Elucidating microplastic adsorption mechanisms in biomass composite materials through interpretable machine learning

As emerging contaminants, microplastics (MPs) pose serious threats to ecosystems and human health, urgently necessitating the development of efficient and sustainable removal strategies, while traditional adsorption techniques face efficiency and complexity challenges. This study employs an interpretable machine learning framework to construct a multidimensional dataset containing 223 structured data points, systematically investigating the adsorption behaviour of biomass composite materials (BCMs) toward microplastics. Adsorption experiments using cellulose composite aerogels were designed under various environmental conditions to validate the model's extrapolation capability. Multiple machine learning algorithms were applied to analyse the comprehensive dataset encompassing microplastic properties, adsorbent characteristics, and environmental factors. Tree based ensemble models demonstrated superior predictive stability and interpretability, revealing that initial concentration and surface potential are the primary features influencing adsorption behavior. Validation experiments confirmed that model predictions accurately reflect the diffusion-controlled kinetics of aerogels. This research elucidates adsorption mechanisms under multifactor coupling effects, providing a data-driven paradigm for efficient biomass adsorbent design and ML applications in environmental remediation.