274 
PACIFIC SCIENCE, VoL VIII, July, 1954 
With regard to the use of only two points, 
practice seems to indicate that they may be 
adequate as estimators from an observed map 
but inadequate when used with prognostic 
charts. This is due to the fact that a prog- 
nostic chart may show considerable skill in 
anticipating the large-scale features of the 
observed map a month later and yet be in 
error at the key point or points on which the 
rainfall forecast is based. Skill scores for Hilo, 
for instance, have been significantly lower 
than those for stations on the Islands farther 
north, and yet the correlation field shows a 
stronger relationship of the observed 700- 
millibar chart to Hilo than to any other 
station. 
CONCLUSIONS 
Objective long-range estimates of Hawaii- 
ian rainfall from observed 700-millibar pat- 
terns can be made with a general accuracy 
at least equal to that attained by a forecaster 
using conventional methods. The correlation 
fields shed some light on the mechanisms of 
mean monthly rainfall processes in Hawaii 
and suggest limits to our ability to make rain- 
fall estimates from a knowledge of the 700- 
millibar field. 
The use of key points as described in this 
study comprises a forecast system which is 
very sensitive to errors on the prognostic 
charts. It may be possible to overcome this 
difficulty by the use of orthogonal methods 
of map description. 
Substantial improvement in monthly rain- 
fall forecast accuracy will depend in part on 
improvement of the prognostic 700-millibar 
charts and in part on the discovery of rainfall- 
related factors other than those found in the 
mean pressure field. 
APPENDIX 
Regression equations for estimating Ha- 
waiian monthly rainfall amounts from prog- 
nostic mean monthly 700-millibar charts. 
(Height differences are expressed in tens of 
feet.) 
(1) W = 0.034A-0.01B 
where W = Oahu-— Wet 
A = difference from normal 700- 
mb. height at 42°N-l63°W 
B = difference from normal 700- 
mb. height at 20°N-l60°W 
(2) D = 0.02C-0.04E 
where D = Oahu — Dry 
C = difference from normal 700- 
mb. height at 49°N-163°W 
E = difference from normal 700- 
mb. height at 22°N-l63°W 
When the Oahu-wet and Oahu-dry indices 
have been computed, an individual station 
can be estimated as follows: 
(3) t =^D+ Ra-23 (W-D) 
® 89 
where t =^the expected standard depar- 
ture from normal (units of one 
standard deviation) of the rain- 
fall for a given station, and 
Ra = the mean annual rainfall of the 
station. 
The constants 23 and 89 are, respectively, the 
rounded mean annual dry-index and wet- 
index-dry-index differences as determined 
from Table 1. 
Some stations, such as Hilo, which may 
not fit into the wet- and dry-index system of 
classification should not be estimated by 
means of formula 3. The formula used for 
Hilo is (4) tH = 0.04F 
where tn^the expected standard depar- 
ture from normal of Hilo rain- 
fall, and 
F = the difference from normal 
700-mb. height at 37°N - 
163°W. 
REFERENCES- 
Aubert, E. J. (MS.) A test of monthly preci- 
pitation forecasting in the Hawaiian Is- 
lands. On file in Extended Forecast Section, 
U. S. Weather Bureau, Washington, D. C. 
Haurwitz, B., and R. A. Craig. 1952. At- 
mospheric How patterns and their repre- 
