Artificial Intelligence-Driven Wildlife Conservation Monitoring Systems for Biodiversity Protection in Rapidly Changing Ecosystems

Abstract

The shift in the ecosystem due to climate change, habitat destruction, urbanization and poaching of wildlife has intensified the rate of biodiversity losses and consequently a need for intelligent and scalable wildlife conservation monitoring system is now a reality. The work in this study is aimed at proposing a novel Artificial Intelligence (AI) based Wildlife Conservation Monitoring System (WCM) to monitor biodiversity in real time and to support decision making in the field of wildlife conservation, using computer vision, deep learning, remote sensing, Internet of Things (IoT) sensor networks, and Explainable Artificial Intelligence (XAI). The main goal is to improve species detection, enable continuous monitoring of habitats, forecast ecological hazards and support conservation strategy decision making based on evidence through automated analytics. It is proposed to use the UAV imagery, satellite data, camera-trap images, acoustic recordings, and observations of the environment by sensors, followed by data preprocessing, multimodal feature fusion, detection of species using the YOLOv11 network, classification of the habitat using the Swin Transformer network, prediction of threats using the strong learning ability of XGBoost, and explainability by the SHAP network to ensure transparent decision-making. The experimental evaluation using publicly available biodiversity data yielded 3.2-6.1% higher accuracy on key evaluation metrics than state-of-the-art baseline methods, with species detection accuracy of 95.84%, habitat classification accuracy of 94.72%, threat prediction accuracy of 92.91%, precision of 94.63%, recall of 93.87%, F1-score of 94.25%, and ROC-AUC of 96.48%. The uniqueness of the proposed framework is its single-platform intelligent conservation that integrates multimodal AI, XAI, and real-time ecological monitoring into a single platform. The proposed system offers an interpretable, scalable, and efficient approach to biodiversity protection that allows for proactive management of wildlife and informed conservation measures in contexts of evolving ecosystems.