Research map / Image generation / Autoregressive and latent generation
SynWeather: Weather Observation Data Synthesis Across Multiple Regions and Variables via a General Diffusion Transformer
Kaiyi Xu, Junchao Gong, Zhiwang Zhou, Zhangrui Li, Yuandong Pu, Yihao Liu, Ben Fei, Fenghua Ling, Wenlong Zhang, Lei Bai
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use SynWeatherDiff instead of WeatherGFM | general model · CONUS CR synthesis | RMSE: 3.124 -> 2.820 | Improved |
| use SynWeatherDiff instead of WeatherGFM | general model · CONUS precipitation synthesis | CSI/15: 0.090 -> 0.113 | Improved |
| use SynWeatherDiff instead of UNet in pixel space | specialized model · East Asia visible light synthesis | SSIM: 0.917 -> 0.868 | Hurt |
| increase CONUS CR sampling ratio to 1/2 (ablation) | SynWeatherDiff · CONUS precipitation synthesis | CSI/15: 0.113 -> 0.137 | Improved |
| increase CONUS CR sampling ratio to 1/2 (ablation) | SynWeatherDiff · CONUS CR synthesis | RMSE: 2.82 -> 2.69 | Improved |
| increase CONUS precipitation sampling ratio to 1/2 (ablation) | SynWeatherDiff · CONUS CR synthesis | CSI/35: 0.158 -> 0.110 | Hurt |
| increase Europe visible light sampling ratio to 1/2 (ablation) | SynWeatherDiff · Europe visible light synthesis | SSIM: 0.864 -> 0.877 | Improved |
| increase MWBT sampling ratio to 1/2 (ablation) | SynWeatherDiff · Europe visible light synthesis | SSIM: 0.864 -> 0.879 | Improved |
| remove any input infrared channel group (SWIR, WV, LWIR, GAS) (ablation) | SynWeatherDiff · six synthesis tasks | task metrics: removing any group worsens all synthesis tasks | Improved |
| use general text-prompted model instead of CONUS-trained specialized model | SynWeatherDiff · East Asia precipitation (OOD) | RMSE: 1.383 -> 1.150 | Improved |
| use SynWeatherDiff instead of UNet and ViT in few-shot fine-tuning | SynWeatherDiff · East Asia precipitation few-shot | SSIM: 0.791 -> 0.828 | Improved |
Closest papers by meaning
- Satellite Observations Guided Diffusion Model for Accurate Meteorological States at Arbitrary Resolution (CVPR 2025 · similarity 0.86)
- TC-Diffuser: Bi-Condition Multi-Modal Diffusion for Tropical Cyclone Forecasting (AAAI 2025 · similarity 0.85)
- DiffDA: a Diffusion model for weather-scale Data Assimilation (ICML 2024 · similarity 0.85)
- Scaling Storm-Resolving Atmospheric AI Simulation to the Entire Planet (NeurIPS 2026 · similarity 0.85)
- WeatherGFM: Learning a Weather Generalist Foundation Model via In-context Learning (ICLR 2025 · similarity 0.84)
- DAWP: A framework for global observation forecasting via Data Assimilation and Weather Prediction in satellite observation space (NeurIPS 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
