TREE-RING INDICES OF RAINFALL, TEMPERATURE, AND RIVER FLOW 
By EDMUND SCHULMAN 
University of Arizona 
INTRODUCTION 
History. Dendrochronology—the construction, based 
on tree-ring widths, of significant indices showing 
changes in climate from year to year and the use of 
these indices in dating and forecasting—had early be- 
ginnings. That the succession of wide and narrow growth 
rings of trees might provide a chronology of the wet and 
dry seasons of the past was noted at least as early as the 
fifteenth century by Leonardo da Vinci. As in other 
fields, research expanded tremendously during the 
1800’s, when botanists and foresters, particularly in 
central Europe and in Scandinavia, examined numer- 
ous details of growth-climate relationships [2]. About 
1880, the Dutch astronomer, Kapteyn [19], succeeded 
in deriving a ring chronology several centuries long, 
which gave a first approximation to the annual rainfall 
of west Germany, but no clear index, which he sought, of 
solar activity. His concepts of tree selection and an- 
alysis, foreshadowing much later work, remained gen- 
erally unknown, however, for some thirty-five years. 
Dendrochronologic work was begun in the southwest- 
ern United States in 1904 by Douglass [5], who for- 
mulated the basic principles and who established the 
superiority of drought areas and coniferous species as 
sources of ring chronologies of rainfall. Since that time 
this research has been continuously developed by him 
and his co-workers at the University of Arizona. Al- 
though the main objectives of this research were cli- 
matic, there have been some fruitful offshoots, such as 
the setting up of a precise archaeological time-scale 
to A.D. 11 for the central part of the Pueblo area, which 
has been accomplished by the backward extension of 
the ring record in living trees, based on the ring dating 
of ancient timbers [6]. 
Dendrochronologic series have thus far been exten- 
sively developed in only three regions. The results of 
work in the semiarid western United States, as they 
bear on the climatology and hydrology of the dry lands, 
will form the bulk of the following survey. In arctic 
areas, Scandinavia has produced in recent years a 
number of long and significant chronologies of temper- 
ature changes, primarily for the June—July interval; 
specially long arctic chronologies, composites of records 
in living trees and ancient timbers after the fashion of 
those in the Southwest, have been derived for northern 
Alaska. Scant justice can be done here to a third region 
of active chronology work, the noncritical areas of the 
eastern and central United States; several thorough 
investigations have shown that chronologies of limited 
climatological significance can be obtained on some 
sites in this region, though many localities apparently 
cannot provide dendroclimatic indices. 
Principles. It is now known that a great range of 
variability exists mn all aspects of ring growth. In 
order to obtain climatic indices of maximum simplicity, 
fidelity, and length—the specific objective of dendro- 
chronologic work in the Southwest—three basic prin- 
ciples have been developed: selection, crossdating, and 
sensitivity. 
Proper field selection [25] of trees for sampling is of 
first importance, for many species, sites, and areas are 
found to yield ring chronologies of no apparent sig- 
nificance as climatic indices. By successive refinement 
of selection criteria, trees may be found in which the 
influence of one dominant factor, such as rainfall in 
the dry lands, is increasingly represented in the ring- 
width fluctuations. 
The principle of crossdating [9] states that the ap- 
proach to parallelism in ring chronology among the 
individual trees of a homogeneous set is a direct measure 
of the influence of some general factor or set of factors 
on the variations in ring-width from year to year. 
Such parallelism provides the scientific control for the 
discovery and evaluation of hidden irregularities in any 
ring series. This rather simple concept has proved very 
far-reaching in its thorough application to drought 
chronologies. 
The approach to parallelism between highly sensitive 
ring series in a small area of the semiarid Southwest, 
where this phenomenon has been found to be near a 
maximum, is illustrated in Fig. 1. Although only a half- 
dozen or so specially long-lived Douglas firs have formed 
the basis of the record in the outer five centuries of the 
Mesa Verde and Tsegi indices, the similarity of these 
two series in light of their 115-mile separation indicates 
that the random term in individual tree growth has 
been largely cancelled in this interval; some earlier 
portions of both series are less well based. The Flag- 
staff index, more distant from Mesa Verde and based 
mainly on a different species (ponderosa pine), shows 
occasional substantial differences from the preceding 
two chronologies, yet is on the whole in good agreement. 
Fifty-year correlation coefficients for the twenty in- 
tervals from A.D. 900 to a.p. 1899 average -+-0.71 for 
Mesa Verde versus Tsegi (115 mi W by 8) and +0.46 
for Mesa Verde versus Flagstaff (220 mi SW). The 
high significance of the Mesa Verde index as a rainfall 
history is illustrated in a later section. 
In ring series which crossdate with one another the 
average proportionate change in width from year to 
year is a quantitative index of sensztivity. Highly sensi- 
tive series have been found to show a close relation to 
rainfall in the dry lands of the Southwest; at the same 
time, even in this region, trees on sites which permit 
significant water carry-over from year to year show low 
sensitivity 
1024 
