Hawaiian Climatology — LEOPOLD and STIDD 
The best correlated pairs of stations had a 
probable deviation from their mean ratio of 
less than 10 per cent. For those pairs having 
poorest correlation, the most probable deviation 
in a single year was 45 per cent. The best corre- 
lated pairs were found among the high rainfall 
stations, while the low rainfall stations showed 
the greatest deviations. 
Wentworth’s study is important also to stu- 
dents of hydrology. He showed how the inter- 
station correlations could be used to synthesize 
missing annual rainfall amounts at stations 
whose records are sufficiently long to establish 
the initial correlation. The method provides 
indices of reliability for such interpolated rec- 
ords and could be used on monthly as well as 
annual values. 
Riehl (MS.) found a double maximum of 
effective precipitation in both summer and 
winter, as had Landsberg (MS.) and Tiillman 
(1936). Riehl attributes the minima in June 
and September to the seasonal shift of the sub- 
tropical ridgeline in the high troposphere. This 
ridge is centered over the latitude of Hawaii 
in June and September, having moved north 
in midsummer and southward again in the fall. 
Secondary maxima of rainfall in winter are 
not due to a change in the number of storms, 
of which there is a single maximum in winter. 
Riehl concludes that at the beginning and end 
of the winter season there is the greatest possi- 
bility of interaction of summer-type tropical 
disturbances with the extra-tropical winter 
storms. This leads to the double maximum of 
rainfall in winter. 
Riehl finds that 80 to 90 per cent of the total 
rain occurs in the general storms or "rainstorms.” 
From this he concludes that present forecasting 
techniques which should allow the forecasting 
of these storms are sufficient to solve the main 
problems of agricultural forecasting for Hawaii. 
The cyclic behavior of rainfall has been 
studied by a number of workers. Cox (1924) 
constructed an index of monthly rainfall values 
based on 10 stations and extending over 44 
years. He concludes that a cycle of 3.7 years 
221 
appears to exist and is even stronger than the 
annual cycle. 
Johnson (1946) analyzed yearly rainfall 
amounts at one station, Kualapuu, Molokai, by 
the method proposed by Alter (1937). Went- 
worth (1947) in reviewing this work concisely 
describes the method as "a process of finding 
that constant interval between pairs of years 
such that the average difference in rainfall be- 
tween the two members of a pair is a minimum.” 
At Kualapuu, the 20-year periodicity was the 
strongest, and the 12 -year cycle was second 
strongest. For the Honolulu record, a 1 4-year 
cycle appeared most probable with the exception 
of cycles of 36 years and certain others too close 
to the length of the 66-year record to be very 
reliable. Using the "Honolulu Rainfall Index” 
(the average of the percentages of the annual 
mean for 10 stations in the Honolulu area), 
Wentworth (1947) applied Alter’s method to 
find the most probable cycles in the annual 
values of this series. He showed that the 16- 
and 20-year cycles were the most promising. 
Working with the monthly values of rainfall 
at Waimanalo, Landsberg (MS.) showed by 
harmonic analysis that the annual variation is a 
wave which is not accidental. He decided that 
the existence of other periodicities in the data 
for that station was not probable. 
The fact that the lengths of most probable 
cycles differ among stations or groups of sta- 
tions in the islands markedly reduces the use- 
fulness of, if not the confidence in, such cycles. 
Wentworth’s analysis was the most practical. 
He showed that the average deviation of the 
"predicted” rainfall index from the actual can 
be reduced by using the 20- and the 12 -year 
cycles. In individual years, however, the differ- 
ence between actual value and the mean value 
was less than that between the actual and the 
value predicted by use of the cycles. Wentworth 
concludes, therefore, that "for practical purposes, 
any statistical, long-run gain is canceled by the 
evident risk of an aberrant estimate for a given 
year.” 
