Despite the fact that the site was at the top of an actual pass, the trees stood in the middle of a wide, essentially flat area. Wood Processing.— The cores and the growth layers they contained were subjected to the following process to prepare them for cross-correlation and precipitation. The thicknesses of the growth layers were measured to hundredths of a millimeter with a measuring microscope.
Groups. – The nine ranges divided themselves geographically into three groups: east of the Pass, on the Pass and west of the Pass. The quality of the correlations between tree growth and rainfall in identical years is wasted by applying a one-year lag and by inverting data.
STUDY OF THE GROWTH LAYERS
These are ignored for now, especially because the stand history and individual tree histories are. The heights of the colored lines are inversely related to the width of the growth layers in the wood, as judged by the eye. The entries in the table where no narrow layer of growth exists in any of the specimens number 55 cases.
SMITHSONIAN MISCELLANEOUS COLLECTIONS
Comparing the average of uniform and variable sequences shows that the uniform has a significantly higher trend coefficient. Site factors at the surface appear to the eye to be nearly identical among the trees in a group, but apparently the factors change within short distances despite appearances. The correlation between different trees and between different groups, as shown not only by the trend coefficients, but also by the trend ratios, is clearly smaller for the period 1850-1897 than for the period 1898-1941.
A comparison of ponderosa pine with other ponderosa pine, of ponderosa with other species, and of other species with each other shows that species have no influence on the trend similarities. Back in 1898, the 48-year period adds only nine parallel trends to the later period's fifteen, despite the fact that the earlier period's record contains one to three fewer trees. However, greater consistency is shown by groupings: for the period, the mean of the variable series is 0.36, of the moderately variable 0.32, and of the uniform 0.28.
All trees from the wetter sites (group 11) had higher mean variations for the period 1850-1897 than they had for the period 1898-1941. Despite their influence, Group 7 shows a slight but progressive decline from the earliest to the latest period. Although the differences between group 10 and II seem striking, they are actually due to the influence of two out of eight trees.
A reexamination of the data shows that the deficiency does not apply as drastically to trees from wetter locations.
TREE GROWTH AND RAINFALL— CLOCK 19 matter of trend agreement appears to be another facet of the general
STUDY OF RAINFALL CHARACTERISTICS
Comparisons that do not include some or all of the summer rainfall are generally higher than comparisons that do. Las Vegas has the closest similarity, one station further away, 2,000 feet lower, and beyond the base of the main mountains. Santa Fe's rainfall correlates with Chacon's to a degree equal to the correlation between Black Lake and Chacon.
The best correlations are for the March-April intervals with Black Lake and Albuquerque showing relationships of opposite trends with respect to. On June 30, 1947, near Lubbock, Tex., 4 to 5 inches of rainfall fell about 2 miles wide, while none fell 2 miles to the west and 0.26 inches 8 miles to the east. Furthermore, it must be remembered in the comparison of tree growth with the rainfall of a station that, as pointed out by Landsberg,^ a rain gauge samples but does not measure rainfall and therefore "the area significance of precipitation amounts captured at a station is very limited.
34; These rainfall features must be given due weight when the growth of selected trees is compared with the record of a station several miles away. Trees can respond to the precipitation they receive themselves, but they are slightly different from what a weather station receives.
TREE GROWTH AND RAINFALL CLOCK
From a qualitative point of view, the trend ratios of Table 9 give a fairly clear indication of the degree of agreement that can be expected between tree growth and rainfall, the two being as far apart as two of the rainfall stations. Good correlations generally result under three conditions: in the absence of summer rainfall in the compared intervals, in the proportion of overlap between the two intervals, and in the length of the intervals. Characteristics of rainfall. To investigate the influence of a single factor, such as rainfall, on tree growth, it is necessary to have long data on the immediate location of the trees.
In this second place, Table 11 shows the precipitation of SantaFe and Albuquerque for the periods 1850-1897 and given for various monthly intervals. Second, there is the fact that average January-August precipitation in Santa Fe was greater in 1850-1897 than in the following 44 years. It was evident that the greater rainfall in August supported by July caused the greater average January-August rainfall in 1850-1897 in Santa Fe.
Given the changes in elevation, the average precipitation change is of the same order of magnitude. Eliminating them from the complete record leaves the homogeneous group consisting of the other two trees of group 10 and all trees of group 1 1 (HPC8 is not included because its sequence is too short). Calculating the average variation of precipitation for different months and monthly intervals shows that only April and May increased their average variations during the period 1898-1941.
It is clear that the relationship between tree growth and precipitation is very complex, not only by but also by the interaction of the whole range of growth factors, a fact which plant physiologists and ecologists emphasize.^'^ The problem requires much more work — it is far from done.
CORRELATION OF TREE GROWTH AND RAINFALL Range of tests. — Fairly extensive correlations were made between
Of these opposite trends, the year 1910 contains 84 percent of the numerical disagreement, and if 1910 is eliminated, the trend coefficient rises to 0.994. Therefore, it can be speculated that the precipitation at the tree site actually corresponded to the growth of the trees; however,. The quality of the correlation between tree growth and precipitation between March and July in the Chacon for the 33-year interval (Table 15) is all that can be ecologically expected given the distance between Holman Passand the Chacon and given the quality of the correlation between precipitation at two approximately equidistant stations like
Tree growth (individual trees) and Chacon precipitation. - Individual trees were correlated with the two rainfall intervals March-July and January-August (table 16). Further, to test the overall relationship, March-July-January-Gustrainfallof Chacon, LasVegas, TaosCanyon, and Black Lake were combined for the period 1909-1941 and tied to group 7. Although these values are high, no advantage results from the use of combined precipitation.
Tree growth and rainfall in Las Vegas. With the exception of Santa Fe, Las Vegas has the longest fall record of any station in the general area, but it is about 40 miles from Holman Pass and 3,000 feet lower. The most striking feature of the table is the clear increase in the correlation of the period 1910-1941 compared to the period 1893-1941. Surprisingly, the trees from the wetter areas, group II, compare most favorably with Santa Fe rainfall during the period 1850-1941.
In view of the quality of the correlation between Holman Passtree growth and Chacon rainfall on the one hand and between Chacon and Santa Fe rainfall on the other, the correlation between tree growth and Santa Fe rainfall has values consistent with the relative distances .
36 SMITHSONIAN MISCELLANEOUS COLLECTIONS
It should be noted that the correlation is higher for 1910-1941 than for 1898-1941; in other words, there is a general increase in the correlation against recent years.
CORRELATION
HOLMAN PASS TREE GROWTH AND RAINFALL RAINFALL INTERVAL
MAR -JUL JAN -AUG
MAY -AUG
MAY-JUN JAN -MAY
This is true for all monthly intervals except March–April, where correlations in groups 4, 7, and 9 are lower for 1898–1941. Before summarizing the contrasts between the periods 1850-1897 and 1898-1941, two points should be mentioned, one of which has to do with the incidence of opposing trends and the other with cumulative variations. First, the incidence of opposite trends was calculated for each annual interval for different groups against the different rainfall intervals.
The graph illustrates the close correspondence between SantaFerainfal variations and tree growth variations represented by group 1 1 which contains trees from humid countries. First rainfall of the period versus the period of increased rainfall; and mean variance, mean departure, and mean departure from mean change decreased. Group 11 contrasted in the same way: mean variation, mean departure, and mean departure from mean variation decreased; and the correlation with precipitation increased to an extent comparable to that with Chacon precipitation considering the much greater distance.
Ill Forgroup lo contrasted similarly: mean variation, mean departure, mean deviation from mean variation, and correlation with rainfall increased. For Group 7, the contrast was similar: mean variation and mean departure increased; mean deviation from mean variation decreased very little; correlation with rainfall increased; and internal agreement on the trends among the several trees is increased. The gist of the above summary is that the characteristics of group II consisting of trees in wet areas and group 7 (limited) agree with the precipitation, whereas most of the characteristics of group 10 and group 7 disagree, except for correlation with precipitation.
Therefore, considering a single wood study for evidence of rainfall changes, the Holman Collection suggests the use of the following methods: (i) amount of agreement in change direction, including correlation and parallelism of trends, among the trees themselves; (2) the difference in mean variation, the mean deviation, and the mean deviation from the mean difference between growth layers of trees grown under conditions, orinazone, at least compare them.
RESUME
Of course, these findings must be translated into a method where the study of wood alone can be done to detect changes in precipitation.
TREE GROWTH AND RAINFALL CLOCK 43 within the Transition Zone up into the Canadian ; the number of trees
The correlations between different trees and between different groups were clearly lower for the period 1850-1897 than for the period 1898-1941. A simultaneous comparison of the trends among all trees yielded nine complete matches among the trees over the 48 years of the 1850–1897 period and 15 over the 44 years of the period. 1898–1941 For group 10 (dry sites), mean interannual variation, mean departure, and mean deviation from mean variation increased, while for group 11 (wet sites) they decreased for the period unlike the period 1850-1897.
However, the average departure of two of the trees from group 10 actually corresponded to group II. The average deviation from the average variation of group 7 (limited) and group 7 itself decreased for the period 1898-1941 in contrast to the period 1850-1897. If the trees responded directly to the rain falling in their immediate location, one might expect them to correlate with the rainfall at Chacon to a degree equal to or slightly greater (due to the distance involved) of the mean of the correlations between Chacon and the other rain stations.
Correlation between the eight different monthly intervals in Chacon ranged from -0.33 to 0.99 for the trend coefficients and from 0.66 to 0.06 for the ratio of opposite trends. For the contrasted periods 1850-1897 and the characteristics of the dry trees contrasted with the rainfall, whereas the trees in the wet areas were parallel. Correlations between tree growth and precipitation in Las Vegas were higher for the period 1910–1941 than for 1893–1941.
The most important information that emerged from the correlation between tree growth and rainfall in Santa Fe from March to July for the periods 1910-1941 was this: The quality of the correlations was lowest for the former period and highest for the latter .
