ficially investigated and found to require a computational 

 treatment far longer than the limited staff of the project 

 could supply. With the aid of machine computation, however, 

 this problem c :iuld be attacked in the future. Therefore, it was 

 decided to proceed with the original assumption. 



Let A]^ be one of the two constituents of the amplitude ot a 

 wave computed from the observed data at one particular time ac- 

 cording to Pjeldstad's theory. Let A2 be the corresponding un- 

 known component for some future time at the same place. At this 

 future time, let one observation of vertical density di.stribution 

 be available so that the w's are known. 



4 2 



Let t (ainWin-agnWgn) =I(a2n) 

 n-i 



n = l,2,...4 (29) 



Now by minimizing the function I we are minimizing the change of 

 the component of the amplitude of +"he wave A]_ (see equation 27). 

 The minimizing of the function I may be accomplished by the so- 

 lution of four simultaneous equations which will yield the un- 

 known a2n» This computation is long and involved and, for the 

 example mentioned below, will not be repeated here. It was car- 

 ried out by desk calculator by using the inverse matrix approach. 

 For the harmonic analys r' 3 and the numerical integration, tech- 

 niques identical with those of Pjeldstad were employed. 



The type of data required to test this method is that ob- 

 tained at a standard anchor station of at least 2l\. hours' dura- 



62 



