High-density recirculating aquaculture systems (RAS) present a persistent challenge: clinically relevant stress and disease signals often remain hidden until pathology is already visible. While recent work has demonstrated the potential of individual sensing modalities - computer vision for biomass estimation, lesion and behaviour recognition, deep learning for water quality time-series prediction, collection of biomarker and passive acoustic monitoring for stress-correlated changes - these streams are studied in isolation. True heterogeneous fusion that jointly reasons across water chemistry, acoustics, visual and behavioural analytics, stress biomarker and operational context within a single inference architecture remains largely unexplored. This talk introduces our work and outcomes related to SAS-C programme on a multimodal early warning framework designed to close that gap. Rather than relying on conventional threshold-based alarms or single-modality classifiers, the approach applies cross-modal representation learning, that includes strategies adapted from biomedical and data fusion and cross-domain early warning and foresight systems. The goal is to surface weak preclinical deviations that unimodal pipelines routinely miss. We discuss the concept and architectural rationale, challenges of aligning modalities with heterogeneous sampling rates and noise profiles, and practical considerations for deployment in production RAS environments. The goal is actionable: the framework targets earlier detection and identification of stress in high-density fish populations within the preclinical window – aiming to reduce stock loss, optimize farming processes and enabling a measurable shift from reactive crisis management toward predictive, data-driven aquaculture operations.
This research is supported by the National Research Foundation, Singapore, under its Campus for Research Excellence and Technological Enterprise (CREATE) programme, and Singapore Food Agency, under its Singapore Food Story R&D 2.0 Programme ‘Seed’ Grant Call (Urban Solutions & Sustainability (“USS”) NRF-SFSRNDCREATE-0002).