Oral Presentation Australian Society for Fish Biology & Australian Freshwater Sciences Society Conference 2026

FROM BEHAVIOUR TO EARLY WARNING: AI-BASED MONITORING OF FISH WELFARE AND DISEASE RISK IN TROPICAL POND AQUACULTURE (144558)

Sunil Kadri 1 , Ly Q Dang 2 , Stefan Sawynok 3 , Doan G Thieu 4 , Tran M Phu 5
  1. Research Institute for Northern Agriculture, Charles Darwin University, Brinkin, NT, Australia
  2. Dept of Sociology, Can Tho University, Can Tho, Vietnam
  3. Infofish Australia Pty. Ltd., Murarrie, QLD, Australia
  4. College of ICT, Can Tho University, Can Tho, Vietnam
  5. College of Aquaculture and Fisheries, Can Tho University, Can Tho, Vietnam

Fish behaviour provides an important, yet underutilised, source of information about health and welfare. Changes in swimming activity, spatial distribution and behavioural patterns can occur before clinical signs or mortality, providing an opportunity for earlier intervention. However, continuous behavioural monitoring remains challenging in pond aquaculture, particularly in turbid environments where visual observation and camera-based systems are often ineffective.

We are developing and testing an AI-enabled behavioural and environmental monitoring platform for striped catfish (Pangasianodon hypophthalmus) aquaculture in the Mekong Delta, Vietnam. The project integrates continuous hydroacoustic monitoring of fish behaviour with water-quality sensing and machine learning to identify behavioural and environmental changes associated with emerging health risks. The approach builds on previous research demonstrating the potential of hydroacoustics and artificial intelligence to assess welfare in whole salmon populations in cages, presented at ASFB 2024 in Newcastle.

The system is currently being deployed on a commercial farm, generating continuous behavioural and environmental datasets across production cycles. Initial data are being used to characterise normal behavioural patterns and investigate relationships between fish behaviour, environmental conditions and health events. We will present the initial predictive model and preliminary results, demonstrating how changes in whole-population behaviour may provide early indicators of deteriorating welfare and increased disease risk.

By combining behavioural monitoring with environmental data, the project aims to shift disease management from reactive treatment towards earlier, evidence-based intervention. The approach has potential to provide a scalable method for continuous welfare and health monitoring in tropical pond aquaculture, with future applications across species and production systems.

GEDSI principles underpin co-designed training and accessible technology, supporting participation by remote and marginalised farming communities.

This project was made possible through the Aus4Innovation program. Launched in 2018, Aus4Innovation is a ten-year A$33.5 million flagship program aimed at strengthening Vietnam’s innovation system to support inclusive and sustainable socio-economic development.