]\gt ; S$ffl ] Harmotric Analysis of 'idal Observations .
270 When a question of this kind is attacked , the solution cannot be deemed complete unless the investigation is left in such a state that an ordinary trained computer is able to use it as a code of inf ; tructions by which to reduce a series of observations , without any knowledgs of tidal theory .
An actual numerical example is thus essential , botb to test method and to serve as ructions to a computer .
The Appendix contains so much of the reduction of three montbs of observation at Bombay as will ssrve as a code .
If the scries be longer than three months , or in such cases as the proper treatment of gaps in series , it is necessary to refer back to the body of the paper for instructions .
I now pass to the theoretical reasons for the rules for reduction .
S2 .
Notation .
The notation of the Report to the Britisb Association for 1883 , and in use in the Indian tidal work and elsewhe ] is here follo The earth 's angular velocity is denoted by ; the mcan motions of the moon , sun , and lunar perigee by ; the mean longitudes of moon , sun , lunar perigee by , and the mean solar hour anglc by .
The R.A. and itude in the lunar orbit of the intersection of the equator the lunar orbit are ; and is the longitude of the moon 's nocle .
The several hal.monic tides are denoted by arbitrarily chosen letters .
Those with which we shall principally have to deal are\mdash ; Semi-diurnal .
Name .
Initial .
Speed .
gument Principal lunar 2 solar 2 Luni-solar Larger elliptic Smaller Diurnal .
Lnni.solar Lunar Solar The symbol denotes the mean semi-range of any one of the tides , and its retardation of pbase behind what it would be accord- lng to the equilibrium theory ; denotes a cerf , ain factor of augmentabion of the lunar and luni-solar tides depending on the value of N. The particnlar tide to which refer will in eral be indicated a subscript small letter , the same as the letter constitut .
take typical terms involving and , and let the rates of incrcase of be , and of be Then we have . .
( 1 ) .
Since at H. or L.W. is a maximum or a minimum , we must have\mdash ; .
( 2 ) .
Let us write . . . . . . . . . . . . . . . . . . . . . .
( ) Then multiply ( 1 ) by and ( 2 ) by , and add ; and .
multiply ( 1 ) by and ( 2 ) by , and subtra , ct , and we ve\mdash ; ( 4 ) lsm .
] Harmom'c of Tidal Observations . . . . .
Also let Then our equations are\mdash ; ( 6 ) . . . . . . . . . . . . . . .
A pair of equations will result from H. and } When a series of tides is sidered , we may take the mean of and substitute a mean The general principle ) is to take the means over such periods that the mean become very .
In fact , shall , in several cases , be able to them so far that these are nogligible , and get simply ; but in other cases , where what typified , the tide is a small one , whilst one or more of the as is , it will bo ecehsary to The finding of these is clearly reducible to the mean -alues of .
useful principle may be illustrated thus : if the tide does not diIJer lch in speed trom the tide , we may put where a small speed .
Then we write If ve , the condition for maximum and imum i iunction with gives , .
, , lvhere are ceriain constants , depending on the sum of a trigonometrical series .
Again , if we take means from to , the second terms hnve their signs changed .
Hence the difference between these two successive sums will give and .
There will be usually two terms such as those typitied by , and we shall then have to take two other means , , one crinning at and ending at , and the other beginning at and ending at .
From the difference of these sums we get and .
From these four tions the two 's and the two 's are found .
soIution is a little complicated in reality by the fact that it is not possible take exactly the beginning of the ser.ies , because the first tide does not occur exactly at noon , but this is a detail which will become clear below .
When all the 's and 's or 's and 's have been found , the posi tion of the sun and moon at the epoch , found from the Nautical Almanac , and certain constants found from the Auxiliary Tables in Baird 's ' Manual of Tidal Observations , are required to complete the evaluation of the 's and The of the processes will become clear when we the various tides .
It may be worth mentioning that I have almost completly evaluated the 's and , which give the perturbation of one on another , in the case considered in the Appendix .
Without any of the details of the laborious arithmetic involved , it may to say that { he conclusion fully justifies the omission of all thow terms , which are neglected in the computation as below .
S 4 .
The tides and L. These the two lunar elliptic tides .
For the sake of brevity all the tides excepting ffl omitted from the analytical expressions .
Since * Taylor and Fraucis , Fleet Street , 1886 .
IS90 .
] Harmonic Analysis of Tidal Observations .
} the expression becomes Hence , taking into account the equation which expresses a maximum or minimum , and neglecting the variation of colllpared with that of , we have\mdash ; .
The mean interval between each tide and the next is houl Then if be the increment of in that period that with equal to per hour , is equal to ) , and if be the values of and at the time of the tide consideration , the equations to the tide approximately\mdash ; If we take tho mean of successive tides , the two latter tern } on the right of ( 9 ) will be multiplied by ' and the in , will bc equal to .
If the tido exactly a semi-lunar-anomalistic period later than the first , .
On account of the incommensurability of the angulal velocity th condition cannot be rigorously satis , but if whoIe series of observations be broken up into such then on the of many such summations it may be taken true .
Then , since is a small angle , , and sirl ; hence the factor is ; thus , if is the mean number of a period , our mean cquations are\mdash ; , , M. . . .
( 11 ) , now denotes summation of the following kind : \mdash ; the numbers ( 1 ) , ( 2 ) , \amp ; c. , indicating the number of semi-lunar anomalistic-periods over the partial sums are taken .
Suppose the whole of observations to be reduced covers quarter-lunar-anomalistic pel'iods , which we denote by \amp ; c First suppose that the semi-period denoted previously by 1 of , that 2 consists of , and so on .
Let be the time of the first tide of the se , and since noon of the day epocb , cannot be more than a ffi hou , a small angle . .
Analysis of rTidal Observations .
ondly , the semi-lun{r-{nomalistic indicated by 1 ( jonSits of iiiii , that 2 consisbs of , and so on .
Obviously the result is by writiog for or , what amounts to the same hing , by outting place of ; / ; but we must also write for , so as to show th the summation begins when is nearly equal to .
Then\mdash ; .
( 14 ) .
Hence , ] .
These eqnations give the four unknowns , and equal to 'rhen if denote the equilibrium arguments of the tides at epoch , we have\mdash ; , where the mean longitudes of moou , sun , and erigee at , and and small angles , functions of the gitude of rll00ll'S node in 's Manual .
Tllen if the factor of reduction ( also tabulated by ) for the tides In this investigation the interferences of solar and ides cted , on the assumption that they are complete eliminated .
The betwoon a lunar period and anomalistic period small that tho eliminatiou of the lrllal tides will be but the effect of the solar tide will probably be sensible , unless we have under reduction l3 istic periods , which only exceed 6 semi-lunations by about 25 hours .
The evaluation of the elliptic tides and from a series of than a quarter year would bo very If we take the mean of successive tides , the equations ( 9 ) give us approximately\mdash ; .
We here assume that in taking this mean over an exact number of -lunations , the lunar elliptic tides , the solar tides , and the diurual tides are eliminated .
Witb respect to the elliptic tides , this condition can only bs proximately satisfied , because no small numbol .
of semi-lunations is equal to a number of anomalistic , and the like is true of the diurnal tides .
In the example given below the diurnal tides are much arger than the elliptic tides , and I have found by actual ( the details of which are not , however , given ) that the disturbauce iu the value of the tide arising from the diurnal tides is quite insen sible , and it may be safely accepted that the same is true of the turbance from the elliptic tides .
With respect to the disturbance arising from the principal solar tide , I find that it is adequately , although not completely , eIiminated by making the number of tides under summation 2 an exact number of semi-lunations .
If the whole series of observations be short , it would be pedantic attempt a close accuracy in results , and we may accept these rmulae ; if the series be long , the residual errors will be completely eliminated .
We have then\mdash ; If ?
be the equilibrium argument at epoch , we have .
, and Analysis of Tidal Observations .
meauings of , have been explained in the last section .
S 6 .
The Tides These are the principal solar and luni-solar semi-diurnal tides .
If the tide is the same phase as at any time , three months later they are in opposite phases .
Hence , for a short series of observations , the two tides cannot be separated , and both must be considered together .
It is proposed t , o treat a series of observations as made up of a succsssion of short series ; hence I begin with a short series .
For the sake of evity the tides and are omitted from analytical expressions .
Since , Hence , taking into account the equation which expresses that is a ximum or minimum , and neglecting the variation of or compared with that of , we have\mdash ; , .
The mean interval between each tide and the next is .
Then be the increment of in that period ( so that with equal to per hour , 9 is equal to ) , the equations corresponding to the tide are approximately\mdash ; ' ( 17 ) .
Now , if be the cube of the ratio of the sun 's parallax to its mean allax , the expression for , together with its paralIactic inequality ( the tides T , R of hatmonic notation ) , is .
Since is the mean solar hour angle , is the same thing as Henoe Also if be the value of at epoch , then for a pel'iod of two three we may take , where is equaI to Prof. CI .
H. Darwin .
, if we put , we have Also since the argnment of tide is is a certain function of the longitude of the moon 's node lated by Baird ) , and since at epoch , it follows that Now , when the means of the equations ( 17 ) are taken for } Huccessive tides , the latter teruls become , wher9 . . . . . . . . . . . . . . . .
( 18 ) .
Also , if we write . . . . . . . . . . . .
( 19 ) , onr equations become\mdash ; ( 20 ) .
It may be observed that the mean value of during the interval embraced by the tides .
In reducing a short series of observations we have to assums what is usually nearly true , , that and , as would bo the case in the equilibrium theory of tides .
With this hypothesis , put which to find and .
Tben from which to find and Lastly , 18 Analysis of 'fidul .
289 In to minimise the disturbance due to the lunal tide , we have to make the tides cover an exact number of semi-lunations , namely , the as that in volved in the evaluation of The elimination of the tide is adequate , although not so complete as the elimination of eifcct of tide on , because is nearly times as as A Lonq Series of .\mdash ; Snppose that is a half year of observations , or two periods of six semi-ltl1lations , each of periods contains exactly the same nlImber of tides .
Then of these periods is to ) educed i with the assumption that and If this assunlption is found subsequently to be rect , it be to amend reductions by by , and by adding to ; but such will not usually be nec , .
From these reductious we geb indcpclldent of si from year , and the mean of those is to be adopted , llroln which to compute and remains to and The factor and the so slowly ] be neglected from one quarter to the next , each quarter supposed to have reduced with oper v idues .
Let and be the 's mean at the tlJey ill clearly by nearly , and we put Hence it is clear that the of in second rter is the four equations , such as ) , to two quarters , may be , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , \ldquo ; , ( 21 ) , where the acccnted symbols apply to the socond quarter , and where , a constallt .
From ( 21 ) , , .
omitted from the analytical expressions .
If denotes , we have , and , Hence , taking account of the equation which expresses that is a maximtlm or miuimum , and neglecting the variation of compmd with that of we have\mdash ; , .
The mcan between each tide and the next is Harmonic Analysis of Tidal Observations .
201 will be mnltiplied by and the last term by the ] function with in place of ; also the in the arguments must be put equal to Tf the tide falls ] a semi-lunar period later than the .
On account of the incommensurability of { he angular velocity , this condition cannot be ] satisfied , but if the whole series of observations be broken up into such semieriods , then , on the average of many such mmations , it may taken as true .
Since is a small angle , , and ; hence the first factor is equal to ' in degrees ; and Therefore , suppose .
Again let .
7048 , . . . . . . . . . . . . . . . .
( 23 ) , and we have , . . . . . . . . . . . . . . . .
( 24 ) .
Thus , if is the mean numbe .
of tides in a semi-lunar period , h means of equations ( 22 ) become . .
, the summations are carried out over the first semi-lnnar , which may be designated as 1 .
If , therefore , we put , we obtain the result from $ by .
the signs on the left and writing for The equations for semi-periods 3 , 4 , 6 , \amp ; c. , tvill be alternately * and \mdash ; on the left , and identical as regards the terms in and with , \amp ; c. , successively in place of Let the observations run over semi-lunar periods ; then doub the equations appertaining to periods 2 , ; , add all th equations together , and divide by .
The in will involve the sel.ies This is equal to .
Then if we pnt , where our equations ( 25 ) become ' ( 3S ) , where now denotes summation of the kind : \mdash ; ffarmonie A nalysis of Tidal Obs .
293 Since ; and from , .
( 28 ) .
If the same notation be adopted as that explained in S4 ( the on being thnt we llow deal with lullar istead of quarter-anomalistic ) , we have , , . . .
( 29 ) .
Secontll the period indicated by 1 consists of consists of , and so .
Then , the result is got by for ; that is fo , write for and , or for .
But is equal to , and we for following the notation used in St for and .
These four 's require correction for the ) ) the to the semi-diurnal terms and , and I retu ] 1t to this point later .
In the antimo write X , and we have\mdash ; , .
If put .
( 34 ) , Prof. G. H. the equations ( 38 ) may be written \mdash ; The four equations ( 82 ) , .
( 38 ) involve six unknown , and are lnsufficient for tlJeir determination .
In reducing a short series of observations it is necessary to assume what is usually nearly t , rue , , that , and as wo1lld be the case in the equilibrium theory of tides .
Then , writing for , we have approximately The argument of the tide is , where is certain function of the longitude of the moon 's node ( tabulated by Baird ) ; and the argument of the tide is At the noon which is as epoch , and the two arguments are equal to and Hence \mdash ; efore Putting ' as explained above , where , a small angle . . . . . . . . . .
( 36 ) .
Then , if ( 37 ) , we have .
Let whence and may be computed ; and ' From these we compute and and Harmonic of idal ) ) ations .
, the equilibrium argument at epoch We have also Returnjng to equations ( 34 ) and ( 35 ) , we -compute and hence , and then and Having these , we compute and from .
Then , if , the argument at epoch of , and where is a certain function of the itude of the moon 's node , tabulated in Baird 's Manual .
A Long Series of Observation \mdash ; Suppose that there is a half year observatiou , or two periods of thirteen quarter-lunar periods , each of which exactly the same number of tides .
Then each of these periods is to be reduced independently the assumpbion that and .
If this assumption be found subsequently to be very incorrect , it might be necessary to amend these reductions by adding to the value of , and by ltiplying , by , but such repetition will usually be necessary .
these reductions we get independent values of from each quarter year , and the means of these are to be adopted from which to compute It remains to evaluate and The factor and the angle vary so slowly the change from one quarter to the next may be neglected , although each is suPposed to have been reduced with its proper values .
Let be the values of the sun 's mean longitude at the two epochs ; then since second epoch is nearly a year latel th an the first , will exceed by about Lst , so that is small .
If be the values of at the second epoch , have , and therefore Again , , and therefore Let be the values of and corresponding to the second epoch , and let ' be the values of those quantities in the second quarter .
Then , replacing by that is its value when is 13 , we have from ( 33 ) , , .
Hence In these tions R is equal to and is equal to The terms involving are clearly small , and approximabc values of and , as derived from the first quarter , will be sutticient to put them .
Afterwards we can compute or ; then if denotes , the equilibrium argument of at the first epoch , The Lues of thus deduced ought not to differ very largely from assumed in the two independent reductions .
The same serves for the evaluation of the tide from any two sets of observations , each consisting of thirteen periods , and with a small change in the analysis we need not suppOS6 to consist of thirteen such periods .
But the two epochs mustbe such that is small and is large , or the formulae , althoug analytically correct , will fail in their object .
Since , the eighteen groups forming ths quadrant may be thrown in with the quadrant by a change of sign ; and the like is true of the and quadrants .
Since and , it that we have to go through the quadrant in reversed order , in order to fall in with the succession which holds in the quadrant , and , moreover , the cosine changes its sign , whilst the sine does not do so .
Hence the following schemss will give us the eighteen group which all have the same cosines and sines:\mdash ; for cosines reversed , for sines reversed .
Thus , one grouping of the heights servss for both cosines and and , save for the last step , the additions are the same .
The combination of ths H.W. and L.W. results is best made at ffi stage where and have been formed .
The negative signs for the L.W. results are introduced before addition to the H.W. results , and total and are formed .
After the eighteen cosine and sine total numbers are thus formk , , hey are to be multiplied by the cosines or sines of . . . . .
The products are then summed so as to give It was noted at the beginning of this section that we also sums of the form 2 These sums are obviously made by unity in place of each height , and , of course , not treating the negative .
Thus , where the H.W. and L.W. are combined it is not sary to change the sign of the L.W. , as was done in the combina : 11840 .
] nalysis of Tidal Observations .
These have to be divided between H. and L.W. For the sake of nvenience , I suppose that we always begin the sel.ies with a H.W. .
then when the integer is odd we put in one more H.W. than L.W. , and thus have the following ruJe :\mdash ; No. ofsemi-lunation . .
1 2 } 4 5 6 No. of last H. W. in the semi-lunation 29 57 86 114 143 I il No. of last L.W. in the semi-lunation. . . . . .
28 57 85 114 ) 171 The H.W. and L.W. are here supposed to be nulubered tively from 1 onwards in separate tables .
The rules of partition given in Appendix are found in the am way .
S .
1 .
On the Ovcr des .
Observations of H. and L.W. are very inappropriate for determination ] of these tides ( of which the im are because they express the ture of the fionl the simple harmonic shalJe , and we are supposed to no information as to what occurs between two tides .
These tides make the interval H. to L.W. longer than from L. 1011.W .
, and is no doubt that , assuming the existence in tho expression of term of form we bhall gct all proximation to and by finding the meall of and .
But computation of the , coefficients for the pertul.bation of by would be .
and hus amount of additiona ] would be very great , whereas in thw analysis of continuo rvation the rtideS are slmost without any additional work .
I am inclined to think that it would be best to obtain hourly lions for seve ] days at several parts of a lunation , and by methods of polation to construct a typical diurnj ) tidewave , from which , by the methods of rmonic a , we could find the ratio of the beights of the over-tides to the fundamental , and the relationship of their phases .
I no atte.mpt at such an tion in this place .
S12 .
On the Annual and Tides .
These tides are of much importance , so that not to be neglected from a point of view .
It is obvimpossible to obtain any results from a series of obscl.vations of leSs thau a year 's duration .
Bules for the partition of tides into monbhs or parts of a year are given in the pondix E. mean of all the H. and L. .
obervations for each month be taken as the height of meam We have supposed in the previous tion that heights are measured from mean sea-level , and although it necessary that this condition should be rigoronsly satisfied , it be well , where there is a large annual tide , to refer the different datum levels in the different quarters of the year .
S13 .
On Gaps in the Series of Observations .
It often happens in sae obser ations that a few tides are through some accident , or are obvionsly vitiated by heavy Now the present method depends for its applicability on the cence of terms in the averages .
It is true that it is orously a plicable even for scattered observations , but it applied to such a case all the coefficients have to be }ulated , and , as every tlde reacts on every other , the computation would be to make the method almost impracticable .
Thus , where there is a observations must be fabricated ( of course noting that they fabrications ) by some sort of interpolation , and even values which en very incorrect are better than none .
* If the interpolation it migbt be well to test its correctness in a few places whsnthe reduction is done .
If a whole week or fortnight be missing , and il the computer cannot find a plausible method of interpolation , I oan only a preliminary reduction from the continuous parts , and the computation of a table for the hiatus .
Each such case musI be treated on its merits , and it is hardly possible to formulate general rules .
APPENDIX .
Tables and Rules of General Applicabilit A. To The following table is for what would be the mean hour-angle , if the moon had been on the meridian at the through which tarth hurnod ratively toonThi 1denoted b per mean solar hour ) since epoch .
Analysis of Tidal Observations .
303 It would be advantageous to extend the table np to 90 days , but it can be used as it is for than 30 days by the division of the time into sets of 30 days .
In the second period of 30 days must be subtracted from the tabular entry , for the third period and so on .
* For example : Find for .
The day is of the third 30 , and the tabular entry for is , and subtracting have gives , so that The correct result is , and it is obvious that an error of may easily be inourred by the use of the table .
The row for day - is given because it may be necessary to use one tide before epoch ; this row used in the example below . .
Observe that the decimals run thus , , 5 , , \amp ; c. , then , 9 , 4 , \amp ; c. , then 8 , 3 , 8 , \amp ; c. , and so on .
The first entry in which the sequence alters I call a " " change The incidence of " " changes " " may be found thus : is ; take CreUe 's multiplication for , and note where the last digit but three , having been 4 , beoomes 5 ; I say that this is a ( change For example , , and 560 ; then a challge occurs at the hour , or at the hour , or at .
If the table be continued to 60 and 90 days , \amp ; c. , cting 5 , \amp ; c. , the changes ill fall a little wrong , but they may easily be corrected by means of Crelle 's table , as here shown.\mdash ; ( Added Aug. 2 , 1890 .
Prof. G. H. Darwin .
Harmonic Arlalysis of Observations .
305 Prof. G. H. Darwin .
To find No table is necessary for the conversion of time into angle at per hour to find , or 2 , since we ltiply bhe hours by and add half the number of minutes .
This rule is the same for evq day .
B. The tides and It is required to compute and from where and , the sun 's parallax referred to being its value at the middle of the period under reduction .
If , for example , February 14 is the middle of the period , II is found thus : Sun 's parx .
Feb. , mean parx .
, diff. Then The period under reduction consists in this case of an exact numM of unations .
The following table gives and , according to the number of semi-lunations:\mdash ; No. of semi- lunations .
1 .
2 .
3 .
4 .
6 .
6 .
is the sun 's mean longitude at epoch , found from Naut .
Al .
; are found from Baird 's Manual iu the tables applicable to tide C. The Tides and L. Summations carried out quarter-lunar-anomalistic numbered , \amp ; c. Grand totals are then made in two ways , [ \mdash ; ( viiviii)]\amp ; c. , to find and r$ [ ( vivii -(viiiix)]\amp ; c. , to find } 18$0 ] Harmonic Analysis of Tidal Observations .
307 , for example , ) denotes summation carried over the half period made up of iardii .
) totals are by certain mean cosines and sines ( whose values are given in F ) , and are summed .
The next process is multiplication by a factor , of the value depends on the number of -anomalisbic periods under treatment .
The following table gives the value of this factor:\mdash ; No. of -lunaranom .
periods . .
ix .
xi .
xiii .
The angle is also required ; it depends on time of the first tide under reduction .
If be the time in hours since epoch to the first tide , For instance , in the example below the first tide is at of day - ; this is , or , before epoch , so that ; then D. The Tides , P. Summations are carried out over quarter-lunar iods numbered I , II , III , \amp ; c. , and totals are formed like those mentioned in , and a factor ( which differs slightly from ) is required in the formation of and .
This factor depends on the number of quarter-lunar periods under treatment , and the following table gives its value:\mdash ; No. of -lunar periods .
III .
V. VII .
IX .
XI .
XIII .
The angles and are required ; they depend on the time of the first tide under reduction .
If be the time in hours of the first tide Since epoch , For instance , in the example below we have , as in , and It is required to compute and from Prof. G. H. Darwin .
where The period under reduction consists in this case of an exact number of quarter-lunar periods , and the following table gives the values of and , according to the number of quarter-lunar periods:\mdash ; No. of -lunar periods .
III .
V. VII .
IX .
XI .
XIIL is the sun 's mean itude at epoch , the formula for is given above , and are found from Baird 's Manual in the tables applicable to the tide E. Rules for the Partition of the vabions into Groups .
If the first event after epoch is a L.W. , either omii it from the reductions , or let the first tide be the H.W. which precefles epoch .
Thus we itxe to begin with a H.W. * The H.W. and L.W. are treal , ed apart in separate tables .
tide ( H.W. or L. W. , as the case may be ) is numbered consecutively , from 1 onwards .
The following are rules for partitions:\mdash ; *This is not necessary , but it makes the statements of the subsequent rulaq simpler , as they have not to be iven in an alternative form .
1890 ] Harmonic Analysis of Tidal Observations . .
: .
Prof. G. H. Darwin .
For and Semi-lunations numbered 1 , 2 , 3 , \amp ; c. No. of semi-iunation . . . . . . . . . . . . . . .
1 2 3 4 5 6 No. of last H.W. in the semi- lunation . . . . . . . . . . . . . . . . . . . . . .
29 57 86 114 143 171 No. of last L.W. in the semilunation . . . . . . . .
.- . . . . . . . . . . . . .
28 57 85 114 142 171 Total No. of tides up to end of each semi-lunation. . . . . . . . . . . . . . . . . .
57 114 228 285 342 For Annua ] and Semi-annual Tides .
Months , or parts of a year , numbered 1 , 2 , 3 .
No. of month . . . . . . . . . . . . . . . . . . . . . . . . . .
1 2 3 No. of last H.W. in month . . . . . . . . . . . . . .
59 118 176 No. of last L.W. in month . . . . . . . . . . . . . .
59 176 Total No. of tides up to end of each month .
No. of tides in each month. . . . . . . . . . . . . . .
118 117 The epoch for the second quarter year should be 91 days after first epoch , that for the third 92 days after the first , for the fourth 91 days after the third , except in leap year , when the last should also be 92 days .
There are six tides ( or about thirty-seven hours ) more in a quarter year than in xiii quarter-lunar-anomalistic periods ; the times of these six tides ( or ten tides in one of the quarters ) are to be omitted from the reduction , and their heights are only required when the annual or semi-annual tides are to be found .
F. Oosine and Sine Factors for all the These are the cosines and sines of , \amp ; c. They are as follows : \mdash ; Cosine and Sine Factors .
Read downwards for cosines , upwards for sines .
1 .
10 .
2 .
11 .
0 .
8 .
12 .
4 .
0 .
13 .
0.462 5 .
14 .
6 .
0 .
15 .
0.301 7 .
16 .
8 .
17 .
9 .
0 .
18 .
0.044 In the evaluation of and only the facta marked are required .
1890 .
] Harmonic Analysis of Tidal Observations .
311 G. Increments of Arguments in arious Tim es .
The following table gives the increments of arguments of the several tides in various periods , multiples of being subtracted .
This table facilitates verification of the calculation of the harmonic constants Place , Time , Datum Level , and Unit of Length .
The case chosen is three montbs of observation ( in reality the tidal predictions of the lment ) at Bombay , and the epoch is , 1887 .
A dabum at or very near mean water-mark is taken , so that all the H.W. are posive and .
negative .
This datum is found by iking the mean of all the H.W. and L.W. of the original observations .
In this case 99 ; was subtracted from all the tide heights .
I might more advantageously have subtracted 102 or 103 inches , but 99 inches was chosen from considerations applicable to my earlier attempts , but which do not apply to the computation in its present form .
At placcs where there is a large annual inequality in the height of water , it would be advisable to use a different datum for each quarter of a year .
It is not , however , important that datum should conform rigorously to mean water-mark , for even the discrepancy of , which occurs in my example , does not materially affect the In recording the heights , a convenient unit of length is to be llsed , and it is advantageous that the H.W. and the L.W. should be expresSible by two figures , so that the larger H.W. and L.W. shall fall into the eighties and nineties .
The unit of length is here the inch .
Prof. G. H. Darwin .
Times and Angles .
The times of H.W. , numbered consecutively , are entered in a as shown on p. 313 .
Since astronomical time is the epoch , the tides will come in the half days which are numbered with and the .
tides in the half days which fall between the inteml numbers .
From time to time there will be a half day with no H.W. ; this row in the table should be left blank , but there happens to be no row in the sample shown .
A computation form for times and angles might be printed , for , although the exigencies of the printer have not the entries to be equally spaced in the sample below , yet the computa .
tion form might be printed with equal spaces , and the dividing lines are to be filled in by hand .
The .
table is similar .
Both H. and .
are to be divided into istic and quarter-lunar iods , and semi-lunations , according to the rules given in E. These partitions and the numbering of the entries could not be printed , becaluse of the occasional blank rows .
The mation of and of , by mearls of Table A and the rule following it , is obvious .
In the subtractions and additions under the headings and is added or subtracted where necessary .
is found by doubling The Heights .
The W. heights are written in columns ( with the same blanks as in the table of times and angles ) , and are so arranged , either on strips of paper , or by folding the paper , that the heights may pinned to the times , bringing each opposite to an angle on th6 same row with the time corresponding to that height .
The heighb will on one occasion have to be pinned opposite the column , on second occasion opposite the column , and on a third occasion column .
The L.W. heights are written in similar columns , bnt the minul signs should be omitted .
Harmonic Analysis of Tidal Observations .
313 None 58 ii iii \amp ; c. , \amp ; c. The sum of all the H.W. entries to the end of xui is 9791 .
sum of all the L.W. to the end of xiii is 8577 .
There H.W. and 173 L.W. Hence mean sea-level is , which is equal to .
It would have been better , thereforn have subtracted 103 inches instead of 99 inches from all the heigw Tbis would have given mean sea level at from the datrg adopted .
Sorting the Heights according to the of the Anglu In the tables below the column belongs to all angles and belongs to to ; and so on .
.316 Prof G. H. Darwin .
Harmon Analysis of Tidal Ob servations .
3 ] 8 Prof. G. H. Darwin .
1890 .
] Harmonic Analysis of Tidal Observations .
319 Table of Sums for and L. ( See p. 320 .
) Maintaining the divisions , \amp ; c. , it is now necessary to sum each of the four times 18 vertical columns in each of the xiii divisions , to subtract the 18 columns of the quadrant from the 18 columns of tlJe 1 , and to subtract the 18 columns of the quadrant from of the .
The columns have then to be reversed .
Since in this case nearly all the H.W. fall in the quadrallt , and .
fall in the quadrant , it is easy to write down at once , and , as shown on the next page .
In this the \mdash ; signs of the L.W. entries have to be reintroduced , but as .
lie mostly in quadrant , which enters with nega , they become positive again .
It thus happens that nearly all the columns come out ; thers , however , a few \mdash ; in Table of Sums for See We now disregard the sub-divisions , \amp ; c. , and sum the 4 times 18 columns into grand totals , stopping the summations , however , at the end of 6 semi-lunations ( i.e. , at 171 H.W. and .
It would hardly be wise to attempt in this case the subtractions , without the intermediate steps .
The following table gives the resuIts .
Rule for Cosine and Sine Summations .
For 'cosines ' the 18 numbers required are reversed ) .
For ' sineH ' the 18 numbers required are derived from roversed ) .
Evaluations of and ( continued ) .
See Table ( f ) of Sums enter the 18 ' cosine ' numbers , in accordance with ( h ) in xiii vertical columns , and perform the operations indicated in the example on page 322 .
The column of ' cosine factors ' are those given in , and is in for xiii 4-1unar-anomalistic periods .
Prof. G. H. Darwin .
: : : : : : : : : .
: : : : ffi : : : : : : : : : Harmonic Analysis of Tidal Observations . .
: : .
: . .
: : : Harmonic Analysis of Tidal Observations .
323 continued )\mdash ; Form , vivii , viii xiii .
Form difference columns ( iiiii)\mdash ; ( ivv ) , ( vivu ) , \mdash ; ( 'iiiix ) , viii , ; add the 5 columns together ; by cosine factors ; sum and multiply by or resuIt is : \mdash ; Sines.\mdash ; From Table ( f ) of Sums , enter 18 " " sine " " numbers , in accordance with ( h ) in xiii vertical columns .
Perform all the same operations as on " " cosine\ldquo ; save that we use sine factors , which are the same as cosine factors in inverse order , beginniug with and with The two results Collecting results , proceed thus : \mdash ; Sum Diff. sum diff. Sum Diff. sum diff. liivaluation ( continued ) .
From the Table ( g ) of for enter in one vertical colum1l 18 cosine nnmbers , in accordance with ( h ) ; multiply them by factors ; add up and divide by the total number of entries for The result is:\mdash ; Then enter in vertical column 18 sine numbers , in ordance with ( h ) ; them by sinc factors , add up , and divide by 342 .
The result Sorbing according to of , and Bvaluation of The H. and L.W. are treated in separate tables , similar in form that the \mdash ; signs of the L.W. heights are omitted .
sine factors , add , and divide by 342 .
The resuIts ( 1 .
) Sorting according to of The whole process is precisely parallel to the sorting according to values of in ( e ) ; the teen divisions are , however , given by the quarter-lunar-periods I , II , . . . .
XIII .
The only difference lies in the substitution of the factor for XIII equal to ) for .
It is unrlecessary to give an example .
The results are:\mdash ; 1890 .
] Analysis of Tidal Observations .
325 way .
The results are Formahon of the Mean Sums of and These may be found with sufficient accuracy from the last Table ( m ) of Sums , part of which is given .
In that table lines are drawn dividing the columns into three divisions of six each .
These are treated in the way shown in the following example:\mdash ; H. and L.W. Refer to preceding sorting ( m ) .
I. six of six of six of . . . .
six of six of . . . . .
, rev .
II . . . . .
six of six of . .
six of . . . . . . .
six of six of . . .
six of \amp ; c. We have now only 6 instead of 18 sub-divisions of the quadrant , the cosine and sine numbers are found in exactly the same way before .
The following example shows part of the treatment , and the factors are those marked* in F. Analysis of Tidal Observations .
L.W. may be sorted on the same paper .
We may also , if it is found convenient , put on it the sorting oi by adopting , say , to denote a H.W. and * aL .
W. , one of these four signs denoting simply unity .
Harmonic Analysis of Tidal Observations .
them in the same way .
The resuIts are ' of , , These 16 coefficienbs are required to correct the four sums , for the influence of the tides and I call corrn.corrn .
, and the other two and Z. The correction to be applied to to get is and the cor1ection to be applied to to get X is and the two other corrections are given by symmetrical formuhe with in place of ( 0 ) .
These coefficients are computed from and of and of , as given in ( m ) ( n ) ( o ) .
It mnst be noticed that we have above in ( m ) computed and of but , so that the signs of our previous results must be changed in these two cases .
If we remark and are constants found by theoretical siderations , that , are already found , and that in ths column we are compelled to omit the affixcs to the letters and 's and , because they indicate various sorts of 's and and 's and CX'S in the different columns , the computations in table are easily followed:\mdash ; .
0140 0074 Sum Diff. Snlm ocno 5 Ilarmonic Analysis of Tidal Observations .
X ( W-Z ) Computation of and other Constants .
Find , the moon 's mean longitude ( see 'Nautical Almanac , ' Moon 's Libration ) , and the sun 's mean longitude ( sidereal time reduced to angle ) from the 'Nautical Almanac , ' and the longitude of moon 's perigee , from Baird 's Manual , * Appendix Table XII ( there called ) , at the epoch , January 1 , 1887 , Bombay mean time , in Hi .
Longitude From Baird , Tables XIV , , XVIII , find the longitude of Moon 's node , and at Dlid-period , ebruary 1 With the value of find from XIX ] for the tides ; from XIX ( 3 ) find for the tide ; from XIX ( 8 ) find for the tide ; from XIX ( 9 ) find for the tide ; from XX find for the tide ; and from XXI find for the tide The results are Then compute initial equilibrium arguments , in the symbol for which the subscript indicate the tides referred to , \mdash ; , for \mdash ; , for \ldquo ; Manual for Iidal Observations , ' by lfajor Baird .
aylor and FIancis , Flcct Stroet , 1886 .
In making these ductions I have really used the of for July 1 , 1887 , I am operating on tidal predictions made for the whole year 1887 , mado with mean for that year .
The difference is almost inseuProf .
G. H. Darwin .
We already shown in the way of In and we have shown how to compute , and By the formula in , with for 6 semi-lunations , By the formula in , with for XIII quarter-lunar periods .
By the formula in .
:\mdash ; for6 semi-lunations , and with the above values of By the formula in :\mdash ; with for XIII quarter-lunar periods , and with the values of and Final luation of From ; is is - , so that lies in second quadrant ; whence ; 1890 .
] Harmonic Anal.ysis of Tidal Observatvo hence , on reducing from inches to feet , Also the value of is taken from ( q ) .
Final valuation of and L. Taking the values of from ( i ) , lies in quad .
, lies in 1st quad .
; whence Then ; n.hence , on reducing inches to feet , Again , since from ( q ) , we have and , where the value of is taken from ( q ) .
Turning to the second pair of equations , Theu ; whence , on reducing from iIlches to fee } , Again , since , we have , and the value of is taken from ( q ) .
Final of and K ; ; , and are , so that lies in quadrant ; The factor is an absolute constant .
Final liivaluation of , P. Taking the values of , from ( p ) , TH ' TH'cos ; , lies in third quadrant , and Then since , from ( q ) , , we have ; and since from ( q ) , therefore ; whence where the value of is taken from ( q ) .
Then , whence , with already found in ( q ) as 9.9241 , and reducing from inches to feet , Also , and The factor is an absolute constant .
We now have to compute 1890 .
] Harmonic Analysis of Tidal Observations .
337 where , an absolute constant for all times and places .
With the values of and and given above in ( q ) and ( p ) , and with the values of and just found , thers resuIts\mdash ; Now We have found in ( p ) 80 that Whence lies in the first quadrant , and Then , wheuce , reducing from inches to fcet , ft. Again , where the value of is taken from ( q ) .
Final Beduction of Mean Water Mark .
We subtracted 99 inches from all the heights before using them , and the mean of the heights was then inches .
Hence mean water is 102,51 inches , or feet above the datum of the original idal observations .
Prof. G. H. Darwin .
of Reduction .
Error of present in inches and of 9 .
obs .
minutes .
in .
in .
too small .
too fast .
in .
too small .
Nil .
in .
too large .
too fast .
in .
too large .
too fast .
in .
too large .
too slow .
in .
too small .
too fast .
in .
too large .
too slow .
in .
too large .
too fast .
1890 .
] Harmonic Analysis of Tidal Observations .
in time with the diurnal tides as with the semi-diurnal .
The errors of fall under the same ory as those of Lastly it is probable that all these errors would have been sensibly if I had subtracted 103 inches from the heights all through instead of 99 , and I krlow that this is to some extent the case .
erification .
In a calculation of this kind some gross error of princip ] may have bsen committed , such , for example , as imputing to some of the a wrong sign ; and this is the kind of mistake which is easily overlooked iu a mere verification of processes .
It is well , therefore , to test whether the tide heights and times are actually given by the computed constants .
This is conveniently done by selecting some three or four tides from amongst those from which the reductions have been , and it makes the calcuIation much shorter if we pick out cases in which it is H. or L.W. withill a few minutes of noon .
example , in the present case it was L.W. on ebruary 1 ( day 46 ) at , and the height was 4 in .
; , it was H.W. March 25 ( day 83 ) at , and the was 13 in .
Now , if denotes the value of any equilibrium argument whose value at the epoch , , January 1 , was denoted in ( q ) by , and if denotes the height of mean sea-level above dalum , the expression the height of water is:\mdash ; .
The time of H.W. depends on a formula involving the sines of the same angles in place of cosines .
Since have chosen cases where it is H. or L.W. at noon , the exceed the 's by an exact number of days ' motion .
The evaluation of the rate terms may be conve11iently made by meaus of an ordinary nautical traverse table , where the decimal point ) is represented by the " " Distance and is givon by " " Latitude and by " " Departure If we know the time of H. or L.W. within or so , the following calculation will give the true time and height .
The computaapplies to the first of the two cases where lve know that should be a L.W. at about of day 46 .
The increments of argument computed from the Table , and bhe 's are subtracted by actual subtraction or by addition of Increment in Ditto Ditto . .
0 .
O.ll 0 .
Time Tabular time Error Mean ater Height

