NO. 4 TEMPERATURE VARIATIONS IN THE NORTH ATLANTIC 3 



west longitude and from thirty-seven degrees to forty-five degrees 

 north latitude (see fig. i and pi. 15). 



Herr Adolf Schroer has given us a statement on the first collec- 

 tion of observational material (March 15 to April 13). From this 

 statement we give the following figures, in which an air tem- 

 perature and the corresponding water temperature rank as one 

 observation : 



Year 1898 1899 1900 1901 1902 1903 



Observations 7S2 878 817 825 1174 868 



Year 1904 1905 1906 1907 1908 1909 1910 



Observations 1215 2229 2293 23S2 2167 2663 2122 



— altogether 20,415 observations. 



It is clear that the numbers of the observations made before 

 and after 1905 are quite different. The reason for this is that 

 prior to 1904 there were generally eight o'clock morning and even- 

 ing observations made, whereas after 1904 we used exclusively 

 journals in which the observations were made at the end of each 

 four-hour watch. 



The observational material at hand from the ships' log-books is 

 A^ery unequal. Formerly the observatory was satisfied with results 

 to 1° or 0.5° but later the results were demanded to 0.1° accuracy. 

 According to many reports, the observations from the meteorologi- 

 cal journals were made on numerous ships, not by the officers, but by 

 the seamen. That in these circumstances the estimation of tenths 

 of a degree did not add to the accuracy is hardly doubtful. The 

 thermometers employed were very unequal. In some journals there 

 are no indications as to the accuracy of the instruments employed. 

 In those thermometers for which corrections are given, we find 

 them mostly only at relatively great intervals, for exatnple at 0° 

 and at 20°. For many thermometers the corrections were altogether 

 too large, even in excess of 1°, hence one would draw the conclusion 

 that the material is quite untrustworthy. If one should reject all 

 the bad thermometers, however, he would have to throw away from 

 thirty to fifty per cent of the material. We have therefore only 

 rejected those journals for which thermometers were used which 

 gave between 0° and 20° corrections larger than 0.5°. For all 

 thermometers for which the reading was only to 1° or to 0.5° in 

 accuracy the small corrections were disregarded. Only for the 

 readings supposed to be of 0.1° accuracy were the corrections 

 employed. 



