Making use of the information in ensemble weather forecasts: comparing the end to end and full statistical modelling approaches
Stephen Jewson
我们讨论了如何集合天气预报可以使用，并突出了两个特定的方法的优点和缺点。
We discuss how ensemble weather forecasts can be used, and highlight the advantages and disadvantages of two particular methods.
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。
继续探索
Weather forecasts, Weather derivatives, Black-Scholes, Feynmann-Kac and Fokker-Planck
本文研究了天气预报、天气导数、Black-Scholes方程、Feynmann-Kac理论和Fokker-Planck方程之间的关系。有一个有用的结果，但总的来说，我们提出的关系似乎比实际有用的更有趣。
Improving probabilistic weather forecasts using seasonally varying calibration parameters
我们发现，概率天气预报的网站特定的温度可以显着提高使用季节性变化，而不是恒定的校准参数。
How does an AI Weather Model Learn to Forecast Extreme Weather?
探究AI气象模型SFNO如何学习极端天气，发现其对大气河流和热带气旋的预报信息在训练后期会丢失，并提出利用早期检查点信息改进预报的训练策略。
Multivariate post-processing of probabilistic sub-seasonal weather regime forecasts
本文针对北大西洋—欧洲地区概率性天气型预报存在的时间和幅度偏差，提出结合单变量集合模式输出统计与集合Copula耦合的两步后处理方法，将逐日天气型预报技巧时效延长至14.5天，并在多元预报质量上优于现有校准做法。
Historical observations of severe weather and simulated severe weather environments (i.e., features) from the Global Ensemble Forecast System v12 (GEFSv12) Reforecast Dataset (GEFS/R) are used in conjunction to train and test random forest (RF) machine learning (ML) models to probabilistically forecast severe weather out to days 4--8. RFs are trained with 9 years of the GEFS/R and severe weather reports to establish statistical relationships. Feature engineering is briefly explored to examine alternative methods for gathering features around observed events, including simplifying features usin…
Hybrid weather prediction using spectral nudging toward machine-learning forecasts
本文研究了一种混合数值天气预测方法，通过将ECMWF集成预测系统谱趋近于机器学习天气模型的预测，以提升大尺度预报能力，同时保持物理模型的小尺度行为。
"What is a realistic forecast?" Assessing data-driven weather forecasts, a journey from verification to falsification
本文针对数据驱动天气预报的“现实性”问题，提出功能、结构和物理三种现实性定义，并建立基于证伪的评估框架以补充传统验证方法。
Evaluating the Predictability of Selected Weather Extremes with Aurora, an AI Weather Forecast Model
Aurora模型在短期极端天气预报中表现优异，但在亚季节性预报中因动力结构丧失而受限。
Forecast error diagnostics in neural weather models
本文利用ConvCastNet模型通过子区域松弛实验诊断神经天气预报误差，发现平流层和边界层对中期中纬度预报技巧改善至关重要，而热带影响可忽略，并提出结合误差敏感性分析指导模型改进。
AI-driven weather forecasts enable anticipated attribution of extreme events to human-made climate change
研究提出一种结合全球气候模型与深度学习天气预报的混合归因方法，实现了对极端天气事件提前归因于人为气候变化，成功预测了热浪、飓风和气旋等极端事件因气候变化而加剧的影响。
Computer-generated forecasts divide the earth's surface into gridboxes, each now ~25% of the size of London, and predict one value per gridbox. If weather varies markedly within a gridbox forecasts for specific sites inevitably fail. A completely new statistical post-processing method, using ensemble forecasts as input, anticipates two gridbox-weather-dependant factors: degree of variation in each gridbox, and bias on the gridbox scale. Globally, skill improves substantially; for extreme rainfall, for example, useful forecasts extend 5 days ahead. Without post-processing this limit is < 1 day.…
Leveraging deterministic weather forecasts for in-situ probabilistic temperature predictions via deep learning
提出一种神经网络方法，通过为多预报时效训练单一模型，从确定性数值天气预报生成概率预报，将CRPS降低15%并改善分布校准，无需模拟多轨迹即可为任意网格预报增加不确定性信息。
Windows of opportunity in subseasonal weather regime forecasting: A statistical-dynamical approach
本研究利用MJO和SPV位相信息并结合神经网络，提出统计-动力方法以改进三周提前期的欧洲天气型活动预报，较ECMWF提升2.9%的准确率，尤其改善了格陵兰阻塞和欧洲阻塞预报。
Rainfall is an important variable to be able to monitor and forecast across Africa, due to its impact on agriculture, food security, climate related diseases and public health. Numerical Weather Models (NWM) are an important component of this work, due to their complete spatial coverage, high resolution, and ability to forecast into the future. In this study, we seek to evaluate the spatiotemporal skill of short-term rainfall forecasts of rainfall across Africa. Specifically, the European Centre for Medium-Range Weather Forecasts (ECMWF) and the National Centers for Environmental Prediction-Gl…
An update to ECMWF's machine-learned weather forecast model AIFS
本文报告了ECMWF机器学习天气预报模型AIFS的更新，通过引入物理一致性约束、更新训练计划和扩展变量集合，显著提升了降水预报和高空评分，并已全面业务化运行。
Evaluation of forecasts by a global data-driven weather model with and without probabilistic post-processing at Norwegian stations
本研究评估了盘古天气等全球数据驱动模型与NWP模型在挪威站点有无概率后处理的预报性能，发现其与ECMWF相当，后处理显著提升预报质量，证实了数据驱动模型的业务应用潜力。
FuXi Weather: A data-to-forecast machine learning system for global weather
本文提出FuXi Weather系统，通过多卫星数据融合实现高分辨率全球天气预测，优于ECMWF HRES，在观测稀疏区域表现更优。
The rise of data-driven weather forecasting
本研究在业务化环境中首次将机器学习模型PanguWeather生成的预报与ECMWF IFS等传统NWP系统进行对比，发现ML预报在准确性和极端事件上表现相当，但存在平滑度增加和偏差漂移等缺陷，标志着依赖ML与再分析数据的新预报范式正在兴起。
Ensemble weather forecasts enable a measure of uncertainty to be attached to each forecast, by computing the ensemble's spread. However, generating an ensemble with a good spread-error relationship is far from trivial, and a wide range of approaches to achieve this have been explored -- chiefly in the context of numerical weather prediction models. Here, we aim to transform a deterministic neural network weather forecasting system into an ensemble forecasting system. We test four methods to generate the ensemble: random initial perturbations, retraining of the neural network, use of random dro…
The Met Office Space Weather Operations Centre was founded in 2014 and part of its remit is a daily Space Weather Technical Forecast to help the UK build resilience to space weather impacts; guidance includes four day geo-magnetic storm forecasts (GMSF) and X-ray flare forecasts (XRFF). It is crucial for forecasters, users, modelers and stakeholders to understand the strengths and weaknesses of these forecasts; therefore, it is important to verify against the most reliable truth data source available. The present study contains verification results for XRFFs using GOES-15 satellite data and GM…
GenCast: Diffusion-based ensemble forecasting for medium-range weather
GenCast是一种基于扩散的机器学习天气预报模型，通过数十年再分析数据训练，以8分钟生成15天全球集合预报，在97.4%的评估目标上超越ECMWF的ENS，尤其在极端天气、热带气旋和风电预测方面表现更优。
VarteX: Enhancing Weather Forecast through Distributed Variable Representation
针对多气象变量高效处理难题，本文提出变量聚合方案与高效学习框架 VarteX，以更少参数和资源实现优于传统模型的预报性能。
Diffusion-LAM: Probabilistic Limited Area Weather Forecasting with Diffusion
本文提出Diffusion-LAM，一种以边界数据为条件的概率有限区域天气模型，并在MEPS数据集上验证了其提供准确概率预报的潜力。
SEEDS: Emulation of Weather Forecast Ensembles with Diffusion Models
本研究提出SEEDS方法，利用扩散模型仿真天气预报集合体以降低计算成本，既能匹配物理集合的统计特性，还能校正预报偏差、提升极端天气概率预报精度，该方法可拓展至气候建模的不确定性量化。
Aardvark weather: end-to-end data-driven weather forecasting
本研究提出端到端数据驱动天气预报系统Aardvark Weather，可替代完整业务化数值天气预报流程，以极低的计算量和数据量实现更优的全球及本地预报性能，为新一代中期预报模型奠定基础。
An ensemble of data-driven weather prediction models for operational sub-seasonal forecasting
提出一套结合数据驱动模型与ECMWF海洋模型的业务化多模型集合预报系统，在1度分辨率下实现4周时效的全球天气预报，其2米温度预测平均超ECMWF原始集合4-17%，证明数据驱动方法可实现接近最先进的次季节预报。
In this paper, we present Pangu-Weather, a deep learning based system for fast and accurate global weather forecast. For this purpose, we establish a data-driven environment by downloading $43$ years of hourly global weather data from the 5th generation of ECMWF reanalysis (ERA5) data and train a few deep neural networks with about $256$ million parameters in total. The spatial resolution of forecast is $0.25^\circ\times0.25^\circ$, comparable to the ECMWF Integrated Forecast Systems (IFS). More importantly, for the first time, an AI-based method outperforms state-of-the-art numerical weather …
We present a data-driven approach for forecasting global weather using graph neural networks. The system learns to step forward the current 3D atmospheric state by six hours, and multiple steps are chained together to produce skillful forecasts going out several days into the future. The underlying model is trained on reanalysis data from ERA5 or forecast data from GFS. Test performance on metrics such as Z500 (geopotential height) and T850 (temperature) improves upon previous data-driven approaches and is comparable to operational, full-resolution, physical models from GFS and ECMWF, at least…
Local Off-Grid Weather Forecasting with Multi-Modal Earth Observation Data
本文提出一种多模态Transformer模型，通过端到端训练将网格化预报降尺度至非网格位置，提升局部天气预测精度，实验显示其优于多种非数据驱动和数据驱动的非网格预报方法。
This popular article provides a short summary of the progress and prospects in Weather and Climate Modelling for the benefit of high school and undergraduate college students and early career researchers. Although this is not a comprehensive scientific article, the basic information provided here is intended to introduce students and researchers to the topic of Weather and Climate Modelling - which comes under the broad discipline of Atmospheric / Oceanic / Climate / Earth Sciences. This article briefly summarizes the historical developments, progress, scientific challenges in weather and clim…
Continuous Ensemble Weather Forecasting with Diffusion models
针对扩散模型集成天气预报计算成本高、高时间分辨率下误差累积的问题，提出连续集成预报方法，可并行生成时间一致轨迹，还能结合自回归滚动实现任意精细分辨率预报，在全球天气预报中取得竞争性结果。
Skillful joint probabilistic weather forecasting from marginals
针对传统数值天气预报精度速度不足的问题，提出FGN建模方法，通过约束模型集合的学习扰动生成预报、直接优化CRPS训练，在多指标下达到SOTA，还能捕捉空间联合结构。
GraphCast: Learning skillful medium-range global weather forecasting
提出基于机器学习的GraphCast方法，直接从再分析数据训练，在1分钟内生成0.25度分辨率的10天全球天气预报，在90%验证目标上超越传统系统，提升极端事件预测能力。
Improving Medium Range Severe Weather Prediction through Transformer Post-processing of AI Weather Forecasts
本研究利用Transformer后处理提升AI天气预报的中等范围严重天气预测能力，通过改进模型结构和训练方法，提高预测准确性和可靠性。
Uncertainty quantification for data-driven weather models
本文旨在系统比较不确定性量化方法，为确定性数据驱动天气模型 Pangu-Weather 生成概率预报。通过扰动初始条件生成集合及事后统计与机器学习方法，实验表明该概率预报在欧洲中期预报中优于传统物理集合模型（5天预报时效内）。
MetMamba: Regional Weather Forecasting with Spatial-Temporal Mamba Model
针对当前DLWP领域有限区域建模和骨干网络研究不足的问题，提出基于Mamba状态空间模型的MetMamba模型，验证其性能优于传统注意力、神经算子骨干网络，并证实与全局模型耦合训练的有限区域深度学习建模可行。
Forecast lead time and the timing of crew deployment for weather-driven power outages using AI weather models
本研究利用GraphCast和FourCastNet两个AI气象模型，分析预报提前期对停电预测人员部署价值的影响，明确提前部署的优势及盈亏平衡可用性比率，还探讨了地表预测因子的技能提升与预测区间校准问题。
Results of direct numerical simulations have been used to show that intensive thermal convection in a horizontal layer and on a hemisphere can be described by the distributed chaos approach. The vorticity and helicity dominated distributed chaos were considered for this purpose. Results of numerical simulations of the Weather Research and Forecast Model (with the moist convection and with the Coriolis effect) and of the Coupled Ocean-Atmosphere Mesoscale Prediction System (COAMPS) were also analysed to demonstrate applicability of this approach to the atmospheric processes. The ensemble foreca…
A primary goal of the National Oceanic and Atmospheric Administration (NOAA) Warn-on-Forecast (WoF) project is to provide rapidly updating probabilistic guidance to human forecasters for short-term (e.g., 0-3 h) severe weather forecasts. Maximizing the usefulness of probabilistic severe weather guidance from an ensemble of convection-allowing model forecasts requires calibration. In this study, we compare the skill of a simple method using updraft helicity against a series of machine learning (ML) algorithms for calibrating WoFS severe weather guidance. ML models are often used to calibrate se…
Generating Probabilities From Numerical Weather Forecasts by Logistic Regression
Logistic模型作为一种工具，将数值天气预报系统（确定性和集合）的输出转换为二进制事件的概率预报。逻辑模型通过将对数优势比等于输入的线性组合来获得。与任何统计模型一样，如果输入的数量与预测实例的数量相当，则逻辑模型将遭受过拟合。计算方法，以避免过度拟合正则化进行了讨论，并提出了有效的方法模型评估和选择。一个logit版本的所谓的套索，这本来是一个线性工具，进行了讨论。在套索模型中，不太重要的输入被识别和丢弃，从而提供了一个有效的和自动的模型简化过程。出于这个原因，套索模型对于诊断目的特别有吸引力。
Do AI models produce better weather forecasts than physics-based models? A quantitative evaluation case study of Storm Ciarán
本研究比较了四个机器学习模型与数值天气预报模型对欧洲风暴Storm Ciarán的预报，发现机器学习模型能准确捕捉气旋大尺度结构但低估风速峰值并难以解析关键细节，表明需通过高影响天气案例研究评估其性能。
Forecasting future weather and climate is inherently difficult. Machine learning offers new approaches to increase the accuracy and computational efficiency of forecasts, but current methods are unable to accurately model uncertainty in high-dimensional predictions. Score-based diffusion models offer a new approach to modeling probability distributions over many dependent variables, and in this work, we demonstrate how they provide probabilistic forecasts of weather and climate variables at unprecedented resolution, speed, and accuracy. We apply the technique to day-ahead solar irradiance fore…
Extreme weather events have significant consequences, dominating the impact of climate on society. While high-resolution weather models can forecast many types of extreme events on synoptic timescales, long-term climatological risk assessment is an altogether different problem. A once-in-a-century event takes, on average, 100 years of simulation time to appear just once, far beyond the typical integration length of a weather forecast model. Therefore, this task is left to cheaper, but less accurate, low-resolution or statistical models. But there is untapped potential in weather model output: …
Super Resolution On Global Weather Forecasts
本研究通过超分辨率技术提升GraphCast全球温度预测的空间分辨率，从1度细化至0.5度，以改进深度学习天气预报的精度。
FuXi: A cascade machine learning forecasting system for 15-day global weather forecast
本文提出FuXi，一个基于39年ECMWF ERA5再分析数据集的级联机器学习天气预报系统，提供15天全球预报，并在15天预报性能上首次实现与ECMWF集合平均相当的水平。
Data driven weather forecasts trained and initialised directly from observations
提出一种纯观测驱动的机器学习天气预报方法，直接从历史观测训练神经网络并初始化，无需再分析数据，初步实现12小时预报，避免数据同化依赖。
Intuitively, one would expect a more skillful forecast if predicting weather averaged over one week instead of the weather averaged over one day, and similarly for different spatial averaging areas. However, there are few systematic studies of averaging and forecast skill with modern forecasts, and it is therefore not clear how much improvement in forecast performance is produced via averaging. Here we present a direct investigation of averaging effects, based on data from operational numerical weather forecasts. Data is analyzed for precipitation and surface temperature, for lead times of rou…
Huracan: A skillful end-to-end data-driven system for ensemble data assimilation and weather prediction
提出Huracan系统，结合集合数据同化与预报模型，仅依赖观测数据实现与ECMWF ENS相当的预报精度，在80.2%的变量和预报时长组合上匹配或超越其连续排序概率评分。
FuXi-2.0: Advancing machine learning weather forecasting model for practical applications
针对现有机器学习天气预报模型时间分辨率粗、变量少的问题，研究提出融合海气模块的FuXi-2.0模型，可逐小时全球预报，在多行业相关气象变量及热带气旋强度预报上优于ECMWF HRES和前代模型。
Scaling transformer neural networks for skillful and reliable medium-range weather forecasting
本文提出简单Transformer模型Stormer，通过天气特定嵌入、随机化动力学预测和压力加权损失，在WeatherBench 2上以极少训练数据和计算量实现领先的中期天气预报性能，并展现出良好的扩展特性。
Global Forecasting of Tropical Cyclone Intensity Using Neural Weather Models
本研究通过神经天气模型结合后处理框架，提升全球热带气旋强度预测的准确性和可及性。
OMG-HD: A High-Resolution AI Weather Model for End-to-End Forecasts from Observations
提出直接基于观测数据的高分辨率AI天气预报模型OMG-HD，无需数据同化，在美国本土12小时预报时效内多项指标优于IFS-HRES和HRRR。
Fixing the Double Penalty in Data-Driven Weather Forecasting Through a Modified Spherical Harmonic Loss Function
针对数据驱动天气预报中均方误差损失导致的“双重惩罚”与精细尺度平滑问题，提出一种无需参数的球谐损失函数修改方法，将去相关与谱振幅误差分离，微调 GraphCast 模型后有效分辨率从 1,250km 提升至 160km，显著改善了预报清晰度、集合散度及极端天气预测。
Post-processing ensemble prediction systems can improve the reliability of weather forecasting, especially for extreme event prediction. In recent years, different machine learning models have been developed to improve the quality of weather post-processing. However, these models require a comprehensive dataset of weather simulations to produce high-accuracy results, which comes at a high computational cost to generate. This paper introduces the ENS-10 dataset, consisting of ten ensemble members spanning 20 years (1998-2017). The ensemble members are generated by perturbing numerical weather s…
Interpolation of mountain weather forecasts by machine learning
本文提出利用机器学习插值山地未来天气，结合周边平原预报数据和历史观测数据，使用LightGBM模型提升山地天气预报精度，减少训练时间并改进RMSE。
GraphDOP: Towards skilful data-driven medium-range weather forecasts learnt and initialised directly from observations
GraphDOP是ECMWF开发的端到端预报系统，直接从地球系统观测学习并初始化，无需物理再分析输入，能预测未来五天的天气参数。
FengWu-W2S: A deep learning model for seamless weather-to-subseasonal forecast of global atmosphere
本文提出FengWu-W2S模型，在FengWu基础上引入海-陆-气耦合与扰动策略，实现最长42天的无缝次季节预报，并提升了对气温、降水及MJO、NAO等信号的预测能力。
SDECast: Probabilistic Weather Forecasting in Continuous Time with Neural SDEs
针对现有机器学习天气预报模型在短时间步长下误差累积和缺乏时间连续性的问题，提出基于神经随机微分方程的SDECast框架，实现连续时间概率预报，并在模拟流和全球小时级预报中验证了其有效性。
FourCastNet, short for Fourier Forecasting Neural Network, is a global data-driven weather forecasting model that provides accurate short to medium-range global predictions at $0.25^{\circ}$ resolution. FourCastNet accurately forecasts high-resolution, fast-timescale variables such as the surface wind speed, precipitation, and atmospheric water vapor. It has important implications for planning wind energy resources, predicting extreme weather events such as tropical cyclones, extra-tropical cyclones, and atmospheric rivers. FourCastNet matches the forecasting accuracy of the ECMWF Integrated F…
Numerical models outperform AI weather forecasts of record-breaking extremes
该研究比较了数值模型HRES与多个先进AI天气预报模型在破纪录极端天气中的表现，发现AI模型误差更大、低估破纪录事件频率和强度，表明其在训练域外外推和高风险应用中仍存在明显局限。
扫码添加微信好友，提出您的宝贵建议 👇
💡 备注请填写：网站反馈
发表机构由 AI 从论文文本抽取，如发现错误，欢迎反馈修正建议。