216 
PACIFIC SCIENCE, Vol. 1, October, 1947 
In an effort to solve the remaining 12 %, the 
writer became entangled in an investigation 
which led far afield.” 
Moreover, any other systematic relation¬ 
ship found in random values for a given 
quantity will, if applied, result in more 
accurate estimates than the use of simple 
averages without such additional terms. For 
example, various stations considered by 
Nakamura in his study of rainfall variation 
on Oahu have a probable departure of the 
rainfall of any given year from the mean of 
about 30 per cent of that mean (Nakamura, 
1933). This figure is true of stations having 
moderate rainfall, such as that of Kualapuu 
or U. S. Weather Bureau at Honolulu. This 
means that, by naming the mean annual 
rainfall, anyone can predict, on the average, 
within a range of 30 per cent. One means 
of improving estimate is offered through a 
recent study of geographic variation (Went¬ 
worth, 1946). In this paper it is shown 
that if the rainfall of near-by, similar sta¬ 
tions for the same year is known, the prob¬ 
able error of estimate is cut down, on Oahu, 
to about 14 per cent. This relationship is of 
no use for future prediction, but is offered 
simply as an example of refinement of esti¬ 
mate due to added knowledge. 
Parallel to this, Johnson says his errors 
were reduced to 12 per cent by use of his 
data on periods. This seems reasonable, 
and it is shown below that harmonic analysis 
of the Honolulu Rainfall Index can furnish 
a basis for somewhat similar improvement of 
average estimate for a near-by, future year, 
over that represented by the simple annual 
mean or normal. 
So far, the studies made by Johnson ap¬ 
pear to be significant. Beyond this point, in 
the search for the remaining 12 per cent, it 
is less clear either that the continued har¬ 
monic analysis of residuals is worth the 
greatly protracted labor involved, or that 
similarities of period found between rainfall 
and other physical phenomena are truly the 
result of cause and effect. Certainly, if 
causation is at work and if the records of 
such correlated sunspots, magnetic values, 
or planetary positions are of long duration 
or can be reliably projected into the future, 
the correlation with rainfall would be of 
much value. 
The writer is not sufficiently acquainted 
with these other phenomena to judge inde¬ 
pendently whether the similarities in period 
do or do not represent a causal relationship. 
The latter half of Johnson’s book constitutes 
an interesting and exhaustive search for re¬ 
lationships between rainfall and other phe¬ 
nomena and there is abundant evidence of 
cyclical variations which can be to some 
extent defined. However, candor compels us 
to conclude that true causal relationship has 
not been proved, and it does not appear that 
use of data from planetary or other changes 
will yield predictions superior to those based 
on one or two empirical rainfall periods 
alone. 
It is noteworthy that in a lecture given in 
Honolulu in 1946, Dr. C. G. Rossby ex¬ 
pressed the view that useful predictions for 
more than 1 month in advance are not now 
practicable. This eminent authority had just 
completed a tour of conference and inspec¬ 
tion in Hawaii, under the sponsorship of the 
Hawaiian Sugar Planters’ Association and 
the Pineapple Research Institute, and al¬ 
though he recommended renewed and more 
extensive study of the problem, he was not 
optimistic of success by available methods in 
the foreseeable future. 2 In line with this 
statement, the paper by Johnson, while it 
probes interesting possibilities, did not, in 
the estimate of the present writer, contain 
anything to show that prediction of the rain¬ 
fall of a single future year could be made 
2 After this paper was completed, the writer 
had opportunity to see a manuscript report by 
H. Landsberg, prepared in collaboration between 
the U. S. Weather Bureau and the University of 
Chicago, in which a similar view was expressed. 
