摘要
主分量分析在大气科学研究中的应用日益广泛,但对其应用的可行性和可靠性的研究则比较欠缺.本文对主分量分析的三种方法,即非旋转主分量、正交旋转的方差极大法和斜交旋转的Harris-Kaiser Case ⅡB^TB法,在区域气候分型研究中的应用进行了计算分析和精度拟合,同时进行了区域分割和Monte-Carlo模拟的统计试验.结果表明,非旋转主分量的分型拟合精度不佳,两种旋转主分量方法的效果很好,其中正交旋转又优于斜交旋转.区域范围的变化对旋转主分量方法的分型结果影响不大,对非旋转主分量分析的影响较大.样本大小的变化对主分量分析的各种方法无明显影响.
The principal component analysis has been applied frequently in the field of atmospheric sciences nowadays. There has not been much work on its app licability and reliability in the Chinese literature of atmospheric research. In this paper, three techniques, i. e., the unrotated principal component analysis, Varimax of orthogonal rotation and Harris-Kaiser Case II BTB of oblique rotation are compared for the regional climate pattern analysis and the accuracy fitted. The Monte-Carlo simulations and area separation tests are also carried out. The results show that, two rotated principal component methods are better while the unrotated principal component is not very accurate. The orthogonal rotation method gives slightly better results than the oblique rotation with the summer rainfall data of Liaoning Province. The patterns of the rotated principal component do not change very much with the change of area, the unrotated principal component fails in the test of changing area. The three techniques are all independent of the selection of sampling sizes.
出处
《大气科学》
CSCD
北大核心
1992年第6期649-658,共10页
Chinese Journal of Atmospheric Sciences
关键词
主分量分析
区域
气候
分型
研究
Principal component analysis
Orthogonal rotation
Oblique rotation
Regional climate pattern analysis.
