Figure 7-2 preseats a composite amplitude spectrum showing data 

 from all three sources In the HEBBLE area. The SASS derived spectrum 

 (from the nearest available trackllne) spans wavelengths of 100 km to 

 200 m. The model regression lines, derived from spectra generated 

 throughout this large province, are shown as straight line segments. 

 The Deep-Tow derived spectrum, spanning wavelengths of 1 km to 10 m. Is 

 very well predicted by the model regression line. The lower frequency 

 estimates begin to diverge at X= 300 m., which may be due to positioning 

 distortions from the transponder navigation system, which Is Interro- 

 gated at 5 minute (or ~300 meter) intervals. Finally, the bottom photo- 

 graph derived spectrum, spanning wavelengths of 25.6 cm to 2 mm, falls 

 approximately .5 orders of magnitude below the regression line. 



The prediction residual of .5 orders of magnitude over 5 decades of 

 frequency indicates some combination of errors in parameter estimates a 

 and b. If all errors were in a, it would be in error by .5 orders of 

 magnitude. Similarly, the error in b would be -.1, using the relation- 

 ships described in the last section. The evaluation of several more 

 bottom photograph-derived profiles from the same region would allow a 

 more quantitative treatment of the model prediction error. It is pos- 

 sible that other high frequency spectra may scatter around the predic- 

 tion line, due perhaps to actual variability of the microrelief in the 

 area rather than error in the prediction model. Visual inspection of 

 the several photograph pairs taken in HEBBLE indicate that other areas 

 are in fact rougher than the relatively featureless photo analyzed here 

 (Arthur Mowell, personal communication, 1983). 



In any case, the prediction of millimeter scale topographic ampli- 

 tude from a surface hull-mounted sonar system to within half an order of 



114 



