Real-time intelligence framework for microplastic monitoring and coastal resilience

The growing threat of microplastic contamination of the marine biosphere requires that monitoring technologies surpass the limitations of conventional approaches. This paper presents the development and validation of an integrated, machine learning-based framework for real-time detection and management of microplastics in coastal environments, whose primary contributions are real-time IoT-edge-cloud integration, multi-sensor fusion, and 72 h predictive modeling. We deployed a multi-sensor platform integrating hyperspectral imaging (400-2,500 nm) with IoT sensor networks at five monitoring locations over 18 months. The analytical pipeline employs a stacked ensemble of CNN, SVM, and random forest algorithms with soft-vote fusion. The framework achieved 96.3% accuracy in classifying microplastic polymers and successfully detected particles as small as 75 μm in diameter. The system demonstrated a strong correlation (r = 0.92, p < 0.001) between predicted and observed pollution levels, enabling 72 h forecasting of pollution events with 84% probability. The framework processes samples 340% faster than traditional laboratory analysis while reducing operational costs by 67%. Through continuous real-time evaluation, this scalable solution will provide a quick instrument in a global network of coastal monitoring tools to create evidence-based policy and focused intervention approaches.