Alaska is home to over a hundred volcanoes, many of which reside in the Aleutian Islands and the Alaska Peninsula, and each volcano has a unique ash fingerprint that helps scientists to better understand past eruptions and prepare for future ones.
Tracking ash and other volcanic research is crucial for evaluating hazards, particularly for aviation since ash can cause engines to fail during flight.
In the past, it’s been grueling work to trace volcanic ash to its source, said Jordan Lubbers, a scientist with the Alaska Volcano Observatory. It’s especially tricky when researching ancient ash that may have traveled thousands of miles during a large volcanic eruption, he said.
“It's like playing this connect the dots linking game,” Lubbers said.
There are many factors when identifying ash, like determining the shape of its particles and chemical makeup, then pairing that information to one source on a long list of volcanoes, said Lubbers.
But now, a type of AI known as machine learning is successfully doing that work much faster than before. Scientists can drop an ash sample in a lab and let a machine thoroughly analyze it.
“Computers can think in many different dimensions at once, and process a lot of information — kind of simultaneously — and learn from all of these relationships,” Lubbers said.
Lubbers said they are going to continue to use the machine learning program, adding more data and improving the software over time.
