is derived by fitting the amplitude estimates with one or two power law 

 functions. 



The distribution of the model parameters is examined for a test area 

 adjacent to the coast of Oregon. The distribution of roughness corre- 

 sponds generally with the various physiographic provinces observed in 

 the region. Additional complexities are apparent in the roughness model 

 which can not be inferred by simply studying the bathymetry. These pat- 

 terns are related to geological processes operating in the region. 



In many cases, the roughness statistics are not constant when 

 observed in different directions, due to the anisotropic nature of the 

 sea-floor relief. A simple model is developed which describes the 

 roughness statistics as a function of azimuth. The parameters of this 

 model quantify the anisotropy of the sea floor, allowing insight into 

 the directionality of the corresponding relief-forming processes. 

 Finally, the model is used to successfully predict the roughness of a 

 surface at scales smaller than those resolvable by surface sonar 

 systems. The model regression line (derived from a hull-mounted sonar) 

 is compared to data from deep-towed sonars and bottom photographs. The 

 amplitude of roughness is predicted to within half an order of magnitude 

 over five decades of spatial frequency. 



