120 



ANNUAL REPORT SMITHSONIAN INSTITUTION, 19 3 5 



Stornoway temperature, and also of the quantities which decrease 

 when the former increase, such as Iceland pressure, reversing these 

 so as to secure similarity of the graphs. We could then draw a 

 graph which is the mean of all these, and could regard it as ex- 

 pressing the variations of the North Atlantic fluctuation as a whole. 

 (See the lowest graph of fig. 1.) If now we were to plot Labrador 

 temperature below it we should see that its variations were, like 

 those of Iceland pressure, strongly opposed, and on reversing Lab- 

 rador there would be very strong similarity. So Labrador becomes 

 a good example of the second group. Now we want to know the 



I S5 I I I I 10 I I I I 95 I I I I 00 I I I I OS I I I I IP I I I 



FiGUKE 1. — North Atlantic oscillation. 



effect of the North Atlantic oscillation on the pressure, temperatures, 

 and rainfall of a large number of places ; and if in this way we put 

 a hundred graphs under one another, some easy to classify and 

 some doubtful in character, it would be difficult to draw satisfactory 

 conclusions in a manner capable of convenient and accurate ex- 

 pression. So instead of graphs we use numbers, Havmg found by 

 preliminary investigation the stations which are most representa- 

 tive, we calculate the figures in successive years for the North At- 

 lantic oscillation as a whole, and then work out the correlation 

 coefficients of this with the pressures, temperatures, and rainfalls 

 of all the places in which we are interested. These coefficients are 

 plotted in figure 2, and in its top chart we see that the rise of 



