]\gt ; S35 IX .
Analysis Resnlts the on Quiet\ldquo ; the Years 1890 to 1900 , with a of Certain Phenomena in the bsolute Observations .
CHARLES CHREE , Sc. D. , .D .
, F. R.S. , endent of the Department of the Laboratory .
( From thoe Physical ) Received May May 28 , 1903 .
Sections 1 .
Introduction 335 2 .
and } ) frequency . . . . . .
336 3-9 .
Secular change . . . . .
337 10-12 .
Annual inequality . . . . .
344 13-17 .
Non-cyclic effect . . . . . .
346 18-98 .
diurnal inequality , tables and their discussion . . . . .
, 30 .
Curves representing diurnal inequality . . . . .
376 31-36 .
Representation of diurnal inequalities by Fourier serieb . . . . . .
380 37 , 38 .
Variation of ranges , \amp ; c. , throughout the year , treated by of Fotlrier . . . . .
394 39 , 40 .
" " Variability\ldquo ; of the declination . . . . .
397 41 .
Sun-spot relations .
Introduction . . . . .
399 42-44 .
Variation of declination and horizontal force ranges with suu-spot frequency . . . .
, . .
399 .
Curves tive of sun-spot relations . . . . .
402 48 , 49 .
Sum-spot relations apparent in Fourier coefficients . . . . .
406 50-60 .
formula numerical relationships . . . . .
415 61-70 .
Relations with meteorological enomena . . . . .
426 .
Nature of the tionship between sum-spots gnetism .
Speculations . . . . . . . . .
433 S1 .
IN 1889 the then Kew Committee agreed to take part in a scheme which that at the end of each year the Astronomel Royal should select five ( netically q days for each month with view to ) tabulation of their records .
has been in continuous operation since 1890 , and the esults from the ( 354 .
) selected quiet days have publisl ) amlually in the 'Report ' made to the Royal Society .
In 1895 I contl.ibuted to the British Association 'Report ' an analysis of the atio and horizontal force records for the years 1890 to 1894 , and described eviously unnoticed " " non-cyclic effects\ldquo ; which appear to be characteristic of netically quiet days at Kew and elsewhere .
In 1901 the development of electric fraction in the West of London rendered it clear that no further records at old be free from suspicion of al.tificial disturbances , and that it would be necessary in the neal future to make arrangements for continuin the magnetic work elsewhere .
Both contingencics pointed to the expediency of a complete of the data obtained from 1890 to 1900 , which I accordingly commenced .
The delay in the execution of the scheme is largely due to the occurrence of various unforeseen difficulties , which , however , I think at last fairly satisfactorily surmounted .
The object of the paper is not so much to put observations on ecord , as to make a critical use of them .
The data used are mainly derived from old Kew 'Reports , ' to which l1lust be made particulars .
S 2 .
One of the questions dealt with is the relation between sun-spots and terrestrial magnetism .
The sun-spot data made use of are from the very important table published by WOLFER in the ' Met .
Zeit .
' * WOLFER'S table gives the mean monthly values to the sun-spot frequency by and ER for a very long series of years .
As frequent reference has to be made to sun-spot data throughout the whole of this , it is convenient to eproduce at once that part of WOLFER'S table which applies to the pel.iod lS90 to 1900 now under review .
This will be found in I. In ddition to WOLFER'S owll figures I give the mean sun-spot frequency obtained for each month of the year when we combine the 11 years 1890 to 1900 .
It will be noticed that the four years lS92 to 1895 are conspicuously years of sunspot maximum , while 1890 , 1899 , and 1900 years of very few sun-spots .
The mean frequency for the 12 months commencin August , , and ending July , 1894 , is ryher than for any other 12 successive months in the 11-year period .
Considering that so many as 11 years are included , the irregularity in the mean values for the several months of the yeal is rather surprising .
March and November in particular show a lemarkably low mean frequency .
This is to be regretted , as it tain e of u1lcertainty into some of the calculations made in the latter part of the paper .
* May , 1902 , p. 193 , et seq. ON ' ' 1890 1900 , . .
84.9 S. .
Few people at all the difficulty cert inty as to the truc value of of .
elements ono year to next .
tion of the ished r of two at such ) and Parc ( Paris ) , .
eyen when so as Greenwich , oses i ulnrities not shared by the two tions .
In the present state of our , we that such ularities r be absolutely ) ntations of the tions of Nature , but to the belief ) they ften owe a good deal to instrumental or UHes .
) ) sration , changes of instrunents , of observerfi , or of of reduction , all to results , and of late ears electric trnlns have been ecking o ) sel.vatories , or uncertainties into their records , over the world .
In the present case , this ) worst cause of uncertainty may be left out of account , but some of the others unfortunately , to be dealt with .
Whenever it appeared neces , , I have refelTed to the original records , and have calculated so far as possible what the results would have been if the methods and instruments loyed had been the same as of late years .
To explain the situation , a brief accounlt is necessary of the present method of deducing results from the measurement of the curves .
To economise space , I shall throughout the rest of the paper distinguish the seyeral magnetic elements by letters , and employ these , when convenient , in the text , and tables .
The ) loyed and their meanings are as follows:\mdash ; CCTT .
declination , ] ortherly component of I cliuation , westerly \ldquo ; horizontal ftotal force ltant of .
of vertical force , ) .
S 4 .
Taking fii the case of , as simplest , we have the scale value\mdash ; that is the equivalent of 1 centim. of curve ordinate\mdash ; dependent only on the distance between the raphic paper and the mirror carried by the declination magnet .
There is no occasion to move this mirror except slightly in azimuth , as the magnet turns with secular change , and no appreciable change of scale value has occurred during the period dealt with .
In this case no temperature compensation is necessary , so far as is known ; and the scale value ed once for all , there remains only to determine the value of the declination which answers to the base ( or time ) line , from which ordinates are measured .
This is done by comparing the results of the absolute observations , taken hree or four times a month , with the corresponding curve ordinates ; each month is dealt with by itself With the exception of a few months early in 1890 , the same magnet has been used throughout for the absolute observations , and an elaborate intercomparison of it with magnet previously accepted as the standard , before the latter was laid aside , gave mean results as nearly as possible identical .
The magnetometer employed has been in use for about half a centu1y .
Certain additions were made to it in January , 1891 , but were discarded after a few weeks ' trial , and the results obtained with it during these weeks were rejected .
Coming next to , the scale value here is affected by change either in the magnetograph magnet itself or in its bifilar suspension .
Durin the whole eleyen years aim has been to keep the scale value at 1 centim. ( where C.G.S. unit ) .
The magnet is an old one , whose moment alters extremely slowly , and when the instrument is left to itself , the change of scale value taking place in twelve months seldom amounts to 1 per cent. As in the case of , the value of the base line is determined by comparing absolute values of , determined three or four times a month , with the corresponding curve ordinates , and each month is treated independently .
A temperature correction is applied , which allows for the difference between the temperature of the magnetograph room at the hour considered and at the times of the absolute determinations of in the month in question .
The temperature coefficient of the graph is very small , and the correction comparatively unimportant .
procedure in the case of and I is less direct .
Here the absolute obseryation gives I , while the curve gives V. The value of 1 centim. of ordinate is found as in case of by direct experiment , and the instrument is adjusted\mdash ; usually once a year\mdash ; so as to keep the sensitiveness as close as possible to 1 centim. .
Three or four absolute observations of I are made each month , usually on the days of the observations of H. observation is corrected to the time of the corresponding ON ' ' QUIET\ldquo ; DAYS DURING 1890 TO 1900 , ETC .
I tion , by allowing for any of cnrve ordinate and for any change of magnetograph room temperature that may have occurred meantime .
Combining the observed values of I and of , as corrected , one obtains an absolute value for V , and so can institute a comparison ith the curve and determine the value of its base line .
The hourly values of inclination are obtained by combining corresponding hourly values of and V as given by the curves .
same inclinometer has been in nse throughout the eleven years .
The needles originally employed were damaged in 1899 and had to be replaced .
The needles then introduced had been vion sl compared with the discarded pair , and no appreciab]e difference was detected .
S 5 .
The methods of reduction just described had not been strictly adhered to in some of the earlier years .
For instance , no satisfactory allowance had been made prior to 1894 for change in between the hours of the and I observations .
This had the effect of depressing the values of V and I by somethin like and 1f respectively .
Trouble also arose from an addition made to the building in 1892 , introducing various discontinuities into the curves , which had not been altogether successfully dealt with at the time .
Reference to the original records , however , enabled these defects to be removed fairly satisfactorily , and when this was done the more striking irregularities apparent in the secular changes of V and I between 1890 and 1894 mainly disappeared .
There still remained rather a prominent discontinuity in the values for 18 .
On into the matter , I found that in March , 1896 , an old weak pair of bar magnets used for stroking the dip needles had been replaced by a much stronger pair .
The secular change obtained by comparing the results for the six months immediately following the introduction of the new } magnets with the results for the corresponding six months of the previous ycar amounted to only 1 ' , whereas by coml ) aring results from the six montbs immediately preceding the change with the results from the corresponding months of previous year , and by comparing results from the six months immediately following the change with results from the corresponding months of the subsequent year , one got values for the secular change which were in good agreement and ayeraged .
This seemed to point pretty conclusively to a sudden discontinuity amounting to about , and simultaneous to all with t ] introduction of the new bar magnets .
For our present it is immaterial whether the of this discontinuity was the introduction of the new bar magnets , or lse occurring approximately at the same time , so long as it is admitted that it is to apply a correction of amount indicated bove .
One 's of tquy s correction must largely depend on one 's estimate of the knowledge and ) the person who decides to apply it .
It may added , however , evidence in its favour that it wiped out discontinuities in the 1896 alike of T. consequence of the several corrections introduced lvas to alter the mean annual values of the elements as published in the annual ' Reports ' to the following extent:\mdash ; 340 .
C. ULTS 1iAlH far be is ] ecessary f personal equation in the ob , servations .
1)uring the whole ) dealt with , the magnetic has been under the of the Chief .
T. W. , who has himself observed on the ( majority of occasions .
) 1891 to Mr. .
S. WHIPPLE etty frequently , and since the1l , during , the horizontal observations have ) mainly taken by myself , and the declination ation ovations by Mr. W. .
No certain diffel.euce has , however , manifested itself between the several observers .
S6 .
The mean annual values of the elements as corl.ected are given Table the corresponding secular changes appear in Table III .
The values of were calculated fiionl those of , and I by the formulae The action of the vertical-force during the last two months of 18 was not considered satisfactory at the time , and the esponding results were discarded .
Thus 1891 was the first year for which satisfactory mean annual values were obtainable for V and I. The mean values at the foot of Table II .
are thus calculated for 10 years only in the case of V , , and I , as ainst 1 years in the other cases .
These meau values are wanted later , they do not possess any immediate significance in connection with secular change .
In Table III .
I added mean values of the secular Parc St. Maur and Potsdam , for comparison with the mean changes at Kew .
The Parc St. Maur data are from publications by MOUREAUX .
As data for 1900 were not available the means in this case are for the period 1890 to 1899 for and , and the period 1891 to 1899 for I and V. The Potsdam data are from a paper ESCHENHAG they efer to exactly the same periods as the Kew data .
'Ann .
Bnroan ) ' 1897 , \ldquo ; .
B. C5 .
'Ann .
der hysik , ' vol. ( i ) , 1901 , p. 424 .
" " \ldquo ; DAYS DUISING THE YEAliS 1890 TO 1900 , ETC .
II.\mdash ; Mean Annual Values .
TABLh .
\mdash ; Secular Changes .
Potsdam .
PKaerwc S .
It will be the es in at and Potsdam are similar to those at Kew .
There is even , in the case of these elements , a fair resemblance in the variations from to year .
All three stations , instance , show a decidedly ckened rate of change in since 1896 .
In 1898 , the changes in at absolutely identical with those at Kew ; but the change at Potsdam was less by 1 ' than in either hereas at Ke it 11 a , s reater in than in the two adjacent years .
and in the change in during as small , and the ] in 1893-4 and also ( at 3.42 DR. C. OF SULTS FROM THE MAGNETOGBAPHS decidedly below the average .
Again , we have specially large of at Kew and Potsdam in , 1895-6 , 1896-7 , and 1899-1900 .
Parc St. Maur grees with and Kew in having a specially small change of in 1893-4 , and specially large changes in 1892--3 and 1896-7 ; but it differf ; in showing an average in 1891-2 and a less than average change in In inclination and vertical force the mean data at the three stations are widely different ; Parc St. Maur shows a nearly stationary ] of V , as against a distinctly value at Potsdam , and a still more distinctly falling at Kew .
The inclination is diminishing at all three places , but the rate of fall appears considerably less at Potsdam than at Parc St. Maur , and considerably less at Parc St. Maur than at Kew .
Speaking generally , the secular changes in I , V , and appear to have been much more irregular at Potsdam and Parc St. Maul han at Kew .
Thus , at Parc St. Maur V rose in 1891-2 and fell 4 in 1893-4 ; while at Potsdam V increased by in 1891-2 , in 1892-3 , and in 1899-1900 , and fell in each of the three years 1893-4 , 1894-5 , and 1898-9 .
There are corresponding fluctuations , of course , in T. At Kew the accuracy of the estimate of secular change in V and is really dependent in the main on the accuracy of the determination of I. An error of 1 ' in I\mdash ; and such an error might arise from very slight deterioration of either circle or needles\mdash ; means an error of over in both V and T. Taking this into account , I do not think much weight can be assigned to the values of the secular change for individual years .
The comparative smoothness of the V and data in Table III .
is probably not more a tribute to the uniformity of Nature than to care and skill of Mr. as an observer .
The theory has been advanced that secular changes are particularly large in of sun-spot maximum .
The data in Table JII .
can hardly be said to support this .
The changes of were certainly larger in 1892-6 when sun-spots were numerous , than in ] 898-1900 when they were few ; but the largest change in occurred in 1890-91 near sun-spot minimum , and in 1893-4 about sun-spot maximum the secular change was about average in and particularly small in and I. S 8 .
To assist in the comprehension of the secular change I have drawn a curve , fig. 1 , which shows the change from year to year in the direction of an inary magnet , freely suspended from its C.G. so as to point along the lin of total force .
The curve may be supposed drawn by a style attached to the dipping end of the magnet , the paper being supposed to lie on a tangential to a sphere , concentric with the magnet , at the point answering to the mean position of the dipping end during the period lS91 to 1900 .
With increasing I , the dipping end would moye towards an. .
observer who stood with his back to the south , looking at the paper .
The west , i.e. , the direction of , would then be on the observer 's left .
The crosses mark the positions of the end of the magnet at the middle of the years specified .
The magnet or style is supposed to be so long that the difference etween ' ' \ldquo ; 1890 TO 1900 , arc and tangent may be ected , so that equal lengths on the paper represent equal of direction in the magnet .
The curve , it will be noticed , does not depart very far from a straight line .
S 9 .
The above graphical method no direct light on absolute nitudes of the of the components of force It is , , of some interest to know Fig. 1 .
the direction and amplitude of the disturbing force to which the secular change may be attributed .
To find this we may proceed as follows:\mdash ; Let represent the values of the northerly , westerly , and vertical components of force at the of a period , at the end of which the values have become .
The total force meantime has changed from to SN , SW , SV represent the components of a force SR ( totally different usually from ST ) which may be regarded as the secular change .
If secular change data for individual years were sufficiently reliable , one would naturally take the SR answering to a single year 's change as measuring in direction and magnitude the force to which the secular change may be attributed .
In general , however , it will be best to calculate a SR for a considerable number of years interval , and to regard its line of action as the general direction of the secular change force throughout period .
In instance we get for the period 1891 to 1900 , at Kew , S , S , S Thence find SR , ( its line of action makes with three rectangular axes drawn respectively to the north , the easf , and vertically ?
] the angles , and The projection of the line of ction of SR on the plane is inclined to the geographical meridian at the in a ierly direction .
S 10 .
If the secular change proceeded at a uniform rate throughout the year , the mean value of any element answering to the middle of the month of the year should be derivable from the mean value -E for the whole year by the formula where is the secular change for the entire year .
This neglects the differences between the lengths of months , as being for the present purpose immaterial .
Conversely , if one applies to correction , one eliminates the effcct of a regularly ) ressive secular , and )tains what is known as the ' ' Annual \ldquo ; In practice , complications arise from the apparent variabihty in the secular changc fro.m to year .
To illustrate the consequences such irregularity , take the simple hypothetical case where the secular change of declination proceeds uniformly from January 1 to December 31 of a year at the rate ] a year , and then proceeds for the next twelye months at the uniform rate of a year .
From the mean values for the two years we should deduce a secular , not of 12 ' or of 6 ' , but of 9 ' year , and if corrected the two years ' monthly values independently , on ) assumption of a secular change of 9 ' a year , we should deduce for each year a wholly fictitious annual inequality .
Combining the monthly values for the two years should in this case conclude , } ightly enough , that there was no true annual inequality , but that is merely an accident of the particular hypothesis .
The illustration will show how uncertain is the physical interpretation to be put on an apparent annual inequality in the case of an element whose secular change is irregular , unless deal with mean monthly values from a large number of years .
The results in Table IV .
have been obtained by combining 10 years ' results for I , , and , and 11 years ' results for the other ments .
The data assigned to the ' ' middle of the month\ldquo ; are the values actually obtained for the differences between the mean lnonthly values and the mean annual value after ) application of the ON ' ' QUIET\ldquo ; DAYS THE ELEVEN YLARS lS90 TO 1900 , corrective term .
The value assigned to in each case was that given in Table III .
as the mean for the 10 or 11-year period considered .
The results under the heading " " .
of month\ldquo ; are each the arithmetic mean of two adjacent mid-month values ; for example , the value assigned to January refers to January 1 , and is the mean of the values for mid-December and mid-January .
The mid-lmonth data are the most suitable for use in a critical examination into the true nature of the phenomena ; but , assuming the existence of a true annual inequality , the of the nlonth data are probably the more relinble measures of its amplitude .
In all the force components , unity represents IV.\mdash ; Annual Inequality .
Middle of month .
January February .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
January Febrtlary .
AprilMarch .
nebIay August September October ovemberDecember .
Beginning of month .
W. N. .
If we take the beginning of the month data in Table IV .
we have the following results:\mdash ; Ran ge of annual i. . .
Range secular change The ratio borne by the range of the annual inequality to the mean secular change is certainly largest in those elements in which the uncertainties are greatest .
In the case of and ( or of and N ) we are dealing with elements whose secular change has been comparatively regular , and there are other grounds for .
regarding the annual inequalities obtained for them as the most reliable .
They show a pretty close resemblance in general character to results obtained VOL. CCII .
\mdash ; A. 2 346 .
C. CHREE : ANALySIS OF RESULTS FROM THE KEW MAGNETOGRAPHS for Kew from the iive years 1890-4 .
At the same time , the ranges now found are somewhat conspicuously less than the ranges , in and in ined on occasion .
In case of and , the variation in the mid-monthly values in Table IV .
seems etlj e too ular to ascribed to chance , and even in I and V something more than chance seems involved .
It must be allowed , however , that in these last two elements the value for March stands out in a way one can hardly suppose to be trnly representative of Nature , and in the case of V , the tendency to negative values in latter of the year , and the apparent jump from December to January , are solnewhat tive of an nnder-estimate of the true seculalnange .
S12 .
One possibility to which attention should be drawn is that an annual equality may be no ascribable to errors of observation , or of estimation of secular , and yet be absolutely fictitious so far as terrestrial magnetism is .
For example , if the distant mark used in absolute observations of shifted in any way dependent on the temperature of the air , or on the temperature or moisture of the soil , an apparent annual inequality would ensue .
A similar result would happen in the case of if the law of variation with temperature in the moment of inertia of the collimator magnet were wrongly assumed .
In the present case I have appIied an elaborate check\mdash ; whose description would take me too far afield\mdash ; calculated to disclose any appreciable uncorrected effect of temperature on observations of , and the conclusion it led to was that if any such effect existed it was small compared to the found for the annual inequality .
It is also that in all the elements the beginning of the month data indicate a semi-annual term in the annual inequality , which is not what one would expect to find if true cause of the phenomena were moisture or temperature .
Non-cyclic Effect .
S 13 .
In virtue of the secular changes shown in Table 1II .
, the value of an element should exceed its value 24 hours earlier , on the average throughout the period now dealt with , by the following amounts:\mdash ; These quantities are all much less than can be directly measured on the magnetograms .
Such as they are , however , they contribute to what I have termed the " " non-cyclic effect 'British Association Report for 1895 , ' p. 226 .
'British Association Report for 1895 , ' p. 210 .
ON ' ' QUIET\ldquo ; DAYS DURING THE ELEVEN YEABS 1890 TO 1900 , Again , the annual inequality , when such exists , nplies an increment or decrement in the course of 24 hours , varying with the time of year ; and if Table IV .
may be trusted , this increment at certain seasons should be cousidel'ably larger than that due to the secular .
But even the most prominent monthly in the table , viz. , from to in the case of V , would epresent all increment of only about in a day .
Any increment , of course , due to annual inequality would in with the season , and Vould cut out for the year as a whole .
There are other known causes tending to introduce a diflerence between the values of the ordinates answering to the beginning and end of a day .
There are the irregular movements visible in most curves , especially at times of sun-spot maximum , and imperfections in the temperature COlTections applied , at least in the case of .
These two last mentioned sources would , however , introduce larities into the monthly values of the non-cyclic effect , without appreciably affecting the mean for year .
A more troublesome source of -cyclic effect is instrumental the case of there does not seem to be any uncertainty 011 this head .
This Jllight be inferred from the nature of the declination , but it is at least desirable to have the eleven dealt with : the magnet had been -adjusted only once , 1898 , when its azimuth was altered so as to increase the ordinate by the equivalent of about 36 ' of declination .
Allowing .
this , the curve diminished the equivalent of about 53 ' , between the taking of ) solute observe ations , ] 8 , and December , 1900 , which showed an actual decl.ease of 63 ' in D. This leaves only about 10 ' of shift unaccounted for by true secular change .
ovel 11 years , this would represent all exceedingly small non-cyclic effect .
the case of the magnetograph also , only one slight artificial alteration had been lllade the distance between the trace and the base line .
Allowing this , the curve ol.dinateiJlCl.eased by about , whilst really eased about .
The ) sents presumably the lllai decrease the lnoment of the bifilal ] nagnet .
ovel 11 years , this } ) esents a Jlleall nou-cyclic of abont } ) .
This , as we shall see , is only about a sixtieth of the average -cyclic eHect actually observed on quiet days .
If we combine the effects due to secular and instrumental causes , we have a mean Doll-cyclic ficct of only about } ) This is in the direction as the non-cyclic effect actually obsel.ved in , but is wholly insignificant .
Results confirmatory of our conclusion as to the smallness of the non-cyclic effect in , due to secular change and instrumental causes obtained from measurements of curve ordinates at the beginning and end of a series of individual years .
In the case of V , the magnet mirror had been altered seven or eight times in the course of 11 years , usually early in January , at the time of the scale determinations .
348 Dl C. THE KEW hlAGNETOGRAPHS As the balance had usually been altered at the same time , it was difficult to arrive at direct conclusions as to the true effect on the curve ordinates .
A comparison , howevel , of the curve ordinates after the scale minations in January with those in the following December , on seve years when no intermediate change had been made , showed a decrease in the ordinate due to instrumental causes at the average rate of a year .
The change was in the same direction each year .
An annual change of gives a mean non-cyclic effect of per day .
instrumental causes with true secular change , we get a non-cyclic effect of about day .
This represents , as we shall see , about 15 per cent. of the average non-cyclic effect actually observed on quiet days .
S 14 .
There is unavoidably some uncertainty in the estimates of artificial noll-cyclic , and I have attempted no correction to the results as observed .
The and nets would probably be more exposed to losses of mag1letism during magnetic than on ordinary occasions , and the chances are that any estimate we should make of the contribution from instrumental causes to the observed -cyclic effect on quiet days would be in excess , in the case at least of .
Again , as will be seen presently , we have to correct the observed readings of the curves for non-cyclic effect , and , so far as the corrections are concerned , what we have to deal with is the effect , whatevel its sources .
It is desirable to have on record the actual corrections made , so that aIlyone can reproduce the actual observation results who wishes to do so .
The quantities then actually recorded under the heading " " non-cyclic effect\ldquo ; in the owing tables V. to X. represent the ] excess of the recorded value of an element at the second midnight over that at the first midnight of a " " quiet ' day , of the source to which the increment is due .
In Table X. , as subsequently , the year is sub-divided into three seasons\mdash ; ' ' Winter comprising the four months mber , Decelnber , January , ; " " luinox \ldquo ; , , , , , , March , Apri ] , September , October ; ' ' Summer\ldquo ; , , , , , , May , June , July , August .
ON ' ' QUIET\ldquo ; DAYS DURDUG GHE yEARS 1890 TO 1900 , OTC ON ' ' QUIET\ldquo ; DAYS THE ELEyEN YEARS 1890 TO 1900 , ETC . . .
" " QUIET\ldquo ; DAYS DURING THE ELEVEN YEARS 1890 TO 1900 , ETC .
TABLE IX.\mdash ; Non-cyclic Effect .
TABLE X.\mdash ; Non-cyclic Effect , Seasonal Values ( Means from 11 Years ) .
Unit in case of orce .
S 15 .
The declination ( see Table V. ) is the element in which the non-cyclic effect smallest and most .
The curves are read only to the nearest so that apparent non-cyclic changes of this amount on days possess little , if any , significance .
If we take the last six years in Table V. , we see that the mean non-cyclic effect for the year in no case exceeds , and if we combine these six years , we obtain a mean non-cyclic effect which absolutely vanishes .
Again , while the mean non-cyclic effects for 1890 and 1891 are of substantial size , they are opposite in sign .
These facts are certainly calculated to rouse suspicions as to the bond fide nature of the non-cyclic effect in D. There are , howeyer , a variety of considerations which point to its having a real physical origin .
In the first place , the contribution of the true secular change to the mean non-cyclic effect in , viz. , , is in the opposite direction to the resultant effect , so that V0L .
CCIr .
354 .
C. CHREE : ANALySIS OF RESULTS ) THI KEW MAGNETOGRAPHS the daily change which we have to account for really amounts on the average of the eleven years to or .
Again , the annual inequality is in general such as to minish the numerical values found for the non-cyclic effect .
In fact we see from IV .
and V. the non-cyclic effect is positive in March and April , the months in which the annual inequality is changing fastest from positive to negative , while it is negative in July and , months in which the annual inequality is from gative to positive .
This brings us to another suggestive feature .
Four of the fiye months in which the mean clic effect is ative are summer months , while it is positive in six successive months , Noyember to April , including all the mid-winter months .
Then , as regards differences of sign in rent years , sun-spots were exceptionally few in 1890 , and though fairly numerous were rapidly diminishing in , 1896 , and 1897 , the other three years in which the mean non-cyclic effect was negative .
On the other hand , 1891 to 1894 , the years in which the non-cyclic effect has its largest positive values , were years in which sun-spots were either very numerous or increasing rapidly in number .
Thus I think that whilst it would be unwise to dogmatise as to the cause of the apparent non-cyclic effect in , the evidence distinctly points to the conclusion that it is a true magnetic phenomenon which varies , however , with the season of the year , and with the position of the year in the sun-spot cycle .
S 16 .
In the case of I ( see Table the non-cyclic effect is evidently no chance phenomenon .
The means for every year of the eleven , and for every month of the twelve , alike come out negative , and the preponderance of individual months in which the effect is negative is overwhelming .
The true secular change is here in the same direction as the apparent non-cyclic effect , and so tends to increase it , but only to an insignificant extent .
In fact it requires only 10 average quiet days to bring about a decrease in I equal to the secular change observed in an entire year .
The absolute magnitude of the non-cyclic change in I , as we see from Table X. , does not vary very conspicuously with the season of the year , but relative to the range of the diurnal inequality it is most important in winter .
In January , in fact , it averaged fully 30 per cent. of the range of the proper cyclic diurnal variation .
S 17 .
The non-cyclic effect in , see Table , is as prominent as in I. Here also it is increased by the contribution from the secular change , but 8 average quiet days suffice to produce an increment in the force equal to the full secular change for the year .
The non-cyclic effect in is on the whole wonderfully uniform throughout the year , but it appears somewhat reduced at midsummer .
Relative to the range of the diurnal inequality , it is most important in winter ; in January , in fact , it over 20 per cent. of the observed range .
In , see Table VIII .
, the non-cyclic effect is considerably less conspicuous than in or I , but more conspicuous than in D. There is a decided preponderance of months in which it has the negative sign , and the mean annual value is positive in only 2 out of the 11 years .
According to Table X. , the magnitude of the nonON ' ' QUIET\ldquo ; DAYS DURING THE ELEVEN YEARS 1890 TO 1900 , ETC .
cyclic eflect in is considerably dependent on the season of the year , reater in summer than in winter .
We have already seen that an fraction of the effect in V is ibable to secular change and instrumental causes : and if these causes are as potent in winter as in summer , the difference between the phenomena observed in and winter is the more } In the case of and , I have confined lnyself to mean seasonal values ( see Table X The equivalent in force of 1f in is nately .
Hence it will be seen that the mean non-cyclic effect in , though much less than in , is larger than in D. The non-cyclic effect in approaches closely that in H. In the non-cyclic effect appears to be but little larger than in , and it also shows the peculiarity of itive in winter but negative in summer .
The secular change and the causes described in dealing with V would tend to produce a slight ( algebraic ) reduction in the results btained for T. Solar Diurnal S 18 .
By the diurnal inequality is meant the periodi change taking place in the value of an element throughout the 24 hours .
In the case of a magnetic element on quiet days , is as we have seen an appreciable non-cyclic effect , in virtue of the value recorded at the second midnight of a day exceeds that recorded at the first midnight by a quantity , say , which may be positive or negative .
To obtain the proper perive change , must be eliminated , for unless this is done a diurnal inequality is obtained which is not periodic , which presents a discontinuity between two successive hourly values , and which is partly pendent on which hour of the 24 we select as the first hour of our day .
If we assume to result from some cause operating uniformly throughout the day , and count ' day from midnight , hours , to midnight , 24 hours , the necessary elimination is cted by applying to each hourly value the correction where is the hour counted from midnight ( hours ) .
This correction brings of course the values answerin to hours and 24 into harmony , and leaves unaltered the mean value of the element for the day as given by where represents the value of the element at hour .
Whell dealing with hourly values from a single day , or as in the present case with hourly means from 5 days a month , we really do not know how the non-cyclic efiect comes in , and our method of elimination introduce a factor not present in diurnal inequalities based on all days of the month .
Another source of slight } ) oticed .
The magnetic elemel .
DR. C. CHREE : ANALYSIS OF RESULTS FROM THE KEW MAGNETOGRAPHS in all probability a lunar as well as a solar diurnal variation .
The incidence of the selected 5 quiet days in the lunar month is largely a matter of accident , and consequently the way in which the lunar variation affects hourly values based on the selected 5 quiet days must vary from month to month .
The lunar influence appears however to be so small that any uncertainty on this ground must be triflin , even in the case of data from a single year , and when we dealing with data from 5 or 6 years , the uncertainty should be no greater than in the case of data derived from all days in a single month .
A further point that is more fully dealt with in S31 need only be mentioned now .
hours to which the curve measurements refer are G.M.T. , whereas a true solar diurnal variation is connected presumab]y with the true local solar time .
) fact that noon at Kew is minutes later than nooll at Greenwich is comparatively unimportant , as the difference is small and affects ] seasons of the year alike , but the considerable annual rango\mdash ; some 31 minutes\mdash ; in the " " equation of time\ldquo ; must be borne in mind .
S 19 .
A really serious difficulty remains to be mentioned .
The Kew ]nagnetographs are placed in a room partly underground in the basement of the Observatory , where the annual variation of temperature , though much less of course than in an unprotected upstairs room , still amounts to some F. At some seasons of the year this implies a somewhat troublesome change of temperature from day to day , but speaking generally there is seldom a difference of more than 1o F. , and only a difference of a few tenths of a degree between the mean temperature on successive days .
It was thus possible to allow pretty satisfactorily for the difference between the mean temperature of the room on the five quiet days a month and on the days of the absolute observations , from eye readings of a mercury thermometer placed under the glass shade containing the vertical-force magnet .
Up to the end of 1896 the readings of this thermometer were the only direct source of information as to tlJe temperature of the netographs .
During 1896 the thermolneter had been read usually thrice a day , at 10 P. , and 10 P. .
Prior to that it had been read twice a day , but in the earlier years of the pel.iod only once a day .
In the end of 1896 the late Kew Committee , on my initiative , introduced a thermograph .
It was so situated and protected as to give readings in close accordance with of mercury thermometer\mdash ; still read thrice daily\mdash ; and temperature corrections ) ased on its been applied from 1897 onwards to the and V curves .
In the of the the temperature coefficient is little ovel for 1o F. , and the temperature correction is very small .
In the case of V the correction is distinctly appreciable .
An examination however of the thermograph results obtained early in 1897\mdash ; at the season , as it has turned out , when the diurnal variatiolt of temperature is least\mdash ; and a comparison of the annual and semi-annual magnetic inequalities for 1896 and 1897 , led me unfortunately to the conclusion that the neglect of the diurnal variation of temperature in previous years was less serious ON ' ' QUIET\ldquo ; DAYS DURING THE ELEVFN 1890 TO ) , ETC . .
I had feared .
Influenced by this ination , I ccepted the data for to as published in the annual ' Reports , ' and ] the tables of diurnal inequalities , and numerous others , were originally calculated on this basis .
The results so obtained were utilised in the " " Preliminary Note on the Relationships between Sun-spots and Terrestrial netism , \ldquo ; published in the ' Roy .
Soc. Proc vol. , p. 221 .
SuLse- ltly I noticed that the ranges in the mean annual inequalities for V for 1898 and 1899 appeared larger relative to the ranges in earlier years than one old 1 ticipated from the relationships I bad observed Sun-spots and Terrestrial Mao.netismo , and I had little difficulty in .
the cause to the of tho temperature correction in the earlier years .
to the smallness of the temI ) rature coefficient in the netograph , the absence of a temperature correction in the data for years prior to 1897 is of little importance .
It became obvious , however , that the earlier years ' data for V must be corrected or dispensed with .
The correction has entailed a reat amount of labour , the complete recalculation of al-l the diurnal inequalities for V and I and of the corresponding Fourier coefficients , \amp ; c. The raph records from 1897 to gave smooth diurnal inequalities of temperature for each month of the year .
The range varied largely with the season , much less at midwinter than at midsummer , but the hours of maximm and minimum and the general features were nearly independent of the time of year .
The raph room and its environment remained unchanged from 1890 to 1901 , and the from the mercury ] ermometer throughout the period firmed the view that if the mean diurnal inequalities of temperature obtained for of the year since 1897 were assumed to to years prior to satisfactory corrections would be obtained .
This has been done , and I do not think that the Jlts now published are affected by any serious source of uncertainty .
S 20 .
Diurnal inequalities for L ) , I , , V , , and appear bless xr .
to XyII .
In the case of first four elements inequalities given for each month of year .
The values ascribed to any one month , say January , obtained by adding together the hourly values from each January of the period included , applying a non-cyclic correction to the sums thus formed , and then the arithmetic means .
The annual inequality for the year was obtained by summing up the mean hourly values for the 12 months\mdash ; as corrected for non-cyclic effect\mdash ; and uaking their mcans .
In the case of and nlcan inequalities are given , not lnerely for the whole 11 years , but for two combinations of years , , 1890 , 189 and 1 years of few sun-spots , and 1892 to 1895 , years of great sun-spot frequency .
The for , and were derived by means of the formulae:\mdash ; SN SH \mdash ; HsinD SD , SW SHH cosD SD , ST SHsinI SV , 358 { .
C. tNALYSIS OF RESULTS THE KEW MAGNETOGRAPHS where SN , for instance , represents the hourly value in the diurnal inequality of ; which corresponds to the values SH and SD , answering to the same hour , in the inequalities of and D. To the degree of accuracy , we may ascribe to , and I in these formulae their values for period considered , and the formulae actually used were ( see Table ) SN SH SD , SW SH SD , ST SH SV .
SD is supposed to ) ured in minutes of arc .
In the case of , and inequalities are given for only three months of the year , selected one from each of the three ; in the of and mean inequalities are for two groups of years , as well as for the whole 11 years .
In a good many cases there are two distinct maxima and two distinct mimma in the inequalities .
All maxima and minima whose existence seems absolutely unmistakable are indicated by heavy type , but in cases where the second maximum and minimum missin or doubtfully represented only the extreme values for the day are so marked .
In the case of the terms " " maximum\ldquo ; and " " minimum\ldquo ; are applied respectively to extreme westerly and an extreme easterly position of the .
This of course purely arbitrary , and is opposed to the practice of a good many writers , measure declination positively to the east .
As a mathematician , I am perfectly alive to the ight of the arguments advanced in support of this practice , which has its conveniences when mathematical calculations are made involving the use of Cartesian co-ordinates .
At a station , however , such as Kew , where the ation is the prominent feature is the rapid forenoon lnovement of needle to the , it is much more convenient for descriptive to take the west the direction .
If one to employ co-ordinate the east as ) ositive direction , it is a simple matter to the sign .
S 21 .
Iu addition to the hourly values , the tables give the of the inequality , as the rence between the algebraically greatest and least of the mean hourly values .
They also give the sum of the differences of the 24 hourly values from their arithmetic mean\mdash ; i. , from the mean for the day\mdash ; taken numerically .
The meaning assigned to the term " " range\ldquo ; is carefully to be borne in mind , because there are a variety of usages , some of which are technically more exact .
If we suppose for simplicity that the phenomena absolutely identical for each day of a month , then the maximum and minimum for each day will occur at fixed times , which will also be the times of occurrence of the maximum and minimum in the mean diurnal inequality for the month .
And if it happen in addition that the times of occurrence of the maximum and minimum each coincides with an exact hour the meaning to be assigned to the term " " range\ldquo ; has no ambiguity .
In practice , " " \ldquo ; DAYS DURING YEARS 1890 TO 1900 , however , the maximum and minimum in any single day seldom at exact hours , so that if we confine ourselves to hourly values we do not as a rule hit on either the absolute maximum or minimum , the differencc the greatest and least of our hourly values as a rule less than the amplitude proper of the diurnal movement .
Again , times of occurrence of the maximum and minimum really vary from day to , so that the we obtain by considering mean hourly values from a of days tends to be less than the arithmetic mean of the ranges deduced from the individual days results .
This last consideration doubly requires to b ) taken into account in connection with inequalities for the year or for a season .
ON ' ' QUIET\ldquo ; DAYS DURING THE ELEVEN yEARS 1890 TO 1900 , ETC .
of Declination .
( West .
) Inequality of Inclination .
VOL. CCII.\mdash ; A. 3 362 DR. C. CHREE : ANALySIS OF RESULTS FROM THE KEW MAGNETOGRAPHS TABLE XIII.\mdash ; Diurnal Inequality of Horizontal TABLE XIV.\mdash ; Diurnal Inequality of Vertical ON ' ' QUIET\ldquo ; DAYS DURING THE ELEyEN YEARS 1890 TO 1900 , ETC .
Force .
Unit C. G. S. Force .
Unit C. G. S. 364 .
C. CHREE : ANALySIS OF RESULTS FROM THE KEW MAGNETOGRAPHS TABLE XV.\mdash ; Diurnal Inequality of Northern TABLE XVI.\mdash ; Diurnal Inequality of Western XVII.\mdash ; Diurnal Inequality of Total Component .
Unit C. G. S. Force .
Unit C. G. S. Declination .
S 22 .
A remark applicable to most if not all of the diurnal inequalities is that turning points , whether maxima or minima , which occur near noon are much less variable in the time of their appearance than those which occur early or late in the day .
particularly good instance of this is afforded by the chief maximum in D. It will be seen in Table XI .
that the largest hourly value appears at 1 .
in 11 out of the 12 months .
The uniformity in the time of appearance of the maximum is perhaps even better shown in the curves in fig. 2 , which represent the inequality for December , March and June .
In the 6 months , October to March , the inequality shows two distinct maxima and minima , and in every month of the year , with the exception of September , Table XI .
shows at least a suggestion of a second maximum and minimum .
There is in all months a well marked morning minimum which , except in January and December , is the chief minimum of the day .
Its time of occurrence shows a pronounced seasonal change , varying from 7 .
in summer to 9 .
in winter .
This is well brought out in fig. 2 , which also shows the rapid swing from east to west between 9 .
and 1 .
so characteristic of the declination needle .
It would require a very large number of years ' records , and possibly the employment of greater sensitiveness , to settle definitely the times of occurrence of the second maximum and minimum , and to place their occurrence or non-occurrence in summer in a position of certainty .
In the mean inequality for the year , we have the somewhat curious phenomenon of a maximum and minimum presenting themselves at adjacent hours , viz. , 1 and midnight .
As this occurs in the inequalities for the two sub-groups of years , as well as in that for the whole period , it is presumably a natural phenomenon .
It will be noticed that though the amplitudes of the mean inequalities for the year are widely different for the two groups of years , the hours of occurrence of the maxima and minima are alike .
A minute comparison of the figures shows , however , a slight tendency for definite features of the inequality to be later in the day in the case of the second group of years ( sun-spot maximum ) .
More definite evidence in this poin is adduced later see .
The month of the year in which the amplitude of the inequality is least is unmistakably December , but the ranges for January , February and November are also very small .
There is a very rapid increase of range in March , and correspondingly rapid decrease in October , whilst in the six months April to September there is comparatively little change .
The sum of the 24 hourly differences , however , is decidedly less for April and September than for the four mid-summer months .
The range appears slightly larger in May and August than in June and July , so that the annual variation of the range seemingly presents two nearly equal maxima , respectively in early and late summer , with an intermediate inconspicuous minimum at midsummer , and a very prominent minimum at midwinter .
ON ' ' QUIIE\ldquo ; DAYS DUDURI THE ELEVEN yEARS 1890 TO 1900 , OTC .
Inclination .
S 23 .
The most distinct phenomenon in the diurnal inequality ( see Table XII .
and fig. 3 ) is the maximum between 9 and 11 , its occurrence being earlier in summer than in winter .
It is difficult to speak with certainty as to the number of maxima and minima in several months .
In the 5 months April to , there appears to only one maximum and minimum , the latter occurring between 7 and 9 In most of the other months there are at least two distinct maxima and minima , and sometimes gestions of a third pair ( see fig. 3 ) .
In the six months October to March , the principal minimum of the day is found in the morning between 5 and 7 In the late evening the value is so nearly stationary in most months , that a very long series of observations might be necessary to elucidate the exact details .
The mean inequality for the year shows clearly only one maximum ( at 10 ) , and one minimum ( at 7 ) ; but from 1 to 5 .
the value is practically stationary .
As regards the range , the phenomena are similar to those noted in the declination , but the existence of two maxima in summer with an intervening minimum is far from clearly shown .
The minimum in December is very strongly marked .
The sum of the 24 differences also shows the December minimum clearly , but the value increases uniformly , though slightly from May to August .
Horizontal Force .
S24 .
Here two distinct maxima and minima appear in the 6 winter months , October to March , but only one is clearly shown in the summer months .
In the mean inequality for the year two maxima and minima are distinctly recognisable .
Of the turning points the forenoon minimum is the best marked ( cf. fig. 4 ) .
Its time of occurrence varies only from about 10 to , being earlier in summer than in winter .
The afternoon maximum is less prominent , especially in winter , but is on the whole fairly definite .
In summer and at midwinter it appears about 7 , but in some of the equinoctial and winter months it is an hour , or even two , later .
The range and the sum of the 24 hourly differences from the mean both show a conspicuous minimum in December .
In the 4 midsummer months it is now sum of the 24 differences that shows two maxima and an intermediate badly-defined minimum .
The range shows only one maximum , in June .
The variation , however , in both cases is so small , from May to August , that the exact nature of the phenomena remains open to doubt .
Force .
S25 .
Here the principal minimum presents itself in the late forenoon .
In the summer months the smallest hourly value usually occurs at noon ; in the winter months it occurs earlier , usually at 11 .
ON " " \ldquo ; TH ELEVLN 1890 TO 1900 , ] VOL. I. ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; \mdash ; " " \ldquo ; DAYS THU INLIVFN AItS lC90 TO 1900 , t71 , C. OF ULTS The is found in the afternoon ; the time varies from 4 to 7 and , as with the minimum , is in winter than in summer ( see fig. 5 ) .
would see1n to ) maxima and minima , the second maximum and in the early morning within a few hours of one another .
This is shown , though intly , even in the mean inequality for the year .
In the winter months , , the values in the earlier morning hours become so nearly stationary true the phenomena is uncertain .
The ( ' and ) ftum of the 24 hourly differences from the mean both show two maxima in sunnner , in May and Jnly , the teryening minimum being more clearly lnarked t in other elements .
The winter minimum in Decemk ) , on the other less clearly marked than in or 1 .
Components .
S 26 .
All the months included in Table XV .
show 2 maxima and minima in the diurnal inequality of , and the same is true of the mean inequality for the year ( cf. fig. 6 ) .
As with , the most conspicuous turning point is the forenoon minimum ; but its time of occurrence is an hour or so later than in , varying from .
to noon .
In December and March the forenoon maximum is the larger ; but in June , and in the inequality for the year , the afternoon maximum\mdash ; occurring usually at 7 or 8 .\mdash ; is decidedly the larger of the two .
The early morning minimum , though not very conspicuous , seems to have a very uniform time of occurrence .
The phenomena met with in the case of ( Table XVI .
, fig. 6 ) naturally resemble those met with in D. The maximum at 1 or 2 , and the minimum at 8 or 9 are somewhat later than the corresponding turning points in the declination .
In June there is no trace of a second maximum and minimum , and in the mean inequality the year their presence is at least doubtful .
In December the evening minimum is the more prominent of the two .
Total Force .
S 27 .
The diurnal inequality of Table XVII .
figs. shows 2 maxima and minima , of which the afternoon maximum at 6 or 7 .
and the forenoon minimum at about 1 .
are the most important .
The early morning maximum and minimum occur pretty close together , and their difference is generally small .
The curves appear exceptionally regular , which is hardly what one would have expected , considering that variations in both and V are involved .
From a mathematical standpoint , is most naturally regarded as the resultant of , and V , and so mean inequalities for the year for these elements are shown alongside of the corresponding inequality for in fig. 6 .
It will be observed in Tables XI .
to XVII .
that the range and the sum of the 24 differences in March generally resemble closely the corresponding quantities from the ON ' DAYS DYSINC ?
THG lA90 TO 1900 , ON ' ' QUIET\ldquo ; DAYS THE 1890 TO 1900 , mean inequality for the year .
In fact , the data fiolll ] give a better general idea of the phenomena than those .
any other month .
S28 .
To facilitate comparison of the law of variation hout the year of the range and the sum of the 24 differences in the diurnal inequalities of I ) , I , , and V , I give a synopsis of the whole of the data in Table XVIII .
Here each quantity dealt with is expl.essed as a percentage of its mean value for the 12 months .
enables us to see more readily points of agreement and diHerence .
It will be observed that there is a close ance between the annual tions in , and that the mean of the results from these three elements for the ) and the sum of the 24 differences are very similar .
On the whole , the sum of the 24 diHerences has the more accentuated annual variation .
The annual variation seems somewhat greater in V the othel , the maxim in May being exceedingly prominent .
TABLE XyIII.\mdash ; Monthly Relative Values .
MeaIl f Mouths .
yJune 1 October S29 .
to 7 should give an adequate idea of the general character of the xlity in individual elements .
A more complete idea of phenomenon whole be obtained other types of curves .
) comnon ] known as the " " vector \ldquo ; has been used by LLOYD , AIRY and others .
It the val.iationtlJroughout the 24 hours of the disturbing force to which may be ibed the chaug.es in the horizontal component of magnetic force .
Tlje 1adius vector a fixed ) represents the horizontal component of the disturbing iorce in itude and direction .
When inequalities for and been calculated , the to draw the is to take the hourly values of the tuld inequalities .
C. CHREE : THI co-ordinates , nlar axes , one in , the other to , the meridian .
has been done in drawing ) vector diagrams for , June , the whole year\mdash ; all for the entire period 1890 to 1900\mdash ; which in to 1 ] .
Fig. 8 .
December .
ON ' ' \ldquo ; DAYS DURING THE ELEVEN YEARS 1890 TO 1900 , Fig. 10 .
March .
DR. C. CHREE : ANALYSIS OF RESULTS FROM THE KEW MAGNETOGRAPHS In the figures , NS represents the astronomical meridian , MM ' the mean etic meridian for the period , the two being inclined at an angle of The hours to which the individual observations refer are shown in the figures , all being reckoned from midnight as When and inequalities are not available , the vector diagram can be readily constructed by axes in and perpendicular to the netic meridian , employing for lates SH and SD , the latter converted into absolute measure .
This was the method which I employed in the ' B. A. Report ' for 1895 .
The difference between the vector diagrams for December and June is very striking , the area described by the radius vector in the latter curve being about 16 times that described in the former .
The diagram for March pretty closely resembles that for the whole , but encloses a somewhat larger area .
In December the portion of the curve answering to the hours near midnight forms a regular closed loop\mdash ; also seen in other midwinter months .
The March curve shows a very tiny loop , and forms a transition to the June curve , where no loop appears , though the portion of the curve answering to the hours from 7 or 8 .
to 3 .
is still markedly indented .
The longest radius vector , and so the largest value of the disturbing force , answers to about 1 .
in all the curves .
The shortest radius vector answers either to about 2 .
or 5 all the curves the radius vector passes through astronomical south between 10 and , the time of crossing being somewhat earlier in December than in March or June .
S 30 .
The second form of curve , illustrated by figs. 12 to 15 , shows the change throughout the day in the direction of an imaginary magnet freely suspended as described in S8 .
The curve may be supposed drawn by a style carried by the dipping end of the magnet , on a plane perpendicular to the mean position of the magnet throughout the day .
MM ' is the trace of the magnetic meridian in this plane .
Fig. 12 .
December .
ON ' QUIET\ldquo ; DAYS ) THE ELEVEN 1890 TO 1900 , Figr .
13 .
June .
Fig. 14 .
i380 .
C. CHREE : ANALySIS OF RESULTS FROM THE KEW MAGNETOGRAPHS Fig. 15 .
Year .
The curve is really drawn from Cartesian co-ordinates , SI parallel to MM ' , and SD ( or here SD ) perpendicular to MM ' .
Increasing I answers to movement towards the foot of the paper , increasing to movement towards the reader 's left hand .
The hours are counted and marked as in figs. 8 , 9 , 10 and 11 , to which the figures now under discussion bear a general resemblance .
The angular distance of the needle from its mean position for the day is seen to be greatest about 1 , though in June the distance at 8 .
is nearly as great .
The approach to the mean position is closest rtly after midnight , or else about 5.30 The movement perpendicular to the netic meridian is considerably larger than that in the ridian .
The curves for March and for the year are oriented , so to speak , approximately magnetic east and west .
The orientations of the June and Decelnber curves are almost equally inclined to this direction , but on opposite sides .
The curves of the second were employed by the late Admiral CAPELLO in a paper in the ' B. A. Report ' for 1898 , p. 750 , dealing with diurnal inequalities at a variety of stations .
They were also used by Sir A. RUCKER in a Rede Lecture to illustrate the actual motion of a freely suspended needle on a magnetic quiet day , showing the average non-cyclic effect experienced at Kew .
In the present case the curves of both types to the strictly periodic part of the diurnal motion .
of Diurnal Inequality in Fourier Series .
S 31 .
The diurnal inequality of an element may be expressed in either of the two equivalent forms .
, ON ' ' \ldquo ; DAYS DURING THE ARS 1890 TO 1900 , ETC .
where epresent G.M.T. counted from the first midnight of the day , an hour being taken as equivalent to ; whilst .
are stants .
Betweell the constants of the two expressions exist the relations The constants with suffix occur in terms in which the period is hours .
periods of the terms in the above expressions are respectively 24 , 12 , 8 , 6 hours .
In a previous paper dealing with Kew data I explained that terms with periods shorter than 6 hours were so small as to be of little account , and similar conclusions have been reached elsewhere .
Thus attention will be confined to the first four periods .
The values of have been found by calculation from the well known formulae ; the values so obtained are , it should be noticed , independent of whether terms of the higher orders are neglected or not .
data employed in the calculations refer to G.M.T. If results are wanted relating to local solar time , all that is necessary is the application of a table of corrections to the values found for the angles , \amp ; c. In the case of , the correction simply represents the angular equivalents of the equation of time and the longitude of the place of observation .
The corrections to , and are obtained by multiplying those for by 2 , 3 , and 4 respectively .
The corrections for the individual months and seasons of the yeal ' applicable to Kew are given in Table XIX .
They represent mean of the results given in the 'Greenwich Magnetical and Meteorological Observations ' for four years 1892 to 1895 , with the addition required to allow for the longitude of Kew , W. The corrections to omitting of course the in close agreement with corresponding results given by General STltAcHE there being sensible differences only in August and September .
In these months , however , they are in good agreement with results given more recently by ANGOT .
There are probably for a mean year errors of 1 ' in the values of in the table , and these of course multiply up for , and ; but the esults are probably sufficiently exact for all practical purposes .
For the seasons , the use of local solar time makes but little difference ; but in several months , llotably and .
November , the distinction between local solar time and G.M.T. is very appreciable .
* ' Phil. Trans. ' , for 1893 , p. 646 .
'Ann .
du Bureau Central Meteorologique de ance , ' 1899 , " " Me'moires , \ldquo ; .
B. 98 .
382 DR. C. CHfiEE : ANALYSIS OF RESULTS FROM THE KEW MAGNETOGRAPHS TABLE XIX.\mdash ; Corrections to be applied to Angles in Fourier Expansions when G.M.T. replaced by Kew Solar Time .
S 32 .
Tables .
to XXIII .
give the values of the two sets of Fourier constants , and \mdash ; using G.M.T. \mdash ; for , I , , and .
Values are given for the inequalities for the several months , for the year as a whole , and for winter , equinox , and summer , each representing , as before , a group of 4 months .
The and results are based on the 11 years 1890 to 1900 , the I and V results on the 10 years 1891 to 1900 .
An or coefficient for a of months is the arithmetic mean of the 's or 's for the individual months ; but this is not in general true of the 's or .
For instance , the arithmetic meaP the 12 values of for individual months of the year , in Table .
is , the mean of the three values for for the three seasons is , whilst the value of from the mean inequality for the year is .
The more variable the angle throughout the season dealt with , the greater is the difference between the corresponding for that season and the arithmetic mean of the 's for the individual months composing it .
In all probability it would require a very large number of years ' data to supply smooth values for the coefficients of the 8-hour and 6-hour terms , and much weight cannot be assigned to apparent irregular fluctuations in the values of these coefficients from month to month .
This is really one of the principal reasons for grouping the months into seasons .
The four mid-winter months stand out rather noticeably from the two adjacent months March and October , so that separation into three seasons seems bettel for purpose than the more usual separation into two 6-month periods .
ON ' ' QUIET\ldquo ; DAYS DURING ELLVEN yEARS 1890 TO 1900 , ETC .
The values of the constants were really calculated to at least one figure farther than appears in the tables , but the retention of the extra figure seemed more likely to mislead than to serve any useful purpose .
It would have been more logical to have similarly restricted the values of the angles , but I have followed the example of the annual Greenwich Tables and given them to minutes .
When it comes to comparing one season or one year with another , the minutes may possess a real significance , but the uncertainties in the monthly data resent dTees rather than minutes .
,384 .
C. CHREE ; ANALySIS OF LESULTS THE MMAGNETOGRAPHS TABLE .\mdash ; Coefficients and Angles in Fourier Series TABLE XXI.\mdash ; Coefficients and Angles in Fourier Serles ON ' ' QUIET\ldquo ; DAYS DURING THE FLEVEN YEARS 1890 TO 1900 , ETC .
Expansion of Diurnal Inequality of Declination ( 1890 to 1900 ) .
Expansion of Diurnal Inequality of Inclination ( 1891 to 1900 ) .
VOL. CCII.\mdash ; A. .386 DR. C. CHREE : ANALySIS OF RESULTS FROM THE KEW MAGNETOGRAPHS TABLE XXII.\mdash ; Coefficients and leso in Fourier Series Expansion TABLE XXIII.\mdash ; Coefficients and Angles in Fourier Series Expansion ON ' ' QUIET\ldquo ; DAYS DURING THE ELEVEN yEARS 1890 TO 1900 , ETC .
of Diurnal Inequality of Horizontal Force ( 1890 to 1900 ) .
Unit of Diurnal Inequality of Vertical Force ( 1891 to 1900 ) .
Unit S33 .
The mode of variation of the coefficients throughout the year is of importance , as its comparison with other phenomena seems one of the most promising methods of investigating the source of diurnal inequality .
To facilitate intercomparisons , I have expressed each monthly value of a " " \ldquo ; as a percentage of the mean value for the 12 months .
The results appear in Tables XXIV .
and .
The tables give also percentage values for the three seasons ; these are based on the seasonal diurnal inequalities , 100 representing in their case the arithmetic mean of the values for the three seasons .
The mode of annual variation of the 's is also shown graphically in figs. 16 and 17 .
It will be readily seen , either from the tables or the curves , that the mode of variation of corresponding coefficients is fairly similar in all the elements .
The coefficient has an annual variation very similar to that shown by the range of the diurnal inequality ( see Table XVIII There is the same very pronounced minimum at midwinter , with at least a suggestion of two nearly equal maxima in summer .
The annual variation in is fairly similar to that in , but the existence of a secondary minimum in summer is usually indicated more clearly .
In the summer minimum is prominent , and in it is in some cases the chief minimum of the year .
The two maxima in and are very prominent , and the interval between their occurrence is so much greater than in or that they fall approximately at the equinoxes .
This question is rther dealt with in S 37 .
TABLE XXIV.\mdash ; Variation of Fourier Coefficients throughout the Year .
Kew , 11 years .
DecemberNovemb .
ON ' QUIUT\ldquo ; DAYS DURING THE ELEVFN YEARS 1890 TO 1900 , DR. C. CHREE : ANALANAL IS RESURES TSOM THE KEW MAGNETOGRAPHS * ON ' ' QUIET DAYS DURING THE ELEyEN yEARS 1890 TO 1900 , TABLE .\mdash ; Variation of Fourier Coefficients throughout the Year .
Kew , 10 years .
JanuaryFebruaryAprilMarchJuneJulyAugustMayNovemb.erSeptemberOctoberDecemberSummerEquinoxWinter .
S 34 .
Table XXVI .
shows the relative importance of the four leading terms of the Fourier series in the case of the mean diurnal inequality for the year .
The importance of the 12-hour term as compared to the 24-hour term appears decidedly greater in , and , which present remarkably similar features , than in or I , which resemble one another .
Table XXVII .
shows how the importance of the higher Fourier terms , as compared to the first , varies with the season of the year .
In I and the relative importance of the higher terms diminishes largely as we pass from winter to equinox , and also conspicuously as we pass from equinox to summer .
In and , on the other hand , and are greatest at the equinox , and is decidedly greater in summer than in winter .
The summer value of is as conspicuously small in and as in the other elements , but the equinoctial value is closely similar to the winter value .
392 DR. .C .
CHREE : ANALySIS OF RESULTS FROM THE KEW MAGNETOGRAPHS TABLE XXVI.\mdash ; Relations between Fourier Coefficients in mean Diurnal Inequality for the Year .
Westerly component DeclinationTotal force .
ears ) Horizontal force Northerly component Inc]ination ( 10 years ) Vertical force TABLE XXVII.\mdash ; Relation between Fourier Coefficients in Diurnal Inequalities for the Seasons ( from 11 or 10 years at Kew ) .
S35 .
Taking the 24-hour term , we see in and V ( see Tables .
to XXIII .
) a fairly regular annual variation of considerable amplitude in .
the angle .
In and is larger in winter than in summer ; in I and it is the other way about .
A larger value of means an earlier hour for the occurrence of the maximum , an increase of representing an advance of one hour .
The angle increases fairly regularly in passing from winter to summer in the case of , I and ; but in V the winter value is the largest .
The angles and show a fairly regular increase from winter to summer in I and , but in and V the annual variation seems small .
S36 .
Table XXVIII .
gives an analysis of the values of for the seasonal inequalities when local solar time is used ; the corresponding hours of occurrence of the first maximum of the day are shown in Table XXIX .
The hours are counted from to 24 , with representing true local midnight .
The table goes only to tenths of hours , but suffices to show the much greater variability in the time of occurrence of the maximum of the 24-hour wave than in the maxima of the waves of shorter periods .
The variability appears greater in I and than in or V. ON ' ' QUIET\ldquo ; DAYS DURING THE ELEyLN 1890 TO 1900 , .
ae .
DR. C. CHREE : ANALYSIS OF SULTS FROM THE KEW MAGNETOGRAPHS A reference to Table XI .
, or , shows that the most prominent turning-point in the diurnal inequality for is the maximum about 1 .
( or 13 ) .
It will be seen that this comes near the time of occurrence of the single maximum in the 24-hour term , of the second maximum in the 12-hour and 8-hour terms , and of the third lnaximum in the 6-hour term .
A similar coincidence will be found in the case of the most prominent turning-points in I , , and V. Variation throughout the Year ( Fourier ) .
S37 .
Fourier series may ) employed to assist in investigating the variation throughout the of the diurnal ange , or the sum of the 24 hourly differences from the mean , or the values of the coefficients in the Fourier series representing the diurr ) inequalities .
A variety of these annual variation series have been culated and the results appear in Table XXX .
In the formulae represents an easing at the rate of per answering to the of January .
In the calculations , the results from the monthly inequalities have been treated as corresponding exactly to the middle of the months , and as separated by equal intervals of time .
Neither assumption is exactly true .
In the selection of the 5 days a month on of the objects kept in view has been that the mean of the 5 days should come near the middle of the month , but in general of course only an approximation is possible .
Again , calendar months are unequal in length .
Still the mean day of a calendar molJth seldom differs by more than hours from the position it would occupy if month were strictly the twelfth of a year , and unless one is dealing with a very , series of years , or with observational data of exceptional accuracy , very little is likely to be ained by replacing calendar months by any theoretically more perfect scheme of days .
In addition to the Fourier series formulae actually found for the annual variation , Table XXX .
gives the ratio existing between : , the amplitude of the term , and the mean of the 12 monthly values of the quantity considered ; ( ii .
) , the amplitude of the semi-annual term , and ; ( iii .
) The amplitudes and The final column in the table giyes the mean difference , irrespective of sign , between the observed and calculated values of each quantity for thoe 12 months , expressed as a percentage of the mean observed value M. In the table , unity represents 1 ' in the case of angles and in the case of force components .
ON ' ' QUIET\ldquo ; DAYS DURING 1890 TO 1900 , ETC . . .
C. : ANALYSIS OF RESULTS FROM THE KEW MAGNETOGRAPHS S38 .
first the last column of Table XXX .
, we see that the agreement between the and calculated monthly values is on the whole best for and , and worst for V. This is in accordance with what we would expect from instrumental considerations .
Even for V the reement is very good .
Except in the case of , the mean percentage error does not exceed 9 , while in the case of the ranges and the sum of the 24 hourly differences from the mean , it does not exceed 7 .
In the case of the quantities concerned are so small that an error of 15 per cent. represents but a small fraction of or , the least quantities actually measured .
This shows that annual and semi-annual terms suffice to give a very accurate representation of the annual variation , and that little would be gained by introducinero higher in the Fourier series.ae Coming next to the ratios , we that in the case of the ranges , the 24 hourly diHerences , and phenomena presented are fairly similar in all the elements ; we have varying only from to , and varyin only from to In the case of the value of is nearly the same for all the elements ; but , compared to the a1mual term , -annual term is relatively much more important in I and than in and V. The relative importance of the semi-annual term is much greater in than in or ; and in the semi-annual term is much the more important of the two , especially in the case of I and H. It will be noticed that , speaking generally , the values of differ less in the veral elements than do the values of , and that there is a close resemblance between the phenomena in and , and again between the phenomena in I and H. next to the phase angles , we see that excluding the different quantities considered show pretty similar results , and that the angles in the annual and semi .
annual terms do not differ much .
It is probably most instructive to consider not so much the themselves , as the results deduced from them for the times of occurrence of the maxima in the several terms .
In the case of the annual terms there is only one maximum .
Taking , for example , the declination range , we find the corresponding to the maximum of the annual term from the equation , or This answers to 177 days from the beginning of the year , going only to the nearest whole day , and may be taken as .
to June 26 .
The minimum occurs six months later .
A semi-annual term has of course two maxima separated by six months , with two minima , at three months intervals from the maxima .
In the case of the declination range , for instance , the first maximum would answer to , or This ) be taken as representing s9 days from the beginning of the year .
*\amp ; e ' Brit. Assoc. Report ' for 1895 , p. 223 .
ON ' ' QUIET ' ) DAYS DURING THE ELEVEN yEARS 1890 TO 1900 , ETC .
The results thus obtained appear in Table XXXL They should be regarded as affected by an uncertainty of the order of a through the treatment of the months as of uniform The maximum in the annual term appears in general near midsummer , and would somewhat earlier in and V than in I and H. The first maximum in the semiannual term appears generally late in March or early in April .
XXXI.\mdash ; Annual yariation .
Date of Occurrence of First Maximum .
' ' riability \ldquo ; of the S39 .
In discussing netic observations made the recent " " Southern Cross\ldquo ; Antarctic Expedition , made a comparison of what I called the Variability of the Declination\ldquo ; at Cape Adare and at Kew in 1899 .
I have made a similar investigation into the data at Kew during the remainder of the period 1890 to 1900 , partly with a view to whether the phenomenon is of sufficient definiteness and regularity to afford a satisfactory basis for inter-comparison of stations .
It is customary at Kew , during absolute observations , to take two declination readings with the magnet 's scale erect , and three with it inverted .
These readings show the changes taking place in during three intervals in which the magnet has remained untouched , the length of an average interval being about 4 minutes .
The observations have mostly been made at a time of the day when the regular diurnal change is slow .
The same instrument has been in use throughout , the procedure has varied but little , and the same observer has been responsible for the great majority of the obseryations .
Thus the results for different months and years are fairly comparable .
The plan I have adopted has been to sum up all the changes in reading in a year , ' Magnetic and Meteorological Observations made by the " " Southern Cross\ldquo ; Antarctic Lxpedition 1898-1900 , ' p. 24 .
or in one month of the 11 years , irrespective of sign , and divide them by the sum of the intervals in minutes between successive readings .
The resulting quantity , measured in seconds of , I have called the " " mean change per minute The results so found for the several years , and the seyeral months of the year , are given iable XXXII .
TABLE XXXII.\mdash ; Variability of the Declination .
S 40 .
If we divide the sum of the nges of reading durmg the whole eleven years by the sum of the intervals between the gs , we find for the " " mean change per minute\ldquo ; the value , which is in practical vreement with the arithmetic mean of the means for the eleve individual years in Table XXXII .
No observation has been omitted to abnormal disturbance , though some 400 to 500 observation days are included in the .
At a polar station no doubt there would be much greater on this round .
The mean values vary on the in a fairly regular way from year to year .
They have certainly been smaller since 1896 than in previous years ; but , as we shall see later , this is only what we should anticipate from sun-spot considerations .
The value for 1893 is somewhat conspicuously low compared to the values in 1892 and 1894 , while the value for 1890 appears somewhat large .
On the whole , however , I think we may conclude that the phenomenon is sufficiently definite to form a satisfactory basis for intercomparison of stations and years .
There is a well-marked annual variation proceeding with very fair regularity .
There are two minima , near midwinter and midsummer , with maxima near the beginning and end of winter .
This is similar to he phenomena observed in declination disturbances , tisme Terrestre , ' Art .
108 .
ON ' ' QUIET\ldquo ; DAYS { ELEVEN 1890 TO 1900 , 39 !
) especially at Greenwich .
When expressed in a FOUBIER the annual variation shown in Table XXXII .
takes the form , where is measured from the of exactly Table XXX .
semi-annual is here the larger of the two .
Its phase angle differs by only 1c from that obtained in Table XXX .
the correS ondintr term in the of the sum of the 24 hourly rences in D. It indicates March 20 as the ) for the first maximum .
The annual term has its first maximum about January :2 , and its mininlu1Jl of course six months later ; it is thus opposite in phase to the annual term in the expressions for the in the of hourly differences in D. -Spot S 41 .
General mention has already been made in S2 , and incidentlv in other parts of the paper , of relations between metic 1 ) henomena and ) frequency , and the numerical esults obtained by WOLF and for sun-spot have been already recorded in Table I. There other methods of estilnating solar disturbance , and it would be of interest to consider ) } ) on these , and to investigate the question a , s to apparent variations in netic phenomena with variation in the position of the spotted areas in the sun 's .
I ve thought it better , , to confine myself wholly on the esent occasion to WOLF and WOLFER'S numbers , and to cover limited portion of the rround i a thorough way , rather than to wander at over the hole field .
My own opinion is that the true inter-relationships of the phenomena likely to be reached , and ) cannot be strated , without a really minute and careful study of the facts .
In .
of sun-spot relations , I shall not adhere strictly to the followed in the first part of the paper , but shall take first the case of the diurnal periodic changes .
S 42 .
It has been established , largely of WART , and ELLIS , that there is a close connection between of maximum nd minimulll in sun-spot frequency and in amplitude of diul.nal inequality of netic elements .
In .
this matter , Mr. LLIs eliminated the annual variation in the of magnetic ranges by to any individual month the mean mplitude deduced from consecutive months , of which it was the central month .
I have pursued a different method .
This consists in the given month as a percent( of the arithl1let mean of the ranges for this ) ticular month of the year throughout the of years with .
This method is vionsly less satisfactory when applied to a short number of years than when applied to a ) eriod many sun-spot cycles , a defect from which ' method is ) It has , however , the advantage of equally satisfactory whether the sun-spot magnetic relationship is or is not variable throughout the year .
Mr. ELLIS ' method tends to smooth down the netic results , and he very properly elnploys slnoothed sun-Spot nnmbel .
S , so that the results he obtains are certain to present fewer irregularities than those obtained by me .
The results I have obtained for and appear in Tables XXXIII .
and XXXIV .
In both cases ' range\ldquo ; means the excess of the greatest over the least hourly value in the diurnal inequality derived by combining the five quiet days of each month , the non-cyclic increment being eliminated .
absolute mean value of the range is given for each month of the year , so that the absolute values for individual months can be at once deduced if equired .
TABLE XXXIII.\mdash ; Variation of Declination Range .
( mean for mean for that month for 11 ) ON ' ' QUIET\ldquo ; DAYS DURING THE ELEVEN YEARS 1890 TO 1900 , ETC .
TABLE XXXIV .
Variation of Horizontal Force Range .
( mean for month mean of that month for 11 years .
) S43 .
Before passing to the immediate question of the sun-spot connection , I would call attention to the fact that the values assigned to the ranges in Tables XXXIII .
and XXXIV .
are without exception greater than those already given in Tables XI .
and XIII .
, and that the differences between the two sets of results are more conspicuous in winter than in summer .
This is an exceedingly good illustration of the principle already discussed in S21 .
A range in Table XXXIII .
is the arithmetic mean of 11 ranges , each based on a combination of only 5 days ' results ; whereas in Table XI .
a range is based on the combination of 55 days ' results .
The range in Table XXXIII .
thus exceeds that in Table XI .
, the excess being most noticeable at those seasons of the year when the hours of occurrence of maxima and mimima are most variable , or when irregular disturbances are largest and most numerous .
S 44 .
Referring to Table I. we see that August , 1893 , was the month in which sun-spot frequency was greatest .
The largest percentage in Table XXXIII .
appears in December , 1892 , while in Table XXXIV .
the largest percentage answers to December , 1893 , and Noyember , 1894 .
The 12 consecutive months for which the sun-spot frequency was largest extended from August , 1893 , to July , 1894 .
The 12 months for which the sum of the percentages is largest run from December , 1892 , to November , 1893 , in the case of , and from January to December , 1893 , in the case of H. Of calendar years 1893 is that having the largest mean value in all three tables .
VOL. CCII.\mdash ; A. 3 402 DR. C. : ANALYSI8 OF RESULTS FROM THE KEW MAGNhTOGRAPHS The years 1892 to 1895 stand out as years of largest magnetic ranges , but their superiority is not quite so prominent as in the case of sun-spot frequencies ; 1890 shows a distinctly smaller sun-spot frequency than any of the 10 following years , but the mean netic ranges in 1890 appear slightly greater than in 1899 and 1900 .
In these two latter years , somewhat curiously , the mean percentages are alike in the case of both and H. The rise to the maximum is more sudden for magnetic ranges , as well as for sun-spot frequency , than the subsequent decline .
This phenomenon has been remarked on by Mr. ELLIS and others .
It will be observed that the percentages in Table XXXIV .
show a wider range than in Table XXXIII .
The full of this will appear presently .
Curves illustrative of Sun-Spot S45 .
In the tables devoted to the diurnal inequalities of and the mean inequality for the year is shown not merely for the combined 11 years , but also for two groups of years , , 1890 , 1899 and 1900 , representing small sun-spot frequency , and 1892 to 1895 representing sun-spot frequency .
The great excess of the ranges and the sums of the 24 hourly differences in the second group as compared to the first is conspicuous , but at first sight there is no appreciable difference in the nature of the inequalities shown in the two cases .
Fairly definite though minor differences do however exist .
If the amplitude of the diurnal inequality alone varied , then curves representing the inequalities for the two groups of years in the ordinary way should transform into one another by merely changing the scale of ordinates , and the same should be true of the difference curve , whose ordinates represent the excess of the ordinates of the sun-spot maximum curve over those of the sun-spot minimum curve .
A glance at the curves for and , in figs. 18 and 19 , shows that this is not the case .
The difference curves , due allowance being made for their smaller amplitude , are very decidedly flatter topped than the curves for either group of years .
The same phenomenon would be equally shown by or curves .
S46 .
Perhaps an even better way of showing the difference in the diurnal inequalities to the eye is to employ vector diagrams .
This has been done in figs. 20 and 21 .
In fig. 20 the mean vector diagram for the whole year has been drawn from the same origin for the two groups of years , while fig. 21 represents a difference vector diagram whose radius vector is the difference between corresponding radii vectors in the two curves of fig. 20 .
The scale in fig. 21 , it should be noticed , is double that in fig. 20 .
The mean magnetic meridians for the two groups of years are not absolutely identical , but could hardly be shown apart on the scale of the figures .
S47 .
If the sole difference between the inequalities in years of sun-spot maximum and sun-spot minimum were one of amplitude , then the points answering to the same hour ON ' ' QUIET\ldquo ; DAYS DURING THE ELEVEN , 1890 TO 1900 , ETC .
ON ' ' QUIET\ldquo ; DAYS DURING THE ELEyEN YEARS 1890 TO 1900 , ETC .
Inner curve 1890 , 1899 , 1900 .
Fig. 20 .
Outer curve 1892 to 1895 . .
Difference curve .
DR. C. ANALYSIS OF RESULTS FROM THE KEW MAGNETOGItAPHS in the two curves of fig. 20 should lie on the same radius vector , and the curves should be similar and similarly situated ; also the curve in fig. 21 should be similar and similarly situated to the curves in fig. 20 .
This is obviously not the case .
At 9 .
corresponding points in the two curves in fig. 20 lie very nearly on the same radius vector , but thereafter the ra\amp ; us vector of the inner curve ( sun-spot minimum ) forges ahead , and continues ahead until the early morning .
The points answering to 1 .
and 6 .
on the sun-spot minimum curve lie almost exactly on the same radii as the points answering to 2 .
and 7 .
on the sun-spot maximum curve .
The difference between the shape of the curves in figs. 20 and 21 is conspicuous ; the latter curve is much more nearly symmetrical with respect to the geographical meridian than the former .
Sun-Spot Relations apparent in Fourier Coefficients .
push the comparison further , I had Fourier coefficients calculated for the monthly inequalities of and from the two groups of years 1890 , 1899 , 1900 and 1892 to 1895 .
Corresponding results for the last group of years were also calculated for Greenwich , making of the values of the \ldquo ; and " " \ldquo ; coefficients in the Greenwich annual publications .
This gave results more strictly parallel to the Kew results than if one had employed the a1mual Greenwich values for the " " \ldquo ; coefficients .
To facilitate intercomparison of the different cases , I have expressed the monthly values of the 's as percentages of their mean .
It is these percentage values that form the main part of Tables XXXV .
and XXXVI .
The absolute values for each month can however be obtained at once by combining the percentage with the absolute mean value given at the foot of the tables .
The data for the 11 years have already appeared in Table XXIV .
, but it was desirable to show them in juxtaposition to the others .
Besides the monthly data , the tables give the values of the 's for the three seasonal inequalities , expressed in percentages of their mean , and likewise the absolute values of the 's in the mean inequalities for the year .
( pp. 408-411 .
) There are of course irregularities in the monthly values , but there is an unmistakable general tendency for the percentage values to rise in winter and fall in summer as we pass from sun-spot minimum years through average years to sun-spot maximum years at Kew .
In the case of the Greenwich data stand on the far side , so to speak , of the Kew sun-spot maximum data .
The same is true for in , but in the case of the other three coefficients in the Greenwich data approach most closely to the average year results for Kew .
It would be of interest to know how much of the differences between Kew and Greenwich is due to difference in geographical position , and how much is due ON " " QUIET\ldquo ; DAYS DURING THE ELEVEN yEARS 1890 TO 1900 , ETC to the fact that the Kew data are from quiet days only , while the Greenwich data are from all days .
Tables XXXVII .
and XXXVIII .
are derived from the same data as Tables XXXV .
and XXXVI .
, but they are intended to show the influence of sun-spot frequency on the relative importance of the differe Fourier Table XXXVII .
deals only with the mean diurnal inequality for the whole year but gives data for and as well as for and H. Table XXXVIII .
gives seasonal values of the ratios , but for and only.\mdash ; ( See p. 412 .
) TABLE XXXV.\mdash ; Annual Variation of Fourier Coefficients in ON QUIET\ldquo ; DAYS DURING THE ELEVEN 1890 TO 1900 , the Diurnal Ineqnalities for Different Groups of Years in D. vears IO S6 0 296 wich .
TABLE XXXVI.\mdash ; Annual Variation of Fourier Coefficients in un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum un-spot laximum ON ' ' QUIET\ldquo ; DAYS DURING THE ELEVEN ARS 1890 TO 1900 , the Diurnal Inequalities for DiHerent Groups of Years in H. XXXVII .
Ratios of Fourier Coefficients in Mean Diurnal Inequality for the Yeal TABLE XXXyIII .
atios of Coefficients in Inequalities for the Three Seasons .
Table XXXVII .
shows a steady decline the relative importance of the higher terms of the Fourier series as the sun-spot frequency increases , and this decline is accentuated in the Greenwich data .
Table XXXVIII .
shows that the decline apparent in the case of the mean inequality the year is shared in general by the seasonal inequalities .
S 49 .
The in the Fourier were also calculated for each month for the seve of years .
It will suffice , however , to record the results obtained in the case of the inequalities for the three seasons and the year in and , adding the values for the year in and N. These appear in Table XXXIX .
The angles all ON ' ' \ldquo ; DAYS DURING THE ELEyEN yEARS 1890 TO 1900 , ETC .
refer to G.M.T. , but the difl'erences between the values at the same season of year would , of course , be the same if local solar time were used .
In the of the mean inequalities for the year , with the exception of the 6-hour in and , we see a decided decrease of the angle in passing from years of sun-spot minimum , , to years of sun-spot maximum .
This means , as already explained , a later hour of occurrence of the maximum in the Fourier series ' term .
In the case of the of the for the sun-spot minimum over the sun-spot maximum years varies from .
to about 11 minutes in time \mdash ; in , to 46'\mdash ; or about 19 minutes in time\mdash ; in ; the difference in time for the four elements , , and , is about minutes .
The corresponddifferences in time in and are about 18 minutes and 14 minutes respectively .
In the algebraic mean of differences for the four elements almost exactly vanishes .
In in and , the erence in angle between sun-spot minimum and sunspot maximum is conspicuously greatest in winter , and is least in summer .
In the difference in in winter is opposite in sign to what it is at the other seasons .
With the exceptions of in , and in , the differences between the angles for sun-spot minimum and sun-spot maximum are Much the same for the equinox as for the whole year .
ON ' ' QUIET\ldquo ; DAYS DURING THE yEARS 1890 TO 1900 , Sun-Spot ; Numerical Relationships .
S 50 .
The previous methods of investigation all point to the conclusion that the relationship between sun-spot frequency and magnetic phenomena varies considerably with the season of .
To give greater less to the results , I have assumed provisionally a linear between the range of a netic element and sun-spot frequency , with " " \ldquo ; and " " )\ldquo ; constants for any one month of the year .
I have applied the formula to quantities other than ranges , , to the sum of the 24 hourly differences in the diurnal inequalities , and to the Fourier coefficients of type .
Various methods have also been adopted for calculatin and The first application was to the data already employed in calculating Tables XXXIII .
and XXXIV .
, and to corresl ) onding data for I and V. In this case the constants were determined from the 11 values of the range answering to the same month of the year , in the different , by least squares .
The results thus found for and and the corresponding values of are given in Table The value , of and are multiplied by so as to avoid decimals .
unit in is 1 ' in the case of and I , and in the case of and V. In addition to the values for the individual months of the , means are also given for the year and the three seasons .
It should be noticed , howeyer , that in this case the " " \ldquo ; or " " \ldquo ; assigned to a season ( or to the year ) is not itself deduced from a seasonal diurnal inequality , but is simply the arithmetic mean of the 's or 's for the months of the season ( or year ) ; also the value of assigned to a ( or year ) is derived from the mean \ldquo ; and the mean " " \ldquo ; for the season ( or year ) .
The value of " " \ldquo ; represents of course the value of the range\mdash ; or other quantity dealt with\mdash ; which corresponds to a total absence of sun-spots .
is measured in the same units as is simply a number , its values individual months of the period dealt with being given in Table I. ; consequently is also a number , and so pendent of the particular units employed the measul.elllent of force or of angles .
* See \ldquo ; for historical references .
ON ' ' QUIET\ldquo ; DAYS DURING THE ELEVEN YEARS 1890 TO 1900 , S 51 .
The variation of " " \ldquo ; throughout the year in Table .
is fairly sin to what we have already observed in the case of the ranges from 11 year means .
There is a conspicuous minimum in winter , usually in December , and at least a trace of second minimum near midsummer .
The fluctuations of and of from month month are somewhat irregular , and are doubtless in part purely accidental .
It would probably require a long series of years , several sun-spot cycles , to give smooth variatiolL general features are clear , and may be easily frolll a compal'ison of the mean values assigned to the three seasons .
In all four elements is decidedly least , but decidedly gl.eatest , winter .
smaller yalue of implies a smaller absolute increase to the value of an element for given increase of sun-spot frequency .
A larger value of means a larger erceutage change in the element for a given change in sun-spot frequency . .
The value of appears less in summer than at the equinox all four elements , and this is even true , though much less conspicuously , of the value of itself in all the elenlents except V. For such practical purposes as navigation and work the value of is probably the most important thing , but for theoretical work , especially when dealing with data a single station , the value of will probably prove to ) ossess more significance .
If we divide the yearly and seasonal values of , in Table XL .
, by 5 , 4 , 3 and 2 respectively , we get the following results , oing to the 1learest integer:\mdash ; This ings out the ilnil a in the seasonal of in the different elements , and closeness between the values of the equinox and for the year as a whole .
Whilst the sun-spot influence as Table .
is least , he difference from the other elements is less than would appear from the Note The results given there for V were appreciably influenced by the neg.lect of the temperature correction in years prior to 1897 , abeady explained in S 19 .
The omission of the correction diminished the range in V , and brought its values in years prior to 1897\mdash ; mostly years of large sun-spot frequency\mdash ; closer to the values in later years than they ought to have been .
VOL. CCII.\mdash ; A. 3 418 DR. C. CHRLE : ANALySIS OF RESULTS FROM THE KEW MAGNETOGRAPHS S 5 based 011 only five quiet days are somewhat uncertain , and calculations by least squares are somewhat laborious , I thus investigated a second method of finding values of and , which seems to work satisfactorily .
It makes use of values deduced from the three groups of years 1892 to 1895 , 1890 , 1899 , 1900 , and I890 to 1900 .
The method will be most easily grasped by considering a concrete case .
For this , let us consider the clination 1ange for January .
We have the following data : Inequality ] to Range Sun-spot frequency \ldquo ; Differences . . . .
Consequent value for Inequaljty 1890 to 1900 Range Sun-spot frequency Hence value Aud value ) The values tained io this for the ranges and for the sums of the 24 hourly differences in the diurnal inequalities of and are given in Tables XLI .
ON ' ' QUIET\ldquo ; ) THI 1890 TO 1900 , ETC .
5 .
C. VSIS OF SULTS F THE ] MAGNETOGRAPHS .
As i11 , the yearly seasonal values of and given in Table XLI .
represent simply arithmetic means of the values for the individual nlonths , and the values assigned to are obtained by .
the for the with the mean .
The units } ) loyed are , as in Table XL .
, 1 ' for , and for H. In comparing Tables .
and XLI .
, it should be remembered that the ranges in former are derived five days in one year , while in the latter they are derived from a group of years at least 15 days .
This implies , as already pointed out , that the average dealt with in .
is larger than the ayerage dealt with in Table XLI .
It will be observed that the yearly and seasonal values of\ldquo ; giyen in Table XL .
, are inyari bly than those given in Table XLI .
, and the same is true of the great majority of individual months .
On the other hand the yearly and seasonal values of , giyen for and , in Table XL .
, are , with one exception , less than the values given in Table XLT .
A third point of difference is , that the equinoctial values of for the in Table XLI .
have not the pre-eminence compared to the summer values that they show in the corresponding cases in Table Whilst minor differences exist , the main features in ] .
and XLI .
are alike .
Both , in wintel , a conspicuous minimum in , but a conspicuous maximum in .
Both make the equinoctial values of early equal to mean values for the year .
The values found for and , in individual months , are evidently affected by considerable uncertainties , but the modes of annual variation in the two tables resemble one another more closely than I had anticipated .
It will be noticed that the values found for for the sum of the 24 hourly differences in Table XLI .
exceed the corresponding values for the ranges in eyery single case in , and in all but two individual months in D. This difference between the two sets of values of is very decidedly larger in winter than in summer .
The inference is that sun-spot influe ] on Terrestrial fnetism is really undernated if we confine our attention to the of the diurnal inequalities , and this is especially true of winter .
phenomena , one need hardly say , absolutely bear out the remarks made in S 45 as to the flat-topped character of the difference curves in figs. 18 and 19 .
S 54 .
The on which Table XLI .
is based , has also been applied to the " " \ldquo ; Fourier coefficients in the diurnal inequalities .
The results thus obtained from the mean annual and seasonal inequalities for the three groups of years are given in Table XLII .
Yearly data are given for and , as as for and H. For arison , yearly results are also given for Wilhelmshaven , based on tables of values of the coefficients for individual years , from 1889 to 1895 , published by Dr. I have employed 1892 to 1895 for the group of sun-spot maximum 'Beob .
aus dem lSIagn .
Obs. zu YVilhelmshavcn , ' Fiinftcr Theil , pp. 46 " " QUIET\ldquo ; DAYS DURING THE yEARS 1890 TO 1900 , ETC .
years , and 1889 , 1890 as the group of sun-spot minimum years in dealing with th Wilhelmshaven data .
The latter group , it will be oticed , differs from that used for Kew .
The Wilhelmshayen data also differ in beinocD all days in the year , and in their cass one had to employ arithmetic means of the 's from individual years instead of employing values to the inequality from the group of years combined .
This last diflerence would inevitably tend to increase the size of the Wilhelmshayen coefficients as compared to the Kew , though probably to but a trifling extent .
As the declination at Wilhelmshaven is some less westerly than at Kew , the results for and at the two places are presumably less directly comparable than the results for H. DR. C. CHREE : ANAL Y OF RESULTS THE KEW MAGNETOGRAPHS .
' .
' .
' ON ' ' QUI \ldquo ; DAYS THE ELFVEN YEARS 1890 TO 1900 , ETC .
S 55 .
Considering iirst the Kew XLTI .
ellsel v see hat the phenomena presented in and by , the 24-hour very similar to those already described in the case of the ngeH .
Absolutely considered , i ) is least in wintel , and is , if anything , slightly the equin oxes than summer ; the value of is conspicuously greatest in least in suIl lner .
The value of is in , but the se of the value appears also in excess of equinoctial .
At Kew , the values of with one exception decidedly smaller than } ) value ; in like manner the yalues for generally less than those for , and the values than those of .
In fact , in sumlnel we have negative values for N. This may be accidental the numerical values appear small , but we may at least conclude that sun-spot fi.equency exerts but a trifling influence on the value of at Kew .
The reduction in the value of ) as we 1)ss to , and to , is also well shown at Wilhelmshaven , though not quite so prominently as at .
Where the Wilhelmshaven data most notably from those at Kew , is in giving a larger value of for than for both and N. The values of found for at Kew appear decidedly larger than those found in Table XLI .
' the , or even as a rule than those found for the sum of the hourly differences .
Comparing the erent elements at Kew , we see that values of for the year are pretty much alike in and , and again in and , the values appearin slightly larger in than in , and slightly smaller in than in D. quantities c and are themselves so small that conclusions based on the apparent differences in their case in the different elements , would possess an uncertain value .
Comparing Wilhelmshaven with Kew , we see that in the case of the Wilhelmshaven values in excess of the Kew ) and ; but the diflerences between the corresponding values for are so small that Kew blco decidedly the larger .
In the case of all the figures for Wilhehnshayen are excess of corresponding figures for Kew , except the value of in N. In the Wilhelmshaven values of are larger in , but the smaller in and N. In the Wilhelmshaven values for and are decidedly larger than the Kew , but still very small .
Whilst differences exist , there is a sufficiently close resemblance between the results at the two places to prove that the phenomena observed at Kew are by no means of an exceptional character .
S56 .
The method of employing three groups of years has also been applied to the ranges and the sums of the 24 hourly diflerences in the mean diurnal inequalities for the year .
The object was partly to obtain a comparison with Parc .
Maur through the intermediary of data published by MOUREAUX .
* These data are the values of * ' Ann. du Bureau Central teorologique de rauce , ' 1899 , " " Memoires .
B.9 .
424 C. CHREE : ANALySIS OF RESULTS FROM THE KEW MAGNETOGRAPHS the in the mean diurnal inequalities for the individual years of the period 1889 to 1899 .
For sun-spot maximum at Parc St. Maur I took the years 1892 to 1895 , and for sun-spot minimum 1889 , 1890 and 1899 .
The value assigned to a group of years was in this case the arithmetic mean of the values for its individual years .
In the case of Kew , the results for I and V obtained exactly like those Parc St. Maur , except that 1900 took the place of 1889 and 1890 ; but in and , I took the values given by the inequalities for the group of years combined , as given in Tables XI .
: XIII .
, XV .
, XVI .
The reason for doing this that individual years ' results for and did not exist , and it was desirable that these two elements and and should be treated exactly alike .
The difference in the method of treatment would tend slightly to reduce the values of the ranges in and at Kew as compared to Parc St. Maur .
For facility of comparison , the results for the ranges at the two places are juxtaposed .
The angular and force units are respectively and 1 XLIII.\mdash ; Values of Constants in Sun-Spot Formula , from mean Diurnal Inequalities for Individual Years or for Combinations of Years ( Years Grouped ) .
S57 .
Colnparing the Kew results in Table XLIII .
with the corresponding results in Table XLI .
, we find an excellent agreement , especially in the case of D. Comparing the Kew data for the ranges in Table XLIII .
with the corresponding data in Table XL .
, we see that the values found for " " \ldquo ; in Table XLIII .
are always slightly the smaller .
This is also true of the values of " " \ldquo ; except in the case of V. On the other hand , the values found for in Table XLIII .
are larger than those given in Table .
The differences between the two tables are by no means large , but they emphasise the necessity of bearing in mind the principle explained in S21 .
Still our attention to the Kew data , we see that , as in the case of in Table XLII .
, the values of for and come close together , the being " " QUIET\ldquo ; ) 1)URING THF TO 1900 , the while the values of difier but , and in the same direction in the two cases .
Coming now to the results for St. Maul see that values for \ldquo ; are all in excess of the corl.esponding values Kew , but the values of at the two places are general ] osely alike , the difference being sometimes in direction , sometimes in other .
The agreement in the values of at the two places is closest in and , and least close in I. The fact that at St. Maur V shows a slightly larger value of than is the whole , the in the ) shown at and Parc St. Maur is exceedingly satisfactory .
The resemblance ) ears decidedly closer than between Kew and Wilhelmshaven .
This is partly due to the that the declinations at Parc St. Maur and Kew difler only about S 58 .
Table XLIV .
gives a final set of results from the mean diurnal inequalities for the year .
They were obtained from the individual inequalities for 11 or 10 years by least squares .
The reement with the results in Table XLIII .
is good .
This is , if only because it shows that it is ) ossible to obtain by means of the much simplel nlethod of Table XLIII .
results .
a good approximation to those found by least TABLL XLIV.\mdash ; Values of in Sun-Spot Inequalities Individual Years by Least Squal.es . .
I have ] applied the ot formula to the results for non-cyclic in I and given in Tables .
and VII .
, and to the monthly esults obtained ] iability of as explained in S 39 .
The years were after the fashion in S 52 .
results are give in Table the whole seasons .
Owin to the smallness of the ) quantities , it old 1 very long series of years to accidental S of .
Thus the seasonal results , more particularly for the non-cyclic effect , affected by considerable uncertainties , especially as regards the values of ) .
In fact , my object in recording the seasonal values was largely to enable the reader to judge for himself what attitude he should adopt towards the mean results for the year .
V0L .
CCII. .
31 426 DR. C. : ANALySIS OF SULTS FROM THE KEW MAGNETOGRAPHS TABLE XLV.\mdash ; Values of Constants in Sun-Spot Formula .
S 60 .
conclusions I myself draw from Table XLV .
are that non-cyclic effect and variability of declination both strongly influenced by sun-spot frequency , and that influence is decidedly most prominent in winter , though hardly I should imagine to the extent indicated by the values found for in the non-cyclic effects .
One exceptional feature in Table XLV .
is that " " \ldquo ; appears greatest , and least , at the equinox .
This might arise from an over-estimate of\ldquo ; It will be noticed that the mean values for the year of in non-cyclic effect in and I , and in the variability in , are remarkably close to the corresponding values found in Table .
for the case of the 24 hourly differences in these respective elements .
When find two phellomena similar laws of variation , whilst the conditions are artificially varied in a definite , we infer at least an intimate between their causes .
One is thus tempted to deduce that the source , or f , ources , of the non-cyclic effect is not absolutely distinct from that of the regular dinrnal inequality .
If , for insbance , the inequality is due to electric currents in the upper atmosphere , one would be disposed to infer that the source of the non-cyclic effect is there also .
It seems improbable that two absolutely distinct loci would be so similarly influenced by the changes , whatever they are , that proceed passu with sun-spot variations .
It is unsafe to base theories on a few numerical coincidences , which after all may be accidental but I hope it will at least be allowed that non-cyclic effect and declination variability are true manifestations of terrestrial magnetism , and that the phenomena they exhibit at different stations and over periods of time form rather a promising field of investigation .
Relations Jleteorological Phenomena .
S61 .
Evidence has been adduced by lnany physicists , including Dr. BALFOUR and Sir NORMAN LOCXYER , suggestive of connection -between sun-spots ON ' ' QUIET\ldquo ; DAYS THE ELEVEN YEARS 1890 TO 1900 , .
427 and various meteorological phenomena , such as rainfall , barometric pressure , \amp ; c. It is thus desirable to see what was happening meteorologically at Kew frum to 1900 , and , more especially , whether there was or was not any appreciable diHerence between the meteorological data for the two groups of years 1892 to lS95 and 1890 , 1899 and 1900 .
Table XLVI .
gives particulars of the mean values of the chief meteorological elements at Kew for the whole period , and for the two shorter groups of years .
In the case of vapour pressure , 1891 was omitted , as reliable data were wanting for two or three months of the year .
The mean daily range of temperature table is the mean excess of maximum over imum on individual days .
What impresses one chiefly on studying the table , is the remarkably close agreement between the means for such comparatively short periods of .
When the means for shorter periods differ , they are as often as llot o11 the me side of the mean for the 11-year period .
TABLE XLVI.\mdash ; Mean Values of Meteorological Elements at Kew .
S 62 .
As results from the diurnal inequalities are more strictly parallel to most of the Kew netic data , mean diurnal inequalities of temperature and barometric pressure\mdash ; the two most important meteorological elements\mdash ; were calculated from the data published by the Meteorological Office in their " " Hourly Ubservations This was done for groups of sun-spot maximum and sun-spot minimum years .
Data for 1900 being not yet available , 1889 was substituted in the sun-spot minimum groul ) .
The following were the results:\mdash ; RANGES in Mean Diurnal Inequalities for the Year .
Group of years .
Barometric pressure . .
C. OF :lities ) ressed in Fourier Series .
ronp of 1892 to 1 . . . . .
1889 , 1890 , 1899 . . . . .
roup of ycars .
of ycars .
pressure ( unit inch ) .
lS92 to 1889 , 1890 , 1899 .
.\mdash ; .
.\mdash ; .\mdash ; .
lS92 to 1889 , 1890 , 1899 .
.\mdash ; .
.\mdash ; .\mdash ; .
Hcre , what strikes one principally is the closeness of reement between results fxom periods containing so few years .
S 63 .
As the results for barometric pressure , attention may be drawn to the important results obtained at the Dutch Colonial Obseryatory at Batavia* for Fourier series expansions of the diurnal tlity .
The unit in all cases is 1 millim. of mercury .
I would specially call attention to the extreme smallness of the variation in the amplitudes of the terms , and to the fact that 1891 to 1895 represents , as compared to the two adjacent groups of years , large sun-spot frequency .
We know from the important researches of M. ANGOT , referred to in the " " Preliminary Note that the connection between magnetic variations and sun-spot frequency is unmistakable at Batayia as in Europe .
'Observations mndc the tlag , and Mot .
Obscrvatory Batayia , ' vol. ) , 1900 , p. 143 .
" " QUIET\ldquo ; DAYS DURING THE lS90 TO 1900 , S64 .
In results as those in Table XLVI .
, allowan ce ) the fnct that in the case of snch element the results ) single station for a single year , or evcn occasionally for of } not be for more } a very limited country .
of the other elements much ) rescntati-c in character , and it fe to conclude that , as between th rous l ) ; of and inimum , there ltot in ] south of generally character at ( of the same order as that manifested in ) nagnctic elemen ts .
From this ever exclude eric e ] potential .
There seem , unfortunately , no data whence one as to changes from year to year in the lneatl potential in the ) .
An attempt is at present ) to fill this , but with what success remains to be seen .
I think nation ( the point , botb ] vels , is much to be desired .
S 65 .
We shall now consider tain rnces between netic and logical phenomena in the year , which merit ation in connection theories as to the source of the netic diurnal variations .
If we take the mean monthly values of temperature range and , used in obtaining Table XLVI .
, and represent each lnonthly value of the mean of the 12 DlOllthly values , we get the results in Table For , this also giveF the } ] ges are means found for the of oyed in calculation o XXXIT I. and XXXIV .
TABLE XLVII .
In the case of ) pressure , we are llot with ) with the mean yalue for the 24 hours .
I have introduced it principally to that 111 seasons of maximum and minimum yapour pressure ) later in tl year than the seasons of maximum and netic r , and that the values of vapour pressure in March and much alike , whereas ranges in the former month are double those in the latter .
It will be noticed that the an nual in in Table XLVII .
, , from 62 to 128 , is less } 430 .
C. : ANALySIS OF ULTS FROM THE KEW MAGNETOGRAPHS winter minimuIn for temperature occurs unmistakably in January , and so is later thau the netic ninimum .
S 66 .
The points of agreement and difference between temperature and netic phenomena brought out more fully by a consideration of the Fourier series for inequalities of temperature .
Such series have been given by General for first order stations of the Meteorological Office employing an 11-year , and for Greenwich employing a 20-year period .
Taking General STRACHEY'S , I have calculated the annual variation of the " " \ldquo ; coefficients\mdash ; i.e. , the amplitudes of the periodic terms\mdash ; expressing the value for each month as a percentage of the mean of the 12 monthly values .
The results are given in XLVIII .
, along with corresponding netic data representing the mean esults for , and V , as given in Tables XXIV .
and XXV .
In the case of and , I give a mean result for Valencia , Falmouth , and Stonyhurst , in addition to values for Greenwich and Kew .
General STRACHEY'S notation is for my TABLE XLVIII .
iation of Fourier Coefficients in the Diurnal Inequality , throughout the Year . .
Temperature alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew . .
Tcmperaturc alcncia , Falmouth , Stonyhurst ) Temperature , Gl.eenwich .
Kew .
Magnetics , Kew .
Temperature , Greenwich .
Kcw .
Magnetics , Kew .
S 67 .
Before discussing Table XLVIII .
I General STRACHEY has pointed out that ' Phil. ' for 1893 , p. 617 . .
ON ' ' QUIET\ldquo ; DAYS THE 1890 TO 1900 , analysed into Fourier terms is no proof of the existence of " " distinct physical influences operating recurrlng of 24 , 12 .
( hours).\ldquo ; He has in fact shown that the simple hypothesis that the heatin effect of the sun varies as the sine of .
altitude , bein nil when the sun is below the horizon , gives rise to Fourler serles which constitute not at all a bad first approximation to the results actually found at certain stations .
Acceptin this contention , we may however get much useful information from the FOURIER'S analysis .
For instance , we know that the consequences of surface on underground temperature can be most easily arrived at by an analysis which treats the Fourier terms as independent entities .
Also the examination of differences between the relative values of coefficients at different seasons of the year , or in different years at the season , is perhaps the most hopeful way of ascertaining community of orlgm different phenomena .
S 68 .
Returning to Table XLVIII .
we see that in the case of there is a very close resemblance between the annual variations in temperature and magnetics .
The reement is just about as good as that between the temperature results for Kew and Greenwich , and decidedly better than that between some of the individual netic elements at Kew .
The difference December and January is cel.tainly less conspicuous in temperature than in netics , but still at both Kew and Greenwich the December temperature mean is the lower .
In the case of and the telnperature results at the different stations differ more widely than in the case of , but clearly the law of annual variation is here altogether different from that appearing in netics .
The temperature data for seemed too small and uncertain to be worth detailed consideration .
As relative importance of the different Fourier terms in temperature and magnetics , the following are the ratios for the yearly means:\mdash ; Magnetics ( mean f This shows the much greater relative importance of the 24-hour term in temperature than in magnetics .
The difference is probably even greater than the figures suggest .
The magnetic values of , \amp ; c. , refer to mean inequalities for the year ; whereas General STRACHEY'S sponding values represent arithmetic means of 12 montbly values , and his phase angles are most variable in the shorter riod terms .
It will l ) seen in fact that the values obtained for , \amp ; c. , from the mean temperature inequalities for 1892 to 1895 , or 1889 , 1890 and 1899 , in S 62 , are less than the values we have just derived from General STRACHEY'S figures .
43 ) .
C. { : ANALYSIS RESULTS THE MAGNETOGRAPH .
of of agreement and just discussed , by the case of V Kew .
As already stated , all the tables to re-calculated , ' allowance for the non-application of a temperature correction in eal.lier years of the .
I ) zzled for a time by the , at first fact that whilst the new values of differed very appreciably from he old , there was exceedingly little difference between the new and old values of the othel coefficients , or of any of hase angles .
The explanation simply that the external telnperature inequality was exactly represented in the \mdash ; of course , 011 much reduced scale\mdash ; with of maxim um and so retarded as to them much closer to the corresponding times for the variation of V. S 70 .
There are othel important to be drawn from the comparison of temperature netics .
The source of the solar diurnal inequality in netism is , of coul.se , the sun , acting some way or othel ; the only question is how .
The esearches of , SCHMIDT , ADAMS , and others all point to the conclusion that metic force at the is derived , if not exclusively , at least in a wholly celltres of force within that surface .
Accepting this conclusion , one would llo3t ally phaps look for the action of the sun on the Earth itself the canse of the diurnal inequality .
MAXWELL and others have ointed out that the tidal effect of the sun by modifying internal strains must exert some influence on onetic strata ; but the trifling size of the lunar netic inequality indicates that any such effect must be very small .
The only other obvious direct effect on the itself is the diurnal heating and cooling of the surface .
It has been pointed out , howevel , by Principal that , owin to the great preponderance of ocean , the portion of the Earth 's surface expo , sed to appreciable temperature variation is comparatively small .
What I wish to point out in addition to this is that any diurnal variation of temperature in the Earth diminishes very rapidly as the depth increases , and the rate of decrease is rapid for the part of the surface heatin represented by the 12-hour and 8-hour Fourier terms than for the part answering to the 24-hour term .
Also any seasonal underground of temperature lags behind the surface change .
Now we have that the ratios borne by the amplitudes of the order terms to the of the 24-houl ltot ] , but much larger , in the mnetic olelnents than in temperature ; hilst Jnagnetic seasonal ] ) lier than of face t , and earlier than ) esented 1 an element to be directly on temperature .
is , of course , the further obvious point that surface temperature show-s no large difference from of sun-spot maximum to years of minimum , and still less will underground temperature .
This may } ) quite to those accept Professor able in the ' ' for 1889 , , as ) ) ' the chiefif ON ' ' QUIET\ldquo ; DAYS DURING THE 1890 TO 1000 , ETC .
not exclusive source of the magnetic diurnal inequality to be external to the I must confess , however , that whilst I consider the conclusion , which ] been at by other physicists from other points of , to be in nain Crect , think that the investigation ought to be repeated when practicable for at least ) diffel.ent groups of years , one representing numerous , the other few pots ; trustworthy simultaneous data at least ten or twelve well distributed observatories would be highly desirable , if not absolutely } ) ) , the vertical force is likely to } ) roy a ) ] between S71 .
It was pointed out in the " " Preliminary Note\ldquo ; that WoLliER'S table of sun-spot frequencies is based on all days for which solar observations exist , whereas the Kew magnetic data refer to five days a month only .
If one divided the days of a month for which sun-spot data exist into two equal groups , the one containing the days of the lowest frequency , and if from this group one selected five days at random , the mean frequency for these days would often be but a small fraction of own mean for the month .
Thus if , as has frequently been suggested , sun-spots were the immediate cause of magnetic disturbances , the disturbance being synchronous with the spot , the presumption would be that the sun-spot frequency for all and for qutet days would be altogether different , the latter being much the smaller .
As explained in the " " Note data which WOLFER does publish for individual days were not used for the following reasons ; ( 1 ) being only they presumably ] trustworthy than the finally accepted monthly values appearing in Table I. ; ( 2 ) not infrequently data were wanting for several of the selected quiet days .
1 have , however , looked carefully into the matter , to see whether the results here arl.ived at would have been materially modified if sun-spot frequencies had been taken from days only .
For this purpose , frequencies were calculated for each month , using values for the selected qniet .
These are } ) ared in Table XLIX .
with the corresponding data derived by WOLFEB from his proyisional values for days .
In the means I left out of account three moIJths when there were less than two quiet days .
TABLE XLIX.\mdash ; WOLFER'S Provisional Sun-Spot Frequencies .
October VOL. CCII.\mdash ; A. 434 DR. .C .
CHREE : ANALySIS OF RESULTS FROM THE KEW MAGNETOGRAPHS I also show side by side in Table L. the mean seasonal and yearly values obtained for sun-spot frequencies Table I. and from the two sets of provisional data made use of in Table XLIX .
Results are given for the groups of sun-spot maximum and minimum , as well as the whole period .
TABLE L. \mdash ; Mean Seasonal and Values of WOLFER'S Sun-Spot Frequencies .
72 .
The smallness of the rences between the means from and from quiet days seems to disI ) absolutely of any theory which regards sun-spot frequency on a given day as any guide whatsoever to the quiet or disturbed character of Terrestrial Magnetism on that particular day .
For our immediate purpose , however , the important thing is the evidence afforded by the above Tables , that the results we have obtained for the and constants in the sun-spot formula , would be but little affected if we took frequencies answering to magnetically quiet days only .
In the methods in which we have employed groups of years , the sun-spot data required for calculating seasonal or mean yearly values of and are those given in Table L. It is thus obvious , so far at least as mean annual values are concerned , that the result of substituting quiet day frequencies would be as nearly as possible nil .
Summer values of and would be slightly raised by the substitution , winter and equinoctial values very slightly diminished .
S73 .
I next investigated whether rises and falls in WOLFER'S frequencies , as given in Table I. , and as deduced from provisional values for quiet days proceeded on parallel lines .
Table I. presents a rather singular feature .
If we attach the signs or \mdash ; according as there is a rise or fall in the frequency in passing from one month to the next , we find that there are no fewer than 90 changes of sign out of a total possible of 130 .
We have long sequences of \mdash ; occurring alternately , e..q .
, from January , lS95 , to March , 1896 .
It looks as if the monthly changes generally went beyond the temporary equilibrium position , or else as if the estimate of them generally overshot the mark .
The quiet day provisional ffequencies present the same feature , the monthly changes they show differ appreciably in general from ON ' ' \ldquo ; DAYS DURING THE ELEVEN YEARS 1890 TO 1900 , ETC .
those deduced from Table I. Omitting the three months already referred to , we have 125 monthly changes common to the two sets of frequencies .
Of these 97 agree , and 28 differ in sign .
If , however , we take only those cases in which the monthly change amounted to at least 10 in one or other of the two sets of data , we find an agreement in in ] than 60 cases out of 7 Passing to the monthly changes in the percentage values of the and ranges , given in Tables XXXIII .
XXXIV .
, find there also equent alternations of rises and lalls .
For , in ] XXXIII .
, rises and falls alternate without a break from January , 1893 , to , 1894 .
the ] , however , there are appreciably fewer transitions of sign in the and H Tables than in ] I. When the monthly changes in the two sets of sun-spot frequencies differ in , the changes in Tables XXXIII .
and XXXIV .
agree sometimes with the set , sometimes with the other , there being in both tables a very slight ponderance of agreements with WOLFER'S day frequencies .
This may be purely accidental , but at all events the magnetic changes agree at least as well with the sun-spot data from all days as with those from quiet days only .
S74 .
There still remains the question whether there is a distinct connection between the mean values for individual of sun-spot frequencies and magnetic ranges .
In the case of there does seem fairly evidence of such a connec- tion .
If we take the changes of sign we have just been considering in Tables I. and XXXIV .
, we find an agreement in sign in 74 cases , as against disagreement in .55 cases ; and if we confine ourselves to cases where the changes between successive months ' values , in Table I. , is at least 10 , we find agreement in 34 cases out of 50 .
This amount of agreement is more than would be at all likely to happen by pure chance .
In the case of , however , the agreements and disagreements in sign are almost equally numerous , and even when we confine ourselves to cases where the sun-spot change was at least 10 the balance of agreements is too small to be relied on .
In the case of a diurnal range of is exceptionally large ; the curve is read only to ; the average non-cyclic effect is , and there may be a small uncertainty through temperature .
Thus the uncertainty in individual figures in Table XXXIV .
cannot well be less than 2 or 3 per cent. even at midsummer .
In the case of , the range often exceeds 10 ' , and seldom falls below 5 ' ; the is read to ; the non-cyclic effect is very small , and there is no temperature correction .
Thus the ainty in individual figures in Table XXXIII .
should be considerably less than in Table XXXIV .
Thus the absence of an unlnistakable connection between ranges and frequencies in indiyidual months has stronger evidence in its fayour than has the apparent connection between ranges and monthly sun-spot frequencies .
If instead of individual months we take years the connection is cleal , and it is manifest enough , even in , for considerably shorter periods than years at times when sun-spot frequency has a rapid general drift in one direction , .
Some light on the degree of intimacy of the connection is derivable from the alnoullt of agreement between the individual monthly ranges observed in and I and those calculated from the values of and in Table .
The probable errors calculated from the differences between the eleyen observed and calculated values for four months representing different seasons of the year were as follows:\mdash ; Greatest .
probable errors by no means that the range for any individual month depends on only five days ' results , and that -cyclic effects in I and are so considerable .
Still the results tend to confirm the conclusion to which other considerations point , that the departures from the mean shown by magnetic phenomena on quiet days calmot be solely determined by the simultaneously existing sun-spot frequency , unless we are prepared to hold that either WOLFER'S sun-spot data or netic data are habitually aflected by considerable errors .
S 7 There a variety of ways of ) laining the phenomena , some of which may indicated iefly .
The solar system may be under the action of some external ency , whose effect on the sun is made manifest by the occurrence of sun-spots .
As the influence simultaneously existent would naturally vary from part to part of the solar system , this hypothesis would not be inconsistent with the view the influence on the sun itself is measured by the sun-spot frequency .
The influence might be of the nature of a radiation making the Earth 's atmosphere offer a lessened esistance to the electric currents due to solar action to which various physicists ascribe the diurnal netic inequality .
Or the sun itself might be the sole agent , if we suppose that sun-spot frequency is a qualitative rather than a quantitative measure of its activity .
The sun-spot might tand to the really active cause somewhat as the smoke from a to the heat of its furnace .
When smoke issues we know the fire is alight , but we cannot deduce the actual heat .
third obvious alternative , , that the sun-spot frequency is a direct measure of the contemporaneous activity , but that the effect at the earth depends appreciably on what has been happening at the sun for some time previously .
This is what we might expect to happen if the sun were the origin of a radiation which took a considerable time to part with all its ionising power , or which travelled so much slower than light that emanations leaving different parts of the sun simultaneously reached the earth at appreciably differen times .
If however this third alternative were true , we should expect magnetic phenomena to show a general tendency to behind sun-spot frequency .
Mr. ELLIS has made investigations on " " QUIET\ldquo ; ) THI ls90 TO 1000 , this point , but has detected no certain lag .
It is , however , to doubt whether a short lag\mdash ; a month or two for instance\mdash ; would with certainty be shown by the method he employed .
Before concluding , I wish to acknowledge the yaluable assistance given several members of the staff of the Observatory Department of the National Physical Laboratory .
Mr. T. W. BAKER , the Chief Assistant , gave much useful help in the interpretation of the earlier magnetic records .
To Mr. G. BADDLRLY and .
B. FRANCIS I much arithmetical , especially in connection with the calculation of Fourier coefficients .
The calculation work has been exceedingly heavy , and some errors have doubtless escaped me , but every reasonable precaution has been taken to secure accuracy .

