Influence of typical anions in brackish water on synergistic remediation of Vanadium(V) by coffee grounds and Ryegrass: Machine learning modeling and co-occurrence network analysis.

Brackish water is widely distributed and highly overlaps with areas contaminated by vanadium (V), presenting multiple challenges for V pollution remediation due to its high salinity and complex anionic environment. This study develops a novel coffee ground (CG)-enhanced ryegrass system for efficient V(V) remediation in brackish water. Ryegrass (52.92%, 4 d) outperformed other several common aquatic plants. Additionally, the study explored the effects of pH and typical coexisting anions on V(V) removal efficiency. The influence of different anion concentrations and types (Cl-, SO42-, CO32-) on plant remediation mechanisms was elucidated through microbial and plant physiological and biochemical indicators. Notably, CG could enhance V(V) removal efficiency and alleviate these adverse effects at specific concentrations. Artificial Neural Networks (ANN) + K-Nearest Neighbors (KNN) (Voting) model accurately predicted V(V) concentration (test R2 = 0.82, MSE = 0.19), with CG addition, pH, and anions as key features. The ANN + KNN (Voting) model was validated through additional experiments, and the model's prediction accuracy showed that the maximum residual was only 0.94 mg/L. Microbial co-occurrence network analysis revealed that the addition of CGpromoted microbial community interactions, with the average clustering coefficient increasing from 0.329 to 0.444 compared to the plant-only group. This also alleviated competition between certain genera, thus facilitating synergistic remediation. Additionally, three major V(V) reducing bacteria-unclassified Enterobacteriaceae, Lactococcus, and Pseudomonas were identified, with each exhibiting varying degrees of adaptability to the environment. This study offers a low-cost, sustainable strategy for V(V) remediation, leveraging waste-derived CG to enhance plant-microbe synergies in brackish water with typical anions. The findings provide new insights into environmental applications of phytoremediation, demonstrating the potential of ML modeling and microbial network analysis in optimizing remediation strategies.