2 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. Ill 



curves for May, it was seen that a decided indication of a variation 

 of about that period occurred, but sometimes from i to (rarely) 



3 days before or after it was expected. However, upon scanning 

 the whole 45 years of May values, it was clear that the period was 

 either a little shorter than 6"^ days (best value found to be 6.6456 ^ 

 days which is about 1/55 of a year) or else it must be about 1/54 

 or 1/56 of a year. Now turning to the tabulation for November, it 

 was found that only the period 6.6456 ^ days could serve. For if 

 either 1/54 or 1/56 of a year were adopted, the November curves 

 would be a half period out of phase with those of May. 



Afterward, the solar-constant values of Montezuma, 1924 to 1945. 

 were critically studied. They proved to exhibit this period of varia- 

 tion, and with perfect regularity. The displacements of from i to 

 (rarely) 3 days in phase in Washington temperature departures 

 from the normal do not occur in the solar variation. Considering the 

 complexity of the earth, its atmosphere, and meteorological reac- 

 tions, it is not surprising that such displacements of the phases of a 

 terrestrial response to regular solar pulses should occur. But they 

 impair the value of the solar period for forecasting purposes. It was 

 shown, however, in Smithsonian Miscellaneous Collections, vol. 107, 

 No. 4, that the temperature departures caused by this periodic solar 

 variation range from 2° to 20° F. at Washington and also at St. Louis 

 and at Helena, Mont. 



PROCEDURE OF INVESTIGATION 



Proceeding with the investigation of New York City temperature 

 departures, I made monthly tables of the four or five recurrences of 

 the 6.6456-day period in each month. These covered 21 years from 

 1928 to 1948, making 1,154 separate recurrences in all. The tables 

 were based on January 17.000, 1946, so as to be comparable directly 

 with the Washington work reported in Smithsonian Miscellaneous 

 Collections, vol. iii, No. 6. The following sample, table i, is for May 

 1929. 



It will be noticed that only six values occur in lines i and 4 of the 

 table and that the vacancies are distributed to the last and first 

 columns. These adjustments take into consideration whether the 

 fractions in the phase dates exceed 0.5 or not. The vacancies are 

 distributed so that, in the year, as many fall in the first column as 

 in the last. Proceeding in this way the New York data were all 

 tabulated from 1928 to 1948 by individual monthly means. It was 



