Deep Learning Detection of Oxygen Stress Behaviour in Freshwater Prawns

Abstract

Introduction: Freshwater systems act as habitats for various forms of aquatic life, where freshwater decapods have significant impacts on benthic functioning and nutrient cycling. Changes in water environments through reduction in dissolved oxygen create physiological and behavioral stress on the animals. Conventional methodologies of observation are difficult as they are cumbersome and unable to detect early stages of behaviors due to rapid environmental degradation. There is therefore a need to use automated intelligence in order to observe stress in animals in relation to falling oxygen concentrations. Materials & Methods: The current study proposes an ecological behavioral assessment method using deep learning to analyze behavioral patterns of oxygen stress on freshwater prawns. Apart from water quality measurements such as dissolved oxygen, temperature, pH, turbidity, and stocking conditions, important behavioral factors are measured, including swimming speed, surface activity, feeding activity, aggregation, and lethargy. Results: The analysis indicates that oxygen depletion causes substantial effects on prawn movement and distribution. At ideal oxygen concentrations, prawns continue with benthic feeding and even distribution. Hypoxia causes active escape to the surface, whereas severe oxygen depletion leads to movement cessation and increased death rate. The proposed prediction model allows us to identify the state of the organisms under various conditions. Conclusion: Deep learning techniques applied to behavioral testing can be a useful tool for monitoring the aquatic environment. Automated detection of stress reactions of prawns is useful for timely detection of hypoxic states.