On June 15, 2013, the ground beneath Parkfield, California shifted — but almost nobody felt it. A section of the San Andreas Fault slipped silently, releasing stress over just a few hours without generating the kind of shaking that would trigger any alarms. Scientists had no idea it happened until now.
A research team has uncovered dozens of these hidden slow slip events using artificial intelligence and ultra-sensitive instruments buried deep in the Earth. The discovery, published in the journal Nature Communications, changes what scientists thought they knew about how California's most famous fault behaves between major earthquakes.
The team was led by Dr. Zahra Zali of the GFZ Helmholtz Centre for Geosciences in Germany, working alongside researchers from Stanford University and the organization EarthScope. They focused on Parkfield, a small town that sits directly on the San Andreas Fault and has been studied more intensively than almost anywhere else on Earth. Despite all that attention, some of the fault's most subtle movements had escaped detection for decades.
"Faults can move in ways that do not generate strong seismic waves and therefore escape traditional earthquake detection methods," Dr. Zali said. The problem is that these silent slips leave only faint traces — tiny stretching and compression of the rock that most instruments miss entirely.
To find these hidden signals, the researchers turned to borehole strainmeters, which are among the most sensitive tools available for watching the Earth's crust breathe. These instruments can detect deformations as small as a fraction of a nanometer. The catch? They produce enormous amounts of data, and subtle patterns can easily get lost in the noise.
That's where artificial intelligence came in. The team trained a deep-learning system to recognize the characteristic fingerprint of slow fault slip within years of continuous strain measurements. Rather than looking for a specific signal, the AI learned to group similar patterns together, allowing it to spot events that human analysts had overlooked.
The results revealed the first-ever catalog of short-duration slow slip events at Parkfield derived directly from strainmeter data. The events occurred at shallow depth along the fault and were consistent with the right-lateral motion typical of the San Andreas Fault. But the real surprise came when the researchers compared their findings to records of low-frequency earthquakes — weak seismic signals that occur when rocks grind against each other deep underground.
They found that every time one of these slow slip events occurred, low-frequency earthquake activity in the area increased afterward. It was a connection nobody had clearly documented before.
"These events are difficult to identify by conventional methods because they are small and often hidden within complex background signals," Dr. Zali said. "Artificial intelligence allowed us to recognize their patterns that would otherwise have gone unnoticed."
The finding suggests that the San Andreas Fault is more dynamic between major earthquakes than scientists realized. Silent slips may be quietly building stress and triggering smaller tremors in ways that could eventually help researchers predict which sections of a fault are most likely to rupture next.