Abstract
Temperature anomaly patterns over Japan and their changes during the past 74 years (1901_??_1974) are examined using principal component analyses (eigenvector analyses) of mean July and August temperature fields. More than 80 percent of the cumulative variance can be explained by the first two components (Table 1). The spatial eigenvector patterns are shown in Fig. 2 and Fig. 3. From an examination of these patterns, it be-comes clear that the first component which is characterized by anomalies of one sign (positive) in all of the area represents the real temperature anomaly patterns of either “hot summer” or “cool summer”, and that the second component which is characterized by anomalies of opposite signs in the north and in the south represents the “north cool-south hot” pattern or the “north hot-south cool” pattern.
Time series of component scores (eigenvector coefficients) are presented in Fig. 4 and Fig. 5. In July, the score of the first component is negative from 1901 to the mid-1910's, but is frequently positive since mid-1950's, while the second component tends to have negative scores since mid-1950's. In August, abrupt changes occurred around 1920 (the first component) and 1950 (the second component). As shown in Fig. 6, different anomaly patterns appearns in 1901_??_1920 and in 1951-1974. During the first period, 1901_??_1920, the “cool summer” pattern appeared frequently and during the third period, 1951_??_1974, the “north cool-south hot” pattern appeared predominantly.
In order to examine the periodicities, a power spectrum analysis is applied to the first two component scores (Fig. 7). The power spectra of the first component show pronounced peaks in the period range 7 to 8 years in August and 5 years in July. In the spectra of the second component, two pronounced peaks at 21-year (July) and 10.5-year (August) periods are found. These periods resemble the 11-year sunspot cycle.
In order to clarify the dynamic cliatological implications of each component, the cor-relation coefficients of the component scores are calculated with the 500 mb heights over the Northern Hemisphere (Figs. 8, 9, 10, 11) and with the occurrence frequencies of various surface pressure patterns over East Asia (Table 2).
The first comoponent is related to the development and decay of the North Pacific anti-cyclone at the 500 mb level over East Asia. At the surface level, it is connected with the occurrence frequencies of the Type V (Summer weather pattern) and Type IVb (Frontal pattern : Front running west-east along the Pacific coast of Japan or in the Pacific).
The second component is related to the development and decay of the subtropical anti-cyclones at the 500mb level which appear over the Tibetan Plateau and the East China Sea. At the surface level, it is connected with the occurrence frequency of the Type IVa (Frontal pattern: Front running west-east over Japan).
