基于Local-BGM法的冷涡暴雨集合预报试验及评估检验  

The ensemble forecasting experiment and assessment of a cold vortex rainstorm based on Local-BGM

作　　者：张海鹏 智协飞[2] 李昊 黎振宇 张志强 黄晶 ZHANG Haipeng;ZHI Xiefei;LI Hao;LI Zhenyu;ZHANG Zhiqiang;HUANG Jing(Electric Power Research Institute,China Southern Power Grid Co.,Ltd.,Guangzhou 510700,China;Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters/Key Laboratory of Meteorological Disaster,Ministry of Education,Nanjing University of Information Science&Technology,Nanjing 210044,China)

机构地区：[1]南方电网科学研究院有限责任公司,广州510700 [2]南京信息工程大学气象灾害预报预警与评估协同创新中心/气象灾害教育部重点实验室,南京210044

出　　处：《气象科学》2023年第1期36-45,共10页Journal of the Meteorological Sciences

基　　金：国家自然科学基金面上资助项目(41575104);南方电网科学研究院有限责任公司基础性前瞻性项目(SEPRI-K213004)。

摘　　要：基于传统增长模繁殖法(Breeding Growing Mode,BGM)和局地增长模繁殖法(Local Breeding Growing Mode,Local-BGM)生成初始扰动成员,对一次冷涡暴雨过程进行集合预报试验,从多方面比较两种方案的预报效果,并且在邻域概率法(Neighborhood Probability,NP)中引入时间邻域,评估概率预报结果。结果表明,引入局地化思想的Local-BGM方案能够生成比传统BGM方案更合理的初始扰动,具有很明显的局地特征。对于扰动变量的预报,Local-BGM方案在均方根误差和离散度等方面均表现更好,同时能够提高各量级降水的预报技巧。邻域集合概率法能够综合各个集合成员预报的降水信息得到优于集合平均的概率预报,分数技巧评分更高。并且在考虑时间不确定性后,无论是控制预报、集合平均还是邻域集合概率法,分数技巧评分均有很大改善,并且降水阈值越大改善效果越明显,能够为极端强降水天气提供较为客观的概率预报信息。Based on traditional Breeding Growing Mode(BGM)and Local Breeding Growing Mode(Local-BGM),the initial perturbation members are generated respectively.The ensemble forecasting experiments for the cold vortex rainstorm process were conducted and the forecasting effects of the two methods were compared in various aspects.A time factor is introduced into the Neighborhood Probability(NP)method to evaluate the probabilistic forecasting results.It is detected that the Local-BGM can generate a more reasonable initial perturbation field than traditional BGM and has strong local characteristics.For the forecast of perturbation variables,the Local-BGM performs better in the aspects of root mean square error,spread,etc.Meanwhile the forecasting effect of each precipitation magnitude is also improved.The NP method can consider precipitation information of each ensemble member to obtain a better probabilistic forecast with higher Fractional Skill Score(FSS)than the ensemble mean forecasting result.After considering the time uncertainty,whether it is control forecasting,ensemble mean forecasting or neighborhood probability method,the FSS is greatly improved and the improvement is more obvious as the precipitation threshold increases.The spatiotemporal NP method can provide more objective probabilistic forecasting results for extreme precipitation weather.

关 键 词：集合预报 局地增长模繁殖法 降水预报 邻域概率法 分数技巧评分 

分 类 号：P457.6[天文地球—大气科学及气象学]