]\gt ; On Rainfall and the Sun-spot Period .
working out the resuIt it will be most convenient to use consistently C.G.S. system .
On this system of measurement the pressure was dynes per square centimetre , and therefore the expended per second in generating the waves was Now the mechanical value of a series of progressive waves is the as the kinetic energy of the whole mass of air concerned , supto be moving with the maximum velocity of vibration ( v ) ; so that , denotes the area of the wave-front considered , be the velocity of and be the density of air , the mechanical vaJue of the waves in a unit of time is expressed by , in which the value of is about 34100 , and that of about 0013 .
In the application is the area of the surface of a hemisphere whose is 82000 centimetres ; and thus , if the whole energy of the escapair were converted into sound , and there were no dissipation on the the value of at the distance of 82000 centimetres be given the equation ' centimetre per second .
result does not require a knowledge of the pitch of the somld .
If be , the relation veen the maximum excursion , and the velocity is present case the note of the whistle was a frequency of 2730 .
Hence the amplitude of the aerial particles less than a ten-millionth of oentimetre .
I am inclined to think that on a still night a sound of this pitch , amplitude is only hundred-millionth of a centimetre , would still audible .
III .
" " On the alleged Correspondence of the Rainfall at Madras with the Sun-spot Period , and on the True Criterion of Periodicity in a series of Variable Quantities .
By General STRACHEY , R.E. , C.S.I. , F.R.S. Received May 3 , 1877 .
A paper has recently been printed by Dr. Humter , the of tistics to the Government of India , having for its object to that the records of the rainfall at Madras , for a peri extending sixty-four years , establish a cycle of rftinfall that ] ) hich rked coincidence with a corresponding cycle of sun-spots\mdash ; the rainfall and sun-spots attaining a minimum iu the ellth , first , second years , and a maximum in the fifth General Strachey on the Correspondence of Irrespective of its scientific interest , the conclusion old , if sound , be of no little practical importance , as it means of indicating the probable recurrence of those seasons drought which produce such terrible results in India , and which the Madras provinces are now suffering in an extreme It is probably generally known that the conclusion which been stated is to be drawn from the Madras observations had sidered to be established some years ago , in a more general Mr. ]Ifeldrum , the Director of the Meteorological Observatory tius , a paper by whom on this subject was read before the in 1873 , and may be found in vol. .
of the ' Procsedings , ' As the numerical results of the method of treating the rainfall which Mr. Meldrum and Dr. Hunter have followed at first may appear to support the conclusions they have adopted , it has me desirable to examine the faots , with a view to arrivin at an opinion as to the trustworthiness or otherwise of those conclusions .
shall first refer to the Madras observations and Dr. Hunter 's resuIts .
The Madras register extends over sixty-four years , beginning with The mean rainfall for the whole period is inches .
from the mean from .
in defect to .
in excess .
arithmetical mean of these deviations the signs is The greatest difference between two consecutive years is the average difference ins .
Dr. Hunter , in order to test the point which he proposes to divides the sixty-four observations into six cycles of and calculates the arithmetical mean of the years series of cycles .
The results shown in the following figures in which represent the differences of the several years ' from the mean of the whole:\mdash ; the Rainfall with the Sun-spot Period .
the above calculation the first year of the of is 1813 , so the average period of maximum sun-spots will be about the third year of the cycle , and the period of minimum will be about the or eleventh of the cycle .
This Tabls certainly seems to indicate a of ximtlm between the third the seventh years , and of between the eighth and the second years .
estimate the true weight of these we must look a Iittle No the only signiIication of the arithmetical mean of a series observed quantities is that it is a quantity above and below there equal amount of deviation in the aggregate of ntions .
conformity to a law of any sort in a series of observations iu to quiLntity is ubviously to be tested by the extent of iation of quantities from the results that such .
Consein the present , the question or not the mean } alues in Table I. can be acccpted as showing a definite of } tion year to year in cycle must be determined by the differveen those and the idual observations which they based .
Treatin the ations in this manner , the results Mean of mean differences inches .
Thus it appears that the mean difference of the idual observations from the calculated means in the last line of Table I. differs but little from mean difference of the ervations from the arithmetical mean of whole series .
In other ords , the supposed law of variation obtained from the means of the six -vear cvcles hardly General Strachey on the Correspondence of gives a closer approximation to the actual observations than taking the simple arithmeticffi mean as the most probable value In order to obtain a practical test of the probable physical the cycle of eleven years , I have calculated a series of mean values sponding to those given in Table I. for a series of cycles of five , six , eight , nine , ten , twelve , and fourteen ye .
1 find that the mean ences between these means and the observed quantities , and corresponding to the mean differences in the last line of are all within a very small fraction of one another and of the tained from the 11-year cycle\mdash ; in short , that one cycle is in this almost as good or as as another .
The mean differences for the several cycles are given in the Table:\mdash ; if in any series of quantities , such as the rainfall Madras , there be a of periodicity , each ] observed quantity may posed to be compoumded of a periodical and a element .
we tnke the sum of a large number of cycles each of hich coincides the cyole of periodicity , the elements will tend to in amount excess defect , and thus to be eliminated , for the successive years of the cycle , or ) hatever the ( which I will term cyclical means ) , ill tend to indicate the elements for the successive .
At the same time , the of these cyclicsd means from the several original quantities theywere obtained will approximate to the several non-periodical These differences I will call differences .
In proportion as the periodical elements are the Rainfau with the Sun-spot Period .
correspondiug non-periodical elements , so the cyclical differences inversely less or more different from the differences between the obscrvations and the mean of the whole of them ; and if there periodicity , the cyclical means will tend to disappear , and the two of differences would , in a sufficiently long series , be identical .
Hence it may be inferred that when the differences closely apin magnitude to the mean difference of the observafrom the arithmetical mean of all of them , the periodical elements in observations must be correspondingly small ; and this applies to the whole of the cycles for which the differences are shown in III .
Further to test the reality of the supposed periodicity shown in Table I. , rearranged the series of 64 years ' observations , in a purely arbitrary in cycles of eleven years , by drawing the actual observations at one after another , and them in succession till the were exhausted .
three arbitrary sets of six cycles thus the meall c difference averaged , and indicate that , by adopting the sequence of the quantities of rain instead of takiug them at random , we prono material effect on the mean cyclical differences , nor any such to a diminution in their numerioal value as necessarily accoma true periodical element .
moreover , important to bear in mind that the mere circumstance series of cyclical means ving a single maximum and single no more real indication that such a result is a truly feature than would be supplied by the ppearance of maxima and .
The ] of periodicity , if it exist at all , can be inferred by the facts indicated by observation ; and it is obviously argue in a circle , first to assume a cycle on which to work , which give a ] maximum minimum , then to infer that there true periodicity be ause of the single maximum minimum .
The of the periodicity is in truth to be sought aItogether outside of the values of the successive cyclical means .
It is , of course , manifest that a tion of periodical elements so mask one another as to prevent positive results being obtained such an examination the cyclical means differences as I in the case before us .
But the whole scope of my present arguis , and it necessarily leads , I think , to the conclusion that variations shown in Table II .
from the mean values in I. are so great as to that any apparent regulnrity or .
tento a maximum iu one part of the cycle , and a mllllmum another , has no real eight , and that there is proof of greater General Strachey on the Correspondence of but an objection possibly be made to the effect that period is not exactly a cycle of eleven.ears , and that a better be obtained by a comparison of the observations corresponding known periods of maximum and minimum sun-spots , without to any special length of cycle .
The following Tnbles show thus arrived at , the figures representing differences from the fall for the whole sixty-four Tbese results are also tive .
The maximum period can ot to be indicated at all .
The minimum shows a diminished mean deviatiol , but this is due to the years furthest from the supposed mum epoch ; if the three central years alone are reckoned , the is much the same obtained from the 11-year cycles .
I have sought for a further test of the character of the that I have been discussing in the rainfall observations at cutta , which have been made for ths greater of the which those at ] extend .
It is hardly conceivable that there be a coincidence with the -spot period , such as is supposed to been ou any physical cause , which should some be discernible in the at Bombay and ing the same cycle for the observations at these in the case of indras , the mean ] for the three are hibited i the followi the Rainfal with the Sun-spot Period .
results entirely negative , and indicate no concordan of the several years of the cycle .
The Bombav and Calcutta treated those of Madras were , to ascertain the observations from the successive means of the cycle , give similar results .
The deviations obtained for Calcutta , manner in Table II .
, as follows:\mdash ; In these cases , in that of Madrns , the menn deviation for the whole years of the cycle very little from the mean variation of single ations from the ithmetical nleau of all of them , these being for Bombay in .
, and for in .
Although my object in the present communication is to the alleged correspondence between the dras minfall and the periods , I have turally turned my nttention to Mr. ] of a similar oharacter , lave tested some of { hem in nuer that I have iust explained .
Taking the observations for fifty-five years , at page .
of the ' of the Mr. Meldrun 's paper before noticed , and a General Strachey on the Correspondence of this case the mean deviation of the original observations from arithmetical mean of the whole series ( inches ) is inches .
result therefore is quite analogous to that obtained from the observations .
I have not attempted to make detailed calculations in other cases , so far as I can judge , the evidenoe of the alleged periodicity will be rally found to fail when it is tested by comparison with the observations on which it has been made to rest .
It will serve to illustrate the argument on which this paper is if we consider what would be the consequence of applying it to a which a well-asoertained periodicity exists , as that of the diurnal metric oscillation .
The Table gives an example , random from an old Madras registel , the intervals being made so as to reduce the number of calculations .
The entries are the ences of the observed barometric heights from the mean of the whole thousandths of an inch .
the Rainfall with the Sun-spot Period .
this case it be found that the mean tion ( disregnrding of the whole of the ations from the arithmetioal mean of whole ( which is here zero ) will be The differences between the mean values , given in the last of , for the two-hourly periods ttnd the tions as follows:\mdash ; Mean of mean reuccs 1 true periodicity existing , the mcml of called it , the , is reduc to 11 .
differences for the observntions for the below the meau tion 3 , and nono of the General Strachey on the Correspondence of Table VIII .
, and had no knowledge of the law of periodicity , and thetically applied a period of ten intervals instead of twelve ( the prooess the results of which for Madras are given in Table III inapplicability of such a period would immediately be made apparent resuIts such as are found in the rainfall observations .
On the just mentioned the mean values , corresponding to those given last line of Table , for the successive intervals of the six cycles that would be formed would be as follows:\mdash ; From these gures the following of tl IX .
The contrast between these results those givon in most striking , and calls for no comment in detaiL In conclusion I would explnin that I do not desire to in the possible or occurrence of the Rainfall with the Sun-spot Period .
to the sun-spot period , but to point out in case of the not has no such correspondence been established , but that has been no sufficient evidence adduced of periodicity all .
Appendix.\mdash ; Reccived Fay 17 , 1877 .
Suppose the numerical values of such a series of observations as that in the paper to be represented by , . . .
for the first year of each cycle of , . . .
for the second years , so on for ( m ) cycles .
Also suppose to represent the arithmetical mean of all the observaand , . . .
to be the periodical riations from the mean , for the several series of the , . .
yeals of the cycles , and the non-periodical tions to be , . . .
, , . . .
, so forth\mdash ; all these quantities being ftffected by their proper signs .
TheJl if esents the value of series , . . .
, should have for ( m ) cycles\mdash ; its represents the llon-periodical portion of the value the A series for years .
With a sufficiently prolonged series of cycles the quantities , . . .
tend to cancel one another and disappear ; so that will become equal to , and If , therefore , thel.e is a truly periodical element , the difference of the of all the observations , and the mean of any lies ( of the of periodicity , will ( in a sufficiently extended series of observittions ) to be identical with the periodical element for cycle .
This holds good of all the series , . .
; and therefore the sum of all the diflerences last mentioned for the several series ill tend to equality the sum of all the periodical elements of the several series .
This be whether we regard the sigu or not .
If we so regard it , the sum of the differences will evidently become equal to zero , will the sum of the periodicnl elements .
If disregard the the sums will have numerical , which call S. This , if by , the number of years the cycle of periodicity , will the meam tion , either excess or defect , of the periodical from the mean of the obselvations .
when there is a periodicity , the sum of these differences regard to sign tends to become an riable quantity , and magnitude of this quantity indicates the nitude of the variation .
next consider the case in which there is no truly periodical Here and .
If replesents the mean alue , irrespectiye of iign , of all the iatiollS of an indefiGeneral Strachey on the Correspon lence of litely prolonged series of observed quantities , . . .
, then the probable vnlue of deduced from the mean of series of observations of , also irrespective of sign , will the known laws of the combination of errors ) equal hold good of all the series , \amp ; c. At the same time , as periodicity , and all the observations are presumablyliable to irregularities of the snme general racter in positive or negative rection , the quantities \amp ; c. will , in a sufficiently prolonged tend to equality one ith another and also with the mean irrespective of sign , of all the observed quantities from the mean of the whole of them , which call E. Hence , whenthere is no periodicit .
, the sum its before defined , to becomo } We are thus led to the conclusion that the consideration of the of the ntily S as the number of cycles of ffords a true criterion of the presence or absence of a clement .
If as increases this ntity is graduallyreduced in a approximaling to may iDfor that the periodicity is small or not exist .
On the other , if the alue of tends to become able , and continues to be of considerable numerical prolonged series of cycles , the existence of a true periodical element apparent .
In the Madras observations the successive values of ( obtained combimDg one nfter nobher the observations of one cycle of 11 years , such cycles , of three , and so on , till the whols six united ) , contrasted with the corl.esponding calculated of 877 ] The last doubtfuI .
As the observations only extend to six cycles , in to test the process I have continued the calculation to twelve , and twenty-four cles , by combining ith the real tions three series hazard by a chance redistribution of the This of course gives support to any conclusions as to ten of peliodicity in the case of dras ; but it , in of periodicity , the theory is completely verified .
The Bombay , Calcuttn , Greenwich ions , similarly treated , results .
In strong contradistinction to the results are those that follow treating a similar manner the barometric observations given in .
; lvhile these tions , their periodicity is , as in Tables X. and XI .
, show results quitc to those seen in Table XII .
The following Table shows these resulfs , both it exists and after it has been :\mdash ; The fact that the figures in ] column but one so much less than the calculated lues of idently arises from the want of conformity he numerical values of the quantities which are found in Table } ith the assumed law of the probable numerical distlibution of errors which the alue rests ; but the mutunl destruction of the non-periodical elements than this theory would reqnired does not affect the gen eral reasoning .

