Prosopis juliflora biochar for hexavalent chromium adsorption: Mechanism and interpretable tree-based ensemble machine learning modelling for predictive removal performances

Widespread heavy metal contamination in water and soil causes severe environmental and life risks. Recently, biochar has demonstrated excellent adsorption performances in removal of various organic and inorganic pollutants. In this study, biochar derived from Prosopis juliflora was prepared and utilized for Cr(VI) removal. Batch adsorption experiments were evaluated for optimizing the chromium removal conditions. Isotherm study revealed Sips isotherm model demonstrated the best fit confirming heterogeneous monolayer adsorption with maximum adsorption capacity of 100.95 mg g −1 . Moreover, adsorption kinetics revealed the best described model fits for pseudo first order kinetics indicating physisorption mechanism. Thermodynamic study confirmed the spontaneous and exothermic nature of Cr(VI) adsorption. Machine learning predictive modelling enables the removal performances of pollutants. Particularly, tree-based ensemble models preferred due to its robustness and superiority in predictive performances. Therefore, the current study integrates six machine learning models to predict the Cr(VI) removal efficiency onto PJ-BC adsorbent. Among ML models, CatBoost demonstrated the best predictive performance with highest R 2 of 0.988 and least RMSE of 2.648. This study also incorporated interpretable ML framework including SHapley Additive exPlanations and Partial Dependence Plots for elucidating the feature importance of input variables. Through experimental and machine learning modelling with Prosopis juliflora biochar for chromium removal, this study contributes towards the development of sustainable and data-driven wastewater treatment strategies for Cr(VI) contaminated effluents.