was made. In this the lower curve represents the average annual growth of 25 trees and the upper curve is the precipitation 12 miles distant. The latter is taken from November 1 to November 1 in order to carry the snowfall into the following season of growth. This study suggested the investigation of the time of year to begin annual means of rainfall, which has already been presented in Chapter II. Figure 4 gives a comparison between Flagstaff rain and the two Flagstaff groups, and also shows how the best time of beginning the year was determined. It proved to be November 1 at Flagstaff and September 1 at Prescott, where the nature of the ground gives more chance of conserving moisture. The great difference between individual trees in response to rain is also shown in figure 5. It is evident that quick-growing trees serve as better indicators.
THE PRESCOTT CORRELATION.
Five subgroups, numbering in all 67 trees, were obtained from different points in the vicinity of Prescott. These all cross-identified among themselves with entire success, both as individuals and as groups. The group curves are shown in figures 6 and 7, but in comparison with the Prescott rainfall they differed greatly, the group nearest the city showing much the best accordance. Accordingly this group is plotted by itself in figures 7 and 15 with the rainfall curve. On the whole there is much agreement, as may be seen by comparing the crests and troughs of one with those of the other. The most conspicuous discrepancy is in 1886, where the rainfall decreases and the growth of the trees increases. In 1873 the growth seems to have responded to the decrease in rainfall, but to a greatly diminished degree. The tree maximum of 1875, one year behind the extreme maximum of 1874 in the rainfall, is entirely reasonable, since the ground may become so saturated that the effects last until the following year. On the whole, the curves shown in figure 7 support the idea of a proportional relation between annual rainfall and annual growth.
Accuracy.—The accuracy with which the pine trees near Prescott represent the rainfall recorded in that city for 43 years is, without correction, about 70 per cent. By a provisional correction for conservation of moisture by the soil this accuracy rises to about 82 per cent. The nature of this conservation correction is very simple; it makes use of the "accumulated moisture" of the meteorologist. It signifies that the rings in these dry-climate trees vary not merely in proportion to the rainfall of the year, but also in proportion to the sum