This action might not be possible to undo. Are you sure you want to continue?
OOAST AND GEODETIO SURVEY
OARLILE P. PATTERSON
A TREATISE
ON
PR 0 J EO TI 0 N S'
BY
THOMAS CRAIG
~.l
WASHINGTON
GOVBRNJlBNT PRINTING OFFIOB 1882
Digitized by Coogle
,
:3?t:JdfB·
E heg' ~28.~?_
•
, 1
o . , I., C I ':,J' ./,. /: ,
)~ , L"V t \._ ,._ "\._, .~ !J .. L..~ I
Digitized by Coogle
i
J
I
,
I .
PART I.
MATHEMATICAL THEORY OF PROJECTIONS.
iii
Digitized by Coogle
..
Digitized by Coogle
CONTENTS.
PART I.
§ I.
PEBBPBOTI'F'E PBOJEOTION.
PRBJ'AOII ••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• INTRoDUCrION •••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• Elementary propertiee of conic aeotioDB .••••••••••.••••••••••••••••••••••••••••••••••••••••••••••••••••••••••
Perspective proJection; plane of projection olliside of ephere .
General values of Ie and ,......... • ••••• • ••• •••••••• • , .
ProjectiOD8 of the meridiaD8 .
ProjectiOD8 of the parallela ••• _ ..
Equatorial projection .
MeridiAD proJection , .
Orthographic projection ..
Orthographic equatorial projection ..
Meridian projection .
Stereographic projection , ..
Stereographic equatorial projection .
Example: To draw the ecliptic with its parallels and ciroJea of longitude ..
Generalized discoasion of perspective projection .
Application to the stereographic equatorial projection of parallols to the ecliptic .
To project the ecJiptio and its parallels .
The angle at which the projection8 of two great cirelee cut is equal to the angle at which the ciroJes thelDllelvee
out .
Extension of this theorem to small ciroles , .
The distance between two points on the epherc and on the PfOjection ..
To find the latitude and longitude of a place from its position on tho chart _ ••
Gnomonic projection .
Gnomonic meridian projeGtion .
Graphical construction of this projeotion •••• __ .~ """ .
Distance between two points ..
§ II.
OBTHQMOBPHIO PROJEOTION.
General definition of a projection _ •• .. ••• •. . .• • ••• •••• 33
Curvilinear coordinates and Gaw8 theory or snrfaces.......... ••.• •••• 3C
Element oflength on any snrfaoe; ratio of original element to its proJection '..... •••• 3G
Orthomorphic projection of the sphere ;... .. .. •••• 36
Mercator'8 projection.... •••••• .. .. •••• .. 37
Harding'8 projection , •••••• 37
Corrected polar di8tanoe.... •••• •••••• •••• 39
Ratio of alteration of lengths _... 41
Lagrange's projeotion ". 41
Hyperbolic fnnctioD8........ •••• •••• 44
Lambert'8 orthomorpl1io copic projection ,;.. •••. • 49
Coordinates or point on projection d.termined by means of seriee . • •. . .. . . . .. . . . .. {)l
v
PIIp.
IX XI
1 13 14 16 16 17 18 18 19 19 20 20 21 2a 23 26
Digitized by Coogle
vi
CONTENTS.
§ III.
ORTHOJlORPHIO PROJEOTION (Continued).
Herschel's projection •••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• Boole'. projection ••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••.
§ IV.
PROJEOTION8 BY DEVELOPMENT.
Conic projection... •••• . ••••• •••••• •••••• •••• •••• •••• •••••• ••• •••• •••• •••• •••• •••••• •••• •••• •••••• • ••••• •••• 66
EDler'. investigation of conic projection.... •••••• •••••••.••••• •••••• .••••• •••••• •••••• •••• •••• •••••• •••• •••• 68
Murdoch's projootion. •. • ••• ••• •••• •••••• •• • ••• •••• ••• • •••••• •• •••• •••• ••• ••• •••• •••• •••••• •••••• ••• • •• • • •••• 71
Bonne's projection...... •••••• •••••• •••••• •••• •••• •••••• •••••• •••• •••• •••• •••• •••••• •••••• •••• •••• •••••• •••• 72
Sanson'. projection.... •••• •••••• •••• •• • ••• •••• ••••• •••• •••• •••• •••• •••• •••• •••• •••••• •••• ••• • •• • ••• •••• •••• • 75
Werner's equivalent projection •••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• ~. 75
Polyoonic projections........... •••• •••• •••• •••••• •••••• •••• •••••• •••• •••. •••• •••• •••• •••• •••• ••••. •••• •••••• 75
Rectangular polyconio projection...... •••••• •••• •• •••• •••• •••• •••••• •• •••• •• •• •• ••• • •••• •••• ••• • .•• ••• •••••• 76
Cylindrio projectlons •••••••••••••••••••••••••••••••••••••••••••••• :..... •••• •••• ••••••• ••• ••• ••• ••• ••• •••••• 79
C_ini's projection •••• •••• •••••• ••••••• ••• •••••• •••••• •••• •••• •••••• •••• •••• •••• •••• •••• •••••• •••• •••• ••••. 80
Projectione of meridians and parallels not orthozonal •••••••••••••• •••• •••••• •• • • •• •••• •• •• •••• ••• • •••• •••••• 83
Mercator's projection •••• '.. •••••••• •••• •••• •••• •••• •••• •••• •••• •••• •••• •••• •••• • •••• • ••• ••••.• •••••••••••••• 84
Formula for, . • • •• • •••• •••• •••• •••• •••• •••• •••• •••• •••• •••••• •••••• ••• ••• •••• ••••• •••• •••• •••• •••• •• •••• •••• 86
Loxodromi08 upon sphere •••••••••••••• _. •••• •••• •••• •••••• •••• • • •• •••• •••• •••• • • ••• •••• •••• •••• •••• •••• •••• 86
Equation of a great circle ••••••••••••••••••••••• : ••• ••• •••• •••• ••• • •••• •• •••• ••• ••• •• •••• •••• • •••••• •••• •••• f!I
Equatione of the projectione of the loxodromio, the great circle, and a parallel to'a meridian...... •••• •••• •••• 88
§ V.
ZENITHAL PROJEOTION8.
Definition and general propertlee... ••• •••••• ••• ••• •• ••• . •••••• •• •••• • •• • • • •• ••• • •• • ••• •••• • •• ••• •••••• •••••• 89
Equidistant zenithal projection...... •••••• •••• •••. •••••• ••••••••• •••• •••••• •••• ••• ••• •••• . •••••. .••••• •••••• 90
Airy's projection by balance of errors.... •••••• ••• ••• •••••• •• •••• •••• •••• ••• ••• •••••• •••••. •••••• ••• ••• •••••• 91
Sir Heury James's projection for areas greater than a hemisphere.. ••• •••• •••• •••••• •••• •••••• •••••. •••• ••.••. 96
Globular projection.... •••••• •••••• •••• •••••• •• •••• •••• •••• •••• •••••• •••• • •• ••• •• • •• • •••• •••••• • •••••• •• •• • • 97
Alteration of lengths...... .••• •••• •••••• •••••. ••••• •••• •••••• •••• •••• •••• •••• •••• •••• •••• . •••• ...... •••• .... 100
§ VI.
EQUIV.A.LENT PROJEOTION.
General eqnationa, _. •• • ••• •••••• .• • .. • 102
Form of i nt egra1s gi vinp; ooordinates in special ClIUIC8.... •••• •• •••••• •••• •. • • •• • •• • ... ••• • •• • ••• •••• 103
Lambert's isocyllndrio proJeotion...... •••. •••••• •••• •••• •••••• 106
Projectione of paraUels and meridiane &8 systems of straight lines : ••. . ••• • 1119
Paralle1e projected into concentrio circle.. •••••• •••• •••• •••••• •• •••• •••• 110
Bonne's projection. •• •••• ...... •• • • • • ••• •• • • • • •••• ...... ...... .... •• • •• • ••• ••• •••••• •• . • • • 112
:rpherical stenoteric projection •••• ••• • • • • • • • • •••••• •••• • •• • • • •• •••• .• • ••• ••• ••• •••• •••• 113
A1ber's projection ' • •••• • •••• • ••••• .. 113
Collignon'. central equivalent projection.... ••• ••• •••• •••••• •• •• •• •••• •••••• •••• •••••• 114
Alteration of angles , ••••• " ••• •••• •••••• 115
Alteration of lengths..... .•.• •••• •••• •••••• •••• •••••• •• •••• •••• •••• •••••• ••.• •••• ••• ••• •••• 116
Polar eqnation of ieoperimetric ourve.... .... .••••• .... .. • • .... •••• 117
Tranllf'ormation of a great oircle •••• •••• .••••• •••• ••• ••• •••• •••• ••••• •••• •••• ... ••• 118
Loxodromio ourvee.. •••••••••••••• ••••• ••• 121
Projection upon the plane of a meridian. • • ••• •••••• • • • • ••• ••• •• .. •••••• 122
Eqnation of a meridian.... •••••• •••••• •••••• •••.• •••• •••••• .... 123
Equation of a paralleL :... • • ••• •••••• .••••• • • •••••• •• •••••• •••••• 124
Mt'Uweide's projection •••••• •••••• •••••• •••••• •••••• •••••. 126
Digitized by Coogle
OONTENTS.
vii
§ VII.
ON THE GENERAL THEORY or ORTHOJlORPHIC PROJECTION.
Pap. 129 130 131 132 133 143 150 156 169
General equations .
Surface of revolut.ion .
Projection of a cone .
Peirce's quincuncial proJection .
Projection of an ellipaoid : ! .
Orthomorpbic projl'ction of any surface upon any other ..
Projection of an ellipsoid upon a sphere .
Caaewhore z+'r=(f + ''1)D ..
lIIothermallines ; .
§ VIII.
GENER4L THEOBY or EQUIV.tl.LENT PROJECTION8.
General considerations...... ...... ...... ...... ... .. .... .. 160
Projection of ellipsoid of revolntion upon a plane...... 161
AlteratioDIi ...... .... . .. 166
CouJugate directions...... 167
§IX.
GENER.4.L THEOBY OF PBOJECTION8 BY DEVELOPMENT.
Deformation of surfaces.... • I.
M_ure of curvature .... .. •••• 176
Method for determining all the IRl1'f'acea applicable to a given 11lrfaoe.... 177
)[eaanre of curvature...... •• 181
Weingarten's investigation.......... 1st
PART I!.
CON8TBUCTION OF PROJECTION8.
Stereographic proJection...... .. .. 187
Stereographic equatorial projection.... •••• .. .. .. 187
Stereographic meridian projection...... .. 188
Distance between two points on the aphereand on the map ~... 190
To find the latitude and longitude of a place from its poIIition on the map.... .. • 190
Gnomonic projection...... . . ... • .. • 191
Guomonic meridian proJection.... •••• 191
Orthographic projection.... .. .. .. 192
Lagrange's projection.. ... .. .. .. .. 193
Projections by development • • 196
Conical projections 196
Enler's investigation.... • 199
Murdoch's projection........ 201
Bonne's projection.... 201
Werner's equivalent projection 206
Polyconio projections.... • • 206
C8118ini'8 projootion ~... !l10
Mercator's proJection...... .. 213
Equivalent projection .. .. .. • !l14
Central equivalent proJection....... 216
Alteration of angles...... ...... ••• 216
Transformation of a great circle........ 219
Construotion of a central equivalent from a etereograpbic proJection .'. 221
Loxodromic ourves.... 223
Projection upon the plane of a meridian .. • .. . • .. 2'.13
EqnatioD of a meridian 224
Equation of a pamllel .. 226
Mollweide's proJection...... .. 227
Tables 230
Digitized by Coogle
•
Digitized by Coogle
PREFACE.
In the following paper an attempt has been made to give a sufficiently eomprehensive account of the theory of projections to auswer the requirements of the ordinary student of that subject. The literature of projectioM is very large, and its btstory presents the names of many of the most eminent mathematicians that have lived between the time of Ptolemy and the present day. In the great mass of papers, memoirs, &c., which have been written upon projections there is much that is of the highest value and much that, though interesting, is tri1ling aud unimportant. Thus many projections have been devised for map construction which are merely elegant geometrical trifles. Although in what follows the author has taken up every method of projection with which he is acquainted, he has not thought it necessary in the cases referred to to do more than mention them and give references to the papers or books in which they may be fonnd folly treated.
As the dift'erent conditions which projections for partioular purposes have to satisfy are so wholly unlike, it is necessary, of COt11'86, to have a diJl'erent method of treatment' for the various cases. Thus no general theory underlying the whole subject of projections can be given. Perhaps the only division of the subjectomitting the simple case of perspective projectionthat has ever been fully treated is that of projection by similarity of infinitely small areas. This is a most important cue, the general theory of which, for the representation of OI4y surfaee upon any other, has been given by Gauss. The mathematical difficulties in the way of suoh a treatment of equivalent projeptious and projections ~y dev~opment seem to be insurmountable, but certainly offer a most attractive field for mathematical research. The author has attempted to add a little to what is already known on these subjects, but feels that what he has done is of little consequence unless, indeed, it should tempt some abler mathematician to take up the subject and develop it as it deserves. A few of the solutions of simple problems in the paper, it is believed by the author, are new and simpler than any he was able to find in the writings of others. The solution of the . problem of the projection of an ellipsoid of three unequal axes upon a sphere by Gauss's method is also believed to be new. With these few exceptions there is no claim to originality in what
..
follows; the attempt having simply been made to present in as simple and natural a form as possi
ble what others have done. The two treatises on projections from which much aid has been obtained are those by Littrow and Germain. Littrow's Chorographie, which appeared in Vienna in 1833, was at that time a most valuable work, but is at the present day too limited in its scope to be of very much use to the student, Unquestionably the most important treatise on the subject at this time is Germain's "TraiU des projrctioM," which contains an account of almost every projection that has ever been invented. The author is under much obligation to this work, both for references to original sources and for solutions of particular problems. In cases where processes or diagrams are taken from this work that are by the author supposed to have been original with M. Germain, special mention of them is made in the text; when, however, Germain has drawn from earlier sources no mention is made of his book, bot as far as possible references to the original papers are given. The opening brief chapter on conic sections has been taken in great part from 'Salmon's Oonle Sections. The object of that chapter is only to give in a simple manner some of the more important and elementary properties of the curves of the second order, so tbat convenient reference could be made in the subsequent part of the paper to the various formulas connected with these carves, and also simple means given for constructing them. At the request of Superintendent Carlile P. Patterson the paper has been divided into two parts. The first part contains
IX
Digitized by Coogle
•
x
PREFAOE
the mathematical theory of projections, while the second part contains merely such a sufficient .account of the various projections as will enable the draughtsm~n to construct them. The principal papers from which excerpts have been made are the following:
Lagrange: "Sur la construction des cartes geographiques," Nouveaux Memoiresde l'Acade· mie de Berlin, 1779.
Gauss: "Allgemeine Anfioaung der Aufgabe, die Theile einer gegebenen Fliiehe auf einer andem gegebenen Fl8ehe so abzubilden, dass die Abbildung dem Ab~bildeten in den kleinsteu Theilen ihnlich wird." Gesammelte Werke. rnittingen edition.
D'Avezac: "Ooup d'ooil historique sur la projection des cartes geographiques." Societe de g60graphie de Paris, 1863.
Tchebychef: "Sur la construction des cartes geographiqnes." Academy of Sciences of St.
Petersburg, 1853.
Oollignon: Journal de l'icole polytoohnique, 41e cahier. Mollweide: Zach's Monatliche Oorrespondenz, 1805.
James and Olarke: "On projections for maps applying to a very large ext.ent of the earth's
surface." Philosophical Magazine, 1865. .
Airy: "ExplBnation of a projection by Balance of Errors applying to a very large extent of the earth's surface," &c. Philosophical Magazine, 1861.
Tissot: "Trouver Ie meilleur mode de projection pour chaque contree partieuliere." Oompte&Bendos, 1860.
An immense list of papers bearing on the subject of projections might be given, but it hardly seems necessary. The above list includes all from which anything of importance has been taken. When minor papers are quoted reference is always made to them in the same place. Special referenee maybe made to a paper by Mr. O. A. Schott, Assistant United States Ooast and Geodetic Survey. This paper is a resome in very compact form of most that is of importance in the subject of projections together with a comparison of the principal methods of prqjecti~n in use at the present day. This paper forms Appendix No.· 15 in theannual report of the Ooast and Geodetic Survey for 1880. In conclusion, the author may say that although he has endeavored to give fnll credit to previous writers on the subject, still it is possible that some reference has been omitted. This should, however, be taken 88 an unintentional oversight, or due to the fact that the author has not been able to trace back to its original source the solution of process in question, and not in any ease to a desire to withhold from any other author his fnll measure of credit.
• THOMAS ORAIG.
OOAS1' .AND GEODETIO SURVEY OFPIOE, August 19, 1880.
Digitized by Coogle
INTRODUOTION.
The history of any science is a history of very gradoal evolotion; so slow at times is the course of this evolotion that often the thread of tradition seems to be broken, and we are left to grope in the darkness of historical uncertainty for the path by which we are to be conducted to the foIl daylight, which always, sooner or later, moots the patient inquirer after truth.
The origin of a science is usually to be sought for not in any systematic treatise, but in the investigation and solution of some particular problem. This is especially the case in the ordinary history of the great improvements in any department of mathematical science. Some problem, mathematical or physical, is proposed, which is found to be insoluble by known mElthods. This condition of iusolubility may arise from one of two causes: Either there exists no machinery powerful enough to effect the required reduction, or the workmen are not sufficiently expert to employ their tools in the performance of an entirely new piece of work. The problem proposed is, however, finally solved, and in its solution some new principle, or new application of old principles, is necessarily introduced, If a principle is brought to light it is soon found that in its application
. it is not of necessity limited to the particular question which occasioned its discovery, and it is then stated in an abstract form and applied to problems of gradually increasing generality.
Other principles, similar in their nature, are added, and the original principle itself receives such modifications and extensions as are from time to time deemed necessary. The same is true of the new application of old principles; the application is at ftrst thought to be merely confined to a particular problem, but it is soon recognized that this problem is but one, and generally a very simple one, out of a large class, to which the same processes of investigation and solution are applicable. The result in both of these' cases is the same. A time comes when these several problems, solutions, and principles are grouped together and found to produce an entirely new and consistent method; a nomenclature and uniform system of notation is adopted, and the principles of the new method become entitled to rank as a distinct science.
In examining the laws which regulate the progress of the human mind in the discovery of truth, the most important points of evidence and data are derived from sciences which have been at their first promulgation the most incomplete, and have owed their subsequent advancement to the successive labors of several, rather than to the unaided efforts of a single mind. It very seldom happens that an individual discoverer gives us the results of his labors in the same form in which they originally presented themselves to his own mind. Still more rarely are the successive steps by which the original investigator conducts the mind of his reader to the perception of new t.ruth identical with those which he himself took in first arriving at it. The early history of any science which hasbeen slow in developing shows us how tedious and inelegant ftrst methods generally are, and also shows how natural it is for the investigator to replace his roughhewn highway wheu steadily, and without stopping to admire the beauty of the surrounding landscape, he has driven over all obstacles, by a beautiful avenue which by easy and natural stages conducts to the desired goal and affords glimpses here and there along its course of immense possibilities in, as yet, unexplored regions. Hence it is easy to see the historical importance of those sciences whose principles have been given to the public, 110t in a complete and systematic form, but gradually, and by methods more or less tedious and imperfect. No science can fornish a better example of this than the science of geography, and in particular that department of it which is the subject of the folJowiog paperthe Theory of Projections.
The name projection itself has a history, and it wonld be curious to trace the development which has occurred in its applications. Borrowed by geographers from gvometers, it has come gradually to signify any method of representation of the surface of the earth upon a plane. In XI
•
Digitized by Coogle
..
·.
XlI
INTRODUOTION.
all rigor the use of the term. projection ought to be confined to representations obtained directly according to the laws of perspective; but it has been extended to take into account representations by development and by other and purely conventional methods. Its mathematical significance is even more extended, as there it is not confined in its application to the representation of the sphere, or spheroid, upon a plane, but of any curve or surface upon any other.
The sphere being nondevelopable, the exact representation of its surface, or even a portion of its surface, upon a plane, is impossible. Oertain conditions can, however, be fulfilled in any projection which will render it sufficiently exact for auy particular purpose. The areas may be preserved, i. e., all areas on the sphere may be reduced in the same proportion, in which case we havean equivalent projection; or the angles may be preserved, in which case we have an orthomorphie projection. The exigencies of any particular use for which the projection is designed give rise to an immense number of other conditions corresponding to which projections have from time to time been invented. It frequently happens that a projection having been constructed to satisfy one impressed condition, is also found to satisfy a number of others; for instance, the stereogrsphic projection at the same time represents the parallels of latitude in their true form, and preserves the angles between the meridians themselves and between these and the parallels; thus this projection, which was originally constructed as a perspective projection, also fulfills the condition of being orthomorphic. The history of projection has been, in consequence of the impossibility of producing a perfect solution of the problem, peculiarly a history of the solution of more or less independent problems.
A method of projection which wiII answer for a country whose extent in latitude is small will not at all answer for another country of great length in a north and south direction; a projection which serves admirably for the representation of the polar regions is not at all applicable for countries near the equator; a projection which is the most convenient for the purposes of the navigator is of little or no value to the geodesist; and so throughout the entire range of the subject particular conditions have constantly to be satisfied, and special rather than general problems to be solved.
It is not, however, the intention in this introductory sketch to give a historical account of the subject, as it would be neither appropriate nor necessary. The complete historical account by M. D' A vesac, in the "Bulletin de 130 Socit~te de Geographic," of Paris, for 1863, leaves absolutely nothing to be said on the subject.
The reader will, however, find in the references to § VII (see Ooast and Geodetic Survey Report for 1880, Appendix No. 15) a valuable bibliography of the subject. This section= was written by Mr. Oharles A. Schott, assistant, United States Coast and Geodetic Survey, with the express object of giving an account of the method of polyconic projection employed in constructing the charts and maps of the Survey. It subserves, however, a double purpose, as it contains a succinct and valuable resum6 of much that precedes, with fnll references to the original sources from which information had to be compiled, and also gives a scientific account of the polyconic projection and a comparison of this, illustrated by examples, with a number of other projections most frequently met with. It is accompanied by six plates and a chart.
We have already spoken of the orthomorphic and equivalent projections, and we may mention in connection with these other general methods and the order in which they are treated in the following pages. This order is not altogether the most seientlfle, bot seems to be better designed for the gradual introduction of the reader to the difficulties of the subject than any other. Dividing the general topic into the following heads:
I. Orthomorphic Projection, n Equivalent Projection, III. Zenithal Projection,
IV. Projection by Development, the arrangement has been as follows:
§ I. After a brief introductory account of the principal properties of the conic sections the subject of perspective projection is taken up. Strictly speaking this should fall under the head of zenithal projection, that is, projections which can be regarded as the .geometrlcel representations
• The paper is not reproduced here, but the reader is referred to Coast and Geodetic Survey Report for 1880, Appendix No. 15.
. .
Digitized by Coogle
INTRODUOTION.
xiii
of the sphere upon the plane of the horizon of any place. The theory of perspective projections is, however, per 8e quite self contained, and is withal the most natural and simple method of representation, so it was thought desirable to open with that method rather than a more general and philosophical, but, at the same time, more difficult method. Sections II and III treat of the &lifferent metbods of orthomorphic projection; § IV treats of projections by development; § V gives an account of zenithal projections, § VI of equivalent projections, and § VII has already been referred to (and for whicb see the report of 1~O, Appendix No. 15) as containing Mr. Schott's account of tbe polyconic projection and its comparison with several other methods. These seetions can for the most part be read by any one possessing a fair acquaintance with the methods of ordinary analytic geometry and the elements of the difterential and integral calculus. The next three seotious are extremely general in their nature, and will require a rather more extensive mathematical knowledge. They were designed to connect the particular problem of the plane representation of a sphere with the much more comprehensive methods of representation of one sur· face upon another which have engaged the attention of the most brilliant mathematicians. Keeping in mind, however, that the book is designed for the use of students, the author bas only stepped across the threshold which leads to the purely transcendental port.ion of the snbject, and bas only given just enough to awaken in the mind of the reader, wbo has a real interest in the general theory, a desire to go himself to the original memoirs for fnller information. A brief section is given on the spheroidal form of the earth. On this snbject very little was either required or desirable; it was not necessary to say much, because the student interested in this subject would natnrally seek for information in a treatise on geodesy rather than in one on projections; it was not desirable to diseuse it very fnlly, because present existing theories, both asto the figure and size of the earth, seem to be in a. transition state.
The United States Ooast and Geodetic Snrvey will undoubtedly soon be able to produce a much better value of the ellipticity than has yet been given. In view of that fact, and also of the fact that the greatest possible change that may take place in the present assigned value of the ellipticity will produce differences in the tables which would be almost inappreciable, it has not been deemed necessary to make any new tables even in the place of old ones, which have been computed on the supposition of an ellipticity as small as rlo.
It is reaclily seen that the general theory of projections touches upon a great number of other subjects in snch a way as to make it a little difficult to decide what is and what is not necessary to incorporate in a treatise having this for its title. Even confining oneself to papers and books entitled "projections, &c.," it is not easy to sift out only that which is of primary importance to the beginner from the immense mass of workgood and badthat has been done upon this subject.
Few departments of mathematics contain more eminent names among those of their founders. From the time of Ptolemy until the present day the most profound mathematicians have devoted time and attention to this subject. The large m~oritY of investigators have, however, had in view the attaining of some particular end, and have deviscd ingenious, but in most cases, rather forced methods of arriving at the desired result. Others, such as Lambert, Lagrange, Euler, GaDS8, and Littrow, have treated the question from a more general theoretical point of view; but even in these cases, as only particnlar divisions of the subject were taken up, the results, as constituting general theories of projection, were very incomplete. The name of Lambert occurs most frequently in the history of this branch of geography, and it is an unquestionable fact that he has done more for tbe advancement afthe subject in the way of inventing ingenious and useful methods than all of those who either preceded or have followed him. The greatest credit is, bowever, due to those princes of the realm of mathematics who, like Euler, Lagrange, and Gauss have done 80 much for the advancement of the theory. Lagrange proposes and resolves the problem of "the representation of a sphere upon a plane in such a way that the smallest parts of the projection shall be similar to the corresponding elements of the sphere, and in wbich the meridians and parallels shall be represented upon the map by circles." Gauss solves a far more general problem in a manner so perfect that it leaves it impossible to add a word to his general tbeory of orthomorphic projection. Lambert, Bonne, Mercator, Mollweide, Oollignon, Airy, and James are but a few of tbose wbo 4ave produced a marked progress in the theory of projections.
Digitized by Coogle
xiv
INTRODUCTION.
In concluding this brief introductory note, the author can do no better than again to refer the reader, desirous of fuller historical information, to D' A vezac's very valuable memoir, and to mention the following most important treatises and memoirs which, having appeared either within the preseJJt or towards the close of the last century, are comparatively easy of access:
TREA.'l'ISES ON PROJECTIONS.
Chorographie oder Anleitung, aUer Arten von Land Bee und HimmelsKarten, Littrow.
Vienna., 1833.
Trait6 des projections des cartes geographiques. Germain. Paris, 1865. Lehrbuch der KartenPrcjeetion, Gretsehel, Weimar, 1873.
MEMOIRS.
Gauss. Algemeine AufiOsung der Aufgabe, die Theile einer Fliiche so absubllden, &0. Bohumacher's Astronomischen Abhandlungen. Altona, 1825. Also in the Gottingen edition of Gauss's
works. .
Lagrange. Snr la construction des cartes geographlques. Mem., de Berlin, 1779. Henry. Memoire sur la projection des cartes. Paris, 1810.
Puissant. Supplement au second livre du Traite de topographie. Paris, 1810.
Euler de repeesentatione superficiei sphmricm super plano. Acta. Acad. Petrop., 1777, pars i. Lambert. Beytriige zum Gebrauche der Mathematik. Berlin, 1772.
Murdoch. Phil. Trans. Vol. I.
Schmidt. Lehrbuch der mathematischen und physischen Geographie. Gottingen, 1829. Zach's Mouat. Corresp. Vols. 11,12,13, 14, 18, 25, 28 contain many papers by MoUweide,
Albers, Textor, and many others.
Crelle's Jonrnal ftir die mine nnd angewandte Ma.thematik, the Mathematische Annalen, the Annali di Matematica, the OomptesBendua of the French Academy, and the Bulletin of the Academy at St. Petersburg, all contain very valuable papers. The same may be said of the Journal de l~cole Poly technique and the Journal de l'~cole Normale Bnperieur of Paris. All of these sources have been consulted in the preparation of the following treatise, and it does not seem to the author as if anything worthy of preservation has been overlooked or left out from any cause whatever.
Digitized by Coogle
A TREATISE ON PROJECTIONS.
MATHEMATIOAL THEORY OF PROJEOTIONS. § 1.
PERSPECTiVE PROJECFION.
A surface in perspective projection has the same appearance as that which it woul<l present to the eye of an observer situated at any determinate point of space. The right line drawnfrom the eye of the observer to the center, in the ease of a central surface, is normal to the plane of projection which may intersect this line at any point of its length. The projection of any point of the surface under consideration is then the point of intersection with this plane of tho line joining the given point to the eye. We will now confine ourselves to the surface of the sphere. Imagine any line drawn on the surface and every point of the line joined to the eye by a strutght line; the aggregate of lines will form the surface of a cone, and the intersection of this cone with the plane will be the projection of the cnrve drawn upon the sphere. If·the cone so formed is of the nth order its intersection with the sphere will be of the 2nth order and with the plane will be of the nth order, so that in general the degree of the curve is lowered by projection, If the cone is a circular cone, its intersection with the sphere will in general be a spheroconic and will be projected in a conic section, as the intersection of a cone of the second degree with a plane is a curve of the second degree or a conic section.
We will in general only be concerned with the projections of circles of the sphere, and they will be projected in conic sections; so before proceeding further with the subject it will be convenient to give a brief statement of the more important properties of these curves as deduced by a study of
their equations. •
The general equation of the second degree in two variables is the equation of a conic section.
This equation in its most general form may be written
. ~
containing five independent constants, viz, tho ratios ~, !If, &c. Five relations between the
• coefficients are sufficient to determine a curve of the second degree; for, though the general equation contains six constants, the nature of the curve depends not 01) the absolute magnitude of these but on their mutual ratios, since if we multiply or divide the equation by !lny constant it remains unaltered. We may, therefore, divide the equation by the quantity C, ma.king the absolute term equal to 1, a.nd there will remain but fire constants to be determined,
Transformation to new coordinate axes frequently has the effect of simplifying YE'ry much the eqnations with which we are dealing; and it will be useful here to find what the general equation becomes on being transformed to a new set of axes, assuming, for a first transformation, that the new axes are parallel to the old,
For this purpose make x=x+.t.', s=s+r: x', y' being the coordinates of the new origin. We
will find that the coefficients of .xl, m.lJ, and r remain aa before, A, 2 H, and B; that
the new G is G'=A.L'+Hy'+G
the new F is F'=Hx'+By'+F
the new constant. term is C'=A';Z'2+2 H.z:'g'+By"+2 Gr.'+2 Fy'+C
Digitized by Coogle
2
TREATISH ON PRO.JECTIONS.
Suppose that we again transform the original equation, this time, however, to polar coordinates. by making X=p eos 0, y=p sin 0; the equation so transformed is
(A· cos' 0+2 B cos 0 sin o+ll sin' 0) p'+2 (G cos o+F sin 0) p+O=O Write this for n. moment as
a being of course = A cos' 0+2 H sin 0 cos o+B sin' 0, &c. The roots of this equation are 1 (1 ± "fJ2ar
= 
p r
that is, the straight line p drawn from the origin meets the conic in two different points.
Suppose here that a=O; then for one root we have ~=O or p=oo; that is, if the coefficient of p
p'=O the line drawn from the origiu meets the curve in two points, oue of which lies at infinity The coefficient of p' is, however,
A cos' 0+2 H sin 0 cos o+B sin' 0
This equated to zero gives
a quadratic in tan 0, and cons' quently we have that there can be drawn through the origin two real, coincident, or imaginary lines, which will meet the curve at an infinite distance; each of which lines also meets the curve at one finite point determined by
2p (G cos o+F sin 0)+0=0
We will now seek the test which will ten us what class of locus is represented by a given equation of the second degree j or we wish to ascertain the form of the curve, whether it is limited in any way or extends to infinity in any direction. Of course if the curve be limited in every direction, 110 radius vector can 00 drawn which will meet it at infinity. For an infinite value of the radius vector we must have
A+2 n tau O+B tan' 0=0
If H~AB<O, the roots of this equation will be imaginary, or no real value of 0 can be found which
will make I *
A (',os' (1+2 H sin 0 cos 0+ B sin' 0=0
Thc curve ill this C80t'e is limited in every direction, and is an Ellipse.
If H'AB>O, there are two real roots to the equation
A+2 H tau o+B tan' 0=0
consequently two reul values of 0, corresponding to which two lines can be drawn from the origin. meeting the curve at infinity. This curve is the Hyperbola.
If H'AJJ=O, the roots of the quadratic arc equal and consequently the two straight lilies which can be drawn to meet the curve at infinity will coincide. The curve in this caseis the Parabola.
If in a quadratic t.c..t'+2 ,Bz+r=O the coefficientp vanishes, the roots are equal with contrary
slgns, Thus then if in the transformed equation .
G cos 0+ F sin 0=0
the two "I\lU('S found for p will be equal and opposite in sign. The points answering to the equal and opposite values of p are equidistant from the origin and on opposite sides of it, and 80 we have thut the chord represented by Gx+Fy=O is bisected at the origin, If we had 0=0 and F=O, then whatever be the value of 0 we should always have
o cos 0+ F sin 0=0 or all chords drawn through the origin would be bisected.
Digitized by Coogle
•
.
3
TREATISI<; ON PROJECTIONS.
Now, by transformation to suitable axes, we can ingeneml cause the coefficients of :e and 11 to vanish. Thus equating to zero the c'oofficients of :e and 11 obtained by a transformation to a new origin, we find that the coordinates of this new origin must fulfill the conditions
A:If+Uy'+U=O
ill order that all chords drawn ·through that point may be then bisected. The point thus determined is called the center of the curve, As these equations for determining the center are linear there can exist only one center to any conic section. The coordinates of the center are found to be
•
For H'AB=O the center lies at infinity, which is the case of the parabola, and this curve is in consequence called 1\ noncentral curve, Obviously the centers of the ellipse and hyperbola lie at a tinite distance.
We have seen that a chord through the origin is bisected if G cos 8+F siu 8=0. Now trans forming the origin to any point, it appears that a parallel chord will be bisected at the new origin Ir
G' cos 0+ F' sin 0=0
or if
This, therefore, is a relation which must be satisfied by the coordinates of the new origin if it be the middle point of a chord making with the axis of x the angle o. Bence the locus of the middle points of parallel chords is
(A:r:+HlI+G) cos 0+ (H:e+By+F) sin 8=0
This line bisecting a system of parallel chords is called a diameter, and we see that it passes through the intersection of
Therefore every diameter passes through the center of the curve.
If two diameters of a conic be 81wl, t1ur.t one of them bisects all chOJ·ds parallel to the other, theA, conversely, the second will bisect all chords parallel to the first.
The equation of a diameter which bisects chords making an angle 0 with the axis of x is
.A:r:+Hy+G+(Hx+B1I+F) tan 8=0 Calling 0' the angle, which this line makes with x, we have
A+H tan 0
tan 0'=  
H+B tan 8
B tan 8 tan O'+H (tan 8+tan 0') +A=O
whence
 And the symmetry of the equation shows that the chords making an angle 0' are also bisected by the diameter making an angle 0 with x.
Diameters 80 related that each bisects all chords parallel to the other are caned conjugate diameters.
The general equation of the second degree in two variables is now, when transformed to the center,
.A:r;2+2 lI:r;y+By' +C'==O where C' is readily found to be equal to
ABC+2 FGHAF2BOSCH2   ABIP
Digitized by Coogle
4
TREATISE ON PROJECTIONS.
If the numerator of this fraction equals 0, the equation would become Ax2+2 IIxy+B1f~O
the equation of a pair of right lines; the condition, then, that the general equation should repre
sent a pair of right lines is .
ABC+2 FGHAF'BG'CH" U
or, in detenninaut form,
lA,
H, G,
H, B, F,
G'I
F, =0
C,
The angles that two conjugate diameters make with the axis of 31 are connected. by the relation
B tan 0 tau o'+H (tan o+t. ... n O')+A=O If the diameters are at right angles,
1 tan 0'=tan 0
lIenee
H tan' O+(AB) tan oH=O
a quadratic equation for the determination of o. Transforming back to rectangular coordinates, this is
H.v2(AB) xyH1f=O
the wellknown equation of two real lines at right angles to each other. These rectangular diameters are called the axes of the curve.
We have seen that when
A cos' 0+2 H sin 0 cos o+B sin' 0=0
the radins vector meets the curve at infinity, and also in one other point determined by C
P=Gcos o+Fsfn °
But if the origin be the center, 'we have G=O and F=Oj hence this distance will also become infinite. Hence two lines can be drawn from the center and meeting the curve in two ooil~'ident points at infinity; these lines are called the asymptotes of the curve and are real in the case of the hyperbola and ima.ginary in the case of the ellipse. The equation of the axes was found to be
Hx2(A:B)xyHy'=O
This is the equation of a pair of lines bisecting the angle between the lines Ax2+2 H.xy+BY=O
Therefore, the axes of the curve bisect the angle between the asymptotes.
The preceding results might all have been obtained by a simple transformation of eoordtnates.' Suppose that, onr original axes being rectangular, we turn the system round through an angle 01, i. e., make
y=x sin OI+y 008 w
the new coeOlcient of x2 will now be
also
A'=A cos' 01+2 H cos 01 sin OI+B sin' 01
H'=B sin 01 cos OI+H (cos' OIsin' OI)A sin 01 008 III B'=A sin' w2 H sin III cos w+B cos' w
Digitized by Coogle
TREATISE ON PROJECTIONS.
5
By pntting H' =0 we get the same equation 8S before for determining tan 0, and in fact this gives us 2H
tan 20=.
AB
for the tangent of the angle made with the given axes by either axis of the curve. Add together A' + B' and we have
Again, write
A'+B'=A+B
2 A'=A+B+2 H sin 2 w+(AB) cos 2 til 2 B'=A+B2 H sin 2 tII(AB) cos 2 til
4 A'B'=(A+B)2[2 H sin 2 til + (AB) cos 2 til]' 4 [1"=[2 H cos 2 O(AB) sin 2 0]'
4 (A'B'B'2)=4 (ABH2) A'B'HI2=ABH' .
hence
but therefore
or
When, therefore, we transpose an equation of the second degree from one set of rectangular axes to another, the quantities A+B and ABH2 remain unaltered.
When, therefore, we want the equation transformed to the axes, we have the Dew H=O, and
From these we can form a quadratic equation to find A' and B'.
We have now for the equation referrerl to the center and axes
A'a:'+ B'y'+C' =0.
Let the intercepts made by the ellipse (or hyperbola) 00 the axes be a on x and b on y. Then
A' 0' =as
B'! lr
and the equation becomes
and for the hyperbola
as the equation of the hyperbola only differs from the ellipse (Inthis case of transformation to the axes) in the sign of the coefficient of ,'.
For the polar equation of the ellipse write
x=p cos a .
,=p ain z
and the equation simply becomes
a2b'
r a2_(a2_b2) cos' ,A
or, by making
e being the eccentricity of the ellipse.
Digitized by Coogle
•
6
TREATISE ON PROJECTIONS.
In like manner we find for the polar equation of the hyperbola, the center being the pole, a21Y
pl= (a2+1J2) cos' A_ai
Making
al+1Y
 ,=e", (e> 1) a
this ia
z 1JI
p = escos""T':Al
In the case of the ellipse the points on the major aDs at the distance va'b"=ad from the center are called the foci. In the case of the hyperbola these are at the distance va'+1J2 from tho center.
In the polar equation of the ellipse we 800 that theleast value the denominator b'+(a2lr) sin' A C IU have is when .1.=0; therefore the greatest value of p is when .1.=0 and is equal to a. Similarly,
. .
we find that the least \'~~e of p is fo~ .1.= 2 a~d this value is equal to b.
These two lines are the
axis major and the axis minor of the curve, I tis also clear that the smaller A is, the larger p will be; hence, tho nearer a.ny diameter is to the axis major the greater it will be. ]f 1=.1., or .1.= A, we will fino the same value of p.; hence two diameters which make equal angles with the axis will be equal. The figure of the ellipse is clearly that giveniu the figure
a F
Flo. A.
F and F' denoting the positions of the foci.
If we solve the equation of the ellipse for 11 we get
b  11= va'xl a
Now, if we describe a concentric circle with radius a, its equation will be 1I=vaci}
Hence, we have the following construction:
Flo. B.
Describe a circlewith radius a and take on each ordinate LP' a length LP, such that ~~,._!; then will P be a point of the ellipse. A similar construction holds for the minor axis, only in that
Digitized by Coogle
TREATISE ON PROJECTIONS.
7
case the ratio of the ordinate of the ellipse to that of the circle eqnals i. The construction is arrived at simply as follows:
•
FIo.O.
Describe concentric circles with radii a and b respectively. Draw any radius CR of the large circle and from the point Q, where it cuts the small circle, draw QP parallel to jihe axis of x; the point of intersection of this line with the line RN, drawn perpendicular to the axis of x, gives l1H P, a point of the ellipse. Similarly, any nnmber of points can be obtained nnd the ellipse drawn through them. Or, again, suppose that we have any line AB constant in length, which moves so that the point B shall always lie on the axis of x and A on the axis of 11.
y
Flo.D.
Now, assume ai"'11 point P, either between A and B or on the prolongation of the line AB, such that AP=a and BP=b; then the locus of P is an ellipse; for calling x and 11 the coordinates of P, we have
~=C08 ABO a
:=sin ABO
therefore
Tangent.
The eqnation of the chord joining any two points x'y' and :x/'!)" 011 the curve is
or
(x'+x").c (y'+y")y_x'.c" y'y" 'a'' + b'  a' + b'  1
which, when x'g' and x"g" approach indefinitely near to each other, becomes xx' + 1!_y' == 1
a' ,,2
the eqnation of the tangent to the ellipse. For the hyperbola the corresponding equation i8
x_x' _ 1111' =1
at b'
Digitized by Coogle
8
TREATISE ON PROJECTIONS.
•
The intercept of the tangent on the axis of x is= ~J'
 ~
The 8ubtangent is the distance from the foot of the ordmate at the poiut of tangency to tho
point of intersection of the tangent and the axis of ». Therefore, subtangent is= a~t2
J:;
2
The quantity ~, beiug independent of b, gives a simple means of drawing a tangent to the
ellipse. If we describe a circle of radius 0, and at that point of the circle which has JI for abscissa draw a tangent, it will intersect the axis of x at the same point as the tangent to the ellipse, 80 that joining the point thus graphically found with point Jly' of the ellipse, we will have the tangent t.o this curve.
Normal.
Forming the equation of the perpendicular to the line ~ +~ =1 at the point Jly', we have
x' y'
a2 (YY')=1ji (xx')
or
a·z biJ . ij 1/ =a'Ir=c2
the equation of the normal to the ellipse,
The intercept of the normal on the axi~ of z equals
tzZ'=e'z'
We can thus draw a normal to an ellipse, for given the intercept of the normal on the axis of x we
can find Z/, the abscissa of the point through which the normal is drawn. . .
The 8ul"iormal is the portion intercepted on the axis between the normal and the ordinate, and equals
Foci.
The square of the distance from tl.ny point x'y' of the ellipse to the focus is equal to
(x' c)'+1f=x"+y'22 ex' +c2
since by deflnltion the eoordinates of the focus are x=c, y=O. But
and
b'+c2=a' Ilenee the distance which we may call FP equals
va'~ cz'+e2z" FP=aex'
We reject the negative value 6:'a, as we are only concerned with the absolute magnitnde of FP and not its direction. Similarly, for the distance from the other focus (c, 0) we find
F'P=a+ex'
Hence
Bence
FP+F'P=2a
or for a fundamental property of the ellipse we have the BUm 01 the distancu Irom a",y point oj' the ellipH to the /oetu is oouta"t ~nd equal to the mqjor axis 0/ the elUp8e.
Digitized by Coogle
TREATISE ON PROJECTIONS.
9
It is not difficult to show that for the hyperbola
FPF'P=~ a
or in tke hyperbola th6 diJferenc6 01 the dista1W88 Iron, tllAJ loci to any point 01 th6 out'1)e is equal to the major (13i8 •.
By help of these theorems the ellipse or hyperbola can be described mechanically.
Ii the extremities of a thread be fastened at two fixed points F and F', it is plain that a pencil moved 1101 out so as to keep the thread always stretched will describe an ellipse whose major axis is the length of the thread. In order to describe a hyperbola, let I. ruler be fastened at oue extremity
II
FIo. E.
F, and capable of moving round it, then if a thread, fastened to a fixed point F' and also to a fixed point on the ruler R, be kept stretched by a ring at P, as the ruler is moved round the point P will describe a hyperbola; for, since the sum of F'P and PR is constant, the difference of FP and F'P will be constant.
Directriz,
. a'
The directrix is a line perpendicular to the axis major at a distance from the center= d::  T' The C
distance of tJ:1e directrix from any point of the curve is
at a 1
C z'= 0 (a  eIl!') =6 (aell!')
Hence we have another fundamental property of these curves which would enable us to construct the curve, viz, that the distance of any point on the curve from the focus is to its distance from the directrix 8S e is to 1.
The length of the focal radius vector we have found to equal aell!'; but z' (being measured from the center) equals p cos A+C. Hence
p=ae p cos 1eo or, solving for p and replacing 0 by its value ae
a (Ie') r= l+e cos A
The double ordinate at the focns is caned the parameter or LatusRectum; its half is found, by makiug1=;, to be=~=a (16'). Denoting the parameter by p the equation may be written
p 1
P=21+e cosI .
The properties that we have discussed 80 far have been common to both the ellipse and hyperbola, hut the hyperbola by virtue of its real asymptotes possesses properties which the ellipse does not P088e88. We saw that in the general case the equation of the asymptotes was obtained by placing the highest powers of the variables=O, the center being the origin. Thns we have for the equation of the asymptotes to the hyperbola
2 T P
Digitized by Coogle
...
10
. TREATISE ON PROJECTIONS.
or
a' and 11 being any pair of conjugate diameters. Hence the asymptotes are parallel to the diagonals of any parallelogram whose sides are any pair of conjugate diameter~.
Parabola ..
We have already seen that when the equation of the second degree represents a parabola, we must have
This is clearly only the condition that the fil'tlt three terms of the general equation should constitute a perfect square, or that the equation might be written
(ax+py)'+2 Gx+2 Fy+C=O
TranRformation o(axes will greatl~: simplify this equation. Suppose that we take for new axes the line ax+fiy and the perpendicular on it fixago Now in the equation of the curve we know tha.t ax+fiy and 2 Gx+2 Fy+c are respectively proportional to the lengths of perpendiculars let fall from the point :cy to the lines
Hence the equation of the curve asserts that the square of the perpendicular from any point of the cnrve on the first of these lines is proportional to the perpendicular from the same poiut on the second line. Now, since the new coordinates x' and y' are to denote the lengths of perpendiculars from any point on the new n,xes, we have
Make a'+P=r; and we have
rx'=13xay rY'= aX+/1g
rx=ay'+ Px' rY=PY'ax'
Making these substitutions in the equatiou of the curve, it becomes r3yI2+2 (G,3Fa) x' +2 (Ga+ Fp) 'II' +rC=O
Or by simply turning the axes through a certain angle we have reduced the equation to the form B'y'+2 G'x+2F'y+C'=O
Again, change to parallel axes through a new origin x'g'; the equation' now becomes
B'g2+2 G'x+2 (B''II'+F')y+B'y'2+2 G'x'+2 F'!/+C'=O
As the coefficient of x has remained unchanged, we evidently cannot make it vanish by this kind of transformation. But we can determine x'g' so that the coefficient of'll and the absolute term
shall vanish. Take for the eoordinates of the new origin •
then the equation reduces to
or simply
Digitized by Coogle
·TREATISE ON PROJECTIONS.
11
when we have
P=~(!,_aG~ (a'+p2),
The quantity p is, in the assumed case of rectangular eoordinates, called the principal paraflleter of the curve. Since every value of x gives two equal and opposite values for ,1, the curve must be symmetrical with respect to the axis of x. None of the curve can lie on the negative side of the origin, since a negative value of x will give imaginary values of 11. The figure of the curve is thut here represented.
FIG. 11'.
The eqnation of the chord joining any two points on the curve is
(1Iy') (1I1I")=rpx (y' +y") 1I=px+1I'1I"
Make y' =y" and write yn :pal and the equation of the tangent to the parabola. is
or
For the intercept on the axis of x we have x= x'; that is, the distance from the foot of the ordinate of contact to the point of intersection of the tangent with the axis of the curve is bisected. at the vertex, or, simply, the .ubtangent is bisected at the vertex.
Normal.
The equation of the normal is
P (1/y')+2 y'(xx')=O
Its intercept on the axis i8
x=x'+lp The Btdmormo,l being defined as before by the relation Subnor. =xx'
we have for the 'parabola that the subnormal is constant and equals l p or i parameter.
F'0C1.UI.
The focus of the parabola is a point situated on the axis of the curve and at a distance from the vertex equal to onefourth of the principal parameter. Calling m this distance, we have for the sqa.re of the distance of any point of the curve from the focns
(x'mS)+1I"=x'22 tnx'+m'+4 mx'=(x'+m)1
Hence the distance of any point from the focus equals x'+m.
The directrix of the parabola is a straight line perpendicular to the axis and at a distance of vertex outside of the curve equal to m;. hence the distance of any point on the curve from the directrix must equal x' +nl. We have, then, as a fundamental property of the parabola, that the dillto1We of an1l poiAt of the curve from tluJ jrlCUlJ iB equal to it. dutance from th,e directrix.
Digitized by Coogle
12
TREATISE ON PROJECTIONS.
The equation of the ellipse is satisfied by making x=a sin 1, lI=b cos 1, when 1 is the complement of the angle PloL (Fig. B), or is the complement of the eccentric anomaly. We have from these values of x and 11
and for the element of are
. d8= v(JXS+d!t= vas cos' l+lJIsin'z dl= va'(a2b') sin'l dx
Taking the eccentricity of the ellipse v~:~fjI as the modulus of an elliptic integral, we ha:ve at once for the length of the entire ellipse,
• •
8=4 a'£2 ./1li' siu'ldl=4 a'£2 .6 (kl}dl
8=4 a E,k
E,k denoting the complete elliptic integral of the second kind. If the eccentricity is small or the ellipse is nearly a circle, the fonction E,k has for value (Oayley's Elliptic Functions, page 46)
or
Ek 71:(1 1 1.2 11.31" 1'.3'.5 1.11 1'.3' .... (2i3}2(~il} • .2l )
1=2 22,.22.4''''22:42:fi2K""···· ~2~42.·..:. (2i}2   IV
The area of the ellipse is well known to be 7I:ab. It can be obtained readily by integrating J J dzdll, the limits of x and 11 being taken from the equation
a;2 11'
(ii+1jZ=1
For the hyperbola
write x=a sec I), 1I=b tan I), where I) is the eccentric anomaly
d:c=a sec I) tan udl)
and thence
dl)
ds= viJ2+a' sm' I) cos u
Here take
k= ~'!_va2+~
the reciprocal of the eccentricity; then k' the complementary modulus equals v a2~ Assume an angle p. such that tan I) =k' p.. Then
a :C=.6p.
C08P.
lI=bk' tan p.
and thence
d8= __!!!!df1. cos' f1.C:.p.
By differentiation of .6p. tau f1. we find
Digitized by Coogle
TREATISE ON PROJECTIONS.
13
and conversely integrating from zero we find
Substituting this in the expression for 8, and remembering that bk=ak' we have
v
a
8= } {tan /L6/L+k"FJLEJL
where 8 denotes the length at' an arc of the hyperbola measured from the vertex.
P The incongruity of explaining the elementary principles of Conic Sections and assuming a knowledge of the more difficult one of Elliptic Functions will perhaps strike all readers; but the object of the explanation in the former case was not so much to teach conics to one who had not studied the subject as it was to give a brief resume of the more important elementary principles, wbich would afford means of praetieally dra,wing these curves. Hereafter elementary explanations will not be given except in particular cases where it ,may be desirable to bring out some important fact in the process.
We will take up now the subject of perspective pr~ectiontaking the plane of projection at first as outside of the sphere. Let C (Fig. I) denote tlie center of the sphere, V the point of sight,Op the trace of the plane of projection upon the plane of the paper, P the pole ot the equator, and M any other point on th~ surface of the sphere, having 0 for latitude and w for longitude, PZZ' being the first meridian. Then we have'
w=ZPM, to these add PZ=;a
0000= PM MZ=9'
It. ;~CZ PZM=",
and also assume VC=c, VO=o'. The projection of Z is 0
}'1O.1.
and of P is p, these being the points in wlrich the projecting lines pierce the plane of projection. Assume til as the projection of M, then the position of this point must be determined with reference to some system of eoordinates, The 1Il0st convenient system to adopt will be the rectangular SYRtem OX, OY, then Un=x, ftm=y, and we have to detennine e and y as functions of the given con
stants c,o', r and the angular magnitudes 0 and w. .
Equating the sum of the three angles about the point C to the sum of the three angles of the triangle CD~I we have, since this triangle is isoscele5,
also
combining these two resnlta, we have
•
Again, in the triangle VDO
•
MVZ ZMDO _ 9' VO~  ~  2
"I>~
MDC=9' V; )\DC=1r{9' V)
sin D c sfnv:::;,.
Digitized by Coogle
14:
TREATISE ON PROJECTIONS.
or
sin (¥' V)=~ sin V; r
Solving this for tan V, we find
r sin ¥' tanV= c+rcos,
and consequently
oq,.. c' .
.. =OVtanV= rSIll¥, c+rcos ¥'
Observing now that MOP=¢, we at once obtain for x and 11 the values
c'r sin ¥' cos ¢, x=     c+rcos,
We have now to determine, and ¢ in terms of 0 and w.
11 e'r sin ¥' sin ¢ c+rcos ¥'
In tbe spherical triangle PZM we have
sin ¥' sin '1'=cos 0 sin w
• sin osin a cos ,
SID, COS I/'=oo~
cos ,=sin a sin o+cos a cos 0 COB til
Combination of the last two gives
sin, COB I/'=cos· a sin osin a COS 0 cos til
Snbstitnting these values of ~ ¥" sin, COS 1/' and sin , sin 1/' in the vslnee of x and 11, these become
(1)
c'r (sin a cos 0 cos wcos a sin 8)
x ~   
c+r (cos a cos 0 COS w+sin a sin (/)
c'r cos 0 sin w
1/= ctr (cos a 0080 cOs w+sin o:8iOO)
Upon these two equations depends the entire construction of the perspective projection of a sphere upon a plane.
For (,'=c+r the plane of projection becomes the tangent plane at the point Zj for c'=c the plane passes through the center of the sphere, and the great circle 80 cut out will be the limiting line of the projection. For the last case the equations become
(2)
or (sin a cos 0 cos wcos a sin 0)
z            
 c+r (cos a cos 0 cos w+sin a sin 8)
or cos 0 sin UI
1/= c+r (cos a cos 0 ~s ~+8in a sin 0)
If the eye be conceived as situated upon the prolongation of the axis of the earth the plane of projection will eotnehle with that of the equator and we have an equatorial projectwn. 'fhe values
of z and 11 for this case are found by making a=; in the IMt formula, thus:
or cos 0 cos til
z= .
c+r 8lU 0
or cos 0 sin w 11= c+r sin 0
If the eye is placed in the plane of the equator the plane of projection will pass through a meridian, and the projection is said to be a meridian projection. For this case we have a=O.
or sin 0 z=  .     a+r cos 0 C08 UJ
or cos 0 sin til 1/= c+r cos 0 Cos u~
By making 0=0 in equations 2, and giving w a series of values, we will determine as mauy points of the projection of the equator as will be necessary to draw that line. And, io like manner, by giving 0 any constant value, and giving UJ any series of values, we can determine the project OU8 of the intersections of all ·the meridians wi h the assumed parallel of o. This }IrOCCS8 is, however, II.
•
Digitized by Coogle
TREATISE ON PROJECTIONS.
15
very lengthy and inelegant one. As we are concerned only with the projections of circles of the sphere we know that, the proieeting curve being of the second degree, these projections will be ClUnS of the second degree. It will consequently be desirable to flnd the equations of these curves, u..d from the equations construct the projections.
To obtain the equatiou of the projections of the meridians, it is only.neceasery to eliminate the latitude o between equations :&; dividing the second of these equations by the flrst, we have
y ~o~w .w
x=siIlaOOs u COs ~C08 a sino=sin a cos w~Cos a tan 8
from which
y sin a cos wxsin w tan 0 =          y cos a
(4)
b rc' cos a sin w
c'rI (lcos'a sini w)
or
Substitnting these in the values of y gives usthe general equation of the projection of meridians.
(3) x2 (&rI sin· a) sin· w+xg (rIe2) sin a sin 2 w+1f (c'c' sin' a sin' wr' cos' ",)
;1; (1"0 sill 2 a sin· w+rc cos a sin 2 wrc' ~I a sin' w=O
This is the eqnation of an ellipse whose semi·axes a and b are given by
rc
a=  .  ====
v&r (1C08·.:c sin'w)
and whose center is at the poiut
cr" sinl a sin' w
(0) ~=2[&r2(1cos'asinlw)]
cr" cos a sin' w 7]=2 [c'rlicost a sin' w}]
For the direction of the axes we have, w being the angle that the major axes makes witli the axis of z,
sin a sinl w
tan 2 w= ; ,_(',os' WSID' a sm' w
and from this, by means of the formutas,
• 2 tan w
tan 2 w=
1 tan' III
sin 2 111=2 sin", COS III
Then easily follows
(6)
tan' w=':;:__!_ S10 a
The quantities ~ and 7] have different values for every meridian, i. e., for every value of w; if, then, we eliminate w between the two equations giving ~ and 7], we will obtain the locos of the centers of all the ellipses
or ...
~ . ta
= sin a n III
7J
ta t.
n w_ sm II
7J
i ' ~
8 n w~..t
~r7J sin 4
Substituting this in the equation giving ~, it becomes
from which
Digitized by Coogle
16
TREATISE ON PROJEOTIONS.
(7)
the equation of an ellipse having ~ and '1J for its current eo ordinates. The ellipse also passes through the origin of coordinates as it lacks an absolute turn. The center of the ellipse lies on the axis of x and is given by
"'C sin a COS a ~=2 (& .... ::r 8in2 a)
and its semiaxes a' and b' are given by
(9)
, ric sin 11 cos a
a 
2(&r sin2 a)
r'c cos a
b' == _ ~====
2 V (&_,.2) (&,., sin2 a)
It is obvious that the major axis coincides with the axis of x, and consequently w=O.
PROJEOTIONS OF THE PARALLELS.
To obtain these projections it is only necessary to eliminate w between equations 2. We have
cx+rc cos a sin 0 
cos w= crsin a cos O=rx cos;;' cosa
,
Dividing the second of equations 2 by the first gives
11 SIn w
x=tlin;; CON wco~~ tan 0
From this we can readily obtain
11 sin a cos 0 cos wy cos a sin O=x cos 0 sin (JJ Square this in order to get rid of sin wand we have
(y' sin' a ~I O+xI eos! 0) cos' w2 11' sin a COS a sin 0 cos 0 4:0S wxI C082 u+r cos' a siu' 0=0 Substituting in this the value of cos (JJ giveu above, and performing several easy but tedious reductions, we come finally to the equation of the projections of the parallels in the form
(Iu) xl [&+2 rc sin a sin Or cos (aO) cos (a+O) J+y2 [c sin a+r sin OJ2+2 rCJ7 (e OOS a sin 0
+r sin a cos a)&r sin (aO) sin (a+O)=O
The curve is an
Ellipse ~ ~ >0 Hyperbola according a& c2+2 rc sin a sin 0'" cos (aO) cos (a+O) <II
Parabola . =0
. .
the qnantit.y H2_AB being .here replaced merely by B since H=O and A itt a perfect square and positive.
By the usual process of transformation to the center and axes we find
b' _ _!,c cos 0 (c_~na+a sin 0) .
 02+2 rc sin a sin 0+'" sin' 0_r2 cos' a
(11)
for the axes, and
~,__ _ or COR a (c sin O+r sin 0) _
 c2+2 or siu a sin 0+& siu' 0  r cos' a

(12)
for the center and direction of the major axis. It follows, then, that the centers of the projections of the parallels an lie upon the axis of x.
For tM projection of the equator 0=0
,
Digitized by Coogle
TREATISE ON PUOJECTIONS.
17
(13)
obviously au ellipse whose axes are
(14 )
I r& sin a
41 =&,2 (.,oS2 ~
and whose center is at
(15)
~ , __ ,.so sin a cos a 1  &r C082 a
The distance p of the projection of the pole from the center of the entire projection is found by making 0= d:;' in the expression for ~', and we have thus
~ .
(16)
two points on the axis of e.
or cos a s=r c+r sin a
or COS a pi or sin a
EQUATORIAL PROJEOTION.
For a=; the general equations become
(17)
or cos 0 cos w x=
c+r sm 0
or cos 8 sin til
1/ c+r sin 0
We have then for the general equation of the projections of the parallels
,
(18)
p'
w being eliminated by the simple process of squaring and adding. This is the equation of a circle whose center is at the origin of coordinates. ,
For the elimination of 8 it is only necessary to divide 1/ by x, thus E k~=l~F7IE
1!.=tanw x
We see from this that the meridians are projected in straight lines, and
that the angle included between the projections of any two meridians is v
equal to the angle between the meridians themselves, Fig. 2 gives an FIG. ll.
idea of this projection.
In the case where the'point of sight is without tbe sphere, i. e., where c>r the projection will extend from the equator to the parallel which passes through the point where the tangent from the point of sight meets the meridian PEP'E'; this latitude is given by
r tan 01 = ,j &'::rs .
and the radios of its projection is equal to
or cos 81 a_rosin 0;
Divide now the circle of projection into degrees, and count upon it from the same point E the lati· tude and longitude, and upon the line of the poles PP' layoff eV =e; join V with the extremities of any parallel which it is desired to construct and the intersections of these projecting lines with the diameter EE', viz, ,. and "', are points in the circumference of the circle into which the given parallel is projected; we have then merely with o as a center to describe circles passing through these points, and they will be the projections of the parallels. The meridians are of course constructed by merely drawing the diameters of the circle of projection.
3 T P
Digitized by Coogle
•
18
TREATISE ON PROJECTIONS.
MERIDIAN PROJECTION.
We have already found for this case
(19) from which
or sin 0
3]=  ,:
o+r eos 0 cos III
or cos 0 sin 01 y= c+r cos 0 cos III
ytan 0
3]=.
SlD UJ
It is to be observed that for the J1egati"e coordinates the values of 3] only change sign, while for negative longitudes the 3] remains unchanged and 1/ changes its sign. For the projections of the meridians climinate 01,
from which follows
ta 0 3] sin III
n _
y
• , :x;t sin' III
SlD 0=    ..  yt+rsw2 OJ
which being substituted in either of the expressions for 3] or '!I would give us the equations of the projections of the meridians. Similarly the projections of the parallels may be found by eliminating til between the expresstona for 3] and 1/.' A further consideration of this projection in the general case would be productive of but little that could interest, 80 we shall leave the subject here, taking it up, however, in the various special cases of· perspective projection that we shall study,
ORTHOGRAPHIC PROJECTION •
.In the case of orthographic projection the eye is supposed to be placed at an infinite distance from the center of the sphere, i. e., 0=00. The projecting' cone becomes then in this case a cylinder, the right section of which is a great circle of the sphere, Here there can no parabolas or hyperbolas occur as the projection of any circle of the sphere, but all circles will be projected in circles, ellipses, or straight lines according to the inclination of their planes to the axis ef the cylinder. This projection is not used for geographical purposes, though it has been for celestial charts, and is commonly employed for architectnml and mechanical drawings.
For 0=00 equation 2 gives
(20)
3]=r (sin a COS 0 COS IIICOS a sin 0)
lI=r cos 8 sin til
The general equation of meridians now becomes
(21)
:c' sin' 1113J1/ sin a sin 2 ",+" (Isin' a sin' III)r COS'a sin' 01=0
The equation of an ellipse whose semiaxes (equations 3) are
(22)
for the center ~=7j=0, and for the direction of the m~or axis tan'lII= :1
SID a
The equation of parallels is now (mde equation 9).
(23)
an ellipse whose semiaxes are
a'=r cos 8
b' =r COS fJ sin II
The center is at
~=r cos a sin 8 and the dh:ection of the major axis 111=0.
Digitized by Coogle
TREATISE ON PROJEOTIONS.
19
The equation of the projectio!!,. of the equator ilt
(24) a:'+1f sin' ar sin' a=O
for which
And finally for the pole
(25) p= r cos a.
These expressions are suftlcient to determine the orthographic projection for any position of
the eye. .
ORTHOGRAPmO EQUATORIAL PROJEOTION.
The oondition that the eye should be on the axis of the earth, aud the plane of the equator that of projection, is arrived at, as in the general case, by making a= 2; we find then
(26)
fl:=r 60S 0 cos w
s=r cos 0 sin w
and, eliminating 8, the meridians are given by • 1!=tanw fl:
. and the parallels by (27)
Thus the meridians are projected in stmight lines passing through the center of projection and the parallels are projected into their true sizes as concentric circles.
If the celestial sphere be thus projected it will be desirable to find the ecliptic. This is simply a great circle whose plane has an inclination of 230 28' to that of the equator. Their line of section has the longitude 00 or 1800. It is obvious that the reqnired projection is an ellipse whose major axis =2r.and is coincident with the projection of the first meridian, and whose minor axis = 2 r cos 230 28' and is coincident with the projection of the meridian of 900.
MERIDIAN PROJEOTION.
In this case the eye is in the plane of the equator, usually also in that of the first meridian; here then a=O and
• (28) fl:=r sin 8 y=r cos 0 sin III
The equation of the projection of meridians is
(29)
a:' 11' 1
4'+f'lsin' w
the equation of an ellipse for which
The equation (30)
b=r sin t t=:O
•
being independent of w, is the equation of the parallels; i. e., the parallels are projected into right ,
linea parallel to the axis of y, Of, the Marne thing, parallel to the equator. •
I:<'or celestial charts the plane of projection is usually that containing the axis of the equator and of the ecliptic, or simply the solstitial eolure, The projections of the equator and ecliptic and all parallels to either will in this case be right lines. The center of the projection will represent the equinoctial points, and the solstices are projected in the extremities of the ecliptic. ])eclination
circles of right ascension ia and meridians of celestiallongitnde w are projected in ellipses whose major axis equals 2 r and whose minor axes respectively equal r cos a and r sin w.
The orthographic projection has the disadvantage of giving the neturalstses only at the center olthe chart. Towards the outside of the projection the portions of the earth's projection are much too small, and at the limit are infinitely small. Moreover, only one hemisphere can be represented upon a single chart.
•
Digitized by Coogle
20
TREAtISE ON PROJECTIONS.
8TEREOGRAPHIC PROJECTION.
In this case the eye is on the surface of the sphere; i. e., in equations Zwe have c=r, and in consequence
(31)
r cos 0 sin w
Y=I+ cos a cos 0 COS w+ sin a sin 8
r (sin a cos o cos w COS a sin 0)
z=   .  .
1+ cos a cos 0 COS w+ SID a SID 0
Our general equation of meridians becomes .
cot « z3+1f2 zr tan a2 yrr=O cos a
(32)
Here H=O and A=B, the well known conditions that the general equation should represent a circle. The center of the circle is at the point
eot e 1/=rcos a
and the radius is
R= r.
cos a sm w
For the proieetion of the pole we have
rCOSa a
P=1+ sin a=rcot_2
a point through which the projections of all tbe meridians must pass. The equation of the IOC118 of centers of meridians is in this case
(33)
a straight line parallel to the axis of z at a distance from it =r tan a.
The eqnation of parallels becomes
FIG. 3.
(34)
cos a r' (sin tJsin a)
z3+r+2 rz.  ... +.  . =0
SID a+sm6 SID 8+SID a
•
a circle whose center is at
whose radius is
R,=.rcos t., slDa+smO
The equation of the equator is 04).
(3D)
• the position and magnitnde ~f this projection being given by
R'=rCOSOOa
e=rcot a
We thns see tlijl.t both meridians and parallels are projected in circles for tbis kind of pnUOOtion, and since, by varying the angle a, we can cause the plane of projection to assume any position relatively to the equator and parallels, it follows that al~ circles oj the sphere are prldected in circlea.
8TEREOGRAPHIC EQUATORIAL PROJECTION.
The plane of the equator is here taken for the plane of projection and 80 a=;; this gives
(36)
rcosOcoscu l+sin 0
Digitized by Coogle
TREATISE ON PROJECTIONS.
21
Calling' the complement of the latitude 0, we have from these equations
In order to do this it is necessary to remember that the stereographie projection of eve,'Y circle is a circle. Draw now, as in Fig, 4, two diameters of a circle perpendicular to one another 88 AB and CD and the chords DE and DF cutting AB in e and I, then fe is the stereograpbic projection of' the arc or chord FE, Now, the angle FED is measured by one halt' of the an' FDand
BD+AF AD+AF
angleefD ~ ~=iFD
o
(37)
~=tan (II :x:
The meridians are thus projected in straight lines passing through the origin, or simply in the diameter of the equator, The parallels are projected in circles of radius =r tan i· For the
/
equator itself '=2 and the radiu~ of this projection =r, the radius of the sphere. Fig. 3 represents
,
this projection, the eye being' placed at the south pole; P is the north pole, and ABCD is the equator, To draw this a. circle of radius, r is described about any point P and its circumference divided into equal portions of 50 or 100, or whatever may be most desirable. The diameters AP ISO, 30 P 210, 900 P 270 are the meridians of 00, 300, and 900, The paralJels are all drawn about
r
P as a center with radii =r tan 2' Table I, which is constructed by means of this formula, gives
the value of p (the radius) for every 50 of latitude 8 (=90°_,); in the table r is assumed =1.
If a perpendicular be erected at the extremity A of the diameter AO the tangents and secants of all the angles necessary to construct the chart may be laid oft' on it. If, for example, the angle APa=23°, then Aa'=r tan 23, Pa'=r sec 230. If r be taken as unity, the construction will of course be quite simple.
TO DRAW THE ECLIPTIC WITH ITS PARALLELS AND CIRCLES OF LONGITUDE.
•
"
Therefore, FED=tf1J and EFD=leD.
If, therefore, FE be the diameter of a circle on the surface of the sphere, the surface of a cone DEF will cut a plane through AB perpendicular to DC, in the figure fe. Since, however, the plane of this section, on account of the equality of the augles FED and tifD, makes a subooutrary section, the curve of intersection fe is a circle. The distance in the plane of projection from the center of this circle to the origin 0 i8
n FIG. 4.
6 O/+Oe ~
and the radius
Let.t denote the distance of the pole of FE from the point of sight c, and I~ the distance e>f the pole from the circumference of FE. Then we have
CF=.t+.u and from the trianglM DOl and DOe
Of:.r tan ~ ().+/~) •
06=r tan i (.t:...p)
Digitize~ by Coogle
22
TREATISE ON PROJECTIONS.
Substitnting these in the above value of 6, this becomes
6=; { tan 1 ().+p) + tan 1 ().p) l
or
r sin). 6
008 ).+('os I'
and, in like manner,
R r sin I'
cos ).+008 I'
ElMmple.Fig., 3.
If it is desired to draw the parallel to the ecliptic, which is 3()0 distant from its pole, it is only necessary to layoff from P, on the diameter PD., the distance PO
_6 r sin 2~0 28'
  cos 230 28' + cos 3()0
from Q thus obtained as a center and with a radius oq
_ R= r stn 3()o _  cos 230 28' + 008 3()o
describe a circle; this is the required projection. For the ecliptic itself 1'=900, and the distance to its center is PO
=r tan 230 28'
and its radius OQ
=r sec 230 28'
AQO is this projection. For the pole of the ecliptic ,,,=0; whence
Pp=6 r sin 23~ 28' =r .tan 110 44/ 1+ cos 230 28'
Further, the distances of the Point P from the two.points in which the diameter BD is cot bythe parallel to the.
8R=r tan * ().p)
·GENERALIZED DISCUSSION.
•
We shall now take up the problem of perspective projection from a more general point of view.
U ntil now the position of the variable point M has been determined by means of the quantities
ZPM=w PM=9000
The position of M will now be determined with reference to any fixed point, say L, on the surface of the sphere. To this end write ML=x, MLP=W. To determine with respect to P the quantities.
ZPL=fl and LP 900 8
Let, for example, P denote the pole of the equator and L that of the ecliptic; then 9008= obliquity of the ecliptic. For a star M the longitude =9()0 W, the latitude =90x, the declination =0, and the right ascension =MPL9()O or p9()O, denoting MPL by {1.
TO DETERMINE 'rHE V A.LUES OF wAND 0 IN 'I'ERMS OF X AND W, 'I'UE LA.TITUDE AND LONGI· TUDE OF M.
In the spherical triangle PLM two sides PL=9008 and LM=x, and the included angle ltlLP = W are known, and from them we have for the determination of MPL={1 and PM=90oo the formulas
(38)
cos 0 sin {1=sin 1 sin W
cos 0 cos {1=cos 8 cos Xsin 8 sin X cos W sin 0=8in 8 cos X+cos 8 sin X cos W •
Digitized by Coogle
,
TREATISE ON PROJECTIONS.
23
p and 8 being found from these, II) is given by the relation lI)=ll+P
The values of these qua~tities II) and 0 must then be substituted in equations 2 to find the values of x and 11, the coordinates of the projection of M in terms of the new variables.
APPLIOATION TO THE STEREOGRAPHIC EQUATORIAL PROJEOTION OF PARALLELS TO THE EOLIPTIO.
For this case a=9()O and the distance of the eye from the plane of projection=r. The values of x and 11 already found are
fr
(39)
r cos 0 cos II)
x
 l+sin 0
r OO8l.sill II) 11 1+sin 0
Now since, according to assumption, the pole L lies in the circle whose plane passes through BD . perpendicnlar to the plane of the paper, we have
a denoting the obliquity of the ecliptic. Tbe distance of the given parallel from the pole is x, and for the points in which it is intersected by the circle perpendicular to BD, "' __ OO and 1800. The above equations .become by the substitution of these values
(40)
cos 8 sin P=O
008 0 cos P=siOt OO81cos£ sin 1 sin 0=008, cos1+sinr sin 1
Whence it follows that
(41)
Similarly for 'F.1800 (42)
In general for the two points
•
(43)
Substituting these values of 0 and OJ in the expressions for x and 11 we have for these coordinates
(44)
r sin (£±1) r tau ~ (r:l:,,)
11 1;+cos ('±1) 2" A
For the pole of the ecliptic 1=0 and
(45) x=O
, Y=T tan 2
. Since in steroographic projection all circles of the sphere are projected as circles, it will only be necessary to find the projections of any three points of the sphere which lie in a circle to. be able to determine the center and radius of the projection of the circle. Calling R the radius of the projection, and; and 'I} the coordinates of its center, also Irl 111 Irt1/s, X3!l3, three points of the circumference, we have then
(46) {xle)2+(1II7j)2=R~
(llIte)'+(1I2'I})2=R' {1r3~)2+ (Y37j )2=R'
from which
Digitized by Coogle
24
TREATISE ON PROJEOTIONS.
(47)
G (a1txl)E (X3XI)
7] (a1tXI) (Y3Yl)(;l;:JXI) T(y._y'I)
R= V(XIe)2+(YI7])'
where
Of course this lengthy analytical process need not be employed, for, baving found the three points 1, 2, 3, it is only necessary to draw lines 12,23,31, and bisect any two of them by perpendiculars which will meet at the center of the circle.
Two points are sufficient to determine the projection if tbey lie at the extremities of a diameter of the sphere. For example: if two points are known on the sphere whose angular distances from the fixed point L are IJf and ISOO'[I' we have for the determination of /1, 0, UJ
cos 0 sin /1 = sin I sin 1[f
cos 0 cos /1=cos 8 cos 1::1: sin 8 sin 1 cos 'l' sin O=sin 0 COS1±COS 8 sin I cos IJf
Tbe opper and lower signs give us two values of /1 and 0, also UJ=n+/1. Now, from the known expressions for x and 71, viz:
(48)
(49)
r cos 0 cos UJ x= I+sin 0
we will be able to determine two points x1111, X,Ys. The coordinates, then, of tbecenter of the projected circle are
and for the radius
STEREOGRAPHIC MERIDIAN PROJECTION •
•
In this projection, which is the one commonly employed when a complete hemisphere is to be projected, the eye is placed at any point of the equator, and the plane of the meridian 9()0 distant from the eye is taken as the plane of projection.
For terrestrial charts the plane of the meridian at Greenwich is usually taken as the plane of projection, and the eye will then be at the point whose longitude is 900 or 2700. But here, as in the former eases, tbe meridian passing through the eye is to be taken as the first meridian in the reckoning of longitude.
This projection ~ill give the means of representing the two terres.trial hemispheres upon two separate charts. If it.is desired to obtain maps of tile polar regions the stereographic equatorial proieetion should be employed. For tItis case we bave x=O, and equations 31 become
(51)
r sin 8
x
I+cos 0 cos UJ
r cos 0 siu w y=1+oos 0 C08 w
The equationof meridiaus thus becomes (equation) 32)
(52)
a circle whose ceuter is at
e=O 7]=r cot UJ
and whose radius R=r cosec w. For the bounding meridian UJ=9()O, and R=r, which determine the bounding circle of the chart. For tbe meridian passing through tbe eye we have w=O, therefore, R=1j=oo , and ';=0; this meridian is thus projected in a straight line, which is of course ob
Digitized by Coogle
TREATISE ON PROJECTIONS.
25
vlous without any proof. For the distance from the center of the map to the intersection of the meridian under consideration with the equator it is eMy to see that we haveh=r tan ~
For the parallels equation 34: becomes
(03) n circle wbose center is given by
~'= r cosec 0
and WhOS6 radius is U'=r cot 0.
The equator is projected in a. straight line, as is obvious from the conditions
R'=.;'= 00
For the distance from the center of the map to the point of intersection of the parallel under consideration with th~ first meridian, we have h' =r tan ;. To construct this projection, draw a circle with radlus BA=BO=r to any convenient scale. The equator and meridian passing through the eye are proieeted in a pair of rectangular diameters. Take AO and DE for these lines, D and E are, of course, the poles of the equator. Lay oW on AO, to opposite directions from B, the distances
7j= ±r cot rAJ
giving rAJ any convenient series of values. The points thna obtained are the centers of projections of meridians. The values r cot rAJ giving the centers of meridians that lie on the + side of DE and the values +r cot rAJ giving the centers of the meridians th~t lie on the  8id~ of DE. With these points as centers, draw eireles of radii = r cosec rAJ and tho meridian projections WIll be constructed.
.
Similarly for the parallels we lay oW distances above and below B on DE= ± r C08('C 0 and
with these points as centers draw circles of radii =r cot 0; these will be the projectlous of the parallels. Here, however, the ± r cosec 0 give the centers of tlte circles lying on the ± side of AU
respectively, which is the opposite of what held iu drawing the meridians. .
TO PROJEOT THE EOLIPTIO AND ITS PARALLELS.
For tbis case we have
.
Letting g denote the distance of the given parallel from the pole of the ecliptic and 1/'0 and 1800, equations 38 give
(54) (1=0 rAJ=O o=9()O(~±1)
•
eubstitutiog these values in the a.bove values of Z and 11 we obtain
(06)
. z= rtan OOO(,±,[) 2
Take, therefore, from the poiut D, on the diameter DE, the dtstances
(56)
The points thus obtained are those in which the parallel to the ecliptic cuts the diameter DE. The distance between these points is conseqoently the diameter 2 R of the projection, and middle point of the distance is the center.
Oall h the distance from the center of the chart B to the center of this projection, then
(57) R_Z,zl s_Z,+Zt
 2 fJ_ 2
4 T P
Digitized by Coogle
26
TREATISE ON PROJECTIONS.
or, 8nbstituting the values of z, and all
(57')
r sin 1 R=.  8m ,:+0081
By means of these equations we can draw all the parallels to the ecliptic by merely giving 1 the proper values. For the ecliptic itself 1=9()O, and
R=rcosooc
For its pole 1=0, and
this being, of eonrse, the distance of its pole from the origin of coordinates, along the line DE.
THE ANGLE AT WHICH THE PROJECTIONS OP TWO GREAT OIROLES OUT IS EQUAL TO THE ANGLE AT WHICH THE CIRCLES THEMSELVES OUT.
Let D (Fig. 4) be the point of sight, and P the point of intersection of two great circles, aR, for example, the circles making with each other an angle equal to cu. The plane of projection passes through the diameter AB of the sphere, and is perpendicular to DC. Let m be the projection of a point 1\1 on the surface of the sphere, Now, from our general equation for the project.ion of meridians by the stereographic method, we have
O"=~=r tan a
Also
tI."
tan mptJ= =cot oa
On+8p
since
aI Op=cot2
From this it follows that mptJ=9()Ooo. If, now, PO is a circular arc whose center is at tI" or the same thing, if pg is the projection of a great circle through P, the angle "pg=cu. Likewise a second great circle, also passing through P and making an angle cu' with the same meridian from which cu was measured, would have for its projection a circle cutting the line AB at an angle cuI. The two projections therefore would make the same angle cu cu' that the circles upon the sphere make with. each other.
THE SAllE PROPOSITION ALSO HOLDS POB SMALL CIRCLES.
Let 71 (Fig. 4) be the center of the projection of a parallel of latitude 0 and m the center of the projection of a meridian of longitnde cu. These circles intersect at right angles on the surface of the sphere, Further let t be one of the two points of intersection of these circles. Then for the parallel
o _~ r cos a 71 sin a+sin (J
r COB (J ·71t=..SID a+SIDI
For the meridian
O,,=rtana
• cot III fltn=rcos a
mt= !".COB a SID III
hence follows:
... _ ,. (l+sin a sin 0)
,.p.      
COB a (sin a+sin 0)
and
or finally:
Digitized by Coogle
TRF.)ATISE ON PROJECTIONS.
and consequently the angle tntp. is a right uugle, as is also the angle of the two circles on these diameters passing through the point t. Hence the projections of the meridians and parallels cut 3t righ( angles.
In stereographio projections we see, then, that all circles on the chart intersect at the same angle that they do on the sphere, and ~ that all angles on the sphere are projected in equal angles on the chart. It follows from this that the projection of auy infinitely small portion of the sphere is similar to the infinitesimal itselfthe only difference being in the relative sizes. This property is one which lies at the foundation of some of the most interesting and elegant investiga.tions of the problem of projection; for the present we shall say no more concerning it, bnt will take it up in another place and fully develop it. The fact that circles are projected in circles, and that the infinitesimal element of surface and its projection are similar, are the reasons why the stereographie projection is the one ftl08t commonly employed for celestial and terrestrial charts. It is, moreover, evident that not only whole hemispheres but also any part of them may be projected in this way, 88, for example, any single country or continent, The point of sight should be chosen as nearly as possible opposite the middle of the part to be projeoted.because the further the part lies from the normal upon the plane of projection from the point of sight the greater is the distortion of the projection.
'I'HE DISTANOE BETWEEN TWO POINTS ON THE SPHERE AND ON THE PROJEOTION.
Let ~ be the distance between two points A and B on the sphere, and b' the distance between A' and B' their projections. Suppose a given point M such that
MA=:t:
MB=y M'B'="
M'A'=:c'
We have thus a spherical triangle MAB 30<1 a plane triangle M' A'B', its projection, with the angles M and M', equal. Now, in the spherical triangle ABM we have
cos ~=cos:t: cos y+sin:t: sin y cos M
and from the plane triangle Observe that
~ft=:c"+"'_2 :c'y' cos :M
Now eliminate M, and after simple reductions we have ~,_ r sin 1 ~ cos i:t:cosl,
From this it follows that if :t: and y are constant, for example, if they ue aaeomed to remain upon the same parallels of latitude, then is ~, proportional to ! sin i~ or to the chord of the arc AB upon the sphere, whatever be the angle M. If M=O, then
and consequently the chord of ~
~'=:c'y'
_~, cluwd (:t:y)
_u :c'y'
from which for ev"ry value of ~, on the chart the corresponding value of b on the sphere can be found. This expression, of course, cannot be used when :c' =" or when :c' is very nearly =". For this case we must make M:_lSOO, then b'=:c'+y', b::z:t:+y,·and chord of I
b' clwrd of (:t:+y)
 :c'+y'
from which the value of I can always be exactly obtained.
Digitized by Coogle
'\
28 •
TREATISE ON PROJEOTIOI4S.
On perspective charts the scale of miles is different at different points. In order to measure small distances and when great accuracy is not required, it will be snfficient to take the length of a degree of longitude or latitude in any part of the chart and cousider that as equal to 60 geographical miles. For greater distances, or where accuracy is important, it will be necessary to take from the chart the latitudes 0 and 0' and the longitudes wand w' of the places and find the distance (J (radius unity) by the known formula
cos 6=sin 0 sin '0'+0080 cos 0' cos (ww') For convenience of logarithmic computation make here
cot 0' cos (ww')=oot Q
then
•
.• • 0 • 0'+ 0 tn' 0' sin (j' sin 0 sin Q+cos 0 cos Q sin 0'
cos u=sm sm cos co ... sm . n..
. sm ...
or finally
\ sin (j' cos (O.a) cos 0=. S111 .a
To find tlUJ longitude and latitude of a place from its position on the chart.
The equation of meridians (3~) is
eot e ~+y3_2 xr tan a2yrr=0 cos a
That of perallels (34) is
~ 11 2 rx _. cos _~_._ + t.J (~in O.siu a) =0
+ + sin a+SlD 0 SID O+SIU a
Make for convenience ~+1I=pl; then the first of these eqnations gives
rI_pl x
cot w= 2 yr cos a+ysin a
the second becomes
These equations give us the means of finding 0 and w if we know x and y. For the stereographic equatorial projection
a=9()O
. _"'_pI smo_ .....  I .+p
For the stereographic meridian projection
. 2x.
SlnO=  l+pl
rpI
cot w= 2 _ yr
GNOMONIC PROJECTION.
This is a perspective projection made upon a plane tangent to the sphere, the point of sight being at the center. It is clear that every great circle will here be projected in straight Iines, A complete hemisphere can obviously not be constructed on this plan, 88 the poiuts of intersection of the projecting lines with the plane of projection will, for the points in the circumference of the complete great circle of the hemisphere, lie at an infinite distance. For gnomonic projection we most have c=O, and in consequence
I .... v C "
X=T (sin a cos 0 cos w":"'cos a sin 0)  cos a cos 0 cos w+sin a siuo
r COS 0 sin w
Y=OO8 ~ cos 0008 w+sin0:8ui 0
•
Digitized by Coogle
TREATISE OS PROJECTIONS.
29
•
"
Tab first tbe simple cases of gnomonic equatorial and gnomonic meridian projection. For
the former of these c:wres
a=9()O
a:=r cot 0 cos UJ
s=r cot 0 sin UJ
The equation of lDeri~ians is thus
1/=a: tan UI
The meridians are tbus projected in straight lines, making the same angles on the projection with the first meridian as the lines themselves do on the sphere. Tbe equation of the parallels is
w+y=rI cot'o
concentric circles having radii proportional to the cotangents of their latitudes.
The construction is extremely simple (Fig. 5). Divide the limiting circle of the chart into any convenient number of parts and join the center to the points which express the latitudes counted from the diameter AA' perpendicular 00 the first meridian; these radii prolonged meet the tangent TT' parallel to this diameter and cut oft' on it distances equal to the radii of the parallels.
GNOMONIC MERIDIAN PROJECTION.
For this case
tan 0 a:=rcos",
The equation of the meridians is then
1/=r tan UJ
w cot' 0  11' ri=:O
The meridians are then straigbt lines parallel to the axis of a: and simply constructed. The parallels are hyperbolas, whose major axis is in the direction of a: and equals 2r tan 0; whose minor axis is equal to 2a: and is perpendicular to the first meridian.
The most convenient metbod of construction by points will be to employ the eoordinate f/J given by
that of the parallels is
_ tan 0
a:==rcos til
and calculate the intersections of the paraUels with the meridians alreacly drawn, by giving 0 a certain value and UI a series of vaJues, 50, 1oo, 150, &c.
We shall now take up the general case where the plane of prQjection is tangent at any point of latitude a.
The equation of the meridians is now , )11
./
."
1/ cos UJf/J sin a sin UI~~COs a sin IIJ
which is the equation of a right line making an angle wj.t,h the first meridian tanI (sin a tan UJ)
Digitized by Coogle
. ~~
"'e'tP "'l~'O
,~
Op=rcot a ~
The equation of the parallels is r: . ~  
a;I (sin'l 0008' a) + 11 si~· 0 + 2 r:e COS a sin a(rt cos' Ocos_' a 0
This is a conic section and is "
an ellipse if sin o>cos a or O>90oa .
an hyperbola if sin O<cos a or 0<900a a parabola if sin o=cos a or 0=9()Oa
. . ~ .
We will consider briefly these three oases. If 0>9()O~01' we have an ellipse whose semiaxes are
. a=r _. __ \8in;~') __ ~~\..·~ &). b~~ .YCOS 0 V .
2 (Inn' ocos' a) "sin' 0  COS' a
•
30
TREATISE ON PROJEOTIONS.
and cntting this meridian in a point whose distance from the center is
. .
The center of the ellipse is at
r sin 2a
e .
 2 (sin' ocos'~)__. ol..
,
For 0<9()Oa, we have an hyperbola whose semiaxes are
and for the center
a r sin 2 0 • j_
cos (Oa) cos (O+a1.. .:t
t . I:: I I
_,\_!: __ . _'_sin 2 a
e  2 2;cos (Oa )cos'("O+"a~)
2 r COS 2 arsin 2 a.:e ~ 11 = 'sln' a
For o=90oa: This gives for the equation of the parallels
This is a parabola whose s~parameter is 9
 .....
p=2 r cot ')
r:
and whose vertex is at the point
'e=r cot 2 a
For the equator we have 0=0, and its equation becomes llI=r tan a
the eqoation of a right line perpendicular to the ftrst meridi3tl and at a distance trom the center= o, (Fig. 6).
FIG. 6.
Digitized' by Coogle
TREATISE ON PROJEOTIONS •
31
.since
r Op=r cot a, p'=pO + Oc= ,
.. C08asma
Then
when q=t&n1 (sin a tan UI).
Iustead of tracing the parallels directly, it will be convenient to determine in the meridians pm,pn', &0., couceived as already drawn, points of latitude 0, and then join these points by a curve.
First, to find the projection m of the point M whose lougitude is UI=OPM and whose latitude is 0=9()OPM. This problem reduces itself to the finding of the distance pm.
In the triangle CpO we have
C,=_?O =.!=
SIDa S10 a
In the triangle Cpm
or
Cmp+Cpm=900+o Cmp=90°toCpm Calling K the angle Cpm and q=OPM, we have readily
. and from the triangle Cpm
cos K=oos q cos a
pm=sin a COS (OK)
r oos s
For the determination of K we have
cos K= cos a
vI + sif'n"'" =a=itan===;.i'=(JI=
which shows that K is constant for all points of the same meridian. For each value of 0 we have, then, for the determination of pm
r cos 0
sin a COS (oK)
For the construction of this projection we may proceed as follows: Layoff from the center 0 (Fig. 7) upon the first meridian Op a length Op=" cot a. This is easily constructed by erecting
FIG, '1.
at 0 a perpendicular 00 to Op, making OC=r, and at C laying off the angle pCO=9()Oa. Similarly construct 0,=1' tan a by erecting Cf perpendicular to Cp. Now draw c6 perpendicular to Ge, and then draw lines from 0, making sngles with Ot equal to the 'longitudes of the meridians whose projections are required. By this means we find upon ee the lengths cp', rv', of the intereepts of the meridians upon the equator Ee perpendicular to pro Then, with, aa a center describe
Digitized by Coogle
,
32
TREATISE ON PROJECTIONS.
arcs of circles passing through the points 11.', ~', &C, and cutting E; in the points Il, v, &c.; joining_ these points to p and we have constructed the projections of the meridians.
We will now determine the point of each meridian of which the latitude 0 is given (Fig. 6).
In the triangle Cpe (1\ triangle in space) the side Ce is in the equator, so that the figure is rightangled at C. Its intersection pm with the sphere is a meridian PM whose projection is in pm, and ill which 1'M=OOoO, the distance of M from the pole P. The right line CM prolonged intersects the line pe ill the projection m. Now, in (Fig'. 7) layoff on Oe the distance Cb=C,l corresponding to the line Ce of the preceding figure; pb will be equal to the distance of the pole from the equator of the map and in consequence to PIJ., and might be constructed by drawing from p as a center an arc of radius PI.I. intersecting Ce in b. It is now only necessary to draw a line C in making with Cr the angle m'C~=O; its intersection n~' with pb gives the distance pm'pm; the latter distance p". being laid off on the line PIl, already drawn. It may be readily verified that
•
rtan w ep=cos a
and
r cos tJ pm= sin a cos (OK)
The gnomonic projection is not much employed in the construction of geographical charts, but is frequently used for celestial proj~ctions. Suppose that we take the plane of projection perpendicular to the horizon of a point C, whose geographical latitude is " and suppose that this plane meet the horizon at the point 0, whose azimuth is w; let Z denote the zenith of C, and P the pole of the earth. Draw the meridian PZR of the point C, and the vertical ZO, making RZO = w. The arc PZ is equal to the colatitude of C, i. e., 000,. In order to make the preceding formulas applicable to this case, call PCO=a. In the spherical triangle PZO, we know PZ=900,; the spherical angle PZO=!SOOw; and the side ZO=900j we can now calculate the angle ZOP='F, and the side PO=900a by means of the formulas
sin a =  sin, cos w
tan IJI'= sin w cot tp
If in the plane of projection Up represent the projection of the meridian OP, and Oz that of the arc of a great circle Oe, then zOp=ZO P= 1/'; in like manner, Ok drawn perpendicular to Oz will represent the projection of the horizon ORR, and ee, perpendicular to pz, the projection of the equator
FlG.8.
Ee'. Draw the line Ok (Fig. S), the projection of the horizon, and at 0 erect the perpendicular Oz, making the angle ~P= IJI', IF being calculated by the formula
tan lJI'_cot if sin w;
Digitized by Coogle
TREATISE ON PROJECTIONS.
38
op will denote the Aret meridian. In order to find p we have
s=r cot a
sin a= sin tp cos OJ
Then finally layoff on Op, Oc=r fan a, and the perpendicular eh to poe will give the projection of the equator.
DISTANCE BETWEEN TWO POINTS.
Since in this projection all great circles are projected in straight lines, it is easy to find the distance between any two points. If we apply here the general solution for all perspective projections, it is obvions that it is only neeessery to draw from 0 (Fig. 9), two perpendieulars to oa
•
FlO. 9.
and ob, the radii of the two points, and make them equal to the radius of the sphere. Then, with the three sides all (known) aO', and bO" thus determined, construct a triangle aCb, and finally, from the point C, with a radius equal to the radius of the sphere, describe an arc AB of a great eirele, which will give the required distance in degrees and fmctions of a degree.
These are the principal perspective projections which have been used for celestial and terrestrial charts. Auy number of modifications might be given, depending upon the position of the point of sight, as c may range anywhere from 0 to 00. It would be difficult, however, by this proc 688, to simplify very much either the eonstrnetion or use of the projections by snch means. The stereographic projection is, from the fact that both meridians and parallels are projected in circles, the most convenient to use. The common fault of all of these projections, and one which is indeed incident to the nature of projection, is that only those portions of the sphere opposite the eye are projected in appronmate1y their true dimeD8lons, those near the boundaries of the map being very much distorted.
§ II.
ORTHOMORPHIO PROJECTION.
From the most general point of view a projection may be defined as the representation of any given surface upon any other surface, whether plane or curved, in such a way as to satisfy certain prescribed conditions. In the representation of any nondevelopable surface (6. g., the sphere) upon a plane certain errors are of course unavoidable, but any of these errors may be diminished, or even made to disappear altogether, at the cost of increasing some other. In the particular case of projections which it is proposed now to study, we will assume that the elements of the sphere are similar to the corresponding elements of the projection, or we shall so construct the projection that corresponding infinitesimal areas upon the sphere and upon the map shan be similar. It will be convenient to use the term given by Germain to such projections, and so we shall call them orthomorpbic.
The nature of a curved snrface is determined by an equation between the three coordinates z, y, z of anyone point of the same. By means of this equation anyone of these eo ordinates can be expressed as a function of the other two, or, more generally, each of the quantities z, y, z may be given as a function of two new independent variables, u and 1', and in eonsequenee each point
OTP
Digitized by Coogle
34
TREATISE ON PROJECTIONS.
of the surface will correspond to definite values of u and e. The general consideration of this case will be reserved for another chapter, As we are here to confine ourselves to the projection of the sphere, it is obvious that the two parameters 1£ and v correspond to the spherical coordinates fP and w, or to the geographical colatitude and longitude (since ,=90°0) of a point on the surface. If, as usual, r denote the radius of the sphere, then
xI+r+z2=r'
is its equation, and the known formulas of transformation to spherical coordinates are
• x=r cos w sin tp
'II=r sin w sin tp
Let, now, ~, 7}, Ii denote the coordinates of a point upon any other surface on which it is desired to project the sphere. 'We make this general assumption here, as it is as easy to obtain the results at present sought for any surface as it is for the plane. The~, 7}, Ii are of course dependent upon one another, and, as in the former case, may each be given as functions of two independent parameters, 1£' and o', If the points (x, 'II, z) and (~, 7), e) correspond, then the coordinates ~,7}, e are dependent upon x, 'II, z, and in consequence upon u and v, or, inthe case under consideration, upon, and w. Now, introducing Gauss' notation, we have
and likewise
the a, b, e, a, p, ri evidently denoting the first differential coefficients of x, 11, z, ~, 7}, Ii, with respect to w, and similarly these same symbols accented denote the derivatives of x, 1/, &c" with respect to e. Imagine, now, these points upon the surface to be projected, which we shall call S, infinitely near to each other; these can then be considered as the vertices of an infinitely small plune triangle. To these three points upon S there will correspond three points upon I (the second surface), likewise infinitely near each other and forming an infinitesimal triangle. As the condition of orthomorphie projection is that the corresponding infinitesimal areas shall be similar, it is obvious that the sides of these two triangles must be proportiona.l. Denoting by tU and dtl corresponding linear
elements of S and I we have '
tlt1=md,;
fit, denoting the ratio of the linear elements of the two surfaces, is in general a function of w and tp and varies from point to point of the surface. In our case fit is a constant, ,and consequently the corresponding elements of area upon S and I are similar.
The ordinary expression for the element of length ds is
which becomes, on substituting the new values of h, dy, ilz, ds'=(a'+bI+c')dwl+(a'A+bl2+c")~2+2(aa'+hb'+cc')dwdfp
Similarly
cltJ2=(a'+P'+r) dw'+(a"+p"+r") d,'+2 (aa' +fJP' +rr') dwdtp . Now, since m is constant, the equation
gives
We can thus write
a2+1"+r=m' (a'+b2+c2)=:.tnSE a"+pl2+r"=m2 (a12+h"+ol3)=mlG aa'+PP'+rr'=mJ (aa'+hb'+cc')=mJF
Digitized by Coogle
TREA.TISE ON PROJEOTIONS.
35
~ da=O, we find by solution of the resulting quadratic
. fLp {F  r.;}
dW=lf =f: "F'I!JG
hm this we derive immediately, i as usual denoting "1, Edw+ Fd,d: td, "EG~F'=O
or
Edw+Fd,+tfLp yEGF'=O Edw+ 1!'fLptfLp YEGF'=O
Oall R and R' the integrating factors of these two differential equations and 8r88nme for the integral of the first
p+iq=const. piq=colfst.
Edw+F~+tfLp "EG=FI=Rl (dp+Uq) Edw+Fd~ YEG F'2=R,1 (dp4dq) Multiplying these two equations together,
and for that of the second
and there follows
or, making
.EU=IRR']l (dp'+dq') lRR'E]' =,.
dr .. ,.(dp'+dq')
In precisely the same way we can find for the surface X the integrals
then
P+iQ_const. and for the element of length
PtQ = coDSt.
"
Tbese two expressions for cia' and dr can be written in the forms
dr=,.(dp+tdq) (dpidq)
and from these, by virtue of the condition d4=mda, we have
, .
tnln (dP+idQ) (dPidQ) lC= (dp+wqndpidq)
It is evident that tbe nnmerator of the rigbt.hand side of this equation is only divisible by the denominator wben dP+idQ is divisible by dp+idq, and dPidQ is divisible by dpidq; or when dP+tdQ is divisible by dpidq, and dPtdQ is divisible by dp+idq.
In the first case dP+idQ will vanish when dp+idq=O, or P+iQ will be constant for p+iqcon. stant; i 8., P+iQ will be merely a fnnctionp+iq and PiQ similarly will be a function of piq. Placing tben
f+iQfl (p+iq) PiQ It (piq)
P+iQfJ (piq) PiQ=I. (p+iq)
It is easy to see that both assnmptions give results which differ only with respect to their signs. The functions /! and 12 must also be of the same form since P+iQ and PiQ differ only in the sign of i, All conditions will then be satisfied it' we take one of the functions, say II, and write
P+iQ=1 (p+iq)
replacing for coDvenience/! by I; P will be the real and iQ the imaginary part ofl (p+iq).
Digitized by Coogle
86
TREATISE ON PROJEOTIONS •
.Assume that in general
Now we have
and
dP+idQ=tlf <p+iq) dP+idQ df(p+iq) dp+idq dp+idq
or, aooonling to the above convention,
The expression for .1 becomes now
dP+idQ
dp+tdq , (p+iq)
dPidQ AI •
d itt =w (ptq) rp q
.I=! ~(p+iq) I (piq)
"
which gives the ratio of the original element to its projection.
The results of the foregoing discU88ion may be briefiy summarized as follows: First, find from the assumed equation d8'=O the two integrals
_ p+iq=const.
p iq=const.
Then, denoting by F any arbitrary function such that P shall be the real part and iQ the imaginary part of F(p+iq), we find at once the two equations which give P and Q in terms of p and q, or we have the sought elements of the projection in terms of the elements of the surface to be projected.
Finally, if
then
which gives the ratio of the length of the linear elements of the surfaoes S and I, where
ORTHOMORPmo PROJEOTION OF THE SPHERE.
Suppose we have a sphere given by the equation
The formulas already given for transformation to spherical C',o.ordinatee are
te= r cos III sin " Differentiating these
'II=r sin III sin "
Squaring and adding
d'll=raw cos til sin ,,+r sin IIJ cos ~ d. ,. sin tpd"
d"_rI BinI "dcuI+r'dtpI
If we then make ,"0, we have E=rlaiB'1I
F==O
Digitized by Coogle
TREATISE ON PROJB()TIONS.
87
and
or
The integral of thi8 is
GI::I:i log cot ~ =const.
Now, if F denote any funotion whatever, we will have e the real and 'I the imaginary part of the function
F (GI+i log cotf)
and these values of E and 'I will be the rectangular ooordinates of the projection of the point on the sphere whose longitude is til and whose colatitude is 'P.
KBROA.TOR'S PROJEOTION.
The simplest snpposition tbatwe can make is that the function F (v) is linear, or that we have F{")==K,, where K is an arbitrary constant. We have then
. e+i"=K(GI+ilogcot~)
from whioh
'I =K log cot ~ 2
the known equations for the Mercator projection.
In order to and the ratio", of the correspondiug elementa of the sphere and plane, make
p+iq=.,+, log eot ~
from which derive as 1l8Ual
We had, however,
d8'
• dP*+fijI
and substituting these values of dp and tlq,
Further,
.== ,., sin' , N_l
,=~~v)=~~=K
and consequently
tIt= I!! '(p+iq) '{P_iq)=JNK1= !
v. ,. rsm'P
BA.ll.DING'S PROJECTION.
Suppose we make the supposition that F{v)=KfIIIw where K and' are constanta. As before, e is the real and 'I the imaginary part of
,t<,
or in the assumed case KB"C·H lo&cott )=K6u..no& t =K, no, t +"=K6u..tan'~
Since
we have
Digitized by Coogle
38
TREATISE ON PROJEOTIONS.
and in consequence, for the equation of the parallels,
a series of concentric circles with radii given by K tan';' In like manner eliminating f/ we have
for the meridiana
7J=e tan ltu the equation of straight lines passing through the origin.
For the determination of m, we observe that
f1 q= log cot 2
N=1
Further,
AI( }_ 8F(u} _ ·nT...a.. II' U ;;ut ... ~.t:I
from which it is clear that
aud consequently
or simply
lK tan': m
r SlO,
For the case of 1=1 we And
....
7J=K tan ~ sin fII
K(lcos f/)
m 'S
rsm"
the formulas that occur in the case of stereographic equatorial projection; and thus we see that this projection has the grea.t advantage of preserving the' similarity of innnitesimal areas. Leaving any further application of this method for another place, we will now revert to the beginning of the subject again and develop the necessary formulas for the orthomorphic projection of a spheroid.
Qt~p~
p~~
FIo. 10,
Suppose in Fig. 10 that PQ denote an element of a meridian upon the spheroid and QP) the element of a parallel .throngh Q, the same letters accented to denote the representations of these quantities upon the projection. The condition of equality of angles gives for these'iJitlnitesimal
Digitized by Coogle
TREATISE ON PROJECTIONS.
39
Squaring and adding nnity to both sides of this equation
P'QI2 P'QI2+Q'Pl"= PQI (PQI+ PQl')
Observe that the f3S'tor ~~; depends only on the latitude and longitude 0 and III of the point P; denote this factor by t' and we have from the figure
P'P?=t'PPl'
The ratio of the corresponding linear elements PPl and p/P'l upon the spheroid and upon the projection depends only upon the ooordinates III and 0 of P and not upon the direction of the element.
By the usual convention we have
P'P?=tk"+d7J'
and we know that, denoting by d8 the element PQ of the meridian a (1") do
d8= ~~i=" 81n2"
• denoting the eccentricity of the spheroid; if p denote the radius of the parallell QP, we have QPl=pdw, and consequently
or
Denoting as before the colatitude (i. e., the angle which the normal makes with the axis of the spheroid) by" we have d.p= dO, and the quantity ~ has fot its value
d, _ (1.')d,
';(1,1 cos' ,) sin ,
d8 is an exact dift'erential and we may denote it by d,,; consequently p
'''=J~f_" J sin ,d.p= ! log_!cos ~+ log (~' cos .f);+logG
sm , 1" cos' , 2 1+cos , 1+, cos 9'
The ftrst term of this with the constant is the value of" for the earth suppoeed spherical; collecting the terms this is
.
"=Iog [G tan ;(t;::: :)i]
If we suppose such an angle, that
tan e =tan '(~' cos ')i
2" 2 1+r cos,
then
which is the same form of " that we have for the earth supposed spherical; , may thus he regarded as the polar distance corrected to allow for the ellipticity of the earth, and, the eccentricity being very small, we have, nearly enough, for'
Digitized by Coogle
40
TREATISE ON PROJEOTIONS.
Returning now to the equation giving the ratio of the linear elements upon the spheroid, which we will again denote by S, and the plane, denoted by E, write
tp=mo
then
tU'+d7j'=mo' (du'+duI')
We have now to determine ~ and 7j in terms of _ and til. Write for brevity
~+i7j=al ~_i,7j=pl
Call Po the value of mo when the q nantities _ and til, on whioh it depends, are replaced by their values in a and p, and we have
Now
and, consequently,
da1 dPl da'+ ~al dpl dP+ (dal dPl+~l dfi1)dadfi= 'dadfi
aa da dfi dfi da dp dp da ,...
from which follows
da1 dfil da1 dpl
da (£1=0  =0
dp dfi
and also
dal dfil
dP=O i14=0
or
dal dpl
aa=O dfi=O
and by integration
a1=F(a) fil=FI(fi)
or
a1=1I(,8) pl=.(a) These are equivalent to the results already obtained on the snpposition that the angle between any two elements of the projection is equal to the angle between the two corresponding elements of the spheroid.
The fnnctions F and. are quite indeterminate and may denote any arbitrary functions of _+ ioI anli uiw. But since the variables ~ and 7j are real, as are u apd til, the functions F and II are not perfectly arbitrary, but have determinate values as soon as values have been assigned to F and II. It is obvious from very simple considerations that if the function F is real, then Fl most denote the same function, and if F is imaginary Fl will denote the conjugate function obtained by the change of i into i in F; of course the same remarks apply to the functions II and 111; and so it is clear that each of the two solutions obtained contains bot one arbitrary function, either real or imaginary. We wili consider merely the first of these solutions, viz:
Direct, solution gives
It is easy now to find the value of m or the ratio of the lengths of two corresponding elements upon Sand E. We had
Digitized by Coogle
TREATISE ON PROJEOTIONS.
41
and in order to preserve 88 mnch 88 possible similarity of the notation in this part of the chapter with that employed in the drst part, we will make t=m. Now
;'
d~+tI:"= .. • (tI .. +tJwI)
Denoting the derivatives of F and FI with respect to u+t. and ut., respectively, by F' and FI', we dnd
then,
and so
·~+tl7J·=F' (u+"") F'I (u.tw) (du·+tJwI) mI=F' ("+"") F'I (uicu)
which gives at once
fA=! ",pi (u+"") F'I (teltu)_ ..... !_ "'(le'cos',,) F' (u+""> F'1 (ut.)
p Sin"
Making for brevity
we have 1lnally
1
,..=pfA
Linear elements in projection are altered in the ratio of 1 :,.., and elementary areas in the ratio 1 :m·.
LAGJUNGK'S PBO.JlI:0TI0N.
Observe that from the equation
we have
fD being for the time constant, and denoting by a the real part and by {I the coefficient of' in the derivative, with respect to u of F (.+iw); further
From these we dnd for the radins of curvature of the meridians .
p.
Now, for any point whatever of the 8tll'face, we know that
Oonsequently
d7J ={ldu+ IICfa, tI{I tla a;=au
From these results
or, finally,
(aI+.8.)1
p. C~+~)
k:'C .f}+P* )
6TP
Digitized by Coogle
42
TREATISE ON PROJEOTIONS.
For the parallels we find, regarding" as constant,
\ ~= a~( ~}+jit)
Remembering the definition of the quantity a, viz:
U= "F' (u+tw) F'I (uim)
these two equations become
1 dU 1 .dU
p;,.= aaJ ~= au
Since _!_ is independent of u, we have that ;;, ~~_ =0. If, then, the meridians are represented by eir
p. utltHM
cles it is clear that the parallels are also circles, for a~ =0 is the condition that ~ shall be con
stant with respect to aJ, as well as the condi~on that : shall be constant with respect to". The projections in these cases are of course circles, as the circle is the only curve which will satisfy the condition that the radius of curvature shall be constant.
Write
1
"F/I(u) = F(u)
then the condition
obviously becomes
the double accents denoting the second derivatives of the corresponding functions. The second number of this equation must, by virtue of the nature of the functions ~ and F, be deduced from the first by the simple interchange of i with i, which equality can only exist when each member of the equation is equal to the sam~ constant k; then
fI)(u+iaJ)=~t't(u+'"')+Bor+'i"(v+,",)
Ae and Bo being constants, real or imaginary, and consequently
F (u+iaJ) H+Aet'f(u+,",+BeYJ:(v+Coo) A, B, and H being constants of the same nature as Ae and Bo
The constant k may be clearly either positive or negative; but if we suppose k negative, say =t', we will evidently arrive at the same result as that which would be obtained by changing u into aJ and aJ into u, 80 we shall only consider a positive value of k, =t', and the above equation becomes, on making H=O,
. • eC (u+,",'
F (u+'aJ) =~+'7j= Ae'(U+'''')+ BtFt(u+,",i
Retaining H is only equivalent to a transformation to parallel axes through a. new origin, 80, of course, nothing Isloat by making H=O. Multiply.the third term of this equation, numerator and
denominator, by 
Ae'(,,CooJ+BeC(uCoo)
This gives
• Be2tu+ Ae2Uoo Be2Cu+A(cos2taJisin2tw) ~+71= A''''+AB (e2Uoo+e2Uoo)+Be2Iu A'e2fu+2 AB cos 2 tw+B'e2Cu
Digitized by Coogle
TREATISE ON PROJECTIONS.
43
Regarding A and B as real and positive, this gives, by equating the real parts and the coefficient of the imaginary parts separately,
A 00fI 2 tw+Br""
e Ate .... +2AB cos 2tw+Bir ....
Asin2tw
Eliminating" from these equations and we find a relation between e, 'I, and ., which will be the equation of the circles representing the meridians, and eliminating fU we find the relation between e, 'I, and ", which represents the parallels, also circles, 88 we have already seen. Square these quantities e aud 'I and add the results;
r ....
e'+'1' A'e'''+2AB cos 2tw+Bir
and
The elimination of" from these gives
e'+'1'~ cot 2tw~=O
or
( 1)' ( cot 2 lui)' ( 1 )'
e2B + '1 ~B = 2Bsin2t(;
a circle whose center is at
1 eO=2B
cot 2 to. 2B
and whose radius is
1
P. 2 B sin 2 to.
The circle obviously passes through the origin of the coordinates (e 'I)' The axis of e is then cut by all the meridians in two fixed pointsthe origin and a point distant from the origin==~ ~. These two points represent the poles. and the axis of e is itself the first meridian.
The elimination of fU from our equations gives then the equation of the parallels 88
~+ ' 2 Be 1 0
..  '1 + A'(J4It'B' A*64NB'
a circle whose center is at
and whose radius is
At!'"
P, A'~B*
Designating by 0 thl' origin of coordinates and bJ P the point distant from the origin i; t. 6., 0 and P are the poles of the earth in projection, and denoting by 0 the center of anyone of the circles representing parallels; we must clearly have
then
It follows from this that
Digitized by Coogle
44
TREATISE ON PROJECTIONS.
or the diameters of the projections of parallels are harmonically divided by the poles or points of intersection of the projections of meridians. By taking the origin of coordinates at the middle of OP we can somewhat simplify the equations of both meridians and paralleJs.
1
OP=:s, =2.1., say ;
then the eqnation of meridians becomes
e'+'1'2.1. cot 2ty.I.'=O
that of parallels
M+ '+2.1. 4A'A'e4IAo+l l' 0
,. '1 4A'A1e4"1 ~+ =
It will be convenient jnst here to introduce the socalled hyperbolic and Gudermannian functions. The hyperbolic functions required as given by
sinh"'=i(e8~)
OOSh6=i(e8 +~)
and
1 coth B tanh •
The derivatives of the first two of these are
dsinh (J h '"
dB 008 V'
d OO8h {} lnh "
dB SI.,.
To these may be added the following expressions for Gudermannian functions:
sgl=tanh 1=i. tan '1
cg1==sech 1=800 '1
tgZ=sinh 1=' sin i1
and
 8h 1
008 'l=co 1 = 0fIl
sgll+cgl1==1
It frequently makes the expressions we are dealing with simpler to introduce these functions in place of the complicated exponential quantities to which they are equal. Assume now
sin 'z=isinh1=itgx
with
and we have for the coefficient of ~ in the above equation
~~"')+ 1 e"<A+to) + trlI(A+to)
2.1.6'<~"J_l 2.1.6"(A+to) _qI'tA+toJ 2 A. coth 2t(u+h)=2.1. coth 2 hi
giving for the equation of parallels
~+'1'+2.1. coth 2tu.;+.I.'=O Resume now the explessions for ~ and '1, viz:
A 008 2tu1+ Btrthl
~ A'leu +2AB 0082 tuI+B'trthl
Asin2tu1
" A ...... ,eu..:;+~2A·B cos 2 tw+B'r
The change to new origin involves writing ~+ 2~ instead of ~, which gives
B'tr"'A'et'" '=2B(AJe"'+2AB oos2tu1+B2rthl)
Digitized by Coogle
TREATISE ON PROJEOTIONS.
45
or introducing the new constanta A and A, this becomes
trICt~}6"<~' 1 tanh 2tD
e=l e"! ...... ,+eb(v+.'+2cos 2tw l+cos 2tw sech2tt1
2 1(/0 sin 2 tw 1 sin 2 III
'1= ~+2 (/0 cos 2 tw+e cosh 2t~+c08 2 tw
sinh 2 tD= t sin 2 ltD
Now, since
cosh 2 w= cos 2 UtI
we have
or
F(u+"")_IA tan t("'+III) F(utw)=t.A tan t(ttlllI)
Of course these can be given in rerms of the hyperbolic functions, but there wonId be no gain in so doing; and similarly the values of e and 'I might be given in terms of the Gudermannian functions by means of the preceding formulas, but the results won1d be interesting only from an entirely theoretical point of view. We have now for the value of Q
and
====:=1===:= cos t(iu+w)cos t(if1w)
Q v'F'(u+tw) FJ'(utw) tl
The value of tI, which ·is to be used in all these formnJas, is
u=log G tao i
, denoting the polar distance, corrected, to allow for etlipticity, in the case of the earth.
In the construction of this projection there are two indeterminate quantities
upon which the scale of the map will depend, and the constant t j there is also indeterminate the position of that point of the earth's surface which is to be taken as the center of the map.
Values for all these iudeterminates should be so found that the alteration in magnitude eonsequent upon the projection of any part of the spherical or spheroidal surface shall be the least possible. The solution of this problem involves finding a point for which .. is a minimum, or the neighborhood of which .. is least altered. Returning for a moment to the equation of the parallels
"', 2 B.; 1 0
"+'1 + A1e4"'B' A'et"'B'=
we will solve the problem of finding the points upon the axis of Ie which will harmonically divide the diameters of these circles, or as we may state the problem, to find the two points npon the axis of Ie whose distances from an arbitrary part upon the circumference of anyone of these circles shall be in a constant ratio.
Take the equation
(leg)'+tI=K'[(Ieg')'+y']
this is the equation of the loens of points wbose distances from two fi%ed points (g,o) (g',o) upon the uis of Ie have a constant ratio K; oomparing this with the above equation we have
Digitized by Coogle
46
TREATISE ON PROJECTIONS.
These are multiplied by
1 g=n
•
•
or we have that the circles which represent the parallel@ are the loci of points whose distances from
the two fixed points P, pi have a constant ratio=~ e". This is rather more general than the principle already obtained, viz, that P and pi divide the diameters harmonically. Oonfining ourselves now to a spherical earth, conceive that the constant t has been chosen and assume upon a horizonta,J. line the points P and pi for the poles, taking the north pole on the right. ThiH line ppl is the meridian ",=0; a line QQI perpendicular io ppl at the middle point is obviously the parallel corresponding to ill, or 11=0. This meridian and parallel can of course be made to pass through any point of the earth's surface, and this point will then be the center of 1 he projection or map. A knowledge of this place infers a knowledge of the meridian from which longitudes are reckoned and affords the means of finding the constant 1. For the center of the chart w=O. and il1=O; calling '1'0 the colatitude of the center (instead of {o as in the case of a spheroid) the value of u at
.
the center will be=log tan ~ (G=l) and ,,,=u+h becomes .
log tan ~ +1=0
Now, to find the meridian of longitude "', draw with ppl for base a segment containing the angle n2w, if", is positive, or n+2w if..,., is negative. And for a parallel of latitude (} or !JOOrp, and for
which u=log tan~, describe a circle the locus of points whose distances fromP and pi have the ratio
Now, take up the subject of the increase ofcoaguitude resulting from the projection of any portion of the surface of the earth. We have for m the value
•
2tA "ltPC08' 'I'
m = a[::A2e'"'=C+"'2=A B cos 2t",+ B=2r':" ... =]rs"'in,
4U "l;Pcos' 'I'
.~~~~~~~r
tan' t f_ tan"~o __ ~2 +2008",,+ __ 2
tan' t fo tan" ~
2 2
for a spherical earth of radius r,
4tl
m=. __ =.:. __ 
tau" , tan" !!_o
2 2
r +2008""+ sin 9'
tanl' 'f~ tanl' '!.2
2
With respect to "', it is obvious that m is a minimum where w=O, that is, the alteration is a minimum along the central meridian. Assume, then, w=O, and confine ourselves to the sphere, we have then for the ftrst meridian
4tA
m:::: .....
r sin 9' [tau' ~ cot' ?,o + tan' ~ cot' f t
2 2 2 2
This is a minimum for the denominator 80 maximum; write
Q_ "sin, [tan' ;'cot'~+tan' ~cot'{]
Digitized by Coogle
TREATISE ON PROJEOTIONS.
47
for a maximum,
this gives readily
tan" ~ cot" flo 2 t008 fI
2 2 2t+008f1
Thus, the distance PP' is divided in the ratio : :~::: by a point at· which the alteration is the least possible. Substitution in the above value of m gives for t~e minimum of this quantity
tA (4 t'C08. ,,)
m.= 4r sin"
This is dependent upon t and fI, and we can again assume that there is a value of , such that the derivative of me with respect to" shall vanish; i. 6., if we give a slight increment to" the resulting change in me will be=O, or
•
dmo_O di
Now
..
this gives
2t= v'l+SlD"fI
an equation for determining t when the colatitude" is given. The practical construction of this projection will be gilen in Part II, and need not be referred to here.
The entire theory of the J4grangian projection might have been obtained from the general considerations at the beginning of this chapter. If we assume F (u)=C08 u, we have
[C08 w+Hog cot ~]=}[ ~,'( .. +HOIOO' ~)+ 6,(·~uogoo'~·)]=![e'tan ~+eWcot;]
=~[(C08 w+i sin at) tan ~ + (008 ,,I sin w) cot;]
Equating separately the real and imaginary parte
sinw 7j=tan"
Or more genera1ly, let
F (u)=K cos (a+,8u)
K, ex, fJ being constants, the resulting values of ~, 7J become in this case
~ ~ cos (a+fJw) (tan; +co~;) And again, if we write
and use the known formula
F (u) K tan (a+ifJu)
e2'Xl tan l=i(e2'K+l)
we will ultimately come to the formulas of Lagrange's projection. But enough has now been said on this subject.
It bas been seen already that, if we 888ume for F a linear function, that is F(u)=:Ku, where K is a constant, we obtain Mercator's projection. .Assume now for F the value
F(u+') Ke'(tJ+w)
Digitized by Coogle
•
48
TREATISE ON PROJECTIONS.
nnd also
Since
F1(u'') = KeI<1 e"=cosltu+' sin ltu .... =OO8ltui sin z.,
and
these values ofF and Fl give us
e =K8'. $lOS ",.
In the case of a spherical earth
and
giving
The parallels are thus projected into circles given by es+7JI=K't~2t~
and the meridians into right lines, all Pa88ing through the center of the concentric circles representing the parallels, whose equation is
~=tan lUI
For the ratio", there resolts
/J!"F' lKtanl; _=_._1 =:::;_
SID, sin,
Since I is an arbitrary eonstant, we are at liberty to assign to it any value that we please For 1=1 we have the stereographte equatorial projection whose equations are
I
meridians ~=tan UI
since ,=9()OtJ. Here K represents the radius of the equator. For this case also 1
m= 2 OO8J~ 2
The general valuee of f and 7J may be put in the form
and also
.~Q
810,
It is obvious from these fonnnlas that the coordlnates of any point whose longitude is w are the same as those which correspond to a longitudeIt in the stereographie equatorial projection for which 1==1. If, then, I is a fraction < 1 the projection of the entire sphere will lie in a sector of a
Digitized by Coogle
TREATISE ON PROJECTIONS.
49
circle which is that fraction of the entire circular area. This projection was proposed by Lambert, but fully elaborated and discussed by Gauss.
Since l is arbitrary, we may determine it so as to satisfy the condition that the lengths of degrees upon two given parallels of the projection shall have the same ratio as they have upon the sphere. Call,o and '1 the colatitndes of the parallels upon the sphere, their degrees are iu the ratio of
sin,o : sin'l
and for the chart it is then necessary to write
(tanf!) .
__ 2_ _sm9'1
tan ~  sin CPt
giving then
l= log sin CPllog sin CPo log tan f!log tan ~
2 2
For '1 and CPo may be taken the extreme values of,. For the construction of this projection, called Lambert's ortkomorphic conic prqjection, draw an indefinite line P A for the central meridian, and with P, the pole, as a center drawcircles of radii
C p=K tan'2
C again denoting the correct polar distance for the spheroid; these circles are the parallels; K is
. an arbitrary constant which fixes the scale of the chart; it may be determined by giving, for example, the value of the latitude, for which the radius e is equal to the corresponding arc of the meridian upon the spheroid. Suppose that O=ok, that is the distance from the pole to the equator
on the map is equal to q the quarter meridian. Now .
a denoting the ellipticity and a the equatorial radius. Then, on the above supposition,
The meridians which on the sphere make angled with the central meridian = w, make on the chart angles with the representation of that meridian =lw, and these are the only angles that are not preserved in their true size. If the arbitrary l be determined by the condition that the degrees of the colatitudes tpo and '1 shall have the same ratio as upon the sphere, we know that
I log sin 'l~(lg siu_~ log tan ~I log tan ~
There yet remains one method for obtaining the values of the coordinates ~ and 1) iu orthomorphic projection, which differs entirely from all that we have so far examined; this is known as the method of indeterminate coefficients. The development of the theory of this system gives rise to quite complicated formulas, and consequently from no practical point of view, the general method is useless, but there is one particular case in which the results simplify themselves to such an extent
7TP
Digitized by Coogle
50
TREATISE ON PROJEOTIONS.
88 to make it worth while to examine briefly the method. The particular case referred to is known as Lambert's orthomorphic cylindric projection.
..
FIo.ll.
In Fig. 11, let P A represent a meridian of longitnde w, and MB a parallel of latitude 0; the longitude of M" is w+ dw and the latitude of lI' is odO. The first condition of this projection makes M"MM' a right angle and consequently the triangles M"m"M, Mlm'M similar. Since also the degrees of longitude and latitude preserve the same ratio that they do upon the sphere we must have
MM' dO
MM" dw cos 0
and consequently,
Mm" dTJ" dw cos 0 MM/=tU' = do
M"m" d~" dw cos 0 M'm' = dTJ' =dO
Now, since e and 1J are functions of wand 0
~=:.: dw+ ~:dO
For a giveu meridian w:const., and dw=O; then _d~/=dedO
do
For a given parallel d~=O, and
:1/:11_ d~d u .. dill w
The above equations of condition are, then, in general
d1J ~ M d1J
d =dcos 0 d=d cos 0
<11·0 W 0
[Multiply the tlrst of these by :~ and the second by :: and add; there results
~ _di+d1J d?l._O
dO dw do ss=
tbe wellknown condition of orthogonality of the lines e =const·. and 1J =co08t.] .
We know that any variable quantity z a function of two independent variables til and (J can be given in the form of the series
z=EAc t1+wzB, O'+w'2'Cf 0'+. . . . .
Digitized by Coogle
TREATISE ON PROJECTIONS.
51
where A, B, C, &CO, are constants to be determined when we know a sufficient nnmber of values of z or it8 sneeesslve derivatives for the given values of 0 and w. The series may be still more com
Pactly written if we denote the A in the above by A(O) the B by A(I)the C by A(I), &c. Theseries
f " , , ,
i8 then
The form chosen by Lambert for this series, in the two particular cases of representing e and 7J in such a form, is
toJ Ita' (J) •
e=,l; UI .I; a, C08 t.0
o 0
<ru I~' A (J). '0 7J=,l; w ,l; .D.( SID t.
o 0
atermA(O)+A(O)w+A(O)w'having to be added in the second formula, since sinOO=O. Now,the
o 0 • ,
equations
give on expanding the above forms
(I) IA(O) ao =2 1
a(l) =! [2AOJ
I 2 I
a(l) =1 [Ao+3AO]
I 2 I I
"
Similarly
Digitized by Coogle
•
52
TREATISE ON PROJECTIONS.
We find readily now
a:I)=~ [(i1)A:I+ (i+1)A~a ~\ [(il) (i2)a~I+2t.Ja:O)+(i+1)(i+2)a~~IJ
a:a) =~ [(i1)A~~1 + (i + 1) A~~IJ =2.!.6 { (iI) (i2) (i3) a': + (il) (3i'3i+2)a~1 +(i+l) (3i'+3i +2)a~~I+ (i+1) (i+2) (i+3)a:a}
It is not necessary to give the general case of a(I). the law being obvious, the reader can readily
• I ,
construct It for himself.
In the particular case mentioned above, La.mbert's orthomorphic cylindric projection, take for the central meridian a stl'flight line (vide figure), and upon it layoff the actual lengths of the degrees of latitude; a second straight line at right angles to the first denotes the equator; other parallels and meridians are orthogonal curves cutting in such a way that the degrees of longitude shall be represented in their true length.
Taking for axes of coordinates the central meridian and the equator, it is clear that, the figure being symmetrical with respect to these axes, '1 should contain even powers of IIJ and odd powers of 0, and that ~ should contain even powers of 0 and odd powers of IIJ. Also for 0=0 we should have '1=0, and e a function of IIJ only; for IIJ=O, "i=0, and ~=O. These series are thus,
'1=KO+1IJ2!:JA(O)~+I+IIJ'!;A(I)rl+ •••
o I 0 I
1:_ 'P.i (OI{J1i + a ru (1)..11+
fiflJ~ a tJ) z: a () .••
o I 0 I
Satisfying, as before, the eqnations of condition
d7J d~
=C080
dw dO
and these series are readily found to become
By properly dividing the numerical coefficients in these series, they are readily found to assume the forms
~=IIJOO80+1IJ8 cosO+_~s30 +~ 4cosO+I0cos30+6OO8I0
4. 1. 3. IIJ 4. 8. 1. 3. 5.
+ 134COSO+I54COS30+210cos50+90cos70+ +
IIJ 4. 8. 12. 1. 3. 5. 7. • .. ...
7J=O+IIJ' sin 20 +m'4 sin 2_0+3'Sin 40+ uJl34 sin 20+60 sin 40+30 sin 60
4 4. 8. 1. 3. 4. 8. 12. 1. 3. 5.
+ w8 496 sin 20+1512 siu 4?+1620 sin 60+680 sin 80+ 4. 8. 12. 16. 1. 3. 5. 7.
Digitized by Coogle
TREATISE ON PROJECTIONS.
53
Again, group the terms in sin 20, sin 40, &c., and those in C08 0, 008 30, &c., and these become e=~costJtan~+i cos 30 tan3;+~co850tan5;+~('.os70tanT;+ &c.
The law of these last two developments is obvious, the general term of e being =:u2 1 cos (2il)0 tanIi
that of 'I
=i \8in2(il)otatll('_I)~
i denoting the number of the term. A final grouping of the terms will conduct us to the formulas
. (1+2 tan; cos o+tanzi) ( sin 20tan.Z; )
e=i log '1=o+tan1
12 tan; cos o+tanZi 100820 tan'i
or, by introducing the colatitude 'P,
e=1lo (1+8~" 8in to) If g Ism 'P 8ID to
or simply
~_11o (cosec 'P+si,n to) t t
.. lJ g 00 '1=oos to an 'P'
cosec ,,8ID to
Forming the dUferential coefficients of e and 1) with respect to 'P and to, we obtain
de sin to cos " d1) cos "
~ Isin' to sin' 'P d" lsinl to sin'"
de COSto~" ~ .to.'Pcos,
dw Isin' to sin' 'P dw =1sin2 to sin',
These obvionsly satisfy the known equations of condition which must exist between these dUferential coefficients, viz:
The ratio of the change in elementary areas is easily arrived at from the above values of the differential coefficients; using the formula dl1= ,/de'+d1)', consider a small quadrilateral on the sphere comprised between two parallels and two meridians; its area is =do dto cos o. Now, making 'P=9000, we have for the arc of a parallel, when dO=O,
dI1= dw cos=o~, = . vlcos2o sm' (AI
and for the arc of a meridian, for which d",=O,
the area of the rectangle is then
dto do cos 0 v1cos2 0 sin' '"
and for the ratio of increase
Digitized by Coogle
'\
54
TREATISE ON PROJEOTIONS.
H w=O, m=l; or, the ratio of areas on the sphere and on the projection is = unity along the central meridian. The principal advantages of this projection are thus seen to be:
(a) That it preserves aU the angles in their true size, and consequently gives orthogonal intersections of the meridians and parallels.
(b) The degrees of latitude are equal upon the central rectilinear meridian, and the degrees of longitude in the neighborhood of this meridian differ but little from their true size. This project.ion will be obtained, as \\"0 have already seen, by passing a cylinder tangent to the sphere along a meridian, projecting the sphere upon the cylinder and developing the latter. It is on this account that tho name orthomorphocylindric has been chosen for this projection.
The general subject of orthomorphic projection will be resumed in another place, and a fuller mathematical theory given of this most interesting problem, but before leaving the subject it is of importance to note that if either of the variables e or TJ be given, the other can be found by simple integration. For, from the equations of condition
we have
a; =d1J cos 0 dw dO
a;=~"1cosoaw a1J ~
aO alii cos 0
a; do d;
~1J= dw cos 0  do cos 0 dw
If, then, either e or 1J is given, forming its differential coefficients, and substituting in the corresponding one of these two equations, we have the means of obtaining the remaiuing eoordinate, For example, let 1J=w; then
and
d;=~
COSO
from which
e=log tan i (OOOO)
the equations of Mercator's projection. Lithrow gives the projection of which one of the equations is
e=tanOcosw
The differential coefficients are
a; to'
=·an Slnw
aw
d;_cosw diJcos''
by means of which we find
a __ (sin 0 sin wdO+cos 0 cos wdw)
1J . cos'w
Integrating this
sin e 1}=cos 0
Oombining the value of e with this value of 1J ill such a way as to eliminate w, we have for the equation of the parallels
and in like manner for the meridians is found
Thus the parallels are projected into ellipses and the meridians into hyperbolas having their common center at the origin of the eoordinatee and their major axes in the direction of the axis 1J. Littrow speaks of both the meridians and parallels as being projected in hyperbolas,· which is evidently a slight error on the part of the eminent astronomer.
Digitized by Coogle
TREATISE ON PROJEOTIONS.
55
Finally assume
';=1 log l+s~n ", cos 0 ISlO'" cos 0
Differentiating this gives
tl~_ sin (JJ sin 0 dOIsin2", cos' 0
d; cos", cos II dill Isin' ", cos' II
From these we have
d cos", tlll + sin III_sin 0 cos 0 tIN
7] Isin'", cos' 0 Isin'", cos' 0
of which the integral is
to. I tan 0 7]= n eos e
which is identical with
cot 1}=cos", tan 'P
the formula already obtained.
§III.
ORTHOMORPHIO PROJEOTION(Oontinued).
We will in this chapter take up two projections closely allied to each other, and very interesting in the methods of development employed by the illustrious authors. The first of these is by Sir John Herschel and is found in Volume XXX ot' the Journal of the Geographical Society of London for I8CO. The paper is entitled "On a New Projection of the Sphere" and the author further calls it "Investiga.tion of the conditions under which a. spherical surface can be projected on a plane, 80 that the representation of any small portion of the surface shall be similar in form to the original;" this is, of course, merely the fundamental proposition of all orthomorphie projections.
Assume the radius of the sphereee l and denote as usual latitude and longitude by 0 and ... , and the plane coordinates by .; and 1}. We must clearly have, since .; and 1} are functions of 0 and (I),
'M, N, P, Q being, of course, functions of 0 and «, The elementary rectangle included between two meridians, whose difference of longitude is dIll, and two parallels whose difference of latitude is dO, will have for its sides do and dIll cos 0, having to each other the ratio
tlw
do cos (J
In passing along the project.ion of anyone meridian", does not vary. In passing then from the point whose projection is defined by';, 1) to the point 0" the small meridian. whose projection is defined by ';+il.;, 1}+d1}, .; and 1} must vary by the variation of 0 alone, or
and the distance between these two projected points is, on the BalI!e meridian =do "N2+QZ
Similarly supposing ourselves to pass along the same parallel we have for the distance between two infinitely near points, do being = 0,
dw"MI+p2
These, then, are the sides of the elementary figure on the plane of projection corresponding to the infinitesimal rectangle on the surface of the sphere, and these two flgures must be Similar; which
Digitized by Coogle
,
56
TREATISE ON PROJECTIONS.
condition, being satisfied, obviously carries with it the similarity of an infiuitesimal figure on the sphere and its projection. The sides then must be in the same ratio and the angle they include a right one. The first of these conditions gives
ilw v Mi+P' ilw "nCOSO VN2+Q2do
or
•
The tangent of the angle made by the projected element of the meridian with the ordinate 7J is evidently represented by
or by!
P
and that of the projected element of the parallel by
d~tl8
do
because, lying on opposite sides of the ordinate 7J, if one tangent be taken positively the other must be taken negatively. The condition, then, of rectangularity requires that the product of these tangents shall be = 1, which gives for the other essential equation
M N pXQ=1
or
PQ=MN
But
or
(P)' .
008'0= N
whence we get the following
M=QcoS8
Assume now
a=J~+w
cosO
p=f~III
coso
which gives
dw = I (da  dP) and by substituting these iu the equations d~ = Mdw + Ndo
do = i (da + dP) C08 fJ
d7J = Pd.w + QdO
we find
d~:ci (P +M)da+i(P M)d{1 whence, adding and subtracting; we obtain d(.$'+7J) = PdaMdfi
.d7J = i (P  M) da  i (P + M) dp
Digitized by Coogle
TREATISE ON PROJEOTIONS.
57
The first members of these equations being exact differentials, the sooond must be so also. The conditions for this are
dP dM ~=da.
dP dM tla = dP
But universally
dP=dP tla+dP dp
tla df1
dP=tlMtla_dMdf1
dp da.
The first member of this being an exact differential, the second must be such also. This gives
whence, substituting,
•
d'M+tJ2M=O (I;;;i dp'
The known form for the integral of this partial differential equation is , and F, denoting arbitrary functions
M='(a+if1) + F(aif1)
Substituting this for M in the expression for dP, reducing and integrating, P=i ['(a+if1) F(aif1]
and putting for brevity
A=a+iP=(! !o+ ",)+{f c!:o"')
we find
which, by writing 2F(A) for/'(A)dA, and 2f(B) for/~(B)dB, affords the following values of e and 7J:
e={l+,)F(A)+(li)f(B)
. 7J=(li)F(A)+{l+i)f(B)
in which F andf are the characteristics of any two functions. both completely arbitrary and independent. Suppose, for example, we take
Then
F(u) f(u)=u e=(l+i) A+(li) B=(A+B)+i (AB)=2 (ap)=4w
and
7J=(li)A+(l+')B=(A+B)QAB =2 (a+f1)=4! do =410g tan i(OOO8) cosO
whi(\h is the law of Mercator's projeCtion.
The equations
e=(l+i)F(A)+(li)f(B)
7J=(li)F(A)+(l+')f(B)
being subject to no restriction, it is evident that we may superadd to the general conditions of the problem any which will suffice either to determine altogether or to limit the generality of the arbitrary functions F and/, in the view of obtaining convenient forms of projected repn8entations. Suppose, for instance, that we assume as a condition that the projected representations of all circles about a fixed pole on the sphere shall be concentric circles about a fixed center on the plane. Since the origin of the eoordinates e and 7J is arbitrary, we will fix it in that center; and since the condition is that when. (J is given, and therefore
f dO =a const., flay k COSO
.
the equation between e and 7J shall be that of a circle about the center, we have
e'+7J'=p'=a function of 0 or k
Digitized by Coogle
58
TREATISE ON PROJECTIONS.
For brevity put Then we have
F(A)=X e=(l+i) X+(li) Y aud substituting and reducing .
f(B)=Y
J}z::(li) X+(l+i) Y
that is to say
F (A)f(B)= ~
Since any function of an arbitrary function is itself an arbitrary function, we may without any loss of generality write eF(.) for F (A), and ellB) for f(B). Now
because
J do
a+p=2 =2k
coso
It does not appear that this equation can be satisfied by any forms of F and f more general than the following, viz:
F(u)=(g+ih)u which give for the value of qI'(a+p)
or, what comes to the same thing,
2 r(gh)+i (g+h)] f 0::0
Practically speaking, this expression is useless unless the imaginary term vanishes, or g+1=0, gh=2g, in which case it reduces itself to
4gfdO =4gk coso
whence also
which, since
f do =Iog tan i (9000) coso
reduces itself to
p=2 ,,2 tan2g i (9000)
Suppose g=1. This is the law of the stereographic projection, and the values of e and J} become e=2p [cos 01+ sin 1II]=2p "2" sin (450+111)
.
7J=2p [cos wsin 1II]=2p "2 cos (450+01)
p=2 "2tan i (9000)
In the more general ~ of 2g=n, we find
e=2p[C08 fku+sin nwj=2p"2 sin (450+fku) 7J=2p [cos nwsin 1&w]=2p "2 cos (450+1&111) p=2 "2 tan"i(9000)
To interpret these expressiona we have only to consider that when III increases by any number of degrees nw increases by n times that number; so that if 01 increases from 0 to 3600, nw increases n
Digitized by Coogle
TREATISE ON PROJECTIONS.
59
times 3600. The coordinates of the projection of any point, therefore, are those corresponding to n times the longitude in the case of the stereographic projection. If, then, n be a fraction less than unity, the projection of the whole spherieal surface will, instead of occupying the whole area of a circle, be comprised within a sector, the same fractional part of the whole area. Thus if n=i, the projection of the whole sphere in longitude will be comprised within a semicircle; if n=1, within a sector of 12)0; if n=i, withiu a sector of 24oo, &c.; and the entire parallels of latitude will in like manner be represented by the portions of concentric circles comprised between the extreme radii of these respective sectors. If 'P be the polar distance of any parallel of latitude, and p the radius of the circular segment representiug that parallel, we have, neglecting the coefficient 2,/2 or taking 1 for the equatorial radius in the projection,
from which it is easy to calculate p for each polar distance from oo to 1800. The valnes for the four cases n=l, i, i, i for 'P = 0°, 1oo, ZOO, &c., are given in the table.
The second case that we take up is in its development very similar to Herschel's projection; it is an investigation of orthomorphic projection by Professor Boole. The idea was suggested to Boole by reading Herschel's paper. The investigation is contained in the supplementary volume to Boole's Differential Equations, published by Todhunter, after the author's death. The general theory as given by Boole is applicable to the projection of any surface upon a plane.
Let x, 1/, z denote the rectangular rectilinear coordinates of any point of the given surface; ~, '1 the coordinates of the corresponding point on the plane of projection. Lett.he equation of the given surface be
or simply
Regarding e and '1 as ultimately functions of x, '11, z we have
d~ de ~
~=ailx+aydy+ (kdz
dz., dy, d,;s being connected by the relation
tlF dF dF
diJdz.+ dydY+didz=O
d = ily dz+dy d +dy d,;s '11 dz dy 11 dz
Now for brevity write
d!_A dF dF
=B u=
dz dy
d~ dt: tU
d:i;=a di=b d,;s=o
~"!a' d'1_b' d'1_o'
lk dy d,;s
then
(1) d~=ad.v+bdy+oibs
(2) d'1=a'da:+b'dy+o'th
(3) O=Alk+Bdy+Cdz The two conditions to be fulfilled are, as we already know, the equalit9 of corresponding angles, and the proportionality of corresponding sides, of the element on the surface and the corresponding element on the plane of projection.
Assuming now any point ~, '1 on the plane of projection, let ~ alone vary, and the infinitesimal line generated by de; and since d7j=O, we have
a'd.x+b'dy+c'dz=O Ad.x:+ Bdy+ Odz=O
Digitized by Coogle
60
TREATISE ON PROJECTIONS.
Denoting by 'V the determinant
a, b, 0
a' bI, 0'
,
A, B, 0
d'V
db=M write
then the above equations give
80 that the directioncosines of the intinitesimal line on the surface F corresponding to the line ~
on the plane are .
L M N
v'1}+M'+N' v'U+Mi+N' v'U+=M~i=+=N=2
In like manner, if 7J alone vary we shall find for the directioncosines of the inftnitesimalline on F corresponding to d7J on the plane
II M' N'
v'I1' +M" +N,I v' Vi + M,I + Nt' v'I11 +M" +N"
where
d'V M,_d'V  N'~~
II = da' db' =s« Since the angle between ~ and 7J is a right angle, the angle on the snrface between the lines whose directioncosines have been found must also be right. This gi\"es at once
The ratio of the element of length de to the corresponding element on the snrface is
or
adz+bdy+cdz v' d:J;2 + dy + dzl
and tinally
aL+bM+cN ,lv+:M2+NS
Equating this ratio to the corresponding ratio of jbe length of d7J to that of its projection on the surface, we have
aL+bM+cN a'L'+b'M'+o'N'
~LI+M'+N'= v'ii+M~+N"
Now we know that
aL+bM+cN='V
and
a'I1 +b'M' +e'N' = 'V
the numerators of the last equation only differing in sign. But 'V, being the determinant of the
system •
Digitized by Google .
TREATISE ON PROJECTIONS.
61
expresses, when equated to zero, the condition that if d~ vanishes el7J must also v .. nish; and tU and d7J being independent, this condition cannot be satistled; 80 that the above equation reduces ~
or
(a)
(P)
and this,. with the alreadyfound condition of orthogonality
will fully express the conditions of similarity. If we multiply p by 2', and add and subtract the result from a, we obtain the eqnivalent system
Now
L'±iL=elF~~_ elFd.; ± i(elF a"j__~F a7J)_elF el(;±i7J)_~~ a (~±i7J)
ay dz dz ay ay dz dz dy  ay dz dz ely
Writing then
we have
~+i7J=u L'+iL=dFau_ dFelu ay dz d; dy
.,..,+.:.,.._dFdu dFdu .w. • .w. dz 'iii  d:c @
Similarly
H' iM= dF df) _ dF df) dz d:JJ dx dz
N'_iN=elFdv _ dFel." d:x; dy dy d:x;
Substituting these in the last equations, there result
(~ au _~~ au)' +(~! du _ dF dU)' +(dF au _~ elU)' =0
dydz dz dY dz dx d:x; dz ax dy dy d:x;
(~elf)_aFaf))3 +(~d1)_aFd,,)3 +(dFdV_~d,,)3 =0
ay dz dz dy dz dx d:x; dz dx ay dy d:x;
Each of these can be written in the form of a symmetrical determinant of the second order. Designating by 0), 0" 031 any three quantities whatever, make U equal to the determiuant
0), 0., 03
• dF dF elF
d:x;' dy' dz
au du du
d"X' ay' dz Then the first of these two di1ferential equations is simply
(au)' (dU\S (dU)'
dO) + d8;} + dO, =0
and, as is well known, can be written
(I)
(~)' + (:;)' +(~¥)',
dF~~+~!~~+~~du d:x; d:x; dy dy dz dz'
Digitized by Coogle
62
TREATISE ON' PROJEOTIONS.
and in like manner for the second
(ll)
(~y +(~:y + (CZY,
dF dv +~! ~~+~F dv dz dz dy dy dz dz'
These are the partial differential equations of the first order, serving to determine v and v as funetions of 11:, y, and e.
But it is not necessary to solve these equations in their general form. It is well known that the coordinates 11:, y, and z of any point on a surface, being connected by the equation of the surface, can always be expressed as functions of two independent variable parameters, and these parameters, when fixed upon, become the independent variables of the problem. Let 0 and 01 represent such parameters, and let their expressions in terms of 11:, y, z give
0='1 (11:,1/, z) 01=',(11:,1/, z)
which equations, combined with that of the given surface, will reciprocally determine 11:, 1/, z as functions of 0 and 01. The differential coefficients
dF all:
dF dy
which are functions of 11:, y, z, now become functions of 0 and 01; and further
dv dv do dv dOl d1/=dody+dUlay
dv_dv~+ dvdUl dzdodz druaz
and as ~~ ~,&c., are known functions of 11:, 1/, and z, they are also expressible in terms of 0 and 01. The result of these substitutions will then be to convert (I) into a partial differential equation in which v is the dependent and 0 and 01 the independent variables, and this equation being, like (I), of the first order and second degree in the differential coefficients of v, will be of the form
p(dV)'+Q dVdV+R(dV)'=o
do do dru dOl
For v we have an exactly similar equation, with the same coefficients.
The above equation is, by the solution of a quadratic, resolvable into two equations of the form
dv dv dv dv
dOAI s: =0 (liJ4 dOl =0
To these correspond the respective auxiliary equations
If the integrals of these are respectively, then we have
Now, v being determinable by an equation of the same form as v, it follows that of the above two values of 11 one must be assigned to v, so that the solution of the problem will be contained in the system
or in the system
v=1f(T) v=IJI'(S)
The particular forms of the arbitrary functions' and IJI' will depend solely upon the nature of the problem under consideration.
The first members of (I) and (II) are obviously essentially positive; and so, if the intermediate transformations are real, is the first member of the equation whose coefficients are P, Q, R. Hence
Digitized by Coogle
TREATISE ON PROJEOTIONS. 63
the quadratic determining AI and ~ will have imaginary roots of the form a ± ip. ffitimately, therefore, it will suffice to integrate one equation of the auxiliary system
dw+AldO=O d(ll+~dO=O
and then to deduce the solution of the other by changing i into i.
Suppose, now, that the surface to be represented is an oblate spheroid, such as the eartll; take the plane of the equator for that of projection and the center for origin. Let the eo ordinates x, y paafj through the meridians of ()O and 900, respectively, and z through the poles. The equation of the surface will be
when a is the earth's equatorial, and b its polar radius. Let also the latitude of the points (x, y, z) be represented by 0 and its longitude by (II. We have
a:'+!f'.i'
F=ac+7)2=l=O
•
and substituting in (I), we have
w+1f+~
a' b'
~ du +y_ du+z d~ a2 ax a' dy 1)1 dz
or, expanding this and writing ~=k'
(III) (w+!f'+k'i') { (~)' +(~;)' +(~Z)' } \_( X X :: +y~; +k'Z: "0)
Now as x, y are coordlnates in the plane of the equator, and x passes through the first meridian we have
q
FIG. 12.
Digitized by Coogle
64
TREATISE ON PROJECTIONS.
Again, representing in Fig. 12 the meridian of the point P, or (x, 1/, z) touched by the straight line QR in the same plane, we have CM= v w+ 11 and MP=z. Therefore if v w+y'=p, the equation of the meridian is
and tHat of the tangent is
pi, z' being current rectangular coordinates of the tangent. Hence
a2z k2z
tan CQR 1JZp v w+"
Bot ()QR=O. Therefore finally
0= tanI k'z
vw+'!I
and we must now transform (III) so as to make 0 and III the independent variables. From the last
_.
equations combined with that of the surface we have readily
ak cos III
x 
.  vk2+tan' 0
and substituting in (III) we obtain
(IV) o=sec'O{ (~Y+(:i)'+(:)2} ( COSIll::+sin 1lI~;+tano:y
ak sin III s= vkz+tau' 0
a tan {I
Again
du_du do + du dill dz do dz dill dz
,
Now
do k'zx
dx= v x2+y2 (.ir+;a+lAzi)
when K=k2+tan' o. In like manner
do sin 0 cos 0 sin IlJ vI{
ify=  aK
Bin 0 cos 0 cos III vK aK
dill sin IlJ vK dx= iiK·
dill coBwv1{ dy=ax
Hence
au ,'K [ . du. dU]
d:l;= aIr S100 cos s eos e dOsm w ~
du v'K[. . au dU]
• dy= aK sm 0 cos 0 SID IlIdiJ + cos w dW
du_ vK K2 2 du
s«: aK cos 0 ao
Substituting these in (IV), and dividing by the common fMtor a~1I we have, on reduction,
(~y + cos' 0 [l+(KIIl) cos' 0]2(~dY =0
which is resolvable into
dlt . du
iI.;1 cos 0 [1+ (K2_1) cos' 0] dO=O
du . du
dw +" cos 0[1+(K21) cos'O]i1.o=O
Digitized by Coogle
TREATISE ON PROJECTIONS.
65
partial differential equations, of which the integrals are included in the common formula U=f(f C080[1+(~~_I)OOSI0]±iUl)
Now
J dO JdO If coso dO
coso [I+(J{II) cos¥O] coso + (IK) l+(K'l) cosiO
J dQ , J cosodo
= cos 0 + (IK ) K'(KZi) sin' 0
_J_dq_+ IJ cosodo (. r_alb"\
 cos 0 ' lr sio'O SIDce  aZ)
U=f[IOg{ (~~::!~Ditani(;+o) }+iw ]
or changing' (UI) into' (6"), since' is an arbitrary function U=, f (I_e s~n 0); ton i(~ +0\6%;' t
l l+e S10 0 .!,) f
,,=1[' f (Ie s~n_!); tan ~(7f + o ")e*;' t
l I+e SID 0 2 J
J et p and (/ denote the polar coordinates of that point in the plane of projection which corresponds to the point whose latitude and longitude on the surface are 0 and UI; and let
S=(Ie s~n 0); tan i(:'+O)
I+e SID 0 2;
then the complete solution assumes the very simple form
(V)
Assume that the parallels of latitude are projected into circles round the pole. This requires that I' be independent of UI, a condition which is satisfied in the most general manner by assuming
.
". (W)=C'W"
We then find
whence, on multiplication and division,
whence, A and B being new arbitrary constants derived from C and O',
•
p=AS" (/= ± nUl + H
Observing that (/ and UI should vanish together, we have B=O, and tbe equation t1=::!::nUl sbows that tbe surface of the sphere will be projected into a sector of a circle, the arc of which is to t be 5 T P
Digitized by Coogle
66
TREATISE ON PROJECTIONS.
circumference of the circle u : 1. Thus if n=!, i, &c., the sphere is projected into a quadrant, a semicircle, &c. These arc of course the results that we have already obtained in several difteront places.
The other equation for p gives'
If 0=0, we fiud
p=A
whence A is the distance of the equator from the pole in the plane of projection; and if that (listance, which is arbitrary, be assumed as the unit, we have
= { tan ~ Cr. +0) (1_ ~~n O)~' }
f' 2 2 1+_ Slll 0
for the distance from the pole of that parallel whose latitude is (I. We can throw this expression into a more familiar form by assuming p=i +0, and introducing an auxiliary quantity q defined
by the relation ..
t cosp=cos q
We have then
_ P)C q)fM
p( tan 2 cot 2
Table IV gives the values of f' for the sphere and for the spheroid whose eccentricity is .08 (about that of the earth), for each ten degrees of polar distance, for the values n=1 and n=!.
The preceding investigation by Boole it'! seen to be much more general than that of Herschel, the latter con1lning himself merely to the projection of a sphere upon a plane.
§ IV.
PROJECTIONS BY DEVELOPMENT.
In order that a surface may be represented upon a plane without any change of angles or areas, it must be such an one as can by actual developmeut be rolled out upon a planeall parts of it coming by a continuous motion to coincide with the planeas, for example, all cones and cylinders. If we desire to make a projection of a comparatively small region, the operation will be rendered quite simple if we can substitute for the actual surface to be projected a certain portion of some developable surface upon which are drawn the meridians and parallels. The construction of these lines upon the developable must of course be such as to make the new elements correspond as closely as possible with the actual elements of the sphere, The attempt to make projections of this kind hl\S naturally given rise to two methods: (1) Conical Projections, (2) C~liDdric Projections. We will first consider the former of these.
Conceive a cone passed tangent to the sphere along the parallel of latitude which is at the middle of the region to be projected. Also imagine the 1) lanes of the different parallels and meridians to. be produced until th ey cut the cone. We will then have npon the surface, of the cone small quadrilaterals corresponding to those of the sphere]' the maguitudes are different, but the angles are obviously the same. Now develop the cone upon a plane; the meridians will clearly become right lines from the vertex of the cone to the different points of the developed parallel of tangency (or any other), and the parallels will be concentriccircles, the vertex of the cone being the common center. The parallel of tangency is obviously the only one unaltered by the development. The quadrilaterals upon the sphere are reproduced upon the plane still as rectangular, but the magnitudes are different, a,.,\ equal distances of latitude upon the sphere are represented by (listanees which diminish towards the pole and increase towards the equator. The differences of
Digitized by Coogle
TREATISE ON PROJECTIONS.
67
longitude are all greater upon the surface of .the cone than upon the sphere, except for the parallel of tangency. The error in latitude may be completely, and that in longitude partially, eliminated by laying oft' along the middle meridian of the development the rectified lengths of the distances between the parallels, and through the points thus obtained, with the vertex of the cone as a center, describing arcs of circles. By this means we obtain for the differences in latitude their true values, and for the difterences in longitude values which are more nearly correct than those ginn by the first method. Fig. 13 shows both methods, the dotted lines corresponding to the second method.
v
v
FIo.13.
We have clearly, from the first figure,
lSOO n
1ff" cos D. mm'
when 8. is the latitude of the middle parallel RM, and n is the difference of longitude of the extreme meridians which are to be projected. Let also V denote the angle of the extreme elements of the cone which appear in the development. The radius VM of the middle parallel is given by
VM=r cot 00
and from figure (2) follows
lSOO V
7ft' cot 80 = mm'
Combination of these two values for mm' gives
V =n sin Do
It is obvious now how to construct the projection: The angle V being determined, we have fur the radius Qf the middle parallel, VM =T cot 00 Layoff from M the distances Ma' and }fbi as obtained by actual rectification. If the distance ab contains n degrees,
and Mb', Ma' each
71'rn ={,(,o
Having then the center and one point in the circumference, we can draw the circles which represent the parallels of latitude. If ye call (J the angle between the projectious of two meridians eorre
Digitized by Coogle
68
TREATISE ON PRO~OTION8.
sponding to UI upon the sphere, we have clearly
(f V . ==smOo UI lr
The radius of the parallel at latitnde 0 will be
=r [cot 00(000)] and the corresponding arc of longitude UI will be
=rUl sin 00 [cot 00(00.)1 The error for each degree of the parallel will then be
=r (000) sin 00
Euler investigated at some length the theory of conic projection and determined a cone fulfllling the following conditions: (1) That the errors at the northern and southern extremities of the chart should be equal. (2) That they shall be equal to the greatest error which occurs near the mean parallel.
The cone in this case is obviously a secant and not a tangent cone to the sphere. Let O. denote the least latitude of the region to be projected, and 0& the greatest value of the latitude. Let AD denote the portion of the middle meridian comprised between these extreme latitudes. Designate
o
1<1.G.14.
by b the length of 10 of the meridian, and let P and Q be the intersections of the central meridian with tle parallels along which the degrees shall preserve upon the map their exact ratio with the actual degrees of latitude; also call 0, and Of tbe latitudes of these two parallels, upon .each of which a degree of longitude has respectively the values b cos 8, and b 008 0,. Layoff these two values of 10 along the Iines Pp and Qq perpendicular to AB, and join pq; this line will represent the meridian removed one degree from AB. The point of intersection 0 will obviously be the common point of meeting of all the meridiaus and the center of all the parallels. The distance from 0 to any parallel is readily found; we have, since OPp is a right angle
PpQq Pp PQ . =PO
or
Digitized by Coogle
TREATISE ON PROJECTIONS.
69
from which
Having determined the center 0, it is only necessary to draw an arc of radius OP and npon it lay off lengths =h cos 0,; these will give the points through which the meridians pass; then laying off, along the middle meridian, distances equal to the number of degrees of latitude of the different parallels to be constructed, draw through the points thus found circles having their centers at 0, and the projections of the parallels will be constructed.
We will now determine the errors resulting from this construction upon the extreme parallelR throngh A and B. Calling til the angle POp, we find
_ Pp _6 (cos O,cos Of>
01_ _
PO O,Of
which becomes
_cos O,cos Of
aJ_ ._ ,
(O,Of)u
if we take 6=10, and express the denominator in parts of radius, which is done by making 1)=0.01745329
the value of 10 in a circle of radius unity. Call z the distance in degrees from the center 0 to the pole. The distance from P to the pole will be 0000., from P to 0 will be 9000,+z; the value of this in parts of radius will be
It is easy to see now that we must h~ve
z~ ~,_ 01') oos 0, 900+0, cos O,cos Of
The distance of the extreme parallel A from 0 will be, in parts of radius, AO=I) (900o.+z)
Mnltiplying this by the value of aJ, we have for the value of the degree upon this parallel
A 6 (OOOD.+z) (oos 0l'oos Of) a_ 00
, ,
instead of " cos 0.. The ditl'erenoo of these two values gives the error along the parallel through A. For B the error is the difference of
~ (OOOO.+z) (cos o,cos Of) d " 0 u 0 0 an cos. , ,
Euler's proposition was to determine the parallels P and Q in such a manner as 'to make the extreme errors at A and B equal. Equating these two errors and reducing, we have
(0.0.) (cos o,cos Of) + (O,O,) (cos O.oos 0.)=0
For the length of one degree upon the parallels of A and B we have
I) (OOOo.+z) aJ and I) (9000.+z)w
We have from these
•
Digitized by Coogle
70
TREATISE ON PROJECTIONS.
from which follows
cos o .. cos 0. w
u(O,o .. }
Further, equate both of these errors to the greatest error which occurs between A and B, su pposin~ in the first instance that it occurs at the point X half way from A to B. The latitude of X is
0,,+0,
=2
The error there is
=[ [u(900 O"tO, z} wcos ~G_to, ]
its sign being opposite to the signs of the errors at A and B. The condition is DOW expressed by the two equations
Uivmg w its value
cos o,,cos 0. u( 0,  0 .. )
we find readily
which reduces to
(1800_.lL ° _10 +2z) 0,0.. [cos ° + cos O.+O'J
2 .. ~. cos o,,cos 0. • 2
from which z is readily found. Applying this to the construction of a map of Bussia, it is only necessary to write
0.=400 The formula fur w gives now at once
0 .. +0·_""'0 Q _tN
OJ
cos 400cos 700
w= 48'44/1
aou
The equation
gives now
(8502z) uw=1.33962 uw=0.0141
Now
therefore
2z 1.33962 85°100
0.0141 
or
So far we have assumed that the maximum error lay at the middle of il, but we will now find the correct point, and assume that for this place the latitude is 0; the error will now be
u (9000+z) wcos 0
Differentiating this with respect to ° and equating to zero we find for the position of maximum
error
sin 0=w=0.8098270
or
Digitized by Coogle
TREATISE ON PROJECTIONS.
71
Equating the error at 0 to those of A and B
u (18000,,0+2z) w=cos o,,+oos 0
from which
The values of z and 0 difter very little from their assumed values of 50 and 550 respectively. The errors at A and B ale then equal to
uw(9000,,+z)cos 0 .. =0.00946
A degree on the parallel of 400 is then expressed by O. 77550 instead of 0.76604, its true value upon the sphere. This degree is, then, about ir greater than the true degree on the parallel of 400; and the degree on the parallel of 70° is about 3\ too great, its true value being 0.34202.
MURDOCH'S PROJECTION.
FIG. 15.
In Fig. 15, let 0" and 06 denote the latitude of two extreme parallels Aa and Bb, which limit a spherical zone whose projection is to be determined. The latitude of M halfway between A and
B is Oa~06. Murdoch's projection consists in making the entire area of the chart equal to the entire area of the zone to be projected. In order to effect this it will be necessary, supposing PN and PO the radii of the extreme parallels of the chart (obtained by rectification), that the surface generated by the revolutiou of ON (=AB) about PC shall be =2lrr( ab), when t"=radius of the sphere expressed in degrees. Let 8 denote the equal angles 7}CM, COM; we must then have
2lrKkAB=2lrt"(ab) From the. similar triangles Kck and MFC, give
Kk KC FO=MO
Consequently
8 +8 Kk=t" cos ~ cos 8
and substituting this in the above equation
This gives for cos 8 the value
cos 8
. 868.
S1O2
068.
Digitized by Google
72
TREATISE ON PROJEOTIONS.
It is easy to see that for the radius Kp=R of the middle parallel we have
The quantities which we have already denoted by 7r and V are here connected by the relation V =7r sin 0.+0.
2
Murdoch, in order to draw the intermediate parallels, divided the right line du into equal parts, giving for tile radius of any parallel °
R+o.+ob=O
2
This method, although perfectly arbitrary, had the effect of diminishing the errors in the chart.
Mayer, who resumed the problem proposed by Murdoch, gave the radii p~ and P"i as
and since
A second method of projection was given by Murdoch, in which the eye is placed at the center of the sphere, as in gnomonic projection, and a perspective is made which is subject to the condition of preserving the entire surface of the zone which is to be represented. Lambert was the first to indicate a method of conic development which should preserve all the angles except the one at the vertex of the cone, when the 3600 having upon the sphere the pole for center will obviously be represented in different manners, according to the different conditions to be fulfilled. A full account of this method is given in the chapter on orthomorphic projections.
BONNE'S PROJECTION.
This method of projection is that which has been almost universally employed for the detailed topographical maps based on the detailed trigonometrical surveys of the several states of Europe. It was originated by Bonne, was thoroughly investigated by Henry and Puissant in connection wil h the map of France, and tables for France were computed by Pleases, In constructing a mup ou this projection a central meridian and It central parallel are first assumed. A cone tangent along the central parallel is then assumed, and the central meridian developed along that element of the cone which is tangent to it, and the cone is then developed on a tangent plane. The parallel falls into an arc of a circle with its center at the vertex, and the meridian becomes a graduated right line. Concentric circles are then conceived to be traced through points of this meridian at elementary distances along its length. The zones of the sphere lying between the parallels through these points are next conceived to be developed, each between its corresponding ares. Thus all the parallel zones of the sphere are rolled out on a plane in their true relations to each other. and to the central meridian, each having in projection the same width, length, and relation to the neighboring zones as on the spheroidal surface. As there are no openings between consecutive developed elements, the total area is unaltered by the development. Each meridian of the
Digitized by Google
•
TREATISE ON PROJECTIONS.
73
projection is so traced as to cut each parallel in the same point in which it intersected it on the sphere,
If the case in hand be that involving the greatest extension of the method, or that of the projection of the entire spheroidal'surface, a prime or central meridian must first be chosen, onehalf of which gives the central straight line of the development, and the other half cuts the zones apart and becomes the outer boundary of the total developed figure. Next the Iatitnde of the governing parallel must be assumed, thus fixing the center of all the concentric circles of development. Having then drawn a straight line and graduated it from 900 north latitude to 900 south latitude, and having fixed the vertex or center of development on it, concentric arcs are drawn from the center through the different graduations. There results from this process 'an oblong kidneyshaped figure, which represents the entire earth's surface, and the boundary of which is the doubledeveloped lower half of the meridian first assumed. This projection preserves in all cases the areas' developed, without any change. The meridians intersect the central parallel at right angles, and along this, as along the central meridian, the map is strictly correct. For moderate areas the intersections approach tolerably to being rectangular. All distances along parallels are correct, but distances along the meridians are increased in projection in the same ratio as the cosines of the angle between the radius of the parallel and the tangent to the meridian at the point of intersection are diminished. Thus, in a full earth projection the bounding meridian is elongated to abont twice its original length.' While each quadrilateral of the map preserves its area unchanged, its two diagonals become unequal; one increasing and the other decreasing in receding towards the corners of the map, the greatest inequality being towards the east and west polar corners.
o
FIG.l6.
Denote the radius of the central parallel by Po; then OAo=po=r cot 00
Denote by 8 the length of the arc AAo, and the arc passing through a given point M; 00 of course denotes the latitude of the central parallel, and 0 that of the parallel BC. The latitude of M is 8
=00 + , and thus. r
MA=p" =wrcos(oo+~) a:=llQ=psinw
It is not difficult in this projection to take account of the spheroidal form of the earth. It is only necessary to multiply cot 00 by the principal normal no, and replace the spherical arc 8 by the elliptic arc S, given by
S=a (le2) [A(Oflo)Bsin (OOo)cos(0+00)+~C8in2(0Oo)C082(U+0,,)J
Digitized by Coogle
74
TREATISE ON PROJECTIONS •
Then
•
p=nocotoS
amo ( b)
w=p cos OO+(i
These give the radii of the projections of the parallels, which are then readily constructed. Lay off from the central meridian, upon the parallels HOW constructed, lengths equal to one degree upon each different parallel, and through these points pass a curve, which will be the projection of the meridians. The lengths are given by the formula.
27t' 7t'a cos 0
"=a60n cos 0=180 (1e2 sin' 0)1
The concave parts of these curves are all turned towards the central meridian.
o
FIG. 17.
The angle x, in Fig. 17, is the angle which the tangent to the meridian at M makes with the radius OM of the parallel through that point. This angle is also the difference between the angle that the meridian makes with the parallel at this point and 900.
We have obviously
tanx ¥
but
therefore Now
dp=dB
pdlJJ tanx=7/ii
Differentiating this gives
aw sin 0 (1&) dO pdw+wdp (1& siii'O)I.
. pdw+wdp=pdw+wdB
dB a(1&) do
(1& sin' 0)1
pdw .
tl8+w=w SID 0
and we have
But we know that
Consequently we may write
Digitized by Coogle
TREATISE ON PROJECTIONS.
75
and
tanflJ=wsinow
aw cos 00 01= Pl(ltJ' sin200)t
a cot 00 Pl={l_e2~siD.i Oo)t
Combining these
and for this case tan flJ=O or flJ=O, which only shows what we already know, viz: that the meridians and central parallel cut at right angles.
If for the central parallel we assume the equator, the vertex of the tangent cone is removed to an infinite distance, the pa~nels all fall into straight lines, and we have the socalled Plamsteed's projection. The kidneyshaped Bonne becomes an elongated oval with the half meridian for one axis and the whole equator for the other. The coordinates for any point in this projection are readily found to be
The form of the equation giving flJ has induced M. d'Avezac to give this projection the name sinusoidtll.
This projection, which should really be called Sanson's projection, is evidently only a particular case of Bonne's method; it is based upon a resolution of the earth's surface into zones or rings by parallels of latitude taken at successive elementary distances laid off along the central meridian of the area to be projected. Having developed this center meridian on a straight line of the plane of projection, a series of perpendiculars is conceived to be erected at the elementary distances along this line. Between these perpendiculars the elementary zones are conceived to be developed in the correct relations to each other and the center meridian. Each zone being of uniform width occupies a constant length along its entire developed length, and consequently the area of the plane projection is exactly equal to that of the spheroidal surface thus developed. The meridians ot the developed spheroid are traced through the same points of the parallels in which they before intersected them. They all cut the parallels obliquely and are concave towards the centre meridian. Thus while each quadrilateral between parallels and meridians contains the same area and points after development as before, the form of the configuration is considerably distorted in receding from the central meridian, and the obliquity of the intersections between parallels and meridians grows to be highly unnatural.
Digitized by Coogle
WERNER'S EQUIVALENT PROJEm'ION.
If the vertex of the cone approaches the sphere instead of receding from it, as in the preeeding case, we have finally, when the tangent con'e becomes a tangent plane, the projection known as Werner's Eqnivalent Projection. The parallels are now arcs of circles described about the pole as a center and with radii equal to their actual distances from the pole, i. e., equal to the rectified arc of the eolatitudes, The meridians are drawn by laying off on the parallels the actual distances between the meridians as they intersect the parallels on the sphere. This projection is not of enough importance to spend any time in obtaining any of the formulas connected with it.
POL YCONIO PBOJEOTIONS.
In all the cases of conic projection that we have treated so far we have supposed that a narrow zone of the earth was to be projected and that for the zone was substituted a developable surface upon which the parallels and meridians were constructed according to any manner that may be desirable. We have seenthat this kind of projection is only available when but a small portion of the earth is to be represented and that to make a projection of a country of great extent in latitude some modification would be necessary.
The system which is used in America and in England replaces each narrow zone of the earth's surface by the corresponding conic zone in such a way as to preserve the orthogonality of the meridians and parallels. This is the projection of which we have already spoken at length in the
76
TREATISE ON PROJECTIONS.
introduction, under the. title of Polyconic Projection. As a very fnll account of trus system has been already given, and comparisona made with the other ordinary methods of projection, we will not say anything on the suhject here, bnt wiU proceed to develop the theory of the system.
The name rectangular polyconic projection is applied to the method in which each parallel of the spheroid is developed symmetrically frOID an assumed central meridian hy means of the cone tangent along its circumference. Supposing each element thus developed relative to the common centralmcridian. it is evident that a projection results in which all parallels and meridians intersect at right angles. The parallels will be projected in circles, and the meridiansin curves which cut those circles at right angles. The radii of the parallels are equal to the cotangents of their latitudes (to radius supposed unity), and the centers are upon the line which has been chosen as the central meridian. Along this meridian the parallels preserve the same distance as they do
upon the sphere. •
FIG. 18.
In Fig. 18 let M be any point of the centralmeridian of which the latitude is O=900u; P the pole, the arc PM=ru. The center of the parallel through the point M is given by OM=rtanu. If M' be a point infinitely near to M (i. 6., MM'=rdu) and C' the center of the corresponding circle, we have C'M'=rtan (u+du), or
p=rt8nu Expanding the second of these, we have
p+dp='Y tan (u+du)
dp=r sec2 ttdu
but
.dp=CC'+MM'=CC'+rdu
Therefore
We have from the triangle CC'D
sin¥, C'B sinB=CC'
or
 ~¥' =tan udu Slll¥,
and integrating
10gcoStt=logtan; +const.
Digitized by Coogle
TREATISE ON PROJEOTIONS
71
or passing to exponentials
Now
tanu=e r
therefore
.y,.'J+p' Substituting this in the equation for the meridians we have
cosu
r
p=r J fit cot' 'P l=_r_Jfit  sin' f (L+o')
2 sin ~ 2
2 .
or
The distance from any point A to the central meridian is =psin,or=rtanusin,; but
rmnu sin =20 rsinu
, l+fitcosu
Foru=90<?, or, at the equator, this becomes
=2cr
The constant 0 must then represent onehalf the longttude of the given meridian, the equator being developed in its true length and divided into equal parts in the same manner as the central meridian. The following construction for this projection is due to Mr. O'Farrell, or the topographical department of the War Office, England. All data being as already given, draw at l[ the tangent nn' perpendienlae to PM. In order to determine the point A, whose longitude is given as w, layoff f'iom M the lengths Mn=Mn' equal to the true length of the required arc on tho par
allel 0, i. e.,=the arc i described with a radius=rBi~u. With n and 'Il' as centers and n'O and flC as radii, draw arcs cutting the given parallel in the points A and A',
M w. .
n= i) BIn U=OSlll u
OJ
and, since we have
OM rtanu
tanMCn=o cosu=tan ~ 2
. or, finally,
ACM='P and the distance from A to the central meridian is
=rtan usin 'P
Tbe radius of the curvature of the meridian whose longitude is w is readily obtained. We have
AA'=ds
Now
also
sec2~ t!r=osin udu 2 2
Digitized by Coogle
78
TREATISE ON PROJECTIONS.
therefore, if p denote.the radius of curvature of the meridian, we have by easy reductions 1+&+& sin2 u
p=r~csinu~
Now consider the distortion in this case, and for this purpose imagine a. small square described on the sphere, having its sides parallel and perpendicular to the meridian. Let u and w (=2 0) define
its position, and let a be the length of the side. If we differentiate the equation tani=ocos u on the supposition that u is constant, we have
sec2 9' d9' =cos udo
also, the length of the representation of 2dc is tan ud'f', or
Hence that side of the square which is parallel to the equator will be represented by a line equal to
Similarly the meridian side will be represented by
a cos2~(1+&+& sin2u)
The square is therefore represented by a rectangle whose sides have the ratio 1+&+& siu2u: 1
and its area is increased in the ratio
1+&+& sin2u. 1 (1+&oos2u)2" .
If we make this ratio = unity, then results the equation
C" cos' tt43c2 cos" u2&=0
which is satisfied either by c=O, i. e., w=O, or by
&,oos2u+3 cos2u2=0
We see from this that there is no exaggeration of area along the meridian or along the curve given by the last equation. This curve crosses the central meridian at right angles in the latitude of about 540 4.1'; it thence slowly inclines southward, and at 900 of longitude from the central meridian reaches 500 26' of longitude; at 1800, or the opposite meridian, it has reached 430 46'. The areas of all tracts of countries lying Oil the north side o_f this curve will be diminished in the representation, and for all tracts of countries south of this curve the areas will be increased in the representation.
If we represent the whole surface of the globe continuously, the area of the representation is
r[(4+1r2) tanI;+21r]
which is greater than the true surface of the globe in the ratio 8 : 5.
The perimeter of the representation is equal to the perimeter of the globe multiplied by v'4+lt21, or 2.72.
It is desirable in certain cases to retain the lengths of the degrees on an the parallels at the sacrifice of their perpendicularity to the meridians. We thus obtain what is known as the ordi
Digitized by Coogle
•
TREATISE ON PROJEOTIONS.
79
nary polyconic projection, which applied to the representation of the entire surface of the globe gives a figure with two rectangular axes and from equal quadrants as in the rectangular polyeonic projection. The central meridian alone is perpendicular to the parallels and is developed in its true length; upon eaelr parallel described with tbe cotangent of its latitude as a radius we layoff the true lengths of the degrees of longitude and draw through the corresponding points so obtained curves wbich will be the projections of the meridians. The ordinary poly conic method bas been adopted by the United States Ooast Survey because its operations being in great part limited to a narrow belt along the seaboard, and not being intended to furnish a map of the country in regular uniform sheets, it is preferred to make an independent projection for each plane table and hydrograpbic sheet, by means of its own central meridian.
The method of projection ill common use in the Coast Survey Office for small areas, such as those of planetable and hydrographic sheets, is called the equidistant polyeonie projection. This is to be regarded as a convenient graphic approximation, admissible within certain limits, rather than as a distinct projection, though it is capable of being extended to the largest areas and with results quite peculiar to itself. In constructing such a projection a central meridian and a central parallel are chosen, and they are constructed as in the rectangular poly conic method. The top or bottom parallel and a sufficient nnmber of intermediate parallels are constrncted by means of the tables prepared for the purpose, and the points of intersection of the different meridians with these parallels ~re then found and the meridians drawn:
Then starting from the central parallel the distance to the next parallel is taken from the central meridian and laid oft' on each other meridian. A parallel is traced through the points thus
. found. Each parallel is constructed hy laying off equal distances on the meridians in like manner, and the tabular auxiliary parallels are, all except the central one, erased. In fact, as only the points of intersection are required, the auxiliary parallels sbould not be actually drawn. From this process of constroction results a projection in which equal meridian distances are intercepted everywhere between the same parallels.
CYLINDRIC PROJEOTION,
So far, in treating of projection by development of some auxiliary surface, we have confined ourselves to the case of intersecting or tangent cones. The next most natural case to consider is when the developable surface is a cylinder. We cannot obviously, as in the case of the cone, pass one cylinder through the upper and lower parallels of a spherical zone, so tbat we cannot here have more than one of these parallels developed in its true size; if the zone is above the equator, the lower parallel may be developed ill its true size by circumscribing a cylinder, and the upper parallel may be represented in its true size by inscribing a cylinder, The better plan, however, and one wbich in general reduces distortion, is to pass the cylinder through some parallel inter. mediate between the extreme parallels of the zone to be projected. We will, however, first consider the case where the cylinder is tangent to the sphere either along the equator or along a meridian.
1.'HE SQUARE PROJEOTION.
The simplest, but rude, method is one in which the cylinder being tangent along the equator, the meridians and parallels appear as equidistant parallel straigbt lines, forming squares. Degrees of latitude and longitude are here all supposed equal in length. Distances and areas, especially in an eastandwest direction, are grossly exaggerated, though for an elementary surface the true proportions of a figure are preserved. This method is occasionally used for representing small surfaces near the equator.
PROJEOTIONS WITH CONVERGING MERIDIANS.
This is a modification of the square projection designed to conform nearly to the condition that arcs of longitude shall appear proportional to the cosines of their respective latitudes. Tho straight line representing the central meridian being properly graduated, that is, the true length of all arc of a degree of latitude (or of a minute or multiple thereof, as the case may be) having been laid off according to the scale adopted, two straight lines are drawn at right angles to the
Digitized by Coogle
•
80
'TREATISE ON PROJECTIONS.
meridian to represent parallels, one near the bottom and the other near the top of the chart. These parallels are next graduated, the arcs representing degrees (multiples or subdivisions) of longitude on each having, by scale, the true length belonging to the latitude. The corresponding points of equal nominal angular distance from the middle meridian thus marked on the parallels, when connected by straight lines, will produce the system of convergent meridians. The disadvantages of this projection are in the facts that but two of the parallels exhibit the lengths of arcs of longitude in their true proportion and that the central meridian is the only one which cuts the parallels at right angles. The projection is suitable for the projection of tolerably large areas, the above defects not being of a serious nature within ordinary limits; it also recommends itself by the ease with which points can be projected or taken off the chart by means of latitude and longitude.
THE RECTANGULAR PROJECTION.
A less defective method of delineation than the square projection consists in presenting the lengths of degrees of longitude along the middle parallel of the chart in their true relation to the corresponding degrees on the sphere; they will therefore appear smaller than the degrees of latitude in the proportion 1 : cos 01' In an eastandwest direction the chart is unduly expanded above and unduly contracted below the middle parallel.
THE RECTANGULAR EQUA.LSURFACE PROJEOTION.
This differs from the last in that the distances of the parallels, instead of being equal, are now drawn parallel to the equator at distances proportional to the sine of the latitude. This gives. it the distinctive property of having the areas of rectangles or zones on the proiection propor"tional to the areas of the corresponding figures on the sphere. The distortion, however, becomes quite excessive in the higher latitudes.
C.ASSINI'S PROJECTION.
This projection makes no use of the parallels of latitude, but substitutes for them a second system of coordinates, viz, one at right angles to the principal or central meridian; it is consequently convenient ill connection with rectangular spherical coordinates having their origin in the middle of the chart; the projection of Oasslm's chart of France consisted of squares and had neither meridians (excepting one) nor parallels. This simple form is, however, not the one which is generally known under Oasstui's name. In the projection commonly called Oasstui's the cylinder is tangent along a meridian; through the different points of division of the equator, planes are passed parallel to the plane of this meridian; and through thtl points of division of the meridian planes are passed intersecting the plane of the equator in a common diameter of the sphere. The first system of planes, of course, cuts small circles, and the second great circles, from the sphere. The cylinder is now developed, the generatives passing through the points of division of the meridian representing the great circles perpendicular to this meridian, while the small circles which are parallel to it have for their projection the .development of these intersections of the cylinder with their planes.
This projection is not now employed, as it offers no facilities for platting positions by latitnde and longitude; moreover, the distortion rapidly increases with the distance from the central meridian of the chart.
We will now obtain formulas which will enable us to find the forms of the projectlons of the parallels and meridians. In Fig. 19, M denotes the center of the sphere of radius MA=r; P is an arbitrary point in the surface, for which we have
AB denotes a quadrant of the equator and AQ a quadrant of the first meridian. The determination of the position of P in the case of Oasslni's projection is effected by means of the great circle passing throngh Band P, and the circle GH whose plane is parallel to that of tbe first meridian AQ; write
Digitized by Coogle
TREATISE ON PROmOTIONS.
81
, We have now, in therigbt triangle MPN,
PN=r sin 8
Q
G "'10. 19.
and from the rigbt triangle PKN but from the triangle MPK we have
PN=PKsinwl
and consequently
Equating the two values of PN and in like manner
PN =r 008 81 sin Ull
sin 8=008 01 sin Ull
sin 81=008 8 sin III From these two equations we obtain readily
oot UI=C08 Ull cot 81 and we also have tLe formulas
sin 01=008 0 sin III cot UI, =cot 8 008 UI P
sin O=oos 0, sin.1 cot .=cot"l cos •
,
\
A M
0
•
C FIG. 20.
Let now, in C&88ini's projection (Fig. 20); 0, 0' denote the center of coordinates, '1=OA=0' A', ~=AM=A'M'; also call 00 the latitnde of 0; then 00+'1 is the quantity denoted by QI, in the preceding formn1as, and ~ is identical with 0, 80 we have
sin ~=C08 8 sin QI cot (Oo+'1)cos QI cotfJ
6TP
Digitized by Coogle
82
TREATISE ON PROJECTIONS.
By elimination of w from these equations we have for tbe equation of the projectionS of the parallels sin' ~+cot' (00+7j) sin' o=cos' 0
ana by elimination of 0 have for the meridians
[COSllw+cot'l (0'+7j)] [sin' w&n' ;]=sinJ fU cos' w
Both meridians and parallels are thus given in this projection as transcendental curves. In these last equations we have regarded the radius of tbe sphere as =1; now make the radius =y; then
for ~ and 7J we must write ~ and ~. When the projection only represents a narrow region included
y ,.
between two meridians and two parallels very near together, the ratios ~ and '! are very small and r r
so is tbe difference 000; so we can write
. . ~
Sln;=r
tan'1='! y
and the equation of the parallels becomes
('1r cot 00)1+~1=r' cot· 00 [cot 00+4 sin i (000)]
and that of the meridians
We 800 that in this case the meridians are projected in parabola and the parallels in circles, 'Write for convenience Oo+'1=A, and let the angle 1, in Fig. 21, denote the angle which the tangent PM to the projection of a meridian makes with the axis 7J; also, let X' denote the angle which the tangent PL to the projection of a parallel makes with the same axis. Then we have
M
tanx=dA
}'IG.21.
The equation of the meridians is easily thrown into the form tan ~=cos A tan w and that of the parallels also becomes very readily
sin O=&n A cos e For these we may substitute in practice the group
I
cot A=cot 0 cos tJJ
sill: ~=cos 0 sin Qr
Digitized by Coogle
TREATISE ON PHOJECTIONS.
8S
We now have
and also
tanl=tan OJ cose sin o Now, from the equation of the parallels, we have
or, since tan ~ = cos A tan OJ
de 1
.tany'
i.1I = tan A tan s ..
\
tan ' 1 cos ~ cot OJ
Z sin A tan OJ sin 0
Combining these values of tan Z and tan t by multiplication, there results tan I tan t =cos· e
The condition that the projections of the meridians and parallels should cut at right angles is tanItant=l
So it is clear that in general in Oassini's projection the meridians and parallels are not represented by orthogonal curves. For e=o we have
tan 1 tan 1'=1
or the projections of all parallels are perpendicular to the central meridian. If A=9CO we have l=OJ, or the projections of the meridians make the same angles with each other as the meridians
themselves. From the equations for the meridians and parallels obtain the values of :~ and ~, and substitute each set of values in the formula
_ l+(:t)~
p~
d~
and by very simple reductions we find for the radius of curvature of the projections of the meridians rsecl
P·sinl [cot_( cosz+2 tan e sinl]
and for the parallels
_ rsin2lsinae.
PPc:OS'l' cote (si1i'~+cotA)
For the case of 1=0 and A=O, or the point where a meridian cuts the equator, the expression for P. becomes indeterminate, but by the ordinary means for finding the value of indeterminate quantitit>,s P. is found to be for this point
The radius PI' becomes infinite for the same latitude, i. e., for 1==0, but is indeterminate for the points at which e=o; one has for these points
or
t"
, PP~ta'n::(}
Digitized by Coogle
84
TREATISE ON PROJECTIONS.
MEROATOR'S PROJEOTION.
If a cylinder be passed tangent to the sphere at the equator, and the planes of the meridians and parallels be produced to cut the surface of the cylinder, the meridians will be represented upon the cylinder by right lines and the parallels by circles, the right sections of the cylinder. If the cylinder be developed, we obtain a projection in which both meridians and parallels arerepresented by right lines, the angles between the lines being right angles.
It will be convenient here to define a. loxodromic curve, or simply a loxodromic; this is a line drawn upon the surface of the sphere in such a manner as to cut all the meridians at the same angle, Any straight line drawn up on a chart constructe¥ above will of course represent a loxodromic upon the sphere. This projection bas already heen alluded to under the head of the square projection; its disadvantages are obviously very great, only eastandwest and northandsouth directions being preserved, and degrees of longitude only preserving their true length upon the line of contact of the cylinder and sphere, i. e., upon the equator.
Beduced charts, or Mercator's charts, are charts whose construction is such that not only are the meridians given as right lines, a necessary condition that the loxodromic curve may ·be represented hy a right line, but so that the angle between any two cnrvilinear elements upon the sphere is represented upon the chart by an equal angle between the representatives of these elements .
. This is effected by a proper spacing of the distances between the parallels, which are also represented as right lines upon the chart. We will now give the means of determining the proper position of any parallel upon the chart by observing the condition that the angles formed by two curvilinear elements upon the sphere shall be preserved upon the cbart. Let ab, Fig. 22, denote
FIG. 22.
\
an element of a loxodromic cutting the two infinitely near meridians Po, and Pb; draw the parallel ab'. Now, in order that the angle A.BB' upon the chart shall=abb' upon the sphere, we must have
BB' AS'
bb' =o,[)'
Now, as the distance between any two meridians is everywhere equal to the distance between the same meridians at the equator, AB is=ap, the element of the equator. Let d4 represent an element of' a meridian upon the earth (hereafter we will write earth instead of sphere, as the intention is to take account of the true shape of spheroid) and dB the element corresponding to this upon the chart . . Let 0 represent the latitude of the extremity of the elliptic arc dtl (which is measured, of course, from the equator), p the radius of the parallel of latitude 0, c the eccentrieity, and, as be4'ore, a the radius of the equator. The condition
BB' AB bb' =aTl
becomes now
Digitized by Coogle
TREATISE ON PROJEOTIONS.
85
but
j 1
and
1 "I
~ '/
: :;:;.,' \ II'
, ,J
C'
•
a cosO
 p (1_e2 sil~2 O)t
a (Ic') dO
ds  
cos 0(1_£2 sinlO)
Multiplying e' in the numerator of thi.xpression by sin' 0+cos2 8, this becomes
therefore
adO eZ cosO
ds=  aiJJJ    
cos 0 1_,' sin' 8
Integrating from 0=0 to 0, i. e., from the equator to latitude 0
i" cos odo i" 'cos odO
8=a J 0 cos' 0 at J 0 1_£2 sini 0
ci I 1..1 l+sino 1.. I l+e sin~
= 12' og lI' og
M: Isin 0 1, sin 0
•
where M =0.4342945 is the modulus of the common logarithm. Since
l+t ( fJ tI "
log I_t=2 t+3+li+' . )
this formula can be written
a r ( 8)6'( (t sin 0)3 )]
S=M L log tan 450 + 2  e sin 0+ a+ .
A· ... ,,~ .• f d writln t1 10800 1 d 10800 thei al h
gam expressmg a III mmutes 0 arc, an wn g or . M' an _ err v mes, we ave
7t 7t
finally for 8
8=7915'.704674 log tan ( 450+;)3437,.7(£' sin 0+" s!na 0 + . . . . .)
where powers of , above the fourth may be neglected. The fnrthM eonsideration of this projection is reserved for Part II.
c
ll'
~'lG. 23.
A brief itlvestigd.t.ion will be given here of the loxodromic curves upon the sphere] at another time a morc rigoroue and general study of tbes6 curves will be given. FrO~ the definition it is
Digitized by Coogle
86
.
TREATISE ON PROJECTIONS.
•
clear that a loxodromic curve upon the sphere is a species of spiral which winds around the sphere approaching indefinitely near but never passing through the poles, which are consequently assymptotic points to the curve. This is obvious when we consider that the loxodromic making equal angles with all the meridians, at the pole it would have to make the same angle with all the meridians, which would be impossible.
Consider two points A and B (Fig. 23) on consecutive meridians, the geographical eoordinates of A being 0 and w, and those of B being 01 and WI. The angle CBA, measured from the north, is the constant amrle that the loxodromic makes with the meridians; call this angle Q. The arc EE' of the equator measures the angle between the meridians. Let dO represent the change of latitude in passing from B to A, and dw the infinitesimal arc EE'. ~raw the parallel Bb; then, rolling the radius of the equator r, we have, since Bb=dw cos 0,
rdO d sin 0
dw cot Q=u=r I . So
cos sm
Integrating this, we have
(A)
For Q=O, or 1800, this reduces to WUlI=O, which gives the curve as the meridian of B. For Q=OOO the equation it! satisfied only by 0=011 or the curve is the parallel passing through B; for tbe particular value 01=0, the loxodromic is the equator itself. For the computation of the arc 01011 in minutes, we have
In the above equation (A) we have, by passing to exponentials,
( 0)
tan 450+
.exp. (WWI)cot Q = 2
r tan(450+~)
Now
•
then
o aI tan·=· 2 a+l
Substituting tbe value of a as obtained. from the above equation, we have finally
0_ mnC 45o+i) expo w;W'cotQ_I tan2 _. 
tan ( 450+~) expo ~;:w, cotQ+I
The introduction of rectangular rectilinear eoordlnates in this oqnntion by means of the formulas
(B)
fl:=r C08 01 COIl (I
y=r sin 01 cos 0 z=r sinO
fl:1 =r cos WI cos 01 YI=r sin (1.11 cos 01 zl=r sin 01
• Exp. i. used 81 au abbreviation of "e to ~e IKJfHf' 0/."
Digitized by Coogle