]\gt ; [ 299 ] VI .
A Discussion of Atmospheric Electric Potential Results at Kew from Selected Days during the Seven Years 1898 to 1904 .
C. CHREE , , LL. D. , F.R. ( From the Physical ) Received February 21 , \mdash ; Read March 15 , 1906 .
CONTENTS .
Sections Page 1-3 .
Nature of observational data . . . . .
299 4 .
Mean monthly and annual values of the potential gradient . . . . .
301 5 .
Non-cyclic effect . . . . .
303 6-7 .
Diurnal inequality . . . . .
305 8 .
Analysis of inequality in Fourier series . . . . .
309 9 .
Annual variation eries ) . . . . . .
311 10 .
Comparison with evious results for Kew . . . . . .
312 11-15 .
Comparison with CHAUVEAU'S results for other stations . . . . .
313 16 .
Irregular diurnal changes . . . . .
319 17 .
Special meteorological features of the selected days . . . . . . . . .
320 18-25 .
onnection of atmospheric potential with ological elements . . . . .
322 26-32 .
Appendix.\mdash ; On the relation between the regular diurnal changes of barometric pressure and potential gradient . . . . .
329 S 1 .
IN 1894 special observations were commenced with the object of determining the influence of its position on the records from the water-dropping electrograph , and also of ascertaining whether any alterations were desirable .
The results were discussed in a paper*communicated to the Society The experiments led certain changes in the apparatus , and an auxiliary system of eye observations was introduced with the object of obtaining quantitative results possessing a definite significance .
The original water tank was replaced by another of much larger diameter and small depth .
The new tank is situated on the wooden roof of a detached portion of the north-west room on the ground floor , which contains the recording part of the electrograph .
The wire connecting the water-dropping tube to the electrograph is * .
Soc. Proc vol. 60 , 1896 , pp. 96-132 .
( 407 .
) 30.6.06 much shorter than under the old arrangement , thus reducing the risk of defective insulation .
Again , the down tube from the tank to the horizontal discharge tube is about feet long , instead of a few inches as formerly , so that the pressure under which the jet issues is practically constant .
Lastly , the jet is emitted parallel to the wall of the building , instead ofperpendicular to it as before .
This tends to increased uniformity in the conditions , as the potential varies rapidly in the direction perpendicular to the building , but little , if at all , in a horizontal direction taken parallel to it .
The discharge tube , which projects from a window in the west wall , is situated about 10 feet to the north of the old tube , but at the same level above the ground .
A flower border used to run the whole length of the west ] , containing shrubs 3 or 4 feet high .
The northern half of this border was cleared of shrubs and laid down in grass , which has been kept short the whole year round .
These changes were made with the object of securing that no appreciable fictitious element should be introduced into the annual change of potential gradient .
A definite test of the insulation of the apparatus was introduced , and is applied each morning .
It is not always possible to improye defective insulation immediately when discovered , but at the worst the test shows when a day 's record is to be regarded with suspicion .
The surface of the ground outside the building is not absolutely level , so the height of the spot where the jet breaks into drops can hardly be given with mathematical precision .
It is very approximately 11 feet ( metres ) .
The perpendicular distance from the general outline of the west wall is inches centims 8 .
2 .
If these distances , the environment , and the insulation of the apparatus could be kept absolutely uniform , the potential recorded on the electrograms would only require multiplication by an absolute constant to become converted into potential gradient per metre of height in the open .
This ideal is not , however , practically attainable , and experience showed the desirability of obtaining a multiplier by reference to absolute eye observations made regularly in the open .
The eye observations are taken with a Kelvin portable electrometer , supported on a small level platform having three vertically projecting points , which prevent the electrometer from slipping .
The platform is carried on a brass rod , fitting inside a hollow brass tube sunk in the .
The rod can be raised in the tube and clamped at any convenient height .
The original idea was to observe at two heights , a given distance apart .
Having regard , however , to time and other considerations , it has beenjudged best not to vary the level , and the electrometer fuse has been used in a horizontal position at a fixed height above the ground .
Thus the potential observed has referred to a fixed level , approximately metres above the general level of the ground in the immediate vicinity .
The ground is practically flat for a considerable space round the stand .
The adjacent grass is kept short , and there is no tree , shrub , or building within what seems a reasonably satisfactory distance .
FROM SELE TED DAYS DURING THE SEVEN YEARS 1898 TO 1904 .
If be the potential shown by the portable electrometer , and that shown by the electrograph at the same instant , ( 1465 P2 ) is a value for the factor by which the potential shown by the electrograph is to be multiplied to collvert curve potentials into volts per metre of height in the open .
Observations are taken with the portable eIectrometer nearly every day when the weather allows , and a mean value for the factor is calculated from each month 's comparisons .
S3 .
When potential changes are very sudden , readings from the electrograph are doubtless partly dependent on its electrical capacity , and when the changes are very large , part of the trace is pretty sure to be lost .
These Iarge and sudden changes are mostly confined to times of rainfall , being specially prominent during thunderstorms .
Further , during rain , drops are falling from all parts of the efflux tube , so that its record may not be solely , or possibly even mainly , determined by the jet at the end .
For these and other reasons , it was decided to confine the tabulation of the electrograms to days when there was no , and to be content with 10 days a month .
In selecting the 10 days one aimed at distributing them throughout the montl ] , avoiding days in which negative potential was recorded , or there was reason to suspect the insulation .
In some months there was a considerable excess of days to chooss from , but in others it was difficuIt , and in a few cases impossible , to get as many as 10 suitable days .
In one very wet month , in order to secure sven 8 days , it was necessary , in two or three instances , to take in place of the natural day successive periods of 24 hours , comprising parts of two successive days .
With this exception , day always meant 24 hours , extending from midnight to midnight , G.M.T. All the data , except in Table IX .
, refer to Greenwich time , but the difference from this of local time is only minutes .
Since 1902 the Annual Reports of the Observatory Department of the National Physical Laboratory have contained tables showing the mean monthly diurnal variation of the potential gradient from the selected days .
Potential gradient is an exceptionaJly fluctuating element , and measurements of the actual photographic trace , taken at exact hours , may be far from representative of the mean value during 60 consecutive minutes .
The practice adopted has been to draw in pencil a continuous curve , following as nearly as the eye can judge the general trend of the trace , and to measure the ordinates of this curve at exact hours , G.M.T. During the past two years , the data for earlier years have been similarly treated , and the present paper deals with the results from the seven years 1898 to 1904 .
Mean MIonthly Vcdues of the Potential Gradient .
S4 .
Table I. gives the mean value of ( the potential gradient in the open in voIts per metre ) for eacb individual month and for each year , and likewise the mean annual inequality based on the seven years ' results .
The annual inequality is exhibited graphically in fig. 1 by the continuous line curve ; the ordinates represent the mean values of for the individual months , expressed as fractions of their arithmetic mean from the 12 months .
TABLE I.\mdash ; Potential dient ( Volts per Metre ) .
Table I. shows at once how variable an element is .
It i@ at least doubtful whether even seven years are sufficient to give an annual inequality which can be relied on as fully representative in its details .
There is unquestionably a well-marked minimum near midsummer .
There seems also to be a second minimum in January , but whether this would be shown\mdash ; or if shown would be as prominent\mdash ; if we had 70 years instead of 7 is open to doubt .
The fact that the largest mean monthly value did not occur in January in any one of the seven years , whilst occurring four times in December , is so far confirmatory .
In November , December , and January , however , fog is a very uncertain while very important influence .
During thick fog in winter is usually abnormally high , 400 to 600 practically double the ordinary gradient\mdash ; being not infrequently met with .
Thus in winter the mean value of varies much from day to day even in fair weather .
For instance , in December , 1898 , the last three of the selected days gave a mean value for of 380 volts , whilst the corresponding mean from the first seven days only slightly exceeded 100 volts .
So , again , in January , 1899 , the mean values of from the first five and the last five of the selected days were roughly 160 and 400 volts respectively .
The figures in Table I. itself , e.g. , those for December in the years 1899 and 1901 , afford examples almost as striking .
The mean value for the year appears lower for the three consecutive years 1900 to 1902\mdash ; which cover the period of sun-spot minimum\mdash ; than for the others .
As this phenomenon also appears when we confine ourselves to the six summer months , April FROM SELECTED DAYS DURING THE SEVEN yEARS 1898 TO 1904 .
to September , when climatic influences are comparatively uniform , it be a real one .
If so , tends to be lower than usual near sun-spot minimum .
The difference shown , however , in Table I. may be accidental ; it is certainly too small to rely on , considering the limited number of years and the many sources of uncertainty .
JAN .
FEB .
MARCH APRIL lIAY JUNE JULy AUG . .
OCT .
NOV .
DEC .
JAII .
Fig. 1 .
\mdash ; Mean monthly values of potential gradient .
( Daily range ) ( mean monthly valus ) .
-Sum 24 hourly differences potential gradient .
-Sum 24 hourly differences barometric pressure .
Non-cyclic S5 .
In meteorological elements such as temperature , where there is a conspicuous annual variation , there is , at certain seasons of the year , a steady tendency to increase , and at other seasons to a decrease , during the 24 hours .
In such a case the value assigned to midnight in the mean data for a particular month varies slightly according as one makes midnight the first or the last hour of the day .
When , instead of all the days in the month , one takes , a limited number of a particular kind , a non-cyclic , or a-periodic , element may present itself in the diurnal inequality , though there is no such element present to an appreciable extent in.the average day of the month .
This phenomenon , for instance , as I have pointed out , is conspicuous in the horizontal component of the earth 's magnetic field on magnetically ' ' quiet\ldquo ; days .
There are two courses open in such a case , viz. , to give the mean hourly values taken direct from the observations , with data for both midnights , or else to eliminate the non-cyclic element from the diurnal variation while stating its amount .
The latter course has been adopted here .
The size of the algebraic mean non-cyclic element for the selected days in individual months is recorded in Table II .
, in volts per metre .
Table II .
shows on the whole a rise of throughout the average selected day , but the mean value of the non-cyclic element is very small , and its sign even varies from year to year .
Coming to individual months , we see that in May the non-cyclic effect was negative in six out of the seven years ; but in three years , 1900 , 1901 , and 1903 , there were more individual days in which it was plus than minus .
The only months in which there seems a decided tendency for to rise during the 24 hours of a selected day are December and January .
In the former month 42 days out of 68 , and in the latter month 38 days out of 65 , showed a rise .
In December there was an excess of days showing a plus effect in six years out of the seven .
TABLE II.\mdash ; Non-cyclic Effect ( Volts per Metre ) .
FROM CTED DAYS DURIN ' SEVEN YEARS 1@98 TO 1904 . .
Table III .
shows the mean diurnal inequality\mdash ; after elimination of the noncyclic effect\mdash ; for the individual months and the year , and also for three seaso1ls , , Winter ( November to February ) , Equinox ( March , April , September and October ) and Surnmer May to .
Each hourly value depends on approximately 70 readings , and as these were taken to at least the nearest 10 , sometimes the nearest 5 volts , the table gives the results to the nearest volt .
Distinct maxima and minima appear in heavy type .
The in the penultimate column represent the excess of the largest mean hourly value over the smallest .
The last column gives the sum of the 24 ences from the meau the day to the nearest volt .
The diurnal inequalities for the individual months and the year are illustrated by the curves in fig. 2 which are drawn to a common scale .
They pass through all the observational points and thus enable the reader to judge for himself of the of the data .
The curve for the year and those for a majority of the months , notably April , June , August and October , possess a smoothness which leaves little to be desired .
A few isolated observational points catch the eye , e.g. , 1 .
in January and in September .
I have satisfied myself that the cause is not arithmetical , but if the number of year 's data were largely increased the outstanding values for individual hours would very probably disappear .
In the four months November to February there is a slight flattening or depression of the curve shortly after the evening maximum which looks hardly natural .
Whether this would disappear if a larger number of year 's data were employed is , however , more doubtful .
It seems to arise from a slight variability in the hour of maximum in the same month of different years .
The double daily period , though most prominent in summer , is shown clearly every instance .
December is the month showing the nearest approach to a single maximum and minimum , November and January coming next ; but even in these three months there is not the slightest doubt as to the existence of an afternoon minimum , while the forenoon maximum exceeds that of the late afternoon in January and February , and is practically equal to it in .
The early morning minimum is the principal one only in the four midwinter months and in August and September .
The equinoctial months show the afternoon minimum just about as prominently as the midsummer The morning minimum occurs near 4 , its time varying very little throughout the year .
The afternoon minimum is only a little more variable , its time of occurrence fallin usually near 2 .
The hours of maxima are ather more variable , the day interval between the two being less in the winter than in the summer months .
When , as in the present case , there is a well-marked double maximum and minimum , the amplitude of the range is probably not so satisfactory a measure of the activity of VOL. CCVI.\mdash ; A. 2 FROM CTED DAYS DURING THE SEVEN 1898 TO 1904 .
Fig. 2 .
the forces to which the diurnal inequality is due as is the size of the 24 hourly differences from the mean .
Both these quantities , as recorded in Table III .
, agree in representing a maximum of activity in February and a minimum in November , though the latter month differs but little from June , July , August and September .
The annual variation of the sum of the 24 differences is represented by the heavy broken line in fig. 1 , unity representing the mean of the 12 monthly values .
S 7 .
In view of theories.as to the cause of the diurnal variation , interest attaches to the comparative size of the potential changes during the day and the night .
The forenoon maximum and afternoon minimum occur at Kew when the sun is above the horizon , the afternoon maximum and the morning minimum when the sun is below the horizon .
I have thus taken the difference between the first-mentioned maximum and minimum as a measure of the change of potential during the day , and the difference between the other maximum and minimum as a measure of the change of potential during the night , terming the first the Day fall the latter the " " Night fall TABLE IV .
ralues , volts Table IV .
shows the day and night falls from the diurnal inequalities of the several months , and also the ratio borne by these two quantities and by the daily range to the mean nlonthly value of P. Considered absolutely , the day fall has a FROM SELECTED DAYS DURING SEyEN YEARS 1898 TO 1904 .
conspicuous minimum in the three lnonths November to January , the night fall a conspicuous maximum from December to March .
The rular causes , whatever they are , to which the day and night falls are due , at Kew nearly equal in intensity in the average month both of equinox and summer .
The daily range when considered absolutely is , as we have already seen , greater in winter than summer ; but relative to the mean value of for the day , the day and night and the range all show a decided minimum in November , December and January .
If the potential gradient is due to a negative charge on the earth , the density of that charge is on the ayerage greater in winter than in summer , but the diminished charge of summer is , relatively considered , more influenced by the causes to which the regular diurnal inequality is due .
The figures in the last column of Table IV .
are shown graphically in the dotted curve of fig. 1 .
Analysis of Inequality Fourier Series .
S 8 .
The diurnal inequalities in Table III .
have been analysed in Fourier series , where represents time counted from midnight ( G.M.T. ) , one hour being taken as equivalent to The amplitudes , \amp ; c. , and the phase angles , \amp ; c. , for the individual months , the year , and the three seasons are shown in Table V. In the case of the amplitudes , unity represents , as before , a potential gradient of one volt per metre .
The variation of and throughout the year is illustrated by the heavier full and dotted line curves respectively of fig. .
The ordinates represent the ratios borne by the monthly values to their arithmetic mean .
The annual variation of is of a most unusual character .
The mid-winter maximum and mid-summer minimum are extraordinarily pronounced .
The phase angle also shows a remarkable annual variation , the times of maximum in " " summer\ldquo ; and " " winter\ldquo ; differing by nearly 8 hours .
In July and December the 24-hour waves are nearly in opposite phases .
The 12-hour term presents very different features .
Its amplitude , , is slightly less for winter than for summer , and there are maxima at the equinoxes , but the annual variation is comparatively small .
The variation shown by the phase angle is also comparatively small , the difference between the times of maximum in winter and equinox being little over half an hour .
The 8-hour and 6-hour terms are so small that considerable uncertainty must attach to the results for individual months .
Still , there is a regularity about the figures , especially those for the 8-hour term , which seems to justify our acceptance of at least the general features .
The annual variation of is somewhat similar to JAN .
FEB .
MARCH APRIL NAY JUNE JULY AUG .
SEPT .
OCT .
DEC .
JAN .
Fig. 3 .
mean of the twelve monthly values .
\mdash ; Annual variation of potential gradient .
barometric pressure .
, , , , potential gradient .
, , , , barometric pressure .
SELECTED DAYS DURING THE SEVEN YEARS 1898 TO 1904 .
that of , the values for the winter months being conspicuously greater than those for the summer months .
There is also a fairly systematic annual variation in the phase angle ; but this angle is largest at the season when is least , and conversely .
The 6-hour term shows an annual variation similar on the whole to that exhibited by the 12-hour term .
In equinox and summer -hour wave seems fully larger than the 8-hour wave .
TABLE V.\mdash ; Diurnal Inequality .
Amplitudes and Phase Angles .
Annual Variation .
Fourier Series .
S9 .
The anllual variation may be represented by a series of the type where is the arithmetic mean of the 12 monthly values , is the time counted from the beginning of the year , 1 month being taken as equivalent to , while , , , are constants .
Table V for a variety of elements the values of the amplitudes and phase angles of the yearly and half-yearly terms , and the ratios borne by these amplitudes to the mean value of the element and to one another .
As before , in the amplitudes unity represents 1 volt per metre .
The angles are given only to the nearest degree .
The representation of the annual variation by two terms of the series is , as a matter of fact , not very close in any one of the elements .
This may mean that other shorter period terms are by no means negligible , or merely that a larger number of year 's data are necessary for high precision .
DR. C. CHREE : ATMOSPHERIC ELECTRIC POTENTIAL RESULTS AT KEW The annual term is , in all cases , the more important of the two .
The relative importance of the semi-annual term is much greater for the day fall than for the other elements ; it is least for the mean value for the day .
TABLE .\mdash ; Annual Variation .
Amplitudes and Phase Angles .
Coraparison with Previous Results for S 10 .
In the ' Phil. Trans. ' for 1868 , pp. 347-361 , atmospheric electricity data from Kew for a 2-year period June , were discussed by Professor J. D. EVERETT .
The data were from all days of the year .
EVERETT calculated Fourier coefficients corresponding to our for the mean diurnal inequality of each individual month , and for each of the two , and then for the 12 months and the year from the two year 's data combined .
His amplitudes were expressed in terms of an arbitrary unit .
The results for corresponding months of the two years vary somewhat largely , and there is considerable difference between the mean results for the two years .
Thus even the values from the two year 's data combined , at least those for individual months of the yea , are clearly exposed to rather ' large uncertainties .
It should be noticed that EVERETT'S hour angles ( loc. cit. , p. 358 ) are counted from noon .
Thus , in comparing his results with mine , must be added to his ( my ) ; his and my are immediately comparable .
The following remarks apply to his values for the two years conbined : My value for is above EVERETT'S from ] September to February , the mean excess being , an below his from March to August , the mean deficiency l0 ' .
In the mean diurnal inequality for the year , however , EvERETT'S value for exceeds mine by only 1o , corresponding to minutes of time .
As , my value is above EvERETT'S in six months , , May and July to November , the mean excess being , and below his in the other six months , the mean deficiency being In the mean diurnal inequality for the year his value of exceeds mine by answering to minutes of time .
The differences in the phase angles in the two inyestigations , though atively small , especially in case of the 12-hour FROM SELECTED RING THE SEVEN EARS 1898 TO 1904 .
term , are too systematic to be ascribed to chance , but several explanations are possible .
EVERETT'S table on his p. 358 agrees with my Table V. in making ( his ) greatest in December and January , and least in July and August ; and his difference between the maximum and minimum values is even greater than mine .
Also his largest values for ( his ) occur , like mine , in February and March , his least values occurring in May , June , and November .
There is thus a fairly close resemblance between our results for the amplitudes .
There is , however , a important difference as regards the relative and .
We have , in short , EVERETT ( 1862-4 ) .
( 1898-1904 ) .
0 .
1.75 0 .
2.95 Some further aspects of this difference are considered later in S13 .
EVERETT also expressed the annual variation in the mean daily potential gradient as a Fourier series with almual and semi-annual terms .
His times were measured from the middle of January , so that his phase angles ( and in his notation ) have to be diminished accordingly for comparison with min .
His results from the two years combined are in the notation of Table VI .
, counting time from the beginning of the year , EVERETT'S results for the annual term are thus fairly similar to mine , but he makes the amplitude of the semi-annual term even less than I do , and his value for its phase angle is quite different from mine .
This is not , however , very surprising , as the phase angles for his two years taken separately differ by , and the amplitude for the second year is only of that for the first .
Also inspection of EVERETT'S annual variation curve in his plate XXI .
shows that its representation by only annual and semi-annual terms must be very rough .
Comparison with CHAUVEAU'S Results for other S 11 .
The intercomparison of results from different stations is difficult owing to uncertainty as to the real significance of the units employed .
The length and the height above ground of the discharge tubes of water droppers vary greatly .
Some are in low , others in lofty buildings , so that even if all were of one length and at one height they would not give comparable results .
To meet this difficulty so far as possible , in the article on " " Atmospheric Electricity\ldquo ; CCVI .
I4 .
C. : ATMOSPHERIC ELECTRIC TS AT the 10th edition of the ' Encyclopaedia Britannica , ' I expressed each hourly value in a diurnal inequality as a percentage of the mean value for the day .
The principal maximum and minimum , for example , in EVERETT'S mean diurnal inequality for the year at Kew were given as 139 and 71 , and thence the range would be .
In an important and very comprehensive paper , .
A. B. CHAUVEAU has employed a method of representing the ranges which differs from the aboye only in taking the mean of the 24 hourly values as 1 instead of 100 .
He compares in particular what he calls day and night ranges mphtude diurne , mplitude octurne ) .
For both ranges he takes the late evening maximum , so that amplitude diurnenight maximum less afternoon minimum , amplitude nocturnenight maximum less early morning minimum .
He did so because in all the cases he considered , with the single exception of Lyons , vening maximum exceeded the forenoon maximum throughout the year .
Finding that at most places the diurnal inequality was nearly the same for May , June , July and August , CHAUyEAU obtains " " summer\ldquo ; values for his day and night ranges from a diurnal inequality based on days selected from these four months .
For corresponding winter values he usually confines himself to days selected from December and January only .
In dealing , howeyer , with the Bureau Central Meteorologique in Paris\mdash ; the station which he considers most fully\mdash ; his winter includes as well days from late Novembel and early February .
CHAUVEAU also records the ratio borne by the mean value of from the selected winter days to the corresponding summer value .
In most cases he confines himself to fine weather days .
Table VTI .
summarises the chief results given by CHAUVEAU , with the addition of corresponding results for Kew from Tables I and III .
CHAUyEAU apparently regards the conditions at the Bureau Central as having somewhat deteriorated since 1893 through the growth of trees .
As against this , the data there for 1893-4 depend on a somewhat limited number of days , and any single yea , may not be fully representatiye .
The site at Perpignan is considered by CHAUyLAU amongst the best .
As to Kew and Greenwich , I shall quote CHAUVEAU'S own words ( loc. cit. p. ) .
" " La modification progressive de la variation d'ete dans les observations de M. EVERETT , de M. WHIPPLE et dans celles de Greenwich nous parait etre en rapport ayec des conditions de moins en moins bonnes dans l'installation du collecteur results at Greenwich and at the College de France show a rather striking ilarity which CHAUVEAU notes , but he regards them both as abnormal in several respects .
A meaIl from the two sets of results for the Bureau Central presents fairly similar 'Annalcs Central Me'teorologique de France , ' Annes , 1900 , I Memoires , pp. to FROM SELECTED DAYS DURING THE SEVEN YEARS 1898 TO 1904 .
.315 features to the latest Kew results , though both the day and night ranges appear to be larger at the former station .
The Kew summer data are from the diurnal inequality in Table III .
for the four summer months combined .
If the quantities had been found for each of the months independently , the arithmetic means would have slightly exceeded the values in the table , viz. , amplitude diurne , amplitude nocturne .
The Kew results for winter are arithmetic means from December , January and February treated independently .
omittin February , the values would be less , , amplitude diurne , amplitude nocturne As regards the different Kew data , it should be taken into account that BIRT'S were obtained with the old Ronalds apparatus , the point at which the potential was measured being some 16 feet above the dome and more than 80 feet above the general level of .
Also the observations were from the even hours only , which inevitably leads to smaller ranges than when hourly observations are taken .
BIRT grouped his data in several classes .
That taken by included only low positive potentials , i.e. , exCluded all cases where the potential was negative or exceptionally high .
EVERETT'S data included all readings positive or negative , WHIPPLE'S all but days of negative potential .
TABLE VII.\mdash ; Abstract of CHAUVEAU'S Results and Comparison .
S 12 .
CHAUVEAU also analyses the several sets of data for the diurnal inequalities at the Bureau Central the Eiffel Tower in Fourier series .
He takes five terms , one more than I have taken for Kew , and says that the additional , or -hour term , distinotly improves the agreement with observation .
I have expressed CHAUVEAU'S values for the amplitudes in terms of the corresponding mean values of taken as unity , and have calculated corresponding results for Kew from EVERETT'S data as well as my own .
In the latter case the values assigned to " " summer\ldquo ; and ' ' winter\ldquo ; are derived from the seasonal data at the foot of Table V. , making use of the mean values of given in Table I. For September and October I have taken the arithmetic means of the 's and 's for the two months .
It seemed hardly worth while analysing a mean diurnal inequality from the two months combined , though that would have been more strictly correct .
In the case of EVERETT'S data the values refer in each case to a mean diurnal inequality for the period stated .
Table VIII .
gives the values thus found for the amplitudes , and Table IX .
applies similarly to the phase angles .
In this table the Kew angles have been modified so as to answer to local solar time , so as to be strictly parallel to CHAUVEAU'S angles .
TABLE VIII.\mdash ; Diurnal Inequality .
Comparison of Amplitudes .
TABLE IX.\mdash ; Diurnal Inequality .
Comparison of Phase Angles .
FROM SELECTED DAYS DURING THE SEVEN 1898 TO 1904 .
.317 S13 .
It will be convenient to discuss the two tables together , and it must be remembered that amplitude now means the ratio borne by the actual amplitude of the periodic term to the mean value of the element during the day .
Comparing corresponding results from the Bureau Central and the Eiffel Tower , we infer that the amplitude of the 24-hour term yaries but little , if at all , with moderate changes of height above the ground .
Its value , however , at both places was much lower in the later series of years than in 1893 .
CHAUVEAU seems disposed to attribute this diminution to the fact that the efflux tube had been shortened , the reduction in the amplitude representing the increased disturbing influence of the building and other surroundings .
A similar phenomenon , hoWever , is seen at Kew in comparing the older and more recent data , and in an enhanced degree especially in winter and autumn , and at Kew the efflux tube was a foot longer in the later riod than in the earlier .
The phase angles in the 24-hour term at Kew and the Bureau Central both vary largely with the season , but are fairly similar at corresponding seasons , and the values from the earlier and later series of years are in fairly close agreement , notwithstanding the great differences in the amplitudes .
On the other hand , the similarity in amplitude at the Bureau Central and the Eiffel Tower is accompanied by a large difference in phase , at least in summer .
CHAUVEAU infers , the similarity of the summer and autumn values of at the Eiffel Tower to the winter value at the Bureau Central , that the large seasonal change in at the latter station is a phenomenon confined to the lowest strata of the atmosphere .
Experimental evidence is , however , necessary to justify the acceptance of this conclusion .
The 12-hour term shows very different phenomena .
The amplitude apparently diminishes rapidly with the height , while the phase is comparatively little affected .
Further , the phase angle at the Bureau Central , as at Kew , varies but little with the season , while both amplitude and phase angle vary but little with the epoch from which the data are derived .
The phase angles at Kew and the Bureau Central do not differ much , but the angles at the former station appear consistently the smaller .
The maxima at Kew are thus somewhat later in the day than at the Bureau Central , the difference in time from the mean of the three seasons in the latest sets of data being about half an hour .
S 14 .
In consequence apparently of the rapid diminution in its amplitude with height CHAItVEAU seems disposed to regard the 12-hour term as of subsidiary importance , and due to special conditions in the very lowest strata , if not to purely local causes .
A somewhat similar view has been expressed by Professor F. xNER , who bases his conclusions largely on observations he has made in th .
tropics .
At some places , e.g. , in Ceylon , he seems to have found but little trace of the ordinary double maximum and minimum in the diurnal inequality .
EXNER attributes the double maximum and minimum to a thin atmospheric layer near the ground , which *Cf .
' Terrestrial Magnetism , ' vol. 7 , 1902 , p. 89 .
exerts , he says , at some stations a large effect in the early afternoon hours through absorbing the sun 's radiation of shortest wave-length .
This absorbsnt stratum he believes especially characteristic of arid and dusty regions , while comparatively non- existent in moist climates , or where foliage is luxuriant .
The conditions at Kew\mdash ; a station situated in an extensive grass-covered park\mdash ; are hardly such as EXNER associates with a large 12-hour wave .
It is possible , of course , that the large extension of London since 1864 may have profoundly altered the electrical conditions even as far west as Kew ; but if so , one would expect an enormous difference to manifest itself when the direction of the wind changes from east to west .
I am not prepared to say that such a change of direction is wholly without effect\mdash ; as it is difficult to distinguish between the direct effect of change of wind direction and the indirect effect of the change of weather conditions accompanying it\mdash ; but if it exists it is by no means consicuouq .
Also the difference between the earlier and later Kew data is not an increase in the 12-hour wave , but a diminution in the 24-hour wave .
Faulty conditions or local peculiarities seem more likely to introduce or enhance fictitious waves than to wipe out real periodic terms .
The observed 24-hour wave , it must be remembered , shows a phase angle which varies largely with the season of the year , but the differences between the phase angles observed in 1862-4 and 1898-1904 are in most months comparatively small .
If , then , the reduction in the amplitude apparent in the later series is due to some disturbing cause , that cause must have itself a phase which nearly keeps step with that of the real 24-hour potential wave throughout the year .
Before attaching much importance to views which after all are mainly theoretical , it should be remembered that the older Kew data refer to days , the later only to fair weather conditions .
Future investigations may show the 24-hour term to be more dependent on the meteorological conditions than the 12-hour term .
Also in considering the relative importance of the two terms we should make allowance for the fact that it is the conditions close to the ground that are at present of most immediate practical importance both to human beings and to agriculture .
S 15 .
Returning to the consideration of Tables VIII .
and IX .
, it will be seen that with the exception of the values of in winter\mdash ; when the amplitude at Kew was so small that considerable uncertainty must exist\mdash ; the Kew and Bureau Central values of and are sufficiently close to raise a strong presumption that the 8-and 6-hour terms represent a true atmospheric electricity effect .
The comparative agreement between the phase angles from the earlier and later series of observations at the Bureau Central and the Eiffel Tower favour the same conclusion .
The 8-and 6-hour waves are decidedly smaller at Kew than at the Bureau Central .
The difference in this case may not improbably be partly due to greater homogeneity in the Kew data .
The fact that the 8-and 6-hour telms are also much less at the Eiffel Tower than at the Bureau Central must , however , have some other explanation 2 it is most natural to ascribe it to the influence of height .
RONI SELECTED DAYS DURING SEVEN YEARS 1898 TO 1904 .
Irregular Diurnal Changes .
S 16 .
Hitherto we have exclusiyely considered the regular diurnal changes , those for each month being derived by taking hourly means from about 70 days and eliminating any non-cyclic element .
The amplitude of these periodic changes gives , however , a very inadequate idea of the phenomena .
Even in the finest weather the regular diurnal change is accompanied by numerous irregulal changes , and very often the latter so predominate that the former is but inconspicuously shown in the photographic trace .
I have thus gone into the individual day 's results with some minuteness .
My object was chiefly to ascertain the effect of different meteoroIogical conditions\mdash ; a subject considered later\mdash ; but also partly to see whether any sort of proportionality existed 1oetween the size of the regular and irregular movements throughout the .
The results of this latter inyestigation are summarised in Table X. TABLE X.\mdash ; Results from Indiyidual Days .
The first numerical column of Table X. gives the monthly and seasonal mean values of the arithmetic mean of the largest and smallest hourly values in individual days ; the fourth column shows the ratio which the value thus found for the month or season bears to the corresponding mean value of P. The second column gives the monthly and seasonal mean values of the excess of the largest over the smallest hourly value in individual days , the non-cyclic effect not being eliminated .
The fifth column shows the ratio borne by the mean range thus found for the month or season to the corresponding mean value of P. The third and sixth columns similarly relate to the non-cyclic effects treated bumerically .
The unit in the first three columns is , as in Tables I. to VI .
, 1 volt per metre .
The last column gives the ratios borne by the ranges in the second column to the corresponding amplitudes of the regular diurnal inequality as given in Table III .
The monthly means in the first three columns of Table X. were really calculated to one significant figure beyond that recorded , and these more exact results were employed in deducing the seasonal means and the ratios .
Several interesting and novel results appear .
The mean of the largest and smallest hourly values exceeds in every case the true mean from the 24 hours ; but the excess not , and , though slightly less in winter than in the other seasons , varies comparatively little throughout the year .
The mean of the individual daily ranges is , on the average , no less than times the amplitude of the regular diurnal inequality .
The ratio borne by the irregular to the regular range is distinctly greater in the winter and equinoctial than in the summer months , showing a fairly well marked maximum in October and November .
Taking the year as a whole , the mean of the individual daily ranges exceeds the mean value of by about 10 per cent.\mdash ; a somewhat remarkable fact ; the ratio borne by the formel quantity to the latter falls slightly below unity in three of the winter months , but seems nearly uniform from March to September .
The ( numerical ) non-cyclic resembles the mean of the individual daily ranges in being largest , absolutely considered , in winter , while showing the largest ratio to the mean value of in the equinoctial season .
The large size of the average noncyclic effect , when taken independent of sign , is an important element in determining the significance to be attached to the results obtained in Table II .
where it is treated algebraically .
Special Features of the Selected S 17 .
In confining ourselves to 10 selected days a month we have followed a precedent set by the annual selection of 5 magnetically " " quiet\ldquo ; days a month by the Astronomer Royal .
In the case of the magnetically quiet days , so far as is known at present , there is no systematic departure from average conditions in anything but the phenomena of Terrestrial Magnetism .
In the present case , however , it is otherwise .
We have eliminated all the rainy days , which constitute a considerable proportion of the days in most winter months , and not infrequently eyen in summer months .
The meteorological conditions on the average selected day thus represent a climate which at some seasons of some years is far from being representative of Kew .
It seemed FROM DAYS THE YEARS 1898 TO 1904 .
important to ascertain what this difference really amounted to .
Fortunately , Kew is a first order meteorological station , so that exceptionally full information is available as to the hourly and the mean daily and monthly values of the meteorological elements .
A comparison was accordingly instituted between mean values deduced for each month from the 10 selected days and the corresponding published taken from all days the month .
was done for the mean barometric pressure of the day , the mean temperature , the mean hourly wind velocity , and the daily record of bright sunshine .
All these elements are recorded directly by self-recording instruments of high accuracy , and the pressure and temperature daily means are based on hourly readings .
The means from all days of the month are regularly calculated and recorded by the Meteorological Office ; the means from the selected days were specially calculated by members of the Observatory staff .
The results of the comparison appear in Table XI .
It is at once apparent that the 10 selected days represent a climate havin more sunshine and a higher barometer than Kew , but which has a distinctly lower winter temperature , and in most months of the year appreciably less wind .
In summer the temperature of the selected days is , if anything , slightly above the normal , but the difference is too small to rely on .
Absolutely considered , the difference as regards sunshine is largest in summer ; but relative to the average amount actually experienced the difference is really greater in winter .
TABLE XI.\mdash ; Mean Excess of Element on the 10 Selected Days as compared to All Days of the Month .
VOL. CCYI. .
2 ection of Atmospheric Potential with Meteorological Elements . .
There are so many possible ways in which a connection may subsist between electric potential and a meteorological element that it is difficult to know how to attack the problem .
Suppose , for instance , that a particular June had a very high mean temperature , then there might be an effect either on the mean value of or on the amplitude of the daily changes , or on both .
Possibly , however , there might be no appreciable effect on either quantity , and yet an appreciable difference exist between the electric phenomena of the hotter and colder days of the average June .
The former phenomenon would usually be interpreted to mean that no connection exists between potential gradient and temperature ; the latter would naturally be interpreted in the opposite sense ; and yet neither interpretation would be justifiable without there investigation .
A very warm June would differ from the average June in other respects besides temperature , and a true direct temperature effect might be practically annulled by some other effect .
Again , during an ordinary June , a hot and a cold day usually present notable differences in other meteorological elements besides temperature .
The first method I have applied is to divide the 7 Januarys , 7 Februarys , and so on , into two groups , one of three the other of four , according to the size of the mean potential gradient , and to find the difference between the meteorological elements for these two groups , taking first the selected days only , and then all the days of the month .
The sub-division of each seven months was done from consideration of the mean potential only , the choice being decided by whether the biggest gap in the values of occurred between the third and fourth , or between the fourth and fifth on the list , when the months were arranged the order of the size .
It is necessary to deal with the months of the year separately .
The combination ' of days from all seasons of the year inevitably leads to error .
The cold days , for instance , then come mainly from winter , and the hot days from summer , and the effects of the regular annual variation get mixed up with any true temperature effect .
Table XII .
gives the results of the ison ; stands for the mean from the selected days only , for that from all days of the month .
The excess is shown for as well as for the meteorological elements , as the size of the potential difference should be taken into account in considering the part that chance may play in the results .
FROM SELECTED DAYS DURINGt THE 1898 TO 1904 .
.\mdash ; Mean Excess of Values of Meteorological ements for Group of Years of Larger as compared to Group of Smaller Potential Gradient from ( selected days ) and ( all days of month ) .
euutabhnmmesar 6 The difference between the mean values of for the two groups is considerably larger in winter than in summer , but in every case it represents a very appreciable fraction of the mean from the whole 7 .
As a rule , the meteorological results have the same sign whether we take all the days of month or the selected days only .
In the four summer months there seems an association of high pote1ltial with hi gh barometer , high temperature , high wind velocity and much sunshine ; but the association with temperature in pal.ticular is very doubtful , two months going one way and two the other .
Equinox agrees on the whole with summer , but the association is less clear .
In winter the results seem the exact opposite of those in summer , but the association with sunshine is small and uncertain .
association of high potential with low temperature in winter is the most consistent .
It will be remembered that the selected days in winter show a markedly lower temperature and less wind velocity than the average of the month , and these meteorological conditions , as we now see from Table XII .
, } ) to be associated at that season with large values of P. S19 .
The second method adopted was to split the ten selected days in each individual month into two groups of 5 , the one containing the days of highest mean the other those of lowest mean P. In the case of March , for instance , we split 70 days into two groups , to the one of which each year contributes the five days in which was greatest , while to the other it contributes the five days in which was least .
Table XIII .
shows the excess of the mean in the former group and of the corresponding mean values of certain meteorological elements as pared to the means from the second group .
This second method gives notably larger differences of potential between the two groups , and so is more likely priori to present decisive results than Table XII .
In Table XIII .
there seems in winter a very distinct association of high potential with low temperature and with low wind velocity .
This is so far in accordance with Table XII .
; but the association of high potent.ial with low temperature in Table XIII .
extends to every month of the year except July , whereas in Table XII .
March , May , August and September are exceptions to the rule .
TABLE XIII.\mdash ; Mean Excess of Values of Meteorological Elements for Groups of Days of Larger as compared to Groups of Days of Smaller Potential Gradient .
Table XIII .
shows also a greater apparent association of high potential with much sunshine than does Table XII .
As regards barometric pressure , Table XIII .
shows from April to August almost no trace of any association with potential gradient , and the results for the six months , September to February , in Tables XII .
and XIII .
differ in even .
The results for barometric pressure are thus exceedingly contradictory .
S20 .
The months which show a high mean value of show also a large average daily range , but it by no means follows that the causes to which the two phenomena FROM SELECTED DAYS DURING THE SEVEN 1898 TO 1904 .
are due are the same .
I thus split the 10 days of each month into two groups , according to the size of the potential range , and have compared the mean results from the two groups of 35 days derived in this way from the seven years .
Except as regards the criterion by which the 10 days were subdivided , the procedure was exactly the same as that followed in calculating Table XIII .
The results of the new subdivision are shown in Table XIV .
TABLE XIV.\mdash ; Mean Excess of Values of Meteorological Elements for Groups of Days of Larger as compared to Groups of Days of Smaller Potential Range .
tncrumynyourieeit 1 Temperature , Wind , Fahrenheit .
miles per houl Comparing Tables XIII .
and XIV .
, we find that the signs of the meteorological differences agree in the large majority of instances .
The meteorological conditions which are associated with high potential thus appear associated with a large diurnal range .
The phenomenon may mean nothing more than that the size of the daily range tends ceteris paribus to increase or diminish according as the mean value of is greater or less .
The association , however , with low temperature , low wind velocity , and long sunshine , seems decidedly less in Table .
than in Table XIII .
There is considerably greater regularity in the results for barometric pressure in Table XIV .
than Table XIII .
, but the relationship with seems decidedly opposite in summer to what it is in winter .
The natural inference is that there is little if any connection between the absolute height of the barometer and the size of either the daily mean or the irregular daily range of P. On the other hand , in winter it seems clear that the weather conditions which accompany a high barometer favour the existence of high values and big diurnal changes in P. .
C. : ATMOSPHLRIC POTENTIAL RESULTS AT KEW S 21 .
Instead of the mean value of , or of the range of , one may take for the criterion in grouping the selected days the value of some one meteorological element .
This is a delicate mode of analysis , so far as the particular element is concerned ; but to carry it out for each element , for each month of the year , would have involved a very serious amount of arithmetic .
The method has thus been applied only to some of the elements , and only for a selection of months .
Table XV .
gives the results thus found in the case of mean barometric pressure and mean daily temperature .
December and January were selected as the months when according to Tables XIII .
and XIV .
the most decisive results might be expected .
June , July and August were selected partly as representing summer , and partly in consequence of the exceptional nature of the July results in Tables XIII .
and XIV .
There appears in Table XV .
, as in the tables just mentioned , a decided association of high potential and large range with high barometer in December and January .
In summer the range of barometric pressure is much less , so we should not in any case expect the effects on the potential to be so conspicuous .
The results , however , are not merely smaller numerically in the three summer months , but they differ in sign amongst themselves .
Table XV .
thus agrees with Tables XIII .
and .
in pointing to the conclusion that any association of mean potential or potential range with barometric pressure in summer is small and doubtful .
In all five months there is in Table XV .
an apparent tendency for to incl.easemol'e ( or diminish less ) throughout the 24 hours when pressul.e is high than when it is low ; curiously , however , this seems much larger in the summer than in the winter months .
The exceptionally large size of the non-cyclic effect , when taken irrespective of sign , in December when pressure is high , and the correspondingly large size of the range , are both probably due in considerable measure to fog .
This is an element whose intensity at two consecutive lnidnights is usually widely different .
FROM ) DAYS DURIN THE SEVEN 1898 TO 1904 .
In the case of temperature , July stands out from the other months exactly as it did in Tables XIII .
and XIV .
The association of high temperature with low potential is again most conspicuous in the winter months .
Accompanying the diminlished value of we have a diminution in the non-cyclic effect treated numerically .
The apparent absence of temperature effect on the potential range in January , and the exceptionalty large tendency to an increase in days when temperature was high , are probably due to some common cause .
S22 .
If there were a direct causal connection between any meteorological element and potential gradient or range , we ought to get the same or at least closely similar results for the ratio ( difference of potential)/ ( difference of element ) when the ratio is derived from a Table such as XV .
, and when it is derived from a Table such as XIII .
or XIV .
If such an influence though not unique were dominant , the two ratios so found should be at least of the same order .
Combining the results from the two months , December and January , I find for the ratios of the differences\mdash ; the units being for temperature , for pressure 1 inch\mdash ; The case of mean potential and temperature is the only one in which the ratios are of the same order .
The mean temperature of the selected days in December and January was F. If instead of December and January we taken June and August , we should have got altogether different results for the potential-temperature ratios , while taking July the sign even would have changed .
Thus , if any direct relationship exists between potential and temperature , it must give a ratio between potential change and temperature change , which alters rapidly , and which changes sign at a temperature between ( the mean temperature for the selected days in July ) and ( the mean temperature for the selected days of June and August ) .
It may be the purest accident , but somewhat similar ideas present themselves when we consider the mean seasonal values of potential and temperature .
For these , taking the temperatures of the selected days , I find\mdash ; WinterEquinoxSummer .
Mean potential .
Mean For the ratio of ( potential decrease)/ ( temperature increase ) we have from\mdash ; The values thus diminish marked ] as the mean temperature alters .
Further , for nearly equal increments in the mean temperatures we have nearly equal decrements in the ratio .
Also , if the ratio were to diminish at the same rate as the mean temperature rose further , we should have it vanishing altogether below , i.e. , below the mean temperature of the selected days in July .
S23 .
The only other element which I have investigated in the same way as pressure and temperature is the duration of sunshine .
Several authorities , e.g. , ELSTER and GEITEL at one time , have regarded ultra-violet radiation from the sun as probably a most important agent in the phenomena of atmospheric electricity .
If this were the case , one would expect to find a conspicuous difference in summer between days of much and of little sunshine .
The grouping of the days according to sunshine in the way described above led to the following results for the excesses in the groups of days of most sunshine:\mdash ; TABLE XVI .
If the July data alone existed , then the conclusion one would naturally draw from the above figures\mdash ; just as from the corresponding figures in Tables XIII .
and xIV.\mdash ; is , that with long duration of sunshine there is a very appreciable rise both in the mean potential and in its daily range .
If June and August however are combined , we get for practically the same excess in hours of sunshine a wholly negligible effect on the value of the mean potential ; whilst the apparent effect on the range , such as it is , is opposite in sign to that observed in July .
S24 .
The effects of temperature , netric pressure , wind and sunshine on the simultaneous values of the potential gradient at Kew were also considered in my FROM SELECTED DAYS DURING THE 1898 TO 1904 .
previous paper*of 1896 , which dealt however only with eye observations at one or two hours during some months of 1895 and 1896 .
The results obtained are generally similar , except that in 1896 I found an apparent association of low potential with long previous sunshine .
Even here the contradiction may be more apparent than real , as the influence of sunshine may depend on the hour of the day .
An intensification for instance of the afternoon minimum may be accompanied by a compensating increase of potential during the night .
Considerations of this kind must in fact be borne in mind in other cases besides that of sunshine .
S25 .
In 1896 I paid special attention to aqueous vapour , as EXNER'S theory , which associated potential gradient and aqueous vapour .
in a definite formula , was then rather in the ascendant .
I advanced grounds\mdash ; whose cogency will now , I think , be hardly questioned\mdash ; for believing the theory in its general form to be untenable .
The 1895-6 data were from a Kelvin portable electrometer , whose insulation was never in doubt .
The present data are from an electrograph , whose insulation may depend , especially at night\mdash ; sometimes it may be feared not inappreciably\mdash ; on the greater or less dryness of the air .
Also , for one and the same amount of aqueous vapour in the air , it seems all important so far as potential is concerned whether fog forms or not , and in the lon nights of winter no satisfactory data exist as to the occurrence of fog , or its thickness when present .
I have thus omitted consideration of the effects of aqueous vapour as too large a subject to be included in the present paper .
A similar remark applies to wind direction .
There are here two main difficulties : first the fact that the direction is perpetually shifting , and the extent of the variation in a single day is often large ; second the fact that when the wind is light\mdash ; and there are few fine weather days when this does not happen sometime during the 24 hours\mdash ; the record of direction is often uncertain , the strength of the wind bein insufficient to turn the vane of the Robinson cup anemograph .
In particular is this true of foggy days in winter , the precise days in which the potential tends to be highest .
APPENDIX .
On the Relation between the Regular Diurnal Changes of Barometric Pressure and Potential S26 .
The only meteorological element whose diurnal inequality presents two prominent maxima and minima is the barometric pressure .
EVERETT seems to have been the first to call attention to the similarity between the regular diurnal inequalities of potential and barometric pressure , and the potential data which he considered more especially in this connection were those from Kew .
Having no Kew barometric data , EVERETT employed instead some from Halle .
Confining himself to the mean .
( for summary of results , see pp. 130 , 131 ) .
VOL. CCVI.\mdash ; A. 2 DR. C. CHREE : ATMOSPHERIC ELECTRIC POTENTIAL RESULTS AT KEW liurnal inequalities for the year , he found that " " the barometric ourve for Halle bears a strong resemblance to the Kew electrical curve , but is upwards of an hour later in phase Several other writers have since called attention to general resemblances of this kind .
Of late years , additional interest has attached to the possibility of the connection , owing to ELSTER and GEITEL'S discovery that air drawn from the soil is ionized , and their consequent suggestion that change of arometric pressure may influence potential gradient , by modifying the rate at which this ionized air escapes into the atmosphere .
diurnal inequality of barometric pressure is an element oonsiderably dependent on local conditions , and it thus appeared essential to an adequate discussion to have barometric data for Kew .
Diurnal inequalities were accordingly got out by the Fig. 4 .
Observatory staff for each month of the year for an 11-year period 1890 to 1900 , making use of the data published by the Meteorological Office in the 'Hourly Means .
' The regular diurnal change is a small quantity , and accordingly means were calculated to inch .
This is perfectly legitimate , as the curves are read to inch , and each hourly value in the mean diurnal inequality for a month was based on over 300 individual readings .
For our present purpose full details of the diurnal inequalities are not absolutely necessary , and I have consequently omitted them to econollise space .
S .
That vERETT had a substantial basis in claiming a resemblance between the mean diurnal inequalities for ths year in barometric pressure and potential gradient is apparent on comparing the two curves of fig. 4 .
The heavy curve , representing FROM SELECTED DAYS DURING THE SEyEN yEARS 1898 TO 1904 .
potential gradient , is drawn on a more open scale than the curyes of fig. 2 ; the light ( jurve representing barometric pressure is drawn on a scale such as to make its apparent range nearly equal to that of the other curve .
Whilst the resemblance is very striking , as between two different elements , there are differences which seem of a fundamental character .
During the forenoon the curves are nearly in the same phase .
The barometer curve lags a little , but very little , behind the other .
In the afternoon , however , the lag in the barometer curve becomes conspicuous , amountin to about two hours at the times of the afternoon maximum and minimum .
This in apparent lag throughout the day does not seem to have been noticed previously , but it persists with wonderful regularity throughout the year .
Treating the maxima and minima as occurring at the hours when the algebraically greatest and least hourly values occur , I find that the hour of the forenoon minimum is the same for the two elements in every month from January to October ; the mean lag for the barometric pressure for the 12 months comes to of an hour , but the exceptionally early hour of the potential minimum in October is responsible for more than half this difference .
In the case of the forenoon maximum there is exact agreement in the hour in eight months , and the mean lag in barometric pressure for the 12 months is only hour .
December is the only month when the afternoon minima accord in time , and the mean lag for the barometric pressure is hours .
The corresponding quantity for the afternoon maxima is hours .
If the relation is a case of cause and , the fact that it is the barometer curve that lags relative to the other would naturally lead one to regard the potential variation rather as the cause than the effect .
S 28 .
When we consider the variation from month to month in the amplitudes of the diurnal inequalities other differences appear .
This will be seen on XVII .
with Tables III .
and .
IV .
The first column in Table XVII .
gives the range , or difference between greatest and least hourly values , in the diurnal inequality of barometric pressure ; the second column the numerical sum of the 24 hourly differences from the mean for the day .
The corresponding potential quantities will be found in the two last columns of III .
The ranges and the sum of the differences show analogous phenomena , but the latter are the better for comparison , because they give a smoother annual variation when the number of years considered is limited .
Comparing , then , the second column of Table XVII .
and the last column of Table III .
, we see that in ( barometric pressure ) the values for the winter months November to February are decidedly the least , and the values for April , May , June , and September the largest .
( potential gradient ) , on the other hand , the largest values occur in December , February , and March , and these are much in excess of the TABLE XVII .
rometric Pressure .
Diurnal Inequality ( Unit inch ) .
values in the summer months .
The annual variations in the sum of the differences for and are shown together in fig. 1 .
To get a smooth curve in either case would probably require some 30 years ' observations , but the differences between the two curves speak for themselves .
S29 .
Whilst the difference of phase in the two curves of fig. 4 is what most readily catches the eye , another difference seems not less important .
In the urve the afternoon maximum and the morning minimum are respectively the largest and smallest values in the day ; but in the curve it is the morning maximum and afternoon minimum that are the extremes .
Comparing the two last columns in Table XVII .
with the day and night falls of in Table IV .
we note again an essential difference .
In the case of the day fall is the larger in every month of the year .
The excess , it is true , is most conspicuous in summel ( the season when the day fall in exceeds the night fall ) , but taking an rithmetic mean from the four winter months , November to FebruaIy , we have day fall/ night fall in in P. S30 .
disclose more exactly the nature of the differences , the monthly and seasonal diurnal inequalities of barometric pressure were analysed in Fourier series with 24 , 12 , 8 , and 6-hour terms .
The appear in Table XVIII .
, the notation used corresponding exactly to that of Table FRO SELFCTED DAYS DURING THL SFVFN YEARS 1898 TO 1904 .
TABL XVIII.\mdash ; Barometric Pressure , Diurnal Inequality , Amplitudes , and Phase Angles Unit Amplitudes , inch ) .
Comparing Tables XVIII .
and V. , we see that in both the amplitude of the 24-hour term varies much with season ; but whereas in the case of it is conspicuously largest in winter , that is precisely the season when it is least in B. Table XVIII .
shows apparently a secondary minimum for in summer , but the smallest value in summer is 13 times the value in January .
The fundamental nature of the difference between the annual variations of in and can in fact be seen at a glance on comparing the heavier and lighter full line curves in fig. 3 , which represent the monthly values expressed as fractions of the arithmetic mean .
The phase angles in Table XVIIT .
show erratic variations in the winter months , which are very probably fictitious .
Judging by the seasonal values , is larger in winter than in summer , but the difference is small compared to the corresponding difference in Table V. In all the seasons is much less for than for , in other words , the hour of maximum for is much the earlier .
S 31. .
In the case of the 12-hour term the differences between the annual variations in the amplitudes for and are comparatively small .
These variations are illustrated by the dotted curves in fig. 3 .
The phase angles in Table XVIII .
show fairly similar characteristics to those exhibited in Table V. In both cases the seasonal variation is comparatively small .
The phase angle is invariably the larger , i.e. , the maxima in the 12-hour curve are again earlier for than for ; the difference in time is however on the average only slightly over hour , though greater in summer than in winter .
The 8-hour term in is unusually large .
In winter is in fact much in excess of The annual variation of is regular and strongly developed , but is of a very unusual character .
The winter and summer values of in Table XVIII .
are opposite in phase .
There is also a considerable difference between the winter and summer of in Table V. , but the difference is twice as big for as for and is exactly in the opposite direction .
In the case of the 6-hour term the phase angles in and vary somewhat similarly with the season , and the difference in the case of the mean diurnal inequalities for the year represents a difference of only some 10 minutes in time .
The almual variation of is somewhat irregular in both elements , and a long series of year 's data would be necessary for a satisfactory comparison .
S 32 .
When comparing the Kew data with those from the Bureau Central and the Eiffel Tower , we found that the 12-hour term the diurnal inequality of potential was the only one which presented under all conditions closely similar features as regards .
times of maxima and minima , and as regards amplitude near the ground .
Similarly , it is well known that the 12-hour term in the diurnal inequality of barometric pressure presents at the ground level phenomena of a much more regular character than does the 24-hour term .
The latter term seems largely dependent on local conditions , whereas the 12-hour term is usually pretty much the same at places of nearly the same latitude .
These phenomena are shown in a striking fashion by comparing the data of Table XyIII .
with corresponding data for Jersey*based on observations by DECHEVRENS .
In view of these facts , it is highly significant that the general resemblance that has been noticed between the diurnal inequalities of potential gradient barometric pressure is now found to proceed almost entirely from the parts that have a 12-hour period .
At Kew , at least , so far as the 24-hour terms are concerned , it is a case not of resemblance , but of marked dissimilarity .
During the seven years whose data have been considered , the charge of the electrograph and the taking of the absolute observations have devolved upon Mr. E. .
CONSTABLE , the Senior Assistant engaged in the meteorological work of the Observatory Department , and his Junior Assistant , Mr. E. BOXALL .
The selected days were chosen by Mr. CONSTABLE in consultation with myself The chief burden of the preliminary work was borne by Mr. CONSTABLE , who drew the free-hand pencil curves and took all the measurements .
A good deal of the arithmetical work was done by other members of the staff , especially by Messrs. BADDERLY and FRANCIS .
All the reductions and calculations have been carefully checked , mostly by myself , and notwithstanding the multiplicity of details I hope no serious errors remain undetected .
It remains to express my thanks to the Meteorological Office\mdash ; to whom the Electrograph belongs\mdash ; and to the Director , Dr. SHAW , for permission to publish the results .
' Meteorologische Zeitschrift , '-August , 1906 ,

