We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Modelling the fate of chemicals of emerging concern in wastewater treatment: From mechanistic models to hybrid data driven approaches
Summary
Everyday chemicals from medicines, cleaning products, and plastics often slip through wastewater treatment plants and end up back in our environment, potentially in our water supply. This review looked at different computer modeling approaches—including newer AI-based methods—that scientists use to predict how well treatment plants remove these pollutants, finding that combining traditional science-based models with machine learning works best. The authors note we still need better tools to track particularly concerning substances like PFAS ("forever chemicals") and microplastics, which matters because understanding how well treatment plants filter these out helps protect the water we drink and use every day.
Chemicals of emerging concern (CECs) pose major challenges for wastewater treatment plants because they are difficult to biodegrade, persist in the environment, and occur at trace concentrations. This review evaluates modelling methodologies for simulating the behaviour and fate of CECs during biological wastewater treatment processes. A systematic review of peer-reviewed literature published between 2000 and 2025 was conducted to assess statistical, mechanistic, machine learning, and hybrid models, focusing on predictive capability, modelling trends, and research gaps. The findings indicate a transition from mechanistically based models, particularly the Activated Sludge Models (ASMs), towards data-driven and hybrid approaches capable of capturing the nonlinear complexities of wastewater treatment systems. Machine learning models demonstrated strong predictive capability and effectively identified operational parameters influencing treatment efficiency, including dissolved oxygen, pH, temperature, solids retention time, and chemical oxygen demand. However, these models remain limited by their reliance on large datasets, reduced transferability across treatment systems, and the interpretability challenges associated with "black box" predictions. Mechanistic models accurately represent biochemical processes but are less effective in simulating dynamic CEC interactions. Hybrid models integrating mechanistic and machine learning approaches emerged as the most promising methodology by combining predictive accuracy with improved process representation. Future research should focus on Per- and polyfluoroalkyl substances (PFAS) and microplastics modelling, uncertainty analysis, explainable artificial intelligence, and region-specific datasets, particularly in developing countries.