NUMBER OF TREES NECESSARY FOR RELIABLE RESULTS.
Fig. 2.—Monthly and yearly precipitation at Prescott and size and character of rings.
In seeking the best curve of tree-growth which a given locality can supply, it might be thought at first that a very large number of trees must be obtained in order to get an average, but experience has shown that the number may be very small. In order to test the accuracy obtained from a small number of trees, a comparison was made between large groups and small. Of the original 25 trees in the first Flagstaff group, 19 were subjected to very careful cross-identification. Averages were then obtained of the oldest 5, going back about 400 years, the oldest 10 (350 years), the oldest 15 (300 years), and the entire 19 reaching back only 200 years. Finally, the record of the oldest 2 was carried back fully 500 years. On plotting the groups of 15, 10, and 5 with its extension of 2, it became immediately evident that 5 trees gave almost the same growth as 15, even to small details. Between these 5 and the oldest 2 taken by themselves the agreement was not quite so perfect, yet was so close that errors thus introduced would not affect the curves. It must not be taken for granted without test that this remarkable agreement between very small groups of trees is true necessarily for other trees or even for
