this presentation, no geographical distinction has been made although 

 results are included from runs a few miles from shore as well as 250 

 m|les at sea. In spite of the heterogeneous collection a general trend of 

 horizontal gradients is clearly indicated. 



Most of the statistical calculations have been limited so far to the 

 early California and Hawaii data. However, Figure 9 shows the variance 

 in degrees centigrade for runs of about 12 nautical miles long. These 

 data indicate a possible difference between data taken in Hawaii and Calif- 

 ornia. A comparison with Figure 8 shows that the horizontal gradient in 

 ° C/nm covers two decades down to 2000 m whereas the variance is con- 

 tained within one decade. In order to estimate the effects of these sound 

 velocity variations on the fluctuation of acoustical signals, we need the 

 auto-correlation function as well as the variance. Only preliminary re- 

 sults are available as yet. Figure 10 is typical for the early California 

 data and the Hawaii data. The first zero crossing is several thousand feet. 

 This has complicated the analysis since long records must be processed. 

 In order to emphasize the higher frequency, short correlation variations, 

 two shallow (50 m depth) Hawaii runs were filtered digitally to remove 

 space wave lengths longer than 20 feet. Figure 11 shows a typical result. 

 The correlation function is not significant since, as noted on the figure, 

 the variance is less than 0.002° C. 



Analysis of these data is continuing with the expectation that acoustical 

 applications will be forthcoming. A large amount of data is required be- 

 fore generalizations are reasonable and adequate summaries can be pre- 

 sented. In addition, we must appreciate that the sea is incomparably 

 larger than the minute areas covered in the three cruises reported here. 



356 



