266 
PACIFIC SCIENCE, Vol. VIII, July, 1954 
tion. Thus, it may be said that, disregarding 
the spatial variations in standard deviation of 
pressure (or, in this case, 700-mb. height), 
the correlation field is strictly analogous to 
an anomalous flow pattern. A tight gradient 
of the isopleths in the correlation field in- 
dicates a strong correlation of the wind tan- 
gent to those isopleths. This view of the 
correlation field is the basis for the discussions 
of the patterns. 
RAINFALL PARAMETERS 
Because the rainfall climate varies enor- 
mously from point to point within the Ha- 
waiian Islands, no single rainfall parameter 
will be representative of conditions in gen- 
eral. Stidd and Leopold (1951) showed that 
two parameters might suffice to describe the 
rainfall patterns over fairly large areas. 
In this study, a group of five stations from 
the wetter areas of Oahu have been combined 
to represent one of these parameters, and four 
dry-area stations represent the other. The 
combination of the two parameter estimates 
into an estimate for some other station is 
explained in the appendix. 
Certain areas under the direct influence of 
mountains higher than 5,000 feet fail to fit 
into this scheme since their rainfall patterns 
change systematically with changes in wind 
direction. In the present study, several such 
places — Hilo, the Hamakua Coast, and Na- 
alehu— have been tested on an individual 
basis. 
To eliminate some of the randomness found 
in single-station rainfall data, each rainfall 
index is composed of the mean of several 
fairly homogeneous station values. These are 
listed in Table 1 and shown in Eigure 2. 
In order to make the rainfall indices have a 
more nearly normal distribution, the cube 
roots were taken of all the group rainfall 
amounts (Stidd, 1953). The assumption that 
the resulting sets of values are normally dis- 
tributed is satisfactory in all these cases, as 
the stations in each group are quite homo- 
geneous and no individual stations reported 
zero rainfall for any of these months. In some 
of the 5-day and 1-day data, which are dis- 
cussed briefly in this report, zeros did occur, 
but not so frequently as to lead to any serious 
departures from normal distributions. Seasonal 
trends were removed by expressing the index 
value of a given month as a difference from 
the mean of all the values for that month. 
The distribution of 700-millibar heights 
was assumed to be so nearly normal that no 
processing was required except for removal of 
the seasonal trend. This was accomplished 
in the same manner as for rainfall. 
Five-Day Means: These were computed only 
for the period from November, 1946, through 
April, 1951. Approximately 255 pairs of values 
(two overlapping 5-day means per week) were 
available for each grid point, although, due 
to the fact of the overlap in 5-day periods 
and to the fact that serial correlations are 
higher in 5 -day data than in monthly data, 
TABLE 1 
Stations Used in Computing Rainfall Indices 
OAHU WET 
OAHU DRY 
HILO 
HAMAKUA 
NAALEHU 
Station 
Mean 
annual 
Station 
Mean 
annual 
Station 
Mean 
annual 
Station 
Mean 
annual 
Station 
Mean 
annual 
Kawai-iki 
Lower Luakaha 
Tantalus 
Nuuanu 4 
Nuuanu 5 
102.19 
137.57 
100.93 
141.77 
87.59 
Ewa 
Wailupe 
Waipahu 
U.S. Mag. Ob. 
20.62 
26.11 
25.07 
19.70 
Hilo 
Papaikou 
Pepeekeo 
Waiakea M. 
138.71 
132.97 
128.57 
200.32 
Kukaiau 
Pauuilo 
Paauhau 
Honokaa 
105.56 
100.81 
67.91 
93.44 
Naalehu 
Kiolakaa 
Moaula Sta. 
Pahala 
45.24 
64.46 
47.60 
43.37 
Group Mean 
112.01 
22.90 
150.14 
92.18 
50.17 
