Explainable AI-based evaluation of factors affecting heavy metal removal by microalgae-based adsorbents

Abstract

The heavy metal (HM) contamination of water poses significant environmental and health risks and requires efficient remediation strategies. This study evaluated the adsorption performance of microalgae-based biochar adsorbents derived from Chlorella and Dpirulina (SPI) for the removal of cationic (Cd(II) and Pb(II)) and anionic (As(V) and Cr(VI)) HMs. Among the prepared types of biochar, SPI-derived SB-200 demonstrated superior adsorption of cationic HMs (> 90%), whereas Fe-modified SPI biochar (FeSB-400) significantly improved anionic HM removal (76%-97%) owing to enhanced electrostatic interactions. The adsorption efficiency was influenced by operational parameters, including the pH, temperature, and coexisting ions, and the removal ratio was maximized under optimal conditions. Scanning electron microscopy, energy-dispersive spectroscopy, and X-ray diffraction analyses confirmed the adsorption mechanisms and biochar surface modifications. Machine learning models (decision trees, random forests, and extreme gradient boosting) were developed to predict the adsorption behavior of the microalgae-based biochar adsorbents; the models achieved high accuracy (R-2 > 0.94). SHapley Additive exPlanations analysis showed that the HM charge, biomass type, Fe functionalization, and pyrolysis temperature were the key adsorption factors. Factor interaction analysis highlighted the influence of charge-Fe functionalization relationships and biomass-dependent adsorption trends. This study integrates experimental and data-driven approaches, thereby providing an optimized framework for biochar-based water treatment. In addition, it demonstrates the potential of explainable artificial intelligence (AI) in adsorption modeling for environmental applications.