Although this represented a duplication of data, it provided a comparison 

 between two different prognostic techniques as well as providing fill-ins 

 for missing data in either of the two sets of data. The duplication procedure 

 was found to be desirable in the operational test program since there were 

 several occasions of missing data in one or the other sets. 



The "real time" operational test program which was started in the 

 summer of 1966 (17 August 1966) continued until 11 March 1967, a total of 

 almost seven months. Data transmissions of sea surface pressures and observed 

 ship reports for six-hourly synoptic times and for six-hourly prognostic times 

 were made by a dataphone link from Monterey, California and Suitland, 

 Maryland to New York City. The transmitted data were used at New York 

 University as the input for data processors to give six-hourly synoptic and 

 prognostic directional wave spectra for periods up to 36 hours. Preliminary 

 detailed reports have been given by Moskowitz (1966, 1967). 



For each six-hourly synoptic time, a first -guess wind field analysis 

 was made based on either the FNWF or the USWB meterological surface 

 pressure field. In both sets of data the "surface" wind was taken to be the 

 wind at the 19.5 meter level since the wave forecasting model was formulated 

 from the recorded wind observations of the Ocean Weather Ships (OWS) which 

 have anemometers at this elevation. The first-guess wind fields were then 

 modified by using all the available ship wind measurements corrected to 19.5 

 meters, so that the wind analyses could be made as accurately as possible. 

 The shipboard wind estimates were used when measurements were not available. 

 The data, however, were weighted according to a predetermined priority 

 scheme. A relatively small error in the wind speed can make fairly large 

 errors in the forecasted wave spectra, hence great care must be taken in 

 constructing the wind field. For the prognostic analyses, of course, no ship 

 observations were used. 



The wave spectra output began with the computations of several days 

 of hindcasts up to the 0000Z analysis for the day on which the prognostic spec- 

 tra were to be made. This up-dating procedure was found to be essential in 

 order to allow the model time to build up realistic wave spectra. The six- 

 hourly prognostic data were made on a daily basis beginning at 0000Z for a 

 total of 30 hours or for five separate forecasts for the FNWF input and for 36 

 hours or six separate forecasts for the USWB input. Each wave spectrum included 

 the following: (1) date, (2) time, (3) gridpoint number, (4) wind speed, 

 (5) wind direction, (6) 180 spectral components in 30-degree intervals of 

 the compass and 15 intervals of frequency, (7) the sums of each of the 15 

 frequency intervals for all directions, and (8) the computed significant wave 



