Til 
a group of five stations from the wet areas 
of Oahu. The group mean annual rainfall of 
112 inches indicates that these stations derive 
the bulk of their rainfall from orographic 
effects, and, as might be expected, the pattern 
of Figure Ga is somewhat similar to that for 
the Hilo area (Fig. 1). There are three impor- 
tant differences, however, as follows: 
1. The positive center is shifted 5 degrees 
northward. 
2. The correlation is negative in the island 
area. 
3. The over-all strength of the relationship 
is not as great as in the case for Hilo. 
The northward shift and negative correlation 
may indicate a somewhat greater importance 
of cyclonic activity to the monthly rainfall 
totals of these wet Oahu stations. The fact 
that the over-all relationship seems weaker 
would indicate, as in the case of the Hamakua 
stations, that rainfall of the wet Oahu stations 
is not as greatly dependent on the strength 
of the trade winds as is that of Hilo. 
Five-day rainfall amounts for these wet 
Oahu stations are shown by Figure Gb to be 
more dependent on cyclonic activity and less 
dependent on the strength of the trades. This 
is indicated by a negative correlation probably 
greater than — 0.30 just to the southwest of 
the Islands, a stronger southerly gradient of 
flow over the Islands, and a much weaker 
positive belt to the north. 
As with Hilo rainfall, summer data fail to 
reveal any useful relationship between the 
wet index and the 700-millibar surface. Figure 
Gc shows the pattern for the season May 
through October. 
DRY INDEX 
In contrast to the wet-index pattern of Fig- 
ure Ga, Figure la shows the pattern for a 
group of four stations chosen from the dry 
areas of the Island of Oahu. In this pattern 
the cyclonic origin of the rainfall is emphas- 
ized and the similarity of this pattern to that 
for Naalehu (Fig. 5c) is marked. The pattern 
of Figure lb, 5-day rainfall amounts for the 
PACIFIC SCIENCE, Vol. VIII, July, 1954 
dry-index stations, indicates that the short- 
period amounts are almost purely dependent 
on cyclonic activity. The correlation coeffi- 
cient — 0.52 is larger than is usually found 
with 5.- day data. 
OBJECTIVE ESTIMATES 
In practice it has been found difficult to 
obtain high multiple correlation coefficients 
by the combination of several selected points 
from the 700-millibar chart. In the case of 
Hilo, for instance, the relationship between 
the 700-millibar surface and rainfall cannot be 
boosted, by any significant amount, above 
the 0.69 expressed by the best simple correla- 
tion coefficient in Eigure 1. In most other 
cases, where reasonably high negative corre- 
lations are found in addition to the positive 
center, a combination of two points — those 
showing greatest positive and greatest nega- 
tive simple correlation— gives a multiple cor- 
relation that is difficult to improve upon by 
the addition of other points. 
Multiple regression equations were devel- 
oped, by means of which the dry index and the 
wet index could each be estimated from a 
knowledge of the heights of two points on 
the 700-millibar surface. Given the wet and 
dry indices, rainfall amounts for various sta- 
tions can be estimated from a knowledge of 
their mean annual rainfalls. The method has 
been used in forecasting by applying the re- 
gression equations to data from the prog- 
nostic monthly mean charts. During the past 
2 years, 24 estimates made by the objective 
method alone had a slightly higher average 
skill score than the final predictions, which 
frequently were modifications of the objective 
estimates. 
For those who may wish to apply the meth- 
od, the regression equations and other essen- 
tial information are given in an appendix. 
COMPARISON TEST 
It was mentioned in the introduction that 
a test had been made on the ability of fore- 
casters to interpret mean charts in terms of 
