INFRAS-CLOUD presents a physics-informed eight-parameter Atmospheric Infrasonic Severity Index (AISI) for real-time classification of severe weather, volcanic, and ocean-atmosphere infrasonic events. Eight governing acoustic parameters — spectral peak frequency (fₚ), microbarom amplitude (Pᵤb), azimuthal arrival angle (θ), stratospheric ducting efficiency (Dₛtr), phase velocity (vₚh), inter-station coherence (γ²), atmospheric absorption coefficient (αₐir), and signal-to-noise ratio (SNR) — are integrated via PCA-regularized logistic regression into a single composite severity score (AISI). Validated across 1, 847 events from 47 International Monitoring System (IMS) stations spanning 2005–2025, INFRAS-CLOUD achieves 93. 1% classification accuracy across six source categories (tropical cyclones, tornadoes, volcanic explosions, microbaroms, mountain-associated waves, anthropogenic). Key results include: tornado precursor lead times of 12–28 minutes prior to NWS confirmation; tropical cyclone detection at ranges up to 4, 200 km; volcanic energy estimation within 10% of independent seismic benchmarks (Hunga Tonga 2022: 38 ± 4 MT TNT) ; and stratospheric wind inversion accurate to ±7. 8 m/s without radiosonde requirements. The AISI framework is released as an open-source Python package (pip install infrascloud) under MIT License.
Samir Baladi (Wed,) studied this question.