36 A. TANAKADATE. 



The 8ame data give in this ease, JX and J(f hailing the same 

 meaning as above expressed in degrees, 



0= .5° 02.'47-8:093J;. + 16:622J^-0.'84ÖJ;:- 

 ±.59 ±.271 ±.2(i5 +.033 



For the co-ordinates j; = ±8° and J(p = ±S°, the probable error 

 becomes ±3/7 instead of ±15/5, although the mean probable 

 error of a single observation is greater, as it should, namely 

 ±6/53 against ±6. '46 of the previous result. 



With this value of o, its annual variation takes the form 



0= 5° 02(5 -8.'09 J/1 +i6.'62Jçc-0:840J^- ut 1895.12 

 ô= 4° .54.'4-7.'9SJ;.+ 14.'46Jç)-0:832J/r at 1887.61 



0= 8:i-0.'llJ/î+ 2.'16J^-0: OU/- fur 7.51 years. 



âl.= im-0m5Jk+ 0:288J^-0:0013J;r per annum 



which is preferable to the provisional fornuüa given in p. 23 

 al)ove in estimating mean declination for few years following the 

 epoch. 



The values of the magnetic elements calculated from these 

 fornmhe for every round number of degrees of longitude and 

 latitude in the country are given in Tables VII to XIV, and 

 in Tables XV and XVI those calculated for each ol' the 

 stations together with the observed values and their differences 

 are given. 



The curves in 1)luo colour in Ma])s 1, 2, and 3 are traced 

 i'runi these etpiations and can be (.-ompared with the isomagnetics 

 drawn 1)y tentative method by laying those sheets closely under 

 them. In Maps 4, 5, 6 and 7 the total force and its rectangular 

 components are drawn, the differences of the observed and cal- 



