AT WINTER QUARTERS. 159 



visible variations in the course of the 2 1 hours. The marks were put on by an assistant using a millimetre 

 scale and a fine pencil. If measurement showed the result of his first attempt to be unsatisfactory, the 

 mark was rubbed out and a fresh one put on. Great care was used, but, of course, absolute accuracy 

 cannot be claimed. The different elements were dealt with at different times, and to some extent by 

 different people, and the time marks put on the three base lines were seldom //.W/r/;/ in a straight line. 

 There are several reasons for this quite apart from human liability to err. The dots of light were not all 

 the same size, and none, of course, were mathematical points. They might not be all absolutely in a line, 

 and the paper was doubtless at times not absolutely even on the cylinder, while the stretching or shrinking 

 of the paper might not be the same at all parts of the sheet. Apart from the hour marks, measurements 

 of time suffered from the fact that the curve and the corresponding base line were often very wide apart. 

 If a measuring scale is imperfect, or if the observer does not get its base line to be absolutely colinear 

 with the curve base line, an error comes into the measurement of the time which is proportional to the 

 length of the ordinate. Uncertainties on this ground were very much greater for the Antarctic curves 

 than for those at most of the co-operating observatories. The reason for dwelling on these points is that 

 any uncertainty in the time increases the difficulty of being absolutely sure of the identification of corre- 

 sponding rapid movements at different stations. As we shall presently see, there was sometimes a double 

 to-and-fro movement in the Antarctic when only a single movement was clearly visible elsewhere, and a 

 2- or 3-minute mistake in the time might suffice to lead to a mistaken identification. All I can say is that 

 in attempting these identifications I had regard to the original hour breaks and all the information 

 available as to both stopping and starting. For the intercomparison of disturbances, if that had been the 

 only object, it would have been more convenient to have used Greenwich time, as in the case of Falmouth, 

 Colaba, Mauritius, and Christchurch. But to show Greenwich time would have entailed putting on a 

 second set of hour marks, because those originally put on gave the local time which was employed in all 

 the hourly measurements and tabulation tables, except those for the international term days and hours. 

 Thus the times shown in all the Antarctic disturbed curves now to be discussed are local time (termed 

 L.T.). The corresponding Greenwich times maybe obtained by subtracting llh. 7m. In practice it is 

 simpler to add 53 minutes, converting p.m. into a.m., and a.m. into p.m. of the previous day. 



The original Antarctic curves were placed in contact with a photographic sheet, and a negative was 

 obtained by transmitted light. Writing made on the original, and unnecessary details as to times, &c., 

 were removed from the negative. The positive from this was intensified, the faint parts being carefully 

 inked in, and the hour gaps in the base lines filled in. The plates are reproductions from these positives. 

 The reproduced curves cannot claim to be absolutely identical with the originals, but the differences 

 between them are much less than if the originals had been tra'ced in the same way as the term-hour curves 

 were. Exact photographic copies of the originals were hardly feasible, because during the very rapid 

 movements characteristic of the larger disturbances the trace was naturally faint. In fact, it was not 

 infrequently so faint that it had to be traced over in the original to be sufficiently clear to be measured. 

 These bracings were mainly done by Mr. BERNACCHI, and to distinguish the Declination and Horizontal 

 Force traces at times he had been obliged to use coloured inks. 



The base line values arid the scales are shown at the margins, the information for one of the elements 

 being on one margin, that for the two other elements on the other. To assist in distinguishing between 

 the different curves all the traces stop just short of a scale-value line, except that of the base line for the 

 element concerned. Arrows are added to mark the direction in which each element increases (V up the 

 sheet, D and II down the sheet), though that can be inferred at once from the scale-value figures. The 

 short intervals on the D scale represent 10', only the 30' intervals are numbered. On the H scale lOy 

 intervals are shown, the 50y intervals being numbered. On the V scale lOOy intervals are shown, but 

 usually only the 200y intervals are numbered; the procedure varies according to the openness of the 

 scale. 



64. Plate XXII gives the Antarctic record of the disturbance of May 8, 1902, corresponding to the 

 records at the co-operating stations shown in Plate XIV. As elsewhere, there is an unusual definiteness 

 about the end as well as the beginning of this storm, it being followed as well as preceded by an 

 unusually quiet time. As already mentioned, it lasted almost exactly 8 hours. At the co-operating 



