higher amplitude and therefore the signal la these frequencies Is "vis- 

 ible" above the RMS noise. The frequency at which the spectrum of the 

 signal Intersects the noise level represents the highest frequency that 

 Is being resolved In the system. 



Figure 4-5 Illustrates these concepts on a spectrum of SASS data. 

 Notice that Figure 4-5 shows an obvious noise level while the spectrum 

 shown In Figure 4-3 does not. Both sets of data were collected using 

 the same sounding system within days of each other, and It Is expected 

 that the Instrument noise In each Is approximately equal. The differ- 

 ence Is therefore that the data In Figure 4-3 represent a rougher area 

 In which the signal energy In the highest frequencies Is sufficient to 

 maintain the resolution of the signal above the noise. The noise level 

 Is present In both sample spectra, however. It was never Intersected In 

 the sample from higher energy sea floor. The ability of a sonar to 

 resolve horizontal features depends not only on the horizontal resolving 

 power limitations of the Instrument (normally limited by the size of the 

 "footprint"), but also by the amplitude of features present In the sea 

 floor. 



In light of the limitations of noise In all measurement systems, a 

 fundamental question In the development of a stochastic model of sea- 

 floor roughness from widely available surface sonar. Is how far Into the 

 high frequencies the model can be extended. One could argue that due to 

 the persistent exponential form of the spectra noted In this study, as 

 well as the work of Bell (1975b), that this functional form can simply 

 be mathematically extrapolated Into higher frequencies. Rirther justi- 

 fication might be provided by the generation of spectra from deep-towed 

 Instrument packages, bottom photographs, and direct observations, to 



30 



