NO. 4 TEMPERATURE VARIATIONS IN THE NORTH ATLANTIC J 



air temperatures unaccompanied by the corresponding- water sur- 

 face temperatures, but such observations we have rejected. 



In spite of all this, it appears from our graphical representations 

 that the values found for the air temperature must on the whole be 

 fairly satisfactory in the eastern part of the region even for the 

 single 2° fields. However, in order to save space we have not given 

 curves of air temperature in the single 2° fields corresponding to 

 those of the surface temperatures given in figures 16-19. On the 

 other hand, we have given in figures 44 and 46 a summary of the 

 surface temperature minus the air temperature for the single fields. 

 These curves show such good corresponding agreement and such 

 completely concordant gradual variation from field to field that they 

 show both for the air temperature and for the water temperatures 

 that the real facts are on the whole determined. 



In order where possible to show a comparison of the values of 

 the variations which we have found in the North Atlantic Ocean 

 south of 50° north latitude with the temperature variations in the 

 northerly regions of this ocean, we have employed the monthly sur- 

 face temperatures published by the Danish Meteorological Institute 

 for the ocean north of 50° (see " Nautisk-Meteorologisk Aarbog " 

 Copenhagen Nautical IMeteorological Annual published by the 

 Danish Meteorological Institute). 



Along the Danish steamship routes north of Scotland to New 

 York, to Iceland and to Greenland, these charts give the mean semi- 

 monthly temperatures for each single degree field for the interval 

 1898 to 1910. The values correspond, one to the first half of each 

 month, and the other to the second half. For the years after 191 1 

 there are simply the mean temperatures for the whole month, but 

 there is given a small figure which shows on how many observa- 

 tions each of these values is founded. Unfortunately the number 

 of the observations in each month for each of these fields is very 

 small. This holds particularly for the months February-April, 

 which we have investigated, when the number of the observations 

 for each field is very often only from one to four or five. The 

 temperatures for the single fields cannot therefore be regarded as 

 of high accuracy. 



In order to reduce the accidental errors as much as possible, we 

 have combined two by two the 1° fields together, so as to make 

 fields of 2° in longitude and 1° in latitude. With the fields thus 

 obtained we have the monthly mean temperatures for the interval 

 from 1898 to 1 9 10, including the month of February as well as 



