pipette
ESEspañol

One2N: Recursive refinement breaks the resolution barrier for Earth-system AI

Yanyan Tan, Xiangjun Tian, Can Wang, Hongqin Zhang, Yinhai Luo, Min Zhao, Zhe Jin, Xingchao Yang, Zaifei Wang, Banglin Zhang, Aiwen Wang, Jing Zhou, Tao Wang, Baosheng An, Xuliang Zhuang, Shilong Piao, Tandong Yao State Key Laboratory of Tibetan Plateau Earth System, Environment, Resources, Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing, China, University of Chinese Academy of Sciences, Beijing, China, College of Mathematical Science, Beijing Information Science, Technology University, Beijing, China, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing, China, College of Meteorology, Oceanography, National University of Defense Technology, Changsha, China, State Key Laboratory of Physical Oceanography, Institute of Oceanographic Instrumentation, Qilu University of Technology, Qingdao, China, Institute of Carbon Neutrality, College of Urban, Environmental Sciences, Peking University, Beijing and 1 more

PreprintBold claims, read criticallyReal-world use

In the authors' words

Earth-system AI models approach kilometre-scale resolution, but halving grid spacing quadruples the grid-point count and end-to-end training cost. We introduce One2N, a recursive incremental framework extending a coarse model through levels of cross-resolution residual correction. At each level the parent is frozen and excluded from backpropagation, so its activations and optimizer states are never retained; only a tier-specific module is trained. A single refinement level on the NeuralGCM-0.7 baseline yields the two-level One2N-0.07, trained on one 80 GB GPU. Against HRES-0.07 and ERA5-0.25 analyses it matches its parent's tropospheric skill while reducing upper-stratospheric errors and resolving finer structure. Tropical-cyclone track errors nearly match the parent's and are lower than Pangu-Weather-0.25 at medium range; RMSE is lower than that model's on 93.46% of common targets. One2N thus turns an existing forecasting system, or an inexpensively trained coarse baseline, into a reusable foundation for high-resolution development, breaking the resolution barrier.

Main resultThe abstract does not state a limitation.

Appeared: Tuesday, November 10. arXiv. Preprint, not yet peer-reviewed.

Authors' comment: 32 pages, 12 figures and 7 tables, including Supplementary Information