Another interestin-^ feature of these results Is the fact 

 that the prediction was much more accurate at greater depths 

 than at the surface. The example shovm in table XII includes 

 only one component of the diurnal wave; actually both components 

 for both the semidiurnal and diurnal wave were computed. In 

 all four cases the results displayed the same inaccuracy near 

 the surface and the same accuracy at the greater depths. Whether 

 this particular inaccuracy is due either to one of the above 

 two reasons or to some near-surface phenomenon not considered 

 in the computations has not as yet been determined. 



It will be noted that the technique outlined above does 

 not, according to strict terminology, constitute a true fore- 

 cast, since either an observation or an assumption of a present 

 density distribution is necessary to employ the method. However, 

 It is considered a large step forward to be able to estimate 

 the amplitude of the wave from a single observation, rather than 

 from a complete anchor station with many time-series observations. 

 If some such method as described above were to be developed, it 

 would automatically increase the chances of predicting the 

 internal wave, for it is not possible to occupy a complete anchor 

 station every time a forecast is desired. If one harmonically 

 analyzed all existing and future anchor stations and computed 

 the Fourier coefficients for each, then at any future- time the 

 internal wave could be estimated for each of these areas. The 

 only data one will need for this operation would be a single 

 density distribution. This observation could be either (a) a 

 recent observation taken in the area, (b) an estimated distri- 



67 



