The new approach to assessing earthquake impacts involves the application of machines to help to determine where aftershocks may occur. It is hoped that the new technique will help to advance knowledge of earthquake behaviour, allowing planners to implemented emergency measures.
Currently forecasting the spatial distribution of aftershocks is challenging. Large earthquakes (the ‘mainshock’) are often followed by thousands of aftershocks. Some of these are, in terms of the damage they cause, indistinguishable from other earthquakes. Aftershocks occur by the same mechanism as earthquakes across the same geological faults. Most of the aftershocks occur within the first hour or day after the main shock.
The need to study aftershocks to address safety concerns is captured by one of the researchers, Dr. Brendan Meade, from Harvard University, who states: “If you think about making forecasts of earthquakes, 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.”
He adds further: “What we wanted to do is to tackle the last leg of this problem – that is where aftershocks are going to be.”
The new technology is a deep-learning approach designed to identify a criterion that could forecast aftershock locations without any prior assumptions about the orientation of the fault.
This was developed by training a neural network using some 131,000 mainshock–aftershock pairs. These were used to predict the locations of aftershocks, with the researchers trying to predict patterns in other earthquakes that the artificial intelligence would not have previously seen. This analysis showed that a typical aftershock pattern is physically interpretable.
The way artificial intelligence application worked was instead of inputting data in relation to a main earthquake through a set of calculations (as currently happens), the neural network used processing power to move through multiple possible pathways in order to make predictions.
Going forwards, this machine-learning-driven insight is set to provide scientists and governments with improved forecasts of aftershock locations. This could allow measures to be identified that could control earthquake triggering during the most active part of the seismic cycle, or to focus on populated areas most likely to be impacted.
Dr. Elizabeth Cochran, who is a seismologist with the U.S. Geological Survey (USGS) and who was not involved in the research, told the BBC: “It does give you a really nice picture of where around the fault you should expect aftershocks to be.”
The new research has been published in the science journal Nature. The research is titled “Deep learning of aftershock patterns following large earthquakes.”
In related earthquake news, the U.S. Geological Survey is developing and testing an earthquake early warning system called ShakeAlert for the west coast of the U.S. The aim is, if funding is secured, to send general public notifications.
Essential Science
This article is part of Digital Journal’s regular Essential Science columns. Each week Tim Sandle explores a topical and important scientific issue. Last week we looked at new step in terms of developing a universal influenza vaccine.
The week before week we explored new research which has found that a moderate carbohydrate intake is optimal for general health and well-being. This goes against other, more recent dietary advice.