AI Lifts Weak Seismic Detection in 30-Year NORSAR Study Using Multi-Sensor Data
Updated
Updated · Eos · Jul 2
AI Lifts Weak Seismic Detection in 30-Year NORSAR Study Using Multi-Sensor Data
1 articles · Updated · Eos · Jul 2
Summary
30 years of seismic-array records showed AI can detect weak earthquake and underground-test signals more reliably than classic methods when it combines readings from multiple sensors.
Three training setups were tested, and the most accurate first amplified weak signals by combining array data before training; a model trained on all stations at once ranked between the others on accuracy.
That all-stations approach was the most computationally efficient, leading researchers to recommend it for real-time monitoring, while slower workflows can combine data before or after model application for higher accuracy.
The model generalized poorly outside its training regions because the dataset was geographically limited, a problem seen mainly for S waves; P-wave detection transferred well, and global training data is expected to improve results.
Beyond spotting nuclear tests, can this AI help us find new geothermal energy sources deep within the Earth?
As AI gets better at detecting secret nuclear tests, are nations finding new ways to hide their explosions?
If an AI mistakes an earthquake for a nuclear test, what prevents it from triggering a global crisis?
NORSAR AI Achieves Record Sensitivity in Seismic Event Detection: Implications for Global Security and Geohazard Monitoring
Overview
NORSAR researchers, led by A. Köhler, have achieved a major breakthrough in seismic monitoring by publishing a study in 2026 that introduces advanced AI models trained on decades of seismic data. These AI models can now identify weak seismic signals, such as faint tremors from small earthquakes or underground nuclear tests, which were previously undetectable by traditional methods. By learning to recognize genuine seismic patterns, the AI brings subtle signals—often hidden by environmental noise—within reach of detection. This marks a new era in geophysical surveillance, revolutionizing how we monitor and understand Earth's subsurface activity.