, however, does appear to have a significant impact on pond performance as measured by LRVs. A pair-wise comparison of the weekly data by season revealed statistically significant differences between all seasons, except for winter and autumn. Summer provided the best performance and spring provided the worst. The results from the health risk assessment (second phase) suggested the microbial quality of the pond effluent would not be suitable for the proposed reuse scenario of irrigating lettuce eaten raw; however alternative, less risky reuse scenarios would be within the scope of the state guidelines. This phase of the research also opened up a broader range of issues such as how guidelines are interpreted and how different modelling approaches (e.g., deterministic versus stochastic) can yield disparate results. The stochastic methods provide more conservative estimates and therefore engender greater assurance in both regulators and the community that safety standards are being met. Overall, this research highlights that stochastic modelling is a powerful tool for risk communication and risk management for the water industry, the community and regulators.