pith. sign in

arxiv: 2403.00142 · v1 · submitted 2024-02-29 · 📊 stat.AP

A Fractional Model for Earthquakes

Louis Davis , Boris Baeumer , Ting Wang This is my paper

Pith reviewed 2026-05-24 03:20 UTC · model grok-4.3

classification 📊 stat.AP
keywords fractional Hawkes processMittag-Leffler kernelearthquake aftershocksETAS modelself-exciting point processseismic cyclemaximum likelihood estimation
0
0 comments X

The pith

A fractional Hawkes process with Mittag-Leffler decay kernel fits aftershock sequences better than the ETAS model on three Southern California catalogs.

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

The paper extends the fractional Hawkes process by using a Mittag-Leffler function for the temporal decay kernel to model mainshock-aftershock earthquake sequences. It develops a maximum likelihood estimation procedure and verifies the consistency of the resulting parameter estimates. When applied to three real earthquake sequences in Southern California, the fractional model outperforms the standard ETAS model. The authors further note that two of the fitted parameters may carry stable geophysical interpretations that depend on the study region and the stage of its seismic cycle. A sympathetic reader would care because the approach replaces an exponential decay assumption with a kernel that can capture different long-term scaling in triggering rates.

Core claim

The fractional Hawkes process is a self-exciting point process whose temporal decay kernel is the Mittag-Leffler function; when fitted by maximum likelihood to three Southern California earthquake catalogs it performs better than the ETAS model, and two of its parameters appear to take on fixed geophysical values that depend on the zone and the current stage of the seismic cycle.

What carries the argument

The Mittag-Leffler function serving as the temporal decay kernel inside a fractional Hawkes process, replacing the exponential kernel of the ETAS model to produce different memory properties in aftershock triggering.

If this is right

  • The model admits a consistent maximum likelihood estimator whose properties can be checked directly on data.
  • It yields lower fitting error than the ETAS model on the three examined Southern California sequences.
  • Two parameters may encode region-specific and cycle-stage-specific geophysical quantities that remain stable across events in the same zone.

Where Pith is reading between the lines

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

  • If the advantage persists on additional catalogs, the model could be used to generate improved short-term aftershock probability forecasts.
  • Treating the two parameters as fixed per region would allow a single calibration step to be reused for successive seismic events in that zone.
  • The same fractional-kernel construction could be tested on other self-exciting processes that exhibit slower-than-exponential decay, such as certain financial or neural spike trains.

Load-bearing premise

That the Mittag-Leffler kernel is a geophysically suitable replacement for the ETAS decay kernel and that any performance gain on the three tested catalogs will generalize.

What would settle it

A side-by-side maximum-likelihood comparison on an independent fourth earthquake catalog from a different region in which the ETAS model achieves equal or lower negative log-likelihood than the fractional model.

Figures

Figures reproduced from arXiv: 2403.00142 by Boris Baeumer, Louis Davis, Ting Wang.

Figure 1
Figure 1. Figure 1: Boxplots (A) λˆ 0, (B) ˆα, (C) ˆγ, (D) βˆ and (E) ˆc with true parameter values α = e −1 , λ0 = 1, β = 0.7, γ = 1.5 and c = 1. The true value is given by the solid black line. The synthetic data was generated by using the thinning method (e.g. Lewis and Shedler, 1979, Ogata, 1981). The magnitude of each event was simulated using i.i.d tapered Pareto random variables as suggested by the “PtProcess” R packag… view at source ↗
Figure 2
Figure 2. Figure 2: (A) Occurrence map of the studied catalogues. (B) Magnitude versus time plot for the [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Residual analysis. (A) corresponds to “JT”, (B) to “LM” and (C) “HM”. For all plots [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Log intensity for the ETAS (black) and fractional Hawkes process (red) on the “LM” [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Theoretical mean removed transformed time for the restricted fractional Hawkes pro [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Transformed time difference residual check for serial correlation. (A), (B) and (C) [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Graphical serial correlation check for the restricted fractional Hawkes process. (A), [PITH_FULL_IMAGE:figures/full_fig_p016_7.png] view at source ↗
read the original abstract

This paper extends the existing fractional Hawkes process to better model mainshock-aftershock sequences of earthquakes. The fractional Hawkes process is a self-exciting point process model with temporal decay kernel being a Mittag-Leffler function. A maximum likelihood estimation scheme is developed and its consistency is checked. It is then compared to the ETAS model on three earthquake sequences in Southern California. The fractional Hawkes process performs favourably against the ETAS model. Additionally, two parameters in the fractional Hawkes process may have a fixed geophysical meaning dependent on the study zone and the stage of the seismic cycle the zone is in.

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 / 1 minor

Summary. The manuscript extends the fractional Hawkes process by replacing the standard decay kernel with a Mittag-Leffler function to model mainshock-aftershock earthquake sequences. It develops a maximum-likelihood estimation procedure, states that consistency of the estimator has been verified, applies the model to three Southern California earthquake catalogs, and reports that the fractional model performs favourably relative to the ETAS benchmark. The paper further suggests that two of its parameters may admit a geophysical interpretation linked to the study region and stage of the seismic cycle.

Significance. A substantiated demonstration that a two-parameter Mittag-Leffler kernel yields statistically reliable improvements over ETAS on representative catalogs would constitute a useful parsimonious alternative for aftershock modeling and could motivate further investigation of fractional point-process kernels in seismology. The explicit mention of an MLE consistency check is a methodological strength that, if documented, would strengthen the work.

major comments (3)
  1. [Abstract] Abstract: the claim that the fractional Hawkes process 'performs favourably' against ETAS supplies no numerical evidence (likelihood ratios, AIC differences, out-of-sample scores, or error bars), leaving the central empirical result unsupported.
  2. [MLE consistency check] MLE section: consistency of the estimator is asserted to have been checked, yet the manuscript provides neither the verification procedure (simulation design, sample sizes, or convergence diagnostics) nor any quantitative results, which is load-bearing for trusting subsequent parameter estimates and model comparisons.
  3. [Empirical comparison] Comparison to ETAS: the evaluation is restricted to three sequences with no reported significance tests, cross-validation, or explicit control for the difference in parameter count (fractional model versus standard ETAS), so it is impossible to assess whether any observed gain exceeds sampling variability or arises from under-parameterization.
minor comments (1)
  1. [Abstract] The abstract should state the exact number of free parameters in the fractional model and whether any are held fixed during fitting.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major point below and indicate the revisions we will make to strengthen the work.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that the fractional Hawkes process 'performs favourably' against ETAS supplies no numerical evidence (likelihood ratios, AIC differences, out-of-sample scores, or error bars), leaving the central empirical result unsupported.

    Authors: We agree that the abstract lacks quantitative support for the performance claim. In the revised manuscript we will insert the key numerical results from the empirical section, specifically the log-likelihood ratios and AIC differences between the fractional model and ETAS on the three catalogs. revision: yes

  2. Referee: [MLE consistency check] MLE section: consistency of the estimator is asserted to have been checked, yet the manuscript provides neither the verification procedure (simulation design, sample sizes, or convergence diagnostics) nor any quantitative results, which is load-bearing for trusting subsequent parameter estimates and model comparisons.

    Authors: The consistency verification was performed via Monte Carlo simulations prior to submission, but the details were omitted. We will add a new subsection (or appendix) that fully documents the simulation design, sample sizes, convergence diagnostics, and quantitative results confirming estimator consistency. revision: yes

  3. Referee: [Empirical comparison] Comparison to ETAS: the evaluation is restricted to three sequences with no reported significance tests, cross-validation, or explicit control for the difference in parameter count (fractional model versus standard ETAS), so it is impossible to assess whether any observed gain exceeds sampling variability or arises from under-parameterization.

    Authors: The three sequences were selected as representative Southern California cases. We will revise the comparison section to explicitly discuss the difference in parameter count and to report bootstrap-based comparisons of predictive performance where feasible. With only three sequences, full cross-validation is limited, but we will add a limitations paragraph addressing sampling variability. revision: partial

Circularity Check

0 steps flagged

No significant circularity; derivation and comparison are externally grounded

full rationale

The paper extends an existing fractional Hawkes process, develops an MLE scheme whose consistency is checked, and reports performance against the independent ETAS model on three external Southern California catalogs. No equations, parameter fits, or self-citations are shown that reduce any claimed result to a tautology or to the same data by construction. The performance claim is therefore falsifiable against an external benchmark rather than self-referential.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The model rests on the domain assumption that a Mittag-Leffler kernel is suitable for earthquake triggering and introduces at least one free fractional-order parameter plus two additional parameters whose geophysical interpretation is asserted but not derived.

free parameters (2)
  • fractional order in Mittag-Leffler kernel
    Controls the shape of the temporal decay; must be estimated from data.
  • two parameters with claimed geophysical meaning
    Stated to depend on study zone and seismic-cycle stage; therefore fitted per catalog.
axioms (1)
  • domain assumption The Mittag-Leffler function supplies a more appropriate decay kernel for earthquake self-excitation than the kernels used in ETAS.
    Invoked when the fractional Hawkes process is positioned as an extension that performs better.

pith-pipeline@v0.9.0 · 5612 in / 1307 out tokens · 48479 ms · 2026-05-24T03:20:02.813940+00:00 · methodology

discussion (0)

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

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. A Multidimensional Fractional Hawkes Process for Multiple Earthquake Mainshock Aftershock Sequences

    stat.AP 2024-04 unverdicted novelty 7.0

    A novel multidimensional fractional Hawkes process with magnitude discretization and Mittag-Leffler kernels outperforms the ETAS model on information criteria, residual diagnostics, and retrospective prediction for Ja...

Reference graph

Works this paper leans on

25 extracted references · 25 canonical work pages · cited by 1 Pith paper

  1. [1]

    Adamopoulos, L. and A. G. Hawkes (1973). Cluster models for earthquakes-regional comparisons. Bull. Int. Stat. Inst. 45 (3), 454–461

    work page 1973

  2. [2]

    Akaike, H. (1974). A new look at the statistical model identification. IEEE transactions on automatic control 19 (6), 716–723

    work page 1974

  3. [3]

    Caputo, M. (1969). Elasticita de dissipazione, Zanichelli, Bologna, Italy,(links). SIAM journal on numerical analysis

    work page 1969

  4. [4]

    Caputo, M. (1976). Vibrations of an infinite plate with a frequency independent Q. The Journal of the Acoustical Society of America 60 (3), 634–639

    work page 1976

  5. [5]

    Hawkes, and E

    Chen, J., A. Hawkes, and E. Scalas (2021). A fractional Hawkes process. In Nonlocal and Fractional Operators, pp. 121–131. Springer

    work page 2021

  6. [6]

    Garrappa, R. (2015). Numerical evaluation of two and three parameter Mittag-Leffler functions. SIAM Journal on Numerical Analysis 53 (3), 1350–1369

    work page 2015

  7. [7]

    Garrappa, R. (2022). The Mittag-Leffler function

    work page 2022

  8. [8]

    Gemant, A. (1936). A method of analyzing experimental results obtained from elasto-viscous bodies. Physics 7 (8), 311–317

    work page 1936

  9. [9]

    Habyarimana, C., J. A. Aduda, E. Scalas, J. Chen, and A. G. Hawkes (2022). A fractional Hawkes process II: Further characterization of the process

    work page 2022

  10. [10]

    (2010, 07)

    Harte, D. (2010, 07). Ptprocess: An R Package for Modelling Marked Point Process Indexed by Time. Journal of Statistical Software 35

    work page 2010

  11. [11]

    Harte, D. (2013). Bias in fitting the ETAS model: a case study based on New Zealand seismicity. Geophysical Journal International 192 (1), 390–412

    work page 2013

  12. [12]

    Hawkes, A. G. (1971). Spectra of some self-exciting and mutually exciting point processes. Biometrika 58 (1), 83–90

    work page 1971

  13. [13]

    Langrock, R., D. L. Borchers, and H. J. Skaug (2013). Markov-modulated nonhomogeneous poisson processes for modeling detections in surveys of marine mammal abundance. Journal of the American Statistical Association 108 (503), 840–851. 14

    work page 2013

  14. [14]

    Lewis, P. W. and G. S. Shedler (1979). Simulation of nonhomogeneous Poisson processes by thinning. Naval research logistics quarterly 26 (3), 403–413

    work page 1979

  15. [15]

    Maechler, M. (2022). sfsmisc: Utilities from ’seminar fuer statistik’ eth zurich. R package version 1.1-13

    work page 2022

  16. [16]

    Marsan, D. and S. S. Nalbant (2005). Methods for measuring seismicity rate changes: a review and a study of how the M w 7.3 Landers earthquake affected the aftershock sequence of the M w 6.1 Joshua Tree earthquake. Pure and Applied Geophysics 162 (6), 1151–1185

    work page 2005

  17. [17]

    Metzler, R. and J. Klafter (2000). The random walk’s guide to anomalous diffusion: a fractional dynamics approach. Physics reports 339 (1), 1–77

    work page 2000

  18. [18]

    Metzler, R. and T. F. Nonnenmacher (2003). Fractional relaxation processes and fractional rheological models for the description of a class of viscoelastic materials. International Journal of Plasticity 19 (7), 941–959

    work page 2003

  19. [19]

    Nutting, P. (1921). A new general law of deformation. Journal of the Franklin Institute 191 (5), 679–685

    work page 1921

  20. [20]

    Ogata, Y. (1981). On Lewis’ simulation method for point processes. IEEE transactions on information theory 27 (1), 23–31

    work page 1981

  21. [21]

    Ogata, Y. (1988). Statistical models for earthquake occurrences and residual analysis for point processes. Journal of the American Statistical Association 83 (401), 9–27

    work page 1988

  22. [22]

    Ogata, Y., L. M. Jones, and S. Toda (2003). When and where the aftershock activity was de- pressed: Contrasting decay patterns of the proximate large earthquakes in southern California. Journal of Geophysical Research: Solid Earth 108 (B6)

    work page 2003

  23. [23]

    Papangelou, F. (1972). Integrability of expected increments of point processes and a related random change of scale. Transactions of the American Mathematical Society 165 , 483–506

    work page 1972

  24. [24]

    Reid, H. F. (1911). The elastic-rebound theory of earthquakes. Univ. Calif. Publ.. Bull. Dept. Geol. 6 (19), 413–444

    work page 1911

  25. [25]

    JT”, “LM

    Wang, T., M. Bebbington, and D. Harte (2012). Markov-modulated Hawkes process with stepwise decay. Annals of the Institute of Statistical Mathematics 64 (3), 521–544. 15 A Supplementary Figures 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.4 0.8 Uk Uk+1 (A) 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.4 0.8 Uk Uk+1 (B) 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.4 0.8 Uk Uk+1 (C) 0.0 0.2 0.4 0.6 0.8...

    work page 2012