SURGES. 



199 



The correlation coefficients between the stations Hut Point, Gauss Station and Snow Hill 

 are very large, all being -8 with a probable error of less than 1. 



A certain amount of the high correlation is probably due to the yearly variation of the 

 pressure being similar at all the stations. Unfortunately, however, we do not know the normal 

 yearly variation at the Gauss Station and Snow Hill. We know it approximately for McMurdo 

 Sound from the smoothed values of the four years' observations. If the correlation coefficients 

 are increased by the yearly variation, it can only be because the variation is similar at all 

 three stations ; we are therefore justified in deducting the yearly variation of McMurdo 

 Sound in order to free the relationship from some at least of the dependency on the yearly 

 variation. 



In the following table the first column contains the normal yearly variation in McMurdo 

 Sound and the three following columns the monthly departures, given in table 116, corrected 

 for this yearly variation. 



Table 118. 

 Pressure lariation corrected for yearly variations. 



These corrected monthly departures have been plotted in the three lower curves of figure 

 60. It will be seen that even now the curves are very similar especially if the changes from 

 month to month are considered, for eight of the eleven changes take place in the same 

 direction on all three curves. 



The correlation coefficients have been reduced somewhat by applying the correction, 

 showing that the annual variation is similar at all three stations.* 



* It should bo remarked that if the stations had not the same or sunilar variation, the effect of dcductino 

 the same variation from all three curves would have been to increase their correlation coefficients. This ia 

 important in relationship to deciding the type of yearly variation at the Gauss Station and Snow Hill for which 

 we have not sufficient observations to form a tnie yearly variation. 



