A super-fast tsunami that ravaged an Indonesian island last year, killing thousands, was almost certainly triggered by underwater landslides, according to a detailed reconstruction of the disaster using surveillance-camera and video footage harvested from amateur posts on YouTube and other social media.
The findings suggest that tsunami warning systems should be upgraded in coastal regions where there is a risk of landslide-triggered events, which can arrive much more quickly than other tsunamis, although some researchers say these events are too fast even for warning tools, and that public education is the key to saving lives.
The results also highlight the power of unconventional research methods. “This is such an important example of citizen science,” says geophysicist Jennifer Haase at the Scripps Institution of Oceanography in La Jolla, California, who led the study.
The Indonesian disaster occurred on 28 September, when a magnitude-7.5 earthquake struck northern Sulawesi. Shortly afterwards, a tsunami hit Palu Bay, a long, narrow and densely populated inlet. The earthquake and tsunami together killed 4,340 people.
Although a tsunami warning had been issued, it predicted only a small wave of about 0.5 metres. In fact, the wave reached 2 metres in much of Palu Bay and reached 8 metres in some places. Eyewitnesses also said it arrived within minutes of the earthquake, rather than tens of minutes as predicted.
The size of the tsunami surprised seismologists because it followed a ‘strike-slip’ tremor, in which continental plates move horizontally. Such quakes should trigger small tsunamis; large waves usually result only when tectonic plates move up and down.
Many tsunami researchers suspected that the quake had triggered a submarine landslide in Palu Bay, causing the tsunami. But there weren’t sufficient data, because Indonesia does not have many of the buoys or tide gauges used to collect them, and researchers struggled to get permits to enter the country to study the aftermath.
So geophysicist Haase and her team went hunting for another source of information.
“We started looking at different social-media platforms like YouTube, Twitter, Facebook, Instagram,” says Matías Carvajal, a seismologist at the Millennium Nucleus The Seismic Cycle Along Subduction Zones, a research collaboration between several universities in Chile. He found 38 amateur video and surveillance clips of the tsunami. Where possible, the team pinpointed the location of the videos on a map and synchronized them. This allowed the researchers to reconstruct how the tsunami had moved through Palu Bay.
The reconstruction shows that the tsunami inundation occurred only a few minutes after the earthquake shaking, and that successive waves came just 1–2 minutes apart. This suggests that the source was close to the shore, which is indicative of a submarine landslide, says Carvajal.
The team published its findings in Geophysical Research Letters1 last month.
Carvajal’s use of “unconventional evidence” to determine how the tsunami behaved is useful because few local data are available for this event, says Anne Socquet, a seismologist at the University of Grenoble in France. Socquet was lead author on one2 of two studies published in February2,3 that used satellite or seismometer data to show that the earthquake had travelled at fast “supershear” velocities.
The study offers a significant piece of evidence supporting the idea that submarine landslides caused the tsunami, says Purna Sulastya Putra, a tsunami specialist at the Indonesian Institute of Sciences in Bandung who has also studied the event.
The results suggest that tsunami warning systems need to be modified to detect landslide-triggered events, says Ignacio Sepúlveda at the Scripps Institution of Oceanography, who co-authored the study. Tsunami warning systems are typically first triggered by seismometers and then verified using recorded changes in sea level. But most verification systems only record long-period waves that displace buoys and tide gauges for several minutes. Indonesia’s systems, like most, aren’t designed to detect landslide-triggered tsunamis with shorter periods.
The Indonesian geophysics agency that issues warnings was widely criticized because it cancelled the Palu tsunami alert half an hour after it was issued. The one tide gauge in Palu Bay that the tsunami passed through didn't pick up the huge waves, says Carvajal.
Accounting for the risk of marine landslides is also difficult because there are no detailed maps of the Indonesian seabed that could be used to pinpoint areas of loose sediment, says Putra. Coastal engineer Mohammad Heidarzadeh of Brunel University London is leading a new effort to map the seabed off Indonesia that will be complete over the next three years, with the goal of forecasting tsunamis more accurately.
But even with maps, tsunamis triggered by landslides usually affect a relatively small area compared to quake-triggered events, and so are particularly hard to predict, says Abdul Muhari of the Ministry of Marine Affairs and Fisheries in Jakarta. He suggests focusing on locations that have experienced landslides before.
And Mika McKinnon, an independent disaster researcher in Vancouver, Canada, questions whether better improved warning systems is the answer to landslide-triggered tsunamis. The best way to save lives is to ensure that people know what to do during a disaster, she says. Tsunami warning systems are useful for places far from the earthquake epicentre, where the wave is minutes away. But when the quake is closer to shore, such as in Palu, the tremor itself is the warning, she says, because even an advanced warning system is unlikely to have gotten an alert out in time. “If you’re on a coastline and feel severe shaking, run for high ground.”
She adds: “We used to think you might have tens of minutes to get to high ground. This event makes it clear you might only have 100 seconds.”
Many coastlines are probably at risk from these tsunamis, says McKinnon. That’s because earthquake-prone areas often include coastlines dotted with steep-sided volcanoes, where landslides can readily occur.
Nature 569, 463-464 (2019)