摘要
天气预报模型中的预报因子一般都有较好的预报能力,因为它们是根据预报对象的特点、预报因子的物理意义和预报经验,使用一定的技术方法而精心筛选的.但是预报因子在每次天气预报中的性能表现有清晰和模糊之分.从集合预报和近邻估计两种方法的基本思路出发,定义了描述因子模糊性的统计量———变异系数,并用此来识别多元回归模型中的模糊因子;从集对分析的基本理论出发,推导了适合于多元回归分析的联系度公式,并借此来处理模糊因子.在此基础上建立基于集对分析的天气预报多元回归模型,比传统模型明显地提高了天气预报准确率.
The predictors of weather forecast models have better forecast ability in general. Because they are specially selected according to the characteristics of predictands, physics meaning of predictors and forecast experience, and by means of certain technical methods. But sometimes when the situation has changed, predictors of multiple regression weather forcast models may appear in clear or fuzzy state in every forcasting. The fuzzy predictors can be identified and handled correctly by means of theory of Set Pair Analysis and method of variability coefficient. Forecast accuracys of multiple regression weather forcast models based on Set Pair Analysis are raised obviously.
出处
《科技通报》
北大核心
2004年第2期151-155,共5页
Bulletin of Science and Technology
基金
浙江省气象局科研课题(200118)
关键词
大气科学
天气预报
多元回归
模糊因子
集对分析
atmospheric science
weather forcasting
multiple regression
fuzzy predictor
Set Pair Analysis
