328 A STUDY OF CORRELATIONS AMONG TERRESTRIAL TEMPERATURES. 



This, substituted in (16), gives 



A = 22,. ,,,j«..w,«,r, (25) 



With equal weights we have, from any one pair of stations, [n = 2) 



A = -IIW (2(5) 



It follows that if we put Aj_ j. the special value of A found for any pair of stations with- 

 out regard to- weights, the final vahie for use in (17) when the weights are taken 



account of is 



A = 2,,,»-,»vA,* (27) 



and we shall then have 



i'robable mean t, - = ^-y^-i ^^-^o (28) 



CHAPTER II. 

 Review of Data and Processes. 

 § 6. Choice and Combination of Material. 



From the preceding exposition of the general method applied it will be seen that, 

 since our result is based on systematic or accidental departures alone, and not on 

 absolute temperatures, our main requirement is long series of observations of tem- 

 perature, at widely separated points of the earth's surface, made and reduced on a 

 plan which should be uniform for each point, but might vary to anj' extent from one 

 point to another. A single observation of temperature on each day would suffice in 

 the long run, provided it were made at the same hour throughout. Of course a better 

 result is reached from a number of daily observations at given hours ; but this is less 

 essential than uniformity of system at each separate station. 



In planning the work it was hoped that the much-criticised labor spent in accu- 

 mulating meteorological observations might be found not so ill-directed as is sometimes 

 thought. Unvaried routine, even if unintelligent, in the method of making and pub- 

 lishing the observations would be an advantage in a case where errors and defects in 

 the instruments and methods are unimportant for the result, so long as they remained 

 unchanged. But, when the actual material was sought out and examined, disap- 

 pointment was nearly everywhere the result. Outside a few government establish- 

 ments suj^ported by civilized nations or other permanent organizations, diversity 

 instead of uniformity was found to prevail, — even unintelligent adherence to any 

 routine system of making, reducing, and publishing the observations being rare. The 

 amount of available material was also diminished by the fact that a very important 

 part of the best-planned meteorological observations are made only to determine the 



