An Energy‐Efficient Space‐Air‐Ground Integrated Network (
SAGIN
) Solution for Forest Threat Detection via Edge
AI
and Link‐Aware Satellite Backhaul
Whai‐En Chen, Shih‐Che Lin, Gwanggil Jeon, Hsin‐Hung Cho ABSTRACT
Forest monitoring in remote regions faces severe connectivity and energy constraints. This study presents an energy‐efficient satellite IoT system for acoustic threat detection, serving as a foundational Space‐Ground segment within the Space‐Air‐Ground Integrated Network (SAGIN) architecture. To overcome bandwidth bottlenecks, we propose a hierarchical edge computing architecture that performs local processing of acoustic data where an optimized TinyML model processes audio locally, achieving 92% detection accuracy with a minimal power draw of 18 mA and an ultra‐low latency of 0.3 s. Furthermore, we introduce a renewal‐based GEO satellite link model to capture temporal state transitions, establishing a theoretical foundation for link‐aware transmission strategies. Based on this model, our analytical projections demonstrate that the link‐aware strategy can reduce packet loss by 43% and decrease energy consumption by 45% compared to conventional blind transmissions. Powered exclusively by implemented solar energy harvesting, the system's robustness and sustainability were successfully validated through a continuous 6‐month field deployment in a mountainous forest, proving its practical applicability for remote environmental monitoring.