Xaurora: Generative Weather Forecasting with Denoising Stochastic Interpolants from a Foundation Model Prior
Abstract
Deep learning has revolutionised weather forecasting in recent years, especially through atmospheric foundation models, which offer competitive skill for a fraction of the computational costs of classic physics-based models. However, most existing foundation models are deterministic, limiting the generation of large ensembles for accurate uncertainty quantification, extreme weather risk assessment, and long-range weather forecasting. Furthermore, these models incur a large, often prohibitive, computational overhead to train from scratch. To address these shortcomings, we turn a pretrained deterministic prior model, namely the Aurora foundation model, into a generative ensemble-prediction model. To that end, we introduce a novel generative method, Denoising Stochastic Interpolants, combined with a replay buffer for Stochastic Differential Equation (SDE) rollout, enabling probabilistic training of SDE trajectories. Our stochastic foundation model, Xaurora, is finetuned from the small Aurora version, yet it approaches the state-of-the-art on global ensemble metrics and is competitive with the large version of Aurora. Our method is parameter and sample efficient, and generates skilful 15-day forecasts in 13 minutes. Our results demonstrate that deterministic foundation models can be efficiently extended into even stronger stochastic models.
Related
- A Review of Deep-learning-based Seismic Data Denoising and Its Promising Paradigm Shift to Foundation Models
- Parameter-Efficient Adaptation of Pre-Trained Vision Foundation Models for Active and Passive Seismic Data Denoising
- Cost-aware evaluation of time-series foundation models for urban air-quality and temperature forecasting
- Dynamical-Model-Informed Stochastic Flow Matching for Global Probabilistic Subseasonal Forecasting of Means and Extremes
- A transparent constrained stochastic framework for urban expansion forecasting and future service-network screening: Cape Town demonstration using annual building evidence, empirical land constraints, spatial hindcasting, and multi-horizon ensembles
BibTeX
@misc{walt2026xaurora,
title = {{Xaurora: Generative Weather Forecasting with Denoising Stochastic Interpolants from a Foundation Model Prior}},
author = {Eliot Walt and Miltiadis Kofinas and Nikolaj Mücke and Efstratios Gavves and Dim Coumou},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.06509},
eprint = {2610.06509},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.06509},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Xaurora: Generative Weather Forecasting with Denoising Stochastic Interpolants from a Foundation Model Prior
AU - Eliot Walt
AU - Miltiadis Kofinas
AU - Nikolaj Mücke
AU - Efstratios Gavves
AU - Dim Coumou
PY - 2026
DA - 2026/10/05
PB - arXiv
DO - 10.48550/arxiv.2610.06509
UR - https://arxiv.org/abs/2610.06509
AB - Deep learning has revolutionised weather forecasting in recent years, especially through atmospheric foundation models, which offer competitive skill for a fraction of the computational costs of classic physics-based models. However, most existing foundation models are deterministic, limiting the generation of large ensembles for accurate uncertainty quantification, extreme weather risk assessment, and long-range weather forecasting. Furthermore, these models incur a large, often prohibitive, computational overhead to train from scratch. To address these shortcomings, we turn a pretrained deterministic prior model, namely the Aurora foundation model, into a generative ensemble-prediction model. To that end, we introduce a novel generative method, Denoising Stochastic Interpolants, combined with a replay buffer for Stochastic Differential Equation (SDE) rollout, enabling probabilistic training of SDE trajectories. Our stochastic foundation model, Xaurora, is finetuned from the small Aurora version, yet it approaches the state-of-the-art on global ensemble metrics and is competitive with the large version of Aurora. Our method is parameter and sample efficient, and generates skilful 15-day forecasts in 13 minutes. Our results demonstrate that deterministic foundation models can be efficiently extended into even stronger stochastic models.
ER -