张华龙,伍志方,肖柳斯,涂静. 2020. 基于因子分析的广东短时强降水预报模型与业务试验[J]. 气象学报, (0):-, doi:10.11676/qxxb2020.076
基于因子分析的广东短时强降水预报模型与业务试验
A probabilistic forecast model of short-time heavy rainfall in Guangdong based on factor analysis and its operational test
投稿时间:2019-11-19  修订日期:2020-09-20
DOI:10.11676/qxxb2020.076
中文关键词:  概率预报模型,短时强降水,预测敏感性,因子分析,暖区降水
英文关键词:Probabilistic forecast model, Short-time heavy rainfall, Prediction sensitivity, Factor analysis, Warm sector rainfall
基金项目:国家重点专项(2019YFC1510404)、公益性行业(气象)科研专项(CYHY201506006)、广东省科技计划项目(2019B02028016)、中国气象局预报员专项(CMAYBY2018-053)、中国气象局强对流预报技术专家创新团队
作者单位E-mail
张华龙 广东省气象台 zhlchris@126.com 
伍志方 广东省气象台中国气象局广州热带海洋气象研究所/广东省区域数值天气预报重点实验室 zhifang_wu@tom.com 
肖柳斯 广州市气象台 xls104@126.com 
涂静 广东省气象台 375834995@qq.com 
摘要点击次数: 22
全文下载次数: 27
中文摘要:
      [资料]利用每日4次0.125°×0.125°的ECMWF-Interim再分析资料和广东2009-2018年地面气象站逐时雨量观测的短时强降水数据集,[目的]针对广东不同季节、不同地域的短时强降水,以提高命中率的同时控制空报率为目的,提出了基于显著性和敏感性评价的物理量优选和因子分析方法,用于构建分期分区的广东短时强降水概率预报模型。[方法]以参数显著性和预测敏感性为标准,在49个待选物理量中挑选18个既与多年平均态存在明显差异,又具有较低空报率的物理量,并应用方差最大正交旋转的因子分析方法将遴选物理量组合成表征大气不同环境条件的6个因子;为使组合因子更具适应性,基于因子偏离度特征对广东前、后汛期不同区域各自独立建模,构建分期分区短时强降水逐6小时格点概率预报模型。[结果与结论]汛期业务试验表明,模型对短时强降水发生概率预报效果较好。对2019年汛期模型每天两次起报的12h预报时效内概率产品进行格点检验,以训练期最优TS对应的固定概率作为预测概率阈值,全省大部分区域TS均达0.25以上,最高达0.42,平均TS较ECMWF-Fine业务模式在前、后汛期的提升幅度分别为0.23与0.21,南部沿海TS提升幅度最大,并且模型在提升命中率与降低空报率之间取得较好的平衡。个例分析表明,对于ECMWF模式常漏报的广东暖区短时强降水,概率预报模型具有明显优势,尤其能为天气尺度弱动力强迫的强降水早期预警提供更多有效信息。
英文摘要:
      [Aim] In order to improve the hit ratio and reduce the false alarm ratio of short-time heavy rainfall (SHR) in Guangdong, a method combing physical parameters selection based on significance and sensitivity evaluation and factor analysis is proposed to construct probabilistic forecast model in different periods and regions. [Data] The ECMWF-Interim reanalysis data of 4 times per day with 0.125°spatial resolution and hourly gauge rainfall data sets from 2009 to 2018 in Guangdong are used. [Method] On the basis of significance and prediction sensitivity evaluation, 18 physical parameters which obviously deviate from multiyear average state and could reduce the false alarm ratio are picked out from 49 parameters. The Varimax Orthogonal Rotation method is employed to recombine the selected parameters into 6 factors. These factors respectively reflect different environmental conditions. In order to optimize the model, factor analysis is separately applied to different regions, the pre-flood and post-flood seasons according to the spatial-temporal features of factor deviation rates. Based on the weighted combination of factors, the per-six-hours probabilistic grid forecast model of SHR in different regions and periods is constructed. [Results and Conclusions] The forecast model gains impressive results in operational experiment during flood season. During the period from April to September 2019, grid verification is carried out on the probabilistic model starting twice a day with forecasting span of 12 hours. A determined probabilistic threshold corresponding to the optimal TS in training period is taken as the forecast probability threshold of SHR in independent region and period, then the calculated threat skill (TS) in most of Guangdong is above 0.25 and the highest one is 0.42. Compared to the operational ECMWF-Fine precipitation forecast, the average TSs of the probabilistic model increase by 0.23 and 0.21 in pre-flood and post-flood seasons, respectively, with the greatest improvement in southern coastal area. Moreover, the model achieves a good balance between increasing the hit rate and decreasing the false alarm rate. Cases analysis shows that the probabilistic forecast model has obvious superiority in SHR forecasting in warm sector which is often missed by ECMWF-Fine precipitation forecast, and can provide more valuable information for the early warning of SHR in weak dynamic force synoptic environment.
查看全文   查看/发表评论  下载PDF阅读器
分享按钮