376 A STUDY OF CORRKLATIONS AMONci TERRKSTRIAL TEMPERATURES. 



long periods of tiiiie. We have shown how a phenomenon of this kind can be 

 detected, even if it lasts in each special case through little more than a single rotation 

 of the sun, by the method of time-correlation. The following considerations may 

 guide our coui'se of thought on the subject. 



Let us grant that on any occasion a region of the sun extending to, at least near 

 the equator, and hotter than the photosphere in general, is carried past the apparent 

 solar meridian by the sun's rotation. During a period of ten days it will be sufficiently 

 near the meridian to produce a rise in terrestrial temperatures. Then, as it disap- 

 pears, the temperature will begin to fall until the region reappears on the sun's eastern 

 limb. Then there will be another rise in the temperature, showing a rhythmical 

 movement of the latter. What we have to do is to inquire into the fluctuations of 

 temperature with a view of determining whether there can be found any rythmical 

 tendency among them to recur at the end of about 26 days. This is most completely 

 and rigorously done by searching for correlations between terrestrial temperatures at 

 any one epoch, or through one term, and during the following terms up to 26 days or 

 more. To discover the effect it seems desirable to take terms as short as five days, and 

 to carry their study continuously forward. It is then certain that, if any exceptionally 

 hot or cool region of the photosphere has been carried past our solar meridian, the 

 effect will be at its maximum during at least some one term. A study of the temper- 

 atures during the five terms following will then show what changes in terrestrial tem- 

 perature have taken place while the special region was moving around and returning 

 again to the solar meridian. 



I have chosen for this research the temperatures at San Diego because they are 

 fairly steady, and it chanced that the data for 5-day terms were available through a 

 period of more than 30 years, and therefore nearly 400 synodic rotations of the sun. 

 The research was confined to this station more through practical considerations than 

 because it was absolutely the best. If the clearest result is to be brought out, stations 

 in some continental interior, where the temperature is little affected by the ocean, and 

 where the irregular fluctuations are as small as possible, should be preferred. More- 

 over, as the effect sought for is common to the whole globe, the mean of the largest 

 practical number of such stations should be used. But the writer conceives that a 

 fairly certain result can be derived from San Diego alone. 



The method by which the periodicity is to be detected is that developed in § 2. 

 We take the departures of temperature during a number of consecutive five-day terms, 

 as great as we choose. In the present case we have cliosen six, making a period of 

 thirty days. The departui'es during the six terms of this period are designated as 



^11' ^P ^2' ^3) ''j; '*5 



