pith. sign in

arxiv: 2603.11108 · v2 · submitted 2026-03-11 · ⚛️ physics.geo-ph

Enhanced Seismicity Monitoring in the Rapid Scientific Response to the 2025 Santorini Crisis

Pith reviewed 2026-05-15 13:01 UTC · model grok-4.3

classification ⚛️ physics.geo-ph
keywords seismicity monitoringdeep learningSantorini crisisvolcanic-tectonic swarmsfluid-driven processesmoment tensor inversiontomographyearthquake catalogue
0
0 comments X

The pith

Deep learning expands Santorini earthquake catalogue from 4,000 to 80,000 events, revealing fluid-driven volcanic-tectonic swarms and a deep magmatic reservoir.

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

The paper applies a deep learning workflow to continuous seismic data to enhance earthquake detection during the 2025 Santorini-Amorgos crisis in near real-time. This increases the catalogue size dramatically, allowing the team to recognize the volcanic-tectonic nature through burst-like seismicity swarms associated with fluid processes from the beginning. Moment tensor analysis shows non-double couple components indicating magmatic or hydrothermal fluids, and tomography locates a third deep reservoir under Anydros Islet. Such patterns with many large events in short bursts have not been seen elsewhere in volcanic-tectonic swarms.

Core claim

By processing daily continuous waveforms with deep learning, the enhanced catalogue of 80,000 earthquakes demonstrates that the unrest involves fluid-driven processes, evident in spasmodic swarms and significant non-double couple moment tensors from early stages, alongside identification of a deep magmatic reservoir beneath Anydros Islet consistent with pressure-driven activity.

What carries the argument

The deep learning workflow for enhanced earthquake detection and location applied to continuous seismic waveforms, integrated with moment tensor inversions and DL-enhanced tomography to characterize the swarms and reservoirs.

If this is right

  • The crisis exhibits a unique pattern of volcanic-tectonic swarms with over 200 events of magnitude over 4 in a few weeks, mostly in episodic bursts.
  • Fluid involvement is indicated from the start, suggesting pressure-driven mechanisms rather than purely tectonic.
  • The third deep reservoir points to a more complex magmatic system than previously recognized.
  • Enhanced real-time monitoring can track the evolution of such crises more effectively.

Where Pith is reading between the lines

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

  • Similar deep learning enhancements could be deployed in other volcanic or seismic crises to detect fluid signals earlier.
  • This may imply that standard monitoring underestimates the scale and character of unrest in island arc settings.
  • Models of Santorini's magmatic plumbing could be updated to include this additional deep reservoir for better hazard assessment.

Load-bearing premise

The deep learning model maintains high precision in detecting true earthquakes without large numbers of false positives or significant location errors amid the noisy and rapidly changing crisis data.

What would settle it

Independent manual verification or comparison with other detection methods revealing a high rate of false positives among the additional detected events would undermine the interpretations of swarm patterns and fluid involvement.

read the original abstract

We used a deep learning workflow to enhance earthquake detection during the 2025 seismic unrest between Santorini and Amorgos islands to track the evolution of the crisis in near real-time. We analysed the continuous seismic waveforms daily (1/2 - 3/3/25) as the crisis unfolded. Our analysis enhanced the earthquake catalogue from around 4,000 to 80,000 earthquakes. The enhanced catalogue allowed this international expert group to identify the volcanic-tectonic character, clearly revealing burst-like, spasmodic seismicity swarms, which is a pattern associated with fluid-driven processes from early stages of the crisis. Detailed moment tensor inversions in early events characterised by a significant non-double couple component indicated the involvement of magmatic or high-pressure hydrothermal fluids driving the unrest. Concurrent DL-enhanced tomography efforts identified a third, deep magmatic reservoir beneath Anydros Islet, consistent with pressure-driven processes. To date, volcanic-tectonic swarms in which >200 earthquakes of ML > 4 occurred within only a few weeks, largely within episodic bursts of seismicity, have not been observed elsewhere.

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

2 major / 2 minor

Summary. The paper describes applying a deep learning workflow to continuous seismic waveforms recorded between 1 February and 3 March 2025 during the Santorini-Amorgos unrest. The workflow expands the earthquake catalogue from roughly 4,000 to 80,000 events, enabling identification of burst-like spasmodic swarms interpreted as fluid-driven, moment-tensor solutions with significant non-double-couple components, and tomographic imaging that reveals a third deep magmatic reservoir beneath Anydros Islet.

Significance. If the detections prove reliable, the work supplies an unusually dense record of a volcanic-tectonic crisis, directly linking episodic swarm patterns and non-double-couple sources to fluid overpressure and documenting a previously unrecognized deep reservoir. The rapid, near-real-time application during an active event offers a practical template for future crisis response and strengthens the empirical basis for fluid-migration models in such settings.

major comments (2)
  1. [Abstract] Abstract and Results: The central claim that the jump to 80,000 events reveals genuine burst-like spasmodic swarms and a third deep reservoir rests on the assumption that the DL detector maintains high precision in the noisy, rapidly evolving 1/2–3/3/25 dataset. No precision-recall figures, manual validation counts on a held-out subset, or false-positive rate estimates are supplied, leaving open the possibility that noise transients inflate the reported episodic bursts and bias the tomography.
  2. [Results] Results section: The interpretation of non-double-couple moment tensors and the deep reservoir as evidence for pressure-driven fluid processes is presented without quantitative uncertainty estimates on event locations or magnitudes for the added detections; this directly affects the load-bearing link between catalogue patterns and magmatic/hydrothermal drivers.
minor comments (2)
  1. [Abstract] The date range in the abstract should be written as 1 February–3 March 2025 for international clarity.
  2. [Figure captions] Figure captions should explicitly state the number of events used in each tomography inversion and the regularization parameters applied.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which have helped us improve the clarity and rigor of our rapid-response analysis. We address each major point below and have revised the manuscript to incorporate additional validation and uncertainty details.

read point-by-point responses
  1. Referee: [Abstract] Abstract and Results: The central claim that the jump to 80,000 events reveals genuine burst-like spasmodic swarms and a third deep reservoir rests on the assumption that the DL detector maintains high precision in the noisy, rapidly evolving 1/2–3/3/25 dataset. No precision-recall figures, manual validation counts on a held-out subset, or false-positive rate estimates are supplied, leaving open the possibility that noise transients inflate the reported episodic bursts and bias the tomography.

    Authors: We agree that explicit performance metrics are necessary to substantiate the catalogue expansion. The revised manuscript includes a new validation subsection reporting precision-recall curves evaluated on a held-out set of 500 manually reviewed events from the 1/2–3/3/25 period, together with false-positive rate estimates obtained by cross-validation against independent short-period network picks. These metrics confirm precision above 0.85 across the relevant magnitude range and support the robustness of the identified burst patterns and tomographic features. revision: yes

  2. Referee: [Results] Results section: The interpretation of non-double-couple moment tensors and the deep reservoir as evidence for pressure-driven fluid processes is presented without quantitative uncertainty estimates on event locations or magnitudes for the added detections; this directly affects the load-bearing link between catalogue patterns and magmatic/hydrothermal drivers.

    Authors: We accept that quantitative uncertainties were insufficiently reported. The revised results section now provides location and magnitude uncertainties for the DL detections, obtained from model confidence scores and standard relocation procedures. These uncertainties are propagated into the discussion of non-double-couple components and the tomographic imaging, thereby strengthening the quantitative link between the observed swarm patterns and fluid-driven processes. revision: yes

Circularity Check

0 steps flagged

No circularity: workflow outputs treated as independent observations

full rationale

The paper applies a deep-learning detection workflow to continuous waveforms (1/2–3/3/25) to produce an expanded catalogue (4k → 80k events), then reports observed swarm patterns and tomography features as direct consequences of that catalogue. No equations, fitted parameters, or self-citations are shown that would make the reported burst-like swarms or third reservoir equivalent to the input waveforms by construction. The derivation chain therefore remains non-circular; any concerns about false-positive rates belong to validation risk rather than circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The paper applies established deep-learning detection, moment-tensor inversion, and tomography methods to new crisis data without introducing new free parameters or entities beyond the interpreted reservoir; relies on standard geophysical assumptions about event detection and source mechanisms.

axioms (2)
  • domain assumption Deep-learning models trained on prior seismic data generalize to the 2025 Santorini waveforms without major domain shift.
    Invoked implicitly by the claim that the workflow enhanced the catalogue during the crisis.
  • domain assumption Non-double-couple moment-tensor components reliably indicate magmatic or hydrothermal fluid involvement.
    Used to interpret early events as fluid-driven.
invented entities (1)
  • Third deep magmatic reservoir beneath Anydros Islet no independent evidence
    purpose: Explains pressure-driven processes inferred from tomography and seismicity
    Identified via DL-enhanced tomography; no independent evidence such as geochemical or geodetic confirmation is provided in the abstract.

pith-pipeline@v0.9.0 · 5600 in / 1374 out tokens · 58084 ms · 2026-05-15T13:01:15.889488+00:00 · methodology

discussion (0)

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