集合预报研究现状与展望

本文系统地论述了集合预报的概念、分类以及国外内的研究现状,在国外的研究现状中重点介绍了目前国际上最先进、最具代表性的两种方法,奇异向量法和增长模繁殖法。国内则介绍了在该方面的最新进展,包括我实验室所做的工作。最后阐述了集合预报的可能发展方向和应用前景。     

Effect of doubling the ensemble size on the performance of ensemble prediction in the warm season using MOGREPS implemented at the KMA

Using the Met Office Global and Regional Ensemble Prediction System (MOGREPS) implemented at the Korea Meteorological Administration (KMA), the effect of doubling the ensemble size on the performance of ensemble prediction in the warm season was evaluated. Because a finite ensemble size causes sampling error in the full forecast probability distribution function (PDF), ensemble size is closely related to the efficiency of the ensemble prediction system. Prediction capability according to doubling the ensemble size was evaluated by increasing the number of ensembles from 24 to 48 in MOGREPS implemented at the KMA. The initial analysis perturbations generated by the Ensemble Transform Kalman Filter (ETKF) were integrated for 10 days from 22 May to 23 June 2009. Several statistical verification scores were used to measure the accuracy, reliability, and resolution of ensemble probabilistic forecasts for 24 and 48 ensemble member forecasts. Even though the results were not significant, the accuracy of ensemble prediction improved slightly as ensemble size increased, especially for longer forecast times in the Northern Hemisphere. While increasing the number of ensemble members resulted in a slight improvement in resolution as forecast time increased, inconsistent results were obtained for the scores assessing the reliability of ensemble prediction. The overall performance of ensemble prediction in terms of accuracy, resolution, and reliability increased slightly with ensemble size, especially for longer forecast times.     

集合敏感性在预报误差及可预报性研究中的应用进展综述

研究大气的可预报性和预报误差产生的原因,对于改进数值预报,提升业务预报技巧具有重要意义。集合敏感性基于具有流依赖特性的集合预报,通过建立预报与初始场或前期预报大气状态之间的统计关系,为揭示与预报对象可预报性相关的动力学特征及理解预报误差来源和传播机制提供了一种新方法。同时,介绍了集合敏感性的定义和度量,并综述了其针对典型天气系统和高影响天气事件研究的进展,并讨论了该方法的优势和局限性。     

集合数值天气预报的研究进展

集合预报是目前国外广泛应用的一种新的数值预报技术,国内由于计算条件的限制,集合预报的研究应用起步较晚,目前虽已取得了一些研究成果,但还没有广泛的应用到实际的业务预报之中。集合预报的应用,在天气预报上主要是概率预报,另外在“目标观测”、资料同化等方面也有广泛应用。集合概率预报的一系列适用的验证,增加了概率预报的信度。     

Validation of Strategies using Clustering Analysis of ECMWF EPS for Initial Perturbations in a Limited Area Model Ensemble Prediction System

One of the main challenges for a skilful Limited Area Model Ensemble Prediction System (LAMEPS) is the generation of appropriate initial perturbations. In most operational LAMEPSs, the initial perturbations are provided by a global Ensemble Prediction System (EPS). Molteni et al. (2001 Marsigli, C., Montani, A., Nerozzi, F., Paccagnella, T., Tibaldi, S., Molteni, F. and Buizza, R. 2001. A strategy for high-resolution ensemble prediction. II: Limited-area experiments in four Alpine flood events. Quarterly Journal of the Royal Meteorological Society, 127: 2095–2115. (doi:10.1002/qj.49712757613)[Crossref], [Web of Science ®] , [Google Scholar]) proposed clustering analysis as an objective selection criterion to choose a member from the global European Centre for Medium-range Weather Forecasts (ECMWF)-EPS model as initial perturbations in LAMEPS. In this article, another strategy for using the clustering method is investigated which ensures that initial perturbations are centred on the control analysis. The main purpose of this article is to study the benefit of cluster analysis and to validate the effect of different clustering strategies on the performance of a 17-member LAMEPS. The system used in this study is the operational Aire Limitée Adaptation Dynamique Développement InterNational-Limited-Area Ensemble Forecasting (ALADIN-LAEF) model.

Three experiments were carried out over a 50-day period to validate different clustering strategies: i) representative members of 16 clusters from the 50-member ECMWF-EPS, where initial perturbations are not necessarily centred; ii) representative members from eight clusters and the symmetric pairs from ECMWF singular vector analysis; and iii) eight arbitrarily chosen ECMWF-EPS singular vector pairs. Results of the verified experiments show that the statistical reliability of ALADIN-LAEF improves when clustering is applied, but no clear improvement can be seen in the skill of LAMEPS. A case study of a heavy precipitation event confirms the result of the 50-day verification. The validation shows that none of the clustering strategies outperforms any other.

RÉSUMÉ?[Traduit par la rédaction] L'un des principaux défis d'un système performant de prévisions d'ensemble de modèle à domaine limité (LAMEPS) réside dans la génération de perturbations initiales appropriées. Dans la plupart des LAMEPS opérationnels, les perturbations initiales sont fournies par un système de prévisions d'ensemble (EPS) global. Molteni et coll. (2001 Molteni, F., Buizza, R., Marsigli, C., Montani, A., Nerozzi, F. and Paccagnella, T. 2001. A strategy for high-resolution ensemble prediction. I: Definition of representative members and global-model experiments. Quarterly Journal of the Royal Meteorological Society, 127: 2069–2094. (doi:10.1002/qj.49712757612)[Crossref], [Web of Science ®] , [Google Scholar]) ont proposé l'analyse par groupement comme critère de sélection objectif pour le choix d'un membre du modèle EPS global du Centre européen pour les prévisions météorologiques à moyen terme (ECMWF) comme perturbations initiales dans le LAMEPS. Dans cet article, nous étudions une autre stratégie d'utilisation de la méthode des groupements, qui assure que les perturbations initiales sont centrées sur l'analyse de contrôle. Le but premier de cet article est d’étudier les avantages de l'analyse par groupement et de valider l'effet des différentes stratégies de groupement sur la performance d'un LAMEPS de 17 membres. Le système utilisé dans cette étude est le modèle opérationnel ALADIN–LAEF (Aire limitée, Adaptation dynamique, Développement InterNational – Prévisions d'ensemble à domaine limité).Nous avons mené trois expériences sur une période de 50 jours pour valider différentes stratégies de groupement: i) membres représentatifs de 16 groupements d'un ECMWF–EPS de 50 membres, où les perturbations initiales ne sont pas nécessairement centrées; ii) membres représentatifs de huit groupements et les paires symétriques issues de l'analyse par vecteurs singuliers du ECMWF; et iii) huit paires issues de l'analyse par vecteurs singuliers du ECMWF-EPS choisies au hasard. Les résultats des expériences vérifiées montrent que la fiabilité statistique du ALADIN–LAEF s'améliore lorsqu'un groupement est appliqué, mais on ne perçoit aucune amélioration nette dans l'habileté du LAMEPS. Une étude de cas d'un événement de fortes précipitations confirme le résultat de la vérification de 50 jours. La validation montre qu'aucune des stratégies de groupement ne surpasse les autres.     

A Comparison of Breeding and Ensemble Transform Vectors for Global Ensemble Generation

To compare the initial perturbation techniques using breeding vectors and ensemble transform vectors,three ensemble prediction systems using both initial perturbation methods but with different ensembl...     

短期气候数值预报中的集合个数问题

利用中国科学院大气物理研究所IAP L2 AGCM1.1模式18年(1980～1997年,每年包含28个积分)的集合后报试验结果,讨论了数值预测中的最小集合个数问题.研究表明预测结果达到稳定所需的最小集合数随区域和年份而变化,对于我国汛期降水跨季度预测,最小集合数在全国范围为19～20个,对于东南区域为10个左右,东北、华北地区不应少于20个.当出现多平衡态时,最小集合数增大,这时应对各平衡态分别进行集合.研究结果同时还显示,对于有预报能力的地区,集合方法可以提高预测技巧及其可信度,而在负技巧区域,集台反使得预测效果更差,因而有必要进行系统性订正.     

Ensemble mean forecast skill and applications with the T213 ensemble prediction system

Ensemble forecasting has become the prevailing method in current operational weather forecasting. Although ensemble mean forecast skill has been studied for many ensemble prediction systems(EPSs) and different cases, theoretical analysis regarding ensemble mean forecast skill has rarely been investigated, especially quantitative analysis without any assumptions of ensemble members. This paper investigates fundamental questions about the ensemble mean, such as the advantage of the ensemble mean over individual members, the potential skill of the ensemble mean, and the skill gain of the ensemble mean with increasing ensemble size. The average error coefficient between each pair of ensemble members is the most important factor in ensemble mean forecast skill, which determines the mean-square error of ensemble mean forecasts and the skill gain with increasing ensemble size. More members are useful if the errors of the members have lower correlations with each other, and vice versa. The theoretical investigation in this study is verified by application with the T213 EPS. A typical EPS has an average error coefficient of between 0.5 and 0.8; the 15-member T213 EPS used here reaches a saturation degree of 95%(i.e., maximum 5% skill gain by adding new members with similar skill to the existing members) for 1–10-day lead time predictions, as far as the mean-square error is concerned.     

风暴尺度集合成员数对预报技巧的影响

利用WRFV2.2模式,对1977年5月20日发生在美国Oklahoma的典型超级单体风暴进行集合预报试验。采用蒙特卡洛法对不同区域初值扰动,对比分析成员个数的变化对预报技巧的影响,检验集合技术应用于风暴尺度系统的可行性及应用价值。结果显示,基于WRFV2.2模式的风暴尺度集合预报（storm-scale ensemble forecasting,SSEF）能够从热力场和动力场上改善单一确定性预报,并成功预报极端降水,表明SSEF具有较高的应用和研究价值;总体上预报技巧随成员数增加而增加,当集合成员数达到5-13时,预报技巧呈饱和特征,不同变量、不同扰动区域时的饱和成员数略有差异。     

Evaluation of a Regional Ensemble Data Assimilation System for Typhoon Prediction

An ensemble Kalman filter(EnKF) combined with the Advanced Research Weather Research and Forecasting model(WRF) is cycled and evaluated for western North Pacific(WNP) typhoons of year 2016. Conventional in situ data, radiance observations, and tropical cyclone(TC) minimum sea level pressure(SLP) are assimilated every 6 h using an 80-member ensemble. For all TC categories, the 6-h ensemble priors from the WRF/EnKF system have an appropriate amount of variance for TC tracks but have insufficient v...     

欧洲业务集合预报系统进展

由于大气初值与数值模式中物理过程存在不确定性等客观事实,集合预报无疑代表着数值天气预报未来前进与发展的方向,它标志着天气预报的预报范式转变,即用户不仅可以得到未来大气状态的单一现实,还可得到未来大气可能出现的一系列场景。文中扼要地梳理了欧洲全球业务集合预报与有限区域模式高分辨业务集合预报的研究动态与技术发展、基本问题及其未来最新发展方向,包括:1)欧洲中期天气预报中心的业务集合预报系统发展沿革及概况;2)欧洲国家主要业务高分辨率集合预报系统概况;3)当前业务集合预报存在的问题、挑战及未来前进的方向。文中除了关注欧洲中期天气预报中心的集合预报应用,还梳理了目前欧洲高分辨业务集合预报取得的成就,以引起有关研究人员的注意。总之,借鉴欧洲业务集合预报的发展思路,不仅有助于集合预报理论创新,还对发展集合预报业务有重要指导意义。     

Evaluation of the NMC Regional Ensemble Prediction System During the Beijing 2008 Olympic Games

Based on the B08RDP (Beijing 2008 Olympic Games Mesoscale Ensemble Prediction Research and Development Project) that was launched by the World Weather Research Programme (WWRP) in 2004, a regional ensemble prediction system (REPS) at a 15-km horizontal resolution was developed at the National Meteorological Center (NMC) of the China Meteorological Administration (CMA). Supplementing to the forecasters’ subjective affirmation on the promising performance of the REPS during the 2008 Beijing Olympic Games (BOG), this paper focuses on the objective verification of the REPS for precipitation forecasts during the BOG period. By use of a set of advanced probabilistic verification scores, the value of the REPS compared to the quasi-operational global ensemble prediction system (GEPS) is assessed for a 36-day period (21 July–24 August 2008). The evaluation here involves different aspects of the REPS and GEPS, including their general forecast skills, specific attributes (reliability and resolution), and related economic values. The results indicate that the REPS generally performs significantly better for the short-range precipitation forecasts than the GEPS, and for light to heavy rainfall events, the REPS provides more skillful forecasts for accumulated 6- and 24-h precipitation. By further identifying the performance of the REPS through the attribute-focused measures, it is found that the advantages of the REPS over the GEPS come from better reliability (smaller biases and better dispersion) and increased resolution. Also, evaluation of a decision-making score reveals that a much larger group of users benefits from using the REPS forecasts than using the single model (the control run) forecasts, especially for the heavy rainfall events.     

台风集合预报研究进展

台风数值预报是防台减灾的关键,而集合预报是体现和减少数值预报不确定性的常用方法。本文对近年来台风集合预报方法的研究进展进行了梳理和总结,涉及初值集合扰动、模式扰动技术以及基于统计的台风集合预报后处理技术。对全球几个主要集合预报系统的发展及我国的区域台风集合预报系统做了回顾。最后,在回顾的基础上,讨论和提出了关于台风集合预报仍存在的问题及未来可能的研究方向。     

中国冬季气温的集合典型相关分析和预报

以欧亚大陆地面温度、北半球500 hPa高度、热带印度洋SST(sea surface temperature)以及北太平洋SST为预报因子,通过典型相关分析(canonical correlation analysis,简称CCA)建立预报关系,然后用集合典型相关分析预报(ensemble canonical correlation prediction,简称ECC)方法预报中国冬季气温,并分析预报技巧及进行独立样本检验.结果表明,不同的预报因子对各个地区有不同的预报技巧,以欧亚大陆地面温度为预报因子预报技巧较高,而ECC模式对中国冬季气温有更好的预报能力,预报技巧高于任何一个单因子场的CCA预报;采用回归法的集合平均比简单的等权集合平均预报技巧更稳定.     

面向台风暴雨的集合预报成员优选订正技术研究及应用试验

本文以2013—2015年主要登陆台风暴雨过程为研究对象,利用ECMWF降水和台风路径集合预报以及中央气象台实时业务台风中心定位资料,在统计分析的基础上,提出一种业务上可用的针对单模式集合预报的台风降水实时订正技术(简称集合成员优选技术)。结果表明,在登陆台风暴雨过程预报中,集合成员优选技术对改进集合统计量降水产品有明显的帮助,并较ECMWF确定性预报产品有一定优势;该方法对改进短期时效预报产品的效果优于中期时效预报,对大暴雨评分的改进高于暴雨和大雨评分。另外,本文基于概率匹配平均(Probability Matching average,PM)和融合(FUSE)产品的计算原理,提出融合匹配平均(Fuse Matching average,FM)产品,结果表明,对36 h时效预报,优选10~15个成员的PM产品TS(Threat Scores)评分可达最优,大暴雨评分较确定性预报提高近10%;对60和84 h时效预报,FM产品大暴雨评分较确定性预报提高超过20%。     

一次特大暴雨过程高分辨率集合预报试验的检验和评估

针对2018年4月22日发生在湖北西部山地的一次特大暴雨过程,采用降尺度方案和显式对流参数化方案模式,开展了高分辨率对流许可尺度(3 km)的集合预报试验,并对全球集合预报(GEFS)和对流尺度集合预报(SSEF)的降水预报进行了对比评估,结果表明:(1)SSEF集合平均的雨量和落区预报均优于GEFS。(2)SSEF各成员的降水离散度分布更合理,因而具有更优的降水区间预报,其“离散度-误差关系”更优,能更好地给出预报误差的分布及其可能的大小。(3)SSEF的概率预报在所有空间尺度上均优于GEFS,且在短历时强降水上的优势更加明显。由此可见,针对此类山地暴雨过程,对流尺度集合预报相对于全球集合预报具有巨大的改进潜力。     

中国夏季降水多模式集成概率预报研究

基于TIGGE资料中的中国气象局（CMA）、欧洲中期天气预报中心（ECMWF）、日本气象厅（JMA）、美国国家环境预报中心（NCEP）以及英国气象局（UKMO）五个中心2007-2011年5月25日-8月31日中国地区逐日12-36 h、36-60 h、60-84 h、84-108 h、108-132 h与132-156 h累积降水集合预报资料,分别利用PoorMan （POOL）和多模式消除偏差（MBRE）两种方法对2011年各中心降水概率预报进行集成,并采用RPS和BS评分方法对预报效果进行评估。结果表明,对于12-156 h逐24 h累积降水量概率预报,多模式集成预报效果优于单模式预报效果,且多模式消除偏差概率预报效果最好;针对小雨、中雨以及大雨以上降水,PoorMan和MBRE概率预报较单中心预报效果均有提高,MBRE概率预报效果优于PoorMan方法。     

The Optimized Short-Range Ensemble Forecast Based on Singular Vector Calculations

In ensemble forecast,by summing up ensemble members,filtering the uncertainty,and retaining the common component,the ensemble mean with a better result can be achieved.However,the filtering works only when the initial perturbation develops nonlinearly.If the initial perturbation propagates in a linear space,the positive and negative members will counteract,leading to little difference between ensemble mean and control forecast and finally insignificant ensemble result.In 1 2-day ensemble forecast,based on singular vector (SV) calculations,to avoid this insignificance,the counteracting members originated from the same SV are advised not to put into the ensemble system together;the only candidate should be the one with the better forecast.Based on the ingredient analysis of initial perturbation development,a method to select ensemble members is presented in this paper,which can fulfill the above requirement.The regional model MM5V1 of NCAR/PSU (National Center for Atmosphere Research/Pennsylvania State University) and its corresponding tangent adjoint model are used.The ensemble spread and forecast errors are calculated with dry energy norm.Two mesoscale lows on the Meiyu front along the Yangtze River are examined.According to the analysis of the perturbation ingredient,among couples of counteracting members from different SVs,those members performing better always have smaller or greater spread compared with other members.Following this thinking,an optimized ensemble and an inferior ensemble are identified.The ensemble mean of the optimized ensemble is more accurate than that of the inferior ensemble,and the former also performs better than the traditional ensemble with positive and negative members simultaneously.As for growth of the initial perturbation,those initial perturbations originated from the summed SVs grow more quickly than those from the single SV,and they enlarge the range of spread of the ensemble effectively,thus leading to better performance of ensemble members.     

欧洲中心集合预报在中国区域中期时效离散度预报技巧关系分析

预测数值模式对于中期天气预报能力是集合预报的重要应用之一。为研究集合预报在中国区域中期时效(96～360h)预报技巧的预测能力,本文基于欧洲中期天气预报中心(ECMWF)全球集合预报系统2007—2011年500 hPa高度场和850hPa温度场预报数据,采取两种不同的离散度-预报技巧关系表征方法进行对比分析,结果表明:(1)以均方根误差为表征的预报技巧(RMS_(ENS))和离散度(SP_(RMSE))表现出季节循环特征,即冬(夏)季值高(低),这是大气内在性质的表现。而以距平相关为表征的预报技巧(AC_(ENS))和离散度(SP_(AC))没有显著的内在季节变化特征。(2)对比分析两种不同表征的离散度-预报技巧关系可知,以距平相关为表征的离散度-预报技巧关系更能反映中期时效的预报技巧,且850 hPa温度场较500 hPa高度场二者的相关性更好。(3)定量分析离散度-预报技巧关系表明,小离散度情况下更能体现高的预报技巧,但这种关系从96～360 h样本百分比下降了20%左右,而在大离散度情况下离散度-预报技巧关系相对弱一些,且随预报时效的延长样本百分比没有显著的降低。(4)样本统计显示中期各时效SP_(RMSE)和SP_(AC)二者一致的样本占59%～66%,并没有显示较高的一致性特征。上述分析结果为集合预报在中期时效预报技巧预测方面提供定性和定量的参考。     

集合数值预报在洪水预报中的应用进展

水文集合预报是近几年正在形成和发展的水文预报分支,其发展大致可分为两个阶段:第1阶段是1970年至20世纪末进行的长期径流预报,第2阶段从21世纪开始,主要学习气象数值预报中集合预报的概念在短期水文集合预报中的应用。目前,除了单一预报中心的集合预报系统在水文集合预报中应用外,多个预报中心的集合预报大集合也逐渐被应用于流域水文预报,甚至一些小流域的洪水预报。如利用TIGGE(THORPEX Interactive Grand Global Ensemble)集合预报驱动形成的大气-水文-水力的串联系统进行早期的洪水预警研究,将全球集合预报作为洪水模型输入的有限区域模式的初始条件和侧边界条件的研究。这些均表明,基于水文集合预报的洪水预报增加了预报附加值,并能够延长预警提前时间。以欧洲中期天气预报中心的欧洲洪水预警系统(EFAS)和美国NOAA的先进水文预报系统(AHPS)为代表,实现了集合预报在洪水中的实时业务预报,但仍存在数据处理和计算量大,以及如何基于集合水文预报做决策等问题。对于水文集合预报的前处理和后处理的各种技术已处于探索和验证阶段,如何更好地理解基于概率预报的洪水预警决策仍存在许多困难和挑战。