Researchers from Harvard University are attempting to harness the power of machine learning and apply it to earthquake aftershock detection.

Aftershocks, the smaller, often less powerful quakes following a “main shock,” can be just as damaging as the “main” quake, striking just about anywhere along fault lines.

As such, researchers are applying machine learning to what they already know about earthquake and aftershock behaviors.

"If you think about making forecasts of earthquakes," said study co-author professor Brendan Meade of Harvard University, "you want to do three things; you want to predict when they're going to be, you want to say something about how large they're going to be and about where they're going to be.

"What we wanted to do is to tackle the last leg of this problem — that is where aftershocks are going to be."

To determine where aftershocks might occur, the team sifted through a database of information concerning roughly 100,000 earthquakes and aftershocks in an effort to train a neural network to detect aftershock patterns.

The researchers then set about using the neural network to predict patterns in earthquakes it hadn’t yet been trained on. Instead of inputting data about the main earthquakes into the system to calculate where aftershocks might occur (a current method for predicting aftershocks), the network has the power to investigate other potential pathways.

Using a machine learning method, the network could successfully locate areas surrounding a fault that were most likely to experience, at the very least, a tremor or other shaking event following a main shock.

"The neural network just did better," said lead author Dr. Phoebe DeVries, from the University of Connecticut.

Dr. Elizabeth Cochran, a seismologist with the United States Geological Survey (USGS), who was not involved in the study, described it as "an interesting approach."

"It does give you a really nice picture of where around the fault you should expect aftershocks to be," she said.

"We're quite far away from having this to be useful in any operational sense at all," said Dr. DeVries. "We view this as a very motivating first step."

The study was published in the journal Nature.