气候场的主分量逐步回归预测模型及应用

MODEL OF PREDICTION OF PRINCIPAL COMPONENT-STEPWISE REGRESSION OF CLIMATIC FIELD AND ITS APPLICATIONS

提出一种气候场的主分量逐步回归预测模型,该模型将气候场的预测变成对该场主分量的预测。提取对气候场变化有重要影响的500 hPa高度、太平洋及印度洋海温和全球海平面气压等含不同区域不同季节的因子场的前若干个含高信息量的优质主分量因子。通过相关筛选和双重检验逐步回归,建立气候场的主分量的预测方程,由此就建立气候场与多个因子场之间的联系,但是模型却仍然保持着两个场相关的特点。根据气候场的特征向量的近似不变性,将其与主分量配合进行反算,从而得到气候场的预测。以广东汛期降水预测作为试验例子,计算了它的主分量,分析了它的时空分布特征及分类,对2003—2005年的分布作预测并与实测分布作比较及进行误差检测,发现预测效能显著。有关预测的年限、区域范围、要素和因子的物理背景等问题有待进一步研究。

A prediction model has been presented of principal component stepwise regression of the climate field. The model changes the prediction of the climate field to that of the principal component of that field. From the 500 hPa height and the sea surface temperature of Pacific and Indian ocean and the sea-level pressure of global in various seasons and regions, excellent principal component factors have been extracted with high information content and important influence on the climate field. Through correlation screening and double test stepwise regression, a prediction equation is developed for the principal component of the climate field, and the relation of the climate field with multiple fields of factors has been established. But the model keeps the correlation between the two element fields. According to the approximate invariability of eigenvectors of the climate field, the prediction of climate field is obtained by return computation, together with the principal component. A test example is predicting the flood period rainfall in Guangdong. The principal component and the characteristics of special-temporal distribution and its classification are computed and analyzed. The prediction of field for 2003--2005 is made and comparisons with the field of observations and errors test are established. The results show that the predictive efficacy is remarkable. Questions concerning the predicted length （years）, regional coverage, physical background of elements and factors will need further research.