l8 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 94 



tuations from normal values of temperature and rainfall, as derived 

 from the tables of " World Weather Records " were so large that they 

 obscured the principal trends which might reveal periodicities cor- 

 »responding to those found in solar radiation already mentioned. As 

 the computations proposed were very laborious and the available com- 

 puters inexperienced, it seemed necessary to restrict the smoothing 

 process to be employed to one of great simplicity. Hence the traveling 

 mean of 5 months was chosen. For instance, the values employed for 

 March and April of any year would be represented as follows : 



March = Ja"- + Feb. + Mar. + Apr. + May 



5 

 . .._ Feb. + Mar. + Apr. + May + June 



In computing the monthly departures themselves, the mean values 

 used throughout were those found in the first volume of " World 

 Weather Records," neglecting those found in the second volume. It 

 was desirable to use the same normals at all times because we wished 

 the departures used to be homogeneous throughout the entire interval 

 of years examined. Departures from these normal values were com- 

 puted for the monthly mean temperatures of a great many stations in 

 all parts of the world, and the 5-month traveling means were computed 

 from these departures as described above. 



With regard to precipitation a modified course was pursued. It 

 is well known that the precipitation at most stations is seasonal, and 

 at many stations the seasons present extreme variations in normal 

 values. Hence a departure from the normal value, expressed in inches 

 or centimeters, which would be moderate if it occurred in the rainy 

 season, would be immense and perhaps unheard of if it occurred in 

 the dry season. But it was indispensable for our purpose that the 

 departures from normal should be comparable whether occurring in 

 the wet or the dry season. Hence the monthly mean precipitations 

 were first expressed in percentages of the normal values, and then 

 smoothed by taking 5-month traveling means. It would perhaps have 

 been preferable to smooth the percentage values by taking the fifth 

 root of the product of five values, but for simplicity the monthly mean 

 percentage values of the normal were smoothed in exactly the same 

 way as the departures from normal temperatures. 



10. Amplitudes of Periodicities Diminished by Smoothing 



It was appreciated that the 5-month traveling means of weather 

 data could not yield the full amplitudes of periodicities as short as 





