]\gt ; PART CERTAIN PROPERTIES 0F THE EARTH'S ATllosPHERE .
S19 .
The solar barometric variation . . . . . .
42 S20 .
The lunar diurnal barometric variation . . . . .
45 S21 .
The electrical conductivity of the upper atmospbere 46 PART VI .
THEORY or THE EXTERNAL SOLAR AND LUNAR DIURNAL MAGNETIC VARIATION FIELDS .
S2 Outline of the mathematical theory for the general law of atmospheric conductivity . . .
48 S23 .
The relative amplitudes of the magnetic variations .
Deduction of the effective atmospheric oscillations and the law of conductivity . . . . .
S24 .
The absolute values of the amplitudes and of the electrical conductivity in the upper atmosphere . . . . .
62 S25 .
The heating effects of the upper air currents . . . . .
64 S26 .
Discussion of the phases of the magnetic variatious . . . . .
67 S27 .
The residual variations and the not dependent solely on local time . . . . .
70 Note . . . . . .
73 S1 .
Introduction .
THE regular daily changes of the earth 's magnetism , considered as a world-wide phenomenon , afford problem of much interest and importance .
It is one , more- over , which the researches of BALFOUR STEWART and SCHUSTER have shown to be more vulnerable to attack than seem most of the problems of terrestrial magnetism .
But in spite of their success , and of the contributions of subsequent writers , the comprehensive study of the subject has suffered undeserved neglect .
It seems an unfortunate fact that the efforts of magneticians are unduly devoted to the accumulation of data , the time and labour spent in their discussion being , proportionately , inconsiderable .
The promising theory that the daily magnetic variations arise mainly from electric currents circulating in the upper atmosphere , under the impulsion of electromotive forces produced by the convective motion of the air across the earth 's permanent magnetic field , was first propounded by BALFOUR STEWART .
* In a simple but penetrating discussion of the features both of the solar and lunar diurnal variations , he showed the power of this theory to account for the facts in a way which none of the other theories then current could do .
The theory was greatly dsveloped , and rendered more , in two important memoirs by SCHUSTER in 1889 and pting a suggestion by GAUSS , he applied the method of spherical harmonic analysis to the solar diurnal magnetic variations , to determine whether they had their origin mainly above or below the *Cf .
his on " " Terrestrial Magnetism " " in the 9th edition of the ' Encyclopaedia Britannica ' ( 1882 ) .
He made considerable use of BROUN'S admirable study of the lunar diurnal variation of ] ination at Trevandrum ( 1874 ) .
SCHUSTER , ' Phil. Trans , vol. 180 , p. 467 ( 1889 ) ; , vol. 208 , DR. S. CHAP AN ON SOLAR AND LUNAR by the same potential function if more harmonics of higher degree were included .
But I doubt whether any improvement thus made would be very substantial or of real value , and I have thereforejudged it best not to discard the above simple and successful representation based on the West force variations alone .
It may be added that if an independent attempt were made to determine , from the North force variations alone , the values of the harmonic functions present in them , of the type which represent the West force data , the results would differ little from those actually calculated from the latter .
The notation of the harmonic functions used in this analysis is described in S9 .
In that notation a function ( 2 ) was used in Tables III .
and III .
to represent the variations at the individual stations of various co-latitudes .
The value of in each case was found to be .
In Tables III .
and III .
two such functions , corresponding to and , were used in each instance .
The constants and are set out in Table S ( 9 ) .
As previous investigations have indicated , both inside and outside causes contribute to the magnetic field at the earth 's surface , so that the vertical force variations cannot be deduced theoretically from the horizontal force variations .
On the contrary , the potential function ( if such exists ) , which represents and is calculated from the vertical force data alone , affords the means of separation of the respectively external and internal parts of the whole variation field .
It proved on examination that functions ( 2 ) , of precisely the same type as were used for the horizontal-force data , serve likewise for the vertical-force variations , only the numerical coefficients ( denoted in this case by and ) being different .
These also are given in Table , alongside the values of and .
The corresponding calculated values of and are given in Table IIL for comparison with the observed data .
For the purpose of the subsequent discussion it was clearly advisable that and and should be determined on some definite plan which would at least give results which were comparable in the different cases .
Where only one harmonic function was involved in the representation of a given set of Fourier coefficients , as in Tables III .
( a ) and III .
( ) , the course adopted was very simple .
The weighted mean of the various values of the function , corresponding to the mean latitude of each of the nine groups of observatories , was taken numerically , i.e. , negative values being treated as positive ; the similarly weighted sum of the group mean.values of .
( or ) was also taken , the signs being reversed where this had been done for the calculated ( negative ) values .
A simple division then gave the required coefficient , or As regards the weighting , the Northern group means were each given unit weight , and the Southern means each half a unit DIURNAL VARIATIO OF TERRESTRIAL MAGNETISM .
of weight .
The fifteen Northern observatories were thus given a total weight six , and the three , four , or five Southern observatories ( according to the number available in the different cases ) received a total weight one or one and a-half .
Where two harmonic functions were involved in the representation of one set of data , as in Tables IIL and III .
( ) , the same general method of weighting and combining the data was used , except that two equations had to be formed , to give the two coefficients .
Usually the middle four or five values of or were used in one equation , and the emainder in the other .
S9 .
The rmonic Representation of the Magnetic Variation Field .
A potential function which varies with the local time , but is otherwise the same at all stations along any parallel of latitude , can always be expanded in a series of spherical harmonic functions of the form ( 3 ) .
Here are numerical coefficients ; is the local time reckoned in angle at the rate of 15 degrees per hour , where is the longitude measured towards the East from some standard meridian , and is the time corresponding to that meridian ) ; is the distance from the earth 's centre to the point considered , at colatitude ( in this paper measured from the North pole as origin ) , and is the earth 's mean radius .
The function , or as it will generally be written , is the ordinary tesseral harmonic of degree and order ; it can readily be calculated from the formula ( 4 ) in which the factors of the form , where is a positive integer , denote .
The part of ( 3 ) which depends on is continuous and satisfies LAPLACE'S equation within the sphere .
In the case of the magnetic variation potential , consequently , it arises from an eIectric current system outside the earth .
The remaining portion of ( 3 ) , which depends on , similarly results from a current system within the earth .
The and I are chosen to indicate the respectively external and internal origins of the corresponding parts of the potential .
A term in the magnetic variation potential would lead to the following DIURNAL VARIATIONS OF TERRESTRIAL MAGNETISM .
concerned in this paper .
These functions were numerically evaluated for each of the twenty-one values of in Table , and the group means I. to IX .
were formed ; by means of these the coefficients and and were determined as described in S8 , the West and vertical force Fourier coefficients being compared with the formulae ( 9 ) and ( 10 ) .
The North force values of and were then calculated from and by means of ( 8 ) .
The observed and calculated values of and for the various components and seasons are tabulated in Tables III .
to ( ) , and the values of \amp ; c. , are given below in Table C. It is clear from the Tables III .
that the above harmonic analysis , although of a very simple character , gives a fair representation of the main features of the daily magnetic variation , except for the 24-hour component of the North force variation ; for the other periodic terms of the latter variation the agreement with the potential calculated the West force is better\mdash ; perhaps even satisfactory , when the irregular " " run\ldquo ; of the North force is considered .
S10 .
Comparison with Previous Harmonic Analyses of the Diurnal Magnetic Variation .
It is of interest to examine how far the varlous studies of the solar diurnal magnetic variation , hich have been made by different methods and with different data , agree in their main results .
The principal terms in SCHUSTER'S , FRITSCHE'S ( 1913 ) , and the present analyses , so far as they are comparable with one another , are collected.in Tables and E. In the former table the potential functions derived from the North and West force variations ( or , in the present paper , from the West force variations only ) are given .
Instead of and , however , the amplitude and phase are given , where ( 14 ) .
All the results have been modified where necessary , to conform to the notation of this paper .
In some cases the authors cited carried their analysis further than in the present instance , in others , as the table indicates , they stopped short of it ; but the principal erms were in all cases the same .
In Tables and , the figures for the present paper , under the heading " " annual terms are obtained from the mean of the solstitial and equinoctial results in Table B. They therefore represent the mean for a whole year , as in the case of SCHUSTER'S and FRITSCHE'S results .
The seasonal terms in the present analysis , on the contrary , refer to a half-year only , i.e. , the two quarters centred at the solstices .
They may be expected of somewhat larger amplitude than if they had been derived from the two half-years , as in the previous discussions .
DIURNAL VARIATIONS OF TERRESTRIAL MAGNETISM .
of the field .
The notation is the same as that explained in SS9 and 11 , only the seasonal divisions and the unit of force ( here in place of ) being different .
In Tables VI .
( b ) to ( d ) the calculated values of and , corresponding to these harmonic functions , are placed for comparison beside the values computed from the observational data .
The agreement between observation and calculation is naturally less good than for the solar diurnal variation , both because the accidental error in the data is greater in the present case ( the whole effect being smaller ) and because single observatories are here used in place of groups of observatories , so that the local irregularities are larger .
The agreement is better , on the whole , in Tables VI .
, ( d ) than in Table VI .
( b ) , which rests on only half as much observational material as the two former ; it is also better for the West than for the North component of force , as in Part II .
The agreement is surprisingly good in Table VI .
( d ) for the horizontal force components .
The vertical force variations are smaller than the horizontal force variations , and some of the values of and , determined from the former , are very uncertain .
On the whole , however , while the present data might be very considerably improved upon , the results prove more satisfactory than I had expected , at any rate for the horizontal force potential and .
In judging the success of the analysis , regard may be had to the agreement of phase between the harmonic components of different periods and the reproduction of other features of the analysis of the solar diurnal magnetic variation , which show the close parallelism of the two phenomena .
Although some of the tables of observed and calculated data in Table VI .
do not seem to show much correspondence between the two , the results in Table suggest that the assumptions underlying the analysis ( i.e. , that the variations can be represented by the functions or ) are sound , and that the discordances from the results of calculation arise from a relatively large amount of accidental error in the observational data .
The use of so small a number of observations is , of course , a fit ground for criticism , and calls for a repetition of this part of the investigation on a larger scale .
The present is to be considered as merely a pioneer attempt .
For this reason the analysis has been narrowly restricted , and the results must be discussed with due recognition that the percentage error is not small .
S14 .
Comparison ?
with VAN BEMMELEN'S As bearing upon the question of the accidental error of the initial data of Tables VI .
( a ) to ( d ) , it is interesting to compare the present values of with those calculated by VAN BEMMELEN for the same observatories ( only the coefficients are given by the latter author ) .
It should be borne in mind , however , that neither the epoch , the amount of observational material dealt with , nor the method of computation , was the same in the two cases .
Dr. VAN BEMMELEN divided DIURNAL VARIATIONS OF TERRESTRIAL MAGNETISM .
'37 The simplest hypothesis , that of a uniformly conducting earth , will first be considered .
LAMB'S theory enables the amplitude ratio , and phase difference , to be calculated for a uniformly conducting sphere of radius and specific resistance for any harmonic term in the potential of the external primary variation field .
Tables equivalent results for certain values of and are to be found in paper , but as they are insufficient for the more extensive observational data of Table I further calculations have been made which are summarized in Table K. Where the two sets of values of and overlap they are in agreement .
The Table gives the values of and corresponding to the two variables and on which they depend ; is defined by the equation ( 20 ) where , the number of seconds in a day ( the period corresponding to ) , is equal to 86,400 in the case of the solar diurnal magnetic variation , and 89,500 ( approximately ) in the case of the lunar diurnal variations .
On the hypothesis that the whole earth is uniformly conducting , we must take ( S9 ) , and cm .
Hence for the solar diurnal magnetic variations ( 21 ) 4 .
' and for the lunar diurnal variations ( 22 ) TABLE K.\mdash ; Amplitude Ratios and Phase Difference between a Primary and Secondary ( Internal ) Magnetic Field , Induced in a Sphere of Uniform Conductivity corresponding to Spherical Harmonics of Various Degrees and for Various Values of , or the Ratio Resistivity .
seasonal harmonics ( cf. Table I ) .
The differences are in most cases easily within the limits of accidental error .
As regards the lunar diurnal magnetic variation , the corresponding value of is 121 , and , as Table shows , the above values of would hardly be affectffi by the change .
The calculated values for the lunar variation are given in Table , alongside the observed data .
Considering the uncertainties in the determination of the small quantities concerned , the agreement between the two sets of values is good .
The observed phase differences are , with one exception , of the right sign , and their.mean is in satisfactory accordance with the mean value of .
The lunar variation , therefore , supports the above hypothesis as fully as the rehability of the data allows one to expect .
The theoretical values of corresponding to the values of in Table were then compared with the observed amplitude-ratios ( which we denote by ) , and in accordance with ( 23 ) , the values of , or , were calculated for each of the annual harmonics .
The four values found were .
Adopting the value , the following calculated values of were deduced from the formula ( 24 ) both for the annual and seasonal harmonics : N.\mdash ; Amplitude Ratios of the Surface Potentials of the External and Internal Diurnal Magnetic Variation Fields .
Again the agreement with observation must be considered satisfactory , so that with the aid of only two disposable constants and a good account has been DIURNAL VARIATIONS OF TERRESTRIAL MAGNETISM .
given of the values of sixteen observational quantities ( eight values of the amplitude ratio and eight of the phase difference ) .
Naturally , however , the hypothesis of a non-uniformly conducting earth such as we have considered must be regarded as giving only a convenient idealized representation of the real facts .
The theoretical values of calculated for the lunar diurnal variation are given in Table ; they differ but little from those of Table N. The observed values in Table are somewhat irregular , but their mean is in satisfactory agreement with the calculated mean , when the accidental error of the lunar data is considered .
S18 .
The Electrical Conductivity of the Earth as the Diurnal Magnetic Variations .
In S17 the following two quantities were determined , in connection with the theory that the earth has a conducting nucleus of radius and specific resistance surrounded by a non-conducting layer:\mdash ; ( 25 ) Here denotes the radius of the earth .
The thickness of the outer layer is given by ( 26 ) m. , or about 160 miles .
The specific resistance of the inner core is similarly found to be as follows:\mdash ; ( 27 ) C.G.S. These values may be compared with those deduced by SCHUSTER in his second memoir .
* The calculation there made was intended to give only a rou estimate , and in order to explain the apparent absence of phase difference between the external and internal fields ( cf. S2 ) it was necessary to assume a high\mdash ; practically infinite\mdash ; conductivity of the inmer core .
Hence no comparison with ( 27 ) is possible .
On this basis , however , the deduced value of the thickness of the non-conducting layer was 1000 km .
, in place of the present value 245 km .
The difference seems altogether beyond the probable limit of error in the latter result , and 1000 km .
must be regarded as definitely too large .
The estimate 245 km .
can hardly be liable to so much as 50 per cent. error , so that the outer layer is probably from 200 to 300 km .
, or 100 to 200 miles in depth .
It should not be forgotten , however , that we have no evidence for a sharp line of demarcation between the outer non-conducting and the inner conducting matter .
'Phil .
Trans , vol. 208 , 1907 , p. 169 .
VOL. CCXVIII.\mdash ; A. As regards the esistivity 2 , we may note that this is considerably less than that of the constituents of the outer rocky crust of the earth .
Ordinary sea water is more conducting ; thus while for distilled water is about , and for rain water 6 .
, SCHMIDT* found that for North Sea salt water is ULLER states that water from the Mediterranean Sea may be only two-thirds as resisting as this .
also finds that the resistivity of moist earth ranges from to ' , but that for dry earth it is about .
Lowy has measured the specific resistance of the ordinary constituents of the earth 's crust ( rock , stone , and so on ) and co1lcluded that for the majority of specimens is greater than , though the results varied somewhat with the moisture in the stone .
It would seem , therefore , that apart from the comparatively shallow oceanic or water-bearing strata at or near the earth 's surface , the outer crust is from 100 to 1000 times as resisting as , from our calculations , the inner core appears to be .
The above data apply only to the solid crust , which geologists consider , on the evidence of seismological and gravity measurements and the study of radio-activity , to extend down to a depth of 30 or 40 miles only .
S Very little is known of the nature or condition of the underlying substance .
The above value of is much greater than the resistivity of metals such as iron at ordinary temperatures ( for which is about ) ; their resistance , however , increases with temperature , and may also be affected by the great pressures to which the interior layers of the earth must be subject .
As regards the depth of the non-conducting layer , 200 to 300 km .
, it may be noticed that HELMEBT and also TITTMANN and , have concluded that the inequalities in the distribution of mass near the earth 's surface extend down to a depth of 120 km .
; variations in the electrical properties of the earth seem therefore to extend below this region of variation of elasticity .
PART V. -ON CrRTAIN PROPERTIES OF THE EARTH'S ATMOSPHERE .
S19 .
the Diurnal arometric Variation .
As was remar ed in S6 , the atmospheric motions to which the daily magnetic variations are to be attributed , on the STEWABT-SCHUSTER theory , are the horizontal and not the vertical movements .
Before proceeding with the study of the magnetic variations , a brief statement will be made of the principal relevant facts regarding the daily circulation and electrical conductivity of the atmosphere .
CHMIDT , ' hrbuch d. Drahtlosen Telegraphie , ' vol. 4 , p. 636 , June , 1911 .
ULLER , ind .
, p. 638 .
, ' Ann. .
Physik , ' 36 , p. 125 , October 3 , 1911 .
S Cf .
Sir A. GEIKIE'S Article on Geology , ' Encyc .
Brit. ' ( 11th ed vol. 11 , p. 664 .
HELMERT , ' Encyc . .
Math. Wiss , 1 , , vol. 2 , 1910 .
" " Geodetic Operations in the U.S.A. , 1906-9.\ldquo ; 'Report to 16th Conferenco of the International Geodetic .
Association , ' by O. H. TITTMANN and T. HAYFORD .
has also remarked ; GOLD has used the law , which fits the observations very closely .
SCHUSTER adopted the law ( or ) , which represents the facts moderately , though distinctly less well than .
SCHNIDT'S expression was ( 29 ) .
The seasonal variation of amplitude has been represented by ANGOT by the formula ( 30 ) where is the sun 's declination and its distance .
Its magnitude is well illustrated by the following results of an analysis of the Bataviau barometric observations for the period 1866 to 1905 :\mdash ; Spring ( February to April ) , Autumn ( August to October ) , Summer ( May to July ) . . . .
, Winter ( November to January ) .
, Mean equinox .
, Mean solstice . .
The mode of origin of the daily barometric variation has been much discussed , but the question whether the important semi-diurnal component is of tidal or thermal origin , or both , seems still open .
If it is fundamentally a tidal effect , resonance with a free atmospheric period of 12 hours must be assumed , since the lunar diur1lal barometric variation ( which can hardly be of other than tidal origin ) is much smaller magnitude .
Probably resonance is necessarily involved.also if the cause is thermal , as the KELVIN-MARGULES theory supposes .
In any case , however , the 12-hour variation is clearly much more fundamental than the 24-hour component , a fact which has interesting bearing on the magnetic variations .
If is the velocity potential of the motion , so that is the velocity in the direction of , the simplest theory connecting and the pressure variation asserts that ( 31 ) where is the velocity of sound .
At the earth 's surface , taking as ( 32 ) in millimetres of mercury ( so that in the same units is 760 ) , we find that ( 33 ) DIURNAL VARIATIONS OF TERRESTRIAL NETISM .
The numerical value of is or where is the radius of the earth .
The calculated values of the semi-diurnal components of velocity to east and south , at the latitude of St. Helena are approximately given by ( 34 ) J. S. DINES has determined the actual values at St. Helena to be { 35 ) The agreement in phase is therefore very good , and also in amplitude for the easterly component ; the southerly component variation is , on the contrary , much larger than the simple theory would predict .
It has already been remarked that seems to diminish upwards , so that in accordance with ( 31 ) , the value of in ( 33 ) should diminish in amplitude with increasing height .
* The phase should also vary with height in the same way as for the pressure variation .
S20 .
, in the 'Mecanique este , ' xiii .
, seem to have been the lirst to lnentio that tidal motions should be present in the as well as in the oceans .
He also discussed a series of barometric observations in France and found evidence of a very small lumar semi-diurnal variation , which he inclined to attribute to the indirect ( rather than direct ) action of the moon working through the lunar tidal motion of the .
SABINE ( ' Phil. Trams .
, ' 1847 ) proved from the discussion of two years ' barometric obsel.vations at St. Helena that the magnitude of the lunar semi-diurnal barometric variation was of the order of mm. of mercury .
The most complete determination of the effect , however , been made at Batavia ( ' Observations , ' 1905 ) ; as in the case of the solar diurnal variations , the * As the magnitudc of the diminutio1l , somo nropean mountain servations discussed by ' Wicn .
hriften , , 1892 , may be quoted , tlthough m observations may not be altogether representave of tho conditions in the free atmosphere at the height .
CoIlsidcring the whole amplitude of the barometric variation , reduced to sea-level according to the formula , the results from heights of to 2S km .
were found to be mm. , or even less , whereas the sea-level value the same latitude is mm. ( ibid. ) shows that , at the mountain referred to , the phase angles range from 110 or 120 degrees to 140 lees , in place of lunar term is variable and irregular , while the semi-diurnal term is constant , its value ( calculated from forty years ' observations ) being , in millimetres of mercury , ( 36 ) .
WAGNIJR ( ' Gottingen Abh .
, ' ix .
, 4 , 1913 ) , has also discussed six years ' hourly barometric observations at Samoa , with the aim of determining the various characteristics of the lunar semi-diurnal barometric variation .
The material was insufficient for the attempted purpose of investigating the effect of season , lunar distance , declination , and phase , but the mean result for the semi-diurnal tide may be , vi ( 37 ) .
The data do not suffice to determine the dependence of phase and amplitude on latitude , but we shall assume that the phase is constant , while the amplitude is specified by the function .
The Samoan result does not support this conclusion very strongly when compared with the Batavian determination , but the material is insufficient to enable a definite judgment to be made as yet .
For the present the Batavian result will be adopted as the basis for discussion in this paper .
The corresponding value of the velocity potential is given by ( 38 ) .
S21 .
Electrical ductivity of the Upper Atmosph ere .
It has already been mentioned ( S6 ) that the electrical conductivity of the upper atmosphere was discussed by SCHUSTpR in his second memoir .
The possibility of the production of a conducting layer such as was suggested in that discussion by the agency of ultra-violet radiation from the sun has recently been considered by SWANN , * in connection with recent physical data bearing on the problem .
Assuming the ionized constituent of the atmosphere to be oxygen , it is possible to determine the rate of supply of energy of ultra-violet radiation necessary to maintain the proposed conductivity C.G.S. electro-magnetic units , in a layer 300 km .
thick , where the average pressure is atmosphere ) .
The ionization potential for oxygen is 9 volts , and only the radiation of wave-length less than 1350 is available for Consid ring the solar spectrum to be that of a black body at C. , it appears that of the whole solar radiation would be thus available ; it is found , however , that even if the total radiation of all wave-lengths were absorbed in the act of ionization , the rate of ionization would still be onIy one-sixteenth of what is equired .
SWANN points out that the method of overcoming the Jagnctism , ' XXI .
, p. 1 , 1916 .
, 'Proc .
Camb .
Phil. Soc 15 , p. 483 , 1910 ; 'Phil .
Mag , p. 685 , 1913 .
IURNAL yARIATIONS OF TERRESTRIAL TISM .
currents .
The STEWART-SCHUSTER theory of their origin will form the basis of this enquiry , and the data to be considered will be those of Table , p. 27 , for the solar diurnal magnetic variations , and Table , p. 35 , for the lunar diurnal variations .
The former data are the more accurate , and are alone suitable for exact numerical comparison with the results of theoretical calculation .
But it will be found that great advantage accrues from the possession of data relating to these two closely similar yet independent sets of magnetic variations .
For convenience later in the discussion it is necessary at this stage to outline the mathematical analysis of the above theory .
The following investigation is a continuation of two earlier studies of the same problem by SCHUSTER* and the present writer .
It is more general than the first of these , and also embodies certain simplifications of the methods of both papers .
The details of the calculations are in all cases similar , however , and will be omitted here .
We suppose that the phenomenon takes its rise in a spherical shell of mean radius and thickness ( small compared with ) .
The conductivity of the air in this shell will be denoted by , and we shall suppose that or is a function ( ( the zenith distance of the sun from the point considered ) expressible in the most general terms as a power series in .
Clearly , if is the declination of the sun , and are the co-latitude and longitude of the point , we shall have ( 89 ) at local time .
We suppose , therefore , that ( 40 ) Consequently ( 41 ) where ( 42 ) It is convenient to transform and into Fourier series in as follows : \mdash ; ( 43 ) sf .
Here ( 44 ) ( 45 ) * SCHUSTXR , ' Phil. Trans , vol. 208 , p. 186 .
'Phil .
Trans , vol. 213 , p. 288 .
VOL. COXVIII .
and ( 46 ) .
so that are power series in and For the present we may consider an atmospheric oscillation of the general harmonic type , for which the velocity potential is ( 47 ) .
The radial magnetic intensity of the earth 's field ( measured positive outwards ) will be .
denoted by V. The components of electric force , X and , measured towards the south and east respectively , are given by ( 48 ) X If we express X and in the form ( 49 ) V the function will be the current function of the electric currents produced by X and In order to obtain from ( 49 ) , SCHUSTER first determined by eliminating and afterwards determined by the use of S. In my own earlier treatment of the problem I sought to avoid the calculation of by using the resistivity in place of the conductivity , so that was the function which was expressed in the form ( 40 ) .
But it is better to keep as the fundamental function , and this is easily effected , and directly determined , by the use of the following method .
On eliminating and multiplying both sides of the resulting equations by \mdash ; this being the step which yields the improvement of method\mdash ; we obtain the following result:\mdash ; ( 50 ) * Here we neglect the effect of self-induction for the present ( see , however , S26 ) .
We define by the property that the flow across an element of length from left to right , is DIURNAL VARIATIONS OF TERRESTRIAL MAGNETISM .
We may suppose that the solution is expressible in the form of a series of spherical harmonic functions , thus ( 51 ) .
In this equation , when is positive , will be defined as equal to .
When is numerically less than is zero .
When the above values of and are substituted on the one side , and of X and on the other , ( 50 ) becomes ( 52 ) sin where ( 53 ) and ( 54 ) By equating corresponding periodic functions of and on the two sides of ( 52 ) , we may determine the values of and For the time being we will now limit the problem to the determination of the part of which depends solely on local time , i.e. , to the case in which is independent of , so that on the left-hand side of ( 52 ) the term will be omitted .
This is the thing as omitting from V the part which depends on .
If we regard the earth as a uniformly magnetized sphere , with its magnetic axis inclined at an angle to the geographical axis , we may write ( 55 ) where is a constant ( approximately equal to having regard to our conventions of sign ) while is the longitude measured from the meridian West Greenwich ) which contains the earth 's north magnetic pole .
The constant has , for convenience , been already introduced in ( 51 ) .
Neglecting , therefore , for the present , the second term in V , ( 45 ) becomes ( 56 ) Hence it appears that for all values of ( on equating the factors of corresponding periodic terms on the two sides of the equation ( 56 ) , we find that ( 58 ) where ( 59 ) There are an infinity of equations of type ( 58 ) , one for each positive and negat integral value of. .
Both sides of ( 58 ) may be expressed as the sum of a series of spherical harnonics of type , where ( cf. 44 , 45 ) may take all integral values .
By equating the factors of corresponding harmonics on the two sides , a doubly infinite set of equations is obtained , from which the doubly infinite set of constants may be determined .
If the atmospheric conductivity is u1liform , .
and are zero except wheu , so that only the central equation of the set ( 58 ) appears , and on the right-hand side vanishes except when , i.e. , when .
Also and .
Thus , comparing the two sides of the equation , the only two values of for which is not zero are .
Consequently , term in the velocity potential of the tmospheric oscillation produces monics of types and others , in the electric current function , when the conductivity is unifonn .
If also , as in the atmospheric oscillations and is zero and only the single harmonic will appear in .
In this case the value of is found to be ( 60 ) .
If the atmospheric conductivity , or , is of the form , in ( 58 ) can take five values , while on the right can take three values to ) .
Hence may range from to , by ( 59 ) , while may range from to Only contain , as before , and they also contain only even powers of contain to the first and odd powers , and so on .
The coefficients \ldquo ; have been worked out by SCHUSTER , for this law of conductivity , to the fourth power of , for the two harmonics and in the velocity potential .
The more important of these coefficients , which will be required in the subsequent discussion , may be obtained from Table by writing For more complicated forms of the calculation of the values of , although straightforward and simple in principle , becomes increasingly laborious .
In my DIURNAL VARIATIONS OF TERRESTRIAL TISM .
previous investigation the calculation was carried as far as and for the law , but this degree of approximation proves to be insufficient for the purposes of this paper .
The terms have therefore been computed as far as the fourth order , for the same law and for the atmospheric oscillation .
In this case ( 58 ) takes the form ( 61 ) .
The values of the more important coefficients calculated from this equation are given in the following table:\mdash ; TABLE O. Annual terms\mdash ; ' Jisonal t sm TABLE ( continued ) .
Terms ( continued ) \mdash ; - We may now also take into account the second term in V , depending on longitude ( cf. 55 ) .
On substituting this term in place of V in ( 52 ) , and taking ( so that we consider only a semi-diurnal atmospheric oscillation of type ) , the left-hand side is found to be ( 62 ) When the conductivity is uniform it follows that the corresponding parts of , the current function , are as follows:\mdash ; ( 63 ) .KC ein DIURNAL yARIATIONS OF TBIAL MAGNETISM .
The general form of may be written thus ( 64 ) in place of ( 51 ) , and from ( 50 ) and ( 62 ) we obtain the following equations for and ( 65 ) , ( 66 ) si .
We shall only consider the simple law for the conductivity , and we shall neglect the seasonal terms containing .
To the first order in the following are the values of and for the principa124-and 12-hour longitude harmonics:\mdash ; ( 67 ) The values in the first line of ( 67 ) necessarily agree with ( 63 ) .
This calculation is very incomplete , but it will sufficiently ilustrate the discussion of the longitude terms in the magnetic variation , and until the main part of the phenomenon is better accounted for , it is hardly worth while to make a more elaborate determination of and The above investigation gives the method by which the electric current function is obtained .
MAXWELL has shown that the magnetic potential corresponding to a term in , at a radius within the spherical current sheet , is given by ( 68 ) Since the lower limit of the current sheet is probably fifty*or more miles above the earth 's surface ( the radius of which we denote by R ) , the mean value of may be perhaps 2 per cent. greater than R. In this case , for some of the higher harmonics in the Tables and , such as , the factor will not be quite negligible , and later it will be again referred to ( S23 ) .
We have so far assumed that are all integers , so that the periodicities , with regard to time , of the atmospheric oscillation and conductivity are commensurate .
In the case of the lunar diurnal atmospheric oscillation and the atmospheric conductivity ( which depends on solar time ) this is not the case .
The difficulty may be overcome in a fairly accurate way by regarding the conductivity as a function whose *Cf .
$ 21 , 24 .
.56 .
CHAPMAN ON OLAR AND LUNAR period is a lunar day , and allowing for the slight difference between this and the time period by supposing it to have a slowly varying phase .
Thus if in the above yeStigation represents solar time , and represents lunar time , we may write ( 69 ) where measures the lunar phase , and increases from to during the interval between successive epochs of new moon .
The atmospheric oscillation ( 47 ) now has the time factor in place of .
Now .
Hence , ifwe replace by , the above investigation remains unchanged .
The result must be interpreted in lunar time , however , so that a term now becomes ( 70 ) .
Thus the phase of the magnetic variation of the same period as the atmospheric oscillation remains invariable , while the phases of other components increase or decrease by whole multiples of , during the lunar month .
This is what is actually observed in the lunar magnetic variation , and the data of Part III .
have been obtained by allowing for this .
It will be noticed , howevpr , that the components for which is negative vary very rapidly in phase ( by 2 per lunar month , 'denoting the numerical value of ) .
These terms are included in the solar diurnal magnetic variation ( where their phase is constant ) , but not in the lunar diurnal magnetic variation as here computed .
S28 .
The Amplitudes of the fagnetic T The theoretical results of S22 will now be applied to the discussion of the relative amplitudes of the magnetic variations , leaving the absolute amplitudes and phases to be considered later .
In the case of the lunar diurnal magnetic variation , the nature of the monthly changes of phase for the different periodic components indicates that the fundamental atmospheric oscillation here concerned is semi-diurnal .
The type is not known ( S20 ) , but our assumption of the form is perhaps not far from the truth .
According to the theory of S22 , the presence of components of other periods ( with varying phases ) indicates that the electrical conductivity of the atmosphere is not uniform and constant throughout the ( solar ) day .
We will first consider what evidence is afforded , concerning the nature of the daily variation of conductivity , by the relative amplitudes of the harmonics of different periods .
The simplest law of variation for the latter is given by ( 71 ) DIURNAL yARIATIONS OF TERRESTRIAL NETISbI .
The amplitudes of the magnetic variations produced by an atmospheric oscillation , corresponding to this law of variation of conductivity , are proportional to the values of , where is obtainable from Table by .
For the present we shall consider only the " " annual\ldquo ; harmonics at the equinoxes , so that we shall also write .
Since the atmospheric conductivity cannot be negative , cannot exceed unity .
Table shows that the subsidiary ( i.e. , non-semi-diurnal ) harmonics are greater the to.reater the value of If we take , the theoretical amplitudes are found to have the following relative values , being an undetermined constant ; the observed equinoctial amplitudes in the lunar diurnal magnetic vari.ation are added for comparison ( cf. Table J ) ; the ratios of the two sets of numbers are also given , assuming to be such that the ratio for is unity:\mdash ; TABLE P. Theoretical relative arved a .
The law ( 71 ) would clearly account for a considerable proportion of the harmonics and , but wholly fails in the.case of the fourth harmonic , both as regards magnitude and sign .
This comparison shows , however , that the daily variation of electrical conductivity is at least as great as that indicated by the formula since if were less than unity the amplitudes of and would be still smaller in comparison with ; also it shows that the sign of must be positive , i.e. , that the conductivity must be great by day and small by night , since the observed phases of and are the same as that of .
If were negative , their theoretical phases would be opposite to that of A more pronounced variation of the atmospheric conductivity can be represented by the law ( 72 ) In my first paper on the lunar diurnal magnetic variation I considered the following special case of this law* ( 73 ) 13 which is never negative , and gives a much greater ratio of day to night conductivity than does ( 71 ) .
With the aid of Table the following values of the rela 'Phil .
Trans , vol. 213 .
On p. a graph of this cxpression is VOL. CCXVIII.\mdash ; A. I theoretical equinoctial amplitudes of the magnetic variations are obtained .
* The ratios of these to the observed values of Table are also given .
TABLE Q. Theoretical relative amplitudes The relative amplitudes of all four components now show very fair agreement .
Clearly ( 72 ) is a great improvement on ( 71 ) as a representation of the law of variation of the electrical conductivity .
Probably the true law is more complicated than ( 72 ) , but we shall not trouble to seek for a closer approximation .
The point gained is that , by taking a law which gives a general representation of the variation known to be probable on other physical grounds , we have been able to explain the presence and order of magnitude of magnetic variations of periods other than that of the primary atmospheric semi-diurnal oscillation .
The calculated amplitudes of the variations arising from , and so on , comparable with those in Table , are found to be ( 74 ) In the case of the lunar diurnal magnetic variations , however , these change their phases with great rapidity , and are not included in the " " observed amplitudes given in Table J. We may proceed further to examine the seasonal changes in the relative amplitudes of the various components .
The numbers in Table relate to the equinoctial variations , and are obtained by taking to be zero in Table O. During the solstitial quarters , however , is approximately 20 degrees .
If this value is substituted , the following numbers , corresponding to those in Table , are obtained:\mdash ; ( 75 ) * The numbers given are 60 } the factor 60 being serted for convenience .
The factor allows for the fact that the magnetic variations are produced at some considerable height abov1 the surface regions whero observations are mado ; is taken as .
@S Formulae resembling those of Table were given in my memoir , but only cnrried as far as .
and ; in this paper some small corrections are made , and a further approximation is made by including the terms , and .
With the nve values of and the resnlts secm to show satisfactory numerical convergcnce .
DIURNAL VARIATIONS OF TERRESTRIAL MA ETIS same relative scale as the results in Table , for 12-honr and 24-hour movements of the same equatorial amplitudes , the following values are obtained Table Phil. Trans ( 77 ) The ratios in Table might be reproduced more closely if a 24-hour atmospheric wave of about one-third the amplitude of , and in phase with , the 12-hour is supposed present in the solar diurnal variations .
The seasonal changes in the amplitudes of the annual rmonics in the solar diurnal magnetic variations are similar to those in the iations .
The diminution at the solstices is partly explicable by the similar decrease in the semi-diurnal barometric variation ( S 19 ) , which in each case takes place without appreciable change of phase .
The further reductions in the higher magnetic harmonics , at the solstices , is to be referred to the effect of the dependence of conductivity on the solar zenith distance , though the law ( 73 ) has been seen to be insufficient to account altogether for the observed changes .
The necessary modification of ( 73 ) would seem to be in the direction of a more rapid diminution of ( the conductivity ) as increases from degree to 90 degrees .
This is probable on other grounds ( S21 ) , but the theoretical discussion of its consequences would be a very laborious task .
The same modification would also increase the theoretical values of the seasonal monics Q , which for the diurnal maguetic variations were found to be three or four times too small , relatively to the annual harmonics , when compared with the observed data .
But a more serious difficulty arises here .
If we examine solar diurnal seasonal harmonics it is found that they are in fair agreement with the theoretical values from ( 73 ) , except in regard to phase .
The seasonal harmonics in the lunar variation are , in fact , nearly as great compared with the annual harmonics as in the solar variation .
This is immediately in } initial data of this cfbles I and yith Vl .
( c ) VI .
, and following ratios of the solar and lunar seasonal lnonics E the same :\mdash ; t$.4 ( 147 ) These numbers are comparable with those of Table ; the same preponderance in the ratio for the diurnal harmonic is seen here , but the first three of these tios a less than great as those Il suppose the electrical conductivity such that the seasonml uonics due to an ] atmospheric oscillation have the relative nitudes shown in the luuar DIURNAL yARIATIONS OF TERRESTRIAL such changes as due to the temperature variations in the lower regions of the atmosphere .
As regards the lunar day , regular temperature variations should be almost or quite non-existent , and should not alter with height .
The relative decrease of the solar as compared with the lunar semi-diurnal atmospheric oscillation , from a ratio of 16 to one of about 10 , may possibly be explained in this way .
In order to obtain a numerical estimate of the electrical conductivity of the region in which the magnetic variations are produced , we will determine the constant in the formula ( 40 ) by a comparison of the lunar diurnal atmospheric velocity potential ( 38 ) with the observed magnetic variation .
Considering the equinoctial " " annual ' harmonic , from Table , we find the amplitude ( cf. 17 ) in C.G.S. units to be ( 78 ) R. The theoretical value ( S23 ) is ( 79 ) ( where ) , and , paying no attention to signs for the present , ( 80 ) Substituting these values in ( 79 ) , and equating the result to ( 78 ) , we find that ( 81 ) Hence , approximately , ( 82 ) .
At points directly beneath the sun the value of thus given is 12 .
This calculation applies to years of low solar activity .
At times of solar maximum Table would rise to 17 .
' or 20 . .
Moreover , as we shall see when we come to take self-induction into account ( S26 ) , all these values must be increased by 30 or 40 per cent. , and the probable maximum value of which has to be explained in any theory of the conducting layer must be , approximately , ( 83 ) 25 .
SCHUSTER'S approximate determination of was 3 .
The larger value here obtained accentuates the difficulties in the explanation of the conducting layer which have been mentioned in S21 ; but there seems no reason to suppose that they are insuperable .
SCHUSTER , ' Phil. Trans , vol. 208 , p. 181 .
DR. .
CHAPMAN ON THE SOLAR AND S25 .
The Heating Effects of the Upper Air Currents .
In his second memoir SCHUSTER remarked that a further consequence of the theory outlined for the magnetic variations would be the production of a sensible heating effect by the electric currents circulating in the low pressure conducting layer .
This , it was suggested , might assist in the explanation of the isothermal layer of the atmosphere .
The primary cause of the approximate constancy of temperature in the stratosphere is now well understood , and the suggested heating effect can have only a secondary influence on the phenomenon .
For other reasons , however , it seems desirable to examine numerically this thermal consequence of the theory .
The results prove to be of some interest , and may explain part of the difference between the solar and lunar diurnal magnetic variations .
If the conductivity of the upper atmosphere is small during the night hours , the electric currents in question will flow entirely or mainly in the sunlit hemisphere .
The heating effect is proportional to the square of the current , so that all the harmonics in the current function , whatever their period , contribute to the heating of this one hemisphere .
As the earth revolves , the temperature of a given portion of the conducting layer will begin to increase at sunrise , and the increase will continue till the time of sunset .
During the night hours cooling , mainly by radiation , must take place in order that the average state may remain steady .
The conducting layer will thus suffer a diurnal change of temperature in which the 24-hour term is of much greater magnitude than any sub-component .
This variation , moreover , is purely solar diurnal , including even the part due to the currents which produce the lunar diurnal magnetic variations .
The temperature variation will be confined mainly to the conducting layer so far as conduction alffi convection are concerned ( the kinematic viscosity will be very igh in regions of such low pressure ) .
A corresponding variation of pressure and of motion , not confined to the conducting layer , will result , and this may possibly account for the diurnal oscillation suggested by the magnetic variations .
The phase should be constant throughout the year , though the amplitude would be expected to show a seasonal change .
The most important question regarding these effects is that of their absolute magnitude , however , and we proceed to a rough numerical calculation with this in view .
For simplicity the solar diurnal magnetic variations will alone be though the lunar variations will slightly increase the heating effect , to a degree depending very little on lunar phase .
The principal terms in the solar diurnal magnetic variation potential at the equinoxes ( mean of 1905 and 1902 ) are as follows ( cf. Table ) ( 84 ) .
DIURNAL yARIATIONS OF TERRESTRIAL Only the external variation field is considered here , of course .
The ) terms in the current function are consequently ( cf. 68 ) given by ( 85 ) neglecting the factors , for simplicity .
The energy expended per second in overcoming the resistance to current flow , in the conducting layer of thickness , as given by ( 86 ) In order to avoid excessive calculation , several approximations will be made .
The current function ( 85 ) is such that the currents in the dark hemisphere are small , the various harmonics largely neutralizing one another there , and reinforcing one another in the sun-lit hemisphere .
We shall imagine the three terms of ( 85 ) combined into one , however , which we shall suppose confined to one hemisphere .
Since the sunl of their squares is less than the square of their sum , we shall represent , for this purpose , by the approximation ( 87 ) Also instead of the variable factor we shall use the constant approximation , where is an average value of ( 82 ) over the daylight hemisphere .
adopted value of will or , allowing for the correction due to selfinduction , approximately When these values are substituted in ( 86 ) , and the integration is taken over one hemisphere , the total ture of energy per second is found to be ( 88 ) or , in watts , ( 89 ) The mean value over this hemisphere , per unit area , expressed in gramme-watercentigrade units , is consequently ( 90 ) *Cf .
the table of values of on , ' Phil. Trans , vol. ) VOL. COXVIn .
As the heating of any volume element proceeds continuously for twelve hours , the total thermal energy communicated during the daylight hours ( and lost during the night time ) is ( 91 ) This refers to one square centimetre of the conducting layer of thickness , and it may be noticed that the calculation is independent of and of the situation of the layer .
The pressure variation produced by this temperature change is much more uncertain , since the heating effect depends greatly on the density of the atmosphere of the conducting layer ; also the variation of pressure will be less than that calculated from the equation , on account of the yielding of the adjacent atmospheric layers .
If we suppose that the conducting layer lies between 90 and 140 km .
above the earth 's surface , ibs mass per square centimetre column is ( 92 ) 760 .
2 .
being density of mercury , and the difference of the pressures at top and bottom being approximately 2 .
atmosphere .
The specific heat of air at the temperature of the atmosphere is about ( at constant pressure ) , while that of hydrogen is about .
For the purpose of an approximate calculation we may take the specific heat as unity .
In this case the total rise of temperature which would be produced by the amount of heat ( 91 ) , in the above mass of gas , provided there were no loss , would be in degrees centigrade .
Hence , which must be less than , cannot be so great as 3 . .
This is negligible compared with the estimated amplitude of the pressure variation due to the atmospheric oscillation , which in the upper air is approximately 1/ 760 or ( assuming a surface amplitude of two millimetres of mercury , and a reduction in of about one-half , in the conducting layer , S24 ) .
In order that the pressure variation due to the electric heating of the conducting layer might be comparable with that due to the main atmospheric oscillation the pressure of the region in which the conducting layer is situated would have to be of the order atmosphere .
Assuming the existence of the hydrogen layers mentioned in S21 , this pressure would be attained only at a height of more than 800 km .
The pressure is of the order required for the ionisation of the conducting layer by ultra-violet radiation , according to SWANN'S calculation ; but it seems more probable that the conducting layer is at a lower level .
IURNAL yARIATIONS OF TERRESTRIAL hIAGNETIShl .
S26 .
Discussion of the Phases of the Magnetic Variations .
The explanation of the phases of the magnetic variations is , perhaps , the most difficult part of the present problem .
The data to be reconciled are as follows .
The various annual harmonics in the solar and lunar diurnal magnetic variations are approximately in agreement amongst themselves , the time factors being approximately as below , in the two cases ( 98 ) ( Solar ) , ( 99 ) ( Lunar ) .
Neglecting self-induction and the small terms ( S 22 ) , the theory of S 22 indicates that the velocity potentials of the atmospheric oscillations responsible for these magnetic variations should have the same phase , the negative sign in ( 68 ) and the negative sign of ( cf. 47 , 51 , 68 ) neutralizing one another .
If the simple relation ( 31 ) holds good between the pressure variation and atmospheric velocity potential , the time factors in the pressure variations corresponding to ( 98 ) and ( 99 ) should be ( 100 ) ( Solar ) , ( 101 ) ( Lunar ) .
The factor is here written as 2 , since the fundamental pressure changes appear to be semi-diurnal .
The observed pressure variations at the earth 's surface ( SS 19 , 20 ) have the time factors ( 102 ) ( Solar ) , ( 103 ) ( Lunar ) .
There is consequently no kind of agreement between the observed and calculated pressure variations in the lunar diurnal case .
In the solar case the two variations agree better , but it must be remembered that the observed phase diminishes with height to a considerable extent ( 90 degrees or possibly so that the agreement between ( 100 ) and ( 102 ) would not hold good if the latter had represented the pressure variation as it is supposed to exist in the upper atmosphere , viz. , approximately ( 104 ) ( Solar ) .
( ; 8 .
S. CHAPMAN ON THE SOLAR AND LUNAR The correction to the calculated magnetic variations produced by a given atmomotion , due to self-induction , may be considered at this stage .
In place of a factor , as given by the method of S22 , the factor must be used , where* ( 105 ) being the degree of the harmonic with which the time factor is associated .
If we take 5 .
as a mean value of throughout the day and night , the following values of are yielded by the formula ( 105 ) .
( 106 ) These figures are only rough , since the non-uniform conductivity may considerably affect the theoretical formula ( 105 ) .
It may be noticed that the lag of phase , owing to self-induction , increases with the frequency of the harmonic , so that self-induction cannot explain the apparent increase of phase with frequency which is observed in the solar diurnal magnetic variation ( Table F ) .
Self-induction will also diminish the amplitudes of the magnetic variations by a factor .
This will not greatly affect the relative magnitudes of the various magnetic harmonics , but it will affect our estimate of the magnitude of in S24 , and of the heating effects in S25 .
The calculated value of would appear to be increased on this account by 30 or 40 per cent. , and this , again , would slightly increase the above values of Taking into account the above phase changes due to self-induction , the calculated time factors in the pressure variations are modified as follows:\mdash ; ( 107 ) ( Solar ) , ( 108 ) ( Lunar ) .
When these.are compared with ( 104 ) and ( 103 ) , the " " calculated -observed \ldquo ; phase differences are found to be ( 109 ) ( Solar ) , ( Lunar ) These values almost suggest that a mistake in sign has crept into the calculation of the magnetic from the pressure variations , but the signs have been carefully .
MAXWELL'S 'Electricity and Magnetism , ' II .
, S672 , or SCHUSTER , 'Phil .
Trans , vol. 208 , p. 172 .
S27 .
The Residual Variations , the Terms not Dependent SOlely on Local Time .
Our discussion of the magnetic variations has so far related entirely to the simple analytical representation of the observed data which has been described in SS10 , 13 .
In this representation the only terms considered were those dependent on local time .
It remains , therefore , to discuss the residuals in Tables III .
and , and to examine how far they are to be attributed to the presence of variations not depending solely on local time ( cf. S22 ) .
In order to abbreviate this investigation , certain general features exhibited by the residuals will be described without setting out the detailed figures .
In the first place , the mean residuals for any of the nin groups of observatories are generally similar for the years 1905 and 1902 , and , in the case of the " " annual\ldquo ; residuals in Tables III .
( a ) and III .
( ) , for the equinoxes and solstices .
They are , however , greater for 1905 than for 1902 , and greater at the equinoxes than at the solstices .
If all the nine group mean residuals from any Table are combined numerically ( counting all signs positive ) , the ratio of the 1905 and 1902 sums , or of the equinoctial and solstitial sums , can readily be determined .
The former ratio is rather less than those of Table , being approximately , for the annual\ldquo ; residuals .
The increase , such as it is , conffims the view that the residuaIs are a real part of the phenomenon , and do not merely represent accidental errors of observations .
The ratio of the equinoctial to .
solstitial " " annual\ldquo ; residuals is greater , being about is shown in the case of all three magnetic elements , and four periodic components , the separate mean ratios for these being and .
These increases roughly correspond to those shown by the harmonics already discussed .
The group-mean residuals , in the mean of equinox and solstice and of 1905 and 1902 , taken from Tables III .
and ( ) , are collected in Table T. The " " seasonal\ldquo ; residuals from Tables III .
and will not be considered .
The largest residuaIs in Table occur in the column relating to th.e24-hour component variation of North force .
The corresponding residuals for the West force are small and may well represent merely local peculiarities at the various observatories .
If the variations indicated by the North force residuals have a potential of simple form , this must be of type , since this is the only type which yields North force terms without contributing also to the West force variations .
In S22 it was shown that the inclination of the magnet to the geographical axis of the earth could give rise in current function to the variation ( 110 ) ( among others ) , and this is of the above type .
The chief coefficient is ' ( cf. ( 67 ) ) .
which is of the order .
The longitude is measured from the meridian West of Greenwich , while is approximately .
Assigning to the value 3 , and to and the values 250 degrees and , deduced roughly from the mean solar semi-diurnal harmonic , ( 110 ) becomes ( 111 ) .
The corresponding term in the magnetic variation is .
The North force variation deducible from this is the unit being C.G.S. This is clearly far too small to account for the residuals referred to above , and , moreover , it is found that the signs do not agree at all consistently with those in the North force column of residuals .
This is owing to the factors and , which vary considerably from group to group for the observatories here dealt with .
The same difficulty is met with in regard to the 12-hour North force residuals ; here also , moreover , the amplitude of the " " longitude\ldquo ; harmonic ( in this case the principal one is is too small to explain the observed residuals .
It may be noted that the magnetic variations depending on longitude would theoretically be smaller if , as suggested in S23 , the main oscillation responsible for the magnetic variations be semi-diurnal , than on .
the hypothesis considered by SCHUSTER ; they are therefore less likely to serve as a check on the .
theory .
I have tried to represent the residuals of Table by harmonics depending on the time of some standard meridian , but with little success , and I am inclined to think that they depend on local time so far as they are not merely irregularities peculiar to particular observatories .
The latter can hardly be the case with regard , at any rate , to the 24-hour North force residuals ; in other cases the position is much less clear .
These North force residuals seem to present a difficult problem , since they apparently cannot be represented by any simple potential function .
It may be that the amplitude of the semi-diurnal pressure variation seems to vary with latitude according to the law , instead of ( or ) as we have supposed ; the true law could only be represented by the introduction of other harmonics besides into our theoletical calculations .
Another fact worth noting is that the South component of the semi-diurnal variation of wind velocity ( at St. Helena ) is markedly larger than that calculated from GOLD'S theory* of the wind , barometric and temperature variations , while the East component seems to be in agreement with theory .
Such .
ante ( S19 ) .

