Figure 5 is an example of how the data are distributed with 

 respect to space and time. Data taken in waters less than 100 

 fathoms in depth were excluded from the analysis. The data 

 distributions for areas B, C, and E for the years 1952, 1953, and 

 1954 are included. The histograms on page 40 show the monthly 

 distributions and the figures on page 41 the geographical distribu- 

 tions. The nonuniform data distribution with respect to time and 

 space is obvious. Geographically, most of the data are in the 

 area nearest shore. The number of data decreases rapidly to the 

 west, many 10-minute-square areas containing no data. Area C 

 is notable for its nonuniform spatial distribution. Temporally, 

 most of the data were taken during the spring and summer. Area 

 E is notable for its nonuniform temporal distribution. 



The distributions of the observations in the other data sets 

 used in this study exhibit similar characteristics. 



Equation (4) was least-squares fitted to 5 years of data taken 

 in Area A from 1951 to 1955 inclusive and in areas B, C, and E 

 from 1950 to 1954 inclusive. Each data set consisted of sea- 

 surface temperatures as recorded in degrees Fahrenheit on 

 bathythermograms taken during the indicated year and in the 

 indicated area. The results of the 20 individual regression 

 analyses are presented in figure 6. For each analysis the number 

 of observations, the multiple correlation coefficient, the percent 

 of the variance explained by regression, and the standard devia- 

 tion of the observations about regression are shown. The last 

 analysis utilized the data taken in the four 1 -degree -latitude 

 strips. As indicated in figure 7 these measurements were treated 

 as a single sample drawn from the area 30 to 49 N and extending 



39 



