When a seismic network failed, citizen science stepped in

  News
image_pdfimage_print
The Raspberry Shake, a simple seismograph based on Raspberry Pi hardware.
Enlarge / The Raspberry Shake, a simple seismograph based on Raspberry Pi hardware.
Mike Hotchkiss, Raspberry Shake

On the afternoon of January 12, 2010, a magnitude-7.0 earthquake struck about 16 miles west of Haiti’s capital of Port-au-Prince. Among the most significant seismic disasters recorded, more than 100,000 people lost their lives. The damage—costing billions of dollars—rendered more than a million people homeless and destroyed much of the region’s infrastructure. The earth tore at the relatively shallow depth of about 8 miles, toppling poorly constructed buildings.

At the time, Haiti had no national seismic network. After the devastating event, scientists installed expensive seismic stations around the country, but that instrumentation requires funding, care, and expertise; today, those stations are no longer functional. In 2019, seismologists opted to try something different and far less expensive—citizen seismology via Raspberry Shakes.

On the morning of August 14, 2021, amidst a summer of COVID-19 lockdowns and political unrest, another earthquake struck, providing the opportunity to test just how useful these Raspberry-pi powered devices could be. In a paper published on Thursday in Science, researchers described using the Raspberry Shake data to demonstrate that this citizen science network successfully monitored both the mainshock and subsequent aftershocks and provided data integral to untangling what turned out to be a less-than-simple rending of the earth.

More powerful earthquake 

The August 2021 event clocked in with a magnitude of 7.2—40 percent more powerful than its 2010 predecessor. It ruptured along the same fault zone but in a more rural region, resulting in comparatively fewer losses. Nevertheless, about 2,500 people lost their lives, 13,000 were injured, and at least 140,000 houses were destroyed or damaged.

One of the Raspberry Shakes installed in 2019 happened to be sited about 21 kilometers from the epicenter, with two more citizen stations near enough to detect the quake. Along with two other seismic stations in Port-au-Prince—one at the US embassy and another educational instrument in a local high school—the Raspberry Shake notification came through within a minute of the earthquake, said Eric Calais. (Calais was among the scientists leading the international response after both the 2010 and 2021 earthquakes and a leader of the citizen science initiative.) Two more Raspberry Shakes near the epicenter, unavailable during the main shock because of Internet connectivity issues, were reconnected by their hosts within two hours.

A less-than-useful network

To detect any rumblings in the ground—including earthquakes—you need a seismic station packed with sensors, a means to record the data, a place to store that information, and power to run the whole contraption. Seismologists typically rely on expensive seismic stations that must be carefully installed to minimize background vibrations caused by people, wind, and even atmospheric pressure changes. To report data in real time, stations need constant communication via cellular networks or satellite links. These stations must be maintained by specialists trained specifically for this purpose.

The Raspberry Shake network in Haiti.
The Raspberry Shake network in Haiti.

After the 2010 earthquake, the newly installed conventional seismic network was to be maintained by Haiti’s Bureau of Mines and Energy. By 2018, when a magnitude-5.9 earthquake killed 17 people, none of these stations was functional, forcing Haiti’s population to rely on information gathered remotely by the US Geological Survey.

According to coauthor and seismologist Anthony Lomax, his impression from Haitian scientists is that a major impediment to a stable seismic network is general lawlessness, ranging from theft of equipment to ransom kidnapping.

“The three main roads out of Port-au-Prince to the provinces are controlled by gangs,” agreed Calais. “The government had to pay them to stop shooting and robbing so that humanitarian help could go through after the quake.”

Raspberry Shake to the rescue  

Raspberry Shakes—cheap plug-and-play seismic stations that require little maintenance—can circumvent many of the problems plaguing the conventional seismic network. Backed by a Raspberry Pi computer that manages the upload of data to servers, Raspberry Shakes need an Internet connection and wall socket to provide data storage and power, respectively. While conventional seismic stations can cost well over $10,000 each, these instruments are a fraction of that: about $400.

Although multiple models exist that can measure different things (like this Raspberry Shake and Boom, which also includes an infrasound detector), the scientists responsible for deploying Haiti’s citizen science network opted for the Raspberry Shake 4D, which includes a vertical velocity detector and accelerometers that measure movement in two horizontal directions as well as up-down. Funding for this project comes mainly from two French institutes, said coauthor Françoise Courboulex. The Raspberry Shake network was mostly installed across Haiti by Steeve Symithe and Calais.

Symithe, Calais, and their fellow citizen scientists placed stations in convenient locations, like the above-mentioned living rooms, which tended to be rather noisy locales. Ambient vibrations picked up by these stations are typically much higher than a conventional seismic station that’s shielded from the tremblings of everyday life by specially designed vaults. In spite of the noise, these Raspberry Shakes still provide valuable information in a country lacking other seismic instrumentation.

Currently, many stations provide data in real time, available on the ayiti-seismes platform. This network can detect much smaller magnitude earthquakes in Haiti than other Caribbean regional networks, with the latest locations and magnitudes available on the website.

Shake test 

The Raspberry Shake station nearest to the earthquake, R50D4, provided invaluable information both during and after the earthquake. First, the peak ground acceleration—the maximum acceleration the ground experienced during an earthquake at the location of that seismic station—was slightly greater than expected. The expected value went into building codes published in 2012. Acceleration and shaking, said Lomax, are typically greater on higher floors. This implies that newer, multistory buildings weren’t designed to withstand the 2021 event.

Building damage in Haiti was extensive after the 2021 earthquake.
Enlarge / Building damage in Haiti was extensive after the 2021 earthquake.

In any case, only about less than 10 percent of Haitian buildings are designed and verified by engineers, said Calais. “It is up to the engineers to follow the code, or not,” he explained. “There is no liability.”

The single well-located station, R50D4, also provided seismologists the opportunity to test whether they could use machine learning to identify aftershocks using only a single seismic station. They trained the algorithm to detect earthquakes greater than magnitude-3.0 using databases of earthquakes and noise. This machine-learning procedure applied to station R50D4 gave the time and approximate magnitude for any subsequent aftershocks near the station, said Lomax.

The resulting catalog compared incredibly well with the catalog of aftershocks produced by the entire citizen seismic network. “AI is quite powerful at finding signals hidden in the noise, provided the algorithm has been properly trained to recognize earthquakes,” said Calais.

This study highlights the importance of observations near the site of rupture, said seismologist Wenyuan Fan, who was not involved in this study. “Even sparse, low-cost, relatively noisy observations could aid hazard mitigation and risk management.”

Clusters of slip, clusters of quakes

One of the other things the Raspberry Shake data provided information on is the complicated nature of the fault system in Haiti. The fault on which the 2010 and 2021 earthquakes struck seems to primarily be strike-slip, in which two tectonic plates grind past each other instead of moving toward or away from one another. But in Haiti, the Caribbean plate scrapes against the North American plate while also pushing toward it. This oblique movement means that earthquakes can be both strike-slip, while also having a reverse component, in which they come together, here at an angle. Those reverse faults hide underground, and unlike strike-slip faults, may never broach the surface while wreaking havoc from below.

The 2010 earthquake contained elements of both these types of plate motion, and analysis of seismic data suggests the same happened in 2021.

Based on their analysis, which includes both Raspberry Shake data and information from conventional seismic stations located at a distance, Calais and his colleagues found that the 2021 earthquake could be split into two separate sub-events. The earthquake began as a thrust, in which one side moved up relative to the other, but this was mostly hidden—this part of the earthquake didn’t breach the surface. The second sub-event was the strike-slip component that occurred west of, and after, the first part. This rupture was shallower and broke the surface, according to Calais.

The thrust-sub event may have “triggered” the strike-slip one. That they are merely coincidence is nearly impossible, said Lomax.

“More work will tell us,” said Calais, what the relationship is between these two parts of the whole earthquake. “The earthquake could have stopped after sub-event one, but the rupture carried enough energy to jump to another nearby segment,” he said, “same as 2010!”

Science, 2022. DOI: 10.1126/science.abn1045

Alka Tripathy-Lang is a freelance science writer with a Ph.D. in geology. She writes about earthquakes, volcanoes, and the inner workings of our planet.

https://arstechnica.com/?p=1840555