Reproduce “DAWP: A framework for global observation forecasting via Data Assimilation and Weather Prediction in satellite observation space” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Junchao Gong, Jingyi Xu, Ben Fei, Fenghua Ling, Wenlong Zhang, Kun Chen, Wanghan Xu, Weidong Yang and 2 more
The DAWP framework enables global observation forecasting by directly processing multi-modal satellite data within a uniform observation space through an AI-driven data assimilation (AIDA) module and a cross-regional conditioned weather prediction model. It achieves a 12-hour lead time advantage over baselines and provides more accurate global precipitation forecasts with improved Critical Success Index (CSI) and False Alarm Ratio (FAR).
Key insights
- AI-driven data assimilation (AIDA) is critical for creating a stable, complete observation space from irregular, sparse inputs, significantly improving AIWP training and multi-step rollout performance.
- Integrating cross-regional boundary conditions (CBC) through a global state cache is essential for maintaining spatial continuity and accuracy in sub-image-based global forecasts.
- Direct operation in a uniform observation space reduces reliance on traditional reanalysis products, offering a path towards more immediate and less biased weather predictions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.