ANALYSIS OF TERRESTRIAL RECORDS 



87 



The Individual versus the Group in Cycle Analysis — A group of tree records 

 represents a geographical locality which may extend from an acre to a square 

 mile or more. Whenever averages of such tree records are made, it means 

 that the trees in the group cross-identify in the spacing of well-characterized 

 rings. Therefore we recognize three methods of reaching a result for the 

 group: (1) Analysis of individual records followed by a frequency periodogram 

 combining all the results (see fig. 35c); (2) simultaneous (or concurrent) 

 analysis of the individuals (described on page 47) and (3) analysis of the mean 

 curve of the group (35a). This last was the method used by the writer in pre- 

 vious published reports. Methods (2) and (3), simultaneous analysis and 

 mean curve analysis, being single analyses result in sharply defined cycle 

 lengths as compared to (1), which consists of analyses of many individual 



5 6 7 8 9 10 II 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 



Fig. 35 — Frequency periodograms by three processes applied to the 

 same data (Central Pueblo Area; 15 groups — 73 trees). 



a. Analysis of one final mean curve. 



b. Simultaneous (concurrent) analysis of 15 curves taken in 2 sets. 



c. Combination of separate analyses of the 15 curves. 



curves and the frequency periodogram combining them. In method (1) 

 there are usually scattered individual values intervening between the obvious 

 maxima, which have the effect of smoothing the periodogram to some extent. 

 We feel that any one of the three methods gives results within the required 

 order of accuracy. In the reports below, the mean curve method (3) is taken 

 as a standard; that is, it has been applied in every case, but many cases 

 have been checked by one or both of the other methods. 



The Two-Year "Scatter" Cycle — In short cycles of the order of two or three 

 years in length, we find the maximum difficulty in separating significant from 

 non-significant cycles, for in a curve of random data, cycles of two or three 

 terms (usually years in our tests) are most frequent, and the frequency dimin- 

 ishes as the length increases. Hence we present here some studies of the 

 Two- Year "Scatter" Cycle. This cycle has characters that make us feel 



