Spatiotemporal Pyramid Flow Matching for Climate Emulation

2025-12-01

AI Summary

Traditional autoregressive methods suffer from computational inefficiency and numerical instability in long-horizon, non-stationary climate simulation under time-varying forcings. To address this, we propose Spatiotemporal Pyramid Flows (SPF), the first framework to introduce hierarchical flow matching into climate modeling. SPF employs a cascaded architecture that jointly decomposes spatiotemporal dynamics across multiple scales and couples them with physical forcings, enabling parallel generation and stable rollout across temporal granularities (yearly, monthly, daily). Crucially, it supports direct sampling from arbitrary initial conditions without iterative autoregression. Trained on ClimateSuite and evaluated on ClimateBench, SPF consistently outperforms strong baselines: it achieves superior accuracy at yearly and monthly scales, accelerates inference by 1–2 orders of magnitude, and demonstrates exceptional generalization to unseen climate scenarios and Earth system models.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Generative models have the potential to transform the way we emulate Earth's changing climate. Previous generative approaches rely on weather-scale autoregression for climate emulation, but this is inherently slow for long climate horizons and has yet to demonstrate stable rollouts under nonstationary forcings. Here, we introduce Spatiotemporal Pyramid Flows (SPF), a new class of flow matching approaches that model data hierarchically across spatial and temporal scales. Inspired by cascaded video models, SPF partitions the generative trajectory into a spatiotemporal pyramid, progressively increasing spatial resolution to reduce computation and coupling each stage with an associated timescale to enable direct sampling at any temporal level in the pyramid. This design, together with conditioning each stage on prescribed physical forcings (e.g., greenhouse gases or aerosols), enables efficient, parallel climate emulation at multiple timescales. On ClimateBench, SPF outperforms strong flow matching baselines and pre-trained models at yearly and monthly timescales while offering fast sampling, especially at coarser temporal levels. To scale SPF, we curate ClimateSuite, the largest collection of Earth system simulations to date, comprising over 33,000 simulation-years across ten climate models and the first dataset to include simulations of climate interventions. We find that the scaled SPF model demonstrates good generalization to held-out scenarios across climate models. Together, SPF and ClimateSuite provide a foundation for accurate, efficient, probabilistic climate emulation across temporal scales and realistic future scenarios. Data and code is publicly available at https://github.com/stanfordmlgroup/spf .

Problem

• Efficiently emulate Earth's climate across multiple timescales.
• Overcome slow autoregressive methods for long climate horizons.
• Enable stable rollouts under nonstationary physical forcings.

Innovation

• Hierarchical flow matching across spatiotemporal scales
• Direct sampling at any temporal level via pyramid structure
• Conditioning on physical forcings for efficient parallel emulation