But, artificial intelligence is now making a difference. Scientists say that they can gather and analyze large amounts of seismic data to better anticipate earthquakes.
Using Neural Networks
AI-based research on earthquake is relying on neural networks – the technology that powers digital assistants and self-driving cars – to find patterns in data that would be impossible, or at the very least, very difficult for scientists to find because of the sheer volume of seismic data.
This data, which is like the audio data that companies use to train digital assistants in interpreting vocal commands, is better studied by neural networks that function like the human brain but are faster.
Quicker Analysis of Voluminous Seismic Data
Seismic data is like audio data. Instead of words, they consist of series of ground-motion measurements that demonstrate similar patterns. Scientists believe that machine learning algorithms and neural networks are making the analysis of this data faster by approximately 500 times.
What took days before now takes merely minutes. And with the prospect of smaller and cheaper sensors in the future, the hope is that scientists will be able to analyze and learn from even larger volumes of seismic data.
Apart from the speed of analysis, artificial intelligence is also garnering new insights. One research paper from Harvard and Google shows that neural networks are effective in predicting the aftershocks of earthquakes.
Research at other institutes like Caltech is using artificial intelligence to monitor earthquakes when they are happening, identify their epicenters and determine where it will spread. These works are changing the nature of earthquake science, and they are centered on artificial intelligence and machine learning.
Earthquake scientists concede that the use of artificial intelligence and neural networks has limitations. For instance, although they can identify signals that they know in data, artificial intelligence is not particularly suited to identifying new signals like the sounds from tectonic plates when they brush each other.
But these limitations are being overcome too. In Mexico, researchers used a machine learning method of classification known as “random forest” to identify a sound with the potential to indicate the arrival of an earthquake, which was previously thought to be ineffectual.
Future of Earthquake Science
Some seismologists and scientists are still skeptical about the use of artificial intelligence in predicting earthquakes. They believe future earthquakes are not dependent on past ones.
Therefore, artificial intelligence and machine learning will only be effective, if they can predict the random chance of an earthquake occurring that does not rely on past data.
Other scientists in Mexico, Japan and the United States, on the other hand, are more optimistic. These countries have systems that warn hospitals and other institutions of earthquakes early so that they are better prepared to face them.
The scientists working in these countries believe artificial intelligence can predict earthquakes and their intensities with greater accuracy. They simply believe that with more data, they will be able to reap better forecasts using neural networks and artificial intelligence.