A Hybrid Response Surface Methodology and Machine Learning Framework for Quantifying Effects of Physicochemical Parameters on PFAS Distribution

Predicting PFAS adsorption across diverse adsorbents and environmental matrices remains challenging because adsorbent physicochemical properties, PFAS molecular descriptors, and operational conditions simultaneously influence adsorption. This study develops and evaluates a unified hybrid modeling framework that integrates Response Surface Model (RSM) with machine-learning algorithms to quantify how six key variables, surface area, Log Kow, pHpzc, pKa, log dose, and log-initial concentration, affect PFAS distribution coefficients (Log Kd). A data set of more than 1000 adsorption observations spanning 15 PFAS compounds, multiple adsorbent types, and a broad operational range was compiled and preprocessed using mode imputation and log transformation. Model performance was evaluated using an 80/20 split and Leave-One-PFAS-Out (LOPO) validation. Gradient Boosting performed best in the 80/20 scenario (R2 = 0.93; RMSE = 0.25), whereas Random Forest achieved the highest performance under LOPO validation (R2 = 0.30; RMSE = 0.78), highlighting the challenge of compound-wise generalization. Feature-importance analyses consistently identify log-dose and log-initial concentration as key predictors, followed by surface area and pHpzc. Partial-dependence analysis and RSM surfaces revealed nonlinear relationships and interaction effects among adsorption drivers. Overall, the framework provides a transparent approach for predicting PFAS adsorption while disentangling the coupled roles of adsorbent properties, PFAS chemistry, and operational conditions.