center would result in large wave-spectra errors. Since the wind speeds at 

 station A have by far the largest bias and RMS error compared with stations I, 

 J, and K (-13 and ±17 knots, respectively, versus -3 to -7 and ±8.5 to ±12.3 

 knots), the discrepancies in the wave spectra would seem to be caused by mis- 

 calculations in the wind field rather than by any fault with the wave-spectra 

 procedure. This points up the fact that the automated numerical forecasting of 

 wave spectra can be no better than the capability of making accurate wind- 

 field predictions. 



CONCLUSIONS AND RECOMMENDATIONS 



1 . The evaluation of the automated numerical wave-spectra prediction 

 program shows that the procedure is feasible and that it produces results which 

 are within a reasonable degree of accuracy with the present state of the art. 



2. The evaluation indicates that there is need for improvement in pre- 

 dicting surface wind fields over ocean areas and that the reliability of wave- 

 spectra predictions closely follows the reliability of wind-field forecasts. 



3. The evaluation discloses that there is no great difference in the 

 accuracy of the prognoses of the two different inputs. If it were necessary to 

 decide between the two, the evaluation suggests that the FNWF might be 

 slightly superior, although this could be due to the maximum forecast interval 

 being only 30 hours for the FNWF input as against 36 hours for the USWB input. 



4. The evaluation reveals some remarkably good correlations between 

 the predicted and the observed wave spectra. On the other hand, for station A 

 and Argus Island the entire set of predictions could be considered as practically 

 a "bust". Some reasons for this were listed in the preceding discussion section. 



5. It is believed the evaluation gives ample proof that the automated 

 numerical wave-spectra predictions would be valuable for operational use in 

 most areas of the North Atlantic Ocean and that this procedure should now take 

 precedence over the less sophisticated automated and manual methods. 



6. As refinements in wind-field forecasting and in wave-spectra models 

 become available, they can be readily adapted into the automated numerical 

 procedures evaluated in the report, resulting in continuous future improvements 

 in the wave-spectra predictions. 



