Predicting a volcanic eruption in time to alert authorities and populations remains a major global challenge.
In a study published in the journal Nature Communications, an international team, involving scientists from CNRS Earth & Universe, proposes a new detection method, named "Jerk," capable of identifying in real-time very early precursor signals of volcanic eruptions using a single seismological instrument.
The "Jerk" method allows for the real-time detection of extremely subtle ground movements related to deep magma injections. These signals, called Jerk signals, manifest as very low-frequency transients observed in the horizontal movements of the ground, both in acceleration and tilt.
The authors show that they are likely generated by dynamic rock fracturing processes preceding an eruption. With an amplitude on the order of a few nanometers per second cubed (nm/s³), these signals can be detected using a single very broadband seismometer, through specific processing that includes the correction of Earth tides.
In April 2014, the tool was implemented at the Piton de la Fournaise Volcanological Observatory of IPGP (OVPF-IPGP, Reunion Island) as a fully automated module of the WebObs system, using data from a seismological station in the global Geoscope network located 5 miles (8 km) from the volcano's summit (Rivière de l'Est).
On June 20, 2014, a first alert was sent 1 hour and 2 minutes before the start of the eruption. For over 10 years, this system for detecting and analyzing Jerk signals has operated continuously 24/7, enabling automatic alerts for 92% of the 24 eruptions that occurred between 2014 and 2023.
Alert times varied from a few minutes to 8.5 hours before magma reached the surface. The method was also tested on data from 24 past eruptions between 1998 and 2010, showing that the Jerk alert works systematically.
The great originality of this work lies in the fact that the Jerk method was tested and validated in real-time automatically and unsupervised for over 10 years, unlike the vast majority of published studies on eruptive precursors, which are based on post-processing of data and a posteriori analysis.
However, the system has occasionally produced "false positives" — clear alerts not followed by an eruption — which have all turned out to be real magma intrusions or "aborted eruptions," an interpretation confirmed by all other observables such as seismicity, ground deformation, and volcanic gas analyses. Beyond the effectiveness of the Jerk alert for eruptions, the tool thus proves to be a perfect and unambiguous detector of magmatic intrusions.
During one of the recent seismic crises at Piton de la Fournaise on December 5, 2025, associated with minor deformations and gas anomalies, a small Jerk signal was emitted (only 0.1 nm/s
3), confirming that a magma intrusion had indeed occurred.
As Piton de la Fournaise is a highly instrumented and monitored laboratory volcano, the Jerk tool is used by OVPF-IPGP as a complementary indicator to the many precursor signs from other observables, allowing confirmation of the reality of a magmatic intrusion. On other less instrumented volcanoes, the Jerk tool could be used as a simple and effective method for early warning of volcanic eruptions.
Much remains to be done, particularly testing the method on other active volcanoes, starting with Mount Etna (Italy) where a project involving the GIPP (Geophysical Instrumental Pool of Potsdam) aiming to detect the Jerk signal with a new network of seismometers is set to begin in 2026, in collaboration with INGV (Italy).