V. CORRELATION WITH RAINFALL.
Result of study of curves. — On completing numerous curves of tree-growth in the manner already described, three characteristics were observed: (1) in arid-climate groups the annual rings are approximately proportional to rainfall; (2) in moist-climate groups they vary with the changes of solar activity; (3) in each they are subject to certain cycles or periodic variation. The first of these is the subject of the present chapter.
Early tests of rainfall correlation. — The earliest comparison with rainfall in this investigation was made between the first Flagstaff subgroup of 6 trees and 43 years of precipitation records at Prescott, 67 miles distant. It was not expected that agreement in individual years would be found; accordingly smoothed curves were used, consisting of overlapping means of 9-year groups. This produced curves of gentle variation, but similarity in the curves was evident. These early curves are presented in figure 13. The best agreement was found
Fig. 13.—Correlation between tree-growth and rainfall in smoothed curves; Flagstaff.
The comparison in figure 13 was made with Prescott records because there were not at that time enough Flagstaff records to be of service. But later, when a Weather Bureau station had been established in Flagstaff for several years, the striking comparison shown in figure 14
by placing each mean of 9 years of rainfall at the end of the 9 years as in this figure instead of in its center. This lag of four years seemed inconsistent with the later results of yearly agreement without lag, and in fact for years it has been accepted with some hesitation by the writer. Yet in the present consideration of the subject it appears to have a special significance. This existence of the lag in long periods agrees in principle with the "accumulated moisture" effects observed in the Prescott trees and with the idea of a tree exhibiting a reserve power or vitality which may run low or be built up by varying environment. The principle will be referred to again below; it is sufficient now to state that it seems quite reasonable to find no lag in yearly correlation with rainfall and at the same time a very considerable lag in the slower variations.