A STUDY OF.CORKKLATIONS AMONG TERRESTRIAL TEMPERATURES. 351 



of the world simultaneously, may be expected to reappear in the discussion and com- 

 parison of the monthly means. But it does not seem objectionable to allow this. We 

 have only to recall the fact in drawing conclusions from any systematic departures 

 that may be found. 



The monthly mean departures which have been selected for discussion are partly 

 those of Dove, and partly those specially collected for the present work. Among the 

 latter are included those subsequently given in connection Avith the ten-day means. 



§11. Dismission of Dove's Dejxirtures. 



In the Memoirs of the Berlin Academy for 1858 Dove gives a great number of 

 tables of observed temperatures at widely separated stations, which are in some points 

 similar in form to those required for the present work. Those best adapted to the 

 present purpose have therefore been used for material. These are found on pp. 364, 

 etc., of the Memoirs. A certain number of regions were selected from Dove's tables 

 so far apart that there seemed to be no possibility of a correlation of their monthly 

 temperatures, except from some cosmical cause. It was also necessary to prefer sta- 

 tions and regions where the temperature was least subject to rapid fluctuations, and 

 for reasons already mentioned, regions of low rather than of high latitude. The 

 regions thus selected were : 



Eastern Asia ; mean of Nagasaki and Pekin. 



Southern Europe ; mean of stations in southern Russia. 



United States; mean of several stations in the southern portion. 



Cape of Good Hope ; one station only. 



Hobartown ; one station only. 



Madras ; one station only. 



In taking the means no distinction of weight was made between the different 

 regions or stations. 



The mean deviations formed from Dove's tables were tabulated and summed 

 separately for each year. The observations at Hobartown terminated with September, 

 1818. 



The results of the summation of the squares of the deviations for the several years 

 ai'e shown in the following table. 



Dove's deviations are given in the Reaumer scale. For convenience these are 

 used without change in the table. 



