Pith. sign in

REVIEW 3 major objections 6 minor 45 references

Data-driven discovery of dynamo cycle equations

T0 review · 3 major / 6 minor · reviewed 2026-07-13 · grok-4.5

Pith's one-line read Data-driven equations for stellar dynamo cycles outperform classic weakly nonlinear analysis, even far from onset and on unseen unstable branches.

desk verdict Solid 1-D demo that constrained SINDy on Hankel-DMD amplitudes can outperform fifth-order WNL on subcritical and stiff dynamos, with real checks—but heavy filtering and a toy PDE keep the claim local. read the letter →

arxiv 2603.29079 v2 pith:A26HRXDU submitted 2026-03-30 astro-ph.SR

classification astro-ph.SR
keywords stellardynamomean-fieldDynamicModeDecompositionSINDyweaklynonlinearanalysissubcriticalbifurcationreduced-ordermodelα-Ω
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

Stars like the Sun generate oscillating magnetic fields that drive activity cycles and space weather, yet full magnetohydrodynamic simulations are too expensive and multi-scale to be practical for many purposes. Reduced-order models that capture the essential cycle dynamics are therefore valuable. This paper shows that a combination of Hankel Dynamic Mode Decomposition (to extract the coherent magnetic wave) and Sparse Identification of Nonlinear Dynamics (to fit a sparse polynomial ODE to its amplitude) recovers the governing amplitude equations for a canonical one-dimensional mean-field dynamo directly from simulation snapshots. The recovered models are compared systematically with the classic weakly nonlinear normal forms obtained by asymptotic expansion about the onset of dynamo instability. The data-driven equations remain accurate far beyond the weakly nonlinear regime, correctly predict saturation amplitudes when the underlying nonlinearities are stiff or non-analytic, and even reconstruct unstable subcritical solution branches that never appear in the training data. The result is a practical route to cycle models that can be extrapolated in dynamo strength and magnetic quenching parameters without requiring an analytic expansion.

What carries the argument

Hankel Dynamic Mode Decomposition followed by adjoint projection and Savitzky–Golay filtering yields a clean complex amplitude time series for the primary dynamo wave; constrained SINDy then identifies a sparse polynomial ODE for that amplitude as a function of the dynamo number D and the magnetic-dissipation parameter κ.

What would settle it

Apply the same DMD–SINDy pipeline to a dynamo whose nonlinearity is strongly non-analytic (large κ2) or to a three-dimensional convective dynamo, and check whether the recovered equations still reproduce saturation amplitudes and the correct supercritical-to-subcritical transition when compared with direct numerical solutions outside the training window.

Watch

Extended reading notes

Core claim

Sparse polynomial models fitted by SINDy to the leading Hankel-DMD modal coefficients of a one-dimensional mean-field dynamo are more robust than the corresponding weakly nonlinear normal forms: they correctly predict magnetic saturation amplitudes far from onset, recover both stable and unstable subcritical branches, and remain usable when the nonlinearity is non-analytic so that weakly nonlinear analysis cannot be applied.

Load-bearing premise

The leading pair of filtered Hankel-DMD modes is assumed to isolate the same slow amplitude that weakly nonlinear theory expands, so that a sparse polynomial fit on those two coefficients is a legitimate reduced model of the full system rather than an over-fit to the filtered signal.

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.

Referee Report

3 major / 6 minor

Summary. The paper develops a data-driven pipeline (Hankel DMD with adjoint modal coefficients, followed by SINDy) to recover reduced-order amplitude equations for oscillatory dynamos from a 1-D mean-field α–Ω model with magnetic-diffusion quenching. The recovered models are compared systematically to weakly nonlinear (WNL) normal forms at third and fifth order. For fixed (D,κ), for D-dependent families, and for a joint (D,κ) model, the authors report that SINDy equations integrate more accurately against DMD trajectories than WNL forms away from onset, recover unstable subcritical branches (cross-checked by Newton–Krylov), and remain usable when the quenching nonlinearity is stiff. Explicit coefficient tables, Pareto plots, bagging/constrained-SR3 tests, and bifurcation diagrams (Figs. 3–5, Tables 1–2) support the comparisons.

Significance. If the claims hold, the work offers a practical route to reduced dynamo-cycle models that do not require a full WNL expansion and that can be trained on simulation (or, in principle, observational) time series. Strengths that raise the contribution above a pure methods demonstration include: (i) side-by-side coefficient and error comparisons with analytically derived WNL forms; (ii) Newton–Krylov confirmation of unstable branches never present in the training trajectories; (iii) an adjoint-based extraction of DMD coefficients that improves isolation of the leading pair; and (iv) explicit tests of robustness (bagging, constrained SR3, polar vs Cartesian libraries). These make the paper a useful benchmark for data-driven reduced modeling of stellar cycles, even if the present demonstration remains on a 1-D mean-field system.

major comments (3)
  1. [Abstract; §7] Abstract and §7 claim that SINDy finds equations 'where the nonlinearity is not analytic and WNL analysis cannot be applied.' In §7 the authors themselves choose κ₂=5·10⁻³ so that |B|<1/√κ₂∼14 remains empirically satisfied, and they defer 'stiff subcriticalities' to future work. The abstract claim should be narrowed to what is actually demonstrated (better extrapolation within a regime where the Taylor expansion is still marginally valid), or an additional experiment at larger κ₂ (where WNL fails) should be added.
  2. [§3.1–4.1; §6.1–6.2] The central identification pipeline (§3.1–4.1, 6.1–6.2) relies on Savitzky–Golay filtering (windows 200–300), radius/κ normalization, and deliberate library restrictions that omit κ² and κD products. Without these steps the authors report overfitting or non-convergence (polar libraries; unrestricted fifth-order libraries). The manuscript should quantify how much of the reported superiority over WNL survives under weaker preprocessing (e.g., no SG filter, or only Hankel delay), and state more clearly that the recovered ODEs are models of the filtered two-mode signal whose fidelity to the full PDE is validated mainly near onset by Newton–Krylov and by integration error on the same filtered coefficients.
  3. [§4.2; Table 2; Eq. (24)] For the subcritical case at fixed D (§4.2, Table 2, Eq. 24), SINDy systematically prefers a combination of r⁴ and r⁵ damping over the pure fifth-order term of the WNL form (22), even under bagging and constrained SR3. The paper treats this as acceptable because integration error remains <1%. That is fine for local prediction, but it weakens the claim that SINDy recovers the 'correct' normal-form structure. Either reframe the claim as structure-agnostic predictive accuracy, or show that the spurious r⁴ term vanishes under a cleaner isolation of H (e.g., exact adjoint-eigenvector projection without DMD).
minor comments (6)
  1. [§5.2; Eq. (28)] Eq. (28) contains the placeholder '[recalculate!]' in the manuscript text; this must be removed and the numerical bounds finalized before publication.
  2. [Figure 1] Figure 1 caption refers to panel (a) as 'y=5/x' and then describes (a)–(d) inconsistently with the body text; renumber and clarify which panels show A, B, eigenvectors, and the DMD spectrum.
  3. [§5.1] In §5.1 the Cartesian model is referred to once as (27) when the displayed equations are (26); check cross-references for (25)–(31) throughout §5–6.
  4. [Abstract; §1; §3.1] The abstract and introduction mention Hankel DMD, but the title and some early paragraphs say only 'DMD'; keep the terminology consistent (Hankel/HODMD vs standard DMD).
  5. [Table 1; Appendices A–B] Table 1 and the WNL appendices are valuable; a short statement of how the adjoint eigenproblem and solvability integrals were discretized in Dedalus would aid reproducibility.
  6. [§2.1; §4.2; Fig. 5] Typographical: 'a a grid-based resolution' (§2.1); 'mildly supercritical regime' used for a subcritical κ case (§4.2); 'intencity' in Fig. 5 caption.

Circularity Check

2 steps flagged · score 3.0 of 10

Library deliberately constrained toward known WNL normal forms and heavy pre-filtering of DMD coefficients introduce mild circularity, but coefficients are fit to independent trajectories and unstable branches are recovered and externally confirmed.

  1. fitted input called prediction [§6.1–6.2, Eqs. (29)–(31); also §4.1 steps 5–6 and §3.3]
    "To remove spurious terms and get a sparser model, we restrict the library only to linear corrections in κ, at the expense of missing the κ^{2} contribution to the coefficient of the fifth-order polynomial. ... As previously, the aim is to obtain a model as close as possible to the normal form and avoid spurious terms, so we restrict parameter dependence of the library to multiples of D and κ with powers of x and y, omitting terms like κ^{2} and κD."

    The authors already possess the fifth-order WNL normal form (Eqs. 7–11, Table 1) and deliberately restrict the SINDy library (and apply radius/κ normalization plus Savitzky–Golay filtering) so that the regression recovers a structure close to that form. Unrestricted libraries produce non-sparse models with large spurious coefficients. The resulting 'discovered' equations are therefore partly steered by knowledge of the target rather than being a fully blind sparse identification; the numerical coefficients themselves remain data-driven, but the equation skeleton is not.

  2. other [§4.1 (data preparation) and §3.1–3.2 (adjoint Hankel DMD)]
    "To remove the remaining contribution of nonlinear interactions to the coefficient, the coefficients are filtered with a Savitsky–Golay filter with a typical window length of 200–300 points and third order polynomials. ... This procedure leads to smoother finite difference derivatives that enables a more robust sparse regression. The radii r are normalized with their steady-state absolute values, r_norm."

    Heavy pre-processing (high-delay Hankel DMD, adjoint projection, Savitzky–Golay filtering of quadratic oscillations, radius normalization) is required before SINDy converges to a WNL-like model. Without it the regression overfits or inserts spurious even-order terms (explicitly noted for polar libraries and for subcritical data). The load-bearing claim that the recovered ODE is a reduced-order model of the full PDE therefore rests on the assumption that the filtered two-mode signal still isolates the same slow amplitude H that WNL expands; that assumption is not independently verified far from onset or for stiff κ.

full rationale

The paper's central claim is that SINDy on Hankel-DMD coefficients yields reduced models more robust than WNL normal forms, including far-from-onset amplitudes, stiff non-analytic nonlinearities, and unstable subcritical branches never present in the training data. That claim is not circular by construction: the coefficients of the sparse models are obtained by regression on independent simulation trajectories (multiple initial amplitudes, ranges of D and κ), the models are integrated and compared to held-out DMD amplitudes, and the unstable branches recovered by the SINDy normal form are later confirmed by independent Newton–Krylov continuation of the full PDE. The mild circularity that remains is methodological rather than definitional. The authors already know the target normal forms from their own WNL calculation (Appendices A–B, Table 1) and explicitly constrain the SINDy library (omit κ^{2} and κD products, force Cartesian rather than polar form, normalize radius and κ) so that the regression recovers a structure close to those forms; without those constraints the unrestricted fifth-order library overfits or fails to converge. Likewise, Savitzky–Golay filtering and high-delay Hankel DMD are required to suppress quadratic oscillations that would otherwise produce spurious terms. These steps make the recovered equations closer to a description of a pre-processed two-mode signal than a fully blind discovery of the PDE reduction. That is a real but limited circularity (score 3): the library is steered toward a known answer, yet the numerical coefficients, the extrapolation, and the recovery of unseen unstable branches remain independent of the WNL calculation and are checked against external data. No self-definitional identity, no fitted-input-called-prediction of a held-out quantity that is forced by the fit, and no load-bearing uniqueness theorem imported from the authors' prior work appear in the derivation chain.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim rests on a standard 1-D mean-field dynamo PDE (Bushby 2003 plus κ-quenching), classical WNL expansions, and the empirical success of Hankel-DMD + constrained SINDy after several hand-tuned preprocessing steps. No new physical entities are postulated; free parameters are the usual dynamo coefficients plus the algorithmic knobs of the discovery pipeline.

free parameters (6)
  • κ1, κ2 (magnetic quenching)
    Control the strength and form of nonlinear magnetic diffusion; chosen by hand to place the system in supercritical or subcritical regimes (κ2 fixed at 0.005).
  • Cα, CΩ, Cη, C1–4, Cuη
    Constant coefficients of the mean-field model, taken from Bushby (2003) or set for numerical convenience; they fix the linear operator and the quadratic nonlinearities.
  • Hankel delay d and rank r
    Ad-hoc DMD hyperparameters (d≈12–20, r=5 or 9) tuned to suppress quadratic oscillations in the modal coefficients.
  • Savitzky–Golay window and polynomial order
    Filtering parameters (window 200–300, order 3) chosen to remove residual oscillations before SINDy; applied once or twice depending on subcriticality.
  • SINDy threshold Θcr and library constraints
    Sparsity threshold and restriction to linear-in-κ / linear-in-D terms are selected via Pareto plots and prior knowledge of the WNL form; they directly determine which terms survive.
  • Normalization radii r_norm and κ_norm
    Rescale modal amplitudes and κ so that library terms are O(1); values taken from particular simulation points (e.g., D=265).
assumptions (4)
  • domain assumption The 1-D mean-field α–Ω system (Eq. 1) with the stated quadratic and quenching nonlinearities is a faithful reduced description of the essential dynamo cycle dynamics for the purpose of method comparison.
    Invoked throughout; the entire study is performed on this PDE rather than on 3-D MHD.
  • ad hoc to paper Hankel-DMD modes of rank 5–9 plus adjoint projection isolate the same slow amplitude that WNL expands about the Hopf bifurcation.
    §§3.1–3.2; required for the SINDy library to be comparable to the analytic normal form.
  • domain assumption A sparse polynomial library (up to fifth order, optionally tensorized with D and κ) is rich enough to capture the saturation mechanisms of interest.
    Standard SINDy modeling choice; justified by the known structure of the WNL normal form but not proved for the stiff non-analytic case.
  • standard math Weakly nonlinear expansions to fifth order (Appendices A–B) correctly give the local normal form of the Hopf bifurcation for this PDE when the nonlinearity is analytic.
    Classical multiple-scale analysis; used as the ground-truth baseline.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Data-driven discovery of dynamo cycle equations." pith.science (2026). https://pith.science/paper/A26HRXDU

@misc{pith2026260329079,
  author       = {Pith},
  title        = {Pith review of: Data-driven discovery of dynamo cycle equations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/A26HRXDU}},
  note         = {Machine review of arXiv:2603.29079}
}
abstract

Many low-mass stars like the Sun host periodic, oscillatory magnetic fields that lead to variable levels of stellar activity, driving space weather that affects the habitability and detection of exoplanets. Owing to the intrinsic difficulty in modeling stellar magnetohydrodynamics across scales, realistic numerical simulations of this process are very challenging, and developing reduced-order models is of interest. In this work, we develop a framework to recover such models directly from numerical data by using a combination of Dynamic Mode Decomposition (DMD) to identify coherent magnetic structures, and the Sparse Identification of Nonlinear Dynamics (SINDy) framework to model their dynamics. We compare these models to those obtained using the classic mathematical method of weakly nonlinear (WNL) analysis. This approach is implemented on a one-dimensional mean-field dynamo model that parameterizes the main components of a convective dynamo in a low-mass star -- helical convection and differential rotation. We recover oscillatory dynamo models as a function of the dynamo strength parameter $D\sim \alpha \Omega'$, magnetic dissipation parameter $\kappa$, and a comprehensive dynamo model that predicts the magnetic state for a combination of these two parameters. Our results suggest that equations discovered with SINDy are more robust than equations from WNL analysis, and can predict the saturation amplitude of magnetic fields in parameter regimes far from the onset of dynamo action characterized by stiff nonlinearities. This includes unstable, and typically unknown, subcritical branches. Further to this, SINDy is able to find equations in parameter regimes where the nonlinearity is not analytic and WNL analysis cannot be applied. These properties of data-driven SINDy models suggest them as a viable alternative for modeling of stellar dynamo cycles directly from the data.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 3 linked inside Pith

  1. [1]

    Tobias and N

    S. Tobias and N. Weiss,The solar dynamo and the tachocline. Cambridge University Press, 2007, p. 319–350

  2. [2]

    Stellar activity cycles,

    S. V. Jeffers, R. Kiefer, and T. S. Metcalfe, “Stellar activity cycles,”Space Science Reviews, vol. 219, no. 7, p. 54, 2023

  3. [3]

    The plato mission,

    H. Rauer, C. Aerts, J. Cabrera, M. Deleuil, A. Erikson, L. Gizon, M. Goupil, A. Heras, T. Walloschek, J. Lorenzo-Alvarezet al., “The plato mission,”Experimental Astronomy, vol. 59, no. 3, p. 26, 2025

  4. [4]

    Evolution of solar and stellar dynamo theory,

    P. Charbonneau and D. Sokoloff, “Evolution of solar and stellar dynamo theory,”Space Science Reviews, vol. 219, no. 5, p. 35, 2023

  5. [5]

    The solar dynamo begins near the surface,

    G. M. Vasil, D. Lecoanet, K. Augustson, K. J. Burns, J. S. Oishi, B. P. Brown, N. Brummell, and K. Julien, “The solar dynamo begins near the surface,”Nature, vol. 629, no. 8013, pp. 769–772, 2024. [Online]. Available: https://doi.org/10.1038/s41586-024-07315-1

  6. [6]

    Hydromagnetic dynamo models

    E. N. Parker, “Hydromagnetic dynamo models.”Astrophys. J., vol. 122, p. 293, 1955

  7. [7]

    Solar-cycle dynamo wave propagation,

    H. Yoshimura, “Solar-cycle dynamo wave propagation,”Astrophysical Journal, vol. 201, Nov. 1, 1975, pt. 1, p. 740-748., vol. 201, pp. 740–748, 1975

  8. [8]

    Reconciling solar and stellar magnetic cycles with nonlinear dynamo simulations,

    A. Strugarek, P. Beaudoin, P. Charbonneau, A. Brun, and J.-D. do Nascimento Jr, “Reconciling solar and stellar magnetic cycles with nonlinear dynamo simulations,”Science, vol. 357, no. 6347, pp. 185–187, 2017

Show all 45 references
  1. [9]

    Oscillatory dynamos and their induction mechanisms,

    M. Schrinner, L. Petitdemange, and E. Dormy, “Oscillatory dynamos and their induction mechanisms,” Astronomy & Astrophysics, vol. 530, p. A140, 2011

  2. [10]

    The turbulent dynamo,

    S. Tobias, “The turbulent dynamo,”Journal of fluid mechanics, vol. 912, p. P1, 2021

  3. [11]

    The solar cycle: parity interactions and amplitude modulation

    S. M. Tobias, “The solar cycle: parity interactions and amplitude modulation.”Astronomy & Astrophysics, vol. 322, pp. 1007–1017, Jun. 1997

  4. [12]

    Chaotic dynamics of the magnetic field generated by dynamo action in a turbu- lentflow,

    F. P´ etr´ elis and S. Fauve, “Chaotic dynamics of the magnetic field generated by dynamo action in a turbu- lentflow,”Journal of Physics: Condensed Matter, vol. 20, no. 49, p. 494203, 2008

  5. [13]

    Modulation and symmetry changes in stellar dynamos,

    E. Knobloch, S. Tobias, and N. Weiss, “Modulation and symmetry changes in stellar dynamos,”Monthly Notices of the Royal Astronomical Society, vol. 297, no. 4, pp. 1123–1138, 1998

  6. [14]

    Subcritical transition to turbulence triggered by a magnetic dynamo,

    F. Daniel, L. Petitdemange, and C. Gissinger, “Subcritical transition to turbulence triggered by a magnetic dynamo,”arXiv preprint arXiv:2311.17654, 2023

  7. [15]

    Finite amplitude cellular convection,

    W. V. R. Malkus and G. Veronis, “Finite amplitude cellular convection,”Journal of Fluid Mechanics, vol. 4, no. 3, p. 225–260, 1958

  8. [16]

    J. T. Stuart, “On the non-linear mechanics of wave disturbances in stable and unstable parallel flows part

  9. [17]

    the basic behaviour in plane Poiseuille flow,”Journal of Fluid Mechanics, vol. 9, no. 3, p. 353–370, 1960

  10. [18]

    Watson, “On the non-linear mechanics of wave disturbances in stable and unstable parallel flows part

    J. Watson, “On the non-linear mechanics of wave disturbances in stable and unstable parallel flows part

  11. [19]

    the development of a solution for plane poiseuille flow and for plane couette flow,”Journal of Fluid Mechanics, vol. 9, no. 3, p. 371–389, 1960

  12. [20]

    Galerkin method for nonlinear dynamics,

    B. R. Noack, M. Schlegel, M. Morzynski, and G. Tadmor, “Galerkin method for nonlinear dynamics,” in Reduced-order modelling for flow control. Springer, 2011, pp. 111–149

  13. [21]

    Discovering governing equations from partial measurements with deep delay autoencoders,

    J. Bakarji, K. Champion, J. Nathan Kutz, and S. L. Brunton, “Discovering governing equations from partial measurements with deep delay autoencoders,”Proceedings of the Royal Society A, vol. 479, no. 2276, p. 20230422, 2023

  14. [22]

    Discovering governing equations from data by sparse identification of nonlinear dynamical systems,

    S. L. Brunton, J. L. Proctor, and J. N. Kutz, “Discovering governing equations from data by sparse identification of nonlinear dynamical systems,”P. Natl. Acad. Sci. USA, vol. 113, no. 15, pp. 3932–3937, 2016

  15. [23]

    On the role of nonlinear correlations in reduced-order modelling,

    J. L. Callaham, S. L. Brunton, and J.-C. Loiseau, “On the role of nonlinear correlations in reduced-order modelling,”Journal of Fluid Mechanics, vol. 938, p. A1, 2022. 18

  16. [24]

    An empirical mean-field model of symmetry- breaking in a turbulent wake,

    J. L. Callaham, G. Rigas, J.-C. Loiseau, and S. L. Brunton, “An empirical mean-field model of symmetry- breaking in a turbulent wake,”Science advances, vol. 8, no. 19, p. eabm4786, 2022

  17. [25]

    Data-driven modeling of the chaotic thermal convection in an annular thermosyphon,

    J.-C. Loiseau, “Data-driven modeling of the chaotic thermal convection in an annular thermosyphon,” Theor. Comp. Fluid Dyn., vol. 34, no. 4, pp. 339–365, 2020

  18. [26]

    A sparse optimal closure for a reduced-order model of wall- bounded turbulence,

    Z. C. Khoo, C. H. Chan, and Y. Hwang, “A sparse optimal closure for a reduced-order model of wall- bounded turbulence,”Journal of Fluid Mechanics, vol. 939, p. A11, 2022

  19. [27]

    Data-driven stabilization of periodic orbits,

    J. J. Bramburger, J. N. Kutz, and S. L. Brunton, “Data-driven stabilization of periodic orbits,”IEEE Access, vol. 9, pp. 43 504–43 521, 2021

  20. [28]

    Data-driven discovery and extrapolation of parameterized pattern-forming dynamics,

    Z. G. Nicolaou, G. Huo, Y. Chen, S. L. Brunton, and J. N. Kutz, “Data-driven discovery and extrapolation of parameterized pattern-forming dynamics,”Physical Review Research, vol. 5, no. 4, p. L042017, 2023

  21. [29]

    Ensemble-sindy: Robust sparse model discovery in the low-data, high-noise limit, with active learning and control,

    U. Fasel, J. N. Kutz, B. W. Brunton, and S. L. Brunton, “Ensemble-sindy: Robust sparse model discovery in the low-data, high-noise limit, with active learning and control,”Proceedings of the Royal Society A, vol. 478, no. 2260, p. 20210904, 2022

  22. [30]

    Convergence of uncertainty estimates in ensemble and bayesian sparse model discovery,

    L. Gao, U. Fasel, S. L. Brunton, and J. N. Kutz, “Convergence of uncertainty estimates in ensemble and bayesian sparse model discovery,”arXiv preprint arXiv:2301.12649, 2023

  23. [31]

    Strong asymmetry in stellar dynamos,

    P. Bushby, “Strong asymmetry in stellar dynamos,”Monthly Notices of the Royal Astronomical Society, vol. 338, no. 3, pp. 655–664, 2003

  24. [32]

    Diffusivity quenching as a mechanism for parker’s surface dynamo,

    S. Tobias, “Diffusivity quenching as a mechanism for parker’s surface dynamo,”Astrophysical Journal v. 467, p. 870, vol. 467, p. 870, 1996

  25. [33]

    Dedalus: A flexible framework for numerical simulations with spectral methods,

    K. J. Burns, G. M. Vasil, J. S. Oishi, D. Lecoanet, and B. P. Brown, “Dedalus: A flexible framework for numerical simulations with spectral methods,”Physical Review Research, vol. 2, no. 2, p. 023068, 2020

  26. [34]

    Dynamic mode decomposition and its variants,

    P. J. Schmid, “Dynamic mode decomposition and its variants,”Annual Review of Fluid Mechanics, vol. 54, pp. 225–254, 2022

  27. [35]

    Higher order dynamic mode decomposition,

    S. Le Clainche and J. M. Vega, “Higher order dynamic mode decomposition,”SIAM Journal on Applied Dynamical Systems, vol. 16, no. 2, pp. 882–925, 2017

  28. [36]

    Data-driven scale identification in oscillatory dynamos,

    A. Guseva, “Data-driven scale identification in oscillatory dynamos,”Monthly Notices of the Royal Astro- nomical Society, vol. 528, no. 2, pp. 1685–1696, 2024

  29. [37]

    Pysindy: A comprehensive python package for robust sparse system identification,

    A. A. Kaptanoglu, B. M. de Silva, U. Fasel, K. Kaheman, A. J. Goldschmidt, J. Callaham, C. B. Delahunt, Z. G. Nicolaou, K. Champion, J.-C. Loiseau, J. N. Kutz, and S. L. Brunton, “Pysindy: A comprehensive python package for robust sparse system identification,”Journal of Open ...

  30. [38]

    Pysindy: a python package for the sparse identification of nonlinear dynamics from data,

    B. M. De Silva, K. Champion, M. Quade, J.-C. Loiseau, J. N. Kutz, and S. L. Brunton, “Pysindy: a python package for the sparse identification of nonlinear dynamics from data,”arXiv preprint arXiv:2004.08424, 2020

  31. [39]

    A unified framework for sparse relaxed regularized regression: Sr3,

    P. Zheng, T. Askham, S. L. Brunton, J. N. Kutz, and A. Y. Aravkin, “A unified framework for sparse relaxed regularized regression: Sr3,”IEEE Access, vol. 7, pp. 1404–1423, 2018

  32. [40]

    Invariant recurrent solutions embedded in a turbulent two-dimensional kolmogorov flow,

    G. J. Chandler and R. R. Kerswell, “Invariant recurrent solutions embedded in a turbulent two-dimensional kolmogorov flow,”Journal of Fluid Mechanics, vol. 722, p. 554–595, 2013

  33. [41]

    On nonlinear transitions, minimal seeds and exact solutions for the geodynamo,

    C. S. Skene, F. Marcotte, and S. M. Tobias, “On nonlinear transitions, minimal seeds and exact solutions for the geodynamo,”Journal of Fluid Mechanics, vol. 1021, p. A37, 2025

  34. [42]

    The Openpipeflow Navier–Stokes solver,

    A. P. Willis, “The Openpipeflow Navier–Stokes solver,”SoftwareX, vol. 6, no. 4, pp. 124–127, 2017

  35. [43]

    JFNK-hookstep,

    ——, “JFNK-hookstep,” https://github.com/apwillis1/JFNK-Hookstep, 2019

  36. [44]

    Chaotically modulated stellar dynamos,

    S. Tobias, N. Weiss, and V. Kirk, “Chaotically modulated stellar dynamos,”Monthly Notices of the Royal Astronomical Society, vol. 273, no. 4, pp. 1150–1166, 1995

  37. [45]

    Data-driven reconstruction of a low-order dynamo model from sunspot data,

    A. Bonanno and R. Arlt, “Data-driven reconstruction of a low-order dynamo model from sunspot data,” Solar Physics, vol. 300, no. 8, p. 116, 2025. 19 A Third-order expansion To third order inϵwe get q= (A, B) T = h ϵGqG exp(iωc i t) +ϵ 3q3,ωc i exp(iωc i t) +q 3,3ωc i exp(3iωc ...

Pith tools

Reviewed July 13, 2026 · model on record in the stance chip above.