本文已被：浏览 1964次 下载 1887次
投稿时间：2014-09-16 修订日期：2015-01-02
投稿时间：2014-09-16 修订日期：2015-01-02
中文摘要: 利用我国黄淮地区1961—2010年50年6—8月的日降水资料，采用REOF和t检验的方法，将中国黄淮地区夏季降水分成Ⅰ—Ⅴ区。5个区域进行差异性t检验表明5个区域之间（彼此）差异显著，说明了区域划分的正确性。在此基础上利用1999—2007年6—8月站点日降水资料以及CFSv2模式后预报的日降水资料建立了5个区域内共5个代表站点的夏季日降水概率预报方程，并进行了确定性、概率性预报检验和业务试用检验。对各区域内5个代表站日降水量的确定性预报检验表明：Logistic回归降水概率预报方程的TS评分要高于CFSv2模式预报和T213的集合预报平均，空报率也低于CFSv2模式预报和T213集合预报平均，但是漏报率却略高。各区域代表站日降水量的概率预报Brier评分检验表明：Brier评分均不超过0.2，大大低于T213集合预报所得概率预报的Brier评分分值，说明本文Logistic回归方程的概率预报较为可靠。Brier技巧评分表明：Logistic回归降水概率预报方程各站的BSS技巧评分都大于0.0，说明各站的预报技巧高于检验样本气候概率的预报技巧，且高于T213集合预报的Brier技巧评分，说明在分区基础上建立Logistic回归降水概率预报方程的方法是有预报意义的。
Abstract:Based on the rain gauge of summer daily precipitation over Huang Huai Region from 1961 to 2010, the REOF method and t test are employed to divide the Huang Huai Region into five areas. The t test of the five areas shows that the five areas have the significant differences, which illustrates the correctness of the division. Based on this, the summer probability regression prediction equations of the five representative sites in the five areas are established using the daily rain gauge data and the daily precipitation reforecast products of CFSv2 numerical model from June to August during 1999-2007, and related verification was carried out by the deterministic and probabilistic way. The deterministic verification of precipitation probability forecast equations established by five representative stations shows that the Threat Score of logistic regression precipitation probability equations is higher than that of the CFSv2 model and the ensemble average of T213 ensemble prediction, the false alarms rate is lower than that of the CFSv2 model and the average of T213 ensemble forecast, but the missing rate of logistic regression is slightly higher. In this paper, the probabilistic forecast verification of precipitation probability forecast equations shows that the Brier Score of logistic regression is lower than 0.2, which is much lower than that of the probability of forecast of T213 ensemble forecast. This implies that the logistic regression has higher reliability. The Brier Score Skills of the logistic regression precipitation probability forecast equations for five stations are bigger than 0.0, which means the prediction skills of precipitation probability forecast for the five representative stations are higher than those of climatical probability, higher than those of the T213 ensemble forecast. Therefore, logistic regression precipitation probability equation based on the partition is an effective and feasible method.
keywords: logistic regression, probabilistic prediction of precipitation, statistics downscaling, REOF
文章编号： 中图分类号： 文献标志码：
基金项目:国家自然科学基金项目(41475044、91437104和91224004)，中国气象局气象关键技术集成与应用项目(CMAGJ2013ZX ZH1)、国家科技重大专项(2013ZX07304 001 1)和中国气象局气候变化专项(CCSF201333)共同资助
引用文本：
赵琳娜,董航宇,吴亮,王彬雁,白雪梅,党皓飞,2015.黄淮地区夏季日降水分区概率预报方法研究[J].气象,41(12):1503-1513.
ZHAO Linna,DONG Hangyu,WU Liang,WANG Binyan,BAI Xuemei,DANG Haofei,2015.Daily Precipitation Probability Forecast Based on Climatic Region over Huang Huai Region in Summer[J].Meteor Mon,41(12):1503-1513.
赵琳娜,董航宇,吴亮,王彬雁,白雪梅,党皓飞,2015.黄淮地区夏季日降水分区概率预报方法研究[J].气象,41(12):1503-1513.
ZHAO Linna,DONG Hangyu,WU Liang,WANG Binyan,BAI Xuemei,DANG Haofei,2015.Daily Precipitation Probability Forecast Based on Climatic Region over Huang Huai Region in Summer[J].Meteor Mon,41(12):1503-1513.