An international team led by the University of Granada has presented a procedure that combines Artificial Intelligence and Signal Theory to anticipate volcanic eruptions with a minimum margin of twelve hours and confirm their end in just three. The approach is designed to reinforce monitoring systems and offer authorities a realistic reaction window.
The research, published in the journal Journal of Volcanology and Geothermal Research, has the participation of the University of Colima (Mexico), the center INVOLCAN (Tenerife) and the University of Canterbury (New Zealand). The goal is to consolidate a new generation of volcanic forecasting tools that are integrated into the operational monitoring.
How it works and what signals it analyzes
The method processes in real time. seismic records to detect patterns before and after eruptive activity. Its core combines three parameters: Shannon entropy (degree of signal disorder), frequency index and kurtosis, whose joint evolution allows the identification of physical changes in the magmatic system.
When entropy decreases steadily, the signals tend to to organize just before an eruptionThe frequency index captures shifts in dominant frequencies associated with different magmatic processes; and kurtosis is especially sensitive to impulsive events (explosions, fractures). This combined reading provides robustness against noise and natural variability.
AI is used to sift through large volumes of data, extract features, and update diagnoses without manual intervention, while Signal Theory It provides explainable metrics that ground the result in physical phenomena. The approach does not aim to replace classical instrumentation, but strengthen surveillance with quantifiable early indicators.
The tests carried out cover volcanoes of Spain, Mexico, Greece, Italy, the United States (Hawaii, Alaska and Oregon), Peru and Russiawith consistent results across different geological contexts and observation networks. This geographical scope suggests potential transfer to other environments in Europe, with particular interest for archipelagos and densely populated areas. In particular, studies have been included in Peru which confirm its applicability in Andean contexts.
Results in La Palma and Colima, and operating profit
During the eruption of Tajogaite (La Palma, 2021)The system advanced the start by more than nine hours and allowed for the identification of the cessation practically in real time thanks to a clear change in entropy. Having a minimum window of 12 hours would have made it easier to activate alerts and preparations with greater ease.
At the Colima Volcano of FireThe analysis of a decade of data (2013-2022) demonstrated that the technique distinguishes the start of intense eruptive phases, the growth of lava domes and the transitions to resting states, which provides context for regulating access, planning resources and assess hazard.
The team, under the direction of Jesus Ibanez (Andalusian Institute of Geophysics and Department of Theoretical Physics and Cosmology) and Carmen Benítez (Department of Signal Theory, Telematics and Communications), has also advised the authorities during the Santorini crisis (Greece), applying the approach to interpret the nature of seismicity.
Their incorporation into European monitoring systems would involve integrating the algorithms into existing seismic networks and harmonizing protocols with civil protection and define local warning thresholds. With a 12-hour warning, crucial time is gained to staggered evacuationsPreventive cuts and verified public communication.
Although the results are promising, more work is needed to calibrate the model. different types of volcanoes, improve interoperability with other signals (deformation, gases) and consolidate operability in monitoring centers, as well as integrate aspects related to the volcanic ashThe advance points to more proactive surveillance, with transparent metrics applicable in Spain and the rest of Europe.
With scientific verification, validations in La Palma and Colima, and the experience of institutional support in Greece, this AI-based methodology It is emerging as a tangible reinforcement for security: it alerts you in 12 hours, confirms the end in 3, and provides objective criteria for managing volcanic crises in a more orderly manner.
