Research map / Time series / Long-term forecasting
Extreme Weather Nowcasting via Local Precipitation Pattern Prediction
Chang hoon Song, Teng Yuan Chang, Youngjoon Hong
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 CDU decoder instead of pixel shuffle (ablation) | exPreCast · KMA (1h) | CSI-M POOL16: 0.4632 -> 0.4841 | Improved |
| use CDU instead of trilinear interpolation (ablation) | exPreCast · KMA (1h) | CSI-M POOL1: 0.2635 -> 0.2794 | Improved |
| use CDU instead of bicubic interpolation (ablation) | exPreCast · KMA (1h) | CSI-M POOL16: 0.4814 -> 0.4841 | Improved |
| use CDU instead of pixel shuffle on extreme threshold without pooling (ablation) | exPreCast · KMA (1h) | CSI-80 POOL1: 0.0372 -> 0.0348 | Hurt |
| fine-tune decoder and TE from frozen 1h encoder instead of training 6h model from scratch (ablation) | exPreCast · KMA (6h) | CSI-M POOL1: 0.0879 -> 0.1214 | Improved |
| exPreCast vs CasCast on KMA with 30x fewer FLOPs | exPreCast · KMA (1h) | CSI-M POOL4: 0.3742 -> 0.3721 | No clear effect |
| exPreCast vs CasCast on SEVIR | exPreCast · SEVIR (1h) | CSI-M POOL16: 0.5525 -> 0.5427 | Hurt |
| exPreCast vs ConvLSTM on MeteoNet without pooling | exPreCast · MeteoNet (1h) | CSI-M POOL1: 0.3076 -> 0.2861 | Hurt |
| exPreCast vs ConvLSTM on MeteoNet with pooling 16 | exPreCast · MeteoNet (1h) | CSI-M POOL16: 0.2683 -> 0.4446 | Improved |
| exPreCast vs SimVP on KMA HSS | exPreCast · KMA (1h) | HSS: 0.3958 -> 0.4042 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- EarthFormer — Earthformer: Exploring Space-Time Transformers for Earth System Forecasting (NeurIPS 2022) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- CasCast: Skillful High-resolution Precipitation Nowcasting via Cascaded Modelling (ICML 2024 · similarity 0.91)
- DiffCast: A Unified Framework via Residual Diffusion for Precipitation Nowcasting (CVPR 2024 · similarity 0.90)
- PostCast: Generalizable Postprocessing for Precipitation Nowcasting via Unsupervised Blurriness Modeling (ICLR 2025 · similarity 0.90)
- Perceptually Constrained Precipitation Nowcasting Model (ICML 2025 · similarity 0.89)
- Integrating Multi-Source Data for Long Sequence Precipitation Forecasting (AAAI 2025 · similarity 0.89)
- AlphaPre: Amplitude-Phase Disentanglement Model for Precipitation Nowcasting (CVPR 2025 · similarity 0.87)
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.
