614 H. MOHN. METEOROLOGY. [norw. pol. exp. 



visible, and also the figures of the thermometer-scale, but not the division- 

 marks. By estimating the distance between the figures, the degrees could 

 nevertheless be read off fairly well, but not the tenths of degrees. I do not 

 therefore consider it advisable to pay too much attention to the temperature- 

 observations during the darkest time, December and January, when the moon 

 was not above the horizon; but as soon as she appeared, the thermometer- 

 scale could be seen better, and with a full moon it could be read off very 

 accurately. We had the moon above the horizon during 24 hours 



in 1895 from October 28. to November 12. Full Moon Nov. 2. 



- — — November 24. - December 10. — — Dec. 2. 



- — — December 22. - 1896, January 6. — — - 31. 



- 1896 — January 18. - February 2. — — Jan. 30. 



- — — February 15. - March 1. — — Feb. 28. 



On February S'^^, 1896, I discovered that the column of Metaxylol C had 

 become broken, and the upper portion indicated — 5'°8, while the thermometer 

 indicated an air-temperature of — 17° after the separated portions of the 

 column had been united (by heating the bulb with the hand and shaking). 

 I write in my diary, that this breaking of the column had probably occurred 

 on January S*'^, 1896, when the wind had carried away the sledge, to which 

 the thermometer was attached, for some distance, and the breaking of the 

 column might have been caused then by the probable violent shaking of the 

 instrument. It seems somewhat difficult to understand how it could have 

 happened later, as the thermometer was then the whole time safely suspended. 



When I discovered the accident on February 3'^'^ there was an interval 

 equal to ll-°2 between the top of the broken off metaxylol and that of the 

 rest of the column, and this interval appeared to remain unaltered during the 

 rise or fall of the column. If this be correct, all readings between January 

 8^^ and February S'^^, 1896, ought consequently to be reduced by a correction 

 of — li-°2; but it is hardly probable that the error has remained the same 

 throughout the time, even if it actually occurred on January 8*'^, which is 

 also doubtful. It is thus to be understood that the temperatures given for 

 this period are not trustworthy." (Fridtjof Nansen). 



The error in the thermometer Metaxylol C found by Prof. Nansen on 

 the 3'^'^ February, 1896, can hardly have been caused by the accident on 



