84 



CLIMATIC CYCLES AND TREE GROWTH 



The periodogram test has been tried out more thoroughly than the other. 

 The three sequences were made into cycleplots and at four different cycle 

 lengths settings were made on each and cyclograms secured and printed on 

 plates, three at a time in an order unknown to me. That made 80 sets of 

 3 patterns each. The correct choice was made of the genuine in every case 

 simply on the basis of relative freedom from short cycles. Three sets (15 

 per cent) were difficult, and the choice held somewhat uncertain. The tests 

 also show the domination of longer cycle lengths in natural sequences. This 

 charaoter may be stated in periodogram terms by saying that the frequency 

 periodograms of natural sequence have a level or upward trend, while those 



GC 



FV 

 SH 



NE 



FLU 



RL 



Mean 



1800 



Flagstaff trees 



1900 



1800 1900 



Lot drawings of same 



Fig. 34 — Effect of smoothing on natural sequences (left) and random 

 sequences (right). Flagstaff area — 20 year running means 



of random sequences have a downward trend. These conditions make it 

 rather evident that short random sequences of 50 or 75 terms become very 

 difficult to distinguish from real ones; and that was found to be the case. 



These two types of triangle tests are illustrated in figure 34, giving an 

 example of smoothing; and Plate 20, showing the cyclogram test. In the 

 latter the right column is genuine. 



Clough has referred to a similar character in natural sequences by observ- 

 ing that in random sequences the curve crosses the mean line more often 

 than in natural sequences. 



