Speaker
Description
Anthropogenic underwater noise is becoming an increasing concern for marine ecosystems, especially in coastal and semi-enclosed areas exposed to intense commercial, industrial, and recreational vessel traffic. Within the DeuteroNoise project, we present an AI-driven framework to detect, characterize, and model vessel-generated underwater noise across five European marine basins: the North Adriatic Sea, the Lagoon of Venice, the Barcelona coast, the North Sea, and the Black Sea. Our approach combines passive acoustic monitoring, in situ measurements, and simulation-oriented analyses to identify representative soundscapes, describe their spatial and temporal variability, and support the controlled reproduction of realistic acoustic environments in laboratory conditions. Particular emphasis is placed on advanced signal processing techniques for denoising, feature extraction, spectral and temporal analysis, event detection, and acoustic pattern characterization from complex underwater recordings. These descriptors are then incorporated into artificial intelligence and machine learning pipelines for vessel-noise identification, acoustic scene classification, and predictive modeling in diverse marine environments. The proposed framework aims to improve the interpretation of large volumes of raw acoustic data, reduce uncertainty in soundscape assessment, and enable the development of computational models that can be transferred across basins with different environmental and traffic conditions. In addition, these results provide a solid basis for controlled experimental studies on marine invertebrate deuterostomes exposed to anthropogenic noise. By bringing together monitoring, signal processing, and AI-based predictive tools, this work contributes to a more comprehensive assessment of underwater noise pollution and supports future mitigation strategies, environmental management, and evidence-based decision-making for healthier and more sustainable marine ecosystems.