578 Proceedings of Societies . [August, 
taken the superintendence of tide-registration around the 
whole sea-board of India and at the port of Aden, and 
also the reduction of the observations by the method of 
harmonic analysis, with the view to the prediction of tides 
for the whole of the ports, it became a matter of necessity, 
in order to save the large outlay which the numerical operation 
of their prediction would have involved, that an instrument 
should be constructed to delineate the predictions. Accordingly, 
on the recommendation of the Surveyor-General of India, Mr. 
Roberts was desired to design, and to undertake the con- 
struction of, an instrument to include such a number of tide- 
components sufficient to predict the Indian Ocean tides with all 
the accuracy necessary for practical purposes. The present 
machine is the outcome of the recommendation. The instru- 
ment combines the following twenty tide-components : — The 
mean lunar semidiurnal, the first and second overtides of the 
mean lunar semidiurnal, two elliptic lunar semidiurnal, two 
evectional lunar semidiurnal, one variational lunar semidiurnal, 
the mean solar diurnal, the mean solar semidiurnal, the luni- 
solar-semidiurnal, the lunisolar diurnal, the lunar diurnal, the 
solar diurnal, one lunisolar elliptic diurnal, one lunar elliptic 
diurnal, one compound (Helmholtz) lunisolar semidiurnal, one 
compound (Helmholtz) lunisolar quarter-diurnal, the solar 
annual, and the solar semiannual. The chief difficulty in the 
construction of the machine is the finding, within reasonable 
limits, of proportions which shall represent with sufficient 
accuracy the periods of the several components, in order that 
the machine may be used for a considerable period of pre- 
diction — say, for twelve months. Very great success has been 
attained in this respect in the present instrument. For 
instance, the error of the period of the chief component (the 
mean lunar semidiurnal) relatively to the mean solar semidiurnal 
is inappreciable during a whole year’s predictions, amounting to 
about o*io° only in a period of fifty years. The largest deviation 
from strict accuracy is 0*37°, after a run representing twelve 
months. This is, however, of one of the very small compo- 
nents, and insensible in its results. This part of the design may 
be therefore regarded as practically perfect. The setting of the 
machine for the prediction of the tide-curves of any part for 
which the tide-components are known is as follows: — The 
dials are first turned so that the epoch or time of maximum is 
exactly under or above the highest or lowest point according as 
the component is situated on the upper or lower row of com- 
ponents. The cranks are set vertically (the slotted cone of the 
wheel on the axis having been first released) and the guide- pin 
thrown out to its proper range to represent the half-amplitude of 
the component. The proper positions of the hands having been 
previously determined by calculation for the time of starting, 
the hands are set and the slotted cones tightened up. The re- 
