Earth-o1: A Grid-free Observation-native Atmospheric World Model
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866913098933207040 |
|---|---|
| author | Gong, Junchao Xu, Kaiyi Wei, Wangxu Tu, Siwei Xu, Jingyi Liu, Zili Fan, Hang Zhou, Zhiwang Han, Tao Xiao, Yi Gu, Xinyu Li, Zhangrui Zhang, Wenlong Chen, Hao Yang, Xiaokang Wang, Yaqiang Cheng, Lijing Gentine, Pierre Ouyang, Wanli Zhang, Feng Tan, Zhe-Min Zhou, Bowen Ling, Fenghua Fei, Ben Bai, Lei |
| author_facet | Gong, Junchao Xu, Kaiyi Wei, Wangxu Tu, Siwei Xu, Jingyi Liu, Zili Fan, Hang Zhou, Zhiwang Han, Tao Xiao, Yi Gu, Xinyu Li, Zhangrui Zhang, Wenlong Chen, Hao Yang, Xiaokang Wang, Yaqiang Cheng, Lijing Gentine, Pierre Ouyang, Wanli Zhang, Feng Tan, Zhe-Min Zhou, Bowen Ling, Fenghua Fei, Ben Bai, Lei |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_06337 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Earth-o1: A Grid-free Observation-native Atmospheric World Model Gong, Junchao Xu, Kaiyi Wei, Wangxu Tu, Siwei Xu, Jingyi Liu, Zili Fan, Hang Zhou, Zhiwang Han, Tao Xiao, Yi Gu, Xinyu Li, Zhangrui Zhang, Wenlong Chen, Hao Yang, Xiaokang Wang, Yaqiang Cheng, Lijing Gentine, Pierre Ouyang, Wanli Zhang, Feng Tan, Zhe-Min Zhou, Bowen Ling, Fenghua Fei, Ben Bai, Lei Computer Vision and Pattern Recognition 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. |
| title | Earth-o1: A Grid-free Observation-native Atmospheric World Model |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2605.06337 |