New Machine Learning Model Helps To Predict Volcanic Eruptions

With the use of artificial intelligence, researchers have shown us how new insights can be gained about volcanoes and the likelihood of an eruption occurring.

Andrew Hooper, a volcanologist at the University of Leeds in the United Kingdom have developed a machine learning algorithms that can automatically predict volcanic eruptions from that data signals of volcanic risk.

The machine learning algorithm will work over the data provided by satellites about the world’s active volcanoes. With the launch of the European Space Agency’s satellites Sentinel 1A and Sentinel 1B, the field of volcanology has received frequent, repeated views of how the ground shifts around the world’s volcanoes.

Radar interferometry is a technology that Sentinel 1 use to compares radar signals sent to and reflected from Earth to track changes in the planet’s surface. The Sentinel 1 satellites revisit each spot on the planet once every 6 days, and the Sentinel team releases those high-resolution observations rapidly.

A group of researchers in the United Kingdom called the Centre for Observation and Modelling of Earthquakes, Volcanoes, and Tectonics (COMET) is developing the database of these ground-movement snapshots, called interferograms.

Hooper, who works with COMET says, “Overlaying this database with automated detection seemed natural given the success machine learning has had in other forms of pattern detection.”

The machine learning system will work with thermal hot spots or ash plumes, which can be automatically detected with weather satellites. The researchers taught the algorithm not to confuse atmospheric shifts for ground motion, something interferograms are prone to do.

Hooper’s and his team use a technique called independent component analysis, which simply learns to break apart a signal into different pieces: such as the stratified atmosphere or short-term turbulence, along with ground shifts in a volcano’s caldera or flank.

The technique allows them to catch both brand-new ground motions, or changes in rate, both of which can be signs of a pending eruption.

In COMET group there is another team of the researcher was led by Juliet Biggs. Juliet Biggs is a volcanologist at the University of Bristol in the United Kingdom. Juliet Biggs and his team created a second algorithm to support the first one.

This second algorithm uses a very popular form of artificial intelligence called convolutional neural networks, which use layers of biologically inspired “neurons” to break apart features of images into ever-more-abstract pools, learning how to tell, for example, cats from dogs.

Although some continued technical hiccups on COMET’s volcano database have prevented the teams from running their algorithms close to real-time on all volcanoes, Hooper has run their technique on select spots, including the volcanic peaks known as Sierra Negra and Wolf on the Galápagos Islands.

Both erupted this past year, and Hooper’s program caught both as their unrest started, he reported yesterday at the meeting.

The two algorithms are complementary; the neural network, for example, cannot catch very slow changes in deformation, but the independent component analysis can. So it’s likely that COMET’s warning system will use both, Hooper says.

For now, the challenge is speeding up how quickly COMET can pull the radar data from Sentinel into its database. Although these data are available from Sentinel within a few hours, it still takes several weeks for them to fully transfer.

It’s painstaking work, Hooper says. “We thought we’d be further along.”

The researchers first trained their neural network using raw interferograms from Envisat, Sentinel’s precursor, for which they had existing examples of eruptions. Although the algorithm had some success on an analysis of 30,000 Sentinel interferograms, it still produced too many false positives.

There were simply too few examples to learn from, says Fabien Albino, a volcanologist who works with Biggs at Bristol. “For machine learning, 100 is nothing. They want thousands and thousands.”

Andrew Hooper says the new algorithms should benefit the roughly 800 million people who live near volcanoes. “About 1400 volcanoes have potential to erupt above the sea,” he says. The vast majority aren’t.” Both methods were presented this week in Washington, D.C., at the semiannual meeting of the American Geophysical Union (AGU).

Without such tools, geoscientists simply can’t keep up with information pouring out the satellites, says Michael Poland, the scientist-in-charge of the U.S. Geological Survey’s Yellowstone Volcano Observatory in Vancouver, Washington, who was not involved in either study. “The volume of data is overwhelming,” he says.

More in AI :

Nvidia’s new GPU TITAN RTX has Monster power for Deep Learning

Google’s Machine Learning Model Decode Humpback Whale Songs

Economic Value Of Artificial Intelligence, It’s Growth and Impact

Leave a Reply

Your email address will not be published.