Five to 10 minutes. That’s what it takes for superheated volcanic ash to shoot 11 kilometers into the sky — reaching altitudes at which commercial jets cruise and potentially harming their engines.
Now scientists have developed a new algorithm that can identify and track explosive ash clouds soon after volcanoes erupt. Using satellite imagery, the program can measure the temperature, height and trajectory of the expanding clouds within about three minutes, researchers report online November 8 in Earth and Space Science.
By tracking these ash plumes in near real time, scientists can alert aviation authorities if there is a need to alter any volcanic ash advisories or change the flight paths of any planes barreling toward hazardous eruptions. “Timely detection is crucial,” says study coauthor Michael Pavolonis, a physical scientist with the National Oceanic and Atmospheric Administration in Madison, Wis.
The new technology could be especially useful for tracking unmonitored volcanoes in remote regions. Of the roughly 1,500 active volcanoes across the globe, fewer than 10 percent are monitored.
The algorithm works by scanning images taken by weather satellites like NOAA and NASA’s joint venture, the Geostationary Operational Environmental Satellite system, and Japan’s Himawari-8. These satellites zip around the equator snapping pictures of large swaths of the Earth as frequently as every 30 seconds.
The challenge is telling different types of clouds apart, Pavolonis says. To distinguish the eruption of volcanic ash clouds and the formation of large thunderstorms, for example, the algorithm analyzes something called brightness temperature. As superheated ash clouds surge into the sky, they cool quickly as they near the stratosphere.
The researchers trained the algorithm on 79 volcanic eruptions seen in satellite data from 2002 to 2017. When the algorithm used data from earlier satellite generations, it accurately identified ash clouds about 55 percent of the time. Using data from newer satellites, the program spotted the clouds in nearly 90 percent of cases.
Mike Burton, a volcanologist at the University of Manchester in England who was not involved in the study, is eager to see artificial intelligence used in monitoring volcanic eruptions. “It’s nascent, but I think there’s enormous potential for machine learning to be applied.”