Research map / Video / Video segmentation and tracking
WIND: Weather Inverse Diffusion for Zero-Shot Atmospheric Modeling
Michael Aich, Andreas Fürst, Florian Sestak, Carlos Ruiz-Gonzalez, Niklas Boers, Johannes Brandstetter
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 |
|---|---|---|---|
| add MMPS thermodynamic (global-mean) guidance to a warm-perturbed initial state (ablation) | WIND · Storm Bernd +2K scenario | mean peak precipitation intensification: +7.0% -> +13.9% | Improved |
| enforce global dry air mass conservation via MMPS guidance (ablation) | WIND · 14-day probabilistic forecast, CRPS | CRPS | No clear effect |
| enforce global dry air mass conservation via MMPS guidance (ablation) | WIND · 4-year rollout | global dry air mass stability | Improved |
| enforce dry air mass conservation (ablation) | WIND · precipitation forecast skill | CRPS (TP6h) | No clear effect |
| train with diffusion forcing (independent per-frame noise) instead of full-sequence diffusion AR-UViT (ablation) | WIND · 20-year unconstrained rollout | physical consistency | Improved |
| use diffusion forcing instead of autoregressive AR-UViT for forecasting | WIND · 14-day probabilistic forecast, CRPS | CRPS | Improved |
| zero-shot WIND instead of task-specific UViT for spatial downscaling | WIND · 4x spatial downscaling ERA5 2021 | RMSE | Hurt |
| zero-shot WIND instead of task-specific UViT for spatial downscaling | WIND · 4x spatial downscaling ERA5 2021 | high-frequency power spectrum | Improved |
| zero-shot WIND instead of task-specific UViT for sparse reconstruction (1% observed) | WIND · sparse reconstruction | RMSE | Improved |
| zero-shot WIND instead of specialized UViT for temporal downscaling | WIND · daily mean to 6-hourly total precipitation | sub-daily skill | Mixed |
Closest papers by meaning
- DiffDA: a Diffusion model for weather-scale Data Assimilation (ICML 2024 · similarity 0.86)
- Genuine Knowledge from Practice: Diffusion Test-Time Adaptation for Video Adverse Weather Removal (CVPR 2024 · similarity 0.86)
- WeatherGFM: Learning a Weather Generalist Foundation Model via In-context Learning (ICLR 2025 · similarity 0.84)
- OmniCast: A Masked Latent Diffusion Model for Weather Forecasting Across Time Scales (NeurIPS 2025 · similarity 0.84)
- Satellite Observations Guided Diffusion Model for Accurate Meteorological States at Arbitrary Resolution (CVPR 2025 · similarity 0.84)
- Predict-Project-Renoise: Sampling Diffusion Models under Hard Constraints (NeurIPS 2026 · 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.
