There are characteristic patterns in the airflow system which relate to distributions of surface climatic elements. Japanese climate has been studied using the airflow patterns (Jacobs, 1946; Maejima, '1954). The airflow patterns were set up from the gradient wind direction. But, the gradient wind direction might not be used, especially in the large area. The airflow system spreads over the Japanese Islands. By classifying the airflow system, the macro-scale airflow patterns in the Japanese Islands are set up.
From Hokkaido to Kyushu, strong winds blow in winter and spring. Then the area is chosen to set up the macro-scale airflow patterns in the Japanese Islands (Fig, 1). In the daytime wind is stronger than in the night, so the observed values of wind are representative for time and space. The airflow patterns are represented by the daily 15 JST wind direction and speed in 1973.
The surface wind distribution was usually classified subjectively. But, if the area was more extensive, the subjective classification was more difficult. While, cluster anaysis was adopted to classify the area (McBoyle, 1971). The cluster analysis, one of the objective methods, can evaluate the difference between the distributions even if they can not be classified subjectively. Then, the surface wind distributions are classified by cluster analysis in this study.
The distance index of two surface wind maps is defined from the sums of wind difference at the same station between them. Also, the revolved definition of distance, the distance between the new cluster and the former cluster, is calculated by the group average method in this ` study. The process of the cluster analysis and the relation between clusters are expressed on dendrogram (Fig. 2). On the dendrogram, the cluster sometimes will not be combined in spite of the increase of distance. Each cluster is well unified on that occasion. The distance index 2. 2 corresponds to that case. There are 12 reasonable clusters. They are grouped into 3 principal clusters.
The macro-scale airflow patterns are set up from the 12 clusters. The resultant winds of them are effective expression of surface airflow patterns (Fig. 3). Characteristics of them are as follows. Southerly winds (a
1_??_a
4 type): a
1, sea breezes blow all over the Japanese Islands; a
2, southwesterly to west-southwesterly strong winds blow all over the Japanese Islands; a
3, sea breezes blow in the Japanese Islands except for the Pacific Ocean side of Hokkaido and Tohoku District; a
4, southwesterly winds blow in southwestern Japan. Northeasterly winds (b
1_??_b
3 type): b
1, easterly-winds-like sea breezes blow on the Pacific Ocean side, while roughly northerly winds blow on the Japan Sea side of southwestern Japan b
2, northeasterly winds blow in southwestern Japan; b
3, northeasterly to northerly winds blow in central and southwestern . Japan. Westerly winds (c
1-c
5 type): c
1, westerly winds blow all over the Japanese Islands; c
2, westerly strong winds blow in the Japanese Islands except for Hokkaido; c
3, northerly winds blow all over the Japanese Islands; c
4, roughly westerly winds blow all over the Japanese Islands; c
5, roughly northwesterly winds blow in northeastern Japan, while roughly northerly winds blow in southwestern Japan.
To clarify the daily macro-scale airflow patterns in other years, discriminant analysis is used. It classifies the unknown sample according to the established type. Then, the macro-scale airflow patterns from 1968 to 1977, are clarified. The annual variation of each airflow pattern's frequency is exhibited (Fig. 4). Southerly winds (a
1_??_ a
4 type) occur in summer. Northeasterly winds (b
1_??_b
8 type) occur mainly in spring and autumn.
View full abstract