Fault based recurrence models and occurrence probabilities of large earthquakes (M6.0) in the Corinth Rift, Greece

C. Kourouklas; E. Papadimitriou; P. Bonatis; V. Karakostas

REVIEW 4 major objections 5 minor 6 references

Fault based recurrence models and occurrence probabilities of large earthquakes (M6.0) in the Corinth Rift, Greece

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read For the 13 major fault segments of the Corinth Rift, mean M≥6.0 earthquake recurrence times range from about 40 to 1,500 years, and renewal-model probabilities split the segments into three groups: most are early in their seismic cycle, a…

desk verdict Route-one recurrence modeling with a useful new Tr table; the headline three-group story rests on a single unvalidated 60% slip-rate factor. read the letter →

arxiv 2506.04382 v1 pith:3NVOPZKG submitted 2025-06-04 physics.geo-ph

C. Kourouklas , P. Bonatis , E. Papadimitriou , V. Karakostas This is my paper

classification physics.geo-ph

keywords earthquakerecurrenceCorinthRiftseismicmomentrateconservationBrownianPassageTimemodelPoissontime-dependenthazardfaultsegmentsM≥6.0earthquakes

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to give each major fault segment of the Corinth Rift its own mean recurrence time for M≥6.0 earthquakes and use it to forecast the next 10, 20, and 30 years. Because observed large earthquakes are too few per fault for direct statistics, the authors estimate $T_r$ from seismic moment rate conservation, then compare a time-independent Poisson model with a Brownian Passage Time renewal model. The result is a three-group split: most segments are early in their cycle and have renewal probabilities well below Poisson, three segments (Eliki, Perachora, Marathias) are nearly equal under both models, and Psathopyrgos is the only segment past its mean recurrence time, giving it the highest short-term occurrence probabilities in the rift. If correct, this provides a fault-specific, time-dependent ranking that could feed probabilistic seismic hazard analysis.

What carries the argument

The argument runs on three linked pieces. Equation (1) computes the mean recurrence time as $T_r = M_{o,\max}/(\mu L w V)$, the ratio of the seismic moment of the largest expected event to the moment accumulated by long-term slip on each fault segment; the paper takes the long-term slip rate $V$ as 60% of the geodetic rate from a published GPS-velocity model. The coefficient of variation $C_v$, estimated through error propagation, sets the aperiodicity of the Brownian Passage Time model, a renewal model in which accumulated strain performs a Brownian random walk toward rupture. Finally, the comparison of the BPT and exponential hazard functions, conditioned on the elapsed time $T_e$ since the last M≥6.0 event, produces the three-group classification.

What would settle it

Recompute the recurrence times using independently measured geological slip rates instead of the 60 percent geodetic factor, and check whether Psathopyrgos still has elapsed time greater than its mean recurrence time; if its true slip rate is smaller than the assumed value, $T_r$ grows, the elapsed time becomes smaller than $T_r$, and the paper's single late-cycle segment with the highest probability disappears.

Watch

Extended reading notes

Core claim

The central claim is that recurrence behaviour of M≥6.0 earthquakes in the Corinth Rift is not uniform across faults and can be classified into three groups using renewal-model probabilities. The southern segments Aigion, Xylokastro, Skinos, and Alepochori, together with the northern segments Kapareli, Delfoi, Makrygialos, and Sykia, are mostly early in their earthquake cycle, and their Brownian Passage Time hazard rates over the next 10, 20, and 30 years fall well below the constant Poisson rates. Eliki, Perachora, and Marathias show almost equal hazard rates under both models, so the two forecast styles nearly coincide there. Psathopyrgos is unique: its elapsed time since the last M≥6.0 earthquake exceeds its estimated mean recurrence time of about 119 years, so the renewal model gives substantially higher occurrence probabilities than the Poisson model, making it the most threatening segment in the study area.

Load-bearing premise

Every recurrence time in the paper scales inversely with the assumed long-term slip rate, and that rate is set as 60 percent of the published geodetic slip rate; if the true slip rate on a fault is not 60 percent of the geodetic value, the group assignments and probabilities change.

Editorial extensions

If this is right

Psathopyrgos would carry the highest 10-, 20-, and 30-year conditional probability of an M≥6.0 earthquake in the Corinth Rift, and time-dependent hazard models should treat it as the priority segment.
For Aigion, Xylokastro, Skinos, Alepochori, Kapareli, Delfoi, Makrygialos, and Sykia, Poisson-based forecasts would overstate the near-term probability relative to a renewal model, because the segments are still early in their cycles.
For Eliki, Perachora, and Marathias, time-dependent and time-independent forecasts nearly coincide over the next 30 years, so model choice matters little for those segments.
The moment-rate conservation approach extends to other active fault systems where individual earthquakes are too rare to estimate recurrence intervals directly from catalogs.
$T_r$ estimates between about 40 and 1,500 years provide a fault-segment baseline that can be plugged directly into long-term earthquake rupture forecast and probabilistic seismic hazard analysis.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The 60% slip-rate factor is the single largest lever on the results; a fault whose true long-term slip rate differs from 60% of its geodetic rate could move between the paper's three groups.
The grouping is sensitive to the catalog-completeness assumption starting in 1714, because an older M≥6.0 earthquake on a northern segment would change the elapsed time and could pull that segment out of the early-cycle group.
A direct test of the renewal probabilities would be to compute the likelihood of the observed 1995-2024 earthquake record under the BPT model versus the Poisson model for each segment; Psathopyrgos, with $T_e/T_r > 1$, is the strongest discriminator.
The same modeling chain could be applied to adjacent rift systems with similar GPS coverage, converting a static slip-rate catalog into time-dependent hazard curves.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The short version: this is a straightforward, route-one application of Field et al.'s (1999) moment-rate conservation to 13 Corinth Rift faults, with Tr values and Poisson/BPT probabilities. The new pieces are the updated Briole et al. (2021) GPS-derived slip rates, the explicit Tr estimates for northern faults like Makrygialos and Sykia that were not clearly listed in previous published work, and the three-group classification of fault behavior. The arithmetic is internally consistent and the error propagation follows Peruzza et al. (2010).

What the paper does well: it puts a single, reproducible parameter set on the table and computes a full suite of recurrence times and 10/20/30-year probabilities. That is genuinely useful for a regional PSHA practitioner who needs a fault-based input. The grouping into early-cycle, mid-cycle, and past-cycle segments is a clean way to see where renewal models matter.

The soft spot is the 60% slip-rate scaling factor. It is introduced in one sentence in Application & Results with no justification, no citation, and no uncertainty. Since Tr is inversely proportional to slip rate V, every Tr and every Te/Tr ratio scales with this constant. The three-group classification, and especially the claim that Psathopyrgos is the only segment past its mean recurrence time, is entirely controlled by that single factor. A sensitivity analysis around, say, 0.4–0.8, or a per-fault justification, would fix this. Without it, the headline grouping is not yet established.

Relatedly, the paper does not validate the calculated Tr against observed recurrence intervals or paleoseismic data, even for faults like Aigion and Eliki where such data likely exist. That is a feasible check and it is missing. The event-to-fault associations come from the Kourouklas (2022) thesis; that is an acceptable source, but the uncertainty in those assignments is not discussed and it propagates into the elapsed times.

The aperiodicity α is derived from the same σ that enters Tr, which is a standard trick, but it means the BPT probabilities restate the input assumptions rather than test them. The paper is aware of its own scope—it excludes Akrata cleanly—and it does not overreach beyond its data in most places. But the conclusions do reach further than the evidence: the three-group classification is a model output, not an empirical discovery.

Who this is for: anyone building a fault-based seismic hazard model for the Corinth Rift will want to see these numbers, but they should treat the grouping as conditional on the 0.6 factor. It deserves a serious referee, but the referee should demand a robustness check on the slip-rate scaling and a comparison against independent recurrence data before publication.

Referee Report

4 major / 5 minor

Summary. This conference paper estimates the mean recurrence time Tr of M≥6.0 earthquakes for 13 normal fault segments in the Corinth Rift using the seismic moment-rate conservation method of Field et al. (1999), with uncertainties propagated following Peruzza et al. (2010). It then applies exponential (Poisson) and Brownian Passage Time (BPT) renewal models to compute hazard functions and conditional occurrence probabilities for the next 10, 20, and 30 years from 1 January 2025. The central claims are that Tr ranges from about 40 to 1500 years, that the segments separate into three groups (early-cycle, near-Poisson, and late-cycle), and that Psathopyrgos is the only segment whose elapsed time exceeds its mean recurrence time, giving it the highest renewal-model probability.

Significance. If the Tr estimates and the three-group classification are reliable, the paper would provide a useful fault-specific input for time-dependent probabilistic seismic hazard analysis in a highly active region. The authors use a standard moment-rate conservation approach, apply formal error propagation, and are explicit that Akrata is excluded because no M≥6.0 event is associated with it. The paper is clearly written and the statistical framework is appropriate. However, the central numerical results are not yet established: the load-bearing slip-rate scaling factor is unjustified, the renewal-model aperiodicity is derived from the same input uncertainties rather than from observed recurrence variability, the three-group pattern is largely a mathematical consequence of Te/Tr and α, and the promised numerical occurrence probabilities are never tabulated. The paper would be significantly strengthened by a sensitivity analysis on the slip-rate factor, a comparison with observed recurrence intervals or independent paleoseismic slip rates, and a table of the actual quantitative probabilities.

major comments (4)

[Application & Results] The 60% scaling of the geodetic slip rates of Briole et al. (2021) is introduced with a single sentence and no derivation, citation, or uncertainty. Because Tr in Eq. (1) is inversely proportional to the long-term slip rate V, this factor directly controls every Tr, every Te/Tr ratio, the BPT-versus-Poisson hazard comparison, and the membership of each segment in the three groups, including the claim that Psathopyrgos is past its mean recurrence time. No robustness test is given (e.g., using 0.4, 0.8, 1.0 of the geodetic rate, or fault-specific fractions). Since the factor is arbitrary and could differ among faults, the reported group memberships and the unique status of Psathopyrgos are not yet established.
[Method] The aperiodicity α is computed as the coefficient of variation Cv = σ/Tr, where σ is obtained by formal error propagation from the same input uncertainties (ΔM and σV) used to compute Tr. Thus α does not measure the observed variability of recurrence intervals; it encodes a propagation assumption. The BPT probabilities therefore depend on an input-derived α, not on data. The paper should test the sensitivity of the hazard and probability estimates to α (e.g., α = 0.5, 0.7, 1.0) and, where possible, compare α against historical recurrence intervals for segments with multiple events (Aigion, Xylokastro, Perachora, Eliki).
[Application & Results] The three-group classification is presented as an empirical finding, but it follows almost algebraically from the estimated Te/Tr and α values: when Te/Tr << 1 with α around 0.6–0.7, the BPT hazard at the elapsed time is much lower than the Poisson hazard; when Te/Tr ≈ 1 the two hazards are nearly equal; and when Te/Tr > 1 the BPT hazard is higher. The paper should state this dependence explicitly and present the actual BPT and Poisson probabilities for the 10-, 20-, and 30-year windows so that the reader can judge how strongly the ranking depends on the uncertain Tr values.
[Conclusions] The abstract and conclusions compare 'renewal model probabilities' with Poisson probabilities, but the paper never reports the numerical occurrence probabilities for the next 10, 20, and 30 years. Figures 3 and 4 show hazard functions, not the probabilities defined by Eqs. (4) and (5). Without a table of these probabilities and their uncertainties, the headline claim that some segments have 'significantly lower' and Psathopyrgos has 'much higher' renewal-model probabilities is not quantitatively supported.

minor comments (5)

[Abstract] The abstract states that Tr ranges from 40 to 1500 years, but the text reports Tr = 35 years for the Akrata segment; the range should be stated consistently.
[Application & Results] The variables ΔM and σV are used in the text but are never explicitly defined; a short definition near Eq. (1) would improve readability.
[Application & Results] The paper would be easier to follow if the estimated Tr, σ, α, Te, and Te/Tr were summarized in a single table, especially since Figure 2a is the only quantitative summary of the input-dependent results.
[References] Some reference entries have inconsistent formatting (e.g., journal names and volume punctuation); a final editorial pass would be helpful.
[Conclusions] The final paragraph about stakeholder decisions and protective measures goes beyond what can be concluded from hazard-rate comparisons alone, since no seismic hazard or risk analysis is performed.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the derivation chain is self-contained; the 60% slip-rate factor and the alpha-from-error-propagation choice are correctness risks, not circular reductions.

full rationale

The paper computes Tr from Eq. (1) using fault dimensions, Mmax_obs, and long-term slip rates; none of these inputs are defined in terms of the target Tr or the three-group classification. The BPT probabilities in Eq. (5) use Tr, alpha, and Te, and the group labels are then read off from the resulting hazard comparisons, so the classification is a model output rather than an input used to fit the model. The reader's concern that alpha is 'derived from the uncertainty in Tr using the same input parameters' identifies a methodological choice (treating parameter-uncertainty propagation as the coefficient of variation of recurrence times), but it is not circular: the paper does not fit alpha to the hazard differences, and the early-cycle/low-hazard relation is a generic property of BPT rather than a fitted coincidence. The '60% of the geodetic slip rates' factor is a strong, unexplained assumption that directly scales every Tr and Te/Tr value; if unjustified, it undermines the accuracy of the absolute probabilities and possibly some group memberships. That is a correctness/falsifiability risk, not equivalence-by-construction. The reliance on Kourouklas (2022) for fault-to-earthquake associations and parameter uncertainties is a self-citation for input data, but the associations are externally checkable historical/paleoseismic assignments and are not invoked as an unverified uniqueness theorem; hence it does not make the derivation circular. Overall, no step in the paper reduces a prediction to its own input by definition.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim depends on multiple input parameters that are not justified or tabulated in the paper. The most significant is the 0.6 slip rate scaling factor, which directly controls the Tr values and the resulting group classification. The model also relies on standard earthquake recurrence assumptions (moment rate conservation, characteristic earthquakes) and on the completeness and association of the historical catalogue.

free parameters (4)

Slip rate scaling factor = 0.6
Applied to the geodetic slip rates of Briole et al. (2021) to obtain long-term fault slip rates. This factor is chosen without stated justification and linearly controls all Tr estimates.
Maximum observed magnitude Mmax_obs per fault = Not tabulated; taken from Kourouklas (2022)
Determines the maximum seismic moment in the recurrence time formula. Values are not listed in the paper.
Fault dimensions L and w = Not tabulated; from Ganas (2024) and Console et al. (2013)
Fault length and width enter the moment and moment rate calculations. Specific values are not provided in the paper.
Uncertainties DeltaM and sigma_V = Not tabulated; from Kourouklas (2022)
Used in error propagation to obtain sigma of Tr and hence aperiodicity alpha. The values are not reported.

assumptions (5)

domain assumption Seismic moment rate conservation (Field et al., 1999)
Assumes the rate of seismic moment release balances tectonic loading and that a characteristic earthquake releases the accumulated moment. This is the core model for Tr estimation.
domain assumption Nearly characteristic earthquake behavior
Each fault segment is assumed to produce Mmax_obs earthquakes that release the entire accumulated moment. This underpins equation (1).
domain assumption Catalog completeness since 1714 AD
The paper relies on the completeness judgement from Console et al. (2013, 2015) to restrict the event catalogue to 22 M≥6.0 earthquakes.
domain assumption Earthquake-to-fault association from Kourouklas (2022)
The assignment of each historical earthquake to a specific fault segment, which determines elapsed times, is taken from the thesis without re-evaluation.
domain assumption BPT distribution as a renewal model for recurrence times
The Brownian Passage Time model with aperiodicity alpha is assumed to describe the recurrence behavior of large earthquakes. This is a standard but unproven statistical choice.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Fault based recurrence models and occurrence probabilities of large earthquakes (M6.0) in the Corinth Rift, Greece." pith.science (2026). https://pith.science/paper/3NVOPZKG

@misc{pith2026250604382,
  author       = {Pith},
  title        = {Pith review of: Fault based recurrence models and occurrence probabilities of large earthquakes (M6.0) in the Corinth Rift, Greece},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3NVOPZKG}},
  note         = {Machine review of arXiv:2506.04382}
}

read the original abstract

The M6.0 earthquakes recurrence times, Tr, exhibit high variability from 40 to 1500 years, with the southern Corinth Rift fault segments reaching values up to 350 years and their antithetic ones ranging from 400 to 1500 years. The fault segments in the Corinth Rift can be divided into three groups, according to their recurrence behaviour, with some of them exhibiting significantly lower renewal model probabilities than the Poisson model, others showing nearly equal probabilities, and one segment where the renewal model probabilities are much higher than the Poisson one.

Figures

Figures reproduced from arXiv: 2506.04382 by the authors.

**Figure 1.** Geomorphological map of Corinth Rift, where the major normal fault segments are represented by red solid lines. Yellow stars depict the M6.0 earthquakes occurred in the study area since 1714. Although the number of large earthquakes (both historical and instrumental) is adequate for regional statistical studies, the number of recurrence intervals become limited when subdivided into subsets associated with specific … view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

6 extracted references · 6 canonical work pages

[1997]

Journal of Seismology 1, 131–150

The Ms = 6.2, June 15, 1995 Aigion earthquake (Greece): Evidence for low angle normal faulting in the Corinth rift. Journal of Seismology 1, 131–150. Doi:https://doi.org/10.1023/A:1009795618839. Briole, P., Ganas, A., Elias, P., Dimitrov, D.,

work page doi:10.1023/a:1009795618839 1995
[2002]

Pace, B., Visini, F., Peruzza, L., (2016)

A Brownian model for recurrent earthquakes, Bulletin of Seismological Society of America 92, 2233–2250. Pace, B., Visini, F., Peruzza, L., (2016). FiSH: MATLAB Tools to Turn Fault Data into Seismic -Hazard Models, Seismological Research Letters, 87, 375-386. Papazachos, B.C., Papazachou, C.C.,

work page 2016
[2013]

Doi:https://doi.org/10.1002/jgrb.50277

Renewal models and coseismic stres s transfer in the Corinth Gulf, Greece, fault system/ Journal of Geophysical Research, Solid Earth 118, 3655 –3673. Doi:https://doi.org/10.1002/jgrb.50277. Console, R., Carluccio, R., Papadimitriou, E., Karakostas, V.,

work page doi:10.1002/jgrb.50277
[2015]

Journal of Geophysical Research 120, 326–343

Synthetic earthquake catalogs simulating seismic activit y in the Corinth Gulf, Greece, fault system. Journal of Geophysical Research 120, 326–343. Doi:https://doi.org.10.1002/2014JB011765. Field, E.H., Jackson, D.D., Dolan, J.F.,

work page doi:10.1002/2014jb011765
[2022]

Determination and simulation of strong earthquakes’ recurrence times in Greece via the application of stochastic models: contribution on seismic hazard assessment. Ph.D. Thesis, Aristotle University of Thessaloniki, Thessalonik i, 329 p. Doi: http://dx.doi.org/10.12681/eadd/53270 Matthews, M.V., Ellsworth, W.L., Reasenberg, P.A.,

work page doi:10.12681/eadd/53270
[2024]

NOAFAULTS KMZ layer Version 6.0 (version 6.0) [Data set]. Zenodo. Doi: https://doi.org/10.5281/zenodo.13168947 Kourouklas, C.,

work page doi:10.5281/zenodo.13168947

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1997]

Journal of Seismology 1, 131–150

The Ms = 6.2, June 15, 1995 Aigion earthquake (Greece): Evidence for low angle normal faulting in the Corinth rift. Journal of Seismology 1, 131–150. Doi:https://doi.org/10.1023/A:1009795618839. Briole, P., Ganas, A., Elias, P., Dimitrov, D.,

work page doi:10.1023/a:1009795618839 1995

[2] [2002]

Pace, B., Visini, F., Peruzza, L., (2016)

A Brownian model for recurrent earthquakes, Bulletin of Seismological Society of America 92, 2233–2250. Pace, B., Visini, F., Peruzza, L., (2016). FiSH: MATLAB Tools to Turn Fault Data into Seismic -Hazard Models, Seismological Research Letters, 87, 375-386. Papazachos, B.C., Papazachou, C.C.,

work page 2016

[3] [2013]

Doi:https://doi.org/10.1002/jgrb.50277

Renewal models and coseismic stres s transfer in the Corinth Gulf, Greece, fault system/ Journal of Geophysical Research, Solid Earth 118, 3655 –3673. Doi:https://doi.org/10.1002/jgrb.50277. Console, R., Carluccio, R., Papadimitriou, E., Karakostas, V.,

work page doi:10.1002/jgrb.50277

[4] [2015]

Journal of Geophysical Research 120, 326–343

Synthetic earthquake catalogs simulating seismic activit y in the Corinth Gulf, Greece, fault system. Journal of Geophysical Research 120, 326–343. Doi:https://doi.org.10.1002/2014JB011765. Field, E.H., Jackson, D.D., Dolan, J.F.,

work page doi:10.1002/2014jb011765

[5] [2022]

Determination and simulation of strong earthquakes’ recurrence times in Greece via the application of stochastic models: contribution on seismic hazard assessment. Ph.D. Thesis, Aristotle University of Thessaloniki, Thessalonik i, 329 p. Doi: http://dx.doi.org/10.12681/eadd/53270 Matthews, M.V., Ellsworth, W.L., Reasenberg, P.A.,

work page doi:10.12681/eadd/53270

[6] [2024]

NOAFAULTS KMZ layer Version 6.0 (version 6.0) [Data set]. Zenodo. Doi: https://doi.org/10.5281/zenodo.13168947 Kourouklas, C.,

work page doi:10.5281/zenodo.13168947