48 



the air and water temperatures in the far north, or variations in 

 precipitation, or perhaps an unnatural phenomenon, such as an ice 

 jam in the Arctic Archipelago. 



Although the investigation is not yet completed at the present 

 writing, the results already indicate a high degree of success for such a 

 method of ice forecasting. Correlation coefficients have been calcu- 

 lated between the following variables: 



(a) Number of bergs (on a scale of to 10). (See fig. 23, p. 76.) 



(6) Amount of field ice (on a scale of to 10). 



(c) Pressure difference (in millibars) between Belle Isle and 

 Ivigtut, combined with a deviation of pressure from normal at 

 Stykkisholm during the period December to March. The mean 

 pressure difference is calculated from the combination: 2XDec. + 

 2X Jan. + 1 xFeb. + l xMarch and this mean is combined with the 

 pressure deviation at Stykkisholm in the proportion of 6 to 1. 



{d) The pressure difference between Stykkisholm and Bergen 

 during the period October to January, inclusive, December being 

 given double weight. 



The correlation coefficients employed in the preparation of the 

 forecast were as follows : 



Between (a) and (6) +0. 85 



Between (a) and (c) — 0. 58 



Between (a) and {d) — 0. 63 



At the end of March a forecast of the number of bergs can be 

 prepared by means of the regression equation: 



Bergs = 4.8 -0.08 (c)-0.12 (d) 

 At the end of the field ice season, April 15, the number of bergs. May 

 to July, can be predicted very closely by making use of the high 

 correlation between field ice and bergs. 



Arrangements have been made with the United States Weather 

 Bureau whereby that organization furnishes the ice patrol with the 

 pressure data for the months October to March, inclusive, and upon 

 which is based the forecast of bergs for the following spring season. 

 The forecast for the ice season of 1926 was "a light ice year," (3.4 

 on scale 0-10), while as a matter of record it developed that we 

 experienced very closely to ''a normal season 4.3." It is fair to add 

 that we were handicapped in making a forecast due to the absence 

 of pressure data from a very critical area, that of Greenland. This 

 difficulty will probably not arise again as Greenland meteorological 

 stations are no,w connected with Europe by means of radio. 



