19 



versa, and a temperature correction afterwards applied to the result. 

 The conversion of barometer readings from English to metric or from 

 metric to English units can only be made correctly after each read- 

 ing has been fully corrected for temperature. A further discussion 

 of this point will be found in the Monthly Weather Review for July, 

 1898, page 302. 



42. Barometer correetion cards. — Each barometer of the Weather 

 Bureau, when sent out, is accompanied by a correction card (Form 

 No. 1059-Metl.) showing the correction for instrumental error, and 

 also the corrections of the attached thenihometer. If these latter cor- 

 rections are as large as half a degree, which is, however, rarely the 

 case, they should be applied to the reading of the attached ther- 

 mometer hefore taking the correction for tem/perat'm'e from the table. 



43. Tables of teTn/peratwre corrections. — Tables of correction for 

 temperature are computed by simple formulae taking into account the 

 known coefficients of expansion of the mercury and of the metal or 

 material of which the scale is made. The scale in this sense includes 

 all the metal parts between the ivory point and the top of the column 

 of mercury. It is generally assumed that the temperatures of the 

 scale and mercury are the same, and that the temperature is given 

 by the indications of the attached thermometer. 



For barometers with brass scales the following formula is used for 

 computing corrections : 



^ ^. ^ , #-28.630 



Correction: C=—h- 



1.1123 # + 10,978 



in which A is the observed reading of the barometer in inches^ and t 

 is the temperature of the mercury and scale in degress Fahrenheit. 



The numerical factors in this equation are obtained by using the 

 following values for the expansion of mercury and brass, viz : 



Cubical expansion of mercury, 0.0001010 per degree Fahrenheit. 



Linear expansion of brass, 0.0000102 per degree Fahrenheit. 



In Section VIII are given full tables of corrections computed by 

 the above formula. 



44. Correction for density of mercv/ry. — If the density of the mer- 

 cury is not the same in two barometers that are exactly alike in every 

 other respect, the heights of the mercurial columns will not be the 

 same for the same pressure. In such a case a reduction to mercury of 

 a standard density will be required. The presence of 1 per cent of 

 lead with mercury causes a change in density that would require a 

 correction of about 0.051 of an inch. On the other hand, mercury 

 containing even so little as one one-hundredth of 1 per cent of lead is 

 rendered so exceedingly foul that it could not be used for barometric 

 purposes. It is therefore easily seen that a correction for standard 

 density is a refinement which need not ordinarily be considered. 



NORMAL BABOMEJTEB — STANDAED BABOMETEK 



45. It is easily understood, after what has been said above about 

 errors of graduation, errors due to capillarity, to imperfect vacuum, 

 to instrumental imperfection, etc., that even the best of ordinary 

 barometers is liable to be quite incorrect until corrections for these 



