Saurabh Singh

Flowchart showing optimization of 176 constructed wetlands using kinetic modelling, Taguchi design, ANOVA and SHAP, yielding operating conditions that enhance BOD, TN and TP removal kinetics. Constructed wetlands (CWs) are sustainable wastewater treatment systems, but their large land requirement limits use in urban and land-constrained areas. This study develops a data-driven kinetic–statistical framework to optimize first-order areal removal-rate coefficients (k-values) for BOD, TN, and TP in full-scale CWs. A secondary dataset comprising 176 operational CWs, including 92 horizontal-flow CWs and 84 vertical-flow CWs, was analyzed across nine design, operational, climatic, and influent-quality parameters. The k-values were estimated using the P-k-C* model standardized to 20°C and optimized using a Taguchi L32 orthogonal array, multivariate regression, sensitivity analysis, interaction analysis, and SHAP-based CatBoost interpretation. Results showed that detention time and organic loading rate mainly controlled BOD removal kinetics, while temperature and media depth strongly influenced TN and TP removal. The optimized operating envelope, including approximately 2-days detention time, 330 g/m3/day organic loading rate, 2.4 m media depth, and 24–34°C temperature, enhanced kinetic coefficients to kBOD = 0.31 m/day, kTN = 0.08 m/day, and kTP = 0.27 m/day. These results indicate that CWs can be shifted toward a high-k, low-footprint operating regime by balancing hydraulic residence time, organic loading, and reactive media capacity. The framework supports compact, hybrid, and climate-resilient CW designs by shifting systems toward a high-k, low-footprint regime, contributing to SDG-6.

Per- and polyfluoroalkyl substances (PFAS) show strongly compound- and condition-dependent adsorption, creating uncertainty in activated-carbon treatment design, particularly for short-chain PFAS with weak and variable affinity. Here, an interpretable machine learning framework is developed to predict the Freundlich adsorption coefficient (KF) as a screening-level indicator of adsorption affinity within a harmonized PFAS–adsorbent–condition data set. A literature database of 256 extracted records was curated into 238 harmonized PFAS–adsorbent–condition entries (IF = 0–85.60). Data were split into 70% training, 15% validation, and 15% testing, with 3-fold cross-validation (k = 3) used for hyperparameter optimization. Seven models (SVR, RF, GB, XGBoost, ANN, and KNN) were evaluated using R2, RMSE, and MAE. Performance increased from conventional to advanced models: XGBoost was the best traditional baseline (R2test = 0.77). Among the evaluated models, KNN showed the highest predictive performance (KNN: R2test = 0.89). SHAP interpretation identified influential descriptor domains consistent with known adsorption behavior, with pH, adsorbent identity, and temperature as key operational controls and molecular descriptors governing size/dispersion effects. Chain-length heatmaps of log10(KF) revealed distinct adsorption regimes and data set gaps, supporting predictive infilling for missing PFAS–adsorbent combinations. These combined approaches help investigate and control PFAS in the environment by supporting source tracking while accurately predicting KF and key drivers for faster screening and improved treatment design.