BROWNING: ENVIRONMENTAL FACTORS AFFECTING LOW FREQUENCY 

 PROPAGATION IN THE OCEAN 



The amount of attenuation can also be varied in the propagation- 

 loss program. Figure 10 shows examples for various values of attenua- 

 tion: one, three, and six times Thorp's coefficient. We took our 

 experimental data (and again we are limited because we are using shots 

 at discrete ranges, so we cannot see the fine detail that is predicted 

 with a model) and chose a value of attenuation for the program which 

 yielded the best fit. Figure 11 is the comparison of our experimental 

 data with the FFP prediction program using an attenuation value of 

 0.054 dB per kiloyard at 400 Hz. A similar fit was obtained at 1000 Hz 

 with attenuation of 0.113. 



Now, we asked, if we get this value of attenuation for our best 

 fit by overlaying the curves, what would we obtain if we applied the 

 same linear fit that we used for our experimental data? That is, if 

 we make a least squares fit, what value of attenuation would we come 

 up with? The answer (see Figure 12) was 0.1126 compared with 0.1135. 

 So there was hardly any difference between taking a straight-line fit 

 to the data and overlaying the curves. Much to our surprise, these 

 data points generated using a sophisticated modeling technique were 

 almost identical to plain cylindrical spreading in the sound channel. 



Figure 13 shows estimates of error. The squares indicate the 

 amount of variation which gave an equivalent fit to our particular 

 data points. You can see there was nothing here to convince us that 

 cylindrical spreading was not as good as any other model for the case 

 that both the receiver and source are on the sound-channel axis. 



We have used the same analysis many times, for example in Lake 

 Superior (a case where we felt there might be a problem) , and we have 

 come up with the same answer . 



781 



