DAWP: A framework for global observation forecasting via Data Assimilation and Weather Prediction in satellite observation space

2025-10-12

AI Summary

Current AI-based weather forecasting heavily relies on reanalysis data, suffering from assimilation biases and temporal inconsistency, while struggling to model spatiotemporal dynamics of cross-platform, irregular, and high-resolution satellite observations. To address these limitations, we propose DAWP—a novel framework that unifies data assimilation and end-to-end prediction directly in observation space for the first time. DAWP employs a masked ViT-VAE encoder to process multi-source satellite observation tokens and constructs a multimodal autoencoder for observation-driven data assimilation. It further integrates a spatiotemporally decoupled Transformer with a cross-regional boundary-condition mechanism to enable sub-image-level global forecasting. Experiments demonstrate significant improvements in rollout accuracy and inference efficiency; notably, DAWP achieves superior performance and practical potential in global precipitation forecasting.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Weather prediction is a critical task for human society, where impressive progress has been made by training artificial intelligence weather prediction (AIWP) methods with reanalysis data. However, reliance on reanalysis data limits the AIWPs with shortcomings, including data assimilation biases and temporal discrepancies. To liberate AIWPs from the reanalysis data, observation forecasting emerges as a transformative paradigm for weather prediction. One of the key challenges in observation forecasting is learning spatiotemporal dynamics across disparate measurement systems with irregular high-resolution observation data, which constrains the design and prediction of AIWPs. To this end, we propose our DAWP as an innovative framework to enable AIWPs to operate in a complete observation space by initialization with an artificial intelligence data assimilation (AIDA) module. Specifically, our AIDA module applies a mask multi-modality autoencoder(MMAE)for assimilating irregular satellite observation tokens encoded by mask ViT-VAEs. For AIWP, we introduce a spatiotemporal decoupling transformer with cross-regional boundary conditioning (CBC), learning the dynamics in observation space, to enable sub-image-based global observation forecasting. Comprehensive experiments demonstrate that AIDA initialization significantly improves the roll out and efficiency of AIWP. Additionally, we show that DAWP holds promising potential to be applied in global precipitation forecasting.

Problem

• Overcoming reanalysis data limitations in AI weather prediction systems
• Learning spatiotemporal dynamics from irregular satellite observation data
• Enabling global observation forecasting across disparate measurement systems

Innovation

• AI data assimilation with mask multi-modality autoencoder
• Spatiotemporal decoupling transformer for observation dynamics
• Cross-regional boundary conditioning enables global forecasting