Swiss researchers develop AI system that listens for mountain sounds to predict and detect avalanches in real time

Swiss AI tool from WSL analyzes infrasound and vibrations to spot avalanches instantly, enhancing safety in the Alps through precise, real-time acoustic monitoring.

By: AXL Media

Published: Feb 23, 2026, 5:28 AM EST

Source: The information in this article was sourced from swissinfo.ch.

Acoustic monitoring targets hidden avalanche risks

Avalanches often occur in remote, high-altitude zones where visual observation proves difficult or impossible, particularly during poor weather or at night. Traditional detection relies on seismometers or infrasound arrays, yet these methods can struggle to differentiate avalanche signals from wind, rockfalls, or distant thunder. The new AI approach processes continuous audio streams to isolate characteristic low-frequency patterns produced by large snow masses in motion.

Machine learning trained on extensive mountain sound library

Researchers compiled a comprehensive dataset of recorded sounds from known avalanche events alongside background mountain noises collected over multiple winter seasons. According to experts involved, the model learns distinctive spectral features, including the rumbling infrasound below 20 Hz and associated ground vibrations, allowing it to classify events with greater precision than rule-based algorithms. Training incorporated both labeled avalanche incidents and negative examples to reduce false positives.

Real-time analysis delivers immediate classification

Once deployed, the AI processes incoming sensor data on edge devices or cloud servers within seconds, flagging potential avalanches as they begin rather than after the fact. This speed represents a significant advance over manual review of seismic records, which can delay alerts by minutes or hours. In test deployments across selected Swiss Alpine sites, the system demonstrated reliable performance even amid complex acoustic environments with overlapping natural sounds.