基于最新全球集合预报系统的可预报性评估

为了提供有价值且可靠的概率(或者不确定性)预报,最新的全球集合预报系统已在美国国家环境预报中心日常业务运行,以满足社会需求。通过对各个关键要素的概率预报统计检验,可为广大用户提供这些概率预报的信心指数。但是预报(或集合预报)能力不仅取决于我们使用的预测要素,而且与时间和空间分辨率,极端事件或者高影响天气,以及预报时效有关。以大尺度天气系统预报为例,通常选择北半球500 hPa位势高度距平相关指数或概率指数表征模式的预报能力。如参照北半球500 hPa位势高度的距平相关指数(60%AC)或概率预报技巧指数(25%CRPSS),美国全球集合预报系统能够提供大约10 d的技巧预报。从全球集合预报系统输出的各预报要素,满足不同时空尺度需求的角度进行讨论,其可预报性(或预报极限)能够为模式研发人员、一线预报员和用户提供参考。尤其是对大气可预报性的深入研究,对于从科学与技术角度全面提升数值预报系统水平非常重要。当能够确定可预报性(或是预报误差)的真实来源时,科学家(包括模式研发人员)就能够有针对性地修改与完善。将传统的可预报性研究与改进的能够更客观地表述预报不确定性的集合预报相结合,所得可预报性将提供另一种有价值的参考。可预报性研究总体表明,全球集合预报系统对行星波、大尺度和天气尺度的系统(或者过程)可能分别具备约15、12、10 d的预报能力。对于热带天气过程的预报,如果进一步改善模式偏差和物理参数化过程,其MJO(Madden-Julian Oscillation)预报技巧可以延长至32.5 d。

An assessment of predictability through state-of-the-art global ensemble forecast system

In order to provide valuable and reliable numerical guidance include probabilistic forecast,state-of-the-art global ensemble forecast system has been implemented into National Centers for Environmental Prediction daily operation to service the general public.The forecast skills for various elements through statistical verification are offering the levels of confidence to general users when they apply these guidance.However,the skills are all differents rely on the forecast elements,spatial and temporal resolutions,and special events or the forecast extremes.As an example of large scale pattern prediction,the global ensemble forecast system (global ensemble mean) could provide about 10 days skillful forecast on average based on 60% Northern Hemisphere 500 hPa geopotential height anomaly correlation.The predictability or limit of prediction we will discuss in this article could be a good reference for model developers,stakeholders and general users in terms of their requirements from various forecast elements to the different spatial-temporal scales.Especially,the investigation of sources predictability are very important to advance our numerical system in terms of science and technowledge.When we understand the source of predictability,the scientists will know where to work on,and how to improve the system.To combine the traditional predictability study and recently developed global ensemble forecast system with full representation of uncertainty,the predictability and limit of prediction through state-of-art global ensemble forecast system could be another valuable reference.As a summary of predictability investigation,there could be about 15-,12-,10-days forecast skills for planetary wave,large scale and synoptic scale patterns respectively.For tropical prediction,a skill of Madden-Julian Oscillation could be extended to 32.5 days if we could reduce model bias and improve others.