Optimization of a Novel Engineered Ecosystem Integrating Carbon, Nitrogen, Phosphorus, and Sulfur Biotransformation for Saline Wastewater Treatment Using an Interpretable Machine Learning Approach

Machine learning predicts selectivity of green synthesized iron nanoparticles toward typical contaminants: critical factors in synthesis conditions, material properties, and reaction process

Green synthesized iron nanoparticles (FeNPs) have gained popularity in contaminant removal due to their low cost and environmentally friendly properties. However, a gap remains in understanding how synthesis conditions (SC), material properties (MP), and reaction processes (RP) affect their removal capacities on typical contaminants. This study utilizes advanced machine learning methods to explore complex dependencies in contaminant removal, achieving high predictive accuracies with R2 rankings of XGBoost (0.9867) > RF (0.9749) > LightGBM (0.8545), and detailed SHAP analyses that elucidate the specific impacts of features. The model revealed that RP significantly influenced FeNPs' removal capacity. Both linear and SHAP analyses demonstrated that SC indirectly affected removal efficiency by influencing MP, thereby weakening their impact on FeNPs' removal capabilities due to their strong linear correlation. For all three contaminants (antibiotics, dyes and heavy metals), the removal capacity of FeNPs was primarily influenced by the C/Fe ratio and the type of plant present in the SC, as well as the pore volume of the MP. Antibiotics removal depends on antibiotic type and FeNPs' Fe content. The interaction time between Fe ions and plant extracts during SC and the specific surface area (SSA) of MP significantly influenced dyes removal, while the pore diameter in MP and the pH in RP were vital for heavy metals removal. MP impacts antibiotics removal more than SC, but SC's indirect effects are more significant for dyes and heavy metals. SHAP analysis clarified the importance and independent roles of specific features in the predictive modeling of removal efficiencies.

Long-term water quality simulation and driving factors identification within the watershed scale using machine learning

Understanding long-term trends and analyzing their driving factors are essential to effectively enhance water quality in watersheds. In China, although the overall quality of surface water continues to improve, significant issues remain in certain regions. The Liao River Basin, a critical industrial hub and key agricultural grain base in northeast China, continues to face unstable water quality conditions, despite over 20 years of management efforts. This study compared several data-driven models (random forest (RF), support vector machine regression (SVR), K-nearest neighbors (KNN), stacking, long short-term memory (LSTM), convolutional-long short-term memory (CNN-LSTM)), to accurately fill the water quality data gaps (i.e., total nitrogen (TN), ammonia nitrogen (NH3-N), total phosphorus (TP), chemical oxygen demand (CODCr), permanganate index (CODMn), electroconductibility (E)) from 1980 to 2022 in Liao River Basin. In addition, the SHapley Additive exPlanations (SHAP) model was employed to quantitatively assess the driving factors of water quality. The results showed that the RF model exhibited robust predictive capabilities. TN showed a steady increase of approximately 20 % from 1980 to 2022, while the other parameters were effectively controlled. Anthropogenic activities, especially in agriculture and urban areas, were found to significantly contribute to water quality deterioration. Additionally, climatic factors such as extreme rainfall, annual average precipitation, and extreme temperatures-along with geographical factors like soil properties and slope, were found to play crucial roles in influencing water quality.

Prediction of airborne bacterial concentrations and identification of critical factors in contaminated waste facilities: Insights into interpretable machine learning models

The efficient prediction of airborne bacterial concentrations is crucial for better understanding and management of environmental sanitation risks in waste facilities. Traditional linear models have proven inadequate in capturing the complex relationships governing the formation of airborne microorganisms. This study developed four machine learning (ML) models to estimate airborne bacterial concentrations in waste facilities regarding the combined dataset as input features. The results revealed that integrating environmental factors, gaseous pollutants, and microbial datasets as input features yielded an improved testing R2 of 0.7369, with a random forest (RF) model identified as the best-performing algorithm. The bacterial populations on the surfaces and handles of waste containers were identified as the most influential parameters in the RF model. The optimal ranges of temperature (32-36 °C) and relative humidity (62 %-80 %), the optimal concentrations of ammonia (< 0.15 mg/m3) and particulate matter 2.5 (0.01-0.07 mg/m3), and the effective disinfection measures of slightly acidic electrolyzed water were recommended for controlling airborne pollution in waste facilities. Overall, the research demonstrates that ML methods have the potential in the prediction of airborne bacterial concentrations in waste facilities. By identifying critical factors with the interpretability analysis, this study offers valuable insights for targeted airborne microorganisms' risk management strategies.

Leveraging big data to elucidate the impact of heavy metal nanoparticles on anammox processes in wastewater treatment

Anammox is a highly efficient nitrogen removal process, yet the effects of metal/metal-oxide nanoparticles (M/MONPs) on these systems remain underexplored. This study investigates the impact of various M/MONPs on the nitrogen removal rate (NRR). Pearson correlation analysis and statistical evaluation indicates that silver and copper oxide nanoparticles exhibit the highest inhibitory effect, with an inhibition rate of 83.4 % and 73.7 %, respectively. Furthermore, Machine learning models, particularly extreme gradient boost (XGBoost), demonstrate superior performance, with R2 values exceeding 0.91. SHapley Additive exPlanations (SHAP) feature importance analysis highlighted nanoparticles concentration, influent ammonia nitrogen concentration as the most influential factors. Additionally, Partial Dependence Plots (PDP) analysis of key features provided further clarity on the optimal ranges for these critical variables. The present study provides a novel predictive methodology and optimization strategies for enhancing the NRR of anammox system under M/MONPs stress, informed by comprehensive big data analysis.