deviations of anomalies at the individual ocean weather stations and the 

 same parameters computed for the five combined Worth Atlantic ocean weather 

 stations. The means and standard deviations of anomalies based on all data 

 are given in column h. Since many of the values included in these- means 

 are insignificant in correlation of the anomalies at the surface and 1+00 

 feet, only the values exceeding 0.5°F have been used. These values are 

 listed in colimn 5, 



The true nature of small anomaly differences between the two levels 

 is not certain. The main difficulty in interpretation arises from the 

 fact that BT error varies between the two levels and that the same instru- 

 ment is often used during a large part of one month at one station. Thus, 

 relationship of small differences of anomalies with different signs at 

 the two levels cannot be determined, because the instrumental error remains 

 in the monthly mean value. The author (5) has shown that 81 percent of 

 instrumental errors between the two levels are less than 0.5°F. Exclusion 

 of anomalies less than 0.5°F permits more conclusive computations. In a 

 total of iK)5 pairs of surface and i^DO-foot anomalies at the five North 

 Atlantic ocean weather stations, 238 pairs exceeded 0.5°F. 



A further attempt to identify data representing reliable linkage 

 between anomalies at the siurface and ^l-OO feet resulted in the values 

 listed in column 6, table h. This column includes only those observations 

 in which anomalies are greater than 0.5°F and of the same sign at both 

 levels. In the total of 238 cases, 201 pairs met these conditions. 



IV. CORRELATION 



Linear correlation coefficients computed for the same three groups 

 of data in table h are shown in table 5« Values in the first column were 

 computed from all available observations. The correlations at all North 

 Atlantic stations, except CHARLIE, agree closely. The value at CHARLIE 

 does not improve substantially by limiting computation to anomalies 

 greater than 0.5°F; however, by using pairs of anomalies exceeding 0.5°F 

 and with the same sign, the value becomes alined with those of the other 

 stations. Some anomalies of different origin at the two levels are 

 eliminated by excluding anomalies with different signs in the computations. 

 However, the remaining anomalies of different origin but coincident in 

 sign constitute a deficiency in computations, because all anomalies are 

 treated as though they were of the same origin. Therefore the third 

 column of correlation coefficients in table 5 can be considered minimal 

 for anomalies of the same origin. With 201 pairs of anomalies for the 

 combined North Atlantic stations the correlation coefficient is O.835. 

 With z-transformation to normeO. distribution, 95 percent confidence 

 limits are O.778 and O.879 for n = Uo (mean number of pairs for individual 

 stations). All correlation coefficients of the individual stations exceed 

 the lower limit. All five stations are located in the western North 

 Atlantic and the number of points is extremely small; however, since their 

 distribution is well balanced with respect to the main surface water masses 

 in the North Atlantic, the computed coefficients may be considered fairly 

 representative of the entire Noarth Atlantic, 



18 



