84 
PACIFIC SCIENCE, Yol. II, April, 1948 
Preparation of Diurnal Rainfall Curves 
Because different synoptic situations are re¬ 
lated to these rainfall types, separation of geo¬ 
graphical areas over which these types occur is 
of some interest. In order to make such separa¬ 
tion, diurnal rainfall curves were prepared from 
recording rain-gage records for six of the sta¬ 
tions on Oahu and the two stations on Lanai. 
These recording gages, with the exception of 
that at Honolulu, were installed by the U. S. 
Soil Conservation Service about 1941, and are 
maintained by the sugar and pineapple planta¬ 
tions. 
Certain lapses in record made it impossible 
to analyze exactly the same record period at all 
gages, but in general, the period January 1, 
1945, to December 31, 1946, was used in all 
cases except Honolulu. The Honolulu quanti¬ 
ties had been counted and tabulated for the 
period 1923 to 1941 by H. P. Parker of the 
U. S. Weather Bureau, who kindly allowed the 
writer to use some of his tabulations for this 
analysis. 
From the original recorder charts, the rain¬ 
fall amounts for stations other than Honolulu 
were tabulated hour by hour for the period of 
record. No attempt was made to adjust the 
chart total to the total recorded in the stand¬ 
ard rain gage which, in most installations, is set 
up adjacent to the recorder. It became apparent 
at an early stage that a short period of record at 
gages in dry localities would not yield statis¬ 
tically significant results comparing number of 
occurrences of various amounts of rain in indi¬ 
vidual hours. Therefore, all individual occur¬ 
rences in a given hour were lumped, regardless 
of amount of rain which fell during the hour. 
"Traces” of rain do not show up on a Ferguson 
reconnaissance type gage, so the minimum 
amount which was counted as an occurrence 
was 0.01 inch. A seasonal breakdown was 
chosen of four periods of 3 months each, start¬ 
ing with December, January, and February as 
the winter period. 
It was desired to determine whether there 
was a significant difference in rain occurrence 
at various parts of the day. In Figure 3A is the 
histogram of rainfall occurrence for individual 
hours at Kawaihapai. This station has a mean 
annual rainfall of less than 30 inches. The total 
number of rainfall occurrences in the 6 winter 
months of the analyzed record (December, Jan¬ 
uary, and February for 2 years) was 110. It can 
be seen that the small number of occurrences 
necessitated grouping to bring out significant 
differences between various times of day. 
For each seasonal group the number of occur¬ 
rences was computed for all combinations of 8 
consecutive hours, i.e., 0000 2 to 0800, 0100 to 
0900, etc. For purposes of statistical analysis, 
the day was broken into three 8-hour periods, 
one of which provided the maximum number 
of occurrences in any 8-hour combination. The 
graphs in Figure 3B show the resulting break¬ 
down for Kawaihapai. The data in Figure 3B 
are the same as in Figure 3A, merely grouped 
into 8-hour totals. 
Statistical Test for Significance 
To determine whether the maximum number 
of occurrences in an 8-hour period was signifi¬ 
cantly different from the number in the other 
periods, a chi-square test was used. Assuming 
that the three 8-hour periods had equal chance 
of rain occurrence, the chi-square computation 
determined the number of occurrences which 
might occur in 8 consecutive hours once in 100 
times as a result of pure chance, and again once 
in 20 times. If a significantly greater number 
of rain occurrences was noted in a given 8-hour 
period than might be expected in random trials, 
then the original hypothesis of equal chance for 
all hours appears untenable. In such cases we 
can reasonably assume that a causal factor 
operates to provide the observed distribution. 
Obviously the statistical procedure does not tell 
us whether the period which the data represent 
was a good sample of a much longer period. 
2 Hour of day in this report is in local standard time 
(LST). Times are shown on basis of a 24-hour clock, 
thus 1300 is 1:00 P.M. 
