]\gt ; December 15 , 1892 .
Sir JOHN EVANS , K.C.B. , D.C.L. , .D .
, Treasurer , in the Chair .
A List of the Presents received was laid on the table , and thanks ordered for them .
The Right Hon. John Morley , a member of Her Majesty 's Most Honourable Privy Council , whose certificate had been lspended as required by the Statutes , was balloted for and:elected a Fellow of the Society .
The following Papers were read:\mdash ; I. ' On an Apparatus for facilitating the Reduction of Tidal Observations By G. H. DARWIN , F.R.S. , Plumian Proand Fellow of Trinity College , Cambridge .
Received November 12 , 1892 .
S 1 .
Introduction .
The tidal oscillation of the ocean may be represented as the sum of a number of simple harmonic waves which go hrough their periods approximately once , twice , , four times in a mean solar day .
But these simple harmonic waves may be as being rigorously diurnal , semi-diurnal , -diurnal , and so forth , if the length of the day referred to be adapted to suit the particular wave under con- sideration .
The idea of a series of special scales of time is thus introduced , each time-scale being appropriate to a special tide .
For example , mean interval between successive ctllminations of the moon is , and this interval may be described as the moan lunar day .
Now there is a series of tides , bearing the initials , \amp ; c. , which go through their periods rigorously once , wics , thrice , four times , \amp ; c. , in a mean lunar day .
The solar tides , , proceed according to mean solar time ; but , besides mean lunar and mean solar times , .
there are special time scales appropriate to the larger ( N ) and smaller ( L ) lunar elliptic tides , to the evectional , to the diurnal and semi-diurnal luni-solar tides , to the lunar diurnal ( O ) , \amp ; c. The process of reduction consisbs of the determination of the mean height of the water at each of 24 special hours , and subsequent harmonic analysis .
The means are taken over such periods of time that the influence of all the tides governed by other special times is eliminated .
346 Prof. G. H. Darwin .
an Apparatus for [ Dec. 15 , The process by which the special hourly heights have hitherto been obtained is the entry of the } observed at the mean solar hours in a schedule so arranged that each entry falls into a column appropriate to the est special hour .
ScheduIes of this kind were epared by Mr. Roberts for the Indian Government .
* The successive rearrangements for each sort of special time were made by recopying the whole of the observations time after time into a series of appropriate schedules .
The mere clerical labour of this work is enormous , and great care is required to avoid mistakes .
All this copying might be avoided if the observed heights were written on movable pieces .
But a year of observation gives 8,760 heights , and the orderly sorting and -sorting of ) early 9,000 of paper or tablets might prove more laborious and more treacherons than recopying the figures .
It occurred to , however , that the marshalIing of movable pieces might be reduced to manageable limits if all the 24 observations pertaining to a single mean solar day were moved together , for the movable pieces be once reduced to 365 , and each piece might be of a size convenient to handle .
The realisation of this plan affords the subject of this paper , and it will appear that not only is all dPsirable accuracy attainable , but that the other requisite of such a ) is satisfied , namely , that the whole computing apparatus shall serve any number of times and for any numker of places .
The first idea which naturally occurred to have narrow sliding tablets which should be thrown inbo their places by a number of templates .
It is unnecessary to recount all my trials and failures , but it .
will suffice to say that the slides and templates require the preclslon of a mathematical strument if they are to work satisfactorily , and that the manufacture would bs so expensive as to make price of the instrument prohibitive .
The idea of making the tablets or strips to slide into their places was then abandoned , and the strips are now made with short pins on their under sides , so that they can be stuck on to a drawing board in any desired position .
The teIItplates , which were also troublesome to make , are replaced by large sheets of paper with numbered marks on edition of these computation forms was reprinted by aid of a grant from the Royal Society , and is sold by the Cambridge ientific Instrument Company , but only about a dozen copies now remain .
In the course of the preparation of the " " guide sheets\ldquo ; of the method proposed in this paper , I found that there are many small mistakes in these Indian forms , but they are fortunately not of such a kind as to produce a sensible vitiation of results .
I learn that the mistakes arose from a misunderstanding on the part of a computer employed to draw up the forms .
The my guide sheets was controlled by aid of Mr. Boberts 's forms , and it was the occasional discrepancy between my results and the forms which led to the detection of the errors referred to .
1892 facilitating the Reduction of Tidal Observations .
347 them to show how the strips are to be set .
The guide sheet is laid on a drawing board , and the pins on the strips pierce the paper and fix them in their proper positions .
The shifting of the strips from one angement to the next is cer*ainly slower than when they slid thsir places automatically , but I find that even without practice it only takes about 7 or 8 minutes to shift 74 of them from any one arrangement to a new one .
The strip belonging to each mean solar day is divided by black lines into 24 equal spaces , intended for the entry of the hourly heights of water .
The strip is 9 in .
by in .
wide and t divisions by are of convenient size for the entries .
There was much difficulty in discovering a good material , bub after various artilicial ivory , or xylonite , was found to serve , the purpose .
Xylonite is white , will take writing with Indian ink or pencil , and can easily be cleaned with a damp cloth .
It is just as easy to write with liquid Indian ink as with ordinary ink , which must not be used , because it stains the surface .
The strips have a great tendency to warp , and I have two methods of overcoming this .
A veneer of xylonite on hard wood serves well , or solid xylonite may be stiffened by sheeb brass let into a slot on the under side .
In the first plan the pins are fixed in the wood , and in the second the is filed to a spike at each end .
Whichever platl is adopted , the strips are expensive , costing about ti7 for a seb , and I do not at prescnt see any way of making them cheaper .
The observations are to be treated in groups of two and a half lunations or 74 day .
A set of strips , therefore , consists of 74 , numbered from to 73 in small figures on their flat ends .
If a set be pinned horizontally on a drawing board in vertical column , we have a form consisting of rows for each mean solar day and columns for each hour .
The observed heights of the water then written on the strips .
the 24 columns are summed and divided by the number of entries we obtain the mean solar hourly mean heights .
The harmonic analysis of these means gives the mean solar .
But for evaluating the other tides the must be rearranged , and to this point we turn our attention .
Let us consider a special case , that of mean Innar time .
A mean lunar hour is about 1h .
time ; hence the of each .
day must lie within .
time of a mean lunar hour .
The sample gives the incidence to the lunar houl of the first few days in a year : 348 Prof. G. H. Darwin .
On an Apparatus for [ Dec. Mean solar Nean lunar time .
tims .
\amp ; c. \amp ; c. The successive of .
time will march retrogressively through all the 24 hours of .
lunar time .
Now , if starting from strip , we push strip 1 ons division to the left , strip 2 two divisions to the left , and so on , the entries on the strips will be arranged in columns of approximately lnnar tims .
The rule for this arrangement is given by marks on a sheet of paper 18 in .
broad ; these marks consist of parallel numbered steps or zigzags showing where the ends of each strip are to be placed so as to bring the hourly values into their proper places .
At the end of a lunation mean solar time has gained a whole day over mean luIJar time and the solar again agrees with the lunar .
On the guide sheet we see that the zigzag which takes its at the left end of strip has descended diagonally from right to left until it has reached the left margin of the paper , and a ) zigzag then begins on the right margin .
When the strips are pinned out following zigzags on the sheet marked , the entries are arranged in 48 lumns , but the number of entries each column is differsnt .
The 48 columns are to be regarded as appertaining to , . .
, , . . . .
, Thus , the numbsr of entries in the left-hand column of any hour added to .
the number of entries in the right-hand column of the same hour is , each case , 74 .
The 48 incomplete columns may , in fact , be regarded as 24 complete ones .
The 24 complete columns are then summed ; the 24 sums would if divided by the total 1lumber of strips , give the 24 mean lunar honrl heights .
The harmonic analysis of these sums certain constants which , when divided by the number of strips , are the required tidal constants .
It must be remarked , however , that , as the incidence of the entrieq is not exact in lunar time , investigation must be made of the corrections arising out of this inexactmess .
The explanation of the guide sheet for lunar time will serve mutatis mutandis , for all the others .
zags h to be placed so as to bring the columns into exact gnJnent , and orinters ' types provide all the accuracy requisite .
1892 .
] facilitating the Reduction of Tidal Observations .
349 Accordingly , the computing strips are made to suit a chosen type .
The standard length for one of the 24 divisions on the strips was chosen as that of a " " 2-em English quadrate 24 of these come to 9 inches , which ] is the length of a strip .
I found the English quadrate a little too narrow , and accordingly between each hne of quadrates there is a " " blind rule of 42 to the inch .
The depth of the guide sheet is that of 74 quadrates and 74 rules , making in .
The computing strips are in .
broad , and 74 of them occupy in .
The excess of 15above , or in .
, is necessary to permit the easy arrangement of the strips .
To guard against the risk of the computer accidentally using the wrollg sheet , the sheets are printed on coloured paper , the sequence of colours being that of the rainbow .
The sheets for days to 73 are all red ; those for days 74 to 7473 , or 147 , are all yellow ; those for days 148 to 14873 , or 221 , .
are green ; those for days 222 to , or 295 , are blue ; and those for days 296 to , or 369 , are violet .
Thus , when the observations for the days of the year are written on the strips all the sheets will be red ; the strips will then be cleaned , and the observations for the second 74 days written in , when all the guide sheebs will be yellow , and so on .
I must now refer to another considerable abridgment of the process of harmonic analysis .
It is independent of the method of arrangement just sketched .
the Indian computation forms the mean solar hourly heights have been found for the whole year , and the observations have been rearranged for the evaluation of certain other tides governed by a time scale which differs but little from the mean ] scale .
I now propose to break the mean solar heights into sets of 30 days , and to analyse them , and next to harmonically analyse the 12 sets of harmonic constituents annual and semi-annual inequa ] ities .
By this plan the harmonlc constants for 11 diflerent tides are obtained by one set of additions .
In fact , we now get the annual , semi-annual , and solar tides , which formerly demanded much troublesome extra computation .
A great saving is secured by this alone , and the results are in close agreement with those derived from tlJe old method .
The guide sheets marked and the computation forms are arranged so that the observations are broken up into the proper groups of 30 days , and they show the computer how to make the subsequent calculations .
I have also devised an abridged method of evaluating the tides of long period , Mf , Mm .
The method is less accurate than that followed erto , but it appears to give fairl ood resuIts and reduces the work to very small dimensions .
Prof. G. H. Darwm .
On an for [ Dec. 15 , Before entering on the details of my plan it is proper to mentiou that Dr. Borgen has devised and used a method for attaining the same end .
He has prepared sheets of tracing paper with diagonal lines 011 them , so arranged that when any sheet is laid On the copy of the observations written in daily rows and hourly columns , the numbers to be snmmed are found written bebween a pair of lines .
his plan is excellenb , but I fear that the difficulty of adding correctly diagonal lines considerable , and the comparative faintness of figures seen through tracing paper may be fatiguing to the .
Dr. Borgen 's plan is simple and inexpensive , and had I not thought that the plan now proposed has considerable advantages I should not have brought it forward .
In the investigations which follow the notation of the Report of 1883 to the British Association on harmonic analysis is used without explanation .
S 2 .
liivaluation of , Sa , Ssa , , P. The 24 mean solar hourly heights of water are entered in a schedule of 24 columns , with one row for each day , extending to days ; the 24 columns are summed , and the sums divided by ; the 24 means are rmonically analysed ; it is required to lind from the results the values of the harmonic constituents .
The speed of any one of the tides differs from a multiple of per hour by a small angle ; thus , any one of the tides is expressible in the form , where is , 1 , 2 , 8 , \amp ; c. , and is small .
When lies between and this formula expresses the oscillation of level due to this tide on the day of the series of days .
If muItiples of from 1 to be added to , the expression gives the height at the same hour , , of mean soIar time on eaoh of the succession of days .
Then if denotes the mean heighl of watsr , as due to this tide alone , at the hour , we have .
( 1 ) .
When is put successiveIy equal to , . . . .
, we get the 24 values of I ) which are to be submitted to harmonic analysis .
The mean valoe of , say ( not to be confused with as written at the head of this section , where is denoted the mean sea level above datum ) is found by taking the mean of the 24 values of By the for the summation of a series of cosines it is easy to prove that 1892 .
] facilitating the Reduction of Tidal Observations .
coss [ . . .
( 2 ) .
We will now find the harmonic constituents .
By the dinary rules .
( 3 ) .
Now Sln and of the sum of the 24 values corresponding to , 1 , . . . .
, is same with sign of changed This expression admits of simplification , because ; making this simplification , and cing the resuIt into ( 3 ) , we obtaiu ( 4 ) .
In the particular case where , we have . . . . . . . . . . . .
( 5 ) .
Prof. G. H. Darwin .
On an Apparatus for [ Dec. } If the number of days be large , will be small unless the denominator of oue of the two tsrms in ( 4 ) be very small .
This last case can only ocour when and when is small .
Hence , in the analysis of a term of the form under consideration , we may neglect all the harmonics except the one .
Accordingly ( 2 ) and ( 5 ) are the only formulae required .
A case , however , which there will be occasion to use hereafter is when , when ( 4 ) becomes .
For the present we have to apply in the two cases and ; now the ratios of to in these two cases are 722 to 1 and 697 to 1 .
In both cases the first term of ( 5 ) .
is .
ible compared with the second .
Now write . . . . . . . . . . . . . . .
and becomes , with sufficient exactness , . . . . . . . . . . . .
If this be compared with ( 2),. .
we see that when this formuIa also comprises ( 2 ) .
In the applicationsto be made is very small , so that is approxinately a function of the form .
This function increases very rapidly when passes , but for considerable values less than it only slightly exceeds unity ; for example , when but when infin .
It follows , therefore , that if the number of days in the series is snch that 12 is less than say , the magnitudes of are but little diminished by division by ( 4 ; but if 12 is nearly become vanishingly small .
If the typical tide here considered be the principal lunar tide and if the number of days be as nearly as possible an exact multiple of a semi-lunation , 12 is nearly , and the corresponding become very small .
No number of whole days can be an exact mul- tiple of a semi-lunation , so that corresponding to cannot be made to vanish completely .
For the present they may be treated as negligible , and we return to this point in the next section .
The above vestigation shows that in the expression for the whole 1892 .
] facilitating the Reduction Ticial Observations .
oscillation of sea level upon which the proposed analysis is performed all those tides may be omitted from which is not very small , and all those whose frequencies are such that the period under consideration 12 is nearly Since the perlod under consideration will be a lunation , it follows that , as far as is now material , the general expression for sea level may be written as follows , denoting mean solar hour angle equal to , annual . . . . . . . . . .
semi-annual. . . . . .
Solar tides , Solar elliptic , . . . . . . . . . . . . . . . .
Lnni-solar , Solar diurnal , . . . .
- . . . . . .
This includes all the tides whose inibials are written at the head of this section .
It is now necessary to break up the year into 12 equidistant lunations of 30 days .
This can be done by the omission of 5 days in ordinary ears , and of 6 days in leap years .
If the days of the year are numbered to 364 ( 365 in leap year ) , the twelve months are as follows : \mdash ; to to ; omit to to ; omit to to ; omit to to 242 ; 8 , to ; omit to 10 , to ; omit to ; in leap year omit The increments of sun 's mean longitude from of month up to of the day numbered of each group of days or month are as follows:\mdash ; ; 1 , ; 2 , ; 3 , ; 4 , ; 5 , ; 9 , Thus if be the sun 's mean longitude at of month , the sun 's mean longitude at of month is , with sufficient approximation .
354 Prof. G. H. Darwin .
On an paratus for [ Dec. 1 Now let with appropriate suffix denote the initial ' ' equilibrium argument\ldquo ; at of month , so that ; then the general expression ( 9 ) for the tide in the month becomes . . . . . . . . . . . . . . . . . . . .
( 10 ) .
Each of these terms falls into the , and is in every case either , or Now , when harmonic analysis of the moan of 30 days is carried out , coefficients are introduced .
Write therefore With the known value of In applying the method investigated above , it lvill be observed that a term of any frequency only contributes to the harmonic constituent of order that 3ritten asThen aying.term b follows :\mdash ; 1892 .
] facilitating the Reduction of Tidal Observations .
; ; . . . . . . . . . . .
( 11 ) .
With the meaning of the term month in the present context , the sun has a mean motion of per month , and each of the first five 's and 's is a function with a constant part and with annual and semi-annual inequalities .
When 7 has successively the 12 values , 1 , . .
, 11 , we have 12 equidistant values of the 's and .
These may be harmonically analysed for annual and semi-annual inequalities .
ppose that the several coefficients to be determined by harmonic analysis are defined by the following equations:\mdash ; aos ; . .
( 12 ) .
Then on comparing ( 12 ) with ( 11 ) we see that:\mdash ; ; 356 Prof. G. H. Darwin .
On an Apparatus for ) cossin ; . .
... . . . . . . . . . . . . . .
( 13 ) .
From these equations we get ; ; ; ; . . .
( 14 ) .
The tidal constants of the several tides enumerated at the head of this section are determinable from these equations .
Our rule is accordingly to allalyse in twelve groups of SO days , and then to analyse the resulting harmonic constituents .
for annual and semi-annual inequalities , combining the final results according to the formula just found .
The edition of\ldquo ; guide sheets\ldquo ; and computation forms which I have drawn up are so arranged as to facilitate the whole process and to render-it quite straightforward .
By this single set of additions it is thus possible to evaluate eleven tides and mean water .
S 3 .
Clearance of from perturbahon by The method of the last section was designed to render all the tides insensible excepting those enumerated .
But is so much larger than any other tide that there is a small residual disturbance which ought to be corrected .
I have made computations , which I do not give , but which show that the disturbance of all the harmonic constituents except 32 , facilitating the Reduction of Tidal Observations .
is insensibls .
It is required then to determine the correction to bo applied to 32 , , and thence those for Ei , Suppose that , when time is counted from mollth tide is expressed by , where When means taken over 30 days are analysed the formula ( 6 ) gives the contributions to 32 , .
As it is now required to obliterate these contributions , the must be changed , and the corrections are . . . .
, , ( 15 ) .
For reasons which will appear below I now write Then introducing the value of into ( 15 ) , I find . .
( 16 ) .
Let denote the value of at of month , and let , and let denote a certain factor whose logarithm is , and let In the harmonic analysis for the tide , considered below in S6 , lve shall have .
Accordingly These values of must now be oduced into ( 16 ) , but the algebraic process need not be given in detail .
If we write VOL. LII .
2 Prof. G. H. Darwin .
On an Apparatus for [ Dec. it follows that Then , when the substitution of the values of is carried out , we find By the definition of it appears that is the increment of twice the mean moon 's hour angle during the time from of month up to of month , that is to say for the time specified .
The following table gives the values of and of its cosine and sine for each month:\mdash ; No. of days Nonth from epoch .
epoch . .
) .
335 If , are regarded as quantities having annual and semiannual inequalities , we may write On .
the numerical values of by the ordinary processes , I find , But in S 2 the .
constituents of when analysed for facilitating the Reduction of Observations .
allnual and semi-annual inequality were denoted by , F2 , and the constituents of were denoted .
Hencs the ten corrections to the 's and 's are ( with an easily intelligible alternative notation ) On substituting the numerical values of , ( 3 , I find Coefflt .
of .
Coeffit .
of Most of these corrections are negligible , but the four which affect the solar elliptic tides must be included , because those tides are so small that a small error affects them sensibly .
Hence we ma take , with sufficient accuracy , , where are the components of the derived from the reduction that tide by the process of S 6 .
for these corrections is made in the computation forms .
S 4 .
Jiivaluation of , Sa , , when a complete year of observation is not ilable .
It is now proposed to consider the case where the period of observation is as much as six complete months and less than a complete year .
Prof. G. H. Darwin .
On an Apparatus for [ Dec. method of the last section apparently depends on the completeness of the year , with certain modffications , it may rendered available for shorter periods .
We suppose that so much of the year as is available is broken into sets of 30 days by the rules of the last section , and that the means are cally analysed .
The resuIts of such harmonic analysis for month are given in ( 11 ) of S3 , but for the purpose ]land they now admit of some simplification .
It is clear that it is not worth while to evaluate the very small solar elliptic tides and from a short period of observation .
If , then , we denote by the ratio of the cube of the sun 's parallax to its mean parallax at the middle of the month , the first three terms of the third of may be included in the expression .
The last term of this equation really does involve the solar parallax to some extent , and we may , with sufficient approximation , write the third pair of equations Let us now consider the value of .
The longitude of the solar perigee is or , and the ratio of the sun 's parallax to its mean parallax is approximately , and tlJe cube of that ratio is or .
Now , the sun 's longitude at the middle of month , is ; hence and .
Thus it is easy to compute the values of for the successive months , when we know the sun 's mean longitude at of the month The semi-annual tide , being usually small , may be neglected in tlJese icomplete o , and the equations ( 11 ) now become , 1892 .
] facilitating the Reduction of Tidal Observations . . . . . . .
When the series of successive values of the a and 's are harmonically analysed ( by processes which we shall consider shortly ) the several coefficients resulting from such analysis will be defined by Mean , Mean Mean , Mean . . . .
( 18 ) .
Then the subsequent procedure as given in ( 13 ) and ( 14 ) holds good , the only differsnce being that we do not obtain the semi-annua ] and solar elliptic tides .
We shall now consider the harmonic analysis of an imperfect series of values .
It must be premised that each monthly value of is to be divided by its corresponding before the analysis is made .
Suppose that denotes a function which is subject to semi inequality , and that Then it is clear that \amp ; c. \amp ; c. I now define tbus : Prof. G. H. Darwin .
On an Appap.atus for ) If there be equations and if they be treated by the method of least squares , we get , .
These are the three equations from which are to be found .
A schedule is given below for the formation of , and \amp ; table of the solutions of these equations to the number of months available .
Next , suppose denotes a function which is subject to annual inequality , and that Then \amp ; c. , \amp ; c. In this case the method of least squares gives , , . .
Tables are given below for the formation of , and of the solutions of the equations according to the number of months available .
1892 .
] th Reduction of Tidal Prof. G. H. Darwin .
On an Apparatus [ Dec. Rule for finding from an incomplete series .
Number of months available .
Coefft .
of .
Coefft .
of .
Coefft .
of Rule for finding annual from an incomplete series .
Number of months available .
Coeffl .
of .
Coefft .
of .
Coefft .
of -0.098 1892 ] facilitating the Rednction of Observatio We thus get the following rule for the evaluation of from 6 , 7 , 8 , 9 , or 10 months of observation:\mdash ; Proceed as though the year were complete and find the 's and 's for as many months as are vailable .
Reduce the by multiphcation by or .
Analyse inequality , and for according to the rules for reduction of incomplete seriesjust given .
Complete the reducXion as in S3 .
These rules for reduction do not include the case of 1 ] months , the case where any month in the series is incomplete ( e.g. , if a fortnight 's observation were wanting in one of the months ) , because these cases may be treated thus:\mdash ; the a and 's return to the same value at the end of a year , and therefore the case of eleven months is the same as that of a missing month at any other part of the year .
In both these cases we may interpolate the missing 's and treat the year as complete .
If three or more weeks of observation missing they might fal ] so as to spoil two months , and in this case we should have an incomplete series .
It is then to be recommended that the equations of least squares be formed and the equations solved .
So many similar cases may rise that it does not seem worth while to solve equations until the case arises .
S 5 .
of from short period of observation .
If the available tidal observations only extend over a few months , it is useless to attempt the independent evaluation of those which we have hitherto found by means of annual and semi-annual inities in the monthly harmonic constants .
We will suppose that 30 days of observations are available .
Then when we neglect the annual tide , and the solar ( meteorological ) tide , we have from ' or ( 17 ) , which give the analysis of 30 days , , .
It is now necessary to assume that tide has the same 366 Prof. G. H. Darwin .
On an Apparatus for [ Dec. lb , amount of retardation as the , and that the ratio of their amplitndes is the same as the equilibrium theory .
We also make the like assumption with respect to the and tides .
Accordingly we put ; Now since we have .
Therefore , Let us put . . . . .
( 19 ) .
Then .
( 20 ) .
If therefore we have ' 1892 .
] the of . .
( 21 ) .
If there be several months available it is recommended that each 30 days be treated quite independently , so that from each group of days we shall get and .
Then the mean vaelue of is to be taken as the final value of that function , and is to be treated similarly ; finaliy are to be found .
The several values of may be treated in the same way .
Of course we assume throughout that , assumptions which are usually nearly correct .
The mean value of must be taken as giving , bnt at ] aces with a considerable annual tide it is impossible to obtain a good value of mean water mark from a short series of observations .
S6 .
On the evaluation of the several tides by of mean solar days .
Let denote the speed in degrees per .
hour of any one tide , being equal to 1 , 2 , 3 , 4 , 5 , or 6 .
Then may be called one " " special hour Since is one .
hour , the ratio of the .
to the special hour is .
Let one .
hour be equal to special hour , then ' special hours .
Let it be required to express the of any .
day of a series of days by reference to special time .
It is clear there .
time will be specified by one of the 24 special hours , with something less half a spscial hour added or subtracted .
Having fixed the of .
time of a particular .
day in the special time scale , let us treat that .
day as a whole , and consider the incidence of the other 23 .
hours in special time .
It is clear that in .
time we work backwards and forwards from by subtracting or adding unity , and that in special time we subtract or add If .
time be , where lies between special time , the following is a schedule of equivalence:\mdash ; Prof. G. H. Darwin .
On an for [ Dec. 15 , Mean solar time .
Special time In the column of special time it is supposed that is added or subtracted , so that the resuIt is less than .
For example , if is 10 , the hour column of special time will rnn , .
, , . . .
, If the series of days be long will have all integral values between and 23 with equal frequency , and since has all values between and - with equal fiequency , the excess of the solar hour above the 1learest exact special hour ( which may be called the error ) will have all its possible values with equal frequency .
If the mean solar hours be arranged in a schedule of columns headed , .
, of special time , each column will be snbject to errors { which follow the same law of fi.equency .
facilitating the of Observations .
369 Let abscissae ( fig. 1 ) measured from along represent magnitude of Since lies between , the limit of the figure is given by OA OA ' If magnitude of error ( .
\mdash ; special hour ) , measured in special time , be represented by ordinates , a line BOB ' at to AOA ' represents all the errors which can in the incidence of the in the schedule of special time .
If a Iine be drawn parallel to and above BB ' by a distance we have a representation of all the errors of incidence of the If a series of equidistant parallel lines be drawn and below until there are 12 above and 11 below , then the rors of all the .
hours are represented , the top one showing the errol'S of the m.s. and the bottom one the of the Any special hour corresponds with equal frequency with each solar hour , and hence each mode of error occurs with equal ency .
It is now necessary to consider in how many ways an of given magnitude can occur .
If in the figure AM represents all of given magnitude , then wherever MN cuts a diagonal line , it shows that an error may arise in one way .
It is thus clear that there are no greater than and no \mdash ; errors reater than , and Errors of magnitude .
to may arise in 1 way .
to , , 2 ways .
to -ays .
to , , ways .
to , , 24 ways .
to , , 23 YYays .
to , , 2 ways .
to , 1 way .
The frequency of error is represented graphically in fig. 2 .
] slope of the two staircases is drawn at , but any other slope would have done equally well .
A frequency curve of this form is not very convenient , and , as there are many steps in the ascending and descending slopes , I substitute the frequency curve shown in fig. 3 .
This is clearly equivalent to the former one .
In fig. 3 all the times shown in fig. 2 are convorted to aIlgle at to the hour ; accordingly denotes 370 Prof. .
H. Darwin .
On an Apparatus for [ Dec. lb Now let be the observed value of a function whose true value is , and suppose that , the error of , has a frequency ; then the mean value of the function deduced from many observations will be In our case is the ordinabe of ths frequency curve whose abscissa is Let OQ , QB , OA , OA ' ; then .
1892 .
] facilitating the Reduction Observaiio The algebraical steps involved in the evaluation of these four integraIs and subsequent simplification are omitted .
Hence the resuIt is By reference to the figure it is clear that Write , then , and we obtain as the mean value of , when found in this way , .
It is obvious that if we had begun with , the argument in the result and the factor would have been the same .
Accordingly , a function would yield the result If 24 equidistant results of this sort are submitted to harmonic analysis to find , we shall get\mdash ; , suppose , , suppose .
Accordingly But it is required to find , so that no .
Thus when the 24 observed hourly tide heights on any .
day are regrouped so that the observed height at .
time is reputed to appertain to an exact special hour , and each of the previous and subsequent hourly values of that .
day are reputed to belong to previous and subsequent exact special hours ; and when a long series of m.s. days are treated similarly , and when the mean heights of water at each of the 24 special hours are harmonically analysed , we shall obtain the required result by augmenting by a factor , and by subtractfrom 372 Prof. G. H. Darwin .
On an Apparatus for [ Dec. 16 , The values and of ne will be different for each kind of tide , and the followlng table gives their numerical values .
Table of and ne .
Initial of tide .
1 2 3 4 6 2 2 2 1 1 1 MS 4 2 2 001 MK 3 MN 4 5 0 .
02632 As it does not appear worth whils to evaluate the tides written below the line , no use will be made of the last six results given in this table .
S 7 .
On the per iods over which the , neans are to be taken in evaluating the tidal constants .
We have considered in previous sections the treatment of the group of tides which are associated with solar time , when the period of observation is less than a year , and we have now to consider the other tides .
It is important that the means be taken over sueh a number of days that the perturbation arising from other tides shall be minimised .
The perturbation between semi-diurnal and diurnal tides is always negligible .
It is thereiore only necessary to consider the action of the tides in the case of semi-diurnal tides , and that of and for diurnal tides .
1892 .
] faciliiating the Reduction of Tidal Observations .
It is easy to see that the influence of a disturbing tide is evanescent when the means are taken over a period such that the excess of the argument of the disturbed over that of the tide has increased { hrough a multiple of .
As , however , we are working with integral numbers of days , and as the speeds of tides are incommensurable , this condition cannot be exactly satisfied .
From this consideration it appears that to minimise the bation of by ( and vice ) we must stop an exact multiple of a semi-lunation .
To minimise tlIe effect of M2 on and , and of on and , we must stop at an exact multiple of a lunar anom- alistic period .
To minimiso the of M2 on , we stop at a multiple of bhe period 2 .
To minimise the effect of on , we must stop at an exact nlultiple of a semi-lunar period .
For the quater-diurnal tide , , it is immaterial where we stop , and so it may as well be taken at a multiple of a semi-lunation .
The table ( p. 374 ) gives the rules derived from these considerations .
S 8 .
On tloe tides of long period .
The annual ( Sa ) and semi-annual ( Ssa ) tides evaluated in the course work by which other important tides are found .
These are the ] two tides of long period which have a importance in respect to tidal prediction , but the luni-solar fortnightly , the lunal ' fortnightly ( Mf ) , and lunar monthly ( Mm ) tides have a theoretical interest .
It will therefore be well to show how they may be found .
The process is short , and , although it is less accurate than the laborious plan followed in the Indian reductions , it appears to give fairly good For the sake of simplicity , let us consider the tide .
Its period is about 14 days , and therefore a day not differ very largely from a twelfth part of the period .
Accordingly , if about two days in a fortnight axe rejected by proper rules , the mean heights of water on the remaining days may be taken as representatives of twelve equidistant values of water height .
-therefore go ough the whole year and reject , according to proper rules , the daily sums of the 24 hourly heights corresponding to certain 69 of the days out of 369 .
The values are written consecutively into a schedule of 12 columns and 25 rows , of which each corresponds to a half .
The 12 columns summed , and the sums are harmonically analysed for the first pair of harmonic components .
These components have be divided by 24 times 25 , or by 600 , because the daily mean water height is of the sum , and ]lere are 25 semi-lunations .
In the same way the semi-lunal period is abo days , and if VOL. Prof. G. H. Darwin .
On am Apparatus for [ Dec. uded in the evaluation of the several over any period up to a year .
For , Q. For with one with one of the following of the following days days ( anom .
periods ) .
( periods we erase by proper rules 45 daily sums out of 369 , we are left with 324 , which may be written consecutively in a schedule of 12 columns and 27 rows , of which each corresponds to , semi-lunar period .
The summing and analysis is the same as in the last case , but the final division is by 24 times 27 , or by 648 .
1892 .
] the Rpduci of Observations .
375 In this way we evaluats the Iuni-solar fortnightly and lunar fortnigbtly inequalities in the height of the water .
The period of the moon is between 27 and 28 days , and if we erase appropriately about one day in eight are left with sets of 24 values which may ) taken as 24 equidistant values of the daily sums .
Accordingly we erase 46 daily sums out of 358 , and write the 312 which remain consecutively into a schedule of 24 columns and rows , of which each corresponds to alunal anomalistic period .
The 24 columns are summed and the sums analysed for the first components .
Finally , the components to be divided by 24 times 13 , or by 312 .
In this way the lunar monthly tide is evaluated .
But the resul obtained in way is , as far as concerns the tide , to some , and it may be to a large , extent fictitious .
It represents , in fact , a residuum of the principal lunar tide .
That this is the case will now be proved .
Suppose that is an integral number of days since epoch , being the time of noon on a certain day ; then the principal lunar tide that day may 'oe written is less than 24 hours .
Then the daily sum for that day will be Now since is an integlal number of days 2 only differs from by an exact multiple of 36 ; hence the argument of the cosine may be written 2 But the true luni-solar fortnightly tide , which we may denote , varies so in the course of a day that the daily some is sensibly equal to It thus appears that the effect of .
is of exactly the same form as that of MSf .
It becomes , therefore , necessary to clear harmonic components , determined as described above , from the effects of In order to determine the values of these clearances , I found the values of cos2 and sin2 for every noon a year of 369 days .
I then erased the values selected for the treatment of MSf and analysed the remaining values .
In this way it was easy to find the effect of the known tide .
Suppose that are the first harmonic components determined by the treatment of a series of daily sums , and are corrections to be applied to them to eliminate the effects of , then I find that if are the two components of as determined by the previous method ( S 6 ) of analysis , Prof. G. H. rwin .
On an Apparatus for [ Dec. 15 , Whence being known from Baird 's manual ( being a function of the longitude of moon 's node ) , and are determinable .
We have also In the set of computation forms which I have prepared for use on the present plan , it is shown what days are to be erased for each of the three analyses , and how they are to be entersd in schedules , summed , and analysed .
S 9 .
On abridgment in the computations .
It seemed probable that one decimal of a foot would suffice to express the hourly tide heights .
In order to test this , I have taken several individual days of observation ab Port Blair , and have found , by harmonic analysis , the time and amplitude of the diurmal and semi-diurnal H.W. , first , when the hourly heights are expressed to two decimal places of a foot , and second ] , when they are only entered to the nearest tenth of a foot .
I find that the times of H.W. agree within less than a minute of time,.and that the amplitudes agree within a fraction of an inch .
If this much be true of individual days , the difference of results arising from two or one place of decimals will clearly entirely disappear when a series of days is considered .
Hence , by taking as unit the tenth of a foot , or the inch , or even two inches at places with large tides , we may always express all , or nearly all , ths heights on which to operate by two signfficant figures .
The adoption of this rule not only saves the writing of a large number of figures , but also enormously diminishes the ]abour of the additions which have to be made .
It also seemed probable that substantial accuracy might be attained from the harmomc analysis of only 12 hourly values insbead of 24 .
In order to test this I took the tidal reductions for Port Blair , Islands ( kindly lent me by the Survey of India ) , and have compared the results which would have been derived from 12 values with those actually obtained from 24 values by the computers of the Indiar Survey .
following tables give the results:\mdash ; 92 .
] facilitating the Reduction of Tidal Observations .
Semi-diurn tides .
tides .
378 Prof. G. H. Darwin .
On an Apparatus for [ Dec. The mean discrepancy in the case of the semi-diurnal tides is , and the greatest is ; in ths case of the diurnal tides the mean discrepancy is , and the greatest is In tidal work results derived from different years of observabion differ far more than do these two sets of results , and hence the analysis of 12 two-hourly values for diurnal and semi-dinrnal tides gives adequate results .
I find that this abbreviation does not give satisfactory resuIts for quater-diurnal tides , and the sixth harmonic is not derivable from 12 values .
Therefore , when these tides are to be evaluated the 24 hourly values must be used .
It will still be necessary to write all the 24 hourly heights on each computing strip , but when the strips are put into any one of the arrangements , except where quater-diurnal tides are required , we need only add up the columns , 2 , 4 , . . . .
, 22 , and may omit the columns 1 , 3 , . . . .
, 23 .
S 10 .
On a trial of the proposed metlwd of reduction . .
already mentioned , I have tidal reductions for one year ( beglnning April 19 , 1880 ) for Port Blair , Andaman Islands .
I am thus able to make a comparison between the results of the old method and of the new .
The computation was , in large part , done for me by .
Wright .
It appeared sufficient to evaluate the tides of the series and those allied with them , the tides of the ssries , and the tide ; also the tides of long period , Mf , Mm .
The series test the new process of harmonic analysis of monthly harmonic components for annual and semi-annual inequalities .
I chose because it is the most important tide , and because it puts the proposed method of grouping to a severe test , and is very small in ude .
In the time scale the day is of mean solar time , from which it follows that one of the 24 mean solar hourly observations may fall as much as away from the exact hour to Tvhich it is reputed to belong .
Thus the hourly observations are arranged in wide groups round the hours , and the hypothesis involved in the method is put to a severe strain .
Lastly , the resuIts for tides of long period test my proposed abridgment .
It will be seen in the table on p. that the two methods give results in close .
There is , vever , a sensible discrepancy in the tide , but in this case I am inclined to accept the new value as better than the old one .
This tide is governed by sidereaI time , which differs ittle from solar time .
Hence , in the Indian 92 .
] facilitating the Reduction of Tidal Observations .
Port Blair .
1880-81 .
method .
eight ) . . . . .
003 000 . . . . . . . . .
hese are derived from 1880-82 .
Prof. G. H. Darwin .
On an Apparatus for [ Dec. 1$ ; Tides of Long Period .
method of grouping , considerable errors of incidence of the hours in the time scale prevail for many days together , and the method seems of doubtful propriety .
The same is true of the tide , and here also the two methods give somewhat different ] The accuracy with which the very small tide comes out , whether from 24 yalues or only from 12 , is surprising , and may perhaps be , to some extent , due to accident .
It shows , however , that the present method may be safely applied , even when the special time scale differs considerably from mean solar time .
The resuIts for the tides of long period are quite as close to the old values as could be expected .
S 11 .
A comparison of he work involved in the new and old methods of reduction .
It has been usual in the Indian reductions to use three digits in expressing the height of er , and there have been 15 series , or even more .
Now is 394000 ; tlencs the computer has had to write that number of figures in reducing a year of observation .
This does not include the evaluation of the annual and semi-annual tides , so that we may say that there have been about 400,000 figures to write .
I propose to express the eights by two digits , and they only have to be written once .
Thus , the present plan , the number of figures to write is , or 17,500 .
Thus the writing of 382,000 figures is saved .
In the old method the computer had to add together all the digits written , say , 394,000 additions of digit to digit .
I propose to use 24 hourly in three series , and MS , and 12 two-hourly values in eight others .
Therefore the number of additions will be or 123000 .
Thus 270,000 additions are saved .
1892 .
] facihtating the Reduction of Observations .
We may say that merly there were about 800,000 operations ( writing and addition ) , and that in the present method there will be about 140,000 .
This estimate does not include a saving of several thousands of operations in obtaining the tides of long period .
I am therefore within the mark when I claim that the work formerly bestowed on one year of observation will now reduce at least five years .
It has been found that the manufacture of my computing strips of xylonite is rather expensive , but as it formerly cost in England rather more than to reduce a year of observation , the cost of the apparatus will be covered by the saving in the reduction of a single year , and it will serYe for any length of time .
S 12 .
On the completion of the record for short and gaps .
In any long series of tidal observations there are usually some breaks in the record in consequence of the of the clock of the tide gauge , or from some other cause .
Now the process of elimination by grouping depends essentially on the completeness of the record , and it is therefore necessary to fill in blanks } ) interpolation .
Such interpolation has not been usual in the opel.ations of the Jndian Survey , and it might be thought that the complete omission of the missing entries is the proper course to take ; but it is easily shown that this treatment is exactly equivalent to the assumption that the water remained stagllant at mean sea level during the whole time of stoppage of gauge .
It is obvious , therefore , that any conjectural values are better than none .
The process by which it is proposed to intel.polate is best shown by an example .
At Port Blair ming April the column of from days 99 to 112 gives the heights in the first column of the table below .
I suppose that the tide broke down on day 103 , and only came into action again on day 110.* There lvas really no break down , and the actuality the supposed hiatus is } in the last column but one .
Now if we look back about a month we find that the water stood about the same height at the same hour of the day .
Then the " " previous record\ldquo ; }hich is complete beginning at is elltered in the next column .
Similarly a " " subsequent record\ldquo ; is found about a month later , and is entered in a third column .
The mean of the previous and subsequent records is then taken as giving values to be interpolated .
* The days are here numbered from 1 , instead of ] .
Tbis has been the usage in India hitherto .
Prof. G. H. Darwin .
On an Apparatus for [ Dec. 15 , The last two columns contain a comparison between the interpolatiou and what in the present case we know to have been actuality .
There is a mean error of ft. Thus it is clear that a fair record may be interpolated even with so long a break as a week .
In this example I have only shown the terpolation for one column , but of course all the other twenty-three columns would really have to be treated similarly .
I find by trial that the result would be a little improved by a iphical method , but that process is slightly more troublesome than the numerical one .
Table of Interpolation .
Defective P It may happen that the hiatus is too long for treatment in this way .
I do not think it would be safe to treat much more than a fortnight by interpolation .
It has been shown in S 4 how the tides associated with are to be treated where the record is deficient , and it remains to consider the other tides .
In S 7 are given the days with which we must stop in the analysis of an incomplete year , and this table aifords us the means of treating a long hiatus in the vation .
We may in fact omit all the entries between any two of the numbels given in the table seriously affecting the resulb .
Let us suppose , as an example , that the tide gauge broke down on day 210 and was only repaired and in operation again on day 226 .
Now 210 is 14862 , and 225 is 2223 .
Then we see by the table in S7 that in finding the means for 1892 .
] facilitating the Reduction of Observations .
383 , MS , when the computing strips are written for whird lime , we must remove strips 59 , 60 , 61 ( which have numbers written on them ) and may leave the remaining strips of that writing are blank .
When the strips are written for the fourth time strips , 1 , 2 , 3 will be blank , but we must remove strips 4 to 13 inclusive .
When all the strips are used in a complete year there are 369 , and this is the divisor used in obtaining the harmonic constants , but when there is this supposed hiatus we do not use 15 strips of the third writing and 14 strips of the fourth writing , so that the divisol ' will be ( 340 .
Again , when we evaluating in the third wliting , strips 57 , 58 , 59 , 60 , ust be removed , and in the fourth writing strips 4 to 9 inclusive .
In a complete year the divisor is 369 , but we now do use 17 strips of the third writing and lO of the fourth writing , so that the divisor becomes 342 .
Again , in evaluating , in the third writing we remove strips 45 to 61 inclusive , and in the fourth strips 4 to 25 inclusivs .
The divisor is reduced from to 303 .
Lastly in evaluating , in the third writing to 61 inclusive are removed , and in the fourth writing xips 4 to 31 inclusive .
The divisor is reduced lrom 350 to 287 .
Any hiatus , be it long or short , may be treated in this way , but it is clear that if it be short enough to treat by interpolation it is best to adopt tlJat method .
INSTRUCTIONS FOR USING THE COMPUTING APPARATUS .
The apparatus for the reduction of tidal observations , together with tion forms , can be purchased from the Cambridge Scientific Instrument Company ab a price ( as far as can be now foreseen ) of about In case of any insufficiency in the following instructions recourse must be taken to the preceding paper .
On the degree of accuracy requisite in the hourly heights .
If will usually be sufficient if the heights be measured to wilhin one-tenth of a foot , and the decimal point may , of course , be omitted in computation .
This gives amply sufficient accuracy at a place where the semirange of the principal lunar tide is 2 , and where spring is from 6 .
to 7 At some places with small tides a smaller unit might be necessary , and at others with very large tides a unit of 2 in .
, or of a fifth of a foot , might suffice .
Prof. G. H. Darwin .
On an for [ Deo .
) , Whatever unit of length be taken it is important , for the @aving of work , and it is sufficient , that all or nearly all the heights should be expressed by two digits .
Completion of record .
If there is an accidenlal break in the record , it is very important that it should be completed according to the method shown in S12 , or by some other equivalent plan .
The computation forms are drawn up on the supposition that the year of observation is complete , but with proper alterations , which will now be indicated , they may be used in other cases .
In S4 it is shown how to treat the tides of the group when the observations have been subject to a long stoppage in the course of the year , and also when the observations extend over any period from six months to a year .
In S5 it is shown how to treat the group for a short period of observation .
If the stoppage be a long one , the method explained in S 12 must be adopted for all the other tides .
The same section also shows the treatment for observations extending over any period , long or short , less than a year .
Entries summations .
The computing strips are intended to take writing in pencil or liquid Indian ink , but not in common ink .
They are to be cleaned with a damp cloth , and a little soda may be put in the water if they become greasy .
Lay the red sheet on one board and set up the strips with their ends abutting against the corresponding numbers .
The strip numbered 60 is also to be put on the board .
Writs the hourly heights for each day on the strip bearing the corresponding number , strip for day , strip 1 for day 1 , and so on up to strip 73 for day 73 .
The 24 hourly heights are to be written in the 24 divisions of each strip , beginning on the left with and ending on the right with Remove strip 60 .
Sum the 24 columns formed by the divisional marks on consecutive sets of 30 strips .
Thus , days to 29 afford 24 sums ; days 30 to 59 afford the second set of 24 sums ; days 61 to 73 afford 24 sums , which are the beginning of a third 30 , to be completed when the second set of 74 days shall have been written on the strips .
The numbers , 1 , 2 , . . .
, 23 , , 1 , . . . .
, 23 at the head and foot of the guide sheet indicate the hours corresponding to the coInmns .
The sums of the columns on the board are to be entered in the 1892 .
] facilitating the Reduction of '17idal .
columns of the form " " Hourly sums of seriss in twelve months Lay the red guide sheet on the other drawing board , and transfer the strips from the first board to the new arrangement shown by the zigzag lines , strip 60 being now reintroduced .
There will now be 48 columns ( more exactly 47 , since one of the columns will be found to have nothing in it ) , numbered at top and bottom , 1 , . . .
, 23 , , 1 , . .
, 23 .
Each of the 48 columns is to be summed bottom to top ( not as for in groups of 30 ) , and the snms are to be entered in the form " " Sums of series M The 24 sums which come from the left half of the board will be entered in the row mark-ed " " red left those from the right in the row marked " " red \ldquo ; Lay the red sheet on the other board , and transfer the strips .
In accordance with S 9 , it will now usually suffice to sum only the columns appertaining to the even hours , 2 , 4 , . .
, 22 ; as these hours repeated twice , there be 24 columns to sum .
The sums are then to be entered on the form " " Sums of serifflN in the alternate columns .
The complete form is , so that all the 24 hourly values may be used if it be thought desirable , but this labour seems unnecessary , at least in a long series of observations .
Lay the successive red sheets on the vacant board , transfer , , and enter , until all the sheets are exhausted .
In the case of , MS the sums of all the columns are necessary , but in the other eight angements only the sums of the alternate columns , those of the even hours , are usually necessary .
For a series of observations it may be best to use all the columns , but in this case it will certainly not be worth while to attempt the evaluation of , which are all small in amount .
If tides of long period , Mf , Mm are required , the 24 numbers written on each strip must be added together , and the sum entered in the form " " Long period tides\mdash ; daily sums Clean the strips .
In exactly the same way work through the ncxt 74 dsys , from to , with yellow guide sheets .
Then clean the strips , and take another 74 days with green guide sheets , and so on with the blue and violet .
In the last ( violet ) set attention must be paid to the rules as to the places where the analysis is to stop in each arrangement .
If the year of observatio1I is so incomplete that the hiatus cannot be made good by interpolation , or if the series does not run over the complete year the series must stop with one of the days specified in the table at the end of S 7 , and a note must be made of the number of days used in each series .
The strips marked for omission on the violet sheets , or those 386 Prof G. H. Darwin .
On an Apparatus .
[ Dec. selected for omission under rules of S 7 , may be hidden by a sheet of paper when the summations are being made .
The additions of , MS may be verified by proving that the grand total of all the numbers ( inclusive of omitted strips in S ) written in each of the sets of 74 days is the same in whatever way they are arranged .
Thus , the sum of the 48 columns should be equal to the sum of the daily sums .
An incomplete verification in the other .
arrangements , when only half the columns are summed , is found by showing that the sum of all the hourly sums of each 74 days is nearly equal to half the grand total of all the numbers written in that period of 74 days .
When the guide sheets become worn with many pin pricks they may easily be patched with adhesive paper .
There seems no reason why this patching should not go on almost indefinitely .
* * It is possible that it may be desired to evaluate the tides OO and , for which guide sheets are provided .
I hsrefore , give instructions for bhe preparation of guide sheets for these cases .
Ihey will be understood by any one has the set of guide sheets before him .
With the instrJctions given below , the compuler might indeed set up the strips without a guide-sheet .
I describe the staircase as descending from left to right or from right to left , and I define a short step as being one space down and one space to the left or right , as the case may be , and a long step as one space down and two to the left or right , as the case may be .
When I say , for example , that a short follows 2 , I mean that 2 to 3 is a short step .
The first mark on sheet is specified by its incidence in the row of hours at the top .
OO ; descending from left to right .
The sequence is long several times repeated and then short .
Red ; between and ; shorts follow 2 , 7 , 13 , 19 , 24 , 30 , 36 , 41 , 47 , 53 , 58 , .
Yellow ; between and ) orts follow 1 , 7 , 12 , 18 , 24 , 29 , 35 , 41 , 46 , .
Green ; between and ; shorts follow , 6 , 12 , 17 , 23 , 29 , 34 , 40 , 45 , 51 , 57 , .
Blue ; between and ; shorts follow , 5 , 11 , 17 , 22 , 28 , 33 , 39 , 45 , 50 , .
Violet ; between and ; shorts follow 5 , 10 , 16 , 21 , 27 , 33 , 38 , 44 , 50 , .
The last strip used for a year is 72 .
: descending from right to left .
The sequence is long , long , short , long , lon , short , and at intervals three longa and a short .
Red ; between and ; shorts follow 1 , 4 , . . . .
, 16 ; 20 , 23 , . . . .
, 35 ; 42 , . . . .
, 57 ; 61 , 64 , ... , 70 .
Yellow ; between 18 ] ) and ; shorts follow 2 ; 6 , 9 , . . . .
, 21 ; 25 , 28 , . . . .
, 40 ; 44 , 47 , . . . .
, 69 ; 63 , 6 . . . .
, 72 .
Green ; between and : shorts follow 1 , 4 ; 8 , 11 , . . . .
, 26 ; 30 , 33 , . . . .
, 45 : 49 , 52 , . . . .
, 64 ; 68 , 71 .
Blue ; between and ; shorts follow , 3 , 6 , 9 ; 13 , 16 , . . . .
, 28 ; 32 , 35 , . . . .
, 50 ; 54 , 57 , . . . .
, 69 ; 1892 .
] facilitatin9 the Tidal Obsert.ations .
Hourly sums and analysis .
Complete the summations in the forms for hourly sums , and copy into the forms for harmonic analysis .
In this copying it will gene-suffice if the last figure in the hourly sums be omitted ; for example , if the observaticns are entered to the nearest tenth of a foot the hourly sums will be given in bhe same unit , and it will suffice if the hourly sums analysed be written to the nearest foot .
There are 12 anftlyses ( one for each month of 30 days ) for the sums in , and one ] ysis for each of the other 10 arrangemenbs .
All the forms are provided with spaces for 24 hourly sums , but in the eight series , where only 12 values will commonly be used , the entries will only be made on the alternate rows of , . . . .
, .
In these cases bhe divisor 12 , which occurs in the penuIbimate stage of finding the 's and , must be replaced by 6 .
The large for for for ; and 350 for v ) represent the number of days under reduction , and must be altered appropriately ( see table , S 7 ) if there be long gaps in the observations , or if the year be incomplete , or if the series be a short one .
If some one of the monthly analyses of is deficient the missing 's are to be made good by interpolation .
' It is then necessary to analyse the monthly values of the 's derived from the 12 analyses of S. We thus obtain .
The rules for these analyses when the year is incomplete are given in S 4 , and the computation forms only apply to case of the complete yeal ' .
Astronomical data final reduction .
Determine from the 'Nautical Almanac ' and Major Baird 's ' Manual of Tidal Observations ' astronomical data at .
on day , and proceed according to the form to find the initial arguments and factors for reduction .
The astronomical data are then to be used in the forms for final reduction .
We have generally ; forms are arranged so that colog A is to be added to to find tall , andthence lies veen - and or between 13 and is added to A to find ; if lies between and or between and ) is added to to find B. Violet ; between and follow 2 , 5 , . . . .
, 14 ; 18 , 21 , . . . .
, 33 ; 40 , . . . .
, 52 ; 56 , 59 , . . . .
, 71 .
The last strip used for a year is 61 .
* Taylor Francis , London , 1886 , price On the Reduction of Tidal Observations .
[ Dec. 15 , Accordingly the computation form has , room being left for the syllable " " \ldquo ; if necessary ; underneath this is written and the computer will insert A or as the case may be .
There is usually required also a numerical factor or , or both ; the logarithms of the are found amongst the astronomical data , and the logarithms of the constant 's are printed in the forms in their proper places .
The treatment of the 21 harmonic components derived from the harmonic analysis of the 5 a and 's is shown in the forms .
The tides of long period .
The processes involved in the evaluation of these tides are sufficiently shown in the forms .
Postscript .
December 17 , 1892 .
Correcfion to previous paper .
An error has been detected on p. 338 of my paper " " On the Harmonic Analysis of Tidal Observations of High and Low Water ' Roy .
Soc. Proc vol. 48 ( 1890 ) .
In the example of reduction certain tabular values , extracted from Baird 's 'Manual , ' have the wrong signs attributed to them .
Since in 1887 the longitude of the moon 's node lay between and ths signs of should be positive , instead of negative as on p. 333 .
When this error is corrected the reduction ieads to the following compal'ison:\mdash ; Computed .
Mean of 9 1892 .
] On some Reptiles from the Elgin Sandstone .
If calculations had been conducted by rigorous methods , the two columns would have agreed nearly with one another .
I may mention that I have copies of a table of up to 90 days ( see p. 304 of the paper here referred to ) which I shall be glad to give to any one actually engaged in the reduchon of H. and L.W. observations .
II .
" " On some new Reptiles from the Elgin Sandstone By .
T. NEWTON , F.G.S. Commuuicated by Sir ARCHIBALD GEIKIE , F.R.S. Beceived November 28 , 1892 .
( Abstracb .
) During the last few years a number of Reptilian remains have been obtained from the EIgin Sandstone at Cnttie 's Hillock , near Elgin , which are now in the possession of the Museum and of the Geological Survey .
These specimens represent at least eight distinct skeletons , seven of which undoubtedly belong to the Dicynodontia , one is a horued Reptile , new to science .
All the remains yet found in this quarry are in the condition of hollow moulds , the bones themselves having eutireIy disappeared .
In order , therefore , to render the specimens available for study , it was necessary , in the first place , so to display and preserve these cavibies that casts might be taken which would reproduce the form of the original bones .
Gutta-percha found to be the most suitable material for taking these impressions ; and in some instances , especially in the case of the skulls , the casts had to be made in several parts and afterwards joined together .
The first specimen described is named Gordonioe Traquairi ; it is the one noticed by Dr. Traquair in 1885 , and referred to the Dicynodontia ; besides the skull , it includes fragmentary portions of other parts of the skeleton , and is contained in a block of ndstone which has been split open so as to divide the skull almost vertically and lonooitudinally .
The two halves have been so deveioped that casts made from them exhibit the left side and upper surface , as well as the main parts of the palate and jaw .
In general appearance this skull resembles those of Dicynodon and Oudenodon .
The nasal are double and directed laterally ; the orbits arelarge and look somewhat orwards and upwards .
The supra-temporal fossa is , and bounded above by the prominent parieto-squamosal crest , belo by the wide supra-temporal bar , which extends downwards posteriorly to form the pedicle for the articulation of the jaw .
There is no lower temporal bar .
The maxilla is directed downlvards and forwar to end in a small tusk .
Seen from above , skull is narrow in the inter-orbital and nasal regions , but wide posteriorly across the temporal bars , although the brain-case itself is very

