ARCT0W8EI, CHANGES IN DISTRIBUTION OP TEMPERATURE 45 



temperature in the United States, I took the tables of Frank H. Bigelow* 

 and, to my great astonishment, found that for some stations the depar- 

 tures of the annual means from the normal values are misleading. 



The yearly mean temperatures of Chicago and Milwaukee, for exam- 

 ple, illustrate my assertion. On Bigelow's tables the means of the first 

 decade of years are 49°.5 and 45°. 7 F. For the years 1873-1882 we 

 have, therefore, a mean difference of 3°. 8. The means of the last decade 

 (1896-1905) are 48°. 3 and 46°. 1, figures which give a difference of only 

 2°. 2. Considering decades of years, the increase is 0°.4 in Milwaukee, 

 whereas there is a decrease of 1°.2 in Chicago. These stations are too 

 close to one another to admit such a disagreement of figures, and it is 

 evident that something is wrong in the records of either Chicago or Mil- 

 waukee. Of course, a difference of 1°.6 is not a very large figure and this 

 difference may be due entirely to the variations of the town influences. 



The records of Port Huron, Detroit and Toledo may serve as another 

 example of misleading discrepancy. Considering again ten yearly means, 

 the differences between the means of 1876-1885 and 1886-1895 are 0°.l 

 for Port Huron, 1°.4 for Detroit and 2°. 4 for Toledo. It seems highly 

 improbable that a decrease of temperature of 2°. 4 could occur at Toledo 

 simultaneously with a decrease of only 0°.l at Port Huron. Now, com- 

 paring the departures of overlapping ten yearly means, departures from 

 the general means or normals, we notice that one or two of these series 

 of observations must certainly be considered non-homogeneous. 



Likewise, the departures of Knoxville compared with those of Cincin- 

 nati, Memphis and Augusta show plainly that the records of Knoxville 

 are unsatisfactory. 



These examples demonstrate how cautious one has to be in dealing with 

 long-range variations of temperature. The changes of the mean tem- 

 peratures due to climatic variations of long duration are small and an 

 apparently insignificant cause of error may modify the values of a series 

 of observations to such an extent that the actual variation will be com- 

 pletely disguised. It is therefore easy to understand that even the best 

 available figures — such as those of Bigelow's tables, for example — lead 

 only to a sort of rough approximation. 



T will pass now to the exposition and discussion of the results of my 

 calculations. 



On Bigelow's tables, there are fifty stations having continuous records 

 from 1873 to 1905, but only five belong to the plateaux and Pacific coast, 

 namely : Cheyenne, Denver, Portland, San Francisco and San Diego. In 



" Report on the temperatnres and vapor tensions of the United States. U. S. Dept. 

 of Agric. Weather Burea\i, Riill. S. 1000. 



