GRAPES-GEPS环流集合预报的分类释用方法研究与检验

Classification interpretation method and verification of circulation ensemble forecasts in GRAPES-GEPS

中文关键词:  集合预报  分类释用  Ward聚类  环流  检验分析

英文关键词:Ensemble prediction  Classification interpretation method  Ward analysis  Circulation  Verification

中文摘要:

      集合预报在数值天气预报体系中具有重要地位，因此如何有效提取集合样本信息以提高集合预报技巧一直是一个重要课题。基于中国全球集合预报业务系统（GRAPES-GEPS）的500 hPa高度场集合资料开展对环流集合预报的分类释用方法研究，并对集合聚类预报结果进行了检验分析。通过在传统Ward聚类法中引入动态聚类的“手肘法”方案，发展了环流集合预报分类释用方法。针对该方法的个例分析表明，对于中国中东部地区环流集合预报的聚类释用方法能够有效地划分出最有可能发生的环流形势类型并提供发生概率。确定性预报综合检验结果显示，集合预报聚类结果中发生概率最高的集合大类相对于集合平均的预报技巧有明显提升，并随着预报时效的延长提升更明显。总体来看，通过集合预报的分类释用方法划分环流形势类型可以为天气预报提供参考依据，具有实际应用价值。

英文摘要:

      Ensemble prediction is playing a vital role in the development of numerical weather prediction system. Hence, how to effectively extract the information of ensemble swatches to improve ensemble forecasting skills has always been an important issue. Based on the 500 hPa geopotential height data from the China global ensemble forecasting operational system (GRAPES-GEPS), the ensemble forecasts have been classified by a cluster analysis approach, and the cluster results are further verified. By introducing the dynamic "elbow" cluster scheme, a classification interpretation method for circulation ensemble forecast is developed, and the related deterministic forecast verification is conducted using the GRAPES-GEPS real-time ensemble forecast dataset. A case study using this method indicates that the clustering of ensemble forecasts of 500 hPa circulation field over central and eastern China can efficiently classify the circulation types and meanwhile provide corresponding probabilities of their occurrence. The deterministic forecast verification results show that with the forecast lead time increase, the primary clusters of ensemble forecasts have significantly improved prediction skill scores together with the highest probability compared to those of the ensemble mean. Generally speaking, the classification interpretation method for circulation ensemble forecasting can provide a quite helpful reference for weather forecast, and it is of great value for potential application to operational weather forecast.