Research map / Probabilistic and scientific modeling / Variational and generative sampling
Spatiotemporal Pyramid Flow Matching for Climate Emulation
Jeremy A. Irvin, Jiaqi Han, Zikui Wang, Abdulaziz Alharbi, Yufei Zhao, Nomin-Erdene Bayarsaikhan, Daniele Visioni, Andrew Y. Ng, Duncan Watson-Parris
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 |
|---|---|---|---|
| SPF vs autoregressive Pyramidal Flow at yearly scale (200M) | SPF · ClimateBench yearly | CRPS: 0.327 -> 0.222 | Improved |
| parallel SPF instead of autoregressive Pyramidal Flow, runtime | SPF · ClimateBench yearly 10-year sampling | runtime (s): 19 -> 3 | Improved |
| SPF vs PixelFlow at yearly RMSE (100M) | SPF · ClimateBench yearly | RMSE: 0.549 -> 0.565 | Hurt |
| SPF vs Multi-Monthly Flow at 200M, monthly | SPF · ClimateBench monthly | CRPS: 0.443 -> 0.453 | Hurt |
| use 3-stage pyramid instead of 5-stage variants (ablation) | SPF · ClimateBench monthly | CRPS: 0.474 -> 0.453 | Improved |
| alternate temporal and spatial transitions (DYYMM vs DYMMM) (ablation) | SPF 5-stage · ClimateBench monthly | CRPS: 0.474 -> 0.463 | Improved |
| add multi-timescale training (DYM-Any vs DYM-Monthly) (ablation) | SPF · ClimateBench monthly | RMSE: 1.064 -> 1.060 | Improved |
| scale to 600M parameters without pretraining (ablation) | SPF · ClimateBench yearly | CRPS: 0.222 -> 0.229 | Hurt |
| scale 200M to 600M (ablation) | SPF · ClimateBench monthly | CRPS: 0.453 -> 0.442 | Improved |
| add ClimateSuite pretraining (ablation) | SPF 600M · ClimateBench yearly | CRPS: 0.229 -> 0.216 | Improved |
| add ClimateSuite pretraining (ablation) | SPF 600M · ClimateBench monthly | CRPS: 0.442 -> 0.432 | Improved |
| SPF vs UNet multi-model emulation | SPF 600M · ClimateSuite held-out SAI UKESM1-0-LL | CRPS: 0.466 -> 0.315 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ClimaX — ClimaX: A foundation model for weather and climate (ICML 2023) · better on 2 of 2 shared results
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
- Scaling Storm-Resolving Atmospheric AI Simulation to the Entire Planet (NeurIPS 2026 · similarity 0.86)
- Probablistic Emulation of a Global Climate Model with Spherical DYffusion (NeurIPS 2024 · similarity 0.86)
- Terra: A Multimodal Spatio-Temporal Dataset Spanning the Earth (NeurIPS 2024 · similarity 0.86)
- GeoCore-9B: Towards Geo-Aware Generative Foundation Models in Earth Observation (NeurIPS 2026 · similarity 0.86)
- PrismFlow: Residual Dynamics for Flow Matching in Time-Series Generation (NeurIPS 2026 · similarity 0.86)
- Causal Deciphering and Inpainting in Spatio-Temporal Dynamics via Diffusion Model (NeurIPS 2024 · 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.
