In the fall of 2017, Colgate Research Associate Dennis Geist noted signs of imminent eruption at the Sierra Negra volcano in the Galápagos Islands. This June, he co-published a paper with program designer Patricia Gregg, associate professor of geology at the University of Illinois, and colleagues, further exploring those signs of instability, the volcano’s subsequent eruption in 2018, and the promise of the technology that helped them to forecast it five months in advance.
Before volcanoes erupt, magma rises from deep within the Earth to shallower levels — less than a mile deep in the case of Sierra Negra. This rising magma causes the volcano to swell anywhere from a few millimeters to a few centimeters. While these changes are invisible to the naked eye, Geist relies on sensors, using modern satellite GPS technology, to precisely note any inflation.
Geist’s work monitoring Sierra Negra began more than two decades ago. A team of geologists, including Geist, Colgate geology department colleague Prof. Karen Harpp, students like Erika Rader ’07, and collaborators from the University of Idaho, installed instruments capable of measuring and recording the mountain’s inflation, deflation, and seismic activity. Packages carried up the slopes, under the equatorial sun, weighed approximately 50 pounds and contained a GPS, antenna, and receiver, as well as car batteries and solar panels to keep sensors up and running.
“It was extremely hot and difficult work, but well worth the challenge,” reflects Geist.
For over a decade, Geist and his colleagues have continued to improve the sensors and monitor the data that they collected, publishing a half-dozen papers based on information mined from the volcano. But it was through a chance meeting at the National Science Foundation that Geist learned of Patricia Gregg’s forecasting software and suggested that she apply it to the highly active Sierra Negra volcano.
Gregg took Geist’s advice and, in January 2018, entered his data into an eruption forecasting model that she and her team had been developing for the past several months. This initial run through yielded a predicted eruption date between Jun. 25 and Jul. 5, a span of 10 days.
On Jun. 26, a 5.4 Mw earthquake occurred, triggering Sierra Negra’s long anticipated eruption. As soon as Geist noted the eruption, he contacted Gregg to confirm the range of dates her forecasting model had predicted. To both Geist and Gregg’s amazement, Sierra Negra had erupted just one day after Gregg’s earliest prediction date — remarkable precision compared to the capabilities of previous forecasting models.
“I never anticipated my data being used in this way, but it has been a fascinating experience,” says Geist.
Prior to Gregg’s innovations, inflation data were used to make broad predictions — forecasts with windows of years. Gregg’s supercomputer model predicts a span of days within which the eruption will occur. Moreover, “Each day, when new measurements come in,” Geist says, “it assimilates new data and improves its predictions going forward.”
While the successful prediction of the Sierra Negra eruption is a promising start, Geist and Gregg’s new article underscores the fact that the forecasting program continues to undergo rigorous real-world testing and improvement. Indeed, one of the reasons that the model worked at Sierra Negra is that the mechanics of its eruptions are relatively simple and well understood — other volcanoes are likely to be much more complicated.
“Continuing to improve upon a model like this is vital,” explains Geist. “Even if the weather report is right today, that doesn’t necessarily mean that it will be right every day for the rest of time.”