]\gt ; XV .
Reduction of Anemograms taken at the Armagh Observatory in the Years 1857-63 .
By T. R. ROBINSON , D.D. , F.R.S. , F.A.S. , Received June 11 , 1875 , \mdash ; Read June 17 , IN the beginning of the year 1845 I erected a self-recording anemometer at the Armagh Observatory , and have a series of its records up to the present time , unbroken except by accidents to the apparatus or occasional illness of the observers .
I , however , soon found it was impossible for me and my single assistant to reduce continuously the mass of materials which was accumulating , without neglecting the primary objects of the establishment ; and I was obliged to content myself with them , in hope that they might be le to future inquirers .
It was thought , however , by some distinguished members of the Royal Society that it was desirable to ascertain how far such observations are able to develop any definite laws amid the seeming lawlessness of the wind ; and a grant was made to me from the Government Grant sufficient to discuss the anemograms for the seven yeals from 1857 to 1863 .
The work has been delayed by the death of one of the computers , migration of another to India , and my own tempolary blindness .
The anemograph is that desclibed by me in the ' Transactions of the Royal Irish Academy , ' vol. xxii .
It differs in nothing essential from that employed by the MeteoCommittee of the Royal Society : the recording-apparatus is different , and the direction is observed by a vane whose excursions are controlled by a peculiar contrivance instead of by a windmill .
The space-records were read to of a mile ( statute ) , and the directions to .
The S. and W. components of the hourly velocity were computed for each to two places of decimals .
Wind is caused by a difference of pressure in the air over adjacent portions of the earth 's surface ; but of the agencies which produce this difference we as yet are imperfectly informed .
Heat is obviously a most important one .
We see that the action of the sun must produce a current from polar towards equatorial regions , and that when the geographical conditions of districts not too far asunder are such as to make their temperatures unequal , air-currents between them will result .
The changes of solar action at a given place depending on the hour of the day and the day of the year , to produce definite periodical modifications of the wind ; and the currents due to the varying tension of aqueous vapour ought to be similarly periodical .
Were these the only causes of the wind , there seems no reason why its force and direction at a time and place might not be predicted as certainly as the sun 's altitude .
But are evidently disturbing agencies of great power which entirely mask the regular .
course of .
T. R. ROBINSON ON TllE REDUCTION OF ANEMOGRAIvIS .
the phenomena , and of whose nature we can only form vague conjectures .
The accumulation of ice in the polar regions forming icebergs may be such an influence ; and what we have learned lecently of the action of the larger planets on the solar spots , and of the connexion of the development of those spots with the magnetic storms and aurolal discharges of our own planet , may suggest the possibility of extra-terrestrial forces ] aying some part in the question before us .
But without following in the track of imagination , this is certain .
that however complicated and irregular a phenomenon may be , if we have a sufficient number of observations , it is possible to determine the values and periods of those parts of it which are subject to definite laws .
Where any of these periods agree with those of agents whose influence is certain , they may be referred to them with certainty , and their effect eliminated , making it much easier to deal with the residual phenomena .
In the present instance the want of self-recording instruments for pressure , temperature , and vapour-tension compelled me to consider the wind solely in reference to time , as depending on the hour of the day and on the month ; and even with this simplification it is not easy to come to precise results .
Were we to seek a velocity and direction which might be considered normal for each hour of the year , such is the irregularity of the air-currents , that I think it could scarcely be obtained in less than 100 years .
Even if we confine ourselves to the west and south components , and take for successive hours the mean of the seven years concerned , it differs so widely from the means of the preceding and following hours , that any existence of law might seem impossible .
But if the hour-means be taken for 20 or 30 successive days , their means present a very different aspect .
I have taken them for months .
Before dealing with these components , I think it may be instructive to present a Table giving a synoptic view of the winds , which may show their general character at Armagh during the seven years concerned .
It gives for each month and for each of the horizon ( S. to S.W. , S.W. to W. , \amp ; c. ) the mean hourly velocity , the mean direction , and the approximate number of hours during which this wind has blown .
At the end of each month is given the maximum hourly velocity for each year , the number of hours when the velocity exceeded 25 miles , and the number of hours during which the anemograph has recorded .
This does not imply that during this time there was no wind , but that there was not enough to move the instrument .
This requires a velocity The direction-vane is much more sensitive ( very much more so than the windmillapparatus now used to record the direction ) , and therefore the records of direction are more numerous than those of velocity .
DR. T. R. ROBINSON ON THE CTION OF ANEMOGRAMS .
TABLE I.\mdash ; January .
ABLE IFebrual y TABLE I.\mdash ; March .
TABLE I.\mdash ; April .
DR. T. R. ROBINSON ON THE REDUCTION OF ANEMOGRAMS .
TABLE I.\mdash ; May .
TABLE I.\mdash ; June .
TABLIB I.\mdash ; July .
TABLE I.\mdash ; August .
DR. T. B. ROBINSON ON THE REDUCTION OF TABLE I.\mdash ; September .
TABLE I.\mdash ; October .
IIDCCCLXXV .
3 I TABLE I.\mdash ; November .
TABLE I.\mdash ; December .
DR. T. R. ROBINSON ON THE REDUCTION OF ANEMOGRAMS .
first thing which strikes one in this Table is the irregularity of the wind .
It varies in each octant ; in each octant it varies with the month , and in each octant and month it varies with the year .
As to the first of these variations , both the velocity of the wind and the number of hours during which it blows are , in general , a maximum in the first octant ( S. to S.W. ) ; they decrease from this to a minimum at octants N. to N.E. , and increase to octant 1 .
The products of the velocity and time at the maximum and minimum are as 6 : 1 .
The predominance of south-westerly winds is what might be expected from the combination of an current with the earth 's rotation ; but it is not obvious why it is not absolute .
Probably much of the of direction arises from circumstances local to the place of observation .
For instance , the ection of the west coast of Ireland , which runs nearly N. and S. , may occasionally turn the S.W. currents northward ; and the mountainous ground of Antrim may divert it here towards the east .
It must also be lemembeled that our only measure of the wind at the earth 's surface , where it is at once letarded and into gigantic eddies and vortices by the effects of friction .
The experience of aeronauts shows that at a few thousand feet elevation the velocity is often far greater than it is , and that the direction is much more uniform .
But I do not see how this error is to be remedied .
The summit of a mountain is not exempt from it ; and though a small and lofty island , like St. Kilda , far from any extensive land , would be bettel , yet even here the fi.iction of the sea 's surface will destroy velocity .
It is possible that an raph at the top of a tall and slendel " " \ldquo ; would give much velocity than one at its base ; the lecord could be easily effected below by telegraphy .
We must remember that a cul'rent of air comports itself like one of watel , and shall be assisted in comprehending the nature of a gale by .
the movements of a river in flood .
There must also be eddies in a vGltical plane .
On the action of these see a valuable paper by Prof. HENNESSEY in Phil. Trans. 1860 .
An anemograph for vertical currents might be made by a set of windmill-vanes placed Secondly , in each octant the amount of wind val.ies with the month .
It is a maximum in January ; decreases from this to July , the l.atio being : 1 .
From this it incleases to the end of the year .
There is an exception to this in March , where the daily amount is greater than in Februal.y in the l.atio of : 1 .
This might seem to countenance the vulgar notion of stormy weather prevailing near the ; but there is no such excess in September above October ; and in March , though the yearly maxima arc than in February , yet the number of hours when the velocity exceeds 25 miles is considerably less .
This monthly change is an obvious consequence of the change of the sun 's declination , for the zone where the easterly winds of low latitudes confine with the westerly ones of more northern regions must shift with that to which the sull is vertical .
For the third of these regularities , that which preyails from year to yeal\ldquo ; there can , in the present state of our knowledge , be no celtain cause assigned .
It will be seen that in the same octants the variation is very different in each month , and that the DR. .
B. ROBINSON ON THE REDUCTION OF ANEMOGRAMS .
maxima in each octant do not belong to the same years ; while the amount of discordance is so great as to almost exclude the idea of any law .
I looked for one in the direction already noticed .
In 1860 the sun-spots were at a maximum , in 1856 at a minimum ; and if they exelt any uence it must have been considerably less in 1857 than in 1860 .
The products of velocity and time were accordingly examined in these years , and that for 1860 is 4167 greater than in the other .
But this result is reversed by 1863 , which exceeds 1860 by a still gleater amount , 5223 ; and evidently many decennial periods must be examined before any reliable conclusion can be attained as to this influence .
The same lawless irregularity may be observed in the maximum velocities of separate years .
The highest in the period before us is 71 miles in November 1861 , the lpwest 19 , in July 1860 .
Far highel velocities than these are sometimes attained , but only for a few minutes .
It holds also as to the number of houls when the velocity exceeds twenty-five miles .
As instances : in Janual y this number is 146 , in1857 it is 23 ; in Apri11859 it is 40 , in 1861 it is ; in Noyember 1861 it is 90 , in 1860 it is 5 .
It occurs also , though not intensely , in the hours of calm .
It may have some interest to the mean velocity for each month irrespective of the direction .
TABLE II .
Here also there seems little indication of equinoctial gales .
March is a trifle more windy than February , but September less so than October .
The yearly sums also do not show any special relation to the solar spots ; the total in ; in ; but in .
The total miles in the seven years and the mean velocity during that time is II .
The most obvious way of dealing with the west and south components of V is to derive from them interpolation formulae for each year involving periodic functions of the time , and deduce from the coefficients of these formulae in successive years some general law .
This , however , seems impracticable , for the components differ so widely in successive yeals as to preclude any hope of reconciling them .
As a specimen of this discordance I give the values for the first hour of the series for January 1:\mdash ; .
T. B. ROBINSON ON THE REDUCTION OF ANEMOGRAMS .
1862 \mdash ; It is -evident that -here is no regular succession ; and equally so that little dependence can be put on even the 5l that day .
But if , as is probable , these discordances are casual , we may expect they will disappear from the mean of a number of observations\mdash ; how large may be estimnted the Probable Error of these observations , though , on account of the magnitude of their discordances , this cannot be detelmined with great precision .
Thele is also this difficulty in the process of finding the Probable Error , that the coordinates daily and monthly variations , which must not be confounded with the casual errors .
It is therefore necessary to confine oulselves to the observations of each individu al hour during the seven years , and combine any mber of these groups of seven .
This is effected by the simple means of .
as the divisor instead of the number of terms in the entire set , and the numbel of groups .
I have only thought it necessary to make the computation for in January and June , and I find of a observation of mean of seven of in Table , mean of 217 of mean of month The dancy in summer is only two thirds of that in winter , and in both is so rreat that the mean of seven is .
to be relied on ; and even the numbers of Table III .
are not sufficiently certain .
Perhaps these seven years may have been exceptionally irregulal .
The discordancy of is still greater than that of W. Evidently single houls were out of the question ; I therefol took for each hour the mean of the month in the first instance ; I then grouped these means for every ten days , but ultimately adopted the entire month as the group .
Before discussing these means individually , it may be useful to give their means for the entire period of seven years .
Supposing the winter from October to March inclusive , the summel from April to September , the day hours from .
to 6 , the night from 7 .
to 6 we find : Winter Day .
Sum ; Sum ; Ann. Translation Winter Night .
Sum ; Sum ; Ann. Translation Summer Day .
Sum ; Sum ; Ann. Translation Summer Night .
Sum ; Sum ; Ann. Translation Both components are more than twice as great in winter as in summer ; the day components are greater than the night ones , except the winter S. The sums of all are Sum ; Sum ; Ann. Trans. ; On examining the records of the components , I find that 630 hours were missed by various accidents , so that the total number of hours is 60714 ; and the above sums , , will give for the mean hourly values ; The value of shows that the wind in the first quadrant is nearly half the total amount .
The monthly means of the components are given in the following Table ( p. 415 ) .
On examining this Table we obserye , First , that all the values both of and are positive ; in other words , that in a considerable number of observations the aerial currents from west and south have at this station a decided predominance over all the others .
Secondly , that , as was anticipated , however discordant the results of individual hours or days may be , yet the means of from 196 to 217 present a notable reement , and the differences which they exhibit are evidently subject to law .
If we look down the vertical columns ( which give approximate values for each hour of the middle day of each month ) we find in each a decided maximum and minimum , and another , or even more than one of each , less in amount .
The hours of these phases vary with the months ; that of the principal maximum occurs in the winter months from noon to 3 .
for ; in the summer from 9 .
to mocn ; for it varies less , being a little before noon .
The principal minimum occurs in the evening , from 6 .
to 10 , both for and S. The extreme diulnal ranges are greatest in March , being for , for ; they are least in November , being and It deserves notice that during the winter months the horary values of for the four afternoon hours exceed those for the four that precede , the sum of the differences being .
In the summer months the reverse is the case , but the \mdash ; differences are only Does this arise from the great extent of land to the east of Ireland as contrasted with the ocean to its west , and the greater evaporation from the latter in summer ?
If we examine the horizontal columns ( which show the monthly variations ) the dominion of Law is still more manifest .
has a maximum in January , a minimum in February ; the greatest maximum is in March , the leas minimum in April : these abrupt changes are remarkable ; but it is possible that the great value of in March is abnormal , and may not occur in subsequent years .
It then increases with a slight DR. T. R. ROBINSON ON THE REDUC'['REN CT NN@OOGRANE .
maximum in August and a slight minimum in November .
The variation is greater here than in the horary columns , being for hour .
The largest is at March ; the least at May The law of is simpler ; it has one maximum in December and one minimum in July ; its range , too , is something being in hour .
The greatest magnitude at December , its least , July here is a general agreement in the change of the two components , with one striking exception , the maximum and minimum which has in March and April .
Such a general agreement might be expected , for any air coming from the south must have a westward motion due to the greater velocity of the earth 's rotation in a southern parallel .
This anomaly , if real , may be caused by the geographical conditions to which I have already alluded .
To them also may be referred the fact that at May , though , from which a sensible magnitude of the other might be expected .
It must , however , be observed that some of the changes exhibited in this Table can scale .cely be regarded as periodical .
I have already pointed out that from the very great discordance of individual cbservations it is evident that a much greater number of them than is afforded by a period of seven years is required to eliminate the barometric and hygrometric influences .
Yet these disturbances might be expected to be distlibuted with some uniformity through the day ; while the changes from hour to hour are sometimes considelable .
Thus in February to ; to ; March to to ; April to Decembel to .
These are the largest ; and it deserves notice that they occur in winter months ; in summer there is much less abruptness of change .
It occm.red to me that some of these irregularities might be due to errors in the records of velocity ; but this seems quite improbable .
Such errors could only arise from three possible causes .
1 .
Referring to my description of the anemograph in the ' Transactions of the Royal Irish Academy , ' vol. xxii .
, it will be obvious that the track of the recording pencil might be excentric to the brass disk which carries the paper .
It was carefully adjusted whenever the clock was cleaned , but was liable to derangement from lough handling .
The error which would thus arise was avoided by an easy adjustment , which made the edge of the reading alidad coincide with the right line drawn by the pencil when the clock was wound up .
It will easily appear that the readings so made are true .
2 .
The paper may be excentric to the centre of rotation .
Let be its excentricity , that of the pencil , the leading of any distance from the winding lin the angle between that line and the line of the two centres , the correction DB .
T. R. ROBINSON ON THE REDUCTION OF NEMOGRAMS .
and calling V the change of in the following hour , correction of ; supposing ( and such an error is not probable ) the maximum error would be .
This , therefore , cannot do much harm .
3 .
A much more serious error may be caused by the rate of the clock which moves the pencils of the instrument .
Suppose it a one , the hour-circles on the paper are less than hours , and the recorded Vs belong , not to the times to which they are ascribed , but to periods a little in advance .
The error is negligible , except for the hour of winding-up .
There the space-curve goes beyond the last hour-circle to a distance equal to the rate in , and the measured V is proportionally too large .
If the velocity were uniform , this would be corrected by multiplying ' by where is the hour-space , its hourly increase ; but as this seldom is the case , the change must be allowed for by interpolation .
In all cases but the last we thus obtain As I never have found greater than , the second term may be neglected , and the coefficient scarcely differs from unity .
In the last we have which may be considerably less than .
The projection of the space-curye beyond the last hour-circle gives .
This excess occurred most frequently in gales from S.W. , was , I think , often caused by the vibration of the lofty structure which supports the instrument .
I have not applied these corrections except in a few cases when the error was glaring .
The .-hour was at 9 .
in 1857 and 1858 , at 10 .
in the other years ; and at these hours this influence might be expected ; but on their values in Table III .
with the formulae of Tables V. and VI .
, they seem as well represented as any of the others* .
The discordances of these quantities would have been less striking had they been grouped as three-hourly means ; and this was my original intention , which I abandoned on account of a difficulty in respect of interpolation to which I will refer presently .
It is , however , necessary to remark that the numbers of Table III .
are merely probable values .
A sensible proportion of the individual values is invariably negative for each hour ; and my first idea was to keep the positive and ?
egative means separate .
I tried it for January and June as extreme cases , and came to the conclusion that this separation would be useless .
The negative values occur so constantly , that they can * I have givcn these details as they will be useful in case it be ever thought desirable to reduce the entire series of their anemograms , which extends from 1847 to 1870 .
scarcely be deemed casual .
In the 744 septimates of January there.are only seven in which all the and are positive ; in the 720 of June there are none .
It might be expected , from the mechanism of the polar and equatorial currents , that both components would change signs simultaneously ; but it is not so .
I find that the proportion of the combinations In ; In June , , ; , , ; , , ; , , The combination of with may arise from the influence of a continent to the east of Ireland , and that of and from a north-east current whose north component has been destroyed by friction ; but I looked for a greater frequency of and -S .
If we confine ourselves to consider , and separately , we find : For January . . .
Sum ; Sum ; Sum ; Sum For June . . .
Sum ; Sum ; Sum ; Sum The amount of negative components does not differ very much in the two months , but that of the positive is nearly triple in January what it is in June .
Were we to attempt to develop separately these nd \mdash ; values , we should be embarrassed by the different numbers of them belonging to each hour .
Thus in January the number for is 47 at , 69 at ; for is 41 at , 56 at .
In June , for it is 63 at , 91 at ; for it is 65 at , 98 at .
Supposing them developed in terms of the time , we should still be unable to obtain any absolute values of the components at a given epoch unless we knew the causes which produce these negative values and the laws of their action .
It is evident that the equatorial current predominates here , but that there coexists with it a polar one , probably above , possibly collateral , which is occasionally mixed with the other by some disturbing barometric .
It seems also that the monthly variation of the components is in a great measure limited to the positive values .
For these reasons I have confined myself to the simple means of the entire set .
But I think it might be well , in a series extending to several periods of iive or seven years , to keep them so far separate as to be able to examine whether the occurrence of the negative values has any relation to time .
A Table like this , whose data refer to dates separated by considerable intervals , will not suffice to give the components generally without some process of interpolation ; and we proceed to consider this .
The form universally adopted where the quantities conDR .
T. R. ROBINSON ON THE REDUCTION OF ANEMOGRAMS .
cerned are periodic functions of the time-angle is that given by BESSEL , in which , calling the quantity and the angle , we have But as the monthly variations must be represented as well as the horary , a formula of this nature including two variables would be very complicated ; and it seems best to obtain , as proposed by BESSEL , the horary formula for each month , and to regard the constants of this formula as themselves periodic functions of the monthly time , and develop them in similar formulas of the month-angle , .
Stopping at terms of the fourth order , we should have nine of these for each component ; and for a given day of the year and hour of the day we must compute the constants for the of the day , and multiply each of the last eight by the cosine or sign of the corresponding multiple of The calculation of the horary constants is shortened by observing that for the angles , and the sines and cosines have the same numerical value ; and hence the calculation need only be made for the first quadrant .
Supposing the circle diyided into equal parts , and that contains of these , the corresponding to any may be characterized as , that corresponding to as , and the sum or difference of these two as As the cosines and sines of odd multiples of and differ in sign , but those of even multiples , the expressions of , and will contain only , those of the others only .
The signs of and are easily determined in each case .
Thus for the first multiples of the cosine and sine the entire ; they are \mdash ; and .
For the second multiples the sine is through the quadrant , the cosine to , \mdash ; through the rest ; for the cosine is the same as for , the sine different .
I take , as in the first instance , the horary division in which , and BESSEL'S formulae become and so on .
These are all combinations of the groups ; and by forming these groups the computation is evidently much simplified .
This simplification is , however , only possible when is an integer , and the first arc of the series Whatever be the value of , BESSEL'S formula fails generally to give and the cosine- and sine-coefficients of the order .
The correspond to and this for the older becomes .
Then cosine ( na ) , sine n ; both odd values of , \mdash ; for even ones .
Thence the coefficient\mdash ; na na , sin na na .
Then summing from to , we get .
Here the divisor of is instead of ; and these coefficients cannot be obtained separately unless or , in which case the cosine or sine How far the series is to be continued depends on the periodic fluctuations of the and may be found by trial , or by BESSEL'S expression for the squares of the residual errors .
In any case it should not be carried further than the order , as after that the coefficients coalesce .
BESSEL has shown this for ; and it can easily be proved to hold good when is a submultiple of and a multiple of For the horary groups find the fourth order sufficient .
These horary groups might be combined in triple sets ; but , as I have said , there is a difficulty in the interpolation due to the fact that while , the mean of any three , is multiplied by cosine or sine of the first and third components of it should be multiplied by the same functions of and .
This , however , may easily be corrected .
Take the case of : the effect of three components to determine this DR T. B. ROBINSON ON THE REDUCTION OF ANEMOGRAMS .
versine Developing the sum and difference of the , which gives \amp ; c. , we obtain the term versine Summing round the circle , calling Su ' .
and remembering that all except and vanish , that each of these , and in ordinary cases , we have versine ) , and ultimately A X 4 ( 1 \mdash ; versine ) is given by the same formula , changing the cosines sines in F. For orders , , it is only necessary to use and .
In the case of , however , the formula must be modified ; for in this instance , and the explession is ( ccs versine 4 ) .
The values of the constants .
The suffixes here are the same as in the .
Thus is is I have compared this formula with the observations of uary and March , the most regular of the whole set , and the results , along with those of the one , given in the following Table .
The numbers are the observed \mdash ; the calculated values .
DR. T. R. ROBINSON ON THE REDUCTION OF ANEMOGRAMS .
TABLE IV . .
The iplet combinations are not much inferior to the others , and might possibly be sufficient ; but I prefer the latter .
Even in the extreme cases of February 16 and March 8 , and , the discordance is not as great as I anticipated from the absence of the constant S. I tried them , omitting the terms of the fourth order , but the results were decidedly inferior .
In considering the magnitude of some of these errors , it must be remembered that the formula expresses only that part of the coordinates which is periodic ; and they are the residues of other effects which do not depend on the time , and which disappear from a larger series of observations ; for the other hours the errors are much smaller .
I thought of grouping the hours in pairs , which would probably have given a better result than the triple combination ; but on deducin the formula , I found it would require more logarithmic work than the complete process .
In it the coefficient of a constant of the order has the coefficient , instead of 6 , as is evident from what precedes . .
T. R. ROBINSON ON REDUCTION OF ANEMOGRAMS .
The horary constants of for the twelve months are given in TABLE V. The similar constants of are given in TABLE VI .
The degree of precision with which these constants represent the observations will appear from the number of errors between certain limits .
has from to inclusive , 177 ; from to , 72 ; from to , 33 ; from to , 3 ; above , 3 .
has from to , 172 ; from to , 85 ; from to ; from to , 3 ; above , 1 .
We now proceed to develop these constants in terms of ; but as four orders do not and with sufficient exactness , I have carried the formula to the sixth order , its utmost extent .
Formula where and SO . .
, for reasons already given , cannot be determined .
It is , however , necessary to obviate two difficulties which interfere in the present instance with the accuracy of this process , but which do not affect the horary interpolation .
It supposes that the employed represent values of the coordinates belonging to dates which correspond with a series of in arithmetical progression .
This is not the case ; for ( 1 ) the means of each month do not represent exactly the coordinates belonging to the middle of that month ; and ( 2 ) the angles representing the distance of the middle of each month from the beginning of each year are not in arith metrical progression , as is evident from the following Table , which gives these angles , and also those belonging to each half month TABLE Both these difficulties are overcome by a process based on a suggestion of Professor STOKES .
DR. T. B. ROBINSON THE BEDUCTIO OF ANEMOGRAMS .
Let the true constants of the formula be denoted by small italic letters , so that then , as the mean of through the space , we have mean Let ; and as all the pairs of terms are of the same form , rating this will do for all .
The integral is . . .
which within the limits . . . .
and by obtain , calling mean ?
Now we might form the equations for and treat them by minimum squares ; but as in this case none of the terms would vanish on summing , though all ( except the one , say ' whose square appears ) are small , the labour of eliminating 12 quantities 13 times over would be truly formidable .
This might be evaded by substituting in each sum for the true constants those given by the series of , which differ little from them , and all , except , are multiplied by small coefficients .
This will give with close approximation .
The process may be repeated with the corrected values , but alone will have any notable effect .
Yet even with this simplification the labour is very great .
But it may be superseded thus .
We have the above equation for ; but we have also 2\amp ; c. ; and equating the two values , It is evident that if we put , and so on , the equation would be satisfied , if the factors .
were equal in every month .
They , however , ' differ so little that I have thought it lawful to take their means for the twelve months .
Though this is fairly warranted , yet it seemed advisable to test it by comparing for the cosine constant , of the fifth order in the series for , with the minimum square process .
It gives for ; the second approximation , using alone , gives , which would be alittle increased by using the corrections of the other constants , so that the agleement is sufficient .
As the factors will answer for any year , I give their logarithms .
It does not seem necessary to give the constants \amp ; c. ; but instead the secondary constants of the formula , deduced from their corrected values , are given as more convenient for computation in Table VIII .
( p. 427 ) .
I have given the constants for the horary coefficients \amp ; c. to the 6th order for symmetry ; but in fact I do not think any of them less than need be attended to .
Even this limit is beyond what can be expected to be available when they are only determined by the observations of seven years , as is evident from what I have already said as to the of the quantities from which they are determined .
Whether the diurnal variation of the coordinates follows the same law in different septennial periods remains to be determined ; probably it does .
The constants belonging to and are larger than the others , and , as derived from larger coefficients , merit more confidence .
The effect of the terms of the first and second orders , which are the chief , are similar ; but the others present opposite phases , and would probably be modified by more accurate determination .
It is here that I think changes in successive years will probably be found ; and were I to pursue this work further , I would combine the observations rather differently from what I have done in the present case .
I would mean the homonymous hours of each month of each year , combine them in pairs , and mean them to get the of each month .
I would then compute the constants , retaining the cosine and form ; and this should be continued through a few periods of the solar spots .
This would decide the question whether the wind is affected by the conditions which modify those phenomena .
At the same time the inspection of the horary means would show whether their laws vary with the time .
Then the final constants could be determined for such intervals as might be considered sufficient .
The sine and cosine formula , though requiring more .
T. B. ROBINSON ON THE REDUCTION OF ANEMOGRAMS .
TABLE IX.\mdash ; Secondary Constants for S. DR. T. R. ROBINSON ON THE REDUCTION OF ANEMOGRAMS .
work in computing , has this advantage , that it permits the combining the constants obtained at different periods by simple meaning , which the sine formula does not .
It also lends itself more easily to an examination of any influence which may be supposed to change the coordinates periodically .
Any such may be developed in a similar series , and the sum or difference of the two will give the residual part which is to be accounted for by other causes .
If this residue be than the original periodic part , the hypothesis must be rejected ; and even it be diminished , this is not sufficient unless there be a priori evidence of a vera causa .
As an example of this may be mentioned one of the elements of the action .
Its heating-power on a given day depends , other , on the sum of the sines of its altitude during that day .
This sum { lat. declin .
lat. decl .
} sin lat. decl .
decl .
being the value of at sunrise .
If the value of this integral be computed for 12 values of , it can be developed in a series ) .
This belongs to the midday of each month , and in strictness to be summed for the entire month by means of the expression of decl .
in terms of ; but it is sufficient for illustration .
is evidently diminished by , and we have what would .
be found if the altitude had no effect , If , the measure of the altitude 's effect on the COOl'dinates , were known , no more would be ; but a probable value of it is that which wonld make the sum of the squares of the periodic parts of the residues or a minimum .
This gives For ; for .
lVith these I computed the sel.ies for and which need not be given , l.emarking merely that the coefficients of the first order are only ones much altel.ed .
It may suffice to give the variable parts of It seems from these numbel .
S that the sun 's nltitude may account for of the variation of , and for of that of S. This discussion the notion that the equatorial current which produccs th ' positive and cooldinates may possibly be more constant than appeal.s at and that a part of these variations may be due to a rent in the opposite cction caused by the solar action in the vicinity of the place of observation and varying with the sun 's declination .
.Supposing to be that part of it due to the altitude , its mean annual value would be , about of ( page 412 ) , and its Other periodical causes , such as the length of air traversed by the sun 's rays at different altitudes , the difference of the earth 's daily and htly radiations , and the amount of watery vapour in the air , might be similarly taken into account .
I have already stated that I thought it useless to deal with the observations of single days ; I , however , tried two experiments in this direction , which may be of some interest , the first of them was unsuccessful .
1 .
In many instances , even when the wind is moderate , there are variations in its direction which suggest the notion that they are due to aerial whirlpools on so small a scale that they are not likely to reach any other meteorological station .
I it be possible to determine the constants of such a motion in the following way .
The curve described on such a supposition by the thread of wind which passes the anemometer at a given station is that which would be traced by a pencil fixed there on a plane reyolving with an angular hourly velocity round a centre which is carlied in a line inclined at the angle to the axis of with the hourly velocity V , and the coordinates of that centre at the origin of the time , and a ) the angular motion there .
It is obvious that we hay Then at successive hours equating to tang , and to , I would be able to get values of the unknown quantities .
But against this is my ignorance of the relation between and this distance from the centre of the circle , which is not given in any book to which I can refer .
NEWTON , in the vortex which he considers , gives it inversely as the distance .
It is probably nearer the inverse square .
Either of these suppositions would make integration impossible , so I gave up the project .
2 .
The other was an attempt to determine from these observations the existence of an atmospheric tidal current .
As in the case of the ocean , so in the atmosphere , the air must be heaped up in the meridian passing through the moon , or a little to the east of it ; and this elevation must be accompanied by a horizontal current .
LAPLACE ( Mcc .
Cel .
) has shown that the maximum air-tidal current is metre in a centesimalsecond* , which in English measure and time is mile in an hour .
He , however , gives no indication of the phase of this maximum , or in what stratum of the atmosphere it occurs .
At the earth 's surface , owing to friction and other causes , it must be considerably less than the above value , and the analogy of sea-tides is too slight to give much assistance in the research .
It may , however , authorize us to assume that on opposite sides of the lunar meridian the directions of this current will be opposite .
is to bo regretted that in this noble work LAPLACE used the centesimal division of the quadrant , and the decimal and centesimal divisions of the day .
Whatever be the fate of the metric system , it is very unlikely that either of tho others will le generally adopted .
DR. T. R. ROBINSON ON THE REDUCTION OF ANEMOGRAMS .
Having no data to guide me in detecting the most favourable Lunar hours , I began by comparing the Ws for , and .
I soon , however , found that this involved too much labour , and confined myself to the last hour .
Calling the current , .
In this I made no attempt to allow for the sun 's elongation the moon , or for their declination , nor for the horary changes of the coordinates , as the selected lunars are nearly distributed in each of the 24 common hours .
By a Table with the moon 's hourly motion in Ascension for ument I found the time which should be added to the Greenwich time of its culmination to obtain the common time of the above-named lunar hours at , and entering the Journal with these I obtained for each day two values of , belonging to the upper and lower culminations .
From the ularity of these values it might seem hopeless to get any result ; but I pursued the inquiry in hopes of ascertaining the limits within which the mean of a considerable number of observations ( even very discordant ) might be depended on .
I only took the first six months of the year , as the results which they gave were quite satisfactory .
TABLE X. The weights are proportional , the least , that for January , being taken as unity .
It will be observed that these probable errors are far less than those given in ; but it should be recollected that here the variations can only occur within lunar hours , whilc in the other case they range months and years .
Even so thel.e are occasionally very great and startling changes when a gale bursts out suddenly or ceases .
There were two values of above 40 , and three above 30 .
Yet with all this I think the result is very remarkable .
I do not to assel.t that this value of really represents the tidal current at these hours , though it is in the l.ight direction and of not improbable amount ; for it may be some esidue of the horary changes .
But it is of great importance , as giving what must be a close approximation to the real value of the air-tidal stream , and as .
my former statements , that casual irlegularities are eliminated from the mean of a large number of observations .
Still I think that a truer result might be obtained by omitting extremely aberrant observations ; but it becomes a question to what extent this should be done .
I think all may be rejected which exceed four times the largest probable error ; in other words , whose probability is less than .
This is for all above 15 .
The number of these is 58 , and the results after their exclusion are given in Table XI .
TABLE XI .
The probable errors are less , and the weights greater than in the other case , so that is probably a better value than It is possible that this mode of proceeding might give the horary changes of the coordinates more correctly than the simple comparison of the numbers in Table III .
; but the labour of computation would be much greater .

