一种新的气候分型区划方法

A New Cluster Method for Climatic Classification and Compartment Using the Conjunction Between CAST and REOF

从理论上证明统计聚类检验(CAST)与旋转经验正交函数或旋转主分量分析(REOF/RPCA)用于气候聚类分型区划的关联性。研究表明,CAST在一定的意义上可认为是REOF/RPCA用于气象要素场(气候)分型区划的理论基础。由此,作者提出CAST与REOF/RPCA相结合的一种新的分型区划方法,并用仿真随机模拟资料和实例计算验证了理论与实际结果的一致性,从而证实了这种分型区划方法的有效性及其优点。

It is theoretically demonstrated that the Cluster Analysis with Statistical Test （CAST） by Yao C. S. （1997） has a close relationship with the rotated empirical orthogonal function or principal component analysis （RE- OF/RPCA） to climatic classification and cluster compartment. Then, a new cluster method for climatic classification and compartment is proposed, which is theoretically based on CAST combined with REOF/RPCA in order to decide to accept or reject the assumption of a climatic category according to the χ^2 test from CAST. Euclidean distance coefficient dij （or Cij ） defined by the standard raw data in classical cluster analysis is related to the correlation coefficient of the raw data, both of them possess certain function relationship. Therefore, the calculation steps of the new cluster method proposed are, first, putting each principal component from RPCA to a meteorological variable field instead of the elected each central station corresponding to climatic classifications, second, determining the number of station （or variation） in every climatic category by using the χ^2 test from CAST. Then, the PCA or EOF is once again made at the stations with large loading, corresponding to every climatic category, and on the basis of the above result, the last climatic cluster and classification may be obtained. The result of the research shows that the CAST may considered a theoretical basis of the REOF/RPCA applied to climatic classification and cluster compartment, because REOF is a data-analytic technique which is based on the Principal Components Analysis （PCA） developed by Hotelling （1933）, and it is not only applied to climatic classification and compartment but also used in other problems. In this paper, the theoretical relationship between CAST and REOF is verified by means of the statistical simulation experiment and real observational data calculations. For example, based on the daily data of maximum and minimum temperatures （1957- 2001 ） from 203 meteorological stations in China by means of CAST combined with REOF/RPCA, the patterns of compartment for the extreme temperature sequences from 203 stations in China are obtained. This example also verifies that the new cluster method in this paper for climatic classification is more reasonable and reliable to determine each climatic category by using REOF （or RPCA） twice.