Constructed wetland-microbial fuel cell integrated with Bayesian calibration: Performance and mechanism of water quality monitoring.

Microbial fuel cell (MFC) biosensors suffer severe time‑varying nonlinear baseline drift caused by synergistic chelation between humic acid (HA) and copper ions (Cu) in natural surface water. Traditional static linear calibration fails to separately quantify three distinct inhibitory pathways, including physical shielding of electroactive biofilms by HA, intracellular cytotoxicity induced by free Cu, and additional signal attenuation derived from insoluble HA-Cu chelated precipitates. In this study, a dual‑functional constructed wetland‑microbial fuel cell (CW‑MFC) system was developed for real‑time water quality sensing and synchronous pollutant remediation. A two‑tier hierarchical Bayesian inversion framework incorporating three independent weighting coefficients (γHA, γCu and γHA-Cu) was established to mathematically decouple multi‑source signal drift under HA-Cu combined contamination. With real‑time Bayesian drift compensation, the concentration‑domain root‑mean‑square error (RMSE) was reduced to 8.77 ± 0.71 mg/L for HA and 0.78 ± 0.06 mg/L for Cu under mixed pollutant conditions. The integrated CW‑MFC system achieved over 80 % chemical oxygen demand (COD) removal, demonstrating its dual capacity for self‑powered real‑time water quality monitoring and decentralized water purification. The proposed Bayesian model is currently calibrated only for binary HA-Cu co‑contamination scenarios; future efforts will expand the interference coefficient library to cover more pollutant types and further improve the practical applicability of this biosensing strategy.