FlowDA: Accurate, Low-Latency Weather Data Assimilation via Flow Matching

2026-02-06

AI Summary

This work proposes FlowDA, the first framework to integrate flow matching into weather-scale data assimilation, addressing the high computational cost of traditional variational methods and the instability of existing generative approaches in long-horizon autoregressive forecasting due to excessive sampling steps and error accumulation. By embedding sparse observations via SetConv and fine-tuning the Aurora foundation model, FlowDA drastically reduces the number of required sampling steps while enhancing the stability and noise robustness of recurrent assimilation. Evaluated under diverse settings with observation rates as low as 0.1%, FlowDA consistently outperforms strong baselines of comparable parameter count, demonstrating superior accuracy, robustness, and long-term temporal consistency.

Technology Category

Machine Learning: Time-Series/Data StreamsPlanning, Routing, and Scheduling: Planning with Language ModelsIntelligent Robots: State Estimation

Application Category

Search and Retrieval-Augmented AI: Vertical and domain-specific searchGraph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphsSystems and Infrastructure for Web, Mobile and WoT: Data management and stream processing for Web, mobile and wireless applications

Abstract

Data assimilation (DA) is a fundamental component of modern weather prediction, yet it remains a major computational bottleneck in machine learning (ML)-based forecasting pipelines due to reliance on traditional variational methods. Recent generative ML-based DA methods offer a promising alternative but typically require many sampling steps and suffer from error accumulation under long-horizon auto-regressive rollouts with cycling assimilation. We propose FlowDA, a low-latency weather-scale generative DA framework based on flow matching. FlowDA conditions on observations through a SetConv-based embedding and fine-tunes the Aurora foundation model to deliver accurate, efficient, and robust analyses. Experiments across observation rates decreasing from $3.9\%$ to $0.1\%$ demonstrate superior performance of FlowDA over strong baselines with similar tunable-parameter size. FlowDA further shows robustness to observational noise and stable performance in long-horizon auto-regressive cycling DA. Overall, FlowDA points to an efficient and scalable direction for data-driven DA.

Problem

• data assimilation
• weather prediction
• machine learning
• generative models
• computational bottleneck

Innovation

• flow matching
• data assimilation
• SetConv
• low-latency
• generative modeling