Image of the Anak Krakatau volcano erupting
After a flank collapse in 2018, the Anak Krakatau volcano, pictured here, was reduced to a third of its former height. Credit: flydime, CC BY 2.0

On 22 December 2018, a tsunami swept across Indonesia’s Sunda Strait without warning, killing more than 400 people and injuring hundreds more. Experts quickly traced the deadly tsunami to the volcanic island of Anak Krakatau, where a large chunk of the volcano’s south flank had collapsed into the sea, sending waves toward Java and Sumatra—whose early warning system was only equipped to detect tsunamis generated by earthquakes.

Volcano-induced tsunamis are a rare but deadly phenomenon. Indeed, they account for a quarter of all fatalities from volcanic activity. A new study in Nature Communications looked at data collected in the months before the 2018 disaster to understand the events that led to the collapse—and to identify signals that could someday be useful for predicting similar events in the future.

“Volcano sector collapses are seldom, but their impact leads to some of the largest volcano-induced disasters,” said Thomas Walter, head of the Volcano Tectonics and Hazards Group at the German Research Centre for Geosciences (GFZ) in Potsdam and lead author of the new study. “It needs to be understood that these events are low probability but high impact and require closer scientific consideration.”

Volcanic islands are unstable formations. They rise rapidly and erode just as quickly, as the weak and fragmented materials they’re built on give way. Anak Krakatau is no exception. The volcano, whose name in Indonesian means “son of Krakatau,” sits atop the caldera of the former Krakatau volcano, which famously and violently erupted over 2 days in 1883, sending multiple tsunamis across the Sunda Strait and causing tens of thousands of casualties.

The active volcano itself is perched on a steep cliff on Anak Krakatau and as a result may be especially susceptible to tsunami-inducing landslides. Still, the rarity of volcano flank collapses means there is a dearth of data on the early signs of impending landslides.

There have been some 130 volcano-induced tsunamis recorded over the past 4 centuries, but few of these events were monitored with modern equipment.

“It’s not like hurricanes [of which there are] so many…in the world every single year that you can study them in great detail, and have a good numerical model for them, and every year become better at predicting how big they will be and where they will go,” said Thomas Giachetti, an assistant professor of volcanology at the University of Oregon who was not involved in the study. “Here it’s a bit different, because you’ve got so few of these events that are monitored and recorded and witnessed.”

The Hazard Cascade

Fortunately, when the collapse occurred last year, researchers at GFZ Potsdam already had equipment in place monitoring the volcano, according to Walter. “On the day of the event, we started looking deeper into available records,” he said.

In just 2 minutes, the volcano was reduced to roughly a third of its former size, according to elevation models; the island’s peak elevation dropped from 320 meters to 120. Thomas and his coauthors used both ground and satellite-based measures to identify the “hazard cascade” that led to the collapse.

A 2D graphic charting the pre-landslide signals such as flank motion and eruptive deposits, the flank collapse, and the post-collapse eruptive activity at Anak Krakatau.
A cascade of activity preceded the December 2018 collapse of the Anak Krakatau volcano that triggered a deadly tsunami. Credit: Walter et al., 2019,, CC BY 4.0

“Any single one of them taken individually cannot tell you that the volcano would collapse, but taken altogether and monitored through time, then you’ve got some warning signs that something may happen.”

The study is unique for the combination of methods and approaches to track the volcano’s behavior, according to Giachetti. “Any single one of them taken individually cannot tell you that the volcano would collapse,” he said, “but taken altogether and monitored through time, then you’ve got some warning signs that something may happen.”

Infrared records from NASA’s satellite-based Moderate Resolution Imaging Spectroradiometer (MODIS) instrument, for example, showed that in late June 2018, thermal activity kicked up in the Anak Krakatau volcano, reaching intensity levels not seen since monitoring began nearly 20 years ago. During this eruptive phase, which lasted 175 days, some 54 million metric tons of new material was deposited on the southern slopes of the volcano, which eventually gave way.

The GFZ team also used interferometric synthetic aperture radar (InSAR), a technique that uses radar pulses from satellites to create a map of the ground, to track the minute movements of the volcano in the year before the collapse. InSAR maps showed that the southern side of the volcano was already slipping toward the ocean at about 4 millimeters per month throughout the first half of the year. When the eruptive activity increased in June, the flank movement reached 10 millimeters per month and maintained that rate through the fall.

Finally, infrasound records showed high levels of volcanic activity on the day of the collapse, including a spike in high-frequency signals just before the landslide, which the authors believe could have been a small earthquake that triggered the collapse. More research is needed to confirm this hypothesis, however.

A Shock but Not a Surprise

Together, the data show that there were clear precursor signals to the collapse of Anak Krakatau.

“I was surprised by the clear deformation occurrence weeks, months, and even years before the collapse,” Walter said, pointing to previous research from Giachetti that anticipated the event.

In 2012, Giachetti and colleagues noticed the precarious position of Anak Krakatau on the caldera of the former Krakatau and sought to find out what might happen if it gave way. “We just said, ‘What if this happened? What will be the size of the wave? What will be the time of arrival? Which cities or coasts will be most affected?’” Giachetti said. “But we never said, ‘This will happen tomorrow, or in a year, or even in 10 years.’ It was really a ‘what if?’ paper.” But it accurately modeled the eventual tsunami.

“We can’t stop the flanks [from] slipping into the ocean, but we can start understanding the processes and signals associated with these motions.”

“So, still today, I am struck by the question,” Walter said: “Why did this event come as a surprise?”

Of course, knowing that an event will eventually occur and predicting precisely when it will happen are two different things. But Walter hopes that with more monitoring, researchers will be able to develop a robust early warning system for volcano-induced tsunamis.

According to Walter, the seismic and infrasound traces from the landslide itself could help researchers make progress toward this end, but more monitoring is needed to fully understand the complex processes that govern flank stability at Anak Krakatau and elsewhere.

“Volcano flank instability and sector collapses can and need to be monitored for reliable early warning,” Walter said. “We can’t stop the flanks [from] slipping into the ocean, but we can start understanding the processes and signals associated with these motions.”

—Kate Wheeling (@katewheeling), Freelance Writer


Wheeling, K. (2019), The hazard cascade that led to the Anak Krakatau landslide, Eos, 100, Published on 04 November 2019.

Text © 2019. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.