A coupled Lagrangian-AI hierarchical and heterogeneous model for predicting bacteria contamination in farmed mussels

Abstract

The quality of coastal waters, particularly aquaculture zones, is crucial to sustainable development and human health. Traditional monitoring methods based on scheduled in-situ sampling are often too slow, costly, and limited in spatial and temporal coverage to meet the needs of large-scale aquaculture management. To overcome these constraints, we introduce the Artificial Intelligence-based Water QUAlity Plus Plus model (AIQUAM++), an AI-based modeling framework designed to predict E. coli contamination directly within farmed mussels. We evaluated and compared a suite of recent and high-performing machine learning architectures, such as K-Nearest Neighbors (KNN), and several Transformer-based architectures (Transformer, Informer, Reformer, TimesNet), to address the complex temporal dependencies within this time series classification (TSC) task. AIQUAM++ was trained with historical microbiological measures of E. coli levels in the mussels provided by the local authorities involved in food safety monitoring. The system architecture, featuring an inference engine completely written in C++ for high performance, leverages hierarchical parallelism to ensure scalability and computational efficiency, incorporating Message Passing Interface (MPI) for inter-process communication on multi-core architectures, OpenMP for multithreaded processing, and CUDA-based acceleration for GPU-optimized computations. This design enables high-throughput inference that is suitable for operational deployment in aquaculture monitoring networks. A test case application of AIQUAM++ was conducted in the Gulf of Naples (Campania, Italy). Empirical results demonstrated that the proposed system achieves classification accuracies that exceed 90 %, supporting its efficacy as a real-time data-driven decision support tool for aquaculture water quality management, minimizing health risks and contributing to sustainable marine resource governance.