- Rating
- 3(+1)
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
Marchuk-S2S is a compact latent flow-matching model that generates multi-day chunks of six-hourly weather states from a single initial condition and time-of-year labels. A single network is trained on variable horizons and composed autoregressively to produce long subseasonal ensembles. The main contributions are showing that training on six-hourly states improves daily forecast skill compared with training on daily means, that composing shorter chunks can outperform direct longer-horizon generation, and that the resulting system is competitive with strong operational and learned subseasonal baselines while remaining computationally efficient.
Critical strengths
-
The model produces long six-hourly forecasts that match strong baselines (ECMWF S2S and FuXi-S2S) at much lower computational cost. This makes high-resolution subseasonal ensembles more practical.
-
Training on six-hourly states and then averaging gives better daily CRPS than training directly on daily means. This is a clear and useful finding.
Critical weaknesses
-
The core components (latent-space generation, flow matching, multi-step/chunk generation, and autoregressive rollouts) are largely combinations of existing ideas. The paper’s main advance is system integration and efficiency rather than a new modeling principle. For a top-tier venue, this level of novelty is borderline.
-
Results are mostly competitive rather than clearly superior, while precipitation skill and late-lead MJO skill, both highly relevant for subseasonal use cases, are noticeably worse than the baselines. Without stronger overall gains or clear progress on these important targets, the practical impact is limited.
Questions for authors
-
Table 2 shows that composing two shorter chunks yields lower CRPS than generating one longer chunk directly, even without a consistency loss. Do you have any analysis or hypothesis for why the direct longer-horizon generation underperforms?
-
The model shows clearly weaker precipitation anomaly correlation than both ECMWF and FuXi-S2S from week 3 onward. Is this mainly caused by information loss in the frozen latent autoencoder, by the flow-matching objective struggling with precipitation’s intermittent distribution, or by something else?
2 comments
Author Responses (2/2)
Q1
We hypothesize that the result reflects a trade-off between autoregressive error accumulation and single-transition difficulty. Direct long-horizon generation must represent a broader and more nonlinear conditional distribution, whereas composition factorizes it into shorter, better-learned transitions. Conversely, very short chunks require more transitions, which can accumulate errors and introduce additional noise realizations.
To examine this trade-off, we evaluated 2-, 4-, and 8-day inference chunks over full rollouts. The 4-day chunk achieves the lowest CRPS in 8 of the 10 variable–lead-time comparisons. The 2-day chunk produces the largest ensemble spread in 9 cases, and the 8-day configuration is optimal under neither metric.
CRPS ↓
| Inference chunk | T2M, 15 | T2M, 45 | T500, 15 | T500, 45 | T850, 15 | T850, 45 | Z500, 15 | Z500, 45 | Z850, 15 | Z850, 45 |
|---|---|---|---|---|---|---|---|---|---|---|
| 2 days | 1.039 | 1.077 | 1.378 | 1.489 | 1.511 | 1.631 | 324.84 | 349.97 | 236.54 | 238.11 |
| 4 days (ours) | 0.991 | 1.048 | 1.360 | 1.449 | 1.500 | 1.591 | 326.98 | 348.58 | 227.04 | 238.97 |
| 8 days | 1.003 | 1.097 | 1.413 | 1.495 | 1.563 | 1.654 | 342.44 | 357.96 | 233.85 | 247.22 |
Spread ↑
| Inference chunk | T2M, 15 | T2M, 45 | T500, 15 | T500, 45 | T850, 15 | T850, 45 | Z500, 15 | Z500, 45 | Z850, 15 | Z850, 45 |
|---|---|---|---|---|---|---|---|---|---|---|
| 2 days | 1.815 | 1.813 | 2.739 | 2.963 | 2.949 | 3.152 | 668.07 | 713.05 | 449.50 | 493.42 |
| 4 days (ours) | 1.726 | 1.879 | 2.616 | 2.822 | 2.850 | 3.065 | 629.53 | 682.09 | 436.84 | 470.45 |
| 8 days | 1.570 | 1.750 | 2.426 | 2.748 | 2.798 | 2.784 | 589.74 | 664.45 | 400.31 | 443.23 |
Short chunks require more model transitions, so errors can accumulate [1, 2]. Long chunks require the model to represent a broader and more nonlinear conditional distribution in one step. The 4-day chunk is the best balance between these effects. The larger spread of the 2-day rollout is consistent with more frequent stochastic sampling: each transition draws an additional noise realization [3]. CRPS jointly evaluates accuracy, calibration, and sharpness [4], so the wider 2-day ensemble is not additional probabilistic skill. The 4-day chunk gives the best overall probabilistic accuracy.
CRPS and ensemble spread by inference chunk horizon. The 4-day inference chunk gives the best overall CRPS, whereas shorter chunks generally produce larger ensemble spread. Columns are lead times in days. Bold marks the best value in each column (lowest CRPS, largest spread).
[1] Pangu-Weather. Bi, Kaifeng, Lingxi Xie, Hengheng Zhang, Xin Chen, Xiaotao Gu, and Qi Tian. Accurate medium-range global weather forecasting with 3D neural networks. Nature, 2023.
[2] Nguyen, Tung, Rohan Shah, Hritik Bansal, Troy Arcomano, Romit Maulik, Veerabhadra Kotamarthi, Ian Foster, Sandeep Madireddy, and Aditya Grover. Scaling transformer neural networks for skillful and reliable medium-range weather forecasting. NeurIPS, 2024.
[3] Price, Ilan, Alvaro Sanchez-Gonzalez, Ferran Alet, Tom R. Andersson, Andrew El-Kadi, Dominic Masters, Timo Ewalds, Jacklynn Stott, Shakir Mohamed, Peter Battaglia, Remi Lam, and Matthew Willson. Probabilistic weather forecasting with machine learning. Nature, 2025.
[4] Gneiting, Tilmann and Adrian E. Raftery. Strictly proper scoring rules, prediction, and estimation. Journal of the American Statistical Association, 2007.
Q2
The weaker precipitation anomaly correlation from week 3 onward is a property of the frozen LaDCast representation. The flow-matching objective is applied only after that representation is fixed. The reconstruction error below is the encode–decode error of the frozen DC-AE on 6-hour precipitation, so it does not depend on the flow model or on lead time.
| Lead time | DC-AE RMSE, mm | Marchuk-S2S RMSE, mm | DC-AE / Marchuk-S2S |
|---|---|---|---|
| 1 day | 0.202 | 0.319 | 63% |
| 5 days | 0.202 | 0.463 | 44% |
| 10 days | 0.202 | 0.508 | 40% |
| 15 days | 0.202 | 0.519 | 39% |
| 30 days | 0.202 | 0.526 | 38% |
| 45 days | 0.202 | 0.530 | 38% |
The frozen DC-AE reconstructs 6-hour precipitation with an RMSE of 0.202 mm at every lead. At 1.5° on 2018, that reconstruction is 63% of the Marchuk-S2S precipitation RMSE at day 1, 44% at day 5, 39% at day 15, and 38% on days 30–45. The decoded forecast therefore starts from a distorted precipitation field, and week-3 anomaly correlation is computed from fields that still pass through the same decoder.
The last column is a ratio of RMSE values, . Independent errors add in the square, so if the reconstruction error were orthogonal to the remaining forecast error, removing it would lower the forecast RMSE by \sqrt[\]{1 - r^2}: about 23% at day 1 and about 8% from day 15 onward.
Author Responses (1/2)
We thank the reviewer for the helpful suggestions. Our point-by-point responses follow.
W1
We agree that the individual components are established. However, dense six-hourly probabilistic forecasting at subseasonal lead times remains underexplored. Marchuk-S2S uniquely combines latent flow matching, variable-horizon joint chunk generation, and autoregressive composition to produce stable 48-day forecasts using a single checkpoint, without lead-time-specific models or gridded climatology. This integration enables efficient dense S2S ensembles and reveals that six-hourly training improves daily forecast skill. In the following, we provide further analysis of these design choices.
W2
On most fields Marchuk-S2S is competitive with FuXi-S2S and the ECMWF S2S ensemble, while precipitation anomaly skill and MJO skill are weaker. This is the question posed in Section 1: whether one compact latent flow-matching model, trained as a single variable-horizon checkpoint on multi-day chunks of six-hourly states and composed to 48 days, can remain competitive with dedicated S2S systems. The experiments show that this architecture, this training pipeline, and six-hourly forecasting at subseasonal leads work well on the variables represented in the latent space. The corresponding scores, significance tests, six-hourly training comparison, and inference cost are reported in Figures 1–2, Table 2, and Section 3.6.
The precipitation and late-lead MJO gaps come from the frozen LaDCast latent space, which does not include outgoing longwave radiation and has limited precipitation reconstruction fidelity. Appendix A.7 specifies the shared precipitation-based RMM-like index and the matched comparison. Replacing OLR with precipitation is established practice for ML weather models [1], and it is noisier, so the absolute COR values lie below the OLR-based correlations reported by FuXi-S2S. The frozen DC-AE already introduces substantial precipitation error before any forecast is made; our response to Q2 shows how this error enters the predicted field. These scores can be improved by an autoencoder with higher precipitation fidelity and with OLR. Improving the autoencoder is beyond the scope of this work.
[1] Chien, M.-T., Barnes, E. A., and Maloney, E. D. Modulation of tropical cyclogenesis on subseasonal-to-interannual timescales in the deep-learning climate emulator ACE2, 2025.