MnOx-based atrazine removal from water: non-linear adsorption modeling and machine learning prediction

Abstract

Atrazine is a widely used triazine herbicide. In this research, manganese oxide (MnOx) was investigated as an effective adsorbent for removal of atrazine from aqueous solutions. The effects of key operational parameters were systematically evaluated. The adsorption performance was strongly influenced by solution pH, with maximum removal efficiency (98.70%) and adsorption capacity (19.7 mg/g) obtained at pH 10 specifically under pH variation experiments. Increasing MnOx dosage enhanced removal efficiency but reduced adsorption capacity. While removal efficiency decreased with increasing initial atrazine concentration, adsorption capacity increased. Temperature had a positive effect, suggesting that adsorption is endothermic in nature. Equilibrium data were analyzed using several isotherms, with the Sips model providing the best fit. Kinetic analyses indicated that the adsorption behavior was best represented by pseudo-second-order model. Overall, the results demonstrate that MnOx is a promising and efficient adsorbent for atrazine removal and may offer a viable alternative for the treatment of pesticide-contaminated waters. To further understand and predict the adsorption behavior, machine learning (ML) models were developed. Among the tested models, Ridge Regression exhibited the highest predictive performance. The integration of experimental data with ML provides a robust framework for predicting adsorption performance and improving process design in water treatment applications.