32 



UNITED STATES COAST AND GEODETIC SURVEY. 



pointer of the height sum pulley and that of the platinum zero link 

 of the time chain in relation to their respective indices noted several 

 times during the prediction of the tides for the 1912 and 1913 tables. 

 No change in^the lengths of the chains could be noticed. 



Accuracy of the macJiine. — At first thought one would probably sup- 

 pose that the best test of the accuracy of a tide-predicting machine 

 would be a comparison of its results with actual observation at the 

 port for which it is set up. Such a comparison is always possible by 

 setting the machine for a preceding period during which automatic 

 tide-gauge readings were made. But the tides of nature are so much 

 affected by meteorological conditions, which can not be predicted, 

 that such a comparison must necessarily be more or less rough. 



After mature consideration it is found that the best test of the 

 accuracy of a tide-predicting machine as a piece of mechanism is to 

 compare its mechanical solution of the equations put into it with the 

 corresponding results obtained by computation from the same 

 elements. 



Such a comparison was made, using the regular predictions for 

 1912 and 1913 tide tables, at the station Aden, Arabia, with 34 com- 

 ponents, and at Hongkong, China, with 30 components, hourly 

 heights being read off the dial, by estimation to hundredths of feet, 

 upon a day at the end of the year for which the predictions were 

 being made. The greatest difference between the predicted and 

 computed 24-hourly heights was 0.02 foot in the case of the Aden 

 tide, and 0.06 foot in that of Hongkong. The results of the latter 

 test are shown in Table A- 



Table A. — Test for accuracy by comparison of 'predicted with computed heights- 

 heights for Hongkong, China, Dec. 31, 1912. 



-Hourly 



