Analysis of Air Temperature Distribution over Northeast of China using Principal Component Analysis

Abstract

The north-west district (Heilungjiang-sheng, Kirin-sheng and Liaoning-sheng) of China covers a wide region, approximately 3.3 times of the total area of Japan, and comprise various types of geographical features. As a result the difference in climatic conditions are considerable.
Thus, a multidimensional time series analysis is useful for the clarification of the distribution patterns of climatic elements and their fluctuations.
In this paper, the monthly data of air temperature and air pressure at 35 sites from 1961 through to 1978, were used for the principal component analysis. The principle patterns of air temperature distribution were obtained from the scattering diagram of eigenvector of each component.
The changeability and periodicity of principle pattern were clarified from the time series of scores of each principal component and its spectral analysis.
The results of principal component analysis of air pressure were used to trace the principle patterns of air temperature to its origin.
The results obtained are summarized as follows:
(i) The first three components can explain more than 80% of the cumulative variance. The spacial eigenvector patterns are shown in Fig. 2. The first component (73.0% of total variance) represents “over the entire area” or “cool over the entire area” patterns of monthly air temperature distribution. The second component (11.4% of total variance), consists of the “east hot—west cool”or“east cool—west hot” patterns.
(ii) The prevailing patterns of air temperature distribution change from year to year and the aspects of the change are different for each month as shown in Fig. 5. For example, the prevailing patterns in summer (July) are “cool over the entire area” pattern in 1961-1966, “north hot—south cool” in 1967-1970, and “hot over the entire area and north hot—south cool” patterns in 1971-1978.
The spectrum analysis of the first and second components for July showed that these two components have the periodicity of 5-6 years and 4-5 years, respectively. From the same analysis for monthly data of the first and second components, it was clarified that these components have a periodicity of 36 months.
(iii) Using the values of the first component, the five patterns of monthly air temperature distribution in the north-east district in China were classified as very hot month (A), slightly hot month (A_B), normal month (C), slightly cool month (B_A) and very cool month (B). The frequency of the hot months (A+A_B), normal months (C) and cool months (B+B_A) were 32%, 40% and 28%, respectively.