Xiang et al. (2025) ADAF: An Artificial Intelligence Data Assimilation Framework for Weather Forecasting

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Journal of Advances in Modeling Earth Systems
• Year: 2025
• Date: 2025-09-01
• Authors: Yanfei Xiang, Weixin Jin, Haiyu Dong, Jonathan A. Weyn, Mingliang Bai, Zuliang Fang, Pengcheng Zhao, Hongyu Sun, Kit Thambiratnam, Qi Zhang, Xiaomeng Huang
• DOI: 10.1029/2024ms004839

Research Groups

Not specified in the provided text.

Short Summary

The study introduces an AI-based data assimilation framework (ADAF) to generate high-quality kilometer-scale analysis fields, significantly reducing computational costs while improving short-term weather forecasts.

Objective

• To develop and verify an artificial intelligence-based data assimilation framework (ADAF) capable of producing high-resolution analysis from multi-source observations to overcome the computational bottlenecks of traditional data assimilation.

Study Configuration

• Spatial Scale: Contiguous United States (CONUS), kilometer-scale.
• Temporal Scale: 3-hour assimilation window; 0–6 hour short-term forecast horizon.

Methodology and Data

• Models used: Artificial Intelligence-based Data Assimilation Framework (ADAF), AI-based weather prediction model.
• Data sources: Multi-source real-world observations, including sparse surface weather observations and satellite imagery.

Main Results

• ADAF generates analysis fields that align closely with actual observations and can accurately reconstruct extreme events, such as tropical cyclone wind fields.
• The framework is robust, maintaining high-quality analysis even when provided with low-accuracy backgrounds or extremely sparse surface observations.
• Computational efficiency is significantly improved, with processing taking approximately 2 s on an AMD MI200 GPU.
• Forecasts initialized with ADAF outperform those initialized with HRRRDAS for the 0–6 hr window.

Contributions

• This is a pioneering work that verifies the efficacy of an AI-based data assimilation method using diverse, real-world multi-source observations for kilometer-scale analysis, demonstrating its potential for operational weather forecasting.

Funding

Not specified in the provided text.

Citation

@article{Xiang2025ADAF,
  author = {Xiang, Yanfei and Jin, Weixin and Dong, Haiyu and Weyn, Jonathan A. and Bai, Mingliang and Fang, Zuliang and Zhao, Pengcheng and Sun, Hongyu and Thambiratnam, Kit and Zhang, Qi and Huang, Xiaomeng},
  title = {ADAF: An Artificial Intelligence Data Assimilation Framework for Weather Forecasting},
  journal = {Journal of Advances in Modeling Earth Systems},
  year = {2025},
  doi = {10.1029/2024ms004839},
  url = {https://doi.org/10.1029/2024ms004839}
}