Pith. sign in

REVIEW 3 major objections 5 minor 47 references

VQLTI: Long-Term Tropical Cyclone Intensity Forecasting with Physical Constraints

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read This paper proposes VQLTI, a two-stage model that reframes tropical-cyclone intensity forecasting as a discrete-latent-space prediction problem and shows that injecting physical knowledge—the FengWu forecast field and a…

desk verdict A promising integration of VQ latents, FengWu forecast fields, and potential intensity for TC intensity forecasting, but the headline SOTA claim is built on an oracle evaluation; the real-time evidence is too thin to support it as stated. read the letter →

arxiv 2501.18122 v1 pith:5QBIHIUH submitted 2025-01-30 cs.LG cs.AI

Xinyu Wang , Lei Liu , Kang Chen , Tao Han , Bin Li , Lei Bai This is my paper · ORCID
classification cs.LGcs.AI
keywords tropicalcycloneintensityvectorquantizationdiscretelatentspacepotentialFengWuforecastphysicalconstraintslong-termforecastingERA5
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

Tropical-cyclone intensity forecasts out to five days are vital for disaster warning, but deep-learning models tend to accumulate error and lose skill beyond 24 hours. VQLTI's central idea is to stop forecasting raw intensity and instead forecast a sequence of discrete latent codes, where each code is learned to represent intensity together with its surrounding spatial weather pattern. The model then decodes those predicted codes into intensity, using the FengWu neural weather model's forecast field as a spatial condition and a potential-intensity (PI) value as a physical upper bound. On a global 2019–2020 test set, the authors report cutting maximum sustained wind (MSW) error by 35.65%–42.51% relative to ECMWF-IFS across 24–120h, and a real-time variant that substitutes FengWu forecasts for future reanalysis fields continues to beat existing deep-learning and operational baselines. If this holds, machine-learning models could move from short-term aids to credible five-day intensity guidance at a fraction of operational NWP cost.

What carries the argument

The engine is a two-stage conditional vector-quantized framework. Stage one is a CVQ-VAE: a Feature Pyramid Network extracts multi-scale feature maps from the 40×40 ERA5 patch, the encoder maps the intensity into a 1,024-entry discrete codebook via nearest-neighbor lookup, and the decoder reconstructs intensity conditioned on the same spatial features. Stage two is an iterative latent-space forecaster: a sliding window of past latent codes is passed through stacked cross-attention and an MLP to produce a future-code score, which is then cross-attended to the FengWu forecast field and to a potential-intensity value $V_{max, pmin} = PI(t2m, msl, t, q)$ (computed with Gilford's pyPI implementation) before being added residually to the last historical code. Each component carries a specific load: the codebook forces discrete, spatially-aware intensity representations; the FengWu field supplies the future atmospheric state in real-time use; and the PI term supplies a thermodynamic upper bound that counteracts the tendency of reanalysis and forecast fields to underestimate cyclone winds.

What would settle it

Run a strict re-forecast where every input is available at forecast time (no future ERA5 anywhere in the pipeline) and compare VQLTI's 120h MSW error to the reported real-time 5.53 m/s; if the error jumps well above that or fails to beat ECMWF-IFS on the same storms, the real-time claim is falsified. A second check is sensitivity: zero out the potential-intensity input and see whether the 120h forecast changes by more than training noise; if not, the physical constraint is not doing the claimed work.

Watch

Extended reading notes

Core claim

The paper's central claim is that tropical cyclone intensity forecasting should be performed over a learned discrete latent space rather than over continuous intensity values, because the same intensity number can correspond to many different storm structures and environmental contexts. VQLTI first trains a conditional vector-quantized variational autoencoder (CVQ-VAE) that encodes the observed intensity into one of 1,024 learnable codebook entries while attending to multi-scale ERA5 spatial features, preserving both the intensity value and its spatial differences. Next, the model freezes the encoder and codebook, unfolds a sliding window of past latent codes, and iteratively predicts future codes with an attention-and-MLP module, using the FengWu forecast field and a computed potential-intensity value as cross-attention conditions and physical constraints. The authors report that this design slows error accumulation, yields state-of-the-art global 24–120h forecasts on 2019–2020 data, and that a fine-tuned real-time variant (with FengWu forecast fields substituted for future ERA5) remains superior to prior deep-learning methods, ECMWF-IFS, and NCEP-GFS through 120h.

Load-bearing premise

The headline accuracy numbers are measured with the model seeing the future atmospheric state as input, so the real-time claim depends on the assumption that a neural weather model's forecast fields can stand in for that future state without leaking the storm's actual intensity.

Editorial extensions

If this is right

  • If the 35.65%–42.51% MSW error reduction transfers to operations, a single learned model could replace or post-process NWP intensity forecasts, lowering the computing cost of early-warning systems.
  • The much slower error accumulation (48h MSLP error grows 16.01% versus 61–70% for prior deep-learning models) implies the latent-space iteration is the key to stable multi-day forecasts.
  • Because forecast length is a parameter and only the latent forecaster is trained, the same checkpoints can produce forecasts at arbitrary lead times and likely transfer to other meteorological target variables.
  • The real-time variant, though degraded relative to the retrospective setup, still beats all compared deep-learning baselines and ECMWF-IFS at 120h, suggesting the FengWu-substitution strategy is a workable path toward operational ML forecasts.
  • Ablation shows that codebook quantization, FengWu conditioning, and the PI constraint each contribute independently, so the architecture's components are modular and may generalize separately.

Reading between the lines

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

  • Editorial inference: the headline 35.65%–42.51% improvement holds in a retrospective evaluation that conditions on future ERA5 fields; the real-time table shows a smaller but still positive advantage, so the operational gain is likely closer to the real-time numbers than to the headline.
  • Editorial inference: because the PI constraint and the FengWu conditioning are both derived from the same forecast field, the physical constraint may be partly redundant with the conditioning signal; using an independently observed sea-surface temperature to compute PI would test whether the constraint adds genuinely new information.
  • Editorial inference: real-time performance depends on how closely FengWu's forecast tracks the storm location, since long-lead forecasts drift spatially; a stress test with deliberately misplaced tracker positions would reveal the model's sensitivity to tracker errors.
  • Editorial inference: the discrete codebook maps intensities to a finite set of codes, which gives a natural categorical (e.g., Saffir-Simpson) reading; the same architecture could be extended to probabilistic forecasts by outputting a distribution over codebook entries.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes VQLTI, a two-stage vector-quantized model for tropical cyclone intensity forecasting. Stage 1 uses a conditional VQ-VAE to map historical intensity into a discrete latent space conditioned on ERA5 spatial fields; stage 2 iteratively forecasts the latent variables with cross-attention and a sliding window, then decodes them into intensity using FengWu forecast fields as conditioning inputs and a potential-intensity (PI) physical constraint. The manuscript claims state-of-the-art 24-120 h global TC intensity forecasts, with MSW error reduced by 35.65%-42.51% relative to ECMWF-IFS, and additionally reports a 2022 real-time forecasting experiment and ablations over the VQ, FengWu-field, and PI components.

Significance. If the claimed results were obtained under a realistic forecast protocol, VQLTI would be a notable advance: it is among the first attempts to condition TC intensity forecasting on forecast fields from an ML weather model, and the discrete-latent formulation with a PI constraint is a plausible way to combat long-term error accumulation. The authors release code, and the ablation design cleanly separates the contributions of the codebook, FengWu fields, and PI. However, the headline quantitative claim is tied to a non-real-time evaluation that appears to condition on future ERA5 fields, which are unavailable at forecast time; the real-time experiment that avoids this is limited to one year, one run, and no significance testing. The significance of the work therefore rests on a protocol that the paper itself does not unambiguously define and that, as written, does not support the stated operational SOTA claim.

major comments (3)
  1. [Real-time Forecast; Table 1] The abstract's headline 35.65%-42.51% MSW error reduction is tied to Table 1, but the Long-term Forecast section never states what conditioning inputs VQLTI receives. The Real-time Forecast section then says, 'Second, future ERA5 fields are also unavailable, so we substitute them with FengWu’s forecast results.' This reveals that the non-real-time evaluation (Table 1) conditioned the decoder on future ERA5 fields. Those fields contain the future atmospheric state of the storm, including wind structure directly correlated with the MSW target, so the comparison against ECMWF-IFS and NCEP-GFS—genuine forecasts made without future data—is not apples-to-apples. The Base ERA5 row in Table 1, with a 24 h MSW MAE of 17.72 m/s, shows how much signal the future box carries by itself. This is a load-bearing flaw for the SOTA claim; the authors must either rerun the 2019-2020 comparison using only inputs available at forecast time, or explicitly relabel Table 1 as an idealized oracle benchmark and remove it from the abstract and conclusion claims.
  2. [Real-time Forecast; Table 2] The real-time experiment is a single year (2022), a single model run, and reports no confidence intervals or significance tests, so the claimed real-time superiority is not statistically established. Moreover, VQLTI is fine-tuned on 2021 data, while the paper does not state that the DL baselines (TC-Pre, MSCAR, TCIF) receive any equivalent additional training or fine-tuning; if those baselines are evaluated with their original checkpoints trained only through 2017, the comparison gives VQLTI an extra-data advantage. The 35.85% reduction over ECMWF-IFS at 120 h in Table 2 therefore needs multi-year evidence and a matched-protocol description of baseline training before it can support an operational claim.
  3. [Forecast; Eq. (1)] The input specification is internally inconsistent. Eq. (1) writes the model input as f(I_{-(n-1):0}, E_{-(n-1):0}, F_{1:m}, m), i.e., it always uses FengWu forecast fields F_{1:m} and no future ERA5, yet the Real-time Forecast section says that 'future ERA5 fields are also unavailable, so we substitute them with FengWu’s forecast results,' implying that earlier experiments used future ERA5. If Table 1 used future ERA5, Eq. (1) is incomplete; if it did not, the substitution sentence is misleading. The manuscript needs one unambiguous protocol statement that identifies, for each table and each row, exactly which future fields were used as conditioning inputs.
minor comments (5)
  1. [Potential Intensity] The sentence 'since FengWu does not perform forecasting' is confusing; the intended meaning is likely that FengWu does not forecast sea-surface temperature, which is why t2m is used as a substitute. Please rephrase.
  2. [Algorithm 1] Line 3 of Algorithm 1 sets 'score = CrossAttn(W[0], W[0], W[0])', which appears to hard-code the first window element rather than the current iteration index; as written, the attention computation over the window is not generally specified. Please clarify or correct the pseudo-code.
  3. [Long-term Forecast; Table 1] There are several typographical errors: 'ECWMF-IFS' should be 'ECMWF-IFS', 'JTCW' should be 'JTWC', and the Table 1 caption contains 'Western North Pacific Pacific'. Please correct these.
  4. [Data description; Figure 2] The text says a 10 degree longitude-latitude patch is cropped around the TC center, while Figure 2's caption mentions a '10 degree diameter'; please make the cropping geometry explicit and consistent.
  5. [Table 2] The 'Base FengWu' row in Table 2 is not defined in the text; state whether it is computed exactly like Base ERA5 in Table 1 (e.g., max wind speed searched in the FengWu forecast box), since that affects the interpretation of VQLTI's improvement over FengWu.

Circularity Check

2 steps flagged · score 6.0 of 10

The headline SOTA claim reduces to a model conditioned on future atmospheric fields; the 'forecast' is a fitted mapping from future wind fields to intensity, so the 35.65-42.51% improvement vs NWP is not an apples-to-apples forecast comparison.

  1. fitted input called prediction [Methodology, 'Forecast' section; Experiments, 'Real-time Forecast' section; Table 1]
    "If we obtain the discrete latent variables corresponding to the future TC intensity, as well as the future ERA5 data, we can then realize accurate forecasting. ... Second, future ERA5 fields are also unavailable, so we substitute them with FengWu’s forecast results."

    The paper's own description identifies the non-real-time protocol as having future ERA5 fields available; the real-time section says the only fix is to 'substitute them with FengWu’s forecast results'. Thus the Table 1 result advertised as a 24-120h forecast is generated by decoding from the future atmospheric state, not from initial-time information. Those future fields include 10-m wind components in the storm-centered box, and the paper's own 'Base ERA5' baseline computes MSW as 'the maximum value of sqrt((u10)^2 + (v10)^2)' (24h MAE 17.72 m/s). The decoder is trained to map that future field to IBTrACS intensity, so the 'prediction' reduces to a fitted correction of the target-bearing input.

  2. other [Methodology, 'Potential Intensity' and Algorithm 1; Ablation Studies, Table 3]
    "we calculate the future PI based on the FengWu forecast data and perform cross-attention between this PI and the MLP output. This provides an upper bound on the possible future TC intensity, which is used to constrain the iterative latent variables physically."

    The physical constraint is not an independent check on the forecast; it is computed from the same future FengWu fields used as the decoder condition. Because those fields contain the future wind state, the PI term injects future information into the latent iteration. The ablation quantifies the dependence: removing the FengWu field degrades 24-120h MSW by 55.29%-70.56%, and the PI component alone improves 120h MSW by 24.35%. So the claimed 'physical knowledge' benefit is entangled with the direct leak of future-state information, not a separate first-principles constraint.

full rationale

The internal equations of VQLTI are not circular in the narrow sense: the encoder-decoder and latent-iteration losses are standard, and the PI formula is an external thermodynamic relation. No load-bearing self-citation chain is present, since FengWu is an externally published model and VQLTI's own design is described independently. The circularity is in the evaluation protocol. The paper's methodology states that accurate forecasting is possible once future ERA5 data are available, and the real-time section reveals that the only change is substituting future ERA5 fields with FengWu forecasts. Therefore the headline Table 1 numbers, which support the abstract's 35.65%-42.51% claim, are produced with future-state fields as conditioning inputs. Since those fields contain 10-m wind components over the storm, the target MSW is directly recoverable (to a first approximation) from the input, and the model is a learned mapping from that future field to IBTrACS intensity. That is a fitted-input-called-prediction design rather than a genuine forecast, and it invalidates the direct comparison to NWP models that do not see future fields. The real-time experiment with FengWu forecasts is a partial mitigation, but it still conditions on a future forecast field and compares against models without such conditioning. Because the central advertised claim is thereby compromised while the model itself remains a plausible learned correction tool, the paper is partially circular rather than wholly vacuous.

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

The model's performance is not derived from first principles. It inherits: (1) a trained prior from VQ-VAE and attention on ERA5 and IBTrACS, (2) the FengWu weather model's forecast fields, and (3) the Emanuel potential intensity theory with an unvalidated t2m-for-SST substitution. The hyperparameters above are hand-selected. The main evaluation also implicitly assumes future reanalysis or FengWu forecast fields are legitimate inputs at forecast time.

free parameters (6)
  • codebook size J = 1024
    Chooses the number of discrete latent vectors; no sensitivity analysis is reported, so the value is hand-selected.
  • latent dimension d_z = 128
    Dimension of discrete embeddings and attention features; hand-selected with no sensitivity analysis.
  • commitment loss coefficient beta = 0.25
    Standard VQ hyperparameter controlling commitment to codebook vectors; chosen by hand.
  • training forecast horizon m = 8
    Stage-2 training predicts 8 steps (48 hours), but evaluation extends to 20 steps (120 hours); the train/inference mismatch is not analyzed.
  • historical window length n = 4
    Uses the past four 6-hourly intensity steps as the initial latent window; chosen without sensitivity analysis.
  • optimizer and regularization settings = lr=1e-4, batch=64, epochs=30, L1=1e-5, EMA
    Hand-selected training settings; no sensitivity or multi-seed variability reported.
assumptions (6)
  • domain assumption IBTrACS USA WIND and USA PRES are accurate enough to serve as ground-truth TC intensity.
    All training and evaluation use these best-track values as truth; biases in best-track wind estimates directly enter the reported errors.
  • domain assumption ERA5 underestimates TC intensity and FengWu forecasts inherit that bias; VQLTI can recover the missing signal from spatial fields.
    Stated in the Introduction to motivate the design; no independent validation of the recovery mechanism is provided.
  • domain assumption The Emanuel potential intensity formula remains a valid upper bound when 2-meter temperature is substituted for sea surface temperature.
    Used in Eq. (4) and the text around it; the substitution error is asserted to be acceptable without quantification.
  • domain assumption ECMWF's TC tracker correctly locates TC centers in FengWu forecast fields.
    The tracker is used to crop the 10-degree FengWu fields; Figure 2 shows spatial misalignment at longer lead times, so tracker errors propagate.
  • domain assumption FengWu forecast fields are valid future atmospheric states containing physical knowledge not already available in ERA5.
    The model is conditioned on these fields, and the ablation attributes most of the skill to them.
  • standard math The temporal train/validation/test split yields representative generalization behavior.
    Standard empirical risk minimization assumption; no distribution shift or regime change analysis is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of VQLTI: Long-Term Tropical Cyclone Intensity Forecasting with Physical Constraints." pith.science (2026). https://pith.science/paper/5QBIHIUH

@misc{pith2026250118122,
  author       = {Pith},
  title        = {Pith review of: VQLTI: Long-Term Tropical Cyclone Intensity Forecasting with Physical Constraints},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5QBIHIUH}},
  note         = {Machine review of arXiv:2501.18122}
}
read the original abstract

Tropical cyclone (TC) intensity forecasting is crucial for early disaster warning and emergency decision-making. Numerous researchers have explored deep-learning methods to address computational and post-processing issues in operational forecasting. Regrettably, they exhibit subpar long-term forecasting capabilities. We use two strategies to enhance long-term forecasting. (1) By enhancing the matching between TC intensity and spatial information, we can improve long-term forecasting performance. (2) Incorporating physical knowledge and physical constraints can help mitigate the accumulation of forecasting errors. To achieve the above strategies, we propose the VQLTI framework. VQLTI transfers the TC intensity information to a discrete latent space while retaining the spatial information differences, using large-scale spatial meteorological data as conditions. Furthermore, we leverage the forecast from the weather prediction model FengWu to provide additional physical knowledge for VQLTI. Additionally, we calculate the potential intensity (PI) to impose physical constraints on the latent variables. In the global long-term TC intensity forecasting, VQLTI achieves state-of-the-art results for the 24h to 120h, with the MSW (Maximum Sustained Wind) forecast error reduced by 35.65%-42.51% compared to ECMWF-IFS.

Figures

Figures reproduced from arXiv: 2501.18122 by the authors.

Figure 1
Figure 1. Past methods, such as LSTM, would map the same [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. For typhoon Hinnamnor at 2022-09-02 18:00, the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. VQLTI Framework. (a) The first stage involves CVQVAE pre-training, which uses ERA5 data as the condition to [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: T-SNE visualization of latent variables. The same [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 33 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Bao, J.; Chen, D.; Wen, F.; Li, H.; and Hua, G. 2017. CVAE-GAN: fine-grained image generation through asymmetric training. In Proceedings of the IEEE international conference on computer vision, 2745--2754

    work page 2017

  4. [4]

    Bi, K.; Xie, L.; Zhang, H.; Chen, X.; Gu, X.; and Tian, Q. 2023. Accurate medium-range global weather forecasting with 3D neural networks. Nature, 619(7970): 533--538

    2023

  5. [5]

    H.; Ebert, B.; Fuentes, M.; Hamill, T

    Bougeault, P.; Toth, Z.; Bishop, C.; Brown, B.; Burridge, D.; Chen, D. H.; Ebert, B.; Fuentes, M.; Hamill, T. M.; Mylne, K.; et al. 2010. The THORPEX interactive grand global ensemble. Bulletin of the American Meteorological Society, 91(8): 1059--1072

    work page 2010

  6. [6]

    Bourdin, S.; Fromang, S.; Dulac, W.; Cattiaux, J.; and Chauvin, F. 2022. Intercomparison of four tropical cyclones detection algorithms on ERA5

    work page 2022

  7. [7]

    A.; Brown, A.; English, S.; Forbes, R.; Geer, A.; Haiden, T.; Leutbecher, M.; Magnusson, L.; Rodwell, M.; et al

    Buizza, R.; Balmaseda, M. A.; Brown, A.; English, S.; Forbes, R.; Geer, A.; Haiden, T.; Leutbecher, M.; Magnusson, L.; Rodwell, M.; et al. 2018. The development and evaluation process followed at ECMWF to upgrade the Integrated Forecasting System (IFS). European Centre for Medium Range Weather Forecasts

    work page 2018

  8. [8]

    Chen, G.; Zhang, X.; Bai, L.; and Wan, R. 2019. Verification of tropical cyclone operational forecast in 2018. Proceedings of the ESCAP/WMO Typhoon Committee, Guangzhou, China, 26

    work page 2019

Show all 47 references
  1. [9]

    Chen, K.; Han, T.; Gong, J.; Bai, L.; Ling, F.; Luo, J.-J.; Chen, X.; Ma, L.; Zhang, T.; Su, R.; et al. 2023 a . Fengwu: Pushing the skillful global medium-range weather forecast beyond 10 days lead. arXiv preprint arXiv:2304.02948

    2023

  2. [10]

    Chen, L.; Zhong, X.; Zhang, F.; Cheng, Y.; Xu, Y.; Qi, Y.; and Li, H. 2023 b . FuXi: A cascade machine learning forecasting system for 15-day global weather forecast. npj Climate and Atmospheric Science, 6(1): 190

    2023

  3. [11]

    Dosovitskiy, A.; Beyer, L.; Kolesnikov, A.; Weissenborn, D.; Zhai, X.; Unterthiner, T.; Dehghani, M.; Minderer, M.; Heigold, G.; Gelly, S.; et al. 2020. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929

    2020 arXiv

  4. [12]

    Emanuel, K. 2003. Tropical cyclones. Annual review of earth and planetary sciences, 31(1): 75--104

    2003

  5. [13]

    Emanuel, K. 2018. 100 years of progress in tropical cyclone research. Meteorological Monographs, 59(1): 15--1

    2018

  6. [14]

    Emanuel, K. A. 1986. An air-sea interaction theory for tropical cyclones. Part I: Steady-state maintenance. Journal of Atmospheric Sciences, 43(6): 585--605

    1986

  7. [15]

    Emanuel, K. A. 1987. The dependence of hurricane intensity on climate. Nature, 326(6112): 483--485

    1987

  8. [16]

    Gilford, D. M. 2021. pyPI (v1. 3): Tropical cyclone potential intensity calculations in Python. Geoscientific Model Development, 14(5): 2351--2369

    2021

  9. [17]

    Graves, A.; and Graves, A. 2012. Long short-term memory. Supervised sequence labelling with recurrent neural networks, 37--45

    2012

  10. [18]

    B.; Ferranti, L.; Prates, F.; and Richardson, D

    Haiden, T.; Janousek, M.; Vitart, F.; Bouall \`e gue, Z. B.; Ferranti, L.; Prates, F.; and Richardson, D. 2018. Evaluation of ECMWF forecasts, including the 2018 upgrade. European Centre for Medium Range Weather Forecasts Reading, UK

    2018

  11. [19]

    Han, T.; Guo, S.; Ling, F.; Chen, K.; Gong, J.; Luo, J.; Gu, J.; Dai, K.; Ouyang, W.; and Bai, L. 2024. Fengwu-ghr: Learning the kilometer-scale medium-range global weather forecasting. arXiv preprint arXiv:2402.00059

    2024 arXiv

  12. [20]

    Hersbach, H.; Bell, B.; Berrisford, P.; Hirahara, S.; Hor \'a nyi, A.; Mu \ n oz-Sabater, J.; Nicolas, J.; Peubey, C.; Radu, R.; Schepers, D.; et al. 2020. The ERA5 global reanalysis. Quarterly Journal of the Royal Meteorological Society, 146(730): 1999--2049

    2020

  13. [21]

    Hsu, K.; Liu, C.-C.; Peng, M.; Chen, D.-S.; Chang, P.-L.; Hsiao, L.-F.; Fong, C.-T.; Hong, J.-S.; Cheng, C.-P.; Lu, K.-C.; et al. 2024. Performance of Five Machine Learning-based Global Weather Prediction Models in the East Asia Region

    2024

  14. [22]

    Huang, C.; Bai, C.; Chan, S.; and Zhang, J. 2022. MMSTN: A Multi-Modal Spatial-Temporal Network for Tropical Cyclone Short-Term Prediction. Geophysical Research Letters, 49(4): e2021GL096898

    2022

  15. [23]

    Huang, C.; Bai, C.; Chan, S.; Zhang, J.; and Wu, Y. 2023. MGTCF: multi-generator tropical cyclone forecasting with heterogeneous meteorological data. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, 5096--5104

    2023

  16. [24]

    Jiang, W.; Zhang, D.; Hu, G.; Wu, T.; Liu, L.; Xiao, Y.; and Duan, Z. 2023. Transformer-based tropical cyclone track and intensity forecasting. Journal of Wind Engineering and Industrial Aerodynamics, 238: 105440

    2023

  17. [25]

    R.; Kruk, M

    Knapp, K. R.; Kruk, M. C.; Levinson, D. H.; Diamond, H. J.; and Neumann, C. J. 2010. The international best track archive for climate stewardship (IBTrACS) unifying tropical cyclone data. Bulletin of the American Meteorological Society, 91(3): 363--376

    2010

  18. [26]

    Krizhevsky, A.; Sutskever, I.; and Hinton, G. E. 2012. Imagenet classification with deep convolutional neural networks. Advances in neural information processing systems, 25

    2012

  19. [27]

    Lam, R.; Sanchez-Gonzalez, A.; Willson, M.; Wirnsberger, P.; Fortunato, M.; Alet, F.; Ravuri, S.; Ewalds, T.; Eaton-Rosen, Z.; Hu, W.; et al. 2023. Learning skillful medium-range global weather forecasting. Science, 382(6677): 1416--1421

    2023

  20. [28]

    Lin, T.-Y.; Doll \'a r, P.; Girshick, R.; He, K.; Hariharan, B.; and Belongie, S. 2017. Feature pyramid networks for object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition, 2117--2125

    2017

  21. [29]

    Magnusson, L.; Majumdar, S.; Emerton, R.; Richardson, D.; Alonso-Balmaseda, M.; Baugh, C.; Bechtold, P.; Bidlot, J.-R.; Bonanni, A.; Bonavita, M.; Bormann, N.; Brown, A.; Browne, P.; Carr, H.; Dahoui, M.; Chiara, G. D.; Diamantakis, M.; Duncan, D.; English, S.; Forbes, R.; Gee...

    2021

  22. [30]

    Malakar, P.; Kesarkar, A.; Bhate, J.; Singh, V.; and Deshamukhya, A. 2020. Comparison of reanalysis data sets to comprehend the evolution of tropical cyclones over North Indian Ocean. Earth and Space Science, 7(2): e2019EA000978

    2020

  23. [31]

    Meng, F.; Yang, K.; Yao, Y.; Wang, Z.; and Song, T. 2023. Tropical cyclone intensity probabilistic forecasting system based on deep learning. International Journal of Intelligent Systems, 2023(1): 3569538

    2023

  24. [32]

    Pan, B.; Xu, X.; and Shi, Z. 2019. Tropical cyclone intensity prediction based on recurrent neural networks. Electronics Letters, 55(7): 413--415

    2019

  25. [33]

    Rasul, K.; Bennett, A.; Vicente, P.; Gupta, U.; Ghonia, H.; Schneider, A.; and Nevmyvaka, Y. 2024. VQ - TR : Vector Quantized Attention for Time Series Forecasting. In The Twelfth International Conference on Learning Representations

    2024

  26. [34]

    N.; and Kim, K.-M

    Rasul, K.; Park, Y.-J.; Ramstr \"o m, M. N.; and Kim, K.-M. 2022. Vq-ar: Vector quantized autoregressive probabilistic time series forecasting. arXiv preprint arXiv:2205.15894

    2022 arXiv

  27. [35]

    R \"u ttgers, M.; Lee, S.; Jeon, S.; and You, D. 2019. Prediction of a typhoon track using a generative adversarial network and satellite images. Scientific reports, 9(1): 6057

    2019

  28. [36]

    Shen, X.; Su, Y.; Zhang, H.; and Hu, J. 2023. New version of the CMA-GFS dynamical core based on the predictor--corrector time integration scheme. Journal of Meteorological Research, 37(3): 273--285

    2023

  29. [37]

    Sohn, K.; Lee, H.; and Yan, X. 2015. Learning structured output representation using deep conditional generative models. Advances in neural information processing systems, 28

    2015

  30. [38]

    Van Den Oord, A.; Vinyals, O.; et al. 2017. Neural discrete representation learning. Advances in neural information processing systems, 30

    2017

  31. [39]

    Van der Maaten, L.; and Hinton, G. 2008. Visualizing data using t-SNE. Journal of machine learning research, 9(11)

    2008

  32. [40]

    N.; Kaiser, .; and Polosukhin, I

    Vaswani, A.; Shazeer, N.; Parmar, N.; Uszkoreit, J.; Jones, L.; Gomez, A. N.; Kaiser, .; and Polosukhin, I. 2017. Attention is all you need. Advances in neural information processing systems, 30

    2017

  33. [41]

    Wang, C.; Li, X.; and Zheng, G. 2024. Tropical cyclone intensity forecasting using model knowledge guided deep learning model. Environmental Research Letters, 19(2): 024006

    2024

  34. [42]

    Wang, X.; Chen, K.; Liu, L.; Han, T.; Li, B.; and Bai, L. 2024. Global Tropical Cyclone Intensity Forecasting with Multi-modal Multi-scale Causal Autoregressive Model. arXiv preprint arXiv:2402.13270

    2024 arXiv

  35. [43]

    D.; Irish, J

    Woodruff, J. D.; Irish, J. L.; and Camargo, S. J. 2013. Coastal flooding by tropical cyclones and sea-level rise. Nature, 504(7478): 44--52

    2013

  36. [44]

    M.; Yang, Z.; Hu, J.; Tang, X.; Mok, K

    Yang, J.; Li, L.; Zhao, K.; Wang, P.; Wang, D.; Sou, I. M.; Yang, Z.; Hu, J.; Tang, X.; Mok, K. M.; et al. 2019. A comparative study of Typhoon Hato (2017) and Typhoon Mangkhut (2018)—Their impacts on coastal inundation in Macau. Journal of Geophysical Research: Oceans, 124(12...

    2017

  37. [45]

    Zhang, X.; Fang, J.; and Yu, Z. 2023. The forecast skill of tropical cyclone genesis in two global ensembles. Weather and Forecasting, 38(1): 83--97

    2023

  38. [46]

    Zhang, Z.; Yang, X.; Shi, L.; Wang, B.; Du, Z.; Zhang, F.; and Liu, R. 2022. A neural network framework for fine-grained tropical cyclone intensity prediction. Knowledge-Based Systems, 241: 108195

    2022

  39. [47]

    Zhou, X.; and Juang, H.-M. H. 2023. A Model Instability Issue in the NCEP Global Forecast System Version 16 and Potential Solutions. EGUsphere, 2023: 1--22

    2023

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.