Asian-Pacific Aquaculture 2024

July 2 - 5, 2024

Surabaya, Indonesia

WELFARE SCORING FOR WHOLE FISH POPULATIONS BASED ON BEHAVIOUR, PRODUCTION AND ENVIRONMENTAL VARIABLES USING ARTIFICIAL INTELLIGENCE

Sunil Kadri*1,2 & Johannes Kvam1

 

1CageEye AS   2Charles Darwin University
Storgata 11       7 Ellengowan Dr

0155 OSLO, Norway          Casuarina NT 0810, Australia

 *sunil.kadri@cdu.edu.au

 



Following on  from our presentation at WAS 2023 (Kadri & Kvam 2023) we will describe advances in the WelfareShield fish welfare monitoring system, showcas ing a newly developed platform and examples from its deployment on fish farms.  This  welfare platform a ggregates data from hydroacoustics , environmental sensors  and  the  farm’s feeding systems .

Welfare indicators are constructed from each of these sources and combined into an overall welfare score , to give farmers an early warning system for changes in welfare status at a cage level (see Fig 1a) .

Using hydroacoustic measurements of the whole population we apply advanced pattern recognition and artificial intelligence algorithms to extract behavioural markers ( see Fig 1b):

  • Activity: Vertical movement
  •  Stability:  Lack of acute s tress.
  • Rhythm: Distinct  day and nighttime activity.
  •  Feeding response

 These behavioural markers are statistically calibrated over a large dataset to establish expected levels. T he platform also incorporates environmental parameters such as temperature and oxygen, as well as f eeding system data  which is used to detect anomalies in appetite levels.

 Kadri, S & Kvam, J. (2023)  Monitoring whole population welfare status in cages using hydroacoustics . World Aquaculture 2023 “Supporting Strength in Aquaculture”, May 29-June 1 Darwin, Australia. p. 146 https://wasblobstorage.blob.core.windows.net/meeting-abstracts/WA2023AbstractBook.pdf