was created uslag a grid spacing of 0.25 minutes of longitude and lati- 

 tude (~460ra X ~300m), well above the resolving capability of the SASS 

 system. 



The area shovra In Figure 6-4 was selected to test the effect of 

 anlsotropy on one-dlmenslonal amplitude spectra. The data set Is 

 located In the vicinity of the ridge crest, which was Identified as a 

 quasl-statlonary province by the methods described In Chapter 5. The 

 llneatlon of the topography Is obvious on the Index chart and trends 

 approximately N25**E. 



Figure 6-5 represents graphically the test area. Although the grid 

 spacing used for the Illustration Is 0.1 minutes of latitude and longi- 

 tude, the profiles were generated from a grid with spacing of 0.05 min- 

 utes (~100m), which approaches the resolving limit of the SASS system 

 for these water depths and noise level conditions. Again the sampling 

 pattern shown In Figure 6-2 was used to generate radial profiles of 256 

 points. No ensemble averaging was used. 



The spectral parameters a( 6) and b(6) are plotted versus azimuth In 

 Figure 6-6. The results agree closely with the theoretical model devel- 

 oped In the previous section. Notice first that the parameter b(6), 

 plotted above In these diagrams and labelled "Slope of Spectrum", shows 

 no systematic fluctuation with azimuth, as predicted by theory. Note 

 also that the mean slope Is -1.24, well below the -1.0 slope of the ran- 

 dom walk (Markov process) model. The large variability of this param- 

 eter could be reduced by ensemble averaging of several spectral esti- 

 mates, created by offsetting the center of the sampling pattern. The RMS 

 variability would be reduced by 1/ / N , where N is the number of esti- 

 mates (see Chapter 7). 



90 



