Summary 



In summarizing the preceding sections, a tentative strategy can be 

 formulated for approaching the stochastic modelling of sea-floor 

 roughness. 



(1) Delineations of homogeneous provinces , using a province 

 picker, such as that developed by Davis (1974), would Insure 

 some degree of statlonarlty, and therefore validity, for the 

 statistics describing the area. As will be discussed in the 

 following chapter, this province-picking algorithm must be 

 based on the same statistical measure that underlies the 

 model, that is, the frequency spectrum. 



(2) Generation of amplitude spectra from available data within the 

 delineated provinces would follow. Sample lengths for spectra 

 generation would be confined to within the province boundaries 

 defined in stage 1. One-dimensional spectra would describe 

 the roughness in several orientations to provide inout for 

 later modelling of topographic anlsotropy. 



(3) Regression analyses would be performed on these amplitude 

 spectra to generate the coefficients of the functional form 

 chosen to represent the spectra. Preliminary indications are 

 that this form will be one or several power law relationships, 

 each of which require only two coefficients to describe. 



(4) Anlsotropy of the bottom would be modelled by studying the 

 variation of the coefficients a and b of the power law model 

 as a function of direction 6. In areas where complete Infor- 



42 



