ent bottom types. Analyses presented In Chapter 5 will link this 

 "universal" relationship 



_1 

 A » a . s 



to the absence of stationary provincing techniques, 



Extension of Spectra into High Frequencies 



The component of random noise present in all empirical data, 

 Includes uncertainties from many possible sources in the total meas- 

 urement system. There are uncertainties associated with the measuring 

 device itself, for example, the errors in the timing of the arrival of a 

 sonar pulse. The interference of external sources, such as radio waves, 

 may also affect the measuring instruments. The nature of the environ- 

 ment between the detection device and the process being sampled may 

 introduce error, such as the variability of sea water sound velocity in 

 bathymetric surveying. The truncation of significant digits in the 

 storage of digital data introduces a finite level of noise, often called 

 "round-off error." All of these sources combine to form a total noise 

 level for a measured data set. 



With the exception of round-off error, which is calculable, the 

 source of these random errors can not be decomposed. However, using 

 spectral techniques the level of the total noise can be estimated. It is 

 a well-known property (see Bracewell, 1978, Chapter 16, for a complete 

 discussion) that the spectrum of a randomly generated signal varies 



28 



