364 A STUDY OF CORRELATIONS AMONG TERRESTRIAT. TEMPERATURES. 



tive steadiness of temperature. The peculiar climate of San Francisco seemed to ren- 

 der it inadvisable to adopt it as a station. The interior points of Salt I^ake City and 

 Phoenix, Arizona, were also selected and used as stations, although the observations 

 at each jDoint have suffered some interruption. 



H. Tlie Argent i7ic Republic. — The main sovu-ce for this region has been, as men- 

 tioned in Chapter II, the publications of the Officina Meteorologica Argentina. 

 The number of stations that could be used was different in different years, and fell off 

 to a single one in 1898. 



4. Samoa. — The Deutsche Uberseeische Meteorologische Beobachtungen contain 

 meteorological observations at a number of coast and island stations, but, for the most 

 part, the observations were not pursued continuously through a sufficient period to be 

 well adapted to the present work. The best station for our purpose proved to be Apia, 

 where the record is nearly complete since 1890. The unpublished results for this 

 station up to 1904 were courteously communicated by the director of the Deutsche 

 Seewarte at Hamburg. 



As no general j^rinciple is illustrated by the process of forming means and finding 

 deviations from them by simple subtraction, the writer conceives that the purpose of the 

 present work will be best subserved by omitting these merely routine details. If, as 

 he earnestly hopes, some authority fully equipped with the necessary computing 

 assistance shall in the interest of meteorology reconstruct the work in question, it can 

 now be more thoroughlj' done than the author has succeeded in doing. Data contin- 

 ually accumulate from year to year and the results of the present work will, it is 

 hoped, be found useful in any such reconstruction. As one of the special purposes 

 now in view is to show the method of determining correlations, that purpose will be 

 best subserved by excluding details not peculiar to the work itself Some remarks on 

 a few special points may however be made. 



After the means were taken for the regions U. S. I, it was found that the acci- 

 dental deviations at St. Louis were so much larger than at the other stations that 

 the means would be more accordant if this station were omitted entirely. Its weight 

 was therefore reduced to one third and new means taken. 



After the definitive means had been formed, it was found that the fluctuations of 

 temperature at Galveston, which were in general quite small, sometimes showed 

 abnormally negative values. When this anomaly was specially noted, and the 

 correctness of the record ascertained, it was too late to modify the work. The most 

 plausible explanation which I can assign for these anomalous temperatures is that 

 they are produced b}'^ the "northers" which are known to occasionally come down 

 from the Rock}' Mountain region into Texas, but which I did not suppose extended 



I 



I 



