magnitude la a surprisingly good result. The ability to predict high 

 frequency roughness from a low frequency model appears to be quite pos- 

 sible, at least for some areas of the world ocean. One serious caution 

 must be taken into account. The successful prediction from the SASS- 

 derlved spectrum was only possible by identifying the break in slope at 

 3 km. Were data only available to that 3 km wavelength, the lower fre- 

 quency estimates would have been used, and prediction error of nearly 

 six orders of magnitude Incurred in estimating the millimeter scale amp- 

 litudes. If it is assumed that such slope breaks are due to changes in 

 relief-forming processes at various frequencies, then any estimates of 

 high frequency roughness exclusively from a lower frequency model pre- 

 sume a continuity of process over all intervening frequencies, and the 

 absence of a significant break in the power law form of the spectrum. 



116 



