Earthquake signs even earlier – could save lives

International the research team developed a technique that could be used to predict large earthquakes months in advance. If it works, it would help prepare for earthquakes in advance and save lives. In the past, earthquakes have hardly been predicted.

The new technology is based on machine learning. Scientist at the Alaska Geophysical Institute Tarsilo Girona and his colleagues developed an algorithm that analyzes extensive data collected from earthquake areas. Among them, it tried to identify signs that predict large earthquakes.

The study focused on two earthquake series, the first of which occurred in Achorage, Alaska in 2018 and the second in Ridgecrest, California in 2019. The magnitude of the earthquakes ranged between 6.4 and 7.1 on the Richter scale.

A magnitude greater than 7 means a very strong earthquake. The Alaska earthquake collapsed a bridge and damaged buildings and roads, among other things. In California, the earthquakes broke houses and water pipes and cut power lines.

Earthquake occurs when the stress stored in the bedrock exceeds the friction of the aggregate. In this case, the bedrock breaks and the boulders move.

Forecasting earthquakes would require accurate information on where and when the stress in the bedrock becomes too great.

When looking at the data, the researchers found that small seismic activity occurred in Alaska and California as early as three months before large earthquakes. These foreshocks were mostly below 1.5 magnitude.

The researchers think that the seismic activity originates from the pressure of the fracture points in the earth’s crust. This pressure is reflected in the so-called pore fluid. Pore ​​liquid is a liquid bound to the aggregate, which fills the empty pores and cavities of the stone.

As the pressure of the pore fluid increases, the friction between the rock blocks decreases. Then they move more easily in relation to each other. This can lead to an earthquake.

The changing pressure of the pore fluid, on the other hand, shows up as an uneven variation in the stress field, which causes small foreshocks that precede the earthquake.

By observing them, the algorithm predicted three months in advance in Alaska that a large earthquake would occur in the area within the next 30 days with an 80 percent probability.

A few days before the earthquake, the probability increased to 85 percent. Similar results were also obtained in California.

Next the researchers intend to test their algorithm in near-real-time situations to identify its potential problem areas and shortcomings.

According to Társilo Girona, who led the research, the reliable prediction of earthquakes still involves big ethical and practical questions.

If earthquakes could be accurately predicted, it would enable early preparation and evacuation. People would be saved and economic damage would be smaller. On the other hand, there is always uncertainty in predicting earthquakes.

“False alarms cause unnecessary panic, disrupt the economy and undermine public confidence. Predictions that go wrong could have catastrophic consequences,” says Girona of the University of Alaska in the bulletin.

Research was published in the scientific journal Nature Communicatios.

Published in Tiede magazine 12/2024