Skip to main content
QUICK REVIEW

[论文解读] Huge Ensembles Part I: Design of Ensemble Weather Forecasts using Spherical Fourier Neural Operators

Ankur Mahesh, William D. Collins|arXiv (Cornell University)|Aug 6, 2024
Meteorological Phenomena and SimulationsEarth and Planetary Sciences被引用 3
一句话总结

本文提出一种基于机器学习的集合天气预报系统,采用球面傅里叶神经算子(SFNO)生成包含1,000至10,000名成员的巨大集合,克服了传统数值模型在计算上的限制。通过结合扰动参数与培育向量初始条件扰动,该方法生成了校准良好、可靠的概率预报,通过了谱分析和极端天气事件诊断,展示了对低概率高影响天气事件的真实模拟。

ABSTRACT

Studying low-likelihood high-impact extreme weather events in a warming world is a significant and challenging task for current ensemble forecasting systems. While these systems presently use up to 100 members, larger ensembles could enrich the sampling of internal variability. They may capture the long tails associated with climate hazards better than traditional ensemble sizes. Due to computational constraints, it is infeasible to generate huge ensembles (comprised of 1,000-10,000 members) with traditional, physics-based numerical models. In this two-part paper, we replace traditional numerical simulations with machine learning (ML) to generate hindcasts of huge ensembles. In Part I, we construct an ensemble weather forecasting system based on Spherical Fourier Neural Operators (SFNO), and we discuss important design decisions for constructing such an ensemble. The ensemble represents model uncertainty through perturbed-parameter techniques, and it represents initial condition uncertainty through bred vectors, which sample the fastest growing modes of the forecast. Using the European Centre for Medium-Range Weather Forecasts Integrated Forecasting System (IFS) as a baseline, we develop an evaluation pipeline composed of mean, spectral, and extreme diagnostics. Using large-scale, distributed SFNOs with 1.1 billion learned parameters, we achieve calibrated probabilistic forecasts. As the trajectories of the individual members diverge, the ML ensemble mean spectra degrade with lead time, consistent with physical expectations. However, the individual ensemble members' spectra stay constant with lead time. Therefore, these members simulate realistic weather states, and the ML ensemble thus passes a crucial spectral test in the literature. The IFS and ML ensembles have similar Extreme Forecast Indices, and we show that the ML extreme weather forecasts are reliable and discriminating.

研究动机与目标

  • 解决传统基于物理的模型在生成包含1,000至10,000名成员的巨大天气集合时计算不可行的问题,以研究低概率高影响(LLHI)天气事件。
  • 设计一种基于机器学习的集合预报系统,准确捕捉初始条件和模型不确定性,同时保持长期动力稳定性。
  • 使用均值、谱分析和极端事件诊断方法,评估机器学习集合预报相对于业务模型的表现,确保概率预报的可靠性与判别能力。
  • 证明基于SFNO的集合能够真实模拟长时间自回归滚动过程中的大气动力学与极端天气事件。
  • 提供一个可复现、开源的框架,用于使用全球再分析数据训练、推理和评分大规模机器学习天气集合。

提出的方法

  • 本研究采用球面傅里叶神经算子(SFNO),一种专为球面上偏微分方程求解设计的深度学习架构,用于模拟欧洲中期天气预报中心(ECMWF)集成预报系统(IFS)。
  • 集合在训练后构建,通过扰动参数表示模型不确定性,通过培育向量在预报误差的最快增长模式上采样初始条件不确定性。
  • 使用ERA5再分析数据作为真实值,训练一个大规模分布式SFNO模型(含11亿参数),以最小化每个时间步的确定性均方误差(MSE)。
  • 评估流程包括均值场诊断、谱诊断(功率谱)以及极端天气事件诊断,如极端预报指数(EFI)和可靠性图。
  • 集合推理采用修改版的earth2mip框架,配备专门用于培育向量生成和概率评分的模块。
  • 该方法利用NERSC和ALCC计划的高性能计算资源,在大规模上进行模型训练与评估。
Figure 1: Overview of ensemble architecture. The ensemble is constructed using two methods: initial condition perturbations and model perturbations. The initial condition perturbations are generated using bred vectors, to sample the fastest growing errors in the initial condition. Model perturbation
Figure 1: Overview of ensemble architecture. The ensemble is constructed using two methods: initial condition perturbations and model perturbations. The initial condition perturbations are generated using bred vectors, to sample the fastest growing errors in the initial condition. Model perturbation

实验结果

研究问题

  • RQ1基于球面傅里叶神经算子的机器学习模型能否生成计算上可行且动力稳定的、包含1,000至10,000名成员的巨大天气集合?
  • RQ2该机器学习集合是否能通过培育向量和扰动参数,分别准确表示初始条件和模型不确定性?
  • RQ3与IFS和ERA5再分析相比,该机器学习集合的谱特性(特别是功率谱随时间的演变)如何?
  • RQ4该机器学习集合的极端天气预报是否可靠且具有判别能力,依据可靠性图和极端预报指数(EFI)衡量?
  • RQ5该机器学习集合能否有效采样气候灾害分布的长尾,从而改善对低概率高影响天气事件的表征?

主要发现

  • 机器学习集合的均值谱随预报时效逐渐退化,符合物理预期;而单个集合成员的谱保持稳定,表明滚动过程中大气动力学模拟真实。
  • 基于SFNO的集合实现了校准的概率预报,可靠性图显示,95百分位温度事件的预测值在48小时和96小时预报时效下与实际发生频率高度一致。
  • 机器学习集合的极端预报指数(EFI)值与IFS相当,证实该模型对极端事件的预报技能水平与业务系统一致。
  • 在预测95百分位温度事件(概率为20–30%)时,实际发生频率略高于预测值,表明预报具有保守但可靠的概率特性。
  • 经训练的SFNO模型(含11亿参数)成功在0.25°水平分辨率下模拟了IFS系统,其确定性预报性能与业务系统相当。
  • 整个流程——包括模型训练、集合推理和评分——均已开源,版本控制的代码与模型权重可通过Zenodo和NERSC门户获取。
Figure 2: Comparing different versions of SFNO. (a) The 850 hPa temperature spread-error ratios are compared for lagged ensembles. A lagged ensemble is created by initializing SFNO using nine adjacent time steps, and the spread-error is shown for each SFNO configurations. (b) Relative power spectra
Figure 2: Comparing different versions of SFNO. (a) The 850 hPa temperature spread-error ratios are compared for lagged ensembles. A lagged ensemble is created by initializing SFNO using nine adjacent time steps, and the spread-error is shown for each SFNO configurations. (b) Relative power spectra

更好的研究,从现在开始

从阅读论文到最终审阅,大幅缩短您的研究时间。

无需绑定信用卡

本解读由 AI 生成,并经人工编辑审核。