Research map / Time series / Long-term forecasting
DuoCast: Duo-Probabilistic Diffusion for Precipitation Nowcasting
Penghui Wen, Mengwei He, Patrick Filippi, Na Zhao, Feng Zhang, Thomas Francis Bishop, Zhiyong Wang, Kun Hu
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 high-frequency diffusion model on top of low-frequency model (ablation) | DuoCast · SEVIR | CSI-M: 0.327 -> 0.338 | Improved |
| add low-frequency model to high-frequency-only model (ablation) | DuoCast · SEVIR | CSI-M: 0.310 -> 0.338 | Improved |
| use convolution instead of self-attention for low-frequency backbone (ablation) | DuoCast low-frequency model · SEVIR | CSI-M: 0.310 -> 0.327 | Improved |
| add thresholded high-intensity forecast Y-dagger (ablation) | DuoCast · SEVIR | CSI-219: 0.048 -> 0.107 | Improved |
| add air mass spatial modeling (ablation) | DuoCast · SEVIR | CSI-M: 0.309 -> 0.338 | Improved |
| add weather front temporal modeling (ablation) | DuoCast · SEVIR | CSI-M: 0.311 -> 0.338 | Improved |
| add Conv_I (large-kernel depthwise) in air mass module (ablation) | DuoCast · SEVIR | CSI-M: 0.316 -> 0.338 | Improved |
| add leading frame A_1 as temporal anchor (ablation) | DuoCast front temporal attention · SEVIR | CSI-M: 0.320 -> 0.327 | Improved |
| use loss weights (0.1, 0.1, 0.5) instead of (0.5, 0.1, 0.1) (ablation) | DuoCast · SEVIR | CSI: 0.3255 -> 0.3375 | Improved |
| use DuoCast instead of DiffCast | precipitation nowcasting · SEVIR | inference time per sample (s): 3.67 -> 0.56 | Improved |
| use DuoCast instead of AlphaPre | precipitation nowcasting · CIKM | CSI-M: 0.3194 -> 0.3179 | No clear effect |
Closest papers by meaning
- PostCast: Generalizable Postprocessing for Precipitation Nowcasting via Unsupervised Blurriness Modeling (ICLR 2025 · similarity 0.89)
- CasCast: Skillful High-resolution Precipitation Nowcasting via Cascaded Modelling (ICML 2024 · similarity 0.89)
- DiffCast: A Unified Framework via Residual Diffusion for Precipitation Nowcasting (CVPR 2024 · similarity 0.88)
- Perceptually Constrained Precipitation Nowcasting Model (ICML 2025 · similarity 0.87)
- Fourier Amplitude and Correlation Loss: Beyond Using L2 Loss for Skillful Precipitation Nowcasting (NeurIPS 2024 · similarity 0.86)
- OmniCast: A Masked Latent Diffusion Model for Weather Forecasting Across Time Scales (NeurIPS 2025 · similarity 0.86)
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.
