Inside China’s Internet addiction centers
Earthquakes, once rare and unpredictable, are becoming increasingly frequent and severe. Scientists have been racing to develop tools to predict these natural disasters, but with limited success.
Now, artificial intelligence is offering a glimmer of hope. By analyzing vast amounts of data, AI is helping scientists identify patterns that could signal an impending quake. Can this technology finally provide the answers we've been seeking? Click on to find out.
Geologists have been attempting to use modern scientific methods to predict earthquakes since at least the 1960s, but with little success.
Many scientists have long considered earthquake forecasting to be impossible—or, at best, they have approached it with extremely tempered optimism.
Chris Marone, a professor of geosciences at Penn State University in Pennsylvania, explains that one factor hindering earthquake prediction is the complexity of the fault systems that crisscross the globe.
Faults are fractures or zones of fractures between two blocks of rock. Movement along these faults can occur rapidly, resulting in an earthquake.
However, faults are not always easily identifiable. Most are underground and are difficult to discover.
We may not hear it over all the daily noise, but the Earth is constantly rumbling and grumbling. This, coupled with traffic, construction, and other noises, makes it difficult to detect clear seismic signals.
According to the United States Geological Survey, a truly useful earthquake prediction requires three elements: location, timing, and magnitude. So far, they say, no one can predict these with certainty.
Instead, geologists have been producing "hazard maps," which estimate the probability of an earthquake occurring within a specific timeframe.
While these maps can aid in planning, such as improving building standards in high-risk areas, they don't offer the precision needed for early warnings that would allow people to evacuate or take shelter.
As natural disasters become increasingly common, scientists are exploring various methods to improve earthquake prediction, from seismic signals to animal behavior and atmospheric changes.
Enter artificial intelligence. Several seismologists (researchers studying earthquakes) are applying this technology in the hope of filling the earthquake prediction gap.
Using AI multiplies and accelerates a single scientist's capabilities. Since earthquake prediction relies on numerous data points and is extremely time-sensitive, this is a significant advantage.
AI programs can process multiple seismic records simultaneously, rendering them with high precision in 3D, faster than any human could.
Geophysicist Zachary Ross and his team at the California Institute of Technology have built an AI program and are already testing it in California. This region is highly geologically active and densely instrumented with seismometers.
Now, they're using a computer program to listen to the Earth's constant hum. This program is searching for any new "songs" that match the ones in their library. If it finds a match, it could be a sign of an impending earthquake.
The algorithm quickly identified nearly two million previously undetected tiny quakes from 2008 to 2017. These quakes revealed an intricate network of faults and fault features that previous earthquake searches had missed.
This program, a precursor to fully-fledged AI software, could only identify earthquakes in the seismic record that matched its training data. Meaning new seismic events were overlooked.
To improve earthquake prediction, Ross turned to more advanced tools: self-learning programs that could use existing information to predict the sound of a wider range of earthquakes.
These programs quickly identified many unfamiliar-sounding quakes, which were later confirmed by human scientists.
These machine learning programs continue to evolve, and they're now doing more than just identifying quiet quakes and hidden faults.
They have already been deployed all over California, where they've identified a new type of long-lasting, slow-moving earthquake swarm.
In Hawaii, the new programs discovered a previously unknown network of pulsing, migrating molten rock beneath two active volcanoes. Traditional seismic analysis methods were unable to detect this hidden activity.
As these programs continue to develop, they could be used to improve the speed and accuracy of earthquake early warning systems. "This is light-years beyond what we could have done a few years ago," says Ross. "It's at superhuman levels now."
In earthquake-prone Tokyo, for example, AI software will scan images from cameras perched at high points throughout the city to recognize fires and building collapses and alert authorities the moment they happen.
In September 2023, researchers at the University of Texas also made progress in earthquake prediction. An AI algorithm correctly predicted 70% of earthquakes a week in advance in a trial in China.
Another group based in Israel recently claimed to have developed a machine-learning model that can predict large earthquakes 48 hours in advance with 83% accuracy. This model analyzes changes in electron content in the ionosphere over the past 20 years.
Despite these advancements, AI programs won't replace human scientists, Ross guarantees. "They're just tools," he says, but they could become as essential as seismometers in the future.
Sources: (National Geographic) (BBC) (Smithsonian Magazine)
See also: The San Andreas Fault: is a big earthquake coming to California?
How to predict the next monster earthquake? Seismologists bet on AI
Artificial intelligence could unlock earthquake prediction
LIFESTYLE Natural disaster
Earthquakes, once rare and unpredictable, are becoming increasingly frequent and severe. Scientists have been racing to develop tools to predict these natural disasters but with limited success.
Now, artificial intelligence is offering a glimmer of hope. By analyzing vast amounts of data, AI is helping scientists identify patterns that could signal an impending quake. Can this technology finally provide the answers we've been seeking? Click on to find out.