Enhanced phytoremediation of vanadium using coffee grounds and fast-growing plants: Integrating machine learning for predictive modeling.

Abstract

Vanadium (V) contamination posed a significant environmental challenge, while phytoremediation offered a sustainable solution. Phytoremediation performance was often limited by the slow growth cycles of traditional plants. A novel approach to enhancing V phytoremediation by integrating coffee grounds with fast-growing plants such as barley grass, wheat grass, and ryegrass was investigated in this study. The innovative use of coffee grounds leveraged not only their nutrient-rich composition but also their ability to reduce oxidative stress in plants, thereby significantly boosting phytoremediation efficiency. Notably, ryegrass achieved 48.7% V⁵⁺ removal within 6 d with initial 20 mg/L V⁵⁺ (0.338 mg/L·d·g ryegrass). When combined with coffee grounds, V⁵⁺ removal by using wheat grass increased substantially, rising from 30.51% to 62.91%. Gradient Boosting and XGBoost models, trained with optimized parameters including a learning rate of 0.1, max depth of 3, and n_estimators of 300, were employed to predict and optimize V⁵⁺ concentrations in the phytoremediation process. These models were evaluated using mean squared error (MSE) and coefficient of determination (R²) metrics, achieving high predictive accuracy (R² = 0.95, MSE = 1.20). Feature importance analysis further identified the initial V⁵⁺ concentration and experimental duration as the most significant factors influencing the model's predictions. The addition of coffee grounds not only mitigated the stress of heavy metals on ryegrass, leading to significant reductions in CAT (87.2%), POD (98.8%), and SOD (39.2%) activities in ryegrass roots, but also significantly altered the microbial community abundance in the plant roots. This research provided an innovative enhancement to traditional phytoremediation methods, and established a new paradigm for using machine learning to predict and optimize V⁵⁺ remediation outcomes. The effectiveness of this technology in multi-metal polluted environments warrants further investigation in the future.