Earth-o1: A Grid-free Observation-native Atmospheric World Model

2026-05-07

AI Summary

Traditional atmospheric modeling relies on fixed spatial grids, which struggle to efficiently assimilate multi-source unstructured observational data and incur high computational costs. This work proposes a grid-free, solver-independent, observation-native continuous atmospheric world model that learns the three-dimensional physical evolution of Earth’s atmosphere end-to-end directly from raw multi-sensor observations, enabling high-fidelity spatiotemporal forecasting. The approach constructs a unified meshless dynamic field capable of real-time prediction and cross-sensor inference. Hindcast experiments demonstrate that the model achieves surface forecast skill comparable to that of operational Integrated Forecasting Systems (IFS), confirming its effectiveness and state-of-the-art performance.

Abstract

Despite the unprecedented volume of multimodal data provided by modern Earth observation systems, our ability to model atmospheric dynamics remains constrained. Traditional modeling frameworks force heterogeneous measurements into predefined spatial grids, inherently limiting the full exploitation of raw sensor data and creating severe computational bottlenecks. Here we present Earth-o1, an observation-native atmospheric world model that overcomes these structural limitations. Rather than relying on conventional atmospheric dynamical modeling systems or traditional data assimilation, Earth-o1 directly learns the continuous, three-dimensional physical evolution of the Earth system from ungridded observational data. By integrating diverse sensor inputs into a unified, grid-free dynamical field, the model autonomously advances the atmospheric state in space and time. We show that this fundamentally distinct paradigm enables direct, real-time forecasting and cross-sensor inference without the overhead of explicit numerical solvers. In hindcast evaluations, Earth-o1 achieves surface forecast skill comparable to the operational Integrated Forecasting System (IFS). These results establish that continuous, observation-driven world models -- a new class of fully observation-native geophysical simulators -- can match the fidelity of established physical frameworks, providing a scalable data-driven foundation for a digital twin of the Earth.

Problem

• atmospheric modeling
• Earth observation
• grid-free
• data assimilation
• multimodal data

Innovation

• grid-free
• observation-native
• atmospheric world model
• continuous dynamical field
• data-driven Earth simulation