NO. l8 TREE GROWTH AND RAINFALL CLOCK 9 



the basic data constitute a continuous time series, the trend method ^ 

 of correlation has been used. This method, in the coefficient t, gives 

 a measure of parallel variation combined with amount of that varia- 

 tion. If t equals i.oo, the trends in both sets of data, tree growth 

 and rainfall, are wholly parallel or in the same direction ; if t equals 

 -1.00, the trends are wholly opposite. Tests were constantly made 

 to detect the undue influence of one or two entries should such be 

 present. The ratios of opposite to parallel trends are included in 

 the tables with the trend coefficients. With the trend method of cor- 

 relation, secular trends or long-period fluctuations did not have to be 

 eliminated. These are ignored for the present especially because the 

 history of the stand and the histories of the individual trees are 

 unknown except as revealed on the wood itself. 



The quality of the correlations between tree growth and rainfall 

 on identical years was tested by application of a one-year lag and by 

 reversal of data. In all cases, the correlations dropped to a value of 

 no significance. 



Although the purpose of the work was the comparison of year-to- 

 year variations, correlations involving smoothed data (second inter- 

 mediate) were nevertheless carried through the main part of the 

 calculations. The majority of the coefficients did not increase signifi- 

 cantly over those using raw data ; in fact, many decreased. 



The initial questions, then, to be answered by use of the trend 

 method, were: If rainfall increases or decreases, does tree growth, 

 as shown by such simply obtained samples as increment cores, increase 

 or decrease in like direction? To which rainfall interval does the 

 tree growth correspond ? As the work progressed new problems came 

 to light and soon carried the study far beyond the original objectives. 



STUDY OF THE GROWTH LAYERS 



Cross-dating. — This process consists of establishing the identity in 

 time of growth layers on different sequences by matching narrow 

 growth layers, in particular, from one specimen to another. Judg- 

 ment as to narrowness depends upon visual comparison with im- 

 mediately adjacent growth layers and should not be confused with 

 or substituted for mathematical expressions. Obviously, cross-dating 

 can possess various degrees of excellence. What constitutes reliable 

 cross-dating is a moot point and may, perhaps, be largely dependent 



'■ A rapid method of correlation for continuous time series. Amer. Journ. Sci., 

 vol. 240, pp. 437-442, 1942. 



