trurrio A wavelength, of 1680 feet would still not be expected, and a wave- 

 length of 560 feet would just barely be beginning to shov/ up. 



On the basis of tiie transformation needed to go from the theoretical 

 Neumann frequency spectrum to the v/ave number spectrum, it was estimated 

 that tlie peak in the spectrum would fall four or five wave numbers away from 

 the origin, and that the range covered would adequately trace out the details 

 of the shape of the spectruin. The full consequences of these decisions will 

 be discussed in a later section^ 



With respect to statistical reliability, tliere are 16 degrees of freedom 

 for each point estimated on the spectrum for each set of data. Fifty degrees 

 of freedom, are desirable so it was decided to do three pairs of stereo photos, j 

 since, with the same grid alignment of all three, the estimates for each set 

 of data could be averaged to obtain final estimates with 48 degrees of freedom. 

 One of the stereo pairs spot heighted by Hydro turned out to have serious 

 "barrel" distortion in addition to tilt, and it had to be discarded so the final 

 .e3.It= will be ba=.,i on 33 d.g,ees of f.eedo^. | 



A choice of a 60 by 90 grid and 20 lags (really 20 to the left, 20 up and 

 20 down plus all conibinations such as, say, 5 to the left and 17 up) implies 

 861 points to be determined .for the co=variance surface and S6l points for 

 the final spectrurn„ About 4,000,000 multiplications and an equal number of 

 additions are needed to get ea,ch of the co-variance surfaces, and about 

 720,000 iTiUltiplications and additions are needed to get each of the raw spectra. 



Much the same considerations entered in the above choices as enter in 



56 



I 



