Connection performance analysis of 3D clustered radiosonde networks in a typhoon

Radiosonde networks are essential for real-time atmospheric profiling during typhoons. However, conventional star-topology networks often suffer from link degradation and data loss due to long-distance attenuation and severe interference in extreme weather. To address these challenges, this paper proposes a Three-Dimensional (3D) clustered networking model, leveraging Cluster Heads (CHs) as relay nodes to shorten communication links and enhance transmission robustness. Focusing on Connection Performance (CP), we establish a theoretical analysis framework by integrating stochastic geometry with a Non-homogeneous Poisson Cluster Process (NPCP). The proposed framework captures the non-uniform spatial aggregation of radiosondes inside typhoons and employs a local uniform approximation to derive analytical CP expressions under three representative typhoon phases. The derived results explicitly account for path loss, rain attenuation, and heterogeneous interference from CHs, Cluster Members (CMs), and Non-Cluster Radiosondes (NCRs). Numerical results and Monte Carlo methods validate the proposed framework, demonstrating that clustered architectures significantly outperform traditional star topologies by optimizing transmission distances and redistributing interference. Furthermore, the impacts of node density, path loss exponent, and power compensation factor on CP are systematically investigated, revealing fundamental trade-offs between observation data density and communication reliability. These findings provide design-oriented insights for deploying and optimizing radiosonde communication architectures to support robust and continuous typhoon observation.