a commerical basis. Wave heights are fore- 
cast, as previously stated, by the Wave 
Spectrum Method developed by Pierson, 
Neumann, and James in1955. The meteoro- 
logical inputs to the method are supplied by 
synoptic data and the 5 day weather forecast 
maps of the Northern Hemisphere issued 
every Monday, Wednesday and Friday by the 
Extended Forecast Section of the Weather 
Bureau, 
The observational network which supplies 
the data for the overlapping 5 day forecasts, 
in addition to continental and island based 
stations, is made up of 10 weather ships in the 
North Atlantic (positioned between 35 and 68 
degrees north latitude), and 3 in the Pacific. 
Augmenting the observations made by these 
stations are those made by naval, coast guard, 
and merchant vessels, The data from these 
sources are, of course, variable both in space 
and time. As a general practice, for instance, 
aircraft carriers report weather observations 
every 6 hours, destroyers every 12 hours, and 
merchant vessels every 24 hours. 
On an average day there is an estimated 
4000 ships at sea in the Atlantic area. Con- 
sidering the fact that weather forecasting over 
the ocean is considered to be simplier than 
over land this coverage, at first glance, would 
seem to be more than adequate. Most of this 
number, however, is made up of merchant 
vessels, only a small percentage of which 
make regular weather observations. This, 
coupled with the fact that ships tend to con- 
centrate along well defined trade routes, con- 
siderably reduces the potential number and 
distribution of mobile weather stations, A 
further restriction on the number of weather 
reports fed to the forecast centers is imposed 
by communication delays. Of the overall total 
of about 10,000 weather observation stations 
in the Northern Hemisphere that could report, 
an average of only 3000 are received. The 
end result is that weather maps muSt occa- 
sionally be extrapolated over considerable 
236 
areas having insufficient data. It should 
also be mentioned that the sea based 
observational network, inadequate now 
from the standpoint of optimum distri- 
bution, will deteriorate rapidly under war 
conditions, 
Very little data is readily available 
on the forecast aCCUEACY. of the present 
program, James reported early in 
the development of the program that a 
comparison of the forecast and observed 
wave heights at some 50 grid points across 
the North Atlantic during winter months 
showed that 85 percent of the forecast 
wave heights were within plus or minus 
4 feet of the observed heights. Since many 
of the waves measured were 25 feet and 
greater in height this was considered an 
acceptable error at the time. An un- 
published average error estimate of plus 
or minus 21 per cent has been expressed by - 
one of the forecast units. Either of these 
errors would have to be reduced by approx- 
imately 50 percent in order to provide the 
accuracy desired for the prediction of those 
sound conditions depending upon the state 
of the sea. Some researchers believe that 
a 90 percent accuracy can be achieved *”, 
but much work will be required before this 
goal is reached, 
A significant increase in the present 
forecast accuracy will require effort on 
several fronts. Weather forecasts can, 
at best, be only as accurate as the data 
upon which they are based. It goes without 
saying that, due to the human factor in- 
volved and such easily corrected things as 
the lack of anamometers on many of the 
reporting merchant vessels, improvements 
can be made in data acquisition, Even with 
adequate and accurate data, however, our 
present level of understanding of the basic 
meteorological and oceanographic forces 
involved is not such that perfect forecasts 
could be made. 
