Science is a constant work in progress; despite breakthroughs, humanity still struggles to fully explain many natural phenomena. That includes predicting volcanic eruptions. While extensive monitoring networks do give researchers some precursors to eruptions, these systems remain imperfect.
But a new monitoring system named “Jerk” might offer a reliable, simple solution to this issue. Described in a December 2025 Nature Communications study, Jerk is a single broadband seismometer capable of automatically identifying very early signals of volcanic eruptions in real time—and it’s fully automated.
Typically, announcements on innovative technologies come before the thing has been out in the real world, with the conclusion being something to the effect of, “We think it should work the way we’re saying it will!” And even if the tool does get tested, it’s rare that experiments go on for years, especially if that tool is meant to survey extreme environments like a volcano.
Well, somehow Jerk is none of these things. The team behind Jerk installed its tool at the Piton de la Fournaise on La Réunion Island, France, in 2014, where it still sits today. The latest paper covers a 10-year survey period, during which Jerk successfully predicted 92% of the volcano’s eruptions, sending alerts between minutes and eight hours in advance.
That couldn’t have been an easy feat. So Gizmodo reached out to François Beauducel and Philippe Jousset, geophysicists who conceptualized, executed, and tested Jerk for over a decade. In this Q&A, Beauducel and Jousset, of the Institut de Physique du Globe de Paris in France and the Helmholtz Centre for Geosciences in Germany, respectively, reflect on the various challenges and insights during their unique decade-long journey—and where things might go from here.
The following conversation has been edited lightly for grammar and clarity.
Gayoung Lee, Gizmodo: Generally speaking, why has it been so difficult for us to predict volcanic eruptions or earthquakes?
François Beauducel: We have many observations of the atmospheric conditions. So we can put all this data into a very complex model and make not perfect, but efficient, predictions in the short term. We don’t have the observation inside the volcano. That’s the main reason. If we were able to measure everything—all the mechanical characteristics, the physical characteristics of the rocks, the magma chamber, and the shape of the pipe from the fissure to the surface—we should probably be able to make nice predictions.
Philippe Jousset: Also, it’s not one measurement technique that will allow us to understand what’s happening. It’s not that there’s an opening and then magma (imagined as a fluid) goes out. It’s more complex because the magma is full of crystals, gas, liquid, and water, whose behavior depends on the composition of the relative materials.
Gizmodo: Okay, so what is Jerk? How does the tool alleviate these challenges?
Beauducel: Presently, volcano stations require [manual operations] by humans to integrate, synthesize, and analyze all the observations, and then you are able to say, “Probably, we think an eruption will be coming.” So this is very specialized work made possible by volcanologists based on geophysics and geochemistry.
The Jerk system is innovative in that it’s fully automatic. It can be implemented using only one station, at a relatively long distance. At the Piton de la Fournaise, Jerk was installed around 8 kilometers [5 miles] from the volcano. The system detects the early fracturing of the rock during the migration of the magma to the surface.
On most volcanoes, the definition of an eruption is magma coming to the surface. And to come to the surface, the magma needs to crack the rock or make a new fissure—and the opening produces a “jerk.” The word used in the paper is magmatic intrusion.
Jousset: If there’s an intrusion, magma may reach the surface, which makes it, by definition, an eruption. But sometimes it gets stuck, because the jerk is not large enough.
Gizmodo: So the intrusion “jerks” the magma to the surface.
Beauducel: Yeah. The word “jerk” here refers to several things, but the first definition is mathematical; it’s a derivative of ground acceleration, as well as the brutal, horizontal movement of the surface.
Gizmodo: And you tested this system at Piton de la Fournaise for ten years. That’s a long time. How were you even able to go through with this and also make sure that everything remained consistent throughout the decade?
Beauducel: That’s a very interesting question. First of all, the team at Piton de la Fournaise is very small, like 15 people. But the observatory was installed about 50 years ago, so that’s almost 50 years of recorded data. I made software specially made to experiment with new signal processing and modeling of real-time data on volcanoes. So we went to the observatory in April of 2014, and it took only 10 days of fieldwork to make the system operational.
As for why it took 10 years to validate, the first reason was that we didn’t know what the “jerk” was. It wasn’t that we made a new theory and tried to apply it. We had the signal and then implemented it as an alert system. At the end, we had 24 eruptions to validate [with the system]. So that’s really the story; we took our time and discussed it a lot. We knew we were the only people looking at this kind of signal, so we weren’t in a rush. And actually, in 2023, the Piton de la Fournaise stopped erupting. This happens every 10 or 15 years for this volcano.
Jousset: Yeah, it was as if the volcano was happy to let us work, and once we wrapped up [our analysis], it was done.
Gizmodo: Philippe, do you remember how your contribution to this project began?
Jousset: I remember exactly where and when we discussed this. That was in Indonesia. At the time Jerk had been implemented for around 5 years, and Piton de la Fournaise had an eruption when we were in Indonesia, and Francois told me, “Oh look, a ‘jerk.’ There will be an eruption.” And I didn’t believe it. I told him, “Yes, it’s working, but what is it?” Francois said this sign was the channels of the seismometer recording the mass position and the velocity [of ground motion].
One of the projects I worked on while in France was at a location where a sudden collapse of material in a mine produced important “tilts” in broadband seismometers installed close to this site. And I saw that in this mass position on the volcano, there is something to discover with this signal.
Gizmodo: So Jerk is quite good at predicting eruptions, at least for Piton de la Fournaise. What’s up for your team next?
Beauducel: At Piton La Fournaise, the next challenge is to improve the Jerk system. Sometimes we have false positives from station maintenance or someone walking close. Eventually we want to have it fully automatic and distinguish false positives from normal alarms.
Jousset: We’d also like to measure these signals at other volcanoes, to see whether there are such signals elsewhere. Mount Etna, [where we’ll be deploying this next], is the best place to do this because it’s an active volcano; there are many eruptions with very small fractures that can’t be seen with conventional instruments [that Jerk might be able to see].
The challenge is that we’ll be deploying at or close to the surface. We’re going to dig holes as deep as possible to insulate the system from other noise sources. With multiple stations, we hope to locate where these tiny signals are happening and whether they’re different compared to Piton de la Fournaise.
