Re-pressurized magma at Mt. Etna, Italy, may feed eruptions for years

The study addresses how to identify, locate, and quantify pressurized magma beneath active volcanoes to improve eruption forecasting, particularly regarding eruption duration and volume. Despite continuous activity, Mt. Etna exhibits phases of inflation and eruptive intensification. After a long steady period beginning in September 2019, activity sharply resumed from December 2020, culminating in paroxysmal lava fountains starting mid-February 2021. This reactivation coincided with GNSS-observed inflation, increased seismicity since summer 2020, and petrological evidence of primitive, gas-rich magma. The research leverages Mt. Etna’s extensive monitoring to image time-varying changes in seismic velocities and infer fresh melt accumulation, thus elucidating the volcano’s evolving magmatic system and its implications for ongoing and future eruptions.

Previous work has identified a persistent high-velocity body beneath Mt. Etna interpreted as a long-lived magma mush extending roughly from 3 to 15 km depth, incorporating intrusive counterparts of non-erupted lavas. Historical episodes show alternations between central explosive activity and flank-instability-dominated dynamics with shallow intrusions. Earlier tomographic and seismological studies documented pre-eruptive dyke intrusions, stress changes, and pressurization of crustal reservoirs, as well as the prognostic value of deep peripheral seismicity as a marker of recharging. Time-lapse tomography at Etna and other volcanoes has previously revealed localized velocity changes near intrusions, but broad, reservoir-scale fresh melt additions had remained elusive in geophysical imaging. This study builds on that foundation, employing repeated tomography to detect transient, spatially extensive low-velocity anomalies indicative of fresh melt accumulation at multiple depths.

The authors applied time-lapse (pseudo-4D) local earthquake tomography to quantify seismic velocity changes between a pre-2019 background model and the period December 2019–February 2021. They used the Simulps code (iterative damped least-squares) to invert for Vp and Vp/Vs in a 3D grid, progressively updating hypocentral and velocity parameters. The starting model was a comprehensive pre-2019 tomographic model derived from data spanning 2005–January 15, 2019. The dataset consists of 3140 local earthquakes recorded by the INGV-OE permanent broadband network (up to 40 stations within 40 km of the summit) from January 15, 2019 to February 2021. Event selection required at least 6 P and 2 S picks, RMS < 0.4 s, location errors < 2.0 km, azimuthal gap < 200° (98% < 180°), and a minimum epicentral distance of 10 km to the nearest station. In total, 36,619 P and 11,789 S arrivals were inverted. The grid used 2 km horizontal spacing; vertical layers 1 km apart from 2 km above sea level to 6 km depth, plus deeper layers at 9, 12, 15, 18, 25, and 30 km. Damping values were 80 (Vp) and 120 (Vp/Vs), chosen by trade-off analysis. After five iterations, the final RMS was 0.13 s with a variance reduction of 34%. To mitigate artefacts from differing ray paths across periods, a conservative two-step inversion (Foulger approach) was also applied: first invert the entire dataset with conservative damping to form a static 3D model, then invert the sub-period using the static model as reference with higher damping to isolate robust transient signals. Resolution and reliability were assessed via the full resolution matrix using the Spread Function (SF) and Derivative Weight Sum (DWS), retaining nodes with SF ≤ 3 in both starting and final models. Synthetic tests with an imposed transient low-Vp anomaly in the 2019–2021 period confirmed recoverability of such features, albeit with some underestimation of amplitude. Additional analyses of time-varying P and S residuals supported the need for a slower model in 2019–2021 and highlighted S-wave attenuation through central anomalies, explaining poorer Vs and Vp/Vs resolution in the core of the melt-rich body.

- Time-lapse tomography reveals three central reservoirs with transient reductions in P-wave velocity since 2019, indicating fresh melt accumulation: R-3 (deep), R-2 (intermediate-shallow), and R-1 (shallow). - R-3: a broad, elongated low-Vp volume centered around ~6 km below sea level (b.s.l.), extending approximately 3–9 km depth, located at the northern edge of the long-known high-velocity magma mush. Seismicity clusters around, but not within, this volume, consistent with re-pressurization. - R-2 and R-1: two smaller low-Vp anomalies beneath the summit craters at roughly ~1 km b.s.l., representing shallow reservoirs that likely feed lava fountain activity. Shallow seismicity exhibits clear on/off behavior around R-2. - Significant S-wave attenuation through the central low-Vp body leads to poor Vs and Vp/Vs resolution in that region, corroborating high melt and/or fluids. - Velocity residuals vs. time indicate the onset of fresh melt accumulation around mid-September 2019, with an acceleration in December 2020, coincident with increased deep (>12 km) and shallow (<3 km) seismicity and inflation. - Estimated melt fraction: a Vp perturbation of approximately −6% in R-3 implies ~4% added melt over ~2 years (using ~1.4% Vp decrease per 1% melt at mid-crustal depths and Etna-compatible compositions). This likely represents a lower bound due to the linearized inversion’s tendency to underestimate anomaly amplitudes. - The altered-velocity volume is large, suggesting substantial new melt injection; erupted lava since late 2020 constitutes only a small proportion of the accumulated melt. - The volcano’s behavior has shifted from flank-collapse-dominated dynamics of the past two decades to re-pressurization of the central system, implying the capacity to sustain eruptions for years.

The findings directly address the need to quantify where and how much melt accumulates beneath Mt. Etna to inform eruption forecasting. The detected transient low-Vp anomalies delineate a vertically connected system, with the principal storage in the mid-crustal reservoir R-3 and shallower reservoirs R-2 and R-1 feeding summit activity. A conceptual model emerges: deep crustal earthquakes (>12 km) mark initial magma ascent from the mantle source, followed by accumulation and pressurization within R-3. Seismicity then clusters around high-fluid-pressure compartments surrounding the aseismic, melt-rich core. Episodic transfer of melt from R-3 to R-2/R-1 is traced by rapid on/off shallow seismicity, leading to explosive summit activity and lava fountains. The magnitude of re-pressurization and added melt explains the resurgence of paroxysmal activity since mid-February 2021 and suggests that the central system now governs Etna’s dynamics more than flank instability. By quantifying a ~4% melt addition associated with a −6% Vp anomaly, the study provides volumetric constraints often missing in operational forecasts, enhancing the capacity to anticipate the duration and intensity of eruptive episodes.

Time-repeated local earthquake tomography at Mt. Etna provides near real-time imaging of transient, reservoir-scale velocity reductions that signify fresh melt accumulation since 2019 in three stacked reservoirs. The primary addition occurs in a deep, mid-crustal reservoir (R-3), with two smaller shallow reservoirs (R-2, R-1) feeding summit activity. The −6% Vp anomaly implies roughly 4% newly added melt over two years, and the size of the affected volumes indicates that recently erupted lava represents only a small fraction of the stored melt. The central magmatic system is re-pressurized, replacing prior flank-controlled dynamics, and could sustain eruptions for years. Future work should integrate ongoing time-lapse tomography with geodetic, petrological, and gas data to refine melt fraction estimates, improve Vp/Vs constraints despite S-wave attenuation, and track temporal evolution at finer time scales to support hazard forecasting.

- Time-lapse tomography can be affected by differences in ray sampling between periods; the study mitigates this with conservative inversion strategies and resolution analyses but residual bias cannot be entirely excluded. - Strong S-wave attenuation through the central melt-rich body leads to poor Vs and Vp/Vs resolution in the most anomalous region, limiting multiparametric interpretation. - Linearized inversions tend to underestimate anomaly amplitudes, implying that inferred melt fractions are likely lower bounds. - Spatial resolution is on the order of kilometers, so smaller-scale features and sharp boundaries may be smeared. - Results depend on the quality and distribution of local earthquakes and station coverage; temporal changes in seismicity can affect image comparability.