Deep Learning Approaches for Automated Detection of Animal Welfare Indicators under Climate Change Scenarios

Abstract

Climate change has emerged as one of the most significant threats to global livestock production, influencing animal health, welfare, productivity, and environmental sustainability. Rising ambient temperatures, prolonged heat waves, fluctuating humidity, deteriorating air quality, and extreme weather events adversely affect livestock physiology and behaviour, resulting in heat stress, reduced feed intake, impaired reproduction, increased disease susceptibility, and economic losses. Conventional animal welfare assessment methods primarily rely on manual observation, periodic veterinary examinations, and subjective behavioural evaluations, which are labour-intensive, time-consuming, and often Deep Learning Approaches for Automated Detection of Animal Welfare Indicators under Climate Change Scenarios incapable of detecting early physiological or behavioural changes associated with climate-induced stress. Consequently, there is a growing demand for intelligent, automated, and real-time welfare monitoring systems capable of supporting climate-resilient livestock management through continuous surveillance and data-driven decision-making. Recent advances in Deep Learning (DL), Computer Vision (CV), Internet of Things (IoT), thermal imaging, and wearable biosensors have revolutionized precision livestock farming by enabling automated recognition of animal behaviour, physiological responses, and environmental conditions. Convolutional Neural Networks (CNNs), You Only Look Once (YOLO) object detection models, Vision Transformers (ViTs), Long Short-Term Memory (LSTM) networks, and hybrid deep learning architectures have demonstrated remarkable capability in detecting animal posture, locomotion, feeding behaviour, lameness, facial expressions, body condition, and heat stress. However, existing approaches generally focus on isolated welfare indicators or individual sensing technologies, while limited attention has been given to developing integrated deep learning frameworks that simultaneously combine multimodal sensing, climate information, behavioural analytics, physiological monitoring, explainable artificial intelligence, and intelligent decision support for comprehensive welfare assessment under changing climate conditions.