Chlorella vulgaris and Arthrospira platensis show distinct removal dynamics for multiple heavy metals in real water matrices

Heavy metal discharge in aquatic systems poses risks to public health and ecosystems. Bioremediation using live microalgae is a sustainable solution, but performance under realistic multi-metal conditions remains insufficiently explored. Chlorella vulgaris and Arthrospira platensis were exposed to a five-metal mixture (Pb 2+ , Zn 2+ , Cd 2+ , Ni 2+ , Cu 2+ ) in three real water matrices (river water, lake water, fish-processing wastewater) at elevated concentrations simulating accidental contamination. Kinetic modeling, Fourier transform infrared (FTIR) spectroscopy and interpretable machine learning (ML) were applied to characterize removal dynamics and associated variables. Chlorella vulgaris demonstrated more stable and progressive removal, while Arthrospira platensis achieved final Day-7 removals of 98.1% for Cd 2+ in fish-processing wastewater and 95.1% for Cu 2+ in lake water. Non-monotonic apparent uptake required use of the first observed uptake maximum for kinetic fitting. Across four input feature configurations, Gaussian process regression generally provided more consistent predictive performance than XGBoost, while reduced models retained predictive information after contemporaneous target-related variables were excluded or replaced. Shapley Additive exPlanations analysis indicated metal and species-specific contributions from temporal, physicochemical, biomass-related and concentration related variables. Overall, the combined analysis revealed distinct, matrix-dependent removal patterns and supports the potential of living microalgae for rapid multi-metal bioremediation under complex multi-metal contamination.