186 MARION- EXPEDITION TO DAVIS STRAIT AND BAFFIN BAY 



The monthly values for (c') were then correlated with (a) giving 

 the following coefficients : 

 Between (a) and {c') — ■ 



It is apparent from this that the November and April gradients 

 have little elfect on the iceberg dispersal, probably because in Novem- 

 ber the berg train is just putting in its appearance while in April 

 the character of the drift is well established. The values of 0.48 and 

 0.43 in December and January respectively indicate, furthermore, 

 that those two months exert twice as strong an influence on the ice 

 as do any of the other months of the year. When the ^^alue of (c') 

 was expressed in monthly miits and combined with the following 

 weights : 



2 X Dec. + 2 X Jan. + 1 X Feb. + 1 X Mar. 



the 4-monthly group values of (c') correlated with («) yielded 

 the high coefficient of -0.58 for the period 1880-1926. 



The Stykkisholm-Bergen pressure gradient gave the following 

 coefficients wdiich show that its iceberg effect does not conunence until 

 October and is mostly ended by February. 



Between (a) and (d) — 



The months were accordingly arranged in the following manner: 

 1 X Oct. + 1 X Nov. + 2 X Dec. + 1 X Jan. 



Emploving seasonal values combined as above, we obtained a coeffi- 

 cient of -0.62 for the series 1880-1926. 



A study was also made of the degree to which the angle that the 

 isobars make with the Labrador coast (see (e)) alfects the number of 

 icebergs passing Newfoundland. It had been assumed that the pres- 

 sure gradient between Ivigtut and Belle Isle was nearly at right 

 angles to the most effective direction of the w ind ; an angle of about 

 38° wdth the Labrador coast. Correlation values indicate that 30° off 

 the general coastal trend is the most effective direction, and since 

 the run of the wind in high-pressure centers is about 30° across 

 the isobars, the mo.st effective w^ind direction is, therefore, approxi- 

 mately 60°. 



The remaining variates tested in the investigation w^ere {(/) and 

 (h), with a lag of one year, which gave the following coefficients: 



Betw^een («) and (g) — 



