METEOROLOGY AND ALLIED SUBJECTS. 247 



The computation, however, may be more simjily arranged if for each 

 day we make the combination : 



4. -jL(&4-2c+4d + 2c+/) 



and call the new means thus found for each successive day cii, &i, Ci, &c. 

 We have then to combine these new means according to the following 

 formula : 



5. fn=Hci + 2(7i + ei) 



The to, thus computed will be the same as given by equation No. 3, 

 and the whole process is reduced to a simple system of summing and 

 halving. {Z. 0. G. M., XIV., p. 380.) 



Pernet, who is now in charge of the International Bureau of Weights 

 and Measures, at Sevres, in a memoir on the determination of the fiducial 

 points of the mercurial normal thermometer.s, and tlie determination of 

 temperatures to the hundredth of a degree centigrade, says : Carefully 

 calibrated thermometers, bandied in various manners, do not agree with 

 each other even in the interval between freezing and boiling ijoint of 

 water. The differences, under some circumstances, amount to several 

 tenths of a degree, and are therefore much greater than the errors of ob- 

 servation. They depend in part upon the irregular expansion of glass, but 

 still more upon the fact that the bulb of the thermometer after being 

 warmed does not immediately return to its original volume, although it 

 may do so in the course of time, and in consequence of this a temporary 

 lowering of the freezing point is produced. Since this depression of the 

 freezing point appears thus far to be subject to no law, we are pre- 

 vented from attempting to make different thermometers agree among 

 themselves. 



This is very much to be lamented, for if it were possible for us to deduce 

 from the observations of various thermometers according to a general 

 method of computation temperatures agreeing with each other, then 

 would every doubt as to the correctness of the measurements of temper- 

 atures be removed. He gives the following methods for determining 

 the fiducial points and for the computations of the corrections depend- 

 ing thereon. 



First. Determination of the freezing point: Fresh fallen snow is the 

 best material; the snow must be thawing throughout its whole mass, 

 must be clean and pure. The next best substitute is fine shavings of 

 natural ice; artificial can only rarely be used. The thermometer and 

 scale must be covered up to above the freezing point with snow or ice. 

 The determination must take place in a cool room, and the ice must not 

 press upon the thin glass of the bulb, else otherwise a sensible error 

 will be produced. In general, the ice point is lower in the vertical posi- 

 tion of the thermometer than in the horizonal position; it is therefore 

 well to determine both ice points. 



Second. The determination of the boiling point: This requires the use 



