The Water Authority of Western Australia, also known as WAWA, was a statutory authority of the state government that was responsible for the water supply, sewerage, and main drainage within Western Australia between 1985 and 1996.
While numerous machine learning (ML) models have been applied to wastewater treatment plant (WWTP) quality prediction tasks, significantly less attention has been paid to the selection and interpretation of performance evaluation metrics. Most studies rely on general-purpose regression metrics such as Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and the Coefficient of Determination (R2). Although widely used, these metrics differ considerably in terms of interpretability, sensitivity to data anomalies, and suitability for dynamic, noisy environments like WWTPs. This review, based on a systematic literature analysis of 27 performance evaluation metrics, critically examines their theoretical foundations, strengths, limitations, and applicability within the context of supervised ML-based WWTP modeling. In addition to statistical metrics, it also explores complementary graphical techniques, such as residual or failure prediction plots, that offer deeper insights into model behavior, insights that purely numerical indicators may overlook. A significant contribution of this paper is the development of a practical, decision-guiding flowchart to assist researchers in selecting appropriate evaluation metrics based on dataset characteristics, modeling objectives, and project constraints. Additionally, it summarizes a reference toolkit of graphical methods that have been used in the literature to assess model performance beyond numerical indicators. Together, these resources not only promote more informed and transparent metric selection in research but also provide wastewater practitioners with actionable tools for interpreting model outputs, comparing predictive approaches, and identifying the models most suitable for reliable process monitoring and operational decision-making.
Accurate prediction of raw wastewater characteristics can reduce laboratory costs, enable optimized chemical dosing, and improve the overall efficiency of wastewater treatment plant (WWTP) operations. This study evaluates machine learning (ML) models as a fast, sustainable, and cost-effective alternative for predicting Biochemical Oxygen Demand (BOD) and Ammonia Nitrogen (NH4+-N) in raw wastewater. Seven ML algorithms, including tree-based, neural, and regression approaches, were trained using standard wastewater indicators, including organic load, nutrient levels, and pH, with hyperparameters tuned via grid search cross-validation. Model performance was evaluated using correlation-based (R2) and error-based metrics (RMSE, MAE, MAPE) to ensure a comprehensive assessment. This study investigates ML-based wastewater prediction using real-world weekly sampling data, in contrast to the high-frequency datasets that dominate existing research. Additionally, data distribution shifts, preprocessing techniques, and splitting strategies were evaluated, with the Smoothness Index used to assess consistency and temporal stability. Despite the limitations of longer sampling intervals, MLP achieved the best overall performance for BOD prediction (RMSE = 29.8 mg/L, MAPE = 7.0 %), while SVR performed best for NH4+-N prediction (RMSE = 2.3 mg/L, MAPE = 3.3 %). These results demonstrate strong same-day predictive capabilities, enabling more efficient monitoring, conserving laboratory resources, and reducing chemical usage.
Naegleria fowleri has been detected in drinking water distribution systems (DWDS) in Australia, Pakistan and the United States and is the causative agent of the highly fatal disease primary amoebic meningoencephalitis. Previous small scale field studies have shown that Meiothermus may be a potential biomarker for N. fowleri. However, correlations between predictive biomarkers in small sample sizes often breakdown when applied to larger more representative datasets. This study represents one of the largest and most rigorous temporal investigations of Naegleria fowleri colonisation in an operational DWDS in the world and measured the association of Meiothermus and N. fowleri over a significantly larger space and time in the DWDS. A total of 232 samples were collected from five sites over three-years (2016-2018), which contained 29 positive N. fowleri samples. Two specific operational taxonomic units assigned to M. chliarophilus and M. hypogaeus, were significantly associated with N. fowleri presence. Furthermore, inoculation experiments demonstrated that Meiothermus was required to support N. fowleri growth in field-collected biofilms. This validates Meiothermus as prospective biological tool to aid in the identification and surveillance of N. fowleri colonisable sites.
Fluoride-bearing apatite minerals such as fluorapatite (FAP: Ca10(PO4)6F2) and related carbonate-rich fluorapatites (CFA: Ca10(PO4)5(CO3,F)F2), which occur ubiquitously as trace components of rocks and sediments, may act as sources for geogenic groundwater fluoride contamination. CFA dissolution often occurs in conjunction with declining dissolved calcium concentrations. Therefore, managed aquifer recharge (MAR) operations using deionised or low calcium source water are at risk of disturbing the naturally persisting geochemical equilibrium between CFA and the ambient groundwater and induce fluoride mobilisation. In this study, we employ reactive transport modelling to investigate how an engineered manipulation of the MAR source water composition might mitigate such groundwater fluoride contamination. Based on a previously developed and calibrated model for Australia's largest groundwater replenishment operation, we investigate the efficiency of (i) raising aqueous calcium concentration through the addition of CaCl2 or Ca(OH)2 amendment, (ii) raising aqueous sodium concentrations through the addition of NaCl or sea salt amendment and (iii) raising the pH. The modelling results illustrate in detail how the geochemical zonation around injection boreholes evolves over time and how this affects fluoride release and attenuation for the different amendment types. Treatments involving the addition of calcium and sodium in the source water are both found to be effective at reducing maximum groundwater fluoride concentrations during MAR, with calcium generally producing the greatest reduction in maximum fluoride concentrations. In contrast, increasing the injectate pH was found to be inefficient in reducing fluoride concentrations significantly due to the strong pH buffering effect of the aquifer sediments.
Groundwater management and allocation planning involves a rigorous assessment of the performance of operational decisions such as extraction/injection rates on community and environmental objectives. Maximizing performance through numerical optimization can be essential for high‐value resources and is often computationally infeasible due to long simulation model run times combined with nonconvex objectives and constraints. In order to mitigate these drawbacks, surrogate models can be used in place of complex models during the optimization process. There exist a number of machine learning techniques that can be used to develop a data‐driven surrogate model. However, the curse of dimensionality, common to groundwater management, limits the use of these techniques due to the necessity for large training data sets. Even though it is now possible to handle large data sets, the generation of these data sets themselves remains computationally prohibitive as they require numerous simulations to produce accurate surrogates. In this study, we integrate a dimensionality reduction method using truncated singular value decomposition to reduce the number of decision variables, thereby reducing the size of the training data set needed. Correspondingly, we demonstrate a simple technique for acquiring an approximate minimax Latin Hypercube design from within the subspace. We also implement a novel technique for adaptive resampling through particle swarm optimization in order to maintain accuracy of the surrogate model throughout the optimization process. The resulting accurate surrogate model for the Perth regional aquifer system of Western Australia runs in a matter of seconds. Adopting this approach can produce timely solutions, making formal optimization tractable for practitioners.
