pith. sign in

arxiv: 2605.21864 · v1 · pith:FWELDYHBnew · submitted 2026-05-21 · ⚛️ physics.ao-ph

A Simulation Methodology Testbed for Typhoon Sensitivity Analysis: Framework Development and Perturbation-Response Experiments with the Pangu Weather Model

Yuehua Peng , Yuchen Zhang , Qin Huang , Chengzhi Ye , Jingsong Yang This is my paper

Pith reviewed 2026-05-22 03:14 UTC · model grok-4.3

classification ⚛️ physics.ao-ph
keywords typhoon sensitivity analysisperturbation-response experimentsAI weather modelPID closed-loop controlsimulation testbedtrack and intensityenvironmental perturbationspredictability limits
0
0 comments X

The pith

A testbed built on an operational AI weather forecasting model allows controlled experiments on typhoon sensitivity to localized perturbations.

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

This paper constructs a simulation testbed by linking an AI weather prediction system with proportional-integral-derivative closed-loop control to examine how typhoons respond to artificial changes in their environment. The setup uses a single-input single-output configuration where perturbations in velocity and temperature serve as inputs, and changes in storm track and intensity are measured as outputs. A reader would care because this method offers a way to explore the boundaries of typhoon predictability and the potential effects of targeted environmental modifications in a more realistic setting than traditional low-order models. The experiments map out response ranges and coupling strengths between inputs and outputs, providing a practical platform for further studies.

Core claim

The central discovery is the development of a modular testbed that incorporates a meteorological prediction module based on an AI model, an artificial perturbation input interface, a typhoon quantitative modeling module, and a PID closed-loop test module. This architecture supports perturbation-response experiments that quantify the feasible ranges for inputs, the tuning of control parameters, and the energy-scale characteristics of the system's responses under different perturbation modes.

What carries the argument

The PID closed-loop test module combined with the artificial perturbation input interface within the overall modular framework.

If this is right

  • The test system identifies feasible perturbation-response ranges for velocity and thermal inputs.
  • Parameter tuning behavior in the PID module is characterized for the test system.
  • Energy-scale response characteristics are quantified for different perturbation modes.
  • Input-output coupling relationships are established for the single-input single-output setup.
  • The framework supports future expansion to multi-input multi-output architectures and advanced control strategies.

Where Pith is reading between the lines

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

  • This approach could link directly to assessing real limits of predictability in typhoon forecasting.
  • Testable extensions might include validating simulated responses against historical typhoon event data.
  • Such a platform might eventually support studies on the feasibility of intensity or track modification in controlled simulations.

Load-bearing premise

The AI weather forecasting model produces physically realistic responses to the localized artificial perturbations that align closely enough with real-world typhoon behavior.

What would settle it

If the magnitude or direction of simulated typhoon track shifts from a given perturbation fails to match observed shifts in real typhoons under comparable environmental changes, this would challenge the testbed's suitability for sensitivity analysis.

Figures

Figures reproduced from arXiv: 2605.21864 by Chengzhi Ye, Jingsong Yang, Qin Huang, Yuchen Zhang, Yuehua Peng.

Figure 1
Figure 1. Figure 1: Comparison of Pangu weather model predictions with ECMWF [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Distribution of pressure relative errors between Pangu and ECMWF at 06:00 on [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Distribution of temperature relative errors around the typhoon calculated by the model [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 7
Figure 7. Figure 7: Closed-loop control of directional artificial regulation of typhoon based on proportional-integral-differential method 4.2 Code Implementation for MATLAB Platform The specific interface of the Pangu weather model is Open Neural Network Exchange (ONNX). ONNX addresses the challenge that neural network models trained on one framework cannot be directly deployed on another. It supports numerous machine learni… view at source ↗
Figure 8
Figure 8. Figure 8: Technical roadmap for MATLAB-based code implementation. As illustrated in [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗
read the original abstract

Understanding how typhoons respond to localized perturbations in their environmental fields is fundamental to assessing the limits of predictability and exploring the potential for track or intensity intervention. This study develops a dedicated simulation methodology testbed for typhoon sensitivity analysis by integrating the Pangu weather model, a high-precision AI forecasting system, with Proportional-Integral-Derivative (PID) closed-loop techniques. The testbed is constructed with modular functional blocks including a meteorological prediction module, an artificial perturbation input interface, a typhoon quantitative modeling module, and a PID closed-loop test module, implemented via a cross-platform MATLAB/ONNX technical framework. A Single-Input Single-Output (SISO) test system was built, with velocity and thermal perturbations set as the core inputs and typhoon track and intensity as the key output targets, to perform controlled perturbation-response experiments. The experiments reveal the feasible perturbation-response range, the parameter tuning behavior of the PID module, and the energy-scale response characteristics under different perturbation modes, and quantify the input-output coupling relationships of the test system. By constructing this testbed on an operational AI weather forecasting model, this study provides a framework that goes beyond idealized sensitivity studies typically validated only on low-order dynamical models. The testbed offers an expandable platform for investigating typhoon sensitivity to artificial environmental perturbations and provides a foundation for subsequent expansion toward multi-input multi-output architectures and advanced analysis strategies such as nonlinear PID or model predictive control.

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 develops a dedicated simulation methodology testbed for typhoon sensitivity analysis by integrating the Pangu AI weather forecasting model with PID closed-loop control in a cross-platform MATLAB/ONNX framework. It defines modular blocks (meteorological prediction, artificial perturbation input, typhoon quantitative modeling, and PID test module) and implements a SISO system with velocity/thermal perturbations as inputs and typhoon track/intensity as outputs. Experiments characterize feasible perturbation ranges, PID tuning behavior, energy-scale responses, and input-output couplings. The central claim is that the testbed, built on an operational AI model, provides a framework that goes beyond idealized sensitivity studies validated only on low-order dynamical models.

Significance. If the Pangu responses to localized artificial perturbations prove physically realistic and consistent with observed typhoon dynamics, the modular testbed could offer a useful expandable platform for perturbation-response studies and potential extensions to MIMO or advanced control strategies. The integration of an operational AI forecaster with standard PID methods is a practical strength, and the explicit SISO architecture allows controlled experiments that are harder to isolate in full physics-based models.

major comments (2)
  1. [Abstract and §3] Abstract and §3 (Framework Development): the claim that the testbed 'goes beyond idealized sensitivity studies typically validated only on low-order dynamical models' is load-bearing for the contribution, yet the manuscript provides no conservation checks, energy-balance diagnostics, or direct comparisons of perturbation responses against observations or physics-based models (e.g., WRF). Without such grounding, the transfer of Pangu's forecasting skill to realistic sensitivity under artificial perturbations remains an assumption rather than a demonstrated property.
  2. [§4] §4 (Perturbation-Response Experiments): the reported feasible ranges, PID tuning behavior, and input-output couplings are described qualitatively with no quantitative metrics, error bars, or statistical measures of robustness; this undermines assessment of whether the observed couplings are reproducible or physically meaningful.
minor comments (2)
  1. [Abstract and §4] Clarify the precise definition and units of 'energy-scale response characteristics' in the abstract and §4; the term appears without an explicit formula or reference to a conserved quantity.
  2. [Figures] Figure captions and axis labels in the perturbation-response plots should explicitly state the perturbation magnitudes and output variables to improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed review. The comments highlight important aspects of validation and quantitative rigor that we address below. We propose targeted revisions to strengthen the manuscript while preserving the core contribution of the modular testbed framework.

read point-by-point responses

  1. Referee: [Abstract and §3] Abstract and §3 (Framework Development): the claim that the testbed 'goes beyond idealized sensitivity studies typically validated only on low-order dynamical models' is load-bearing for the contribution, yet the manuscript provides no conservation checks, energy-balance diagnostics, or direct comparisons of perturbation responses against observations or physics-based models (e.g., WRF). Without such grounding, the transfer of Pangu's forecasting skill to realistic sensitivity under artificial perturbations remains an assumption rather than a demonstrated property.

    Authors: We acknowledge that the manuscript does not present explicit conservation checks, energy-balance diagnostics, or side-by-side comparisons against physics-based models such as WRF or observational datasets. The primary contribution lies in the construction of a modular, cross-platform testbed that enables controlled SISO perturbation-response experiments on an operational AI forecasting model. While Pangu's documented skill in standard forecasting tasks provides a foundation, we agree this does not automatically guarantee physical consistency under artificial perturbations. In the revised manuscript we will expand §3 with a dedicated limitations subsection that discusses these gaps and incorporates preliminary energy-balance diagnostics derived from the existing model outputs. We will also add a forward-looking paragraph outlining planned comparisons with WRF and reanalysis data. These changes clarify the current scope without overstating the demonstrated realism. revision: partial

  2. Referee: [§4] §4 (Perturbation-Response Experiments): the reported feasible ranges, PID tuning behavior, and input-output couplings are described qualitatively with no quantitative metrics, error bars, or statistical measures of robustness; this undermines assessment of whether the observed couplings are reproducible or physically meaningful.

    Authors: The experiments in §4 were designed as an initial characterization of the testbed's behavior rather than a comprehensive statistical study. We recognize that the purely qualitative presentation limits evaluation of reproducibility and physical significance. In the revised version we will augment §4 with quantitative metrics, including standard deviations across repeated runs for the reported ranges, integral-absolute-error values for PID tuning performance, and Pearson correlation coefficients together with p-values for the input-output coupling relationships. These additions will allow readers to assess robustness more directly while retaining the exploratory character of the current results. revision: yes

Circularity Check

0 steps flagged

No circularity: methodological framework built on external model

full rationale

The paper develops a simulation testbed by integrating the external Pangu AI weather model with standard PID closed-loop control and modular MATLAB/ONNX blocks for SISO perturbation-response experiments on typhoon track and intensity. No equations, fitted parameters, or predictions are defined in terms of outputs from the same study; the reported feasible ranges, coupling relationships, and energy-scale characteristics are observational results from running the pre-existing Pangu model under controlled inputs. The central claim of providing a framework beyond low-order dynamical models rests on the choice of an operational AI forecaster rather than any self-referential derivation or self-citation chain that reduces the result to its own inputs by construction. This is a self-contained engineering contribution with no load-bearing steps that collapse into fitted quantities or renamed ansatzes.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The framework rests on the assumption that the Pangu model can be treated as a black-box dynamical system suitable for closed-loop perturbation experiments and that standard PID tuning applies directly to meteorological outputs.

axioms (2)
  • domain assumption Pangu weather model responses to localized perturbations are sufficiently realistic for sensitivity analysis
    Invoked when the testbed is constructed on the operational AI model rather than a low-order dynamical system.
  • domain assumption PID controllers can be tuned to produce stable and interpretable input-output relationships in this meteorological setting
    Central to the closed-loop test module and the reported parameter tuning behavior.

pith-pipeline@v0.9.0 · 5807 in / 1454 out tokens · 45008 ms · 2026-05-22T03:14:51.591701+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

27 extracted references · 27 canonical work pages

  1. [1]

    R., Gila, M., Klein, M., Martinasso, M., and Schulthess, T

    Alam, S. R., Gila, M., Klein, M., Martinasso, M., and Schulthess, T. C., 2023: Versatile software-defined HPC and cloud clusters on ALPS supercomputer for diverse workflows. The International Journal of High Performance Computing Applications , 37(3–4), 288 –305, https://doi.org/10.1177/10943420231167811

    work page doi:10.1177/10943420231167811 2023

  2. [2]

    Nature, 525(7567), 47–55, https://doi.org/10.1038/nature14956

    Bauer, P., Thorpe, A., and Brunet, G., 2015: The quiet revolution of numerical weather prediction. Nature, 525(7567), 47–55, https://doi.org/10.1038/nature14956

    work page doi:10.1038/nature14956 2015

  3. [3]

    D., Hoefler, T., Quintino, T., Schulthess, T

    Bauer, P ., Dueben, P . D., Hoefler, T., Quintino, T., Schulthess, T. C., and Wedi, N. P ., 2021: The digital revolution of Earth -system science. Nature Computational Science , 1(2), 104 –113, https://doi.org/10.1038/s43588-021-00023-0

    work page doi:10.1038/s43588-021-00023-0 2021

  4. [4]

    Nature Reviews Earth & Environment, 4(8), 507–509, https://doi.org/10.1038/s43017-023-00468-z

    Bauer, P ., Dueben, P ., Chantry, M., Doblas-Reyes, F., Hoefler, T., McGovern, A., and Stevens, B., 2023: Deep learning and a changing economy in weather and climate prediction. Nature Reviews Earth & Environment, 4(8), 507–509, https://doi.org/10.1038/s43017-023-00468-z. Ben Bouallègue, Z., Weyn, J. A., Clare, M. C. A., Dramsch, J., Dueben, P ., and Chan...

    work page doi:10.1038/s43017-023-00468-z 2023

  5. [5]

    Bulletin of the American Meteorological Society , 105, E864 –E883, https://doi.org/10.1175/BAMS-D-23-0162.1

    Sandu, I., Dueben, P., Chantry, M., and Pappenberger, F., 2024: The rise of data-driven weather forecasting: A first statistical assessment of machine learning -based weather forecasts in an operational-like context. Bulletin of the American Meteorological Society , 105, E864 –E883, https://doi.org/10.1175/BAMS-D-23-0162.1

    work page doi:10.1175/bams-d-23-0162.1 2024

  6. [6]

    arXiv preprint arXiv:2211.02556 , year=

    Bi, K., Xie, L., Zhang, H., Chen, X., Gu, X., and Tian, Q., 2022: Pangu -weather: A 3D high-resolution model for fast and accurate global weather forecast. arXiv preprint arXiv:2211.02556, https://doi.org/10.48550/arXiv.2211.02556

    work page doi:10.48550/arxiv.2211.02556 2022

  7. [7]

    doi: 10.1038/s41586-023-06185-3

    Bi, K., Xie, L., Zhang, H., et al., 2023: Accurate medium-range global weather forecasting with 3D neural networks. Nature, 619, 533–538, https://doi.org/10.1038/s41586-023-06185-3

    work page doi:10.1038/s41586-023-06185-3 2023

  8. [8]

    Journal of Advances in Modeling Earth Systems , 12(12), e2020MS002232, https://doi.org/10.1029/2020MS002232

    Bonavita, M., and Laloyaux, P ., 2020: Machine learning for model error inference and correction. Journal of Advances in Modeling Earth Systems , 12(12), e2020MS002232, https://doi.org/10.1029/2020MS002232

    work page doi:10.1029/2020ms002232 2020

  9. [9]

    Wiley Interdisciplinary Reviews: Climate Change, 9(5), e535, https://doi.org/10.1002/wcc.535

    Carrassi, A., Bocquet, M., Bertino, L., and Evensen, G., 2018: Data assimilation in the geosciences: An overview of methods, issues, and perspectives. Wiley Interdisciplinary Reviews: Climate Change, 9(5), e535, https://doi.org/10.1002/wcc.535

    work page doi:10.1002/wcc.535 2018

  10. [10]

    Sustainability, 15(24), 16681 , https://doi.org/10.3390/su152416681

    Cheon, M., and Mun, C., 2023: The climate of innovation: AI’s growing influence in weather prediction patents and its future prospects. Sustainability, 15(24), 16681 , https://doi.org/10.3390/su152416681

    work page doi:10.3390/su152416681 2023

  11. [11]

    Y ., Yang, J., Zhang, Z

    Fan, P. Y ., Yang, J., Zhang, Z. P., Zang, N. H., Li, Y . F., and Feng, G. L., 2023: Summer precipitation prediction in eastern China based on machine learning. Climate Dynamics, 60(9–10), 2645– 2663, https://doi.org/10.1007/s00382-022-06464-1

    work page doi:10.1007/s00382-022-06464-1 2023

  12. [12]

    G., Kim, J

    Ham, Y . G., Kim, J. H., and Luo, J. J., 2019: Deep learning for multi-year ENSO forecasts. Nature, 573(7775), 568–572, https://doi.org/10.1038/s41586-019-1559-7

    work page doi:10.1038/s41586-019-1559-7 2019

  13. [13]

    G., Kim, J

    Ham, Y . G., Kim, J. H., Kim, E. S., and On, K. W., 2021: Unified deep learning model for El Niño/Southern Oscillation forecasts by incorporating seasonality in climate data. Science Bulletin, 66(13), 1358–1366, https://doi.org/10.1016/j.scib.2021.03.009

    work page doi:10.1016/j.scib.2021.03.009 2021

  14. [14]

    M., and Boers, N., 2022: Physically constrained generative adversarial networks for improving precipitation fields from Earth system models

    Hess, P., Drüke, M., Petri, S., Strnad, F. M., and Boers, N., 2022: Physically constrained generative adversarial networks for improving precipitation fields from Earth system models. Nature Machine Intelligence, 4(10), 828–839, https://doi.org/10.1038/s42256-022-00540-1

    work page doi:10.1038/s42256-022-00540-1 2022

  15. [15]

    W., Jamshidi, M

    Hipel, K. W., Jamshidi, M. M., Tien, J. M., and White III, C. C., 2007: The future of systems, man, and cybernetics: Application domains and research methods. IEEE Transactions on Systems, Man, and Cybernetics, 37, 726–743, https://doi.org/10.1109/tsmcc.2007.900671

    work page doi:10.1109/tsmcc.2007.900671 2007

  16. [16]

    G., Joo, Y

    Kim, H., Ham, Y . G., Joo, Y . S., and Son, S. W., 2021: Deep learning for bias correction of MJO prediction. Nature Communications, 12, 3087, https://doi.org/10.1038/s41467-021-23406-3

    work page doi:10.1038/s41467-021-23406-3 2021

  17. [17]

    Deep learning,

    LeCun, Y ., Bengio, Y ., and Hinton, G., 2015: Deep learning. Nature, 521, 436 –444, https://doi.org/10.1038/nature14539

    work page doi:10.1038/nature14539 2015

  18. [18]

    I: Direct and inverse or optimal controlling problem for precipitation enhancement operation

    Lei, H., Lei, W., and Zeng, Q., 2012: Cybernetics in the artificial weather modification. I: Direct and inverse or optimal controlling problem for precipitation enhancement operation. Climatic and Environmental Research (in Chinese), 17(6), 968 –978, https://doi.org/10.3878/j.issn.1006-9585.2012.06.33

    work page doi:10.3878/j.issn.1006-9585.2012.06.33 2012

  19. [19]

    Y ., Yu, Y ., Huang, D

    Li, P. Y ., Yu, Y ., Huang, D. N., Wang, Z. H., and Sharma, A., 2023: Regional heatwave prediction using graph neural network and weather station data. Geophysical Research Letters , 50(7), e2023GL103405, https://doi.org/10.1029/2023GL103405

    work page doi:10.1029/2023gl103405 2023

  20. [20]

    H., Luo, J

    Ling, F. H., Luo, J. J., Li, Y ., Tang, T., Bai, L., Ouyang, W. L., and Yamagata, T., 2022: Multi-task machine learning improves multi -seasonal prediction of the Indian Ocean Dipole. Nature Communications, 13(1), 7681, https://doi.org/10.1038/s41467-022-35412-0

    work page doi:10.1038/s41467-022-35412-0 2022

  21. [21]

    J., 2021: ENSO-ASC 1.0.0: ENSO deep learning forecast model with a multivariate air -sea coupler

    Mu, B., Qin, B., and Yuan, S. J., 2021: ENSO-ASC 1.0.0: ENSO deep learning forecast model with a multivariate air -sea coupler. Geoscientific Model Development , 14(11), 6977 –6999, https://doi.org/10.5194/gmd-14-6977-2021

    work page doi:10.5194/gmd-14-6977-2021 2021

  22. [22]

    X., Anderson, G

    Pan, B. X., Anderson, G. J., Goncalves, A., Lucas, D. D., Bonfils, C. J. W., and Lee, J., 2022: Improving seasonal forecast using probabilistic deep learning. Journal of Advances in Modeling Earth Systems, 14, e2021MS002766, https://doi.org/10.1029/2021MS002766

    work page doi:10.1029/2021ms002766 2022

  23. [23]

    Nature, 597(7878), 672–677, https://doi.org/10.1038/s41586-021-03854-z

    Ravuri, S., and Coauthors, 2021: Skilful precipitation nowcasting using deep generative models of radar. Nature, 597(7878), 672–677, https://doi.org/10.1038/s41586-021-03854-z

    work page doi:10.1038/s41586-021-03854-z 2021

  24. [24]

    Journal of Meteorological Research , 34, 675 –698, https://doi.org/10.1007/s13351-020-9847-6

    Shen, X., Wang, J., Li, Z., Chen, D., and Gong, J., 2020, Research and operational development of numerical weather prediction in China. Journal of Meteorological Research , 34, 675 –698, https://doi.org/10.1007/s13351-020-9847-6

    work page doi:10.1007/s13351-020-9847-6 2020

  25. [25]

    and Wan, K., 2011, A spectral mixture model analysis of the Kuroshio variability and the water exchange between the Kuroshio and the East China Sea

    Song, J., Xue, H., Bao, X., Wu, D., Chai, F., Shi, L., Y ao, Z., Wang, Y ., Nan, F. and Wan, K., 2011, A spectral mixture model analysis of the Kuroshio variability and the water exchange between the Kuroshio and the East China Sea. Journal of Oceanology and Limnology , 29, 446 –459, https://doi.org/10.1007/s00343-011-0114-7

    work page doi:10.1007/s00343-011-0114-7 2011

  26. [26]

    and Li, H., 2017: Definition of water exchange zone between the Bohai Sea and Y ellow Sea and the effect of winter gale on it

    Song, J., Guo, J., Li, J., Mu, L., Liu, Y ., Wang, G., Li, Y . and Li, H., 2017: Definition of water exchange zone between the Bohai Sea and Y ellow Sea and the effect of winter gale on it. Acta Oceanologica Sinica, 36, 17–25, https://doi.org/10.1007/s13131-017-0989-z

    work page doi:10.1007/s13131-017-0989-z 2017

  27. [27]

    H., and Kozma, R., 2021: Systems science and engineering research in the context of systems, man, and cybernetics: Recollection, trends, and future directions

    Pedrycz, W., Smith, M. H., and Kozma, R., 2021: Systems science and engineering research in the context of systems, man, and cybernetics: Recollection, trends, and future directions. IEEE Transactions on Systems, Man, and Cybernetics: Systems , 51, 5 –21, https://doi.org/10.1109/tsmc.2020.3043192

    work page doi:10.1109/tsmc.2020.3043192 2021