20 



SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 1 28 



available in the original data, I think we should expect nothing 

 smoother, for 21 other periods interfere with their own features, 



as well as 45^ months. Turning to curve a-^. the periods of and 



■ are successively removed, as shown in the dotted curves C and 



D. Here, again, nothing more can be removed. I now shift curve 

 B to 5 months later, and take the mean of it and curve D, yielding 



Fig. II. — The 452-month periodicity in Peoria precipitation as modified by 

 shorter periodicities, aliquot parts of 452 months. Left, Wolf numbers above 

 20; right, Wolf numbers below 20. 



the dotted curve E, to represent the 45^-month periodicity in pre- 

 cipitation at Peoria, 111., 1856 to 1939, when Wolf sunspot num- 

 bers > 20. The heavy full curve smooths out the interferences by 

 other periods, and the accidental errors, and shows an amplitude of 

 22 percent for the 45i-month periodicity.^^ 



11 I felt some doubt at one time whether this procedure, of computing subordi- 

 nate periods from mean values for long periods, would give approximately correct 

 results. But I satisfied myself in two ways. First, since the later stages of my 



