The regression line relating iceberg count as dependent variable 

 to December-March Belle Isle/Ivigtut pressure difference as independent 

 variable is shown in Figure 2 for the three periods selected. There 

 is no substantial difference in the slope of the regression line for 

 the three periods, however it is observed that Smith in recompiling 

 Mecking's data possibly introduced bias through considering the 

 December-March Belle Isle/Ivigtut pressure difference to some extent 

 which cannot now be exactly reconstructed. Assuming that the 1946- 

 1969 data are most correct, then the 1880-1903 relationship is high 

 by about 66 icebergs or 19%, and the 1919-1942 relationship is high 

 by about 42 icebergs or 11%. It is concluded that, in terms of 

 average values over about twenty years, differences in iceberg count 

 of more than those shown in Table 2 are probably significant. 



Frequency Distribution of Iceberg Counts 



Smith (1931) states that iceberg count data fail to follow the 

 requirement that a normally distributed variable show the greatest 

 number of deviations proportionately near the mean. In order to 

 correct this he arranged the data in conformity with the shape of 

 a normal probability curve, and from this constructed a transform 

 curve to convert from actual iceberg counts to normally distributed 

 counts on a scale of 0-10. Smith (1931) and Schell (1952, 1962) 

 made use of normally distributed iceberg data on a scale of 0-10 

 in deriving their iceberg prediction formulas. Figure 3 shows the 

 frequency distribution of iceberg count data. The distribution is 

 clearly not normal. Low count data are most frequent, with frequency 

 falling off rapidly with increasing count. There are indications of 

 a second frequency mode at 200-400. About 30% of the years since 

 1880 have seen less than 100 icebergs even though the average count 

 since 1880 is 337 icebergs. In about 50% of the years since 1946 

 the count has been less than 100 even though the average since 

 then is 222. Thus a normal distribution centered on the average 

 fails to emphasize the fact that iceberg intensity is most commonly 

 very light. 



17 



