1028 
the Southwest, are almost nonexistent for the simple 
though narrow-ringed arctic series discussed by Hrl- 
andsson [12], Ording [23], and others. On the other 
hand, the fluctuations in ring-widths im arctic series 
tend to be relatively small, and the chronologies vary 
from site to site more than those in the Southwest. 
Almost all arctic studies have noted the relation of 
ring-width fluctuations to the fluctuations in mean or 
average maximum summer temperature, particularly 
that of the June-July interval, the highest correlation 
coefficients being of the order of +0.7. As one would 
expect, precipitation is in general not an important 
limiting factor m growth. The longer tree indices of 
arctic temperatures are noted in Table I. 
Good series for numerous areas in northern Alaska 
have been developed by Giddings [13, 14], who finds 
that the more sensitive tree records represent June— 
July mean temperatures with some fidelity. Extending 
living-tree records by the use of archaeological beams, 
he has developed one series almost a thousand years in 
length. 
Indices of Physiological Drought in the Eastern 
United States. In the absence of a dominant climatic 
variable, it does not appear possible to derive a simple 
meteorological index from the rmg-growth in the non- 
marginal areas, although some correlations of larger 
growth departures with rainfall or temperature of vari- 
ous spring and summer months are reported by Lyon 
[21, 22] and others. An example of a fair degree of cross- 
dating over a considerable area is furnished by the 
analysis of hemlock growth in New England—an in- 
dication that trees in relatively moist areas can provide 
a significant history representing the resultant effect 
on plant growth of a complex of climatic conditions. 
LONG-TERM CLIMATIC FLUCTUATIONS 
Changing Areas of Drought. A sufficient number of 
ring chronologies of rainfall in the western United 
States have been developed to permit a preliminary 
view of this phenomenon over the centuries. The tend- 
ency towards a variability in the area affected by 
drought in different years, a striking characteristic of 
gage records of rainfall and runoff, applies also to 
longer drought intervals and is evident over the entire 
range of the data. Centuries-long chronology transects 
for 1500 miles or more of the Pacific Slope and the 
front ranges of the Rocky Mountains reveal no syste- 
matic time displacements in maxima or minima from 
one area to another of sufficient significance to permit 
their use in long-term weather forecasting. Perhaps the 
most important element emphasized by this analysis is 
the need for caution in generalizing, either in time or 
areal extent, any local indications of cyclic recurrence, 
drought frequency, trends, and similar phenomena. 
Dry and Wet Years and Intervals. The three regional 
chronologies noted above, for the Colorado River Basin, 
southern California, and eastern Oregon, each on a 
broad statistical base of sensitive rmg records, may be 
used to derive long-term statistical parameters of in- 
terest to the climatologist. It appears that specially 
dry winters and low runoff in the Colorado River Basin, 
of the severity of the “dust-bowl year” 1934, occurred 
CLIMATOLOGY 
about once in twenty years during the 658 years of ring 
record (not every twenty years!); instances of two suc- 
cessive years of such critically small runoff, of special 
importance in the management of the reservoir level at 
Hoover Dam, seem to have occurred only three times 
during that record. In the Colorado River index, after 
simple three-term smoothing of the type noted in Table 
II, the median length of an interval of general excess 
or deficit is about six years, intervals of deficiency or 
excess more than three times this length occurring 
once in about 200 years. In southern California the 
median length is about 8 years in the 560 years of 
similarly smoothed record. Intervals of excess and deficit 
are in general somewhat longer in Keen’s Oregon chro- 
nology, an effect in part, at least, of the heavier smooth- 
ing. 
The Colorado Basin index indicates that the interval 
of greatest deficiency in rainfall and runoff in this 
region since A.D. 1300 occurred in A.D. 1573-1593, 
when the average deficiency was well below that in the 
most severe intervals recorded by modern gages. It is 
significant that the greatest minimum in the southern 
California index occurred in 1571-1597, when the ac- 
cumulated deficit approached twice that of any mterval 
in the century or so of rain-gage records. The eastern 
Oregon index shows only a weak general deficiency 
from 1565 to 1599, broken by a short interval of excess 
near 1587; the most pronounced minimum in that index 
began in 1917 and was followed by ‘gradual en- 
croachment of desert conditions into what were once 
thriving pine stands” [20] during the next two decades. 
Relevant to this is the interregional correlation coeffi- 
cient, computed for the interval 1650-1935, of +-0.46 
for the Colorado Basin versus southern California, 
which compares with a coefficient of +-0.21 for the same 
interval between the Colorado Basin and eastern Ore- 
gon; very similar results are obtained when winter 
rainfall data in the three regions are correlated. No 
significant correlation between British Columbia growth 
and that in the Colorado River Basin has been noted 
for the last three centuries or so of data. 
No long-term trend has yet been found in any of the 
major ring indices which could be reliably interpreted 
as a secular change in climate, though temporary fluc- 
tuations of considerable magnitude and duration are 
evident throughout the records. Beams in South- 
western ruins, which were cut as much as 1700 years 
ago, show about the same ring sensitivity as living 
trees on the same sites, a strong hint that no decidedly 
more moist or dry conditions prevailed at those times. 
Cycle Analysis. Douglass has poimted out the per- 
sistent tendency of many tree-ring series to show fluc- 
tuations of 22-23 years and fractions and multiples of 
this cycle, which seem related to sunspot or possibly 
other extraterrestrial phenomena. Cycle lengths of this 
order are to be found in the ring series in fossil woods 
and have been reported in the Scandinavian ring chro- 
nologies of temperature. The cycles found in tree growth 
seem, however, to be characterized by variability in 
length, amplitude, and form, to tend to appear and 
disappear according to no discernible law, and to occur 
in almost any combination. A satisfactory physical 
Oi eee 
