The researchers’ approach is based on a simple idea: under normal conditions, the background ‘noise’ recorded by a seismic station follows a predictable pattern known as a Gaussian distribution. The technique is designed to detect subtle, long-term changes in the statistical properties of this background signal, which can result from environmental phenomena such as atmospheric disturbances.

The researchers demonstrate the approach through a specific application: detecting changes that indicate problems with the monitoring equipment.

Instead of relying on fixed thresholds, which can vary between different stations and environments, the team developed a method that automatically separates the normal background signal from the unusual data that doesn’t belong.

Testing the technique on four permanent seismic stations in France showed it could detect both subtle and severe changes in signal quality. In some cases, it even highlighted problems before they became obvious using traditional monitoring methods, allowing issues caused by humidity and sensor degradation to be identified earlier.

The researchers suggest that a simple measure of how much a signal departs from its expected Gaussian behaviour could become a valuable new tool for monitoring seismic networks around the world.

Importantly, the method isn’t limited to seismology. Because it can identify tiny departures from the normal behaviour of almost any signal without relying on arbitrary thresholds, it could also prove useful in fields ranging from acoustics and pressure monitoring to magnetism and other types of sensor data, opening the door to a wide range of future applications.