Decoding impact of antibiotics stress to nitrogen removal in constructed wetlands: Overestimated?

Antibiotics and nitrogen presented in aquatic environments posing significant ecological risks, which could be addressed through constructed wetlands (CWs). However, the complex removal mechanisms of antibiotics and nitrogen influenced by various factors, had remained difficult to elucidate with traditional univariate experiments. To overcome these limitations, machine learning models were employed to decoding a database comprising 4218 data points covering diverse input features such as wetland characteristics, influent water quality, antibiotics categories, and microbial community composition. The code of core structure (CCS), hydrophobicity reference values (XLogP3), and Wiener Index (WI) were used to represent different antibiotics categories. The results demonstrated that antibiotics removal efficiency was primarily governed by the molecular structure and concentration of antibiotics, with WI and antibiotics concentration accounting for over 65 % of the removal variance. Influent water quality and constructed wetlands volume significantly influenced nitrogen removal (52.4 % and 10.6 %, respectively), with system size and dissolved oxygen dynamics offering potential areas for optimization. Additionally, Actinobacteria played a crucial role in both nitrogen and antibiotics removal, underscoring microbial community composition as a key mechanism. Interestingly, antibiotics had little effect on TN removal efficiency (5.1 %). These insights would provide a foundation for optimizing the design and operation of constructed wetlands under antibiotics stress, offering a novel framework for improving wastewater treatment performance.