The earth rumbled under sleeping giant tortoises on Isabela Island in the Galápagos on June 26, 2018. Soon after, a volcano towering over the island, Sierra Negra, erupted. The volcano's fissures spat out enough lava over the next two months to cover about 19 square miles.

The Sierra Negra's eruption wasn't it's first: It has erupted at least seven more times in the past century alone. Geologists had predicted the eruption's date as early as January of last year, which made the 2018 event so special. They came pretty close to predicting it exactly.

Certainly, the forecast was favorable. Scientists have now figured out how to get their simulations back on track after hitting the mark with their estimates in a paper published in Science Advances today.

In a sparsely populated archipelago, Sierra Negra is just one volcano, but when millions of people live in volcanic danger zones around the world, applying these predictions to other craters could help save countless lives.

The paper's author, Patricia Gregg, a geologist at the University of Illinois Urbana-Champaign, says there is still much to do before volcano forecasting becomes a reality.

It's like forecasting the weather when it comes to eruptions. With so many variables and moving parts, it becomes increasingly difficult to project further into the future while painting a picture of the present. Forecasts for tomorrow might be trusted, but forecasts for a week away may not.

Gregg and her colleagues were incredibly lucky with their Sierra Negra forecast five months prior to the eruption. The forecasters accept that it was a gamble even though the volcano was grumbling by then, with spikes in seismic activity. The test was always intended to be a test, says Gregg. 'We had little faith in our forecast being accurate.'

Sierra Negra, however, is an ideal laboratory to refine volcanic predictions. Considering it erupts every 15 or 20 years, scientists from Ecuador and around the world are constantly monitoring it. In 2017, their instruments detected renewed rumblings that indicated a future eruption was imminent.

Experts know that volcanoes like Sierra Negra blow their top when magma builds up in the reservoir below. As more magma strains against the surrounding rock, it puts the earth under ever-mounting pressure. Eventually, something has to give.

The rocks break, and magma begins to burst through. If geologists could understand exactly how the rocks crumble, they could forecast when that breaking point was likely to occur.

Gregg and colleagues relied on methods familiar to weather or climate forecasters: They combined observational data of the volcano’s ground activity with predictions from simulations.

They then used satellite radar images of the ground beneath Sierra Negra to watch what the bloating magma reservoir was doing and ran models on supercomputers to learn what happens next.

Based on how the magma was inflating by January 2018, their forecasts highlighted a likely eruption between June 25 and July 5. The levels kept rising at the same rate over the next few months—and the eruption began on June 26, right on schedule.

“If anything had changed during those months, our forecast would not have worked,” says Gregg.

“The very tight coincidence of the author’s forecast with the eruption onset must involve some good fortune, but that in itself tells us something,” says Andrew Bell, a geologist at the University of Edinburgh, who has studied Sierra Negra but wasn’t an author on the paper.

The Sierra Negra volcano is seemingly at rest. So, in the years afterward, Gregg and her colleagues combed back over their calculations to determine what they’d gotten right—and what that “something” might be. They ran more simulations using data from the actual eruption to see how close they could get to reality.

What they found was that the magma buildup remained relatively constant over the first part of 2018. By late June, the reservoir had placed enough pressure against the volcano’s underside to trigger a moderately strong earthquake. That seemed to have been the final straw, cracking the rock and letting the magma flow through.

This practice of simulating historic phenomena to check the accuracy of forecast models is sometimes called “hindcasting” in meteorology. In addition to Sierra Negra, Gregg and her colleagues have examined old eruptions from Sumatra, Alaska, and underwater off the coast of Cascadia. 

But is it possible to use the same forecasting techniques in different areas of the world? Every volcano is unique, which means geologists need to adjust their models. By doing so, however, the authors behind the Sierra Negra study found some commonalities in how ground motions translate into the chance of an eruption.

Better forecasting models also mean that scientists learn more about the physical processes that cause volcanoes to rumble to life as they try to match simulations to real-world conditions. “Making genuine quantitative forecasts ahead of the event happening is a challenging thing to do,” Bell says, “but it’s important to try.”