A STUDY OF CORRELATIONS AMONG TERRESTRIAL TEMPERATURES. 363 



CHAPTER V. 

 Study of Ten-day Terms. 

 § 13. Stations and Material Used. 



The term of ten days was chosen because it has been extensively adopted, espe- 

 cially in the Dehadenherichte of the German Seevarte. Mean temperatures for this pur- 

 pose being available in a number of cases, the labor of forming them for the entire 

 work was not necessary. A term of one fourth or one fifth the sun's rotation would 

 have been better adapted to bringing out fluctuations having the period of that rota- 

 tion ; but a lesser period than ten days would Ije subject to the drawback that small 

 fluctuations in the radiation require time to produce their full etfect upon the tem- 

 perature, so that little indication of their effect could be expected. 



Strictly speaking, the period is not ten days but one third of a month. When it 

 was necessary to form independent mean temperatures from daily records, the year 

 was divided into thirty-six parts as nearly equal as possible. There were, therefore, 

 thirty or thirty-one periods of ten days each, and five or six of eleven days in each 

 year. But when the ten-day means had been taken on a different system, the month 

 for example being divided into three parts, I adopted these means without modifica- 

 tion, deeming slight defects in coincidence not sufficiently important to be taken 

 account of 



The period chosen for the research commenced with the year 1872, because 

 although observations of the United States Weather Bureau date from 1871, when 

 they were commenced by the Army Signal Service, the data for that year were insuf- 

 ficient. This consideration was paramount in preparing the work because, in first 

 planning the work, it was not intended to include any stations but those for which 

 uniform records were readily obtainable. It was also intended to include as many 

 regions as possible in the investigation, but the circumstances mentioned in § 6 led to 

 the omission of several regions which might have been included had the data been 

 available. It was also believed that definitive results would be obtained by confining 

 the discussion to those regions where the data were easily accessible and undoubted. 



The regions and stations finally chosen were as follows : 



1. The United States East of the Rocky Mountains, Called U. 8. I. — In order to 

 lessen the effect of accidental fiuctuations at a single point several stations as widely 

 separated as possible are preferable. Guided by the consideration that stations near 

 the tropics were to be preferred, the four finally chosen for this region were Washing- 

 ton, Key West, Galveston and Saint Louis. 



2. The United States West of the Rocky ^fountains, or U. S. II. — The best station in 

 this region was San Diego owing not only to its southern position, but to its compara- 



I 



