74 INSTRUMENTS AND RECORDS. 



reducing the curves, are seldom a source of real uncertainty, provided they occur when the instrument is 

 running. It is, unfortunately, otherwise with changes that take place when the instrument is not in 

 action. At an ordinary station, it is true, supposing only a few minutes' trace lost whilst changing the 

 sheets, an artificial shift of a trace can usually be detected at a glance, and at least a very approximate 

 estimate be obtained of its amount. But in the Antarctic the conditions were not favourable for detecting 

 artificial changes. The lamp had to be filled and trimmed, and there was not infrequently an interval of 

 over 20 minutes between successive days' records. At times a trace was off the sheet when papers were 

 changed. It will thus be readily understood that with elements perpetually altering an alteration of 

 1' per minute in the Declination was quite an ordinary one a very appreciable artificial shift in the trace 

 of an element might occur during the changing time without the traces themselves suggesting it. In the 

 case of the temperature trace and the Vertical Force, the observer on a good many occasions made a shift 

 intentionally when changing the papers. He would, for instance, find the temperature trace off the sheet 

 and bring it back. The extent of these changes was seldom much in doubt, because readings from a 

 mercury thermometer afforded a check on the temperature record, while the Vertical Force was seldom 

 much disturbed. 



3. When a change was detected or suspected, the conditions were examined into by myself, and when the 

 existence of a change was accepted, allowance was made for it after full consideration of the circumstances. 



In March, April, and September, 1902, and September, 1903, there were a good many interruptions in 

 the record. Subsequent to September, 1903, only a few records were taken, as the photographic paper 

 was nearly exhausted. At times, in very cold weather, the oil did not burn properly, and the trace 

 became gradually fainter and finally invisible. On one or two occasions, during a blizzard, snow got into 

 the magnetograph room and interrupted the record, and on one occasion a magnetograph suffered through 

 the fall of some ice which had formed on the ceiling. The only satisfactory way of treating discontinuities 

 in the base-line values that may occur on such occasions is by reference to the absolute observations. The 

 primary object, of course, of absolute observations at a magnetic station furnished with a magnetograph is 

 to determine base-line values. Under ordinary conditions, if one suspects a discontinuity, one simply 

 compares the values given by the absolute observations for the base line before and after the date of the 

 supposed discontinuity. In the Antarctic, unfortunately, with Declination changing half a degree or more 

 during the course of an absolute observation, the results of a single observation are of no very high 

 precision. To obtain information as reliable as that existing, for instance, at Kew where absolute 

 observations are taken once a week it would have been necessary to observe at least once a day. As will 

 be seen, however, on reference to Commander Chetwynd's discussion of the absolute observations, 

 "Physical Observations," p. 133, the number of absolute observations available was really small. 



4. After these remarks it is perhaps unnecessary to say that the uncertainties entering into the 

 absolute values of the elements as given in the various tables are much greater than would be the case at an 

 ordinary European station. They are probably greatest in the case of the Vertical Force, V. This is 

 derived from the Horizontal Force, H, and the Inclination, I, by the formula 



V= H tanl. 



The absolute value of V corresponding to an observed Inclination I is deduced by multiplying tanl by 

 the corresponding value of H, as obtained from measurement of the Horizontal Force curve at the time of 

 the dip observation. If v be the ordinate of the V curve in centimetres at this same time, and s the scale 

 value (i.e. the equivalent in C.G.S. measure of 1 cm. of curve ordinate), then the base-line value VQ is given by 



Vo = H tanl - ra. 



H was a quantity not far from '065, while I averaged about 84|. Thus an error of 1' in I means an 

 error of some 200y (ly EE O'OOOOl C.G.S.) in the value of H tanl. Accuracy to 0' - 5 in the dip derived 

 from observations with two needles is considered good under the most favourable conditions, at places 

 where the dip is from 60 to 70. What it is reasonable to expect from a dip circle in the Antarctic it is 

 impossible to say on our present knowledge. There can, however, be little doubt that even if we neglect 

 the uncertainties in the values of H and the uncertainty in an absolute observation was probably nearer 

 1/500 than 1/1000 of H the uncertainty in the base-line values of the V curves was too large to admit of 

 discontinuities of the order of, lOOy being detected by reference solely to the base-line values. 



