用AI天气模型学习多尺度动力学的挑战:对稳定性的启示及一种解决方案
Challenges of learning multi-scale dynamics with AI weather models: Implications for stability and one solution
AI总结:
该研究揭示AI天气模型长期积分不稳定的普遍成因是深度学习的谱偏差(仅学大尺度、忽略小尺度),结合湍流物理与积分器无收敛性会放大误差,进而提出缓解方案,实现了兼具准确短期预报与数百年稳定物理一致积分的气候数据驱动模型。
摘要:
若要将基于AI的天气模型用于次季节至季节预报乃至更长期的场景(例如气候变化预估),长期稳定性和物理一致性是其核心关键属性。然而,当前的AI天气模型仅能提供准确的短期预报,当时间积分时长超过数周或数月时,模型就会变得不稳定或出现物理不一致性:要么出现数值爆炸,要么生成不符合实际的大气变量动力学特征,这一问题与当前自回归大语言模型的缺陷类似。这类不稳定性的成因尚不明确,现有用于提升模型稳定预报时长的方法多为临时性方案,缺乏严谨的理论支撑。本文揭示了任意湍流流动中这类不稳定性的普遍因果机制:其根源是**谱偏差**,即*任何*深度学习架构都存在学习偏差,只会学习大尺度动力学特征,而完全忽略小尺度特征。本文进一步阐明了湍流物理特性、以及基于深度学习的时间积分器缺乏收敛性这两个因素如何放大该偏差,进而导致不稳定的误差传播。最后,本文以准地转流动和欧洲中期天气预报中心(ECMWF)再分析数据为测试案例,打通深度学习理论与数值分析之间的壁垒,针对这类非物理行为提出了一种缓解方案。我们构建了适用于气候系统的长期物理一致性数据驱动模型,验证了其既能实现准确的短期预报,又能完成数百年时长的时间积分,且均值和变率均保持准确。
英文摘要:
Long-term stability and physical consistency are critical properties for AI-based weather models if they are going to be used for subseasonal-to-seasonal forecasts or beyond, e.g., climate change projection. However, current AI-based weather models can only provide short-term forecasts accurately since they become unstable or physically inconsistent when time-integrated beyond a few weeks or a few months. Either they exhibit numerical blow-up or hallucinate unrealistic dynamics of the atmospheric variables, akin to the current class of autoregressive large language models. The cause of the instabilities is unknown, and the methods that are used to improve their stability horizons are ad-hoc and lack rigorous theory. In this paper, we reveal that the universal causal mechanism for these instabilities in any turbulent flow is due to \textit{spectral bias} wherein, \textit{any} deep learning architecture is biased to learn only the large-scale dynamics and ignores the small scales completely. We further elucidate how turbulence physics and the absence of convergence in deep learning-based time-integrators amplify this bias, leading to unstable error propagation. Finally, using the quasi-geostrophic flow and European Center for Medium-Range Weather Forecasting (ECMWF) Reanalysis data as test cases, we bridge the gap between deep learning theory and numerical analysis to propose one mitigative solution to such unphysical behavior. We develop long-term physically-consistent data-driven models for the climate system and demonstrate accurate short-term forecasts, and hundreds of years of time-integration with accurate mean and variability.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。