SCIENTIFIC RESULTS 185 



111 vioAv of the f()ri>L^)injr. it was considered desirable to iiivesti<i:ate 

 quantitatively tlie etiects of the foll()win«>' variates, be<i;iiiniiig with 

 those which appeared to bear the closest relationship to the number 

 of icebergs so far as available data might allow : 



{(f) Represents the number of iceberg's south of Newfoundland 

 (the forty-eighth parallel of latitude). 



We also Inne : 



{/j) Amount of i)ack ice ott Newfoundland, February to May. 



(c) Atmospheric pressure ditferences (mb) Ivigtut minus Belle 

 Isle in November to Ai)ril. preceding the iceberg season off New- 

 foundland. 



(d) Atmospheric pressure difference (mb) Stykkisholm minus 

 Bergen. August to January, jireceding the iceberg season off New- 

 foundland. 



(() Cross gradient pressure Ivigtut minus Belle Isle, December 

 to March. 



(/) The i^ressure anomaly at Stykkisholm, December to March. 



{(/) The monthly precipitation at Upernivik, Greenland. 



{h ) The temperature during June, July, and August at Upernivik. 



The correlation coefficient between (a) and {h) Avas found to be 

 very high, +0.86, proving the correctness of the theory that pack ice 

 is a vital factor in the dispersal of icebergs to the North Atlantic. "^"^ 



If the assumptions previously set forth, viz, that an abnormal 

 amount of nortlnvesterly winds during the winter on the American 

 coast south of Baffin Land results in a heavy iceberg season past 

 Newfoundland, then we should expect to find good correlation co- 

 efficients between (a) and (c) ; actually they are as follows: 



Between (a) and (c) — 



Several other natural factors Avere tested in the hopes that coeffi- 

 cients found in the case (a) and (c) might be further raised, wdth the 

 result that a correlation of —0.49 existed between (a) and (/). In 

 order to test this effect Avlien divorced from the Ivigtut-Belle Isle 

 influence, partial correlations were calculated as follows: 



rac.f=-0.30 and raf.c= -0.11 

 and the regression equation for bergs on a scale 0-10: 

 (a) = 0.33 (c) - 0.05 (/) + 4.8 



The foregoing equation indicates that the Ivigtut-Belle Isle gra- 

 dient has about six times as strong an influence upon the southward 

 distriliution of icebergs as have atmospheric pressure conditions at 

 Stykkisholm. Both factors, however, are real and Avere employed 

 in the following })roportions: 



G(e) + if) 



(c') 



""A correlation between numbers of lceberg.s and sun spots gave negligible results. 



