to overestimate the vand speeds and above 20 knots, they underestimate 

 them (Fig. ll) . There is furthermore a lack of accuracy due to the use 

 of the Beaufort scale which gives quite a wide range of wind speed 

 values with no ira-y to determine the exact speed. A study made by the 

 Oceanographic Office listed many other reasons for wind inaccuracies of 

 ship reports '"•'. 



Problems Encountered - New York University 



The problem of ship wind observation variability was also studied by 

 New York University. After looking at all aspects of the problem the 

 conclusion was reached that much of the variability may be eliminated 

 by correcting all wind observations to a standard level of 10 m, 19-5 ^^ 

 or some other level above the sea. Anemometers on ships have been found 

 at varying heights anyi/here from 10 to 25 m or more above the sea surface. 

 If all ship wind observations were corrected to the same level by some 

 appropriate method and these used to produce a wind field, then con- 

 sistent wave spectra could be computed. For the work done by New York 

 University the decision was made to correct all wind observations to the 

 19.5 m level since the model being used was based on observations made 

 at this height. 



There were several problems encountered in the computer part of the 

 project. Tvro programs had to be worked out: 



1. An automatic data processing program for eliminating gross errors 

 from ship observations, sorting the ships into priority categories and 

 eliminating all but the priority reports for each given gridpoint 

 location, and finally correcting the wind observations to a level of 

 19.5 m wherever this was required. All this had to be done before the 

 observations could be integrated into the initial regression-wind field 

 obtained from the gridpoint pressure data. 



2. A machine program for interpolating surface winds at inter- 

 mediate time ( 6-hour ly) intervals between the 12-hour ly fields provided 

 by the gridpoint pressure data. 



Both problems were solved by consultation with U. S. Weather Bureau 

 personnel where programs of a similar nature had already been produced. 

 These programs with appropriate modifications were adapted to the New 

 York University computer system after considerable testing. 



In computing the wave spectral fields from the wind fields there 

 were several problems that had to be solved. One of the most serious 

 was the poor specification of the initieil wave conditions. Without any 

 prior knowledge of the shape of the spectr\jm at each of the gridpoints, 

 an assumption had to be made as to how much of the spectrum was due to 

 swell. Since the spectrum of the swell was unknown, it must of necessity 

 be ignored originally. The best solution found was to start the machine 

 computations for several days prior to the desired time for the wave 



^h 



