Logistic回归及其在概率降水预报中的应用  

Logistic Regression and Its Application for Probability Forecast of Precipitation

作　　者：纪玲玲[1] 王昌雨[1] 张志华[1] 

机构地区：[1]解放军理工大学气象学院,江苏南京211101

出　　处：《解放军理工大学学报（自然科学版）》2003年第5期92-94,共3页Journal of PLA University of Science and Technology(Natural Science Edition)

摘　　要：介绍了一种气象上较少应用的 Logistic回归的原理和计算方法 ,并利用 1 999年 3月 1日至 2 0 0 1年5月 30日期间 ,国家气象局 T1 0 6模式输出的 80个物理量场及其导出场资料和福州地区降水的 0、1化资料 ,采用 Logistic回归方法 ,制作春季福州 2 4 h降水概率预报 ,并与事件概率回归方法作了比较。结果表明 ,该方法对二分类变量有较好的预报效果。This paper introduced the principle and method of logistic regression, which is seldom used in weather forecasting. The predictive equations of the precipitation in Fuzhou station in spring for 24 hours are made by the Logistic regression approach. It employs 80 physical field data by T106 numerical model of Stated Metorological Administration and rainfall 0-1 data from Fuzhou station during 24 hours from 1 March 1999 to 40 May 2001, and compares it with REEP method. The fitting results of the equations are ideal for the dichotomous variables.

关 键 词：概率预报 LOGISTIC回归 事件概率回归 

分 类 号：P456.8[天文地球—大气科学及气象学]