lO SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 95 



errors in solar-constant work which, combined, produce still more dis- 

 turbing distortions in these more meager averages than those occurring 

 in the general mean curves representing more numerous data. On the 

 whole, however, it seems clear that the first half and last half of the 

 data are in fair accord as to the course of temperature departures 

 following sequences of solar change. 



Reverting to figure 2, the principal contrasted temperature features 

 corresponding to rising and falling solar sequences are large, and even 

 surprisingly so. In many of the months, temperature changes at 

 Washington exceeding 5° Fahrenheit are found in each curve of the 

 pair. In some months, as October, December, and February, the larger 

 ranges even exceed 10° Fahrenheit. Thus we have come upon de- 

 partures from normal temperature which have nearly as large ranges 

 as the largest of those ordinarily to be met with from day toi day in 

 the weather. It is easy to conceive, furthermore, that since the ap- 

 parent effects last for many days, there may have been individual 

 cases when previous or immediately succeeding solar causes would 

 have produced temperature features in unison with those of the mean 

 curves, whereby individual departures produced by solar causes in 

 such cases may have been nearly or quite twice as large as those here 

 shown. In short, if we admit, as seems justified, that these are the 

 temperature effects produced by solar changes, then we must concede 

 that solar changes are a main if not the principal cause of tempera- 

 ture changes in weather. That average changes of temperature rang- 

 ing sometimes as much as 10° Fahrenheit should be thus produced is 

 perhaps difficult to account for in theory, for the average solar changes 

 discussed here cannot much exceed 0.5 percent. But the curves speak 

 for themselves as to the facts. 



We may now inquire whether the fourth criterion is fulfilled. Fig- 

 ures 5 and 6 show the general averages for 12 years, 1924 to 1935, for 

 St. Louis, Missouri, and Helena, Montana. The same dates given in 

 table I as incipient dates of solar changes were used, of course, for all 

 stations. It will be seen at once that the oppositeness of features that 

 confronted us in figure 2 is also found almost universally in figures 

 5 and 6. Without actually reproducing here the curves to show it, 

 the reader may also be assured that our criteria numbered 2 and 3 

 are also fulfilled in the St. Louis and Helena data. 



So the observed variations of the sun, hitherto unrecognized in mak- 

 ing forecasts, seem to be main causes of temperature changes in 

 weather. Their effects appear to last for at least 2 weeks. Unfor- 

 tunately, this cannot be immediately tested as a new method of fore- 



