METHOD OF ESTIMATING RAINFALL BY GROWTH OF TREES. 



105 



separated from that of the next. Nevertheless, the rings may be so thin that they can 

 not easily be distinguished, and seem to merge into one, but a microscopic examination 

 usually shows indications of a soft, white ring as well as of a hard, red ring for each year. 

 As a rule, therefore, each annual ring is extremely well marked, and there is no doubt 

 as to its purely annual or seasonal character. In some few cases, however, rings die out 

 completely, while in others they are double. In the first investigation of the trees at Flag- 

 staff, it was estimated that the results were subject to an error of 2 per cent, most of 

 which occurred near the center of the tree. The more rigorous methods subsequently 

 employed, however, proved that the error of unchecked counting in these trees was 4 per 

 cent and lay ahnost entirely in the recent years. It was due chiefly to the omission of 

 rings or the merging of several together, apparently from lack of nutrition. The number 

 of trees in which serious errors are found is not sufficient to prevent the curves of growth 

 and of rainfall from showing close agreement. Mistakes can be guarded against only by 

 a process of cross-identification which will be described shortly. The effect of the unde- 

 tected omission or the doubling of rings in individual trees is to lessen the intensity of the 

 variations in the curve of growth obtained by the averaging of many trees. The errors 

 may be divided into two classes: first, local errors of identity in small groups of rings in 

 a few individual trees, which simply flatten the curve without affecting the final count; 

 second, cases in which a given ring, in spite of attempts at cross-identification, is still in 

 doubt, showing perhaps in half of the trees, and not in the other half. Such cases affect 

 the final count but do not flatten the curve. One case of this sort will be noted below. 

 It leaves a question of one year in the dating of all the earUer portions of the curve. 



THE TREES OF PRESCOTT. 



The problem of cross-identification is well illustrated in the trees of Prescott. These 

 were measured in 1911 for the purpose of testing the conclusions derived from the Flagstaff 

 trees some years earlier. Prescott was chosen because, as has already been said, the 

 weather records there go back to 1867 with only slight breaks. From that date until 1898 

 the observations were made at Fort Whipple, about a mile northeast of the town, and from 

 1898 to the present time they have been taken on the southwest edge of town. The small 

 breaks referred to were chiefly in the summer of 1869. These have been supplied approxi- 

 mately by comparison with the records in other parts of Arizona during the years 1866 to 

 1870, but there is still a question of several inches for the total July and August rains for 

 1869. The cuttings from tree stumps in the Prescott region were procured through the 

 assistance of Mr. C. H. Hinderer, supervisor of the Prescott National Forest. The region 

 about Prescott has been in the Forest Reserve since 1898, and no cutting has been allowed 

 except by special permit, but by the records he was able to tell just when the trees had 

 been cut. The trees used were all of average size, being several hundred years of age; 

 the cuttings were made from the edges of the stumps and were intended to include the 

 last fifty years or so. Sixty-four were measured and the data in regard to them are 

 shown in table 1. 



Besides the cuttings shown in table 1, three others were measured, two in the first group 

 and one in the thud, but were finally omitted because theii- oldest rings did not date back 



