It is one of the most important but rather paradoxical problems of the world-meteorology that the seasonal correlation changes sign suddenly and the reversed correlation becomes to exist in the new epoch as good as it did in the preceeding period. The present investigation will certainly throw some lights on this point, though it is not yet quite conclusive.
The author examined the seasonal correlation for 40 years (1885-1924) between the summer temperature (June, July and August) in Tohoku district (Average temperature for Akita, Miyako, Yamagata, Ishinomaki and Niigata is taken) and the winter temperature at Nemuro. He found that the winter temperature to be correlated with must be that of January for a certain group of years and that of February in the other group. Moreover he found that the sense of correlation must be reversed in some years. All these, according to his idea, are ruled by the activity of the sunspot.
The laws proposed by the author are as follows:-
(I) According to that the annual Wolf-Wolfer's number for the previous year was above or below 5, January or February temperature must be taken as to be correlated with. When, however, the number was more than 50 and in the decreasing phase, then the January temperature must be referred to.
(II) Generally the correlation is positive. It is reversed when (1) the sunspot number of the previous year was less than 10 and (a) increasing very slowly within the limit of 2 per year, or (b) decreasing very rapidly over the limit of 9 per year, or (2) when the number of the previous year was in a domain between 26 and 50, and in the decreasing phase.
These laws are only applicable to the case above mentioned. The coefficient for the positive correlation was 0.79±0.04 in the case concerned, the number of years being 31 and the negative correlation was -0.71±0.11, the number of years being, 9.
View full abstract