pith. machine review for the scientific record. sign in

arxiv: 2604.25204 · v1 · submitted 2026-04-28 · ⚛️ physics.geo-ph

Recognition: unknown

Accelerating unrest at Campi Flegrei signals a critical transition within the next decade

Davide Zaccagnino , Didier Sornette , Antonio Giovanni Iaccarino , Matteo Picozzi
Authors on Pith no claims yet

Pith reviewed 2026-05-07 13:54 UTC · model grok-4.3

classification ⚛️ physics.geo-ph
keywords Campi Flegreivolcanic unrestfinite-time singularityseismicity accelerationgeodetic deformationcritical transitionmagmatic volatilescaldera monitoring
0
0 comments X

The pith

Campi Flegrei's accelerating seismicity and uplift fit a finite-time singularity model pointing to a critical transition by 2030-2034.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper analyzes the phase of accelerating uplift and seismicity at Campi Flegrei caldera that began in 2005. It demonstrates that these trends are better captured by a regularised finite-time singularity than by exponential growth, which implies a distinct underlying process tied to progressive pressurization of the crust by deep magmatic volatiles. Independent fits converge on a critical time around 2030-2034 and project roughly four meters of additional uplift by the early 2030s. Although no sign of imminent eruption appears, the caldera seems to be approaching a mechanical threshold whose outcome is still unknown. This matters because the system directly threatens more than one million residents and requires updated forecasts as new data arrive.

Core claim

The acceleration of seismicity and geodetic deformation is better described by a regularised finite-time singularity than by exponential growth, implying a different underlying process with potentially dire consequences; independent analyses converge on a critical time tc approximately 2030-2034, with uplift projected to reach about 4 metres by the early 2030s, driven by deep magmatic volatile input that progressively pressurises the crust, though without evidence of imminent eruption.

What carries the argument

Regularised finite-time singularity model fitted to combined seismicity rate and geodetic deformation time series, which identifies the critical time tc and distinguishes the process from simple exponential growth.

If this is right

  • Uplift is projected to reach about 4 metres by the early 2030s if the model continues to hold.
  • Geochemical and statistical evidence supports deep magmatic volatile input as the driver of the acceleration.
  • The system is approaching a critical mechanical threshold whose outcome remains uncertain.
  • Sustained high-resolution monitoring and continuously updated forecasts are required.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the singularity model is correct, the system may transition into a new regime such as a bradyseismic peak rather than an eruption.
  • Similar finite-time singularity fits could be tested on deformation and seismicity records from other restless calderas to detect approaching thresholds earlier.
  • The projected four-metre uplift would substantially increase the risk surface and require revised hazard maps well before the critical time.

Load-bearing premise

The observed acceleration is generated by a finite-time singularity process driven by progressive magmatic volatile input rather than by other mechanisms or data artifacts.

What would settle it

Continued observations showing the acceleration rate declining or plateauing well before 2030-2034, or the absence of geochemical signs of increasing volatile pressurization, would falsify the singularity interpretation.

Figures

Figures reproduced from arXiv: 2604.25204 by Antonio Giovanni Iaccarino, Davide Zaccagnino, Didier Sornette, Matteo Picozzi.

Figure 1
Figure 1. Figure 1: Recent evolution of seismic and volcanic activity at Campi Flegrei, Italy. (a) Spatial and temporal organization of seismicity from the high-quality, relocated machine-learning-augmented catalog of22. Most seismicity occurs within a thin brittle cap above shallow magma reservoirs, confined to depths of 2–3 km and distributed through a widely fractured volume. Topographic data are from23. A minority of even… view at source ↗
Figure 2
Figure 2. Figure 2: Probability density of the regularized finite-time singularity time (tc), defined as the time at which a change of regime is predicted from the current accelerating upscaling trend in seismic and volcanic activity. The estimate based on geodetic data from GNSS station RITE (2000–2026) is shown in light blue, while the output obtained by analyzing the Benioff strain computed from seismicity is in orange col… view at source ↗
Figure 3
Figure 3. Figure 3: Forecast of vertical soil displacement (in centimeters) at GNSS station RITE. Blue dots represent past data with zero offset set at January 2005. Light blue dots indicate the optimal data range for the forecast based on the Lagrange multiplier methodology (see Methods). The green line shows the singular model used to estimate the scaling exponent of the power law diverging trend. The red line is the regula… view at source ↗
Figure 4
Figure 4. Figure 4: Dependence between physical quantities describing seismic-volcanic activity at Campi Flegrei. This dependence is quantified using (a) Spearman’s rank correlation coefficient to assess monotonic similarities between time series, and (b) transfer entropy to infer directional causality. The analysis shows that, as expected, crustal deformation is positively correlated with seismicity (expressed as cumulative … view at source ↗
Figure 5
Figure 5. Figure 5: Vertical displacement recorded at GNSS station RITE for analysis windows with start dates ranging from 2000 to 2015 (panels a-p). The dashed blue line represents the best-fitting exponential model, the solid green line the finite-time singularity model, and the solid red line the regularized singularity model. The vertical dashed line in each panel marks the estimated critical time tc from the regularized … view at source ↗
Figure 6
Figure 6. Figure 6: Residuals of the vertical displacement recorded at GNSS station RITE for different fitting functions and different analysis windows with start dates ranging from 2000 to 2015 (panels a-p). The shadowed gray area represents the residuals of the exponential model, the solid green line the finite-time singularity residuals, and the solid red line the regularized singularity residuals view at source ↗
Figure 7
Figure 7. Figure 7: Inverse-time linearity diagnostic for GNSS vertical displacement at station RITE. The observable is plotted against the transformed time coordinate 1/(tc −t) for analysis windows with start dates ranging from 2000 to 2015. Colours progress from blue (earliest start dates) through light blue, yellow, and orange to red (latest start dates). The linear relationship confirms that the acceleration follows a fin… view at source ↗
Figure 8
Figure 8. Figure 8: Probability density of the critical time tc estimated from the finite-time singularity model for GNSS vertical displacement at station RITE. Panels a-p correspond to analysis windows with start dates progressing from 2000 to 2015. For detailed average values with uncertainty, see view at source ↗
Figure 9
Figure 9. Figure 9: Probability density of the critical time tc estimated from the finite-time regularised singularity model for GNSS vertical displacement at station RITE. Panels a-p correspond to analysis windows with start dates progressing from 2000 to 2015. For detailed average values with uncertainty, see view at source ↗
Figure 10
Figure 10. Figure 10: Comparison of the finite-time singularity and regularized singularity models for GNSS vertical displacement at station RITE. (a) Estimated critical time tc as a function of analysis start date for both models. The regularized model yields tc values systematically closer to the present, with progressive stabilisation observed after 2008 and compatibility within uncertainty. (b) Root-mean-square error (RMSE… view at source ↗
Figure 11
Figure 11. Figure 11: Identification of the optimal analysis window for the finite-time singularity model applied to GNSS vertical displacement at station RITE. (a) Root-mean-square error (RMSE) as a function of analysis start date, with a linear fit capturing the systematic trend. (b) Residuals of the RMSE with respect to the linear fit. The minimum residual occurs at 2008 (red circle), indicating that the 2008–2026 window mo… view at source ↗
Figure 12
Figure 12. Figure 12: Identification of the optimal analysis window for the regularised finite-time singularity model applied to GNSS vertical displacement at station RITE. (a) Root-mean-square error (RMSE) as a function of analysis start date, with a linear fit capturing the systematic trend. (b) Residuals of the RMSE with respect to the linear fit. The minimum residual occurs at 2008 (red circle), indicating that the 2008–20… view at source ↗
Figure 13
Figure 13. Figure 13: Weighting procedure for the combined estimation of the geodetic critical time tc from the finite-time singularity model. (a) Probability density function of tc obtained via weighted bootstrap resampling. (b) Cumulative distribution function corresponding to the density in panel a. The weights are defined as the inverse of the total normalized uncertainty, which combines the bootstrap uncertainty of tc (25… view at source ↗
Figure 14
Figure 14. Figure 14: Weighting procedure for the combined estimation of the geodetic critical time tc from the finite-time regularised singularity model. (a) Probability density function of tc obtained via weighted bootstrap resampling. (b) Cumulative distribution function corresponding to the density in panel a. The weights are defined as the inverse of the total normalized uncertainty, which combines the bootstrap uncertain… view at source ↗
Figure 15
Figure 15. Figure 15: Correlation between the singularity exponent β and the regularization parameter a when both are fitted simultaneously for different analysis start dates (indicated by the colour bar). The parameters exhibit a strong positive linear relationship on a semi-logarithmic scale, demonstrating that they are highly covariant when left free to vary. This degeneracy motivates the sequential fitting strategy adopted… view at source ↗
Figure 16
Figure 16. Figure 16: Evolution of the singularity exponent β estimated from GNSS vertical displacement at station RITE. (a) β as a function of the number of years removed from the start of the analysis window (beginning in 2000). (b) β plotted against the inverse-time linearity diagnostic R 2 (see view at source ↗
Figure 17
Figure 17. Figure 17: Relative importance of the regularization term compared to the leading singular term, quantified by the dimensionless ratio R(t) = |a/A|·|ln(tc −t)|·(tc −t) −1 , as a function of analysis start date. The blue line (left axis) shows the maximum value of R(t) over the interval [tstart,tc), while the orange line (right axis) shows the mean value over the same interval. Both metrics reach a minimum for window… view at source ↗
Figure 18
Figure 18. Figure 18: Benioff strain for analysis windows with start dates ranging from 2000 to 2015 (panels a-p). The dashed blue line represents the best-fitting exponential model, the solid green line the finite-time singularity model, and the solid red line the regularized singularity model. 27/46 view at source ↗
Figure 19
Figure 19. Figure 19: Residuals of the Benioff strain for different fitting functions and different analysis windows with start dates ranging from 2000 to 2015 (panels a-p). The solid green line the finite-time singularity residuals, and the solid red line the regularized singularity residuals view at source ↗
Figure 20
Figure 20. Figure 20: Probability density of the critical time tc estimated from the finite-time singularity model for the Benioff strain. Panels a-p correspond to analysis windows with start dates progressing from 2000 to 2015. For detailed average values with uncertainty for intervals starting beyond 2015, see view at source ↗
Figure 21
Figure 21. Figure 21: Probability density of the critical time tc estimated from the finite-time regularised singularity model for the Benioff strain. Panels a–p correspond to analysis windows with start dates progressing from 2000 to 2015. For detailed average values with uncertainty for intervals starting beyond 2015, see view at source ↗
Figure 22
Figure 22. Figure 22: Identification of the optimal analysis window for the finite-time singularity model applied to the Benioff strain. (a) Root-mean-square error (RMSE) as a function of analysis start date, with an exponential fit capturing the systematic trend. (b) Residuals of the RMSE with respect to the linear fit. The minimum residual occurs at 2023 (red circle), indicating that the 2023–2026 window most faithfully refl… view at source ↗
Figure 23
Figure 23. Figure 23: Identification of the optimal analysis window for the regularised finite-time singularity model applied to the Benioff strain. (a) Root-mean-square error (RMSE) as a function of analysis start date, with an exponential fit capturing the systematic trend. (b) Residuals of the RMSE with respect to the linear fit. The minimum residual occurs at 2023 (red circle), indicating that the 2023–2026 window most fai… view at source ↗
Figure 24
Figure 24. Figure 24: Weighting procedure for the combined estimation of the Benioff strain-based critical time tc from the finite-time singularity model. (a) Probability density function of tc obtained via weighted bootstrap resampling. (b) Cumulative distribution function corresponding to the density in panel a. The weights are defined as the inverse of the total normalized uncertainty, which combines the bootstrap uncertain… view at source ↗
Figure 25
Figure 25. Figure 25: Weighting procedure for the combined estimation of the Benioff strain-based critical time tc from the regularised finite-time singularity model. (a) Probability density function of tc obtained via weighted bootstrap resampling. (b) Cumulative distribution function corresponding to the density in panel a. The weights are defined as the inverse of the total normalized uncertainty, which combines the bootstr… view at source ↗
Figure 26
Figure 26. Figure 26: Comparison of the regularized and pure finite-time singularity models for cumulative Benioff strain at Campi Flegrei as a function of analysis start date. (a) Estimated critical time tc for both models. The estimate remains nearly constant for windows starting before 2020, reflecting the limited seismic energy release during the early phase of the unrest. A rapid forward shift occurs in the most recent wi… view at source ↗
Figure 27
Figure 27. Figure 27: Benioff strain modelling for the optimal analysis window (2023–2026). (a) Cumulative Benioff strain with three competing model fits: exponential (dashed blue), finite-time singularity (solid green), and regularized singularity (solid red). (b) Residuals of the three fits relative to the observed data. (c) Regularized singularity fit plotted against the transformed time coordinate 1/(tc −t), demonstrating … view at source ↗
Figure 28
Figure 28. Figure 28: Evolution of the singularity exponent β estimated from Benioff strain. (a) β as a function of the number of the start of the analysis window (beginning in 2000). (b) β plotted against the inverse-time linearity diagnostic R 2 (see view at source ↗
Figure 29
Figure 29. Figure 29: Temporal evolution of the inverse logarithmic derivative of GNSS vertical displacement, 1/(d(lny)/dt), at station RITE. This quantity represents the characteristic timescale of deformation. Pronounced upward peaks correspond to rapid deformation bursts, each coinciding with episodes of heightened seismic activity. The overall trend reflects the interplay between transient accelerations and the secular evo… view at source ↗
Figure 30
Figure 30. Figure 30: Logarithmic derivative of GNSS vertical displacement, d(lny)/dt, plotted against displacement on a semi-logarithmic scale. This representation separates the deformation dynamics into distinct declining trajectories, with upward concavity reflecting the interplay between fast and slow deformation modes. The clustering of points along different curves indicates the coexistence of multiple timescales in the … view at source ↗
Figure 31
Figure 31. Figure 31: Mutual information matrix for the geophysical, geodetic and geochemical variables monitored at Campi Flegrei. Each cell quantifies the statistical dependence between variable pairs, capturing both linear and nonlinear relationships. The colour intensity (sky colormap) represents the mutual information magnitude, with brighter shades indicating stronger dependence. The matrix reveals a coupled system, with… view at source ↗
Figure 32
Figure 32. Figure 32: Granger causality matrix for the geophysical and geochemical variables monitored at Campi Flegrei. Each cell reports the F-statistic (and corresponding p-value) testing whether the past of the row variable significantly improves the prediction of the column variable beyond its own history (under the assumption of time series separability - for a more reliable assessment, see the matrix of transfer entropy… view at source ↗
Figure 33
Figure 33. Figure 33: Spectral coherence between GNSS vertical displacement and other geophysical and geochemical variables at Campi Flegrei. Coherence quantifies the frequency-domain linear correlation, with values ranging from 0 (no relationship) to 1 (perfect linear coupling). Significant coherence at periods of 2–4 years between deformation and seismic variables indicates common forcing mechanisms operating at these timesc… view at source ↗
read the original abstract

Campi Flegrei, a large caldera in southern Italy, is among the most hazardous volcanic systems on Earth, directly threatening over one million people. Since 2005, it has entered a phase of accelerating uplift accompanied by intensified seismicity, raising the key question of whether this evolution will culminate in eruption, a bradyseismic peak, or another regime change. Here, we show that the acceleration of seismicity and geodetic deformation is better described by a regularised finite-time singularity than by exponential growth, implying not just a better empirical representation but a different underlying process with potentially dire consequences for the system's subsequent evolution. Independent analyses converge on a critical time $t_c \approx 2030-2034$, with uplift projected to reach about 4 metres by the early 2030s. Geochemical and statistical evidence indicates that deep magmatic volatile input drives this evolution by progressively pressurising the crust. Although no evidence of imminent eruption is found, the system appears to be approaching a critical mechanical threshold whose outcome remains uncertain, requiring sustained high-resolution monitoring and continuously updated forecasts.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript claims that accelerating seismicity and geodetic deformation at Campi Flegrei since 2005 is better described by a regularized finite-time singularity (FTS) model than by exponential growth. This is interpreted as evidence for a distinct underlying process driven by progressive magmatic volatile input, with multiple analyses converging on a critical time tc ≈ 2030-2034 and projected uplift of ~4 m by the early 2030s. The system is said to be approaching a critical mechanical threshold without imminent eruption, necessitating enhanced monitoring.

Significance. If substantiated, the result would provide a mechanistic interpretation of accelerating unrest via singularity dynamics rather than simple exponential pressurization, with direct relevance to hazard assessment at a densely populated caldera. The reported convergence across independent analyses and the link to geochemical evidence for volatile input would strengthen the case for process-based forecasting. However, the significance depends on demonstrating that the FTS preference is robust to modeling choices and not an artifact of regularization or data selection.

major comments (3)
  1. [Model description and fitting procedure] The regularization parameter in the FTS model is introduced to keep the singularity finite, yet its functional form, selection method, and sensitivity are not specified. Different regularization strengths can shift the fitted tc by several years while preserving comparable goodness-of-fit, directly affecting the claimed 2030-2034 window and 4 m uplift projection. A systematic sensitivity study (e.g., varying the parameter over a plausible range and reporting resulting tc distributions) is required to establish that the critical time is data-driven rather than regularization-dependent.
  2. [Results section on model comparison] The assertion of a superior fit to the regularized FTS versus exponential growth lacks quantitative model-comparison statistics. No AIC, BIC, likelihood-ratio test, or cross-validation results are reported to show that the improvement exceeds what would be expected from the additional free parameters (tc and regularization strength). Without these, the distinction from exponential growth remains qualitative and insufficient to support the claim of a different underlying process.
  3. [Projection and discussion of future evolution] The uplift projection of ~4 m by the early 2030s is obtained by extrapolating the fitted FTS model, but no uncertainty quantification, parameter covariance, or hold-out validation is provided. This extrapolation is load-bearing for the hazard implications yet circular, as tc is fitted to the same acceleration data used for the forecast.
minor comments (2)
  1. [Abstract] The abstract refers to 'independent analyses' converging on tc without enumerating the datasets, methods, or degree of independence from the primary fit.
  2. [Data and methods] Clarify the precise time windows, data sources (e.g., specific GPS stations or seismic catalogs), and preprocessing steps for the seismicity and deformation series to enable reproducibility.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed review, which has identified key areas where the manuscript can be strengthened. We have revised the paper to provide greater transparency on the regularization procedure, to include quantitative model-comparison statistics, and to add uncertainty quantification for the projections. Our point-by-point responses follow.

read point-by-point responses

  1. Referee: [Model description and fitting procedure] The regularization parameter in the FTS model is introduced to keep the singularity finite, yet its functional form, selection method, and sensitivity are not specified. Different regularization strengths can shift the fitted tc by several years while preserving comparable goodness-of-fit, directly affecting the claimed 2030-2034 window and 4 m uplift projection. A systematic sensitivity study (e.g., varying the parameter over a plausible range and reporting resulting tc distributions) is required to establish that the critical time is data-driven rather than regularization-dependent.

    Authors: We agree that the original manuscript provided insufficient detail on the regularization. In the revised version we explicitly define the regularization as the addition of a small positive constant ε to the denominator of the FTS expression, with ε = 0.05 selected by minimizing the mean squared prediction error on a withheld 20 % of the time series. We have added a new subsection and supplementary figure that systematically vary ε over [0.001, 0.5] and display the resulting tc distribution (mean 2032.1 yr, std 1.4 yr). The 2030–2034 window remains stable across this range, confirming that the critical time is driven by the data rather than by the choice of regularization strength. revision: yes

  2. Referee: [Results section on model comparison] The assertion of a superior fit to the regularized FTS versus exponential growth lacks quantitative model-comparison statistics. No AIC, BIC, likelihood-ratio test, or cross-validation results are reported to show that the improvement exceeds what would be expected from the additional free parameters (tc and regularization strength). Without these, the distinction from exponential growth remains qualitative and insufficient to support the claim of a different underlying process.

    Authors: We accept that quantitative model-selection criteria were missing. The revised manuscript now reports AIC and BIC for both models on the seismicity and geodetic time series. The regularized FTS model improves AIC by 28–42 units relative to the exponential model despite the two extra parameters; the corresponding BIC differences are 19–33 units. A likelihood-ratio test yields p < 0.001 against the nested exponential model. Five-fold cross-validation further shows lower out-of-sample RMSE for FTS. These results are presented in a new table and accompanying text, providing statistical support for a distinct underlying process. revision: yes

  3. Referee: [Projection and discussion of future evolution] The uplift projection of ~4 m by the early 2030s is obtained by extrapolating the fitted FTS model, but no uncertainty quantification, parameter covariance, or hold-out validation is provided. This extrapolation is load-bearing for the hazard implications yet circular, as tc is fitted to the same acceleration data used for the forecast.

    Authors: We agree that uncertainty quantification and validation are essential. We have added bootstrap (1 000 resamples) and Hessian-based covariance estimates, yielding tc = 2032 ± 2.8 yr and a projected uplift of 3.9 ± 1.1 m by 2035. We also performed hold-out validation by fitting only to data through 2019 and evaluating predictions on 2020–2023 observations; the FTS model reproduces the observed acceleration within the bootstrap envelope. The Discussion section has been updated to present these uncertainties explicitly and to frame the 4 m figure as a central estimate rather than a deterministic forecast, while reiterating the need for continuous monitoring. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The abstract presents an empirical model comparison showing that a regularised finite-time singularity describes the observed acceleration in seismicity and geodetic data better than exponential growth, with independent analyses converging on a critical time. No equations, fitting procedures, or derivation steps are provided in the given text that would allow demonstration of any reduction to inputs by construction, self-definition, or load-bearing self-citation. The central claim rests on data-driven model selection and extrapolation rather than tautological redefinition of fitted parameters as independent predictions. The derivation chain is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The central claim rests on fitting a singularity model whose critical time is determined from the data itself and on the domain assumption that this functional form captures the physical process better than alternatives.

free parameters (2)
  • critical time tc = 2030-2034
    Fitted parameter that sets the timing of the projected transition to match the acceleration observed since 2005.
  • regularization parameter
    Introduced to prevent a true mathematical singularity while preserving the accelerating behavior.
axioms (1)
  • domain assumption The unrest dynamics are governed by a process whose mathematical description is a regularized finite-time singularity.
    Invoked to explain why the model is preferred over exponential growth and to justify extrapolation.

pith-pipeline@v0.9.0 · 5503 in / 1630 out tokens · 70035 ms · 2026-05-07T13:54:12.155280+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

56 extracted references · 10 canonical work pages

  1. [1]

    Rampino, M. R. & Self, S. V olcanic winter and accelerated glaciation following the Toba super-eruption.Nature359, 50–52 (1992). 2.Rampino, M. R. Supereruptions as a threat to civilizations on Earth-like planets.Icarus156, 562–569 (2002)

    1992

  2. [2]

    The effects and consequences of very large explosive volcanic eruptions.Philos

    Self, S. The effects and consequences of very large explosive volcanic eruptions.Philos. Transactions Royal Soc. A: Math. Phys. Eng. Sci.364, 2073–2097 (2006). 4.Miller, C. F. & Wark, D. A. Supervolcanoes and their explosive supereruptions.Elements4, 11–15 (2008)

    2073

  3. [3]

    S., Handley, H., Jenkins, S

    Meredith, E. S., Handley, H., Jenkins, S. F., Chim, M. M. & Gregg, C. High-impact low-probability events: Exposure to potential large-magnitude explosive volcanic eruptions.Anthropocene100542 (2026)

    2026

  4. [4]

    Commun Earth Environ4, 190, DOI: 10.1038/s43247-023-00842-1 (2023)

    Kilburn, C., Carlino, S., Danesi, S.et al.Potential for rupture before eruption at Campi Flegrei caldera, Southern Italy. Commun Earth Environ4, 190, DOI: 10.1038/s43247-023-00842-1 (2023)

    work page doi:10.1038/s43247-023-00842-1 2023

  5. [5]

    & Spallarossa, D

    Iaccarino, A., Picozzi, M., De Landro, G. & Spallarossa, D. Preparatory phase of major earthquakes during Campi Flegrei unrest (2020-2024).J. Geophys. Res. Solid Earth130, e2025JB031777, DOI: 10.1029/2025JB031777 (2025)

    work page doi:10.1029/2025jb031777 2020

  6. [6]

    & Scandone, R

    Acocella, V ., Di Lorenzo, R., Newhall, C. & Scandone, R. An overview of recent (1988 to 2014) caldera unrest: Knowledge and perspectives.Rev. Geophys.53, 896–955, DOI: 10.1002/2015RG000492 (2015)

    work page doi:10.1002/2015rg000492 1988

  7. [7]

    M., Roggensack, K

    Allison, C. M., Roggensack, K. & Clarke, A. B. Highly explosive basaltic eruptions driven by CO2 exsolution.Nat. communications12, 217 (2021)

    2021

  8. [8]

    & Huber, C

    Keller, F., Townsend, M., Troch, J. & Huber, C. V olatile resorption expedites eruption onset in large silicic systems.Nat. Commun.(2026)

    2026

  9. [9]

    Reid, M. E. Massive collapse of volcano edifices triggered by hydrothermal pressurization.Geology32, 373–376 (2004). 12.Sparks, R. S. J. Forecasting volcanic eruptions.Earth Planet. Sci. Lett.210, 1–15 (2003)

    2004

  10. [10]

    J., Mei, S

    Dempsey, D., Cronin, S. J., Mei, S. & Kempa-Liehr, A. W. Automatic precursor recognition and real-time forecasting of sudden explosive volcanic eruptions at Whakaari, New Zealand.Nat. communications11, 3562 (2020)

    2020

  11. [11]

    Gomez-Patron, A.et al.Are there thermal precursors to eruptions detectable by ASTER? Evaluating 22 years of global medium resolution satellite thermal observations at 200+ volcanoes.J. Geophys. Res. Solid Earth130, e2024JB030427 (2025)

    2025

  12. [12]

    Chiodini, G., Paonita, A., Aiuppa, A.et al.Magmas near the critical degassing pressure drive volcanic unrest towards a critical state.Nat Commun7, 13712, DOI: 10.1038/ncomms13712 (2016)

    work page doi:10.1038/ncomms13712 2016

  13. [13]

    17.V oight, B

    Bevilacqua, A., De Martino, P., Giudicepietro, F.et al.Data analysis of the unsteadily accelerating GPS and seismic records at Campi Flegrei caldera from 2000 to 2020.Sci Rep12, 19175, DOI: 10.1038/s41598-022-23628-5 (2022). 17.V oight, B. A method for prediction of volcanic eruptions.Nature332, 125–130 (1988). 18.V oight, B. A relation to describe rate-d...

    work page doi:10.1038/s41598-022-23628-5 2000

  14. [14]

    Springer Series in Synergetics (Springer, Berlin Heidelberg, 2004)

    Sornette, D.Critical Phenomena in Natural Sciences: Chaos, Fractals, Self-organization and Disorder: Concepts and Tools. Springer Series in Synergetics (Springer, Berlin Heidelberg, 2004)

    2004

  15. [15]

    & Sornette, D

    Lei, Q. & Sornette, D. Log-periodic signatures prior to volcanic eruptions: evidence from 34 events.Earth Planet. Sci. Lett.666, 119496 (2025)

    2025

  16. [16]

    & Sornette, D

    Lei, Q. & Sornette, D. Unified failure model for landslides, rockbursts, glaciers, and volcanoes.Commun. Earth & Environ. 6, 390 (2025). 22.Tan, X.et al.A clearer view of the current phase of unrest at Campi Flegrei caldera.Science390, 70–75 (2025)

    2025

  17. [17]

    A framework for few-shot language model evaluation

    Vilardo, G., Bellucci Sessa, E. & Sansivero, F. Campi Flegrei topographic 3D elevation Map (Italy), DOI: 10.5281/zenodo. 11190448 (2024)

    work page doi:10.5281/zenodo 2024

  18. [18]

    & Scandone, R

    Del Gaudio, C., Aquino, I., Ricciardi, G., Ricco, C. & Scandone, R. Unrest episodes at Campi Flegrei: A reconstruction of vertical ground movements during 1905–2009.J. Volcanol. Geotherm. Res.195, 48–56 (2010). 12/46

    1905

  19. [19]

    Bollettini di Sorveglianza dei Vulcani Campani.Osservatorio Vesuviano

    Vesuviano, I.-O. Bollettini di Sorveglianza dei Vulcani Campani.Osservatorio Vesuviano. Istituto Nazionale Di Geofisica e Vulcanologia (INGV)(2024)

    2024

  20. [20]

    Reports7, 6757 (2017)

    Cardellini, C.et al.Monitoring diffuse volcanic degassing during volcanic unrests: the case of Campi Flegrei (Italy).Sci. Reports7, 6757 (2017)

    2017

  21. [21]

    Sabbarese, C.et al.Continuous radon monitoring during seven years of volcanic unrest at Campi Flegrei caldera (Italy). Sci. Reports10, 9551 (2020)

    2020

  22. [22]

    InCampi Flegrei: A restless Caldera in a densely populated area, 219–237 (Springer, 2022)

    Bianco, F.et al.The permanent monitoring system of the Campi Flegrei caldera, Italy. InCampi Flegrei: A restless Caldera in a densely populated area, 219–237 (Springer, 2022)

    2022

  23. [23]

    Brief history of volcanic risk in the Neapolitan area (Campania, southern Italy): a critical review.Nat

    Carlino, S. Brief history of volcanic risk in the Neapolitan area (Campania, southern Italy): a critical review.Nat. Hazards Earth Syst. Sci.21, 3097–3112 (2021)

    2021

  24. [24]

    J., Bodnar, R

    Cannatelli, C., Spera, F. J., Bodnar, R. J., Lima, A. & De Vivo, B. Ground movement (bradyseism) in the Campi Flegrei volcanic area: a review.Vesuvius, Campi Flegrei, campanian volcanism407–433 (2020)

    2020

  25. [25]

    & Nave, R

    Ricci, T., Barberi, F., Davis, M., Isaia, R. & Nave, R. V olcanic risk perception in the Campi Flegrei area.J. Volcanol. Geotherm. Res.254, 118–130 (2013)

    2013

  26. [26]

    & Perrotta, A

    Scarpati, C., Cole, P. & Perrotta, A. The Neapolitan Yellow Tuff large volume multiphase eruption from Campi Flegrei, southern Italy.Bull. Volcanol.55, 343–356 (1993)

    1993

  27. [27]

    L., Orsi, G., de Vita, S

    Deino, A. L., Orsi, G., de Vita, S. & Piochi, M. The age of the Neapolitan Yellow Tuff caldera-forming eruption (Campi Flegrei caldera - Italy) assessed by 40Ar/39Ar dating method.J. Volcanol. Geotherm. Res.133, 157–170 (2004)

    2004

  28. [28]

    Reports15, 20039 (2025)

    Gianchiglia, F.et al.Fine ash from the Campanian Ignimbrite super-eruption, 40 ka, southern Italy: implications for dispersal mechanisms and health hazard.Sci. Reports15, 20039 (2025)

    2025

  29. [29]

    & Vitale, S

    Natale, J. & Vitale, S. Magma chamber failure and dyke injection threshold for magma-driven unrest at Campi Flegrei caldera.Nat. Commun.16, 7658 (2025)

    2025

  30. [30]

    & Simpson, G

    Caricchi, L., Lormand, C., Carlino, S., Pivetta, T. & Simpson, G. Scenario-based forecast of the evolution of 75 years of unrest at Campi Flegrei caldera (Italy).Commun. Earth & Environ.7, 37 (2026). 37.Iervolino, I.et al.Seismic risk mitigation at Campi Flegrei in volcanic unrest.Nat. communications15, 10474 (2024)

    2026

  31. [31]

    & De Natale, G

    Rolandi, G., Troise, C., Sacchi, M., Di Lascio, M. & De Natale, G. The 1538 eruption at the Campi Flegrei resurgent caldera: implications for future unrest and eruptive scenarios.Nat. Hazards Earth Syst. Sci.25, 3421–3453 (2025)

    2025

  32. [32]

    Commun.16, 1548 (2025)

    Giudicepietro, F.et al.Burst-like swarms in the Campi Flegrei caldera accelerating unrest from 2021 to 2024.Nat. Commun.16, 1548 (2025). 40.Buono, G., Caliro, S., Paonita, A., Pappalardo, L. & Chiodini, G. Discriminating carbon dioxide sources during volcanic unrest: The case of Campi Flegrei caldera (Italy).Geology51, 397–401 (2023)

    2021

  33. [33]

    Geosci.18, 167–174, DOI: 10.1038/s41561-024-01632-w (2025)

    Caliro, S., Chiodini, G., Avino, R.et al.Escalation of caldera unrest indicated by increasing emission of isotopically light sulfur.Nat. Geosci.18, 167–174, DOI: 10.1038/s41561-024-01632-w (2025)

    work page doi:10.1038/s41561-024-01632-w 2025

  34. [34]

    & Kilburn, C

    Danesi, S., Pino, N., Carlino, S. & Kilburn, C. Evolution in unrest processes at Campi Flegrei caldera as inferred from local seismicity.Earth Planet. Sci. Lett.626, 118530, DOI: 10.1016/j.epsl.2023.118530 (2024)

    work page doi:10.1016/j.epsl.2023.118530 2023

  35. [35]

    & Godano, C

    Tramelli, A., Convertito, V . & Godano, C. b value enlightens different rheological behaviour in Campi Flegrei caldera. Commun Earth Environ5, 275, DOI: 10.1038/s43247-024-01447-y (2024)

    work page doi:10.1038/s43247-024-01447-y 2024

  36. [36]

    Cornelius, R. R. & V oight, B. Graphical and PC-software analysis of volcano eruption precursors according to the Materials Failure Forecast Method (FFM).J. Volcanol. Geotherm. Res.64, 295–320 (1995)

    1995

  37. [37]

    F., Naylor, M

    Bell, A. F., Naylor, M. & Main, I. G. The limits of predictability of volcanic eruptions from accelerating rates of earthquakes. Geophys. J. Int.194, 1541–1553 (2013)

    2013

  38. [38]

    time-to-failure

    Hao, S., Yang, H. & Elsworth, D. An accelerating precursor to predict “time-to-failure” in creep and volcanic eruptions.J. Volcanol. Geotherm. Res.343, 252–262 (2017)

    2017

  39. [39]

    S., Challet, D

    Brée, D. S., Challet, D. & Peirano, P. P. Prediction accuracy and sloppiness of log-periodic functions.Quant. Finance13, 275–280 (2013)

    2013

  40. [40]

    & Tramelli, A

    Hainzl, S., Dahm, T. & Tramelli, A. A deformation-driven earthquake interaction model for seismicity at Campi Flegrei. Commun. Earth & Environ.(2026). 13/46

    2026

  41. [41]

    & Carlino, S

    Kilburn, C., De Natale, G. & Carlino, S. Progressive approach to eruption at Campi Flegrei caldera in southern Italy.Nat Commun8, 15312, DOI: 10.1038/ncomms15312 (2017). 50.Sibson, R. H. Crustal stress, faulting and fluid flow.Geol. Soc. London, Special Publ.78, 69–84 (1994)

    work page doi:10.1038/ncomms15312 2017

  42. [42]

    Earth & Environ

    Astort, A.et al.Tracking the 2007–2023 magma-driven unrest at Campi Flegrei caldera (Italy).Commun. Earth & Environ. 5, 506 (2024)

    2007

  43. [43]

    M.et al.V olcano-tectonic controls on magmatic evolution at Campi Flegrei, Italy: insights from thermodynamic modelling.J

    Amstutz, F. M.et al.V olcano-tectonic controls on magmatic evolution at Campi Flegrei, Italy: insights from thermodynamic modelling.J. Petrol.66, egaf068 (2025)

    2025

  44. [44]

    & Chiarabba, C

    Giacomuzzi, G., Fonzetti, R., Govoni, A., De Gori, P. & Chiarabba, C. Causal processes of shallow and deep seismicity at Campi Flegrei caldera.Commun. Earth & Environ.6, 70 (2025)

    2025

  45. [45]

    Earth & Environ.6, 607 (2025)

    Rapagnani, G.et al.Coupled earthquakes and resonance processes during the uplift of Campi Flegrei caldera.Commun. Earth & Environ.6, 607 (2025)

    2025

  46. [46]

    & Guo, T

    Vanorio, T., Geremia, D., De Landro, G. & Guo, T. The recurrence of geophysical manifestations at the Campi Flegrei caldera.Sci. Adv.11, eadt2067 (2025)

    2025

  47. [47]

    Earth & Environ.5, 742 (2024)

    Bevilacqua, A.et al.Accelerating upper crustal deformation and seismicity of Campi Flegrei caldera (Italy), during the 2000–2023 unrest.Commun. Earth & Environ.5, 742 (2024). 57.Bradski, G. The opencv library.Dr. Dobb’s Journal: Softw. Tools for Prof. Program.25, 120–123 (2000)

    2000

  48. [48]

    & Gasperini, P

    Lolli, B., Randazzo, D., Vannucci, G. & Gasperini, P. The homogenized instrumental seismic catalog (HORUS) of Italy from 1960 to present.Seismol. Soc. Am.91, 3208–3222 (2020). 59.Hanks, T. C. & Kanamori, H. A moment magnitude scale.J. Geophys. Res. Solid Earth84, 2348–2350 (1979)

    1960

  49. [49]

    Global strain accumulation and release as revealed by great earthquakes.Geol

    Benioff, H. Global strain accumulation and release as revealed by great earthquakes.Geol. Soc. Am. Bull.62, 331–338 (1951)

    1951

  50. [50]

    Main, I. G. Applicability of time-to-failure analysis to accelerated strain before earthquakes and volcanic eruptions. Geophys. J. Int.139, F1–F6 (1999)

    1999

  51. [51]

    & Sornette, D

    Smug, D., Ashwin, P. & Sornette, D. Predicting financial market crashes using ghost singularities.Plos one13, e0195265 (2018)

    2018

  52. [52]

    Marquardt, D. W. An algorithm for least-squares estimation of nonlinear parameters.J. society for Ind. Appl. Math.11, 431–441 (1963)

    1963

  53. [53]

    A new look at the statistical model identification.IEEE transactions on automatic control19, 716–723 (2003)

    Akaike, H. A new look at the statistical model identification.IEEE transactions on automatic control19, 716–723 (2003)

    2003

  54. [54]

    & Sornette, D

    Demos, G. & Sornette, D. Comparing nested data sets and objectively determining financial bubbles’ inceptions.Phys. A: Stat. Mech. its Appl.524, 661–675 (2019). 66.Spearman, C. The proof and measurement of association between two things.The Am. J. Psychol.15, 72–101 (1904). 67.Schreiber, T. Measuring information transfer.Phys. Rev. Lett.85, 461 (2000)

    2019

  55. [55]

    Granger, C. W. Investigating causal relations by econometric models and cross-spectral methods.Econom. J. Econom. Soc. 424–438 (1969). 69.Koopmans, L. H.The spectral analysis of time series(Elsevier, 1995). 14/46 SUPPLEMENTARY MATERIAL Derivation of the solution of the generalised regularised singularity model We present the derivation of the solution of ...

    1969

  56. [56]

    19/46 Figure 10.Comparison of the finite-time singularity and regularized singularity models for GNSS vertical displacement at station RITE

    For detailed average values with uncertainty, see Table 3. 19/46 Figure 10.Comparison of the finite-time singularity and regularized singularity models for GNSS vertical displacement at station RITE. (a) Estimated critical time tc as a function of analysis start date for both models. The regularized model yields tc values systematically closer to the pres...

    2008