February 2, 1922] 



NATURE 



137 



even apart from its greater economy in time, the direct- 

 reading machine has a still further advantage in that 

 it is possible to secure more accurate results than with 

 the use of the curve-tracing type. The dials on the 

 direct-reading tide predictor are graduated to single 

 minutes and to single tenths of a foot ; it is, therefore, 

 not difficult to estimate to within half a minute or 

 five hundredths of a foot. To estimate as closely 

 from a curve drawn to a moderate time-and-height 

 scale is well-nigh impossible. 



Since the tide predictor sums a number of con- 

 tinuous functions, it cannot be made to give the exact 

 results of an adding or a multiplying machine. And 

 because of the large number of moving parts that 

 enter into the construction of a tide predictor, we can- 

 not hope to secure the accuracy that may be obtained 

 with a planimeter. In fact, the errors of the tide 

 predictor may be ascribed almost wholly to the vary- 

 ing tensions on the numerous moving parts, but it 

 appears that these errors should not be large enough 

 to be serious. 



In 1910 the U.S. Coast and Geodetic Survey put 

 into operation a direct-reading tide predictor which 

 had been constructed in its instrument division. Prior 

 to the use of this machine for the prediction of tides 

 for the annual tide tables it was carefully tested, one 

 of the tests consisting in the comparison of the hourly 

 heights of the tide as given by the machine and as 

 computed analytically "by hand." The machine was 

 set with 30 components for Hong Kong, China, 

 January i, 1912, and run through to Decefnber 31. 

 For that day hourly heights of the .tide were read 

 off for the entire twenty-four hours. 



At the beginning of this year, after it had been in 

 use about twelve years, the machine was again set for 

 Hong Kong, January i, iqi2, and run through to 

 December 31, and the hourly heights read off. 

 December 31 was chosen so that all errors due to the 

 incommensurability of the speeds of the various com-, 

 ponents might accumulate. The table below shows 

 the differences between the computed and the pre- 

 dicted heights :— 



Hourly Height'; of Tide. Hong Kong, China, 

 December 31, 1912. 



Predicted in Differences 



predictor is graduated to tenths, so that the hun- 

 dredths were estimated. The differences between the 

 heights as predicted and as computed are relatively 

 small, in no case exceeding 006 ft. It is also in- 

 teresting to note the close agreement between the pre- 

 dictions made in 1910 and those in 1922, these pre- 

 dictions being made not only twelve years apart, but 

 also by different persons. It is only proper to add 

 that these predictions were undoubtedly carried out 

 with more than ordinary care, owing to the psycho- 

 logical effect of the knowledge that a test was involved. 

 Furthermore, it must be stated that with a single- 

 component tide the accuracy attained was not so 

 gratifying as in the example given above, the reason 

 being that with a number of components there are 

 compensating tensions of the moving parts, while 

 with a single component there are no such compensa- 

 tions. It is to be borne in mind, however, that in the 

 prediction of tides there is always a considerable 

 number of components involved. 



It would be unfortunate if the impression that tide 

 predictors are subject to serious errors gained cur- 

 rency, for, apart from the use of such machines for 

 the prediction of tides for the tide-tables, there are 

 numerous tidal problems involving time-consuming 

 computations which may be very easily made with the 

 tide predictor. In the elimination of the effects of 

 short-period tides on daily mean sea-level, in the com- 

 putation of the changes in sea-level due to tides of 

 long period, in the elimination from the observed tide 

 of the tide due to a number of constituents, and in 

 similar problems, the tide predictor should very 

 materially lessen the laborious computations involved. 



H. A. Marmer. 



U.S. Coast and Geodetic Survey, Washington, 

 D.C., January 5. 



The errors found in the British tide-predicting 

 machines referred to in the article were more serious 

 than those described by Mr. Marmer. The machines 

 were of the curve-tracing type, and therefore inevit- 

 ably less accurate (with the time and height scales 

 ordinarily used) than direct-reading machines ; they 

 had zero errors, and also diminished the apparent 

 range of the tide. The direct-reading machine used 

 by Mr. Marmer is both more accurate and quicker 

 in use ; for many purjwses, though perhaps not for 

 all, such a machine is a valuable substitute for 

 numerical computation. 



The Writer of the Article. 



The Oxidation of Ammonia. 



The following details of the early history of the 

 oxidation of ammonia, a process which became of 

 great importance during the war, do not appear 

 generally to be known, and may be of interest. The 

 first clear statement of the oxidation of ammonia 

 which I have seen is contained in a paper by the Rev. 

 Isaac Milner, B.D., F.R.S., president of Queens' Col- 

 lege, Cambridge, published in the Philosophical Trans- 

 actions for 1789 (vol. 79, pt. 2, pp. 300-13), and re- 

 publishe<nn Crell's Annalen (1795, pt. i, pp. 550-62). 

 The title of the English paper is " On the Production 

 of Nitrous Acid and Nitrous Air," and the German 

 paper is a translation, in which "nitrous acid" is 

 rendered " Saltpetersaure " and "nitrous air" (i.e. 

 nitric oxide, NO) " Saltpeterluft." 



Milner remarks that, although the relation between 

 nitrous acid and the volatile alkali was known, there 

 was no known case in which the latter was used in 

 the production of nitrous acid or nitrous gas. In 



