NATURE 



[May 8, 19 19 





cury column) in the course of a day, far exceeding 

 not only the lunar, but also the solar, diurnal 

 variation. 



Even after abstracting the latter periodic oscil- 

 lation from the hourly values, the elimination of 

 the irregular changes requires the use of a large 

 amount of observational material. Airy's dis- 

 cussion shows that even twenty years' data might 

 prove insufficient. The Greenwich records of 

 atmospheric pressure now extend over sixty years, 

 but this threefold enlargement of the available 

 material does not by itself ensure very much reduc- 

 tion in the accidental error affecting the deter- 

 mination. Hence, in attempting a new investiga- 

 tion, improvement was sought by excluding all 

 but relatively " quiet " days from its scope, on 

 the ground that the diminution in the number of 

 days used is outweighed in advantage by their 

 better quality for the purpose in hand. Rather 



lunar diurnal inequality of pressure to be deduced. 

 Wherever possible, simplifying devices were used 

 in computation, and the solar diurnal variation 

 was duly removed from the data to rid the results 

 of this important source of error. 



The accompanying figure (taken from the 

 Q.J. Roy. Met. Soc, vol, xliv., p. 271, 1918) 

 represents the mean lunar daily inequality of 

 atmospheric pressure which was finally obtained. 

 The unbroken curve, which is almost entirely 

 semidiurnal, as tidal theory would predict, is the 

 one deduced from the observations (the inner two 

 vertical lines mark out a complete lunar day, on 

 either side of which a small portion of the curve 

 is repeated) ; on harmonic analysis its semidiurnal 

 component proves to be 



o'oogo sin(2t+ 114°) mm. of mercury, 



represented in the figure by the broken curve. 



Fig. I.— The lunar semidiurnal tide in the atmosphere at Greenwich, as determined from the Greenwich Records of 

 Barometric Pressure, 1854-1917. 



less than one-third of the whole number of days 

 in the sixty-four-year period 1854-1917 were re- 

 tained, being those on which the range of pressure 

 did not exceed o'l in. The hourly values conse- 

 quently totalled about 160,000, as in Airy's work. 

 There are approximately twenty-five solar hours 

 in a lunar day, so that the twenty-four-hourly 

 values on each " quiet " solar day were sup- 

 plemented by the last hourly value on the pre- 

 ceding day. Each such series of twenty-five 

 observations was broken into two parts, pre- 

 ceding and following the lunar transit on the day 

 in question. The preceding portion was trans- 

 posed so as to succeed the other, in order that the 

 rearranged series might correspond with intervals 

 of, in the average, oj, i^, 2^, . . . 24^ solar hours 

 after lunar transit. These series were written in 

 rows, and the numbers in each hourly column 

 were then added up, so as to enable the mean 

 NO. 2584, VOL. 103] 



The whole range of this, in inches, is 0*0007 1, 

 appreciably less than one-thousandth of an inch 

 (indicated in the diagram by AA'). The original 

 observations were made to o"ooi in. of mercury 

 (from a photographic record giving a fourfold 

 magnification) ; in the computations, however, the 

 last figure in each hourly value was omitted (the 

 previous digit being raised when necessary), the 

 entries on the lunar sheets being made to o"oi in. 

 only. In the circumstanQes, considering the (rela- 

 tively) large irregular changes of pressure, even 

 on these "quiet" days, it is somewhat remark- 

 able that so small a variation can be detected so 

 clearly. 



In such an investigation it is needful to guard 

 against obtaining a fictitious result which merely 

 happens to be of semidiurnal type. This point may 

 be tested by subdividing the data and examining 

 the internal agreement of the results from the 



