]\gt ; On Newcomb 's Method of Investigating Periodicities .
349 dissipation of due to magnetic hysteresis is lower than in the normal ltaUoy specimen for a given value of the magnetic induction .
The above experiments were made in the Sir William Siemens Electrical Engineering Laboratory , University of London , King 's College .
I wish Go bhank my laboratory attendant , .
George Jones*for the assistance which he rendered .
On 's Method of Investigating iodicities its Application to Briickner 's Weather Cycle .
By Prof. ARTHUR SCHUSTER , Sec. R.S Received April 2Read During the last few years of his life Prof. Simom Newcomb was keenly interested in the problem of periodicities , and devised a new method for their investigation .
This method is explained , and to some extent applied , in a paper entitled " " A Search for Fluctuations in the Sun 's Thermal Radiation through their Influence on Terrestrial Temperature The importance of the question justifies a critical examination of the relation.ship of the older methods to that of Newcomb , and though I do not agree with his contention that his process gives us more than can be obtained irom Fourier 's analysis , it has the advantage of great simplicity in its numerical work , and should prove useful in a certain , though I am afraid , very limited field .
Let represent a function of a variable which we may take to be the time , and let the average value of the function be zero .
Newcomb examines sum of the series where , etc. , are definite values of the variable which are taken to lie at equal distances from each other .
If the function be periodic so as to repeat itself after an interval , the products are all squares and each term .
is positive .
If , on the other hand , the periodic time be , each product will be negative and the sum itself therefore negative .
It is easy to see that if be varied continuously the sum of the series passes throug.h maxima and minima , and the maxima will indicate the periodic time , or any of its multiples .
VOL. XC.\mdash ; A. 2 350 Prof. A. Schuster .
Newcomb 's Method of Investigating Newcomb takes , for convenience , successive values of to be equal multiples of the interval , so that if , etc. , now represent the successive values of the function , it is sufficient to calculate the sums the general term being If the sum .
be the first to give a maximum , a periodicity with time is indicated .
The relationship of Newcomb 's method to that based on Fourier 's analysis , is easily established .
Writing give to successive values , etc. , and perform the summation .
If the total range of time included be large compared to , the sum involving the second term of the right-hand side will be sensibly equal to the corresponding sum of the third term and independent of .
The variation in the value of the sum of the left-hand term , which is twice Newcomb 's sum of products , depends therefore on the variation of The interpretation of this is obvious , if we make use of the optical analogy to ' which I have drawn attention in previous communications .
For in any train of waves we may identify with some vector such that energy passing unit surface at that point is proportional to the integral of taken between the proper limits .
If in the expression ( 1 ) the sum is replaced by an integral , we see that it can be identified with the energy of two superposed equal trains of disturbances , one retarded with reference to the other through a time interval , and the variation of ives us the interference pattern formed by a grating with only two reflecting spaces .
There is , therefore , no difference in principle between Newcomb 's method and that of the periodogram , both having their analogies in the spectroscopic resolution of light ; but while the periodogram aims at separating true periodicities from accidental ones by high resolvingpowers , Prof. Newcomb restricts the resolving power to its lowest possible .
value .
This renders it quite ineffective as a detector of hidden periodicitiea Its use , if any , must lie in a different direction .
The class of problems for which Newcomb devised his method is that , which contains what he calls a " " tendency\ldquo ; towards a periodicity .
As an example , he gives the fluctuation of ocean waves , or the swing of Periodicities and its Application to Bruckner 's Weather Cyde .
351 pendulum which is subjected , at intervals , to impulses altering its phase and amplitude , and argues that here Fourier 's method might give " " no period whatever This statement is not correct .
In the case contemplated , the harmonic analysis would give the resultant ( taking account of difference of phase ) of all the Fourier terms obtained when each group of undisturbed oscillations is treated separately ; such group , consisting of a limited number of oscillations , does not constitute a homogeneous system , but contains a number of periods adjacent to the principal one .
The resultant of the combined groups for each period can be estimated by applying a theorem due to Lord which reduces the question to one of probabilities .
According to this theorem , there is a definite expectancy for the resultant amplitude , which can be expressed in a simple manner , if the amplitudes due to separate groups are equaL In that case , the expectancy for the resultant is times the common amplitude , if be the number of groups , and although it may happen that for one period the resultant is zero , this is compensated by amplitudes greater than the in the adjacent periods , the average for all these periods being perfectly definite .
In the periodogram method the average intensity is taken for closely adjoining frequencies , and this avoids all ambiguity or possibility of failure .
In optical language , the disturbance contemplated by Newcomb has a spectrum containing one band which may be broad if the number of undisturbed oscillations be small .
For a spectrum of this kind small resolving powers may be as good as large ones , and much better if account be taken of the diminution of labour involved .
But in cases such as occur most frequently , we must be careful to remember that what may look like a single broad band may really be a superposition of several bands which the low power is not capable of resolving .
We should therefore use the method of low resolving powers cautiously , as a first exploration , to see what kind of disturbance we are dealing with .
For definite periods standing out feebly against a continuous background high powers are essential .
There is one distinction between an optical or acoustical in of waves and the disturbances which are contemplated in the present discussion .
In the former cases we have no knowledge of the actual displacements and the separation of periods by optical analysis or by resonance is the only method available .
III hunting for periodicities in cosmical phenomena we start on the other hand with a direct knowledge of the vector to be anaJysed , and it might be urged that this additional source of information allows us to apply a more direct treatment than that based on Fourier 's analysis .
My reply is * On the Resultant of a Number of Vibrations of the same Pitch and Arbitrary Phase ' Phil. Mag vol. 10 , p. 73 1880 ) .
352 Prof A. Schuster .
Newcomb 's Method of Investig that in the which I have mainly had in view , the additional information is of no assistance , the periodicities to be looked for being hidden by accidental or at any rate irregular disturbances .
There are , however , interte cases between obvious periods , such as the diurnal change of temperature , and hidden periods such as that involved in the possible synchronism of sunspots and meteorological phenomena .
A case in point is the lnspot curve .
A period in the neighbourhood of eleven years is clearly indicated without recourse to analysis , by direct inspection of the variations and the number of spots , but the regular cycle is interfered with and considerable variations in the distances between successive maxima are observed .
Fourier 's analysis here may possibly be supplemented by other investigations , though it cannot be replaced by them .
When , however , the curve of variability gives us no indication by direct inspection of the periodicity looked for , the periodogranl method seems to me to be the only available one .
About 25 years ago , Dr. Eduard Bruckner , investigating secular changes of climate , came to the conclusion that " " the variations of climate consist of fluctuations of temperature , barometric pressure and rainfall , which occur simultaneously over the whole earth in a period of 35 years Examining the evidence on which this very definite conclusion is based , it occurred to me that the periodicity claimed\mdash ; if it has a reality\mdash ; must be of the kind for which Newcomb 's method is specially applicable .
The average distances between snccessive maxima or minima in the series of observations on which Bruckner bases his conclusions vary very considerably , and though there can be no question of a strict period , there seemed sufficient evidence of groups of periods near 35 years to render further investigation desirable .
In the meantime meteorologists generally have accepted Bruckner 's cycle , and sometimes used it as a basis for predicting a succession of cold or warm winters .
The records collected by Bruckner for the frequency of exceptionally cold winters date back to the year 800 and are summarised in a table in which progressively from 6 to 5 years the numbers of these winters in an interval of 20 years are given .
Thus the figure attached to the year 1430 represents the number of cold winters between 1420 and 1440 , and the next figure attached to the year ] is the number of cold winters in the interval 1425 to 1445 .
It is clear that this process eliminates any period of 20 years or of its submultiples , and considerably diminisfles the amplitude of periodicities up to 26 or 30 years .
In applying Newcomb 's procedure to Bruckner 's series of figures , I have formed the summation of the products , increasing the values of.p

