that the higher values form a zone directly off southern Baja 

 California. The low -value contours occurring in the northeast- 

 southwest zone lie in the same position. Of all isotherms con- 

 sidered, those recorded from the region 60 miles to the south- 

 southeast from the extremity of the peninsula are generally 

 steeper and show more variability than those elsewhere in the 

 area. 



Autocorrelation of Depth Values 



Another method of measuring subsurface temperature 

 variability is by means of autocorrelation coefficients. 18 By 

 using the same half-minute isotherm depth data, autocorrelations 

 were computed for each selected isotherm and section of Cruise 

 8. Successive pairs of points at equal but overlapping time 

 intervals were tested for correlation with each other, and the 

 process was repeated for each time interval, increasing by half- 

 minute steps from one-half minute to 72 minutes (144 lags) or 

 10 percent of the number of data points. Autocorrelation, R x , 

 was computed for increasing intervals of X, each interval in- 

 creased by 304 feet (half minute), using the expression: 



N- A N- A 



{N-X) Y ZiX i + x ^Xi J X < + 



1 i=i 



R, = 



(N-X) J X, 2 - 



"\- A 



2 



1/2 f 



I* 





■ 



i= 1 







(N-X) Jx 2 H 



X^-K 



where, in general, X = 0, 1, 2, . . . 144 lag intervals andN = 

 total number of depth recordings in a run. N is usually 1440 and 

 the maximum value of X is N/10. The computed autocorrelations 

 of the selected isotherms on each sample section were plotted 

 for comparison (Appendix C). 



One example of the 52 autocorrelations (fig. 24) is from 

 data derived from section Q (fig. 15). Starting with zero lags 

 (zero minutes) the autocorrelation it A is 1.0, but as the lags 

 increase, the correlation becomes less. In some cases the 



48 



