Close

Precision Farming in Aquaculture: Use of a non-invasive, AI-powered real-time automated behavioural monitoring approach to predict gill health and improve welfare in Atlantic salmon (Salmo salar) aquaculture farms

Meredith Burke, Dragana Nikolic, Pieter Fabry, Hemang Rishi, Trevor C. Telfer, Sonia Rey-Planellas

Posted on: 11 September 2024 , updated on: 12 September 2024

Preprint posted on 6 April 2024

Keeping an eye on the prize: Using Artificial Intelligence (AI) behavioural monitoring to improve Atlantic salmon farming welfare

Selected by Jasmine Talevi

Background:

When we are stressed, we sometimes bite our nails or bounce our legs. In Atlantic Salmon, stress can present itself as changes in feeding or swimming behavior. Shoaling is one example which refers to changes in swimming behavior leading multiple fish to group together. Shoaling, in addition to other changes in feeding and swimming behavior, can occur in response to stress from predators, changes in the environment, or health status.

In a recent study, Burke and collaborators have leveraged our current understanding of swimming and feeding behavioral changes to improve Atlantic Salmon farming practices. By installing video cameras in fish net pens (Fig. 1 of the preprint), the researchers could use Artificial Intelligence (AI) machine learning technology to recognize real-time patterns in fish behavior.

Detecting patterns in fish behavior indicative of stress and compromised health could provide farmers with early warning signs of stress, specifically gill disease in the case of this study. Gill disease is a major concern in salmon farming worldwide. Two types of gill disease, Amoebic Gill Disease (AGD) and Proliferative Gill Disease (PGD), are especially common and both of these cause breathing difficulties, hinder growth, and cause mortality. Unfortunately, with increasing sea temperatures, occurrences of AGD and PGD may also increase, highlighting the need for the early detection of fish stress to help prevent disease rather than treat symptoms.

Methods:

As part of this preprinted study, the researchers selected two fish farms (Farm A and Farm B) in Scotland and deployed video camera(s) in 9-10 cages per farm. Several behavioral characteristics were monitored including depth of fish activity (Depth of Max Activity) and proximity of the fish to one another (Activity). Specific Feeding Rate (SFR) were also calculated as ((feed given/fish biomass) X 100). Weekly gill health examinations were carried out where signs of AGD or PGD were scored on a scale of 0 (healthy tissue) – 5 (necrotic tissue) and mortality was monitored throughout the experiment.

 

Figure 1. The map shows Scotland highlighting the location of the Atlantic salmon aquaculture farms on the Western Isles (a). Below are the camera orientations for the study cages at each farm (b). Reproduced with permission from the preprint authors.

 

Highlighted Key Findings:

1) Proliferative Gill Disease (PGD) was the main driver of increased activity levels

The researchers observed an increase in group fish activity at both farms, which coincided with poorer gill health status, specifically PDG (Fig. 2 of the preprint). On average, there was a 15% increase in activity at Farm A (which experienced poorer gill health), while there was a 7.7% increase at Farm B.

2) Increased fish activity indicates shoaling behavior

Increases in fish activity was concentrated towards the centre of the cage indicating fish shoaling in this area (Fig. 2 of the preprint). This behaviour is often a stress response, commonly triggered by the presence of predatory stimuli or environmental stressors

3) Mortality increased and Specific Feeding Rate (SFR) decreased after the onset of poor gill health

At both farms mortality increased with the onset of poor gill health (Fig. 3 of the preprint). SFR also increased with the onset of poor gill health though more prominently at Farm A where gill health was poorer than Farm B.

4) Group-level assessments are more informative than individual fish assessments

When behavioral changes began to emerge and gill health scores became worse, individual-level assessments of fish did not significantly change. Though individual assessments for signs of illness (changes in eyes, fins, skin) are important, by the time the outward signs of poor health emerge the fish may already be compromised. Thus, group assessments, such as behavioral change, are important for predicting impacts to fish.

 

Figure 2. Comparative figure for Farm A (left) an Farm B (Right) with: depth of maximum activity where colours indicate the hourly-averaged amount of activity present at that depth (a,e), weekly gill health scores (0 = healthy gills, 5 = necrotic tissue), where red indicates Proliferative Gill Disease (PGD) and blue indicates Amoebic Gill Disease (AGD; b, f), daily mortality as a percent loss within the cage (c, g), and hourly temperature at 8 m depth (d, h). Reproduced with permission from the preprint authors.

 

Figure 3. The health data from each cage in farm A (left) and farm B (right) including: The weekly PGD scores (A. E), the daily activity of the fish (B, F), the daily specific feeding rate (SFR; C, G), and the daily mortality (D, H). The colours indicate pre- (white) and post- (grey) increase in activity (July 7 [Farm A]; August 18 [Farm B]). Different letters denote significant differences within and among cages per farm. Reproduced with permission from the preprint authors.

 

Why I Highlight This Preprint:

This preprint presents an innovative solution to a globally relevant issue. The technology tested offers a practical, easily deployable device that farms can readily implement. Additionally, this technique has the potential to enhance salmon welfare, contributing to improved food security.

Questions for the Authors:

  1. In comparison to individual fish assessments, approximately how much extra time could farmers gain by implementing this group-level behavioural assessment and what does this mean in terms of fish welfare?
  2. Did the net pen with one strategically placed camera capture enough fish activity to serve as an effective early warning sign? Will farmers need to deploy more than one camera per net pen?

Tags: aquaculture, salmo salar

doi: https://doi.org/10.1242/prelights.38340

Read preprint (No Ratings Yet)

Author's response

The author team shared

1. In comparison to individual fish assessments, approximately how much extra time could farmers gain by implementing this group-level behavioral assessment and what does this mean in terms of fish welfare?
It’s hard to say exactly time-wise, but if we can use behaviour as a metric to measure fish health and welfare rather than the current observational methods, then farmers may reduce the amount of handling they do to the fish. Right now the farmers conduct these surveys about once a week in the summer (maybe once every 2 weeks in the winter), taking 10 fish out of each cage, so it can take 1-2 days to complete depending on the size of the farm. This is a cause of stress to the fish, and time-consuming for the farmers. Moreover, if the farmers happen to select 10 healthy fish or 10 fish in poor health, then you can see there is some room for human error and maybe missing how the group is doing as a whole. Using behaviour as a non-invasive form of monitoring can be hugely beneficial in this way.
2. Did the net pen with one strategically placed camera capture enough fish activity to serve as an effective early warning sign? Will farmers need to deploy more than one camera per net pen?
The ultimate goal is to be able to use this with one camera because most farms already have this camera installed for feeding. We don’t want the farmers to have to install anything extra to be able to use this technology. From this study we see in Figure 4 that the activity increased in all cages and aside from the one cage that had 4-5 cameras, the rest of the cages just had 1 camera. As this camera was positioned in the centre of the cage and the fish tended to congregate there when stressed, we believe that this 1 camera model can work well. Our future studies will go into validating this at other sites and properly push forward the idea of using this as an early warning indicator.

Have your say

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Sign up to customise the site to your preferences and to receive alerts

Register here

Also in the animal behavior and cognition category:

Pharyngeal neuronal mechanisms governing sour taste perception in Drosophila melanogaster

Bhanu Shrestha, Jiun Sang, Suman Rimal, et al.

Selected by 23 September 2024

Matthew Davies

Cell Biology

Impaired 26S proteasome causes learning and memory deficiency and induces neuroinflammation mediated by NF-κB in mice

Christa C. Huber, Eduardo Callegari, Maria Paez, et al.

Selected by 22 August 2024

Gustavo Stelzer, Marcus Oliveira

Biochemistry

Blue appendages and temperature acclimation increase survival during acute heat stress in the upside-down jellyfish, Cassiopea xamachana

Megan E. Maloney, Katherine M. Buckley, Marie E. Strader

Selected by 30 April 2024

Maitri Manjunath

Animal Behavior and Cognition

Also in the ecology category:

Precision Farming in Aquaculture: Use of a non-invasive, AI-powered real-time automated behavioural monitoring approach to predict gill health and improve welfare in Atlantic salmon (Salmo salar) aquaculture farms

Meredith Burke, Dragana Nikolic, Pieter Fabry, et al.

Selected by 11 September 2024

Jasmine Talevi

Animal Behavior and Cognition

Gestational exposure to high heat-humidity conditions impairs mouse embryonic development

Avinchal Manhas, Amritesh Sarkar, Srimonta Gayen

Selected by 08 July 2024

Girish Kale, preLights peer support

Developmental Biology

How the liver contributes to stomach warming in the endothermic white shark Carcharodon carcharias

David C. Bernvi, Geremy Cliff

Selected by 22 April 2024

Sarah Young-Veenstra

Physiology

Also in the physiology category:

Precision Farming in Aquaculture: Use of a non-invasive, AI-powered real-time automated behavioural monitoring approach to predict gill health and improve welfare in Atlantic salmon (Salmo salar) aquaculture farms

Meredith Burke, Dragana Nikolic, Pieter Fabry, et al.

Selected by 11 September 2024

Jasmine Talevi

Animal Behavior and Cognition

Gestational exposure to high heat-humidity conditions impairs mouse embryonic development

Avinchal Manhas, Amritesh Sarkar, Srimonta Gayen

Selected by 08 July 2024

Girish Kale, preLights peer support

Developmental Biology

Modular control of time and space during vertebrate axis segmentation

Ali Seleit, Ian Brettell, Tomas Fitzgerald, et al.

AND

Natural genetic variation quantitatively regulates heart rate and dimension

Jakob Gierten, Bettina Welz, Tomas Fitzgerald, et al.

Selected by 24 June 2024

Girish Kale, Jennifer Ann Black

Developmental Biology
Close