some elevation other than that from which the spectr^,! model .was origi- 

 nally derived. Thus, the lack of wind profile considerations in using 

 ship wind reports may well result in some considerable errors of the 

 wave spectra. 



Problems Encountered - Travelers 



The best wind estimates obtained by the screening regression 

 technique contained rms errors of 8.5 knots for each wind component. 

 Since small errors have a rather large effect on the spectral results 

 it was felt that this technique could not be used for the wind field 

 in computing wave spectra. The errors in the regression technique were 

 most likely inherent because of the coarseness of the gridpoint pressiare 

 data and because the machine interpolation techniques tended to flatten 

 out the presstire gradients in the vicinity of pressure centers. 



The solution to the above serious problem was to use the regression 

 winds as an initial estimate of the wind field and to integrate selected 

 ship wind observations with the regression winds to obtain an objective- 

 analysis wind. A machine program for integrating the selected ship wind 

 observations with the initial guess regression winds was formulated by 

 Travelers. The technique finally used by New York University was a 

 modification of the Travelers programming. In the N.Y.U. program the 

 ship wind observations to be used were selected for use in the initial 

 guess field in the following priority: 1. All weather ship wind obser- 

 vations, 2. Observations of U. S. Navy ships with anemometers, 3- 

 Observations of all remaining ships not located within a certain grid- 

 point range of the first two categories with anemometers, h. Observa- 

 tions of transient ships without anemometers. Before these ship wind 

 observations were used, the low centers on each 12-hourly synoptic . 

 map were located according to the nearest gridpoint and the corresponding 

 vyind fields at these positions were translated to the point half way 

 between and averaged inside a circle of appropriate radius. Before the 

 ship wind observations were entered in the priority order stated above 

 they were all corrected to the standard height of 19-5 m when the 

 anemometer heights could be ascez-tained. Only one ship 'wind observation, 

 supposedly the most correct report possible for that location, was used 

 to correct the initial guess wind at any given gridpoint. 



During the time before the final procedure was adopted, there was 

 considerable variance of opinion on the validity of many ships' wind 

 observations as compeired to the true wind field. One viewpoint was that 

 the variance was at least 5 knots or greater and the other that it VTas 

 less than 5 knots. At present there is no simple way to determine which 

 viewpoint is correct. When wind observations are estimated on the basis 

 of sea appearance, as is done on most ships without anemometers, the 

 winds are usually underestimated when the sea is building up and over- 

 estimated when the waves are dying. down because of the lag effects on 

 the sea for increasing or decreasing winds. A study by the U. S. Weather 

 Bureau^ 5) also found that up to about 20 knots the transient ships tend 



23^ 



